<%BANNER%>

LC/MS Optimization and Implementation for Metabolomic Applications

Permanent Link: http://ufdc.ufl.edu/UFE0044559/00001

Material Information

Title: LC/MS Optimization and Implementation for Metabolomic Applications
Physical Description: 1 online resource (163 p.)
Language: english
Creator: Elliott, Noelle Marie
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2012

Subjects

Subjects / Keywords: blood -- chromatography -- liquid -- mass -- metabolomics -- plasma -- spectrometry -- urine
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: The study of metabolomics aims to detect and identify asmany compounds as possible in biofluids or tissues in order to determine thechanges occurring within an organism in response to some perturbance.  This is commonly performed by liquidchromatography/mass spectrometry (LC/MS) or nuclear magnetic resonancespectroscopy (NMR) followed by statistical analysis of the data.  The field of metabolomics has growndrastically over the last decade, including the formation of the MetabolomicsSociety in 2004 and the start of the Metabolomicsjournal in March of 2005. This work employed LC/MS to investigate the metabolomics ofseveral different biological studies on biofluids.  The metabolomics of piglet plasma and redblood cells were compared over the first seven days of life and plasma wasindicated to be a better sampling technique, depending on the application.  Another study compared the plasma from humanpatients with colon cancer to healthy controls and five potential biomarkerswere identified including dodecenoylcarnitine, myristoleylcarnitine, myristoylcarnitine,palmitoleylcarnitine, and linoleoylcarnitine. Manatee urine was also examined to look for potential biomarkers in estrousurine that may cause the male manatee to easily find a female manatee in estrus.  In manatee urine three potential biomarkerswere identified as creatine, glutamic acid, and cytidine.  LC with MS/MS instrumentation was used inconjunction with accurate mass to identify metabolites determined to besignificant in the colon cancer and manatee urine projects. In addition to the metabolomics applications, some LC/MSoptimization was performed.  The maingoal of the optimization was to increase the number of detectable molecular features,while maintaining an efficient chromatographic run time.  Particularly, the positive mode electrospraysettings were optimized and the LC run time was shortened.
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 Noelle Marie Elliott.
Thesis: Thesis (Ph.D.)--University of Florida, 2012.
Local: Adviser: Yost, Richard A.

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2012
System ID: UFE0044559:00001

Permanent Link: http://ufdc.ufl.edu/UFE0044559/00001

Material Information

Title: LC/MS Optimization and Implementation for Metabolomic Applications
Physical Description: 1 online resource (163 p.)
Language: english
Creator: Elliott, Noelle Marie
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2012

Subjects

Subjects / Keywords: blood -- chromatography -- liquid -- mass -- metabolomics -- plasma -- spectrometry -- urine
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: The study of metabolomics aims to detect and identify asmany compounds as possible in biofluids or tissues in order to determine thechanges occurring within an organism in response to some perturbance.  This is commonly performed by liquidchromatography/mass spectrometry (LC/MS) or nuclear magnetic resonancespectroscopy (NMR) followed by statistical analysis of the data.  The field of metabolomics has growndrastically over the last decade, including the formation of the MetabolomicsSociety in 2004 and the start of the Metabolomicsjournal in March of 2005. This work employed LC/MS to investigate the metabolomics ofseveral different biological studies on biofluids.  The metabolomics of piglet plasma and redblood cells were compared over the first seven days of life and plasma wasindicated to be a better sampling technique, depending on the application.  Another study compared the plasma from humanpatients with colon cancer to healthy controls and five potential biomarkerswere identified including dodecenoylcarnitine, myristoleylcarnitine, myristoylcarnitine,palmitoleylcarnitine, and linoleoylcarnitine. Manatee urine was also examined to look for potential biomarkers in estrousurine that may cause the male manatee to easily find a female manatee in estrus.  In manatee urine three potential biomarkerswere identified as creatine, glutamic acid, and cytidine.  LC with MS/MS instrumentation was used inconjunction with accurate mass to identify metabolites determined to besignificant in the colon cancer and manatee urine projects. In addition to the metabolomics applications, some LC/MSoptimization was performed.  The maingoal of the optimization was to increase the number of detectable molecular features,while maintaining an efficient chromatographic run time.  Particularly, the positive mode electrospraysettings were optimized and the LC run time was shortened.
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 Noelle Marie Elliott.
Thesis: Thesis (Ph.D.)--University of Florida, 2012.
Local: Adviser: Yost, Richard A.

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2012
System ID: UFE0044559:00001


This item has the following downloads:


Full Text

PAGE 1

1 LC/MS OPTIMI ZATION AND IMPLEMENTATION FOR METABOLOMIC APPLICATIONS By NOELLE MARIE ELLIOTT A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE D EGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2012

PAGE 2

2 2012 Noelle Marie Elliott

PAGE 3

3 To my loving and supportive family

PAGE 4

4 ACKNOWLEDGMENTS I would like to thank those people who have supported me both academically and emotionally throughou t my graduate studies. First, I would like to thank my advisors. I am thankful to Dr. David Powell for the patience shown to me during my time in his lab, the insightful discussions, and the knowledge that was shared. I am grateful to Dr. Richard Yost f or the academic advice that was offered to me and the interesting ideas that were offer ed. I also appreciate the time spent discussing metabolomics with Dr. Peggy Borum and the samples provided to me by her. I would also like to thank Dr. Ben Smith and D r. Nick Polfer for their helpful comments while serving on my committee. I am grateful for the samples provided to me by Dr. John Koomen and Dr. Erin Siegel at Moffitt Cancer Center, as well as the opportunity to work in lab for several days. The f riends that I have made in graduate school deserve to be acknowledged for their support both in the lab and outside of it. The Powell and Yost group s have been a supportive second family to me during my time at UF. I am grateful for the thorough training given to me by Dr. Soledad Cerutti. Dr. Julia Rummel should also be thanked for her scientific insight and help early in my graduate studies. The Yost group was particularly helpful with editing o f docume nts and presentations. I would especially like to thank Dr. Kyle Ann Mino and Robert Menger for their assistance with editing and more importantly, their friendship s Lastly I would like to thank my family for their love and never ending faith in me. ways believed in me, and I will always be grateful for that. My husband has also been very supportive of me; this is most apparent in his agreement to relocate from Pennsylvania to Florida so that I could

PAGE 5

5 continue my education I cannot imagine being her e without him next to me. I look forward to starting the next chapter of our lives together.

PAGE 6

6 TABLE OF CONTENTS page ACKNOWLEDGMENTS ................................ ................................ ................................ .. 4 LIST OF TABLES ................................ ................................ ................................ ............ 9 LIST OF FIGURES ................................ ................................ ................................ ........ 12 ABSTRACT ................................ ................................ ................................ ................... 16 CHAPTER 1 INTRODUCTION ................................ ................................ ................................ .... 18 Metabolomics ................................ ................................ ................................ .......... 18 Background ................................ ................................ ................................ ...... 18 Approaches ................................ ................................ ................................ ...... 19 Platforms ................................ ................................ ................................ .......... 20 Biological Samples ................................ ................................ ................................ 20 High Performance Liquid Chromatography ................................ ............................. 22 LC Theory ................................ ................................ ................................ ......... 22 Monolithic and Small Particle HPLC Columns ................................ .................. 25 Electrospray Ionization ................................ ................................ ............................ 26 Time of Flight Mass Spectrometry ................................ ................................ .......... 26 Resolution and Mass Accuracy ................................ ................................ ........ 27 TOF MS Da ta Acquisition ................................ ................................ ................. 28 TOF MS System ................................ ................................ ............................... 28 Data Analysis ................................ ................................ ................................ .......... 29 Principal Compo nent Analysis ................................ ................................ .......... 30 Partial Least Squares Discriminant Analysis ................................ .................... 31 t test ................................ ................................ ................................ 32 Analysis of Variance ................................ ................................ ......................... 33 Benjamini and Hochberg False Discovery Rate ................................ ............... 34 2 COMPARISON OF THE METABOLOMICS OF PLASMA AND RED BLOOD CELLS FROM PIGLETS IN THE FIRST SEVEN DAYS OF LIFE .......................... 42 Background ................................ ................................ ................................ ............. 42 Experimental ................................ ................................ ................................ ........... 43 Sample Collection and Preparation ................................ ................................ .. 43 Liquid Chromatography Conditions ................................ ................................ .. 44 Electrospray Source ................................ ................................ ......................... 44 Mass Spectrometer ................................ ................................ .......................... 45 Data Analysis ................................ ................................ ................................ ... 45 Results and Discussion ................................ ................................ ........................... 46

PAGE 7

7 Piglet Plasma ................................ ................................ ................................ ... 46 Piglet Red Blood Cells ................................ ................................ ...................... 48 Database Search of Significan t Piglet Plasma and RBC Masses ..................... 48 Targeted Analysis ................................ ................................ ............................. 49 Summary ................................ ................................ ................................ ................ 50 3 METABOLOMICS OF COLON CANCER BY ANALYSIS OF HUMAN PLASMA ... 65 Overview ................................ ................................ ................................ ................. 65 Experimental ................................ ................................ ................................ ........... 69 Sample Collection and Preparation ................................ ................................ .. 69 Liquid Chromatography Conditions ................................ ................................ .. 69 Electrospray Source ................................ ................................ ......................... 70 Mass Spectrometer ................................ ................................ .......................... 70 MS/MS Instrumentation ................................ ................................ .................... 70 Data Analysis ................................ ................................ ................................ ... 70 Results and Discussion ................................ ................................ ........................... 71 Influential PLS DA Masses ................................ ................................ ............... 72 Significant Masses with High est FC ................................ ................................ 73 Database Search of Significant Masses ................................ ........................... 74 MS/MS of Significant Masses ................................ ................................ ........... 75 Biological Significance ................................ ................................ ...................... 77 Challenges Faced ................................ ................................ ............................ 78 Summary ................................ ................................ ................................ ................ 80 4 METHOD OPTIMIZATION: TOWARDS MORE EFFICIENT IONIZATION AND SHORTER ANALYSIS TIME ................................ ................................ .................. 97 Overview ................................ ................................ ................................ ................. 97 Experimental ................................ ................................ ................................ ........... 98 Chromatography Methods ................................ ................................ ................ 98 Electrospray Source ................................ ................................ ......................... 99 Mass Spectrometer ................................ ................................ .......................... 99 Capillary Voltage ................................ ................................ .............................. 99 Fragmentation Voltage ................................ ................................ ................... 100 Feature Extraction of Urin e Samples ................................ ....................... 100 Feature Extraction of Plasma Samples ................................ .................... 101 Comparison of Monolithic and Small Particle Columns ................................ .. 101 Evaluation of Different Mobile Phases for use with the Zorbax Column ......... 102 Optimization of Solvent Systems for use with HILIC ................................ ...... 103 Results and Discussion ................................ ................................ ......................... 104 Capillary and Fragmentor Voltages ................................ ................................ 104 Feature Extraction of B iofluids ................................ ................................ ........ 105 Implementing Different LC Columns ................................ ............................... 106 Use of Different Solvents with Zorbax Column ................................ ............... 107 Optimization of HILIC ................................ ................................ ..................... 108 Summary ................................ ................................ ................................ .............. 109

PAGE 8

8 5 METABOLOMICS OF ESTROUS AND ANESTROUS URINE FROM MANATEES ................................ ................................ ................................ .......... 132 Background ................................ ................................ ................................ ........... 132 Experimental ................................ ................................ ................................ ......... 134 Sample Collection and Preparat ion ................................ ................................ 134 Liquid Chromatography Conditions ................................ ................................ 134 Electrospray Source ................................ ................................ ....................... 135 Mass Spectrometer ................................ ................................ ........................ 135 MS/MS Instrumentation ................................ ................................ .................. 135 Results and Discussion ................................ ................................ ......................... 135 RP/MS ................................ ................................ ................................ ............ 135 HILIC/MS ................................ ................................ ................................ ........ 138 Summary ................................ ................................ ................................ .............. 140 6 CONCLUSIONS AND FUTU RE DIRECTIONS ................................ .................... 154 Piglet Plasma and RBC Comparison ................................ ................................ .... 154 Metabolomics of Colon Cancer ................................ ................................ ............. 155 Method Optimization for Metabolomic Applications ................................ .............. 155 Metabolomics of Estrous and Anestrous Manatee Urine ................................ ...... 155 LIST OF REFERENCES ................................ ................................ ............................. 157 BIOGRAPHICAL SKETCH ................................ ................................ .......................... 163

PAGE 9

9 LIST OF TABLES Table page 1 2 Typical se t up for ANOVA. ................................ ................................ ................ 35 1 3 Possible outcomes of individual tests. ................................ ................................ 35 2 1 Liquid chromatography gradient. ................................ ................................ ........ 52 2 2 Results from ANOVA for piglet plasma. ................................ .............................. 52 2 3 PLS DA prediction results for piglet plasma. ................................ ...................... 53 2 4 Results from ANOVA for piglet RBCs. ................................ ................................ 53 2 5 PLS DA prediction results for piglet RBCs. ................................ ........................ 54 3 1 Sample demographics ................................ ................................ ........................ 81 3 2 PLS DA prediction results for human plasma from healthy controls and colon cancer patients. ................................ ................................ ................................ .. 81 3 3 Potential formulas from [M+H] + d atabase search of influential PLS DA masses. ................................ ................................ ................................ .............. 82 3 4 Potential formulas from [M+Na] + database search of influential PLS DA masses. ................................ ................................ ................................ .............. 82 3 5 Potential formulas from [M+H]+ database search of masses with the highest FC. ................................ ................................ ................................ ...................... 83 3 6 Potential formulas from [M+Na] + database search of masses with the highest FC. ................................ ................................ ................................ ...................... 83 3 7 List of metabolites identified by accurate mass and MS/MS fragmentation from the PLS DA influential mass list. ................................ ................................ 84 4 1 Chromatography methods used to optimize ESI settings for standards and biofluids. ................................ ................................ ................................ ........... 110 4 2 Chromatography methods used to compare the monolithic and small particle column. ................................ ................................ ................................ ............. 111 4 3 Chromatography methods used to optimize a HILIC method. .......................... 112 4 4 Average number of features extracted with manatee urine by the monolithic and short monolithic methods at v arying capillary/fragmentor voltages. ........... 113

PAGE 10

10 4 5 Average number of features extracted with piglet plasma by the monolithic and short monolithic methods at varying capillary/fragmentor voltages. ........... 113 4 6 List of retention times, peak widths and column efficiencies for the piglet plasma analysis with the monolithic method. ................................ .................... 114 4 7 List of retention times, peak widths and column efficiencies for the piglet plasma analysis w ith the short monolithic method ................................ ............ 114 4 8 Average number of features extracted with piglet plasma by four LC methods. ................................ ................................ ................................ ........... 115 4 9 List of retention times, peak widths and plate heights for carnitine standard mixture using the MC method. ................................ ................................ .......... 115 4 10 List of retention times, peak widths and plate heights for carnitine standard mixture using the MC with ZC Conditions method. ................................ ........... 116 4 11 List of retention times, peak widths and pla te heights for carnitine standard mixture using the ZC method. ................................ ................................ ........... 116 4 12 List of retention times, peak widths and plate heights for carnitine standard mixture using the ZC with MC Conditions met hod. ................................ ........... 117 4 13 List of retention times, peak widths and plate heights for five randomly chosen m/z ................................ ................................ 117 4 14 List of retention times, peak widths and plate heights for five randomly chosen m/z ................................ .. 118 4 15 List of retention times, peak widths and plate heights for five randomly chosen m/z ................................ ................................ 118 4 16 List of retention times, peak widths and plate heights for five randomly chosen m/z ................................ .. 119 4 17 Average number of features extracted from piglet plasma analyzed with the Zorbax column and three different solvent sets. ................................ ............... 119 4 18 List of retention times, peak widths and theoretical plates for carnitine standard mixture using 1% AA in H2O and ACN. ................................ ............. 120 4 19 List of retention times, peak widths and theoretical plates for carnitine standard mixture using 0.1% AA in H2O and ACN. ................................ .......... 120 4 20 List of retention times, peak widths and theoretical plates for carnitine standard mixture using 0.2% AA in H2O and MeOH. ................................ ....... 121

PAGE 11

11 4 21 List of retention times, peak widths and theoretical plates for five randomly chosen m/z 2 O and ACN. ................................ ................... 121 4 2 2 List of retention times, peak widths and theoretical plates for five randomly chosen m/z 2 O and ACN. ................................ ................ 122 4 23 List of retention times, peak widths and theoretical plates f or five randomly chosen m/z 2 O and MeOH. ................................ ............. 122 5 1 PLS DA prediction results for urine from manatees in estrus and anestrus by RPLC/MS. ................................ ................................ ................................ ......... 142 5 2 PLS DA prediction results for urine from manatees in estrus and anestrus by HILIC/MS. ................................ ................................ ................................ ......... 142 5 3 Overview of identified significant masses elevated in estrous urine by RP/MS and HILIC/MS. ................................ ................................ ................................ .. 142

PAGE 12

12 LIST OF FIGURES Figure page 1 1 Depiction of the number of publications in metabolomics each year, for the past 10 years ................................ ................................ ................................ ...... 36 1 2 The separation of whole blood into its three components ................................ ... 36 1 3 Comparison of the diffusional path of monolithic based colu mns and small particle columns ................................ ................................ ................................ .. 37 1 4 Depiction of dead time, retention time and peak widths in LC ............................ 37 1 5 Schematic of the electro spray source ................................ ................................ 38 1 6 Schematic of the Agilent 6210 time of flight mass spectrometer ........................ 39 1 7 Generic 3D PCA plot of a metabolomics ex periment with three groups ............. 40 1 8 Threshold p values for three different multiple testing corrections. ..................... 41 2 1 Overlay of piglet pl asma chromatograms. ................................ .......................... 55 2 2 Overlay of piglet RBC chromatograms. ................................ .............................. 55 2 3 3D PCA plot of ANOVA masses in piglet plasma ................................ ............... 56 2 4 2D PCA plot of ANOVA masses in piglet plasma ................................ ............... 56 2 5 3D PLS DA plot of ANOVA masses in piglet plasma. ................................ ........ 57 2 6 2D PLS DA plot of ANOVA masses in piglet plasma ................................ ......... 57 2 7 PLS DA loadings plot of ANOVA masses in piglet plasma ................................ 58 2 8 3D PCA plot of ANOVA masses in piglet RBCs ................................ ................. 58 2 9 2D PCA plot of ANOVA masses in piglet RBCs ................................ ................ 59 2 10 3D PLS D A plot of ANOVA masses in piglet RBCs ................................ ............ 59 2 11 2D PLS DA plot of ANOVA masses in piglet RBCs ................................ ............ 60 2 12 PLS DA loadings plot of ANOVA masses in piglet RBCs ................................ ... 60 2 13 Bar graphs of carnitine response in piglet plasma and RBC .............................. 61 2 14 Bar graphs of octanoylcarnitine re sp onse in piglet plasma and RBC ................ 61

PAGE 13

13 2 15 Bar graphs of decanoylcarnitine response in piglet plasma and RB C ............... 62 2 16 Bar graphs of lauroylcarnitine response in piglet plasma and RBC ................... 62 2 17 Bar graphs of myristoylcarnitine response in piglet plasma and RBC. ............... 63 2 18 Bar graphs of palmitoylcarnitine response in piglet plasma and RBC ............... 63 2 19 Bar graphs of stearoylcarnitine resp onse in piglet plasma and RBC ................. 64 3 1 Areas affected by a nd used in staging colon cancer ................................ .......... 85 3 2 3D PCA plot of the significant masses in plasma from colon canc er patients and healthy controls ................................ ................................ ........................... 85 3 3 2D PCA plot of the significant masses in plasma from colon cancer patients and healthy controls ................................ ................................ ........................... 86 3 4 3D PLS DA plot of t test ( p = 0.0 1) masses in human plasma from healthy controls and colon cancer patients ................................ ................................ ..... 86 3 5 2 PLS DA plot of t test ( p = 0.01) masses in human plasma from healthy controls and colon cancer patients ................................ ................................ ..... 87 3 6 PLS DA loadings plot of the significant masses determined by a t test ( p = 0.01) in human plasma from healthy controls and colon cancer patients ........... 87 3 7 Bar graph of the average abundances of masses furthest from W C0 = 0 in the PLS loadings plot ................................ ................................ .......................... 88 3 8 Bar graph of the average abundances of the significant masses with the highest fold change ................................ ................................ ............................ 89 3 9 Schematic of the Velos dual pressure LIT ................................ .......................... 90 3 10 MS/MS spectrum of m/z 370 in plasm a from a colon canc er patient. ................. 91 3 11 MS/MS spectra of m/z 372 in plasma from a colon cancer patient (top) and a standard (bottom). ................................ ................................ .............................. 92 3 12 MS/MS sp ectrum of m/z 398 in plasm a from a colon cancer patient. ................. 93 3 13 MS/MS spectrum of m/z 424 in plasm a from a colon cancer patient. ................. 94 3 14 MS/MS spectra of m/z 344 in plasma from a colon cancer patient (top) and a stan dard (bottom) ................................ ................................ ............................... 95 3 15 Carnitine palmitoyltransferase (CPT) enzyme activity in the mitochondria ......... 96

PAGE 14

14 4 1 Average peak areas of different carnitines as a function of capillary voltage. .. 123 4 2 Average peak areas of different carnitines as a function of fragmentor vo ltage. ................................ ................................ ................................ ............. 124 4 3 Base peak chromatograms of standard mixtures of carnitines using four different LC methods ................................ ................................ ........................ 125 4 4 Base peak chromatograms of piglet plasma using four different LC methods 126 4 5 Base peak chromatograms of standard mixtures of carnitines using three different solvent sets and Zor bax column. ................................ ........................ 127 4 6 Base peak chromatograms of piglet plasma using three different solvent sets an d Zorbax column ................................ ................................ .......................... 128 4 7 Average efficiencies (N) of carnitine standards for HILIC with four different solvent systems ................................ ................................ ................................ 129 4 8 Average efficiencies (N) of amino acid standards for HILIC with f our different solvent systems ................................ ................................ ............................... 130 4 9 Average retention times of standards for HILIC with four different solvent systems ................................ ................................ ................................ ........... 131 5 1 Chromatograms of manatee urine f rom animals in estrus using RPLC. ........... 143 5 2 Chromatograms of manatee urine from animals in anestrus using RPLC. ....... 143 5 3 3D PCA plot of the significant masses in urine from manatees in estrus and anestrus (9 masses, p = 0.05) by RPLC/MS ................................ .................... 144 5 4 2D PCA plot of the significant masses in urine from manatees in estrus and anestr us (9 masses, p = 0.05) by RPL C/MS ................................ ................... 144 5 5 3D PLS DA plot of t test ( p = 0.05) masses in urine from manatees in estrus and anestrus (9 masses) by RPLC/MS. ................................ ........................... 145 5 6 2 PLS DA plot of t test ( p = 0.05) masses in manatees in estrus and anestrus (9 masses) by RPLC/MS. ................................ ................................ ................. 145 5 7 Bar graph comparing the average peak areas of the 9 significant masses in estrous and anestrous urine by RPLC/MS.. ................................ ..................... 146 5 8 MS/MS spectrum of m/z 175 in estrous urine, determined by RPLC/MS ........ 146 5 9 MS/MS spectrum of m/z 143 in estrous urine determined by RPLC/MS ........ 147 5 10 Chromatograms of manatee urine from animals in estrus using HILIC. ........... 147

PAGE 15

15 5 11 Chromatograms of manatee urine from animals in anestrus using HILIC. ....... 148 5 12 3D PCA plot of the significant masses in urine from manatees in estrus and a nestrus by HILIC/MS (30 masses, p = 0.05) ................................ .................. 148 5 13 2D PCA plot of the significant masses in urine from manatees in estrus and anestrus (30 masses, p = 0.05) by HILIC/MS ................................ .................. 149 5 14 3D PLS DA plot of t test ( p = 0.05) masses in urine from manatees in estrus and anestrus (30 masses) by HILIC/MS. ................................ .......................... 149 5 15 2 PLS DA plot of t t est ( p = 0.05) masses in manatees in estrus and anestrus (30 masses) by HILIC/MS.. ................................ ................................ .............. 150 5 16 Bar graph comparing the average peak areas of the 10 significant masses elevat ed in estrous urine by HILIC/MS ................................ ............................ 150 5 17 MS/MS spectra of m/z 132, determined by HILIC/MS, in estrous urine (top) and a standard (bottom). ................................ ................................ .................. 151 5 18 MS/M S spectrum of m/z 148 in estrous urine, determined by HI LIC/MS. ........ 152 5 19 MS/MS spectrum of m/z 176 in estrous urine, determined by HILIC/MS.. ........ 152 5 20 MS/MS spectrum of m/z 244 in estrous urine, determined by HILIC/MS. ......... 153

PAGE 16

16 Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment o f the Requirements for the Degree of Doctor of Philosophy LC/MS OPTIMI ZATION AND IMPLEMENTATION FOR METABOLOMIC APPLICATIONS By Noelle Marie Elliott August 2012 Chair: Richard A. Yost Major: Chemistry The study of metabolomics aims to detect and identi fy as many compounds as possible in biofluids or tissues in order to determine the changes occurring within an organism in response to some perturbance. This is commonly performed by liquid chromatography/mass spectrometry (LC/MS) or nuclear magnetic reso nance spectroscopy (NMR) followed by statistical analysis of the data. The field of metabolomics has grown drastically over the last decade including the formation of the Metabolomics Society in 2004 and the start of the Metabolomics journal in March of 2005. This work employ ed LC/MS to investigate the metabolomics of several different biological studies on biofluids. The metabolomics of piglet plasma and red blood cells were compared over the first seven days of life and plasma was indicated to be a bet ter sampling technique, depending on the application A nother study compared the plasma from human patients with colon cancer to healthy controls and five potential biomarkers were identified including dodecenoylcarnitine, myristoleylcarnitine, myristoylc arnitine, palmitoleylcarnitine, and linoleoylcarnitine Manatee u rine was also examined to look for potential biomarkers in estr o us urine that may cause the male manatee to easily find

PAGE 17

17 a female manatee in estrus In manatee urine three potential biomarke rs were identified as creatine, glutamic acid, and cytidine. LC with MS/MS instrumentation was used in conjunction with accurate mass to identify metabolites determi ned to be significant in the colon cancer and manatee urine projects. In addition to the m etabolomics applications, some LC/MS optimization was performed. The main goal of the optimization was to increase the number of detectable molecular features while maintaining an efficient chromatographic run time. Particularly, the positive mode elect rospray settings were optimized and the LC run time was shortened.

PAGE 18

18 CHAPTER 1 INTRO DUCTION Metabolomics Background A biological marker, or biomarker, is defined as an indicator of normal biological processes, pathogenic processes, or pharmacologic respons es to administered therapeutics that are objectively measured and evaluated. 1 Biomarker discovery has changed over the last decade due to the introduction of high throughput technologies. 2 The metabolomics workflow has become much more automated in sampling, analysis, and data processing. Despite this change the overall purpose remains to identify compounds directly associated with diseases, or compounds that are respon ses to different stimuli Biomarkers may or give new insight into various diseases. Furthermore, biomarker s may not appear to be directly associated with the disease ; thus new biological pathways may need to be considere d. 3 There are many categories of biomarkers, as well as identification techniques (Tabl e 1 1 ) The biomarker category that will be focused on in this research is metabolomics as it is the most predictive of ph enotype 4 Metabolomics is a relatively new field of science, which has grown significantly over the last 10 years (Figure 1 1) and only recently formed a Metabolomics Society (2004) and a Metabolomics journal (2005). Metabolomic studies evaluate and compare changes in biological systems due to disease, diet, or other co nditions to an unchanged system ( i.e., the control ) 5 Metabolomics is one of several omic approaches but is the mo st predictive of phenotype. 6 Typically, small endogenous compounds (<1000 Daltons) present in tissues or biofluids are monitored for change. For example, a recent

PAGE 19

19 study determined metabolite changes between health y individuals and patients with either methylmalonic aci demia (MMA) or propionic ac i demia (PA), two inborn errors in metabolism. 7 Precipitated plasma was used in this example but other biological fluids can be used, such as urine, saliva, tissue biopsies, and cerebral spinal fluid (CSF). 3 After choosing a biological sample, an analytical technique must also be chosen for analyses. There are various techniques used to study metabolomics, but two common techniques are mass spectrometry ( MS ) and nuclear magnetic resonance spectroscopy ( NMR ) Approaches Two general approaches are used in metabolomics: profiling and fingerprinting. Metabolic profiling is a targeted analysis o f a limited set of metabolites whereas fingerprinting is an unbiased global approach. 4 Global approaches in metabolomics are ; the general approach is discussed below. A global fingerprinting approach beg in s with sample collection and preparation. Sample preparation is important for the removal of proteins as will be discussed later in this chapter. Next, an anal ytical technique is implemented (e.g., high performance liquid chromatography (HPLC or LC) c oupled to MS ) in order to separate and detect the large number of compounds present within biological samples. LC/MS generates a large amount of data, which can then be extracted and exported for statistical multivariate data analyses. The statistical an alyses are utilized to determine significant mass to charge ratios ( m/z ) from the raw data set. O nce significant m/z are determined, database searches and other comple mentary techniques, such as MS/MS or NMR, can also be employed to confirm the identity of the compounds. 4

PAGE 20

20 Platforms MS and NMR offer comple mentary information in metabolomics ; MS is used to determine accurate masses and NMR provides structural information based on hydrogen or carbon locations. Each plat form has its strengths and weaknesses, but MS has the advantage of being highly sensitive. 8 In contrast, NMR performed as an initial acquisition tool risks potentially significant, low concentration metabolites go ing unmeasured. 9 Therefore, it is often beneficial to begin a metabo lomics workflow with MS, since MS is capable of detecting compounds in a larger dynamic range than NMR. Following MS analysis, NMR can be used to analyze concentrated fractions collected within the LC run to assist in compound identification 10 11 Both platforms generate large datasets that can be analyzed with multivariate p attern recognition techniques to determine potential biomarkers. Similar to LC being used to decrease the complexity of the mass spectra, LC fractions can be collected to simplify the data collected by NMR. 12 Bioinfo rmatics technology continues to grow in the field of metabolomics, improving the software available to handle the large datasets generated in MS or NMR based metabolomic experiments. Biological Samples Blood and urine are common biofluids used for metabo lomic experiments, especially since collecting CSF is an invasive procedure. 13 15 Although sampling urine is less labor intensive, blood contains metabolites that travel from organ to organ ; therefore, blood provides information on fluxes caused by diseases or other perturbations that urine cannot. Metabolites are maintained in the blood until their concentrations exceed the relevant renal threshold (at which point they become present in urine), so blood is a more fav orable sampling medium than urine. 16

PAGE 21

21 Due to the complexity of whole blood, plasma is often used in metabolomics. Plasma, composed of mostly water that co ntains metabolites and proteins requires less sample prepara tion than blood c omponents containing cells, and is less complex than using whole blood. For example, e rythrocyt es, or red blood cells (RBC), require lysing in order to empty their contents. 17 For these reasons, plasma was used in most of our experiments. Plasma is separated from whole blood by centrifugation T he separation depicted in Figure 1 2 causes plasma to separate to the top, followed by a buffy coat, and the red blood cells on the bottom. The buffy coat contains platelets and white blood cells and is also called platelet rich plasma ( PRP ) Proteins are commonly precipitated from the plasma by organic solvent, such as acetonitrile, methanol, acetone, or a mixture of similar solvents. After precipit ation, the supernatant is removed for direct analysis or is dried down and reconstituted in a solvent of choice. 18 R emoving proteins from plasma can also remove smaller molecules; however, it is a necessary step if HP LC is to be performed. This is due to the increased back pressure that can be experienced, in addition to the potential deterioration of the HPLC column due to protein precipitation from organic mobile phases and buffers. 19 Additionally, proteins may interfere with the ion signal of smaller metabolites if present in the samples. Additionally, plasma should not be confused with serum. While the two biofluids are similar, plasma is prepared by the addition of an anticoagulant, such as heparin or ethylenediaminetetraacetic acid ( EDTA ) After the addition of the anticoagulant the blood is centrifuged as described above and the plasma is removed from the top layer. Serum, on the other hand, is blood plasma that has had the clotting factors such as

PAGE 22

22 fibrinogen a glycoprotein, removed 20 Once the serum or plasma is removed from whole blood the precipitation process can be performed. Also, it was recent ly suggested that plasma and serum are equally useful biofluids for use in a metabolomics study. 21 High Performance Liquid Chromatography LC Theory To understand liquid chromatography a basic knowledge of analytical separation s is needed In a liquid liquid extraction a solute will spend a certain amount of time in each solvent or phase. This relationship can be described by the partition coefficient ( K ), described in Eq 1 1 below. In the equation A s 1 and A s 2 are the activi ty of the solute in phase 1 and phase 2, respectively. 22 More commonly the concentration of the solute in each phase is used ([ S ] 1 and [ S ] 2 ). This extraction theory is directly applicable to HPLC separati ons. Similar to liquid liquid extractions, LC operates based on the partitioning of a solute between two phases, one phase being the stationary phase and the other a liquid mobile phase. The mobile phase continually flows through the stationary phase, if a solute has a higher affinity for the m obile phase it will spend less time on the stationary phase and elute earlier. Similarly, if a solute has a higher affinity for the stationary phase it will spend more time there and be less affected by the mobile phase causing later elution 22 (1 1) When performing LC, the data output is in the form of a chromatogram, which is simply the detector signal as a function of time (Figure 1 4). Once the sample is injected, the amount of time it takes the mobile phase (or unretain ed analyte) to reach

PAGE 23

23 the detector is called the dead time (labeled t 0 in Figure 1 4 ). The time it takes for an individual retained analyte to reach the detector is its retention time or t r In Figure 1 4, two different compounds are shown, A and B, that have two different retention times ( t A and t B ). The chromatographic peaks for compounds A and B have a certain peak width ( w ) that can be measured at the baseline or at half of the maximum width ( w 0.5 ). Once the retention times and peak widths are deter mined other calculations can be performed relevant to the separating power of the column. In Chapter 3, numbers of theoretical plates ( N ) and plate heights ( H ) will be examined for different LC methods to determine the optimal method. The numbers of t heo retical plates are often used to determine the chromatographic efficiency of a column. The number of theoretical plates ( N ), which is related to the retention time and width of a chromatographic peak, can be calculated by using Eq 1 2. After the number of theoretical plates is determined, the plate height, which is dependent upon column length ( L ), can also be determined using Eq 1 3. 23 (1 2) (1 3) The efficiency of chromatography met hods utilizing the same column can be compared by calculating N for individual analytes. If different columns are used, the n H allows for a better comparison as the column length is considered. Another way to estimate the quality of separation, which is a rgued to be more versatile than the number of theoretical plates ( N ), is to calculate the peak capacity. 24 Peak capacity is defined as the maximum number of resolvable peaks 25 In gradient elution the peak capacity ( n c ) is defined by Eq 1 4 and Eq. 1 5. Equation 1 4

PAGE 24

24 considers the entire gradient run time ( t g ) and Eq 1 4 removes unused space at the beginning or end of a chromatogram. When determining the peak ca pacity from Eq 1 4 and Eq. 1 5, is the average peak width and t 1 and t f are the retent ion times of the first and last eluting peaks, respectively. (1 4) (1 5) In LC, the resolution ( R s ) is a mea sure of how well two analytes are separate d Resolution is defined below in Eq 1 6, where t a and t b are the retention times of two peaks a and b, and w a and w b are their res pective baseline widths. 23 (1 6) Due to the complex nature of biological samples, LC separations are often performed prior to MS or NMR analyses. LC allows for direct injection of biofluids with minimal sample preparation and is capable of high throughput ana lyses. 26 Furthermore, LC columns have a relatively long life span and are routinely used to analyze hundreds of samples daily. 23 Chromatographic methods allow optimization through varyin g different parameters, such as columns, solvents, and gradients. Although c olumns are chosen based on iofluids are composed of compounds that cover a wide range of polarities ; thus all compounds cannot be efficiently retained and separated with a single column. The most common column employed is a reverse phase (RP) column, which retains more polar compounds poorly, but can separate compounds with low to moderate polarities with reproducibility. RP chromatography use s a non polar column (stationary phase) and polar solvents (mobile phases) compatible with the biological samples and the mass spectrometer. Solvents

PAGE 25

25 can be used in an isocratic (constant composition of solvents) or gradient elution (composition of solven ts varies with time). Gradient elution is more common due to its ability to remove non polar compounds more rapidly as well as its self cleaning nature. Monolithic and Small Particle HPLC Column s Early chromatography columns were composed of larger packe d particles leading to longer diffusional path lengths. With time, smaller particles were made that allowed shorter diffusional path lengths at the expense of higher back pressures. Monolithic colu mns were developed with the intention of having a shorte r diffusional path length as well as allowing column permeability. 27 A monolithic column is essentially a one piece continuous phase of porous silica rods modified with various functional groups, such as C18 28 Pressures observed with a Phenomenex C18 Onyx monolithic column (Phenomonex, Torrance, CA) are comparable to packed columns with 11 m particles but have a chromatographic efficiency of packed columns with 3 m part icles 29 Small particle columns, as already mentioned, have the advantage of efficient chromatography at the expense of a high back pressure ; however, back pressure may be decreased by applying heat which in turn decrease the solvent viscosity While the monolithic column is useful, small particle columns are more commonly implemented. In general, typical columns are approximately 100 mm x 2 mm with 3 m particles. 20 The s mall particle column used in this work is packed with 1.8 m particles and is half the length (50 mm) as what is commonly used. By heating the column ultra high performance LC (UPLC) is mimicked without the ultra high pressure. Figure 1 3 visualizes the stationary phases in both monolithic and small particle columns.

PAGE 26

26 Electrospray Ionization The electrospray ionization (ESI) 1985 and is now one of the most widely used ionization sources. 30 ESI is performed by passing a solution through a capillary, typically held at a high potential (2 4 kV). The high potential causes the solution to break into highly charged droplets that reduce in 31 ESI is performed at atmospheric pressure and the desolvated ions are transferred through ion optics by potential and pressure gradient s. The ESI source used in this work has an orthogonal spra y (Figure 1 5 ) with respect to the mass spectrometer capillary entrance. Additionally, th e ESI capillary is held at ground and a non zero potential is instead applied to the MS capillary (Vcap). Time of Flight Mass Spectrometry Time of flight mass spectr ometry ( TOF MS ) analysis can be explained mathem atically by considering E quation 1 7 : (1 7) where m is the mass of the ion, v is the velocity of the ion, z is the number of charges, e is the charge of an electron and E is the voltage ap plied in order to accelerate all the ions to the same kinetic energy. 32 Solving for v shows the inverse square root dependence on mass ; thus the larger the mass, the slower the velocity of the ion. The flight time ( t ) can be determined using the flight tube length ( D ) (E q 1 8 ) and then the m/z value can be found by E q 1 9 (1 8)

PAGE 27

27 (1 9) Commonly, TOF mass spectrometers are equipped with an ion mirror or re flectron. Reflectrons are beneficial because they allow for longer flight tube lengths by simply changing the direction of the ion path within the flight tube. For instance, a one meter flight tube becomes a flight path of two meters by adding a reflectr on. Increasing the flight path becomes beneficial to mass resolution, as will be discussed in the next section. How the reflectron operates is relatively straight forward. Ions that arrive at the flight tube begin their journey at the ion pulser. The i on pulser is a stack of plates with a center hole, except for the back plate. A pulse is applied to the back plate of the pulser and the ions are accelerated through the flight tube. At the opposite end of the flight tube the ions reach the reflectron. There are two potential gradients applied to the ion mirror that allow improved second order time focusing. The focusing is a result of horizontal momentum of the ions allows the reflection path to reach the detector instead of directly back to the pulser. 33 Resolution and Mass Accuracy Mass spectrometers are often evaluated based on two fig ures of merit: resolution and mass accuracy. Resolution defines how well two mass spectral peaks are separated and mass accuracy defines how far from the true or theoretical mass an observed mass is. The e quations for resolution ( R ) and mass accuracy ( A ) are shown below (Eq 1 10 and Eq. 1 1 1 ) : m 2 and m 1 are two adjacent masses and m m and m t are the measured mass and the theoretical mass. Although R is unitless, A is measured in parts per million (ppm). The flight tube length of the TOF MS used in this w ork is

PAGE 28

28 approximately two meters. This is a relatively large flight tube length yielding i mproved resolution, as resolution is proportional to flight time and the ions will maintain a constant time interval ( ) between them ( Eq 1 1 2 ). Therefore, the resolution increases with the flight time, which is proportional to the flight tube length ( D ), (E q 1 8 ). 32 (1 10) (1 11) (1 12) TOF MS Data Acquisition When using TOF begin s when the ions are pulsed through the flight tube and ends when all ions are detected. For the range of m/z 100 transient length is ~100,000 ns. The detector samples every nanosecond, thus one transient will have 100,000 sample intervals. About 10,000 transients are summed together for one spectrum, so at this mass range the sampling rate is one spectrum per second. 34 The mass range that w as used in these metabolomic studies wa s m/z 1 12 1700. The transient length wa s ~ 72,640 ns and 9,652 transients we re summed ; t herefore, the s ampling rate wa s 1.41 spectra/s and one spectrum wa s generated every 0.7 s. TOF MS System The TOF MS system operated in this research is characterized by relatively high mass accuracy (typically less than 5 ppm) an d resolution (5000 13000 over m/z 100 3200). Since thousands of compounds can be present in a single biological sample,

PAGE 29

29 high resolving power and accurate mass are desirable to reduce the number of possible candidates for compounds of interest. Most MS exp eriments in this work were performed on an Agilent 6210 orthogonal acceleration Time of Flight (oa TOF) mass spectrometer (Agilent Technologies, Santa Clara, CA). As shown in Figure 1 6, ions generated by ESI travel through a capillary to the skimmer. Li ghter ions and drying gas are then pumped away in vacuum stage 1. Ions that proceed through the skimmer are guided into a DC quadrupole, which shapes the beam of ions into a parallel packet. The ion pulser applies a DC bias that causes a packet of ions t o accelerate towards the ion mirror with equal kinetic energ ies At the ion mirror, the direction of the ions is reversed and the ions move toward the detector. Data Analysis Data analysis (extraction, alignment and statistics) can be performed with vario us software packages, some of which are freely available online. 35 One of the most popular methods implements XCMS ( The Scripps Research Institute, La Jolla, California) for data processing in combination w ith SIMCA P (Umetrics). 36 38 XCMS is website and SIMCA P is a statistical software package. 39 While this combination is widely utilized there are a few disadvantages. SIMCA P is rather expensive and XCMS requires extensive knowledge and understanding of the programming language R, making it more difficult to implement than other software provided by instrument companies. An online versi on of XCMS was recently launched (summer 2011), requiring no prior knowledge of R. The online program allows files to be uploaded and analyzed by an appropriate generic R scr ipt of you r choosing. For example, ultra high performance liquid chromatography (UPLC) or HPLC separations can be chosen in

PAGE 30

30 conjunction with a quadrupole TOF (Q TOF), or orbitrap mass analyzer. Once the online process is started the pr ogram proceeds to a lign all chromatograms and determine p values for each feature. Box and whisker plots are created, as well as extracted ion chromatograms for each feature. When all the processing is finished an email is sent to the user. By logging into the Scripps web site the results of the data set can be visualized Unfortunately, the online version of XCMS does not perform any t test. In addition to the packages mentioned above, most mass spectrometer manufacturer s ha ve a program available for metabolomics data processing/analysis. For example, AB SCIEX has created Marker View and Waters uses MassLynx. 40 Agilent Technologies created MassHunter software that performs pretreatment (ex traction of all m/z and retention time pairs over a certain abundance threshold) and an additional software package called Mass Profiler Professional (MPP), capable of multivariate statistical analysis. t test, analysis of variance (ANOVA), principal component analysis (PCA), and partial least squares discriminant analysis (PLS DA). MPP w as used frequently throughout this work and Qualitative Analysis w as used for all pretreatment (part of Agilent MassHu nter) of the LC/MS data collected. After extraction of the molecular features through Qualitative Analysis, files in the form of .CEF (formatted for MPP use) and .CSV (Excel format) can be exported Principal Component Analysis PCA is an unsupervised stat istical meth od, which reduces the d imensionality of a sample set by identifying the variance in data and projecting it into a lower dimensionality 41 In general, the results of a study can be considered as a matrix an d

PAGE 31

31 the covariance matrix of the initial matrix can be calculated After the covariance matrix is determined the Eigenvectors and Eigenvalues of the matrix are calcu lated. The Eigenvalues are organized from largest to smallest and the first two or three c orrespon ding Eigenvectors are plotted (an Eigenvector is termed a principal component). 42 If two component s are plotted the outcome will be two dimensional (2D), whearas if three components are plotted it will be three dimensional (3D). PCA plots are commonly used to determine if there are any obvious differences occurring within a sample set. For example, if a metabolomic s study is composed of three groups, a control group and two different treated groups, one would e xpect to see the PCA plot resulting in three individual groups, such as that shown in Figure 1 7 In addition to being used as a data visualization tool in metabolomics, PCA can also be used to determine the f eatures/masses (or loadings) that caus e those differences observed in the PCA plot (or scores plot). Although PCA is popular within the field of metabolomics it is increasing ly being replaced by partial least squares discriminant analysis. Partial Least Squares Discriminant Analysis Partial least squ ares discrim inant analysis is a statistical technique that has recently gained more attention in the field of metabolomics. In an ISI Web of Science 2008 only 34 results were foun d, while a search of those same terms for the years 2008 2011 yielded 142 results. This increase in metabolomics publications utilizing PLS DA shows its usefulness as a discrimination and dimension reduction tool. PLS DA differs from PCA in the fact that it is a supervised method. This means it is directed by a response matrix Y, which is then used to derive components from the

PAGE 32

32 data set matrix X that best describe the Y matrix. 43 PLS DA and other supervised methods are commonly used for complex metabolomics studies, where important changes between study groups may be masked by other large changes (e.g. human urine can vary greatly between individuals, which may hinder elucidation of more subtle changes). 43 Additionally, PLS DA can be used as a prediction model. Since the method is supervised the conditions given can be used to determine how individual samples would be grouped. While PCA is commonly used solely to determine h ow samples group, PLS DA is often used to determine the masses/features that are able to describe the grouping. As PLS DA is a supervised method, analysis of the loadings plot from PLS DA is more straightforward than the loadings plot from P CA. Thus, PLS DA can be used to model the sample separation as well as to easily determine significant influences on that separation. t test t test is used when only two experimental groups exist (e.g. diseased and control). Signifi cant masses are determined by comparing the means of the two groups f or each individual feature. The t statistic is calculated by applying Eq 1 1 3 where x is the average for the two groups (denoted by the subscript), s is the pooled standard deviation (Eq 1 1 4 ) s n is the standard deviation of a sample grouping and n is the number of samples in either group one or group two, depending on the subscript. 44 (1 13) (1 14)

PAGE 33

33 T he t statistic is compared to a critical value to determine if a significan t difference exists between sample groupings If the t statistic exceeds the specified critical value within a certain probability ( p value) the null hypothesis (that the two averages are equal) can be rejected. 44 Once significant masses are determined by the t test more analyses can be performed to either filter the list further, or identify those significant masses. Analysis of Variance The analysis of variance (ANOVA) is commonly used in metabolomics when there t test inefficient. ANOVA u ses the sum of squares (SS) w hich is the variance after mean centering, as well as F tests to calculate p values and determine significant features. The sum of squares is shown below in Eq 1 1 5 where x is the average of a set of values and x i is an individual value 41 (1 15) The general format for ANOVA is illustrated in Table 1 2 The among group variation ( SS A ) is calculate d by determining the SS of the means multip lied by the number of samples for each group. The within gr oup variation ( SS W ) is calculated using the SS for each group. Once the SS is determined for among and within the groups the degrees of freedom are used to calculate the mean squares and the F te st value (Table 1 2 ). 41 The p F density plots (some of which are freely available online).

PAGE 34

34 Benjamini and Hochberg False Discovery Rate It should be noted that p values can be calculated and corrected by various methods. Most commonly p values are asymptotically calculated (as opposed to permutative calculation) and a multiple testing correction is used in metabolomics. The testing correction that will be used in this work is the Benjamini and Hochberg false discovery rate (FDR). The goal of multiple testing corrections such as FDR is to prevent type I errors from occurring (the reject ion of a true null hypothesis, more commonly known as a false positive) Unfortunately, controlling for the type I errors often leads to an increase in type II errors (the failure to accept a false null hypothesis, a false negative). 44 The relationship of type I and type II errors i s often referred to as the familywise error rate (FWER). The possible outcomes of individua l tests can be seen in Table 1 3 The procedure for the Benjamini and Hochberg FDR is as follows: the p values (P) for m tests are ranked in ascending order (P 1 P 2 P m ) and each p value, beginning with P 1 is checked to meet the requirement P i i where i is the rank. The term m is the type I error rate. Eventually k, the largest i where P i i is true, is determined and all the corresponding null hypotheses (H (1) to H (k) ) are rejected. 45 The ability of the Benjamini and Hochberg FDR to potentially reduce type II errors compared to other methods is shown in Figure 1 8. Because of the desire to have n ot even one type I error, FWER correction methods require a lower threshold setting (thus increasing the likelihood of typ e II errors) tha n the FDR method (Figure 1 8).

PAGE 35

35 Table 1 1. Biomarker categories and examples of methods used within each category. 2 3 Biomarker Category Methods Genomics DNA arrays DNA Based Copy number/loss of heterozygosity, sequence and epigenetic variations, and genome rearrangements Seque ncing RNA Based mRNA and miRNA signatures Proteomics Mass spectrometry Proteins Liquid chromatography Peptides Protein arrays Metabolomics Mass spectrometry Metabolites Liquid chromatography Lipids Nuclear magnetic resonanc e Carbohydrates Table 1 2 Typical set up for ANOVA Source of Variation S um Degrees of Freedom Mean Squares F T est Among sample groups SS A m 1 s A 2 s A 2 / s W 2 Within sample replicates SS W N m s W 2 Total Variation SS T N 1 s T 2 SS is the sum of squares, m is the number of groups, N is the number of samples and s is the average (sum/degrees of freedom) for each source of variation Table 1 3. Possible outcomes of individual tests. 46 Truth Decision Tot al Not Significant Significant Null Hypothesis U V m 0 Alternative Hypothesis T S m m 0 Total m r r m V is the number of type I errors, T is the number of type II errors and the variables m, r, and m r are observed. U, T, and S are unknown.

PAGE 36

36 Figur e 1 1. Depiction of the number of publications in metabolomics each year for the past 10 years (database search of ISI Web of Science). Figure 1 2 The separation of whole blood into its three components. 47 0 200 400 600 800 1000 1200 1400 2002 2004 2006 2008 2010 Number of Publications Year

PAGE 37

37 Figure 1 3 Comparison of the diffusional path of monolithic bas ed columns (A ) and small particle columns (B ). 29 Figure 1 4 Depiction of dead time, retention time and peak widths in LC. 48

PAGE 38

38 Figure 1 5 Schematic of the electrospray source. The LC flow enters the source through the nebulizer, where it is sprayed orthogonally with respect to the mass spectrometers capillary inlet (Vcap). 49

PAGE 39

39 Figure 1 6 Schematic of the Agilent 6210 time of flight mass spectrometer. 50

PAGE 40

40 Figure 1 7 Generic 3D PCA plot of a metabolomics experiment with three groups (indicated by the colors red, bl ue and yellow ) 51

PAGE 41

41 Figure 1 8. Threshold p values for three different multiple testing corrections. Conditions include 50 tests being performed and FWER = FDR = 0.05 (that is, m = 50 and = 0.05). The Benjamini and Hochberg FDR tests from the largest to smallest p value while the two Bonferroni methods test from the smallest to the largest p value. 45

PAGE 42

42 CHAPTER 2 COMPARIS ON OF THE METABOLOM ICS OF PLASMA AND RE D BLOOD CELLS FROM PIGLETS IN THE FIRST SEVEN DAYS OF LIFE Background Piglets in the first seven days of life are excellent animal models for neonatal disorders. 52 While much is already know n about neonatal disorders, additional knowledge could only improve our methods of testing for these diseases. 53 For instance, o ctanoylcarnitine (C8) is currently used as a biomarker to screen newborns for medium chain acyl CoA dehydrogenase deficiency (MCAD). 54 One problem with this method is the lack of specificity for MCAD, since C8 can also be elevated in other disorders. Global metabolomic profiling and targeted studies of carnitines and acylcarnitines in piglets are hypothesized to indicate how levels of metabolites change throughout the first week of life in humans ; this could allow further insight into neonatal disorders. For these analyses, whole blood is separated i nto plasma, red blood cells, and a buffy layer, but typically only plasma is analyzed. In this work, we extend the study to both plasma and red blood cells (RBC) It is common in metabolomic studies to utilize biofluids such as plasma or serum (as discuss ed in Chapter 1). In fact, one paper in the field discusses whether plasma or serum is a better choice for intersubject comparisons The conclusion to this paper is that neither biofluid is better than the other; there are very little variations between plasma and serum allowing them to be compared relatively easily 21 There is no comparison in the literature of RBCs to plasma. T herefore, t his work is beneficial to determining if RBCs are a valid way of looking at the metabolome of piglets As mentioned, c arnitines and acylcarnitines are of particular interest in this study due to their importance in early human development. This is because acylcarnitines are

PAGE 43

43 the catabolic end products of fatty acids and some br anched chain amino acids used to generate energy. Acylcarnitines are derived from the exchange of acyl groups between their corresponding acyl coenzyme A and L carnitine. However, acylcarnitines are capable of crossing mitochondrial and cell membranes, w hile their corresponding acyl coenzyme A cannot. Additionally, the acylcarnitine pattern in a fasting individual, under normal circumstances, is relatively stable. Because of this, the plasma acylcarnitine profile can be used to detect many metabolic dis orders. In fact, all babies born in the U.S., and most of Europe, are screened by a dried blood spot analysis for their acylcarnitine profile. 55 For these important reasons, some carnitines and acylcarnitines wi ll be targeted in the piglet plasma and red blood cells in order to observe changes occurring over the first seven days of life. 5 5 Experimental Sample Collection and Preparation Plasma and red blood cell samples were prepared from whole blood of piglets in the first seven days of life The whole blood was separated into its components by centrifugation and the plasma and RBC were frozen. The frozen RBC and plasma were then prepared by two different procedures. Frozen plasma was thawed and 100 L portions were mixed with 1 mL of acetonitrile:methanol (3:1) and frozen overnight. After the overnight freeze, the samples were thawed, centrifuged, and the supernatants removed. The supernatants were then centrifuged two more times and evapora ted to dryness und er nitrogen. Frozen RBCs were prepared by adding equal volumes of ice cold de ionized water to each sample. The RBC samples were then re froze n overnight and later centrifuged. The supernatants were removed and the samples dried under nitrogen.

PAGE 44

44 The dried products from both plasma and RBC samples were each dissolved in 100 L of 40% acetonitrile, 40% methanol, 10% water and 10% isopropanol. 56 Three piglets per day of life were sampled and separated into tw o vials per piglet and run in duplicate for plasma analyses (12 runs per day, 84 files in total). For RBC analyses, three piglets were sampled per day and run in duplicate (6 runs per day, 42 files in total). Liquid Chromatography Conditions All c hromato grams were acquired using an Agilent 1200 Series LC system (Agilent Technologies, Inc., Santa Clara, California). Two, in series, Phenomenex C 18 Onyx monolithic columns (Phenome nex, Torrance, CA), with individual dimensions of 4.6 x 100 mm, were used. Wa ter was used for solvent A and acetonitrile for solvent B, both solvents contained of 1% (v/v) acetic acid. The gradient performed began with 95% A and 5% B and held for 6.5 minutes, followed by a ramp to 100% B at 25.5 minutes. Solvent B was held at 100 % for 9.5 minutes and then returned to 95% A and 5% B over 10.0 minutes (Table 2 1) The column was then re equilibrated with 95% A and 5% B for 5.0 minutes. The total gradient time was 50.0 minutes, with only the first 37.0 minutes being used for data a nalysis. The flow rate was set to 0.8 mL/min and an injection volume of 15.0 L was used for all RBC and plasma samples. Electrospray Source Ionization of the LC flow was performed by an orthogonal dual spray electrospray source operating in positive mode (Agilent Technologies, Inc., Santa Clara, California). The dual spray is used to constantly deliver calibrant ions into the system. The dual spray ESI source allows for easy post analysis recalibration, and when used properly, no recalibration is necess ary since a real time calibration will be performed. A gas

PAGE 45

45 temperature of 300 C and a drying gas flow of 10 L/min was used. The nebuli zer pressure was set to 50 psi. Mass Spectrometer Mass spectra were acquired using an Agilent 6210 Time of Flight mass spectrometer ( Agilent Technologies, Inc., Santa Clara, California). The capill ary, fragmentor, and skimmer voltage were set as 4000, 180, and 60 V, respectively. The instrument was operated in extended dynamic range (2 GHz), as opposed to high resolution mode (4 GHz) in order to increase the linear dynamic range and decrease satura tion of the detector. Data Analysis Agilent Technologies, Inc., Santa Clara, California) and molecular features were extracted. The abundance threshold for the feature extraction was set to 2 ,500 arbitrary units (a.u.). Only singly charged species were considered. To be considered a molecular feature at least two ions had to be present (this could be two ions of the isotope pattern). After feature extraction the files were exported in .CEF and .CSV formats. The .CEF files ( Agilent Technologies, Inc., Santa Clara, California). In MPP, features over the first 37 minutes were considered. Mass an d retention time alignment was performed on features with thresholds 5 ppm and 0.5 minutes, respectively. Normalization of each file was done by baselining to the median. Additionally, Quantitative Analysis software ( Agilent Technologies, Inc., Santa Clar a, California) was also used to extract 11 carnitines/acylcanitines (carnitine, acetylcarnitine, propionylcarnitine, butyrylcarnitine, hexanoylcarnitine, octanoylcarnitine,

PAGE 46

46 decanoylcarnitine, lauroylcarnitine, myristoylcarnitine, palmitoylcarnitine, and st earoylcarnitine) This was done by entering the know n retention times and accurate masses into the software so that a spreadsheet could be automatically generated containing important information for those compounds. This information included but was n ot limited to retention time, chromatographic peak area and height, and peak width. From this spreadsheet the relative behavior of the different carnitines/acylcarnitines could be monitored over the seven days of life Results and Discussion As can be seen by comparing the chromatograms of the piglet plasma and RBC samples (Figures 2 1 and 2 2) the plasma chromatograms are much more complex. This may mean there is m ore information to be found by analyzin g plasma than RBC s After visual comparison of the chromatograms, the extracted features from the plasma and RBC samples were considered. Piglet Plasma In the piglet plasma 15,989 masses were determined to be present within the first seven days of life. T hese 15, 989 masses were filtered on frequenc y of observation (80 %) in order to retain the masses more frequently observed, as well as to decrease the number of compounds to a more manageable number By doing this an individual feature had to be present in 80 % of the samples tested in an individual day, but not necessarily every day. For example, a feature would pass the filter if it was present in 80 % of the samples from day 2 and day 5 but not in days 3, 4, 6, 7, or 8. A total of 568 masses passed this frequency cut off. ANOVA was then performe d to determine which of those 568 features were statistically significant. Using a p value of 0.0 1, a post hoc test (Tukey HSD), and a multiple testing correc tion (Benjamini Hochberg FDR), 82

PAGE 47

47 masses were determined to be significant. Table 2 2 depicts th e statistically significant masses between days. This number of features is not overwhelming and may make it easier to find identifications. The 3D PCA plot is shown in Figure 2 3 for the 82 ANOVA masses. The x axis (component 1) accounts for 25 15% of the variation, while the y and z axes (components 2 and 3) account for 11 93% and 7 53 %, respectively. There does appear to be some separation in Figure 2 3. In particular it seems that the earlier days of life are on the top right side of the plot and the later days of life drift towards the bottom left. A two dimensional plot is also shown in Figure 2 4, as some prefer this view. Previous work in our lab with only piglet plasma, f ound day 2 and days 7 and 8 grouping away from the middle days, with da ys 7 and 8 grouping together on the PCA plot. 56 These results are similar to what is shown here but with less distinct separation. Also, the data analysis process is not the same for this work and the pre vious work. A PLS DA was also performed and the plot can be seen in Figure 2 5 (3D view) and Figure 2 6 (2D view). The prediction model had an accuracy of 80.95%, most of the inaccuracies stemmed from day 6 as well as some other days of life in the middle of the time period tested (Table 2 3). The day of life that was most accurate (100%) was day 2. This means that significant changes are occurring from day two compared to the later days. The loadings plot from the PLS DA (Figure 2 7) is often used to de termine what masses are influencing the separation observed. This can be done by examining the masses that are farthest from WC0 = 0. These masses (30) will be discussed in more detail later.

PAGE 48

48 Piglet Red Blood Cells Piglet RBC samples were put through the same data analysis procedure as described above for piglet plasma. A total of 8,7 39 masses were determined to be present in the red blood cells of piglets during the first seven days of life. After filtering on the frequency of observation (80 % as desc ribed above) 879 masses passed the filter. The filtered masses were then subjected to ANOVA ( p value = 0.0 1 ) with a post hoc test (Tukey HSD) and a multiple testing correction (Benjamini Hochberg FDR) to determine the statistically significant masses. O nly 23 features were determined to be significant from day to day within the RBC samples (Table 2 4) A 3D and 2D PCA plot of the 23 ANOVA masses is shown in Figure 2 8 and Figure 2 9, respectively. The x axis (component 1) accounts for 37.87% of the var iation, while the y and z axes (components 2 and 3) account for 14.16% and 7.95%, respectively. As can be seen in the PCA plots, the m ost separation is observed for days 7 and 8 of life. The 3D PLS DA plot is shown in Figure 2 10, as well as the 2D PLS DA plot in Figure 2 11. The piglet RBC prediction model had the same accuracy as the plasma model, 80.95% (Table 2 5). Days 2, 7, and 8 had 100% prediction accuracy while days 4 and 6 were lacking in accuracy. It is important to stress that there are h alf as many samples in the RBC model as in the plasma model; there fore, the RBC model may be lacking the appropriate statistical power. The loadings plot (Figure 2 12) was used to examine the masses that were farthest from WC0 = 0. These masses (13) will be discussed below. Database Search of Significant Piglet Plasma and RBC Masses As already discussed, PLS DA is a powerful and popular model The loadings plot allows easy determination of masses that are influencing the observed separation. The

PAGE 49

49 ANOVA ma sses were further filtered by considering masses in the PLS D A plot that satisfied the conditions of WC0 < 0.1 or WCO > 0.1 for both the plasma and RBC data sets. The piglet plasma dataset resulted in 30 masses and the piglet RBC dataset resulted in 13 masses, which were then applied to an accurate mass database search. This was performed using METLIN 57 58 The mass of the positively charged ion was used in the database search and a mass accuracy of 10 ppm was applied. Two different ion options were searched, [M+H] + and [M+Na] + The positive ion that is outputted as significant in MPP is supposed to be an [M+H] + ion, however, the [M+Na] + ion was still searched as a pr ecaution. Piglet plasma masses yielded a formula for 3 ions as [M+Na] + ions. Piglet RBC masses yielded only one formula as an [M+H] + ion. Targeted Analysis In this study 11 carnitines / acylcarnitines were studied further to determine their relative chang e over the first seven days of life. This was done by determining the average abundance within the two replicates of three piglets (n = 6) for the RBC dataset, and for two replicates of thee piglets with two vials each (n = 12) for the plasma dataset Of the 11 compounds all were found to be present in the plasma samples, but only 7 were present in the RBC samples. Those seven compounds were carnitine, octanoylcarnitine, decanoylcarnitine, lauroylcarnitine, myristoylcarnitine, palmitorylcarnitine and ste aroylcarnitine. Th e fact that only 7 of the targeted compounds were found in the RBC samples (compared to the 11 found in plasma) further supports the appropriateness of plasma for an experiment of this type. Bar graphs of the average abundance of the 7 compounds found in both plasma and RBC are shown in Figures 2 13 to 2 1 9 In Figure 2 13 the abundance of carnitine is observed to be lowest on day 2 in both plasma and RBC samples. The abundance of

PAGE 50

50 carnitine over the rest of the days overlapped when the standard deviation of the mean was considered but showed a slight increase from day 2 thru day 8 This same trend is apparent in the other 6 compounds, with the exception of some of the RBC samples. The RBC trends from day 2 to day 8 seem to randomly g o up and down, although day 2 was consistently the lowest abundance day for all 7 compounds except octanoylcarnitine (Figure 2 14 ). There may be a few reasons the plasma and RBC samples do no show the same trends over the 7 days. One reason may be the e xtraction efficiency. If all the contents of the red blood cells are not efficiently emptied than there will be different results when compared to plasma. Since some of the lower molecular weight carnitines and aceylcarnitnes are missing from the RBC sam ples, this may indicate a lack of extraction of these from the red blood cells. Also, it may simply be that certain carnitines and acylcarnitines are present in the red blood cells at lower concentration s than in plasma Summary P iglet plasma yielded 82 s ignificant masses, while piglet red blood cells only resulted in 23 significant masses This shows that there are more changes to compare in plasma than in RBCs of piglets Only two statistically significant masses matched between the ANOVA results of th e piglet plasma and RBC samples. Again, this may be the result of poor com pound extraction from the RBC. It would be beneficial to reproduce this experiment with the correct number of replicates (at least three) as well as an equal number of samples betwe en the plasma and RBC groups. Unfortunately, at the time of analysis, the chromatographic method was still 50 minutes, requiring any long metabolomic experiment to be broken into two separate analyses, usually about a week apart. This allowed for the ins trument to be

PAGE 51

51 used for service samples and by other lab mates. It also led to irreproducibility for one set of RBC samples to the other, which is why there are only half as many as the plasma samples. This was amplified by the fact that the instrument re ceived maintenance in between the two weeks that the experiment was broken into.

PAGE 52

52 Table 2 1. Liquid chromatography gradient Time (minutes) A:B 0.0 6.5 95:5 25.5 35.0 0:100 45.0 50.0 95:5 A:B represents the ratio of mobile phase A to mobile ph ase B at a given time. Table 2 2. R esults from ANOVA for piglet plasma p = 0.01. Day of Life 8 7 6 5 2 3 4 8 82 4 3 10 44 29 14 7 78 82 5 12 48 28 15 6 79 77 82 4 34 16 9 5 72 70 78 82 34 10 6 2 38 34 48 48 82 14 30 3 53 54 66 72 68 82 8 4 68 6 7 73 76 52 74 82 Valu es above the diagonal indicate those masses that are differentially expressed between the two indicated days. The valu es below the diagonal indicate the masses that are not differentially expressed. The diagonal valu es (82) show the total number of statistically significant masses between all the days of life.

PAGE 53

53 Table 2 3. PLS DA prediction results for piglet plasma Predictions Day 2 Day 3 Day 4 Day 5 Day 6 Day 7 Day 8 Accuracy (%) 12 0 0 0 0 0 0 100 .00 1 11 0 0 0 0 0 91.67 0 2 10 0 0 0 0 83.33 0 0 0 10 0 2 0 83.33 0 0 2 1 5 1 3 41.67 0 0 0 0 0 11 1 91.67 0 0 0 1 0 2 9 75 .00 80.95 There are 12 files for each day of life, thus each should have a prediction of 12, those that vary cause the accuracy of the model to dec rease. Table 2 4 Results from ANOVA for piglet RBCs, p = 0.01. Day of Life 8 7 6 5 2 3 4 8 23 4 5 4 8 9 8 7 19 23 8 9 15 14 10 6 18 15 23 0 2 4 2 5 19 14 23 23 4 4 4 2 15 8 21 19 23 4 3 3 14 9 19 19 19 23 1 4 15 13 21 19 20 22 23 Valu es above the diagonal indicate those masses that are differentially expressed between the two indicated days. The valu es below the diagonal indicate the masses that are not differentially expressed. The diagonal valu es (23) show the total number of statistically significant masses between all the days of life.

PAGE 54

54 Table 2 5. PLS DA prediction results for piglet RBCs Predictions Day 2 Day 3 Day 4 Day 5 Day 6 Day 7 Day 8 Accuracy (%) 6 0 0 0 0 0 0 100 0 5 0 1 0 0 0 83.33 0 4 2 0 0 0 0 33.33 0 0 0 6 0 0 0 100 0 0 0 0 3 3 0 50 0 0 0 0 0 6 0 100 0 0 0 0 0 0 6 100 80.95 There are 6 files for each day of life, thus each should have a prediction of 6, those that vary cause the accuracy of the model to decrease.

PAGE 55

55 Figure 2 1. Overlay of 84 piglet plasma chromatograms. Figure 2 2. Overlay of 42 piglet RBC chromatograms.

PAGE 56

56 Figure 2 3. 3D PCA plot of ANOVA masses in piglet plasma filtered for 80% frequency (82 masses). Each color represents a different day of life Each point of the PCA plot depicts on e replicate. Figure 2 4. 2D PCA plot of ANOVA masses in piglet plasma filtered for 80% frequency (82 masses) Each color represents a different day of life and each point is one replicate. The color code is the same as Figure 2 3.

PAGE 57

57 Figure 2 5 3D P LS DA plot of ANOVA masses in piglet plasma filtered for 80% frequency (82 masses). Each color represents a different day of life. Each point on the plot depicts one replicate. Figure 2 6 2D P LS DA plot of ANOVA masses in piglet plasma filtered fo r 80% frequency (82 masses). Each color represents a different day of life. Each point on the plot depicts one replicate. The x axis (Identifier) shows individual replicates. The color code is the same as Figure 2 5.

PAGE 58

58 Figure 2 7 P LS DA loadings plo t of ANOVA masses in piglet plasma filtered for 80% frequency (82 masses). WC0 indicates how influential a m/z value is (largest and smallest WC0 values are more influential in PLS DA separation). Figure 2 8 3D PCA plot of ANOVA masses in piglet RBCs filtered for 80% frequency (23 masses). Each color represents a different day of life and each point is one replicate.

PAGE 59

59 Figure 2 9. 2D PCA plot of ANOVA masses in piglet RBCs, filtered for 80% frequency (23 masses) Each color represents a different day of life and each point is one replicate. The color code is the same as Figure 2 8. Figure 2 10 3D P LS DA plot of ANOVA masses in piglet RBCs, filtered for 80% frequency (23 masses). Each color represents a different day of life. Each point on th e plot depicts one replicate.

PAGE 60

60 Figure 2 11 2D P LS DA plot of ANOVA masses in piglet RBCs, filtered for 80% frequency (23 masses). Each color represents a different day of life. Each point on the plot depicts one replicate. The x axis (Identifier) sho ws individual replicates. The color code is the same as Figure 2 10. Figure 2 12 P LS DA loadings plot of ANOVA masses in piglet RBCs, filtered for 80% frequency (23 masses). WC0 indicates how influential a m/z value is (largest and smallest WC0 value s are more influential in PLS DA separation).

PAGE 61

61 Figure 2 13 Bar graphs of carnitine response in piglet plasma and RBC. Error bars indicate the standard deviation of the mean. Figure 2 14 Bar graphs of octanoylcarnitine response in piglet plasma an d RBC. Error bars indicate the standard deviation of the mean. 0.0E+00 2.0E+06 4.0E+06 6.0E+06 8.0E+06 1.0E+07 1.2E+07 1.4E+07 Day 2 Day 3 Day 4 Day 5 Day 6 Day 7 Day 8 Average Abundance (a.u.) Day of Life Carnitine Response in Plasma 0.0E+00 2.0E+05 4.0E+05 6.0E+05 8.0E+05 1.0E+06 1.2E+06 Day 2 Day 3 Day 4 Day 5 Day 6 Day 7 Day 8 Average Abundance (a.u.) Day of Life Carnitine Response in RBC 0.0E+00 5.0E+04 1.0E+05 1.5E+05 2.0E+05 2.5E+05 3.0E+05 3.5E+05 4.0E+05 4.5E+05 5.0E+05 Day 2 Day 3 Day 4 Day 5 Day 6 Day 7 Day 8 Average Abundance (a.u.) Day of Life Octanoylcarnitine Response in Plasma 0.0E+00 1.0E+04 2.0E+04 3.0E+04 4.0E+04 5.0E+04 6.0E+04 7.0E+04 Day 2 Day 3 Day 4 Day 5 Day 6 Day 7 Day 8 Average Abundance (a.u.) Day of Life Octanoylcarnitine Response in RBC

PAGE 62

62 Figure 2 15 Bar graphs of decanoylcarnitine response in piglet plasma and RBC. Error bars indicate the standard deviation of the mean. Figure 2 16 Bar graphs of lauroylcarnitine res ponse in piglet plasma and RBC. Error bars indicate the standard deviation of the mean. 0.0E+00 1.0E+05 2.0E+05 3.0E+05 4.0E+05 5.0E+05 6.0E+05 7.0E+05 8.0E+05 9.0E+05 Day 2 Day 3 Day 4 Day 5 Day 6 Day 7 Day 8 Average Abundance (a.u.) Day of Life Decanoylcarnitine Response in Plasma 0.0E+00 1.0E+04 2.0E+04 3.0E+04 4.0E+04 5.0E+04 Day 2 Day 3 Day 4 Day 5 Day 6 Day 7 Day 8 Average Abundance (a.u.) Day of Life Decanoylcarnitine Response in RBC 0.0E+00 1.0E+05 2.0E+05 3.0E+05 4.0E+05 5.0E+05 6.0E+05 Day 2 Day 3 Day 4 Day 5 Day 6 Day 7 Day 8 Average Abundance (a.u.) Day of Life Lauroylcarnitine Response in Plasma 0.0E+00 1.0E+04 2.0E+04 3.0E+04 4.0E+04 5.0E+04 6.0E+04 7.0E+04 Day 2 Day 3 Day 4 Day 5 Day 6 Day 7 Day 8 Average Abundance (a.u.) Day of Life Lauroylcarnitine Response in RBC

PAGE 63

63 Figure 2 17 Bar graphs of myristoylcarnitine response in piglet plasma and RBC. Error bars indicate the standard deviation of the mean. Figure 2 18 Bar grap hs of palmi toylcarnitine response in piglet plasma and RBC. Error bars indicate the standard deviation of the mean. 0.0E+00 2.0E+05 4.0E+05 6.0E+05 8.0E+05 1.0E+06 1.2E+06 Day 2 Day 3 Day 4 Day 5 Day 6 Day 7 Day 8 Average Abundance (a.u.) Day of Life Myristoylcarnitine Response in Plasma 0.0E+00 3.0E+05 6.0E+05 9.0E+05 1.2E+06 1.5E+06 Day 2 Day 3 Day 4 Day 5 Day 6 Day 7 Day 8 Average Abundance (a.u.) Day of Life Myristoylcarnitine Response in RBC 0.0E+00 1.0E+06 2.0E+06 3.0E+06 4.0E+06 5.0E+06 6.0E+06 7.0E+06 Day 2 Day 3 Day 4 Day 5 Day 6 Day 7 Day 8 Average Abundance (a.u.) Day of Life Palmitoylcarnitine Response in Plasma 0.0E+00 5.0E+06 1.0E+07 1.5E+07 2.0E+07 Day 2 Day 3 Day 4 Day 5 Day 6 Day 7 Day 8 Average Abundance (a.u.) Day of Life Palmitoylcarnitine Response in RBC

PAGE 64

64 Figure 2 1 9 Bar graphs of stearoylcarnitine response in piglet plasma and RBC. Error bars indicate the standard deviation of the mean 0.0E+00 1.0E+06 2.0E+06 3.0E+06 4.0E+06 5.0E+06 6.0E+06 7.0E+06 8.0E+06 Day 2 Day 3 Day 4 Day 5 Day 6 Day 7 Day 8 Average Abundance (a.u.) Day of Life Stearoylcarnitine Response in Plasma 0.0E+00 5.0E+06 1.0E+07 1.5E+07 2.0E+07 2.5E+07 3.0E+07 3.5E+07 Day 2 Day 3 Day 4 Day 5 Day 6 Day 7 Day 8 Average Abundance (a.u.) Day of Life Stearoylcarnitine Response in RBC

PAGE 65

65 CHAPTER 3 METABOLOMICS OF COLO N CANCER BY ANALYSIS OF HUMAN PLASMA Overview Colorectal cancer (CRC) affects the colon and/or rectum and is the second highest cancer related cause of death when con sidering both males and females 59 When men and women are considered individually CRC is the third highest cancer related cause of death after lung cancer and breast and prostate cancers in women and men respectively. 59 For 2012 the number of new colon and rectal cancer ca ses is estimated to be 103 170 and 40 290 respectively. 59 Although colon cancer and rectal cancer are frequently grouped together, 72% of the estimated new CRC cases are colon cancer, proving how important it is to better understand the disease. Colon cancer results from changes in polyps in the lining of the bowel. It takes approximately 10 15 years for an adenoma polyp to change to a carcinoma polyp; however, the chance of a benign one centimeter polyp to change to a malignant polyp is only 10%. 60 There are some things that can be done to try to protect oneself against CRC, including exercising an d maintaining a diet that avoids a high amount of fat as well as eating plenty of fruits and vegetables Because of the high mortality rate with CRC as well as the readily available screening tools, early detection is the key to overcoming both colon a nd rectal cancers. It is recommended that anyone at average or slightly above average risk (98% of the population) undergo a colonoscopy at the age of 50 to screen for CRC 61 While the colonoscopy can remove 90% of polyps, it is an uncomfortable and invasive procedure that requires preparation in order to remove the contents of the colon and rectum. 60 A less invasive screening tool is the fecal occult blood test. The fecal occul t blood test

PAGE 66

66 detects blood in stool samples, but it is not a very sensitive test. It is recommended that after more than 4/6 positive fecal tests the patient receives a colonoscopy. The fecal occult blood test can detect 72% of cancers with screening eve ry two years. 60 The lack of a non invasive, sensitive screening tool is one reason why the metabolomics of colon cancer in plasma should be examined. If a plasma biomarker can be determined, it could be possible for c olon cancer to be screened by a simple blood drawing. In order to better understand the samples investigated in this chapter, the method of staging colon cancer should be mentioned. The staging system used in this work is that given by the American Join t Committee on Cancer (AJCC). 62 By the AJCC standards there are five stages of cancer (0 IV). The AJCC staging is determined by considering how far the primary tumor has progressed through th e colon wall (T), how many nearby lymph nodes have been afflicted (N), and how far the disease has metastasized (M). Because of this, the AJCC staging system is also sometimes called the TNM system. 63 A s the staging of colon cancer is discussed below many areas in and around the colon will be mentioned; a cross section o f the colon is shown in Figure 3 1 to better understand the complexity of the stages. Essentially, the inner most area of the colon is called the mucosa and is affected first with colon cancer cells. The second inner most region of the colon is called the submucosa, followed by muscle layers. The outer layer of the colon is called the serosa. In stage 0, abnormal cells are in the mucos a of the colon wall. The cancer then spreads from the mucosa to the muscle layer in stage I of colon cancer. With this early stage of cancer the five year survival rate for stage I colon cancer is 93.2%. 64 Within stage II there are three sub stages: stage IIA, stage IIB and stage IIC. If the cancer has spread through the muscle

PAGE 67

67 to the serosa then it is considered stage IIA. Stage IIB indicates that the colon cancer spread through the serosa but not to an y lymph nodes/organs. The cancer is considered stage IIC if it has spread through the serosa, and to nearby organs. Stage II five year survival rates for IIA and IIB are 84.7% and 72.2% respectively. No survival rate for stage IIC was reported but is ex pected to be lower than stage IIB. 64 Colon cancer at stage III is again divided into three sub stages: stage IIIA, stage IIIB and stage IIIC. Stage IIIA is characterized by two different conditions: 1) th e cancer has spread through the mucosa to the submucosa and a muscle layer, and has spread to 1 3 nearby lymph nodes or tissue OR 2) the cancer has spread through the mucosa to the submucosa and 4 6 nearby lymph nodes. Three different scenarios can be see n in stage IIIB of colon cancer: 1) cancer spread through the muscle and the serosa, as well as to 1 3 nearby lymph nodes or tissue OR 2) cancer spread through the muscle to the serosa and 4 6 nearby lymph nodes OR 3) cancer spread through the mucosa to th e submucosa/muscle layer and 7 or more nearby lymph nodes. The final sub stage of stage III is stage IIIC, where there are the following three possibilities: 1) cancer has spread through the serosa to 4 6 nearby lymph nodes but to no nearby organs OR 2) cancer has spread to or through the serosa and 7 or more nearby lymph nodes but to no nearby organs OR 3) cancer has spread through the serosa to nearby organs and to one or more nearby lymph nodes/tissues. Five year survival rates are much lower for stag e III colon cancer with the exception of stage IIIA. Stage IIIA, IIIB, and IIIC have five year survival rates of 83.4%, 64.1% and 44.3%, respectively. 64 When colon cancer has spread through the blood and l ymph nodes to other parts of the body it is considered to be in the last stage, stage IV. At this point, the five year survival rate is

PAGE 68

68 only 8.1%. 64 All the descriptions of the stages of colon cancer were taken from the 65 Most previous research has focused on transcriptional regulation of genes associated with CRC; therefore, the effect of CRC on metabolites and protein expression is not well understood. 60, 66 ] Previous research has shown that different cancers could be distinguished (colon and rectal), but successful discrimination between stages of the cancer could not be observed. 66 67 One study that compared CRC tissue to normal control tissue found higher up take of gluc ose in CRC tissues, as well increased glycolysis in those same tissues. 68 Interestingly, Warburg proposed the idea of cancer cells using glycolysis to create energy in 1956. 66, 6 9 This phenomenon was shown in another study of CRC and normal control tissues as well. 70 The majority of metabolomic CRC studies have been on tissues 67 71 72 but some stu dies have been performed on serum from persons with CRC. 73 Advantages of plasma analyses, compared to tissue analyses include: simple sample collec tion the potential for rapid screening and no extraction method is needed (after protein precipitation) For these reasons, we have focused on studying the metabolomics of human plasma from 18 healthy controls and 18 colon cancer patients. Colon and rec tal cancers are different from each other and therefore may have different metabolic profiles. Our study focuses on colon cancer to avoid complicating the experiment with two cancers. Due to the nature of this disease, both the large number of individuals that are affected by it and the large number of deaths caused by it each year, new research on colon cancer is very important. This is the first global metabolomics work to focus only

PAGE 69

69 on colon cancer patient plasma compared to healthy controls. Comparing these results to previous CRC studies may provide more insight about the metabolomics of colon cancer. This may allow for a better understanding of such a fatal disease so that better treatments can eventually be implemented. Experimental The experiment al conditions are the same as those described in C hapter 2 unless otherwise indicated below. Sample Collection and Preparation Whole blood was collected between May 2005 and May 2009 from 36 individuals ; 18 healthy controls and 18 colon cancer patients wi th stage II or stage III cancer (each stage is composed of level A, B, and C) Within each group (controls and cases) there were equal numbers of males and females (9 each). The m ean patient age was 65.6 years old and the median age was 67.5 years old The control group had a mean age of 65.5 years old and a median age of 67.5 years old. The sample demographics are shown in Table 3 1. The samples were prepared as discussed in C hapter 2 Plasma was protein precipitated and evaporated to dryness. The d ri ed products were dissolved in 100 L of 40% acetonitrile, 40% methanol, 10% water and 10% isopropanol. Liquid Chromatography Conditions The LC conditions are described in C hapter 2 and Table 2 2 An injection volume of 5 L was used. All samples were an alyzed in triplicate for a total of 108 LC/MS files.

PAGE 70

70 Electrospray Source The source co nditions used are indicated in C hapter 2 The only change made was a gas temperature of 325 C was used instead of 300 C, as recommended by the manufacturer. Mass Spectro meter The TOF mass spect rometer settings are listed in C hapter 2 MS/MS Instrumentation Analyses were also performed in positive ESI mode on a Velos dual pressure linear ion trap (LIT) mass spectrometer ( Thermo Scientific Waltham, MA ) capable of MS/MS ana lyses Chromatography conditions used for the LC/TOF analysis were copied on an Accela 600 LC system (Thermo Scientific Waltham, MA). The MS/MS method included a parent mass list of 64 masses determined to be influential to the PLS DA plot, as well as t hose masses with the highest fold change. In addition to the parent mass list a data dependent acquisition was implemented to perform MS/MS on the most intense ion if no parent mass was found. Other settings included: source voltage of 3 000 V, source te mperature of 350 C, CID activation, collision ene rgy of 35%, isolation width of two, and a signal threshold of 10 000 counts. By using both accurate mass and MS/MS fragmentation compound identification was possible for some ions of interest. Data Analysis As described in C hapter 2 Qualitative analysis ( Agilent Technologies, Inc., Santa Clara, California). The abundance threshold was set to 2,500 a.u., as determined by examining the intensity of the background.

PAGE 71

71 Files did not require recalibration since a reference solution was sprayed along with the LC flow for constant real time calibration. No targeted approach was taken in this study b ut some statistically significant m/z Agilent Technologies, Inc., Santa Clara, California). This allowed for the calculation and comparison o f average abundances of particular compounds in healthy controls com pared to cancer patients and vice versa. MS/MS data analysis was performed with Ther m Browser ( Thermo Scientific, Waltham, MA). Results and Discussion A total of 3,858 aligned compounds were determined to be present in the 108 d ata files. After performing an asymptotic t test ( p = 0.01), with a Benjamini Hochberg FDR multiple testing correction, 165 masses were determined to be significant. A 3 D PCA plot of the 165 significant masses can be seen in Figure 3 2 In the 3D PCA th e x axis accounts for 39.17% variation, while the y and z axes represent 20.67% and 6.11% of the variations, respectively. The 2D PCA is also shown in Figure 3 3 By examining the PCA plots it can be noted that all the controls group together, as well a s a few cases, but most of the cases seem to spread out away from the controls. There are six case samples that make up the small cluster near the control samples in the PCA plots. This suggests that although the colon cancer is present in these six pati ents, the samples may not contain as many of the significant features found in the rest of the case samples. This is why the six samples group closer to the controls tha n the remainder of the case samples. W hen the significant masses were examined more c losely, this was

PAGE 72

72 exactly the cause. Many of the significant masses determined to be more intense in the cancer patients than in the controls are missing from these six individuals. The significant masses determined by the t test ( p = 0.01) were also used to create a 3D and 2D PLS DA plot (Figure 3 4 and 3 5 ) The PLS DA yielded a prediction accuracy of 93.52% (Table 3 2 ) In the PLS DA model the case samples had a prediction accuracy of 100%, while the control samples were only predicted correct for 47 of the 54 replicates. The loadings plot (Figure 3 6 ) was used to examine the masses that were farthest from WC0 = 0 ( WC0 < 0. 1 or WCO > 0. 1 ) These masses ( 22 ) will be discussed below. Influential PLS DA Masses From the PLS DA plots, 22 masses were dete rmined to be the most influential in the separation observed. These masses were further investigated Quantitative Analysis software ( Agilent Technologies, Inc., Santa Clara, California). More specifically, the extracted ion chromatographs (EICs) for each of the 22 m/z were plotted for all 108 data files and the peak areas were extracted. The extraction was performed using Quantitative Analysis and was manually inspected. Any ions that had unusual repeatability between replicates, or wer Analysis For each ion, two average peak areas were calc ulated: one for the cases and one for the controls. The standard deviation of the mean was also calculated for each average. Of the 22 masses, 11 were used to create a bar graph comparing the average peak areas for each group (control and case). The reason for only 11 masses being depicted in Figure 3 7 is that a mass had to be present in at least 60% of the samples in at least one of the groups (in at least 11 of the 18 samples for case or control) to be

PAGE 73

73 considered. Additionally, a signal to noise ratio (S/N) of at least 3 was necessary to be considered present in any sample. By examining Figure 3 7 it is obvious that many of the significant masses influencing the PLS DA plots are present only in the cancer These masses were investigated further by a database search as well as with MS/MS analysis. Significant Ma sses with Highest FC Fold change is often used to determine what masses are changing most from one group to another. Additionally, it can be used to filter a large number of significant masses down to a more manageable number. This is sometimes done by c oncentrating on significant masses that also have a fold change of at least 2 from one group to another. The masses with the highest fold change (FC) from either case to control or control to case were considered in this work. The masses that passed th e t test and also had the highest FC (FC >1400) are compared in Figure 3 8 (14 masses) This was done in the same manner described under the heading Influential PLS DA Masses, which is above As can be seen in Figure 3 8 all of the masses are found in e ither the case or control samples, but not both. Also, a majority of the masses are DA masses. A database search and MS/MS analysis of these masses will be discussed lat er. Th e fact that the FC masses were only present in either the case or control samples is both a con and a pro for data analysis. The negative aspect is that the statistical software is putting in a very small value for the abundance of a mass that is ac tually not present in either the case or the control samples. Thus, when the fold change is used to filter masses it results in very high changes and is essentially only

PAGE 74

74 determining masses present in only one of the groups. This makes it difficult to loo k at masses that are actually present in both groups and changing greatly. The advantage of this is that a compound not present in con trol samples but present in cancer patient samples may be a very easy biomarker to screen for. This is because the conce ntration of the biomarker present would not be as important as just determining if it is detectable. Often times, when a biomarker is found it is either elevated or found to be decreased in the diseased samples. This means developing a diagnostic tool for colon cancer would be more difficult, since the concentration would need to be taken into account. Database Search of Significant Masses As mentioned, t he t test masses were further filtered by considering masses in the PLS DA loadings plot that satisfied the conditions of WC0 < 0. 1 or WC0 > 0. 1 (Figure 3 6 ). This resulted in 11 masses to consider which were then applied to an accurate mass database search. The database search was performed using METLIN. 58 Th e mass of the positively charged ion was used in the database search and a mass accuracy of 10 ppm was applied. Two different ion options were searched, [M+H] + and [M+Na] + as discussed in C hapter 2 The database search of [M+H] + ions for the PLS DA mass es resulted in hits for 8 of the 11 masses searched. These are listed in Table 3 3 Most of the results generated multiple potential compounds for one m/z with the same formula. No ID is given for those search results. As would be expected, some ions r esulted in multiple potential formulas, such as m/z 303.2159 in Table 3 3 These possible identifications may help make the MS/MS characterization easier. The [M+Na] + ion search results are listed in Table 3 4 for the PLS DA influential masses. Only 4 o f the 11 masses received hits in the database search. This may be because the masses are more likely [M+H] + ions.

PAGE 75

75 Masses that passed the t test were also filtered by considering the significant masses that had the highest FC from control to case or case to control (Figure 3 8 ). These 14 masses were also applied to an accurate mass database search, as described above and in C hapter 2 A search of [M+H] + ions resulted in hits for 4 of the 14 masses, three of which ha d single formula hits (Table 3 5 ). Th e [M+Na] + database search resulted in hits for 7 of the 14 ions searched. The results for the [M+Na] + ion search are shown in Table 3 6 when characterizing the significant masses by MS/MS. MS/MS of Significant Masses Identifying compounds based on accurate mass alone is im possible. By implementing MS/MS fragmentation patterns were found for most of the significant ions. Identification was done for some ions by use of accurate mass and MS/MS databa ses as well as literature searches for specific ions or neutral losses observed in the spectra. Before describing the resulting data the MS/MS instrumentation will be mentioned briefly A schematic of the dual pre ssure LIT is shown in Figure 3 9 In gene ral, ions formed at the source are transferred into the MS and go through the S lens. The S lens is a series of flat ring electrodes with radio frequency (RF) voltage applied to them. Every other electrode has an opposite phase RF and each ring is an inc reased distance from the last. The S lens openings were designed to allow optimal ion transfer while minimizing the RF voltage required for focusing. 74 After ions are guided through two quadrupoles and an o ctopole they reach two trapping cells maintained at different pressures. The first cell is held at approximately 5 X 10 3 Torr. One benefit of higher pressure in the first cell is improved fragmentation efficiency over typical LITs. The

PAGE 76

76 second cell is m aintained at about 4 X10 4 Torr and allows improved mass resolution for a given scan rate, or this resolution can be sacrificed for a higher scan rate. 74 The dual pressure LIT is faster and more sensitive th an its previous models but still lacks the mass accuracy that is desired for metabolomic applications. When examining the PLS masses four ions were quickly identified due to correct accurate mass database results and corresponding MS/MS spectra. All four of these compounds were carnitines. More specifically, myristoleylcarnitine (C14:1) myristoylcarnitine (C14:0) palmitoleylcarnitine (C16:1) and linoleoylcarnitine (C18:2) were identified. Table 3 7 lists the experimental masses observed with the LC/TO F instrumentation, the theoretical masses, as well as the calculated retention times observed. The MS/MS spectra for each of the four identified carnitines in plasma from colon cancer patients can be seen in Figures 3 10 to 3 1 3 Carnitines are relativel y easy to identify by MS/MS fragmentation due to a characteristic neutral loss of 59, corresponding to the trimethyl amine, as well as the neutral loss of 161 from the loss of carnitine. For carnitines, an ion is also observed at m/z 144 that is dehydrate d carnitine Thi s ion can be seen in Figure 3 1 1 although it is present in all four of the MS/MS spectra. A standard was analyzed for C14:0 and is compared to the MS/MS spectrum from plasma in Figure 3 1 1 MS/MS spectra were acquired for several of the fold change masses. Unfortunately, no compounds have been able to be identified. Some characteristics of the masses could be determined, such as if it was containing a hydroxy group, a phosphate group or a fatty acid chain. However, enough information t o identify the comp ounds was not obtained.

PAGE 77

77 Spectra was acquired for several of the remaining significant masses (not grouped with the FC masses or the influential PLS mass es) as well. Another carnitin e which was observed to be elevated in cancer cases, w as identified at m/z 344 as dodecanoylcarnitine (C12:0). The same neutral losses and similar fragment ions were observed in the MS/MS spectrum of m/z 344 from colon cancer patient plasma as other carnitines already discussed (Figure 3 1 4 ). A standard was obtained for C12:0 and the standard spectrum matched the experimental spectrum from colon cancer patient plasma (Figure 3 1 4 ). Biological Significance It is interesting that five significant compounds were identified as carnitines with varying fatty acid tail lengths. Previous work with CRC patient serum samples compared to healthy controls has suggested a down regulation of fatty acids such as myristic acid (C14 chain ) and palmitic acid (C16 chain ) in cancer samples 73 The conclusion of this finding was a dysfunction in the oxidation pathway used for fatty acid metabolism. Another study of CRC tissue compared to healthy tissue resulted in up regulation of oleic acid (C18:1 chain ) and stearic acid (C18 chain ) in CRC samples. 67 The autho rs suggested that this increase occurs with increased cell death that may be occurring more frequently in the malignant tissue. 75 Another study of colon tissue and healthy tissue showed elevated palmitic and oleic acids in healthy tissue. 71 These results match those already discussed. In is unclear if the carnitines identified in this work would be associated with the cell death mentioned by Lehtimaki et al., but it is l ikely that that some type of defect is occurring in the oxidation pathway. In particular, when medium and long chain carnitines are elevated metabolism disorders are often the culprit. These disorders can

PAGE 78

78 sometimes be the result of a defect in carnitin e palmitoyltrasferase (CPT2) in the mitochondria. As shown in Figure 3 15, free carnitine is used to transfer fatty acids into the mitochondria via the translocase. The acylcarnitines (carnitine with varying fatty acid tails) are then converted back to f ree carnitine and Acyl CoA by CPT2. 76 If a defect occurs in CPT2 then one might suspect an increase in acylcarnitines, as was a result in this research. Finally, Wright et. al. found an increase in the myristoyl ation of proteins in a rat colon cancer model. 77 If increased myristoylation is occurring in individuals with colon cancer then an increase in the transportation of C14 chains may be necessary. Thus, an elevation o f myristoylcarnitine may be observed in colon cancer patients, such as was found here. Obviously more biological understanding of the processes of the identified metabolites must be acquired. With a better understanding treatments for colon cancer can be considered and experiments can be designed for these treatments. Challenges Faced There were several challenges associated with the MS/MS aspect of this work. Firstly, although no data was presented here, a significant amount of time was spent attempting to utilize a quadrupole TOF (Q TOF) instrument located at Moffitt Cancer Center. A major flaw in the Q TOF system was the ionization source. The Q TOF was equipped capable of mobile phase flow rates of approximately 2 40 L/min Due to the higher flow rate used in this work (800 L/min) a split was required for compatibility. However, when the split was used there was a large decrease in the sensitivity of the instrument. This was a result of most of the analyte going to waste. In order to solve this problem a capillary version of the monolithic column was purchased

PAGE 79

79 from Phenomenex. Unfortunately, the capillary column offered its own difficulties, especially with sample overloading. It is believed that the column was quickly damag ed from the manufacturer and configured to the Q TOF. The new source was capable of up to 1 mL/min flow rates. However, when the chromatography method used in this w as implemented with the new source desolvation suffered. This was evident by the accumulation of solvent within the source, even with a flow rate of only 0.6 mL/min. The lack of desolvation is believed to have caused a loss of signal that was observed at the end of one day of research. The periodic loss of signal resulted in the necessity of finding a different MS/MS capable instrument to use. In addition to instrumental problems, the process of identification of compounds by accurate mass and MS/MS frag mentation proved to be difficult. While accurate mass libraries such as HMDB and METLIN are useful they are still incomplete. This means that many significant masses that are searched result in no hits. Also, the amount of MS/MS spectra available in the se databases is even more limited. Due to this, more literature searching is necessary to try to find MS/MS work that has already been done on database results from accurate masses. Additionally, specific neutral losses or ions observed in MS/MS can be s earched in the literature. Another challenge that was faced when examining the MS/MS spectra of the significant masses was that many were related to each other by either a loss of water or the addition of sodium. This is supposed to be addressed in the so ftware, or least the addition of sodium should be easily determined. Obviously this is not always the case so careful consideration of retention times and ions present in the MS/MS spectra should be used A recent

PAGE 80

80 publication also mentioned the many cha llenges associated with metabolomics research, and the bottleneck of the work flow is the actual identification of features. 78 With this in mind, t here are many advances to be made in software capable of processin g metabolomics data as well as in expanding accurate mass and MS/MS libraries. Summary LC/TOF analysis of plasma from colon cancer patients and healthy controls resulted in 165 significant masses. These masses were also analyzed by LC/LIT to obtain MS/MS fragmentation patterns. To simplify the identification process, masses that were found to be influencing the PLS DA plot or to have the highest fold change were considered initially. This resulted in the identification of f ive carnitines. Considering a ll other significant features one more compound was identified. Some of these results suggest a disturbance in the oxidation pathway.

PAGE 81

81 Table 3 1. Sample demographics. The median age for both groups is 67.5 years old. Colon Cancer Patients Healthy Co ntrols # of Samples 18 18 Age (mean, range) 65.6, 40 86 65.5, 40 88 Sex (male, female) 9, 9 9, 9 Stage 2 (A, B, C) 7, 0, 0 N/A Stage 3 (A, B, C) 2, 4, 5 N/A Table 3 2 PLS DA prediction results for human plasma from healthy controls and colon cancer patients Predictions Case Control Accuracy (%) 54 0 100 .0 7 47 8 7 04 93 52 There are 54 files for each group, thus each should have a prediction of 54, those that vary cause the accuracy of the model to decrease.

PAGE 82

82 Table 3 3 Potential for mulas from [M+H] + database search of influential PLS DA masses. m/z Searched Formula Theoretical m/z Error (ppm) ID 289.2005 C 15 H 28 O 5 289.2010 1 1,2 Dihexanoyl sn glycerol 303.2159 C 12 H 26 N 6 O 3 C 16 H 30 O 5 C 17 H 26 N 4 O 303.2139 303.2166 303.2179 6 2 6 E medastine 372.3111 C 21 H 41 NO 4 372.3108 0 Tetradecanoylcarnitine 398.3264 C 23 H 43 NO 4 398.3265 0 Hexadec enoyl carnitine 424.3417 C 25 H 45 NO 4 424.3421 1 Linoleyl carnitine 427.2694 C 23 H 38 O 7 C 19 H 41 NO 7 P 427.2690 427.2693 0 0 431.2640 C 23 H 34 N 4 O 4 431.2653 2 467.2619 C 22 H 43 O 6 PS C 25 H 38 O 8 467.2591 467.2639 6 4 (2S) OMPT A mass accuracy of 10 ppm was required. ID was given if no other compounds with the same formula resulted. Table 3 4 Potential formulas from [M+Na ] + database search of influential PLS DA masses. m/z Searched Formula Theoretical m/z Error (ppm) ID 431.2640 C 26 H 36 N 2 O 2 431.2669 6 O 2545 (hydrochloride) 467.2619 C 23 H 40 O 8 467.2615 0 485.3108 C 28 H 46 O 3 S C 30 H 42 N 2 O 2 485.3060 485.3138 9 6 Arachidonoyl Serotonin 489.3037 C 27 H 46 O 4 S 489.3009 5 Cholesterol Sulfate A mass accuracy of 10 ppm was required. ID was given if no other compounds with the same formula resulted.

PAGE 83

83 Table 3 5 Potential formulas from [M+H]+ database search of masses with the highest FC. m/z Searched Formula Theoretica l m/z Error (ppm) ID 277.0626 C 13 H 12 N 2 O 3 S 277.0641 5 Sulfabenzamide 283.0705 C 13 H 9 F 3 N 2 O 2 283.0689 5 Niflumic Acid 445.2800 C 23 H 40 O 8 445.2 796 0 597.3603 C 33 H 48 N 4 O 6 597.3647 7 L Urobilinogen A mass accuracy of 10 ppm was required. ID was given if no other compounds with the same formula resulted. Table 3 6 Potential formulas from [M+Na] + database search of masses with the highest FC m/z Searched Formula Theoretical m/z Error (ppm) ID 261.0885 C 16 H 14 O 2 261.0886 0 4' Methoxychalcone 277.0626 C 1 1 H 14 N 2 O 3 S 277.0617 3 485.3108 C 28 H 46 O 3 S C 30 H 42 N 2 O 2 485.3060 485.3138 9 6 Arachidonoyl Serotonin 485.3107 C 28 H 46 O 3 S C 30 H 42 N 2 O 2 485.3060 485.3138 9 6 Arachidonoyl Serotonin 513.3416 C 30 H 50 O 3 S 513.3373 8 537.2824 C 30 H 42 O 7 C 31 H 46 O 2 S 2 537.2823 537.283 1 0 1 Probucol spiroquinone 583.3451 C 25 H 48 N 6 O 8 583.3426 4 Deferoxamine A mass accuracy of 10 ppm was required. ID was given if no other compounds with the same formula resulted.

PAGE 84

84 Table 3 7. List of metabolites identified by accurate mass and MS/MS f ragmentation from the PLS DA influential mass list. Metabolite Experimental m/z Theoretical m/z Error (ppm) Retention (min) Dodecanoylcarnitine* C 19 H 37 NO 4 344.279 9 344.279 5 1.16 22.9 Myristoleylcarnitine C 21 H 39 NO 4 370.2961 370.2952 2.43 23.7 Myrist oylcarnitine* C 21 H 41 NO 4 372.3111 372.3108 0.81 24.7 Palmitoleylcarnitine C 23 H 43 NO 4 398.3264 398.3265 0.25 25.2 Linoleoylcarnitine C 25 H 45 NO 4 424.3417 424.3421 0.94 25.6 Indicates verification by standard

PAGE 85

85 Figure 3 1 Areas affected by and used in staging colon cancer 65 Figure 3 2 3D PCA plot of the significant masses in plasma from colon cancer patients and healthy controls (165 masses, p = 0.01). Each point represents one replic ate.

PAGE 86

86 Figure 3 3 2D PCA plot of the significant masses in plasma from colon cancer patients and healthy controls (165 masses, p = 0.01). E ach point represents one replicate. The color code is the same as Figure 3 3. Figure 3 4 3D P LS DA plot of t test ( p = 0.01) masses in human plasma from healthy controls and colon cancer patients (165 masses). Each point on the plot depicts one replicate.

PAGE 87

87 Figure 3 5 2 PLS DA plot of t test ( p = 0.01) masses in human plasma from healthy controls and colon can cer patients (165 masses). Each point on the plot depicts one replicate. The x axis (Identifier) values are the sample replicates. The color code is the same as Figure 3 5. Figure 3 6 PLS DA loadings plot of the significant masses determined by a t test (165 masses p = 0.01 ) in human plasma from healthy controls and colon cancer patients. WC0 indicates how influential a m/z value is (largest and smallest WC0 values are more influential in PLS DA separation).

PAGE 88

88 Figure 3 7 Bar graph of the average abundances of masses furthest from WC0 = 0 in the PLS loadings plot. A mass had to be present in 60% of the samples in one condition (case or control) to be considered. Error bars indicate the standard deviation of the mean. Metabolites identified were m/z 370.2961, 372.3111, 398.3264, and 424.3417 (all carnitines). 0.00E+00 1.00E+05 2.00E+05 3.00E+05 4.00E+05 5.00E+05 6.00E+05 7.00E+05 8.00E+05 9.00E+05 1.00E+06 Average Abundance (a. u.) m/z Average Intensities of Influential PLS Masses Control Case

PAGE 89

89 Figure 3 8 Bar graph of the average abundances of the significant masses with the highest fold change. A mass had to be present in 60% of the samples in one condition (case or control) to be considered. Error bars indicate the standard deviation of the mean. 0.00E+00 5.00E+05 1.00E+06 1.50E+06 2.00E+06 2.50E+06 3.00E+06 Average Abundance (a.u.) m/z Average Intensities of Significant Masses With Highest FC Control Case

PAGE 90

90 Figure 3 9 Schematic of the Velos dual pressure LIT. 74

PAGE 91

91 Figure 3 10 MS/MS spectrum of m/z 370 in plasma from a colon cancer pa tient. The neutral losses (NL) of 59 and 161 are due to the loss of the trimethyl amine (below blue arrow) and the loss of carnitine (left of orange arrow), respectively.

PAGE 92

92 Figure 3 1 1 MS/MS spectra of m/z 372 in plasma from a colon cancer patient (top) and a standard (bottom) The neutral losses (NL) of 59 and 161 are due to the loss of the trimethyl amine (below blue arrow) and the loss of carnitine (left of orange arrow), respectively. The ion at m/z 144 is indicative of dehydrated carnitine (left of green arrow).

PAGE 93

93 Figure 3 12 MS/MS spectrum of m/z 398 in plasma from a colon cancer patient. The neutral losses (NL) of 59 and 161 are due to the loss of the trimethyl amine (below blue arrow) and the loss of carnitine (left of orange arrow), respectiv ely.

PAGE 94

94 Figure 3 1 3 MS/MS spectrum of m/z 424 in plasma from a colon cancer patient. The neutral losses (NL) of 59 and 161 are due to the loss of the trimethyl amine (below blue arrow) and the loss of carnitine (left of orange arrow), respectively.

PAGE 95

95 Fig ure 3 1 4 MS/MS spectra of m/z 344 in plasma from a colon cancer patient (top) and a standard (bottom). The neutral losses (NL) of 59 and 161 are due to the loss of the trimethyl amine (below blue arrow) and the loss of carnitine (left of orange arrow), respectively. The ion at m/z 144 is indicative of dehydrated carnitine (left of green arrow).

PAGE 96

96 Figure 3 1 5 Carnitine palmitoyltransferase (CPT) enzyme activity in the mitochondria 76

PAGE 97

97 CHAPTER 4 METHOD OPTIM I Z ATION: TOWARDS MORE EFFICIENT IONIZATION AND SHORTER ANALYSIS TIM E Overview In metabolomic experiments it is important to be able to ionize and detect as many compounds as possible. Compounds that are not detected could still be potential biomarker s. F r equently researchers will run experiments in both positive and negative modes as well as with ESI and atmospheric pressure chemical ionization (APCI) sources to broaden the types of compounds detected. 9 36 79 Each of these ionization techniques should be optimized separately. Efficient ionization is often achieved by good chromatographic separation prior to ionization ; this preve nts or reduces ion suppression. Although chromatography is necessary, it should be timely to allow all samples to be analyzed together instead of in multiple batches where instrument variability is increased. In particular, shared instruments have limi ted free time so shorter chromatographic run times are helpful. It is also more economical since solvent consumption is reduced. This chapter concentrates on the optimization of positive ESI as well as different LC time gradient lengths and LC column s to be coupled to the source. One technique that is used in this work, in order to compare the various optimization conditions and LC methods, is the extraction and comparison of features detected with LC/MS analyses This technique is also used in the lab o f one of the leaders in metabolomic research, Gary Siuzdak, at The Scripps Research Institute. 9 80 81 Metabolomics software pack ages typically begin data analysis by extracting all features (or m/z and retention time pairs). These features can therefore be used to represent the number of different components that were able to be detected by the

PAGE 98

98 mass spectrometer. Thus, one can co mpare the number of features from one method to another to determine the conditions that allow the highest number of compounds to be detected. As previously mentioned, in global metabolomics it is desirable to detect the largest number of compounds. Beca use of this, conditions that achieve the highest n umber of features were implemen ted in this work. Finally, a hydrophilic interaction liquid chromatography (HILIC) column was optimized. As can be seen by the chromatograms in Figures 5 1 and 5 2, urine con tains many polar analytes that are not well retained with RP chromatography. Due to the nature of HILIC, the elution order of analytes is essentially reversed, allowing polar compounds to be better retained and in return non polar compounds are less well retained. 82 Sometimes both HILIC and RP chromatography are implemented in order to achieve complimentary results. 83 84 The opti mized HILIC method w as later implemented for the separation of manatee urine prior to mass spectrometric detection (Chapter 5). Experimental Chromatography Methods Three chromatography methods were used for ionization optimization and feature extraction of urine and plasma samples. Feature extractions were done to determine the parameters that led to the highest number of features being detected. The three methods are shown in Table 4 1 and described below. The monolithic method uses two Phenomenex (Phenom enex, Torrance, CA) monolithic columns and a 50 minute gradient elution analysis. The short monolithic method uses similar conditions and the same columns but the analysis time is only 27 minutes. The final chromatography method the Zorbax method uses an Agilent Zorbax column (Agilent Technologies, Santa Clara, CA). The Zorbax column is a small particle column as opposed to a

PAGE 99

99 monolithic column and is more frequently used with fast chromatography. The Zorbax method is a 25 minute analysis. Four diffe rent methods were also used to compare the Zorbax and monolithic columns to determine if one was more chromatographically efficient than the other. As described in Table 4 2, each column was implemented under its normal conditions and then a second method was created to mimic the other columns conditions. Feature extractions were compared for piglet plasma analyzed by each of the methods. Finally, work was done to op timize the solvents used with a HILIC column. Four different solvent systems were impleme nted with the HILIC column and amino acids and carnitines were examined for chromatographic efficiency and retention. The four methods are shown in Table 4 3 The column implemented was a Halo Penta HILIC ( Advanced Materials Technology, Inc. Wilmington, DE). Similar to the Zorbax method, the HILIC column used a 25 minute time analysis, allowing for a more time efficient analysis of large sample sets. Electrospray Source The conditions used a re the same as those listed in Chapter 3 or are otherwise state d (Capillary Voltage and Fragmentation Voltage sections). Mass Spectrometer The mass spectrometer condit ions are identical to those of Chapter 2 and Chapter 3 Capillary Voltage The capillary voltage is applied to the front end of the capillary (Vcap) at t he entrance of the MS as shown in Figure 1 3. For the Agilent TOF MS, common capillary setting are between 3000 4000 V.

PAGE 100

100 The f ragmentor and all other settings were held constant while the capillary was varied between 3000 and 4200 V Carnitine standards were used to determine the optimal capillary voltage. The standards included: carnitine (C0), hexanoylcarnitine (C6), octanoylcarnitine (C8), lauroylcarnitine (C12), myristoylcarnitine (C14), and stearoylcarnitine (C18) The short monolithic method was u sed for this experiment (Table 4 1) The average peak area (n = 3) of the extracted ion chromatograms (EIC) was determined and plotted as a function of the capillary voltage. Fragmentation Voltage The fragmentor voltage is applied at the back end of the capillary (fragmentor), as shown in Figure 1 3. For the Agilent TOF MS, common fragmentor voltages are 120 180 V. For this study the c apillary and all other settings were held constant while the fragmentor was varied between 70 and 240 V Carnitine stan dards of carnitine (C0), hexanoylcarnitine (C6), octanoylcarnitine (C8), decanoylcarnitine (C10), lauroylcarnitine (C12), and myristoylcarnitine (C14) were injected (n = 3) and the Zorbax method was implemented (Table 4 1) The average peak area from the EIC was plotted as a function of the fragmentor voltage. Feature Extraction of Urine Samples Manatee urine was analyzed in triplicate to find the optimal conditions to produce the largest number of features in urine Conditions that were considered were different capillary/fragmentor voltages as well as a longer and shorter LC time gradient. This was done by using both the monolithic a nd short monolithic methods (Table 4 1) and capillary/fragmentor settings of 3000 V/120 V, 3000 V/180 V, 4000 V/120 V, 4 000 V/180 V, and 3600 V/150 V for each of the LC methods. The f eatures were extracted using

PAGE 101

101 Mass Profiler Professional (Agilent Technologies) from each data file and averaged for each group (n = 3). ANOVA was implemented to determine if there was a sign ificant difference between the average number of features detected with the short monolithic method compared to the monolithic method. Feature Extraction of Plasma Samples Piglet plasma was analyzed in triplicate to find the optimal conditions to produce t he largest number of features in plasma. Conditions that were considered were different capillary/fragmentor voltages as well as a longer and shorter LC time gradient. This was done by using both the monolithic and short monolithic methods (Table 4 1) an d capillary/fragmentor settings of 3000 V/120 V, 4000 V/120 V, 4000 V/180 V, and 3600 V/150 V for each of the LC methods. The features were extracted using MPP from each data file and averaged for each group (n = 3). ANOVA was also used to determine if a significant difference existed between the average number of features detected with the short monolithic method compared to the monolithic method. In order to compare the methods further, five different compounds were also evaluated in detail. The EIC f or each m/z value were used to determine the retention times (t r ) and base peak widths (w). These values were then used to calculate the column efficiency (N). This was performed for the capillary/fragmentor settings of 4000 V/120 V only. Comparison of M onolithic and Small Particle Columns In order to determine if a different column would be more efficient, the monolithic column was compared to a small particle Zorbax column. Table 4 2 shows the four

PAGE 102

102 different LC methods that were used with the two diffe rent columns 79 In addition to each individual method (MC and ZC methods) another method was made that attempts to use similar conditions for a fair comparison (MC with ZC conditions and ZC with MC condtions me thods). Piglet plasma was injected in triplicate for each of these methods. Capillary and fragmentor voltages were maintained at 4000 V and 120 V, respectively. It is important to note that since the Zorbax column contains small particles it was always heated to 60 The number of features for each analysis was determined with MPP and the same five compounds were explored as in the Feature Extraction of Plasma Samples section The EIC s for each m/z value w ere used to determine the average retention times (t r ) and base peak widths (w). These values were then used to calcul ate the column efficiency (N). In addition to the five compounds chosen at random, standard mixtures of carnitines were analyzed using e ach LC method and the same calculations were performed. Carnitines that were in the mixture include: carnitine (C0), hexanoylcarnitine (C6), octanoylcarnitine (C8), lauroylcarnitine (C12), myristoylcarnitine (C14), and stearoylcarnitine (C18). Evaluation of Different Mobile Phases for u se with the Zorbax Column Different mobile phase compositions were explored for uses with the Zorbax column method. The LC method used was the ZC method shown in Table 4 2. The solvents were the only condition changed for this study. T hree different pairs of mobile phases were used The first set of mobile phases analyzed were the same as those listed on Table 4 2. The second set of mobile phases used water (H 2 O) as mobile phase A and acetonitrile (ACN) as mobile phase B, the amount of AA added to each (A and B) was 0.1% (v/v) instead of 1% (v/v) used in the first set. The final set of mobile

PAGE 103

103 phases used water as mobile phase A and methanol (MeOH) as mobile phase B t he amount of AA added to each (A and B) was 0.2% (v/v ). 79 These three different LC solvent sets were used with the Zorbax column to analyze piglet plasma in triplicate. The number of features were determined with the MPP software and averaged for the three repli cates in each solvent set. The same five compounds that were explored as in the Feature Extraction of Plasma Samples section were examined. The EICs for each m/z value were used to determine the average retention times (t r ) and base peak widths (w). The se values were then used to calcu late the column efficiency (N). Optimization of Sol vent Systems for use with HILIC Standards of carnitine (C0), hexanoylcarnitine (C6), octanoylcarnitine (C8), decanoylcarnitine (C10), lauroylcarnitine (C12), myristoylcarni tine (C14) L anserine, L arginine, L carnosine, creatine, L histidine, L homocystine, gamma DL hydroxylysine, L lysine, 1 methyl L histidine, 3 methyl L histidine, L ornithine and L tryptophan were examined by HILIC ESI/TOF MS with four different solvent systems. The same HILIC column, flow rate, column temperature, and gradient were used, only the solvents differed. The four solvent systems are described in Table 4 3. The first two solvent systems had 5 mM ammonium formate (A m F) in each mobile phase A and B and were at pH 3.75. Both A m F methods used the same mobile phase B, 50% water and 50% acetonitrile with 5 mM A m F. The first method, named A m F 95:5 had mobile phase A of 95% acetonitrile and 5% water with 5 mM A m F. The second method (A m F 90:10) onl y differed from the first by mobile phase A, which was 90% acetonitrile and 10% water with 5 mM A m F. The second two solvent systems used with HILIC had 5 mM ammonium acetate (A m A) in each mobile phase A and B and were at pH 5.75. Both A m A methods used th e same mobile phase B, 50% water and 50% acetonitrile with 5

PAGE 104

104 mM A m A. The first A m A method, named A m A 95:5, had mobile phase A of 95% acetonitrile and 5% water with 5 mM A m A. The second A m A method, called A m A 90:10, differed from the previous A m A method b y having a mobile phase A of 90% acetonitrile and 10% water with 5 mM A m A. The gradient used for all methods held 100% mobile phase A for three minutes and then changed to 90% mobile phase B over 17 minutes and then immediately headed back to 100% mobile phase A over three minutes and held 100% mobile phase A for 2 minutes for a total analysis time of 25 minutes. The 18 standards were mixed together and injected in triplicate with each solvent system. The peak areas for each analyte were determined from t he EIC of each + ion. The average retention time and peak widths were also noted so that the efficiency (N) could be calculated for each analyte. Results and Discussion Capillary and Fragmentor Voltages In general, the carnitine standards appeared to have a maximum peak area at a capillary voltage of 3000 V and a fragmentor voltage of 120 V (Figure 4 1 and Figure 4 2). It is clear that a lower fragmentor voltage would promote more efficient ionization, as it w ould be gentler on the ions an d less likely to br eak them. However, it is less apparent why a lower capillary voltage would lead to more ionization. It seems more likely that a higher fragmentation voltage would lead to more ionization. While this information is useful, carnitines ar e not the only compounds of interest in metabolomic experiment s Different types of compounds may not ionize efficiently under these types of conditions. Thus, this study was extended to different biological fluids in order to determine what conditions p roduce the largest number of features.

PAGE 105

105 Feature Extraction of Biofluids Table 4 4 depicts the average number of features (three replicates) detected in manatee urine for each chromatography method and capillary/fragmentor setting. T he short monolithic meth od and capillary/fragmentor voltages of 4000 V/120 V yield the largest average number of features in manatee urine. The piglet plasma results can be seen in Table 4 5 shown as the average number of features (three replicates) detected for each chromatogr aphy method and capillary/fragmentor setting. The short monolithic method and capillary/fragmentor voltages of 4000 V/120 V yield the largest average number of features in piglet plasma. This verifies that a shorter time gradient can be implemented witho ut losing features when working with plasma or urine A NOVA verified that no significant differences existed between the monolithic method and short monolithic method in both pi glet plasma and manatee urine. The column efficiencies in the monolithic and s hort monolithic methods were calculated and compared for the piglet plasma analysis. This was done to verify the chromatographic efficiency did not suffer to o extensively with the shorter time gradient. Table 4 6 lists the column efficiencies for five m/ z plasma analyses with the monolithic method. The same calculations were performed for the piglet plasma analyses with the short monolithic method and listed in Table 4 7 These results show little change in the column efficiency when the time gradient is reduced. This is likely because the peak widths are also decreasing with the shorter gradient, maintainin g the efficiency. The short monolithic method can thus be used for metabolomic experiments with capillary and fr agmentor voltages of 4000 V and 120 V, respectively.

PAGE 106

106 Implementing Different LC Columns To decrease the time length of a gradient and also maintain chromatographic efficiency it is common to either increase the pressure or temperature or to use different types of particles. 23 For these reasons we compare d the monolithic column to a column that utilizes th o se properties (Zorbax column) The results of the feature extraction for this study can be seen in Table 4 8 The numbers of features detected for each of the three replicates of piglet plasma, analyzed with four LC methods were averaged. As noted in Table 4 8 the largest number of features was observed while the ZC method was implemented. The carnitine standard mixture produced resolved chromatographic peaks for all four LC methods shown in Figure 4 3. Addtionally, Tables 4 9 to 4 1 2 list retention times, peak widths and the calculated plate heights of the carnitines for each of the four methods. The Zorbax me thods allowed for smaller plate heights (except in the first two minutes of the analysis) than the methods that equipped the monolithic column. The two Zorbax column methods also allowed for sharper peaks than the two monolithic column methods. This can be seen by comparing the peak widths shown in Tables 4 9 to 4 1 2 T he chromatograms for individual piglet plasma injection s under each of the four methods are shown in Figure 4 4. Much sharper peaks are observed for the two Zorbax column methods than fo r the monolithic column methods. Tables 4 1 3 to 4 1 6 list the retention times, peak widths, and calculated plate heights for five randomly chosen compounds. Similar trends were observed for the carnitine standards as were seen with the five random masses Those trends include smaller peak widths and plate heights f or the methods that use the Zorbax column.

PAGE 107

107 When considering the number of features found for each of the four LC methods, as well as the chromatography results for both carnitine standards and piglet plasma, it is obvious that the ZC methods allows for efficient chromatography and the detection of t he largest number of features. Use of Different Solvents with Zorbax Column The results of the feature extraction from the three solvent sets tested are shown in Table 4 1 7 The largest average number of features was observed for the third solvent set that implemented water (A) and methanol (B) with 0.2% AA (v/v) in each. For the two solvent sets that used acetonitrile instead of methanol, the method that used 0.1% AA instead of 1% AA (v/v) resulted in more extracted f eatures The chromatographic results are shown in Tables 4 1 8 to 4 2 3 Tables 4 1 8 to 4 20 are the chromatographic results of carnitine standard mixtures for each of the three solvent s ets. The most column efficiency is observed for the method using water and m ethanol with 0.2 % AA (Table 4 20 ). This method (0.2% AA) also has longer retention times than the two methods using ACN, adding to the efficiency of the method. The chromatograp hic results of five randomly chosen compounds are shown in Tables 4 21 to 4 2 3 for each of the solvent systems. Other than the first two m/z in Table 4 23 a much higher column efficiency was observed for the method using methanol and 0.2% AA. T he two methods using ACN yielded similar column efficiencies but in general, much lower than the method implementing methanol and 0.2% AA Figure s 4 5 and 4 6 show the chromatograms for carnitine mixtures and piglet plasma, respectively. Similar chromat ographic profiles are observed with ACN with 1%

PAGE 108

108 AA and ACN with 0.1% AA for both the carnitine standard mixtures and the piglet plasma. However, the use of methanol and 0.2% AA shows a shift to longer retention times in the carnitine standard mixture and the piglet plasma sample. I ncreased column efficiency for the carnitine standard mixture as well as the p iglet plasma was observed with the solvent system utilizing methanol and 0.2% AA. Therefore, the solvent system using methanol and 0.2% AA will be imp lanted in metabolomic studies. This will allow for efficient chromatography and more features to be determined based on the work in this section. Optimization of HILIC Analysis of the standard mixture resulted in higher efficiencies for the two solvent sy stems that used ammonium acetate. The efficiencies observed for carnitine standards, which experienced much lower efficiencies than the amino acids, are shown in Figure 4 7. C0 was the only carnitine to be retained with solvents containing ammonium forma te, explaining the low efficiencies obtained. The carnitines were retained better with the ammonium acetate solvents, but still had lower efficiencies than most of the amino acids since the retention was not as long for most of the carnitines as it was fo r the amino acids. The efficiencies calculated for amino acid standards are depicted in the bar graph in Figure 4 8. Most of the efficiencies for the amino acids with ammonium acetate solvents were three times that of the carnitines despite what method w as used. Figure 4 9 shows the retention of all the standards with each of the four solvent systems explored. For each analyte, an ammonium acetate solvent system led to the most retention and the highest efficiency. More frequently, the method labeled A mA 95:5 had more efficiency than AmA 90:10 and will be used in future studies of urine.

PAGE 109

109 Summary The work in this chapter has shown that a capillary/fragmentor voltage of 4000 V/120 V is optimal when studying the metabolomics of urine or plasma. It was als o shown that a shorter time gradient can be implemented with the monolithic column, for urine and plasma, while maintaining efficient chromatography and not losing any molecular features. Additionally, more efficient chromatography (smaller plate heights and peak widths) is seen when studying plasma with a Zorbax column, as com pared to a monolithic column. Also a solvent system utilizing methanol and 0.2% AA allow ed for more efficient chromatography in addition to the detection of more molecular features than with acetonitrile and 1% AA or 0.1% AA. Finally, experiments with HILIC using a Halo column resulted in optimal retention and efficiency with solvents containing ammonium acetate at pH 5.75. Particularly, any work with polar analytes would benefit from analysis with the AmA 95:5 HILIC method, since more analytes had the highest efficiency with this method than the other three methods. In conclusion, future metabolomics work will be performed with capillary/fragmentor voltages of 4000 and 120 V, resp ectively. A 25 minute LC method will be used that implements the Zorbax column. The LC method will use solvents containing 0.2% AA in water (A) and methan ol (B). The metabolomic study of manatee urine, discussed in C hapter 5 used this method and the Ha lo column for HILIC

PAGE 110

110 Table 4 1. Chromatography methods used to optimize ESI settings for standards and biofluids. Method Column Column Dimensions (mm) Mobile Phase A Mobile Phase B Flow Rate (m L/min) Gradient [A:B(min)] Monolithic Phenomenex O nyx C18 Monolithic (2) 4.6 x 100 1% acetic acid (AA) in water 1% AA in acetonitrile 0.8 95:5(0.0 6.5) 0:100(25.5 35.0) 95:5(45.0 50.0) Short Monolithic Phenomenex Onyx C18 Monolithic (2) 4.6 x 100 1% AA in water 1% AA in acetonitrile 0.8 95:5( 0.0 4.0) 0:100(13.0 17.0) 95:5(24.0 27.0) Zorbax Agilent Zorbax SB Aq and SB C8 guard particles) and particles) 0.2% AA in water 0.2% AA in methanol 0.35 95:5(0.0 13.0) 0:100(13.0 19.2) 95:5(24.0 25.0)

PAGE 111

111 Table 4 2 Chromatography methods used to compare the monolithic and small particle column Met hod Column Column Dimensions (mm) Mobile Phase A Mobile Phase B Flow Rate (mL/min) Gradient [A:B(min)] MC Phenomenex Onyx C18 Monolithic (2) 4.6 x 100 1% acetic acid (AA) in water 1% AA in acetonitrile 0.8 95:5(0.0 4.0) 0:100(13.0 17.0) 95:5(2 4.0 27.0) MC with ZC Conditions Phenomenex Onyx C18 Monolithic (2) 4.6 x 100 1% AA in water 1% AA in acetonitrile 0.8 98:2(0.0) 2:98(13.0 19.0) 98:2(24.0 25.0) ZC Agilent Zorbax SB Aq and SB C8 guard 2.1 x 50 (1.8 2.1 x 30 1% AA in water 1% AA in acetonitrile 0.6 98:2(0.0) 2:98(13.0 19.0) 98:2(24.0 25.0) ZC with MC Conditions Agilent Zorbax SB Aq and SB C8 guard 2.1 x 50 (1.8 2.1 x 30 1% AA in water 1% AA in acetonitrile 0.6 95:5(0.0 4.0) 0:100(13.0 17.0) 95:5(24.0 27.0) The column is heated to 60 C in the ZC and ZC with MC Conditions methods. MC and ZC are abbreviated for monolithic column and Zorbax column, respectively.

PAGE 112

112 Table 4 3. Chromatography met hods used to optimize a HILIC method. Method Column Column Dimensions (mm) Mobile Phase A Mobile Phase B Flow Rate (mL/min) Gradient [A:B(min)] A m F 95:5 Halo Penta HILIC 2.1 x 100 (2.7 m) 95% acetonitrile and 5% water with 5 mM A m F 50% acetonit rile and 50% water with 5 mM A m F 0.5 100 : 0(0.0 3 .0) 10 : 90(20.0 ) 100 : 0(23.0 25 .0) A m F 90:10 Halo Penta HILIC 2.1 x 100 (2.7 m) 90% acetonitrile and 10% water with 5 mM A m F 50% acetonitrile and 50% water with 5 mM A m F 0.5 100 : 0(0.0 3 .0) 10 : 90(20.0 ) 1 00 : 0(23.0 25 .0) A m A 95:5 Halo Penta HILIC 2.1 x 100 (2.7 m) 95% acetonitrile and 5% water with 5 mM A m A 50% acetonitrile and 50% water with 5 mM A m A 0.5 100 : 0(0.0 3 .0) 10 : 90(20.0 ) 100 : 0(23.0 25 .0) A m A 90:10 Halo Penta HILIC 2.1 x 100 (2.7 m) 90% ac etonitrile and 10% water with 5 mM A m A 50% acetonitrile and 50% water with 5 mM A m A 0.5 100 : 0(0.0 3 .0) 10 : 90(20.0 ) 100 : 0(23.0 25 .0) The column is heated to 35 C in all methods. A m F and A m A are abbreviated for ammonium formate and ammonium acetate, resp ectively. All A m F methods were at pH of 3.75 and A m A methods were at pH 5.75.

PAGE 113

113 Table 4 4 Average number of features (n =3) extracted from LC analysis with manatee urine by the monolithic and short monolithic methods at varying capillary/fragmentor volta ges. Capillary V/Fragmentor V Average # of Features Monolithic Method Average # of Features Short Monolithic Method 3000/120 121 129 3000/180 97 99 4000/120 120 138 4000/180 104 105 3600/150 115 129 Table 4 5 Average number of features (n =3) extr acted from LC analysis with piglet plasma by the monolithic and short monolithic methods at varying capillary/fragmentor voltages. Capillary V/Fragmentor V Average # of Features Monolithic Method Average # of Features Short Monolithic Method 3000/120 133 144 4000/120 137 154 4000/180 109 125 3600/150 131 141

PAGE 114

114 Table 4 6 List of retention times, peak widths and column efficiencies for the piglet plasma analysis with the monolithic method (standard deviation = ) The number of significant figures is limited by the standard deviation. m/z Average t r (min) of t r (min) Average w (min) of w (min) N 132.0639 4.072 0.007 0.54 0.02 917.87 205.0910 14.51 0.03 0.44 0.01 17741.34 304.3033 23.220 0.004 0.48 0.04 37028.60 332.3243 24.719 0.004 0.522 0.007 35879.04 360.3658 26.281 0.004 0.54 0.02 37664.18 Table 4 7 List of retention times, peak widths and column efficiencies for the piglet plasma analysis with the short monolithi c method (standard deviation = ) The number of significant figures is limited by the standard deviation. m/z Average t r (min) of t r (min) Average w (min) of w (min) N 132.0639 4.075 0.004 0.55 0.03 880.30 205.0910 11.122 0.003 0.267 0.007 27762.80 304.3033 15 .011 0.004 0.29 0.03 43869.79 332.3243 15.796 0.003 0.34 0.02 35081.44 360.3658 16.635 0.004 0.33 0.02 40741.12

PAGE 115

115 Table 4 8 Average number of features (n =3) extracted from LC analysis with piglet plasma by four LC methods. Table 4 9 List of retention times, peak widths and plate heights for carnitine standard mixture using the MC m ethod Carnitine t r (min) w (min) H (mm) C0 3.701 0.447 0.182 C6 11.659 0.318 0.009 C8 12.589 0.330 0.009 C10 13.378 0.341 0.008 C12 14.202 0.400 0.010 C14 15.132 0.730 0.029 Average # of Features MC Average # of Features MC with ZC Conditions Average # of Features ZC Average # of Features ZC with MC Conditions 135 214 301 294

PAGE 116

116 Table 4 10 List of retention times, peak width s and plate heights for carnitine standard mixture using the MC with ZC Conditions method. Carnitine t r (min) w (min) H (mm) C0 3.831 0.459 0.179 C6 8.870 0.294 0.014 C8 10.259 0.365 0.016 C10 11.413 0.413 0.016 C12 12.543 0.483 0.0 19 C14 13.779 0.765 0.039 Table 4 1 1 List of retention times, peak widths and plate heights for carnitine standard mixture using the ZC method. Carnitine t r (min) w (min) H (mm) C0 0.396 0.153 0.466 C6 1.785 0.836 0.685 C8 5.199 0.3 18 0.012 C10 6.317 0.235 0.004 C12 7.224 0.271 0.004 C14 8.001 0.271 0.004

PAGE 117

117 Table 4 1 2 List of retention times, peak widths and plate heights for carnitine standard mixture using the ZC with MC Conditions method. Carnitine t r (min) w (min) H (mm) C0 0.411 0.212 0.831 C6 3.849 0.789 0.131 C8 8.322 0.188 0.002 C10 9.099 0.235 0.002 C12 9.723 0.235 0.002 C14 10.253 0.294 0.003 Table 4 1 3 List of retention times, peak widths and plate heights for five rand omly chosen m/z The number of significant figures is limited by the standard deviation. m/z Average t r (min) of t r (min) Average w (min) of w (min) H (mm) 132.0639 4.483 0.003 0.95 0.03 0.57 205.0910 10.424 0.005 0.27 0.01 0.01 304.3033 14.835 0.008 0.279 0.007 0.004 332.3243 15.659 0.008 0.37 0.03 0.01 360.3658 16.56 0.01 0.35 0.02 0.01

PAGE 118

118 Table 4 1 4 List of retention times, peak widths and plate heights for five randomly chosen m/z with ZC Conditions The number of significant figures is limited by the standard deviation. m/z Average t r (min) of t r (min) Average w (min) of w (min) H (mm) 132.0639 5.96 0.03 1.36 0.08 0.65 205.0910 7.484 0.004 0.220 0.00 7 0.01 304.3033 13.461 0.002 0.28 0.02 0.01 332.3243 14.571 0.008 0.37 0.02 0.01 360.3658 15.756 0.003 0.349 0.007 0.01 Table 4 1 5 List of retention times, peak widths and plate heights for five randomly chosen m/z method The number of significant figures is limited by the standard deviation. m/z Average t r (min) of t r (min) Average w (min) of w (min) H (mm) 132.0639 0.485 0.007 0.251 0.007 0.84 205.0910 0.606 0.002 0.216 0.007 0.40 304.3033 8.20 0.03 0.287 0.006 0.004 332.3243 8.94 0.03 0.30 0.02 0.004 360.3658 9.64 0.04 0.34 0.01 0.004

PAGE 119

119 Table 4 1 6 List of retention times, peak widths and plate heights for five randomly chosen m/z with MC Conditions The number of significant figures is limited by the standard deviation. m/z Average t r (min) of t r (min) Average w (min) of w (min) H (mm) 132.0639 0.514 0.008 0.35 0.02 1.41 205.0910 1.134 0.003 0.463 0.0 07 0.52 304.3033 10.41 0.01 0.24 0.01 0.002 332.3243 10.924 0.009 0.26 0.02 0.002 360.3658 11.419 0.009 0.239 0.007 0.001 Table 4 1 7 Average number of features (n =3) extracted from piglet plasma analyzed wit h the Zorbax column a nd three different solvent sets Average # of Features H 2 O & ACN, 1% AA Average # of Features H 2 O and ACN, 0.1% AA Average # of Features H 2 O and MeOH, 0.2% AA 278 33 8 455

PAGE 120

120 Table 4 1 8 List of retention times, peak widths and theoretical plates for ca rnitine standard mixture using 1% AA in H2O and ACN. Carnitine t r w N C0 0.424 0.141 144.68 2 C6 3.108 0.906 188.289 C8 5.451 0.247 7792.529 C10 6.499 0.247 11076.923 C12 7.346 0.224 17207.818 C14 8.1 00 0.271 14293.923 Table 4 1 9 List of retention times, peak widths and theoretical plates for carnitine standard mixture using 0. 1% AA in H2O and ACN. Carnitine t r w N C0 0.441 0.153 132.927 C6 3.666 0.789 345.423 C8 5.562 0.259 7378.744 C10 6.668 0.235 12881.767 C12 7.598 0.235 16725.644 C14 8.3 87 0.33 10334.879

PAGE 121

121 Table 4 20 List of retention times, peak widths and theoretical plates for carnitine standard mixture using 0.2 % AA in H2O and MeOH Carnitine t r w N C0 0.458 0.153 143.373 C6 4.731 0.424 1992.022 C8 7.25 0 0.318 8316.522 C10 9.0 05 0.306 13856.213 C12 10.252 0.259 25069.03 7 C14 11.171 0.318 19744.668 Table 4 2 1 List of retention times, peak widths and theoretical plates for five randomly chosen m/z 1% AA in H 2 O and ACN The number of significant figures is limited by the standard deviation. m/z Average t r (min) of t r (min) Average w (min) of w (min) N 132.0639 0.5 5 0.0 2 0.33 0.0 3 43.05 205.091 0 0.68 0.03 0.25 0.06 118.16 304.3033 8.3 1 0.02 0.28 0.01 14224.84 332.3243 9.0 3 0.02 0.306 0 .000 13924 .00 360.365 8 9.73 0.02 0.353 0 .000 12165.32

PAGE 122

122 Table 4 2 2 List of retention times, peak widths and theoretical plates for five randomly chosen m/z 0.1% AA in H 2 O and ACN T he number of sign ificant figure s is limited by the standard deviation. m/z Average t r (min) of t r (min) Average w (min) of w (m in) N 132.0639 0.588 0.006 0.26 0.01 82.37 205.091 0 0.80 0.01 0.267 0.007 144.96 304.3033 8.25 0.02 0.33 0.01 10283.12 332.3243 8.99 0.02 0.48 0.02 5642.98 360.3658 9.76 0.02 0.47 0.01 6991.40 Table 4 2 3 List of retention times, peak widths and th eoretical plates for five randomly chosen m/z 0.2% AA in H 2 O and MeOH The number of significant figures is limited by the standard d eviation. m/z Average t r (min) of t r (min) Average w (min) of w (min) N 132.0639 0.64 0.01 0.381 0.0 3 45.60 205.091 0 1.56 0.0 6 0.683 0.02 83.30 304.3033 10.34 0.02 0.294 0.02 19758.87 332.3243 11.0 9 0.02 0.334 0.006 17658.98 360. 3658 11.7 4 0.0 2 0.298 0.00 7 24817.17

PAGE 123

123 Figure 4 1. Average peak areas of different carnitines as a function of capillary voltage. Error bars indicate the standard deviation. 0.0E+0 2.0E+6 4.0E+6 6.0E+6 8.0E+6 1.0E+7 1.2E+7 1.4E+7 1.6E+7 1.8E+7 2.0E+7 2900 3100 3300 3500 3700 3900 4100 4300 Average Peak Area (n=3) Capillary Voltage (V) Effect of Capillary Voltage on Carnitine Mixture Using +ESI Short Monolithic Method C0 C6 C8 C12 C18

PAGE 124

124 Figure 4 2. Average peak areas of different carnitines as a function of frag mentor voltage. Error bars indicate the standard deviation 0.0E+00 5.0E+06 1.0E+07 1.5E+07 2.0E+07 2.5E+07 3.0E+07 3.5E+07 4.0E+07 4.5E+07 50 70 90 110 130 150 170 190 210 230 250 Average Peak Area, (n = 3) Fragmentor Voltage (V) Effect of Fragmentor Voltage on Carnitine Mixture Using +ESI Zorbax Method C0 C6 C8 C10 C12 C14

PAGE 125

125 Figure 4 3. Base peak chromatograms of standard mixtures of carnitines using four different LC method s From top to bottom: MC method, MC with ZC conditions method, ZC method, ZC with MC condi tions method.

PAGE 126

126 Figure 4 4. Base peak chromatograms of piglet plasma using four different LC methods. From top to bottom: MC method, MC with ZC conditions method, ZC method, ZC with MC conditions method.

PAGE 127

127 Figure 4 5. Base peak chromatograms of standard mixtures of carnitines using three different solvent sets and Zorbax column. From top to bottom: H 2 0 and ACN with 1% AA, H 2 O and ACN with 0.1% AA, H 2 O and MeOH with 0.2% AA.

PAGE 128

128 Figure 4 6. Base peak chromatograms of piglet plasma using three different so lvent sets and Zorbax column. From top to bottom: H 2 0 and ACN with 1% AA, H 2 O and ACN with 0.1% AA, H 2 O and MeOH with 0.2% AA.

PAGE 129

129 Figure 4 7. Average efficiencies (N) of carnitine standards for HILIC with four different solvent systems. AmF and AmA are abbreviations of ammonium formate and ammonium acetate, respectively. Error bars indicate the standard deviation of the mean (n = 3). 0 500 1000 1500 2000 2500 3000 3500 4000 4500 C0 C6 C8 C10 C12 C14 Efficiency (N, unitless) Carnitines Column Efficiency for Carnitines AmF 95:5 AmF 90:10 AmA 95:5 AmA 90:10

PAGE 130

130 Figure 4 8. Average efficiencies (N) of amino acid standards for HILIC with four different solvent systems. AmF and A mA are abbreviations of ammonium formate and ammonium acetate, respectively. Error bars indicate the standard deviation of the mean (n = 3). 0 5000 10000 15000 20000 25000 30000 35000 40000 45000 Efficiency (N, unitless) Amino Acids Column Efficiency for Amino Acids AmF 95:5 AmF 90:10 AmA 95:5 AmA 90:10

PAGE 131

131 Figure 4 9. Average retention times of standards for HILIC with four different solvent systems. AmF and AmA ar e abbreviations of ammonium formate and ammonium acetate, respectively. Error bars indicate the standard deviation of the mean (n = 3). 0 2 4 6 8 10 12 Retention Time (min) Standard Retention of Carnitines and Amino Acids with HILIC AmF 95:5 AmF 90:10 AmA 95:5 AmA 90:10

PAGE 132

132 CHAPTER 5 METABOLOMICS OF ESTR O US AND ANESTR O US URINE FROM MANATE ES Background The West Indian manatees, or Trichechu s manatus reside around Florida and the Caribbean Manatees are marine mammals that typically weigh 800 1,200 pounds and reach a length of about 10 feet. It is common for these large herbivores to eat more than 32 pounds of plants a day, a task that tak es 5 8 hours. 85 87 Manatees are often Manatees are semi social, in that they interact with one another but do not form any permanent bo nds. Th e only exception to the bonding behavior is when a female manatee births a calf. After birth it is typical for the calf to remain with the mother 1 2 years. 88 Since manatees do not form permanent bonds it is common for male manatees to compete with one another in order to mate with a single female. When a female manatee is in estrus male manatees will pursue for as long as 2 4 weeks, creating a mating herd. 89 It is assumed that a signal or pheromone, is given off by the female that allows male manatees to know when she is receptive to mating. Manatees have limit ed vision and lack communication othe r than between mother and calf; because of this the pheromon e is thought to be chemical. 89 90 Pheromones are substances utilized for intra specific communication, or communication between the same species They are emitted outside the body of an in dividual and received by other individuals of that species. The pheromone can contain information about gender, dominance, or reproductive status, as well as other characteristics. The reactions to the pheromone should be species wide and not independent of the individual sending or receiving it. 91 It is hypothesized that the

PAGE 133

133 pheromones allowing male manatees to know when a female manatee is receptive to mating are transmitted through urine. This is a common way f or pheromones to be secreted, along with feces and through glands. 91 However, manatees do not have any known glands, thus urine seems the most obvious pathway. Two compounds known to be associated with the manatee r eproductive status p rogesterone estradiol are commonly monitored to determine where in the reproductive cycle a female manatee is 92 93 The urine samples obtained for this work were determined to be from estrus or anestrus periods by radio immunoassays that measured the above urinary hormones. Additionally, some of the estr o us urine samples were shown to invoke a behavioral response in some male manatees, suggesting that the pheromone i s present i n urine. Elephants are mammals that are closely related to manatees and progesterone has also been monitored to determine where in the reproductive cycle female elephants are. 94 Additionally, it has been observed that mating behavior does not occur at the maximum progesterone level, indicating that progesterone is not the pheromone responsible for showing the readiness of a female to mate. 94 In fact, the sex pheromone that indicat es a female elephant is ready for mating has been identified for elephants. The compound is (Z) 7 dodecen 1 yl acetate. 95 The same compound is the sex pheromone for about 140 different species of moths. 96 The pheromone in manatees may be structurally similar, in which case GC/MS analysis may be necessary. However, little is known about what type of compound the sex pheromone may be, so initial experiments will use LC/MS analyses. In addition potentially identifying a sex

PAGE 134

134 pheromone we will be able to learn more about the profiling of different metabolites during the manatees estrus cycle. T he metabolomics of urine from estrus and anestrus female manatees was perfor med using two different LC columns. LC with MS/MS was performed for the compounds found to be significant between the estr o us and anestr o us groups. In this work, particular attention was paid to those compounds found to be elevated in estr o us urine. The goal of this project was to potentially identify the sex pheromone in female manatee urine, but to also understand the reproductive cycle of manatees. This understanding would allow the possibility of better breeding techniques for captive manatees and p otentially preserve this endangered species. 92 Experimental The experimental conditions are the same as those described in C hapter 2 and Chapter 3 unless otherwise indicated below. Sample Collection and Prepa ration Manatee urine was collected directly from four different females at varying times between May 2003 and August 2010. There were a total of ten estr o us samples and ten anestr o us samples. The urine was stored frozen until the time of analysis and wa s directly injected ( no sample preparation ) Liquid Chromatography Conditions The Zorbax chromatogr aphy method (ZC) introduced in C hapter 3 was implemented in this work (Table 4 2). The solvents used were methanol (A) and water (B), both with 0.2% acetic acid. The column was heated to 60 C. An injection volume of 5 L of urine was used and the samples were analyzed in triplicate for a total of 60 data files.

PAGE 135

135 HILIC chromatography was also performed after noting that all 9 significant compounds found by statistical analysis of the Zorbax c hromatography method were at a retention time of 0.3 0.4 minutes. The HILIC chromatography method used was AmA 95:5, explained in Chapter 4 and implementing the Halo column. The injection volume for the HILIC method was 5 L. Electrospray Source The cond itions used a re the same as those listed in C hapter 3 Mass Spectrometer The mass spectrometer conditions are identical to those of C hapter 2 and Chapter 3 MS/MS Instrumentation The MS/MS i nstrumentation is described in C hapter 3 MS/MS events were creat ed to scan for the significant ions. Data Analysis The LC/MS data analysis process is described in both C hapter 2 and Chapter 3 ( Thermo Scientific, Waltham, MA). Results and Discussion RP/MS T he chromatograms of estr o us and anestr o us urine samples obtained with the Zorbax column are shown in Figures 5 1 and 5 2 In general, the chromatograms for the estr o us samples appear to be more complex in the first couple of minutes tha n the anestr o us chromatograms.

PAGE 136

136 The manatee urine analyzed by LC/MS with the Zorbax method was determined to have 1,472 masses present. T hese masses were subjected to an asymptotic t test (p = 0.05) with a Benjamini Hochberg FDR multiple testing correcti on. The t test resulted in only 9 significant masses between the two groups of estrus and anestrus. This small number is very easy to work with and MS/MS can be performed on all 9 of the masses. The reason for this low of a number of masses may be due t o only subtle changes between estr o us and anestr o us urine. Additionally, if an anestr o us sample is collected close to the beginning of estrus then fewer changes would be expected between the two groups. It is also important to note that all the 9 masses have early retention times, indicati ng a high solubility in water. The 3D PCA plot created with the statistically significant masses is shown in Figure 5 3 The x axis (component 1) accounts for 60.50 % of the variation, while the y and z axes (componen ts 2 and 3) account for 15.73 % and 7.90 %, respectively. Distinct separation is not observed between the two groups but the estr o us samples do appear to group towards the top right area of the plot. A two dimensional plot is also shown in Figure 5 4 The 2D view allows better visualization of the grouping. The estr o us samples group tightly on the left side of the 2D plot. A PLS DA was also performed and the plot can be seen in Figure 5 5 (3D vie w) and Figure 5 6 (2D view). The predic tion model had an ac curacy of 83 33 %, most of the inaccuracies resulted from 9 an estr o us replicates being predicted as estrus (Table 5 1 ). Only one estr o us replicate was predicted to be anestrus. The loadings plot from the PLS DA was not considered in this experiment, as th e low number of significant features allowed further examination of all the masses.

PAGE 137

137 The 9 masses determined to be significant were extracted from each of the three replicates for the estr o us and anestr o us urine samples. The extracted ion chromatograms w ere then manually integrated and the peak areas were averaged for all the estr o us replicates as well as for the anestr o us replicates. A bar gr aph was then created (Figure 5 7 ) comparing the average abundance of the 9 masses in estr o us and anestr o us urine. As can be seen, three masses show an increase from anestr o us to estr o us urine. However, no mass is only present in the estr o us urine, as we would expect if it was a pheromone released only during estrus. The three masses found to be elevated in estr o us urine were the focus of MS/MS experiments. However, the two ions with m/z 143 could not be differentiated with MS/MS due to the lack of high resolution data. The significant ion at m/z 175 resulted in a loss of 32 mass units that could be a loss of a met hoxy group (Figure 5 8). There were two accurate mass library results, neither of which would likely ionize in positive mode or have a loss of 32 Da. Formulas were calculated based on accurate mass and one resulted in a structure that would likely lose a methoxy group. That formula was C 7 H 10 O 3 S. The MS/MS spectrum for m/z 143 is shown in Figure 5 9 a loss of 42 mass units is observed at m/z 101 It is not known if this is mixture of th e two 143 ions found by RP LC /MS or one of them. An accurate mass l ibrary search resulted in two compound for m/z 143.0242 but none for m/z 143.0197. However the two compounds in the library did not have matching fragment ions and one would only ionize in negative mode. The formula for the compound that would ionize in positive mode is C 5 H 6 N 2 OS All data base searching was performed for [M+H] + ions with a mass tolerance of 30 ppm.

PAGE 138

138 H ILIC/MS The results of the HILIC chromatography can be seen in Figure 5 10 for estr o us urine samples and in Figure 5 9 for anestr o us sample s. Again, the estr o us samples seem to generate a more complicated chromatogram than the anestr o us samples. It is obvious that more compounds are being retained on the HILIC column than were with the Zorbax column if Figures 5 1 and 5 2 are compared to Fi gures 5 10 and 5 11 When the HILIC/MS results for manatee urine were examined with MPP, more than two times the number of aligned compounds were found for HILIC than were for RP/MS with the Zorbax column. The manatee urine analyzed by HILIC/MS had 3,209 masses present, versus the 1, 472 masses found aligned with the Zorbax method. This is not surprising considering polar nature of urine; better separation should be achieved with HILIC than RP chromatography and thus allow for better statistical analyses. The 3,209 aligned compounds were subjected to an asymptotic t test (p = 0.05) with a Benjamini Hochberg FDR multiple testing correction. The t test resulted in 30 significant masses between the two groups of estrus and anestrus. Again, this is much high er, greater than 3 times the number of significant masses found using RPLC. Although there are more significant masses found with HILIC than with RPLC 30 masses still allow s for relatively easy MS/MS analyses. The 3D PCA plot, created with the 30 statist ically significant masses, is shown in Figure 5 1 2 The x axis (component 1) accounts for 44.70% of the variation, while the y and z axes (components 2 and 3) account for 11.93% and 9.59%, respectively. Fair separation is observed between the two groups where the estr o us samples appear to group towards the back right area of the plot. A two dimensional plot is also shown in

PAGE 139

139 Figure 5 1 3 The 2D view allows better visualization of the grouping. The estr o us samples group on the left side of the 2D plot. A PLS DA was also performed and the plot can be seen in Figure 5 1 4 (3D view) and Figure 5 1 5 (2D view). The prediction model had an accuracy of 98.33%, the inaccuracy resulted from one anestr o us replicate being predicted as estrus (Table 5 2). No estr o us replicates were predicted to be anestrus. The loadings plot from the PLS DA was not considered in this experiment, as the low number of significant features allowed further examination of all the masses. The masses determined to be significant and also elevated (10 masses) in the estr o us samples were extracted from each of the three replicates for the estr o us and anestr o us urine samples. The extracted ion chromatograms were then manually integrated and the peak areas were averaged for all the estr o us r eplicates as well as for the anestr o us replicates. A bar graph was then created (Figure 5 1 6 ) comparin g the average abundance of the 10 masses in estr o us and anestr o us urine. As can be seen when manually extracted one m/z is actually not elevated in the estr o us urine despite being identified as up regulated in the estr o us urine with MPP Similar to the RP LC /MS results no mass is only present in the estr o us urine, as we would expect if it was a pheromone released only during estrus. The 10 masses found to be elevated in estr o us urine were t he focus of MS/MS experiments. One of the accurate masses matched that of creatine and the fragmentation pattern observed in the urine samples matched the fragmentation pattern of a standard of creatine (Figure 5 17). In the MS/MS spectrum the ion at m/z 90 is due to the loss of the C N 2 H 2 group. The retention time of creatine with HILIC/MS also

PAGE 140

140 matched the ion of interest at m/z 132. Thus the significant mass at m/z 132 was determined to be creatine. Another signi ficant mass, at m/z 148, had a result ba sed on accurate mass as glutamic acid and also had fragmentation that matched that observed for a standard of glutamic acid In Figure 5 18, the transition of m/z 148 to m/z 130 and 102 is due to the loss of water a nd the loss of a carboxylic acid, respectively. The MS/MS spectrum for the significant mass at m/z 176 is shown in Figure 5 19. The loss of 44 mass units is observed at m/z 132 and indicates the loss of a CN 2 H 4 group. One accurate mass library result wa s argin in ic acid, and although no MS/MS data was available the structure of argininic acid support the MS/MS fragmentation observed Finally, the significant mass at m/z 244 resulted in a large fragment ion at m/z 112 (Figure 5 20). One of the results of the accurate mass library search was cytidine. The fragmentation observed in the library matched that observed for the ion at m /z 112 as well as a synthetic standard indicating a loss of the oxygen containing ring. A summary of the identified metabolite s found by both RP LC /MS and HILIC/MS can be seen in Table 5 3. Compounds not included on this table resulted in: no hits from metabolite libraries, no conclusive calculated formulas based on accurate mass, MS/MS fragments that did not match for any hits f or potential metabolites, and/or were not detected by the LIT. Summary Other than the possible identification of C 7 H 10 O 3 S as m/z 175 found by RP/MS, most of the other identified metabolites are responsible for energy supply or waste removal. The exceptio n is cytidine, which is a nucleoside. It is likely that none of the identified compounds are the pheromone of interest. The fact that none of the significant ions were exclusive to estrous urine supports that claim. Additionally, m/z

PAGE 141

141 values of known sex pheromones, including the elephant pheromone, were extracted from the data files but yielded no significant peaks. Those urine samples that did invoke a response in male manatees were also compared to those that did not, but no additional potential phero mones could be found. This work is still novel, as no one has explored the metab olites of manatee urine. It is also not common that urine is directly injected, but this work has shown differences can still be seen between estr o us and anestr o us urine with out sample preparation. It has been shown that HILIC is a better chromatographic technique for the separation of urine but different ionization modes/sources can still be explo red, such as ESI and +/ APCI Any potential pheromone would also need to be t ested for behavior responses within male manatees. Finally, this work may help our understanding of the manatee estrus cycle, so that reproduction can be assisted to better maintain this endangered species. 92

PAGE 142

142 Table 5 1. PLS DA prediction results for urine from manatees in estrus and anestrus by RPLC/MS Predictions Estrus Anestrus Accuracy (%) 29 1 96.67 9 21 70 00 83 33 T here are 30 files for each group, thus each should have a prediction of 30 those that vary cause the accuracy of the model to decrease Table 5 2 PLS DA prediction results for urine from manatees in estrus and anestrus by HILIC/MS. Predictions Estrus Anestrus Accuracy (%) 30 0 100.0 1 29 96.67 98.33 T here are 30 files for ea ch group, thus each should have a prediction of 30 those that vary cause the accuracy of the model to decrease Table 5 3 Overview of identified significant masses elevated in estr o us urine by RP LC /MS and HILIC/MS. Metabolite Formula Experimental m/z T heoretical m/z Error (ppm) Retention (min) Unknown C 7 H 10 O 3 S 175.0441 175.0423 10.3 0.3 RP LC Creatine C 4 H 9 N 3 O 2 132.0760 132.0768 6.1 8.1 HILIC Glutamic acid C 5 H 9 NO 4 148.0596 148.0604 5.4 10.0 HILIC Argin in ic Acid C 6 H 13 N 3 O 3 176.1021 176.1030 5.1 9.5 HILC Cytidine C 9 H 13 N 3 O 5 244.0917 244.0928 4.5 2.8 HILIC Metabolites are noted with RP LC or HILIC under retention to indicate if they were identified as significant with RP LC /MS or HILIC/MS. An asterisk indicates verification with a synthetic stan dard.

PAGE 143

143 Figure 5 1 Overlay of 30 c hromatograms of manatee urine from animals in estrus using RPLC Figure 5 2 Overlay of 30 c hromatograms of manatee urine from animals in anestrus using RPLC

PAGE 144

144 Figure 5 3 3D PCA plot of the significant masses in urine from manatees in estrus and anestrus (9 masses, p = 0.05) by RP LC /MS Each point represents one replicate. Figure 5 4 2D PCA plot of the significant masses in urine from manatees in estrus and anestrus (9 masses, p = 0.05) by RP LC /MS E ach poin t represents one replicate. The color code is the same as Figure 5 5.

PAGE 145

145 Figure 5 5 3D P LS DA plot of t test ( p = 0.05) masses in urine from manatees in estrus and anestrus (9 masses) by RP LC /MS Each point on the plot depicts one replicate. Figure 5 6 2 PLS DA plot of t test ( p = 0.0 5 ) masses in manatees in estrus and anestrus (9 masses) by RP LC /MS Each point on the plot depicts one replicate. The x axis (Identifier) values are the individual sample replicates. The co lor code is the same as Figu re 5 7

PAGE 146

146 Figure 5 7 Bar graph comparing the average peak areas of the 9 significant masses in estr o us and anest r o us urine by RP LC /MS A possible formula was identified in Table 5 3 for m/z 175.0441. Error bars indicate the standard deviation of the mea n (n = 30). Figure 5 8. MS/MS spectrum of m/z 175 in estr o us urine, determined by RP LC /MS The ion at m/z 143 (blue arrow) is likely due to the loss of a methoxy group if the proposed structure is correct. The proposed structure is based on accurate m ass. 0.00E+00 5.00E+05 1.00E+06 1.50E+06 2.00E+06 2.50E+06 Average Peak Area m/z Significant Masses by RPLC/MS Estrus Anestrus

PAGE 147

147 Figure 5 9. MS/MS spectrum of m/z 143 in estr o us urine, determined by RP LC /MS A loss of 42 mass units is observed at m/z 101. Figure 5 10 Overlay of 30 c hromatograms of manatee urine from animals in estrus using HILIC.

PAGE 148

148 Figure 5 11 Overlay of 30 c hromatograms of manatee urine from animals in anestrus using HILIC Figure 5 1 2 3D PCA plot of the significant masses in urine from manatees in estrus and anestrus by HILIC/MS (30 masses, p = 0.05). Each point represents one replicate.

PAGE 149

149 Figur e 5 1 3 2D PCA plot of the significant masses in urine from manatees in estrus and anestrus (30 masses, p = 0.05) by HILIC/MS. E ach point represents one replicate. The color code is the same as Figure 5 10. Figure 5 1 4 3D P LS DA plot of t test ( p = 0.05) masses in urine from manatees in estrus and anestrus (30 masses) by HILIC/MS. Each point on the plot depicts one replicate.

PAGE 150

150 Figure 5 1 5 2 PLS DA plot of t test ( p = 0.0 5 ) masses in manatees in estrus and anestrus (30 masses) by HILIC/MS Each po int on the plot depicts one replicate. The x axis (Identifier) values are individual sample replicates. The co lor code is the same as Figure 5 12 Figure 5 1 6 Bar graph comparing the average peak areas of the 10 significant masses elevated in estr o us urine by HILIC/MS. Identifications were made for m/z values 132.0760, 148.0596, 176.1021, and 244.0917 in Table 5 3. Error bars indicate the standard deviation of the mean (n = 30). 0.0E+00 5.0E+05 1.0E+06 1.5E+06 2.0E+06 2.5E+06 3.0E+06 3.5E+06 4.0E+06 4.5E+06 Average Peak Area m/z Significant Masses Elevated in Estrous Urine Estrus Anestrus

PAGE 151

151 Figure 5 1 7 MS/MS spectra of m/z 132, determined by HILIC/MS, in est r o us urine (top) and a synthetic standard (bottom) The ion at m/z 90 ( loss of 42 mass units ) indicates the loss of the CN 2 H 2 group.

PAGE 152

152 Figure 5 1 8 MS/MS spectra of m/z 148 in estr o us urine (top) and a synthetic standard (bottom) determined by HILIC/MS The ion at m/z 130 (blue arrow) is due to the loss of water and the ion at m/z 102 (loss of 46 mass units, orange arrow) is from the loss of a carboxylic acid. Figure 5 1 9 MS/MS spectrum of m/z 176 in estr o us urine, determined by HILIC/MS The ion a t m/z 13 2 (blue arrow) is due to the loss of the CN 2 H 4 group

PAGE 153

153 Figure 5 20 MS/MS spectra of m/z 244 in estr o us urine (top) and a synthetic standard (bottom) determined by HILIC/MS The ion at m/z 112 (blue arrow) is due to the loss of the oxygen con taining ring

PAGE 154

154 CHAPTER 6 CONCLUSIONS AND FUTU RE DIRECTIONS Summary The field of metabolomics is continually growing. However, bottlenecks in metabolomics still exist, which include data processing and metabolite identification. Robust software is necessa ry to process the complex data files generated from the analyses of biological samples in such large numbers. As shown in this work, identification of metabolites is very difficult. T he expansion of metabolite and MS/MS libraries is necessary for easier identification using accurate mass and fragmentation patterns. Piglet Plasma and RBC Comparison This work has shown that the metabolomics of piglet plasma leads to more statistically significant changes over the first seven days of life compared to piglet RBCs. Piglet plasma led to 82 statistically significant masses while piglet RBCs only resulted in 23 statistically significant masses ( p = 0.01). More c arnitines were found to be present in plasma than RBC s A general increase in carnitines was observed in plasma samples over seven days, whereas no trend was observed in the RBCs. Future work could include th is experiment being rerun with an equal number of samples and three replicates per sample. The result would be more beneficial to the metabolomics community, as (to the best of my knowledge) no one has examined the metabolomics of plasma compared to RBCs in any organism. I f nothing else it may conclude that little useful information can be determined from RBC samples compared to plasma. Additional ly, work can be done to identify (MS/MS analyses) the significant masses determined by ANOVA in both the piglet plasma and RBCs.

PAGE 155

155 Metabolomics of Colon Cancer This work has compared the plasma of healthy human controls to plasma from patients with colon can cer. The potential biomarkers determined by statistics of the LC/TOF data were then analyze d by LC/LIT instrumentation T he MS/MS capabilities of the LIT along with the accurate mass from the TOF allowed fo r structural elucidation and identification of five of the potential biomarkers as carnitines No other studies have indicated these carnitines as potential biomarkers for colon cancer. More work can be done to determine if these potential biomarkers are biologically relev ant to the disease of colon c ancer. Also, the identification of unkno wn significant masses can be investigated further. Method Optimization for Metabolomic Applications Methods for positive mode ESI were optimized in order to achieve the largest number of features. LC run time was d ecreased for a more time efficient method without injuring the resulting number of features. In addition to the LC run time being optimized, the results from this chapter were used to select the best LC column for further applications. A HILIC method was also optimized. Future work may include the optimization of the APCI source. This would allow both ESI and APCI analysis, generating a larger number of potential biomarkers. Also, negative mode ESI and APCI can be implemented and optimized. Metabolomics of Estr o us and Anestr o us Manatee Urine Since many metabolites in urine are very polar the RP LC /MS method resulted in all of the significant masses being at a retention time of about 0.3 minutes, or the void volu me. This implies that the chromatogram was very congested in the first few minutes of analysis, leading to less sensitive MS detection due to ion suppression, and

PAGE 156

156 thus making statistical analyse s more difficult. This was shown by the fact that only 9 metabolites were found to be significant by R P LC /MS whereas the HILIC/MS method led to 30 significant compounds. The HILIC/MS method allowed better separation but poorer peak shapes than the RP LC /MS method. Cytidine, creatine, and glutamic acid were identified as significant compounds elevated in es trous urine. Two putative identifications of significant metabolites elevated in estrous urine were made with accurate mass and MS/MS, but without verification with a standard, as argininic acid and an unknown metabolite (formula calculation). However, n one of these metabolites appear to be the pheromone of interest. The pheromone was expected to be present in only estr o us urine, or at least very elevated. None of the elevated metabolites in estr o us urine fit this description well. Although a pheromone was not found, no one has examined the met abolomics of manatee urine, thus any data collected is valuable and may help us better understand this endangered species. Future work that would benefit this project includes introducing the male manatees to the elephant pheromone (not observed with +ESI) and noting the behavior, as well as acquisition of LC/MS by either negative mode ESI or by APCI. Additionally, targeted studies could be performed by gas chromatography/MS to see if the elephant pheromone is pre sent in the manatee urine, although this may require special sample preparation.

PAGE 157

157 LIST OF REFERENCES (1) Atkinson, A. J.; Colburn, W. A.; DeGruttola, V. G.; DeMets, D. L.; Downing, G. J.; Hoth, D. F.; Oates, J. A.; Peck, C. C.; Schooley, R. T.; Spilker, B. A.; Woodcock, J.; Zeger, S. L. Clin. Pharmacol. Ther. 2001 69 89 95. (2) Nass, S. J.; Moses, H. L. Cancer Biomarkers: The Promises and Challenges of Improving Detection and Treatment ; The National Academies Press: Washington, D.C., 2 007 (3) Kaddurah Daouk, R.; Kristal, B. S.; Weinshilboum, R. M. Annu. Rev. Pharmacol. Toxicol. 2008 48 653 683. (4) Dettmer, K.; Aronov, P. A.; Hammock, B. D. Mass Spectrom. Rev. 2007 26 51 78. (5) Madsen, R.; Lundstedt, T.; Trygg, J. Anal. Chim Acta. 2010 659 23 33. (6) Fiehn, O. Plant Mol. Biol. 2002 48 155 171. (7) Wikoff, W. R.; Gangoiti, J. A.; Barshop, B. A.; Siuzdak, G. Clin. Chem. 2007 53 2169 2176. (8) Robertson, D. G. Toxicol. Sci. 2005 85 809 822. (9) Nordstrom, A.; Wa nt, E.; Northen, T.; Lehtio, J.; Siuzdak, G. Anal. Chem. 2008 80 421 429. (10) Lin, Y.; Schiavo, S.; Orjala, J.; Vouros, P.; Kautz, R. Anal. Chem. 2008 80 8045 8054. (11) Beckonert, O.; Keun, H. C.; Ebbels, T. M.; Bundy, J.; Holmes, E.; Lindon, J. C.; Nicholson, J. K. Nat. Protoc. 2007 2 2692 2703. (12) Viant, M. R. Biochem. Biophys. Res. Commun. 2003 310 943 948. (13) Lewis, G. D.; Wei, R.; Liu, E.; Yang, E.; Shi, X.; Martinovic, M.; Farrell, L.; Asnani, A.; Cyrille, M.; Ramanathan, A.; Shaham, O.; Berriz, G.; Lowry, P. A.; Palacios, I. F.; Tasan, M.; Roth, F. P.; Min, J.; Baumgartner, C.; Keshishian, H.; Addona, T.; Mootha, V. K.; Rosenzweig, A.; Carr, S. A.; Fifer, M. A.; Sabatine, M. S.; Gerszten, R. E. J. Clin. Invest. 2008 118 3503 3512. (14) Zeng, M.; Liang, Y.; Li, H.; Wang, M.; Wang, B.; Chen, X.; Zhou, N.; Cao, D.; Wu, J. J. Pharm. Biomed. Anal. 2010 52 265 272. (15) Pereira, H.; Martin, J.; Jo ly, C.; Sbdio, J.; Pujos Guillot, E. Metabolomics 2010 6 207 218. (16) Gibney, M. J.; Walsh, M.; Brennan, L.; Roche, H. M.; German, B.; van Ommen, B. Am. J. Clin. Nutr. 2005 82 497 503.

PAGE 158

158 (17) van Haandel, L.; Becker, M. L.; Leeder, J. S.; Williams T. D.; Stobaugh, J. F. Rapid Commun. Mass Spectrom. 2009 23 3693 3702. (18) Bruce, S. J.; Tavazzi, I.; Parisod, V.; Rezzi, S.; Kochhar, S.; Guy, P. A. Anal. Chem. 2009 81 3285 3296. (19) Hewavitharana, A. K.; Hyde, C.; Thomas, R.; Shaw, P. N. J. Chromatogr. B. Analyt Technol. Biomed. Life. Sci. 2006 834 93 97. (20) Villas Bas, S. G., Roessner, U., Hansen, M., Smedsgaard, J. and Nielsen, J. Metabolome Analysis : An Introduction ; Wiley Interscience: Hoboken, N.J, 2007. (21) Wedge, D. C.; All wood, J. W.; Dunn, W.; Vaughan, A. A.; Simpson, K.; Brown, M.; Priest, L.; Blackhall, F. H.; Whetton, A. D.; Dive, C.; Goodacre, R. Anal. Chem. 2011 83 6689 6697. (22) Harris, D. C.; Freeman, W. H., Ed.; Quantitative Chemical Analysis, 6th ed.; Palgrav e Macmillan: New York, 2002. (23) Snyder, L. High Performance Gradient Elution : The Practical Application of the Linear Solvent Strength Model; John Wiley: Hoboken, NJ, 2007. (24) Franois, I.; Sandra, K.; Sandra, P. Anal. Chim. Acta. 2009 641 14 31. (25) Horvath, C. G.; Lipsky, S. R. Anal. Chem. 1967 39 1893 1893. (26) Lu, X.; Zhao, X.; Bai, C.; Zhao, C.; Lu, G.; Xu, G. J. Chromatogr. B. Analyt Technol. Biomed. Life. Sci. 2008 866 64 76. (27) Svec, F.; Huber, C. G. Anal. Chem. 2006 78 2100 2107. (28) Cabrera, K.; Lubda, D.; Eggenweiler, H.; Minakuchi, H.; Nakanishi, K. Journal of High Resolution Chromatography. 2000 23 93 99. (29) Phenomenex. http:/ /www.phenomenex.com/cms400min/product.aspx?id=119 (accessed September 2, 2010). (30) Whitehouse, C. M.; Dreyer, R. N.; Yamashita, M.; Fenn, J. B. Anal. Chem. 1985 57 675 679. (31) Gaskell, S. J. Journal of Mass Spectrometry. 1997 32 677 688. (32) Cotter, R. E. Time of Flight Mass Spectrometry ; American Chemical Society: Washington, DC; 1994. (33) Agilent Technologies. Publication Number 5990 9207EN 2011. (34) Agilent Technologies. Publication Number G3335 90028 2003.

PAGE 159

159 (35) Blekherman, G.; Laubenbacher, R.; Cortes, D. F.; Mendes, P.; Torti, F. M.; Akman, S.; Torti, S. V.; Shulaev, V. Metabolomics. 2011 7 329 343. ( 36 ) Wikoff, W. R.; Anfora, A. T.; Liu, J.; Schultz, P. G.; Lesley, S. A.; Peters, E. C.; Siuzdak, G. Proc. Natl. Acad. Sci. U .S.A. 2009 106 3698 3703. ( 37 ) Dervilly Pinel, G.; Weigel, S.; Lommen, A.; Chereau, S.; Rambaud, L.; Essers, M.; Antignac, J.; Nielen, M. W. F.; Le Bizec, B. Anal. Chim. Acta. 2011 700 144 154. ( 38 ) Dai, Y.; Li, Z.; Xue, L.; Dou, C.; Zhou, Y.; Zhan g, L.; Qin, X. J. Ethnopharmacol. 2010 128 482 489. ( 39 ) Smith, C. A.; Want, E. J.; O'Maille, G.; Abagyan, R.; Siuzdak, G. Anal. Chem. 2006 78 779 787. ( 40 ) Katajamaa, M.; Oresic, M. J. Chromatogr. A. 2007 1158 318 328. ( 41 ) Bremer, M. ; Doerge, R. W. Statistics at the Bench : A Step by Step Handbook for Biologists ; Cold Spring Harbor Laboratory Press: Cold Spring Harbor, N.Y.; 2009. ( 42 ) Smith, L. A. www.cs.otago.ac.nz/cosc453/student_tutorials/principal_components.pdf (accessed on October 14, 2011). ( 43 ) Fonville, J. M.; Richards, S. E.; Barton, R. H.; Boulange, C. L.; Ebbels, T. M. D.; Nicholson, J. K.; Holmes, E.; Dumas, M. J. Chemomet rics. 2010 24 636 649. ( 44 ) Miller, J. ; Miller, J. Statistic and Chemometrics for Analytical Chemistry 5th ed.; Prentice Hall: New York; 2005. ( 45 ) Verhoeven, K. J. F.; Simonsen, K. L.; McIntyre, L. M. Oikos. 2005 108 643 647. ( 46 ) Benjamini, Y. ; Hochberg, Y. Journal of the Royal Statistical Society.Series B (Methodological). 1995 57 pp. 289 300. ( 47 ) Hall, M. P.; Band, P. A.; Meislin, R. J.; Jazrawi, L. M.; Cardone, D. A. J. Am. Acad. Orthop. Surg. 2009 17 602 608. ( 48 ) Ardrey, R. E. Liq uid C hromatography Mass Spectrometry : An Introduction ; J. Wiley: West Sussex, England ; Hoboken, NJ; 2003. ( 49 ) Agilent Technologies. Publication Number G3335 90028 2003. (50) Fjeldsted, J. Agilent Technologies. Publication Number 5989 0373EN 2003 ( 51 ) Schnackenberg, L.; Sun, J.; Espandiari, P.; Holland, R.; Hanig, J.; Beger, R. BMC Bioinformatics. 2007 8 S3.

PAGE 160

160 ( 52 ) Mersmann, H. J. ; Pond, W. G. Biology of the Domestic Pig ; Cornell University Press: Ithaca, N.Y. 2001. ( 53 ) Fellman, V.; Lemmel S.; Sajantila, A.; Pihko, H.; Jrvel, I. Journal of Human Genetics. 2008 53 554 558. ( 54 ) Chace, D. H.; Hillman, S. L.; Van Hove, J. L.; Naylor, E. W. Clin. Chem. 1997 43 2106 2113. ( 55 ) Metz, T. O. Metabolic Profiling : Methods and Protocols ; S pringer: New York; 2011. ( 56 ) Cerutti, E. S. LC/MS Based Targeted and Global Metabolomic Methodologies and their Application to Biomarker Discovery ; University of Florida: Gai nesville, FL ; 2009. ( 57 ) The Scripps Research Institute. http://metlin.scripps.edu/metabo_batch.php (accessed on May, 8, 2011). ( 58 ) Smith, C. A.; Maille, G. O.; Want, E. J.; Qin, C.; Trauger, S. A.; Brandon, T. R.; Custodio, D. E.; Abagyan, R.; Siuzdak, G. Ther. Drug Monit. 2005 27 ( 59 ) American Cancer Society. http://www.cancer.org/Research/CancerFactsFigures/CancerFactsFigu res/most requested tables figures 2012 (accessed on January 11, 20 12 ). ( 60 ) Kerr, D., Young, A. and Hobbs, R. ABC of Colorectal Cancer .; BMJ Books: London; 2001. ( 61 ) Lefere, P. ; Gryspeerdt, S. Virtual Colonoscopy: A Practical Guide, 2nd ed; Springer Verlag: Berlin; 2010. ( 62 ) Ame rican Joint Committee o n Cancer; Edge, S. B. AJCC Cancer Staging Manual, 7th ed. ; Springer: New York; 2010. ( 63 ) American Cancer Society. http://www.cancer.org/Cancer/ColonandRectumCancer/DetailedGuide/colorectal cancer staged (accessed on January 17, 2012) ( 64 ) Journal of the National Cancer Institute. 2004 96 1420 1425. ( 65 ) National Cancer Institute. http://www.cancer.gov/cancertopics/pdq/treatment/colon/Patient/page2 (accessed on January 13, 2012). ( 66 ) Na mbiar, P. R.; Gupta, R. R.; Misra, V. Mutation Research/Fundamental and Molecular Mechanisms of Mutagenesis. 2010 693 3 18.

PAGE 161

161 ( 67 ) Chan, E. C.; Koh, P. K.; Mal, M.; Cheah, P. Y.; Eu, K. W.; Backshall, A.; Cavill, R.; Nicholson, J. K.; Keun, H. C. J. Prot eome Res. 2009 8 352 361. ( 68 ) Mal, M.; Koh, P. K.; Cheah, P. Y.; Chan, E. C. Y. Rapid Communications in Mass Spectrometry. 2009 23 487 494. ( 69 ) W arburg O. Science. 1956 123 309 314. ( 70 ) Hirayama, A.; Kami, K.; Sugimoto, M.; Sugawara, M.; To ki, N.; Onozuka, H.; Kinoshita, T.; Saito, N.; Ochiai, A.; Tomita, M.; Esumi, H.; Soga, T. Cancer Res. 2009 69 4918 4925. ( 71 ) Denkert, C.; Budczies, J.; Weichert, W.; Wohlgemuth, G.; Scholz, M.; Kind, T.; Niesporek, S.; Noske, A.; Buckendahl, A.; Diet el, M.; Fiehn, O. Mol. Cancer. 2008 7 72. ( 72 ) Piotto, M.; Moussallieh, F. .; Dillmann, B.; Imperiale, A.; Neuville, A.; Brigand, C.; Bellocq, J. .; Elbayed, K.; Namer, I. Metabolomics. 2009 5 292 301. ( 73 ) Qiu, Y.; Cai, G.; Su, M.; Chen, T.; Zhe ng, X.; Xu, Y.; Ni, Y.; Zhao, A.; Xu, L. X.; Cai, S.; Jia, W. J. Proteome Res. 2009 8 4844 4850. ( 74 ) Pekar Second, T.; Blethrow, J. D.; Schwartz, J. C.; Merrihew, G. E.; MacCoss, M. J.; Swaney, D. L.; Russell, J. D.; Coon, J. J.; Zabrouskov, V. Anal. Chem. 2009 81 7757 7765. ( 75 ) Lehtimki, K. K.; Valonen, P. K.; Griffin, J. L.; Visnen, T. H.; Grhn, O. H. J.; Kettunen, M. I.; Vepslinen, J.; Yl Herttuala, S.; Nicholson, J.; Kauppinen, R. A. Journal of Biological Chemistry. 2003 278 45915 459 23. ( 76 ) McGarry, J. D.; Brown, N. F. European Journal of Biochemistry. 1997 244 1 14. ( 77 ) Wright, M. H.; Heal, W. P.; Mann, D. J.; Tate, E. W. J. Chem. Biol. 2009 3 19 35. ( 78 ) Patti, G. J.; Yanes, O.; Siuzdak, G. Nat. Rev. Mol. Cell Biol. 2012 13 263 269. ( 79 ) Sana, T. R.; Waddell, K.; Fischer, S. M. Journal of Chromatography B. 2008 871 314 321. ( 80 ) Crews, B.; Wikoff, W. R.; Patti, G. J.; Woo, H.; Kalisiak, E.; Heideker, J.; Siuzdak, G. Anal. Chem. 2009 81 8538 8544. ( 81 ) Nordstro m, A.; O'Maille, G.; Qin, C.; Siuzdak, G. Anal. Chem. 2006 78 3289 3295. ( 82 ) Spagou, K.; Wilson, I. D.; Masson, P.; Theodoridis, G.; Raikos, N.; Coen, M.; Holmes, E.; Lindon, J. C.; Plumb, R. S.; Nicholson, J. K.; Want, E. J. Anal. Chem. 2011 83 382 390.

PAGE 162

162 ( 83 ) Gika, H. G.; Theodoridis, G. A.; Wilson, I. D. J. Sep. Sci. 2008 31 1598 1608. ( 84 ) Cubbon, S.; Bradbury, T.; Wilson, J.; Thomas Oates, J. Anal. Chem. 2007 79 8911 8918. ( 85 ) Perrin, W. Encyclopedia of Marine Mammals ; Elsevier Inc: To ronto, ON; 2009. ( 86 ) B est R. C. Mamm. Rev. 1981 11 3 29. ( 87 ) Horikoshi Beckett, C.; Schulte, B. A. Zoo Biol. 2006 25 285 301. ( 88 ) Young Harper, J.; Schulte, B. A. Zoo Biol. 2005 24 135 144. ( 89 ) Hartman, D. S. Ecology and Behavior of the Manatee (Trichechus Manatus) in Florida ; Am erican Society of Mammalogists: Pittsburgh, PA; 1979. ( 90 ) Mann, D.; Colbert, D.; Gaspard, J.; Casper, B.; Cook, M.; Reep, R.; Bauer, G. Journal of Comparative Physiology A: Neuroethology, Sensory, Neural, and Behavioral Physiology. 2005 191 903 908. ( 91 ) D ehnhard M. International Zoo Yearbook. 2011 45 55 79. ( 92 ) Francis Floyd, R.; White, J. R.; Chen, C. L.; Cardeilhac, P. T.; Cichra, C. E. J. Aquat. Anim. Health. 1991 3 70 73. ( 93 ) Tripp, K. M.; V erstegen, J. P.; Deutsch, C. J.; Bonde, R. K.; Rodriguez, M.; Morales, B.; Schmitt, D. L.; Harr, K. E. Theriogenology. 2008 70 1030 1040. ( 94 ) Carden, M.; Schmitt, D.; Tomasi, T.; Bradford, J.; Moll, D.; Brown, J. Anim. Reprod. Sci. 1998 53 133 142. ( 95 ) Rasmussen, L. E. L.; Lee, T. D.; Zhang, A.; Roelofs, W. L.; Daves, G. D. Chemical Senses. 1997 22 417 437. ( 96 ) Wyatt, T. D. Nature. 2009 457 262 263.

PAGE 163

163 BIOGRAPHICAL SKETCH Noelle Marie Elliott was born in Pottstown, Pennsylvania. While at tending undergraduate school at Muhlenberg College in Allentown, Pennsylvania, she performed research under Dr. Christine Ingersoll. Noelle also participated in a Research Experience for Undergraduates (REU) program during the summer of 2007 at the Univer sity of Michigan in Ann Arbor, Michigan. At the University of Michigan, she worked on data analysis and extractions for metabolomic applications with Dr. Robert Kennedy. In May of 2008 Noelle graduated from Muhlenbe rg College with a Bachelor of Science i n chemistry and a minor in mathematics. Noelle entered graduate school at the University of Florida (UF) in the fall of 2008, shortly after she married Nathan Elliott. At UF, Noelle performed research in the field of mass spectrometry, with applications in metabolomics, under the direction of Dr. David Powell and Dr. Richard Yost. In the summer of 2012, Noelle will graduate with a Doctor of Philosophy in chemistry.