<%BANNER%>

Application of the Microarray Technology to Study the Genomics of the Ovine Fetal Brain During the Last Stage of Gestation

Permanent Link: http://ufdc.ufl.edu/UFE0045039/00001

Material Information

Title: Application of the Microarray Technology to Study the Genomics of the Ovine Fetal Brain During the Last Stage of Gestation
Physical Description: 1 online resource (188 p.)
Language: english
Creator: Rabaglino, Maria Belen
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2012

Subjects

Subjects / Keywords: brain -- fetus -- genomics -- microarray -- parturition -- sheep
Animal Molecular and Cellular Biology -- Dissertations, Academic -- UF
Genre: Animal Molecular and Cellular Biology thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: The main purpose of the series of studiespresented in this dissertation was to apply microarray technology to measuregene expression in the late gestation fetal brain in different types ofexperiments. Chapters 3 and 4 describe in vivoexperiments to evaluate the effect of estradiol-3-sulfate (predominantcirculating form in fetal plasma) or estradiol infusion, respectively, on thegenomics of the ovine fetal hypothalamus. The main result of Chapter 3 was thatestradiol-3-sulfate strongly induced genes related to feeding behavior whilethe unexpected result of Chapter 4 was to reveal that ICI 182 780, a putativeestrogen-receptor blocker, stimulates estrogen agonist genomic responses. Somebiological processes were similarly induced by estradiol-3-sulfate orestradiol/ICI 182 780 while others were only caused by one or other treatment.Therefore, the genomic actions of estradiol-3-sulfate and estradiol in theovine fetal hypothalamus can be independent of each other. Chapters5 and 6 describe, respectively, time course experiments with a single conditionor comparison of two conditions. In Chapter 5, the objective was to identifygroups of co-expressed genes that follow an increasing or decreasing expressionpattern toward the end of pregnancy in each region of the ovine fetal brain. Genes in which expression increased over time were related withneuronal differentiation and gliogenesis while genes displaying a decrease inexpression over time were related to cell cycle, as described previously. Anovel finding is that genes related with hematopoiesis progressively increasedin expression toward the end of gestation. In Chapter 6, the objective was to identifygenes highly expressed temporally in hypothalamus compared to pituitary duringthe very last period of gestation. As expected, genes related with a functionalfetal hypothalamus or pituitary were highly expressed in one region but notexpressed in the other.  In conclusion, the studies reported in thisdissertation have helped to reveal the dynamics of gene expression in the fetalbrain during the last stage of gestation, and have shown that microarraytechnology is a powerful approach suitable to measure the expression of thousandsof genes in ovine fetal brain tissues.
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 Maria Belen Rabaglino.
Thesis: Thesis (Ph.D.)--University of Florida, 2012.
Local: Adviser: Wood, Charles E.
Electronic Access: RESTRICTED TO UF STUDENTS, STAFF, FACULTY, AND ON-CAMPUS USE UNTIL 2014-12-31

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2012
System ID: UFE0045039:00001

Permanent Link: http://ufdc.ufl.edu/UFE0045039/00001

Material Information

Title: Application of the Microarray Technology to Study the Genomics of the Ovine Fetal Brain During the Last Stage of Gestation
Physical Description: 1 online resource (188 p.)
Language: english
Creator: Rabaglino, Maria Belen
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2012

Subjects

Subjects / Keywords: brain -- fetus -- genomics -- microarray -- parturition -- sheep
Animal Molecular and Cellular Biology -- Dissertations, Academic -- UF
Genre: Animal Molecular and Cellular Biology thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: The main purpose of the series of studiespresented in this dissertation was to apply microarray technology to measuregene expression in the late gestation fetal brain in different types ofexperiments. Chapters 3 and 4 describe in vivoexperiments to evaluate the effect of estradiol-3-sulfate (predominantcirculating form in fetal plasma) or estradiol infusion, respectively, on thegenomics of the ovine fetal hypothalamus. The main result of Chapter 3 was thatestradiol-3-sulfate strongly induced genes related to feeding behavior whilethe unexpected result of Chapter 4 was to reveal that ICI 182 780, a putativeestrogen-receptor blocker, stimulates estrogen agonist genomic responses. Somebiological processes were similarly induced by estradiol-3-sulfate orestradiol/ICI 182 780 while others were only caused by one or other treatment.Therefore, the genomic actions of estradiol-3-sulfate and estradiol in theovine fetal hypothalamus can be independent of each other. Chapters5 and 6 describe, respectively, time course experiments with a single conditionor comparison of two conditions. In Chapter 5, the objective was to identifygroups of co-expressed genes that follow an increasing or decreasing expressionpattern toward the end of pregnancy in each region of the ovine fetal brain. Genes in which expression increased over time were related withneuronal differentiation and gliogenesis while genes displaying a decrease inexpression over time were related to cell cycle, as described previously. Anovel finding is that genes related with hematopoiesis progressively increasedin expression toward the end of gestation. In Chapter 6, the objective was to identifygenes highly expressed temporally in hypothalamus compared to pituitary duringthe very last period of gestation. As expected, genes related with a functionalfetal hypothalamus or pituitary were highly expressed in one region but notexpressed in the other.  In conclusion, the studies reported in thisdissertation have helped to reveal the dynamics of gene expression in the fetalbrain during the last stage of gestation, and have shown that microarraytechnology is a powerful approach suitable to measure the expression of thousandsof genes in ovine fetal brain tissues.
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 Maria Belen Rabaglino.
Thesis: Thesis (Ph.D.)--University of Florida, 2012.
Local: Adviser: Wood, Charles E.
Electronic Access: RESTRICTED TO UF STUDENTS, STAFF, FACULTY, AND ON-CAMPUS USE UNTIL 2014-12-31

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2012
System ID: UFE0045039:00001


This item has the following downloads:


Full Text

PAGE 1

1 APPLICATION OF THE MICROARRAY TECHNOLOGY TO STUDY THE GENOMICS OF THE OVINE FETAL BRAIN DURING THE LAST STAGE OF GESTATION By MARIA BELEN RABAGLINO A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2012

PAGE 2

2 2012 Maria Belen Rabaglino

PAGE 3

3 To my parents, Adriana Garay and Hugo Rabaglino, for all the support and love I have received from them my whole life and especially during these last years that I have been far away from home

PAGE 4

4 ACKNOWLEDGMENTS I would like to express my most sincere gratitude to my advisor, Dr. Charles E. Wood, for his superv ision and help during my Ph.D. program. He is not only an excellent person but also an outstanding mentor. The astonishing positive attitude and professionalism that he uses in front of every situation have showed me the correct way to confront any challen ge. I hope I can state with these words how grateful I am to him for the enriching experience of this Ph.D. program, both for my professional and personal life. I will strive to follow his example in the future I want to acknowledge the members of my co mmittee: Dr. Peter Hansen, Dr. Nancy Denslow and Dr. Carlos Risco, for their suggestions and directions that contributed to shape this dissertation. In addition, I want to thanks Dr. Hansen for his friendship and the precious help I received from him when I was looking for a Ph.D. program during the time I was finishing my M.Sc. program. I really appreciate his support during that transition period. Also, I would like to thanks Dr. Denslow for her kindness and the invaluable inputs s he has contributed to my doctoral research. Dr. Denslow provided the bases for this dissertation, which introduced me to the bioinformatics world, an area that I had not imagined to enjoy this much. I am very grateful for that. Last but not least, I want to give my most special t hanks to Dr. Carlos Risco, my mentor during my M.S program. He provided me the initial training and invaluable guidance on my first years of graduate school. I wish to express my deepest gratitude to the incredible contribution that he has made to my prof essional and personal development. I would like to give immense thanks to Dr. Elaine Richards, for the wonderful help I have received from her during the years of my Ph.D. program. She has been

PAGE 5

5 responsible for all the steps involving sample preparation fo r microarray hybridization. Her excellent job in this area has allowed the generation of the satisfactory results found in the studies described in this dissertation I want to ackn owledge also Dr. Maureen Keller Wood, for her efforts, support, and constru ctive criticism of my research. Both Dr. Richards and Dr. Keller Wood have represented a priceless aid to my career as a scientist I hope I can reflect in my future everything I have learned from them. Special ng, for all the patience she has had while help ing me in the laboratory techniques, always with a smile in her face. I want to give many thanks to my other office mate and friend, Eileen Chang. I not only shared the office with her but also conferences, su rgeries, sheep care hours and many other activities that have left unforgettable moments in m y memory. I would like to thank the members of the Dr. Wood and Dr. Keller Wood labs (Ashley Grapes, Teresa Collins, Xiaodi Feng, Heidi Straub and Tatiana Ramirez Hiller) for the excellent environment I have had during the time spent there. Dolores and Alberto Gochez, and their sons Nicolas and Francisco (my godson), Federico Martin, Esteban Rios and Anna Denicol, also an amazing partner of conference trips. I really appreciate the friendship I cultivated with all of them and I am sure that i t will last forever. I also want to give m y most sincere thanks to all my other good friends from the Department of Animal Sciences I would like to express an especial acknowledgement to the authorities from the National University of Rio Cuarto; Rio Cua rto, Cordoba, Argentina. This University is

PAGE 6

6 where not only I have graduated as D.V.M., but also I have continued working as a faculty member. Finally I wish to offer my deepest gratitude to my parents: Hugo and Adriana to my brother and sister in law E zequiel and Rossana, to my beloved niece and nephew: Francesca and Bruno and to the rest of my family. The distance from them during these years has represented one of the hardest things to get used to I know how difficult it was for them too so I apprec iate all the support they have given to me during these years. At the end, I wish I could put in words how grateful I am to my fianc Oscar Queiroz. I would just say that he personifies the best attributes that I have ever imagined in someone. Thanks to Go d to allow me to find him and to join our lives. Thanks to God also for give me the opportunity to live and to enjoy this unique experience.

PAGE 7

7 TABLE OF CONTENTS page ACKNOWLEDGMENTS ................................ ................................ ................................ .. 4 LIST OF TABLES ................................ ................................ ................................ .......... 11 LIST OF FIGURES ................................ ................................ ................................ ........ 12 ABSTRACT ................................ ................................ ................................ ................... 15 CHAPTER 1 INTRODUCTION ................................ ................................ ................................ .... 17 2 LITERATURE REVIEW ................................ ................................ .......................... 22 Gene Exp ression ................................ ................................ ................................ .... 22 Gene Expression and Development ................................ ................................ ....... 22 Methods to Measure Gene Expression ................................ ................................ ... 23 Microarrays ................................ ................................ ................................ ....... 24 Microarrays platforms ................................ ................................ ................ 25 Affymetrix platform ................................ ................................ ..................... 25 Agilent platform ................................ ................................ .......................... 26 RNA Sequencing (RNA seq) ................................ ................................ ............ 27 Deep sequencing platforms ................................ ................................ ....... 27 Roche 454 life science ................................ ................................ ............... 27 Illumina/Solexa analyzer ................................ ................................ ............ 28 Applied biosystem SOLiD ................................ ................................ .......... 28 The Importance of Microarray and RNA Seq Technologies ............................. 29 Pros and cons of RNA seq ................................ ................................ ......... 30 Pros and cons of microarrays ................................ ................................ .... 30 Use of the Microarray Technology to Measure Gene Expression during Fetal Life ................................ ................................ ................................ ....................... 32 Steps Involved in a Microarray Experiment ................................ ............................. 33 Sample Size in Microarray Experime nts ................................ ........................... 34 Sample Preparation for Microarrays ................................ ................................ 35 Pre processing Microarray Raw Data ................................ ............................... 36 Imaging acquisition and background subtraction ................................ ....... 36 Normalization ................................ ................................ ............................. 37 Examples of pre processing microarray data using the R software ........... 39 Creating an expression data set ................................ ................................ 42 Quality Assessment ................................ ................................ .......................... 43 Identification of Differentially Expressed Genes ................................ ............... 45 Comparison of two or more groups ................................ ............................ 47 Analysis of a time course experiment ................................ ........................ 49 Software for Microarray Data Analysis ................................ ............................. 50

PAGE 8

8 Data Mining ................................ ................................ ................................ ...... 51 Hierarchical clustering ................................ ................................ ................ 52 Cytoscape ................................ ................................ ................................ .. 53 Weighted gene co e xpresssion network analysis ................................ ...... 53 Microarray Validation ................................ ................................ ........................ 54 Summary of Microarray Analysis ................................ ................................ ............ 55 3 COMPARISON OF TWO GROUPS EXPERIMENT: GENOMICS OF ESTRADIOL 3 SULFATE ACTION IN THE OVINE FETAL HYPOTHALAMUS ..... 63 Introduction ................................ ................................ ................................ ............. 63 Materials and Methods ................................ ................................ ............................ 64 Animal Procedures ................................ ................................ ........................... 64 Blood Collection and Plasma Hormone Assays ................................ ............... 66 Sample Collection ................................ ................................ ............................ 66 RNA Extraction and Preparation ................................ ................................ ...... 67 Microarray Hybridization ................................ ................................ ................... 67 Statistical Analysis ................................ ................................ ............................ 68 Function al Annotation ................................ ................................ ....................... 69 Clustering Analysis ................................ ................................ ........................... 69 Quantitative Real Time (qRT) PCR Validation ................................ ................. 70 Results ................................ ................................ ................................ .................... 72 Plasma Hormones ................................ ................................ ............................ 72 Microarray Results ................................ ................................ ........................... 72 Functional Annotation ................................ ................................ ....................... 72 Clustering Analysis ................................ ................................ ........................... 73 Upregulated network ................................ ................................ .................. 73 Downregulated network ................................ ................................ ............. 73 Quantitative Real Time (qRT) PCR ................................ ................................ .. 73 Overlap with ESR 1 and HIF1A Regulated Genes ................................ ........... 74 Discussion ................................ ................................ ................................ .............. 74 Neuropeptides Related to Feeding Behavior ................................ .................... 75 Mediators of Vascularization and Hypoxia Response ................................ ...... 76 COPI System ................................ ................................ ................................ .... 78 Transforming Gro wth Factor Beta 1 ................................ ................................ 78 4 COMPARISON OF MORE THAN TWO GROUPS EXPERIMENT: AGONIST ESTRADIOL AND ICI 182 780 IN THE LATE GESTATION OVINE FETAL HYPOTHALAMUS ................................ .................... 90 Introduction ................................ ................................ ................................ ............. 90 Materials and Methods ................................ ................................ ............................ 91 Animal Procedures ................................ ................................ ........................... 91 Experimental Procedure and Blood Sample Collection ................................ .... 92 Tissue Sample Collection ................................ ................................ ................. 93 RNA Extraction and Preparation ................................ ................................ ...... 93 Microarray Hybridization ................................ ................................ ................... 94

PAGE 9

9 Functional Annotation ................................ ................................ ....................... 94 Statistical Analysis ................................ ................................ ............................ 95 Clustering Analysis ................................ ................................ ........................... 96 Quantita tive Real Time (qRT) PCR Validation ................................ ................. 97 Results ................................ ................................ ................................ .................... 98 Plasma Hormones ................................ ................................ ............................ 98 Microarray Results ................................ ................................ ........................... 99 Network An alysis ................................ ................................ .............................. 99 Quantitative Real Time (qRT) PCR ................................ ................................ 100 Discussion ................................ ................................ ................................ ............ 100 5 TIME COURSE EXPERIMENT WITH A SINGLE CONDITION: GENE CO EXPRESSION ANALYSIS OF THE OVINE FETAL BRAIN ONTOGENY DURING THE LAST STAGE OF GESTATION AND FIRST D AY OF EXTRA UTERINE LIFE ................................ ................................ ................................ ..... 114 Introduction ................................ ................................ ................................ ........... 114 Materials and Methods ................................ ................................ .......................... 115 Tissue Collection ................................ ................................ ............................ 115 RNA Extraction and Preparation ................................ ................................ .... 115 Microarray Hybridization ................................ ................................ ................. 116 Microarray Data ................................ ................................ .............................. 11 6 Principal Component Analysis ................................ ................................ ........ 117 Statistical Analysis ................................ ................................ .......................... 117 Superv ised Weighted Gene Co expression Network Analysis (WGCNA) ...... 117 Consensus WGCNA ................................ ................................ ....................... 118 Gene Ontology Analysis ................................ ................................ ................. 119 Quantitative Real Time (qRT) PCR Validation ................................ ............... 119 Results ................................ ................................ ................................ .................. 121 Principal Components Analysis ................................ ................................ ...... 121 Statistical Analysis ................................ ................................ .......................... 121 Supervised Weighted Gene Co expression Network Analysis ....................... 122 Consensus WGCNA ................................ ................................ ....................... 123 Quantitative Real Time (qRT) PCR ................................ ................................ 123 Discussion ................................ ................................ ................................ ............ 123 6 TIME COURSE EXPERIMENT WITH TWO CONDITIONS: COMPARATIVE GENOMIC ANALYSIS OF THE OVINE FETAL HYPOTHALAMIC AND PITUITARY ONTOGENIES DURING THE LAST STAGE OF G ESTATION AND FIRST DAY OF EXTRA UTERINE LIFE ................................ ............................... 147 Introduction ................................ ................................ ................................ ........... 147 Materials And Methods ................................ ................................ ......................... 148 Tissue Collection ................................ ................................ ............................ 148 RNA Extraction and Preparation ................................ ................................ .... 148 Microarray Hybridization ................................ ................................ ................. 149 Microarray Data ................................ ................................ .............................. 149

PAGE 10

10 Statistical Analysis ................................ ................................ .......................... 149 Clustering Analysis ................................ ................................ ......................... 150 Network Analysis ................................ ................................ ............................ 150 Quantitati ve Real Time (qRT) PCR Validation ................................ ............... 151 Results ................................ ................................ ................................ .................. 152 Statistical and Clustering Analysis ................................ ................................ .. 152 Network and Gene Ontology Analysis ................................ ............................ 152 Quantitative Real Time (qRT) PCR ................................ ................................ 153 Discussion ................................ ................................ ................................ ............ 154 7 GENERAL DISCUSSION ................................ ................................ ..................... 163 Applic ation of the Microarray Technology ................................ ............................. 163 Biological Interpretation of the Results ................................ ................................ 168 LIST OF REFERENCES ................................ ................................ ............................. 172 BIOGRAPHICAL SKETCH ................................ ................................ .......................... 186

PAGE 11

11 LIST OF TABLES Table page 2 1 Examples of target files for different type of experiments ................................ ... 59 2 2 Examples of data files for different type of experiments ................................ ..... 60 2 3 Examples of experimental design files for different type of experiments ............ 61 2 4 Quality control parameters for the Affymetrix bovine platform ........................... 62 3 1 Sequences of primers and probes for real time PCR analysis ........................... 87 3 2 Statistically overrepresented biological processes found on the significant clusters of the treatment network ................................ ................................ ........ 88 3 3 Statistically overrepresented biological processes found on the significant clusters of the control network ................................ ................................ ............ 89 4 1 Sequences of primers and probes for real time PCR analysis ......................... 111 4 2 Statistically over represented biological processes found of the on the significant clusters of subnetworks from the up regulated networks for all treatme nts. ................................ ................................ ................................ ........ 112 4 3 Statistically over represented biological processes found of the on the significant clusters of subnetworks from the down regulated networks for all treatments. ................................ ................................ ................................ ........ 113 5 1 Sequences of primers and probes for real time PCR analysis ......................... 143 5 2 Representative biological processes significantly enriched with the genes forming part of the top modules with highest correlation in each of the ovine fetal brain regions ................................ ................................ ............................. 144 5 3 Significantly enriched KEGG pathways in all ovi ne fetal brain regions analyzed ................................ ................................ ................................ ........... 145 5 4 Genes highly positive correlated with ovine fetal age and with high connectivity within respective modules in all brain regions ............................... 146 6 1 Sequences of primers and probes for real time PCR analysis ......................... 161 6 2 Statistically over represented biological processes found of the on the networks inferred with genes with the highest expression in hypothalamus or pituitary compared to each other ................................ ................................ ...... 162

PAGE 12

12 LIST OF FIGURES Figure page 2 1 Box plots of absolute rates of expression c hange within each stage of life ........ 56 2 2 Main steps of the microarray technique ................................ .............................. 57 2 3 Plot of power versus sample size ................................ ................................ ....... 58 3 1 Plasma levels of estradiol and estradiol 3 sulfate during the infusion period in treated and control fetuses ................................ ................................ ................. 80 3 2 Volcano plot showing the log odds of differential expression versus the log fold change in gene expression between estradiol 3 sulfate treatment and control arrays ................................ ................................ ................................ ...... 81 3 4 Estradiol 3 sulfate induced changes in mRNA expression of HIF1A; ARNT2, TGFB1 and COPB1 ................................ ................................ ............................ 83 3 5 Estradiol 3 sulfate induced strong changes in mRNA expression of AGRP and NPY ................................ ................................ ................................ ............. 84 3 6 Clusters of genes upregulated and do wnregulated by E 2 SO 4 and clusters of genes known to be transcrip ................................ ....... 85 3 7 Overlapping clusters of genes upregulated by E 2 SO 4 and of genes known to be transcriptionally regulated by HIF1A ................................ .............................. 86 4 1 Plasma levels of estradio l and ACTH during the infusion period in treated and control fetuses ................................ ................................ ................................ .. 104 4 2 Network resulting from the merge of the network composed for up regulated genes for estradiol with the networks composed for down regulated genes for ICI 182 780 low dose and high dose ................................ ................................ 105 4 3 Network resulting from the merge of the network composed for down regulated genes for estradiol with the networks composed for down regulated genes for ICI 182 780 low dose and high dose ................................ ................ 106 4 4 Network resulting from the merge of the network composed for up regulated genes for estradiol with the networks composed for up regulated genes for ICI 182 780 low dose and high dose ................................ ................................ 107 4 5 Network resulting from the merge of the network composed for down regulated genes for estradiol with the networks composed for down regulated genes for ICI 182 780 low dose and high dose ................................ ................ 108

PAGE 13

13 4 6 Estradiol and ICI both doses treatments induced changes in mRNA expression of VEGFA and JAG1 ................................ ................................ ...... 109 4 7 Venn diagram of overlapped genes between the up regulated genes induced by treatments with estradiol and ICI 182 780 5 ug/kg or 5 mg/kg dose. ........... 110 5 1 Principal component analysis for gene expression on each sample of the data set ................................ ................................ ................................ ............. 131 5 2 Venn Diagram of the number of differentially expressed genes following a temporal profile during the last stage of gestation in different regions of the ovine fetal brain. ................................ ................................ ............................... 132 5 3 Weighted gene co expression network analysis identifies modules of co expressed genes following an increasing or decreasing expression temporal pattern du ring last stage of gestation in ovine fetal cortex, brainstem, hippocampus and hypothalamus ................................ ................................ ...... 133 5 4 Module membership versus gene sig nificance plots ................................ ........ 134 5 5 Consensus weighted gene co expression analysis done in the four brain regions of the fetal ovine brain ................................ ................................ .......... 135 5 6 Trajectories of average expression for genes involved in the sphingolipid metabolism, cell cycle, response to hypoxia, and response to estrogen, for all brain regions analyzed. ................................ ................................ .................... 136 5 7 Trajectories of average expression for genes involved in the hemat opoietic cell lineage and activation of the immune system for all brain regions analyzed. ................................ ................................ ................................ .......... 137 5 8 Gene expression of PTGS2 measur ed by microarray at 80, 120, 145 days of gestation, and 1 and 7 days of extra uterine life and corresponding fold changes in mRNA concentration relative to 80 days, measured by qRT PCR in samples from ovine fetal hypothalamus ................................ ........................ 138 5 9 Gene exp ression of CD34, CD109, CD44, CD5 and CD9 measured by microarray at 80, 100, 120, 130, 145 days of gestation and 1 day of extra uterine life and corresponding fold changes in mRNA concentration relative to 80 days, measured by qRT PCR in samples from ovine fetal brainsteam. .. 139 5 10 Gene e xpression of CSF1R, CSF1, IL34, CD11B, and CD81 measured by microarray at 80, 100, 120, 130, 145 days of gestation and 1 day of extra u terine life and corresponding fold changes in mRNA concentration relative to 80 days, measured by qRT PCR in samples from ovine fetal cortex ........... 140

PAGE 14

14 5 11 Gene expression of CD3G, CD3D and CD3E measured by microarray at 80, 100, 120, 130, 145 days of gestation and 1 day of extra uterine life and corresponding fold changes in mRNA concentration relative to 80 days, measured by qRT PCR i n samples from ovine fetal hippocampus .................. 141 5 12 Gene expression of CD24, MBP FGRIIB, IL10 and TGFB measured by microarray at 80, 100, 120, 130, 145 days of gestation and 1 day of extra uterine life and corresponding fold changes in mRNA concentration relative to 80 days, measured by qRT PCR in samples from ovine fetal cortex ........... 142 6 1 Hierarchical clustering of differentially expressed genes between hypothalamus and pituitary. ................................ ................................ .............. 157 6 2 Network inferred with the genes with the highest expression in hypothalamus compared to pituitary ................................ ................................ ........................ 158 6 3 Network inferred with the genes with the highest expression in pituitary compared to hypothalamus ................................ ................................ .............. 159 6 4 Fold change in mRNA concentrations at 120, 130, 145 days of gestational life and 1 day of extrauterine life in the fetal hypothalamus relative to the fetal pituitary for GRIN1; GRIA1; GRIA3 and GRM3 ................................ ................ 160

PAGE 15

15 Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy APPLICATION OF THE MICROARRAY TECHNOLOGY TO STUDY THE GENOMICS OF THE OVINE FETAL BRAIN DURING THE LAST STAGE OF GESTATION By Maria Belen Rabaglino December 2012 Chair: Charles E. Wood Major: Animal Molecular and Cell u l a r Biology The main purpose of the series of studies presented in this dissertation was to apply microarray technology to measure gene expression in the late gestation fetal brain in different typ es of experiments. Chapters 3 and 4 describe in vivo experiments to evaluate the effect of estradiol 3 sulfate (predominant circulating form in fetal plasma) or estr adiol infusion, respectively, on the genomics of the ovine fetal hypothalamus The main re sult of Chapter 3 was that estradiol 3 sulfate strongly induced genes related to feeding behavior while the unexpected result of Chapter 4 was to reveal that ICI 182 780, a putative estrogen receptor blocker, stimulates estrogen agonist genomic responses. Some biological processes were similarly induced by estradiol 3 sulfate or estradiol/ICI 182 780 while others were only caused by one or other tr eatment. Therefore, the genomic actions of estradiol 3 sulfate and estradiol in the ovine fetal hypothalamus ca n be independent of each other. Chapter s 5 and 6 describe, respectively, time course experiments with a single condition or comparison of two conditions. In Chapter 5, the objective was to identify

PAGE 16

16 groups of co expressed genes that follow an increasing or decreasing expression pattern toward the end of pregnancy in each region of the ovine fetal brain Genes in which expression increased over time were related with neuronal differentiation and gliogenesis while genes displaying a decrease in expression ove r time were related to cell cycle, as described previously. A novel finding is that genes relat ed with hematopoiesis progressively increased in expression toward the end of gestation In Chapter 6, the objective was to identify genes highly expressed tempo rally in hypothalamus compared to pituitary during the very last period of gestation. As expected, genes related with a functional fetal hypothalamus or pituitary were highly expressed in one region but not expressed in the other. In conclusion, the studies reported in this dissertation have helped to reveal the dynamics of gene expression in the fetal brain during the last stage of gestation and have shown that microarray technology is a powerful approach suitable to measure the e xpression of thousands of genes in ovine fetal brain tissues

PAGE 17

17 CHAPTER 1 INTRODUCTION The last stage of gestation is the most critical developmental period in the life span of an individual. Dramatic physiological changes take place during this period to prepare the newborn for an adverse environment outside the uterus. Parturition culminat es fetal life, thus, regulation of the parturition process is important since the offspring has to be mature enough to survive in the new environment before parturition is initiated In ruminants, initiation of parturition is a consequence of the neuroen docrine cascade between the hypothalamus pitui tary adrenal (HPA) axis (74) The hypothalamus is responsible for synthesis of corticotro phin releasing hormone (CRH) that control s the secretion of adrenocorticotropic hormone (ACTH) fro m the pituitary. In the ewe, the gestation length varies from 142 to 152 days, with an average of 147 days. The ovine fetal adrenal becomes responsive to ACTH stimulus at about day 120 of gestation (144) leading to a rinse in fetal cortisol secretion and an increasing activity of the fetal HPA axis (147) Increased cortisol secretion at the end of gestation induces placental CYP17 gene which encodes for an enzyme that ha hydroxylase and 17, 20 lyase activities, inducing the conversion of circulating progesterone to estrogen (41, 122) Estrogen in turn stimulates the fetal HPA axis, generating a posit ive feedback loop that increases ACTH and cortisol secretion (109) Estrogen circulates in the fetal plasma as sulfoconjugated estrogen and the conjugated is far more abundant than unconjugated estrogens (149) The fetal brain expresses both steroid sulfatase and estrogen sulfotransferase, allowing for the bidirectional interconversion of estradiol and estradiol

PAGE 18

18 sulfate (102, 103) The action of unconjugated and conjugated forms over the HPA axis is not exactly similar. Therefore, each compound could affect the expression of different genes on the fetal hypothalamus. The general objectives of Chapter s 3 a nd 4 were to study the genomics of exogenous administration of estradiol sulfate and estradiol, respectively, in the ovine fetal hypothalamus. As mentioned above, the fetal period involves pronounced physiological changes, which are coordinated by changes in gene expression. Accordingly, the rate of change in gene expression during fetal life is incredibly high compared with changes in other periods of life, and the rate of change is even more pronounced in the last stage of gestation (23, 65) This fact reflects the importance of the fetal period to establish physiological parameters that will have repercussions throughout adult life. These changes are critical in the brain, which cont rols endocrine and visceral function (63) The study of gene expression in the fetal brain close to the end of gestation can provide insights into the complicated mechanisms that prepare the newborn for extra uteri ne life, and can help to understand pathological processes originated from disruption in the normal gene expression. The general objectives of Chapter s 5 and 6 were to study the genomics of the ovine fetal brain ontogeny during the last stage of gestation and first day of extra uterine life. The study of genomics of a given organism was limited years ago because routine laboratory techniques allowed the study of relatively few genes. Fortunately, the development of high throughput technologies, such as th e microarray technique or deeps sequencing method, made possible the investigation of global changes in gene expression caused by a treatment, drug or physiological status. The accession to high

PAGE 19

19 throughput technologies has caused a great impact in the biol ogical sciences and can give insights into the gene expression dynamics of a type of cell or tissue, and they can also generate new hypotheses for future experiments. Mi croarray technology was established much earlier than deep sequencing technology. Microarrays can measure the expression of thousands of genes but they are limited to the genes present in the arrays. However, microarrays are still more accessible and easie r to analyze than deep sequencing technologies (83) This is especially true with regard to the sheep, which does not have a fully annotated genome. The studies presented in this dissertation were performed usin g an Agilent microarray platform developed for Ovis aries, the animal model used in our lab. The main purpose of the series of studies presented in this dissertation was to apply the microarray technology to measure gene expression in the fetal brain close to parturition in different types of experiment: Chapter s 3 and 4 describe, respectively, comparison of two or more than two groups experiments. In Chapter 3, chronically catheterized ovine fetuses of around 125 days of gestational age were infused intra cerebrally with estradiol 3 sulfate, a sulfoconjugated estrogen that circulates in fetal plasma at a concentration 40 100 times greater than unconjugated estrogens (149) Control ovine fetuses were infused with saline solution. The objective of this Chapter was to measure the changes in gene expression caused by the estradiol 3 sulfate treatment compared to the control in the fetal hypothalamus. In Chapter 4, chronically catheterized ovine fetuses of around 125 days of gestational age were infused intravenously with estradiol (estradiol group), or

PAGE 20

20 ICI 182 780 5 mg/kg (ICI high dose group) or ICI 182 780 (ICI low dose group) or saline solution (Control group). ICI 182 780 is an anti estrogen compound that has been shown t o interfere with estrogen receptors (28) The objective of this Chapter was to measure the changes of genes expression induced with the estradiol treatment but blocked with ICI 182 780, to identify genes that are p urely activated by estrogen receptor interaction, in the ovine fetal hypothalamus. Unexpectedly, this objective could not been reached because ICI 182 780 in this experiment behaved as an estrogen agonist, an action that has been barely reported for ICI 18 2 780. Chapter s 5 and 6 describe, respectively, time course experiments with a single condition or comparison of two conditions. In Chapter 5, samples were taken from normal fetal brains at five fetal ages and one day of extra uterine life during the last stage of gestation. The objective of this study was to identify groups of co expressed genes that follow an increasing or decreasing expression pattern toward the end of pregnancy en each brain region analyzed. The genes were selected as those showing a correlated differential expression between time points compared to the first baseline measurement, that was 80 days of fetal age. Results from this study can provide insights in the spatial tempora l dynamics of gene express ion during the late gestation fetal period. In Chapter 6, samples from hypothalamus and pituitary taken at three gestational ages and one day of extra uterine life were used for the analysis. The objective of this study was to determine the differentially expressed genes in the fetal hypothalamus compared to pituitary, and vice versa, during the last period of gestation. Genes were selected as those showing a correlated differential expression between time points but comparing two conditions (hypothalamus a nd pituitary). Results from this

PAGE 21

21 study are obvious; since hypothalamus and pituitary have different origins thus functional genes expressed in one organ are not expressed in the other. However, these results are valid to corroborate the microarray as a tec hnology to correctly identify the differentially expressed genes in fetal tissues. In summary, the studies presented in this dissertation successfully employed the microarray technique as a high throughput technology to study the genomics of the ovine fet al brain in different types of experiment. Results from these studies could also provide bases to other researches involving the fetal period, which despite its importance for the future life of an individual still remains unexplored, at least from the

PAGE 22

22 CHAPTER 2 LITERATURE REVIEW Gene Expression Gene expression is defined as the process where information from a gene is used for the synthesis of a functional gene product or in most cases a protein. Gene expression occurs in two major stages: transcription and translation. During transcription the gene is copied to produce an RNA molecule (a primary transcript) that has essentially the same sequence as the gene. These transcripts need to be processed by splicing to remove intron s equences and generate the final transcripts, or messenger RNA (mRNA), that only contains exons, which carry the information for the protein synthesis. This process takes place in the nucleus. Translation refers to the process of de codification of the mRNA for the synthesis of proteins and occurs in the cytoplasm. Genes comprise only about 2% of the human genome; the remainder consists of noncoding regions. The human genome is estimated to contain 20,000 25,000 genes (123) The transcriptome is defined as the entire repertoire of transcripts in a species and represents the link between the information encoded in DNA and the phenotype of an individual (83) Gene Expression and Deve lopment Temporal dynamics of gene expression in the cells that constitute tissues and organs are essential for the functional development of a complex individual, arising from a relatively simple DNA blueprint.

PAGE 23

23 The most complex organ in the body is the br ain, which exerts centralized control over the other organs of the body. Development of the brain circuitry depends on the diversity and precise spatiotemporal regulation of its transcriptome. In the human brain, the greatest changes in spatio temporal g e ne expression occur during fetal life (65) The same pattern was observed for the human pre fontal cortex (Figure 2 1) (23) The fetal period is particularly decisive to de fine physiological changes that can be manifested in the individual after birth, as stated in the Barker Hypothesis (29) Interestingly, both Kang et al. (65) and Colanto uni et al. (23) showed that postnatal changes in gene expression remain relatively low until 50 years of age, when they rise again to mirror the changes in gene expression that occurred in early postnatal life. C haracterization of the transcriptome in the aging brain could provide insights into age related cognitive changes and neurodegeneration. These studies mostly focused on the first half of gestation (around 25 weeks of pregnancy), not on late gestation, or preterm period, probably because the scarce availability of samples from this period. A better understanding of the genomics of the fetal brain in late gestation can help to improve our understanding of the physiological changes necessar y to prepare the ne wborn for extra uterine life. Methods to Measure Gene Expression Tools to measure mRNA expression have been available for years. Some of these tools are reverse transcrip tion PCR (RT PCR), Western blotting expressed sequence tags (ESTs), and serial analysis of gene expression (SAGE) (83) Although RT PCR and Western blot are powerful techniques for measur ing gene or protein expression respectively they require m anual selection of genes or proteins of interest.

PAGE 24

24 Thus, the number of genes or proteins that can be studied at one time is limited with these approaches. The development and improvement of high throughput technologies allowed a more global quantification of the transcriptome than older techniques. The first technology that made possible the measurement of gene expression in large scale was DNA microarrays. Now, it is possible to directly sequence the transcriptional output of the genome through the RNA seq technique. The following is an explanation of both techniques and the pros and cons of one technique respect to the other. Microarrays Microarrays consist of small spots of probes that are immobilized on a solid substrate. These probes are short oligonucl eotide probes representing genomic DNA, and they should be complementary to the transcripts whose presence is to be investigated. The d esign of these probes is based on the genome sequence or on known or predicted open reading frames. Usually, multiple pro bes are used to represent one gene. Transcripts (mRNAs), are extracted from the samples to be investigated and labeled with fluorescent dyes. These dyes are cyanines, which are able to increase the sensitivity range of photographic emulsions (increase the range of wavelengths). The cyanine dyes used for mRNA labelling are Cy3 and Cy5. Cy3 dyes fluoresce yellow green (~550 nm excitation, ~570 nm emission), while Cy5 is fluorescent in the red region (~650/670 nm) but absorbs in the orange region (~649 nm). T ranscripts can be labeled either one (Cy3) or the two dyes (Cy3 and Cy5), corresponding to one channel or two channel microarrays respectively.

PAGE 25

25 These labeled transcripts are hybridized to the probes in the arrays: probes that correspond to transcribed RNA will hybridize to their complementary target. Transcripts are labeled with the fluorescent dyes that emit a light when are excited with a laser. The more a gene is expressed, the greater will be the hybridization with the probe, and the more fluorescent d ye will be associated with that probe. Thus, light intensity is used as a measure of gene expression. Finally, the arrays are washed and scanned to measure the signal (light intensity). Signal for each probe is processed and normalized for subsequent stati stical analysis. The main steps of microarray technique are shown in Figure 2 2. Microarrays platforms This is performed with a robotic arm, which harvests the sample by immers ion and then deposits pico to nanoliter volumes of it (spot) on the slide by touching the glass surface (6) The other type of microarray fabrication consists on the in situ synthesis of DNA on a solid surf ace. This method is employed by two of the most important commercial companies dedicated to fabrication of microarrays: Affymetrix and Agilent (49) Affymetrix platform Affymetrix (Santa Clara, CA, USA) was the fir st company to commercialize arrays fabricated by in situ synthesis of the desired oligonucleotide sequences. These arrays are called GeneChip, and are based on a technique that enables the direct synthesis of high density oligonucleotide arrays directly o n a solid surface via a series of chemical coupling reactions (40) This method consists of the use of a solid support that is linked to molecules protected by a photolabile group. Exposure to ultraviolet (UV) light

PAGE 26

26 causes deprotection and activation of the molecules, which can be coupled to new protected nucleotides b y a chemical reaction. UV light can be precisely located so molecules are activated in specific positions, allowing a stepwise synthesis of the oligos. The length of the oligos (probes) synthesized with this method is relatively short: 25 nucleotides. To overcome problems of low specificity due to the short probes, Affymetrix technology synthesizes multiple probes for the same target gene. In addition, Af fymetrix technology has developed a method to monitor non specific binding: for each probe that has a perfect match (PM) with the sequence of the target gene, there is a probe that has one mismatching (MM) nucleotide with the sequence of the target genes (49) In other words, the sequences of both probes are the same except for a change to the Watson Crick complement in the middle of the MM probe sequence. The intensities generated by the PM and MM hybridization are used to calculated the expression value of the transcript and generate detection calls to classify the transcript as Present, Absent or Marginal. Agilent platform Agilent technology (Santa Clara, CA, USA) manufactures commercial and custom arrays using a c ombination of in situ sy nthesis of oligonucleotides and an inkjet printing process. The method consists of the uniform deposition of the first layer of nucleotide onto specially prepared glass slides followed by a stepwise synthesis (base by base) of 60 me r length oligonucleotide probes. The precise inkjet process delivers accurate volumes (picoliters) of the chemicals to be spotted, following a digital sequence file. Nucleotides are coupled by phosphoramidite chemistry, which allows for very high coupling efficiencies to be maintained at each step in the synthesis of the full length oligonucleotide. A real time quality control inspection system verifies the

PAGE 27

27 chemical deposition at each step in order to avoid premature truncation of the oligonucleotide probes RNA Sequencing (RNA seq) RNA Seq uses deep sequencing technologies to sequence cDNA directly, obtaining complete information about the transcripts contained in one sample. Basically, the technique consists of a conversion of a population of RNA to a libr ary of cDNA fragments with adaptors attached to one or both ends. These fragments can be amplified and then are sequenced in a high throughput manner to obtain short sequences from one end (single end sequencing) or both ends (pair end sequencing). These s equences contains typically between 30 400 bp, depending on the DNA sequencing technology used (138) Deep sequencing platforms So far, there are three main platforms that are actually employed for DNA sequencing: Roc he 454 Life Science, Illumina/Solexa analyzer and Applied Biosystem SOLiD. The other two platforms, the Polonator G.007 and the Helicos HeliScope, have just recently been introduced and are not widely used. Below there is a brief description of the three m ost currently used platforms: Roche 454 life s cience This platform ( http://www.454.com ) was the first commercial platform introduced in 2004 (157) The sequencing method used by 454 is the single nucleotide addition or pyrosequencing. This is a non electrophoretic, bioluminescence method. Fragments of cDNA are amplified and attached to wells of a fiber optic plate. The wells are exposed to a flow of one unlabeled nucleotide at a time allowing synthesis of the complementary strand of DNA to proceed. Incorporation of a nucleotide releases inorganic

PAGE 28

28 pyrophosphate, which is converted into visible light using a series of enzymatic reactions. The light is recorded as a series of peaks call ed a flowgram. The read length is around 400 bp. Illumina/Solexa analyzer This platform ( http://www.illumina.com ) was the second to reach the market and currently, is the most used technology (157) The sequencing method is cyclic reversible termination, using reversible terminators in a cyclic method that comprises nucleotide incorporation, fluorescence imaging and cleavage. Fragments of cDNA are bound through an adapter to a solid surface of a flow cell that is already coated with a dense layer of the adapters. The free end forms a bridge to hybridize with complementary adapters on the surface, initiating the synthesis of the complementary strand, resulting in cluster growth. Multiple cycles of this solid phase amplification followed by denaturation create clusters of ~1,000 copies of single stranded DNA molecules distributed randomly on the surface of the flow cell. Then, the flow cell is loaded with primers, DNA polymerase an d four differently labeled, reversible terminator nucleotides. Following nucleotide incorporation, the remaining unincorporated nucleotides are washed away and imaging is then performed to determine the identity of the incorporated nucleotide. The nucleoti read length of 30 35 nucleotides. Applied b iosystem SOLiD This platform ( http://www.appliedbiosystems.com ) uses a cyclic method similar to the Illu mina analyzer. The difference with this method is the use of DNA ligase and two base encoded probes (oligonucleotide sequence in which two interrogation bases are

PAGE 29

29 associated with a particular dye). Fragments of cDNA are amplified and bound through an adapter to a glass slide. A primer is added to hybridize with the adapter. Then, four fluorescently labeled di base probes compete for ligation to the sequencing primer. Specificity o f the di base probe is achieved by interrogating every 1st and 2nd base in each ligation reaction. The DNA ligase joins the dye labeled probe to the primer. Non ligated probes are washed away, followed by fluorescence imaging to determine the identity of t he ligated probe. Multiple cycles of ligation, detection and cleavage are performed with the number of cycles determining the eventual read length. Following a series of ligation cycles, the extension product is removed and the template is reset with a pri mer complementary to the n 1 position for a second round of ligation cycles. The current read length is between 30 and 35 nucleotides. The Importance of Microarray and RNA Seq Technologies Microarray was the first high throughput technology developed, and its biological application was published more than 10 year earlier than RNA seq. The first report of the use of miniaturized microarrays for gene expression profiling was in 1995 (112) while the first article mentioning the use of RNA sequencing is from 2008 (92) Undoubtedly, the development of both technologies has caused a great impact in the biological sciences and represented a major input to the study the genomics of an organism. Both techniques are currently accessible for the scientific community, and allow the investigation of global changes in gene expression caused by a treatment, drug or physiological status. Results from the ap plication technologies can give insights into the gene expression dynamics of a type of cell or tissue, and they can even generate new hypotheses for future experiments. Thus, high throughput technologies are an incredible resource for the biological resea rch field.

PAGE 30

30 Pros and cons of RNA seq Certainly, the main advantage of RNA seq is the possibility of having direct access to the transcripts sequences. Thus, RNA seq is able to detect polymorphism of a gene, which can provide direct measure of allele specific expression (83) Also, it can quantify individual transcript isoforms generated by alternative splicing, which allows to study the expression of different isoforms for a gene and to make comparisons of isoform diversity and abundance (106) However, RNA seq technology has many biases. For example, there is not yet an ideal standardized methodology for dealing with errors introduced by the technique itself (4) as de veloped for microarrays. Another bias is the depth of the sequencing required to effectively sample the transcriptome, or in other words, how many times sequence a sample. Small amounts of sequencing are sufficient if the gene is highly expressed but more reads will be needed if the gene is moderately or modestly expressed. In addition, the high cost of deep sequencing is another concern for researche r s. Th e cost can decrease the necessary number of reads to accurately sample the transcriptome and also redu ce detection of the number of biological replicates, conducting to inaccurate estimates of gene expression level and probably false inferences (4) Pros and cons of microarrays Contrary to RNA seq, detection of gen e expression with the microarray technology relies on the probes that are already attached to array platform. Thus, the number of the genes to study is limited. Also, it cannot easily detect allele specific differences in gene expression. These disadvantag es are greater if the array was designed for species where the full genome sequence is not completely known.

PAGE 31

31 technology. As mentioned above, microarray was developed much earli er than RNA seq. Thus, it is more standardized and the quality of the data generated with microarrays has improved dramatically. The MicroArray Quality Control (MAQC) consortium, from the food and drug administration (FDA), collects biases and failures ide ntified on the microarrays platforms and helps in their improvement, leading to the development of quality control standards that operate to ensure the utility of a well performed microarray experiment (115) The resul ts from the microarray technique are raw data files (.CEL or text files) that have around 30 MB or 10 MB respectively and can be easily handled with spreadsheet software. Instead, deep sequencing generates files of around 20 30 GB of data. These large file s are handled with encrypting operative system such as UNIX or PYTHON and would need a separate storage rather than regular computers. Microarrays will provide expression measurement for all the genes that hybridize with the transcripts that are present i n the microarray platform. So, coverage should not represent a problem for microarrays, as it could be with RNA seq (83) Despite these advantages, a main benefit of microarrays is the low cost per sample comp ared to sequencing. For example, in the Interdiciplinary Center for Biotechnology Research (ICBR) at the University of Florida, the current service fee to label the targets for microarray (Agilent platform), hybridize and scan is less than $400 per sample. The fee for next generation sequencing depends on the amount of sample to sequence but varies from $1100 (1/8 plate to run) to $7500 (Full plate run) for the

PAGE 32

32 454 pyrosequencing or $1338 (1/8 flow cell 100 cycles) to $9372 (full flow cell 100 cycles) for the Illumina sequencing ( http://www.biotech.ufl.edu/servicefees.html# ). In summary, RNA seq is a powerful technology that can provide answers to researchers interested in genomics details or w orking with some atypical organism. Microarrays could be more limited to study a full genome but it is more adequate to provide a general idea of gene expression in determined samples, since it is cheaper to perform and the resulting data is easier to proc ess and analyze than RNA seq. It can be predicted though that in a few years RNA seq technology will be very accessible and standardized, and it probably will be chosen over microarrays. Use of the Microarray Technology to Measure Gene Expression during Fe tal Life As mentioned before, the fetal period is the most critical term in the life span of an individual with regard to changes in gene expression, at least in the brain. This fact reflects the importance of this period for development of neuron al networ ks and the regulatory circuits that influence function of other organs in the body. Pathological processes that cause disruptions of the brain genomics during the fetal life can have consequences in adulthood (54) The rate of change in gene expression increments during the fetal life to decreases after parturition. The rate of change is even more pronounced during the last period of gestation, as expected, since the fetus needs to adapt for a new life outside the u terus (23, 65) Our lab oratory has used the sheep as an animal model to study the genomics in the fetal brain during the last period of gestation. The ovine fetus is an excellent mode l to study brain development since the entire gestational equivalent of human brain development occurs in utero (104)

PAGE 33

33 The study of gene expression in the ovine fetal bra in was carried out employing microarray te chnology, given the accessibility of this technique, as explained before. The platform used was a newly available one channel array from Agilent, designed for Ovis aries This platform is a sheep 8 X 15 K array slide (Agilent 019921), containing 8 arrays w ith 15208 oligomers with a length of 60 bases. Samples from the ovine fetal brain were hybridized with this platform, washed and scanned with an Agilent G2505B 2 dye scanner at the ICBR at the University of Florida. Features were extracted with Agilent Fea ture extraction 9.1 software. Raw files with the intensities results for each sample were delivered as text files, and were used for posterior analysis. The first challenge faced with this newly developed array was the lack of annotation for the ovine geno me. In other words, the only information for almost all the probes in the array was the 60 nucleotide sequence, but there was no information about the gene complementary to that probe. To solve that, the Blast2go software was used to annotate each probe of the array, as explained in Chapter 3. A total of 12497 probes were annotated, out of the 15208 probes. These probes correspond to a total of 8487 unique genes. The newly annotated platform was submitted to the Gene Expression Omnibus (GEO) website, a pub lic repository that archive high throughput functional genomic data submitted by the scientific community. The platform number is GPL14112 (Agilent 019921 Sheep Gene expression microarray 8x15K, G4813A), public on August 3, 2011. Steps Involved in a Microa rray Experiment The first step in a microarray experiment is common to any biological experiment: correct planning of the experimental design, which involves the consideration of the

PAGE 34

34 number of replicates per condition or sample size. Samples obtained from the experiment need to be carefully prepared for microarray. This step is particularly important since it has a strong influence in the microarray quality. Once the microarray f the signal intensity for each probe, including data normalization. Then, an unsupervised analysis is carried out to evaluate the quality of the data and the distribution of the samples in the array. If the data are acceptable, the statistical analysis co rresponding to the type of experiment is performed, in order to identify the differentially expressed genes (DEG) for the experiment. The final steps are data mining, to extract important information from the DEG, and interpr etation of the results. Next wi ll follow an explanation of each step in the full analysis of microarray data. Sample Size in Microarray Experiments Microarray is a very effective technique. However, the available budget can dictate that the microarray experiment is performed with few r eplicates per group which can result in high false positive (Type I error) and false negative rates (Type II error) (141) Generally, the calculation of sample size is performed to control the type I error rate (79) The calculation of the sample size depends mainly on three factors: the magnitude of the expression change that is biologically meaningful (or desirable to detect), the power to detect the expression change, and the False Discovery Rate (FDR). The concept of FDR is developed below but in general terms it is defined as the expected proportion of false positives among all rejected hypotheses (136) A method proposed by Liu et al. (79) consists in the calculation of the sample size while the FDR is controlled. First, a rejection region for each sample size is defined

PAGE 35

35 according the estimate of the proportion of non differentially expressed genes and the level of FDR to control. Then, the power is calculated for the selected rejection region for each sample size. Finally, the sample size is decided according to the desired power. For example, for a two sample comparison by t test, if the exp ected proportion of DEG with a 2 fold change difference were 50% and the FDR were 0.05, a desired power of 80% would require 5 samples per group. However, if the expected proportion of DEG with a 2 fold change difference were the 10% of the genes in the ar ray, then a power of 80% would require a sample size of 9 per group (FDR=0.05) Figure 2 3 shows the plot of power versus sample size for different expected proportion of DEG with a 2 fold change difference, at a fixed FDR of 0.05. Sample Preparation for M icroarrays Samples for microarray should be collected in RNA free tubes and frozen in liquid nitrogen immediately. Storage of these samples should be done at 80 C. For RNA extraction, it is recommended to use Trizol Reagent (Invitrogen, Carlsbad, CA) fol lowed by a DNA cleanup step for preparing RNA from tissue, using a RNase DNase kit (e.g. Qiagen RNeasy) (31) The absorbance should be checked at 260 and 280 nm for determination of sample purity and concentration. The A260/A280 should be close to 2.0 for pure RNA (ratios between 1.9 and 2.1 are acceptable). Excessive absorbance at 280 nm indicates the presence of protein in the sample while excessive absorbance at 230 nm may indicate the presence of residual phenol in the sample. The next step is to determine the quality of RNA samples prior to microarray analysis, to ensure that differential degradation of samples is not later confounded with differential expression. Sample quality is determined using a Bioanalyzer, wh ich determines an RNA Integrity Number

PAGE 36

36 (RIN), an objective measure of RNA quality. RIN scores vary from 1 10 with 10 being the highest quality samples showing the least degradation (31) Finally, samples are labeled and Pre processing Microarray Raw Data After microarray hybridization, the s canner generates the intensity images that isy signaling is also generated that has to be corrected from the intensity measures. Then, the corrected measures are normalized to remove possible bias in the intensity measures. Each of these steps is explained below, followed by examples of pre processing data from the most used microarray platforms: Agilent and Affymetrix. Imaging acquisition and background subtraction The first step after the hybridization procedure is the image acquisition. This refers to the generation of digital images from the assays e.g., using a phosphorimager (for radioactively labeled targets) or a scanning laser microscope (for fluorescently labeled targets). Once the images have been generated, it is necessary to obtain the corresponding quantified values for further analysis. This process is referred to as feature extraction, which is done using feature extraction software (50) These values need to be subjected to background correction, since part of the signal in the array image comes f rom background noise (nonspecific binding or fluorescence from other chemicals on the slide) and does not reflect actual labeled extract hybridized to reporter (136) The current platforms estimate background locally (i.e., estimating background as a function of location on the chip). So, the signal can be corrected by direct subtraction of the background signal for the corresponding location on the chip.

PAGE 37

37 Normalization Microarray intensity data can have a lot o f variation, which in general derives from technical causes rather than biological causes. Non biological sources of variation can be introduced during sample preparation (e.g., dye effects, labeling), array manufacture (e.g., probe concentration), and hyb ridization (e.g., amount of sample, mRNA quality) and in the measurement process (e.g., scanner inaccuracies) (56) An acceptable method to remove the technical variation and retain the informative biological var iation is data normalization. In this way, normalization puts the data on equal footing before intensity comparisons are made. Different methods have been developed for microarray normalization, either for one or two channels microarray data. Below there is a brief explanation of the normalization methods for each type of microarray Normalization of two channel microarray data. Two values used for normalization and quality control of two channel microarray data are M and A, where M represents the differe nce in log2 intensities in the two channels and A the average log2 intensity in the two channels. Thus, if two identical samples are hybridized to the two channels of an array and the biases are removed, the points should be close to the line M=0 in the M versus A plot. There are several methods to remove the bias within arrays in a two channel microarray data. Typical methods are: normalization by a global constant, print tip loess normalization and global loess normalization. Normalization by a global con stant is a very simplistic approach and it is not generally recommended (50) The print tip loess normalization method consist of subtracting a value m(p,A ) which depends on the value of A for each probe as well a s on the print

PAGE 38

38 print tip to an appropriate subset (possibly all) of the probe on that print tip in the M versus A plot (153) This method is recommended as a routine normalization method for microarrays (120) The global loess normalization consists of subtract ing a value m(A) from the M value. The value m(A) depends on the A value of the probes, and it is computed by array variation and it is not recommend as a routine method (120) Both normalization method s ha ve the effect of shifting points in the plot up or down by different amounts based on their A value, potentially allowing a greater number to lie near M=0. With in normalization of two channel microarray data can be performed using the limma package for the R software (120) The procedure to normalize between arrays is similar to the normalization of one channel microarra y data, explained below. Normalization of one channel microarray data. A simple and effective method for normalization of one channel microarray data is to standardize the numerical values using a mathematical method such as average or the median. In other words, the intensities are transformed so the average or the median of the probes is the same for all the arrays. The assumption of this type of normalization is that the biases are the same for all probes on the assay (50) A more sophisticated normalization method for one channel microarray is the quantile normalization, proposed by Bolstad in 2003 (11) This method transforms the

PAGE 39

39 intensities of all probes on the array into one standard distribution shape, typically with some differences in the distribution tails (which might reflect actual biological differences). The standard distribution is determined by pooling all the individual array distributions. The algorithm mapped every value on any one array to the corresponding This simple between array procedure works as well as most of the more complex procedures. However, one issue that is presented by this method was that genes that were in the upper range of intensity were forced into the same distribution shape, which could reduce technical differences but also biological differences. Fortunately, this problem was fixed in an adjustmen t to the quantile procedure in the latest versions of the affy package (including limma) for the R software (105) Thus, if this method is selected for normalization, the recommendation is to implement it using the affy o r limma package. The assumption of the quantile normalization is that the samples hybridized to the different assays have roughly the same distribution of RNA abundance over the transcripts represented on the array. Examples of pre processing microarray data using the R software R is a free software language and environment for statistical computing and graphics ( http://www.r project.org ). R is powerful software that is easily extended via packages. Bioconductor ( http://www.bioconductor.org ) is open source software that employs R statistical programming language to provide tools for the analysis and comprehension of high throughput genomic data, and it is extensively used for the analysis of microarray data. So far, Bioconductor offers 554 software packages to work with genomic data.

PAGE 40

40 The examples below are developed for data generated with the use of the single channels Affymetrix and Agilent platforms. Preprocessing Affym etrix data. A package that imports, subtracts background, normalizes and summarizes probe level intensity of Affymetrix data is gcRMA (62) This method converts .CEL files (Affymetrix files) into an expression set using the Robust Multi array Average (RMA) method with the help of probe sequence and GC content background correction. The background correction used in gcRMA accounts for background noise, as well as non specific binding. Probe affinity is modeled as a sum of position dependent base effects, and can thus be calculated for each PM and MM value, based on its corresponding sequence information. Each PM value is adjusted by subtracting a shrunken MM value that has been corrected for its affinity. Normalization is performed with the quantile normalization method. Once the probe level PM values have been background corrected and normalized, they are summarized into expression measures. The result is a single expression measure per probe set, per array. The gcRMA pack age needs to be installed with dependences (packages employed by the gcRMA package, such as Affy and MASS). A simple example to import and pre process Affymetrix data with R is the following R code: rectory to the folder that contains the .CEL files library(gcrma) Affy < ReadAffy() eset < gcrma(Affy) data < exprs (eset) # To calculate the expression write.table(data, file = "Data.csv", sep = ",", col.names = NA)

PAGE 41

41 The result is a spreadsheet table (Data.csv) that contains the processed expression measures for all the probes in each array. Preprocessing Agilent data. Import and normalization of Agilent one channel data can be performed with the limma package (118) Th e first step before data input is to create a target file in order to define the microarray raw files of the experiment. The target file can be created in tab delimited text format. Each row represents one array of the experiment. The column that specifies the name of the raw file has to be called Table 2 1 shows examples of target files for different types of experiments. The raw files are read in the order listed in the text file. The R code for import, subtract background and normalize one ch annel Agilent data is the following: contains the raw files and the target file in a text format. library(limma) targets < readTargets("targets.txt") x < read.maimages(target s, source="agilent", green.only=TRUE) y < backgroundCorrect(x,method="normexp") y < normalizeBetweenArrays(y,method="quantile") neg95 < apply(y$E[y$genes$ControlType== 1,],2,function(x) quantile(x,p=0.95)) #This code is applied to filter out low expres sed probes and control probes, retaining the probes that are at least 10% brighter than the negative controls. cutoff < matrix(1.1*neg95,nrow(y),ncol(y),byrow=TRUE) isexpr < rowSums(y$E > cutoff) >= 3 #This last number depends on the number of replicate s per group. y0 < y[y$genes$ControlType==0 & isexpr,]

PAGE 42

42 yave < avereps(y0,ID=y0$genes[,"ProbeName"]) #This code is to average intensities values for the probes that have the same ID in the array. write.table(yave, file = "Data.csv", sep = ",", col.names = NA) This data file (Data.csv) contains the processed expression measures for all the probes in each array. The name of the column for each array is the same than the name of the raw file for that array. It is convenient to replace that name for the group n ame, data file. The final data file should have the first column listing the probe ID and the rest of the columns detailing the expression measure for that probe on each array. Examples of the data file format for different types of experiment are shown in Table 2 2. Creating an expression data set ing with the rest of the analyses since it combines diff erent sources of information into a single convenient structure. In other words, it will combine the expression data from the microarray experiment with the experimental design data. The expression data set can be constructed with the Biobase package from Bioconductor (46) At least two files are necessary to create the expression data set: one is the arrays, obtained as explained above, and the other file is the experimental design file. This file should be tab delimited text file with the first column listing the arrays (with the same name than in the Data file) and the rest of the columns specifying experimental details Examples of experimental design file for different types of experiments are shown in Table 2 3. The R code to create the Expression Data set is the following:

PAGE 43

43 library(Biobase) exprs < as.matrix(read.csv("Data.csv", header = TRUE, row.names = 1, as.is = TRUE)) dim(exprs) #To check that the data file is correctly read, it can be ask the number of rows and columns in the data. colnames(exprs) # Or the name of the columns. pData < read.table("experimentdata.txt", row.names = 1, header = TRUE, sep = \ t") d im(pData) #The same than for the data file can be ask for the experimental design file. rownames(pData) # the row names of the first column should be the same than the column names in the data file. all(rownames(pData) == colnames(exprs)) #The same row names in the experimental design file and column names in the data file can be checked with this code, that should returns TRUE metadata < data.frame(labelDescription = c("Experimental groups", "Replicate"), row.names = c("Group", "Replicate"))#The names listed here are the same than the column names in the experimental design file. phenoData < new("AnnotatedDataFrame", data = pData, varMetadata = metadata) DataSet < new("ExpressionSet", exprs = exprs, phenoData = phenoData) Quality Assessment The qualit y of the data imported and processed should be checked to confirm that the normalization step has removed possible bias. Quality assessment is done through visualization methods that differ if the microarray is one or two channels. Quality of Affymetrix da ta can also be evaluated with quality metrics. Two channel microarray. For two channel microarrays is essential to plot M versus A for each array, in order to compare the distribution of the points that should be close to M=0 (120)

PAGE 44

4 4 One channel microarray. For one channel microarrays or affymetrix arrays, quality can be evaluated observing the distribution of the arrays. The distribution can be assessed through graphical techniques, such us histogram, kernel densi ty estimates and boxplots of the arrays. The R codes that are useful to obtain the corresponding plots are detailed below. The package RColorBrewer (38) allows to give different colors to the resulting bar or boxes. library (RColorBrewer) cols < brewer.pal(8, "Set1"); hist(exprs), freq=FALSE, col=cols); plot(density(exprs), main="Density Plot"); boxplot(exprs, col=cols) Affymetrix: One advantage of the Affymetrix platform is that it provides a number of metrics, known as qu ality control (QC) metrics, which can be used to evaluate the array quality. The most commonly used QC metrics are: average actin. These quality metrics are calculated using the package sympleaffy (27) The QC parameters depend on the specie for which the microarray was designed (QC environment). The QC environment for the corresponding specie is automatically s et up in the sympleaffy package. However, there are certain species that need to be manually set up, in other words, it is necessary to load a configuration file detailing the QC parameters for that specie. The cow ( Bos taurus ) is one of the species that n eeds to be manually set up of the QC environment. Table 2 4 shows the QC environment I constructed for the bovine species. In order to use it, it is necessary to save it as a tab up the QC environment is:

PAGE 45

45 library(simpleaffy) setQCEnvironment("bovinecdf", "/path to the .qcdef file") Data.qc = qc(Data) #This is to generate the QC stats from the imported raw Affymetrix data, before the pre processing with the gcRMA package. The code s to check the QC parameters: average background, scale factor, proportion of avbg(Data.qc); sfs(Data.qc); percent.present(Data.qc); ratios(Data.qc) According to Afymetrix guidelines, the average b ackground values should be fold of each other. The percent present should be similar, especially for replicates. Very low percentage of presents (<20%) can be indicating poor quali ty sample. Finally, the 3' 5' ratios should fall below 3 to show acceptable degradation. A high ratio indicates possible degraded RNA or an inefficient reaction during sample preparation. Identification of Differentially Expressed Genes Almost all the st atistical tests can be applied to microarray data, and the selection of the method depends on the type of experimental design. One important factor to consider when applying the statistical method to determine the DEG is the effects of multiple comparisons since thousands of genes are being studied. In other words, if the statistical test approaches one gene at a time then multiple tests are run in likely to be observed in at least one of the parallel tests (50) For example, the probability that a gene is false positive (identify the gene as DEG when it is not) is 0.05 for that single gene. However, if 100 genes are tested, the pr obability to falsely identify a DEG is 0.994 and the results will have at least one gene that is false positive. Therefore, any p values that could be derived from tests on the individual probe should

PAGE 46

46 be interpreted in the context of thousands of probes th at are being tested simultaneously. Thus, p values need to be modified in order to have sensible error rate control for all of the tests in parallel. The Bonferroni method is a simple method to correct for multiple testing that can be used in microarray d ata analysis (78) This method divides the p value cutoff by the number of probes in the array, generating a new p value threshold, and considers a DEG those genes with a p value smaller than the new threshold. The probabilit y of correctly identifying differentially expressed genes is very small with this method, so many potentially interesting genes may be missed (136) Another method that is widely used to correct for multiple testin g in microarray studies and yield acceptable results, is the false discovery rate (FDR) (152) The FDR measures the (expected) proportion of false positives among the set of all predictions. In other words, the FDR is the expected proportion of false positives among all rejected hypothesis. Instead of trying to avoid any false positives, the FDR controls the proportion of positive calls that are false positives. The method to control the FDR was developed by Benjami ni and Hochberg (9) This method consists in the following: the p values are calculated for each of the probes and then, the p values are ordered from smallest to largest. The ordered p values are plotted v n desired significance level and n is the number of probes in the array. The last p value that lies below the line is used to reject the hypotheses, corresponding to all p valu es less than or equal to the last p value.

PAGE 47

47 This is the most basic version of the Benjamini Hochberg procedure, and it can control the FDR but under certain assumptions (9) This method has been modified to require less strong assumptions about the data (45, 125, 128) Comparison of two or more groups Comparison of two or more groups can be performed with classic tests such as t test and ANOVA. Smyth proposed a variation of the classical t test, which consists in the use of an empirical Bayes method to deal with the few replicates per group, typical of microarray experiment (119) Basically, t his method calculates a moderated t statistic for differential expression for each gene by performing a linear model fit on the data. Then, an empirical Bayes step is applied to moderate the standard errors of the estimated log fold changes and produce mor e stable estimates, especially when the number of replicates is small. This method could be applied to the microarray data using the limma package (118) for the R software. Below are the R codes to apply the linear method, with the limma package, for comparison of two or more than two groups in a microarray experiment. Remember first to create an expression data set as explained above. library(limma) factors < factor(pData(DataSet)$Group, levels=c("Treatment","Control")) design < model.matrix(~0+factors) #create the design matrix. colnames(design) < c("T","C") design #check that the design matrix is correct. fit < lmFit(exprs, design) #perform the linear model fit of the design matrix. contrast.matrix < makeContrasts(T C, levels=design)

PAGE 48

48 contrast.matrix #check the contrast between groups. In this case, the contrast should be T=1 and C= 1 fit2 < contrasts.fit(fit, contrast.matrix) ebayes < eBayes(fit2) #Apply the empirical Bayes step. results < topTable(ebayes, adjust= "BH", number=nrow(exprs) ) #The p values are adjusted with the Benjamini and Hochberg's method to control the false discovery rate (9) write.table(results, file = "Results.csv", sep = ",", col.names = NA) For comparison of more than two groups experiments the codes are the same except that the levels should be fitted to the number of groups. If the experimental design file looks like in Table 2 3 for comparison of more than two groups, after setting the exp ression data set, the R code should be written as: library(limma) factors < "Control")) design < model.matrix(~0+factors) colnames(design) < fit < lmFit(exprs, design) c ontrast.matrix < makeContrasts(TA C, TB C, levels=design) #Of course, the researcher can modify here the contrasts between groups. In this case the differences are calculated with the control group. For all pairwise differences the contrast should be: TA C, TB C, TA TB. fit2 < contrasts.fit(fit, contrast.matrix) ebayes < eBayes(fit2) results < topTable(ebayes, adjust="BH", coef=1, number=nrow(exprs) ) # The results are only for the first contrast (TA C). For results of the second contrast specify coef=2 ) write.table(results, file = "ResultsTA C.csv", sep = ",", col.names = NA)

PAGE 49

49 Analysis of a time course experiment The analysis of developmental time course experiments is challenging, because the time series are usually short, there are very few replicates (one to five replications) per gene, and thousands of genes in an experiment (127) Time course experiments can be longitudinal, if the samples on the different time points are obtained from the same individual, or cross sectional, if the samples are obtained from different individuals. They are also classified as one condition, where the DEG are those that change over time respect the first time point (baseline), or two conditions, where the DEG are those that chang e over time respect the other condition. Linear approaches for the analysis of non time course experiments can be applied to time course experiments. However, these methods are not totally appropriate for a time course data since they treat the time point s as unordered, which can decrease the power to detect DEG (2) Several authors have developed different approaches to analyze time course microarray experiments. For example, Guo et al (53) applied an estimating equation technique but only suitable for longitudinal time course data. Efron et al (36) and Eckel et al (35) proposed an empirical Bayes met hod to detect DEG but the correlation between time points needed to be uniform. Tai and Speed (127) order of differential expression. The ir method can be applied to cross sectional or longitudinal experiments at uniform or arbitrary time points. However, it is not applied a statistical significance to the genes, such as FDR. One method that overcomes many of these limitations is the bayesia n estimation of temporal regulation (BERT), developed by Ayree et al (2) This algorithm uses the time dependent structure of the data and employs an empirical Bayes procedure to stabilize estimates derived from the small

PAGE 50

50 sample sizes. The method is suitable for one or two channel microarrays, longitudinal or cross sectional time course experiments with one or more than one conditions, and allows the specification an alpha value for the FDR. The formulas are applied u sing the BETR package (2) for the R software. The corresponding R code is detailed below. An expression data set should be created first. Example of the experimental design file for a time course experiment with two conditions and two time points is shown in Table 2 3. The R code to apply the BETR package is: library(betr) prob < betr(eset=DataSet, cond=pData(DataSet)$Group, timepoint=pData(DataSet)$Time, replicate=pData(DataSet)$Replicate, twoColor=FALSE, alpha=0.05 ) write.table(prob, file = "betrfinal.csv", sep = ",", col.names = NA) If the experiment is a single condition then the first baseline measurement (time from the code. This method returns the probabilities of differential expression for each gene in the data set. Genes with the best evidence for differential expression will have values close to 1. Software f or Microarray Data Analysis The examples presente d so far were explained for the R software, which uses a programming language. Of course, another software programs have been developed, point and click applications, such us JMP Genomics from SAS. JMP Genomics was employed for the analysis of the microarray data in Chapter 3. Compared to R, JMP Genomics has the great advantage that generates dynamically interactive graphics that

PAGE 51

51 facilitates the exploration of data relatio nship. However, it requires an annual license that needs to be renewed every year. Data Mining Once that the microarray data was properly pre processed and statistically analyzed to identify the DEG, the next step is to perform data mining. Data mining r efers to the process of analyzing data from different perspectives and summarizing it into useful biological information. In other words, data mining is the process of finding correlations or patterns between the DEG. For example, the researcher could won der if there is statistical evidence for the DEG to have some specified biological property larger than the expected if the genes had been chosen at random from the gene population. One of the most popular and powerful associations used for this purpose ar e Gene Ontology (GO) terms. The three organizing principles of GO are molecular function, biological process, and cellular component ( www.geneontology.org/ ). Also, the Kyoto Encyclopedia of Genes and Genomes (KE GG) is another database resource extensively used to associate the list of genes to biological pathways ( http://www.genome.jp/kegg/ ) Then, one approach for data mining could be to take the list of up regulated or down regulated DEG and input them into open bioinformatics tools, such as WebGestat (156) or DAVID database (59) that determine the GO categories or pathways significantly enriched in the list of genes through the hypergeometric test. However, if the list of DEG is relatively large (hundred to thousand genes), it could be difficult to identify re levant biological information.

PAGE 52

52 Therefore, it is advisable to use a methodology to cluster the genes first and then determine the enriched GO terms or KEGG pathways in the genes belonging to the cluster. Below are detailed the methods I have used to cluste r the DEG genes. Hierarchical c lustering Hierarchical clustering method uses measurement method to decide the degree of similarity between two profiles. The most commonly similarity metric methods used are the Euclidean distance the Pearson correlation co efficient and the correlation (50) The Euclidean distance is the "ordinary" distance between two points that one would measure with a ruler, and is given by the Pythagorean formula T he Pearson corre lation coefficient (or centered correlation) is a measure of the correlation between two variables X and Y and measure of the strength of linear dependence between two variables. The k correlation ranks the data in order of intensity first and then applies standard correlation to the profiles of the ranks. Once that the distance is computed, the data is clustered according the degree of similarity. H ierarchical clustering is a method of cluster analysis that build s a hierarchy of clusters v isualized as a tree with the profiles at the leaves Strategies for hierarchical clustering fall into two types: agglomerative and divisive. Agglomerative, or bottom up approaches, start with each profile alone in in its own cluster Similar cluster s are t hen gradually merged until all profiles are in one cluster. Divisive, or top down approaches, start with a single cluster including all profiles. This is split into two clusters which are in turn cluster and splits are performed recursively as one moves down the hierarchy The output is the same for both approaches: a tree with the profiles at the leaves.

PAGE 53

53 Cytoscape Cytoscape is an open source platform that allows the visualization of complex networks, and it is extensible trough plugins that allow the a ddition of features to this bioinformatics software (117) One of these plugins is GeneMania, which can be employed to infer a network from the input list of genes, using a very large set of functional association da ta (140) Clusters inside the network can be obtained using ClusterOne plugin, which identifies densely connected and possibly overlapping regions within the network (clusters) (5) These highly connected network regions can indicate protein complexes or fractions of them. A p value is calculated for each cluster, based on the one sided Mann Whitney U test performed on the in weights and out weights of the vertices. A n in weights value significantly larger than the out weights value would indicate a valid cluster and not the result of random fluctuations. Thus, a p value is assigned to the cluster. Finally, the statistically enriched GO terms in the list of genes belon ging to a significant cluster can be assessed with the use of the BiNGO plugin (81) Weighted gene co expresssion network analysis The nodes of a weighted gene co expression network (WGCN) correspond to gene expressi on profiles, while the edges are determined by the pairwise correlations between gene expressions. Weighted network means that the edges specify the connection strength between the nodes that are connecting. Then, the pairwise correlations between gene exp ression profiles are used to create modules (clusters) of co expressed genes (155) These modules can be associated to external traits, such us fetal age, to measure the correlation between the gene expression profile and the trait.

PAGE 54

54 Gene significance (GS) is defined as the correlation of i th gene with the associated trait. The higher the value of the GS, the more biologically significant is the i th gene. Module membership (MM) is defined as the correlation of the i th gene respects its corresponding module (the higher is the MM the more connected is the i th gene with the other genes of the corresponding modules). The correlation coefficient of MM and GS can be measure for each module, plotting MM versus GS. Hi gher correlation between MM and GS means that genes that are highly associated with the biological trait are also the central elements of the given module (155) Extensive information about the methodology of WG NA for network construction is detailed in http://labs.genetics.ucla.edu/horvath/CoexpressionNetwork/ Complete tutorials to perform WGCNA using the WGCNA package for R can be found at: http://labs.genetics.ucla.edu/horvath/CoexpressionNetwork/Rpackages/WGCNA/ Microarray Validation The final steps after the complete analysis of the microarray d ata are interpretation and verification of the results. Interpretation of the results should be performed in the biological context of the aims proposed for the experiment. If any of the identified GO terms or pathways have biological significance for the given experiment, the researcher should verify with other method the expression of some of the genes involved in such biological process or pathway. Gene expression measured with microarray can be validated using a quantitative or semiquantitative method. Some of the methods employed for the mRNA quantification are Northern blotting and PCR (134) Real time PCR is the most sensitive quantitative method and requires relative low amount of mRNA. Microarray results can

PAGE 55

55 be also validated with qualitiative methods, such as immnuhistochemistry, which allows the direct detection of the gene product in the tissue. Summary of Microarray Analysis In general terms, a microarray experiments begin with the experimental design and the selection of an appropriate platform to measure global gene expression in the samples from the experiment. The species used for the experiment could decide the platform to use, especially if they are non typical species. The experiment should have enough number of replicates, or sample size, to increase the power in the statistical analysis. Samples obtained from the experiments need to be properly processed for mRNA extraction. Extracted mRNA is labeled and hybridized with the microarray. The inte nsity signaling is obtained and converted into intensity measures, corrected for background signaling and normalized. It is important to perform a quality check before and after normalization, to detect possible outliers. If the data is suitable, the next step is to detect the DEG through the corresponding statistical analysis. The list of DEG should be analyzed with different methods to find the biological significance of those genes. The biological findings should be interpreted in the context of the expe riment. Finally, the expressions of the most interesting genes are validated with other quantitative methods, such us real time PCR.

PAGE 56

56 Figure 2 1. Box plot s of absolute rates of expression change with in each stage of life. Changes in gene expression during fetal life are particularly high, so the y axis scale is different for fetal and infant stages than for all other stages. Open points represent the mean expression correlation across subjects within each age stage Filled points represent the mean e xpression correlation across subjects between adjacent age stages. The grey histogram displays the distribution of ages that marked a change in the trajectory of expression for genes across the postnatal lifespan Reprinted with permission from the Nature Publishing Group. (Source: Colantuoni et al., 2011. Temporal dynamics and genetic control of transcription in the human prefrontal cortex Nature. 478:519 523. Figure 1b, page 520).

PAGE 57

57 Figure 2 2. Main steps of the microarray technique. Transcripts (mRNA) are extracted from the samples to study, converted to cDNA and labeled with fluorescent dyes. Labeled transcripts are hybridized to the probes in the array. After washing, the light intensity of the hybridized transcripts is quantified with a scanner. The data is then processed for statistical analysis. (Source: http://en.wikipedia.org/wiki/DNA_microarray Last accessed: September, 2012).

PAGE 58

58 Figure 2 3 Plot of power versus sample size. The plot s are calculated at different proportions of non 0 ). Fold change for Oxford University Press (Source: Liu P. and Gene Hwang J., 2007. Quick calculation for sample size while controlling false discovery rate with application to microarray analysis. Bioinformatics. 23:739 746. Figure 1a, page 741)

PAGE 59

59 Table 2 1. Examples of target files for differe nt type of experiments Comparison of two groups experiment Array FileName a ArrayName 1 RawFileNameforTreatment1.txt Treatment1 2 RawFileNameforTreatment2.txt Treatment2 3 RawFileNameforTreatment3.txt Treatment3 4 RawFileNameforControl1.txt Control1 5 RawFileNameforControl2.txt Control2 6 RawFileNameforControl3.txt Control3 Comparison of more than two groups experiment Array FileName a ArrayName 1 RawFileNameforTreatmentA1.txt TreatmentA1 2 RawFileNameforTreatmentA2.txt TreatmentA2 3 RawFileNameforTreatmentB1.txt TreatmentB1 4 RawFileNameforTreatmentB2.txt TreatmentB2 5 RawFileNameforControl1.txt Control1 6 RawFileNameforControl2.txt Control2 Time course experiment Array FileName a ArrayName 1 RawFileNameforTreatment _t ime 0_1 .txt Treatment_t ime 0_1 2 RawFileNameforTreatment _t ime0_2 .txt Treatment_t ime 0_2 3 RawFileNameforTreatment _t ime1 _1 .txt Treatment_t ime 1_1 4 RawFileNameforTreatment _t ime1_2 .txt Treatment_t ime 1_2 5 RawFileNameforControl t ime 0_1 .txt Control_t ime 0_1 6 RawFileNameforControl t ime0_2 .txt Control_t ime 0_2 7 RawFileNameforControl t ime1 _1 .txt Control_t ime 1_1 8 RawFileNameforControl t ime1_2 .txt Control_t ime 1_2 a The including the file extension. For Agilent, the raw files are .txt files and for Affymetrix are .CEL files

PAGE 60

60 Table 2 2. Examples of d ata files for different type of experiments Comparison of two groups experiment ProbeID Treatment1 Treatment2 Treatment3 Control1 Control2 Control3 A 2.1 2.2 2.3 4.1 4.2 4.3 B 3.1 3.2 3.3 2.1 2.2 2.3 C 4.1 4.2 4.3 4.1 4.2 4.3 Comparison of more than two groups experiment ProbeID TreatmentA1 TreatmentA2 TreatmentB1 TreatmentB1 Control2 Control3 A 4.2 4.3 3.1 3.2 4.2 4.3 B 3.1 3.2 4.1 4.2 2.2 2.3 C 4.1 4.2 4.3 4.1 4.2 4.3 Time course experiment ProbeID Treatment_time 0_1 Treatment_time 0_2 Treatment_time 1_1 Treatment_time 1_2 Control_time 0_1 Control_time 0_2 Control_time 1_1 Control_time 1_2 A 2.1 2.2 4.2 4.3 2.1 2.2 3.1 3.2 B 3.1 3.2 4.2 4.3 2.2 2.3 2.1 2.2 C 4.1 4.2 4.3 4.1 3.1 3.2 3.3 3.4

PAGE 61

61 Table 2 3. Examples of experimental design files for different type of experiments Comparison of two groups experiment a Group Replicate Treatment1 Treatment 1 Treatment2 Treatment 2 Treatment3 Treatment 3 Control1 Control 1 Control2 Control 2 Control3 Control 3 Comparison of more than two groups experiment a Group Replicate TreatmentA1 TreatmentA 1 TreatmentA2 TreatmentA 2 TreatmentB1 TreatmentB 1 TreatmentB2 TreatmentB 2 Control1 Control 1 Control2 Control 2 Time course experiment a Group Replicate Time Treatment_t ime 0_1 Treatment_time0 1 0 Treatment_t ime 0_2 Treatment_time0 2 0 Treatment_t ime 1_1 Treatment_time1 1 1 Treatment_t ime 1_2 Treatment_time1 2 1 Control_t ime 0_1 Control_time0 1 0 Control_t ime 0_2 Control_time0 2 0 Control_t ime 1_1 Control_time1 1 1 Control_t ime 1_2 Control_time1 2 1 a The row names of the first column in the experimental design file should match with the names of the columns containing the intensity measures in the data file.

PAGE 62

62 Table 2 4 Quality control (QC) parameters for the Affymetrix bovine platform. To use this table as QC environment the contents should be copied to an excel file and then saved as a text file. Make sure that the cursor is below the last raw in the text file. Finally, change the txt extension for qcdef array bovinecdf alpha1 0.05 alpha2 0.065 spk bioB AFFX r2 Ec bioB 3_at spk bioC AFFX r2 Ec bioC 3_at spk bioD AFFX r2 Ec bioD 3_at spk creX AFFX r2 P1 cre 3_at ratio actin3/actin5 AFFX Bt actin 3_at AFFX Bt actin 5_at ratio actin3/actinM AFFX Bt actin 3_at AFFX Bt actin M_at ratio gadph3/gadph5 AFFX Bt gapd 3_at AFFX Bt gapd 5_at ratio gadph3/gadphM AFFX Bt gapd 3_at AFFX Bt gapd M_at

PAGE 63

63 CHAPTER 3 COMPARISON OF TWO GROUPS EXPERIMENT: GENOMICS OF ESTRADIOL 3 SULFATE ACTION IN THE OVINE FETAL HYPOTHALAMUS Introduction In the fetal sheep and pregnant ewes sulfoconjugated estrogens are far more abundant than unconjugated estrogens (19, 133, 149) High concentrati ons of estrone sulfate in uterine vein plasma (compared to peripheral vein plasma) of pregnant sheep suggested a high secretion rate for this steroid by placenta in late gestation (30) Plasma concentrations of est radiol 3 sulfate (E 2 S) are approximately 40 100 times those of estradiol (E 2 ) in the late gestation fetal sheep. E 2 S is taken up by the fetal brain and stimulates responses that are both similar to and distinct from the responses to E 2 (143, 149) Sulfoconjugation increases the half life in the blood (107) and supplies a ready source of E 2 in tissues ( e.g., hypothalamus) that express steroid sulfatase (STS) (103, 143) E 2 S can only bind estrogen receptor after deconjugation by STS. We have proposed that, while the function of the sulf oconjugated estrogens in fetal and maternal sheep is unknown, deconjugation can increase estrogen action in specific tissues that express STS and estrogen receptor (143, 148) The fetal brain ex presses both STS and estrogen sulfotransferase (STF) allowing for the bidirectional interconversion of estradiol (E 2 ) and E 2 S (102, 103) although the ratio of expression for thes e enzymes favors deconjugation (143) The fetal brain also expresses transporters that have the capacity to transport the sulfoconjugated steroid across the blood brain barrier (25) We have recently observed that E 2 S has some actions that are similar to E 2 and some actions that are distinct from E 2 on the HPA (143, 148) C hronic infusions of E 2 to fetal shee p produce sustained increases in both ACTH and cortisol causing a potent

PAGE 64

64 stimulation of the fetal hypothalamus pituitary adrenal (HPA) axis (150) In contrast, long term infusion (2 3 weeks) of E 2 S inhibits the pe ri parturient rise in ACTH and reduces HPA activity, even when plasma cortisol levels are not affected (148) Given the multifactorial response to E 2 S, it is probable that the response includes activation of genes that are not directly responsive to the estrogen receptor The present study was designed to reveal the genomics of E 2 S action in the fetal brain. Using a newly available ovine array, we tested the hypothesis that E 2 S both stimulates and inhibits genes involved in the neuroendocrine pathways that direct or facilitate fetal develo pment at the end of gestation. We also hypothesize d that E 2 S w ould significantly alter the activity of genes involved in late gestation fetal development. Materials a nd Methods Animal Procedures A total of 4 sets of chronically catheterized ovine twin fetuses were studied with one infused with estradiol 3 sulfate intracerebroventricularly (1 mg/day) for 7 12 days using an osmotic mini pump implanted in the fetus, and the othe r s erved as an untreated control. The gestational age at the time of surgery was 120 127 days of gestation a developmental window of time that is prior to the preparturient rise in ACTH. Twin fetuses were randomly assigned to the two groups at the time of surgery. All animals were housed in individual pens located in the Animal Resources Department at the University of Florida and all of these experiments were approved by the University of Florida Institutional Animal Care and Use Committee. The rooms main tained controlled lighting and temperature and sheep were given food and water ad libitum

PAGE 65

65 Food was withheld from the pregnant ewes for 24 hours before surgery. Ewes were intubated and anesthetized with halothane (0.5 to 2%) in oxygen before and during s urgical preparation as previously described (149) Surgery and catheter placement for all fetuses was performed using aseptic technique as previously described, with lateral cerebral ventricle, femoral arterial, and ve nous catheters as well as amniotic fluid catheters (111, 149) For placement of the catheter into the lateral cerebral ventricle, the scalp was retracted and a small catheter (outside diame ter, 0.05 in.; inside diameter, 0.03 in.) attached to an osmotic mini pump (size 2mL2 Alza Corp., Palo Alto, CA ) was inserted thr ough a hole made in the skull. This catheter was held in place using Vet Bond (3M Corp., St. Paul, MN). T he exposed catheter and osmotic mini pump were placed subcutaneously before closing the incision on the head. All minipumps in the treated fetuses were filled with E 2 SO 4 (Sigma Aldrich, St. Louis, MO) in vehicle (water) and minipumps in the control fetuses were filled with vehicle only. The position of the catheters and the function of the pumps w ere verified by visual inspection at the time of sa crifice and tissue collection. At the end of the surgery, antibiotics (750 mg ampicillin) were administered into the amnioti c cavity via direct injection. Vascular catheters were exteriorized through the flank of the ewe using a trochar, where they were maintained in a re movable synthetic cloth pocket. Ewes were treated with 1 mg/kg flunixin meglumine (Webster Veter inary, Sterling, MA) for analgesia and returned to their pens where they were monitored until they could stand on their own. If needed, a second treatment with flunixin meglumin e was administered 24 48 hours after the first treatment with this drug. Twice daily during a 5 day recovery period ewes were treated with antibiotic (ampicillin, 750 mg im : Polyflex, Fort Dodge

PAGE 66

66 Laboratories, Fort Dodge, IA) and rectal temperatures were mo nitored for indication of post operative infection. None of the animals in this study showed any signs of postoperative infection. Blood Collection and Plasma Hormone Assays Blood samples were collected to measure plasma hormone levels for both groups, in order to prove the efficacy of E 2 S treatment and check for fetal fetal transfer between treatment and control twin fetuses. Blood samples from treated and control fetuses were collected from the arterial catheter every other morning after the recovery per iod. Plasma from each sample was obtained by centrifugation at 3,000 xg for 15 minutes at 4C and stored at 20C until analysis. Assays to measure E 2 S and E 2 levels were performed as previously described (149) Mean v alues for E 2 S and E 2 for these experiments have been reported previously (143) Sample Collection Pregnant ewes and t win fetuses of known gestational age (130 to 134 days) were euthanized with an overdose of sodium pentobarbital. Brains were rapidly removed, dissected into distinct regions, and snap frozen in liquid nitrogen. Tissues were collected from hypothalamus, pituitary, hippocampus, medullary brainstem, cerebellum, and cerebral cortex and were stored a t 80C until processed for mRNA. These tissue samples were originally collected for other experiments (143) In the present microarray experiment, the mRNA isolated from hypothalamus was analyzed. This region wa s selected for being a critical component of the HPA axis, a major feature of the fetal stress response and crucial for initiation of parturition in the sheep (84)

PAGE 67

67 RNA Extraction a nd Preparation RNA was extracted from the hypothalam us using Trizol (Invitrogen, Carlsbad, CA) following The RNA was resuspended in RNAsecure, and stored at 80 o C in aliquots until us e. For microarray analysis 20 g of these RNAs, which had been s tored for seven years, were DNase treated using the Turbo RNase free DNase kit (Ambion, Foster City, CA), the concentration determined with a Nanodrop spectrophotometer (ND 1000, ThermoFisher, Wilmington DE) and the integrity of the RNA was measured using an Agilent Bioanalyzer, 2100 model One RNA sample had a n RNA Integrity Number ( RIN ) value of 5.1 and was excluded from further analysis; the remaining RNAs had RIN values of 6.9 through 8.0. One g of the DNase treated RNA was labeled with Cyanine 3 (Cy3) CTP with the Agilent Quick Amp kit (5190 0442, New Castle, DE) according to their methodology, purified with the Qiagen shown in the Quick Amp kit protocol except that the microcentrifugation spins were performed at room temperature instead of 4 o C. The resulting labeled cRNA was analyzed with the Nano D rop spectrophotometer, and the specific activities and the yields of the cRNAs were calculated; these ranged from 10.41 to 1 9.73 pmol Cy3/g RNA and from 5.0 to 12.8 g, respectively. The labeled cRNA was stored at 80 o C until use. Microarray Hybridization This was performed following protocols from Agilent, but briefly 600 ng of each labeled cRNA was fragmented and then mix ed with hybridization buffer using the Agilent gene expression hybridization kit. These were applied to a sheep 8 X 15 K array slide (Agilent 019921), containing 8 arrays with 15, 208 oligomer s with a length of 60

PAGE 68

68 bases and hybridized at 65 o C for 17 h at 10 rpm. The arrays were washed, dried, stabilized, and scanned with an Agilent G2505B 2 dye scanner at the Interdisciplinary Center for Biotechnology Research at the University of Florida. Features were extracted with Agilent Feature extraction 9.1 software. Features flagged as Feature Non uniform outliers were excluded from further analysis Statistical Analysis Transcript levels were normalized to the chip median and log transformed, in order to obtain more power in discovering differences between groups and compensate for systematic differences between the arrays. To identify the genes that are differentially regulated between the treated and control fetus es the normalized and transformed intensities were analyzed by one way ANOVA To estimate whether mi croarray observations were able to predict the categorical outcome (treatment or control groups), class prediction was performed using Distance Scoring, a nonparametric discriminant method that bases predictions for an observation on distances between it a nd observations in a training set. Distances were computed from class centroids, and the statistical test employed was T test (p<0.05). All the statistical procedures were carried out using JMP Genomics 5 software (SAS Institute Inc, Cary, NC, USA). A gene was considered to be significantly differentially expressed (over or under expressed) if both of the following conditions were met: 1) the ratio of the normalized intensity of the treatment fetus sample to normalized intensity in the control fetus sample was higher or lower than a 2 fold change (up or down regulation respectively) ; and 2) d ifferences were considered statistically

PAGE 69

69 Functional Annotation We used func tional annotation of the genes a s a u seful tool to categorize the genes in functional classes, leading to a better understanding of the physiological relevance of altered gene expression The most common method to access this analysis is to place each gene in the gene ontology (GO) hierarchy, developed at the GO Consort ium (3) The GO information for e very gene is not available for the ovine genome. Thus, the Blast2Go software (V 4.2.2) (24) a tool for functional annotation of (novel) s equences, was employed. This software uses the Basic Local Alignment Search Tool (BLAST) to find seq uences similar to the queries. For each probe, the sequence of 60 nucleotides (as supplied by Agilent) was first compared to the nucleotide sequence database using blastn The most congruent genes were selected and their accession numbers were input into the NCBI site (EntrezBatch) to o btain their complete sequences. These sequences were blasted again using blast x in order to compare the nucleotide query sequence translated in all reading frames against the known protein sequence database nr Clustering Analysis The network inference and clustering analysis was performed using CytoScape version 2.7.0 (117) through the following plugins: GeneMania ClusterONE, and BINGO. GeneMania was used to infer network data (140) The set of functional association data between genes was downloaded from the Homo sapiens database. The list of human official symbols for the genes of interest was input into the GeneMania plugin to retrieve the corresponding associati on network. The association data employed was protein protei n and protein DNA interaction. The network was inferred for both up regulated and down regulate d genes (treatment vs. control).

PAGE 70

70 ClusterONE was used to discover densely connected and possibly over lapping regions within the network (clusters) (5) (81) These highly connected network regions can indicate protein complexes or fractions of them. A p value is calculated f or each cluster, based on the one sided Mann Whitney U test performed on the in weights and out weights of the vertices. An in weights value significantly larger than the out weights value would indicate a valid cluster and not the result of random fluctua tions. Thus, a p value is assigned to the cluster. Only the clusters with a p value less than 0.05 were considered in further analyses. BiNGO was run to determine which biological processes are statistically overrepresented in the set of genes correspondin g to the identified cluster (81) The statistical test employed was the hypergeometric test (equivalent to the Fisher test). The threshold p value was 0.05, after correction by the Bonferroni method. Quantitative Rea l Time (qRT ) PCR Validation The mRNA samples e xtracted from the hypothalami of the four set s of twin fetuses (E 2 S /control) were converted to cDNA with a High Capacity cDNA Archive kit using the methodology recommended by the kit manufacturer (Applied Biosystems, Foster City, Calif., USA). The newly synthesized cDNA was stored at 20 C until qRT PCR was performed. A total of 11 se lected genes from the significant clusters identified in the up regulated and down regulated networks were tested by qRT PCR to validate the microarray results The up regulated selected genes were: agouti related protein (AGRP), neuropeptide y (NPY), hyp oxia inducible factor 1, alpha subunit (HIF1A), aryl hydrocarbon receptor nuclear translocator 2 (ARNT2), coatomer protein complex, subunit alpha (COPA) and subunit beta 1 (COPB1) and transforming growth factor,

PAGE 71

71 beta 1 (TGFB1). The down regulated selected genes were: chemokine (C C motif) ligand 3 (CCL3), interleukin 12B (IL12B), interleukin 18 (IL18) and tumor necrosis factor (TNF). Relative expression of AGRP, HIF1A, TGFB1 and TNF were determined by qRT PCR using FAM Taqman probes (HIF1A, TGFB1), VIC Ta qman probes (AGRP, TNF) or MGB probe (NPY) and primers (Sigma Aldrich, St Louis, MO) and Taqman qRT PCR master mix (Applied Biosystems, Foster City, CA). Relative expression of ARNT2, COPA, COPB1, CCL3, IL12B and IL18 were determined using primers (Sigma Aldrich, St Louis, MO) and Syb r Green PCR Master Mix (Applied Biosystems, Foster City, CA). Probes and primers were designed with Primer Express software (Applied Biosystems). Primers for ARNT2, IL12B and IL18 and primers and probes for NPY, AGRP, HIF1A, TGFB1 and TNF were designed from the corresponding ovine mRNA. Primers for COPA, COPB1 and CCL3 wer e designed from the bovine mRNA Sequences for primers and probes and accession numbers are reported in Table 3 1. All primer or probe and primer pairs had e fficiencies greater than 95%. The abundance of actin mRNA was determined in each sample, using primers and VIC Taqma n probes designed from the ovine actin sequence and Taqman qRT PCR master mix (Applied Biosystems, Foster City, CA). All samples were r un in triplicate for each gene and for actin. Relative mRNA between the mean Ct for each gene and the mean Ct for actin mRNA from the same sample. The effect of E 2 S o n each gene was compared by one way ANOVA using the

PAGE 72

72 group; fold change in each sample was calculated as 2 statistical analyses, the criterion for achieving statistical significance was P< 0.05. A p value of 0.1 was considered a tendency. Results Plasma Hormones Treatment with E 2 S caused an increase in plasma levels of E 2 S and E 2 with peaks of 1.36 0.11 ng/mL and 301 67 pg/mL respectively, compared to plasma levels of these hormones in the control fetuses (Figure 3 1). Microarray Results Using one way ANOVA analysis, the expression of a total of 2442 genes were Of these genes, expression of 526 had a fold change higher than 2 (up regulated), and 1916 had a fold change lower than 2 (down regulated). The volcano plot generated from this analysis is shown in F igure 3 2 Class pre diction analysis showed that the probability for each observation of correctly predicting whether it belongs to the treatment or the control group was higher than 91.18%, with a mean of 98.89% and a median of 99.99%. Functional Annotation The gene represented by an oligomer on the microarray could be identified for 1830 oligomers (out of 2442 with significantly altered expression) After duplicate genes were removed 1544 genes remained. The final gene lists that were submitted to the GeneMania plugin contained 3 63 genes in the up regulated category and 1201 genes for the down regulated.

PAGE 73

73 Clustering Analysis Upregulated network The network inferred with the up regulated genes contained 218 nodes and 697 edges The network as visualized using the BioLay out Express 3D software (129) is shown in Figure 3 3. Nodes with similar color in this network represent the clusters selected by Biolayaout Express 3D, that were consistent with the clusters found by ClusterO ne plugin in Cytoscape A total of 9 significant clusters with overrepresented biological processes were identified on th is network, as showed in Table 3 2 after redundant biological processes were simplified. Downregulated network In this case, the netwo rk was composed of 756 nodes and 4668 edges. There were 9 significant clusters with overrepresented biological processes identif ied on this network, Table 3 3 (redundant biological processes were simplified). Quantitative Real Time ( qRT ) PCR In concordance with the microarray results, e xpression of AGRP, NPY, ARNT2, COPB1 and TGFB1 mRNAs were significantly induced by the E 2 S treatment (p<0.05), while the expression of HIF1A mRNA showed a tendency of induction by the E 2 S treatment (p =0.1) (Figure 3 4). T he expression for NPY and AGRP genes showed the greatest increase (5 15 fold) in expression with E 2 S treatment compared to control twins (Figure 3 5 ). There w as no significant increase for COPA mRNA expression and there were no statistically significant c hanges in the mRNA expression for the down regulated genes in the microarray analysis

PAGE 74

74 Overlap w ith ESR 1 a nd HIF1A Regulated Genes To identify genes in the up and down regulated networks whose expression 1), we t ested for overlap of these networks with the known ESR 1 sensitive genes (76) There was no overlap of ESR 1 sensitive genes with genes whose expression were up or down regulated by E 2 SO 4 (Figure 3 6). To identify genes whose expression could be controlled by HIF1A, we performed a similar analysis with genes whose expression is known to be regulated by HIF1A (8) In this case, we found significant overlap with the genes whose expression was upregulated by E 2 SO 4 (Figure 3 7). Discussion Circulation of E 2 in the sulfoco njugated form in the ovine fetus represents an important source of biologically active E 2 after deconjugation by STS. The fetal hypothalamus expresses STS, and the ratio of STS to STF (the enzyme that catalyzes the reverse reaction) is high in the fetal hy pothalamus (102, 143) The fetal hypothalamu throughout the latter half of gestation (110) Because E 2 S cannot bind ER directly and must be converted to E 2 by STS, we have proposed that it functions a s a precursor hormone. The increase of STS levels in the latter stages of gestation suggests an increasing capacity for converting E 2 S to E 2 as the fetus approaches spontaneous parturition. R ecent results from this laboratory have indicated that the action s of E 2 S are not identical to those of E 2 (116) It is therefore possible that E 2 S exerts actions that are not ER mediated. A limitation of the present study is that the design cannot distinguish ER mediated from ER independent mechanisms and cannot distinguish responses to E 2 from possible direct responses to E 2 S (via a novel mechanism). Gene expression

PAGE 75

75 could be influenced by E 2 S, E 2 and/or both since intra cerebral administration of E 2 S in the treated animals inc reased the plasma levels of both E 2 S and E 2 compared to the control (Figure 3 1). Interestingly, expressions of genes known to be directly controlled by ER were not upregulated by E 2 S (Figure 3 6). This suggests that some, if not all, of the fetal hypoth alamic genomic response to E 2 S are mediated by other mechanisms. As discussed below, these mechanisms can include non genomic actions of E 2 (after deconjugation of E 2 S), or could include yet undiscovered signal transduction responses to E 2 S. The expressio n of many of the genes is likely to be influenced by the cellular responses to changes in neurotransmission, which comprise a part of the downstream result of E 2 S administration. Using this logic, it is important to acknowledge that genomic responses, per se are not necessarily direct responses to the infused hormone. Neuropeptides Related t o Feeding Behavior A striking result in this experiment is the dramatic up regulation of NPY and AGRP gene expression in the E 2 S treated fetuses confirmed by qRT PCR ( Figure 3 5) The neural network regulating appetite and energy balance in the adult sheep is established during fetal life: both NPY and AGRP, together with propiomelanocortin (POM C) and cocaine and amphetamine regulated transcript (CART), are expressed i n the arcuate nucleus of the ovine fetal hypothalamus from at least 110 days of gestation, as well as the NPY projections from the arcuate nucleus to the paraventricular nucleus (90) Our data suggest that E 2 S treatment might stimulate alterations in appetite and/or energy balance in fetal sheep, however the mechanism by hich this might occur is unclear Estrogen receptors are present on AGRP and NPY neurons in vivo (39) and in

PAGE 76

76 vitro (7, 131, 132) and can have a dual effect on NPY induction in th e adult female. During proestrus, E 2 stimulates NPY expression and release, whi ch in turn contributes to the stimulation of preovulatory GnRH secretion into the hypophysial portal vessels (39) Despite this stimulatory action in these pathways, chronic administration of E 2 has predominately anor ectic properties (12, 151) in rats and monkeys that depend on the inhibition of NPY and AGRP expression and release (12 26) In contrast to the present results with E 2 S, several reports have demonstrated inhibition of NPY by E 2 E 2 treatment of ovariectomized rats decreased both NPY and AGRP mRNA in the arcuate nucleus (116) Sim ilarly, in rhesus monkeys E 2 decreased AGRP and increased POMC secretion into cerebrospinal fluid (151) The upregulation of NPY and AGRP by E 2 S in the present study could be the consequence of an unknown stimulatory action of E 2 in subpopulations of NPY AGRP neurons or of non ER mediated E 2 S action, although differences of E 2 action between fetal and adult life and of between species cannot be discounted. While the mechanism is unknown, E 2 S might be and important phys iological stimul us for feeding behavior and regulation of energy balance necessary for the survival of the newborn. Mediators of Vascularization a nd Hypoxia Response E 2 S effects on genes related to the response to hypoxia more closely mirrored the expected effects of E 2 E 2 treatment in rat pituitary autografts increase s the expression of pro angiogenic factors, specially vascular endothelium growth factor ( VEGF ) its receptor, fms related tyrosine kinase ( FLT1 ) and HIF1A (80) and E 2 increases the expression of mRNA for HIF1A in endometrium of ovariectomized ewes (64) Our microarray results showed an up regulation o f FLT1 and HIF1A (PCR validation for HIF1A was not statistically significant,

PAGE 77

77 p=0.1) -but not VEGF expression by E 2 S It is therefore not clear whether E 2 S promotes angiogenesis in the fetal hypothalamus. Another up regulated gene from the same cluster was ARNT2, validated by qRT PCR (Figure 3 4 ) The encoded protein from the ARNT2 gene complexes with HIF1A in the nucleus and this co mplex binds to hypoxia respons e elements in enhancers and promoters of oxygen responsive genes (124) HIF1A and ARNT2 are members of the bHLH PAS family together with ARNT and aryl hydrocarbon receptor (AHR) (52) ARNT and AHR can modulate ER depe ndent transcription by protein protein association (14) Cho et al., showed that both the N terminus and C terminus of ER interact with the bHLH PAS domain of HIF 1 decreasing 2 (22) However, th integral to the upregulation of VEGF expression in response to hypoxia (67) Why we did not observe an increase in VEGF expression in the present experiments is unclear, but could be the result of competing responses to E 2 S. The apparent upregulation of both HIF1A and ARNT2 by E 2 S in the present experiments suggests that during the last stage of gestation before spontaneous parturition, the fetal hypothalamus increases the expression of hypoxic related genes in response to increased hypothalamic E 2 and E 2 S concentrations or increased (110) even when the fetus in not undergoing clinical hypoxia. Given that the genes of the bHLH PAS family can interact with ER, the mediator of this effect might be E 2 in jection of

PAGE 78

78 E 2 increased the expression of ARNT2 mRNA in the hypothalamus of rats without affecting the expression of AHR and ARNT mRNA, possibly through interaction with (76) COPI System COPI coated vesicles are vesicular carriers that function in the early secretory pathway especially the retrograde transport of luminal and membrane proteins in the ER Golgi segment of the secretory pathway (10 72) The COPI system is present in the CNS but the effects of estrogen action on the expression of this system in the hypothalamus are unknown. The qRT PCR validation showed a significant difference in mRNA expression between treatment and control onl y for COPB1 (Figure 3 4) but not for COPA. Consequently, this action of E 2 S treatment cannot be confirmed in the present study. Transforming Growth Factor Beta 1 It has been previously documented that E 2 treatment from hypothalamic astrocytes in vitro (17) and increases the expression of TGFB 1 gene in the hypothalamus of ovariectomized female rats in vivo (73, 85) Inc reased TGB1 expression after E 2 stimulation is mediated via ER dependent mechanism that involves the PI3K/Akt signaling pathway (33) The role of the estrogen hypothalamus may be that of a m ediator of the ability of astroctytes to modulate GnRH release and as an intermediary of the neuroprotective effect of E 2 (82) In support of this last action, the main biological processes that were down regulated by the E 2 S treatment in the present study were those related to inflammatory/immune response. However, changes in expression of none of the genes involved in these biological processes could be validated by qRT PCR (CCL3, IL12B, IL18 and TNF ).

PAGE 79

79 We have sugg ested that the actions of E 2 S are likely to be mediated by E 2 liberated after deconjugation of the E 2 S and binding of E 2 to the estrogen receptor. While the present results reveal some genomic responses to E 2 S that are reminiscent of responses to E 2 the NPY and AGRP response does not appear to mimic the response to E 2 reported in the adult. Nevertheless, a limitation of this study is that we did not compare responses to those after treatment with E 2 alone, and the question of whether E 2 S can act through a novel mechanism is still an open question. In conclusion, E 2 S treatment of ovine fetuses near the end of gestation induces an up regulation of genes encoding factors involved in feeding behavior, response to hypoxia and possibly provided neuroprotection in the hypothalamus. The effect of E 2 S treatment on the orexigenic peptides observed in this study was not predicted by studies in adult animals. Our results demonstrate that E 2 S induced a strong increase in the expression of NPY and AGRP genes. The incre ased appetite induced by these neuropeptides could be an important component for the survival of the newborn, and could have an effect on the regulation of energy balance regulation before and after birth.

PAGE 80

80 Figu re 3 1 Plasma levels of estradiol (A) and estradiol 3 sulfate (B) during the infusion period in treated and control fetuses. The difference in hormone concentration between both groups was significant for both hormones (P<0.001)

PAGE 81

81 Figure 3 2. Volcano plot showing the log odds of differenti al expression versus the log fold change in gene expression between estradiol 3 sulfate treatment and control arrays. Same color means same cluster of expression for both groups. Genes located above the significance cutoff (p value=0.05 indicated by the dashed lines) are those with statistically significant difference of expression between the treatment and control groups.

PAGE 82

82 Figure 3 3. Network obtained from the list of up regulated genes as visualized using Biolayout Express 3D software. Nodes with similar colors represent a cluster. Blue nodes denote those that do not belong to any cluster.

PAGE 83

83 Figure 3 4. Estradiol 3 sulfate induced changes in mRNA expression of HIF1A (P=0.1); ARNT2, TGFB1 and COPB1 (P<0.05). Expression of each gene was normalized actin expression in the same sample. Data are mean SE fold differences relative to control group expression. (*) represents statistically significant difference in mRNA expression between treatment and control groups.

PAGE 84

84 Figure 3 5. Estradiol 3 s ulfate induced strong changes in mRNA expression of AGRP actin expression in the same sample. Data are mean SE fold differences relative to control group expression.

PAGE 85

85 Figure 3 6. Clusters of genes upregulated (left top) and downregulated (right top) by E 2 SO 4 and (left and right bottom) clusters of genes known to be (76) Note that there is no overlap between th e known ESR1 sensitive genes and the genes differentially regulated by E 2 SO 4

PAGE 86

86 Figure 3 7. Overlapping clusters of genes upregulated by E 2 SO 4 (top) and of genes known to be transcriptionally regulated by HIF1A (bottom) (8)

PAGE 87

87 Table 3 1. Sequences of primers and probes for real time PCR analysis Gene forward primer reverse primer probe accession number oARNT2 GGTCCACTGCACAGGCTA CA CGGCATCTTCTTCAGGTA TAGTCA CAAGGCTTGGCCGCC AGCA NM_001009452 oIL12B CATCAGGGACATCATCAA ACCA CCACCTGCCGAGAATTCT TTAG Sybr EU340264 IL18 AATCAACCTGTCTTTGAG GATATGC CAGACCTCTAGTGAGGCT GTCCTTATA Sybr NM_001009438 NPY CGGAGGACTTGGCCAGAT AC TGCCTGGTGATGAGATTG ATG ACTCAGCGCTGCGAC NM_001009263 oAGRP GAGGTGCTAGATCCGGAA GGA TGGTGTCCCAGACAGGAT TCA CCCACGTCGCTGCGT AAGGCTG AY310396 oHIF1A GCCACAACGTCACCATAT AGTGA TCTGTCTGTTCTATGACTC CTTTTCC AAGTCGGACAGCCTC ACCCAACAGAG AY485676 oTGFB1 CAGTAAGGATAACACGCT TCAAGTG CCGGTTCATGCCGTGAAT ACATCAACGGGTTCA GTTCCGGCC NM_001009400 oTNF CCCTTCCACCCCCTTGTT ATGTTGACCTTGGTCTGG TAGGA CCACACCATCAGCCG CATTGCA NM_001024860 bCOPA GAAAAACCCCACAGATGC CTAT CCGATAAGATGCAGCACA GATG Sybr NM_001105645 bCOPB1 GGTCTGTCATGCTAATCC ATCAGA AGCAGGGCTAGATGACTG TAGTAAGTT Sybr NM_001078007 bCCL3 GGTGTCATCTTCCAGACC AAAAA TCCTGGACCCAGTCCTCA GT Sybr NM_174511

PAGE 88

88 Table 3 2 Statistically overrepresented biological processes found on the significant clusters of the treatment network (p value< 0.05). Bolded genes were validated by quantitative real time PCR Cl Biological Process Genes 1 feeding behavior; positive regulation of response to stimulus NPY GHRL AGRP 2 response to hypoxia; blood vessel development; positive regulation of cell pr oliferation; response to chemical stimulus HIF1A FLT1 COL3A1 EDN1 ARNT2 C JUN 3 COPI coating of Golgi vesicle; vesicle targeting, to, from or within Golgi; Golgi vesicle budding; membrane budding; organelle localization COPB2 COPA COPB1 CO PZ1 PAFAH1B1 4 transforming growth factor beta receptor signaling pathway; pathway restricted SMAD protein phosphorylation; response to prostaglandin E stimulus; response to estrogen stimulus; response to hypoxia; immune system development; immune system process CAV1 TGFBR1 TGFBR2 TGFBR3 TGFB1 5 RNA splicing; nucleic acid metabolic process PAPOLA HNRNPH2 SFRS5 EFTUD2 CDC40 RBM5 SRP75 NFX1 6 establishment of localization in cell XPO1 ARHGEF2 DERL1 HTATIP2 CENPA PAFAH 1B1 RANBP2 CLTCL1 7 translational initiation EIF3A EIF3G EIF3 P36 8 macromolecule catabolic process ERCC5 DCP2 USP12 USP46 9 mitotic cell cycle CDC6 CDKN1A NCAPD2

PAGE 89

89 Table 3 3 Statistically overrepresented biological processes found on the significant clusters of the control network (p value<0.05) Bolded genes were validated by quantitative real time PCR Cl Biological Process Genes 1 regulation of immune response; signaling pathway FCER1A LAT2 CD36 FYN ITGB6 FCER1G MS4A2 SKAP1 SPN 2 regulation of leukocyte proliferation and activation IL4 IL2RA SLA2 IFNG IL12B CD28 IL2 IL18 3 immune system process; inflammatory response CCL3 CXCL9 CCL15 D PP4 TNF 4 response to wounding FGF7 IL8 CD46 ITGA2 SDC2 5 nucleic acid metabolic process POLR2F SNAPC2 SF3A1 SF3B3 POLR2A SLBP SC 35 PHAX TAF10 GTF2E1 GTF2F2 SNRPB LSM4 GTF3C5 LSM10 SUPT4H1 CPSF2 NFIA ERCC1 GEMIN4 SNRPG 6 protein metabolic process COPS5 UBA3 NEDD8 CUL4B FBXO22 NAE1 7 lipopro tein catabolic process; cholesterol metabolic process APOB LDLR APOE 8 regulation of macromolecule metabolic process SREBF1 MED30 NR4A2 MED12 CDC23 MED14 K35 MED13L ERB 9 cellular protein localization GRPEL1 SEC24A SRPR SSR2 S SR3

PAGE 90

90 CHAPTER 4 COMPARISON OF MORE THAN TWO GROUPS EXPERIMENT: AGONIST GENOMIC ESTRADIOL AND ICI 182 780 IN THE LATE GESTATION OVINE FETAL HYPOTHALAMUS Introduction In ruminants, initiation of parturition is consequence of the neuroendocrine cascade between the hypothalamus pitui tary adrenal (HPA) axis (74) The hypothalamus is responsible for synthesis of the corticotro phin releasing hormone (CRH) that control secretion of adrenocorticotropic hormone (ACTH) from the pituitary. In the ewe, the gestation length varies from 142 to 152 days, with an average of 147 days. The ovine fetal adrenal becomes responsive to ACTH stimulus toward the day 120 of gestation (144) leading to a rise in the fetal cortisol and an increasing activity of the fetal HPA axis (147) Increased cortisol secretion at the end of gestation induces placental CYP 17 hydr oxylase and 17, 20 lyase activities, inducing the conversion of the circulating progesterone to estrogen (41, 122) Estrogen in turn stimulates fetal HPA axis, generating a positive feedba ck loop that increases ACTH and cortisol secretion (109) An anti estrogen compound that has been shown to interfere with estrogen receptors (ER) is ICI 182 780 (28) This drug is largely use for the treatment of breast cancer, since it is an efficacious antagonist of the proliferative actions of estrogen ER dependent in reproductive organs such as the breast and uterus (58) Our la b oratory has administered ICI 182 780 and estradiol into the ovine fetal brain to compare the effect on mRNA expression of prostaglandin synthase 2 (PGHS 2), which affects ACTH secretion, in different regions of the ovine fetal brain (111) However, this

PAGE 91

91 study did not measure global changes in gene expression induced by the treatments with estradiol or ICI 182 780. The obj ective of the present study was to evaluate the genomic effect of exogenous estradiol and ICI 1 82 780 at low and high dose i.v. administration in the ovine fetal hypothalamus. Using a newly available ovine array, we tested the hypothesis that genes purely activated by ER interaction will be up regulated with estradiol and blocked and down regulated with ICI 182 780 treatment, and the same for the reciprocal action. Genes stimulated or inhibited with estradiol and blocked with ICI 182 780 will be involved in the neuroendocrine pathways that direct or facilitate fetal development at the end of gestatio n. Materials a nd Methods Animal Procedures A total of 11 pregnant ewes were used for this study. Three ewes were carrying singleton while the other 8 were pregnant with twin fetuses. The gestational age at the time of surgery was 124 126 days of gestation. The fetuses were randomly assigned to the groups at the time of surgery. If the fetuses were twins both animals were assigned to the same treatment group. Three sets of twin fetuses were infused with saline, two estradiol infusion (500 ug/kg/day), another two sets were receiving ICI 182 780 5 ug/kg/ day infusion, three singletons and one set of twin were treated with ICI 182 780 5 mg/kg/day infusion. However, only one of the twins was considered in the study. Thus, 6 fetuses were assigned to the control group and 4 fetuses for each of the treatment groups. All animals were housed in individual pens located in the Animal Resources Department at the University of Florida. All of these experiments were approved by the University of Florida Institutional Animal Care and

PAGE 92

92 Use Committee. The rooms maintained controlled lighting and temperature and sheep were given food and water ad libitum Food was withheld from the pregnant ewes for 24 hours before surgery. Ewes were i ntubated and anesthetized with halothane (0.5 to 2%) in oxygen before and during surgical preparation as previously described (149) Surgery and catheter placement for all fetuses was performed using aseptic technique as previously described, with femoral arterial, and venous catheters as well as amniotic fluid catheters (111, 149) Vascular catheters were exteriorized through the flank of the ewe using a trochar, where they were maintained in a removable synthetic cloth pocket. Ewes were treated with 1 mg/kg flunixin meglumine (Webster Veterinary, Sterling, MA) for analgesia and returned to their pens where they were monitored until they could stand on their own. If needed, a second treatment with flunixin meglumine was administered 24 48 hours after the first treatment with this drug. Twice daily during a 5 day recovery period ewes were treated with antibiotic (ampicillin, 750 mg im : Polyflex, Fort Do dge Laboratories, Fort Dodge, IA) or Cefazolin (11 22 mg/kg, iv or im) and rectal temperatures were monitored for indicati on of post operative infection. None of the animals included in this study showed any signs of postoperative infection. Experimental P rocedure and Blood Sample Collection After the 5 day recovery period, the corresponding compounds (estradiol, ICI 5 ug/kg, ICI 5 mg/kg and saline) were infused into the fetuses intravenously. Infusions were continued for 48 hours, allowing enough time for genomic actions after alteration of steroid action. The infusions were delivered with an SAI MiniBT Infusion pump (Strategic Applications Inc., Chicago, IL) which was calibrated to dispense the 20 ml

PAGE 93

93 volume, where the compounds were dissolved, continuousl y and uniformly during the 48 h period. Blood samples were collected to measure plasma hormone levels for all the groups. Thee blood samples were collected from the arterial catheter during the experimental period. The first sample was taken before placin g the pump to start the infusion, the second was collected at the 24 hrs of infusion and the last sample at 48 hrs of infusion, before proceed with the sample collection. Plasma from each sample was obtained by centrifugation at 3,000 xg for 15 minutes at 4C and stored at 20C until analysis. Assays to measure estradiol and ACTH levels were performed as previously described (111) Tissue Sample Collection Pregnant ewes and fetuses were euthanized with an overdose of sodium pentobarbital after the 48 hrs of the experimental period. Brains were rapidly removed, dissected into distinct regions, and snap frozen in liquid nitrogen. Tissues were collected from al l brain regions and different body regions, and will be employed for other experiments. In the present microarray experiment, the mRNA isolated from hypothalamus was analyzed. This region was selected for being a critical component of the HPA axis, a major feature of the fetal stress response and crucial for initiation of parturition in the sheep (84) RNA Extraction a nd Preparation RNA was extracted from the hypothalamus using Trizol (Invitrogen, Carlsba d, and stored at 80 o were DNase treated using the Turbo RNase free DNase kit (Ambion, Foster City, CA),

PAGE 94

94 the concentration determined with a Nanodrop spectrophotometer (ND 1000, ThermoFisher, Wilmington DE) and the integrity of the RNA was measured using an Agilent Bioanalyzer, 2100 model. The RNA Integrity Number (RIN) value for the RNAs ranged from 7.1 to 8 .6. One g of the DNase treated RNA was labeled with Cyanine 3 (Cy3) CTP with the Agilent Quick Amp kit (5190 0442, New Castle, DE) according to their methodology, purified with the Qiagen RNeasy kit (Valencia, CA) according to agen protocol as shown in the Quick Amp kit protocol except that the microcentrifugation spins were performed at room temperature instead of 4 o C. The resulting labeled cRNA was analyzed with the NanoDrop spectrophotometer, and the specific activities and t he yields of the cRNAs were calculated; these ranged from 10.22 to 12.38 pmol Cy3/g RNA and from 5.6 to 8.9 g, respectively. The labeled cRNA was stored at 80 o C until use. Microarray Hybridization This was performed following protocols from Agilent, but briefly 600 ng of each labeled cRNA was fragmented and then mixed with hybridization buffer using the Agilent gene expression hybridization kit. These were applied to a sheep 8 X 15 K array slide (Agilent 019921), containing 8 arrays with 15,208 oligom ers with a length of 60 bases and hybridized at 65 o C for 17 h at 10 rpm. The arrays were washed, dried, stabilized, and scanned with an Agilent G2505B 2 dye scanner at the Interdisciplinary Center for Biotechnology Research at the University of Florida. Fe atures were extracted with Agilent Feature extraction 9.1 software. Functional Annotation The Gene Ontology (GO) information was not available for the ovine genome. Thus, the Blast2Go software (V 4.2.2) (24) a too l for functional annotation of (novel)

PAGE 95

95 sequenc es, was employed as described in Chapter 3 The newly annotated platform was submitted to the Gene Expression Omnibus (GEO) website, a public repository that archive high throughput functional genomic data subm itted by the scientific community. The platform number is GPL14112 (Agilent 019921 Sheep Gene expression microarray 8x15K, G4813A), public on August 3, 2011. Statistical Analysis The limma package was employed to import the raw data into R ( http://www.r project.org ), perform background correction and normalize the data using the quantile normalization method (118) Control probes and low expressed probes were filtered out retaining for further analysis the probes that were at least 10% brighter than the negative controls on at least four arrays. The same package was used for the statistical analysis, applying the empirical Bayes method proposed by Smith (119) This method calculates a moderated t statistic for differential expression for each gene by performing a linear model fit on the data. Then, an empirical Bayes step is applied to moderate the standard errors of the estimated log fold changes and produce more stable estimates, especially when the number of replicates is small. Each treatment group was compared to the control. A gene was considered to be significantly differentially expressed (over or under expressed) if both of the fol lowing conditions were met: 1) the ratio of the normalized intensity of the treatment fetus sample to normalized intensity in the control fetus sample was higher or lower than a 2 fold change (up or down regulation respectively); and 2) differences were co nsidered

PAGE 96

96 Clustering Analysis The network inference and clustering analysis was performed using CytoScape version 2.8.3 (117) through the following plugins: GeneMania, Cluster ONE, and BINGO. GeneMania was used to infer network data (140) The set of functional association data between genes was downloaded from the Homo sapiens database. The list of human official symbols for the ge nes of interest was input into the GeneMania plugin to retrieve the corresponding association network. The network was inferred for the up regulated and down regulated genes for each of the treatments compared to the control. These networks were merged us ing the Advance Network Merge plugin, included in the Cytoscape core, in order to detect overlapped genes between networks. Four types of merges were performed 1) between the networks composed of up regulated genes for each treatment; 2) between networks c omposed of down regulated genes for each treatment, 3) between networks inferred with up regulated genes for the estradiol treatment and down regulated genes for ICI (both doses) treatments and 4) between networks containing down regulated genes for the es tradiol treatment and up regulated genes for ICI (both doses) treatments. ClusterONE was used to discover densely connected and possibly overlapping regions (clusters) within sub networks (5) A p value is calcula ted for each cluster, based on the one sided Mann Whitney U test performed on the in weights and out weights of the vertices. An in weights value significantly larger than the out weights value would indicate a valid cluster and not the result of random fl uctuations. Thus, a p value is assigned to the cluster. Only the clusters with a p value less than 0.05 were considered in further analyses.

PAGE 97

97 BiNGO was run to determine which biological processes are statistically overrepresented in the set of genes corres ponding to the identified cluster (81) The statistical test employed was the hypergeometric test (equivalent to the Fisher test). The threshold p value was 0.05, after correction by the Bonferroni method. Quantitati ve Real Time (qRT ) PCR Validation The mRNA samples extracted from the 18 hypothalami samples were converted to cDNA with a High Capacity cDNA Archive kit using the methodology recommended by the kit manufacturer (Applied Biosystems, Foster City, Calif., US A). The newly synthesized cDNA was stored at 20 C until qRT PCR was performed. A total of 5 selected genes from the significan t clusters identified in the up regulated and down regulated networks were tested by qRT PCR to validate the microarray results. The up regulated genes were: vascular endothelial growth factor A (VEGFA), jagged 1 (JAG1) and signal transducer and activator of transcription 5B (STAT5b). The down regulated genes were: ATP synthase, H+ transporting, mitochondrial Fo complex, subunit C1 ( ATP5G1 ) and subunit d ( ATP5H) Relative expression of VEGFA w as determined by qRT PCR using FAM Taqman probes and primers (Sigma Aldrich, St Louis, MO) and Taqman qRT PCR master mix (Applied Biosystems, Foster City, CA). Relative expression of JAG1, STAT 5b, ATP5G1 and ATP5H w as determined using primers (Sigma Aldrich, St Louis, MO) and Syb r Green PCR Master Mix (Applied Biosystems, Foster City, CA). P rimers and probe were designed with Primer Express software (Applied Biosystems). Primers and probe for VEGFA and primers for ATP5G1 and ATP5H were designed from the corresponding ovine mRNA. Primers for JAG1 and STAT5b were

PAGE 98

98 designed from the bovine mRNA Primers s equences and accession numbers are reported in Table 4 1. All primer pairs had efficiencies gre ater than 95%. actin mRNA was determined in each sample, using primers actin sequence and Taqman RT PCR master mix (Applied Biosystems, Foster City, CA). All samples were run in triplicate actin. Relative mRNA actin mRNA from the same sample. The expression of each gene was com pared by one values. Data are graphed as the mean fold change in mRNA relative to the control group; fold change in each sample was calculated as 2 he control group. For all statistical analyses, the criterion for achieving statistical significance was P< 0.05. Results Plasma Hormones As expected, treatment with estradiol caused a significant increase in plasma levels of estradiol at the second and t hird day of administration compared to plasma levels of these hormones in the other treatment group (Figure 4 1A). Treatment with estradiol caused at significant increase in plasma ACTH levels at the third day of administration (Figure 4 1B). The differenc e in plasma ACTH levels was significant for ICI 182 780 5mg/kg on day 1. However, this is a baseline measurement, taken before the drugs were administered. So, this effect cannot be attributed to ICI 182 780 5mg/kg treatment.

PAGE 99

99 Microarray Results The numbe r of up regulated and down regulated genes for each of the treatments was, respectively: estradiol: 349 and 251 genes; ICI 5 mg/kg: 401 and 358 genes and ICI 5 ug/Kg: 457 and 378 genes. Network Analysis The number of nodes in the networks inferred with the up regulated genes was 344 for the estradiol treatment, 388 for the ICI 5 mg/kg treatment and 438 for the ICI 5 ug/kg treatment. For the down regulated genes, the number of nodes in each network was 242 for the estradiol treatment, 343 for the ICI 5 mg /kg treatment and 367 for the ICI 5 ug/kg treatment. The merged networks are shown in figures 4 2 to 4 5. Contrary to the expected, there was a strong overlap for the up regulated genes for each treatment and the same for the down regulated genes for each treatment The overlap was even more evident between the networks resulting from the E2 group and ICI high dose group (Figure 4 4 and 4 5) Thus, the effects of ICI both doses treatments are agonist to the estradiol treatment effects. We interrogated for clusters of genes in the sub network of overlapped genes and not overlapped genes for each of the treatments for the networks composed for up regulated genes for all the treatments (Figure 4 4) and down regulated genes for all the treat ments (Figure 4 5). The overrepresented biological processes for these clusters are summarized in Table 4 2 for the up regulated networks and Table 4 3 for the down regulated networks (redundant biological processes were simplified).

PAGE 100

100 Quantitative Real Time (qRT ) PCR In concordance with the microarray results, expression of VEGFA mRNAs wer e significantly induced by the estradiol and ICI both doses treatment s (p<0.05), while the expression of JAG1 mRNA s were significantly induced by the estradiol and ICI low dose treatments (p<0.05) but showed a tendency of induction by the ICI high dose treatment (p=0.1) (Figure 4 6). There was no significant increase for STAT5b mRNA expression and there were no statistically significant changes in the mRNA expression for th e down regulated genes in the microarray analysis. Discussion plasma ACTH at the third day of infusion, in agreement with previous studies done in our lab (150) Infusion of the ER blocker ICI 182 780 in different doses did not affect ACTH secretion, as also previously found in this lab (111) However, an unexpected finding of this study is th estradiol (Figures 4 4 and 4 5). This effect was not only observed in hypothalamus but also in other tissues collected from this experiment, such us cotyledon and pituitary (data not published). Thus, although the estradiol effect differed, the agonist action was not exclusive of the brain tissue. Figure 4 7 is a Venn Diagram for the number of overlapped genes between the up regulated genes induced by estradiol and ICI both doses treatments, showing tha t there are few common overlapped genes among hypothalamus, pituitary and cotyledon. Despite the high number of same genes regulated in similar way with estradiol or ICI treatment, the biological processes affected with both treatments are also alike

PAGE 101

101 (Tab le 4 2 and Table 4 3 ). Thus, treatment with ICI 182 780, either at low or high dose, agonize the treatment with estradiol, in relation with the genomics effect in the fetal hypothalamus. ER ar affinity remodeling of the classical ER which could impair the dimerization of the ER (16) and nuclear uptake, necessary for ER binding to the DNA (28) However, this mechanism is still controversial. The effect of estradiol and ICI 182 780 on the ER was evaluated on yeast genetic system (34) Yeasts were selected for this study because they lack endogenous nuclear receptors and other receptor co regulatory proteins, characteristics of mammalian cells (137) ICI 182 780 was able to induce ER dimerizatio n and transcriptional activity of the ER and it did not antagonize the effect of estradiol. Agonist up regulation of ER 780 differs with the cell type. Also, other mechanisms specific to mammalian cells can trigger the antiestrogen activity of ICI 182 780, such as differences in compound receptor binding caused by different cytoplasmic components or cytoplasm to nuclear transport systems (34) Nevertheless, the effect of ICI 182 780 in mammalian cells has been shown not to be exclusively estrogen antagonist, at leas t in tissues other than adult reproductive tissues such as breast and uterus. Interestingly, agonist effects of ICI 182 780 have been observed in the brain in several studies in vitro. Thus, ICI 182 780 activates to induce neuronal plasticity and

PAGE 102

102 neuroprotection in fetal rat hippocampal neurons in vitro (158) ICI 182 780 were shown to activate ERK1/2 MAPK signaling in developing and mature cerebellar neurons, and continues exposure to estradiol had anti mitotic effect (145) neurons in vitro, and this effect was not blocked with I CI 182 780 (18) Therefore, these in vitro studies demonstrated estradiol and ICI 182 780, which was one of the over represented biological processes for the up regulated genes by all treatments (Table 4 2 ). Other over represented biological processes listed in Table 4 2 are angiogenesis and cell migration, critical during fetal development. Estrogen has shown to induce angiogenesis in brain tissues (69) or to stimulate actin cytoskeleton remodeling, cell adhesion and cell migration (48) But, there is no report of agonist effect of ICI 182 780 for these actions. Tab le 4 3 details the over represented biological processes for the down regulated genes by all the treatments. Most of these processes are related to the mitochondrial respiratory chain. Estradiol has contradictory effects on the respiratory chain. Several w efficiency and induces the transcription of mitochondrial respiratory chain protein, as reviewed by Chen et al. (21) mitochondrial respiratory activity and oxidative phosphorylation in liver cells, and these effects were not blocked with ICI 182 780 (87) electron transport in hom ogenates of rat uterus, liver, and skeletal muscle (70) tissue (88) and on the ind (1) Down regulation of genes related with the

PAGE 103

103 cell cycle occurs physiologically in the ovine fetal brain during the last stage of gestation (Chaper 5). It is possible that physiologically, ci rculating estradiol decreases the mitochondrial respiratory function as part of adaptive mechanism t hat balance s the increasing oxygen demand with the limited supply though the umbilical artery. In summary, these results has shown that intravenously admin istration of ICI 182 780 in ovine fetuses for 48 hours has agonist genomics effect than the administration of molecular mechanism to explain the agonist effect. However, i t could be hypothesized that in the fetal brain, binding of ICI 182 780 to the ER activates similar downstream such us the transmemb rane G protein coupled receptor GPR30 for which ICI 182 780 exhibits high binding affinity (130) So far, only our lab has published a previous report involving the effect of ICI 182 780 in the ovine fetal brain on late gestation (111) Few genes were evaluated in this study and ICI 182 780 significantly down regulated the expression of PGHS 2 in pituitary and hippocampus compared to estradiol administration. Bu t, the dose and via of administration in that report differed from the actual study (25 ug/day administered intracerebroventricularly ) Therefore, results found in this genomic study suggest a novel and unexplored effect of ICI 1 82 780 in the ovine fetal h ypothalamus when it is administered intravenously and continuously for 48 hrs.

PAGE 104

104 Figure 4 1. Plasma levels of estradiol (A) and ACTH (B) during the infusion period in treated and control fetuses. (*) Plasma values of estradiol were significantly higher for the estradiol treatment group than the other groups at the second and third day of infusion. (*A) Plasma values of ACTH were significantly higher for the ICI 182 780 treatment group than the oth er groups at day one. (B*) Plasma values of ACTH were significantly higher for the estradiol treatment group than the other groups at the third day of infusion.

PAGE 105

105 Figure 4 2. Network resulting from the merge of the network composed for up regulated gene s for estradiol (top) with the networks composed for down regulated genes for ICI 182 780 low dose (bottom left) and high dose (bottom right). Blue nodes denote the overlapped genes between the networks.

PAGE 106

106 Figure 4 3. Network resulting from the merge of the network composed for down regulated genes for estradiol (top) with the networks composed for down regulated genes for ICI 182 780 low dose (bottom right) and high dose (bottom left). Blue nodes denote the overlapped genes between the networks.

PAGE 107

107 Figure 4 4. Network resulting from the merge of the network composed for up regulated genes for estradiol (top right) with the networks composed for up regulated genes for ICI 182 780 low dose (bottom) and high dose (top left). Blue nodes denote the overlapped genes between the networks.

PAGE 108

108 Figure 4 5. Network resulting from the merge of the network composed for down regulated genes for estradiol (top right) with the networks composed for down regulated genes for ICI 182 780 low dose (top left) and high dose (bottom). Blue nodes denote the overlapped genes between the networks.

PAGE 109

109 Figure 4 6. Estradiol and ICI both doses treatments induced changes in mRNA expression of VEGFA and JAG1 (p<0.05). ICI high dose treatment te nded to induce changes in mRNA expression of JAG1 (p=0.1). Expression of each actin expression in the same sample. Data are mean SE fold differences relative to control group expression. (*) represents statistically significant d ifference in mRNA expression between treatment and control groups.

PAGE 110

110 Figure 4 7 Venn diagram of overlapped genes between the up regulated genes induced by treatments with estradiol and ICI 182 780 5 ug/kg or 5 mg/kg dose.

PAGE 111

111 Table 4 1. Sequence s of primers and probes for real time PCR analysis Gene forward primer reverse primer probe accession number oVEGFA GCTCTCTTGGGTGCATTGGA TGCAGCCTGGGACCACTT CCTTGCCTTGCTG CTCTACCTTCACCA NM_001025110 bJAG1 CAACACCTTCGACCTCAAAGC GGCCAGGCGAAACTGAAAG Sybr NM_001191178 bSTAT5b GGCTATCTTGGGTTTCGTGAAC AGGTCCCGTCTGGCTTGTT Sybr NM_174617 oATP5G1 TGGCTCGGCGGGAAT AACTTGGCCGCTGTGTCAAT Sybr NM_001009396 oATP5H CCAGAAGGCCGTGGCTAA AGGTGGCTTCTCAGGCAGAGT Sybr NM_001142891

PAGE 112

112 Table 4 2 Statistically over represented biological processes found of the on the significant clusters of subnetworks from the up regulated networks for all treatments Subnetwork Overrepresented biological processes Genes Overlap Regulation of cell migration/locomotion CDH13 LAMA4 PTPRM TIE1 JAG1 IGFBP3 VCL Angiogenesis CDH13 LAMA4 COL4A1 LEPR VEGFA FOXC1 TIE1 JAG1 VEZF1 CDH5 Cell adhesion CDH13 LAMA4 STAT5B NID1 Cell differentiation/development NBN HTATIP2 CADM1 STAT5B NNAT JAG1 ACAT1 SLC7A5 CDH4 PCSK1 PEG10 SLC1A3 PBXIP1 TEAD4 MSI2 BCL6 PATZ1 TIE1 VEZF1 SATB1 COL4A1 PTPRM PLXNB2 LEF1 WWTR1 NOTCH2 LAMA4 VEGFA FOXC1 GGNBP2 Myeloid cell differentiation L3MBTL1 STAT5B GNAS RARA JAG1 Not overlap E strogen G lucose metabolic process GPI LDHA PGAM2 ENO1 Response to external stimulus ALPL LDHA AXL ADNP USF1 PTEN SLIT2 CXCL10 TGFB2 ASGR1 PITPNM1 NBR1 PDGFRB BMP7 CTSH CLOCK ACSL5 Not overlap ICI 5 ug/kg Positive regulation of synaptic transmission CCL2 IFNG OXTR GRIA4 Response to stress CCL2 EDN1 FOS CTGF IFNG THBS1 PTX3 FGF2 ANGPTL4 MAFF SELP HERPUD1 CEBPB FLT1 IL8 SOCS3 SMAD7 NR4A2 DDIT4 S100A12 THBD ADM BTG2 DUSP1 JUN DNAJB1 CD14 Response to oxygen levels FLT1 CCL2 ADM SOCS3 CTGF EDN1 NR4A2 THBS1 ANGPTL4 DDIT4 Angiogenesis SELP FLT1 IL8 SMAD7 SOCS3 EDN1 TIPARP FOXO1 JUNB JMJD6 CTGF JUN THBS1 FGF2 CYR61 Response to progesterone stimulus FOS CCL2 SOCS3 THBS1 JUNB Phosphoinositide 3 kinase cascade IGF1R EDN1 PIK3R1 Not overlap ICI 5 mg/kg No significant over represented biological processes at corrected p value<0.05

PAGE 113

113 Table 4 3 Statistically over represented biological processes found of the on the significant clu sters of subnetworks from the down regulated networks for all treatments Subnetwork Overrepresented biological processes Genes Overlap Oxidative phosphorylation FXN ATP5B ATP5C1 ATP5G1 NDUFS3 ATP5H Mitochondrial transport SLC25A20 ATP5B TIMM17B ATP5C1 ATP5H Energy coupled proton transport, down electrochemical gradient ATP5B ATP5C1 ATP5G1 ATP5H Not overlap Estrogen Oxidative phosphorylation ATP5F1 ATP5A1 NDUFS1 Oxidation reduction GPD1 PECR PHYH BDH1 NDUFS1 ETFA ATP metabolic process ATP5S ATP5F1 ATP5A1 NDUFS1 Not overlap ICI 5ug/kg RNA splicing RALY FUS POLR2E RNPS1 SF3A2 DDX5 XAB2 SART1 PRPF6 SF3B2 DHX38 DDX23 PRPF8 THOC6 GEMIN6 RBM10 CPSF3 PUF60 CPSF1 Cell cycle CDC23 HMG20B ILF3 MBD3 SART1 GPS2 NCAPH MCM7 GSPT1 POLD1 CHTF18 PSMD3 DYNC1H1 SUPT5H Mitotic cell cycle NCAPH FZR1 GSPT1 POLD1 KATNB1 CDC23 PSMD3 DYNC1H1 Not overlap ICI 5mg/kg Electron transport chain CYB561D2 UQCR10 UQCR11 UQCRC1 TXN2 NDUFA13 Mitochondrial ATP synthesis coupled electron transport UQCR10 UQCR11 UQCRC1 Oxidative phosphorylation UQCR10 UQCR11 UQCRC1 ATP6V0B Oxidation reduction CYB561D2 HSD17B10 UQCR10 UQCR11 UQCRC1 TXN2 NDUFA13 SCO2 ALKBH4

PAGE 114

114 CHAPTER 5 TIME COURSE EXPERIMENT WITH A SINGLE CONDITION: GENE CO EXPRESSION ANALYSIS OF THE OVINE FETAL BRAIN ONTOGENY DURING THE LAST STAGE OF GESTATION AND FIRST DAY OF EXTRA UTERINE LIFE Introduction Temporal dynamics of gene expression in the cells that constitute tissues and organs are essential for the functional development of an individual. In the human brain, the major changes in spatio temporal gene expression occur during the fetal life (65) This period is particularly decisive to define physiological changes that can be manifested in the individual after birth, as stated in the Ba r ker Hypothesis (29) So far, only two studies have analyzed the temporal transcriptome of the human brain (65) or human cortex (23) during fetal development. These studies mostly focused on the first half of gestation (around 25 weeks of pregnancy), not on late gestation, or preterm period, probably because the scarce availability of samples from this period. A better understanding of the genomics during late gestation can help to improve our understanding of the physiological changes necessary to prepare the newborn for the extra uterine life. In this study we use the sheep as an animal model to identify co expressed genes in diffe rent regions of the ovine fetal brain, from the period previous to the hypothalamus pituitary adrenal (HPA) axis activation (around 120 days of gestation) to one day of postnatal life. In the ewe, gestation length varies from 142 to 152 days, with an avera ge of 147 days. The ovine fetus is an excellent model to study brain development since the entire gestational equivalent of human brain development occurs in utero (104) Using a newly available ovine array and we ighted gene co expression network analysis (WGCNA), we tested the hypothesis that the resulting products of

PAGE 115

115 genes expressed in a similar pattern through the last stage of gestation in different brain regions are functionally related and could play an impor tant role in the normal fetal development. The purpose of this study is to analyze the ontogeny of global gene expression in cortex, brainstem, hippocampus and hypothalamus in order to identify clusters of co expressed genes showing a pattern of increased/ decreased expression from 80 days of gestation to 1 day of extra uterine life in each brain region and accordingly in the whole brain. Materi als a nd Methods Tissue Collection Tissues were collected from fetuses at 80 (80 d, n =4), 96 100 (100 d, n= 4), 120 (120 d, n =4), 130 (130 d, n =4), and 142 144 (145 d, n= 4) days of gestation and on the first (1 d, n= 4) day after delivery. Each group included one set of twin fetuses. None of the ewes showed any signs of impending labor. For collection of fetal tissues, e wes were killed with 20 ml of Euthasol solution (7.8 g pentobarbital and 1 g phenytoin sodium; Virbac AH, Fort Worth, TX) administered intravenously, the fetus was quickly removed, and the fetal brain was removed. Fetal tissues were rapidly frozen in liqui d nitrogen and stored at 80C. The use of animals in this project was approved by the University of Florida Institutional Animal Care and Use Committee. RNA Extraction a nd Preparation RNA was extracted from the cortex, hippocampus, hypothalamus and brain stem RNA was resuspended in RNAsecure, and stored at 80 o C in aliquots until use. For microarray analysis 20 ug of RNA was DNase treated using the Turbo RNase free

PAGE 116

116 DN ase kit (Ambion, Foster City, CA), the concentration determined with a Nanodrop spectrophotometer (ND 1000, ThermoFisher, Wilmington DE) and the integrity of the RNA was measured using an Agilent Bioanalyzer, 2100 model. One g of the DNase treated RNA was labeled with Cyanine 3 (Cy3) CTP with the Agilent Quick Amp kit (5190 0442, New Castle, DE) according to their methodology, purified with the Qiagen RNeasy kit (Valencia, CA) as shown in the Quick Amp kit protocol except that the microcentrifugation spins were performed at room temperature instead of 4 o C. The resulting labeled cRNA was analyzed with the Nanodrop spectrophotometer, and the specific activities and the yields of the cRNAs were calculat ed. The labeled cRNA was stored at 80 o C until use. Microarray Hybridization This was performed follo wing protocols from Agilent. B riefly 600 ng of each labeled cRNA was fragmented and then mixed with hybridization buffer using the Agilent gene expres sion hybridization kit. These were applied to sheep 8 X 15 K array slides (Agilent 019921), containing 8 arrays with 15,208 oligomers with a length of 60 bases and hybridized at 65 o C for 17 h at 10 rpm. A total of 3 slides (24 arrays) were employed per each brain region ( six gestational ages times four replicates per gestational age) The arrays were washed, dried, stabilized, and scanned with an Agilent G2505B 2 dye scanner at the Interdisciplinary Center for Biotechnology Research at the University of Florida. Features were extracted with Agilent Feature extraction 9.1 software. Microarray Data The limma package was employed to import the raw data into R ( http://www.r project.org ), perform background correction and normalize the data using the quantile

PAGE 117

117 normalization method (118) Control probes and low expressed probes were filtered out, retaining for further analysis the probes that were at least 10% brighter than the negative co ntrols on at least four arrays. Principal Component Analysis Principal component analysis (PCA) was performed in the data from the 4 brain regions to determine relatedness between the 96 samples. The function prcomp from the stats package for R was used to calculate the principal components of the data (135) The first and second principal components were plotted in a 2 D plot using the plot function from the graphics package for R. The 96 data points (gestational age/b rain region) contained in the resulting scatter plot were colored according the brain region or gestational age. Statistical Analysis The Bayesian Estimation of Temporal Regulation (BETR) algorithm was used to identify the differentially expressed genes ( DEG) for each brain region at a False Discovery Rate (FDR) <0.05. The first gestational age (80 days) was considered as baseline measurement and compared to the subsequent time points, to correlate the differential expression between time points. The algor ithm was applied using the BETR package for R software. A detailed explanation of the mathematical model can be found in Aryee et al. (2) This method returns the probabilities of differential expression for each gen e in the data set. Genes with a probability higher than 99.99% were considered as DEG. Supervised Weighted Gene Co expression Network Analysis (WGCNA) The DEG for each brain region was subjected to signed WGCNA. Rows of each data set were collapsed formin g an average expression of the genes with the same

PAGE 118

118 official symbol, in order to obtain unique identifiers for each gene in the working data set. The automatic method was employed for block wise network construction and module detection. The co expression s imilarity was raised to a soft thresholding power a ij =|(1+cor( x i ,x j ))/2| (155) The resulting modules for each network were r elated with the gestational age to identify modules, or clusters, of co expressed genes with increasing or decreasing expression pattern. Gene significance (GS) was defined as the correlation of i th gene with a temporal pattern. Module membership (MM) was defined as the correlation of the i th gene respects its corresponding module (the higher is the MM the more connected is the i th gene with the other genes of the corresponding modules). The correlation coefficient of MM and GS was measure for each modul e, plotting MM versus GS. Higher correlation between MM and GS means that genes that are highly associated with the temporal pattern are also the central elements of the given module (155) Modules that showed t he highest correlation between MM and GS, either with increasing pattern (positive GS) or decreasing pattern (negative GS), were selected for Gene Ontology analysis. To find key genes in modules with positive GS for each network, a gene screening was perfo rmed selecting the genes that showed a GS higher than 0.5 and were between the 50% of mostly connected genes. The resulting lists of genes from each network were compared to identify common key genes for all the brain regions. Consensus WGCNA The common D EG for the 4 brain regions were subjected to this analysis, which consist in the construction of modules consensus modules that are present in each network. For our data, the purpose was to identify modules that were preserved in each

PAGE 119

119 brain region. These modules were related to the gestational age, to determine GS for each consensus module. Enriched biological processes were determined for the modules with positive GS, in other words, modules composed by genes with increased expression pattern. All the an alyses were performed with the WGCNA package for R software. More details about the methodology for WGNA for network construction can be found at: http://labs.geneti cs.ucla.edu/horvath/CoexpressionNetwork/Rpackages/WGCNA/ Gene Ontology Analysis The bioinformatics tool applied for this purpose was WebGestalt ( WEB based GE ne S e T A na L ysis T oolkit). This program is designed for functional genomic studies that generate large number of gene list. Then, it organizes the large set of genes based on common functional features, like GO categories or biochemical pathways (156) The list of human official symbols for the genes composing the top modules for each network was submitted for enrichment biological processes and KEGG pathway analysis, selecting H. Sapiens as the organism of interest. The statistical me thod employed was the hypergeometric test, adjusting the p values for Benjamini & Hochberg method and using the human genome as reference set. Significantly enriched Kegg pathways (p<0.05) were compared to ident if y common enriched pathways between all brai n regions. Quantitative Real Time ( qRT ) PCR Validation The mRNA samples extracted from the four brain regions at the six different gestational ages (4 fetuses/gestational age; 96 samples in total) were converted to cDNA with a High Capacity cDNA Archive kit using the methodology recommended by

PAGE 120

120 the kit manufacturer (Applied Biosystems, Foster City, Calif., USA). The newly synthesized cDNA was stored at 20 C until qRT PCR was performed. A total of 19 genes were selected from the analysis of the four brain regions for further validation by qRT PCR. The cDNA from each brain region was used to measure the mRNA of the following genes: Hypothalamus: prostaglandin endoperoxide synthase 2 (PTGS2) Brainstem: CD34, CD109, CD44, CD5 and CD9. Hippocampus: CD 3 gamma (CD3G), CD3 delta (C3D) and CD3 epsilon (CD3E) Cortex: colony stimulating factor 1 (macrophage) (CSF1), colony stimulating factor 1 receptor (CSF1R), interleukin 34 (IL34), integrin, alpha M (CD11B), CD81, Fc fragment of IgG, low affinity IIb, receptor (F CGR2B), interleukin 10 (IL10), transforming g rowth factor, beta 1 (TGFB), CD24 and myelin basic protein (MBP) Relative expression of selected genes were determined using primers (Sigma Aldrich, St Louis, MO) and Sybr Green PCR Master Mix (Applied Biosyste ms, Foster City, CA). Primers were designed with Primer Express software (Applied Biosystems. Primers for CD109, CD44, CSF1, CSF1R, IL34 CD24 and MBP were designed from the corresponding bovine mRNA. Primers for PTGS2, CD34 (100) CD5, CD9, CD3G, CD3D, CD3E CD81, CD11B, FCGR2B, IL10 and TGFB were designed from the ovine mRNA. Primers s equences and accession numbers are reported in Table 5 1. All primer pairs had efficiencies greater than 95%. The abundance of B actin mRNA was determined in each brain region sample, using primers and VIC Taqman probes designed from the ovine B actin sequence and Taqman qRT PCR master mix (Applied Biosystems, Foster City, CA). All samples were run in triplicate f or each gene and for B actin. Relative mRNA expression of each gene

PAGE 121

121 each gene and the mean Ct for B actin mRNA from the same sample. The effect of gestational age on each Pairwise Data were graphed as the mean fold change in mRNA relative to the 80d group; fold change in each sample was calculated as 2 criterion for achieving statistical significance was P< 0.05. Results Principal Components Analysis Figure 5 1 shows the results from the PCA analysis. Samples colored by br ain region are shown in Figure 5 1A. Visually, samples from cortex, brainstem and hypothalamus tended to be more related than samples from hippocampus, whi ch were more dispersed. Figure 5 1B displays the microarray samples colored by gestational age. The distinction between the gestational ages is not so clear as for brain regions. However, samples from 80 and 100 days of gestational age tended to be grouped in the upper part of the plot. Thus, the first principal component (PC1) appears to separate samples according the brain region, while the second principal component (PC2) seems to align with the gestational age. Statistical Analysis The number of DEG for each brain region is represent ed in the Venn D iagram (Figure 5 2). Cortex and brainstem were the brain regions presenting the larger number of DEG, followed by hypothalamus and finally hippocampus.

PAGE 122

122 Supervised Weighted Gene Co expression Network Analysis WGCNA determines pair wise correlations betwee n gene expression profiles to create modules (clusters) of co expressed genes. These modules can be associated to an external trait to measure the correlation between the gene expression profile and the trait. In this study, a co expression network was cre ated for each brain region (Figure 5 3). Modules for each brain region were associated to gestational age to determine the modules that had a positive correlation, composed by genes with increasing expression pattern, or negative correlation, composed by g enes with decreasing expression pattern. Top modules were defined as those having the highest positive or negative correlation between GS and MM. The increasing and decreasing pattern for the top modules was visually confirmed with a dendogram of the respe ctive module. Figure 5 4 shows the selected plots with the highest positive or negative MM GS correlation and the respecti ve dendograms for each network. Genes within the top modules were interrogated for enriched biological processes and KEGG pathways. Si gnificantly enriched KEGG pathways in each brain region were compared to identify common enriched pathways. Table 5 2 shows the most representative enriched BP (p value<0.05) in the top modules with positive correlation for each network. The top modules wi th negative correlation for all the networks were constantly enriched for cell cycle an M phase of the cell cycle. Table 5 3 details the common KEGG enriched pathways for all brain regions (p value<0.05). Redundant pathways and those enriched in both tempo ral patterns were excluded from this table. Genes with high connectivity are known as hub genes and are expected to be functionally important within the module. Also, genes with high GS should be highly

PAGE 123

123 biologically significant. In this study, genes with high GS are those that tend to increase (or decrease) strongly toward the end of gestation. The 50% upper quartile of highly connected genes with high GS was extracted from all the positive correlated modules for each network. These genes were combined and compared to determine common key genes between the four brain regions. A total of 25 common genes were identified in this analysis, which were part of the main BP found for each network (Table 5 4 ). Consensus WGCNA This analysis determines clusters or mo dules of genes that are preserved in each brain regions. In other words, we can identify modules of genes that have similar temporal expression during the last stage of gestation in each region of the ovine fetal brain. Figure 5 5 shows the network obtaine d from this analysis, the dendogram of the modules with positive correlation with fetal age and the main biological processes that genes in these modules were enriching. These BP were also enriched in different regions of the fetal brain, suggesting that g enes related with that BP are temporally expressed during the last stage of gestation in the whole ovine fetal brain. Modules with negative correlation with fetal age were constantly enriched for cell cycle process. Quantitative Real Time (qRT ) PCR In conc ordance with the microarray results, the mRNA expression of all the validated genes measured by qRT PCR were following a similar temporal expression pattern that the ones observed with the microarrays (Figure 5 8 to 5 12). Discussion This study explores t he temporal gene expression in the ovine fetal brain during the last stage of gestation, a period where the major developmental events take place in order to prepare the newborn for the extrauterine life. We have identified groups of

PAGE 124

124 genes that are tempora lly co expressed in the fetal brain regions and their functional implication. In agreement with other studies, groups of co expressed genes that increase in expression toward the end of gestation are related with gliogenesis, myelination, generation of ene rgy and response to stimulus, while genes that decrease in expression are mainly related to cell cycle. These findings suggests that during the last steps of fetal brain development the cell machinery declines cell division (quiescence) while emphasizes ne uronal developmental and differentiation (23, 63, 97) A main question that rises from these analyses is what genes are responsible for cell cyc le arrest. The present study shows that there is a point during the developmental period where certain groups of genes start increasing in expression while genes related with cell cycle start to decrease in expression, as appreciated in the dendrograms for the modules with highest positive or negative corr elation with fetal age (Figure 5 4). Thus, up regulation of certain genes should be influencing the expression of genes encoding for cell cycle progression. These regulatory genes could be important compon ents of the physiological fetal brain development. To explore on this concept we search for enriched KEGG pathways in the top modules of co expressed genes with increasing temporal profile, and compare the results to find commonly enriched KEGG pathways (T able 5 3 ). One of the common enriched pathways was the sphingolipid metabolism. Sphingolipids, a component of the eukaryotic cell membrane, have important roles in membrane biology and produce bioactive metabolites that regulate cell function (43) Two main effector molecules products of the sphingolipid metabolism are ceramide and sphingosine 1 phosphate (S1P). Ceramide mediates antiproliferative responses, and is

PAGE 125

125 related with cell cycle arrest and apoptosis, while S1P is involved with transformation, angiogenesis and cell motility (96) The temporal expression pattern of genes involved with the sphingolipod metabolism is opposite to the expression pattern of genes inv olved with cell cycle, and the difference in expression is evident at 130 days of gestational age (Figure 5 6). The mirror pattern suggests that products of this metabolism, such as ceramide, could be affecting the expression of cell cycle genes. Different pathways have been proposed to explain ceramide induction of cell cycle arrest, although the complete mechanism is not totally elucidated. One study showed that ceramide induces a concentration dependent inhibition of the cyclin dependent kinase CDK2, wit h less effect over CDC2 and CDK4 (71) These small proteins tightly regulate the progression of the eukaryotic cell cycle by the binding and phosphorylation of cyclins (89) In our stu dy, these three cyclin dependent kinases have a temporal decreasing expression pattern and were part of the top modules negatively correlated with fetal age, together with some members of the cyclin family. Ceramide can also mediate alpha B crystallin (CRY AB) transcription (20) 5 4 ). CRYAB in fact, can induce the turnover of cyclin D1 when complexed with FBX4 (77) The o ther main product of the sphingolipid metabolism is S1P, which binds a G protein coupled receptor and plays numerous important roles in the functional development of neurons and glial cells (15) S1P is able to stimu late proliferation of neuronal progenitor cells via extracellular signal regulated kinase (ERK) (55) and to activate the MAPK pathway in oligodendrocytes (myelinating cell of the central nervous system), regulating their development and differentiation (154) Also, S1P can regulate

PAGE 126

126 glutamate release in the CA3 region of hippocampus and has profound effects on functions such as spatial learning ( 66) Formation of S1P is result of the sphingosine phosphorylation by sphingosine kinases (SPHK). There are two isoforms of SPHK, SPHK1 and SPHK2, which are encoded in different chromosomes. Expression of SPHK1 is associated with cell survival and proli feration (15) and it was suggested as the primary isoform in the mice brain (42) Furthermore, SPHK1 is critical for neuronal development: knockout (KO) ( sphk1 / ) mice exhi bited a deficiency of S1P which severely disturbed neurogenesis, including neural tube closure and angiogenesis and caused embryonic lethality (86) Accordingly, SPHK1 is one of the genes identified as key gene in the modules positively correlated with age (Table 5 4 ). Many stimuli have been reported to activate SPHK1. Hypoxia can increase the mRNA expression and activity of SPHK1, but not SPHK2 (113) Both hypoxia induci ble factors, HIF 1 alpha and HIF 2 alpha, increase SPKH1 transcription by binding the multiple hypoxia inducible factor responsive region of the SPHK1 gene (113) Interes tingly, other stimuli that can increase SPHK1 expression are pregnancy hormones such us progesterone and estradiol (121) SPHK1 mRNA is increased at late pregnancy in the rat myometrium, coincident with an ele vated progesterone/estradiol ratio (114) SPHK1 in turns induces an up regulation of PTGS2 though a Gi independent pathway. Thus, SPHK1 and PTGS2 have similar expression profile in the rat myometrium: they were present at the end of gestation but declined after parturition (121) In our data, PTGS 2 gene was present in the top modules positive correlated with age in the networks from all brain regions, indicating that this gene was strongly incremented in expression by the end of gestation. mRNA

PAGE 127

127 expression measured by RT PCR showed that PTGS2 expression in th e ovine fetal hypothalamus increases toward the end of gestation but decreases after one week of born (Figure 5 8). In breast cancer cell, estradiol 17 beta stimulates SPHK1 activity by dual action: rapid activation through membrane associated estrogen rec eptor (ER) coupled with Gi proteins, and delayed action by transcriptional activity of nuclear ER (126) Evidence that ER is necessary to induce SPHK activation is given by the fact that activation is blocked wit h the ER ant agonist ICI 182 780 (94) Results from this study found that both estrogen and oxygen responsive genes showed a similar temporal expression pattern than genes related with the sphingolipid metabolism (Figu re 5 6). Thus, it is possible that the high levels of estrogen in the fetal brain at the end gestation and the pr ogressive brain hypoxia secondary to the balance between fetal growth (oxygen demand) and a limitation in the ability of the uteroplacental ci rculation to keep up the oxygen delivery induce the transcription of estrogen and hypoxia responsive genes and in turn, stimulate the expression of genes related to the sphingolipid metabolism. Products of this metabolism could be at least in part respons ible for cell cycle arrest and neuronal differentiation in the ovine fetal brain. Another striking result from this study is the development of the immune system together with the neuronal maturation of the ovine fetal brain, since hematopoietic cell line age was one of the commonly enr iched KEGG pathways (Table 5 3) Both estradiol and SPHK1 are important regulators of the immune response. Estrogen is known to be neuroprotective (99) while SPHK1 negatively regula tes neuroinflammation. Challenges with a strong inducer of inflammation, such as lipopolisacaride greatly increase microglia activation and release of pro inflammatory cytokines in sphk1 / mice (51)

PAGE 128

128 Accordingl y, genes related to hematopoiesis increase in expression during fetal age but genes related to immune system activation are strongly induced after birth (Figure 5 7), suggesting a regulated development of the immune system in the fetal brain. Marker genes for progenitor immune cells were found in the top positive correlated module of the ovine fetal brainstem. These markers are CD34, CD109 and CD44 (Figure 5 9 A) and CD5 and CD9 (Figure 5 9 B). CD109 antigen is expressed by a subpopulation of CD34+ HSC and pr ogenitor cells in the fetal and adult bone marrow (91) An isoform of CD44 is expressed on CD34+ cells: HSC and progenitor cells (95) Progenitors cells of the myeloid lineage, such us colony forming unit macrophage/dendritic cells, express the CSF1R for differentiation. The resident myeloid inflammatory cells of the CNS parenchyma are microglial cells. Thus, e xpression of CSF1R seems critical for microglia development. In mice, the absence of CSF 1R (Csf1r / ) results in essentially no macrophages/microglia (<99% depletion) in the embryonic and early post natal brain (37) In the mice, IL34 binds to the CSF 1R receptor with high affinity t o regulate myeloid development and it can substitute for CSF 1 in vivo IL34 is differentially more highly expressed than CSF 1 mRNA in most of the regions of the developing brain (142) These genes were part of the to p positive correlated module in the cortex network (Figure 5 10 A) together with other genes markers of antigen presenting cells, such as CD1D, CD11B, and CD81 (Figure 5 10 B), showing the development of microglia in the fetal CNS parenchyma. As mentioned above, genes related with immune system activation are strongly induced after birth. Accordingly, mRNA expression of genes markers for mature T cells such us the three chains of the CD3 marker (CD3G CD3E CD3D) (Figure 5 11) were strongly incr eased after birth in

PAGE 129

129 the ovine fetal hippocampus. In addition, two CD markers genes that were following a temporal decreasing expression patter n in all br ain regions were CD24 and CD93 CD93 is considered a myeloid cell marker. However, CD93 was expressed on nave T lymphocytes on neonatal umbilical cord blood (61) CD24 is a surface glycoprotein that is strongly expressed by immature B and T lymphocytes, and disappears from T cell surface as T cell mature. Howeve r, activation of T cells resulted in a rapid induction of CD24 expression (60) CD24 mRNA expression measured by RT PCR was remarkably decreased from 80 days of gestation to 1 day of extra uterine life (Figure 5 12A) in the ovine fetal cortex. Thus, it is possible that cells of the immune system in the fetal brain develop certain type of tolerance to avoid an autoimmune reaction of T cells against self proteins that are being produced at a highly increased rate by the end of gestation, such us MBP. The MBP gene was not only part of the top positive correlated modules for all brain regions but also showed a strong increasing mRNA expression when measured by RT PCR in the ovine fetal cortex (Figure 5 12B) In conco rdance with this concept, genes related with immune tolerance were following an increasing expression pattern. The FCGRIIB for example, is an inhibitory receptor on myeloid cells that can regulate T cell tolerance and promotes T regulatory cell induction b y increasing IL10 production (32, 108) Both IL10 and TGFb, another cytokine involved in T regulatory cell induction, showed an increased expression pattern in the cortex ontogeny measure d by microarray and qRT PCR (Figure 5 12C ). Thus, these results suggest a regulated development of the immune system in the ovine fetal brain. Activation of the immune response after birth could be part of the physiological adaptation of the newborn for th e extrauterine life.

PAGE 130

130 In conclusion, this study provides insight into the dramatic changes in gene expression that take place in the brain during the fetal life, especially during the last stage of gestation. Results from the present study propose the pro ducts of sphingolipid pathway as possible mediator s that dictate the dynamics of gene expression in the fetal brain. Also, they suggest a novel and important role of the immune system in the physiological development of the fetal brain. These findings coul d provide the bass for new investigation in this area, in order to understand key regulators of fetal brain development.

PAGE 131

131 A) B) Figure 5 1. Principal component analysis for gene expression on each sample of the data set. Each dot represents a gestational age per ovine brain regions. Dots were colored according the brain regions (A) or gestational age (B).

PAGE 132

132 Figure 5 2. Venn Diagram of the number of differentially expressed genes following a temporal profile during the last stage of gestation in different regions of the ovine fetal brain.

PAGE 133

133 Figure 5 3. Weighted gene co expression network analysis identifies modules of co expressed genes f ollowing an increasing or decreasing expression temporal pattern during last stage of gestation in ovine fetal cortex (A), brainstem (B), hippocampus (C) and hypothalamus (D). Modules are colored according the number of genes contained in the module. For e turquoise if contains the largest number of genes.

PAGE 134

134 Figure 5 4. Module membership versus gene significance plots. Modules composed by highly connected genes with the highest positive or negative correlation with gestation al age and the respective dendogram, identified en network analysis from the ovine fetal cortex (A), brainstem (B), hippocampus (C) and hypothalamus (D). Each plot is colored according the corresponding module.

PAGE 135

135 Figure 5 5. Consensus weighted gene co expression analysis done in the four brain regions of the fetal ovine brain. Dendograms show the gene expression for the 3 modules more positive correlated with fetal age in each brain region, and the main biological processes enriched with these modules. BS: brainstem, Hippo: hippocampus, Hypo: hypothalamus. Modules are colored according the number of genes contained in the module.

PAGE 136

136 Figure 5 6. Trajectories of average expression for genes involved in the sphingolipid metabolism (blue), cell cycle (red), response to hypoxia (green), and response to estrogen (orange), for all brain regions analyzed.

PAGE 137

137 Figure 5 7. Trajectories of average expression for genes involved in the hematopoietic cell lineage (blue) and activation of the immune system (red ) for all brain regions analyzed.

PAGE 138

138 Figure 5 8 Gene expression of PTGS2 measured by microarray at 80, 120, 145 days of gestation and 1 and 7 day s of extra uterine life and corresponding fold changes in mRNA concentration relative to 80 days, measured by qRT PCR in samples from ovine fetal hypothalamus Data are fold differences relative to mean expression at 80d. a different from 80d values; b different from 120d values; c different from 145d values. For all statistical comparisons, P<0.05 was used as the criterion for significance.

PAGE 139

139 Figure 5 9. Gene expression of CD34, CD109, CD44 (A), CD5 and CD9 (B) measured by microarray at 80, 100, 1 20, 130, 145 days of gestation and 1 day of extra uterine life and corresponding fold changes in mRNA concentration relative to 80 days, measured by qRT PCR in samples from ovine fetal brainsteam Data are fold differences relative to mean expression at 8 0d. a different from 80d values; b different from 100d values; c different from 120d values; d different from 130d values; e different from 145d values. For all statistical comparisons, P<0.05 was used as the criterion for significance.

PAGE 140

140 Fi gure 5 10. Gene expression of CSF1R, CSF1, IL34 (A), CD11B, and CD81 (B) measured by microarray at 80, 100, 120, 130, 145 days of gestation and 1 day of extra uterine life and corresponding fold changes in mRNA concentration relative to 80 days, measured b y qRT PCR in samples from ovine fetal cortex. Data are fold differences relative to mean expression at 80d. a different from 80d values; b different from 100d values; c different from 120d values; d different from 130d values; e diff erent from 145d values. For all statistical comparisons, P<0.05 was used as the criterion for significance

PAGE 141

141 Figure 5 11. Gene expression of CD3G CD3D and CD3E measured by microarray at 80, 100, 120, 130, 145 days of gestation and 1 day of extra uterine life an d corresponding fold changes in mRNA concentration relative to 80 days, measured by qRT PCR in samples from ovine fetal hippocampus Data are fold differences relative to mean expression at 80d. a different from 80d values; b different from 100d valu es; c different from 120d values; d different from 130d values; e different from 145d values. For all statistical comparisons, P<0.05 was used as the criterion for significance

PAGE 142

142 Figure 5 12. Gene expression of CD24 (A), MBP (B) and FGRIIB, IL10,TGFB (C) measured by microarray at 80, 100, 120, 130, 145 days of gestation and 1 day of extra uterine life and corresponding fold changes in mRNA concentration relative to 80 days, measure d by qRT PCR in samples from ovine fetal cortex. Data are fold differences relative to mean expression at 80d. a different from 80d values; b different from 100d values; c different from 120d values; d different from 130d values; e different f rom 145d values. For all statistical comparisons, P<0.05 was used as the criterion for significance.

PAGE 143

143 Table 5 1. Sequences of primers and probes for real time PCR analysis Gene forward primer reverse primer accession number oPTGS2 GCACAAATCTGATGTTTGCATTCT CTGGTCCTCGTTCATATCTGCTT NM_001009432 oCD34 TGGAGCCGTGAACTCTTCTGT CACAAAGCTGATGAGGGTAGAAGAC (100) bCD109 CACAAGATGCTTCAGTGTCCATAGT CTGCGCTCTGGAGTTGTAGCT XM_002690022 bCD44 CGGGTTCATAGAAGGGCATGT TGTTCGCAGCACAGATGGA NM_174013 oCD5 CAGTGTGGCTCCTTCCTGAAG TTTGGCCTCCTGGCTTTG NM_173899 oCD9 CTGAAGCCATCGACGAGATCT CCACGGCAATCCCAATACC NM_001114764 oCD3G ATTGCTGGACAGGAAGGAGTTC TGGTAGAGCTGGTCATTGTTCAA X52994 oCD3D ATCGAATGTGCCAGAACTGTGT GGCAATGATGTCGGTGATGA NM_001009382 oCD3E GAGGTGGCCACAATCATCGT TTTCGGCTCTTGCTCCAGTAA NM_001009418 bCSF1 GACTGGAACATTTTCAGCAAGAACT TCAGGCTTGGTCACCACATC NM_174026 bCSF1R ACACAAAACTCGCAATCTCTCAAC TCGAGTTCGAGAGTCAGGACTTT NM_001075403 bIL34 GATTCCTGCGGGACAAGCT CACCCCCTCATAAGGCACACT NM_001100324 oCD11b CTCCCTCTGCTCCGTGGAT TCGCATCTGCTCATAAAAATGG NM_001082593 oCD81 CCTCCTGTATCTGGAGCTTGGA CAATAAGGATGTAGATGCCCACATA NM_001127281 oFCGR2b CTGCAGTGGCTGTTGTTGCT CGGCTGAAATTGGCTTTCTC NM_001139453 oIL10 CCAGGATGGTGACTCGACTAGAC TGGCTCTGCTCTCCCAGAAC NM_001009327 oTGFb CAGTAAGGATAACACGCTTCAAGTG CCGGTTCATGCCGTGAAT NM_001009400 bCD24 GCCCCTCATCCAGCCAAT GACTGGCTGTTGACTGCAGAGT GJ060586 bMBP AAAACCCTGTAGTGCACTTCTTCA CCCTTTCCTTGCGATGGA NM_001206674

PAGE 144

144 Table 5 2. Representative biological processes significantly enriched with the genes forming part of the top modules with highest correlation genes with the highest increasing temporal profile during the last stage of gestation in each of the ovine fetal brain reg ions Cortex Generation of precursor metabolites and energy Response to steroid hormone stimulus Response to hypoxia Angiogenesis Brainstem Steroid biosynthethic process C ellular homeostasis Apoptosis Hippocampus Myelination Gliogenesis Prostaglandin biosynthethic process Hypothalamus Response to steroid hormone stimulus Response to hypoxia Apotosis Angiogenesis

PAGE 145

145 Table 5 3. Significantly enriched KEGG pathways in all ovine fetal brain regions analyzed. Table in the top details the KEGG pathways enriched with genes belonging to top modules with positive correlation with gestational age, while the table in the bottom refers to the KEGG pathways enriched with genes belonging to top modules with negative correlation with fetal age KEGG pathways (positive correlated top modules) Endocytosis Regulation of actin cytoskeleton Insulin signaling pathway Tight junction Sphingolipid metabolism Hematopoietic cell lineage Leukocyte transendothelial migration KEGG pathways (Negative correlated top modules) Ribosome Spliceosome Cell cycle

PAGE 146

146 Table 5 4. Genes highly positive correlated with ovine fetal age and with high connectivity within respective modules in all brain regions. These genes are expected to be functionally important. Most of these key genes were significantly part of the main biological processes enriched with the top positive correlated modules for each network Gene Symbol Gene Name Enriched BP COX7A1 cytochrome c oxidase subunit VIIa polypeptide 1 (muscle) Generation of precurso r s metabolites and energy CYP1A2 cytochrome P450, family 1, subfamily A, polypeptide 2 PYGM phosphorylase, glycogen, muscle PACSIN3 protein kinase C and casein kinase substrate in neurons 3 Regulation of transport SLC1A2 solute carrier family 1 (glial high affinity glutamate transporter), member 2 Glucose transport ERBB3 v erb b2 erythroblastic leukemia viral oncogene homolog 3 (avian) Gliogenesis GFAP glial fibrillary acidic protein NKX6 1 NK6 homeobox 1 MBP myelin basic protein Mielination ACSBG1 acyl CoA synthetase bubblegum family member 1 C16orf5 chromosome 16 open reading frame 5 Response to stimulus SPHK1 sphingosine kinase 1 Regulation of cell growth CRYAB crystallin, alpha B FAM107A family with sequence similarity 107, member A SDC4 syndecan 4 Protein binding INF2 inverted formin, FH2 and WH2 domain containing SLC12A4 solute carrier family 12 (potassium/chloride transporters), member 4 C ell homeostasis ANKH ankylosis, progressive homolog (mouse) Development TMOD1 tropomodulin 1 TM7SF3 transmembrane 7 superfamily member 3 ENPP5 ectonucleotide pyrophosphatase/phosphodiesterase 5 (putative function) TNFAIP6 tumor necrosis factor, alpha induced protein 6 PPIP5K1 histidine acid phosphatase domain containing 2A SPARCL1 SPARC like 1 (hevin) FBXO32 F box protein 32

PAGE 147

147 CHAPTER 6 TIME COURSE EXPERIMENT WITH TWO CONDITIONS: COMPARATIVE GENOMIC ANALYSIS OF THE OVINE FETAL HYPOTHALAMIC AND PITUITARY ONTOGENIES DURING THE LAST STAGE OF GESTATION AND FIRST DAY OF EXTRA UTERINE LIFE Introduction In ruminants, activation of the fetal hypothalamus pituitary adrenal (HPA) axis is critical to initiate the neuroendocrine cascade that culminates in parturition (75) Several stimuli are involved in the mechanism of hypothalamic stimulation (147) which in turn releases peptide hormones to the portal vessels that connects with the anterior pituitary, to regulate the release of pituitary hormones. This action establishes the communica tion between a brain region (hypothalamus) and an endocrine organ (pituitary), which is not part of the brain. Also, the hypothalamus secretes peptide hormones such us oxytocin (OXT) and vasopressin (VP) that are stored and released from the posterior pitu itary. Therefore, the last stage of gestation implicates a sustained development of the fetal hypothalamus pituitary neuroendocrine communication. The ontogeny of gene expression in both regions dictates the physiological maturity of the HPA axis, a key ev ent in the parturition process. Hence, It can be hypothesized that genes that are differentially expressed between both regions during the last stage of gestation are related with neuroendocrine functions involved in the HPA axis activation. The objective of this study was to determine the differentially expressed genes (DEG) in the fetal hypothalamus compared to pituitary, and vice versa, during the last period of gestation. This study also has the aim to validate the microarray technology as a method to identify the correct DEG in fetal tissues.

PAGE 148

1 48 Materials A nd Methods Tissue Collection Tissues were collected from fetuses at 120 (120d, n =4), 130 (130d, n =4), and 142 144 (145d, n= 4) days of gestation and on the first (1d, n= 4) day after delivery. Each group included one set of twin fetuses. None of the ewes showed any signs of impending labor. For collection of fetal tissues, ewes were killed with 20 ml of Euthasol solution (7.8 g pentobarbital and 1 g phenytoin sodium; Virbac AH, Fort Worth, TX) administere d intravenously, the fetus was quickly removed, and the fetal brain was removed. Fetal tissues were rapidly frozen in liquid nitrogen and stored at 80C. The use of animals in this project was approved by the University of Florida Institutional Animal Car e and Use Committee. RNA Extraction a nd Preparation RNA was extracted from the hypothalamus and pituitary using Trizol (Invitrogen, in RNAsecure, and stored at 80 o C in aliquots until use. For microarray analysis 20 ug of these RNAs, were DNase treated using the Turbo RNase free DNase kit (Ambion, Foster City, CA), the concentration determined with a Nanodrop spectrophotometer (ND 1000, ThermoFisher, Wilmington DE) and th e integrity of the RNA was measured using an Agilent Bioanalyzer, 2100 model. The RNA Integrity Number (RIN) value for the RNAs ranged from 6.5 to 8 One g of the DNase treated RNA was labeled with Cyanine 3 (Cy3) CTP with the Agilent Quick Amp kit (5190 0442, New Castle, DE) according to their methodology, purified with the Qiagen RNeasy kit (Valencia, CA) protocol except that the microcentrifugation spins were performed at room temperature

PAGE 149

149 instead of 4 o C. The resulting labeled cRNA was analyzed with the Nanodrop spectrophotometer, and the specific activities and the yields of the cRNAs were calculated. The labeled cRNA was stored at 80 o C until use. Microarray Hybridiza tion This was performed following protocols from Agilent, but briefly 600 ng of each labeled cRNA was fragmented and then mixed with hybridization buffer using the Agilent gene expression hybridization kit. These were applied to sheep 8 X 15 K array slide s (Agilent 019921), containing 8 arrays with 15,208 oligomers with a length of 60 bases and hybridized at 65 o C for 17 h at 10 rpm The arrays were washed, dried, stabilized, and scanned with an Agilent G2505B 2 dye scanner at the Interdisciplinary Center for Biotechnology Research at the University of Florida. Features were extracted with Agilent Feature extraction 9.1 software. Microarray Data The limma package was employed to import the raw data into R ( http://w ww.r project.org ), perform background correction and normalize the data using the quantile normalization method (118) Control probes and low expressed probes were filtered out, retaining for further analysis the probes that were at least 10% brighter than the negative controls on at least four arrays. A total of 6967 of 8487 unique genes were retained. Statistical An al ysis The Bayesian Estimation of Temporal Regulation (BETR) algorithm was used to identify the differentially expressed genes (DEG) between hypothalamus and pituitary at a False Discovery Rate (FDR) <0.05. The method takes the correlations between time poi nts into account and it can be used to make comparisons between two conditions

PAGE 150

150 hypothalamus and pituitary in this case. The algorithm was applied using the BETR package for R software. A detailed explanation of the mathematical model can be found in Aryee et al., (2009) (2) This method returns the probabilities of differential expression for each gene in the data set. Genes with a probability of 100 % were considered as DEG. Clustering Analysis The selected DEG were clustered using an agglomerative hierarchical clustering. On this procedure, each observation begins in a cluster by itself. The two closest clusters are merged to form a new cluster that replace the two old clusters, and this process is repeated until on e cluster is left. The method used to join the clusters was the Fast Ward method (139) where the criterion for choosing the pair of clusters to merge at each step is based on the squared Euclidean distance between points The methodology was implemented with the JMP Genomics 5.0 software. Network Analysis Selected DEG from the clustering analysis were subjected to network inference using the GeneMania plugin (140) The set of functional association data between genes was downloaded from the Homo sapiens database. The list of human official symbols for the genes of interest was input into the GeneMania plugin to retrieve the corresponding a ssociation network. The networks wer e inferred for the DEG showing the highest difference between hypothalamus and pituitary. The BiNGO plugin was employed to determine the statistically overrepresented biological processes on each network (81) The st atistical test employed was the hypergeometric test (equivalent to the Fisher test). The threshold p value was 0.05, after correction by the Bonferroni method.

PAGE 151

151 Quantitative Real Time ( qRT ) PCR Validation The mRNA samples extracted from the four hypothalami and the four pituitaries at each developmental age were converted to cDNA with a High Capacity cDNA Archive kit using the methodology recommended by the kit manufacturer (Applied Biosystems, Foster City, Calif., USA). The newly synthesized cDNA was stored at 20 C until qRT PCR was performed. Four genes were tested by qRT PCR These genes were: glutamate receptor, ionotropic, N methyl D aspartate 1 (GRIN1), glutamate receptor, ionotropic, AMPA 1 (GRIA1), glutamate receptor, ionotropic, AMPA 3 (GRIA3) and glutamate receptor, metabotropic 3 (GRM3). Relative expression of these genes w as determined using primers (Sigma Aldrich, St Louis, MO) and Sybr Green PCR Master Mix (Applied Biosystems, Foster City, CA). P rimers were designed with Primer Express softwar e (Applied Biosystems). Primers for GRIN1, GRIA1 and GRIA3 were designed from the corresponding ovine mRNA. Primers for GRM3 were designed from the bovine mRNA Primers s equences and accession numbers are reported in Table 6 1. All primer pairs had efficie ncies greater than 95%. actin mRNA was determined in each sample, using primers actin sequence and Taqman qRT PCR master mix (Applied Biosystems, Foster City, CA). All samples were run in actin. Relative mRNA actin mRNA from the same sample. The expression of each ge ne was compared by one

PAGE 152

152 values. Data are graphed as the mean fold change in mRNA for each gestational age relative to the pituitary ; fold change in each sample was calculated as 2 where ch hypothalamic the pituitary sample for that gestational age For all statistical analyses, the criterion for achieving statistical significance was P< 0.05. Results Statistical a nd Clustering Analysis As expected, there was a h igh number of significant DEG between hypothalamus and pituitary: 1818 genes out of 6967 (FDR<0.05). The clustering analysis revealed that 913 genes were highly expressed in hypothalamus compared to pituitary and 905 genes were highly expressed in pituitar y compared to hypothalamus (Figure 6 1). Clusters of genes with the highest expression differences between both organs were selected for further analysis. These clusters contained 237 genes for the highest expression in hypothalamus (red box, Figure 6 1) a nd 78 genes for the highest expression in pituitary (blue box, Figure 6 1). Network a nd Gene Ontology Analysis The network inferred with the genes with the highest expression in hypothalamus compared to pituitary contained 232 nodes (Figure 6 2). Emphasized genes in this network were the hypothalamic hormones: gonadotropin releasing hormone (GNRH1), growth hormone releasing hormone (GHRH), somatostatin (SST), OXT, neuropeptide Y (NPY) and agouti related protein (AGRP). Unfortunately, the gene for CRH, critical for the parturition period, is lacking in the Agilent array. Thus, we cannot determine the expression of th is gene with the current microarray, reflecting one of the disadvantages of the microarray technique. However, one gene of this network was the corticotropin

PAGE 153

153 releasing hormone binding protein (CRHBP), which binds CRH. The network also contained genes encod ing for intermediates of the prostaglandin metabolism, such us prostaglandin endoperoxide synthase 2 (PTGS2) and genes encoding for the predominant excitatory neurotransmitter receptors in the mammalian brain: ionotropic glutamate receptors, N methyl D asp artate 1 (GRIN1) AMPA 1 (GRIA1), AMPA 2 (GRIA2) AMPA3 (GRIA3) and AMP4 (GRIA4) and metabotropic glutamate receptors 3 (GRM3) and 8 (GRM8). The network inferred with the genes with the highest expression in pituitary compared to hypothalamus contained 84 nodes (Figure 6 3). Emphasized genes in this network were the pituitary hormones: prolactin (PRL), growth hormone 1 (GH1), and proopiomelanocortin (POMC), the precursor of adrenocorticotropin (ACTH), the main player of the HPA axis. Other nodes in this net work were the genes encoding for the beta subunit of the glycoprotein hormones: thyroid stimulating hormone, beta (TSHB) luteinizing hormone beta polypeptide (LHB) and follicle stimulating hormone beta polypeptide (FSHB); and the gene encoding for the comm on alpha subunit for these glycoprotein hormones (CGA). Also, this network contained genes encoding for the receptors of the hypothalamic hormones, such us thyrotropin releasing hormone receptor (TRH), growth hormone releasing hormone receptor (GHRHR) and gonadotropin releasing hormone receptor (GNRHR). Table 6 2 details the main biological processes enriched with all the genes in each network. Quantitative Real Time (qRT ) PCR In concordance with the microarray results, fold changes in mRNA concentration o f GRIN1, GRIA1, GRIA3 and GRM3 mRNAs were significantly higher (p<0 05) in the fetal hypothalamus relative to pituitary at all the gestational ages studied (Figure 6 4 ).

PAGE 154

154 The fold change was dramatically higher for GRIN1 at all ages than for the other gluta mate receptors (Figure 6 4 A ). Discussion The present genomic study found that 26% (1818/6967) of the evaluated genes were significantly DEG (FDR< 0.05) between hypothalamus and pituitary, main components of the HPA axis. The large number of DEG was expected, since both organs have different origins. The hypothalamus is part of the brain while the pituitary is an endocrine gland. Notably, the DEG tha t showed the highest differences in expression between both organs were those encoding for important factors of the neuroendocrine communication, including components of the HPA axis activation. Thus, these results demonstrate that while these functional g enes are increasingly expressed in one of the organs, the same genes are not expressed in the other organ, emphasizing that there is precise regulation of gene expression in specific tissues. For example, genes that were highly expressed in hypothalamus bu t down regulated in pituitary were those encoding for hypothalamic releasing hormones, which travel to pituitary to induce the expression of genes related with the corresponding pituitary hormones. Accordingly, genes that were highly expressed genes in pit uitary and minimally or not expressed in hypothalamus were receptors for the hypothalamic hormones and pituitary hormones. Genes highly expressed in the hypothalamus were those encoding for the glutamate receptors. Glutamate receptors are classified as io notropic or metabotropic receptor s Ionotropic glutamate receptors are coupled to ion channels, which allow flux of ions across the pre or postsynaptic membranes. These receptors are further classified according to agonist selectivity and amino acid homol ogy into N methyl D

PAGE 155

155 aspartate receptor (NMDA), and alpha amino 3 hydroxy 5 methyl 4 isoxazole propionate (AMPA) Metabotropic glutamate receptors are coupled to a G protein activating second messengers (57) Synaptic glutamate receptors are found in magnocellular neurons of the supraoptic nucleus of the hypothalamus, and glutamate transmission of the ionotropic receptors mediates the OXT and VP release from the magnocellular neurons (98) Interestingly, glutamatergic synapses of the NMDA receptor (GRIN1) is involved in the fetal ACTH release (93) Accordingly, the mRNA fold change for GRIN1, measured with RT PCR, was fro m 700 times greater at 120 days of gestation to more than 2000 times greater at 145 days relative to pituitary (Figure 6 4A) A previous study in our lab showed that blockade of NMDA receptors inhibits ACTH release in response to bradycardia during cerebra l hypoperfusion (101) demonstrating that NMDA receptors are involved in the afferent signaling from the cardiovascular centers of the medulla and the ACTH release from the fetal hypothalamus. Our lab also found th at NMDA receptor mediated glutamatergic neurotransmission stimulates PTGS2 mediated prostanoid generation (68) The gene for this enzyme was part of the highly expressed genes in hypothalamus (Figure 6 2). Presence of PTGS2 has been localized within hypothalamic regions known to be important for controlling the activity of the HPA axis (13) Results from our lab showed that b lockade of PTGS 2 chronically in the fetal brain reduces fetal HPA axis activity and delays parturition but does not completely abolish fetal HPA axis activity (47, 146) In summary, the application of bioinformatics tools to the results of this genomic time course study comparing ovine fetal hypothalamus and pituitary allowed a global identification of the main players related with the maturity and activity of the fetal HP A

PAGE 156

156 axis close to parturition. Interestingly, most of the hypothalamic genes detected in the present study were not recognized as DEG in the study from Chapter 5. In Chapter 5, the DEG were selected as those showing a correlated differential expression betwe en time points compared to the first baseline measurement (within a single condition: each brain region). In the present study, the DEG were selected as those showing a correlated differential expression between time points but comparing two conditions (hy pothalamus and pituitary). Therefore, it is possible that genes for glutamate receptors, for example, are not changing dramatically in expression from 80 days of fetal age to one day of life but, they are indeed strongly expressed in the hypothalamus compa red to the expression in the pituitary from 120 days of fetal age to one day of life. Thus, even when the results found in this study were obvious, the take home message of the present Chapter is that the methodology employed to analyze the huge amount of microarray data allowed the identification of the correct DEG in the fetal tissues.

PAGE 157

157 Figure 6 1. Hierarchical clustering of differentially expressed genes between hypothalamus and pituitary. Clusters of genes with the highest expression for hypothalamu s (red box) or pituitary (blue box) compared to each other were selected for further analysis.

PAGE 158

158 Figure 6 2. Network inferred with the genes with the highest expression in hypothalamus compared to pituitary. The network contains genes encoding f or hypothalamic hormones (red nodes), prostaglandin metabolism (green nodes) and glutamate receptors (light blue nodes).

PAGE 159

159 Figure 6 3 Network inferred with the genes with the highest expression in pituitary compared to hypothalamus. The network contain s genes encoding for pituitary hormones (red nodes) and hypothalamic receptors ( pink nodes).

PAGE 160

160 Figure 6 4. Fold change in mRNA concentrations at 120, 130, 145 days of gestational life and 1 day of extrauterine life in the fetal hypothalamus relative to the fetal pituitary for GRIN1 (A); GRIA1 (B); GRIA3 (C) and GRM3 (D). (*) Statistically significant difference in mRNA expression between hypothalamus and pituitary at each fetal age (p<0.05).

PAGE 161

161 Table 6 1. Sequences of primers and probes for real time PCR analysis Gene forward primer reverse primer accession number oGRIN1 CAAGAGCATCCACCTGAGCTT TAGACGCGCATCATCTCGAA AY434689 oGRIA1 GAAGCAAGGACTCCGGAAGTAA CCTCCGATCAGGATGTAGAACAC AY346122 oGRIA3 GCGGAAGTCCAAGGGAAAGT GGTTTTCTCTGCTCAATGTACTCATT AY346124 bGRM3 GCCCTGCTGACCAAGACAAA GGCCTCTGAGCGCCATT NM_001098123

PAGE 162

162 Table 6 2. Statistically over represented biological processes found of the on the networks inferred with genes with the highest expression in hypothalamus or pituitary compared to each other N etwork Overrepresented biological processes Genes Hypothalamus Steroid biosynthetic process TM7SF2 APOA1 MVD SQLE CYP51A1 DHCR7 FDPS HMGCS1 HSD17B6 LSS SC4MOL FDFT1 Response to steroid hormone stimulus PLAT ARSB TF GNRH1 PTGS2 CRYAB OXT MAOB MFGE8 IL10 ACSBG1 CD38 AGT ENO2 PDGFRA CA4 IL1B SST Response to corticosteroid stimulus PLAT GNRH1 PTGS2 AGT OXT MAOB IL1B IL10 ACSBG1 Nervous system development ARSB LPPR4 TUBB2B APC2 ERBB3 TUBB2A NCS1 GJA1 AQP4 HP CNP GPM6B CLDN11 SOX9 RTN1 MBP ACSBG1 CKB SEMA5B SLC1A3 APOE BCL11B AGT PTN NDRG2 TUBB3 PLP1 STMN2 GRIN1 NKX6 1 YWHAG S100B GHRH FYN NTRK2 PHGDH PDGFRA ID4 SEMA4D FABP7 CHL1 Neurogenesis LPPR4 APC2 TU BB2B ERBB3 TUBB2A NCS1 GJA1 CNP SOX9 RTN1 MBP SLC1A3 APOE BCL11B AGT TUBB3 PLP1 STMN2 GRIN1 NKX6 1 YWHAG S100B FYN PDGFRA ID4 SEMA4D Transmission of nerve impulse PLAT SYT1 PLP1 OXT GRIN1 CNP CLDN11 MBP ACSBG1 SLC1A4 GRM3 SLC1A3 NPY GRIA2 APOE GRM8 GRIA1 SST Regulation of cell migration TF CORO1A ENPP2 APOE RRAS2 AGT PDGFRA HP SST Axon ensheathment PLP1 CLDN11 MBP ACSBG1 Synaptic transmission, glutamatergic SLC1A4 PLAT GRM8 Glutamate signaling pathway GRIN1 GRIA3 GRIA4 Anti apoptosis PRKCZ TF PEA15 GNRH1 CRYAB APOE IL1B SEMA4D IL10 Feeding behavior NPY FYN AGT OXT AGRP Pituitary Endocrine system development CGA CREB1 NEUROD1 PBX1 CDH1 GHRHR Pituitary gland development CREB1 CDH1 GHRHR Response to peptide hormone stimulus GH1 PTPN2 MC4R FBP1 GHRHR TEC SCAP Regulation of hormone levels CGA MC4R NEUROD1 FSHB LHB GHRHR C21 steroid hormone biosynthetic process FSHB LHB

PAGE 163

163 CHAPTER 7 GENERAL DISCUSSION The series of studies performed in this dissertation have helped to improve our understand ing of the dynamics of gene expression in the fetal brain during the last stage of gestation. Also, they have shown that the microarray is a powerful technology suitable to measure the expression of several genes in fe tal tissues. Therefore, there are two main points to discuss in this section: the application of the microarray technology and the biological interpretation of the results. Application of t he Microarray Technology Several considerations should be made when microarray is applied to a given experiment. The first issue to decide is the software and the methodology to use in order to analyze the microarray data. The studies from this thesis were analyzed with JMP Genomics or the R software. Results from t he employment of one or other software have been equally satisfactory. The advantages of JMP Genomics are that the complete microarray data analysis can be done with point and click applications and it allows the interact ion between graphics and data. For example, in Chapter 6, clusters of genes that were highly expressed in one region versus the other (blue and red box, Figure 6 1), were selected by clicking over the cluster in the graphic of the hierarchical clustering, which highlights the corresponding rows in the input dataset. This type of graphic data interaction cannot be done in R, at least with the packages used for microarray analysis. However, a main advantage of the R software is that it can be extended with pa ckages added by other authors. Most of the packages related to microarray data analysis have the corresponding reference thus it is possible to familiarize with the method and decide if it is appropriate to analyze the

PAGE 164

164 data. For example, the statistical me thod employed to compare the two groups on Chapter 3 was one way ANOVA using empirical Bayes method to shrink the variances, which results in more stable variance component estimates (36) This approach was run with the JMP Genomics software. One way ANOVA t test identifies diff erentially expressed genes (DEG) associated with biological or experimental conditions but assumes that each gene is independent from the other. However, gene expression levels are often non n ormally distributed and have non identical and dependent distributions between genes (119) One approach that deals with these concepts is the moderated t statistics, which pool the global information from all genes r ather than estimating within group variability for each gen e (119) One advantage of this method is that large t statistics are less likely to arise merely f rom under estimated sample vari ances that can occur with small sample size. In other words, diminished the possibility of false positives in experiments with small sample size. Thus, the em pirical Bayes moderation is especially useful in cases with few replicates This method can be applied using the limma package for R (118) and it was used to compare the experimental groups in Chapter 4. Another package from the R software was employed to analyze the time course studies from Chapters 5 and 6. In this case t he package selected was betr, or bayesian estimation of temporal regulation (2) This algorithm uses the time dependent structure of the data and employs an empirical Bayes procedure to stabilize estimates derived fr om the small sample sizes. Therefore, the experiments in Chapter s 4, 5 and 6 were analyzed using the R software, because it provided more appropriate methods to statistically analyze the microarray data.

PAGE 165

165 Once the statistical method is chosen, the second i ssue to deal is the effect of unlikely to be observed in the data for one test becomes more likely to be observed in at least one of the parallel tests (50) Thus, p values need to be adjusted with methods such as FDR in order to have sensible error rate control f or all of the tests in parallel. However, adjusting the p values could lead to have very few or none significant genes, since the adjusting method reduces the power available t o detect changes in expression for individual genes (44) This effect is more evident when the number of replicates per treatment group is low. Thus, if we expect 5% of the genes to be differentially expressed at a fold change higher than 2 (around 700 genes for the ovine Agilent array) we need 10 replicates per group to detect them at a FDR<0.05, with a power of 80% (Figure 2 3) (79) Depending on the type of experiment this high sample size could be difficult to achieve. For example, the in vivo experiments performed in our lab consist in the in vivo catheterization of sheep fetuses, which involves high cost per animal in pregnant ew es. Therefore, the number of replicates per treatment group needs to be lower than 10. For example, the experiments done in this thesis have a sample size of 4. Then, the amount of DEG detected at FDR <0.05 depends on the degree of change in expression cau sed by the treatment or physiological condition. In Chapters 5 and 6 the DEG were those changing in a temporal fashion towards the end of gestation. As the rate of gene expression changes during fetal life is particularly high, the significant DEG were sel ected according the adjusted p value (FDR<0.05). However, the treatments performed in Chapters 3 and 4 are not expected to induce such strong variation in the expression of the genome. Thus, the significant DEG were selected according the p value (p value< 0.05) and the fold change (>2), since the

PAGE 166

166 adjusted p value (FDR<0.05) returned none significant DEG. The findings of both experiments suggest that the treatments were indeed inducing gene expression, as confirmed with RT PCR (Figures 3 4 and 3 5). Therefor e, the use of not adjusted p value is valid if the sample size is lower than the required to detect DEG and the rate of change in gene expression is not too strong. Validation of the results with another method (real time PCR in the present experiments) wi ll confirm then if some of those genes were truly affected. The last issue to consider with the microarray analysis is the handling of the relative large number of DEG, which could be hundreds. A productive approach is to group these genes with some cluste ring method, and then interrogate for enriched biological processes or pathways in that group of genes. There are several methods that can be used to group or classify the genes. The methods employed in this thesis were hierarchical clustering, Cytoscape/C lusterOne plugin and weighted gene co expression network analysis (WGCNA). Hierarchical clustering is a useful method to classify the genes according the degree of similarity between the profiles. Also, it can be experimental replicates within the same condition. Ideally, profiles of samples from the same experimental condition should be closer to each other than they are to profiles of samples from the other conditions. This method was applied to the data analyzed in Chapter 6 and it was very effective to group the rows and columns of the data according the gene expression and biological condition, respectively (Figure 6 1). Cytoscape is an open source platform that allows the visualization of complex networks, and it is exten ded t h rough plugins that allow the addition of features to this bioinformatics software (117) GeneMania is a Cytoscape

PAGE 167

167 plugin that infers networks according the relationship between the input genes. The plugin us es a very large set of functional association data that include protein and genetic interactions, pathways, co expression, co localization and protein domain similarity (140) The clusters in the resu lting network are found with the help of ClusterOne (5) This plugin determine clusters according the physical relation of the gene in the network. Thus, the clusters rely on the network structure. This methodolog y was used in Chapters 3 and 4 inferring separate networks for the up or down regulated genes. However, the methodology was not appropriate to cluster the genes on Chapter 5, since the aim was to group the genes according their change in expression across the time. An approach that was useful in this situation was WGCNA (155) WGCNA searches pair wise correlations between gene expression profi les and associates the resulting clusters to external quantitative data (trait). Therefore, in Chapter 5 the clusters of co expressed DEG were related to the gestational age, which allowed the identification of group of genes following a similar increasing or decreasing pattern toward the end of gestation. Once that the genes have been organized or clustered, the enriched ontological terms or KEGG pathways can be determined with several databases, such as BiNGO (81) DAVID (59) or WebGelstat (156) These mentioned databases are just examples, since the number of available tools to analyze the DEG is quite large, and they can be found at http://www.geneontology. org Therefore data mining the process of analyzing data from different perspectives and summarizing it into useful biological information is probably what takes longer in a microarray experiment.

PAGE 168

168 Biological Interpretation of t he Results Chapters 3 and 4 describe in vivo experiments to evaluate the genomics effect of estradiol 3 sulfate (predominant circulating form in fetal plasma) or estradiol infusion, respectively, in the ovine fetal hypothalamus. The main result of Chapter 3 was that estradiol 3 su lfate caused a strong induction of genes related to feeding behavior, such as neuropeptide Y (NPY), agouti related protein (AGRP) and ghrelin (GHRL). The increased appetite induced by these neuropeptides could be an important component for the survival of the newborn, and could have an effect on the regulation of energy balance regulation before and after birth. The main unexpected result of Chapter 4 was to reveal an agonist genomic effect of estradiol and ICI 182 780 in the ovine fetal hypothalamus. ICI 1 82 780 is classified as a pure anti estrogen compound (58) A possible explanation for this effect is that ICI 182 780 induces the same signaling cascade than estradiol when binds estrogen receptors in fetal tissues Or, ICI 182 780 binds a novel membrane receptor that mimics the action of estrogen binding, such as the transmemb rane G protein coupled receptor GPR30 for which ICI 182 780 exhibits high binding affinity (130) D espite the mechanism of action, the enriched biological process with the list of upregulated genes induced by estradiol/ICI 182 780 were related to estrogenic action (Table 4 2). Some of these processes were enriched with the list of up regulated genes ind uced by estradiol 3 sulfate (Table 3 2), such as angiogenesis, response to oxygen levels or response to stress. Genes related to feeding behavior were not affected by estradiol or ICI 182 780 treatments. Thus, these results might suggest that NPY, AGRP and GHRL are induced by estradiol 3 sulfate independently of estradiol action.

PAGE 169

169 The study described in Chapter 5 consisted of the analysis of gene ontogeny in different regions of the ovine fetal brain during the last stage of gestation Genes showing a diffe rential expression across the time were exposed to WGCNA to find clusters of co expressed genes. Results from this research demonstrated that for all brain regions, genes related to cell cycle decreased in expression while genes related to neuronal differentiation and myelination increased in expression similar to previous reports that studied the temporal dynamics of gene expression in the human brain (23, 65) A probable expla nation for this effect is the increased expression of genes related with the sphingolipid metabolism (Figure 5 6). Two main effector molecules products of the sphingolipid metabolism are ceramide and sphingosine 1 phosphate (S1P). Ceramide mediates antipro liferative responses, and is related with cell cycle arrest and apoptosis, while S1P is involved with neuronal development (96) Furthermore, intermediaries of the sphingolipid metabolism, such as sphingosine kinases (SPHK) are induced by different stimulus such us the pregnancy hormones: estrogen and progesterone (121) or low oxygen levels (113) Both estrogen and oxygen responsive genes showed a similar temporal expression pattern that genes related with the sphingolipid metabolism (Figure 5 6). Therefore, the shingolipid metabolism may play a central role dictating the dynamics of gene expression in the fetal brain. Als o, both estrogen and SPHK1 have shown to be important regulators of the immune response (51, 99) Consistent with this concept, another interesting finding from this study reveals that genes related with hematopoiesis were progressively increasing toward the end of gestation in a regulatory fashion. Thus, genes related with immune tolerance were increasing in expression, while some genes related with T cell activation are decreasing

PAGE 170

170 in expression. Furthermore, genes related with general activation of the immune system are induced right after birth but not during the gestational period (Figure 5 7). The s e result s suggest a novel and important role of the immune system in the physiolog ical development of the fetal brain. Finally, the study described in Chapter 6 identified the genes highly expressed in a temporal manner in hypothalamus compared to pituitary and vice versa during the very last period of gestation and first day of extra u terine life. This study was useful to validate the microarray technology and the followed methodology to analyze the data. As expected, genes related with a mature fetal hypothalamus or pituitary were highly expressed in one region but not expressed in the other, such as hypothalamic hormones for hypothalamus or pituitary hormones and receptors for the hypothalamic hormones for pituitary. In addition, genes encoding for the NMDA receptor (GRIN1) and prosta glandin endoperoxide synthase 2 (PTGS2) were also hi ghly expressed in hypothalamus but not in pituitary. NMDA receptor mediated glutamatergic neurotransmission stimulates PTGS2 mediated prostanoid generation (68) Either blockage of NMDA receptor (101) or PTGS2 (47, 146) declines fetal hypothalamus pituitary adrenal axis activity and ACTH release. Interestingly, DEG identified in this study were not detec ted as DEG in the analysis of the hypothalamus performed in Chapter 5. The difference is that in Chapter 5, DEG were those showing a correlated differential expression between time points compared to the first baseline measurement (80 days) while in Chapte r 6 DEG were those showing a correlated differential expression between time points but comparing two conditions (hypothalamus and pituitary). Therefore, genes involved with the hypothalamus maturity are probably not

PAGE 171

171 changing dramatically in expression fro m 80 days of fetal age to one day of life but they are indeed strongly expressed in the hypothalamus compared to the expression in the pituitary from 120 days of fetal age to one day of life. In conclusion, the studies done for this thesis have analyzed th e expression of thousands of genes in the fetal brain during last gestation. Results have shown that hormonal treatments are able to modify gene expression and maybe influence in the behavior of the animal after born. Also, we have revealed several feature s of the temporal dynamics of the ovine fetal brain transcriptome toward the end of gestation and first day of extra uterine life These findings could provide the basis of new research in this area, aiming at a better understanding of the complexity of fe tal brain development.

PAGE 172

172 LIST OF REFERENCES 1. Alvarez Delgado, C., C. A. Mendoza Rodriguez, O. Picazo, and M. Cerbon 2010. Different expression of alpha and beta mitochondrial estrogen receptors in the aging rat brain: interaction with respirator y complex V. Exp Gerontol 45 :580 585. 2. Aryee, M. J., J. A. Gutierrez Pabello, I. Kramnik, T. Maiti, and J. Quackenbush 2009. An improved empirical bayes approach to estimating differential gene expression in microarray time course data: BETR (Bayesian E stimation of Temporal Regulation). BMC Bioinformatics 10 :409. 3. Ashburner, M., C. A. Ball, J. A. Blake, D. Botstein, H. Butler, J. M. Cherry, A. P. Davis, K. Dolinski, S. S. Dwight, J. T. Eppig, M. A. Harris, D. P. Hill, L. Issel Tarver, A. Kasarskis, S. Lewis, J. C. Matese, J. E. Richardson, M. Ringwald, G. M. Rubin, and G. Sherlock 2000. Gene ontology: tool for the unification of biology. The Gene Ontology Consortium. Nat Genet 25 :25 29. 4. Auer, P. L., and R. W. Doerge 2010. Statistical design and ana lysis of RNA sequencing data. Genetics 185 :405 416. 5. Bader, G. D., and C. W. Hogue 2003. An automated method for finding molecular complexes in large protein interaction networks. BMC Bioinformatics 4 :2. 6. Barbulovic Nad, I., M. Lucente, Y. Sun, M. Zha ng, A. R. Wheeler, and M. Bussmann 2006. Bio microarray fabrication techniques -a review. Crit Rev Biotechnol 26 :237 259. 7. Belsham, D. D., F. Cai, H. Cui, S. R. Smukler, A. M. Salapatek, and L. Shkreta 2004. Generation of a phenotypic array of hypothal amic neuronal cell models to study complex neuroendocrine disorders. Endocrinology 145 :393 400. 8. Benita, Y., H. Kikuchi, A. D. Smith, M. Q. Zhang, D. C. Chung, and R. J. Xavier 2009. An integrative genomics approach identifies Hypoxia Inducible Factor 1 (HIF 1) target genes that form the core response to hypoxia. Nucleic Acids Res 37 :4587 4602. 9. Benjamini, Y., and Y. Hochberg 1995. Controlling the false discovery rate: a pra ctical and powerful approach to multiple testing. J Roy Statist Soc Ser B 57 :289 300. 10. Bethune, J., F. Wieland, and J. Moelleken 2006. COPI mediated transport. J Membr Biol 211 :65 79. 11. Bolstad, B. M., R. A. Irizarry, M. Astrand, and T. P. Speed 200 3. A comparison of normalization methods for high density oligonucleotide array data based on variance and bias. Bioinformatics 19 :185 193.

PAGE 173

173 12. Bonavera, J. J., M. G. Dube, P. S. Kalra, and S. P. Kalra 1994. Anorectic effects of estrogen may be mediated b y decreased neuropeptide Y release in the hypothalamic paraventricular nucleus. Endocrinology 134 :2367 2370. 13. Breder, C. D., W. L. Smith, A. Raz, J. Masferrer, K. Seibert, P. Needleman, and C. B. Saper 1992. Distribution and characterization of cycloox ygenase immunoreactivity in the ovine brain. J Comp Neurol 322 :409 438. 14. Brunnberg, S., K. Pettersson, E. Rydin, J. Matthews, A. Hanberg, and I. Pongratz 2003. The basic helix loop helix PAS protein ARNT functions as a potent coactivator of estrogen re ceptor dependent transcription. Proc Natl Acad Sci U S A 100 :6517 6522. 15. Bryan, L., T. Kordula, S. Spiegel, and S. Milstien 2008. Regulation and functions of sphingosine kinases in the brain. Biochim Biophys Acta 1781 :459 466. 16. Brzozowski, A. M., A. C. Pike, Z. Dauter, R. E. Hubbard, T. Bonn, O. Engstrom, L. Ohman, G. L. Greene, J. A. Gustafsson, and M. Carlquist 1997. Molecular basis of agonism and antagonism in the oestrogen receptor. Nature 389 :753 758. 17. Buchanan, C. D., V. B. Mahesh, and D. W Brann 2000. Estrogen astrocyte luteinizing hormone releasing hormone signaling: a role for transforming growth factor beta(1). Biol Reprod 62 :1710 1721. 18. Cambiasso, M. J., and H. F. Carrer 2001. Nongenomic mechanism mediates estradiol stimulation of axon growth in male rat hypothalamic neurons in vitro. J. Neurosci. Res. 66 :475 481. 19. Carnegie, J. A., and H. A. Robertson 1978. Conjugated and unconjugated estrogens in fetal and maternal fluids of the pregnant ewe: a possible role for estrone sulfat e during early pregnancy. Biol Reprod 19 :202 211. 20. Chang, Y., A. Abe, and J. A. Shayman 1995. Ceramide formation during heat shock: a potential mediator of alpha B crystallin transcription. Proc Natl Acad Sci U S A 92 :12275 12279. 21. Chen, J. Q., P. R Cammarata, C. P. Baines, and J. D. Yager 2009. Regulation of mitochondrial respiratory chain biogenesis by estrogens/estrogen receptors and physiological, pathological and pharmacological implications. Biochim Biophys Acta 1793 :1540 1570. 22. Cho, J., D Kim, S. Lee, and Y. Lee 2005. Cobalt chloride induced estrogen receptor alpha down regulation involves hypoxia inducible factor 1alpha in MCF 7 human breast cancer cells. Mol Endocrinol 19 :1191 1199.

PAGE 174

174 23. Colantuoni, C., B. K. Lipska, T. Ye, T. M. Hyde, R. Tao, J. T. Leek, E. A. Colantuoni, A. G. Elkahloun, M. M. Herman, D. R. Weinberger, and J. E. Kleinman 2011. Temporal dynamics and genetic control of transcription in the human prefrontal cortex. Nature 478 :519 523. 24. Conesa, A., S. Gotz, J. M. Garci a Gomez, J. Terol, M. Talon, and M. Robles 2005. Blast2GO: a universal tool for annotation, visualization and analysis in functional genomics research. Bioinformatics 21 :3674 3676. 25. Cousins, R., and C. E. Wood 2010. Expression of organic anion transpo rters 1 and 3 in the ovine fetal brain during the latter half of gestation. Neurosci Lett 484 :22 25. 26. Cowley, M. A., N. Pronchuk, W. Fan, D. M. Dinulescu, W. F. Colmers, and R. D. Cone 1999. Integration of NPY, AGRP, and melanocortin signals in the hyp othalamic paraventricular nucleus: evidence of a cellular basis for the adipostat. Neuron 24 :155 163. 27. Crispin, J. M simpleaffy: Very simple high level analysis of Affymetrix data. http://www.bioconductor.org/packages/2.12/bioc/manuals/simpleaffy/man/si mpleaffy.pdf 28. Dauvois, S., R. White, and M. G. Parker 1993. The antiestrogen ICI 182780 disrupts estrogen receptor nucleocytoplasmic shuttling. J Cell Sci 106 :1377 1388. 29. de Boo, H. A., and J. E. Harding 2006. The developmental origins of adult dis ease (Barker) hypothesis. Aust N Z J Obstet Gynaecol 46 :4 14. 30. Deayton, J. M., I. R. Young, and G. D. Thorburn 1993. Early hypophysectomy of sheep fetuses: effects on growth, placental steroidogenesis and prostaglandin production. J Reprod Fertil 97 :513 520. 31. Demougin, P., G. Hardiman, A. Ponton, and S. Tenenbaum Microarray Sample Prep Tech Guide. http://www.genomeweb.com/node/939243#q2 32. Desai, D. D., S. O. Harbers, M. Flores, L. Colonna, M. P. Downie, A. Bergtold, S. Jung, and R. Clynes 200 7. Fc gamma receptor IIB on dendritic cells enforces peripheral tolerance by inhibiting effector T cell responses. J Immunol 178 :6217 6226. 33. Dhandapani, K. M., F. M. Wade, V. B. Mahesh, and D. W. Brann 2005. Astrocyte derived transforming growth factor {beta} mediates the neuroprotective effects of 17{beta} estradiol: involvement of nonclassical genomic signaling pathways. Endocrinology 146 :2749 2759. 34. Dudley, M. W., C. Q. Sheeler, H. Wang, and S. Khan 2000. Activation of the human estrogen receptor by the antiestrogens ICI 182,780 and tamoxifen in

PAGE 175

175 yeast genetic systems: implications for their mechanism of action. Proc Natl Acad Sci U S A 97 :3696 3701. 35. Eckel, J. E., C. Gennings, V. M. Chinchilli, L. D. Burgoon, and T. R. Zacharewski 2004. Empiri cal Bayes gene screening tool for time course or dose response microarray data. Journal of biopharmaceutical statistics 14 :647 670. 36. Efron, B., R. Tibshirani, J. D. Storey, and V. Tusher 2001. Empirical Bayes analysis of a microarray experiment. Journa l of the American Statistical Association 96 :1151 1160. 37. Erblich, B., L. Zhu, A. M. Etgen, K. Dobrenis, and J. W. Pollard 2011. Absence of colony stimulation factor 1 receptor results in loss of microglia, disrupted brain development and olfactory defi cits. PLoS One 6 :e26317. 38. Erich, N 2011. RCol orBrewer: ColorBrewer palettes. http://CRAN.R project.org/package=RColorBrewer 39. Estrada, K. M., S. Pompolo, M. J. Morris, A. J. Tilbrook, and I. J. Clarke 2003. Neuropeptide Y (NPY) delays the oestrogen induced luteinizing hormone (LH) surge in the ovariectomized ewe: further evidence that NPY has a predominant negative effect on LH secretion in the ewe. J Neuroendocrinol 15 :1011 1020. 40. Fodor, S. P., J. L. Read, M. C. Pirrung, L. Stryer, A. T. Lu, and D. Solas 1991. Light directed, spatially addressable parallel chemical synthesis. Science 251 :767 773. 41. France, J. T., R. R. Magness, B. A. Murry, C. R. Rosenfeld, and J. I. Mason 1988. The regulation of ovine placental steroid 17 alpha hydroxylase an d aromatase by glucocorticoid. Mol Endocrinol 2 :193 199. 42. Fukuda, Y., A. Kihara, and Y. Igarashi 2003. Distribution of sphingosine kinase activity in mouse tissues: contribution of SPHK1. Biochem Biophys Res Commun 309 :155 160. 43. Gault, C. R., L. M. Obeid, and Y. A. Hannun 2010. An overview of sphingolipid metabolism: from synthesis to breakdown. Adv Exp Med Biol 688 :1 23. 44. Ge, Y., S. Dudoit, and T. P. Speed 2003. Resampling based multiple testing for microarray data analysis. Test 12 :1 77. 45. Ge, Y., S. Dudoit, and T. P. Speed 2004. Linear models and empirical bayes methods for assessing differential expression in microarray experiments. Stat Appl Genet Mol Biol 3 :Article3.

PAGE 176

176 46. Gentleman, R. C., V. J. Carey, D. M. Bates, B. Bolstad, M. Dettlin g, S. Dudoit, B. Ellis, L. Gautier, Y. Ge, J. Gentry, K. Hornik, T. Hothorn, W. Huber, S. Iacus, R. Irizarry, F. Leisch, C. Li, M. Maechler, A. J. Rossini, G. Sawitzki, C. Smith, G. Smyth, L. Tierney, J. Y. Yang, and J. Zhang 2004. Bioconductor: open soft ware development for computational biology and bioinformatics. Genome Biol 5 :R80. 47. Gersting, J., C. E. Schaub, M. Keller Wood, and C. E. Wood 2008. Inhibition of brain prostaglandin endoperoxide synthase 2 prevents the preparturient increase in fetal a drenocorticotropin secretion in the sheep fetus. Endocrinology 149 :4128 4136. 48. Giretti, M. S., and T. Simoncini 2008. Rapid regulatory actions of sex steroids on cell movement through the actin cytoskeleton. Steroids 73 :895 900. 49. Gorreta, F., W. Car bone, and D. Barzaghi 2012. Genomic profiling: cDNA arrays and oligoarrays. Methods Mol Biol 823 :89 105. 50. Grant, G. R., E. Manduchi, and C. J. J. Stoeckert 2007. Analysis and management of microarray gene expression data. Curr Protoc Mol Biol Chapter 19 :Unit 19.6. 51. Grin'kina, N. M., E. E. Karnabi, D. Damania, S. Wadgaonkar, I. A. Muslimov, and R. Wadgaonkar 2012. Sphingosine kinase 1 deficiency exacerbates LPS induced neuroinflammation. PLoS One 7 :e36475. 52. Gu, Y. Z., J. B. Hogenesch, and C. A. Bradfield 2000. The PAS superfamily: sensors of environmental and developmental signals. Annu Rev Pharmacol Toxicol 40 :519 561. 53. Guo, X., H. Qi, C. M. Verfaillie, and W. Pan 2003. Statistical significance analysis of longitudinal gene expression data. Bioinformatics 19 :1628 1635. 54. Hagberg, H., P. Gressens, and C. Mallard 2012. Inflammation during fetal and neonatal life: implications for neurologic and neuropsychiatric disease in children and adults. Ann Neurol 71 :444 457. 55. Harada, J., M. Foley, M. A. Moskowitz, and C. Waeber 2004. Sphingosine 1 phosphate induces proliferation and morphological changes of neural progenitor cells. J Neurochem 88 :1026 1039. 56. Hartemink, A., D. Gifford, T. S. Jaakkola, and R. A. Young 2001. Maximum likelihood es timation of optimal scaling factors for expression array normalization. p. 132 140. In M. Bittner, Y. Chen, A. Dorsel, and D. Bellingham (eds.), Microarrays: Optical Technologies and Informatics. SPIE International Society for Optical Engineering,

PAGE 177

177 57. Hol lmann, M., and S. Heinemann 1994. Cloned glutamate receptors. Annual review of neuroscience 17 :31 108. 58. Howell, A., C. K. Osborne, C. Morris, and A. E. Wakeling 2000. ICI 182,780 (Faslodex): development of a novel, "pure" antiestrogen. Cancer 89 :817 8 25. 59. Huang da, W., B. T. Sherman, and R. A. Lempicki 2009. Systematic and integrative analysis of large gene lists using DAVID bioinformatics resources. Nat Protoc 4 :44 57. 60. Hubbe, M., and P. Altevogt 1994. Heat stable antigen/CD24 on mouse T lymph ocytes: evidence for a costimulatory function. Eur J Immunol 24 :731 737. 61. Ikewaki, N., H. Yamao, J. K. Kulski, and H. Inoko 2010. Flow cytometric identification of CD93 expression on naive T lymphocytes (CD4(+)CD45RA (+) cells) in human neonatal umbili cal cord blood. J Clin Immunol 30 :723 733. 62. Jean, W., and I. w. c. f. J. M. J. G. Rafael gcrma: Background Adjustm ent Using Sequence Information. http://bioconductor.wustl.edu/bioc/html/gcrma.html 63. Johnson, M. B., Y. I. Kawasawa, C. E. Mason, Z. Krs nik, G. Coppola, D. Bogdanovic, D. H. Geschwind, S. M. Mane, M. W. State, and N. Sestan 2009. Functional and evolutionary insights into human brain development through global transcriptome analysis. Neuron 62 :494 509. 64. Johnson, M. L., A. T. Grazul Bils ka, D. A. Redmer, and L. P. Reynolds 2006. Effects of estradiol 17beta on expression of mRNA for seven angiogenic factors and their receptors in the endometrium of ovariectomized (OVX) ewes. Endocrine 30 :333 342. 65. Kang, H. J., Y. I. Kawasawa, F. Cheng, Y. Zhu, X. Xu, M. Li, A. M. Sousa, M. Pletikos, K. A. Meyer, G. Sedmak, T. Guennel, Y. Shin, M. B. Johnson, Z. Krsnik, S. Mayer, S. Fertuzinhos, S. Umlauf, S. N. Lisgo, A. Vortmeyer, D. R. Weinberger, S. Mane, T. M. Hyde, A. Huttner, M. Reimers, J. E. Kle inman, and N. Sestan 2011. Spatio temporal transcriptome of the human brain. Nature 478 :483 489. 66. Kanno, T., T. Nishizaki, R. L. Proia, T. Kajimoto, S. Jahangeer, T. Okada, and S. Nakamura 2010. Regulation of synaptic strength by sphingosine 1 phospha te in the hippocampus. Neuroscience 171 :973 980. 67. Kazi, A. A., J. M. Jones, and R. D. Koos 2005. Chromatin immunoprecipitation analysis of gene expression in the rat uterus in vivo: estrogen induced recruitment of both estrogen receptor alpha and hypox ia inducible factor 1 to the vascular endothelial growth factor promoter. Mol Endocrinol 19 :2006 2019.

PAGE 178

178 68. Knutson, N., and C. E. Wood 2010. Interaction of PGHS 2 and glutamatergic mechanisms controlling the ovine fetal hypothalamus pituitary adrenal axis Am J Physiol Regul Integr Comp Physiol 299 :R365 70. 69. Krause, D. N., S. P. Duckles, and D. A. Pelligrino 2006. Influence of sex steroid hormones on cerebrovascular function. J Appl Physiol 101 :1252 1261. 70. Kuss, E., and G. Jutting 1967. [Specificit y and localization of the effect of estrogens on electron transport]. Hoppe Seylers Z Physiol Chem 348 :913 920. 71. Lee, J. Y., A. E. Bielawska, and L. M. Obeid 2000. Regulation of cyclin dependent kinase 2 activity by ceramide. Exp Cell Res 261 :303 311. 72. Lee, M. C., E. A. Miller, J. Goldberg, L. Orci, and R. Schekman 2004. Bi directional protein transport between the ER and Golgi. Annu Rev Cell Dev Biol 20 :87 123. 73. Levings, M. K., R. Sangregorio, F. Galbiati, S. Squadrone, R. de Waal Malefyt, and M. G. Roncarolo 2001. IFN alpha and IL 10 induce the differentiation of human type 1 T regulatory cells. J Immunol 166 :5530 5539. 74. Liggins, G. C., R. J. Fairclough, S. A. Grieves, J. Z. Kendall, and B. S. Knox 1973. The mechanism of initiation of parturition in the ewe. Recent Prog Horm Res 29 :111 159. 75. Liggins, G. C., P. C. Kennedy, and L. W. Holm 1967. Failure of initiation of parturition after electrocoagulation of the pituitar y of the fetal lamb. Am J Obstet Gynecol 98 :1080 1086. 76. Lin, C. Y., A. Strom, V. B. Vega, S. L. Kong, A. L. Yeo, J. S. Thomsen, W. C. Chan, B. Doray, D. K. Bangarusamy, A. Ramasamy, L. A. Vergara, S. Tang, A. Chong, V. B. Bajic, L. D. Miller, J. A. Gust afsson, and E. T. Liu 2004. Discovery of estrogen receptor alpha target genes and response elements in breast tumor cells. Genome Biol 5 :R66. 77. Lin, D. I., O. Barbash, K. G. Kumar, J. D. Weber, J. W. Harper, A. J. Klein Szanto, A. Rustgi, S. Y. Fuchs, a nd J. A. Diehl 2006. Phosphorylation dependent ubiquitination of cyclin D1 by the SCF(FBX4 alphaB crystallin) complex. Mol Cell 24 :355 366. 78. Lin, D. Y 2005. An efficient Monte Carlo approach to assessing statistical significance in genomic studies. Bi oinformatics 21 :781 787. 79. Liu, P., and J. T. G. Hwang 2007. Quick calculation for sample size while controlling false discovery rate with application to microarray analysis. Bioinformatics 23 :739 746.

PAGE 179

179 80. Lombardero, M., A. Quintanar Stephano, S. Vidal E. Horvath, K. Kovacs, R. V. Lloyd, and B. W. Scheithauer 2009. Effect of estrogen on the blood supply of pituitary autografts in rats. J Anat 214 :235 244. 81. Maere, S., K. Heymans, and M. Kuiper 2005. BiNGO: a Cytoscape plugin to assess overrepresent ation of gene ontology categories in biological networks. Bioinformatics 21 :3448 3449. 82. Mahesh, V. B., K. M. Dhandapani, and D. W. Brann 2006. Role of astrocytes in reproduction and neuroprotection. Mol Cell Endocrinol 246 :1 9. 83. Malone, J. H., and B Oliver 2011. Microarrays, deep sequencing and the true measure of the transcriptome. BMC Biol 9 :34. 84. McDonald, T. J., and P. W. Nathanielsz 1991. Bilateral destruction of the fetal paraventricular nuclei prolongs gestation in sheep. Am J Obstet Gyne col 165 :764 770. 85. Melcangi, R. C., I. Cavarretta, V. Magnaghi, L. Martini, and M. Galbiati 2001. Interactions between growth factors and steroids in the control of LHRH secreting neurons. Brain Res Brain Res Rev 37 :223 234. 86. Mizugishi, K., T. Yamash ita, A. Olivera, G. F. Miller, S. Spiegel, and R. L. Proia 2005. Essential role for sphingosine kinases in neural and vascular development. Mol Cell Biol 25 :11113 11121. 87. Moreira, P. I., J. Custdio, A. Moreno, C. R. Oliveira, and M. S. Santos 2006. T amoxifen and estradiol interact with the flavin mononucleotide site of complex I leading to mitochondrial failure. J. Biol. Chem. 281 :10143 10152. 88. Moreira, P. I., J. B. A. Custdio, E. Nunes, P. J. Oliveira, A. Moreno, R. Seia, C. R. Oliveira, and M. S. Santos 2011. Mitochondria from distinct estradiol and tamoxifen. J. Steroid Biochem. Mol. Biol. 123 :8 16. 89. Morgan, D. O 1995. Principles of CDK regulation. Nature 374 :131 134. 90. Muhlhausler, B. S., I. C. Mc Millen, G. Rouzaud, P. A. Findlay, E. M. Marrocco, S. M. Rhind, and C. L. Adam 2004. Appetite regulatory neuropeptides are expressed in the sheep hypothalamus before birth. J Neuroendocrinol 16 :502 507. 91. Murray, L. J., E. Bruno, N. Uchida, R. Hoffman, R. Nayar, E. L. Yeo, A. C. Schuh, and D. R. Sutherland 1999. CD109 is expressed on a subpopulation of CD34+ cells enriched in hematopoietic stem and progenitor cells. Exp Hematol 27 :1282 1294.

PAGE 180

180 92. Nagalakshmi, U., Z. Wang, K. Waern, C. Shou, D. Raha, M. G erstein, and M. Snyder 2008. The transcriptional landscape of the yeast genome defined by RNA sequencing. Science 320 :1344 1349. 93. Nardo, L., Y. Soong, D. Wu, I. R. Young, D. Walker, and H. H. Szeto 2002. Site and mechanism of action of dynorphin A (1 13) and N methyl D aspartate on ACTH release in fetal sheep. Am J Physiol Endocrinol Metab 282 :E1301 7. 94. Nava, V. E., J. P. Hobson, S. Murthy, S. Milstien, and S. Spiegel 2002. Sphingosine kinase type 1 promotes estrogen dependent tumorigenesis of brea st cancer MCF 7 cells. Exp Cell Res 281 :115 127. 95. Neu, S., A. Geiselhart, M. Sproll, D. Hahn, S. Kuci, D. Niethammer, and R. Handgretinger 1997. Expression of CD44 isoforms by highly enriched CD34 positive cells in cord blood, bone marrow and leukapher eses. Bone Marrow Transplant 20 :593 598. 96. Ogretmen, B., and Y. A. Hannun 2004. Biologically active sphingolipids in cancer pathogenesis and treatment. Nat Rev Cancer 4 :604 616. 97. Oldham, M. C., G. Konopka, K. Iwamoto, P. Langfelder, T. Kato, S. Horva th, and D. H. Geschwind 2008. Functional organization of the transcriptome in human brain. Nat Neurosci 11 :1271 1282. 98. Pak, C. W., and M. C. Curras Collazo 2002. Expression and plasticity of glutamate receptors in the supraoptic nucleus of the hypotha lamus. Microsc Res Tech 56 :92 100. 99. Papenfuss, T. L., N. D. Powell, M. A. McClain, A. Bedarf, A. Singh, I. E. Gienapp, T. Shawler, and C. C. Whitacre 2011. Estriol generates tolerogenic dendritic cells in vivo that protect against autoimmunity. J Immun ol 186 :3346 3355. 100. Porada, C. D., D. D. Harrison Findik, C. Sanada, V. Valiente, D. Thain, P. J. Simmons, G. Almeida Porada, and E. D. Zanjani 2008. Development and characterization of a novel CD34 monoclonal antibody that identifies sheep hematopoiet ic stem/progenitor cells. Exp Hematol 36 :1739 1749. 101. Powers, M. J., and C. E. Wood 2007. Ketamine inhibits fetal ACTH responses to cerebral hypoperfusion. Am J Physiol Regul Integr Comp Physiol 292 :R1542 9. 102. Purinton, S. C., H. Newman, M. I. Castr o, and C. E. Wood 1999. Ontogeny of estrogen sulfatase activity in ovine fetal hypothalamus, hippocampus, and brain stem. Am J Physiol 276 :R1647 52.

PAGE 181

181 103. Purinton, S. C., and C. E. Wood 2000. Ovine fetal estrogen sulfotransferase in brain regions important for hypothalamus pituitary adrenal axis control. Neuroendocrinology 71 :237 242. 104. Ramadoss, J., E. R. Lunde, W. J. Chen, J. R. West, and T. A. Cudd 2007. Temporal vulnerability of fetal cerebellar Purkinje cells to chronic binge alcohol exposure: ovine model. Alcohol Clin Exp Res 31 :1738 1745. 105. Reimers, M 2010. Making informed choices about microarray data analysis. PLoS Comput Biol 6 :e1000786. 106. Richard, H., M. H. Schulz, M. Sultan, A. Nurnberger, S. Schrinner, D. Balzereit, E. Dagand, A. Rasche, H. Lehrach, M. Vingron, S. A. Haas, and M. L. Yaspo 2010. Prediction of alternative isoforms from exon expression levels in RNA Seq experiments. Nucleic Acids Res 38 :e112. 107. Ru der, H. J., L. Loriaux, and M. B. Lipsett 1972. Estrone sulfate: production rate and metabolism in man. J Clin Invest 51 :1020 1033. 108. Samsom, J. N., L. A. van Berkel, J. M. van Helvoort, W. W. Unger, W. Jansen, T. Thepen, R. E. Mebius, S. S. Verbeek, a nd G. Kraal 2005. Fc gamma RIIB regulates nasal and oral tolerance: a role for dendritic cells. J Immunol 174 :5279 5287. 109. Saoud, C. J., and C. E. Wood 1997. Modulation of ovine fetal adrenocorticotropin secretion by androstenedione and 17beta estradi ol. Am J Physiol 272 :R1128 34. 110. Schaub, C. E., J. A. Gersting, M. Keller Wood, and C. E. Wood 2008. Development of ER alpha and ER beta expression in the developing ovine brain and pituitary. Gene Expr Patterns 8 :457 463. 111. Schaub, C. E., M. Keller Wood, and C. E. Wood 2008. Blockade of estrogen receptors decreases CNS and pituitary prostaglandin synthase expression in fetal sheep. Neuroendocrinology 87 :121 128. 112. Schena, M., D. Shalon, R. W. Davis, and P. O. Brown 1995. Quantitative monitoring of gene expression patterns with a complementary DNA microarray. Science 270 :467 470. 113. Schwalm, S., F. Doll, I. Romer, S. Bubnova, J. Pfeilschifter, and A. Huwiler 2008. Sphingosine kinase 1 is a hypoxia regulated gene that stimulates migration of hu man endothelial cells. Biochem Biophys Res Commun 368 :1020 1025. 114. Serrano Sanchez, M., Z. Tanfin, and D. Leiber 2008. Signaling pathways involved in sphingosine kinase activation and sphingosine 1 phosphate release

PAGE 182

182 in rat myometrium in late pregnancy: role in the induction of cyclooxygenase 2. Endocrinology 149 :4669 4679. 115. Shi, L., L. H. Reid, W. D. Jones, R. Shippy, J. A. Warrington, S. C. Baker, P. J. Collins, F. de Longueville, E. S. Kawasaki, K. Y. Lee, Y. Lu o, Y. A. Sun, J. C. Willey, R. A. Setterquist, G. M. Fischer, W. Tong, Y. P. Dragan, D. J. Dix, F. W. Frueh, F. M. Goodsaid, D. Herman, R. V. Jensen, C. D. Johnson, E. K. Lobenhofer, R. K. Puri, U. Schrf, J. Thierry Mieg, C. Wang, M. Wilson, P. K. Wolber, L. Zhang, S. Amur, W. Bao, C. C. Barbacioru, A. B. Lucas, V. Bertholet, C. Boysen, B. Bromley, D. Brown, A. Brunner, R. Canales, X. M. Cao, T. A. Cebula, J. J. Chen, J. Cheng, T. M. Chu, E. Chudin, J. Corson, J. C. Corton, L. J. Croner, C. Davies, T. S. Da vison, G. Delenstarr, X. Deng, D. Dorris, A. C. Eklund, X. H. Fan, H. Fang, S. Fulmer Smentek, J. C. Fuscoe, K. Gallagher, W. Ge, L. Guo, X. Guo, J. Hager, P. K. Haje, J. Han, T. Han, H. C. Harbottle, S. C. Harris, E. Hatchwell, C. A. Hauser, S. Hester, H. Hong, P. Hurban, S. A. Jackson, H. Ji, C. R. Knight, W. P. Kuo, J. E. LeClerc, S. Levy, Q. Z. Li, C. Liu, Y. Liu, M. J. Lombardi, Y. Ma, S. R. Magnuson, B. Maqsodi, T. McDaniel, N. Mei, O. Myklebost, B. Ning, N. Novoradovskaya, M. S. Orr, T. W. Osborn, A. Papallo, T. A. Patterson, R. G. Perkins, E. H. Peters, R. Peterson, K. L. Philips, P. S. Pine, L. Pusztai, F. Qian, H. Ren, M. Rosen, B. A. Rosenzweig, R. R. Samaha, M. Schena, G. P. Schroth, S. Shchegrova, D. D. Smith, F. Staedtler, Z. Su, H. Sun, Z. Sza llasi, Z. Tezak, D. Thierry Mieg, K. L. Thompson, I. Tikhonova, Y. Turpaz, B. Vallanat, C. Van, S. J. Walker, S. J. Wang, Y. Wang, R. Wolfinger, A. Wong, J. Wu, C. Xiao, Q. Xie, J. Xu, W. Yang, L. Zhang, S. Zhong, Y. Zong, and W. J. Slikker 2006. The Micr oArray Quality Control (MAQC) project shows inter and intraplatform reproducibility of gene expression measurements. Nat Biotechnol 24 :1151 1161. 116. Silva, L. E., M. Castro, F. C. Amaral, J. Antunes Rodrigues, and L. L. Elias 2010. Estradiol induced hy pophagia is associated with the differential mRNA expression of hypothalamic neuropeptides. Braz J Med Biol Res 43 :759 766. 117. Smoot, M. E., K. Ono, J. Ruscheinski, P. L. Wang, and T. Ideker 2011. Cytoscape 2.8: new features for data integration and net work visualization. Bioinformatics 27 :431 432. 118. Smyth, G. K 2005. Limma: linear models for microarray data. p. 397 420. In V. C. R. Gentleman, S. Dudoit, R. Irizarry, W. Huber (ed.), Bioinformatics and Computational Biology Solutions using R and Bioco nductor, R. Springer, New York. 119. Smyth, G. K 2004. Linear models and empirical bayes methods for assessing differential expression in microarray experiments. Stat Appl Genet Mol Biol 3 :Article3. 120. Smyth, G. K., and T. Speed 2003. Normalization of cDNA microarray data. Methods 31 :265 273.

PAGE 183

183 121. Spiegel, S., and S. Milstien 2007. Functions of the multifaceted family of sphingosine kinases and some close relatives. J Biol Chem 282 :2125 2129. 122. Steele, P. A., A. P. Flint, and A. C. Turnbull 1976. A ctivity of steroid C 17,20 lyase in the ovine placenta: effect of exposure to foetal glucocorticoid. J Endocrinol 69 :239 246. 123. Stein, L. D 2004. Human genome: end of the beginning. Nature 431 :915 916. 124. Stolze, I., U. Berchner Pfannschmidt, P. Frei tag, C. Wotzlaw, J. Rossler, S. Frede, H. Acker, and J. Fandrey 2002. Hypoxia inducible erythropoietin gene expression in human neuroblastoma cells. Blood 100 :2623 2628. 125. Storey, J. D 2003. The positive false discovery rate: A Bayesian interpretation and the q value. Annals of Statistics 2013 2035. 126. Sukocheva, O., C. Wadham, and P. Xia 2009. Role of sphingolipids in the cytoplasmic signaling of estrogens. Steroids 74 :562 567. 127. Tai, Y. C., and T. P. Speed 2009. On gene ranking using replicate d microarray time course data. Biometrics 65 :40 51. 128. Tang, Y., S. Ghosal, and A. Roy 2007. Nonparametric bayesian estimation of positive false discovery rates. Biometrics 63 :1126 1134. 129. Theocharidis, A., S. van Dongen, A. J. Enright, and T. C. Fre eman 2009. Network visualization and analysis of gene expression data using BioLayout Express(3D). Nat Protoc 4 :1535 1550. 130. Thomas, P., Y. Pang, E. J. Filardo, and J. Dong 2005. Identity of an estrogen membrane receptor coupled to a G protein in huma n breast cancer cells. Endocrinology 146 :624 632. 131. Titolo, D., F. Cai, and D. D. Belsham 2006. Coordinate regulation of neuropeptide Y and agouti related peptide gene expression by estrogen depends on the ratio of estrogen receptor (ER) alpha to ERbet a in clonal hypothalamic neurons. Mol Endocrinol 20 :2080 2092. 132. Titolo, D., C. M. Mayer, S. S. Dhillon, F. Cai, and D. D. Belsham 2008. Estrogen facilitates both phosphatidylinositol 3 kinase/Akt and ERK1/2 mitogen activated protein kinase membrane signaling required for long term neuropeptide Y transcriptional regulation in clonal, immortalized neurons. J Neurosci 28 :6473 6482. 13 3. Tsang, C. P 1974. Changes in plasma levels of estrone sulfate and estrone in the pregnant ewe around parturition. Steroids 23 :855 868.

PAGE 184

184 134. VanGuilder, H. D., K. E. Vrana, and W. M. Freeman 2008. Twenty five years of quantitative PCR for gene expression analysis. Biotechniques 44 :619 626. 135. Venables, W. N., Ripley, B. D 2002. Modern Applied Statistics with S. 136. Verducci, J. S., V. F. Melfi, S. Lin, Z. Wang, S. Roy, and C. K. Sen 20 06. Microarray analysis of gene expression: considerations in data mining and statistical treatment. Physiol Genomics 25 :355 363. 137. Walfish, P. G., T. Yoganathan, Y. F. Yang, H. Hong, T. R. Butt, and M. R. Stallcup 1997. Yeast hormone response element assays detect and characterize GRIP1 coactivator dependent activation of transcription by thyroid and retinoid nuclear receptors. Proc Natl Acad Sci U S A 94 :3697 3702. 138. Wang, Z., M. Gerstein, and M. Snyder 2009. RNA Seq: a revolutionary tool for tran scriptomics. Nat Rev Genet 10 :57 63. 139. Ward Jr, J. H 1963. Hierarchical grouping to optimize an objective function. Journal of the American statistical association 58 :236 244. 140. Warde Farley, D., S. L. Donaldson, O. Comes, K. Zuberi, R. Badrawi, P. Chao, M. Franz, C. Grouios, F. Kazi, C. T. Lopes, A. Maitland, S. Mostafavi, J. Montojo, Q. Shao, G. Wright, G. D. Bader, and Q. Morris 2010. The GeneMANIA prediction server: biological network integration for gene prioritization and predicting gene funct ion. Nucleic Acids Res 38 :W214 20. 141. Wei, C., J. Li, and R. E. Bumgarner 2004. Sample size for detecting differentially expressed genes in microarray experiments. BMC Genomics 5 :87. 142. Wei, S., S. Nandi, V. Chitu, Y. G. Yeung, W. Yu, M. Huang, L. T. Williams, H. Lin, and E. R. Stanley 2010. Functional overlap but differential expression of CSF 1 and IL 34 in their CSF 1 receptor mediated regulation of myeloid cells. J Leukoc Biol 88 :495 505. 143. Winikor, J., C. Schlaerth, M. B. Rabaglino, R. Cousins M. Sutherland, and C. E. Wood 2011. Complex actions of estradiol 3 sulfate in late gestation fetal brain. Reprod Sci 18 :654 665. 144. Wintour, E. M 1984. Developmental aspects of the hypothalamic pituitary adrenal axis. J Dev Physiol 6 :291 299. 145. Wo ng, J. K., H. H. Le, A. Zsarnovszky, and S. M. Belcher 2003. Estrogens and ICI182,780 (Faslodex) modulate mitosis and cell death in immature cerebellar neurons via rapid activation of p44/p42 mitogen activated protein kinase. J Neurosci 23 :4984 4995. 146. Wood, C. E., M. P. Fraites, and M. Keller Wood 2009. Blockade of PGHS 2 inhibits the hypothalamus pituitary adrenal axis response to cerebral

PAGE 185

185 hypoperfusion in the sheep fetus. American Journal of Physiology Regulatory, Integrative and Comparative Physiol ogy 296 :R1813 R1819. 147. Wood, C. E 1999. Control of parturition in ruminants. J Reprod Fertil Suppl 54 :115 126. 148. Wood, C. E 2011. Fetal hypothalamus pituitary adrenal responses to estradiol sulfate. Endocrinology 152 :4966 4973. 149. Wood, C. E., K. E. Gridley, and M. Keller Wood 2003. Biological activity of 17beta estradiol 3 sulfate in ovine fetal plasma and uptake in fetal brain. Endocrinology 144 :599 604. 150. Wood, C. E., and C. J. Saoud 1997. Influence of estradiol and androstenedione on ACTH and cortisol secretion in the ovine fetus. J Soc Gynecol Investig 4 :279 283. 151. Xiao, E., A. J. Kim, R. Dutia, I. Conwell, M. Ferin, and S. L. Wardlaw 2010. Effects of estradiol on cerebrospinal fluid levels of agouti related protein in ovariectomized rhesus monkeys. Endocrinology 151 :1002 1009. 152. Yang, J. J., and M. C. Yang 2006. An improved procedure for gene selection from microarray experiments using false discovery rate criterion. BMC Bioinformatics 7 :15. 153. Yang, Y. H., S. Dudoit, P. Luu, an d T. P. Speed 2001. Normalization for cDNA microarray data. In M. Bittner, Y. Chen, A. Dorsel, and D. Bellingham (eds.), Microarrays: Optical Technologies and Informatics. SPIE International Society for Optical Engineering, 154. Yu, N., K. D. Lariosa Wil lingham, F. F. Lin, M. Webb, and T. S. Rao 2004. Characterization of lysophosphatidic acid and sphingosine 1 phosphate mediated signal transduction in rat cortical oligodendrocytes. Glia 45 :17 27. 155. Zhang, B., and S. Horvath 2005. A general framework for weighted gene co expression network analysis. Stat Appl Genet Mol Biol 4 :Article17. 156. Zhang, B., S. Kirov, and J. Snoddy 2005. WebGestalt: an integrated system for exploring gene sets in various biological contexts. Nucleic Acids Res 33 :W741 8. 157. Zhang, J., R. Chiodini, A. Badr, and G. Zhang 2011. The impact of next generation sequencing on genomics. J Genet Genomics 38 :95 109. 158. Zhao, L., K. O'Neill, and R. D. Brinton 2006. Estrogenic agonist activity of ICI 182,780 (Faslodex) in hippoca mpal neurons: implications for basic science understanding of estrogen signaling and development of estrogen modulators with a dual therapeutic profile. J. Pharmacol. Exp. Ther. 319 :1124 1132.

PAGE 186

186 BIOGRAPHICAL SKETCH Maria Belen Rabaglino was born in the nice city of Rio Cuarto, Cordoba, Argentina. After a wonderful childhood and a great high school experience at the Agronomy School of Rio Cuarto, in 1998 she enrolled in the National University of Rio Cuarto (UNRC) to obtain a degree as veterinarian. Early in her veterinary education she devel oped an interest in reproductive physiology and the use of the laboratory as a tool to investigate physiologic process or its application in the clinical setting. Consequently, she trained to become Laboratory Technician, to complement her education in veterinary medicine. Maria Belen obtained her degree as Laboratory Technician in February, 2003; and as Veterinarian in August, 2003. After this, from July 2004, she attended graduat e courses at the National University of Cordoba (UNC) and the Animal Reproduction Institution of Cordoba (IRAC) and in August 2006, received the degree of Specialist in bovine reproduction. As a professional, Maria Belen has always been interested in acad emic and research activities. She obtained a scholarship for two years from the government at the UNRC as an undergraduate student to be a research assistant in the Animal Pathology Department. After graduation as Laboratory Technician, she obtained an emp loyment contract in the department of Clinical Analyses at the UNRC. After obtaining her veterinary degree, Maria Belen acquired a position in May 2004 as head teaching assistant in the Animal Reproduction Department at the UNRC and a contract to teach Phy siopathology of Reproduction at a private University in Mendoza, Argentina.

PAGE 187

187 During her tenure at UNRC and the private University, Maria Belen participated in research and service activities related to canine, equine and ruminant theriogenology, as well as organizing and teaching of graduate courses. All these activities provided her with an excellent opportunity to develop her personal skills conducive for a teamwork approach for research and the stimulus to continue improving her knowledge in a foreign co untry. It is known that USA is at the cutting edge in research and technology so, in order to fulfill her goals, Maria Belen applied and obtained in August 2006, a special Fulbright scholarship (Faculty Development program). This scholarship gave her the o pportunity to select any University in the United States to pursue a Master program beginning in August 2007, which opened a new chapter in her life. Maria Belen chose University of Florida due to its prestige and, as she has been always interested in the events related with gestation, parturition and the postpartum period, to conduct research in these areas with Dr. Carlos Risco as ad visor. Maria Belen received her m aster degree from University of F lorida in the summer 2009. The m aster program stimula ted her to continue with gradua te studies. So, she joined a doctoral program in Animal Molecular and Cell Biology at the University of Florida in August 2009, with Dr. Charles E. Wood as advisor. The doctoral research was developed in the area of the bioin formatics and fetal biology, analyzing the changes in gene expression in the ovine fetal brain close to parturition. The overall doctoral program has resulted absolutely satisfactory to Maria Belen. She not only has gained invaluable skills in the area of gene analysis and fetal surgeries but also, she has acquired incredible knowledge in the field of scientific investigation. Now that this whole

PAGE 188

188 amazing experience in USA is concluding, Maria Belen will continue with her position as a teacher and investigat or at the UNRC in Argentina, where she hopes to contribute with the scientific development of that country.