EPA/600/R-04/039F
                                                 March 2006
  Summary of the NCEA Colloquium on Current Use and
Future Needs of Genomics in Ecological and Human Health
                       Risk Assessment
                            Prepared by

                          Rebecca Klaper
                       AAAS Fellow 2002-2004
                          and Susan Euling
                            NCEA, EPA
                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.
                                       NOTICE

       This document is a general record of discussions during the workshop. The document
captures the main points and highlights of the discussions and may include brief summaries of
work group sessions.  It is not a complete record of all details discussed, nor does it interpret or
elaborate upon matters that were incomplete or unclear. Statements represent the individual
views of the workshop participants; except as specifically noted, none of the statements represent
analyses by or positions of the EPA.
Preferred Citation:
U.S. Environmental Protection Agency (EPA). (2006) Summary of the NCEA colloquium on
current use and future needs of genomics in ecological and human health risk assessment.
National Center for Environmental Assessment, Washington, DC; EPA/600/R-04/039F.
Available online at http://www.epa.gov/ncea.

-------
                               CONTENTS


1. EXECUTIVE SUMMARY	1

2. INTRODUCTION	2
2.1. COLLOQUIUM PURPOSE	2
2.2. INITIAL INVESTIGATION OF KNOWLEDGE AND USE OF GENOMICS IN
            RISK ASSESSMENT	3
2.3. COLLOQUIUM PARTICIPANTS	4
2.4. COLLOQUIUM FORMAT AND SCOPE	4

3. BREAKOUT GROUP DISCUSSIONS: RESPONSES TO CHARGE QUESTIONS	5
3.1. ECOLOGICAL RISK ASSESSMENT BREAKOUT GROUP	5
3.2. HUMAN HEALTH RISK ASSESSMENT BREAKOUT GROUP	7
3.3. RISK ASSESSMENTS OF SUSCEPTIBLE/SENSITIVE POPULATIONS
            (WILDLIFE AND HUMANS) BREAKOUT GROUP	10
3.4. HAZARD IDENTIFICATION: SCREENING AND PRIORITIZATION BREAKOUT
            GROUP	12

4. COLLOQUIUM DISCUSSION CONCLUSIONS	16

REFERENCES	19

APPENDIX A: INFORMATION GATHERED PRIOR TO THE COLLOQUIUM
            ABOUT THE USE AND KNOWLEDGE OF GENOMICS ACROSS EPA
            OFFICES	20
APPENDIX B: COLLOQUIM PARTICIPANTS	24
APPENDIX C: COLLOQUIUM AGENDA	26
APPENDIX D: ABSTRACTS FOR SOME OF THE PRESENTATIONS	27
                                in

-------
                             1. EXECUTIVE SUMMARY

       Many federal agencies recognize the great potential of genomics technologies to change
the way in which human health and environmental exposure and effects are measured.  It is
anticipated that the use of genomics technologies may improve risk assessment by providing
more sensitive measures of toxic agent-induced changes in the physiology of various organisms,
including humans. It is also anticipated that data from genomic studies may identify genes that
cause susceptibility and biomarkers of exposure and/or effect and provide information to
extrapolate across species. Although some immediate applications of genomics have been
defined (e.g., as biomarkers of disease), more work is needed to determine how this information
will be used  in risk assessment.
       In 2002, the U.S. Environmental Protection Agency's (EPA's, or the Agency's) Science
Policy Council (SPC) developed the Interim Policy on Genomics, allowing genomics to be used
on a case-by-case basis in a weight-of-evidence approach for risk assessments (U.S. EPA, 2002).
Genomics technologies currently need further refinement (e.g., reduced experimental
variability), development, and validation before data from these experiments can be used in risk
assessment.  Even when these issues have been addressed, it is unclear how genomics data will
affect individual risk assessments and whether the current risk assessment process will
accommodate the integration of genomics data.
       Because one role of the National Center of Environmental Assessment (NCEA) is to
develop and  improve EPA risk assessment methods, NCEA held a colloquium to provide a
forum to assess current thinking about the use of genomics in risk assessment (e.g., how this
technology will be applicable to risk assessors) and to determine the needs of the EPA program
offices and regions that NCEA serves. NCEA defined two goals for the colloquium: (1) to
identify how genomics data may improve  risk assessment, and (2) to identify the current and
future needs  (e.g., tools, data, case studies) of the EPA program offices and regions in the area of
genomics and risk assessment.
       Colloquium participants consisted of scientists, risk assessors, and managers from various
EPA offices  and laboratories, including NCEA; the Office of Water; the Office of Prevention,
Pesticides and Toxic Substances; EPA regional offices 2 and 4; the National Health and
Environmental Effects Laboratory; the National Exposure Research Laboratory; the Office of the
Administrator; and the Office of the Chief Financial Officer.
                                          1

-------
       In preparing for and developing the colloquium, information was gathered from several
EPA offices to gauge the understanding of genomics technologies and how these types of data
could be used-currently and in the future-in risk assessments. The responses to questions
indicated that many offices recognized the future importance of genomics in risk assessment in
general.
       The colloquium was designed to initiate discussions among risk assessors, managers, and
scientists developing and using genomic technologies. Participants recognized the need for
future, extended interaction between laboratory scientists and risk assessors, so that experiments
are designed to produce data in a form that will be useful for risk assessment purposes.
       An overall conclusion  of the colloquium discussions was that genomics data will most
likely play a role in several aspects of the risk assessment process, including hazard
identification, defining mode(s)  and mechanism(s) of toxicity, identification of genetic
susceptibilities, and prioritization for screening and testing of environmental agents. However,
participants  considered it unlikely that gene expression data will be used as the sole indicator of
an adverse effect. Rather,  such data will be used in conjunction with in vivo endpoints. For
"omics" (defined as genomics, proteomics, and metabonomics) data to be useful in risk
assessment,  several issues will need to be addressed, including (1) validation of methodology,
including data analysis; (2) development of interpretation tools for risk assessors; (3)
development of criteria for cross-species extrapolation from model organisms to humans;  (4)
linkage of traditional in vivo endpoints to genomics  data; (5) development of a method for
communicating this information both within and outside the Agency; and (6) development of
criteria for the inclusion of genomics data in risk assessment.

                                  2.  INTRODUCTION

2.1.  COLLOQUIUM PURPOSE
       A colloquium entitled "Current Use and Future Needs of Genomics in Ecological and
Human Risk Assessment" was held on May 8, 2003, in Alexandria, VA, to discuss how data
from current genomics technologies and their future refinements (e.g., reduced experimental
variability) could be used in risk assessment.  The overall goal of the colloquium was to provide
Agency scientists, researchers who are using genomics, and risk assessors an opportunity to

-------
share perspectives, to discuss how genomics may improve risk assessment, and to identify
current needs.
       Because the colloquium represented a scoping step in identifying future needs in the area
of genomics data in risk assessment, this summary report of the colloquium was not peer-
reviewed.

2.2.  INITIAL INVESTIGATION OF KNOWLEDGE AND USE OF GENOMICS IN
RISK ASSESSMENT
       In preparing for and developing this colloquium, Rebecca Klaper, an American
Association for the Advancement of Science (AAAS) fellow at NCEA, requested responses to
seven questions sent via e-mail to several EPA program and regional office contacts.  The
questions were designed to gauge an office staffs level of understanding of genomics and to
determine whether they had received or were currently using genomics data and whether they
had discussed how genomics data might be used in their health or risk assessments.
       Overall, the responses to the questions indicated that many offices recognize the future
importance of genomics for their needs and for risk assessment in general. However, only a few
staff members in each office had a reasonable comprehension of the technologies and types of
data that will be generated and the current limitations of the technology. Those who responded
indicated that they believe genomics will eventually contribute to risk assessment through a
better understanding of the mechanisms and/or modes of chemical toxicity, the shape of dose-
response curves for many pollutants, the basis for extrapolations from model organisms to
species of interest, identification of susceptible populations, and estimates of uncertainty  factors.
At the time of the colloquium, none of the program offices had received microarray data to
support a risk assessment or decision; however, genomics data, in the form of single gene
expression changes, as well as protein and genetic biomarker data, had been submitted in the
past. All respondents expressed an interest in having genomics training provided to their office.
The questions and the peer consultant's answers and comments are included in Appendix A.
       The EPA Science Policy Council's definition of genomics was used at the colloquium
and is used in this document: Genomics is the study of all the genes of a cell, or tissue, at the
DNA (genotype), mRNA (transcriptome), or protein (proteome) level (U.S. EPA, 2002).  By
extension, toxicogenomics is defined as the study of gene expression (mRNA and/or protein
products) changes after exposure to a toxic agent.
                                          3

-------
2.3.  COLLOQUIUM PARTICIPANTS
       The colloquium was an internal EPA meeting designed to bring together risk assessors,
scientists, and managers at the EPA program offices and regions and laboratory scientists within
EPA's Office of Research and Development who are currently developing or using genomics
technologies. Forty-two participants attended the colloquium, including scientists and risk
assessors from many different EPA offices. Their areas of expertise spanned a broad spectrum
of knowledge and agency understanding.  The colloquium participants are listed in Appendix B.

2.4.  COLLOQUIUM FORMAT AND SCOPE
       Presentations in the morning session included discussions on how genomics data might
be used in risk assessment and EPA's policy on their use (see Appendices C and D). Each
presentation was followed by a question-and-answer period.
       For the afternoon session, participants were divided into four breakout groups to discuss
the implications of genomics technologies in four specific risk assessment areas: (1) ecological
risk assessment, (2) human health risk assessment, (3) identification of sensitive or susceptible
subpopulations, and (4) screening and prioritization of chemicals and microbes. The breakout
group discussions focused on charge questions designed by the conference organizers and
session co-chairs. Summaries of each breakout group responses to the charge questions are
presented in Chapter 3.
       Topics  discussed included the potential use of genomics data in risk assessments and how
to make these data useful from the risk assessor's perspective. The following questions were
addressed:  (1) If we assume that issues such as standardization of genomics techniques have
been solved, then how will this information be used in risk assessment in the future? (2) What
genomics data set format will be most useful to risk assessors? (3) How can experiments be
designed that will provide the most useful information for risk assessment?

-------
  3. BREAKOUT GROUP DISCUSSIONS: RESPONSES TO CHARGE QUESTIONS

3.1.  ECOLOGICAL RISK ASSESSMENT BREAKOUT GROUP
Moderators: Bob Frederick, Sig Degitz
Participants: Rebecca Klaper, Tala Henry, Greg Toth, Ann Miracle, Thomas Baugh, Michael
Brody, Greg Susanke

Question 1.  Given that it is critical to establish a link between gene expression data and an
endpoint of concern, what specific information or links would be needed to use
toxicogenomics data in risk assessment?

Response:
       Ecological risk assessment focuses on the population rather than the individual,  so a
primary goal for ecological risk assessment will be to extend genomics data on exposure and
effects from the individual level to the population level. Ecological risk assessment, by its
nature, involves determining exposure and effects for many different organisms in an
environment rather than for just one, as in human health risk assessment. Because it is
impossible to determine the exposure and effects for every species in an ecosystem, species
extrapolation is a key issue. Therefore, it is necessary to determine the genomic (i.e., the global
gene expression profile) homologies and similarities in biochemical mechanisms and metabolism
among species to be able to extend genomic technologies developed for one species to another.
For a given chemical, it would be helpful to understand the degree of cross-species conservation
among genes whose expression pattern has changed after chemical exposure.
       How can genomics be used in assessing population effects in general? Genetic diversity
is critical to  sustaining populations. Thus, genomics will likely provide insight into the role that
genetic diversity plays in sustainability.  Currently, it is difficult to gain information for a
wildlife species using genomics tools unless the species of interest is genetically well-defined,
which is rare. Researchers are defining stressor-response relationships for markers known to be
relevant to population viability (e.g., survival, development, fitness). When these relationships
have been defined, then links between genomics changes (e.g., gene expression patterns of
response) and the response can be assessed.  For species within an ecosystem, genomics data

-------
from a well-defined species may be used to predict physiologic responses in a less well-defined
but related species.  Eventually, genomics will provide a sensitive means to directly compare
diverse species within an ecosystem.
       Patterns of gene expression will likely provide new and more specific indicators of
exposure or effects. It is necessary to define what genomic changes (e.g., in what genes and at
what level of change)  are relevant to adverse outcomes. The amount and type of information
needed to link gene expression to an endpoint of interest will need to be determined for each
scenario.
       Case studies demonstrating the linkage (i.e., proof of concept) between gene expression
and adverse outcomes would be valuable. For example, it would be useful to begin by linking a
well-defined stressor to a genomic response to a known adverse outcome. As one example, there
are data linking estrogen exposure, vitellogenin gene expression, and male feminization effects.
In addition, it will be important to link genomic changes to toxicity pathways and use this
information to inform  mechanism or mode of action (MOA). Genomic endpoint information
will not be used in isolation but may be used to inform other, higher-level effects.

Question 2.  What are the current limitations of the technology for use in ecological risk
assessment?  Will these be overcome in the near term (less than 5 years) or in the long term (5
or more years)?

Response:
       Currently, information connecting genomics data and effects data is lacking at the
individual and the population levels. In order for genomics to  become a viable tool for
ecological risk assessment, genomics data must be developed for species that represent
organisms of ecological interest (for chemical testing  and field work), and data will need to be
extrapolated to population- and community-level effects.
       For ecological  risk assessment, there is a need for genomic information for ecologically
relevant species and resources to support this need. At present, there are genomics data for a few
species relevant to ecological risk assessment, including the fathead minnow (Pimephales
promelas), the African clawed frog (Xenopus laevis\  daphnia  (Daphniapulex\ zebrafish (Danio
rerio), and Japanese medaka (Oryzias latipes). However, genomics information is lacking for

-------
other ecologically relevant species used in chemical testing by many of the program offices,
including the Office of Water and the Office of Prevention, Pesticides and Toxic Substances.
       Technical issues for genomics technologies include an inadequate reproducibility within
and across laboratories, expression level variability, and the ability of genomics data to be
quantitative.  For this technology to be useful in the near future, further validation of the
techniques is needed. Reproducibility of data will improve as the understanding of the
techniques and experimental  variables improves, and this is currently being addressed by several
studies.
       Another limitation is the lack of interaction between scientists using genomics and risk
assessment scientists, contributing to roadblocks in use and acceptance of the technology in risk
assessment.  In the short term, the Agency is unlikely to use genomics data for quantitative
aspects of risk assessment. However, the data may be used to inform qualitative aspects of risk
assessment, including mode or mechanism of action or exposure.

3.2.  HUMAN HEALTH RISK ASSESSMENT BREAKOUT GROUP
Moderators: Vicki Dellarco, Ines Pagan
Participants: Linda Birnbaum, David Bussard, Chao Chen, David Dix, Karen Hamernik, Oscar
Hernandez, Robert McGaughy, Julian Preston, Vickie Wilson

Question 1.  Where in the overall risk assessment process (e.g., hazard identification, dose-
response, exposure assessment, risk characterization) do you think omics data are more likely
to play an important role? How can the data from omics be potentially used in risk
assessment?  Consider the mechanism oftoxicity as well as treatment conditions (e.g., route,
duration, magnitude of exposure) that are important for expression of the toxic effect.

Response:
       Several areas were mentioned, including the following:
       •  Identifying hazards.
       •  Defining the type oftoxicity of various chemicals (e.g., genotoxic versus hepatotoxic
          chemicals).
       •  Acquiring MOA and mechanism of action information of toxicants.

-------
       •  Using gene expression patterns in the future to prioritize chemicals for screening and
          testing.

Question 2.  To use omics data in human health risk assessment, what issues need to be
addressed (e.g., handling the breadth and scope of data, interpreting biological and statistical
meaning, training Agency risk assessors) ?  What criteria should be considered for those data
to be useful to risk assessors?

Response:
       It is unlikely that gene expression will be used as the sole indicator of an adverse effect,
but it will be used in conjunction with other endpoints. To use omics data in human health risk
assessment, issues that need to be addressed include the following:
       •  Validation of methods, including validation of data analysis methods such as
          Minimum Information about a Microarray Experiment (MIAME) (see
          www.mged.org/Workgroups/MIAME/miame.html for MIAME=compliant data/study
          methods).
       •  Development of interpretation tools, including computer software, statistics, and
          bioinformatics tools for risk assessors.
       •  Development of criteria for use of omics data (e.g., criteria for extrapolation of
          information for model organisms to humans).
       •  Determination of whether there is a link between histopathology data and traditional
          endpoints of toxicity to omics data.
       •  Identification of sentinel genes (i.e., biomarkers of effect) that are good predictors of
          toxic response.  Case studies to serve as examples of the use of omics in risk
          assessment and development of "lessons learned" across agencies and within EPA
          (across offices). Recommendations for case studies included
              1.  Use a simple case study with few confounding factors as a proof of principle.
              2.  Use existing research from the Office of Research and Development to
                 develop "lessons learned" and research needs; then develop a second phase of
                 research.
              3.  In the end, develop a quality assurance filter (i.e.,  a practice set of
                 guidelines/standards/criteria and guidance for reviewers to interpret data).
       •  Statistical approaches to analyze genomics data for use within risk assessment.
       •  Overall guidance for methods to use this type of information in risk assessments.

-------
       •  Risk communication issues: How can genomics data be translated into information
          that is easy to understand for the general public and applicable for a regional
          assessment?

       •  EPA involvement in partnerships to address the preceding issues and to develop a set
          of guidelines that will actively support the development of both criteria (see below)
          and tools to enable EPA risk assessors to display and analyze omics data.


Criteria that need to be developed include

       •  A framework for use of genomics data in risk assessment that is similar in scope to
          other EPA frameworks, including the framework for evaluating a hypothesized
          carcinogenic MO A within the guidelines for carcinogen risk assessment (U.S. EPA,
          2005); the framework for human health risk assessment (U.S. EPA, 1998); the
          framework for application of the toxicity equivalence methodology for
          polychlorinated dioxins, furans, and biphenyls in ecological risk assessment (U.S.
          EPA, 2003a); and the framework for cumulative risk assessment (U.S. EPA, 2003b).
          As for any new technologies, test validation needs to be done with genomics,
          including several levels of quality assurance (QA) and quality control (QC).

       •  For dose-response data, the level of exposure that will induce a genomics response
          and determination of what that response indicates about the effects of that chemical
          on the organism are needed. Genomics data currently provide only a "snapshot" in
          time; thus, criteria will need to be developed as to how many snapshots, and at what
          time periods and intervals, are needed to provide the data necessary to link an
          exposure to an effect.

       •  Other questions that were raised included

              1.  Can biomarkers be found using omics?

             2.  What is the state of metabonomics  technologies?

             3.  How will metabonomics data be used?

             4.  Can omics be an appropriate tool for the identification of surrogate tissues to

                 test toxicity endpoints?  If so, what types of cells are needed to identify a

                 specific response?

       •  To use these types  of data, genomics data need to be linked to an MOA that is
          relevant to humans in the proper time course and duration, and the data need to be
          accurate in extrapolations from high to low doses.  In addition, research efforts need
          to establish correlations between omic response and adverse effect.  It will be  critical
          to determine the normal, or unperturbed, biological variability to establish the gene
          expression patterns for the normal versus the disease or toxic response state (i.e.,
          validation).

-------
3.3.  RISK ASSESSMENTS OF SUSCEPTIBLE/SENSITIVE POPULATIONS
(WILDLIFE AND HUMANS) BREAKOUT GROUP
Moderators: Margaret Chu, Les Touart
Participants: Ross Highsmith, Elizabeth Mendez, Marian Olsen, Brenda Percovich Foos, Chris
Saint, Bob Sonawane, Ravi Subramaniam, Larry Valcovic, Vanessa Vu

       The group recognized that the definitions for susceptibility, sensitivity, and omics
technologies could affect the responses to the questions addressed.  In general, the group thinks
that omics data have great potential for identifying susceptible and/or sensitive individuals and/or
species. Currently, omics data are being applied in clinical medicine. For example, broadly
defined genomic technologies are used clinically in determining susceptibility in complex
diseases such as cancer. As omics technologies develop, they may be used in determining
susceptibility to,  and reducing uncertainty in, assessing environmental and health risk
assessment.
       This summary should be viewed as the breakout group's discussion of how current omics
has or can be applied to ecological and human health risk assessments only, not to other areas of
biomedical applications.

Question 1. How have omics been applied to identify sensitive/susceptible subpopulations
and/or species?

Response:
       Overall, very few examples have been identified where genomics data have been used to
identify susceptible populations. Genomic techniques are currently used to define
polymorphisms in humans (mostly through animal models for humans, such as those for mice
and rats) by looking for genetic variation associated with diseases (e.g., genes associated with
diabetes and cancer).  Identification of these polymorphisms is currently aiding in the
development of drugs to counteract the effects of genetic susceptibility or in the development of
alternative therapies for susceptible genotypes.  Genomics is being used to improve the
effectiveness  and specificity of pharmaceuticals by directing the action to a molecular target
known to play a role in susceptibility.  However, to use this information in risk assessment, the
                                          10

-------
risks and levels of susceptibility will need to be quantitative, whereas currently they are merely a
qualitative identification of genes.
       Genomic technologies at present are not reproducibly quantitative.  In applying these new
tools, it is important to understand which genes are being affected and by what exposures. These
tools offer promise in screening for patterns of gene expression. Understanding the underlying
disease mechanisms more fully will allow investigation and identification of relevant gene
patterns. To elucidate which genes cause susceptibility, one can look at patterns to see whether
connections appear. Different genes are turned on or off at different life stages, and this
complicates interpretation. Many genes activated in the cancer process, for example, are very
active in early development but not in mature individuals.  For risk assessment, it is expected that
the critical pathways can be identified by comparing treated to untreated states.

Question 2.  What types or combinations ofomics technologies are most likely to have the
greatest impact on developing biomarkers of susceptibility?

Response:
       To date, the science is not at the point where a recommendation can be made as to the
type of technology that will have the greatest impact on development of biomarkers of
susceptibility. However,  compared with conventional biomarkers, omic-based measures may be
more  advantageous.  In the best case, they could provide sequential connections between the
different levels of physiology, from the gene, to gene transcription, to protein formation,  and to
physiological function. Genomics may provide a link between a specific biomarker and the
cause of a particular susceptibility, and subsequently the biomarker may be an mRNA, protein,
or metabolite. The type of genomics techniques that will be  the most useful will be determined
by which marker provides the appropriate information.
       A critical question that arises for the application ofomics as biomarkers is how to
effectively deal with the inherent increased sensitivity of the technology. Changes in gene
expression may not necessarily translate into an adverse consequence, but understanding the
linkage is important. Perhaps the greatest benefit is in identifying, under certain conditions,
which populations or subpopulations are the most susceptible.
                                           11

-------
Question 3. How can ontics help reduce uncertainty?

Response:
       In the future, genomics will be important for interspecies extrapolations, helping to
determine the physiological relationship among species and, therefore, the extent of uncertainty
when extrapolating results of toxicological tests from model organisms to organisms of interest.
Genomics will likely be more useful for identifying patterns of response (i.e., qualitatively) than
for defining a dose-response curve. This technology can be used to identify critical gene(s)
linked to an effect. Genomics may also be useful for  determining the factors associated with
low-dose responses to compounds, providing evidence to support or reject hypotheses
surrounding hormetic effects. Genomics is a powerful tool, but to be useful it has to be applied
to a reasonable and tractable question.

3.4. HAZARD IDENTIFICATION:  SCREENING AND PRIORITIZATION
BREAKOUT GROUP

Moderators: Susan Euling, Jennifer Seed
Participants: Nancy McCarroll, Cynthia Nolt-Helms, Robin Oshiro, Devon Payne-Sturges, Phil
Sayre, Rita Schoeny, Deborah Segal

Question la. What types of genomics  technologies might be useful for chemical screening
purposes in the future (e.g.,for the Endocrine Disruptor Screening Program [EDSP],
pesticide inerts, Toxic Substances Control Act's High Production Volume [HPV] chemicals,
and Office of Water's Candidate Contaminant List [CCL])?

Response:
       A large number of chemicals need to be prioritized for testing in a number of screening
programs, such as the EDSP (http://www.epa.gov/scipolv/oscpendo/edspoverview/index.htm).
Prioritization is a separate, complex exercise and consists of a number of different criteria.  For
example, prioritization could be based on a chemical's MOA or quantitative structure-activity
relationship (QSAR).  Genomics technologies could help to determine the MOA of a chemical
and therefore assist in prioritization.  Genomics technologies could also be used to gain
information about structure-activity relationships and could then be incorporated into the dataset
                                           12

-------
for QSAR modeling. If a gene expression profile has been linked to an adverse effect after
chemical exposure, then chemicals could be screened by their gene expression profile.
Chemicals without this link to an adverse effect could be assigned a lower toxicity testing
priority.  In addition, genomics could be used to develop a gene expression fingerprint for the
response of an organism to chemical mixtures.
       Genomics technologies need to be validated before routine use in a screening program.
Specifically, microarray or proteomic data need to be linked to an adverse effect (in the case of
human health risk assessment) or endpoint of concern (in the case of ecological risk assessment).
Verifying this link is complicated by a number of factors,  including high inter-experiment and
intra-experiment variability of response, that have made it difficult to replicate genomics
experiments in some cases. The group members thought that microarray analysis is probably the
furthest along for liver-mediated toxicity and  estrogen receptor-mediated toxicity.
       Microarrays  may not be the best tool for chemical  screening purposes if proven enzyme
screening methods already exist (e.g., cholinesterase activity). Eventually, this technology may
lead to more rapid screens and to a decrease in the use of animals in testing, but initially, as the
techniques are being developed and validated, an increase in cost and animal usage may occur.
In the future, genomics will most likely provide a more  sensitive and specific means with which
to measure effects from chemicals than current methods. Proteomics, in particular, may be the
most useful technology for chemical screening because  proteins are, in most known cases, the
actual functional component within the cell and organism (e.g., small RNAs have been found to
be the functional molecule for some genes [Lee et al., 1993]).  Therefore, measuring the global
proteomic response  may be the optimum indicator of the physiological response of the organism
after chemical exposure.

Question Ib. How do these genomics technologies compare with other screening technologies
currently used?  Identify the strengths and weaknesses.

Response:
       Current screening technologies include the following:
       •  QSARs,  which are currently limited to use in screening for ecotoxicology and
          mutagenic carcinogens.
       •  Various in vitro tests (e.g., receptor binding assays, Ames test).
                                           13

-------
       •  Short-term in vivo assays that focus on a specific response (e.g., uterotrophic).
       •  Short-term in vivo assays that have a broad focus (e.g., The Organisation for
          Economic Co-operation and Development [OECD] combined repeat/reproduction,
          acute toxicity study).

       The group discussed the question, "What are the requirements for a screening assay?"
The group recommended that the assay must be:
       1.  Sensitive (erring on the side of false positives so that problem chemicals are
          identified).
       2.  Somewhat specific to the endpoint of concern.
       3.  Fast and efficient (relative measure).
       4.  Inexpensive (relative measure).

       Currently, genomics technologies are not as rapid, inexpensive, or efficient as the
currently used screening technologies. In some cases, they may be more sensitive and  specific
than other assays, but some current screening assays are quite selective and specific to  the MOA
of interest and rapid (e.g., cholinesterase activity for organophosphate pesticides). Therefore, the
state of the technology is not ready for screening or prioritization because the omics technologies
are  currently not as rapid or inexpensive as the other currently available methods.

Question 2.  What types of genomics technologies/experimental design could be useful for
microbial screening for drinking water quality?

Response:
       Current microbial screening techniques include culturing microorganisms from water
samples and using polymerase chain reaction (PCR) techniques to examine a water sample for a
genetic component of a bacterium of interest (e.g., 16sRNA). Culturing is the gold standard for
identification and quantification of bacteria, but the technique does have limitations.  It may be
slow and the cultures may contain both virulent and nonvirulent forms of the bacteria of interest.
Additionally, under certain  conditions, viable bacteria may be present but noncultureable. For
most viruses, there are few or no culture techniques available.
       PCR techniques rely on known differences in segments of genetic material within each
strain to identify the presence of a species of interest.  Microarray technologies, like PCR

                                           14

-------
techniques, rely on identifying genetic components of known pathogens or the virulence factors
within those pathogens. Microarray technology has an added benefit in that it will also screen
for emerging pathogens that happen to have the same virulence factor.  However, both of these
techniques are limited because they may not identify a specific pathogen due to genetic
similarities between strains, leading to false positive results and small mutations that do not
change the virulence but do change the rate of detection. In addition, Genbank
(http://ncbi.nlm.nih.gov/Genbank/index.html), the repository for sequence information on many
species, including bacterial pathogens, is subject to scientific and clerical errors.  Sequenced
samples may be impure or slightly inaccurate, which could lead to a greater variability when
developing a diagnostic test.  Finally, water samples need to be large enough to detect the
bacteria, and samples may need to be amplified using PCR before detection methods can be
effective.
       Proteomics may be useful in the future but may not be  as sensitive as genomics for water
sampling because the production of proteins may not have occurred by the time of sampling.
EPA's regulations at 40 CFR 136.4, 136.5, and 40 CFR 141.27 allow one to apply for permission
by EPA to use an alternate test procedure instead of an EPA-approved reference method. The
Alternative Test Procedure Program (http://www.epa.gov/waterscience/methods) in the Office of
Water has received inquiries regarding submissions for the use of genetic techniques for
detecting microbes in recreational waters, but these data cannot be accepted yet because there are
no approved methods for genetic-based tests.  Such methods cannot be accepted for review until
methods are published in 40 CFR Part 136 (ambient water, wastewater, or biosolids methods) or
40 CFR Part 141 (drinking water methods) as approved methods.

Question 3.  What are the current limitations of the technology for use in screening assays?

Response:
       The following limitations to the use of omics in chemical screening were noted:
       •  Linkage to adverse effect of concern is needed.  There are not many cases where this
          has been established.
       •  Sensitivity of the technologies (erring on the side of false positives so problem
          chemicals do not slip through) is not well established.
       •  Specificity/selectivity is not established.

                                          15

-------
       •  Speed/efficiency.
       •  Cost.
       •  Animal usage compared with some in vitro screens.

       The following limitations to the use of omics in microbial screening were noted:
       •  Sensitivity is not well established. Microarray analysis may not identify a specific
          pathogen owing to similarities in strains leading to false positive results and small
          mutations that do not change the virulence but do change the rate of detection.
       •  Specificity/selectivity is not established.
       •  Cost.
       •  Currently a need for a large water sample using PCR.
       •  Procedures are less well established for viruses.
       •  Microarrays may be preferable to proteomics.

                   4.  COLLOQUIUM DISCUSSION CONCLUSIONS

       In the future, genomics data could play a significant role in several aspects of the risk
assessment process, including hazard identification, definition of mode(s) and mechanism(s) of
toxicity, identification of genetic susceptibilities, and prioritization for screening and testing of
environmental chemicals. However, it is unlikely that gene expression data will be used as the
sole indicator of an adverse effect; rather, such data will be used in conjunction with in vivo
endpoints. To use omics data in risk assessment, several issues will need to be addressed,
including
       1.  Validation of methodology and data analysis, including several levels of QA/QC
          (similar to validation of other new technologies).
       2.  Development of interpretation tools for risk assessors.
       3.  Development of criteria for cross-species extrapolation from model organisms to
          humans.
       4.  Linkage of traditionally used in vivo endpoints to genomics data.
       5.  Development of a method to communicate this information both within and outside
          the Agency.
       6.  Development of criteria for the inclusion  of genomics data in risk assessment (similar
          to criteria that have been developed for new technologies in the past).
                                           16

-------
       It will be important to develop case studies to demonstrate linkage (i.e., proof of concept)
between exposure, gene expression, and adverse outcomes. Studies need to start by
characterizing and supporting the links between a well-defined stressor (e.g., toxic agent) and a
well-established environmental effect.  For example, one can build  a case for the links between
estrogen exposure, vitellogenin protein expression, and male feminization effects in fish.  Then,
gene or protein expression data will need to be assessed for whether they are linked to exposure
to the stressor as well as the response or effect in the organism or population.
       Eventually, genomics will also be useful for screening and testing, but there will be
limitations and issues similar to those described above. Genomics technologies could be used to
gain information about QSARs and could then be incorporated into the dataset for QSAR
modeling. This technology may lead to more rapid screening assays and to decreased use of
animals in testing. But initially, as the techniques are being developed and validated, animal
usage may increase.
       In the future, genomics will most likely provide a more sensitive and specific means to
measure effects after chemical exposure than those offered by current methods. Proteomics in
particular may be the most useful technology for chemical screening because proteins are
typically the functional component within the cell and organism.  Currently, the state of the
technology has not been optimized for screening and/or prioritization purposes because
genomics is not as consistent in response (i.e.,  high variability), efficient, or low in cost as some
of the other currently available methods. For example, cholinesterase activity assays are rapid
and specific to the MOA of interest.
       Genomics techniques may be particularly useful for microbial screening because culture
techniques are limited.  For example, under certain conditions viable bacteria may be present but
noncultureable. Thus, gene expression products of the species of interest may  be a method to
circumvent this problem.  An added benefit to the microarray approach is that it can
simultaneously screen for emerging pathogens that happen to have  a similar virulence factor.
The Alternative Test Procedure Program has received  inquiries regarding submissions for the use
of genetic techniques for detecting microbes, but these data cannot be accepted until methods for
genetic-based tests have been approved. A current Agency focus is to begin developing these
genetic methods.
       An additional point was raised for the incorporation of genomics into ecological risk
assessment (ERA). ERA focuses on the population rather than on the individual, so one of the
                                           17

-------
primary goals will be to extend genomics data at the individual level to the population level with
regard to both detection and effects.  Thus, exposure and effects for many different species in an
environment need to be determined.  Because it is impossible to determine the exposure and
effects for every species in an ecosystem, species extrapolation is important.  Therefore, it is
necessary to determine the genomic (i.e., global gene expression profiles) homologies, degree of
conservation of genes whose expression is altered,  and similarities in biochemical mechanisms
and metabolism among species to be able to extend genomic technologies developed for one
species to another.
       This colloquium and future EPA activities will provide opportunities for risk assessors to
learn more  about this field and to exchange ideas about the use of genomics in risk assessment.
Participants recognized the need for further interactions among laboratory scientists and risk
assessors to inform the development and possible uses of this technology for risk assessment. It
is clear that risk assessors must be included in discussions of genomics within the Agency so that
data will be designed and presented in a form that will be useful for risk assessment purposes.
                                           18

-------
                                          REFERENCES

Lee, RC; Feinbaum, RL; Ambros, V. (1993) The C. elegans heterochronic gene lin-4 encodes small RNAs with
antisense complementarity to lin-14.  Cell 75(5):843-54.

U.S. EPA (Environmental Protection Agency). (1998) Summary of the U.S. EPA colloquium on a framework for
human health risk assessment (Volume 2, 1998). Risk Assessment Forum, Washington, DC.

U.S. EPA (Environmental Protection Agency). (2002) Interim policy on genomics. Prepared by the Science Policy
Council, Washington, DC. Available online at http://www.epa.gov/osa/spc/genomics.htm.

U.S. EPA (Environmental Protection Agency). (2003a) Framework for application of the toxicity equivalence
methodology for polychlorinated dioxins, furans and biphenyls in ecological risk assessment, external review draft.
Risk Assessment Forum, Washington, DC. EPA/630/P-03/002A.

U.S. EPA (Environmental Protection Agency). (2003b) Framework for cumulative risk assessment. Office of
Research and Development, National Center for Environmental Assessment, Washington, DC.  EPA/600/P-
02/00 IF.

U.S. EPA (Environmental Protection Agency). (2005) Guidelines for carcinogen risk assessment.  Federal Register
70(66)17765-17817.  Available online at http://www.epa.gov/cancerguidelines.
                                                 19

-------
APPENDIX A: INFORMATION GATHERED PRIOR TO THE COLLOQUIUM
ABOUT THE USE AND KNOWLEDGE OF GENOMICS ACROSS EPA OFFICES

      Rebecca Klaper of the National Environmental Assessment (NCEA) designed seven
questions to assess the level of understanding and knowledge of genomics technologies as they
relate to risk assessment.  Various EPA office contacts were asked in January of 2003 to
comment on each question and were encouraged to get input and have discussions about the
questions from others in their offices. In addition, contacts were asked to inform their office
about the upcoming colloquium.

      The peer consultants, listed below, responded to the questions.
NATIONAL CENTER FOR ENVIRONMENTAL
ASSESSMENT
Susan Euling
Bob Frederick
Inez Pagan
Bob Sonawane
Cindy Sonich-Mullin
Michel Stevens

OFFICE OF WATER
Joyce Donohue
Jafrul Hasan
Tala Henry
Tony Maciorowski
Edward Ohanian
Rita Schoeny

INTEGRATED RISK INFORMATION SYSTEM
Mike Broder
Lynn Flowers

OFFICE OF AIR
Tom Curran
Carl Mazza
Maria Pimentel
OFFICE OF POLLUTION.
PREVENTION AND TOXICS
Phil Sayre
Jennifer Seed

OFFICE OF PREVENTION AND
TOXIC SUBSTANCES
Gary Timm
Les Touart
Maurice Zeeman

EPA REGION 9
Bobbye Smith (RSL)

EPA REGION 4
Thomas Baugh (RSL)

OFFICE OF SOLID WASTE
AND EMERGENCY REPSONSE
Lee Hoffman
                                     20

-------
Summary of Responses to Questions about Knowledge and Experience with Genomics
from Peer Consultants


       The questions were sent to the contacts via e-mail. The summaries of all responses

received are described below.  The intention was to present the range of responses received.

Individual responses are not presented here.


Question 1. Do you think that people in your office understand the definition and types of
data associated with genomics andproteomics?

Response:

       Comments ranged from offices with some staff having a basic awareness of the
technology  of genomics and the potential use in risk assessment to offices reporting that no one
had heard of the term "genomics."  Among the offices with a basic awareness of genomics, it
was commented that very few staff had a detailed understanding of the "omics" technologies.

Question 2. Are people in your office familiar with some of the limitations in analyzing this
data?

Summary of Responses:

       Most offices stated that they were not familiar with limitations. Among the offices that

reported being familiar with limitations, the following limitations were noted:

       •      New, not validated technologies.

       •      Little data to understand the patterns of gene expression after exposure to
              chemicals (i.e., little toxicogenomics data).


Question 3. Would your office be most interested in data to enhance information on effects,
exposure, or identifying susceptible populations?


Summary of Responses:

       •      Effects were mentioned most often, and the areas of dose-response, reduction of
              uncertainty, and risk assessment were highlighted. Exposure and susceptible
              populations were mentioned equally. It was noted that all three areas, effects,
              exposure, and susceptible population identification, were of interest.

Question 4. What are your office's training needs in the area of omics?
                                          21

-------
Summary of Responses:
       The following course topics were suggested:
             Basic "Genomics 101" training course (mentioned most frequently).
             Data analysis, use in risk assessment (when this becomes available).
             Statistical analysis employed in the analysis of these data.
             Ethical implications and legal issues.
             Training for risk communicators.
       Other needs mentioned:
       •      Consider the needs of states and tribes.
       •      Discussion of when these data will come to fruition.
       •      Statistical analysis employed in the analysis of these data.
       •      Case-study: Risk assessment that supports regulatory decision making.
Question 5.  Has your office received any type of these data?

Summary of Responses:
       All queried offices stated they had not received genomics data. However, some offices
stated that they had received or used single gene expression data (the Science Policy Council's
definition of genomics).

Question 6.  What impact do you see this having on the work in your office?

Summary of Responses:
       It was noted that genomics research has the potential to improve human and ecological
risk assessments.  Areas that genomics will contribute to were noted:
       •      Mechanisms of chemical toxicity.
       •      Biological interaction of chemicals and chemical mixtures.
       •      Signal transduction pathways.
       •      Induced  gene expression and, therefore, development of biomarkers of human
             exposure.
       •      Understanding mechanisms for genetic damage and/or DNA repair, among other
             mechanisms.
       It was noted that the application of this new technology could assist EPA risk assessors
to:
       •      Characterize the shape of dose-response curve for a number of pollutants in a
              timely and cost-effective manner.
                                          22

-------
             Allow comparisons between animal mode species and humans (i.e., interspecies
             extrapolation).

             Reduce uncertainty factors through the understanding of individual susceptibility
             to environmental stressors.

             Incorporate genomics data into both human health and ecological risk assessments
             and monitoring programs. Additional comments: Exposure assays for endocrine
             disrupters have been the single most requested molecular biology tool from
             California and tribes.  Genomics assays may eventually replace some of the
             current screens used in EPA's Endocrine Screening and Testing Program.
             However, with any new technology there is considerable training and new
             infrastructure costs.

             Analyze qualitative information on mode of action and susceptible populations
             (probably from genetic polymorphisms) to support rulemaking efforts. Additional
             comments: Such data/tools will require clear definition of programmatic
             problems in order to evaluate when and where such data could be incorporated
             into risk assessments and decision-making processes.
Question 7.  Other interests/comments?


Summary of Responses:

       Two comments were received:
              The Office of Pollution Prevention and Toxics (OPPT) is currently working with
              the Office of Research and Development laboratories to look at the feasibility of
              fingerprinting for certain chemical classes, and OPPT has a Cooperative
              agreement with International Life Sciences Institute.

              The Office of Water is using genomics data within its Chemical Contaminants
              List program.
                                          23

-------
                   APPENDIX B: COLLOQUIM PARTICIPANTS
Name

Thomas Baugh
Linda Birnbaum
Michael Brody
David Bussard
Chao Chen
Margaret Chu
Sig Degitz
Vicky Dellarco
David Dix
Susan Euling
Bob Fredrick
Karen Hamernik
Jafrul Hansan
Tala Henry
Oscar Hernandez
Ross Highsmith
Rebecca Klaper
Nancy McCarroll
Robert McGaughy
Elizabeth Mendez
Ann Miracle
Cynthia Nolt-Helms
Marian Olsen
Robin Oshiro
Ines Pagan
Devon Payne-Sturges
Office/Division             Breakout
                          Group
Region 4                   ERA
NHEERL-RTP              HHRA
OCFO/OPPA               ERA
NCEA-W                  HHRA
NCEA-W                  S/SP
NCEA-W                  S/SP
NHEERL-Duluth            ERA
OPPTS                    HHRA
NHEERL-RTP              HHRA
NCEA-W                  SC
NCEA-W                  ERA
OPPTS/OPPT              HHRA
OW/OST                  ERA
OW/OST                  ERA
OPPTS/OPPT              HHRA
NERL                     S/SP
AAAS fellow at NCEA-W    FLOAT
OPPTS/OPP                SC
NCEA-W                  HHRA
OPPTS/OPP                S/SP
NERL-Cincinnati            ERA
NCER                     SC
Region 2                   S/SP
OW/OST                  SC
NCEA-RTP                HHRA
OPEI                     SC
Role(s)

Participant
Participant
Participant
Participant
Participant
Co-chair
Co-chair
Co-chair
Participant/Speaker
Co-chair/Organizer
Co-chair/Organizer
Participant
Participant
Participant
Participant
Participant
Organizer/Speaker
Participant
Participant
Participant
Participant
Participant
Participant
Participant
Co-Chair
Participant
                                       24

-------
Brenda Percovich Foos
Julian Preston
Chris Saint
Phil Sayre
Rita Schoeny
Jennifer Seed
Deborah Segal
Bob Sonawane
Ravi Subramaniam
Greg Susanke
Shirlee Tan

Greg Toth
Les Touart
Larry Valcovic
Vanessa Vu
Vickie Wilson
OA/OCHP
NHEERL-RTP
NCER
OPPTS/OPPT
OW/OST
OPPTS/OPPT
NCER
NCEA-W
NCEA-W
OAA/OSP
AAAS fellow at
OPPTS/OSCP
NERL-Cinci
OPPTS/OSCP
NCEA-W
OA/SAB
NHEERL-RTP
S/SP         Participant
HHRA       Participant
S/SP         Participant
SC          Participant
SC          Participant
SC          Co-Chair
SC          Participant
S/SP         Organizer
S/SP         Participant
ERA         Participant
ERA         Participant

ERA         Participant/Speaker
S/SP         Co-Chair
S/SP         Participant
S/SP         Participant/Speaker
HHRA       Participant/ Speaker
 ERA, Ecological Risk Assessment; HHRA, Human Health Risk Assessment; S/SP, Risk
 Assessments of Susceptible/Sensitive Populations; SC, Hazard Identification: Screening and
 Priori tization
                                        25

-------
APPENDIX C:  COLLOQUIUM AGENDA

9:00 -9:15 a.m.            David Bussard (Director, NCEA-W),
                         Rebecca Klaper (AAAS fellow at NCEA)
                         Welcome and Introduction to the Purpose of the Colloquium

9:15-9:50 a.m.             Vickie Wilson (NHEERL-RTP)
                         The Nuts and Bolts of Genomics Research; Microarrays and
                         Proteomics

9:50-10:25 a.m.            David Dix (NHEERL-RTP)
                         Integration of Toxicogenomics and Risk Assessment:
                         Common Modes of Action and Biomarkers

10:25-10:35 a.m.          Break

10:35-11:40 a.m.          Greg Toth (NERL-Cincinnati)
                         Ecological Risk Assessment Example of Genomics Data

11:40 a.m.-12:15 p.m.      Vanessa Vu (OA/SAB)
                         Development and Implementation of EPA Genomics Action Plan

12:15-1:15 p.m.            Break for Lunch

1:15-1:30 p.m.             Rebecca Klaper (AAAS fellow at NCEA)
                         Information Gathered Regarding the Use of Genomics Data Across
                         the EPA Offices

1:30-1:40 p.m.             Susan Euling (NCEA-W)
                         Charge to the Breakout Groups

1:40-3:40 p.m.             Breakout Group Discussions

3:40-3:50 p.m.             Break

3:50-4:50 p.m.             Reports from Breakout Groups

4:50-5:00 p.m.             Bob Frederick (NCEA-W)
                         Summary of the Day and Close of the Meeting
                                       26

-------
         APPENDIX D: ABSTRACTS FOR SOME OF THE PRESENTATIONS

David Dix (NHEERL-RTP)
Integration of Toxicogenomics and Risk Assessment:  Common Modes of Action and
Biomarkers

       Genomics and proteomics will greatly improve the accuracy of risk assessments by
informing dose and species extrapolations, guiding cumulative assessments based on common
modes of action, and identifying sensitive subpopulations. Integration of toxicogenomic data
into risk assessments will require applicable genomic methods developed by regulatory agencies
and their partners, knowledge of the laboratory and bioinformatic methods that affect outcomes,
and the ability to evaluate the quality of genomic data.  Critical questions to be addressed include
the following:  Can toxicogenomic data identify NOAELs and LOAELs for risk assessment
purposes? Can toxicogenomics be applied to human epidemiology investigations? What
problems/methods in toxicogenomic data analysis have the greatest effect on use in risk
assessments?  EPA should consider establishing ORD-OPPTS working groups to develop the
necessary tools for risk assessors to use toxicogenomics.  This proposed tool kit would include a
regulatory toxicogenomics database, toxicogenomics data quality evaluation software, and
prototype data sets and risk assessments centered on ORD's strengths in carcinogenesis and
reproductive toxicology.

Greg Toth (NERL-Cincinnati)
Ecological Risk Assessment Example of Genomics Data

       Diagnostic and prognostic risk assessments of chemical and biological stressors in
aquatic ecosystems stand to be improved significantly by application of data from the omic
technologies.  Aquatic organisms especially offer the potential to serve as models for the linkage
of exposure and effects models for the prediction of adverse outcomes all the way to the
population level.  EPA/ORD omics research with aquatic organisms, structured significantly by
the emerging framework for computational toxicology, incorporates all of the elements of the
source-to-outcome paradigm. Integration of metabonomics, proteomics, and genomics data to
more completely test hypotheses has become a realistic goal for molecular ecologists in the
immediate future.  EPA/ORD approaches this goal with the potential for huge sequence
resources, instrumentation for advanced proteomics and metabonomics, and an awareness of the
complexities being revealed at the systems biology level. This presentation lays out several
hypotheses to address quantitative risk assessment in this overall context.
                                         27

-------