The Toxicologist - Society of Toxicology

489 USE OF ‘OMICS DATA TO IMPROVE PAH MIXTURES
CANCER RISK ASSESSMENT.
P. McClure 1 , H. Carlson-Lynch 1 , J. Stickney 1 , K. Salinas 1 , I. Cote 2 and L.
Flower 2 . 1 Chemical, Biological, and Environmental Center, SRC, Inc., North
Syracuse, NY and 2 Office of Research and Development, U.S. EPA, Washington, DC.
Cancer risks from PAH mixtures are estimated using slope factors for whole mixtures
(e.g., coke oven emissions) or, when whole-mixture slope factors are not available,
an approach using relative potency factors (RPFs) and a slope factor for
benzo[a]pyrene (BaP). Environmental degradation and compositional variance of
PAH mixtures are key sources of uncertainty in whole-mixture approaches. In the
RPF approach, uncertainty is compounded by the lack of cancer data for many
PAHs. Efforts to decrease uncertainty in these approaches with more rodent bioassay
data are hampered by several issues including high expense. Because omic technologies
hold promise as alternatives or supplements to rodent bioassays, we analyzed
data from recently published in vitro studies comparing gene expression and
post-transcriptional changes from carcinogenic and non-carcinogenic PAHs or
PAHs of differing genotoxic or carcinogenic potencies. In human MCF-7 and
HepG2 cells, carcinogenic PAHs [e.g., BaP, dibenzo[a,l]pyrene (DBalP)] modulated
many more genes than noncarcinogenic PAHs and a correspondence was apparent
between genotoxic and carcinogenic potency and the total number of genes
modulated [DBalP > BaP > benzo[b]fluoranthene > fluoranthene (FA)]. PAHs with
the highest potency (e.g., DBalP, BaP) modulated a number of genes in the p53 signaling
pathway, while noncarcinogenic or low potency PAHs (benzo[e]pyrene, FA)
modulated no genes in this pathway. Studies of proteins in HepG2 cells exposed to
PAHs or PAH-containing soil extracts provide supporting evidence that consideration
of post-transcriptional changes of gene products in the mdm2-p53 network
may enhance the prediction of carcinogenic potential of PAHs or PAH mixtures.
The analysis provides impetus for further research to develop omics-based testing
schemes to predict the carcinogenic potential of untested PAHs or PAH mixtures.
[The views expressed in this abstract are those of the authors and do not represent
the policy of the US EPA.]
490 SOME TRICKS FOR STATISTICAL ANALYSIS OF
NEOPLASTIC LESIONS USING PROC MULTTEST IN SAS.
A. K. Thakur and H. Chen. Nonclinical Safety Assessment, Covance Laboratories,
Vienna, VA.
There are two approaches to statistical evaluation of neoplastic lesions for a carcinogenicity
study- regression techniques using variations of logistic distribution
and interval based or the so-called IARC stratified method. Because of simplicity of
the method and the ease of incorporating exact permutation techniques, the interval
based method has gained popularity among investigators. Various regulatory
agencies also prefer this latter method. The statistical analysis package SAS provides
a procedure called the MULTTEST, which is capable of performing both asymptotic
and exact tests for neoplastic lesions incorporating cause of death information.
The tests are both for association or trend as well as heterogeneity where one compares
the treated group incidences versus control. There are several issues regarding
this procedure under MULTTEST. Irrespective of whether one is interested in asymptotic
or exact trend probability, if one wishes to use the nominal dose levels
which are often unequally spaced, the trend probability one gets is actually the one
based on ordinal or equally spaced dose levels unless one uses the “CONTRAST”
statement specifying the actual dose levels. Also, if the nominal dose levels are not
integers, then the procedure reverts back to equally spaced or ordinal dose levels irrespective
of whether one uses the “CONTRAST” statement. In this latter case, to
get the correct result for trend, one has to integerize the dose levels (such as changing
a dose of 0.01 mg/kg/day to 10 μg/kg/day by multiplying by 100). This conversion
with the “CONTRAST” statement as before will produce the correct trend
probability. Integerizing the dose levels does not change the strength of the trend.
The heterogeneity probabilities are not affected. We will provide examples and
show how to proceed appropriately.
491 A GROUP APPROACH FOR HUMAN HEALTH SAFETY
ASSESSMENT OF FRAGRANCE INGREDIENTS.
A. Api, S. Bhatia, C. Letizia, D. McGinty, J. Scognamiglio and L. W. Smith.
Research Institute for Fragrance Materials, Inc., Woodcliff Lake, NJ.
While chemical structure assessments have been used for some time (Hall,1968,
Ford,2000, Bickers,2003) the safety evaluation of fragrance ingredients presents
unique challenges. Over 2500 chemically defined ingredients in commerce, many
used at low levels, were classified into 92 parent structural groups based on chemical
reactivity. Subsequent subdivisions of major groups were established using the
same principles in order to create structure-activity groups of reasonable size. Some
generalizations are possible: 1) 88% are structurally simple, low molecular weight,
predominantly semi-volatile substances consisting of carbon, hydrogen and oxygen,
2) The majority can be assigned to homologous groups with predicted consistency
of metabolism and toxicity, 3) For most groups, detoxication yields innocuous
metabolites, 4) Systemic exposure levels are typically below thresholds of toxicological
concern. A primary assumption is that the structural homologies allow safety issues
to be considered for several materials within the context of information which
exists for the structural group as a whole, eg, reading across among individual materials
within the group. In many cases existing information for a structural group
may obviate the need to submit a particular individual substance to full toxicological
testing. In other cases, it may be necessary to test one or more members of a
structural class to obtain more robust data to solidify assessment of the class. To
date, 10 groups of fragrance ingredients have been evaluated and published, the
most current being the “Alcohols Branched Chain Saturated” group. Currently 27
groups are in preparation: 6 of these are in final review, 11 are in data analysis and
10 are in data collection and preliminary assessment.
492 IMPLEMENTING COMPUTATIONAL METHODS IN AN
INSTITUTIONAL KNOWLEDGE-BASE AT FDA’S
CENTER FOR FOOD SAFETY AND APPLIED
NUTRITION: A MODE-OF-ACTION-BASED APPROACH
TO BUILDING QSAR MODELS.
L. Ye, R. Brown, E. Busta, K. Arvidson and C. Yang. CFSAN, U.S. FDA, College
Park, MD.
The Chemical Evaluation and Risk Estimation System (CERES) project under development
at FDA emphasizes providing reviewers with easy-to-use decision support
tools for their safety reviews. As part of the knowledge-base, a series of structural
rules and computational models are being developed and implemented. At
FDA CFSAN, a team of toxicologists and chemists provides input and feedback on
computational methods development including the QSAR models. The CERES
project adopted a philosophy of transparent training sets and mode-ofaction(MoA)-driven
models for the development of Quantitative Structure Activity
Relationship (QSAR) models. In this poster, salmonella reverse mutation and in
vitro chromosome aberration endpoints are discussed. These assays are accepted by
FDA Redbook and ICH guidelines for evaluating potential mutagenicity and clastogenicity,
respectively, and further interpreted for potential carcinogenicity. For
salmonella mutation, models were built for classes where chemical reactivity and
metabolic transformations play a role. For chromosomal aberrations, structural
classes that can result in direct DNA damage or interrupt DNA replication were
added to reflect mechanisms. In addition, the experimental conditions of cell types
(CHL or CHO cell lines) and treatment method (pulsed or continuous) were modeled
separately. The individual MoA models were then pulled together to give a
final overall weight-of-evidence result. The training sets are compiled from public
databases and publications as well as FDA CFSAN and CDER data. The QC
process of the dataset preparation and the rules to resolve conflicts from different
data sources will be discussed along with the prediction performance.
493 ASSESSMENT OF MITOCHONDRIAL TOXICITY OF
ENVIRONMENTAL CHEMICALS USING A
QUANTITATIVE HIGH-THROUGHPUT SCREENING
APPROACH.
M. S. Attene-Ramos 1 , R. Huang 1 , S. Sakamuru 1 , S. Shahane 1 , L. Shou 1 , K. L.
Witt 2 , R. R. Tice 2 , C. P. Austin 1 and M. Xia 1 . 1 National Institutes of Health
Chemical Genomics Center (NCGC), Rockville, MD and 2 National Toxicology
Program (NTP), National Institute of Environmental Health Sciences (NIEHS),
Research Triangle Park, NC.
As part of the U.S. Tox21 initiative, the NCGC is developing and optimizing cellbased
and biochemical assays suitable for quantitative high throughput screening
(qHTS) in a 1536-well format. This effort will generate pathway profiles for environmental
compounds that will facilitate the evaluation of mechanisms of toxicity
and prioritization for more extensive testing, as well as the development of predictive
models for in vivo toxicity. In this study, we optimized a mitochondrial membrane
potential assay using the water-soluble JC-10 sensor to detect mitochondrial
depolarization in HepG2 cells and we then used this method to evaluate the mitochondrial
toxicity of 1408 environmental compounds provided by the NTP. In response
to mitochondrial depolarization, the ratio of the cytosolic green fluorescent
monomeric form to the mitochondrial red fluorescent aggregate form increases. Of
the 1408 compounds screened over a 14-point concentration curve (0.5 nM to 92
μM), 44 compounds disrupted the mitochondrial potential in HepG2 cells after
treatment for one hour. We selected 33 compounds for further studies, including
SOT 2011 ANNUAL MEETING 105