EPA/600/R-09/028F | September 2009 | www.epa.gov
United States
Environmental Protection
Agency
An Approach to Using Toxicogenomic Data in
U.S. EPA Human Health Risk Assessments:
A Dibutyl Phthalate Case Study
-------
Document cover:
The document cover was designed by Katherine Loizos of SRA International, Creative Services,
Cincinnati, OH. Permission was obtained to use the following images:
Human chromosome image: Elsevier Press. The source is the article published in Genomics
9(4) by Ward, DC and Baldini, A in the article, in situ hybridization banding of human
chromosomes with Alu-PCR products: a simultaneous karyotype for gene mapping studies, pp.
770-774. Copyright Elsevier (1991).
Microarray experiment output image: Poirazi Laboratory, Institute of Molecular Biology and
Biotechnology, Foundation for Research and Technology - Hellas, Crete, Greece.
Gene network image: Justen Andrews, Indiana University. The data for the gene network
image are from the paper: Costello JC, Dalkilic MM, Beason SM, et al. (2009). Gene networks
in Drosophila melanogaster: integrating experimental data to predict gene function. Genome
Biol 2009 Sep 16;10(9):R97. [Epub ahead of print]
-------
EPA/600/R-09/028F
September 2009
An Approach to Using Toxicogenomic Data
In U.S. EPA Human Health Risk Assessments:
A Dibutyl Phthalate Case Study
National Center for Environmental Assessment
Office of Research and Development
U.S. Environmental Protection Agency
Washington, DC 20460
-------
DISCLAIMER
This document has been reviewed in accordance with U.S. Environmental Protection
Agency policy and approved for publication. Mention of trade names or commercial products
does not constitute endorsement or recommendation for use.
Preferred citation:
U.S. Environmental Protection Agency (EPA). (2009) An approach to using toxicogenomic data
in U.S. EPA human health risk assessments: a dibutyl phthalate case study. National Center for
Environmental Assessment, Washington, DC; EPA/600/R-09/028F. Available from the National
Technical Information Service, Springfield, VA, and online at http://www.epa.gov/ncea.
11
-------
CONTENTS
LIST OF TABLES vii
LIST OF FIGURES ix
LIST OF ABBREVIATIONS AND ACRONYMS xi
PREFACE xiv
AUTHORS, CONTRIBUTORS, AND REVIEWERS xv
ACKNOWLEDGMENTS xvi
1. EXECUTIVE SUMMARY 1-1
1.1. APPROACH 1-1
1.2. DBF CASE STUDY 1-2
1.3. RECOMMENDATIONS 1-5
1.4. RESEARCH NEEDS 1-7
2. INTRODUCTION 2-1
2.1. PURPOSE 2-1
2.2. REPORT OVERVIEW 2-3
2.3. USE OF TOXICOGENOMICS IN RISK ASSESSMENT 2-4
2.3.1. Definitions 2-4
2.3.2. Current Efforts to Utilize Toxicogenomic Data in Risk Assessment 2-7
2.3.2.1. Toxicogenomics Informs TD 2-7
2.3.2.2. Toxicogenomics Informs Dose-Response 2-9
2.3.2.3. Toxicogenomics Informs Interspecies Extrapolations 2-10
2.3.2.4. Toxicogenomics Informs Intraspecies Variability 2-11
2.3.2.5. TK/TD Linkages Informed by Toxicogenomic Data 2-11
2.3.2.6. Toxicogenomic Activities at the U.S. Food and Drug
Administration (FDA) 2-12
2.3.2.7. Toxicogenomic Activities at EPA 2-14
2.3.2.8. Toxicogenomic Activities at Other Agencies and Institutions 2-16
2.3.3. Current Challenges and Limitations of Toxicogenomic Technologies 2-18
2.4. INTRODUCTION TO THE CASE STUDY 2-19
2.4.1. Project Team 2-19
2.4.2. Chemical Selection 2-19
2.4.2.1. Six Candidate Chemicals 2-20
2.4.2.2. Selection of the Case-Study Chemical 2-21
2.4.3. Case-Study Scope 2-23
3. DBF CASE-STUDY APPROACH AND EXERCISE 3-1
3.1. EVALUATING THE EXTERNAL REVIEW DRAFT OF THE IRIS
TOXICOLOGICAL REVIEW (TOX REVIEW) OF DBF 3-1
3.2. CONSIDERATION OF RISK ASSESSMENT ASPECTS THAT
TOXICOGENOMIC DAT A MAY ADDRESS 3-3
3.2.1. Informing TK 3-7
in
-------
CONTENTS (continued)
3.2.1.1. Identification of Potential Metabolic and Clearance
Pathways 3-7
3.2.1.2. Selection of Appropriate Dose Metrics 3-8
3.2.1.3. Intra- and Interspecies Differences in Metabolism 3-8
3.2.1.4. TK/TDLinkages andFeedback 3-9
3.2.1.5. Research Needs for Toxicogenomic Studies to Inform TK 3-9
3.2.1.6. DBF Case Study: Do the Available Toxicogenomic Data
Inform TK? 3-10
3.2.2. Informing Dose-Response 3-13
3.2.2.1. DBF Case Study: Do the Toxicogenomic Data Inform Dose-
Response? 3-13
3.2.3. Informing TD 3-14
3.2.3.1. General Considerations: TD Portion of Mechanisms of
Action and MO As 3-14
3.2.3.2. DBF Case Study: MO As for Male Reproductive
Developmental Effects 3-14
3.3. IDENTIFYING AND SELECTING QUESTIONS TO FOCUS THE DBF
CASE STUDY 3-17
4. EVALUATION OF THE REPRODUCTIVE DEVELOPMENTAL TOXICITY
DATA SET FOR DBF 4-1
4.1. CRITERIA AND RATIONALE FOR INCLUSION OF TOXICOLOGY
STUDIES IN THE EVALUATION 4-2
4.2. REVIEW OF THE TOXICOLOGY DATA SET 4-10
4.3. UNEXPLAINED MO As FOR DBF MALE REPRODUCTIVE TOXICITY
OUTCOMES 4-21
4.4. CONCLUSIONS ABOUT THE TOXICITY DATA SET EVALUATION:
DECISIONS AND RATIONALE 4-26
5. EVALUATION OF THE DBF TOXICOGENOMIC DATA SET 5-1
5.1. METHODS FOR ANALYSIS OF GENE EXPRESSION: DESCRIPTION OF
MICRO ARRAY TECHNIQUES AND SEMI-QUANTITATIVE RT-PCR 5-1
5.1.1. Microarray Technology 5-1
5.1.2. Reverse Transcript!on-Polymerase Chain Reaction (RT-PCR) 5-2
5.2. REVIEW OF THE PUBLISHED DBF TOXICOGENOMIC STUDIES 5-3
5.2.1. Overview of the Toxicogenomic Studies 5-3
5.2.2. Microarray Studies 5-3
5.2.2.1. Shultzetal. (2001) 5-3
5.2.2.2. Bowman etal. (2005) 5-7
5.2.2.3. Liu et al. (2005) 5-8
5.2.2.4. Thompson et al. (2005) 5-10
5.2.2.5. Plummer et al. (2007) 5-12
IV
-------
CONTENTS (continued)
5.2.3. RT-PCR Studies 5-14
5.2.3.1. Barlow et al. (2003) 5-14
5.2.3.2. Lehmann et al. (2004) 5-15
5.2.3.3. Thompson et al. (2004) 5-16
5.2.3.4. Wilson et al. (2004) 5-17
5.2.4. Study Comparisons 5-18
5.2.4.1. Microarray Study Methods Comparison 5-18
5.2.4.2. RT-PCR Study Methods Comparison 5-19
5.3. CONSISTENCY OF FINDINGS 5-22
5.3.1. Microarray Study Findings 5-22
5.3.2. RT-PCR Gene Expression Findings 5-25
5.3.3. Protein Study Findings 5-25
5.3.4. DBF Toxicogenomic Data Set Evaluation: Consistency of Findings
Summary 5-26
5.4. DATA GAPS AND RESEARCHNEEDS 5-30
5.5. PATHWAY ANALYSIS OF DBF MICROARRAY DATA 5-31
5.5.1. Objective of the Reanalysis of the Liu etal. (2005) Study 5-31
5.5.2. Pathway Analysis of Liu et al. (2005) Utilizing Two Different Methods
to Generate Hypotheses for MO As Underlying the Unexplained Testes
Endpoints 5-32
5.5.2.1. Two Methods for Identifying Differentially Expressed
Genes (DEGs) 5-34
5.5.2.2. Pathway Analysis 5-37
5.6. CONCLUSIONS 5-46
6. EXPLORATORY METHODS DEVELOPMENT FOR ANALYSIS OF GENOMIC
DATA FOR APPLICATION TO RISK ASSESSMENT 6-1
6.1. OBJECTIVES AND INTRODUCTION 6-1
6.2. PATHWAY ANALYSIS AND GENE INTERACTIONS AFTER IN UTERO
DBF EXPOSURE 6-2
6.2.1. Pathway Activity Approach 6-2
6.2.1.1. Significance Analysis of Pathway Activity Levels 6-3
6.2.1.2. Pathway Activity Analysis 6-4
6.2.2. Developing a Temporal Gene Network Model 6-9
6.3. EXPLORATORY METHODS: MEASURES OF INTERSPECIES (RAT-TO-
HUMAN) DIFFERENCES IN TOXICODYNAMICS 6-12
6.4. CONCLUSIONS 6-19
7. CONCLUSIONS 7-1
7.1. APPROACH FOR EVALUATING TOXICOGENOMIC DATA IN
CHEMICAL ASSESSMENTS 7-1
7.2. DBF CASE-STUDY FINDINGS 7-4
-------
CONTENTS (continued)
7.2.1. MOA Case Study Question: Do the DBF Genomic Data Inform
Mechanism(s) of Action andMOA(s)? 7-5
7.2.2. Interspecies MOA Case Study Question: Do the DBF Genomic Data
Inform Interspecies Differences in TD? 7-8
7.2.3. Application of Genomic Data to Risk Assessment: Exploratory
Methods and Preliminary Results 7-9
7.2.4. Application of Genomic Data to Risk Assessment: Using Data
Quantitatively 7-10
7.3. LESSONS LEARNED 7-12
7.3.1. Research Needs 7-13
7.3.1.1. Data Gaps and Research Needs: DBF 7-13
7.3.1.2. Research Needs for Toxicity and Toxicogenomic Studies for
Use in Risk Assessment 7-14
7.3.2. Recommendations 7-16
7.3.3. Application of Genomic Data to Risk Assessment: Future
Considerations 7-18
APPENDIX A: SUPPORTING TABLES FOR CHAPTER 5 A-l
APPENDIX B: SUPPORTING TABLES AND FIGURES FOR CHAPTER 6 B-l
APPENDIX C: QUALITY CONTROL AND ASSURANCE C-l
GLOSSARY G-l
REFERENCES R-l
VI
-------
LIST OF TABLES
2-1. Information available July 2005 on the selection criteria for the six candidate
chemicals affecting the androgen-mediated male reproductive developmental
toxicity pathway 2-22
4-1. Studies with exposures during development that have male reproductive
outcomes (limited to reproductive organs and/or reproductive function) and
were considered adequate for reference value determination 4-4
4-2. Reporting and study size characteristics of male reproductive studies following
in utero exposure to DBF 4-12
4-3. Life stage at observation for various male reproductive system outcomes
assessed in studies of developmental exposure to DBF 4-15
4-4. Age of assessment for individual endpoints across studies of the male
reproductive system following developmental exposure to DBF 4-16
4-5. Incidence of gross pathology in Fl male reproductive organs in one continuous
breeding study with DBF 4-22
4-6. Evidence for MO As for the observed effects in the male reproductive system after
in utero DBF exposure 4-24
5-1. Study comparisons for the toxicogenomic data set from male tissues after in utero
DBF exposure 5-4
5-2. Lehmann et al. (2004) DBF dose-response gene expression data measured by RT-
PCR showing statistically significant changes from control 5-16
5-3. Method comparisons for DBF microarray studies 5-19
5-4. Method comparisons among the RT-PCR DBF studies 5-20
5-5. Evaluation of the published protein studies after DBF in utero exposure (testes
only) 5-27
5-6. Common pathways between the REM and SNR analyses of differentially
expressed genes (DEGs) after in utero DBF exposure from the Liu et al. (2005)
data 5-39
5-7. Genes involved in cholesterol biosynthesis/metabolism that were identified by
both the REM and SNR analyses of Liu et al. (2005) 5-42
6-1. The KEGG pathways ordered based on their/7-value for pathway activity 6-7
vn
-------
LIST OF TABLES (continued)
6-2. The amino acid sequence similarity of the enzymes in the steroidogenesis
pathway between rat and human 6-17
7-1. DBF dose-response progression of statistically significant events illustrated with
a subset of precursor event data (steroidogenesis gene expression, T expression)
and in vivo endpoints with the reduced T MOA 7-11
7-2. Research needs for toxicogenomic studies to be used in risk assessment 7-15
7-3. Research needs for toxicity studies for utilizing toxicogenomic and toxicity data
together in risk assessment 7-17
Vlll
-------
LIST OF FIGURES
2-1. The relationship between the project process, goals, and products for the
development of an approach and case study for the use of toxicogenomic data in
risk assessment 2-2
2-2. Androgen-mediated male reproductive developmental toxicity pathway 2-20
3-1. DBF case-study approach for evaluating toxicogenomic data for a health
assessment 3-2
3-2. Exposure response array for candidate endpoints and PODs for RfD derivation
presented in the external review draft IRIS Tox Review for DBF (U.S. EPA,
2006a) 3-4
3-3. Potential uses of toxicogenomic data in chemical screening and risk assessment 3-5
3-4. Potential uses of toxicogenomic data in understanding mechanisms of action 3-6
3-5. The fetal Ley dig cell in the fetal testis 3-12
3-6. Approach to utilizing toxicity and toxicogenomic data for identifying affected
pathways and candidate modes and mechanisms of action 3-15
3-7. The proposed DBF mechanism of action for the male reproductive developmental
effects 3-16
4-1. The process for evaluating the male reproductive developmental toxicity data set
for low-dose and low-incidence findings 4-3
4-2. The process for evaluating the MOA for individual male reproductive system
outcomes following developmental DBF exposure 4-23
5-1. Venn diagram illustrating similarities and differences in significant gene
expression changes observed in three recent microarray studies of the testes:
Thompson et al. (2005), Plummer et al. (2007), and Liu et al. (2005) 5-24
5-2. Summary of DBF-induced changes in fetal gene and protein expression 5-29
5-3. Schematic of the two analysis methods (REM and SNR) for identifying
differentially expressed genes and subsequent pathway analysis using GeneGo 5-33
5-4. Heat map of 1,577 DEGs from SNR analysis method 5-37
5-5. Mapping the Liu et al. (2005) data set onto the canonical androstenedione and
testosterone (T) biosynthesis and metabolism pathway in MetaCore (GeneGo) 5-43
IX
-------
LIST OF FIGURES (continued)
6-1. An illustration of the adapted version of pathway activity level analysis for the
tryptophan metabolism pathway, a nonactive pathway for DBF 6-5
6-2. Metabolic pathway network for DBF (Liu et al., 2005 data) using the pathway
activity method and the KEGG database 6-8
6-3. The relationship between differential expression of individual genes and
pathway activity using the Liu et al. (2005) DBF data 6-10
6-4. A gene network for DBF data of Liu et al. (2005) generated using Ingenuity
Pathway Analysis (IPA) 6-11
6-5. A temporal gene network model created by IP A from the informative
gene list based on time-course data after in utero DBF exposure 6-13
6-6. The phylogenetic relations among eight organisms based on enzyme presence, for the
biosynthesis of steroids pathway, and based on information available on the NCBI
taxonomy website (Sayers etal., 2008) 6-15
7-1. Approach for evaluating and incorporating genomic data into future chemical
assessments 7-2
-------
LIST OF ABBREVIATIONS AND ACRONYMS
Please note that most gene and protein name abbreviations are not included in this list
because of the large number of genes and proteins described in the report. The gene and protein
names have been standardized using information from the Rat Genome Project.
ADH alcohol dehydrogenase
ADME absorption, distribution, metabolism, and excretion
AGD anogenital distance
AMH anti-mullerian hormone
ANOVA analysis of variance
AR androgen receptor
BBDR biologically based dose-response
BBP butyl benzyl phthalate
BMD benchmark dose
BMDL benchmark dose lower confidence limit
BPA bisphenol A
cDNA complementary DNA
CNPs copy-number polymorphisms
DBF dibutyl phthalate
DEG differentially expressed gene
DEHP di-(2-ethylhexyl) phthalate
DEP diethyl phthalate
DMP dimethyl phthalate
DOTP diocytyl tere-phthalate
DPP dipentyl phthalate
EDC endocrine disrupting chemical
EPA Environmental Protection Agency
ER estrogen receptor
ESTs expressed sequence tags
FDA Food and Drug Administration
GAPDH glyceraldehyde-3-phosphate dehydrogenase
GD gestation day
GO Gene Ontology
GSH glutathione
XI
-------
LIST OF ABBREVIATIONS AND ACRONYMS (continued)
HESI Health and Environmental Sciences Institute
ILSI International Life Sciences Institute
IPA Ingenuity® Pathway Analysis
IPCS International Programme on Chemical Safety
IRIS Integrated Risk Information System
KEGG Kyoto Encyclopedia of Genes and Genomes
LC Leydig cell
LMW low molecular weight
LOAEL lowest-observed-adverse-effect level
LOEL lowest-observed-effect level
MAPK/ERK mitogen-activated protein kinase/extracellular signal-regulated kinase
MAQC MicroArray Quality Control
MAS microarray suite
MBP monobutylphthalate
MMP matrix metalloproteinase
MOA mode of action
mRNA messenger RNA
NCCT National Center for Computational Toxicology
NCEA National Center for Environmental Assessment
NIEHS National Institute of Environmental Health Sciences
NOAEL no-observed-adverse-effect level
NOEL no-observed-effect level
NRC National Research Council
NTP National Toxicology Program
PA pathway activity
PBPK physiologically-based pharmacokinetic
PCA principal component analysis
PCR polymerase chain reaction
PFOA perfluorooctanoic acid
PND postnatal day
POD point of departure
PPAR peroxisome proliferator-activated receptor
PPS preputial separation
xn
-------
LIST OF ABBREVIATIONS AND ACRONYMS (continued)
RA risk assessment
RACE reproductive assessment by continuous breeding
RfD reference dose
RT-PCR reverse transcription-polymerase chain reaction
SD Sprague-Dawley
SLR signal log ratio
SNPs single nucleotide polymorphisms
SNR signal-to-noise ratio
SPC Science Policy Council
STAR Science to Achieve Results
T testosterone
TD toxicodynamics
TF transcription factor
TK toxicokinetics
Tox Review Toxicological Review
UFH intraspecies uncertainty factor
UMDNJ University of Medicine and Dentistry of New Jersey
VLI valine, leucine, isoleucine
WD Wolffian duct
WOE weight-of-evidence
Xlll
-------
PREFACE
The U.S. Environmental Protection Agency (EPA) is interested in developing methods to
use genomic data most effectively in risk assessments performed at EPA. The National Center
for Environmental Assessment (NCEA) within the Office of Research and Development (ORD)
prepared this document for the purpose of describing and illustrating an approach for using
toxicogenomic data in risk assessment. The approach and dibutyl phthalate (DBF) case study
described in this document were developed by a team of scientists at EPA laboratories and
centers, and outside organizations including The Hamner Institutes for Health Sciences, the
National Institute of Environmental Health Sciences (NIEHS), and the EPA National Center for
Environmental Research (NCER) Science to Achieve Results (STAR) Environmental
Bioinformatics and Computational Toxicology (Comp Tox) Center at the University of Medicine
and Dentistry of New Jersey (UMDNJ) and Rutgers University. The intended audience for this
document includes risk assessors as well as scientists with expertise in genomics, bioinformatics,
toxicology, and statistics. The approach outlined in this document is expected to be useful to
EPA risk assessors in the Integrated Risk Information System (IRIS) Program and other program
offices and regions, as well as the scientific community at large. The review of the literature on
the use of genomic data in risk assessment, as well as discussions of issues, recommendations,
and methods for evaluating and analyzing toxicogenomic data, could be useful to scientists and
risk assessors within and outside of EPA. The research needs identified in this document will be
useful to scientists performing toxicology and toxicogenomic research studies for application to
risk assessment. The DBF case study presented in this document is a separate activity from the
IRIS DBF health assessment. The review of the literature included in this document was last
updated in July 2007.
xiv
-------
AUTHORS, CONTRIBUTORS, AND REVIEWERS
AUTHORS
Susan Euling, NCEA-W, U.S. EPA, Washington, DC (Lead)
loannis Androulakis, NCER STAR Bioinformatics Center, ebCTC, Rutgers/UMDNJ,
Piscataway, NJ
Bob Benson, Region 8, U.S. EPA, Denver, CO
Weihsueh Chiu, NCEA-W, U.S. EPA, Washington, DC
Paul Foster, NIEHS, Research Triangle Park (RTF), NC
Kevin Gaido, U.S. FDA, Rockville, MD; formerly of *The Hamner Institutes for Health
Sciences, RTF, NC
L. Earl Gray Jr., NHEERL, U.S. EPA, RTF, NC
Susan Hester, NHEERL, U.S. EPA, RTF, NC
Channa Keshava, IRIS, U.S. EPA, Washington, DC
Nagalakshmi Keshava, NCEA-W, U.S. EPA, Washington, DC
Andrea Kim, Allergan, Inc., Irvine, CA; formerly of *NCEA-W, U.S. EPA, Washington, DC
Susan Makris, NCEA-W, U.S. EPA, Washington, DC
Meric Ovacik, NCER STAR Bioinformatics Center, ebCTC, Rutgers/UMDNJ, Piscataway, NJ
Banalata Sen, NIEHS; formerly of *NCEA-RTP, U.S. EPA, RTF, NC
Chad Thompson, ToxStrategies, Katy, TX; formerly of *NCEA-W, U.S. EPA, Washington, DC
Lori White, *NIEHS; formerly of *NCEA-RTP, U.S. EPA, RTF, NC
Vickie Wilson, NHEERL, U.S. EPA, RTF, NC
CONTRIBUTORS
Stan Barone, NCEA-W, U.S. EPA, Washington, DC
Marianthi lerapetritou, NCER STAR Bioinformatics Center, ebCTC, Rutgers/UMDNJ,
Piscataway, NJ
*Affiliation at the time of work on this project for scientists with more than one affiliation listed.
xv
-------
AUTHORS, CONTRIBUTORS, AND REVIEWERS (CONTINUED)
REVIEWERS
Internal
Maureen R. Gwinn, NCEA-W, ORD, Washington, DC
Michael Hemmer, NHEERL, ORD, Gulf Breeze, FL
Nancy McCarroll, Office of Pesticide Programs (OPP), Washington, DC
Gregory Miller, Office of Children's Health Protection and Environmental Education
(OCHPEE), Washington, DC
Marian Olsen, Region 2, New York, NY
Santhini Ramasamy, Office of Water (OW), Washington, DC
Jennifer Seed, Office of Pollution Prevention and Toxics (OPPT), Washington, DC
Imran Shah, National Center for Computational Toxicology (NCCT), ORD, RTF, NC
Jamie Strong, IRIS, ORD, Washington, DC
Dan Villeneuve, NHEERL, ORD, Duluth, MN
External
Jeanne Manson (chair), Exponent, Inc.
Robert Chapin, Pfizer, Inc.
Julia Gohlke, NIEHS
Hisham Hamadeh, Amgen, Inc.
Poorni Iyer, California EPA (Note: Dr. Iyer conducted the peer review as a private consultant
and not as a representative of California EPA)
ACKNOWLEDGMENTS
This project was funded by NCEA and the EPA NCCT's Research Program under their
new starts grants. We thank the outside partners, NIEHS and The Hamner Institutes for Health
Sciences, for allowing team members at these institutions to work on this project. Some of the
work described was performed at the STAR Bioinformatics Center at UMDNJ and Rutgers
University that is supported by the grant R832721 from the EPA's STAR program. We
gratefully acknowledge Dr. Kevin Gaido for providing the data from the Liu et al. (2005) study
performed in his laboratory at The Hamner Institutes for Health Sciences, Terri Konoza of
NCEA for her detailed editorial contribution to this document, and Sarah Burgess-Herbert for her
thoughtful review of Chapter 6.
xvi
-------
1. EXECUTIVE SUMMARY
We developed a systematic approach for evaluating and utilizing toxicogenomic data in
health assessment. This report describes this approach and a case study conducted for dibutyl
phthalate (DBF) to illustrate the approach. As a result of the case-study exercise, we refined the
initial case-study approach for general use in new chemical assessments. In this report, we
reviewed some of the recent and ongoing activities regarding the use of genomic data in risk
assessment, inside and outside of the U.S. Environmental Protection Agency (EPA). We also
identified research needs, recommendations, and issues for future consideration when using
genomic data in risk assessments.
Toxicogenomics is the application of genomic technologies (e.g., transcriptomics,
proteomics, metabolomics, genome sequence analysis) to study the effects of environmental
chemicals on human health and the environment. The EPA Interim Genomics Policy (U.S. EPA,
2002a) encourages the use of genomic data, on a case-by-case basis, in a weight-of-evidence
(WOE) approach. Currently, EPA provides no guidance for incorporating genomic data into risk
assessments of environmental agents. However, EPA's Science Policy Council (SPC) has
developed interim guidance regarding other aspects of the use of microarray data at EPA,
entitled Interim Guidance for Microarray-Based Assays: Data Submission, Quality, Analysis,
Management, and Training Considerations (U.S. EPA, 2006b).
DBF was selected for the case study because it has a relatively large genomic data set and
phenotypic anchoring of certain gene expression data to some male reproductive developmental
outcomes. The scope of the case study was limited to the male reproductive developmental
outcomes of DBF, and this effort was limited to evaluating the available published toxicity and
toxicogenomic data for the DBF case study. The DBF case study is a separate endeavor with
distinct goals from EPA's Integrated Risk Information System (IRIS) assessment of DBF.
1.1. APPROACH
Genomic data have the potential to inform toxicodynamics (TD), toxicokinetics (TK),
inter- and intraspecies differences in TD and TK, exposure assessment, and dose-response
assessment. Our strategy was to design an approach for evaluating genomic data for risk
assessment that is both systematic and flexible enough to accommodate different health and risk
1-1
-------
assessment practices. The first step of the approach is to evaluate the available genomic data set
for its application to a broad range of information types (e.g., TD, TK, intra-and interspecies TD
and TK differences) that are useful to risk assessment as well as the steps of health assessment
(e.g., hazard characterization, dose-response assessment). Through this iterative process, the
potential use of the available genomic data is determined. As part of the scoping step, the
available human, toxicology, and genomics studies are reviewed to determine their use to the
genomic data set evaluation. For instance, the toxicity, human, and toxicogenomic data sets are
considered together to determine the relationship (i.e., degree of phenotypic anchoring) between
gene and pathway changes to health or toxicity outcomes. As a result of the scoping step,
questions are posed to direct and focus the evaluation of the genomic data set.
The next steps include detailed evaluations directed by the formulated questions of the
toxicity and/or epidemiological data sets and the toxicogenomic data set. For example, when
genomic data are available to inform mechanisms of action or modes of action (MOAs), the
toxicogenomic and toxicity data sets can be evaluated together, relating the affected endpoints
(identified in the toxicity data set evaluation) to the genes and/or pathways (identified in the
toxicogenomic data set evaluation) to establish or formulate hypotheses about an MOA. In
addition to informing the mechanisms of action and the MO As (TD and TK steps), genomic data
also have the potential to inform inter- and intraspecies TD differences, and dose-response
assessment, depending on the genomic study design (e.g., species, organ, single dose vs. multiple
doses, genomic method) of the available data. The approach also includes new analyses of the
genomic data for the purpose of risk assessment when data are available and such new analyses
may address questions that are relevant to the risk assessment.
1.2. DBF CASE STUDY
For the DBF case-study example, we utilized the data set summaries and data gaps
identified in the external review draft IRIS Tox Review for DBF (U.S. EPA, 2006a) and asked
whether the genomic data set could inform any of these data gaps. In parallel, the DBF genomic
data set was considered, in light of all risk assessment aspects that these data might inform. As a
result of following these two processes, we formulated two specific case-study questions that the
available genomic data for DBF had the potential to inform:
1-2
-------
• Do the toxicogenomic data inform the mechanisms of action and/or MO As for DBF?
• Do the toxicogenomic data inform interspecies differences in TD?
The team considered it highly likely that the DBF toxicogenomic data set could inform the
modes or mechanisms of action. The team considered it possible, but less certain, that the cross-
species differences in one or more DBF MO As could be informed by evaluating genomic data
(e.g., DNA sequence data).
Additional questions were excluded because appropriate data were lacking. For example,
one question of great interest is, Do the toxicogenomic data inform dose-response? However,
this question could not be addressed in this case study because there were no dose-response
genomic data for DBF. Few chemicals have available dose-response genomic data and DBF is
not unusual in this respect. The evaluation of the one available DBF dose-response gene
expression study, although not global, is discussed in the report. As a result of the DBF genomic
data set limitations, the case study focuses on the qualitative application of genomic data to risk
assessment. In addition, exposure assessment was not considered in this approach because the
case study was performed using the IRIS chemical assessment model, which only includes
hazard identification and dose-response steps of the risk assessment paradigm.
We found that the DBF toxicogenomic data did inform the mechanism of action, and
generated hypotheses about possible additional MO As, for DBF and male reproductive
developmental outcomes. There is substantial evidence in the published literature that a number
of the gene expression changes observed in genomic studies are phenotypically anchored for a
number of the male reproductive developmental outcomes observed after in utero DBF exposure
in the rat. The available genomic and other gene expression data, hormone level data, and
toxicity data for DBF are instrumental in the establishment of two MOAs: (1) a decrease in fetal
testicular testosterone (T); and (2) a decrease in Insulin-like 3 (Insl3} expression. A decrease in
fetal testicular T is a well-established MOA for a number of the male reproductive
developmental effects observed in the male rat after in utero DBF exposure. The genomic and
single gene expression data, after in utero DBF exposure, identified changes in genes involved in
steroidogenesis and cholesterol transport, consistent with the observed decrease in fetal testicular
T. Decreased Insl3 expression is a second well-established MOA responsible, in conjunction
with reduced T, for the undescended testis effect observed following in utero DBF exposure.
1-3
-------
Reverse transcription-polymerase chain reaction (RT-PCR) and in vivo toxicology study results
support the role oflnslS in one of the two steps of testis descent.
Evaluating genomic and toxicity data together also provides information on putative
novel MO As. A number of the DBF toxicity and toxicogenomic studies were performed in the
same strain of rat using similar doses and exposure intervals that allowed for comparisons across
studies. In this case study, rodent reproductive developmental toxicity studies were evaluated for
low incidence and low-dose findings and for the male reproductive developmental effects that
currently do not have an explained MOA (termed "unexplained endpoints"). In the case study,
we focused on the outcomes in the testes because all, but one, of the DBF toxicogenomic studies
were performed on testes. We identified five testicular endpoints without a known MOA that
were pursued further in the evaluation of the toxicogenomic data set.
The nine published RT-PCR and microarray studies in the rat were evaluated as part of
the toxicogenomic and associated gene expression data set to identify genes and pathways
affected after in utero DBF exposure. Both the microarray data set alone and the entire gene
expression data set (including all gene expression studies including microarray studies) were
evaluated for consistency of findings. At the gene level, the findings from the DBF genomic
studies (i.e., microarray, RT-PCR, and protein expression) were relatively highly correlated with
one another in both the identification of differentially expressed genes (DEGs) and their
direction of effect. The evaluation of the published toxicity and toxicogenomic studies
corroborates the two known MO As for DBF.
The published microarray studies for DBF focused primarily on pathways related to the
reduced fetal testicular T MOA, such as the steroidogenesis pathway. We performed new
analyses of the data from one rat testes microarray study in order to identify all possible
pathways significantly affected by in utero DBF exposure. Using two different analytical
methods, pathways associated with the two well-established MO As (decreased Insl3 and fetal
testicular T), as well as new processes (e.g., growth and differentiation, transcription, cell
adhesion) and pathways (e.g., Wnt signaling, cytoskeleton remodeling) not associated with either
Insl3 or steroidogenesis pathways, were identified. The newly identified putative pathways may
play a role in the regulation of steroidogenesis (i.e., related to a known MOA for DBF) or,
alternatively, may inform additional MO As for one or more unexplained outcomes in the testes.
1-4
-------
The new analyses and the approach allowed us to develop hypotheses about possible DBF
MO As for some male reproductive developmental outcomes.
To address the question of whether the available genomic data for DBF could inform the
interspecies TD part of the interspecies uncertainty factor, genomic data were evaluated to
inform interspecies differences in the steroidogenesis pathway, relevant to the decreased fetal
testicular T MOA. We explored the development of new methods to evaluate interspecies TD
differences. To evaluate cross-species similarity metrics for the steroidogenesis pathway
between rats and humans, we explored two approaches: protein sequence similarity and enzyme
presence. Preliminary results from applying each method suggest that steroidogenesis genes are
relatively highly conserved between rats and humans. However, we do not recommend utilizing
these data to inform interspecies uncertainty for DBF because it is difficult to make unequivocal
conclusions regarding a "high" versus "low" degree of conservation for the genes in this pathway
based on these data alone. With further refinement and improved data sources, these methods
could potentially be applied to other chemical assessments.
New methods for evaluating microarray data for the purposes of risk assessment were
explored and developed during the DBF case study. A new pathway analysis method, the
pathway activity level method, was developed and tested with two DBF study data sets. The
pathway activity level method determines pathway level changes as the initial step as opposed to
standard pathway analysis methods in which DEGs are first identified, followed by mapping of
the DEGs to pathways, as a second step. Further, the pathway activity level method was used to
evaluate time-course microarray data. A preliminary gene network model for DBF, based on the
results from one time-course study, identified a temporal sequence of gene expression and
pathway interactions that occur over an 18-hour interval within the critical window of exposure
for DBF and testicular development effects.
1.3. RECOMMENDATIONS
In addition to following the principles of the approach (i.e., systematically consider all
types of information with respect to the steps of risk assessment, identify questions to direct the
evaluation, and evaluate genomic data and toxicity data together), several specific
methodological recommendations arose from the DBF case-study experience. The first two
recommendations are straightforward and could reasonably be performed by a risk assessor with
1-5
-------
basic training in genomics data evaluation and interpretation. The third recommendation
requires expertise in genomic data analysis methods for implementation. The recommendations
are presented below:
1. Evaluate the genomic and other gene expression data for consistency of findings across
studies to provide a WOE evaluation of the affected gene expression and pathways.
Some simple methods, such as using Venn diagrams and gene expression compilation
approaches, can be applied to risk assessment. When evaluating the consistency of
toxicogenomic data findings, it is advantageous to include all available gene expression
data (single gene, global gene expression, protein, RNA) because single gene expression
techniques have been traditionally used to confirm the results of global gene expression
studies and because single gene expression data add to the database.
2. Perform benchmark dose (BMD) modeling on high-quality RT-PCR dose-response
studies of genes known to be in the causal pathway of an MO A or outcome of interest.
Obtaining a BMD and benchmark dose lower confidence limit (BMDL) is a useful
starting point for both linear low-dose extrapolation and reference value approaches. We
are not indicating which approach is appropriate to take for making predictions about the
potential risk below the BMD or BMDL. "High quality" is defined in this context as a
well-conducted study that assessed enough animals and litters for sufficient statistical
power for characterizing the mean responses and the variability (interlitter and intralitter).
3. Perform new analysis of toxicogenomic data in cases when the new analysis is likely to
yield new information that would be useful to the risk assessment. Examples include:
• Perform a new pathway analysis in order to identify all affected pathways or other
risk assessment applications. When the available published microarray studies
have been conducted for purposes (e.g., basic science, pharmaceutical
development) other than risk assessment, it may be useful to reanalyze the raw
data for risk assessment purposes. Information about all affected pathways may
contribute to an understanding of the mechanisms and MO As.
• Identify the genes and pathways affected over a critical window of exposure if
global gene expression time-course data are available. Specifically, by
developing a gene network over time, it may be possible to identify the earliest
affected genes and/or pathways, which in turn may represent the earlier or
initiating events for the outcome of interest.
Based on these recommendations, we refined our initial case-study approach to produce a
generalizable approach that can be used to evaluate genomic data in new chemical assessments.
1-6
-------
1.4. RESEARCH NEEDS
The following research needs could potentially improve the utility of genomic data in risk
assessment:
• Perform parallel toxicity and toxicogenomic studies with similar design characteristics
(i.e., dose, timing of exposure, organ/tissue evaluated) in order to obtain comparable
results which would aid our understanding of the relationship between gene expression
changes and phenotypic outcomes.
• Test multiple doses, with increased numbers of animals, in microarray and toxicity
studies (see bullet above) in order to relate the dose to the gene expression and pathway
response, and to the in vivo response.
• Perform a time-course global gene expression study over a relevant exposure interval
(e.g., critical window of development) in order to identify the earlier and possibly,
initiating gene expression events.
• Generate TK data in an appropriate study (e.g., time, dose, tissue), and obtain a relevant
internal dose measure to derive the best internal dose metric.
• Further develop bioinformatic methods for analyzing genomic data for the purpose of use
in risk assessment.
As a result of considering how to best use genomic data in risk assessment, we identified
a number of issues for future consideration. As more and various types of genomic studies are
performed, genomic data will likely inform multiple steps of the risk assessment process beyond
MOA. To facilitate the advancement of the use of genomics in risk assessment, first, we need
approaches to utilize genomic data quantitatively, specifically, the application of genomic data to
dose-response, intraspecies variability, and TK. Second, analytical methods tailored to use in
risk assessment are needed. Bioinformatics methods development work, some initiated in this
project, continues to evolve. The goal is to develop and/or adapt existing bioinformatic tools
currently used for hypothesis generation to the express purpose of utilizing genomic data for risk
assessment. The pathway activity level method presented in this report is a promising approach
for application to risk assessment. However, continued efforts, with input from both statistical
modeling and biology experts, is required to validate, test, and refine these methods. Third,
training risk assessors in genomic data analysis methods would assist EPA in the evaluation and
interpretion of complex, high-density data sets and in performing new analyses when necessary.
1-7
-------
Finally, some of the issues in utilizing genomic data in health and risk assessment are not
unique to genomic data but apply to precursor event information in general. Two of these issues
are (1) defining adversity and (2) establishing biological significance of gene expression changes
or pattern. The design and performance of appropriate studies, with both genomic and toxicity
components, may help to address the scientific aspects of these two important issues (see
research needs above).
To the best of our knowledge, this is the first systematic approach for using genomic data
in health assessment at EPA. We believe that this report can be used by risk assessors when
considering a large range of potential applications, issues, and methods to analyze genomic data
for future assessments. This approach advances efforts in the regulatory and scientific
communities to devise strategies for using genomic data in risk assessment, and it is consistent
with the pathway-based risk assessment vision outlined in the National Research Council's
(NRC's) report, Toxicity Testing in the 21st Century. We also anticipate that the research needs
and future considerations described herein will advance the design of future toxicogenomic
studies for application to risk assessment, and as a result, benefit the bioinformatic,
toxicogenomic, and risk assessment communities.
1-8
-------
3. DBF CASE-STUDY APPROACH AND EXERCISE
This chapter presents a description of the approach used to evaluate toxicogenomic data
in risk assessment, and a description of the first three steps of the DBF case study. Our strategy
for evaluating genomic data for risk assessment was to design a flexible yet systematic approach
that would provide a thorough evaluation of the genomic data set for a particular chemical, while
still accommodating different risk assessment practices. The discussion includes both
(1) general (i.e., not chemical-specific) considerations for evaluating a genomic data set, and
(2) consideration of the DBF genomic data set as part of the DBF case study.
3.1. EVALUATING THE EXTERNAL REVIEW DRAFT OF THE IRIS
TOXICOLOGICAL REVIEW (TOX REVIEW) OF DBF
The case-study approach begins with an evaluation of the 2006 external review draft IRIS
Tox Review for DBF (see Figure 3-1). Use of this draft assessment as the starting point allowed
us to take advantage of (1) the compilation of the toxicity and human data sets, allowing us to
focus on the toxicogenomic data set evaluation, and (2) data gaps that were identified, thus,
providing possible questions that the toxicogenomic data may be able to address.
The IRIS assessment of DBF was in progress and the internal review draft was available
when the DBF toxicogenomic case study project was initiated in 2005. The external review draft
of the Tox Review for DBF and IRIS Summary were released for public comment and peer
review on June 27, 2006 (U.S. EPA, 2006a;
http://cfpub.epa.gov/ncea/cfm/recordisplay.cfm?deid=l55707). The External Review Peer
Review Panel meeting was held July 28, 2006.
There are extensive studies documenting developmental toxicity of DBF and its primary
metabolite, monobutyl phthalate, in rodents (Barlow et al., 2004; Barlow and Foster, 2003;
Mylchreest et al., 2002, 2000, 1999, 1998; Ema and Miyawaki, 2001a, b; Ema et al., 2000a, b
1998, 1997, 1996, 1995, 1994, 1993; see Chapter 4 for further details). DBF exposure to the
developing male rat fetus during a critical window of development in late gestation causes a
variety of structural malformations of the reproductive tract (e.g., hypospadias); a decrease in
anogenital distance (AGD); delayed preputial separation (PPS); agenesis of the prostate,
epididymis, and vas deferens; degeneration of the seminiferous epithelium; interstitial cell
hyperplasia of the testis; and retention of thoracic areolas and/or nipples (Bowman et al., 2005;
3-1
-------
Evaluate
DBF IRIS
Assessment ERD
(Chapter 3)
Consider RA Aspects
I that Genomic Data Set
may Address
(Chapter 3)
Identify Case Study Questions
(Chapter 3)
Do the genomic data inform
(1) Mode or mechanism of action for male repro dev effects?
(2) Interspecies differences in MOA?
Toxicity Data Set
Evaluation
(Chapter 4)
iGenomic Data Set
Evaluation
(Chapter 5)
IGenomic Data Set
New Analyses
(Chapters 5 & 6)
Case Study Findings
(Chapter 7)
1) Putative additional pathways to further understand MOA
2) Cross-species conservation information for
steroidogenesis pathway
Application to RA
(Chapter 7)
•Generic Approach
•Research Needs
•Recommendations
^M_ _^M
Figure 3-1. DBF case-study approach for evaluating toxicogenomic data for
a health assessment. Evaluation steps in the case-study process are shown in
rectangles. Findings or products of the case study are shown in ovals. ERD,
external review draft. Numbers in parentheses indicate report chapters where the
case-study step is described.
5-2
-------
Kleymenova et al., 2005a; Barlow et al., 2004; Kim et al., 2004b; Barlow and Foster, 2003;
Fisher et al., 2003; Higuchi et al., 2003; Mylchreest et al., 2002, 2000, 1999, 1998; Ema et al.,
2000b, 1998, 1997, 1994; Saillenfait et al., 1998).
Figure 3-2 shows the studies that were candidates for the development of the reference
dose (RfD) presented in the 2006 external review draft IRIS Tox Review for DBF (U.S. EPA,
2006a). The point of departure (POD) selected for derivation of the RfD for all exposure
durations (acute, short-term, subchronic, and chronic) was the no-observed-adverse-effect level
(NOAEL) of 30 mg/kg-d for reduced fetal testicular T (Lehmann et al., 2004). In this study, a
statistically significant decrease in T concentration in the fetal testis was detected at 50 mg/kg-d.
The reduction in fetal testicular T is one of the well-characterized MO As for DBF that occurs
after in utero DBF exposure (during the critical window), initiating the cascade of events for a
number of malformations in the developing male reproductive tract. Studies using
radioimmunoassay of T levels in fetal testes and studies using RT-PCR, microarrays, and/or
immunochemical staining found a decrease in the expression of protein and mRNA for several
enzymes in the biochemical pathways for cholesterol metabolism, cholesterol transport, and
T biosynthesis (also called steroidogenesis more generally) in the fetus (Plummer et al., 2005;
Thompson et al., 2004, 2005; Lehmann et al., 2004; Liu et al., 2005; Barlow et al., 2003; Fisher
et al., 2003; Shultz et al., 2001). Collectively, these studies document that exposure to DBF
disrupts T synthesis in the fetal testis. Thompson et al. (2004) established that following in utero
exposure to 500 mg/kg-d, the T levels in the testes return to normal after the metabolites of DBF
are cleared from the circulation. However, the malformations induced by 500 mg/kg-d exposure
persist into adulthood (Barlow et al., 2004; Barlow and Foster, 2003). Thus, although the
inhibition of T synthesis can be reversed, the biological effects resulting from the decrease in T
during the critical developmental window are irreversible.
3.2. CONSIDERATION OF RISK ASSESSMENT ASPECTS THAT
TOXICOGENOMIC DATA MAY ADDRESS
While microarray and RT-PCR data have been used to inform the MO As of a chemical,
the many types of -omic data have the potential to inform TK, dose-response, interspecies and
intraspecies differences in TK or TD, and be utilized as biomarkers of exposure or effect (see
Figure 3-3).
5-3
-------
3000 -
1000 -
_^ 300 -
£ 100 -
*• 30 -
1 10 -
y>
o i —
Q '
0.3 -
0.1 -
j
<
4
L
1
J
t ~*
i
t-
» -4
~r T T
-i
L
<
>-
t T JL
1 0 ° A J
> — ~ •*- -4
t '
^ •<
-j
^
>- <
t -•-
>
--
Developmental Testicular Hepatic Neuro
CD ^r ^~
^ O CM
I 0 I
CM CM O
'-" ~m Q
° OJ °
c
ffl
E
TT
UD ^ — O ^ —
CD CM O CM
CD 1 CD ~
^ CM CM LJ
o_" Tii -^r cE
' Q 4-1 o
~Z. Q 0) ^
o3 Q
2 O
u
— - ^ — - O1O CDLO U1LO CDLO CD
LD Ul
X CMX >*CD CJ1CD ^^CD C7JCD CDCD ^
H ,-,?S "^CD COCD 1>CD COCD CDCD d>
LJ^-^ TU^ 14—^ CD^ 4—^ 4—^ CL>
" 2^ LO- °- 3- °- °
' Q_ T ^ := olQ- rv^Q- OlCL 01
CD ora — Zj ^"' — ^' — ^"' — ^^
C!)^CD<™5 -§ 3 3
roCl^j S^ ^ ^~ °°
( — (T) Ql O
N J S S Q
•; — f — i < — i _
B:" «
1 — ^
!g
<
LO T- „ - 01 CM
CD CM Q ^ CD
^^ O O ^ CD
^ O
1 Q CD 4-1 UJ
IO ^ "0 CD
*-
-------
Screening & Prioritization Programs
Risk Assessment
Exposure TGxdata
biomarhe
Hazard Identification
(Screening & Characterization
Exposure Assessment
TK < TGxdata
Risk Characterization
Intra & mterspecies
uncertainty and variability
Risk Communication/Management
Figure 3-3. Potential uses of toxicogenomic data in chemical screening and
risk assessment. -Omic data from appropriately designed studies have the ability
to inform multiple types of information and in turn, steps in screening and
prioritization, and risk assessment. Arrows with TGx data indicate the types of
information these data can provide. Shaded boxes indicate some of the types of
information that are useful in risk assessment. TGx, toxicogenomic.
5-5
-------
However, in this case study, chemical screening and exposure assessment were not
considered. Instead, we considered the many types of information useful to hazard
characterization, dose-response, and risk characterization. Toxicogenomic data have been
successful in providing information about the molecular events altered in the mechanisms of
action, and, at times, information about the TD or TK key events of a MO A (see Figure 3-4; see
Chapter 2). Further, data from appropriately designed toxicogenomic studies could be used to
inform intra- and interspecies differences in molecular responses.
TK
TD
i r
=
a
3
-M
s
Exposure
— *
»v
5 T"6* rfato >
k
Exposure
A
E
*
*
Intra & Interspecies
differences
Target
tissue
dose
*
Biol.
effective
dose
^
Altered
molecular
events
+•
Altered
cellular
events
+•
Altered
tissue
^
Altered
organ
Observed
outcome/
endpoint
Observed
outcome/
endpoint
Figure 3-4. Potential uses of toxicogenomic data in understanding
mechanisms of action. The process from exposure to outcome encompasses all
of the steps of a mechanism of action, including both TK and TD steps. Available
TGx data, such as microarray data and other gene expression data, can provide
information about altered molecular events, at the gene expression level. In turn,
appropriate TGx data can be used to inform intra- and interspecies differences in
molecular responses. Appropriate TGx data could also inform internal dose and
intra- and interspecies differences in internal dose. ADME, absorption,
distribution, metabolism, and excretion.
-------
3.2.1. Informing TK
Characterizing the absorption, distribution, metabolism, and excretion (ADME) of
environmental toxicants is important for both the understanding and application of MO A
information in predicting toxicity. Differences in TK across species, individuals, and exposure
patterns (routes, level, duration, frequency) can lead to different biological effects for the same
total exposure to a chemical. It is well-established that a quantitative understanding of chemical
TK (e.g., using PBPK models) can be useful in analyzing dose-response data and extrapolating
across species, individuals, and exposure patterns (U.S. EPA, 2006e). The principles of these
uses for TK are the same regardless of the types of response data utilized (i.e., in vivo toxicity
endpoints [e.g., pup weight] or molecular precursor events [e.g., toxicogenomic changes]), and
will not be reviewed here. However, the inverse question of how toxicogenomic data can inform
TK has not been fully explored. Here we consider whether toxicogenomic data could be useful
for understanding four aspects of a chemical's TK: (1) identification of potential metabolic and
clearance pathways; (2) selection of an appropriate dose metric; (3) intra- and interspecies
differences in metabolism; and (4) TK/TD linkages and feedback. Each of these applications is
discussed below. Finally, the available toxicogenomic data for DBF are evaluated for use in
informing TK.
3.2.1.1. Identification of Potential Metabolic and Clearance Pathways
While TK studies, themselves, are designed to help understand the pathways for
metabolism and clearance of xenobiotics, toxicogenomic data may provide important
complementary information as to what enzymes and tissues may be involved in metabolism. For
example, many xenobiotics induce the expression of the Phase I and II enzymes that are
responsible for their clearance. Thus, toxicogenomic data showing changes in the expression of
genes, such as cytochrome P450s in a particular tissue, may implicate their involvement in
metabolizing the compound. While such toxicogenomic data may confirm the major sources of
metabolism or clearance, they may also identify minor TK pathways relevant for inducing
toxicity. However, data indicating gene expression changes alone are insufficient to conclude
that there is a corresponding increase in a protein levels or activity, or are relevant to the ADME
of the chemical of interest. Ultimately, toxicogenomic data may be most useful for generating
5-7
-------
hypotheses about metabolism and clearance pathways that can be tested with additional TK
studies.
3.2.1.2. Selection ofAppropria te Dose Metrics
Due to inherent differences in TK across species, individuals, and exposure patterns,
dose-response relationships are best established based on an internal measure of a biologically
effective dose as opposed to an external or applied dose. However, an understanding of TK
alone may provide multiple options for the internal "dose metric," such as blood or tissue
concentrations of the parent or metabolites, or rates of formation of reactive compounds. Thus, a
key question in utilizing TK data for dose-response analyses and extrapolation is dose-metric
selection, which depends on the determination of the active chemical species and the MO As of
toxicity. There often may be more than one biologically plausible choice of dose metric, which
contributes to the uncertainty in the dose-response analysis. The potential utility of
toxicogenomic data is that gene expression data may demonstrate earlier biological effects, and,
thus, are closer both spatially and temporally to the interaction between the active chemical
species and endogenous cellular molecules than traditional toxicological outcomes (see
Figure 3-4). Thus, toxicogenomic data can, in principle, provide biological support for the
choice dose metric. Different predictions for internal dose can be statistically analyzed along
with toxicogenomic changes that inform TD to determine the dose metric that is best correlated.
3.2.1.3. In tra - and In tersp ecies Differ en ces in Met a b olism
Data from polymorphisms is one type of genomic data that can be extremely useful to
informing intraspecies differences. Across species, data on differential expression of different
isozyme genes may be indicative of differences in overall metabolizing capacity and affinity. In
addition, toxicogenomic data may be informative about whether the tissue distribution of
metabolizing enzymes may be different across species. Within species, interindividual
variability in metabolizing capacity and/or affinity due to differences in enzyme expression or
genetic polymorphism can greatly influence the overall TK of a chemical. For example, genetic
polymorphisms in aldehyde dehydrogenase-2 (ALDH2) can result in an increase in blood
acetaldehyde levels following alcohol consumption, thereby leading to overt health effects
(Ginsberg et al., 2002). Similarly, data on CNPs can provide information (Buckley et al., 2005)
-------
that directly informs TK. For example, some individuals possess different copy numbers of
CYP2D6 that influence their response to pharmaceuticals (Bodin et al., 2005). When the
impacts of gene expression levels and polymorphisms on enzyme levels and function have been
established (i.e., preferably confirmed by measurement of enzyme level), this information can
either be used to characterize the difference in a predicted dose metric for a subpopulation
relative to the most common alleles, or it can be used in probabilistic (e.g., Monte Carlo)
analyses to characterize the impact on population variability.
3.2.1.4. TK/TDLinkages andFeedback
Ultimately, toxicogenomic data may be useful for linking together TK and TD models
into more comprehensive biologically based dose-response (BBDR) models (Daston, 2007).
With an appropriate dose metric, one can link the TK predictions for a chemical (e.g., tissue
concentration of a metabolite) with toxicogenomic changes (e.g., change in mRNA level) that, in
turn, are linked through a TD model to alterations in cellular constituents and, ultimately, frank
effects. Furthermore, toxicogenomic data may be useful in providing the link by which the TD
feedback of gene and protein expression changes on TK (e.g., enzyme induction) can be
modeled.
3.2.1.5. R esearch Needs for Toxicogen omic Studies to Inform TK
Changes in gene expression can be highly labile and vary as a function of dose and time.
Thus, identification of appropriate dose metrics involves detection of relevant gene changes as
well as the moiety that caused the changes. Therefore, simultaneous data collection of
toxicogenomic data and tissue concentrations of the relevant chemical species would be
beneficial. In order to inform interspecies extrapolation, it is important to mine toxicogenomic
data for potential indicators of species differences in metabolism. For intraspecies variability, it
is important to assess the potential impact of polymorphisms in Phase I and II enzymes.
Microarray data may also be useful for identifying life stage and gender differences in relative
expression of enzymes involved in the TK of the chemical of interest.
5-9
-------
3.2.1.6. DBF Case Study: Do the Available Toxicogenomic Data Inform TK?
We considered whether the available toxicogenomic data set for DBF informs TK. A
greater level of detail is presented for TK in this chapter than for TD because the latter subject is
considered in greater detail in the subsequent chapters. This chapter also provides examples of
considerations that may be helpful to risk assessors examining whether the available
toxicogenomic data can inform TK for their chemical of interest.
The TK of DBF is reviewed in U.S. EPA, 2006a, and is summarized briefly here for
context. Following ingestion, DBF is primarily hydrolyzed to monobutylphthalate (MBP) in the
gastrointestinal tract and enters systemic circulation through the portal blood. MBP undergoes
glucuronidation in the liver, and both free and glucuronidated MBP circulate in serum and are
subsequently excreted in urine. While there are a number of TK studies in rats, little human TK
data are available, particularly for known exposures to DBF. The available data suggest that free
MBP is responsible for the effects on T biosynthesis. In terms of TK pathways, the data set did
not lead to the identification of alternative metabolic pathways for DBF.
Toxicogenomic data could inform dose-metric selection in two broad ways: relating the
metabolite to the gene expression or using gene expression as the dose metric. In a more
traditional approach, changes in the expression for genes of interest could be related to a
chemical moiety in a target tissue of relevance (or convenience). For example, Lehmann et al.
(2004) provides a dose-response analysis of gene expression following DBF exposure.
However, these data are limited for use in extrapolation without TK data (e.g., tissue
concentrations of MBP). Ideally, TK data could be collected at various time points following
various doses, but this would require a large number of fetuses. In the absence of such empirical
data, analyses could be performed using physiologically based TK modeling, but none have yet
been attempted. Such an approach might utilize the available published TK studies for DBF and
attempt to reconstruct the exposure scenarios in the toxicogenomic studies with the intent to
predict the MBP concentration in a target tissue (or blood) at the developmental time points
where toxicogenomic samples were obtained.
A second and more complex approach might be to use a toxicogenomic change as a
dosimeter (or "biomarker"), which may obviate the need for TK data altogether. For example,
the microarray study of Wyde et al. (2005) reports changes in maternal liver Cyp2bl and
Cyp3al, and estrogen sulfotransferase mRNA levels following DBF exposure. Not only do
3-10
-------
these gene expression changes serve as potential biomarkers, but they also suggest that there may
be related changes in metabolic biomarkers (i.e., metabonomics) because these enzymes have
roles in lipid and hormone synthesis, in addition to xenobiotic metabolism. Although it is not
clear whether these changes have a relationship to a toxic endpoint of interest, it may be possible
to establish, for instance, that an increase in a specific maternal liver mRNA is correlated with a
decrease in a specific mRNA in the fetal testis. Indeed, Wyde et al. (2005) show that maternal
liver estrogen sulfotransferase gene expression increases in a dose-dependent manner from
10-500 mg/kg-d, while Lehmann et al. (2004) observed a dose-dependent decrease in Scarbl,
Star, Cypllal, Hsd3b, or CyplJal mRNA levels in fetal testes from 0.1-500 mg/kg-d.
With respect to interspecies extrapolation and interindividual variability, the lack of
adequate human TK data precludes quantitative extrapolation, a situation that cannot be
remedied with the available toxicogenomic data (unless, as discussed above, a
toxicogenomic-based dosimeter/biomarker was developed). For instance, available blood
measurements of MBP in humans were taken from spot samples in the general population where
the individual exposure patterns were unknown. Although differences were observed in the ratio
of free to conjugated MBP in human serum as compared to the rat, these data are insufficient for
quantitative interspecies extrapolation, because in order to replace administered dose as a dose
metric, it is necessary to determine the absolute., not the relative, level of free MBP in serum as a
function of exposure. The Wyde et al. (2005) study suggests that DBF-induced enzyme
induction occurred. Specifically, this study reported that exposure to 50 and 500 mg/kg-d DBF
leads to an increase in rat liver UDP glucuronsyltransferase 2B1 (Ugt2bl) mRNA levels. More
TK analysis would be required to ascertain whether this induction in rats occurs at levels that are
relevant to low-dose exposures. This enzyme induction may occur in humans and such a
response may increase interindividual sensitivity to DBF toxicity. With regard to human TK,
none of the available toxicogenomic data on DBF were performed in humans and thus, do not
provide any information on DBF interindividual TK variability. For example, there are no
available data on polymorphisms in glucuronyltransferases responsible for metabolizing MBP.
Finally, we considered the potential for TK/TD linkages with the available data and concluded
that in order for TK and toxicogenomic data to be integrated for use in quantitative
dose-response analysis, more sophisticated BBDR models are needed. Using such an approach,
3-11
-------
it may be feasible to relate changes in expression of genes involved in T production to quantify
testicular T levels (see Figure 3-5).
HDLCholesteryl Ester
Figure 3-5. The fetal Leydig cell in the fetal testis. The boxes represent genes
involved in the biosynthesis of T; the percentages (%) represent % control gene
expression in fetal testis of dams treated with 500 mg/kg-d DBF.
Source: Adapted from Barlow et al. (2003).
The male reproductive developmental effects of DBF appear to be mediated by MBP,
which causes a down-regulation of cholesterol transporters across the cell membrane (SCARE 1)
and mitochondrial inner membrane (STAR), as well as the down-regulation of two enzymes
involved in converting cholesterol to T, CYP11 Al, and CYP17A1 (Liu et al., 2005; Lehmann et
al., 2004; Barlow et al., 2003; Shultz et al., 2001). Thus, it may be possible to relate DBF and/or
MBP levels to reductions in cholesterol transporter (e.g., SCARE 1 and STAR) and the levels of
steroidogenic enzymes (e.g., CYP11A1 and CYP17A1) at the mRNA, protein, and/or activity
3-12
-------
levels. Changes in these parameters may then be modeled to predict changes in testicular T
levels, which may subsequently be correlated to developmental toxicity endpoints.
3.2.2. Informing Dose-Response
Toxicogenomic data that informs TK can be useful for informing or improving
dose-response analysis because it may improve the dose metric selection among alternative dose
metrics. However, use of toxicogenomic data as an endpoint in dose-response analysis has not
been extensively explored. For example, BMD analysis of some dose-response studies
determined PODs based on the GO categorization of gene expression changes (based on an
approach of Yu et al., 2006) as a function of dose (Andersen et al., 2008; Thomas et al., 2007).
3.2.2.1. DBF Case Study: Do the Toxicogenomic Data Inform Dose-Response?
Unfortunately, there are currently no available dose-response microarray studies to assess
the genome-wide gene expression over a dose range. However, there is one available
dose-response gene expression study for DBF. Specifically, Lehmann et al. (2004) conducted a
dose-response study evaluating testicular T, RT-PCR and protein expression for a subset of
genes thought to underlie the male reproductive developmental outcomes. This study reported a
significant reduction in fetal testicular T at 50 mg/kg-d DBF or higher. Western analysis found
that steroidogenic acute regulatory protein (STAR) and scavenger receptor class B, member 1
(SCARB1) were significantly decreased at 50 mg/kg-d while cytochrome P450, family 11,
subfamily a, polypeptide 1 (CYP11 Al) was only reduced at 500 mg/kg-d. Further, RT-PCR
analysis findings confirmed that the mRNA of these three genes was statistically significantly
reduced at 50 mg/kg-d. The results of this study support the role of steroidogenesis enzymes and
cholesterol transport proteins in the decreased testicular T MOA after in utero DBF exposure.
However, without first establishing the biologically significant level of change in gene
expression and the critical subset of genes that constitute a well-established precursor event, it is
difficult to use these data in a dose-response assessment (see Chapter 7). It would be helpful to
have dose-response microarray or proteomic studies to assess mRNA and protein expression on a
genome-wide level.
3-13
-------
3.2.3. Informing TD
There are numerous examples where toxicogenomic data have been used to inform the
TD steps within mechanisms of action or MO As for a chemical, and there are a small number of
examples where such data have been used corroboratively for risk assessment decisions (see
Chapter 2).
3.2.3.1. General Considerations: TD Portion of Mechanisms of Action and MO As
One feature of the approach (see Figure 3-1) is the evaluation of the toxicity and
toxicogenomic data sets in conjunction in order to consider the relevance of gene expression
changes with respect to specific endpoints of interest identified in the toxicity data set. In this
manner, data on affected pathways may generate hypotheses and inform the mechanisms of
action for a chemical for specific endpoints. In addition, using this approach could provide
connections between affected pathways (identified from the toxicogenomic data set) and
endpoints (identified from the toxicity data set), which may, in turn, inform modes or the
mechanisms of action, as illustrated by Figure 3-6. Chapter 2 and the glossary describe the
distinction between the definitions for mechanisms of action and MO As.
This approach is best suited to instances where comparable study designs between the
toxicity/epidemiology and toxicogenomic data sets are available. For example, toxicogenomic
and toxicity studies performed in the same species, using similar doses, similar exposure
intervals, and assessing the same organ or tissue would be ideal for utilizing this approach. For
the DBF toxicity (see Chapter 4) and toxicogenomic data sets (see Chapter 5), there is some
comparability across some of the studies—i.e., some toxicity and toxicogenomic studies were
performed at the same doses with similar exposure intervals, in the same species and strain, and
assessed some of the same organs (e.g., testis). However, no two studies are the same for all
study-design aspects, such as precise timing of exposure and time of assessment.
3.2.3.2. DBF Case Study: MO As for Male Reproductive Developmental Effects
Developmental toxicity studies (reviewed in Chapter 4) and toxicogenomic studies
(reviewed in Chapter 5) have contributed to a good understanding of DBF as a chemical that has
multiple MO As. Two well-characterized MO As: a reduction in fetal testicular T and a reduction
in Insl3 signaling activity, explain a number of the observed male reproductive developmental
3-14
-------
Pathways
TGx
DATA SET
i
Comparable study design characteristics, e.;
-Species -Organ
-Strain -T of exposure
-Dose
Figure 3-6. Approach to utilizing toxicity and toxicogenomic data for
identifying affected pathways and candidate modes and mechanisms of
action. Toxicogenomic data can be analyzed for differentially expressed genes
(DEGs) and, in turn, grouped into affected pathways. Toxicity data can provide
information about affected endpoints. Evaluating toxicogenomic and toxicity data
together can provide a level of phenotypic anchoring between gene and pathway
changes, and in vivo outcomes. The identification of affected pathways can
inform mechanisms of action, including MO As, for a chemical. Such an
approach requires similar study parameters (e.g., dose, species, duration of
exposure) for the toxicity and toxicogenomic studies. TGx, toxicogenomic.
abnormalities. Some other observed abnormalities are not explained by these two MO As,
suggesting that there are additional MO As for DBF. Acknowledging that there are additional
data not presented in Figure 3-7, this figure attempts to show where there is agreement in the
scientific community, based on reproducibility of microarray and RT-PCR studies, about
affected pathways and the well-characterized MO As for DBF. There are some endpoints and
pathways that need further characterization and, as a result, we were interested in determining
whether the toxicogenomic data could be used to associate the DBF MO As and endpoints, and/or
form hypotheses about additional MO As for DBF.
3-15
-------
Sertoli Cell
Cholesterol
T Cholesterogenesis
I Lipogenesis
Steroidogenesis 3P-HSD
|p450c!7
Disrupted
Sertoli Cell \^ Gonocyte
evelopment
a inhibition
Fetal Leydig Cell
Testosterone
Male developmental
reproductive effects:
Undescended testes
Altered repro tract dev
Multmucleated gonocyte
Figure 3-7. The proposed DBF mechanism of action for the male reproductive developmental effects. The
mechanism of action is defined as all of the steps between chemical exposure at the target tissue to expression of the
outcome. The steps shown are based on male reproductive developmental toxicity and toxicogenomic studies. Some
of the affected pathways and individual genes whose expression was significantly affected after DBF exposure in
multiple studies are included. By contrast, MO As are shown in purple letters including two well-characterized MO As
and one example of an unidentified MOA.
Source: Figure adapted from Liu et al. (2005), Thompson et al. (2004), Wilson et al. (2004), Barlow et al. (2003), and
Shultzetal. (2001).
-------
3.3. IDENTIFYING AND SELECTING QUESTIONS TO FOCUS THE DBF CASE
STUDY
In reviewing the external review draft IRIS Tox Review for DBF, data gaps in the
assessment were noted. Then the DBF toxicogenomic data set was evaluated for these data
could potentially address any of the gaps (see Figure 3-1). The identified data gaps led to
formulation of questions of interest whose answers may be able to contribute valuable
information to a risk assessment. The following questions were identified:
Can the DBF toxicogenomic data set inform the
• biologically significant level of reduction in fetal T? As the external review draft IRIS
Tox Review for DBF used a reduction in fetal testicular T as the critical effect, we
considered whether the toxicogenomic data set could aid in determining the biologically
meaningful level of T reduction for the male reproductive developmental effects.
• dose-response assessment in risk assessment? The microarray and RT-PCR studies
have identified genes and pathways associated with the reduced fetal testicular T. Thus,
there is the potential for evaluating these genes and pathways in a dose-response
assessment.
• modes and mechanisms of action for male reproductive developmental outcomes?
Not all of the male reproductive developmental outcomes after in utero DBF exposure
are a consequence of reduced fetal testicular T or reduced Insl3 expression. Therefore,
additional MO As for these endpoints may be identified from pathway analysis of the
microarray data.
• interspecies (rat-to-human) differences in MOAs that could, in turn, inform the TD
part of the UFH? There is evidence from toxicogenomic studies that a reduction in gene
expression of some of the steroidogenesis genes underlies the reduction in fetal testicular
T observed after in utero DBF exposure. Unfortunately, there are no genomic studies in
appropriate human in vitro cell systems to make comparisons to in vivo rat MOA
findings. Using available DNA sequence data and other methods, we would like to
assess the rat-to-human conservation of the steroidogenesis pathway genes.
The existing genomic data for DBF had the potential to inform two of the questions:
informing modes and mechanisms of action and interspecies differences for the reduced T MOA
(see above). It was highly likely that the DBF toxicogenomic data set could aid in hypothesis
generation of DBF modes or mechanisms of action. Using genomic data, such as DNA sequence
data, it may be possible, but less likely, to inform cross-species differences in TD for the reduced
T MOA. Although the other two questions (see list above) were of great interest, the available
genomic data were not considered appropriate to address them.
3-17
-------
The remaining steps of the DBF case study are presented in the subsequent chapters. The
evaluations of the toxicity data set for the male reproductive developmental effects after
developmental exposure to DBF (see Chapter 4) and the toxicogenomic data set including new
analyses of one microarray study (see Chapter 5). Exploration of pathway analysis methods
development for applying microarray data to risk assessment and the use of available methods to
evaluate rat-to-human differences for the reduced T MO A are presented in Chapter 6. Chapter 4
follows with an in-depth evaluation of the DBF toxicity data set.
3-18
-------
4. EVALUATION OF THE REPRODUCTIVE DEVELOPMENTAL TOXICITY
DATA SET FOR DBF
This chapter presents the evaluation of the available toxicity data for the development of
the male reproductive system following DBF exposure and the MOA(s) (see Chapter 2 and
glossary for definition) that contribute to these outcomes. We used the compilation of the male
reproductive toxicology literature cited in the 2006 external peer review draft IRIS Tox Review
for DBF (U.S. EPA, 2006a) as a starting point for our toxicology literature review for this case
study. Each toxicology study was examined for the lowest dose and low-incidence effects in
order to identify the full spectrum of male reproductive developmental effects. In a second
evaluation, we used available mechanistic information for each endpoint to identify potential
MO As. Endpoints with MO A information have support for phenotypic anchoring to some of the
observed DBF gene expression changes (further discussed in Chapter 5). Endpoints with
unexplained MO As were used to identify and focus future research needs to study the
mechanisms that underlie those endpoints using genomics and other techniques.
An extensive toxicological data set exists for DBF that includes acute and subchronic
studies in multiple species, multigeneration reproduction studies in rodents, and studies that
assess developmental outcomes following in utero or perinatal/postnatal exposures. Following
DBF exposure during the critical stages of development, the male reproductive system
development is perturbed in rodent studies (Gray et al., 1999, 2001; Mylchreest et al., 1998,
1999, 2000). Two MO As of DBF, for a number of these outcomes, have been well established
(David, 2006; Foster, 2005). The 2006 external draft IRIS Tox Review for DBF (U.S. EPA,
2006a) selected reduced fetal testicular T levels, observed in Lehmann et al. (2004), as the
critical effect for the derivation of acute, short-term, subchronic, and chronic reference values for
DBF. This case study evaluated information from genomic and other gene expression studies to
target and further elucidate the molecular events underlying these developmental outcomes (see
Chapter 5). The intent of performing this evaluation of the toxicology studies was to examine
the usefulness of the toxicogenomic data in characterization of the MOA(s) that contribute to the
adverse outcomes. We also examined the data for low-dose or low-incidence findings because
such data may aid the interpretation of toxicological outcomes that can be misinterpreted as
transient (e.g., AGD), or nonadverse due to low incidence or magnitude (e.g., not statistically
4-1
-------
significant incidences of gross pathology findings in male offspring reproductive organs, or
alteration of fetal T levels).
4.1. CRITERIA AND RATIONALE FOR INCLUSION OF TOXICOLOGY STUDIES
IN THE EVALUATION
Figure 4-1 illustrates the process for evaluating the DBF toxicology data set for the case
study (Section 4.2 discusses the later steps of the evaluation process in more detail). The first
step in the process was the identification of studies to be included for consideration in the case
study. We identified a number of study selection criteria in Step 1. One criterion of prime
importance was that the studies should include exposures to DBF during sensitive periods of
male reproductive system development. Secondly, a no-observed-effect level (NOEL),
lowest-observed-effect level (LOEL), or BMDL would need to be identified for presumably
adverse outcomes in the reproductive organs and/or function of male offspring. Additionally, the
studies would need to be of adequate quality in order to establish confidence in the study
conduct, methods, and results. These criteria, taken together, define a subset of the available
toxicology studies that were considered possible candidates for determining the POD for
derivation of reference values of various exposure durations in the 2006 external peer review
draft Tox Review for DBF (see Tables 4-1, 4-2, and 4-3 in U.S. EPA, 2006a). These candidate
study lists were considered during the external peer review of the IRIS document, conducted in
July 2006, thereby providing a measure of confidence in their inclusiveness and veracity for the
purpose of this case study. Though there are observable adverse effects on male reproductive
system development in multiple species, the only available and relevant genomic studies with
DBF (i.e., those that addressed effects on male reproductive system development following
prenatal exposures) were conducted in rats. Table 4-1 lists the studies that were identified for
inclusion as of July 2006. For each study, the following information was summarized: a
description of the dose and exposure paradigm, the treatment-related outcomes observed at each
dose level, and the experimentally derived reproductive NOEL and/or LOEL. The terms
NOAEL and LOAEL are not used in this case-study report, although these terms are commonly
used in risk assessment, because some study reports do not address the issue of adversity of
observed study outcomes. In addition, some study reports do not specifically define NOELs or
LOELs. For that reason, Table 4-1 presents those outcomes that could be considered biomarkers
4-2
-------
of effects on the male reproductive system that were reported by the study authors, without
specific consideration or judgment of adversity.
STEP 1: Evaluate toxicology studies cited in IRIS
assessment, for inclusion in case study
STEP 2: For each study, identify NOEL/LOEL
and examine data for low incidence and
low dose findings
Problems with
quality of
reported data;
individual data
not available
Individual data
available
STEP 3: Examine individual data for each male repro
outcome/endpoint
A /"
Thorough
assessment of
data
Issues in
evaluating data
II
Study or
reporting
deficits prevent
further analysis
of data
Data at NOEL
appear to be
biologically
relevant
J \.
Interpret or
corroborate
study outcomes
& endpoints
Figure 4-1. The process for evaluating the male reproductive developmental
toxicity data set for low-dose and low-incidence findings. IRIS assessment, the
2006 external peer review draft IRIS Tox Review for DBF.
-------
Table 4-1. Studies with exposures during development that have male reproductive outcomes (limited to
reproductive organs and/or reproductive function) and were considered adequate for reference value
determination
Study3
Barlow and
Foster, 2003
Barlow et al.,
2003
Barlow et al.,
2004
Bowman etal.,
2005
Carruthers and
Foster, 2005
Emaetal., 1998
Ema et al., 2000b
Ferraraetal.,
2006
Species (strain), duration,
and exposure
Rat (SD); CDs 12-21; 0 or
500 mg/kg-d
Rat (SD); CDs 12-19;
500 mg/kg-d
Rat (SD); CDs 12-21; 0,100,
or 500 mg/kg-d
Rat (SD); CDs 12-19 or 21; 0
or 500 mg/kg-d
Rat (SD); CDs 14-15, 15-16,
16-17, 17-18, 18-19, 19-20; 0
or 500 mg/kg-d
Rat (Wistar); CDs 11-21;
0,331, 555, or 661 mg/kg-d
Rat (Wistar); CDs 15-17;
0, 500, 1,000, or 1,500 mg/kg-d
Rat (Wistar); CDs 12-14, or
GD 20; 0, 1,000, or
1,500 mg/kg-d
Rat (Wistar); CDs 13.5-21.5;
0 or 500 mg/kg-d
Reproductive system effects
Large aggregates of Leydig cells, multinucleated gonocytes, and an
increased number of gonocytes in fetal testes; a decreased number
of spermatocytes on PNDs 16 and 21; epididymal lesions
(decreased coiling of the epididymal duct, progressing to mild
[PND 45], and then severe [PND 70] seminiferous epithelial
degeneration).
Large aggregates of Leydig cells with lipid vacuoles.
Testicular dysgenesis (proliferating Leydig cells and aberrant
tubules); decreased AGO; areolae retention; small incidence of
Leydig cell adenomas.
Marked underdevelopment of the Wolffian ducts (characterized by
decreased coiling).
Decreased AGO; retained areolae and nipples; reduced epididymal
weights, increased testes weight due to edema; malformations of
the seminal vesicles, agenesis of various regions of the epididymis,
small or flaccid testes; malformation of the coagulating gland.
At 555 and 661 mg/kg-d, increased incidences of cryptorchidism
and decreased AGO.
At 1,500 mg/kg-d, cryptorchidism observed in 80% of litters.
At 500, 1,000, and 1,500 mg/kg-d, decreased AGO.
At 1,500 mg/kg-d (CDs 12-14), cryptorchidism observed in 50%
of litters.
At 1,000 and 1,500 mg/kg-d, decreased AGO.
Delayed entry of gonocytes into quiescence, increase in gonocyte
apoptosis, and subsequent early postnatal decrease in gonocyte
numbers (exposures: GDs 13.5-17.5); >10% increase in
multinucleated gonocytes (exposures: GDs 19.5-21.5).
NOEL
mg/kg-d
100
331
LOEL
mg/kg-d
500
500
500
500
500
555
500
1,000
500
-------
Table 4-1. (continued)
Study3
Fisher etal.,
2003
Gray etal., 1999
Kim et al., 2004
Ab
Kleymenova et
al., 2004 Ab
Kleymenova et
al., 2005aAb
Kleymenova et
al., 2005b
Species (strain), duration,
and exposure
Rat (Wistar); CDs 13-21; 0 or
500 mg/kg-d
Rat (Long-Evans) (PO); PND
21— adult; 0, 250, 500, or
1,000 mg/kg-d
Rat (Long-Evans) (Fl);
GD 0-PND 21; 0, 250, 500, or
1,000 mg/kg-d
Rat (Long-Evans) (Fl); GD 14
to PND 3; 0 or 500 mg/kg-d
Rat (SD); CDs 10-19; 0, 250,
500, or 700 mg/kg-d
Rat (strain not specified);
CDs 12-17, 19, 20; 0 or
500 mg/kg-d
Rat (SD); CDs 12-20; 0, 0.1, 1,
10, 30, 100, or 500 mg/kg-d
Rat (SD); CDs 12-21; 0 or
500 mg/kg-d
Reproductive system effects
Cryptorchidism, hypospadias, infertility, and testis abnormalities
similar to human testicular dysgenesis syndrome; abnormal Sertoli
cell-gonocyte interaction.
At 250, 500, and 1,000 mg/kg-d, delayed puberty.
At 500 and 1,000 mg/kg-d, reduced fertility related to testicular
atrophy and reduced cauda epididymal sperm numbers.
At 250 and 500 mg/kg-d, reproductive malformations (low
incidences of hypospadias, testicular nondescent, and uterus
unicornous); reduced fecundity.
Reduced AGO, retained nipples, permanently reduced
androgen-dependent tissue weights.
Decreased testes and accessory sex organ weight; delayed testis
descent; increased expression of estrogen receptor in testes.
Altered proliferation of Sertoli and peritubular cells; multinucleated
gonocytes; changes in Sertoli cell-gonocyte interactions.
At 30 and 50 mg/kg-d, disruption of Sertoli-germ cell contact.
At 50 mg/kg-d, Sertoli cell hypertrophy, decreased total cell
number and number of seminiferous tubules.
At 100 mg/kg-d, increased multinucleated gonocytes.
Cytoplasmic changes in Sertoli cells with abnormal cell-cell contact
with gonocytes, clustering of gonocytes in the middle of the
tubules, altered morphometry of seminiferous tubules, clusters of
interstitial cells, decreased number of tubular cross sections per
testicular section; increased number of multinucleated gonocytes.
NOEL
mg/kg-d
10
LOEL
mg/kg-d
500
250
250
500
250
(presumed) b
500
30
500
-------
Table 4-1. (continued)
Study3
Species (strain), duration,
and exposure
Reproductive system effects
NOEL
mg/kg-d
LOEL
mg/kg-d
Lee etal, 2004
Rat (SD); GD 15 to PND 21; 0,
1.5, 14.4, 148, or 712 mg/kg-d
(converted from 0, 20, 200,
2,000, and 10,000 ppmDBP in
diet)
At 712 mg/kg-d, decreased percent males; decreased AGD and
retained nipples, decreased relative testis weight.
At 1.5, 14.4, 148, and 712 mg/kg-d, on PND 21, reduction in
spermatocyte development, increased foci of aggregated Leydig
cells, and decreased epididymal ductular cross section.
At 148 and 712 mg/kg-d, atwk 11, loss of germ cell development.
At 1.5 mg/kg-d, degeneration and atrophy of mammary gland
alveoli in males at 8-11 wks of age.
1.5
Lehmannetal.,
2004
Rat (SD); CDs 12-19; 0,0.1,1,
10, 30, 50, 100, or 500 mg/kg-d
At >50 mg/kg-d, decreased fetal T concentration.
At 500 mg/kg-d, a reduction in oil red O staining of lipids in fetal
testes.
30
50
Liu et al., 2005
Rat (SD); CDs 12-19; 0,
500 mg/kg-d
Significant reduction in AGD at GD 19.
500
Mahood et al.,
2005
Rat (Wistar); CDs 13.5-20.5; 0
or 500 mg/kg-d
Aggregation of fetal Leydig cells; reduced Leydig cell size; reduced
T levels at GDs 19.5 and 21.5 (early event in testicular dysgenesis);
cryptorchidism; partial absence of epididymis at PND 90.
500
MylchreestetaL
1998
Rat (SD); GD 30 to PND 20;
0, 250, 500, or 750 mg/kg-d
At 500 and 750 mg/kg-d, decreased AGD.
At 250, 500, and 750 mg/kg-d, absent or underdeveloped
epididymis, associated with testicular atrophy and germ cell loss,
hypospadias, ectopic or absent testes.
At 500 and 750 mg/kg-d, absent prostate and seminal vesicles,
small testes, and seminal vesicles.
250
MylchreestetaL,
1999
Rat (SD); GDs 12-21; 0, 100,
250, or 500 mg/kg-d
At 500 mg/kg-d, hypospadias; cryptorchidism; agenesis of the
prostate, epididymis, andvas deferens; degeneration of the
seminiferous epithelium; interstitial cell hyperplasia and adenoma;
decreased weight of prostate, seminal vesicles, epididymis, and
testes.
At 250 and 500 mg/kg-d, retained areolae or thoracic nipples,
decreased AGD.
At 100 mg/kg-d, delayed preputial separation (attributed to highly
affected litter, and not repeated in subsequent study).
100
250
-------
Table 4-1. (continued)
Study3
Species (strain), duration,
and exposure
Reproductive system effects
NOEL
mg/kg-d
LOEL
mg/kg-d
Mylchreestetal.,
2000
Rat (SD); CDs 12-21; 0, 0.5, 5,
50, 100, or 500 mg/kg-d
At 500 mg/kg-d, decreased AGD, hypospadias, cryptorchidism,
absent or partially developed epididymis, vas deferens, seminal
vesicles, and ventral prostate; decreased weights of testes,
epididymis, dorsolateral and ventral prostates, seminal vesicles, and
levator anibulbocavernosus muscle; seminiferous tubule
degeneration, focal Leydig cell hyperplasia, and Leydig cell
adenoma.
At 100 and 500 mg/kg-d, retained thoracic areolae or nipples in
male pups.
50
100
Mylchreestetal.,
2002
Rat (SD); CDs 12-21; 0 or
500 mg/kg-d
In GDs 18 and 21 fetuses, testicular atrophy, Leydig cell
hyperplasia, enlarged seminiferous cords with multinucleated
gonocytes; decreased testicular T; fewer epididymal ducts.
500
NTP, 1991
Rat (SD); continuous breeding
(16 wks) (gestation and
lactation); 0, 80, 385, or
794 mg/kg-d in dams (converted
from 0.1, 0.5, and 1.0 % DBF in
feed)
Fl adults:
At 80, 385, and 794 mg/kg-d: Increased incidence of absent, poorly
developed, or atrophic testis and underdeveloped or absent
epididymis.
At 385 and 794 mg/kg-d: Increased incidence of seminiferous
tubule degeneration.
At 794 mg/k-d: Decreased mating, pregnancy, and fertility indices;
decreased epididymal, prostate, seminal vesicle and testis weights;
decreased cauda epididymal sperm concentration; decreased
average spermatid count, total spermatid heads/testis or total
spermatid heads/g testis; increased incidence of absent,
small/underdeveloped/poorly developed, or atrophic penis, seminal
vesicles, epididymis, and prostate; interstitial/Leydig cell
hyperplasia; delayed testicular descent or cryptorchidism.
80
NTP, 1995 (some
of this is also
reported in Wine
etal., 1997)
Rat (SD); continuous breeding
(16 wks) (gestation and
lactation); 0, 80, 385, or
794 mg/kg-d in dams (converted
from 0.1, 0.5, and 1.0% DBF in
feed)
At 794 mg/k-d: Decreased mating, pregnancy, and fertility indices;
decreased epididymal, prostate, seminal vesicle, and testis weights.
385
794
-------
Table 4-1. (continued)
Study3
NTP, 1995
NTP, 1995
Plummer et al.,
2005 Ab
Shultzetal.,
2001
Thompson etal.,
2004a
Thompson etal.,
2005
Wilson etal.,
2004
Species (strain), duration,
and exposure
Rat (Fischer 344); perinatal and
lactation plus 17 wks; 0, 138,
279, 571, 1,262, or
2,495 mg/kg-d in dam
(converted from 0 or
10,000 ppm during gestation
and lactation; 0, 1,250, 2,500,
5,000, 7500, 10,000,
20,000 ppm for 4 wks PN;
0, 2,500, 5,000,10,000, 20,000,
and 40,000 for last 13 wks PN)
Rat (Fischer 344); perinatal and
lactation plus 4 wks; 0, 143,
284, 579, 879, or 1,115 mg/kg-d
in dam (converted from 0,
1,250, 2,500, 5,000, 7,500,
10,000, and 20,000 ppm)
Rat (strain not specified);
gestation; 0 or 500 mg/kg-d
Rat (SD), CDs 12-21; 0 or
500 mg/kg-d
Rat (SD); CDs 12-17, 12-18,
or 12-19; 0 or 500 mg/kg-d
Rat (SD); GD 19; 0 or
500 mg/kg-d
Rat (SD); CDs 14-18; 0 or
1,000 mg/kg-d
Reproductive system effects
At 571, 1,262, and 2,495 mg/kg-d: Degeneration of germinal
epithelium.
At 1,262 and 2,495 mg/kg-d: Decreased testes and epididymal
weights, fewer sperm heads per testis, and decreased epididymal
sperm concentration.
At 879 and 1,115 mg/kg-d: Moderate epididymal hypospermia in
all males.
At 579 mg/kg-d, mild epididymal hypospermia in 2 of 10 males.
Decreased fetal T levels.
Decreased fetal testicular T and androstenedione; increased
progesterone.
Decreased fetal T.
Decreased fetal T.
Decreased fetal T, expression oflnslS.
NOEL
mg/kg-d
279
284
LOEL
mg/kg-d
571C
579d
500
500
500
500
1,000
oo
-------
Table 4-1. (continued)
Study3
Zhang etal.,
2004
Species (strain), duration,
and exposure
Rat(SD);GD ltoPND21;0,
50, 250, or 500 mg/kg-d
Reproductive system effects
At 250 and 500 mg/kg-d, decreased AGD; underdeveloped
epididymides; decreased epididymis or prostate weight atPND 70;
decreased percent motile sperm and total sperm heads;
degeneration of the seminiferous epithelium.
At 500 mg/kg-d, cryptorchidism, absent epididymides, decreased
total number of sperm.
NOEL
mg/kg-d
50
LOEL
mg/kg-d
250
Ab, Abstract only; AGD, anogenital distance; GD, gestation day; PND, postnatal day; LOEL, lowest-observed-effect level for male reproductive system
outcomes found in the study; NOEL, no-observed-effect level for male reproductive system outcomes; SD, Sprague-Dawley; T, testosterone. Note: These terms
are used solely in a descriptive manner in this table, they may not reflect the terminology of the source study, and they are not intended to convey any regulatory
implication.
aAll studies used an oral route of exposure. Lee et al. (2004) and NTP (1995, 1991) administered DBF in the diet. All other studies used oral gavage.
bThe abstract states that the effects were "dose dependent," but a LOEL is not specifically indicated.
'Overall, the study NOEL and LOEL are lower based on liver peroxisome activity.
dOverall, the study NOEL and LOEL are lower based on increased liver weight.
-------
It is also noted that although BMDL values were calculated for specific developmental
endpoints identified by Lehmann et al. (2004), Mylchreest et al. (2000), and the NTP (1995) (see
Table 4-4 of the 2006 external peer review draft Tox Review for DBF), these values were not
used as a POD for reference value derivation.
4.2. REVIEW OF THE TOXICOLOGY DAT A SET
Figure 4-1 illustrates the stepwise approach taken in the evaluation of the toxicity studies,
focusing on low-dose and low-incidence outcomes. First, for each toxicology study, we
examined the data at the lowest dose levels (as defined by the study NOELs and LOELs)
(Step 2). If there was any indication of insurmountable problems with the quality of the reported
data (e.g., excessive variability, critical methodological concerns, lack of peer review as with
abstracts, etc.), or if there were no individual animal data reported (as is often the case for poster
abstracts as well as for many published studies which only contain extracted summary data), the
review of that study would be terminated. However, if individual data were available, the review
could proceed (Step 3). The individual animal data were examined for evidence of reproductive
system outcomes in the males. Although for most studies the exposures were only administered
during the perinatal developmental period, we recognized that an adverse treatment-related
outcome might be identified at any life stage that was assessed in the study. There were three
possible courses that the data review could take from this point forward. In cases where
problems were identified in the data, we attempted to analyze the extent of the issues and
determine the ability to move forward with the study analysis. In some cases, the analysis
stopped at this point, due to deficits in the study data or to inadequate reporting of individual
animal data. However, if the data in the report appeared to be thoroughly assessed, then the
study outcomes and endpoints were examined. Alternatively, in some cases where adequate
individual study data were available for analysis (NTP, 1991, 1995), further examination of the
study could identify effects at the lowest dose levels that had been considered biologically
irrelevant in the original review, but it might require further consideration. At any point in this
stepwise process where data were deemed insufficient to proceed further, we identified research
needs (discussed in Chapter 7).
To begin the characterization and evaluation of the published studies according to this
stepwise model, important aspects of each study protocol, conduct, and reporting were
4-10
-------
summarized (see Table 4-2). Examination of this table demonstrates that approximately half of
the studies that were selected for analysis (i.e., 14 of 29) were limited to a single dose group,
which eliminated them from further examination for lower-dose level effects. It is also
important to note that individual animal data were reported in only 2 of the 29 studies, thereby
severely limiting, and in some cases even preventing, more rigorous evaluation of the study
findings. These two characteristics alone tend to overshadow any of the other listed study
attributes that might contribute to confidence in study findings (i.e., evidence that the study was
conducted according to quality laboratory standards, description of statistical analysis of the data,
and/or specific information regarding the number of litters and offspring assessed, which would
provide an indicator of statistical power). Of the studies listed, only those conducted by
NTP (1995, 1991) were considered suitable for extended examination.
In order to create a profile of outcomes to the male reproductive system following
developmental exposures, which might then serve as a baseline for further comparison and
analysis of toxicological findings across the studies, a list of observed effects was compiled (see
Table 4-3). The content of this list is very clearly defined by the study protocols, both in terms
of what endpoints were examined in each study and when (i.e., at what life stage) they were
examined. For some endpoints, the precise GD or postnatal day (PND) of evaluation may even
be critical. For example, fetal T should peak at approximately GD 18, so assessments made at
earlier or later time points may be less sensitive in detecting adverse outcomes, and the effects
will not be directly comparable across fetal ages. Decreases in T levels may not be observed
postnatally unless treatment is continued or if testicular malformations disrupt T level (which is a
different mechanism of perturbation than alterations to the steroidogenesis pathway). In
neonates, examination for nipple retention is generally conducted at around PND 13 when the
structure is readily visible but before it is obscured by hair growth. Cryptorchidism, even though
present at birth, may not be readily observable in neonates until they reach the age of
PND 16-21 (and of course, it should be detectable at postweaning ages and in adults). Preputial
separation (PPS) delays can only be observed at the time of sexual maturation, which, in the
male Sprague-Dawley (SD) rat, occurs at approximately PND 42; therefore, this effect cannot be
detected at an earlier life stage, nor will it be observed in sexually mature adults. On the other
hand, sperm alterations (count, morphology, or motility) and perturbations in male fertility can
only be assessed in adult males, not in immature individuals at earlier life stages
4-11
-------
Table 4-2. Reporting and study size characteristics of male reproductive studies following in utero exposure to
DBF
Study
Barlow and Foster, 2003
Barlow etal, 2003
Barlow etal, 2004
Bowman etal., 2005
Carruthers and Foster, 2005
Etna etal., 1998
Ema et al., 2000b
Ferrara et al., 2006
Fisher et al., 2003
Gray etal., 1999
Kim etal., 2004 Ab
Kleymenova et al., 2004 Ab
Kleymenova et al., 2005a
Ab
Kleymenova et al., 2005b
>One
high
dose
y
y
y
y
PPS
only
y
y
y
Individual
data
publicly
available
^subset13
Statistical
analysis
method
reported
y
y
y
y
y
y
y
y
y
y
y
Study
conduct
level
reported
y
y
y
y
y
y
y
y
Number evaluated/group
Litters
l-9a
NR
8-lla
18
l-14d
1 1 DBF treated
73 DBF treated
"in most instances" ~3-6
NR
4 (LE); 8 (SD)
NR
NR
NR
3
Offspring
7-60a
3
35-74a'c
All male fetuses
l-91d
AGO :NR; crypt: 144
~770e
1-3/litter
Testis wt: 5-10 animals/age group (4);
hyp. and crypt.: 10 adults
LE: 30 male pups; 13 adult males
SD: 48 male pups; 17 adult malesf
NR
NR
NR
14-21 pups/evaluation
to
-------
Table 4-2. (continued)
Study
Lee etal, 2004
Lehmann et al, 2004
Liu et al., 2005
Mahood et al., 2005
Mylchreest et al., 1998
Mylchreest et al., 1999
Mylchreest et al., 2000
Mylchreest et al., 2002
NTP, 1995, 1991
Plummer et al., 2005 Ab
Shultz etal., 2001
Thompson et al., 2004a
Thompson et al., 2005
Wilson etal., 2004
Zhang etal., 2004
>One
high
dose
y
y
y
y
y
y
y
Individual
data
publicly
available
y
Stat
analysis
method
reported
y
y
y
y
y
y
y
y
y
y
y
y
y
y
Study
conduct
level
reported
y
y
y
y
y
y
y
y
y
y
y
y
Number evaluated/group
Litters
6-8
1-4
3
2-7
7-10
10
11-20
5-6
20
NR
3
4
4
3
14-16
Offspring
11-20 adults
3-4 fetuses/group
3 fetuses/litter
NR
All males/litter
All males/litter
All males/litter
23-49 fetuses
All pups/litter in-life thru necropsy;
histopath: 10/selected groups
NR
1 male/litter
1 male/litter
3 fetuses/litter
All males/litter
20 pups/group
-------
Table 4-2. (continued)
Ab, Abstract only; LE, Long Evans; NR, Not reported; PPS, preputial separation; •/, present.
aLitters and pup numbers not reported for AGD and areolae retention.
bData for three individual animals were reported for LC and Sertoli cell staining. The other results are not reported in this table because they were from
toxicogenomic studies (see Chapter 5).
°57-100% of these pups survived to necropsy so for malformations that required necropsy, the number of pups is less than shown.
dLitters for AGD were the statistical unit; neither litter nor pup numbers for AGD were reported.
eNumber derived from the mean number of live fetuses/litter.
fln some cases, data from two experiments were combined.
-------
Table 4-3. Life stage at observation for various male reproductive system
outcomes assessed in studies of developmental exposure to DBF
Findings
Decreased T
Malformations
Decreased AGD
Hypospadias
Retained nipples/areolae
Cryptorchidism
Delayed PPS
Organ weights
Histopathology of male
reproductive organs
Abnormal sperm
Decreased fertility
Life stage of animals (rats) at observation
Fetus
/
/
/
Neonate through
puberty
/
/
/
/
/
/
/
/
/
Adult
/
/
/
/
/
/
/
/
/
/
T, Testosterone; AGD, Anogenital distance; PPS, Preputial separation.
Using the list in Table 4-3 as a guide, a more extended analysis was conducted for each
of the selected studies. Table 4-4 presents the detailed results. In this table, the various observed
outcomes are arrayed across three general life stage categories: prenatal (i.e., observations
conducted in fetuses), neonatal through puberty (i.e., observations conducted in pups), and adult
(i.e., observations conducted in young, sexually mature animals). These life stage categories do
not represent the period of exposure for the study. While all studies include exposures during
late gestation (i.e., during the critical window of male reproductive system development), some
studies also maintained exposures during later life stages. For reference, Table 4-1 provides
general descriptions of exposure durations.
4-15
-------
Table 4-4. Age of assessment for individual endpoints across studies of the male reproductive system following
developmental exposure to DBF
Barlow and Foster, 2003
Barlow et al., 2003
Barlow et al., 2004
Bowman et al., 2005
Carruthers and Foster, 2005
Ema et al., 1998
Ema et al., 2000b
Fisher etal., 2003
Gray et al., 1999b
Kim etal., 2004 Ab
Kleymenova et al., 2004 Ab
Kleymenova et al., 2005a
Ab
Kleymenova et al., 2005b
Lee et al., 2004
Lehmann et al., 2004
Liu etal., 2005
Mahood et al., 2005
Mylchreest et al., 1998
Fetus
I
Ta
/
/
/
Histo-
pathb
/
/
/"
/
/
/
/
/
/
Neonate through puberty
I
AGD
/
/
/'
/'
/'
/
/
/s
/
Hyp
/
/
—
— *
/
Ret.
nip/
areolae
/
/
/j
/
/
Cryptc
/
/'
/'
/
—
— *
/'
/
Del.
PPS11
/—
/°
—
—
I
Org
wt
/
/
—
NR
Histo-
pathb
/
/
/
/
I
T
SI— m
/"
Adult
Malf
/
/
/
/
/
— *
/
/
I
Org
wt
/e
/f
/
/
/
/
NR
/
Histo-
pathb
/
/
/
/
/
/
/
Ab.
Sperm
/
/
/°
U
|Fert
/
/"
Hyp
/
/8
—
/
/
NR
/
Ret.
nip/
areolae
/
/k
/
Crypt
/
/
—
/
/
/
/
A
AGD
/|
/t
4
Ta
m
/"
-------
Table 4-4. (continued)
Mylchreest et al., 1999
Mylchreest et al., 2000
Mylchreest et al., 2002
NTP, 1991
Plummer et al., 2005
Shultzetal.,2001
Thompson et al., 2004a
Thompson et al., 2005
Wilson etal., 2004
Zhang et al., 2004
Fetus
4
Ta
/
/
/
/
/
/
Histo-
pathb
/
Neonate through puberty
I
AGD
/
/
NRS
NRS
/
Hyp
/
/
/
—
Ret.
nip/
areolae
/
/
Crypt0
/
/
/v
/
Del.
PPSd
/
—
4
Org
wt
—
Histo-
pathb
4Ta
Adult
Malf
/
/
/
/
4
Org
wt
/
/
/
/
msto-
pathb
/
/
/
/
Ab.
Sperm
/
/u
4Fert
/
Hyp
/
/
/
—
Ret.
nip/
areolae
Crypt
/
NR
/
/
A
AGD
4
Ta
•/ Observed; —, Not observed; white box, Not evaluated; shaded box, Evaluated; NR, Not reported, although the study indicates that the endpoint was evaluated;
Ab, Abstract only; PPS, preputial separation; AGD, anogenital distance; Hyp, hypospadias; A, change; Ret. nip/areolae, retained nipples and/or areolae; Cryp,
cryptorchidism; Del, delay; Org wt, organ weight decrease (absolute or relative) in at least one reproductive organ; Malf, malformations including
ventral/dorsal/lateral prostate, seminal vesicles, androgen dependent muscles, (accessory sex organs) epididymis, vas deferens external genitalia, cryptorchidism,
small or flaccid testes; Pert, fertility; Ab. Sperm, abnormal changes in sperm count, motility, and/or morphology.
"Decreased testicular testosterone (T) was measured in the fetus; Decreased serum T was measured postnatally and in adults.
bHistological changes—Leydig cell hyperplasia (aggregation); multinucleated gonocytes; Wolffian duct increased coiling (can be measured in fetus, neonate
through puberty, or adult).
°Cryptorchidism can be observed between PNDs 16-21 and older.
dDelayed preputial separation normally observed ~PND 42.
Enlargement of the seminiferous cords was observed at PNDs 19-21.
fln addition to the observed decreases and absences of male reproductive organs, "occasional enlargement" of the testes was observed only in the 500 mg/kg-d
group.
-------
Table 4-4. (continued)
8Assessed in adult animals at PNDs 180, 370, and 540. Hypospadias only observed in the 500-mg/kg-d group.
hWolffian ducts smaller, more fragile, adipose tissue surrounding duct was more gelatinous, and decreased coiling.
'Assessed at PNDs 1 and 13. Reduction in AGD observed in animals exposed to DBF on GDs 16 and 17, GDs 17 and 18, or GDs 19 and 20; no change in AGD
in animals exposed GDs 14 and 15.
JAssessed on PND 13; assessed on an individual animal basis, significant increase in nipple retention was observed after dosing on GDs 15-16; 16-17; 17-18; or
19-20.
kAssessed at PND 90; significant increase in nipple retention only for males dosed GDs 16 -17 (individual animal basis).
'AGD and cryptorchidism were assessed in fetuses on GD 21. Exposed pregnant dams were sacrificed on day 21, and live fetuses were removed.
m Blood plasma T levels significantly reduced on PND 25 but not on PNDs 4, 10, or in adult.
"Delayed PPS only reported for parental generation (PO) males exposed from weaning through to puberty.
"Reduced epididymal sperm numbers; not necessarily abnormal sperm.
p In PO males.
q Evaluated T levels at 31 and 42 days (not fetus) and found decreased at 42 days.
rlt is presumed that specific malformations would have been observed if present based on the study design and methods.
"Examined in GDs 19 or 21 fetuses.
'Observed at PNDs 25 and 90; nonscrotal testes were not evaluated histopathologically.
"Only motility was evaluated in Mylchreest et al. (1998); in Zhang et al. (2004), sperm number, motility, and morphology were evaluated, but only count was
affected.
v Study mentions that adult cryptorchidism was observed, but study methods do not indicate that offspring were retained until adult age.
-------
Table 4-4 summarizes the outcomes and presents a broad representation of positive and
negative observations in a manner that demonstrates that not all relevant endpoints were
evaluated at all life stages or even in each study. To facilitate summarization of the myriad
individual study findings, information was often combined by category (e.g., "histopathology"
includes a broad variety of outcomes in various reproductive organs), and for the sake of brevity,
the minute details and nuances of the study design and observations, although quite interesting,
are not typically presented. In a few cases, negative outcomes presented in the table are
extrapolations based on the presumption that specific findings would have been observed if they
were present. For example, with methods that include detailed external and internal
(macropathology) examination of pups and/or adults, the absence of reported malformations at
either of these life stages was presumed to indicate that no gross malformations were observed
because they should have been readily detectable (Lee et al., 2004).
Tables 4-1, 4-3, and 4-4 clearly illustrate that the study protocols varied quite extensively.
In general, with the exception of the NTP studies, the protocols were not designed to conform to
a particular regulatory guideline. Rather, the majority of the studies were focused research
efforts that were verifying and/or expanding upon previously observed outcomes; therefore, the
differences across study methods are understandable. As a result, the apparent lack of
consistency in male reproductive system observations across studies is generally attributable to
differences in protocol design and implementation. Some examples are discussed in detail as
follows:
Although all of these studies used exposures during late gestation (i.e., a critical period of
male reproductive system development in the rat), the specific endpoints that were
assessed and/or the life stages at which endpoints were examined varied extensively
across the studies. Obviously, treatment-related alterations of life-stage-specific events
require examination during the most appropriate or optimal life stage (e.g., increased
multinucleated gonocytes can only be observed in fetal testes, delays in PPS can only be
observed in juvenile animals at the time of sexual maturation, and disturbances in
reproductive function can only be observed in sexually mature adults). Other permanent
structural abnormalities may be detected across multiple life stages (e.g., hypospadias or
cryptorchidism could theoretically be observed in late gestation fetuses, in adolescents,
and in adults). For some outcomes, it is difficult to predict a priori the optimal time point
for evaluation. For example, DBF-related increases in the estrogen receptor (ER) were
observed at 31 days but not at 42 days (Kim et al., 2004).
4-19
-------
• It is important to realize that not all available offspring are evaluated in every study;
therefore, identification of adverse outcomes may rely, in part, on sampling protocols and
the statistical power of the sample size for detection of rare or low-incidence events.
Calculations of statistical power are rarely provided in study reports.
• In some cases, apparent differences in studies may result because the report contains an
insufficient level of detail on a particular endpoint or life stage—often because the
emphasis of the scientific review lies in a slightly different direction. For example, if
high doses of DBF are administered during sensitive periods of male reproductive system
development, and the males are maintained and terminated as adults, at which time
histopathological evaluation is performed, it might be assumed that various male
reproductive system malformations and/or cryptorchidism would have been present in
some of the males at necropsy. Yet, these findings may not be reported because the
histopathological findings are the primary focus of the investigation and/or the
publication (e.g., Lee et al., 2004).
• In other situations, the description of the findings at various life stages may vary. For
example, evidence of cryptorchidism may be described as "testis located high in the
abdomen" in a fetus, as "undescended testis(es)" in an adolescent rat, or as "unilateral
testis" upon noninvasive clinical examination of an adult. To some extent, this lack of
consistency in terminology may result from laboratory Standard Operating Procedures
that direct technical staff to avoid the use of diagnostic terminology.
Overall, in spite of numerous differences in the study designs, the toxicity data set for
DBF clearly demonstrates that exposure to DBF during critical stages of male reproductive
system development can result in adverse structural and functional reproductive outcomes.
When specific critical aspects of study design and implementation were similar, consistent
outcomes were almost universally observed. The WOE embodied by the data described above is
further supported by studies in rats that demonstrated similar incidences of cryptorchidism and
decreased AGD in male pups of dams treated with either DBF or MBP, the metabolite of DBF
(Ema and Miyawaki, 200 la). The ability of MBP to cross the placenta and reach the fetus has
also been conclusively demonstrated (Fennell et al., 2004; Saillenfait et al., 1998), and these two
TK events (metabolism and placental transport) are key to the MOA of reduced fetal testicular T
(David, 2006). Available toxicogenomic data, described elsewhere in this case-study report,
further elucidate the MOA(s) of DBF in producing adverse effects on male reproductive system
development and are an important consideration in the WOE analysis of the toxicological data
set.
4-20
-------
In the selected DBF toxicology study data set, the presentation of extensive individual
offspring data was limited to the NTP (1991) study conducted as a reproductive assessment by
continuous breeding (RACE) in SD rats. The individual data from this study were carefully
examined in order to confirm the NOEL and LOEL described in the study report. This analysis
was conducted under the presumption that statistical and/or biological significance noted in the
summary compilations of male reproductive system outcomes might not identify low incidence
effects in individual offspring at lower dose levels. To further aid the identification of treatment-
related outcomes, the male reproductive system outcomes were grouped by organ instead of
individual animal. This analysis revealed apparently treatment-related findings in the testis and
epididymis of Fl male offspring, as summarized in Table 4-5. At the highest dose tested
(794 mg/kg-d, equivalent to 1.0% DBF in the diet), additional findings in the male reproductive
organs of Fl offspring included single incidences of (1) underdeveloped prepuce; (2) mild
secretion and severe vesiculitis of the prostate; (3) a mass on the testis; and (4) a focal granuloma
with fluid and cellular degeneration in the epididymis; none of these findings were observed at
the lower dose levels. Understandably, the findings at the low- and mid-dose groups were
originally interpreted as not being treatment-related (Wine et al., 1997; NTP, 1991). However,
consideration of MO A information for DBF, including toxicogenomic data, resulted in a more
conservative interpretation of the toxicity data both by NTP researchers (conference call in 2008
between Paul Foster [NTP/NIEHS], Susan Makris [EPA/NCEA], and Susan Euling
[EPA/NCEA]) and by the EPA IRIS program (U.S. EPA, 2006a). Consequently, further analysis
of individual offspring data in the current case study did not identify any additional sensitive
toxicological outcomes; the study LOEL was confirmed to be the lowest treatment level tested in
the NTP RACE study (80 mg/kg-d).
4.3. UNEXPLAINED MOAs FOR DBF MALE REPRODUCTIVE TOXICITY
OUTCOMES
Figure 3-6 illustrates the broad conceptual approach for consideration and interpretation
of toxicogenomic and toxicology data to inform MO A. The toxicogenomic data can be
evaluated to identify altered genes, gene products, and pathways; this information can lead to a
more complete understanding of the mechanism of action or MOA(s) for the chemical toxicity.
From the opposite perspective, the toxicity data can provide information critical to identifying
4-21
-------
Table 4-5. Incidence of gross pathology in Fl male reproductive organs in
one continuous breeding study with DBPa
Gross finding"1
Testis: absent, poorly developed, atrophic,
undescended
Penis: small/underdeveloped
Epididymis: underdeveloped/absent
Dose (% in diet)
0
0/20
0/20
0/20
0.1
1/20
0/20
1/20
0.5
1/20
0/20
1/20
1.0
6/20
4/20
12/20
"Incidences were compiled from reported individual animal macroscopic pathology data; statistical analysis was not
performed.
bSome animals have more than one type of malformation, and these animals were counted separately for each of the
three outcome categories.
Source: (NTP, 1991).
the relevant MOA(s) involved in the toxicological outcomes, and thereby inform the
interpretation of gene alterations and relevant pathways.
Each male reproductive system outcome was evaluated for consistency with either or
both of the two well-established MO As using expert judgment based on the available published
literature for DBF (see Figure 4-2). This exercise helped to identify the unexplained endpoints
for which the evaluation of the toxicogenomic data set may suggest potential MO As (see Chapter
5). For the DBF case study, Table 4-6 presents a compendium of the specific findings noted in
the male reproductive system following exposures at critical windows of development. While
reduced fetal testicular T and reduced Insl3 signaling can be linked to some of the observed
outcomes on the basis of available data, potential key events cannot specifically be identified for
other outcomes. The unexplained MO As are good candidates for further study, both in
toxicology and toxicogenomic studies, to elucidate the underlying mechanism of action.
4-22
-------
:E^Jh|:mSl6:t6p^dtibtive
following developmental
xDBF :exposu:re| x
Consider MOA evidence
Explained by J,T?
Explained by |/ns/3?
|TMOA
unexplained
MOA
l/ns/3
MOA
Figure 4-2. The process for evaluating the MOA for individual
male reproductive system outcomes following developmental
DBF exposure. The available data for MOA for each male
reproductive outcome following developmental DBF exposure
were evaluated by a team of experts. For each outcome, the
current WOE of the data either support the MOA ("YES"), support
that this is not the MOA ("NO"), or are inconclusive for the MOA,
i.e., either unlikely or unclear ("?"). "Unexplained MO As" include
both "?" and "NO" conclusions.
4-23
-------
Table 4-6. Evidence for MOAs for the observed effects in the male
reproductive system after in utero DBF exposure
Organ/
Function
Testes
Gubernacular
ligament
Epididymis
Mammary gland
Effect
Multinucleated gonocytes; increased number of
gonocytes in fetal testes
Altered proliferation of Sertoli and peritubular
cells; fewer Sertoli cells
Gonocyte apoptosis increase; early postnatal
decrease in gonocyte number
Abnormal Sertoli cell-gonocyte interaction
Small incidence of Ley dig cell adenomas,
aggregates, and hyperplasia
Decreased number of spermatocytes or cauda
epididymal sperm concentration.
Small or flaccid; other abnormalities; decreased
weight
Increased weight due to edema
Decreased number or degeneration of
seminiferous cords/tubules; altered
morphology; degeneration of the epithelium;
enlarged cords/tubules
Testes descent: none (cryptorchid) or delayed
Gubernacular ligament development effects:
agenesis or elongation
Lesions and agenesis; partial to complete
absence; decreased epididymal ductular cross
section
Reduced weights
Nipple and/or areolae retention in males
Degeneration and atrophy of alveoli in males
MOA
Reduced
fetal
testicular
T
?b
?b
?b
?b
/
/
/
?e
?b
/f
X
/
/
/
?b
Reduced
InslS
signaling
?c
?c
?c
?c
?c
/d
/
?
?c
/f
/
X
/
X
X
4-24
-------
Table 4-6. (continued)
Organ/
Function
Wolffian ducts
Seminal vesicles
Coagulating gland
Penis
Accessory sex
organ
Prostate
Vas deferens
Levator
anibulbocavernosus
muscle
Male/female ratio
Perineum
Repro function
Effect
Underdeveloped
Malformations or absent; decreased weight
Malformations
Small, underdeveloped
Hypospadias
Delayed preputial separation
Decreased weight
Decreased wt or absent
Agenesis
Decreased weight
Decreased % male offspring as determined by
AGO at birth
Decreased AGO
Infertility
MOA
Reduced
fetal
testicular
T
/
/
/
/
/
/
/
/
/
/
/
/
/
Reduced
Iml3
signaling
X
X
X
X
X
X
X
X
X
?c
X
X
/d
AGD, anogenital distance; ?, Current data indicate that it is unlikely the MOA; •/, Current weight of evidence of the
data support this MOA leading to the effect; X, Current weight of evidence of the data indicate that this MOA is not
the MOA for this outcome.
aMOA is defined as one or a sequence of key events upon which the outcome is dependent (see glossary).
bReduced fetal testicular T may play a role, but current data indicate that reduced T is not solely responsible for this
outcome.
°The Insl3 knockout mouse phenotype suggests that Insl3 is specifically required for gubernacular ligament
development and, therefore, testis descent in mice since these mice do not have other defects.
dDecreased fertility in males is a result of reduced Insl3 signaling since reduced Insl3 signaling leads to undescended
testes, which, in turn, reduces sperm count (presumably by increasing the temperature) and can cause infertility.
4-25
-------
Table 4-6. (continued)
eln some animals, increased weight, due to edema, can result in animals that have epididymal agenesis, which is a
consequence of reduced testosterone (T).
flnsl3 signaling is required for development of the gubernacular ligament and through this mechanism—the 1st stage
of testis descent from the kidney region to the inguinal region. Testosterone is required for the 2nd stage of testis
descent, from the inguinal region to the scrotum (reviewed in Klonisch et al., 2004). After in utero DBF exposure,
the cryptorchid phenotype resembles the Insl3 knockout. A delay in testis descent can result from reduced Insl3
andT.
4.4. CONCLUSIONS ABOUT THE TOXICITY DATA SET EVALUATION:
DECISIONS AND RATIONALE
The review of the toxicology data set identified a number of issues and limitations that
are evident in the study descriptions and endpoint summaries presented in this chapter. These
include the following:
• Lack of dose-response information: A number of studies conducted with DBF used a
single high-dose treatment level (often at 500 mg/kg-d) in order to produce readily
observable adverse outcomes to male reproductive system development that could be
examined. In such studies, the absence of lower dose levels prevents the evaluation of
dose-dependent responses and does not allow the identification of study-specific NOELs
or LOELs. While this approach is useful for hazard characterization, it does not facilitate
other aspects of risk assessment (e.g., dose-response assessment or risk characterization).
Thus, studies utilizing a single high-dose level may provide important information for a
WOE assessment of the toxicology profile, but they have diminished usefulness in
identifying outcomes for use in risk calculations at environmentally relevant doses.
• Insufficient information on study methods: Even though every study report includes a
section on study methods, there can be a great deal of unevenness in the amount of
detailed information provided. Consequently, important questions may arise during study
review that cannot be readily resolved. In some cases, this can have an impact on
individual study interpretation or on conclusions that rely on a thorough WOE evaluation
of the data set.
• Unavailable individual outcome data: A full range of individual animal data is seldom
included in studies published in the open literature and is almost never available when the
only available publication is a presentation abstract. Conversely, individual animal data
are generally included in toxicology reports generated in response to a regulatory
mandate or conducted by a federal agency (e.g., NTP). The availability of individual
animal data can be quite important in interpreting the study findings, because it can
reveal problems or inadequacies in the data, but it can also help identify low incidence
adverse outcomes. In the case of DBF, the individual offspring data presented in the
NTP study report (1991) include alterations in the reproductive system of Fl males that
4-26
-------
had been exposed during development. These findings are similar to outcomes identified
at higher dose levels, are consistent with the proposed MO A, and, consequently, are used
to establish a LOEL for the study.
• Protocol limitations: Unless studies are designed to meet the recommendations of a
standardized testing protocol (e.g., NTP or U.S. EPA/Office of Prevention, Pesticides and
Toxic Substances reproductive toxicity study guidelines), there may be a high degree of
variability among the protocols used for testing any one chemical. Between two studies,
there can be differences in the treatment regimen or in the assessment of outcomes that
render them incomparable. DBF provides a good example of a chemical that targets a
very specific critical prenatal window of reproductive system development in males, and
results in adverse outcomes that could go unidentified if the appropriate endpoint(s) are
not assessed at the optimal life stage or time point.
• Specific study's limitations: Even when a study design optimizes the detection of adverse
outcomes from chemical treatment, there may be challenges in study analysis and
interpretation. Such is the case with the NTP study on DBF, which was conducted in
several phases and reported both in the open literature (Wine et al., 1997) and by the
Institute that conducted the experiments (NTP, 1995, 1991).
The analysis of the toxicology data in this chapter has provided a firm basis for expanded
consideration of the toxicogenomic data for DBF as depicted in Figure 3-6. The extensive
analysis of the toxicology data set and consideration of MOA(s) provide a source of information
for use in phenotypic linking of known and potential MO As. This chapter also provides an
example of steps one can take to develop a toxicological data source, in particular, examining
(1) the individual toxicity studies; (2) the WOE for the studies; (3) potential low incidence and
low-dose effects; and (4) the MO A for the affected endpoints. All of these steps are useful
exercises for evaluating toxicogenomic data in future risk assessments. The evaluations of both
the toxicity and toxicogenomic data sets (detailed in Chapters 4 and 5) provide strong support for
phenotypic anchoring for a number of gene expression changes occurring after in utero DBF
exposure for several of the male reproductive outcomes. The available toxicogenomic studies
for DBF are evaluated in Chapter 5.
4-27
-------
5. EVALUATION OF THE DBF TOXICOGENOMIC
DATA SET
This chapter presents an evaluation of the DBF toxicogenomic data set from the
published literature and some new analyses of one of the microarray studies for DBF. The
toxicogenomic studies include nine published RT-PCR and microarray studies in the rat after in
utero DBF exposure. First, we evaluated the toxicogenomic data set from the published
literature for the consistency of findings. Second, evaluating the published literature and
performing new pathway analyses of the Liu et al. (2005) data set, we generated hypotheses
about pathways/mechanism affected by DBF in utero exposure that may explain the testis
endpoints for which there is no established MOA (these "unexplained" endpoints were identified
in Chapter 4). The DBF genomic data set includes nine papers published through July 2007.
The microarray studies all exposed animals to DBF doses of 500-1,000 mg/kg-d during the
critical window for male reproductive development, which is during late gestation and correlates
with the time of peak T production. The chapter first discusses the methodologies utilized in the
nine studies and provides a brief overview of each study. The chapter then presents an
evaluation of the consistency of the findings for the microarray, RT-PCR, and protein studies
performed in the rat testes. The findings of the Lehmann et al. (2004) study, the one available
dose-response RT-PCR study for DBF, are discussed. In addition, the pathway reanalysis of the
Liu et al. (2005) study is presented, and data gaps and research needs are identified.
5.1. METHODS FOR ANALYSIS OF GENE EXPRESSION: DESCRIPTION OF
MICROARRAY TECHNIQUES AND SEMI-QUANTITATIVE RT-PCR
5.1.1. Microarray Technology
Microarray technology allows for analysis of genome-wide expression of thousands of
genes from the organ or tissue of interest. In principle, there are two main types of microarrays:
the cDNA microarray and the oligonucleotide array. The cDNA microarray contains DNA from
each open reading frame spotted onto glass microscope slides or nylon membranes. These
probes are used to detect cDNA, which is DNA synthesized from a mature, fully spliced mRNA
transcript. For example, Clontech's Atlas Arrays contain DNA sequences from thousands of
genes immobilized on nylon membrane or glass slides. Each gene found on these arrays is well-
characterized. These arrays use a radiolabelled detection system for analyzing the changes in
5-1
-------
gene expression and have been optimized for high-quality expression profiling using a limited
set of genes. Moreover, such arrays allow for the use of 32P, and, therefore, offer a sensitive
measure of gene expression. The second type of microarray is the oligonucleotide array. Here,
short DNA sequences or oligonucleotides (oligos) are synthesized directly onto the glass slide
via a number of different methods. For example, Affymetrix® uses 'Photolithographic'
technology, where probes are directly synthesized onto the arrays. Briefly, the slide is coated
with a light-sensitive chemical compound that prevents the formation of a bond between the slide
and the first nucleotide of the DNA probe being created. Then, chromium masks are used to
either block or transmit light onto specific locations on the surface of the slide. A solution
containing thymine, adenine, cytosine, or guanine is poured over the slide, and a chemical bond
is formed in areas of the array that are not protected by the mask (exposed to light). This process
is repeated 100 times in order to synthesize probes that are 25 nucleotides long. This method
allows for high-probe density on a slide.
Affymetrix uses an antibody detection system with horseradish peroxidase and
streptavidin conjugates, and a 2-dye system (Cy3- and Cy5- labeled fluorescein dyes), which is
unique to this platform. The Agilent scanner detects the relative intensities of the red and green
labels and gives a relative measure of the gene expression changes between the control and
treated samples. In the case of Affymetrix and Clontech, the detection system measures the
absolute intensity of the individual probes of the treated and control samples. These values are
then used to calculate the relative gene expression change between the treated and control
samples.
5.1.2. Reverse Transcription-Polymerase Chain Reaction (RT-PCR)
Polymerase Chain Reaction (PCR) is a method that allows exponential amplification of
short DNA sequences within a longer double stranded DNA molecule using a thermo-stable
DNA polymerase called Taq polymerase. RT-PCR is a semi quantitative technique for detection
of expressed gene transcripts or mRNA. Over the last several years, the development of novel
chemistries and instrumentation platforms enabling detection of PCR products on a real-time
basis has led to widespread adoption of real-time RT-PCR as the method of choice for
quantitating changes in gene expression. Real-time RT-PCR is a kinetic approach in which the
reaction is observed in the early, linear stages. Furthermore, real-time RT-PCR has become the
5-2
-------
preferred method for confirming results obtained from microarray analyses and other techniques
that evaluate gene expression changes on a global scale.
5.2. REVIEW OF THE PUBLISHED DBF TOXICOGENOMIC STUDIES
5.2.1. Overview of the Toxicogenomic Studies
We evaluated nine studies published prior to July 2007 that characterized altered gene
expression in rats following prenatal DBF exposure. Among these nine studies, four are based
on the analysis of preselected genes by real-time RT-PCR, while the other five are based on the
analysis of global gene expression by microarray technology. Table 5-1 summarizes general
information (e.g., DBF dose, exposure route, exposure window, tissue type) for these nine
studies, and each study is briefly reviewed. Section 5.2.3.2 presents information about the
similarities and differences among these studies.
5.2.2. Microarray Studies
5.2.2.1. Shultz et al (2001)
Six SD rats per group were treated by gavage with corn oil, DBF (500 mg/kg-d), or
flutamide (reference antiandrogen, 50 mg/kg-d) from GDs 12-16, GDs 12-19, or GDs 12-21.
Testes were then isolated on GD 16, 19, or 21. Global changes in gene expression were
determined by Clontech cDNA expression array (588 genes). Shultz et al. (2001) isolated total
RNA from testis of control and treated animals. Reverse transcription reactions were performed
using total RNA, [32P]-dATP, and superscript IIMMLV-RT. Following purification, the probes
were counted, and equal numbers of counts per minute were added to each rat gene cDNA
expression array. The arrays were hybridized with cDNA using 1 fetus per dam. Hybridization
and washing were performed according to manufacturer's instructions. Digital images were
collected on a BioRad phosphorimager and analyzed using Clontech's Atlas Image software.
Eight genes were further examined by real-time RT-PCR. Total RNA was isolated from both
testes using RNA STAT60, and then the RNA was treated with DNase I in the presence of
RNasin. cDNA was then synthesized using random primers and TaqMan reverse transcription
reagents. Quality of RT reactions was confirmed by comparison of RT versus no enzyme control
for each RNA sample using the glyceraldehyde-3-phosphate dehydrogenase (GAPDH) primer
5-3
-------
Table 5-1. Study comparisons for the toxicogenomic data set from male tissues after in utero DBF exposure
Study3
Barlow et
al., 2003
Bowman et
al., 2005
Lehmann et
al., 2004
Liu et al.,
2005C
Plummer et
al., 2007d
Shultz et al.,
2001
Thompson
et al., 2004
Strain and
species
SDrat
SDrat
SDrat
SDrat
Wistar rat
SDrat
SDrat
DBF doses
500 mg/kg-d
500 mg/kg-d
0.1, 1.0, 10,50,
100, or
500 mg/kg-d
500 mg/kg-d
500 mg/kg-d
500 mg/kg-d
500 mg/kg-d
Treatment interval13
GDs 12-19
GDs 12-19 or 19-21
GDs 12-19
GDs 12-19
GDs 12.5-15.5;
12.5-17.5, or 12.5-19.5
GDs 12-16, 12-19, or
12-21
GDs 12-17, 18, or 19;
13-19, 14-19, 15-19,
16-19, 17-19, 18-19, or
19
Toxicogenomic method
Microarray
(Platform)
No
Yes (Clontech
cDNA arrays)
No
Yes
(Affymetrix®
GeneChip®
oligo arrays)
Yes (Agilent
22K and 44K
oligo arrays)
Yes (Clontech
cDNA arrays)
No
RT-PCR
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Tissues
collected
Testis
Wolffian
ducts
Testis
Testis
Testis: whole,
seminiferous
cord, and
interstitial
regions
Testis
Testis
-------
Table 5-1. (continued)
Study3
Thompson
et al., 2005
Wilson et
al., 2004e
Strain and
species
SDrat
Rat, SD
DBF doses
500 mg/kg-d
1,000 mg/kg-d
Treatment interval13
0.5-24 hrs on GDs
18-19 or GD 19
GDs 13-17
Toxicogenomic method
Microarray
(Platform)
Yes
(Affymetrix
GeneChip oligo
arrays)
No
RT-PCR
Yes
Yes
Tissues
collected
Testis
Testis
aln all studies, oral gavage was the route of exposure.
bGD 0 = sperm positive.
°Study assessed 7 different phthalates.
dPlummer et al. (2007) reported dosing intervals spanning GDs 12.5-19.5, which is comparable to GDs 12-19 in the other studies due to differences in reporting of
GD and sperm positive at GD 0.5.
eWilson et al. (2004) reported a dosing interval of GDs 14 -18, which is comparable to GDs 13-17 in the other studies due to differences in reporting of GD
and sperm positive at GD 1.
-------
set. Fourteen rat-specific primer sets were used for analyses. The ABI PRISM 7700 and the
ABI PRISM 7900HT Sequence Detection System was used for RT-PCR with the SYBR Green
PCR and TaqMan Universal PCR Master Mix reagents. GAPDH was used as an on-plate
internal calibrator for all RT-PCR reactions.
Genes analyzed by real-time RT-PCR include clusterin (C7w), cytochrome P450,
family 11, subfamily a, polypeptide 1 (Cypllal), myristoylated alanine-rich C-kinase substrate
(Marcks), proliferating cell nuclear antigen (Pcna), cytochrome P450, family 17, subfamily a,
polypeptide 1 (Cypl7al}, steroidogenic acute regulatory protein (Star), scavenger receptor class
B, member 1 (Scarbl), and v-kit Hardy Zuckerman 4 feline sarcoma viral oncogene homolog
(Kit). Radioimmunoassay of steroid hormones and immunocytochemical analysis of certain
proteins (i.e., CLU and b-cell leukemia/lymphoma 2 [BCL2]) in the fetal testes were also
performed.
Of the 588 genes examined, -45 genes had at least a 2-fold change in the average
expression values in DBF-treated rats relative to the average values in control rats. DBF
exposure led to a reduced expression of steroidogenic enzymes at GD 19, such as Cypllal,
Cypl7al, ScarbJ, and Star. These genes were upregulated at GD 19 following flutamide
exposure, suggesting that DBF does not act as an androgen antagonist at this time point.
Flutamide and DBF demonstrate patterns of gene expression that overlap, though both have
distinctly expressed genes. This suggests to Shultz et al. (2001) that there are both common and
distinct molecular pathways within the developing fetal testes.
Other genes affected after DBF exposure were Clu (upregulated) and Kit
(downregulated). Using immunocytochemical staining of CLU and BCL2 protein in the fetal
testes, increased amounts of both proteins were observed in the Ley dig and Sertoli cells of
GD 21 testes. Decreases in testicular T and androstenedione in testes isolated on GDs 19 and 21
were observed, while increases in progesterone in testes isolated on GD 19 in DBF-exposed
testis were observed.
Shultz et al. (2001) suggest that the antiandrogenic effects of DBF are due to decreased
T synthesis. Furthermore, enhanced expression of cell survival proteins, such as CLU and
BCL2, may be involved in DBF-induced Leydig cell (LC) hyperplasia, while downregulation of
Kit may play a role in gonocyte degeneration.
5-6
-------
5.2.2.2. Bowman et al (2005)
Four to seven SD rats per group were treated by gavage with corn oil or DBF at
500 mg/kg-d from GDs 12-19 or GDs 12-21. The animals were sacrificed on GD 19 or 21, and
Wolffian ducts (WD) were pooled from three to four fetuses (to obtain enough RNA for
analysis) within the same litter for gene expression analysis. Global changes in gene expression
were determined by Clontech Atlas Rat Toxicology 1.2 cDNA expression array (1,185 genes).
Images were collected using a Phosphorimager and then imported into Atlaslmage 2.01 and
GeneSpring 4.2 for analysis. Selected genes were further examined by real-time quantitative
RT-PCR using the GeneAmp 5700 Sequence Detection System. Total RNA was isolated,
DNAse-treated, and reverse-transcribed using TaqMan reagents. Twenty-three primer sets were
used for RT-PCR analysis. Reactions were standardized using GAPDH-specific primers. The
genes analyzed by RT-PCR include those in the insulin-like growth factor (Igf) pathway, the
matrix metalloproteinase (Mmp) family, the extracellular matrix, and other developmentally
conserved signaling pathways: bone morphogenetic protein 4 (Bmp4), collagen, delta like (Dlk),
mitogen-activated protein kinase 12 (Map3kl2), epidermal growth factor receptor (Egfr),
fibroblast growth factor 10 (FgflO), FGF receptor 2 (Fgfr2), fibronectin, insulin-like growth
factor 1 (Igfl), insulin-like growth factor 2 (/g/2), insulin-like growth factor 1 receptor (Igflr),
insulin-like growth factor binding protein 5 (IgfbpS), integrinAS, integrinBl, matrix Gla protein
(Mgp\ matrix metallopeptidase 2 (Mmp2), matrix metallopeptidase 14 (Mmp 14), matrix
metallopeptidase 16 (Mmp 16), Notch2 receptor (Notch2), and tissue inhibitors of MMPs (Timpl,
Timp2, and Timp3). Immunohistochemistry was also performed to evaluate changes in
localization and/or intensity of IGFLRP and androgen receptor (AR) protein expression.
Microarray data were not presented due to considerable variability in gene expression
levels within the treatment group at each age. Based on real-time RT-PCR analysis, compared
with controls, prenatal exposure to DBF from GDs 12-19 or GDs 12-21 increased mRNA
expression of different members of the IGF family including Igfl (on GDs 19 and 21), Igf2(on
GD 19), Igfrlr (on GD 19), and IgfbpS (on GD 21) in the developing WD, while Egfr was
unchanged on GDs 19 and 21. Additionally, mRNA expression of^4r, Bmp4, integrmAS, Mmp2,
andMap3k!2 was increased on GD 19; mRNA expression ofFgflO, Fgfr2, Notch2, Mmp2,
Timpl, and Mgp was increased on GD 21. IGFLRP immunostaining was higher in the cytoplasm
of the ductal epithelial cells and increased in the cytoplasm of mesenchymal cells in
5-7
-------
DBF-exposed fetuses compared with that in controls. In general, reduction of AR
immunostaining in the nuclei of ductal epithelial cells of DBF-exposed WD was observed on
GD 19. Compared with controls, WDs dissected from GD 19 DBF-exposed fetuses were slightly
smaller in size (underdeveloped) and appeared to be more fragile. By GD 21, control fetus WDs
were markedly coiled, while those from the exposed fetuses exhibited less coiling.
Prenatal DBF exposure appears to alter the mesenchyme-epithelial signaling of growth
factors (e.g., IGFs) and other developmentally conserved pathways (e.g., BMP4) in WDs.
Bowman et al. (2005) contend that the effect of DBF on WD differentiation is likely a
consequence of decreased fetal testicular T, although direct effects of DBF on the developing
WD independent of T are also possible.
5.2.2.3. Liu et al (2005)
Five to ten SD rats per group were treated by gavage with corn oil, DBF (500 mg/kg-d),
or one of six other phthalate esters (500 mg/kg-d) daily from GDs 12-19. The six other
phthalate esters include diethyl phthalate (DEP), dimethyl phthalate (DMP), diocytyl tere-
phthalate (DOTP), diethylhexyl phthalate (DEHP), dipentyl phthalate (DPP), and butyl benzyl
phthalate (BBP). Testes were collected on GD 19, homogenized, and then total RNA was
isolated. RNA integrity was assessed using an Agilent 2100 Bioanalyzer. cDNA was
synthesized from 2.5 ug total RNA and purified using RiboAmp OA. The BioArray High-Yield
RNA Transcript Labeling Kit was used for cRNA amplification and biotin labeling. Affymetrix
GeneChip Sample Cleanup Module was used for purifying and fragmenting the cRNA. The
Complete GeneChip® Instrument System was then used to hybridize, wash, stain, and scan the
GeneChip arrays (RAE230A and RAE230B; -30,000 genes). The data were analyzed using
analysis of variance (ANOVA [one-way, two-way, nested one-way]), Dunnett test (post hoc),
Tukey test, and Bonferroni adjustment.
Image files obtained from the scanner were analyzed with the Affymetrix Microarray
Suite (MAS) 5.0 software and normalized by global scaling. Absolute analysis was performed
for each array prior to comparative analysis. To identify differentially expressed transcripts,
pair-wise comparison analyses were carried out with MAS 5.0 (Affymetrix). The ^-values were
determined by the Wilcoxon's signed rank test and denoted as "increase," "decrease," or "no
change." A transcript is considered significantly altered in relative abundance when/? < 0.05.
Analysis using MAS 5.0 provides a signal log ratio (SLR), which estimates the magnitude and
5-8
-------
direction of change of a transcript when two arrays are compared (experimental versus control).
The SLR output was converted into "fold-change" as recommended by Affymetrix.
Furthermore, stringent criteria were used to identify robust signals as follows: (1) software call
of "present," and (2) > 2-fold change or SLR 1.0, in both replicates. Average and standard
deviations were calculated for all the fold-change values. In general, only transcripts induced or
suppressed by > 2-fold were considered as differentially expressed.
Selected genes were further examined by real-time quantitative RT-PCR using 18 primer
sets. The genes analyzed by RT-PCR include epididymal secretory protein 1 (reJ), low-density
lipoprotein receptor (Ldlr), 17p-hydroxysteroid dehydrogenase 3 (HsdJ7b3), l?p-hydroxysteroid
dehydrogenase 7 (HsdlJbT), luteinizing hormone/choriogonadotropin receptor (Lhcgr),
CCAAT/enhancer-binding protein (C/EBP), beta (Cebpb)., early growth response 1 (EgrJ),
nuclear receptor subfamily 4, group A, member 1 (Nr4al), nuclear factor, interleukin 3,
regulated (Nfil3), nuclear receptor subfamily 0, group B, member 1 (NrObl), transcription factor
1 (Tc/7), insulin-induced gene 1 (Insigl), protein kinase C-binding protein (Prkcbpl), decay-
accelerating factor (Daf), dopa decarboxylase (Ddc), seminal vesicle secretion 5 (Svs5), and
testis-derived transcript (Testin). Anogenital distance (AGD) was measured and
immunohistochemistry was performed for NROB1, TESTIN, GEB14, DDC, and CEBPB
proteins.
Of-30,000 genes examined, 391 were statistically significantly altered following
exposure to the four developmentally toxic phthalates (DBF, BBP, DPP, and DEHP) relative to
the controls. While the four developmentally toxic phthalates were indistinguishable in their
effects on global gene expression, no significant changes in gene expression were detected in the
phthalates that do not lead to developmental effects (DMP, DEP, and DOTP). Of the 391 genes
altered by the developmentally toxic phthalates, 225 were unknown and uncharacterized
transcribed sequences. Of the remaining 166 genes, the largest GO classification (31 genes) was
of genes related to lipid, sterol, and cholesterol homeostasis. Additional GO classification
groups include genes involved in lipid, sterol, and cholesterol transport (10 genes);
steroidogenesis (12 genes); transcription factors (9 genes); signal transduction (22 genes);
oxidative stress (11 genes); and cytoskeleton-related (13 genes). RT-PCR results indicated that
the developmentally toxic phthalates reduced the mRNA levels ofHsd!7b7, Lhcgr, Ldlr, rel,
Svs5, Insigl, and Ddc. Additionally, the RT-PCR results indicated that the developmentally
5-9
-------
toxic phthalates induced the mRNA levels ofGrb!4, Prkcbpl, and Testin. RT-PCR results also
indicated that gene expression of several transcription factors including Dax-1, Cebpb, Nfil3,
Nr4al, and Tcfl were significantly changed by at least one of the toxic phthalates. Based on
immunohistochemical analysis, DAX-1 expression was reduced in the gonocyte population of
DBF-treated testis compared with that of controls. Additionally, the expression of nuclear
CEBPB, GRB14, and DDC proteins was reduced in interstitial cells of DBF-treated testis, while
TESTIN and GRB14 expression levels were increased in Sertoli cells of DBF-treated testis. An
AGO reduction was observed in male fetuses exposed to any of the developmentally toxic
phthalates.
This study showed that the four phthalates (DBF, DEHP, BBP, and DPP) that have
similar effects on the developing male rat reproductive tract are indistinguishable in their
genomic signature for the developing fetal testis. These phthalates targeted pathways in LC
production of T and other pathways that are important for normal interaction and development
between Sertoli cells and gonocytes. By contrast, a different genomic signature was observed in
animals exposed to any of the four phthalates that do not exhibit developmental toxicity.
5.2.2.4. Thompson et al (2005)
Four SD rats per group were gavaged with corn oil or DBF daily at 500 mg/kg-d. In the
first study, DBF treatment was 30 minutes, 1 hour, 2 hours, 3 hours, 6 hours, 12 hours, 18 hours,
or 24 hours before sacrifice on GD 19. Global changes in gene expression were determined by
Affymetrix GeneChips (the specific GeneChips used in the study were not reported). The
methods were similar to Liu et al. (2005)—with the exception of the statistical analysis.
Thompson et al. (2005) used IMP statistical software to perform Student t-tests or one-way
ANOVAs with Tukey post hoc analysis. Selected genes were further examined by real-time
quantitative RT-PCR. An ABI Prism 7900HT Detection System, the SYBR Green PCR Master
Mix, and 30 primer pairs were used for analysis of DBF-induced changes in gene expression.
The genes analyzed by RT-PCR included Cypllal, Scarbl, Star, Cypl7al, Egrl, Egr2, Nr4al,
Nfil3, Tcfl, serum/glucocorticoid regulated kinase (Sgk), tumor necrosis factor receptor
superfamily, member 12a (Tnfrsfl2a), sclerostin domain containing 1 (Sostdcl), Wnt oncogene
homolog 4 (Wnt4), B-cell translocation gene 2, antiproliferative (Btg2), C/EBP, delta (Cebpd),
FBI murine osteosarcoma viral oncogene homolog (Fos), dual specificity phosphatase 6
5-10
-------
(Duspd), Hes6_predicted, interferon-regulated developmental regulator (Ifrdl\ Ldlr, nuclear
receptor subfamily 4, group A, member 3 (Nr4a3\ Pawr, NrObl, Jun-B oncogene (Junb),
endothelial differentiation sphingolipid G-protein-coupled receptor 3 (Edg3), thrombospondin 1
(Tspl), and stanniocalcin 1 (Stcl). Immunoblotting by SDS-PAGE was performed for SCARE 1,
CYP1 lal, STAR, and CYP17al. Fetal testicular T concentration was measured by
radioimmunoassay.
Based on microarray analysis, 106 genes in the DBF-treated groups were significantly
different from time-matched controls. Six genes were significantly elevated within 1 hour after
DBF exposure. An additional 43 genes were upregulated, and five genes were downregulated
3 hours after DBF exposure. The rapid induction of these genes at 1 hour was a transient effect;
none of the genes that were upregulated after 1 hour of DBF treatment remained significantly
different than the controls 6 hours after treatment. Only nine genes showed significant changes
from the control group between the 3- and 6-hour time points. After 1 and 3 hour DBF
exposures, the majority of the changes in expression had reflected increased transcription. At
6 hours after exposure, 19 genes were downregulated and 17 were upregulated. Based on
RT-PCR analysis, the immediate early gene, Fos, and the putative mRNA destabilizing gene,
zinc finger protein 36 (ZJp36), were at peak expression level 1 hour after DBF exposure. Other
immediate early genes were at peak expression at 2 hours after DBF exposure. At 3 hours after
exposure, the expression of Cebpd, Cxcll, and Nr4a3 increased rapidly, while other genes
showed a more gradual increase. Tspl expression was increased 25-fold at 3 hours after
exposure and returned to baseline at 6 hours after exposure. Genes involved in testicular
steroidogenesis were first noticeably affected 2 hours after DBF exposure. Inhibition of Star
transcription was detected ~2 hours after DBF exposure. Scarbl, Cypllal, and Cypl7al
showed a significant decrease in expression at about 6 hours after DBF exposure. At 6 hours
after exposure, the T concentration dropped to approximately the level observed after long-term
DBF treatment. At 12 hours after exposure, steroidogenesis-associated genes, NrObl andNr4aJ,
were elevated. Tcfl and Sgk were downregulated soon after DBF exposure, but values returned
to control levels by 3 hours after DBF exposure. Sostdcl and Hes6_predicted returned to control
levels at 6 hours after exposure. Based on radioimmunoassay, a decrease in fetal testicular T up
to 50% was observed within an hour after DBF exposure.
5-11
-------
In a second experiment to compare the effect of DBF on steroidogenesis in the fetal
adrenal gland, DBF treatment at GDs 12-19 was followed by analysis of gene expression in this
tissue. A decrease (but not statistically significant) of corticosterone after GDs 12-19 DBF
exposure was observed in the fetal adrenal. The expression of genes involved in steroidogenesis
was less affected in the adrenal (males and females) than in the testes. This study indicates that
the effect of DBF exposure on steroidogenesis gene expression is specific to the fetal testis and
not in other steroidogenic organs.
Rapid transcriptional changes after DBF exposure in a number of genes could be
responsible for the reduction in steroidogenesis. Peroxisome proliferator-activated receptors
(PPAR) activation is ruled out since changes in expression of genes targeted by PPAR a and y
are not observed until 3 hours after DBF treatment. Many of the genes whose upregulation was
detected within the first hour after treatment were "immediate early genes," meaning genes
involved in cell growth and differentiation. One possible mechanism for DBF's repression of
steroidogenesis is that DBF may initially stimulate the mitogen-activated protein
kinase/extracellular signal-regulated kinase (MAPK/ERK) pathway in the fetal testis. Increased
expression ofEgrJ and ZJp36 could, in turn, lead to degradation of the transcripts involved in
testicular steroidogenesis. Consistent with this possibility, the Star mRNA contains the AU-rich
element, which are regions with many A and U bases that target the RNA for degradation, in
target transcripts of ZJp36.
5.2.2.5. Plummer et al. (2007)
Five Wistar rats per group were gavaged with corn oil or DBF at 500 mg/kg-d from
GD 12 until the day prior to sacrifice. Animals were sacrificed on GD 15, 17, or 19 and used for
immunolocalization, Western analysis, or RNA quantification (of whole testes, seminiferous
cord, or interstitial regions using laser capture microdissection). Samples for laser capture
microdetection were collected from sections of single testes from GD 19 animals. RNA samples
from three treated litters were compared to a pool of RNA samples from control animals to
lessen errors due to biological variation. The Agilent 22K rat and 44K whole-rat oligonucleotide
arrays were used for analysis of the whole-fetal testes and microdissected tissue, respectively.
RNA was isolated from the homogenized whole-fetal testes using the RNeasy mini kit (Qiagen)
and from laser capture microdissected samples using RNeasy micro kit (Qiagen). Isolated RNA
5-12
-------
was labeled using the Agilent Low Input Linear Amplification Labeling kit according to the
manufacturer's instructions. Specific activity of the labeled cRNA was measured using the
microarray analysis program on aNanoDrop ND1000 spectrophotometer (Montchanin, USA).
Microarray analysis with whole-fetal testis RNA was performed using Agilent 22K rat
oligonucleotide arrays (Agilent #G4110A). Regional microarray analysis on RNA isolated from
laser capture microdissected fetal testis tissue was performed using Agilent 44K whole-rat
genome oligonucleotide microarrays (Agilent #G4131 A). Microarray data analysis was
conducted using Agilent feature extraction (v7.1) and Rosetta Luminator software (Rosetta
Biosoftware, Kirkland, USA) to generate "signature" lists, defined as significantly (p < 0.01)
different. The compare biosets function in Luminator was used to compare signature lists from
different fetal testis regions. Pathway analysis used Ingenuity Pathways Analysis software.
DBF induced statistically significant changes in gene expression at all three time points.
At GD 15 in whole testes, expression of genes regulating lipid metabolism, redox homeostasis,
cell proliferation, and apoptosis were altered. At GDs 17 and 19, these four main gene clusters
were altered: steroidogenesis (e.g., Cypl7al, Cypllal), lipid metabolism, cholesterol (e.g., Star,
ScarbJ), and redox homeostasis. In laser capture microdissection studies of GD 19 tissue, both
regions demonstrated altered expression of genes associated with steroidogenesis (e.g.,
CypJ7aJ\ cholesterol transport (e.g., Scarbl), cell/tissue assembly, and cellular metabolism. In
the interstitial regions only, genes involved in fatty acid oxidation, testes morphogenesis, and
descent (e.g., InsI3) were altered. In the cord samples, genes associated with stress responses,
chromatin bending, and phagocytosis were altered.
RT-PCR analysis was performed on RNA from GD 19 testes from five rats/group using
sequence-specific primers for the orphan nuclear receptor, nuclear receptor subfamily 5, group
A, member 1 (Nr5al; also known as steroidogenic factor 1 fSflJ), Star, Cyplla, andlnslS. The
data were analyzed using a one-way ANOVA, followed by the Bonferroni post-test using
GraphPad Prism. These studies showed a statistically significant reduction in the expression of
Star, Cypllal, and Insl3 but notNrSaL
Analysis of protein expression at GD 19 showed DBF-induced reduction in levels of
CYP11 A, inhibin-a, cellular retinoic acid binding protein 2 (CRABP2), and
phosphatidylethanolamine binding protein (PEBP) in LCs, and no change in Sertoli
cells/seminiferous cords. These data correlated with microarray data for the genes coding for
5-13
-------
these proteins. Immunoreactivity for antimullerian hormone (AMH) was slightly increased in
Sertoli cells following DBF treatment. Western blot analysis and immunolocalization of NR5A1
demonstrated no effects of DBF on protein expression in Sertoli or LCs. Using time plots to
assess time-dependent changes in gene expression, a coordinate down-regulation oflnhibin-a,
Scarbl, Star, andCypllal was observed between GDs 15 and 19.
This study confirms other study results, showing down-regulation of Scarbl, Star,
Cypllal, and Cypl7al. The authors suggest that DBF induces LC dysfunction indirectly
through sequestration of cofactors used in key signaling pathways and not through decreases in
NR5A1 protein expression. They further state that the use of Wistar rats could be important, as
Wistar rats may be more susceptible than SD rats to testicular effects of DBF.
5.2.3. RT-PCR Studies
5.2.3.1. Barlow et al (2003)
Six to seven SD rats per group were gavaged with corn oil or DBF at 500 mg/kg-d from
GDs 12-19. Testicular RNA was then isolated from three randomly selected male fetuses per
litter. RT-PCR studies were performed as described in Shultz et al. (2001).
The mRNA of 13 preselected genes in the steroid biosynthetic pathway was analyzed by
real-time RT-PCR; immunohistochemical and oil red O histochemical analyses were performed to
further confirm mRNA changes. The 13 genes analyzed were Scarbl, Star, Cypllal,
hydroxyl-delta-5-steroid dehydrogenase, 3 beta- and steroid delta-isomerase 1 (Hsd3b), Cypl7al,
hydroxysteroid (17-beta) dehydrogenase 3 (Hsdl7b3), Ar, luteinizing hormone receptor (Lhr),
follicle-stimulating hormone receptor (Fshr), Kit, stem cell factor (Scf), Pcna, and Clu.
Compared with controls, mRNA expression was downregulated for Scarbl, Star, Cypllal,
Hsd3b, Cypl7al, and Kit in DBF-treated testes; mRNA expression was upregulated for Clu following
DBF exposure. These changes in mRNA expression were supported by immunohistochemical
localization of selected proteins and by staining for lipids.
The results in the study of Barlow et al. (2003) confirm the gene expression changes
observed in a previous study (Shultz et al., 2001). Furthermore, the data support alterations in
cholesterol synthesis, transport, and storage that likely play a role in decreased T production by
fetal LCs. The decreased level of mRNA expression for P450scc indicates another possible
contributor, as P450scc conversion of cholesterol to pregnenolone is the rate-limiting enzymatic
step in T biosynthesis.
5-14
-------
5.2.3.2. Lehmann et al (2004)
To date, Lehmann et al. (2004) is the only dose-response gene expression study on the
testis performed with DBF. The other studies used a single high dose shown to affect male
reproductive system development.
Five to seven SD rats per group were treated by gavage with corn oil or DBF at 0.1, 1.0,
10, 50, 100, or 500 mg/kg-d from GDs 12-19. Testes were then isolated on GD 19, and changes
in gene and protein expression were measured by real-time RT-PCR (as described in Shultz et
al., 2001) and Western analysis. Ten preselected genes in the steroid biosynthetic pathway were
analyzed by RT-PCR: Scarb, Star, Cypllal, Hsd3bl, Cypl7al, Kit, benzodiazepine receptor,
peripheral (Bzrp), Insl3, Clu, and sterol regulatory element binding factor 1 (Srebfl). Fetal
testicular T concentration was determined by radioimmunoassay in a separate group of animals
using doses of 0.1, 1.0, 10, 30, 50, 100, or 500 mg/kg-d.
The aim of this study was to determine the DBF doses at which statistically significant
alterations in the expression of a subset of genes and a reduction in fetal testicular T occur. As
summarized in Table 5-2, Lehmann et al. (2004) established 50 mg /kg-d as an LOEL and
10 mg/kg-d as an NOEL for reductions in genes and proteins associated with T production as
well as genes associated with other MO As (e.g., Kit, InslS) together with reductions in
intratesticular T. The Lehmann et al. (2004) study demonstrated a decrease mHsd3b (also called
3J3-HSD) gene expression involved in T synthesis was detected at levels as low as 0.1 mg/kg-d.
DBF exposure resulted in a dose-dependent decline in expression of the genes involved
in cholesterol transport and steroidogenesis: Scarb 1, Star, Cypllal, Hsd3b, Cypl7al, andlnslS.
Expression of Bzrp and Clu were increased in response to DBF. Furthermore, fetal testicular T
was significantly reduced at DBF doses > 50 mg/kg-d and reduced by 26% at 30 mg/kg-d. This
study reported a LOEL of 50 mg DBP/kg-d and a NOEL of 10 mg DBP/kg-d for reductions in
genes and proteins associated with T production together with reductions in intratesticular T. It
demonstrates the coordinated reduction in genes and corresponding proteins involved in
steroidogenesis and cholesterol transport, concurrent with a decrease in testicular T.
Importantly, the study results identify changes in T concentration and gene expression at DBF
doses lower than the observed effects on male reproductive development in toxicology studies
reviewed in this report (see Chapter 4).
5-15
-------
Table 5-2. Lehmann et al. (2004) DBF dose-response gene expression data
measured by RT-PCR showing statistically significant changes from control
Gene Symbol
(reported gene name)
Scarbl (Sr-Bl)
Star
Cypllal (P450ssc)
CyplJal
Hsd3b (3p-HSD)
Bzrp (PER)
Trpm2
Kit (c-Kit)
Iml3
Dose (mg/kg-d)
0.1
NC
NC
NC
NC
|0.3
NC
NC
|0.3
NC
1
40.6
NC
NC
NC
40.4
NC
NC
40.5
NC
10
NC
NC
NC
NC
NC
NC
NC
NC
NC
50
40.5
40.4
40.6
NC
40.5
NC
NC
40.3
NC
100
40.3
40.3
40.7
NC
40.3
NC
NC
40.5
NC
500
40.2
40.1
40.2
40.3
40.5
T2.0
tl.6
40.1
40.3
NC, no statistically significant change. Gene expression values are from DBF-exposed testes expressed
relative to control values and are the statistically significant (p < 0.05)averages from five separate rat fetuses
from different dams per treatment group.
For Scarbl, Hsd3b, and Kit, significant reductions in mRNA levels were observed at
DBF doses that approach 0.1 mg/kg-d. Thus, alterations in the expression of Scarbl, Hsd3b, and
Kit are at least sensitive indicators of DBF exposure. However, it is not clear whether alterations
in any one of these three genes alone or together can cause one or more reproductive
developmental effects of DBF.
5.2.3.3. Thompson et al. (2004)
Four to five SD rats per group were gavaged with corn oil or DBF at 500 mg/kg-d from
GDs 12-19. Testes were isolated on GD 17, 18, or 19. Testes mRNA was isolated, and four
preselected genes (Scarbl, Star, Cypllal, and Cypl7al) in the cholesterol and steroidogenesis
pathways were analyzed by real-time RT-PCR as described by Shultz et al. (2001).
Immunoblotting was performed using the total protein extracted from paired testis, and the
5-16
-------
expressed protein levels were quantified using FluorChem. Fetal testicular T concentration was
determined by radioimmunoassay, and whole-cell cholesterol uptake assessment was performed
on overnight cultures.
A significant decrease in fetal testicular T concentration was observed as early as GD 17
after in utero exposure to DBF. On GD 18, the decrease in T levels, as measured by the percent
difference in testicular T between treated and control testes, was much higher (17.8% of control
T levels) than on GD 17 (46.6% of control T levels). Furthermore, significant decreases in
mRNA expression of Scarbl, Star, Cypllal, and Cypl7al were observed as early as GD 17. In
agreement with T levels, the percentage difference of gene expression between control and
treated testes was higher on GD 18 than on GD 17. The suppression of the transcription by DBF
was a reversible effect, as the mRNA levels for all genes returned to control levels 48 hours after
DBF withdrawal. When protein expression was analyzed, results similar to the gene expression
data were obtained (i.e., strong expression in controls, decreased expression in treated animals
with 24-hour DBF withdrawal, and rising expression after the 48-hr DBF withdrawal).
Additionally, there was a significant decrease in the amount of cholesterol transported across the
mitochondrial membrane in the testes from DBF-treated fetuses as assayed in overnight cultures
of testis explants. This observation indicates that the decrease in Star mRNA correlated with
diminished protein function as transport of cholesterol from the outer to the inner mitochondrial
membrane by the STAR protein is a rate-limiting steps of steroidogenesis (Miller, 2008).
The results of this study demonstrate that DBF-induced suppression of T production in
the fetal testis correlates with diminished transcription of several genes in the cholesterol
transport and steroidogenesis pathways as early as GD 17. This diminished effect was
reversible, suggesting that DBF directly interferes with the signaling processes necessary for
maintenance of steroidogenesis or with the transcriptional regulators required to maintain
coordinate expression of the genes involved in cholesterol transport and T biosynthesis.
5.2.3.4. Wilson et al (2004)
In the study by Wilson et al. (2004), SD rats were treated by gavage with corn oil or a
developmental toxicant daily from GDs 14-18 in two separate experiments. In the first
experiment, five rats were treated with DEHP at 750 mg/kg-d and five rats were treated with
vehicle. In the second experiment, three rats were treated with one of six chemicals, each known
5-17
-------
to induce male reproductive malformations and three rats were treated with vehicle. The
chemicals used for the second study were three AR antagonists (vinclozolin [200 mg/kg-d],
linuron [100 mg/kg-d], and prochloraz [250 mg/kg-d]) and three phthalate esters (DEHP
[1 g/kg-d], DBF [1 g/kg-d], and BBP [1 g/kg-d]). Dams were sacrificed on GD 18, and testes
were removed and pooled by litter. In the first study, RNA was prepared to quantify expression
of one preselected gene, InslS, by real-time RT-PCR. In the second study, both steroid hormone
production (ex vivo incubation) and Insl3 expression were assessed. Total RNA was isolated
using Trizol, digested using Dnase I, and quantitated with RiboGreen. ImProm-II Reverse
Transcriptase was used for RT, followed by amplification using Taql. They completed RT-PCR
for Insl3 using a Bio-Rad iCycler.
In the first study, the mRNA expression oflnslS was reduced by -80% in DEHP litters
compared with that in control litters. In the second study, among the six chemicals tested, only
phthalate esters (DEHP, DBF, or BBP) reduced mRNA levels in the fetal testis, with DBF and
BBP being more effective than DEHP. In contrast, prochloraz, linuron, DEHP, DBF, or BBP
significantly reduced ex vivo T production.
In a previous study with antiandrogenic chemicals that alter male sexual differentiation
(Gray et al., 2000), phthalate esters were the only class that produced agenesis of the
gubernacular ligaments; some of the phthalate ester-exposed rats had a phenotype similar to that
seen in the Insl3 knock-out mouse. The study of Wilson et al. (2004) confirms this hypothesis
since only the three phthalates reduced Insl3 gene expression. The authors proposed that the
effects of DEHP, DBF, or BBP on Insl3 mRNA and T production result from a delay in
maturation of fetal LCs, resulting in hyperplasia as they continue to proliferate rather than
differentiate.
5.2.4. Study Comparisons
5.2.4.1. Microarray Study Methods Comparison
Table 5-3 compares the study design and method of determining statistical significance
across the five microarray studies. Because the Bowman et al. (2005) paper assessed changes in
gene expression in WD rather than testis, and because the microarray data were not presented in
the paper, the discussions will focus on the four other microarray studies. The Plummer et al.
(2007) study pooled control tissue and used the Agilent platform, which differed from the
5-18
-------
platforms used in the other studies. Liu et al. (2005), Schutz et al. (2001), and Thompson et al.
(2005) all assessed mRNA levels in rat testis—but with somewhat differing significance criteria.
All studies included vehicle-treated controls.
Table 5-3. Method comparisons for DBF microarray studies
Study
Bowman et
al., 2005
Liu et al.,
2005
Plummer et
al., 2007
Shultz et al.,
2001
Thompson et
al., 2005
Tissue
collected
Wolffian
ducts
Testis
Testis (whole,
laser captured
interstitial
tissue, or laser
captured
seminiferous
cord tissue)
Testis
Testis
Significance criteria
ND (microarray data not
presented)
p < 0.05 compared to
control by either 1-way
ANOVA, post hoc
Dunnett test, or Tukey
test
p < 0.01 using Agilent
feature extraction
software and then
Rosetta Luminator
software by performing
one-way ANOVA on log
fold change in the
replicates
2-fold change in average
expression value
compared to control
p < 0.05 multiple
comparison using
Bonferroni correction
Individual animals («) used?
No, pooled (3-4 fetuses/litter;
67 dams/treatment group)
Yes, (6 fetuses/litter;
6 dams/treatment group)
Yes for DBF-treated (3 pups
from 3 different dams); Control
RNAs were pooled
GDs 19 and 21 time points:
Yes, 1 fetus/litter;
3 dams/treatment group.
GD 16 time point: pooled RNA
from 5 fetuses/1 litter; 3 arrays
hybridized/treatment group.
Yes (NR)
ANOVA, analysis of variance; NR, not detected.
5.2.4.2. RT-PCR Study Methods Comparison
Table 5-4 compares the RT-PCR methods across the nine toxicogenomic published
studies. There were many similarities among the studies. With the exception of Bowman et al.
(2005), all groups extracted RNA from testis. All studies used a vehicle-treated control.
5-19
-------
Table 5-4. Method comparisons among the RT-PCR DBF studies
Study
Barlow et al.,
2003
Bowman et
al., 2005
Lehmann et
al., 2004
Liu et al.,
2005
Plummer et
al., 2007
Shultz et al.,
2001
Thompson et
al., 2004
Thompson et
al., 2005
Wilson et al.,
2004
Tissue
collected
Testis
Wolffian ducts
Testis
Testis
Testis (whole,
laser-captured
interstitial
tissue, or
laser-captured
seminiferous
cord tissue)
Testis
Testis
Testis
Testis
Significance criteria
(p values)
p < 0.05 compared to
control
p < 0.05 compared to
control
p < 0.05 compared to
control
p < 0.05 compared to
control by either 2-way
nested ANOVA or
Dunnett
p < 0.05 compared to
control, normalized to
1.0. Expressed as mean
+/- SEM; one-way
ANOVA followed by
Bonferroni post test using
GraphPad Prism software
p < 0.05 compared to
control
p < 0.05 compared to
control (Student's t-test
or 1-way ANOVA)
p < 0.05 normalized mean
of 3-5 fetuses/treatment
group relative to control
p < 0.01 compared to
control (means on a litter
basis)
Individual animals («) used?
Yes (3 fetuses/litter;
5 dams/treatment group)
No, pooled (3-4 fetuses/litter;
6-7 dams/treatment group)
Yes (5 fetuses/litter;
4-5 litters/treatment group)
Yes (control: 6 fetuses/dam;
6 dams for control. Treated:
3 fetuses/dam; 3 dams)
NDa; assessed GD 19.5 fetal
testes
GDs 19 and 21 time points:
Yes, 1 fetus/litter;
3 dams/treatment group.
GD 16 time point: pooled
RNA from 5 fetuses/1 litter;
3 arrays hybridized/treatment
group.
ND
Yes, 3-5 fetuses/litter
No, pooled for each litter
(3 dams/treatment group)
aNot clear from the Materials and Methods.
ANOVA, analysis of variance; ND, not detected.
5-20
-------
Most of the studies used the same significance criteria (p < 0.05). There were some differences
in the number of fetuses used per experiment while some studies pooled tissues.
There were also important similarities among the nine toxicogenomic studies. Eight of
the studies used the same strain of rat (SD), all purchased from the same vendor (Charles River,
Raleigh, NC). All studies described dissolving the DBF in corn oil, using a corn oil vehicle
control, and using oral gavage as the route of exposure. Six of the studies (Plummer et al., 2007;
Bowman et al., 2005; Liu et al., 2005; Thompson et al., 2004; Barlow et al., 2003; Shultz et al.,
2001) treated the animals by gavage with 500 mg/kg-d from GDs 12-19. This dose has been
shown to adversely affect male reproductive development without causing maternal toxicity or
fetal death. Lehmann et al. (2004) completed a dose-response during the GDs 12-19 period,
using 0, 0.1, 1.0, 10, 50, 100, or 500 mg/kg-d. Bowman et al. (2005) and Shultz et al. (2001)
included an additional exposure duration of GDs 12-21. Wilson et al. (2004) exposed for a
slightly shorter duration (GDs 13-17) and at a higher dose (1,000 mg/kg-d). This paper reports
exposures on GDs 14-18; however, these authors consider GD 1 as the day a sperm-positive
smear was identified in dams, whereas the other studies consider the sperm-positive day as
GD 0. Therefore, to be comparable with the other reports, we are reporting the exposure period
as GDs 13-17. Similarly, Plummer et al. (2007) reports exposures ranging from GDs 12.5-19.5,
which are equivalent to GDs 12-19 as the authors consider GD 0.5 to be the sperm-positive day,
adjusted to facilitate comparison.
All of the other selected studies collected testes for RNA extraction, with the exception of
Bowman et al. (2005), which collected WDs. Bowman et al. (2005) focused on the WD because
they were interested in characterizing the mechanisms responsible for prenatal DBF-induced
epididymal malformations. WD tissue from three to four fetuses was obtained to ensure enough
RNA for analyses (see Table 5-3). Since WDs are the precursor of the vas deferens, epididymis,
and seminal vesicles, the tissue assayed by Bowman et al. (2005) is different from the tissue
evaluated in the other seven studies (RNA from the testes of 1-3 fetuses). The studies used a
variety of toxicogenomic methodologies to assess changes in gene expression. General
descriptions of these methods utilized by the studies were presented in Section 5.1.
An important consideration is the reliability of the data being generated and compared in
these nine DBF studies. As discussed, the MAQC project (Shi et al., 2006) has recently
completed a large study evaluating inter- and intraplatform reproducibility of gene expression
5-21
-------
measurements (see Chapter 2). Six commercially available microarray platforms and three
alternative gene expression platforms were tested. Both Affymetrix microarrays and RT-PCR
assays were included in the MAQC testing. Affymetrix and the other one-color platforms
showed similar coefficients of variation of quantitative signal values (5-15%) when used to
detect 8,000 to 12,000 genes. When comparing variation within and between test sites, the
one-color assays demonstrated 80-95% agreement.
Although it is difficult to compare expression values generated on different platforms
because of differences in labeling methods and probe sequences, MAQC was able to show good
agreement between the Affymetrix platform and the other platforms. This was particularly true
when using the same biological sample (and, thus, removing variability introduced by the sample
or sample preparation method). It is worth noting that Affymetrix displayed high correlation
values with RT-PCR based on comparisons of-500 genes. The results of the MAQC report
suggest that the comparisons made in this case study are valid due to the reliability of the data.
Additionally, since seven out of the nine experiments in the case study were performed in the
same laboratory, interlaboratory variability is not an issue with these studies.
5.3. CONSISTENCY OF FINDINGS
In the assessment of consistency of findings, a potential source of incongruence is the
decreased sensitivity for low-expression genes in the microarray platforms as compared to the
gene expression technologies and differences in probe location.
5.3.1. Microarray Study Findings
An evaluation of the consistency across the four microarray studies of the testis was
performed. Bowman et al. (2005) is not included because the microarray study results were not
reported. In order to enhance comparability, the data from the whole testis microarray study of
Plummer et al. (2007) are included in the evaluation, but the data from the microdissected
regions of the testis are excluded because the lack of comparison to any other study.
Three of the four microarray studies used the same strain, SD, and all nine used the same
species (rat). Plummer et al. (2007) was the only study to use the Wistar rat strain because it is
considered more susceptible to effects on the testis than SD. Table A-l in Appendix A includes
those genes whose expression was reported to be significantly altered, as reported by Shultz et al.
(2001), Thompson et al. (2005), Plummer et al. (2007) (for the whole testis only), or Liu et al.
5-22
-------
(2005). Also presented in Table A-l are the official gene names, exposure times, and directional
response changes. It should be noted that some differences are to be expected in these
comparisons because no two studies had identical study designs or platforms, or applied the
same statistical cut-offs. For example, Thompson et al. (2005) used a very short duration of
exposure, whereas the other three studies had longer exposure durations. In addition, the
Affymetrix microarray platform was used only by Thompson et al. (2005) and Liu et al. (2005).
The three testis microarray studies (Plummer et al., 2007; Liu et al., 2005; Thompson et
al., 2005) that used the "second generation chips" containing a much larger number of probes
(therefore, covering many more genes) than the Clontech platform were compared. The Venn
diagram, developed for these three studies, shows some unique gene expression changes for each
study as well as a number of common gene expression changes (see Figure 5-1). Nevertheless,
significant corroboration in the direction of effect among the common genes was observed in
these three studies (see Appendix A). Additionally, most of the common genes were
downregulated after in utero DBF exposure. Further, two genes in the steroidogenesis pathway,
Cypllal, and Scarbl, are common among all four microarray studies. These findings indicate
that the microarray data set for DBF is relatively consistent and findings are reproducible.
A number of genes involved in steroidogenesis (Cypllal, Scarbl, Star, and Cypl7al)
were found to be downregulated by DBF in all three studies (see Figure 5-1). Other genes
significantly altered include a downregulation of the serotonin and catecholamine pathway
enzyme, Ddc, and the myosin, heavy polypeptide 6, cardiac muscle, alpha (Myh6), and the
androgen-regulated structural protein, Svs5.
Other genes were significantly altered in two of the three studies. For example, in
comparing the results of the two studies that utilized the same platform (Affymetrix), the Liu et
al. (2005) and Thompson et al. (2005) studies observed a downregulation of the steroidogenesis
genes Sqle and Hsd3bl_predicted, cyclin-dependent protein kinase inhibitor (Cdknlc), the
cellular retinoic acid binding protein 2 (Crabp2\ the FGF receptor activating protein 1 (Fragl),
and the fatty acid binding protein (Fabp3). These same two studies found upregulation of the
steroidogenesis gene Nr4al.
There are a number of genes for which the different studies found a similar significant
alteration but the direction of effect varied. For example, GSH S-transferase, mu 2 (Gstm2), a
gene involved in xenobiotic metabolism, was found to be significantly downregulated by Liu et
5-23
-------
Thompson et al. (2005)
Plummer et al. (2007)
| Cypllal
\,Cypl7al
Scarbl
[Ddc
[Fdxl
[Myh6
I Prdx3
| Star
[Svs5
[Idhl
[Lhgcr
t Nr4al
[Sqle
Stcl
\Tpml
Liu et al. (2005)
Figure 5-1. Venn diagram illustrating similarities and differences in
significant gene expression changes observed in three recent microarray
studies of the testes: Thompson et al. (2005), Plummer et al. (2007), and Liu
et al. (2005). Numbers within each circle indicate genes whose expression was
statistically significantly altered and that are unique to the study (i.e., not
replicated by either of the other two studies). Gene numbers do not include
expressed sequence tags (ESTs). The red circle indicates the Thompson et al.
(2005) study; the green circle indicates the Plummer et al. (2007) study; and the
blue circle indicates the Liu et al. (2005) study; Black arrows indicate the
direction of effect, which was the same for all three of these studies.
al. (2005) and Thompson et al. (2005) and significantly upregulated by Shultz et al. (2001). The
microsomal GSH S-transferase 1 gene (Mgstl) was downregulated in Liu et al. (2005) and
upregulated in Shultz et al. (2001). Appendix A presents a table of the statistically significant
gene expression changes in the Thompson et al. (2005), Shultz et al. (2001), Liu et al. (2005),
5-24
-------
and Plummer et al. (2007) studies. These differences in microarray results can be explained by a
number of factors including study differences (e.g., duration of exposure, platform, and rat
strain) and/or variability of microarray study results.
Overall, the data indicate that there are some unique gene expression changes for each
study as well as a number of common gene expression changes. Significant corroboration in the
direction of effect among the common genes was observed in at least three studies. In addition,
most of the common genes among these three studies were downregulated after in utero DBF
exposure. These findings indicate that the microarray data set for DBF is very consistent and
reliable although certain uncertainties remain when comparing data from different platforms with
different study design.
5.3.2. RT-PCR Gene Expression Findings
Comparisons were also made of RT-PCR data (see Table A-2; Appendix A). All nine
studies performed RT-PCR, and in the case of Liu et al. (2005), Shultz et al. (2001), Plummer et
al. (2007), and Thompson et al. (2005), RT-PCR was performed following identification of the
genes of interest from microarray studies. A number of genes were found to be similarly up- or
downregulated by in utero DBF exposure. In the steroidogenesis pathway, five genes (Cypllal,
Cypl7al, Hsdl7b3, Scarbl, and Star) were found to be downregulated by more than one
laboratory. Some commonalities were also observed in altered gene regulation of transcription
factors. Egrl, Nfil3, and Nr4al were shown in two different studies to be upregulated. Two
studies reported similar downregulation ofNrObl and Tcfl.
Three studies (Plummer et al., 2007; Lehmann et al., 2004; Wilson et al., 2004) observed
reduced Insl3 gene expression. As discussed, Insl3 has a role in sexual differentiation and testis
descent. Reduced fetal Insl3 has been shown to produce agenesis of the gubernacular ligaments.
Two other genes have been shown to have DBF-induced altered expressions as assessed by
RT-PCR in two laboratories: Clu (upregulated) and Kit (downregulated).
5.3.3. Protein Study Findings
All nine studies completed either Western analysis (immunoblotting) or
immunohistochemistry to characterize fetal DBF-induced changes in protein expression.
Usually, protein analysis was conducted for proteins that had demonstrated changes in mRNA
5-25
-------
expression. However, up- or downregulation of genes and proteins does not always occur
simultaneously, so a disparity between these two experimental results is quite common.
Table 5-5 presents the protein-expression data from these studies.
Four proteins in the steroidogenesis pathway were shown to be downregulated by DBF
exposure. These findings are fairly consistent with the gene expression data presented earlier.
STAR was shown to be downregulated by Western blotting in three separate experiments, and by
immunolocalization in another experiment. STAR expression was found only in LCs in both the
control and DBF-treated testes, with the DBF-treated testes having decreased staining intensity
(Barlow et al., 2003). Quantitatively, three experiments demonstrated reduced SCARE 1 protein
levels in DBF-treated fetal testes; however, immunolocalization showed DBF-induced increased
staining of Sertoli cells and decreased staining of LCs. Both CYP1 lal and CYP17al protein
levels were shown in several separate experiments to be reduced following DBF exposure, which
correlated with microarray and PCR findings. Immunolocalization was completed for CYP1 lal
and found to be downregulated in LCs (Plummer et al., 2007). Using immunolocalization, CLU
was found to be increased in Sertoli cells and LCs. One study has shown that DBF lowers
INSL3 protein immunoexpression levels in the fetal testis (McKinnell et al., 2005). The
expression of NR5A1/SF1 was unchanged in Wistar rats, however, four proteins regulated by
NR5A1 (CYP1 lal, INHA, CRABP2, and PEBP) and AMH were reduced in LCs following DBF
exposure (Plummer et al., 2007).
5.3.4. DBF Toxicogenomic Data Set Evaluation: Consistency of Findings Summary
A comprehensive summary of the DBF toxicogenomic data set assessed in this case
study, including all microarray, RT-PCR, and protein data from the nine studies, is presented in
Figure 5-2. The genes and protein included in the figure are limited to those for which two or
more studies detected statistically significant results. In many cases, when comparing across
RT-PCR and microarray studies, a DEG is found in one or even several studies that is not
identified in another study. For example, Kit was downregulated in the Barlow et al. (2003),
5-26
-------
Table 5-5. Evaluation of the published protein studies after DBF in utero exposure (testes only)
Official
Gene
Symbol
Amh
Bcl2
Bzrp
Cebpb
Crabp2
Clu
Clu
Cypllal
Cypllal
Cypllal
Cypl7al
Cypl7al
Cypl7al
Ddc
Grbl4
Inha
Insl3
Kit
KM
Protein Name
Reported in
Paper
AMH
bcl-2
PER
CEBPB
CRABP2
PEBP
TRPM-2
TRPM-2
CYPllal
P450ssc
P450ssc
CYP17al
CYP17
cypl?
Dopa
decarboxylase
GRB14
INHA
InslS
c-kit
SCF
Exposure
Interval
CDs 12-19
CDs 12-21
CDs 12-19
CDs 12-19
CDs 12-19
CDs 12-19
CDs 12-21
CDs 12-19
GD 18 for 18 hrs
CDs 12-19
CDs 12-17 or 18
GD 18 for 18 hrs
CDs 12-17 or 18
CDs 12-19
CDs 12-19
CDs 12-19
CDs 12-19
CDs 12-19
GDsl2-19
CDs 12-19
Dose
(mg/kg-d)
500
500
500
500
500
500
500
500
500
500
500
500
500
500
500
500
500
500
500
Method Used
Immunolocalization
Immunolocalization
Immunolocalization
Immunolocalization
Immunolocalization
Immunolocalization
Immunolocalization
Western analysis
Western analysis
Western analysis
Western analysis
Western analysis
Western analysis
Immunolocalization
Immunolocalization
Immunolocalization
Immunolocalization
Immunolocalization
Immunolocalization
Change in Protein Expression Compared
to Controls
t slightly in Sertoli cells
t in Sertoli and Leydig cells
j in interstitial cells
J, in interstitial cells
J, in Leydig cells
t in Sertoli and Leydig cells
t in Sertoli cells
|(0.6 of control)
|(0. 5 of control)
|(0. 15 at 24 hrs; 0.5 at 48 hrs)
|(0.6 of control)
| (ND at 24 hrs; 0.4 of control at 48 hrs)
|(0.2 of control)
J, in interstitial cells
J, in interstitial cells and t in Sertoli cells
J, in Leydig cells
J, in interstitial cells
J, in gonocytes
t in Sertoli cells
Reference
Plummer et al., 2007
Shultzetal., 2001
Lehmann et al., 2004
Liu etal, 2005
Plummer et al., 2007
Shultzetal., 2001
Barlow et al., 2003
Thompson et al., 2005
Lehmann et al., 2004
Thompson et al., 2004
Thompson etal., 2005
Thompson etal., 2004
Lehmann et al., 2004
Liu etal., 2005
Liu etal., 2005
Plummer etal., 2007
Lehmann et al., 2004
Barlow et al., 2003
Barlow et al., 2003
fj\
to
-------
Table 5-5. (continued)
Official
Gene
Symbol
NrObl
Pebp
Scarbl
Scarbl
Scarbl
Scarbl
Star
Star
Star
Star
Testin
Protein Name
Reported in
Paper
DAX-1
PEBP
SCARE 1
SR-B1
SR-B1
SRB1
STAR
STAR
STAR
STAR
testin
Exposure
Interval
CDs 12-19
CDs 12-19
GD 19for6hrsor
GD 18 for 18 hrs
CDs 12-17 or 18
CDs 12-19
CDs 12-19
GD 18 for 18 hrs
CDs 12-17 or 18
CDs 12-19
CDs 12-19
CDs 12-19
Dose
(mg/kg-d)
500
500
500
500
50, 100, 500
500
500
500
50, 100, 500
500
500
Method Used
Immunolocalization
Immunolocalization
Western analysis
Western analysis
Western analysis
Immunolocalization
Western analysis
Western analysis
Western analysis
Immunolocalization
Immunolocalization
Change in Protein Expression Compared
to Controls
4 in gonocytes
I in Leydig cells
4 (0.3 of control)
| (0. 15 at 24 hrs; (0.7 of control at 48 hrs)
| (0.6, 0.5, and 0.1 of control)
1 in Leydig; t in Sertoli cells
|(0.4 of control)
| (ND at 24 hrs; 0.4 of control at 48 hrs
1(0.1,0.2, 0.1 of control)
4 in Leydig cells
t in Sertoli cells and gonocytes
Reference
Liu etal., 2005
Plummer et al., 2007
Thompson et al., 2005
Thompson et al., 2004
Lehmann et al., 2004
Barlow et al., 2003
Thompson et al., 2005
Thompson et al., 2004
Lehmann et al., 2004
Barlow et al., 2003
Liu etal., 2005
to
oo
ND, not detected.
-------
Gene
fj\
to
Steroidogenesis
Pathway/Function
Key
m=
M] =
fPl =
RT-PCR
microarray
protein
upregulation
downregulation
Figure 5-2. Summary of DBF-induced changes in fetal gene and protein expression. M = microarray; R =
RT-PCR; P = protein. Red indicates upregulation; green indicates downregulation. Genes and protein included in the
figure are limited to genes that were statistically significantly altered in two or more studies. Gene symbols are
indicated at the top of the figure. The pathway or function of each gene is listed on the bottom of the figure. This
information has been taken from the case-study articles or from the DAVID (Database for Annotation, Visualization
and Integrated Discovery http://david.abcc.ncifcrf.gov/list.jsp) entry for that gene.
-------
Lehmann et al. (2004), and Schultz et al. (2001) studies; by contrast, it was not altered
significantly in the Liu et al. (2005) study even though it is represented on the Affymetrix array.
Data from the Bowman et al. (2005) paper were not included because it evaluated
changes in DBF-induced gene expression in the WD rather than testes. There are no other WD
studies for comparisons. If an increase or decrease was reported at any time point, it was
included. Multiple time points from the Thompson et al. (2005) study were not all included; if
several time points showed a change, then it was recorded as one study showing a change. For
protein data, descriptions of immunohistochemical studies suggesting an increase, though
without real quantitation, were still counted. For the dose-response study (Lehmann et al.,
2004), data from only the 500 mg/kg-d dosing were used to allow better comparisons with the
other studies.
Figure 5-2 presents a summary of the changes in gene and protein expression following
in utero DBF exposure across studies. What is most striking is the consistency of evidence for
the DBF-induced downregulation of the steroidogenesis pathway. Both microarray and RT-PCR
analysis show consistent downregulation ofCypllal, Cypl7al, Star, and Scarbl mRNA
expression. Protein expression ofCypllal, Cypl7al, Star, and Scarbl is concurrently
downregulated. Downregulation of both Hsd3b and Lhcgr mRNA expression is demonstrated
consistently. Significantly, two genes involved in lipid/sterol/cholesterol transport, Npc2 and
Ldlr, also show downregulation. Three transcription factors (Nfil3, Egrl, and Nr4al)
demonstrate DBF-induced upregulation, while two genes (NrObl and Tcfl) show downregulation
in a number of experiments. Three immediate early genes (Fos, Egr2, and Zfp36) are
upregulated by DBF exposure. Interestingly, Clu (also known as T repressed prostate
message-2) is upregulated, as shown by two microarray, two RT-PCR, and two protein assays.
5.4. DAT A GAP SAND RESEARCH NEEDS
Based on the evaluation of the nine toxicogenomic studies, a number of research needs
became apparent. There are genomic data gaps for many environmental chemicals. For DBF,
confirmatory RT-PCR studies for all of the genes identified from microarray studies, would give
additional credence to the microarray results. Similarly, additional protein analysis, with
quantitation by Western blotting and with immunolocalization, could further characterize
DBF-induced effects on the male reproductive system. Looking at DBF-induced changes in
5-30
-------
gene expression in additional reproductive and nonreproductive (Thompson et al., 2005) tissues
could also add information about mechanism(s) of action and tissue specificity. As testes are
comprised of a number of cell types, evaluating additional homogeneous cell populations within
the testes, as Plummer et al. (2007) reported, could be useful.
In order to fully consider the question about informing the modes or mechanism of action
(see Chapters 1 and 3), using the toxicogenomic data to determine whether there are other MO As
responsible for some of the male reproductive developmental effects, we decided that it would be
helpful to analyze the raw data to assess all affected pathways. The published studies, while all
of excellent quality, focused their pathway analyses and descriptions on particular pathways of
interest to basic science. The following section describes efforts to reanalyze some of the DBF
microarray studies with this goal in mind.
5.5. PATHWAY ANALYSIS OF DBF MICROARRAY DATA
We determined that it would be advantageous to reanalyze the raw data utilizing multiple
analytical approaches (see Figure 3-1) because most of the DBF microarray studies in the
published literature were focused on further delineation of the mechanism of action relevant to
one MO A, the reduction in fetal testicular T. In fact, it was the microarray and RT-PCR study
results that identified the modulation of the steroidogenesis pathway as leading to reduced fetal
testicular T, one of the DBF MO As, and then, leading to a number of the male reproductive
developmental effects. Further, a second DBF MOA of reduced Insl3 gene expression has also
been identified (Wilson et al., 2004; see Chapter 3) leading to testis descent defects. Not all
pathways for the identified DEGs were discussed (or presented) in detail in the published studies
because of this focus. Therefore, a reanalysis that looks more broadly to define all pathways
affected by DBF may inform additional pathways related to MO As that could be linked to the
unexplained male reproductive developmental outcomes identified in Chapter 4. Thus, the
purpose of this reanalysis of the existing data set was to identify and characterize additional
molecular pathways affected by DBF, beyond a reduction in fetal T and Insl3 gene expression.
5.5.1. Objective of the Reanalysis of the Liu et al. (2005) Study
The goal was to reanalyze DBF microarray data to address the Case Study Question: Do
the genomic data inform DBF additional MOAs and the mechanism of action for the male
5-31
-------
reproductive developmental effects? The purpose for the reanalysis of the existing data sets is to
identify and characterize additional molecular pathways affected by DBF, beyond the effects on
the androgen-mediated male reproductive developmental toxicity pathways. This exercise was
designed to generate hypotheses about mechanisms/pathways that could underlie the unexplained
testicular endpoints after in utero DBF exposure (see Chapter 4).
The Liu et al. (2005) study was selected for reanalysis because the data set had a
comprehensive exposure scenario that covered the critical window for developmental exposure
to DBF (GDs 12-19). The Affymetrix chip was used (compatible with the proprietary and free
software programs used for pathway-level analysis), and the data were provided by Dr. Kevin
Gaido, a collaborator on this project. Some limitations of the Liu et al. (2005) data set are the
small number of samples (i.e., 3 controls and 3 DBF-treated) and the lack of characterization of
variance for treated and control. This study was a comparative analysis of six phthalate esters.
However, only the DBF treatment and vehicle control data were used for this analysis. The
Liu et al. (2005) study investigated global gene expression in the fetal testis following in utero
exposure to a series of phthalate esters, including both developmentally toxic phthalates (DBF,
BBP, DPP, and DEHP) and nondevelopmentally toxic phthalates (DMP, DEP, and DOTP)
(Liu et al., 2005). The original analysis was based on a two-way nested ANOVA model using
Bonferroni correction that identified 391 significantly expressed genes from the control out of
the approximately 30,000 genes queried. In their analysis, two classes of phthalate esters were
distinguished based on the gene expression profiles. The authors also showed that
developmentally toxic phthalates targeted gene pathways associated with steroidogenesis, lipid
and cholesterol homeostasis, insulin signaling, transcriptional regulation, and oxidative stress.
We can assume that the differentially expressed genes in common among the "developmental
phthalates" assessed in the Liu et al. (2005) study are due to phthalate exposure and not general
toxicity, providing internal positive controls.
5.5.2. Pathway Analysis of Liu et al. (2005) Utilizing Two Different Methods to Generate
Hypotheses for MO As Underlying the Unexplained Testes Endpoints
Pathway analysis methods and software have been previously developed for analysis of
microarray data for basic and applied research. Pathway-level analysis mainly depends on the
definition of the pathways (database) and significance level uses to measure the differential
expressions. Using these validated methods, a pathway analysis was performed. Differentially
5-32
-------
expressed genes that were input into the pathway analysis (GeneGo) were identified by two
different methods, Signal-to-Noise Ratio (SNR) and Rosetta Error Model (REM). By assessing
the intersection of the pathways identified by each approach provides a more conservative list of
pathways than using one approach. The overall process for generating hypotheses about
pathways that may be relevant to the testis endpoints using pathway analysis is illustrated in
Figure 5-3.
Liuetal. (2005) DBP data
Rosetta
Error Model
Signal to
Noise Ratio
I
Differentially .s ^"\ Differentially
unique Expressed f commonJ Expressed unique
Genes ^~^^^ Genes
I
I
Input filtered
gene list into
GeneGo
Input filtered
gene list into
GeneGo
I
Significant
Pathways
Significant
Pathways
Figure 5-3. Schematic of the two analysis methods (REM and SNR) for
identifying differentially expressed genes and subsequent pathway analysis
using GeneGo. Two separate analyses, REM and SNR statistical filters, were
performed to identify common and unique genes from the Liu et al. (2005) data.
The two separate filtered gene lists were input into GeneGo to identify
statistically significantly affected pathways. Common and unique pathway lists
were generated.
5-33
-------
5.5.2.1. Two Methods for Identifying Differentially Expressed Genes (DEGs)
5.5.2.1.1. Rosetta error model (REM)
The data set for the vehicle-treated and DBF-treated samples were input into the
proprietary software, Rosetta Resolver. A principal component analysis (PCA) of the entire data
set shows a distinct treatment response (i.e., the control and treated samples are clearly separated
into two distinct groups) but also demonstrates the variance in the data set between similarly
treated samples.
Next, the gene expression data were normalized using error-model algorithm in Rosetta
Resolver®, in part, because this software was available, but more importantly, because we
performed an internal evaluation of this algorithm compared to four other normalization
methods. The REM is a method for identifying DEGs that takes into account the variance of the
color intensity outputs from microarray studies. The error model conservatively estimates
intensity error and uses this approach to decrease the likelihood of identifying a change in gene
expression that is the result of intensity variance. When the results of REM were compared to li-
test and fold-change methods, the REM provided higher detection power (Weng et al., 2006).
The Rosetta Resolver system is a comprehensive gene expression analysis solution that
incorporates analysis tools with a robust, scalable database. Using the reference microarray data
set, Choe et al. (2005) compared a number of normalization methods including the quantile,
constant, invariant set, Loess, and error models. Receiver-operator characteristic curves were
generated to evaluate the sensitivity and specificity. Results showed that the REM identified 40-
50% more true positives compared to the other four methods (personal communication on June
2009 between Bill Ward [EPA/NHEERL] and Susan Hester [EPA/NHEERL]).
The annotated genes of the rat genome on the Affymetrix gene chip, -30,000 genes, were
input into the significance analysis using the Benjamini and Hochberg false discovery rate (FDR)
for multiple testing correction applied atp < 0.01, a relatively stringent statistical cut-off. Of the
-30,000 genes, the analysis passed 118 genes as being significantly altered following DBF
exposure. Of these, 17,496 genes did not pass the statistical filter and 13,428 genes were not
affected by the treatment. One possible reason that only 118 genes passed the multiple-testing
correction filter is that there is a high variance between individual samples, as demonstrated by
the PCA.
5-34
-------
Using the error-model algorithm analysis, the filtering criterion was changed top < 0.05
without applying an FDR because so few genes passed thep < 0.01 plus FDR filter which would
be limiting for pathway-analysis purposes. It is often the case that after correcting for multiple
hypothesis testing, few or no genes pass the threshold of statistical significance because the
biological variances are modest relative to the noise inherent in a microarray experiment
(Tomfohr et al., 2005). In performing DEG and pathway analysis, professional judgment is
required to determine when to use a highly stringent statistical significance filter and when to
focus on the available information regarding the biological significance of gene expression
changes. We considered it appropriate to use ap < 0.05 without applying an FDR in order
obtain a greater number of genes because the objective was to perform a pathway analysis in
order to gain new information about DBF toxicity. The DEGs identified using the REM are
shown in Table A-3 in Appendix A.
The set of 1,977 genes was deemed suitable to perform a comprehensive pathway-level
analysis because about one third of the DEGs (999) did not meet the statistical cut-off criteria (a
/>-value < 0.05). The list of 1,977 genes was input into the data analysis software program,
GeneGo, for pathway analysis. MetaCore's™ analytical tools enable the identification and
prioritization of the most relevant pathways, networks, and cellular processes affected by a given
treatment.
5.5.2.1.2. Signal-to-noise ratio (SNR)
We also identified DEGs by analyzing the Liu et al. (2005) data via SNR (Golub et al.,
1999), a method that differentiates between gene expression levels of two sample groups relative
to the standard deviation within each group. Consequently, a high SNR indicates that the two
sample groups are statistically more distinct whereas a low SNR indicates that the two sample
groups are less statistically distinct.
For a given gene, gt SNR is evaluated as in Eq. 5-1
' &
__
5-35
-------
where the means and standard deviations of the expression levels of gene g, are evaluated for the
samples in group 1, gtj (control), and group 2, gii2 (DBF treated).
SNR is used in quantitative noise analysis for microarray experiments (Tu et al., 2002)
and feature selection in classification gene expression studies (Goh et al., 2004; Shipp et al.,
2002). Here, SNR enables us to rank genes based on the assumption that genes whose
expression is related to DBF treatment should exhibit higher SNR values than genes whose
expression is unaffected by DBF. In order to identify DEGs, we evaluated a permutation test.
The multiple testing of-30,000 gene expressions poses a problem as the probability of Type I
errors increases with the number of hypotheses (Dudoit et al., 2003). To address this issue, we
executed thousands of comparisons by randomly permuting the gene expression levels from the
chip for each gene expression. Following this randomization process, ^-values were obtained as
the fraction of the randomized SNR values that are higher than the actual SNR. The genes that
were assigned a/>-value < 0.05 were characterized as DEGs (see Appendix A; the algorithm for
selecting DEGs [Figure A-l] and the list of identified DEGs [Table A-4]). 1,559 probe sets were
identified as DEGs. The heat map (see Figure 5-4) illustrates the distinction between the control
and DBF treated samples.
5-36
-------
Control 1 Control 2 Control 3 DBF 1
DBF 2
DBF 3
Figure 5-4. Heat map of 1,577 DEGs from SNR analysis method. The three lanes
on the left are vehicle treated and the three lanes on the right are DBF treated. Data
used for analysis from Liu et al. (2005). Control 1-3 lanes correspond to three replicate
control samples. DBF 1-3 lanes correspond to three replicate DBF-treated samples.
Rows represent the different 1,577 DEGs. The color red represents upregulation of
gene expression, and green represents downregulation of gene expression.
5.5.2.2. Pathway Analysis
Analysis of DBF toxicogenomic studies was carried out using many proprietary
databases and software packages with enhanced bioinformatic capabilities for pathway and
functional level analysis (Rosetta Resolver, MetaCore GeneGo, Ingenuity® Pathway
Knowledgebase). These software tools accept lists of genes of interest and then, using their
database of knowledge about these gene elements, map them to cellular pathways known to exist
from experimental literature. The advantage of trying to understand groups of genes acting in
the same cellular process, such as the cell cycle, is that effects on a pathway or biological process
likely provide meaningful biological information. In contrast, information about effects on
expression of one gene does not necessarily capture the relationship of the exposure to a
5-37
-------
chemical on a biological process or pathway. The rationale behind the exercise was that
interrogation of multiple databases would result in a more complete mining of the microarray
data sets, which may provide an understanding of all of the potential DBF MO As underlying the
testes reproductive developmental effects. Analysis using different statistical tools provides
information about the similarities and differences in results.
The GeneGo analysis normalized data set revealed that 131 biological processes
(p < 0.05) were associated with the 1,977 DEGs. The pathways with ap < 0.05 using the Rosetta
Error Model (REM) are listed in Appendix A (see Table A-5). Comparisons made on the level
of gene lists obtained by different statistical methods often do not converge (Manoli et al., 2006).
We decided to perform a comparison of methods based on the assumption that biologically
related groups of genes, such as metabolic or signaling pathways, may be more valid if identified
using different microarray analysis methods. Towards this effort, we input the gene list
(1,559 genes) using SNRto a pathway-level analysis using GeneGo, similar to the analysis
performed on the REM results. The pathway-analysis results of significant genes identified by
SNR are listed in Table A-6 of Appendix A. Table 5-6 lists the common pathways when two
different statistical filters for DEGs were conducted using the GeneGo pathway analysis (i.e., the
union of the two separate pathway lists; see Tables A-5 and A-6). In addition to the already
established changes in the steroidogenesis pathway, this analysis highlights biological processes
and pathways that are affected by DBF exposure to fetal testis. An assessment of linkages
between the unique pathways and processes identified to the DBF-induced male reproductive
toxicity outcomes can be made by querying the published literature.
5-38
-------
Table 5-6. Common pathways between the REM and SNR analyses of
differentially expressed genes (DEGs) after in utero DBF exposure from the
Liu et al. (2005) dataa'b'c
Biological Process
Cell adhesion
Cell signaling*
Disease
Growth and differentiation
Hormones
Immune response
Pathways
Cytoskeleton remodeling
ECM remodeling
Endothelial cell contacts by junctional mechanisms
Ephrins signaling
Integrin inside-out signaling
Integrin outside-in signaling
Integrin-mediated cell adhesion
Reverse signaling by ephrin B
Activation of PKC via G-Protein coupled receptor
CCR3 signaling in eosinophils
ChREBP regulation pathway
G-Protein beta/gamma signaling cascades
G-Proteins mediated regulation p. 38 and JNK signaling
Leptin signaling via JAK/STA T and MAPK cascades2
Regulation of actin cytoskeleton by Rho GTPases
Role PKA in cytoskeleton reorganization
NF-AT signaling in cardiac hypertrophy
NTS activation of IL-8 in colonocytes
Angiotensin activation of ERK
Angiotensin signaling via STATs
EPO-induced Jak-STAT pathway
MAG-dependent inhibition of neurite outgrowth
Regulation of acetyl-Co A carboxylase 2 activity in muscle
WNT signaling pathway
Ligand-dependent activation of the ESR1/SP pathway
MIF - the neuroendocrine-macrophage connector
CXCR4 signaling pathway
5-39
-------
Table 5-6. (continued)
Biological Process
Metabolism*
Metabolism*
Pathways
Androstenedione and testosterone biosynthesis and metabolism p.l2
Cholesterol biosynthesis2
Cholesterol metabolism2
dATP/dlTP metabolism
dGTP metabolism
Estrone metabolism
Fructose metabolism
G-alpha(q) regulation of lipid metabolism
Gamma-aminoburyrate (GABA) biosynthesis and metabolism
Glutathione metabolism
Glycolysis and gluconeogenesis (short map)
Glycolysis and gluconeogenesis p. 1
Glycolysis and gluconeogenesis p. 2
Histamine metabolism
Histidine-glutamate-glutamine and proline metabolism
Leucine, isoleucine and valine metabolism p. 2
Lysine metabolism
Mitochondrial ketone bodies biosynthesis and metabolism
Mitochondrial long chain fatty acid beta-oxidation
Mitochondrial unsaturated fatty acid beta-oxidation
Peroxisomal branched chain fatty acid oxidation
Phenylalanine metabolism
PPAR regulation of lipid metabolism2
Propionate metabolism p. I2
Propionate metabolism p. 22
Regulation of fatty acid synthesis: NLTP and EHHADH
Regulation of lipid metabolism by niacin and isoprenaline
Regulation of lipid metabolism via LXR, NF-Y, andSREBP2
Regulation of lipid metabolism via PPAR, RXR, and VDR2
Serotonin — melatonin biosynthesis and metabolism
TCA
Triacylglycerol metabolism p. 1
Tryptophan metabolism
5-40
-------
Table 5-6. (continued)
Biological Process
Transcription*
Pathways
Brcal as transcription regulator
Role of VDR in regulation of genes involved in osteoporosis
Transcription factor Tubby signaling pathways
""Statistically significant gene lists from SNR and REM methods were input into the GeneGo pathway
analysis program (www.genego.com). The Gene ontology process/pathway list was generated using a
cut-off ofp < 0.05 for each analysis. From those lists, the common pathway list was generated.
bPathways that are part of, or overlap with, the testosterone synthesis pathways are indicated by bold italics.
These pathways were identified by performing a PubMed literature search
(http://www.ncbi.nlm.nih.gov/sites/entrez?db=PubMed) for "testosterone" and the name of each pathway
(listed in the table).
°Entrez Gene indicates that Insl3 is the ligand for the LGR8 receptor, but the Insl3 pathway is not fully
defined
(http://www.ncbi.nlm.nih.gov/sites/entrez?Db=gene&Cmd=ShowDetailView&TermToSearch=114215&or
dinalpos=3&itool=EntrezSystem2.PEntrez.Gene.Gene_ResultsPanel.Gene_RVDocSum).
*Biological processes identified in Liu et al. (2005).
Functions shown to be related to the InslS pathway are G-protein-coupled receptor binding and hormone
activity. Processes identified are G-protein signaling, adenylate cyclase inhibiting pathway, gonad
development, in utem embryonic development, male gonad development, negative regulation of apoptosis,
negative regulation of cell proliferation, oocyte maturation, positive regulation of cAMP biosynthetic
process, and positive regulation of cell proliferation. While a number of G-protein pathways were
identified in this analysis, none are considered exclusive to InslS and are, therefore, not listed in bold
italics.
Cholesterol biosynthesis/metabolism and associated pathways underlie one of the MO As
of DBF. To determine a metric for statistical analysis protocols of toxicogenomic data, we chose
to compare the genes that are involved in the cholesterol biosynthesis/metabolism as identified
by the three independent analysis methods (described herein) and the published data set from Liu
et al. (2005) (see Table 5-7). These results show that there is a high degree of overlap in the
most biologically relevant pathway/process involved in DBF toxicity, even when different
statistical procedures are used for analysis of the same data set. These are in agreement with the
published literature, giving the approaches used in this exercise biological confidence.
By utilizing databases such as GeneGo, additional canonical pathways and biological
processes were identified that may play an important role in DBF male reproductive
developmental toxicity. Regulation of steroidogenesis requires multiple signaling pathways and
growth factors (Stocco et al., 2005). Signaling pathways, like the protein kinase C pathway,
arachidonic acid metabolism, growth factors, chloride ion, and the calcium messenger system are
5-41
-------
Table 5-7. Genes involved in cholesterol biosynthesis/metabolism that were
identified by both the REM and SNR analyses of Liu et al. (2005)
REM (GeneGo)
Cyp27al
CypSlal
Cyp7bl
Dhcr7
Hmgcr
Hmgcsl
Hsdllbl
HsdSbl
Idil
Sqle
Sc4mol
Soatl
SNR (GeneGo)
Acatl
CypSlal
Dhcr7
Dhcr24
Ebp
Fdftl
Fdps
Hmgcr
Hmgcsl
Idil
Mvd
Nsdhl
Sqle
Sc4mol
Tm7sf2
SNR (KEGG)
Acatl
Dhcr7
Ebp
Fdftl
Fdps
Hmgcr
Hmgcsl
Idil
Mvd
Sqle
capable of regulating/modulating steroid hormone biosynthesis. It is possible that some of the
pathways and processes identified by the two methods may play a role in the regulation of
steroidogenesis, a pathway that underlies one of the well-established MO As by DBF. Another
scenario could be that these pathways and processes have yet to be associated with DBF-induced
toxicity. The androstenedione and T biosynthesis and metabolism pathway was one of the
common pathways in the GeneGo analysis of the two different methods gene list (see
Figure 5-5).
5-42
-------
•2.4 Androstenediol
sulWe
&
62 ^
_ HSD17B8
i •=»• <
62 dehydnoepiandro
sterone sul fate
'
Homodimer
c
& ST2M
Homodimer
| »
2. 8 2.2
I
-3^2
1.14.14.1
dehydroepiandro
sterone
1.14.99.9
16 alpha-hydroxy-
dehydfo-
epiandrosterone
17-alpha-fiyd'axy-
cregnenolone
1.1.1.145/5.3.3.1
16 alpha-hydroxy-
androstenedone
CVP17
*
«
!
HSD
Conversion of pregnenolone and progesterone to
their 17- alpha-hydroxylated products and
subsequently to dehydroepiandrosterone (DHEA)
and androstenedione. Catalyzes both the 17-
alpha-hydroxylation and the 17j20-lyase reaction.
Involved in sexual development during fetal life
and at puberty.
Pregnenolone and
progesterone
biosynthesis and
metabolism
17-alpha-
hydroxy progesterone
Figure 5-5. Mapping the Liu et al. (2005) data set onto the canonical
androstenedione and testosterone (T) biosynthesis and metabolism pathway
in MetaCore (GeneGo). Key enzymes activated by DBF are identified by red
thermometers.
It has been reported in the literature (MAQC-I, see Chapter 2) that the results of
microarray experiments often depend on the data analysis protocol and the biological pathway-
analysis tools available to interpret the list of statistically significant genes. Dissimilar sets of
gene expression signatures with distinct biological contexts can be generated from the same raw
data by different data analysis protocols. Distinct biological contexts can also be generated from
5-43
-------
the same gene expression signatures by different biological pathway protocols. Therefore, it
becomes important to determine and understand the relationship between the gene expression
and pathway changes and a biological outcome of interest.
In order to do a thorough investigation, it is necessary to use many sources of gene and
pathway annotation. The intent of using multiple sources is to gain an enriched analysis. In
practice, analysis is carried out with the suite of tools available to the analyst. In this case, the
STAR Center primarily used KEGG (a resource rich in enzymatic and metabolic reactions but
weak in signaling pathways); whereas the EPA used Rosetta Resolver, GeneGo, and Ingenuity
Pathway Analysis, resources that are populated with signaling as well as metabolic pathways.
This exercise demonstrates that multiple approaches to microarray data analysis can yield
similar biologically relevant outcomes and some differences. The differences observed in the
results could be due to a number of factors including (1) the different data normalization
procedures used in the two separate analyses; and (2) different data interpretation tools such as
the software for pathway analyses. However, it cannot be ruled out that the differences may
reflect differences in biological significance (i.e., one approach is superior).
We performed a number of reanalyses of the Liu et al. (2005) data because the pathway
analysis presented in the article was not performed for risk assessment purposes. While the
authors of this and other microarray studies support two MO As for DBF, a reduction of fetal
testicular T via affects on steroidogenesis and cholesterol transport genes, not all pathways
associated with the differentially expressed genes were discussed in detail.
Two different bioinformatics tools to analyze the same data were utilized. Each analysis
used multiple statistical filters to parse the noise from the signal in the microarray data set and to
assess the quality of the data set. Ideally, for a high-quality study data set, there would be a
minimum of variance between similarly treated samples, and the variance would lie between the
control and treated sample data. PC A shows the quality of the Liu et al. (2005) data set to be of
moderate quality based on the observed variance among similarly treated data sets (control and
treated groups). One analysis utilized multiple proprietary software packages (GeneGo, Rosetta
Resolver). The rationale for looking at the effect of DBF on the pathway level, as opposed to a
cluster of genes, is that DBF is most likely affecting multiple pathways within a cellular
environment. This exercise allowed us to generate a list of affected common pathways between
the two methods, and in this way, provided more confidence about these pathways.
5-44
-------
The results of the new pathway analyses both corroborate the previously identified two
MO As for DBF male reproductive development toxicity, and provide putative novel pathways
affected by in utero DBF exposure that may play a role in DBF-mediated toxicity. The results of
the new pathway analyses provide hypotheses for MOA that could be tested in new experimental
studies. Future research could investigate the role of these pathways in DBF-induced toxicity.
In addition, a gene network was developed for DBF based on the Liu et al. (2005) data. The
GeneGo analysis corroborated prior findings for the role of the steroidogenesis pathway and
identified the modulation in Cypl? and^4r that are involved in the androgen biosynthetic
process. This is a new hypothesis that requires follow-up with new studies to confirm this
observation. Performing new analyses was useful for the purposes to further our understanding
of the DBF mechanism of action.
Analyzing any given data set multiple ways and arriving at the same conclusion provides
confidence in the analytical approach; however, there is no "gold standard" analytical method.
Applying stringent statistical filters in pathway analysis (e.g.,/? < 0.05, Benjamini Hochberg
multiple testing correction) can limit the number of genes that are identified. Interpretation of
the biology of the system using only a limited gene set is restrictive. It is important to remember
that the genes that do not pass the statistical stringency cut-off that may be crucial for
understanding the biology of the system, as statistical significance and biological significance are
not necessarily the same. Therefore, it becomes incumbent upon the researcher to analyze the
data in multiple ways in order to maximize the benefits of microarray data.
In summary, by identifying differentially expressed genes by two different approaches,
performing pathway analysis, and compiling a list of common pathways between the two
approaches, a list of corroborated pathways has been identified. The pathways (see Table 5-6)
and processes identified have some overlap with those presented in the Liu et al. (2005) article as
well as some differences. Comparisons of our results to those of Liu et al. (2005) are difficult
because they presented differentially expressed genes and their associated process, not pathways.
In Liu et al., 2005, oxidative stress and cytoskeleton processes were unique findings. Our results
identified cell adhesion, disease, immune response, hormone, and growth and differentiation
processes as unique findings. In addition, the reanalysis of the Liu et al. (2005) study identified
common and unique pathways (see Table 5-6) with the tabulation of affected pathways from the
published literature that we performed including all of the DBF gene expression studies (see
5-45
-------
Table 5-2). This exercise has generated hypotheses about mechanisms/pathways that could
underlie the unexplained testicular endpoints after in utero DBF exposure (see Chapter 4) that
need to be tested in additional studies.
5.6. CONCLUSIONS
In this chapter, evaluations of the published studies and a reanalysis of pathways from
one microarray study was performed. Nine toxicogenomic studies from the published literature
were evaluated for study comparability and study result consistency. This was done by utilizing
Venn diagrams and a visual method for looking at the consistency across all of the gene
expression studies (see Figure 5-2). These methods could be applied in a new assessment for a
chemical with genomic data.
The reanalysis of the Liu et al. (2005) data set provides some examples of methods for
identifying differentially expressed genes and performing pathway analysis using either
proprietary or publicly available methods and databases. In performing the reanalysis,
hypotheses were generated about possible pathways underlying some of the known and unknown
MO As for the testes outcomes observed after in utero DBF exposure.
5-46
-------
6. EXPLORATORY METHODS DEVELOPMENT FOR ANALYSIS OF GENOMIC
DATA FOR APPLICATION TO RISK ASSESSMENT
6.1. OBJECTIVES AND INTRODUCTION
The overall goal of this chapter is to describe exploratory methods developed for analyzing
and applying toxicogenomic data in risk assessment. The three objectives of the methods
development projects were to
1. Explore the development of new methods to analyze microarray data for application to
risk assessment.
The motivation was to develop methods for performing gene expression analyses of
microarray data for use in risk assessment. Microarray studies for basic research
purposes do not necessarily require as high a level of stringency as for risk assessment
purposes because the analyses are often performed to generate hypotheses (e.g., for
developing MOA hypotheses) that are subsequently tested in additional studies.
2. Utilize existing DBF genomic data to develop a temporal gene network model for use in
risk assessment.
We asked whether there are data to understand gene expression changes over time. By
modeling the gene and pathway interactions across the critical window of exposure to
DBF, it may be possible to understand the relationships among genes and pathways over
time, and possibly, to identify the initiating event(s) for the decreases in fetal testicular T
or Insl3 expression. Identifying the initiating event would be very useful to risk
assessment, as this would provide a biologically significant gene whose expression is
critical to the outcome.
3. Utilize genomic and other molecular data to address the Case Study Question: Do the
toxicogenomic data inform inter species differences in TD?
We utilized the available gene sequence data, protein sequence, and pathway cross-
species data to assess the rat-to-human conservation of the genes involved in the
steroidogenesis pathway that underlie the reduced fetal testicular T MOA for DBF.
The work to address the objectives of this chapter is the result of a collaborative effort
among scientists at the STAR Bioinformatics Center at UMDNJ and Rutgers, and the EPA. The
analyses were performed at Rutgers University.
The work presented in this chapter is highly technical and thus, is intended to be
beneficial to scientists with expertise in bioinformatics. The technical details of the analyses are
6-1
-------
provided in order that scientists could apply these methods to their work. Such an approach will
allow the risk assessor proficient in microarray analysis methodology an opportunity to apply
these methods. The last section of this chapter (Section 6.4) summarizes the findings for a
scientific audience that does not have a strong background in microarray analysis methods.
6.2. PATHWAY ANALYSIS AND GENE INTERACTIONS AFTER INUTERO DBF
EXPOSURE
6.2.1. Pathway Activity Approach
Usually, to identify significant biological pathways from transcriptional data, pathway
analysis is performed after determining the DEGs using a statistical filter. Two examples of this
approach are described in Chapter 5 (Section 5.5). An alternative approach is the use of
"pathway scoring" methods, which begin with projecting gene expression changes onto
pathways (Rahnenfuhrer et al., 2004; Moothaet al., 2003; Hanisch et al., 2002). The main
advantage of applying pathway scoring methods to microarray data is that changes can be
identified at the pathway level that may not be detected by first identifying individual DEGs.
Most of these methods calculate the average correlation between pairs of genes within pathways
(Rahnenfuhrer et al., 2004; Sohler et al., 2004; Hanisch et al., 2002; Zien et al., 2000). Another
pathway scoring method tests for association between gene expression and a phenotype (e.g.,
Gene Set Enrichment Analysis [GSEA]; Mootha et al., 2003). In GSEA, all genes are ranked
with respect to some measure that quantifies the gene expression associated with a phenotype
(i.e., differentiation between healthy vs. disease samples). Tomfohr et al. (2005) introduced a
pathway-based approach that is similar in spirit to GSEA. Their method translates the overall
gene expression levels within a pathway to a "pathway activity level," which is derived from
singular value decomposition (SVD), described below. Hence, pathway activity levels can be
used in the same kinds of applications as gene expression levels (Tomfohr et al., 2005).
Tomfohr et al. (2005) compared their pathway activity method to GSEA using expression data
from two different studies, one that studied Type 2 diabetes and one that studied the influence of
cigarette smoke on gene expression in airway epithelia. They found similar results to those
obtained using GSEA in the diabetes set, and further, improved results for identifying
differentially expressed pathways in the cigarette smoke data.
We applied a pathway activity level approach to DBF microarray data. Since pathway
activity levels are a reduced form of the overall gene expression matrix (represented by the
6-2
-------
largest deviation in the overall gene expressions within a pathway) Alter et al. (2000) and
Cangelosi (2007) raised the critical issue that pathway activity levels (represented by the largest
deviation in the overall gene expressions within a pathway) may be attributed to random
deviations in the data. Therefore, we use a significance analysis to distinguish the information
captured by pathway activity levels from random deviation.
6.2.1.1. Significance Analysis of Pathway Activity Levels
The procedure begins with mapping genes to the KEGG pathway database. The entire
gene set represented by the Liu et al. (2005) data set (i.e., using the Affymetrix RAE230 A and
B chips) maps to 199 pathways in the KEGG database with 4,772 associated genes.
Pathway activity formulation starts with SVD of the gene expression matrix of a given
pathway. SVD involves a mathematical procedure that transforms a number of possibly
correlated variables into a smaller number of uncorrelated variables. It mathematically
transforms the data to a new coordinate system such that the greatest variance by any projection
of the data lies on the first coordinate (called the eigenvector), the second greatest variance on
the second coordinate, and so on. Associated with each of these coordinate eigenvectors is a
weight term (called the eigenvalue) that represents the variance in the data. The eigenvalues are
normalized such that they express the fraction of the variance along their corresponding
eigenvector. In this study, SVD is used to calculate pathway activity levels for each
experimental condition where each pathway activity level represents the most significant gene
expression pattern within each pathway. The details of SVD analysis are as follows:
Using Eq. 6-1, let Sp(k,t) be the gene expression data associated with a given pathway,/*,
composed of & genes measured at t different conditions (time, treatment, dose, etc.), normalized
(i.e., to a mean of zero mean and unit standard deviation). The SVD ofSp(k,t) is given as
follows:
Zp(k,f) = Up(k,K)xSp(k,f)xVp(ttf (6-1)
Eq. 6-1 states that the columns of the matrix Up(k,k) are the orthonormal eigenvectors ofEp(k,t).
Sp(k,t) is a diagonal matrix containing the associated eigenvalues, and the columns of the matrix
Vp(t,t) are projections of the associated eigenvectors of Ep(k,t). As the elements ofSp(k,t) are
6-3
-------
sorted from the highest to the lowest, the first row of Vp(t,t) represents the most significant
pattern within a pathway across different samples. Hence, PALP is mathematically defined as the
first vector of the Vp(t,t) (given in Eq. 6-2 ).
PALp=Vp(n,Tf (6-2)
The fraction of the overall gene expression that is captured by PALP is evaluated through Eq. 6-3.
s (i,
,
fp= L (6-3)
An additional analysis is needed to evaluate whether PALP represents significant
information about the pathway. As a standard procedure for evaluating significance of
microarray data, random sampling is used. For each pathway, an equal number of gene
expression values are permutated 1,000 times. The/>-value is computed as the permutated^, that
exceeded the actual^, (p-value < 0.05). Next, the pathways are filtered based on the associated
/7-value of their fp value.
We illustrate the importance of the significance analysis ofPALp in Figure 6-1 using the
gene expression matrix for the tryptophan metabolism pathway. Panel A of Figure 6-1 depicts
both the fraction of the overall gene expression captured by each eigenvector,^,, and the average
fraction of the overall gene expression captured by each eigenvector of the randomized data. We
observe that thefp value captured by the PALP of the tryptophan metabolism pathway can be
retrieved with a randomly selected gene set and thus, the tryptophan metabolism pathway is not
significantly affected by DBF exposure. We applied a significance analysis of PALpio improve
the confidence of Tomfohr's pathway activity level formulation for further calculations.
6.2.1.2. Pathway Activity Analysis
The main goal of pathway analysis is to identify significantly affected pathways, based
on gene expression data, due to DBF exposure. For this purpose, as described above,
6-4
-------
1) 0.35
3
S. 0.3
A)
0.2
« 0.15
0.1
"5
0.05
] tract I on of total variation captured by the associated eigenvector
_ fraction of total variation captured by the associated eigenvector
retrlved from the permutated data
3 4
Number of eigenvector
C CCT TT
C CCT TT
CCCT T T
CCCT TT
CC CTTT
CCCT TT
Figure 6-1. An illustration of the adapted version of pathway activity level
analysis for the tryptophan metabolism pathway, a nonactive pathway for
DBF. In panel A, the boxes indicate the variability in the actual gene expression
data, associated with the tryptophan metabolism for each individual eigenvector.
For comparison, the solid line represents the fraction of data variability captured
by the corresponding eigenvectors when randomly generated data were used. No
apparent distinction between the actual data and randomly generated data was
identified, as quantified by the calculatedp-va\ue of 0.25. In panel B, the
projection of the gene expression on each eigenvector is depicted for each sample
of the control (C) and DBF-treated (T) groups. PALP is the first vector that
corresponds to the largest variation in the data.
overall gene expressions within a pathway are reduced to PALP. The differentiation between
PALP of different samples is denoted as pathway activity and is determined through a process
analogous to SNR analysis.
6-5
-------
If n\ samples are associated with vehicle treatment (control) and n^ samples with
chemical treatment (DBF), then the activity levels associated with treatment groups are given in
Eqs. 6-4 and 6-5, respectively.
(6_4)
(6-5)
Pathway activity is calculated using Eq. 6-6 where ju and a represent the mean and
standard deviation respectively.
PA,=
A high pathway activity represents a better differentiation between control and treated
pathway activity levels. The statistical significance of pathway activity is determined using the
randomization process. For each pathway, an equal number of genes within a given pathway are
randomly assigned and gene expression changes are generated (from the chip) 10,000 times. The
/7-value of the pathway activity is computed as the fraction of the randomized pathway activity
that exceeded the actual pathway activity. In this analysis, the pathways that have both
statistically significant (p-value < 0.05) pathway activity and pathway activity level are defined
as "active" pathways.
"Active" pathways are those for which the overall change in gene expression in a
pathway of treated samples compared to control samples was statistically significant. For
example, an active pathway could be one for which gene expression was downregulated or
turned off after DBF exposure. Alternatively, a pathway that is not identified as active would
still have gene expression occurring, but might not exhibit a significant difference in gene
expression following DBF exposure compared to the control samples. Thus, the term active does
not refer to gene expression from a particular pathway. The algorithm for selecting active
pathways using the pathway activity method is shown in Appendix B, Figure B-l.
6-6
-------
We identified 15 active pathways from querying the KEGG database (see Table 6-1).
The pathway activity method identified pathways such as biosynthesis of steroids (C21 Steroid
hormone metabolism pathways known to be biologically relevant to T levels) as well as other
pathways including butanoate metabolism, pyruvate metabolism, and biosynthesis of unsaturated
fatty acids (PPAR signaling pathway and fatty acid metabolism).
Table 6-1. The KEGG pathways ordered based on their />-value for pathway
activity3
Pathway name
Reductive carboxylate cycle (CO2 fixation)
Valine, leucine and isoleucine degradation
Biosynthesis of steroids
Citrate cycle (TCA cycle)
Glutathione metabolism
Tryptophan metabolismf
Pentose phosphate pathway
Glycolysis / Gluconeogenesis
Butanoate metabolism
Pyruvate metabolism
C21 Steroid hormone metabolism
Glyoxylate and dicarboxylate metabolism+
Biosynthesis of unsaturated fatty acids
Fatty acid metabolism
Nicotinate and nicotinamide metabolism
Propanoate metabolism
Cyanoamino acid metabolism+
PPAR signaling pathway
p-value of
PAb
0.001
O.001
0.001
0.002
0.002
0.002
0.002
0.003
0.004
0.004
0.006
0.012
0.012
0.020
0.028
0.030
0.032
0.042
p-value of
PALC
0.002
O.001
0.001
O.001
0.006
0.250
0.001
O.001
0.006
O.001
0.048
0.480
0.048
0.030
0.068
0.018
0.074
O.001
aPathway activity quantifies the difference between control and DBF-treated samples from Liu et al. (2005)
(see Eq. 6-6). PAL is the pathway activity level for both the control and treated samples (see Eq. 6-2).
The statistical significance of PA and PAL values are evaluated through a randomization procedure. The
p-value of PAL is used as an additional filtering process to eliminate potentially nonactive pathways.
bThe p-value of the PA is computed as the fraction of the randomized PA that exceeded the actual PA. In
the event that the PA of the randomly generated matrices exceeds the actual PA by more than 5 % of the
randomization process, then the actual PA is attributed to a random variable (p-value < 0.05).
The p-value of PAL quantifies the significance of fraction of the overall gene expression captured by
PAL. It is computed as the fraction of the randomized fp exceeding the actual^,. In the event that the PA
of the randomly generated matrices exceed the actual PA by more than 5 % of the randomization process,
then the actual PA is attributed to a random variable (p-value < 0.05).
PA, pathway activity; PAL, pathway activity level.
6-7
-------
To explore the biological significance of the active pathways, a metabolic pathway
network of the active pathways illustrating their connections via metabolites was built
(Figure 6-2). This process includes the integration of the statistical outcome of the pathway
activity analysis and the relationships among these pathways by querying the KEGG database.
After DBF in utero exposure, the pathways related to cholesterol biosynthesis exhibit more
significant changes in their gene expression compared to the rest of the active pathways. This
finding is consistent with the hypothesis that an early decrease in T level might be due to
cholesterol unavailability (Thompson et al., 2005).
PA=Difference between
control and DBF treated
samples
Valine, Leucine and Isoleucine Degradation
Biosynthesis of Steroids
Steroid Hormone Metabolism
Pyruvate Metabolism
Fatty Acid Metabolism
Proponate Metabolism
Figure 6-2. Metabolic pathway network for DBF (Liu et al., 2005 data) using
the pathway activity method and the KEGG database. Active pathways
connected to each other via metabolites are ordered from the most active pathway
(top of the figure) to the less active pathways (bottom of the figure). The
connections between the active pathways were retrieved from KEGG (Kanehisa
and Goto, 2000).
6-8
-------
We explored the contribution of DEGs to the pathway activity for a given pathway
(Figure 6-3 A, B, C, and D). The pathway activity of each pathway is calculated by adding
genes one-by-one starting with the gene with the highest SNR and adding genes sequentially in
the order of their SNR until all genes in the pathway have been added. Figure 6-3 A and B
illustrate examples of active pathways, whereas Figure 6-3 C and D are examples of pathways
that were not identified as active in our analysis. For pathways that were identified as active or
not active, the cumulative pathway activity value undergoes a decrease as genes of lower SNRs
are added. Yet for the active pathways, the cumulative pathway activity remains high enough to
be statistically significant. For pathways identified as not active, the cumulative pathway
activity reaches a low level when all of the genes are added. Accordingly, their pathway activity
value is not statistically significant. The four pathways are composed of a similar number of
genes; therefore, the number of genes in the pathway is not an issue in this comparison. We
hypothesize that there is a subset of genes that maintain the pathway activity value high enough
within active pathways, even when all genes are added. The cumulative behavior of this subset
enables us to differentiate the active and nonactive pathways. Differentially expressed genes in
active pathways are defined as "informative genes" (see Table B-l). We identified a relatively
small number of genes as informative, and these may represent genes that DBF has most greatly
affected.
One of our goals was to utilize existing DBF genomic data to develop a gene network
model useful to risk assessment. Gene network models illustrate interactions between genes and
their products (e.g., mRNA, proteins). We used IPA software to construct a gene network model
after DBF in utero exposure. IPA adds nodes (i.e., genes) to the input gene list (i.e., informative
genes) and then, builds edges (i.e., relations) based on the literature. The interactions among the
informative genes from the Liu et al. (2005) data were retrieved using IPA, and the resulting
preliminary gene network model is shown in Figure 6-4.
6.2.2. Developing a Temporal Gene Network Model
The Thompson et al. (2005) study was selected to develop a temporal gene network
because it was the only available time-course study. The study had the advantages of using the
rat Affymetrix chip, which has -30,000 gene transcripts represented, and availability of the data
(i.e., kindly provided by Dr. Kevin Gaido). In the study, animals were exposed to DBF for 0.5,
6-9
-------
A) Biosynthesis of Steroids
B) Butanoate Metabolism
15
10
C) Pentose and Glucuronate
Interconversions
m
5
0
lllll
)
'
nun
j i
ih
0 1
5 2
0 2
5 3
D) Ether Lipid Metabolism
15
10
0 10
Number of Genes, Ranked by SNR
30 35
Figure 6-3. The relationship between differential expression of individual genes and
pathway activity using the Liu et al. (2005) DBF data. The pathway activity of a given
pathway is first evaluated using the gene that has the highest SNR. Subsequently, the
genes are added in the order of their SNR, from highest to lowest. Pathways identified as
active for DBF, such as biosynthethis of steroids (A) and butanoate metabolism (B),
maintain high pathway activity values even when all genes in the pathway are added.
Alternatively, pathways not identified as active for DBF such as pentose and glucuronate
interconversions (C) and ether lipid metabolism (D), exhibit a decrease in pathway
activity as the less discriminating genes (i.e., those with a lower SNR value) are added.
6-10
-------
Valine, Leucine, and Isoleucine Degradation
Glycolysis/Gluconegenesis
Citrate Cycle
C21-Steroid Hormone Metabolism
Xenobiotic Metabolism Signaling
Figure 6-4. A gene network for DBF data of Liu et al. (2005) generated using Ingenuity Pathway Analysis (IPA).
The figure illustrates the interactions among informative genes following in utero DBF exposure in the rat testis from
Liu et al. (2005). Interactions among genes (shown in Appendix B, Table B-l) are derived from the Ingenuity
database. Genes or gene products are represented as nodes. Diamonds, enzymes; Horizontal ovals, transcription
regulators; Squares, cytokines; Rectangles, nuclear receptors. Solid lines represent direct relationships between nodes
(i.e., molecules that make physical contact with each other, such as binding or phosphorylation). Dashed lines
represent indirect interactions (i.e., not requiring physical contact between the two molecules, such as signaling events).
CP, Canonical pathway.
-------
1, 2, 3, 6, 12, 18, or 24 hours before sacrifice on GD 19. The limitations of the Thompson et al.
(2005) study include (1) the dosing was initiated on GD 18, late in the critical window, and (2)
the shortest duration exposure (30 minutes) began at the latest developmental time (i.e., GD 19);
thus, developmental stage and duration of exposure do not coincide (see Chapter 5). Given this
caveat, we utilized the available to test algorithms to build a prototype of a temporal gene
network model.
We used the pathway activity level method described earlier to identify biologically
active pathways at each time point. We evaluated the informative genes at each time point and
the resulting preliminary temporal gene network, based on the Thompson et al. (2005) data, is
shown in Figure 6-5. The analysis showed a preponderance of signaling pathways such as
JAK/STAT, PPAR, and MAPK perturbed at the earlier exposure durations. After the longest
DBF exposures (18 hours), the metabolic pathways, including amino acid metabolism, lipid
metabolism, and carbohydrate metabolism, were affected. Thompson et al. (2005) hypothesized
that the decrease in T level after a short duration of DBF exposure might be due to cholesterol
unavailability and their findings support this hypothesis. To have a complete understanding of
the temporal sequence of gene expression and pathway affect events after in utero DBF
exposure, data from an exposure-duration series across the entire critical window of exposure are
needed.
6.3. EXPLORATORY METHODS: MEASURES OF INTERSPECIES (RAT-TO-
HUMAN) DIFFERENCES IN TOXICODYNAMICS
The goal of this section is to address whether genomic and mechanistic data could inform
the interspecies (rat-to-human) differences TD for one of the DBF MO As reduced fetal testicular
T (one of the DBF case-study questions). Although progress has been made in understanding the
MO As of chemical toxicants, it is important to evaluate the mechanistic relevance of these
MO As to humans. The genomic data set for DBF does not include human genomic data of any
type, including studies from in vitro cell lines. Even if such data were available, extrapolation of
in vivo data (rat genomic) to in vitro data (human genomic) may confound the ability to generate
accurate interspecies comparison. In the absence of DBF genomic data in human cell lines, we
considered genetic sequence data and other data from rats and humans for making species
comparisons. It is significant that the role of T in male reproductive development during sexual
6-12
-------
steroid Hormone Metabolism |
Glutatmone Metabolism
Added nodes by Ingenuity
Exposure Time
-18 hour
- 6 Hour
-3 Hour
-1 Hour
Figure 6-5. A temporal gene network model created by IPA from the informative gene list based on time-course
data after in utero DBF exposure (Thompson et al., 2005). The informative genes were evaluated at each time point
and mapped onto a global molecular network using the Ingenuity Pathways Knowledge Base. Diamonds, enzymes;
Horizontal ovals, transcription regulators; Squares, cytokines; Rectangles, nuclear receptors. Solid lines represent
direct relationships (also called edges) between nodes (i.e., molecules that make physical contact with each other, such
as binding or phosphorylation). Dashed lines represent indirect interactions (i.e., not requiring physical contact
between the two molecules, such as signaling events). CP, Canonical pathway. Low (green), downregulated
expression with respect to control. High (red), upregulated expression with respect to control.
-------
differentiation is conserved among vertebrates, thus providing a measure of human relevance of
the reduced fetal testicular T observed in the rat after in utero DBF exposure.
Phylogenetic analysis, the reconstruction of evolutionary relations, is based on shared,
derived characters presumed to have a common origin. Taxonomy of organisms is one method
for determining species relatedness. However, since DBF perturbs the activity of the
steroidogenesis pathway and leads to the decreased T MOA for DBF, we were interested in
developing metrics by comparing this pathway between the rat (for which we have data) and
human. Previous phylogenetic analyses of individual pathways have included assessing: the
number of common enzymes and their conservation across different species (Forst, 2002; Forst
and Schulten, 1999); the topology of the underlying enzyme-enzyme relational graphs including
their sequence conservation (Heymans and Singh, 2003); and the intersection of compounds,
reactions, and enzymes across species (Clemente et al., 2005).
We reconstructed the phylogenetic relationships for biosynthesis of steroids among eight
species based on enzyme presence (Forst and Schulten, 1999; see Figure 6-6). The enzyme
presence method is based on information available in the KEGG database about the presence of
an enzyme (defined as catalyzing a specific reaction) in the pathway for a given species. As a
result, a pathway topology can be represented and compared across species. In this
representation of pathways, a vector containing binary information (where "1" is for presence,
"0" is for absence of the enzyme) is created for a given pathway. Then, the similarity between
pathways for two different species is defined as the ratio of the number of common enzymes to
the number of unique enzymes. The results suggest that the steroidogenesis pathway is quite
similar between rat and human. Further, we found that the species differences based on enzyme
presence were different from those based on the NCBI taxonomy (Sayers et al., 2008) of the
organisms, which is not surprising based on previous findings (Searls, 2003). In order to utilize
more complete information about a pathway, cross-species pathway comparisons should include
other biologically relevant information such as gene regulatory information and pathway
interactions.
Sequencing of the human, mouse, and rat genomes and their comparison has increased
our understanding of cross-species similarities and differences in genes and proteins. Co-
expressed genes across multiple species are most likely to have a conserved function. The rat
genome project reported that almost all human genes known to be associated with disease have
6-14
-------
A) Enzyme Presence for the Biosynthesis of Steroids Pathway
B) Phylogenetic and Taxonomic Knowledge
dre
eco
see
eel
ath
rno
mmu
hsa
Figure 6-6. The phylogenetic relations among eight organisms based on enzyme presence, for the biosynthesis of
steroids pathway, and based on information available on the NCBI taxonomy website (Sayers et al., 2008). Panel
A shows the results of evaluating the phylogenetic relations for the biosynthesis of steroids pathway, based on enzyme
presence (KEGG database), among eight model species (hsa, Homo sapiens; mmu, Mouse; rno, Rat; dre, Zebra fish;
ath, Arabidopsis; eel, C. elegans, see, Yeast; eco, E. coif). Panel B shows the phylogenetic relations among the same
eight organisms based on taxonomic and phylogenetic information retrieved from the NCBI taxonomy database
(http://www.ncbi.nlm.nih.gov/Taxonomy/taxonomyhome.html/index.cgi).
-------
orthologous genes in the rat genome, and that the human, mouse, and rat genomes are
approximately 90% homologous (Gibbs et al., 2004). While the function of certain genes and
their involvement in disease might not be conserved across species, the function of a pathway is
likely to be more highly conserved among species that perform similar functions (Fang et al.,
2005). Thus, cross-species pathway conservation metrics may be more useful.
Similarity among species can be investigated by phylogenomics analysis that involves a
comparison of genes and gene products across a number of species, characterizing homologues
and seeking further insights about evolutionary relationships. Analyzing the similarities between
phylogenetic gene trees and their associated protein trees can reveal additional information. For
example, a reconstruction of the CYP2A family of cytochrome P450 enzymes indicates that the
rat liver isoform (CYP2A1) has diverged significantly from the human (CYP2A6) and mouse
(CYP2A4) enzymes, having a distinct branch of the tree rooted outside the rest of the family
(Searls, 2003). This considerable deviation is associated with a well-known functional shift that
the rat enzyme causes the coumarin to be metabolized to a hepatotoxic epoxide, whereas the
human and mouse enzymes act on the same substrate by way of a more harmless hydroxylation.
The same principles can be extended to amino acid sequence comparisons for the genes
that make up a pathway. Utilizing the predicted amino acid sequence information for genes in
the steroidogenesis pathway from rats and humans, we evaluated the similarity among this set of
genes. Preliminary results suggest that proteins involved in the biosynthesis of steroids are
highly conserved across rats and humans, with the average sequence similarity of enzymes
between human and rat being -87% as presented in Table 6-2. However, it is difficult to
unequivocally determine a "high" versus "low" degree of conservation for the genes in this
pathway—especially in light of the fact that events important to the effect of DBF on
steroidogenesis are not well-understood. For example, initiating event after DBF exposure is not
known. Additionally, there are likely differences between identifying a gene that is statistically
highly conserved versus understanding whether or not the biologically meaningful regions of the
predicted protein sequence, active sites, are conserved. However, endocrinological,
developmental, and genetic studies in many vertebrate species indicate that the role of androgens
is highly conserved across vertebrates, as androgens are critical for sexual differentiation in the
male. Thus, taken together, this information suggests a high conservation of steroidogenesis and
androgen synthesis in rats and humans.
6-16
-------
Table 6-2. The amino acid sequence similarity of the enzymes in the steroidogenesis pathway between rat and
human.
Gene
symbol
Dhcr7
Ml
Fdps
Fdftl
Hmgcr
Mvd
Sqle
Ebp
Lss
ScSd
Mvk
Cyp27bl
Nqol
Vkorcl
Entrez gene
ID
64191
89784
83791
29580
25675
81726
29230
117278
81681
114100
81727
114700
24314
309004
mRNA and protein IDs
NM_022389.2^NP_071784.1
NM_053539.1^NP_445991.1
NM_03 1840. 1 ^NP_1 14028. 1
NM_019238.2^NP_0621 1 1.1
NM_013134.2^NP_037266.2
NM_03 1062. 1 -^NP_1 12324. 1
NM_017136.1^NP_058832.1
NM_057 137. 1 ^NP_476478. 1
NM_03 1049. 1^NP_1 123 11.1
NM_053642.2^NP_446094. 1
NM_03 1063 . 1 ^NP_1 12325. 1
NM_053763.1^NP_446215.1
NM_017000.2^NP_058696.2
NM_203335.2^NP_976080.1
Human
homolog IDs
Q9UBM7
AF003835
M34477
AAP36671
AAH33692
AAP36301
NP_003120
NP_002331
NP_002331
NP_008849
BAD92959
NP_000776
NP_000894
AAQ 13668
Average similarity scores
Identities3
412/475 (86%)
196/227 (86%)
301/353 (85%)
356/413 (86%)
738/890 (82%)
338/398 (84%)
481/574 (83%)
618/732 (84%)
618/732 (84%)
246/299 (82%)
323/393 (82%)
413/508 (81%)
234/274 (85%)
83/94 (88%)
84%
Positives'"
443/475 (93%)
215/227 (94%)
326/353 (92%)
393/413 (95%)
768/890 (86%)
357/398 (89%)
528/574 (91%)
673/732 (91%)
673/732 (91%)
275/299 (91%)
355/393 (90%)
453/508 (89%)
250/274 (91%)
88/94 (93%)
94.14%
Gapsc
4/475 (0%)
0/227 (0%)
0/353 (0%)
0/413 (0%)
58/890 (6%)
1/398 (0%)
1/574 (0%)
1/732 (0%)
1/732 (0%)
0/299 (0%)
0/393 (0%)
7/508 (1%)
0/274 (0%)
0/94 (0%)
-------
Table 6-2. (continued)
Identities, The number and fraction of total residues in the HSP which are identical.
Positives, The number and fraction of residues for which the alignment scores have positive values.
°Gap, A space introduced into an alignment to compensate for insertions and deletions in one sequence relative to another. To prevent the accumulation
of too many gaps in an alignment, introduction of a gap causes the deduction of a fixed amount (the gap score) from the alignment score. Extension of
the gap to encompass additional nucleotides or amino acid is also penalized in the scoring of an alignment.
The HSP (high-scoring segment pair) is the fundamental unit of BLAST algorithm output. Alignment: The process of lining up two or more sequences
to achieve maximal levels of identity (and conservation, in the case of amino acid sequences) for the purpose of assessing the degree of similarity and
the possibility of homology.
Source: http://searchlauncher.bcm.tmc.edu/help/BLASToutput.html#anchorl4684156.
oo
-------
The same principles can be extended from amino acid sequence comparisons to
structures, pathways, and expression patterns.
6.4. CONCLUSIONS
The exploratory projects presented in this chapter include efforts to develop methods for
analyzing genomic data for use in risk assessment and examples of genomic data analyses
available to the risk assessor with expertise in bioinformatics. These methods include pathway
level analysis (including the newly described pathway activity method), gene network analysis,
and tools to assess cross-species similarities in pathways. A summary for a less technical reader
is presented below, grouped by the three objectives for the work.
1. Explore the development of new methods for pathway analysis ofmicroarray data for
application to risk assessment.
Quality-control requirements for microarray study analysis for use in risk assessment are
distinct from basic research. In traditional pathway level analysis, differentially
expressed genes are first identified and then mapped to their respective pathways.
Depending on the number of genes that map to a given pathway, the role of the pathway
can be over- or underestimated. To overcome this problem, we used the pathway activity
method. This method scores pathways based on the expression level of all genes in a
given pathway.
The pathway activity analysis identified valine, leucine, isoleucine (VL1) degradation,
sterol biosynthesis, citrate cycle, and fatty acid metabolism as the most active pathways
following DBF exposure. These findings support the hypothesis of Thompson et al.
(2005), that an early decrease in T levels may be a result of cholesterol unavailability.
However, for this approach to be useful, knowledge of tissue-specific pathways is
required. For example, even though bile acid biosynthesis does not take place in the
testis, a pathway related to bile acid biosynthesis was identified as statistically significant
in this analysis. This method shows promise for use in risk assessment.
2. Utilize existing DBF genomic data to develop a gene network model for use in risk
assessment.
Determining a sequence of gene expression changes and pathway level effects over time
can be very useful for understanding the temporal sequence of critical biological events
perturbed after chemical exposure, and thus, useful to a risk assessment. We developed a
method for developing a gene network model for DBF based on the available data. The
availability of time-course data (Thompson et al., 2005) enabled our group to model the
series of events that occurred between exposure to DBF and the onset of toxic
reproductive outcomes. However, given the limitations of the Thompson et al. (2005)
study design, we did not draw conclusions about genes and pathways affected over time
6-19
-------
for DBF. Instead, the Thompson et al. (2005) data was used to build a prototype of a
temporal gene network model and thus, the exercise allowed us to develop methods for
analyzing time-course data.
3. Utilize genomic and other molecular data to address the Case Study Question: Do the
toxicogenomic data inform inter species differences in TD?
Extrapolation from animal-to-human data is critical for establishing human relevance of
MOA(s) in risk assessment. Co-expressed genes across multiple species could have a
conserved function. The human, mouse, and rat genomes have been reported to be 90%
homologous (Gibbs et al., 2004). However, because it is not certain whether the function
of a specific gene is conserved across species, conservation of pathways across species
can be one important factor in establishing cross species concordance of one or more
MO As. In addition, a common critical role of androgens in both rodent and human male
development of reproductive organs has been well-established.
Using the available DNA, sequence, and protein similarity data for the steroidogenesis
pathway, we used three different methods to assess rat-to-human conservation as metrics
that may inform the interspecies differences in TD for one MO A, the reduced fetal
testicular T. The pathways for the biosynthesis of steroids have similarity between
human and rat. Comparing the predicted amino acid sequences for the steroidogenesis
pathway genes, we found that the average sequence similarity between rat and human is
-87%, and the average promoter region similarity of genes is 52%. Some of the
challenges in using similarity scores to estimate the cross-species relevance of a MOA
are described (see Section 6.3).
In summary, the preliminary analytical efforts described in this chapter address and raise
a number of issues about the best approaches for analyzing microarray and other genomic data
for risk assessment purposes. Traditional pathway analysis methods, while useful, also restrict
the incorporation of all genes for determining relevant pathways that are affected by DBF. There
is a substantial amount of background noise generated in a typical microarray experiment (i.e.,
gene expression variability even among the controls; see Smith, 2001). For use in risk
assessment, it is important to be able to identify and separate the signal from the noise.
Innovative approaches, such as the pathway activity method described in this chapter, may
provide more confidence when evaluating microarray data for use in risk assessment. These
efforts reveal some of the promises and challenges of analyzing and interpreting genomic data
for application to risk assessment.
6-20
-------
7. CONCLUSIONS
This chapter describes the general approach for systematically evaluating genomic data
for risk assessment. This general approach is a result of refining the DBF case-study approach
(see Figure 3-1). In addition, conclusions from the DBF case study, recommendations, research
needs, and future considerations for applying genomic data to risk assessment are described.
7.1. APPROACH FOR EVALUATING TOXICOGENOMIC DATA IN CHEMICAL
ASSESSMENTS
There were two goals of this project (see Chapter 2):
• Develop a systematic approach that allows the risk assessor to utilize the available
toxicogenomic data in chemical-specific health risk assessments performed.
• Perform a case study to illustrate the approach.
The first goal was to develop an approach for evaluating toxicogenomic data in future
assessments. In the DBF case study, we had the benefit of the 2006 external peer-review draft
IRIS Tox Review of DBF, including data summaries and gaps. Additionally, DBF has a more
extensive toxicological and toxicogenomic database than most chemicals. The DBF published
literature and the draft Tox Review provided a focus to the case study on one set of endpoints
(the male reproductive developmental endpoints), that occur in the lower dose range. The case-
study approach (see Figure 3-1) needed refinement because the case-study chemical and process
had some differences from that of a new assessment. A generalized approach (Figure 7-1) was
developed for use in future chemical assessments.
The steps of the approach are
• STEP 1: Compile the available epidemiologic, animal toxicology, toxicogenomic, and
other studies.
7-1
-------
STEP 1: Compile Data Sets for Assessment
STEP 2: Consider Quantitative and Qualitative
Aspects that Genomic Data Set May Address
STEP 3: Identify Questions to Direct the Evaluation
Do the Genomic Data Inform
• Toxicokinetics • Intraspecies Differences
• Hazard • Interspecies Differences
• Toxicodynamics • Dose-Response
• Exposure * Other Data-Dependent Issues
STEP 4:
Toxicity
Data Set
Evaluation
phenotypic
anchoring
MO A/pathways
study
comparability
STEP 5:
Genomic
Data Set
Evaluation
STEP 6: Results of Evaluation
• New Pathways Identified
• Other Results
STEP 7: Conclusions in Assessment
• MOA and Other Sections
• Data Gaps
• Research Needs
Figure 7-1. Approach for evaluating and incorporating genomic data into
future chemical assessments. "Toxicity Data Set Evaluation" may include
evaluation of animal toxicity data and/or human outcome data, depending on the
available data for the chemical.
7-2
-------
• STEP 2: Consider the quantitative and qualitative aspects of the risk assessment that
these data may address.
A thorough and systematic consideration of the types of information, in light of the
available genomic data, will identify the potential utility of the genomic data and whether
these data can be used quantitatively or qualitatively (see Section 3.2). The genomic data
set is considered in light of whether these data could inform any risk assessment
components (e.g., dose-response) and information (e.g., MOA information, interspecies
TK differences) useful to risk assessment. The type of information that these data will
provide to a risk assessment depends in part on the type of the available genomic studies
(e.g., species, organ, design, and method). This step helps focus the genomic data
evaluation and ensure that an important application is not overlooked.
• STEP 3: Formulate questions to direct the toxicogenomic data set evaluation.
Questions are formulated that can direct and focus the genomic data evaluation, and thus,
improve efficiency. This step is similar to a scoping exercise performed in ecological
and cumulative risk assessment. Some examples of questions considered in the DBF case
study were: Do the data inform the MO As for multiple outcomes (e.g., male and female
reproductive outcomes)? Do the data inform dose-response? For example, if microarray
data are available, then one of the questions will likely include whether the genomic data
can inform the mechanisms and/or MO As for the chemical as microarray studies
typically inform the mechanism of action of a chemical. The DBF case study describes
some examples and considerations for determining the risk assessment components that
may be informed by a particular genomic data set (see Section 3.3).
• STEPS 4 and 5: Evaluate the toxicity and/or human outcome and genomic data sets.
The approach includes an integrated assessment of the toxicogenomic and toxicity data
set to relate the affected endpoints (identified in the toxicity data set evaluation) to the
pathways (identified in the toxicogenomic data set evaluation) as a method for
Phenotypic Anchoring
Determining the level of support for phenotypic anchoring of genomic changes to
in vivo outcomes is critical for appropriate interpretation of genomic data for use
in risk assessment. In particular, determining whether gene expression changes
are associated with or in the causal pathway for an outcome of interest.
Informing the Mechanisms of Action/MOAs
Depending on the type of assessment performed, risk assessors may want to
utilize aspects of the approach defined herein along with the MOA Framework in
the EPA Cancer Guidelines (U.S. EPA, 2005a) and/or other risk assessment
decision-logic frameworks for establishing MO As.
7-3
-------
Study Comparability
Another principle of the approach is comparing toxicity and toxicogenomic data
study designs in order to identify a set of comparable studies. It is important to
compare the study designs among studies. Study design aspects include the time
of exposure (in light of critical windows), dose, species, strain, and time of
assessment. As a result of assessing study comparability for a given data set, one
can select studies for the best comparisons across the outcome and toxicogenomic
genomic data sets. For example, in the DBF case study, all toxicogenomic studies
were performed in the rat, and, in most cases, the testis. Therefore, the genomic
data set was compared with the rat toxicity data and focused on effects in the
testis. Broadening beyond the DBF example, the available toxicogenomic data
are best considered in light of the toxicity or epidemiologic study data that share
study design similarities with the toxicogenomic data. For example, if
toxicogenomic data from human tissue or cells are available, then these data are
best considered with the human epidemiologic outcome data for the chemical.
However, even in the absence of comparable data in the same species, the
genomic data may still be used, but with less confidence. See Chapters 4 and 5
for further details of the DBF case-study toxicity and toxicogenomic data set
evaluations.
Chapter 5 includes a number of simple methods for assessing the consistency
across the toxicogenomic studies. Venn diagrams have been used for illustrating
the similarities and differences of DEG findings across genomic studies (see
Figure 5-1). Figure 5-2 provides an excellent example of another method for
assessing the consistency of findings across all types of gene expression data.
New Analyses
New analyses of toxicogenomic data may be valuable for the assessment
depending on the questions asked and the nature of the analyses presented in the
published studies. However, new analyses of the original data are not always
needed. For instance, reanalysis may not be needed when available published
data have been analyzed for application to risk assessment questions of interest.
See Section 5.5 for more details of the new case study analyses methods and
results, and Chapter 6 for exploratory methods development.
• STEP 6: Describe results of evaluations and analyses to answer the questions posed in
Step 3.
• STEP 7: Summarize the conclusions of the evaluation in the assessment.
7.2. DBF CASE-STUDY FINDINGS
The second goal of the project was to develop a case study. The case-study findings are
summarized below and the details of the case-study evaluation and analyses are presented in
7-4
-------
Chapters 4-6 (with supplemental material in Appendices A and B). Three advantages to using
DBF as the case-study chemical are as follows:
1. The temporal aspects (e.g., time of dosing and time of evaluation) could be considered
because a number of well-designed studies exist.
2. A causal connection (i.e., a high degree of phenotypic anchoring) between gene
expression changes for some of the steroidogenesis pathway genes with a number of the
male reproductive developmental effects has been well-established.
3. Two well-established MO As for DBF have been defined at the molecular level. A
number of endpoints resulting from in utero DBF exposure have MO As that have not
been identified or established, thus allowing for a query of the genomic data for possible
additional MO As.
7.2.1. MOA Case Study Question: Do the DBF Genomic Data Inform Mechanism(s) of
Action and MOA(s)?
In the DBF case study, we found that toxicogenomic data did inform the TD steps of the
mechanisms of action and MO As. The available genomic and other gene expression data,
hormone measurement data, and toxicity data for DBF were instrumental in establishing two of
its MO As: (1) a decrease in fetal testicular T, and (2) a decrease in Insl3 expression. A decrease
in fetal testicular T is the MOA responsible for a number of the male reproductive developmental
effects in the rat. The genomic and other gene expression data identified changes in genes
involved in steroidogenesis and cholesterol transport, providing evidence for the underlying basis
for the observed decrease in fetal testicular T. A decrease in Insl3 expression is one of the two
MO As responsible for the undescended testis effect, and this MOA is well-established from the
results of RT-PCR and in vivo toxicology studies. RT-PCR studies identified reduced Insl3
expression after in utero DBF exposure (Wilson et al., 2004) as an MOA for agenesis or
abnormalities in the gubernaculum, effects that are not seen after exposure to anti-androgens
(i.e., chemicals that affect T synthesis or activity). These results provided support for the Insl3
MOA as a second well-established MOA for the male reproductive developmental effects of
DBF.
The rodent reproductive developmental toxicity data set is robust, having a high quantity
and relatively high quality of studies. Additionally, there are a number of rodent toxicity studies
that used similar study designs (e.g., dose, species, strain, timing of exposure) as some of the
7-5
-------
toxicogenomic studies. This aspect of the DBF data set is exceptional for the case study,
allowing for the establishment of the relationship between dose, pathways, and outcomes. We
evaluated the rodent reproductive toxicity data set for low incidence and low-dose findings but
due to data limitations (see Chapter 4), no new findings were identified. We also evaluated the
male reproductive developmental toxicity data set for effects that currently do not have a well-
established MOA (see Chapter 4). The testes outcomes were the focus of the case study because
the DBF toxicogenomic studies were all performed on testicular tissue. Five effects in the testes
effects associated with DBF exposure that do not have well-defined MO As were identified in
this evaluation.
The toxicogenomic and other gene expression studies, including nine published RT-PCR
and microarray studies in the rat after in utero DBF exposure (Plummer et al., 2007; Bowman et
al., 2005; Liu et al., 2005; Thompson et al., 2005; Lehmann et al., 2004; Thompson et al., 2004;
Wilson et al., 2004; Barlow et al., 2003; Shultz et al., 2001), were evaluated. The review of the
toxicogenomic data set focused on an evaluation of the consistency of findings from the
published studies, both across microarray studies and all gene expression data, and on whether
any additional pathways may illuminate the unexplained endpoints. The evaluation of the
published literature found that the gene level findings from the DBF genomic studies (i.e.,
microarray, RT-PCR, and protein expression) were highly consistent in both the identification of
DEGs and their direction of effect.
New analyses of the Liu et al. (2005) microarray study were performed because we were
interested in performing a complete pathway analysis of these data (which had not been the
purpose of the published study). These evaluations (see Chapter 5) indicate that there are a
number of pathways affected after in utero DBF exposure; some of these pathways are related to
new MO As that are distinct from the reduced fetal testicular T or the Insl3 signaling MO As. The
Liu et al. (2005) DBF data set was reanalyzed using two different methods, the SNR and REM,
both using a statistical cut-off of p < 0.05. Each method identified the steroidogenesis and
cholesterol transport pathways, corroborating prior study conclusions. Each analysis also
identified putative new pathways and processes that are not associated with either Insl3 or
steroidogenesis pathways; some were similar across analytical methods and some were different.
The common pathways identified between the two methods (see Table 6-3) fall into eight
processes (characterized by Ingenuity): cell signaling, growth and differentiation, metabolism,
7-6
-------
transcription, immune response, cell adhesion, hormones, and disease. Among these, 54 putative
new pathways that are not related to the two known MO As, reduced T or Insl3 expression, were
identified. Further, a subset of pathways (e.g., WNT signaling and cytoskeleton remodeling) was
identified in our analysis that had not previously been identified in the published literature for
DBF. One or more of these putative new pathways may be related to the toxicity endpoints
without identified MO As in the rat testes, but additional hypothesis testing studies are needed.
Evaluating the genomic and toxicity data sets together provided information on potential,
heretofore unexplored, MO As.
There are a number of possible reasons for the differences in findings between our
reanalysis and the published analysis of the Liu et al. (2005) data. These include but are not
limited to
• The analyses had different purposes. Liu et al. (2005) was interested in determining
whether there is a developmental phthalate genomic signature. The purpose of our
analysis was to identify all affected pathways.
• In the four years since the study was published, gene and pathway annotation has
increased.
Repeated identification of DEGs and pathways via different analysis methods provides an
additional level of confidence regarding the importance of "common" DEGs and pathways.
However, it is important to note that the lack of repeated identification of a gene or pathway does
not necessarily indicate a lack of biological importance for these genes or pathways.
We also asked whether there were appropriate data to develop a temporal gene network
model, a sequence of the gene and pathway interactions over time, for DBF. Using the data from
Thompson et al. (2005), the only time-course study available at the time of the project, changes
in gene expression and pathways were modeled (see Figure 6-5). Two limitations of these data
are that (1) the exposure interval was at the tail end of the critical window of exposure, GD 18, a
time that most consider too late to induce the full spectrum of male reproductive developmental
effects; and (2) the duration of exposure and developmental time were not aligned because all
animals were sacrificed on GD 19 (i.e., the 1 hr time point was the latest in development; see
Sections 5.3.1 and 6.2.2 for more discussion). The more recent study of Plummer et al. (2007)
could provide more appropriate data for building a temporal and spatial network model as both
7-7
-------
time-course of exposure over the critical window of development and microdissection of the
testis cell types were employed in their study.
7.2.2. Interspecies MOA Case Study Question: Do the DBF Genomic Data Inform
Interspecies Differences in TD?
Human gene expression data are not available for DBF. Therefore, the case study used
information on interspecies similarities of the affected pathways from other available data and
methodologies. We explored the interspecies (rat-to-human) differences in the TD part of MOA,
focusing on the steroidogenesis pathway underlying one of the DBF MO As, the decrease in fetal
testicular T MOA. Comparisons of the steroidogenesis genes and pathway were performed to
evaluate cross-species similarity metrics using three approaches: (1) protein sequence similarity;
(2) pathway network similarities; and (3) promoter-region conservation (see Chapter 6). Results
from all three approaches indicate that steroidogenesis pathways are relatively highly conserved
across rats and humans and, thus, qualitatively, the rat and human mechanisms for
steroidogenesis are highly similar.
These results further corroborate what is known about the similar roles for androgens
during normal male development in both rats and humans. However, the data sources used for
all three approaches have gaps in the knowledge bases. The pathway network diagramming data
source is not of high enough quality or comprehensive enough to utilize quantitatively. In fact, it
is difficult to use any of the three new lines of evidence to quantitatively inform the relative
sensitivity to DBF across species. It is possible that the small differences across species have a
strong penetrance, leading to significant differences in the specific enzymes that may become
more sensitive to DBF and thus, affecting T production. We further considered whether some
steroidogenesis genes are of higher relative importance and, thus, should be weighted higher in a
cross-species assessment of the steroidogenesis pathway. The initiating event for DBF action in
the male reproductive developmental outcomes has not been established. However, some
information about the rate-limiting steps for steroidogenesis, in the unperturbed scenario, is
available (reviewed in Miller, 2008). Some studies have identified CYP11 Al (also called
P450SCC) as a limiting enzymatic step for T production (Omura and Morohashi, 1995; Miller,
1988). However, the available information on kinetics reflects the unperturbed state because the
rate-limiting step was defined in assays without DBF exposure. Additionally, the rate-limiting
step information is limited in scope to steroidogenic enzymes and not all upstream activities
7-8
-------
leading to T production, such as STAR, a protein that impacts the availability of cholesterol (by
transporting cholesterol to the inner mitochondrial membrane for cleavage by P450SCC) for T
production. Thus, there is no a priori knowledge to argue for placing more weight on a
particular gene leading to T production.
Because there are some questions as to the reliability of the data used to generate the
pathway comparisons used for each species and the relative importance of individual
steroidogenesis enzymes, there is no basis on which to transform a measure of conservation to a
quantitative measure of sensitivity. While the confidence in these cross-species comparisons of
the steroidogenesis pathway were not high enough to use the findings quantitatively, for the DBF
example, the findings do add to the WOE suggesting that the role of T in male fetal development
in rats and humans is well-conserved. These methods, however, when based on high-quality
data, could be applied quantitatively to future chemical assessments. Further, the exploratory
methods for developing metrics for cross-species pathway similarities described in this document
(see Chapter 6) could be developed further and validated in the future for quantitative use in risk
assessment.
7.2.3. Application of Genomic Data to Risk Assessment: Exploratory Methods and
Preliminary Results
Chapter 6 describes exploratory methods and preliminary results for analyzing genomic
data for risk assessment application, developing a DBF gene network model, and measuring
cross-species differences for a given pathway.
None of the DBF genomic studies were designed with the application to risk assessment
in mind. Methods for analyzing microarray and other -omic data were originally developed for
screening purposes (i.e., designed to err on the side of false positives over false negatives). For
risk assessment application, genomic analytical tools are needed that are different from those
used in screening that can reliably separate signal from noise. In traditional pathway level
analysis, first, DEGs are identified by a statistical filter, and second, significant genes are
mapped to their respective pathways. Typically, the presence of three affected genes (DEGs)
within a pathway is the cut-off for identifying a particular pathway. Depending on the number of
genes that map to any given pathway, the role of the pathway can be over- or underestimated.
To overcome this problem, we explored using the pathway activity level method (calculating
PALp) that identifies affected pathways in the single step. This method ranks pathways based on
7-9
-------
the expression level of all genes in a given pathway and shows promise for use in risk
assessment and further validation is underway.
Gene network models can be very useful for understanding the temporal sequence of
critical biological events perturbed after chemical exposure, and thus, useful to a risk assessment.
We developed a method for developing a gene network model for DBF based on the available
data. The availability of one time-course study (Thompson et al., 2005) enabled our group to
model the series of events that occurred between exposure to DBF and the onset of reproductive
outcomes. However, given the limitations of the Thompson et al. (2005) study design, we could
not determine the genes and pathways affected by DBF exposure earliest in the critical window
from this study. However, the exercise allowed us to develop methods for analyzing time-course
data for use in gene network modeling.
We also explored the use of three different methods to assess rat-to-human conservation
as metrics that may inform the interspecies differences for one MO A, reduced fetal testicular T
(Section 7.2.2). More work in the area of cross-species metrics is needed. Efforts to address the
challenges in using similarity scores to quantitatively estimate the human relevance of an MOA
are ongoing (Section 6.3).
7.2.4. Application of Genomic Data to Risk Assessment: Using Data Quantitatively
This case study was limited to qualitative uses of genomics in risk assessment due
to the absence of dose-response, global gene expression studies (i.e., microarray studies)
for DBF. EPA and the larger scientific community working with genomics are interested
in methods to use genomic data quantitatively in risk assessment. There is one dose-
response RT-PCR study that, although not a genomic (i.e., not global) study, was
considered for use quantitatively in risk assessment (Lehmann et al., 2004; see
Table 7-1). Some strengths of the Lehmann et al. (2004) study include the following:
• The study includes multiple doses ranging from low to high.
• Some of the genes assessed in this study were first identified in microarray studies,
providing a relatively high level of confidence in the connection between the expression
of some of the genes and particular outcomes, as well as demonstrating reproducibility
across studies. For example, findings for Star gene expression are reproduced across
protein expression, RT-PCR, and microarray studies.
7-10
-------
Table 7-1 DBF dose-response progression of statistically significant events illustrated with a subset of precursor
event data (steroidogenesis gene expression, T expression) and in vivo endpoints with the reduced T MOA
Precursor event3
In vivo endpoint
0.1 mg/kg-d
| HsdSb
1 mg/kg-d
| HsdSb
I Scarbl
10 mg/kg-d
NC in gene exp.
NC in [T]
30 mg/kg-d
ND for gene
exp.
NC in [T]
50 mg/kg-d
[Scarbl
IHsdSb
[Star
[Cypllal
4[T]
80 mg/kg-d
ND for gene exp.
t incidence of absent,
poorly developed, or
atrophic testis and
underdeveloped or
absent epididymisb
100 mg/kg-d
[Scarbl
[3/3-HSD
[StAR
\P450scc
4[T]
Retained nipples
and areolae0
NC, no statistically significant change; ND = not determined (Lehmann et al. [2004] did not test 80 mg/kg-d).
"Lehmann et al. (2004).
bNTP (1991).
TVIylchreest et al. (2000).
-------
However, there are a number of issues in applying these dose-response RT-PCR
data with confidence to BMD modeling. These limitations include the following:
• Some of the gene expression changes are not reproducible. For example, Kit was
observed to be significantly altered in the Lehmann et al. (2004) study but was not
observed to be significantly reduced after in utero DBF exposure in a microarray study
(Liu et al., 2005) utilizing the Affymetrix gene chip, yet Kit is on the Affymetrix rat chip.
• The relationship between statistical and biological significance is not known for these
gene expression data. For example, the expression ofHsdSb mRNA is statistically
significantly altered at lower doses than a statistically significant T decrease was
observed. Thus, Lehmann et al. (2004) argued that the changes in Hsd3b at 0.1 and
1.0 mg/kg-d were not biologically significant. Alternatively, Hsd3b gene expression
changes could be a precursor to T level changes in time and thus, be a valid precursor
event. It is also not known whether changes in the expression of a single or multiple
steroidogenesis genes would lead to a significant alteration in T and the phenotype.
• Interlitter variability could not be characterized from the Lehmann et al. (2004) data
because the RT-PCR data were collected on five individual pups representing four to
five litters per treatment group (i.e., ~1 pup/litter). In order to have appropriate data for
BMD modeling, litter mean values calculated from a study with a greater sample size and
multiple litters are needed to allow characterization of intra- and interlitter variability.
The use of the litter as the statistical unit is generally agreed upon because of the high
variability in gene expression for pups within one litter (Barlow et al., 2003).
We concluded that the available dose-response RT-PCR data for DBF are not of sufficient
quality due to the lack of information about interspecies variability. Additionally, there is not
sufficient knowledge about the biological significance of a gene expression change (and the level
of change that is biologically significant), for one or a subset of genes, that would invariably lead
to a reduction in T and in turn, lead to the observed male reproductive developmental outcomes.
7.3. LESSONS LEARNED
The lessons learned from the case study are grouped by research needs that are useful to
research scientists and recommendations that are useful to risk assessors.
7-12
-------
7.3.1. Research Needs
7.3.1.1. Data Gaps and Research Needs: DBF
There are some research needs that would be very useful, specifically for a DBF risk
assessment including the following:
1. Develop a gene network model for DBF using the Plummer et al. (2007) data. This data
set would be an excellent source of temporal and spatial gene expression information
because one of its studies includes three time intervals, thus covering the entire critical
window for male reproductive outcomes, and a second study used microdissection of the
cord and interstitial cells of the testis. This study was not modeled because it was not
published until after the modeling work had been completed, and we had not obtained the
data. By comparing gene expression, Plummer et al. (2007) hypothesized the MO A
underlying the gonocyte and LC effects.
2. Perform microarray studies in male reproductive tissues, other than the testis, affected by
DBF in order to understand the similarities and differences in DBF-affected pathways
across reproductive organs and tissues in the male rat. Bowman et al. (2005) performed
such a study in the WDs, but studies in other male tissues are needed.
3. Perform microarray studies in human tissues (either cell lines or from aborted male fetal
tissue), along with parallel in vitro and in vivo studies in rats for validation and
comparison. Such data would provide valuable information about interspecies
differences in TD sensitivity. Some human studies found an association between in utero
phthalate exposure and newborn male reproductive developmental measures (Main et al.,
2006; Swan et al., 2005) that indicate human relevance for some of the DBF effects
observed in male rat studies.
4. Perform w ell-designed proteomic and metabolomic studies to understand the effects of in
utero DBF exposure on the function of expressed proteins and on cellular metabolites.
These data may provide complementary data to the available transcriptomic data, which
could yield some new insights.
5. Perform genomic studies to identify early, critical, upstream events as a means to identify
the initiating event for DBF's action in the testis. This would require performing studies
much earlier in gestation, at the beginning of sexual differentiation. In addition, such
studies may require greater sensitivity regarding gene expression change identification
because a statistically significant change may be greater than a biologically significant
change. If identified, the initiating event could be utilized in the risk assessment, thereby
reducing uncertainty.
6. Perform genomic studies to under stand whether the female reproductive tract
malformations after DBF exposure have common or different MOAs with the male
development reproductive effects. This line of research would identify pathways affected
in the developing female reproductive tracts after early gestational DBF exposure.
7-13
-------
7. Compare the affected DEGs and pathways between the phthalates with and without
developmental effects could be useful for a cumulative risk assessment of the
developmental phthalates. All of the data from the Liu et al. (2005) data set could be
utilized to evaluate this issue. Further, evaluating consistency of findings across
chemicals in the same MOA class that do and do not produce the same set of effects
could be useful for improving specificity of the pathway and MOA findings for DBF.
8. Studies to distinguish affected genes and pathways that may be compensatory vs. those in
the causal pathway for DBP-toxicity.
7.3.1.2. Research Needs for Toxicity and Toxicogenomic Studies for Use in Risk Assessment
EPA and the larger scientific community are interested in methods to use genomic data
quantitatively in risk assessment. This case study was limited to qualitative uses of genomics in
risk assessment due to the absence of dose-response global gene expression studies for DBF.
This is the case for many chemicals as multiple dose studies are very costly. However, multiple
dose microarray or other global gene expression studies are needed (see Table 7-2). Such studies
need to be designed properly such that the identification and interpretation of lower dose effects
is possible. Gene expression changes at the lower dose may not be affected in every organ,
tissue or cell sample assessed. High single dose microarray studies have been performed such
that all organs are affected and one can assess a smaller sample size than for a dose-response
study. In a dose-response study including low- to high doses, the sample size per dose group
would need to be high enough to increase statistical power (i.e., the detection of gene expression
changes when only a few animals are affected). For example, if an endpoint is affected in 20%
of the animals at lower doses, then the sample size for microarray studies must be large enough
to identify the affected animals (with affected gene expression). Perhaps the highest priority
study is one that assesses global gene expression and toxicity endpoints of interest as
components of the same experiment; the organ or tissue of interest would be collected at the
appropriate age in one group of animals and a second group would be followed through to
evaluation of the endpoint of interest. In this manner, such a study would generate data that
could define the relationships between dose, time of exposure, gene expression level changes,
pathway level changes, and in vivo changes.
Table 7-2 describes some of the priority research needs for toxicogenomic studies for
developmentally toxic chemicals, including DBF. First, appropriate time-course gene expression
data over the critical window, using a small subset of genes whose altered expression is linked to
7-14
-------
Table 7-2. Research needs for toxicogenomic studies to be used in risk
assessment
Purpose
1) Develop a gene network model
2) Improve pathway analysis statistical
power
3) Use of toxicogenomic data to inform
toxicokinetics in dose-response
analysis
4) Use of toxicogenomic data in dose-
response analysis
5) Determining the degree of phenotypic
anchoring; informing MO As (see
Figure 3-4)
6) Assess intraspecies differences
7) Assess interspecies differences
8) Appropriate statistical pathway
analysis methods for use in risk
assessment
9) Screening and categorizing chemicals
by MOA in risk assessment (e.g.,
cumulative risk assessment)
Study needed
Exposure time-course microarray data.
Number of replicates increased.
Genomic and toxicity studies with same study
design: Generate TK data in relevant study (time,
dose, tissue), and obtain relevant internal dose
measure to derive best internal dose metric.
Multiple doses in microarray studies in parallel
with toxicity studies for phenotypic anchoring.
Similar study design characteristics for genomic
and toxicity studies (i.e., dose, timing of
exposure, organ/tissue evaluated). This includes
assessing whether genes and pathways are due to
compensatory mechanisms and/or general toxic
responses.
A study assessing multiple doses across rat
strains (e.g., Wistar vs. SD); endpoint and
microarray component of the study.
A study to assess whether different species with
similar pathways (genes and sequence of steps)
have a similar sensitivity to a given chemical.
The findings could potentially enhance the utility
of toxicogenomic data to aid species
extrapolation in risk assessments.
Further comparisons and evaluations of different
methods.
Genomic (transcriptomic, proteomic, and/or
metabolomic) signatures can be particularly
useful for screening and categorizing chemicals
by MOA in risk assessment.
the outcome of interest, would be very relevant for developing a regulatory network model.
These studies need to be carefully designed based on the information on the critical window of
exposure and the relationship to the particular outcome of concern. Second, the statistical power
of pathway-analysis methods for global expression techniques, including microarrays,
7-15
-------
proteomics and metabolomics, could be improved by designing and performing studies with
more replicates. Thus, variability would be better characterized. Third, it would be helpful to
design genomic studies that could inform both TK and dose-response (see Table 7-2, #3 and #4).
Performing genomic and toxicity studies with similar designs would provide useful
information. These studies would be designed at the most relevant time of exposure, include low
to high doses, and assess the relevant tissues. Relevant internal dose measurements could be
obtained on which to base the internal dose metric. These studies, employing genomic and
toxicity studies of comparable designs, would provide information about the relationship of dose,
gene expression, and outcome, and thus, could potentially be used in dose-response analysis.
Studies with both a toxicity and toxicogenomic component would obviously require assessment
of a large sample size to be informative. These same studies could be used to inform MO As
(Table 7-2, #5) and could be adapted to comparing species (Table 7-2, #6). Regarding
quantitative measures of intraspecies and interspecies differences, it should be noted that the
same information which is necessary for quantitative assessment of interspecies differences
(Section 7.2.2) may be useful for characterizing intraspecies variability, and vice versa. In
particular, factors that explain or predict interstrain differences in rodent sensitivity to DBF, such
as those noted between Wistar and SD rats, may be hypothesized to contribute to human
variability. Further, lexicologically important interstrain differences identified from the
toxicogenomic data could be an excellent data source for investigating whether they are also
important for modulating interspecies sensitivity.
Finally, further development and comparison studies to identify appropriate statistical
pathway analysis methods for use in risk assessment are needed (Table 7-2, #8). It is important
to note that such studies require research funding and laboratories with expertise in both
genomics and toxicology. Research needs for toxicity studies that would improve the utility in
risk assessment are also described in Table 7-3. As was noted for the DBF case (see Chapter 4),
complete reporting is necessary for studies that are intended for use in risk assessment.
7.3.2. Recommendations
Based on the lessons learned from performing the DBF case-study exercise, we
developed some recommendations or best practices for evaluating genomic data in new
assessments. The approach includes systematic consideration of
7-16
-------
Table 7-3. Research needs for toxicity studies for utilizing toxicogenomic and
toxicity data together in risk assessment
Study aspect
Study design
Reporting
Research need
Exposing animals during optimal
developmental stage/time (i.e., for the
critical window).
Assessing outcome at optimum
developmental stage/time for that outcome.
Parallel study design characteristics with
toxicogenomic studies (i.e., dose, timing of
exposure, organ/tissue evaluated) to obtain
comparable toxicity and toxicogenomic
studies to aid in determining relationships
between gene expression changes and
outcomes.
Individual animal data to aid identification
of low incidence effects, correlate gene
expression changes and outcomes, and
characterize intraspecies variability.
All endpoints that were evaluated
(independent of whether the outcome was
positive or negative).
the genomic data for whether they could inform risk assessment steps, identification of questions
to direct the evaluation, and evaluation of the genomic data and toxicity data to assess
phenotypic anchoring. In addition, we have some specific recommendations. The first two
recommendations are straightforward and could reasonably be performed by a risk assessor with
basic training in genomics data evaluation and interpretation while the third recommendation
requires expertise in genomic data analysis methods for implementation. The recommendations
are presented below:
1. Evaluate the genomic and other gene expression data for consistency of findings across
studies to provide a WOE evaluation of the affected gene expression and pathways.
Some simple methods, such as using Venn diagrams and gene expression compilation
approaches, can be applied to risk assessment. When evaluating the consistency of
toxicogenomic data findings, it is advantageous to include all available gene expression
data (single gene, global gene expression, protein, RNA) because single gene expression
7-17
-------
techniques have been traditionally used to confirm the results of global gene expression
studies and because single gene expression data add to the database.
2. Perform BMD modeling on high-quality RT-PCR dose-response studies of genes known
to be in the causal pathway of an MO A or outcome of interest. Obtaining a BMD and
BMDL is a useful starting point for both linear low-dose extrapolation and reference
value approaches. We are not indicating which approach is appropriate to take for
making predictions about the potential risk below the BMD or BMDL. "High quality" is
defined in this context as a well-conducted study that assessed enough animals and litters
for sufficient statistical power for characterizing the mean responses and the variability
(interlitter and intralitter).
3. Perform new analysis of toxicogenomic data in cases when the new analysis is likely to
yield new information that would be useful to the risk assessment. Examples include:
• Perform a new pathway analysis in order to identify all affected pathways or other
risk assessment applications. When the available published microarray studies
have been conducted for purposes (e.g., basic science, pharmaceutical
development) other than risk assessment, it may be useful to reanalyze the data
for risk assessment purposes. Information about all affected pathways may
contribute to an understanding of the mechanisms and MO As.
• Identify the genes and pathways affected over a critical window of exposure if
global gene expression time-course data are available. Specifically, by
developing a gene network over time, it may be possible to identify the earliest
affected genes and/or pathways, which in turn may represent the earlier or
initiating events for the outcome of interest.
7.3.3. Application of Genomic Data to Risk Assessment: Future Considerations
A number of the issues that emerged in evaluating the DBF genomic data set are relevant
to using genomic data in risk assessment in general. Some issues regarding the use of genomic
data are to the same as for the use of precursor information in risk assessment, regardless of the
technique used to gather the information. Two outstanding questions are
How is the biologically significant level of change in a precursor marker determined?
And, specifically for toxicogenomic data, what are the key genes (i.e., a key gene, a
handful of genes associated with the outcome of interest, a genomic signature) whose
altered expression leads to an adverse outcome? Currently, decisions about the degree of
change of a precursor event tend to be based on statistical significance because data to
address biological significance are typically lacking (as is the case for T levels and male
development of the testis). Genes are identified as DEGs in microarray studies based on
statistical-significance criteria that may not reflect biological significant changes (i.e.,
identified genes may not be biologically meaningful while unidentified genes may be
7-18
-------
meaningful). This point is also relevant to the question: What pathway analysis methods
are most appropriate for risk assessment? As noted in Section 5.5, it is difficult to know
whether one has identified the biologically relevant DEGs and pathways. Statistically
significant changes and repeated findings of the same genes and pathways across studies
and using different analytical methods, while providing corroboration, do not necessarily
provide a greater confidence regarding biological significance of these genes and
pathways over other genes and pathways. Further, there is a bias towards the well-
annotated genes as biologically significant when, in fact, the unannotated genes could be
of greater importance.
• What are the requirements for linkage of precursor events to in vivo endpoints? Studies
to assess the relationship between the gene expression and outcomes are needed to
establish a causal connection. It is important to note that DBF has two well-established
MO As and strong phenotypic anchoring of some gene expression changes, which is not
typical.
There are also a number of technical issues in utilizing microarray data in EPA risk
assessments that have not fully been surmounted. The primary technical issue is the validation
of the reproducibility of microarray study results. Reproducibility depends on biological sample
preparation, interlaboratory (presumably related to operator and protocol differences),
intralaboratory (presumably related to operator differences), and batch and platform variability.
The results of the MAQC-I project (see Chapters 2 and 5) revealed that reproducibility was
achieved when using the same biological sample. This is very encouraging for using microarray
data in risk assessment. However, biological sample variability still needs to be addressed in
order that protocols and details of the underlying reasons for the variability can be understood.
MACQ-II and III projects are underway to address additional technical issues (see Chapter 2).
A number of the issues stem from the complexity of the data output from the global
expression techniques (e.g., microarrays, proteomics, and metabolomics). This is in part a
training issue. To address the training needs, the EPA Risk Assessment Forum held introductory
and intermediate level training in genomics in 2007, and the FDA held genomics training
(http://www.fda.gov/cder/genomics/Default.htm). However, it would be advantageous for
organizations that perform risk assessments to embark on further training of risk assessors to
enable them to perform analyses of microarray and other genomic data analysis techniques, and
to understand the issues in applying traditional analytical methods to risk assessment.
If additional case studies are performed using the approach outlined in Figure 7-1, then
we recommend a chemical whose exposure leads to both cancer and noncancer outcomes to
7-19
-------
explore use of these data for multiple outcomes, as well as the impacts on the different risk
assessment paradigms and processes (e.g., cancer vs. noncancer). Further, performing case
studies on data-rich and data-poor chemicals would aid in further evaluating the approach
described herein. For instance, performing a case study on a chemical with dose-response data
and on a chemical with human polymorphism data would address issues in evaluating these types
of data for risk assessment, allowing further refinement of the approach.
The approach for utilizing toxicogenomic data in risk assessment outlined in this
document may be applied to other chemical assessments. This document advances the effort to
devise strategies for using genomic data in risk assessment by defining an approach, performing
a case study, and defining critical issues that need to be addressed to better utilize these data in
risk assessment. This case study serves as an example of the considerations and methods for
using genomic data in future risk assessments.
7-20
-------
APPENDIX A.
SUPPORTING TABLES FOR CHAPTER 5
Appendix A contains additional tables that support the work shown in Chapter 5.
A-l
-------
Table A-l. Weight of evidence (WOE) for statistically significant gene expression changes after in utero
exposure to dibutyl phthalate (DBF) from the whole rat testis microarray studies'1 as reported in Thompson et
al. (2005)b, Shultz et al. (2001)b, Liu et al. (2005)c d, and Plummer et al. (2007)e
Official
gene
symbol
Aacs
Aadat
Abcgl
Acaal
Acaca
Acadl
A cads
AcsU
AdamlS
Adamtsl
Admr
Adralb
Akt2
Alasl
Alasl
Official gene namef
Acetoacetyl-CoA synthetase
Aminoadipate aminotransferase
ATP-binding cassette, sub-family G (WHITE),
member 1
Acetyl-Coenzyme A acyltransferase 1
Acetyl-Coenzyme A carboxylase alpha
Acetyl-Coenzyme A dehydrogenase, long-chain
Acyl-Coenzyme A dehydrogenase, short chain
Acyl-CoA synthetase long-chain family member 4
A disintegrin and metallopeptidase domain 15
(metargidin)
A disintegrin-like and metallopeptidase (reprolysin
type) with thrombospondin type 1 motif, 1
Adrenomedullin receptor
Adrenergic receptor, alpha Ib
Thymoma viral proto-oncogene 2
Aminolevulinic acid synthase 1
Aminolevulinic acid synthase 1
Exposure
window
CDs 12-19
CDs 12-19
CDs 12-19
CDs 12-19
CDs 12-19
CDs 12-19
CDs 12.5-15.5
CDs 12-19
CDs 12.5-17.5
CDs 12.5-19.5
CDs 12-19
CDs 12-19
CDs 12-21
CDs 12-19
CDs 12.5-17.5
Up or
down
Down
Down
Up
Down
Down at
GD19
Down at
GD19
Up
Down
Up
Down
Down
Down
Down at
GD21
Down
Down
Fold
change
-0.37 Iog2
-0.38 Iog2
0.38 Iog2
-0.37 Iog2
>2
>2
1.50
-0.60 Iog2
1.20
-1.35
-0.90 Iog2
-0.30 Iog2
>2
-1.01 Iog2
-1.33
Cutoff used (method)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
2-fold
2-fold
p< 0.01 (ANOVA)
p < 0.05 (ANOVA)
p< 0.01 (ANOVA)
p< 0.01 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
2-fold
p < 0.05 (ANOVA)
p< 0.01 (ANOVA)
Reference
Liu et al., 2005
Liu et al., 2005
Liu et al., 2005
Liu et al., 2005
Shultz etal., 2001
Shultz etal., 2001
Plummer etal.,
2007
Liu et al., 2005
Plummer etal.,
2007
Plummer etal.,
2007
Liu et al., 2005
Liu et al., 2005
Shultz etal., 2001
Liu et al., 2005
Plummer etal.,
2007
to
-------
Table A-l (continued)
Official
gene
symbol
Alasl
Aldhla3
Aldh2
Aldh2
Aldh2
Aldoa
Aldoc
AnxaS
Aoxl
Aqpl
Arfl
Arrb2
Asns
Ass
At/2
Official gene namef
Aminolevulinic acid synthase 1
Aldehyde dehydrogenase family 1, subfamily A3
Aldehyde dehydrogenase 2
Aldehyde dehydrogenase 2
Aldehyde dehydrogenase 2
Aldolase A, fructose-bisphosphate
Aldolase C
Annexin A5
Aldehyde oxidase 1
Aquaporin 1
ADP-ribosylation factor 3
Arrestin, beta 2
Asparagine synthetase
Argininosuccinate synthetase
Activating transcription factor 2
Exposure
window
CDs 12.5-19.5
CDs 12-19
CDs 12-19
CDs 12.5-17.5
CDs 12.5-19.5
CDs 12.5-19.5
CDs 12-19
CDs 12.5-19.5
CDs 12-19
CDs 12.5-15.5
CDs 12.5-17.5
CDs 12-21
CDs 12-19
CDs 12-19
CDs 12-21
Up or
down
Down
Down
Down
Down
Down
Down
Down
Down
Down
Down
Down
Down at
GD21
Down
Down
Up at
GD21
Fold change
-1.44
-0.43 Iog2
-0.82 Iog2
-1.50
-1.91
-1.24
-0.44 Iog2
-1.20
-0.50 Iog2
-1.29
-1.23
>2
-0.24 Iog2
-0.82 Iog2
>2
Cutoff used (method)
p< 0.01 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p< 0.01 (ANOVA)
p< 0.01 (ANOVA)
p< 0.01 (ANOVA)
p < 0.05 (ANOVA)
p< 0.01 (ANOVA)
p < 0.05 (ANOVA)
p< 0.01 (ANOVA)
p< 0.01 (ANOVA)
2-fold
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
2-fold
Reference
Plummeretal.,
2007
Liu et al., 2005
Liu et al., 2005
Plummeretal.,
2007
Plummer et al.,
2007
Plummeretal.,
2007
Liu et al., 2005
Plummeretal.,
2007
Liu et al., 2005
Plummeretal.,
2007
Plummeretal.,
2007
Shultzetal.,2001
Liu et al., 2005
Liu et al., 2005
Shultzetal.,2001
-------
Table A-l (continued)
Official
gene
symbol
Atf4
Atplbl
Atp4b
AtpSfl
Baiap2
Bhlhb2
Bhmt
BircS
Btg2
Btg2
C4a
Cadps
Calb2
Cd63
Cdknlc
Official gene namef
Activating transcription factor 4
ATPase, Na+/K+ transporting, beta 1 polypeptide
ATPase, H+/K+ exchanging, beta polypeptide
ATP synthase, H+ transporting, mitochondria! FO
complex, subunit B 1
Brain-specific angiogenesis inhibitor 1 -associated
protein 2
Bhlhb2 basic helix-loop-helix domain containing,
class B2
Betaine-homocysteine methyltransferase
Baculoviral IAP repeat-containing 5
B-cell translocation gene 2, anti-proliferative
B-cell translocation gene 2, anti-proliferative
Complement component 4a
Ca2+-dependent secretion activator
Calbindin 2
CD63 antigen
Cyclin-dependent kinase inhibitor 1C (P57)
Exposure
window
GD19for3hr
CDs 12-19
CDs 12-19
CDs 12.5-15.5
CDs 12-19
GD 19 for 3 hr
CDs 12-19
CDs 12.5-15.5
GD 19 for 1 hr
GD 19 for 3 hr
GD 19 for 6 hr
CDs 12-19
CDs 12-19
CDs 12.5-19.5
CDs 12-19
Up or
down
Up after
3hr
Down
Down
Up
Down
Up after
3hr
Down
Up
Up
afterl hr
Up after
3hr
Down after
6 hr
Up
Down
Down
Down
Fold change
0.67
-0.24 Iog2
-0.60 Iog2
1.22
-0.22 Iog2
0.88
-0.24 Iog2
1.68
1.30
1.88
-0.77
0.311og2
-0.77 Iog2
-1.36
-0.81 Iog2
Cutoff used (method)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p< 0.01 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p< 0.01 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p< 0.01 (ANOVA)
p < 0.05 (ANOVA)
Reference
Thompson et al.,
2005
Liu et al., 2005
Liu et al., 2005
Plummer et al.,
2007
Liu et al., 2005
Thompson et al.,
2005
Liu et al., 2005
Plummer et al.,
2007
Thompson et al.,
2005
Thompson et al.,
2005
Thompson et al.,
2005
Liu et al., 2005
Liu et al., 2005
Plummer etal.,
2007
Liu et al., 2005
-------
Table A-l (continued)
Official
gene
symbol
Cdknlc
Cdknlc
Cebpb
Cebpd
Clu
Clu
Cmklrl
Cnrl
Cnbp
Cpal
Cpal
Cpd
Cpe
Cptla
Official gene namef
Cyclin-dependent kinase inhibitor 1C (P57)
Cyclin-dependent kinase inhibitor 1C (P57)
CCAAT/enhancer binding protein (C/EBP), beta
CCAAT/enhancer binding protein (C/EBP), delta
Clusterin
Clusterin
Chemokine-like receptor 1
Cannabinoid receptor 1 (brain)
Cellular nucleic acid binding protein
Carboxypeptidase Al
Carboxypeptidase Al
Carboxypeptidase D
Carboxypeptidase E
Carnitine palmitoyltransferase la, liver
Exposure
window
GD19for6hr
CDs 18-19 for
18 hr
CDs 12-19
GD19for3hr
CDs 12-21
GD 18 for 18 hr
CDs 12.5-19.5
GD 19 for 3 hr
CDs 12.5-19.5
CDs 12.5-17.5
CDs 12.5-19.5
CDs 12-21
CDs 12-19
CDs 12-19
Up or
down
Down after
6hr
Down after
18 hr
Down
Up after
3hr
Up at
GD21
Up after
18 hr
Down
Up after
3hr
Down
Down
Down
Up at
GD21
Up
Down at
GD 19
Fold change
-1.08
1.63
-0.6 Iog2
1.62
>2
1.03
-1.17
0.99
-1.29
-1.73
-2.33
>2
0.59 Iog2
->2
Cutoff used (method)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
2-fold
p < 0.05 (ANOVA)
p< 0.01 (ANOVA)
p < 0.05 (ANOVA)
p< 0.01 (ANOVA)
p< 0.01 (ANOVA)
p< 0.01 (ANOVA)
2-fold
p < 0.05 (ANOVA)
2-fold
Reference
Thompson et al.,
2005
Thompson et al.,
2005
Liu et al., 2005
Thompson et al.,
2005
Shultzetal.,2001
Thompson et al.,
2005
Plummer et al.,
2007
Thompson et al.,
2005
Plummer et al.,
2007
Plummer et al.,
2007
Plummer et al.,
2007
Shultzetal.,2001
Liu et al., 2005
Shultzetal.,2001
-------
Table A-l (continued)
Official
gene
symbol
Cptla
Cptlb
Cpz
Crabp2
Cmbp2
Crem
Crispld2
Cryab
Ctgf
Ctgf
Ctsb
Ctsd
CxcllO
Cyb5
Cypllal
Official gene namef
Carnitine palmitoyltransferase la, liver
Cptlb carnitine palmitoyltrans-feraselb, muscle
Carboxypeptidase Z
Cellular retinoic acid binding protein 2
Cellular retinoic acid binding protein 2
cAMP responsive element modulator
Cysteine-rich secretory protein LCCL domain
containing 2
Crystallin, alpha B
Connective tissue growth factor
Connective tissue growth factor
Cathepsin B
Cathepsin D
Chemokine (C-X-C motif) ligand 10
Cytochrome b-5
Cytochrome P450, family 11, subfamily a,
polypeptide 1
Exposure
window
CDs 12-21
CDs 12-19
CDs 12-19
CDs 12-19
GD 19 for 6 hr
GD 19 for 3 hr
CDs 12-19
CDs 12-19
GD 19 for 3 hr
GD 19 for 6 hr
CDs 12.5-15.5
CDs 12.5-19.5
GD 19 for 3 hr
CDs 12-19
CDs 12-19
Up or
down
Down at
GD21
Up
Up
Down
Down after
6hr
Up after
3hr
Down
Up
Up after
3hr
Up after
6hr
Up
Down
Up after
3hr
Down
Down
Fold change
->2
0.23 Iog2
0.21 Iog2
-0.311og2
-1.24
0.58
-0.27 Iog2
0.22 Iog2
2.10
2.37
1.53
-1.22
2.07
-0.30 Iog2
-1.07 Iog2
Cutoff used (method)
2-fold
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p< 0.01 (ANOVA)
p< 0.01 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
Reference
Shultzetal.,2001
Liu et al., 2005
Liu et al., 2005
Liu et al., 2005
Thompson et al.,
2005
Thompson et al.,
2005
Liu et al., 2005
Liu et al., 2005
Thompson et al.,
2005
Thompson et al.,
2005
Plummer et al.,
2007
Plummer et al.,
2007
Thompson et al.,
2005
Liu et al., 2005
Liu et al., 2005
>
-------
Table A-l (continued)
Official
gene
symbol
Cypllal
Cypllal
Cypllal
Cypllal
Cypllbl
Cypl7al
Cypl7al
Cypl7al
Cypl7al
CypSl
CypSl
CypSl
Dab2
Dafl
Official gene namef
Cytochrome P450, family 11, subfamily a,
polypeptide 1
Cytochrome P450, family 11, subfamily a,
polypeptide 1
Cytochrome P450, family 11, subfamily a,
polypeptide 1
Cytochrome P450, family 11, subfamily a,
polypeptide 1
Cytochrome P450, subfamily 1 IB, polypeptide 1
Cytochrome P450, family 17, subfamily a,
polypeptide 1
Cytochrome P450, family 17, subfamily a,
polypeptide 1
Cytochrome P450, family 17, subfamily a,
polypeptide 1
Cytochrome P450, family 17, subfamily a,
polypeptide 1
Cytochrome P450, subfamily 5 1
Cytochrome P450, subfamily 5 1
Cytochrome P450, subfamily 5 1
Disabled homolog 2 (Drosophila)
Decay accelerating factor 1
Exposure
window
CDs 12-19
GD 18 for 18 hr
CDs 12.5-17.5
CDs 12.5-19.5
GD 18 for 18 hr
CDs 12-19
GD 18 for 18 hr
CDs 12.5-17.5
CDs 12.5-19.5
GD 18 for 18 hr
CDs 12.5-17.5
CDs 12.5-19.5
CDs 12-19
CDs 12-19
Up or
down
Down at
GD19
Down after
18 hr
Down
Down
Down after
18 hr
Down
Down after
18 hr
Down
Down
Down after
18 hr
Down
Down
Up
Up
Fold change
->2
-1.93
-1.71
-2.85
-1.63
-1.76 Iog2
-2.1
-2.15
-3.08
-1.06
-1.59
-1.81
0.27 Iog2
0.191og2
Cutoff used (method)
2-fold
p < 0.05 (ANOVA)
;?< 0.01 (ANOVA)
p< 0.01 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p< 0.01 (ANOVA)
p< 0.01 (ANOVA)
p < 0.05 (ANOVA)
p< 0.01 (ANOVA)
p< 0.01 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
Reference
Shultzetal.,2001
Thompson et al.,
2005
Plummeretal.,
2007
Plummeretal.,
2007
Thompson et al.,
2005
Liu et al., 2005
Thompson et al.,
2005
Plummeretal.,
2007
Plummeretal.,
2007
Thompson et al.,
2005
Plummeretal.,
2007
Plummeretal.,
2007
Liu et al., 2005
Liu et al., 2005
>
-------
Table A-l (continued)
Official
gene
symbol
Dbi
Dbi
Dec
Ddc
Ddc
Ddc
Ddit4
Ddit4
Ddt
Decrl
Dhcr?
Dhcr?
Dhcr?
Dnm3
Duspl
Dusp6
Official gene namef
Diazepam binding inhibitor
Diazepam binding inhibitor
Deleted in colorectal carcinoma
Dopa decarboxylase
Dopa decarboxylase
Dopa decarboxylase
DNA-damage-inducible transcript 4
DNA-damage-inducible transcript 4
D-dopachrome tautomerase
2,4-dienoyl Co A reductase 1, mitochondrial
7-dehydrocholesterol reductase
7-dehydrocholesterol reductase
7-dehydrocholesterol reductase
Dynamin 3
Dual specificity phosphatase 1
Dual specificity phosphatase 6
Exposure
window
CDs 12-19
CDs 12.5-19.5
CDs 12-19
CDs 12-19
GD 18 for 18 hr
CDs 12.5-19.5
CDs 12-19
GD 18 for 18 hr
CDs 12.5-19.5
CDs 12-19
CDs 12-19
GD 19 for 6 hr
CDs 18-19 for
18 hr
CDs 12-19
GD19for3hr
CDs 12-19
Up or
down
Down
Down
Down at
GD19
Down
Down after
18 hr
Down
Down
Down after
18 hr
Down
Down
Down
Down after
6hr
Down after
18 hr
Down
Up after
3hr
Up
Fold change
-0.38 Iog2
-1.28
->2
-1.141og2
-1.38
-1.44
-1.02 Iog2
-1.57
-1.22
-0.21 Iog2
-0.73 Iog2
-1.34
-1.18
-0.27 Iog2
0.91
0.39 Iog2
Cutoff used (method)
p < 0.05 (ANOVA)
;?< 0.01 (ANOVA)
2-fold
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p< 0.01 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p< 0.01 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
Reference
Liu et al., 2005
Plummeretal.,
2007
Shultzetal.,2001
Liu et al., 2005
Thompson et al.,
2005
Plummeretal.,
2007
Liu et al., 2005
Thompson et al.,
2005
Plummeretal.,
2007
Liu et al., 2005
Liu et al., 2005
Thompson et al.,
2005
Thompson et al.,
2005
Liu et al., 2005
Thompson et al.,
2005
Liu et al., 2005
>
oo
-------
Table A-l (continued)
Official
gene
symbol
Dusp6
Ebp
Echsl
Egrl
Egr2
Egr2
ElovlS
Elovl6
Emp3
Enol
Enpep
EntpdS
Epasl
Ephxl
Erbb2
Official gene namef
Dual specificity phosphatase 6
Phenylalkylamine Ca2+ antagonist (emopamil)
binding protein
Enoyl Coenzyme A hydratase, short chain 1,
mitochondria!
Early growth response 1
Early growth response 2
Early growth response 2
ELOVL family member 5, elongation of long chain
fatty acids (yeast)
ELOVL family member 6, elongation of long chain
fatty acids (yeast)
Epithelial membrane protein 3
Enolase 1, alpha non-neuron
Glutamyl aminopeptidase
Ectonucleoside triphosphate diphosphohydrolase 5
Endothelial PAS domain protein 1
Epoxide hydrolase 1, microsomal
v-erb-b2 erythroblastic leukemia viral oncogene
homolog 2, neuro/glioblastoma derived oncogene
homolog (avian)
Exposure
window
GD19for3hr
CDs 12-19
CDs 12-19
CDs 12-19
GD 19 for 1 hr
GD 19 for 3 hr
CDs 12-19
CDs 12-19
CDs 12.5-19.5
CDs 12.5-19.5
CDs 12-19
CDs 12-19
CDs 12-19
CDs 12-19
CDs 12.5-17.5
Up or
down
Up after
3hr
Down
Down
Up
Up after
Ihr
Up after
3hr
Down
Down
Down
Down
Up
Down
Down
Down
Up
Fold change
1.28
-0.64 Iog2
-0.181og2
0.77 Iog2
1.93
1.53
-0.171og2
-0.40 Iog2
-1.24
-1.63
0.48 Iog2
-0.52 Iog2
-0.21 Iog2
-0.57 Iog2
1.26
Cutoff used (method)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p< 0.01 (ANOVA)
p< 0.01 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p< 0.01 (ANOVA)
Reference
Thompson et al.,
2005
Liu et al., 2005
Liu et al., 2005
Liu et al., 2005
Thompson et al.,
2005
Thompson et al.,
2005
Liu et al., 2005
Liu et al., 2005
Plummer et al.,
2007
Plummer et al.,
2007
Liu et al., 2005
Liu et al., 2005
Liu et al., 2005
Liu et al., 2005
Plummer etal.,
2007
-------
Table A-l (continued)
Official
gene
symbol
Etfdh
Ezr
Ezr
F10
Fabp3
FabpS
Fabp3
FabpS
Fabp3
FabpS
Fabp6
Fadsl
Fadsl
Fadsl
Fads2
Fatl
Official gene namef
Electron-transferring-flavoproteindehydrogenase
Ezrin
Ezrin
Coagulation factor X
Fatty acid binding protein 3
Fatty acid binding protein 3
Fatty acid binding protein 3
Fatty acid binding protein 3
Fatty acid binding protein 3
Fatty acid binding protein 5, epidermal
Fatty acid binding protein 6, ileal (gastrotropin)
Fatty acid desaturase 1
Fatty acid desaturase 1
Fatty acid desaturase 1
Fatty acid desaturase 2
FAT tumor suppressor homolog 1 (Drosophila)
Exposure
window
CDs 12-19
CDs 12-19
CDs 12-19
CDs 12-19
CDs 12-19
CDs 12-19
GD 19 for 3 hr
GD 19 for 6 hr
CDs 18-19 for
18 hr
CDs 12-19
CDs 12-19
CDs 12-19
CDs 12.5-15.5
CDs 12.5-19.5
CDs 12-19
CDs 12.5-15.5
Up or
down
Down
Up
Down at
GD19
Down
Down
Down at
GD 19
Down after
3hr
Down after
6hr
Down after
18 hr
Down at
GD 19
Down
Down
Up
Down
Down
Down
Fold change
-0.39 Iog2
0.20 Iog2
->2
-0.51 Iog2
-0.49 Iog2
->2
-0.78
-1.68
-1.09
->2
-0.23 Iog2
-0.80 Iog2
1.42
1.47
-0.42 Iog2
-1.32
Cutoff used (method)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
2-fold
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
2-fold
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
2-fold
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p< 0.01 (ANOVA)
p< 0.01 (ANOVA)
p < 0.05 (ANOVA)
p< 0.01 (ANOVA)
Reference
Liu et al., 2005
Liu et al., 2005
Shultzetal.,2001
Liu et al., 2005
Liu et al., 2005
Shultzetal.,2001
Thompson et al.,
2005
Thompson et al.,
2005
Thompson et al.,
2005
Shultzetal.,2001
Liu et al., 2005
Liu et al., 2005
Plummeretal.,
2007
Plummeretal.,
2007
Liu et al., 2005
Plummeretal.,
2007
-------
Table A-l (continued)
Official
gene
symbol
Fbp2
Fdftl
Fdftl
Fdps
Fdps
Fdps
Fdxl
Fdxl
Fdxl
Fdxl
Fdxr
Fdxr
Fgfr4
Folrl
Fos
Fos
Official gene namef
Fructose-l,6-bisphosphatase 2
Farnesyl diphosphate farnesyl transferase 1
Farnesyl diphosphate farnesyl transferase 1
Farensyl diphosphate synthase
Farensyl diphosphate synthase
Farensyl diphosphate synthase
Ferredoxin 1
Ferredoxin 1
Ferredoxin 1
Ferredoxin 1
Ferredoxin reductase
Ferredoxin reductase
Fibroblast growth factor receptor 4
Folate receptor 1 (adult)
FB J murine osteosarcoma viral oncogene homolog
FB J murine osteosarcoma viral oncogene homolog
Exposure
window
CDs 12-19
CDs 12-19
CDs 12.5-19.5
CDs 12-19
CDs 12.5-17.5
CDs 12.5-19.5
CDs 12-19
GD 18 for 18 hr
CDs 12.5-17.5
CDs 12.5-19.5
CDs 12-19
CDs 12.5-17.5
CDs 12-19
CDs 12-19
GD 19 for 1 hr
GD 19 for 3 tu-
lip or
down
Up
Down
Down
Down
Down
Down
Down
Down after
18 hr
Down
Down
Down
Down
Down
Down
Up after
Ihr
Up after
3hr
Fold change
0.28 Iog2
-0.58 Iog2
-1.40
-0.73 Iog2
-1.49
-1.41
-1.65 Iog2
-2.53
-2.06
-2.97
-0.37 Iog2
-1.41
-0.191og2
-0.48 Iog2
3.28
2.70
Cutoff used (method)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p< 0.01 (ANOVA)
p < 0.05 (ANOVA)
p< 0.01 (ANOVA)
p< 0.01 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p< 0.01 (ANOVA)
p< 0.01 (ANOVA)
p < 0.05 (ANOVA)
p< 0.01 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
Reference
Liu et al., 2005
Liu et al., 2005
Plummeretal.,
2007
Liu et al., 2005
Plummeretal.,
2007
Plummeretal.,
2007
Liu et al., 2005
Thompson et al.,
2005
Plummeretal.,
2007
Plummeretal.,
2007
Liu et al., 2005
Plummeretal.,
2007
Liu et al., 2005
Liu et al., 2005
Thompson et al.,
2005
Thompson et al.,
2005
-------
Table A-l (continued)
Official
gene
symbol
Fragl
Fragl
Fthfd
Fthfd
Fthfd
Fzd2
Gaa
GgtlS
Gjal
Glrxl
Gnrhr
Gnrhr
Gpsn2
Grbl4
Grbl4
Grbl4
Official gene namef
FGF receptor activating protein 1
FGF receptor activating protein 1
Formyltetrahydro-folate dehydrogenase
Formyltetrahydro-folate dehydrogenase
Formyltetrahydro-folate dehydrogenase
Frizzled homolog 2 (Drosophila)
Glucosidase, alpha, acid
Gamma-glutamyltransferase-like 3
Gap junction membrane channel protein alpha 1
Glutaredoxin 1 (thioltransferase)
Gonadotropin releasing hormone receptor
Gonadotropin releasing hormone receptor
Glycoprotein, synaptic 2
Growth factor receptor bound protein 14
Growth factor receptor bound protein 14
Growth factor receptor bound protein 14
Exposure
window
CDs 12-19
GD 18 for 18 hr
CDs 12-19
GD 19 for 6 hr
CDs 18-19 for
18 hr
GD 19 for 3 hr
CDs 12-19
CDs 12-19
CDs 12-19
CDs 12-19
GD 19 for 3 hr
GD 19 for 6 hr
CDs 12-19
CDs 12-19
GD19for6hr
CDs 18-19 for
18 hr
Up or
down
Down
Down after
18 hr
Down
Down after
6hr
Down after
18 hr
Down after
3hr
Down
Down
Down
Down
Up after
3hr
Up after
6hr
Down
Up
Up after
6hr
Up after
18 hr
Fold change
-0.48 Iog2
-0.65
-1.03 Iog2
-0.98
-0.83
-0.7
-0.30 Iog2
-0.32 Iog2
-0.36 Iog2
-0.20 Iog2
1.38
2.03
-0.42 Iog2
0.68 Iog2
1.78
0.93
Cutoff used (method)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
Reference
Liu et al., 2005
Thompson et al.,
2005
Liu et al., 2005
Thompson et al.,
2005
Thompson et al.,
2005
Thompson et al.,
2005
Liu et al., 2005
Liu et al., 2005
Liu et al., 2005
Liu et al., 2005
Thompson et al.,
2005
Thompson etal.,
2005
Liu et al., 2005
Liu et al., 2005
Thompson et al.,
2005
Thompson et al.,
2005
-------
Table A-l (continued)
Official
gene
symbol
Grina
Gsta2
Gsta2
GstaS
Gsta3
Gsta3
Gstm2
Gstm2
Gstm2
Gstol
Gstpl
Hao2
Hmgcr
Hmgcr
Hmgcsl
Official gene namef
Glutamate receptor, ionotropic, N-methyl
D-aspartate-associated protein 1 (glutamate
binding)
Glutathione-S-transferase, alpha type2
Glutathione-S-transferase, alpha type2
Glutathione S-transferase A3
Glutathione S-transferase A3
Glutathione S-transferase A3
Glutathione S-transferase, mu 2
Glutathione S-transferase, mu 2
Glutathione S-transferase, mu 2
Glutathione S-transferase omega 1
Glutathione-S-transferase, pi 1
Hydroxyacid oxidase 2 (long chain)
3-hydroxy-3-methylglutaryl-Coenzyme A reductase
3-hydroxy-3-methylglutaryl-Coenzyme A reductase
3-hydroxy-3-methylglutaryl-Coenzyme A
synthase 1
Exposure
window
CDs 12.5-15.5
CDs 12.5-17.5
CDs 12.5-19.5
CDs 12-19
CDs 12.5-17.5
CDs 12.5-19.5
CDs 12-19
CDs 12-21
CDs 18-19 for
18 hr
CDs 12-19
CDs 12.5-15.5
CDs 12-19
CDs 12-19
CDs 12.5-19.5
CDs 12-19
Up or
down
Up
Down
Down
Down
Down
Down
Down
Up at
GD21
Down after
18 hr
Down
Up
Down
Down
Down
Down
Fold change
1.59
-1.48
-2.23
-0.96 Iog2
-1.75
-2.63
-0.42 Iog2
>2
-0.47
-0.42 Iog2
1.34
-0.58 Iog2
-0.47 Iog2
-1.83
-1.03 Iog2
Cutoff used (method)
;?< 0.01 (ANOVA)
;?< 0.01 (ANOVA)
;?< 0.01 (ANOVA)
p < 0.05 (ANOVA)
;?< 0.01 (ANOVA)
p< 0.01 (ANOVA)
p < 0.05 (ANOVA)
2-fold
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p< 0.01 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p< 0.01 (ANOVA)
p < 0.05 (ANOVA)
Reference
Plummeretal.,
2007
Plummeretal.,
2007
Plummeretal.,
2007
Liu et al., 2005
Plummeretal.,
2007
Plummeretal.,
2007
Liu et al., 2005
Shultzetal.,2001
Thompson et al.,
2005
Liu et al., 2005
Plummeretal.,
2007
Liu et al., 2005
Liu et al., 2005
Plummeretal.,
2007
Liu et al., 2005
-------
Table A-l (continued)
Official
gene
symbol
Hmgcsl
Hmgcsl
Hmoxl
Hpgd
Hprt
Hrasls3
Hsdllb2
Hsdl7b3
Hsdl7b7
Hsd3bl_
predicted
Hsd3bl_
predicted
Hspb?
Idhl
Idhl
Idil
Mil
Official gene namef
3-hydroxy-3-methylglutaryl-Coenzyme A
synthase 1
3-hydroxy-3-methylglutaryl-Coenzyme A
synthase 1
Heme oxygenase (decycling) 1
Hydroxyprostaglandin dehydrogenase 15 (NAD)
Hypoxanthine guanine phosphoribosyl transferase
HRAS like suppressor 3
Hydroxy steroid (11-beta) dehydrogenase 2
Hydroxysteroid (17-beta) dehydrogenase 3
Hydroxysteroid (17-beta) dehydrogenase 7
Hydroxysteroid dehydrogenase- 1, delta< 5 >-3-beta
(predicted)
Hsd3bl_predicted hydroxy steroid dehydrogenase- 1,
delta< 5 >-3-beta (predicted)
Heat shock 27kD protein family, member 7
(cardiovascular)
Isocitrate dehydrogenase 1 (NADP+), soluble
Isocitrate dehydrogenase 1 (NADP+), soluble
Isopentenyl-diphosphate delta isomerase
Isopentenyl-diphosphate delta isomerase
Exposure
window
CDs 12.5-17.5
CDs 12.5-19.5
CDs 12-19
CDs 12-19
CDs 12-19
CDs 12-19
GD 19 for 6 hr
CDs 12-19
CDs 12-19
CDs 12-19
GD 18 for 18 hr
CDs 12-19
CDs 12-19
GD 18 for 18 hr
CDs 12-19
CDs 12.5-17.5
Up or
down
Down
Down
Down
Down
Down at
GD 19
Down
Down after
6hr
Up
Down
Down
Down after
18 hr
Up
Down
Down after
18 hr
Down
Down
Fold change
-1.72
-1.87
-0.27 Iog2
-0.46 Iog2
->2
-0.45 Iog2
-1.16
0.28 Iog2
-0.32 Iog2
-0.50 Iog2
-0.7
0.41 Iog2
-0.52 Iog2
-0.67
-0.85 Iog2
-1.57
Cutoff used (method)
;?< 0.01 (ANOVA)
;?< 0.01 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
2-fold
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p< 0.01 (ANOVA)
Reference
Plummeretal.,
2007
Plummeretal.,
2007
Liu et al., 2005
Liu et al., 2005
Shultzetal.,2001
Liu et al., 2005
Thompson et al.,
2005
Liu et al., 2005
Liu et al., 2005
Liu et al., 2005
Thompson et al.,
2005
Liu et al., 2005
Liu et al., 2005
Thompson et al.,
2005
Liu et al., 2005
Plummeretal.,
2007
-------
Table A-l (continued)
Official
gene
symbol
Igfbp2
IgfbpS
Il6st
Ifitm2
Inha
Inha
Insigl
Insl3
interim
symbol:
Loc31432
3
interim
symbol:
Ratsg2
Kcnj'8
Khk
Kit
Krt2-8
Official gene namef
Insulin-like growth factor binding protein 2
Insulin-like growth factor binding protein 3
Interleukin 6 signal transducer
Interferon induced transmembrane protein 2
Inhibin alpha
Inhibin alpha
Insulin induced gene 1
Insulin-like 3
Interim full name: transporter
Interim name: Ratsg2
Potassium inwardly -rectify ing channel, subfamily J,
member 8
Ketohexokinase
V-kit Hardy -Zuckerman 4 feline sarcoma viral
oncogene homolog
Keratin complex 2, basic, gene 8
Exposure
window
CDs 12-19
CDs 12-21
CDs 12-21
CDs 12.5-17.5
CDs 12-19
CDs 12.5-19.5
CDs 12-19
CDs 12-19
CDs 12-19
CDs 12-19
CDs 12-21
CDs 12.5-17.5
CDs 12-21
CDs 12-19
Up or
down
Down
Up at
GD21
Down at
GD21
Down
Down
Down
Down
Down
Down
Down
Down at
GD21
Up
Down at
GD21
Up
Fold change
-0.39 Iog2
>2
->2
-1.11
-1.00 Iog2
-1.64
-0.77 Iog2
-1.56 Iog2
-0.35 Iog2
-0.13 Iog2
->2
1.30
->2
0.28 Iog2
Cutoff used (method)
p < 0.05 (ANOVA)
2-fold
2-fold
p< 0.01 (ANOVA)
p < 0.05 (ANOVA)
p< 0.01 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
2-fold
p< 0.01 (ANOVA)
2-fold
p < 0.05 (ANOVA)
Reference
Liu et al., 2005
Shultzetal.,2001
Shultzetal.,2001
Plummeretal.,
2007
Liu et al., 2005
Plummeretal.,
2007
Liu et al., 2005
Liu et al., 2005
Liu et al., 2005
Liu et al., 2005
Shultzetal.,2001
Plummeretal.,
2007
Shultzetal.,2001
Liu et al., 2005
-------
Table A-l (continued)
Official
gene
symbol
Ldha
Ldlr
Ldlr
Lhcgr
Lhcgr
Lhcgr
Lhcgr
Limkl
Lnk
Lr8
Lss
Mapkl
Marcks
Mdk
Mel
Official gene namef
Lactate dehydro-genase A
Low density lipoprotein receptor
Low density lipoprotein receptor
Luteinizing hormone/choriogonadotropin receptor
Luteinizing hormone/choriogonadotropin receptor
Luteinizing hormone/choriogonadotropin receptor
Luteinizing hormone/choriogonadotropin receptor
LIM motif-containing protein kinase 1
Linker of T-cell receptor pathways
LR8 protein
Lanosterol synthase
Mitogen activated protein kinase 1
Myristoylated alanine rich protein kinase C
substrate
Midkine
Malic enzyme 1, NADP(+) -dependent, cytosolic
Exposure
window
CDs 12.5-19.5
CDs 12-19
CDs 12-19
CDs 12-21
GD 19 for 6 hr
CDs 18-19 for
18 hr
CDs 12-19
CDs 12-21
GD 19 for 3 hr
CDs 12-19
CDs 12-19
CDs 12-21
CDs 12-19
CDs 12-19
CDs 12-19
Up or
down
Down
Down
Down at
GD19
Down at
GD21
Down after
6hr
Down after
18 hr
Down
Down at
GD21
Up after
3hr
Down
Down
Up at
GD21
Up at
GD 19
Up
Down
Fold change
-1.30
-0.79 Iog2
->2
->2
-1.00
-1.51
-1.39 Iog2
->2
1.17
-0.45 Iog2
-0.48 Iog2
>2
>2
0.20 Iog2
-0.67 Iog2
Cutoff used (method)
p< 0.01 (ANOVA)
p < 0.05 (ANOVA)
2-fold
2-fold
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
2-fold
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
2-fold
2-fold
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
Reference
Plummeretal.,
2007
Liu et al., 2005
Shultzetal.,2001
Shultzetal.,2001
Thompson et al.,
2005
Thompson et al.,
2005
Liu et al., 2005
Shultzetal.,2001
Thompson et al.,
2005
Liu et al., 2005
Liu et al., 2005
Shultzetal.,2001
Shultzetal.,2001
Liu et al., 2005
Liu et al., 2005
-------
Table A-l (continued)
Official
gene
symbol
Mel
Menl
Mgatl
Mgp
Mgstl
Mgstl
Mir 16
Mlxipl
Mmp2
Mtusl
Mtusl
Mvd
Mydll6
Myh6
Myh6
Official gene namef
Malic enzyme 1, NADP(+) -dependent, cytosolic
Multiple endocrine neoplasia 1
Mannoside acetylglucosaminyltransferase 1
Matrix Gla protein
Microsomal glutathione S-transferase 1
Microsomal glutathione S-transferase 1
Membrane interacting protein of RGS 16
MLX interacting protein-like
Matrix metallopeptidase 2
Mitochondrial tumor suppressor 1
Mitochondrial tumor suppressor 1
Mevalonate (diphospho) decarboxylase
Myeloid differentiation primary response gene 116
Myosin, heavy polypeptide 6, cardiac muscle, alpha
Myosin, heavy polypeptide 6, cardiac muscle, alpha
Exposure
window
CDs 12.5-17.5
CDs 12.5-15.5
CDs 12-19
GD19for6hr
CDs 12-19
CDs 12-21
CDs 12-19
CDs 12-19
CDs 12-21
GD 19 for 3 hr
GD 19 for 6 hr
CDs 12-19
GD 19 for 3 hr
CDs 12-19
CDs 18-19 for
18 hr
Up or
down
Down
Down
Up
Up after
6hr
Down
Up at
GD21
Down
Down
Up at
GD21
Up after
3hr
Up after
6hr
Down
Up after
3hr
Down
Down after
18 hr
Fold change
-1.36
-1.17
0.28 Iog2
1.66
-0.36 Iog2
>2
-0.56 Iog2
-0.311og2
>2
0.67
0.55
-0.41 Iog2
0.58
-0.72 Iog2
-1.52
Cutoff used (method)
;?< 0.01 (ANOVA)
;?< 0.01 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
2-fold
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
2-fold
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
Reference
Plummeretal.,
2007
Plummeretal.,
2007
Liu et al., 2005
Thompson et al.,
2005
Liu et al., 2005
Shultzetal.,2001
Liu et al., 2005
Liu et al., 2005
Shultzetal.,2001
Thompson et al.,
2005
Thompson et al.,
2005
Liu et al., 2005
Thompson et al.,
2005
Liu et al., 2005
Thompson et al.,
2005
-------
Table A-l (continued)
Official
gene
symbol
Myh6
Myom2
Myrip
Nalp6
Nexn
Nfl
Nfil3
Nfkbia
Npc2
Nppc
NrObl
NrObl
Nr4al
Nr4al
Nr4a3
NrSal
Official gene namef
Myosin, heavy polypeptide 6, cardiac muscle, alpha
Myomesin 2
Myosin VIIA and Rab interacting protein
NACHT, leucine rich repeat and PYD containing 6
Nexilin
Neurofibromatosis 1
Nuclear factor, interleukin 3 regulated
Nuclear factor of kappa light chain gene enhancer
in B -cells inhibitor, alpha
Niemann pick type C2
Natriuretic peptide precursor type C
Nuclear receptor subfamily 0, group B, member 1
Nuclear receptor subfamily 0, group B, member 1
Nuclear receptor subfamily 4, group A, member 1
Nuclear receptor subfamily 4, group A, member 1
Nuclear receptor subfamily 4, group A, member 3
Nr5al nuclear receptor subfamily 5, group A,
member 1
Exposure
window
CDs 12.5-19.5
CDs 12-19
CDs 12-19
CDs 12-19
CDs 12-19
CDs 12-21
CDs 12-19
GD 19 for 3 hr
CDs 12-19
CDs 12-19
CDs 12-19
CDs 12.5-19.5
CDs 12-19
GD19for3hr
GD19for3hr
CDs 12.5-19.5
Up or
down
Down
Up
Down
Up
Up
Down at
GD21
Up
Up after
3hr
Down
Down
Down
Down
Up
Up after
3hr
Up after
3hr
Down
Fold change
-1.64
0.64 Iog2
-0.27 Iog2
0.45 Iog2
0.26 Iog2
->2
0.311og2
0.79
-0.26 Iog2
-0.56 Iog2
-0.37 Iog2
-1.15
0.3 Iog2
1.83
2.25
-1.18
Cutoff used (method)
p< 0.01 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
2-fold
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p< 0.01 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p< 0.01 (ANOVA)
Reference
Plummeretal.,
2007
Liu et al., 2005
Liu et al., 2005
Liu et al., 2005
Liu et al., 2005
Shultzetal.,2001
Liu et al., 2005
Thompson et al.,
2005
Liu et al., 2005
Liu et al., 2005
Liu et al., 2005
Plummeretal.,
2007
Liu et al., 2005
Thompson et al.,
2005
Thompson et al.,
2005
Plummeretal.,
2007
oo
-------
Table A-l (continued)
Official
gene
symbol
Ntf3
Okl38
Olfml
P2ryl4
Park?
Pcnvr
Pcna
Pcyt2
Pdapl
Pdyn
Pebpl
Pebpl
Penkl
Penkl
PJkp
Official gene namef
Neurotrophin 3
Pregnancy -induced growth inhibitor
Olfactomedin 1
Purinergic receptor P2Y, G-protein coupled, 14
Parkinson disease (autosomal recessive, early
onset) 7
PRKC, apoptosis, WT1, regulator
Proliferating cell nuclear antigen
Phosphate cytidylyltransferase 2, ethanolamine
PDGFA associated protein 1
Prodynorphin
Phosphatidylethanolamine binding protein 1
Phosphatidylethanolamine binding protein 1
Proenkephalin 1
Proenkephalin 1
Phosphofructokinase, platelet
Exposure
window
CDs 12.5-17.5
CDs 12-19
CDs 12-19
CDs 12-19
CDs 12.5-17.5
GD 19 for 3 hr
CDs 12-21
CDs 12-19
CDs 12-21
CDs 12-19
CDs 12-19
CDs 12.5-19.5
CDs 12.5-17.5
CDs 12.5-19.5
CDs 12.5-19.5
Up or
down
Up
Down
Down
Down
Down
Up after
3hr
Up at
GD21
Down
Up at
GD21
Down
Down
Down
Down
Down
Down
Fold change
1.34
-0.33 Iog2
-0.141og2
-0.37 Iog2
-1.32
1.02
>2
-0.20 Iog2
>2
-1.06 Iog2
-0.36 Iog2
-1.67
-1.41
-1.86
-1.41
Cutoff used (method)
;?< 0.01 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p< 0.01 (ANOVA)
p < 0.05 (ANOVA)
2-fold
p < 0.05 (ANOVA)
2-fold
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p< 0.01 (ANOVA)
p< 0.01 (ANOVA)
p< 0.01 (ANOVA)
p< 0.01 (ANOVA)
Reference
Plummeretal.,
2007
Liu et al., 2005
Liu et al., 2005
Liu et al., 2005
Plummeretal.,
2007
Thompson et al.,
2005
Shultzetal.,2001
Liu et al., 2005
Shultzetal.,2001
Liu et al., 2005
Liu et al., 2005
Plummeretal.,
2007
Plummeretal.,
2007
Plummeretal.,
2007
Plummeretal.,
2007
-------
Table A-l (continued)
Official
gene
symbol
Pgaml
Pgkl
Phb
Phb
Phyh
Plat
Plaur
Pmp22
Pmp22
Pmp22
Pnliprp2
For
For
Ppib
Official gene namef
Phosphoglycerate mutase 1
Phosphoglycerate kinase 1
Prohibitin
Prohibitin
Phytanoyl-CoA hydroxylase
Plasminogen activator, tissue
Plasminogen activator, urokinase receptor
Peripheral myelin protein 22
Peripheral myelin protein 22
Peripheral myelin protein 22
Pancreatic lipase-related protein 2
P450 (cytochrome) oxidoreductase
P450 (cytochrome) oxidoreductase
Peptidylprolyl isomerase B
Exposure
window
CDs 12.5-19.5
CDs 12.5-19.5
CDs 12-21
CDs 12-19
GD19for6hr
CDs 12-19
GD 19 for 3 hr
CDs 12-19
GD 19 for 3 hr
GD 19 for 6 hr
CDs 12-19
CDs 12-19
CDs 12.5-19.5
CDs 12.5-17.5
Up or
down
Down
Down
Down at
GD21
Down at
GD19
Down after
6hr
Up at
GD19
Up after
3hr
Up at
GD 19
Down after
3hr
Down after
6hr
Down
Down
Down
Down
Fold change
-1.26
-1.25
->2
->2
-1.02
>2
0.86
>2
-0.75
-0.59
-0.28 Iog2
-0.64 Iog2
-1.39
-1.21
Cutoff used (method)
;?< 0.01 (ANOVA)
;?< 0.01 (ANOVA)
2-fold
2-fold
p < 0.05 (ANOVA)
2-fold
p < 0.05 (ANOVA)
2-fold
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p< 0.01 (ANOVA)
p< 0.01 (ANOVA)
Reference
Plummeretal.,
2007
Plummeretal.,
2007
Shultzetal.,2001
Shultzetal.,2001
Thompson et al.,
2005
Shultzetal.,2001
Thompson et al.,
2005
Shultzetal.,2001
Thompson et al.,
2005
Thompson et al.,
2005
Liu et al., 2005
Liu et al., 2005
Plummeretal.,
2007
Plummeretal.,
2007
to
o
-------
Table A-l (continued)
Official
gene
symbol
Ppplcb
Prdx3
Prdx3
Prdx3
Prgl
Prkar2b
Prkcbpl
Prlr
Ptma
Ptp4al
PVR
PVR
Rabep2
Rasdl
Rlnl
Official gene namef
Protein phosphatase 1, catalytic subunit,
beta isoform
Peroxiredoxin 3
Peroxiredoxin 3
Peroxiredoxin 3
Plasticity related gene 1
Protein kinase, cAMP dependent regulatory,
type II beta
Protein kinase C binding protein 1
Prolactin receptor
Prothymosin alpha
Protein tyrosine phosphatase 4al
Poliovirus receptor
Poliovirus receptor
Rabaptin, RAB GTPase binding effector protein 2
RAS, dexamethasone-induced 1
Relaxin 1
Exposure
window
CDs 12.5-17.5
CDs 12-19
CDs 18-19 for
18 hr
CDs 12.5-19.5
CDs 12-19
CDs 12-19
CDs 12-19
CDs 12-19
CDs 12-19
CDs 12-21
GD 19 for 3 hr
GD 19 for 6 hr
GD 19 for 3 hr
CDs 12-19
CDs 12-19
Up or
down
Down
Down
Down after
18 hr
Down
Down
Down
Up
Down
Down at
GD 19
Up at
GD21
Up after
3hr
Up after
6hr
Down after
3hr
Down
Down
Fold change
-1.37
-0.53 Iog2
-0.86
-1.63
-0.97 Iog2
-0.33 Iog2
0.32 Iog2
-1.02 Iog2
->2
>2
1.26
0.92
-0.48
-0.52 Iog2
-0.36 Iog2
Cutoff used (method)
;?< 0.01 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p< 0.01 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
2-fold
2-fold
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
Reference
Plummeretal.,
2007
Liu et al., 2005
Thompson et al.,
2005
Plummeretal.,
2007
Liu et al., 2005
Liu et al., 2005
Liu et al., 2005
Liu et al., 2005
Shultzetal.,2001
Shultzetal.,2001
Thompson et al.,
2005
Thompson et al.,
2005
Thompson et al.,
2005
Liu et al., 2005
Liu et al., 2005
>
-------
Table A-l (continued)
Official
gene
symbol
Rnhl
Rpa2
Rpll3
Rpl32
Rpl37
Rpl36a
Rpl36a
Rpn2
Rpsl3
Rpsl7
Rpsl9
Rps29
Sc4mol
Official gene namef
Ribonuclease/angiogenin inhibitor 1
Replication protein A2
Ribosomal protein L 13
Ribosomal protein L32
Ribosomal protein L37
Large subunit ribosomal protein L36a
Large subunit ribosomal protein L36a
Ribophorin II
Ribosomal protein S 13
Ribosomal protein S 17
Ribosomal protein S 19
Ribosomal protein S29
Sterol-C4-methyl oxidase-like
Exposure
window
CDs 12.5-17.5
CDs 12-21
CDs 12.5-15.5
CDs 12.5-19.5
CDs 12.5-19.5
CDs 12-19
CDs 12.5-15.5
CDs 12.5-19.5
CDs 12.5-15.5
CDs 12.5-19.5
CDs 12.5-17.5
CDs 12.5-19.5
CDs 12-19
Up or
down
Down
Down at
GD21
Up
Up
Up
Down at
GD19
Up
Down
Up
Up
Up
Down
Down
Fold change
-1.20
->2
1.17
1.13
1.13
->2
1.22
-1.19
1.30
1.25
1.25
-1.13
-1.02 Iog2
Cutoff used (method)
^<0.01(ANOVA)
2-fold
^<0.01(ANOVA)
^<0.01(ANOVA)
^<0.01(ANOVA)
2-fold
^<0.01(ANOVA)
^<0.01(ANOVA)
^<0.01(ANOVA)
^<0.01(ANOVA)
^<0.01(ANOVA)
^<0.01(ANOVA)
p < 0.05 (ANOVA)
Reference
Plummeretal.,
2007
Shultzetal.,2001
Plummeretal.,
2007
Plummeretal.,
2007
Plummeretal.,
2007
Shultzetal.,2001
Plummeretal.,
2007
Plummeretal.,
2007
Plummeretal.,
2007
Plummeretal.,
2007
Plummeretal.,
2007
Plummeretal.,
2007
Liu et al., 2005
to
to
-------
Table A-l (continued)
Official
gene
symbol
Sc4mol
Sc4mol
ScSd
Scarbl
Scarbl
Scarbl
Scarbl
Scarbl
Scarbl
Scdl
Scn3b
Scp2
Scp2
Sdf4
Seppl
Official gene namef
Sterol-C4-methyl oxidase-like
Sterol-C4-methyl oxidase-like
Sterol-C5-desaturase (fungal ERGS,
delta-5-desaturase) homolog (S. cerevisae)
Scavenger receptor class B, member 1
Scavenger receptor class B, member 1
Scavenger receptor class B, member 1
Scavenger receptor class B, member 1
Scavenger receptor class B, member 1
Scavenger receptor class B, member 1
Stearoyl-Coenzyme A desaturase 1
Sodium channel, voltage-gated, type III, beta
Sterol carrier protein 2
Sterol carrier protein 2
Stromal cell derived factor 4
Selenoprotein P, plasma, 1
Exposure
window
CDs 12.5-17.5
CDs 12.5-19.5
CDs 12-19
CDs 12-19
CDs 12-19
GD 19 for 6 hr
CDs 18-19 for
18 hr
CDs 12.5-17.5
CDs 12.5-19.5
CDs 12-19
GD 19 for 6 hr
CDs 12-19
CDs 12.5-19.5
CDs 12-19
CDs 12-19
Up or
down
Down
Down
Down
Down
Down at
GD19
Down after
6hr
Down after
18 hr
Down
Down
Down
Up after
6hr
Down
Down
Down
Down
Fold change
-1.82
-2.36
-0.32 Iog2
-1.91 Iog2
->2
-1.60
-2.72
-2.23
-2.85
-0.58 Iog2
1.49
-0.171og2
-1.24
-0.27 Iog2
-0.45 Iog2
Cutoff used (method)
;?< 0.01 (ANOVA)
;?< 0.01 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
2-fold
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p< 0.01 (ANOVA)
p< 0.01 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p< 0.01 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
Reference
Plummeretal.,
2007
Plummeretal.,
2007
Liu et al., 2005
Liu et al., 2005
Shultzetal.,2001
Thompson et al.,
2005
Thompson et al.,
2005
Plummeretal.,
2007
Plummeretal.,
2007
Liu et al., 2005
Thompson et al.,
2005
Liu et al., 2005
Plummeretal.,
2007
Liu et al., 2005
Liu et al., 2005
to
-------
Table A-l (continued)
Official
gene
symbol
Serpinel
Serpinhl
Sgk
Slc3a2
Slcl2a2
Slcl6a6
Slc25al
Slc25a20
Slc7a8
Slc7a8
Smpx
Sod2
Sod3
Sqle
Sqle
Ssr4
Official gene namef
Serine (or cysteine) peptidase inhibitor, clade E,
member 1
Serine (or cysteine) peptidase inhibitor, clade H,
member 1
Serum/glucocorticoid regulated kinase
Solute carrier family 3 (activators of dibasic and
neutral amino acid transport), member 2
Solute carrier family 12 (sodium/potassium/chloride
transporters), member 2
Solute carrier family 16 (monocarboxylic acid
transporters), member 6
Solute carrier family 25, member 1
Solute carrier family 25 (mitochondria!
carnitine/acylcarnitine translocase), member 20
Solute carrier family 7 (cationic amino acid
transporter, y+ system), member 8
Solute carrier family 7 (cationic amino acid
transporter, y+ system), member 8
Small muscle protein, X-linked
Superoxide dismutase 2, mitochondrial
Superoxide dismutase 3, extracellular
Squalene epoxidase
Squalene epoxidase
Signal sequence receptor 4
Exposure
window
GD19for3hr
CDs 12.5-15.5
CDs 12-19
CDs 12-19
CDs 12.5-17.5
CDs 12-19
CDs 12-19
CDs 12-19
CDs 12.5-17.5
CDs 12.5-19.5
CDs 12-19
CDs 12-19
CDs 12-19
CDs 12-19
GD 18 for 18 hr
CDs 12-19
Up or
down
Up after
3hr
Down
Down
Down
Down
Down
Down
Down
Down
Down
Up
Down
Down
Down
Down after
18 hr
Down
Fold change
1.56
-1.32
-0.45 Iog2
-0.48 Iog2
-1.39
-0.38 Iog2
-0.27 Iog2
-0.23 Iog2
-1.82
-2.18
0.21 Iog2
-0.51 Iog2
-0.33 Iog2
-0.59 Iog2
-1.26
-0.23 Iog2
Cutoff used (method)
p < 0.05 (ANOVA)
;?< 0.01 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p< 0.01 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p< 0.01 (ANOVA)
p< 0.01 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
Reference
Thompson et al.,
2005
Plummeretal.,
2007
Liu et al., 2005
Liu et al., 2005
Plummeretal.,
2007
Liu et al., 2005
Liu et al., 2005
Liu et al., 2005
Plummeretal.,
2007
Plummeretal.,
2007
Liu et al., 2005
Liu et al., 2005
Liu et al., 2005
Liu et al., 2005
Thompson et al.,
2005
Liu et al., 2005
to
-------
Table A-l (continued)
Official
gene
symbol
Ssrpl
Star
Star
Star
Star
Stcl
Stcl
Stc2
Stc2
Sts
Suclgl
SvsS
SvsS
SvsS
SvsS
Official gene namef
Structure specific recognition protein 1
Steroidogenic acute regulatory protein
Steroidogenic acute regulatory protein
Steroidogenic acute regulatory protein
Steroidogenic acute regulatory protein
Stanniocalcin 1
Stanniocalcin 1
Stanniocalcin 2
Stanniocalcin 2
Steroid sulfatase
Succinate-CoA ligase, GDP -forming, alpha subunit
Seminal vesicle secretion 5
Seminal vesicle secretion 5
Seminal vesicle secretion 5
Seminal vesicle secretion 5
Exposure
window
CDs 12-19
CDs 12-19
CDs 18-19 for
18 hr
CDs 12.5-17.5
CDs 12.5-19.5
CDs 12-19
GD 19 for 6 hr
CDs 12-19
CDs 12.5-19.5
CDs 12-19
CDs 12.5-19.5
CDs 12-19
CDs 18-19 for
18 hr
CDs 12.5-17.5
CDs 12.5-19.5
Up or
down
Down at
GD19
Down
Down after
18 hr
Down
Down
Up
Up after
6hr
Down
Down
Down at
GD 19
Down
Down
Down after
18 hr
Down
Down
Fold change
->2
-2.45 Iog2
-2.33
-2.19
-2.53
0.98 Iog2
1.61
-1.181og2
-1.59
->2
-1.21
-3.75 Iog2
-3.36
-5.89
-3.75
Cutoff used (method)
2-fold
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p< 0.01 (ANOVA)
p< 0.01 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p< 0.01 (ANOVA)
2-fold
p< 0.01 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p< 0.01 (ANOVA)
p< 0.01 (ANOVA)
Reference
Shultzetal.,2001
Liu et al., 2005
Thompson et al.,
2005
Plummer et al.,
2007
Plummer et al.,
2007
Liu et al., 2005
Thompson et al.,
2005
Liu et al., 2005
Plummer et al.,
2007
Shultzetal.,2001
Plummer et al.,
2007
Liu et al., 2005
Thompson et al.,
2005
Plummer et al.,
2007
Plummer etal.,
2007
to
-------
Table A-l (continued)
Official
gene
symbol
Syngrl
Tcfl
Tc/21
Tec
Testin
Tfrc
TgfbS
Timpl
TimpS
Tkt
Tkt
TmedlO
Tnfrsfl2a
Tnnil
TnniS
Tnnt2
Official gene namef
Synaptogyrin 1
Transcription factor 1
Transcription factor 2 1
Tec protein tyrosine kinase
Testin gene
Transferrin receptor
Transforming growth factor, beta 3
Tissue inhibitor of metallopeptidase 1
Tissue inhibitor of metalloproteinase 3 (Sorsby
fundus dystrophy, pseudoinflammatory)
Transketolase
Transketolase
Transmembrane emp24-like trafficking
protein 10 (yeast)
Tumor necrosis factor receptor superfamily,
member 12a
Troponin I, skeletal, slow 1
Troponin I type 3 (cardiac)
Troponin T2, cardiac
Exposure
window
CDs 12-19
CDs 12-19
CDs 12-19
GD 19 for 3 hr
CDs 12-19
CDs 12-19
CDs 12-19
GD 19 for 6 hr
CDs 12-21
CDs 12.5-17.5
CDs 12.5-19.5
CDs 12.5-19.5
GD 19 for 6 hr
CDs 12-19
CDs 12-19
CDs 12-19
Up or
down
Down
Down
Up
Up after
3hr
Up
Down
Down at
GD 19
Up after
6hr
Down at
GD21
Down
Down
Down
Up after
6hr
Up
Up
Up
Fold change
-0.161og2
-0.141og2
0.171og2
0.69
0.59 Iog2
-0.23 Iog2
->2
1.04
->2
-1.19
-1.28
-1.20
1.34
0.33 Iog2
0.26 Iog2
0.77 Iog2
Cutoff used (method)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
2-fold
p < 0.05 (ANOVA)
2-fold
p< 0.01 (ANOVA)
p< 0.01 (ANOVA)
p< 0.01 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
Reference
Liu et al., 2005
Liu et al., 2005
Liu et al., 2005
Thompson et al.,
2005
Liu et al., 2005
Liu et al., 2005
Shultzetal.,2001
Thompson et al.,
2005
Shultzetal.,2001
Plummer et al.,
2007
Plummer et al.,
2007
Plummer et al.,
2007
Thompson et al.,
2005
Liu et al., 2005
Liu et al., 2005
Liu et al., 2005
to
a\
-------
Table A-l (continued)
Official
gene
symbol
Tpil
Tpml
Tpml
Tsc22dl
Tsn
Tst
Tubal
Tubal
Txnl
Txnll
Uba52
UncSb
Vapa
Vcaml
Vdacl
Official gene namef
Triosephosphate isomerase 1
Tropomyosin 1, alpha
Tropomyosin 1, alpha
TSC22 domain family, member 1
Translin
Thiosulfate sulfurtransferase
Tubulin, alpha 1
Tubulin, alpha 1
Thioredoxin 1
Thioredoxin-like 1
Ubiquitin A-52 residue ribosomal protein fusion
product 1
Unc-5 homolog B (C. elegans)
VAMP (vesicle-associated membrane protein)-
associated protein A
Vascular cell adhesion molecule 1
Voltage-dependent anion channel 1
Exposure
window
CDs 12-19
CDs 12-19
GD19for6hr
CDs 12.5-19.5
CDs 12.5-17.5
CDs 12-19
CDs 12-21
CDs 12.5-19.5
GD 18 for 18 hr
CDs 12.5-15.5
CDs 12.5-19.5
CDs 12-21
CDs 12.5-19.5
CDs 12-19
CDs 12.5-19.5
Up or
down
Down
Up
Up after
6hr
Down
Up
Down
Down at
GD21
Down
Down after
18 hr
Up
Up
Down at
GD21
Down
Down
Down
Fold change
-0.24 Iog2
0.36 Iog2
1.04
-1.34
1.54
-0.33 Iog2
->2
-1.26
-0.62
1.20
1.10
->2
-1.37
-0.63 Iog2
-1.13
Cutoff used (method)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p< 0.01 (ANOVA)
p< 0.01 (ANOVA)
p < 0.05 (ANOVA)
2-fold
p< 0.01 (ANOVA)
p < 0.05 (ANOVA)
p< 0.01 (ANOVA)
p< 0.01 (ANOVA)
2-fold
p< 0.01 (ANOVA)
p < 0.05 (ANOVA)
p< 0.01 (ANOVA)
Reference
Liu et al., 2005
Liu et al., 2005
Thompson et al.,
2005
Plummer et al.,
2007
Plummer et al.,
2007
Liu et al., 2005
Shultzetal.,2001
Plummer et al.,
2007
Thompson et al.,
2005
Plummer et al.,
2007
Plummer et al.,
2007
Shultzetal.,2001
Plummer et al.,
2007
Liu et al., 2005
Plummer etal.,
2007
to
-------
Table A-l (continued)
Official
gene
symbol
Vim
Vnnll
Vsnll
Ywhae
Zfp36
Zyx
Not found
Not found
Not found
Not found
Official gene namef
Vimentin
Vanin 1
Visinin-like 1
Tyrosine 3-monooxygenase/tryptophan
5-monooxygenase activation protein, epsilon
polypeptide
Zinc finger protein 36
Zyxin
Listed as "Tppc" and 289920_Rn in article, and
Genbank #BF400584 (Plummer, personal
communication) does not match a gene name.
Listed as "Similar to mouse IAP -binding protein"
and 2055 10 Rn in article, and Genbank
#:BG378907 (Plummer, personal communication)
does not match a gene name.
LOC499942 similar to WAP four-disulfide core
domain protein 8 precursor (Putative protease
inhibitor WAP8) (Rattus norvegicus).
LOC497726 hypothetical gene supported by
NM_138518 (Rattus norvegicus). This record was
discontinued.
Exposure
window
CDs 12.5-19.5
CDs 12-19
CDs 12-19
CDs 12.5-19.5
GD 19 for 1 hr
GD 19 for 3 hr
CDs 12.5-17.5
CDs 12.5-15.5
CDs 12-19
CDs 12-19
Up or
down
Down
Down
Down
Down
Up after
Ihr
Up after
3hr
Down
Up
Down
Down
Fold change
-1.60
-0.32 Iog2
-0.62 Iog2
-1.37
1.79
1.03
-1.39
1.26
-0.25 Iog2
-0.27 Iog2
Cutoff used (method)
;?< 0.01 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p< 0.01 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
p< 0.01 (ANOVA)
p< 0.01 (ANOVA)
p < 0.05 (ANOVA)
p < 0.05 (ANOVA)
Reference
Plummer etal.,
2007
Liu et al., 2005
Liu et al., 2005
Plummer etal.,
2007
Thompson et al.,
2005
Thompson et al.,
2005
Plummer etal.,
2007
Plummer etal.,
2007
Liu et al., 2005
Liu et al., 2005
to
oo
aThe four studies dosed at 500 mg/kg-d DBF in the Sprague-Dawley (SD) rat.
bThompson et al. (2005) and Shultz et al. (2001) dosed with DBF alone; gene expression changes for DBF were relative to vehicle control expression.
-------
Table A-l. (continued)
°Liu et al. (2005) presented microarray data for all five developmental phthalates, including DBF, since they did not find any differences in statistical significance among
the five phthalates. Thus, we present the data for all five phthalates, which should be the same as for DBF.
dThe Affy ID 1387057_at was found to be significantly down-regulated by Liu et al. (2005). This Affy ID was listed as the gene Slc7a8 (solute carrier family 7 [cationic
amino acid transporter, y+ system], member 8) at the time of their publication. As of January 2007, Affy now lists both Slc7a8 and Syngapl. This probeset is apparently
capable of hybridizing with two different genes. Thus, this Affy ID was not incorporated in the case study evaluation since it is not clear which gene was altered after
DBF inutero exposure.
eThe Plummer et al. (2007) data from the whole testis are included in this table. The data from microdissection of testicular regions are not presented since no other
studies were comparable. Plummer et al. (2007) performed their study in the Wistar rat whereas the other three microarray studies were performed in the SD rat.
fGene function and pathway information was gathered from GeneGo (www.genego.com).
ANOVA, analysis of variance; GD, gestation day; hr, hour.
to
-------
Table A-2. WOE for statistically significant gene expression changes after in utero exposure to DBF from
whole-rat testis reverse transcription-polymerase chain reaction (RT-PCR) studies
Official
gene
symbol
Ar
Bmp4
Btg2
Bzrp
Cebpb
Cebpd
Clu
Clu
Clu
Cxcll
Official gene name*
Androgen receptor
Bone morphogenetic protein 4
B-cell translocation gene 2,
anti-proliferative
Benzodiazepine receptor,
peripheral
CCAAT/enhancer binding
protein (C/EBP), beta
CCAAT/enhancer binding
protein (C/EBP), delta
Clusterin
Clusterin
Clusterin
Chemokine (C-X-C motif)
ligand 1
Dose
500 mg/kg-d
500 mg/kg-d
500 mg/kg-d
500 mg/kg-d
500 mg/kg-d
500 mg/kg-d
500 mg/kg-d
500 mg/kg-d
500 mg/kg-d
500 mg/kg-d
Exposure window
CDs 12-19
CDs 12-19
GD 19for30minto6hr
timepoints and GD 18 for
12, 18, and 24 hr time
points
CDs 12-19
CDs 12-19
GD 19for30minto6hr
timepoints and GD 18 for
12, 18, and 24 hr time
points
CDs 12-19
CDs 12-19
CDs 12-16, 12-19, or
12-21
GD 19for30minto6hr
timepoints and GD 18 for
12, 18, and 24 hr time
points
Up or down
Up
Up
Up after
-1-6 hr
(peak ~2 hr)
Up
Down
Up after
-1-6 hr
(peak ~3 hr)
Up
Up
Up
Up after
-1-12 hr
(peak at
~3hr)
Statistical analysis method
t-test,^<0.05
t-test,^<0.05
Relative expression determined
using mean Ct; triplicate
samples; GADPH control; SE;
but/) value not calculated
Dunnett's test, ANOVA (one
way), p< 0.05
One way and two-way nested
ANOVA;^<0.05
Relative expression determined
using mean Ct; triplicate
samples; GADPH control; SE;
but/) value not calculated
ANOVA, nested design,
^<0.05
Dunnett's test, ANOVA (one
way), p< 0.05
^<0.05
Relative expression determined
using mean Ct; triplicate
samples; GADPH control; SE;
but/) value not calculated
Reference
Bowman et
al., 2005
Bowman et
al., 2005
Thompson
et al., 2005
Lehmann et
al., 2004
Liuetal.,
2005
Thompson
et al., 2005
Barlow et
al., 2003
Lehmann et
al., 2004
Shultz et
al., 2001
Thompson
et al., 2005
oo
o
-------
Table A-2. (continued)
Official
gene symbol
Cypllal
Cypllal
Cypllal
Cypllal
Cypllal
Cypllal
Cypllal
Cypl7al
Cypl7al
Official gene name*
Cytochrome P450,
family 11, subfamily a,
polypeptide 1
Cytochrome P450,
family 11, subfamily a,
polypeptide 1
Cytochrome P450,
family 11, subfamily a,
polypeptide 1
Cytochrome P450,
family 11, subfamily a,
polypeptide 1
Cytochrome P450,
family 11, subfamily a,
polypeptide 1
Cytochrome P450,
family 11, subfamily a,
polypeptide 1
Cytochrome P450,
family 11, subfamily a,
polypeptide 1
Cytochrome P450,
family 17, subfamily a,
polypeptide 1
Cytochrome P450,
family 17, subfamily a,
polypeptide 1
Dose
500 mg/kg-d
50 mg/kg-d
100 mg/kg-d
500 mg/kg-d
500 mg/kg-d
500 mg/kg-d
500 mg/kg-d
500 mg/kg-d
500 mg/kg-d
Exposure window
CDs 12-19
CDs 12-19
CDs 12-19
CDs 12-19
CDs 12-16, 12-19, or
12-21
CDs 12-17 and 12-18
CDs 12.5-19.5
CDs 12-19
CDs 12-19
Up or down
Down
Down
Down
Down
Down
Down at
GD 18
Down
Down
Down
Statistical analysis method
ANOVA, nested design,
^<0.05
Dunnett's test, ANOVA (one
way), p< 0.05
Dunnett's test, ANOVA (one
way), p< 0.05
Dunnett's test, ANOVA (one
way), p< 0.05
^<0.05
t-test, ANOVA (one-way) with
Tukey post hoc analysis;
^<0.05
One-way ANOVA followed by
Bonferroni post test using
GraphPad Prism; p < 0.05
Repeated measure ANOVA,
nested design, p < 0.05
Dunnett's test, ANOVA (one
way), p< 0.05
Reference
Barlow et
al, 2003
Lehmann et
al., 2004
Lehmann et
al., 2004
Lehmann et
al., 2004
Shultz et
al., 2001
Thompson
et al., 2004
Plummer et
al., 2007
Barlow et
al., 2003
Lehmann et
al., 2004
>
-------
Table A-2. (continued)
Official
gene symbol
Cypl7al
Cypl7al
Dafl
Ddc
Dusp6
Edg3
Egfr
Egrl
Egrl
Official gene name
Cytochrome P450,
family 17, subfamily a,
polypeptide 1
Cytochrome P450,
family 17, subfamily a,
polypeptide 1
Decay accelerating factor 1
Dopa decarboxylase
Dual specificity
phosphatase 6
Endothelial differentiation
sphingolipid
G-protein-coupled
receptor 3
Epidermal growth factor
receptor
Early growth response 1
Early growth response 1
Dose
500 mg/kg-d
500 mg/kg-d
500 mg/kg-d
500 mg/kg-d
500 mg/kg-d
500 mg/kg-d
500 mg/kg-d
500 mg/kg-d
500 mg/kg-d
Exposure window
CDs 12-16, 12-19, or
12-21
CDs 12-17 and 12-18
CDs 12-19
CDs 12-19
GD 19for30minto6hr
timepoints and GD 18 for
12, 18, and 24 hr time
points
GD 19for30minto6hr
timepoints and GD 18 for
12, 18, and 24 hr time
points
CDs 12-19 and 12-21
GD 19for30minto6hr
timepoints and GD 18 for
12, 18, and 24 hr time
points
CDs 12-19
Up or down
Down at
GD 19
Down at
GD 17 and
18
Up
Down
Up after
-1-12 hr
(peak at ~3
hr)
Up after
-1-6 and
18 hr (peak
~3hr)
Un-changed
Up after
-1-7 hr
(peak -2 hr)
Up
Statistical analysis method
^<0.05
t-test, ANOVA (one-way) with
Tukey post hoc analysis;
^<0.05
One way and two-way nested
ANOVA,^<0.05
One way and two-way nested
ANOVA,^<0.05
Relative expression determined
using mean Ct; triplicate
samples; GADPH control; SE;
but/) value not calculated
Relative expression determined
using mean Ct; triplicate
samples; GADPH control; SE;
but/) value not calculated
t-test, p< 0.05
Relative expression determined
using mean Ct; triplicate
samples; GADPH control; SE;
but/) value not calculated
One way and two-way nested
ANOVA;^<0.05
Reference
Shultz et
al., 2001
Thompson
et al., 2004
Liuetal.,
2005
Liuetal.,
2005
Thompson
et al., 2005
Thompson
et al., 2005
Bowman et
al., 2005
Thompson
et al., 2005
Liuetal.,
2005
oo
to
-------
Table A-2. (continued)
Official
gene symbol
Egr2
FgflO
Fgfr2
Fos
Grbl4
Hes6
Hsdl7b3
Hsdl7b7
Hsd3bl_
predicted
Official gene name
Early growth response 2
Fibroblast growth factor 10
Fibroblast growth factor
receptor 2
FBJ murine osteosarcoma
viral oncogene homolog
Growth factor receptor
bound protein 14
Hairy and enhancer of
split 6 (Drosophila)
Hydroxysteroid (17-beta)
dehydrogenase 3
Hydroxysteroid (17-beta)
dehydrogenase 7
Hydroxysteroid
dehydrogenase- 1,
delta< 5 >-3-beta
(predicted)
Dose
500 mg/kg-d
500 mg/kg-d
500 mg/kg-d
500 mg/kg-d
500 mg/kg-d
500 mg/kg-d
500 mg/kg-d
500 mg/kg-d
500 mg/kg-d
Exposure window
GD 19for30minto6hr
timepoints and GD 18 for
12, 18, and 24 hr time
points
CDs 12-21
CDs 12-19 and 12-21
GD 19for30minto6hr
timepoints and GD 18 for
12, 18, and 24 hr time
points
CDs 12-19
GD 19for30minto6hr
timepoints and GD 18 for
12, 18, and 24 hr time
points
CDs 12-19
CDs 12-19
CDs 12-19
Up or down
Up after
-1-12 hr
(peak ~2 hr)
Up
No stat.
change
Up after
30 min and
6 hr (peak at
Ihr)
Up
Down after
1-3 hr (peak
at3hr)
Up
Down
Down
Statistical analysis method
Relative expression determined
using mean Ct; triplicate
samples; GADPH control; SE;
but/) value not calculated
t-test,^<0.05
t-test,^<0.05
Relative expression determined
using mean Ct; triplicate
samples; GADPH control; SE;
but/) value not calculated
One way and two-way nested
ANOVA,^<0.05
Relative expression determined
using mean Ct; triplicate
samples; GADPH control; SE;
but/) value not calculated
One way and two-way nested
ANOVA,^<0.05
One way and two-way nested
ANOVA,^<0.05
ANOVA, nested design,
^<0.05
Reference
Thompson
et al., 2005
Bowman et
al., 2005
Bowman et
al., 2005
Thompson
et al., 2005
Liuetal.,
2005
Thompson
et al., 2005
Liuetal.,
2005
Liuetal.,
2005
Barlow et
al., 2003
-------
Table A-2. (continued)
Official
gene symbol
Hsd3bl_
predicted
Hsd3bl_
predicted
Hsd3bl_
predicted
Hsd3bl_
predicted
Official gene name
Hsd3b l_predicted
hydroxysteroid
dehydrogenase-1,
delta< 5 >-3-beta
(predicted) orHsdSbl
hydroxy-delta-5 -steroid
dehydrogenase, 3 beta- and
steroid delta-isomerase 1
Hsd3b l_predicted
hydroxysteroid
dehydrogenase-1,
delta< 5 >-3-beta
(predicted) orHsdSbl
hydroxy-delta-5 -steroid
dehydrogenase, 3 beta- and
steroid delta-isomerase 1
Hsd3b l_predicted
hydroxysteroid
dehydrogenase-1,
delta< 5 >-3-beta
(predicted) orHsdSbl
hydroxy-delta-5 -steroid
dehydrogenase, 3 beta- and
steroid delta-isomerase 1
Hsd3b l_predicted
hydroxysteroid
dehydrogenase-1,
delta< 5 >-3-beta
(predicted) orHsdSbl
hydroxy-delta-5 -steroid
dehydrogenase, 3 beta- and
steroid delta-isomerase 1
Dose
0.1 mg/kg-d
1 mg/kg-d
10 mg/kg-d
50 mg/kg-d
Exposure window
CDs 12-19
CDs 12-19
CDs 12-19
CDs 12-19
Up or down
Down
Down
Down
Down
Statistical analysis method
Dunnett's test, ANOVA (one
way), p< 0.05
Dunnett's test, ANOVA (one
way), p< 0.05
Dunnett's test, ANOVA (one
way), p< 0.05
Dunnett's test, ANOVA (one
way), p< 0.05
Reference
Lehmann et
al., 2004
Lehmann et
al., 2004
Lehmann et
al., 2004
Lehmann et
al., 2004
-------
Table A-2. (continued)
Official
gene symbol
Hsd3bl_
predicted
Hsd3bl_
predicted
Ier3
Ifrdl
Igfl
Igfl
Igflr
Official gene name
Hsd3b l_predicted
hydroxysteroid
dehydrogenase-1,
delta< 5 >-3-beta
(predicted) orHsdSbl
hydroxy-delta-5 -steroid
dehydrogenase, 3 beta- and
steroid delta-isomerase 1
Hsd3b l_predicted
hydroxysteroid
dehydrogenase-1,
delta< 5 >-3-beta
(predicted) orHsdSbl
hydroxy-delta-5 -steroid
dehydrogenase, 3 beta- and
steroid delta-isomerase 1
Immediate early response 3
Interferon-related
developmental regulator 1
Insulin-like growth factor 1
Insulin-like growth factor 1
Insulin-like growth factor 1
receptor
Dose
100 mg/kg-d
500 mg/kg-d
500 mg/kg-d
500 mg/kg-d
500 mg/kg-d
500 mg/kg-d
500 mg/kg-d
Exposure window
CDs 12-19
CDs 12-19
GD 19for30minto6hr
timepoints and GD 18 for
12, 18, and 24 hr time
points
GD 19for30minto6hr
timepoints and GD 18 for
12, 18, and 24 hr time
points
CDs 12-21
CDs 12-19
CDs 12-19
Up or down
Down
Down
Up after
1-12 hr
(peak ~2 hr)
Up after
~l-6 and
18hr(peak
~3hr)
Up
Up
Up
Statistical analysis method
Dunnett's test, ANOVA (one
way), p< 0.05
Dunnett's test, ANOVA (one
way), p< 0.05
Relative expression determined
using mean Ct; triplicate
samples; GADPH control; SE;
but/) value not calculated
Relative expression determined
using mean Ct; triplicate
samples; GADPH control; SE;
but/) value not calculated
t-test,^<0.05
t-test,^<0.05
t-test,^<0.05
Reference
Lehmann et
al., 2004
Lehmann et
al., 2004
Thompson
et al., 2005
Thompson
et al., 2005
Bowman et
al., 2005
Bowman et
al., 2005
Bowman et
al., 2005
-------
Table A-2. (continued)
Official
gene symbol
IgP
Ig/bpS
Insigl
Insl3
Insl3
Insl3
Itgav
Junb
Kit
Kit
Official gene name
Insulin-like growth factor 2
Insulin-like growth factor
binding protein 5
Insulin induced gene 1
Insulin-like 3
Insulin-like 3
Insulin-like 3
Integrin alpha V
Jun-B oncogene
v-kit Hardy -Zuckerman
4 feline sarcoma viral
oncogene homolog
v-kit Hardy -Zuckerman
4 feline sarcoma viral
oncogene homolog
Dose
500 mg/kg-d
500 mg/kg-d
500 mg/kg-d
500 mg/kg-d
1000
mg/kg-d
500 mg/kg-d
500 mg/kg-d
500 mg/kg-d
500 mg/kg-d
0.1 mg/kg-d
Exposure window
CDs 12-19
CDs 12-21
CDs 12-19
CDs 12-19
CDs 13-17 (CDs 14-18 in
Wilson etal., 2004 was
changed to CDs 13-17 to
make the GD comparable to
the other 7 studies)
CDs 12.5-19.5
CDs 12-19
GD 19for30minto6hr
timepoints and GD 18 for
12, 18, and 24 hr time
points
CDs 12-19
CDs 12-19
Up or down
Up
Up
Down
Down
Down
Down
Up
UP after
-1-12 hr
(peak
-2-3 hr)
Down
Down
Statistical analysis method
t-test,^<0.05
t-test,^<0.05
One way; and two-way nested
ANOVA,^<0.05
Dunnett's test, ANOVA (one
way), p< 0.05
ANOVA followed by
LSMEANS, p < 0.01 or less
One-way ANOVA followed by
Bonferroni post test using
GraphPad Prism; p < 0.05
t-test,^<0.05
Relative expression determined
using mean Ct; triplicate
samples; GADPH control; SE;
but/) value not calculated
ANOVA, nested design,
^<0.05
Dunnett's test, ANOVA (one
way), p< 0.05
Reference
Bowman et
al., 2005
Bowman et
al., 2005
Liu etal.,
2005
Lehmann et
al., 2004
Wilson et
al., 2004
Plummer et
al., 2007
Bowman et
al., 2005
Thompson
et al., 2005
Barlow et
al., 2003
Lehmann et
al., 2004
-------
Table A-2. (continued)
Official
gene symbol
Kit
Kit
Kit
Kit
Kit
KM
Ldlr
Lhcgr
Map3kl2
Marcks
Mgp
Official gene name
v-kit Hardy -Zuckerman
4 feline sarcoma viral
oncogene homolog
v-kit Hardy -Zuckerman 4
feline sarcoma viral
oncogene homolog
v-kit Hardy -Zuckerman
4 feline sarcoma viral
oncogene homolog
v-kit Hardy -Zuckerman
4 feline sarcoma viral
oncogene homolog
v-kit Hardy -Zuckerman
4 feline sarcoma viral
oncogene homolog
Kit ligand
Low density lipoprotein
receptor
Luteinizing
hormone/choriogonadotropi
n receptor
Mitogen activated protein
kinase kinase kinase 12
Myristoylated alanine rich
protein kinase C substrate
Matrix Gla protein
Dose
1 mg/kg-d
50 mg/kg-d
100 mg/kg-d
500 mg/kg-d
500 mg/kg-d
500 mg/kg-d
500 mg/kg-d
500 mg/kg-d
500 mg/kg-d
500 mg/kg-d
500 mg/kg-d
Exposure window
CDs 12-19
CDs 12-19
CDs 12-19
CDs 12-19
CDs 12-19
CDs 12-19
CDs 12-19
CDs 12-19
CDs 12-19
CDs 12-16, 12-19, or
12-21
CDs 12-21
Up or down
Down
Down
Down
Down
Down at
GD 19
Down
Down
Down
Up
No stat.
Change
Up
Statistical analysis method
Dunnett's test, ANOVA (one
way), p< 0.05
Dunnett's test, ANOVA (one
way), p< 0.05
Dunnett's test, ANOVA (one
way), p< 0.05
Dunnett's test, ANOVA (one
way), p< 0.05
^<0.05
ANOVA, nested design,
^<0.05
One way and two-way nested
ANOVA,^<0.05
One way and two-way nested
ANOVA,^<0.05
t-test,/?<0.05
^<0.05
t-test,/?<0.05
Reference
Lehmann et
al., 2004
Lehmann et
al., 2004
Lehmann et
al., 2004
Lehmann et
al., 2004
Shultz et
al., 2001
Barlow et
al., 2003
Liuetal.,
2005
Liuetal.,
2005
Bowman et
al., 2005
Shultz et
al., 2001
Bowman et
al., 2005
-------
Table A-2. (continued)
Official
gene symbol
Mmp2
Mmp2
Nfil3
Nfil3
Notch2
Npc2
NrObl
NrObl
Nr4al
Nr4al
Official gene name
Matrix metallopeptidase 2
Matrix metallopeptidase 2
Nuclear factor, interleukin 3
regulated
Nuclear factor, interleukin 3
regulated
Notch gene homolog 2
(Drosophila)
Niemann Pick type C2
Nuclear receptor
subfamily 0, group B,
member 1
Nuclear receptor
subfamily 0, group B,
member 1
Nuclear receptor
subfamily 4, group A,
member 1
Nuclear receptor
subfamily 4, group A,
member 1
Dose
500 mg/kg-d
500 mg/kg-d
500 mg/kg-d
500 mg/kg-d
500 mg/kg-d
500 mg/kg-d
500 mg/kg-d
500 mg/kg-d
500 mg/kg-d
500 mg/kg-d
Exposure window
CDs 12-19
CDs 12-21
CDs 12-19
GD 19for30minto6hr
timepoints and GD 18 for
12, 18, and 24 hr time
points
CDs 12-21
CDs 12-19
CDs 12-19
GD 19for30minto6hr
timepoints and GD 18 for
12, 18, and 24 hr time
points
CDs 12-19
GD 19for30minto6hr
timepoints and GD 18 for
12, 18, and 24 hr time
points
Up or down
Up
Up
Up
Up after
-2-24 hr
(peak ~6 hr)
Up
Down
Down
Down at
2hr,Up
12 hr (peak
at 12 hr)
Up
Up after
~6 and 18 hr
(peak at
12 hr)
Statistical analysis method
t-test,/?<0.05
t-test,/?<0.05
One way and two-way nested
ANOVA,^<0.05
Relative expression determined
using mean Ct; triplicate
samples; GADPH control; SE;
but/) value not calculated
t-test,/?<0.05
One way and two-way nested
ANOVA,^<0.05
One way and two-way nested
ANOVA,^<0.05
Relative expression determined
using mean Ct; triplicate
samples; GADPH control; SE;
but/) value not calculated
One way and two-way nested
ANOVA,^<0.05
Relative expression determined
using mean Ct; triplicate
samples; GADPH control; SE;
but/) value not calculated
Reference
Bowman et
al., 2005
Bowman et
al., 2005
Liuetal.,
2005
Thompson
et al., 2005
Bowman et
al., 2005
Liu et al.,
2005
Liuetal.,
2005
Thompson
et al., 2005
Liuetal.,
2005
Thompson
et al., 2005
oo
oo
-------
Table A-2. (continued)
Official
gene symbol
Nr4a3
Pawr
Pcna
Prkcbpl
Scarbl
Scarbl
Scarbl
Scarbl
Scarbl
Scarbl
Scarbl
Official gene name
Nuclear receptor
subfamily 4, group A,
member 3
PRKC, apoptosis, WT1,
regulator
Proliferating cell nuclear
antigen
Protein kinase C binding
protein 1
Scavenger receptor class B,
member 1
Scavenger receptor class B,
member 1
Scavenger receptor class B,
member 1
Scavenger receptor class B,
member 1
Scavenger receptor class B,
member 1
Scavenger receptor class B,
member 1
Scavenger receptor class B,
member 1
Dose
500 mg/kg-d
500 mg/kg-d
500 mg/kg-d
500 mg/kg-d
500 mg/kg-d
1 mg/kg-d
50 mg/kg-d
100 mg/kg-d
500 mg/kg-d
500 mg/kg-d
500 mg/kg-d
Exposure window
GD 19for30minto6hr
timepoints and GD 18 for
12, 18, and 24 hr time
points
GD 19for30minto6hr
timepoints and GD 18 for
12, 18, and 24 hr time
points
CDs 12-16, 12-19, or
12-21
CDs 12-19
CDs 12-19
CDs 12-19
CDs 12-19
CDs 12-19
CDs 12-19
CDs 12-16, 12-19, or
12-21
CDs 12-17 and 12-18
Up or down
Up after
-1-12 hr
(peak at
-3-6 hr)
Up after
-2-24 hr
(peak -6 hr)
No stat.
change
Up
Down
Down
Down
Down
Down
Down
Down at
CDs 17 and
18
Statistical analysis method
Relative expression determined
using mean Ct; triplicate
samples; GADPH control; SE;
but/) value not calculated
Relative expression determined
using mean Ct; triplicate
samples; GADPH control; SE;
but/) value not calculated
^<0.05
One way and two-way nested
ANOVA,^<0.05
ANOVA, nested design,
^<0.05
Dunnett's test, ANOVA (one
way), p< 0.05
Dunnett's test, ANOVA (one
way), p< 0.05
Dunnett's test, ANOVA (one
way), p< 0.05
Dunnett's test, ANOVA (one
way), p< 0.05
^<0.05
t-test, ANOVA (one-way) with
Tukey post hoc analysis,
^<0.05
Reference
Thompson
et al., 2005
Thompson
et al., 2005
Shultz et
al., 2001
Liuetal.,
2005
Barlow et
al., 2003
Lehmann et
al., 2004
Lehmann et
al., 2004
Lehmann et
al., 2004
Lehmann et
al., 2004
Shultz et
al., 2001
Thompson
et al., 2004
oo
VO
-------
Table A-2. (continued)
Official
gene symbol
Sgk
Sostdcl
Star
Star
Star
Star
Star
Star
Star
Official gene name
Serum/glucocorticoid
regulated kinase
Sclerostin domain
containing 1
Steroidogenic acute
regulatory protein
Steroidogenic acute
regulatory protein
Steroidogenic acute
regulatory protein
Steroidogenic acute
regulatory protein
Steroidogenic acute
regulatory protein
Steroidogenic acute
regulatory protein
Steroidogenic acute
regulatory protein
Dose
500 mg/kg-d
500 mg/kg-d
500 mg/kg-d
50 mg/kg-d
100 mg/kg-d
500 mg/kg-d
500 mg/kg-d
500 mg/kg-d
500 mg/kg-d
Exposure window
GD 19for30minto6hr
timepoints and GD 18 for
12, 18, and 24 hr time
points
GD 19for30minto6hr
timepoints and GD 18 for
12, 18, and 24 hr time
points
CDs 12-19
CDs 12-19
CDs 12-19
CDs 12-19
CDs 12-16, 12-19, or
12-21
CDs 12-17 and 12-18
CDs 12.5-19.5
Up or down
Down and
Up; Down
after 2 hr;
Up after 4
and 10 hr
(peak at
6hr)
Down after
2-6 hr; Up
atlShr
(peak)
Down
Down
Down
Down
Down at
CDs 16, 19,
and 21
Down at
CDs 17 and
18
Down
Statistical analysis method
Relative expression determined
using mean Ct; triplicate
samples; GADPH control; SE;
but/) value not calculated
Relative expression determined
using mean Ct; triplicate
samples; GADPH control; SE;
but/) value not calculated
Repeated measure ANOVA,
nested design, p < 0.05
Dunnett's test, ANOVA (one
way), p< 0.05
Dunnett's test, ANOVA (one
way), p< 0.05
Dunnett's test, ANOVA (one
way), p< 0.05
^<0.05
t-test, ANOVA (one-way) with
Tukey post hoc analysis;
^<0.05
One-way ANOVA followed by
Bonferroni post test using
GraphPad Prism; p < 0.05
Reference
Thompson
et al., 2005
Thompson
et al., 2005
Barlow et
al., 2003
Lehmann et
al., 2004
Lehmann et
al., 2004
Lehmann et
al., 2004
Shultz et
al., 2001
Thompson
et al., 2004
Plummer et
al., 2007
-------
Table A-2. (continued)
Official
gene symbol
Stcl
SvsS
Tcfl
Tcfl
Testin
Thbsl
Timpl
Tnfrsfl2a
Wnt4
Official gene name
Stanniocalcin 1
Seminal vesicle secretion 5
Transcription factor 1
Transcription factor 1
Testin gene
Thrombospondin 1
Tissue inhibitor of
metalloproteinase 1
Tumor necrosis factor
receptor superfamily,
member 12a
Wingless-related MMTV
integration site 4
Dose
500 mg/kg-d
500 mg/kg-d
500 mg/kg-d
500 mg/kg-d
500 mg/kg-d
500 mg/kg-d
500 mg/kg-d
500 mg/kg-d
500 mg/kg-d
Exposure window
GD 19for30minto6hr
timepoints and GD 18 for
12, 18, and 24 hr time
points
CDs 12-19
CDs 12-19
GD 19for30minto6hr
timepoints and GD 18 for
12, 18, and 24 hr time
points
CDs 12-19
GD 19for30minto6hr
timepoints and GD 18 for
12, 18, and 24 hr time
points
CDs 12-21
GD 19for30minto6hr
timepoints and GD 18 for
12, 18, and 24 hr time
points
GD 19for30minto6hr
timepoints and GD 18 for
12, 18, and 24 hr time
points
Up or down
Up after
-3-24 hr
(peak ~6 hr)
Down
Down
Down after
1-3 hr (peak
atlhr)
Up
Up after
2-4 hr (peak
~3hr)
Up
Up after
-1-12 hr
(peak at
~3hr)
Up after
-12 and
18 hr(peak
12 hr)
Statistical analysis method
Relative expression determined
using mean Ct; triplicate
samples; GADPH control; SE;
but/) value not calculated
One way and two-way nested
ANOVA,^<0.05
One way and two-way nested
ANOVA,^<0.05
Relative expression determined
using mean Ct; triplicate
samples; GADPH control; SE;
but/) value not calculated
One way and two-way nested
ANOVA,^<0.05
Relative expression determined
using mean Ct; triplicate
samples; GADPH control; SE;
but/) value not calculated
t-test,/?<0.05
Relative expression determined
using mean Ct; triplicate
samples; GADPH control; SE;
but/) value not calculated
Relative expression determined
using mean Ct; triplicate
samples; GADPH control; SE;
but/) value not calculated
Reference
Thompson
et al., 2005
Liuetal.,
2005
Liuetal.,
2005
Thompson
et al., 2005
Liuetal.,
2005
Thompson
et al., 2005
Bowman et
al., 2005
Thompson
et al., 2005
Thompson
et al., 2005
>
-------
Table A-2. (continued)
Official
gene symbol
Zfp36
Official gene name
Zinc finger protein 36
Dose
500 mg/kg-d
Exposure window
GD 19for30minto6hr
timepoints and GD 18 for
12, 18, and 24 hr time
points
Up or down
Up after
30 min and
6 brand
15 and 20 hr
(peak at
Ihr)
Statistical analysis method
Relative expression determined
using mean Ct; triplicate
samples; GADPH control; SE;
butp value not calculated
Reference
Thompson
et al., 2005
*Gene function and pathway information was gathered from GeneGo (www.genego.com).
to
-------
Table A-3. Genes identified using the Rosetta Error Model statistical filter
and mapping to the five most significant biochemical functions and /or
pathways using Ingenuity
Gene symbol
Gene name
Genes mapped to integrin pathway
F2r
Src
GngS
Gnai3
Gng7
Mapk3
Gnaol
Actcl
Camk2d
Gnaq
CxclJ2
Prkce
Coagulation factor II (thrombin) receptor
Rous sarcoma oncogene
Guanine nucleotide binding protein (G protein), gamma 5 subunit
Guanine nucleotide binding protein, alpha inhibiting 3
Guanine nucleotide binding protein, gamma 7
Mitogen activated protein kinase 3
Guanine nucleotide binding protein, alpha o
Actin alpha cardiac 1
Calcium/calmodulin-dependent protein kinase II, delta
Guanine nucleotide binding protein
Chemokine (C-X-C motif) ligand 12
Protein kinase C, epsilon
Genes mapped to cholesterol biosynthesis/metabolism
Hmgcsl
HsdSbl
Dhcr?
Sqle
Soatl
CypSlal
Cyp27aJ
Hsdllbl
Hmgcr
Idil
Sc4mol
Cyp7bl
3-Hydroxy-3-methylglutaryl-Coenzyme A synthase 1
Hydroxyl-delta-5-steroid dehydrogenase
7-Dehydrocholesterol reductase
Squalene epoxidase
Sterol O-acyltransferase 1
Cytochrome P450, family 51, subfamily a, polypeptide 1
Cytochrome P450, family 27, subfamily a, polypeptide 1
Hydroxy steroid 11 -beta dehydrogenase 1
3-Hydroxy-3-methylglutaryl-Coenzyme A reductase
Osopentenyl-diphosphate delta isomerase
Sterol-C4-methyl oxidase-like
Cytochrome P450, family 7, subfamily b, polypeptide 1
A-43
-------
Table A-3. (continued)
Gene symbol
Gene name
Genes mapped to chemokine mediated signaling
Src
Gng5
Hmgcsl
Serpine2
ItgbS
Dhcr?
Gnai3
Gng7
Sqle
Mapk3
Gnaol
Actnl
Actcl
Cav2
CypSlal
Rous sarcoma oncogene
Guanine nucleotide binding protein (G protein), gamma 5 subunit
3-Hydroxy-3-methylglutaryl-Coenzyme A synthase 1
Serine (or cysteine) proteinase inhibitor, clade E, member 2
Integrin, beta 5
7-Dehydrocholesterol reductase
Guanine nucleotide binding protein, alpha inhibiting 3
Guanine nucleotide binding protein, gamma 7
Squalene epoxidase
Mitogen activated protein kinase 3
Guanine nucleotide binding protein, alpha o
Actinin, alpha 1
Actin alpha cardiac 1
Caveolin 2
Cytochrome P450, family 51, subfamily a, polypeptide 1
Genes mapped to chemokine mediated signaling
Colla2
Cfll
Cavl
Hmgcr
Mmp2
Mm
Gsk3b
Idil
Plat
Sdc2
Sc4mol
Lefl
Procollagen, type I, alpha 2
Cofilin 1, non-muscle
Caveolin 2
3-Hydroxy-3-methylglutaryl-Coenzyme A reductase
Matrix metallopeptidase 2
Moesin
Glycogen synthase kinase 3 beta
Isopentenyl-diphosphate delta isomerase
Plasminogen activator, tissue
Syndecan 2
Sterol-C4-methyl oxidase-like
Lymphoid enhancer binding factor 1
A-44
-------
Table A-3. (continued)
Gene symbol
Vegf
Gene name
Vascular endothelial growth factor
Genes mapped to glycolysis/gluconeogenesis
Pgkl
Hmgcsl
Tpil
Fbp2
Dhcr?
Pfkm
P/kp
Mdhl
Sqle
Pgaml
Aldoa
CypSlal
Hmgcr
Hkl
Gpi
Gapdh
Idil
Sc4mol
Pfkl
Phosphoglycerate kinase 1
3-Hydroxy-3-methylglutaryl-Coenzyme A synthase 1
Triosephosphate isomerase 1
Fructose- 1,6-bisphosphatase 2
7-Dehydrocholesterol reductase
Phosphofructokinase, muscle
Phosphofructokinase, platelet
Malate dehydrogenase 1, NAD (soluble)
Squalene epoxidase
Phosphoglycerate mutase 1
Aldolase A
Cytochrome P450, family 51, subfamily a, polypeptide 1
3-Hydroxy-3-methylglutaryl-Coenzyme A reductase
Hexokinase 1
Glucose phosphate isomerase
Glyceraldehyde-3 -phosphate dehydrogenase
Isopentenyl-diphosphate delta isomerase
Sterol-C4-methyl oxidase-like
Phosphofructokinase, liver
A-45
-------
Table A-4. Differentially expressed genes that mapped to statistically
significant pathways identified using the Signal to Noise Ratio (SNR)
statistical filter
Gene
symbol
Aadat
Acadm
Acads
Acatl
Aco2
Acsl4
Akrlb4
Alasl
Aldhla4
Aldh2
Aldh6al
Aldoa
Aldoc
Ass
Bhmt
Chkb
Cypllal
CyplJal
Dcxr
Ddc
Dhcr?
Ebp
Ephxl
Fbp2
Fdftl
Fdps
Fhl
Entrez
gene ID
29416
24158
64304
25014
79250
113976
24192
65155
29651
29539
81708
24189
24191
25698
81508
29367
29680
25146
171408
24311
64191
117278
25315
114508
29580
83791
24368
Gene name
Aminoadipate aminotransferase
Acetyl-Coenzyme A dehydrogenase, medium chain
Acyl-Coenzyme A dehydrogenase, short chain
Acetyl-Coenzyme A acetyltransferase 1
Aconitase 2, mitochondrial
Acyl-CoA synthetase long-chain family member 4
Aldo-keto reductase family 1, member B4 (aldose reductase)
Aminolevulinic acid synthase 1
Aldehyde dehydrogenase family 1, subfamily A4
Aldehyde dehydrogenase 2
Aldehyde dehydrogenase family 6, subfamily Al
Aldolase A
Aldolase C, fructose-biphosphate
Arginosuccinate synthetase
Betaine-homocysteinemethyltransferase
Choline kinase beta
Cytochrome P450, family 11, subfamily a, polypeptide 1
Cytochrome P450, family 17, subfamily a, polypeptide 1
Dicarbonyl L-xylulose reductase
Dopa decarboxylase
7-dehydrocholesterol reductase
Phenylalkylamine Ca2+ antagonist (emopamil) binding protein
Epoxide hydrolase 1
Fructose- 1,6-bisphosphatase 2
Farnesyl diphosphate farnesyl transferase 1
Farnesyl diphosphate synthase
Fumarate hydratase 1
A-46
-------
Table A-4. (continued)
Gene
symbol
G6pdx
Gad2
Gapdh
Gatm
Ggtl3
Gsta2
Gsta3
Gstm2
Gstm3
Hmgcr
Hmgcsl
Idhl
Mel
Mgstl
Mif
Mvd
Nosl
Pycr2
Sqle
Suclgl
Tpil
Entrez
gene ID
24377
24380
24383
81660
156275
24422
24421
24424
81869
25675
29637
24479
24552
171341
81683
81726
24598
364064
29230
114597
24849
Gene name
Glucose-6-phosphate dehydrogenase
Glutamate decarboxylase 2
Glyceraldehyde-3 -phosphate dehydrogenase
Glycine amidinotransferase (L-arginine:glycine amidinotransferase)
Gamma-glutamyltransferase-like 3
Glutathione-S-transferase, alpha type2
Glutathione S-transferase A5
Glutathione S-transferase, mu 2
Glutathione S-transferase, mu type 3
3-hydroxy-3-methylglutaryl-Coenzyme A reductase
3-hydroxy-3-methylglutaryl-Coenzyme A synthase 1
Isocitrate dehydrogenase 1 (NADP+), soluble
Malic enzyme 1
Microsomal glutathione S-transferase 1
Macrophage migration inhibitory factor
Mevalonate (diphospho) decarboxylase
Nitric oxide synthase 1, neuronal
Pyrroline-5-carboxylate reductase family, member 2 (predicted)
Squalene epoxidase
Succinate-CoA ligase, GDP-forming, alpha subunit
Tpil protein
A-47
-------
Table A-5. GeneGo pathway analysis of significant genes identified by REM
Pathway
NF-AT signaling in cardiac hypertrophy
MIF — the neuroendocrine-macrophage
connector
Lysine metabolism
Cholesterol metabolism
Glycolysis and gluconeogenesis (short
map)
Integrin-mediated cell adhesion
Tryptophan metabolism
Cholesterol biosynthesis
ECM remodeling
Regulation of lipid metabolism via
PPAR, RXR, and VDR
Propionate metabolism p. 2
PPAR regulation of lipid metabolism
Mitochondrial long chain fatty acid
beta-oxidation
Role of VDR in regulation of genes
involved in osteoporosis
ChREBP regulation pathway
Androstenedione and testosterone
biosynthesis and metabolism p. 1
Arginine metabolism
Regulation of fatty acid synthesis:
NLTP and EHHADH
Angiotensin signaling via STATs
Cytoskeleton remodeling
dGTP metabolism
TCA
Glycolysis and gluconeogenesis p. 1
Peroxisomal branched chain fatty acid
oxidation
Biological process
Disease
Immune response
Amino acid metabolism
Steroid metabolism
Carbohydrates metabolism
Cell adhesion
Amino acid metabolism
Steroid metabolism
Cell adhesion
Transcription
Carbohydrates metabolism
Regulation of lipid metabolism
Lipid metabolism
Transcription
G-protein coupled receptor
signaling
Steroid metabolism
Amino acid metabolism
Regulation of lipid metabolism
Growth and differentiation
Cell adhesion
Nucleotide metabolism
Amino acid metabolism
Carbohydrates metabolism
Lipid metabolism
p-Valuea
2.23E-04
3.00E-04
3.05E-04
6.95E-04
7.40E-04
8.44E-04
9.56E-04
1.44E-03
1.64E-03
1.96E-03
1.96E-03
2.04E-03
2.28E-03
3.16E-03
3.82E-03
4.30E-03
4.45E-03
5.02E-03
5.18E-03
5.19E-03
5.34E-03
5.70E-03
5.70E-03
5.70E-03
No. of
genesb'c
19/90
19/92
9/27
6/14
10/36
18/92
9/31
7/21
13/60
7/22
7/22
8/28
6/17
12/57
10/44
6/19
9/38
4/9
11/53
26/176
9/39
6/20
6/20
6/20
A-48
-------
Table A-5. (continued)
Pathway
Gamma-aminobutyrate (GABA)
biosynthesis and metabolism
Ligand-dependent activation of the
ESR1/SP pathway
Integrin inside-out signaling
Reverse signaling by ephrin B
G-protein beta/gamma signaling
cascades
Activation of PKC via G-Protein
coupled receptor
Gap junctions
WNT signaling pathway
Angiotensin activation of ERK
Role of Akt in hypoxia induced HIF1
activation
Regulation of actin cytoskeleton by Rho
GTPases
CCR3 signaling in eosinophils
MAG-dependent inhibition of neurite
outgrowth
Endothelial cell contacts by junctional
mechanisms
Fructose metabolism
Regulation of lipid metabolism via LXR,
NF-Y and SREBP
CXCR4 signaling pathway
Serotonin-melatonin biosynthesis and
metabolism
Glycolysis and gluconeogenesis p. 2
Oxidative phosphorylation
Urea cycle
G-proteins mediated regulation p. 3 8 and
JNK signaling
Biological process
Metabolism of mediators
Response to hormone stimulus
Cell adhesion
Cell adhesion
G-protein coupled receptor
protein signaling pathway
G-protein coupled receptor
protein signaling pathway
Cell adhesion
Proteolysis
G-protein coupled receptor
protein signaling pathway
Proteolysis
Small GTPase mediated signal
transduction
Immune response
Response to extracellular
stimulus
Cell adhesion
Carbohydrates metabolism
Transcription
Cytokine and chemokine
mediated signaling pathway
Metabolism of mediators
Carbohydrates metabolism
Energy metabolism
Amino acid metabolism
G-protein coupled receptor
protein signaling pathway
p-valuea
5.70E-03
6.38E-03
6.85E-03
6.86E-03
6.94E-03
7.65E-03
8.51E-03
8.59E-03
9.12E-03
9.83E-03
1.18E-02
1.22E-02
1.47E-02
1.80E-02
1.80E-02
1.80E-02
1.89E-02
2.04E-02
2.15E-02
2.37E-02
2.58E-02
2.60E-02
No. of
genesbc
6/20
9/40
14/78
15/86
11/55
15/87
10/49
7/28
11/57
10/50
11/59
18/117
10/53
7/32
7/32
7/32
10/55
5/19
4/13
15/99
6/27
11/66
A-49
-------
Table A-5. (continued)
Pathway
Transcription factor tubby signaling
pathways
Role PKA in cytoskeleton reorganization
Ephrins signaling
Propionate metabolism p. 1
Estrone metabolism
Regulation of acetyl-CoA carboxylase 2
activity in muscle
Chemokines and adhesion
Arachidonic acid production
dCTP/dUTP metabolism
Regulation of lipid metabolism by niacin
and isoprenaline
Ubiquinone metabolism
Phenylalanine metabolism
Leptin signaling via JAK/STAT and
MAPK cascades
IMP biosynthesis
EPO-induced Jak-STAT pathway
Integrin outside-in signaling
Brcal as transcription regulator
P53 signaling pathway
Bile acid biosynthesis
Histidine-glutamate-glutamine and
proline metabolism
NTS activation of IL-8 in colonocytes
Biological process
Transcription
Protein kinase cascade
Cell adhesion
Carbohydrates metabolism
Steroid metabolism
Response to extracellular
stimulus
Cytokine and chemokine
mediated signaling pathway
Lipid metabolism
Nucleotide metabolism
Regulation of lipid metabolism
Vitamin and cofactor
metabolism
Amino acid metabolism
Response to hormone stimulus
Nucleotide metabolism
Response to extracellular
stimulus
Cell adhesion
Cell cycle
Transcription regulation
Steroid metabolism
Amino acid metabolism
Immune response
p-valuea
2.63E-02
2.64E-02
2.66E-02
2.81E-02
2.81E-02
2.81E-02
2.82E-02
2.87E-02
2.99E-02
3.01E-02
3.01E-02
3.05E-02
3.57E-02
3.70E-02
3.78E-02
3.95E-02
4.15E-02
4.28E-02
4.43E-02
4.79E-02
4.85E-02
No. of
genesbc
8/42
13/83
10/58
4/14
4/14
4/14
23/174
7/35
8/43
9/51
9/51
6/28
6/29
3/9
7/37
12/79
6/30
8/46
5/23
8/47
10/64
aOrdered from most significant (lowest />-value) to less significant.
lumber of genes from the DBF-exposed gene list mapping to the GeneGo pathway.
°Total number of genes in the GeneGo pathway.
A-50
-------
Table A-6. Significant biological pathways corresponding to differentially expressed genes (DEGs) obtained
from SNR analysis input into GeneGo
Pathway
Cholesterol Biosynthesis
Propionate metabolism p.2
MIF — the neuroendocrine-macrophage connector
Tryptophan metabolism
Lysine metabolism
Cholesterol metabolism
NF-AT signaling in cardiac hypertrophy
Glycolysis and gluconeogenesis (short map)
G-alpha(q) regulation of lipid metabolism
Activation of PKC via G-protein coupled receptor
Fructose metabolism
Regulation of lipid metabolism by niacin and isoprenaline
ATP metabolism
Angiotensin activation of ERK
NTS activation of IL-8 in colonocytes
Leucine, isoleucine, andvaline metabolism.p.2
Biological Process
Steroid metabolism
Carbohydrates metabolism
Immune response
Amino acid metabolism
Amino acid metabolism
Steroid metabolism
Disease
Carbohydrates metabolism
Regulation of lipid metabolism
G-proteins/GPCR
Carbohydrates metabolism
Regulation of lipid metabolism
Nucleotide metabolism
Growth and differentiation
Immune response
Amino acid metabolism
p-Valuea
1.81E-09
5.54E-06
3.22E-04
3.78E-04
3.93E-04
1.09E-03
1.38E-03
1.77E-03
1.93E-03
2.00E-03
2.06E-03
2.08E-03
2.09E-03
2.55E-03
3.60E-03
3.64E-03
No. ofgenesbc
15/21
12/22
25/92
12/31
11/27
7/14
23/90
12/36
13/41
22/87
11/32
15/51
16/56
16/57
17/64
9/25
>
-------
Table A-6. (continued)
Pathway
Reverse signaling by ephrin B
Cortisone biosynthesis and metabolism
CXCR4 signaling pathway
G-Protein beta/gamma signaling cascades
Glutathione metabolism
Mitochondria! ketone bodies biosynthesis and metabolism
Integrin inside-out signaling
Propionate metabolism p. 1
Role of VDR in regulation of genes involved in osteoporosis
Endothelial cell contacts by junctional mechanisms
EPO-induced Jak-STAT pathway
A3 receptor signaling
Angiotensin signaling via STATs
MAG-dependent inhibition of neurite outgrowth
Phenylalanine metabolism
Androstenedione and testosterone biosynthesis and metabolism p. 1
Cytoskeleton remodeling
Biological Process
Cell adhesion
Steroid metabolism
Immune response
G-proteins/GPCR
Vitamin and cofactor metabolism
Lipid metabolism
Cell adhesion
Carbohydrates metabolism
Transcription factors
Cell adhesion
Cell survival
G-proteins/GPCR
Growth and differentiation
Growth and differentiation
Amino acid metabolism
Steroid metabolism
Cell adhesion
p-Valuea
3.92E-03
4.31E-03
4.63E-03
4.63E-03
5.77E-03
5.96E-03
6.07E-03
6.51E-03
6.63E-03
7.02E-03
7.24E-03
8.08E-03
8.28E-03
8.28E-03
8.48E-03
8.76E-03
9.69E-03
No. of genesbc
21/86
7/17
15/55
15/55
11/36
5/10
19/78
6/14
15/57
10/32
11/37
19/80
14/53
14/53
9/28
7/19
35/176
to
-------
Table A-6. (continued)
Pathway
ChREBP regulation pathway
Leptin signaling via JAK/STAT and MAPK cascades
dGTP metabolism
TCA
Glycolysis and gluconeogenesis p. 1
Gamma-aminobutyrate (GABA) biosynthesis and metabolism
BAD phosphorylation
Ligand-dependent activation of the ESR1/SP pathway
RAB5A regulation pathway
Integrin outside-in signaling
Hedgehog and PTH signaling pathways participation in bone and
cartilage development
G-Proteins mediated regulation MARK-ERK signaling
Integrin-mediated cell adhesion
Mitochondria! long chain fatty acid beta-oxidation
CCR3 signaling in eosinophils
Regulation of lipid metabolism via PPAR, RXR, and VDR
Biological Process
Regulation of transcription
Growth and differentiation
Nucleotide metabolism
Amino acid metabolism
Carbohydrates metabolism
Metabolism of mediators
Apoptosis
Hormones
G-proteins/RAS-group
Cell adhesion
Growth and differentiation
G-proteins/GPCR
Cell adhesion
Lipid metabolism
Immune response
Transcription factors
p-Valuea
1.08E-02
1.09E-02
1.10E-02
1.20E-02
1.20E-02
1.20E-02
1.21E-02
1.34E-02
1.49E-02
1.50E-02
1.62E-02
1.64E-02
1.78E-02
1.88E-02
2.02E-02
2.07E-02
No. of genesbc
12/44
9/29
11/39
7/20
7/20
7/20
19/83
11/40
5/12
18/79
11/41
17/74
20/92
6/17
24/117
7/22
-------
Table A-6. (continued)
Pathway
Glycolysis and gluconeogenesis p. 2
Regulation of fatty acid synthesis: NLTP and EHHADH
Role PKA in cytoskeleton reorganization
Arginine metabolism
ECM remodeling
Ca (2+)-dependent NF-AT signaling in cardiac hypertrophy
WNT signaling pathway
PPAR regulation of lipid metabolism
Insulin regulation of the protein synthesis
CXCR4 signaling via second messenger
Angiotensin signaling via beta-Arrestin
Estrone metabolism
Regulation of acetyl-Co A carboxylase 2 activity in muscle
Prolactin receptor signaling
Triacylglycerol metabolism p. 1
Serotonin-melatonin biosynthesis and metabolism
Angiotensin signaling via PYK2
Biological Process
Carbohydrates metabolism
Regulation of lipid metabolism
Kinases
Amino acid metabolism
Cell adhesion
Disease
Growth and differentiation
Regulation of lipid metabolism
Translation regulation
Immune response
Growth and differentiation
Steroid metabolism
Growth and differentiation
Growth factors
Lipid metabolism
Metabolism of mediators
Growth and differentiation
p-Valuea
2.16E-02
2.30E-02
2.43E-02
2.44E-02
2.45E-02
2.55E-02
2.64E-02
2.64E-02
2.67E-02
2.67E-02
2.71E-02
2.99E-02
2.99E-02
3.19E-02
3.23E-02
3.27E-02
3.32E-02
No. of genesbc
5/13
4/9
18/83
10/38
14/60
15/66
8/28
8/28
13/55
13/55
11/44
5/14
5/14
14/62
8/29
6/19
16/74
-------
Table A-6. (continued)
Pathway
G-Protein alpha-i signaling cascades
dATP/dlTP metabolism
Brcal as transcription regulator
Ephrins signaling
Mitochondria! unsaturated fatty acid beta-oxidation
GDNF signaling
Aspartate and asparagine metabolism
Peroxisomal branched chain fatty acid oxidation
Histidine-glutamate-glutamine and proline metabolism
TGF-beta receptor signaling
Regulation of actin cytoskeleton by Rho GTPases
G-Protein alpha-s signaling cascades
Al receptor signaling
Membrane-bound ESR1 : interaction with growth factors signaling
Transcription factor Tubby signaling pathways
Histamine metabolism
PPAR pathway
Biological Process
G-proteins/GPCR
Nucleotide metabolism
Cell-cycle control
Cell adhesion
Lipid metabolism
Growth and differentiation
Amino acid metabolism
Lipid metabolism
Amino acid metabolism
Growth and differentiation
G-proteins/RAS-group
G-proteins/GPCR
G-proteins/GPCR
Growth and differentiation
Regulation of transcription
Metabolism of mediators
Transcription factors
p-Valuea
3.36E-02
3.86E-02
3.90E-02
3.99E-02
4.01E-02
4.08E-02
4.15E-02
4.15E-02
4.24E-02
4.51E-02
4.51E-02
4.51E-02
4.61E-02
4.64E-02
4.64E-02
4.83E-02
4.86E-02
No. of genesbc
12/51
12/52
8/30
13/58
5/15
7/25
6/20
6/20
11/47
13/59
13/59
13/59
16/77
10/42
10/42
4/11
11/48
-------
Table A-6. (continued)
Pathway
Cross-talk VEGF and angiopoietin 1 signaling
EPO-induced MAPK pathway
Biological Process
Growth and differentiation
Growth and differentiation
p-Valuea
5.08E-02
5.08E-02
No. of genesbc
9/37
13/60
aOrdered from most significant (lowest />-value) to less significant.
bNumber of genes from the DBF exposed gene list mapping to the GeneGo pathway.
°Total number of genes in the GeneGo pathway.
-------
For a given gene
expression
Evaluate SNR between
control and treated
samples
.
/ Select /
—'$/ random gene fi—
/ expression /
Evaluate SNR for randomly selected
gene expression
Random Expression = random Expression +1
No
Statistical Significance = (Random Expression/1000)* 100
\
,
Given gene expression is
statistically significant
Figure A-l. Algorithm for selecting differentially expressed genes
(DEGs) using signal-to-noise ration (SNR). 1,000 random gene
expressions were generated for each probe set, and then, SNRs were
calculated. The ratio of the randomly generated SNR that was higher
than the actual SNR determined whether individual probe set's
expression was discriminating or not.
A-57
-------
APPENDIX B.
SUPPORTING TABLES AND FIGURES FOR CHAPTER 6
Appendix B contains additional tables and figures supportive of the work described in
Chapter 6.
Table B-l. Nodes added by using Ingenuity® Pathway Analysis (IPA)
software in developing the gene network model for DBP
Gene
symbol
Acol
Esrra
Fgf4
Insigl
Kcnjll
Lep
Lnpep
Nfrc
Nmel
Nr2fl
NrSal
Pld2
Ppargclb
Srebfl
Srebf2
Zdhhc23
Gene name
Aconitase 1, soluble
Estrogen-related receptor alpha
Fibroblast growth factor 4
Insulin induced gene 1
Potassium inwardly-rectifying channel, subfamily J, member 1 1
Leptin
Leucyl/cystinyl aminopeptidase
Nuclear factor I/C (CCAAT -binding transcription factor)
Non-metastatic cells 1, protein (NM23A) expressed in
Nuclear receptor subfamily 2, group F, member 1
Nuclear receptor subfamily 5, group A, member 1
Phospholipase D2
Peroxisome proliferative activated receptor, gamma, coactivator 1, beta
Sterol regulatory element binding transcription factor 1
Sterol regulatory element binding transcription factor 2
Zinc finger, DFfflC-type containing 23
B-l
-------
For a given pathway
Evaluate P/L
Select a random
set of gene
expression
Evaluate PAp for randomly
selected set of genes
No
Yes
Random Pathway Activity - Random Pathway Activity +1
\
t
No
i= 1000
Yes
Statistical Significance = (Random Pathway Activity/1000)* 100
Statistical Significance < 0.05
Given pathway
activity is random
Given pathway activity is
statistically significant
Figure B-l. Algorithm for selecting significant pathways using the pathway
activity method. 1,000 random sets of gene expressions were generated for each
pathway, then pathway activity, PAp, was evaluated. The/>-value of each PAp is
computed as the fraction of the randomized PAp that exceeded the actual PAp.
B-2
-------
APPENDIX C.
QUALITY CONTROL AND ASSURANCE
Appendix C contains quality assurance/quality control (QA/QC) information for the work
described in Chapters 5 and 6. The work described in this Appendix (C) is secondary data
analysis. The studies include exploratory studies using new methods for analyzing genomic data
for risk assessment purposes as well as some preliminary analyses using well-established of the
raw data from two published studies.
Three projects were performed:
(1) A qualitative analysis and presentation of the 9 toxicogenomic DBF studies. No
statistical analyses were performed by members of our team.
(2) In-house analysis of the raw data from Liu et al. (2005) study performed at both
NHEERL, US EPA by Drs. Susan Hester and Banalata Sen, and by by collaborators, Dr.
loannis Androulakis and Meric Ovacik, STAR Grantees at the STAR Bioinformatics
Center at Rutgers/UMDNJ.
(3) New analyses of Thompson et al. (2005) data performed by collaborators, Dr. loannis
Androulakis and Meric Ovacik, STAR Grantees at the STAR Bioinformatics Center at
Rutgers/UMDNJ.
C.I. PROJECT 1
The data presented in 9 published toxicogenomic studies for DBF were compared. No
additional analyses were performed. Data were entered directly into an excel spreadsheet from
the published literature. Study descriptions in tables and figures were developed. The data entry
process included team members entering in the data from the published articles into tables for
differentially expressed genes and pathways affected. One person entered the data for a subset of
genes. A second person checked the results in the table against the articles.
C.2. PROJECT 2
The data source was the DBF treatment only data from the Liu et al. (2005) study. The
Liu et al. (2005) data were kindly provided by Dr. Kevin Gaido, a collaborator on this project.
The study was performed in his laboratory at The Hamner Institutes for Health Sciences
(formerly CUT). His QA statement for the collection and analysis of the data is provided below.
C-l
-------
C.3. PROJECT 3
The data source was the Thompson et al. (2005) study. The Thompson et al. (2005) data
were kindly provided by Dr. Kevin Gaido, a collaborator on this project. The study was
performed in his laboratory at The Hamner Institutes for Health Sciences (formerly CUT). His
QA statement for the collection and analysis of the data is provided below.
C.4. PROJECTS 2 AND 3: DATA SOURCES
The sources of the data used in the secondary analyses were the Liu et al. (2005) and
Thompson et al. (2005) studies. Both of these studies were performed in the laboratory of Dr.
Kevin Gaido. The QA details for the two studies are presented below. The Hamner Institute's
Quality Assurance Director is Patricia O. Pomerleau, M.S., RQAP (pomerleau@thehamner.org).
C.4.1. Sample Handling Procedures
Virgin female SD outbred CD rats, 8 weeks old, were time mated. Dams were assigned
to a treatment group by randomization using Provantis NT 2000 and subsequently be identified
by an ear tag and cage card. Dams were kept in the Association for Assessment and
Accreditation of Laboratory Animal Care International accredited animal facility at The Hamner
Institute (at the time of the two studies, The Hmaner was named CUT) in a humidity- and
temperature-controlled, high-efficiency particulate-air-filtered, mass air-displacement room.
Dams were treated by gavage daily from gestation day (GD) 12-19 with corn oil (vehicle
control) and dibutyl phthalate. Body weights were recorded daily before dosing (GDs 12-19).
The oral treatments were administered on a mg/kg-body weight basis and adjusted daily for
weight changes. Animal doses were calculated through Provantis NT 2000. All calculations
were checked by a second individual and recorded in the investigators' The Hamner Institute
notebooks. Analytical support staff confirmed appropriate dose solutions at the beginning of the
dosing period. Body weights and doses administered were recorded each day in Provantis NT
2000. Pups and dams were euthanized by carbon dioxide asphyxiation.
Fetal tissues for RIA's and RNA isolation were snap frozen in liquid nitrogen and stored
at -80°C. The remaining tissues were either be embedded in optical coherence tomography and
frozen or fixed in formalin for 6 to 24 hours followed by 70% ethanol and then processed and
embedded in paraffin for histological examination within 48 hours. The embedded tissues were
C-2
-------
sectioned at approximately 5 microns and stained with hematoxylin and eosin. The study
pathologist in consultation with the histology staff determined the gross trim, orientation, and
embedding procedure for each tissue. RNA were isolated from the frozen male reproductive
tract, and changes in gene expression were identified by real-time reverse
transcription-polymerase chain reaction (RT-PCR) analysis (following manufacturer's protocols
P/N 402876 and P/N 4304965, Applied Biosystems, Foster City, CA) and in some cases, by
complementary DNA (cDNA) microarray (following manufacturers protocol PT3140, Clontech,
Palo Alto, CA).
Total RNA were treated with DNase I at 37°C for 30 minutes in the presence of RNasin
to remove DNA contamination before cDNA synthesis, followed by heat inactivation at 75°C for
5 minutes. Primer pairs were selected using the program Primer Express and optimized for use
prior to quantification. cDNA were synthesized using random hexamers and TaqMan Reverse
Transcription Reagents according to the manufacturer's suggested protocol. Real-time PCR
(TaqMan) were performed on a Perkin-Elmer/Applied Biosystems 7500 Prism using TaqMan
probe chemistry according to the manufacturer's instructions for quantification of relative gene
expression. Relative differences among treatment groups were determined using the CT method
as outlined in the Applied Biosystems protocol for reverse transcriptase(RT)-PCR. A CT value
was calculated for each sample using the CT value for glyceraldehyde-3-phosphate
dehydrogenase (or an appropriate housekeeping gene) to account for loading differences in the
RT-PCRs.
C.4.2. Microarray Hybridization
Testes from individual fetuses were homogenized in RNA Stat 60 reagent (Tel-Test, Inc.,
Friendswood, TX) and RNA was isolated using the RNeasy Mini Kit (Qiagen, Valencia, CA)
following manufacturer's protocol. RNA integrity was assessed using the Agilent 2100
Bioanalyzer (Agilent Technologies, Palo Alto, CA), and optical density was measured on a
NanoDrop ND 1000 (NanoDrop Technologies, Wilmington, DE). cDNA was synthesized from
2.5 or 3 jig total RNA and purified using the Affymetrix® One-Cycle Target Labeling and
control reagents kit (Affymetrix, Santa Clara, CA) according to manufacturer's protocol. Equal
amounts of purified cDNA per sample were used as the template for subsequent in vitro
transcription reactions for complementary RNA (cRNA) amplification and biotin labeling using
C-3
-------
the Affymetrix GeneChip® IVT labeling kit (Affymetrix) included in the One-Cycle Target
Labeling kit (Affymetrix). cRNA was purified and fragmented according to the protocol
provided with the GeneChip® Sample Cleanup module (Affymetrix). All GeneChip® arrays
were hybridized, washed, stained, and scanned using the Complete GeneChip® Instrument
System according to the Affymetrix Technical Manual.
For immunocytochemistry, tissues were rapidly removed, immersed in 10% (v/v)
neutral-buffered formalin for 24-48 hours, and then stored in ethanol 70% (v/v) until processed.
The reproductive tissues were embedded in paraffin, sectioned at 5 ji, and processed for
immunohistochemistry or stained with hematoxylin and eosin.
Experimental notes and data were entered into uniquely numbered Hamner Institute
laboratory notebooks and three-ring binders along with descriptions of procedures used,
according to SOP# QUA-007. Specimens (RNA and frozen tissue) were retained until analysis
or discarded after a maximum of 1 year after collection. Formalin-fixed tissues, blocks, and
slides were archived at the end of the study. Retention of these materials will be reassessed after
5 years.
€.4.3. Quality Assurance
Both QA and QC procedures are integral parts of our research program. The research
was conducted under the The Hamner Institute Research Quality Standards program. These
standards include (1) scientifically reviewed protocols that are administratively approved for
meeting requirements in data quality, animal care, and safety regulations; (2) standardized
laboratory notebooks and data recording procedures; (3) documented methods or standard
operating procedures for all experimental procedures—including calibration of instruments; (4) a
central managed archive for specimens and documentation; and (5) internal peer review for
scientific quality of abstracts and manuscripts. The Hamner Institute QA and QC processes
assessing overall study performance and records ensure that conduct of the proposed research
satisfies the intended project objectives.
C.4.4. Statistical Analysis
RT-PCR data were analyzed using JMP statistical analysis software (SAS Institute, Gary,
NC). RNA were isolated from at least 3 pups from 3 different dams for each treatment group.
C-4
-------
PCR reactions, radioimmunoassays, and protein analysis were repeated 3-5 times for each
sample. Based on our experience, the number of animal replicates has the statistical power to
detect a significant change in gene expression >20% atp < 0.05. The effect of treatment was
analyzed using a general-linear model regression analysis. Posthoc tests were conducted when
the overall analysis of variance is significant at the/? < 0.05 level using the LS-means procedure
and adjusted for multiple comparisons by Dunnett's method.
Microarray data were analyzed by a linear mixed model with SAS Microarray Solution
software. Perfect-match only data were normalized to a common mean on a Iog2 scale, and a
linear mixed model was then applied for each probe set. Restricted maximum likelihood was
used for estimating the parameters for both the fixed and random effects. Significance was
determined using mixed-model based F-tests (p < 0.05).
C.4.5. Procedures used to Evaluate Success
Uniquely numbered written protocols were prepared and reviewed internally prior to the
start of this study. The content of a protocol includes study design, materials, laboratory
methods, sample collection, handling and custody, record keeping, data analysis and statistical
procedures, animal care requirements, and safety measures. Numbered standardized laboratory
notebooks and guidelines for date recording ensures completeness of data and the ability to
reconstruct the study. An independent QA department manages the overall research data quality.
Manuscripts describing the results of our study were prepared at the completion of each stage of
this study. All manuscripts undergo a rigorous internal peer review that includes review by all
authors, at least two additional PhD- level scientists, the science editor, the division manager,
and the vice president for research.
C.5. PROJECT 2: DATA REVIEW, VERIFICATION, AND VALIDATION
Banalata Sen received the Liu et al. (2005) raw data files from Dr. Kevin Gaido. Two
team members, Dr. Banalata Sen (National Center for Environmental Assessment, Research
Triangle Park [NCEA-RTP]) and Dr. Susan Hester (National Health and Environmental Effects
Research Laboratory [NHEERL]) performed the data analysis at NHEERL, RTF. Barbara
Collins (collins.barbara@epa.gov) at NHEERL-RTP has agreed to serve as the Quality
Assurance Manager (QAM) for the project. Dr. Hester and Sen performed analyses of the "DBF
C-5
-------
only" data that is a subset of the data presented in Liu et al. (2005). The analyses at NHEERL
included statistical filtering to identify of differentially expressed genes and pathway analysis.
C.5.1. Verification of Data upon Receipt
Upon receiving data from Kevin Gaido at the Hamner Institute, EPA NHEERL scientisits
conducted a QA review of the data by gross inspection of the eel files to confirm that the data
had been transmitted successfully. The scientists at the STAR Bioinformatics Center/Rutgers
received the data files from Susan Euling at EPA NCEA who had received the data from Kevin
Gaido at the Hamner Institute. Kevin Gaido gave permission to Susan Euling to provide the data
for these analyses. A review of the data was performed by inspection of the txt files and the
published data to confirm that the data had been transmitted successfully.
C.5.2. Verification of Data Analysis Calculations
EPA NHEERL used a principal component analysis (PCA) to evaluate the within-group
and across-group variance of the six samples. PCA elucidates the separation of different
treatment groups and provides information about whether the data contain significant
information. This was conducted using the raw data eel files in Rosetta Resolver Software. The
analyses were in silico without functional validation (RT-PCR of individual genes).
The Star Bioinformatics Center also performed a principal component analysis (PCA)
and displayed a 3-D plot to evaluate the within-group and across-group variance of the samples.
This was conducted using the txt files in MATLAB® Software. This was an in silico analysis.
The data were normalized to a zero mean and a unity standard deviation over samples. They
assessed the degree of separation for Liu et al. (2005) data. A regular regular t-test and ANOVA
analyses of the data were performed. The filtered data were visualized in a heatmap to determine
the statistically significant subset of genes to provide a differentially expressed gene (DEG) list.
Drs. Susan Hester and Banalata Sen also performed some comparative analyses between
the two outpus (above). The two independent analyses of the same dataset were contrasted with
one another. Correlation plots comparing the LoglO average intensities of control samples vs.
DBF treated samples was performed in order to determine the noise in both groups. Average
background signal and scaling factors will be applied based on the vendor recommendations.
QC plots will be made to determine the relationship between light intensity and each genechip.
C-6
-------
C.6. PROJECT 3: DATA REVIEW, VERIFICATION, AND VALIDATION
This project analyzed the time-course data from Thompson et al. (2005) dataset to then
build a regulatory network model. The STAR Center's internal QA/QC procedures are
implemented and monitored by a QA official, Clifford Weisel (weisel@eohsi.rutgers.edu), at
Rutgers University that reports to the National Center for Environmental Research (NCER), the
granting organization for the STAR program.
C.6.1. Verification of Data upon Receipt
Data were received from Susan Euling at EPA who had received the data from Kevin
Gaido at the Hamner Institute. Kevin Gaido gave permission to Susan Euling to provide the data
for these analyses. A review of the data was performed by inspection of the txt files and the
published data to confirm that the data had been transmitted successfully.
C.6.2. Verification of Data Analysis Calculations
A principal component analysis (PCA) was performed and a 3-D plot was displayed to
evaluate the within-group and across-group variance of the samples. This was conducted using
the txt files in MATLAB® Software. This was an in silico analysis. The data were normalized
to a zero mean and a unity standard deviation over samples. They assessed the degree of
separation for the Thompson et al. (2005) data. A regular regular t-test and ANOVA analyses of
the data were performed. The filtered data will be visualized in a heatmap to determine the
statistically significant subset of genes to provide a differentially expressed gene (DEG) list.
C-7
-------
GLOSSARY
Amplified Fragment Length Polymorphism Polymerase Chain Reaction (AFLP-PCR or
AFLP): A PCR-based DNA fingerprinting tool that is a highly sensitive method for detecting
DNA polymorphisms.
Benchmark Dose (BMD) or Concentration (BMC): A dose or concentration that produces a
predetermined change in response rate of an adverse effect (called the benchmark response or
BMR) compared to background.
Complementary DNA (cDNA): DNA synthesized from a mature mRNA template in a reaction
catalyzed by the enzyme reverse transcriptase.
Copy Number Polymorphism (CNP): The normal variation in the number of copies of a gene
or of sequences of DNA in the genome of an individual.
Exposure: Contact made between a chemical, physical, or biological agent and the outer
boundary of an organism. Exposure is quantified as the amount of an agent available at the
exchange boundaries of the organism (e.g., skin, lungs, gut).
Exposure Assessment: An identification and evaluation of the human population exposed to a
toxic agent, describing its composition and size, as well as the type, magnitude, frequency, route
and duration of exposure.
Expressed Sequence Tag (EST): A short subsequence of a transcribed cDNA sequence,
produced by sequencing of a cloned mRNA representing portions of expressed genes, which can
be used to identify gene transcripts.
Gene Network: An illustration of the interactions between genes and gene products based on
gene expression and other molecular information curated from the published literature.
G-l
-------
Gene Ontology (GO): A bioinformatics initiative of the GO Consortium with the goal of
standardizing terminology for describing gene and gene product characteristics across species
and databases. The GO has developed three structured vocabularies (ontologies), independent of
species, to describe gene products in terms of their associated: 1) biological processes; 2) cellular
components; and 3) molecular functions. The GO also provides tools to access and process these
data.
Genomics: The study of the structure and function of the whole genome. This term can also
refer to "genomic technologies," defined as methods to study the genome at the level of DNA
(including genome sequencing and genotype analysis). Sometimes this term refers more
generally to all of the methods to study the genome (see -omics).
Genomic Technologies: Methods to study the genome including genome sequencing
technologies and genotype analysis.
Hazard Assessment: The process of determining whether exposure to an agent can cause an
increase in the incidence of a particular adverse health effect (e.g., cancer, birth defect) and
whether the adverse health effect is likely to occur in humans.
Hazard Characterization: A description of the potential adverse health effects attributable to a
specific environmental agent, the mechanisms by which agents exert their toxic effects, and the
associated dose, route, duration, and timing of exposure.
Human Health Risk Assessment: The evaluation of scientific information on the hazardous
properties of environmental agents (hazard characterization), the dose-response relationship
(dose-response assessment), and the extent of human exposure to those agents (exposure
assessment). The product of the risk assessment is a statement regarding the probability that
populations or individuals so exposed will be harmed and to what degree (risk characterization).
Key Event: An empirically observable precursor step that is, itself, a necessary element of the
mode of action or is a biologically based marker for such an element.
G-2
-------
Lowest Observed Adverse Effect Level (LOAEL): The lowest exposure level at which there
are biologically significant increases in the frequency or severity of adverse effects between the
exposed population and its appropriate control group.
Lowest Observed Effect Level (LOEL): In a study, the lowest dose or exposure level at which
a statistically or biologically significant effect is observed in the exposed population compared
with an appropriate unexposed control group.
Mechanism of Action: The complete molecular sequence of events between the interaction of
the chemical with the target site and observation of the outcome. Thus, the mechanism of action
can include toxicokinetic and/or toxicodynamic steps.
Metabolic Pathway Network: An illustration of interactions between metabolites derived from
pathway information curated from the published literature.
Metabolomics: The analysis of collections of small molecule metabolic intermediates and
products of diverse biologic processes.
Microarray: A transcriptomics tool for analyzing gene expression that consists of a small
membrane or glass slide containing samples of many genes arranged in a regular pattern.
Microarray Quality Control (MAQC): An FDA project that was developed to provide
quality-control tools, guidelines, and standard operating procedures (SOPs) to the microarray
community in order to avoid procedural failures. To facilitate this effort, the MAQC has
provided the public with large reference data sets and reference RNA samples.
Mode of Action (MOA): One or a sequence of key events, that a particular outcome is
dependent upon (i.e., part of the causal pathway and not a coincident event).
No Observed Adverse Effect Level (NOAEL): The highest exposure level at which there are
no biologically significant increases in the frequency or severity of adverse effect between the
G-3
-------
exposed population and its appropriate control; some effects may be produced at this level, but
they are not considered adverse or precursors of adverse effects.
No Observed Effect Level (NOEL): An exposure level at which there are no statistically or
biologically significant increases in the frequency or severity of any effect between the exposed
population and its appropriate control.
-omics: A suffix that is used as a general term for the genome-wide study of biological
information objects (or "omes"), such as toxicogenome, proteome, and metabolome; a term
referring to all of the methods for assessing the genome including genomics, metabolomics,
proteomics, and transcriptomics.
Physiologically Based Pharmacokinetic (PBPK) Model: A model that estimates the dose to a
target tissue or organ by taking into account the rate of absorption into the body, distribution
among target organs and tissues, metabolism, and excretion.
Principal Component Analysis (PCA): A technique for analysis of multivariate data involving
a mathematical procedure that transforms a number of possibly correlated variables into a
smaller number of uncorrelated variables, called principal components.
Proteomics: The study of the protein complement of the genome of an organism.
Reverse Transcription-Polymerase Chain Reaction (RT-PCR): A two-step process for
converting mRNA to cDNA, using the enzyme reverse transcriptase, and the subsequent PCR
amplification of the reversely transcribed DNA using the enzyme DNA polymerase.
Serial Analysis of Gene Expression (SAGE): A technique based on sequencing strings of short
expressed sequence tags (ESTs) representing both the identity and the frequency of occurrence
of specific sequences within the transcriptome. This method allows the entire collection of
transcripts to be catalogued without assumptions about which transcripts are actually expressed.
G-4
-------
Single-Nucleotide Polymorphism (SNP): A DNA sequence variation occurring when a single
nucleotide in the genome (or other shared sequence) differs between members of a species or
between paired chromosomes in an individual.
Singular Value Decomposition (SVD): A technique for the analysis of multivariate data where
a rectangular, real or complex matrix, is factorized. SVD has been extensively used in
microarray data analysis in order to achieve a linear projection of the data and represent these
data in a reduced dimensionality space which further enables clustering and visualization of gene
expression data patterns.
Toxicogenomics: The application of genomic technologies to study the adverse effects of
environmental and pharmaceutical chemicals on human health and the environment.
Transcriptomics: A set of techniques to measure genome-wide mRNA expression that are used
to understand the expression of genes and pathways involved in biological processes; also called
"gene expression profiling."
G-5
-------
REFERENCES
Aardema, MJ; MacGregor, JT. (2002) Toxicology and genetic toxicology in the new era of "toxicogenomics":
impact of "-omics" technologies. Mutat Res 499:13-25.
Alter, O; Brown, PO; Botstein, D. (2000) Singular value decomposition for genome-wide expression data
processing and modeling. Proc Natl Acad Sci USA, 97(18): 10101-10106.
Andersen, ME; Thomas, RS; Gaido, KW; et al. (2005) Dose-response modeling in reproductive toxicology in the
systems biology era. Reprod Toxicol 19(3):327-337.
Andersen, ME; Clewell, HJ 3rd; Bermudez, E; et al. (2008) Genomic signatures and dose-dependent transitions in
nasal epithelial responses to inhaled formaldehyde in the rat. Toxicol Sci. 105(2):368-383.
Anderson, NL; Anderson, NG. (1998) Proteome and proteomics: new technologies, new concepts, and new words.
Electrophoresis 19(11):1853-1861.
Ball, CA; Brazma, A; Causton, H; et al. (2004) Submission of microarray data to public repositories. PLoS Biol
2(9):E317.
Barlow, NJ; Foster, PM. (2003) Pathogenesis of male reproductive tract lesions from gestation through adulthood
following in utero exposure to di(n-butyl) phthalate. Toxicol Pathol 31(4):397-410.
Barlow, NJ; Phillips, SL; Wallace, DG; et al. (2003) Quantitative changes in gene expression in fetal rat testes
following exposure to di(n-butyl) phthalate. Toxicol Sci 73(2):431-441.
Barlow, NJ; Mclntyre, BS; Foster, PM. (2004) Male reproductive tract lesions at 6, 12, and 18 months of age
following in utero exposure to di(n-butyl) phthalate. Toxicol Pathol 32:79-90.
BASF. (1992) Study on the oral toxicity of dibutyl phthalate in Wistar rats. Administration via the diet over
3 months. Prepared by BASF Corporation, Ludwigshafen, Germany for Eastman Kodak Company;
Project No. 31S0449//89020; Submitted under TSCA Section 8D; EPA Document No. 86-920000903;
NTIS OTS0535640 (As cited in NTP-CERHR, 2000).
Bodin, L; Beaune, PH; Loriot, MA. (2005) Determination of cytochrome P450 2D6 (CYP2D6) gene copy number
by real-time quantitative PCR. J Biomed Biotechnol 2005(3):248-253.
Boverhof, DR; Zacharewski, TR. (2006) Toxicogenomics in risk assessment: applications and needs. Toxicol Sci
89(2):352-360.
Bowman, CJ; Turner, KJ; Sar, M; et al. (2005) Altered gene expression during rat Wolffian duct development
following di(n-butyl) phthalate exposure. Toxicol Sci 86:161-174.
Brazma, A; Hingamp, P; Quackenbush, J; et al. (2001) Minimum information about a microarray experiment
(MIAME)—toward standards for microarray data. Nat Genet 29(4):365-371.
Buckley, PG; Mantripragada, KK; Piotrowski, A; et al. (2005) Copy-number polymorphisms: mining the tip of an
iceberg. Trends Genet 21(6):315-317.
Canales, RD; Luo, Y; Willey, JC; et al. (2006) Evaluation of DNA microarray results with quantitative gene
expression platforms. Nat Biotechnol 24(9): 1115-1122. doi:10.1038/nbt!236. Available online at
http://www.nature.com/nbt/journal/v24/n9/pdf/nbtl236.pdf.
Cangelosi, R; Goriely, A. (2007) Component retention in principal component analysis with application to cDNA
microarray data. Biol Direct 2:2.
R-l
-------
Carruthers, CM; Foster, P. (2005) Critical window of male reproductive tract development in rats following
gestational exposure to di-n-butyl phthalate. Birth Defects Res B Dev Reprod Toxicol 74(3):277-285.
Casanova, M; Morgan, KT; Gross, EA; et al. (1994) DNA-protein cross-links and cell replication at specific sites in
the nose of F344 rats exposed subchronically to formaldehyde. Fundam Appl Toxicol 23(4):525-536.
Casciano, DA; Woodcock, J. (2006) Empowering microarrays in the regulatory setting (forward). Nat Biotechnol
24(9): 1103.
Ceder, R; Merne, M; Staab, C; et al. (2007) The application of normal, SV40 T-antigen-immortalized and tumor-
derived oral keratinocytes, under serum-free conditions, to the study of the probability of cancer progression as a
result of environmental exposure to chemicals. AlternLab Anim 35(6):621-639.
Chan, VS; Theilade, MD. (2005) The use of toxicogenomic data in risk assessment: a regulatory perspective. Clin
Toxicol 43(2): 121-126.
Choe, SE; Boutros, M; Michelson, AM; et al. (2005) Preferred analysis methods for Affymetrix GeneChips revealed
by a wholly defined control dataset. Genome Biol 6(2):R16.
Clemente, JC; Satou, K; Valiente, G. (2005) Reconstruction of phylogenetic relationships from metabolic pathways
based on the enzyme hierarchy and the gene ontology. Genome Inform 16(2):45-55.
Coen, M; Lenz, EM; Nicholson, JK; et al. (2003) An integrated metabonomic investigation of acetaminophen
toxicity in the mouse using NMR spectroscopy. Chem Res Toxicol 16(3):295-303.
Cunningham, ML; Bogdanffy, MS; Zacharewski, T; et al. (2003) Workshop overview: use of genomic data in risk
assessment. Toxicol Sci73(2):209-215.
Daston, GP. (2007) Genomics and developmental risk assessment. Birth Defects Res A Clin Mol Teratol 79: 1-7.
Daston, GP. (2008) Gene expression, dose-response, and phenotypic anchoring: applications for toxicogenomics in
risk assessment. Toxicol Sci. 105(2):233-234.
David, RM. (2006) Proposed mode of action for in utero effects of some phthalate esters on the developing male
reproductive tract. Toxicol Pathol 34(3):209-19. Available online at
http://tpx.sagepub.eom/cgi/reprint/34/3/209.pdf.
Dix, D J; Gallagher, K; Benson, WH; et al. (2006) A framework for the use of genomics data at the EPA. Nat
Biotechnol 24(9):1108-1111. Available online at http://www.nature.com/nbt/journal/v24/n9/pdf/nbt0906-1108.pdf.
Donate, MT; Lahoz, A; Castell, JV; et al. (2008) Cell lines: a tool for in vitro drug metabolism studies. Curr Drug
Metab 9(1): 1-11.
Dudoit, S; Shaffer, JP; Boldrick, JC. (2003) Multiple hypothesis testing in microarray experiments. Stat Sci
Ekins, S; Nikolsky, Y; Nikolskaya, T. (2005) Techniques: application of systems biology to absorption, distribution,
metabolism, excretion and toxicity. Trends Pharmacol Sci 26(4):202-209.
Ellinger-Ziegelbauer, H; Stuart, B; Wahle, B; etal. (2005) Comparison of the expression profiles induced by
genotoxic and nongenotoxic carcinogens in rat liver. Mutat Res 575(l-2):61-84.
El-Masri, H; Constan, AA; Ramsdell, HS; et al. (1996) Physiologically based pharmacodynamic modeling of an
interaction threshold between trichloroethylene and 1, 1-dichloroethylene in Fischer 344 rats. Toxicol Appl
Pharmacol 141:124-132.
El-Masri, H; Bell, DA; Portier, CJ. (1999) Effects of glutathione transferase theta polymorphism on the risk
estimates of dichloromethane to humans. Toxicol Appl Pharmacol 158(3):221-230.
R-2
-------
Ema, M; Miyawaki, E. (2001a) Adverse effects on development of the reproductive system in male offspring of rats
given monobutyl phthalate, a metabolite of dibutyl phthalate, during late pregnancy. Reprod Toxicol
15(2): 189-194.
Ema, M; Miyawaki, E. (200 Ib) Effects of monobutyl phthalate on reproductive function in pregnant and
pseudopregnant rats. Reprod Toxicol 15(3):261-267.
Ema, M; Amano, H; Itami, T; et al. (1993) Teratogenic evaluation of di-n-butyl phthalate in rats. Toxicol Lett
69(2): 197-203.
Ema, M; Amano, H; Ogawa, Y. (1994) Characterization of the developmental toxicity of di-n-butyl phthalate in rats.
Toxicology 86(3): 163-174.
Ema, M; Kurosaka, R; Amano, H; et al. (1995) Developmental toxicity evaluation of mono-n-butyl phthalate in rats.
Toxicol Lett 78(2):101-106.
Ema, M; Kurosaka, R; Harazono, A; et al. (1996) Phase specificity of developmental toxicity after oral
administration of mono-n-butyl phthalate in rats. Arch Environ Contam Toxicol 31(2): 170-176.
Ema, M; Harazono, A; Miyawaki, E; et al. (1997) Developmental effects of di-n-butyl phthalate after a single
administration in rats. JAppl Toxicol 17(4):223-229.
Ema, M; Miyawaki, E; Kawashima, K. (1998) Further evaluation of developmental toxicity of di-n-butyl phthalate
following administration during later pregnancy in rats. Toxicol Lett 98(l-2):87-93.
Ema, M; Miyawaki, E; Kawashima, K. (2000a) Effects of dibutyl phthalate on reproductive function in pregnant and
pseudopregnant rats. Reprod Toxicol 14(1): 13-19.
Ema, M; Miyawaki, E; Kawashima, K. (2000b) Critical period for adverse effects on development of reproductive
system in male off spring of rats given di-n-butyl phthalate during later pregnancy. ToxicolLett lll(3):271-278.
Fang, H; Tong, W; Perkins, R; et al. (2005) Bioinformatics approaches for cross-species liver cancer analysis based
on microarray gene expression profiling. BMC Bioinformatics 6(Suppl 2):S6.
Fennell, TR; Krol, WL; Sumner, SC; et al. (2004) Pharmacokinetics of dibutylphthalate in pregnant rats. Toxicol
Sci82(2):407-418.
Ferrara, D; Hallmark, N; Scott, H; et al. (2006) Acute and long-term effects of in utero exposure of rats to
di(n-butyl) phthalate on testicular germ cell development and proliferation. Endocrinology 147(11):5352-5362.
Fisher, JS; Macpherson, S; Marchetti, N; et al. (2003) Human 'testicular dysgenesis syndrome': a possible model
using in-utero exposure of the rat to dibutyl phthalate. Hum Reprod 18(7): 1389-1394.
Forst, CV. (2002) Network genomics~a novel approach for the analysis of biological systems in the post-genomic
era. MolBiolRep29(3):265-280.
Forst, CV; Schulten, K. (1999) Evolution of metabolisms: a new method for the comparison of metabolic pathways
using genomics information. J Comput Biol 6(3-4):343-360.
Foster, PMD. (2005) Mode of action: impaired fetal Leydig cell function—effects on male reproductive
development produced by certain phthalate esters. Crit Rev Toxicol 35(8-9):713-719.
Frueh, FW. (2006) Impact of microarray data quality on genomic data submissions to the FDA. Nature Biotechnol
24(9): 1105-1107. Available online at http://www.nature.com/nbt/journal/v24/n9/pdf/nbt0906-l 105.pdf.
Frueh, FW; Huang, SM; Lesko, LJ. (2004) Regulatory acceptance of toxicogenomics data. Environ Health Perspect
112(12):A663-A664.
R-3
-------
Geng, J; Strobel, HW. (1995) Identification of inducible mixed function oxidase system in rat glioma C6 cell line. J
Neurochem 65(2):554-563.
Gibbs, RA; Weinstock, GM; Metzker, ML; et al. (2004) Genome sequence of the Brown Norway Rat yields insights
into mammalian evolution. Nature 428(6982):493-521.
Ginsberg, G; Smolenski, S; Hattis, D; et al. (2002) Population distribution of aldehyde dehydrogenase-2 genetic
polymorphism: implications for risk assessment. Regul Toxicol Pharmacol 36(3):297-309.
Golub, TR; Slonim, DK; Tamayo, P; et al. (1999) Molecular classification of cancer: class discovery and class
prediction by gene expression monitoring. Science 286(5439):531-537.
Gray, LE, Jr; Wolf, C; Lambright, C; et al. (1999) Administration of potentially antiandrogenic pesticides
(procymidone, linuron, chlozolinate, p,p'-DDE, and ketoconazole) and toxic substances (dibutyl- and diethylhexyl
phthalate, PCB 169, and ethane dimethane sulphonate) during sexual differentiation produces diverse profiles of
reproductive malformations in the male rat. Toxicol Indust Health 15(1-2):94-118.
Gray, LE, Jr; Ostby, J; Furr, J; et al. (2000) Perinatal exposure to the phthalates DEHP, BBP, and DINP, but not
DEP, DMP, or DOTP, alters sexual differentiation of the male rat. Toxicol Sci 85(2):350-365.
Gray, LE, Jr.; Ostby, J; Furr, J; et al. (2001) Effects of environmental antiandrogens on reproductive development in
experimental animals. Hum Reprod Update 7(3):248-264.
Guerreiro, N; Staedtler, F; Grenet, O; et al. (2003) Toxicogenomics in drug development. Toxicol Pathol
31(5):471-479.
Guo, L; Lobenhofer, EK; Wang, C; et al. (2006) Rat toxicogenomic study reveals analytical consistency across
microarray platforms. Nat Biotechnol 24(9): 1162-1169. Available online at
http://www.nature.com/nbt/journal/v24/n9/pdf/nbtl238.pdf.
Hackett, JL; Lesko, LJ. (2003) Microarray data—the US FDA, industry, and academia. Nat Biotechnol
21(7):742-743.
Hamadeh, HK; Bushel, PR; Jayadev, S; et al. (2002a) Gene expression analysis reveals chemical-specific profiles.
Toxicol Sci 67(2) :219-231.
Hamadeh, HK; Bushel, PR; Jayadev, S; et al. (2002b) Prediction of compound signature using high density gene
expression profiling. Toxicol Sci 67(2):232-240.
Hanisch, D; Zien, A; Zimmer, R; et al. (2002) Co-clustering of biological networks and gene expression data.
Bioinformatics 18 Suppl 1:S145-S154.
Hedberg, JJ; Grafstrom, RC; Vondracek, M; et al. (2001) Micro-array chip analysis of carbonyl-metabolizing
enzymes in normal, immortalized and malignant human oral keratinocytes. Cell Mol Life Sci 58(11): 1719-1726.
Heymans, M; Singh, AK. (2003) Deriving phylogenetic trees from the similarity analysis of metabolic pathways.
Bioinformatics 19 (Suppl I):il38-il46.
Higuchi, TT; Palmer, JS; Gray, LE, Jr; et al. (2003) Effects of dibutyl phthalate in male rabbits following inutero,
adolescent, or postpubertal exposure. Toxicol Sci 72(2):301-313.
Ji, H; Davis, RW. (2006) Data quality in genomics and microarrays. Nat Biotechnol 24(9):1112-1113. Available
online at http://www.nature.com/nbt/journal/v24/n9/pdf/nbt0906-l 112.pdf.
Kanehisa, M; Goto, S. (2000) KEGG: kyoto encyclopedia of genes and genomes. Nucleic Acids Res 28(1):27-30.
Kavlock, R; Ankley, GT; Collette, T; et al. (2005) Computational toxicology: framework, partnerships, and program
development: September 29-30, 2003, Research Triangle Park, North Carolina. Reprod Toxicol 19(3):265-280.
R-4
-------
Kennedy, S. (2002) The role of proteomics in toxicology: identification of biomarkers of toxicity by protein
expression analysis. Biomarkers 7(4):269-290.
Kienhuis, AS; van de Poll, MCG; Wortelboer, H; et al. (2009) Parallelogram approach using rat-human in vitro and
rat in vivo toxicogenomics predits acetaminophen-induced hepatotoxicity in humans. Toxicol Sci 107(2):544-552.
Kim, T; Shin, J; Lee, S; et al. (2004) Effects of inutero exposure of diethylstilbestrol and dibutyl phthalate on the
testis descent in rat offspring. Abstract # 574. Toxicologist 78:118.
Kleymenova, E; Dei Rossi, E; Liu, K; et al. (2004) Developmental exposure to di-n-butyl phthalate affects cord
organization and Sertoli cell-gonocyte interactions in the fetal rat testis. Abstract # 926. Toxicologist 78:190.
Kleymenova, E; Swanson, C; Gaido, KW. (2005a) Low dose cellular responses in the fetal rat testis in utero
exposed to di-n-butyl phthalate. Abstract #1372. Toxicologist 79:281.
Kleymenova, E; Swanson, C; Boekelheide, K; et al. (2005b) Exposure in utero to di(n-butyl) phthalate alters the
vimentin cytoskeleton of fetal rat Sertoli cells and disrupts Sertoli cell-gonocyte contact. Biol Reprod
73(3):482-490.
Klonisch, T; Fowler, PA; Hombach-Klonisch, S. (2004) Molecular and genetic regulation of testis descent and
external genitalia development. Dev Biol 270:1-18.
Lee, K-Y; Shibutani, M; Takagi, H; et al. (2004) Diverse developmental toxicity of di-n-butyl phthalate in both
sexes of rat offspring after maternal exposure during the period from late gestation through lactation. Toxicology
203(l-3):221-238.
Lehmann, KP; Phillips, S; Sar, M; et al. (2004) Dose-dependent alterations in gene expression and testosterone
synthesis in the fetal testes of male rats exposed to di (n-butyl) phthalate. Toxicol Sci 81(l):60-68.
Leighton, JK. (2005) Application of emerging technologies in toxicology and safety assessment: regulatory
perspectives. Int J Toxicol 24(3): 153-155.
Liu, K; Lehmann, KP; Sar, M; et al. (2005) Gene expression profiling following in utero exposure to phthalate esters
reveals new gene targets in the etiology of testicular dysgenesis. Biol Reprod 73(1): 180-192.
Lobenhofer, EK; Bushel, PR; Afshari, CA; et al. (2001) Progress in the application of DNA microarrays. Environ
Health Perspect 109(9):881-891.
Mahood, IK; Hallmark, N; McKinnell, C; et al. (2005) Abnormal Leydig cell aggregation in the fetal testis of rats
exposed to di (n-butyl) phthalate and its possible role in testicular dysgenesis. Endocrinol 146(2):613 -623.
Main, KM; Mortensen, GK; Kaleva, MM; et al. (2006) Human breast milk contamination with phthalates and
alterations of endogenous reproductive hormones in infants three months of age. Environ Health Perspect
114(2):270-276.
Manoli, T; Gretz, N; Grone, HJ; et al. (2006) Group testing for pathway analysis improves comparability of different
microarray datasets. Bioinformatics 22(20):2500-2506.
McKinnell, C; Sharpe, RM; Mahood, K; et al. (2005) Expression of insulin-like factor 3 protein in the rat testis
during fetal and postnatal development and in relation to cryptorchidism induced by in utero exposure to di (n-
Butyl) phthalate. Endocrinology 146(10):4536-4544.
Miller, WL. (1988) Molecular biology of steroid hormone synthesis. Endocr Rev 9(3):295-318.
Miller, WL. (2008) Steroidogenic enzymes. Endocr Dev 13:1-18.
Moggs, JG. (2005) Molecular responses to xenoestrogens: mechanistic insights from toxicogenomics. Toxicology
213(3):177-193.
R-5
-------
Moggs, JG; Tinwell, H; Spurway, T; et al. (2004) Phenotypic anchoring of gene expression changes during
estrogen-induced uterine growth. Environ Health Perspect 112(16): 1589 -1606.
Mootha, VK; Lindgren, CM; Eriksson, KF; et al. (2003) PGC-lalpha-responsive genes involved in oxidative
phosphorylation are coordinately downregulated in human diabetes. Nat Genet 34(3):267-73.
Mylchreest, E; Cattley, RC; Foster, PMD. (1998) Male reproductive tract malformations in rats following
gestational and lactational exposure to di(n-butyl) phthalate: an antiandrogenic mechanism? Toxicol Sci 43:47-60.
Mylchreest, E; Sar, M; Cattley, RD; et al. (1999) Disruption of androgen-regulated male reproductive development
by di(n-butyl) phthalate during late gestation in rats is different from flutamide. Toxicol Appl Pharmacol
156(2):81-95.
Mylchreest, E; Wallace, DG; Cattley, RC; et al. (2000) Dose-dependent alterations in androgen-regulated male
reproductive development in rats exposed to di(n-butyl) phthalate during late gestation. Toxicol Sci 55:143-151.
Mylchreest, E; Sar, M; Wallace, DG; et al. (2002) Fetal testosterone insufficiency and abnormal proliferation of
Leydig cells and gonocytes in rats exposed to di(n-butyl) phthalate. Repro Toxicol 16:19-28.
Naciff, JM; Jump, ML; Torontali SM; et al. (2002) Gene expression profile induced by 17alpha-ethynyl estradiol,
bisphenol A, and genistein in the developing female reproductive system of the rat. Toxicol Sci 68(1): 184-199.
Naciff, JM; Overmann, GJ; Torontali, SM; et al. (2003) Gene expression profile induced by 17 alpha-ethynyl
estradiol in the prepubertal female reproductive system of the rat. Toxicol Sci 72(2):314 -330.
Naciff, JM; Hess, KA; Overmann, GJ; et al. (2005) Gene expression changes induced in the testis by transplacental
exposure to high and low doses of 17a-ethynyl estradiol, genistein, or bisphenol A. Toxicol Sci 86(2):396-416.
NRC (National Research Council) (2007a) Applications of toxicogenomic technologies to predictive toxicology and
risk assessment. Washington, DC: National Academy Press.
NRC (National Research Council). (2007b) Toxicity testing in the 21st century. Washington, DC: National
Academy Press.
NTP (National Toxicology Program). (1991) Final report on the reproductive toxicity of di-n-butyl phthalate (Cas.
No. 84-74-3) in Sprague-Dawley rats. U.S. Department of Health and Human Services, Public Health Service,
National Institutes of Health, National Toxicity Program. Research Triangle Park, NC. NTP Report # RACB89012.
NTIS#PB92-111996.
NTP (National Toxicology Program). (1995) NTP technical report on toxicity studies of dibutyl phthalate
administered in feed to F344/N rats and B6C3F1 mice. Toxicity Report Series Number 30. U.S. Department of
Health and Human Services, Public Health Service, National Institutes of Health, National Toxicity Program.
Research Triangle Park, NC. Available online at http://ntp.niehs.nih.gov/ntp/htdocs/ST_rpts/tox030.pdf.
NTP-CERHR (National Toxicology Program - Center for the Evaluation of Risks to Human Reproduction). (2000)
NTP-CERHR monograph on the potential human reproductive and developmental effects of di-n-butyl phthalate
(DBF). National Toxicology Program, U.S. Department of Health and Human Services, Research Triangle Park,
NC. Available online at http://cerhr.niehs.nih.gov/chemicals/phthalates/dbp/DBP_Monograph_Final.pdf.
Oberemm, A; Onyon, L; Gundert-Remy, U. (2005) How can toxicogenomics inform risk assessment? Toxicol Appl
Pharmacol 207(2 Suppl):592-598.
Omura, T; Morohashi, K. (1995) Gene regulation of steroidogenesis. J Steroid Biochem Mol Biol 53(l-6):19-25.
Pandey, A; Mann, M. (2000) Proteomics to study genes and genomes. Nature 405(6788):837 -846.
R-6
-------
Patterson, TA; Lobenhofer, EK; Fulmer-Smentek, FB; et al. (2006) Performance comparison of one-color and two-
color platforms within the Microarray Quality Control (MAQC) project. Nat Biotechnol 24(9): 1140 -1150.
Available online at http://www.nature.com/nbt/journal/v24/n9/pdf/nbtl242.pdf.
Pennie, WD; Pettit, SD; Lord, PG. (2004) Toxicogenomics in risk assessment - An overview of an ILSIHESI
collaborative research program. Environ Health Perspect 112(4):417-419.
Pettit, SD; Lord, PG; Pennie, WD. (2003) The application of genomics to mechanism-based risk assessment—
overview of an ILSI HESI collaborative research program. J Toxicol Sci 28(4):220 -221.
Plummer, S; Hallmark, N; Sharpe, R; et al. (2005) Identification of gene clusters and signaling pathways affected by
dibutyl phthalate—nuclear receptor interactions in fetal rat testes. Abstract #2258. Toxicologist 79:463.
Plummer, S; Sharpe, RM; Hallmark, N; et al. (2007) Time-dependent and compartment-specific effects of in utero
exposure to di(n-butyl) phthalate on gene/protein expression in the fetal rat testis as revealed by transcription
profiling and laser capture microdissection. Toxicol Sci 97(2):520-352.
Rahnenfuhrer, J; Domingues, FS; Maydt, J; et al. (2004) Calculating the statistical significance of changes in
pathway activity from gene expression data. Stat Appl Genet Mol Biol 3:Article 16.
Raunio, H; Hakkola, J; Hukkanen, J; et al. (1999) Expression of xenobiotic-metabolizing CYPs in human pulmonary
tissue. Exp Toxicol Pathol 51(4-5):412-417.
Reynolds, VL. (2005) Applications of emerging technologies in toxicology and safety assessment. Int J Toxicol
24(3):135-137.
Robinson, DE; Pettit, SD; Morgan, DG. (2003) Use of genomics in mechanism based risk assessment. In: Inoue, T;
Pennie, WD; eds., Toxicogenomics. Tokyo: Springer-Verlag, pp. 194-203.
Saillenfait, AM; Payan, JP; Fabry, JP; et al. (1998) Assessment of the developmental toxicity, metabolism, and
placenta! transfer of di-n-butyl phthalate administered to pregnant rats. Toxicol Sci 45(2):212-224.
Sayers, EW; Barrett, T; Benson, DA; et al. (2008) Database resources of the National Center for Biotechnology
Information. Nucleic Acids Res. 37(9):3124.
Schlosser, PM; Lilly, PD; Conolly, RB; et al. (2003) Benchmark dose risk assessment for formaldehyde using
airflow modeling and a single-compartment, DNA-protein cross-link dosimetry model to estimate human equivalent
doses. Risk Anal 23(3):473-487.
Searls, DB. (2003) Pharmacophylogenomics: genes, evolution and drug targets. Nat Rev Drug Discov. 2(8):613-
623.
Shastry, BS. (2006) Pharmacogenetics and the concept of individualized medicine. Pharmacogenomics J 6:16 -21.
Shi, L; Tong, W; Goodsaid, F; et al. (2004) QA/QC: challenges and pitfalls facing the microarray community and
regulatory agency. Expert Rev Mol Diagn 4(6):761-777.
Shi, L; Reid, LH; Jones, WD; et al. (2006) The MicroArray Quality Control (MAQC) project shows inter- and
intraplatform reproducibility of gene expression measurements. Nat Biotechnol 24(9): 1151 -1161.
Shippy, R; Fulmer-Smentek, S; Jensen, RV; et al. (2006) Using RNA sample titrations to assess microarray platform
performance and normalization techniques. Nat Biotechnol 24(9): 1123 -1131. Available online at
http://www.nature.com/nbt/journal/v24/n9/abs/nbtl241.html.
Shultz, VD; Phillips, S; Sar, M; et al. (2001) Altered gene profiles in fetal rat testes after in utero exposure to
di(n-butyl) phthalate. Toxicol Sci 64(2):233-242.
R-7
-------
Sladek, R; Bader, JA; Giguere, V. (1997) The orphan nuclear receptor estrogen-related receptor alpha is a
transcriptional regulator of the human medium-chain acyl coenzyme A dehydrogenase gene. Mol Cell Biol
17(9):5400-5409.
Simmons, PT; Portier, CJ. (2002) Toxicogenomics: the new frontier in risk analysis. Carcinogenesis 23(6):903-905.
Sohler, F; Hanisch, D; Zimmer, R. (2004) New methods for joint analysis of biological networks and expression
data. Bioinformatics 20(10):1517-1521.
Smith, LL. (2001) Key challenges for toxicologists in the 21st century. Trends Pharmacol Sci 22(6):281-285.
Srivastava, SP; Srivastava, S; Saxena, DK; et al. (1990) Testicular effects of di-n-butyl phthalate (DBF):
biochemical and histopathological alterations. Arch Toxicol 64:148-152.
Staab, CA; Ceder, R; Roberg, K; et al. (2008) Serum-responsive expression of carbonyl-metabolizing enzymes in
normal and transformed human buccal keratinocytes. Cell Mol Life Sci 65(22):3653-3663.
Stocco, DM; Wang, X; Jo, Y; et al. (2005) Multiple signaling pathways regulating steroidogenesis and steroidogenic
acute regulatory protein expression: More complicated than we thought. Mol Endocrinol 19(11): 2647-2659.
Swan, SH; Main, KM; Liu, F; et al. (2005) Decrease in anogenital distance among male infants with prenatal
phthalate exposure. Environ Health Perspect 113(8): 1056-1061. Erratum in: Environ Health Perspect
113(9):A583.
Swanson, HI. (2004) Cytochrome P450 expression in human keratinocytes: an aryl hydrocarbon receptor
perspective. Chem Biol Interact 149(2-3):69-79.
Thomas, RS; Allen, BC; Nong, A; et al. (2007) A method to integrate benchmark dose estimates with genomic data
to assess the functional effects of chemical exposure. Toxicol Sci 98(l):240-248.
Thompson, CJ; Ross, SM; Gaido, KW. (2004) Di(n-butyl) phthalate impairs cholesterol transport and
steroidogenesis in the fetal rat testis through a rapid and reversible mechanism. Endocrinology 145(3): 1227-1237.
Thompson, CJ; Ross, SM; Hensley, J; et al. (2005) Differential steroidogenic gene expression in the fetal adrenal
versus the testis and rapid and dynamic response of the fetal testis to di(n-butyl) phthalate. Biol Reprod
73(5):908-917.
Thybaud, V; Le Fevre, A; Boitier, E. (2007) Application of toxicogenomics to genetic toxicology risk assessment.
Environ Mol Mutagen 48:369-379.
Tilton, SC; Orner, GA; Benninghoff, AD; et al. (2008) Genomic profiling reveals an alternate mechanism for
hepatic tumor promotion by perfluorooctanoic acid in rainbow trout. Environ Health Perspect 116(8): 1047-1055.
Available online at http://www.ehpomine.org/docs/2008/! 1190/abstract.html.
Tomfohr, J; Lu, J; Kepler, TB. (2005) Pathway level analysis of gene expression using singular value
decomposition. BMC Bioinformatics 6:225.
Tong, W; Lucas, AB; Shippy, R; et al. (2006) Evaluation of external RNA controls for the assessment of microarray
performance. Nat Biotechnol 24(9): 1132-1139. Available online at
http://www.nature.com/nbt/journal/v24/n9/pdf/nbtl237.pdf.
U.S. EPA (Environmental Protection Agency). (1995) Reregistration eligibility decision (RED): Linuron. Office of
Prevention, Pesticides and Toxic Substances, Washington, DC. EPA 738-R-95-003. Available from the National
Technical Information Service, Springfield, VA, and online at http://www.epa.gov/oppsrrdl/REDs/0047.pdf.
U.S. EPA (Environmental Protection Agency). (1998) Reregistration eligibility decision (RED): Iprodione. Office of
Prevention, Pesticides and Toxic Substances, Washington, DC. EPA 738-R-98-017. Available from the National
R-8
-------
Technical Information Service, Springfield, VA, and online at
http://www.epa.gov/oppsrrdl/REDs/factsheets/2335fact.pdf.
U.S. EPA (Environmental Protection Agency). (2000a) Reregistration Eligibility Decision (RED): Vinclozolin.
Office of Prevention, Pesticides and Toxic Substances, Washington, DC. EPA 738-R-00-023. Available from the
National Technical Information Service, Springfield, VA, and online at
http://www.epa.gov/opp00001/reregistration/REDs/2740red.pdf.
U.S. EPA (Environmental Protection Agency). (2002a) Interim genomics policy. Prepared by the Science Policy
Council, Washington, DC. Available from the National Technical Information Service, Springfield, VA, and online
at http://www.epa.gov/osa/spc/genomics.htnx
U.S. EPA (Environmental Protection Agency). (2002b) Tolerance reassessment progress and risk management
decision (TRED) for linuron. Office of Prevention, Pesticides and Toxic Substances, Washington, DC. Available
online at http://www.epa.gov/oppsrrdl/REDs/linuron_tred.pdf.
U.S. EPA (Environmental Protection Agency). (2004a) A framework for a computational toxicology research
program. Office of Research and Development, National Center for Computational Toxicology, Washington, DC.
Available online at http://www.epa.gov/comptox/comptox_framework.html..
U.S. EPA (Environmental Protection Agency). (2004b) Potential implications of genomics for regulatory and risk
assessment applications at EPA. Genomics Task Force Workgroup, Science Policy Council, Washington, DC; EPA
100/B-04/002. Available online at http://www.epa.gov/OSA/pdfs/EPA-Genomics-White-Paper.pdf.
U.S. EPA (Environmental Protection Agency). (2004c) Final Report for Acetochlor. Part 2: Mode of Action
Assessment Document: Evaluation of the Mode of Action of Acetochlor for Nasal Olfactory Epithelium Tumors in
Rats and its Relevance to Human Cancer Risk Assessment. August 31, 2004. TXR No. 0052727. Cancer
Assessment Review Committee, Health Effects Division, Office of Pesticide Programs, Washington, DC. Available
online at http://www.epa.gov/opp00001/foia/reviews/121601/121601-2004-08-3 la.pdf.
U.S. EPA (Environmental Protection Agency). (2005a) Supplemental guidance for assessing susceptibility from
early-life exposure to carcinogens. Risk Assessment Forum, Washington, DC; EPA/630/R-03/003F. Available
from the National Technical Information Service, Springfield, VA, and online at
http ://cfpub. epa.gov/ncea/cfm/recordisplay. cfm?deid= 160003.
U.S. EPA (Environmental Protection Agency). (2005b) Tolerance reassessment progress and risk management
Decision (TRED) for procymidone. Office of Prevention, Pesticides and Toxic Substances, Washington, DC.
Available online at http://www.epa.gov/oppsrrdl/REDs/procymidone_tred.pdf.
U.S. EPA (Environmental Protection Agency). (2006a) Toxicological review of dibutyl phthalate (di-n-butyl
phthalate). [External review draft]. Integrated Risk Information System, Washington, DC; NCEA-S-1755.
Available online at http://cfpub.epa.gov/ncea/cfm/recordisplay.cfm?deid=155707.
U.S. EPA (Environmental Protection Agency). (2006b) Interim guidance for microarray-based assays: data
submission, quality, analysis, management and training considerations. [External review draft]. Genomics Task
Force Workgroup, Science Policy Council, Washington, DC. Available online at
http://www.epa.gov/osa/spc/genomicsguidance.htm.
U.S. EPA (Environmental Protection Agency). (2006c) Summary of the NCEA colloquium on current use and
future needs of genomics in ecological and human health risk assessment (final report). National Center for
Environmental Assessment, Washington, DC; EPA/600/R-04/039. Available from the National Technical
Information Service, Springfield, VA, and online at http://cfpub.epa.gov/ncea/cfm/recordisplay.cfm?deid=149984.
U.S. EPA (Environmental Protection Agency). (2006d) Revised reregistration eligibility decision document (RED)
for MSMA; DSMA, and CAMA and cacodylic acid. Office of Prevention, Pesticides and Toxic Substances,
Washington, DC; EPA/738/R-06/021. Available online at from the National Technical Information Service,
Springfield, VA, and online at http://www.epa.gov/oppsrrdl/REDs/organic_arsenicals_red.pdf.
R-9
-------
U.S. EPA (Environmental Protection Agency). (2006e) Approaches for the application of physiologically based
pharmacokinetic (PBPK) models and supporting data in risk assessment. National Center for Environmental
Assessment, Washington, DC; EPA/600/R-05/043F. Available from the National Technical Information Service,
Springfield, VA, and online at http://cfpub.epa.gov/ncea/cfm/recordisplay.cfm?deid=157668.
U.S. FDA (Food and Drug Administration). (2003) Guidance for industry. Pharmacogenomic data submissions.
Draft guidance. Washington, DC. Available online at www.fda.gov/cder/guidance/5900dft.pdf [accessed 26 July
2004].
Vondracek, M; Xi, Z; Larsson, P; et al. (2001) Cytochrome P450 expression and related metabolism in human
buccal mucosa. Carcinogenesis 22(3):481-488.
Vondracek, M; Weaver, DA; Sarang, Z; et al. (2002) Transcript profiling of enzymes involved in detoxification of
xenobiotics and reactive oxygen in human normal and simian virus 40 T antigen-immortalized oral keratinocytes.
Int J Cancer 99(6):776-782.
Waters, MD; Fostel, JM. (2004) Toxicogenomics and systems toxicology: aims and prospects. Nat Rev Genet
5(12):936-948. Available online at
http://www.cmbi.ru.nl/edv^ioinf4/articles/pdf/tox_ReviewToxicogenomicsNaturedec2004.pdf.
Weng, L; Dai, H; Zhan, Y; et al. (2006) Rosetta error model for gene expression analysis. Bioinformatics
Wilson, VS; Lambright, C; Furr, J; et al. (2004) Phthalate ester-induced gubernacular lesions are associated with
reduced ins!3 gene expression in the fetal rat testis. Toxicol Lett 146(3):207-215.
Wine, RN; Li, L-H; Barnes, LH; et al. (1997) Reproductive toxicity of di-w-butylphthalate in a continuous breeding
protocol in Sprague-Dawley rats. Environ Health Perspect 105: 102-107.
Wu, W; Hu, W; Kavanagh, JJ. (2002) Proteomics in cancer research. Int J Gynecol Cancer 12(5):409-423.
Wyde, ME; Kirwan, SE; Zhang, F; et al. (2005) Di-n-butyl phthalate activates constitutive androstane receptor and
pregnane X receptor and enhances the expression of steroid-metabolizing enzymes in the liver of rat fetuses.
Toxicol Sci 86(2):281-290.
Yu, X; Griffith, WC; Hanspers, K; et al. (2006) A system-based approach to interpret dose- and time-dependent
microarray data: quantitative integration of gene ontology analysis for risk assessment. Toxicol Sci 92(2): 560 -577.
Zhang, Y; Jiang, X; Chen, B. (2004) Reproductive and developmental toxicity inFl Sprague-Dawley male rats
exposed to di-n-butyl phthalate in utero and during lactation and determination of its NOAEL. Reprod Toxicol
18(5):669-676.
Zhou, T; Chou, J; Watkins, PB; etal. (2009) Toxicogenomics: transcription profiling for toxicology assessment.
Mol Cli Envi Tox 1:325-366
Zien, A; Kuffner, R; Zimmer, R; et al. (2000) Analysis of gene expression data with pathway scores. Proc Int Conf
Intell SystMol Biol 8:407-417.
URLS AND WEBSITES:
http://cfpub.epa.gov/ncea/cfm/recordisplay.cfm?deid=149984
http://www.niehs.nih.gov/research/supported/centers/trc/
http://cerhr.niehs.nih.gov/news/phthalates/report.html
R-10
-------
http ://david. abcc .ncifcrf.gov/listj sp
http://intranet.epa.gov/ncea/pdfs/qmp/ncea_qmp.pdf
http://searchlauncher.bcm. tmc.edu/help/BLASToutput. html#anchorl4684156
http://www.ehponline.org/txg/docs/admin/txg-n-press.html?section=toxicogenomics
http://www.genome.jp/kegg
http://www.hesiglobal.org/Co mmittees/TechnicalCommittees/Genoniics/EBI+Toxicogenoniics.htm
http://www.ncbi.nlm.nih.gov/sites/entrez?db=gene
www.epa.gov/iris/whatsnewarchive.htm
http://www.epa.gov/ncea/iris/subst/0170.htm
http://www.epa.gov/ncea/iris/subst/0378.htm
http://www.epa.gov/ncea/iris/subst/0126.htm
www.genego.com
www.omics.org
http://www.fda.gov/ScienceReserch/BioinformaticsTools/MicroarrayQualityControlProject/default.htm
R-ll
-------
United States
Environmental Protection
Agency
National Center for Environmental Assessment
Office of Research and Development
Washington, DC 20460
Official Business
Penalty for Private Use
$300
Printed with vegetable-based ink on paper that
contains a minimum of post-consumer fiber
PRESORTED STANDARD
POSTAGE & FEES PAID
PERMIT NO. G-35
------- |