<%BANNER%>

Lc/ms-Based Targeted and Global Metabolomic Methodologies and Their Application to Biomarker Discovery

Permanent Link: http://ufdc.ufl.edu/UFE0024947/00001

Material Information

Title: Lc/ms-Based Targeted and Global Metabolomic Methodologies and Their Application to Biomarker Discovery
Physical Description: 1 online resource (137 p.)
Language: english
Creator: Cerutti, Estela
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2009

Subjects

Subjects / Keywords: biomarker, blood, chromatography, electrospray, hilic, metabolomics, ms, rplc
Chemistry -- Dissertations, Academic -- UF
Genre: Chemistry thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: LC/MS-BASED TARGETED AND GLOBAL METABOLOMIC METHODOLOGIES AND THEIR APPLICATION TO BIOMARKER DISCOVERY Small-molecule profiling, termed metabolomics, is a valuable tool to study phenotype and changes in phenotype caused by environmental influences, disease, or changes in genotype. The metabolome can be defined as the complete complement of all small molecule ( < 1500 Da) metabolites found in a specific cell, organ or organism. The metabolome represents a vast number of components that belong to a wide variety of compound classes, and these compounds are very diverse in their physical and chemical properties and occur in a wide concentration range. Consequently, studying the metabolome is a major challenge to analytical chemistry. Mass spectrometry (MS) is used in metabolomics to detect, quantify, and identify enzymatic substrates and byproducts from biological and clinical samples. MS-based metabolomics offers qualitative and quantitative analyses with high selectivity and sensitivity, wide dynamic range, and the ability to analyze biofluids with extreme molecular complexity. The combination of liquid chromatography with MS reduces the complexity of the mass spectra, decreases ion suppression, provides isobar resolution, and delivers information on the properties of the metabolites. In this study, the hypothesis that detectable changes will occur in the blood plasma metabolic profile of healthy female and male adults before and after a ketogenic diet has been tested. In addition, changes in the plasma metabolome of piglets between days 2 through 8 of life have been evaluated. Novel complementary chromatographic approaches?reversed phase and hydrophilic interaction liquid chromatography, directly coupled to a time-of-flight mass spectrometer operating under electrospray conditions in positive ion mode, have been developed and optimized. The performance/contribution of each separation strategy, identification of unique m/z features, and technical variability have been evaluated. The studies involved a large number of samples that required powerful data processing/analysis capabilities. In this sense, the raw data were processed using commercial instrument software. From the obtained chromatograms, features were extracted, aligned, normalized, filtered, and then analyzed by different statistical methods, including analysis of variance, principal component analysis, and volcano plots. Finally, using the accurate mass criterion of 2 ppm mass error, putative biomarkers responsible for the metabolic differences in the samples were identified using several databases.
General Note: In the series University of Florida Digital Collections.
General Note: Includes vita.
Bibliography: Includes bibliographical references.
Source of Description: Description based on online resource; title from PDF title page.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility: by Estela Cerutti.
Thesis: Thesis (Ph.D.)--University of Florida, 2009.
Local: Adviser: Yost, Richard A.
Local: Co-adviser: Powell, David H.

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2009
System ID: UFE0024947:00001

Permanent Link: http://ufdc.ufl.edu/UFE0024947/00001

Material Information

Title: Lc/ms-Based Targeted and Global Metabolomic Methodologies and Their Application to Biomarker Discovery
Physical Description: 1 online resource (137 p.)
Language: english
Creator: Cerutti, Estela
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2009

Subjects

Subjects / Keywords: biomarker, blood, chromatography, electrospray, hilic, metabolomics, ms, rplc
Chemistry -- Dissertations, Academic -- UF
Genre: Chemistry thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: LC/MS-BASED TARGETED AND GLOBAL METABOLOMIC METHODOLOGIES AND THEIR APPLICATION TO BIOMARKER DISCOVERY Small-molecule profiling, termed metabolomics, is a valuable tool to study phenotype and changes in phenotype caused by environmental influences, disease, or changes in genotype. The metabolome can be defined as the complete complement of all small molecule ( < 1500 Da) metabolites found in a specific cell, organ or organism. The metabolome represents a vast number of components that belong to a wide variety of compound classes, and these compounds are very diverse in their physical and chemical properties and occur in a wide concentration range. Consequently, studying the metabolome is a major challenge to analytical chemistry. Mass spectrometry (MS) is used in metabolomics to detect, quantify, and identify enzymatic substrates and byproducts from biological and clinical samples. MS-based metabolomics offers qualitative and quantitative analyses with high selectivity and sensitivity, wide dynamic range, and the ability to analyze biofluids with extreme molecular complexity. The combination of liquid chromatography with MS reduces the complexity of the mass spectra, decreases ion suppression, provides isobar resolution, and delivers information on the properties of the metabolites. In this study, the hypothesis that detectable changes will occur in the blood plasma metabolic profile of healthy female and male adults before and after a ketogenic diet has been tested. In addition, changes in the plasma metabolome of piglets between days 2 through 8 of life have been evaluated. Novel complementary chromatographic approaches?reversed phase and hydrophilic interaction liquid chromatography, directly coupled to a time-of-flight mass spectrometer operating under electrospray conditions in positive ion mode, have been developed and optimized. The performance/contribution of each separation strategy, identification of unique m/z features, and technical variability have been evaluated. The studies involved a large number of samples that required powerful data processing/analysis capabilities. In this sense, the raw data were processed using commercial instrument software. From the obtained chromatograms, features were extracted, aligned, normalized, filtered, and then analyzed by different statistical methods, including analysis of variance, principal component analysis, and volcano plots. Finally, using the accurate mass criterion of 2 ppm mass error, putative biomarkers responsible for the metabolic differences in the samples were identified using several databases.
General Note: In the series University of Florida Digital Collections.
General Note: Includes vita.
Bibliography: Includes bibliographical references.
Source of Description: Description based on online resource; title from PDF title page.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility: by Estela Cerutti.
Thesis: Thesis (Ph.D.)--University of Florida, 2009.
Local: Adviser: Yost, Richard A.
Local: Co-adviser: Powell, David H.

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2009
System ID: UFE0024947:00001


This item has the following downloads:


Full Text

PAGE 1

1 LC/MS-BASED TARGETED AND GLOBAL METABOLOMIC METHODOLOGIES AND THEIR APPLICATION TO BIOMARKER DISCOVERY By ESTELA SOLEDAD CERUTTI A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLOR IDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2009

PAGE 2

2 2009 Estela Soledad Cerutti

PAGE 3

3 To my son and beloved family

PAGE 4

4 ACKNOWLEDGMENTS This work would not have been possible without the guidance and encouragem ent I received from my academic advisors, Dr. Rich ard A. Yost and Dr. David H. Powell, under whose supervision I chose this topic and began the research. I would like to sincerely express that I have enriched my life from this experi ence in both personal and professional aspects. I also like to express my gr atitude to Dr. Peggy Borum, whose thoughtful advice often served to give me a sense of direction dur ing my PhD studies. I would like to extend my gratitude to Dr. Ben Smith and Dr. John Eyler for their assistance and for spending their valuable time serving on my committee. I also want to tha nk the members of Powells (Julia, Jodie, Cris, Basri, Noelle, and Jacina) and Yosts groups for th eir friendship and helpful advice in situations related to science, research, and life. I am deeply grateful to my academic in stitution in Argentina and the Fulbright Commission for the trust and support that they gave me in order to study in the United States. I am tempted to thank all of my friends individually; however, b ecause the list might be too long and by fear of leaving someone out, I will simply say thank you very much to you all. I give thanks from the bottom of my heart to my parents, Lila and Luis, for providing me with the values to discover and achieve my goals in life. My gratitude also goes to my beloved sister, Laura, for her continuous support and encouragement to find the positive side of everything. My family and family-to-be deserve so much of the credit for my success. I cannot end without thanking God for givi ng me the strength and the opport unity to achieve one of the most important dreams in my life.

PAGE 5

5 TABLE OF CONTENTS page ACKNOWLEDGMENTS ............................................................................................................... 4LIST OF TABLES ...........................................................................................................................8LIST OF FIGURES .........................................................................................................................9ABSTRACT ...................................................................................................................... .............12 CHAP TER 1 INTRODUCTION .................................................................................................................. 14Metabolomics .................................................................................................................. .......14Background .................................................................................................................... ..14Metabolomic Approaches ................................................................................................14Analytical Instrument Platforms ...................................................................................... 15Time-of-Flight Mass Spectrometer ......................................................................................... 22Background .................................................................................................................... ..22Orthogonal Acceleration Time-ofFlight Mass Spectrometer ......................................... 23Electrospray Ionization ....................................................................................................... ....23Liquid Chromatography ..........................................................................................................25Monolithic Columns ........................................................................................................ 25HILIC Columns ...............................................................................................................27Target Compounds ..................................................................................................................30Carnitine and Acylcarnitines ...........................................................................................30Amino Acids ................................................................................................................... .312 DEVELOPMENT AND COMPARISON OF T WO CHROMATOGRAPHIC APPROACHES .................................................................................................................... ..40Liquid Chromatography ..........................................................................................................40Experimental .................................................................................................................. .........42Mass Spectrometer ..........................................................................................................42TOF Calibration ............................................................................................................... 42HPLC System ..................................................................................................................43Standard Solutions ...........................................................................................................43Mobile Phase Preparation ................................................................................................43Plasma Sample Preparation .............................................................................................44MS and Chromatographic Conditions ............................................................................. 44Analytes of Interest ..........................................................................................................45

PAGE 6

6 Results and Discussion ........................................................................................................ ...46Coupling of Two C18-Monolithic Columns ..................................................................... 46HILIC Method Development and Optimization .............................................................. 47Effect of acetonitrile content and change on retention ............................................. 47Effect of salt type and concentration ........................................................................ 47Effect of temperature: Vant Hoff plots ................................................................... 48Effect of flow rate: van Deemter plots .....................................................................49Evaluation of ionizat ion suppression .......................................................................49Repeatability ................................................................................................................. ...50Applicability of HILIC to the Separati on of Endogenous Metabolites in Plasma Samples ....................................................................................................................... .50Specificity of HILIC vs. C18-Monolithic Chromatography ............................................. 50Conclusions .............................................................................................................................523 APPLICATION OF LC/MS METHODOLOGIES TO HUMAN PLASMA SAMPLES FOR GLOBAL AND TARGETED METABOLOMIC STUDIES .......................................65Metabolomics Applied to Biomarker Discovery .................................................................... 65Ketogenic Diet Therapy ..................................................................................................66Experimental Workflow .................................................................................................. 68Sample analysis ........................................................................................................68Identification ............................................................................................................ 74Biological interpretation ...........................................................................................74Experimental Methods .......................................................................................................... ..75Sample Collection ...........................................................................................................75Sample Preparation .......................................................................................................... 76Results and Discussion ........................................................................................................ ...76LC/MS Experiments ........................................................................................................ 76Global Metabolomics Strategy ........................................................................................ 77Hierarchical clustering analysis ............................................................................... 77Principal component analysis ................................................................................... 78Analysis of variance .................................................................................................78Volcano analysis ......................................................................................................79Database search for identification ............................................................................79Targeted Metabolomic Strategy ...................................................................................... 80Conclusions .............................................................................................................................814 APPLICATION OF LC/MS METHODOLOGIES TO PIGLET PLASMA SAMPLES FOR GLOBAL AND TARGETED METABOLOMIC STUDIES .....................................103The Plasma Metabolome of Pigl ets from Days 2-8 of Life ..................................................103Experimental Workflow ................................................................................................ 105Experimental Methods .......................................................................................................... 105Sample Collection .........................................................................................................105Sample Preparation ........................................................................................................ 105Results and Discussion ........................................................................................................ .105LC/MS Experiments ...................................................................................................... 105

PAGE 7

7 Global Metabolomics Approach .................................................................................... 106Hierarchical clustering analysis ............................................................................. 106Principal component analysis ................................................................................. 107Analysis of variance ...............................................................................................107Targeted Metabolomics ................................................................................................. 108Conclusions ...........................................................................................................................1095 CONCLUDING REMARKS AND FUTURE WORK ........................................................128REFERENCES .................................................................................................................... ........130BIOGRAPHICAL SKETCH .......................................................................................................137

PAGE 8

8 LIST OF TABLES Table page 1-1 Comparison of meta bolic technologies ..............................................................................341-2 Methods for data processing and analysis .........................................................................351-3 MS, NMR, and metabolic pathway databases ................................................................... 362-1 Chemical formula, RTs, and m/z of the [M+H]+ ions of carnitine, acylcarnitines, and amino acids ................................................................................................................... .....534-1 Piglet ID, day of life, and MS ID. .................................................................................... 111

PAGE 9

9 LIST OF FIGURES Figure page 1-1 Representation of the number of m etabolomic publications per year. .............................. 371-2 Main components of typical oa-TOF system with reflecting mass analyzer. ....................381-3 Chemical structures of free carnitine, acylcarnitines, and amino acids. ............................ 392-1 Base peak chromatogram obtai ned with a monolithic column. ......................................... 542-2 Effect of acetonitrile content.. .......................................................................................... ..552-3 Effect of buffer addition.. ...................................................................................................562-4 Extracted ion chromatograms for some model compounds showing the effect of buffer addition. ...................................................................................................................572-5 Extracted ion chromatograms for some model compounds demonstrating the effect of buffer addition. ..............................................................................................................582-6 Vant Hoff plots for evaluating the eff ect of the temperature on the retention efficiency.. .................................................................................................................. ........592-7 van Deemter plots for evaluating the eff ect of the mobile phase flow rate on the retention efficiency. ...........................................................................................................602-8 Total ion chromatograms for the application of C18-monolithic/ and HILIC/MS on piglet plasma samples. ....................................................................................................... 612-9 Extracted ion chromatograms of carnitine, acetylcarnitine, propionylcarnitine, and butyrylcarnitine under HILIC and C18-monolithic conditions. .......................................... 622-10 Extracted ion chromatograms of alan ine, proline, valine, threonine, and leucine/isoleucine under HILIC and C18-monolithic conditions. ...................................... 632-11 Extracted ion chromatograms of glutamic acid, methionine, histidine, phenylalanine, and tyrosine under HILIC and C18-monolithic conditions. ................................................ 643-1 Study design for the evaluation of the me tabolomic effects of a ketogenic diet on healthy adults. ............................................................................................................... .....823-2 Experimental workflow ..................................................................................................... 833-3 Total ion chromatograms (T ICs) for the overlapping of all female samples before and after ketogenic diet using C18-monolithic chromatography. ....................................... 84

PAGE 10

10 3-4 Total ion chromatograms (T ICs) for the overlapping of all fem ale samples before and after ketogenic diet us ing HILIC chromatography.. ................................................... 853-5 Total ion chromatograms (T ICs) for the overlapping of all male samples before and after ketogenic diet using C18-monolithic chromatography ...............................................863-6 Total ion chromatograms (T ICs) for the overlapping of all male samples before and after ketogenic diet usin g HILIC chromatography. ........................................................... 873-7 Hierarchical clustering anal ysis (monolithic column). ...................................................... 883-8 Hierarchical clustering an alysis (HILIC column). ............................................................. 893-9 Principal component analysis on C18-monolithic column data files. ................................. 903-10 Principal component analysis on HILIC data files. ........................................................... 913-11 One-way analysis of variance (ANOVA) for C18-monolithic and HILIC approaches ...... 923-12 C18-chromatography Volcano plot s on differential abundance ....................................... 933-13 HILIC-chromatography Volcano plots on differential abundance. .................................943-14 C18-chromatography Venn diagrams. ............................................................................. 953-15 HILIC chromatography Venn diagrams. ........................................................................ 963-16 C18-chromatography Classes of compounds identified before and after diet. ................ 973-17 HILIC chromatography Classes of compounds. ............................................................. 983-18 A/B (after diet/befor e diet) ratio response vs. carnitine and acylcarnitines in each of the ten female samples. ......................................................................................................993-19 A/B (after diet/befor e diet) ratio response vs. carnitine and acylcarnitines in each of the ten male samples. ....................................................................................................... 1003-20 A/B (after diet/before diet) ratio versus carnitine and acylcarnitines profiles for all the female samples. .......................................................................................................... 1013-21 A/B (after diet/before diet) ratio versus female profiles for all the carnitine and acylcarnitines under study. ...............................................................................................1013-22 A/B (after diet/before diet) ratio versus carnitine and acylcarnitines profiles for all the male samples. ............................................................................................................. 1023-23 A/B (after diet/before diet) ratio versus males for all the carnitine and acylcarnitines. .. 102

PAGE 11

11 4-1 Study design for the evaluation of the me tabolom ic changes from day 2 to 8 in the life of piglets. ...................................................................................................................1124-2 Total ion chromatograms (T ICs) for the overlapping of all piglet samples using C18monolithic chromatography ............................................................................................. 1134-3 Total ion chromatograms (TICs) for the overlapping of all piglet samples using HILIC chromatography. ...................................................................................................1144-4 Hierarchical clustering anal ysis (monolithic column). .................................................... 1154-5 Hierarchical clustering an alysis (HILIC column). ........................................................... 1164-6 Principal component analys is (monolithic column). ....................................................... 1174-7 Principal component anal ysis (HILIC column). .............................................................. 1184-8 One-way analysis of variance (ANOVA) for C18-monolithic. ........................................ 1194-9 One-way analysis of va riance (ANOVA) for HILIC. ...................................................... 1204-10 Targeted metabolomics analysis. Carnitine profile. ........................................................ 1214-11 Targeted metabolomics analysis. Acetylcarnitine profile. ............................................... 1214-12 Targeted metabolomics analysis. Propionylcarnitine profile. .......................................... 1224-13 Targeted metabolomics analysis. Butyrylcarnitine profile. ............................................. 1224-14 Targeted metabolomics analysis. Hexanoylcarnitine profile. .......................................... 1234-15 Targeted metabolomics analysis. Octanoylcarnitine profile. ........................................... 1234-16 Targeted metabolomics analysis. Decanoylcarnitine profile. .......................................... 1244-17 Targeted metabolomics analysis. Lauroylcarnitine profile. ............................................. 1244-18 Targeted metabolomics analysis. Myristoylcarnitine profile. .......................................... 1254-19 Targeted metabolomics analysis Palmitoylcarnitine profile. ..........................................1254-20 Targeted metabolomics analysis. Stearoylcarnitine profile. ............................................ 1264-21 Targeted metabolomics analysis. Carnitine and acylcarnitines profiles between days 2 through 8 in the life of piglets. ...................................................................................... 127

PAGE 12

12 Abstract of Dissertation Pres ented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy LC/MS-BASED TARGETED AND GLOBAL METABOLOMIC METHODOLOGIES AND THEIR APPLICATION TO BIOMARKER DISCOVERY By Estela Soledad Cerutti August 2009 Chair: Richard A. Yost Cochair: David H. Powell Major: Analytical Chemistry Small-molecule profiling, termed metabolomics, is a valuable tool to study phenotype and changes in phenotype caused by environmental infl uences, disease, or changes in genotype. The metabolome can be defined as the complete co mplement of all small molecule (<1500 Da) metabolites found in a specific cell, organ or organism. The metabolome represents a vast number of components that belong to a wide variety of comp ound classes, and these compounds are very diverse in their physical and chemical properties and occur in a wide concentration range. Consequently, studying the metabolome is a major challenge to analytical chemistry. Mass spectrometry (MS) is used in meta bolomics to detect, quantify, and identify enzymatic substrates and byproducts from bi ological and clinical samples. MS-based metabolomics offers qualitative and quantitative analyses with high selectivity and sensitivity, wide dynamic range, and the ability to analyze biofluids with extreme molecular complexity. The combination of liquid chromatography with MS reduces the complexity of the mass spectra, decreases ion suppression, provide s isobar resolution, and delivers information on the properties of the metabolites.

PAGE 13

13 In this study, the hypothesis that detectable changes will occur in the blood plasma metabolic profile of healthy female and male adults before and after a ke togenic diet has been tested. In addition, changes in th e plasma metabolome of piglets between days 2 through 8 of life have been evaluated. Novel complementary chromatographic a pproachesreversed phase and hydrophilic interaction liquid chromatography, directly coupl ed to a time-of-flight mass spectrometer operating under electrospray conditions in pos itive ion mode, have been developed and optimized. The performance/contribution of each separation strategy, id entification of unique m/z features, and technical variability have been evaluated. The studies involved a large number of samples that required powerful data processi ng/analysis capabilities. In this sense, the raw data were processed using commercial instrument software. From the obtained chromatograms, features were extracted, aligned, normalized, filtered, and then analyzed by different statistical methods, including analysis of variance, prin cipal component analys is, and volcano plots. Finally, using the accurate mass criterion of 2 ppm mass error, putative biomarkers responsible for the metabolic differences in the samples were identified using several databases.

PAGE 14

14 CHAPTER 1 INTRODUCTION Metabolomics Background Metabolom ics is one of the most rapidly gr owing and changing fi elds of contemporary science and it is one of those rela tively new omic disciplines. Th is field is evolving toward the intention of extracting useful knowledge from metabolite pools [1, 2]. Metabolomics is complementary to the other omic technologies as it is downstream of genomics, transcriptomics, and proteomics and, is closest to cells physio logical state [3]. Consequently, metabolomic analysis holds the promise to contribute to a be tter understanding of the interactions from genes to phenotypes. The rising number of publications in the field (Figure 1-1) demonstrates that metabolomics is not just a new omics word, but a valuable emerging technology to study the phenotype (which refers to the observable physic al or biochemical characteristics of an organism, as determined by both genetic makeup and environmental influences) and changes in the phenotype caused by environmental influe nces, disease, or changes in phenotype. The metabolome was first described by Oliver et al. in 1998 [4] as the set of metabolites found in or produced by an organism. This definition has been limited to the quantitative complement of all of the low molecular weight molecules (< 1500 Da) present in cells in a particular physiological or deve lopmental state [5]. Fiehn made distinctions between different metabolite analysis techniques and defined metabolomics as a comprehensive analysis in which all metabolites of a biological system were identified and quantified [6]. Metabolomic Approaches Metabolom ics defined as the unbiased identification and quantification of all the metabolites present in a specific biological sample cannot be carried out in its totality. Thus,

PAGE 15

15 different analytical approaches can help to redu ce the complexity of the analysis, these categories include target metabolomics, me tabolite profiling, and metabolite fingerprinting. Target analysis is constrained to the qualitative and quantitative analysis of a particular set of metabolites of known identity. Metabolite profiling involves the identification and quantitation of a predefined set of metabolites of known or unknown identity and belonging to a class or a selected metabolic pathway. Metabolite fingerprinting aims to rapidly classify numerous samples using multivariate statistics, typically with out differentiation of individual metabolit es or their quantit ation [5, 7-9]. The metabolome represents a vast number of components that belong to a wide variety of compound classes. These compounds are very divers e in their physical and chemical properties and occur in a wide concentrati on range [1, 10]. There have been many attempts to estimate the number of metabolites in a biological system. However, the size of the metabolome varies greatly, depending on the organism studied [11]. Current estimates put the mammalian metabolome at about 3,000 different co mpounds [12] whereas the yeast and E. coli metabolomes are believed to consist of betw een 600 to 800 compounds [13], and it is estimated that the plant kingdom may encode more than 200,000 metabol ites [14]. These molecules can be ionic inorganic species, hydrophilic co mpounds or amphoteric analytes. Furthermore, the elemental composition, the order of atoms, an d the stereochemical orientation may have to be elucidated de novo for metabolites [2]. Thus, there is no singl e technology platform that has the ability to profile the metabolome simultaneously a nd a combinatorial approach is used. Analytical Instru ment Platfor ms Analytical variance is a major factor when using different technology platforms and refers to the coefficient of variance related to a par ticular experimental approach. Therefore, the knowledge and correction of analy tical variance between experiment al approaches is necessary. Numerous analytical techniques have been used in the field of metabolomics to monitor and

PAGE 16

16 explore metabolic differences between biological samples, such as nuclear magnetic resonance (NMR) [15], Fourier tr ansform-infrared spectroscopy (FT-IR ) [16, 17] and mass spectrometry (MS) [18, 19] coupled to sepa ration techniques or using di rect flow injection. Magnetic resonance spectroscopy has several advantages relative to mass spectrometry: it does not require sample preparation and it produces signals that correlate directly and linearly with compounds abundances. However, only medium to high abundan ce metabolites will be detected with this technique and the identification of individual metabolites is ch allenging in complex mixtures [18]. On the other hand, mass spectrometry offers quantitative analyses w ith high selectivity and sensitivity and the potential to identify metabo lites [18]. However, mass spectrometry-based techniques usually require a sample preparati on step, which can cause metabolite losses and, based on the sample introduction system and the ionization technique used, specific metabolite classes may be discriminated against. There has been tremendous progress in mass spectrometry metabolomics in recent years, leaving researchers with a variety of choices for chromatographic separation, ionization, and mass spectrometry analysis. Separations may be achieved by gas chromatography (GC) [20], capillary electrophoresis (CE) [9 ], or liquid chromatography (LC) [21], with LC approaches continuously evolving. A comparis on of different commonly used metabolic technologies is shown in Table 1-1; specificity, sensitivity and structural range of the different methods vary substantially. Later discussion will be focused on those specific tec hnologies that were applied to this research work. In order to overcome the drawbacks of di rectly injecting complex samples, liquid chromatography can be associated with the MS detector for metabolomic analysis. Liquid chromatography can reduce ion suppression caused by coeluting compounds, isobaric

PAGE 17

17 interferences in the case of low-resolving mass analyzers, and often can separate isomers [1]. In addition, a good analytical separation will result in better detection limits and MS data quality due to reduced background noise. Important technological advances in liquid chromatography over the past few years have forged a new era of research. Recently, liquid chromatography has shifted from the standard hi gh-performance liquid chromatography (HPLC) to the ultra-high pressure liquid chromatography (UHPLC) which can increase resolution sensitivity and peak capacity while decreasing sample volumes and mobile phases [22]. The downside of this approach is the high pressure (10,000,000 psi) n eeded to operate these columns, and thus special UHPLC systems are required. Metabolomics deals with a great diversity of small molecules that differ greatly in their physical and chemical properties of size, pol arity/hydrophobicity, and ch arge [18]. While no single chromatographic method is suitable for all classes of metabolites, different alternatives exist. Reversed-phase liquid chromatography (RPL C) is a standard tool for the separation of medium polar and non-polar analytes. However, very polar metabolites are not retained on classical reversed-phase stationary phases a nd elute with the void volume. An interesting alternative to RPLC is the use of hydrophilic in teraction liquid chromatography (HILIC) for the separation of highly polar and hydrophilic compounds. HILIC is orthogonal to RPLC chromatography and uses polar stationary phase materials and low aqueous/high-organic solvent systems. In summary, technological a dvances in NMR and MS have introduced metabolomics as an approach to study the metabolism and its regulation in relation to disease, genetic, and environmental factors. With regard to human health alone, multiple benefits of metabolomics investigations can be visualize d. They can deliver new tools to diagnose disease or monitor the

PAGE 18

18 success of nutritional, pharmacological, and th erapeutic interventions. Metabolomics can also provide new biomarkers to assess human healt h, and over time a powerful list of diagnostic markers will likely be discovered, which can be measured using high-throughput assays. However, in order to proceed from the single biomarker concept to the global metabolome evaluation outside a research environment, the technology has yet to be developed to provide the clinician with the tools to assess entire wide clas ses of metabolites in biofluids and automatically process the data to evaluate the biochemical status of an individual. Many technical and methodological issues have to be addressed to cr eate analytical platforms that readily answer biological questions efficiently. In general, for every type of MS-based me tabolomics experiment the following steps need to be addressed during method development and validation. Sampling. Sample acquisition is driven by the ex perimental design and the experiment type. A sufficient number of samples are required in order to reduce the influence of biological variability. In particular when studying human samples, the influences of gender, diet, age, and genetic factors have to be considered. In addition, representative quality control samples (replicates) and blanks have to be analyzed. Sample preparation The main goal of the sample prepar ation procedure is to extract the analytes from the complex biological matrices such as serum, plasma, whole blood, urine, tissue, saliva, cell pellets, etc. ; and bring them into a format that it is compatible with the analytical technique used while removing matrix components that will interfere with the analysis. In terms of metabolic ta rget analysis the sample prepar ation procedure can be tailored for the target metabolites, because the analyt es are known and surrogate compounds or stable isotope-labeled standards can be utilized to optimize the extraction procedure and matrix

PAGE 19

19 removal. On the other hand, if a global metabolom ic analysis is pursued, the sample preparation step should be as simple and universal as possi ble. Sample preparation and sample introduction methods can include direct in jection, liquid-liquid extraction (LLE), solid-phase extraction (SPE), supercritical fluid ex traction, accelerated solvent ex traction, microwave-assisted extraction, protein precipita tion, and membrane methods, such as dialysis and ultracentrifugation. Sample analysis. This point has already been discu ssed in the Analytical Instrument Platforms section, vide supra. Fu rther details for the applied analyt ical strategies will be given later in the thesis. Data processing and analysis. Besides the technological asp ects, data processing is an important factor for making sense of the results and th e steps taken depend on what question was initially asked (hypothesis approach). In all cases, statis tical analysis must be performed in order to ensure analytical ri gor [23]. How this is done remains a debate, since different methods could give varying results and no consensus approach has been selected by the research community. Table 1-2 shows some of the common methods used for data processing, all the steps taken before any statistical methodology is applied, and analysis or a pplication of supervised or unsupervised statistical treatments. To deal with large datasets, it is very important to carry out the correct data processing steps as batch to batch differences in data become a clear problem. This process involves feature extraction (a feat ure is a molecular entity characterized by a specific mass and retention time), alignment (e.g. mass and/or retenti on time), normalization (e.g. to adjust the intensities within each samp le run by reducing the systematic error), filtering (e.g. using the most intense peaks across the ex periments), and quality control (e.g. clustering trees). On the other hand, data analysis uses multivariate techniques to reduce the dimensionality

PAGE 20

20 of the measurement vector. This can be done by se lecting a subset of va riables directly as a representation of the total proce ss, combining (transforming) the original measurements to a new set with fewer features. The probability distribu tion is estimated and this probability density function would help in the proce ss of validation and testing the sy stem against the available data. A classification system is determined during this process and now derives rules that allocate new observations to pre-defined categories. Different methods include non-linear mapping (NLM), hierarchical clustering analysis (HCA), principal component analysis (PCA), k-means clustering and self-organization maps (SOM), independent component analysis (ICA), partial least squares-discriminant analysis (PLSDA), multilevel component analysis (MCA); probabilistic neural networks (PNNs), artifici al neural networks (ANNs), and orthogonal projection on latent structure-discriminant analysis (O-P LS-DA), among others. Once discriminant analysis is done, statistical significance can be determined by applying classical statistics like Students t-test, analysis of variance (ANOVA), or multiple analyses of variance (MANOVA). Another way of dealin g with data after the processing step is to detect for significant correlations of components, which is based on the covariance and/or correlations within a data matrix. Specific details of the applied data processing and statistical tools are given in Chapter 3. Identification and biological interpretation The identification of metabolites remains a major bottleneck for metabolomics. The unamb iguous identification of a differentiating biomarker(s) is important from a biological point of view. A careful scrutiny on the biochemistry of differentiating metabolites is an essential part of the biomarke r discovery process [24]. For the proper assignment of metabolites, whether in terms of identi ty or models of networks, bioinformatic tools and data base s are required [2]. There are few tools that can automatically

PAGE 21

21 produce a list of possible metabolites from the mass signals at a part icular retention time (MS) or chemical shift (NMR). In fact, the connection be tween experimental data (MS and NMR spectra, retention time, fragmentation pa ttern, chemical shift, coupling constant) and the available chemical databases (Table 1-3) is still weak, let alone automatic. In summary, many technical a nd methodological issues need to be addressed to create analytical platforms that readily answer biologi cal questions efficiently. In this sense, data analysis and visualization tools, libraries, and databases for metabolomics have yet to be developed. A major obstacle in this type of inve stigation is the high diversity and variability encountered in the metabolome. The physico-chemi cal properties of metabolites are so diverse that none of the currently avai lable techniques can analyze all metabolites simultaneously. In addition, the quantification of the metabolites enc ountered is a challenge because signal intensity is not only a function of concentration or mass bu t also depends on the chemical structure of the analytes and can be influenced by matrix interf erence (especially when electrospray ionization associated with mass spectrometry is used). On the other hand, the concentration of the components in the metabolome varies over a trem endous dynamic range, which can be as high as 100,000. Although metabolomics is still in its infanc y and the framework is still developing, the greater synergy between organisms will provide a much clearer picture of the function of cells, organs, and organisms, bringing us closer to understanding thei r roles in nature. The principle aim of this study is to contribute to the develo pment of global and targeted metabolomic strategies for the evaluation and id entification of metabolomic changes in human and animal biological samples. Therefore, this work focuses on the development of chromatographic approaches for the separation of polar and nonpolar compounds, their coupling to mass spectrometric detection, and application to real metabolomic studies.

PAGE 22

22 The remainder of Chapter 1 presents an overview of mass spectrometry, focusing on the time-of-flight mass spectrometer, a brief descripti on of the type of ionization applied for this projectelectrospray ionizati on, some general principles of liquid chromatographyfocusing on C18-monolithic and hydrophilic interaction liquid chromatographic approaches, a succinct explanation of the metabolic ro le of the targeted metabolites, and, finally, the scope of the present work. Time-of-Flight Mass Spectrometer Background In a tim e-of-flight (TOF) analyzer, ions are separated on the basis of differences in their velocities as they move in a straight path toward a co llector. The time-of-flight mass spectrometer (TOF-MS) is fast, capable of high re solving power and high accuracy, applicable to chromatographic separation, and it is used fo r the mass determination of small and large biomolecules [25]. William E. Stephens of the Un iversity of Pennsylvania proposed the concept of TOF-MS in 1946 [26]. Two year s later, in 1948, Cameron and Eggers [27] reported the first time-of-flight mass spectra, demonstrating a mass resolution of around 5. In 1955, Wiley and McLaren [28] developed a technique to focus th e spread of ionization position and initial ion energy, achieving a mass resolving power of 300 or higher. This modification provided TOF-MS with a practical mass resolving power, and th e Bendix Corporation be gan distribution of commercial models [29]. In the 1970s, a reflectron TOF-MS [ 30] was developed, featuring a mass resolving power of a few thousand. In the late 1980s, Dawson and Guilhaus [31], and Dodonov [32] developed orthogonal acceleration (oa), which allowed an efficient combination of TOF-MS and continu ous ionization sources.

PAGE 23

23 Orthogonal Acceleration Time-of-Flight Mass Spectrometer In the basic structure of an orthogonal acc eleration tim e-of-flight mass spectrometer ( oaTOF-MS) the sample is continuously ionized by the ion source. Then, ions from the ion source pass into the flight tube, which is perpendicular to the incident beam direction. The perpendicular orientation of the source with respect to the flig ht tube reduces the kine tic energy dispersion in the direction of the flight tube improving the ma ss resolving power [33]. To initiate a pulse of ions into the TOF analyzer an electrostatic fi eld is created in nanosec onds (typically) and as a consequence, the ions in this field experience a force. The direction of the force is strictly orthogonal to the ion beam axis. The resulting orthogonal acceleration imparts a new component of velocity to the sampled ions and this com ponent is vectorially independent of the axial velocity of the ion beam. Vectorial decoupling of the velocity of ions in the ion beam and TOF directions is an im portant feature of oa-TOF-MS. Figure 1-2 shows a schematic of an oa-TOFMS. The independence of the TOF and source axes l eads to advantages such as reduction of the spread of the velocity distribu tion in the TOF analysis direct ion and the orthogonal acceleration region fill time can be matched to the longest flight of the ions in the TOF analysis region insuring that there are no ions wasted and thereby maximizing sensitivity. Electrospray Ionization Electrospray ionization (ESI) is a simple and elegant m ethod that handles big molecules, operates at atmospheric pressure and at a moderate temperature, and is probably the most gentle ionization technique available for MS [34]. It has also become the most successful interface for LC/MS and CE/MS applications [1, 9]. Although th e concept of electrosp ray was put forward by Malcolm Dole in 1968 [35], the development of ESIMS is credited to John Fenn [36-39], who was awarded the 2002 Nobel Prize in Chemistry fo r that contribution. Elect rospray applications account for most of the activity in oa -TOF-MS [40]. The coupling of electrospray with oa-TOF-

PAGE 24

24 MS has important advantages: dutycycle-related sensitivity, speed to facilitate MS and MS/MS with online liquid chromatography, high m/z cap ability, and excellent mass accuracy [41]. As the name implies, electrospray ionization is a process that produces a fine spray of highly charged droplets under the influence of an intense elect ric field. Evaporation of the solvent converts those charged dr oplets into gas-phase ions. The mechanism of ESI is a highly debated topic. It is generally believed that ionization in electr ospray involves three different processes: droplet formation, droplet shrinkage, and desorption of gaseous ions [42]. Advantages of ESI are as follows: analysis of compounds with a molecular weight up to about 310,000 Da is possible [43]; ESI is very se nsitive; typical detectio n limits range from low attomole to picomole levels [44]; ESI is a very mild ionization techniqu e; usually only sample molecules carrying multiple protons are generate d. It is possible to obs erve native biological complexes bound by noncovalent interactions. Fragmentation can be induced by increasing the potential difference in the i on source and on-line coupling with liquid chromatography or capillary electrophoresis (CE) equipment is possi ble because ESI is a solvent-based ionization technique. Disadvantages of ESI are as follows: the pres ence of salts, buffers, detergents, and other additives reduces the sensitivity dramatically. If buffers are needed, volatile buffers, such as ammonium acetate or ammonium fo rmate are preferred; and analysis of mixtures is difficult, because each compound gives rise to several si gnals corresponding to sample molecules carrying a range of protons [45]. In addition, signal suppre ssion (the signal intensity of an analyte ion may decrease)/enhancement (the signal intensity of an analyte ion ma y increase) is common in ESI. Signal suppression/enhancement originates from a competition between an analyte and coeluting species (both with the same charge) for placement in the surface excess charge layer

PAGE 25

25 which ultimately transfers into the gas phase. Ion suppression or enhancement affects precision and accuracy of an assay and it is most commonly induced by matrix components. Liquid Chromatography Although chrom atography has been around for more than a century and column chromatography for about half a century, new colu mns, new chemistry, new materials as well as new instrumentations are continuously introduced [11]. Highly efficient separations are the cornerstone of many analytical methods that d eal with extremely complex mixtures. These new analytical tools need to be fast and of sufficient resolving power to detect minute qualitative and quantitative changes in a metabolic profile. For the separation and identific ation of metabolites GC/MS, LC/MS, or CE/MS techniques have been primarily employed. Among these techniques, HPLC/MS is most widely applicable to metabolomics. HPLC separations are better suited for the analysis of labile and high molecular weight compounds and for the analysis of nonvol atile polar compounds in their natural form. However, since metabolomics deals with a diversity of small molecules, there is no single ideal chromatographic approach that can be applied to all classes of metabolites. We have found that two methods: one for nonpolar compounds and one for polar compounds (described below)-, provide a reasonable br eadth of coverage. Monolithic Columns High perform ance liquid chromatography has tr aditionally been performed in columns packed with 3 or 5 m particle diameters. The internal di ameter of these columns is typically between 2 and 4.6 mm, although smaller column diameters are gaining popularity. High separation efficiency with a concomitant reduc tion in analysis time is achieved by reducing particle size [46]. Chromatograp hic packing materials with di ameters in the range of 1-2 m are

PAGE 26

26 now commercially available and known as st ationary phases for ultra-high pressure (performance) liquid chromatography. These sma ller particles dramatically increase column efficiency, which in turn increases mass sensiti vity, chromatographic re solution, and speed [47]. However, columns packed with such particles need a special LC system. The injection volume should be small as well as the volume of the dete ctor cell, in case UV det ection is applied [48]. In addition, smaller column particle size results in higher system pressures and requires components, such as pumps, with high pr essure ratings (2,000 bar or 30,000 psi). Emerging technology in separation sciences is enabling high separa tion efficiency and speed of analysis, surpassing th e conventional particle-packed columns in HPLC. This includes the use of monolithic columns [48-50]. As compared to packed particle columns, m onolithic columns consist of a single piece of porous cross-linked polymer or porous silica. Mono liths are made in different formats as porous rods, generated in thin capillarie s or made as thin membranes or disks. As a result of their internal design, all the mobile phase flows through the sta tionary phase. The convective flow, which is the typical driving force for mass transfer in this type of columns, enables a substantial increase in the speed of separation [51]. Monolit hic columns have high column permeability and small-sized skeletons that decrease the diffusion path length of molecules in the stationary phase, resulting in a reduced contribu tion to band broadening [52]; thus van Deemter curves for some monolithic columns are much flatter at high flow rates compared to conventional columns. These features allow the operation of monolithic columns at very high flow rates for fast separations with no significant loss in efficiency [52-55]. Based on the nature of their construction materials, monolithic columns can be classified as organic polymeror silica-based columns [49]. The organic polymer-based monoliths have

PAGE 27

27 always done a better job of separating larger molecules; while silica-based monolithic columns enable fast separations of smaller molecules [49]. Monolithic silica columns were introduced much later than polymer monoliths. After conventional-size columns became commercially available in 2000, the publication rate on mono lithic silica columns exceeded that on polymer monoliths [56]. In this research, a reversed-phase C18 silica-based monolithic column was used for the separation of the nonpolar metabolites in biological samples. Thus, the discussion will be briefly focused on this particul ar type of column material. Monolithic silica columns : these columns contain a tailor-made bimodal pore structure with both macropores or through pores and mesopores [57]. The large macropores are responsible for a low flow resistance, and therefore allow the a pplication of high eluent flow rate, while the small pores ensure sufficient surface area (300 m2/g approximately) for separation efficiency [57]. One of the impor tant features of these columns is their high permeability, which is nearly twice as high as that of packed colu mns [58]. Therefore, monolithic silica columns can be operated at high flow rates of up to 10 mL min-1, thus allowing fast separations of various mixtures [59]. Monolithic silica columns are suitable for high throughput analysis, e.g. metabolomics analysis, as well as for two-dimensional HPLC methods [56, 60]. HILIC Columns Reversed-phase m ode is most often em ployed in HPLC, where chemically bonded stationary phases (C8, C18, C30, etc.) have advantages in rapid equilibration with mobile phase, and high separation efficiency and high reproducibility in gradient separations, based on the hydrophobic properties of the stationary phases. However, the highest concentrations for endogenous metabolites in body fluids are found for very polar, small molecules [61, 62]; these compounds exhibit poor retention on reversed-phase columns, resulting in co-e lution of many

PAGE 28

28 polar compounds [63, 64]. These compounds are eluting early in the chromatographic run, under highly aqueous mobile phase conditions and without sufficient compound separation, and ionization efficiency of ESI is often poor [65]. The retention and separati on of polar compounds is an on-going challenge for chromatographers [66-67]. Ion exchange or ion pairing, mobile phase pH manipulation, and reversed-phase chromatography with specially designed columns are techni ques that traditionally have been used for retention of polar analytes However, each of these techniques has certain drawbacks. Ion exchange or ion pairing work well onl y if the analytes of interest are ionizable. In addition, ion pairing is difficult with mass spectro metry because ion pairi ng reagents will cause signal suppression. Manipulation of mobile phase pH is a technique that also works for ionizable compounds, because the retention characteristics of ionizable compounds are a function of pH of the mobile phase; however, manipulation of mobi le phase pH is not always successful in retaining analytes that are very polar. Also some compounds might not be stable outside a narrow pH range. Reversed-phase chromatography is versatile and able to retain and resolve many classes of compounds; however, the retenti on of polar analytes often requires a highly aqueous mobile phase to achieve retention, wh ich can cause a number of issues such as dewetting of the stationary phase. In addition, the highly aqueous mobile phases that are required for polar retention are not ideal for mobile phase desolvation by ESI-MS, and thus result in poor sensitivity [68]. Hydrophilic interaction chromatography is a mode of chromatography that can address these issues. HILIC was first introduced by Al pert for the separation (amongst other polar compounds) of amino acids, which were eluted in opposite order to that found in reversed-phase chromatography [69]. HILIC is similar to normal phase liquid chromatogra phy in that elution is

PAGE 29

29 promoted by the use of polar modifiers (e.g. H2O) in less polar mobile phases (e.g. acetonitrile). However, HILIC is unique in that the presence of water in the mobile phase is crucial for the establishment of a stagnant enriched aqueous la yer on the surface of the stationary phase into which analytes may selectively partition. Alpert considered that dipole-dipole interactions (hydrogen bonds) might contribute to partitioning into the stationary phase layer. Hydrogen bonding, especially when using mobile phases of low water content, has also been proposed as a retention mechanism [70]. Clearly, the mechanism of HILIC separations is complex and different phenomena contribute to various degrees. Despite the complexity of the mechanism, HILI C is simple and its general advantages can be summarized as follow: reasonable peak shapes obtained for bases; mass spectrometer sensitivity is enhanced due to the high orga nic content in the mob ile phase and the high efficiency of spraying and desolv ation techniques; direct injecti on can often be made of extracts eluted from C18 columns with solvents of high organic co ntent; the orde r of elution is generally the opposite of that found in reversed phase sepa rations, giving useful al ternative selectivity; good retention of polar compounds is obtained in HILIC, wher eas poor retention is often obtained in reversed-phase chromatography; and higher flow rates are po ssible due to the high organic content of typi cal mobile phases [70]. The HILIC separations of proteins [ 69], peptides [71-73], amino acids [69], oligonucleotides [69], carbohydrat es [74], histones [75], and natural products [76, 77], among others compounds [78], have been reported. The us e of HILIC chromatography as a tool for the evaluation of polar compounds in different metabolomic applications is increasing in popularity and acceptance [63, 76, 79-81]. One of th e aims of this study has been to illustrate the utility of HILIC for global and targeted metabolomics.

PAGE 30

30 Target Compounds Carnitine and Acylcarnitines Carnitine and acylcarnitines are endogenous m etabolites present in most mammalian tissues. Carnitine, (3-hydroxy-4-( N ,N ,N -trimethylammonio) butanoate) is a small, water soluble, quaternary nitrogen-conta ining compound (see Figure 1-3) that is involved in the transport of activated fatty acids into the mitochondrial matrix, where they are metabolized via -oxidation [82]. Fatty acid metabolism occurs in the mito chondrial matrix; however, the mitochondrial inner membrane is impermeable to fatty acids. Thus, carnitine plays a vital transport role in fatty acid metabolism and in cellular energy production. In mammals, carnitine functions through the re versible esterification of its 3-hydroxyl group, with subsequent transloc ation of the acylcarnitines produced from one cellular compartment to another. Carn itine acyltransferases are th e enzymes responsible for the production of acylcarnitines, which can have di ffering chain-lengths, commonly designated as short-, medium-, and long-chain length; and wh ich vary by their cellula r location and metabolic functions. Carnitine is also implicated in the maintena nce of the cellular pool of free coenzyme A (CoA) and in the elimination of pot entially toxic acyl-CoA, originating from exposure to xenobiotics and/or from blockage of metabolic pathways. In contrast to acyl-CoAs, the corresponding acylcarnitines can be excreted via the urine [82, 83]. In healthy subjects, carnitine and acetylcarnitin e are the dominant concentration carnitines of the body fluid and tissue carnitine pools [82, 84]. Skeletal muscle contains >95% of the total carnitine body stores, and the tissue concentrations are considerably higher than the concentration in plasma; transport systems ensure its widesp read distribution from sites of absorption and synthesis throughout the body. The kidneys play a crucial role in car nitine homeostasis, since

PAGE 31

31 they reabsorb >90% of the filtered carnitine [82], so that the plasma levels of free carnitine are maintained at 30 mol L-1 [85, 86]. Reported plasma concentrations in healthy humans are 29-50 mol L-1 for free carnitine, 2.5-8.6 mol L-1 for acetylcarnitine, 0.18-0.6 mol L-1 for propionylcarnitine, 0.03-0.17 mol L-1 for isovalerylcarnitine, 0.02-0.05 mol L-1 for hexanoylcarnitine, 0.01-0.13 mol L-1for octanoylcarnitine, and 2.2-4.9 mol L-1 for the longchain acylcarnitines [85]. The rela tive amounts of acylcarnitines are often expressed as a ratio of acylcarnitine to free carnitine. In plasma sample s, a ratio greater than 0.4 is indicative of a carnitine deficiency [87]. In many metabolic disorders, carnitine meta bolism is greatly disturbed, leading to a redistribution of the carnitine and acylcarnitine pools. The determination of individual acylcarnitines in biological flui ds is a powerful means to dia gnose and monitor these disorders [88-91]. Amino Acids An a mino acid is a molecule containing both amine and carboxyl functional groups. In the alpha am ino acids (H2NCHRCOOH, where R is an organic substituent), the amino and carboxylate groups are both attached to the carbon. The various alpha am ino acids differ in which side chain (R group) is a ttached to their alpha carbon. Twenty standard amino acids are used by cells in protein biosynthesis, and these are specified by the general genetic code. These 20 amino acids are biosynthesized from other molecules, but organisms differ in which ones they can synthesize and which ones m ust be provided in their diet. The eight amino acids (isoleucine, leucine, lysine, methionine, phenylalanine, threonine, tryptophan, and valine) that cannot be synthesized by an organism are known as essential amino acids and must be obtaine d from food. However, the essentiality depends on many factors, e.g. cysteine, taurine, tyrosine, histidine and arginine are semi-essential

PAGE 32

32 amino acids in children because the metabolic pathways that synthesize these amino acids are not fully developed. The amounts required also depend on the age and health of the individual [92, 93]. Depending on the polarity of the side chain, amino acids vary in their hydrophilic or hydrophobic character [94]. These properties are im portant in protein structure and proteinprotein interactions. The im portance of the physical properties of the side chains comes from the influence this has on the amino acid residues inte ractions with other structures, both within a single protein and between pr oteins. The distribution of hydrophilic and hydrophobic amino acids determines the tertiary structure of the protein, and their physical location on the outside structure of the proteins influences their quaternary structure. The anal ytical separation of a mino acids is also influenced by their polar or nonpolar character. Chemical structures of carnitine, acylcarnitines, and amino acids are shown in Figure 1-3. This research focused on the application of liquid chromatography with mass spectrometry to global and targeted metabolomic studies. De tailed description of the development and optimization of novel chromatographic approaches and their direct application to biological samples (plasma from animal and human sources) is presented in the following chapters. In the case of target metabolomic studies, special attent ion to carnitine, acylcarnitines (short-, medium-, and long-chain) and amino acids wa s given. Commercially available C18-monolithicand hydrophilic interaction liquid chromatography-colu mns for the separation of nonpolar and polar analytes were used. In addition, an oa -TOF mass spectrometer was util ized for the accurate mass analyses of the mentioned metabolites. Data were acquired in positive electrospray ionization mode.

PAGE 33

33 There are two emerging column technologies that have attracted much attention in recent years, one is the monolithic column and the ot her is the use of hydrophilic interaction liquid chromatography. Herein, we explore these two chromatographic approaches. The analytical conditions for the development of a reversed-phase chromatographic strategy using two in-series C18-monolithic columns for the efficient separa tion of nonpolar compounds is presented in Chapter 2. In addition, the steps corresponding to the optimization of a HILIC separation methodology for the chromatographic resolu tion of polar compounds are presented. A comparison of the selectivity of both chromatogr aphic methods and application to real samples are detailed at the end of Chapter 2 as well. The application of the optimized LC/MS parame ters to the human plasma samples before and after a ketogenic diet therapy for global and targeted metabolomic studies is discussed in Chapter 3. The application of the same LC/MS methodolog ies to piglet plasma samples to evaluate metabolome changes during days two to ei ght of life is described in Chapter 4. Finally, Chapter 5 summarizes the work presented here, as well as some future research directions.

PAGE 34

34 Table 1-1. Comparison of metabolic techno logies (adapted from reference [7]) Criteria NMR GC/MS LC/MS Sensitivity Limit of detection Poor 10-6 mol Good 10-12 mol Excellent 10-15 mol Metabolites detected High concentration of organic compounds in a solution Volatile, nonpolar compounds Nonvolatile compounds in a solution Robustness Good Reasonable Reasonable Speed Rapid Depending on chromatography Depending on chromatography Quantification Good Poor, standards needed Poor, standards needed Identification By chemical shift calibration By mass, fragmentation, and retention time By mass, fragmentation, and retention time Problems Peak overlap Volatility of metabolites Ionization of metabolites Disadvantages -Exact adjustment of pH required after metabolite extraction -Different methods can circumvent the need for extraction, but result in a further loss of sensitivity -Nonvolatile compounds have to be derivatized -Some metabolites cannot be made volatile even with derivatization -Problem with ion supression and adduct formation -Different metabolites detected in positive and negative scanning mode -Lack of libraries for metabolite identification

PAGE 35

35 Table 1-2. Methods for data processing and analysis (ada pted from reference [2]) 1ANOVA: analysis of variance; 2MANOVA: multiple anal ysis of variance; 3PCA: principal component analysis; 4ICA: independent component analysis; 5HCA: hierarchical clustering analysis; 6ANNs: artificial neural networks; 7PNNs: probabilistic neural networks Data processing methods Data analysis methods Normalization Coefficient of variation Baseline correction, peak shif ting, and noise removal ANOVA1 or MANOVA2 Missing value correction PCA3, ICA4, and subtypes Deconvolution of peak Clustering-HCA5, k-means Data reduction Self organizing maps Limited data analysis to speci fied representative region of data Fisher discriminant analysis Exclude variables or sample outlie rs Partial least squares (PLS) Neural networks (ANNs6, PNNs7) Genetic programming and algorithms

PAGE 36

36 Table 1-3. MS, NMR, and metabolic pathway databases (DB) Name Source/URL MS-based databases Golm Metabolome Database (GMDB@CSB.DB) Max Planck Institute of Molecular Plant Physiology Human Metabolome Database (HMDB) Genome Alberta and Genome Canada KNApSAcK (Comprehensive Species Metabolite Relationship Database) Nara Institute of Science and Technology (NAIST) Metlin The Scripps Research Institute NIST/EPA/NIH Mass Spectral Library (NIST 0. 5) National Institute of Standards and Technology (NIST) SpecInfo Daresbury Laboratory Spectral Database for Organic Compounds (SDBS) National Institute of Advanced Industrial Science and Technology (AIST) NMR-based databases ACD Databases Advanced Chem istry Development, Inc. Human Metabolome Database (HMDB) Genome Alberta and Genome Canada NMRShiftDB University of Koeln SpecInfo Daresbury Laboratory Spectral Database for Organic Compounds (SDBS) National Institute of Advanced Industrial Science and Technology (AIST) Standard Compounds on Biological Magnetic Resonance Bank (BMRB) University of Wisconsin Metabolic pathway databases BRENDA BRaunschweig Enzyme Database (http://www.brenda.uni-koeln.de/) Chemical Entities of Biological Inte rest (ChEBI) http://www.ebi.ac.uk/chebi/ HumanCyc Encyclopedia of Human Metabolic Pathways (http://humancyc.org/) KEGG Kyoto Encyclopedia of Genes and Genomes (http://www.genome,jp/kegg/) Lipid Maps http://www.lipidmaps.org/ MetaCyc Encyclopedia of Metabolic Pathways (http://metacyc.org/) Nicholsons Metabolic Minimaps h ttp://www.tcd.ie/Biochemistry/IUBMBNicholson/ PUMA2 Evolutionary Analysis of Metabolism (http://compbio.mcs.anl.gov/puma2/cgibin/index.cgi) Reactome A Curated Knowledgebase of Pathways (http://www.reactome.org/) Roche Applied Sciences Biochemical Pathways Chart http://www.expasy .org/cgi-bin/searchbiochem-index

PAGE 37

37 Figure 1-1. Representation of the number of me tabolomic publications per year from 1998 to 2008 (obtained by searching metabolomic s using SciFinder, Version 2007, (American Chemical Society)).

PAGE 38

38 Figure 1-2. Main components of typical oa-TOF system with reflecting mass analyzer. The ion beam enters from a source at the left (ion formation region) and it is accelerated to enter the orthogonal accelerator (oa or ion pulser in this figure). The beam optics make the beam more parallel before it enters the oa. The beam fills the first stage of the oa until a bipolar push-out pulse pair is applied. A packet of ions is thus sampled and accelerated through grids to enter the drif t region. Conventional reflecting TOF optics are used to bring the ions to a sp ace-time focus on the detector. During the time that the ions are in the drift-region (and i on mirror) the oa is refilled with new beam. Figure adapted from www.chem.agilent.com.

PAGE 39

39 Figure 1-3. Chemical structures of free car nitine, acylcarnitines, and amino acids. The R1 group in the acylcarnitine stru cture is esterified by di fferent chain lengths giving place to acylcarnitines with different chain lengths. The R2 group in the amino acid structure determines its identity. Free carnitine N+CH3CH3CH3OHO OH Acylcarnitines N+CH3CH3CH3OO OH R1 Amino acids N+H H H R2O OH

PAGE 40

40 CHAPTER 2 DEVELOPMENT AND COMPARISON OF TW O CHR OMATOGRAPHIC APPROACHES Liquid Chromatography In liqu id chromatography analytes are separate d by virtue of differing partition coefficients between a liquid mobile phase and a solid st ationary phase [95]. High-performance liquid chromatography (HPLC) is one mode of chromatography in which a liquid mobile phase is used to separate the analytes. Differe nt steps take place during this separation process: the components are first dissolved in a solvent; they are forced to flow thr ough a chromatographic column under a high pressure; and once in the column, the mixtur e is resolved into its components, which elute from the column sequentially over time The interaction between the components and the column stationary phase directly determines the chromatographic resolution. The interaction of the solute with mobile and stationary phases can be manipulated through diffe rent choices of both solvents and stationary phases. As a result, HPLC acquires a high degree of versatility not found in other chromatographic systems and it has the ability to easily separate a wide variety of chemical mixtures [96, 97]. HPLC is characterized by small diameter (2 5 mm), reusable stainless steel columns, column packings with very small (1, 2, 3, 5 and 10 m) particles and the continual development of new substances to be used as stationary pha ses, relatively high inlet pressures and controlled flow of the mobile phase, precis e sample introduction without the need for large samples, special continuous flow detectors capable of handling small flow rates and detecting very small amounts, automated standardized instruments, rapid analysis, and high chromatographic resolution [95-97].

PAGE 41

41 The high performance obtained with these chroma tographic systems is the result of many factors: very small particles of narrow distribution range and uni form pore size and distribution, high pressure column slurry packing techniques, accurate low volume sample injectors, sensitive low volume detectors and, of cour se, good pumping systems. Naturally, pressure is needed to permit a given flow rate of the mobile phase; otherwise, pressure is a negative factor not contributing to the improvement in separation [95-97]. Today there is a competition be tween two means for fast LC analysis, namely, HPLC with monolithic phases and small particle phases (< 2 m) used in ultra-high pressure liquid chromatography. With both phase types a substantially faster analysis, reduced solvent consumption, and increased mass sensitivity can be achieved. It is important to understand the basic princi ples of a rapid reso lution system; however, since the chromatographic technologies used in this work are not rela ted to UHPLC, a further discussion will be avoided. Attention will in stead be focused on monolithic and hydrophilic interaction chromatography columns. Monolithic stationary phases are continuous separation media in the format that can be compared to a single large par ticle that does not contain inte rparticle voids. This type of material exhibits dual pore arrays named throughpores and mesopores. The throughpores are responsible for the free flow of the mobile pha se through the column and the mesopores for the retention performance. Therefore, the efficiency of this column is controlled by the average sizes of the throughpores and mesopores. The major advantage of this new approach is that we might be able to choose these two sizes independently of each other, but it is still the beginning of a long period of research endeavor s during which the immense number of possible approaches to design and preparation of these monolithic columns will be optimized.

PAGE 42

42 Hydrophilic interaction liquid ch romatography is a technique su itable for the separation of very polar and hydrophilic compounds. Therefore, solutes that have little or no retention on reversed-phase columns genera lly experience strong retention on HILIC columns. HILIC is a variation of normal-phase chromatography without the disadvantages of using solvents that are immiscible with water. Besides, HILIC enhances the chromatographic retention and separation of those analytes which are poorly retained by re versed-phase chromatography. In this type of column, the stationary phase consists of a polar material (e.g. silica, cyano, diol, amino, amide, polymeric coatings, etc.), the mobile phase is highly organic (> 80%) with a small amount of aqueous/polar solvent, and the order of elution goes from least to most polar. Experimental Mass Spectrometer Positive ES I [(+)ESI] spectra were collected on an Agilent 6210 Time-of-Flight mass spectrometer (Agilent Technologies, Inc., Santa Clara, California) configured with a dual nebulizer electrospray source. With this source, ions were simultaneously generated from the LC stream by one capillary nebuliz er and from mass calibrant solution, via a second capillary nebulizer. Data acquisition and processing we re conducted using the Agilent MassHunter Workstation software. TOF Calibration An Agilent Technologies ES Tuning Mix (P /N G2421A) containing betaine (CAS #: 10743-7, MW : 117.1), hexamethoxyphosphazene (CAS #: 957-13-1, MW: 321.0), hexakis(2,2difluoroethoxy)phosphazene (CAS #: 186817-57-2, MW: 621.0), hexakis(1H, 1H, 3Htetrafluoropropoxy)phosphazene (CAS #: 58943-98-9, MW: 921.9), hexakis(1H, 1H, 5Hoctafluoropentoxy)phosphazene (CAS #: 1605916-8, MW: 1521.0), hexakis(1H, 1H, 7Hdodecafluoroheptoxy)phosphazene (CAS #: 383074-8, MW: 2121.4), and hexakis(1H, 1H, 9H-

PAGE 43

43 perfluorononyloxy)phosphazene (CAS #: 186043-67-4, MW: 2720.1), provided mass reference ions for a baseline calibra tion of the mass axis. HPLC System An Agilent 1200 Series Rapi d Resolution LC system equipped with a binary pump provided solvent delivery For HILIC chromatography, a Luna HILIC column (Phenomenex, Torrance, California; 2. 0 mm 150 mm, with 3.0 m particle diameter) was employed. In order to improve chromatographic resolution, two in-series C18-monolithic columns (Phenomenex Onyx, each column was 100 mm 4.6 mm) were used fo r conventional reversed-phase chromatographic separation. Standard Solutions Laborato ry and commercial-made standards for carnitine, acylcarnitines, and amino acids were used for all the experiments. Standard stock solutions for each analyte were prepared in methanol and were stored at -4C. Standard mixes were freshly prepared on the day of analysis by mixing known volumes of the different standard stock solutions to creat e a final solution with 0.5 mg L-1 of each analyte. Whenever necessary, dilutions of these stock solutions were used. LC-MS grade acetonitrile (ACN), HPLC grade water, methanol, and isopropanol were obtained from Burdick and Jackson (Muskegon, Michigan). Acetic acid (glacial, TraceMetal grade), HPLC grade ammonium acetate, and cer tified ammonium formate were obtained from Fisher Chemical (Fisher Scientific Inc., Pitts burgh, Pennsylvania). Formic acid was purchased from Acros Organics (Fisher Scientific). Mobile Phase Preparation For HILIC chrom atography, the mobile phase was prepared by first dissolving a known amount of ammonium acetate or ammonium format e in water (stock solu tion), and then mixing with the desired volume of acetonitrile. The salt concentration in the te xt and figure captions

PAGE 44

44 refers to the final concentration in the mobile phase. The pH of the salt solution was not adjusted. The mobile phases A and B for C18-monolithic chromatography were prepared by adding 1% (v/v) of acetic acid to water and acetonitrile. Plasma Sample Preparation Piglet p lasma samples were treated acco rding to the standard operating procedure established to provide standards that are used in the Metaboli c Assessment Laboratory (MAL) as required by section 58.81 of the Good Laboratory Practice for the Non-clini cal Laboratory [98]. Frozen plasma was gently thaw ed with swirling to maintain cold temperature throughout the sample. Plasma (100 L) was pipetted into 1 mL of acetonitrile:methanol (3:1), mixed vigorously and frozen overnight The plasma/solvent samples we re thawed, mixed, centrifuged for 15 min at 12,000 rpm, the supernatant collecte d and the pellet discarded. The supernatant was centrifuged a second and a third time and evaporated to dryness using nitrogen gas. The sample replicates were all assayed together in order to reduce error. In order to solubilize polar and less polar compounds, the pellets resulting from the preparation of the plasma samples were dissolved in 100 L solvent composed of 10.0 L of water, 10.0 L of isopropanol, 40.0 L of methanol, and 40.0 L of acetonitrile. Blank soluti ons were prepared in the same way. It is important to note that no deriva tization procedure was used before or after column separation. This fact differentiates this procedure from othe r approaches reported in the literature for the separation and/or determination of carnitine and carnitin e-based compounds in clinical samples [99]. MS and Chromatographic Conditions The electrospray source was operated in posit ive ionization mode. The capillary was set to -4.0 kV, the nebulizer gas was operated at 50 psi, and the drying gas was set to 10 L min-1 at a temperature of 300 C. The capi llary exit was set to 180 V with skimmer set to 60 V. The

PAGE 45

45 octopole offset voltage was set to 250 V. The TOF provided resolving power and mass accuracy (less than three ppm) with high sensitivity, scanning time (40 full scan spectra from m/z 50 to 1650 per second), and dynamic range (3-4 orders of magnitude.) Signal was acquired using a fast analog-to-digital converter (ADC); the ADC did not require any dead time corrections and was configured to record up to 10,000 transients per second. For HILIC chromatography, the mobile phase A was 7.5 mM ammonium formate in water and mobile phase B was 7.5 mM ammonium formate in 8:92 water: acetonitrile. HILIC chromatography was carried out wi th a flow rate of 0.3 mL min-1 and the following gradient program: 10% A held for 5 min, then increased lin early to 50% A over 20 min and then held at 50% A for 5 min. A return to the initial conditions was accomplished by 10-minute linear gradient from 50 to 10% A where it was held for additional 5 min. The column was held at a temperature of 30 C. The injection volumes for the standard mixture and plasma samples were 10 and 20 L, respectively. For C18-monolithic chromatography, the mobile phase A was 1% acetic acid in water and mobile phase B was 1% acetic acid in acetonitrile. The C18 gradient at 1.0 mL min-1 was as follows: 95% A held for 6.5 min, 19 min linear grad ient to 0% A, held at 0% A for 9.5 min. A return to the initial conditions was accomplished by 10-minute linea r gradient to 95% A where it was held for additional 5 min. The column was he ld at a temperature of 30 C. The injection volume was 20 L for the analysis of standards and plas ma samples. The optimization of these parameters has previously been reported [100]. Analytes of Interest Due to the z witterionic nature of carnitine, short-chain acylcarnitines, and amino acids, conditions were used to prom ote protonation of the carboxylic group, which resulted in the

PAGE 46

46 production of positively charged ions. Thus, the positive ESI mass spectra of these compounds were dominated by their [M+H]+ ions with minimal fragmentat ion. The pertinent characteristics of the analytes under study as well as the m/z of the [M+H]+ ions are given in Table 2-1. Results and Discussion Coupling of Two C18-Monolithic Columns It is important to note that the HPLC paramete rs, such as effect of the composition of the mobile phase, effect of mobile phase buffer, temp erature, flow rate, and gradient elution profile related to this approach were optimized previ ously during my Masters degree research work [100]. As a summary, the previous results indicated that a chro matographic gradient system composed of water and acetonitril e as organic mobile phase, wh en adding 1% acetic acid, could sharpen peak shapes and improve analytical sens itivity and resolution for the HPLC analysis of carnitine and acylcarnitine compounds (Figure 21). The optimum gradient was summarized in the MS and chromatographic conditions section. The novel aspect of this study was the coupling of two in-series C18-monolithic columns with the purpose of improving the chromatographic resolution of the compounds. Therefore, experiments were designed to evaluate the system performance using carnitine, acylcarnitines, and amino acids as target analytes (amino acids were not included in the list of target compounds in the Masters degree research work). As seen from Figure 2-1, the polarities and corresponding retention tim es of carnitine and acylcarnitines are quite different with the long-c hain acylcarnitines being much less polar than free carnitine and short-chain (< hexanoylcarnitine, C6) acylcarnitines. The non-polar character of the medium(hexanoylcarnitine decanoylca rnitine) and long-chain (> decanoylcarnitine) acylcarnitines due to their ex tended hydrocarbon chain results in strong retention on the C18functionalized stationary phase (Table 2-1). However, the short chain, highly polar

PAGE 47

47 acylcarnitines, and most of the polar am ino acids, are poorly retained on the C18 column. Thus, there is significant interest in developing a chroma tographic approach for the separation of more polar compounds. Consequently, experiments were designed to optimize the parameters that affect the retention efficiency on a HILIC column. There were three main objectives to these studies: the first was to optimize the experimental conditions to achieve an efficient se paration of the target metabolites (carnitine, acylcarnitines, and amino aci ds), the second to compare the efficiency of both chromatographic methodologies, and third to demonstrate the applicability of the methodologies to the analysis of the ta rgeted compounds in plasma samples. HILIC Method Development and Optimization Effect of acetonitrile content and change on retention HILIC often em ploys acetonitrile and water mi xtures as the mobile phase [101-103]. The choice of organic component and its concentration have a significant imp act on the retention of polar compounds on HILIC [104]. The effect of the acetonitrile content on retention under isocratic separation was investigated by changing it from 50 to 97% (v/v) in 10% increments while keeping the ammonium formate concentra tion unchanged. All of the compounds exhibited typical HILIC behaviors of incr easing retention betw een 9 and 17 minutes as the ACN content increased (Figure 2-2). A content of 90% of acet onitrile was chosen for further experiments. Effect of salt type and concentration When charged m olecules are analyzed with HILIC, the buffer is an essential component in the mobile phase [104]. The buffer type (ammon ium formate or acetate) and concentration (5, 7.5, and 10 mM) were varied for isocratic 90% a cetonitrile HILIC separation of a mixture of carnitine, acylcarnitines, and amino acids. The use of ammonium formate led to improved peak shape and increased retention times compared to ammonium acetate. The use of either buffer

PAGE 48

48 gave improved retention time, peak shape, and sensitivity compared to no buffer (Figures 2-3 and 2-4). The effect of the ammonium formate concentration on retention was evaluated by varying its concentration from 5 to 10 mM in a mobile phase of 90% acetonitrile. The ammonium formate concentration could not be further in creased due to its solubility limitations in the mobile phase. The retention time of all of the compound s increased as the conc entration of ammonium formate increased from 5 to 7.5 mM; and then decreased as the concentration of the buffer increased to 10 mM (Figure 2-5 ) As a result, a 7.5 mM concentr ation of ammonium formate was selected for all further experiments. Effect of temperature: Vant Hoff plots Colum n temperature is anothe r important parameter that a ffects the retention of polar compounds in HILIC [104]. Operation at elevated temperatures or at sub-ambient temperatures has been shown to influence the retention mechan ism between the solute and the stationary phase material, often resulting in selectivity changes [105]. A Vant Hoff plot of the natural logarithm of the retention factor, ln k, which is expressed as: where tr and t0 are the retention times of the solute and the void time; vs. the reciprocal temperature, 1/T, is often obtained when operating at various temperatures, with a non-linear plot indicating a shift in the retent ion mechanism involved [104]. Vant Hoff plots for carnitine, short-chain acy lcarnitines and some model amino acids on the HILIC column over the temperature interval 20 C, are displayed in Figure 2-6. Carnitine and acylcarnitines showed decreased retention with increasing column temperature. In contrast, amino acids demonstrated increased retention at higher temperature. The largest range of (2-1)

PAGE 49

49 retention times among all of the analytes was obtained when the temperature was fixed at 30 C. This temperature was selected for further experiments. Effect of flow rate: van Deemter plots The effect of the m obile phase flow rate on th e separation of carnitine, polar acylcarnitines, and amino acids was evaluated using van Deemter pl ots [52-55]. Ten microliters of the standard sample mixture was injected onto the HILIC syst em at varying flow rates, from 0.1 to 0.5 mL min-1 with a gradient separation. The height equivalent to a theoretical plate (H) was determined using the following equations for N (number of plates) and H: where L is the column length ( m), tr is the retention time fo r the analyte (min), and W1/2 is the peak width at half height (min). The plate he ights for twelve analytes at varying flow rates are shown in Figure 2-7. A flow rate of 0.3 mL min-1 gave the smallest H values for all analytes, and was the flow rate chosen for analytical separation. Evaluation of ionization suppression Since all the am ino acids were very close in retention time, the potential for ionization suppression among them was investigated. Standa rd solutions containi ng individual compounds and a solution made of a mixtur e of the standards were analyzed without chromatographic separation. Each standard was at a concentration of 5 mg L-1, except for phenylalanine, proline, valine, and tyrosine which were each at 2.5 mg L-1. The ESI molar responses (peak area per moles injected) of each an alyte in the individual solutions were compared to the values in the (2-2) (2-3)

PAGE 50

50 standard mixture; at the level of concen tration under evaluati on, no ion suppression was observed. Repeatability The intra-day precision of the proposed m ethod wa s tested with six repeated injections of standard solution mixture at the concentration level of 0.5 mg L-1. The inter-day precision was studied by analyzing 0.5 mg L-1 standard solution mixture, with six injections randomly executed in a 30-day period. For retention time and mola r response, the intra-da y relative standard deviations were below 2 and 8%, respectively. Similar results were obtained for inter-day relative standard deviations. Applicability of HILIC to the Separation of E ndogenous Metabolites in Plasma Samples Piglet plasma samples were prepared as described in the experimental section. The abovementioned optimized parameters for the separatio n and determination of carnitine, short-chain acylcarnitines and amino acids by HILIC/ESI-TOF were applied. Total ion chromatograms for both methodologies are shown in Figure 2-8. Extracted ion chroma tograms of the analytes under HILIC and C18-monolithic conditions are shown in Figur es 2-9 2-11. While the separation of the carnitine and the C2-, C3and C4acylcarnitines was much impr oved over that obtained with C18 RPLC, the peak shapes of the acylcar nitines showed some peak splitting (C2-acylcarnitine) and fronting (Figure 2-9). In contrast, most of the am ino acids gave well-shaped chromatographic peaks with good separation (F igures 2-10 and 2-11). Proline, valine, methionine, and tyrosine did show some peak splitting (Figures 2-10 and 2-11.) Specificity of HILIC vs. C18-Monolithic Chromatography The high sensitivity afforded by coupling HPLC with ESI-MS is subject to some limitations when RPLC is used: decreases in the ES I efficiency and stability as a consequence of

PAGE 51

51 polar compound elution in a highly aqueous mobile phase, and ion s uppression due to poor retention and co-elution of polar compounds. In this study, retention effici encies on both HILIC and RPLC columns were evaluated. As shown in Table 2-1, fourteen of the twenty-eig ht analytes were weak ly retained under RPLC conditions and eluted near the solvent front (3.6 min). On the HILIC column, the same compounds were strongly retained and eluted in a high organic mobile phase (~90-65%), which facilitated analyte desolvation and resulted in enhanced (+)ESI-MS sensitivity. In general, the concentration of acetonitrile in the mobile phase re sulted in a greater than 10 times increase in the MS signal when compared to the C18 phase. From a global metabolomics perspective and using the C18-column approach, 810 compounds were isolated by using the software Molecular Feature Extractor (MFE). Identification of the metabolites was performed by searching METLIN database with the accurate m/z of the detected ions. As a result, 8% of the metabolites matched compounds in the database agree within two pp m mass accuracy. Positive mass error was frequently observed, which was in agreement with the mass errors exhibite d for the calibrant ions at the specific scan. On the other hand, when the HILIC column information was analyzed, 105 compounds were isolated by the MFE software. A total of 54 (51%) compounds were tentatively identified in the METLIN database within two ppm mass accu racy. Negative mass error was most common, which was in agreement with the mass errors displayed for the calibrant ions at the specific scan. In summary, 65 compounds were tentativel y identified when using the monolithic C18HPLC-based method. On the other hand, 54 co mpounds were tentatively identified when performing HILIC-based chromatography. After comparing the results, 18 compounds were found by METLIN in both chromatographic approaches.

PAGE 52

52 Conclusions Two novel separation methodologies for the orthogonal resolution of m etabolites in biological samples were developed and su ccessfully applied. The two in-series C18-monolithic columns demonstrated adequate performance fo r the separation of nonpolar compounds. On the other hand, a simple, relatively rapid, a nd accurate LC-MS methodology was successfully developed and validated for sepa rating carnitine, short-chain acy lcarnitines, and amino acids by hydrophilic interaction chromatography. In addi tion, the conditions required for the HILIC separation were favorable to ESI-MS dete ction. The optimized methodology showed applicability to biological samples. Based upon an extensive literature search using SciFinder Scholar and Web of Science, this is the first time that a HILIC separation approach has been applied to the simultaneous separation of carnitine, polar acylcarnitines, and amino acids.

PAGE 53

53 Table 2-1. Chemical formula, RTs, and m/z of the [M+H]+ ions of carnitine, acylcarnitines, and amino acids ID Formula [M+H]+,1m/z Retention time (RT) (min) RPLC-C18 column RRT (min) RPLCC18 column2 Retention time (RT) (min) HILIC column RRT (min) HILIC column2 Carnitine C7H15NO3 162.1125 4.3 0.7 18.2 15.4 Acetylcarnitine C9H17NO4 204.1230 4.6 1.0 16.4 13.6 Propionylcarnitine C10H19NO4 218.1387 4.8-6.1 1.2 14.5 11.7 Butyrylcarnitine C11H21NO4 232.1543 6.0-11.4 2.4 13.5 10.7 Hexanoylcarnitine C13H25NO4 260.1856 17.1 13.5 13.6 10.8 Octanoylcarnitine C15H29NO4 288.2169 19.2 15.6 12.7 9.9 Decanoylcarnitine C17H33NO4 316.2482 20.1 16.5 11.9 9.1 Lauroylcarnitine C19H37NO4 344.2795 22.4 18.8 11.1 8.3 Myristoylcarnitine C21H41NO4 372.3108 23.2 19.6 10.4 7.6 Palmitoylcarnitine C23H45NO4 400.3421 24.2 20.6 9.9 7.1 Stearoylcarnitine C25H49NO4 428.3734 26.1 22.5 9.3 6.5 Alanine C3H7NO2 90.0555 4.4 0.8 18.2 15.4 Aspartic acid C4H7NO4 134.0448 4.5 0.9 20.4 17.6 Asparagine C4H8N2O3 133.0608 4.6 1.0 13.8 11.0 Taurine C2H7NO3S 126.0219 4.0 0.4 11.2 8.4 Glycine C2H5NO2 76.0393 4.6 1.0 13.1 10.3 Glutamic acid C5H9NO4 148.0604 4.5 0.9 20.8 18.0 Histidine C6H9N3O2 156.0768 4.7 1.1 19.2 16.4 Isoleucine C6H13NO2 132.1019 4.7 1.1 14.5 11.7 Leucine C6H13NO2 132.1019 4.7 1.1 14.6 11.8 Methionine C5H11NO2S 150.0583 5.2 1.6 14.8 12.0 Phenylalanine C9H11NO2 166.0863 6.0 2.4 14.5 11.7 Proline C5H9NO2 116.0706 4.6 1.0 14.7 11.9 Threonine C4H9NO3 120.0655 4.7 1.1 18.2 15.4 Tyrosine C9H11NO3 182.0812 4.6 1.0 17.2 14.4 Valine C5H11NO2 118.0863 4.7 1.1 12.1 9.3 Tryptophan C11H12N2O2 205.0972 11.1 7.5 14.7 11.9 Carnosine C9H14N4O3 227.1139 4.5 0.9 21.2 18.4 1Monoisotopic mass/charge (m/z); 2RRT (relative retention time) = RRT = RT(analyte) RT(void)

PAGE 54

54 Figure 2-1. Base peak chromatogr am obtained with a monolithic C18 column and acetonitrile as orga nic mobile phase; concentration standard mixture: 0.5 mg L-1; injected volume: 15 L; mobile phase flow rate: 1.0 mL min-1; C0: carnitine; C2: acetylcarnitine; C3: propionylcarnitine; C4: butyrylcarnitine; C6: hexanoylcarnitine; C8: octanoylcarnitine; C10: decanoylcarnitine; C12: lauroylcarnitine; C14: myristoylcarnitine; C16: palmitoylcarnitine; C18: stearoylcarnitine (adapted from reference [100].)

PAGE 55

55 Figure 2-2. Effect of acetonitril e content on the isocratic HILI C of acylcarnitines and amino acids. Mobile phase: ACN:H2O; Flow rate: 0.3 mL min-1; Temperature: 30 C; Concentration of standa rd mixture: 0.5 mg mL-1.

PAGE 56

56 Figure 2-3. Effect of buffer addition to the mobile phase fo r HILIC separation. Mobile phase: ACN:H2O; Flow rate: 0.3 mL min-1; Temperature: 30 C; Concentration of standard mixture: 0.5 mg mL-1.

PAGE 57

57 Figure 2-4. Extracted ion chromatograms for some model compounds showing the effect of buffer addition to the mobile phase fo r HILIC separation. Mobile phase: ACN:H2O; Flow rate: 0.3 mL min-1; Temperature: 30 C; Concentration of standard mixture: 0.5 mg mL-1.

PAGE 58

58 Figure 2-5. Effect of concentr ation of buffer on the retention time of some representative acylcarnitines and amino aci ds. Mobile phase: ACN:H2O; Buffer: ammonium formate; Flow rate: 0.3 mL min-1; Temperature: 30 C; Concentration of standard mixture: 0.5 mg mL-1.

PAGE 59

59 Figure 2-6. Vant Hoff plots fo r evaluating the effect of th e temperature on the retention efficiency. Mobile phase: ACN:H2O; Buffer: ammonium form ate; Flow rate: 0.3 mL min-1; Concentration of standard mixture: 0.5 mg mL-1.

PAGE 60

60 Figure 2-7. van Deemter plots for evaluating the effect of the mobile phase flow rate on the retention efficiency. Mobile phase: ACN:H2O; Buffer: ammonium formate; Temperature: 30 C; Concentration of standard mixture: 0.5 mg mL-1.

PAGE 61

61 Figure 2-8. Total ion chromatogr ams for the application of C18-monolithic/ and HILIC/MS on piglet plasma samples.

PAGE 62

62 Figure 2-9. Extracted ion chromatograms of carn itine, acetylcarnitine, propionylcarnitine, and butyrylcarnitine under HILIC and C18-monolithic conditions, respectively. Note: the extracted ion chromatograms for those compounds listed in Table 2-1, but not detected in the plasma samples, are not shown. The shaded peaks in each chromatogram correspond to those matching the retention times of the standards.

PAGE 63

63 Figure 2-10. Extracted ion chromatograms of alanine, proline, valine, threonine, and leucine/isoleucine under HILIC and C18-monolithic conditions, respectively. The shaded peaks in each chromatogram corres pond to those matching the retention times of the standards.

PAGE 64

64 Figure 2-11. Extracted ion chromatograms of glutamic acid, methionine, histidine, phenylalanine, and tyrosine under HILIC and C18-monolithic conditions, respectively. The shaded peaks in each chromatogram correspond to those matching the retention times of the standards.

PAGE 65

65 CHAPTER 3 APPLICATION OF LC/MS METHODOLOGIES TO HUMAN PLASMA SAMPLES FOR GLOB AL AND TARGETED METABOLOMIC STUDIES Metabolomics Applied to Biomarker Discovery The m etabolome is the most predictive metr ic of phenotype; thus, metabolomics holds the promise to extensively contribu te to the understanding of phenotypi c changes as an organisms response to disease, genetic changes, and nutritional, toxi cological, environmental, and pharmacological influences [1]. With respect to human health, multiple benefits of metabolomics investigations can be envisioned. This field can deliver new tools to diagnose disease or monitor the success of nutritional and pha rmacological interventions, provi de new biomarkers to assess human health, and over time a powerful list of diagnostic markers will evolve, which can be measured using high-throughput assays. Medical diagnosis and treatment efficacy wi ll improve significantly when a more personalized system for health assessment is im plemented. This system will require diagnostics that provide sufficiently detailed information about the metabolic st atus of individuals such that assay results will be able to guide food, drug and li festyle choices to maintain or improve distinct aspects of health without compromising othe rs [106]. The technologi es to accurately and quantitatively understand how the integrated meta bolism affects human health are available and they are now been brought into practice for this purpose. To make sense of metabolite data, the metabol ites must first be understood in the context of their biochemical pathways [ 107, 108]. To explain why the levels of particular metabolites are outside a normal range, and more importantly to predict how to alter i nputs e.g. diet, drugs and lifestyle to accomplish a selected change in metabolism, an understanding of how pathways and their respective reactions f unction is required [ 106]. This is the body of knowledge that the field of biochemistry built throughout the 20th century and continues toda y. Metabolomics is thus

PAGE 66

66 ideally situated to integrate th e existing knowledge of biochemistry into a single, comprehensive strategy to address the h ealth challenges facing the modern world [106]. One of the important objectives of metabolomics is to measur e all or a substantial fraction of the metabolites within a biological sample and to quantify each relative to an absolute index of the sample. Using modern analytic al technologies, it is often as easy to monitor many analytes as it is to measure one. In this sens e, two different analyt ical strategies can be defined: global and targeted metabolomic approaches. The global or non-targeted approach tries to comprehensively analyze all known and unknown metabolites in a given sample with the ultimate goal of identifying discriminating metabolites. On the other hand, targeted metabolomics has the more modest goal of quantitati on of selected metabolites, most typically dozens to hundreds of known compounds. The use of these two complementary appr oaches will be presented in this chapter. Ketogenic Diet Therapy A ketogenic diet (KD) is a high-fat, low-carbohydrate/ade quate-protein diet designed to increase the bodys dependence on fat rather than glucose for energy and to treat disorders of the brain [109]. The two disorders m ore commonly treat ed are epilepsy and cert ain inborn errors of metabolism involving glucose utilization. Despite its long history of clinical use, it is still not entirely clear what mechanism(s) underlie its epileptic seizure-suppre ssive action and how the KD affects the brains general metabolism. Anti-epileptic medications (AEDs) are commonly the first choice of treatment for epilepsy due to their efficacy and convenience in admi nistration. But despite th e large array of (and combinations of) AEDs available, there remain 25% of patients that continue to have seizures, and are deemed to have intractable epilepsy [109]. Many of these intractable epileptic patients are on a heavily concentrated multiple AED regimen and their quality of life has been diminished due to large list of detrimental side effects of AEDs such as sedation, insomnia,

PAGE 67

67 hyperactivity, hepatotoxicity, nausea, vomiting, d ecreased bone density, metabolic acidosis and behavioral issues [109-112]. Thus the goal of the KD is to improve the quality of life by reducing or eliminating seizures and by reducing the adve rse side effects of AEDs by decreasing their dosage. In the US, ketogenic diet therapy for seizures has been used clinically for more than seven decades with an efficacy similar to that of the best AEDs [109-112]. During those years numerous proposed mechanisms of action have been evaluated, but the mechanism remains unknown [113, 114]. The ketogenic diet clearly increases blo od ketones which have the advantage over most other intermediates in en ergy generating pathways in that they are transportable from the mitochondria of the organ of synthesis to blood and then to the brain to regenerate acetyl-coenzyme A, which can enter the tricarboxylic acid cycle and be used for energy [115]. Acetylcarnitine is also synthesized in energy generating pathways and shares all the transportable advantages of ketones [116]. Whereas expenditure of adenosine triphosphate (ATP) is required for conversion of ketones to acetyl-coenzyme A; acetylcarnitine does not require ATP to produce acetyl-coenzyme A, which is critically important in a pathological condition characterized by a low metabolic energy conditions. In addition to acetylcarnitine possibly being an important source of energy for the brain in KD, many of its other newly discovered metabolic roles may position it in an important role in the mechanism of ketogenic diet therapy [115]. To our knowle dge, this is a novel hypothesis concerning the mechanism of ketogenic diet therapy. To test this hypothesis, plasma samples we re collected. The carnitinome and untargeted metabolome of fasting plasma collected from h ealthy female and male adults (n=10) before beginning the ketogenic diet were compared to the data of fasting plasma obtained in the same

PAGE 68

68 people, after a week of ketogenic thera py. Liquid chromatography interfaced with mass spectrometric analyses was conducted to permit m easurement of the targeted carnitinome (free carnitine and the common acylcarnitines from acetyl carnitine to stearoylcarnitine). The objective was to test the hypothesis that ketogenic diet alte rs the plasma carnitinome and other metabolites. In this sense, nontargeted global metabolomic analyses were used to identify any metabolites that increased or decreased when a KD was followed. Th e experimental design utilized for this study is presented in Figure 3-1. Experimental Workflow The workflow utilized for this study consisted of the followin g steps: Sample analysis The high-throughput screening with LC/MS techniques generates large volum es of analytical data that require advanced softwa re for data mining. MassHunter and GeneSpring MS software (Agilent Technologies, Wilmington, DE) were used to perform high-throughput processing and automated quality analysis of hun dreds of gigabytes of MS-based metabolomics data. These platforms also include sophisticated statistica l analysis algorithms that allow for data comparison. Data processing. There is no universal metabolomic software due to the variety of commercial and custom-built instruments and the num ber of data formats. Therefore, instrument manufacturers (or independent soft ware developers) create their ow n software [1]. Software is able to perform data extraction to allow da ta comparison and discovery of differentially expressed metabolites. In our study, data proces sing was performed by the instrument software. It consisted in the following steps: feature ex traction, peak alignment, normalization, filtering and quality control of the data (Figure 3-2).

PAGE 69

69 Feature extraction consists in the detection and extraction of three-dimensional signals, socalled features, which are caused by chemical entities [117]. From an LC/MS perspective, each chemical entity is characteriz ed by a mass-to-charge ratio (m/z ratio), its chromatographic retention time and integrated area. Peak alignment is the process of finding significant peaks across samples in multiple files. A significant peak is one that is prevalent across sa mples and is the most intense peak in a certain m/z and retention time range [117]. In the software used for this study, peak alignment was carried out on the compounds read from the data files. The file that has the maximum number of compounds (most populated sample) is picked as a reference and is used for finding significant peaks from the remaining samples. The peak s/compounds of the most populated sample are placed in a two dimensional grid (array), then the compounds from the remainder of the samples are placed at the place of best match. The result is a set of buckets containing compounds found from all the samples. In addition, internal sta ndards can be used to correct for retention time drift. When a feature is found in a particular file whose RT and mass match the specifications of a standard, the standard is considered found. Normalization is the process of applying math ematical modifications to the values of a variable. The data values are recalculated and used in subsequent analysis. Normalization is applied to avoid neglecting signi ficant, but low-abundance peaks. In other words, normalization is used to adjust the intensities within each sample by reducing the systematic error [117]. Different normalization steps can be applied on the data. Per-run normalizations control for runwide variations in intensity. Per mass normalizat ion accounts for the differences in detection efficiency between runs and compare the relative change in mass abundance levels.

PAGE 70

70 Filtering and quality control allows the analyst to sort through the large amounts of abundance data, to evaluate the quality of sample data before performing data analysis, and to identify interesting masses for fu rther study after analys is [117]. Filters can be applied to single or multiple data objects. Masses can be filtered on specific abundance criteria and the masses that pass a filter are made into a mass list. The fi lters can be based on any data associated with the masses including raw or normali zed intensity values, fold cha nge comparisons, flag values, statistical information, or raw data from the scanning software. Filtering can be applied to identify masses that fall below a given intensity value threshold, data that exceed recommended signal-to-noise or signal-to-b ackgrounds measurements, outliers that fall outside the range of standard deviations from the mean, random quant ification errors, masses that do not show any changes in abundance during the experiment, interesting masses suit able for additional analysis, etc. The ability to restrict a mass list based on the behavior of its masses in experiments or in individual samples is an impor tant quality control tool. Data analysis. The statistical tools for the metabolomic data analysis should be selected according to the aim of the study. Commonly used methods include unsupervised and supervised learning techniques. In unsupervised learning problems, the object is often to identify previously unknown structure in the data. In supervised lear ning, the structure (e.g., classes or groups in the data) is assumed to be known at the outset an d this knowledge is used in the statistical analysis [118]. A brief description of the statistical tools applied to this study is given in the following paragraphs. Hierarchical clustering analysis (HCA) is an iterative agglom erative clustering method that can be used to produce condition trees. Thes e resulting trees group samples or conditions together based on the similarity of their abundan ce data (across a specified mass list) and allow

PAGE 71

71 picking out groups of similar samples/conditions [117]. In these structures, individual samples are grouped together at the base with larger clusters forming as the tree is grown toward the top. Samples that are more similar to one another are shown closer together in the tree and the branching structure shows where masses and group of masses were joined during the clustering steps. At the bottom of the tree all samples exis t as the most independent grouping of the objects. Moving upward shows the samples joining into larg er branches until all the samples are joined. Some of the advantages of cluste ring trees are: clustering struct ure is determined completely by abundance data and if confidence levels are calc ulated, the significance of clusters can be assessed. On the other hand, some of the disadvantages of this analysis tool are: it is computationally intensive and can be slow for a large starting number of samples; there can be multiple ways to display the branching structure for a given clustering. The generated display is not unique, and there is no easy way to generate a set of discrete clusters from a condition tree. Principal component analysis (PCA) is a cova riance analysis of different factors [117]. Covariance is always measured between two fact ors. So with three factors, covariance is measured between factor x and y, y and z, a nd x and z. When more than two factors are involved, covariance values can be placed into a matrix. This is where PCA becomes useful. PCA finds eigenvectors and eige nvalues relevant to the data using the covariance matrix. Eigenvectors can be thought as of as preferential direct ions of a data set, or in other words, main patterns in the data. For PCA on masses, an eigenvector would be represented as a mass profile that is most represen tative of the data. For PCA on co nditions, an eigenvector is a condition profile. For either PCA, there cannot be more eigenvectors than there are masses or conditions in the data.

PAGE 72

72 Eigenvalues can be thought of as quantitat ive assessment of how close an eigenvector represents the data. The higher th e eigenvalue of an eigenvector, th e more representa tive it is of the data. Eigenvalues can also indicate the level of explained variance as a percentage of total variance and the percent of variance explained is dependent on how well all the components summarize the data. In a typical MS experiment, the abundance of thousands of m/z values is measured across many conditions or time points. Therefore, it beco mes impossible to make a visual inspection of the relationship between m/z values or conditions in such a multi-dimensional matrix. One way to make sense to this data is to reduce its di mensionality. Several data decomposition techniques are available for this purpose: principal componen t analysis is among these techniques, it reduces the data to just a few dimensions. PCA reduces data dimensionality by performing a covariance analysis between factors. If each of the thousands of integrated m/z intensitie s measured in an LC/MS experiment is plotted in a multi-dimensional scatter plot, the result is a cloud (grouping) of values in a multidimensional space. To characterize the trends exhibited by th is data, PCA extracts directions where the grouping, which has the appearance of a cloud, is more extended; this is called the first principal component. PCA then looks for the next direct ion, orthogonal to the first one, reducing the multidimensional cloud grouping into a two-dimensi onal space. Two components may be able to explain most of the clouds groupings trends. Ho wever, in a more complex data set, the third, fourth, and higher components might reveal additiona l information about interesting trends in the data.

PAGE 73

73 PCA is recommended as an exploratory tool to uncover unknown trends in the data. In the case of LC/MS experiments, PCA on integrated mass intensities masses provides a way to identify predominant mass abundance patterns When applied on conditions, PCA explores correlations between samples or conditions. PCA is not a clustering tool and does not attempt to group masses by user-specified criter ia as do the clustering tools. Analysis of Variance (ANOVA) is based on a pr obability value (p-value), and allows the analyst to determine if one given factor, such as drug treatment, has a significant effect on mass abundance behavior across any of the groups under study [117]. The p-value indicates the probability of getting a mean diff erence between the groups as high as or higher than what is observed. The lower the p-value, the more signi ficant the difference between groups. In other words, a significant p-value resulting from a 1-way ANOVA test would i ndicate that a mass is differentially expressed in at leas t one of the groups analyzed. However, if there are more than two groups being analyzed, the one-way ANOVA does not specifically i ndicate which pair of groups is statistically differentially expressed. Post hoc tests can be applied in this specific situation to determine which spec ific pair/pairs are differentially expressed [117]. If the Tukey post hoc test is applied, all means for each condition are ranked in order of magnitude, from the lowest to the largest. The group with the lo west mean gets a ranking of 1. The pairwise difference between means, starting with the largest mean compared with the smallest mean are tabulated between each group pair and divided by th e standard error. This value, q, is compared to a t-test range critical value. If q is larger than the critic al value, the pair of groups is considered to be statistically differentially expressed. Volcano plots are useful tools for visualizing differential ab undance between two different conditions [117]. They are constructed using fold -change values and p-values, and thus allow

PAGE 74

74 visualizing the relationship between fold (ma gnitude of change) and statistical significance (which takes both magnitude of change and vari ability into consideration). They also allow subsets of masses to be isolated, based on those values. A volcano plot is ba sically a scatter plot, where each mass is plotted on two axes correspo nding to p-value (p-value for a t-test of differences between samples) and fold-change (ma gnitude o change). Both axes are typically on a log scale, a negative log scal e for the p-value, so smaller pvalues appear higher up; and a positive and negative log scale for the fold-change, so the changes appear symmetric. The foldchange indicates biological impact of the change; the p-value axis indicates the statistical evidence, or reliability of the change. Identification Num erous tools are normally used to identi fy metabolites: mass accuracy, retention times, UV signals, fragmentation patterns, elemental co mposition, structural information, and library and database searches. Searching a database of me tabolite information can help to narrow the list of possible candidates. Additionally, accurate ma ss data make this database searching more effective by narrowing the mass window and thus reducing the number of po ssible identities that need to be searched. In these experiments, a database search was conducted on the significant masses for identification of putative biomarkers and it was constrained by the two ppm mass tolerance specification for the Ag ilent 6210 ESI-TOF. Two different databases were utilized and compared: METLIN (METaboliteLINk) and HMDB (Human Metabolome DataBase). Biological interpretation Biological interpretation is where technology, identification, and knowledge need to be integ rated. However, no metabolite ontology has been developed yet for metabolites and the interpretation of correlations is still a challenge [10]. Figure 3-2 lists all th e steps that form part of the experimental workflow used for this study.

PAGE 75

75 The above-mentioned workflow has been applie d to the evaluation of the hypothesis that the metabolome will change as a consequence of a ketogenic diet therapy. It is well-known that the mechanism behind KD is connected with the radically altered macronutrient metabolism [115]. This shift in metabolism is expected to cause major alterations in the plasma metabolites including acylcarnitines, which have been noted to change du ring such macronutrient changes [119]. The changes in metabolites and acylcarnitines can be measured and evaluated using the LC/MS, data processing, and data analysis globa l and targeted metabolomic approach. Since the study is designed to assess the eff ects of a ketogenic diet on hea lthy adults, the results found will reflect only changes due to KD, unlike the metabolome to be examined from patients with the added complications of seizures and AEDs. The ultimate goal is to understand the effect of the KD on the metabolome/carnitinome of intractable ep ileptic patients, which will be addressed in subsequent studies. Experimental Methods Sample Collection The study was conducted at the General Clinic al Research C enter (GCRC) of Shands Hospital at the University of Florida and was a part of a larger study evaluating the KDs effects on dyslipidemia and insulin sensitivity [120]. Th e protocol was approved by the Scientific Advisory Committee of the GCRC and the Institu tional Review Board (IRB) of College of Medicine. All subjects signed the informed consent form. Twenty young (10 males and 10 females), healthy first year medical students volunteered to partic ipate in the study. The subjects were screened and excluded if they were pregnant, vegetarian or an endurance athlete, or if they had cardiovascular disease, diabetes mellitus, ne urological diseases, kidney disease, any food allergies, or if on any medications. Subject s were matched for gender and body mass index (BMI) and randomized by parallel design into tw o groups, polyunsaturated fatty acids (PUFA) or

PAGE 76

76 saturated fatty acids (SFA). There were no significant differences between age, height, and BMI between the two groups. On the first day of study the subjects were admitted to the GCRC and fasted overnight. The following morning two vials of venous blood we re obtained for later carnitinomics and metabolomics analysis. The subj ects then received thei r respected diets of 60% PUFA, 15% SFA, and 25% MUFA (monounsaturat ed fatty acids, the PUFA diet) or 60% SFA, 15% PUFA, and 25% MUFA (the SFA diet). Both diets consisted of 70% of fatty acids, 15% of carbohydrates and 15% of proteins, which is the equivale nt of a 2.33:1 ratio. The diets were designed to maintain the subjects weight This was done through th e collection of a 4 day non-consecutive diet recall before in itiating the diet and careful monitoring of weight during the diet. The subjects only ate food pr ovided by the GCRC and on the 5th day the subjects venous blood was again collected for acylcarnitine and metabolomic analysis. Sample Preparation The volunteers cam e in fasting and their blood wa s on ice or in a freezer at all times during preparation. Once blood was received, it was immedi ately centrifuged at 4,750 rpm and 4-6 C to separate the red blood cells and plasma compartments. It was th en brought back to the Food Science and Human Nutrition Depart ment (FSHN) lab on ice and stored at -20 C until further preparations took place. The experi mental section as detailed in Ch apter 2 is also applicable to Chapter 3. Results and Discussion LC/MS Experiments In this study, the hypothesis that the plasm a metabolic profile of healthy female and male adults before and after ketogenic diet therapy will differ was tested. Si nce ten female and ten male participated in this study and each sample was run in triplicate, 30 files were obtained for the experiments related to the female sample s before and 30 files after the KD using the

PAGE 77

77 monolithic column approach. In addition, 30 files were obtained for the female samples using the HILIC column. On the other hand, 30 files per chromatographic method were acquired for the male samples before and after the ketogenic diet. Thus, 120 files per chromatographic method were used for further data analysis and processing. Total ion chromatograms (TICs) corresponding to the overlapping of all files for the female and male samples before and after the ketogenic diet using monolithic and HILIC ch romatography are shown in Figures 3-3 3-6. Changes in the chromatographic profiles can be observed from these figures. In this sense, reproducibility was noticeab le better for the TICs obtained with the C18-approach. The variation in the TICs shape for the HILIC methodology can be a consequence of runto-run changes in the interaction between samples and stationary phase. Global Metabolomics Strategy This approach focused on the com prehe nsive analysis of all known and unknown compounds present in the samples under study. Fo r data processing and analysis, the abovementioned statistical tools were applied to the LC/MS files. Hierarchical clustering analysis As m entioned previously, in a hierarchical clustering anal ysis samples that are more similar to one another are shown cl oser together in the tree structur e. In the case of MS data, the tree structures are determined by abundance data. Cluster analysis was a pplied to the raw data (163,309 masses corresponding to C18-monolithicand 94,070 masses to HILIC-experiments) in order to evaluate the female and male samples before and after ketogenic diet using C18 monolithic and HILIC strategies. The dendrograms gene rated revealed two major groups: male and female samples; however, no clear differe ntiation based on diet-before and afterwas observed. These results are shown in Figures 3-7 and 3-8, in which the female samples before

PAGE 78

78 and after the diet are highlighted in yellow and red, respectively; and the male samples before and after the diet are highlight ed in bright and dark blue. Principal component analysis PCA is able to reveal relationships between different experim ent interpretations, e.g. gender, diet type, etc. Unlike cluster analysis, PCA does not have the constraint of generating a hierarchic classification. Therefore, 120 files were subjected to PCA. Prior to performing PCA, the raw masses were filtered based on relative frequency tool, which allowed only the most frequent masses across all the samples to remain part of the analysis. PCA plots were obtained and they are shown in Figures 3-9 and 3-10. Th e PCA plots revealed the presence of gender differences, which was visualized as two well-defined groups and these results agreed for both C18-monolithic and HILIC chromatography. However, PCA was not able to find diet correlations across the samples. Analysis of variance ANOVA is applied to c ompare between multiple sample classes. The so-called post-hoc tests are used to further locate the differences between any of these classes. Both methods are dependent on the number of replications and th e distribution properties of measurements. The significance of differences is typically e xpressed by a probability value (p-value). ANOVA applied to the above-mentioned data sets (masses from C18-monolithic and HILIC approaches after being filtered on relative frequency) demons trated that the minor variances of a data set may contain valuable discriminatory informati on. In this sense, not only differences between genders were observed, but also between sample s before and after diet Figure 3-11 summarizes the results for the ANOVA test for both types of column approaches. In the case of the C18approach, a total of 7,385 masses were found stat istically significant by the ANOVA analysis, 593 masses were different between females before and after diet and 853 masses between males.

PAGE 79

79 On the other hand, when the results for the HILI C samples were obtained, a total of 1571 masses were found differentially expressed, 468 masses we re different between female samples before and after diet and 67 between male samples. Volcano analysis In order to evaluate the m agnitude of cha nge in those statistically significant masses obtained by the ANOVA tests, Volcano analysis was used. As mentioned earlier, the Volcano analysis uses the p-value (proba bility value) and the magnitude of change, or fold-change, to determine the significantly expressed masses. In this study, a p-value cutoff of 0.05 (0.05 false discovery rate) and a fold change equal or greater than two were ut ilized to perform the analysis. The ANOVA masses obtained from the comparison between females and males before and after diet were submitted to Volcano analysis. Thus, for the female samples analysis using C18monolithic columns, 41 (7%) masses out of 593 pa ssed the filter; whereas 112 (13%) masses out of 853 passed the filter for the male samples (F igure 3-12). On the other hand, when the dataset from HILIC column separation was used, 64 (1 4%) masses out of 468 passed the filter for the female samples and 25 masses out of 67 (37%) for the males (Figure 3-13). Database search for identification The m asses that passed the Volcano test were submitted for identification to two databases: METLIN and HMDB. Using a mass tolerance equal to 2 ppm, the Venn diagrams in Figures 314 and 3-15 show the number of mass es identified in each of the da tabases. In all the cases, the HMDB database contributed a higher percentage of identities for the putative biomarkers. The tentative biomarkers were sort ed in different classes of co mpounds based on the biochemical information provided by the HMDB database (Figur es 3-16 and 3-17). In this sense, no classes of compounds were common to males a nd females results when using the C18-separation

PAGE 80

80 strategy. However, three classes of compounds -acyl glycines, amino acid conjugates, and acylcarnitineswere common for both males and females results when HILIC was applied. Targeted Metabolomic Strategy Carnitinom ics: Carnitines intimate role in fatty acid metabolism makes it and other acylcarnitines potential biomarkers for understanding changes in macronutrient metabolism. Thus, a subset of compounds composed of carniti ne and some important acylcarnitines was used for the targeted approach. These compounds have already been introduced in Chapter 1 and their structures shown in Figure 1-3. In this targeted study, the atten tion was focused on evaluation of the effects of the ketogenic diet on 11 acylcarnitines: carnitine (C0), acetylcarnitine (C2), propionylcarnitine (C3), butyrylcarnitine (C4), hexanoylcarnitine (C6), octanoylcarnitine (C8), decanoylcarnitine (C10), lauroylcarnitine (C12), myristoylcarnitine (C14), palmitoylcarnitine (C16), and stearoylcarnitine (C18); respectively. The instrument respons e, expressed as peak area counts, was extracted for each compound in each of the da ta files. Then, these peak areas for each compound in each of the data files were added to obtain the total area. Thus, the areas of each compound were compared to the total area and th e corresponding percentages of the total area were obtained. A ratio of the percentages of each person after (A) and before (B) the ketogenic diet was calculated and the A/B ratio was plotted versus carnitine a nd acylcarnitines; thus, profiles for all of the females and males were obta ined (Figures 3-18 and 3-19). From these plots, it is possible to observe that th e ketogenic therapy affected carnitine and acylcarnitines profiles and these changes were more pronounced for some of the male samples. On the other hand, a higher variability can be observed in the case of the female samples. This is the first time that a metabolomic approach is applied to evaluate th e changes in the carnitinome of healthy adults before and after a ketogenic diet and there is no evidence or expl anation in the literature about the observed profiles. Our research related to the biochemical significan ce of the results is

PAGE 81

81 underway. In addition to the mentioned plots, aver aged values of the prev iously calculated A/B ratios versus targeted carnitine and acylcarnitines profiles and averaged values of the A/B ratios versus total carnitine and acylcarni tines profiles were plotted and th ey are shown in Figures 3-20 and 3-21 for the female samples, and in Figures 3-22 and 3-23 for the male samples. In summary, it is possible to conclude that carnitine and almost a ll of the acylcarnitines changed due to the effects of the ketogenic diet. Conclusions Two com plementary HPLC column technologies, C18-monolithic and HILIC, were applied to underivatized human plasma samples successfully. The extensive amount of information generated was processed and analyzed with sophis ticated software tools. Clustering and principal component analyses showed a dist inct grouping of the males and fe males, but no diet effect was observed. After applying analysis of variance, statistically significant masses for females and males before and after ketogenic diet ther apy were identified. As shown from the ANOVA results, five percent of the tota l number of masses distinguished th e diet for the female samples, and a similar number for the males. Volcano analysis allowed determining the masses that increased or decreased by a factor equal or gr eater than two. Those masses were submitted for identification, based on mass accuracy, to METLIN and HMDB databases, different percentages of the compounds were found as illustrated with the Venn diagrams and lists of classes of compounds were generated. In addition to the global st rategy, a targeted metabolomic approach was applied and carnitine and acylcarnitines profiles were generated. These pr ofiles demonstrated the significant effect of the ketogenic diet on the targeted compounds.

PAGE 82

82 Figure 3-1. Study design for the evaluation of th e metabolomic effects of a ketogenic diet on healthy adults. Plasma samples from ten he althy female and ten healthy male adults were collected before and after a ketogenic diet therapy. LC/MS approaches were applied for the analysis and th ree technical replicates per sample were analyzed. Both monolithic C18-RPLC and HILIC column were employed followed by positive ion electrospray MS.

PAGE 83

83 Figure 3-2. Experimental workflow. The raw instru ment data were processed in several steps by the instrument software. From the obtain ed chromatograms, features (integrated mass-intensity pairs) were first extracted, aligned, normalized, filtered and visualized for quality control evaluation (data processing sequence), and then analyzed by different statistical methods, including hi erarchical clustering analysis (HCA), principal components analysis (PCA), analysis of va riance (ANOVA), and Volcano plots. Finally, using the accurate mass criterion of 2 ppm mass error, putative biomarkers were identified in the databases METLIN and HMDB.

PAGE 84

84 Figure 3-3. Total ion chromatograms (TICs) for the overlapping of all female samples ( 30 files) before (top) and after ketogen ic (bottom) diet using C18-monolithic chromatography. Solvent A: 1% (v/v) acetic acid in water; Solvent B: 1% (v/v) acetic acid in acetonitrile; Flow rate (mL min-1): 1.0. Injection volume (L): 15.0; Grad ient: A:B (min): 95:5 (0.0-6.5); 0:100 (25.5-35.0); 95:5 (55.0-65.0).

PAGE 85

85 Figure 3-4. Total ion chromatograms (TICs) for the overlapping of all female samples ( 30 files) before and after ketogenic die t using HILIC chromatography. Solvent A: 7.5 mM ammonium format e in water; Solvent B: 7.5 mM ammonium formate in acetonitrile; Flow rate (mL min-1): 0.3. Injection volume (L): 10.0; Gradient: A:B (min): 10:90 (0.0-5.0); 50:50 (25.030.0); 10:90 (40-45).

PAGE 86

86 Figure 3-5. Total ion chromatograms (TICs) for the overlapping of all male samples (30 files) before (top) and after (bottom) ketogenic diet using C18-monolithic chromatography. Solvent A: 1% (v/v) acetic acid in water; Solvent B: 1% (v/v) acetic acid in acetonitrile; Flow rate (mL min1): 1.0. Injection volume (L): 15.0; Grad ient: A:B (min): 95:5 (0.0-6.5); 0:100 (25.5-35.0); 95:5 (55.0-65.0)

PAGE 87

87 Figure 3-6. Total ion chromatograms (TICs) for the overlapping of all male samples (30 files) before (top) and after (bottom) ketogenic diet using HILIC chromatography. Solvent A: 7.5 mM ammonium formate in water; Solvent B: 7.5 mM ammonium formate in acetoni trile; Flow rate (mL min-1): 0.3. Injection volume (L): 10.0; Gradient: A:B (min): 10:90 (0.0-5.0); 50:50 (25.0-30.0); 10:90 (40-45).

PAGE 88

88 Figure 3-7. Hierarchical clustering analysis (C18-monolithic column, #masses: 163,309) result ing in grouping by different experiment interpretations (gender, sample type, samp le number, patient number, and file name). The tree is clearly divided into two groups: female and male samples (see color key on the right), but no differentiation based on diet was obtained. CT all samples Gender and sample type Sample Number File Name Gender Sample type Patient number and gender Patient number, gender, and sample type

PAGE 89

89 Figure 3-8. Hierarchical cluste ring analysis (HILIC column, #masses: 94,070) resulting in grouping by different experiment interpretations (gender, sample type, samp le number, patient number, and file name ). The tree is also divided into two groups: female and male samples (see color key on the right), but once again no diet differentiation was observed. CT all masses Gender_Sample type Sample Number File Name Gender Sample type Patient number_Gender Patient number, gender, sample type

PAGE 90

90 Figure 3-9. Principal component analysis on C18-monolithic column data files. The PCA plot shows correlations across gender.

PAGE 91

91 Figure 3-10. Principal component analysis on HILIC data files. The PCA plot show s correlations across gender.

PAGE 92

92 Figure 3-11. One-way analysis of variance (ANOVA) for C18-monolithic and HILIC approaches. Test type: parametric, do not assume variances equal; False discovery rate: 0.05; Multiple testing co rrection: Benjamini and Hochberg false discovery rate; Post Hoc test: Tukey; BD: before diet; AD: after diet. The red boxes represent the statistically different masses, the grey boxe s the total number of masses, and the blue boxes the statistica lly similar masses. C18-monolithic HILIC

PAGE 93

93 Figure 3-12. C18-chromatography Volcano plots on di fferential abundance (fold changemagnitude of changeand p-value-magnitude of change and variability-) between females A) and males B) before and afte r diet. Red dots correspond to masses that pass the test. A B

PAGE 94

94 Figure 3-13. HILIC-chromatograp hy Volcano plots on differen tial abundance (fold changemagnitude of changeand p-value-magnitude of change and variability-) between females A) and males B) before and afte r diet. Red dots correspond to masses that pass the test. B A

PAGE 95

95 Figure 3-14. C18-chromatography Venn diagrams. The masses that passed the Volcano test were submitted for identification to several databases. The diagrams show the number of masses identified in METLIN and HMDB The pink-orange diagrams correspond to the group of masses obtained after comp aring female samples and the light-dark blue ones to the masses from the male analyses.

PAGE 96

96 Figure 3-15. HILIC chromatography Venn diag rams. The masses that passed the Volcano test were submitted for identification to several databases. The diagrams show the number of masses identified in METLIN and HMDB The pink-orange diagrams correspond to the group of masses obtained after comp aring female samples and the light-dark blue ones to the masses from the male analyses.

PAGE 97

97 Figure 3-16. C18-chromatography Classes of compounds id entified as changing as the result of the ketogenic diet.

PAGE 98

98 Figure 3-17. HILIC chromatography Classes of compounds identifi ed as changing as the result of the ketogenic diet.

PAGE 99

99 Figure 3-18. A/B (after diet /before diet) ratio response vs. carnitine and acylcarnitines in each of the ten female samples. C0: carnitine, C2: acetylcarnitine; C3: propionylcarnitine; C4: butyrylcarnitine; C6: hexanoylcarnitine; C8: octanoylcarnitine; C10: decanoylcarnitine; C12: lauroylcarnitine; C14: myristoylcarnitine; C16: palmitoylcarnitine; C18: stearoylcarnitine.

PAGE 100

100 Figure 3-19. A/B (after diet /before diet) ratio response vs. carnitine and acylcarnitines in each of the ten male samples. C0: carnitine, C2: acetylcarnitine; C3: propionylcarnitine; C4: butyrylcarnitine; C6: hexanoylcarnitine; C8: octanoylcarnitine; C10: decanoylcarnitine; C12: lauroylcarnitine; C14: myristoylcarnitine; C16: palmitoylcarnitine; C18: stearoylcarnitine.

PAGE 101

101 Figure 3-20. A/B (after diet/before diet) ratio versus carnitine and acylcarnitines profiles for all the female samples. Figure 3-21. A/B (after diet/befor e diet) ratio versus female pr ofiles for all the carnitine and acylcarnitines under study.

PAGE 102

102 Figure 3-22. A/B (after diet/before diet) ratio versus carnitine and acylcarnitines profiles for all the male samples. Figure 3-23. A/B (after diet/before diet) ra tio versus males for all the carnitine and acylcarnitines.

PAGE 103

103 CHAPTER 4 APPLICATION OF LC/MS METHODOLOGIES TO PIGLET PLASMA SAMPLES FOR GLOB AL AND TARGETED METABOLOMIC STUDIES The Plasma Metabolome of Pigl ets from Days 2-8 of Life Metabo lomics is the study of all small molecules and carnitinomics is the targeted metabolomic analysis of car nitine and acylcarnitines. Th e importance of studying the metabolome has been described in previous chap ters. The main objective of this study was the application of a global and target ed metabolomic workflow (refer to Chapter 3 and Figure 3-2) to an animal model for neonatal metabolomic fluc tuations over time. The experimental design consisted of twenty-one plasma samples obtained fr om three piglets on each day of life from day 2 to day 8 which were analyzed using the LC/M S methodologies described previously (Chapter 2). Targeted acylcarnitine analysis (carnitinomics) is often used in diagnosis of inborn errors of metabolism (IEM) such as medium chain acyl-CoA dehydrogenase deficiency (MCAD), one of the most common inborn errors of fatty acid metabolism [121, 122]. Esters of fatty acyl-CoA of 4-8 carbon chain length are broken down by medium chain acyl-CoA dehydrogenase, and a problem with this enzyme leads to an increase in medium chai n fatty acyl groups available for binding to carnitine [121]. Newborn screening for MCAD deficiency has been implemented in many regions. The metabolic marker for MCAD deficiency, octanoylcarnitine (C8), can be detected with a high degree of sensitivity in newborns by tandem mass spectrometry [123]. However, there is no clear evidence of metabolic information associated to changes in other acylcarnitine levels. It is here where the applic ation of carnitinomics (targeted and untargeted) becomes clinically useful. It is important to search for not only t hose acylcarnitines found previously, but also possible acylcarnitines that may be specific for the patient population that is being studied.

PAGE 104

104 Obtaining a sample of plasma from an adult is a relatively simple and minimally invasive procedure. However, analysis of metabolic profiles in human tissues or in infant plasma is much more difficult and impractical and may require the use of a model. Questions involving compartmentalization of compounds (e.g. acylcarnitin es) in tissues, such as how increases in levels of a specific metabolite in one organ may affect another, or how this metabolite levels change over the first few weeks of life may be be st suited to a non-human model in experimental design. A number of animal species have been used to study aspects of postnatal development, particularly rodents, sheep and primates. The pig (Sus scrofa domestica) is a popular model for many human conditions and its value as a laborat ory animal has already been assessed [124]. Some useful comparisons can be made between the newborn pigs and the human infants, e.g. their comparable level of maturity at birth, some of their anatomi cal similarities, their susceptibility to hypothermia, their ability to shiver, thei r increase in metabolic rate in the first few days after birth, and their limited thermal in sulation [124, 125]. Thus, th ese animals are ideal for study in experiments that do not require rapid reproductive cap abilities [126]. On the other hand, human neonates have an increased need for dietary carnitine due to the reduced stores and impaired synthesis of th e compound [127], and at birth both the neonatal human and piglet must switch from carbohydrat e to lipid metabolism involving the use of carnitine. The piglet can also be a good m odel for neonatal carnitine analysis because concentrations of plasma and tissue carnitine have been show n to be similar during early development [128]. Despite the fact that one day in the life of a pig does not correspond to one day in the life of a human, the life span and other developmental parameters correspond more with that of a human than almost any other an imal used for experimental purposes [128].

PAGE 105

105 In this study, the workflow of global and targeted metabolomics was applied to a piglet model to assess changes in neonatal metabolite le vels over the first week of life (days 2 through 8). The ultimate goal was to find biomarkers fo r metabolic and mitochondrial function for their clinical application to assess human neonates metabolic changes over time. Experimental Workflow The workflow utilized for this study consisted in the sam e steps described in Chapter 3. Experimental Methods Sample Collection Twenty-one m ale piglets born betwee n March 20, 2001 and January 29, 2003 were sacrificed in Dr. Peggy Borums lab from day 2 to day 8 of life (Table 4-1). The animals were fed by the sow until a short time before euthanization. Three piglets were euthanized on each day of life using isoflurane and oxyge n as anesthesia, and blood was extracted from the heart and quickly placed on ice for a short time before centrifugation. Plasma and red blood cells were separated using centrifugation for 15 minutes at 8,450 rpm and 4-6 C then stored at -20C. Piglets were also weighed and hematocrit was analyzed from the mean of the data from a hematocrit reader using three capillary tubes centrifuged for 5 minutes at 12,400 rpm. Sample Preparation Piglet plasm a sample preparation assay has been described in Chapter 2. In addition, the experimental section as detailed in Chap ter 2 is also applicable to Chapter 4. Results and Discussion LC/MS Experiments In this study, metabolic change s between days two and eight in the life of piglets were evaluated. The experimental design is presented in Figure 4-1. A total of twenty-one piglets were part of this study, three biological replicates ( i.e. three plasma samples from the same piglet) and

PAGE 106

106 three technical replicates (each sample was run three times) were analyzed; thus a total of 189 files were acquired per LC/MS methodology. Tota l ion chromatograms corresponding to the overlapping of all files using C18-monolithic and HILIC chromatography are shown in Figures 42 and 4-3, respectively. Global Metabolomics Approach This study focused on the com prehensive analysis of known and unknown compounds present in the piglet samples. Once the LC/ESI-M S data files were acquired, all the steps related to data processing and analysis were performed and plots for each of the statistical tools were generated. Hierarchical clustering analysis In Chapter 3, a brief description of the princi ples behind each of th e statistical algor ithms was given. Results obtained for the pigl et experiments are described here. Clustering analysis enables visu alization of hierarchical rela tions between different groups. Ideally, samples that are simila r are shown closer together in the dendograms. HCA was applied to the raw data: 15,260 mass-intensity pairs for the C18-monolithicand 34,736 mass-intensity pairs for the HILIC-experime nts were obtained. For the C18-approach, the generated clusters revealed single groups for the data files from days of life 3, 4, 7, and 8. However, data files related to days 2, 5, and 6 showed a greater vari ability, thus causing these groups to split (Figure 4-4). On the other hand, the data files corresponding to days 3-7 for the HILIC strategy grouped in individual clusters, but the files corresponding to days 2 and 8 divided into two groups each (Figure 4-5). It was found that biological and tech nical variability affected the clustering of the data files.

PAGE 107

107 Principal component analysis The fi rst few principal components re present a relevant part of the total data variance. Thus, when plotting principal com p onent scores, the data structure can be visually inspected in two or three dimensions in orde r to identify groups of objects (e.g. data files, conditions, etc.) Prior to performing PCA, the raw mass-intensity pairs were filtered out using the tool filter on relative frequency, which allowed only the most frequent mass-intensity pairs across all the samples to remain part of the analysis. After f iltering, the mass-intensity pairs from all the data files were subjected to PCA and the corresponding plots were genera ted (Figures 4-6 and 4-7). In the case of the C18-monolithic data, the PCA plot reveal ed the presence of four groups of samples; group 1 corresponding to days 2, 3, 5, and 6 in the life of the piglets, group 2 to day 4, group 3 to day 3, and group 4 to day 8. However, if a diagonal line is traced across this two dimensional plot, the four groups can be reduced into two: one for those data files corresponding to days 2-6 in the life of the piglets and the other for the files relate d to the days 7 and 8. Therefore, PCA was able to identify important changes in abundance across the first seven days of life. Similar results were obt ained when PCA was applied to the HILIC data, which confirmed the above-mentioned trends. Analysis of variance ANOVA was applied to m ake comparison betwee n multiple sample classes and extract the statistically significant mass-intens ity pairs. A Tukey post-hoc test was used to further locate the differences between any of these classes and th e significance of differen ces was expressed by the p-value. ANOVA was applied to the C18-monolithic and HILIC filtered mass-intensity pairs. Statistically significant ma sses between the different days in the life of the piglets were found (Figures 4-8 and 4-9). In the case of the C18-chromatography, a total of 1,672 mass-intensity pairs were found statistically si gnificant, these mass-intensity pairs distributed across the

PAGE 108

108 different days. On the other hand, for the HI LIC mass-intensity pairs, a total of 800 massintensity pairs were found differentially expressed. For both groups of statistical ly significant mass-intensit y pairs, a list called ANOVA masses was generated and submitted to database identification using mass accuracy-2 ppm mass toleranceas the crite ria for searching. The METLIN database search generated 313 (18.7%) hits for the C18-monolithic ANOVA masses and 198 (24.8%) hits for the HILIC ANOVA masses. Targeted Metabolomics Carnitinom ics : One of the aims of this study was to develop and apply a metabolomic workflow to targeted carnitinomics analysis usin g piglets as a model for neonatal acylcarnitine metabolism. The hypothesis to test was that the plasma carnitinom e would change between day 2 and day 8 of life. As mentioned earlier (Chapter 3), the interest was focu sed on the evaluation of the changes occurred on 11 acylcarnitines: carnitine (C0), acetylcarnitine (C2), propionylcarnitine (C3), butyrylcarnitine (C4), hexanoylcarnitine (C6), octanoylcarnitine (C8), decanoylcarnitine (C10), lauroylcarnitine (C12), myristoylcarnitine (C14), palmitoylcarnitine (C16), and stearoylcarnitine (C18); respectively. The instrument respon se, expressed as peak area counts, was extracted for each of the compounds in each of the data files. Using this information, the peak area counts (averaged) versus the day of life for each carnitine were plotted; thus profiles for carnitine and the different acy lcarnitines were generated (Fi gures 4-10 4-20); error bars correspond to 1 standard devia tion and the number of areas (n) extracted per analyte per day was 27. The day of life markedly af fected the relative levels of carnitines and acylcarnitines and different profiles were obtained; some of them showed a progressive increase of acylcarnitines (C0, C8, and C14) from day 2 to 8, whereas others show ed some interesting fluctuations (C2-C6, C10, C12, C16 and C18). In addition, the total area (%) versus the day of lif e was plotted for all the carnitines under study (Figure 4-21). In su mmary, carnitine and acylcarnitines changed

PAGE 109

109 significantly across the first se ven days of life and the biol ogical interpretation of this information is underway. Conclusions In this study, LC/MS-based gl obal and targeted m etabolomic workflows were successfully applied to assessing change s in the piglet plasma metabolome dur ing the first seven days of life. Results obtained after data processing and compound identification reve aled that there are differences in metabolite levels from day 2 to day 8. These results confirmed the initial hypothesis that the plasma metabolome w ould change within this time period. In the case of the global strategy, although stat istical differences were observed day after day, the most important changes happened for days 7 and 8 in life. When using different statistical tools, the da ta files corresponding to these two days discriminated very well against the others. After birth both piglets and humans must make the transition from using primarily carbohydrate as an energy source to the metabo lism of lipids from milk [126]. However, neonatal piglets differ from humans in the fact that they have very small amounts of adipose tissue at birth and experience an extremely rapid accumulation of fat before weaning. Also in contrast to humans, piglets do not experience hyperketonaemia as a result of dietary fat absorption. Studies have suggested the tendency of fatty acyl-CoA towards re-esterification for fat accumulation directly after birth instead of producing energy by beta-oxidation requiring the use of carnitine. This suggests that changes in acy lcarnitine levels may begin later in life as the piglet begins to accumulate fat and move towards increased lipid metabolism for energy. However, this trend was not directly observed in this study. Differences in acylcarnitines were observed over time, but the most differences were seen on days 4 and 7 with a decrease in some acylcarnitine differences on days 5, 6 and 8. Thes e unexpected decreases suggest that while the

PAGE 110

110 differences in acylcarnitines may be attributed to age there seem to be other factors influencing carnitinome changes after day 2. Also while th ese changes are not linear and differ among acylcarnitines they may still be important considerations in the analysis and diagnosis of metabolic diseases that result in small fluctua tions in metabolites according to disease state.

PAGE 111

111 Table 4-1. Piglet ID, da y of life, and MS ID. Piglet ID # Day of life MS ID # 362 2 7 370 2 7 371 2 7 357 3 8 361 3 8 369 3 8 358 4 9 363 4 9 368 4 9 356 5 10 359 5 10 376 5 10 360 6 11 364 6 11 365 6 11 373 7 12 374 7 12 375 7 12 366 8 13 367 8 13 372 8 13

PAGE 112

112 Figure 4-1. Study design for the evaluation of th e metabolomic changes from day 2 to 8 in the life of piglets. Plasma samples from tw enty one piglets were obtained. LC/MS approaches were applied for the analysis of three biological and three technical replicates per piglet the dataset consiste d in 189 files per chro matographic approach.

PAGE 113

113 Figure 4-2. Total ion chromatograms (TICs) for the overlapping of all pi glet samples (189 files) using C18-monolithic chromatography. Solvent A: 1% (v/v) acetic acid in water; Solv ent B: 1% (v/v) acetic acid in acetonitrile; Flow rate (mL min-1): 1.0. Injection volume (L): 15.0; Gradient: A:B (min): 95:5 (0.0-6 .5); 0:100 (25.5-35.0); 95:5 (55.0-65.0)

PAGE 114

114 Figure 4-3. Total ion chromatograms (TICs) for the overlapping of all piglet sample s (189 files) using HILIC chromatography. Solvent A: 7.5 mM ammonium formate in water; Solvent B: 7.5 mM ammonium formate in acetonitrile; Flow rate (mL min-1): 0.3. Injection volume (L): 10.0; Gradient: A:B (min): 10:90 (0.0-5.0); 50: 50 (25.0-30.0); 10:90 (40-45).

PAGE 115

115 Figure 4-4. Hierarchical clustering analysis (monolithic column, #masses: 15,260) resulting in grouping by MS ID# (color key b elow clustering tree) or day of life (color key on the right). The tree is clearly divided into several groups, some of them clustered all the data files for a specific day of life (days: 3, 4, 7, and 8) and some of them were divided into different subgroups (days 2, 5, and 6).

PAGE 116

116 Figure 4-5. Hierarchical cluste ring analysis (HILIC column, #m asses: 34,736) resulting in groupi ng by day of life (color key o n the right). The tree is divided into several groups, data files co rresponding to all of the days, except for days 2 and 8, were clustered together. Batches 1-7_HILIC (All Samples)

PAGE 117

117 Figure 4-6. Principal component analysis (m onolithic column). The two dimensional PCA plot shows four different group of samples, which corresponds to days 2, 3, 4, and 5 for group 1, day 4 for group 2, day 7 for group 3, and day 8 for group 4. If a red diagonal is traced, two major groups can be visualized: days 2-6 in the life of the piglets below the diagonal and days 7 and 8 above the diagonal.

PAGE 118

118 Figure 4-7. Principal component analysis (HILIC column). The rotated three dimensional PCA plot shows correlations across samp les from days 2-6 and between days 7 and 8. These resu lts are consistent with those obtained for the C18-monolithic approach.

PAGE 119

119 Figure 4-8. One-way analysis of variance (ANOVA) for C18-monolithic. Test type: parametric, do not assume variances equal; False discove ry rate: 0.05; Multiple testing correction: Benjamini and Hochberg false discovery rate; Post Hoc tes t: Tukey. The red boxes represent the sta tistically different masses, the grey boxes the total number of masses, and the blue boxes the sta tistically similar masses.

PAGE 120

120 Figure 4-9. One-way analysis of variance (ANOVA) for HILIC. Test type: parametric, do not assume variances equal; False discovery rate: 0.05; Multiple testing correction: Benjamini and Hochberg false discovery rate; Post Hoc tes t: Tukey. The red boxes represent the sta tistically different masses, the grey boxes the total number of masses, and the blue boxes the sta tistically similar masses.

PAGE 121

121 Figure 4-10. Targeted metabolomics an alysis. Carnitine profile (m/z = 162.1125). Figure 4-11. Targeted metabolomics analys is. Acetylcarnitine profile (m/z = 204.1230).

PAGE 122

122 Figure 4-12. Targeted metabolomics analysis Propionylcarnitine profile (m/z = 218.1387). Figure 4-13. Targeted metabolomics analysis. Butyrylcarnitine profile (m/z = 232.1543).

PAGE 123

123 Figure 4-14. Targeted metabolomics analys is. Hexanoylcarnitine profile (m/z = 260.1856). Figure 4-15. Targeted metabolomics analys is. Octanoylcarnitine pr ofile (m/z = 288.2169).

PAGE 124

124 Figure 4-16. Targeted metabolomics analys is. Decanoylcarnitine profile (m/z = 316.2482). Figure 4-17. Targeted metabolomics analys is. Lauroylcarnitine profile (m/z = 344.2795).

PAGE 125

125 Figure 4-18. Targeted metabolomics analysis Myristoylcarnitine profile (m/z = 372.3108). Figure 4-19. Targeted metabolomics analysis Palmitoylcarnitine pr ofile (m/z = 400.3421).

PAGE 126

126 Figure 4-20. Targeted metabolomics analys is. Stearoylcarnitine profile (m/z = 428.3734).

PAGE 127

127 Figure 4-21. Targeted metabolomics analysis. Carnitine and acylcarnit ines profiles between days 2 through 8 in th e life of piglets.

PAGE 128

128 CHAPTER 5 CONCLUDING REMARKS AND FUTURE WORK LC-MS m ethods were developed for improving the efficiency of separation of nonpolar and polar compounds in a metabolic context. High precision and accuracy of the proposed chromatographic approaches were demonstrated. Th e results clearly suggest that the use of both C18-monolithic and HILIC columns in HPLC ar e complementary and can serve as an advantageous complement to conve ntional methods for separation. The chromatographic methodologies describe d in the present manuscript dealt with samples without any kind of derivatization (neith er prior, nor after column separation), thus achieving a less complex, less expensive, and less time-consuming way to carry out the analysis of the metabolites. In addition, this fact different iates this study and its procedures from other approaches normally reported in the literature for the separation and/or determination of carnitine and carnitine-based compounds in clinical samples. Using the features provided by the proposed chromatographic techniques plus the massaccuracy and sensitivity characteristics of the oa-TOF mass spectrometer, targeted and untargeted metabolites were efficiently separated and accurately detected. The aim of this study was to achieve a comprehensive understanding of the metabolic role of known and unknown metabolites. The development of a reliable anal ytical workflow based on the utilization of cutting-edge technology together wi th sophisticated data processing and analysis software tools enabled the accomplishment of this goal. Putative biomarkers for the different applica tions were generated and their unambiguous identification is still underway. In fact, the uneq uivocal identification of metabolites is still one of the bottlenecks of any metabolomics study. In th is sense, it is wellknown that tandem mass spectrometry provides the ability to positively id entify the analytes with the correct compound

PAGE 129

129 assignments using the information provided by characteristics fragmentation patterns. In addition, NMR can generate valuable structural information, but much less sensitively than MS. Therefore, implementation of these analytical me thodologies to improve the identification of the tentative biomarkers is part of the future wo rk. In addition, different ionization methodologies have been demonstrated to contribute to the detection of unique metabolites. The application of ionization techniques other than ESI needs to be carried out. This study also generated many biological qu estions to be answered through further experimentation. Ideas have previously been prop osed for reasons why there may be more or less difference in levels of certain compounds. Th ese compounds may have important diagnostic and biological implications previous ly unknown, and the hypotheses generated from these analyses must be tested through other studies. While the changes in metabolites in plasma are important due to thei r clinical relevance, compartmentalization of metabolites in other tissues may also play a role in the interpretation of the data. Plasma fluctuations may be influe nced by the uptake of compounds into other biological media. Further studies on differences in putative metabolites in other tissues need to be conducted to apply fully the pr oposed analytical models to c linical care. Understanding the changes in metabolites and acylcarnitines as a re sult of a ketogenic diet may lead to further understanding of the mechanis m of ketogenic therapy. The fluctuations in differences among targeted metabolites seen in th is study may indicate a need for a more controlled experimental design when using humans or piglets as a model for patient populations. The findings de scribed would allow further biological implications to be made and investigated. These data are preliminary and must be investigated further in order to accurately analyze the results.

PAGE 130

130 REFERENCES [1] K. Dettm er, P. Aronov, B.D. Hammock, Mass Spectrom. Rev. 26 (2007) 51. [2] S.H. Gary Khoo, M. Al-Rubeai, Cell E ngineering 5, Systems Biology, Springer, Netherlands (2007). [3] C.W. Schmidt, Environ. Health Perspect. 112 (2004) A410-A415. [4] S.G. Olivier, M.K. Winson, D.B. Kell, F. Banganz, Trends Biotechnol. 16 (1998) 373. [5] D. Ryan, K. Robards, Anal. Chem. 78 (2006) 7954. [6] Fiehn, Comp. Funct. Genom. 2 (2001) 155. [7] M. Mayr, Circ. Cardiovasc. Genet. 1 (2008) 58-65. [8] W.B. Dunn, D.I. Ellis, Trends Anal. Chem. 24 (2005) 285. [9] R. Ramautar, G.W. Somsen, G.J. de Jong, Electrophoresis 30 (2009) 276. [10] R. Goodacre, S. Vaidyanathan, W.B. Dunn, G.H. Harrigan, D.B. Kell, Trends Biotechnol. 22 (2004) 245. [11] S. G. Villas-Boas, J. Nielsen, J. Smedsgaard, M.A.E. Hansen, U. Roessner-Tunali, Hoboken, New Jersey, John Wiley & Sons, Inc. (2007). [12] K. Dettmer, B.D. Hammock, Enviro n. Health Perspect. 112 (2004) A396. [13] I.M. Keseler, J. Collado-Vides, S. Gama-Castro, J. Ingraham, S. Paley, I.T. Paulsen, M. Peralta-Gil, P.D. Karp, Nu cl. Acids Res. 33 (2005) D334. [14] K-M. Oksman-Caldentey, D. Inze, M. Oresic, Proc. Natl. Acad. Sci. 101 (2004) 9949. [15] I.F. Duarte, J. Marques, A.F. Ladeirinha, C. Rocha, I. Lamego, R. Calheiros, T.M. Silva, M. P.M. Marques, J.B. Melo, I.M. Carre ira, A.M. Gil, Anal. Chem. 81 (2009) 5023. [16] G.G. Harrigan, R.H. LaPlante, G.N. Cosma, G. Cockerell, R. Goodacre, J.F. Maddox, J.P.Luyendyx, P.E. Ganey, R.A. Roth, Toxicol. Lett. 146 (2004) 197. [17] H.E. Johnson, D. Broadhurst, D.B. Kell, M. K. Theodorou, R.J. Me rry, G.W. Griffith, Environ. Microbiol. 70 (2004) 1583. [18] W. Lu, B.D. Bennett, J.D. Rabinowitz J. Chromatogr. B 871 (2008) 236. [19] A. Koulman, G. Woffedin, V.K. Narayana, H. Welchman, C. Crone, D. Volmer, Rapid Commun. Mass Spectrom. 23 (2009) 1411.

PAGE 131

131 [20] J. A, J. Trygg, J. Gullberg, A.I. Johansson, P. Jonsson, H.Antti, S.L. Marklund, T. Moritz, Anal. Chem. 77 (2005) 8086. [21] W. Wikoff, A.T. Anfora, J. Li u, P.G. Schultz, S.A. Lesley, E.C. Peters, G. Siuzdak, Proc. Natl. Acad. Sci. 106 (2009) 3698. [22] D.T.-T Nguyen, D. Guillarme, S. Heinisch, M-P. Barrioulet, J.-L. Rocca, S. Roudaz, J.-L. Veuthey, J. Chroma togr. A 1167 (2007) 76. [23] L.W. Summer, P. Mendes, R.A. Dixon, Phytochem. 62 (2003) 817. [24] A. Koulman, G.A. Lane, S.J. Harrison, D.A. Volmer, Anal. Bioanal. Chem. 394 (2009) 663. [25] R.J. Wenzel, U. Matter, L. Schultheis, R. Zenobi, Anal. Chem. 77 (2005) 4329. [26] W.E. Stephens, Phys. Rev. 69 (1946) 691. [27] A.E. Cameron, D.F. Eggers, Rev. Sci. Instrum. 19 (1948) 60. [28] W.C. Wiley, I.H. McLaren, Re v. Sci. Instrum. 26 (1955) 1150. [29] W.C. Wiley, Science 124 (1956) 817 [30] B.A. Mamyrin, V.I. Karataev, D.V. Shmikk, V.A. Aagulin, Sov. Phys. JETP 37 (1973), 45. [31] J.H.J. Dawson, M. Guilhaus, Rapi d Commun. Mass Spectrom. 3 (1989) 155. [32] A.F. Dodonov, I.V. Chernushevich, V.V. La iko, Proceeding of 12th International Mass Spectrometry Conference, Amst erdam, Netherlands, (1991) 153. [33] M. Guilhaus, D. Selby, V. Mlynski Mass Spectrom. Rev. 19 (2000) 65. [34] A.P. Bruins, J. Chromatogr. A 794 (1998) 345. [35] M. Dole, L.L. Mack, R. Hines, M.B. Alice, L.D. Ferguson, R.C. Mobley, J. Chem. Phys. 49 (1968) 2240. [36] J.B. Fenn, M. Yamashita, J. Chem. Phys. 88 (1984) 4451. [37] J.B. Fenn, C.K. Meng, Org. Mass Spectrom. 26 (1991) 542. [38] J.B. Fenn, T. Nohmi, J. Am. Chem. Soc. 114 (1992) 3241. [39] S. Nguyen, J.B. Fenn, Proc. Natl. Acad. Sci. 23 (2007) 1111. [40] N.J. Haskins, J. Pharm. Biomed. Anal. 25 (2001) 767.

PAGE 132

132 [41] C.E.C.A. Hop, R. Bakhtiar, Biospectroscopy 34 (1998) 259. [42] Chhabil Dass, Hoboken, Fundamentals of C ontemporary Mass Spectrometry, John Wiley & Sons, Inc., New Jersey, (2007). [43] S. Sanglier, H. Ramstrom, J. Haiech, E. Lei ze, A. Van Dorsselaer, Int. J. Mass Spectrom. 12208 (2002) 1. [44] W.E. Haskins, Z. Wang, C.J. Watson, R.R. Rostand, S.R. Witowski, D.H. Powell, R.T. Kennedy, Anal. Chem. 73 (2001) 5005. [45] J.H. Gross, Mass Spectrometry. A Text B ook, Springer, Heidelberg, Germany, (2004). [46] J.J. Kirkland, J. Chromatogr. Sci. 38 (2000) 535. [47] Y. Shen, R.D. Smith, El ectrophoresis 23 (2002) 3106. [48] K.K. Unger, R. Skudas, M.M. Schul te, J. Chromatogr. A 1184 (2008) 393. [49] S. Svec, C.G. Huber, Anal. Chem. 78 (2006) 2101. [50] E. Klodzinska, D. Moravcova, P. Jandera, B. Buszeewski, J. Cromatogr. A 1109 (2006) 51. [51] F. Svec, L. Geiser, R.E. Majors, LC GC N. Am. 24 (2006) 22. [52] L. Rieux, D. Lubda, H.A.G. Niederlnder, E. Verpoorte, Rainer Bischoff, J. Chromatogr. A 1120 (2006) 165. [53] A.M. Siouffi, J. Chromatogr. A 1126 (2006) 86. [54] N. Wu, J. Dempsey, P.M. Yehl, A. Dovletoglou D. Ellison, J. Wyvratt, Anal. Chim. Acta 523 (2004) 149. [55] J. Billen, P. Gzil, G.V. Baron, G. De Smet, J. Chromatogr. A 1077 (2005) 28. [56] K. Cabrera, J. Sep. Sci. 27 (2004) 843. [57] M. Al-Bokari, D. Cherrak, G. Guiochon, J. Chromatogr. A 975 (2002) 275. [58] A.M. van Nederkassel, A. Aerts, A. Dieric k, D.L. Massart, Y.Van der Heyden, J. Pharm. Biomed. Anal. 32 (2003) 233. [59] S. El Deeb, H. Watzig, Turk. J. Chem. 30 (2006) 543. [60] V.V. Tolstikov, A. Lommen, K. Nakanishi, N. Tanaka, O. Fiehn, Anal. Chem. 75 (2003) 23.

PAGE 133

133 [61] J.K. Nicholson, P.J. Foxall, M. Spraul, R.D. Farrant, J.C. Lindon, Anal. Chem. 67 (1995) 793. [62] E. Holmes, P.J. Foxall, M. Spraul, R.D. Fa rrant, J.K. Nicholson, J.C. Lindon, J. Pharm. Biomed. Anal. 15 (1997) 1647. [63] M. Godejohann, J. Chro matogr. A 1156 (2007) 87. [64] T. Ikegami, H. Fujita, K. Horie, K. Hosoya N. Tanaka, Anal. Bioanal. Chem. 386 (2006) 578. [65] F. Beaudry, P. Vachon, Biomed. Chromatogr. 20 (2006) 200. [66] B.A. Olsen, J. Chromatogr. A 913 (2001) 113. [67] J. Layne, J. Chroma togr. A 957 (2002) 149. [68] E.S. Grumbach, D.M. Wagrowski-Diehl, J.R. Mazzeo, B. Alden, P.C. Iraneta, LC GC N. Am. 22 (2004) 1010. [69] A.J. Alpert, J. Chromatogr. 499 (2007) 46. [70] D.V. McCalley, J. Chro matogr. A 1171 (1996) 378. [71] A.R Oyler, B.L. Armstrong, J.Y. Cha, M. X. Zhou, Q. Yang, R.I. Robinson, R. Dunphy, D.J. Burinsky, J. Chromatogr. A 724 (1996) 378. [72] T. Yoshida, J. Biochem. Biophys. Methods 60 (2004) 265. [73] B.Y. Zhu, C.T. Mant, R.S. Hodges, J. Chromatogr. 594 (1992) 75. [74] S.C. Churms, J. Chromatogr. A 720 (1996) 75. [75] H. Lindner, B. Sarg, C. Meraner, W. Helliger, J. Chromatogr. A 743 (1996) 137. [76] V.V. Tolstikov, O. Fiehn, Anal. Biochem. 301 (2002) 298. [77] J.K. Troyer, K.K. Stephenson, J.W. Fahey, J. Chromatogr. A 919 (2001) 299. [78] P. Hemstrom, K. Irgum, J. Sep. Sci. 29 (2006) 1784. [79] Y. Lin, D. Si, Z. Zhang, C. Liu, Toxicol. 256 (2009) 191. [80] J.J. Pesek, M.T. Matyska, S.M. Fischer, T.R. Sana, J. Chromatogr. A 1204 (2008) 48. [81] F.B. Collings, V.S. Vaidya, Toxicol. 245 (2008) 167. [82] C. Hoppel, Am. J. Kidney Dis. 41 (2003) S4S12.

PAGE 134

134 [83] C.R. Roe, C.L. Hoppel, T.E. Stacey, R.A. Chalmers, B.M. Tracey, D.S. Millington, Arch. Dis. Child 58 (1983) 916. [84] L. Vernez, M. Dickenmann, J. Steiger, M. Wenk, S. Krahenbuhl, Nephrol. Dial. Transplant. 21 (2006) 450. [85] C.L. Hoppel, Techniques in Diagnostic Hu man Biochemical Genetics: A Laboratory Manual, Wiley-Liss, New York, (1991). [86] L. Vernez, M. Wenk, S. Krahenbuhl, Ra pid Commun. Mass Spectrom. 18 (2004) 1233. [87] S. Ahmad, Semin. Dial. 14 (2001) 209. [88] P.E. Minkler, J. Kerner, K. North, C.L. Hoppel, Clin. Chim. Acta 352 (2005) 81. [89] M. Yoshino,Y. Tokunaga,Y. Watanabe,I. Yoshida, M. Sakaguchi, I. Hata,Y. Shigematsu, S. Yamaguchi, J. Chromatogr. B 792 (2003) 73. [90] S. Vijay, A. Patterson, S. Olpin, M.J. Henderson, S. Clark, C. Day, G. Savill, J.H. Walter, J. Inher. Metab. Dis. 29 (2006) 627. [91] T. Werner, I. Treiss, D. Kohlmueller, P. Mehlem, M. Teich, E. Longin, T. Gerstner, A. Schulze, Epilepsia 48 (2007) 72. [92] K. Imura, A. Okada, Nutrition 14 (1998) 143. [93] R. Lourenco, M.E. Camilo, Nutr. Hosp. 17 (2002) 262. [94] T.E. Creighton, Protein Folding, W.H. Freeman & Co. New York, (1992). [95] R.P.W. Scott, Techniques and Practices of Chromatography, Marcel Dekker, New York, (1995). [96] V.R. Meyer, Practical High-performance Liquid Chromatography, Wiley, New York, (1994). [97] M.E. Swartz, LC GC N. Am. (2005) 8. [98] Department of Health and Human Serv ices, Food and Drug Administration. Good laboratory practice regulati ons. Final Rule Federal Register 52 (1987) 33768. [99] P.E. Minkler, S.T. Ingalls, C.L. Hoppel, Anal. Chem. 77 (2005) 1448. [100] S. Cerutti, Masters Thesis, University of Florida, Florida (2007). [101] P. Hemstrom, K. Irgum, J. Sep. Sci. 29 (2006) 1784. [102] Y. Guo, J. Liq. Chromatogr. & Rel. Tech. 28 (2005) 497.

PAGE 135

135 [103] X. Wang, W. Li, H.T. Rasmusse n, J. Chromatogr. A 1083 (2005) 58. [104] Y. Guo, S. Gaiki, J. Chromatogr. A 1074 (2005) 71. [105] P. Molander, R. Trones, K. Haugla nd, T. Greibrokk, Analyst 124 (1999) 1137. [106] J.B. German, B.D. Hammock, S.M. Watkins, Metabolomics 1 (2005) 3. [107] S.M. Watkins, P.R. Reisfsnyder, H.J. Pan, J.B. Leiter, J. Lipid Res. 43 (2002) 1809. [108] S.M. Watkins, X. Zhu, S.H. Zeisel, J. Nutr. 133 (2003) 3386. [109] J.M. Freeman, E.H. Kossoff, A.L. Hartman, Pediatrics 119 (2007) 535-543. [110] J.M. Freema, E.P. Vining, D.J. Pillas, J.C. Casey, L.M. Kelly, Pediatrics 102 (1998) 1358. [111] C.B. Henderson, F.M. Filloux, S.C. Alder, J. L. Lyon, D.A. Caplin, J. Child Neurol. 21 (2006) 193. [112] D.L. Keene, Pediatr. Neurol. 35 (2006) 1. [113] K.J. Bough, J.M. Rho, Epilepsia 48 (2007) 43. [114] A.L. Hartman, M. Gasior, E.P. Vining, M.A. Rogawski, Pediatr. Neurol. 36 (2007) 281. [115] A.A. Morris, J. Inher. Met. Dis. 28 (2005) 109. [116] A. Virmani, Z. Binienda, Mol. Aspects Med. 25 (2004) 533. [117] GeneSpring MS Manual, Agilent Technologies, California, (2006). [118] M.J. Baxter, Archaeometry 4 (2006) 671. [119] X. Zhao, A. Peter, J. Fritsche, M. Elcnerova A. Fritsche, H.U. Haring, E.D. Schleicher, G. Xu, R. Lehmann, Am. J. Physiol. Endocrinol. Metab. 296 (2009) 384. [120] B.S. Fuehrlein, M.S. Rutenberg, J.N. Silver, M.W. Warren, D.W. Theriaque, G.E. Duncan, P.W. Stacpoole, M.L. Brantly. J. Clin. Endocrinol. Metab. 89 (2004) 89 1641. [121] Bodamer, A. Muhl, Monatshefte fur Chemie 136 (2005) 1293. [122] A. Schulze, M. Lindner, D. Kohlmuller, K. Olgemoller, E. Mayatepek, G.F. Hoffmann, Pediatrics 111 (2003) 1399. [123] H-W. Hsu, T.H. Zytkovicz, A.M. Comeau, A.W. Strauss, D. Marsden, V.E. Shih, G.F. Grady, R.B. Eaton, Pediat rics 121 (2008) e1108. [124] J.E. Cooper, Lab. Anim. 9 (1975) 329.

PAGE 136

136 [125] S.A. Book, L.K. Bristad, J. Anim. Sci. 38 (1974) 997. [126] E.M. Glauser, Exp. Med. Surg. 24 (1966) 181. [127] P.R. Borum, Nutr. Rev. 39 (1981) 385. [128] J.K. Baltzell, F.W. Bazer, S.G. Miguel, P.R. Borum, J. Nutr. 117 (1987) 754.

PAGE 137

137 BIOGRAPHICAL SKETCH Estela So ledad Cerutti was born in the beauti ful province of San Luis, Argentina. She was a curious child and discovered he r fascination for chemistry at an early age. The search for explanations, the magic colors behind the chem ical reactions, and the understanding that we all are a chemistry factory led Soleda d to choose chemistry as her major. Soledad attended Universidad Nacional de Sa n Luis, graduating with a Bachelors degree in chemistry in March, 2002. After graduation, sh e immediately started working toward a Ph.D. and completed her studies with a doctora l degree in chemistry on December 15, 2006. Soledad realized the importance of travel to United States to expl ore the field of mass spectrometry. After applying and being awarded with a Fulbright scholarship, she moved to Gainesville and started with her Masters program at University of Florida and completed these studies in August, 2007. Positive personal and sc ientific experiences convinced Soledad to continue with her research toward her Ph.D. de gree in mass spectrometry under the supervision of Dr. Richard Yost and Dr. David Powell at University of Florida.