Addressing human variability in next-generation human health risk assessments of environmental chemicals Lauren Zeise, Frédéric Y. Bois, Weihsueh A. Chiu, Dale B. Hattis, Ivan Rusyn, Kathryn Z. Guyton

To cite this version:

Lauren Zeise, Frédéric Y. Bois, Weihsueh A. Chiu, Dale B. Hattis, Ivan Rusyn, et al.. Addressing human variability in next-generation human health risk assessments of environmental chemicals. En- vironmental Health Perspectives, National Institute of Environmental Health Sciences, 2013, 121 (1), pp.23-31. ￿10.1289/ehp.1205687￿. ￿ineris-00961796￿

HAL Id: ineris-00961796 https://hal-ineris.archives-ouvertes.fr/ineris-00961796 Submitted on 20 Mar 2014

HAL is a multi-disciplinary open access L’archive ouverte pluridisciplinaire HAL, est archive for the deposit and dissemination of sci- destinée au dépôt et à la diffusion de documents entific research documents, whether they are pub- scientifiques de niveau recherche, publiés ou non, lished or not. The documents may come from émanant des établissements d’enseignement et de teaching and research institutions in France or recherche français ou étrangers, des laboratoires abroad, or from public or private research centers. publics ou privés. Review

Addressing Human Variability in Next-Generation Human Health Risk Assessments of Environmental Chemicals Lauren Zeise,1 Frederic Y. Bois,2 Weihsueh A. Chiu,3 Dale Hattis,4 Ivan Rusyn,5 and Kathryn Z. Guyton3 1Office of Environmental Health Hazard Assessment, California Environmental Protection Agency, Oakland, California, USA;2 Institut National de l’Environnement Industriel et des Risques (INERIS), Verneuil en Halatte, France; 3National Center for Environmental Assessment, U.S. Environmental Protection Agency, Washington, DC, USA; 4George Perkins Marsh Institute, Clark University, Worcester, Massachusetts, USA; 5Department of Environmental Sciences and Engineering, University of North Carolina–Chapel Hill, Chapel Hill, North Carolina, USA

In this review, we explore how next- Background: Characterizing variability in the extent and nature of responses to environmental­ ­generation (“NexGen”) human health risk exposures is a critical aspect of human health risk assessment. assessments of chemicals­ might take advantage Objective: Our goal was to explore how next-generation human health risk assessments may of novel data to better characterize and quan­ better characterize variability in the context of the conceptual framework for the source-to-outcome tify variability in susceptibility, by using and ­continuum. expanding upon current analytical methods. Methods: This review was informed by a National Research Council workshop titled “Biological We begin by describing biological variabil­ Factors that Underlie Individual Susceptibility to Environmental Stressors and Their Implications ity through the conceptual framework of the for Decision-Making.” We considered current experimental and in silico approaches, and emerg- source-to-outcome continuum. Next, the util­ ing data streams (such as genetically defined human cells lines, genetically diverse rodent models, human omic profiling, and genome-wide association studies) that are providing new types of infor- ity of that framework is illustrated in a review mation and models relevant for assessing interindividual­ variability for application to human health of current approaches to describing variability risk assessments of environmental­ chemicals.­ in susceptibility in human health risk assess­ ments. Then, emerging data streams that may Di s c u s s i o n : One challenge for characterizing variability is the wide range of sources of inherent biological variability (e.g., genetic and epigenetic variants) among individuals. A second challenge is be informative in characterizing human vari­ that each particular pair of health outcomes and chemi­cal exposures involves combinations of these ability in susceptibility are described. Finally, sources, which may be further compounded by extrinsic factors (e.g., diet, psychosocial stressors, we consider the opportunities, challenges other exogenous chemi­cal exposures). A third challenge is that different decision contexts present and methods for using emerging data to help distinct needs regarding the identification—and extent of characterization—of inter­individual varia­ assess inter­individual variability in responses bility in the human population. to environmental­ ­chemicals­ across different Conclusions: Despite these inherent challenges, opportunities exist to incorporate evidence from decision contexts. emerging data streams for addressing interindividual­ variability in a range of decision-making contexts. Susceptibility as a Function Key words: environmental­ agents, genetics, human health risk assessment, modeling, omics tech- of the Source-to-Outcome nologies, susceptible populations, variability. Environ Health Perspect 121:23–31 (2013). http:// dx.doi.org/10.1289/ehp.1205687 [Online 19 October 2012] Continuum and Biological Variability The “source-to-outcome continuum” [U.S. Human variability underlies differences in the authority; the available time, resources, and Environmental Protection Agency (EPA) degrees and ways in which people respond expertise to collect data and conduct analyses; 2007; NRC 2007] is a conceptual frame­ to environ­mental chemi­cals, and address­ and stakeholder concerns. work for human health risk assessment of ing these differences is a key consideration Over the past decade, efforts to systemati­ environmental­ chemicals­ in which changes in in human health risk assessments for chemi­ cally “map” human variability have expanded the sources of chemi­cals in the environment cals [Guyton et al. 2009; Hattis et al. 2002; dramatically, focusing mainly on genetic National Research Council (NRC) 2009]. variation (Schadt and Björkegren 2012). In Address correspondence to L. Zeise, California A large array of possible health outcomes is addition to genetic differences, omics stud­ Environmental Protection Agency, 1515 Clay St., of concern for such assessments, and many ies have examined the impact of epigenetic, 16th Floor, Oakland, CA 94612 USA. Telephone: sources of variation can influence the sever­ transcriptomic, proteomic, and metabolo­ (510) 622-3195. Fax: (510) 622-3211. E-mail: ity and frequency of the adverse effects at dif­ mic variation on disease susceptibility, [email protected] This review was informed by the discussions and ferent exposure levels. These sources may be prognosis, or options for pharmacotherapy presentations at a National Research Council (NRC) intrinsic (e.g., heritable traits, life stage, aging), (Chen et al. 2008; Emilsson et al. 2008; Illig workshop titled “Biological Factors that Underlie or extrinsic, exogenous, and acquired (e.g., et al. 2010; Manolio 2010; Schadt 2009). Individual Susceptibility to Environmental Stressors background health conditions, co-occurring Tailored chemotherapy treatment based on and Their Implications for Decision-Making” held in chemi­cal exposures, food and nutrition status, patient (Phillips and Mallal 2010) or tumor April 2012 in Washington, DC. psychosocial stressors). Interactions between (La Thangue and Kerr 2011) genetics is an We thank the staff, particularly K. Sawyer and M. Shelton-Davenport, and members of the NRC’s inherent and extrinsic factors create the large example of a significant success in applying Committee on Emerging Science for Environmental range of biological variation exhibited in such discoveries; however, for many diseases, Health Decisions. We also thank I. Cote for her response to a chemical­ exposure (NRC 2009). the substantial nongenetic variation in disease thoughtful comments. Given that biological variability in susceptibil­ or treatment outcomes has limited their util­ The views in this article are those of the authors, ity is context-­dependent, so too is the extent to ity. Thus, the characterization of the broad and do not necessarily reflect the views or policies which it needs to be described and quantified set of environ­mental factors, including those of the U.S. Environmental Protection Agency or the California Office of Environmental Health Hazard to inform any particular environ­mental deci­ related to chemical­ exposures, that may con­ Assessment. sion. The salience of variability information for tribute to disease is directly relevant to both The authors declare they have no actual or potential specific choices is affected by the range of avail­ personal­ized medicine and environ­mental competing financial interests. able risk management options; the regulatory health protection (Khoury et al. 2011). Received 28 June 2012; accepted 19 October 2012.

Environmental Health Perspectives • v o l u m e 121 | n u m b e r 1 | January 2013 23 Zeise et al. are further propagated within the individual • Internal doses are the amounts/concentrations altered by interactions with environ­mental through a series of biological and physiologi­ of environmental­ chemicals­ or their metabo­ chemicals­ or their metabolites, and are related cal steps that may ultimately manifest as an lites at the target site(s) of interaction with to internal doses by pharmaco­dynamic (PD) adverse health outcome (Figure 1): biological molecules, and are related to exter­ processes. Variation leading to differential • Source/media concentrations are measures nal doses by pharmaco­kinetic (PK) processes. susceptibility can stem from differences in of the chemi­cal, which may change under Susceptibility may arise from differences in transport systems, receptors and/or proteins specific risk management options being con­ compartment sizes and composition (e.g., fat in other toxicity pathways, as well as repair sidered. A given risk management decision concentration in plasma, which rises during capacity (of, for example, DNA), which in may differentially affect media concentra­ pregnancy) (Roy et al. 1994), as well as dif­ turn are affected by intrinsic and extrinsic tions depending on local conditions. ferences in the rates of uptake (e.g., fraction factors such as genetics and life stage. • External doses are measures of exposure (e.g., absorbed from diet or air), metabolism, elim­ • Physiological/health status reflects the overall concentration in air × breathing rate per body ination, and transport to sites of action (e.g., state, structure, or function of the organ­ weight) to or intake (e.g., amount ingested the blood–brain barrier). Such differences ism and is related to biological responses per body weight) of environmental­ chemicals,­ may be due, for example, to genetics (e.g., through systems dynamics, the underlying and are related to source/media concentra­ via polymorphisms­ in metabolic enzymes, physiological status of the host to which tions by exposure pathways. Sources of vari­ uptake and efflux transporters), other chemi­ the chemical­ -specific perturbation is added. ability that may confer susceptibility include cal exposures (via metabolic­ enzyme induc­ Examples include maintenance and adapta­ differences in behaviors, such as breathing tion and inhibition), and preexisting health tion processes (associated with preexisting rates, water consumption, and dietary habits conditions and life stage (e.g., via metabolism health conditions, sex hormone levels, for (e.g., the amount of fish consumed), and, and mobilization from tissue storage). example), and the accumulation of dam­ in an occupational context, use of personal • Biological responses are measures of biological age events from past exposures (e.g., loss ­protective equipment. state (e.g., the concentration of glutathione)­ of alveolar septa from past cigarette smoke

Source-to-outcome continuum Types of biological Source/media concentrations Susceptibility variability indicators Multiple sources leading to chemicals in multiple media

Exposure Heredity Exposure (genetic and parameters epigenetic) External doses Background and Multiple chemicals via coexposure doses Sex, Modifying how multiple routes lifestage, and changes in aging source/media Pharmacokinetics PK concentrations are parameters propagated to changes in outcome. InternalIl concentrationsi Existing health Endogenous conditions Multiple chemicals (including metabolites) concentrations at multiple target sites

Pharmacodynamics PD Coexposures parameters (sources outside decision context) BiologicalBi lig l responsep measurements For fixed Baseline biomarker source/media Multiple biological responses in values concentrations, multiple tissues/biological media modifying the Food/nutrition background or Systems dynamics Systems baseline parameters conditions.

Physiological/healthy g / status Outcome latency, Psychosocial likelihood, and e stressors ase d e as severity isease is ise Normal d d function y arl Late disease Early d Perturbed Perturbe E Late

Figure 1. Framework illustrating how susceptibility arises from variability. Multiple types of biological variability intersect with the source-to-outcome continuum, either by modifying how changes to source/media concentrations propagate through to health outcomes or by modifying the baseline conditions along the con- tinuum. The aggregate result of all these modifications is variability in how a risk management decision impacts individual health outcomes. The parameters­ and initial conditions along the source-to-outcome continuum serve as indicators of differential susceptibility, some of which are more or less influential to the overall outcome (see Figure 2).

24 v o l u m e 121 | n u m b e r 1 | January 2013 • Environmental Health Perspectives Addressing human variability in NexGen assessments

exposure). Variation in these can confer data-driven estimation of the likely impact of through independent experiments, clinical susceptibility by altering the likelihood of interindividual­ variability on human health measurements, or surveillance. Table 1 lists progression from normal function to mild risk assessments. some examples of data sources for developing perturbations, early disease, and late disease. For presumed nonthreshold cancer end a priori parame­ter distributions. Monte Carlo Systems dynamics describes the propaga­ points, interindividual­ variability is not cur­ simulations are used to propagate the distribu­ tion of biological perturbations regardless rently addressed when risk is estimated from tions from model parame­ters to model pre­ of whether they are due to chemi­cal expo­ animal studies, with the exception that for dictions (Portier and Kaplan 1989; Spear and sure, thus distinguishing it from pharmaco­ mutagenic compounds exposures occurring Bois 1994). A third approach, the “Bayesian dynamics, which describes how chemi­cal early in life are weighted more heavily (by a fac­ PBPK” approach, offers a synthesis of the other exposure causes biological perturbations.­ tor of 10 between birth and 2 years of age, and two, applying mechanism- or chemical­ -specific Figure 2 illustrates the distinct effects of a factor of 3 between 2 and 16 years of age). parame­ter variability data from a variety of different sources of variability on external Cancer risk for susceptible populations, such as independent sources while using population dose, internal dose, or biological response. The smokers who have been exposed to radon, may observations of relevant biomarkers of internal first category of biological variability is indi­ be calculated in addition to that for a general exposure and effect to further inform parame­ cated by differences in the parame­ters gov­ population (U.S. EPA 2003). Alternatively, ter variability (Allen et al. 2007; Bernillon erning the relationship of one measureable adjustments may be made to address suscep­ and Bois 2000; Hack 2006). Parameter cova­ quantity to the next (e.g., external to internal tible subgroups, such as the sex-specific effects riance can be modeled by multivariate prior dose, and internal dose to biological response) of 1,3-butadiene (U.S. EPA 2002). There have distributions (Burmaster and Murray 1998) (Figure 2A,B). In addition, there may be bio­ been calls to formally account for variability in or joint posterior distributions obtained by logical variability in the initial conditions for cancer dose response (NRC 2009). Bayesian multilevel­ modeling (Bois et al. 1990; each measureable quantity, as well as the con­ Over the past 30 years, several strategies Wakefield 1996). A Bayesian PBPK model- tribution from the source of environ­mental to characterize (predominantly PK) variability based analysis of the population toxicokinetics chemi­cal exposure under consideration for combining mathematical models and statistical of trichloroethylene­ (TCE) and its metabolites risk management (Figure 2C,D). For exam­ distributions have developed in parallel. The in mice, rats, and humans provides a practical ple, increases in background exposure to the first strategy, mostly used for data-rich pharma­ example of how a systematic­ method of simul­ same or a different chemi­cal(s) may result ceuticals, couples empirical PK models and taneously estimating model parame­ters and in saturation of metabolic activation and/or multilevel (random effect) statistical models to characterizing their uncertainty and variability clearance processes, or temporary depletion extract a posteriori estimates of variability from can be applied to a large database of studies on of cofactors involved in detoxification, such clinical data on patients or volunteers. This a chemical­ with complex toxicokinetics (Chiu as glutathione, resulting in either attenuation “population PK” approach (Beal and Sheiner et al. 2009). or amplification of the effect of additional 1982) seeks to measure variability and to dis­ PBPK models have been often used to increments of chemi­cal exposure on internal cover its determinants. The second, the “pre­ assess variability on the basis of prior parame­ dose (Figure 2C). Nonetheless, a biological dictive PK,” approach takes advantage of the ter distributions obtained from in vitro experi­ response with a low background level may be predictive capacity of mechanistic models and ments or the physiological literature (Bois much less altered by additional exposure than assigns a priori distributions to their parameters­ et al. 2010; Jamei et al. 2009) and can include one with a high background because of to the (e.g., blood flows, organ volumes). The parame­ genetic information regarding variability. cooperativity associated with a relatively higher ters having biological meaning can be observed For example, PBPK models can inform the baseline internal dose (Figure 2D). 1.0 0.25 Current Approaches 0.9 Different biological 0.8 0.2 to Addressing Variable response depending 0.7 on PD parameters Susceptibility 0.6 0.15 Variability for assumed threshold-like dose– 0.5 0.4 Different 0.1 response relationships is currently addressed internal dose

Internal dose 0.3 by applying an “uncertainty” or “adjustment” depending on 0.2 PK parameters 0.05 factor (U.S. EPA 2011). The factor to account 0.1 Biological response for interindividual­ variability in human popu­ 0 0 lation has typically been 1, 3, or 10. In some 05 10 15 20 25 30 0 0.5 1.0 1.5 cases, the factor is further divided to separately External dose Internal dose account for variation in PK and PD (U.S. 0.8 Fixed change in 0.25 Fixed change in EPA 2011; International Programme for 0.7 external dose internal dose due to source 0.2 due to source Chemical Safety 2001). In this context, PD 0.6 has included both PD and systems dynamics 0.5 0.15 processes described above and in Figure 1. 0.4 Different change Data permitting, the PK component can 0.3 0.1 in internal dose Different change be addressed through physiologically based Internal dose 0.2 depending on in response background/ 0.05 Biological response depending on pharmaco­kinetic (PBPK) modeling, in which 0.1 coexposure doses endogenous internal dose case a factor addressing only PD is applied 0 0 (U.S. EPA 2011). Occasionally, exposure– 0 5 10 15 20 25 30 0 0.5 1.0 1.5 effect observations are available for particularly Total external dose Total internal dose susceptible human populations, such as with Figure 2. Effects of variability in PK (A), PD (B), background/coexposures (C), and endogenous concentra- ozone and persons with asthma (U.S. EPA tions (D). In (A) and (B), individuals differ in PK or PD parame­ters. In (C) and (D), individuals have different 2006), or those sensitive to chronic beryllium initial baseline conditions (e.g., exposure to sources outside of the risk management decision context; disease (U.S. EPA 1998), which allows for a endogenously produced compounds).

Environmental Health Perspectives • v o l u m e 121 | n u m b e r 1 | January 2013 25 Zeise et al. implications of polymorphisms in metabolism and focus on PD rather than system dynam­ populations and for additional compounds genes (Johanson et al. 1999). The effects of ics elements of the disease process and do not (Watson et al. 2011). Furthermore, such stud­ such polymorphisms on PK of environ­mental attempt to model the full process from tissue ies may potentially inform the prioritiza­tion of toxicants and drugs have been the subject of exposure to disease outcome. chemotherapeutic drugs with a sizable genetic many empirical studies (reviewed by Ginsberg response component for future investigation et al. 2009c, 2010). These polymorphisms are Emerging Data Streams (Peters et al. 2011) and assist in identifying of particular concern for xenobiotics whose on Biological Variability germline predictors of cancer treatment out­ metabolic fate or mechanism(s) of action is Experimental population-based paradigms comes (Huang et al. 2011). controlled by a particular enzyme (Ginsberg to address intrinsic variability in response to The utility of such in vitro models to et al. 2010), and in such cases genetic variabil­ exposure comprise multiple levels of biological toxicol­ogy, especially for exploring the extent ity can profoundly influence enzyme function organization, from molecules to whole bodies. and nature of genetic components of inter­ with implications for internal dose (Figure 1). Published examples, reviewed by Rusyn et al. individual variability in PD and systems However, because enzymatic pathways with (2010), include animal models and large- dynamics, was recently demonstrated (Lock overlapping or redundant function and other scale in vitro screening platforms to study et al. 2012; O’Shea et al. 2011). Quantitative pharmacokinetic­ factors (e.g., blood flow population-­based genetic determinants. Those high-throughput screening (qHTS) pro­ limitation) can also influence metabolic fate studies have also aided in the identification of duced robust and reproducible data on intra­ (Kedderis 1997), PBPK models are needed to genetic susceptibility factors that underlie toxi­ cellular levels of adenosine triphosphate and evaluate the implication of genetic polymor­ city phenotypes. Complementary to these are caspase-3/7 activity (i.e., biological response) phisms in metabolizing enzymes in human genome-wide (Hutter et al. 2012) and expo­ indica­tive of general cytotoxicity and activa­ health risk assessment (Ginsberg et al. 2010). sure-wide (Patel et al. 2010) association stud­ tion of apoptosis (i.e., physiological status), The situation is somewhat different for PD ies for assessing human population variability.­ with utility for variability assessment as fol­ and systems toxicology models. The biologically Experimental in vitro data on genetic lows. First, standardized and high-quality based dose–response models describe apical or variability. Human cell lines obtained from concentration–response profiling, with repro­ intermediate end point responses as a function genetically diverse subjects and multiple popu­ ducibility confirmed by comparison with pre­ of PK-defined internal doses (Crump et al. lations (Durbin et al. 2010) hold the promise vious experiments, enables prioritization of 2010). However, models designed purely from of providing data for assessing genetic deter­ chemi­cals based on inter­individual variability our understanding of the disease process, such minants of different components of toxic in cytotoxicity. Second, genome-wide associa­ as the role of cytotoxicity and regenerative­ pro­ response. Many recent studies have used tion analysis of cytotoxicity phenotypes allows liferation in carcinogenesis (Luke et al. 2010b), human lymphoblastoid cell lines, represen­ exploration of the potential genetic determi­ or the effect of dietary iodide and thyroid hor­ tative of the genetic diversity in populations nants of that variability. Finally, the highly mones on the hypothalamic–­pituitary–­thyroid of European, African, Asian, and North and significant associations between basal gene axis (McLanahan et al. 2008), require further South American ancestry, to quantify inter­ expression variability and chemi­cal-induced development to reliably predict an adverse out­ individual and interpopulation variability in toxicity suggest plausible mode-of-action come from tissue exposure (the last two arrows response to drugs (Welsh et al. 2009). Dozens hypotheses for follow-up analyses. in Figure 1), or its variability. Understanding of studies published in the past 5 years have Several extensions of these studies can be a disease process at the pathway level (i.e., profiled the cytotoxicity of single to as many envisioned to advance the identification of PD and systems dynamics components of the as 30 drugs (mostly chemotherapeutics) in determinants of genetic susceptibility and vari­ source-to-­outcome continuum) is in itself not hundreds of cell lines. Diverse applications ability in toxic response. Opportunities include sufficient to define reliable and informative for such a population-based cell model has the testing of additional, and more diverse, mechanistic models because of great model sen­ been suggested. Drug class–specific signatures chemicals­ (including major metabolites) and sitivity to uncertain inputs. Most such models of cytotoxicity, which could indicate possi­ concentrations (to account for lower meta­ are based on equations derived from the classi­ ble shared mechanisms, have been identified bolic capacity of these cells). Other specific end cal receptor theory (Csajka and Verotta 2006) and replicated in both cell lines from different points could also be assessed. Further, these studies could be expanded to include larger Table 1. Examples of data sources for modeling PK and PD variability. panels of lymphoblasts and other cell types Example References from genetically and geographically diverse Variability in human phase I and phase II metabolism and renal excretion, Dorne 2010; Ginsberg et al. 2002, 2004; populations. Development of related assay sys­ including in different age groups–neonates, children, and the elderly Hattis et al. 2003 tems to monitor differences in susceptibility to Compilations of genetic polymorphisms of specific metabolic enzyme activities: perturbation of communication between cells Paraoxonase Ginsberg et al. 2009a (e.g., neurotransmission or differentiation sig­ N-Acetyltransferase 1 and 2 Bois et al. 1995; Walker et al. 2009 nals) could address other aspects of variability Glutathione transferases Ginsberg et al. 2009b not present in cultures comprising only one CYP2D6 (cytochrome P450 2D6) Neafsey et al. 2009b kind of cell. The development and use of these CYP2E1 (cytochrome P450 2E1) Neafsey et al. 2009a and other types of in vitro assays would be fur­ ALDH2 (acetaldehyde dehydrogenase 2) Ginsberg et al. 2009c ther informed by quantitative comparisons of Human biomonitoring observations of interindividual­ differences in Bois et al. 1996 the PD inter­individual variability measured biomarkers of exposure (e.g., chemi­cal-protein adducts) or in levels of parent/metabolite in vitro with observable human pharmaco­ Variability in physiological parameters­ for older adults: bodymass, Thompson et al. 2009 dynamics variability in vivo. Candidate chemi­ surface area, body mass index, health status cals for this comparison would be selected Indicators of PD variability environ­mental toxicants (such as ozone) and Human DNA repair enzyme XRCC1 Ginsberg et al. 2011 pharmaceuticals that have been tested for Human host defense enzymes Ginsberg et al. 2010 responses in appreciable numbers of human Lung function response to particulate matter Hattis et al. 2001 subjects at different known exposure levels. The Susceptibility to infectious organisms Hattis 1997 extent of interindividual­ variability in response

26 v o l u m e 121 | n u m b e r 1 | January 2013 • Environmental Health Perspectives Addressing human variability in NexGen assessments that was observed for different chemi­cals in potential interplay with the genetic back­ have long had important roles in quantifying in vitro assays could also be compared with ground, in susceptibility. For example, human variability in the risks of exposures previously collected sets of in vivo human PD Koturbash et al. (2011) demonstrated that to widespread toxicants such as ozone and variability data (Hattis et al. 2002). interstrain differences in susceptibility to 1,3- airborne particulates. The addition of GWAS Experimental in vivo data. Several proof- butadiene–induced genotoxicity may be due to these established tools has the potential of-concept studies that utilized a “mouse model to strain-specific epigenetic events that are also to widen the capability for quantification of of the human population” have demonstrated­ part of a PD response. effects on susceptibility of many individual the potential for translation to clinical appli­ Practical use of this type of experimental genotypic variants that individually have rela­ cations and for addressing both PK and PD information is possible mainly when the mech­ tively modest effects (Holloway et al. 2012). components of variability (Guo et al. 2006, anistic pathways to human adverse responses Establishing the roles of individual pathways 2007; Harrill et al. 2009b; Kleeberger et al. are better established. More general application in affecting susceptibility via genetic analysis, 1997; Prows et al. 1997). For example, the will also depend on the development of suites in turn, has the potential to advance the assess­ extent and nature of TCE metabolism is an of rodent models that more fully represent ment of effects of other exposures during life important consideration in relating adverse human diversity in both genetics and other that also affect the same pathways. Elucidating health effects in rodents to humans. Bradford factors, such as age (Hamade et al. 2010). Such these determinants for prominent toxicants, et al. (2011) measured variability in PK for studies can, in turn, provide important insights however, requires a very considerable research TCE using a panel of inbred mouse strains, concerning the identity and extent of sources effort. Nonetheless, this research paradigm revealing marked differences among indi­ of variability that may arise in the source-to- provides opportunities to explore variability in vidual mice (e.g., a greater than 4-fold differ­ outcome continuum for a given chemi­cal class, adverse responses that is due to physiological ence in peak serum concentrations of TCE physiologic state, or adverse response. states for which in vitro and experimental ani­ metabolites). These experimental data on intra­ Human clinical and observational data. mal models are lacking. species differences in TCE metabolism may be Genome-wide association studies (GWAS) Variability in human response to an agent used to calibrate the variability in outputs of with disease severity as the phenotypic trait stems in part from differences in the under­ PBPK models, and thus inform quantitative are used to associate genetic loci with risk for lying exposures that contribute to a given dis­ assessment of variability in TCE metabolism complex diseases (Rosenberg et al. 2010). ease response prevalence within the population. across species. Even though GWAS approaches have uncov­ A person’s internal “chemi­cal environment” With regard to PD variability, genetically ered numerous genomic loci that may affect may be as important for possible disease asso­ diverse mouse strains can be used to under­ the risk of human disease (Manolio 2010), ciations as exposures to the variety of chemicals­ stand and predict adverse toxicity in hetero­ the identified variants explain only a small in the external environment. Under this “expo­ geneous human populations. For example, proportion of the heritability of most complex some” concept (Wild 2005), exposures include Harrill et al. (2009a) evaluated the role diseases (Manolio et al. 2009). Some have sug­ environmental­ agents and internally generated of genetic factors in susceptibility to gested that unexplained heritability could be toxicants produced by the gut flora, inflamma­ acetaminophen-­induced liver injury in a panel partly due to gene × environment interactions, tion, oxidative stress, lipid peroxidation, infec­ of inbred mouse strains and two cohorts of or complex pathways involving multiple genes tions, and other natural biological processes human volunteers. The authors identified and exposures (Schadt and Björkegren 2012). (Rappaport and Smith 2010). genes associated with differential susceptibility The GWAS concept is now being applied to toxicity in a preclinical phase. This finding to identify additional genotype-dependent Advances in in Silico Methods has the potential to focus further toxico­ metabolic phenotypes and to gain insight to Address Human Variability genetics research, overcome the challenges of into nongenetic factors that contribute to the Modeling of variability is expected to be studies in small human cohorts, and shorten effects of xenobiotics on system dynamics. In needed for both data-rich and -sparse chemi­ the validation period. The data acquired animal studies, metabolic phenotype-related cals. Recent advances in software, publicly with this model may be used in analyses of quantitative trait loci were shown to be use­ available data and ongoing computational individual risk to toxicants. Furthermore, ful in understanding genome × ­phenotype activities in biomedical research should facili­ when combined with omics data collected on relationships and how extended genome tate the development and use of the results of an exposed population of individual strains, (microbiome) perturbations may affect dis­ this type of modeling. it may be possible to explore underlying ease processes through transgenomic effects Modeling the PK dimension of human genotype-­dependent and -independent (Dumas et al. 2007). In a series of human variability. Commercial software prod­ toxicity pathways involved in PD response studies (Gieger et al. 2008; Illig et al. 2010; ucts [e.g., by Simcyp (http://www.simcyp. (Bradford et al. 2011; Harrill et al. 2009a). Suhre et al. 2011), serum collected from two com), Bayer Technology (http://www.pksim. Experiments such as these afford the large European cohorts (2,820 individuals in com)] are available to explicitly address vari­ opportunity to quantitatively understand the total) was analyzed with nontargeted metabo­ ability for pharmaceutical or human health interplay between genetics, PD, and systems lomics, focusing on endogenous metabolites risk assessment applications to, for example, dynamics. In addition, genetically defined and covering 60 biochemical­ pathways. Ratios adjust dosing for different target patient popu­ mouse models may be used to supplement of metabolites to parent chemical­ concentra­ lations (Jamei et al. 2009; Willmann et al. the limited data from human studies to not tions served as surrogates for enzymatic rate 2007). Several of these offer generic PBPK only discover the genetic determinants of constants. Thirty-seven genes were associated models, applicable to “any” substance; how­ susceptibility and understand the molecular with blood metabolite concentrations and, in ever, their substance-specific parameters­ have underpinnings of toxicity (Harrill et al. some cases, explained a substantial fraction to be obtained from in vitro experiments 2009a; Koturbash et al. 2011) but also to of the variance. Endogenous and xenobiotic (particularly on metabolism) or quantitative develop descriptions of variability for use in metabolites (mostly of drugs) were studied. structure–property relationships. The variabil­ dose–­response and mechanistic evaluation Clinical (Brown et al. 2008; Hernandez ity of subject-specific physiological parameters­ components of human health risk assessments. et al. 2010) and epidemiological (Jia et al. can be informed by compiled databases (see Such rodent systems can also be used to 2011; Wood et al. 2010) studies of acute above) and literature searches (Bois et al. 2010; assess the role of epigenetics, as well as its and chronic effects of ambient air exposures Ginsberg et al. 2009c), and could include

Environmental Health Perspectives • v o l u m e 121 | n u m b e r 1 | January 2013 27 Zeise et al. adjustments or protocols to address limitations applications other than toxicant risk evaluation. many markers (e.g., of apoptosis, cell divi­ in data availability. Quantitative structure–­ Further, because of these large-scale efforts, the sion), the linkage to risk is highly uncertain property relationship models or in vitro data necessity of sharing and standardization is well (Woodruff et al. 2008), so the ranges of pos­ can also be used to derive substance-specific understood in the United States. The systems sible variability may be very large. Further, parameters.­ These models are being applied in biology markup language (Hucka et al. 2003), the ability to reinforce information by linking an exploratory fashion in in vitro–based assess­ for example, is a high-level language developed with the impact of injury on multiple targets ments (Judson et al. 2011; Rotroff et al. 2010). explicitly to provide a common intermediate is also limited because such links are generally Using a Bayesian multilevel population format for representing and exchanging sys­ not well understood. approach, some of the key parame­ters of tems biology models. Predictive toxicology will these generic models could be calibrated by benefit from these developments. Implications for NexGen Human integrating human observational data with data The frontier for both PK and PD is in the Health Risk Assessments from lower levels of biological organization. integration of the rapidly growing informa­ Multiple “tiers” of human health risk assess­ This presents a computational challenge on tion about metabolic networks, receptors, and ment needs, requiring different levels of a chemi­cal-specific basis, because those their regulation with toxicity pathways. The precision, can be envisioned. These include models are neither particularly parsimonious models so far most amenable to quantitative screening-level analyses of multiple chemicals­ nor quickly evaluated. Yet an extensive predictions are differential equation models. to inform the prioritization of management calibration of a complex generic model for a PBPK models will likely be merged with and enforcement actions across communities, selected number of data-rich environmental­ systems biology and virtual human models. ensuring protection across the population to or pharma­ceutical chemi­cals could be used The boundary between PK and PD actually widespread exposure to legacy contaminants, as support to develop generic approaches for tends to blur as metabolism becomes more or identifying subpopulations for which differ­ PK variability­ treatment in human health risk and more integrated into detailed models of ing risk management options might be applied. assessment. For example, generalizations could toxicity pathways when, for example, model­ In the lowest (simplest) tier of assessments, be made about the extent to which particular ing enzymatic induction by xenobiotics (Bois evaluations are expected to primarily rely on enzymes may contribute to overall human PK. 2010; Luke et al. 2010a). The variability of the the results of high- and medium-­throughput Extensions of the approach of Hattis et al. different components of those models will be in vitro screening tests in mostly human cell (2002) can also be developed to construct directly informed by time series of genomic, lines, as well as complementary in silico predic­ “bottom up” quantitative descriptions of PK proteomic, metabolomic data on the chemi­ tive methods. The Tox21 collaboration (Collins variability that can be applied as defaults across cal species considered. This may provide et al. 2008) is leading the field in exploring how classes of chemicals.­ a framework for assessing the variability in a broad spectrum of in vitro assays, many in Modeling the PD dimension of human susceptibility to chemically­ induced effects qHTS format, can be used to screen thousands variability. Semi-empirical PD models can as influenced by possible metabolic interac­ of environ­mental chemi­cals for their potential include observed biomarkers of susceptibil­ tions as well as preexisting disease. In time to disturb biological pathways that may result ity as covariates. Such models are increas­ this may facilitate computing the impact of, in human disease (Xia et al. 2008). Such data ingly applied in predictive toxicity and for example, single nucleotide polymorphisms on toxicologically relevant in vitro end points human health risk assessment. Environmental on the reaction rates of enzymes and recep­ can be used as toxicity-­based triggers to assist epidemiol­ogy also routinely models quantal tors and translating these calculations to esti­ in decision making (Reif et al. 2010), as predic­ types of biomarker data in logistic regressions. mates of human variability (Mortensen and tive surrogates for in vivo toxicity (Martin et al. Harmonizing the tools and models of toxico­ Euling 2011). Ongoing work on simula­tions 2010; Zhu et al. 2008), to generate testable logical risk assessment with those of epidemio­ of enzymatic reactions or receptor binding at hypotheses on the mechanisms of toxicity (Xia logical risk assessment, and reconciling their the atomic level (e.g., the potassium channel et al. 2009), and to develop screening assays data and results, should facilitate the develop­ pore) shows the way forward for predicting based on pathway perturbations. The extent of ment of better approaches for background fundamental reaction rates by physical chem­ inter­individual variability in toxic response to and variability descriptions in NexGen human istry approaches. Prediction of the quantitative be estimated from these types of assays can be health risk assessments. impact of sequence or amino-acid variation on informed by empirical data and PK/PD models Integrating PK and PD into a systems the function of the reactive species involved in that address multiple factors in the source-­to- biology framework. The link between toxicity systems biology models is coming within reach ­outcome continuum as described in Figure 1. pathway and “normal cell physiology” models (Giorgino et al. 2010; Sadiq et al. 2010). The genomic component of variability may of systems biology could also be further devel­ Biologically based PD models, such as the be partially informed by test data from geneti­ oped and used as the basis to explore poten­ systems biology models of response networks cally diverse but well-defined human cell lines, tial ranges of human variability. The potential (Schuster 2008), models of toxicity pathway such as from the HapMap (http://hapmap. of publicly accessible and curated biomodel perturbations, and biologically based dose– ncbi.nlm.nih.gov/) and 1000 Genomes (http:// and database repositories will be increasingly response models proposed to link biochemical­ www.1000genomes.org/) projects. For exam­ exploited as familiarity increases in the risk responses to apical effects, clearly hold promise ple, emerging data based on standardized and assessment and risk management communi­ (Csajka and Verotta 2006; Jonsson et al. 2007; high-quality concentration–­response profiling ties. Importantly, systems biology models can Nong et al. 2008) but face challenges similar can help inform characterizations of the extent describe background biological processes and to those that hampered the use of biologically of inter­individual variability in cytotoxicity. the impact of their perturbation and provide a based cancer models (Bois and Compton- When chemi­cal-specific estimates are lack­ framework for exploring human variability and Quintana 1992; Chiu et al. 2010). To explore ing, the range of interindividual­ variability for identifying susceptible populations for targeted the extent of human variability in response to structurally related compounds may be infor­ assessment and management efforts. Although toxicant and stressor exposures, the various mative, in a read-across approach. Quantitative they come at the price of tremendous complex­ steps in the relevant causal path need to be data characterizing the range in response (e.g., ity, their development can leverage the consid­ modeled quantitatively and on a population size and variance) may be integrated with erable ongoing effort by the biomedical and basis. A problem is that the quantitative link­ probabilistic default distribu­tions addressing pharmaceutical research community to support ing of omics biomarkers to risk is missing. For the remaining key sources of interindividual­

28 v o l u m e 121 | n u m b e r 1 | January 2013 • Environmental Health Perspectives Addressing human variability in NexGen assessments

variability. Quantitative estimates of PK vari­ hypothesis-based molecular clinical and epide­ tetra­chloroethylene in mice, rats, and humans. Toxicol ability would be also incorporated. In addition, miological approaches that integrate emerging Appl Pharmacol 102(2):300–315. Bradford BU, Lock EF, Kosyk O, Kim S, Uehara T, Harbourt D, factors such as life stage and background expo­ biological knowledge of pathways with obser­ et al. 2011. Interstrain differences in the liver effects of sures may be particularly important consider­ vations of physiological disease status, mark­ trichloroethylene in a multistrain panel of inbred mice. ations for approaches accounting for baseline ers of early biological response, and genetics Toxicol Sci 120(1):206–217. differences in the spectrum of the “chemical­ are likely to provide the way forward with Brown JS, Bateson TF, McDonnell WF. 2008. Effects of exposure to 0.06 ppm ozone on FEV1 in humans: a secondary analysis environment” (Rappaport and Smith 2010), ­population-based descriptions of variability. of existing data. Environ Health Perspect 116:1023–1026. in interpreting results from the omics assays, Burmaster DE, Murray DM. 1998. A trivariate distribution Conclusions for the height, weight, and fat of adult men. Risk Anal and in evaluating the potential contributions of 18(4):385–389. nongenetic­ variability factors. Emerging data streams can inform multiple Chen Y, Zhu J, Lum PY, Yang X, Pinto S, MacNeil DJ, et al. At these lower tiers, a probability distri­ aspects of biological variability, be used in 2008. Variations in DNA elucidate molecular networks that cause disease. Nature 452(7186):429–435. bution may best acknowledge the many uncer­ different modeling approaches addressing Chiu WA, Euling SY, Scott CS, Subramaniam RP. 2010. tainties involved in making inferences with PK and/or PD variability, and have applica­ Approaches to advancing quantitative human health limited data. Systematic analyses of chemi­cal tion across different chemi­cal screening and risk assessment of environ­mental chemi­cals in the post- sets will be needed to refine distributions for evaluation schemes. Successful examples of genomic era. Toxicol Appl Pharmacol; doi.org/10.1016/j. taap.2010.03.019 [Online 29 March 2010]. the chemi­cal-specific and general case. For addressing PK variability include the devel­ Chiu WA, Okino MS, Evans MV. 2009. Characterizing uncertainty instance, external comparisons of in vitro mea­ opment and application of a Bayesian PBPK and population variability in the toxicokinetics of trichloro- sures based on genetic variability in pharmaco­ model–based analysis systematically­ estimat­ ethylene and metabolites in mice, rats, and humans using an updated database, physiologically based pharmaco­ dynamics to in vivo observations may inform ing model parameters­ and characterizing their kinetic (PBPK) model, and Bayesian approach. Toxicol the choice of distribution used for a particular uncertainty and variability for TCE, a chemi­ Appl Pharmacol 241(1):36–60. chemical­ or chemical­ category. Standard cate­ cal with complex toxicokinetics (Chiu et al. Collins FS, Gray GM, Bucher JR. 2008. Transforming environ­ mental health protection. Science 319(5865):906–907. gories, comprising different size and variance 2009). Additionally, data from animal models Crump KS, Chen C, Chiu WA, Louis TA, Portier CJ, distributions for multiple variability factors and large-scale in vitro screening platforms Subramaniam RP, et al. 2010. What role for biologically that can then be applied to other chemicals,­ that have incorporated population-­based based dose–response models in estimating low-dose risk? Environ Health Perspect 118:585–588. may emerge from these analyses.­ The ranking genetic determinants (reviewed by Rusyn et al. Csajka C, Verotta D. 2006. Pharmacokinetic–pharmaco­dynamic and grouping of chemi­cals for the applica­ 2010), have provided insight into the extent modelling: history and perspectives. J Pharmacokinet tion of these distributions may be based on of genetic variability in response to a diversity Pharmacodyn 33(3):227–279. Dorne JL. 2010. Metabolism, variability and risk assessment. structural class, the relative extent of observed of toxicants, as well as aided in the identi­ Toxicology 268(3):156–164. variability,­ a common determinant of vari­ fication of genetic susceptibility factors that Dumas ME, Wilder SP, Bihoreau MT, Barton RH, Fearnside JF, ability (e.g., as identified in GWAS analysis of underscore the development of toxic pheno­ Argoud K, et al. 2007. Direct quantitative trait locus mapping cytotoxicity­ phenotypes),­ or other factors (e.g., types. Hypothesis-based molecular clinical of mammalian metabolic phenotypes in diabetic and normo- glycemic rat models. Nat Genet 39(5):666–672. likelihood of coexposures or confounders). and epidemiological approaches to integrating Durbin R, Abecasis G, Altshuler D, Auton A, Brooks L, Gibbs R. Compounds demonstrated or predicted to genetics, molecular pathway data, and clini­ 2010. A map of human genome variation from population- have highly variable toxic responses may also cal observations and biomarkers are likely to scale sequencing. Nature 467(7319):1061–1073. Emilsson V, Tholeifsson G, Zhang B, Leonardson AS, Zink F, be given a higher priority for further study, in contribute to population-based descriptions Zhu J, et al. 2008. Genetics of gene expression and its effect combination with chemical­ and other expected of variability. Complementary to these are on disease. Nature 452(7186):423–428. modifiers of susceptibility. genome-wide (Hutter et al. 2012) and expo­ Gieger C, Geistlinger L, Altmaier E, de Angelis MH, Kronenberg F, Meitinger T, et al. 2008. Genetics meets metabolomics: a At higher tiers of NexGen human health sure-wide (Patel et al. 2010) approaches for genome-wide association study of metabolite profiles in risk assessments, animal and in some cases assessing human population variability in toxic human serum. PLoS Genet 4(11):e1000282; doi:10.1371/jour- human data are available for evaluating dose– response. Opportunities exist to employ these nal.pgen.1000282 [Online 28 November 2008]. Ginsberg G, Angle K, Guyton K, Sonawane B. 2011. Poly­morphism response relationships, major pathways for emerging data streams in the development of in the DNA repair enzyme XRCC1: utility of current database some of the critical toxicities for risk assess­ in silico predictive models for application in a and implications for human health risk assessment. Mutat ment can be reasonably well understood, and range of decision-making contexts. Res 727(1–2):1–15. Ginsberg G, Guyton K, Johns D, Schimek J, Angle K, Sonawane B. some in vivo human data relevant to those 2010. Genetic polymorphism in metabolism and host defense pathways may be available. For some chemi­ Re f e r e n c e s enzymes: implications for human health risk assessment. Crit cals, sensitive populations may have been iden­ Rev Toxicol 40(7):575–619. tified and studied using omics technologies. In Allen BC, Hack CE, Clewell HJ. 2007. Use of Markov Chain Monte Ginsberg G, Hattis D, Sonawane B. 2004. Incorporating pharmaco­ Carlo analysis with a physiologically-based pharmaco­ kinetic differences between children and adults in assessing the case of ozone, for example, gene expression kinetic model of methylmercury to estimate exposures in children’s risks to environmental­ toxicants. Toxicol Appl data and genomic markers may be collected US women of childbearing age. Risk Anal 27(4):947–959. Pharmacol 198(2):164–183. on individuals of high and average sensitivity. Beal SL, Sheiner LB. 1982. Estimating population kinetics. Crit Ginsberg G, Hattis D, Sonawane B, Russ A, Banati P, Kozlak M, Rev Biomed Eng 8(3):195–222. et al. 2002. Evaluation of child/adult pharmacokinetic­ differ- Toxicity pathways exhibited in cultured air­ Bernillon P, Bois FY. 2000. Statistical issues in toxicokinetic ences from a database derived from the therapeutic drug way epithelial cells exposed to ozone may also modeling: a Bayesian perspective. Environ Health Perspect literature. Toxicol Sci 66(2):185–200. be compared with those in humans exposed 108:883–893. Ginsberg G, Neafsey P, Hattis D, Guyton KZ, Johns DO, Bois FY. 2010. Physiologically based modelling and prediction Sonawane B. 2009a. Genetic polymorphism in paraoxonase 1 in vivo to ozone. Such data will aid a better of drug interactions. Basic Clin Pharmacol 106(3):154–161. (PON1): Population distribution of PON1 activity. J Toxicol characterization of the dose–time–­response Bois FY, Compton-Quintana PJ. 1992. Sensitivity analysis of a Environ Health B Crit Rev 12(5):473–507. severity relationships at low doses. In other new model of carcinogenesis. J Theor Biol 159(3):361–375. Ginsberg G, Smolenski S, Hattis D, Guyton KZ, Johns DO, Bois FY, Gelman A, Jiang J, Maszle DR, Zeise L, Alexeef G. Sonawane B. 2009b. Genetic polymorphism in glutathione cases, where individuals are studied epidemio­ 1996. Population toxicokinetics of tetrachloroethylene. transferases (GST): population distribution of GSTM1, T1, logically, the current bioinformatics analyses Arch Toxicol 70(6):347–355. and P1 conjugating activity. J Toxicol Environ Health B Crit lack power and require pooling of subjects to Bois FY, Jamei M, Clewell HJ. 2010. PBPK modelling of inter- Rev 12(5):389–439. individual variability in the pharmacokinetics­ of environ­ Ginsberg G, Smolenski S, Neafsey P, Hattis D, Walker K, detect trends, losing variability estimation in mental chemi­cals. Toxicology 278(3):256–267. Guyton KZ, et al. 2009c. The influence of genetic poly­ the process. In such cases, there will be a need Bois FY, Krowech G, Zeise L. 1995. Modeling human inter­ morphisms on popu­lation variability in six xenobiotic- to couple default descriptions of PD variabil­ individual variability in metabolism and risk: the example of metabolizing enzymes. J Toxicol Environ Health B Crit Rev ity with PBPK modeling to obtain an overall 4-aminobiphenyl. Risk Anal 15(2):205–213. 12(5):307–333. Bois FY, Zeise L, Tozer TN. 1990. Precision and sensitivity Giorgino T, D’Abramo M, Gervasio F, De Fabritiis G. 2010. prediction of variability. In the future, new of pharmaco­kinetic models for cancer risk assessment: Exploring the kinetics of drug binding to the hERG channel

Environmental Health Perspectives • v o l u m e 121 | n u m b e r 1 | January 2013 29 Zeise et al.

through large-scale simulations [Abstract]. In: Proceedings Johanson G, Jonsson F, Bois F. 1999. Development of new PLoS ONE 5(5):e10746; doi:10.1371/journal.pone.0010746 of the 2010 Virtual Physiological Human Conference, technique for risk assessment using physiologically based [Online 20 May 2010]. 30 September—1 October 2010, pp. 248–250. Bruxelles, toxicokinetic­ models. Am J Ind Med 1:101–103. Peters EJ, Motsinger-Reif A, Havener TM, Everitt L, Hardison NE, Belgium. Available: http://www.vph-noe.eu/vph-repository/ Jonsson F, Jonsson EN, Bois FY, Marshall S. 2007. The applica- Watson VG, et al. 2011. Pharmacogenomic characterization doc_download/204-book-of-abstracts-for-vph20107 tion of a Bayesian approach to the analysis of a complex, of US FDA-approved cytotoxic drugs. Pharmacogenomics [accessed 8 June 2012]. mechanistically based model. J Biopharm Stat 17(1):65–92. 12(10):1407–1415. Guo Y, Lu P, Farrell E, Zhang X, Weller P, Monshouwer M, et al. Judson RS, Kavlock RJ, Setzer RW, Hubal EA, Martin MT, Phillips EJ, Mallal SA. 2010. Pharmacogenetics of drug hyper- 2007. In silico and in vitro pharmacogenetic analysis in Knudsen TB, et al. 2011. Estimating toxicity-related biological sensitivity. Pharmacogenomics 11(7):973–987. mice. Proc Natl Acad Sci USA 104(45):17735–17740. pathway altering doses for high-throughput chemical­ risk Portier CJ, Kaplan NL. 1989. Variability of safe dose estimates Guo Y, Weller P, Farrell E, Cheung P, Fitch B, Clark D, et al. 2006. assessment. Chem Res Toxicol 24(4):451–462. when using complicated models of the carcinogenic pro- In silico pharmacogenetics of warfarin metabolism. Nat Kedderis G. 1997. Extrapolation of in vitro enzyme induction data cess. A case study: methylene chloride. Fundam Appl Biotechnol 24(5):531–536. to humans in vivo. Chem Biol Interact 107(1–2):109–121. Toxicol 13(3):533–544. Guyton KZ, Kyle AD, Aubrecht J, Cogliano VJ, Eastmond DA, Khoury MJ, Bowen MS, Burke W, Coates RJ, Dowling NF, Prows DR, Shertzer HG, Daly MJ, Sidman CL, Leikauf GD. 1997. Jackson M, et al. 2009. Improving prediction of chemi­cal Evans JP, et al. 2011. Current priorities for public health prac- Genetic analysis of ozone-induced acute lung injury in sensi- carcinogenicity by considering multiple mechanisms and tice in addressing the role of human genomics in improving tive and resistant strains of mice. Nat Genet 17:471–474. applying toxicogenomic approaches. Mutat Res Rev Mutat population health. Am J Prev Med 40(4):486–493. Rappaport SM, Smith MT. 2010. Epidemiology. Environment and Res 681(2–3):230–240. Kleeberger SR, Levitt RC, Zhang LY, Longphre M, Harkema J, disease risks. Science 330(6003):460–461. Hack CE. 2006. Bayesian analysis of physiologically based Jedlicka A, et al. 1997. Linkage analysis of susceptibility to Reif DM, Martin MT, Tan SW, Houck KA, Judson RS, Richard AM, toxico­kinetic and toxicodynamic models. Toxicology 221(2– ozone-induced lung inflammation in inbred mice. Nat Genet et al. 2010. Endocrine profiling and prioritization of environ­ 3):241–248. 17:475–478. mental chemi­cals using ToxCast data. Environ Health Hamade AK, Misra V, Rabold R, Tankersley CG. 2010. Age-related Koturbash I, Scherhag A, Sorrentino J, Sexton K, Bodnar W, Perspect 118:1714–1720. changes in cardiac and respiratory adaptation to acute ozone Swenberg JA, et al. 2011. Epigenetic mechanisms of mouse Rosenberg NA, Huang L, Jewett EM, Szpiech ZA, Jankovic I, and carbon black exposures: interstrain variation in mice. interstrain variability in genotoxicity of the environ­mental Boehnke M. 2010. Genome-wide association studies in Inhal Toxicol 22(supp 2):84–94. toxicant 1,3-butadiene. Toxicol Sci 122(2):448–456. diverse populations. Nat Rev Genet 11(5):356–366. Harrill AH, Ross PK, Gatti DM, Threadgill DW, Rusyn I. 2009a. La Thangue NB, Kerr DJ. 2011. Predictive biomarkers: a paradigm Rotroff DM, Wetmore BA, Dix DJ, Ferguson SS, Clewell HJ, Population-based discovery of toxicogenomics biomarkers shift towards personalized cancer medicine. Nat Rev Clin Houck KA, et al. 2010. Incorporating human dosimetry and for hepatotoxicity using a laboratory strain diversity panel. Oncol 8(10):587–596. exposure into high-throughput in vitro toxicity screening. Toxicol Sci 110(1):235–243. Lock EF, Abdo N, Huang R, Xia M, Kosyk O, O’Shea SH, et al. Toxicol Sci 117(2):348–358. Harrill AH, Watkins PB, Su S, Ross PK, Harbourt DE, Stylianou IM, 2012. Quantitative high-throughput screening for chemi­cal Roy AC, Loke DF, Saha N, Viegas OA, Tay JS, Ratnam SS. 1994. et al. 2009b. Mouse population-guided resequencing reveals toxicity in a population-based in vitro model. Toxicol Sci Interrelationships of serum paraoxonase, serum lipids and that variants in CD44 contribute to acetaminophen-induced 126(2):578–588. apolipoproteins in normal pregnancy. A longitudinal study. liver injury in humans. Genome Res 19(9):1507–1515. Luke NS, Devito MJ, Shah I, El-Masri HA. 2010a. Development of Gynecol Obstet Invest 38(1):10–13. Hattis D. 1997. Human variability in susceptibility: how big, how a quantitative model of pregnane X receptor (PXR) mediated Rusyn I, Gatti DM, Wiltshire T, Wiltshire T, Kleeberger SR, often, for what responses to what agents? Environ Toxicol xenobiotic metabolizing enzyme induction. Bull Math Biol Threadgill DW. 2010. Toxicogenetics: population-based Pharmacol 4(3–4):195–208. 72(7):1799–1819. testing of drug and chemi­cal safety in mouse models. Hattis D, Baird S, Goble R. 2002. A straw man proposal for a quanti­ Luke NS, Sams R II, DeVito MJ, Conolly RB, El-Masri HA. 2010b. Pharmacogenomics 11(8):1127–1136. tative­ definition of the RfD. Drug Chem Toxicol 25(4):403–436. Development of a quantitative model incorporating key Sadiq SK, Wright DW, Kenway OA, Coveney PV. 2010. Accurate Hattis D, Ginsberg G, Sonawane B, Smolenski S, Russ A, events in a hepatotoxic mode of action to predict tumor ensemble molecular dynamics binding free energy ranking Kozlak M, et al. 2003. Differences in pharmaco­kinetics incidence. Toxicol Sci 115(1):253–266. of multidrug-resistant HIV-1 proteases. J Chem Inf Model between children and adults: II. Childrens variability in drug Manolio TA. 2010. Genomewide association studies and assess- 50(5):890–905. elimination half-lives and in some parameters­ needed for ment of the risk of disease. N Engl J Med 363(2):166–176. Schadt EE. 2009. Molecular networks as sensors and drivers of physiologically-based pharmaco­kinetic modeling. Risk Anal Manolio TA, Collins FS, Cox NJ, Goldstein DB, Hindorff LA, common human diseases. Nature 461(7261):218–223. 23(1):117–142. Hunter DJ, et al. 2009. Finding the missing heritability of Schadt EE, Björkegren JL. 2012. NEW: network-enabled wis- Hattis D, Russ A, Goble R, Banati P, Chu M. 2001. Human inter­ complex diseases. Nature 461(7265):747–753. dom in biology, medicine, and health care. Sci Transl Med individual variability in susceptibility to airborne particles. Martin MT, Dix DJ, Judson RS, Kavlock RJ, Reif DM, Richard AM, 4(115):115rv111; doi:10.1126/scitranslmed.3002132 [Online Risk Anal 21(4):585–599. et al. 2010. Impact of environmental­ chemicals­ on key 4 January 2012]. Hernandez ML, Lay JC, Harris B, Esther CR Jr, Brickey WJ, transcription regulators and correlation to toxicity end Schuster P. 2008. Modeling in biological chemistry. From Bromberg PA, et al. 2010. Atopic asthmatic subjects but not points within EPA’s ToxCast program. Chem Res Toxicol biochemical­ kinetics to systems biology. Monatshefte fur atopic subjects without asthma have enhanced inflammatory 23(3):578–590. Chemie / Chemical Monthly 139(4):427–446. response to ozone. J Allergy Clin Immunol 126(3):537–544.e1; McLanahan ED, Andersen ME, Fisher JW. 2008. A biologically Spear RC, Bois FY. 1994. Parameter variability and the interpreta- doi:10.1016/j.jaci.2010.06.043. based dose-response model for dietary iodide and the tion of physiologically based pharmaco­kinetic modeling Holloway JW, Savarimuthu Francis S, Fong KM, Yang IA. 2012. hypothalamic-pituitary-thyroid axis in the adult rat: evalua- results. Environ Health Perspect 11:61–66. Genomics and the respiratory effects of air pollution expo- tion of iodide deficiency. Toxicol Sci 102(2):241–253. Suhre K, Shin SY, Petersen AK, Mohney RP, Meredith D, sure. Respirology 17(4):590–600. Mortensen HM, Euling SY. 2011. Integrating mechanistic and poly- Wägele B, et al. 2011. Human metabolic individuality Huang RS, Johnatty SE, Gamazon ER, Im HK, Ziliak D, Duan S, et al. morphism data to characterize human genetic susceptibility in biomedical and pharmaceutical research. Nature 2011. Platinum sensitivity-related germline poly­morphism dis- for environmental­ chemical­ risk assessment in the 21st cen- 477(7362):54–60. covered via a cell-based approach and analysis­ of its asso- tury. Toxicol Appl Pharmacol; doi:10.1016/j.taap.2011.01.015 Thompson CM, Johns DO, Sonawane B, Barton HA, Hattis D, ciation with outcome in ovarian cancer patients. Clin Cancer [Online 1 February 2011]. Tardif R, et al. 2009. Database for physiologically based Res 17(16):5490–5500. Neafsey P, Ginsberg G, Hattis D, Johns DO, Guyton KZ, pharmaco­kinetic (PBPK) modeling: physiological data for Hucka M, Finney A, Sauro HM, Bolouri H, Doyle JC, Kitano H, Sonawane B. 2009a. Genetic polymorphism in CYP2E1: healthy and health-impaired elderly. J Toxicol Environ Health et al. 2003. The systems biology markup language (SBML): Population distribution of CYP2E1 activity. J Toxicol Environ B Crit Rev 12(1):1–24. a medium for representation and exchange of biochemi­cal Health B Crit Rev 12(5):362–388. U.S. EPA (U.S. Environmental Protection Agency). 1998. network models. Bioinformatics 19(4):524–531. Neafsey P, Ginsberg G, Hattis D, Sonawane B. 2009b. Genetic Toxicological Review of Beryllium and Compounds (CAS Hutter CM, Chang-Claude J, Slattery ML, Pflugeisen BM, Lin Y, polymorphism in cytochrome P450 2D6 (CYP2D6): Population No.7440-41-7) in Support of Summary Information on the Duggan D, et al. 2012. Characterization of gene–environment distribution of CYP2D6 activity. J Toxicol Environ Health B Integrated Risk Information System (IRIS). EPA/635/R-98/008. interactions for colorectal cancer susceptibility loci. Cancer Crit Rev 12(5):334–361. Washington, DC:U.S. EPA. Available: http://www.epa.gov/ Res 72(8):2036–2044. Nong A, Tan YM, Krolski ME, Wang J, Lunchick C, Conolly RB, iris/toxreviews/0012tr.pdf [accessed 8 June 2012]. Illig T, Gieger C, Zhai G, Römisch-Margl W, Wang-Sattler R, et al. 2008. Bayesian calibration of a physiologically based U.S. EPA (U.S. Environmental Protection Agency). 2002. Prehn C, et al. 2010. A genome-wide perspective of genetic pharmacokinetic/pharmaco­ dynamic­ model of carbaryl Health Assessment of 1,3-Butadiene. EPA/600/P- variation in human metabolism. Nat Genet 42(2):137–141. cholinesterase inhibition. J Toxicol Environ Health A 98/001F. Washington, DC:U.S. EPA, Office of Research International Programme for Chemical Safety. 2001. Guidance 71(20):1363–1381. and Development, National Center for Environmental Document for the Use of Data in Development of Chemical- NRC (National Research Council). 2007. Toxicity Testing in Assessment. Available: http://cfpub.epa.gov/ncea/cfm/ Specific Adjustment Factors (CSAFs) for Interspecies the 21st Century: A Vision and a Strategy. Washington, recordisplay.cfm?deid=54499 [accessed 8 June 2012]. Differences and Human Variability in Dose/Concentration– DC:National Academies Press. Available: http://www.nap. U.S. EPA (U.S. Environmental Protection Agency). 2003. EPA Response Assessment. WHO/PCS/01.4. Available: http:// edu/catalog.php?record_id=11970 [accessed 8 June 2012]. Assessment of Risks from Radon in Homes. EPA/402/R- www.who.int/ipcs/publications/methods/harmonization/en/ NRC (National Research Council). 2009. Science and Decisions: 03/003. Washington, DC:U.S. EPA. Available: http://www. csafs_guidance_doc.pdf [accessed 8 June 2012]. Advancing Risk Assessment. Washington, DC:National epa.gov/radiation/docs/assessment/402-r-03-003.pdf Jamei M, Marciniak S, Feng K, Barnett A, Tucker G, Rostami- Academies Press. Available: http://www.nap.edu/catalog. [accessed 8 June 2012]. Hodjegan A. 2009. The Simcyp population-based ADME php?record_id=12209 [accessed 8 June 2012]. U.S. EPA (U.S. Environmental Protection Agency). 2006. Air simulator. Expert Opin Drug Metab Toxicol 5(2):211–223. O’Shea SH, Schwarz J, Kosyk O, Ross PK, Ha MJ, Wright FA, Quality Criteria for Ozone and Related Photochemi­cal Jia X, Song X, Shima M, Tamura K, Deng F, Guo X. 2011. et al. 2011. In vitro screening for population variability in Oxidants. EPA/600/R-05/004aF-cF. Washington, DC:U.S. Acute effect of ambient ozone on heart rate variability chemi­cal toxicity. Toxicol Sci 119(2):398–407. EPA. Available: http://cfpub.epa.gov/ncea/isa/recordisplay. in healthy elderly subjects. J Expo Sci Environ Epidemiol Patel CJ, Bhattacharya J, Butte AJ. 2010. An Environment-Wide cfm?deid=149923 [accessed 15 June 2012]. 21(5):541–547. Association Study (EWAS) on type 2 diabetes mellitus. U.S. EPA (U.S. Environmental Protection Agency). 2007.

30 v o l u m e 121 | n u m b e r 1 | January 2013 • Environmental Health Perspectives Addressing human variability in NexGen assessments

Framework for Assessing the Public Health Impacts of Risk using CEPH pedigrees. PloS one 6(5):e17561; doi:10.1371/ Woodruff TJ, Zeise L, Axelrad DA, Guyton KZ, Janssen S, Management Decisions. Washington, DC:U.S. EPA, Office of journal.pone.0017561 [Online 5 May 2011]. Miller M, et al. 2008. Meeting report: Moving upstream— Research and Development. Welsh M, Mangravite L, Medina MW, Tantisira K, Zhang W, evaluating adverse upstream end points for improved risk U.S. EPA (U.S. Environmental Protection Agency). 2011. Huang RS, et al. 2009. Pharmacogenomic discovery using assessment and decision-making. Environ Health Perspect Toxicological Review of Trichloroethylene (CASRN 79-01-6) cell-based models. Pharmacol Rev 61(4):413–429. 116:1568–1575. in Support of Summary Information on the Integrated Risk Wild CP. 2005. Complementing the genome with an “exposome:” Xia M, Huang R, Witt KL, Southall N, Fostel J, Cho MH, et al. 2008. Information System (IRIS). EPA/635/R-09/011F. Washington, the outstanding challenge of environmental­ exposure mea- Compound cytotoxicity profiling using quantitative high- DC:U.S. EPA. Available: http://www.epa.gov/iris/ surement in molecular epidemiology. Cancer Epidemiol throughput screening. Environ Health Perspect 116:284–291. supdocs/0199index.html [accessed 8 June 2012]. Biomarkers Prev 14(8):1847–1850. Xia MH, Huang RL, Guo V, Southall N, Cho MH, Inglese J, et al. Wakefield J. 1996. Bayesian individualization via sampling- Willmann S, Höhn K, Edginton A, Sevestre M, Solodenko J, 2009. Identification of compounds that potentiate CREB based methods. J Pharmacokinet Biopharm 24(1):103–131. Weiss W, et al. 2007. Development of a physiology-based signaling as possible enhancers of long-term memory. Proc Walker K, Ginsberg G, Hattis D, Johns D, Guyton KZ, Sonawane B. whole-body population model for assessing the influence Natl Acad Sci USA 106(7):2412–2417. 2009. Genetic polymorphism in N-acetyltransferase (NAT): of individual variability on the pharmaco­kinetics of drugs. Zhu H, Rusyn I, Richard A, Tropsha A. 2008. Use of cell viability population distribution of NAT1 and NAT2 activity. J Toxicol J Pharmacokinet Pharmacodyn 34(3):401–431. assay data improves the prediction accuracy of conventional Environ Health B Crit Rev 12(5–6):440–472. Wood AM, Harrison RM, Semple S, Ayres JG, Stockley RA. 2010. quantitative structure–activity relationship models of animal Watson VG, Motsinger-Reif A, Hardison NE, Peters EJ, Outdoor air pollution is associated with rapid decline of lung carcinogenicity. Environ Health Perspect 116:506–513. Havener TM, Everitt L, et al. 2011. Identification and replica- function in a-1-antitrypsin deficiency. Occup Environ Med tion of loci involved in camptothecin-induced cytotoxicity 67(8):556–561.

Environmental Health Perspectives • v o l u m e 121 | n u m b e r 1 | January 2013 31