T?TULO: Nuevas plataformas anal?ticas en metabol?mica AUTOR: Beatriz ?lvarez S?nchez ? Edita: Servicio de Publicaciones de la Universidad de C?rdoba. 2012 Campus de Rabanales Ctra. Nacional IV, Km. 396 A 14071 C?rdoba www.uco.es/publicaciones publicaciones@uco.es UNIVERSIDAD DE C?RDOBA FACULTAD DE CIENCIAS DEPARTAMENTO DE QU?MICA ANAL?TICA NUEVAS PLATAFORMAS ANAL?TICAS EN METABOL?MICA NEW ANALYTICAL PLATFORMS IN METABOLOMICS Beatriz ?lvarez S?nchez C?rdoba, enero de 2012 Mediante la defensa de esta Memoria se pretende optar a la menci?n de Doctora Europea, habida cuenta de que la doctoranda re?ne los requisitos exigidos para tal menci?n, a saber: 1. Informes favorables de dos doctores pertenecientes a Instituciones de Ense?anza Superior de otros pa?ses europeos: ? Prof. Dr. Carlos Cordeiro, Departamento de Qu?mica e Bioqu?mica, Faculdade de Ci?ncias da Universidade de Lisboa, Portugal. ? Prof. Dr. Walter J?ger, Department of Clinical Pharmacy and Diagnostics, University of Vienna, Austria. 2. Uno de los miembros del tribunal que ha de evaluar la Tesis pertenece a un centro de Ense?anza Superior de otro pa?s europeo: ? Prof. Dr. Oleg Mayboroda, Leiden University Medical Center, Biomolecular Mass Spectrometry Unit, Leiden, The Nederlands. 3. La defensa de parte de esta Memoria se realizar? en la lengua oficial de otro pa?s europeo: Ingl?s. 4. Estancia de tres meses en un centro de investigaci?n de otro pa?s europeo: ? Biomedical Proteomics Group (BPRG) dirigido por el Prof. Jean- Charles S?nchez, Department of Structural Biology and Bioinformatics Faculty of Medicine, University of Geneva, Geneva, Switzerland. ?NDICE Index ?NDICE OBJETIVOS??????????????????????????? 1 OBJECTIVES??????????????????????????.. 9 INTRODUCTION?????????????????????????. 15 1. Systems biology and development of ?omics? disciplines?????......... 17 2. Metabolomics within the systems biology context?????????...? 19 3. Strategies of analysis in metabolomics?.??????????.............. 20 4. Analytical processes in metabolomics???????????????. 24 5. Detection techniques??????????????????????. 25 6. Data treatment techniques and databases????????????..... 45 7. Applications and future trends??????????????????... 48 8. References??????????????????????????. HERRAMIENTAS Y EQUIPOS ANAL?TICOS?????????????. 60 79 PARTE EXPERIMENTAL????????????????....................... 87 PART 1: Sample preparation in Metabolomics????..????????. 89 Chapter 1. Metabolomics analysis I. Selection of biological samples and practical aspects preceding sample preparation???????????.... 95 Chapter 2. Metabolomics Analysis II: Preparation of biological samples prior to detection ????????????????????.??????... 117 PART 2: Target analysis??????..????..??????????.... 143 Chapter 3: Automated solid-phase extraction for concentration and clean-up of female steroid hormones prior to liquid chromatography?electrospray ionization?tandem mass spectrometry: An approach to lipidomics??........... 149 Chapter 4: Ultrasound-enhanced enzymatic hydrolysis of conjugated female steroids as pretreatment for their analysis by LC?MS/MS in urine?????????????????????????????.... 177 Chapter 5: Targeted analysis of sphingoid precursors in human biofluids by solid phase extraction with in situ derivatization prior to ?-LC?LIF determination??....???????????????????????.. Chapter 6: Automated determination of folate catabolites in human biofluids (urine, breast milk and serum) by on-line SPE?HILIC?MS/MS??????... 203 229 Chapter 7: Ultrasonic enhancement of leaching and in situ derivatization of haloacetic acids in vegetable foods prior to gas chromatography?electron capture detection. ??????????????????..?????? Chapter 8: Quantitative targeted profiling of compounds with nutraceutical interest in tomato by mass-spectrometry-based analytical methodologies?????????????????????????? 257 279 PART 3: Global Profiling Analysis??????????????????.. 307 Chapter 9: Human salivary metabolomics profiling by LC?TOF/MS in accuracy mode.????????????????????????? Chapter 10: Comparison of sample preparation protocols for metabolomics profiling of human breast milk by high-accuracy LC?TOF/MS and 1H- 313 NMR?????????????????????????????? 347 PART 4: Metabolomics Fingerprinting????????????????... Chapter 11: Near-infrared spectroscopy and partial least squares-class modelling (PLS-CM) for metabolomics fingerprinting discrimination of obese individuals after Intake of intervention breakfasts???????.......???... Chapter 12: Comparative study of the Influence of fried edible oils intake on the urinary Metabolic Fingerprint by Nuclear Magnetic Resonance Spectrometry (1H-NMR)???????????????????..??.. Chapter 13: Global metabolomics profiling of human urine by LC?TOF/MS in accurate mode to evaluate the intake of breakfasts prepared with fried edible oils??????????????????????????????. DISCUSI?N DE LOS RESULTADOS????????????????... DICUSSION OF THE RESULTS??????????????????? CONCLUSIONES????????????????????????... CONCLUSIONS?????????????????????????. ANEXOS????????????????????????????.. 381 387 425 453 499 523 543 551 557 LISTA DE ABREVIATURAS????????????????????? LIST OF ABBREVIATIONS????????????????????.? 601 605 OBJETIVOS 3 Objetivos Objetivos El crecimiento exponencial que ha experimentado la metabol?mica en los ?ltimos a?os es un hecho incuestionable, debido a su significativa aportaci?n en ?reas de dominio exclusivo de ?micas desarrolladas con anterioridad; lo que pone de manifiesto que se trata de una disciplina totalmente aceptada por la comunidad cient?fica. De las estrategias t?picas de la metabol?mica (an?lisis orientado o ?targeting analysis?, perfil metabol?mico global o ?global metabolomics profiling? y huella metab?lica dactilar o ?metabolomics fingerprinting?) cada una de ellas est? orientada a la obtenci?n de un tipo de informaci?n caracter?stico. Dado el inter?s de estas estrategias, el objetivo gen?rico planteado al inicio de la investigaci?n que constituye la presente Tesis Doctoral fue realizar aportaciones utilizando cada una de estas estrategias para contribuir a eliminar las lagunas existentes en metabol?mica, de forma que se incremente la posibilidad de resoluci?n de problemas por ella misma o integrada con otras ?micas (biolog?a de sistemas). Apoyada en el equipamiento anal?tico del grupo de investigaci?n en el que se integra la doctoranda {sistemas miniaturizados de preparaci?n de muestra (laboratorio en una v?lvula, lab-on-valve o LOV), estaciones autom?ticas de extracci?n en fase s?lida (SPE) acopladas en l?nea con cromatograf?a l?quida (LC) o microcromatograf?a l?quida (?-LC) con detector de fluorescencia inducida por l?ser (LIF), de masas de triple cuadrupolo (QqQ), o de tiempo de vuelo?masas (TOF/MS); cromatograf?a de gases con detector de captura electr?nica (GC?ECD); o en colaboraci?n con otros grupos cuando no se ha dispuesto del equipo adecuado (espectrometr?a de resonancia magn?tica nuclear ?NMR, espectrometr?a de infrarrojo cercano, NIRS), as? como en la b?squeda y estudio de la informaci?n bibliogr?fica Nuevas plataformas anal?ticas en metabol?mica 4 sobre metabol?mica, se planificaron los siguientes objetivos concretos, que se fueron materializando y se recogen en los diferentes cap?tulos que componen esta Memoria: 1. Revisar de forma exhaustiva y cr?tica las etapas previas ?desde el muestreo hasta la introducci?n en el analizador? al an?lisis mediante cualquiera de las estrategias en metabol?mica. La experiencia del grupo en el que se integra la doctoranda en la preparaci?n de la muestra y su formaci?n personal en este ?mbito han dado como resultado las dos publicaciones que se recogen como Cap?tulos 1 y 2. Los objetivos concretos en investigaci?n de laboratorio se dividieron en funci?n de la estrategia utilizada. As?, en an?lisis orientado los objetivos fueron: 2. Innovar en la preparaci?n de muestras cl?nicas, tanto con el dise?o de las etapas a escala micro (utilizando estaciones LOV), como en su aceleraci?n mediante el uso de energ?as auxiliares, como ultrasonidos. Tras la adecuada preparaci?n de la muestra, la etapa de an?lisis mediante LC? QqQ permite la cuantificaci?n de familias de analitos diana en muestras complejas, como orina, y la orientaci?n a analitos de inter?s cl?nico, como hormonas femeninas de tipo esteroide, dando lugar a la investigaci?n que constituye los Cap?tulos 3 y 4. 3. Llevar a cabo la determinaci?n sensible y selectiva de compuestos lip?dicos apolares, como los precursores de esfingol?pidos, en suero y orina, mediante la automatizaci?n completa de la preparaci?n de muestra (incluyendo extracci?n y derivatizaci?n) previa a la separaci?n individual/detecci?n mediante ?-LC?LIF. 4. Desarrollar autom?ticamente el an?lisis completo mediante un acoplamiento en l?nea de la etapa de SPE con la separaci?n cromatogr?fica y la detecci?n por QqQ fue otro de los objetivos cumplidos (Cap?tulo 6). Validar el m?todo mediante su aplicaci?n a diferentes fluidos biol?gicos 5 Objetivos (orina, suero y leche materna) fue otro de los logros perseguido y conseguido en esta investigaci?n. 5. Innovar en la preparaci?n de muestras vegetales para su an?lisis por CG mediante la reducci?n dr?stica ?con auxilio de ultrasonidos? del tiempo requerido para esta etapa. Tras esta mejora dr?stica, los analitos (?cidos haloac?ticos) se determinan mediante GC?ECD (Cap?tulo 7). 6. Obtener informaci?n amplia y complementaria sobre las familias de compuestos t?picas del tomate mediante an?lisis orientados m?ltiples fue el objetivo cuyo desarrollo se recoge en el Cap?tulo 8. Partir de muestras liofilizadas y utilizar ultrasonidos para acelerar?mejorar la preparaci?n de un extracto en el que determinar carotenoides y fenoles mediante LC?QqQ y mono- y disac?ridos mediante GC?MS/MS constituy? el grueso de la planificaci?n. La informaci?n obtenida y los m?todos para llevarla a cabo se consideran de gran inter?s en el ?rea agr?cola dentro de programas de mejora de cultivos. Los objetivos concretos de la investigaci?n recogida en los Cap?tulos 9?10 fueron la obtenci?n de perfiles metab?licos globales en fluidos humanos, en su mayor parte poco estudiados, como es el caso de la leche y la saliva. Estos objetivos fueron los siguientes: 7. Proporcionar un m?todo ?ptimo para el an?lisis metabol?mico de saliva teniendo en cuenta los estudios que se?alan el inter?s cl?nico de este biofluido para la b?squeda de biomarcadores de ciertas enfermedades. La comparaci?n de diferentes protocolos de preparaci?n de muestra y la obtenci?n de su perfil metab?lico por LC?TOF/MS fueron los aspectos m?s destacados, tal como se recoge en el Cap?tulo 9. Conseguir un protocolo de preparaci?n de muestra ?ptimo que permitiera determinar un n?mero grande de compuestos identificados (az?cares, l?pidos, amino?cidos, antioxidantes, etc.), de muchos de los cuales no se hab?a encontrado informaci?n en la bibliograf?a, constitu?a el hito m?s destacado de este objetivo. Nuevas plataformas anal?ticas en metabol?mica 6 8. Desarrollar y comparar diferentes protocolos de preparaci?n de la muestra de leche materna para conseguir, mediante LC?TOF/MS, un perfil metabol?mico lo m?s completo posible (Cap?tulo 10), que permita la evaluaci?n nutricional de este biofluido. La separaci?n de las fases polar y no polar y el estudio de cada una de ellas en los modos positivo y negativo para obtener un perfil amplio de esta muestra poco estudiada, y el an?lisis directo mediante RMN, menos resolutivo, pero complementario estuvo dentro de lo planificado en este caso. 9. La investigaci?n recogida en los Cap?tulos 11?13 completa los objetivos concretos perseguidos, ya que est? dedicada al an?lisis de huellas dactilares metabol?micas. Esta estrategia constituye una herramienta id?nea para llevar a cabo la comparaci?n gen?rica de espectros de poblaciones amplias obtenidos con diferentes t?cnicas y que, tras un indispensable, generalmente amplio, tratamiento quimiom?trico, proporciona informaci?n de enorme inter?s para detectar?identificar variaciones o diferencias entre los miembros de la poblaci?n en estudio. La poblaci?n seleccionada para esta investigaci?n (compuesta por individuos obesos a los que se administraron de forma controlada desayunos preparados con cuatro tipos de aceites sometidos a fritura con diferente tipo y concentraci?n de antioxidantes) proporcion? muestras de orina que constituyeron la diana de esta investigaci?n, dedicada a: 10. Proporcionar diferentes niveles de informaci?n para la comparaci?n del efecto de la ingesta de cada uno de los desayunos en el metabolismo de los individuos que formaban la poblaci?n en estudio. Para ello se planific? la utilizaci?n de 3 tipos de t?cnicas anal?ticas que han dado lugar a cada uno de los cap?tulos que forman este bloque. As?, NIRS ?una t?cnica que se caracteriza por la alta frecuencia de an?lisis, reducido coste de an?lisis y facilidad de uso? permite el desarrollo de modelos quimiom?tricos con capacidad de predicci?n de la ingesta de cada uno de los desayunos administrados (Cap?tulo 11). Por otra parte, NMR, permite generar huellas dactilares para la discriminaci?n temporal entre muestras, 7 Objetivos as? como la identificaci?n b?sica de familias de metabolitos significativos (Cap?tulo 12). Finalmente, LC?TOF/MS constituye una plataforma que proporciona el nivel de informaci?n m?s avanzado para la identificaci?n de metabolitos con mayor significado estad?stico, a la vez que posee una enorme capacidad de discriminaci?n (Cap?tulo 13). 11. La formaci?n de la futura doctora, que constituye el objetivo ?ltimo de toda tesis doctoral, ha incluido el m?ster en ?Qu?mica Fina?, en el que la doctoranda ha cursado el n?mero de cr?ditos correspondientes. Paralelamente a la investigaci?n recogida en la parte principal de la Memoria, se ha pretendido una formaci?n m?s amplia de la doctoranda con la realizaci?n de otras actividades que se recogen como anexos, tales como: Cap?tulo de la ?Encyclopedia of Analytical Chemistry? sobre plataformas anal?ticas en metabol?mica, estrechamente relacionado con el tema de la Tesis. (Anexo I). Investigaci?n simult?nea con la de la tesis realizada en colaboraci?n con otros grupos de la UCO. (Anexo II). Cap?tulo del libro ?Advances in Flow Injection Analysis and Related Techniques?, sobre separaciones mediante filtraci?n y extracci?n l?quido? l?quido basadas en membranas. (Anexo III). Comunicaciones en conferencias nacionales e internacionales (Anexo IV). Cursos de formaci?n especializada en t?cnicas de an?lisis (HPLC y NIR) y en t?cnicas quimiom?tricas para el dise?o y optimizaci?n de experimentos (Anexo V). OBJECTIVES 11 Objectives Objectives The exponential growing of Metabolomics in the last years is an irrefutable fact that shows the significant and accepted contribution of this discipline to a broad variety of field of research, together with other older omics. Among the typical strategies in metabolomics (targeting analysis, global metabolomics profile and fingerprinting), each of them is focused on obtaining a characteristic type of information. Taking into acount the interest of these strategies, the general objective in planning the research to be developed was to use of them to contribute to fill the existing gaps in metabolomics, in order to increase the possibility to solve the existing problems, by metabolomics itself or integrated with other omics (systems biology). Supported by the analytical equipment of the research team where the PhD student is integrated {miniaturized systems for sample preparation (lab-on-valve or LOV), automatic stations for solid-phase extraction (SPE) on line coupled to liquid chromatography (LC), or liquid microchromatography (?-LC) with laser-induced fluorescence detector (LIFD), triple-quad mass detector (QqQ), or time-of-flight? mass detector (TOF/MS); gas chromatography with electron capture detector (GC? ECD}; o in colaboration with other groups when suitable equipment was not available in the laboratory of the group of the PhD student (nuclear magnetic resonance, NMR, near-infrarred spectroscopy, NIRS); as well as the searching for and exhaustive study of the bibliographic information on metabolomics, the following specific objectives were planned; then, they were carried out and constitute the different chapters of this book: 1. To review, in an exhaustive and critical manner, the previous steps ? from sampling to insertion into the analyzer? to analysis by the metabolomics strategies. The experience in sample preparation of the group where the PhD student is integrated and her personal formation in this field have yielded two publications, which constitute Chapters 1 and 2. The specific objectives in laboratory research were classified as a function of the given strategy. Thus, in targeting analysis, the objectives were as follows: Nuevas plataformas anal?ticas en metabol?mica 12 2. To innovate in sample preparation of clinical simples, both by the design of microscale steps (by using LOV stations) and by accelerating them with the assistance of auxiliary energies such as ultrasound. After sample preparation, the analysis by LC?QqQ allowed application of the resulting method to complex samples such as urine for the determination of femail hormones of steroid type, giving place to the research which constitutes Chapters 3 and 4. 3. To determine, in a sensitive and selective manner, lipidic non-polar compounds, such as sphingolipid precursos, in serum and urine with the complete automatization of simple preparation (including extraction and derivatization) prior to individual separation and detection by ?-LC?LIF. 4. To develop the entire analysis in an automatic way by coupling on-line the SPE step with the chromatographic separation and detection by QqQ was other of the fulfilled objectives (Chapter 6). To validate the method by application to different biological fluids (urine, serum and breast milk) was another pursued achievement in this research. 5. To innovate in sample preparation involving vegetal samples for subsequent introduction into GC was an objective achieved by a drastic reduction of the time required for this step with the help of ultrasound. After this drastic improvement, the analytes (haloacetic acids) are determined by GC?ECD (Chapter 7). 6. To obtain wide and complementary information on other families of typical compounds in tomato by multiple targeting analyses was the objective of the research in Chapter 8. An extract from lyophilized samples (also with the help of ultrasound) was used for determination of carotenoids and phenols by LC?QqQ, and mono- and disaccharides by GC?MS/MS. The information thus obtained, with no similar precedent in the literature, and the methods for its development are of great interest in the agriculture programs to improve this field. The concrete objetives of the research in Chapters 9 and 10 were to obtain global metabolomics profiles in human fluids, mainly those scarcely studied, as is the case of breast milk and saliva. Therefore, the objectives were: 13 Objectives 7. To provide an optimum method for sample preparation of saliva, prior to obtaining its metabolomic profile by LC?TOF/MS, as shown in Chapter 9. To identify a wide number of compounds, extending the previous knowledge about composition of this biofluid, was also a milestone of this research. 8. To develop and compare different protocols for sample preparation of breast milk in order to achieve, by LC?TOF/MS, a metabolomics profile as complete as posible (Chapter10). Separation of polar and non-polar phases and the study of each in the positive and negative ionization modes would provide a wide profile of this poorly studied sample, which could be completed by direct analysis by NMR, less resolutive than LC?TOF/MS,. 9. The research in Chapters 11?13 completes the concrete objectives, as it was devoted to the analysis of metabolomics fingerprintings. This strategy constitutes a suitable tool to carry out generic comparison of spectra from wide populations, which can be obtained by different techniques. After a thorough, mandatory chemometrics treatment, the comparisons provide suitable information to detect?identify changes and/or differences among the members of the population under study. The population selected to implement this research was integrated by obese individuals, who ingested, in a controlled design, breakfasts prepared with four frying oils with different type and concentration of antioxidants. The target individuals provided urine samples which were the target of this research, which had as main aim: 10. To provide different information levels to compare the intake effect of each breakfast in the metabolism of the individuals forming the population under study. With this aim, three types of analytical techniques was planned to be used, the research developed with each constituting one of the chapters which make this part of the book. Thus, NIRS, a technique characterized by a high frequency of analysis, small analysis costs and easy handling, allows the development of chemometric models capable for predicting the intake of each of the supplied breakfasts (Chapter 11). The second technique, NMR, allows generation of fingerprintings for temporary discrimination among samples, as well as basic identification of significant metabolite families (Chapter 12). Finally, LC?TOF/MS constitutes an instrumental platform capable of providing a higher information level Nuevas plataformas anal?ticas en metabol?mica 14 for identification of analytes with more statistical significance, together with an excellent capacity for discrimination (Chapter 13). 11. Formation of the future doctor, which constitutes the final objective of a doctoral thesis, has also included a master on ?Fine Chemistry?. Simultaneously to the research which constitutes the main part of this book, a wider formation of the PhD student was pursued by participation in other activities, included in the book as annexes, such as: Chapter of the ?Encyclopedia of Analytical Chemistry? on analytical platforms in metabolomics, closely related to the thesis subject. (Annex I). Research simultaneous with that of the thesis resulting from collaborations with other groups in UCO. (Annex II). Chapter of the book ?Advances in Flow Injection Analysis and Related Techniques? on membrane-based separations by filtration and liquid?liquid extraction. (Annex III). Poster communications in national and international meetings and conferences (Annex IV). Attendance to specialized courses on analytical techniques (NIRS and HPLC) and on chemometrics techniques applied to experimental design and optimization (Annex V). I NTRODUCCI? N Introductio n 17 Introduction 1. Systems biology and development of ?omics? disciplines Since the 1990s, and particularly since the elucidation of the human and other genomes, there has been a revolution of techniques and approaches used in molecular biology [1]. Such changes have fuelled a renewed interest in applying the information obtained from biological systems to different areas of knowledge, such as clinics, agriculture and pharmacy. However, molecular biology has failed in obtaining practical information about the physical state of organisms, as a direct relationship between the genome and the phenotype cannot be clearly observed [2]. In fact, within a molecular biology framework, it is generally accepted that there is a linear downstream flux of information that goes from genes to transcripts, proteins and down to metabolites through the action of enzymes. However, this traditional scheme is insufficient for interpreting biological systems, where metabolic processes are intimately networked with interactions among these structural levels [3]. The construction, visualization and understanding of these networks certainly pose huge challenges for modern biology, as does a full understanding of the metabolic fluxes and their responses to genetic variations or external stimuli [4]. In this scenario, a new system-level approach, in contrast to the classical molecular-level approach, has emerged as a result of the urgent need to bridge this genotype-to-phenotype gap? ?Systems biology? ??? is a relatively recent term arisen from the development of the ?omics? technologies such as genomics, proteomics or metabonomics/- metabolomics. In these fields of study, large amounts of data are obtained at each level of biomolecular organization [6]. One of the expectations of systems biology is that, in some way, such data can be integrated to give a 1 8 Nuevas plataformas anal?ticas en metabol?mica holistic picture of the state of a biological system, ultimately enabling the understanding of biology at both individual and population levels. Figure 1 shows this change of scheme from molecular to systems biology [7], the latter regarding the phenotype of the organism under study as the final goal of knowledge. As can be observed in the figure, this evolution is intimately related to a rapid development in available analysis instruments used at each level. Provided that the state of a living system, be it a cell, organ or whole organism, is a combination of genotype, physiology (e.g,. age), diseases, nutritional and environmental state, among other influencing factors, the complexity faced by systems biology is clearly enormous and can obviously be accomplished only by highly sensitive and accurate analytical and computational tools as well as by complete databases compiled in continuous feedback by devoted research organizations. Figure 1. A. Traditional central dogma of molecular biology. B. General scheme of systems biology. 19 Introduction 2. Metabolomics within the systems biology context Since the objective of systems biology is to obtain an integrative model of systems and their interactions, the experimental techniques that better suit systems biology are those that are system-wide and attempt to be as complete as possible. Therefore, genomics, transcriptomics, metabolomics, proteomics and the main high-throughput techniques are used to collect quantitative data for the construction and validation of models. Hence, the suffix ?omics? comes from the Latin form ?ome?, which means ?massive?? This has been used in a wide range of scientific areas to highlight the huge number of data handled in all of them. Nowadays, there are nearly ??? different named ?omics? ???, many of them related to systems biology sub-disciplines. This number is even expected to increase, as there is a tendency to dilute the knowledge from cellular components into different emergent sub-omics, so it is important to bear that there is an urgent need to integrate the knowledge from them. Therefore, existing omics are capable to provide comprehensive descriptions of nearly all components and interactions within the cell. Omics data sets can be generally classified into three categories [8,9]: components, interactions and functional-states data. Thus, components refer to the molecular content of cells or systems, interactions describe links between molecular components, and functional-states provide an integrated readout of all omics data types by revealing the overall cellular phenotype. To highlight some of the most relevant omics, DNA (genomics) is first transcribed to mRNA (transcriptomics) and translated into proteins (proteomics), which can catalyse reactions that act on and give rise to metabolites (metabolomics) such as oligosaccharides (glycomics) or lipids (lipidomics). Many of these components can be tagged and localized within the cell (localizomics). The processes responsible for generating and 2 0 Nuevas plataformas anal?ticas en metabol?mica modifying these cellular components are generally dictated by molecular interactions, for example by protein?DNA interactions in the case of transcription, and protein?protein interactions in translational processes as well as enzymatic reactions. Ultimately, the metabolic pathways comprise integrated networks, or flux maps (fluxomics), which dictate the cellular behaviour or phenotype (phenomics) [10]. Among omics sciences, metabolomics is one of the youngest disciplines but its impact is clearly justified by a rapid increase in the number of publications. Figure 2 shows the evolution in the number of publications on metabolomics as compared to proteomics. It is clear that there has been a tendency to increase its impact at least at the same pace than proteomics, which leads to the conclusion that its impact on systems biology will expectedly increase. Figure 2. Evolution of proteomics and metabolomics in the last fifteen years by number of devoted articles. To this point, metabolomics is mature enough to be granted the same status as other omics, supported by the existing technology and the 21 Introduction emerging of metabolite database in an attempt to compile the existing knowledge to be applied to different organisms from bacterial to mammals, including humans. The usefulness of metabolomics is justified by the direct connection between metabolome and phenotype. Intuitively, metabolites are further down the line from gene to function and so more closely reflect the activities of the cell at the phenotypic (functional) level, as highlighted in Figure 1B. In fact, physiological changes resulting from gene deletion or overexpression are amplified through the hierarchy of the transcriptome and proteome with a final result in the metabolome [11]. The formalism of Metabolic Control Analysis (MCA) [12,13] indicates that little changes in the expression level of individual proteins have large effects on the concentrations of intermediary metabolites. As a result, the metabolome is more sensitive to overall perturbations than the transcriptome and the proteome, because the activities of metabolic pathways are reflected more directly in the concentrations of metabolites than in those of the relevant enzymes or, indeed, those of their encoding mRNAs. As suggested by Kell [14], metabolomics is often deemed more discriminating than transcriptomics and proteomics for reasons such as the following: ? The number of metabolites is smaller than the numbers of genes and proteins, which reduces processing complexity. As an example, the human genome contains 31,896 genes (http://eugenes.org/) and the expression and alternative splicing of mRNAs indicates that humans may be able to produce 106 different proteins [15]; meanwhile, it has been estimated that a typical eukaryotic organism contains between 4,000 and 20,000 metabolites (approximately 7,900 metabolites for humans) [16]. 2 2 Nuevas plataformas anal?ticas en metabol?mica ? Metabolic profiling is more inexpensive and expeditious than proteomics and transcriptomics analyses. This favors high- throughput analysis. It is estimated that metabolomics experiments cost two-to-three times less than proteomics and transcriptomics experiments [17]. ? Unlike a transcript or a protein, a metabolite is the same for every organism that contains it. ? Finally, the technology involved in metabolomics is more generic and versatile. One major problem in the analysis of metabolites in comparison with the analysis of proteins and genes is the lack of totally comprehensive approaches. Here, it is important to design the analytical method according to the final aim of the metabolomics experiment, which may vary from the quantitative profile of few metabolites to the complete elucidation of the existing metabolic profile (metabolome) of one or a given set of biological samples under certain physiological conditions. 3. Strategies of analysis in me tabolomics There is a controversy about the use of the terms metabonomics and metabolomics. Metabonomics was formally defined in 1999 by J. ?icholson et al? ???? as? ?the quantitative measurement of the dynamic multiparametric metabolic response of living systems to patophysiological stimuli or genetic modifications?? On the contrary, the term metabolomics was introduced by Oliver Fiehn in 2001 [19] and defined differently as ?a comprehensive and quantitative analysis of all metabolites in a system?, which means that while metabonomics is mainly focused on changes of certain metabolite levels with external stimuli, metabolomics deals with the non-biased identification of all 23 Introduction metabolites in the biological system. Although some differences in the concept still remain, there is now an overlap in methodologies and the two terms are often used interchangeably by scientists. The information provided by both disciplines would lead to the complete metabolome, which is defined as the quantitative complement of low-molecular-weight metabolites present in a biological fluid, cell or organism under a given set of physiological conditions or under different perturbations (e.g., genetic variations, pathological states or responses to external stimuli) [20,21]. Metabolomics analysis encompasses different strategies depend- ing on the situation under study [17]: (1) target ing an a l ysis [22,23], which aims at qualitative and quantitative study of one or, more frequently, a small group of chemically similar metabolites; ( 2 ) g lobal m etabolomics profilin g [24,25], which allows detection of a broad range of metabolites by using a single analytical platform or a combination of complementary analytical platforms ?based on GC?MS, LC?MS, capillary electrophoresis (CE)?MS or NMR? to obtain a comprehensive profile of the metabolome; and (3) metabolomics fingerp rinting [26,27], a high-throughput, rapid strategy for analysis of biological samples that provides fingerprints for sample classification and screening. Figure 3 shows the differences between the three strategies in terms of data quality and number of covered metabolites. Data quality encompasses overall sensitivity, accuracy, precision of quantification, and metabolite identification rate [30]. Metabolite coverage is defined as the total number of detected candidates to metabolites (features), identified or not, achieved by one or various analytical platforms from a target biological sample. 2 4 Nuevas plataformas anal?ticas en metabol?mica Therefore, targeting analysis calls for highly selective and sensitive methods, as the information required is eminently quantitative. The most sensitive and precise analyses are typically those for single metabolites. Targeted methods for metabolite analysis provide high- quality data on a single class of compounds using dedicated and optimized methods. Depending on the field of application, it covers classical analytical methods used in many areas, e.g., in clinical analysis, quantification of biomarkers in biological fluids such as urine or plasma, or in food analysis, for target quantification of highly-valuable compounds such as antioxidants in food matrices. Thanks to its enormous development, there is a wide range of available analytical techniques as well as, in many cases, reference methods for validation. Figure 3. Comparison of metabolomics approaches through the binomial data quality-number of covered metabolites [24]. 25 Introduction Unlike targeting analysis, global metabolomics profiling can cover around several hundreds of metabolites, with concentrations encompassing several orders of magnitude. The purpose of metabolomics profiling is usually to obtain qualitative information, although quantitative or semi-quantitative analyses are also possible when different subjects or biological states are compared [31]. Broad metabolite profiling provides data for a wide range of chemical classes, but the methods represent a compromise and do not provide the same data quality for all of the metabolites covered. Therefore, analysis methods are preferably non-selective so as to cover the maximum number of detected compounds notwithstanding their chemical properties. These methods may be semi- or quantitative, but sensitivity is needed to be of several orders of magnitude. One of the main downside of global profiling is that available technology is usually more expensive and complex than for target analysis as more resolution capability is required. On the other hand, the scope of metabolomics fingerprinting is to compare patterns, signatures or ??fingerprints?? of metabolites that change in response to external stimuli (e.g., toxin or drug exposure, or environmental or genetic alterations [32]). For this purpose, selective (never specific) and quantitative information about metabolites is not necessarily provided. Metabolomics fingerprinting is a promising tool in clinical diagnosis, drug screening and toxicology studies [33,34]. Metabolite-fingerprinting approaches also offer the widest coverage, but profiles, and not metabolite levels, are compared by these approaches, and the methodologies can be subject to artifacts that undermine the quantitative assessment of the data. Thus, metabolic fingerprinting can be used as a diagnostic tool, for example, by evaluating a patient?s metabolic fingerprint in comparison to healthy and diseased subjects. In addition, the success of treatment strategies can be monitored by observing if the metabolic phenotype shifts back to the healthy state or, in other words, if a sample after treatment falls in the cluster of healthy 2 6 Nuevas plataformas anal?ticas en metabol?mica subjects. However, using metabolomics exclusively for fingerprinting without identifying the metabolites that cause clustering of experimental groups will provide only a classification but not directly contribute to biochemical knowledge and understanding of underlying mechanisms of action [35]. In a recent attempt to overcome this problem, multivariate discriminant analysis together with high resolution analysis techniques are being currently combined. 4. A nalytical pro cesses in metabolomics Identification, detection and quantification of large numbers of metabolites present at widely differing concentrations require the operating conditions of metabolomics methods to be optimized. In addition, variability in chemical and physical properties makes unfeasible to analyze the whole metabolic profile by using a single analytical platform. Figure 4 shows the general scheme followed in analytical chemistry, which also applies to metabolomics experiments with certain peculiarities depending on the used approach. Figure 4. General workflow of an analytical process. Firstly, sampling involves selection of the most suitable biological material for the aim of the analysis and should seek to obtain 27 Introduction representative analytical samples and ensure to preserve them. Sample preparation comprises extraction of the metabolites into a suitable solvent, preconcentration, clean up, and/or derivatization. These steps must be compatible with the detection step and their complexity varies with sample nature and technique used. Sample preparation deserves particular consideration due to its importance in metabolomics analyses. For this reason, Chapters 1 and 2 of this Thesis are devoted to critically review the existing sample preparation methodologies and their peculiarities in metabolomics with regards to the nature of the sample and type of approach followed. Not less importantly, detection techniques used in metabolomics, mainly NMR [36?38] and MS [35,39,40], have been widely reviewed in literature. Both are the most widespread detection techniques in metabolomics. The main analytical techniques used in the experimental development of this Thesis are discussed in the following sections. 5. Detection techniques Nuclear magnetic resonance spectrometry and mass spectrometry are the prevailing techniques in metabolomics thanks to their sensitivity and resolution capabilities. The latter technique usually requires to be hyphenated to another technique for separation of the metabolic components such as GC, usually after chemical derivatization, or LC. The use of CE coupled to MS has also shown to be promising. Other more specialized techniques such as Fourier transform infrared (FT-IR) spectrometry and fluorescence spectrometry are currently being exploited, but applications are by far less extended. 2 8 Nuevas plataformas anal?ticas en metabol?mica 5.1. NMR spectrometry 5.1.1. Principles of NMR In NMR spectrometry, liquid (or solid) samples are placed inside a detection coil, and exposed to an external, static and homogeneous magnetic field referred as to Bo (z-axis). NMR process is depicted in Fig. 5 [41]. The magnetic moments in the sample align along Bo according to a Boltzmann distribution (a). According to quantum mechanics, magnetic moments can only align parallel or anti-parallel with respect to this external field, leading to a macroscopic magnetic moment M. After a short application of a high frequency B1 field orthogonal to Bo (b), the magnetization is aligned with the rotating B1 field. Thus M, now aligned along x-direction (c), preccesses with a resonance frequency fo , which is proportional to Bo ( ). This preccessing process induces a voltage, Uind, modulated with fo (d) [42,43]. Due to the diamagnetism of the electron clouds of the molecules the local magnetic field experienced by a spin is changed by a small amount compared to Bo. Thus, the frequency fA measured for spin A directly reports the electronic/chemical neighborhood of the nucleus observed. The value (fA-fo)?106/fo, measured in parts per million with respect to Bo, is the chemical shift of this spin. This chemical shift is independent of Bo, thus data obtained at different values of field strength can be compared easily [44]. Detection of NMR signals can be only from atomic nucleuses possessing unpaired number of electrons (responsible for the resultant spin of these atoms, which align parallel or anti-parallel to the applied magnetic field), property termed as paramagnetism. In the absence of 29 Introduction magnetic field, nucleuses are rotating with randomized trajectories, but if a suited magnetic field is applied these nucleuses are aligned with it. Most elements have at least one isotope with a magnetic spin number greater than zero, necessary for the NMR effect. In the case of biological samples, the two more common nucleuses, 1H and 13C, give place to the two most frequent NMR techniques, 1H-NMR and 13C-NMR. 1H-NMR spectrometry is the most sensitive technique because of the natural presence of this nucleus. Thus, magnetization of hydrogen atoms experiencing a magnetic field of 14.1 T will rotate at 600 MHz aligned with the magnetic field. Such a spectrometer is called a 600 MHz equipment. This is the reason why NMR spectrometers are normally classified in frequencies. With the same magnet, a 13C spin will precess at about 150 MHz due to the natural presence of this nucleus (1.1%). Nowadays, there are available NMR magnets up to 1 GHz available, as the recently launched by Bruker [45] that operate at magnetic field strengths of 23.5 Tesla and proton NMR frequency of 1000 MHz with tremendously high field stability. Figure 3. NMR process. (a) A sample with unpaired spins is placed inside a strong magnetic field Bo (yellow) along z-direction. After short application of B1 (b), magnetization is aligned along x (c) and starts to precess around B, inducing a voltage U detected in the detection coil. (a) (b) Figure 5. NMR process. (a) A sample with unpaired spins is placed inside a strong magnetic field Bo (vertical arrow) along z-direction. After short application of B1 (b), magnetization is aligned along x (c) and starts to precess around B, inducing a voltage U detected in the detection coil (d). (a) (b ) (c) (d) 3 0 Nuevas plataformas anal?ticas en metabol?mica An NMR spectrum is composed by spectral amplitude plotted versus chemical shift. This method, called 1D-NMR, allows the determination of composition of a sample only if there is no severe overlap of signals. A method called 2D-NMR is used to overcome this limitation. It consists of a sequence of B1 pulses under systematic variation of an inter-pulse delay, leading to a 2Dmap where the amplitude is plotted as a function of two frequency dimensions. A cross- signal in this map indicates a pair of spins with a transfer of magnetization during the pulse sequence. This allows determining related signals of the spectrum (such as signals corresponding to a same molecule). 2D-NMR spectrometry can encompass homonuclear and heteronuclear techniques if the couplings are between the same or different nucleuses such 1H?1H and 1H?13C, as the most frequent couplings. 5.1.2. Characteristics of NMR Benefits and downsides of NMR spectrometry with respect to its use in metabolomics are depicted in Table 1. The great advantages of NMR are the potential for high throughput fingerprinting, minimal requirements for sample preparation, and the non-discriminating and non-destructive nature of the technique. In fact, 1H-NMR spectra can be acquired rapidly in 3?15 min with minimum sample preparation that usually just entails buffering and internal standard addition [46]. Furthermore, 1H-NMR has been largely used to unequivocally determine metabolite structures and, perhaps due to its non-invasive nature, is more commonly used in mammalian systems than mass-spectrometry technologies. However, only medium-to-high abundance metabolites will be detected with this approach and the identification of individual metabolites based on multivariate analysis of signals that cause sample clustering is difficult in complex mixtures. Indeed, the 1H-NMR spectrum of a complex sample is composed of several hundred resonances (with 31 Introduction chemical shifts between 0?10 ppm), which are estimated to arise from a hundred or more metabolites, resulting in a highly congested spectrum, given that each compound usually provide more than one signal each with a certain multiplicity. On the other hand, the reduced sensitivity of 13C-NMR has practically limited its application to the study of metabolic fluxes in cells and tissues by suited fed with isotopically labeled metabolites, thus given place to fluxomics. Table 1. Strengths and weaknesses of NMR for its use in metabolomics. Limitations of 1D-NMR have been partly overcome by the use of 2D-NMR and by the recently developed ultrahigh-field NMR spectrometers, which allow increasing spectral dispersion and reduce STRENGTHS WEAKNESSES ? Rapid and simple sample preparation required. ? Does not require chromatographic separation. ? High reproducibility. ? Relatively high throughput in 1D- NMR experiments. ? Useful for identification purposes. ? Versatility. Possibility of direct analysis of solids and in-vivo studies. ? Non-destructive. ? Difficult automatization despite robotized systems. ? Relatively poor resolution in complex samples. ? Poor sensitivity and selectivity. ? Very low throughput in 2D-NMR experiments. ? Difficult to unequivocally assign signals. ? Useful for fingerpringting, less suitable for global profiling and target analyses. ? Large sample volume (0.5 mL). 3 2 Nuevas plataformas anal?ticas en metabol?mica strong coupling-associated distortions. This has allowed its use in global profiling analysis. Together with ultrahigh-field equipment, the development of a technique called high resolution magic angle spinning (MAS), NMR spectrometry has broadened versatility of NMR in metabolomics. This technique is based on a rapid spinning of the sample (at 4?6 kHz) at an angle of 54.7? to the applied magnetic field, reducing peak broadening caused by nuclear dipole-dipole interaction. The reduction of this effect, which is more acute in solids, also reduce effects caused by sample heterogeneity; thus allowing direct analysis of tissues [47,48]. Finally, it is generally accepted that sensitivity in 1H-NMR is lower than in MS. This depends on magnetic field strength, with higher field magnets providing greater signal-to-noise ratio. This ratio can be increased by increasing acquisition time, which means increasing the number of scans. On the other hand, integration can be directly carried out by comparison with an internal standard by simple integration. Although recent methodologies that couple NMR to liquid chromatography and solid-phase extraction columns to improve separation and sensitivity have been reported [30,49,50], it is unlikely that NMR will attain the sensitivity of mass spectrometry. 5.1.3 Applications and perspectives of NMR In-vivo NMR analysis is probably the most outstanding application of NMR in metabolomics due to its use in diagnosis of diseases such as cancer. Interestingly, in-vivo ?MR could result in ??metabolite mapping?? of tissues after biopsy or before local ablative procedures. As an example, 1H-NMR analysis of breast biopsy samples have been used to identify over 30 endogenous metabolites in breast tissue. In addition, breast carcinogenic tissue reliably showed elevated total choline (tCho) levels (resulting from increased phosphocholine), low glycerophosphocholine, and low glucose 33 Introduction compared with benign tumors or healthy tissue ???????? Thus, when 1H- NMR of the breast is in-vivo performed before biopsy, precise differentiation of cancer and benign tissue is possible based on choline detection, which could lead to biopsy prevention on the choline-negative tissue. Similar to breast cancer, prostate cancer exhibits a distinct metabolic profile characterized by high tCho and phosphocholine levels, along with an increase in the glycolytic products lactate and alanine ???????? Metabolomics fingerprinting by NMR has proved to be an excellent approach to discover metabolic changes associated with diseases in humans [57] and genetic manipulation in plants [58]. NMR fingerprinting has been used for many years to authenticate foodstuffs, especially in the beverage industry. A recent study has employed the method to investigate grape quality [59]. The aim was to investigate the effect on grape berry skin metabolites of three cultivars grown over three seasons at five different geographical locations. Using standard methods of 1H-NMR data collection, followed by principal component analysis (PCA) and partial least squares (PLS) methods, the predictive modelling was able to pinpoint the spectral areas responsible for a separation according to vintage. Another remarkable application of 1H-NMR fingerprinting is to identify unintended effects from genetic modifications (GMs) in plants. Thus, 1H-NMR fingerprinting with multivariate analysis has been used to identify and classify maize seeds obtained from transgenic plants into different classes according to changes in metabolites [60]. Prediction to latent structures-discriminant analysis methods were used to build a predictive model that could identify GMs material by means of only 13 variables that explained over 90% of the variability. Metabolite profiling by 1H-NMR has recently been used for a genetic study of strawberry fruit quality, a functional study of tomato transformants and a study of Arabidopsis thaliana phosphoenolpyruvate transformants [61]. In the tomato study, a comparison of the roots of transformants with wild types showed that environmental factors significantly modified the 3 4 Nuevas plataformas anal?ticas en metabol?mica metabolic profile of plants, masking the expression of a given genetic background. Studies on the Arabidopsis transformants showed that a decrease in phosphoenolpyruvate carboxylase activity impacted on metabolic profile without compromising the plant growth. Although 1D-NMR studies are extremely useful in classifying similar groups of samples, it hinders unequivocal identification due to problems with large numbers of overlapping peaks, making actual identification of large numbers of metabolites difficult, and thus limiting its use for metabolic profiling. In 2D-NMR spectra, overlapping resonances are spread into a second dimension. There are various alternatives to carry out 2D- measurements, the most commonly employed being the Correlation Spectroscopy (COSY), Total Correlation Spectroscopy (TOCSY), Heteronuclear Single-Quantum Correlation Spectroscopy (HSQC), Heteronuclear Multiple-Bond Correlation Spectroscopy (HMBC) and Nuclear Overhauser Effect Spectroscopy (NOESY). The main limitation of these experiments is the high acquisition time required, tipically within 16?20 hours to achieve a comparable sensitivity to 1D spectra. This is not clearly feasible when hundreds of samples must be processed, but appears to be useful for identification and quantification and, thus, to develop target and global profiling analyses. However, it could be stated that identification in complex mixtures is still more difficult than in MS and requires large user interpretation. The use of 2D-NMR for metabolomics is usually restricted to the characterization of unidentified compounds from the 1D spectra. Alternatively, the increased resolution provided by the second dimension can allow characterization of components in a non-fractionated or partially fractionated mixture. Examples of this include the characterization of tomato juice [62] and the identification of the phenylpropenoids produced by methyl jasmonate treated Brassica rapa [63] and Arabidopsis thaliana [64]. Even with the improved resolution of 2D-NMR techniques, complete characterization of complex mixtures such as plant extracts by NMR is often 35 Introduction impossible. Hyphenating NMR to LC alleviates some of the problems by allowing NMR data to be collected on individual components of a mixture. Such on-flow NMR has been used as a means of screening the LC profiles of crude lipophilic extracts of aquatic plants for potential algacides [65,66]. The compounds of interest could not be completely identified from the LC?NMR analysis. However, it did give good clues as to the chemical nature of the constituents of the extract, thereby allowing targeted isolation of the most interesting compounds (labdane diterpenes) to be carried out. The routine use of on-flow LC?NMR for phytochemical analysis is limited by its lack of sensitivity and low throughput (the previous examples used LC runs of >?12 h). The advent of automated solid-phase extraction (SPE) peak trapping [67] has circumvented this problem and allowed LC? NMR to achieve its full potential. The technique has been used to investigate the composition of an African medicinal plant Kanahia ianiflora [68]. Alcoholic extracts of the plant were investigated by analytical scale LC?SPE? NMR using multiple peak trapping, to give sufficient of each of the major peaks to allow their complete characterization using 1D- and 2D-NMR techniques. Four flavanol-glycosides and three ??-cardenolides were successfully identified. Despite its relatively low sensitivity, targeting NMR analysis is gaining increased attention thanks to the broad range of applications that has recently appeared. In fact, quantification of biomarkers related to diseases or human metabolism, including schizophrenia [69], multiple sclerosis [70], diabetes [71], organ rejection [72], and rheumatoid arthritis [73] have already been published and exploited. Interestingly, NMR-based serum-lipoprotein quantification is now a mainstream clinical diagnostic tool. In-vivo identification and quantification is nowadays well implemented, with more than 80 papers published on the subject [74]. Currently, NMR-based metabolomics allows semi-automated identification and quantification of metabolites with concentrations as low 3 6 Nuevas plataformas anal?ticas en metabol?mica as ? ?M? This means that up to ?? different metabolites can be detected (in- vivo) in certain tissues and up to 100 metabolites (in-vitro) in certain biofluids. Given that the metabolome of many biofluids and tissues is of the order of thousand of metabolites [75,76], a major future challenge will be to find ways of increasing that number. In addition, provided that quantitative NMR metabolomics is carried out almost exclusively in aqueous conditions, it is not surprising that most of the quantifiable metabolites are water soluble, thus limiting considerably its applicability. Among its strengths, quantitative NMR has been applied in-vivo, with negligible sample preparation. In addition, no internal standard addition, apart from D2O (to serve as a frequency lock signal) and a chemical-shift reference standard (DSS or TSP) are required. It is worth mentioning the use of quantitative NMR for in-vivo applications. This approach was used for the first time to quantify brain metabolites [77] (glutamine, glutamate and gamma butyric acid) in patients suffering from epilepsy, that were proved to increase as compared to controls [78]. In other study, precise quantification of the major brain anti- oxidants (vitamin C and glutathione) at ?M levels is described, given that altered antioxidant profiles in the brain have been linked to schizophrenia [79]. Analysis of N-acetyl aspartate, creatine, choline and myo-inositol concentrations in the spinal cord are shown to significantly differ from those found in the brain stem. This may have implications in the diagnosis of multiple sclerosis and the monitoring of spinal cord injuries [80]. 5.2. Mass spectrometry 5.2.1. Principles of MS Mass spectrometry is a detection technique based on the differential displacement of ionized molecules through vacuum by applying an electrical field. Simplistically, a mass spectrometer consists 37 Introduction of an ion source, a mass analyzer, a detector and a data system. Sample molecules are inserted to the ion source, where they become ionized. The ions, which are in the gas phase, are separated according to their mass- to-charge ratio (m/z) in the mass analyzer and are then detected. Mass spectrometers are usually coupled to a separation unit, either chromatograph or capillary electrophoresis equipment ???????? To generate the gas-phase ions, there are different types of ionization sources, such as electron ionization (EI) and chemical ionization (CI), frequently used in GC?MS, while electrospray ionization (ESI) and atmospheric pressure chemical ionization (APCI) are frequently employed in LC?MS. Polar and ionic compounds are best suited for this type of ionization. For the purpose of this research, electrospray ionization analyzers have been mainly used. This type is particularly useful to be coupled to LC as ionization occurs at atmospheric pressure. Polar and ionic compounds are best suited for this type of ionization. In ESI, the sample is sprayed from a metal or fused silica capillary. An electrospray is achieved by raising the potential on the spray capillary to 4 kV in positive or negative ionization modes. The resulting spray of charged droplets is directed toward a counter electrode at a lower electrical potential, where the droplets lose solvent leading to ionic species into the gas phase. The counter electrode contains an orifice through which ions are transmitted into the vacuum chamber of the mass spectrometer, traversing differentially pumped regions via skimmer lenses [84]. Once the sample has been ionized, it is transported to the mass analyzer via an electric or magnetic field. There is currently a wide range of available mass analyzers. The most relevant to the research developed in this Thesis are here briefly described. i) Triple quadrupole mass spectrometers 3 8 Nuevas plataformas anal?ticas en metabol?mica A diagram of the design of a Triple Quadrupole MS is shown in Figure 6 [85]. Figure 6. General scheme of a Triple Quadrupole Mass Spectrometer. The triple quadrupole mass spectrometer consists of an ion source (usually ESI source) followed by ion optics that transfer the ions to the first quadrupole. The quadrupole is formed by four parallel rods to which specific DC and RF voltages are applied. These rods filter out all ions except those of one or more particular m/z values as determined by the voltages applied. The applied voltage is variable so that ions with other m/z values are allowed to pass through. Afterwards, selected ions reach a collision cell where they are fragmented. The collision cell is typically called the second quadrupole, but it is actually a hexapole, filled with an inert gas such as nitrogen or argon. The fragment ions formed in the collision cell are then sent to the third quadrupole for a second filtering stage to enable a user to isolate and examine multiple precursor to product ion transitions. This is called selected reaction monitoring or SRM. Since the fragment ions are pieces of the precursor, they represent portions of the overall structure of the precursor molecule.The voltages applied to the collision cell must be different from those applied to the 39 Introduction quadrupoles to enhance the movement of all of the product ions toward the third quadrupole. Due to the low-mass accuracy achieved with respect to other mass analyzers, triple quadrupole spectrometers are preferably used for target analyses, as they allow quantify with high sensitivity and selectivity in SRM mode. The most sensitive mode of operation for the triple quadrupole MS instrument is to fix both belts and only monitor a specific precursor ion and a specific product ion (SRM). In normal operation, a triple quadrupole MS instrument involves running multiple SRMs for the same precursor ions. ii) Quadrupole-Time-of-Flight mass spectrometers Figure 7 shows the diagram of the instrument used in the research of this Thesis [86]. Figure 7. General scheme of a Quadrupole Time-of-Flight Mass Spectrometer. 4 0 Nuevas plataformas anal?ticas en metabol?mica The Q?TOF mass analyzer is based on the same configuration as the QqQ but replacing the last quadrupole by an acceleration tube as mass analyzer (usually in orthogonal configuration) to filter out ions according to the equation of kinetic energy. The Q?TOF can operate in MS mode with the TOF by taking benefit from the high mass accuracy or in MS/MS mode for structural elucidation. This hybrid mass analyzer offers better selectivity than triple quadrupoles, meanwhile sensitivity is considerably lower. On the other hand, thanks to its great mass accuracy (below 2 ppm) highly reliable identification can be carried out, thus allowing its use for global metabolic profiling. 5.2.2 Characteristics of MS Benefits and downsides of MS with respect to NMR for its use in metabolomics are depicted in Table 2. The strongest point of MS is its sensitivity. In fact, mass spectrometry-based metabolomics offers quantitative analyses with high selectivity and sensitivity and the potential to identify metabolites with high mass accuracy, taking into account that sensitivity in MS is affected by both ionization and type of detector but, in general terms, it is much more sensitive than NMR. ESI ionization sources have sensitivities within the range of high femtomole to low picomole, with some nanoESI techniques having sensitivities as low as high zeptomole. MALDI and APCI sources have sensitivities within the femtomole range. CI and EI sources are less sensitive, with picomole sensitivities [87,88]. On the other hand, quantification requires suitable internal standards with similar ionization and fragmentation efficiencies. Quantitative information on a metabolite peak can be obtained in one of three ways: (i) by integrating it against a reference sample of the same compound (this requires the identity of the metabolite to be known and 41 Introduction compared between separate runs, which can introduce error); (ii) by comparing the relative ratios of a set of peaks across a series of spectra; and, (iii) by addition of a stable-isotope version of the metabolite of interest to the sample. Concerning qualitative analysis, tandem MS fragmentation patterns can give clues as to the connectivity and the presence of specific functional groups in an unknown metabolite. Identification of metabolites is often expedited by comparison to internal standards and searching against mass spectral libraries. Fragmentation of compounds under a certain applied voltage is very reproducible, thus MS/MS measurements are highly selective. STRENGTHS WEAKNESSES ? Highly sensitive in its different modes. ? Very selective (especially in MSn). ? Suitable for quantitative target analysis, global profiling and fingerprinting. ? Structural information throughout fragmentation patterns ( MSn). ? Accurate m/z measurements ideal for identification of metabolites. ? Excellent resolution (improved by separation). ? Detection depends on ionization efficiency. ? More efficient when coupled to separation techniques. ? Quantification requires chemically-related internal standard. ? Destructive, particularly complicated for analysis of valuable samples. ? Limited to liquid samples or extracts isolated from solid samples. ? Laborious sample preparation protocols. Table 2. Strengths and weaknesses of NMR for its use in metabolomics. 4 2 Nuevas plataformas anal?ticas en metabol?mica Although MS measurements are rapid, the overall run time depends on the chromatographic step, that may vary from minutes to hours (but never to days, as in 2D-NMR). However, combination with chromatographic or electrophoretic equipment reduces the complexity of mass spectra due to metabolite separation in a time dimension, besides delivering additional information on the physicochemical properties of metabolites. Among the weakest points, mass spectrometry-based techniques usually require sample preparation, which can cause metabolite losses. Complexity of sample preparation steps depends on the type of mass spectrometer but, on the whole, it is more tedious than in NMR. Thus, MALDI?MS techniques require the sample to be mixed with a suitable matrix prior to analysis. In liquid spray ionization sources, choice of solvent is important: ESI requires use of volatile organic solvents to aid in evaporation. GC?MS requires, most times, samples to be derivatized prior to detection. Furthermore, LC?MS is, by nature, unsuitable for in-vivo analysis. On the other hand, global profiling may be limited by the ionization type, due to the fact that specific metabolite classes may be discriminated based on the sample introduction system and the ionization efficiency. Despite of this fact, mass spectrometry is currently the most reliable platform for identification. Comparatively, it can be stated that mass spectrometry-based metabolomics is characterized by a number of annual publications exceeding those based on NMR. 5.2.3 Applications and perspectives of MS Mass spectrometry has also been used as detection technique in metabolomics fingerprinting and global profiling, apart from being the preferred technique for target analysis in several areas (i.e., pharmaceutics, 43 Introduction clinics, forensics, food, agriculture, etc.). MS is a powerful technique for molecular identification, especially through the use of tandem MS. The recent introduction of highly accurate mass spectrometers (Orbitrap and TOF analyzers) has enabled its use for global profiling with a relatively high throughput. Mass-spectrometric metabolic profiling has been countless used for diagnosis of metabolic disorders in different biofluids. Thus, the analysis of the urinary metabolome for diagnostics of metabolic disorders using GC? MS has recently been reviewed by Kuhara [89]. In addition, electrospray ionization with tandem mass spectrometry (MS/MS) has become an important tool for the assessment of disorders of amino acids metabolism, fatty acids, and organic acids biosynthesis, among others. Several reviews have recently been published, describing in detail the methodology and biochemical interpretation of data [90,91]. One of the stronger features of MS for global profiling is the possibility to characterize the non-polar fraction, and particularly, the lipidic fraction, that is generally termed as ?lipidomics?? The lipidome, comprising lipid classes, subclasses, and lipid signaling molecules, is an important sub- compartment of the metabolome and has been well studied through mass spectrometry [92] giving insights in biochemical mechanisms of lipid metabolism, lipid?lipid and lipid?protein interactions. As an example, metabolic profiling of eicosanoids, a class of lipid signaling molecules, was recently termed eicosanomics [93]. Target analysis usually requires sample preparation based on preconcentration and cleanup steps, which should be performed automatically by fully automated systems as the handly development is tedious, in addition to be properly optimized and validated. GC coupled to MS and other detectors has been used extensively for metabolomics and analysis of metabolites in a variety of samples, especially from plants [95]. However, the nature of GC makes its use only suitable for volatile compounds, or those derivatized to be volatile. Thus, non-volatile and thermally labile molecules are better analyzed by LC?MS (or by one of the other atmospheric pressure ionization modes) or by MALDI?MS. 4 4 Nuevas plataformas anal?ticas en metabol?mica Nevertheless, the application of suitable derivatization methods allows a significant proportion of the metabolome to be observed by GC?MS. Despite the need for derivatization, GC?MS offers a number of advantages over LC? MS. GC columns still offer higher resolution than conventional LC columns, and spectra obtained by using an electron ionization (EI) detector contain a wealth of structural information. The existence of extensive libraries of EI spectra (for example, http://www.nist.gov/srd/nist1a.htm), and reproducible retention indices often make compound identification easier by GC?MS than by LC?MS. A disadvantage of GC?MS when using EI is that the molecular ions may be minor or absent from the spectra; thus, information on the molecular weight of the starting molecule is absent. This can be overcome by recording chemical ionization (CI) spectra, where [M+H]+ or [M+NH4]+ ions, for example, tend to dominate. 5.3. Other techniques Although usually ignored in most of the reviews devoted to metabolomics, there are other detection techniques extensively exploited for metabolomics purposes. For instance, spectrophotometric techniques have traditionally been used for target analysis in an endless number of applications, including clinical routine analysis. Fluorescence and infrared spectrometries are briefly discussed here for their application in this Thesis. 5.3.1. Fluorescence spectrometry The key advantages of fluorescence detection for its use in metabolomics are high sensitivity, suitability for kinetics studies due to the rapid analytical response of instruments, possibility to use in probe format, nondestructive nature, and high-throughput. The main limitation of 45 Introduction fluorescence detection is that only very few metabolites can be analyzed directly by native fluorescence, making unsuitable this technique for global profiling. In most cases, fluorometric detection has been applied in targeted analysis of small-molecule compounds [95,96]. For example, An et al. [97] developed a method for label-free quantitative analysis of intracellular carotenoids in cells of red yeast (Phaffia rhodozyma) based on the intrinsic fluorescence of these compounds. Nevertheless, fluorescence detection usually requires derivatization, and chromatographic or electrophoretic separation. The use of laser-induced fluorescence (LIF) is quite extended in targeted analysis due to the high sensitivity of this technique, required for the low concentration of most of the existing metabolites. In an early study by Kennedy et al. [98], amino acids could readily be analyzed in extracts from individual neuron cells obtained from snails by CE with LIF detection after chemical derivatization. In other study, LIF detection was used to quantify gangliosides (large molecular-weight sphingolipids) in pituitary tumor cells [99] with acceptable sensitivity. As a novel and promising trend in metabolomics, vibrational-based spectroscopic techniques are rapidly gaining popularity in metabolomics. To date, the vast majority of metabolomic studies undertaken using vibrational spectroscopy have been carried out with FT-IR spectroscopy. However, some work has been carried out using Raman and, in terms of metabolomics, this is an emerging technology with significant potential for monitoring metabolites [100,101]. 5.3.2. Infrared spectrometry (IRS) FT-IR spectrometry is a well established analytical technique which allows for the rapid, high-throughput, non-destructive analysis of a wide range of sample types. FT-IR is based on the principle that when a sample is excited with an infrared beam, the functional groups within the sample will 4 6 Nuevas plataformas anal?ticas en metabol?mica absorb the infrared radiation and vibrate in one of a number of ways, either stretching, bending, deformation or combination vibrations. These absorptions/vibrations can then be correlated directly to (bio)chemical species and the resultant infrared absorption spectrum can be described as an infrared ?fingerprint? characteristic of any chemical or biochemical material [17]. With the requirements for a routine spectroscopic approach that could be made portable, minimum or null sample preparation and non- invasive, IR spectrometry appears as a candidate to metabolomics studies, particularly for disease diagnosis. Whilst selectivity and sensitivity are poor as compared with NMR and MS, the rapidity and reproducibility of FT-IR cannot be overstressed and it has been recognised as a valuable tool for metabolic fingerprinting/footprinting due to its ability to analyze simultaneously carbohydrates, amino acids, fatty acids, lipids, proteins and polysaccharides. In addition, portable FT-IR instruments are also available, allowing the analyst much scope in terms of spectral collection in a variety of environments. However, FT-IR does have some drawbacks, such as intense IR absorption of water and the need for external calibration. Most of metabolites give signals in the mid-IR (from 4000?600 cm-1) and near-IR (14000?4000 cm-1), both including fundamental vibration, overtones or harmonics. However, one potential disadvantage of mid-IR is that the absorption of water is very intense; problem that can be overcome in one of several ways such as; dehydration of samples (impractical in most of biological samples), subtraction of water signal, or by application of attenuated total reflectance (ATR) [102]. IR spectrometry has been widely used in metabolic fingerprinting applied to different areas. Therefore, it has been used in microbiology for the rapid and accurate identification of bacteria to the sub-species level [103] differentiation of clinically relevant species [104] and discrimination or identification of a range of bacterial genera [105]. Clinical applications of NIRS include analysis of faeces [106], follicular fluids to provide a biomarker 47 Introduction for oocyte quality [107], and synovial fluid to aid in the diagnosis of arthritic disorders [108]. In terms of analysis of serum, IR spectra could be used to discriminate between type 1, type 2 diabetes and healthy donors [109]. Metabolic profiling of athletes to detect doping and overtraining has also been investigated using a variety of body fluids, demonstrating that FT-IR could be applied to routine clinical analyses [110]. Attempts at cancer diagnostics include analysis of cell maturation in cervical tissue [111] and of colorectal adenocarcinoma [112]. On the other hand, quantitative target analysis is restricted to relatively abundant metabolites, due to the inherent low sensitivity of the technique. In addition, quantitative data can only be obtained by calibration using other techniques. As an example, it has been used to quantify lactate in human blood [113]. With respect to global profiling, there is little chance to use IR spectrometry for metabolite identification. Thus, applicability of IR is almost limited to fingerprinting [17]. 6 . Data treatment techniques an d databases 6.1. Data treatment techniques Metabolomics studies typically generate large amounts of data that complicates the use of univariate statistical analysis. However, if the concentration of a particular metabolite is found to be significantly altered through multivariate analysis, univariate analysis can be used to test the statistical significance of the change. This typically involves the use of the Student?s t-test or one-way analysis of variance (ANOVA). A more useful statistical approach is multivariate analysis, which can be applied to reduce large volumes of data into a few dimensions for classification and prediction of outcomes. It is worth noting that the 4 8 Nuevas plataformas anal?ticas en metabol?mica predictive power of multivariate analysis comes from patterns in the data, and identification of specific metabolites is not necessary to discriminate different classes of data. Both unsupervised and supervised learning methods have been devised [114]. In unsupervised strategies, no prior knowledge of the data is built into the statistical model. In contrast, supervised learning techniques require a training set of data to be used for prediction and classification. However, this also requires a separate set of data to test the predictions of the model, which can sometimes be problematic when not enough independent data to build both training and validation sets are available. This can be circumvented by using a data-splitting method such as crossvalidation, where the data are continually split into training and validation sets and the predictions of the data sets are averaged [115]. The most common unsupervised technique for identifying patterns and trends in metabolomics data is PCA, and this is often used as a starting point for analysis. PCA attempts to reduce multi-dimensional data so that it can be plotted in a two- or three-dimensional Cartesian coordinate system, with the axes (principal components) representing the greatest variations in the data. The popularity of this technique stems from the easy graphical interpretation of the data, with clustering often observed between data points based on class-type (age, gender, etc.). Contributions of individual variables to the separation of samples can be identified by corresponding loading plots, which plot the contribution of each variable based against selected principal components. PCA is usually followed by a supervised learning technique, such as partial least squares discriminant analysis (PLS-DA), which attempts to maximize the covariance between the independent and dependent variables to discriminate amongst samples. This method often discriminates between classes better than PCA does. Moreover, the lack of class information when determining the principal components in PCA can lead to discrimination of samples based on factors non related to the classes of interest. A specialized 49 Introduction form of PLS-DA is orthogonal projections to latent structures (OPLS), in which non-correlated systematic variance is removed from the model. Other less extended supervised learning techniques such as support vector machines (SVMs), artificial neural networks (ANNs) and classification and regression trees (CARTs) are by far less extended in metabolomics, although they find some applications [56,87]. 6.2. Standardization and database indexing The whole metabolomics experiment has a direct impact on the quality of the generated information, so it should be planned and validated as an integrated unit [116]. It is also important to consider that loads of metabolomics data are generated daily. In order to facilitate the exchange of information and the creation of databases, all metabolic profiling analyses should ideally be performed by a standardized method using identical instruments and operating conditions, as recommended by the the Metabolomics Standard Initiative (MSI). This initiative led to the MIAMET, or Minimum Information About a Metabolomics Experiment, which would act as an equivalent of MIAME for microarrays in transcriptomics (http://www.mged. org/Workgroups/ MIAME/miame.html). In contrast to genomics and proteomics where a number of databases are currently available, metabolomics databases are still growing. Biochemical databases can be used to identify unknown metabolites, for example, to identify structures from known elemental compositions, or to determine the biological function of the identified metabolite. Information about biochemical pathways and metabolites involved in given pathways are available in the KEGG (http://www.genome.jp/kegg/) and BioCyc (http://biocyc.org/) databases. More specific lipidomics databases exist, such as LipidMaps (http://www.lipidmaps.org/data/index.html), SphinGOMAP (http://sphingomap.org/), and Lipid Bank 5 0 Nuevas plataformas anal?ticas en metabol?mica (http://lipidbank.jp/index00.shtml), which contains structural and nomenclatural information as well as standard analytical protocols. General information about physico-chemical properties of metabolites can be obtained by searching general chemical databases such as PubChem or CAS. The development of specific metabolomics mass spectrometry databases is also in progress. The Golm metabolome database (GMD) (http://csbdb.mpimp-golm.mpg.de/csbdb/gmd/gmd.html) contains publicly available mass spectra and retention index metabolite (MSRI) libraries and GC?MS metabolic profiles of plant samples. The METLIN database (http://metlin.scripps.edu/) catalogues metabolites, MS/MS spectra, and LC?MS profiles, and also NMR spectra of human plasma and urine metabolites. To date, existing metabolomics databases aim primarily at the structural identification of metabolites in various biological samples. However, once a better annotation of the metabolome in various organisms is achieved, the generation of databases containing quantitative metabolite data can be expected. An example for this type of database is the human metabolite database that contains more than 1,400 metabolites found in the human body. Each metabolite is described by a MetaboCard designed to contain chemical data, clinical data and molecular biology/biochemistry data (http://www.hmdb.ca/) [117]. 7 . Ap p lications and f utu re trends Until recently, most of the studies devoted to metabolomics have mainly focused on clinical or pharmaceutical applications such as drug discovery, drug assessment, clinical toxicology, and clinical chemistry, cellular assays, environmental studies and plant metabolomics. However, over the past few years, metabolomics has also emerged as a field of increasing interest to many other areas, such as food chemistry and nutrition. Although the fields of application of metabolomics are 51 Introduction enormous, the most interesting with regards to this Thesis are highlighted below. 7.1. Plant metabolomics Plant metabolomics has been used in diverse applications including phenotyping of species, fingerprinting of genetically or environmentally modified plants, identification of metabolic differences between genetically modified and parental lines of crops, quantitative analysis of nutraceuticals or other potentially interesting compounds for human health or food industry. Main uses of metabolomics in plants analysis have been summarized in Figure 8. One of the most challenging aspects of plant metabolomics is to characterize the large number of metabolites present in plants, which are estimated to produce, across the plant kingdom, up to 200,000 different metabolites [11]. This global profiling approach, traditionally called phenotyping, has been used to study the metabolic profile of genetically modified plants such as tomato or potato. From an analytical point of view, global metabolic profiling usually requires complicated sample preparation steps as occurs in solid sample, especially in plant tissues as cells are surrounded by a thick wall that is necessarily broken before metabolite extraction. As for metabolic profiling of genetically modified plants, the potential for crop improvement by genetically modified traits is also enormous as has recently been demonstrated in studies on the so-called genetic metabolomics (linking metabolic profiles with genomic modification) [118] using Arabidopsis and tomato [119,120]. In Arabidopsis, F1 progeny derived from a same-species ecotype cross were surprisingly shown, when analyzed using LC?MS, to contain approximately 30% new mass peaks that were not present in either parent. These extensively new 5 2 Nuevas plataformas anal?ticas en metabol?mica chemical signatures, simply resulting from different intra-specific allelic combinations (transgressive segregation), reveal how variable genotypes are and the enormous potential of metabolomics to detect transgressive segregants before a trait is expressed. Another use of genetic metabolomics is to help to identify functions of genes. In the legume Medicago both isoflavonoid and triterpine pathways have been elucidated using LC?MS- based metabolomics [121,122]. Figure 8 . Applications of plant metabolomics. Fingerprinting in plant analysis has been extensively used for sample classification. The aim was usually to evaluate environmental factors on plant growth such as fluctuating growth conditions [123], exposure to cadmium [124] and herbicides, as well as to compare ecotypes, i.e., of Arabidopsis thaliana [125], between wild-type and transgenic genotypes of tomato [126]. Fingerprinting has also been applied to plant-derived products, for instance, to assess the authenticity of cocoa, used to produce high quality chocolate, or to detect adulterations or classify varieties of plant-derived products such as wine, juice [127] or oil. In plant fingerprinting studies, IR and NMR spectrometries are the outstanding 53 Introduction techniques due to the low sample preparation requirements and the possibility of direct analysis [128]. Applications of plant metabolomics aim to assess the quality of crops by evaluating both their safety and nutritional values. One of the key issues in the safety evaluation of transgenic crop plants is whether so-called ?unintended effects? may have occurred in plants as a result of genetic engineering, which could affect the safety or nutritional status of the crop or derived products. This is based on the hypothesis that potentially unintended effects of genetic modifications on human health or the environment will be connected to changes in metabolite composition. With regards to the nutritional value, it has been generally accepted the use of target analysis of highly valuable compounds (nutraceuticals) so as to improve crops quality. The capacity of some plant-derived foods to reduce the risk of chronic diseases has been in part associated to the occurrence of secondary metabolites (phytochemicals) that have been shown to exert a wide range of biological activities. Tomato has been selected in this Thesis as a model plant material for plant metabolomics due to its impact on Mediterranean diet and its well- known nutraceutical properties. Thus, tomato (Solanum lycopersicum) is widely consumed and well known for its health benefits, many of which have been associated with its high content in antioxidants. In fact, tomato is rich in carotenoids (lycopene), provitamins (provitamin A), polyphenols (flavonoids) and some vitamins (ascorbic acid). The large consumption of tomatoes implies a high production and also an increasing demand of quality. In this sense, tomato quality can be defined by both nutritional and organoleptics features, which are directly associated to metabolite composition. Therefore, the quality of tomato has a direct relationship with its metabolic profile, which is affected by different factors such as the variety, environmental effects, cultivating conditions, ripening stage and post-harvest storage. Phenotyping and target analysis of antioxidant compounds have been carried out with the final aim of obtaining a complete 5 4 Nuevas plataformas anal?ticas en metabol?mica profile of tomato metabolites by using different analytical techniques and by optimizing sample preparation. As tomato is rich in both polar and non- polar fractions, it is important to evaluate the effect of metabolites extraction on the final coverage of compounds. On the other hand, the continuous development of sensitive techniques makes necessary the implementation of more suitable target analysis protocols for antioxidant species, which have been traditionally carried out by non-selective photometric methodologies. 7.2. Nutrimetabolomics Over the past century great strides have been made by food scientists and biochemists in identifying the essential nutrients needed for human growth and health. Improved dietary guidelines and mandated food supplementation with essential minerals and vitamins has been remarkably successful in treating most nutritional ??deficiencies??? Nowadays, nutritional scientists are challenged with finding new ways of treating or preventing diseases brought on by nutritional ??oversufficiencies?? such as obesity, diabetes, chronic inflammation and cardiovascular diseases. It should also be worthwhile identifying bioactive food components that potentially increase life expectancy, reduce weight, enhance physical or mental performance and prevent diseases such as atherosclerosis, heart disease, cancer and arthritis. [129]. There should be highlighted some particular characteristics to be considered in nutrimetabolomics: ? There is a chemical transformation of food constituents after foods are cooked or digested. In addition, food intake may have an impact over metabolism of endogenous metabolites (i.e., fat mobilization) that may be reflected in the metabolic profile. 55 Introduction ? The effect of gut microflora is often associated only with the large bowel, but, depending on the biofluid, the roles of the oral microflora and of gastric colonization by Helicobacter pylori may also need to be factored into nutritional metabolomics. The microflora can change constituents in food and make them available to themselves or to the host for additional metabolism. ? The number of different molecules in the food supply that are not nutrients outweigh the number that are nutrients by orders of magnitude. For example, plants accumulate secondary metabolites for defense, reproduction, and so forth; however, none of these are essential nutrients. In traditional nutrition, these phytochemicals were mostly ignored until recently, when the potential metabolic effects of plant compounds were noted. ? Certain foods are known to produce obvious changes in urine in some individuals, which indicates a genotype interaction. For instance, in some individuals, beetroot produces red urine; in others, asparagus give rise to malodorous urine. ? The effect of food intake will persist over a certain period of time and may be different among metabolites. For example, metabolites of coffee are detected in urine collected 4?5 h after coffee ingestion, whereas heterocyclic amines from grilled meat persist after 48?72 h after intake [130]. There are three main approaches in nutrimetabolomics: (1) food component analysis; (2) food quality/authenticity detection; (3) changes caused by diet intervention studies. The most challenging and unexploited area in nutrimetabolomics is probably its use for monitoring diet intervention. This would be used to asses metabolic changes related to administration of certain nutrients that have an impact on human health throughout changes in metabolic profiles of biofluids. In fact, metabolism alterations derived from these 5 6 Nuevas plataformas anal?ticas en metabol?mica treatments can be effectively monitored through metabolic alterations in biological fluids, such as urine or serum. Metabolomics fingerprinting has been used to determine the effects of drug administration and diet variations on metabolism. Thus, the scheme followed in these studies entails: ? Selection of an adequate food material for diet intervention as well as the appropriate population to participate in the study. In order to obtain statistically representative results, sample size and inclusive criteria for both the action and control groups have to be preselected. Another key aspect is to establish the appropriate sample collection time after intervention. ? Analysis of biofluids after drug or diet intake. This analysis involves the group of patients, besides untreated individuals (control group) under the same analysis conditions. ? Supervised or non-supervised statistical analysis to find diet- related differences between groups. ? Determination of analytical features responsible of these differences. These may be a region in the NMR or NIR spectra or m/z from mass spectra. Identification of related metabolites and (relative) quantification of the resulting compounds, which implies to query against available databases. ? Elucidation of metabolic pathways in order to give as much biological information as possible. As an example, an NMR-based metabolomics analysis was used to demonstrate the metabolic impact of chronic cysteamine (CS) supplementation in rats [131]. CS has been successfully applied for many years to treat children with cystinosis. It may also be used as an endogenous regulator of immune system activity and it is also a potential therapeutic agent for the treatment of Huntington disease. Examination of PC loadings and subsequent inspection of the corresponding 1H NMR 57 Introduction spectra of urine samples enabled identification of endogenous metabolites, whose levels were perturbed by CS exposure. Thus, the approach allowed for the identification of biochemical markers related to CS supplementation. A key observation in this study is the impact of CS on intermediary metabolism in rats. Of the identified differential metabolites in the CS treatment group, succinate, an important intermediate in the tricarboxylic acid (TCA) cycle, was decreased. This suggests that the TCA cycle is downregulated by CS supplementation, which also implies decreased levels of citrate, another intermediate in the TCA cycle. In this Thesis, metabolomic fingerprinting has been used to explore the effect of the intake of different oils-rich meals on the organism. The aim was to determine if oils composition and, particularly, if the presence of antioxidants, would affect metabolism and how it is reflected in the urinary profile. The presence of antioxidants, naturally existing (or added), such as sterols, fatty alcohols, triterpenic dialcohols and unsaturated fatty acids, are known to exert a protective effect against coronary heart disease. Particularly olive oil, the principal source of fat in the Mediterranean diet, is known to have an adequate fatty acid profile and a rich composition in antioxidants [134]. Although benefits of olive oil and, concretely, of its antioxidants are gaining interest, the mechanisms involved in this protective effect are still unknown. Therefore, it could be worth exploring how metabolism is affected by oils intake and, particularly, how olive oil contribute to coronary heart disease (CHD) prevention. One of the main challenges pursuit in this research was to explore the composition of biological fluids that has not been received much attention in metabolomics. Thus, within the nutrimetabolomics context, biological fluids such as plasma, serum and urine [132,133] have been commonly used, as they are easily and non-invasively collected, directly reflect the global state of an individual and/or the response to drug treatment, and usually require the application of simple sample preparation 5 8 Nuevas plataformas anal?ticas en metabol?mica protocols. In contrast to the vast literature devoted to urine and blood metabolomics, there is a gap in studies on the analysis of other biological fluids, such as saliva, maternal milk, semen or cerebral spinal fluid. The main contributions of this Thesis to nutrimetabolomics have just been the metabolic profiling of biofluids that have been scarcely studied but with great potential in human nutrition, e.g., saliva and maternal milk. Particularly, maternal milk is the main or unique source of feeding for children during early stages of growing, which are marked by a rapid development and continuous intake of nutrients. Therefore, its high nutritional value should reflect a rich and varied composition; so metabolomics of maternal milk could be of great interest to assess its nutritional value, to identify biomarkers of certain diseases or to monitor its variability as a consequence of lactation state, food, supplementation, drugs intake or circadian variations. On the other hand, the use of saliva for metabolic profiling in order to identify candidate biomarkers has also been proposed. Saliva is not widely used in human nutrition research, but its inclusion could open new perspectives. In fact, saliva is a readily attainable biofluid that is rich in hormones such as 17-OH progesterone, testosterone, estradiol, and free cortisol. Its fatty acid composition has been used as a biomarker of plasma arachidonic acid, and it has been extensively studied for its antioxidant capacity [135]. Although saliva has not been used in metabolomic studies, its potential for distinguishing between metabolic profiles and for monitoring changes induced by diet would be worth exploring. The limited number of studies with this biofluid is reflected by the absence of databases and the limited bibliography, only target analyses have been addressed to date. This leads to the possibility to explore the global profile of saliva to complete current knowledge of the human metabolome [136]. 59 Introduction Figure 9. Workflow of metabolic fingerprinting to identify differences between control and action groups. 6 0 Nuevas plataformas anal?ticas en metabol?mica 7.3. Clinical analysis Metabolomics, with its impressive and ever increasing coverage of endogenous compounds and its intrinsic high-throughput capacity, provides a much more comprehensive assessment of human health and can be used in the identification, qualification, and development of biomarkers. An ever increasing number of metabolomics studies are used to identify pathomechanisms of complex diseases, to characterize phenotypes, and to develop diagnostic biomarkers [4]. Applications of this Thesis to the clinical area have been focused on the target analysis of biomarkers in human biofluids. Folic acid (vitamin B9) is a water-soluble vitamin involved in a broad variety of biological processes, as long as it acts as enzymatic cofactor in transference of methyl- group reactions [137]. It is known that severe deficiency of folate leads to numerous diseases associated to hindered cell division processes and deficiency of Red Blood Cells (RBCs), such as megaloblastic anemia, bone marrow, or fetal diseases (spina bifida, neural tube defects, etc.) [138]. Folic acid deficiency has also been associated to an increased risk factor for cardiovascular disease, atherosclerosis, and coronary heart disease [139,140]. Among the target analysis in clinics, lipids have been the bottleneck due to their particular characteristics. Lipids are eminently non-polar compounds that play widely diverse roles in metabolism, as expected by the chemical differences in this family. The importance of lipids is so enormous that lipidomics has evolved as a separate discipline because of the extraordinary structural diversity of lipids and their key roles in the pathophysiology of diseases. Lipidomics has been defined as the full characterization of lipid molecular species and their biological roles concerning expression of proteins involved in lipid metabolism and function, including regulation [139]. Although still an emerging field, lipidomics has 61 Introduction already provided promising new research possibilities; nevertheless, technological challenges remain. Here are some key aspects to consider about lipidomics analyses: ? Cellular lipids are widely diverse among species, cells organs and cellular microdomains. This class comprises eminently non-polar compounds with great differences in their chemical structures and concentrations. This complicates enormously lipidomics analyses. ? The lipidome of a given species varies dynamically depending on nutritional state, health conditions, lifestyle, exercise and disease. Lipids are usually present at very low concentrations. ? Most of the currently available analytical platforms are by nature unsuitable to non-polar compounds. For instance, analytical methods that involve reconstitution of a metabolite extract into an aqueous HPLC solvent are likely to be strongly discriminatory against its lipid content, for the simple reason that many lipids are not soluble in aqueous solvents. The aforementioned problems have made mandatory the development of dedicated target methods rather than global profiling approaches. Contributions of this Thesis to lipidomics have been focused on the target analysis of two lipid families, i.e., estrogens and sphingolipids: Estrogens: estrogens and progestogens are a group of female sterol lipid hormones derived from cholesterol, which are widely distributed in animals and humans. These hormones have mainly reproductive functions, being thus mainly distributed in breast, ovary, vagina and uterus [141]. Endogenous estrogens have a protective function against various diseases, such as osteoporosis, atherosclerosis and cardiovascular and neuro- degenerative diseases [142,143]. The analysis of steroid hormones in biological samples can be employed as a diagnostic tool in diseases promoted by disorders in the steroids profile. One of the main problems of 6 2 Nuevas plataformas anal?ticas en metabol?mica existing methods is that estrogens are usually excreted in a wide variety of conjugated compounds (mainly sulfates and glucuronides). This makes necessary a hydrolysis step to release the free forms before target detection. Sphingolipids: sphingolipids (SLs) are a family of lipidic compounds endowed with a long amino-alcohol chain, known as sphingoid base. Sphingolipids exert different functions in signal transmission and cell recognition. Sphingoid precursors ?sphinganine (Sa) and sphingosine (So)? act as structural and signalling molecules in membranes [144]. Sphingosine 1-phosphate, commonly present in human plasma and platelets, is an intermediate of sphingolipids catabolism [145], which inhibits the reproduction of certain tumoral cells, and also plays a key role both in the mobilization of intracellular calcium [144,146] and in cellular growth, differentiation, senescence, and apoptosis [147]. Few methods have been reported for simultaneous determination of sphingoid bases and phosphate derivatives. Current analytical methods for target analysis of sphingoid precursors are manual and, in many cases, they require large reagent and sample consumption, adding the complexity of a derivatization step when needed. For this reason, automation and miniaturization of sample preparation is desirable. 8 . References [1] The ?International ?Human Genome-Mapping ?Consortium. A physical map of the human genome. Nature 409 (2001) 934. [2] H. Kitano, Computational systems biology. Nature 420 (2002) 206. [3] J. Lin, J. Qian. Systems biology approach to integrative comparative genomics. Expert Rev. Proteomics 4 (2007) 107. [4] J.C. Lindom, J.K. Nicholson, E. Holmes. The Handbook of Metabonomics and Metabolomics. Elsevier, (2007). 63 Introduction ??? ???? ?icholson, I??? ?ilson? ?nderstanding ??lobal? systems biology? Metabonomics and the continuum of metabolism. Nat. Rev. Drug Disc. 2 (2003) 668. [6] H.V. Westerhoff, B.O. Palsson. The evolution of molecular biology into systems biology. Nat. Biotechnol. 22 (2004) 1249. [7] R. Goodacre. Metabolomics of a superorganism. J. Nutr. 137 (2007) 1S. [8] A.R. Joyce, B.?. Palsson. The model organism as a system: integrating ?omics? data sets. Nat. Rev. Mol. Cell Biol. 7 (2006) 198. [9] E.C. Butcher, E.L. Berg, E.J. Kunkel. Systems biology in drug discovery. Nat. Biotechnol. 22 (2004) 1253. [10] H. Ge, A.J.M. Walhout, M. Vidal. Integrating ?omic? information? a bridge between genomics and systems biology. Trends in Genetics 19 (2003) 551. [11] B. ?lvarez-S?nchez, F. Priego-Capote, M.D. Luque de Castro. Metabolomics analysis I. Selection of biological samples and practical aspects preceding sample preparation. Trends Anal Chem. 29 (2010) 111. [12] D.B. Kell, P. Mendes, A. Cornish-Bowden, M.L. Cardenas. Technological and medical implications of Metabolic Control Analysis. Kluwer Academic Publishers, London, UK (1999). [13] C.L. Winder, W.B. Dunn, S. Schuler, D. Broadhurst, R. Jarvis, G.M. Stephens, R. Goodacre. Global metabolic profiling of Escherichia coli cultures: an evaluation of methods for quenching and extraction of intracellular metabolites . Anal. Chem. 80 (2008) 2939. [14] D.B. Kell. Systems biology, metabolic modelling and metabolomics in drug discovery and development. Drug Discov. Today. 11 (2006) 1085. [15] J.E. Oh, K. Krapfenbauer, M. Fountoulakis, T. Frischer, G. Lubec. Evidence for the existence of hypothetical proteins in human bronchial epithelial, fibroblast, amnion, lymphocyte, mesothelial and kidney cell lines. Amino Acids. 26 (2004) 9. 6 4 Nuevas plataformas anal?ticas en metabol?mica [16] J.R. Bain, R.D. Stevens, B.R. Wenner, O. Ilkayeva, D.M. Muoio, C.B. Newgard. Metabolomics Applied to Diabetes Research Moving From Information to Knowledge. Diabetes 58 (2009) 2429. [17] W.B. Dunn, D.I. Ellis. Metabolomics: current analytical platforms and methodologies. Trends Anal. Chem. 24 (2005) 285. [18] J.K. Nicholson, J. Conelly, J.C. Lindon, E. Holmes. Metabonomics: a platform for studying drug toxicity and gene function. Nat. Rev. Drug Disc. (2002) 1153. [19] O. Fiehn. Metabolomics ?the link between genotypes and phenotypes. Plant Mol. Biol. 48 (2002) 155. [20] S.G. Oliver, M.K. Winson, D.B. Kell, F. Baganz. Systematic functional analysis of the yeast genome. Trends Biotechnol. 16 (1998) 373. [21] D.B. Kell, M. Brown, H.M. Davey, W.B. Dunn, I. Spasic, S.G. Oliver, Metabolic footprinting and systems biology: the medium is the message. Nat. Rev. Microbiol. 3 (2005) 557. [22] K.T. Myint, K. Aoshima, S. Tanaka, T. Nakamura, Y. Oda. Polar anionic metabolome analysis by nano-LC/MS with a metal chelating agent. Anal. Chem. 81 (2009) 1121. [23] H. Yoshida, J. Yamazaki, S. Ozawa, T. Mizukoshi, H. Miyano. Advantage of LC-MS Metabolomics methodology targeting hydrophilic compounds in the studies of fermented food samples. J. Agric. Food Chem. 57 (2009) 1119. [24] G. Theodoridis, H.G. Gika, I.D. Wilson. LC-MS-based methodology for global metabolite profiling in metabonomics/ metabolomics. Trends Anal. Chem. 27 (2008) 251. [25] C. Bottcher, E. von Roepenack-Lahaye, E. Willischer, D. Scheel, S. Clemens. Evaluation of matrix effects in metabolite profiling based on capillary liquid chromatography electrospray ionization quadrupole time-of- flight mass spectrometry. Anal. Chem. 79 (2007) 1507. 65 Introduction [26] E. Allard, D. Backstrom, R. Danielson, P.J.R. Sjoberg, J. Bergquist. Comparing Capillary Electrophoresis-Mass Spectrometry fingerprints of urine samples obtained after intake of coffee, tea, or water. Anal. Chem. 80 (2008) 8946. [27] H.G. Gika, G.A. Theodoris, I.D. Wilson. Liquid Chromatography and Ultra Performance Liquid Chromatography-Mass Spectrometry fingerprinting of human urine: sample stability under different handling and storage conditions for metabonomics studies. J. Chromatogr. A 1189 (2008) 314. [28] N.N. Kaderbhai, D.I. Broadhurst, D.I. Ellis, R. Goodacre, D.B. Kell. Functional genomics via metabolic footprinting: monitoring metabolite secretion by Escherichia coli tryptophan metabolism mutants using FT?IR and direct injection electrospray mass spectrometry. Comp. Funct. Genom. 4 (2003) 376. [29] O. Fiehn, J. Kopka, P. D?rmann, T. Altmann, R. N. Trethewey, and L. Willmitzer. Metabolite profiling for plant functional genomics. Nat. Biotechnol. 18 (2000) 1157. [30] A.R. Fernie, R.N. Trethewey, A.J. Krotzky, L. Willmitzer. Metabolite profiling: from diagnostics to systems biology. Nat. Rev. Mol. Cell Biol. (2004) 1. [31] R.N. Trethewey, A.J. Krotzkzy, The Handbook of Metabonomics and Metabolomics, Elsevier, Amsterdam, The Netherlands, (2007). [32] I. Garc?a-P?rez, M. Vallejo, A. Garc?a, C. Legido-Quigley, C. Barbas. Metabolic fingerprinting with capillary electrophoresis. J. Chromatogr. A 1204 (2008) 130. [33] D.I. Ellis, W.B. Dunn, J.L. Griffin, J.W. Allwood, R. Goodacre. Metabolic fingerprinting as a diagnostic tool. Pharmacogenomics 8 (2007) 1243. [34] D.I. Ellis, R. Goodacre. Metabolic fingerprinting in disease diagnosis: biomedical applications of Infrared and Raman Spectroscopy. Analyst 131 (2006) 875. 6 6 Nuevas plataformas anal?ticas en metabol?mica [35] K. Dettmer, P.A. Aronov, B.D. Hammock. Mass Spectrometry-based metabolomics. Mass Spectrom Rev. 26 (2007) 51. [36] M.R. Viant, E.S. Rosenblum, R.S. Tjeerdema. NMR-based metabolomics: a powerful approach for characterizing the effects of environmental stressors on organism health. Environ. Sci. Technol. 37 (2003) 4982. [37] M. Aalim, Weljie, J. Newton, P. Mercier, E. Carlson, C.M. Slupsky. Targeted profiling: quantitative analysis of 1H NMR metabolomics data. Anal. Chem . 78 (2006) 4430. [38] I.F. Duarte, I. Lamego, C. Rocha, A.M. Gil. NMR Metabonomics for mammalian cell metabolism studies. Bioanalysis 9 (2009) 1597. [39] C. Chen, F.J. Gonzalez and J.R. Idle. LC-MS-Based metabolomics in drug metabolism. drug metabolism Rev. 39 (2007) 581. [40] W. Lu, B.D. Bennett, J.D. Rabinowitz. Analytical strategies for LC?MS- based targeted metabolomics. J. Chromatogr. B 871 (2008) 236. [41] A. Ross, G. Schlotterbeck, F. Dieterle. H. Senn. NMR Spectroscopy techniques, The Handbook of Metabonomics and Metabolomics. Elsevier, 2007. [42] R. Ernst, G. Bodenhausen, A. Wokaun. Principles of Nuclear Magnetic Resonance in one and two dimensions. Oxford University Press, Oxford, 1990. [43] M. Goldman. Quantum description of high-resolution NMR in liquids. Oxford Universitiy Press, Oxford, 1991. [44] P.T. Callaghan. Principles of Nuclear Magnetic Resonance Microscopy. Oxford University Press, Oxford, 1991. [45] http://www.bruker-biospin.com/av1000-dir.html [46] M.R. Viant, C. Ludwig, U.L. G?nther. Metabolomics, Metabonomic and Metabolite Profiling. RSC Publishing, Cambridge, 2008. 67 Introduction [47] E.C. Yong Chan, P.K. Koh, M. Mal, P.Y. Cheah, K.W. Eu, A. Backshall, R. Cavill, J.K. Nicholson, H.C. Keun. Metabolic profiling of human colorectal cancer Using High-Resolution Magic Angle Spinning Nuclear Magnetic Resonance (HR-MAS NMR) Spectroscopy and Gas Chromatography Mass Spectrometry (GC/MS). J. Proteome Res. 8 (2009) 352. [48] J.C. Lindon, J.K. Nicholson. Analytical technologies for metabonomics and metabolomics, and multi-omic information recovery. Trends Anal. Chem. 27 (2008) 194. [49] M. Waisim, M.S. Hassan, R.G. Brereton. Evaluation of chemometric methods for determining the number and position of components in High- Performance Liquid Chromatography detected by Diode Array Detector and on-flow 1H Nuclear Magnetic Resonance Spectroscopy. Analyst 128 (2003) 1082. [50] J.C. Lindon. HPLC-NMR-MS: past, present and future. Drug Discov. Today 8 (2003) 1021. [51] T.F. Bathen, L.R. Jensen, B. Sitter. MS-determined metabolic phenotype of breast cancer in prediction of lymphatic spread, grade, and hormone status. Breast Cancer Res. Treat. 104 (2007) 181. [52] I.S. Gribbestad, B. Sitter, S. Lundgren. Metabolite composition in breast tumors examined by Proton Nuclear Magnetic Resonance Spectroscopy. Anticancer Res. 19 (1999) 1737. [53] K. Glunde, C. Jie, Z.M. Bhujwalla. Molecular causes of the aberrant choline phospholipid metabolism in breast cancer. Cancer Res. 64 (2004) 4270. [54] L.L. Cheng, C. Wu, M.R. Smith. Non-destructive quantitation of spermine in human prostate tissue samples using HRMAS 1H NMR spectroscopy. T. FEBS Lett, 494 (2001) 112. 6 8 Nuevas plataformas anal?ticas en metabol?mica [55] M.G. Swanson, A.S. Zektzer, Z.L. Tabatabai. Quantitative analysis of prostate metabolites using 1H HR-MAS Spectroscopy. Magn. Reson. Med. 55 (2006) 1257. [56] J.L. Spratlin, N.J. Serkova, S.G. Eckhardt. Clinical applications of metabolomics in oncology: a review. Clin. Cancer Res. 15 (2009) 431. [57] D.I. Ellis, W.B Dunn, J.L. Griffin, J.W. Allwood, R. Goodacre. Metabolic Fingerprinting as a diagnostic tool. Future Med. 8 (2007) 1243. [58] Y.L. Wang, H.R. Tang, J.K. Nicholson, P.J. Hylands, J. Sampson, I. Whitcombe, C.G. Stewart, S. Caiger, I. Oru, E. Holmes. Metabolomic strategy for the classification and quality control of phytomedicine: a case study of chamomile flower (Matricaria recutita L.). Planta Med. 70 (2004) 250. [59] G.E. Pereira, J.P. Gaudillere, C. Van Leeuwen, G. Hilbert, M. Maucourt, C. Deborde, A. Moing, D. Rolin. 1-H NMR metabolite fingerprints of grape berry: comparison of vintage and soil effects in Bordeaux grapevine growing areas. Anal Chim. Acta 563 (2006) 346. [60] C. Manetti, C. Bianchetti, M. Bizzarri, L. Casciani, C. Castro, G.D. Ascenzo, M. Delni, M.E. Di Cocco, A. Lagana, A. Miccheli et al. NMR-based metabonomic study of transgenic maize. Phytochemistry 65 (2004) 3187. [61] A. Moing, M. Maucourt, C. Renaud, M. Gaudillere, R. Brougisse, B. Leboutellier, A. Gousset-Dupont, J. Vidal, D. Granot, B. Denoves-Rothan et al. Quantitative metabolic profiling by 1-dimensional H-1-NMR analyses: application to plant genetics and functional genomics. Func. Plant Biol. 31 (2004) 889. [62] A. Sobolev, A. Segre, R. Lamanna. Proton High-Field NMR study of tomato juice. Magn. Reson. Chem. 41 (2003) 237. [63] Y.S. Liang, H.K. Kim, A.W.M. Lefeber, C. Erklens, Y.H. Choi, R. Verpoorte. Metabolic differentiation of Arabidopsis treated with methyl jasmonate using nuclear magnetic resonance spectroscopy. Plant Sci. 170 (2006) 1118. 69 Introduction [64] O. Hnedrawati, Q. Yao, H.K. Kim, H.J.M. Linhorst, C. Erklens, Y.H. Choi, R. Verpoorte. Identification of phenylpropenoids in methyl jasmonate treated Brassica rapa leaves using two-dimensional nuclear magnetic resonance spectroscopy. J Chrom. A 1112 (2006) 148. [65] P. Waridel, J.L. Wolfender, J.B. Lachavanne, K. Hostettmann. ent- Labdene diterpenes from the aquatic plant Potamogeton pectinatus . Phytochemistry 64 (2003) 1309. [66] P. Waridel, J.L. Wolfender, J.B. Lachavanne, K. Hostettmann. Ent- Labdene glycosides from the aquatic plant Potamogeton lucens and analytical evaluation of the lipophilic extract constituents of various Potamageton species. Phytochemistry 65 (2004) 945. [67] V. Exarchou, M. Godejohann, T.A. Van Beek, I.P. Gerothanassis, J. Vervoort. LC-UV-solid phase extraction-NMR-MS combined with a cryogenic flow probe and its application to the identification of compounds present in Greek oregano. Anal. Chem. 75 (2003) 6288. [68] C. Clarkson, D. Staerk, S.H. Hansen, J.W. Jaroszewski. Hyphenation of Solid-Phase Extraction with Liquid Chromatography and Nuclear Magnetic Resonance: application of HPLC-DAD-SPE-NMR to identification of constituents of Kanahia laniflora. Anal. Chem . 77 (2005) 3547. [69] M. Terpstra, M. Marjanska, P.G. Henry, I. Tkac and R. Gruetter. Detection of an antioxidant profile in the human brain in vivo via double editing with MEGA-PRESS. Magn. Reson. Med. 56 (2006) 1192. [70] A.F. Marliani, V. Clementi, L. Albini-Riccioli, R. Agati and M. Leonardi. Magn. Reson. Med. 57 (2007) 160. [71] A. Festa, K. Williams, A.J. Hanley, J.D. Otvos, D.C. Goff, L.E. Wagenknecht and S.M. Haffner. Nuclear magnetic resonance lipoprotein abnormalities in prediabetic subjects in the insulin resistance atherosclerosis study. Circulation 111 (2005) 3465. 7 0 Nuevas plataformas anal?ticas en metabol?mica [72] N.J. Serkova, Y. Zhang, J.L. Coatney, L. Hunter, M.E. Wachs, C.U. Niemann and M.S. Mandell. Early detection of graft failure using the blood metabolic profile of a liver recipient. Transplantation 83 (2007) 517. [73]A.M. Weljie, R. Dowlatabadi, B.J. Miller, H.J. Vogel and F.R. Jirik. A inflammatory arthritis-associated metabolite biomarker pattern revealed by 1H NMR Spectroscopy. J. Proteome Res. 7 (2007) 3456. [74] D.S. Wishart. Quantitative metabolomics using NMR. Trends Anal. Chem. 27(2008) 228. [75] D.S. Wishart, D. Tzur, C. Knox, R. Eisner, A.C. Guo, N. Young, D. Cheng, K. Jewell, D. Arndt, S. Sawhney, C. Fung, L. Nikolai, M. Lewis, M.A. Coutouly, I. Forsythe, P. Tang, S. Shrivastava, K. Jeroncic, P. Stothard, G. Amegbey, D. Block, D.D. Hau, J. Wagner, J. Miniaci, M. Clements, M. Gebremedhin, N. Guo, Y. Zhang, G.E. Duggan, G.D. Macinnis, A.M. Weljie, R. Dowlatabadi, F. Bamforth, D. Clive, R. Greiner, L. Li, T. Marrie, B.D. Sykes, H.J. Vogel, L. Querengesser. HMDB: the human metabolome database. Nucleic Acids Res. 35 (2007) D521. [76] D.S. Wishart, S.G. Villas-Boas, Editor, Metabolome Analysis: An Introduction, John Wiley & Sons, New York, USA, 2007. [77] S.W. Provencher. Estimation of metabolite concentrations from localized in vivo proton NMR spectra. Magn. Reson. Med. 30 (1993) 672. [78] R.J. Simister, M.A. McLean, G.J. Barker and J.S. Duncan. A proton magnetic resonance spectroscopy study of metabolites in the occipital lobes in epilepsy. Epilepsia 44 (2003) 550. [79] U.E Emir, S. Raatz, S. McPherson, J.S Hodges, C. Torkelson, P. Tawfik, T. White, M. Terpstra. Noninvasive quantification of ascorbate and glutathione concentration in the elderly human brain. NMR Biomed. 7 (2011) 888. [80] A.F. Marliani, V. Clementi, L. Albini-Riccioli, R. Agati and M. Leonardi. Magn. Reson. Med. 57 (2007) 160. 71 Introduction [81] R. Ramautar, G.W. Somsen G.J. de Jong. CE-MS in metabolomics. Electrophoresis 30 (2009) 276. [82] C. Chen F.J. Gonzalez, J.R. Idle. LC-MS-based metabolomics in drug metabolism. Drug Metab Rev. 39 (2007) 581. [83] S. Ma, S.K. Chowdhury, K.B. Alton. Application of mass spectrometry for metabolite identification. Curr. Drug. Metab. 7 (2006) 503. [84] W.J. Griffiths. Metabolomics, Metabonomics and Metabolite Profiling. RSC Publishing, 2007. [87] R.J. Mishur, S.L. Rea. Applications of Mass Spectrometry to Metabolomics and Metabonomics: detection of biomarkers of aging and of age-related diseases, Mass Spectrometry Reviews. DOI: 10.1002/mas.20338. [88] P. Krishnan, N.J. Kruger, R.G. Ratcliffe. Metabolite fingerprinting and profiling in plants using NMR. J. Experim. Botany 56 (2005) 255. [89] T. Kuhara. Gas Chromatographic-Mass Spectrometric urinary metabolome analysis to study mutations of inborn errors of metabolism. Mass Spectrom. Rev. 24 (2005) 814. [90] D.H. Chace, T.A. Kalas, E.W. Naylor. The application of tandem mass spectrometry to neonatal screening for inherited disorders of intermediary metabolism. Annu. Rev. Genomics Hum. Genet. 3 (2002) 17. [91] A. Schulze, M. Lindner, D. Kohlm?ller, K. Olgem?ller, E. Mayatepek, G.F. Hoffmann. Expanded newborn screening for inborn errors of metabolism by electrospray ionization-tandem mass spectrometry: results, outcome, and implications. Pediatrics 111 (2003) 1399. [92] J.B. German, M.A. Roberts, S.M. Watkins. Genomics and metabolomics as markers for the interaction of diet and health: lessons from lipids. J Nutr. 133(2003) 2078S. [93] M. Balazy. Eicosanomics: targeted lipidomics of eicosanoids in biological systems. Prostaglandins Other Lipid Mediat. 73 (2003) 173 . 7 2 Nuevas plataformas anal?ticas en metabol?mica [94] K. Dettmer, P.A. Aronov, B.D. Hammock. Mass spectrometry-based metabolomics. Mass Spectrom Rev. 26 (2007) 51 . [95] J. Berg, Y.P. Hung, G. Yellen. A genetically encoded fluorescent reporter of ATP:ADP ratio. Nat Meth 6 (2009) 161. [96] H.T. Chang, E.S. Yeung. Determination of catecholamines in single adrenal medullary cells by capillary electrophoresis and laser-induced native fluorescence. Anal Chem 67 (1995) 1079. [97] G.H. An, O.S. Suh, H.C. Kwon, K. Kim, E.A. Johnson. Cultivation of the carotenoid-hyperproducing mutant 2A2N of the red yeast Xanthophyllomyces dendrorhous (Phaffia rhodozyma) with molasses. Biotechnol. Lett. 22 (2000) 1031. [98] R.T. Kennedy, M.D. Oates, B.R. Cooper, B. Nickerson, J.W. Jorgenson. Microcolumn separations and the analysis of single cells. Science 246 (1989) 57. [99] C.D. Whitmore, O. Hindsgaul, M.M. Palcic, R.L. Schnaar, N.J. Dovichi. Metabolic cytometry. Glycosphingolipid metabolism in single cells. Anal. Chem. 79 (2007) 5139. [100] D.L. Utzinger, A. Heintzelman, A. Mahadevan-Jansen, M. Malpica, M. Follen, R. Richards-Kortum. Near-Infrared Raman Spectroscopy for in Vivo Detection of Cervical Precancers. Appl. Spectrosc. 55 (2001) 955. [101] A. Amantonico, P.L. Urban, R. Zenobi, Analytical techniques for single- cell metabolomics: state of the art and trends. Anal. Bioanal. Chem. 398 (2010) 2493. [102] E.B. Hanlon, R. Manoharan, T.-W. Koo, K.E. Shafer, J.T. Motz, M. Fitzmaurice, J.R. Kramer, I. Itzkan, R. R. Dasari, M.S. Feld. Prospects for in vivo Raman spectroscopy. Phys. Med. Biol. 45 (2000) R1. [103] D.I. Ellis, D. Broadhurst, D.B. Kell, J.J. Rowland, R. Goodacre. Rapid and Quantitative detection of the microbial spoilage of meat by Fourier 73 Introduction Transform Infrared Spectroscopy and Machine Learning. Appl. Environ. Microbiol. 68 (2002) 2822. [104] N.N. Kaderbhai, D.I. Broadhurst, D.I. Ellis, R. Goodacre, D.B. Kell. Functional genomics via metabolic footprinting: monitoring metabolite secretion by Escherichia coli tryptophan metabolism mutants using FT?IR and direct injection electrospray Mass Spectrometry. Comp. Funct. Genom. 4 (2003) 376. [105] M. Diem, S. Boydston-White, L. Chiriboga. Infrared Spectroscopy of cells and tissues: shining light onto a novel Subject. Appl. Spectrosc. 53 (1999) 148A. [106] T. Nakamura, T. Takeuchi, A. Terada, Y. Tando and T. Suda. Near- infrared spectrometry analysis of fat, neutral sterols, bile acids, and short- chain fatty acids in the feces of patients with pancreatic maldigestion and malabsorption. Int. J. Pancreatol. 23 (1998) 137. [107] R. Goodacre, E.M. Timmins, M. Gaudoin, R. Fleming. Fourier Transform Infrared Spectroscopy of follicular fluids from large and small antral follicles. Hum. Reprod. 15 (2000) 1667. [108] H.H. Eysel, M. Jackson, A. Nikulin, R.L. Somorjai, G.T.D. Thomson, H.H. Mantsch. A novel diagnostic test for arthritis: multivariate analysis of infrared spectra of synovial fluid Biospectroscopy 3 (1997). 161. [109] W. Petrich, B. Dolenko, J. Fruh, M. Ganz, H. Greger, S. Jacob, F. Keller, A.E. Nikulin, M. Otto, O. Quarder, R.L. Somorjai, A. Staib, G. Warner, H. Wielinger. Disease pattern recognition in Infrared spectra of human sera with diabetes mellitus as an example. Appl. Opt. 39 (2000) 3372. [110] C. Petibois, G. Cazorla, G. Deleris. Chemical mapping of tumor progression by FT-IR imaging: towards molecular histopathology. Appl. Spectrosc. 56 (2002) 10. 7 4 Nuevas plataformas anal?ticas en metabol?mica [111] L. Chiriboga, P. Xie, H. Yee, V. Vigorita, D. Zarou, D. Zakim and M. Diem. Infrared spectroscopy of human tissue I. Differentiation and maturation of epithelial cells in the human cervix. Biospectroscopy 4 (1998) 47. [112] P. Lasch, W. Haensch, E.N. Lewis, L.H. Kidder, D. Naumann. Characterization of colorectal adenocarcinoma sections by spatially resolved FT-IR microspectroscopy. Appl. Spectrosc. 56 (2002) 1. [113] D. Lafrance, L.C. Lands, D.H. Burns. Measurement of lactate in whole human blood with Near-Infrared Transmission Spectroscopy. Talanta 60 (2003) 635. [114] J.C. Lindon, J.K. Nicholson. Spectroscopic and statistical techniques for information recovery in metabonomics and metabolomics. Annual Rev. Anal. Chem. 1 (2008) 45. [115] G. Schlotterbeck, A. Ross, F. Dieterle, H. Senn. Metabolic profiling technologies for biomarker discovery in biomedicine and drug development. Pharmacogenomics 7 (2006) 1055. [116] K. Saito, F. Matsuda. Metabolomics for Functional Genomics. Systems Biology, and Biotechnology. Annu. Rev. Plant Biol. 61 (2010) 463. [ 117] K. Dettmer, A.P. Aronov, B.D. Hammock, Mass Spectrometry-based Metabolomics. Mass Spectrom. Rev. 26 (2007) 51. [118] J.J. Keurentjes, Genetical metabolomics: closing in on phenotypes. Curr. Opin. Plant. Biol. 12 (2009) 223. [119] J.J. Keurentjes, Y. Fu, C.H.R. De Vos, A. Lommen, R.D. Hall, R.J. Bino, L.H.W. van der Plas, R.C. Jansen, D. Vreugdenhil, M. Koornneef. The genetics of plant metabolism. Nat. Genet. 38 (2006) 842. [120] N. Schauer, Y. Semel, U. Roessner, A. Gur, I. Balbo, F. Carrari, T. Pleban, A. Perez-Melis, C. Bruedigam, J. Kopka, L. Willmitzer, D. Zamie, A.R. Fernie. Comprehensive metabolic profiling and phenotyping of interspecific introgression lines for tomato improvement. Nat. Biotechnol. 24 (2006) 447. 75 Introduction [121] R.D. Hall, D.I. Brouwerc, M.A. Fitzgerald. Plant metabolomics and its potential application for human nutrition. Physiologia Plantarum 132 (2008) 162. [122] A. Lommen, J.M. Weseman, G.O. Smith, H.P.J.M. Noteborn. On the detection of environmental effects on complex matrices combining off-line liquid chromatography and 1H NMR. Biodegradation 9 (1998) 513. [123] N.J.C. Bailey, M. Oven, E. Holmes, J.K. Nicholson, M.H. Zenk. Metabolomic analysis of the consequences of cadmium exposure in Silene cucubalus cell cultures via 1H NMR spectroscopy and chemometrics. Phytochemistry 62 (2003) 851. [124] J.L. Ward, C. Harris, J. Lewis, M.H. Beale. Assessment of 1H NMR spectroscopy and multivariate analysis as a technique for metabolite fingerprinting of Arabidopsis thaliana. Phytochemistry 62 (2003) 949. [125] H.P.J.M. Noteborn, A. Lommen, R.C. van der Jagt, J.M. Weseman. Chemical fingerprinting for the evaluation of unintended secondary metabolic changes in transgenic food crops. J. Biotechnol. 77 (2000) 103. [126] P.S. Belton, I.J. Colquhoun, E.K. Kemsley, I. Delgadillo, P. Roma, M.J. Dennis, M. Sharman, E. Holmes, J.K. Nicholson, M. Spraul. Application of chemometrics to the 1H NMR spectra of apple juices: discrimination between apple varieties. Food Chem. (1998) 207. [127] P. Krishnan, N.J. Kruger, R.G. Ratcliffe. Metabolite fingerprinting and profiling in plants using NMR. J. Exper. Botany 56 (2003) 410. [128] X. Zhang, Y. Yap, D. Wei, G. Chen, F. Chen. Novel omics technologies in nutrition research. Biotechnol Adv. (2008) 169. [129] D.S. Wishart. Metabolomics: applications to food science and nutrition research. Trends Food Sci. Technol. 19 (2008) 482. [130] S. Watanabe, M. Yamaguchi, T. Sobue, T. Takahashi, T. Miura, Y. Arai, et al. Pharmacokinetics of soybean isoflavones in plasma, urine and faeces of 7 6 Nuevas plataformas anal?ticas en metabol?mica men after ingestion of 60 g baked soybean powder (kinako). J. Nutr. 128 (1998) 1710. [131] R. Llorach, I. Garrido, M. Monagas, M. Urpi-Sarda, S. Tulipani, B. Bartolome, C. Andr?s-Lacueva. Metabolomics study of human urinary metabolome modifications after intake of almond (Prunus dulcis (Mill.) D.A. Webb) Skin Polyphenols). J. Proteome Res. 9 (2010) 5859. [132] R. Llorach, I. Garrido, M. Urpi-Sarda, O. Jauregui, M. Monagas, C. Andr?s-Lacueva. An LC-MS-Based Metabolomics Approach for Exploring Urinary Metabolome Modifications after Cocoa Consumption. J. Proteome Res. 8 (2009) 5060. [133] M.J. Gibney, M. Walsh, L. Brennan, H.M. Roche, B. German, B. van Ommen. Metabolomics in human nutrition: opportunities and challenges. Am. J. Clin. Nutr. 82 (2005) 497. [134] R. Jap?n-Luj?n, M.D. Luque de Castro. Liquid?liquid extraction for the enrichment of edible oils with phenols from olive leaf extracts. J. Agric. Food Chem. 56 (2008) 2505. [135] P.V. de Almeida, A.M. Gregio, A.M. Machado, A.A. de Lima, L.R. Azevedo, Saliva. Composition and functions: a comprehensive review. J. Contemp. Dent. Practice 9 (2008) 72. [136] C.J.L. Silwood, E. Lynch, A.W.D. Claxson, M.C. Grootveld. 1H and 13C NMR Spectroscopic Analysis of Human Saliva. J. Dent. Res. 81(2002) 422. [137] L.B. Bailey, J.F. Gregory. Folate metabolism and requirements. J. Nutr. 129 (1999) 779. [138] T. Tamura, M.F. Picciano. Folate and human reproduction. Am. J. Clin. Nutr. 83 (2006) 993. [139] B. Van Guelpen, J. Hultdin, I. Johansson, C. Witth?ft, L. Weinehall, M. Eliasson, G. Hallmans, R. Palmqvist, J.-H. Jansson, A. Winkvist. Plasma folate and total homocysteine levels are associated with the risk of 77 Introduction myocardial infarction, independently of each other and of Renal Function. J. Intern. Med. 266 (2009) 182. [140] E. Fahy, S. Subramaniam, H.A. Brown, C.K. Glass, A.H. Merrill, R.C. Murphy, C.R.H. Raetz, D.W. Russell, Y. Seyama,W. Shaw, T. Shimizu, F. Spener, G. van Meer,M.S. VanNieuwenhze, S.H. White, J.L. Witztum, E.A. Dennis. A comprehensive classification system for lipids. J. Lipid Res. 46 (2005) 839. [141] S.C. Sala, V. Martineti, A.M. Carossino, M.L. Brandi. Genetics and pharmacogenetics of estrogen response. Expert Rev. Endocrinol. Metabolism 2 (2007) 503. [142] R.A. Khalil. Sex hormone replacement therapy and modulation of vascular function in cardiovascular disease. Future Cardiol. 3 (2007) 283. [143] H.J. Tede. Sex hormones and the cardiovascular system: effects on arterial function in women. Clin. Exp. Pharmacol Physiol. 34 (2007) 672. [144] T.K. Gosh, J. Bian, D.L. Gill. Intracellular calcium release mediated by sphingosine derivatives generated in cells. Science 248 (1990) 1653. [145] S. Spiegel, S. Milstien. Sphingosine 1-phosphate, a key cell signaling molecule. J. Biol. Chem. 277 (2002) 25851. [146] A. Oliviera, S. Spiegel. Sphingosine-1-phosphate as second messenger in cell proliferation induced by PDGF and FCS mitogens. Nature 365 (1993) 557. [147] A. G?mez-Mu?oz, D.W. Waggoner, L. O?Brien, D.N. Brindley. Short- chain ceramide-1-phosphates are novel stimulators of DNA synthesis and cell division: antagonism by cell-permeable ceramides. J. Biol. Chem. 265 (1990) 21309. H ERRAMIENTAS Y EQUIPOS ANA L ?TICO S Analytical tools and equipment 81 Herramientas anal?ticas y equipos En este apartado de la Memoria se describen someramente los diferentes instrumentos y aparatos usados durante el desarrollo experimental de la Tesis. En los cap?tulos se incluye una explicaci?n m?s detallada de los que se han utilizado en cada uno de ellos. 1. Sistemas autom?ticos y/o continuos para la preparaci?n de muestra En el bloque 2 de la Memoria se ha tratado de enfatizar el uso de sistemas continuos para el tratamiento de muestra en el an?lisis orientado (targeting analysis) de distintas familias de metabolitos. Estos sistemas permiten llevar a cabo de manera reproducible, parcial o totalmente automatizada y con a veces dr?stica reducci?n del volumen de muestra y reactivos, esta etapa crucial del proceso anal?tico, que es una de las principales fuentes de error de los m?todos anal?ticos cuantitativos. Para la extracci?n l?quido?s?lido asistida por ultrasonidos de ?cidos haloac?ticos de muestras vegetales (Cap?tulo 7) se utiliz? un sistema de inyecci?n en flujo (FIA, actualmente FI), por el cual, mediante el paso de extractante en repetidos ciclos por la c?mara de extracci?n durante la irradiaci?n con ultrasonidos, se consigui? aumentar la eficacia del proceso. Se emplearon bombas perist?lticas y v?lvulas de inyecci?n y de selecci?n (siempre de baja presi?n) y tubos de tefl?n para la construcci?n del sistema FIA. La muestra, mezclada con arena inerte para evitar la compactaci?n y la formaci?n de caminos preferenciales, se situ? en una c?mara de extracci?n de acero inoxidable. La bomba perist?ltica trabajaba en los dos sentidos de circulaci?n del extractante para evitar sobrepresi?n en el sistema. La sonda de ultrasonidos utilizada fue una Branson 450 digital, que permite la 82 Nuevas plataformas anal?ticas en metabol?mica selecci?n de la amplitud de la radiaci?n as? como el modo irradiaci?n, continuo o discontinuo. Para el an?lisis de estr?genos (Cap?tulos 3 y 4) y esfingol?pidos (Cap?tulo ?) se utili?? un sistema de ?laboratorio en v?lvula? (lab-on-valve, LOV) para la preconcentraci?n y limpieza de la muestra basado en extracci?n en fase s?lida (SPE). Los sistemas LOV constituyen la tercera generaci?n de las t?cnicas de inyecci?n en flujo (progresivamente FIA, SIA y LOV), y est?n basados en la inyecci?n secuencial de muestra y reactivos en el sistema, utilizando vol?menes entre 100 ?L y 1 mL. Por este motivo el sistema se denomina ?meso-flu?dico? por utili?ar vol?menes a niveles ?meso? de muestra, entre micro- y mililitros. El sistema LOV utiliza los seis puertos de una v?lvula multiposici?n como canales de entrada y salida del flujo, que pueden usarse de forma secuencial, permitiendo al usuario programar de forma sencilla todas las etapas requeridas para la preparaci?n de muestra; lo que lo dota de una particular versatilidad. El sistema est? constituido por una microjeringa de un volumen m?ximo de 1 mL, que permite aspirar, dispensar, parar, acelerar y ralentizar el flujo; una v?lvula de 6 canales de plexigl?s, en la que los puertos est?n conectados entre s? y con la posici?n central de la v?lvula; un bucle de llenado entre la bomba y la v?lvula y una v?lvula de selecci?n de dos v?as para la introducci?n de un flujo adicional de disolvente portador o ?carrier?? La secuencia de pasos est? completamente automatizada y optimizada mediante el software FIAlab 5.0 para Windows. Una v?lvula externa adicional, tambi?n controlada por el software FIAlab, se us? para la recolecci?n del eluido despu?s de la SPE en el sistema. Las columnas de extracci?n fueron fabricadas en el laboratorio a partir de tubo Peek y material sorbente y se conectaron a uno de los puertos del LOV. Un sistema m?s sofisticado para llevar a cabo SPE es el Prospekt-2, que se us? para el pretratamiento de muestra en el an?lisis de folato y sus catabolitos, como se recoge en el Cap?tulo 6. Adem?s de trabajar en modo din?mico, la eluci?n directa con la fase m?vil cromatogr?fica y la posibilidad 83 Herramientas anal?ticas y equipos de trabajar a alta presi?n permite la conexi?n en l?nea de este sistema con el conjunto cromatogr?fo/detector. De esta forma se consigui? la automatizaci?n completa del m?todo anal?tico. El sistema Prospekt est? compuesto por tres m?dulos, un muestreador (MIDAS), una unidad de extracci?n en fase s?lida (ACE) y una bomba de alta presi?n dispensadora de disolventes de 2 mL de capacidad (HPD). 2. Sistemas n o continuos para la preparaci?n de muestra Dada la naturaleza no selectiva, semi- o no cuantitativa de las plataformas anal?ticas empleadas en los bloques 3 y 4 de la Tesis (dedicados al an?lisis metab?lico global y mediante huella dactilar, respectivamente), se enfatiz? el empleo de m?todos con m?nima o nula preparaci?n de muestra, no haci?ndose uso, por tanto, de sistemas continuos de pretratamiento de muestra. En estos casos, el pretratamiento de muestra se bas? en etapas de centrifugaci?n, extracci?n e hidr?lisis asistida por ultrasonidos o preconcentraci?n mediante evaporaci?n a vac?o. 3. Separaci?n cromatogr?fica y/o detecci?n Durante el desarrollo experimental de esta Tesis Doctoral se han desarrollado m?todos basados en separaci?n cromatogr?fica de l?quidos y gases, y en detecci?n mediante espectrometr?a de masas, de resonancia magn?tica nuclear y mediante detectores ?pticos (espectrometr?a de reflectancia en el infrarrojo y de fluorescencia molecular inducida por l?ser), as? como en captura electr?nica. 84 Nuevas plataformas anal?ticas en metabol?mica As?, en los Cap?tulos 3 y 4, dedicados al an?lisis orientado de estr?genos y progest?genos, en el Cap?tulo 6, para la cuantificaci?n de metabolitos del ?cido f?lico, y en el Cap?tulo 8, para compuestos nutrace?ticos de tomate (carotenoides y fenoles), la separaci?n por cromatograf?a l?quida y posterior detecci?n por espectrometr?a de masas en t?ndem por triple cuadrupolo se llev? a cabo con un cromat?grafo Agilent 1200 Series LC equipado con una bomba binaria, un desgasificador, un automuestreador y un compartimento de columna termostatizados, y un espectr?metro de masas Agilent 6410 con una fuente de ionizaci?n por electrospray (ESI). El software Agilent MassHunter Workstation se us? para la toma de datos y el an?lisis cuali- y cuantitativo. En el Cap?tulo 6, el an?lisis de folatos se realiz? con el mismo equipo; mientras que en m?todos de an?lisis de estr?genos y progest?genos, fenoles y carotenoides se usaron columnas cromatogr?ficas con fase estacionaria reversa C-18; para la separaci?n de folatos se utiliz? una fase estacionaria de interacci?n hidr?fila ?HILIC?, especialmente dise?ada para compuestos polares o i?nicos que tienen poca o ninguna retenci?n en las columnas de fase reversa. En los m?todos dedicados al perfil metabol?mico (metabolomic profiling) de muestras de saliva (Cap?tulo 9) y leche materna (Cap?tulo 10), y en el m?todo utilizado para obtener huellas dactilares a partir de muestras de orina (Cap?tulo 13) se utiliz? un equipo de HPLC acoplado a un detector de masas de tiempo de vuelo de alta resoluci?n, Agilent 6540, en todos los casos utiliz?ndose separaci?n cromatogr?fica en fase reversa. En todos los casos se us? el software MassHunter para la adquisici?n de espectros, el an?lisis cualitativo y semicuantitativo y la identificaci?n de metabolitos. En el Cap?tulo 5, dedicado al an?lisis de esfingol?pidos, se utiliz? un cromat?grafo ? -LC Agilent 1100, compuesto por una bomba binaria capilar, una columna micro en fase reversa (C-18), una v?lvula de inyecci?n autom?tica y un detector de diodos en fila. El sistema estaba conectado por tubos capilares y se usaron micro-vol?menes de muestra y de fase m?vil. Tras la salida de la c?lula de flujo del detector DAD, el sistema se conect? a 85 Herramientas anal?ticas y equipos un detector de fluorescencia inducida por l?ser (Zetalif 2000 con un l?ser de He-Ne, longitud de onda de emisi?n de 325 nm). El control del sistema cromatogr?fico y los detectores y la integraci?n de las se?ales se llevaron a cabo mediante el software Chemstation de Agilent. En cuanto a la cromatograf?a de gases se utiliz? acoplada a un detector de captura electr?nica para el an?lisis de ?cidos haloac?ticos (Cap?tulo 7), y a un detector de masas de trampa i?nica para el an?lisis de az?cares del tomate (Cap?tulo 8), ambos de Varian y en los dos casos usando el software de Varian de control de sistema e integraci?n de se?ales. En los m?todos de an?lisis no cromatogr?ficos se utilizaron un detector de resonancia magn?tica nuclear (RMN) de 500 MHz y un detector espectrofotom?trico de reflectancia en el infrarrojo cercano (NIR) ambos de ?ruker, para el an?lisis de ?huellas dactilares? (fingerprinting) de muestras de orina, llevados a cabo en el Instituto de Estudios Biofuncionales de la Universidad Complutense de Madrid y en el Servicio Central de Apoyo a la Investigaci?n de la Universidad de C?rdoba, respectivamente. 4. T?cnicas quimiom?tricas De acuerdo con la importancia que ha adquirido la quimiometr?a en los m?todos usados en metabol?mica, en la tesis doctoral se han utilizado extensamente herramientas quimiom?tricas, tanto para el desarrollo y optimizaci?n de m?todos anal?ticos como para el tratamiento de datos. La metodolog?a del dise?o de experimentos se ha utilizado, cuando ha sido posible, en la optimizaci?n de algunas etapas de la extracci?n en fase s?lida, aunque la naturaleza de algunas variables, generalmente discontinuas, hizo que su optimizaci?n fuera obligatoriamente univariante. La precisi?n de los m?todos propuestos para el an?lisis cuantitativo (targeting analysis) de familias de compuestos se estudi? como reproducibilidad dentro del 86 Nuevas plataformas anal?ticas en metabol?mica laboratorio y repetibilidad mediante series de experimentos por triplicado usando an?lisis de varianza (ANOVA) a diferentes niveles de concentraci?n de los analitos. El tratamiento de datos se realiz? con distintos programas inform?ticos: Statgraphics, The Unscrambler, Mass Profiler Professional y MatlAb, necesarios para realizar estudios univariantes mediante regresi?n lineal simple y regresi?n log?stica simple, as? como estudios de tratamientos multivariantes, mediante regresi?n log?stica multivariante, an?lisis por componentes principales y regresi?n por m?nimos cuadrados parciales, este ?ltimo en dos modalidades diferentes: An?lisis discriminante (PLS-DA) o modelado de clases (PLS-CM). Todos los estudios anteriormente citados se explican con mayor detalle en los siguientes cap?tulos de esta Tesis. 5. Bases de datos Existe una extensa variedad de bases de datos de metabolitos a disposici?n del usuario que contienen informaci?n para la identificaci?n y caracterizaci?n de compuestos presentes en diversas matrices biol?gicas. As?, las bases de datos Metabolites and Tandem MS Database (METLIN) y Human Metabolome Database (HMDB) se han usado para la ident ificaci?n de compuestos presentes en biofluidos (saliva, leche materna y orina) a partir de los datos espectrales obtenidos mediante espectrometr?a de masas y de resonancia magn?tica nuclear en las partes 3 y 4 de esta Memoria. P A RTE E XPERIMENTAL Experimental part P A RT 1 : Sample preparation in metabolomi cs Esta Parte I de la Memoria recoge investigaci?n no experimental realizada por la doctoranda mediante documentaci?n bibliogr?fica fundamentalmente con tres prop?sitos: (i) Conocer la situaci?n actual de las etapas que preceden a la detecci?n en metabol?mica, un aspecto al que generalmente se le concede poca importancia en los ?mbitos no anal?ticos y que no se hallaba sistematizado en ninguna publicaci?n. (ii) Sistematizar la investigaci?n existente de forma cr?tica y poner de manifiesto los aspectos de inter?s, los pobremente desarrollados y las lagunas existentes en otros. (iii) Ofrecer una visi?n cr?tica de la forma de resolver los problemas detectados en esta ?rea, bien con ejemplos basados en la investigaci?n desarrollada por la doctoranda, o con proposiciones de desarrollo de nueva investigaci?n a realizar en este campo. Con estos criterios, y dada su extensi?n, el tema a tratar se dividi? en dos partes: La primera dedicada a los aspectos previos a la preparaci?n de la muestra propiamente dicha (es decir, a la selecci?n de la muestra, clave para la obtenci?n de unos resultados representativos del problema en estudio, y la forma de almacenamiento de la muestra para una conservaci?n apropiada de las especies de inter?s). La segunda dedicada a la preparaci?n de la muestra en s? que, al tratarse siempre de muestras biol?gicas, entra?a una gran complejidad y puede verse afectada por numerosas fuentes de variabilidad. Part I of this book is devoted to the non-experimental research carried out by the PhD student, with three main purposes: (i) to know the forefront in sample preparation and other key steps developed in metabolomics prior to detection, aspects to which little importance is traditionally given; therefore, it had not been systematically studied in publications in the field. (ii) to critically systematize the existing research, clearly showing the interesting aspects, those poorly developed and the existing gaps in others. (iii) to offer a critical overview of the ways to solve the problems detected in this area, either with examples based on the research carried out by the PhD student or with proposals for the development of new research in this field. With these premises and taking into account the large extent of the subject, it was divided into two parts: the first (Chapter 1) was devoted to the steps prior to sample preparation ( viz. sample selection, a key aspect to obtain representative results on the target problem; and sample storage for proper preservation of the target species). The second part (Chapter 2) was devoted to sample preparation as such that, in this case ?always biological samples? entails a great complexity and can be affected by a number of sources of variability. CHAPTER 1: Metabolomics analysis I: Selection of biological samples and practical aspects preceding sample preparation Metabolomi cs analysis I: Selection of biologi cal samples and practi cal aspects preceding samp le preparati on B. ?lvarez - S?nc hez, F. Prie go - C apo te, M.D. Luque de Castro Department of Anal yti cal Chemistry, University of C?rdoba, Annex C - 3 Building, Campus of Rabanal es, E - 1 4 07 1 C?rdoba, Spain; and, In stitute of Bio medical Research Maim? nides (IMIBIC), Reina Sof?a Hospit al , Univers it y of C?rdoba, E - 14 07 1 C?rdoba, Spain Trends in Analytical Chemistry, 29 (2010) 111?119 99 Cha p t er 1 Trends A na l. Che m. , 29 (201 0 ) 111 ? 1 1 9 Met abo lomics analy sis I : Selectio n of bio l ogi cal sample s and practi cal aspect s prece di ng sam ple pre pa rat io n B. ?lvarez - S?nchez, F. Pri e go - C apo te, M.D. Luque de Castro Abst ra ct Metabolomics is one of the most recently emerged ??-omics?? sciences? Its significance in systems biology is gaining interest to levels similar to proteomics, transcriptomics and genomics. One of the main limitations in metabolomics analysis is the lack of totally comprehensive approaches and in-depth studies, as individuals or laboratories with different skill sets usually develop these. This variability particularly affects sample preparation due to the extensive heterogeneity of biological samples. The first part of this chapter focuses on analytical criteria for sample selection and operations that precede sample preparation. 100 N uevas plat a for ma s ana l ?t ica s en met a bol ?m ica 1. Intr oductio n One major problem in metabolomics analysis is the lack of totally comprehensive approaches, given the diversity and complexity of biological systems. Identification, detection and quantification of large numbers of metabolites present at widely differing concentrations require the operating conditions of metabolomics methods to beoptimized. The most widespread strategy for addressing these problems involves integration of various analytical platforms [e.g., gas chromatography combined with mass spectrometry (GC-MS), liquid chromatography combined with MS (LC-MS) and nuclear magnetic resonance (NMR)] in order to improve metabolite coverage and expand the range of identification. Great efforts are currently being made to improve the steps of detection and identification as well as the treatment of data generated to obtain biological information. However, other steps in the general workflow on metabolomics analysis have not received the same attention. These steps fall within the concept of sample preparation, which continues to be the bottleneck in the development of analytical methods. The main drawback of sample preparation in metabolomics is the lack of reference or conventional approaches to be applied. This can be justified by the heterogeneity of applications based on, e.g., sample diversity, variability of stimuli or perturbations, and metabolite composition. One example is found in protocols for interrupting metabolism in cell cultures that have to be optimized for each cell line in order to preserve the integrity of cellular membranes [1 ]. It is evident t hat there is a demand for the development of robust strategies for preparation of biological samples to be reproduced in different laboratories. The purpose of this chapter is to offer a guide about the main steps to be followed in preparing biosamples for 101 Cha p t er 1 Trends A na l. Che m. , 29 (201 0 ) 111 ? 1 1 9 detection (or separation?detection if a chromatographic or electrophoresis step is implemented). In contrast to sample preparation, the detection step has been reviewed for techniques {e.g., NMR [2 ] and MS (also combined with chromatography and capillary electrophoresis) [3 ]} applied to metabolomics analysis. This chapter aims at covering this unjustified gap in the metabolomics field and to show researchers in this field the need for more in-depth studies. Due to the heterogeneity of target samples in metabolomics experiments, sample preparation is an extensive topic. For this reason, those steps required for sample selection and operations prior to sample preparation are reviewed in this first part, and those steps involved in sample preparation prior to metabolite detection in the second part. Figure 1 illustrates a general workflow of the main steps involved in conventional metabolomics analysis. An experiment in metabolomics starts with selection of the biological material ? usually biofluids (e.g., blood or urine, cells, and, less frequently, tissues) ? and sampling, which is usually the limiting step in metabolomics and, ideally, should seek to be non-invasive and ensure representativeness. In addition, significant variations in chemical and physical properties or in the concentration of metabolites must be avoided. Although direct analysis is an ideal option pursued for characterization of liquid and solid samples, it is an infrequent option mostly linked to metabolomics fingerprinting, so most analytical methods applied to metabolomics involve a sample-preparation protocol, which is usually initiated with a rapid interruption (quenching) of the metabolism, particularly when working with cell cultures or biological materials that can be affected by enzyme action. Quenching provides a real snapshot of the metabolic state [1 ]. After metabolic quenching, cells are separated from the surrounding medium (containing extracellular metabolites) and, subsequently, 102 N uevas plat a for ma s ana l ?t ica s en met a bol ?m ica intracellular metabolites are extracted by effective permeation of the cell membrane. Figure 1. General workfl ow of the main steps invol ved in conventio nal metabolo mics anal ysis. Solid samples (e.g., tissues or plant material) require an extraction step for the metabolites of interest (target metabolites or profiling analysis). When working with biological fluids, liquid-liquid extraction is commonly employed to isolate metabolites, with optional steps for preconcentration and clean up [e.g., solid -phase extraction (SPE)]. The aim of SPE or equivalent tech techniques is to improve sensitivity and/or selectivity and SPE can also be applied to extracts from solid samples [4 ]. 103 Cha p t er 1 Trends A na l. Che m. , 29 (201 0 ) 111 ? 1 1 9 The complexity of the analytical method clearly depends on the purpose of the metabolomics study. If the goal is to develop a targeted method, it is evident that sample preparation should focus on the metabolites of interest. By contrast, global metabolomics profiling is generally carried out by non-selective methods in order to maximize coverage. After sample preparation, the resulting solutions are analyzed, analytical techniques such as NMR or MS being used most frequently [5,6 ]. Direct analysis by MS and NMR [7 ] are amenable for fingerprinting analysis without exhaustive, time-consuming sample-preparation steps. However, global and target metabolomics analyses usually require metabolite separation prior to detection, typically by conventional LC or the ultraperformance LC (UPLC) [8 ], or by GC after derivatization [9,10 ]. Less frequently used, capillary electrophoresis is considered a promising tool in metabolomics as it provides complementary information to that obtained by chromatographic techniques [11,12 ]. Elucidation of the metabolome is still a difficult task with unclear results. The complexity of the metabolome underlies the wide range of compounds with different physic chemical properties that makes unlikely obtaining reliable results from a single analytical platform. The metabolome for various well-studied organisms has been estimated from their genome. It is supposed to be approximately 800 metabolites for Escherichia coli [13 ], 600 metabolites for Saccharomyces cerevisiae [14 ], and up to 200,000 metabolites for the plant kingdom [15,16 ]. However, these estimates can be far from the real metabolome size, as there are still many genes with unknown functions [17 ]. The most feasible approach for comprehensive analysis of an organism metabolome is to integrate the information obtained from different analytical methodologies, aiming to increase metabolite coverage. The wide range of metabolites ?which include ionic (organic acids), polar, neutral (sugars) and non-polar (lipid) compounds? usually demand 104 N uevas plat a for ma s ana l ?t ica s en met a bol ?m ica application of complementary sample-preparation protocols. This results in several ??analytical samples?? ???] to be subjec ted to the detection step after individual separation, if required. Figure 2 shows a Venn diagram of the number of metabolites obtained using four different extraction procedures: methanol, ethanol, chloroform?methanol and potassium hydroxide. The diagram includes information about the number of metabolites detected with each protocol independently and with different protocols simultaneously for the metabolome analysis of E. coli [19 ]. The integrated analysis of all the results obtained by different methodologies obtains more complete results. Thus, a total of 264 peaks were obtained by combining the extraction methods. Similarly, the separation capability achieved by different chromatographic separation techniques (e.g., reversed-phase LC, hydrophilic interaction chromatography (HILIC) or normal-phase LC [20 ]) can be combined. The ionization source and polarity in MS analysis can also influence metabolite coverage. In general terms, the coverage achieved by electrospray ionization (ESI) is wider than by atmospheric pressure chemical ionization (APCI), but the overall coverage is considerably improved (estimated to be about 30% higher) when the results provided by both techniques are combined [21 ]. It is clear that metabolomics coverage is always a compromise between quality of data and throughput of analysis. Thus, while global metabolic profiling and metabolic fingerprinting are tools for large- metabolite coverage, the quality of the data is lower than in targeted profiling, where the method developed is exclusively optimized for one metabolite or a few metabolites [22 ]. However, a large volume of data is required for high coverage, which is clearly opposed to high-throughput analysis. 105 Cha p t er 1 Trends A na l. Che m. , 29 (201 0 ) 111 ? 1 1 9 Figure 2. Venn diagram of the co verage obt ained with different extract i o n pro to co l s for the anal ysis of E. col i . 2. Sample selecti on as the start ing point of metabo lomics analysi s One of the critical steps in metabolomics is sample selection, as the results generated will depend on its suitability. Thus, metabolomics analysis requires previous knowledge ofthe biological system, in some cases obtained from predictive models by means of metabolic modeling [23 ], which is based on existing databases (http://www.husermet.org) created from -omics data, and on informatics and chemometrics tools [24 ]. This information aids in the design of a suitable analytical protocol based on the nature and the particular characteristics of the sample. The existence of more than one biological material available for sampling enables the selection of the sample most suited for the analytical 106 N uevas plat a for ma s ana l ?t ica s en met a bol ?m ica problem under study. As an example, most clinical analyses are performed on biological fluids. Specifically, plasma, serum and urine have traditionally been used for prognosis or diagnosis of many diseases, as they are easily collected, reflect directly the global state of an individual and allow the biological response to drug therapy to be monitored. In addition, plasma and urine provide complementary information about the state of an organism [25 ]. Thus, pl asma gives an ??instantaneous?? readout of the metabolic state at the time of collection, and its composition directly reflects catabolic and anabolic processes occurring in the whole organism. However, urine provides an ??averaged?? pattern of easily-excreted polar metabolites discarded from the body as a result of catabolic processes. There are also other biofluids (e.g., saliva [26 ], amniotic fluid [27 ], cerebral spinal fluid [28 ], breast milk, synovial fluid, seminal plasma [29 ], bile, digestive fluids, or breathing air) that can provide valuable information, especially in the discovery of biomarkers for certain diseases. This is the case of monitoring biomarkers in cerebrospinal fluid for the diagnosis of ictus and in seminal serum for the diagnosis of male infertility [30 ]. It is worth mentioning that, although metabolic profiling of these biofluids is a challenging task, as they potentially provide complementary information to that from serum, plasma or urine, there are few reported attempts to date on this subject [28,31,32 ]. As an example, the brain is a challenging organ to work with due to its complex location. It is encapsulated within the blood-brain barrier, a membrane that limits the passage of many metabolites. In fact, it is questionable whether a biomarker produced within the brain may be detected in easily accessible biofluids, such as blood or urine [33 ]. Within the field of tissue metabolomics, the analysis of these materials offers particular benefits over biofluids as the spatial description of metabolite distribution can be accomplished. For example, direct study of tumor tissues results in the profile of existing metabolites distributed in the affected tissue. Monitoring some drugs in certain tissues (e.g., brain) gives information about their mechanisms of action and effects [34 ]. In the field of biomarkers, the greatest chance of discovering a novel biomarker resides in 107 Cha p t er 1 Trends A na l. Che m. , 29 (201 0 ) 111 ? 1 1 9 its screening within the target tissue [35 ]. Some remarkable drawbacks of tissue metabolomics are sample heterogeneity, the small availability of tissues and the invasive character of sampling techniques. An alternative can sometimes be to use microdialysates. The study of cell cultures is one of the most extended approaches in metabolomics. Metabolome analysis of cells usually distinguishes between extracellular and intracellular metabolites, which constitute the endometabolome and the exometabolome, respectively [36 ]. Thus, the samples are separated into two fractions, usually by a filtration step. Analysis of the liquid fraction obtains the profile of extracellular metabolites; meanwhile the separated cells, containing the intracellular metabolites, are subjected to extraction steps. Concerning plant metabolomics, the large variety of available samples [e.g., leaves , roots, sap, fruits, stalks, tubers, flowers, derived materials (e.g., oil, wine, resins)] and species is reflected in the wide number of reported studies in this field [37 ?39]. Thus, plants synthesize compounds in roots that are transported to shoots via the xylem sap. Some of these compounds are vital for signaling and adaptation to environmental stress, such as drought. The analysis of sap has therefore proved crucial to understanding how alterations in composition may lead to changes in development and signaling during adaptation to drought [40 ]. Another example is the analysis of fruits to elucidate the spatial metabolite analysis (e.g., the study developed by Biais et al. on melon [41 ]). Direct 1H NMR profiling of juice or GC-TOFMS profiling of tissue extracts collected from different locations in the fruit flesh revealed several gradients of metabolites, which can be related with differences in metabolism. The final aim of metabolomics is to integrate raw data into databases that serve as potent tools for diagnosis or development of predictive models. Accordingly, the management of a large number of samples is required in order to obtain robust, transferrable results to establish comparisons between organisms. In addition, this involves minimizing the main sources 108 N uevas plat a for ma s ana l ?t ica s en met a bol ?m ica of variability between samples [25 ], which can be due to intraindividual or inter-individual factors ?mainly caused by physiological state, or dietary, environmental, genetic or pathophysiological conditions. Variability between subjects is more noticeable in some types of biological samples. Some matrices (e.g., plasma, serum, tissues and cerebral spinal fluid) are physiologically regulated by homeostatic control, the metabolite profile being relatively constant within healthy specimens. However, composition of urine samples can widely vary inter-individually and even intra- individually, depending on urine volume, water and food intake and other physiological conditions (e.g., age, sex, weight and environment) [42 ]. 3. Sample storage and pre limina ry operati ons The control of the potential sources of variability exposed above is crucial to avoid errors in data interpretation. After inter-individual and intra-individual variations, the main source of error in metabolomics is associated with sampling and post-collection procedures (e.g., freeze-thaw cycles or inadequate storage conditions [ 42] ). It is known that unsuitable sampling and samplepretreatment protocols can lead to biased results due to conversion or degradation of metabolites [43 ]. There is increased interest in rapid collection and handling of samples for metabolomics purposes, while turnover kinetics of some metabolites is known to be extremely fast. For example, for intermediates in energy metabolism (e.g., ATP, ADP and glucose 6-phosphate), turnover rates are 1.5?2.0 s for Saccharomyces cerevisiae [44 ]. Thus, sampling techniques, particularly in the case of cells and tissues, need to be fast enough to ensure that the metabolic profile reflects in vivo conditions. Accordingly, the time window between sampling and analysis has to be as short as possible. In some cases, rapid inactivation of metabolism (also known as quenching) is performed with this aim. Nevertheless, immediate analysi of the samples is impossible in some 109 Cha p t er 1 Trends A na l. Che m. , 29 (201 0 ) 111 ? 1 1 9 occasions so storage is required (e.g., in banks of biological samples for research purposes). Sample storage is another critical reason for errors in metabolomics analysis. Most metabolites are preserved if samples are immediately frozen at below -80? C (e.g., by using liquid nitrogen). However, it is worth noting that differences in storage time or frequent thaw/freeze cycles may have a strong influence on the development of metabolomics models. As an example, Lauridsen et al. found that human-urine samples can be stored at or below 25? C for 26 weeks without changes in the 1H NMR fingerprints. Formation of acetate, presumably due to microbial contamination, was occasionally observed in samples stored at 4? C without addition of a preservative. Freeze-drying of urine and reconstitution in D2O at pH 7.4 resulted in the disappearance of the creatinine CH2 signal due to deuteration [57]. The influence of sample storage has been widely studied in biofluids (e.g., serum and plasma). The strategies for sampling and storage of biofluids for metabolomics studies are especially important, compared to proteomics and transcriptomics. This is justified by the metabolic activity time-scale (metabolic reaction halflives are often <1 s). Metabolic activity during sampling and storage requires stopping or minimizing changes in the metabolic profile either in concentration or structure. For this purpose, reduced temperatures during sample preparation (4?C) and storage ( -80? C) are common [20,25 ]. It is important to keep in mind the purpose of metabolomics analysis. Thus, the variability associated with sample storage or preliminary operations can be critical in targeted metabolomics approaches focused on a restricted set of metabolites that could be seriously affected. However, small changes observed for a reduced number of metabolites in a global profiling analysis can be considered acceptable in studies involving large sample sizes. This can be justified by greater inter-individual variability than that associated with sample storage and preparation. 110 N uevas plat a for ma s ana l ?t ica s en met a bol ?m ica A well-planned study based on global profiling with GC?TOF/MS has been performed by the UK Biobank to assess the stability of serum and urine [45 ]. The study comprised analysis of biofluids stored at 4? C for two different time periods (0 and 24 h) before storing at -80? C. The profiling analysis involved the detection of more than 700 metabolite peaks, which were studied by statistical analysis to assess possible changes in the urine metabolome. Figure 3 illustrates the combination of metabolomics analysis for serum and urine samples subjected to this study. Principal components analysis showed metabolic differences for a small number of serum and urine samples. Nevertheless, univariate analysis revealed that these differences were not statistically significant, as they were associated with a small number of metabolites. As a result, the variance in the metabolome of a single subject stored at -80?C or 4? C for 24 h is small, when compared with the variance in the metabolomes of 40 healthy volunteers. This inter- individual variance can be ascribed to genotypic differences, but also to many phenotypic factors (e.g., diet, health and lifestyle, and diurnal and estrus cycles). The concluding report of this research was that the UK Biobank sample coll ection, transport and fractionation protocols, involving the storage of serum and urine samples at 4? C for 24 hours and well- controlled UKbased transport, are suitable for high-resolution metabolomics studies. The scenario changes completely in the analysis of a group of target metabolites. The potential effects of pre-analytical variables associated with storage and pretreatment of human blood were studied for analysis of endocannabinoids and metabolites. Storing fresh blood at 4? C selectively enhanced the concentration of ethanolamide without altering monoglycerides and nonesterified fatty acids (e.g., arachidonic acid). By contrast, ethanolamide and monoglyceride concentrations were not altered through three plasma freeze/thaw cycles, whereas plasmatic arachidonic acid increased, probably due to reflection of ongoing metabolism [46 ]. 111 Cha p t er 1 Trends A na l. Che m. , 29 (201 0 ) 111 ? 1 1 9 Similar studies have been carried out to evaluate the influence of preliminary steps. Saude et al. [47 ] found that metabolite composition of the sample is maintained throughout the collection and analytical process with the addition of a bacteriostatic preservative (e.g., sodium azide). Urine filtration enhanced metabolite preservation, as ascribed to removal of bacterial contamination. Finally, freeze/thaw cycles should be avoided in metabolomics protocols. For samples with high-humidity content (e.g., those from plants), lyophilization (drying under vacuum at reduced heating) is recommended to improve sample stability. Lyophilized samples stored for a preset time should be lyophilized again before analysis as these samples can absorb moisture during storage. 4. Conc lusio ns Sample preparation has not received enough attention in metabolomics compared to detection, particularly by NMR and/or MS. However, metabolomics analysis cannot be efficiently completed without a well-planned sample-preparation protocol. We have reviewed preliminary steps to be considered in a metabolomics analysis by encompassing the different criteria for sample selection and storage and those operations to be carried out prior to analysis. The relevance of these steps to the quality of the final results is evident. These preliminary steps are an important part in MIAMET (Minimum Information About METabolomics Experiment) [53] reports to ensure the reproducibility of results. 112 N uevas plat a for ma s ana l ?t ica s en met a bol ?m ica A ckno wledgments Figure 3. (a) Gl o bal metabo lo mics pro filing of serum and urine anal yzed wi th GC ? TOF / MS by the UK Bio bank to assess the stabilit y of biofl uids. (b) Principal Compo nents Anal ysis scores (PC) plots fro m 40 vol unteers obt ained by anal ysis of GC ?TOF / MS pro fil es. The two sets of sampl es were sto red at 4 ? C (0 h and 24 h) befo re sto rag e at ?8 0 ? C. PC 1 and 2 represent 8% and 5% of the variance, respect ivel y. c) Loadings plots fo r Principal Compo nents PC1 and PC2 for serum sampl es. Shaded area indi cates the metabo lit es respo nsible for the erro neo us sampl es in the PC loadings plo t. (Repro duced with permission of Oxfo rd University Press, Reference [60 ]). 113 Cha p t er 1 Trends A na l. Che m. , 29 (201 0 ) 111 ? 1 1 9 5. Ackno wledgement s The Spanish Ministerio de Ciencia e Innovaci?n (MICINN) is acknowledged for financial support through project CTQ2009-07430. F.P.C. is grateful to MICINN for a Ram?n y Cajal contract (RYC -2009-03921). B.A.S. is also grateful to Ministerio de Ciencia y Tecnolog?a for an FPI scholarship (BES -2007-15043). 6 . Referenc es [1 ] C.L. Winder, W.B. Dunn, S. Schuler, D. Broadhurst, R. Jarvis, G.M. Stephens, R. Goodacre, Anal. Chem. 80 (2008) 2939. [2 ] D.S. Wishart, Trends Anal. Chem. 27 (2008) 228. [3 ] M. Bedair, L.W. Sumner, Trends Anal. Chem. 27 (2008) 238. [4 ] V. Exarchou, M. Godejohann, T.A. van Beek, I.P. Gerothanassis, J. Vervoort, Anal. Chem. 75 (2003) 6288. [5] W.B. Dunn, D.I. Ellis, Tre nds Anal. Chem. 24 (2005) 285. [6 ] W.B. Dunn, N.J.C. Bailey, H.E. Johnson, Analyst (Cambridge, UK) 130 (2005) 606. [7 ] N.E. Simpson, Z. He, J.L. Evelhoch, Magn. Reson. Med. 42 (1999) 42. [8 ] S.J. Bruce, I. Tavazzi, V. Parisod, S. Rezzi, S. Kochhar, P.A. Gu y, Anal. Chem. 81 (2009) 3285. [9 ] J.K. Nicholson, J.C. Lindon, Nature (London) 455 (2008) 1054. [10 ] T.O. Metz, J.S. Page, E.S. Baker, K. Tang, J. Ding, Y. Shen, R.D. Smith, Trends Anal. Chem. 27 (2008) 205. 114 N uevas plat a for ma s ana l ?t ica s en met a bol ?m ica [11 ] B. Timischl, K. Dettmer, H. Kaspar, M. Thi eme, P.J. Oefner, Electrophoresis 29 (2008) 2203. [12 ] I. Nobeli, H. Ponstingl, E.B. Krissinel, J.M. Thornton, J. Mol. Biol. 334 (2003) 697. [13] G. Theodoridis, H.G. Gika, I.D. Wilson, Trends Anal. Chem. 27 (2008) 251. [ 14] J. Forster, I. Famili, P. Fu, B .O. Palsson, J. Nielsen, Genome Res. 13 (2003) 244. [ 15] O. Fiehn, Plant Mol. Biol. 48 (2002) 155. [16] M. Bedair, L.W. Sumner, Trends Anal. Chem. 27 (2008) 238. [17 ] A. Goffeau, B.G. Barrell, H. Bussey, R.W. Davis, B. Dujon, H.Feldmann, F. Galibert, J.D. Hoheisel, C. Jacq, M. Johnston, E.J. Louis, H.W. Mewes, Y. Murakami, P. Philippsen, H. Tettelin, S.G. Oliver, Science (Washington, DC) 274 (1996) 546. [18 ] M.D. Luque de Castro, F. Priego -Capote, Analytical Applications of Ultrasound, Elsevier, Amsterdam, The Netherlands, 2006. [19] C.L. Winder, W.B. Dunn, S. Schuler, D. Broadhurst, R. Jarvis, G.M. Stephens, R. Goodacre, Anal. Chem. 80 (2008) 2939. [20 ] S. Cubbon, J.W. Bradbury, J. Thomas -Oates, Anal. Chem. 79 (2007) 8911. [21 ] T.R. Sana, K. Waddell, S.M. F ischer, J. Chromatogr., B 871 (2008) 314. [22 ] A.R. Fernie, R.N. Trethewey, A.J. Krotzky, L. Willmitzer, Nat. Rev. Mol. Cell Biol. 5 (2004) 763. [23 ] I. Borodina, P. Krabben, J. Nielsen, Genome Res. 15 (2005) 820. [24 ] E. Holmes, H. Antti, Analyst (Cambrid ge, UK) 127 (2002) 1549. [25 ] A.D. Maher, S.F.M. Zirah, E. Holmes, J.K. Nicholson, Anal. Chem. 79 (2007) 5204. 115 Cha p t er 1 Trends A na l. Che m. , 29 (201 0 ) 111 ? 1 1 9 [26 ] I. Takeda, C. Stretch, P. Barnaby, K. Bhatnager, K. Rankin, H. Fu, A. Weljie, N. Jha, C. Slupsky, NMR Biomed. 22 (2009) 577. [27 ] E. Seli, L . Botros, D. Sakkas, D.H. Burns, Fertil. Steril. 90 (2008) 2183. [28 ] D.S. Wishart, M.J. Lewis, J.A. Morrissey, M.D. Flegel, K. Jeroncic, Y. Xiong, D. Cheng, R. Eisner, B. Gautam, D. Tzur, S. Sawhney, F. Bamforth, R. Greiner, L. Li, J. Chromatogr., B 871 ( 2008) 164. [29 ] N. Salsabili, A.R. Mehrsai, S. Jalaie, Andrologia 41 (2009) 24. [30 ] F. Deepinder, H.T. Chowdary, A. Agarwal, Expert Rev. Mol. Diagn. 7 (2007) 351. [31 ] G. Grac?a, I.F. Duarte, B.J. Goodfellow, I.M. Carreira, A.B. Couceiro, M.R. Domingues, M. Sprau, Anal. Chem. 80 (2008) 6085. [32 ] J.L. Bock, Clin. Chem. 40 (1994) 56. [33 ] J.L. Griffin, R.M. Salek, Expert Rev. Proteomics 4 (2007) 435. [34 ] G.A. McLoughlin, D. Ma, T.M. Tsang, D.N.C. Jones, J. Cilia, M.D. Hill, M.J. Robbins, I.M. Benzel, P.R. Maycox, E. Holmes, S. Bahn, J. Proteome Res. 8 (2009) 1943. [35 ] T. Soga, R. Baran, M. Suematsu, Y. Ueno, S. Ikeda, T. Sakurakawa, Y. Kakazu, T. Ishikawa, M. Robert, T. Nishioka, M. Tomita, J. Biol. Chem. 281 (2006) 16768. [36 ] V. Mapelli, L. Olsson, J. Ni elsen, Trends Biotechnol. 26 (2008) 490. [37 ] L.W. Sumner, P. Mendes, R.A. Dixon, Phytochemistry 62 (2003) 817. [38 ] E. Fukusaki, A. Kobayashi, J. Biosci. Bioeng. 100 (2005) 347. [ 39] R.J. Bino, R.D. Hall, O. Feihn, M.H. Beale, R.N. Trethewey, B.M. Lange, E.S. Wurtele, L.W. Sumner, Trends Plant Sci. 9 (2004) 418. [40 ] S. Alvarez, E.L. Marsh, S.G. Schroeder, D.P. Schachtman, Plant Cell Environ. 31 (2008) 325. 116 N uevas plat a for ma s ana l ?t ica s en met a bol ?m ica [41 ] B. Biais, J.W. Allwood, C. Deborde, Y. Xu, M. Maucourt, B. Beauvoit, W.B. Dunn, D. Jacob, R. Go odacre, D. Rolin, A. Moing, Anal. Chem. 81 (2009) 2884. [42 ] M. Lauridsen, S.H. Hansen, J.W. Jaroszewski, C. Cornett, Anal. Chem. 79 (2007) 1181. [43 ] O. Teahan, S. Gampble, E. Holmes, J. Waxman, J.K. Nicholson, C. Bevan, H.C. Keun, Anal. Chem. 78 (2006) 4 307. [44 ] U. Theobald, W. Mailinger, M. Baltes, M. Rizzi, M. Reuss, Biotechnol. Bioeng. 55 (1997) 305. [45 ] W.B. Dunn, D. Broadhurst, D.I. Ellis, M. Brown, A. Halsall, S. O _ Hagan, I. Spasic, A. Tseng, D.B. Kell, Int. J. Epidemiol. 37 (2008) i23. [46 ] T. Wo od, A. Joddi, J.S. Williams, L. Pandarinathan, A. Courville, M.R. Keplinger, D.R. Janero, P. Vouros, A. Makriyannis, C. Lammi -Keefe, J. Clin. Chem. Lab. Med. 46 (2008) 1289. [47 ] E.J. Saude, B.D. Sykes, Metabolomics 3 (2007) 19. CHAPTER 2: Metabolomics analysis II: Preparation of biological samples prior to detection Metabolomics Analysis II: Preparation of biological samples prior to detection B. ?lvarez -S?nchez, F. Priego -Capote, M.D. Luque de Castro Department of Analy tical Chemistry, University of C?rdoba, Annex C-3 Building,Campus of Rabanales, E-14071 C?rdoba, Spain; and, Institute of Biomedical Research Maim?nides (IMIBIC), Reina Sof?a Hospital, University of C?rdoba, E-14071 C?rdoba, Spain Trends in Analytical Chemistry, Vol. 29, No. 2, 2010, 120? 127 121 Chapter 2 Trends Anal. Chem., 29 (2010) 120 ?128 Metabolomics analysis II: Preparation of biological samples prior to detection B. ?lvarez -S?nchez, F. Priego -Capote, M.D. Luque de Castro Abstract After discussing the significance of preliminary operations, such as selection or storage of biological samples, we critically review the other steps of the analytical process prior to metabolite detection. First comes interruption of metabolism or quenching, which is of crucial importance in the analysis of biological samples. Then, we consider quantitative extraction of metabolites (selective or total for targeted or global metabolomics analysis, respectively) as a function of sample characteristics: solid or liquid samples and cell culture. Finally, we comment on additional steps, such as preconcentration of metabolites, clean up of extracts and fractionation for metabolite separation, and discuss their inclusion in analytical methods. Nuevas plataformas anal?ticas en metabol?mica 122 1. Introduction As discussed in the previous chapter, great efforts are currently devoted to improving the detection and identification steps as well as the bioinformatics and statistical treatment of data generated to obtain biological information. As an example, metabolomics has taken extraordinary benefits from the accurate mass measurements and high resolving power of recent mass analysers such as LTQ-Orbitrap. The combination of these high-performance features with MS/MS capabilities further facilitates the identification and structure elucidation of a large number of metabolites providing extra information about the biochemical connectivity between them [1]. Similarly, the recent developed ultra-high- field NMR (at 900 MHz and beyond) gives new potential to overcome limitations associated to other instruments operating at lower frequencies by increasing spectral dispersion and reducing strong coupling-associated distortions [2]. Despite these advances in analytical instrumentation, the major cause of errors in the generation of analytical results is still linked to sample preparation for metabolites detection. In fact, sample preparation protocols used in metabolomics are almost exclusively based on conventional steps such as maceration or solid?liquid extraction by stirring for solid samples, and liquid?liquid extraction for liquid samples. The lack of homogeneous standard protocols for sample preparation means also a barrier to compare results among laboratories and reproduce metabolomics experiments. The MIAMET (Minimum Information About a METabolomics experiment) program [3] represents a first positive step in the direction of standardization of metabolomics analysis to ensure comparability of results. In this sense, the role played by sample preparation in metabolomics analysis is expected to be reduced by MIAMET reports. 123 Chapter 2 Trends Anal. Chem., 29 (2010) 120 ?128 In the previous chapter a practical guide about particular considerations for selection and storage of biosamples as well as for operations developed prior to metabolomics analysis was described. This chapter deals with the different analytical operations carried out prior to metabolites detection for identification and/or quantification. 2. Fast sampling and metabolism quenching Sample representativeness in metabolomics can only be achieved when metabolism is efficiently interrupted during the sampling process. This enables to provide a snapshot view of the metabolic state and to assess a reliable metabolite profile. Quenching aims at stopping instantly the metabolism by inhibition of endogenous enzymes. In this way, changes in the metabolic profile during sampling processes are suppressed. Quenching strategies should fulfil the following requirements [4]: ? Inactivation of the metabolism should be faster than metabolic changes occurring in the sample. The effectiveness of the quenching process is crucial since the turnover rates of many primary metabolites are usually in the range of 1 mM s-1 (taking into account that concentration of metabolites vary from few molecules per cell, as in the case of certain signalling molecules, to milimolar concentrations for some primary metabolites such as glucose) [5,6]; ? The sample integrity should be carefully preserved during the process, particularly in the case of cells (i.e. preservation of the cell envelope) where leakages of intracellular metabolites should be avoided or minimized; ? Quenching should not induce significant variations in chemical and physical properties or in the concentration of metabolites; Nuevas plataformas anal?ticas en metabol?mica 124 ? The resulting quenched sample should be amenable to the subsequent steps of the analytical process. Quenching of biological fluids, such as blood, usually refers to the storage of the samples at low temperatures (<20 ?C) [7,8]. In most cases, the serum must be rapidly separated from the blood cells prior to the storage, which is usually accomplished by centrifugation or fast filtration under vacuum. Both procedures require low temperatures (4 ?C) to slow down the enzymatic activity and the releasing of intracellular metabolites during the process. Freezing of samples during centrifugation must be avoided as it may cause the disruption of cell envelopes as a consequence of the formation of ice crystals. On the other hand, metabolism deactivation is particularly critical in the analysis of cells and tissues aiming at elucidating the metabolic profile inside (endometabolome) and outside (exometabolome) the cell. The common strategies for quenching of cells and tissues are based on a rapid modification of sample conditions, usually pH or temperature. In the former case, quenching is achieved by instantly changing to extreme pH, either to high alkali (e.g. by addition of KOH or NaOH) or to high acid pH ( e.g. by addition of perchloric, hydrochloric or trichloroacetic acid) [4]. Concerning temperature, quenching is mainly carried out by cooling at values usually lower than ?20 ?C, assuming that sample integrity is not endangered by the cold shock [7]. The most popular method is cold methanol quenching, which allows a rapid interruption of the metabolism in the sub-second time scale [8] and ca n be implemented in protocols destined to discrimination of intra- and extracellular metabolites [8]. This approach has been extensively applied to yeast [8,10,11], but it is also suitable for application to other microorganisms, such as bacteria, unicellular algae or filamentous fungi, and, less frequently, to mammalian cells. The quenching solution is normally constituted by an aqueous?methanol mixture precooled at low temperature (usually ?40 ?C or lower). Methanol seems to be the perfect organic solvent to prepare quenching solutions as it is miscible with water, possesses a low 125 Chapter 2 Trends Anal. Chem., 29 (2010) 120 ?128 freezing-point (much lower than ethanol or glycerol) [11], and methanol ? water solutions are not very viscous. For these reasons, other organic solvents are rarely used in quenching protocols [12]. One of the main limitations of the cold methanol method is the possibility of affecting cellular membranes with metabolite leakages [9]. In fact, some types of cells ( i.e. bacterial cells) are known to be highly sensitive to osmotic changes of the surrounding medium, with subsequent changes in the concentration of intracellular metabolites by cellular membrane disruption [12]. Many protocols include a quenching buffer (e.g. tricine, HEPES or ammonium carbonate [5,8]) in order to control t he ionic strength, thus avoiding damages to the cell envelopes. One other quenching approach based on temperature change is the use of liquid nitrogen (?196 ?C). This approach is specially applied to animal and plant cells, but also to bacteria [13]. In a n optimization study performed by H. Hajjaj et al., cold methanol and liquid nitrogen quenching methods were compared for the fungus Monascus ruber. The cold methanol quenching solution was 10 mM HEPES in 60% methanol/water solution pre - cooled at ?40?C. Fo r the nitrogen quenching method, the sample was transferred directly from the fermentor into liquid nitrogen. The comparison was made by using the concentration of piruvate metabolites, which were extracted after the quenching step with a boiling ethanol solution (75% (v/v) ethanol with 10 mM HEPES at 80 ?C) . The conclusion of this study was that the quenching efficiency of both methods was similar, but the cold methanol quenching method its easier and allows sampling within a time window of milliseconds [1 4]. Nevertheless, cellular envelopes could also be damaged using this alternative due to the formation of ice crystals during freeze/thaw cycles [ 4] . Interruption of the metabolism has scarcely been performed by a heating shock with a fast increase of temperature. One example can be the addition of ethanol at 90 ?C [5]. However, this method seems not to be a Nuevas plataformas anal?ticas en metabol?mica 126 competitive alternative due to potential degradation of thermolabile metabolites and increased cell permeability [9,15]. Apart from in batch protocols, a variety of approaches for automated quenching has recently been reported [16,17]. These are based on on -line coupling of a fast-sampling device to a bioreactor where the metabolism is stopped. One example is found in Fig. 1, in which a heat exchanger was used for sampling and quenching of mammalian cells [17], which is a field characterized by a lack of validated methods. With this automated approach, the sampling device is pre-cooled at ??? ?C by filling the container with the quenching mixture (methanol/water 60:40, v/v) and the pressure is reduced to 200 mbar. This device is on-line coupled to the bioreactor, which enables the development of sampling and quenching steps simultaneously without metabolite losses. Figure 1. Automated approach for quenching of mammalian cells by on-line coupling of a a heat exchanger used as fast-sampling device to a bioreactor where the metabolism is stopped (Reproduced with permission of Springer- Verlag, Reference [17]). 127 Chapter 2 Trends Anal. Chem., 29 (2010) 120 ?128 It is worth mentioning that there are significant physico-chemical differences between cell envelopes of eukaryotic and prokaryotic organisms. This fact explains that metabolite leakages caused by quenching with organic solvents are less severe for eukaryotic cells than for prokaryotic ones [5,18,19] . For this reason, quenching protocols cannot be directly transferred among different organisms and require proper optimization and validation as a function of the target organism [4,19]. One example is found in the study developed by Faijes et al. for metabolome analysis of Lactobacillum plantarum [5], in which f our different aqueous quenching solutions, all containing 60% methanol (1, 60% MeOH; 2, 60% MeOH 70 mM HEPES; 3, 60% MeOH and 0.85% NaCl; 4, 60% MEOH and 0.85% w/v ammonium carbonate pH 5.5), were tested to compare their efficiency. Only solutions containing either 70 mM HEPES or 0.85% (w/v) ammonium carbonate provoked less than 10% cell leakage and the energy charge of the quenched cells (estimated by ATP measurement in the supernatant after quenching) was severe, indicating rapid inactivation of the metabolism. Five different extraction protocols (based on, (i) cold methanol, (ii) perchloric acid, (iii) boiling ethanol, (iv) 1:1 chloroform?methanol, and (v) 1:1chloroform?water) were compared for isolation of metabolites from L. Plantarum. Targeting quantification of representative intracellular metabolites showed that the best extraction efficiencies were achieved with cold methanol, boiling ethanol or perchloric acid. In most cases, quenching is followed by division of the sample in two fractions, usually by fast filtration or cold centrifugation. This step allows minimizing the dilution effect of cells in the supernatant and quantifying separately intra- and extracellular metabolites from cells and supernatant, respectively [20]. Fast filtration has been widely used, either as a separation step after quenching or as a fast sampling mode. The last option is not frequently used due to the relatively long time required for sample filtration; therefore, this sampling method is limited to the analysis of metabolites with turnover rates from minutes to hours and high intracellular levels (i.e. amino acids or tricarboxilic acid metabolites). Interestingly, the risk of leakages or Nuevas plataformas anal?ticas en metabol?mica 128 cellular damage is considerably lower than in the methanol-quenching method [8]. It is worth emphasizing that the washing solution has to be properly selected, as this step is probably the bottleneck of fast filtration regarding to leakage effects. Finally, an extraction step is usually performed aiming at permeabilizing the membrane and extracting intracellular metabolites with the highest efficiency and minimum degradation [12]. Extraction can be performed with organic solvents at high or low temperatures, either pure (e.g. ethanol, methanol) or in mixture (e.g. methanol?cloroform) or by acid or basic solutions. The extraction efficiency of the former is limited by the solubility of the metabolites in the organic solvent, the posibility of side-reactions or by degradation of the metabolites [5,15]. However, these methods can be of interest for the extraction of target compounds (e.g. organic acids with KOH). With these premises, alcohols (particularly methanol) are the most suitable extractants for this step as they precipitate proteins, effectively permeate cells, are easy to evaporate? concentrate and do not add undesirable salts in mass spectrometry analyses. 3. Metabolites extraction Metabolite extraction is a key step in the metabolomics analytical process and its effectiveness directly affects the quality of the final data. Ideally, metabolite extraction aims to: (i) efficiently release metabolites from sample (ii) remove interferents that may difficult the analysis, such as salts and proteins (iii) make the extract compatible with the analytical technique and, (iv) when necessary, concentrate trace metabolites before analysis. 129 Chapter 2 Trends Anal. Chem., 29 (2010) 120 ?128 As previously mentioned, the required selectivity of the extraction step will depend on the aim of the study. Therefore, while targeting metabolomics calls for a highly selective extraction that provides clean and concentrated extracts, extraction is eminently non-selective in metabolic profiling, so, only salts and macromolecules (e.g. proteins) are considered potential interferents to be removed in this step. The extraction protocol mainly depends on the target biological sample, which, in the case of solids, is performed by solid?liquid extraction (e.g,. maceration, Soxhlet extraction, Folch extraction, supercritical fluid extraction, ultrasound-assisted extraction or microwave-assisted extraction). On the other hand, metabolites from liquid samples are mainly extracted by liquid?liquid extraction (LLE), solid-phase extraction (SPE) or solid phase microextraction (SPME). These approaches are described below. 3.1. Liquid samples Most biological fluids are aqueous, so they can be directly analysed by MS or LC?MS (also CE?MS) with minimum sample preparation [21]. This is the case of urine, microdyalisates and digestive fluids, for which sample preparation usually entails buffering, dilution or evaporation and centrifugation [22,23]. Direct analysis minimizes metabolite losses, but the high-salt content can cause ionization suppression, lead to adduct formation, and also negatively affect the instrument performance by the presence of non-volatile residues. These aspects can be minimized by including, when possible, an effective extraction step, such as LLE [24] or SPE [25] to clean and desalt the target sample. Concerning NMR, direct analysis is the most suited alternative with minimum sample preparation such as buffering with deuterated solutions [26]. On the other hand, the analysis of serum or plasma samples is limited by the presence of a large number of proteins that interfere in mass Nuevas plataformas anal?ticas en metabol?mica 130 spectra. This makes the analysis of low-molecular weight metabolites difficult, as long as most of them are present at very low concentrations. Although possible, direct analysis of plasma by NMR or MS is rarely performed, and an extraction step is usually included in the analytical process for high efficient removal of proteins. Deproteinization can be achieved by lowering the pH or using organic solvents such as acetonitrile or methanol [27,28]. The number of detected metabolites is higher with the latter, which is usually the preferred choice. Interestingly, classical extraction techniques such as LLE or SPE are of great relevance in metabolomics analysis of liquid samples. Liquid?liquid extraction enables extraction of metabolites in two fractions that separately contain polar and non-polar compounds, which can be independently analysed by NMR, LC?MS or GC?MS [6]. Solid -phase extraction in its different formats is widely used for selective extraction of target compounds from biological samples with excellent results in terms of precision and sensitivity [29,30]. Solid -phase microextraction is a widely extended approach for analysing volatile metabolites. For instance, SPME has been used for the extraction of volatile compounds in human sweat aiming at establishing the metabolic basis of the human odour [31], and for the extraction of volatile metabolites from wine for characterizing flavour and aroma [32]. Due to their selectivity, SPE and SPME are widely implemented in targeting metabolomics methods [16,17]. 3.2. Solid samples As mentioned before, the first step of sample preparation is to quench the metabolism by using one of the alternatives exposed in section ?Fast sampling and quenching of metabolism?? Then, the next step is to mill the sample to obtain a homogeneous powder. This step is critical, especially in plant tissues in which cells are surrounded by a thick wall that is 131 Chapter 2 Trends Anal. Chem., 29 (2010) 120 ?128 necessarily broken before metabolite extraction. The most common procedures for plant-cell wall-breakage and homogenization are performed by a mortar and pestle [33], ball mill [34], vibration mill [35], Ultra Turrax [36], ultrasonic probe and thermomixer. When possible, this step is performed under liquid nitrogen to prevent tissue defrosting. Finally, the powder is lyophilized and stored at low temperatures until analysis. Although there are some methods in which the powder is directly analysed by MS or NMR [37], an extraction step prior to analysis is usually necessary. Accordingly, solid?liquid extraction (better expressed as leaching or lixiviation) is performed in order to permeate the membrane and extract the metabolites with the highest efficiency and minimum degradation. With this aim, a suitable extractant is added to the solid material, and the contact between them is favoured by shaking, vortexing or magnetic stirring for a preset time. Conventional protocols of leaching (such as Soxhlet or Folch extraction) are well established for the extraction of metabolites from solid samples, and their usefulness has been extensively proved [38,39 ]. The extraction time may vary from minutes to hours, depending on the characteristics of the sample, so the extraction process can be accelerated by means of microwaves, ultrasound or by supercritical fluids [40,41]. Particularly, Focused Microwave-Assisted Soxhlet Extraction (FMASE) has proved to be highly efficient for the extraction of weakly polar and non-polar metabolites in biological solid samples [42]. Thus, while conventional extraction requires several hours for quantitative extraction of the target metabolites, FMASE is efficiently performed within minutes. Due to the complexity of solid samples, their preparation protocols must be necessarily optimized in order to avoid degradation or modification of metabolites during this step, caused either by extraction conditions or by enzymatic activity. The best extraction conditions depend on the aim of the analysis. In metabolic profiling, due to the large variety of metabolites and their differences in physical and chemical properties, there is no one ideal procedure that allows simultaneous extraction of all metabolites with high Nuevas plataformas anal?ticas en metabol?mica 132 efficiency. There are some reported studies on the metabolome of solid samples (especially in plant metabolomics) in which various extraction procedures are compared in order to maximize metabolite coverage [43]. In most of them, extraction is performed with organic solvents at high or low temperatures, either pure (e.g. ethanol, methanol, acetonitrile, chloroform, hexane), in mixture (e.g. methanol?cloroform) or by acid?basic solutions. The efficiency of the extraction with organic solvents is limited by solubility of the metabolites in the extractant, possibility of side-reactions or degradation of the metabolites [32]. It is worth mentioning that deuterated solvents are required for metabolites extraction when NMR is the determination technique [44]. On the other hand, extraction with perchloric acid is widely used prior NMR analysis either after a quenching step or during cell inactivation [45]. Although this extractant is highly efficient to precipitate proteins and to extract some hydrophilic metabolites, acidic treatment is limited to the extraction of acid-stable metabolites. With these premises, alcohols (particularly methanol) and acetonitrile have demonstrated to be the best extractants for metabolic profiling as they precipitate proteins, effectively permeate the cell, are easy to evaporate? concentrate and they do not add salts, undesirable in mass spectrometry analyses. In addition, they are likely to extract a wide range of metabolites when used in mixtures. For instance, methanol?water?chloroform mixtures have been successfully used for simultaneous extraction in a single step of polar, weakly apolar and non-polar metabolites [46]. Hydrophilic metabolites such as sugars, amino acids and organic acids are extracted in the methanol?water phase, while lipophilic compounds, such as lipids, chlorophylls and waxes are extracted in the chloroform phase. After extraction, the two phases, separated by simple centrifugation, are treated and analysed separately by LC?MS or NMR and GC?MS, respectively [47,48]. As previously mentioned, solid biological materials are likely to give an insight of the spatial distribution of metabolism in complex organisms. Accordingly, the use of solid samples is not limited to the field of plant 133 Chapter 2 Trends Anal. Chem., 29 (2010) 120 ?128 metabolomics, being also noteworthy the existing protocols for the treatment of organ tissues (such as brain, liver and tumoural tissues), faeces or whole organisms such as earthworms. The sample preparation method developed by Lin et al. [43] for the study of the metabolome of different organ tissues (namely salmon muscle and liver) reflects the main steps that are usually performed in the treatment of organ tissues. The process starts with a quenching step, which is carried out by freezing the samples in liquid nitrogen. Then, tissues are lyophilized, effectively disrupted and milled to homogeneity, keeping the temperature constant throughout the process. Metabolites are then extracted and the extract is dried and reconstituted into an appropriate solution for NMR analysis, which consists of buffered D2O and a chemical shift standard such as 3-(trimethylsilyl)proprionate-2-2- 3-3-d4, TMSP. The method was optimized by comparing different tissue disruption procedures and different extractant mixtures. As in the case of plant tissues, a mixture of water?methanol?chloroform was the best extractant with regards to the number of covered metabolites. 3.3. Cells Determination of metabolite levels in intact cells has been performed by NMR or MALDI-TOF-MS [49]. Although NMR provides in vivo data for well-established conditions [50], the low sensitivity achieved by this technique restricts its use to abundant metabolites and/or to high cell- density cultures [49]. Thu s, like in solid samples, determination of intracellular metabolites is performed after an extraction step. Extraction from cells can be performed by two approaches [3], the main difference between them being that quenching and extraction are either simultaneous or sequential. In the former, the quenching solution also serves as extractant, while in the latter, cells are separated from the surrounding medium and then metabolites are isolated from cells with an Nuevas plataformas anal?ticas en metabol?mica 134 appropriate extractant. Figure 2 shows the scheme for the steps involved in each approach [51]. In the sequential mode, quenching is followed by division of the sample in two fractions, usually by fast filtration or cold centrifugation. This step allows diminishing the dilution effect of cells in the supernatant and separately quantifying intra- and extracellular metabolites from the cell and the supernatant, respectively. Figure 2. Scheme of the steps followed in the simultaneous or sequential approaches used for extraction of metabolites from cells (Reproduced with permission of the American Chemical Society, Reference [51]). Cold centrifugation [52] is usually the preferred option due to the relatively long time required for fast filtration. After cells isolation, an extraction step is usually performed in a way similar to that applied to solid samples, using a cold or hot extractant (e.g. methanol, ethanol, methanol? 135 Chapter 2 Trends Anal. Chem., 29 (2010) 120 ?128 water or methanol?water?chloroform), acid or basic solutions (e.g. perchloric acid, acetic acid, KOH solutions). In the simultaneous mode, intra- and extracellular metabolites are determined together, without leakage problems. Nevertheless, the exo- and endometabolome of cells cannot be distinguished, and detection and quantification of many metabolites is difficult, as they are extremely diluted in the large volume constituted by the extracellular medium and quenching solution, which also contains a high salt concentration [53]. Shaub et al. developed a simple method for the simultaneous quenching and extraction of metabolites from polysaccharide-producing bacteria using cold ethanol [51]. After extraction, an evaporation step was performed in order to concentrate the extracts prior to analysis. 4. Other additional steps of the metabolomics analytical process After extraction, metabolites are usually diluted in large volumes of solvent that are necessarily removed, either partially or totally, aiming at concentrating the extract prior to determination. Accordingly, a solvent evaporation step is usually performed, either by rotary evaporation, centrifugal concentration or lyophilization. It should be emphasized that the whole sample preparation process is designed around the analytical platform, which, in some cases, demands for additional preliminary steps. As an example, 1H NMR analysis comprises dilution of the extracts in deuterated solvents for providing a frequency lock for the spectrometer [54]. On the other hand, the low volatility of many metabolites makes necessary a derivatization step prior to GC?MS analyses, which is usually performed by silylation [55]. By this method, functional Nuevas plataformas anal?ticas en metabol?mica 136 groups containing active hydrogen atoms, such as ?OH, NH, ?COOH and ?SH are trimethylsilylated by a silylation reagent, usually N-methyl-N-trimethyl- silyltrifluoroacetamide (MSTFA) or N,O -bis trimethylsilyltrifluoroacetamide (BSTFA). Therefore, very common metabolites such as organic acids, amino acids, sugars, fatty acids, and steroids, among others, can be determined by GC?MS. Even though an extensive discussion of the chromatographic techniques used in metabolomic experiments is out of the scope of this review, it is noteworthy the implementation of new chromatographic approaches with promising perspectives in metabolomics. Therefore, despite traditional GC and LC separations has been extensively used in metabolomics so far [56], the use of new chromatographic phases, such as monolithic capillary columns [57] or ultra performance liquid chromatography (UPLC) columns, as well as new chromatographic formats, such as comprehensive multidimensional chromatography or preparative chromatography, are gaining importance in the recent investigations. This can be justified by the need of an extremely high resolution power, demanded by the large number of compounds present in biological samples. In UPLC the use of chromatographic columns with smaller particle sizes and higher pressures leads to an increased resolution and sensitivity as compared to conventional HPLC. Grata et al. developed a UPLC?MS method for the metabolite fingerprinting of the plant Arabidopsis Thaliana obtained after subjected to stress conditions [58]. The hyphenation of the UPLC with a TOF mass spectrometer allowed the tentative identification of the stress- induced ions, which can be potentially used as biomarkers. In other study, the robustness and repeatability of the UPLC?MS methodology was evaluated [59] to ensure the comparability of the data obtained from the analysis of more than 7000 serum samples within several years, in a long- term study which is planned to be performed by the Human Serum Metabolome Project (HUSERMET). 137 Chapter 2 Trends Anal. Chem., 29 (2010) 120 ?128 One other alternative is to combine two or more separation techniques in a single analysis, usually GCxGC [22] LCxLC and, less frequently, LCxCE, which is known as comprehensive multidimensional chromatography [60]. Well implemented in proteomics, multidimensional chromatography is gaining attention in the field of metabolomics in the last years as long as it offers tremendous resolution capability and, in some cases, increased sensitivity as compared to one-dimensional chromatography. In fact, the maximum resolution achieved in an orthogonal approach (which is performed when the separation mechanism of the two dimensions is essentially different) is equal to the product of the individual peak capacities of each independent dimension. In GCxGC, the sample components are separated throughout two GC columns, which are in series connected by means of a temperature-controlled interface. Thus, the sample is firstly loaded on a conventional high resolution capillary column, which is usually non-polar. Then, small fractions of eluate are sequentially transferred via a cryogenic trap, called modulator, to a second column, which is usually polar and notably shorter than the former. Lu et al. have developed a method for the analysis of cigarette smoke condensates by GCxGC?TOF/MS [61], which allowed identification of over 1000 compounds present in the exhaled smoke. Comprehensive GCxGC has also been used in the separation of complex samples such as herbal oils [62]. On the o ther hand, multidimensional liquid chromatography offers potential versatility thanks to the large variability of stationary phases available for LC separations [60]. However, applications of multidimensional liquid chromatography are by far less exploited than those of GC, which is probably attributed to the lack of MS libraries, and the less resolution capability associated to this technique. 5. Conclusions Nuevas plataformas anal?ticas en metabol?mica 138 The different steps involved in preparation of biosamples in metabolomics experiments have been reviewed here. These steps have encompassed from sample selection, a critical task to extract the maximum level of biological information, to isolation of metabolites, derivatization or concentration prior to detection. Practical considerations have been provided for each step with significant examples for different types of biosamples: solids, liquids or cells. With the recent evolution experienced by detectors such as MS or NMR, it is expected that special attention will be paid to sample preparation in order to reduce its decisive impact on the quality of the final results. 6. Acknowledgments The Spanish Ministerio de Ciencia e Innovaci?n (MICINN) is acknowledged for financial support through project CTQ2009-07430. F.P.C. is grateful to MICINN for a Ram?n y Cajal contract (RYC -2009-03921). B.A.S. is also grateful to Ministerio de Ciencia y Tecnolog?a for an FPI scholarship (BES -2007-15043). 7. References [1] R. Breitling, A.R. Pitt, M.P. Barrett, Trends Biotechnol. 24 (2006) 543. [2] R.J. Bino, R.D. Hall, O. Fiehn, J. Kopka, K. Saito, J. Draper, B.J. Nikolau, P. Mendes, U. Roessner-Tunali, M.H. Beale, R.N. Trethewey, B.M. Lange, E.S. Wurtele, L.W. Sumner, Trends Plant Sci. 9 (2004) 418. [3] M.R. Mashego, K. Rumbold, M. De Mey, E. Vandamme, W. Soe taert, J.J. Heijnen, Biotechnol. Lett. 29 (2007) 1. [4] M. Faijes, A.E. Mars, E.J. Smid, Microb. Cell Fact. 6 (2007) 27. 139 Chapter 2 Trends Anal. Chem., 29 (2010) 120 ?128 [5] S.G. Villas -Bo? s, J. Nielsen, J. Smedsgaard, M.A.E. Hansen, U. Roessner - Tunali, Metabolome Analysis. An Introduction, Wiley Interscience, New York, USA, 2007 p. 15 ?38. [6] C. Wittmann, J.O. Kromer , P. Kiefer, T. Binz, E. Heinzle, Anal. Biochem. 327 (2004) 135. [7] C. Bolten, C. Wittmann, Biotechnol. Lett. 30 (2008) 1993. [8] C.L. Winder, W.B. Dunn, S. Schuler, D. Broadhurst, R. Jarvis , G.M. Stephens, R. Goodacre, Anal. Chem. 80 (2008) 2939. [9] M.O. Loret, L. Pedersen, J. Francois, Yeast 24 (2007) 47. [10] A.B. Canelas, C. Ras, A. ten Pierick, J.C. van Dam, J.J. Heijnen, W.M. van Gulik, Metabolomics 4 (2008) 226. [11] S.G. Villas -Bo? s, P. Bruheim, Anal. Biochem. 370 (2007) 87. [12] H. Wu, A.D. Southam, A. Hines, M.R. Viant, Anal. Biochem. 372 (2008) 204. [13] H. Hajjaj , P.J. Blanc, G. Goma, J. Franc ois, Microbiol. Lett. 168 (1998) 195. [14] B. Gonz ?lez, J. Francois, M. Renaud, Yeast 13 (1997) 1347. [15] C. Wiendahl, J.J. Brandner, C. K? ppers, B. Luo, U. Schygulla, T. Noll, M. Oldiges, Chem. Eng. Technol. 30 (2007) 322. [16] J. Hiller, E. Franco -Lara, V. Papaioannou, D. Weuster-Botz, Biotechnol. Lett. 29 (2007) 1161. [17] U. Nasution, W.M . Van Gulik, R.J. Kleijn, W.A. Van Winden, A. Proell, J.J. Heiinen, Biotechnol. Bioeng. 94 (2006) 159. [18] S.G. Villas -Bo? s, J. Hojer -Pedersen, M. Akesson, J. Smedsgaard, J. Nielsen, Yeast 22 (2005) 1155. [19] C.J. Bolten, P. Kiefer, F. Letisse, J. -C. Portais, C. Wittmann, Anal. Chem. 79 (2007) 3843. Nuevas plataformas anal?ticas en metabol?mica 140 [20] H.L. Kirschenlohr, J.L. Griffin, S.C. Clarke, R. Rhydwen, A.A. Grace1, P.M. Schofield, K.M. Brindle1, J.C. Metcalfe, Nat. Med. 12 (2006) 705. [21] M.P. Hodson, G.J. Dear, J.L. Griffin, J.N. Haselden, Meta bolomics 5 (2009) 166. [22] M.C.Y. Wong, W.T.K. Lee, J.S.Y. Wong, G. Frost, J. Lodge, J. Chromatogr., B 871 (2008) 341. [23] S. Bajad, V. Shulaev, Trends Anal. Chem. 26 (2007) 625. [24] A.M. Flores -Valverde, E.M. Hill, Anal. Chem. 80 (2008) 8771. [25] M.R. Viant, C. Ludwig, U.L. Gu ? nther, in: William Griffiths (Editor), Metabolomics, Metabonomics and Metabolite Profiling, RCS Publishing, Cambridge, UK, 2008, p. 44. [26] E.J. Want, G. O ? Maille, C.A. Smith, T.R. Brandon, W. Uritboonthai, C. Qin, S.A. Trauger, G. Siuzdak, Anal. Chem. 78 (2006) 743. [27] S. Souverain, S. Rudaz, J.L. Veuthey, J. Pharm. Biomed. Anal. 35 (2004) 913. [28] S. Rezzi, F.A. Vera, F. -P.J. Martin, S. Wang, D. Lawler, S. Kochhar, J. Chromatogr., B 871 (2008) 271. [29] J.M. Mata -Granados, J.M. Quesada G? mez, M.D. Luque de Castro, Clin. Chim. Acta 403 (2009) 126. [30] A.M. Curran, S.I. Rabin, P.A. Prada, K.G. Furton, J. Chem. Ecol. 31 (2005) 1607. [31] M. Bonino, R. Schellino, C. Rizzi, R. Aigotti, C. Delfini, C. Baiocchi, Food Chem. 80 (2003) 125. [32] A. Erban, N. Shauer, A.L. Fernie, J. Kopka, in: W. Weckwerth (Editor), Metabolomics Methods and Protocols, Humana Press Inc., Totowa, NJ, USA, 2007, p. 19. [33] W. Weckwerth, K. Wenzel, O. Fiehn, Proteomics 4 (2004) 78. 141 Chapter 2 Trends Anal. Chem., 29 (2010) 120 ?128 [34] P. Jonsson, J. Gul lberg, A. Nordstrom, M. Kusano, M. Kowalczyk, M. Sjostrom, T. Moritz, Anal Chem. 76 (2004) 1738. [35] U. Roessner, C. Wagner, J. Kopka, R.N. Trethewey, L. Willmitzer, Plant J. 23 (2000) 131. [36] D. Leibfritz, W. Dreher, W. Willker, in: J.C. Lindon, J.K. N icholson, E. Holmes (Editors), The Handbook of Metabonomics and Metabolomics, Elsevier, Oxford, UK, 2007, pp. 487 ?586. [37] S.J. Iverson, S.L.C. Lang, M.H. Cooper, Lipids 36 (2001) 1283. [38] K. Sch? fer, Anal. Chim. Acta 358 (1998) 69. [39] C.W. Huie, Anal . Bioanal. Chem. 373 (2002) 23. [40] W. Weckwerth, Annu. Rev. Plant Biol. 54 (2003) 669. [41] J.L. Luque -Garc?a, M.D. Luque de Castro, Talanta 64 (2004) 571. [42] C.Y. Lin, H. Wu, R.S. Tjeerdema, M.R. Viant, Metabolomics 3 (2007) 55. [43] R. Verpoorte, Y.H . Choi, N.R. Mustafa, H.K. Kim, Phytochem. Rev. 7 (2008) 525. [44] N.J. Kruger, M.A. Troncoso -Ponce, R.G. Ratcliffe, Nat. Protoc. 3 (2008) 1001. [45] H. Wu, A.D. Southam, A. Hines, M.R. Vian, Anal. Biochem. 372 (2008) 204. [46] K. Dettmer, P.A. Aronov, B.D . Hammock, Mass Spectrom. Rev. 26 (2007) 51. [47] J.L. Edwards, R.T. Kennedy, Anal. Chem. 77 (2005) 2201. [48] A.A. de Graaf, R.M. Wittig, U. Probst, J. Strohhaecker, S.M. Schobert, H. Sahm, J. Magn. Res. 98 (1992) 654. [49] A. Hartbrich, G. Schmitz, D. We uster-Botz, A.A. de Graaf, C. Wandrey, Biotechnol. Bioeng. 51 (1997) 624. Nuevas plataformas anal?ticas en metabol?mica 142 [50] J. Schaub, C. Schiesling, M. Reuss, M. Dauner, Biotechnol. Prog. 22 (2006) 1434. [51] J. Hiller, E. Franco -Lara, D. Weuster-Botz, Biotechnol. Lett. 29 (2007) 1169. [52] R.P. Mah arjan, T. Ferenci, Anal. Biochem. 313 (2003) 145. [53] K. Odunsi, R.M. Wollman, C.B. Ambrosone, A. Hutson, S.E. McCann, J. Tammela, J.P. Geisler, G. Miller, T. Sellers, W. Cliby, F. Qian, B. Keitz, M. Intengan, L. Shashikant, J.L. Alderfer, Int. J. Cancer 113 (2005) 782. [54] J.M. Bu? cher, D. Czernik, J.C. Ewald, U. Sauer, N. Zamboni, Anal. Chem. 81 (2009) 2135. [55] D.I. Ellis, W.B. Dunn, Trends Anal. Chem. 24 (2005) 285. [56] V.V. Tolstikov, A. Lommen, K. Nakanishi, N. Tanaka, O. Fiehn, Anal. Chem. 75 (2003) 6737. [57] L.M. Blumberg, J. Chromatogr., A. 985 (2003) 29. [58] X. Lu, M. Zhao, H. Kong, J. Cai, J. Wu, M. Wu, R. Hua, J. Liu, G. Xu, J. Sep. Sci. 27 (2004) 101. [59] X. Di, R.A. Shellie, P.J.Marriott,C.W.Huie, J. Sep. Sci.27(2004) 451. P A RT 2 : Targeting analysis En esta Parte II de la Memoria se recoge la investigaci?n realizada utilizando una de las estrategias caracter?sticas de la metabol?mica que permite profundizar en el conocimiento de un metabolito o, m?s com?nmente, un grupo concreto o una familia de metabolitos que comparten una ruta o son caracter?sticos de un determinado comportamiento del sistema en estudio. Esta estrategia se orienta al an?lisis cualitativo y/o cuantitativo y recibe en ingl?s el nombre de ?targeted analysis? o "targeting analysis" , pero es dif?cil de expresarla con s?lo dos palabras en espa?ol, por lo que, en lo sucesivo y por brevedad, se adoptar? la terminolog?a inglesa. La investigaci?n realizada utilizando esta estrategia ha permitido el desarrollo de un total de 6 m?todos anal?ticos, con sus correspondientes aplicaciones, que ha dado lugar a otras tantas publicaciones. Estas aplicaciones se han ordenado atendiendo al tipo de matriz en el que se ha llevado a cabo el ?targeting analysis?? As?, en una matri? compleja, como es la orina, se han determinado hormonas esteroideas femeninas utilizando cromatograf?a de l?quidos y espectrometr?a de masas en t?ndem, con analizador de triple cuadrupolo para la separaci?n y la detecci?n, haciendo hincapi? en la etapa de preparaci?n de muestra. En concreto, se ha pretendido innovar en las etapas de limpieza de la muestra y preconcentraci?n de los analitos de forma autom?tica y en l?nea con la etapa de separaci?n?detecci?n (Cap?tulo 3), o en la hidr?lisis enzim?tica, aceler?ndola mediante la utilizaci?n de ultrasonidos como energ?a auxiliar (Cap?tulo 4). En dos matrices biol?gicas (suero y orina) se han determinado precursores de esfingol?pidos mediante separaci?n por microcromatograf?a y detecci?n por fluorescencia inducida por l?ser, tras las etapas de extracci?n en fase s?lida y derivatizaci?n in situ realizadas en un sistema LO? (?lab-on-valve?) (Cap?tulo ?)? Tres han sido las matrices (suero, orina y leche humana) en las que se han analizado vitamina B9 y sus catabolitos mediante un sistema de extracci?n en fase s?lida autom?tico, conectado en l?nea con un cromat?grafo de l?quidos y un detector de masas en t?ndem (Cap?tulo 6). Tambi?n se ha reali?ado ?targeting analysis? en matrices vegetales, como en el m?todo desarrollado para la determinaci?n de ?cidos haloac?ticos en acelgas y espinacas, basado en lixiviaci?n asistida por ultrasonidos con derivatizaci?n in situ previa a la separaci?n mediante cromatograf?a de gases y detecci?n por captura electr?nica (Cap?tulo 7); as? como en el estudio dedicado al tomate, en el que se han determinado carotenos, fenoles, ?cido asc?rbico y carbohidratos. Esta amplia gama de compuestos requiri?, tras su lixiviaci?n asistida por ultrasonidos, la separaci?n mediante cromatograf?a de gases o de l?quidos, con detecci?n en ambos casos mediante espectrometr?a de masas (Cap?tulo 8). Part II of this book contains the research developed using one of the strategies characteristic of metabolomics that allows to go in depth into the knowledge of a metabolite or, more frequently, a given group of metabolites, or a family of them which shares a metabolic pathway, or are characteristic of the behavior of the system under study; that is: targeting analysis. The developed research based on this strategy has resulted in 6 analytical methods and their corresponding applications, which have given place to equal number of publications. The applications have been ordered as a function of the type of matrix to which targeting analysis has been applied. Thus, in a complex matrix as urine, female steroid hormones have been determined using liquid chromatography and in-tandem mass spectrometry (triple quad analyzer) for separation and detection, and making special emphasis on sample preparation. In fact, the main aim in Chapter 3 was innovation both in sample cleanup and analytes preconcentration in an automatic manner and in-line with the separation? detection step; while in Chapter 4 acceleration of enzymatic hydrolysis by using ultrasound as auxiliary energy was the main achievement. Sphingoid precursors have been determined in two biological matrices such as serum and urine by microchromatography separation and detection by laser- induced fluorescence, both after solid-phase extraction and in situ derivatization carried out in an LOV (lab-on-valve) system, as shows Chapter 5. Vitamin B9 and their catabolites have been determined in three matrices (serum, urine and breast milk) by using an automatic solid-phase extraction system on-line connected to a liquid chromatograph and tandem mass detector (see Chapter 6). Also targeting analysis has been the strategy applied to vegetal matrices, as in the method for determination of haloacetic acids in chard and spinach, based on ultrasound-assisted leaching with in situ derivatization prior to gas-chromatography separation and electron- capture detection, which is the subject of Chapter 7. In addition, a vegetable matrix (tomato) has been the sample for determination of carotenes, phenols, ascorbic acid and carbohydrates; a game of compounds that made necessary two types of chromatography (liquid and gas chromatography) with mass spectrometry detection in both cases, and after ultrasound- assisted extraction, as Chapter 8 shows. CHAPTER 3: Automated solid-phase extraction for concentration and clean-up of female steroid hormones prior to liquid chromatography?electrospray ionization?tandem mass spectrometry: An approach to lipidomics Automated solid- phase extraction for concentration and clean- up of female steroid hormones prior to liquid chromatography?electrospray ionization ?tandem mass spectrometry: An approach to lipidomics B. ?lvarez-S?nchez, F. Priego-Capote, M.D. Luque de Castro Department of Analytical Chemistry, University of C?rdoba, Annex C-3 Building, Campus of Rabanales, E-14071 C?rdoba, Spain Journal of Chromatography A, 1207 (2008) 46?54 153 J. Chromatogr. A, 1207 (2008) 46?54 Chapter 3 Automated solid- phase extraction for concentration and clean- up of female steroid hormones prior to liquid chromatography?electrospray ionization ?tandem mass spectrometry: An approach to lipidomics B. ?lvarez-S?nchez, F. Priego-Capote, M.D. Luque de Castro Abstract A method for determination of free and glucuronide-conjugated female steroid hormones in urine at the pg mL?1 level is here presented. For this purpose, a dual approach with or without ?-glucuronidase hydrolysis has been developed to succeed in this analysis. The target analytes were two progestogens ?progesterone and pregnenolone? and three endogenous estrogens ?estradiol, estriol and estrone?. Separation and detectionwere carried out by liquid chromatography electrospray ionization and tandem mass spectrometry (LC?ESI?MS/MS) with a triple quadrupole (QqQ) mass detector. The determination stepwas optimized by multiple reaction monitoring for highly selective identification and sensitive quantification of female hormones in a complex sample such as human urine. As these compounds are present in urine at very low concentration (ngmL?1 level), a preconcentration and clean-up step by solid-phase extraction was automatically carried out prior to the chromatographic step in order to improve the sensitivity of the method. This sample pretreatment was performed using a lab-on-valve (LOV) manifold which provided preconcentration factors ranging from 59.1 to 72.3 for 10 mL urine. The detection and quantification limits were in the ranges 1.8?18 pg and 6?61 pg on-column, respectively, with precision values from 1.93 to 10.99%, expressed as relative standard deviation. These results enable to conclude the suitability of the LOV?LC?QqQ approach for determination of the lipidomic profiling of the main female steroid hormones in a difficult matrix as human urine. The method can be potentially applied to the clinical and other metabolomic areas. 154 Nuevas plataformas anal?ticas en metabol?mica 1. Introduction Systems biology, which involves the integration of genomic, transcriptomic and proteomic techniques, provides a powerful approach to understand mechanisms of diseases and develop new therapeutic agents [1,2]. Even in simple systems, statistical relationships between gene expression and protein levels can be considerably weak [3]. This fact has stimulated the use of metabonomics to provide quantitative measurement of the multivariate metabolic responses of multicellular systems to pathophysiological stimuli or genetic manipulations [4]. In metabonomics, lipidomics has evolved as a separate discipline because of the extraordinary structural diversity of lipids and their key roles in the pathophysiology of diseases. Lipidomics has been defined as the full characterization of lipid molecular species and their biological roles concerning expression of proteins involved in lipid metabolism and function, including regulation [5]. Although still an emerging field, lipidomics has already provided promising newresearch possibilities; nevertheless, technological challenges remain unsolved. Crucial differences between different biological states can be distinguished mainly by quantitation of overall or targeted lipids profiling. Estrogens and progestogens are a group of female sterol lipid hormones [6] derived from cholesterol, which are widely distributed in animals and humans. The highest levels of these hormones are found in tissues with reproductive function, such as breast, ovary, vagina and uterus [7]. Estrogens exert diverse biological functions in female organisms, i.e. female sexual differentiation, arterial vasodilatation and maintenance of bone density [8]. In addition, endogenous estrogens have a protective function against various diseases, such as osteoporosis, atherosclerosis and cardiovascular and neurodegenerative diseases [9,10]. Progesterone plays a key role in the female menstrual cycle (mainly produced after ovulation) and 155 J. Chromatogr. A, 1207 (2008) 46?54 Chapter 3 during pregnancy, when its production causes suppression of further ovulation and provides the correct environment for the developing embryo [11]. Pregnenolone, together with other steroids, is found at high concentrations in certain areas of the brain, where it acts as a neuroprotective agent, affects synaptic functioning and enhances myelinisation [12]. The steroidogenesis of female sex hormones from cholesterol (Figure 1) involves the participation of six isoforms of cytochrome- dependent monooxygenase (CYP) and two hydroxysteroid dehydrogenases [13]. The first step is the cleavage of the side chain of cholesterol mediated by CYP11A to form the C21 steroids pregnenolone and progesterone. The C19 steroids, named androstenedione (precursor of testosterone) and dehydroepiandrosterone, are formed by hydroxylation and subsequent cleavage of the two-carbon side chain of their precursors with the participation of the CYP17 isoform, which have 17-hydroxylase and C17?20 lyase activity. Finally, estrone and estradiol are formed by aromatization of androstenedione and testosterone, catalyzed by the aromatase CYP19. The steroidogenesis pathway may occur in different tissues such as those forming the ovary, testis or placenta. Steroids are liberated to the bloodstream bound to plasma proteins or as conjugated forms, particularly, sulphates or glucuronides [14]. Biotransformation of estrogens to their conjugated forms occurs in the liver and renders to molecules less lipophilic, which are more easily excreted in urine and bile. The analysis of steroid hormones in biological samples can be employed as a diagnostic tool in diseases promoted by disorders in the steroids profile. Thus, the determination of estrone aids the diagnosis of Turner?s syndrome ????? The determination of estradiol is used in the diagnosis of precocious puberty in girls, amenorrhea and to monitor the follicle development in the ovary in the days prior to in-vitro fertilization [16]. Estriol, along with other hormones, enables to assess fetal diseases during pregnancy, such as Smith-Lemli-Opitz syndrome [17], or steroid 156 Nuevas plataformas anal?ticas en metabol?mica sulfatase deficiency [18]. Progesterone is suitable to be monitored during treatment of infertility [19]. Elevated levels of pregnenolone can be produced in polycystic ovarian disease [20] and hyperplasia [21]. Exogenous estrogens and progestogens have been administered for years in hormone replacement therapy aiming at the treatment of hormonal disorders, i.e. during menopause. Estrogens are also used in the treatment of several other diseases such as infertility or endometriosis, and in menstrual disorders, among others [22]. In addition, estrogens have encountered other applications in treatment and/or prevention of diseases that affect women in a major extension, or in the case of Alzheimer disease [23?25]. One of the most negative effects of the administration of exogenous estrogens is their contribution to the development and evolution of breast cancer and endometrial cancer [8,26,27]. Therefore, monitoring steroid levels in urine throughout hormone therapy or during pregnancy is also of clinical interest. In most cases, the biological samples used for determination of female steroid hormones are serum and urine. Due to the ease of collection and the abundant presence of metabolites, urine is an ideal bio-fluid for human and animalmetabolomic studies [28]. The purpose of this research is the development of a method for determination of glucuronide-conjugated and free steroids in human urine by a dual approach with or without ?- glucuronidase hydrolysis. This application, marked by the low concentration of steroids in urine and the complexity of this sample matrix, demands for the development of highly selective and sensitive analysis methods. The method proposed here is based on liquid chromatography tandem mass spectrometry (LC?MS/MS) with a triple quadrupole mass detector (QqQ). Before separation, an automated preconcentration and clean-up step has been optimized by solid-phase extraction (SPE) with a laboratory-on-valve (lab-on-valve or LOV) device. The automation of sample preparation enables to scaledown this step and improves precision by minimizing human intervention [29]. 157 J. Chromatogr. A, 1207 (2008) 46?54 Chapter 3 2. Experimental 2.1. Reagents and samples Acetonitrile and ammonia of LC?MS grade (Scharlab, Barcelona, Spain) and deionized water (18?cm) from a Millipore Milli-Q water purification system (Millipore, Bedford, MA, USA) were used for preparation Figure 1. Principal pathway of steroid hormone biosynthesis in the human ovary. Cytochrome P450 isoforms: CYP11A1 (cholesterol side chain cleavage), CYP17 (17?-hydroxylase), CYP19 (aromatase), CYP17 (17,20-lyase). 17?-HSD: 17?-hydroxysteroid dehydrogenase; 3?-HSD: 3?-hydroxysteroid dehydrogenase. 158 Nuevas plataformas anal?ticas en metabol?mica of chromatographic phases. HPLC grade methanol and water acidified with phosphoric acid (pH 4.5) were used in the SPE step. ?-Glucuronidase from Escherichia coli (200 UmL?1) was provided by Boehringer (Mannheim GmbH, Germany). Estrone, estradiol, 5-pregnen-3?-ol-20-one, estriol and progesterone were fromSigma?Aldrich (St. Louis, MO, USA). (E)- Diethylstilbestrol, also from Sigma?Aldrich, was used as internal standard. Stock standard solutions 1000 ? gmL?1 of each steroidwere prepared in methanol and can be stored at ?20 ? C in the dark for a month without alteration. Individual and multistandard solutions of steroids for the optimization study were daily prepared by dilution of the stock standard solutions also in methanol. Urine samples were voluntary donated by healthy women aged 23? 28. The samples were collected in sterile containers, buffered with 1mL of 2Msodium acetate buffer (pH 4.8), stored at 4 ? C until analysis within 24 h after collection. 2.2. Instruments and apparatus Analyses were carried out by reversed-phase liquid chromatography (RP-LC) followed by electrospray ionization (ESI) and tandem mass spectrometry detection. Separation was carried out with an Agilent (Palo Alto, CA, USA) 1200 Series LC system equipped with a binary pump, vacuum degasser, autosampler and thermostated column compartment, and detection with an Agilent 6410 Triple Quadrupole mass detector. An Agilent Zorbax Eclipse XDB-C18 (4.6mm? 150mm, 5? m particle size) chromatographic column was used for separation. Agilent Mass HunterWorkstation was the software for data acquisition, qualitative and quantitative analysis. The lab-on-valve manifold for sample preparation was a FIAlab 3000 sequential injection analyzer (Medina, WA, USA), equipped with a bi-directional microsyringe pump (1000 ? L), a six-way selector valve, 159 J. Chromatogr. A, 1207 (2008) 46?54 Chapter 3 a PVC monolithically structured LOV mounted on-top of the six-way selector valve, and an external 10-port twoposition selection valve (VICI, Valco, Houston, USA). The manifold is fully automated and controlled by the FIAlab for Windows version 5.0 software. Homemade mini-columns (50 mm? 1.5 mm) packed with 0.05 g Chromabond C18 or C18-Hydra? (Macherey Nagel, D?rem, Germany) as sorbent material (30?40 ? m particle diameter) were used for clean-up and preconcentration of the target analytes by SPE. Polyetheretherketone (PEEK) tubing (0.5 mm and 0.8 mm inner diameter) was used to connect the manifold components to the ports of the selection valve, including the SPE column. The total volume of the LOV system was 300 ? L. A thermostated water bath from Selecta (Barcelona, Spain) was used to develop the enzymatic hydrolysis step. A centrifuge (Selecta) was used after the hydrolysis step. 2.3. Sample preparation The approach for determination of free and glucuronideconjugated steroids is schematized in Figure 2A. The protocol for determination of glucuronides and total steroids required an enzymatic hydrolysis with ?- glucoronidase (40 ? L) by heating at 37 ?C for 18 h [30]. The determination of steroids was initiated by centrifuging 10mL urine at 3000rpm for 5 min in order to facilitate the performance of the SPE mini-column. Then, the internal standard (10 ngmL?1) was added to follow with the SPE step. The automated SPE process was carried out in the LOV manifold shown in Figure 2B. The operational sequence for automatic SPE in the LOV manifold can be summarized as follows: (1) activation of the mini-column with 2 mL methanol and conditioning with 1 mL carrier (both steps at 3 mLmin?1); (2) equilibration of the column with 1 mL carrier solution at 1.2 mLmin?1; (3) loading 10 mL sample at 1.2 mLmin?1; (4) washing the mini- column with 2 mL carrier and flushing of air to eliminate the rest of the 160 Nuevas plataformas anal?ticas en metabol?mica sample matrix; (5) elution of steroids with 150 ? L methanol in order to solubilise them. Urine samples were handled with the special care required for analysis of biological samples. Figure 2. Approach for determination o free and glucuronide-conjugated steroids. (A) Flow diagram for determination of the target analytes. (B) Configuration of the LOV manifold used for the SPE step. C, carrier; SP, syringe pump; SPV, syringe pump valve; HC, holding coil, S, sample; W, waste; A, air; E, eluent; C, mini-column; SV, switching valve. 161 J. Chromatogr. A, 1207 (2008) 46?54 Chapter 3 2.4. LC?MS/MS separation?determination method Mobile phases A and B were 0.1% ammonia in water and 0.1% ammonia in 95:5 acetonitrile?water (v/v), respectively. After sample injection (20 ? L), a linear gradient was programmed for 12 min from 80:20 A?B to obtain 20:80 A?B composition, which was held for 1 min. Then, the concentration of B was increased to 95% in 0.5 min and held for 6.5 min. The total analysis time was 20 min while 4 min was required for re- establishing and equilibrating the initial conditions. The flow rate was set at 1 mLmin?1 during the chromatographic process and the temperature of the analytical column was 12 ? C. The entire eluate was electrosprayed, ionized and monitored by MS?MS detection in MRM mode. Ionization was positive for progesterone and pregnenolone, and negative for estrone, estriol, estradiol and diethylstilbestrol. For this purpose, the MS polarity was switched by time segments according to the retention of the target analytes. The flow rate and temperature of the drying gas (N2) were 13 Lmin?1 and 350 ? C, respectively. The nebulizer pressure was 35 psi and the capillary voltage 4000 V. The dwell time was set at 200 ms. 3. Results and discussion 3.1. Optimization of the determination method The lack of methods for determination of female hormones in human urine has been ascribed to the low concentrations at which these compounds are found in this biological fluid. Therefore, the application of highly selective and sensitive detection techniques is crucial to achieve the low concentration levels of these hormones to be determined. Mass spectrometry after gas chromatography separation has provided excellent 162 Nuevas plataformas anal?ticas en metabol?mica results for determination of estrogens in sediments [31,32], water [33,34] or urine [35,36]. However, a derivatization step prior to chromatographic separation is mandatory for most of these compounds because they are thermally unstable, non-volatile or polar [35]. Derivatization makes sample preparation laborious and time-consuming, and may lead to losses [37,38] or degradation of the target steroids [39]. LC?MS/MS can be a competitive option because of the recent improvements in this hyphenated technique, particularly in interface performance. The use of a triple quadrupole mass detector together with the multiple reaction monitoring (MRM) mode enables the development of methods with low detection and quantification limits and a great identification capability even in complex samples. For this reason, this was the alternative selected in this research. This combination is based on the selection of transitions for each analyte from the precursor ion to its most characteristic product ions resulting in highly selective methods. Thus, the identification of a certain compound is based on the presence of these transitions in its mass spectrum; transitions which are optimized in order to select the most intense for quantification of the target analytes and favoured by practical removal of background noise in the chromatograms. The optimization study was carried out with a steroid multistandard under the chromatographic conditions explained in Section 2, which enables complete separation within 20 min. (E)-Diethylstilbestrol, used as internal standard, was also considered in this optimization. The criteria for selection of diethylstilbestrol were its physical and chemical properties similar to those of the analytes and its absence in the samples. A parallel study was carried out to test the influence of the ionization mode for each analyte. This study was performed with the multistandard solution in scan mode in which only the second mass filter is operating to detect ions within a limited mass range between 100 and 1000 m/z with a scan time of 500 ms. Total ion chromatograms corresponding to this study are shown in Figures 3 A y B, from which it can be concluded that progesterone and pregnenolone should be monitored in positive mode, while negative ionization was better for estriol, estrone and estradiol. Table 163 J. Chromatogr. A, 1207 (2008) 46?54 Chapter 3 1 shows the precursor and product ions for each steroid as well as the optimum values of MS?MS parameters: voltage of the first quadrupole for isolation of the precursor ion and collision energy for efficient fragmentation. Figure 4 shows the MRM chromatograms obtained with amultistandard solution and the quantification transition. The study of the detection step finished with the selection of the dwell time, tested within the range 10?250 ms, being 200 ms the best value. The application of the MRM approach enables identification and confirmation of the presence of the target steroids in a urine sample. Thus, monitoring the three most characteristic transitions for each analyte provides an extra level of selectivity. Table 1 shows the strategy to be followed for confirmatory analysis of the target female hormones in urine just before quantitative analysis. Table 1. (A) Optimization of the MS?MS step for qualitative (confirmatory analysis) and (B) quantitative analysis of female steroid hormones in human urine by multiple reaction monitoring. (A) q1: first quadrupole Analyte Voltage q1 (V) Precur sor ion (m/z) Collision voltage (V) Product ions (m/z) Quantification transition Estriol 120 287 60 255/171/145 287?145 Estradiol 140 271 50 239/183/145 271?183 Estrone 140 269 45 253/183/145 269?145 IS 140 267 35 237/222/131 267?222 Progesterone 120 315 20 273/160/109 315?109 Pregnenolone 120 317 20 256/159/109 317?109 164 Nuevas plataformas anal?ticas en metabol?mica (B) Precursor ion (m/z) Qualifier for steroid identification Female hormone Qualifier for confirmatory analysis Qualifier for confirmatory analysis Confirmatory analysis 287 145 Estrogen 255 171 Estriol 271 145 Estrogen 239 183 Estradiol 269 145 Estrogen 253 183 Estrone 315 109 Progestogen 273 160 Progesterone 317 109 Progestogen 256 159 Pregnenolone Figure 3. (A) Total ion chromatograms obtained with a 10 ?gmL?1 standard solution in full scan mode with positive (A) and negative (B) ionization. (B) 165 J. Chromatogr. A, 1207 (2008) 46?54 Chapter 3 166 Nuevas plataformas anal?ticas en metabol?mica 3.2. Optimization of the solid-phase extraction step An automated SPE step was the selected approach for clean-up and preconcentration of steroids from urine prior to determination by LC?MS? MS. Several studies based on SPE for isolation and enrichment of steroid hormones have previously been reported [38,40]. Different sorbent materials have been tested in these studies concluding that octadecyl-C18 phase provides the best results. In this research, two types of C18 sorbents, conventional C18 and C18-Hydra? , were compared to assess the location of silanol active groups. While silanol active groups are randomly distributed in conventional C18, C18-Hydra? is characterized by higher activity of silanol groups located on the silica surface. Figure 4. Multiple reaction monitoring chromatograms obtained with a standard solution (10 ngmL?1, including internal standard) at the most sensitive transition for each analyte (working conditions, under Section 2). 167 J. Chromatogr. A, 1207 (2008) 46?54 Chapter 3 The main variables involved in the SPE step were optimized in order to study their influence on the clean-up and preconcentration efficiency. The variables were sample volume, sample loading flow-rate, washing-solution volume, eluent flow-rate and elution volume. Table 2 shows these variables, the ranges studied and the resulting optimum values. Acetonitrile, methanol and ethanol were tested for elution of the retained compounds from the SPE column, methanol providing a more uniform elution. The fraction of eluate collected for chromatographic analysis was another key variable in the SPE. The elution profile was studied by LC?MS/MS analysis of consecutive 50 ? L fractions of the eluate. The elution was homogeneous in the first 150 ? L of eluate and, for this reason, this fraction was isolated for analysis. No elution of the target analytes was observed in the subsequent fractions and, thus, these fractions were discarded. Under the optimum working conditions, preconcentration factors for 10 mL urine were from 59.1 to 72.3, as shows Table 3. This concentration parameter was estimated with spiked samples by comparison with direct analysis without hormones enrichment. 3.3. Optimization of the hydrolysis step Variable Tested range Optimum value Sorbent C18, C18-Hydra? C18-Hydra? Loading flow-rate (mLmin-1) 0.3???? 1.2 Sample volume (mL) 5??? 10 Washing volume (mL) ??? 2 Washing flow-rate (mLmin-1) ????? 3 Volume of eluate (mL) 50????? 200 Elution flow-rate (mLmin-1) ????? 0.3 Table 2. Optimization of the solid-phase extraction variables. 168 Nuevas plataformas anal?ticas en metabol?mica Estrogens can be excreted in the urine as glucuronides or sulphate- conjugated forms. Nevertheless, the content of sulphates estrogens in human urine is really low. Therefore, the determination of total or glucuronide-conjugated steroids requires implementation of a hydrolysis step prior to other steps of the analytical process. Non-specific chemical hydrolysis with hydrochloric acid and heating at 70 ? C for 1 h in a thermostatic water bath has been reported for total estrogens. The alternative is the use of ?-glucuronidase for specific hydrolysis of steroids conjugated with glucuronic acid. The procedure for enzymatic hydrolysis is described in Section 2. The main limitation of this procedure is the time required for hydrolysis completion (18 h). This reaction is performed under physiological conditions preserving in this way the integrity of the sample [39]. 3.4. Characterization of the method Calibration plots were run under the optimum working conditions with eight spiked samples at different concentrations of analytes using the relative peak area (peak area of each compound vs. that of the IS) as a function of the concentration of each compound. The regression coefficients for the dynamic ranges are shown in Table 3. As can be seen, they range from 0.9937 to 0.9999. The limits of detection (LODs) and quantification (LOQs) were calculated from the MRM chromatograms obtained with urine samples on the basis of a minimal accepted value of the signal-to-noise (S/N) ratio of 3 and 10, respectively. The background noise was estimated by the peak-to-peak baseline near each analyte peak. As shows Table 3, LODs range from1.8 to 18 pg on-column while LOQs are from 6 to 61 pg. In order to evaluate the precision of the proposed method, intra-day and inter-day variability were evaluated in a single experimental set-up with duplicates [41]. 169 J. Chromatogr. A, 1207 (2008) 46?54 Chapter 3 The experiments were carried out using 10 mL of urine sample spiked with 2 ngmL?1 of each of the target analytes under the optimum working conditions. Two measurements per day were performed on seven days. As can be seen in Table 3, the results obtained, expressed as relative standard deviation (RSD), are from 1.93 to 5.05% for intra-day variability and from 4.04 to 10.99% for inter-day variability. The absence of reference materials for the determination of steroids in urine led to spike urine samples with the target analytes in order to test the suitability of the proposed method. With this aim, stock standard solutions were added to three aliquots of urine at 2 ngmL?1. Three aliquots of a non-spiked sample were also analysed. As can be seen in Table 3, the spiked steroids were quantified at all concentrations with an error below 6% resulting in recoveries indexes from 95.29 to 103.1%. 3.5. Application of the method to urine samples In order to demonstrate the applicability of the proposed method, urine samples from five volunteer women were analysed following the procedure here proposed. For each sample, six replicates were analysed in order to obtain the content of free and glucuronide female steroids. Table 4 Preconcentration factor LOD (pg) LOQ (pg) R2 Recovery ?SD a (%) Intra-day variability (%) Inter-day variability (%) Estriol 72.3 5.4 17.9 0.9988 99.7?1.3 3.22 6.73 Estradiol 71.4 18.3 61.1 0.9982 98.7?0.7 1.93 4.04 Estrone 70.9 8.6 29.8 0.9968 95.3?5.8 3.41 5.35 Pregnenolone 60.7 1.8 5.9 0.9999 103.1 ?6.1 5.05 10.99 Progesterone 59.1 10.5 35.1 0.9987 96.2?4.1 4.09 8.41 a SD: standard deviation. Table 3. Characterization of the method. 170 Nuevas plataformas anal?ticas en metabol?mica shows the results found for each analyte in the five samples. Pregnenolone was the only hormone detected as free form in one of the samples. The absence of free hormones is justified by biotransformation of these compounds in the liver for its excretion. Therefore, this method is able to detect abnormal values, thus resulting in an interesting tool for clinical diagnosis. On the other hand, estrone and pregnenolone were quantified in most of the samples as conjugated forms. Figure 5 shows the TIC chromatogram for one of these samples after enzymatic hydrolysis to quantify glucuronide-conjugated steroids. Figure 5. Total ion chromatogram provided by a urine sample using the proposed MRM method. Estriol Est rone Pregnenolone Progesterone Estradiol Free Free + glucuronide conjugated Free Free + glucuronide conjugated Free Free + glucuronide conjugated Free Free + glucuronide conjugated Free Free + glucuronide conjugated N.D. 0.184 ? 0.010 N.Q. 2.2386? 0.050 N.Q. 5.3934 ? 0.15 N.Q. 0.486 ? 0.015 N.Q. 0.741? 0.005 N.D. N.Q. N.D. N.Q. N.D. N.Q. N.Q. N.Q. N.D. N.D. N.D. N.Q. N.D. 1.319 ? 0.030 N.Q. 4.9112 ? 0.2 N.Q. 0.303 ? 0.006 N.D. N.Q. N.D. N.Q. N.D. 0.329 ? 0.009 2.224 ? 0.050 3.271 ? 0.08 N.Q. N.Q. N.D. N.Q. N.D. N.D. N.D. 0.837 ? 0.011 N.Q. 2.189 ? 0.10 N.Q. N.Q. N.D. N.Q. Table 4. Concentration of steroids in urine samples (ngmL?1 ?SDa) 172 Nuevas plataformas anal?ticas en metabol?mica 4. Conclusions A method allowing identification and quantification of estrogens and progestogens in urine at low ngmL?1 levels has been here proposed. It should be emphasized that the use of a lab-on-valve approach led to automated sample preparation with minimum human intervention. Thus, the contact with biological samples as well as contamination and irreproducibility problems are practically avoided by precise delivery of small volumes. The high selectivity and sensitivity of the method demonstrate its suitability for determination of these compounds in urine samples. This can be of interest to study the status of these hormones or their metabolic evolution in the human body. Thus, it constitutes a useful method in studies of fertility, variability of the endogenous hormones within a menstrual cycle or during hormone replacement therapy, and also in disease diagnosis, i.e. in the diagnosis of endocrine disorders. The method can also be applied to any biological fluid after modification of the sample preparation step, if required. The present research is an example of how the improvements in mass spectrometers interfaced to LC (e.g., MALDI?TOF equipment [42]) increase the availability of these detectors to characterize a number of lipids with minimal sample preparation, of great interest to study the profound influence of these compounds on systems biology. 5. Acknowledgments The Spanish Ministerio de Ciencia e Innovaci?n (MICINN) is acknowledged for financial support through project CTQ2009-07430. F.P.C. 173 Chapter 3 J. Chromatogr. A, 1207 (2008) 46?54 is grateful to MICINN for a Ram?n y Cajal contract (RYC-2009-03921). B.A.S. is also grateful to Ministerio de Ciencia y Tecnolog?a for an FPI scholarship (BES-2007-15043). 6. References [1] H. Kitano, Science 295 (2002) 1662. [2] L. Hood, D. Galas, Nature 421 (2003) 444. [3] S.P. Gygi, Y. Rochon, B.R. Franza, R. Aebersold, Mol. Cell. Biol. 19 (1999) 1720. [4] J.K. Nicholson, I.D. Wilson, Nat. Rev. Drug Discov. 2 (2003) 668. [5] G.L. Bannenberg, J. Aliberti, S. Hong, A. Sher, C.N. Serhan, J. Exp.Med. 199 (2004) 515. [6] E. Fahy, S. Subramaniam, H.A. Brown, C.K. Glass, A.H. Merrill, R.C. Murphy, C.R.H. Raetz, D.W. Russell, Y. Seyama,W. Shaw, T. Shimizu, F. Spener, G. van Meer,M.S. VanNieuwenhze, S.H. White, J.L. Witztum, E.A. Dennis, J. Lipid Res. 46 (2005) 839. [7] S.C. Sala, V. Martineti, A.M. Carossino, M.L. Brandi, Expert Rev. Endocrinol. Metabolism 2 (2007) 503. [8] Y. Tsuchiya, M. Nakajima, T. Yokoi, Cancer Lett. 227 (2005) 115. [9] R.A. Khalil, Future Cardiol. 3 (2007) 283. [10] H.J. Tede, Clin. Exp. Pharmacol. Physiol. 34 (2007) 672. [11] R.C. Tuckey, Placenta 26 (2005) 273. [12] S. Veiga, L.M. Garc?a-Segura, I. Azcoitia, J. Neurobiol. 56 (2003) 398. [13] G.E. Ackerman, B.R. Carr, Rev. Endocr. Metab. Dis. 3 (2002) 225. [14] R. Raftogianis, C. Creveling, R. Weinshilboum, J. Weisz, J. Natl. Cancer I. Monographs 27 (2000) 113. 174 Nuevas plataformas anal?ticas en metabol?mica [15] C.H. Gravholt, B. Svenstrup, P. Bennett, J.S. Christiansen, Clin. Endocrinol. 50 (1999) 791. [16] D.J. Cahill, P.G. Wardle, C.R. Harlow, L.P. Hunt, M.G.R. Hull, Hum. Reprod. 15 (2000) 1909. ???? E? Schoen, C? ?orem, ?? O??eefe, R? ?rieger, ?? ?alton, T?T? To, Obstet? Gynecol. 102 (2003) 167. [18] M.N. Ahmed, A. Killam, K.H. Thompson, M.B. Qumsiyeh, Obstet. Gynecol. 92 (1998) 687. [19] D. de Ziegler, R. Fanchin, Steroids 65 (2000) 671. [20] R.J. Chetkowski, R.J. Chang, J. DeFazio, D.R. Meldrum, H.L. Judd, Obstet. Gynecol. 64 (1984) 27. [21] H. Ueshiba, M. Segawa, T. Hayashi, Y. Miyachi, M. Irie, Clin. Chem. 37 (1991)1329. [22] M.J. L?pez de Alda, D. Barcel?, J. Chromatogr. A 892 (2000) 391. [23] A. Morinaga, M. Hirohata, K. Ono, M. Yamada, Biomed. Biopharm. Res. Commun. 359 (2007) 697. [24] A. Paganini-Hill, V.W. Henderson, Arch. Intern. Med. 156 (1996) 2213. [25] R. Li, Y. Shen, L.B. Yang, L.F. Leu, C. Finch, J. Rogers, J. Neurochem. 75 (2000) 1447. [26] M.C. Pike, D.V. Spicer, L. Dahmoush, M.F. Press, Epidemiol. Rev. 15 (1993) 17. [27] H.S. Feigelson, B.E. Henderson, Carcinogenesis 17 (1996) 2279. [28] M. Lauridsen, S.H. Hansen, J.W. Jaroszewski, C. Cornett, Anal. Chem. 79 (2007) 1181. [29] J. Ruzicka, Analyst 125 (2000) 1053. [30] M.J. L?pez de Alda, D. Barcel?, J. Chromatogr. A 938 (2001) 145. 175 Chapter 3 J. Chromatogr. A, 1207 (2008) 46?54 [31] M.S. D?az-Cruz, M.J. L?pez de Alda, R. L?pez, D. Barcel?, J. Mass Spectrom. 38 (2003) 917. ???? ?? ??jkov?, ?? ?ulkrabov?, ?? Schu? rek, ?? ?ajs?lov?, ?? ?oustka, M? N?pravn?kov?, V. Kocourek, Anal. Bioanal. Chem. 387 (2007) 1351. [33] U. Knust, T. Strowitzki, B. Spiegelhalder, H. Bartsch, R.W. Owen, Rapid Commun. Mass Spectrom. 21 (2007) 2245. [34] A. Stopforth, B.V. Burger, A.M. Crouch, P. Sandra, J. Chromatogr. B 856 (2007) 156. [35] C. Desbrow, E.J. Routledge, G.C. Brightly, J.P. Sumpter,M.Waldock, Environ. Sci. Technol. 32 (1998) 1549. [36] D. Arroyo, M.C. Ortiz, L.A. Sarabia, J. Chromatogr. A 1157 (2007) 358. [37] A.A.M. Stolker, U.A.T. Brinkman, J. Chromatogr. A 1067 (2005) 15. [38] A. Stafiej, K. Pyrzynska, F. Regan, J. Sep. Sci. 30 (2007) 985. [39] L. Mao, C. Sun, H. Zhang, Y. Li, D.Wu, Anal. Chim. Acta 522 (2004) 241. [40] A. Salvador, C. Moretton, A. Piram, R. Faure, J. Chromatogr. A 1145 (2007) 102. [41] D.L. Massart, B.G.M. Vadeginste, L.M.C. Buydens, S. de Jong, P.J. Lewi, P.J. Smeyers-Verbeke, Handbook of Chemometrics and Qualimetrics Part A, Elsevier, Amsterdam, 1997. [42] J. Schiller, R. Suss, B. Fuchs, M. Muller, O. Zschornig, K. Arnold, Frontiers Biosci. 12 (2007) 1. C H A PTER 4 : Ultrasound - enhanced enzymatic hydrolysis of conjugated female steroids for their analysis by LC?MS/MS in urine Ultrasound - enhanced enzymatic hydrolysis of conjugated female steroids as pretreatment for their analysis by LC ?MS/MS in urine B. ?lvarez-S?nchez, F. Priego-Capote, M.D. Luque de Castro Department of Analytical Chemistry, University of C?rdoba, Annex C-3 Building, Campus of Rabanales, E-14071 C?rdoba, Spain The Analyst, 134 (2009) ???????22 181 The Analyst, 134 (2009) 1416?1422 Chapter 4 Ult rasound- enhanced enzymatic hydrolysis of conjugated female steroids as pretreatment for their analysis by LC ?MS/MS in urine B. ?lvarez-S?nchez, F. Priego-Capote, M.D. Luque de Castro Abstract A fast, selective and sensitive method is here proposed for the analysis of female steroid hormones as conjugated forms (mainly, glucuronides and sulfates). The method implements an enzymatic hydrolysis (?-glucuronidase with sulfatase activity) kinetically enhanced by ultrasonic energy in order to generate the free steroid forms. This enables a drastic shortening of the time required for this step as compared with conventional protocols (from 12?18 h to 30 min). After hydrolysis, the free steroid hormones were isolated and preconcentrated by automated solid-phase extraction and the eluate was subsequently analysed by liquid chromatography?tandem mass spectro- metry (LC?MS/MS). The target analytes were confirmed and quantified by multiple reaction monitoring (MRM). The detection and quantification limits were within 0.06?0.8 ng mL-1 and 0.19?2.69 ng mL-1, respectively. The precision of the method, expressed as intra-day and inter-day variability, ranged between 2.1 and 5.2% and between 4.9 and 8.0%, respectively. A complementary study was carried out to assess the storage conditions of urine samples. This study is crucial in those applications involving metabolic processes as the integrity of the sample has to be preserved. Nuevas plataformas anal?ticas en metabol?mica 182 1. Introduction Estrogens and progestogens are a group of steroid-sex hormones with essential functions in female metabolism. In the body, these are all derived from cholesterol in a biosynthetic pathway,1 which mainly occurs in the ovaries, corpus luteum and placenta and, in a lesser extent, in liver, adrenal glands and breast.2 The three major naturally present estrogens in women are estradiol, estriol and estrone. Endogenous estrogens mainly exert reproductive functions as they promote the development of female secondary sex characteristics, and are also involved in the thickening of the endometrium and other aspects of menstrual cycle regulation.3 Exogenous estrogens are frequently administrated as part of oral contraceptives, in estrogen replacement therapy of postmenopausal women and in treatment of menstrual disorders.4,5 Progesterone is the most abundant progestogen in metabolism, secreted by the corpus luteum and the placenta. Progesterone is required in embryo implantation, pregnancy maintenance and development of mammary tissue for milk production.3 Progesterone is also an intermediate in the steroidogenesis of estrogens, androgens and adrenal corticosteroids. Pregnenolone, the metabolic precursor of progesterone, belongs to the group of neurosteroids, which are found in high concentrations in certain areas in the brain, where they affect the synaptic functioning, are neuroprotective and enhance myelinization.6 Furthermore, pregnenolone potentially improves cognitive and memory functioning. Female steroid hormones are present in urine almost exclusively as conjugated metabolites such as glucuronides, sulfates, diglucuronides, disulfates and sulfoglucuronides.7 Conjugation of endogenous steroids 183 The Analyst, 134 (2009) 1416?1422 Chapter 4 (excepting progesterone as the hydroxyl group is substituted by a ketone group) constitutes an important metabolic pathway of female hormones biotransformation.1 The purpose of this metabolic pathway is to increase the polarity of steroids in order to favour their transport into biological fluids. The most abundant conjugated derivatives are sulfates and glucuronides (estrone sulfate is found at concentrations about ten-times higher than unconjugated estrone8). As they are less active ligands for steroid receptors (SR), conjugated forms are considered as hormone precursors, suitable to cross cellular membranes of target tissues via organic ion transporters, being hydrolyzed by intracellular membrane-bound enzymes. Finally, its union to SR causes transcriptional activation by SR-activated union to especific DNA sequences. Conjugated female steroids also exert regulatory effects by means of sulfation?desulfation within target tissues. Despite the free steroids are the active forms of these metabolites, the analysis of the conjugated hormones is of great interest to assess their metabolism. However, due to the lack of a complete profile of conjugated standards and the variety of existing forms, determination of total steroid hormones requires hydrolysis prior to other steps of the analytical process.8 Hydrolysis has traditionally been carried out by enzymatic incubation. This step usually takes 12?18 h,?,???? being performed under physiological conditions in order to preserve sample stability. Non-specific chemical hydrolysis has also been reported.12 which is simpler and less time- consuming than enzymatic hydrolysis. However, the reaction is usually carried out with hydrochloric acid by heating at 70 ? C for 1 h in a thermostatic water bath. Under these drastic conditions, either the stability of the target analytes and sample integrity are not ensured. One alternative to enhance enzymatic hydrolysis is the assistance with ultrasonic energy,15 still an emerging field. The mechanism through which ultrasound enhances enzymatic reactions has not been elucidated. Nevertheless, cavitation seems to promote the reaction rates by an increased analyte transport from the bulk to the enzyme surface, which Nuevas plataformas anal?ticas en metabol?mica 184 leads to an effective contact between substrate and active sites of the enzyme.16 There are few precedents in the literature on the application of ultrasound to accelerate enzymatic reactions. Most of the reported methodologies aim to liberate protein-bonded compounds by means of protease digestion with unspecific bonds breaking.????? The drastic reduction in the analysis time and the simplicity of the methodology justify the extension of ultrasound-assisted enzymatic reactions to other analytical methodologies. With these premises the applicability of ultrasound energy to accelerate the enzymatic hydrolysis of conjugated estrogens present in female urine is here tested. To achieve this goal the optimization of the main variables involved in the enzymatic step and kinetic characterization of the ultrasound-assisted reaction are mandatory steps. Apart from enzymatic hydrolysis, other limitation in the analysis of female steroids hormones is their low concentration in urine (ng mL?1 level). Matrix effects associated to urine samples can also be problematic. There is a clear demand for methods implementing both preconcentration/clean-up steps as well as highly selective and sensitive detection. The coupling between solid-phase extraction (SPE) and LC?MS/MS is a powerful solution because of its possibilities being selected in this research. Multiple reaction monitoring (MRM) mode in MS/MS was used for confirmatory and quantification analysis of female steroid hormones. 2. Experimental 2.1. Instruments and apparatus Figure 1 shows an overview of the pre-treatment steps including ultrasound-assisted enzymatic hydrolysis of female steroids and lab-on- 185 The Analyst, 134 (2009) 1416?1422 Chapter 4 valve (LOV) preconcentration. The eluate was subsequentlly individual determination by HPLC?triple quad. Ultrasonic irradiation was applied by means of a Branson 450 digital sonifier (20 KHz, 450 W) with tunable amplitude and duty cycle, which was equipped with a cylindrical titanium alloy probe (12.70 mm in diameter). The sample was placed in a 20-mL container with the probe tip placed at a fixed distance from the bottom (2 mm), totally immersed into the solution in order to avoid formation of aerosol and foam that create dead zones where cavitation does not occur.15 The temperature was kept at 37 ? C during the enzymatic hydrolysis by means of a thermostated water-bath. A FIAlab 3000 sequential injection analyzer (Medina, WA, USA) was used for sample preparation, The manifold was equipped with a bi- directional microsyringe pump (1000 ? L), a six-way selector valve, a PVC monolithically structured lab-on-valve (LOV) mounted on-top of the 6-way selector valve, and an external 10-port 2-position selection valve (VICI, Valco, Houston, USA). The manifold is fully automated and controlled by the FIAlab for Windows version 5.0 software. Chromabond C18 Hydra? , 30?40 ? m particle diameter, (Macherey Nagel, D?rem, Germany) was used as sorbent material in the SPE step. Polyetheretherketone (PEEK ) tubing (0.5 mm and 0.8 mm inner diameter) was used to connect the manifold Figure 1. Workflow of the sample pretreatment. Nuevas plataformas anal?ticas en metabol?mica 186 components to the ports of the selection valve, including the SPE column. Polytetrafluorethylene (PTFE) tubing (1.5 mm inner diameter) from An?lisis V?nicos (C. Real, Spain) was used for making the minicolumn. The analyses were carried out by reversed-phase liquid chromatography (RP-LC) followed by electrospray ionization (ESI) and tandem mass spectrometry detection. The separation was carried out with an Agilent (Palo Alto, CA, USA) 1200 Series LC system equipped with a binary pump, vacuum degasser, autosampler and thermostated column compartment, and detection with an Agilent 6410 Triple Quadrupole mass detector. An Agilent Zorbax Eclipse XDB -C18 (4.6 x 150 mm, 5 ? m particle size) chromatographic column was used for separation. Agilent MassHunter Workstation was the software for data acquisition, qualitative and quantitative analysis. A thermostated water-bath from Selecta (Barcelona, Spain) was used to develop the overnight incubation for the optimization of the hydrolysis step. A centrifuge (Selecta) was used for cleaning the sample after hydrolysis. 2.2. Reagents and samples Acetonitrile and ammonia of LC?MS grade (Scharlab, Barcelona, Spain) and ultrapure water (?? m??cm) from a Millipore Milli-Q water purification system (Millipore, Bedford, MA, USA) were used for preparation of the chromatographic phases. HPLC grade methanol and water acidified with phosphoric acid (pH 4.5) were used for SPE. ?ure standards of steroids ?estrone, estradiol, ?-pregnen-??-ol-20- one, estriol and progesterone? and internal standard ?diethylstilbestrol? were from Sigma?Aldrich? Individual stock solutions of ???? ?g mL-1 were prepared in methanol and can be stored at ??? ? C in the dark for a month 187 The Analyst, 134 (2009) 1416?1422 Chapter 4 without alteration. Individual and multistandard solutions of steroids for the optimization study were daily prepared by dilution of the stock standard solutions also in methanol. ?-Glucuronidase from Helix Pomatia (Type H-1), of 21900 kU g-1 and ???,??? U g-1 sulfatase activity from Sigma?Aldrich (St? Louis, MO, USA) was dissolved in 5-mL 1 M ammonium acetate (pH 5). Urine samples were obtained from premenopausal women. All participant subjects were healthy and non-pregnant. No considerations about life style, intake of contraconceptives or stage within the menstrual cycle were made. The samples were collected in sterile containers, buffered with 1-mL 2 M sodium acetate buffer (pH 5) and stored at ??? ? C in the dark until analysis. Calibrations were run with blank male urine, collected and stored as described. For optimization of the hydrolysis step, 100-mL urine from seven women was pooled and stored at ??? ?C for a week? 2.3. Ultrasoun d- assisted enzymatic hydrolysis (USAEH) Aiming at performing the USAEH, 10-mL human urine were mixed with 1-mL sodium acetate buffer (0.15 M, pH 5), 10 ?L internal standard (stock solution 1000 ng mL?1 in methanol) and the appropriate volume of enzyme solution (500 U mL?1 urine) was added; then, the mixture was immersed in a thermostated water-bath at 37 ? C. The ultrasonic probe was dipped into the bulk solution, with the tip horn at the minimum distance from the bottom of the container without contacting it. Hydrolysis of conjugated steroids was performed under ultrasonic irradiation for 30 min at 35% duty cycle (fraction of ultrasound application per each second), and output amplitude 50% (converter applied power 400 W). For convencional hydrolysis of conjugated estrogens, the incubation mixture was prepared as described for the USAEH procedure, and incubated at 37 ?C for 18 h. After the hydrolysis step, the sample was centrifuged at 3000 rpm for 5 min and the supernatant subjected to the SPE step. Nuevas plataformas anal?ticas en metabol?mica 188 2.4. Solid - phase extraction (SPE) The SPE minicolumn was made by packing 0.05 g of C18-hydra in an 8-cm PTFE tubing (3 mm E.D.). The SPE step was performed as follows: (1) activation of the mini-column with 2-mL methanol and conditioning with 1- mL carrier (both steps at 3 mL min-1); (2) equilibration of the column with 1- mL carrier solution at 1.2 mL min-1; (3) loading of 10-mL sample at 1.2 mL min-1; (4) washing of the minicolumn with 2-mL carrier and flushing air to eliminate the rest of sample matrix; (5) elution of steroids with 150 ?L methanol. Steps (1) and (2) enables the renewal of the column for analysis of next sample. In this way, cross-contamination effects are avoided. 2.5. Chromatographic separation and detection Mobile phases A and B wer e 0.1% ammonia in water and 0.1% ammonia in 95:5 acetonitrile?water (v/v), respectively. Initially, the mobile phase was ????? A??? After sample injection (20 ?L), a linear gradient was programmed for 12 min to obtain a 20:80 A?B composition, which was held for 1 min. Then, the concentration of B was increased up to 95% in 0.5 min and held for 6.5 min. The total analysis time was 20 min, while 4 min was required for re-establishing and equilibrating the initial conditions. The flow rate was set at 1 mL min-1 during the chromatographic process and the temperature of the analytical column was 12 ? C. The entire eluate was electrosprayed, ionized and monitored by MS/MS in MRM mode. Ionization was positive for progesterone and pregnenolone, and negative for estrone, estriol, estradiol and diethylstilbestrol. For this purpose, the MS polarity was switched by time segments according to the retention of the target analytes. The flow rate and temperature of the drying gas (N2) were 13 L min-1 and 350 ? C, respectively. 189 The Analyst, 134 (2009) 1416?1422 Chapter 4 The nebulizer pressure was 35 psi and the capillary voltage 4000 V. The dwell time was set at 200 ms. 3. Results and Discussion 3. 1. Optimization of the SPE and LC ?MS/MS steps The conditions for the SPE and LC?MS/MS steps were optimized using urine samples hydrolysed by the conventional enzymatic protocol. The optimum results for both steps are shown in Table 1. Different volumes of sample were tested in order to assess the capacity of the home-made column, which provided an excellent behaviour up to 10 mL of sample. The elution of the retained target steroids was tested with different solvents such as acetonitrile, methanol or ethanol because of the non-polar properties of the analytes after hydrolysis. The optimum elution was performed with 150 ? L of methanol, which was dried in a speed-vac with subsequent reconstitution in 20 ? L methanol. Although methanol was used as reconstitution solvent for chromatographic analysis, the performance was not affected because of the low injection volume. Figure 2, which corresponds to the analysis of the urine pool using the conventional enzymatic protocol shows this fact. Figure 2. Total ion chromatogram with negative and positive ionization modes for estrogens and progestogens, simultaneously. Nuevas plataformas anal?ticas en metabol?mica 190 Figure 3. Analysis of the urine pool using the conventional enzymatic protocol for hydrolysis of the conjugated steroids: MRM chromatograms for each free steroid generated with the quantitation transition (operating LC?MS/MS ??????????????????????????????????????????????????? 191 The Analyst, 134 (2009) 1416?1422 Chapter 4 Figure 3 shows the MRM chromatogram for each compound obtained with the transition used for quantification. The transitions that were monitored for confirmatory analysis of the target analytes are included in Table 1. In this way, the conjugated steroids can be determined with a high level of sensitivity and selectivity. Table 1. Optimum working conditions of the SPE and MS/MS detection steps. 3.2. Analytical features of the determination method Calibration plots were obtained by using the standard peak?internal standard peak ratio as a function of the standard concentration. Calibration runs were performed with spiked male urine with no detectable levels of SPE Sorbent C18 Hydra (0.05 g) Eluant Methanol (150 ?L) Sample volume* 10 mL Analyte Voltage q1 (V) Polarity Collision voltage (V) Qualitative ions (m/z) Quantification transition ( m/z) Estriol 120 Negative 60 255.1/171.1/144.9 287.1?145.1 Estradiol 140 Negative 50 238.9/183.1/145.1 271.1?183.1 Estrone 140 Negative 45 253.2/183.2/145.1 269.1?145.1 Diethylstilbestrol (IS) 140 Negative 35 237.1/222.0/131.7 267.1?222.0 Progesterone 120 Positive 20 273.1/160.0/109.4 315.0?109.2 Pregnenolone 120 Positive 20 256.0/158.8/109.4 317.0?109.2 MULTIPLE REACTION MONITORING MS/MS MODE *Initial sample volume. Volume used for SPE , 11 mL. Nuevas plataformas anal?ticas en metabol?mica 192 female hormones, which was proved by applying the proposed method to 10 mL of urine to the overall process? Ten concentration levels ?between ??? and 500 ng mL-1? were used to build the calibration curves by subjecting them to the complete analytical process. The regression coefficients (between 0.992 and 0.999) and the linearity range for all the steroids are shown in Table 2. Table 2. Features of the proposed method. The limits of detection (LOD), expressed as the mass of analyte which gives a signal that is 3? above the mean blank signal (where ? is the standard deviation of the blank signal) ranged between 0.06 and 0.8 ng mL-1 (1.2?16 pg on column), as shows Table 2. The limits of quantification, expressed as the mass of analyte which gives a signal 10? above the mean blank signal, ranged from 0.19 to 2.69 ng mL-1 (3.8?53.8 pg on column). 3.3. Multivariate optimization of ultrasound - enhanced enzymatic hydrolysis (USAEH) ?-Glucosidase is an enzyme commonly used for hydrolysis prior to the analysis of metabolites in urine. As previously indicated, the main limitation of this conventional protocol is the time required for hydrolysis Analyte R2 LOD (ng mL - 1 ) LOQ (ng mL - 1 ) Linear dynamic range (ng mL - 1 ) Estriol 0.9976 0.66 2.23 ???????? Estradiol 0.9983 0.36 1.12 ???????? Estrone 0.9996 0.06 0.19 ???????? Pregnenolone 0.9997 0.81 2.69 ???????? Progesterone 0.9956 0.12 0.39 ???????? 193 The Analyst, 134 (2009) 1416?1422 Chapter 4 completion which usually ranges from 12 to 18 h. We performed a kinetics study of the evolution of the hydrolysis process in the conventional protocol using the urine pool. Figure 4 shows this evolution for all steroids but progesterone, which is not conjugated. As can be seen, 5.5 hours were required for hydrolysis of all compounds except for estrone that needed 18 hours. This can be explained by the high concentration of estrone. The alternative proposed here is based on the assistance of the hydrolysis process with ultrasound energy. The optimization sequence followed consists of a multivariate approach to find the significant variables involved in the enzymatic hydrolysis. A half-fraction type-V resolution design allowing three degrees of freedom and involving 8 randomized runs plus 3 centre points was built for a screening study of the main factors affecting the USAEH step, namely: ultrasound radiation amplitude, percentage of duty cycle of ultrasound exposure, reaction time and enzyme concentration (Table 3). The upper and lower values given to each factor were selected from preliminary experiments. The conclusions of this first screening study were that the percentage of duty cycle of ultrasound, irradiation time and radiation amplitude were not statistically influential factors within the ranges under study. However, the results showed better efficiencies with the maximum values of radiation amplitude, namely 50%, thus selected for further experiments. The working value of duty cycle was fixed at 35%, as no effect on the reaction efficiency was observed by changing this variable. Table 3. Tested ranges in the half-fraction type-V design and optimal values of the variables involved in the USAEH. Factor Half-fraction design Univariate design Optimum value Duty cycle (%) ????? 35 35 Radiation amplitude (%) ????? 50 50 [ ?-Glucuronidase] (U mL-1 urine) ??????? ??????? 500 Hydrolysis time (min) ????? ???? 30 Nuevas plataformas anal?ticas en metabol?mica 194 Figure 4. Kinetics of the conventional enzymatic protocol for hydrolysis of conjugated steroids using the urine pool. The effect of the enzyme concentration was studied by a univariate design within the range 400???? ? mL-1, after fixing the other variables at their optimum values. The irradiation time was fixed at 60 min. The results from this study showed that hydrolysis was complete with 500 U mL-1 of enzyme, without shortening of the hydrolysis time at higher values. Thus, 500 U mL-1 was fixed at optimum value. A kinetics study was made testing different irradiation times in order to determine the time necessary for total hydrolysis of the target compounds. One mL of sodium acetate buffer, 10 ?L of intern al standard (stock solution 1000 ng mL-1 in methanol) and the appropriate volume of enzyme solution (500 U mL-1 urine) were added to 10 mL of pooled urine. Each sample was irradiated under the optimum working conditions (50% radiation amplitude, 35% duty c ycle) at different sonication times within ???? min? The extraction efficiency was calculated as the ratio between the area obtained by the USAEH and that obtained by overnight incubation al 195 The Analyst, 134 (2009) 1416?1422 Chapter 4 37? C. As can be seen in Figure 5, the reaction was complete in 10 min for estriol, in 20 min for estrone and in 30 min for pregnenolone and estradiol. In view of these results, 30 min was selected and used for subsequents experiments as times longer than those required for their hydrolysis do not affect to the other steroids. Figure 5. Kinetics of the ultrasound-enhanced enzymatic protocol for analysis of conjugated steroids in the urine pool. The conclusion of this study is that ultrasound-assisted hydrolysis is complete at the optimum irradiation time as the reaction efficiency is similar, o even higher, for all the target analytes. Figure 6 corresponds to the chromatogram provided by analysis of the urine pool with the ultrasound- assisted protocol under optimized conditions. As can be seen, there is no modification caused by the assistance with auxiliary energy as compared to that obtained with the conventional protocol. Nuevas plataformas anal?ticas en metabol?mica 196 Figure 6. Total ion chromatogram provided by analysis of the urine pool after application of the ultrasound-assisted protocol under optimized conditions. 3.5. Complementary studies Additional experiments were performed in order to prove the effect of ultrasound over the enzymatic hydrolysis. Firstly, free estrogens were determined in the pooled urine, for which 10 mL of urine, 1 mL of sodium acetate buffer and 10 ?L of internal standard (10 ng) were mixed and directly subjected to the overall process without enzymatic hydrolysis. The results showed that only progesterone can be quantified in the pooled sample form as this progestagen is not conjugated. In addition, levels of estrone were detected below the quantification limit. One other experiment consisted of the application of ultrasonic irradiation under the optimum operation conditions in the absence of enzyme. The conclusion of this study was that no bond-breaking takes place in the absence of enzyme as only free progesterone was found. 3.6. Study of sample stability 197 The Analyst, 134 (2009) 1416?1422 Chapter 4 Urine analysis has proved a useful tool for the assessment of metabolism both for specific group of compound and in global terms. In most studies involving urine, the samples were stored at 4 ? C and analysed within the day of collection. For optimization studies, sample storage at ??? ? C was necessary, being a week the maximum storage time. These considerations make necessary a study of the storage conditions in order to ensure sample integrity and thus reliability of the results. With this aim, 250 mL of fresh female urine were stored at ??? ?C. Aliquots were daily analysed for a week. The results of this study were that the analytes are stable under the storage conditions, as shows Figure 7.A. In a second test, aliquots of urine were stored at 4 ? C and analysed every two hours within a day. Figure 7.B shows that variations in the concentration of analytes throughout a day were not found when stored at 4 ? C. 3.7. Application of the method The usefulness of the proposed method was demonstrated by application to female urine samples. With this aim, first morning urine was obtained from six female voluntaries. Ten mililiter aliquots were separately analysed in triplicate (results shown in Table 4). The results indicate that estrone and pregnenolone were the most abundant estrogens in all cases. Progesterone and estradiol were also found in most samples. Sample Estriol Estradiol Estrone Pregnenolone Progesterone 1 3.62 ? 0.15 1.58 ? 0.03 7.18 ? 0.18 5.23 ? 0.30 0.48 ? 0.02 2 4.47 ? 0.18 1.45 ? 0.05 3.92 ? 0 11 10.15 ? 0.15 0.59 ? 0.02 3 2.68 ?0.09 0.73 ? 0.01 4.66 ? 0.15 7.64 ? 0.32 0.52 ?0.05 4 0.89 ? 0.05 0.31 ? 0.01 2.66 ? 0.08 2.62 ? 0.10 0.45 ? 0.05 5 8.95 ? 0.12 1.16 ? 0.02 5.98 ? 0.06 4.22 ? 0.20 0.17 ? 0.01 6 2.07 ? 0.12 0.55 ? 0.03 4.12 ? 0.09 4.20 ? 0.08 0.60 ? 0.04 Table 4. Concentration(ng mL-1) of female hormones in urine. Nuevas plataformas anal?ticas en metabol?mica 198 In order to evaluate the precision of the proposed method, intra-day and inter-day variability were evaluated in a single experimental set-up with duplicates.25 The experiments were carried out using 10 mL of female urine buffered with 1 mL of sodium acetate (0.15 M, pH 5), with internal standard at 10 ng mL-1, and under the optimum working conditions. Two mesurements per day were performed on seven days. The results obtained, Figure 7. Study of urine stability for analysis of conjugated steroids. (A) Daily ??????????????????????????????????????????????????????????????????????????????????? samples stored at 4 ?C within a day. 199 The Analyst, 134 (2009) 1416?1422 Chapter 4 expressed as relative standard deviation (RSD), ranged from 2.15 to 5.16 % for intra-day variability and from 4.90 to 8.01 % for inter -day variability. 4. Conclusions An ultrasound-assisted enzymatic hydrolysis method has been proposed for deconjugation of estrogens in human urine samples. A drastic shortening of the time required for complete reaction is achieved as compared with traditional methods based on enzymatic incubation, which usually require 12?18 h. Extraction of free estrogens from urine is carried out automatically by S?E, while quantification is performed by LC?MS/MS with multiple reaction monitoring. After a multivariate optimization of the variables involved in the USAEH, a kinetics study showed that the hydrolysis reaction of all conjugated steroids is complete in only 30 min. The study included stability experiments of urine stored at ? ?C and ??? ?C? Since estrogens and progestogens are involved in the development of some diseases such as breast cancer, the profile of these compounds can be used to assess the metabolism of this compounds. For this reason, rapidity, sensitivity and selectivity are desirable properties for determination of these compounds in biological fluids, which makes the proposed method suitable for clinical analysis. 5. Acknowledgments The Spanish Ministerio de Ciencia e Innovaci?n (MICINN) is acknowledged for financial support through project CTQ2009-07430. F.P.C. is grateful to MICINN for a Ram?n y Cajal contract (RYC -2009-03921). B.A.S. Nuevas plataformas anal?ticas en metabol?mica 200 is also grateful to Ministerio de Ciencia y Tecnolog?a for an FPI scholarship (BES -2007-15043). 6. References 1 G. E Ackerman, B. R. Carr, Rev. Endocr. Metab. Dis., 2002, 3, 225?230. S. C. Sala, V. Martineti, A. M. Carossino, M. L. Brandi, Expert Rev. Endocrinol. Metabolism 2007, 2, 503?516. 2 R. C. Tuckey, Placenta, 2005 , 26 , 273?281. 3 J. W. Simpkins, S. -H. Yang, Y. Wen, M. Singh, Cell. Mol. Life Sci., 2005, 62, ???????? 4 L. S. Goodman, A. Gilman, The pharmaceutical basis of therapeutics, Macmillan, ????, ?????????? 5 S. Veiga, L. M. Garc?a-Segura, I. Azcoitia, J. Neurobiol., 2003, 56, 398-406. 6 J. I. Taylor, P. B. Grace, S. A. Bingham, Anal. Biochem., 2005, 341, 2??????? 7 R. Raftogianis, C. Creveling, R. Weinshilboum, Weisz, J. Natl. Cancer I. Monographs, 2000, 27, 113-124. 8 X. Xu, T. D. Veenstra, S. D. Fox, J. M. Roman, H. J. Issaq, R. Falk, J. E. Saavedra, L. K. Keefer, R. G. Ziegler, Anal. Chem., 2005, 77, ???????54. 9 D. Arroyo, M. C. Ortiz, L. A. Sarabia, J. Chromatogr. A, 2007, 1157, 358? 368. 10 X. Xu, J. M. Roman, H. J. Issak, L. K. Keefer, T. D. Veenstra, R. G. Ziegler, Anal. Chem., 2007, 79 , ?????????? 11 H. Adlercreutz, P. Kiuru, S. Rasku, K. W?h?l?, T. Fotsis, J. Steroid Biochem. Mol. Biol., 2004, 92 , 399?411. 12 L. Mao, C. Sun, H. Zhang, Y. Li, D. Wu, Anal. Chim. Acta, 2004, 522, ???????? 201 The Analyst, 134 (2009) 1416?1422 Chapter 4 13 J. L. Capelo, P. Xim?nez -Emb?n, Y. Madrid -Albarr?n, C. C?mara, Anal. Chem. 2004, 76, ???????? 14 C. Pe?a-Farfal, A. Moreda-Pi?eiro, P. Bermejo -Barrera, H. Pinochet - Cancino, I. de Gregori-Henr?quez, Anal. Chim. Acta, 2005, 548 , 183-191. 15 E. Sanz, R. Mu?oz-Olivas, C. C?mara, J. Chromatogr. A, 2005, 1097, ???? 16 V. G. Mihuez, E. T?tar, I. Vir?g, C. Zang, Y. Jao, G. Z?ray, Food Chem., 2007, 105 , ?????????? 17 G. Vale, S. Pereira, A. Mota, L. Fonseca, J.L. Capelo, Talanta, 2007, 74, ???????? 18 G. Vale, R. Rial-Otero, A. Mota, L. Fonseca, J. L. Capelo, Talanta, 2008, 75, ???????? 19 M. Siwek, A. B. Noubar, J. Bergmann, B. Niemeyer, B. Galunsky, Anal. Bioanal. Chem., 384 , 2006, 244?249. 20 M. M?guez-Framil, A. Moreda-Pi?eiro, P. Bermejo -Barrera, P. L?pez, M. J. Tabernero, A. M. Bermejo, Ana. Chem, 2007, 79 , 8564?8570. 21 D. L. Massart, B. G. M. Vadeginste, L. M. C. Buydens, S. de Jong, P. J. Lewi, J. Smeyers-Verbeke, Handbook of Chemometrics and Qualimetrics, Part A, Elsevier, Amsterdam 1997, 383?390. CHAPTER 5 : Targeted analysis of sphingoid precursors in human biofluids by solid phase extraction with in situ derivatization prior to ? - LC?LIF determinatio n Targeted analysis of sphingoid precursors in human biofluids by solid phase extraction with in situ derivatization prior to ?-?????? determination B. ?lvarez-S?nchez, F. Priego-Capote, M.D. Luque de Castro Department of Analytical Chemistry, University of C?rdoba, Annex C-3 Building, Campus of Rabanales, E-14071 C?rdoba, Spain; and, Institute of Biomedical Research Maim?nides (IMIBIC), Reina Sof?a Hospital, University of C?rdoba, E- 14071 C?rdoba, Spain Analytical and Bioanalytical Chemistry, 400 (2011 ) 757 ? 7 6 5 207 Anal. Bioanal. Chem,. 400 (2011) ??????? Chapter 5 Targeted analysis of sphingoid precursors in human biofluids by solid phase extraction with in situ derivatization prior to ?-?????? determination B. ?lvarez-S?nchez, F. Priego-Capote, M.D. Luque de Castro Abstract A method for determination of two relevant sphingoid precursors such as sphingosine and sphinganine and the corresponding conjugates sphingosine 1-phosphate and sphinganine 1-phosphate in human urine and serum is here presented. The method is characterized by a solid phase extraction step with in situ derivatization of the sphingolipids in the eluate (o-phtaldialdehyde derivatives) to obtain fluorescent compounds. In this way, sample preparation was completely performed in a single automated step by means of a lab-on-valve system. Derivatized analytes were injected into a liquid chromatography system operating at micro regime and detected by laser induced fluorescence. For determination of sphingoid phosphates, they were enzymatically converted to free sphingoids to obtain stable fluorescent derivatives. The detection limits were in the range 4.2? 10.2 ng mL -1 for serum and 0.56 ?1.36 ng mL -1 for urine, with repeatablility ranging from 3.9 to 6.2 % expressed as relative standard deviation. The method was validated by direct infusion tandem mass spectrometry in multiple reaction monitoring (MRM) to compare results provided by analysis of biofluids and to confirm the identity of the target compounds. Sensitivity and precision were better than or similar to those provided by the confirmatory method. The automation of sample preparation enables to scale-down this step and improves precision by minimization of human intervention, being thus suitable for clinical analysis. 208 Nuevas plataformas anal?ticas en metabol?mica 1. Introduction Sphingolipids (SLs) are a family of lipidic compounds endowed with a long amino-alcohol chain, known as sphingoid base. Sphingolipids are components of biological membranes, being widely distributed in certain tissues, such as brain and nerve tissue. Sphingolipids exert different functions in signal transmission and cell recognition. Sphingoid precursors ?s phinganine (Sa) and sphingosine (So)? act as structural and signalling molecules in membranes [1]. Sphingosine 1 -phosphate, commonly present in human plasma and platelets, is an intermediate of sphingolipids catabolism [2], which inhibits the reproduction of certain tumoral cells, and also plays a key role both in the mobilization of int racellular calcium [1,3] and in cellular growth, differentiation, senescence and apoptosis ?????. Sphingolipids metabolism involves a number of synthesis and degradation pathways, as expected from the variety of sphingolipids present in cells. Sphingoids are present in the organism by degradation of complex sphingolipids (gangliosides, ceramides, etc.) or by de novo synthesis [7]. The latter starts with the formation of sphinganine from palmitoyl CoA and serine. Sphinganine is then acylated to form dihydroceramide, which is desaturated to form ceramide. Ceramide can undergo sereval pathways, one of them being a breakdown to generate sphingosine. The sphingoid phosphate metabolites, sphingosine 1-phosphate and sphinganine 1- phosphate are formed by phosphorilation via two kinases, sphingosine kinase types 1 and 2 [8]. The main group of diseases associated to sphingolipid metabolism are sphingolipidoses [9], which entail alterations in the catabolism of these compounds as a consequence of genetic disorders, leading to the storage of free or glucide-bonded sphingolipids inside cells. Sphingoids are also involved in diseases related with apoptosis and cellular signalling, such as diabetes, neuropathies, Alzheimer's disease, Parkinson's 209 Anal. Bioanal. Chem,. 400 (2011) ??????? Chapter 5 disease, atherosclerosis or cancer [10]. Therefore, the role of sphingoid precursors as potential metabolites to correlate their concentration levels with the diagnosis or prognosis of these diseases should be emphasized. The detection of sphingoid precursors as free or conjugated fo rms in plasma makes this biofluid specially suited for quantitative analysis. However, the non-invasiveness of urine sampling and the presence of sphingoid precursors in this biofluid justifies its alternative or complementary use for sphingoids analysis. Few methods have been reported for simultaneous determination of sphingoid bases and phosphate derivatives. The most consolidated approach involves a derivatization step with o-phtaldialdehyde (OPA) to form isoindol derivatives prior to chromatographic separation with fluorescent detection ??,??????? ?ith this strategy, sensitive and selective methodologies can be developed. However, phosphate OPA derivatives are not kinetically stable, which involves a significant source of irreproducibility. Caligan et al. used EDTA buffer as derivatization solvent to obtain stable phosphate OPA derivatives for 5 h if the samples are stored below 4 ?C [5]. However, the derivatives were not stable at room temperature, which is a considerable limitation. Additionally, the pH of the samples has to be modified by dilution with the EDTA buffer that could alter the integrity of the sample. Another way to proceed is to take benefit from enzymatic selectivity of alkaline phosphatase by a dual determination with and without enzyme hydrolysis. A less frequent approach for detection of sphingoid bases and phosphates is the use of radio-labeled substrates [15], which increases considerably the cost of the analysis? Recently, new methodologies based on LC?MS/MS detection have been developed [16 ?18], although mass analy zers are not always present in routine laboratories ?additionally, the structure of sphingoid precursors does not seem to be ideal for optimum fragmentation by tandem mass spectrometry. Due to the non-polar nature of sphingoids and their low concentration in biological fluids, sample preparation is usually the 210 Nuevas plataformas anal?ticas en metabol?mica bottleneck of sphingoids analysis. In fact, most of the methods developed so far have been based on liquid?liquid extraction, evaporation of the organic solvent and reconstitution in a suitable phase prior to LC separation ???????? The use of solid-pha e extraction (SPE) allows obtaining clean extracts with high concentration factors, leading to a considerable increase of sensitivity [19]. However, these methods are m anual and, in many cases, they require large reagent and sample consumption adding the complexity of a derivatization step when this is needed. For this reason, automation and miniaturization of sample preparation is desirable to avoid all limitations associated to long protocols. A promising manifold that allows attaining the benefits associated to automation and miniaturization is the lab-on-valve (LOV). This system is considered the third generation of Flow-Injection Analyzers (FIA), that clearly revolutionized the analysis automation. Thus, the applicability of LOV configurations in clinical analysis has been widely discussed [20]. A method for determination of sphingosine (So), sphinganine (Sa), sphingosine 1-phosphate (So-1P) and sphinganine 1-phosphate (Sa-1P) in human biofluids is here proposed. The low concentration of sphingoids in urine and serum and the complexity of both types of biofluids demand for the development of highly-selective and sensitive analysis methods. The proposed method enables to extract the target compounds with in situ derivatization during elution with OPA, generating the fluorescent derivatives in a fully-automated format. Separation of sphingoids, carried out by LC at micro scale is coupled to laser-induced fluorescence detection (LIF). 2. Experimental 2.1. Instruments and apparatus 211 Anal. Bioanal. Chem,. 400 (2011) ??????? Chapter 5 Figure 1 illustrates the experimental setup used for the development of this analytical method. The extraction and derivatization of the sphingolipids were automatically performed in a single step using a sequential injection (SI) lab -on-valve (LOV) microfluidic system. The LOV manifold consists of an 1-mL microsyringe pump, which allows aspirating and dispensing micro-volumes of sample and reagents; a 6 -port valve made of Plexiglass, in which the ports are interconnected and also with the syringe pump by a center port; a holding coil between the syringe pump and the 6 -port valve; and a 2 -position selection valve. Polyetheretherketone (PEEK ) tubings (0.5 mm x 0.8 mm inner diamete r) were used to connect the manifold components to the ports of the selection valve. Polytetrafluorethylene (PTFE) tube (1.5 mm inner diameter) from An?lisis V?nicos (C. Real, Spain) was used as minicolumn. Chromabond C18 (30 ?40? m particle diameter) from Macherey-Nagel (D?rem, Germany) was used as sorbent material in the SPE step. Other sorbent materials tested were Chromabond C18 Hydra, Bond elut NH2 (Varian, Palo Alto, USA), Chromabond HR -P and C8 Strata (Phenomenex, Torrance, California, USA). An external 10 -port 2-position selection valve (VICI, Valco, Houston, USA), Figure 1. Experimental workflow of the LOV system. 212 Nuevas plataformas anal?ticas en metabol?mica connected to the microcolumn channel, enables the collection of the eluate after the SPE sequence. The LOV and the external selection valves are fully automated and controlled by the FIAlab for Windows version 5.0 software. Separation of sphingolipids was performed by an Agilent (Palo Alto, CA, USA) 1100 micro liquid chromatograph ( ? -LC ) equipped with a binary capillary pump and an automatic injection valve (1 -?L sample loop) and a diode array detector (200 ?800 nm). The analytical column was a reversed - phase Zorbax SB -C18 (150 mm ? 0.5 mm I.D., 5?m) from Agilent. The overall system was mounted on capillary tubes 75 ?m i.d. ? 375 ?m O.D. from An?lisis V?nicos. After chomatographic separation, the derivatized analytes were detected by a Zetalif 2000 325 nm/CE laser -induced fluorescence (LIF) detector from Picometrics (Toulouse, France). The ? -LC was connected to the LIF detector by a 75 ?m i.d. ? 375 ?m capillary tube in which the detection window, of 5 mm length, was made for collecting the overall emitted light. The signal acquisition from the LIF detector is monitored and integrated by the Agilent Chemstation software. Validation of the proposed method was performed by direct infusion to an Agilent 6410 triple quadrupole mass analy zer. An Agilent 1200 Series system furnished with a binary pump, a vacuum degasser, and an autosampler were used for injecting the sample and pumping the carrier solution. The data were processed using a MassHunter Workstation Software for acquisition, qualitative and quantitative analysis. 2.2. Chemicals, reagents and working solutions D-Sphingosine (So), D-erythro-dihydrosphingosine (sphinganine, Sa), sphingosine 1-phosphate (So1P), D-erythro-dihydrosphingosine 1-phosphate (sphinganine 1-phosphate, Sa1P) and tocopherol acetate were purchased from Sigma Aldrich. Stock standard solutions at 1000 ?g mL -1 (So and tocopherol acetate), 500 ?g/mL (S a) and 200 ?g mL -1 (So1P and Sa1P), were 213 Anal. Bioanal. Chem,. 400 (2011) ??????? Chapter 5 prepared in 25:75 cloroform ?acetonitrile. Multistandard solutions were prepared by diluting the stocks in a ????? acetonitrile?water mixture? All the above solutions were stored at ??? ?C in glass vials and kept in the dark until use. ?eioni?ed water (?? M??cm) from a Millipore Milli-Q water purification system (Millipore, Bedford, MA, USA) was used to prepare all aqueous solutions? Acetonitrile of LC?MS grade from ?anreac (?arcelona, Spain) and cloroform from Scharlau (Barcelona, Spain) were used for standard stock solutions. The car rier solution for preconditioning the cartridge in the solid phase extraction was deionized water acidified with phosphoric acid (pH=3). Acetonitrile of LC grade (Panreac) and deionized water were used for chromatographic mobile phases. Ammonium formate was used as additive in the carrier solution to enhance ionization required for mass spectrometry detection. Free sphingoids were eluted and derivatized with 250 ?g mL -1 of o- phthaldialdehyde from Sigma?Aldrich in 0.1% 5 -mercaptoethanol (Merck, Darmstadt, Germany) and ????? acetonitrile??? mM boric acid (p? ????)? This reagent was daily prepared and stored at 4 ?C until use due to its instability. For the optimization of the chromatographic separation and LIF detection, a multistandard solution at 100 ng L -1 was prepared and stored at 4 ?C in the dark until use. Alkaline phosphatase from bovine intestinal mucosa (Sigma?Aldrich) was prepared in buffer Tris-HCl 200 mM pH 7.4 (400 U mL -1). The reaction buffer for the enzymatic dephosphorilation was ??? mM Tris??Cl (pH 7.4), 75 mM MgSO 4 in 2 M glycine buffer (pH 9). 2. 3. Sample collection and preparation of spiked samples Urine samples were obtained from volunteers, collected in sterile containers, and stored at ??? ?C in the dark until analysis? For optimi?ation purposes, 100 mL of sample from five volunteers was pooled and divided 214 Nuevas plataformas anal?ticas en metabol?mica into 1.5 -mL aliquots. 0.5 -mL sodium phosphate 50 mM (pH =7) was added to each aliquot and all them were stored at ??? ?C? Venous blood was collected into a Vacutainer? tube (Becton Dickinson). The tube was placed in ice or kept refrigerated until processed. Blood samples were centrifuged at 4000 ? g for 10 min. All steps from blood extraction to analysis were performed in compliance with the guidelines dictated by the World Medical Association Declaration of Helsinki (2004), which were supervised by the ethical committee of Reina Sofia Hospital (C?rdoba, Spain) that approved the experiments. Individuals selected for this study were informed to obtain consent prior to this research. The resulting serum was divided into 1-mL aliquots in sterile containers and stored at ??? ?C in the dark? ???-?L serum aliquots were diluted to a final volume of 2 mL with 10 mM phosphate buffer (pH= 7) in order to prevent protein precipitation, being thus ready for the SPE step. Samples were spiked with 50 ng mL -1 tocopherol acetate as external standard (final amount injected into the chromatograph 1 ?L). After vortexing for 10 min, the samples were subjected to the overall method. Analyses were carried out within 24 h of sample preparation. One aliquot of the sample was directly analyzed whereas a second one was used to hydrolyze phosphate derivatives. In the latter, sphingolipids So1P and Sa1P were hydrolyzed by means of an alkaline phosphatase in order to re lease free sphingolipids, So and Sa, respectively. With this aim, the sample aliquot was mixed with 200 ?L of reaction buffer and 125 -?L alkaline phosphatase (50 U) and incubated for 30 min at 37 ?C under agitation. 2.4 . Solid- phase extraction and deriva tization step Before chromatographic separation, derivatization of the target compounds was performed to obtain fluorescent products suitable to be 215 Anal. Bioanal. Chem,. 400 (2011) ??????? Chapter 5 measured by LIF. In this method, derivatization was carried out in situ with elution from the sorbent cartridge by means of a LOV automatic manifold, in a sequence of programmed steps that can be summarized as follows: (a) activation of the mini-column (prepared by packing 0.05 g of C18 -hydra in 80 mm PTFE tubing) with 2 -mL 90:10 acetonitrile ?water solution and conditioning with 1-mL carrier (both steps at 3 mL min-1); (b) equilibration of the column with 1-mL carrier solution at 1.2 mL min-1; (c) loading of sample (2 mL diluted urine and serum) at 1.2 mL min-1; (d) washing of the mini-column with 0.5 -mL carrier and flushing air to eliminate the rest of sample matrix; (e) 400 ?L of eluant (90:10 acetonitrile ?water mixture in which the OPA reagent was dissolved) was passed through the mini-column and discarded to prevent dilution of the analytes; (f) switching t he external selection valve and (g) elution with 200 ?L of eluant at 5 ?L min -1. 2.5. Chromatographic separation and detection The extract was loaded into the microinjection valve by filling of the sample loop (total volume, 1 ?L). Separation was perf ormed by ? -LC. Mobile phase (90:10 acetonitrile/water) was maintained at 10 ?L min -1 isocratically during the chromatographic run (9 min). The excitation wavelength for LIF detection was 360 nm. The external standard was on -line monitored by diode array detection and quantified at 360 nm. 2.6. Mass spectrometry analysis Validation of the proposed method was carried out by direct injection ?tandem mass spectrometry (di-MS/MS) with electrospray ionization (ESI) in positive mode. 5 -?L aliquots were injecte d and pumped to the triple quadrupole mass detector by means of a 90:10 acetonitrile ?water 216 Nuevas plataformas anal?ticas en metabol?mica 2 mM ammonium acetate and 0.1% formic acid carrier solution, which was pumped at 1 mL min-1. The electrospray ionization source was operated at a capillary voltage of 4000 V. The working conditions of the mass spectrometer were: nebulizer pressure 35 psi and drying gas (N 2) conditions at 10 mL min -1 flow-rate and 300 ?C temperature. The mass detector was operated in multiple reaction monitoring (MRM) mode. Table 1 summarizes the working conditions for the di -MS/MS mehod. Precursor ions corresponded to [M+H] + ions for sphingosine and sphinganine metabolites, which were preferentially fragmented to generate ?M???(?2O)] + by dehydratation and ?M???(?2O)?(C?2OH)] + by fragmentation of the alcoholic group. Table 1. MRM parameters for sphingosine, sphinganine and the IS (tocopherol acetate). A calibration curve was built by injecting pure standards at different concentrations ranging from limits of quantification to 8000 ng mL -1. For the analysis of serum and urine samples, 5 -?L of eluate from the solid -phase extraction step was injected under the optimum working conditions, in which the analytes were eluted with 200 ?L of 90:10 acetonitrile ?water. Compound Molecular weight MRM transitions Fragmentor Voltage (V) Collision energy (eV) Dwell time (ms) Sphingosine (So) 299.5 300.3 ? 282. 3 300.3 ? 252.3 100 100 7 7 20 Sphinganine (Sa) 301.5 302.3 ? 284.3 302.3 ? 254.3 120 120 10 10 20 Tocopherol acetate (IS) 472.7 490.5 ? 207.1 100 18 20 217 Anal. Bioanal. Chem,. 400 (2011) ??????? Chapter 5 3. Results and discussion 3.1. Optimization of the derivatization ?extraction step Taking into account the particularity of the process and the involvement of discontinuous variables, a univariant strategy was selected to study the sample preparation process. The variables under study, the tested range and the optimum value are shown in Table 2. For this study, human biofluids were spiked with the analytes at 10 ?g mL -1 each (50 ?g mL - 1 external standard). In addition, non-spiked samples were analysed under the resulting optimum operating conditions to evaluate accuracy of the method. A strategy based on direct analysis of biofluids was selected for this research, which demanded for two different dilution factors according to the sample matrix. Thus, an 1:10 dilution factor was needed for serum analysis owing to the high concentration of proteins, which were not previously precipitated. On the other hand, urine was practically analysed without dilution except for the addition of buffer solution to adjust the pH. The optimization study was divided into two parts: first, those variables involved in the SPE step were initially optimized with subsequent off-line derivatization and LC?LIF analysis; then, the variables affecting derivatization were studied with the optimum SPE variables. Thus, the first study included the optimization of the sorbent material, sample loading rate, pH of the sample, volume of discharged eluate, elution volume and composition of eluant. The first study aimed at obtaining the most suitable sorbent for the SPE. Six sorbents were tested taking into accoun t the chemical properties of the analytes under study, namely: C18 (45 ?m particle size), a non -encapped octadecyl modified silica phase; C18 Hydra (45 ?m), a non -endcapped 218 Nuevas plataformas anal?ticas en metabol?mica octadecyl phase for relatively low polar analytes; C8 strata (45 ?m), an octyl - phase based on non-specific retention of non-polar compounds with the surface endcapped to minimize secondary polar silanol interactions; aminopropyl NH 2 BondElut (40 ?m ?120 ?m particle size), a very polar sorbent based on both hydrogen bonding and anion exchange interactions; HR -P (50 ?100 ?m) highly porous resin based on polystyrene - divinylbenzene (PS/DVB); and Chromabond SB (45 ?m), strongly basic anion exchanger. Table 2. Optimization of the SPE step with in situ derivatization of sphingolipids The home-made cartridges were prepared and fixed by packing 0.05 g of sorbent material in a 80 mm PTFE tubing, which perfectly suited to the lab-on-valve system without overpressure. Due to the lipidic character of the target analytes and their small molecular size, the best results regarding retention capability were obtained with C18, particularly for urine because of the high efficiency to remove salts and polar interferences. Another key variable was the pH of the sam ple. To test its influence, the pH of the sample was adjusted with different buffer solutions, namely: 100 mM ammonium acetate (pH =5), 100 mM potassium phosphate (pH=7), 200 mM sodium Variable Tested range Optimum value Sample loading rate (?L min - 1) 0 ? 60 5 pH of sample 2 ? 10 7 Volume of eluate discharged (?L) ?????? 400 Elution volume (?L) 50 ? 1000 200 Concentration of OPA (mg mL-1) 0.05 ? 1 0.125 Composition of eluant (% acetonitrile) 50 ? 100 90 Sorbent C18, C18 Hydra, C8 strata, DMA, NH2 BondElut, HR -P, Chromabond SB C18 219 Anal. Bioanal. Chem,. 400 (2011) ??????? Chapter 5 borate (pH= 9), 1 M sodium carbonate (pH= 10), being pH 7 the optimum value. After this study, the concentration of OPA was optimized together with the volume of discharged eluate, the elution volume and the composition of eluant to check if they influenced the derivatization of sphingoids. The optimum values for each variable are included in Table 2. The volume and the number of the eluate fraction were tested by off-line analysis of 200 -?L fractions together with the non -retained material after sample loading. Figure 2 illustrates the elution profile resulting from the analysis of the different fractions. As can be seen, the maximum recovery was obtained in the third fraction with isolation of approximately 85% of the initial amount spiked. These high recovery values can be ascribed to the in situ derivatization step that favors a uniform elution profile as concluded from this test. The target compounds were not detected in the non-retained material after the retention step. Therefore, the first 400 ?L were discharged to avoid dilution of the target analytes. It is worth emphasizing that similar recoveries were obtained for serum and urine. These results were combined with the analysis of non-spiked samples to verify the recoveries obtained with spiked biofluids, which assesses accuracy of the method (85%). These results also demonstrate the absence of matrix effects with both types of biofluids and, for this reason, calibration plots can be built with standard solutions. Tocopherol acetate was selected as external standard due to its similar non-polarity to the target metabolites with an aliphatic chain in its structure. Additionally, this compound is not physiologically present in humans. 3.2. Optimization of the laser - induced fluorescence detection Concerning the laser device, the excitation wavelength was set at 325 nm, close to the optimum excitation wavelength for the OPA derivatized 220 Nuevas plataformas anal?ticas en metabol?mica sphingolipids, which is ??? nm ?????? On the other hand, the whole fluorescence emitted from the flow-cell was collected. The LIF instrument works in three different detection ranges, nam ely 0 ?2, 0 ?20 and 0 ?200 relative fluorescence units (RFU). The lower RFU range was selected for analysis to avoid saturation of the detection system. Figure 3 shows the chromatograms obtained with two serum aliquots, without (A) and with (B) hydrolysis catalized by alkaline phosphatase. (B) (A) (C) Figure 2. Optimization of the elution volume in the SPE step. Chromatograms obtained for the elution of the first (A) second (B) and third (C) fraction. Tim e (m in ) Tim e (m in ) Tim e (m in ) 221 Anal. Bioanal. Chem,. 400 (2011) ??????? Chapter 5 3.3. Features of the method As a dual analysis with an enzymatic hydrolysis step is required for complete analysis of the target sphingolipids, the sphingoid phosphates So1P and Sa1P were quantified by the difference between concentrations corresponding to So and Sa after and before hydrolysis. Thanks to the absence of matrix effects calibration plots were run by using the standard Figure 3. Chromatograms obtained with two 50 ng mL-1 spiked serum aliquots, without (A) and with (B) phosphatase alkaline hydrolysis. (A) (B) 222 Nuevas plataformas anal?ticas en metabol?mica peak area ?external standard peak area rat io as a function of the standard concentration. Calibration was performed with multistandard solutions at ten concentration levels ?within 5 and 10000 ng mL -1? which were analyzed in triplicate. The regression coefficients and the linear range are shown in Table 3. The sensitivity of the method was evaluated for each studied biofluid according to its dilution factor. In any case, the limits of detection (LOD) and quantification (LOQ) were expressed as the mass of analyte which gives a signal that is 3? and 10 ?, respectively, above the mean blank signal (where ? is the standard deviation of the blank signal). Thus, LODs ranged between 4.2?10.2 ng mL -1 (or pg on-column) in the case of human serum, while LOQs ranged between 13.9 and 34.1 ng mL -1 (or pg). As there was no concentration factor with serum, LODs and LOQs practically coincided with those obtained with standard solutions. In the case of urine, there was a concentration factor of 7.5 that influenced positively the sensitivity of the analysis. Therefore, LODs ranged between 0.56 ?1.36 ng mL-1 for human urine, while LOQs ranged between 1.85 and 4.55 ng mL -1. Precision was assessed by analysis of five aliquots of urine and serum samples spiked at 100 ng mL -1 in a single experimental set. As can be seen in Table 3, the results obtained, expressed as relative standard deviation (RSD), were within 3.9 and 6.2%, which are quite acceptable for clinical analysis. It is worth emphasizing that similar precision levels were found for both biofluids. 3.4. Applicat ion of the method to human biofluids The proposed method was tested by application to the two most frequent biofluids used in clinical analysis: serum and urine. Preparation of serum samples is limited by the presence of proteins while the high concentration of salts is the main obstacle for urine analysis. Therefore, the cleanup effect associated to the preparation approach is critical to succeed 223 Anal. Bioanal. Chem,. 400 (2011) ??????? Chapter 5 in the quality of the results. An experimental strategy was designed for analysis of two sets of three independent replicates with and without alkaline phosphatase treatment for each biofluid (n=3). Table 4 shows the mean value and the standard deviation obtained for the target compounds in each sample. Table 3. Features of the proposed method. Analyte Linear regression LOD serum LOQ serum LOD urine LOQ urine Linear dynamic range Precision (R2) (ng mL-1) (ng mL-1) (ng mL-1) (ng mL-1) (ng mL-1) (RSD, % ) So 0.9935 5.2 17.4 0.69 2.32 LOQ????? 3.9 Sa 0.9987 4.2 13.9 0.56 1.85 LOQ????? 4.5 Sa1P 0.9979 9.1 30.5 1.21 4.07 LOQ????? 6.2 So1P 0.9898 10.2 34.1 1.36 4.55 LOQ????? 5.1 The sphingoid phosphate So1P was the most abundant metabolite in serum, which can be justified because the conjugate is the suited form to be transported in serum and also because serum does not contain the enzymes for So1P degradation [21]. Concentrations reported in the literature for sphingoid precursors and phosphate derivatives in serum are characterized by a great variability, as can be seen in Table ? ???,??????? Thus, levels detected in this research were within the range reported by previous studies. Lower values reported in some cases are below LOQs determined with the proposed method. However, LOQs could be considerably decreased by implementation of an additional preconcentration step by evaporation since the injection volume into the chromatograph was only 1 ? L while the elution volume was 200 ? L. In addition, So was the only sphingoid quantified in urine, while Sa was detected but at concentration below the LOQ. To complete this analysis, the amount of the sphingoid phosphates in urine was 224 Nuevas plataformas anal?ticas en metabol?mica negligible (below the LOD). According to literature, this fact can be explained by the catabolic reaction pathway of So-1P based on its degradation in the liver by the action of a lyase, giving ethanolamine phosphate and a fatty aldehyde as final products for secretion in urine [21,26,27]. * belo w t he LOQ , - belo w t he LOD 3.5. Validation of the repor ted method by comparison with the di - MS/MS method The results obtained by the method reported in this research were validated with those from the analysis by di-MS/MS. With this aim, ten Biofluid Sphingosine (ng mL-1) Sphinganine (ng mL-1) Serum Found Reported Found Reported 145.0 (11.3) ????? [23,24] 36.8 (11.0) ????? [22,24] Urine 16.8 (2.6) ????????? ???? * 0.006 ???? [23] Biofluid Sphingosine 1-P (ng mL-1) Sphinganine 1-P (ng mL-1) Serum Found Reported Found Reported 184.7 (29.1) ?????? [12,24,25] 143.5 (4.5) ???? [24] Urine - - - - Table 4. Mean concentration of sphingoid precursors found in human serum and urine from volunteers (mean, n=3, SD in brackets) and range of concentration reportedin bibliography. 225 Anal. Bioanal. Chem,. 400 (2011) ??????? Chapter 5 calibration solutions were prepared and analyzed in triplicate with the method based on mass spectrometry detection. The LOD and LOQ, expressed as the mass of analyte which gives a signal that is 3??and 10 ? above the mean blank signal (where ? is the standard deviation of the blank signal), ranged between 2.6 and 6.1 and 8.6 and 20.3 pg injected amount, respectively, which are similar to those obtained with the proposed method. Three replicates of urine and serum aliquots spiked at 100 ng mL -1 were analyzed by both the proposed method and the di-MS/MS. Differences between the results obtained by both methods were not significant as checked by application of the statiscal t -value test (P=0.05), which enables to validate the reported approach. 4. Conclusions The proposed method allows determining sphingoid precursors in two human biofluids: urine and serum. For this purpose, a simple and automated sample preparation protocol based on SPE has been developed. In this step, the elution and derivatization of the analytes are simultaneously developed in an LOV manifold, thus shortening the overall time of analysis. The LOV manifold is automated and fully computer-controlled, thus avoiding contact of the operator with biological samples, which is crucial in clinical analysis. The developed LOV application proves the utility of this manifold in clinical analysis by following previous generation of injection analy zers. ?erivati?ed sphingolipids were injected to a ??LC, which had been hyphenated to an LIF detector. This allows management of very low volumes of sample and organic solvents with a high sensitivity, as required by biological samples. The simplicity, sensitivity and precision of the developed method are key characteristics for determination of these compounds in urine and 226 Nuevas plataformas anal?ticas en metabol?mica serum samples. This application can be of interest to correlate the levels of sphingoid precursors with diseases related with the sphingolipid metabolism. Validation has been carried out by tandem mass spectrometry with direct infusion of treated samples. 5. Acknowledgements The Spanish Ministerio de Educaci?n y Ciencia is gratefully acknowledged for financial support (project no. CTQ 2006 -01614). B. ?lvarez -S?nchez is also grateful to the Ministerio de Ciencia y Tecnolog?a for an FPI scholarship. 6. References [1] Gosh TK, Bian J, Gill DL (1990) Scie nce 248: 1653. [2] Spiegel S, Milstien S (2002) J Biol Chem 277: 25851. [3] Oliviera A, Spiegel S, (1993) Nature 365: 557. [4] A G?mez -Mu?oz, DW Waggoner, L OB?rien, DN Brindley, (1990) J Biol Chem. 265: 21309. [5] TB Caligan, K Peters, J Ou, E Wang (2000) Anal Biochem 281: 36. [6] YA Hannun, LM Obeid (2008) Nat Rev Mol Cell Bio 9: 139 . [7] AH Merrill, MC Sullards, E Wang, KA Voss, RT Riley, (2001) Environ Health Persp 109: S2. [8] CE Chalfant, S Spiegel (2009) J Lipid Res 50: S91. 227 Anal. Bioanal. Chem,. 400 (2011) ??????? Chapter 5 [9] JEM Groener, BJHM Poo rthuis, MTJ. Helmond, CEM Hollak, JMFG Aerts, (2007) Clin Chem 53: 742. [10] B Ogretmen, YA Hannun (2004) Nat Rev Cancer 4: 604 . [11] JJ Butter, RP Koopmans, MC Miche, (2005) J. Chromatogr B 824: 65. [12] P Andr?ani, MH.Gr?ler, (2006) Anal Biochem 358: 23 9. [13] AH Merrill Jr, E Wang, RE Mullins, WCL Jamison, S Nimkar, DC Liotta, (1988) Anal Biochem 171: 373 . [14] M Solfrizzo, J. Avantaggiato, A. Visconti, (1997) J. Chromatogr B 692: 87. [15] S Aoki, Y Yatomi, M Otha, M Osada, F Kazama, K Satoh, K Nakahar a, Y Ozaki, (2005) J Biochem 138: 47. [16] J Bielawski, ZM Szulc, YA Hannun, A Bielawska, (2006) Methods 39: 82. [17] M Scherer, K Leuth?user -Jaschinski, J Ecker, G Schmitz, G Liebisch, (2010) J Lipid Res 51: 2001. [18] CA Haynes, JC Allegood, H Park, MC Sullards, (2009) J Chromatogr B 877: 2696. [19] J Bodennec, C Famy, G Brichon, G Zwingelstein, J Portoukalian, (2000) Anal Biochem 279: 245. [20] MD Luque de Castro, J Ruiz -Jim?nez, JA P?rez-Serradilla, (2008) Trends Anal Chem 27: 118. [21] Y Yatomi, Y Iga rashi, L Yang, N Hisano, R Qi, N Asazuma, K Satoh, Y Ozaki, S. Kume,(1997) J Biochem 121: 969. [22] A Laganhl, A Marino, G Fago, C Terregino, B Pardo -Mart?nez, (1994) Chromatographia 39: 85. [23] S Ribar, M Mesaric, M Bauman, (2001) J Chromatogr B 754: 51 1. [24] H Schmidt, R Schmidt, G Geisslinger, (2006) Prostag Oth Lipid Med 81: 162. 228 Nuevas plataformas anal?ticas en metabol?mica [25] X He, C -L Huang, EH Schuchman, (2009) J Chromatogr B 877: 983. [26] Y Yatomi, Y Ozaki, T Ohmori, Y Igarashi, (2001) Prostag Oth Lipid M. 64: 107 . [27] S Spiegel, S M ilstien, (1995) J Membr Biol 146: 225. Chapter 6 : Automated determination of folate catabolites in human biofluids (urine, breast milk and serum) by on -line SPE ?HILIC ? MS /MS Automated determination of folate catabolites in human biofluids (urine, breast milk and serum) by on-line SPE?HILIC?MS/MS B. ?lvarez-S?ncheza,b, F. Priego-Capote*a,b, J. M. Mata-Granadosa,b,c and M. D. Luque de Castroa,b aDepartment of Analytical Chemistry, University of C?rdoba, Annex C-3 Building, Campus of Rabanales, E-14071 C?rdoba, Spain; bInstitute of Biomedical Research Maim?nides (IMIBIC), Reina Sof?a Hospital, University of C?rdoba, E-14071, C?rdoba, Spain cSanyres I+D+i Department, Sanyres group, C?rdoba, Spain Journal of Chromatography A, 1217 (2010) 4688?4695 233 Chapter 6 J. Chromatogr. A, 1217 (2010) 4688?4695 Automated determination of folate catabolites in human biofluids (urine, breast milk and serum) by on - line SPE ?HILIC ?MS/MS B. ?lvarez-S?nchez, F. Priego-Capote , J. M. Mata-Granados , M.D. Luque de Castro Abstract A rapid, precise and fully-automated method for analysis of folate (vitamin B9) and its catabolites (viz. p-aminobenzoylglutamate, and its acetamide derivative) in biofluids is here presented. The method is based on on-line hyphenation of solid-phase extraction (SPE) by a Prospekt-2 system with hydrophilic interaction liquid chromatography?tandem mass spectrometry (HILIC?MS/ MS). The method was analytically characterized by estimation of repeatability (RSD, n=5, between 0.5 and 4.1%), accuracy (between 96 and 105%), and sensitivity (limits of quantification between 0.3 and 8.3 ng/mL (1.1 and 18.8 pmol/mL) or 0.03 and 0.83 ng (0.11 and 1.88 pmol). The proposed method is suited for routine analysis of folate catabolites as biomarkers to monitor deficiency of vitamin B9. 234 Nuevas plataformas anal?ticas en metabol?mica 1. Introduction Folic acid (vitamin B9) is a water-soluble vitamin involved in a broad variety of biological processes, as long as it acts as enzymatic cofactor in transference of methyl-group reactions [1]. Folic acid is responsible for the synthesis of DNA bases and chains; therefore, it is essential for the formation of new cells, especially during periods of rapid cell division and growth, such as infancy and pregnancy [2]. Folic acid is also involved in the synthesis of the haeme group of haemoglobin, being required for growth and maturation of red blood cells (RBCs) [3]. Accordingly, severe deficiency of folate leads to numerous diseases associated to hindered cell division processes and deficiency of RBCs, such as megaloblastic anemia, bone marrow, or fetal diseases (spina bifida, neural tube defects, etc.) [4]. On the other hand, folate takes part in the synthesis of amino acids serine and methionine. In this sense, there is also evidence of the relationship between folate deficiency and accumulation of homocystein, which is the substrate in the folate- mediated synthesis of methionine. This is currently considered an increased risk factor for cardiovascular disease, atherosclerosis and coronary heart disease [5,6]. Folic acid, mainly present as folate under physiologically normal conditions, is excreted in urine as more polar catabolites [7,8]. Catabolic transformation of folate involves reduction to its chemically unstable tetrahydro form, easily transformed to p-aminobenzoylglutamate (pABGA). The final product of the folate catabolism is the acetamide derivative of p- aminobenzoylglutamate (a-pABGA), the most abundant folate metabolite in urine and other biofluids, as it lacks of metabolic activity. The metabolic profile of folate catabolites in biological fluids can provide valuable clinical information about the abundance of this essential 235 Chapter 6 J. Chromatogr. A, 1217 (2010) 4688?4695 vitamin in humans. This can be of especial interest in individuals with increased risk of deficiency, such as neonates or pregnant women. In fact, levels of folate catabolites pABGA and a-pABGA have been found considerably increased in serum and urine during pregnancy [1], when the metabolism of this vitamin is enhanced. Determination of folic acid has been classically performed by microbiological assays [9,10]. However, these methods are not able to discriminate between different folate species and are also affected by the presence of interferences, including growth inhibitors and some antibiotics [11]. Most of the chromatographic methods developed so far are based on reversed-phase liquid chromatography?tandem mass spectrometry (RP-LC? MS/MS) [12?15]. However, due to their relatively polar nature, folate catabolites are weakly retained in RP chromatographic columns and poor separation is achieved. Hydrophilic interaction liquid chromatography (HILIC) is gaining popularity in metabolomics since biofluids are mainly aqueous and, therefore, likely to contain polar compounds such as carbohydrates, their phosphorylated derivatives, glycolytic intermediates and organic acids [16]. From the best of our knowledge, only one method for determination of folates by HILIC?MS/MS has so far been reported and applied to human plasma after in-batch sample preparation steps, including standard addition, protein precipitation, filtration, evaporation and reconstitution [17]. Here we present a fully-automated method for the analysis of folate and catabolites, pABGA and a-pABGA, in biofluids. This method is based on the on-line hyphenation of a solid-phase extraction step, automatically carried out by a Prospeckt-2 system, with HILIC?MS / MS chromatographic separation and detection. This hyphenation allows complete automatic performance of the analytical method with a drastic reduction of the analysis time. 236 Nuevas plataformas anal?ticas en metabol?mica 2. Materials and methods 2.1. Reagents LC?MS grade acetonitrile (Panreac, Barcelona, Spain), ammonium formate (Sigma-Aldrich, St. Louis, MO, USA), and deionized water (18 M??cm) from a Millipore Milli-Q water purification system (Millipore, Bedford, MA, USA) were used for preparation of the chromatographic phase. Pure standards of folic acid (FA), its metabolite N-(p-aminobenzoyl)- L-glutamic acid (pA??A) and the internal standard (IS) ?aminopterin? were from Sigma?Aldrich. Individual stock solutions were prepared by dissolving 10 mg of standard in 10-mL 0.2 M potassium phosphate buffer, pH 7.3, containing 0.03% ascorbic acid [17]. The acetamide derivative was prepared from the pABGA metabolite with total conversion [7] by adding 140 ?L 50% acetic acid (Panreac) in deionized water (v/v) and 20 ?L acetic anhydride (Panreac) to 10 mg pABGA. The reaction mixture was vortexed in the dark at room temperature for 1 h. The remaining acetic acid was evaporated under nitrogen at room temperature, and the a-pABGA standard was reconstituted in 10-mL 0.2 M potassium phosphate buffer, pH 7.3, with 0.03% ascorbic acid. Multistandard solutions were daily prepared for optimization of the analytical steps by dilution of the stock standard solutions in deionized water or 80:20 acetonitrile?water v/v. Standard solutions were stable at ??? ?C for at least one month. 2.2. Sample handling Samples were kindly donated by healthy volunteers. No restrictions about diet, life style, sex or age were taken into account. Venous blood was 237 Chapter 6 J. Chromatogr. A, 1217 (2010) 4688?4695 collected into a BD Vacutainer? Plus SST tube containing spray-coated silica and polymer gel for serum preparation (Becton Dickinson). The tube was not opened to ambient air and placed in ice or kept refrigerated until processing. Blood samples and milk samples were centrifuged at 4000? g for 10 min at 4 ?C. The resulting serum was divided into 1-mL aliquots in sterile containers and stored at ??? ?C? Ascorbic acid was immediately added (0.03% w/v) to prevent oxidation of metabolites. For urine samples, the pH was adjusted to 7.3?0.1 with 3 M NaOH or 1 M HCl and ascorbic a cid was added (0.03% w/v) to 1-mL urine aliquots, which were also fro?en at ??? ?C until analysis without further preparation. 2.3. Instruments A centrifuge from Selecta (Barcelona, Spain) was used to separate serum from milk and blood samples. A Prospekt-2 system (Spark Holland, Emmen, The Netherlands), equipped with a MIDAS autosampler, an automatic solid-phase extraction unit (automatic cartridge exchanger, ACE) and a high pressure dispenser syringe (HPD), was used for automated solid- phase extraction. Hysphere MM anion exchange (10 mm? 2 mm, 25?35 ?m particle size) cartridges from Spark Holland were used for the SPE step. A Hysphere method development tray, also from Spark Holland, was used to optimize the type of sorbent used in the SPE protocol. The Prospekt-2 system was on-line connected to an Agilent (Palo Alto, CA, USA) 1200 Series LC system, which consists of a binary pump, a vacuum degasser and a thermostated column compartment. After chromatographic separation, detection was performed by means of an Agilent 6410 triple quadrupole mass detector (QqQ), furnished with an electrospray ion (ESI) source. Agilent MassHunter Workstation was the software for data acquisition, qualitative and quantitative analysis, while the Sparklink System controller v. 2.1 Software was used to control the Prospekt units. 238 Nuevas plataformas anal?ticas en metabol?mica Chromatographic separation was carried out by hydrophilic interaction liquid chromatography, using a Luna HILIC column (100? 4.6 mm, 3 ?m particle size) from Phenomenex Inc. (Torrance, California, USA), furnished with a HILIC 2? 4 mm (5 ?m particle size) Guard Cartridge System. 2.4. Solid-phase extraction After addition of the internal standard to a final concentration of 20 nmol/mL, biological fluids were placed at the MIDAS tray, being ready for analysis without further preparation. The entire procedure was automatically performed, in a sequence of steps schemed in Figure 1 and summarized as follows: (A) the sample loop (100 ?L) is filled by passing 200 ?L sample, the cartridge is clamped into the A CE unit, which is then solvated and equilibrated by means of the HPD unit with 1 mL methanol at 5 mL/min (step 1) and 2 mL loading solution (2 mM potassium phosphate, pH 7.3?0.1) at 5 mL/min, respectively; then, 1-mL loading solution is passed through the cartridge at 0.5 mL/min (step 2); (B) after switching the autosampler valve, the sample is loaded by passing 1-mL loading solution at 0.5 mL/min through the sample loop and then through the SPE cartridge (step 3 (A) 239 Chapter 6 J. Chromatogr. A, 1217 (2010) 4688?4695 (B) (C) Figure 1: General scheme of the automatic steps in the Prospeckt 2 system: (A) cartridge loading and solvatation and equilibration, (B) sample loading and cartridge washing (C) elution and washing. HPD: high pressure delivery unit, AV: autosampler valve, ACE automatic cartridge exchange unit, SV: selection valve. 240 Nuevas plataformas anal?ticas en metabol?mica the cartridge is subsequently washed with 1-mL of 10% acetonitrile?water at 2 mL/min (step 4); an input/output signal is established between the Prospekt module and the HPLC system (step 5), which is waiting for an external start,(C) the elution starts by switching the ACE valve 1 to the load position; thus, the chromatographic mobile phase passes through the SPE cartridge and the analytes are eluted to the chromatographic column; after 2-min elution (step 6), the ACE valve 1 is switched again; a cleaning step of the cartridge, with 2-mL methanol (step 7) and 2-mL water (step 8) is finally performed. Thus, the SPE cartridge is ready for reuse. Each SPE cartridge was used for three analyses. The whole SPE process is performed at room temperature in 10 min. The SPE and chromatographic steps are synchronized in order to increase the throughput analysis, which is 5 samples/h. 2.5. Chromatographic separation and mass-spectrometry detection Chromatographic separation was performed isocratically in 14 min. The mobile phase was 20 mM ammonium formate in 80:20 (v/v) acetonitrile?water (pH 7.3). The flow rate and column oven temperature were set at 1 mL/min and 35 ?C, respectively. Analyses were carried out in multiple reaction monitoring (MRM) positive ionization mode with nitrogen as drying and nebulizing gas. The operating conditions of the ESI?QqQ, were: flow rate and temperature of drying gas 10 mL/min and 300 ?C, respectively, nebulizer pressure 45 psi, capillary voltage 4000 V, dwell time 200 ms and delta EMV (potential of the electron multiplier) 700 V. The quantification transition for each compound was: 266.8?119.9 m/z for pABGA, 308.8?162.0 m/z for a-pABGA, 441.2?294.2.2 m/z for FA and 442.2?295.1 m/z for aminopterin (IS). The mass spectra obtained with the individual standard solutions pABGA (A) and a-pABGA is shown in Figure 2. 241 Chapter 6 J. Chromatogr. A, 1217 (2010) 4688?4695 Figure 2. Mass spectra obtained with individual standards (A) pABGA and (B) a-pABGA. 242 Nuevas plataformas anal?ticas en metabol?mica 3. Results and discussion 3.1. Optimization of chromatographic separation and mass spectrometry detection Study of the chromatographic separation was performed in positive fullscan detection mode, being the peak separation and shape the parameters considered for optimization. Due to the polar character of the target compounds, HILIC was selected for chromatographic separation as long as it provided an optimum retention and separation. Separation in the HILIC column is performed under high-organic mobile-phase conditions, giving an enhanced ESI performance, which leads to an increased sensitivity and precision. Subsequently, the ionization agent, flow rate and temperature of the column oven were optimized. With this aim, ammonium formate, ammonium acetate and formic acid were tested as ionization reagents, over the concentration range of 5?20 mM. The best results in terms of peak area were obtained with 20 mM ammonium formate in 80:20 acetonitrile?water, which was used further on. The pH of the mobile phase strongly affects the chromatographic separation; thus, pH values between 3 and 9 were investigated, being 7.3 pH the optimum value. Flow rates between 0.5 and 1.5 mL/min and temperatures between 20 and 40 ?C were also tested, being 1 mL/min and 40 ?C, respectively, the optimum values. The influence of the operating conditions of the mass spectrometer ?namely the temperature of the drying gas, capillary voltage and nebulizer pressure? on the peak area was studied by a multivariate design in order to take into account possible dependence relationships of these variables, being the optimum 250 ?C, 3700 V and 45 psi, respectively. Optimization of the best MRM transitions for the analytes and the internal standard was performed with 10 ?g/mL multistandard solutions in a sequence of steps that involved the selection of the polarity, precursor and product ions, 243 Chapter 6 J. Chromatogr. A, 1217 (2010) 4688?4695 voltage of the first quadrupole, collision energy and dwell time. The optimization of detection is summarized in Table 1. Table 1. Optimization of mass spectrometry operation parameters for the multiple reaction monitoring . Analyte Molecular weight (g/mol) Q1 m/z Q3 m/z Retention time (min) Cone voltage (V) Collision Energy (eV) Dwell time (ms) FA 441.4 442.2 295.1 8.5 140 17 200 p-ABGA 266.0 266.8 119.9 5.0 120 15 200 ap-ABGA 308.0 308.8 162.0 5.2 120 8 200 Aminopterin(IS) 440.4 441.2 294.2 10.5 140 22 200 3.2. Optimization of solid-phase extraction and on-line elution to the liquid chromatograph The complexity of biological fluids demands for a suitable sample preparation protocol able to remove the effect of potential interferences prior to chromatographic separation. In this sense, analysis of urine samples is marked by its high-salt content that causes ionization suppression and also negatively affect the instrument performance due to the formation of non-volatile residues. On the other hand, analysis of serum and milk samples is limited by the presence of a large number of proteins that interfere in the chromatographic separation and mass detection if they are not removed [18]. This interference makes the analysis of folate and catabolites especially difficult as long as they are present at very low concentrations. Accordingly, 244 Nuevas plataformas anal?ticas en metabol?mica direct analysis of biofluids is rarely performed. Garbis et al. developed a method for determination of folates in human serum based on direct analysis after deproteinization, filtration, vacuum evaporation and reconstitution of the sample [17]. However, the manual performance of these steps enhances the risk of contamination and analyte losses as biological materials are directly handled in all operations. Solid-phase extraction provides clean and concentrated extracts and the possibility of fully-automated on-line implementation. Optimization of the SPE is summarized in Table 2. Firstly, selection of the SPE sorbent was carried out with the following cartridges: silica-based cyanopropyl phase (CN-SE), silica-based ethyl phase (C2-SE), end-capped silica-based octyl phase (C8 EC-SE), high-density end-capped silica-based octadecyl phase (C18 HD), polydivinyl-benzene phase (Resin GP), strongly hydrophobic polystyrene-divinylbenzene phase (Resin SH) and mixed-mode phase containing a strong anion exchange functional group (MM anion). This study was performed with both 10 ? g/L multistandard solutions and spiked samples in order to check the influence of matrix effects. The poorest retention was obtained with CN-SE, C2-SE, Resin SH and Resin GP, while MM anion gave the best results in terms of area and peak shape, as the analytes are present as anionic species at weakly basic pH. The MM anion phase is specially suited for extraction of organic acids since it combines a dual interaction based on ionic and polymeric mechanisms responsible of a high retention capacity. Other relevant parameters optimized were those related with the selection of the sample loading conditions such as pH, concentration of acetonitrile and loading flow rate. For optimization of the sample pH, values within 3 and 9 were tested, for which the appropriate amounts of formic acid, ammonium phosphate buffer (0.2M pH 7.3) or NaOH were added. The maximum retention capability was at pH values above 6. Therefore, physiological blood and milk serum pH at 7.3 was selected. In case of urine samples with a pH range from 4 to 8, samples were buffered to pH 7.3 by 245 Chapter 6 J. Chromatogr. A, 1217 (2010) 4688?4695 adjusting with NaOH or formic acid. On the other hand, many protocols for analysis of serum and milk include the addition of acetonitrile or other reagents that favor proteins precipitation. Pre-precipitation of proteins was tested by adding different amounts of acetonitrile to biofluid samples spiked with the target compounds. However, similar results were obtained with and without protein precipitation, which suggests that proteins do not interfere in the extraction process, so this step was not required. Finally, the influence of the flow rate for sample loading was studied by testing flow rates between 0.2 and 2 mL/min, being 0.5 mL/min the optimum value. Optimization of the washing step involved the selection of the optimum flow rate, volume, and composition of the solution. Flow rate and volume were not influential, so that they were fixed at 2 mL/min and 1 mL. The washing solution was 10% acetonitrile in water, in order to improve the effectiveness of the washing step without analyte losses. Concerning recovery of the target compounds from the SPE cartridge, the elution time was studied within 0.2 and 4 min. The recovery was approached to 100% with 2-min elution times and levelled off for longer times, thus, being minimized to 2 min. Characterization of SPE protocol was completed by assessment of the breakthrough volume. For this purpose, multistandards at increased concentrations of the target compounds were analysed by using an experimental setup with two cartridges in an on-line configuration. For this purpose, a second MM anion cartridge was clamped in series by means of an additional selection valve [19]. Thus, the breakthrough volume, expressed as the maximum mass of compounds that can be loaded without giving a quantifiable signal in the elution of the second cartridge, was 10 ?g for folic acid, 30 ?g for pABGA and a-pABGA and 5 ?g for the internal st andard. The same study applied to the target biofluids limited the sample loop volume at 100 ?L in order to preserve the retention capability. 246 Nuevas plataformas anal?ticas en metabol?mica Table 2. Optimization of the main variables involved in the SPE step 3.3. Features of the method Recovery was also evaluated with the double-cartridge configuration. Firstly, the cartridges were conditioned and equilibrated prior to sample loading. If the analytes were not completely retained in cartridge 1 or saturation is produced, retention in cartridge 2 occurs. Then, the second valve is switched and the mobile phase is loaded through cartridge 1 onto the chromatographic column. Valve 2 is finally switched for elution of the analytes from the second cartridge. Therefore, the maximum recovery is obtained when the MRM signals from the elution of the second cartridge are minimized. Figure 3 shows the chromatograms obtained with a Variable Tested range Optimum condition Selected condition SPE sorbent CN-SE, C2-SE, C8 EC-SE, C18 HD, resin GP, resin SH, MM anion and MM cation MM anion MM anion pH 4-9 Above 6 7.3 Amount of acetonitrile in the sample (%) 0?30 Non-influential 0 Loading flow rate (mL/min) ????? 0.5 0.5 Amount of acetonitrile in the washing solution (%) 0?30 10 10 Washing volume (mL) 0?3 Non-influential 1 Washing flow rate (mL/min) 0.2?2 Non-influential 2 Sample loading flow rate (mL/min) 0.2?2 0.5 0.5 Elution time (min) 0.2?4 2 2 247 Chapter 6 J. Chromatogr. A, 1217 (2010) 4688?4695 1 ?g/mL multistanda rd solution from the first (A) and second (B) cartridge elution. Signals obtained from the second cartridge were not quantifiable; thus, the retention capability was supposed to be 100% for all the target compounds. As Figure 3(A) illustrates, a band broadening effect is observed in FA peak signal, which can be ascribed to the mixed interaction of MM anion phase. Figure 3. Chromatograms obtained with a 1 ?g/mL multistandard solution from the first (A) and second (B) cartridge elution. 0 1 2 3 4 5 6 7 8 1 2 3 4 5 6 7 8 9 10 11 12 Coun ts vs. Acqu isitio n Tim e ( min ) x10 4 x10 1 0 1 2 3 4 5 6 7 8 9 1 2 3 4 5 6 7 8 9 10 11 12 a - p ABG A, p ABGA IS FA ( B) ( A ) 248 Nuevas plataformas anal?ticas en metabol?mica On the other hand, Figure 4 shows the chromatograms obtained with a mustistandard solution at concentrations above the breakthrough volume (20 ?g/mL of each analyte) from the first and the second cartridge. In this case, retention in the first cartridge was not complete as detected in the chromatogram generated by analysis of the elution fraction from the second cartridge. In order to assess the recovery, urine samples spiked at two concentration levels ?50 and 250 ng/mL? and a blank were analysed. Accuracy was investigated by comparing theoretically and experimentally measured analyte concentrations obtained from spiked samples at the two levels. The results obtained for both parameters are summarized in Table 3. Calibration plots were run by using the standard peak?internal standard peak ratio as a function of the standard concentration. Calibration was performed with multistandard solutions at ten concentration levels ? between 0.1 and 10000 ng/mL? which were analysed in triplicate. The regression coefficients and the dynamic range are shown in Table 3. The lower limits of detection (LLOD), expressed as the mass of analyte which gives a signal that is 3??above the mean blank signal (where ? is the standard deviation of the blank signal) ranged between 0.3 and 5.5 pmol/mL (0.03 and 0.55 pmol on column). The lower limits of quantification (LLOQ), expressed as the mass of analyte which gives a signal that is 10??above the mean blank signal, ranged between 1.1 and 18.8 pmol/mL (0.11 and 1.88 pmol on column). Precision, expressed as repeatability, was estimated by analysis of five aliquots of spiked samples at 50 ng/mL (FA: 0.12 nmol/mL, pABGA: 0.19 nmol/mL and a-pABGA: 0.16 nmol/mL) and 250 ng/mL (FA: 0.60 nmol/mL, pABGA: 0.94 nmol/mL and a-pABGA: 0.81 nmol/mL) in a single working session. As can be seen in Table 3, the results obtained, expressed as relative standard deviation (RSD), were from 2.5 to 4.1% and 1.5 to 3.8% for the low and high level, respectively. 249 Chapter 6 J. Chromatogr. A, 1217 (2010) 4688?4695 Figure 4. TIC MRM chromatograms obtained with a multistandard solution at 20 ?g/mL of each analyte e luted from the first (A) and the second cartridge (B). Table 3. Features of the proposed method. 1 low level: urine pool spiked at 50 ng/mL (FA: 0.12 nmol/mL, pABGA: 0.19 nmol/mL and a-pABGA: 0.16 nmol/mL). 2 high level: urine pool spiked at 250 ng/mL (FA: 0.60 nmol/mL, pABGA: 0.94 nmol/mL and a-pABGA: 0.81 nmol/mL). 3 Accuracy (%): 100 x (measured concentration ? blank concentration) / spiked concentration. Analyte LLOD (pmol/mL) LLOQ (pmol/mL) Linear regression (r2) Dynamic range Precision at low level1 (RSD%, n=5) Precision at high level2 (RSD%, n=5) Accuracy at low level1(%) Accuracy at high level2 (%) pABGA 0.3 1.1 0.9999 103 2.5 1.7 96 100 a- pABGA 0.6 1.9 0.9996 103 3.2 3.8 102 100 FA 5.5 18.8 0.9998 104 4.1 2.8 101 105 251 Chapter 6 J. Chromatogr. A, 1217 (2010) 4688?4695 3.4. Application of the method to human biofluids The proposed method was tested by application to human biofluids in order to check its suitability for clinical analysis. Three biofluids such as serum, urine and human milk were selected for analysis following the developed method. Five independent replicates were made for each biofluid (n=5). Table 4 shows the mean value and the standard deviation obtained for the target compounds in each sample, where all the folates were present at quantifiable levels. Most studies dealing with folate deficiency have been carried out by radioimmunoassay kits for serum analysis. The application of these kits has set a deficiency cut-off for folate concentrations below 11 pmol/mL [20]. As indicated in the introduction section, these assays are not specific and, therefore, respond to associated metabolites. Therefore, taking into account LLODs and LLOQs reported here for folate and metabolites, the proposed method is specially suited for serum analysis. As shows Figure 5(A), folates were successfully analysed in serum without chemical effect of interferences caused by high concentration of proteins. Similarly, the proposed method can be applied to human breast milk as Figure 5(B) proves. In both cases, a-pABGA was the metabolite found at higher concentration. Folic acid is predominantly excreted in urine as more polar catabolite. Therefore, the analysis of folates in this biofluid could be interesting. Urine is a representative biological matrix with a high salt content and, for this reason, sample preparation is a critical task. As Figure 5(C) shows, cleaning and desalting were efficiently performed by the SPE- based approach proposed here. Mass spectrometry analysis was not affected by ionization suppression. 252 Nuevas plataformas anal?ticas en metabol?mica (A) (B) (C) Figure 6. MRM chromatograms provided by blood serum (A), milk serum (B) and urine (C) samples analysed by the proposed method. 253 Chapter 6 J. Chromatogr. A, 1217 (2010) 4688?4695 4. Conclusions A method for analysis of folic acid and its main catabolites, p- aminobenzoylglutamate and the acetamide derivative in biofluids has been developed. The method, based on the on-line hyphenation of a solid-phase extraction step with HILIC?MS / MS chromatographic separation and detection, is fully automated, and no in-batch operations, such as protein precipitation or filtration, were required. The entire analytical process has been carefully optimized and characterized. Subsequently, it has been applied to the analysis of the target compounds in urine, breast milk and serum samples. Although the target analytes were present in the three biofluids, the most abundant folate catabolite in all samples is a-pABGA. On the other hand, the highest concentration of folic acid is found in serum. In fact, it is well established that serum is likely to reflect the metabolic state of an organism as a consequence of both catabolic and anabolic processes occurring in the whole organism, while excreted biofluids, such as urine, provides readout of catabolic processes, in which polar metabolites without metabolic activity are more readily discarded from the body. A metabolic profile of these compounds in the target samples offers valuable information about the abundance of this essential vitamin in humans. This can be of especial interest in groups of individuals with increased risk of deficiency (such as neonates or pregnant women). Comparison between the levels of these related catabolites in the three types of biofluids gives a realistic readout of the occurrence and fate of this water-soluble vitamin in humans. 5. Acknowledgments 254 Nuevas plataformas anal?ticas en metabol?mica The Spanish Ministerio de Ciencia e Innovaci?n (MICINN) is acknowledged for financial support through project CTQ2009-07430. F.P.C. is grateful to MICINN for a Ram?n y Cajal contract (RYC-2009-03921). B.A.S. is also grateful to Ministerio de Ciencia y Tecnolog?a for an FPI scholarship (BES-2007-15043). 6. References [1] M.A. Caudill, J.F. Gregory, A.D. Hutson, L.B. Bailey, J. Nutr. 128 (1998) 204. [2] L.B. Bailey, J.F. Gregory, J. Nutr. 129 (1999) 779. [3] C.E. Butterworth Jr., T. Tamura, Am. J. Clin. Nutr. 50 (1989) 353. [4] T. Tamura, M.F. Picciano, Am. J. Clin. Nutr. 83 (2006) 993. [5] B. Van Guelpen, J. Hultdin, I. Johansson, C. Witth?ft, L. Weinehall, M. Eliasson, G. Hallmans, R. Palmqvist, J.-H. Jansson, A. Winkvist, J. Intern. Med. 266 (2009) 182. [6] M.R. Malinow, Clin. Chem. 41 (1995) 173. [7] A.A.H. Sokoro, M.L. Etter, J. Lepage, B. Weist, J. Eichhorst, D.C. Lehotay, J. Chromatogr. B 832 (2006) 9. [8] H. McNulty, J. McPartlin, D. Weir, J. Scott, J. Chromatogr. 614 (1993) 59. [9] Z. Fazili, C.M. Pfeiffer, M. Zhang, R. Jain, Clin. Chem. 51 (2005) 2318. [10] C.M. Pfeiffer, Z. Fazili, L. McCoy, M. Zhang, E.W. Gunter, Clin. Chem. 50 (2004) 423. [11] R. Hannisdal, A. Svardal, P.M. Ueland, Clin. Chem. 54 (2008) 665. [12] Z. Fazili, C.M. Pfeiffer, M. Zhang, Clin. Chem. 53 (2007) 781. [13] R.J. Pawlosky, V.P. Flanagan, J. Agric. Food Chem. 49 (2001) 1282. 255 Chapter 6 J. Chromatogr. A, 1217 (2010) 4688?4695 [14] J.D.M. Patring, J.A. Jastrebova, J. Chromatogr. A 1143 (2007) 72. [15] E.P. Quinlivan, A.D. Hanson, J.F. Gregory, Anal. Biochem. 348 (2006) 163. [16] S. Cubbon, C. Antonio, J. Wilson, J. Thomas-Oates, Mass Spectrom. Rev. (2009), doi:10.1002/mas.20252. [17] S.D. Garbis, A. Melse-Boonstra, C.E. West, R.B. van Breemen, Anal. Chem. 73 (2001) 5358. [18] B. ?lvarez-S?nchez, F. Priego-Capote, M.D. Luque de Castro, Trends Anal. Chem. 29 (2010) 120. [19] Spark Holland B.V., Prospeckt Technical note 6, Strategy for accelerated on-line solid phase extraction method development; serial extraction, 1996. [20] H.X. Wang, ?. Wahli n, H. Basun, J. Fastbom, B.Winblad, L. Fratiglioni, Neurology 56 (2001) 1188. CHAPTER 7: Ultrasonic enhancement of leaching and in situ derivatization of haloacetic acids in vegetable foods prior to gas chromatography? electron capture detection Ultrasonic enhancement of leaching and in situ derivatization of haloa cetic acids in vegetable foods prior to gas chromatography ?electron capture detection B. ?lvarez S?nchez, F. Priego Capote*, M. D. Luque de Castro Department of Analytical Chemistry, Annex C-3 Building, Campus of Rabanales, University of C?rdoba, C?rdoba, Spain Journal of Chromatography A, 1201 (2008) 21?26 261 J. Chromatogr. A, 1201 (2008) 21?26 Chapter 7 Ultrasonic enhancement of leaching and in situ derivatization of haloacetic acids in vegetable foods prior to gas chromatography ?electron capture detection B. ?lvarez-S?nchez, F. Priego-Capote, M.D. Luque de Castro Abstract A continuous ultrasound-assisted approach to enhance the extraction of nine haloacetic acids (HAAs) from vegetables with in situ derivatization to methyl esters for their gas chromatography (GC) analysis is presented. The optimization of simultaneous extraction (using acidic methanol as extractant) and derivatization enabled the completion of both steps in 15 min. Ultrasound assistance has proved to enhance both linked steps, which results in a considerable shortening of the overall analysis time (i.e. 552.1 and 552.3 EPA methods for analysis of these compounds in drinking water require 1 and 2 h, respectively, only for derivatization). After sample preparation, the esterified HAAs were isolated by liquid?liquid extraction with n-hexane and analyzed by GC?electron capture detection. The proposed method is an interesting alternative to present methods for the determination of HAAs in vegetable foods. This is an area unjustifiably forgotten by reference laboratory organisms as proves the absence of official methods for analysis of the target compounds in these samples. The proposed method can be applied to the analysis of HAAs in any solid sample after optimization of the main variables involved in the extraction? derivatization step. 262 Nuevas plataformas anal?ticas en metabol?mica 1. Introduction Haloacetic acids (HAAs) are a group of chlorinated, brominated and mixed bromo?chlorinated organic compounds included in the category of ?disinfection by-products? (???s)? This category includes a heterogeneous variety of compounds with significant differences in polarity, volatility and chemical structures. Haloacetic acids are commonly formed in water as a result of the reaction between organic matter, naturally present in water, and oxidizing compounds formed during disinfection with chlorine gas (Cl2) ?e.g. hypochlorous acid (HOCl) [1]. Disinfection is a common treatment for water pre-conditioning in order to avoid the spread of diseases, such as cholera or poliomyelitis. Chlorination is the most extended alternative to obtain drinkable water because of its low level of chlorine residuals, protection against microbial recontamination and relative low cost [2,3]. Despite chlorinated derivatives are the most common DBPs, brominated organic compounds can also be formed due to reaction between disinfectants and bromine ions present in water. The formation of brominated by-products in water is favoured by high bromine levels [4]. There are epidemiological evidences of a positive relationship between consumption of chlorination by-products in drinking water and bladder, liver or rectal cancer in humans [5,6]. For this reason, the Environmental Protection Agency (EPA) has set strict regulatory limits for the sum of five HAAs (chloroacetic, dichloroacetic and trichloroacetic acids, and bromoacetic and dibromoacetic acids) in water [7,8]. Concerns about HAAs are exclusively focused on their presence in drinking water. However, there are other potential contamination sources for humans and animals such as direct exposition of foods to disinfected water. For instance, irrigation can be a way to expose humans to these 263 J. Chromatogr. A, 1201 (2008) 21?26 Chapter 7 carcinogenic agents. Despite the toxic effects and potential contamination of HAAs, methods for the determination of haloacetic acids in foods have not so far been reported, probably due to the difficulty to handle solid samples. In fact, most of the developed methods for the determination of haloacetic acids are based on extraction from an aqueous sample to an organic extractant, derivatization with diazomethane or acidic methanol and determination of the methyl esters by gas chromatography (GC) coupled with either mass spectrometry (MS) ?????? or electron capture detection (EC?) ???????? The ?C?ECD coupling has provided higher sensitivity than GC?MS [1] and, for this reason, constitutes the basis for 552 [15] and 552.2 [16] EPA Methods, and standard method 6251B [17]. Ultrasonic-assisted leaching has proved to be an effective way to extract a number of analytes from different types of solid samples ???????? This high extractive power can be ascribed to two main reasons: (1) the localized influence of extremely high temperatures and pressures under cavitation, which increases solubility and diffusivity of target analytes, and also favours penetration and transport at the interface between a solution and a solid matrix; and (2) the oxidative energy of species created during sonolysis of the liquid phase (?? and O?? and hydrogen peroxide in water) [21]. In this research, the aim was to study the influence of ultrasonic energy on the extraction of HAAs from vegetable foods, but also on their in situ derivatization to methyl esters. This dual ultrasound assistance enables displacement of the leaching equilibrium by esterification of the leached compounds. This is expected to result in a high leaching efficiency together with a considerable shortening of the time for sample preparation. A dynamic manifold was designed to accomplish this objective. After sample preparation, the extract was subject to liquid?liquid extraction (LLE) prior to analysis by GC?ECD as the EPA Methods suggest. Figure 1 illustrates the workflow of the method proposed in this research and that used in the EPA Method 552.2 for analysis of HAAs in drinking water. 264 Nuevas plataformas anal?ticas en metabol?mica 2. Materials and methods 2.1. Instruments and apparatus Ultrasonic irradiation was applied by means of a Branson 450 digital sonifier (20 KHz, 450 W) with digital timer, temperature controller and tunable amplitude and duty cycle, which was equipped with a cylindrical titanium alloy probe (12.70 mm in diameter), immersed into a water bath. The sample was placed in the extraction?derivatization chamber, which consists of a stainless steel cylinder (10 cm x 10 mm i.d.) closed with screw-caps at either end, thus permiting circulation of the leaching fluid through it. Sea sand, which was previously washed with hydrochloric acid and calcined, was used in order to avoid sample compaction in the extraction chamber. Figure 1. Workflow of the proposed method and EPA Method 552.2. MTBE: metyl-terc-butyl-??????????????????????????????????????????????????? detection. 265 J. Chromatogr. A, 1201 (2008) 21?26 Chapter 7 A Gilson Miniplus-3 peristaltic pump, fitted with PTFE 0.8 mm i.d. tubing, was programmed for changing the rotation at preset intervals. A Rheodyne 5041 low-pressure injection valve and Teflon tubing were used to build the dynamic manifold shown in Figure 2. A Varian star 3400 CX gas chromatograph (Varian, Palo Alto, CA, USA) equipped with a 63Ni electron capture detector (ECD) and a Varian 8200 CX autosampler was used to sep arate and quantify the analytes. The chromatographic column was a Factor Four Capillary Column (VF-5ms 60 m ? 0.25 mm) i.d. DF = 0.25) from Varian. Helium and nitrogen (from Carburos Met?licos, Spain) were used as carrier and make-up gases, respectively. A conventional meat grinder (Moulinex, Barcelona, Spain) was used to homogenize the vegetable foods. A thermostated water bath and electrical stirrer from Selecta (Barcelona, Spain) were also used. Figure 2. Experimental set-up for dynamic ultrasound-assisted extraction (USAE) and derivatization of haloacetic acids in vegetable foods. PPP: programmable peristaltic pump; SV: switching valve; LC: leaching carrier; ER: extract reservoir; UP: ultrasonic probe; EC: extraction chamber; WB: water bath; C: coil; GC: gas chromatograph; ECD: electron capture detector. PC: personal computer. 266 Nuevas plataformas anal?ticas en metabol?mica 2.2. Reagents and solutions ?eioni?ed water (?? ??cm) from a Millipore MilliQ water purification system (Bedford, MA, USA) was used to prepare the water ? methanol extractant mixtures. Methanol and hexane of Hplc grade were both from Scharlau (Barcelona, Spain), while sulphuric acid was supplied by Panreac (Ba rcelona, Spain). Acids monochloroacetic (MCAA), monobromoacetic (MBAA), dichloroacetic (DCAA), dibromoacetic (DBAA), bromochloroacetic (BCAA), trichloroacetic (TCAA), tribromoacetic (TBAA), chlorodibromoacetic (CDBAA), bromodichloroacetic (BDCAA) were obtained at a purity higher than 98% from Supelco (Madrid, Spain). The stock solutions of haloacetic acids were prepared by dissolving 0.05 g of each standard in 50 mL of hexane, except for bromodichloroacetic acid, prepared at a concentration of 500 ?g/ mL. Anhydrous Na2SO4 from Merck (Darmstadt, Germany) was used as drying agent after the liquid?liquid extraction step. All standard solutions were stored at 4 ?C in glass containers to minimize losses by volatilization or adsorption, as recomended by EPA method 552.2 [1,13]. 2.3. Samples ?egetable foods ?chard and spinach? were obtained from a local market. Samples were milled and kept at ?20 ?C in dark until use. Spiked samples were prepared by adding 800 ?l of standard solution at the suitable concentration to 2 g of milled sample; then shaked vigorously for 2 h in a mechanical stirrer until evaporation of the organic solvent. 2.4. Ultrasound-assisted extraction of HAAs with in situ derivatization 267 J. Chromatogr. A, 1201 (2008) 21?26 Chapter 7 The experimental set-up for dynamic ultrasound-assisted extraction (USAE) and derivatization of haloacetic acids is depicted in Figure 2. Two grams of milled vegetable food was placed in the extraction chamber (EC) and three glass balls were added to avoid clogging and thus high-pressure problems. Then, the extraction chamber was assembled to the dynamic system and filled with ??? sulphuric acid in methanol ?extractant and derivati?ing reagent (E?R)? by means of the peristaltic pump (??)? The dynamic system becomes a closed circuit by switching the selection valve (SV). After filling the system, the extraction chamber was immersed into the water bath (WB) at 35 ?C. The extractant carrier was then circulated through the solid sample under ultrasonic irradiation (duty cycle 0.4 s, output amplitude 20% of the converter applied power) for a 10-min preset time with the probe (UP) placed at the minimal distance to the top surface of the extraction cell, but without contacting it. During extraction, the direction of the leaching carrier (at flow-rate = 2 mL/min) was changed each two min to minimize increased compactness of the sample in the extraction chamber that could cause overpressure in the system. After the extraction time was elapsed, the selection valve was switched and the extract collected for subsequent analysis steps. 2.5. Analysis of target compounds by gas chromatography?electron capture detection After the ultrasonic-assisted step, the extract was subject to LLE for solvent exchange and cleanup prior to injection into the gas chromatograph. First, the methanol extract was mixed with 1 mL hexane and shaken for 2 min in a mechanical electrical stirrer with subsequent isolation of the organic phase. Then the hexane extract was cleaned with 1 mL water for three times. Then, the resulting organic phase was dried with Na2SO4 before individual separation of the target analytes by GC and detection by ECD. 268 Nuevas plataformas anal?ticas en metabol?mica The carrier gas employed was helium (99.999% purity) at a flow - rate of 1.6 mL/min. Nitrogen (99.999% purity) was used as the ECD make - up gas, at a flow-rate of 30 mL/min. Splitless injections were made by injecting 1.5 ? L of extract. The injection port temperature and detector were set at 200 and 260 ?C, respectively. The oven temperature program for the chromatographic separation was as follows: initial temperature of 35 ?C, held 9 min, increased to 40 ?C at 1 ?C/min, held 3 min, and then to 220 ?C at 6 ?C/min, and then to 205 ?C at 20 ?C/min, then kept at this temperature for 7 min. 3. Results and Discussion The development of the method proposed here was initiated with the optimization of the chromatographic analysis and the optimization of the leaching?derivatization steps in the sample-preparation manifold. Then, the method was characterized in terms of sensitivity and precision. Finally, a validation study was developed by using spiked samples. 3.1. Optimization of the chromatographic analysis of the target compounds Because of the polar nature and acidity of HAAs, they cannot be directly injected into the GC column, so a derivatization step is mandatory. This step is usually carried out by esterification of the carboxylic group ?typically methylation with dia?omethane or acidic methanol? prior to liquid?liquid extraction to an organic phase for chromatographic separation [18]. In this research, acidic methan ol was selected for derivatization, as suggested in the EPA Method 552.2, instead of diazomethane, which has 269 J. Chromatogr. A, 1201 (2008) 21?26 Chapter 7 proved to be highly toxic. Identification of HAA esters and optimization of their individual separation was carried out using standard solutions, which were derivatized with the protocol employed in this EPA method. The experimental chromatographic variables were optimized resulting in the operating conditions described in Section 2. Splitless and split injections were tested to check the influence of the split ratio on the GC?ECD analysis. Injection without split-flow provided the best sensitivity, which is a key factor in the analysis of these compounds taking into account their low allowable levels in water. Identification of the chromatographic peaks was accomplished by comparison of retention times provided by standard solutions of each HAA. Complete separation of them was achieved in less than 30 min. 3.2. Optimization of the variables involved in the ultrasound-assisted extraction with in situ derivatization of haloacetic acids from vegetable foods The composition of the liquid phase (extractant) was crucial for the simultaneous extraction and derivatization to methyl esters of the target acids. Therefore, this study was aimed at the assessment of the influence of the main variables involved in the sample-preparation approach for isolation of HAAs from vegetable foods. The influencial variables were the composition of the liquid phase (expressed as percentage of sulphuric acid in methanol), flow-rate of the liquid phase in the dynamic system, duty cycle of ultrasound irradiation, sonication amplitude, temperature of the water bath where the extraction chamber is immersed, and sonication time. A multivariate approach was applied in order to take into account possible dependence relationships of these variables. The response factors were the peak areas of the GC chromatograms obtained after analysis of the derivatized analytes. Therefore, the objective was to maximize the response 270 Nuevas plataformas anal?ticas en metabol?mica factors resulting from high efficiencies of the extraction and derivatization steps without degradation or losses of the target compounds. The vegetable food selected for the optimization study was spinach, although the resulting optimal conditions were tested with other vegetable foods observing a similar behaviour. A Plackett?Burman design 26x3/16 type III resolution allowing eight degrees of freedom and involving 12 randomized runs plus three centre points was built for a screening study of the seven factors. The upper and lower values for each factor were set according to preliminary experiments. The tested range and the optimum value for each variable are shown in Table 1. The conclusions of this first study were that the concentration of sulphuric acid in methanol and the sonication time were the most influential factors on this dual-step sample preparation approach. Both factors have a positive effect on all HAAs, except for dibromoacetic acid, for which the sonication time was not significant in the range studied. For this reason, higher values of these two variables were tested in a second study. VARIABLE STUDIED RANGE OPTIMUM VALUE Plackett-Burman design Full factor design Bath temperature (?C) ????? 35 35 Duty cycle (%) ????? 40 40 Radiation amplitude (%) 20??? 20 20 Flow-rate (mL/min) ??? 1 1 [H 2SO4 ] (%) ???? ????? 20 Extraction time (min) ???? ????? 10 Table 1. Optimization of the variables involved in the ultrasonic leaching with in situ derivatization step. 271 J. Chromatogr. A, 1201 (2008) 21?26 Chapter 7 Concerning the other variables, the flow-rate was not a significant variable within the studied range for any analyte showing a negative effect in all cases. Therefore, low flow-rates are required to ensure a better sample?liquid phase contact. For this reason, 1 mL/min was selected for subsequent experiments. The irradiation amplitude showed a similar behaviour to that of the flow-rate, so its optimum value was set at 20%. In the case of duty cycle, no effect was observed by changing the duty cycle, so an intermediate value, 40%, was selected that means 0.4 s of ultrasound irradiation per second. The last non-significant variable was the water bath temperature, the increase of which, in contrast to the previous ones, yielded a positive effect favouring the performance of this approach. In the second step, higher values for the sonication time and concentration of sulphuric acid were tested using a two-level full factorial design involving eight randomized runs plus three centre points. The concentration of sulphuric acid was significant with a positive effect. However, the increase from 15 to 20% sulphuric acid was accompanied by fast degradation of both the pump tubes and seals in the dynamic system. For this reason, its value was set at 15% for further experiments. Also the sonication time was positively significant. After setting the optimum values of the other variables, a kinetics study was carried out by testing different sonication times. The results, which enable to monitor the evolution of the extraction?derivatization process, are shown in Figure 3. As can be seen, the kinetic curves have the same behaviour for all the target compounds, having a maximum at 5 min for monochloroacetic acid, dibromoacetic acid and monobromoacetic acid, and at 10 min for bromodichloroacetic acid, dichloroacetic acid, trichloroacetic acid, tribromoacetic acid, bromochloroacetic acid, chlorodibromoacetic acid. After 10 min, a considerable reduction of the signal is observed for some HAAs, probably due to degradation or re- absorption of the target analytes in the sample matrix. In view of these results, ten minutes was selected as the operating sonication time. Figure 4 272 Nuevas plataformas anal?ticas en metabol?mica shows a chromatogram obtained after application of the proposed method to a chard sample spiked with the target analytes at 10 ? g/mL. Figure 3. Kinetics study of the extraction?derivatization step. MCAA: monochloroacetic acid; MBAA: monobromoacetic acid; DCAA: dichloroacetic acid; DBAA: dibromoacetic acid; BCAA: bromochloroacetic acid; TCAA: trichloroacetic acid; TBAA: tribromoacetic acid, CDBAA: chlorodibromoacetic acid; BDCAA: bromodichloroacetic acid. 3.3. Characterization of the method for quantification of haloacetic acids Calibration plots were obtained by using the peak area of each compound as a function of the concentration of the standard conentration. Seven concentration levels were used to build the calibration curves. The regression coefficients (between 0.983 and 0.999) for all analytes are shown in Table 2. The limit of detection (LOD) for each analyte was estimated from vegetable samples spiked with the target HAAs, and expressed as the mass of analyte which gives a signal that is 3? above the mean blank signal (where 273 J. Chromatogr. A, 1201 (2008) 21?26 Chapter 7 ? is the standard deviation of the blank signal). The LODs obtained ranged between 0.4 and 9.7 ng/g. The limits of quantification, expressed as the mass of analyte which gives a signal ??? above the mean blank signal, ranged from 1.6 to 34.3 ng/g. Table 2. Limits of detection (LOD), limits of quantification (LOQ) and regression coefficients (R2) for the proposed method. 3.4. Applicability of the proposed method to spiked vegetable samples The absence of both reference methods for analysis of HAAs in vegetable foods and certified materials made mandatory the use of spiked food samples in order to demonstrate the applicability of the proposed method? Two different types of vegetable ?spinach and chard? were spiked at two concentration levels, 2 and 10 ?g/ mL, and analysed in triplicate with the proposed method. Previous analyses proved that none of the tested Analyte LOD (ng/g) LOQ (ng/g) R2 MCAA 9.7 34.3 0.992 MBAA 4.9 15.8 0.997 DCAA 0.5 1.9 0.999 TCAA 1.0 3.1 0.997 BCAA 0.7 2.3 0.997 DBAA 0.5 1.9 0.989 BDCAA 1.1 3.3 0.995 CDBAA 0.4 1.6 0.983 TBAA 0.5 1.8 0.996 274 Nuevas plataformas anal?ticas en metabol?mica samples contained detectable levels of the target analytes according to the determination method used in this research. The results obtained are shown in Table 3 for each sample and concentration level. As can be seen, the values found were in agreement with the concentrations spiked for each sample. The recoveries ranged from 80 to 115% for the target HAAs, which demonstrates the utility of the overall method for analysis of these compounds. The assessment of the precision provided acceptable values in terms of within-day variability. These values were similar for both vegetable foods, ranging from 1.97 and 4.93% for 2 ? g/mL and from 2.42 to 6.12% for 10 ? g/mL, expressed as relative standard deviation. Figure 4. Chromatogram obtained after application of the proposed method to a chard sample spiked with the target analytes at 10 ?g/mL. MCAA: monochloroacetic acid; MBAA: monobromoacetic acid; DCAA: dichloroacetic acid; DBAA: dibromoacetic acid; BCAA: bromochloroacetic acid; TCAA: trichloroacetic acid; TBAA: tribromoacetic acid, CDBAA: chlorodibromoacetic acid; BDCAA: bromodichloroacetic acid. 275 J. Chromatogr. A, 1201 (2008) 21?26 Chapter 7 4. Conclusions The sample preparation approach proposed allows leaching of HAAs from vegetable foods with in situ derivatization to methyl esters. The target analytes were isolated by ultrasound-assisted leaching in a dynamic system, then converted into methyl esters. Ultrasound energy provided an enhance ment over both linked steps resulting in an acceleration effect that completes the process in only 10 min (it is worth noting that EPA methods for analysis of these compounds require at least 1 h for quantitative derivatization). After this treatment, the esterified HAAs are transferred to a hexane phase by liquid?liquid extraction for subsequent analysis by GC? ECD. The overall results qualify the proposed method as suitable for routine analysis of these compounds in vegetable foods as it is endowed with remarkable characteristics such as short extraction times, low energy and acquisition costs and low handling of toxics and hazardous organics and reagents. The method has been applied to two vegetable foods, spinach and chard, although it can also be applied to the analysis of HAAs in any solid sample after optimization of the main variables involved in the extraction? derivatization step. 5. Acknowledgements The Spanish Ministerio de Educaci?n y Ciencia is gratefully acknowledged for financial support (project no. CTQ 2006-01614). B. ?lvarez -S?nchez is also grateful to the Ministerio de Ciencia y Tecnolog?a for an FPI scholarship. Sample Spiked con cen tra tio n Fou n d MCAA MBAA DCAA TCAA BCAA DBAA BDCAA CDBAA TBAA Spin ach 2 ?g/ mL Concen tra tio n (?g /mL ) 2.2 (2.65) 2.0 (3.92) 2.1 (4.14) 2.2 (3.97) 2.0 (3.15) 2.3 (1.97) 2.1 (3.05) 2.1 (3.25) 2.1 (4.21) Reco very (% ) 110 100 105 110 100 115 105 105 105 Chard Concen tra tio n (?g /mL ) 1.8 (3.65) 2.2 (4.92) 1.9 (3.58) 1.8 (4.80) 1.9 (3.32) 1.6 (2.98) 2.2 (3.2 2) 2.1 (4.35) 2.2 (4.93) Reco very (% ) 90 110 95 90 95 80 110 105 110 Spin ach 10 ?g/ mL Concen tra tio n (?g /mL ) 8.8 (5.13) 9.0 (5.58) 9.2 (6.12) 9.8 (4.15) 10.9 (5.35) 9.7 (2.42) 9.5 (5.14) 10.5 (3.95) 10.6 (5.80) Reco very (% ) 88 90 92 98 109 97 95 105 106 Chard Concen tra tio n (?g /mL ) 8.9 (4.78) 9.2 (5.75) 9.8 (2.89) 9.9 (4.95) 10.9 (4.24) 10.8 (3.12) 10.4 (5.6) 9.6 (4.55) 9.5 (5.08) Reco very (% ) 89 92 98 99 109 108 104 96 95 Table 3. Determination of haloacetic acids in spiked vegetable food samples at different concentrations by the proposed method (n = 3 replicates) and corresponding error in brackets expressed as percentage. 277 J. Chromatogr. A, 1201 (2008) 21?26 Chapter 7 6. References [1] E.T. Urbansky, J. Environ. Monit. 2 (2000) 285?291. [2] R. Sadiq, M.J. Rodr?guez, Sci. Total Environ. 321 (2004) 21 ?46. [3] K. Gopal, S.S. Tripathy, J.L. Bersillon, S.P. Dubey, J. Hazard. Mater. 140 (2007) 1?6. [4] L. Heller-Grossman, J. Manka, B. Limoni -Relis, M. Rebhun, Wat. Res. 27, 8 (1993) 1323?1331. [5] K.P. Cantor, C.F. Lynch, M. Hildesheim, M. Dosemeci, J. Lubin, M. Alavanja, G. Craun, Epidemiology 9(1) (1998) 21?28. [6] R.L. Melnick, A. Nyska, P.M. Foster, J. H. Roycroft, G. E. Kissling, Toxicology, ???(???) (????) ???????? [7] US Environmental Protection Agency, 1998. Disinfectants and disinfection by-products; final rule. Federal Register 63 (241), 69478. [8] US Env ironmental Protection Agency, 2006. Stage 2 Disinfectants and disinfection byproducts rule. EPA-HQ-OW-2002-0043. [9] Y. Chie, Wat. Res. 35(6) (2001), 1599 ?1602. [10] A.D. Nikolau, S.K. Golfinopoulos, M.N. Kostopoulou, T.D. Lekkas, Wat. Res. 36(4) (2002) 1????????? [11] Yee -Chung. Ma; Chi-Yung Chiang, J. Chromatogr. A 1076(1 ?2) (2005) 216?219. [12] E. Malliarou, C. Collins, N. Graham, M. Nieuwenhuijsen, Wat. Res. 39(12) (????) ?????????? [13] M.N. Sarri?n, F.J. Santos, M.T. Galcer?n, J. Chromatogr. A 859 (1 999) ???????? ???? M??? Rodr?gue?, ?? Serodes, ?? Roy, ?at? Res? ??(??) (????) ?????????? 278 Nuevas plataformas anal?ticas en metabol?mica [15] US Environmental Protection Agency, EPA Method 552, Environmental Monitoring Systems Laboratory, Cincinnati, OH (1990). [16] US Environmental Protection Agency, EPA Method 552.2 National Exposure Research Laboratory, Cincinnati, OH, 1995.) [17] A.D. Eaton, L.S. Clesceri, A.E. Greenburg (Eds.), Standard Methods for the Examination of Water and Wastewater, 19th ed. (1995. [18] J.A. P?rez -Serradilla, F. Priego-Capote, M.D. Luque de Castro, Anal. Chem? ??(??) (????) ?????????? [19] A. Collasiol, D. Pozebon, S.M. Maia, Anal. Chim. Acta 518(1 ?2) (2004) ???????? [20] J.L. Luque -Garc?a, M.D. Luque de Castro, Trends Anal. Chem. 22(1) (????) ?????? [21] M.D. Luque de Castro and F. Priego-Capote, Analytical Applications of Ultrasound, Elsevier, Amsterdam (2006). CHAPTER 8: Quantitative targeted profiling of compounds with nutraceutical interest in tomato by mass- spectrometry-based analytical methodologies Q uantitative targeted profiling of compounds with nutraceuti cal interest in tomato by mass - spectrome try - based analytical methodologi es B. ?l varez S?nchez, F. Prie go Capo te , M. D. Luque de Castro * Department of Anal yti cal Chemistry, Annex C - 3 Building, Campus of Rabanal es, University of C?rdoba, C?rdo ba, Spain Insti tut e of Bio medical Resear ch Maim? nides (IMIBIC), Reina Sof?a Hospit al , University of C? rdo ba, E - 1 4 07 1 , C?rdoba, Spain Sent to Phyto chemical Anal ysis for publicati o n 283 Sent to Phytoche m ica l A na lys is Cha p t er 8 Quanti ta tive ta rgeted p ro filing of compo unds with nutra ceut ica l int erest in tomato by mass - sp ectr ometry - bas ed ana lytica l methodolo gies B. ?l varez - S?nchez, F. Prie go - C apo te , M.D. Luque de Castro * Abst ra ct Int roduct ion ? Tomato (Sola num lyc opersi cum ) is a widely cons um ed - frui t rich in nutraceu ti c als such as carotenoi ds , provi tam i ns , vitami ns or phenolic com pound s , whic h are well - known bec aus e of thei r phytoc hemi c al and organoleptical properti es. Obje ctive ? Quantitati ve analysis of a panel of compounds with contri bution to the nutri tional quali ty of tomato speci es has been carri ed out. These com pounds may be us e d as potential markers for selec tion of culti vars and harves t time to improve the quali ty of frui ts in tom ato breedi ng programs . Monitored metaboli tes were carotenes (hydrophobi c nutrac eu ti cals), phenols and asc orbic acid (hydrophi lic nutrac eu tic als ) and c arbohydrates (organoleptical properti es ). Me thodology ? The methodologi es used were bas ed on liqui d chrom atography ?m as s spec trom etry for the analysis of carotenoids , as corbic aci d and phenols , and on gas chrom atography ?mass spec trom etry for the analysi s o f sugars . The target com pou nds were isolated by ultrasou nd - as si s ted extrac ti on. Results ? Analyti cal methods of interes t in tom ato breedi ng program s were reported for quanti tati on of nutrac eu ti cals in frui ts. High selec ti vi ty and sensi ti vi ty levels were a ttai ned by hyphe nati ng MS detec tors to gas - or liquid chrom atographs . C on clusion ? The sensi tive and selec ti ve metho ds reported here for analysis of the target com pou nds may be us ed to ass es s the nutri ti onal value of tom ato frui t. 284 N uevas plat a for ma s ana l ?t ica s en met a bol ?m ica 1. Intr oductio n T om ato ( So l anum lycopersicum ), a mai n sourc e of natural anti oxi dants , is a widely cons um ed frui t pres ent in many healthy diets inc ludi ng the Medi terranean diet. In fac t, tom ato is ric h in nutrac eu ti cals suc h as carotenoi ds (lyc opene), provi tam i ns (provi tam i n A), vitami ns (ascorbic aci d) and phenols (flavonoi ds ), among others . The term "nutrac eutic al" was coined in 1989 by DeFelic e from "nutri ti on" and "pha rmaceu tic al". Acc ordi ng to DeFeli ce, nutrac euti c al can be defi ned as, "a food (or part of a food) that p rovi des medic al or health benefi ts , inc ludi ng the prevention and/or treatm ent of a dis eas e" (Kalra, 2003). Among nutrac eu tic als pres ent in tom ato, the main anti oxi dants are carotenoi ds , a fami ly consti tuted by over 600 compounds, among whic h the mos t abun dant are lyc opene, ?- carotene, ?- carotene, lutei n, and zeaxanthi n ( Rold?n - Guti ?rrez, Luque de Cas tro, 2007). They are natural fat - s oluble pigments present in fruits and vegetables , but als o synthesi zed by other plants , algae and photos ynthetic organism s . C hemi c ally, carotenoids are formed by eight isoprene units and, thus , com prise a long central chai n of conjugated dou ble bonds and coul d have a hydroc arbon ring in one or both ends . Along with their anti oxi dant acti vi ty, some carotenoi ds als o exert other es s ential biolo gical func ti ons ( i.e. acti ng as reti nol prec urs ors , imm unores ponse modulators and gap - junc ti on comm uni c ati on inducers ). Carotenoi ds have a protec ti ve effec t agai ns t som e hum an dis eas es suc h as cancer, heart dis eas e, mac ular degenerati on and ca tarac ts ( Rao and Rao , 2007 ). Furthe rm ore, carotenoi ds have als o been us ed in the treatm ent of photos ensi ti vi ty dis ease ( Oli ver and Palou, 2000) . Other remarkable group of natural com pou nds is that of phenols , secondary metaboli tes synthesi zed in plants wh ere they act as phytoalexi ns, anti oxi dants , UV light protec tors and natural pigm ents ( Slim es tad and 285 Sent to Phytoche m ica l A na lys is Cha p t er 8 V erheu l, 2009) . The mos t abundant phenoli c com pounds in tom ato are flavonoids , consti tuted by two arom ati c rings linked by a three - c arbon chai n. Thi s arom at ic - bas ed conjugated struc ture renders flavonoi ds , good hydrogen and elec tron donors , and, therefore, good antioxi dants . In additi on, flavonoi ds have proved to protec t agai nst allergi es, inflam mation, hypertensi on and arthriti s . Moreover, they are thought t o reduc e the proliferati ve ac ti vi ty of certai n types of tum or cells and be involved in the apoptos is of HL - 60 leukem ia cells, as propos ed by Surh (1998) and Verhaegen et al. (1995) . Vitami n C, als o known as asc orbi c aci d, is a water - s olu ble vitam i n non - s ynthesi zed by humans, bei ng henc e obtai ned through diet (mai nly by citri c frui ts and tom ato). Vitami n C is involved in many physi ologic al process es suc h as the synthesi s of bile aci ds from cho les terol ( Roc k et al., 1996) , the regenerati on of vitam i n E afte r oxidati ve process es , or the synthes is of the ami no acid carni ti ne ( Rebouc he, 1991). Low or nil intakes of vi tami n C are known to cause scurvy. Moreover, num erous reports have sho wn that vitam i n C is involved in the preventi on of oxidati ve stress - related illnes s es such as cancer, cardiovasc ular dis eas es and the capi llary pluggi ng proces s es produc ed in seps is , as rec ently sugges ted by Wi ls on (2009). Seas onal vari ations of vitami n C conc entrati on in tom ato are related to light expos ure together with sugar co ntent ( Mas sot , 2010). In this sens e, low - m olec ular wei ght sugars are held to parti ci pate in vitam i n C bios ynthesi s both by ac ting as subs trate and as signali ng molecul e in the regul ation of gene express i on of enzymes involved in vitami n C metabolism ( Wh eel e r et al., 1998). Besides partic i pati ng in as corbic aci d synthesi s, carboh ydrates are an important frac ti on to be quanti fi ed in tom ato as they contri bute to both organoleptic (sweetness ) and nutri ti onal properties , partic ularly reduc i ng sugars , due to th eir abi li ty to reac t with ami no acids and protei ns to form Mai llard Reac ti on Produc ts (MRPs ) ( Morales and Jim ?nez - P?rez, 2001) . Des pi te MRPs are known to negati vely affec t som e organoleptic and 286 N uevas plat a for ma s ana l ?t ica s en met a bol ?m ica nutri ti ve properties of edible plants ( i.e. browni ng or chelat ion of mic ronutri ents such as trac e metals ), they are endowed with anti oxi dant, anti mi crobial or cytotoxi c ac ti vi ties (Ch evali er et al., 2001). From the bes t of our knowledge, there are scarc e reports on the analys is of sugars in tom ato, none of them devot ed to the i r indi vi dual quanti fi cati on ( Mas sot et al., 2010) . Tom ato breedi ng program s have been mai nly foc us ed on genetic research to improve yield and resis tance to dis eas es. However, in the last years an inc reasi ng attenti on has been paid to quali ty by improvi ng flavor, organoleptic al properti es and nutri tional quali ty (Mathieu et al., 2009; Eybi s htz et al., 2009; Com lekc i oglu et al., 2010). In thi s sens e, the proposals of strategies to evalua te the nutri ti onal quali ty by monitori ng panels of nutrac eutic al com pounds are well - rec ei ved. Selec ti ve and sensi ti ve MS - bas ed metho ds for the analysis of ascorbic aci d and carotenoi ds ( Garri do Freni ch et al., 2005; Mertz et al., 2009) in tom ato are sti ll scarc e, as rec ently revi ewed by G?m ez Rom ero et al. (2007), w hi le the GC ? MS /MS and LC ? MS /MS determi nation of phenols in a vari ety of plants , inc lu di ng tomato, has been well - im plem ented (G?m ez Rom ero et al., 2007). In fac t, mos t of the exi s ti ng metho ds applied to the quanti tati ve targeted analys is of tom ato are exc lu s i vely devoted to the analysi s of this frac ti on. A recent study has charac teri zed the phenolic frac ti on extrac ted from three tom ato vari eti es by LC ?T OF/MS with quanti tati ve analysis of a panel of charac teri s tic com pounds (G?m ez Romero et al., 2010). To the authors ' knowledge no analyti cal metho ds based on MS have been reported for determi nation of sugars in tomato frui ts . In the pres ent study, the quanti tati ve profi li ng of tom ato com pounds with nutrac eutic al properti es as potential markers for culti var sele c ti on and harves ti ng has been carried out. The sensi ti ve and selec ti ve analys is of the target com pou nds may give a com plete ins i ght of the nutraceu ti c al potential of tomato and ass es s its nutri ti onal value, thus bec om i ng a potential tool to be used in tom a to breedi ng programs . 287 Sent to Phytoche m ica l A na lys is Cha p t er 8 2. Materia ls and methods 2.1. Instruments and apparatus Ultras onic irradi ati on was applied by a Branso n 450 digi tal sonicator (20 kHz, 450W ) with tunable ampli tude and duty cyc le, equipped with a cyli ndri c al titani um alloy probe (12.70 mm in diameter), whi ch was direc tly imm ers ed in the extrac tion beaker. A centri fu ge from Selec ta (Barc elona, Spai n) and a rotary - evaporator B?c hi R - 200 equipped with a heati ng bath B?chi B - 490 (B?chi , Switzerla nd) were used after extrac tion to sepa rate the extrac t and remove the solvent, respec ti vely. An Eppendorf Vac ufuge Conc entrator was used to evaporate the extrac tant in ali quo ts of tomato extrac ts prior to the chrom atographic steps . Analyse s of carotenoids , phenols and asc orbic acid were perfor m ed by revers ed - phas e LC follo wed by MS detec ti on, dependi ng on the case. Chrom atographi c separati on was perform ed with an Agi lent (Palo Alto, CA, USA) 1200 Seri es LC system , whi ch consi s ts of a binary pum p, a vac uum degass er, an autos am pler and a thermos t ated colum n com partm ent. Detec ti on and quanti fi c ati on were carried out with an Agi lent 6410 Tri ple Quadrupole (QqQ) mas s analyzer. Data were proc ess ed by a Mas s Hunter Works tation Software from Agi lent for quali tati ve and quanti tati ve analys is . An Inertsi l ODS - 2 C18 analyti c al colum n (4.0 mm i.d. ? 250 mm; 5 ?m particle size, GL Sciences Inc., Tokyo, Japan) was used for chrom atographic separati on. Gas chromatography ?m as s spec trom etry analys es were carri ed out by a Vari an CP 3800 gas chrom atograph cou pled to a Saturn 2200 ion trap mas s spec trom eter (Sugar Land, TX, USA) equipped with a Fac torFou r capi llary colum n (VF - 5 ms 30 m ? 0.25 mm, 0.25 ?m) from Vari an (Palo Alto, USA). Heli um at a constant flow rate of 1 mL/mi n was used as carrier gas . 288 N uevas plat a for ma s ana l ?t ica s en met a bol ?m ica 2.2. Reagents and samples Dani ela vari ety tom ato sam ples ( Lyco persicum esculentum ) were purchased in a loc al market. Upon ac quis i ti on, sam ples were milled, pooled and lyophi li zed; then, the produc t was vacuum - packed and stored at ? 80 ?C until analysis . The leac hi ng agen t for the soli d ?li quid extrac tion step was a 3:1 (v/v) mixture of LC - grade tetrahydrofuran and metha nol (Panreac , Barcelona, Spai n). Chrom atographic - grade methanol, acetonitri le, tetrahydrofu ran (Panreac), and Milli - Q water (Millipore, Bedford, MA, USA) we re used for the LC mobile phases . Aceti c acid and amm onium form ate (Si gma ?Aldri ch, St. Louis , MO, USA ), used as ioniz ati on agents in the LC?MS/MS analyses, were of MS grade? ?erivati?ation- grade N , O ??is(trimethylsilyl)trifluoroacetamide (?STFA) and trimethylchloro- s i lane (TMCS ) from Sigm a ?Aldri ch and pyri di ne from Merc k (Darm s tadt, Germany) were used in the deriv ati?ation step prior to ?C?MS analyses, for which appropri ate safety prec auti ons (gloves, hood - fume, etc.) were taken. Lycopene, ?- c arotene, lutei n and xanthophylls such as neoxanthi n, viola xanthi n, mutatoxanthi n, antheraxanthi n, zeoxanthi n, ?- c ri ptoxant hin and cantaxanthi n were ac qui red from Carotenature (Lups i ngen, Switzerland). P ure standards of querc eti n, myri c eti n, catec hi n, ruti n, querc eti n 3 - ?- D - gluc os i de, and aci ds dihydroxybenzoic (proc atec huic ), chlorogenic , caffeic , p - c oum ari c , ferulic and asc orbic were from Sigm a. Stoc k indi vidual soluti ons were prepared by dis solving the appropri ate amou nt of each com pound in metha nol to a final conce ntratio n of 2000 ?g/m L. Carboh ydrate standards D - (+) - arabi nose, D - (?)- xylos e, D - (?)- mannose, D - fruc tos e, D - (+ ) - galac tose, D - (?)- glu cose, D - ( ?)- s ucros e, D - ( ?)- lac tos e, were also from Si gm a ? Aldric h. 289 Sent to Phytoche m ica l A na lys is Cha p t er 8 2.3. Solid?liquid extraction In thi s step, 0.5 g of the lyophi li zed sam ple and 25 mL of leac hi ng solvent (a 3:1 tetrahydrofu ran ?m ethanol mixture) were poured into a 100 - m L beaker. Extrac ti on was perform ed by direc t imm ers ion of the ultras oni c probe tip into the mixture at a fixed dis tance of 1 cm from the bottom. The sonic ati on ampli tude and duty cyc le were set at 30% of the nomi nal value of the ultrasonic converter (450 W) and 0.5 s/s , respec ti vely. The extrac ti on process was repeated, and the resul ti ng 50 - m L extrac t was rota - evaporated to drynes s and rec o ns ti tuted in 2 - m L 3:1 tetrahydrofu ran ?m etha nol. The extract was stored at ??? ?C in dark until analysis? ???????????????????????????????????????????? For the analysis of carotenoi ds, 100 ?L of extrac t was evaporated to drynes s and reconsti tuted in 1000 - ?L 5 mM amm onium formate in 10% water ?ac etoni trile, from whic h 10 ?L was injec ted. Chrom atographic mobile phases were 5 mM ammonium form ate in 10% water ? ac etoni trile (mobil e phase A) and tetrahydrofu ran (mobi le phas e B). Initially, 100% mobile phase A was mai ntai ned isoc ratic ally for 6 min. Then, a linear gradi ent of 7 min was programm ed to 75% mobi le phas e B, and to a final com pos i ti on of 100% in 1 min, whic h was mai ntai ned 11 min and, fi nally, a pos t - run of 4 min was program med to re - es tablish the ini ti al condi tions. The flow rate and the colum n oven temperature were set at 1.5 mL/mi n and 30 ?C, respec ti vely. Analyse s were carried out by MS in single ion monitori ng (SIM) mode after ionization in posi ti ve mode. Nitrogen was used as dryi ng and nebuli zi ng gas . The operati ng condi tions of the ESI ? Q were: flow rate and temperature of dryi ng gas 10 mL/m i n and 350 ?C, respec ti vely, nebuli zer press ure 35 psi , capi llary voltage 4000 V, dwell tim e 10 ms and delta EMV (potenti al of the elec tron multi pli er) 700 V. The quanti fi c ati on ions and 290 N uevas plat a for ma s ana l ?t ica s en met a bol ?m ica qua drupole voltage for eac h carotenoi d are sho wn in Table 1. The rest of carotenoids included under ?Reagents and samples? were not found in tom ato extrac ts . Tabl e 1. Mass spect ro metry para meters based on SIM for the anal ysis of ca ro teno ids by LC ?MS. Com pound Quantitati ve ion (m/z) Filteri ng Q voltage (V) Viola xanthi n 600.4 180 Lutei n 568.4 180 Cryptoxanthi n 552.8 220 Lyc opene 536.4/537.4 60 ?- Ca rotene 536.4 60 ???????????????????????????????????????????????????????????? Pri or to LC/MS analysi s of phenols , 100 ?L of extrac t was evaporated to dryness and reconsti tuted into 100 ?L of ini ti al mobi le phas e, and 10 ?L of the rec ons ti tuted ext rac t was injec ted. Chrom atographi c separati on was perform ed in 71 min, bei ng the mobi le phas es A and B 0.4% formic aci d in water and 50% acetoni trile ?m ethanol, res pec ti vely. The flow rate and the colum n oven temperature were set at 1 mL/mi n and 25 ?C, resp ec ti vely. The chromatographi c program was as follows : the ini ti al mobile phas e was set at 4% of B, whi ch was increased to 50% in 40 mi n and, then, to 60% B in 5 min. Finally, B was rais ed up to 100% in 3 min, and mai ntai ned 17 min. A re - equi li bration step of 6 min was programm ed after each chromatographic run. Analyse s were carri ed out in selec ted reac tion monitori ng (SRM) after ioni zation in negati ve mode with nitrogen as dryi ng 291 Sent to Phytoche m ica l A na lys is Cha p t er 8 and nebuli zi ng gas . The operating conditi ons of the ESI ?QqQ, were as follows : flow rate and tem perature of dryi ng gas 10 mL/m i n and 325 ?C, respec ti vely, nebuli zer pressure 40 psi , capi llary voltage 2700 V and dwell time 200 ms. The quanti fi c ati on trans i ti on and the operating condi tions for each phenol are sho wn in Table 2. Tabl e 2. Quantificatio n ions and MS/MS operati ng parameters for anal ysis of pheno l s and asco rbic acid. 2.6. Quantification of mono- ????????????????????????????? Before GG ?MS analysi s, the target com pounds were silylated, for whic h a n 150 - ?L ali quot of extrac t was evaporated to drynes s and C ompou nd Prec u rs or ion ( m/ z ) Produc t ion ( m/ z ) Fra gme ntor (V) Col l ision Ene rgy ( eV) Rutin 609 300 220 45 609 151 220 52 Myric eti n 317 179 100 30 317 151 100 30 317 137 100 30 Que rc eti n 301 151 100 30 301 121 100 30 301 107 100 30 Que rc eti n 3 - ?- D gl uc oside 463 300 200 30 Ca tec hin 289 203 140 20 289 123 140 30 289 109 140 30 Ferul ic acid 193 178 120 15 193 134 120 15 p - C ouma ric acid 163 119 80 15 163 93 80 20 Proc a tec hu ic acid 153 109 120 15 Chl oroge nic acid 353 191 100 10 Ca ffeic acid 1 79 135 120 15 Asc orbic acid 175 115 80 8 292 N uevas plat a for ma s ana l ?t ica s en met a bol ?m ica rec ons ti tuted i n 150 - ?L of the deri vati zation solu tion, whi ch consi s ted of 50 - ?L pyri di ne, 98 - ?L BSTFA and 2 - ?L TMCS . The reac tion mixture was vortexed at room tem perature for 1 h, thus bei ng ready for GC ?MS analys is . As for the GC ?MS condi tions, the tem perature program for the gas chrom atographi c separati on started at 50 ?C, held for 2 min ; then , a temperature gradient of 4 ?C/m i n to 300 ?C, and, finally, held for 10 min, as r eported by G?m ez - Gonz?le z et al. (2010) . Therefore, the overall tim e for each chrom atographic run was 74.5 min. Qualifi ers and quanti fi ers for each com pound were moni tored in SRM mode usi ng the conditi ons desc ri bed in Table 3. Tabl e 3. Working parameters of the GC ?MS metho d used for the anal ysis of mono - and disacc harides. Comp ound Rete ntion time (min ) Qua ntifi e r ion ( m/z ) Excita tio n stora ge ( m / z ) Excita tio n amp litud e (V) Qua lif ie r ions ( m/z ) D - ( - ) - A ra bin ose 19 .6 9 21 7 76 .5 60 12 9 , 143 , 14 7 D - ( - ) - Ga la ct ose 20 .3 43 5 12 0 60 21 8 , 305 ,3 31 D - ( +) - Ma nn ose 22 .8 5 20 4 71 50 14 9 , 151 , 16 3 D - ( - ) - Fructose 23 .2 9 43 7 12 0 60 22 9 , 257 , 34 5 D - ( +) - Glucose 24 .8 3 20 4 71 .8 50 14 3 , 152 , 16 3 D - ( +) - Sucrose 35 .1 5 36 2 96 70 15 5 , 183 , 27 1 3. Results and Discus sion 3.1. Optimization of the ultrasound-assisted extraction S oli d ?li qui d extrac tion was selec ted in thi s research for sam ple preparation to isolate the com pounds of interes t. The extrac ti on proc ess was 293 Sent to Phytoche m ica l A na lys is Cha p t er 8 as sis ted by ultrasou nd to enhance the leachi ng kineti cs . Ultraso nic energy was chos en ins tead of mic rowaves as auxiliary energy to avoid a signi fic ant inc reas e of tem perature of the leachi ng medi um that could prom ote analytes degradation. In fac t, ultras onic - ass is ted extrac ti on is frequently refe rred as ?cold extraction? because of the reduced effect of irradiation on the medium tem perature. Optim i zati on of the mai n vari ables involved in thi s step was carri ed out to ens ure the preci se and quanti tati ve extrac ti on of the target com pounds . Pri or leac hi ng, tom ato sam ples sh ou ld be lyophi li zed sinc e the water content in this frui t can reac h levels from 90 to 95%. In thi s way, extrac tion effic iency can be express ed as dry wei ght and enzyme acti vi ty is mini mi zed. Due to differences in polari ty of the target com pou nds (from non - polar charac ter of carotenoi ds to polar charac ter of asc orbi c aci d), the optim um leac hi ng extrac tant com posi tion was properly optim i zed to obtain the maxim um leachi ng effi ci ency in a single extrac tion step. Mixtures of THF and metha nol have been reported as sui ted extrac ti on medi a for isola ti on of carotenoi ds from diffe rent types of food (Di as et al., 2010; Garrido - Frenich et al., 2005). Addi tionally, metha nol seems to be one of the mos t effi ci ent solvents for extrac tion of phenolic hydrophi lic anti oxi dant s from vegetable foods (Garc? a - S alas et al., 2010). Sim i larly, THF and ethanol ?water mixtures have been used as leac hing medi a pri or to quantifi cati on of small organic carboxylic and dic arboxyli c aci ds (oxali c, succ i ni c , glycolic and malonic ac i ds ) (Di etze n et al., 1997) and water - s olu ble vitam i ns (Genti li et al., 2008), respec ti vely. Therefore, THF ?alc oho li c mixtures shou ld lead to quanti tati ve extrac tion of asc orbi c acid at norm al concentrations in tomato. With these premi ses , THF ?m etha nol mixtures withi n 25 and 90% were tes ted to evaluate the optimum com pos i ti on for isola tion of lipophi li c and hydrophi lic anti oxi dants from lyophi li zed tom ato. As res pons e variables , the concentrations of the diffe rent com pounds moni tored in thi s res earch grouped into fam i l i es (phe nols , carotenoids , carboh ydrates and ascorbic ac i d, whic h was independently evaluated) were us ed. Two trends in extrac tion effi ci enc y were observed whi ch dis crim i nated two groups of 294 N uevas plat a for ma s ana l ?t ica s en met a bol ?m ica c om pounds : on the one side, non - pola r carotenoi ds and, on the othe r side, phenols and asc orbic acid. A statis tic al signi fi cant extrac ti on of carotenoi ds was reported at THF conc entrations above 75% (95% confi dence level). However, highe r conce ntratio ns of THF led to a decreas e in extrac tion effic iency of phenols and asc o rbic aci d. For this reas on, the extrac tant com pos i ti on was set at 75:25 (v/v) THF ?m etha nol. Apart from extrac tant com pos i ti on, othe r optim ized fac tors were the probe posi tion (dis tance of the tip to the water - bath bottom ) by immersi on in the leac hi ng medi um , percent of ultras ound expos ure or duty cycle (fracti on of each sec ond during whic h ultras onic irradi ati on is appli ed), ultrasou nd ampli tude and irradi ation tim e. No statis ti c al influe nce was observed for probe pos i tion (from 1 to 3 cm) and duty cyc le ( from 0.5 to 1.0 s/s ) on the extrac ti on of anti oxi dants and sugars by com pari son of total concentrations. The ampli tude of ultrasou nd irradiati on was signi fi cant for isolation of carotenoi ds affec ti ng negati vely to thei r stabi li ty. By contras t, there was no stati s ti cal signi fic anc e of ultras oni c am pli tude to extract phenols and sugars in the range studied (from 10 to 70% ampli tude). Nevertheles s, there was a posi ti ve effec t of ultrasonic am pli tude to extrac t sugars from 50% ultras oni c ampli tude as com pared t o lower ampli tudes . After evaluati on of thes e resul ts , a com promi se soluti on was establishe d by setti ng the ampli tude at the maxi mum valu e without potenti al degradati on of carotenoi ds, which was 30%. Under the com prom is ed condi tions (0.5 s/s duty cycle, 3 0% am pli tude, probe posi tion 1 cm from the bottom , 3:1 THF ?metha nol extrac tant), a kinetic s study of the extrac ti on of carotenoi ds was carri ed out. No degradation of carotenoi ds was observed up to 10 min extrac tion, when a sli ght degradati on of lycopene an d carotene was obs erved. For thi s reas on, the sam ple was extrac ted for this period and three equal cyc les were repeated to chec k leaching com pleti on. Two cyc les under the opti mi zed conditi ons were requi red for quanti tati ve extrac ti on of the com pounds of in terest, whi ch was vis ualized by tomato dec oloration. 295 Sent to Phytoche m ica l A na lys is Cha p t er 8 3.2. Characterization of nutraceuticals in tomato A panel of metaboli tes was us ed for quanti tati ve analysi s of nutrac eutical c om pounds in tom ato, the selec ti on of which met the followi ng cri teri a: (a) the com pounds are known to be pres ent in mos t of the tomato vari eties accordi ng to the informati on gathe red from the literature; (b) they are known to be involved in phytoc hemi cal properti es of tomato, thus bei ng of interes t from a nutriti onal poi nt of view; (c) they contri bute to the anti oxi dant ac ti vi ty of tomato. Quanti tati on of com pounds was supported in all cas es by tri plicate analys is . Caro tenoids. T he absence of stable produc t ions for carotenoids in tandem mas s spec trom etry jus ti fi es the optimi zati on of an SIM metho d for thei r analys is . Cali bration plots for quanti fi cati on were run by usi ng the peak area of the extrac ted ion chrom atogram for eac h monitored ion as a func tion of the standard concentration. Cali bration was perform ed with multis tand ard solu ti ons at ten conce ntrati ons ? between 30 ng/m L and 1000 ?g/m L. The limi ts of detec tion (LO Ds ), express ed as the mass of com pound whi ch gives a signal that is ?? above the background noise signal (where ? is the standard devi ati on of the bac kground noi s e signal) ranged between 0.01 and 0.05 ? g/g. The limi t of quantific ation, express ed as the mas s of com pou nd whi ch gives a signal ??? above the background noise signal, was within ???? and 0.16 ? g/g. Table 4 shows the analytic al features of the metho d for determ i nati on of carotenoi ds in tom ato. The method was appli ed to charac teri ze the levels of c arotenoi ds in tom ato. For thi s purpose, the conc entrati on of carotenoi ds in the extrac t was obtained by interpolation withi n the linear portion of the given cali brati on curve by triplic ate experim ents. Figure 1 sho ws the extrac ted ion c hrom atogram in SIM m ode for the target carotenoi ds . Lyc opene was clearly the mos t conc entrated carotenoid in extrac ts from lyophi li zed tom ato with a 296 N uevas plat a for ma s ana l ?t ica s en met a bol ?m ica c oncentration of 935 ? g/g. This level is in agreem ent with other studi es previ ous ly reported in the literature suc h as thos e de velo ped by Oli ves et al. (2006), and Toor et al. (2005). At a sec ond level, ?- c arotene, whic h is widely us ed as supplem ental nutrac eutic al, was fou nd at 29.0 ? g/g. Althou gh the lyc opene - to - carotene ratio vari es withi n tom ato vari eties , lyc opene is fou nd at cons iderably hi gher conc entrations than carotene (G?mez - Rom ero et al., 2007). Other quanti fied xantho phylls were viola xanthi n, lutei n and cryptoxanthi n found at 7.3, 6.6 and 4.5 ? g/g dry matter, respec ti vely. Other studi es , suc h as thos e carri ed out by M? ller (1997) and Burns et al. (2003), have als o reported repres entati ve conce ntrati ons of thes e xantho phylls in tom ato extrac ts , altho ugh cryptoxanthi n was only detected in tom ato fruits obtained from plants cultivated in greenhouses. Hydro phi lic pheno l s and asco rbic acid. At least two trans i ti ons from precurs or to produc t ions were used for detec ti on of hydrophi li c anti oxi dants monitored in this research, except for querc eti n - 3 - ?- D - glu c os i de, 3,4 - di hydrobenzo ic aci d, chlorogenic ac i d, caffei c aci d and as corbic aci d (Table 2), whic h were detec ted with only one transi tion. Sim i larly to lipophi li c anti oxi dants , ten multi s tandards at conc entrati ons rangi ng from 0.01 to 1000 ?g/m L were used to run the cali bration curves for quantific ati on of hydrophi lic anti oxi dants . By the sam e way, these analys es enabled to estim ate LODs , whic h ranged from 13 to 312 ng/g. Conc erni ng LOQs, they were from 40 to 1039 ng/g, setti ng the lower level of the cali brati on curve for each analyte. These res ults can be seen in Table 5, whic h lis ts the quanti tati on transi tion for each hydrophi lic antioxi dant togethe r with quanti tati ve param eters such as calibrati on curve, LODs and LOQs. Table 5 also inclu des the concentration of the hydrophi lic antioxi dants found in tom ato extrac t express ed as dry wei ght taking into ac c ount that tom ato was lyophili zed. Figure 2 illus trates a chrom atogram obtained by analys is of lyophi li zed tom ato extrac t. Com pou nds are organized as flavonoids , phenoli c aci ds and, finally, asc orbi c aci d representi ng a polar 297 Sent to Phytoche m ica l A na lys is Cha p t er 8 vi tami n. In fac t, asc orbi c aci d was the unique non - phenol analysed in this method and, partic ularly, the mos t conce ntrated hydrop hi lic anti oxi dant detected, with a conce ntrati on above 1 mg/g dry weight. Lyophi lization of the origi nal sam ple allowed analysi s of asc orbic aci d in tom ato extrac t with mini mum degradati on. Asc orbi c aci d concentrati on level found in this charac teri zati on e mphas izes the nutrac eutic al properties of tomato bas ed on the anti oxi dant capaci ty of this com pound. Figure 1. Extract ed ion ch ro mato gram in single ion monito ring (SIM) mode for the target caro tenoids from a tomato extract . T he mai n phenoli cs detec ted in tom ato extrac ts were flavonoids catec hi n, querc eti n and a conjugated form of the latter suc h as ruti n ( que rc eti n - 3 - O - ruti no s i de), with conce ntrati ons rangi ng from 81.8 to 89.9 ? g/g dry weight; and phenoli c aci ds such as chlorogenic aci d at 119.1 ? g/g and caffeic aci d at 51.7 ? g/g both express ed as dry weight. It is worth mentioning the relationshi p between chlorogenic ac i d and caffei c aci d since the former is form ed by es teri fic ation of aci ds caffe ic and qui nic (1,3,4,5 - tetrahydroxyc yc lohe xanec arboxylic acid). Thes e res ul ts are fai rly cons is tent 298 N uevas plat a for ma s ana l ?t ica s en met a bol ?m ica with other studi es previou s ly reported in bibliography pointi ng out the relevance o f ruti n and chlorogenic aci d as two of the mos t abundant phenols in different studied tomato spec ies (Caris - Veyrat, et al., 2004; Mart?nez - V alverde et al., 2002; Mui r et al., 2001). In fac t, mos t of the anti oxi dant ac ti vi ty of the phenolic frac ti on of tom a to is asc ri bed to thes e two com pounds (B?nard et al., 2009), whi ch are mainly dis tributed in tomato peel tis sue . Apart from caff ei c aci d, other hydroxyc i nnam ic aci ds suc h as feruli c aci d and p - c oumari c aci d were also detec ted in lyophi li zed tom ato extrac ts although at lower conce ntrati ons than the other monitored phe nolic ac i ds . Simi larly, other conjugated forms of querc eti n, querc eti n - 3 - ?- D - glu c os i de and myri c eti n were found at low conc entratio ns as com pared to rutin, que rc eti n and catechi n. 3.3 Characterization of sugars as compounds with organoleptical properties T he selec tion of highl y - valu able sugars for thei r quanti tati ve analys is was bas ed on thei r contri bution to tom ato organoleptic properti es , particul arly to the sweetnes s and arom a, but also to thei r connec ti on with the anti oxi dant acti vi ty of tom ato (reduc i ng sugars ) and metabolis m of vitami n C (mai nly suc ros e). Addi ti onally, it is known that n on - s truc tural carboh ydrates , among other solu tes, ac t as osmoregulators and osm oprotec tors of the tolerance res pons e to abi otic stres s es ( G?mez - Gonz?l ez et al., 2010). Sim i larly to LC ?MS analyses, cali bration plots for GC ? MS determi nati on of sugars were run by usi ng the peak area of the standard as a func tion of standard concentration. Multi standard solu tions at ten concentrations between ??? and ??? ?g/mL were analysed after silylation by usi ng the deri vatization protoc ol previous ly des c ri bed. The cali brati on equations and regress i on coef f ic i ents are shown in Table 6. Tabl e 4. Quantitatio n of caro teno ids in tomato ?????????????????????????????????????????????????????????????????? Tabl e 5. Quantitatio n of hydro phi lic antio xidants in tomato extract s obt ained under the co nditio ns described in the experimental sect io n . C ompou nd Qua ntita tiv e tra ns itio n ( m/ z ) Conc e ntra tio n (?g / g m ea n, n = 3) SD (%) Ca l ibra tion curve Regr es s ion (R 2 ) LOD (ng/ g) LOQ (ng/ g) Rutin 609 ? 300 89.92 4.36 y= 18.21x ? 1196 .7 0.994 100 336 Myric eti n 317 ? 179 1.24 3.95 y=1922.9x + 45 8.37 0.9998 46 154 Que rc eti n 301 ? 121 81.80 0.48 y=5936.7x + 33 5.12 0.9999 31 102 Que rc eti n 3 - ?- D gl uc oside 463 ? 300 10.33 0.15 y= 4.2x ? 1935.7 0.9962 312 1039 Ca tec hin 289 ?2 0 3 83.26 4.16 y= 1023 .2x + 520.8 3 0.9997 20 67 Ferul ic acid 193 ? 134 19.24 5.00 y= 3490.7 x ? 576 .5 0.9999 86 287 p - C ouma ric acid 163 ?1 1 9 11.35 1.68 y= 40. 217 x ? 30. 57 0.9981 58 193 Proc a tec hu ic acid 153 ? 109 6.40 5.83 y = 1522 .3 x + 44 0.7 0.9997 13 44 Chl oroge nic acid 353 ? 191 119.13 2.85 y= 25.3 x ? 146 .7 0.9999 66 219 Ca ffeic acid 179 ? 135 51.72 0.25 y= 17.21x ? 1040 .57 0.9993 52 173 Asc orbic acid 175 ? 115 1031 6.17 y= 25635.4 x ? 14 82 0.9937 62 208 Com pound Cali brati on curve R 2 LOD (?g/g) LOQ (?g/g) Conc entrati on (?g/g) Standard devi ati on (%) Viola xanthi n y= 155191x+ 6440.1 0.9987 0.01 0.03 7.3 2.10 Lutei n y = 7817.5x + 1884.2 0.99 95 0.05 0.16 6.6 2.36 Cryptoxanthi n y=6243x + 29128 0.9959 0.01 0.03 4.5 3.12 Lyc opene y = 309832x + 2929.3 0.9931 0.01 0.03 935.2 4.38 ?- Ca rotene y = 120355x + 6464.7 0.9930 0.01 0.03 29.0 3.85 300 N uevas plat a for ma s ana l ?t ica s en met a bol ?m ica The LO?s, expressed as for phenols, ranged between ???? and ??? ?g/g? The LOQs, also expressed as for phenols, ranged between ??? and ? ?g/g . Figure 2. Mul ti pl e react ion monito ring (MRM) chromato gram of pheno l s and asco rbic acid fro m a tomato extract . Figure 3 sho ws the tim e segments for the identified com pounds in SRM mode, reported by analysis of a tom ato extract after silylati on. Table 6 sho ws the mean conc entratio n value and the standard devi ation for the target carbohydrates found in the pooled tomato by triplicate analysi s . 301 Sent to Phytoche m ica l a na lys is Cha p t er 8 Tabl e 6. Features of the meth o d and concentratio n of carbo hydrates in tomato fruit. Thes e ranged from 28.2 to 2020.6 ?g/g dry wei ght and from 2.12 to 5.95 %, res pec ti vely. As the table shows, the mos t abundant sugars were two monos ac cha ri des : gluc os e (the prim ary pho tos ynt he tic produc t) and fruc tos e , whi ch are the mai n res ponsi ble for tom ato sweet flavou r and contri bute signi fi c antly to osm oti c adjus tm ent (Handa et al., 1983). The fruc tos e and gluc os e content is known to be strongly related to maturati on state and growing ( partic ul arly to light expos ure), ripeni ng and storage conditi ons ( Mass ot et al., 2010). In fac t, the glucose - to - fruc tos e rati o varies through frui t growth, being approxi m ately two duri ng the ini tial stages of develo pm ent, but dec reas i ng to less than unity when approac hi ng to maturi ty (Mass ot et al., 2010). Suc rose is the les s abundant sugar, als o respons i ble for sweet flavor and with seasonal vari ation ( Dum as et al., 2003). Minor sugars ( e.g. mannos e, galac tos e and arabinos e) were found at concentrations ab out tenfold lower than fruc tos e and gluc os e. Variati ons in concentrations of som e of these monos ac chari des in tom ato as a consequenc e of plant treatm ent have als o been reported ( Stakhova et al., 2001). C ompou nd Conc e ntra tio n (?g/g) Standa r d devia tio n (rs d,%) Ca l ibra tion curve R? LOD (?g/g) LOQ ( ?g/g) Ara bin ose 72.0 2 . 79 y = 1201 2x + 8968 0 . 986 0. 30 1 Gal a c tose 123.2 5. 15 y = 1729 x ? 29 25 0. 997 0. 30 1 Mannose 267.8 3. 92 y = 601. 7x + 642,4 0. 993 1. 21 4 Fruc tose 1519.6 5. 95 y = 452. 3x + 11432 0. 968 0. 12 0. 4 Gl uc ose 2020.6 4. 87 y = 1189 .x + 11256 0. 9 83 0. 06 0. 2 Sucrose 28.2 2.12 y = 9578 . x + 15524 0. 967 0. 06 0. 2 302 N uevas plat a for ma s ana l ?t ica s en met a bol ?m ica Figure 3. Time segments in singl e react io n monit o ring (SRM) obt ained fro m the anal ysis of a tomato extract . 4. Conc lu sio ns Q ua nti tati ve analysi s of nutraceu tic al valuable compounds pres ent i n tom ato has been carri ed out. F irs tly, an ultrasound - as sis ted soli d ?li qui d extrac tion step was performed to l each the target com pounds . Liqui d chromatography?m as s spec trom etry was used for the quanti tati ve analysis of phenols , carote noi ds and asc orbi c aci d, whi le g as Chromatography ?m ass s pec trom etry was employed for the analys is of sugars. T he sens i ti ve and selec ti ve analys is by MS methodologi es may give a representati ve ins i ght of the nutrac eu ti c al potential of tomato to assess its nutri ti onal value with poss i bili ti es of im plem entati on in tom ato breedi ng programs . 5. A ckno wledgment s The Spanis h Minis teri o de C ienc ia e Innovaci ?n (MICI NN ) is ac knowledged for financial support through projec t CTQ2009 - 07430. F.P.C. 303 Sent to Phytoche m ica l a na lys is Cha p t er 8 i s grateful to MICINN for a Ram?n y Cajal contrac t (RYC - 2009 - 03921). B.A.S . is als o grateful to Minis teri o de Cienc i a y Tec nolog? a for an FPI scholars h ip (BES - 2 0 0 7 - 15043). 6. References 1 . B?nard, B. C., Gauti er, H., Bourgau d, F., Grass elly, D., Navez, B., Caris - V eyrat, C., Wei ss , M., G?nard , M. (2009). Effec ts of low nitrogen supply on tomato ( Solanum lyc opersi c um ) frui t yield and quali ty with spec ial em phasi s on sugars, aci ds , asc orbate, carotenoi ds , and phenolic com pounds . Journal of Agri c ul tural and Food Chemis try , 57, 4112 ? 4123. 2 . Burns , J., Fras er, P. D., Bram ley, P. M. (2003). I denti fic ation and quanti fi c ati on of carotenoi ds , toc ophe rols and chlor ophylls in com monly cons um ed fruits and vegetables . Phytoc hemi s try, 62, 939 ? 9 4 7 . 3. Ca ris - V eyrat, C., Ami ot, M. J., Tys sandier, V., Gras selly, D., Bure, M., Mikolajc zak, M., Gui lland, J. C., Bouteloup - Dem ange, C. (2004). Influence of organi c versus conventi onal agric ul tural prac ti ce on the anti oxidant mic roc ons ti tuent content of tom atoes and derived purees ; consequenc es on anti oxi dant plas ma status in hum ans. Jou rnal of Agri c ul tural and Food Chem is try , 52 , 6503 ?6509. 4. Ch evali er, F., Cho bert, J. M., Genot, C. , Haertl?, T. (2001). Scavengi ng of free radic als , antimi c robi al, and cytotoxi c ac ti vi ti es of the Mai llard reac ti on produc ts of ?- Lac toglobuli n glyc ated with several sugars. Journal of Agric ul tural and Food Chemi s try, 49, 5031 ? 5038. 5. Com lekci oglu, N., Sims ek, O., Bonc uk, M., Aka - Kacar, Y. (2010). Genetic charac teri zati on of heat tolerant tom ato (Solanum lyc opersic on) genotypes by SRAP and RAPD markers . Geneti cs and Molec ul ar Research, 9, 2263 ?2274. 304 N uevas plat a for ma s ana l ?t ica s en met a bol ?m ica 6 . Di as , M. G., Olivei ra, L., Cam oes , M. F., Nunes , B., Vers lo ot P., Hulsho f, P. J. (2010). Criti cal ass essm ent of three hi gh perform ance liquid chrom atography analyti cal metho ds for food carotenoi d quanti fi cati on. Journal of Chrom atography A, 1217, 3494 ?3502. 7. Di etzen, D. J., Wilhi te, T. R., Kenagy, D. N., Milli ner, D. S., Smi th C. H., Landt, M. (1997). Extrac ti on of glyc eri c and glycoli c aci ds from uri ne with tetrahydrofuran: uti li ty in detec tion of prim ary hyperoxaluria. Cli nic al Chemi s try, 43, 1315 ?1320. 8. Du mas , Y., Dadomo, M., Lucc a, G., Groli er, P. (2003). Effec ts of envi ronmental fac tors and agri c ul tural tec hnique s on anti oxi dant content of tom atoes. Journal of the Science of Food and Agri c ul ture, 83, 369 ?382 . 9. Eybi s htz, A., Peretz, Y., Sa de, D., Akad, F., Czos nek, H. (2009). Silenci ng of a single gene in tom ato plants resis tant to Tomato yello w leaf curl virus renders them susc epti ble to the virus . Plant Molec ular Biology, 71, 157 ? 1 7 1 . 10. Gahle r, S., Otto, K., Bohm , V. (2003). Alterations of vitami n C, total phenoli cs , and anti oxidant capaci ty as affec ted by process i ng tom atoes to different produc ts . Journal of Agric ultural and Food Chem is try, 51, 7962 ?7968. 11. Garc?a - S alas , P., Morales - S oto, A., Segura - Ca rretero, Fern?ndez - Guti ?rrez, A. (2010). Phenolic - c om pound - extrac tion sys tems for fruit and vegetable sam ples. Mole cul es, 15, 8813 ?8 8 2 6 . 12. Garri do - Freni ch, A., Hern?ndez - T orres, M. E., Belmonte - Vega, A., Mart?nez - V i dal, J. L. Plaza - Bola?os , P. (2005). Determ i nati on of asc orbic ac i d and carotenoids in food com modi ti es by liqui d chrom atography with mas s spec trom etry detecti on. Jou rnal of Agri c ul tural and Food Chem is try, 53, 7371 ?7376. 305 Sent to Phytoche m ica l a na lys is Cha p t er 8 1 3 . ?entili, A?, Caretti, F?, ??Ascen?o, ??, Marchese, S?, ?erret, ??, ?iCorcia ??, Mai nero - Rocc a, L. (2008). Sim ul taneou s determi nati on of water - s oluble vitami ns in selec ted food matrices by liqui d chrom atography/elec trospray ionization tandem mas s spec trom etry. Rapid Comm uni cati ons in Mass Spec trom etry, 22, 2029 ? 2043. 14. Georg e , S., Tou rni ai re, F., Gautier, H., Goupy, P., Roc k, E., Caris - V eyrat, C. (2011). Changes in the contents of carotenoi ds , phenoli c com pounds and vitami n C during techni cal process i ng and lyophi li s ati on of red and yello w tom atoes . Food Chemi s try, 124, 1603 ? 1611. 15. G?m ez - Gonz?lez, S., Rui z - Jim ?nez, J., Pri ego - Ca pote, F., Luque de Cas tro, M. D. (2010). Quali tati ve and quanti tati ve sugar profi li ng in oli ve frui ts, leaves, and stems by gas chromatography?tandem mass spectrometry (GC - MS /MS ) after ultras ound - as sis ted leac hi ng. Journal of Agri cul tural and Food Chemi s try, 58, 12292 ?12299. 16. G?m ez - Romero, M., Arraez - Rom?n, D., Segura - Ca rretero A. Fern?ndez - Guti ?rrez A. (2007). Analyti c al determi nation of anti oxi dants in tom ato: typic al com ponents of the Mediterranean diet . Jou rnal of Separati on Sci ences, 30, 452 ?4 6 1 . 17. G?m ez - Romero, M., Segura - Carretero A. Fern?ndez - Guti ?rrez A. (2010). Metaboli te profi li ng and quanti fi c ati on of phenoli c com pound in methanol extrac ts of tomato frui ts. Phytoc hemis try, 71, 1848 ?1864 . 18. Handa, S., B ress an, R. A., Handa, A. H., Carpi ta, N. C., Hasegawa, P. M. (1983). Solu tes contri buti ng to osm oti c adjus tm ent in cultured plant cells adapted to water stres s. Plant Physi ology, 73, 834 ?843. 19. Kal ra, E. K. (2003). Nutrac eu tic al ? Defi ni ti on and introduc tion . The Ameri c an Ass oci ation Pharm ac eu ti cal Sc i ence Journal, 5, 25. 20. Le Gall, G. Colquho un, I. J., Davis , A. L., Colli ns , G. J., Verhoeyen, M. E., (2003), Metaboli te Profili ng of Tom ato ( Lyc opers ic on esc ulentum) Usi ng 1 H NMR Spec tros copy as a Tool To Detec t P otenti al Uni ntended 306 N uevas plat a for ma s ana l ?t ica s en met a bol ?m ica Eff ec ts Followi ng a Geneti c Modific ation, of Agri cul tural and Food Chem is try, 51 , 2447 ?2456. 21. Lenuc ci , M. S., Cadi nu, D., Tauri no, M., Piro, G., DAles sandro, G. (2006). Anti oxi dant com posi tion in cherry and hi gh - pi gment tom ato culti vars. Journal of Agric ul tural and Food Chemi s try, 54, 2606 ? 2613. 22. Mart? nez - V alverde, I., Peri ago, M. J., Provan, G., Ches son, A. (2002). Phenolic com pou nds, lycopene and antioxi dant acti vi ty in com merci al vari eties of tom ato ( L y c o p e r s i c u m e s c u l e n t u m ). Journal of Sci ence and Food Agri cul ture, 82, 323 ?330. 23. Mass ot, C., G?nard, M., Stevens , R., Gauti er, H. (2010). Fluc tuati ons in sugar content are not determi nant in explai ni ng vari ations in vitam i n C in tom ato frui t. Plant Physi olo ly and Bioc hem is try , 48, 751 ?757. 24. M a thieu , S., Ci n, V. D., Fei , Z., Li, H., Blis s , P., Taylor, M. G., Klee, H. J., Tiem an, D. M. Flavour com pounds in tom ato frui ts: identi fi cati on of loci and potenti al pathways affec ti ng volati le compos i ti on. Jou rnal of Experim ental Botany, 60, 325 ?337. 2 5 . M c Lauc hlan, W. R., Sanders on, J., Qui nlan, M., Willi am son, G. (1998). Measurem ent of the Total Anti oxi dant Acti vi ty of Hum an Aque ous Humor. Cli nical Chem is try, 44, 888 ?889 . 26. Mel? ndez - Mart? nez, A. J., Fraser, P. D., Bramley, P. M. (2010). Acc umul ation of heal th prom oti ng phytoc hemi c als in wild relati ves of tom ato and their contri bution to in vitro anti oxi dant ac ti vi ty . Phytoc hemis try, 71, 1104 ?1114. 27. Mertz, C., Ganc el, A. L., Gunata, Z., Alter, P., Dhui que - Mayer, C., Vai llant, F., P?rez, A. M., Ruales, J., Bra t, P. (2009). Phenoli c com pounds, carotenoi ds and antioxi dant ac ti vi ty of three tropi c al frui ts . Journal of Food Com pos i ti on and Analysis, 22, 381 ?3 8 7 . 307 Sent to Phytoche m ica l a na lys is Cha p t er 8 2 8 . Morales , F. J., Jim ?nez - P?rez, S. (2001). Free radic al scavengi ng capaci ty of Maillard reac ti on produc ts as related to colour and fluoresc enc e . Food Chemi s try, 72, 119 ?1 2 5 . 29. Mui r, S. R., Colli ns, G. J., Robins on, S., Hughe s, S., Bovy, A., Ric de Vos , C. H., van Tunen, A. J., Verhoeyen, M. E. (2001). Overexpres si on of petunia chalc one isomerase in tomato res u lts in frui t contai ni ng increased levels of flavonols . Nature Bi otec hnology, 19, 470 ? 4 7 4 . 30. M?l ler, H. (1997). Determ i nati on of the carotenoid content in selec ted vegetables and frui t by HPLC and photodi ode array detecti on. Zei tsc hrift f?r Lebensm i ttelu nters uc hu ng und Fors chu ng A, 204, 88 ?9 4 . 31. Oli ver, J., Palou, A. (2000). Chrom atographi c determi nati on of carotenoi ds in foods . Journal of Ch romatography A, 881, 543 ? 555. 32. Oli ves Barba, A. I. C?m ara Hurtado, M., S?nche z Mata, M. C., Fern?ndez Rui z, V., L?pez S?en z de Tejada, M. (2006). Appli c ati on of a UV - vis detecti on - HPLC method for a rapi d determ i nation of lycopene and [beta] - carotene in vegetables . Food Chemi s try, 95, 328 ? 336. 33. Rao , A. V. Rao , L. G. (2007). Carotenoi ds and hum an health. Pharmacolo gy Research, 55, 207 ?2 1 6 . 34. Rebou che , C.J. (1991). Asc orbi c aci d and carni ti ne bios ynthesi s. Ameri c an Journal of Cli ni cal Nutri tion, 54, 1147S ? 1152S . 35. Roc k, C. L., Fada, R. D., Jac ob, R. A., Bowen, P. E. (1996). U pdate on the Biologi c al Charac teris tics of the Anti oxi dant Mic ronutri ents : Vitam i n C, Vitami n E, and the Carotenoi ds . Journal of the Americ an Dietetic Ass oci ation, 96, 693 ?702. 36. Rold?n - Guti ?rrez , J. M., Luque de Cas tro, M. D. (2007). Lycopene: The need for better metho ds for charac terization and determ i nati on . TrAC Trends in Analytical Chem is try, 26, 163 ?170. 308 N uevas plat a for ma s ana l ?t ica s en met a bol ?m ica 3 7 . S lim es tad, R, Verheul , M. (2009). Review of flavonoi ds and other phenoli cs from frui ts of different tom ato ( Ly cop ersicon escu len tu m Mill.) culti vars. Jou rnal of Sc i ence of Food Agri cul ture, 89, 1255 ?1270. 38. S takh ova, L. N., Ladygi n, V. G., Stakhov, L. F. (2001). Changes in the Content of Sugars and Ami no Aci d in the Frui t of Lycopers ic on es cul entum after Que rc etin Trea tm ent. Biolo gy Bulleti n, 28, 471 ?476. 39. S urh, Y. J. (1998). Cancer ch em oprevention by dietary phyto c hemi c als : a mec ha ni s ti c viewpoi nt. Nature Reviews Cancer, 11 , 6 ? 1 0 . 40. T oo r, R. K., Savage, G. P. (2005). Antioxi dant ac ti vi ty in different frac ti ons of tom atoes . Food Research Internati onal , 38, 487 ?4 9 4 . 41. V eli oglu , Y. S., Mazza, G., Gao, L., Oom ah , B. D. (1 998). Anti oxi dant ac ti vi ty in different frac ti ons of tom atoes . Journal of Agri c ul tural and Food Chemi s try, 46, 4113 ?4 1 1 7 . 42. V erhaegen, S., Mcgowan, A. J. Brophy, A. R., Fernandes , R. S., Cotter, T. G . (1995). I nhi bi ti on of apoptosi s by anti oxidants in the hu man HL - 60 leukem i a cell li ne. B iochem is tr y and P harm ac ology, 50, 1021 ?1029. 43. Wh eele r, G.L., Jones , M. A., Smi rnoff, N. (1998). The bios ynthes is pathway of vi tami n C in highe r plants . Nature, 393 , 365 ?369. 44. Wi ls on, J. X. (2009). Mec hanism of ac ti on of vitami n C in sepsi s: as corbate modulates redox signali ng in endothelium . Biofac tor, 35, 5 ? 1 3 . P A RT 3 : Global p rofil ing a nalysis La Parte 3 de la Memoria (Cap?tulos 9 y 10) abarca la investigaci?n realizada en el ?mbito de perfiles metabol?micos, estrategia que proporciona una informaci?n global, no cuantitativa, de los metabolitos existentes en la muestra. Esta herramienta es la m?s adecuada para obtener una primera aproximaci?n de la composici?n metab?lica de muestras desconocidas o poco estudiadas, entre las cuales, y considerando los fluidos biol?gicos, se encuentran la saliva y la leche materna. Ambos fluidos biol?gicos est?n ganando inter?s por ser una fuente reconocida de potenciales biomarcadores de ciertas enfermedades, en el caso de la saliva, y por el valor nutricional en una etapa clave de desarrollo, en el caso de la leche materna. La tarea primordial para obtener un perfil ?ptimo de muestras poco estudiadas es desarrollar un protocolo para su preparaci?n que permita, tras la aplicaci?n de la plataforma anal?tica correspondiente, obtener un perfil lo m?s rico posible en metabolitos. Esta tarea ha constituido la base de la investigaci?n recogida en los Cap?tulos 9 y 10, dedicados a saliva y leche materna, respectivamente. Tras el desarrollo exhaustivo de los protocolos de preparaci?n de la muestra, la utilizaci?n de la plataforma LC?TOF/MS en ambos casos (suplementada con la plataforma NMR en el caso de la leche) ha proporcionado unos excelentes perfiles de estas muestras, que ponen de manifiesto la calidad de los protocolos de preparaci?n de muestra desarrollados. Part 3 of this Thesis, which covers Chapters 9 and 10, deals with the research developed in global profiling of different types of biological samples. This analytical strategy provides global, non-quantitative information of the metabolites present in biological samples. This is of particular interest for exploring samples such as saliva and breastfeeding milk, taking into account the gap in the hitherto metabolomics studies of this biofluid, in contrast to the vast literature devoted to urine and blood metabolomics. Both biofluids are gaining special interest because of being a known source of potential biomarkers for certain diseases, in the case of human saliva, and the nutritional value in a key growing step for breastfeeding milk. The main task for obtaining a reliable global profile is to develop a non-selective analytical protocol to cover the maximum number of metabolites. Therefore, the aim of the research in Chapters 9 and 10 was, on the one hand, to obtain a suitable sample preparation protocol for global profiling of saliva (Chapter 9) and human breast milk (Chapter 10) so as to achieve the maximum metabolite coverage, and, on the other hand, to apply the resulting method to obtain accurate identification of salivary and breast milk metabolites. LC?TOF/MS was used in both cases (together with NMR in the case of breast milk) providing complete profiles and ensuring that the proposed method complies the requisites for global profiling. CHAPTER 9: Human salivary metabolomics profiling by LC?TOF/MS in accuracy mode Human salivary metabolom i cs profiling by LC ?TOF/MS in accuracy mode B. ?lvarez S?nchez* , F. Priego Capote, M. D. Luque de Castro Department of Analytical Chemistry, Annex C-3 Building, Campus of Rabanales, University of C?rdoba, C?rdoba, Spain Institute of Biomedical Research Maim?nides (IMIBIC), Reina Sof?a Hospital, University of C?rdoba, E-14071, C?rdoba, Spain Sent to Journal of Chromatography A for publication 319 Sent to Journal of Chromatography A Chapter 9 Human salivary meta bo lo m ic s pro filing by LC ?TOF/MS in accuracy mo de B. ?lvarez-S?nchez* , F. Priego-Capote, M.D. Luque de Castro Abst ra ct S ali va is a biolo gic al fluid produc ed in the sali vary glands with a com pos i ti on marked by a 99% of water and several minor com ponents inc ludi ng elec trolytes (mai nly sodi um , potass ium , calci um , magnesi um , bic arbonate, and phosphate salts ), a vari ety of protei ns suc h as imm unoglobuli ns , hydrolas es or muc i ns , and other low - m olecul ar wei ght com ponent s such as urea and amm oni a. Saliva has rec ently gai ned populari ty as a potenti al tool for diagnosis and biom arkers monitori ng. In fac t, unli ke blood drawing, sali va sam pli ng is eas y by virtue of its non ? i nvasi veness charac ter and does not requi re health ca re pers onnel. In additi on, sali va com posi tion may potenti ally reflec t plasma metaboli te levels and, therefore , may be used as an indic ator of the phys iolo gic al state. The ai m of the present study is to opti mi ze the sam ple preparati on protocol to obtain a m etaboli te profili ng analys is of hum an sali va by Liqui d Chrom atography ?T im e - o f - Fli ght/Mass Spec trom etry (LC ?T OF/MS ) in hi gh - ac curac y mode. Under the optim um sam ple preparation condi tions, identific ation of potenti al molecul ar features was carri ed out. T h i s r e s u l t e d i n 12 identi fi ed com pounds in negati ve mode and 91 in posi ti ve mode with a mas s toleranc e window below 10 ppm from the aci dic extrac t, whi ls t 13 in negati ve mode and 52 in posi ti ve were identi fi ed from the alkali ne extrac t. A m i n o a c i d s , s u g a r s , l i p i d s a n t i o x i d a n t s a n d o t h e r p o t e n t i a l l y i n t e r e s t i n g b i o m a r k e r s s u c h a s p o l y a m i n e s , vi t a m i n B 3 , a n d e t h y l p h o s p h a t e h a v e b e e n i d e n t i f i e d . T h i s s t u d y c o v e r s t h e g a p o f k n o w l e d g e a b o u t t h i s b i o f l u i d a n d o p e n s ne w p o s s i b i l i t i e s f o r t h e s e l e c t i o n o f s a l i v a a s s o u r c e o f m e t a b o l i t e bi o m a r k e r s r e p r e s e n t a t i v e o f s p e c i f i c d i s o r d e r s . 320 Nuevas plataformas anal?ticas en metabol?mica 1. Intr oductio n S ali va is a biolo gic al fluid produced in sali vary glands marked by a com pos i ti on with 99% water and several minor com ponents inc lu di ng elec trolytes (mai nly sodium , potas si um , calci u m , magnesi um , bicarbonate, and phos pha te salts ), a vari ety of protei ns ( i.e. immunoglobuli ns , hydrolas es or muc i ns ), and other low - m olec ular weight com ponents ( i.e. urea and amm onia) [1]. These com ponents carry out a vari ety of func tions suc h as diges ti on, lubric ation, bufferi ng, cleans i ng, aggregati on of mi c roo rganism s to form plaque, and acti ng as imm une protec ti ve agains t exogenous patho gens [2]. Therefore, sali va is cri tic al to pres erve and mai ntai n the health of oral tiss ues . Sali va com posi tion is af fec ted by phys i ologic al, pathologi c al, and envi ronmental fac tors . Thu s, mas ti c ati on, food intake, nutri ti onal changes , age [3], circ adi an and circ annua l vari ati ons [4], and medi cation intake [5] may caus e changes in the flow and com posi tion of sali va. Thes e variati ons affec t the metaboli te profi le and, therefore, make the readou t obtai ned from its analys is unrealis tic . Sali va is gai ni ng populari ty as a com plementary or a standalone sam ple for target analys is of certai n biom arkers due to the fac t that conc entration of several metaboli tes direc tly reflec ts plas ma levels. Moreover, unli ke blood drawing, sali va sam pli ng is non - i nvas i ve, more cos t - eff ec ti ve and does not requi re health personnel care. Therefo re, sali va cou ld be us ed to diagnose certai n disorders , monitor the evolution of pathologies or to adjus t the dos age of medi ci nes [6,7]. On the othe r hand, sali va has been largely overloo ked for metabolic profi li ng [8], with the exc eption of an extrem ely lim i ted num ber of studies [9]. In fac t, the hum an metab olom e database (HMDB) does not inc lude comm on sali vary metaboli tes withi n its 321 Sent to Journal of Chromatography A Chapter 9 data index [10]. Therefore, the study of the global profi le of sali vary consti tuents is jus ti fi ed. In general term s , the final goal of metabolomics is to obtai n a reproduc i ble and reli able insi ght of the metaboli te com posi ti on, either referred to a defi ned set of com pou nds (target analys is ) or to the who le set of small - m olecul ar wei ght metaboli tes (global metabolic profi li ng). In the latter, the mai n problem is the vas t num ber o f com ponents and thei r wide vari ety in chem ic al properties [11]. The two mai n platfo rms used in metabolom ic s are nuc lear magneti c resonanc e spec tros c opy (NMR) and mas s spec trom etry (MS ) [12]. NMR allows direc t analys is of mos t types of sam ples , inc ludi ng t iss ues ; allows the rec overy of sam ples for further analys es , and provi des detai led informati on on molec ular struc ture [13]. However, NMR is charac teri zed by a poor sensi ti vi ty as com pared to mass spec trom etry - based tec hni que s, and spec tra can be very diffi c ul t to interpret for com plex mixtures as biologic al sam ples due to overlappi ng among metabolite signals ???????? On the other hand, mass spectrometry possesses hi gh sensi ti vi ty and selecti vi ty and accurate mass meas urements thanks to recent improvements o f mass analyzers. This tec hni que , usually coupled with ups tream separati on methodolo gi es like gas chrom atography (GC) [17] or liqui d chrom atography (LC) enables unequi voc al identi fi c ati on of com pounds and struc ture eluci dati on of a large num ber of metaboli tes, provi di ng extra inform ati on about the bioc hemical relati ons hi p between them [18]. Despi te these advanc es in analytic al ins trumentati on, the major sou rc e of errors in metabolo mi cs is still linked to sam ple preparati on and the lac k of optimi zed metho ds desi gned for each type of biologic al sam ple. In this sense, if a com plete eluc idation of the metabolom e or, more realis tic ally, the widening of the metaboli te coverage is intended, the ideal sam ple preparation protocol shoul d allow to cover the larges t ra nge of metaboli tes regardles s thei r chemi c al ch arac teris tic s and differenc es in conc entrati on. 322 Nuevas plataformas anal?ticas en metabol?mica T he aim of the present study was to optim ize the analyti cal sam ple preparation protoc ol to obtain a metaboli te profile by LC ?T im e - of - Fli ght/Mas s Spec trom etry (L C ?T OF/MS ) of human sali va. Additionally, identific ation of the maxi mum num ber of extrac ted features or potenti al metabolic peaks is also addres s ed. The pres ent study aims to cover the exi s ti ng gap in metabolom ic s experim ents devoted to sali va, whi ch could be of interes t in the study of sali va com posi ti on vari abi li ty and identi fi cation of potential biom arkers as a biofl ui d com plem entary to serum . 2. Materia ls and methods 2.1. Reagents and samples LC ? MS grade acetonitri le, MS - grade form ic aci d (Scha rla b, Barc elo na, Spai n) and dei onized water (18 M??cm) from a Millipore Milli- Q water purifi cati on sys tem (Milli pore, Bedford, MA, USA) were us ed in sam ple preparation and to prepare the chrom atographi c phas es . Chloroform and methanol from Scha rlab were us ed in the optim i zation of sam ple preparation. Five volu nteers were recrui ted to provi de sam ples ; no one of them referred any disease or health dis order and were not subjec ted to any drug treatment. Thes e subjec ts ranged in age from 21 to 29 yrs (mean age ? standard devi ation, 25.4?3 yr). Sam ples corres ponded to a whole day, taken each six hours. Owi ng to the presenc e of food debri s and soli d matter, a previ ous centri fu gati on step was needed. Thi s was perform ed at 12000 g for 15 min, at 6 ?C. Then, sam ples we re ali quo ted in 2 - m L Eppendorfs and stored at ? 20 ?C unti l analysis . For the opti mi zati on study, sam ples were pooled to a final volume of 1 mL and stored at ?20 ?C. 323 Sent to Journal of Chromatography A Chapter 9 2.2. Instruments and apparatus Ultras onic irradi ati on was applied by a Branso n 450 digi t al sonicator (20 kHz, 450 W) with tunable am pli tude and duty cyc le, equi pped with a cyli ndri c al titani um alloy probe (12.70 mm in diameter), whi ch was direc tly imm ers ed in the extrac tion beaker. Centri fugation was carri ed out with a thermos tated centri fuge Thermo Sorvall Legend Mic ro 21 R from Thermo (The rmo Fishe r Sci entifi c , Bremen, Germ any). A Vacuf uge centri fu gal vacuum concentrator from Eppendorf ( Eppendorf , Inc ., Ham burg, Germany) was used to evaporate sam ples to drynes s . All sam ples were analyzed us i ng a 1200 Seri es HPLC sys tem (Agi lent Tec hnologi es , Waldbronn, Germ any) equi pped with a binary pum p, a degass er, a well plate autos am pler, a therm os tated colum n com partm ent, and coupled to an Agi lent 6530 TOF/MS provided with Jets tream ESI source. Mass Hu nter Works tati on Data acqui si tion software (Agilent Tec hnologi es ) was used to control the ins trum ent. Data were proc es sed usi ng Mass Hunter Qualitati ve Analys is software (Agi lent Technologi es ). Extrac tion of unknown featu res from raw data was carri ed out wi th the Molec ul ar Feature Extracti on (MFE) algori thm in Mass Hunter Quali tati ve analys is software. Analyses were process ed usi ng Mass Hunter Quali tati ve software and com pound identi fi cati on was perform ed usi ng the MET LIN Personal Metaboli te Databas e and Mole c ul ar Form ul a Generati on software (Agi lent Tec hnologi es ). 2.3. Sample preparation protocols Different experim ental protocols were com pared in order to obtain the maxim um metaboli te coverage. On the one hand, hydrolysi s of sali va cou ld release metaboli te s from protei ns and cells pres ent in sali va. Thus, 324 Nuevas plataformas anal?ticas en metabol?mica two different hydrolys is condi ti ons were tes ted, aci dic and mild bas ic conditi ons , so as to com pare the effec t of pH. Wi th this aim , 100 - ?L ali quots were mixed 1:1 with 0.1 M NaOH or HCl in water and vorte xed at room temperature for 30 min, then, evaporated to drynes s and reconsti tuted with mobile phase A. On the other hand, it has been previ ous ly dem ons trated that ultrasou nd (US ) irradi ation ai ds hydrolys is reac ti ons [19,20], helpi ng to releas e metaboli te s that are linked to protei ns or othe r sam ple consti tue nts , and maki ng them amenable for subs equent LC ? MS analys is . Therefo re, US has been previ ous ly used with different types of biolo gic al sam ples , suc h as uri ne [21] and hai r [22], inc reas i ng hydrolys is e ffi ci enc y. Therefore, both ac i dic and basi c hydrolysi s protoc ols were repeated with US irradi ation for whic h the tip of the US probe was imm ers ed direc tly in a 5 - m L vial contai ni ng, 2 - m L sali va mixed with 2 - m L 0.1 M NaOH or HCl. Condi ti ons for US irradi ati on were: irradi ation time, 10 mi n; 50% of total output ampli tude, and 50% duty cyc le. After hydrolys is , 100 ?L of the res ul ti ng solu ti on was evaporated to dryness and rec onsti tuted in mobile phas e A before injec ti ng into the LC ?MS analyzer. Finally, a prec oncentration step (3:1) was carri ed out in order to inc rease the overall coverage. 2.4. Liquid Chromatography ?Mass Spectrometry Chrom atographi c separati on was carri ed out with a Zorbax Ec li pse XDB - C18 colum n (4.6x150 mm, 5 ?m partic le size). 10 ?L of s am ple was injec ted in eac h chromatographi c run. Mobi le phas es , deli vered at 1 mL/mi n, consis ted of 0.1% formi c aci d either in water (A) or in acetoni trile (B). The chrom atographi c gradi ent was as follows : 2% B 5 min followed by a gradi ent to 100% B in 55 min and an isoc rati c step at 100% B for 10 mi n. The ESI Jets tream sourc e operated in pos i tive and negati ve ionization, usi ng the followi ng condi ti ons : needle voltage set at +/ ? 4 kV, 325 Sent to Journal of Chromatography A Chapter 9 nebuli zer gas at 40 psi g, dryi ng and sheath gas flow rate and tem perature at, 8 and 11 L/m i n and 325 and 350 ?C, res pec ti vely. Data were acqui red in both centroid and profi le mode at a rate of 1 spec trum per s in the extended dynam ic range mode ( 2 GHz). The mass range and detection window were set at m/z 100 ?1100 and 100 ppm, respec ti vely. Reference mas s correc tion on each sam ple was perform ed with a conti nuo us infus i on of agi lent tof biopolym er analysis mixture contai ni ng purine (m/z 121.0508) an d hexametho xyphos phazyne (m/z 322.0481) with resolution of 45 000. 2.5. MS data processing Once the sam ples were analyzed by LC ?MS , data were extrac ted into features and the calc ul ated neutral mass was used to calcul ate the molec ul ar formul a by the Mo lecul ar Form ul a Generator Software. Featu res extrac tion was bas ed on the molec ular feature extrac ti on algori thm (MFE), that loc ates and groups all ions related to a sam e neutral molec ule. This relation is referred as to the covariance of peaks withi n the s ame chromatographic retenti on tim e, the charge - s tate envelo pe, isotopi c dis tribution, and/or the presenc e of adduc ts and dimers. The MFE took all ions exc eedi ng 500 cou nts into acc ount, with charge state lim i ted to a maxim um of two and a peak spaci ng toler anc e of 0.0025 m/z (plus 7 ppm ). Each feature was given by a mini mum of two ions . The extrac tion algori thm was bas ed on a common organic model with chrom atographic separation. The allowed pos i ti ve ions were protonated spec ies and sodium, potass ium , formi c aci d and ac etoni trile adduc ts, and the negati ve ions form ed by form ate adduc ts and proton loss es. Deh ydratati on neutral loss es were also allowed. Molecul ar form ulas were generated usi ng the Molec ul ar Formul a Generator algori thm . Both mas s accurac y and is otope inform ati on (abundanc e and spaci ng) were consi dered to lim i t the num ber of hits. Only 326 Nuevas plataformas anal?ticas en metabol?mica the com mon elements C, H, N, O, P and S were cons i dered in the generation of form ulas . The calc ulated neutral mass of eac h feature was queri ed agai ns t the MET LIN da tabase [23] for matchi ng to com pounds withi n mass tolerance window of a maxi mum of 10 ppm . The MET LIN database matched the neutral mas s to the monoi sotopi c mass valu e calcul ated from the em pi ric al form ul a of com pounds in the database. Identi fi ed com pounds were eventual ly contras ted agai ns t the Hum an Metabolom e Databas e (HMDB) [24] to confi rm thei r occ urrence in hum ans . 3. Results and dis cus sion 3.1. Comparison of acidic and basic hydrolysis In a firs t approach, the sali va pool was direc tly analys ed b y LC ? T OF/MS in posi ti ve and negati ve ioni zati on modes after reconsti tuti on in the chrom atographi c mobi le phase A. This analysi s enabled the detec ti on of 5 and 40 potenti al molecul ar features in negati ve and posi ti ve ioni zati on modes , res pec ti vely. This low - detec ti on capabili ty cou ld be asc ri bed to the need for optimi zi ng sam ple treatm ent that wou ld expec tedly improve metaboli te coverage. With thi s regard, diffe rent steps enc om pas si ng hydrolysi s and preconce ntratio n were tes ted. It is well - known that aci dic natural pH of sali va has the role of helpi ng to hydrolyze nutri ents suc h as sugars , whic h are usually form i ng polym ers , as a firs t step of food diges tion. In additi on, certai n metaboli tes bonded to protei ns , ac ting as cofac tors or as a res ul t of pos ttrans lati onal modi fi c ati ons , woul d be releas ed after sui ted hydrolys is . On the othe r hand, hydrolysi s is a comm on proc edure in drug analysi s as it leads to the release of the target spec ies bounded to polar com pounds to inc reas e thei r polari ty and faci li tate th ei r excretion, for ins tance, as glucuronide - li nked com pounds. 327 Sent to Journal of Chromatography A Chapter 9 In thi s sens e, hydrolysi s is a com mon step in preparation of biolo gic al sam ples . Therefo re, a hydrolys is study was planned with the aim of inc reasi ng metaboli te coverage. Both aci dic and bas ic hydrolysi s reac tions were tes ted in order to maxim ize metaboli te identi fic ation in sali va. The protocol selec ted for this study was bas ed on the inc ubati on of 100 - ?L sali va ali quots in 1:1 proportion with 0.1 M NaOH or HCl aque ous soluti on at room tem perat ure for 30 min. The res ul ti ng solu ti ons were evaporated and rec ons ti tuted in mobile phase A as a preconce ntrati on step pri or chrom atographic analys is. Aci di c hydrolys is led to a total of 48 and 74 molecul ar features in negati ve and posi ti ve ioniz ati on mode s, respec ti vely; while bas ic hydrolys is provided 27 and 80 features in negati ve and posi ti ve, res pec ti vely. The resul ts from both types of hydrolys is were com pared with thos e from the protoc ol non - bas ed on hydrolys is (5 and 40 features obtai ned in negati ve and pos i ti ve ionization). Res ul ts from this study were used to bui ld the Venn diagram , sho wn in Figure 1(A ) (negati ve ionizati on) and (B ) (posi ti ve ionizati on) to com pare the metaboli tes profi le of both approache s. As can be seen, 4 com mon features for al l the experim ents were obtai ned in negati ve mode, while bas ic and aci di c hydrolysis provided 23 and 44 uni que featu res . The num ber of common features (23) is highe r in pos i ti ve experim ents, while the num ber of new featu res with res pec t to the non - hydrolyz ed sam ple is greatly enhanc ed, givi ng 47 and 55 features asc ri bed to hydrolys is . Thus , the implem entati on of a hydrolys is step is jus ti fied to inc reas e the metaboli te coverage . Attendi ng to these res ul ts , the identific ation capabi li ty cou ld be inc reased wi th a dual analys is involving hydrolysi s both in aci d and alkali ne medi a. 3.2. Ultrasound assisted hydrolysis 328 Nuevas plataformas anal?ticas en metabol?mica As previous ly stated, the use of auxi li ary energi es to enha nce hydrolysi s reac tions in biolo gic al sam ples is gai ni ng populari ty, as the reac ti o n kinetics are cons i derably enha nced. In this study, the aim was to evaluate the inci denc e of ultrasou nd ass is tanc e to the hydrolysi s proc es s in terms of num ber of metaboli tes identifi ed. Thus , both hydrolysis protocols (aci d and bas ic hydrolysis ) were tes ted by appli c ati on of US energy under moderate irradi ati on condi tions (10 min, 50% of total output ampli tude and 50% duty cyc le). (A ) (B ) Results obtai ned from this study in po si ti ve mode are sho wn in Figure 2. While US energy does not have any impac t on the num ber of detected features in negati ve ioniz ati on, thes e is a consi derably inc reas e of detected features in both ac i dic and basi c hydrolyses in pos i ti ve mode. Thus , US - as s i s ted hydrolys is led to 38 unique features in bas i c hydrolysis , in contras t to 16 enti ti es obtai ned from the hydrolysi s witho ut US ass is tanc e ( A ). On the other hand, 36 and 11 unique features were obtained in aci di c Figure 1. Venn diagram of the extracted features from the analysis of 100-?L saliva after basic and acidic hydrolysis as compared with a non-hydrolyzed sample:(A) negative ionization;(B) positive ionization. 329 Sent to Journal of Chromatography A Chapter 9 hydrolysi s with and witho ut US ass is tanc e, res pec ti vely ( B ). With thes e resul ts , US - ass is tanc e reduc es the hydrolys is proc es si ng tim e , but apart from that, inc reas es the metaboli tes coverage. (A) (B ) Figure 2. Venn diagram of features obtained from the analysis of 100-?L saliv a after ultrasound-assisted hydrolysis reactions (basic and acidic) as compared with a non-hydrolyzed sample. (A): negative ionization, (B): positive ionization. Therefo re, its implementati on for charac teri zati on of sali va metabolom e is of great interes t . Figure 3 shows the total com pound chrom atogram obtained with US - assi s ted basi c (blac k line) and aci di c (grey line) hydrolysi s in posi ti ve mode showi ng a cons i derably highe r num ber of unique peaks in the aci di c hydrolysi s in the fi rs t 14 mins of the chrom atogram. 3.3. Preconcentration When com pared with other biolo gic al flui ds , the num ber of detec ted features from sali va is relati vely small, whi ch can be due to the low 330 Nuevas plataformas anal?ticas en metabol?mica c oncentration at which the target com pounds are present (99% of sali va com pos i ti on is water). Thes e low levels make nec es sary to inc lude a preconc entrati on step. For this reas on, the US - as si s ted hydrolys ate was preconc entrated in a 4:1 ini tial ?fi nal rati o. This res ul ted in an inc reas e to 174 molec ular featu res in pos i ti ve and 84 in negat i ve mode for aci di c hydrolysi s and to 150 and 70 for bas ic hydrolys is . Identifi cati on was done with the features extrac ted from thes e experim ents, the sam ple preparation conditi ons of whi ch were proved to be the optim um. Figure 3. Total ion chromatogram obtained with an acidic (grey) and basic (black) hydrolysate from saliva in positive ionization mode. 3.4. Identification of features In order to eluc idate the sali vary metabolome, a final identi fic ation step was carried out for whic h the extrac ted f eatures were queri ed agai ns t the MET LIN database, whic h inc ludes endogenous metaboli tes found in plasm a besi des com mon hum an drugs and thei r metaboli tes . Finally, further identific ation of the res ulti ng features was done by searc hi ng in the HMDB by com pari ng the experim ental base peak mas s agai ns t the M+H or M ?H 331 Sent to Journal of Chromatography A Chapter 9 peaks as well as othe r adduc ts from the databas e. The HMDB databas e, that inc ludes over 7900 metaboli tes fou nd in human biofluids and tiss ues ac cordi ng to literature res ul ts , does not inc lude inform ati on about sali vary com pos i ti on. This com plic ates interpretation of the resul ts obtai ned in the present res earch. In addi ti on, it is a major issue to dis ti nguis h between metaboli tes from food intake or resul ti ng from endogenous sali vary com pounds . The re s ul ts of the databas e search are sum marized in Tables 1 - to - 4, whic h inc lu de the molecul ar form ula, the detec ted adduc t mas s, matc hi ng tolerance, and the num eri c reference in HMDB. A total of 12 com pou nds in negati ve mode and 91 in posi ti ve mode with a mass tolerance window below 10 ppm were found from the aci dic extrac t, whi ls t 13 in negati ve mode and 52 in pos i ti ve were identi fied from the alkali ne extrac t. Among the identi fi ed metaboli tes , sugars, horm ones , lipi ds , ami no ac i ds , anti oxi dants and other mi nor com ponents were identi fi ed. The occ urrence of these com pou nds in sali va is furthe r disc us sed here. 3.4.1. Sugars Vari ous chrom atographi c peaks corres ponding to C 6 H 12 O 6 compounds were fou nd in chrom atograms from both experim ents , which might corres po nd to the isom ers D - fruc tos e, D - gala c tose, D - gluc os e, D - m annos e, scylli tol or myoi nosi tol. Sali vary sugars are hi ghl y related to cari es and teeth demi nerali zati on. Thus , sugars provide subs trate for the ac ti ons of oral bac teri a, loweri ng plaque and sali var y pH, res ul ti ng in teeth demi nerali zati on [25]. In previ ous studi es , galac tos e, gluc os e, mannose, fucos e, gluc os am i ne and galac tos ami ne have been found in hum an sali va [26]. Bes i des sugars, the sweetener aspartam e has been detec ted; thi s is an artifi ci al, non - c arbohydrate addi ti ve in foods and beverages chem ic ally formed by es teri fic ati on of the ami no aci ds asparti c aci d and phenylala ni ne. Di sacc hari des suc h as suc ros e and lac tos e, maltose or the synthetic 332 Nuevas plataformas anal?ticas en metabol?mica s weetener aspartam e, whi ch are likely to be obtaine d from food, were also found in the aci dic extrac t. 3.4.2. Lipids Among the lipi ds pres ent in sali va, the arac hi donic aci d derivati ves 11 - hydroxyeic os atetraenoate glyceryl ester (11(R) - HET E) 12 - hydroxy - he ptadec atri enoate glyceryl es ter (12(S) - HHT rE) and 2 - (14,15 - epoxy - ei cosatri enoyl) glyc erol and derivati ves from HEPE, KETE and EpET E, were identifi ed . 11(R) - HET E is produc ed from arac hi doni c aci d by two cyc looxygenas es , whi le 12(S) - HHT rE is one of the primary arac hidonic acid metaboli tes of the hum an plat ele ts . Pres enc e of these com pounds in sali va has not been previous ly reported and thei r biologi c al role is unc ertai n. On the other hand, an ami de of the fatty aci d oleic aci d, oleam i de, was fou nd . It is an endogenous subs tance that accumul ates in the cereb ros pinal fluid during sleep depri vation [27]. Pres enc e of phos pholi pi ds in sali va has also been previ ous ly reported [28]. Two groups of phos pholi pi ds, glyc erophos phoeth anolam i nes and glyc erophos phoc holi nes , have been detected in this study. In glyc eropho s phoc holi nes (PC), phos pho rylcho li ne moi ety occ upi es a glyc erol subs ti tuti on site, whi le in phos phati dyletha nolamine (PE) it is occ upi ed by phos pho ryletha nolami ne. Many diffe rent com bi nati ons are pos si ble, with fatty aci ds of varyi ng lengths and saturati on attac hed at the C - 1 and C - 2 posi ti ons of glyc erol. Fatty ac i ds contai ni ng 16, 18 and 20 carbons are the com mones t. This wide vari ety lead to a great num ber of isom ers with identical m olecul ar wei ght that cannot be dis ti nguishe d by single - s tage mass spec tr om etry: some poss ible c andidates are lis ted in Tables 1 ??? Other remarkable group of identi fi ed lipi ds is that form ed by es trogens . Thus , 6 - ketoestrio l, estrone glucuro nide, 16 - oxoes trone, estrone 3,4 - (or 2,3 - ) qui ni ne, estradi ol, and 17 - ethynyles tradi ol were detec ted. Sali va has proved to be an exc elle nt b ioflu id for estrogens moni tori ng [7]. It 333 Sent to Journal of Chromatography A Chapter 9 i s even preferred to bloo d due to the fac t that mos t blood hormones (approxim ately 95%) are protein - bou nded. By contras t, sali va contai ns a majori ty of free horm ones that can be eas i ly measured. As an exam ple, sali v a has been used for more than 30 years for monitori ng es trogens changes throughou t menstrua l cyc le to ass es s ferti li ty. Actual ly, it has been demons trated that hum an sali va sho ws cyclic vari ation in its com pos i ti on during the menstrua l cyc le. Concretely, e stradi ol and its conjugated metaboli tes are found at hi gh levels in women during thei r reproduc ti ve years [29,30]. Predi c ti on of premature birth can als o be detec ted through sali vary es tradiol meas urem ent, a test approved by the FDA (Foo d and Drug Adm i ni s t rati on). Addi ti onally, it is known that horm onal changes can lead to sys temi c dis orders suc h as, diabetes melli tus , cancer or Cushi ng's syndrome [31]. Finally, cortis ol and its derivati ves 18 - hydroxycortic os terone and derivati ve 18 - hydroxycorti s ol was al s o detec ted. While this com pou nd has not been previous ly found in uri ne, sali vary corti s ol has proved to be an exc elle nt indic ator of the plasm a corti sol conc entrati on, whic h is a biomarker for changes in adrenal func tion [32]. 3.4.3. Amino acids T he fre e ami no aci ds , argi ni ne, carni ti ne, leuci ne and proline, were identifi ed. It is well known that free ami no aci ds are pres ent in hum an sali va at relatively low concentrati ons (10 ? ?? 1 0 ? ? M ) [33]. The role and sou rc e of sali vary free ami no aci ds are only partially known. Som e of them are known to be involved in the metabolism of mic roorganisms and protei n decom posi ti on proc ess es. Thei r contri buti on to infla mm atory periodontal dis ease has also been sugges ted [34]. Ami no aci d metabolis m produc ts , suc h as phenylpyruvic aci d, enol - phenylpyruvate, 3 - dehydroxyc arni ti ne , methylglu tarylc arni ti ne, propyi onyl carni ti ne and L - ac etylc arni ti ne have also been identi fi ed. Phenylpyruvi c acid 334 Nuevas plataformas anal?ticas en metabol?mica i s a keto - ac i d produced by oral bac teri a, mai nly by B. gingivalis, B. endodontalis and B. loescheii. Thus, phenylpyruvi c aci d may be us ed as biomarker for ami no acid metabolic ac ti vi ty of oral bacteria . On the other hand, p ropionylc arni ti ne, L - ac etylcarni ti n e and dehydoroxycarni tni ne are ac ylcarni ti nes , produc ed from fatty aci ds oxi dati on to produc e energy. Thus , ac ylcarni ti nes may be used as markers of energy balanc e, exerc is e intensi ty and fat uti li zati on. Acylc arni ti nes als o provide valua ble inform ati on fo r the study, diagnosi s and unders tandi ng of metabolic dis orders , myopathi es, conges ti ve heart fai lu re and end stage of renal disease . Though no previ ous ly been reported in sali va, hi gh levels of p ropi onylc arni ti ne in uri ne are found in pati ents with methyl m alo nyl - CoA mutas e (MUT ) defici ency [35]. 3.4.4. Antioxidants Other interes ti ng group of com pounds detec ted in sali va is that formed by antioxi dants , whic h are likely to be obtained from diet. Anti oxi dants are believed to reduc e the ris k of certai n dis e as es by reac ti ng with free radi cals , whic h are involved in the developm ent of a wide vari ety of dis eas es , inc ludi ng canc er, stroke, chroni c inflam mation and rheum atism . Among them , alkaloi ds caffe i ne, theobrom i ne, theophylli ne, paraxanthi ne, 3 - hydroxypheny lac etic aci d, hydroxic i nnam ic aci d, benzoic aci d and uri c aci d were identifi ed. Alkaloi ds are the mos t widely consum ed psyc hos timul ants in the world, mai nly from tea and coffee . Furtherm ore, hydroxic ynnami c ac i ds can be fou nd in a wide variety of edi ble pl ants suc h as peanuts, tom atoes, carrots , and garli c, and are also used in flavours , perfumes and pharmaceu tic als . Benzoi c aci d occ urs natural ly free and bou nd as benzoic ac i d esters in many plant and anim al spec ies . Benzo ic aci d is a fungi s tatic com pound w idely us ed as a food preservati ve. 3 - Hydroxyphenylacetic is a rutin metaboli te that also exhi bi ts anti oxi dant ac ti vi ty. In addi tion, it is an intermediate in the pathway of tyrosi ne metabolis m. Vanilli n is a cataboli c produc t of ferul ic aci d degradati on. L ike many polyphenols fou nd in plants, 335 Sent to Journal of Chromatography A Chapter 9 vanill i n has antioxi dant and anti - tumor acti vi ties [24]. Uric aci d is one of the mos t important antioxi dants in sali va, and its lev e l s in hum an sali va are related to thos e in bloo d. Interesti ngly, quanti fic ati on of uri c aci d in sali va has been used as a biomarker of gou t [36]. 3.4.5. Other compounds found in saliva It is als o worth noti ng the presenc e of ali phatic polyam i nes sperm i ne ( SP ), sperm idi ne ( SPD ), spermi c aci d and N - acetylc adaveri ne , tradi tionally as soci ated with bad breath. Thes e com pounds also play an important role as cofac tors in several bioc hemi c al proc es ses ass oc iated with cellu lar acti vi ties and proliferati on. Sinc e the presenc e of polyam i nes in bucc odental flui ds has been well establis he d, it is concei vable that thei r levels cou ld inc reas e in several patho logi cal proliferati ve proc ess es . Thus , thei r quanti fi c ati on have been reported as a tool for moni tori ng patholo gies of oral cavity tiss ues as soci ated with an intense metabolic ac ti vi ty, such as gingi val hypertrophy or tumors of the maxi llofac i al area [37]. Niac i nami de or vitam i n B3 can be used as biomarker of pellagra. Ethyl phos phate was als o found, whic h has be en sugges ted as alcohol consum pti on biom arker in sali va, uri ne and serum [38]. Proli ne betai ne is an osm oprotec ti ve com pound and anti bac terial com pou nd fou nd in uri ne. It is thought to serve as osm oprotec ti ve role for the kidney. Proli ne betai ne is a glyc i ne betai ne analogue found in many citrus foods. Elevated levels of proline betai ne in hum an uri ne are found after cons um ptio n of citrus frui ts and jui c es [25]. Other potenti al biom arkers such as creati ne, and nic oti ne gluc uronide (an exc retion produc t of n i coti ne) were also identi fi ed. 4. Conclusi ons Di f f e r e n t s a m p l e pr e p a r a t i o n p r o t o c o l s h a v e b e e n c o m p a r e d f o r g l o b a l m e t a b o l o m i c s pr o f i l i n g a n a l y s i s o f s a l i v a . Altho ugh sali va is scarcely 336 Nuevas plataformas anal?ticas en metabol?mica us ed in metabolomi cs studi es , its potenti al for disc rimi nati ng betw een metabolic profi les correlated with different patho logi es and for moni tori ng changes induc ed by diet would makes this biofluid worth explori ng. Among the protoc ols tested, hydrolys is (both bas ic and acidi c ) of sali va was found to be a sui ted strategy to maxi mi ze the num ber of detec ted featu res . In additi on, ultrasou nd energy was used to enhanc e hydrolysi s with an as soci ated effec t of inc reas ed metaboli te coverage. Fi n a l l y , a p r e c o n c e n t r a t i o n s t e p w a s c a r r i e d o u t t o a t t a i n a n e x t r a l e v e l o f i d e n t i f i c a t i o n . O n c e t h e o p t i m u m s a m p l e p r e p a r a t i o n c o n d i t i o n s w e r e o p t i m i z e d , t h e r e s u l t i n g f e a t u r e s we r e i d e n t i f i e d b y q u e r y i n g a g a i n s t t h e H M D B . A total of 12 compounds in negati ve mode and 91 in posi ti ve mode with a mass toleranc e window below 10 ppm were found from the aci dic extrac t, whi ls t 13 in negati ve mode and 52 in positi ve were identi fied from the alkali ne extrac t. Am i n o a c i d s , s u g a r s , l i p i d s a n t i o x i d a n t s a n d o t h e r p o t e n t i a l l y i n t e r e s t i n g bi o m a r k e r s s u c h a s po l y a m i n e s vi t a m i n B3 a n d e t h y l p h o s p h a t e we r e i d e n t i f i e d . T h i s s t u d y c o v e r s t h e g a p o f k n o w l e d g e a b o u t t h i s bi o f l u i d a n d o p e n s n e w po s s i b i l i t i e s t o th e u s e o f s a l i v a f o r b i o m a r k e r s d i s c o v e r y a n d m o n i t o r i n g . T h e l i m i t e d n u m b e r o f s t u d i e s w i t h s a l i v a i s re f l e c t e d b y t h e a b s e n c e o f da t a b a s e s a n d t h e l i m i t e d b i b l i o g r a p h y , o n l y t a r g e t a n a l y s e s h a v e be e n a d d r e s s e d t o d a t e . By c o n t r a s t , t h i s g l o b a l pr o f i l i n g a p p r o a c h a i m s t o c o m p l e t e t h e c u r r e n t k n o w l e d g e o f t h e h u m a n m e t a b o l o m e . 5. Ackno wledgement s The Spanish Mi nis terio de Ciencia e Innovac i?n (MICI NN ) and FE DER program are thanked for fi nanci al support through projec t CTQ2009 - 07430. B.A. - S. and F.P. - C. are also grateful to the MIC INN for an FPI scho larshi p (BES - 2 0 0 7 - 15043) and a Ram?n y Cajal contrac t (RYC - 2009 - 03921). Table 1. Identification of metabolites in saliva after acidic hydrolysis (negative ionization) HMDB ID Commo n Na me Chemic a l Formul a Theo re tic a l adduc t m/z Experim en ta l er ro r (ma s s units ) Adduc t Acc ura c y er ror (pp m) HMDB 12942 Estron e - 3 , 4 - qui no ne C18H20O3 285.148 4 0.00140 4 M+H [ 1+] 4.923 HMDB 12535 12S - HHT C17H27O3 280.203 2 0.00158 7 M+H [1+] 5.663 HMDB 00372 16 - Oxo es tro ne C18H20O3 285.148 4 0.00140 4 M+H [1+] 4.923 HMDB 01926 17a - E thy nyl es tra diol C20H24O2 297.178 8 0.00116 M+H [1+] 3.903 HMDB 00418 18 - Hydroxyc o rtisol C21H30O6 379.21 1 5 0.00250 2 M+H [1+] 6.597 HMDB 13651 2 - ( 14,15 - E poxye ic osa tri eno yl ) Gl yc erol C23H38O5 395.279 1 0.00167 8 M+H [1+] 4.245 HMDB 02641 2 - Hydroxyc i nna mic acid C9H8O3 165.054 6 0.00012 2 M+H [1+] 0.739 HMDB 06831 3 - Dehy dr oxyc a rni tin e C7H15N O2 146.117 5 0.00006 1 M+H [1+] 0.417 HMDB ID Commo n Na me Chemic a l Formul a T heo re tic a l adduc t m/z Experim en ta l er ror (ma s s units ) Adduc t Acc ura c y er ror (pp m) HMDB 00300 Urac il C4H4N 2O2 195.087 6 4 6 0.00006 1 M+2A C N +H [1+] 0.312 HMDB 00718 Val eric acid C5H10O2 144.101 8 9 8 0.00009 2 M+A C N +H [1+] 0.638 HMDB 13128 Val eryl c a rniti ne C12H23N O 4 144.101 8 9 8 0.00009 2 M+A C N+2H [2+] 0.638 HMDB 00378 2 - M ethy l buty royl c a rniti ne C12H23N O4 144.101 8 9 8 0.00009 2 M+A C N+2H [2+] 0.638 HMDB 02923 Isoma l tose C12H22O11 170.050 8 2 7 0.00148 M - 2H [2 - ] 8.703 HMDB 00740 La c tul ose C12H22O11 170.050 8 2 7 0.00148 M - 2H [2 - ] 8.703 HMDB 00258 Sucrose C12H22O11 170.050 8 2 7 0.00148 M - 2H [2 - ] 8.703 HMDB 11595 La c tosy l c era mide (d 18:1/ 24 :0) C54H103N O13 954.724 5 4 8 0.00103 8 M - H20 - H [1 - ] 1.087 HMDB 00319 18 - Hydroxyc o rtic ost er on e C21H30O5 383.183 9 9 0 0.00122 1 M+N a - 2H [1 - ] 3.186 HMDB 06758 17a ,21 - Dihy dro xy - 5 b - pr eg n a ne - 3 , 1 1 , 2 0 - t rio ne C21H30O5 383.183 9 9 0 0.00122 1 M+N a - 2H [1 - ] 3.186 HMDB 00063 Cortisol C21H30O5 383.183 9 9 0 0.00122 1 M+N a - 2H [1 - ] 3.186 HMDB 03441 Cinna ma l dehy d e C9H8O 167.026 9 1 7 0.00068 7 M+Cl [1 - ] 4.113 Table 2. Identification of metabolites in saliva after acidic hydrolysis (positive ionization) HM DB 02035 4 - Hydroxyc i nna mic acid C9H8O3 165.054 6 0.00012 2 M+H [1+] 0.739 HMDB 00530 6 - Keto es tr iol C18H22O4 303.152 1 0.00030 5 M+H [1+] 1.009 HMDB 01894 Aspa rta me C14H18N 2O5 295.123 0.00140 4 M+H [1+] 4.773 HMDB 01870 Benzoic acid C7H6O2 123.044 0 0.00006 1 M+H [1+] 0.417 HMDB 03449 Beta - D - G a l a c tose C6H12O6 181.070 6 0.00123 6 M+H [1+] 6.826 HMDB 00516 Beta - D - G l uc ose C6H12O6 181.070 6 0.00123 6 M+H [1+] 6.826 HMDB 01847 Ca ffein e C8H10N 4O2 195.087 6 0.00013 7 M+H [1+] 0.702 HMDB 00660 D - F ruc tose C6H12O6 181 .0706 0.00123 6 M+H [1+] 6.826 HMDB 00143 D - Ga l ac tose C6H12O6 181.070 6 0.00123 6 M+H [1+] 6.826 HMDB 00122 D - Gl uc ose C6H12O6 181.070 6 0.00123 6 M+H [1+] 6.826 HMDB 00169 D - M a nnose C6H12O6 181.070 6 0.00143 4 M+H [1+] 6.826 HMDB 12225 Enol - phenyl py ruv a te C9H8O3 165.054 6 0.00012 2 M+H [1+] 0.739 HMDB 00068 Epin ephr in e C9H13N O3 184.089 1 0.00044 3 M+H [1+] 2,419 HMDB 04483 Estron e gl uc uro nid e C24H30O8 447.194 1 0.00003 1 M+H [1+] 0,069 HMDB 12228 Ethy l phos pha te C2H7O4P 127.008 1 0.00009 2 M+H [1+] 0.730 HMDB 02497 Gl yc oc h enod eoxyc hol a te - 3 - s ul fa te C26H43N O8S 530.278 1 0.00061 M+H [1+] 1.150 HMDB 00201 L - A c etyl c a rnitin e C9H17N O4 204.123 0 0.00022 9 M+H [1+] 2.869 HMDB 03416 L - A rginin e C6H14N 4O2 175.118 9 0.00015 3 M+H [1+] 0.873 HMDB 00062 L - Ca rnitin e C7H15N O3 162.112 4 0.00160 2 M +H [1+] 9.943 HMDB 00687 L - L euc ine C6H13N O2 132.101 9 0.00079 3 M+H [1+] 6.002 HMDB 00162 L - Prol ine C5H9N O2 116.070 6 0.00089 3 M+H [1+] 7.693 HMDB 00158 L - Tyrosin e C9H11N O3 182.081 1 0.00140 4 M+H [1+] 7.753 HMDB 13133 Methy l gl uta ryl c a rniti ne C11H19N O6 262.128 5 0.00079 3 M+H [1+] 3.025 HMDB 00211 Myoinositol C6H12O6 181.070 6 0.00123 6 M+H [1+] 6.826 HMDB 02496 N - [(3 a ,5b,7a ) - 3 - hy droxy - 2 4 - oxo - 7 - ( s ul fooxy) c hol a n - 2 4 - yl ] - G l ycine C26H43N O8S 530.278 1 0.00061 M+H [1+] 1.150 HMDB 02284 N - A c etyl c a da verine C7H16N 2O 145.133 5 0,00085 4 M+H [1+] 5.884 HMDB 01406 Nia c ina mide C6H6N 2O 123.055 2 0.00000 8 M+H [1+] 0.065 HM DB 02117 Olea mid e C18H35N O 282.279 1 0.00247 2 M+H [1+] 8.757 HMDB 01860 Para xa nthi ne C7H8N 4O2 181.072 0 0.00021 4 M+H [1+] 0.564 HMDB 07885 PC(14:0/ 20:5( 5Z ,8Z ,11Z ,14 Z ,17 Z ) ) C42H74N O8P 752.522 4 0.00573 7 M+H [1+] 7.623 HMDB 08427 PC(x:a / y: b) C42H74N O8P 752.522 4 0.00573 7 M+H [1+] 7.623 HMDB 11273 PC(P - 18:1( 11Z ) / 16:1( 9Z ) ) C42H80N O7P 742.574 5 0.00390 6 M+H [1+] 5.332 HMDB 00205 Phenyl py ruv ic acid C9H8O3 165.054 6 0.00012 2 M+H [1 +] 0.739 HMDB 00020 p - Hydroxy phe nyl a c etic acid C8H8O3 153.054 6 0.00003 1 M+H [1+] 0.202 HMDB 04827 Prol in e beta in e C7H13N O2 144.101 9 0.00001 5 M+H [1+] 0.104 HMDB 00824 Prop ionyl c a rni tin e C10H19N O4 218.138 6 0.00189 2 M+H [1+] 8.673 HMDB 06088 Scyl l itol C6H12O 6 181.070 6 0.00123 6 M+H [1+] 6.826 HMDB 13075 Spe rmic acid C10H20N 2O4 233.149 5 0.00122 1 M+H [1+] 5.525 HMDB 01257 Spe rmi din e C7H19N 3 146.165 1 0.00102 2 M+H [1+] 6.992 HMDB 01256 Spe rmi ne C10H26N 4 203.223 0 0.00117 5 M+H [1+] 5.781 HMDB 02825 Theo br omi ne C7H8N 4O2 181.072 0 0.00010 7 M+H [1+] 0.590 HMDB 01889 Theo phyl l ine C7H8N 4O2 181.072 0 0.00010 7 M+H [1+] 0.590 HMDB 00289 Uric ac id C5H4N 4O3 169.035 6 0.00010 7 M+H [1+] 0.633 HMDB 12308 Vanil l in C8H8O3 153.054 6 0.00015 3 M+H [1+] 0.202 HMDB 04883 Trih exosy l c era mid e (d18:1/ 24:1( 15Z ) ) C60H111N O18 576.410 6 0.00006 1 M+H+N H4 [2+ ] 0.105 HMDB 10326 Thyrox in e gl uc uro nid e C21H19I4N O10 488.357 6 0.00024 4 M+H+N a [2+] 0.499 HMDB 13208 9 - Hexa d ec en oyl c hol ine C21H42N O2 363.310 7 0.00030 5 M+N a [1+] 0.839 HMDB 13582 PGP(18:3( 9Z ,12Z ,15Z ) / 18:2 ( 9Z ,12Z ) ) C42H74O13 P2 425.237 4 0.00051 9 M+2H [2+] 1.220 HMDB 13567 PGP(18:3( 6Z ,9Z ,12Z ) / 18:2( 9Z ,12Z ) ) C42H74O13 P2 425.237 4 0.00051 9 M+2H [2+] 1.220 HMDB 13553 PGP(18:2( 9Z ,12Z ) / 18:3( 6Z , 9Z ,12Z ) ) C42H74O13 P2 425.237 4 0.00051 9 M+2H [2+] 1.220 HMDB 13497 P GP(16: 1( 9Z ) / 20:4( 5Z ,8Z ,1 1Z ,14Z ) ) C42H74O13 P2 425.237 4 0.00051 9 M+2H [2+] 1.220 HMDB 00064 Crea ti ne C4H9N 3O2 195.085 2 0.00116 M+A C N +N a [1+] 5.946 HMDB 13222 Beta - G ua nidin opr opio nic aci d C4H9N 3O2 195.085 2 0.00116 M+A C N +N a [1+] 5.946 HMDB 00300 Urac il C4H4N 2O 2 195.087 6 0.00125 1 M+2A C N +H [1+] 6.410 HMDB 06027 Oxym es te ro ne C20H30O3 664.462 0 0.00128 2 2M+3H2O+2 H [2 +] 1.929 HM DB 08236 PC(18:4( 6Z ,9Z ,12Z ,15Z ) / 18: 1( 9Z ) ) C44H78N O8P 452.320 3 0.00146 5 M+3A C N+2H [2+] 3.238 HMDB 08235 PC(18:4( 6Z ,9Z ,12Z ,15Z ) / 18: 1( 11Z ) ) C44H 78N O8P 452.320 3 0.00146 5 M+3A C N+2H [2+] 3.238 HMDB 08204 PC(18:3( 9Z ,12Z ,15Z ) / 18:2( 9Z ,12Z ) ) C44H78N O8P 452.320 3 0.00146 5 M+3A C N+2H [2+] 3.238 HMDB 08171 PC(18:3( 6Z ,9Z ,12Z ) / 18:2( 9Z ,12Z ) ) C44H78N O8P 452.320 3 0.00146 5 M+3A C N+2H [2+] 3.238 HMDB 08141 PC(18:2( 9Z ,12Z ) / 18:3( 9Z ,12Z ,15Z ) ) C44H78N O8P 452.320 3 0.00146 5 M+3A C N+2H [2+] 3.238 HMDB 08140 PC(18:2( 9Z ,12Z ) / 18:3( 6Z ,9Z ,12Z ) ) C44H78N O8P 452.320 3 0.00146 5 M+3A C N+2H [2+] 3.238 HMDB 12230 Ga mma - gl uta myl - L - putres c i ne C9H19N 3O3 452.319 0 0.00271 6 2M+N H4 [1+] 6.073 HM DB 06478 Iso - Va l era l dehy d e C5H10O 104.106 9 0.00011 4 M+N H4 [1+] 1.095 HMDB 01272 Nic otine gl uc uro nid e C16H22N 2O6 181.072 1 0.00024 4 M+H+N a [2+] 1.347 HMDB 00826 Pen ta dec a no ic acid C15H30O2 144.101 5 0.00030 5 M+2N a [2+] 2.116 HMDB 01786 Ethen ode oxya de nosi ne C12 H13N 5O3 578.226 8 0.00030 5 2M+3H2O+2 H [2 +] 5.274 HMDB 10326 Thyrox in e gl uc uro nid e C21H19I4N O10 488.357 6 0.00042 7 M+H+N a [2+] 0.874 HMDB 02212 Ara c hidic acid C20H40O2 330.336 6 0.00054 9 M+N H4 [1+] 1.661 HMDB 10217 5 - KETE C20H30O3 664.462 0 0.00061 2M+3H2O+2 H [ 2 +] 0.918 HMDB 10212 17,18 - E pETE C20H30O3 664.462 0 0.00061 2M+3H2O+2 H [2 +] 0.918 HMDB 10210 15 - KETE C20H30O3 664.462 0 0.00061 2M+3H2O+2 H [2 +] 0.918 HMDB 10209 15 - HE PE C20H30O3 664.462 0 0.00061 2M+3H2O+2 H [2 +] 0.918 HMDB 10205 14,15 - E pETE C20H30O3 664.462 0 0.00061 2M+3H2O+2 H [2 +] 0.918 HMDB 10202 12 - HE PE C20H30O3 664.462 0 0.00061 2M+3H2O+2 H [2 +] 0.918 HMDB 05081 5 - HE PE C20H30O3 664.462 0 0.00061 2M+3H2O+2 H [2 +] 0.918 HMDB 13633 12 - KETE C20H30O3 664.462 0 0.00061 2M+3H2O+2 H [2 +] 0.918 HMDB 01337 Leuk otri en e A4 C20H30O3 664.462 0 0.00061 2M+3H2O+2 H [2 +] 0.918 HMDB 04029 11 - Dehy dr oc ortic ost e ron e C21H28O4 195.088 5 0.00088 5 M+2N a [2+] 4.536 HMDB 00700 Hydroxy pro pio nic acid C3H6O3 181.070 6 0.00123 6 2M+H [1+] 6.826 HMDB 00429 17a - E s tra diol C18H24O2 295.166 8 0.00167 8 M+ N a [1+] 5.684 Table 3. Identification of metabolites in saliva after basic hydrolysis (negative ionization) HMDB ID Commo n Na me Chemic a l Formul a Theo re tic a l adduc t m/z Adduc t Experim en ta l er ro r (ma s s units ) Acc ura c y er ror (pp m) HMDB 12135 1 - ( 3 - A mino pr opyl ) - 4 - a mino buta na l C7H16N 2O 125.107 9 M - H20 - H [1 - ] 0.0004 3.357 HMDB 11768 Cer( d18:0/ 2 4:0) C42H85N O3 686.622 4 M+Cl [1 - ] 0.0037 5.422 HMDB 11839 Ga ngl ioside GD2 (d18: 0/ 16: 0) C77H137N 3O34 548.295 5 M - 3H [3 - ] 0.0032 5.900 HMDB 12252 Linol eoyl etha nol a mid e C 20H37N O2 358.251 8 M+Cl [1 - ] 0.0027 7.430 HMDB 11514 Lys oPE( 20:3( 11Z ,14Z ,17Z ) / 0:0) C25H46N O7P 548.299 4 M+F A - H [1 - ] 0.0036 6.567 HMDB 11515 Lys oPE( 20:3( 5Z ,8Z ,11Z ) / 0:0) C25H46N O7P 548.299 4 M+F A - H [1 - ] 0.0036 6.567 HMDB 02284 N - A c etyl c a da verine C7H16N 2O 125.10 7 9 M - H20 - H [1 - ] 0.0004 3.357 HMDB 13561 PGP(18:2( 9Z ,12Z ) / 22:6( 4Z , 7Z ,10Z ,13Z ,16Z ,19Z ) ) C46H76O13 P2 298.484 8 M - 3H [3 - ] 0.0021 7.179 HMDB 13574 PGP(18:3( 6Z ,9Z ,12Z ) / 22:5( 4Z ,7Z ,10Z ,13Z ,16Z ) ) C46H76O13 P2 298.484 8 M - 3H [3 - ] 0.0028 9.524 HMDB 12342 PS( 14:1( 9Z ) / 14: 1( 9Z ) ) C34H62N O10P 710.380 6 M+Cl [1 - ] 0.0029 4.124 HMDB 06403 Pseudo ec go nyl - C oA C30H49N 8O18 P3S 971.158 4 M+K - 2H [1 - ] 0.0027 2.828 HMDB 10346 Triio dothy roni n e gl uc uro nid e C21H20I3N O10 807.803 8 M - H20 - H [1 - ] 0.0010 1.285 HMDB 12296 Trim ethy l a mi noa c eto ne C6H14N O 231.207 8 2M - H [1 - ] 0.0019 8.183 HMDB ID Commo n Na me Chemic a l Formul a Theo re tic a l adduc t m/z Adduc t Experim en ta l er ro r (ma s s units ) Acc ura c y er ror (pp m) HMDB 03424 1 - Hexa d ec a nol C16H34O 122.137 8 M+2H [2+] 0.0006 4.626 HMDB 02091 Isova l eryl gl uc uroni d e C11H18O8 140.057 4 M+2H [2+] 0.0012 8.825 HMDB 02916 4 - N itroc a tec hol C6H5N O4 140.058 0 M+3A C N+2H [2+] 0.0006 4.141 HMDB 04827 Prol in e beta in e C7H13N O2 161.128 5 M+N H4 [1+] 0.0008 4.778 HMDB 06009 Isoputr ea ni ne C7H16N 2O2 161.128 5 M+H [1+] 0.0012 7.261 HMDB 134 7 2 PGP(16:0/ 16:0) C38H76O13 P2 1627.94 1 5 2M+N a [1+] 0.0029 1.799 HMDB 11880 Ga ngl ioside GM1 (d18:0/ 18: 1( 9Z ) ) C74H132N 2O31 1627.94 1 8 M+2A C N +H [1+] 0.0027 1.649 Table 4. Identification of metabolites in saliva after basic hydrolysis (positive ionization) HM DB 02345 Hen eic osa no ic acid C21H42O2 164.166 5 M+2H [2+] 0.0001 0.091 HMDB 01352 Hydroxy pyruv i c acid C3H4O4 168.026 7 M+A C N +N a [1+] 0.0016 9.444 HMDB 00691 Ma l onic acid C3H4O4 168.026 7 M+A C N +N a [1+] 0.0016 9.444 HMDB 02780 Ca tec hin C15H14O6 168.028 7 M+2N a [2+] 0.0004 2.541 HMDB 00232 Quinol i nic acid C7H5N O4 168.029 1 M+H [1+] 0.0008 4.903 HMDB 11131 MG( 18:0/ 0:0/ 0:0) C21H42O4 180.161 4 M+2H [2+] 0.0009 5.167 HMDB 13076 Spe rmi ne dial dehy de C10H22N 2O2 225.157 3 M+N a [1+] 0.0021 9.149 HMDB 00392 2 - Octen oic acid C8H14O2 225.159 8 M+2A C N +H [1+] 0.0004 1.625 HMDB 05049 10 - N itrol inol eic acid C18H31N O4 225.159 8 M +3A C N+2H [2+] 0.0004 1.625 HMDB 11626 Dodec a nol C12H26O 225.161 5 M+K [1+ ] 0.0021 9.486 HMDB 00560 5,8 - Tet ra dec a di en oic acid C14H24O2 225.184 9 M+H [1+] 0.0017 7.522 HMDB 06954 2 - M ethy l - 3 - hy droxy - 5 - for m yl pyridin e - 4 - c a rboxyl a te C8H7N O4 226.008 7 M+2N a - H [1+] 0 .0014 6.344 HMDB 01024 Phosphohy d roxypy ruv ic aci d C3H5O7P 226.011 1 M+A C N +H [1+] 0.0010 4.455 HMDB 01358 Retina l C20H28O 326.247 8 M+A C N +H [1+] 0.0007 2.056 HMDB 03112 (3R ,3' R ,13 - c is) - b,b - C a roten e - 3 , 3 ' - diol C40H56O2 326.247 8 M+2A C N+2H [2+] 0.0007 2.056 HMDB 032 3 3 Lutein C40H56O2 326.247 8 M+2A C N+2H [2+] 0.0007 2.056 HMDB 07080 DG( 15:0/ 20:2( 11Z ,14Z ) / 0:0 ) C38H70O5 326.250 4 M+2N a [2+] 0.0019 5.799 HMDB 07416 DG( 20:2( 11Z ,14Z ) / 15:0/ 0:0 ) C38H70O5 326.250 4 M+2N a [2+] 0.0019 5.799 HMDB 07090 DG( 15:0/ 22:5( 4Z ,7Z ,10Z ,13 Z ,16Z ) / 0:0) C40H68O5 326.251 6 M+H+N a [2+] 0.0031 9.446 HMDB 07735 DG( 22:5( 7Z ,10Z ,13Z ,16Z ,19 Z ) / 15:0/ 0:0) C40H68O5 326.251 6 M+H+N a [2+] 0.0031 9.446 HMDB 11655 2 - ( 3 - C a rboxy - 3 - a minop ropy l ) - L - his tidine C10H16N 4O4 513.241 6 2M+H [1+] 0.0033 6.421 HMDB 10359 17 - h ydroxya n drosta ne - 3 - gl uc uronid e C25H40O8 513.243 5 M+2N a - H [1+] 0.0014 2.735 HMDB 10321 3,17 - A ndr osta ne diol gl uc ur onid e C25H40O8 513.243 5 M+2N a - H [1+] 0.0014 2.735 HMDB 02057 Prista noyl - C oA C40H70N 7O18 P3S 531.692 8 M+2H [2+] 0.0028 5.281 HMDB 12913 CoA - 20 - C OO H - L TE4 C44H64N 8O22 P3S2 607.646 7 M+2H [2+] 0.0009 1.405 HMDB 11153 MG ( P - 18:0e/ 0:0/ 0:0) C21H42O3 685.634 0 2M+H [1+] 0.0040 5.874 HMDB 04852 Ga ngl ioside GM3 (d18:1/ 25: 0) C66H122N 2O21 701.974 1 M+3A C N+2H [2+] 0.0030 4.260 HMDB 04913 Ga ngl ioside GD3 (d18: 1/ 16: 0) C68H121N 3O29 741.889 5 M+H+K [ 2+] 0.0029 3.867 HMDB 00941 Chol es t - 5 - en e C27H46 768.743 0 2M+3H2O+2 H [2 +] 0.0004 0.476 HM DB 07413 DG( 20:1( 11Z ) / 24:1( 15Z ) / 0:0) C47H88O5 774.697 0 M+A C N +H [1+] 0.0038 4.884 HMDB 07441 DG( 20:2( 11Z ,14Z ) / 24:0/ 0:0 ) C47H88O5 774.697 0 M+A C N +H [1+] 0.0038 4.884 HMDB 07805 DG( 24:0/ 20:2( 11Z ,14Z ) / 0:0 ) C47H88O5 774.697 0 M+A C N +H [1+] 0.0038 4.884 HMDB 07814 DG( 24:0/ 22:4( 7Z ,10Z ,13Z ,1 6Z ) / 0:0) C49H88O5 774.697 0 M+N H4 [1+] 0.0038 4.884 HMDB 08972 PE(16:1( 9Z ) / 20:5( 5Z ,8Z ,11 Z ,14Z ,17Z ) ) C41H70N O8P 81 8.544 3 M+2A C N +H [1+] 0.0024 2.982 HMDB 09096 PE(18:2( 9Z ,12Z ) / 18:4( 6Z ,9 Z ,12Z ,15Z ) ) C41H70N O8P 818.544 3 M+2A C N +H [1+] 0.0024 2.982 HMDB 09452 PE(20:5( 5Z ,8Z ,11Z ,14Z ,17Z ) / 16:1( 9Z ) ) C41H70N O8P 818.544 3 M+2A C N +H [1+] 0.0024 2.982 HMDB 11391 PE(P - 18:0/ 22:4( 7Z ,10Z ,13Z , 16Z ) ) C45H82N O7P 818.546 0 M+K [1+ ] 0.0042 5.144 HMDB 11704 TG(15:0/ 18:0/ 18:3( 9Z ,12Z , 15Z ) ) [iso6] C54H98O6 887.707 5 M+2N a - H [1+] 0.0025 2.818 HMDB 11708 TG(15:0/ 18:1( 9Z ) / 18:2( 9Z , 12Z ) ) [iso6] C54H98O6 887.707 5 M+2N a - H [1+] 0.0025 2.818 HMDB 12561 13' - Hydr oxy - ga ma - toc oph er ol C28H48O3 887.709 9 2M+N a [1+] 0.0001 0.137 HMDB 08546 PC(22:0/ 22:1( 13Z ) ) C52H102N O8P 944.705 4 M+2N a - H [1+] 0.0046 4.910 HMDB 08768 PC(24:0/ 20:1( 11Z ) ) C52H102N O8P 944.705 4 M+2N a - H [1+] 0.0046 4.910 HMDB 08750 PC(22:6( 4Z ,7Z ,10Z ,13Z ,16Z , 19Z ) / 24:1( 15Z ) ) C54H94N O8P 998.732 1 M+2A C N +H [1+] 0.0013 1.344 HMDB 08814 PC(24:1( 15Z ) / 22:6( 4Z ,7Z ,10Z ,13Z ,16Z ,19Z ) ) C54H94N O8P 998.732 1 M+2A C N +H [1+] 0.0013 1.344 344 Nuevas plataformas anal?ticas en metabol?mica 6. References [1] H.A. Soi ni , I. Klouc kova, D. Wi es ler, E. Oberzauc her, K. Gram mer, S. J. Di xon Y. Xu, R.G. Brereton D.J. Penn, M.V . Novotny, J. Chem. Ec ol. (2010) 36, 1035 ?1042. [2] P. del V. de Almei da, A.M. Gregio, A.M. Macha do, A.A. de Lim a, L.R. Azevedo, J. Contemp. Dent. Prac tice (2008). 9, 72 ? 80. [3] D. Fishe r, J.A. Shi p, Age and Agei ng (1 999) 28 (6), 557 - 561. E. Zussm an, A.L. Yari n,R.M. Nagler, J. Dental Res. 86 (3), 281 - 285. [4] M.J. Lars en, A.F. Jensen, D.M. Mads en, E.I.F. Pearc e, Arc h. Oral Biol. (1999) 44 (2), 111 - 117. [5] S.P. Hum phrey, R.T . Williams on, J. Pros thet. Dent. (2001), 85, 2, 162 - 169. [6] S. Chi appin, G. Antonelli , R. Gatti, E.F. De Palo, Cli n. Chi m. Acta (2007), 383, 1 - ?, ?????? [7] E. Kaufm an , I.B. Lams te, Cri t Rev Oral Biol Med. (2002), 13,1 97 ?2 1 2 . [8] C.J.L. Silwoo d, E. Lync h, A.W. D. Claxs on, M.C. Groo tveld, J. Dent. Res . (2002), 81(6), 422 - 427. [9] M. Sugim oto, D. T. Wong, A. Hirayama, T. Soga, M. Tomi ta, Metabolom ics (2010) 6, 78 ?9 5 . [10] D.S . Wisha rt, D. Tzu r, C. Knox et al. 2007, HMDB: the Human Metabolom e Databas e, Nuc lei c Aci ds Res. (2007), 35, (suppl 1): D521 ? D 5 2 6 . [11] B. ?lvarez - S ?nc hez, F . Priego - Ca pote, M.D. Luque de Cas tro, TrAC Trends Anal Chem . (2010) 29, 2, 111 ?1 1 9 . [12] J.C. Lindom , J.K. Nichols on, E. Holmes . The Handboo k of Metabonomi cs and Metabolomi cs . Els evier, (2007) 1 ? 3 3 . [13] T.N . Kolokolo va, O. ?u? Savel?ev, ??M? Sergeev, J. Anal. Chem . 63(2008) 104 ?120. 345 Sent to Journal of Chromatography A Chapter 9 [14] A. Giovane, A.Bales tri eri , C. Napoli , J. Cell.. Bioc hem., 105 (2008) 648 ? 654. [15] M. Betz, K. Saxena, H. Schwalbe, Curr Opin Chem Biol. 10 ( 2006) 219 ? 2 2 5 . [16] E.K. Kems ley, G. Le Gall, J .R. Dai nty, A.D. Wats on, L.J. Harvey, H.S. Tapp, I.J. Colquh oun, Br. J. Nu tr. 98 (2007) 1 ?1 4 . [17] M.P. Styczyns ki, J.F. Moxley, L.V . Tong, J.L. Walther, K.L. Jensen , G.N. Stephanopou los, Anal. Chem. 79 (2007) 966 ? 973. [18] B.S. Kri s tal, Y.I . Shu rubor, R . Kaddurah - Daouk, W.R. Mats on, Methods Mol. Biol. 358 (2007) 159 ?1 7 4 . [19] J.L. Capelo, P. Xim ?nez - Em b?n, Y. Madrid - Albarr?n, C. C?m ara, Anal. Chem . 76 (2004), 233 ?237 . [20] C. Pe?a - Farfal, A. Moreda - Pi?ei ro, A. Bermejo - Barrera, P. Berm ejo - Barrera, H. Pin oc he t - Canc i no, I. de Gregori - Henr? que z, Anal. Chem . , 76 (2004) 3541 ?3547. [21] M. M?guez - Fram il, A. Moreda - Pi ?ei ro, P. Bermejo - Barrera, P. L?pez, M.J. Tabernero, A.M. Berm ejo, Anal. Chem . , 79 ( 2007 ) 8564 ?8570. [22] B. ?lvarez - S?nchez, F. Priego - Ca pote, M. D. Luque de Cas tro, Analyst 134 (2009) 1416 - 14 2 2 . [23] http://m etli n.s c ri pps.edu / [24] http://www.hm db.ca/ [25] R. Touger - Dec ker, C. van Loveren, Am. J.Cli n. Nutr. 78 (2003) 881S ? 8 9 2 S . [26] I.D. Mande l, R. Thomps on Jr., S. Arch? Oral ?iol? ? (????) ???????? [27] S. Hui tron - Res endiz, L. Gom bart, B.F. Cravatt, S.J. Henri kse n, Exp. Neurol. 172 (2001) 235 ?243. 346 Nuevas plataformas anal?ticas en metabol?mica [28] B.L. Slom iany, A. Slomi any, Bioc hem. Biophys . Res . Comm un. 318 (2004), 247 ?52. [29] Y. - M. Che n, N.M. Cintron, P.A. Whi tson . Cli n. Chem. 38 (1992) 304 ?306 . [30] L.F. Hofman, J. Nutr.131 ( 2001) 1621S ? 1625S . [31] D. Perei ra Lima, D. Garc ?a Diniz, S. A. Sali ba Moim az, D. His sako Sum ida, A.C. Okam oto, Int. J. Infec t. Di s . 14 (2010) e184 ?e188. [32] T. Tan, L. Chang, A. Wood ward, B. McWhi nney, J. Galligan, G.A. Mac donald, J. Coh en, B. Venkates h, J. Hepatol. ?? (????) ???????? [33] E. Linkos alo, H. Markkanen, S. Syrjanen, J. Nutr , 115 (1985) 588 ? 592. [34] E. ?obo?y, ?? Czarkowska, M. Trojanowic z J. Biochem . Bioph. Metho ds , ?? (????) ?????? [35] A.K. Ghosha l, T. Guo, N. Soukho va, S.J. Soldin , Cli n. Chim . Ac ta , 358 (????) ???????? [36] Y. Tanaka, N. Narui shi , H. Fukuya, J. Sakata, K. Sai to, S. Waki da, J. Chromatogr? A ???? (????) ???????? [37] M.L. Hannuks ela, M.K. Liis anantti, A.E.T . Nis sinen, M.J. Savolai nen, Cli n. Chem. Lab. Med. 45 (2007) 953 ?6 1 . [38] M . Venza, M. Vis alli , D. Cicc iu, D. Teti , J. Chro m atogr. B, 757 (2001), ???????? CHAPTER 10: Comparison of sample preparation protocols for metabolomics profiling of human breast milk by high- accuracy LC?TOF/MS and 1H-NMR C omparison of sa mp le prep aration protocols for metabolomics p rofi ling of human breast milk by hi gh - accuracy LC ?TOF/MS and 1 H - NMR B. ?lvarez S?nchez* , F. Priego Capote, M. D. Luque de Castro Department of Analytical Chemistry, Annex C-3 Building, Campus of Rabanales, University of C?rdoba, C?rdoba, Spain Institute of Biomedical Research Maim?nides (IMIBIC), Reina Sof?a Hospital, University of C?rdoba, E-14071, C?rdoba, Spain Sent to Metabolomics for publication 351 Sent to Metabolomics Chapter 10 Compa rison of sample prepara tion proto co ls for Metabo lomics Pro fili ng of human breast milk by hig h - acc??acy ????????S and 1?- NMR B. ?lvarez-S?nchez*, F. Priego -Capote, M.D. Luque de Castro Abst ra ct Breas t milk is the mai n or unique sou rc e of feeding for chi ldren at early growing stages . Milk metabolomi cs cou ld be of great interes t to ass ess its nutri ti onal valu e, to identify bioma rkers of certai n patholo gies or to monitor its vari abili ty as a cons equence of lac tation state, food, supplem entati on, drugs intake or circ adi an vari ations. However, a com plete charac teri zati on of maternal milk has so far not been reported, fac t that cou ld be asc ri bed to the lac k of opti mi zed standard operati on protoc ols . In this study, a com paris on between different analyti c al sam ple preparation protocols has been carri ed out, aimi ng at obtai ni ng a repres entati ve metabolic profile of human breas t milk. The influence of a centri fu gati on step for phas es separation, pH adjus tment and liqui d ? li qui d extrac ti on on the metabolomic s coverage has been tes ted. The study reveale d that ac i di fi cati on reported a negati ve effec t on the coverage of polar metaboli tes, but i t seems to posi ti vely affec t the coverage of non - polar metaboli tes . On the other hand, centri fu gati on proved a signi fi cant influ enc e on coverage as milk is com pos ed by a fat - ri ch phas e sus pended in an aqueous serum . Finally, analysi s by LC ?TOF/MS and 1 H - N M R platforms has been used for identific ation of 148 compounds, inc ludi ng sugars , glyc eri des, fatty aci ds and anti oxi dants, am ong others . 352 Nuevas plataformas anal?ticas en metabol?mica 1. Intro ductio n Metabolomi cs is defi ned as the dis ci pli ne responsi ble for the analysis of low molec ular - wei ght com p ou nds in biolo gi c al sam ples under certain phys iolo gic al conditi ons (Wis ha rt et al. 2007). Therefo re, metabolomic s can be referred to biologic al flui ds , tiss ues, cells or whole organis ms . With regards to human metabolomi cs , biolo gi c al flui ds suc h as plas m a, serum and uri ne have been com monly us ed, as they are eas i ly and minim ally or non - i nvasi vely collec ted. Additi onally, these bioflui ds direc tly reflec t the global state of an indi vi dual and/or the res pons e to drug treatment, and usually requi re the applic at ion of simple sam ple preparati on protocols ( Li ndon et al. 2007) . In contras t to the vas t literature devoted to urine and blood metabolom ic s , there is a gap in studi es on the analysis of other biolo gic al flui ds , such as sali va, maternal milk, sem en or cereb ral spinal flui d (?lvarez - S ?nche z et al. 2010; Gibney et al. 2005). Partic ul arly, metabolomi cs of maternal milk cou ld be of great interes t to assess its nutri ti onal valu e, to identify biom arkers of certai n dis eas es or sim ply to moni tor its variabili ty as a cons equence of lac tati on state, food, supplem entati on and drugs intake or circ adi an vari ati ons (Mi toul as et al. 2002). Maternal milk is the primary or unique sou rc e of feedi ng for children during the early stages of growing, whic h are marked by a rapid develo pm ent and conti nuou s intake of nutrients (Sha h 2000). Therefore, its hi gh nutri ti onal value shoul d reflec t a ric h and varied com pos i ti on, inc ludi ng proteins (cas ei ns or imm unoglobul i ns ) (Mitou las et al. 2002), calcium phos pha te, citrate, minor ions, lipovitam i ns , antioxi dants (Canfi eld et al. 2003), and lipi ds (such as cho les terol and phos pholi pi ds ) ( Hamosha et al. 1992) , hydros oluble vitam i ns (Pi cc iano et al. 1995) and sugars (Co ppa et al. 2004). Whereas the imm unoprotec ti ve func tion of certai n prote i ns ( i.e. 353 Sent to Metabolomics Chapter 10 i mm unoglobuli ns ) and peptides agai ns t virus and bac teri a has been widely studi ed (Sha h, N.P. 2000; Van de Perre P . 2003; Palm ei ra et al. 2009), scant attenti on has been paid to low - m olec ular weight metaboli tes whi c h are also es senti al in the deve lopm ent of the indi vi dual. In thi s sense, lipi ds (Jens en R.G. 1999; Anderson et al. 1999; Blaas et al. 2011) and oli gosacc hari des (Kunz et al. 2000) are by far the bes t - c harac terized frac ti ons. Target analysis of antioxi dants (Canfield et al. 2003), lipovi tamins (Ti jeri na - S?enz et al. 2009; Kam ao et al. 2007), and hydros oluble vitam i ns (Hoppu et al. 2005) have als o rec ently been reported. However, sinc e the bes t of our knowledge, charac teri zati on of maternal milk has not been develo ped so far, whi ch cou ld b e attri buted to the lac k of optim ized protoc ols . If a com plete eluci dati on of the metabolom e or, more reali s ti cally, the wideni ng of the metaboli te coverage is intended, the ideal method of analys is sho uld allow coveri ng the larges t range of metaboli tes r egardless thei r chemi c al charac teristi cs and differenc es in conc entration. In this sense, the mos t feas i ble approac h is to integrate the inform ati on obtained by different analytic al methods , for ins tanc e, com pari ng the resul ts provi ded by different sam ple preparati on protoc ols , different analyti cal techni ques, or applyi ng different stati s tical treatm ents (?lvarez - S?nch ez et al. 2010b). In general term s, the fi nal goal of metabolom ics woul d be to obtain a com plete, reproduci ble and reli able insi ght of the m etaboli c com pos i ti on, whic h is usua lly acc om plis he d by applic ati on of a sui table protoc ol based on the use of sensi ti ve and acc urate analytic al equi pm ent, usually mass spec trom etry (MS) or nuc lear magneti c res onance spec trometry (NMR) (Dieterle et al. 2011 ). The form er can be cons i dered as the preferred choi ce for metabolom ic s analysis . In fac t, accurate mass spec trom etry analyzers com bi ned with complete mas s spec tra databas es and hi gh - res olution chrom atography offer enorm ous capabi li ty for identi fic ati on a nd provide extraordi nary informati on about biolo gic al proc es s es taking plac e in the sam ple. On the other hand, NMR - bas ed metabolom ic s is als o widely used due to its versati li ty, the pos si bi li ty of direc t analysis or with very low 354 Nuevas plataformas anal?ticas en metabol?mica s am ple preparation require m ents and the recent introduc ti on of hi gh - fi eld NMR analyzers with improved res olution capabili ty ( Vi ant et al. 2003; Lin et al. 2007). The mai n drawbac ks of thi s tec hni que are the com plexi ty for interpretati on and assignm ent of spec tral inform ati on, the n arrower dynam ic range and the lower sens i ti vi ty as com pared to mas s spec trom etry techni ques . Despi te the rapi d advances in analyti c al ins trum entati on, the major cause of errors in metabolomi cs analys is is still linked to sam ple preparati on steps due to s am ple handli ng and analys t interventi on. For this reas on, different sam ple preparati on protoc ols are usua lly tested to ass es s thei r influ enc e on metabolomic s coverage. The aim of the present study was to com pare different analyti c al sam ple preparati on prot oc ols to obtai n a com plete metabolic profi li ng of human breas t milk. Addi ti onally, identific ation of the maxi mum num ber of extrac ted molec ul ar features or potenti al metaboli c peaks is als o address ed. Therefo re, LC ?T OF/MS and unidim ens ional nuc lear magnetic resonance ( 1 H - N MR) are used to obtai n a com plem entary profi le of thes e sam ples . Informati on obtai ned from the present study would cover the exis ti ng gap in metabolo mi cs experim ents on breas t milk, whic h cou ld be of interes t in the ass es sm ent of the nutri t i onal state of breas tfed neonates and infants or in the study of milk com pos i ti on vari abili ty. 2. Ma terials and methods 2.1. Reagents and samples LC ? MS grade acetonitri le, MS - grade form ic aci d (Scha rlab, Barc elo na, Spai n) and dei onized water (18 M??cm) from a Millipore Milli- Q water purifi cati on sys tem (Mi llipore, Bedford, MA, USA) were used to prepare the 355 Sent to Metabolomics Chapter 10 c hrom atographi c phases . Chloroform and methanol from Scha rlab were used for the charac terizati on of the non - polar frac tion. Sam ples were kindl y donated by fi ve volu nteers, in the firs t term of breas tfeedi ng. No one referred any dis eas e or health disorder. Each sam ple corres ponded to a who le intake. After collec ti on, all sam ples were pooled, ali quo ted in 2 - m L Eppendorf tubes, and stored at ?80 ?C until proc es si ng. 2.2. Instruments and apparatus Centrifugation was carri ed out with a thermos tated centri fu ge Thermo Sorvall Legend Mic ro 21 R from Thermo (T he rmo Fis he r Sci enti fic, Brem en, Germ any) . A Vac ufuge centri fu gal v a c u u m c o n c e n t r a t o r from Ep pendorf ( Eppendorf , Inc ., Hamburg, Germany ) was us ed t o evaporate sam ples to dryness . All sam ples were analyzed by an 1200 Seri es HPLC sys tem (Agi lent Tec hnologi es , Waldbronn, Germ any) equi pped with a binary pum p, a degass er, a well plate autos am pler and a therm os tated colum n com partm ent, whi ch was cou pled to an Agi lent 6530 TOF mass spec trom eter equi pped with a Jetstream ? ESI sou rc e . Mass Hunter Works tation Data ac quis iti on software (Agi lent Tec hnologi es ) was used to control the ins trum ent. Data were proce ss ed usi ng Mass Hunter Quali tati ve Analysis software (Agi lent Tec hnologi es ). Extrac ti on of unknown features from raw data was carri ed out with the Molec ular Featu re Extrac tion (MFE) algori thm in Mas s Hunter Quali tati ve analys is software. Analys es were proce ss ed usi ng MassHunter Quali tati ve software and com pou nd identific ation was perform ed usi ng the METLIN Personal Metaboli te Database, the Hum an Metabolom e Databas e and the Molecul ar Form ula Generation algori thm (Agi lent Tec hnolo gies ). 1 H - NMR spectroscopy wa s performed at 500.13 MHz using a Bruker AMX500 spectrometer 11.7 T (Bruker BioSpin GmbH, Rheinstetten, Germany). 356 Nuevas plataformas anal?ticas en metabol?mica 2.3. Sample preparation protocols The scheme followed for optimi zati on of sam ple preparation is sho wn in Figure 1. Diffe rent experim ental p rotoc ols were com pared in order to obtain the maxim um metaboli te coverage. For eac h analys is , 100 ?L of pooled milk were us ed. In a firs t study, the effec t of centri fugatio n was tested, for whic h an aliquo t was 1:1 mixed with metha nol for protei n precipita ti on and direc tly analyzed (1). This study was com pared with other that inc lu ded a final centri fu gati on step at 12000 g for 10 min, with cons tant temperature at 6 ?C (2). Charac teri zati on of the polar phase started with a preconce ntrati on step, whic h was carri ed out by mixi ng an ali quo t with pure metha nol, then evaporati ng to drynes s and recons ti tuti ng in mobi le phas e A (0.1% formic ac i d in water) (3). Afterwards, the effec t of pH on the polar phas e was evaluated. With this aim, ac i di fi cati on of sam ples wa s carried out by mixi ng 1:1 an aliquo t with methanol 10% (v/v) ac idi fi ed with formi c aci d. The polar phase was then centri fuged and reconsti tuted in mobile phase A (4). Finally, the non - pola r phas e was als o charac teri zed, for whic h an aliquot was 1:2 mixed with 50% methanol ?chl oroform . Then, the extrac t was centri fu ged and the polar and non - polar phas es were separated, evaporated and recons ti tuted in the chrom atographic mobi le phase A (0.1% formic aci d in water) (5) and in 50% (v/v) ac etonitri le ?H 2 O soluti o n 0.1% aci difi ed with formic aci d (6), respec tively. The same experiment was repeated with a 10:40:50 (v/v/v) form ic aci d ?m etha nol ?chl oroform mi xture (7 and 8). 2.4. Liquid chromatography ?mass spectrometry 357 Sent to Metabolomics Chapter 10 Chrom atographi c separati on was carri ed out with a Zorbax Ec li pse XDB - C18 colum n (4.6?150 mm, 5 ?m partic le size), in whi ch 10 ?L of sam ple was injec ted. Mobi le phas es , deli vered at 1 mL/m i n, cons is ted of 0.1% formic ac id in water (A) or in ac etonitri le (B). The chrom atographic gradi ent was as follows : 2% B for 5 min followed by a gradient to 100% B in 55 min and an isoc rati c step at 100% B for 10 min. The ESI Jetstream source operated in pos i ti ve and negati ve ionizati on usi ng the followi ng condi ti ons : needle voltage set at +/ ? 4 kV, nebuli zer gas at 40 psi , dryi ng and sheath gas flow rate and tem perature at 8 and 11 L/m in and 325 and 350 ?C, respec ti vely. Data were collec ted in both centroi d and profi le mode at a rate of 1 spec trum per s in the extended dynami c range mode (2 GHz). Accurate mas s spec tra were adqui red in the m/z range 60 ?1100. The ins trument gave typic al res olution 15000 FWHM (Ful l Width at Half Maxim um ) at m/z 121 and 30000 FWHM at m/z 922. To ass ure the des ired mass accurac y of recorded ions , conti nuous inter nal cali brati on was perform ed during analys es with the use of signals at m/z 121.0509 (protonated puri ne) and m/z 922.0098 [protonated hexakis (1H, 1H, 3H - tetrafluoro - propoxy) - phos pha zi ne or HP - 921] in pos i ti ve ion mode; in negati ve ion mode ions with m/z 11 9.0362 (proton abstrac ted puri ne) and m/z 966.000725 (formate adduc t of HP - 921) were used. 2.5. MS data processing Onc e samples were analy?ed by LC?TOF/MS, data were extracted i nto features and the calc ulated neutral mas s was queri ed agai ns t the MET LI N and HMDB databas es of known metaboli tes. Features extrac ti on was based on the molec ular feature extrac ti on algori thm (MFE) that loc ates Figure 1. Optimization of sample preparation steps for global metabolic profiling of milk by LC?TOF/MS. 359 Sent to Metabolomics Chapter 10 and groups all ions related to the sam e neutral molec ule. Thi s relati on is referred as to the covariance of peaks withi n the same chromatographic retenti on tim e, the charge - s tate envelo pe, isotopi c dis tribution, and/or the presenc e of adduc ts and dimm ers . The MFE took into acc ount all ions exc eedi ng 1000 cou nts, with charge state lim i ted to a maxim um of two and a peak spaci ng tolerance of 0.0025 m/z (plus 7 ppm). Eac h feature was given by a minim um of two ions. The extrac ti on algori thm was bas ed on a com mon organic model with chrom atographi c separati on. The allowed posi ti ve ions were protonated speci es and sodi um adduc ts , and the negati ve ions form ed by form ate adduc ts and proton losses . Dehydratati on neutral loss es were also allowed. Molec ul ar form ulas were generated usi ng the Molec ul ar Formul a Generator algori thm . Both mas s accurac y and isotope inform ati on (abundanc e and spaci ng) were consi dered to lim i t the num ber of hits. Only the com mon elements C, H, N, O, P and S were cons i dered in the generation of form ulas . The calc ulated neutral mass of eac h feature was quer i ed agai ns t the MET LIN databas e for matchi ng to com pounds withi n a maxi mum mass tolerance window of 10 ppm . The MET LIN databas e matche d the neutral mas s to the monois otopic mass valu e calc ul ated from the empi ric al formul a of com pou nds in the databas e. Identi fied compounds were eventually contras ted agai ns t the Hum an Metabolom e Databas e (HMDB) to confi rm thei r occ urrence in hum ans . 2.6. 1H -NMR spectrometry analysis of milk samples: data acquisition and processing Milk serum, obtained after centrifugation at 12000 g for 10 min at 4 ?C (400 ?L) was mixed w ith with 150 ?L of 0.1 mM solution of trimethylsilyl propionate (TSP), used as chemical shift reference, in deuterium water (D 2 O). The sample was examined at 4 ?C to minimize metabolic changes. No gross 360 Nuevas plataformas anal?ticas en metabol?mica degradation was noted in the signals of multiple spec tra acquired under the same conditions Standard solvent suppressed spectra were grouped into 16 00 0 data points, averaged over 256 acquisitions. The data acquisition lasted in total 13 min using a sequence based on the first increment of the Nuclear Over hauser Effect Spectroscopy (NOESY) pulse sequence to effect suppression of the water resonance and limit the effect of B 0 and B 1 inhomogeneities in the spectra (relaxation delay - 90? - t1 - 90? - tm - 90? - acquire Free Induction Decay (FID) signal), in which a secon dary radio frequency irradiation field was applied at the water resonance frequency during the relaxation delay of 2 s and during the mixing period (tm = 150 ms), with t1 fixed at 3 s. The acquisitions were performed using a spectral width of 8333.33 Hz. P rior to Fourier transformation, the FID signals were multiplied by an exponential weight function corresponding to a line broadening of 0.3 Hz. Spectra were referenced to the TSP singlet at 0 ppm chemical shift. 3. Results and discus sion 3.1. Characterization of milk serum by LC ? TOF/MS 3.1.1. Influence of centrifugation Milk is consti tuted by an aqueous serum with colloi dal parti c les formed by lipids and cas ei n mic elles . This heterogeneous com posi ti on makes diffic ul t to obtain a com plete profi le of met aboli tes as a charac teris tic fi ngerprinti ng. Thus , the mai n drawbac k found in the analysi s of untreated milk is attri buted to sample ins tabi li ty sinc e it is rapi dly separated into two phases leadi ng to irreprodu ci ble res ul ts . In additi on, metaboli c profi li ng analys is is conditi oned by the partiti on of com pounds between both phas es ac cordi ng to thei r parti tion coeffi ci ent. 361 Sent to Metabolomics Chapter 10 An alternative to avoid this lac k of preci si on is the inc lusi on of a centrifu gation step for sam ple frac ti onati on. The influe nc e of mil k centrifu gation was evaluated in two ali quo ts that were subjec ted to a deprotei ni zati on step with metha nol in an 1:1 (v/v) ratio. The hi gh content of proteins makes mandatory a step for thei r removal sinc e they are clear respons i ble for ioniz ati on suppres s ion phenomena with los s of sens i ti vi ty. After that, one of the ali quo ts was centri fuged and the resul tant serum was analyzed in both posi ti ve and negati ve ioni zati on modes while the other ali quo t was direc tly analys ed . Non - c entri fu ged aliquo t sho uld give a larger num ber of molec ular features as it contai ns both polar and non - polar metaboli tes . However, a simi lar num ber of molec ular featu res were obtained. Thus , algori thm for extrac ti on of molec ul ar features allowed the isolation of 20 molec ular features in the serum phas e obtai ned after centrifu gation in the negati ve ioni zati on mode versus 26 molecul ar features detected after direc t analys is of the non - c entri fu ged ali quo t. This resul t can be jus ti fi ed by lower ionizati on effi ci ency in the non - polar phas e, p robably caus ed by ionization suppress i on of lipidi c molecul es. Owi ng to this , both phases shoul d be analyzed separately and sam ple preparati on opti mi zed independently. Figure 2 sho ws a Venn diagram repres enti ng the molecul ar features obtai ned from both ali quots , in negati ve (2A) and posi ti ve (2B) ionization. As can be seen in Fig 2.A, 18 molecul ar features were commonly detected in both sam ple preparation tes ts, whi le 2 and 8 exc lusi ve featu res were detected in negati ve mode for the centri fu ged and non - c ent ri fu ged sam ples , res pec ti vely. On the other hand, 24 molec ular features were com monly detec ted in pos i ti ve mode, whi le 10 and 19 unique featu res were detected in the centri fuged and non - centri fuged sam ples , res pec ti vely. Therefo re, a total of 53 molecul ar features were detec ted in posi ti ve ionization (49% of them in comm on for the two ali quo ts ) while 28 molec ular features were detected in negati ve mode (64% in comm on for the two ali quots ). 362 Nuevas plataformas anal?ticas en metabol?mica Figure 2. Venn diagram of molecular features obtained from the analysis of a centrifuged and a non-centrifuged aliquots, in negative (2A) and positive (2B) ionization modes. 3.1.2. Characterization of the polar phase The aforem entioned protoc ol impli es dilu ti ng ali quo ts in metha nol, which mi ght led to a dec rease i n the coverage. Therefore, a prec oncentration step was carried out, for whic h an aliquot was 1:1 mixed with metha nol and subsequently centri fuged. The polar phas e was then evaporated to dryness and reconsti tuted in 100 - ?L mobile phase A (0.1% form ic aci d i n water) to enhance ioni zati on and improve chrom atographic separati on. Analysi s of these ali quo ts in posi ti ve and negati ve ioni zati on modes provi ded the base peak chromatogram s (BPCs ) illus trated in Fig ure 3. The conc lus ion of thi s test is that prec oncentr ati on inc reased the num ber of detec ted molec ul ar features from 20 up to 60 by negati ve ionization. One of the mai n drawbac ks of milk metabolic profi li ng is the large dynam ic range at whic h metaboli tes can be found and the pres enc e of mac rom olec ular units , mai nly protei ns , whi ch might interfere not only in elec tros pray ionizati on but als o in chrom atographi c res olution. As res ul ted in the previ ous sec ti on, these drawbac ks can be mini mi zed after protei n removal by prec i pi tati on with an organic solvent such as metha nol. (A) (B ) 363 Sent to Metabolomics Chapter 10 Figure 3. Base peak chromatograms obtained without (grey line) and with preconcentration (black line) in positive (3A) and negative (3B) ionization. (A) (B) 364 Nuevas plataformas anal?ticas en metabol?mica Protei n preci pitati on c an be ai ded by sam ple ac i di fi cati on since casei ns and othe r protei ns preci pitate at pH below 5. Additi onally, pH ac i di fi cati on could inc reas e the solu bi li ty of metaboli tes with ionizable groups . With thes e premis es, the follo wi ng test involved the ass ess me nt of pH aci di fic ation on the deprotei ni zati on process usi ng metha nol with 10% (v/v) form ic aci d. This tes t was com pared with that of the sam e process without aci dific ation. The two ali quo ts were evaporated and reconsti tuted i n 100 ? L mobile phas e A. Figure 4 shows the Ven n diagram s for negati ve (A) and pos i ti ve ( B) ioni zation modes with the molec ular featu res extrac ted from the LC ?T OF/MS analysis of the aci di fi ed sam ple (A), a non - aci di fied sam ple (NA). In negati ve mode, 48 molecul ar features were comm on to both the aci di fied and non - acidi fied ali quo ts , whi le the total num ber of detec ted features was 6 0 (NA) and 65 (A), res pec ti vely, for both analys is . Acc ordi ngly, the total num ber of detected features rem ai ned prac tic ally cons tant with the decrease of pH, thus demonstrati ng to be non - c ruci al for ioni zati on. Resul ts differ signific antly in positi ve mode , as aci di fi cati on leads to a decreas e from 49 to 22 uni que features . Figure 4. Venn diagrams for negative (A) and positive (B) ionization modes with the molecular features extracted from the LC?TOF/MS analysis of the acidified sample (A), a non-acidified sample (NA). (A) (B) 365 Sent to Metabolomics Chapter 10 This dec reas e can be due to the fac t that aci di fi c ati on favors hydrolysi s and metaboli te co - prec i pi tation, thus dec reasi ng solu bi li ty. Globally, the total of molec ular enti ti es detec ted in posi ti ve mode was 118 and 104 for A and NA, res pec ti vely. Bas e peak chrom atograms from thes e experim ents are illus trated in Figure 5 reveal i ng the differences in the metabolic profile as a resul t of the aci d medium . (A) (B) Figure 5. Base peak chromatograms obtained from the non-acidified (black line) and acidified (grey line) aliquots in positive (A) and negative (B) following the experimental protocols (3) and (4), respectively. 366 Nuevas plataformas anal?ticas en metabol?mica 3.1.3. Characterization of the non-polar phase The mai n limi tation of the previ ous sam ple preparati on strategi es is that metha nol is not able to solu bi lize the non - polar frac tion com pletely and, for this reas on, characteri zati on of lipi ds is not sui ted. Aimi ng at charac teri zi ng the n on - pola r cons ti tue nts, a methanol ? chl orofo rm mixture was added to a breas t milk ali quot and the mixture was shaken for liqui d ? li quid extrac tion. The two resul ti ng phases were subsequently analyzed after centrifu gation, evaporation and rec ons ti tuti on as des c ri bed in the experim ental sec tion. In negati ve mode, a total of 27 and 35 molec ul ar features were extrac ted from the polar and non - pola r phas e, res pec ti vely, bei ng only 5 of them com mon to both phas es , whic h is a logic al behavior taking into acc ou nt diffe renti al solubi li ty between phases . The sam e experim ent was repeated by usi ng 10:40:50 form ic aci d ?m ethanol ? c hl oroform (7 and 8 in Fig ure 1), whi ch led to a cons i derable inc reas e in the num ber of unique features in the non - pola r phase (chl oroform ). In fac t, an inc reas e from 35 to 94 uni que featu res in the non - pola r phase was obtai ned, agai ns t a dec reas e from 27 to 17 features in the polar phas e. Therefore, the effec t of pH was more acute on the non - pola r phas e than on the polar one, whic h cou ld be linked to the releas e of metaboli tes by hydrolys is . In order to know whi ch metaboli tes were affec ted by pH, thi s experi ment was com pleted with a charac teri zati on of the non - pola r phas e by identi fic ati on of the extrac ted features with MET LIN and HMDB resources . There fore, 94 com pounds were identi fi ed in the non - polar phas e by meas urement of ac curate mass and a maxim um mass diffe renc e of 10 ppm . Thes e com pounds are lis ted in supplementary Tables 1 and 2 for pos i ti ve and negati ve ioni zati on modes, res pec ti vely. 3.2. Identification of metabolites in breast milk analysis 367 Sent to Metabolomics Chapter 10 Extrac ted molec ular features from each experi ment were searc hed agai ns t the MET LIN and HMDB databases supported on meas urem ent of ac curate mas s . The databas e searc h res ul ts are expos ed in Tables 1 and 2 that inc ludes the HMDB code, comm on nam e, chem ic al formul a, m/z ratio and detec ted adduc t, acc urac y error expres s ed as ppm and, finally, the analys is in whi ch the metaboli te was identi fied. Thus, polar frac ti on corres ponded to the analys is of the sam ple a fter methanol deprotei ni zati on in 10% form ic acid (1:1 v/v ratio), evaporati on of the supernatant phas e and recons ti tuti on in mobile phase A. On the other hand, the non - polar frac ti on corres ponded to the analys is of the sam ple after liqui d ?li qui d extrac tio n with metha nol ?chl orofo rm in 10% form ic aci d, evaporation of the non - polar phase and rec ons ti tuti on in ac etoni trile ?H 2 O (50:50 v/v) 0.1% acidi fi ed with formic aci d. In total, 148 metaboli tes were identified, 105 of them in posi ti ve ionization and 48 in ne gati ve mode (5 of them were comm on to both ionization modes ). Conc erni ng the frac tion where they were detec ted, 94 metaboli tes were identi fied in the non - polar phas e, while 85 were identi fied in the polar phas e. These resul ts mean that a 63% of total featu res were identifi ed in the non - polar phase whi le 57% of them were found in the polar phase. Thus, it is worth mentioning the com plementari ties of both sam ple preparation strategies. As can be seen, the non - polar phas e was marked by the pres enc e of mono - , d i - and tri glyc eri des , glyc erophos pholi pi ds suc h as glyc erophos pho etanolam ine, glyc erophos phoc holi ne, glyc erophos phos eri ne and phos phati dyli nosi tol metaboli tes . Spec ial menti on des erves the identific ation of carni ti nes for trans port of fatty aci ds , ganglios i des , steroi d horm ones suc h as aldos terone, inflamm atory medi ators suc h as leukotri ne E4, and metaboli tes of phytyl diphos pha te as an interm edi ate in the bios ynthesi s of steroi ds. Dis ac cha ri des and polys acc hari des were als o detected in thi s phas e with inte res t of lac tose deri vati ves that were pres ent at hi gh conce ntrati on. On the other hand, the polar phase was charac teri zed by the pres enc e of free fatty aci ds and derivati ves such as HETEs (hydroxyeic os atetraenoic ac i ds ), othe r carboxyli c ac i ds such as citr i c acid 368 Nuevas plataformas anal?ticas en metabol?mica and pantothe ni c aci d, monos acc ha ri des such as gluc os e and fruc tos e with thei r corres pondi ng phosphate i nterm edi ates , among others. 3.3. 1H -NMR spectrometry analysis of breast milk Charac teri zati on of breas t milk was com pleted by unidim ens ional 1 H - N MR analys is of breas t milk and charac teristi c cross - peaks from 2D spec tra to help in unequivoc al assi gnation of the metaboli tes . In this cas e, an analyti cal strategy bas ed on direc t analys is of milk was em ployed. The NMR spec tra were marked by the pre s enc e of major signals corres ponding to lac tos e and lac tos e - bas ed polysacc hari des . Thus , Fig ure 6.A illus trates a charac teris tic spec trum for breas t milk where, apart from the main disacc ha ri des , other carboh ydrates such as gluc os e and fuc os e, and carboxyl ic aci ds suc h as citric aci d and glutam ic acid were identifi ed. Figure 6.B shows a zoom in y - axis in whi ch minor signals cou ld be assi gned. In this way, other monosacc harides suc h as xylos e and arabinose, ami no aci ds suc h as creati ne and alanine and signal s corres pondi ng to fatty ac i ds were identifi ed. The res onanc es were identi fied with the Chenom x NMR software ac cordi ng to the Human Metabolom e Database. Altho ugh it is evident that the identific ation capability of NMR cannot be compared with that reported by MS in the analys is of com plex sam ples , the NMR spec tra provi ded by breas t milk consti tute a charac teris ti c fi ngerpri nting that cou ld be used as a fas t snapsho t view of its metabolic state. 4. Conc lusio ns The pres ent study aims at testi ng diffe rent sam ple preparation strategies for analys is of the metabolic profile of hum an breas t mi lk. A sui ted 369 Sent to Metabolomics Chapter 10 optim i zation of the experi mental protoc ol has been plan ned as a func ti on of the metaboli te coverage in order to maxi mi ze the num ber of identi fi ed metaboli tes . The variables taken into accou nt were the dilu ti on solvent, pH, centrifu gation and liqui d ?li quid extrac tion. It can be concluded that the bes t. strategy shou ld be the com bi nati on of a dual protocol: a deprotei ni zati on with metha nol, evaporati on and rec onsti tuti on in ac idi fi ed water for analys is of the polar phas e, and liqui d ? li qui d extrac tion with metha nol ? c hl oroform for analys is of the non - polar phas e On the other hand, centrifu gation leads to more repeatable analyses as sam ple stabili ty is marked by the pres ence of two phases that progress i vely separate over tim e. Therefo re, it is advi sable to centri fu ge the sam ple and independently analyze the res ul ti ng phas es in an appropri ate solvent. In addi ti on, it is hi ghl y rec om mendable to carry out a solvent exc ha nge to the ini ti al chrom atographi c mobile phas e to favor elec tros pray ionization as the coverage is inc reased up to three ti mes. Figure 6A. Spectrum for breast milk obtained by 1H-NMR. 370 Nuevas plataformas anal?ticas en metabol?mica Figure 6B. Spectrum zoomed in y-axis. 5. Ackno wledgement s The Spanish Mi nis terio de Ciencia e Innovac i?n (MICI NN ) and FEDER program are thanked for fi nanci al support through projec t CTQ2009 - 07430. B.A. - S. and F.P. - C. are also grateful to the MIC INN for an FPI scho larshi p (BES - 2 0 0 7 - 15043) and a Ram?n y Cajal cont rac t (RYC - 2009 - 03921). Table 1. Metabolites identified in the characterization of the non-polar and polar phases by LC?TOF/MS analysis in positive ionization mode. HM DB ID Commo n Na me Chemic a l Formul a Theo re tic a l mas s m/z Adduc t Experim en ta l er ror (ma s s units ) Acc ura c y er ro r (ppm) Non - pol a r pha s e Pola r pha s e HMDB 06478 Iso - Va l era l dehy d e C5H10O 104.107 0 M+N H4 [1+] 0.0002 1.8347 X HMDB 00076 Dihydrou ra c il C4H6N 2O2 132.076 8 M+N H4 [1+] 0.0005 3.4677 X HMDB 00064 Crea ti ne C4H9N 3O2 132.076 8 M+H [1+] 0.0005 3 .4677 X HMDB 10715 2 - Phenyl a c eta mi de C8H9N O 136.075 7 M+H [1+] 0.0003 2.2414 X HMDB 03052 La c ta l dehy de C3H6O2 138.052 6 M+A C N +N a [1+] 0.0010 7.5696 X HMDB 01392 p - A mino be nzoic acid C7H7N O2 138.055 0 M+H [1+] 0.0003 1.8761 X X HMDB 04827 Prol in e beta in e C7H 13N O2 144.101 9 M+H [1+] 0.0003 0.0000 X HMDB 00005 2 - Keto buty ric acid C4H6O3 144.065 5 M+A C N +H [1+] 0.0002 1.4854 X HMDB 01259 Succ inic acid se mia l dehy d e C4H6O3 144.065 5 M+A C N +H [1+] 0.0002 1.4854 X HMDB 00826 Pen ta dec a no ic acid C15H30O2 144.101 5 M+2N a [ 2+] 0.0002 1.3740 X HMDB 00148 L - Gl uta mic acid C5H9N O4 148.060 4 M+H [1+] 0.0003 2.2693 X HMDB 02931 N - A c etyl s erin e C5H9N O4 148.060 4 M+H [1+] 0.0003 2.2693 X HMDB 02035 4 - Hydroxyc i nna mic acid C9H8O3 165.054 6 M+H [1+] 0.0006 3.7927 X HMDB 00660 D - F ruc tos e C6H12O6 181.070 7 M+H [1+] 0.0009 0.0000 X HMDB 00143 D - Ga l ac tose C6H12O6 181.070 7 M+H [1+] 0.0009 5.1416 X HMDB 00122 D - Gl uc ose C6H12O6 181.070 7 M+H [1+] 0.0009 5.1416 X HMDB 02825 Theo br omi ne C7H8N 4O2 181.072 0 M+H [1+] 0.0002 1.1819 X HMDB 00469 5 - H ydroxy methy l u ra c il C5H6N 2O3 184.071 7 M+A C N +H [1+] 0.0009 5.1393 X HMDB 06831 3 - Dehy dr oxyc a rni tin e C7H15N O2 184.073 4 M+K [1+ ] 0.0005 2.8195 X HMDB 00159 L - Phenyl a l a nine C9H11N O2 188.068 2 M+N a [1+] 0.0020 10.7089 X HMDB 01565 Phosphoryl c hol in e C5H15N O4P 2 02.107 7 M+N H4 [1+] 0.0009 4.4531 X HMDB 00355 3 - Hydroxy methy l gl uta ric aci d C6H10O5 204.086 6 M+A C N +H [1+] 0.0005 2.6165 X HMDB 00201 L - A c etyl c a rnitin e C9H17N O4 204.123 0 M+H [1+] 0.0004 2.0919 X X HMDB 13227 cis - 5 - Dec en edioic acid C10H16O4 218.138 7 M+N H4 [ 1+] 0.0004 1.7466 X X HM DB 00824 Prop ionyl c a rni tin e C10H19N O4 218.138 7 M+H [1+] 0.0004 1.7466 X X HMDB 12150 2 - Keto - 6 - a c eta mi doc a proa t e C8H13N O4 220.118 0 M+CH3OH+H [1+] 0.0000 0.2090 X HMDB 00210 Pantoth enic acid C9H17N O5 220.118 0 M+H [1+] 0.0000 0.2090 X X HMDB 02013 Buty ryl c a rniti ne C11H21N O4 232.154 3 M+H [1+] 0.0001 0.5901 X HMDB 09452 PE(20:5( 5Z ,8Z ,11Z ,14Z ,17Z ) / 16:1( 9Z ) ) C41H70N O8P 246.168 6 M+3H [3+] 0.0015 6.1990 X HMDB 13128 Val eryl c a rniti ne C12H23N O4 246.170 0 M+H [1+] 0.0001 0.4956 X HMDB 00982 5 - M ethy l c ytidin e C10H15N 3O5 258.108 4 M+H [1+] 0.0013 4.9669 X HMDB 00086 Gl yc erophosph oc hol ine C8H20N O6P 258.110 1 M+H [1+] 0.0005 1.8907 X X HMDB 07975 PC(16:0/ 18:3( 9Z ,12Z ,15Z ) ) C42H78N O8P 260.183 4 M+2H+N a [3+] 0.0011 4.2239 X HMDB 08006 PC(16:1( 9Z ) / 18:2( 9Z ,12Z ) ) C42H78N O8P 260.183 4 M+2H+N a [3+] 0.0011 4.2239 X HMDB 08017 PC(16:1( 9Z ) / 20:5( 5Z ,8Z ,11Z ,14Z ,1 7Z ) ) C44H76N O8P 260.184 2 M+3H [3+] 0.0003 1.0569 X X HMDB 08142 PC(18:2( 9Z ,12Z ) / 18:4( 6Z ,9Z ,12Z ,1 5Z ) ) C44H76N O8P 260.184 2 M+3H [3+] 0.0003 1.0569 X HMDB 0 82 0 6 PC(18:3( 9Z ,12Z ,15Z ) / 18:3( 9Z ,12Z , 15Z ) ) C44H76N O8P 260.184 2 M+3H [3+] 0.0008 2.9325 X HMDB 00705 Hexa n oyl c a rniti ne C13H25N O4 260.185 6 M+H [1+] 0.0009 3.5206 X X HMDB 00455 Al l oc ys ta thionine C7H14N 2O4S 261.030 6 M+K [1+ ] 0.0006 2.2220 X HMDB 00124 Fruc t ose 6 - ph osph a te C6H13O9P 261.037 0 M+H [1+] 0.0004 1.5209 X HMDB 01401 Gl uc ose 6 - phos pha te C6H13O9P 261.037 0 M+H [1+] 0.0004 1.5209 X HMDB 08051 PC(18:0/ 22:0) C48H96N O8P 282.903 0 M+3H [3+] 0.0012 4.1003 X HMDB 00573 El aidic a c id C18H34O2 283.263 2 M+H [1+ ] 0.0006 2.2629 X HMDB 00207 Oleic acid C18H34O2 283.263 2 M+H [1+] 0.0006 2.2629 X HMDB 01235 5 - A minoim ida zol e ri bonuc l e otide C8H14N 3O7P 296.064 2 M+H [1+] 0.0011 3.7120 X HMDB 06038 3' - O - M ethy l gua nosin e C11H15N 5O5 298.114 6 M+H [1+] 0.0000 0.0000 X HMD B 09740 PE(24:0/ 24:1( 15Z ) ) C53H104N O8P 312.917 9 M+2H+N a [3+] 0.0014 4.4868 X HMDB 03797 Bovinic acid C18H32O2 313.273 7 M+CH3OH+H [1+] 0.0001 0.2937 X HMDB 00673 Linol eic acid C18H32O2 313.273 7 M+CH3OH+H [1+] 0.0001 0.2937 X HMDB 00480 7,10 - Hexa d ec a die noi c acid C16H28O2 313.274 3 M+Is oPro p+H [ 1+] 0.0003 0.8778 X HM DB 00186 Al pha - L ac tose C12H22O11 343.123 5 M+H [1+] 0.0009 2.5792 X X HMDB 00163 D - M a l tose C12H22O11 343.123 5 M+H [1+] 0.0009 2.5792 X X HMDB 00740 La c tul ose C12H22O11 343.123 5 M+H [1+] 0.0009 2.5 792 X HMDB 06603 3 - b - G a l a c topyra nosy l gl uc ose C12H22O11 365.105 4 M+N a [1+] 0.0005 1.5037 X HMDB 03947 8Z,11Z ,14Z - eic osa tri en oyl - C oA C44H72N 7O17 P 3 S 366.137 9 M+3H [3+] 0.0003 0.9177 X HMDB 01557 Ribofl a vin re duc ed C15H16N 4O6 366.140 8 M+N H4 [1+] 0.0026 7.0 847 X HMDB 06323 Tet ra c osa pen ta en oic acid (2 4:5n - 3) C24H38O2 381.276 4 M+N a [1+] 0.0038 9.9246 X HMDB 12356 PS( 16:0/ 18:0) C40H78N O10P 391.288 7 M+H+N H4 [2+ ] 0.0036 9.2029 X HMDB 08939 PE(16:0/ 20:5( 5Z ,8Z ,11Z ,14 Z ,17Z ) ) C41H72N O8P 391.739 0 M+2N a [2+] 0.0037 9.5038 X HMDB 09095 PE(18:2( 9Z ,12Z ) / 18:3( 9Z ,1 2Z ,15Z ) ) C41H72N O8P 391.739 0 M+2N a [2+] 0.0037 9.5038 X HMDB 09191 PE(18:4( 6Z ,9Z ,12Z ,15Z ) / 18: 1( 9Z ) ) C41H72N O8P 391.739 0 M+2N a [2+] 0.0039 9.9709 X HMDB 09451 PE(20:5( 5Z ,8Z ,11Z ,14Z ,17Z ) / 16:0) C41H72N O8P 391.73 9 0 M+2N a [2+] 0.0039 9.9709 X HMDB 01397 Gua nosin e mo no phos pha te C10H14N 5O8P 427.073 8 M+A C N +N a [1+] 0.0040 9.3614 X HMDB 12380 PS( 18:0/ 18:2( 9Z ,12Z ) ) C42H78N O10P 456.315 2 M+3A C N+2H [2+] 0.0023 5.1499 X HMDB 09765 PE(24:1( 15Z ) / 22:0) C51H100N O8P 462.844 5 M +H+K [ 2+] 0.0031 6.7906 X HMDB 11593 La c tosy l c ermide (d18:1/ 20: 0) C50H95N O13 478.825 4 M+H+K [ 2+] 0.0001 0.2548 X HMDB 12642 20 - Oxo - l euk ot rie ne E4 C23H34N O6S 491.173 9 M+K [1+ ] 0.0043 8.8217 X HMDB 10034 PIP2(16:0/ 16:2( 9Z ,12Z ) ) C41H77O19 P3 506.202 8 M+2N a [2+] 0.0038 7.4753 X HMDB 07852 LPA( 0:0/ 18:2( 9Z ,12Z ) ) C21H39O7 P 511.162 4 M+2K+H [1+] 0.0004 0.7767 X HMDB 02235 O - 6 - deoxy - a - L - ga l a c topyra nosy l - ( 1 - > 2 ) - O - b - D - ga l a c topyra nosy l - ( 1 - > 3 ) - 2 - ( a c etyl a mino) - 1 , 5 - a nhy dro - 2 - deoxy - D - a ra bi no - H ex - 1 - e nitol C20H33N O14 512. 197 4 M+H [1+] 0.0010 1.9075 X X HMDB 01064 Linol eoyl - C oA C39H66N 7O17 P3S 524.193 0 M+H+N H4 [2+ ] 0.0034 6.5205 X HMDB 00121 Fol ic acid C19H19N 7O6 524.200 1 M+2A C N +H [1+] 0.0038 7.2186 X HMDB 06620 Fuc osy l la c tose C18H32O15 527.137 3 M+K [1+ ] 0.0013 2.5477 X X HMDB 11913 Ga ngl ioside GM3 (d18:0/ 12: 0) C53H98N 2O21 572.322 3 M+2N a [2+] 0.0047 8.3187 X HMDB 11484 Lys oPE( 0:0/ 20:3( 11Z ,14Z ,1 7Z ) ) C25H46N O7P 587.356 3 M+Is oPro p+N a +H 0.0015 2.5981 X [1+] HMDB 06569 6' - Sia l yl la c tose C23H39N O19 634.218 9 M+H [1+] 0.0019 3.079 4 X HMDB 11801 Ga ngl ioside GD1a (d18:1/ 23 :0) C90H159N 3O39 636.358 9 M+3H [3+] 0.0007 1.0544 X X HMDB 11765 Cer( d18:0/ 2 2:0) C40H81N O3 646.610 8 M+N a [1+] 0.0019 3.0204 X HMDB 02430 (3a ,5b,7a ) - 2 3 - C a rboxy - 7 - hy droxy - 2 4 - norc hol a n - 3 - y l - b - D - G l uc opyra nosidu r onic a cid C30H48O10 647.346 0 M+DM SO+H [1+] 0.0019 2.9227 X HMDB 06834 D - Pa ntoth enoyl - L - c ys tein e C12H22N 2O6S 667.228 9 2M+N a [1+] 0.0026 3.8413 X HMDB 01373 Dephos pho - C oA C21H35N 7O13 P2S 705.182 7 M+N H4 [1+] 0.0003 0.4325 X HMDB 06537 (a - D - ma nnosy l ) 2 - b - D - ma nn osy l - N - a c etyl gl uc osa mine C26H45N O21 730.237 6 M+N a [1+] 0.0019 2.5909 X HMDB 13647 Aden osin e thia m in e di phos p ha te C22H30N 9O10 P2S 738.146 9 M+A C N +N a [1+] 0.0014 1.9021 X HMDB 07324 DG( 18:3( 9Z ,12Z ,15Z ) / 22:6( 4Z ,7Z ,1 0Z ,13Z ,16Z ,19Z ) / 0:0) C43H66O5 739.410 0 M+2K+H [1 +] 0.0010 1.4038 X HMDB 07577 DG( 20:5( 5Z ,8Z ,11Z ,14Z ,17Z ) / 20:4( 8Z ,11Z ,14Z ,17Z ) / 0:0) C43H66O5 739.410 0 M+2K+H [1+] 0.0010 1.4038 X HMDB 07744 DG( 22:5( 7Z ,10Z ,13Z ,16Z ,19 Z ) / 18: 4( 6Z ,9Z ,12Z ,15Z ) / 0:0) C43H66O5 739.410 0 M+2K+H [1+] 0.0010 1.4038 X HMDB 06566 La c to - N - tetra ose C26H45N O21 746.211 5 M+K [1+ ] 0.0018 2.3720 X X HMDB 10580 PG(16:0/ 20:4( 5Z ,8Z ,11Z ,14 Z ) ) C42H75O10 P 803.543 3 M+CH3OH+H [1+] 0.0002 0.2277 X HMDB 10621 PG(18:1( 11Z ) / 18:3( 6Z ,9Z ,1 2Z ) ) C42H75O10 P 803.543 3 M+CH3OH+H [1+] 0.0002 0.2277 X HMDB 07941 PC(15:0/ 18:3( 6Z ,9Z ,12Z ) ) C41H76N O8P 818.450 0 M+2K+H [1+] 0.0012 1.4918 X HMDB 08935 PE(16:0/ 20:3( 5Z ,8Z ,11Z ) ) C41H76N O8P 818.450 0 M+2K+H [1+] 0.0012 1.4918 X HMDB 11852 Ga ngl ioside GD2 (d18: 1/ 14: 0) C75H131N 3O34 818.451 3 M+H+N H4 [2+ ] 0.0026 3.1315 X HMD B 02211 Uropo rphy rino ge n I C40H44N 4O16 854.309 0 M+N H4 [1+] 0.0003 0.3570 X HMDB 06705 La c to - n - fuc openta os e I C32H55N O25 854.313 6 M+H [1+] 0.0020 2.0718 X X HMDB 06576 La c to - N - fuc openta ose III C32H55N O25 854.313 6 M+H [1+] 0.0020 2.0718 X X HMDB 06706 La c to - N - fuc openta ose V C32H55N O25 854.313 6 M+H [1+] 0.0020 2.0718 X X HMDB 06577 La c to - N - fuc openta ose - 2 C32H55N O25 854.313 6 M+H [1+] 0.0020 2.0718 X X HMDB 06805 Beta - A l a nyl - C oA C24H41N 8O17 P3S 861.141 5 M+N a [1+] 0.0023 2.6221 X HM DB 10422 TG(16:1( 9Z ) / 14:0/ 18:2( 9Z , 12Z ) ) [i s o6] C51H92O6 864.705 1 M+A C N +N a [1+] 0.0008 0.9876 X X HMDB 09493 PE(22:0/ 20:0) C47H94N O8P 864.705 2 M+CH3OH+H [1+] 0.0001 0.1411 X HMDB 06487 Pen ta gl uta myl fola te C39H47N 11O18 999.343 9 M+A C N +H [1+] 0.0023 2.2595 X HMDB 06589 Sial yl l a c to - N - tetra ose a C37H62N 2O29 999.351 1 M+H [1+] 0.0049 4.9472 X HMDB 10033 PIP2(16:0/ 16:1( 9Z ) ) C41H79O19 P3 1007.40 5 9 M+K [1+ ] 0.0021 2.0597 X HMDB 06602 La c to - N - hexa ose C40H68N 2O31 1073.38 7 9 M+H [1+] 0.0019 1.8195 X HMDB 06532 Pal mitol eyl CoA C37H64N 7O17 P3S 1080.24 8 3 M+2K+H [1+] 0.0002 0.2259 X Table 2. Metabolites identified in the characterization of the non-polar and polar phases by LC?TOF/MS analysis in negative ionization mode. HM DB ID Commo n Na me Chemic a l Formul a Theo re tic a l mas s m/z Addu c t Experim en ta l er ror (ma s s units ) Acc ura c y er ro r (ppm) Non - pol a r pha s e Pola r pha s e HMDB 00444 3 - F uroic acid C5H4O3 111.008 8 M - H [1 - ] 0.0039 6.7177 X X HMDB 01259 Succ inic acid se mia l dehy d e C4H6O3 161.045 5 M+Ha c - H [1 - ] 0.0042 7.9327 X HMDB 02825 Theo br om i ne C7H8N 4O2 161.046 3 M - H20 - H [1 - ] 0.0027 9.5082 X X HMDB 12883 Adre noc hro me o - s e miqui no ne C9H10N O3 161.047 7 M - H20 - H [1 - ] 0.0027 9.7266 X X HMDB 03441 Cinna ma l dehy d e C9H8O 167.026 9 M+Cl [1 - ] 0.0007 2.5677 X HMDB 00641 L - Gl uta mine C5H10N 2O3 181.038 5 M+Cl [ 1 - ] 0.0028 9.8205 X HMDB 00094 Citric acid C6H8O7 191.019 7 M - H [1 - ] 0.0047 8.9247 X HMDB 01014 4 - Imida zol one - 5 - p ropi onic a cid C6H8N 2O3 191.022 9 M+Cl [1 - ] 0.0015 2.0308 X HMDB 00210 Pantoth enic acid C9H17N O5 218.103 4 M - H [1 - ] 0.0049 6.3102 X X HMDB 02396 Trim ethy l tr id ec a noic acid C16H32O2 255.233 0 M - H [1 - ] 0.0019 8.5372 X HMDB 00220 Pal mitic acid C16H32O2 255.233 0 M - H [1 - ] 0.0023 2.9002 X HMDB 03797 Bovinic acid C18H32O2 279.233 0 M - H [1 - ] 0.0021 9.5138 X HMDB 06270 Linoel aidic a c id C18H32O2 279.233 0 M - H [1 - ] 0.0021 3.5406 X HM DB 11711 TG(15:0/ 18:2( 9Z ,12Z ) / 18:2 ( 9Z ,12Z ) ) [iso3] C54H96O6 279.233 0 M - 3H [3 - ] 0.0021 4.1397 X HMDB 10737 (R ) - 3 - Hydroxy - Octa dec a noi c acid C18H36O3 281.248 0 M - H20 - H [1 - ] 0.0016 9.3159 X HMDB 00573 El aidic a c id C18H34O2 281.248 6 M - H [1 - ] 0.0021 4.0829 X HMDB 00207 Oleic acid C18H34O2 281.248 6 M - H [1 - ] 0.0021 4.3844 X HMDB 03529 Inos itol 1, 3,4,5,6 - p en ta k isph osph a te C6H17O21 P5 288.940 2 M - 2H [2 - ] 0.0013 1.7249 X HMDB 11563 MG( 15:0/ 0:0/ 0:0) C18H36O4 297.243 0 M - H20 - H [1 - ] 0.0038 10.56 86 X HMDB 07578 DG( 20:5( 5Z ,8Z ,11Z ,14Z ,17Z ) / 20:5( 5Z ,8Z ,11Z ,14Z ,17Z ) / 0:0) C43H64O5 329.230 4 M - 2H [2 - ] 0.0042 7.9326 X HMDB 07353 DG( 18:4( 6Z ,9Z ,12Z ,15Z ) / 22:6( 4Z ,7 Z ,10Z ,13Z ,16Z ,19Z ) / 0:0) C43H64O5 329.230 4 M - 2H [2 - ] 0.0042 8.2795 X HMDB 04710 9,10,13 - Tr iHOM E C18H34O5 329.233 4 M - H [1 - ] 0.0012 2.0834 X HMDB 10222 9 - HE TE C20H32O3 355.204 5 M+Cl [1 - ] 0.0009 2.2632 X HMDB 03876 15(S) - HE TE C20H32O3 355.204 5 M+Cl [1 - ] 0.0009 3.1467 X HMDB 00037 Al dostero n e C21H28O5 381.168 3 M+N a - 2H [1 - ] 0.0012 2.5700 X HMDB 03164 Chl oroge nic acid C16H18O9 391.043 7 M+K - 2H [1 - ] 0.0015 3.6910 X HMDB 06701 3 - O - a - L - F uc opyra nosy l - D - gluc ose C12H22O10 405.040 2 M+B r [1 - ] 0.0019 1.9192 X HMDB 02200 Leuk otri en e E4 C23H37N O5S 420.220 9 M - H20 - H [1 - ] 0.0042 4.3029 X HMDB 00215 N - A c etyl - D - gl uc osa mine C8H15N O6 487.178 1 2M+F A - H [1 - ] 0.0009 0.8977 X HMDB 11116 Phytyl diphos pha te C20H42O7 P2 501.238 8 M+F A - H [1 - ] 0.0004 2.6514 X HMDB 12501 10,11 - Dihy dr o - 1 2 R - hy droxy - l euk otri en e E4 C23H38N O6S 501.240 2 M+F A - H [1 - ] 0.0010 6.0665 X HMDB 04270 D - Gl uc osa minide C18H35N 3O13 522.191 7 M+N a - 2H [1 - ] 0.0027 5.2605 HMDB 11164 L - beta - a s pa rtyl - L - gl uta mic acid C9H14N 2O7 523.153 0 2M - H [1 - ] 0.0007 4.0433 X X HMDB 00585 Gl uc osy l gal a c tosy l hydroxyl ys ine C18H34N 2O13 523.154 7 M+K - 2H [1 - ] 0.0010 1.0512 X X HMDB 03351 GD P - gl uc ose C16H25N 5O16 P2 586.058 8 M - H20 - H [1 - ] 0.0013 8.3393 X HMDB 01391 GDP - 4 - Dehy dr o - 6 - L - deoxyg a l a c tose C16H23N 5O15 P2 586.059 3 M - H [1 - ] 0.0008 2.9052 X HM DB 13617 Lipoyl - G M P C18H26N 5O9P S2 586.060 4 M+Cl [1 - ] 0.0002 0.8939 X HMDB 01260 ADP - R ibosy l - L - a rg inin e C21H35N 9O15 P2 736.147 5 M+N a - 2H [1 - ] 0.0021 2.7435 X HMDB 09280 PE(20:1( 11Z ) / P - 1 8 : 0 ) C43H84N O7P 778.573 2 M+N a - 2H [1 - ] 0.0044 5.6449 X X HMDB 09249 PE(20:0/ P - 18:1( 9Z ) ) C43H84N O7P 778.573 2 M+N a - 2H [1 - ] 0.0044 5.6449 X HMDB 09491 PE(22:0/ 18:3( 9Z ,12Z ,15 Z ) ) C45H84N O8P 778.575 1 M - H20 - H [1 - ] 0.0039 5.0169 X X HMDB 09232 PE(20:0/ 20:3( 5Z ,8Z ,11Z ) ) C45H84N O8P 778.575 1 M - H20 - H [1 - ] 0.0039 5.0169 X HMDB 09106 PE(18:2( 9Z ,12Z ) / 22:1( 13Z ) ) C45H84N O8P 778.575 1 M - H20 - H [1 - ] 0.0039 5.0169 X HMDB 11391 PE(P - 18:0/ 22:4( 7 Z ,10Z ,13Z , 16Z ) ) C45H82N O7P 778.575 6 M - H [1 - ] 0.0020 5.1238 X HMDB 03324 Biotri pyrr in - b C25H27N 3O6 929.372 7 2M - H [1 - ] 0.0029 8.7143 X HMDB 10466 TG(18:1( 9Z ) / 18:1( 9Z ) / 22:1( 13Z ) ) C61H112O6 985.844 1 M+F A - H [1 - ] 0.0024 7.2290 X HMDB 05468 TG(18:2( 9Z ,12Z ) / 20: 0/ 20:1 ( 11Z ) ) C61H112O6 985.844 1 M+F A - H [1 - ] 0.0024 5.6637 X HMDB 05457 TG(18:1( 9Z ) / 20:1( 11Z ) / 20:1( 11Z ) ) C61H112O6 985.844 1 M+F A - H [1 - ] 0.0024 5.9069 X 378 Nuevas plataformas anal?ticas en metabol?mica 6. References 1. ?lvarez - S ?nche z B. , Pri ego - Capote F. , Luque de Castro , M.D. (2010a) Metabolomi cs analys is I. Selec ti on of biological s amples and prac ti cal as pec ts prec edi ng sam ple preparati on . Trends Anal . Chem . , 29 ( 2 ): 111 ?119. 2. ?lvarez - S ?nche z B. , Pri ego - Capote F. , Luque de Cas tro M.D. (2010b) Metabolomi cs analys is II. Preparation of biologi ca l sam ples prior to detecti on, Trends Anal . Chem . , 29 ( 2 ): ???????? 3. Anders on J.W. , Joh ns tone, B.M., Rem ley D.T. (1999) Breas tfeedi ng and cogni ti ve develo pm ent: A meta analys i s. Am . J . Cli n . Nutr . , 70 ( 4 ): 525 ?535. 4. Blaas N . , Sch?? rm ann C . , Bartke N . , Stahl B . , Hum pf H . U . (2011) Struc tural profili ng and quanti fi c ati on of sphi ngom yeli n in hum an breas t milk by HPLC - MS /MS . Agric . Food Chem . , 59(11): 6018 ?60 2 4 . 5. Ca nfield L.M., Clandi ni n M.T ., Davies D.P., Fernandez M.C., Jacks on J., Hawkes J., Goldm an W.J., Pram uk K., Reyes H., Sablan B. , S onobe T., Bo X. (20 03) Multi nati onal study of major breas t milk carotenoi ds of health y mothe rs . Eu r . J . Nutr . , 42: 133 ?141. 6. Co ppa G.V . , Bruni S., Morelli L., Soldi S., Gabri elli O. (2004) T he First Prebio ti cs in Hum ans , Human Milk Oli gos acc ha ri des . J. Cli n . Gastroenter . , 38 : S80 ?S 83. 7. Di eterle F . , Riefke B . , Schlotterbec k G . , Ross A . , Senn H . , Amberg A . (2011) NMR and MS metho ds for metabonom ic s. Methods Mol . Biol. , 691 : ???????? 8. Gibney M.J., Walsh M., Brennan L., Roche H.M., German B., van Omm en B. (2005) Metabolom ic s in hum a n nutri tion: opportuni ti es and challenges . Am . J . Cli n . Nutr . , 82, 497 ? 503. 379 Sent to Metabolomics Chapter 10 9 . Ham os ha M., Bitm an J. (1992) Hum an milk in dis eas e: Lipid Com posi ti on . Lipi ds , 27 : 848 ?8 5 7 . 10. Hoppu U . , Rinne M . , Salo - V ??n?nen P . , Lam pi A . M . , Pii ronen V . , Is olau ri , E. ( 2005) Vitam in C in breas t milk may reduce the ris k of atopy in the infant. Eur . J . Cli n . Nutr. , 59(1): 123 ?1 2 8 . 11. Jens en R.G. (1999) Lipids in hum an milk . Lipi ds , 34 ( 12 ): 1243 ?1271 . 12. Kamao M . , Tsugawa N . , Suh ara Y . et al. (2007) Quanti fic ati on of fat - s olu ble vitam i ns in human breas t milk by liquid chrom atography - tandem mas s spec trom etry . J . Chrom atogr . B , 859 (2): 192 ?2 0 0 . 13. Kunz C., Rudloff S., Baier W., Klei n N., Strobel S. (2000) O li gosacc hari des in h um an m ilk : Struc tural , Functi onal, and Metaboli c Aspec ts . Ann . Rev . Nutr . , 20: 699 ?7 2 2 . 14. Li n C.Y., Wu H., Tjeerdem a R.S., Viant M.R. (2007) Evalua tion of metaboli te extrac ti on strategi es from tiss ue sam ples usi ng NMR metabolom ic s . Metabolom ic s , 3 ( 1 ) : 55 ?67 . 15. Li ndon J.C., Nicho ls on J.K., Holm es E. (2007) The Handboo k of Metabonom ics and Metabolo mi cs , p p. 1 ? 33, Elsevi er, Ams terdam , The Netherlands, (IS BN:0 - 4 4 4 - 52841 - 5). 16. Mi toulas L.R., Kent J.C. , Cox D.B., Owens R.A., Sherri ff J.L., Hartm ann P.E. (2002) Vari ati on in fat, lac tose and protei n in hum an milk over 24h and throu ghout the firs t year of lac tati on, Br . J. Nutr . , 88 : 29 ? 37. 17. Palm ei ra P . , Cos ta - Ca rvalh o B.T ., Ars lania n C ., Pontes G.N ., Nagao A . T . , Carnei ro - S am pai o M.M . (2009) Transfer of anti bodies ac ross the plac enta and in breas t milk from mothe rs on intravenous imm unoglobuli n . Pediatr . Allergy Imm unol. , 20(6) : 528 ?535. 18. Picc i ano, M.F. (1995) Milk Lipids : Water soluble Vitami ns in Human Milk. Handboo k of Milk Com posi ti on, pp. 675 ?688. San Diego, CA: Academ ic Pres s. 380 Nuevas plataformas anal?ticas en metabol?mica 1 9 . S hah, N.P. (2000) Effec ts of milk - deri ved bioac ti ves : an overvi ew, Bri t . J. Nutr . , 84 (1): S3 ?S 10. 20. T i jeri na - S ?enz A . , Inni s S . M . , Ki tts DD . (2009) Antioxi dant capac i ty of hum an milk and its associ ation with vitam i ns A and E and fatty acid com pos i ti on . Acta Pedi atr . , 98(11): 1793 ?179 8 . 21. V an de Perre P . (2003) Trans fer of anti body via mothe r's milk. Vacci ne , 21(24) : 3374 ? 3376. 22. V i ant M.R., Rosenblum E.S ., Tjeerdem a R.S . (2003) NMR - Bas ed Metabolomi cs : A Powerful Approach for Charac terizi ng the Effec ts of E nvi ronmental Stress ors on Organism Health . Environ. Sci . Tec hnol . , 37 (21) : 4982 ?4989. 23. Wis ha rt D.S ., Tzu r D., Knox C. et al. ( 2007 ) HMDB: the Hum an Metabolom e Databas e . Nuc lei c Aci ds Res . , 35 (1): D521 ? D526. PART 4: Metabolomics fingerprinting La Parte IV de la Memoria la constituye la investigaci?n dedicada al estudio de huellas dactilares metab?licas mediante tres t?cnicas anal?ticas diferentes, pero complementarias. Adem?s del denominador com?n que constituye la estrategia, la poblaci?n en estudio estuvo constituida por individuos obesos no sometidos a medicaci?n y sin ninguna patolog?a que pudiese alterar el estudio. Se realiz? un plan para la administraci?n controlada de desayunos preparados con cuatro tipos de aceites de origen vegetal con diferente perfil de antioxidantes sometidos a un proceso de fritura simulado. La muestra utilizada en los tres cap?tulos que componen este bloque fue orina, con la que se obtuvo informaci?n del efecto de la ingesta de cada uno de los aceites en el metabolismo de cada individuo. El criterio considerado para la ordenaci?n de los cap?tulos que componen este bloque ha sido el nivel de informaci?n proporcionada por los diferentes equipos anal?ticos utilizados para obtener los espectros correspondientes que, a su vez, condiciona el tipo de tratamiento quimiom?trico a utilizar. Con este criterio, el Cap?tulo 11 recoge la investigaci?n realizada mediante NIRS con posterior an?lisis multivariante de tipo supervisado (PLS-CM) para predecir el efecto causado en el metabolismo de los individuos por la ingesta de cada uno de los desayunos administrados. El Cap?tulo 12, dedicado a la utilizaci?n de NMR, junto al an?lisis no supervisado y supervisado (PCA y PLS-DA, respectivamente), para discriminar entre los dos muestreos realizados despu?s de la ingesta frente al realizado en condiciones basales, as? como la identificaci?n de familias de metabolitos influenciados por la administraci?n de cada desayuno. Finalmente, el Cap?tulo 13 recoge el estudio realizado mediante una t?cnica como LC?TOF/MS con el fin de no s?lo igualar los niveles de discriminaci?n conseguidos en los dos cap?tulos anteriores, sino tambi?n de identificar los metabolitos que han experimentado un mayor cambio asociado a la ingesta de cada desayuno. Part IV of the book is devoted to the research on metabolic fingerprintings by three different and complementary techniques. In addition to the common denominator of the given strategy, the population under study was constituted by obese individuals, non subjected to drugs intake and without pathologies that can influence the study. A plan was designed for controlled supply to the population of breakfasts prepared with four types of vegetal oils with different profile of antioxidants and subjected to a simulated frying process. The target sample used in all three chapters in this part of the book was urine, from which information about the effect on the metabolism of each individual of the intake of each of the breakfast was obtained. The criterium to order the chapters in this part was the level of information provided by the different instrumental plaftorms used to obtain the given spectra, which, in turn, determined the chemometric treatment to be applied. With this criterium, Chapter 11 contains the research carried out by NIRS, with subsequent supervised multivariate analysis (PLS-CM) to predict the effect on the individuals metabolism of the intake of the different breakfasts. Chapter 12 describes the use of NMR, together with the non supervised and supervised analysis (PCA and PLS-DA, respectively) for discrimination between the two samplings developed after intake (taking as blank or control a sampling developed before breakfasts intake), and also identification of metabolites families affected by breakfast intake. Finally, Chapter 13 is devoted to the study based on LC?TOF/MS intending, not only to reach levels of discrimination similar to those in Chapters 11 and 12, but also to identify those metabolites that experienced a higher change associated to breakfast intake. CHAPTER 1 1 : Near - infrared spectroscopy and partial least squares - class modelling (PLS - CM) f or metabolomics fingerprinting discrimination of obese Individual s after intake of i ntervention b reakfasts Near - i nfrared spe ctroscopy and partial l e ast squares - cl ass mode lli ng (PLS - CM) for metabolomi cs f i nge rpri nti ng discrimination of obese individuals after intake of intervention b reakfasts B. ?lvarez - S?nchez 1 , , 2 F. Priego - C apote, J. Garc?a - Olmo , M.C. Orti z, L. Sarabia , M.D. Luque de Castro * 1 Department of Anal y ti cal Chemistry, Annex Marie Curie Building. Campus of Rabanal es, University of C?rdoba, E - 1 40 71 , C?rdo ba, Spain 2Insti tut e of Bio medical Research Maim? nides (IMIBIC ), Reina Sof?a Hospit al , University of C? rdo ba, E - 1 4 07 1 , C?rdoba, Spain 3S.C .A.I., NIR/MIR Spect ro sco py Divisio n, Ram? n y Cajal Buil ding, Campus of Rabanal es, University of C?rdoba, E - 1 40 71 C?rdoba, Spain 4Chemistry Department, P laza Misael Ba?uelo s, s/n, University of Burgo s, Burgo s, Spain. 5Mathematics and Computatio n Department. Plaza Misael Ba?uelo s, s/n, University of Burgo s, Burgo s, Spain Sent to Metabol o mics for publicatio n 391 Sent to Met a b olo mics Cha p t er 11 Near - In fra red Spectr osc opy and Partial Least Squares - Class Modellin g (PLS - CM) for Meta bolo mics Fin gerprint ing Discri minatio n of Obese Ind ivid ua ls after Inta k e of Intervention Brea kfasts B. ?lvarez - S?nchez, F. Prie go - C apo te, J. Garc?a - Olmo , M.C. Orti z, L. Sar abias, M.D. Luque de Castro * Abst ra ct N ear - i nfrared spec trometry (NIRS) has been used in nutri ti onal metabolom ic s fi ngerpri nti ng for the ass ess ment of th e intake of four interventi on breakfas ts prepared with four different vegetable oils that were previ ous ly subjec ted to a deep fryi ng proc es s for 20 cyc les of 5 min at 180?C. Target oils were an extra virgi n oli ve oil and three varieti es of refi ned sunflowe r oil. Of the three las t, one of them was us ed as such, othe r was spiked with a syntheti c oxi dation inhi bi tor such as dim ethyls iloxane and, fi nally, the las t one was enri che d with an extrac t of phenolic com pounds from oli ve pom ac e, whic h are well - known bec aus e of thei r anti oxi dant properties . Uri ne sam pled from indi vi dua ls before intake and 2 and 4 h after intake, was direc tly analyzed by NIRS to obtai n fingerprintings charac teris tic of the metabolome com pos i ti on. Res ul ti ng uri nary patterns were com bi ned fo r statis tic al analysis by unsupervis ed and supervi s ed approache s. Parti al leas t squa res - c las s modeli ng (PLS - CM) enabled to develo p class - m odels for each interventi on breakfas t that enabled disc rim i nati on of uri nary fi ngerprinti ngs from indivi duals after br eakfas t intake. Thes e models were statis tic ally charac terized by es tim ati on of sens i ti vi ty and speci fi ci ty parameters for the trai ni ng and vali dation (predic ti on) steps, whic h reported qui te acc eptable valu es consi deri ng a nutri tional study deali ng with hu m ans . The appli cati on of Variable Im portance in Projec ti on (VIP) algori thm enabled to detec t thos e spec tral regions with hi gher signific ance to explai n the vari abi li ty observed in the PLS clas s - m odels . Quanti tati ve differences of VIP scores disc rimi nated a mong the different class es under study. 392 N uevas plat a for ma s ana l ?t ica s en met a bol ?m ica 1. Intr oductio n M e t a b o l o m i c s i s ga i n i n g p o p u l a r i t y i n t h e l a s t d e c a d e s i n a w i d e r a n g e o f re s e a r c h a r e a s a s i t p r o v i d e s a n i n s i g h t o f th e m e t a b o l i c s t a t e o f b i o l o g i c a l s y s t e m s . Thus , targeti ng metabolomi cs , whic h deals with analysis of a limi ted set of metaboli tes , is sens i ti vely carried out by tandem mas s spec trom etry after chromatographi c or elec trophoreti c separati on although other les s - sensi ti ve detecti on sys tems can als o be implemented dependi ng on the requi r ed sensi ti vi ty [1]. Global metabolomi cs profi li ng, whi ch aim s at d e t e c t i n g a b r o a d r a n g e o f m e t a b o l i t e s , is usually carried out by hi gh - res olution separati on tec hni ques [(hi gh or ultra perform anc e liqui d chrom atography (HPLC or UPLC), gas chrom atography (G C) and capi llary elec trophoresis (CE)] cou pled to a mass analyzer [Ion Trap (IT ), Quadrupole - Tim e - o f - Fli ght (Q - T OF) and Orbi trap] [2,3]. The las t mode, known as metabolomics fi ngerprinti ng, is widely us ed to provide charac teris tic patterns or "fingerprints " of the biologi c al sys tem. Analytic al methods for metabolomic s fi ngerpri nti ng are usua lly sim ple and requi re mini mum or null sample preparati on [4,5]. The mai n limi tati on of fi ngerprinti ng studi es is that direc t analysis is not always pos si ble and sam ple dilution, extrac ti on of metaboli tes or deprotei ni zation are frequently mandatory steps . Infrared spec trom etry (IRS) is starti ng to gai n wide acc eptanc e withi n metabolomi cs for fi ngerpri nti ng analys is [6] as it allows direc t analys is of a wide range of bi olo gic al samples , inc lu di ng tiss ues , cells and biofl ui ds . T h e m a i n l i m i t a t i o n o f I R S f o r i t s u s e i n m e t a b o l o m i c s i s th e r e l a t i v e l y p o o r s e l e c t i v i t y and sensi ti vi ty as com pared to Nuc lear Magneti c Resonanc e (NMR) and Mas s Spec trom etry (MS), whic h lim i ts its sui tabi li ty for global profili ng and quanti tati ve target analysis . However, rapi di ty and reproduc i bi li ty of Fou ri er trans form - IR (FT - IR), the nil sam ple treatm ent 393 Sent to Met a b olo mics Cha p t er 11 requi red and the pos si bi li ty of i n - vivo studi es even with soli d sam ples are relevant benefi t s that have contri buted to rec ognize IRS as a valu able tool for metaboli c fi ngerpri nti ng . Thus, IRS - fi ngerprinti ng has been used in mic robi olo gy for the identi fic ation of bac teria to the sub - speci es level [7], differenti ati on and identi fi c ati on of cli nic al ly relevant bac teri al speci es [8,9]. Cli nic al appli cations of near - IR spec trom etry (NIRS ) inclu de analys is of faeces for diagnosi s [10], follic ular flui ds to provide a biom arker for VOO yte quali ty [11], and synovi al flui d to aid in the diagnos is of arthri t i c disorders [12]. In terms of serum analys is , IR spec tra have enabled to disc ri mi nate between diabetes type 1, diabetes type 2 and healthy donors [13]. Metabolic fi ngerprints of athletes to detec t dopi ng and overtraini ng has als o been inves ti gated usi ng a variety of body flui ds , dem ons trating that FT - IR cou ld be applied to rou ti ne clini c al analysi s [14]. Attem pts at cancer res earch inc lude the correc t identifi cati on and early diagnos is of (pre)c anc er stages, enabli ng prompt therapeutic intervention and lea di ng to desirable prognos is . A signific ant num ber of studies have been undertaken usi ng IRS techni ques to detec t several form s of canc ers with varyi ng degrees of suc c ess . However, it shoul d be stated that thes e studi es mus t be vali dated with his patho logic a l data on the ass ayed biolo gi c al sam ple . The use of IRS for metabolic fi ngerprinti ng can only be address ed in com bi nati on with multivari ate statis tic al tec hnique s such as princ i pal com ponent analysi s (PCA), partial leas t squa res dis c rim i nant analysi s (P LS - DA) or partial leas t squares class modeli ng (PLS - CM) ???????? Agai ns t a disc rimi nati ng tec hni que , clas s - m odelli ng tec hni que s allo w unders tandi ng the spec tral properties of the sam ple and develo ping clas si fi cati on models , bei ng poss i ble to estim ate sens i ti vi ty and spec i fi city for eac h clas s - model. Sensi ti vi ty meas ures the capaci ty of the com puted model to rec ognize its own objec ts whereas spec i fi ci ty meas ures the capac i ty of the model to reject forei gn objec ts . Thus , both parameters joi ntly desc ri be the p erform anc e of the com puted models and allow evaluation of the poss i ble confus ion among categori es or detec ti on of outli er data. From the defi ni ti ons it is clear that clas s - models with high sens i ti vi ty and high speci fi ci ty are needed. However, 394 N uevas plat a for ma s ana l ?t ica s en met a bol ?m ica these param et ers generally show opposi te beha viou r: when one inc reas es the othe r dec reas es, so it is usefu l to evaluate the sensi ti vi ty and speci fici ty of a fam i ly of clas ses both in fitting and in predic tion [28,29]. In this researc h NIR spec trom etry has been used to obtain metabolic fi ngerprints of urine obtai ned from a nutri tional study in humans. The nutri tional study consi sted of intake of fou r independent interventi on breakfas ts prepared with fri ed edi ble oils contai ni ng diffe rent concentrations of oxi dation inhi bi tors naturally pres ent or added to them. The analysis of the effec ts caus ed by the intake of the different breakfas t was supported on a class - m odelli ng (CM) stati s tic al tec hni que (PLS - CM). The aim of this study was to extend the applic abi li ty of IRS - bas e d fi ngerpri nti ng to hum an nutri tion by explori ng the effec t of food intake on metabolism . In fac t, IRS has been previ ous ly used to assess the nutri ti onal charac teri s ti cs of food and for quali ty control during food produc ti on [17,18], whi le there are not ap plic ati ons of IRS to evalua te the final effec t of food intake on metabolis m. This coul d be addres sed in a sim i lar way to studi es propos ed to evaluate the biologic al effec t of toxi ns or drugs . One exam ple is the analys is of urinary fingerprints, us ed to de m o n s t r a t e t h e e f f e c t o f c h r o n i c c y s t e a m i n e (C S ) s u p p l e m e n t a t i o n [2 4 ] o n th e u r i n a r y m e t a b o l i c pr o f i l e , t h u s e n a b l i n g i d e n t i f i c a t i o n o f e n d o g e n o u s m e t a b o l i t e s w h o s e l e v e l s w e r e d i s t u r b e d b y C S e x p o s u r e . B y e x t r a p o l a t i o n t h e pu r p o s e s h o u l d be t o a n a l y z e t h e c hange in the metabolic profi le of biofl ui ds after food consum pti on. Biologi c al flui ds such as plasm a, serum or urine [22,23] have been com monly used for fi ngerprinting metabolomi cs , as they are easi ly and non - i nvasi vely collec ted. Additi onally, they direc tly reflec t the global state of an indi vidual and/or his respons e to drug treatm ent or diet intake, with the only requi rem ent of sim ple sam ple preparation protoc ols . It is expec ted that metabolic changes due to the intake of breakfas ts prepared with diffe r ent oi ls would lead to changes in the uri nary profi le. 395 Sent to Met a b olo mics Cha p t er 11 2. Materia ls and methods 2.1. Oils and heating procedure T he fou r edi ble oi ls used for this study were: (1) Extra virgi n oli ve oil prepared by mixi ng diffe rent comm erc ial extra - vi rgi n oli ve oils (VOO ). The mixture was optim ized for a final concentration of total phenols of 400 ?g/m L, expres sed as ?g/m L of caf feic aci d by the Fo lin?Ciocalteu test, and had the followi ng fatty ac i ds com pos i ti on: 70.5% monou nsaturated fatty ac i ds (MUFAs ), 11.1% PUFAs , 18.4% saturated fatty aci ds ( SFAs ). (2) Com merci al pure refi ned sunflower oil with nil content in phenolic com pounds ( S ), and had the fo llowi ng fatty aci ds com pos i ti on: 34.3% MUFAs , 58.3% PUFAs and 7.3% SFAs . (3) R efi ned hi gh - oleic sunflower oi l that was spi ked at 400 ? g/m L with a synthetic lipophi li c oxi dati on inhi bi tor (dim ethylsi loxane, DSO) . This prepared oil had the followi ng fatty aci ds com pos i ti on: 71.8% MUFAs , 18.0% PUFAs and 10.2% SFAs . (4) R efi ned hi gh - ole ic sunflower oi l that was enric hed with an extrac t of hydrophi lic phenols isola ted from oli ve pomac e by a protocol sim i lar to that develo ped by Gir?n et al. [30]. The enric hm ent was carr ied out at 400 ? g/m L of total phenols expres sed as caffeic aci d. The concentration of fatty aci ds was as follows : 76.7% MUFAs , 17.6% PUFAs and 5.8% SFAs . E ach oil (2 L) was plac ed in a stai nles s - s teel deep fryer, and was heated at 180 ?C ? 5 ?C for 5 min a total of 20 cyc les. The purpos e of the enri chment of refi ned edi ble oils with antioxi dants was to enha nce oils stabi li ty and improve healthy properties . 396 N uevas plat a for ma s ana l ?t ica s en met a bol ?m ica 2.2. Subjects and samples The experim ent was planned followi ng the gui deli nes dic tated by the Worl d Medic al As S iation Dec larati on of Helsi nki (2004), whic h were supervi sed by the ethic al revi ew board of Reina Sofi a Hos pi tal (C? rdoba, Spai n) that approved the experi ments . Ei ghte en obese indi vi dua ls with a body mass index between 30 ?40 kg/m 2 formed the c oh ort in thi s study. All of them gave their inform ed conse nt and underwent a com prehe nsi ve medi cal his tory, phys ic al exam i nati on and cli nic al chemi s try analysi s befo re enrolment. Partic i pants with evidence of kidney, panc reas , lung, liver or thyroi d dis eas es were exc lu ded. All subjec ts were non - diabetic s, non - s mokers and did not manifes ted cli ni c al evidences of cardi ovasc ul ar dis eas e. The target coho rt was com pos ed by 9 pos t - m enopau s al wom en, age 48 ?70 years , and 9 men, age 39 ?70 years. None of the subjec ts was taki ng medi cati on or supplem entary vitam i ns with influe nti al effec t on urine metabolom e . All volu nteers recei ved fou r breakfas ts in muf fi n form at prepared with the fou r different oils (0.45 mL of oil per kilogram of body weight), previ ous ly subjec te d to the sim ul ated fryi ng proces s. The admi nis trati on of each breakfas t was random ized and cross ed followi ng a Lati n squa re des i gn, whic h inc reas ed the power of the study. The volu nteers ate one of the breakfas ts every two weeks (4 oils , 8 weeks ). During t he sam pli ng period (4 hours ) subjec ts did not cons ume any food. Sam pli ng was performed follo wi ng the Rec om mendations on Biobanki ng Procedures for urine proc es si ng and management recently publishe d by the European Cons ens us Expert Group Report [31]. Accord i ng to this docum ent, biobanking procedures for uri ne shoul d cons i der the followi ng general cons ens us rec om mendations: (i) cells and partic ul ate matter shou ld be rem oved ( e.g . by centri fugation); (ii ) sam ples shoul d be 397 Sent to Met a b olo mics Cha p t er 11 stored at ????C or below? (iii) time limi ts for proces si ng shou ld have been defi ned experim entally and shoul d be appropriate to the analytes to be meas ured; (iv) unless spec ifi ed for a parti cul ar downs tream analysi s , uri ne sam ples should be stored without addi ti ves [32]. In order to obtain sui table control sam ples and carry out a tim e - c ou rs e study of uri nary exc reti on after dietary interventi on, sam ples were obtained at 0 (basal state) and 2 and 4 h after intake . Uri ne was collec ted in sterile contai ners , ali quo ted in 2 - m L Eppendorfs , centri fu ged at 1500 g 4 ?C for 10 min and stored at ?80 ?C until analysi s. This protocol ens ures rem oval of parti cul ates that may interfere with both analys is and quenchi ng bac teri al/enzym ati c acti vi ty in uri ne duri ng storage. 2.3. Instrumen t s and apparatu s The ins trument used for spec tra collec ti on was a Spec trum One NTS FT - N IR spec trophotom eter (Perkin Elm er LLC, Shelton, USA) equi pped with an integrati ng sphere module. Sam ples were analyzed by trans flec tance usi ng a glas s Petri dis h and a hexagonal reflec t or with a total transf lec tance path length of approximately 0.5 mm. A diffus e reflec ti ng stai nless steel surfac e plac ed at the bottom of the cup reflec ted the radiati on bac k through the sam ple to the reflectanc e detec tor. The spectra were collec ted usi ng S pec trum Software 5.0.1 (Perkin Elmer LLC, Shelton, USA). Before analysi s, ali quo ts were thawed and therm os tated at 24 ?C. The reflec tanc e (log1/R) spec tra were collec ted in duplicate and averaged for chem om etri c analysi s . Centrifugation was carri ed out wi th a thermos tated centri fu ge Thermo Sorvall Legend Mic ro 21 R from Thermo (T he rmo Fis he r Sci enti fic, Brem en, Germany) . 2.4. Data acquisition and soft ware 398 N uevas plat a for ma s ana l ?t ica s en met a bol ?m ica Two reflec tance spec tra were acqui red for each sam ple at the wavele ngth range ???????? nm at ??? nm intervals? The two spectra per s am ple were averaged, res ul ti ng in 8501 reflec tance measurem ents for eac h sam ple. Therefore , the prelim i nary data set was com pos ed by 2 16 files (corres pondi ng to the total spec tra obtained for each uri ne sam ple taken before and 2 and 4 h after breakfas t intake) ? 8501 meas urements . Unsc ram bler software (versi on 9.2, Process AS, Oslo, Norway) was us ed for data process i ng. To fit and vali date the PLS models , the PLS Toolbox 3.5 (Ei genvec tor Research, Inc ) for MAT LAB TM was used. The MAT LAB codes us ed for com puti ng the RC and CVRC proc edeu res were developed by the authors . 3. Results and discus sion 3.1. Dat a pret reat ment Dee p fat fryi ng is one of the mos t comm on process es used worldwi de for preparati on of cooked food. However, fryi ng lead to oxi dati on reac ti ons res ulti ng in los s of nutri tional valu e as well as in changes of organoleptic al properties [19,20]. It has been demons trated that the presenc e of anti oxi dants, naturally exis ti ng in (or added to ) oils , exerts benefic i al effec ts by avoidi ng or delaying oxi dation of com pounds such as sterols , fatty alc oho ls , tri terpeni c dialc oho ls and unsaturated fatty acids. Phenolic com pou nds are naturally occ urri ng anti oxi dants present in vegetable oi ls [21]. In the present study, we pos tul ate that com pos i ti on of vegetable oi ls and thei r anti oxi dant profi le would have an impac t on hum an 399 Sent to Met a b olo mics Cha p t er 11 he alth and metaboli sm , whic h woul d be reflec ted in the metabolic profi le and, conc retely in the urinary fingerprint. Pri or to m ulti vari ate analysis , the data set was norm ali zed by application of the standard norm al vari ate (SN V ) and fi rs t Norris deri vati ve. A comm on fac tor ass oc iated to metabolom ic s fingerprinting analys is in cli ni c al and nutri ti onal studi es deali ng with hum ans is the biologi cal vari abili ty among indi vidual s. This vari abili ty is frequently highe r than vari abili ty ass oc i ated to the effec t caus ed by internal fac tors (dis eas es, metabolic dis orders ) or external fac tors (diet, lifes tyle). For this reas on, vari abili ty am ong indi vidual s can be mas ked by thei r influe nce. In this research, data reduc ti on was carri ed out in order to minim ize the biolo gic al vari abili ty not asc ri bed to breakfas ts intake. For this purpos e, the spec trum ac qui red in basal state (t 0 ) was subtrac ted from spec trum acqui red 2 and 4 h after the consum pti on of the prepared meal. Therefore , the sam ple set was fi n ally com pos ed by 144 sam ples (18 indi vi duals ? 4 breakfas ts ? 2 subtrac ted spec tra) and 8501 meas urements . By thi s data pretreatm ent mini mi zati o n of inter - i ndivi duals vari abi li ty sources shoul d be expec ted, and thus, vari abi li ty asc ri bed to the type of interventi on breakfas t be disc rimi nated. Supplem entary Figure 1 sho ws the acc um ulation of 144 spec tra hi ghli ghti ng those correspondi ng to each br eakfas t. As can be seen, uri ne spec tra behave as charac teris ti c fingerprinti ngs of the metabolic profi le. Di fferences in the spec tra as s oc i ated to each breakfas t can be clearly vis ualized reveali ng a vari abili ty source in speci fic zones of the spec trum. Th e only regi on that did not show vari abili ty among indi vi dua ls was that in the wavele ngth range 800 ????? nm? therefore it was removed from the data set to avoid redundant inform ati on. A total of 1600 colum ns were elim i nated from the data set leadi ng t o a final matri x com pos ed by 144 sam ples ? 6901 meas urem ents. 400 N uevas plat a for ma s ana l ?t ica s en met a bol ?m ica 3.2. Stati sti cal analysis by unsupervised and supervised an alysis The com paris on of metabolic fingerprinti ngs obtai ned for each indi vidual after intake of the four interventi on breakfas ts was supported on a dual strategy integrati ng two multi vari ate analysis tec hni ques . The strategy was based on unsupervis ed analys is , to detec t the pres ence of clus ter of indi viduals that coul d be correla ted with the interventi on, and supervi sed analys is , incorporati ng inform ati on for the develo pm ent of predic ti ve models . Multi vari ate analys is started by unsupervi s ed PCA analys is to reduce dim ens i onali ty of the data set. The purpos e was to detec t the presence of sample clus ters that coul d be as s oc i ated to the intake of certai n fried oils. As previou s ly explai ned, data were normali zed and, in thi s case, the data set was autosc ale d pri or to the appli c ati on of the algori thm . Figure 1 shows the scores plot corres ponding to the PC1 ? PC2 spac e. As can be seen, no clus ter or sam ple groupi ngs associated to the intervention breakfas t intake by indi vidual s coul d be detec ted by com bi nation of the two fi rs t PCs or by othe r combi nati on. The explai ned vari abili ty in this case with the PC1 and PC2 was 34.89%. Influe nce of anthropom etric fac tors such as age and gender was discarded sinc e the indi vi duals were nei the r class ifi ed ac cord i ng to these two vari ables . Attendi ng to these res ults , the data set was analyzed by supervi sed analys is usi ng PLS - CM. The propos ed stati s ti cal metho dolo gy couples a PLS regress ion with a binary respons e and a hypothes is tes t to bui ld a set of different m odels for eac h clas s . The PLS - CM technique has been previous ly applied to evalua te screening methods in chem ic al analysi s [33]. PLS - CM seems to be speci ally sui ted for the study here reported sinc e the correlations of abs orbanc e valu es at diffe rent wavele n gths or the pos si ble colli neari ti es do not limi t PLS regres si on. Additi onally, spec tral regions with hi ghe r statis tic al relevance for the develo pm ent of the model for eac h class can be identi fied. 401 Sent to Met a b olo mics Cha p t er 11 Figure 1. S co res plot co rrespo nding to the PC1 ?PC 2 spac e obt ained fro m the PXA anal ysis of the four classes . Suppl ementary Figure 1.A. Acc umulatio n of 14 4 spectra; red: NSO, blue: PSO,DSO, VO O 402 N uevas plat a for ma s ana l ?t ica s en met a bol ?m ica Suppl ementary Figure 1.B . Acc umulatio n of 14 4 spectra; r ed: VOO, blue: PSO,DSO, NS O . Suppl ementary Figure 1.C . A cc umulatio n of 14 4 spectra; r ed: DSO, blue: PSO, VO O, NSO. 403 Sent to Met a b olo mics Cha p t er 11 Suppl ementary Figure 1.D . Acc umulatio n of 14 4 spectra; r ed: PSO, blue: VOO , DSO, NSO . Fi nally, and also importantly, the Q and T 2 s tatis tic al param eters allow detec ti ng outli ers that can be elim i nated from the data set to avoi d erroneous applic ation of the model. Wi th thes e premi s es , the data set was as soci ated to a vec tor y (respons e) with 144 lines to categori ze the vari able to be modeled. Thus , the valu e of thi s vari able was "1" when the indi v idual belonged to the modeled class , and "0" when the indi vidual belonged to another class . Thi s operati on was perform ed for eac h interventi on breakfas t to develo p an independent model for each one. Prior to analysis the raw data set was subjec ted to the f ollowi ng pretreatment: (i) fi rs t deri vati ve of predic tor variables by means of a Savitzky - Golay filter with a sec ond order polynomi al and a window of 15 points ; (ii ) standard norm al vari ate (SNV ); and, (iii ) no pretreatm ent for the 1/0 binary response . Fig ure 2 illus trates the PLS plots provi ded for VOO interventi on breakfas t for the 1st and 2nd 404 N uevas plat a for ma s ana l ?t ica s en met a bol ?m ica and 3rd and 4th latent vari ables . As can be seen, indi vi duals after intake of the modele d class were disc rimi nated from the rest of indi vi dua ls although partial ove rlappi ng can be visua li zed. Sim i lar PLS - CM disc rimi nations were observed for the resti ng interventi on breakfas ts . Therefo re, the different effec ts res ulti ng from the intake of the interventi on breakfas ts can be confi rmed. 3.3. Deve lopmen t of stat ist ical m odels by PLS - CM for each interven tion breakfast Once disc rim i nati on between indi vi duals after intake of eac h breakfas t was acc om pli she d, the next purpos e was the charac teri zati on of the param eters that stati s ti cally defi ne each class . The mai n aspec t d is ti nc ti ve of PLS - CM is that the class - m odels are bui lt from es tim ations of the probabili ty dis tribution of each category, whic h means that sensi ti vi ty and spec i fi ci ty can be computed. The graphic al plot of both parameters is the ris k curve (RC). In this s ense, PLS - CM can be diffe rentia ted from PLS - ?A ???????, although they s ha re som e of thei r properties . Barker and Rayens [34] provi ded som e insi ghts and form al support to the PLS - DA method and sugges ted its use for dimensi on reduc ti on aimed at disc rimi nation. Also, Gonz?lez - Arjona et al. [37] establis hed the relati on between PLS - DA and procrus tes dis c rim i nant analys is . However, the resul ti ng metho ds were exclu si vely aim ed at disc rimi nation tasks . A disc us si on about the relation between PLS - CM and hypothes is tes t can be seen in ref. [ 3 8]. Usua lly, the class - models are com puted to be us ed in predic ti on of future sam ples . In that sense, statis tic al vali dati on about the expec ted behaviour in predi c ti on sho ul d be given. When there are not many sam ples available, cros s - validation is the preferre d cho ic e, i.e. , to sys tem ati cally spli t the available data set into different subs ets of size t , which are used as an 405 Sent to Met a b olo mics Cha p t er 11 evaluati on set (vali dati on or predic ti on set), whi le the rem ai ni ng n ? t samples bec ome the trai ni ng set. Figure 2.A. Sco res of the first and seco nd latent variabl es for t he cl as s model VOO vs. the rest. Figure 2.B. Scores of th ird and fourth latent variabl es for the class PSO, DSO , NSO VOO PSO, DSO , NSO VOO 406 N uevas plat a for ma s ana l ?t ica s en met a bol ?m ica The proc edure ends when all sam ples have been evaluated onc e. However, the usua l cross - vali dati on proc edures are desi gned to evalua te the predic ti on capac i ty of a classi fic ati on model. On the contrary, PLS - CM generates a set of class models, and thus , a proc edure for this tas k needs to be defi ned. Ortiz et al. [38] proposed a segm ented dou ble cross - vali dati on of the ris k curve in the case of PLS - CM. In thi s res earch, the cros s - validation for each PLS regress ion class - m odel has been carried out with three sets of evaluati on, selec ti ng samples accordi ng to the veneti an bli nds procedure. 3 .3 .1 . Class - m odels fo r VOO - interventio n breakf ast The detai ls of PLS regressi on with all the sam ples for traini ng, X , and the three cross - vali dati on sets are in Table 1. It is important to menti on that in none of the fou r models expos ed in Table 1, indivi duals with v alues of T 2 and Q parameters above the thres hold value at 95% confi denc e were found; therefore, no outliers were detected. As Table 1 sho ws, with an explai ned vari ance in X between 42.3 and 44.4% (cons i dering four latent vari ables ), the four PLS class - m ode ls explai ns a variance between 73.7 and 81.6% of the indi vi duals after intake of VOO - i ntervention breakfas t. Thi s means that less than a half of the variabili ty of the spec tra can be related to the intake of this intervention breakfast. The means and stan dard deviati ons were 0.74 and 0.26 for VOO clas s and 0.001 and 0.18 for the non - V OO clas s (formed by individu als after intake of the res ti ng breakfas ts ), respec ti vely. Usi ng two norm al dis tributi ons with these param eters , the RC obtained is shown with a co nti nuous line in Figure 3.A. The cross - vali dati on ris k curve (CV RC) is depic ted by open circ les . As can be seen, models with 95.3% of sensi ti vi ty and 95.3% of speci fi ci ty were obtained in the trai ni ng step. Charac teri zati on was completed with cross - vali da tion sinc e a more balanc ed model (71.4% and 71.3% of sensi ti vi ty and spec i fi ci ty, res pec ti vely) was obtained. As previous ly em phasi zed, sens i ti vi ty and speci fi ci ty present an 407 S e n t to Met a b olo mics Cha p t er 11 Tabl e 1. Statistic al eval uatio n of PLS - CM model fo r VOO interventio n breakf ast v ersus the resting breakf asts. Root mean squared err o r in cro ss - val idati o n (RMSEC V) and in fitt ing (RMSEC ) and cumulative variances of predicto rs (X) and respo nse (y) expl ained by the PLS model as a funct io n of the number of latent variabl es. The val ues sel ect ed for the mo del are indicated with ital ic charact ers . N? of latent vari ables RMSECV RMSEC Explai ned vari ance in X (% accumul ated) Explai ned vari ance in y (% ac cumul ated) PLS with all 108 sampl es (21 VOO and 87 non VOO ) 1 0.4430 0.4094 25.83 13.79 2 0.4425 0.3231 34.12 46.31 3 0.4184 0.2814 39.65 59.28 4 0.4 341 0.2 261 42.29 73.70 5 0.4388 0.1871 44.81 82.00 PLS for the first training set 1T X , with 72 sampl es (14 VOO and 58 non VOO ) 1 0.4509 0.3943 25.26 20.02 2 0.4402 0 .2908 34.90 56.50 3 0.4404 0.2523 40.43 67.26 4 0.4 553 0.2 150 44.43 76.22 5 0.5005 0.1779 47.27 83.73 PLS for the seco nd training set 2T X , with 72 sampl es (14 VOO and 58 nonVOO ) 1 0.5064 0.3386 7.65 41.05 2 0.4973 0.3110 34.05 50.26 3 0.5181 0.2525 38.99 67.21 4 0.5 242 0.1 994 42.72 79.55 5 0.5376 0.1526 45.08 88.03 PLS for the third training set 3T X , with 72 sampl es (14 VOO and 58 non VOO ) 1 0.4349 0.4102 27.64 13.46 2 0.4402 0.3124 34.43 49.82 3 0.4215 0.2461 40.01 68.86 4 0.4 185 0.1 893 43.13 81.58 5 0.4291 0.1533 45.98 87.91 408 N uevas plat a for ma s ana l ?t ica s en met a bol ?m ica opposi te beha vi or. Thus , if a predi c ti on model with 95% of sensi ti vi ty was requi red, spec i fi ci ty would decreas e up to 25%, whi le a model with 95% of speci fi ci ty would re duce sensi ti vi ty up to 29%. Figure 3.A. Risk curve (continuo us line) and c ro ss - validati o n risk curve (open circles) for VOO - cl ass model . 3.3 .2 . Class - models fo r N S O - interventio n breakf ast T he statis ti c al param eters of PLS regress ion models for N S O - i nt erventi on breakfas ts is pres ented in Table 2. Sim ilarly, four latent vari ables were used to defi ne the class - m odel and no outli ers at 95% of confi dence were detec ted. The perc entage of variance explai ned in X (or iT X i= 1, 2, 3) blo c k was sim i lar to VOO class - m odel rangi ng from 42.6 to 45.0%.The vari anc e explai ned for the response vari able ranged from 66.3 and 78.7%, a little bit lower that the vari ance explai ned in the VOO - class model. As a res ult, the mean and standard deviati on we re 0.66 and 0.31 for S 409 S e n t to Met a b olo mics Cha p t er 11 c las s , whi le for the non - S were 0.06 and 0.31, res pec ti vely. Therefore, the means are closer and the RC is further from the 100%, as Figure 3.B shows . Figure 3.B. Risk curve (continuo us line) and c ro ss - validati o n risk curve (open circles) for NS O - cl ass model . In this case, an equili brated model was obtained in the trai ni ng step with 88.3 of sens i ti vi ty and 87.4% of spec i fi ci ty. On the other hand, the predic ti on model reported lower valu es for both parameters , with 71.4 and 71.2% for sens i ti vi ty and speci fi ci ty, res pec ti vely. Altho ugh both param eters were lower than in the VOO class - m odels , both param eters were sim i lar in predic ti on for the two studi es . The model can be forced to attai n a 95% of speci fi ci ty with a subsequent decre as e of sensi tivi ty up to 75.5% in the traini ng step and up to 22.9% in predic tion. Simi larly, a model with 95% sensi ti vi ty cou ld be forc ed with spec ific i ty values of 75.5 and 46.6% in the traini ng and predi c ti on steps, res pec ti vely. 410 N uevas plat a for ma s ana l ?t ica s en met a bol ?m ica Tabl e 2. Statistic al ev al uatio n of PLS - CM develo ped model fo r S interventio n breakf ast versus the resting breakf asts. Root mean squared erro r in cro ss - val idati o n (RMSECV) and in fitt ing (RMSEC ) and cumulative variances of predicto rs (X) and respo ns e (y) expl ained by the PLS mode l as funct io n of the number of latent variabl es. The val ues sel ected for the model are indicated with italic charact ers. N ? of latent vari ables RMSECV RMSEC Explai ned vari ance in X (% accumul ated) Explai ned vari ance in y (% ac cumul ated) PLS with all 108 sampl es (35 S and 73 non S ) 1 0.5460 0.5284 27.75 13.85 2 0.5195 0.4425 35.78 39.57 3 0.5078 0.3781 40.10 55.89 4 0.5 165 0.3 304 43.18 66.32 5 0.5197 0.2845 45.64 75.03 PLS for the first training set 1T X , with 70 sampl es (23 S and 47 non S ) 1 0.5366 0.5043 27.24 10.09 2 0.5194 0.3733 34.38 55.65 3 0.5354 0.3183 39.59 67.76 4 0.5 460 0.2 589 42.65 78.66 5 0.5838 0.2280 46.68 83.45 PLS for the seco nd training set 2T X , with 72 sampl es (23 S and 49 non S ) 1 0.5412 0.5152 30.27 16.90 2 0.5918 0.4116 34.77 46.96 3 0.6140 0.3512 39.85 61.38 4 0.6 391 0.2 858 43.09 74.43 5 0.6585 0.2219 45.67 84.59 PLS for the third training set 3T X , with 73 sampl es (24 S and 49 non VOO ) 1 0.5241 0.4811 14.28 29.60 2 0.5094 0.4360 37.13 42.17 3 0.5035 0.3427 41.36 64.27 4 0.4 959 0.2 979 45.04 73.00 5 0.5304 0.2566 47.58 79.98 411 S e n t to Met a b olo mics Cha p t er 11 3 .3 .3 . Class - models fo r DSO - interventio n breakf ast The PLS model for intake of DSO - i ntervention breakfas t was develo ped res ul ti ng in the param eters lis ted in Table 3. No anomalo us data were found at 95% confi denc e level acc ordi ng to T 2 and Q param eters. Exc ept for PLS model for the sec ond cros s - validati on set, the explai ned vari ance of predi c tor bloc k ranged from 42.5 t o 47.5% whi le the variance explai ned for the res ponse y was from 82.0 to 85.3%. When the traini ng set was 2T X , the values of thes e vari ances were lower (parti cul arly, 36.9 and 47.5%, res pec ti vely). This behavi our cou ld be ascri bed to the pres enc e of two subs ets of indi vi dua ls affec ted by a different trend. Figure 3.B. Risk curve (continuo us line) and c ro ss - validati o n risk curve (open circles) for NS O - cl ass model . 412 N uevas plat a for ma s ana l ?t ica s en met a bol ?m ica Tabl e 3. Statistic al eval uatio n of PLS - CM model fo r DSO interventio n breakf ast versus the resting breakf asts. Root mean squared err o r in cro ss - val idati o n (RMSEC V) and in fitt ing (RMSEC ) and cumulative variances of predicto rs (X) and respo nse (y) expl ained by the PLS model as a funct io n of the number of latent variabl es. The val ues selected fo r the mo del are indicated with ital ic charact ers. N ? of latent vari ables RMSECV RMSEC Explai ned vari ance in X (% accumul ated) Explai ned vari ance in y (% ac cumul ated) PLS with all 108 sampl es (28 DSO and 80 non DSO ) 1 0.4910 0.4707 28. 08 14.56 2 0.4746 0.3913 36.05 40.94 3 0.4581 0.3053 39.76 64.06 4 0.4410 0.2633 43.81 73.26 5 0.4 236 0.2 159 45.64 82.01 6 0.4344 0.1751 47.12 88.18 PLS for the first training set 1T X , with 70 sampl es (18 DSO and 52 non DSO ) 1 0.4896 0.4660 31.10 15.56 2 0.5019 0.3654 36.87 48.08 3 0.4988 0.2614 39.87 73.43 4 0.4864 0.2351 44.99 78.51 5 0.4 849 0.1 943 47.48 85.32 6 0.5221 0.1606 49.71 89.97 PLS for the seco nd training set 2T X , with 73 sampl es (19 D SO and 54 non DSO ) 1 0.5010 0.2563 25.37 19.67 2 0.3 899 0.1 267 36.90 47.53 3 0.4425 0.0634 41.21 66.11 4 0.4425 0.0448 45.42 77.48 5 0.3543 0.0093 48.41 85.78 6 0.3991 0.0093 50.94 90.90 PLS for the third training set 3T X , w ith 73 sampl es (19 DSO and 54 non DSO ) 1 0.5039 0.4577 24.71 19.51 2 0.5030 0.3725 35.16 46.70 3 0.5042 0.2555 39.12 74.91 4 0.4 982 0.2 043 42.53 83.96 5 0.5048 0.1514 44.26 91.19 6 0.5131 0.1220 46.35 94.28 413 S e n t to Met a b olo mics Cha p t er 11 The mean and standard deviation of PLS va lu es for DSO class models were from 0.24 to 0.82, respec ti vely. Sam ples not belongi ng to the class formed by indi viduals after intake of DSO - i nterventi on breakfas t pres ented mean and standard devi ation values of 0.01 and 0.18, res pec ti vely. In com pari s on w ith the VOO class - m odel, the mean valu es were more separated, whi le the standard devi ati ons were simi lar, whic h explai ned the closer pos i ti on to the axes of the RC. Thi s can be chec ked in Figure 3.C. The resul ti ng model was more balance d in the traini ng step, 97.6% of sens i ti vi ty and 97.7% of spec ific i ty, whi le in predic ti on thes e values were 78.6 and 80.0%, res pec tively. Sensi ti vi ty could be improved up to a 95% of speci fi ci ty with sensi ti vi ty of 98.8% in the trai ni ng step and a reduc tion up to 28.6% in the predi c ti on step. Analogous ly, a model with 95% sens i ti vi ty cou ld be attained with spec ific i ty values of 99.2 and 38.8% for the traini ng and predic tion steps, res pec ti vely. Figure 3.C . Risk curve (conti nuo us line) and c ro ss - validati o n risk cu rve (open circles) for DS O - cl ass model. 414 N uevas plat a for ma s ana l ?t ica s en met a bol ?m ica Tabl e 4. Statistic al eval uatio n of PLS - CM develo ped model fo r PSO intervention breakf ast versus the resting breakf asts. Root mean squared erro r in cro ss - val idati o n (RMSECV) and in fitt ing (RMSEC ) and cumulative vari ances of predicto rs (X) and respo ns e (y) expl ained by the PLS model as funct io n of the number of latent variabl es. The val ues sel ected for the model are indicated with italic charact ers. N ? of latent vari ables RMSECV RMSEC Explai ned vari ance in X (% accumu lated) Explai ned vari ance in y (% ac cumul ated) PLS with all 108 sampl es (28 PSO and 80 non PSO ) 1 0.4634 0.4441 28.96 11.23 2 0.4463 0.3599 34.90 41.72 3 0.4363 0.2876 39.04 62.77 4 0.4 142 0.2 499 43.46 71.90 5 0.4268 0.2147 45.92 79.25 6 0.4320 0.1 829 48.14 84.95 PLS for the first training set 1T X , with 70 sampl es (14 PSO and 56 non PSO ) 1 0.4504 0.3752 20.41 29.61 2 0.4291 0.2834 37.01 58.43 3 0.4205 0.2138 40.81 77.15 4 0.4036 0.1698 44.65 85.89 5 0.3 982 0.1 399 48.23 90.21 6 0.3999 0.1124 50.70 93.68 PLS for the seco nd training set 2T X , with 72 sampl es (16 PSO and 56 non PSO ) 1 0.4695 0.4357 26.02 14.56 2 0.4650 0.3577 35.83 42.42 3 0.4509 0.2750 40.05 65.96 4 0.4 366 0.2 216 42.86 77.90 5 0.4456 0.1872 46.10 84.24 6 0.4574 0.1583 48.98 88.72 PLS for the third training set 3T X , with 71 sampl es (15 PSO and 56 non PSO ) 1 0.4427 0.4190 26.21 16.90 2 0.4438 0.3191 31.57 51.80 3 0.4358 0.2260 34.45 75.82 4 0.4 294 0.1 959 40.19 81.83 5 0.4301 0.1668 44.96 86.83 6 0.4335 0.1414 48.38 9054 415 S e n t to Met a b olo mics Cha p t er 11 3 .3 .4 . Class - models fo r PSO - interventio n breakf ast T he las t PLS - CM was that for charac teri zati on of indi vi duals after intake of PSO - i ntervention breakfas ts (Table 4). As in all previ ous cas es , not outliers at 95% of confi dence level were found, which means that no anomalous cases were found in the com plete experi ment. The percentage of vari ance explai ned in X and y were between 40.2 and 48.2% and between 71.9 and 85.9%, res pec tiv ely. As com pared to the three previous PLS - CM studi es , simi lar values were found . The mean and standard deviati on values for PSO class - models were 0.72 and 0.25, res pec ti vely, whi le thes e parameters were 0.03 and 0.2 for indi vi duals that did not intake the referred interventi on breakfas t. As can be seen in Figure 4 , the RC presented with these data, the RC curve is that of Fig. 3.D, that is equally pos i ti oned as the VOO RC. The optim um model (more equi li brated) in the trai ni ng step reported a 93.7% in sens iti vi ty and 93.5% in spec i fic i ty, whi le the predi c ti on model provi ded that sens i ti vi ty and speci fi ci ty valu es of 70.8 and 66.7%, respec ti vely. The model cou ld be forc ed up to a 95% of spec ific i ty with a decrease of sens i ti vi ty up to 97.3% in traini ng and 3 3.3% in predic ti on. A model with 95% of speci fic i ty woul d involve a 97.45 of sens i ti vi ty in traini ng and a relevant reduc tion up to 20.2% in predi c ti on. As a global summ ary, the PLS - CM study has verified that metabolic disc rimi nation of indi vi duals after intake of different interventi on breakfas ts can be attained by co m pari son of uri ne fingerprint s by NIR spec trometry. The different class es (interventi on breakfas ts ) could be ordered in term s of speci fi ci ty and sensi ti vi ty as follows : DSO - i nterventi on break fas t gave place to the clas s - m odel with the hi ghes t speci fi ci ty and sensi ti vi ty parameters (RC closer to the axis ), then S - c las s model, PSO - c las s model and, fi nally, the VOO - c lass model. Conc erni ng predic ti on capabi li ty, thi s was qui te acc eptable for the f ou r class - m odels studi ed. Among them , the DSO - model was the best in predic tion capabili ty, then VOO - and S - m odels were qui te sim i lar and, fi nally, the PSO - model. Altho ugh evalua ted in the plots and tables, the 416 N uevas plat a for ma s ana l ?t ica s en met a bol ?m ica qua nti tati ve differenc es in the statis tic al pa rameters used to charac teri ze the class models were below 10%. The diffe rences obs erved in sens i ti vi ty and speci fic i ty for traini ng and predi c ti on steps coul d be asc ri bed to inter - i ndi vidual and intra - i ndivi dual differences . Wi th the data pretreatm ent em pl oyed in this research, inter - i ndi vi dual vari abi li ty befo re the intake of breakfas ts is minim ized. However, each indi vi dua l responds in a diffe rent way to eac h intake, and this type of vari abi li ty is partic ul arly relevant in cli ni c al and nutri ti onal studi es deali ng with humans . Apart from this inter - i ndi vidual variabili ty, there is also contri buti on of intra - i ndi vi dua l vari abi li ty sinc e eac h indi vi dua l is influ enced by a great vari ety of internal and external fac tors . In this study, som e of thes e fac tors suc h as medic ati on or dis eases were controlled but, there are other fac tors that cannot be controlled. Taking into accou nt that a nutri ti onal study was carri ed out involving hum ans, the models are quite ac ceptable for the purpose of the res earch. Figure 3.C . Risk curve (conti nuo us line) and c ro ss - validati o n risk curve (open circles) for PS O - cl ass model . 417 S e n t to Met a b olo mics Cha p t er 11 3.4. Ident ificat ion of spectral region s respon sible of int ervent ion breakfast discrimi n at ion T he statis tic al charac teri zati on of the different class - m od els develo ped for eac h interventi on breakfas t enables the pos si bi li ty to detec t spec tral regions res ponsi ble for the disc rimi nation of eac h clas s . These spec tral regi ons can be modi fi ed both in quali tati ve and quanti tati ve term s. In this res earc h, the iden ti fic ati on of si gnific ant spectral regi ons contri buti ng to explai n the obs erved vari abili ty in each PLS - CM develo ped was carried out by the Variable Importanc e in Projec tion (VIP) algori thm . This algori thm provi des a projec ti on of the vari ables used to dev elop the PLS model with a score to quanti fy thei r stati s ti cal signi fi c anc e. A vari able with a VIP score clos e to or greater than one can be consi dered important. Vari ables with VIP scores less than one are signi fi cantly less important and might be good can di dates for exc lusi on from the model. It shoul d be noted that the nature of the VIP calc ulati on is suc h that when the model is rebui lt, new variables will always be below the thresho ld so an iterati ve vari able exc lusi on is not recomm ended [40]. As can b e seen, the same spec tral regions are res ponsi ble to explai n the variabili ty of each PLS class - m odel. Thes e spec tral regi ons range from 1800 to 2000 cm - 1 and from 2350 to 2500 cm - 1 . Therefore, quali tati vely, the intake of interventi on breakfas ts is simi lar ly mani fes ted attendi ng to the urine NIR fi ngerpri nti ngs . Nevertheless , there are quanti tati ve differenc es in the score valu es for thes e spec tral regions. Thus , in the spec tral regi on rangi ng from 1800 to 2000 cm - 1 , scores values are the hi ghes t for PSO - i n terventi on breakfas t, then for DSO - i nterventi on breakfas t, VOO - i nterventi on breakfas t and, fi nally for S - i ntervention breakfas t. Thi s region is freque ntly ass oc i ated to hydroxyl functional groups , the firs t sobretone of carbonyl functi onal groups but als o ami no grou ps coul d be involved. 418 N uevas plat a for ma s ana l ?t ica s en met a bol ?m ica T he refo re, fami lies of com pounds that cou ld contri bute to thes e signals range from oxi di zed compounds suc h as fatty acids , aldeh ydes or ketones and als o ami no com pou nds pres ent in uri ne such as ami no acids metaboli tes or car ni ti ne deri vati ves (metaboli tes of fatty ac ids ). Als o, phenolic com pounds coul d contri bute to signals obtained in thi s region with spec ial em phasi s on VOO and PSO interventi on breakfas ts , although tocophe rols are present in sunflower oils even at hi gher co ncentrati on than VOO. On the other hand, the spec tral regi on from 2350 to 2500 cm - 1 shows two different bands . In the area clos e to 2350, bands are more relevant for VOO class , bei ng practi cally abs ent for PSO, DSO and S class es. However, the area closer t o 2500 cm - 1 presents a com mon profi le for VOO and S class es whi le it is more attenua ted for DSO and PSO clas ses . Thi s regi on can be assi gned to ? C H 3 , ?C H 2 ? and ?CH=CH ? groups pres ent in chemi cal struc tures of metaboli tes proc eedi ng from the metaboli sm of f ri ed oils used to prepare the interventi on breakfasts . 4. Co nclusi ons Metabolomi cs fi ngerprinti ng is practi cally dom i nated by the utili zati on of NMR and MS tec hni ques due to thei r potenti al in quali tati ve analys is with identi fi c ati on purpos es and the be nefi ts in terms of selec ti vi ty/s ens i ti vi ty. However, ins trum entati on bas ed on these tec hni ques is expensi ve and requi res personnel with a high form ati on level. On the other hand, near - i nfrared spec trom etry (NIRS ) is a fas t and reproduc i ble techni que speci a lly sui ted for the developm ent of class i fi cation methods to disc rimi nate among classes . Thus , in this research NIRS has been employed to obtain urine fi ngerpri nti ngs for com pari son of the metaboli c effec t caus ed by diffe rent interventi on breakfas ts . In thi s sens e, the potential of NIRS as a tool for developm ent of screeni ng methods has been proved. Thus , after screeni ng analysis , extensi ve researc h cou ld be carri ed out with othe r more 419 S e n t to Met a b olo mics Cha p t er 11 Figure 4.A. VIP scores for VOO PLS - cl ass model . Figure 4. B. VIP scor es for N S O PLS - cl ass model. 420 N uevas plat a for ma s ana l ?t ica s en met a bol ?m ica Figure 4.C . VIP scores for DSO PLS - cl ass model Figure 4.D. VIP scores for P SO - PLS c lass model . 421 Sent to Metabolomics Chapter11 sophisticated techniques. The main benefits of PLS-CM have been taken for the development of this research. 5. Acknowledgements The Spanish Ministerio de Ciencia e Innovaci?n (MICINN) and FEDER program are thanked for financial support through project CTQ2009-07430. B.A.-S. and F.P.-C. are also grateful to the MICINN for an FPI scholarship (BES-2007-15043) and a Ram?n y Cajal contract (RYC-2009-03921). 6. References [1] Dunn, W.U., Ellis D.I. (2008) Trends Anal. Chem. 24 (4) ???????? [2] Theodoridis G., Gika H.G., Wilson I.D. (2008) Trends Anal. Chem. 27 (3) ???????? [3] Goodacre R., Vaidyanathan S., Dunn W., Harrigan G.G., Kell D.B., Trends Biotechnol. 22 (5) ???????? [4] Ward J.L., Baker J.M., Beale M.H. (2008) FEBS Journal 274 (5): 1126? 1131. [5] Shulaev V., Cortes D., Miller G., Mittler R. (2008) Physiologia Plantarum 132 (2): 199?208. [6] R.G. Dambergs, A. Kambouris, I. L. Francis, M. Gishen, J. Agric. Food Chem. 50, 11 (2002) 3079?3084. 422 Nuevas plataformas anal?ticas en metabol?mica [7] Ellis D., Broadhurst D., Kell D.B., Rowland J.J., Goodacre R. (2002) Appl. Environ. Microbiol. 68: ?????????? [8] Kaderbhai N.N., Broadhurst D.I., Ellis D.I., Goodacre R., Kell D.B. (2003) Comp. Funct. Genom. 4: ???????? [9] Diem M., Boydston-White S., Chiriboga L. (1999) Appl. Spectrosc. 53: ???A????A? [10] Nakamura T., Takeuchi T., Terada A., Tando Y., Suda T. (1998) Int. J. Pancreatol. 23: ???????? [11] Goodacre R., Timmins E.M., Gaudoin M., Fleming R. (2000) Hum. Reprod. 15: ?????????? [12] Eysel H.H., Jackson M., Nikulin A., Somorjai R.L., Thomson G.T.D., Mantsch H.H. (1997) Biospectroscopy 3: ???????? [13] Petrich W., Dolenko B., Fruh J., Ganz M., Greger H., Jacob S., Keller F., Nikulin A.E., Otto M., Quarder O., Somorjai R.L, Staib A., Warner G., Wielinger H. (2000) Appl. Opt. 1: ?????????? [14] Petibois C., Cazorla G., Deleris G. (2002) Appl. Spectrosc. 56: ?????? [15] Chiriboga L., Xie P., Yee H., Vigorita V., Zarou D., Zakim D., Diem M. (1998) Biospectroscopy 4: ?????? [16] Lasch P., Haensch W., Lewis E.N., Kidder L.H. Naumann D. (2002) Appl. Spectrosc. 56: ???? [17] Cen H., He Y. (2007) Trends Food Sci. Technol. 18 (2): ?????? [18] Infrared Spectroscopy for Food Quality Analysis and Control. Ed. Da- Wen Sun, 2009, Elsevier. [19] Tyagi V.K. Vasishtha A.K. (19996) J. Am. Oil Chem. S. 73: ???????? 423 Sent to Metabolomics Chapter11 [20] Saguy I.S., Danaa D. (2003) J. Food Eng. 56: ???????? [21] Jap?n-Luj?n R., Luque de Castro M.D. (2008) J. Agric. Food Chem. 56: (2008) 2505?2511. [22] Llorach R., Urpi-Sarda M., Jauregui O., Monagas M., Andres-Lacueva C. (2009) J Proteome Res. 8 (11): ?????????? [23] Gibney M.J., Walsh M., Brennan L., Roche H.M., German B., van Ommen B. (2005) Am. J. Clin. Nutr. 82: 497?503. [24] Llorach R., Garrido I., Monagas M., Urpi-Sarda M., Tulipani S., Bartolome B., Andres-Lacueva C. (2010) J. Proteome Res. 9 (11): ?????????. [25] Cozzolino D., Flood L., Bellon J., Gishen M., De Barros Lopes M. (2006) Yeast 23 (14-15): ???????? [26] Westad F., Martens H. (2000) J. Near Infrared Spectrosc. 8: 117?124. [27] Cozzolino D., Holdstock M., Dambergs R.G., Cynkar W.U., Smith P.A. (2009) Food Chem. 116 (3): ???????? [28] Ortiz M.C., Sarabia L.A., Garc?a-Rey R., M.D. Luque de Castro M.D. (2006) Anal. Chim. Acta 558: 125?13. [29] D?ez R., Sarabia L.A., Ortiz M.C. (2007) Anal. Chim. Acta 585: 350?360. [30] ?ir?n M .V., Ruiz-?im?ne? ?., Luque de Castro M.D. (2009) J. Agric. Food Chem. 57: 2797?2802. [31] M. Yuille M., Illig T., Hveem K., Schmitz G., Hansen J., Neumaier M., Tybring G., Wichmann E., OllierB. (2010) Biopreserv Biobank 8 (1): 65?69. [32] Bernini P., Bertini I., Luchinat C., Nincheri P., Staderini S., Turano P. (2011) J. Biomol. NMR 49: 231?243. [33] Indahl U.G., Martens H. T. N?s. (2007) J. Chemomet r. 21: 529?536. [34] Stahle L., Wold S. (1987) J. Chemometr. 1: 185?196. 424 Nuevas plataformas anal?ticas en metabol?mica [35] Nocairi H., Qannari E.M., Vigneau E., Bertrand D. (2005) Comput. Stat. Data Anal. 48: 139?147. [36] Barker M., Rayens, W. (2003) J. Chemometr. 17: 166?173. [37]Gonz?lez-Arjona D., L?pez-P?rez G., Gonz?lez A.G. (1999) Talanta 49: 189?197. [38] Ortiz M.C., Sarabia L.A., S?nchez M.S. (2010) Anal. Chim. Acta 674: ????142. [39] S?nchez M.S., Ortiz M.C., Sarabia L.A, Busto V. (2010) Chemom. Intell. Lab? Syst? ???? ???42. [40] Jun C. (2005) Chemo. Intell. Lab. Sys. 78: ??????? CHAPTER 12: Comparative study of the Influence of fried edible oils intake on the urinary metabolic fingerprint by nuclear magnetic resonance spectrometry (1H-NMR) Comparative study of the influence of fried edible oi ls intake on the urinary metabolic fingerp ri nt by nuclear magneti c resonance s pectrome try ( 1 H - NMR) B. ?lvarez - S?nchez, F. Prie go - C apo te, J.L. Izquierdo - Garc?a, M.D. Luque de Castro * Department of Anal yti cal Chemistry, Annex Marie Curie Building. Campus of Rabanal es, University of C?rdo ba, E - 1 40 71 , C?rdo ba, Spain Insti tut e of Bio medical Research Maim? nides (IMIBIC), Reina Sof?a Hospit al , University of C? rdo ba, E - 1 4 07 1 , C?rdoba, Spain bCIBER de Enfermedades Respiratorias, CIBERES, Madrid 429 Cha p t er 12 Comp a ra t ive st udy of the inf lu ence o f fried edible oils int a ke on the u rinary Metabolic Fingerpri nt By Nucl ear Ma gnet ic Resona nce Spec t romet ry ( 1 H - NM R) B. ?lvare z - S? nche z, F . Prie go - Ca p ot e, J. L Izq u ier d o - G a rc ?a , M. D . Luq ue de Ca s t ro * Abst ra ct In this study, com pari son of uri nary fingerprints obta i ned by unidim ens ional nuclea r magneti c res onance was carri ed out to disc rimi nate between 144 sam ples obtai ned from 12 volunteers . Uri ne sam ples from obes e indivi duals were taken after intake of meals prepared with four fried vegetable oi ls that differed in thei r anti oxi dant and fatty aci ds com pos i ti on. The aim was to find poss i ble statis tic al differences in the metabolic com posi tion of uri ne related with the consum ption of these oils, as a conse que nce of a different effec t on indi vi duals metabolis m . The application of chemometri c techni ques (PCA and PLS - DA) to NMR spec tra allowed the evaluati on of changes in the levels of metaboli tes after intake of fried vegetable oils . Usi ng PCA, disc rimi nation of controls (bas al state) and sam ples taken 4 h after intak e was pos si ble, letting to conclude that the intake of thes e oils had an effec t on the uri nary metabolic fi ngerpri nt. PLS - DA was used to disc rim inate am ong diffe rent clas s es . Thi s disc rimi nation was pos si ble by com paring clas ses in pairs , with predic ti on a ccurac ies from 84% to 100%. In a final step, chem ic al shi fts res pons i ble for dis c rim i nati on were extrac ted. After a tentati ve identi fi cati on of the mai n signals from NMR spec tra, thes e chem ic al shi fts were attributed to groups of endogenous or exogenous me taboli tes present in uri ne. The resul ti ng approach is a com peti ti ve alternati ve to metho dolo gies based on MS as it is rapid and requi res mini mum sam ple preparati on. 430 N uevas plat a for ma s ana l ?t ica s en met a bol ?m ica 1. Intro duction Metabolomi cs , defi ned as the study of all meas urable metaboli tes in a given biolo gic al sample, provi des a view to understand metabolic vari ations in com plex biolo gic al organism s as a consequenc e of a given stim ulus, such as a patholo gic al condi ti on, drug or diet adm i ni s trati on or genetic vari ation [1,2]. Metabolomi cs analysi s can be address ed by three different approaches dependi ng on the aim of the study (i.e. targeti ng analys is , global profi li ng or metabolic fi ngerpri nti ng). Thus , targeti ng analys is , the mos t sensi tive and selec ti ve approach, aim s at quantifying a relativel y small group of metaboli tes . On the othe r hand, global profi li ng is focus ed on obtaining as muc h biologi cal informati on of the sam ple as poss i ble, usi ng non - selec ti ve approac hes with non - qua nti tati ve or sem i quanti tati ve purposes . Finally, metabolom ic s fi n gerprinting is widely us ed to provi de charac teris tic patterns or "fi ngerpri nts " of the biologi cal sys tem [3]. The aim of metabolom ic s fi ngerpri nti ng is to find statis ti c al differences among groups subjec ted to a concrete stim ulus and to relate these differ enc es to the fi ngerprints (for example to spec tral regi ons ). Succ ess fu l applic ati ons of this metabolomi c approac h in toxic i ty screeni ng, biomarker disc overy [4], in vivo mapping [5,6], and func ti onal genom ic s [7] have been well doc um ented. Metabolomi cs studi es typic ally employ ei the r Mas s Spec trometry (MS) or Nuclea r Magneti c Res onanc e (NMR) to interrogate biologi cal sam ples , due to their great sensi ti vi ty/s elec ti vi ty binomi al and their vers ati li ty [8]. Althou gh MS is clearly more sensiti ve than NMR, the latter offers great a d v a n t a g e s f o r i t s u s e i n h i g h t h r o u g h p u t f i n g e r p r i n t i n g , s u c h a s m i n i m a l r e q u i r e m e n t s f o r s a m p l e p r e p a r a t i o n , a n d t h e no n - d i s c r i m i n a t i n g a n d n o n - d e s t r u c t i v e na t u r e o f th e t e c h n i q u e [ 9 ] . I n f a c t , 431 Cha p t er 12 1 H - N M R s p e c t r a c a n b e a c q u i r e d r a p i d l y i n 3 ? 1 5 m i n wi t h m i n i m u m s a m p l e p r e p a r a t i o n , u s u a l l y j u s t e n t a i l s bu f f e r i n g a n d i n t e r n a l s t a n d a r d a d d i t i o n [ 1 0 ] . Fu r t h e r m o r e , 1 H - N M R h a s be e n l a r g e l y u s e d t o u n e q u i v o c a l l y d e t e r m i n e m e t a b o l i t e s t r u c t u r e s , a n d p e r h a p s d u e t o i t s n o n - i n v a s i v e n a t u r e , i s m o r e c o m m o n l y u s e d i n m a m m a l i a n s y s t e m s t h a n M S m e t h o d o l o g i e s . Thus , NMR metabolom ic s has already demonstrated consi derable potenti al as fi ngerprinti ng tool with extra capabili ty to identi fy biomarkers in biolo gi c al fluids [11], or in toxi c ologi cal studi es [12 ]. Nevertheles s, its use in hum an nutri ti on is sti ll lim i ted [13] des pite metabolom ic s is more sensi ti ve than proteom ics and genom ics to subtle differences in biochem ic al profi les followi ng dietary interventi on. Urine is an ideal biological flui d for NMR fingerpri nti ng ??????? in nutri tional studies , as it direc tly reflec ts the global state of an indi vidual after food intake, and requi res minim um sam ple preparation. Nevertheless , uri ne NMR spec tra are com plex, maki ng to visua lize the effec t of diet through ou t the levels of metaboli tes diffic ul t. On the othe r hand, it has been demons trated that uri ne com posi tion is affec ted by more sources of vari abili ty than other bioflu i ds (e.g. plasm a), inc ludi ng circ adi an vari ati on, hydration state, and es trous cycle [17 ]. In additi on, inter - i ndi vi dual vari ation in metabolism is generally the mai n sou rc e of variabili ty due to thei r great diversi ty in genetic and envi ronm ental fac tors among indi viduals [18]. For these reas ons, pattern recogni ti on tec hni ques , such as princ i pal com ponents analys is (PCA) and partial leas t squares disc rimi nant analys i s (PLS - DA) are frequently us ed to com pare metaboli te levels with class ific ation purposes [19,20]. N a t u r a l a n t i o x i d a n t s a r e k n o w n t o e x e r t a p r o t e c t i v e e f f e c t a g a i n s t c o r o n a r y h e a r t d i s e a s e [2 1 , 2 2 ] . P a r t i c u l a r l y , t h e h e a l t h be n e f i t s o f o l i v e o i l , t h e p r i n c i p a l s o u r c e o f f a t i n th e M e d i t e r r a n e a n d i e t , a r e m a i n l y a t t r i b u t e d t o i t s e q u i l i b r a t e d f a t t y a c i d p r o f i l e a n d r i c h c o m p o s i t i o n i n p h e n o l i c a n t i o x i d a n t s [ 2 3 , 2 4 ] . A l t h o u g h b e n e f i t s o f o l i v e o i l a n d , 432 N uevas plat a for ma s ana l ?t ica s en met a bol ?m ica c o n c r e t e l y , o f i t s a n t i o x i d a n t s , a r e g a i n i n g i n t e r e s t , t h e m e c h a n i s m s i n v o l v e d i n th e p r o t e c t i v e e f f e c t a r e s t i l l u n k n o w n . T h i s s t u d y i s f o c u s e d o n c o m p a r i n g t h e e f f e c t o f th e i n t a k e o f f r i e d v e g e t a b l e o i l s wi t h a na t u r a l o r a d d e d c o n t e n t i n o x i d a t i o n i n h i b i t o r s t h r o u g h t h e u r i n e m e t a b o l o m e o f o b e s e i n d i v i d u a l s . F o r t h i s pu r p o s e , uri nary fingerprints were obtained by 1 H - N M R t o e x p l o r e t h e e f f e c t o f d i f f e r e n t b r e a k f a s t s p r e p a r e d wi t h t a r g e t o i l s o n th e s e l e c t e d c o h o r t . S t a t i s t i c a l a n a l y s i s with PCA and PLS - DA would allow fi ndi ng poss i ble signi fic ant differenc es in the metabolic com posi tion of urine related with the consum pti on of thes e oils , as a consequenc e of a different effec t on metabolis m. T h e o i l s u s e d i n t h i s s t u d y h a d a d i f f e r e n t pr o f i l e i n a n t i o x i d a n t s . 2. Materials and methods 2.1. Reagen t s and equipment Centrifugation was carri ed out with a thermos tated centri fu ge Thermo Sorvall Legend Mic ro 21 R from Thermo (T he rmo Fis he r Sci enti fic, Brem en, Germany); 1 H - NMR spectroscopy w as performed at 500.13 MHz using a Bruker AMX500 spectrometer 11.7 T (Bruker BioSpin GmbH, Rheinstetten, Germany), and trimethylsilyl propionate (TSP) was used as chemical shift reference. A pH 7.2 buf fer contai ni ng 200 mM sodi um phos phate, 0.25 mM TSP, 0.025% sodium azi de, and 25% deuterium oxi de, all from Sigm a ?Aldri ch (St Lou is, MO) was prepared. All reagents were stored at 4 ?C until us e. 433 Cha p t er 12 2.2. Preparati on of oils T he fou r edi ble oi ls used for this study were: (1) Extra virgi n oli ve oil prepared by mix i ng different comm erc ial extra - vi rgi n oli ve oils (VOO). The mixture was optim ized for a final concentration of total phenols of 400 ?g/m L, expres sed as ?g/m L of caffeic acid by the Folin?Ciocalteu test, and had the followi ng fatty ac i ds com pos i ti on: 70.5% monou nsaturated fatty ac i ds (MUFAs ), 11.1% PUFAs , 18.4% saturated fatty aci ds ( SFAs ). (2) Com merci al pure refi ned sunflower oi l with nil content in phenolic com pounds (NS O), and had the followi ng fatty aci ds composi ti on: 34.3% MUFAs , 58.3% PUFAs and 7.3% SFAs . (3) R efi ned hi gh - oleic sunflower oi l that was spi ked at 400 ? g/m L with a synthetic lipophi li c oxi dati on inhi bi tor (dim ethylsi loxane, DSO) . This prepared oil had the followi ng fatty aci ds com pos i ti on: 71.8% MUFAs , 18.0% PUFAs and 10.2% SFAs . (4) R efi ned hi gh - olei c sunflower oi l that was enric he d with an extrac t of hydrophi lic phenols isola ted from oli ve pomac e by a protocol sim i lar to that develo ped by Gir?n et al. [30]. The enric hm ent was carried out at 400 ? g/m L of total phenols , express ed as caff ei c acid. The c oncentration of fatty aci ds was as follows : 76.7% MUFAs , 17.6% PUFAs and 5.8% SFAs . E ach oil (2 L) was placed in a stai nless - s teel deep fryer, and heated at 180 ?C ? 5 ?C for 5 min a total of 20 cyc les . The purpos e of the enric hm ent of refi ned edi ble oi ls with anti oxi dants was to enha nc e oils stabi li ty and improve health y properties . 434 N uevas plat a for ma s ana l ?t ica s en met a bol ?m ica 2.3. Sub jects and samples Experim ent was planned followi ng the guideli nes dic tated by the World Medi c al Ass oc i ation Dec larati on of Hels inki (2004), whi ch were supe rvi sed b y the ethic al revi ew board of Reina Sofi a Hos pi tal (C? rdoba, Spai n) that approved the experi ments . Seventeen obese indi vi duals with a body mass index between 30 ?40 kg/m 2 formed the coh ort in thi s study. All of them gave their inform ed conse nt and underwent a com prehe nsi ve medi cal his tory, phys ic al exam i nati on and cli nic al chemi s try analysi s befo re enrolment. Partic i pants with evidence of kidney, panc reas , lung, liver or thyroi d dis eas e were exc lu ded. All subjec ts were non - diabetic s, non - s mokers and did not mani fes t cli nic al evidences of cardi ovas c ular dis eas e. The target coh ort was com pos ed by 17 post - m enopaus al women, age 48 ?70 years , and 9 men, age 39 ?70 years. None of the subjec ts was taki ng medi cati on or supplem entary vitam i ns with influe nti al effec t on urine metabolom e . All volu nteers recei ved fou r breakfas ts in muffi n form at prepared with the fou r different oils (0.45 mL of oil per kilogram of body weight), previ ous ly subjec ted to a sim ulated fryi ng proces s . The adm i ni s tration of each breakfas t was he ld at random ization and cross followi ng a Lati n squa re desi gn, which inc reased the power of the study. The volu nteers ate one of the breakfas ts every two weeks (4 oils , 8 weeks). Duri ng the sam pli ng period (4 h) the volunteers did not consume any food. 2 . 3 .1 . Sampl ing Sam pli ng was performed follo wi ng the Recom mendations on Biobanki ng Procedures for urine proc es si ng and management recently publishe d by the European Cons ens us Expert Group Report [25]. Accordi ng 435 Cha p t er 12 to thi s doc um ent, protocols for pretreatm ent u ri ne shoul d avoid addi ti ves for stori ng, inc lude c entri fugatio n for rem oval of parti cul ates, and freezi ng at ?80 ?C or belo w. In order to obtain sui table control sam ples and carry out a tim e - c ou rs e study of urinary exc reti on after intake, sam ples were ob tai ned at 0 (jus t before intake, bas al state) and 2 and 4 h from intake (pos t - basal s tates ). Uri ne was collec ted in steri le contai ners, ali quoted in 2 - m L Eppendorfs , centrifu ged at 1500 g, 4 ?C for 10 min and stored at ? 80 ?C until analys is. Thi s protoc ol ensures elim i nati ng parti c ulates that may interfere analysi s and quenc hi ng bac teri al/enzym ati c acti vi ty in uri ne duri ng storage. 2.4. Sample preparati on After thawi ng for analysi s , ali quo t s of 400 ?L were centrifuged at 12000 g for 10 min at 4 ?C for sp i nni ng down non - s olu ble partic les and 360 ?L of supernatant were mixed with 180 ?L of buf fer to minim ize pH vari ati on and with 60 ?L of deuterium water (D 2 O) [8]; afterwards, uri ne sam ples were vortexed and plac ed into an 1 - m L probe for analysis . 2.5. Da ta acquisition The samples weres analyzed at 4 ?C to minimize metabolic changes. Sample stability at 4?C was proved by analyzing a urine sample in duplicate with a time interval of 15 h. No gross degradation was noted in the signals of multiple spectra a cquired under the same conditions. Standard solvent suppressed spectra were grouped into 16.000 data points, averaged over 256 acquisitions. The data acquisition lasted in total 13 min using a sequence based on the first increment of the Nuclear Overhaus er 436 N uevas plat a for ma s ana l ?t ica s en met a bol ?m ica Effect Spectroscopy (NOESY) pulse sequence to effect suppression of the water resonance and limit the effect of B 0 and B 1 inhomogeneity in the spectra (relaxation delay - 90? - t1 - 90? - tm - 90? - acquire Free Induction Decay ?FID ? signal), in which a secondary r adio frequency irradiation field was applied at the water resonance frequency during the relaxati on delay of 2 s and during the mixing period (tm = 150 ms), with t1 fixed at 3 s. The acquisitions were performed using a spectral width of 8333.33 Hz. Prior t o Fourier transformation, the FID signals were multiplied by an exponential weight function corresponding to a line broadening of 0.3 Hz. Data proc es si ng was done with the Mes tRenova software. Chenom x. The "Metabonom ic " GUI bas ed on the R - Tc l/T k interfac e [26], and the Agi lent Mass Profi ler Softwares were used for stati s ti cal analyses . Chenom x NMR Sui te 5.0Profi ler software was used for peak as si gnement. 2.6. NMR spectral dat a processin g Firstly, spectra were referenced to the t ri m ethylsi lyl propionate ( TSP) singlet at 0 ppm chemical shift. Afterwards , regi ons contai ni ng non - relevant inform ati on or spec tral artifac ts were rem oved. This step is nec es s ary bec aus e the spec tral width to acqui re NMR data is usua lly wider than nec es s ary to digi tize all chem ic a l shifts associ ated with endogenous metaboli tes [27]. Thus , downfi eld and upfield spec tral areas witho ut endogenous metaboli tes are ini ti ally exc luded. On the othe r hand, spec tral regi ons highl y dependi ng on the experi m ental parameters , suc h as water and r efe rence regi ons are also deleted, as thes e regi ons are sens i ti ve to spec tral arti fac ts, such as inadequate phasi ng. Therefo re, the spec trum outside the 0.2 - 10 - ppm window was exc luded. 437 Cha p t er 12 2 .6 .1 . Basel ine and phase co rrect io n Baseli ne correc ti on is an ess en ti al step to obtai n hi gh quali ty NMR spec tra. Rolli ng bas eli nes can make diffic ul t to identi fy peaks and can introduc e signi fi cant errors into quanti tati ve meas urements [26]. On the other hand, the baseli ne of the spec trum shoul d be flat and symm etri cally dis posed on eithe r side of eac h of the peaks . If the bas eli ne about the peaks is as ym metric al, then the spec trum requi res phasi ng. All spec tra were manually phas e - and bas eli ne - c orrec ted usi ng the Mes tRenova software. 2.6.2 . Binning T he mos t com mon method for reduc i ng the influe nc e of shifti ng peaks is the so - c alled binni ng or buc keti ng metho d, whic h reduc es spec trum resolution [28]. Thus, the spec tra are integrated withi n small spec tral regi ons, called "bi ns " or "buc kets". Subsequent data analysi s proc edu res applied to the binned spec tra are not influe nced by peak shi fts , as long as these shi fts rem ai n within the borders of the corres pondi ng bins. Therefo re, NMR spec tral data were autom atic ally reduc ed into regi ons of equal width (0.02 ppm and 0.04 ppm) an d the integral of each regi on was determ i ned. The spec tral regi on from 4.0 ?6.0 ppm was exc luded from analysis to rem ove the effec t of vari ati ons in the suppress i on of the water resonanc e and vari ations in the urea si gnal. 2.6.3 . Normal izati o n A cruc ial st ep in pre - proc essi ng of spec trum data in metabonom ic studi es is the so - calle d norm ali zati on step [27]. This step tri es to acc ount for poss i ble vari ati ons in sam ple concentrations. Norm ali zati on may als o be nec es s ary for tec hnic al reas ons . If spec tra are re c orded usi ng a diffe rent num ber of scans or different devic es, the absolu te valu es of the spec tra vary, 438 N uevas plat a for ma s ana l ?t ica s en met a bol ?m ica and renderi ng a joi nt analysis of spec tra without prior norm ali zati on is imposs i ble. 2 .6 .4 . Statistical anal ysis In this study, pri nci pal com ponents a nalysi s (PCA) and partial leas t squa res disc rimi nant analys is (PLS - DA) were us ed for sam ple class i fi cation and pattern rec ogni ti on. The former is a non - supervis ed techni que that allows explori ng data groupings or patterns among sam ples . Thus , in PCA, data are trans form ed to a K - di m ens ional spac e (whe re K is equal to the num ber of vari ables ), and are subsequently projec ted into a few pri nci pal com ponents that desc ri be the maximum vari ati on withi n the data. The firs t PC is the linear com bi nati on of the origi n al vari ables , whi ch explai ns the maxim um amou nt of variance in the data. The scores plots of the firs t two or three PCs provide bi - or tridi mensi onal maps sho wing the relationshi p between data and dis playi ng groupings [19,29]. The loadi ngs plot shows the r elationshi p between vari ables and in com bi nation with the scores plot provi de inform ati on concerni ng the importanc e of vari ables . PCA provides an eas i ly unders tandable graphi cal approac h to explore poss i ble sam ple grouping and identi fy the spec tral regions of difference between the clas ses. On the other hand, PLS - DA is a linear regres si on metho d in whic h the multi variate variables corres pondi ng to the observations (spec tral desc ri ptors ) are ass oc iated with the clas s mem bershi p for each sam ple [29]. In cont ras t to PCA, the method is supervis ed (i.e. the clas s mem bershi p of the sam ples is inc lu ded in the calc ul ati on). PLS - DA is used when clus ters are not dis ti nc tly separated in the scores plot and grou ps overlap so that it is desi rable to fine - tune the models to allow com plete dis ti nc ti on between groups . As the method is bas ed on the creati on of a clas si fi c ati on model, this allows predic tions by assi gni ng probabili ti es to new observati ons for class mem bers hi p. Vali dati on of the res ul ti ng PLS models was support ed on 439 Cha p t er 12 c ros s - validation tes ts due to the num ber of indi vi duals com posi ng the cohort under study. 3. Results and discus sion 3.1. Prin cipal compon ent s analysis T he matri x contai ni ng the reduc ed spec tral data was exported to the MPP software for statis tic a l analysi s . The preli mi nary data set was com pos ed by 245 poi nts per spec trum with a total of 144 spec tra. In a first step, data were reduc ed by applyi ng a fold change (FC) filter, whi ch was em ployed to keep spec tra regi ons with high abundanc e rati os betwee n two clas s es of indi vi dua ls. Afterwards, sam ples were subjec ted to PCA analys is to determ i ne stati s ti cal differences between sam ples after oils intake and controls for each interventi on breakfas t. For thi s study, sam ples were clas si fied as t0 (basal state , obtained befo re intake) for control indi viduals , t2 (pos t - basal state, obtai ned 2 h from intake) and t4 (pos t - bas al state, obtained 4 h from intake). Figure 1 shows the PCA scores plots for each study. These analys es showed that t2 sam ples were not dis c r im i nated from controls for any of interventi on breakfas ts , exc ept for VOO that sho ws a partial separati on between control indi vi dua ls and t2. On the other hand, t4 sam ples were clearly grouped and separated from controls in all cas es . This study led to con c lu de that intake has a metaboli c effec t on the uri nary fi ngerprints for all types of cons um ed breakfas ts but this effec t cou ld be only detected after 4 h from intake. Therefore, t4 sam ples were subs equently us ed to study stati s tic al diffe renc es among diet s by PCA and PLS - DA. Control sam ples were not inc luded as independent group in furthe r studi es to reduce the num ber of clas ses cons i dered in each analys is (namely, VOO, NSO, PSO and DSO). However, spec tral regions res pons i ble for differences (D) 440 N uevas plat a for ma s ana l ?t ica s en met a bol ?m ica am ong class es were evalua ted by com pari son of t4 and control groups in a final step. In a sec ond analys is , samples from the four interventi on breakfas ts were used in an ini tial PCA to explore poss i ble grou pings. The scores plot Figure 1. PCA s co res plot s obt ained by compariso n of t0 , t2 and t4 sampl es for each interventio n breakf ast . (C) (D) (A) (B) 441 Cha p t er 12 (Fi gure 2.A) dem ons trated that differenc es exhi bi ted among the four clas ses did not caus e any separati on and indi vi duals formi ng the cohort were dis persed in the three - dim ensi onal space, whic h explai ned only 40% of the overall vari abi li ty. A sli ght clus teri ng was obs er ved for VOO interventi on breakfas t. With thes e premi s es , PCA cou ld not explai n differenc es among breakfas ts . A simi lar pattern has been found in the literature, where mos t of studi es carried out with uri ne sam ples have requi red supervi sed analysis (mai nly by PLS - DA) for developm ent of clas si fic ation and predic tion models . Thi s can be jus ti fi ed bec aus e metaboli tes are freque ntly found at low concentration and the large vari abi li ty among indi viduals may mask vari abili ty from other sourc es or sti mul i. 3.2. P art ial least squares discrimi n an t analysis After non - s upervis ed analysis , PLS - DA was needed to attain disc rimi nation am ong groups and to create class i fi cation and predic ti on models . The PLS - DA scores plot obtai ned with the three firs t latent vari ables is shown in Figure 2.B. A certai n clus teri ng of indi vi duals attendi ng to the interventi on breakfas t can be vis ua lized. However, only indi vi dua ls after intake of breakfas t prepared with VOO can be clearly separated. The predic ti ve acc urac y was us ed to evaluat e the capabili ty of the PLS model , whic h is defi ned as the perc entage of correc tly clas si fi ed experim ents in a given clas s. This acc urac y is repres ented as a confusi on matri x, with the true clas s in rows and the predi c ted clas s in colum ns . The diagonal con tai ns the correc tly class ifi ed sam ples. The acc urac y matri x is diffe rent for trai ni ng and cros s - validation, and both joi ntly desc ri be the predic ti on capabili ty of the model. Table 1 shows the acc u rac y matri x for cross - vali dati on (a) and traini ng (b) for e ach clas s and the overall (mean) ac c urac y. PSO led the worst accurac y for trai ning (20%), whereas NSO did for vali dati on (54%). 442 N uevas plat a for ma s ana l ?t ica s en met a bol ?m ica T he overall accurac y of the model was 49% and 87% for trai ni ng and predic ti on; therefore, the model was inadequa te for predic ti o n. To the view of thes e res ul ts , the following schem e was to simpli fy the model to achi eve total separation and fi nally, to find spec tra regi ons that caus es differenc es am ong clas ses . Acc ordingly, a sec ond PLS study was carri ed out by com pari ng class es in groups of three. Table 2 shows the ranges and overall ac curac i es obtai ned for eac h study. As can be seen, the overall acc uraci es ranged from 40.5 to 70.5% for vali dati on and from 92 to 100% for trai ni ng. Interesti ngly, PSO, DSO and VOO, whi ch are the group s bas ed on oils with riche r oxi dati on inhi bi tors content, gave the best resul ts , indic ati ng that thei r com pos i ti on affec t differently to the uri nary fi ngerprint. By explori ng the PLS scores graphic s, DSO sam ples were random ly dis pers ed in the three analys e s givi ng mos t of fals e posi ti ves by misc las si fi c ati on in the othe r three groups . F ig ure 2. Scores plot o bta i ne d wit h the PCA (A) an d PLS ( B) fr om the ana l ysis of the four int er ve nt io n br ea kfas t s . As the PLS models usi ng three clas ses did not lead to com plete clas si fic ati on, groups were com pared in pai rs with the final aim of obtai ni ng ( A ) (B) 443 Cha p t er 12 Tabl e 1. Charact erizati o n of the PLS model obt ained with the anal ysis of the four classes for val idatio n (a) and training (b). (a) Tabl e 2. Ranges and overal l acc uracies obt ained for val idati o n an d training for the PLS - DA models obtained fro m co mpariso ns in gro ups of three clas s es [NSO] (Pred ic ted) [PSO] (Pre d ic ted) [DSO] (Pred ic ted) [VOO ] (Pred ic ted) Acc ura c y [NSO] (T ru e) 6 1 3 3 46.1538 5 [PSO] (T rue) 1 5 5 1 55.5555 6 [DSO ] (Tru e) 3 5 2 0 20.0000 0 [VOO ] (T rue) 2 1 2 10 66.6666 6 Ove ra l l acc ura c y 48.9361 7 [NSO] (Pred ic ted) [PSO] (Pred ic ted) [DSO] (Pred ic ted) [VOO ] (Pred ic ted) Acc ura c y [NSO] (T ru e) 7 1 3 2 53.8461 5 [PSO] (T rue) 0 9 0 0 100.000 [DSO] (Tru e) 0 0 10 0 100.000 [VOO ] (T rue) 0 0 0 15 100.000 Ove ra l l acc ura c y 85.455 Accuracy (%) NSO, PSO,DSO NSO,PSO, DSO NSO,VOO, DSO PSO,NSO, VOO Validation Range 20-53.8 38.4-73.3 23-77 66.6-88.8 Overall 40.625 60.5 43.2 73.5 Training Range 92.3-100 76.9-100 93.3-100 100 Overall 96.875 92.1 97.3 100 (b) 444 N uevas plat a for ma s ana l ?t ica s en met a bol ?m ica Figure 3. PCA - DA s co res plo t s of the comparati ve study of cl asses in pairs. 445 Cha p t er 12 the spec tra regi ons dis tinc ti ve of each clas s. Thus , six PLS analys es were carri ed out, whose ac curac y resul ts are sum marized in Table 3. Separati on was exc ellent for all cas es , with overall ac curaci es rangi ng from 84% to 100%. Figure 3 contai ns the PLS scores plots for each study, sho wi ng a com plete separati on and grouping among class es. Loa di ngs from these studi es were used to extrac t the chemi cal shi fts that explai ned vari abi li ty as soci ated to oils intake. 3.3. Peaks assign men t After explori ng the PLS models , the bins that caus ed separati on am ong clas ses served to know the spec tra regi ons that caus ed stronger vari abili ty associ ated with breakfas t intake. The total num ber of bins was ini tially 151 and are lis ted in Table 3 for each PLS - DA study. Then, bins were fi ltered accordi ng to an algori thm bas ed on fold change rati o , so the hi ghe r fold change, the larger diffe renc es between com pared interventi on breakfas ts . After this filtrati on step, bins were reduc ed to 61, then used for develo pm ent of PLS models. Table 4 lis ts the mos t signi fi cant bins , express ed as mean chemi c al shi ft valu es in ppm , contri buti ng to disc rimi nate among interventi on breakfas ts . The sign of the fold change ratio is als o inc luded for each bin. Figure 4 corresponds to the NMR spec tra obtained with an urine sam ple. The spec trum has been zoomed and divi ded in three parts , corres pondi ng to 0.5 to 2.5 ppm , 2 to 4.5 ppm and 6.5 to 9 ppm (spec tra regi ons exc lu ded in the stati s ti c al analys is were not consi dered). The mai n signals were assi gned with the Chenom x NMR Profi ler software, by fi tti ng the experim ental spec trum with re ferenc e spec tra from the database. The software profi led all peaks corres ponding to the sam e com pound cons i deri ng thei r rati o as a function of conc entrati on. Among the identi fi ed peaks, creati ne, creati ni ne, hyppurate and citrate were the mos t abundant met aboli tes . Other minor com pounds , such as ami no 446 N uevas plat a for ma s ana l ?t ica s en met a bol ?m ica ac i ds (alani ne, tryptopha n, glutam ate) methylhis ti di ne, phenilacetate, gluc os e, lac tos e, niac iamide and form ate were also detected. Exam i ni ng the chemi c al shift range (ppm ) of the signific ant bins indic ated th at they were mai nly loc ated withi n 0.5 ?1.8 ppm , where glutam ate and signals from lipids were ass i gned; 8 ?9 ppm , with contri buti on of formate and monos ubs ti tuted am i no groups (1 - methylni c oti nam i de, niaci nami de and methylhis ti di ne); 2 ? 3 ppm that acc ou nted fo r citrate, creati ne, creati nine, and other com pounds contai ni ng ami no groups (ami no aci ds previ ous ly mentioned); and around 6 ?7 ppm, tha t corres ponded to arom ati c metaboli tes. 4. Co nclusi ons A metabolomi cs fi ngerpri nti ng approac h bas ed on NMR analys i s of uri ne sam ples has been us ed to com pare the metabolic effec t caus ed by intake of interventi on breakfas ts prepared with different fried oils . Uri ne sam ples were taken in bas al state and 2 and 4 h after intake of meals . Multi vari ate analys is of sam ples w as carri ed out with PCA and PLS - DA to reveal : (i) Pos t - basal s amples were only dis c rim i nated from bas al sam ples at 4 h after breakfas ts intake; (ii ) PLS - DA enabled to dis crim i nate among interventi on breakfas ts if two class es were com pared, with predic tion ac curac i es from 84% to 100%. In a final step, chemi c al shi fts responsi ble for disc rimi nation were extrac ted. After a tentati ve identi fi c ati on of the main signals from NMR spec tra, thes e chem ic al shi fts were attri buted to exogenous or endogenous groups of m etaboli tes pres ent in uri ne. Fig ure 4. Peaks assignment in the NMR anal ysis of a urine sampl e 4 hours aft er intake of breakf ast prepared with fried VOO. N uevas plat a for ma s ana l ?t ica s en met a bol ?m ica Fig ure 4. Peaks assignement in the NMR anal ysis of a urine sampl e 4 hours aft er intake of breakf ast prepared with fried VOO (continued) . Tab l e 3. Ranges and overal l acc uracies obt ained for val idati o n a nd training for the PLS - DA models and total bins obt ained used f or the mo del s fro m compariso ns in pairs Accura cy (% ) NSO vs PSO NSO vs D SO NSO vs VOO PSO vs VO O PSO vs DS O DSO vs VO O Val ida tion Ra nge ??????? ???????? ?????? ?????? ??????? ???????? Overa l l 73.9 86.3 85.7 80 89.4 83.3 Tra ini ng Ra nge 100 100 100 100 ???????? 100 Ove ra l l 100 100 100 100 94.7 100 Bins 41 60 38 54 42 60 VOO vs PS O NSO vs PSO VOO vs DS O NSO vs D SO DSO vs PS O NSO vs VO O 9.363 u p 6.307 up 1.583 dow n 6.477 up 6.296 dow n 5.404 up 1.617 dow n 9.735 up 9.227 dow n 6.443 up 6.169 dow n 4.421 dow n 1.583 dow n 9.702 up 1.549 dow n 8.039 up 4.64 dow n 4.333 up 1.549 dow n 9.43 up 0.701 dow n 9.193 up 4.529 dow n 3.506 up 6.376 up 0. 26 up 0.294 dow n 8.446 up 4.098 up 3.353 dow n 0.531 dow n 1.821 up 9.193 dow n 9.363 up 3.994 dow n 3.048 up 6.24 up 1.617 up 0.633 dow n 3.111 dow n 3.947 dow n 3.042 up 6.41 up 1.685 up 6.376 up 8.344 up 3.917 dow n 3.040 dow n 6.477 up 9.193 up 2.228 dow n 6 .953 up 3.897 dow n 3.004 up 9.057 up 8.616 dow n 0.361 dow n 6.715 up 3.815 up 2.919 dow n 2.194 dow n 8.344 up 9.464 dow n 6.307 up 3.78 dow n 2.796 up 0.735 up 0.599 up 9.532 dow n 8.616 dow n 3.742 dow n 2.783 up 8.446 up 8.955 up 9.091 dow n 0.667 up 3.31 up 2.654 dow n 0.769 up 0.667 up 8.955 dow n 6.647 up 3.281 dow n 2.534 up Tabl e 4. More significant bins and the sign of regul atio n for each compari son obtained fro m compariso ns in pairs 450 N uevas plat a for ma s ana l ?t ica s en met a bol ?m ica 5. Ackno wledgement s The Spanish Mi nis terio de Ciencia e Innovac i?n (MICI NN ) and FEDER program are thanked for fi nanci al support through projec t CTQ2009 - 07430. B.A. - S. and F.P. - C. are als o gratefu l to the MICINN for an FPI scho larshi p (BES - 2 0 0 7 - 15043) and a Ram ?n y Cajal contrac t (RYC - 2009 - 03921), respec ti vely. 6. Referenc es [1] R. Kaddurah - Daou k, B.S. Kri s tal, R. M. Wei ns hi lboum , Ann u. Rev. Pharmacol. Toxi col. 48 (2008) 653 . [2] O. Fi eh n. Plant Mol. Biol. 48 (2002) 155. [3] W.B. Dunn, D.I. Elli s. Trends Anal. Chem ., 24 (2005) 285. [4] D.I. Elli s , W.B Dunn , J.L. Gri ffi n , J.W. Allwood , R. Goodac re , Futu re Medic i ne, 8 (2007) 1243. [5] T.F. Bathen, L.R. Jensen, B. Sitter, Breas t Cancer Res. Treat. 104 (2007) 181. [6] I.S. Gri bbes tad, B. Si tter, S. Lundgren, Antic ancer Res. 19 (1999) 1737. [7 ] Y.L. Wang, H.R. Tang, J.K. Nicho ls on, P.J. Hylands, J. Sam pson, I. Whi tc om be, C.G. Stewart, S. Cai ger, I. Oru, E . Holmes , Planta Med 70 (2004) 250. [8] J.L McCla y , D.E. Ad kins , N.G. Isern, T.M. O'Connell, J.B. Wooten , B.K. Zedler, M.S . Das i ka, B.T. Webb, B.J. Webb - Robertso n, J.G. Pounds , E.L. Murrelle, M.F. Leppert, E.J. van d en Oord, J. Pr oteome Res. 9 (2010) 3083. 451 Cha p t er 12 [9] E.C. Yong Chan, P.K. Koh, M. Mal, P.Y. Cheah, K.W . Eu, A. Bac ksha ll, R. Cavi ll, J.K. Ni cho ls on, H.C. Keun, J. Proteome Res. 8 ( 2009) 352. [10] M.R. Viant, C. Ludwig, U.L. G?nther, Metabolomi cs , Metabonomi c and Metabolite Pro fi li ng. RS C Publi shi ng, Cam bri dge. (2008) 44. [11] E. Holm es , P.J.D. Foxall, M. Spraul, R.D. Farrant, J.K. Nic holson, J.C. Lindon, J. Pharm. B iomed. Anal. 15 (1997) 1647 . [12] E. Holm es , P.J.D. Foxall, J.K. Nic holson, G.H. Nei ld, S.M. Brown, C.R. Beddell, B.C. Sweatm an, E. Rahr, J.C. Lindon, M. Spraul, P. Nei di g, A nal. Biochem. 220 (1994) 284 . [13] K.S . Solanky, N.J. Bai ley, B.M. Bec kwi th - Hall, S. Bingham , A. Davis , E. Holmes, J.K. Ni cho ls on, A. Cass idy, J. Nu tri tional Bioc hem . 16 (2005) 236 . [14] R. Llorac h , I. Garrido , M. Monagas , M. Urpi - S arda, S. Tuli pani , B. Bartolom ? , C. Andr?s - La cue va , J. Proteome Res. 9 (2010) 5859 . [15] R. Llorach , M. Urpi - Sard? , O. Jauregui , M. Mo nagas , C. Andr?s - Lacue va, J . Proteome Res . 8 (2009) 5060. [16] M.A. Cons tanti nou , E. P apakons tanti nou, M. Spraul, S. Sevasti adou , C. Cos talos , M.A. Koupparis , K. Shu lpi s , A. Tsantili - Kakou li dou , E. Mikros , Anal. Chim . Acta 542 (2005) 169. [17] A.D. Maher, S.F.M. Zirah, E. Holm es , J.K. Nic ho ls on, Anal. Chem . 79 (2007) 5204. [18] M. Lauri ds en , S.H. Hansen, J.W. Jaroszews ki, C. Cornett, Anal. Chem . 79 (2007) 1181. [19] M.L. Anthony, B.C. Sweatm an, C.R. Beddell, J.C. Lindon, J.C. Nic holson, Mol. Pharm ac ol. 46 (1994) 199. [20] E. Holm es , J.K. Nicho lson, A.W. Nicho lls , J.C. Lindon, S.C. Connor, S. Polley, J. Connelly, Chem om. Intell. Lab. Syst. 44 (1998) 245. 452 N uevas plat a for ma s ana l ?t ica s en met a bol ?m ica [ 2 1 ] F . P?rez - Ji m?nez , A. Es pi no , F. L?pez - S egura , J. Blanc o , V. Rui z - Guti ?r rez, J.L. Prada, J. L?pez - Mi randa , J. Jim ?nez - Perez, J.M. Ordovas Am . J . Cli n . Nutr . 62 (1995) 769 . [22] D.M. H egs t ed, R.B. McGandy, M.L. Myers , J. St are, Am . J . Cli n . Nutr . 17 (1965) 281 . [23] O. Morei ras - V arela Eur . J . Cli n . Nutr . 43 (1989) 83 . [24] R. Jap ?n - Luj?n , M.D. Luque de Cas tro, J . Agric . Food Chem . 56 (2008) 2505. [25] M. Yui lle, T. Illi g , K. Hveem , G. Schmi tz , J. Hans en , M. Neum ai er , G. Tybri n g , E. Wi chm ann , B. Ollier, Biopres erv Biobank 8 (2010) 65. [26] J.L Izqui erdo - Garc ?a, I. Rodr? gue z, A. Kyriazi s, P. Villa, P. Barrei ro, M. Desc o, J. Rui z - Cabello, BMC Bioi nform ati cs 10 (2009) 363 . [27] A. Ross , G. Schlo tterbec k , F. Dieterle , H. Senn , The Handboo k of Metabonom ics and Metabolo mi cs Edi ted by: J.C. Lindon , J.K. Nic hols on, E. Holmes , Ams terdam, ELSEV IER; 2007, 96 . [28] E. Holm es , P.J.D. Foxall , J.K. Ni cho ls on, Anal . Bioc hem . 220 (2002) 284 . [29] C.L. Gavagha n , E. Holm es , E. Lenz, I.D. Wils on, J.K. Nic ho ls on, FEBS Lett. 484 ( 2000) 169 . CHAPTER 13: Global metabolomics profiling of human urine by LC?TOF/MS in accurate mode to evaluate the intake of breakfasts prepared with fried edible oils Global metabolomics p rofili ng of human urine by LC ?TOF/MS in accurate mode to evaluate the inta ke of b reakfasts prepared with fried e dible o ils B. ?lvarez-S?nchez, F. Priego-Capote, M.D. Luque de Castro Department of Analytical Chemistry, Annex Marie Curie Building, Campus of Rabanales, University of C?rdoba, E-14071 C?rdoba, Spain Institute of Biomedical Research Maim?nides (IMIBIC), Reina Sof?a Hospital, University of C?rdoba, E-14071 C?rdoba, Spain Sent to Analytical and Bioanalytical Chemistry for publication 457 Sent to Anal. Bioanal. Chem. Chapter 13 Glo bal metabo lomics p rofili ng of human urine by LC ?TO F/M S in accur ate m o de to evaluate the intake of b reakfasts p rep are d wit h frie d edib le o ils B. ?lvarez-S?nchez, F. Priego-Capote, M.D. Luque de Castro* Abst ra ct Epidemi olo gi c studi es have dem ons trated a direc t relationshi p between coronary heart dis eas e and diet as a con sequenc e of fats consum pti on. The aim of this study was to obtai n the metaboli c profi le of uri ne after a single intake of breakfas ts prepared with fri ed edi ble oils from different sou rc es . Thus , meals prepared with four different fri ed oils (oli ve oi l, pur e sunflower oi l, sunflo wer oil enri che d with oli ve phenols as natural anti oxi dants and sunflower oi l enri che d with the syntheti c oxi dati on inhi bi tor dim ethylsi loxane) were us ed in this study. A metabolom ic s global profili ng approach has been us ed to reveal stati s ti c al differences in the metabolom e of uri ne from indi vi duals after intake of the diffe rent breakfas ts and to identi fy com pou nds responsi ble for such vari abili ty. Uri ne sam ples obtained befo re (blank) and 2 and 4 hours after intake were analyzed by LC ? T OF/MS in accurate mode. Pri nci pal Com ponent Analysis (PCA) and Parti al Leas t Square Dis c rim i nant Analysi s (PLS - DA) were carri ed out to establi sh differences among clas ses (intake of breakfas t prepared with eac h oil) and to bui ld a class i fi cation and pr edic tion model. The PCA analys is showed a clear separation between uri ne from control indi vi duals befo re breakfas t consum pti on and uri ne from indi vi duals 4 h after intake. On the other hand, the PLS - DA model allowed the development of class i fi cation models with ac ceptable acc urac y param eters . The mass features obtai ned from this analys is were used to identi fy metaboli tes res ponsible for dif ferences among clas s es and to infer the pos si ble metaboli c pathways . 458 Nuevas plataformas anal?ticas en metabol?mica 1. In t r o d u c t i o n O v e r t h e p a s t c e n t u r y , n u t r i t i o n a l s t u d i e s we r e f o c u s e d o n alimentary ?deficiencies? by identifying compounds that were essential f o r h u m a n g r o w t h a n d h e a l t h . Ho w e v e r , t h e c h a n g e i n l i f e s t y l e h a s b r o u g h t a b o u t a d r a m a t i c c h a n g e i n n u t r i t i o n a l m e d i c i n e , a n d c u r r e n t n u t r i t i o n a l s c i e n t i s t s a r e n o w c h a l l e n g e d wi t h f i n d i n g ne w w a y s o f treating or preventing diseases caused by nutritional ?oversufficiencies? s u c h a s o b e s i t y , d i a b e t e s , c h r o n i c i n f l a m m a t i o n a n d c a r d i o v a s c u l a r d i s e a s e s . T h i s n e w a p p r o a c h wo u l d i n t e n d t o i d e n t i f y b i o a c t i v e f o o d c o m p o n e n t s t h a t po t e n t i a l l y i n c r e a s e l i f e e x p e c t a n c y , r e d u c e w e i g h t , e n h a n c e ph y s i c a l o r m e n t a l p e r f o r m a n c e a n d p r e v e n t t h e a b o v e d i s e a s e s c h a r a c t e r i s t i c o f de v e l o p e d c o u n t r i e s [ 1 ] . Among them , coronary heart disease (CHD) is the most prevale nt caus e of death in develo ped cou ntries [2]. Epidemi olo gi c studi es have proved a direc t connecti on between diet and CHD, as is the cas e of fat cons um ption, whic h has shown to affec t hi gh - and low - densi ty - li poprotei n cho les terol (HDL and LDL) concentrations. Des pi t e cons um ption of fat - ric h diets (35 ?40% of total energy) [3], Mediterranean popul ations have been charac teri zed by a low prevale nc e of CHD , sugges ti ng that diet in this area is healthi er or, at leas t, more likely to protect agai ns t CHD throughou t the effec t of bioac ti ve com pou nds [4]. In fac t, oli ve oil, the princ i pal sou rc e of fat in the Mediterranean diet, is known to exert a protec ti ve effec t agai ns t heart dis eas es. The caus e of these potenti al benefi ts has been as c ri bed to both an adequa te fatty ac id pr ofi le and the pres ence of nutrac eu ti cals suc h as phenols, many of whi ch are well - known anti oxi dants [5]. Thu s, oli ve oil is mai nly cons ti tuted by glyc erols and other minor com ponents suc h as ali pha tic and triterpenic alc ohols , sterols , carotenoi ds and phen olic anti oxi dants , that can be both lipophi lic and 459 Sent to Anal. Bioanal. Chem. Chapter 13 hydrophi lic [6]. Interes ti ngly, whereas lipophi li c phenols such as tocophe rols can be found in other vegetable oils, mos t hydrophi li c phenols [7,8] found in oli ve oil are absent in other oils [9]. Evidence dem onstrates that these hydrophi lic phenols are powerfu l anti oxi dants , both in vitro and in vivo [10]. Although further studies are requi red, thes e com pounds might be goo d candi dates as func ti onal food com ponents to protec t agai ns t CHD. Although benefi t s of oli ve oil and, conc retely, of phenolic anti oxi dants are gai ni ng interes t, the mecha nism s involved in this protec ti ve effec t are sti ll unknown. Therefo re, it coul d be worth explori ng how metabolis m is affec ted by oi ls intake and, parti c ularly, how oli v e oil contri butes to CHD preventi on. I n th i s s e n s e , m e t a b o l o m i c s po s e s p r o m i s i n g pr o s p e c t s , a s d e m o n s t r a t e d b y th e i n c r e a s i n g n u m b e r o f a p p l i c a t i o n s o n h u m a n n u t r i t i o n , t h a t h a s l e d t o gi v e t h e n a m e o f nutrimetabolomics to this area of metabolomics) ??????] . T h e m o s t p r o m i s i n g a n d u n e x p l o i t e d a p p l i c a t i o n o f nu t r i m e t a b o l o m i c s i s pr o b a b l y i t s u s e f o r m o n i t o r i n g d i e t i n t e r v e n t i o n . T h i s wo u l d a l l o w d e t e r m i n i n g m e t a b o l i c c h a n g e s r e l a t e d t o a d m i n i s t r a t i o n o f c e r t a i n nu t r i e n t s t h a t h a v e a n i m p a c t o n h u m a n h e a l t h . C o n c r e t e l y , m e t a b o l o m i c s f i n g e r p r i n t i n g a n d pr o f i l i n g h a v e be e n u s e d t o de t e r m i n e t h e e f f e c t s o f f o o d s u p p l e m e n t a t i o n a n d d i e t i n t e r v e n t i o n o n m e t a b o l i s m . T h u s , t h e s c h e m e f o l l o w e d i n t h e s e s t u d i e s e n t a i l s : (i ) s e l e c t i n g a n a d e q u a t e f o o d m a t e r i a l f o r d i e t i n t e r v e n t i o n a s we l l a s t h e a p p r o p r i a t e po p u l a t i o n t o p a r t i c i p a t e i n t h e s t u d y i n o r d e r t o o b t a i n s t a t i s t i c a l l y r e p r e s e n t a t i v e r e s u l t s , (i i ) a n a l y z i n g bi o s a m p l e s t a k e n a f t e r f o o d i n t a k e u n d e r t h e s a m e w o r k i n g c o n d i t i o n s ; (i i i ) u s i n g no n - s u p e r v i s e d o r s u p e r v i s e d s t a t i s t i c a l a n a l y s i s t o f i n d d i e t - r e l a t e d d i f f e r e n c e s b e t w e e n g r o u p s ; (i v ) d e t e r m i n i n g a n a l y t i c a l f e a t u r e s r e s p o n s i b l e f o r t h e s e d i f f e r e n c e s , wh i c h m a y be a r e g i o n i n th e nu c l e a r m a g n e t i c r e s o n a n c e (N M R ) s p e c t r a o r a m o l e c u l a r e n t i t y l i n k e d t o a n m/z v a l u e f r o m a m a s s s p e c t r u m , ( v ) i d e n t i f y i n g r e l a t e d m e t a b o l i t e s , wh i c h i m p l i e s t o qu e r y a g a i n s t a v a i l a b l e d a t a b a s e s ; a n d ( v i ) e l u c i d a t i n g m e t a b o l i c p a t h w a y s i n o r d e r t o g i v e a s m u c h b i o l o g i c a l i n f o r m a t i o n a s po s s i b l e [ 1 7 ] . T h e s e s t e p s a r e p o s s i b l e i f a 460 Nuevas plataformas anal?ticas en metabol?mica s e n s i t i v e a n d a c c u r a t e a n a l y t i c a l pl a t f o r m i s u s e d s i n c e m e t a b o l i c d i f f e r e n c e s a m o n g f i n g e r p r i n t s a r e u s u a l l y s m a l l . In thi s sense, mass spec trom etry (MS ) or NMR are usually the preferred choi ces [18]. NMR allows direc t analysis of mos t types of sam ples, inc lu di ng tis sues ; allows the recovery of sam ples for furthe r analyses , and provi des detai led inform ation on molec ular struc ture [19]. However, NMR is charac teri zed by a poor sensi ti vi ty as com pared to MS, and spec tra from com plex mixtures as biolo gic al sam pl es can be very diffic ul t to interpret due to overlapping among metabolite signals ???????? On the other hand, MS possesses high s ensi ti vi ty and selec ti vi ty and provi des accurate mas s meas urements thanks to rec ent improvem ents of mas s analyzers . This tec hni que , usua lly cou pled with ups tream separation techni ques like gas chrom atography (GC) [23] or liquid chrom atography (LC), enables unequi voc al identi fi cati on of com pounds and struc ture eluci dati on of a large num ber of metaboli tes, provi di ng extra informati o n about the bioc hemi cal relati ons hi ps between them [24]. For thi s reason, MS is the preferred choi ce ei the r to obtain metabolic profi les or to identi fy com pounds res ponsi ble for metabolic changes . Wi thi n the nutri metabolomi cs context, biolo gic al flui ds su c h as plasm a, serum and uri ne [12,13,16] have been com monly used, as they are easi ly and non - (or scant) invas i vely collec ted, direc tly reflec t the global state of an indi vidual and/or the res ponse to drug treatment or diet intake; us ually requi ri ng the ap pli c ati on of sim ple sam ple preparation protocols . It is expec ted that metabolic changes due to diet lead to changes in the uri nary profile. The aim of this res earc h was to determi ne short - term changes in the metabolic profi le of uri ne after intake of a b reakfas t prepared with fri ed oi ls . Wi th this aim , oi ls with different com pos i ti on, particul arly with different anti oxi dants profi le, were us ed in order to detect diffe rences in the metabolis m reflec ted in the uri nary profile. The cohort selec ted for thi s s tudy was com pos ed by obese indi vi dua ls to find metabolic differences 461 Sent to Anal. Bioanal. Chem. Chapter 13 c orrelated with fat cons um ption. A final identi fic ati on of metaboli tes explai ni ng the vari abi li ty observed among groups was also intended to obtain informati on about the metaboli c pathway s affe c ted by oils intake. 2. Experimenta l 2.1. Reagents LC ?MS grade acetonitri le, MS - grade form ic aci d (Scha rlab, Barc elo na, Spain) and deioni?ed water (??M??cm) from a Millipore Milli- Q water purifi cati on sys tem (Mi llipore, Bedford, MA, USA) were used to prepare the chrom atographi c phases . 2.2. Instruments and apparatus Centrifugation was carri ed out with a thermos tated centri fu ge Thermo Sorvall Legend Mic ro 21 R from Thermo (The rmo Fis he r Sci enti fic, Brem en, Germ any) . A Vac ufuge centri fugal vacuum concentrator from Eppendorf ( Eppendorf , Inc ., Hamburg, Germany ) was us ed to evaporate sam ples to dryness . All sam ples were analyzed by an 1200 Seri es HPLC sys tem (Agi lent Tec hnologi es , Waldbronn, Germ any) equi pped with a binary pum p, a degass er, a well plate autos am pler and a therm os tated colum n com partm ent, whi ch was cou pled to an Agi lent 6530 TOF mass spec trom eter equi pped with a Jetstream ? ESI sou rc e . Mass Hunter Works tation software (Agi lent Technologi es ), inclu di ng Data ac qui si ti on, Quali tati ve Analys is and Mass Profi ler Profes si onal, was used for proc ess i ng raw MS data, inc lu di ng feature extrac ti on, data fi lteri ng, stati s ti cal analysi s 462 Nuevas plataformas anal?ticas en metabol?mica by ANOV A and PCA, followed by the cons truc tion of PLS predi c ti on model, molec ular formul a and database searchi n g. Compound identi fic ation was perform ed usi ng the MET LIN Personal Metaboli te Database throu ghout the IDbrowser tool and the Molec ul ar Formul a Generati on algori thm (Agi lent Tec hnologi es ). The Human Metabolome Database (HMDB) was used to confi rm and extend identifi cati on. 2.3. Oils and heating procedure Four edi ble oils were prepared for thi s study. Two of them were com merci ally obtained and the others were enric he d with two types of oxi dation inhi bi tors : a synthetic oxi dation inhibi tor (dim ethylsi loxan e, DMSO, at 400 ?g mL ?1 ) and a natural extrac t from oli ve pom ac e contai ni ng phenoli c anti oxi dants obtai ned by a protoc ol sim i lar to that developed by Gir?n et al . [25]. The purpos es of the enri chm ent of refi ned edi ble oi ls with oxi dation inhi bi tors were to enhance oi ls stabili ty. The edible oi ls were: (1) V OO: Di fferent extra - vi rgi n oli ve oils mixed up to a total concentration of phenoli c anti oxi dants of 400 ?g mL ?1 , express ed as ?g mL ?1 of caffeic acid by the Folin?Ciocalteu (F ?C) test. The oi l was com pos ed by 70.5% monouns aturated fatty ac i ds (MUFAs ), 11.1% PUFAs , 18.4% saturated fatty ac i ds ( SFAs ); (2) N S O: non - enric hed refi ned sunflower oi l with 34.3% MUFAs , 58.3% PUFAs and 7.3% SFAs . Du e to the refi ni ng proc ess , the concentration of hydrophi lic anti oxi dants was null. (3) PSO: refi ned high - oleic sunflower oi l enric hed with phenols from oli ve - pom ace extract up to 400 mg L ?1 , als o e xpress ed as caffei c acid, 76. 7 % MUFAs , 17.6% PUFAs and 5.8% S FAs . 463 Sent to Anal. Bioanal. Chem. Chapter 13 (4) DS O: refi ned high - olei c sun flo wer oil enri che d with 400 mg L ? 1 of dim ethyls i loxane as an oxi dati on inhi bi tor, 71.8% MUFAs , 18.0% PUFAs and 10.2% SFAs . Two liters of the target oil was plac ed in a stai nles s - s teel deep fryer. The oil was heated at 180 ? C ? 5 ?C for 5 min ten t im es every day for two days (total heati ng cyc les : 20) . The fried oils were used to prepare four types of baked muf fi ns (one per fried oil) with conventi onal ingredi ents and clas si fied for the experi mental plan. 2.4. Sub jects and samples Experim ent was planned followi ng the guideli nes dic tated by the World Medi c al Ass oc i ation Dec larati on of Hels inki (2004), whi ch were supe rvi sed by the ethic al revi ew board of Reina Sofi a Hos pi tal (C? rdoba, Spai n) that approved the experi ments . Se venteen obese indi vi duals with a body mass index between 30 ?40 kg m ?2 form ed the cohort in thi s study. All of them gave their inform ed conse nt and underwent a com prehe nsi ve medi cal his tory, phys ic al exam i nati on and cli nic al chemi s try analysi s befo re enrolm ent. Partic i pants with evidence of kidney, panc reas , lung, liver or thyroi d dis eas e were exc lu ded. All subjec ts were non - diabetic s, non - s mokers , without cli nic al mani fes tations of cardi ovas cul ar dis eas e and off treatment. The target cohort was com pos ed by 10 post - menopaus al women, age 48 ?70 years , and 7 men, age 39 ?70 years . None of the subjec ts was taking medi cati on or supplem entary vitam i ns with influ enti al effec t on uri ne metabolom e. All volu nteers recei ved fou r breakfas ts in muffi n form at prepared wit h the fou r different oils (0.45 mL of oil per kilogram of body weight), previ ous ly subjec ted to the sim ul ated fryi ng proces s. The admi nis trati on of each breakfas t was held at random ization and followi ng a Lati n cross ed squa re desi gn, whi ch increas ed the po wer of the study. The volu nteers ate 464 Nuevas plataformas anal?ticas en metabol?mica each breakfas t every two weeks (4 oils , 8 weeks ). Duri ng the sam pli ng period (4 hours ) the volunteers did not cons ume any food. Sam pli ng was performed follo wi ng the Recom mendations on Biobanki ng Procedures for urine pr oc es si ng and management recently publishe d by the European Consensus Expert Group [26]. Accordi ng to this doc um ent, biobanki ng procedures for uri ne should cons i der the followi ng general cons ens us recomm endati ons : (i) cells and particul ate matter sho uld be removed ( e.g. by centri fu gati on); (ii ) sam ples shou ld be stored at ?80 ?C or below; (ii i ) tim e lim its for the proc ess i ng shou ld have been defi ned experim entally and sho uld be appropri ate for the analytes to be meas ured; (iv) unles s spec ifi ed for a partic ul ar downs tream analys is , uri ne sam ples shoul d be stored wi thout addi ti ves [27]. In order to obtain sui table control sam ples and carry out a tim e - c ou rs e study of uri nary exc reti on after dietary interventi on, sam ples were obtained at 0 (jus t before) and 2 a nd 4 h from intake. Uri ne was collec ted in sterile contai ners , ali quo ted in 2 - m L Eppendorfs , centri fu ged at 1500 g, 4 ?C for 10 min and stored at ?80 ?C unti l analys is. This protocol ens ures elim i nati ng partic ul ates, cells and bac teri a that may interfere a nalys is and quenc hi ng enzyma tic activi ty in uri ne duri ng storage. After thawing for analysis , sam ples were 1:4 diluted with dei oni zed water , and thei r pH adjusted to 4.8 with ammonium acetate. Thus , the uri ne sam ples were ready to be injec ted in the analy tic al ins trum ent 2.5. LC? TOF/MS analysis The LC ?T OF/MS ins trument used in this study was an Agi lent Seri es 1200 SL rapi d res olu tion LC sys tem (cons is ti ng of vac uum degass er, cooled autos am pler, rapi d resolu ti on binary pum p, and therm os tatted colum n 465 Sent to Anal. Bioanal. Chem. Chapter 13 c om partm ent) hyphe nated to an Agi lent 6530 accurate - m ass TOF MS with elec tros pray ionization (ESI) via Jet Stream Tec hnology (Agi lent Tec hnologi es , Santa Clara, CA, USA). Chrom atographic separation was perform ed usi ng a Teknokrom a Medi terranean Sea C18 anal yti cal colum n (5 mm ? 0.46 mm i.d., 3 ?m parti c le size, from Agilent Tec hnologi es , Santa Clara, CA, USA) kept at a temperature of 25 ?C. Uri ne sam ples were eluted with a binary gradient of (A) 0.1 % formic aci d in Milli Q water and (B) 0.1% formic aci d in ACN. The eluti on program was as follows : 0 ?1 min 2% B, 1 ? 20 min 100% B. A post - run of 5 min was inc luded to equili brate the colum n. The flow rate was mai ntai ned at 0.6 mmi n ?1 . The volum e of injec ted sam ple was 10 ?L. The operati ng conditi ons were as follo ws: gas tem perature, 325 ?C; dryi ng gas , nitrogen at 8 L min ?1 ; nebuli zer pres sure, 40 psi ; sheath gas temperature, 350 ?C; sheath gas flow, nitrogen at 11 L min ?1 ; capi llary voltage, 4000 V; skimm er, 65 V; oc topole radiofrequency voltage, 750 V; focusi ng voltage, 90 V. Data ac qui si tion (2.5 spec tra s ? 1 ; mass range 60 ?1100 m/z) was governed via the Agi lent Mass Hunter Works tation software. The ins trum ent gave typical resolu ti on 15000 FWHM (Ful l Wi dth at Half Maxi mum ) at m/z 112.9856 and 30000 FWHM at m/z 1 033.9881. To assure the desi red mas s accuracy of recorded ions , conti nuo us internal cali brati on was perform ed duri ng analyse s with the use of signals at m/z 121.0509 (protonated puri ne) and m/z 922.0098 [protonated hexakis (1H, 1H, 3H - tetrafluoropropoxy) p hos phazi ne or HP - 921] in posi ti ve ion mode. In negati ve ion mode, ions with m/z 119.0362 (proton abstrac ted puri ne) and m/z 1033.988109 adduc t of HP - 921 were used. The ins trum ent was cali brated and tuned accordi ng to proc edures recom mended by the manufac tu rer. 2.6. Data processing and statistical analysis 466 Nuevas plataformas anal?ticas en metabol?mica Onc e the sam ples were analyzed by LC/MS , data were extrac ted into molec ular features . Features extrac tion was bas ed on the extrac tion algori thm (MFE) that loc ates and groups all ions related to the same neutral molec ule. This relati on is referred as to the covariance of peaks withi n the sam e chromatographic retenti on tim e, the charge - s tate envelo pe, isotopic dis tri buti on, and/or the presenc e of adduc ts and dimm ers . The MFE took into accou nt all ions exc e edi ng 1000 counts, with charge state lim i ted to a maxim um of two and a peak spaci ng tolerance of 0.0025 m/z (plus 7 ppm ). Each feature was given by a minimum of two ions . The extrac ti on algori thm was bas ed on a comm on organic model with chrom atographi c se parati on. The allowed pos i ti ve ions were protonated spec i es and sodi um adduc ts , and the negati ve ions form ed by form ate adduc ts and proton loss es . Dehydratati on neutral los s es were als o allowed. Figure 1 shows the base peak chrom atograms obtai ned in negati ve ioni zati on mode with a blank sam ple (blac k line) and a sam ple obtai ned 4 h after oli ve oil intake (grey line). Figure 1. Base peak chromatograms, negative ionization mode, provided by a control sample (black line) and obtained 4 h after breakfast prepared with fried VOO (grey line). After extrac ti ng molec ular features , files in com pound excha nge format (.c ef fi les ) were created for each sam ple and exported into the Mass 467 Sent to Anal. Bioanal. Chem. Chapter 13 Profi ler Professi onal (MPP) software for stati s ti c al analys is . Molec ular formul a s were generated usi ng the Molec ular Formul a Generator algori thm . Both mass acc urac y and isotope inform ation (abundanc e and spaci ng) were consi dered to lim i t the num ber of hits. Only the com mon elem ents C, H, N, O, P and S were cons i dered in the generati on of formul as. The calc ulated neutral mass of each feature was queri ed agai ns t the MET LIN databas e [28] for matc hi ng to com pounds withi n a maxi mum mas s tolerance window of 10 ppm . The MET LIN databas e matche d the neutral mas s to the monois otopic mas s valu e c alcul ated from the em pi ri c al formula of com pou nds in the database. Identi fi ed compounds were eventua lly contras ted agai ns t the Human Metabolom e Databas e (HMDB) [29] to confi rm thei r occ urrence in hum ans. 3. Results and discussion 3.1. Scheme for data processing 3.1.1. Filtering Multi vari ate statis tic al analysis was carri ed followi ng a stepwi se schem e desc ri bed here. The proc ess is initi ated with a data set bui lt from the exported raw .cef files . The data set inc ludes all the molecul ar featu res extrac ted from each LC ?TOF/MS analysis , leadi ng to a huge num ber of data obtained from all sam ples (above 12000 molecul ar features ). Therefore, it is important to elim i nate redundant featu res to reduc e the dim ens ionali ty of the data set prior to Pri nc i pal Com pon ents Analys is (PCA) and Partial Leas t Squa re Dis c rim i nant Analys is (PLS - DA). For this purpose, the firs t step for data process i ng was to filter out all the non - i nfluenc i ng molec ular features from the data matri x, remai ni ng only thos e that could explai n the vari abi li ty as soci ated to breakfas ts intake and that wou ld allow sam ple clas si fic ati on 468 Nuevas plataformas anal?ticas en metabol?mica and screeni ng. Accordi ngly, three process i ng filters ?namely, by frequency or occ urrence in sam ples, by ANOV A acc ordi ng to a preset p- value , and by fold - c hange algori th m ? were applied and are di scussed here. Firs tly, the frequenc y filter aims at keepi ng all the MFs defi ni ng of a clas s . Thus , a ll the MFs that were pres ent in at leas t 50% of sam ples in at leas t one class (fi ve clas ses enc om pas si ng indi vi duals after intake of the four breakfas ts and the controls form ed by indi vi dua ls pri or breakfas t intake). Thi s means that MFs that were pres ent in the majori ty of sam ples of one group were kept sinc e they are supposedly defi ni ng the clas s. The cut - off of 50% in the frequenc y fi lter was supported on the exi s ti ng biologic al vari abili ty among indi vi dua ls who coul d be influ enc ed by num erous internal and external fac tors non - related with metabolism . The ANOV A - bas ed fi lter ai ms at keepi ng MFs stati s ti c ally signific ant to diffe rent iate between two clas s es by es tim ati on of p- valu es calc ulated for each MF by one - way ANOV A. The p- value cut - off of 0.05 was chos en to ensure that only MFs whic h differ am ong class es with statisti c al signific ance (95% in thi s partic ul ar cas e) are passed on and furthe r proc ess ed (the lower the p- valu e, the more signi fi c ant difference between the vari eti es ). The final MFs filtrati on step was perform ed usi ng fold change (FC) analys is , whi ch was employed to fi nd MFs with hi gh abundanc e ratios between two class es of indi vi duals (in this case diets and controls ). 3.1.2. Chemometric analysis PCA is freque ntly used as unsupervis ed pattern recogni ti on techni que enabli ng data dimensi onali ty reduc ti on, while retai ni ng maxim um vari abili ty of the data. This is perform e d via the trans form ati on of measured vari ables into unc orrelated princ i pal com ponents , each bei ng a linear com bi nati on of the origi nal variables [21]. Follo wing the above data pretreatm ent, PCA was employed in the firs t step of chem om etric analysi s for bot h posi ti ve and negati ve ioni zati on to evalu ate sam ple clus teri ng 469 Sent to Anal. Bioanal. Chem. Chapter 13 ac cordi ng to breakfas ts intake and to find trends of MFs (after fi lteri ng) explai ni ng the obs erved vari abili ty. Afterwards , PLS - DA, a widely used supervis ed pattern recogni tion tool capable of sam ple class predic tion, was used to bui ld and vali date a statis tic al model for clas s clas si fic ati on and predi c tion. The resul ts of sam ple clas si fic ati on are pres ented in terms of acc uracy in predic tion abili ti es , express ed as the perc entage of the sam p les correc tly class ifi ed duri ng model traini ng and cross - vali dati on, res pec ti vely. 3 .2. Time - course study It has been dem ons trated that the effec t of diet over metaboli sm persi s t s a certai n peri od of tim e, whi ch may be different among metaboli te clas s es . For exam ple, metaboli tes of coff ee are detec ted in uri ne collec ted 4 ? 5 h after coff ee inges ti on, whereas heteroc yc li c am i nes from gri lled meat persi s t after 48 ?72 ho after intake [14]. Therefore, it is important to carry out a tim e cou rs e study to sel ec t the tim e after breakfas ts intake that sho ws a hi gher inc i denc e on the uri nary profi le. As oils are comm on com ponents of meals, sam pli ng was carried out duri ng a fas t peri od (4 h) after oils intake to avoid influ ence of other sou rces . For the tim e - c ou rse study, sam ples were class i fi ed as t0 (blank sam ples , obtained before intake) as control indi vi duals, t2 (sam ples obtained two h from intake) t4 (sam ples obtai ned 4 h from intake). Sam ples were analyzed by both pos i ti ve and negati ve ionizati on modes, an d the data from eac h analys is proces sed followi ng the aforem enti oned schem e for extrac ti ng MFs . Enti ti es were filtered by frequency (50% cut - off ) and by fold change (2 - fold cut - off ). The num ber of molec ul ar features remai ni ng after application of thes e two algori thms ranged from 51 for DSO - bas ed breakfas t to 136 for PSO - bas ed breakfas t. After that, PCAs for each breakfas t, where times 0, 2 and 4 corres pond to the three class es in this study, were carried 470 Nuevas plataformas anal?ticas en metabol?mica ou t. The firs t conclu si on of this study was that nega ti ve ionization led to m ore pronounc ed clus teri ng and signi fi cantly better separati on among sam pli ng times , whi le pos i ti ve ioni zati on modes did not report a clear disc rimi nation between uri ne sam ples collec ted at different peri ods . Figure 3 sho ws the PCA m odel obtai ned in pos i ti ve mode for the tim e cou rs e study, whic h led to worse separati on as com pared to thos e obtained in negati ve mode (Fi g ure 3 ). Wi th these prem is es , only the data set obtai ned from negati ve ioni zati on mode was used furthe r on. On the oth er hand Figure 3 sho ws the scores graphs obtained from PCAs for eac h time - c ourse study in negati ve ionizati on. As can be seen, separati on between uri ne metaboli tes profili ng at tim es 0 and 4 was clear for every diet, whereas there was overlappi ng of sam ple s taken at 2 h with a di fferent behavio r. Figure 3. Scores plots obtained from PCAs for each time-course study in negative ionization mode. (A) DSO; (B) PSO; (C) NSO; (D) VOO. (A) (B) (C) (D) 471 Sent to Anal. Bioanal. Chem. Chapter 13 Thus , PCAs over sam ples corres pondi ng to DSO and PSO - based breakfas ts (Fi g. 2.A and 2.B) reveal ed an overlapping of sam ples taken after intake (t2 and t4) with a clear separation from sam ple s before intake (t0). Therefo re, the urinary metabolom e changes after single intake of a breakfas t prepared with fri ed PSO and DSO. In the cas e of the breakfas t prepared with non - enric hed sunflower oi l, only sam ples taken at t4 showed a profi le different f rom sam ples taken at t2 and before intake, whic h were grouped (Fi g. 2.C). Finally, the only breakfas t prom oti ng the separation of the three sam pli ng tim es was that prepared with fri ed VOO. In fact, it can be vi s ualized a gradual change from urine sam ples t aken before intake to thos e sam pled 4 h after the planned intake (Fi g. 2.D). Attendi ng to these res ul ts , differenc es am ong clas ses are likely to be obtai ned from samples 4 h after intake, as they are statis tic ally different from the control indi vi duals and , therefore, it is the optim um sam pli ng tim e to evaluate the inc idence of breakfas ts prepared with the diffe rent fried oils. Additionally, PSO and DSO - bas ed breakfas ts sho wed clearer separati ons from intake as com pared to NSO and VOO. Thi s separati on is ac c eptable for urine sam ples , as the effec t of diet is dilu ted in uri ne sec retion and non - supervis ed PCA expec tedly gives a more realis tic view of biolo gi c al sam ples. The applic ati on of ANOV A fi lteri ng led to better separations by rem oval of molec ul ar featur es that coul d produc e a com mon pattern of non - inform ati ve signals . Thus, one - way ANOV A with a Bonferroni - Holm correc tion was tes ted in thi s study. A probabili ty level of p < 0.05 was cons i dered statis tic ally signi fi cant. The PCA scores graph obtained after this fi lteri ng led to conclude that separation of the three clas s es (tim e) was com plete for the fou r diets after applyi ng the ANOV A fi lter, but this reduc ed the num ber of molec ular features to les s than 10%. Thi s would hi nder a final identific ation, so th i s filter was not furthe r applied, consi deri ng that freque nc y and fold change filters were enough to obtai n good separation. 472 Nuevas plataformas anal?ticas en metabol?mica 3 .3. Comparison of u rin ary met abolome profiles after breakfast s in t ake Once independent disc rim i nati on of indi vi duals after intake of the different breakfas ts was attained, the next step was to establis h statis tic al differences associ ated to breakfas t cons um ption. For this study, t4 sam ples were used to create the raw data matri x, whic h was cons ti tuted by 12800 molec ular enti ti es . In a firs t step, this matri x was treated to elim i nate random vari abili ty, by applyi ng frequency and fold change filters . Thus, after fi lteri ng by frequenc y, only MFs pres ent in at leas t 50% of samples pertaini ng to one group were kept. This led to a re duc ti on to 510 MFs provi ng the influence of redundant informati on and contri buti on to other internal and external fac tors . By the sec ond fi lter, only MFs that diffe r 2 - fold am ong class es were kept, leadi ng to a total of 223 MFs . Figure 3 repres ents PCA sco res plot carri ed out after filteri ng. As can be seen, the scores plot did not show clear clus ters of sam ples acc ordi ng to the breakfas t consum ed fi ndi ng a high dis pers ion. In a final study, PLS - DA was used to create a clas si fic ati on and predic ti on model. The scores plot obtai ned with the latent vari ables after application of this tool is shown in Figure 4. The robus tnes s of the model can be evaluated through the predic ti ve accurac y, defined as the percentage of correc tly class ified experi ments in a given c lass . Thi s acc urac y is diffe rent for trai ni ng and vali dation. Thus , the traini ng set produces a confusi on matri x, whi ch is a matri x with the true class in rows and the predi c ted class in colum ns . The diagonal represents correc tly clas si fi ed sam ples . Table 1 sho ws the acc urac y for eac h clas s and the overall (mean) acc urac y. These ranged from 58 to 100% for the diffe rent breakfas ts in the trai ni ng step and the overall acc urac y was 85%. 473 Sent to Anal. Bioanal. Chem. Chapter 13 T he next step was to tes t the model for vali dati on. The same set of sam pl es was used for cros s - validation of the trained model. Although redundant, this is consi dered a vali d stati s tic al proc edure when the num ber of avai lable sam ples is lim i ted. The predic ti on matri x obtai ned from vali dati on shows an overall acc urac y of 65%. Bo th the traini ng and vali dati on steps poi nted out that indi vi duals after DSO and PSO - bas ed meals intake were not properly predic ted accordi ng to the urine metabolomi cs analys is . Table 1. Characterization of the PLS model obtained with the analysis of the four classes for cross-validation (a) and training (b). (a) (b) [NSO] (Pred ic ted) [PSO] (Pred ic ted) [DSO] (Pred ic ted) [VOO ] (Pred ic ted) Acc ura c y [NSO] (T ru e) 10 1 0 1 83.3333 4 [PSO] (T rue) 2 7 5 3 41.1764 7 [DSO ] (Tru e) 2 3 9 0 64.2857 1 [VOO ] (T rue) 1 1 0 10 83.3333 4 Ove ra l l a cc ura c y 65.4545 4 After explori ng these matric es , it can be observed that the lowes t ac curac y coeffi ci ents were obtained for DSO and PSO prepared breakfas ts , with 41 and 64%, res pec ti vely, in the vali dati on step esti m ated by cros s - vali dati on. As the mode l was unlikely to predic t the two previ ous groups of [NSO] (Pred ic ted) [PSO] (Pred ic ted) [DSO] (Pred ic ted) [VOO ] (Pred ic ted) Acc ura c y [NSO] (T ru e) 12 0 0 0 100.000 [PSO] (T rue) 1 10 4 2 58.824 [DSO ] (Tru e) 0 1 13 0 92.857 [VOO ] (T rue) 0 0 0 12 100.000 Ove ra l l acc ura c y 85.455 474 Nuevas plataformas anal?ticas en metabol?mica i ndi viduals , the study was sim pli fi ed by explori ng the differences of VOO and NSO - prepared breakfas ts independently with PSO and DSO - bas ed meals by analys i ng diffe renc es in groups of three breakfas ts . Da ta sets were consti tuted by 119 and 261 molecul ar features for PLS studi es deali ng with PSO and DSO - breakfas ts. Therefore, a highe r num ber of features was requi red to disc rimi nate DSO - prepared breakfasts than to disc rim i nate between PSO - bas ed breakfas ts . T his strategy cons i derably improved the clas si fic ati on res ul ts , as sho ws Fig. 5 repres enti ng sam ples with the three fi rs t latent vari ables . Figure 5.A shows the scores plot that inc lu des the vari abili ty for indi vi duals after intake of VOO, NSO and PSO - prepa red breakfas ts . As can be seen, indi vi dua ls after intake of VOO and NSO - bas ed breakfas ts are clearly separated. In between, it can be clearly vis uali zed the group of invi di dua ls after intake of PSO prepared breakfas t, whi ch seems to be logic al. Additi onall y the model was able to predic t with good accurac y levels. Thus , the trai ni ng model achieved acc uracy levels of 100% for the three groups of indi vi duals , whi le cros s - vali dation reported acc urac y levels from 69% (for indi vi dua ls after intake of NSO - based br eakfas t) to 92% (for indi viduals after intake of the breakfas t prepared with fried VOO). In sum mary, thi s PLS - DA model was exc ellent to predic t thos e indi viduals after intake of VOO - bas ed breakfas t but, at the sam e time, had a good predic tion capabili ty to dis c rim i nate thos e indi viduals after intake of NSO and PSO breakfas ts . Quali tati vely, indi viduals after intake of PSO - prepared meal provi ded a uri nary metabolo me close to indi vi dua ls who inges ted VOO - based breakfas t. Com plem entari ly, Fig. 5.B sho ws the scores plot that represents the vari abili ty for indi vi duals after intake of VOO, NSO and DSO - prepared breakfas ts . Sim i larly to the previ ous case, sam ples corres ponding to VOO and NSO - based breakfas ts are clearly separated. The differenc e here is that the group of sam ples ass oc i ated to DSO are not loc ated between the two non - enric hed oils . Therefore, it can be consi dered as a com pletely different clus ter or class in this model. This sec ond model had agai n a good predic ti ve capabili ty with good accurac y leve ls , as can be seen in Table 3. Thus, the 475 Sent to Anal. Bioanal. Chem. Chapter 13 trai ni ng model simi larly ac hi eved acc urac y levels of 100% for the three group of indi vi duals, while cros s - vali dation reported acc urac y levels from 50% for indi viduals after intake of DSO - bas ed breakfas t to 92% for i ndi viduals after intake of the breakfas t prepared with fri ed VOO. In this case, thi s PLS - DA model was exc ellent to predict thos e indi viduals after intake of VOO - bas ed breakfas t and had a good predic ti on capabili ty to disc rimi nate NSO. However, cros s - vali da tion was not good enough to differenti ate indi viduals after intake of sunflower oi l with the synthetic oxi dation inhi bi tor. In fact, 50% of indi vi duals were mis c lass ifi ed sinc e some of them were ass i gned to the intake of breakfas t prepared with fri ed sunfl ower oi l and others as indi vi duals after the intake of the muf fi n prepared with VOO. In quali tati ve terms , indi vi duals after intake of DSO - prepared meal provi ded a urinary metabolom e different from the two othe r groups of indi vi duals, but predic ti on of the m as an independent class was not goo d. Table 2. Characterization of the PLS model for the study involving VOO, NSO and PSO-prepared breakfasts; (a) validation and (b) training. (a) (b) [NSO] (Pred ic ted) [PSO] (Pr edic t ed) [VOO ] (Pred ic ted) Acc ura c y [NSO] (T ru e) 9 2 2 69.231 [PSO] (T rue) 3 12 1 75 [VOO ] (T rue) 1 0 11 91.667 Ove ra l l acc ura c y 78.049 [NSO] (Pred ic ted) [PSO] (Pr edic t ed) [VOO ] (Pred ic ted) Acc ura c y [NSO] (T ru e) 13 0 0 100 [PSO] (T rue) 0 16 0 100 [VOO ] (T rue) 0 0 12 100 Ove ra l l acc ura c y 10 0 476 Nuevas plataformas anal?ticas en metabol?mica 3. 4. Ident ificati on of sign ifican t molecular feat ures Table 3. Characterization of the PLS model for the study involving VOO, NSO and DSO-prepared breakfasts; (a) validation and (b) training. (a) (b) [NSO] (Pred ic ted) [DSO] (Pred ic ted) [VOO ] (Pred ic ted) Acc ura c y [NSO] (T ru e) 15 0 0 100 [DSO] (Tru e) 0 14 0 100 [VOO ] (T ru e) 0 0 12 100 Ove ra l l acc ura c y 100 A final identific ati on step was carri ed out, for whic h the features obtained after the PLS analys is were queri ed agai ns t the METLIN databas e, whic h inc lu des endogenou s metaboli tes found in plasm a and, in a less er exte nt, exogenous hum an drugs and thei r metaboli tes . Finally, furthe r identific ation of the res ulti ng features was done by searc hi ng in the HMDB by com pari ng the experim ental bas e peak mass agai ns t the [M ?H] ? peaks as well as other adduc ts suc h as thos e form ed with formi c aci d or by dehydrati on. The database searc h res ul ts are summ ari zed in Tables 3 and 4, whic h inc lu de the molecul ar form ula, the detec ted adduc t mas s, matc hi ng tolerance, and the num eri c refe rence in HMDB. A total of 50 com pounds [NSO] (Pred ic ted) [DSO] (Pred ic ted) [VOO ] (Pre d ic ted) Acc ura c y [NSO] (T ru e) 11 1 3 73.333 [DSO] (Tru e) 5 7 2 50.000 [VOO ] (T rue) 0 1 11 91.667 Ove ra l l acc ura c y 68.085 477 Sent to Anal. Bioanal. Chem. Chapter 13 were identi fi ed from the enti ti es lis t created after statis tic al analysis in the PLS - DA study involving VOO, NSO and PSO - prepared breakfas ts with a mass tolerance window below 10 ppm (Table 4). Sim i larly, a total of 114 com pounds were identi fied in the cas e of the study inc lu di ng VOO, NSO and DSO - prepared breakfas ts (Table 5). The 42 and 44% of metaboli tes identifi ed in both PLS enti ti es lis ts were in comm on reveali ng the sim i lari ty of both studies. These com pou nds are expected to be related with differences in metabolis m produced by oils intake. Among the identi fi ed metaboli tes , sugars, horm ones , lipi ds , ami no ac i ds , and other minor com ponents were identifi ed. The occ urrence of thes e com pounds in uri ne as a consequenc e of oils intake is furthe r dis c uss ed by com pari s on of fold change (FC) rati os of indi vi dua ls after intake of diffe rent breakfas ts (equi vale nt to concentration changes ). Lipids Lipids are the mai n cons ti tuents of oils . Thus , they are likely to appear in uri ne after oi ls intake. Among lipids , several mono - and diac ylglyc eri des were identi fied, des pite thei r non - polar charac ter. The intake of NSO and VOO breakfas ts led to sim ilar uri ne levels of diac ylglyc erols ( DG ) linked to C20 or longer fatty ac i ds , whi le indi vi dua ls after DSO and PSO breakfas ts reported FC rati os of ? 2 and 2.9, res pec ti vely. In the case of monoglyc erols (MG), the intake of fri ed sunflower oi ls inc reas ed the conc entrati on of detec ted MG as com pared to the intake of VOO breakfas t. Thi s inc reas e was maxim um for indi vi duals after intake of NS O breakfas t, whi le it was minim um for PSO indi viduals . It has been demons trated that a high percent of mono and diacylglyc erols in oi ls aids to prevent and control obes i ty [30]. Concerni ng phos pho li pids , different metaboli tes suc h as phos phati di l glycerols (PG), phos phati di l glycerol phos pha te deri vati ves (PGP) or phos phati dyli nositol phos pha te deri vati ves PIP suffe red conce ntrati on changes com pari ng NSO, DSO and VOO 478 Nuevas plataformas anal?ticas en metabol?mica breakfas ts whi le only PIP metaboli tes contri buted to explai n the vari abi li ty observed in t he PLS - DA of NSO, PS O and VOO breakfas ts. Additionally, the acylglycines isobuti rylglu yci ne and hexanoylglyc i ne were als o identi fi ed. Thes e are minor metaboli tes of fatty aci ds conjugated with the ami no aci d glyci ne that are norm ally excreted in uri ne [3 1]. As a consequenc e of the breakfas ts intake, uri ne from indi viduals who inges ted NSO and DSO prepared breakfas ts pres ented higher concentrations of both ac ylglyc i nes than indi vidual s after intake of VOO and PSO breakfas ts. Therefo re, thes e metabolites al lowed disc ri mi nation as a res ul t of the levels of phenolic anti oxidants . Hydroxylated fatty aci ds such as hydroxybutyric ac i d or hydroxys ebac ic aci d and dic arboxyli c aci ds suc h as hexenedioic acid or hexadecadienoic aci d were also signific ant metaboli tes t o explai n the vari abili ty in both PLS studi es . These compounds were at lower concentration in uri ne from NSO and DSO indi vi duals , whi le thei r concentration was highe r in the uri ne from PSO indivi duals . Other minor lipids suc h as oxi dati on produc ts of eic os anoi ds and prostanoi ds contri buted to explai n the vari abi li ty in the study com pari ng DSO, NSO and VOO indi vi duals. As sho uld be expec ted, thes e oxi di zed metaboli tes involved in the inflamm ati on casc ade were found with FC rati os between 4 and 6 by compari so n of indi vi dua ls after intake of DSO and NSO breakfas t versus VOO breakfas t. Thi s is a really interes ti ng resul t since these metaboli tes were not crucial to di sc rimi nate PS O indi vi dua ls. Amino acids The pres enc e of ami no aci ds and deri vati ves among the of metaboli tes detec ted in uri ne with contri buti on to diffe renti ate among indi viduals after intake of breakfas ts is suppos edly due to protein breakdown. Therefore, dipeptides, L - gamm a - glu tam yl - L - is oleuci ne and L - beta - as partyl - L - leuci ne and the free ami noac i d N(6) - methyllysi ne contri buted to differenti ate indi vi duals after intake of PSO - bas ed breakfas t. 479 Sent to Anal. Bioanal. Chem. Chapter 13 Oxi di zed forms of ami no ac i ds suc h as tryptopha n, tyros i ne or leuc i ne were signi fic ant to dis ti ngui sh indi vi duals after intake of DSO and NSO breakfas ts. How ever, the mos t abundant metaboli tes related to ami no aci d metaboli sm are cataboli c produc ts, whic h are norm ally fou nd in uri ne. Among them , hydroqui none, serotoni n, ac etam i dobutanoic aci d, indol deri vati ves indoxyl sulpha te and 3 - i ndolebutyri c acid and val eric ac i d deri vati ves alpha - ketois ovaleric aci d and 2 - oxovaleric were identi fied. Other compounds There are other minor consti tuents that were fou nd to have a signi fic ant relationshi p with oi ls intake; for ins tance , nucleo ti de catabolic produc ts and su gars were als o detec ted. Thus, methylgua ni nes are naturally occ urri ng modi fied puri nes deri ved from tRN A. Thes e are naturally fou nd in biofl ui ds and abnormally elevated in serum and urine of cancer patients and smokers [32]. 5 - Hydroxym ethylurac i l is an oxi dation produc t deri ved from thymi ne or 5 - methylc ytosi ne [33]. Among sugars , pentosi des (xylos e, ribose or arabinos e) were also detec ted. Othe r norm ally occ urri ng uri nary metaboli tes suc h as 2 - methylglu tari c aci d, siali c ac i d and riboflavin, were identifi ed . 2 - Methylglu tari c aci d is an organi c acid us ua lly found in hum an uri ne. 2 - Methylglu taric aci d is a metaboli te of suc cini c acid, a citri c aci d cyc le intermediate [34]. N - ac etylneurami nic aci d (Neu Ac) or sialic aci d is an acetyl derivati ve of the ami no suga r neurami nic aci d, whic h occ urs in many glyc oprotei ns , glyc oli pids , and polys ac cha rides in human tiss ues flui ds, inc ludi ng serum, cerebros pinal fluid, sali va, uri ne, am niotic flui d, and breas t milk [35]. Ribofla vi n or vitami n B2 is an eas i ly absorbed, wate r - solu ble mic ronutri ent involved in energy produc ti on by aidi ng in the metabolis m of fats , carboh ydrates , and proteins [35]. Phenolic com pounds and deri vati ves suc h as tyros ol, hom ovani lli c acid, trim ethoxyc i nnami c acid, phenylacetaldehyde or vanill ylam i ne , among others , were detec ted in the uri ne from indi vi duals after intake of breakfas ts prepared with fri ed VOO and sunflower oil enric hed with natural phe nols . HMDB ID Common Name Che mica l Formula Neutra l Mass MW Add uct MW Cha rge Add uct, MW Dif fe re nce (D a ) MW Dif fe re nce (pp m) 1 HMDB0 2 43 4 Hyd roquin on e C6 H6 O2 11 0 .03 68 10 9 .02 95 1 - M - H 0.0 0 03 28 3.0 0 2 HMDB0 0 62 0 Gluta con ic acid C5 H6 O4 13 0 .02 66 11 1 .00 82 1 - M - H2 0 - H 0.0 0 07 32 6.5 9 HMDB0 2 09 2 Ita con ic acid C5 H6 O4 13 0 .02 66 11 1 .00 82 1 - M - H2 0 - H 0.0 0 07 32 6.5 9 3 HMDB0 0 80 8 N - Butyry lglyci ne C6 H1 1 NO3 14 5 .07 39 14 4 .06 66 1 - M - H 0.0 0 02 59 1.7 9 HMDB0 3 68 1 4 - A ce ta midobuta no ic acid C6 H1 1 NO3 14 5 .07 39 14 4 .06 66 1 - M - H 0.0 0 02 59 1.7 9 4 HMDB0 0 00 8 2 - Hyd roxybutyri c acid C4 H8 O3 10 4 .04 73 14 9 .04 55 1 - M+FA - H 0.0 0 05 95 3.9 9 5 HMDB 0 0 09 8 D - Xy lose C5 H1 0 O5 15 0 .05 28 14 9 .04 55 1 - M - H 0.0 0 05 8 3.8 9 HMDB0 0 64 6 L - A ra bin ose C5 H1 0 O5 15 0 .05 28 14 9 .04 55 1 - M - H 0.0 0 05 8 3.8 9 6 HMDB0 0 89 3 Sube ric acid C8 H1 4 O4 17 4 .08 92 17 3 .08 19 1 - M - H 0.0 0 07 48 4.3 2 7 HMDB0 0 11 8 Homov a ni llic acid C9 H1 0 O4 18 2 .05 79 1 8 1 .05 06 1 - M - H 0.0 0 06 41 3.5 4 HMDB0 2 64 3 3 - ( 3 - hyd roxyp he ny l) - 3 - hyd roxyp rop ano ic acid C9 H1 0 O4 18 2 .05 79 18 1 .05 06 1 - M - H 0.0 0 06 41 3.5 4 8 HMDB0 2 14 2 Phosp hori c acid H3 O4 P 97 .9 76 89 8 19 4 .94 65 1 - 2M - H 0.0 0 08 39 4.3 0 9 HMDB0 0 25 9 Seroton in C1 0 H1 2 N2 O 17 6 .09 50 19 7 .06 96 1 - M+Na - 2 H 0.0 0 09 4.5 6 HMDB0 2 03 8 N(6 ) - Me thyllysin e C7 H1 6 N2 O2 16 0 .12 12 19 7 .06 98 1 - M+K - 2 H 0.0 0 10 68 5.4 1 10 HMDB0 0 68 2 Ind oxyl sulf a te C8 H7 NO4 S 21 3 .00 96 21 2 .00 23 1 - M - H 0.0 0 12 82 6.0 5 1? 1 HMDB0 0 70 1 Hexan oy lglyci ne C8 H1 5 NO3 17 3 .10 52 21 8 .10 34 1 - M+F A - H 0.0 0 04 12 1.8 9 12 HMDB0 1 86 5 2 - Oxova le ric acid C5 H8 O3 11 6 .04 73 23 1 .08 74 1 - 2M - H 0.0 0 13 89 6.0 1 13 HMDB1 1 17 0 L - ga mma - gluta my l - L - isole ucin e C1 1 H2 0 N2 O5 26 0 .13 72 24 1 .11 88 1 - M - H2 0 - H 0.0 0 14 34 5.9 5 Table 4. Identification of molecular features related with oils intake after statistical analysis in the PLS-DA study involving VOO, NSO and PSO-prepared breakfasts. HMDB1 1 17 1 L - ga mma - gluta my l - L - le ucin e C1 1 H2 0 N2 O5 26 0 .13 72 24 1 .11 88 1 - M - H2 0 - H 0.0 0 14 34 5.9 5 14 HMDB0 3 33 2 3 - Me thoxy - 4 - Hyd roxyp heny lglyco l sulf a te C9 H1 2 O7 S 26 4 .03 04 24 5 .01 20 1 - M - H2 0 - H 0.0 0 24 26 9.9 0 15 HMDB1 1 16 6 L - be ta - a sp a rtyl - L - le ucin e C1 0 H1 8 N2 O5 24 6 .12 16 28 3 .07 02 1 - M+K - 2 H 0.0 0 01 83 0.6 5 16 HMDB0 2 51 1 3,4 ,5 - Trime thoxyci nn a mic acid C1 2 H1 4 O5 23 8 .08 41 28 3 .08 23 1 - M+FA - H 0.0 0 09 16 3.2 4 17 HMDB0 0 39 3 3 - He xen ed io ic acid C6 H8 O4 14 4 .04 23 28 7 .07 72 1 - 2M - H 0.00 1 37 3 4.7 8 18 HMDB0 0 42 2 2 - Me thylgluta ric acid C6 H1 0 O4 14 6 .05 79 29 1 .10 86 1 - 2M - H 0.0 0 15 56 5.3 5 19 HMDB0 0 23 0 N - A ce tylne ura min ic acid C1 1 H1 9 NO9 30 9 .10 60 30 8 .09 87 1 - M - H 0.0 0 22 58 7.3 3 20 HMDB1 3 24 0 Ind ole a ce tyl gluta min e C1 5 H1 7 N3 O4 30 3 .12 19 34 0 .07 05 1 - M+K - 2 H 0.0 0 11 9 3.5 0 21 HMDB1 2 28 1 Portula ca xan thin II C1 8 H1 8 N2 O7 37 4 .11 14 35 5 .09 30 1 - M - H2 0 - H 0.0 0 15 56 4.3 8 2 2 HMDB0 0 24 4 Ribo f la vin C1 7 H2 0 N4 O6 37 6 .13 83 35 7 .11 99 1 - M - H2 0 - H 0.0 0 06 71 1.8 8 23 HMDB0 0 73 4 Ind ole a crylic acid C1 1 H9 NO2 18 7 .06 33 37 3 .11 94 1 - 2M - H 0.0 0 35 1 9.4 1 24 HMDB0 0 89 7 7 - Me thylgua ni ne C6 H7 N5 O 16 5 .06 51 37 5 .12 83 1 - 2M+ FA - H 0.0 0 29 3 7.8 1 HMDB0 1 56 6 3 - M e thylgua ni ne C6 H7 N5 O 16 5 .06 51 37 5 .12 83 1 - 2M+ FA - H 0.0 0 29 3 7.8 1 HMDB0 3 28 2 1 - Me thylgua ni ne C6 H7 N5 O 16 5 .06 51 37 5 .12 83 1 - 2M+ FA - H 0.0 0 29 3 7.8 1 25 HMDB0 2 30 2 3 - Ind ole p rop io ni c acid C1 1 H1 1 NO2 18 9 .07 90 37 7 .15 07 1 - 2M - H 0.0 0 36 93 9.7 9 26 HMDB1 1 55 4 MG(0 :0 / 22 :4 ( 7 Z,1 0 Z,13 Z,1 6 Z) /0 :0) C2 5 H4 2 O4 40 6 .30 83 38 7 .28 99 1 - M - H2 0 - H 0.0 0 14 34 3.7 0 HMDB1 1 58 4 MG(2 2:4 (7 Z,1 0 Z,1 3 Z,16 Z) / 0 :0 /0 :0) C2 5 H4 2 O4 40 6 .30 83 38 7 .28 99 1 - M - H2 0 - H 0.0 0 14 34 3.7 0 27 HMDB1 1 69 0 7 - A min ome thyl - 7 - ca rba gua nine C7 H9 N5 O 17 9 .08 07 40 3 .15 96 1 - 2M+ FA - H 0. 0 0 28 38 7.0 4 28 HMDB0 2 09 6 3 - Ind ole butyri c acid C1 2 H1 3 NO2 20 3 .09 46 40 5 .18 20 1 - 2M - H 0.0 0 22 28 5.5 0 29 HMDB0 1 45 1 Lip oi c acid C8 H1 4 O2 S2 20 6 .04 35 41 1 .07 98 1 - 2M - H 0.0 0 33 87 8.2 4 30 HMDB0 0 35 0 3 - Hyd roxyse ba cic acid C1 0 H1 8 O5 21 8 .11 54 43 5 .22 36 1 - 2M - H 0.0 0 15 87 3.6 5 31 HMDB0 6 07 0 Pre gna ne trio l C2 1 H3 6 O3 33 6 .26 60 44 9 .25 16 1 - M+TFA - H 0.0 0 43 33 9.6 4 32 HMDB1 0 32 1 3,1 7 - And rosta ned io l glucuron id e C2 5 H4 0 O8 46 8 .27 23 44 9 .25 39 1 - M - H2 0 - H 0.0 0 20 14 4.4 8 HMDB1 0 33 9 3 - a lp ha - a nd rosta ned io l glucuron id e C2 5 H4 0 O8 46 8 .27 23 44 9 .25 39 1 - M - H2 0 - H 0.0 0 20 14 4.4 8 HMDB1 0 35 9 17 - hyd roxya nd rosta ne - 3 - glucuron id e C2 5 H4 0 O8 46 8 .27 23 44 9 .25 39 1 - M - H2 0 - H 0.0 0 20 14 4.4 8 HMDB0 0 44 7 7a,1 2a - D ihyd roxy - 3 - oxo - 4 - chole no ic acid C2 4 H3 6 O5 40 4 .25 63 44 9 .25 45 1 - M+FA - H 0.0 0 14 34 3.1 9 33 HMDB1 1 14 4 LPA ( 18 :0e /0 :0) C2 1 H4 5 O6 P 42 4 .29 54 46 1 .24 40 1 - M+K - 2 H 0.0 0 25 63 5.5 6 34 HMDB1 1 53 9 MG(0 :0 / 18 :3( 6 Z,9 Z,1 2 Z) /0 :0 ) C2 1 H3 6 O4 35 2 .26 14 46 5 .24 69 1 - M+TFA - H 0.0 0 37 23 8.0 0 HMDB1 1 54 0 MG(0 :0 / 18 :3( 9 Z,1 2 Z,15 Z) / 0:0 ) C2 1 H3 6 O4 35 2 .26 14 46 5 .24 69 1 - M+TFA - H 0.0 0 37 23 8.0 0 HMDB1 1 56 9 MG(1 8:3 (6 Z,9 Z,1 2 Z) /0 :0 /0 :0 ) C2 1 H3 6 O4 35 2 .26 14 46 5 .24 69 1 - M+TFA - H 0.0 0 37 23 8.0 0 HMDB1 1 57 0 MG(1 8:3 (9 Z,1 2 Z,1 5 Z)/ 0 :0 / 0:0 ) C2 1 H3 6 O4 35 2 .26 14 46 5 .24 69 1 - M+TFA - H 0.0 0 37 23 8.0 0 35 HMDB0 2 82 9 And roste ron e glucuron id e C2 5 H3 8 O8 46 6 .25 67 46 5 .24 94 1 - M - H 0.0 0 12 82 2.7 6 36 HMDB0 0 45 1 cis - 4 - Hyd roxycy clohexyla ce tic acid C8 H1 4 O3 15 8 .09 43 47 3 .27 56 1 - 3M - H 0.0 0 36 62 7.7 4 37 HMDB0 3 18 0 Cortol C2 1 H3 6 O5 36 8 .25 63 48 1 .24 19 1 - M+TFA - H 0.0 0 41 81 8.6 9 38 HMDB1 0 35 1 11 - be ta - hyd roxya nd roste ron e - 3 - glucuron id e C2 5 H3 8 O9 48 2 .25 16 48 1 .24 43 1 - M - H 0.0 0 17 4 3.6 2 39 HMDB0 9 93 7 PIP ( 16 :0/ 22 :5( 4 Z,7 Z,1 0 Z,1 3 Z,16 Z) ) C4 7 H8 2 O1 6P 2 96 4 .50 78 48 1 .24 66 2 - M - 2 H 0.0 0 05 8 1.2 1 HMDB0 9 93 8 PIP ( 16 :0/ 22 :5( 7 Z,1 0 Z,13 Z,1 6 Z,1 9 Z) ) C4 7 H8 2 O1 6P 2 96 4 .50 78 48 1 .24 66 2 - M - 2 H 0.0 0 05 8 1.2 1 HMDB0 9 94 5 PIP ( 16 :2( 9 Z,12 Z ) / 22 :3( 10 Z,1 3Z,1 6 Z) ) C4 7 H8 2 O1 6P 2 96 4 .50 78 48 1 .24 66 2 - M - 2 H 0.0 0 05 8 1.2 1 HMDB0 9 96 7 PIP ( 18 :1( 11 Z) /2 0 :4 (5 Z,8 Z,1 1 Z,1 4 Z) ) C4 7 H8 2 O1 6P 2 96 4 .50 78 48 1 .24 66 2 - M - 2 H 0.0 0 05 8 1.2 1 HMDB0 9 96 8 PIP ( 18 :1( 11 Z) /2 0 :4 (8 Z,1 1 Z,1 4Z,1 7 Z) ) C4 7 H8 2 O1 6P 2 96 4 .50 78 48 1 .24 66 2 - M - 2 H 0.0 0 05 8 1.2 1 HMDB0 9 98 0 PIP ( 18 :1( 9 Z)/ 20 :4( 5 Z,8 Z,1 1 Z,14 Z) ) C4 7 H8 2 O1 6P 2 96 4 .50 78 48 1 .24 66 2 - M - 2 H 0.0 0 05 8 1.2 1 HMDB0 9 98 1 PIP ( 18 :1( 9 Z)/ 20 :4( 8 Z,1 1 Z,1 4 Z,1 7 Z) ) C4 7 H8 2 O1 6P 2 96 4 .50 78 48 1 .24 66 2 - M - 2 H 0.0 0 05 8 1.2 1 HMDB0 9 99 1 PIP ( 18 :2( 9 Z,12 Z) / 20 :3( 5 Z,8 Z,11 Z ) ) C4 7 H8 2 O1 6P 2 96 4 .50 78 48 1 .24 66 2 - M - 2 H 0.0 0 05 8 1.2 1 HMDB0 9 99 2 PIP ( 18 :2( 9 Z,12 Z) / 20 :3( 8 Z,1 1 Z,1 4 Z) ) C4 7 H8 2 O1 6P 2 96 4 .50 78 48 1 .24 66 2 - M - 2 H 0.0 0 05 8 1.2 1 HMDB1 0 00 9 PIP ( 20 :3( 5 Z,8 Z,1 1 Z)/ 18 :2( 9 Z,12 Z) ) C4 7 H8 2 O1 6P 2 96 4 .50 78 48 1 .24 66 2 - M - 2 H 0.0 0 05 8 1.2 1 HM DB1 0 01 4 PIP ( 20 :3( 8 Z,11 Z,1 4 Z) /1 8 :2 ( 9 Z,1 2 Z) ) C4 7 H8 2 O1 6P 2 96 4 .50 78 48 1 .24 66 2 - M - 2 H 0.0 0 05 8 1.2 1 HMDB1 0 01 7 PIP ( 20 :4( 5 Z,8 Z,1 1 Z,14 Z) / 18 :1(1 1 Z) ) C4 7 H8 2 O1 6P 2 96 4 .50 78 48 1 .24 66 2 - M - 2 H 0.0 0 05 8 1.2 1 HMDB1 0 01 8 PIP ( 20 :4( 5 Z,8 Z,1 1 Z,14 Z) / 18 :1(9 Z) ) C4 7 H8 2 O1 6P 2 96 4 . 50 78 48 1 .24 66 2 - M - 2 H 0.0 0 05 8 1.2 1 HMDB1 0 02 1 PIP ( 20 :4( 8 Z,11 Z,1 4 Z,1 7 Z) /1 8 :1( 1 1 Z)) C4 7 H8 2 O1 6P 2 96 4 .50 78 48 1 .24 66 2 - M - 2 H 0.0 0 05 8 1.2 1 HMDB1 0 02 2 PIP ( 20 :4( 8 Z,11 Z,1 4 Z,1 7 Z) /1 8 :1( 9 Z) ) C4 7 H8 2 O1 6P 2 96 4 .50 78 48 1 .24 66 2 - M - 2 H 0.0 0 05 8 1.2 1 HMDB1 0 02 4 PIP ( 22 :3( 10 Z,1 3 Z,1 6 Z)/ 1 6:2 (9Z,1 2 Z) ) C4 7 H8 2 O1 6P 2 96 4 .50 78 48 1 .24 66 2 - M - 2 H 0.0 0 05 8 1.2 1 HMDB1 0 02 9 PIP ( 22 :5( 4 Z,7 Z,1 0 Z,13 Z,1 6 Z) /16 :0) C4 7 H8 2 O1 6P 2 96 4 .50 78 48 1 .24 66 2 - M - 2 H 0.0 0 05 8 1.2 1 HMDB1 0 03 0 PIP ( 22 :5( 7 Z,10 Z,1 3 Z,1 6 Z,1 9 Z)/1 6 :0 ) C4 7 H8 2 O1 6P 2 96 4 .50 78 48 1 .24 66 2 - M - 2 H 0.0 0 05 8 1.2 1 40 HMDB0 2 00 4 5 - Me thoxyd ime thyltryp ta min e C1 3 H1 8 N2 O 21 8 .14 19 48 1 .28 20 1 - 2M+ FA - H 0.0 0 01 53 0.3 2 41 HMDB0 0 08 9 Cytid in e C9 H1 3 N3 O5 24 3 .08 55 48 5 .16 38 1 - 2M - H 0.0 0 40 89 8.4 3 42 HMDB1 2 16 7 4alp ha - f ormyl - 4 be ta - me thyl - 5 alp ha - chole sta - 8 ,2 4 - d ie n - 3 be ta - o l C2 9 H4 6 O2 42 6 .34 98 50 5 .26 87 1 - M+Br 0.0 0 04 88 0.9 7 43 HMDB0 0 03 0 Biotin C1 0 H1 6 N2 O3 S 24 4 .08 82 53 3 .17 46 1 - 2M+ FA - H 0.0 0 12 82 2.4 0 44 HMDB1 3 04 3 Prosta gla nd in PGE2 1 - glyce ryl este r C2 3 H3 8 O7 42 6 .26 17 53 9 .24 73 1 - M+TFA - H 0.0 0 44 56 8.2 6 45 HMDB1 0 32 0 Co rtolone - 3 - glucuron id e C2 7 H4 2 O1 1 54 2 .27 27 54 1 .26 54 1 - M - H 0.0 0 15 26 2.8 2 46 HMDB0 3 52 9 Ino sitol 1,3 ,4 ,5 ,6 - p entakisphosp ha te C6 H1 7 O21 P5 57 9 .89 50 61 6 .84 36 1 - M+K - 2 H 0.0 0 18 31 2.9 7 47 HMDB0 0 29 6 Urid ine C9 H1 2 N2 O6 24 4 .06 95 73 1 .20 14 1 - 3M - H 0.0 0 34 18 4.6 7 48 HM D B0 7 41 1 DG( 20 :1 ( 11 Z) /2 2 :6 (4 Z,7 Z,1 0 Z,1 3 Z,1 6 Z, 1 9 Z) /0 :0) C4 5 H7 4 O5 69 4 .55 37 80 7 .53 92 1 - M+TFA - H 0.0 0 03 05 0.3 8 HMDB0 7 43 8 DG( 20 :2 ( 11 Z,1 4 Z) /2 2:5 (4 Z,7 Z,1 0 Z,1 3 Z, 1 6 Z) /0 :0) C4 5 H7 4 O5 69 4 .55 37 80 7 .53 92 1 - M+TFA - H 0.0 0 03 05 0.3 8 HMDB0 7 46 6 DG( 20 :3 ( 5 Z,8 Z,1 1 Z)/ 22 :4( 7 Z ,10 Z,1 3 Z,1 6 Z) / 0 :0 ) C4 5 H7 4 O5 69 4 .55 37 80 7 .53 92 1 - M+TFA - H 0.0 0 03 05 0.3 8 HMDB0 7 49 5 DG( 20 :3 ( 8 Z,11 Z,1 4 Z) /2 2 :4 (7 Z,1 0 Z,1 3 Z, 1 6 Z) /0 :0) C4 5 H7 4 O5 69 4 .55 37 80 7 .53 92 1 - M+TFA - H 0.0 0 03 05 0.3 8 HMDB0 7 58 1 DG( 20 :5 ( 5 Z,8 Z,1 1 Z,14 Z,1 7 Z) /22 :2( 1 3 Z, 1 6 Z) /0 :0) C4 5 H7 4 O5 69 4 .55 37 80 7 .53 92 1 - M+TFA - H 0.0 0 03 05 0.3 8 HMDB0 7 66 5 DG( 22 :2 ( 13 Z,1 6 Z) /2 0:5 (5 Z,8 Z,1 1 Z,1 4 Z, 1 7 Z) /0 :0) C4 5 H7 4 O5 69 4 .55 37 80 7 .53 92 1 - M+TFA - H 0.0 0 03 05 0.3 8 Table 5. Identification of molecular features related with oils intake after statistical analysis in the PLS-DA study involving VOO, NSO and DSO-prepared breakfasts HMDB ID Common Name Che mica l Formula Neutra l Mass MW Add uct MW Cha rge Add uct, MW Dif fe re nce ( D a ) MW Dif fe re nc e (pp m) 1 HMDB0 1 16 9 4 - A min op he nol C6 H7 NO 10 9 .05 3 10 8 .04 54 9 - 1 M - H 0.0 0 02 21 2.0 5 2 HMDB0 2 43 4 Hyd roquin on e C6 H6 O2 11 0 .03 7 10 9 .02 95 0 - 1 M - H 0.0 0 02 06 1.8 9 3 HMDB0 0 62 0 Gluta con ic acid C5 H6 O4 13 0 .02 7 11 1 .00 82 - 1 M - H2 0 - H 0.0 0 08 32 7.4 9 HMDB0 2 09 2 Ita con ic acid C5 H6 O4 13 0 .02 7 11 1 .00 82 2 - 1 M - H2 0 - H 0.0 0 08 32 7.4 9 4 HMDB0 0 56 2 Cre a tin in e C4 H7 N3 O 11 3 .05 9 11 2 .05 16 4 - 1 M - H 0.0 0 04 12 3.6 8 HMDB0 7 69 0 DG( 22 :4 ( 7 Z,10 Z,1 3 Z,1 6 Z) /2 0 :3( 5 Z,8 Z,1 1 Z) / 0 :0 ) C4 5 H7 4 O5 69 4 .55 37 80 7 .53 92 1 - M+TFA - H 0.0 0 03 05 0. 3 8 HMDB0 7 69 1 DG( 22 :4 ( 7 Z,10 Z,1 3 Z,1 6 Z) /2 0 :3( 8 Z,1 1 Z, 1 4 Z) /0 :0) C4 5 H7 4 O5 69 4 .55 37 80 7 .53 92 1 - M+TFA - H 0.0 0 03 05 0.3 8 HMDB0 7 71 8 DG( 22 :5 ( 4 Z,7 Z,1 0 Z,13 Z,1 6 Z) /20 :2( 1 1 Z, 1 4 Z) /0 :0) C4 5 H7 4 O5 69 4 .55 37 80 7 .53 92 1 - M+TFA - H 0.0 0 03 05 0.3 8 HMDB0 7 74 7 DG( 22 :5 ( 7 Z,10 Z,1 3 Z, 1 6 Z,1 9 Z)/2 0 :2 ( 11 Z ,1 4 Z) / 0:0 ) C4 5 H7 4 O5 69 4 .55 37 80 7 .53 92 1 - M+TFA - H 0.0 0 03 05 0.3 8 HMDB0 7 77 5 DG( 22 :6 ( 4 Z,7 Z,1 0 Z,13 Z,1 6 Z,1 9Z) / 2 0 :1 ( 1 1 Z) /0 :0) C4 5 H7 4 O5 69 4 .55 37 80 7 .53 92 1 - M+TFA - H 0.0 0 03 05 0.3 8 49 HMDB0 2 72 8 Thyro xine sulf a te C1 5 H1 1 I4 NO7 S 85 6 .64 35 89 3 .59 21 1 - M+K - 2 H 0.0 0 31 13 3.4 8 50 HMDB1 2 56 2 13 ' - Hyd roxy - ga ma - toco trie no l C2 8 H4 2 O3 42 6 .31 34 89 7 .62 50 1 - 2M+ FA - H 0.0 0 31 74 3.5 4 5 HMDB0 0 53 5 Cap roi c acid C6 H1 2 O2 11 6 .08 4 11 5 .07 64 5 - 1 M - H 0.0 0 03 66 3.1 8 6 HMDB0 0 25 4 Succini c acid C4 H6 O4 11 8 .02 7 11 7 .01 93 3 - 1 M - H 0.0 0 03 28 2.8 0 7 HMDB0 4 28 4 Tyro sol C8 H1 0 O2 13 8 .06 8 11 9 .04 97 0 - 1 M - H2 0 - H 0.0 0 06 87 5.7 7 HMDB0 6 23 6 Phe ny la ce ta lde hy d e C8 H8 O 12 0 .05 8 11 9 .05 02 4 - 1 M - H 0.0 0 01 45 1.2 2 8 HMDB1 1 72 4 4 - Hyd roxybe nzy l alcohol C7 H8 O2 12 4 .05 2 12 3 .04 51 5 - 1 M - H 0.0 0 00 46 0.3 7 9 HMDB0 3 90 3 2 - Hyd roxye tha ne sulf on a te C2 H6 O4 S 12 5 .99 9 12 4 .99 14 0 - 1 M - H 0.0 0 04 88 3.9 0 10 HMDB0 0 39 3 3 - He xen ed io ic acid C6 H8 O4 14 4 .04 2 12 5 .02 38 7 - 1 M - H2 0 - H 0.0 0 05 11 4.0 9 11 HMDB0 0 48 2 Cap rylic acid C8 H1 6 O2 14 4 .11 5 12 5 .09 66 5 - 1 M - H2 0 - H 0.0 0 07 17 5.7 3 HMDB0 0 48 0 7,1 0 - He xad e cad ie noi c acid C1 6 H2 8 O2 25 2 .20 9 12 5 .09 71 8 - 2 M - 2 H 0.0 0 01 83 1.4 6 12 HMDB1 1 65 5 2 - ( 3 - Ca rboxy - 3 - a min op rop yl) - L - his tid in e C1 0 H1 6 N4 O4 25 6 .11 7 12 7 .05 13 0 - 2 M - 2 H 0.0 0 01 6 1.2 6 13 HMDB0 0 14 8 L - Gluta mic acid C5 H9 NO4 14 7 .05 3 12 8 .03 47 8 - 1 M - H2 0 - H 0.0 0 07 48 5.8 4 14 HMDB0 4 81 3 3 - Me thylurid in e C1 0 H1 4 N2 O6 25 8 .08 5 12 8 .03 53 1 - 2 M - 2 H 0.0 0 02 14 1.6 7 15 HMDB1 2 30 9 Vani llyla min e C8 H1 1 NO2 15 3 .07 9 13 4 .06 05 9 - 1 M - H2 0 - H 0.0 0 09 6.7 1 16 HMDB0 2 46 6 Hyd roxybe nzoi c acid C7 H6 O3 13 8 .03 2 13 7 .02 44 1 - 1 M - H 0.0 0 04 58 3.3 4 17 HMDB0 2 73 0 Nicotina mid e N - oxide C6 H6 N2 O2 13 8 .04 3 13 7 .03 56 5 - 1 M - H 0.0 0 00 61 0.4 5 18 HMDB0 1 97 8 5 - Hyd roxyp yra zina mide C5 H5 N3 O2 13 9 .03 8 13 8 .03 09 0 - 1 M - H 0.0 0 00 46 0.3 3 19 HMDB1 2 15 3 3,4 - D ihyd roxybe nzy la min e C7 H9 NO2 13 9 .06 3 13 8 .05 60 5 - 1 M - H 0.0 0 02 75 1.9 9 20 HMDB0 0 73 0 Isobutyry lg lyci ne C6 H1 1 NO3 14 5 .07 4 14 4 .06 66 2 - 1 M - H 0.0 0 00 15 0.1 0 HMDB0 3 68 1 4 - A ce ta midobuta no ic acid C6 H1 1 NO3 14 5 .07 4 14 4 . 06 66 2 - 1 M - H 0.0 0 00 15 0.1 0 HMDB1 2 15 1 2 - Ke to - 6 - a min oca p roa te C6 H1 1 NO3 14 5 .07 4 14 4 .06 66 2 - 1 M - H 0.0 0 00 15 0.1 0 21 HMDB0 0 64 1 L - Gluta min e C5 H1 0 N2 O3 14 6 .06 9 14 5 .06 18 6 - 1 M - H 0.0 0 00 15 0.1 0 22 HMDB1 2 81 5 5 - A min ope ntan a l C5 H1 1 NO 10 1 .08 4 14 6 .08 22 6 - 1 M+FA - H 0.0 0 0 0 31 0.2 1 23 HMDB0 1 54 5 Pyri d oxa l C8 H9 NO3 16 7 .05 8 14 8 .03 98 6 - 1 M - H2 0 - H 0.0 0 06 1 4.1 2 24 HMDB0 0 00 8 2 - Hyd roxybutyri c acid C4 H8 O3 10 4 .04 7 14 9 .04 55 3 - 1 M+FA - H 0.0 0 06 26 4.2 0 2 5 HMDB0 0 28 3 D - R ibo se C5 H1 0 O5 15 0 .05 3 14 9 .04 55 5 - 1 M - H 0.0 0 06 1 4.0 9 HMDB0 0 64 6 L - A ra bi no se C5 H1 0 O5 15 0 .05 3 14 9 .04 55 5 - 1 M - H 0.0 0 06 1 4.0 9 26 HMDB0 0 23 9 Pyri d oxine C8 H1 1 NO3 16 9 .07 4 15 0 .05 55 1 - 1 M - H2 0 - H 0.0 0 06 26 4.1 7 27 HMDB0 1 18 2 6,8 - D ihyd roxyp urin e C5 H4 N4 O2 15 2 .03 3 15 1 .02 61 5 - 1 M - H 0.0 0 01 37 0.9 1 28 HMDB1 2 20 9 Die thylphosp ha te C4 H1 1 O4P 15 4 .03 9 15 3 .03 22 1 - 1 M - H 0.0 0 10 07 6.5 8 29 HMDB0 0 17 7 L - Histid in e C6 H9 N3 O2 15 5 .06 9 15 4 .06 22 0 - 1 M - H 0.0 0 01 37 0.8 9 30 HMDB0 0 15 9 L - P he ny la la ni ne C9 H1 1 NO2 16 5 .07 9 16 4 .07 17 0 - 1 M - H 0.0 0 02 9 1.7 7 31 HMDB0 0 33 7 (S) - 3 ,4 - D ihyd roxybutyri c acid C4 H8 O4 12 0 .04 2 16 5 .04 04 7 - 1 M+FA - H 0.0 0 07 93 4.8 0 32 HMDB0 1 36 6 Purin e C5 H4 N4 12 0 .04 4 16 5 .04 17 9 - 1 M+FA - H 0.0 0 01 98 1.2 0 33 HMDB0 0 28 9 Uric acid C5 H4 N4 O3 16 8 .02 8 16 7 .02 10 6 - 1 M - H 0.0 0 05 19 3.1 1 34 HMDB0 2 65 6 Prosta gla nd in A1 C2 0 H3 2 O4 33 6 .23 0 16 7 .10 77 6 - 2 M - 2 H 0.0 0 05 04 3.0 2 HMDB0 2 98 2 P rosta gla nd in B1 C2 0 H3 2 O4 33 6 .23 0 16 7 .10 77 6 - 2 M - 2 H 0.0 0 05 04 3.0 2 HMDB0 4 24 3 12 ( S) - HP ETE C2 0 H3 2 O4 33 6 .23 0 16 7 .10 77 6 - 2 M - 2 H 0.0 0 05 04 3.0 2 HMDB0 4 24 4 15 ( S) - HP ETE C2 0 H3 2 O4 33 6 .23 0 16 7 .10 77 6 - 2 M - 2 H 0.0 0 05 04 3.0 2 HMDB0 4 69 2 12 (R ) - HP ETE C2 0 H3 2 O4 33 6 .23 0 16 7 . 10 77 6 - 2 M - 2 H 0.0 0 05 04 3.0 2 HMDB0 4 69 3 11 H - 1 4 ,1 5 - EETA C2 0 H3 2 O4 33 6 .23 0 16 7 .10 77 6 - 2 M - 2 H 0.0 0 05 04 3.0 2 HMDB0 4 69 6 11 (R ) - HP ETE C2 0 H3 2 O4 33 6 .23 0 16 7 .10 77 6 - 2 M - 2 H 0.0 0 05 04 3.0 2 HMDB0 4 69 9 8(S) - HP ETE C2 0 H3 2 O4 33 6 .23 0 16 7 .10 77 6 - 2 M - 2 H 0.0 0 05 04 3.0 2 HMDB0 5 05 0 15 H - 1 1 ,1 2 - EETA C2 0 H3 2 O4 33 6 .23 0 16 7 .10 77 6 - 2 M - 2 H 0.0 0 05 04 3.0 2 HMDB1 0 20 4 14 ,1 5 - D iHETE C2 0 H3 2 O4 33 6 .23 0 16 7 .10 77 6 - 2 M - 2 H 0.0 0 05 04 3.0 2 HMDB1 0 21 1 17 ,1 8 - D iHETE C2 0 H3 2 O4 33 6 .23 0 16 7 .10 77 6 - 2 M - 2 H 0.0 0 05 04 3.0 2 HMDB1 0 21 6 5,1 5 - DiHETE C2 0 H3 2 O4 33 6 .23 0 16 7 .10 77 6 - 2 M - 2 H 0.0 0 05 04 3.0 2 HMDB1 0 21 9 8,1 5 - DiHETE C2 0 H3 2 O4 33 6 .23 0 16 7 .10 77 6 - 2 M - 2 H 0.0 0 05 04 3.0 2 HMDB1 1 13 5 5 - HP ETE C2 0 H3 2 O4 33 6 .23 0 16 7 .10 77 6 - 2 M - 2 H 0.0 0 05 04 3.0 2 36 HMDB0 2 24 3 Picolin ic acid C6 H5 NO2 12 3 .03 2 16 8 .03 02 3 - 1 M+FA - H 0.0 0 04 43 2.6 4 37 HMDB0 0 89 3 Sube ric acid C8 H1 4 O4 17 4 .08 9 17 3 .08 19 2 - 1 M - H 0.0 0 02 44 1.4 1 38 HMDB0 1 49 4 Ace tylphosp ha te C2 H5 O5 P 13 9 .98 7 17 4 .95 68 6 - 1 M+Cl 0.0 0 02 44 1.3 9 39 HMDB0 0 71 4 Hip p uric acid C9 H9 NO3 17 9 .05 8 17 8 .05 09 6 - 1 M - H 0.0 0 01 22 0.6 9 40 HMDB0 0 12 2 D - Glucose C6 H1 2 O 6 18 0 .06 3 17 9 .05 61 1 - 1 M - H 0.0 0 08 85 4.9 4 HMDB0 0 14 3 D - Ga la ctose C6 H1 2 O6 18 0 .06 3 17 9 .05 61 1 - 1 M - H 0.0 0 08 85 4.9 4 HMDB0 0 16 9 D - Ma nn ose C6 H1 2 O6 18 0 .06 3 17 9 .05 61 1 - 1 M - H 0.0 0 08 85 4.9 4 HMDB0 0 66 0 D - Fructose C6 H1 2 O6 18 0 .06 3 17 9 .05 61 1 - 1 M - H 0.0 0 08 85 4.9 4 41 H MDB0 0 15 8 L - Tyro sin e C9 H1 1 NO3 18 1 .07 4 18 0 .06 66 2 - 1 M - H 0.0 0 01 22 0.6 8 42 HMDB0 1 97 0 Methyluric acid C6 H6 N4 O3 18 2 .04 4 18 1 .03 67 1 - 1 M - H 0.0 0 05 49 3.0 3 43 HMDB0 0 20 9 Phe ny la ce tic acid C8 H8 O2 13 6 .05 2 18 1 .05 06 3 - 1 M+FA - H 0.0 0 03 05 1.6 8 44 HMDB0 1 39 2 p - A min obe nzoi c acid C7 H7 NO2 13 7 .04 8 18 2 .04 58 8 - 1 M+FA - H 0.0 0 00 31 0.1 7 45 HMDB0 1 90 8 19 - Hyd roxy - P GE2 C2 0 H3 2 O6 36 8 .22 0 18 3 .10 26 6 - 2 M - 2 H 0.0 0 07 17 3.9 2 HMDB0 1 97 9 6,1 5 - Dike to,1 3 ,14 - d ihyd ro - P GF1 a C2 0 H3 2 O6 36 8 .22 0 18 3 .10 26 6 - 2 M - 2 H 0.0 0 07 17 3.9 2 HMDB0 3 24 7 20 - Hyd roxy - P GE2 C 2 0 H3 2 O6 36 8 .22 0 18 3 .10 26 6 - 2 M - 2 H 0.0 0 07 17 3.9 2 HMDB0 4 24 1 6 - Ke top rosta gla nd in E1 C2 0 H3 2 O6 36 8 .22 0 18 3 .10 26 6 - 2 M - 2 H 0.0 0 07 17 3.9 2 HMDB0 4 24 2 11 - De hyd ro - thromboxa ne B2 C2 0 H3 2 O6 36 8 .22 0 18 3 .10 26 6 - 2 M - 2 H 0.0 0 07 17 3.9 2 HMDB1 2 11 0 5(6) - Epo xy Prosta gla ndin E1 C2 0 H3 2 O6 36 8 .22 0 18 3 .10 26 6 - 2 M - 2 H 0.0 0 07 17 3.9 2 46 HMDB0 6 02 9 N - A ce tylgluta min e C7 H1 2 N2 O4 18 8 .08 0 18 7 .07 24 3 - 1 M - H 0.0 0 10 68 5.7 1 47 HMDB0 0 32 5 3 - Hyd roxysube ric acid C8 H1 4 O5 19 0 .08 4 18 9 .07 68 4 - 1 M - H 0.0 0 08 7 4.6 0 48 HMDB0 0 09 4 Citric acid C6 H8 O7 19 2 .02 7 19 1 .01 97 3 - 1 M - H 0.0 0 07 02 3.6 8 49 HMDB0 2 12 3 1,3 ,7 - Trime thyluric acid C8 H1 0 N4 O3 21 0 .07 5 19 1 .05 69 0 - 1 M - H2 0 - H 0.0 0 03 36 1.7 6 5 0 HMDB0 0 12 7 D - Glucuron ic acid C6 H1 0 O7 19 4 .04 3 19 3 .03 53 7 - 1 M - H 0.0 0 01 68 0.8 7 HMDB0 2 54 5 Ga la cturon ic acid C6 H1 0 O7 19 4 .04 3 19 3 .03 53 7 - 1 M - H 0.0 0 01 68 0.8 7 51 HMDB0 2 14 2 Phosp hori c acid H3 O4 P 97 .9 77 19 4 .94 65 2 - 1 2M - H 0.0 0 09 31 4.7 8 52 HMDB0 0 56 5 Ga la cton ic acid C6 H1 2 O7 19 6 .05 8 19 5 .05 10 3 - 1 M - H 0.0 0 10 07 5.1 6 HMDB0 0 62 5 Glucon ic acid C6 H1 2 O7 19 6 .05 8 19 5 .05 10 3 - 1 M - H 0.0 0 10 07 5.1 6 53 HMDB 0 1 98 2 Dime thyluric acid C7 H8 N4 O3 19 6 .06 0 19 5 .05 23 7 - 1 M - H 0.0 0 02 59 1.3 3 54 HMDB0 0 44 0 3 - Hyd roxyp he ny la ce tic acid C8 H8 O3 15 2 .04 7 19 7 .04 55 5 - 1 M+FA - H 0.0 0 06 87 3.4 9 55 HMDB0 0 79 2 Seba cic acid C1 0 H1 8 O4 20 2 .12 1 20 1 .11 32 4 - 1 M - H 0.0 0 05 49 2.7 3 56 HMDB0 0 21 2 N - A ce tylga la ctosa min e C8 H1 5 NO6 22 1 .09 0 20 2 .07 15 5 - 1 M - H2 0 - H 0.0 0 08 09 4.0 0 HMDB0 0 21 5 N - A ce tyl - D - glucosa min e C8 H1 5 NO6 22 1 .09 0 20 2 .07 15 5 - 1 M - H2 0 - H 0.0 0 08 09 4.0 0 57 HMDB0 0 92 9 L - Tryp top ha n C1 1 H1 2 N2 O2 20 4 .09 0 20 3 .08 26 0 - 1 M - H 0.0 0 04 73 2.3 3 58 HMDB0 0 35 0 3 - Hyd roxy se ba cic acid C1 0 H1 8 O5 21 8 .11 5 21 7 .10 81 4 - 1 M - H 0.0 0 04 73 2.1 8 59 HMDB0 6 69 5 Prolylhyd roxyp roline C1 0 H1 6 N2 O4 22 8 .11 1 22 7 .10 37 3 - 1 M - H 0.0 0 05 65 2.4 9 60 HMDB0 0 01 9 Alp ha - ke toi sov a le ric acid C5 H8 O3 11 6 .04 7 23 1 .08 74 0 - 1 2M - H 0.0 0 13 89 6.0 1 HMDB0 0 31 0 Methyla ce to a ce tic acid C5 H8 O3 11 6 .04 7 23 1 .08 74 0 - 1 2M - H 0.0 0 13 89 6.0 1 HMDB1 2 23 3 Gluta ra te semia ld e hyde C5 H8 O3 11 6 .04 7 23 1 .08 74 0 - 1 2M - H 0.0 0 13 89 6.0 1 61 HMDB0 0 29 6 Urid ine C9 H1 2 N2 O6 24 4 .07 0 24 3 .06 22 6 - 1 M - H 0.0 0 04 12 1.7 0 62 HMDB0 0 49 7 5,6 - D ihyd rourid in e C9 H1 4 N2 O6 2 4 6 .08 5 24 5 .07 79 1 - 1 M - H 0.0 0 07 63 3.1 1 63 HMDB1 0 57 6 PG( 16 :0 / 18 :3( 6 Z,9 Z,1 2 Z)) C4 0 H7 3 O1 0P 74 4 .49 4 24 7 .15 74 4 - 3 M - 3 H 0.0 0 16 33 6.6 1 HMDB1 0 57 7 PG( 16 :0 / 18 :3( 9 Z,1 2 Z,15 Z) ) C4 0 H7 3 O1 0P 74 4 .49 4 24 7 .15 74 4 - 3 M - 3 H 0.0 0 16 33 6.6 1 HMDB1 0 59 0 PG( 16 :1 ( 9 Z)/ 18 :2( 9 Z,1 2 Z)) C 4 0 H7 3 O1 0P 74 4 .49 4 24 7 .15 74 4 - 3 M - 3 H 0.0 0 16 33 6.6 1 HMDB1 0 64 6 PG( 18 :2 ( 9 Z,12 Z) / 16 :1 ( 9 Z)) C4 0 H7 3 O1 0P 74 4 .49 4 24 7 .15 74 4 - 3 M - 3 H 0.0 0 16 33 6.6 1 HMDB1 0 66 0 PG( 18 :3 ( 6 Z,9 Z,1 2 Z)/ 16 :0) C4 0 H7 3 O1 0P 74 4 .49 4 24 7 .15 74 4 - 3 M - 3 H 0.0 0 16 33 6.6 1 HMDB1 0 67 5 PG( 18 :3 ( 9 Z,12 Z,1 5 Z) /1 6 :0 ) C4 0 H7 3 O1 0P 74 4 .49 4 24 7 .15 74 4 - 3 M - 3 H 0.0 0 16 33 6.6 1 64 HMDB0 0 00 5 2 - Ke tobutyri c acid C4 H6 O3 10 2 .03 2 24 9 .06 15 8 - 1 2M+ FA - H 0.0 0 17 7 7.1 1 65 HMDB0 0 69 5 Ketole ucin e C6 H1 0 O3 13 0 .06 3 25 9 .11 87 1 - 1 2M - H 0.0 0 10 38 4.0 1 66 HMDB0 3 22 4 Deoxyri bose C5 H1 0 O4 13 4 .05 8 26 7 .10 85 5 - 1 2M - H 0.0 0 26 25 9.8 3 67 HMDB0 0 15 3 Estrio l C1 8 H2 4 O3 28 8 .17 3 26 9 .15 41 4 - 1 M - H2 0 - H 0.0 0 24 41 9.0 7 HMDB0 0 33 8 2 - Hyd roxye stra dio l C1 8 H2 4 O3 28 8 .17 3 26 9 .15 41 4 - 1 M - H2 0 - H 0.0 0 24 41 9.0 7 68 HMDB0 0 61 3 Erythron ic acid C4 H8 O5 13 6 .03 7 27 1 .06 70 8 - 1 2M - H 0. 0 0 12 82 4.7 3 69 HMDB0 0 39 3 3 - He xen ed io ic acid C6 H8 O4 14 4 .04 2 28 7 .07 72 4 - 1 2M - H 0.0 0 09 16 3.1 9 70 HMDB0 0 87 2 Tetra d e ca ne d io ic acid C1 4 H2 6 O4 25 8 .18 3 29 5 .13 17 1 - 1 M+K - 2 H 0.0 0 00 92 0.3 1 71 HMDB0 0 10 1 Deoxyad e no sine C1 0 H1 3 N5 O3 25 1 .10 2 29 6 .10 00 4 - 1 M+FA - H 0.0 0 07 93 2.6 8 HMDB0 1 56 3 1 - Me thylgua no sin e C1 1 H1 5 N5 O5 29 7 .10 7 29 6 .10 00 7 - 1 M - H 0.0 0 07 63 2.5 8 72 HMDB0 0 81 4 N - A ce tylglucosa min e 6 - sulf a te C8 H1 5 NO9 S 30 1 .04 7 30 0 .03 94 9 - 1 M - H 0.0 0 14 34 4.7 8 73 HMDB0 0 23 0 N - A ce tylne ura min ic acid C1 1 H1 9 NO9 30 9 .10 6 30 8 .09 87 2 - 1 M - H 0.0 0 2 0 45 6.6 4 74 HMDB0 1 96 1 1,7 - D ime thylgua no sin e C1 2 H1 7 N5 O5 31 1 .12 3 31 0 .11 56 9 - 1 M - H 0.0 0 01 83 0.5 9 75 HMDB0 0 39 8 3 - Oxoa d ip ic acid C6 H8 O5 16 0 .03 7 31 9 .06 70 8 - 1 2M - H 0.0 0 03 36 1.0 5 76 HMDB0 1 43 4 3 - Me thoxytyrosine C1 0 H1 3 NO4 21 1 .08 4 32 4 .07 00 4 - 1 M+TFA - H 0.0 0 30 82 9.5 1 77 HMDB1 1 65 8 2,8 - D ihyd roxyquin oline - be ta - D - glucuron id e C1 5 H1 5 NO8 33 7 .08 0 33 6 .07 25 1 - 1 M - H 0.0 0 16 48 4.9 0 78 HMDB1 0 32 4 Be nzoy l glucuron id e C1 3 H1 4 O8 29 8 .06 9 34 3 .06 70 8 - 1 M+FA - H 0.0 0 12 51 3.6 5 79 HMDB0 0 15 0 Glucon ola cton e C6 H1 0 O6 17 8 .04 8 35 5 .08 82 3 - 1 2M - H 0 .0 0 26 55 7.4 8 HMDB0 1 35 3 2 - Ke to - 3 - d e oxy - D - glucon ic acid C6 H1 0 O6 17 8 .04 8 35 5 .08 82 3 - 1 2M - H 0.0 0 26 55 7.4 8 80 HMDB0 0 03 0 Biotin C1 0 H1 6 N2 O3 S 24 4 .08 8 35 7 .07 37 6 - 1 M+TFA - H 0.0 0 33 26 9.3 1 81 HMDB0 3 26 9 Nicotinuric acid C8 H8 N2 O3 18 0 .05 3 35 9 .09 97 3 - 1 2M - H 0.0 0 01 22 0 .3 4 82 HMDB0 0 24 4 Ribo f la vin C1 7 H2 0 N4 O6 37 6 .13 8 37 5 .13 10 1 - 1 M - H 0.0 0 00 61 0.1 6 83 HMDB0 0 16 3 D - Ma ltose C1 2 H2 2 O1 1 34 2 .11 6 37 7 .08 56 0 - 1 M+Cl 0.0 0 16 78 4.4 5 8 4 HMDB1 1 55 4 MG(0 :0 / 22 :4( 7 Z,1 0 Z,13 Z,1 6 Z) /0 :0) C2 5 H4 2 O4 40 6 .30 8 38 7 .28 99 2 - 1 M - H2 0 - H 0.0 0 18 92 4.8 9 HM D B1 1 58 4 MG(2 2:4 (7 Z,1 0 Z,1 3 Z,16 Z) / 0 :0 /0 :0) C2 5 H4 2 O4 40 6 .30 8 38 7 .28 99 2 - 1 M - H2 0 - H 0.0 0 18 92 4.8 9 85 HMDB0 4 07 6 5 - Hyd roxykynure na mine C9 H1 2 N2 O2 18 0 .09 0 40 5 .17 79 5 - 1 2M+ FA - H 0.0 0 02 75 0.6 8 86 HMDB0 1 45 1 Lip oi c acid C8 H1 4 O2 S2 20 6 .04 4 41 1 .07 97 7 - 1 2M - H 0.0 0 26 25 6.3 9 87 HMDB0 1 90 3 Calcitrio l C2 7 H4 4 O3 41 6 .32 9 41 5 .32 17 8 - 1 M - H 0.0 0 25 02 6.0 2 88 HMDB1 3 50 1 PGP (1 6:1 (9 Z) /2 2 :6 (4 Z,7 Z,1 0 Z,1 3 Z,1 6 Z, 1 9 Z) ) C4 4 H7 4 O1 3P 87 2 .46 0 43 5 .22 29 6 - 2 M - 2 H 0.0 0 21 06 4.8 4 HMDB1 3 57 2 PGP (1 8:3 (6 Z,9 Z,1 2 Z) /2 0 :4 (5 Z,8 Z,1 1 Z,1 4 Z) ) C4 4 H7 4 O1 3P 87 2 .46 0 43 5 .22 29 6 - 2 M - 2 H 0.0 0 21 06 4.8 4 HMDB1 3 58 7 PGP (1 8:3 (9 Z,1 2 Z,1 5 Z)/ 2 0:4( 5 Z,8 Z,1 1 Z, 1 4 Z) ) C4 4 H7 4 O1 3P 87 2 .46 0 43 5 .22 29 6 - 2 M - 2 H 0.0 0 21 06 4.8 4 89 HMDB1 3 48 6 PGP (1 6:0 /2 2 :6 (4 Z,7 Z,1 0 Z,1 3 Z,1 6 Z,1 9 Z) ) C4 4 H7 6 O1 3P 87 4 .47 6 43 6 .23 08 0 - 2 M - 2 H 0.0 0 21 67 4.9 7 HMDB1 3 49 9 PGP (1 6 :1 (9 Z) /2 2 :5 (4 Z,7 Z,1 0 Z,1 3 Z,1 6 Z) ) C4 4 H7 6 O1 3P 87 4 .47 6 43 6 .23 08 0 - 2 M - 2 H 0.0 0 21 67 4.9 7 HMDB1 3 50 0 PGP (1 6:1 (9 Z) /2 2 :5 (7 Z,1 0 Z,1 3 Z,1 6 Z,1 9 Z ) ) C4 4 H7 6 O1 3P 87 4 .47 6 43 6 .23 08 0 - 2 M - 2 H 0.0 0 21 67 4.9 7 HMDB1 3 55 7 PGP (1 8:2 (9 Z,1 2 Z) /2 0 :4 (5 Z,8 Z,1 1 Z,1 4 Z) ) C4 4 H7 6 O1 3P 87 4 .47 6 43 6 .23 08 0 - 2 M - 2 H 0.0 0 21 67 4.9 7 HMDB1 3 57 0 PGP (1 8:3 (6 Z,9 Z,1 2 Z) /2 0 :3 (5 Z,8 Z,1 1 Z) ) C4 4 H7 6 O1 3P 87 4 .47 6 43 6 .23 08 0 - 2 M - 2 H 0.0 0 21 67 4.9 7 HMDB1 3 57 1 PGP (1 8:3 (6 Z,9 Z,1 2 Z) /2 0 :3 (8 Z,1 1 Z,1 4 Z) ) C4 4 H7 6 O1 3P 87 4 .47 6 43 6 .23 08 0 - 2 M - 2 H 0.0 0 21 67 4.9 7 HMDB1 3 58 5 PGP (1 8:3 (9 Z,1 2 Z,1 5 Z)/ 2 0:3( 5 Z,8 Z,1 1 Z) ) C4 4 H7 6 O1 3P 87 4 .47 6 43 6 .23 08 0 - 2 M - 2 H 0.0 0 21 67 4.9 7 HMDB1 3 58 6 PGP (1 8:3 (9 Z,1 2 Z,1 5 Z)/ 2 0:3( 8 Z,1 1 Z,1 4 Z ) ) C4 4 H7 6 O1 3P 87 4 .47 6 43 6 .23 08 0 - 2 M - 2 H 0.0 0 21 67 4.9 7 90 HMDB0 0 47 2 5 - Hyd roxy - L - tryp top ha n C1 1 H1 2 N2 O3 22 0 .08 5 43 9 .16 23 2 - 1 2M - H 0.0 0 00 61 0.1 4 9 1 HMDB0 0 13 8 Glycocholic acid C2 6 H4 3 NO6 46 5 .30 9 44 6 .29 06 5 - 1 M - H2 0 - H 0.0 0 15 26 3.4 2 92 HMDB0 0 63 1 Deoxycholic acid glyci ne conjuga te C2 6 H4 3 NO5 44 9 .31 4 44 8 .30 68 5 - 1 M - H 0.0 0 08 24 1.8 4 93 HMDB1 0 33 9 3 - a lp ha - a nd rosta ned io l glucuron id e C2 5 H4 0 O8 46 8 .27 2 44 9 .25 39 1 - 1 M - H2 0 - H 0.0 0 15 26 3.4 0 94 HMDB1 3 60 9 D - Tryp top ha n C1 1 H1 2 N2 O2 20 4 .09 0 45 3 .17 79 5 - 1 2M+ FA - H 0.0 0 13 73 3.0 3 95 HMDB0 1 09 3 5a - Chole sta - 8 ,2 4 - d ie n - 3 - o n e C2 7 H4 2 O 38 2 .32 4 46 1 .24 24 6 - 1 M+Br 0.0 0 03 36 0.7 3 HMDB0 2 39 4 Chole sta - 4 ,6 - d ie n - 3 - one C2 7 H4 2 O 38 2 . 32 4 46 1 .24 24 6 - 1 M+Br 0.0 0 01 83 0.4 0 96 HMDB1 1 53 9 MG(0 :0 / 18 :3( 6 Z,9 Z,1 2 Z) /0 :0 ) C2 1 H3 6 O4 35 2 .26 1 46 5 .24 69 5 - 1 M+TFA - H 0.0 0 37 84 8.1 3 HMDB1 1 53 9 MG(0 :0 / 18 :3( 6 Z,9 Z,1 2 Z) /0 :0 ) C2 1 H3 6 O4 35 2 .26 1 46 5 .24 69 5 - 1 M+TFA - H 0.0 0 38 45 8.2 6 HMDB1 1 54 0 MG(0 :0 / 18 :3( 9 Z,1 2 Z,15 Z ) / 0:0 ) C2 1 H3 6 O4 35 2 .26 1 46 5 .24 69 5 - 1 M+TFA - H 0.0 0 37 84 8.1 3 HMDB1 1 54 0 MG(0 :0 / 18 :3( 9 Z,1 2 Z,15 Z) / 0:0 ) C2 1 H3 6 O4 35 2 .26 1 46 5 .24 69 5 - 1 M+TFA - H 0.0 0 38 45 8.2 6 HMDB1 1 56 9 MG(1 8:3 (6 Z,9 Z,1 2 Z) /0 :0 /0 :0 ) C2 1 H3 6 O4 35 2 .26 1 46 5 .24 69 5 - 1 M+TFA - H 0.0 0 37 84 8.1 3 HMDB1 1 56 9 MG(1 8:3 (6 Z,9 Z,1 2 Z) /0 :0 /0 :0 ) C2 1 H3 6 O4 35 2 .26 1 46 5 .24 69 5 - 1 M+TFA - H 0.0 0 38 45 8.2 6 HMDB1 1 57 0 MG(1 8:3 (9 Z,1 2 Z,1 5 Z)/ 0 :0 / 0:0 ) C2 1 H3 6 O4 35 2 .26 1 46 5 .24 69 5 - 1 M+TFA - H 0.0 0 37 84 8.1 3 HMDB1 1 57 0 MG(1 8:3 (9 Z,1 2 Z,1 5 Z)/ 0 :0 / 0:0 ) C2 1 H3 6 O4 35 2 .26 1 46 5 .24 69 5 - 1 M+TFA - H 0.0 0 38 45 8.2 6 97 HMDB0 2 82 9 And roste ron e glucuron id e C2 5 H3 8 O8 46 6 .25 7 46 5 .24 93 9 - 1 M - H 0.0 0 13 43 2.8 9 98 HMDB1 3 04 3 Prosta gla nd in PGE2 1 - glyce ryl este r C2 3 H3 8 O7 42 6 .26 2 47 1 .25 99 5 - 1 M+FA - H 0.0 0 17 7 3.7 6 HMDB1 3 04 5 Prosta gla nd in PGE2 glyce ryl este r C2 3 H3 8 O7 42 6 . 26 2 47 1 .25 99 5 - 1 M+FA - H 0.0 0 17 7 3.7 6 HMDB1 3 65 3 Prosta gla nd in D2 - 1 - glyce ryl este r C2 3 H3 8 O7 42 6 .26 2 47 1 .25 99 5 - 1 M+FA - H 0.0 0 17 7 3.7 6 99 HMDB0 9 92 7 PIP ( 16 :0/ 20 :1( 11 Z) ) C4 5 H8 6 O1 6P 2 94 4 .53 9 47 1 .26 23 0 - 2 M - 2 H 0.0 0 05 8 1.2 3 HMDB0 9 94 9 PIP ( 18 :0/ 18 :1( 11 Z) ) C4 5 H8 6 O1 6P 2 94 4 .53 9 47 1 .26 23 0 - 2 M - 2 H 0.0 0 05 8 1.2 3 HMDB0 9 95 0 PIP ( 18 :0/ 18 :1( 9 Z)) C4 5 H8 6 O1 6P 2 94 4 .53 9 47 1 .26 23 0 - 2 M - 2 H 0.0 0 05 8 1.2 3 HMDB0 9 96 0 PIP ( 18 :1( 11 Z) /1 8 :0 ) C4 5 H8 6 O1 6P 2 94 4 .53 9 47 1 .26 23 0 - 2 M - 2 H 0.0 0 05 8 1.2 3 HMDB0 9 97 2 PIP ( 18 :1( 9 Z)/ 18 :0) C4 5 H8 6 O1 6P 2 94 4 .53 9 47 1 .26 23 0 - 2 M - 2 H 0.0 0 05 8 1.2 3 HMDB0 9 99 8 PIP ( 20 :1( 11 Z) /1 6 :0 ) C4 5 H8 6 O1 6P 2 94 4 .53 9 47 1 .26 23 0 - 2 M - 2 H 0.0 0 05 8 1.2 3 1 0 0 HMDB0 9 93 7 PIP ( 16 :0/ 22 :5( 4 Z,7 Z,1 0 Z,1 3 Z,16 Z) C4 7 H8 2 O1 6P 96 4 .50 8 48 1 .24 66 - 2 M - 2 H 0.0 0 04 27 0.8 9 HMDB0 9 93 8 PIP ( 16 :0/ 22 :5( 7 Z,1 0 Z,13 Z,1 6 Z,1 9 Z) ) C4 7 H8 2 O1 6P 2 96 4 .50 8 48 1 .24 66 - 2 M - 2 H 0.0 0 04 27 0.8 9 HMDB0 9 94 5 PIP ( 16 :2( 9 Z,12 Z) / 22 :3( 10 Z,1 3Z,1 6 Z) ) C4 7 H8 2 O1 6P 2 96 4 .50 8 48 1 .24 6 6 - 2 M - 2 H 0.0 0 04 27 0.8 9 HMDB0 9 96 7 PIP ( 18 :1( 11 Z) /2 0 :4 (5 Z,8 Z,1 1 Z,1 4 Z) ) C4 7 H8 2 O1 6P 2 96 4 .50 8 48 1 .24 66 - 2 M - 2 H 0.0 0 04 27 0 .8 9 HMDB0 9 96 8 PIP ( 18 :1( 11 Z) /2 0 :4 (8 Z,1 1 Z,1 4Z,1 7 Z) ) C4 7 H8 2 O1 6P 2 96 4 .50 8 48 1 .24 66 - 2 M - 2 H 0.0 0 04 27 0.8 9 HMDB0 9 98 0 PIP ( 18 :1( 9 Z)/ 20 :4( 5 Z,8 Z,1 1 Z,14 Z) C4 7 H8 2 O1 6P 2 96 4 .50 8 48 1 .24 66 - 2 M - 2 H 0.0 0 04 27 0.8 9 HMDB0 9 98 1 PIP ( 18 :1( 9 Z)/ 20 :4( 8 Z,1 1 Z,1 4 Z,1 7 Z) ) C4 7 H8 2 O1 6P 2 96 4 .50 8 48 1 .24 66 - 2 M - 2 H 0.0 0 04 27 0.8 9 HMDB0 9 99 1 P IP ( 18 :2( 9 Z,12 Z) / 20 :3( 5 Z,8 Z,11 Z) C4 7 H8 2 O1 6P 2 96 4 .50 8 48 1 .24 66 4 - 2 M - 2 H 0.0 0 04 27 0.8 9 HMDB0 9 99 2 PIP ( 18 :2( 9 Z,12 Z) / 20 :3( 8 Z,1 1 Z,1 4 Z) ) C4 7 H8 2 O1 6P 2 96 4 .50 8 48 1 .24 66 - 2 M - 2 H 0.0 0 04 27 0.8 9 HMDB1 0 00 9 PIP ( 20 :3( 5 Z,8 Z,1 1 Z)/ 18 :2( 9 Z,12 Z) ) C4 7 H8 2 O1 6P 2 96 4 .50 8 48 1 .24 66 - 2 M - 2 H 0.0 0 04 27 0.8 9 HMDB1 0 01 4 PIP ( 20 :3( 8 Z,11 Z,1 4 Z) /1 8 :2 ( 9 Z,1 2 Z) ) C4 7 H8 2 O1 6P 2 96 4 .50 8 48 1 .24 66 - 2 M - 2 H 0.0 0 04 27 0.8 9 HMDB1 0 01 7 PIP ( 20 :4( 5 Z,8 Z,1 1 Z,14 Z) / 18 :1(1 1 Z) ) C4 7 H8 2 O1 6P 2 96 4 .50 8 48 1 .24 66 - 2 M - 2 H 0.0 0 04 27 0.8 9 HMDB1 0 01 8 P IP ( 20 :4( 5 Z,8 Z,1 1 Z,14 Z) / 18 :1(9 Z) C4 7 H8 2 O1 6P 2 96 4 .50 8 48 1 .24 66 - 2 M - 2 H 0.0 0 04 27 0.8 9 HMDB1 0 02 1 PIP ( 20 :4( 8 Z,11 Z,1 4 Z,1 7 Z) /1 8 :1( 1 1 Z)) C4 7 H8 2 O1 6P 2 96 4 .50 8 48 1 .24 66 - 2 M - 2 H 0.0 0 04 27 0.8 9 HMDB1 0 02 2 PIP ( 20 :4( 8 Z,11 Z,1 4 Z,1 7 Z ) /1 8 :1( 9 Z) ) C4 7 H8 2 O1 6P 2 96 4 .50 8 48 1 .24 66 - 2 M - 2 H 0.0 0 04 27 0.8 9 HMDB1 0 02 4 PIP ( 22 :3( 10 Z,1 3 Z,1 6 Z)/ 1 6:2 (9Z,1 2 Z) ) C4 7 H8 2 O1 6P 2 96 4 .50 8 48 1 .24 66 - 2 M - 2 H 0.0 0 04 27 0.8 9 HMDB1 0 02 9 PIP ( 22 :5( 4 Z,7 Z,1 0 Z,13 Z,1 6 Z) /16 :0) C4 7 H8 2 O1 6P 2 96 4 .50 8 48 1 .24 66 4 - 2 M - 2 H 0.0 0 04 27 0.8 9 HMDB1 0 03 0 PIP ( 22 :5( 7 Z,10 Z,1 3 Z,1 6 Z,1 9 Z)/1 6 :0 ) C4 7 H8 2 O1 6P 2 96 4 .50 8 48 1 .24 66 - 2 M - 2 H 0.0 0 04 27 0.8 9 10 1 HMDB0 3 13 4 Biocy tin C1 6 H2 8 N4 O4 S 37 2 .18 3 48 5 .16 87 3 - 1 M+TFA - H 0.0 0 00 92 0.1 9 10 2 HMDB1 1 54 4 MG(0 :0 / 20 :2( 11 Z,1 4 Z) /0 :0) C2 3 H4 2 O4 38 2 .30 8 49 5 .29 39 2 - 1 M + TFA - H 0.0 0 41 5 8.3 8 HMDB1 1 57 4 MG(2 0:2 (1 1 Z,14 Z) / 0 :0 /0 :0) C2 3 H4 2 O4 38 2 .30 8 49 5 .29 39 2 - 1 M+TFA - H 0.0 0 41 5 8.3 8 10 3 HMDB0 6 34 4 Alp ha - N - P he ny la ce tyl - L - gluta min e C1 3 H1 6 N2 O4 26 4 .11 1 52 7 .21 47 2 - 1 2M - H 0.0 0 16 48 3.1 3 10 4 HMDB0 0 03 0 Biotin C1 0 H1 6 N2 O3 S 24 4 .08 8 53 3 .17 45 6 - 1 2M+ FA - H 0.0 0 07 93 1.4 9 10 5 HMDB1 0 35 7 Tetra hyd roa ld oste ron e - 3 - glucuron id e C2 7 H4 0 O1 1 54 0 .25 7 53 9 .24 98 2 - 1 M - H 0.0 0 25 02 4.6 4 10 6 HMDB0 1 52 6 S - A ce tyld ihyd rolip oa mid e C1 0 H1 9 NO2 S2 24 9 .08 6 54 3 .16 96 8 - 1 2M+ FA - H 0.0 0 13 43 2.4 7 10 7 HMDB0 2 59 6 Deoxycholic acid 3 - glucuron id e C3 0 H4 8 O1 0 56 8 .32 5 58 9 .29 94 4 - 1 M+Na - 2 H 0.0 0 03 66 0.6 2 10 8 HMDB0 0 82 5 3' - Sia lylla ctose C2 3 H3 9 NO1 9 63 3 .21 2 63 2 .20 43 5 - 1 M - H 0.0 0 15 26 2.4 1 10 9 HMDB0 6 58 4 6 - Sia lyl - N - a ce tylla ctosa min e C2 5 H4 2 N2 O19 67 4 .23 8 67 3 .23 09 0 - 1 M - H 0.0 0 39 67 5.8 9 11 0 HMDB0 0 1 1 5 Glycolic acid C2 H4 O3 76 .0 16 75 .0 08 77 - 1 M - H 0.0 0 04 12 5.4 9 11 1 HMDB0 1 03 3 Hyd roge n sulf ite [HSO3 ] - 80 .9 65 79 .9 57 36 - 1 M - H 0.0 0 00 53 0.6 6 11 2 HMDB0 7 41 1 DG( 20 :1 ( 11 Z) /2 2 :6 (4 Z,7 Z,1 0 Z,1 3 Z,1 6 Z, 1 9 Z) /0 :0) C4 5 H7 4 O5 69 4 .55 4 80 7 .53 92 5 - 1 M+TFA - H 0.0 0 06 71 0.8 3 HMD B0 7 43 8 DG( 20 :2 ( 11 Z,1 4 Z) /2 2:5 (4 Z,7 Z,1 0 Z,1 3 Z, 1 6 Z) /0 :0) C4 5 H7 4 O5 69 4 .55 4 80 7 .53 92 5 - 1 M+TFA - H 0.0 0 06 71 0.8 3 HMDB0 7 43 9 DG( 20 :2 ( 11 Z,1 4 Z) /2 2:5 (7 Z,1 0 Z,1 3 Z,1 6 Z ,1 9 Z) / 0:0 ) C4 5 H7 4 O5 69 4 .55 4 80 7 .53 92 5 - 1 M+TFA - H 0.0 0 06 71 0.8 3 HMDB0 7 46 6 DG( 20 :3 ( 5 Z,8 Z,1 1 Z)/ 22 :4( 7 Z,1 0 Z,1 3 Z,1 6 Z) / 0 :0 ) C4 5 H7 4 O5 69 4 .55 4 80 7 .53 92 5 - 1 M+TFA - H 0.0 0 06 71 0.8 3 HMDB0 7 49 5 DG( 20 :3 ( 8 Z,11 Z,1 4 Z) /2 2 :4 (7 Z,1 0 Z,1 3 Z, 1 6 Z) /0 :0) C4 5 H7 4 O5 69 4 .55 4 80 7 .53 92 5 - 1 M+TFA - H 0.0 0 06 71 0.8 3 HMDB0 7 58 1 DG( 20 :5 ( 5 Z,8 Z,1 1 Z,14 Z,1 7 Z) /22 :2( 1 3 Z, 1 6 Z) /0 :0) C4 5 H7 4 O5 69 4 .55 4 80 7 .53 92 5 - 1 M+TFA - H 0.0 0 06 71 0.8 3 HMDB0 7 66 5 DG( 22 :2 ( 13 Z,1 6 Z) /2 0:5 (5 Z,8 Z,1 1 Z,1 4 Z, 1 7 Z) /0 :0) C4 5 H7 4 O5 69 4 .55 4 80 7 .53 92 - 1 M+TFA - H 0.0 0 06 71 0.8 3 HMDB0 7 69 0 DG( 22 :4 ( 7 Z,10 Z,1 3 Z,1 6 Z) /2 0 :3( 5 Z,8 Z,1 1 Z) / 0 :0 ) C4 5 H7 4 O5 69 4 .55 4 80 7 .53 92 - 1 M+TFA - H 0.0 0 06 71 0.8 3 HM D B0 7 69 1 DG( 22 :4 ( 7 Z,10 Z,1 3 Z,1 6 Z) /2 0 :3( 8 Z,1 1 Z, 1 4 Z) /0 :0) C4 5 H7 4 O5 69 4 .55 4 80 7 .53 92 - 1 M+TFA - H 0.0 0 06 71 0.8 3 HMDB0 7 71 8 DG( 22 :5 ( 4 Z,7 Z,1 0 Z,13 Z,1 6 Z) /20 :2( 1 1 Z, 1 4 Z) /0 :0) C4 5 H7 4 O5 69 4 .55 4 80 7 .53 92 - 1 M+TFA - H 0.0 0 06 71 0.8 3 HMDB0 7 74 7 DG( 22 :5 ( 7 Z,10 Z,1 3 Z,1 6 Z,1 9 Z)/ 2 0 :2 ( 11 Z ,1 4 Z) / 0:0 ) C4 5 H7 4 O5 69 4 .55 4 80 7 .53 92 - 1 M+TFA - H 0.0 0 06 71 0.8 3 HMDB0 7 77 5 DG( 22 :6 ( 4 Z,7 Z,1 0 Z,13 Z,1 6 Z,1 9Z) / 2 0 :1 ( 1 1 Z) /0 :0) C4 5 H7 4 O5 69 4 .55 4 80 7 .53 92 - 1 M+TFA - H 0.0 0 06 71 0.8 3 11 3 HMDB0 0 35 7 3 - Hyd roxybutyri c acid C4 H8 O3 10 4 .04 7 85 .0 28 95 - 1 M - H2 0 - H 0.0 0 06 71 7.8 9 HMDB0 0 34 1 2 - Octe ne dio ic acid C8 H1 2 O4 17 2 .07 4 85 .0 29 50 - 2 M - 2 H 0.0 0 01 14 1.3 4 11 4 HMDB0 0 13 9 Glyceric acid C3 H6 O4 10 6 .02 7 87 .0 08 22 - 1 M - H2 0 - H 0.0 0 08 24 9.4 7 HMDB0 0 24 3 Pyruvi c acid C3 H4 O3 88 .0 16 87 .0 08 77 - 1 M - H 0.0 0 02 82 3.2 4 495 Sent to Anal. Bioanal. Chem. Chapter 13 4. Conclusions In this study, the metabolic profile of urine after intake of four different oils was obtained. These oils were olive oil and sunflower oil, pure and enriched with natural antioxidants or a synthetic oxidation inhibitor such as dimethylsiloxane. Urine samples obtained before (blank) and 2 and 4 h after intake were analyzed by LC?TOF/MS. PCA and PLS-DA were applied to establish differences among individuals after ingestion of prepared breakfasts and to build classification and prediction models. This study demonstrates the effect of oils composition on the urinary metabolic profile, showing statistical differences among samples obtained after intake of oils with different composition. 5. Acknowledgements The Spanish Ministerio de Ciencia e Innovaci?n (MICINN) and FEDER program are thanked for financial support through project CTQ2009-07430. B.A.-S. and F.P.-C. are also grateful to the MICINN for an FPI scholarship (BES-2007-15043) and a Ram?n y Cajal contract (RYC-2009-03921). 5. References [1] Wishart DS (2008) Trends Food Sci Technol 19:482 [2] P?rez -Jim?nez F, Espino A, L?pez -Segura F, Blanco J, Ruiz-Guti?rrez V, Prada JL, L?pez-Miranda J, Jim?nez -Perez J, Ordovas JM (1995) Am J Clin Nutr 62:769 496 Nuevas plataformas anal?ticas en metabol?mica [3] Hegsted DM, McGandy RB, Myers ML, Stare FJ (1965) Am J Clin Nutr 17:281 [4] Moreiras-Varela O (1989) Eur J Clin Nutr 43:83 [5] Jap?n-Luj?n R, Luque de Castro MD (2008) J Agric Food Chem 56:2505 [6] Servili M, Selvaggini R, Esposto S, Taticchi A, Montedoro GF, Morozzi G (2004) J Chromatogr A 1054:113 [7] Laporta O, Perez-Fons L, Mallavia R, Caturla N, Micol V (2006) Food Chem 101:1425 [8] Zhong H, Bedgood D R, Bishop A G, Prenzler PD, Robards K (2006) Food Chem 100:1544 [9] Shahidi F (1996) Natural Antioxidants Chemistry, Health Effect and Applications, AOCS Press. USA [10] Manna C, Migliardi V, Golino P, Scognamiglio A, Galletti P, Chiariello M, Zappia, V (2004) J Nutr Biochem 15:461 [11] Watanabe S, Yamaguchi M, Sobue T, Takahashi T, Miura T, Arai Y, et al (1998) J Nutr 128:1710 [12] Llorach R, Garrido I, Monagas M, Urpi-Sarda M, Tulipani S, Bartolom? B, Andr?s -Lacueva C (2010) J Proteome Res 9:5859 [13] Llorach R, Urpi-Sarda M, Jauregui O, Monagas M, Andr?s -Lacueva C (2009) J Proteome Res 8:5060 [14] Gibney MJ, Walsh M, Brennan L, Roche HM, German B, van Ommen B (2005) Am J Clin Nutr 82:497 [15] Zhang X, Yap Y, Wei D, Chen G, Chen F (2008) Biotechnol Adv 26:169 [16] Liu G, Wang Y, Wang Z, Cai J, Lv X, Zhou A (2011) J Agric Food Chem 59:5572 [17] ?lvarez -S?nchez B, Priego-Capote F, Luque de Castro MD (2010) Trends Anal Chem 29:111 497 Sent to Anal. Bioanal. Chem. Chapter 13 [18] Lindom, JC, Nicholson JK, Holmes E (2007) The Handbook of Metabonomics and Metabolomics, Elsevier, Oxford [19] Kolokolova TN, Savelev OY, Sergeev NM (2008) J Anal Chem 63:104 [20] Giovane, A, Balestrieri, A, Napoli, C (2008) J Cell Biochem 105:648 [21] Betz M, Saxena K, Schwalbe H (2006) Curr Opin Chem Biol 10:219 [22] Kemsley EK, Le Gall G, Dainty JR, Watson AD, Harvey LJ, Tapp HS, Colquhoun IJ (2007) Br J Nutr 98:1 [23] Styczynski MP, Moxley JF, Tong LV, Walther JL, Jensen KL, Stephanopoulos GN (2007) Anal Chem 79:966 [24] Kristal BS, Shurubor YI, Kaddurah-Daouk R, Matson WR (2007) Methods Mol Biol 358:159 [25] ?ir?n M?, Ruiz-?im?ne? ?, Luque de Castro MD (2009) J Agric Food Chem 57:2797 [26] Yuille M, Illig T, Hveem K, Schmitz G, Hansen J, Neumaier M, Tybring G, Wichmann E, Ollier B (2010) Biopreserv Biobank 8:65 [27] Bernini P, Bertini I, Luchinat C, Nincheri P, Staderini S, Turano P (2011) J Biomol NMR 49:231 [28] METLIN database http://metlinscrippsedu/ [29] Wishart DS, Tzur D, Knox C et al (2007) Nucleic Acids Res 35:D521 [30] Morita O, Knapp JF, Tamaki Y, Nemec MD, Varsho BJ, Stump DG (2008) Food Chem Toxicol 46:3059 [31] Tan B, O'Dell DK, Yu YW, Monn MF, Hughes HV, Burstein S, Walker JM (2010) J Lipid Res 51:112 [32] Mandel HG, Kusmierz JJ, Dickens BF, Anderson LW (1994) Anal Biochem 217:292 [33] Chen HJ, Wu CF, Huang JL (2005) Toxicol Lett 155:403 498 Nuevas plataformas anal?ticas en metabol?mica [34] Guneral F, Bachmann C (1994) Clin Chem 40:862 [35] Geigy Scientific Tables, 8th Rev edition, 130 Ed. C Lentner, West Cadwell, NJ: Medical education Div, Ciba-Geigy Corp Basel, Switzerland c1981 D I S C U S I ? N DE LO S RESU LTADOS 501 Discusi?n de los resultados La normativa vigente en la Universidad de C?rdoba para la elaboraci?n de una Memoria de Tesis Doctoral en la modalidad en la que se incluyen los art?culos (publicados o pr?ximos a su publicaci?n) como tales requiere una discusi?n conjunta de los resultados, lo que puede ser m?s o menos factible dependiendo de la homogeneidad de la investigaci?n realizada. La investigaci?n recogida en esta Memoria tiene como denominador com?n la metabol?mica y, en ella, el desarrollo de plataformas anal?ticas basadas en las diversas estrategias caracter?sticas de esta disciplina. Por tanto, se discutir?n de forma separada las plataformas basadas en cada estrategia Discusi?n independiente merecen las publicaciones sobre las revisiones bibliogr?ficas dedicadas a las operaciones previas y a la preparaci?n de muestra frecuentemente utilizadas en an?lisis metabol?mico. Pa rte 1: Revisi?n sobre la preparaci?n de la muestra La investigaci?n sobre la documentaci?n bibliogr?fica realizada por la doctoranda durante los primeros a?os de desarrollo de su Tesis alert? sobre la situaci?n de las etapas previas al an?lisis en metabol?mica: Etapas que han recibido una atenci?n desigual, pobre en general. Este hecho se hizo patente en la reuni?n organizada en 2005 por la Sociedad de Metabol?mica (www.metabolomicssociety.org) y que dio lugar a la conocida como ?Metabolomics Standard Initiative? enfocada a la estandari?aci?n de las diferentes etapas y procesos necesarios para la obtenci?n de informaci?n biol?gica en experimentos en metabol?mica. El resultado de esta iniciativa es asegurar la transferibilidad de m?todos de an?lisis y de estudios con el fin de posibilitar la reproducci?n exacta de ?stos en cualquier laboratorio. La alerta condujo a la doctoranda al desarrollo de un estudio en profundidad de este aspecto que, a pesar de su enorme repercusi?n en los resultados finales de una investigaci?n, no recibe la atenci?n que merece. Una especie de 502 Nuevas plataformas anal?ticas en metabol?mica recopilaci?n cr?tica para un uso m?s acertado de lo ya existente, junto con una llamada de atenci?n de las deficiencias y carencias en esta parcela, es lo que se ha pretendido conseguir con la elaboraci?n y publicaci?n de la revisi?n realizada y que constituye la Parte I de esta Memoria de Tesis. La extensi?n de esta revisi?n condujo a su divisi?n en dos publicaciones en el mismo volumen de la revista TrAC (Trends in Analytical Chemistry), que lideraba en el momento de la publicaci?n la clasificaci?n de la Journal Citation Reports (ISI Web of Knowledge) en base al ?ndice de impacto. En la primera de estas revisiones, con el t?tulo ?Metabolomics analysis I. Selection of biological samples and practical aspects preceding sample preparation?, se describen las caracter?sticas de la metabol?mica en comparaci?n (y en relaci?n) con otras ?micas menos recientes. Se pone tambi?n de manifiesto el escaso inter?s que se ha prestado a las primeras etapas del proceso anal?tico en metabolomica; lo que queda patente por el n?mero de revisiones que se han publicado sobre la instrumentaci?n utilizada en metabol?mica y la ausencia de ellas sobre las etapas anteriores al uso de esta instrumentaci?n, que son claves para que los instrumentos proporcionen una respuesta sin errores ni sesgos. Hablar de las diferentes estrategias en metabol?mica resulta imprescindible en estos casos, ya que la preparaci?n de la muestra es diferente para cada una de ellas; como imprescindible es tambi?n un esquema en el que el lector ubique cada una de las etapas a desarrollar. Ambos aspectos conducen a la discusi?n sobre la selecci?n del material biol?gico m?s interesante para el muestreo (generalmente sangre u orina, pero tambi?n c?lulas o tejidos, adem?s de otros fluidos como aliento, saliva, l?grimas, l?quido cefalorraqu?deo, leche materna, l?quido sinovial o plasma seminal) y la forma m?s adecuada de detener la actividad enzim?tica en la muestra (quenching) para conseguir representatividad del estado del sistema en estudio en el momento del muestreo. La forma de conservaci?n de la muestra dependiendo de sus caracter?sticas ha sido objeto de muy escasa investigaci?n, a pesar de su papel clave en los cambios que se producen durante el almacenamiento y en los casos de descongelaci?n/congelaci?n repetida. Los errores a que da 503 Discusi?n de los resultados lugar un mal desarrollo de esta etapa se enfatizan en la publicaci?n, donde tambi?n se incluye la orientaci?n sobre la forma m?s adecuada de realizarla en cada caso. La segunda parte de la revisi?n, publicada con el t?tulo ?Metabolomics analysis II. Preparation of biological samples prior to detection? se dedica a la preparaci?n de la muestra propiamente dicha? A pesar de que esta etapa engloba todas las sub-etapas durante las que los analitos se adecuan para la detecci?n, el hecho de que los detectores utilizados en metabol?mica generalmente est?n conectados en l?nea con un equipo de separaci?n (cromat?grafo o, menos com?nmente, equipo de electroforesis capilar), hace que se suela considerar terminada la etapa de preparaci?n de la muestra cuando ?sta (o lo que de ella queda tras su preparaci?n, que recibe el nombre de ?muestra anal?tica?) llega al sistema de separaci?n. Se hace hincapi? en la escasa innovaci?n que se ha producido en esta etapa, que generalmente se desarrolla por m?todos estrictamente convencionales, frente a los grandes logros en separaci?n?detecci?n, y se realiza una descripci?n m?s completa de c?mo llevar a cabo la inactivaci?n enzim?tica para el an?lisis r?pido tras el muestreo. Todo ello precede a las etapas t?picas de separaci?n de los analitos de la matriz de la muestra: Extracci?n l?quido ?l?quido o s?lido?l?quido, dependiendo del estado f?sico de la muestra, son sub-etapas que se requieren generalmente previas al an?lisis. Las formas m?s adecuadas para cada estrategia en metabol?mica, funci?n tambi?n del binomio matriz? analito, se discuten cr?ticamente. El caso especial de muestreo?preparaci?n? an?lisis en cultivos celulares es considerado de forma separada, dada su naturaleza y especiales caracter?sticas. Etapas adicionales, como derivatizaci?n, se discuten tambi?n en esta revisi?n considerando las diferentes alternativas existentes para su aceleraci?n. Son de destacar los criterios anal?ticos de los autores de la revisi?n, junto con la experiencia en preparaci?n de la muestra del grupo en el que se integra la doctoranda, que permiten tener una visi?n de estas operaciones que puede ser de utilidad 504 Nuevas plataformas anal?ticas en metabol?mica manifiesta para bioqu?micos, cl?nicos u otros investigadores no familiarizados con su consideraci?n cr?tica ni con el objetivo de su mejora. Pa rte 2: An?lisis orientado (targeting analysis) Este tipo de an?lisis, como cualquier estrategia en metabol?mica, tiene sus ventajas e inconvenientes y proporciona una informaci?n limitada cuyo inter?s puede ser pobre (cuando se requiere obtener una informaci?n completa del metaboloma) o demasiado prolija (cuando s?lo se persigue una huella dactilar, sin entrar en concentraciones o identificaciones individuales). Una vez definido el problema en cuesti?n y si ?ste se refiere a unos compuestos concretos, generalmente los precursores y los metabolitos de unas determinadas rutas metab?licas, el an?lisis orientado es la v?a id?nea para el estudio. Desarrollar un m?todo basado en esta estrategia implica una ?ptima preparaci?n de la muestra, obviamente tras las adecuadas etapas de muestreo y conservaci?n de la muestra, seguida de una separaci?n individual de los analitos (exhaustiva o no, dependiendo del analizador utilizado finalmente) y un sistema de detecci?n tan sensible como requiera la concentraci?n de los analitos. Una etapa de preconcentraci?n puede resolver el problema de una detecci?n poco sensible. Por tanto, se habla del desarrollo de m?todos muy selectivos para obviar interferencias que pudiesen dificultar el an?lisis cualitativo/cuantitativo de los compuestos de inter?s, pero a su vez de elevada sensibilidad para su detecci?n incluso a bajos niveles de concentraci?n. Hasta 6 m?todos se han desarrollado utilizando esta estrategia en los que, junto con una instrumentaci?n variada que ha contribuido a la formaci?n de la doctoranda, se ha pretendido en todo momento innovar en la etapa de preparaci?n de la muestra con vistas a su aceleraci?n y automatizaci?n y en la forma que mejor se ajustara a la subsiguiente etapa de an?lisis. El desarrollo de un m?todo siempre ha de completarse con su validaci?n y aplicaci?n a muestras naturales que ponga 505 Discusi?n de los resultados de manifiesto su utilidad. Los m?todos desarrollados siempre se han aplicado a estos tipos de muestras; cl?nicas, cuando los analitos han sido los caracter?sticos de muestras de orina, sangre o incluso leche materna; o vegetales, cuando los analitos han sido t?picos de esta ?rea. Cap?tulos 3 y 4 Estr?genos y progest?genos constituyen un grupo de hormonas esteroideas con funciones esenciales en el metabolismo femenino. De los primeros, los tres mayoritarios en mujeres son el estradiol, el estriol y la estrona, que pueden ser end?genos (formados a partir del colesterol mediante una ruta biosint?tica) o ex?genos, administrados como parte de contraceptivos orales, en etapas postmenop?usicas y en el tratamiento de des?rdenes en la menstruaci?n. La progesterona es el m?s abundante de los progest?genos y un compuesto intermedio en la esteroidog?nesis de estr?genos, andr?genos y corticosteroides; mientras que la pre gnenolona es el precursor metab?lico de la progesterona y pertenece al grupo de los neurosteroides. Estas hormonas, que han sido el objeto de 2 de los art?culos publicados, se encuentran en la orina, el tipo de muestra para la que se ha desarrollado el m?todo, casi exclusivamente como metabolitos conjugados, tales como glucur?nidos, sulfatos, diglucur?nidos, bisulfatos y sulfoglucu- r?nidos. Los dos problemas m?s importantes que plantea la determinaci?n de estos metabolitos son el tiempo requerido para la hidr?lisis enzim?tica de los conjugados (entre 12 y 18 h) y las bajas concentraciones a las que se encuentran en orina, adem?s de la complejidad intr?nseca de este tipo de muestra. El problema abordado en primer lugar fue el de conseguir una adecuada sensibilidad para su determinaci?n mediante una etapa de limpieza y preconcentraci?n. Para ello se parti? de un volumen de muestra relativamente grande en estudios cl?nicos (10 mL de muestra), pero no en el caso de muestras de orina que no presentan limitaci?n. La preconcentraci?n 506 Nuevas plataformas anal?ticas en metabol?mica se realiz? en un volumen de muestra anal?tica de 150 ?L tras la etapa de extracci?n en fase s?lida, con factores de preconcentraci?n entre 60 y 72. Para esta preparaci?n de la muestra se utiliz? un sistema autom?tico de ?laboratorio en una v?lvula (lab-on-valve o LO?)? con un tratamiento paralelo de muestras sometidas y no sometidas a hidr?lisis enzim?tica para la determinaci?n de los esteroides inicialmente libres y conjugados y la de s?lo los libres, respectivamente. La optimizaci?n de la etapa de separaci?n cromatogr?fica (cromatograf?a de l?quidos, LC) y la del subsiguiente an?lisis mediante ionizaci?n por electroespray y espectrometr?a de masas de triple cuadrupolo, permiti? alcanzar l?mites de detecci?n de 1.8 pg en columna, ? 0.06 ng/mL en la muestra, con desviaciones est?ndar relativas en el intervalo 1.93 ?10.99%. El siguiente reto de este estudio fue la reducci?n del tiempo requerido para la hidr?lisis enzim?tica mediante su aceleraci?n por ultrasonidos, aspecto en el que el grupo en el que se integra la doctoranda tiene una reconocida experiencia. Un a optimizaci?n multivariante de las variables implicadas en la hidr?lisis (tanto las caracter?sticas de la sonda de ultrasonidos como las del sistema bioqu?mico) dio lugar a una disminuci?n dr?stica del tiempo requerido para esta etapa, que se complet? en 30 minutos. La aplicabilidad del m?todo, en cuyo desarrollo se utilizaron muestras de orina, se ratific? aplic?ndolo a una serie de muestras de este tipo. El tratamiento autom?tico de la muestra con m?nima intervenci?n humana y la ausencia de contaminaci?n por el desarrollo en l?nea son aspectos que soportan la reproducibilidad y utilidad del m?todo para an?lisis sistem?tico de estos metabolitos. Cap?tulo 5 Los esfingol?pidos son una familia de compuestos lip?dicos que contienen una larga cadena amino-alcohol conocida como base esfingoide y son componentes de las membranas biol?gicas. Los esfingol?pidos ejercen en 507 Discusi?n de los resultados c?lulas funciones de transmisi?n y reconocimiento de se?ales y su metabolismo implica un gran n?mero de rutas de s?nTesis y de degradaci?n, como es de esperar de la variedad de esfingol?pidos presentes en las c?lulas. Existe un gran n?mero de enfermedades asociadas a los esfingol?pidos y su metabolismo, por lo que un m?todo que permita determinarlos en un rango de concentraciones que abarque los valores extremadamente bajos a los que se encuentran en fluidos humanos es extremadamente ?til. Su baja concentraci?n y su naturaleza no polar constituyen los principales cuellos de botella para el an?lisis de estos compuestos. Los m?todos existentes han estado basados mayoritariamente en extracci?n l?quido?l?quido, evapora- ci?n del disolvente org?nico y reconstituci?n previa a su separaci?n mediante LC. El uso de la extracci?n en fase s?lida se ha circunscrito a m?todos manuales, que han requerido una gran cantidad de muestra y de reactivos y la complejidad a?adida de una etapa de derivatizaci?n. El m?todo que se recoge como Cap?tulo 5 para el an?lisis orientado de precursores de esfingol?pidos (a saber, D -esfingosina, D -eritro- dihidroesfinganina, esfingosina 1-fosfato, D -eritro-dihidroesfingosina 1- fosfato) se ha aplicado a su determinaci?n en suero y en orina y est? basado en una etapa de preparaci?n de la muestra en un sistema LOV, que consiste en una etapa de extracci?n en fase s?lida y eluci?n con un eluyente en el que se ha disuelto el reactivo derivatizante (o-ftaladialdeh?do), previa a la introducci?n de los productos de la reacci?n en un ?-LC y a la detecci?n de su fluorescencia inducida por l?ser. La excelente sensibilidad que proporciona este tipo de detecci?n se pone de manifiesto en los bajos l?mites de detecci?n que presenta el m?todo en los dos fluidos biol?gicos (en los rangos 4.2 ?10.2 y 0.56 ?1.36 ng/mL para suero y orina, respectivamente) y la automatizaci?n de la preparaci?n de la muestra lo hacen muy adecuado para an?lisis cl?nico. La validaci?n del m?todo se llev? a cabo por infusi?n directa de las muestras en un analizador de masas de triple cuadrupolo en el modo de seguimiento de reacciones m?ltiples (SRM), tanto para comparar los resultados con los del m?todo propuesto, como para confirmar la identidad de los compuestos en estudio. Es de destacar que la sensibilidad y 508 Nuevas plataformas anal?ticas en metabol?mica precisi?n del m?todo propuesto es mejor que ?o similar a, dependiendo de los compuestos? las del m?todo de validaci?n. Cap?tulo 6 En contraposici?n a los metabolitos objeto de los anteriores m?todos ?todos ellos de naturaleza lip?dica?, los del m?todo recogido como Cap?tulo 6 son compuestos hidrosolubles, ya que se trata de la vitamina B9 (?cido f?lico) y de sus metabolitos. Esta vitamina est? implicada en una amplia variedad de procesos biol?gicos, ya que act?a como cofactor enzim?tico en las reacciones de transferencia del grupo metilo. El hecho de que la deficiencia de ?cido f?lico est? relacionada con numerosas enfermedades, as? como con la s?ntesis de amino?cidos como la serina y la metionina, hace necesaria su determinaci?n y la de sus metabolitos en diferentes fluidos biol?gicos en funci?n del tipo de anomal?a org?nica en la que puedan estar implicados. El desarrollo totalmente en l?nea de este m?todo se realiz? en tres matrices biol?gicas (orina, suero y leche materna) con el fin de demostrar su versatilidad e implic? las siguientes etapas: (i) Preparaci?n de la muestra mediante un equipo comercial de extracci?n en fase s?lida (Prospekt) que trabaja a alta presi?n. Una vez retenidos los analitos en el sorbente de relleno del cartucho, el eluyente, que es la fase m?vil del sistema cromatogr?fico con el que est? conectado en l?nea, se bombea y los compuestos eluidos se conducen a la cabeza de la columna cromatogr?fica. De esta forma, todo el eluido alcanza la columna ?no s?lo el que puede ocupar el bucle de la v?lvula de inyecci?n del cromat?grafo, que en este caso no se utiliza. Se consigue de esta forma una m?xima sensibilidad y un m?nimo deterioro de los analitos, que no est?n en contacto en ning?n momento con la atm?sfera. (ii) La separaci?n cromatogr?fica se realiz? en r?gimen isocr?tico, utilizando una columna de interacci?n hidrof?lica (HILIC) y una fase m?vil con alto contenido org?nico, lo que favoreci? la etapa 509 Discusi?n de los resultados siguiente. (iii) La detecci?n, basada en ionizaci?n mediante electroespray previa a un analizador de triple cuadrupolo, se vio positivamente influenciada por las condiciones de la etapa anterior. Basado en las anteriores etapas, el m?todo permite la determinaci?n de ?cido f?lico y sus metabolitos ??cido L-glut?mico N-(p-aminobenzoilo) y su derivado de acetamida? con una excelente sensibilidad (del orden de las fracciones de picomol); lo que lo hace muy adecu ado para an?lisis cl?nico en situaciones de deficiencia de estos compuestos. Cap?tulos 7 y 8 Los m?todos que conforman estos dos cap?tulos de la Memoria tienen en com?n la naturaleza de las matrices en las que se han aplicado: Vegetales en ambos casos. El m?todo recogido en el Cap?tulo 7, si bien responde al denominador com?n de la Parte II de la Memoria (an?lisis orientado) no se engloba en el ?mbito de la metabol?mica, ya que los compuestos problema son, estrictamente hablando, productos de degradaci?n, no pertenecientes a una ruta bioqu?mica. Los analitos corresponden al tipo denominado ?subproductos de la desinfecci?n?, ?cidos haloac?ticos de los que se seleccionaron los 9 m?s significativos (?cidos monocloro y monobromo ac?ticos, dicloro y dibromo ac?ticos, tricloro y tribromo ac?ticos, clorobromo ac?tico, clorodibromo ac?tico y bromodicloro ac?tico). El objetivo fue la proposici?n de un m?todo que mejorara significativamente el de referencia de la EPA para estos compuestos, pero basado en las mismas etapas y con el mismo sistema de separaci?n individual y de detecci?n (cromatograf?a de gases y detecci?n por captura electr?nica ?GC ?ECD). Por tanto, la aceleraci?n y automatizaci?n de la etapa de preparaci?n de la muestra, extremadamente larga en el m?todo de la EPA, fue el objetivo primordial. La experiencia del grupo en el que se integra la doctoranda en el dise?o de sistemas din?micos y en el uso de energ?as auxiliares fue el 510 Nuevas plataformas anal?ticas en metabol?mica soporte para la construcci?n de un sistema din?mico en el que el lixiviante, conteniendo un reactivo derivatizante, se recirculaba a trav?s de la muestra que estaba sometida a la acci?n de una sonda de ultrasonidos. La inversi?n a tiempo programado del sentido de circulaci?n del lixiviante?derivatizante, elimin? el efecto de compacidad creciente de la muestra s?lida creado cuando el lixiviante circula en un ?nico sentido. La acci?n conjunta de los ultrasonidos y la transformaci?n in situ de los analitos lixiviados en sus derivados metilados permiti? reducir las m?s de 2 horas requeridas por el m?todo EPA para ambas etapas (lixiviaci?n y derivatizaci?n secuencial) a s?lo 10 min. La eliminaci?n justificada del est?ndar externo (utilizado en el m?todo EPA) y una exhaustiva optimizaci?n de la etapa de extracci?n l?quido?l?quido previa a la separaci?n cromatogr?fica y la detecci?n complet? un m?todo que se aplic? con ?xito a la determinaci?n de los analitos en muestras vegetales como espinacas y acelgas, a las que los compuestos de inter?s podr?an haber llegado a trav?s de agua de riego. El Cap?tulo 8 abarca un amplio an?lisis orientado , realizado en tomate, de numerosos metabolitos tales como carotenoides, provitaminas, vitaminas, fenoles y az?cares, que proporciona un perfil completo de estas familias de metabolitos. Estos compuestos, agrupados en funci?n de su naturaleza en nutrac?uticos (hidrof?licos y lipof?licos) y carbohidratos (mono y disac?ridos), se separaron por LC y GC, respectivamente, y se determinaron por espectrometr?a de masas en ambos casos (triple cuadrupolo y masas en t?ndem, respectivamente). La etapa de lixiviaci?n, acelerada por ultrasonidos, se llev? a cabo con una mezcla de tetrahidrofurano?metanol 3:1, que se evapor? a sequedad. El extracto seco se reconstituy? en el medio adecuado para cada etapa cromatogr?fica, que s?lo en el caso de los az?cares requiri? derivatizaci?n (sililaci?n) antes de la inyecci?n en el cromat?grafo. La optimizaci?n de la etapa cromatogr?fica y la de detecci?n, exhaustiva para cada grupo de compuestos, junto con la r?pida preparaci?n de la muestra, proporcionan m?todos sensibles, selectivos y precisos para la determinaci?n de estos grupos de compuestos. Cada uno de los grupos, o su conjunto, pueden utilizarse como marcadores 511 Discusi?n de los resultados para la selecci?n de variedades de tomate o del tiempo para su recogida en programas de mejora de la calidad de este producto. Pa rte 3: Perfil metabol?mico global Esta estrategia, como cada una de las que se utilizan en metabol?mica, tiene sus objetivos caracter?sticos y sus herramientas propias. Los objetivos se pueden resumir en la obtenci?n de informaci?n lo m?s completa posible sobre la naturaleza de los metabolitos existentes en la muestra, que, seg?n las caracter?sticas de ?sta y del objetivo final, se puede reducir a metabolitos de car?cter polar, apolar, etc. En los m?todos de obtenci?n del perfil metabol?mico global se pretende maximizar el n?mero de metabolitos detectados e identificados en una determinada muestra. Lo ideal ser?a conseguir el perfil completo de metabolitos presentes en la muestra, lo que es inviable en organismos considerados superiores, como caso de plantas (las estimaciones apuntan hasta 200.000 metabolitos) o humanos (7900 metabolitos seg?n la HMDB ?Human Metabolome Data Base ) debido a los rangos de concentraci?n a los que estos metabolitos est?n presentes, que pueden ser de varios ?rdenes de magnitud. Con el fin de maximizar el n?mero de metabolitos identificados en una muestra se pueden combinar diferentes estrategias de preparaci?n de muestra (extracci?n l?quido?l?quido con diferentes extractantes, extracci?n en fase s?lida con diferentes fases estacionarias, reacciones de hidr?lisis para la identificaci?n de metabolitos conjugados, etc.) o de detecci?n (LC?MS/MS con diferentes formas de ionizaci?n y con distintas modalidades cromatogr?ficas, que a su vez pueden implicar diferentes fases estacionarias; GC ?MS/MS con diferentes programas de temperatura, fases estacionarias o reacciones de derivatizaci?n; NMR, unidimensional o bidimensional, homonuclear o heteronuclear, etc.) de forma que se consigan resultados complementarios. A?n combinando diferentes m?todos 512 Nuevas plataformas anal?ticas en metabol?mica anal?ticos, en la mayor?a de las ocasiones s?lo se consigue identificar una peque?a fracci?n de metabolitos. Las herramientas quimiom?tricas y bioinform?ticas utilizadas tienen una aplicaci?n posterior a la de las plataformas anal?ticas. Las primeras de estas herramientas est?n constituidas por diferentes algoritmos de tratamiento de los datos para alinear cromatogramas y espectros de masas con los que realizar una adecuada comparaci?n entre muestras, eliminar ruido de fondo o filtrar aquellas entidades moleculares no asociadas a la muestra. Las herramientas bioinform?ticas est?n constituidas por las diferentes bases de datos con las que confrontar los resultados obtenidos para identificar metabolitos a partir de los espectros obtenidos en la aplicaci?n de la correspondiente plataforma anal?tica. Bases de datos como METLIN (Scripps Center for Metabolomics and Mass Spectrometry) o HMDB son la clave para la obtenci?n de esta informaci?n una vez deducidas las f?rmulas moleculares en el caso de la espectrometr?a de masas, que ha sido la opci?n utilizada principalmente en esta investigaci?n. Para la generaci?n de estas f?rmulas se requiere la aplicaci?n de una serie de algoritmos que conduzcan a ellas, tales como: (i ) La localizaci?n y agrupaci?n de todos los iones relacionados con la misma mol?cula (e.g. covarianza de picos con el mismo tiempo de retenci?n cromatogr?fico, la distribuci?n isot?pica y/o la presencia de aductos y d?meros) mediante el algoritmo MFE (molecular feature extraction); (ii) L a generaci?n de las f?rmulas moleculares propiamente dichas utilizando el algoritmo correspondiente (Molecular Formula Generator), una vez realizadas las correcciones oportunas en cada caso (exactitud de masa e informaci?n isot?pica ?abundancia y distribuci?n espacial en este ?ltimo caso). Una vez obtenidas las f?rmulas moleculares, la b?squeda mediante la base de datos METLIN o HMDB permite ajustar cada una de ellas con una determinada tolerancia en el valor de la masa (10 ppm como error m?ximo en la plataforma instrumental utilizada) y lleva a confirmar o no su existencia en los estudios cl?nicos o nutricionales realizados en humanos. 513 Discusi?n de los resultados La investigaci?n realizada con esta estrategia se orient? al estudio de perfiles metab?licos en fluidos de muestreo no invasivo, como es el caso de dos fluidos humanos poco estudiados: La saliva y la leche materna. Todo ello se recoge en los Cap?tulos 9 y 10, cuyo conjunto constituye la Parte III de esta Memoria. Cap?tulos 9 y 10 Teniendo en cuenta los escasos estudios realizados sobre perfiles metabol?micos en saliva y en leche materna, la investigaci?n se centr? mayoritariamente en desarrollar protocolos de preparaci?n de muestra ?ptimos para maximizar el n?mero de metabolitos detectados e identificados mediante LC?TOF/MS en esos fluidos. Con este prop?sito, y teniendo en cuenta la muy diferente naturaleza de la saliva y la leche, la optimizaci?n implic? etapas distintas. La saliva es un fluido actualmente considerado como una interesante fuente de biomarcadores que pueden asociarse a diferentes patolog?as, y que implica un muestreo simple, no invasivo y sin requerimiento de personal sanitario para su obtenci?n. Adem?s, la composici?n de la saliva puede reflejar los niveles de ciertos metabolitos en sangre, por lo que es susceptible de uso como indicador del estado fisiol?gico del individuo. La principal etapa y a la que se dedic? la parte m?s importante de la investigaci?n recogida en el Cap?tulo 9 fue la optimizaci?n de la preparaci?n de la muestra para la obtenci?n de un perfil LC?TOF/MS lo m?s completo posible. La primera parte en el establecimiento del protocolo de preparaci?n de la muestra m?s adecuado se dedic? a la optimizaci?n de la etapa de hidr?lisis (?cida o b?sica) de los metabolitos potencialmente enlazados a otras mol?culas. Los resultados de este estudio se compararon con los proporcionados por muestras no sujetas a hidr?lisis (en todos los casos el seguimiento de esta etapa se realiz? mediante LC?TOF/MS en los modos de ionizaci?n positivo y negativo). Los diagramas de Venn correspondientes 514 Nuevas plataformas anal?ticas en metabol?mica ponen de manifiesto de forma inequ?voca la necesidad de la etapa de hidr?lisis, la presencia de un mayor n?mero de entidades moleculares al trabajar en modo de ionizaci?n negativo y un mayor n?mero de metabolitos comunes al trabajar en modo positivo. La necesidad de la etapa de hidr?lisis y su duraci?n (30 min) indujo a su aceleraci?n mediante ultrasonidos, con lo que se redujo a 10 min. Adem?s, la hidr?lisis asistida po r ultrasonidos dio lugar a un n?mero de entidades moleculares mayor que el modo convencional. Este hecho tuvo una especial influencia en el modo de ionizaci?n positivo en ambos tipos de hidr?lisis, ?cida y b?sica (de nuevo comparadas con las proporcionadas por la hidr?lisis convencional mediante diagramas de Venn). La baja concentraci?n de metabolitos en saliva hizo necesaria una etapa de preconcentraci?n, con la que se consigui? un incremento de entidades moleculares (174 en el modo de ionizaci?n positivo y 84 en el negativo para la hidr?lisis ?cida y 150 y 70, respectivamente, para la hidr?lisis b?sica), y por tanto la capacidad de detecci?n. A la etapa de comparaci?n de protocolos de preparaci?n de muestra sigui? la de identificaci?n de las entidades moleculares con las bases de datos METLIN y HMDB, que no incluyen informaci?n acerca de la composici?n de la saliva. Este hecho dificult? la interpretaci?n de los resultados, ya de por si complicada por la presencia de compuestos procedentes de la ingesti?n de alimentos y por los resultantes de la reacci?n con compuestos end?genos de la saliva. Un total de 12 compuestos en el modo de ionizaci?n negativo y 91 en el positivo (tolerancia en exactitud de masa por debajo de 10 ppm) fuer on encontrados en la muestra sometida a hidr?lisis ?cida, mientras que en la sometida a hidr?lisis b?sica se encontraron 13 y 52 en los modos de ionizaci?n negativa y positiva, respectivamente. La presencia de los metabolitos identificados (az?cares, l?pidos, amino?cidos, antioxidantes y otros componentes minoritarios) se ha justificado de forma exhaustiva. 515 Discusi?n de los resultados En el estudio sobre leche materna (Cap?tulo 10), una primera etapa de desproteinizaci?n, con o sin subsiguiente etapa de centrifugaci?n, puso de manifiesto las diferencias en el n?mero y naturaleza de los metabolitos existentes dependiendo de que se aplique o no esta ?ltima etapa. Los diagramas de Venn permiten tanto visualizar claramente este comporta- miento como poner m?s claramente de manifiesto el diferente comporta- miento cuando la etapa de desproteinizaci?n se realiza en un medio ?cido (metanol conteniendo un 10% de ?cido f?rmico). Por otra parte, la existencia de compuestos polares y no polares hizo necesaria la extracci?n de la muestras con mezclas metanol?cloroformo para la separaci?n de las fracciones polar?apolar, respectivamente. El escaso n?mero de metabolitos comunes en ambas fracciones ratifica la necesidad de esta etapa de separaci?n. La combinaci?n de estas estrategias de preparaci?n de muestra permiti? incrementar el n?mero de metabolitos identificados. La obtenci?n de los espectros de masas en los modos de ionizaci?n positivo y negativo para cada fracci?n condujo a la aplicaci?n de los algoritmos correspondientes para conocer las entidades moleculares y, finalmente, a la b?squeda en las bases de datos METLIN y HMBD para la adjudicaci?n a un determinado compuesto y la confirmaci?n o no de su existencia en humanos, respectivamente. Todo este tratamiento dio como resultado total la identificaci?n de 29 metabolitos en la s fracciones polar y apolar estudiadas en modo de ionizaci?n positivo, y a la de 33 en el modo de ionizaci?n negativo. El perfil realizado mediante 1H -RMN unidimensional permiti? obtener una huella digital que corrobor? parcialmente los resultados obtenidos mediante LC?TOF/MS con una m?nima preparaci?n de la muestra. Este estudio ha puesto de manifiesto por primera vez la importancia del pH, la centrifugaci?n y la extracci?n l?quido ?l?quido en la preparaci?n de este tipo de muestra. Finalmente, un cambio del medio en el que se encuentra la muestra anal?tica tras la preparaci?n a la fase m?vil 516 Nuevas plataformas anal?ticas en metabol?mica cromatogr?fica permite conseguir un incremento de la sensibilidad en la etapa de ionizaci?n por electroespray de hasta 3 veces. Caracter?sticas destacables de estas muestras son la necesidad de una etapa de desproteinizaci?n, el amplio rango de concentraci?n de los analitos y el variable n?mero de compuestos susceptibles de solubilizaci?n en medio ?cido por la presencia de grupos ionizables. Por tanto, con este estudio global se ha propuesto una estrategia con la que acometer el an?lisis metabol?mico global de un fluido biol?gico de gran inter?s, y que puede ser de utilidad con fines de evaluaci?n nutricional. Pa rte 4: Huella dactilar metabol?mica La ?ltima de las estrategias utilizadas en metabol?mica tuvo como objetivo obtener perfiles representativos de un grupo de muestras (huella dactilar metabol?mica) que permitieran su discriminaci?n frente a otros grupos de muestras, definidos o no. Las t?cnicas anal?ticas utilizadas con mayor frecuencia en esta disciplina son las espectrom?tricas, que generan un espectro representativo de la composici?n de la muestra. En concreto, NMR, MS mediante inyecci?n directa o previa separaci?n cromatogr?fica o electrofor?tica y, en menor medida, FT-IR son las t?cnicas preferentemente utilizadas en metodolog?as destinadas a comparar huellas dactilares metabol?micas. En la investigaci?n recogida en este bloque de la Tesis se han utilizado las tres t?cnicas mencionadas: FT -IR en la regi?n del infrarrojo cercano (NIRS), NMR en r?gimen unidimensional y LC?TOF/MS a partir de una matriz de datos compuesta por un perfil de metabolitos. Normalmente, los m?todos de an?lisis de la huella dactilar metabol?mica suelen utilizar, siempre que es posible, protocolos directos con el fin de simplificar el an?lisis de un gran n?mero de muestras y 517 Discusi?n de los resultados minimizar la alteraci?n de su composici?n. Este aspecto suele ser, en la mayor?a de los casos, perfectamente viable en FT-IR y NMR. Sin embargo, el an?lisis directo mediante espectrometr?a de masas es poco frecuente, excepto en muestras l?quidas de naturaleza principalmente acuosa y con una moderada concentraci?n salina. En este bloque de la Tesis, la muestra seleccionada fue orina humana, que se analiz? de forma directa en el caso de NIR y se diluy? como etapa previa en el caso de NMR y LC?TOF/MS debido a la peque?a concentraci?n de prote?nas (muestras con concentraci?n significativa de prote?nas requieren una etapa de desproteinizaci?n). La estrategia de huella dactilar metabol?mica carece de sentido sin la implementaci?n de t?cnicas de an?lisis multivariante que permitan reducir la dimensionalidad de las matrices de datos que usualmente se manejan (n muestras ? m variables). Estas t?cnicas tienen como objeti vo detectar la presencia de grupos o clases en la poblaci?n en estudio en base a alg?n criterio que no se aporta para el desarrollo del modelo (an?lisis no supervisado) o se aporta con el objetivo de desarrollar modelos con capacidad de predicci?n (an?lisis supervisado). En este bloque de la Tesis se ha aplicado como estrategia general la combinaci?n de t?cnicas de an?lisis no supervisado y de an?lisis supervisado con los siguientes objetivos: (i) Detectar la presencia de agrupamientos de muestras de acuerdo con el objetivo del estudio o por causas externas al mismo (caracter?sticas antropol?gicas de la poblaci?n) mediante an?lisis por componentes principales (PCA); (ii) Desarrollar modelos con capacidad de predicci?n de una o varias variables respuesta (informaci?n aportada al modelo) mediante an?lisis por m?nimos cuadrados parciales utilizando diferentes algoritmos. Una vez terminado el an?lisis, el objetivo fue en todos los estudios la b?squeda y posterior identificaci?n de los metabolitos o familias de metabolitos que permiten explicar la variabilidad observada en cada uno de los estudios. As?, diferentes algoritmos estad?sticos como ANOVA o de estimaci?n de cambio relativo (en ingl?s fold change analysis) han permitido la identificaci?n de las variables (metabolitos) que contribuyen en mayor medida a explicar la variabilidad observada. 518 Nuevas plataformas anal?ticas en metabol?mica El esquema de trabajo descrito se ha aplicado en nutrimetabol?mica para evaluar el efecto de la ingesta por individuos obesos de desayunos preparados con diferentes aceites vegetales sometidos a fritura controlada. Para ello, se tomaron muestras de orina en condiciones basales y 2 y 4 h despu?s de la ingesta de cada uno de los desayunos. Con el objetivo de aumentar el inter?s estad?stico de este estudio, los pacientes ingirieron cada desayuno despu?s de un per?odo de 2 semanas con el fin de normalizar las condiciones basales. Los aceites utilizados fueron: (i) Oliva virgen extra con un contenido natural de antioxidantes fen?licos de 400 ? g/mL, expresado en contenido de ?cido g?lico seg?n el m?todo de Folin-Ciocalteu (VOO); (ii) aceite de girasol comercial con un contenido nulo de antioxidantes fen?licos debido al proceso de obtenci?n ?aceite refinado? (NSO); (iii) A ceite de girasol alto oleico refinado enriquecido con un inhibidor de la oxidaci?n sint?tico como es el dimetilsiloxano a una concentraci?n de 400 ? g/mL (DSO); y, finalmente, (iv) aceite de girasol alto oleico refinado enriquecido con un extracto rico en fenoles procedente de alperujo hasta una concentraci?n fen?lica de 400 ? g/ml, de acuerdo con el m?todo de Folin- Ciocalteu (PSO). Tres estudios utilizando las tres t?cnicas de detecci?n previamente indicadas componen los Cap?tulos 11?13. El orden en el que se exponen es el de nivel de informaci?n creciente que proporcionan. Cap?tulo 11 Este primer cap?tulo describe la utilizaci?n de la t?cnica NIRS que es, de las tres utilizadas, la que menos se ha aplicado en estudios de huella dactilar metabol?mica. De hecho, se ha utilizado ?nicamente con prop?sitos de clasificaci?n y predicci?n puesto que su capacidad de identificaci?n de metabolitos es escasa. Entre sus ventajas destaca la rapidez de medida, la posibilidad de an?lisis directo, incluso en el caso de muestras s?lidas (tejidos), facilidad de uso del instrumento y reducido coste de adquisici?n y 519 Discusi?n de los resultados mantenimiento del equipo. Esta primera investigaci?n en huella dactilar metabol?mica se bas? en la utilizaci?n de una t?cnica de modelado de clases en an?lisis supervisado como es la PLS-CM (del ingl?s Partial Least Squares- Class Modeling) con el que se desarrollaron cuatro modelos, uno por cada desayuno ingerido, con el fin de predecir su ingesta frente al resto. Frente a las t?cnicas PLS basadas en an?lisis discriminante (PLS-DA), la t?cnica de PLS-CM permite interpretar las propiedades espectrales de las muestras que componen cada clase y estimar los par?metros de sensibilidad (capacidad de un modelo para predecir individuos pertenecientes a la clase modelada) y especificidad (capacidad de un modelo para predecir la no pertenencia de un individuo a la clase modelada) para las etapas de desarrollo y validaci?n cruzada. Ambos par?metros son de enorme utilidad puesto que dan idea de la robustez de cada uno de los modelos. Otra novedad de esta investigaci?n es el pretratamiento de datos realizado. As?, los espectros NIR no se utilizaron tal cual, sino que los obtenidos en condiciones basales para cada individuo fueron substra?dos de los obtenidos en condiciones post-basales. El objetivo de este tratamiento fue minimizar la variabilidad entre individuos para evitar el efecto de enmascaramiento de la variabilidad asociada a la ingesta de los desayunos. La aplicaci?n del algoritmo VIPs (Variables Important in Projection) permiti? identificar aquellas regiones espectrales de mayor peso estad?stico para explicar la variabilidad observada en cada modelado de clases. Estas regiones espectrales coincidieron para todos los desayunos administrados, pero se detectaron diferencias en el nivel de significado de cada una de ellas. Cap?tulo 12 La investigaci?n que compone el segundo cap?tulo de esta partetuvo como objetivo evaluar la utilizaci?n de la NMR en el estudio anteriormente descrito. En este caso las muestras se diluyeron con un tamp?n para ajustar el pH y con agua deuterada para la estabilizaci?n del campo magn?tico. El estudio se inici? con la aplicaci?n de una t?cnica de an?lisis no supervisado, 520 Nuevas plataformas anal?ticas en metabol?mica PCA, que permiti? discriminar las muestras tomadas 4 h despu?s de la ingesta de cada desayuno frente a las muestras control. Una vez detect ado el mayor cambio metab?lico ocasionado por la ingesta de cada desayuno, la siguiente etapa fue el desarrollo de modelos de discriminaci?n entre pares de clases (muestras tomadas a las 4 h despu?s de la ingesta de cada desayuno) basados en PLS-DA. Con este an?lisis se consiguieron capacidades de predicci?n entre 84 y 100% para los diferentes estudios comparando clases dos a dos. Una ventaja frente al estudio anterior fue que las variables con mayor significado estad?stico para explicar la variabilidad observada en cada estudio se correspond?an con se?ales de desplazamiento qu?mico, que pueden permitir la identificaci?n de metabolitos mediante bases de datos adecuadas (en este caso Chenomx). Con estas consideraciones fue posible la identificaci?n de metabolitos tales como ?cidos glut?mico y c?trico, l?pidos y compuestos con grupos amino, tales como niacinamida, metilhistidina, creatina o creatinina como los principales responsables de las discriminacio- nes estudiadas. Cap?tulo 13 El ?ltimo de los cap?tulos de este bloque contiene la informaci?n sobre el estudio realizado utilizando LC?TOF/MS de alta resoluci?n. En este caso tambi?n se ha utilizado una estrategia combinada de an?lisis no supervisado mediante PCA y supervisado mediante PLS-DA. Previamente, la matriz de datos se simplific? mediante la aplicaci?n de un filtro por frecuencia al 50% (eliminaci?n de aquellas entidades moleculares no presentes en al menos el 50% de las muestras pertenecientes a cada clase) y un an?lisis de cambio relativo (fold change analysis). El an?lisis mediante PCA permiti? discriminar de forma clara los perfiles metab?licos pertenecientes a individuos 4 h despu?s de la ingesta de forma independiente de cada desayuno preparado con los 4 aceites utilizados. Por 521 Discusi?n de los resultados otro lado, el an?lisis mediante PLS-DA permiti? discriminar entre los perfiles de metabolitos detectados en el suero extra?do 4 h despu?s de la ingesta de los desayunos preparados con los aceites fritos. Para ello, se compararon las clases de tres en tres. As?, en el primer estudio se compararon los perfiles metab?licos proporcionados por la ingesta de VOO, NSO y PSO consigui?ndose una capacidad de predicci?n por encima del 69% en validaci?n cruzada, mientras que en el desarrollo del modelo fue del 100% para todos los aceites. En el segundo estudio, dise?ado para la comparaci?n de los perfiles metab?licos obtenidos tras la ingesta de VOO, NSO y DSO, se consiguieron capacidades de predicci?n mediante validaci?n cruzada del 92, 73 y 50%, respectivamente, mientras que la predicci?n en el desarrollo del modelo fue del 100%. Por tanto, se puede concluir que la capacidad de predicci?n para cada uno de los desayunos fue buena (a excepci?n del desayuno preparado con DSO, que present? una baja capacidad de predicci?n en validaci?n cruzada) teniendo en cuenta que se trata de un estudio biol?gico sobre nutrici?n. La alta resoluci?n que se consigue con la t?cnica LC?TOF/MS incrementa la capacidad de identificaci?n de los metabolitos de mayor influencia en la variabilidad observada en cada uno de los estudios mediante PLS-DA. As?, entre los metabolitos identificados se pueden citar az?cares, hormonas, l?pidos, amino?cidos y otros compuestos minoritarios. El potencial de identificaci?n de los metabolitos con mayor peso estad?stico es pues la principal ventaja de la plataforma utilizada en esta investigaci?n. D I S C U S S I O N OF THE RESU LT S 525 Discussion of the results The present regulation of the University of C?rdoba that deals with the writing of the Book of a Doctoral Thesis, in the modality in which the articles (published or close to publication) are included themselves, indicates that the report must include a section of discussion of the results. This regulation can be accomplished in a different extent, depending on the homogeneity of the developed research. Thus, as research in this Thesis has metabolomics as a common denominator, sections in this book are devoted to the development of analytical platforms based on the different strategies characteristic of this discipline. Therefore, the discussion of the results encompasses, a separated discussion of the developed platforms based on each strategy. An independent consideration deserves the publications on reviews devoted to operations previous to metabolomics analysis. PART 1: Sa mple preparation in metabolomics During the first years of the Thesis the PhD student carried out an exhaustive bibliographic research in metabolomics, which led to alert about the poor attention paid to the analytical steps prior to detection. This fact was clearly highlighted in the meeting organized in 2005 by the Metabolomics Society (www.metabolomicssociety.org) that created the ?Metabolomics Standard Initiative?, focused on standardi?ation of the different steps and processes necessary to obtain information from metabolomics experiments. This initiative should allow transferability of analytical methods, thus making possible to exactly reproduce analytical conditions in any laboratory. This led to an in-depth study on these steps that, despite of being of enormous importance in the final results of the metabolomics experiment, have not received the deserved attention. Therefore, the purposes of Part 1 of this book was to develop a critical compilation of the reported methods in metabolomics, aimed at making a 526 Nuevas plataformas anal?ticas en metabol?mica better use of the existing information, together with a call for attention on the deficiencies and lags in this area. The extent of the overall review led to two publications in the same volume of TrAC (Trends in Analytical Chemistry) journal, which, with the highest impact index in Analytical Chemistry, leaded the Journal Citation Report (from the ISI Web of Knowledge). In the first part of the review, entitled ?Metabolomics analysis I? Selection of biological samples and practical aspects preceding sample preparation?, the characteristics of metabolomics analysis as compared to (and in relation with) other less recent omics are described. This also highlights the scant interest paid to the first steps of the analytical process in metabolomics, as demonstrated by the large number of publications on analytical instrumentation as compared to the few publications on sample preparation, despite of being a key aspect to obtain minimum errors or bias from the instruments. It is also worth describing the different strategies used in metabolomics, as sample preparation differs for each strategy; the review also includes a scheme where the reader can localize every step to be developed. These aspects led to a discussion on the selection of the most interesting biological material for sampling (mainly blood or urine, but also cells and tissues, in addition to other fluids such as breath, synovial fluid, bile, amniotic fluid, saliva, tears, breast milk or seminal plasma) and the way to quench the enzymatic activity in the sample to achieve representativeness in sampling. In addition, the most suitable sample storage protocol for each biological material depending on its characteristics have also been discussed, as it is the key to preserve sample integrity and avoid metabolites degradation, which mainly occurs because of repeated freezing/thaw steps. Consequences of inadequate storage protocols, as well as some advice about this step are emphasized in the review. The second part of the review, published under the title ?Metabolomics analysis II? ?reparation of biological samples prior to detection?, is devoted to sample preparation itself? ?espite this step involves 527 Discussion of the results all sub-steps to prepare analytes for detection, the fact that most of the detectors used in metabolomics are on-line connected to separation equipment (viz. a chromatograph or, less commonly, capillary electro- phoresis equipment) leads to consider the end of sample preparation when the sample (or, more exactly, that remaining after preparation, with is called as ?analytical sample?) reaches the separation system? The scarce innovation introduced in sample preparation so far is emphasized, as this step is usually developed by conventional approaches. This fact makes more noticeable the great achievements on separation?detection equipment . Enzymatic quenching for a fast analysis after sampling is described in detail. All this aspects precede the typical steps for analytes extraction from the sample matrix, usually carried out by liquid?liquid extraction or solid? liquid extraction, depending on the physical state of the sample. The most appropriate treatment for each metabolomics strategy, as a function of the matrix?analyte binomial, are critically discussed. The special case of sampling?sample preparation?analysis on cell cultures is separately considered, as required by its nature and special characteristics. Additional steps, such as derivatization, are also discussed in this review. It is worth noting the analytical criteria of the authors of this review, together with the experience on sample preparation of the team in which the PhD student is integrated. Both aspects allow a vision of these operations that can be of clear utility for biochemists, clinicians or other researchers who are not familiar with the critical study or the improvement of sample preparation in metabolomics. Part 2: Targeting ana lysis This type of analysis, similarly to other strategies in metabolomics, has advantages and disadvanges, as it provides information limited to a group of compounds, which can be poor (when the complete information on 528 Nuevas plataformas anal?ticas en metabol?mica the metabolome is required) or too prolix (when a fingerprinting is the aim, with no interest in neither individual identification or quantitation). After defining the concrete analytical problem, targeting analysis is the best approach if quantification of a limited set of compounds (usually the precursors and metabolites of some pathways) is the aim. Development of a method based on this strategy involves an appropriate sample preparation (which necessarily consider optimization of sampling and storage), followed by an individual separation of the analytes (which can be exhaustive or not, depending on the analyzer finally used) and a detection system as sensitive as required by the concentration of the analytes. However, a preconcentration step can solve the problem caused by low-sensitive detection. Therefore, targeted analysis refers to methods enough selective to circumvet the potential interferents that can difficult qualitative? quantitative analysis of the target compounds, but also to endow the methods with high sensitivity to detect very low concentrations. A total of 6 methods have been developed based on this strategy using an assorted instrumentation, which has contributed to the formation of the PhD student. The aim in all cases has been to innovate in acceleration and automatization of sample preparation in order to improve the resulting characteristics of the overall method. After optimizing the main variables, a final application of the resulting method to natural samples is mandatory to demonstrate its usefulness. The methods developed in this Thesis have always been applied to either vegetables or clinical samples (in the latter case when the analytes have been those characteristic of urine, blood or even breast milk). Chapters 3 and 4 Estrogens and progestogens are a group of steroid sex hormones with essential functions in female metabolism. Among estrogens, the most abundant in women are estradiol, estriol and estrone, which can be endogenous (formed from colesterol through a biosynthetic pathway) or exogenous, frequently administrated as a part of oral contraceptives, in 529 Discussion of the results estrogen replacement therapy of post-menopausal women, and in the treatment of menstrual disorders. Progesterone is the most abundant progestogen and also an intermediate compound in the steroidogenesis of estrogens, androgens and corticosteroids; while pregnenolone, the metabolic precursor of progesterone, belongs to the group of neurosteroids. These hormones are the object of two of the published articles, which are present in urine (the type of sample for which the method has been developed) almost exclusively as conjugated metabolites such as glucuronides, sulfates, diglucuronides, bisulfates and sulfoglucoronides. The two more significant problems in the determination of these metabolites are the time required for enzymatic hydrolysis of the conjugates (between 12 and 18 h) and the low concentrations at which they are in urine, in addition to the intrinsic complexity of this type of sample. The problem to be firstly overcome was to achieve the required sensitivity for their determination by using a cleanup/preconcentration system. Thus, solid-phase extraction allowed reducing sample volume from 10 mL (a relatively high sample volume in clinical studies) to 150 ?L of the analytical sample, with preconcentration factors ranging from 60 to 72. A ?laboratory on a valve (lab-on-valve or LO?)? automated system was used for in parallel treatment of samples, which could be previously subjected to enzymatic hydrolysis, for determination of total steroids present in the sample (free and conjugated), or directly injected into the LOV system, to determine the free fraction. Optimization of the chromatographic step (liquid chromatography, LC) and subsequent analysis by electrospray ionization and triple-quad mass spectrometry, allowed reaching detection limits of 1.8 pg on column (0.06 ng/mL in sample), with relative standard deviations within 1.93 ?10.99%. The next challenge in this study was to shorten the required time for enzymatic hydrolysis by using ultrasound energy, taking advantage of the experience of the research group. Multivariate optimization of the variables involved in each step (both those characteristic of the ultrasound probe and 530 Nuevas plataformas anal?ticas en metabol?mica those of the biochemical system), led to a dramatic decrease of the time for this step, which was complete after 30 min. Automation of all steps for sample preparation support the reproducibility of the method and its usefulness for routine analysis of these metabolites. Chapter 5 Sphingolipids, a family of lipids endowed with a long amino-alcohol chain, known as sphingoid base, are components of biological membranes. Sphingolipids exert different functions in signal transmission and cell recognition, and their metabolism involves a number of synThesis and degradation pathways, as expected from the variety of sphingolipids present in cells. There are several diseases associated to sphingolipids and their metabolism; therefore, it would be of great interest to develop a sensitive method that allows their determination in human biofluids, as they are present at extremely low concentrations. Together with these low concentrations, their non-polar nature are the main bottlenecks for their analysis. The existing methods have been based mainly on liquid?liquid extraction, evaporation of the organic solvent and reconstitution prior to separation by LC. The use of phase-solid extraction has been restricted so far to manual methods that have required a high amount of both sample and reagents, and the added complexity of a derivatization step. The method in Chapter 5 for the targeted analysis of spingolipid precursors (namely, D-sphingosine, D-erytro-diydosphingosine, sphingosine 1-phosphate, and D-erytro-dihydrosphingosine 1-phosphate) has been applied to their determination in serum and urine. It is based on a sample preparation step, using an LOV system, which consists of solid-phase extraction and elution with the derivatizing reagent (o-phthaldialdehyde) dissolved in the eluent, prior to the chromatographic separation of the ?-LC and detection by laser-induced fluorescence. This type of detection provides excellent sensitivity, as demonstrated by the low limits of detection achieved 531 Discussion of the results in the two biological fluids (within the ranges 4.2 ?10.2 and 0.56 ?1.36 ng/mL for serum and urine, respectively); the sample preparation was fully automated, which makes the method very appropriate for clinical analysis. Validation of the method was carried out by direct infusion of samples into a triple-quad mass-detector in mode of multiple reaction monitoring (MRM), aiming at comparing the results with those provided by the proposed method and confirming the identity of the target compounds. It is worth emphasizing that both sensitivity and precision of the proposed method is better than ?or similar to, depending on the compounds? those of the validation method. Chapter 6 By contrast to the previously discussed metabolites ?all them of lipidic nature? Chapter 6 deals with the analysis of a family of hydrophilic compounds: vitamin B9 (folic acid) and its metabolites. This vitamin is involved in a wide variety of biological processes, as long as it acts as enzymatic cofactor in the transference of methyl-group reactions. Due to the fact that a deficiency of folic acid is related with a number of diseases, and that it is involved in the synThesis of amino acids such as serine and methionine, quantification of this vitamin and its related catabolites in different biological fluids may allow to determine anomalies in their levels as a consequence of a disease or abnormal function. With this aim, a method for quantification of these target metabolites in three different biological matrices (urine, serum and breast milk) was proposed, demonstratintg its versatility. It involved three steps: (i) Sample preparation by a commercial solid-phase extraction (SPE) system (Prospekt) working at high pressure. Once the analytes are retained in the sorbent cartridge, the eluent, which is the mobile phase of the chromatograph on-line connected to the Prospekt system, is pumped and the eluted compounds are led to the chromatographic column. Maximum 532 Nuevas plataformas anal?ticas en metabol?mica sensitivity is thus obtained and minimum analytes degradation, which are never in contact with the atmosphere. (ii) The chromatographic separation is carried out by an isocratic regime with high content of organic phase, thanks to the use a hydrophilic interaction column (HILIC); this favoured the following step. (iii) Detection, based on electrospray ionization prior to a triple-quad mass analyzer, was positively influenced by the conditions in step ii. Thus, the resulting method allows the determination of folic acid and its metabolites ?N(p-aminobenzoyl)-L-glutamic acid and its acetamide derivative? with an excellent sensitivity (in the order of picomol fractions); thus making this method very appropriate for clinical analysis in cases of deficiency of these compounds. Chapters 7 and 8 The methods that constitute these chapters of the book have been applied to vegetable matrices. By contrar y to the previous studies, these are solid matrices, making necessary to lixiviate analytes in a quantitative and reproducible manner. Despite it is under the common denominator of Part II of the book (targeted analysis), method developed in Chapter 7 is not properly within the field of metabolomics, as the target compounds are not metabolites but degradation products formed from disinfection processes, so they do not belong to an endogenous biochemical pathway, but to the category of ?disinfection by-products?? Among these compounds, ? of the most significant haloacetic acids were selected (monochloro- and monobromo acetic acids, dichloro- and dibromo acetic acids, trichloro- and tribromo acetic acids; chrorobromo acetic - chlorodibromo acetic- and bromodichloro acetic acids). The objective was to develop a method adapted from the EPA reference method for drinking water to vegetable matrices that can be contamined by the effect of watering The method is based on the same steps and with the same individual separation and detection systems (gas 533 Discussion of the results chromatography and electron capture detection, GC ?ECD) but with acceleration and automation of the sample preparation step ?very long in the EPA method? as the basic objective. The experience of the research group in which the PhD student is integrated in the design of dynamic systems and in the use of auxiliary energies was the support to construct a dynamic system in which the lixiviant, containing the derivatizing reagent, was recirculated through the sample, which was subjected to the action of an ultrasound probe. The change of direction at preset intervals of the lixiviating?derivatizing mixture avoided increased compactness of the solid sample, which occurs when the lixiviant circulates in a single direction. The joint effect of ultrasound and in situ derivatization of the lixiviated analytes into their methylated derivatives allowed to shorten the step from more than 2 h (required in the EPA method for extraction and subsequent derivatization) to only 10 min. An ex haustive optimization of the liquid? liquid step prior to chromatographic separation and detection completed the method, which was successfully applied to the determination of the target analytes in vegetables such as spinach and chard. Chapter 8 is devote d to a wide targeted analysis in tomato encompassing a number of metabolites such as carotenoids, provitamins, vitamins, phenols and sugars providing a complete profile of the target families. The compounds, grouped as a function of their nature into nutraceuticals (both hydrophilic and lipophilic) and carbohydrates (mono and disaccharides), were separated by LC and GC, respectively, and determined by mass spectrometry (triple-quad and ion-trap tandem-mass, respectively). The lixiviation step, accelerated by ultrasound energy, was carried out with a 3:1 tetrahydrofuran ?methanol mixture, then evaporated to dryness. The dry extract was reconstituted in the appropriate medium for each chromatographic steps, which required derivatization (silylation) only in the case of sugars prior to injection into the gas chromatograph. Optimization of the chromatographic and detection steps, exhaustive for each group of compounds, together with the fast sample preparation, provide sensitive, selective, and precise methods for determination of these 534 Nuevas plataformas anal?ticas en metabol?mica groups of compounds, which can be used as markers to select either tomat varieties or collection time in cultivar improvements. Pa rt 3: Global metabolic profiling This strategy differs from others in metabolomics in its scope and analytical tools used. Thus, the scope of global metabolic profiling is to obtain the maximum information about metabolites existing in the biological system under study, which may require different approaches depending on their characteristics, i.e., if they are polar or non-polar and their concentration ranges. Global profiling methods seek to maximize the metabolite coverage, which is the number of detected metabolites in a given sample. Ideally, a complete metabolic profile should be obtained, which is unviable in the case of complex organisms such as plants (with 200000 estimated metabolites) or humans (with 7900 metabolites indexed in the HMDB) due to the wide concentration ranges at which metabolites are present, usually within several orders of magnitude. With the aim of maximizing the number of identified metabolites in a given sample, it is possible to combine several sample preparation strategies (e.g. liquid?liquid extraction with different solvents, solid-phase extraction with different stationary phases, hydrolysis reactions for identification of conjugated metabolites) and/or detection techniques (liquid chromatography?mass spectrometry in simple (MS) or multiple (MS/MS) mode, by different ionization modes or different chromatographic separations; for instance, by using different stationary phases? gas chromatography?mass spectrometry (GC MS or ?C?MS/MS) with different temperature gradients, stationary phases or derivatization protocols; nuclear magnetic resonance (NMR) mono- or bidimensional, homo- or heteronuclear, etc., so as to achieve complementary results. Even after combining different analytical methods, it 535 Discussion of the results is only possible identify a small fraction of the overall components in most cases. In this research, chemometrics and bioinformatics have eventually been used to extract information after the application of the analytical platforms. In the former, different algorithms for data pretreatment are used to align chromatograms and MS or NMR spectra for an adequate comparison between samples, eliminate background noise or filter molecular entities non-associated with sample composition. With respect to the bioinformatic tools used in this research, these are constituted by a variety of metabolite databases to query results for a final identification. Therefore, the use of databases such as METLIN (Scripts Center for Metabolomics and Mass Spectrometry) or HMDB (Human Metabolome Database) are essential to obtain accurate identification. In the case of MS, which is the main analytical platform used in this part of the research, a first step in identification involves elucidation of molecular formulas from composite spectra. Thus, different algorithm are applied for generating these formulas, entailing: (i) location and grouping of all the ions related to a given molecule (e.g. peak covariance for all ions with the same retention time, the charge envelope and/or the presence of dimmers and adducts ions) by means of the molecular features extraction algorithm (MFE), and (ii) generation of molecular formulas themselves by the Molecular Formula Generator algorithm (MFG) after correcting mass accuracy, isotopic dist ribution and abundance. Afterwards, database search allows adjusting every feature with a mass abundance window (in this research, below 10 ppm of difference between experimental and theoretical masses) and confirming biological occurrence of compounds in the examined biofluids, for nutritional or clinical applications. Following this scheme, research carried out by this strategy was focused on obtaining metabolic profiles of biofluids with non- invasive sampling, such as saliva and milk, which are fluids still unexploited in metabolomics. All these aspects are described in Chapters 9 and 10, constituting part 3 of this book. 536 Nuevas plataformas anal?ticas en metabol?mica Chapters 9 and 10 Bearing in mind the scarce reports dealing with metabolic profiling of saliva and human breast milk, research was focused on the development of optimum protocols for sample preparation to maximize the number of metabolites detected and identified by LC?TOF/MS? ?ith this aim, and considering the very different nature of saliva and milk, optimization involved different steps. Saliva is a biofluid that is gaining importance as an interesting source of biomarkers without needing healthcare personnel for sampling. Moreover, saliva composition may reflect values of several metabolites in blood, which makes this biofluid suitable for indicating the physiological state of an individual. The main contribution of the research enclosed in Chapter 9 was the optimization of sample preparation to obtain a complete LC?TOF/MS metabolic profile? Thus, the protocol started with a hydrolysis (acidic or basic) step to release metabolites that may be conjugate to other molecules (such as proteins). The results were compared to other protocols without en?ymatic hydrolysis, in every case followed by LC?TOF/MS detection in both positive and negative modes. Venn diagrams shows unequivocally the need of a hydrolysis step, which was accelerated (as it lasted 30 min) by ultrasound, and reduced to 10 min. Furthermore, US - assisted hydrolysis led to an increase of the molecular features with respect to the non-hydrolyzed sample, especially for positive ionization mode. Afterwards, a preconcentration step was carried out to increase the number of molecular features (174 in positive mode and 84 in negative mode for acidic hydrolysis and 150 and 70, respectively, for basic hydrolysis). Once sample preparation protocols were optimized, identification of the extracted molecular features was carried out by comparing with the METLIN and HMDB, being worth noting the absence of metabolites indexed for saliva. This fact, as well as the presence of exogenous compounds from foods intake, complicated identification. 537 Discussion of the results Conclusively, a total of 12 compounds in negative ionization and 91 in positive (with a mass tolerance below 10 ppm) were identified from acidic hydrolysis, while 13 and 52 were found from basic hydrolysis, respectively. The detected metabolites, including sugars, lipids, amino acids, antioxidants and other minor components, were exhaustively justified and contrasted. On the other hand, Chapter 10 focuses on global profiling of maternal milk. In this case, sample preparation started with a deproteinization, with or without subsequent centrifugation, with considerable differences between them. Venn diagrams clearly show these differences in the number and nature of detected metabolites depending on the inclusion of a centrifugation step, although showing a different pattern when deproteinization was carried out in acidic medium (10% formic acid in methanol). On the other hand, the presence of both polar and non-polar fractions involved liquid?liquid extraction of samples with methanol? chloroform, for polar and non-polar extraction, respectively. The presence of common metabolites in both phases justified the use of liquid?liquid extraction, and combination of these strategies allowed increasing the number of detected metabolites. Finally, preconcentration by evaporation and reconstitution in an organic solvent increased three times the number of detected metabolites. After applying the extraction algorithm, mass spectra from both positive and negative ionization modes provided molecular features. Finally, to search in the METLIN and HMDB for identification and confirmation of the presence in humans was carried out. This led to a total of 29 metabolites in polar and non-polar fraction in positive and 33 in negative modes. The metabolic profile obtained from monodimensional 1H -NMR allowed to obtain a fingerprint that complemented results from LC?TOF/MS, in this case with minimum sample preparation. 538 Nuevas plataformas anal?ticas en metabol?mica This study highlights for the first time the importance of pH, centrifugation and liquid?liquid extraction in sample preparation of this type of samples. Significant characteristics of this research was the need for deproteinizing samples, the wide range of concentration of metabolites and the variable number of compounds likely to be solved in acidic medium by the presence of ionisable groups. Therefore, the global profiling approach offers the possibility to analyze a global profile of a biological fluid that can be interesting from a clinical or nutritional point of view. Pa rt 4: Metabolomics fingerprinting The last strategy used in metabolomics aims at obtaining representative profiles or fingerprints from a group of samples that allow discriminating against other defined or non-defined classes. Analytical platforms frequently used in this discipline are mainly spectrometrics, as they generate a spectrum representative of sample composition. Concretely, NMR, MS by direct infusion or with short chromatographic or electrophoretic separation, and, less frequently, FT-IR are the leading techniques for fingerprinting. The research developed in this part of the Thesis was based on these three analytical platforms; FT -IR in near-infrared region (NIRS), mono-dimensional ?MR and LC?TOF/MS? Metabolomics fingerprinting methods usually entail direct analysis methods, as far as it is possible, to simplify the analysis of a great number of samples for representative results and minimize alterations in sample composition. This is quite usual in NMR and NIRS, whereas direct analysis by MS is not frequent, except for aqueous samples with moderate salts content. Therefore, in this last part of the book, sample selected for fingerprinting analysis was urine, involving direct analysis for NIRS, and 539 Discussion of the results dilution for NMR and LC?TOF/MS? Thus, deproteini?ation was not necessary due to the very low proteins content in urine. Metabolomics fingerprinting cannot be carried out without employing multivariate analysis techniques that allow reducing dimensionality of the raw data usually obtained, which forms matrices of n samples x m variables. These techniques have in common the detection of classes or groups within the population following a certain criterium, which may be included (supervised) or not (non-supervised) during the model development. In this part of the Thesis both supervised and non-supervised strategies have been combined with the following aims: (i) to detect the presence of clustering according to the aim of the study or by extrinsic causes (for instance, anthropologic characteristics) by principal component analysis(PCA); (ii) to build prediction models with one or more response variables (information that has to be included in the model) by partial least squares (PLS) by different algorithms. Afterwards, the aim was to find metabolites or families of metabolites that enabled to explain the observed variability in each model. Thus, different tools such as ANOVA or Fold Change analysis allowed the identification discriminating metabolites with higher influence in the explained variability. This scheme has been applied to nutrimetabolomics to evaluate the effect of the intake by obese individuals of breakfasts prepared with different vegetable oils after deep-frying. Urine samples from these subjects were obtained before and 2 and 4 h after the intake. With the aim of increasing the statistical relevance of the study, breakfast intake was carried out after a period of 2 weeks to recovering basal conditions. The oils used in this study were: (i) extra virgin olive oil (VOO) with a final concentration of total phenols of 400 ?g/mL expressed as ?g/mL of caffeic acid by the Folin?Ciocalteu test? (ii) pure refined sunflower oil with nil content in phenolic compounds (NSO); (iii) r efined high-oleic sunflower oil that was spiked at 400 ? g/mL with a synthetic lipophilic oxidation inhibitor (dimethylsiloxane, DSO); and (iv) r efined high-oleic sunflower oil (PSO) that 540 Nuevas plataformas anal?ticas en metabol?mica was enriched with an extract of hydrophilic phenols isolated from olive pomace at 400 ? g/mL of total phenols expressed as caffeic acid. Therefore, three different analytical strategies were used, as explained in Chapters 11 to 13. The order followed was based on the increasing level of information obtained from each instrumental platform. Chapter 11 This chapter is devoted to the use of NIRS to obtain urinary fingerprints. This technique is the less applied to fingerprinting. In fact, it has been only used with classification and prediction purposes, as it has scarce capacity to unequivocally identify metabolites. Among its strengths, it is worth noting its rapidity, the possibility to directly analyze samples (even solid samples such as tissues), the easiness to use and the low acquisition and maintenance costs of the equipment. The research was based on the use of a modelling technique by supervised PLS-CM (Partial-Least Squares- Class Modelling), which served to develop four models, one for each breakfast intake. In comparison to other PLS techniques based on discriminant analysis (PLS-DA), PLS-CM gives information about spectral characteristics of each class and estimates sensitivity (capacity of the model to predict samples from the modelled class) and specificity (capacity of the model to discriminate from samples belonging to other classes) for the model development and class validation. Both parameters are very useful as they describe the robustness of each model. Other novelty of this research is the data pretreatment. Thus, spectral data were not used themselves, but blank spectra were subtracted from the spectra obtained after intake. The aim was to minimize variability among individuals to avoid the effect of hindering variability associated to breakfast intake. The use of VIP algorithm (Variables Important in Projection) led to identify spectral regions with the highest statistical 541 Discussion of the results relevance to explain variability observed in each model. These regions were the same for all classes, but differences in the significance levels were found. Chapter 12 This chapter was focused on the use of NMR for the aforementioned nutritional study. In this case, samples were diluted for pH adjustment with buffer and deuterated water to lock the magnetic field strength. The study started with the application of a non-supervised analysis technique (PCA) to discriminate samples taken 4 h after intake against control samples . Once identified the biggest metabolic change associated to breakfast intake for each breakfast, the next step was the development of discrimination models between pairs of classes (with samples taken 4 h af ter intake) based on PLS- DA. By this analysis, accuracy levels of prediction between 84 and 100% were achieved. One advantage of this study as compared with the previous one is that signals associated to a larger variability between classes corresponded to chemical shifts in the NMR spectra, which were used to identify metabolites by peak assignement with Chenomx and HMDB database. With these premises, it was possible to identify glutamic acid, citric acid, lipids and compounds with amino groups, such as niacinamide, methylhistidine creatine and creatinine, as responsible for variability among classes. Chapter 13 The last chapter of this book was devoted to the analysis by high resolution LC?TOF/MS? In this case, a combined strategy based on supervised and non-supervised analysis by PLS-DA and PCA, respectively, was also used. In a previous step, the raw data matrix obtained from the analysis of all samples was simplified by applying a filter by frequency, thus 542 Nuevas plataformas anal?ticas en metabol?mica eliminating all molecular features that were not present in at least 50% of samples in each class, and a Fold Change filter. The PCA analysis allowed discriminating metabolic profiles detected in urine taken 4 h after intake. On the other hand, PLS-DA was used to discriminate metabolites from classes comparing breakfastss in groups of three, using samples after 4 h from intake. Thus, the first study provided discrimination between NSO, PSO and VOO with more than 69% of prediction accurac y in cross validation. In the second study, VOO, NSO and DSO classes were compared, giving prediction capabilities above 92, 73 and 50% by cross validation, with accuracy levels for model development of 100%. Therefore, it can be concluded that prediction capability for each class was good taking into account the use of biological samples, which implies a high degree of variability. The less favorable prediction for cross validation was obtained for DSO. Finally, the high resolution achieved by this technique increases the identification capability; thus, metabolites with a higher significance in explained variability were identified. It is worth noting the presence of sugars, hormones, lipids, amino acids and other minor compounds, the metabolism of which is affected by intake of these oils. The potential for unequivocally identifying these compounds was the greatest advantage of this application. CONCLUSIONES 545 Conclusiones Un estudio en profundidad de la bibliograf?a en metabol?mica ha permitido evaluar cr?ticamente el estado actual de las etapas que preceden al an?lisis (selecci?n de la muestra, muestreo, conservaci?n y preparaci?n), mostrando sus deficiencias o carencias y proporcionando as? pautas para la investigaci?n en esta parcela de la metabol?mica, que, hasta la fecha, no hab?a recibido una adecuada atenci?n. Se ha planteado el uso de las tres estrategias caracter?sticas de la metabol?mica con objetivos t?picos de cada una de ellas. As?, el an?lisis orientado se ha usado en el ?rea cl?nica (principalmente en lipid?mica, pero tambi?n en otras sub-disciplinas de la metabol?mica) y en metabol?mica vegetal para realizar los siguientes estudios: (i) El desarrollo de un m?todo para la determinaci?n de hormonas esteroideas femeninas en orina cuyas caracter?sticas anal?ticas est?n basadas en la reducci?n dr?stica del tiempo requerido para la hidr?lisis enzim?tica gracias a la acci?n de una sonda de ultrasonidos, el desarrollo miniaturizado y automatizado de la etapa de limpieza y concentraci?n de los analitos mediante un sistema lab-on-valve previa a la adecuada separaci?n por cromatograf?a de l?quidos y a la cuantificaci?n sensible y selectiva mediante espectrometr?a de masas de triple cuadrupolo tras la ionizaci?n por electroespray. (ii) El dise?o de una configuraci?n instrumental para la determinaci?n de precursores de esfingol?pidos basada en un sistema lab-on-valve para la preparaci?n de la muestra ?sangre u orina? mediante extracci?n en fase s?lida, y posterior derivatizaci?n de los analitos de forma simult?nea con la eluci?n 546 Nuevas plataformas anal?ticas en metabol?mica (inclusi?n del agente derivatizante ?o-ftaldialdeh?do? en el eluyente). Esta configuraci?n aceler? de forma dr?stica la etapa de preparaci?n de la muestra. Se llev? a cabo la separaci?n individual de los productos de derivatizaci?n de los analitos mediante el uso de un micro-cromat?grafo de l?quidos con detecci?n por fluorescencia asistida por l?ser. La aplicaci?n de esta configuraci?n dio lugar a un m?todo que result? ser tanto o m?s sensible que el basado en el uso de un detector de masas de triple cuadrupolo utilizado para la validaci?n. (iii) Tambi?n en el ?rea cl?nica, pero para analitos hidrosolubles, se ha puesto a punto un m?todo de determinaci?n de ?cido f?lico (vitamina B9) y sus metabolitos en suero, orina y leche materna en el que se han alcanzado l?mites de detecci?n muy bajos con los que poder diagnosticar estados de deficiencia de esta vitamina. Para ello se ha utilizado un sistema de extracci?n en fase s?lida autom?tico que trabaja a alta presi?n y est? conectado en l?nea con el cromat?grafo, lo que permite que alcance la cabeza de la columna de interacci?n hidrof?lica todo el eluido del cartucho utilizando fase m?vil cromatogr?fica como eluyente. Un detector de masas de triple cuadrupolo completa el equipo anal?tico para el desarrollo de este m?todo. Se han desarrollado dos m?todos basados en an?lisis orientado para compuestos en muestras vegetales. (iv) El primero para la determinaci?n en plantas de los subproductos de desinfecci?n constituidos por los ?cidos haloac?ticos, en el que la etapa de preparaci?n de la muestra se acort? enormemente con respecto al de referencia de la EPA gracias a la lixiviaci?n asistida por ultrasonidos y a la integraci?n de esta etapa con la de derivatizaci?n, para la posterior separaci?n? determinaci?n mediante cromatograf?a de gases y detecci?n por captura electr?nica. 547 Conclusiones (v) El desarrollo de tres m?todos basados en la utilizaci?n de cromatograf?a de gases y de l?quidos con detecci?n por espectrometr?a de masas ha permitido obtener perfiles de compuestos de inter?s nutrac?utico tales como carotenoides, provitaminas, vitaminas, fenoles e hidratos de carbono (mayoritariamente az?cares). El inter?s de estas metodolog?as radica en su uso potencial como marcadores en programas de mejora de la calidad del tomate mediante la selecci?n de variedades y el establecimiento del tiempo ?ptimo de recolecci?n. La obtenci?n del perfil metab?lico se ha orientado a fluidos biol?gicos poco estudiados con lo que: (vi) Se han puesto a punto protocolos de preparaci?n de muestra complementarios para la obtenci?n de perfiles metab?licos mediante LC?TOF/MS de fluidos biol?gicos de muestreo no invasivo y poco o nada utilizados hasta la fecha en este tipo de estrategia: S aliva y leche materna. El estudio exhaustivo de las etapas implicadas en la preparaci?n de la muestra y la aplicaci?n de la plataforma anal?tica utilizada (complementada en el caso de leche con NMR) han permitido obtener resultados excelentes en t?rminos de n?mero de metabolitos identificados en estos fluidos. La composici?n de metabolitos presentes en ambos biofluidos, no considerados en las actuales bases de datos de metabol?mica, incrementa el inter?s de los estudios realizados poniendo de manifiesto las buenas caracter?sticas de los protocolos desarrollados. 548 Nuevas plataformas anal?ticas en metabol?mica La estrategia de la huella dactilar metabol?mica ha sido utilizada en un estudio de nutrimetabol?mica destinado a la comparaci?n de los efectos metab?licos causados por la ingesta independiente de cuatro desayunos preparados con aceites vegetales sometidos a fritura. Este bloque ha sido abordado con la utilizaci?n de tres plataformas anal?ticas que han permitido obtener las siguientes conclusiones: (vii) La espectrometr?a de NIR junto a la t?cnica multivariante adecuada, en este caso PLS-CM, es de especial utilidad para el desarrollo de modelos predictivos en estudios de nutrimetabol?mica. En este caso se ha demostrado la utilidad de un pretratamiento de datos dirigido a la substracci?n de los espectros obtenidos para cada individuo en estado basal de los obtenidos tras la ingesta de los desayunos preparados con cada uno de los aceites. La utilizaci?n de la t?cnica de PLS-CM ha demostrado un gran potencial para el modelado de cada una de las clases estudiadas. (viii) La espectrometr?a de NMR combinada con un t?cnica multivariante, en este caso PLS-DA, ha permitido avanzar un nivel respecto al estudio anterior en t?rminos de informaci?n biol?gica. En concreto, se ha podido detectar la mayor variabilidad frente al estado basal en las huellas dactilares correspondientes al tiempo de muestreo de 4 h despu?s de la ingesta. Adem?s, el mayor poder de resoluci?n de la NMR ha quedado patente con la identificaci?n de aquellos metabolitos con mayor significado estad?stico para explicar la variabilidad observada. 549 Conclusiones (ix) El uso de LC?TOF/MS ha hecho posible la obtenci?n del m?ximo nivel de informaci?n biol?gica mediante el desarrollo de modelos discriminantes que permiten predecir el desayuno ingerido comparando dietas tres a tres. Adem?s, la alta resoluci?n, soportada en el an?lisis cromatogr?fico y en la espectrometr?a de masas en modo exactitud de masa, ha permitido alcanzar el mayor n?mero de metabolitos identificados, adem?s de poder diferenciar aquellos compuestos con mayor significado estad?stico para cada una de las ingestas. CONCLUSIONS 553 Conclusions An in-depth study of the literature in metabolomics has allowed a critical assessment of the present situation of the steps previous to analysis (sample selection, sampling, storage and preparation), showing its deficiencies or lacks; thus providing patterns for research in this step which had not received appropriate attention so far. The three characteristic strategies of metabolomics have been used with the typical objectives of each. Targeting analysis has been used in the clinical area (mainly in lipidomics, but also in other metabolomics sub-disciplines) and in plant metabolomics to develop the following studies: (i) The development of a method for the determination of female esteroid hormones in urine, with analytical characteristics based on the drastic reduction of the time required for enzymatic hydrolysis thanks to the action of an ultrasound probe, the miniaturized and automatized development of cleanup and concentration of the analytes by a lab-on-valve system prior to the appropriate individual separation of the analytes by liquid chromatography, sensitive and selective quantitation by electrospray ionization and triple-quad mass detection. (ii) The use of an instrumental configuration for the individual determination of sphingoid precursors, based on a lab-on-valve system for sample preparation of serum and urine ?by solid- phase extraction and derivatization of the target analytes, carried out simultaneously to elution (by solving the derivatizing agent, o-phthaldialdehyde, in the eluant)? and final quantitation by laser-induced fluorescence. This configuration drastically accelerated sample preparation. By means of a micro -LC, the analytes were individually separated and finally identified and 554 Nuevas plataformas anal?ticas en metabol?mica quantified by laser induced fluorescence. The resulting method was equal to, or more sensitive than, that based on a triple-quad mass detector, used for confirmatory analysis. (iii) Also in the clinical area, but for hydrosoluble analytes, a method for the determination of folic acid (vitamin B9) and its metabolites in serum, urine and breast milk has been developed. The very low limits of detection achieved allow determination of deficiency states of this vitamin. For this achievement, an automated system for solid phase extraction working at high pressure has been coupled on-line to a liquid chromatograph, thus allowing that all the eluate reached the chromatographic column of hydrophilic interaction by using the chromatographic mobile phase as eluent. A triple-quad mass spectrometer completed the analytical equipment for development of this method. Two methods have been developed for targeting analysis of compounds in vegetable samples. (iv) The first method for the determination in plants of disinfection byproducts, constituted by haloacetic acids, in which the sample preparation step was drastically shortened as compared with the reference EPA method. The shortening of this step was achieved by assisting leaching with ultrasound and integrating this step with derivatization as overall step prior to individual separation?determination by gas chromatography and electron- capture detection. (v) The development of three anaylytical methods based on the use of gas and liquid chromatography with mass spectrometry led to obtain profiles of compounds with great interest as nutraceuticals, such as carotenoids, provitamins, vitamins, polyphenols, and carbohydrates (mainly sugars). The interest of 555 Conclusions these methods relies on their potential use as biomarkers in programs of tomato improvements, to enhance product quality by cultivar selection and know the optimum time for harvesting. Metabolomic profiling approaches have been devoted to the analysis of samples scarcely studied; thus: (vi) Sample preparation protocols have been developed to obtain complementary metabolomic profiles by LC?TOF/MS of biological fluids of non-invasive sampling that have been scarcely or never used so far for this type of metabolomic strategy: saliva and breast milk. The exhaustive study of the steps involved in sample preparation and the application of the analytical platform LC?TOF/MS (complemented with NMR in the case of milk samples) have made possible to obtain excellent metabolomics profiles. The presence of metabolites that had not been identified previously demonstrated the great potential of the developed protocols. Metabolomics fingerprinting approaches have been used in a nutrimetabolomics study aimed at comparing the metabolic effects of the intake of four different meals, prepared with fried vegetable oils. This part has bee carried out by using three different analytical platforms, which led to the following conclusions: (vii) Near-infrared spectrometry, together with a multivariant technique (PLS-CM) was used to build predictive models. In this nutritional, the effectiveness of sample pretreatment based on substraction of spectra obtained in basal conditions to those obtained after meals intake. The use of PLS-CM has demonstrated to provide great results for modelling every class. 556 Nuevas plataformas anal?ticas en metabol?mica (viii) NMR spectrometry, combined with a multivariant technique (PLS-DA), allowed to obtain a step forward in the information obtained for this study. Thus, the metabolic fingerprints obtained 4 h after intake were significatively more different from blank samples than in the previous study. Moreover, the higher resolution capability of NMR was evident for identification of several compounds related to oils intake. (ix) LC-TOF/MS allowed to obtain the maximum level of biological information by the development of discriminant models by PLS- DA, which allowed predicting meals intake by comparing breakfasts in groups of three. Furthermore, the higher resolution of this technique and the use of high-accuracy mass spectrometry led to identify a great number of compounds that where statistically related to the intake of each breakfast. A NEXOS Anexe s A NEXO 1 : Art?culo enviado para su publicaci?n An approach to the phytochemical profile of rocket (Eruca sativa (Mill.) Thell) Myriam Villatoro-Pulido1, Feliciano Priego-Capote2, Beatriz ?lvarez -S?nchez 2, Shikha Saha3, Mark Philo4, Sara Obreg?n -Cano5, Antonio De Haro -Bail?n5, Rafael Font6, Mercedes Del R?o -Celestino6 1 IFAPA Centro-Alameda del Obispo, Department of Plant Breeding and Biotechnology, C?rdoba, Spain. 2 Department of Analytical Chem istry, Annex Marie Curie Building, Campus of Rabanales, University of C?rdoba, C?rdoba, Spain. 3 Phytochemicals and Health Programme, Institute of Food Research, Norwich Research Park, NR4 7UA Norwich, United Kingdom. 4 Metabolomics and Mass Spectrometry, Institute of Food Research, Norwich Research Park, NR4 7UA Norwich, United Kingdom. 5 Department of Plant Breeding, Institute of Sustainable Agriculture (IAS - CSIC), Alameda del Obispo s/n, 14080 C?rdoba, Spain. 6 IFAPA Centro La Mojonera, Department of Pl ant Breeding and Biotechnology, La Mojonera, Almer?a, Spain. Sent to Phytochemical Analysis for publication 563 Annex 1 Sent to phytochemical analysis An approach to the phytochemical profile of rocket (Eruca sativa (Mill.) Thell) Myriam Villatoro-Pulido, Feliciano Priego-Capote, Beatriz ?lvarez -S?nchez, Shikha Saha, Mark Philo, Sara Obreg?n -Cano, Antonio De Haro -Bail?n5, Rafael Font, Mercedes Del R?o -Celestino Abstract The purpose of this study was to determine the profile of different families of compounds with nutraceutical and organoleptical properties in leaves or four rocket accessions (Eruca vesicaria subsp. sativa). The target families were glucosinolates, isothiocyanates, phenolic compounds, carotenoids and carbohydrates. The four accessions were named according to the total content of glucosinolates that ranged from 14??? to ????? ?mol/ g of dry weight. Glucoraphanin represented up to 52% of the total glucosinolate content in leaves (high glucosinolate content 1 accession). Accessions showed differences in the hydrolysis of glucoraphanin and formation of the isothiocyanate, sulforaphane. Data showed no correlation between both compounds in leaves, which suggested differences in the myrosinase activity within accessions. In addition leaves of rocket had variable phenolic profiles represented by quercetin-3-glucoside, rutin, traces of myricetin, quercetin and phenolic acids such as ferulic and p- coumaric acids. The total carotenoid content ranged from 16.2 to ??? ?g/g of dry weight revealing a high variability. Lutein was the main carotenoid ranging from ??? to ????? ?g/g dw? The low glucosinolate content ? accession is a good candidate for future breeding programs because of its pattern of healthy beneficial related compounds. However, further research is essential to evaluate the biological activity of these four accessions, assessing the possible non-desirable effects before planning strategies to design functional food and improving consumer?s health? A NEXO 2 : Cap?tulos de libros multiautor 567 Annex 2 Chapters in multiauthors book 568 Nuevas plataformas anal?ticas en metabol?mica 569 Annex 2 Chapters in multiauthors book 571 Annex 2 Chapters in multiauthors book Book: ENCYCLOPEDIA OF ANALYT ICAL CHEMISTRY ?ont?ib?to??s details Mar?a Dolores Luque de Castro, Ph.D., Full Professor, Feliciano Priego-Capote Beatriz ?lvarez -S?nchez Title of Chapter Analytical platforms in metabolomics Name of Contributor(s) Feliciano Priego-Capote, Mar?a Dolores Luque de Castro * , Beatriz ?lvarez - S?nchez *Corresponding author. Affiliation of the Contributor(s) University of C?rdoba, C?rdoba, Spain. 572 Nuevas plataformas anal?ticas en metabol?mica Abstract The present importance of metabolomics to go inside the unsolved secrets of living organisms, as a single discipline or together with other omics, has open the interest in the analytical platforms to obtain proper metabolomic information. The steps involved in developing these platforms (viz. sampling, sample storage, sample preparation, individual metabolites separation, if required, and detection) are critically discussed as a function of the type of organism under study, the type of sample and the families of metabolites to be analyzed. An overview of the fundamentals of key equipment in metabolomics, such as that based on nuclear magnetic resonance or mass spectrometry is also given for better understanding of its usefulness in this discipline. Key words Metabolomics; analytical platforms; mass spectrometry; nuclear magnetic resonance; biological samples; sample preparation; gas chromatography; liquid chromatography. A NEXO 3 : Carteles presentados en congresos C arteles presentados en congresos 575 C arteles presentados en congresos 577 578 Nuevas plataformas anal?ticas en metabol?mica C arteles presentados en congresos 579 580 Nuevas plataformas anal?ticas en metabol?mica C arteles presentados en congresos 581 C arteles presentados en congresos 583 C arteles presentados en congresos 585 C arteles presentados en congresos 587 C arteles presentados en congresos 589 C arteles presentados en congresos 591 L ISTA DE ABREVIATURAS Lista de abreviaturas 595 Lista de abreviaturas ACE unidad de extracci?n en fase s?lida ANOVA an?lisis de varianza a- pABGA derivado acetamida de ?cido L-glut?mico N-(p-aminobenzoilo) BDCAA ?cido bromodicloro ac?tico CBAA ?cido clorobromo ac?tico CDBAA ?cido clorodibromo ac?tico CE electroforesis capilar DAD detector de diodos en fila DBAA ?cid dibromo ac?tico DCAA ?cido dicloro ac?tico DSO aceite de girasol enriquecido con dimetilsiloxano ECD detecci?n por captura electr?nica ESI ionizaci?n por electrospray FA ?cido f?lico FC estimaci?n de cambio relativo FIA An?lisis por inyecci?n en flujo FT - IR espectrometr?a de reflectancia en el infrarrojo por transformadas de Fourier GC cromatograf?a de gases HILIC cromatograf?a l?quida con fase estacionaria de interacci?n hidr?fila HPD bomba de alta presi?n dispensadora de disolventes LC cromatograf?a l?quida LIF fluorescencia inducida por l?ser L L E extracci?n l?quido?l?quido 596 Nuevas plataformas anal?ticas en metabol?mica LOV laboratorio en v?lvula MBAA ?cido monobromo ac?tico MCAA ?cido monocloro ac?tico MS espectrometr?a de masas MS/MS espectrometr?a de masas en t?ndem NIRS espectrometr?a de reflectancia en el infrarrojo cercano NSO aceite de girasol natural OPA o-ftaladialdeh?do pABGA ?cido L-glut?mico N-(p-aminobenzoilo) PCA an?lisis por componentes principales PLS - CM an?lisis discriminante de componentes principales por modelado de clases PLS - DA an?lisis discriminante de componentes principales PSO aceite de girasol enriquecido con polifenoles qQq espectr?metro de masas de triple cuadrupolo RMN resonancia magn?tica nuclear RP- C18 cromatograf?a l?quida con fase estacionaria reversa C-18 Sa D-eritro-dihidroesfinganina Sa1 - P D-eritro-dihidroesfingosina 1-fosfato So D-esfingosina So - 1P esfingosina 1-fosfato SPE extracci?n en fase s?lida SRM seguimiento de reacciones m?ltiples TBA ?cido tribromo ac?tico TCAA ?cid tricloro ac?tico TOF espectr?metro de masas de tiempo de vuelo VOO aceite de oliva virgen A BBREVIATIONS LIST Abbreviations list 599 Abbreviations list ? LC micro-liquid chromatograpy 1D - NMR monodimensional Nuclear Magnetic Resonance 2D - NMR bidimensional Nuclear Magnetic Resonance ACE automatic cartridge exchange a- pABGA para-aminobenzoglutamic acid acetamide-derivative AU absorbance units BDCAA bromidichloroacetic acid BPC base peak chromatogram BSTFA N,O-Bis(trimethylsilyl)trifluoroacetamide CDBAA chlorodibromoacetic acid CE capillary electrophoresis CHD coronary heart disease CI chemical ionization COSI correlation spectroscopy CV coefficient of variation D2O deuterated water DAD diode array detector DBPs disinfection by products DSO sunflower oil enriched with dimethylsiloxane ECD electron capture detection EI electron impact EMV electron multiplier voltage EPA environmental protection agency ESI electrospray ionization FA folic acid 600 Nuevas plataformas anal?ticas en metabol?mica FC fold change FDA food and drug a dministration FIA flow-injection analysis FMASE focused microwave-assisted Soxhlet extraction FTIRS Fourier transform-infrared spectroscopy GC gas chromatography GMs genetic modifications HAA haloacetic acids HILIC hydrophilic interaction chromatography HMBC heteronuclear multiple-bond correlation spectroscopy HMDB human metabolome database HPD high-pressure delivery unit HSQC heteronuclear single-quantum correlation spectroscopy HUSERMET human serum metabolites database IS internal standard IT ion trap KEGG Kyoto encyclopedia of genes and g enomes LC liquid chromatography LIF laser-induced detection LOD limit of detection LOQ limit of quantification LOV lab-on-valve MALDI matrix-assisted laser desorption/ionization MAS magic angle spin MCA metabolic control analysis MCAA monochloroacetic acid Abbreviations list 601 MFE molecular features extraction MIAMET minimum Information about a metabolomics experiment MRM multiple reaction monitoring MS mass spectrometry MSI metabolomics standard initiative MS /MS tandem mass spectrometry MUFAs mono-unsaturated fatty acids NIRS near-infrared spectrometry NMR nuclear magnetic resonance NOESY nuclear overhauser effect spectroscopy NSO non-treated sunflower oil OPLS - DA orthogonal projections to latent structures discriminant analysis pABGA para-aminobenzoglutamic acid PCA principal components analysis PLS - CM partial least squares-class PLS - DA Modelling PP peristaltic pump PSO poliphenols-enriched sunflower oil PUFA poliunsaturated fatty acid Q1 first Quadrupole Q3 third Quadrupole QqQ triple quadrupole mass detector RBC red blood cells RFU relative fluorescence units RP- LC reverse-phase-liquid chromatography RSD relative standard deviation 602 Nuevas plataformas anal?ticas en metabol?mica S/N signal-to-noise ratio Sa sphinganine Sa1P sphinganine-1-phosphate SIM single-ion-monitoring SLs sphingolipids So sphingosine So1P sphingosine 1-phosphate SPE solid phase extraction SRM single reaction monitoring SV switching valve TBAA tribromoacetic acid TCAA trichloroacetic acid TIC total ion chromatogram TMCS trimethylchlorosilane TOCSY total correlation spectroscopy TOF time-of-flight mass spectrometer TSP trimethylsilyl-propionate U enzimatic activity UPLC ultra-high performance liquid chromatography US ultrasound USAEH ultrasound-assisted enzymatic hydrolysis VOO virgin olive oil