Stochastic Simulation of Hepatic Preneoplastic Foci Development forFour Chlorobenzene Congeners in a Medium-Term Bioassay

Ying C. Ou,* Rory B. Conolly,† Russell S. Thomas,‡ Daniel L. Gustafson,§ Michael E. Long,§ Ivan D. Dobrev,¶

Laura S. Chubb,¶ Yihua Xu,|| Smadar A. Lapidot,¶ Melvin E. Andersen,† and Raymond S. H. Yang¶ ,1

*Preclinical Development, Human Genome Sciences, Inc., Rockville, Maryland 20850; †CIIT Centers for Health Research, Research Triangle Park, NorthCarolina 27709; ‡Kalypsos, Inc., La Jolla, California 92037; §School of Pharmacy, University of Colorado Health Science Center, Denver, Colorado80262; ¶Quantitative and Computational Toxicology Group, Center for Environmental Toxicology and Technology, Department of Environmental and

Radiological Health Sciences, Colorado State University, Fort Collins, Colorado 80523; and ||Department of Oncology, McArdle Laboratory for CancerResearch, Madison, Wisconsin 53706

Received November 20, 2002; accepted March 3, 2003

A combination of experimental and simulation approaches wasused to analyze clonal growth of glutathione-S-transferase �(GST-P) enzyme-altered foci during liver carcinogenesis in aninitiation-promotion regimen for 1,4-dichlorobenzene (DCB),1,2,4,5-tetrachlorobenzene (TECB), pentachlorobenzene (PECB),and hexachlorobenzene (HCB). Male Fisher 344 rats, eight weeksof age, were initiated with a single dose (200 mg/kg, ip) of dieth-ylnitrosamine (DEN). Two weeks later, daily dosing of 0.1 mol/kgchlorobenzene was maintained for six weeks. Partial hepatectomywas performed three weeks after initiation. Liver weight, normalhepatocyte division rates, and the number and volume of GST-Ppositive foci were obtained at 23, 26, 28, 47, and 56 days afterinitiation. A clonal growth stochastic model separating the initi-ated cell population into two distinct subtypes (referred to as Aand B cells) was successfully used to describe the foci developmentdata for the four chlorobenzenes. The B cells are initiated cells thatdisplay a selective growth advantage under conditions that inhibitthe growth of initiated A cells or normal hepatocytes. The simu-lation exercise for the four chlorobenzenes indicates a positivecorrelation between the estimated net growth rate of B cells duringthe 2-week regeneration period following partial hepatectomy andfinal foci volume at the end of the bioassay. This observation isconsistent with the sensitivity analysis of model parameters. WhileTECB, PECB, and HCB all significantly increased foci volume,only HCB increased normal hepatocyte proliferation. Together,these results indicate that examining effects of chemicals on re-generative responses following partial hepatectomy may be ameans for understanding the carcinogenicity potential of chloro-benzene compounds.

Key Words: preneoplastic foci; simulation; liver carcinogenesis;clonal growth model; chlorobenzene.

Chlorobenzenes are significant environmental contaminants

manufactured worldwide for both industrial and domestic uses(Morita, 1977; Tobin, 1986). The chlorobenzene family iscomprised of 12 congeners differing by the number and posi-tion of chlorine substitution on a single benzene ring. Usage ofchlorobenzene compounds ranges from room deodorizer andmoth repellent to chemical intermediates in commercial pro-duction of pesticide (Hill et al., 1995; Morita, 1977) andsignificant human exposure to chlorobenzenes is documented(Carey et al., 1986; Peattie et al., 1984). The potential carci-nogenicity of various chlorobenzene congeners in rodents hasbeen studied. Results from a two-year bioassay performed bythe National Toxicology Program (NTP) have shown that1,4-dichlorobenzene (DCB) is a carcinogen (NTP, 1987);monochlorobenzene shows equivocal results (NTP, 1985b),and 1,2-dichlorobenzene is negative (NTP, 1985a). The carci-nogenicity of hexachlorobenzene (HCB) has been demon-strated in rats and mice using the two-year chronic animalbioassay (Cabral and Shubik, 1986; Cabral et al., 1996). Stud-ies using an initiation-promotion hepatocarcinogenesis proto-col also suggest that HCB and pentachlorobenzene (PECB) areliver tumor promoters (Cabral et al., 1996; Stewart et al., 1989;Thomas et al., 1998b). However, carcinogenicity of otherchlorobenzene congeners has not been systematically evalu-ated.

We have used the medium-term bioassay of Ito et al.(1989a,b) to study the carcinogenicity of chlorobenzenes(Gustafson et al., 1998, 2000). The Ito assay involves thesequential administration of a potent initiator, diethylnitro-samine (DEN), followed by chemical treatment and mitogenicstimulation of hepatocyte growth via partial hepatectomy. Thisprotocol allows the evaluation of carcinogenic potential withineight weeks by identification of glutathione-S-transferase �(GST-P) positive preneoplastic foci as end point marker le-sions. A large number of chemicals have been tested using thisprotocol. When compared with the two-year chronic bioassay,results from the Ito medium-term bioassay have correctly iden-

1 To whom all correspondence should be addressed at the Department ofEnvironmental and Radiological Health Sciences, Colorado State University,CETT, Foothills Campus, Fort Collins, CO 80523. Fax: (970) 491-8304. E-mail:raymond.yang@colostate.edu.

TOXICOLOGICAL SCIENCES 73, 301–314 (2003)DOI: 10.1093/toxsci/kfg078Copyright © 2003 by the Society of Toxicology

301

by guest on May 25, 2014

http://toxsci.oxfordjournals.org/D

ownloaded from

tified 97% of genotoxic hepatocarcinogens and 86% of theknown nongenotoxic hepatococarcinogens (Ogiso et al., 1990).

Knowledge on the sequential alteration in growth controland cell dynamics of foci may contribute to an understandingof chemical carcinogenesis. Furthermore, to facilitate interpre-tation of the Ito assay results from our study (Gustafson et al.,1998, 2000), we measured time-course development of foci,cell proliferation rates, and chemical tissue concentrations ofdosed chemicals during the Ito assay. Molecular pathwayspotentially involved in the development of GST-P foci werealso studied. We showed that both 1,2,4,5,-tetrachlorobenzene(TECB) and PECB treatments resulted in an increased ratio ofreduced to oxidized glutathione (GSH:GSSG; Thomas et al.,1998a), and for PECB a low incidence of GST-P foci in thecentrilobular region was accompanied by increased glutathionereductase and �-glutamylcysteine synthetase (Thomas et al.,1998a). PECB and HCB (Thomas et al., 1998b), but not TECBand DCB (Carlson, 1977; Chu et al., 1983), also appear tostimulate production of porphyrin; in addition, TECB, PECB,and HCB increase protein expression of c-fos, c-jun, CYP 2E1,CYP 2B1/2, and CYP 1A1 (Gustafson et al., 2000).

The two-stage Moolgavkar, Venzon, and Knudson (MVK)model (Moolgavkar and Luebeck, 1990; Moolgavkar and Ven-zon, 2000) has been proposed as an improvement over severalexisting models for estimating carcinogenic risks to humanhealth, because it incorporates more biological considerationsthan previous models, notably cell population kinetics. Severalquantitative and statistical methods based on the MVK two-stage framework have been used to analyze foci developmentdata (Conolly and Andersen, 1997; Luebeck et al., 1991;Portier et al., 1996). We adopted a discrete-time numericalapproach (Ou et al., 2001; Thomas et al., 2000) of the clonalgrowth model (Conolly and Kimbell, 1994), similar to thatimplemented by Cohen et al. (Cohen and Ellwein, 1990; Ell-wein and Cohen, 1992), which does not require the analyticalsolution implemented in the original MVK model. In thiscurrent model, time axis is decomposed into a series of timeintervals, where parameters are allowed to change between butnot within segments. This approach differs from the MVKmodel, where differential equations are used with variables ofthe model changed continuously over time. The simulationmodel described here may be a useful tool for the analysis ofthe complex data surrounding the development of focal lesions.In particular, this simulation model allows examination ofdynamic changes in foci development under different chemi-cal/biological perturbations.

A simulation model incorporating a two-cell hypothesis wassuccessfully used to analyze 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD) dose-response foci data (Conolly andAndersen, 1997) and also for PECB and HCB time course focidata in the Ito medium-term bioassay (Ou et al., 2001; Thomaset al., 2000). Inclusion of two foci populations in the frame-work of the two-stage model (Fig. 1) was based on the negativeselection hypothesis of tumor promotion (Andersen et al.,

1995; Jirtle et al., 1994). The initiated cells were partitioned inthe simulation model into A and B cell populations, where Bcells represent the population of cells with a growth advantagein an otherwise mitoinhibitory or cytotoxic environment toinitiated A or normal uninitiated cells. Exposure to chemicalslike phenobarbital can result in increased production of growthfactors, such as transforming growth factor (TGF)-�1 to con-strain proliferation (mitoinhibition; Jirtle et al., 1994). Thisselective environment may lead to outgrowth of clones that areresistant to mitoinhibition (Jirtle et al., 1994).

The goal of the current study is threefold. First, we deter-mined whether the previous clonal growth simulation model(Ou et al., 2001) for PECB or HCB could also be employed toanalyze foci growth kinetics following exposure to two otherchlorobenzene congeners, TECB and DCB. Second, we per-formed a sensitivity analysis on the clonal growth model togain a better understanding of model behavior. Finally, simu-lation modeling for four chlorobenzene congeners provided anopportunity for comparing model parameter values and effectsof four chlorobenzene congeners on molecular endpoints in-volved in the development of GST-P foci. To our knowledge,this is the first demonstration of combined experimental andsimulation analysis on the time-course foci development forfour structurally related compounds.

MATERIALS AND METHODS

The method used in the present study consists of two main parts: experi-mentation and stochastic simulation modeling. Experimental data collectionincluded time course measurements of liver growth, hepatocyte proliferation,and foci development. The following sections describe how we collected theexperimental data for TECB and DCB. Similar data collections for PECB and

FIG. 1. A simple two-stage model of carcinogenesis. Within the frame-work of the two-stage model, the possibility of two initiated populations foradequately describing the kinetics of foci growth was evaluated. The celldivision and death rates of normal hepatocytes are denoted by � and � (1/day),respectively. The probability of mutation per cell division to A or B initiatedcells per day are denoted by �a and �b. The division and death rates of A, Bcells are denoted by �a, �a, �b, �b, respectively.

302 OU ET AL.

by guest on May 25, 2014

http://toxsci.oxfordjournals.org/D

ownloaded from

HCB have been reported earlier (Ou et al., 2001). Together, these data wereused for comparative simulation modeling of all four congeners.

Experimental Data Collection

Chemicals. Chlorobenzenes were purchased from Aldrich Chemical (Mil-waukee, WI). DEN was purchased from Sigma Chemical (St. Louis, MO).

Animals and treatment. Male F344 rats, 30 days of age, from HarlanSprague-Dawley (Indianapolis, IN) were acclimated for four weeks before thestart of the experimentation. The rats were randomized by weight and dividedinto three treatment groups (Fig. 2). At week 0, animals received a single ipinjection of DEN (200 mg/kg) dissolved in 0.9% saline. After two weeks, therats received daily gavage administration of corn oil or 0.1 mmol/kg TECB orDCB in a corn oil vehicle through the remainder of the eight-week study. Atweek 3 (Day 21), a partial hepatectomy was performed on all animals. Animalswere given food (Harlan Teklad NIH-07 Diet, Madison, WI) and water adlibitum and lighting was set on a 12-h light/dark cycle. On days 23, 26, 28, 47,and 56, at least five animals from each treatment group were sacrificed byaortic exsanguination (Fig. 2). Whole livers were removed, tissues were fixedin either 10% neutral-buffered formalin or ice-cold acetone, and embedded inparaffin. Several sections from each liver lobe were taken and then serialsectioning was done at a thickness of 5 �m each. One serial sectioning wasused for the BrdU analysis, the other for GST-P foci identification. The studieswere conducted in accordance with the National Institutes of Health (NIH)guidelines for the care and use of laboratory animals. Animals were housed ina fully accredited American Association for Accreditation of LaboratoryAnimal Care (AAALAC) facility.

Quantification of GST-P positive foci. Acetone fixed tissues were used forthe immunohistochemical identification of GST-P foci (Thomas et al., 1998a).Liver sections were deparaffinized in xylene and rehydrated by passage

through an alcohol series. Endogenous peroxidase was quenched in 3% hy-drogen peroxide for 10 min. The slides were rinsed with deionized water andplaced in phosphate-buffered saline (PBS; pH 7.4, 2.7 mM KCl, 0.14 M NaCl,1.5 mM KH2PO4, 8.1 mM Na2PO4). A standard avidin/biotin (ABC) protocol(Vector Labs, Burlingame, CA) was followed, and foci were detected withGST-P primary antibody (Binding Site, San Diego, CA). GST-P foci weremeasured using a Leitz light microscope coupled with the BioQuant imageanalysis system (Version 4; R & M Biometrics, Nashville, TN). Foci contain-ing more than two cells were recorded. One serial sectioning from each of thefour lobes was analyzed. The number and area of foci recorded beforeconversion to three-dimensional representations is the sum total from all fourlobes.

Stereology methods. Quantitative stereology data for each rat were ob-tained by using STEREO, a Windows 98/NT program developed at theMcArdle Laboratory for Cancer Research, University of Wisconsin (Xu et al.,1998). STEREO uses a data file containing tissue information (tissue areas andindividual focal transections) as the input to provide quantitative stereologyresults on a three-dimensional basis. Numbers of foci per cubic cm of liverwere calculated according to the method of Saltykov (2000). Calculations wereperformed at the truncation value of 50.12 microns. Preneoplastic foci wereclassified according to Saltykov size classes 1 to 11 with a maximal diameterof 63, 79, 100, 125, 154, 199, 251, 316, 398, 501, and 630 microns in eachclass, respectively. Data files from the same group were combined togetherfirst on a two-dimensional level and then three-dimensional data (number ofthe foci in each size class) were calculated according to the method ofSaltykov. The volume fraction of the liver occupied by GST-P foci wascomputed by the method of Delesse (1848).

Determination of cell division rate. Osmotic minipumps (Alzet model2ML1, 10 ml/h; Alza Corporation, Palo Alto, CA), filled with 5-bromo-2�-deoxyuridine (BrdU; 20 mg/ml), were implanted subcutaneously over thedorsal midscapular region. Animals were anesthetized with isofluorane(Anaquest, Madison, WI) and the incision closed using stainless steel woundclips. To avoid saturation of labeling, pumps were implanted one day prior totissue collection on time points soon after partial hepatectomy (days 23, 26,28). For other time point collection, pumps were implanted three days prior tothe sacrifice. Control animals not hepatectomized were given BrdU for 3 days.Detection of BrdU-labeled cells was performed on formalin-fixed liver sec-tions using standard avidin/biotin (ABC) immunoperoxidase kits (Vector Labs,Burlingame, CA) with primary BrdU antibody (Biogenex Labs, San Ramon,CA) and 3-amino-9-ethylcarbozole (AEC, Biomeda, Foster City, CA). At least1000 cells per animal and four animals per group were counted. The labelingindex (LI) was calculated as the number of cells labeled divided by the totalnumber of cells counted. The cell division rate (�, /day) was calculated fromthe LI using Equation 1 below (Moolgavkar and Luebeck, 1992) where t is thenumber of days of exposure to BrdU.

� �1

2t� ln� 1

1 � LI� (1)

Stochastic Clonal Growth Modeling

Description of the clonal growth model. The simulation model used forthe current analysis was based on the clonal growth model as previouslydescribed (Conolly and Andersen, 1997; Conolly and Kimbell, 1994; Ou et al.,2001). A summary of the basic modeling framework is provided here.

1. Proliferation of normal hepatocytes. Proliferation of normal hepatocytesis described as a function of cell division (�; 1/day) and death rate (�; 1/day)deterministically by,

dN

dt� N�� � �� (2)

where N is the number of normal hepatocytes per cm3, � represents all modes

FIG. 2. Experimental design for the initiation/promotion study. The ini-tiation agent, DEN, was administered ip (200 mg/kg) on Day 0. TECB or DCBwas delivered by gavage starting on day 14 at a dose of 0.1 mmol/kg per day,seven days/week. A two-thirds partial hepatectomy was performed on allanimals on day 21. Liver tissues were collected 23, 26, 28, 47, 56 days (days23, 26, 28, 47, 56) following initial DEN dosing.

303STOCHASTIC SIMULATION OF FOCI DEVELOPMENT

by guest on May 25, 2014

http://toxsci.oxfordjournals.org/D

ownloaded from

of cell death including apoptosis and necrosis, and t is simulation time (days)beginning with DEN treatment on day 0.

2. Mutation to initiated cells. The expected number of initiated cells gen-erated is modeled stochastically as a Poisson process and is linked to thenumber of hepatocyte cell divisions during each time step �t by Equation 3.The model assumes that cells act independently and have an equal chance ofmutation in every time interval. Since mutations are rare events drawing fromroughly an infinite pool, the number of mutations will be approximatelyPoisson distributed.

Nm � N���t (3)

Nm represents the number of normal hepatocytes mutating during the timeinterval �t, where t is simulation time (days) beginning with DEN treatment onday 0, � is the probability of mutation per cell division. A random deviateabout Nm denoting the number of mutations during �t is drawn from a Poissondistribution using the function PODEV (Bratley et al., 2000). Inputs toPODEV are the mean of the Poisson distribution and a pseudorandom numberbetween 0 and 1 generated with the algorithm UNFL (Bratley et al., 2000). Thecurrent work describes the first stage of the two-stage model (from normal toinitiated cells) for the GST foci data (Fig. 1), using the two-cell modelhypothesis as described previously (Ou et al., 2001; Thomas et al., 2000). Inthis case, the probability of mutation per cell division to A or B initiated cellsare denoted by �a, �b. The division and death rates of A and B cells are denotedby �a, �a, �b, and �b, respectively (Fig. 1).

3. Proliferation of initiated cells. Division of single mutated cells deriveddirectly from normal cells may give rise to clones of initiated cells (preneo-plastic foci). The growth of initiated cells is described by a multinomialdistribution, a generalization of the binomial distribution for two category(binary) outcomes. We assumed that the three possible events of initiated cellsduring each time step-cell division, cell death, or no change, are independentof each other. Thus, the probability distribution of the sums in each event isgiven by the multinomial distribution. The program keeps track of each cloneover time allowing description of average clone size (cells/clone) and number(clones/cm3). The simulation results can then be compared with three-dimen-sional foci volume (the percent volume of liver occupied by the foci) and focinumber data converted from two-dimensional focal transection data using theSTEREO program as described above (Xu et al., 1998). For comparing thesimulation output with the foci data collected, only clones larger than two cellsin size were considered detectable.

Throughout the simulation, the model keeps track of the total hepatocytenumber in the liver. Simulated liver weight is calculated on the basis of totalhepatocyte number divided by the corresponding hepatocyte density (numberof hepatocytes/unit volume). Partial hepatectomy is described as an instanta-neous decrease in liver weight and cell number.

Modeling strategies. Model parameters were obtained in a stepwise man-ner as described in our previous report (Ou et al., 2001). Briefly, the samepiecewise constant implementation was used in six time intervals (Days 1–7,7–14, 14–21, 21–28, 28–35, 35–120). Division and death rates of normalhepatocytes at the various time points were obtained based on the study ofKato et al. (1993) and the present study. The probability of mutation of normalhepatocytes and the growth characteristics (division and death rates) of initi-ated cells are not known, and have to be estimated based on fitting of the modelto experimental data. Parameterization of mutation rates for the different timeintervals was based on the time course of DEN-induced DNA adduct levels(Dragan et al., 1994) and consistency of modeling outputs to the observed totalfoci number. Thus, the highest number of detectable foci in any of thetreatment groups (DEN, DEN � TECB, and DEN � DCB groups) in anygiven time point was used as an approximation of the number of mutated cellsavailable for subsequent expansion of initiated cells. We have previouslyshown that by visual inspection, model parameters assuming one population ofinitiated cells with homogeneous growth characteristics could not be used to fitthe time-course foci development of DEN controls with partial hepatectomy(Ou et al., 2001). Time-course foci data could only be adequately described by

partitioning GST-P foci into two populations of focal lesions (the two-cellhypothesis). This observation was further supported by the published reportshowing that DEN treatment alone creates heterogeneity of initiated cells, andresistant cells (referred to as B cells here) can account for 5–23% of totalGST-P foci (Yusuf et al., 1999). The two-cell hypothesis (i.e., A and B cells)assuming two populations of focal lesions was therefore used as a basis for thepresent modeling exercise. The percentage of resistance clones (B cells)following DEN initiation is in a range of 5–23% (Yusuf et al., 1999), thus themutation rate to initiated A cells was confined to at least three times higherthan that to initiated B cells. Division and death rates of initiated cells wereestimated based on consistency with several published data sets on the time-course appearance of GST-P foci following DEN treatment (Jang et al., 1993;Kato et al., 1993; Satoh et al., 1989; Tiwawech et al., 1991) and time coursechanges of foci number and volume obtained in the present study. In addition,the cell division rate of initiated hepatocytes are parameterized to conform tothe experimental observation showing that after a sufficiently large dose ofDEN, the cell division rate of initiated hepatocytes declines over time (Traviset al., 1991), whereas the death rate of initiated cells increases with time (Rotsteinet al., 1986). The foci growth parameters were obtained by iterative optimizationto the time course changes of foci number and volume simultaneously using thetwo-cell hypothesis. Both cell division and death rates of A and B initiated cellswere parameterized to ensure that these values were within biologically plausibleintervals for putative preneoplastic cells (Rotstein et al., 1984; Travis et al., 1991).

Parameter estimation. Parameter estimation was conducted by iterative fit-ting of time-course foci number and volume data simultaneously. The parametervalues were identified in a two-step procedure. First, an “initial value” wasobtained by visual observation of the model fit to the data. During this iterativeprocess, an understanding of the impact of parameter changes on the modeloutputs was gained. Next, multiple neighboring values around the “initial values”were picked and were subjected to comparison of their weighted sum of squarevalues calculated from the mean of 100 runs for both the foci volume and foci numberdata. The optimized parameters were identified with the sum of least squares.

Modeling analysis of DEN � TECB and DEN � DCB data. The growthparameters defined in the DEN group were used to evaluate experiments involvingthe administration of TECB and DCB. Thus, we were interested in identifying thenecessary parameter changes from those of DEN controls in order to describetime-course foci development of the DEN � TECB and DEN � DCB group.Modeling exercises were performed assuming that TECB and DCB act as muta-gens, as agents affecting cell proliferation, or as both. Because we did not measurecell division and death rates in the GST-P foci, effects of TECB or DCB on focigrowth could be assumed as either increasing cell division rates, decreasing celldeath rates, or a combination of both. Parameter estimation based on fitting themodel to the data could not distinguish whether TECB or DCB affect the celldivision or death pathway of initiated cells. Thus, the current model only permittedaccurate estimation on how TECB or DCB treatments affect the net growth rate(cell division rate minus death rate) of A and B cells (i.e., an identifiabilitydilemma). For convenience, we present the data with TECB or DCB affecting onlythe cell death rate of initiated cells. Parameter estimation was conducted byiterative fitting of time-course foci number and volume data in the DEN � DCBor DEN � TECB groups. The parameter values were again identified in a two-stepprocedure as described above.

Sensitivity analysis. Sensitivity analysis determines the changes in a mea-surement, m, given small changes in a given input parameter, p, i.e., the partialderivative of the measurement with respect to the parameter, m/p. In the currentanalysis, we used the slope around the parameter for an approximation of thesensitivity coefficient, m/p � �m/�p (Beck and Arnold, 1977). Furthermore, wecalculated the normalized sensitivity coefficient (NSC) with the following formula.

NSC �

�m

m

�p

p

(4)

The larger the sensitivity coefficient, the larger the impact on the measure-

304 OU ET AL.

by guest on May 25, 2014

http://toxsci.oxfordjournals.org/D

ownloaded from

ment given a change in the parameter. The current analysis was to identify thesensitive parameters for the final total foci volume at the end of the eight-weekmedium-term bioassay. A 2.5, 5, 10, 20, and 30% change in each modelparameters was tested for 160, 80, 20, 10, or 5 runs, respectively. A smallernumber of runs (e.g., 5 and 10 runs) were also tested in this study, although ingeneral smaller changes in model parameters (e.g., 2.5 or 5% changes) re-quired more runs to observe consistent results due to the stochastic nature ofsimulation.

Software and hardware. The simulation model was written in AdvancedContinuous Simulation Language (ACSL; The Aegis Technology Group, Inc.,Huntsville, AL), and run on a 300 mHz Intel Pentium II (Gateway 2000, SiouxCity, SD). The least square optimization method and sensitivity analysis wereimplemented in the ACSL Math (ACSL; The AEgis Technology Group, Inc.,Huntsville, AL) environment. A listing of the simulation program and sensi-tivity analysis/optimization routine is available from one of the authors(rconolly@ciit.org). STEREO, a software application for converting two-dimensional tissue transection data to quantitative stereology results on athree-dimensional basis is available from Dr. Xu (xu@oncology.wisc.edu).

Statistical analysis. Time-dependent and treatment-dependent changes inliver weight, cell proliferation, foci volume, foci number, and hepatocytedensity were analyzed using a mixed-effects ANOVA model, followed bypost-hoc Tukey-Kramer multiple comparison tests at p � 0.05. The analysiswas programmed using the SAS system (SAS Institute Inc., Cary, NC).

RESULTS

Experimental Data Collection

Body weight and general histopathology. The effects ofTECB or DCB on final body and liver weight are summarizedin Table 1. Following DEN initiation, exposure to TECB, at 0.1mmol/kg per day for six weeks led to a significant increase inliver weight (expressed as % total body weight) at the end ofthe study compared to the DEN controls (p � 0.05). Exposureto DCB at the same dose and duration did not cause a signif-icant increase in liver weight. Treatment with DCB or TECBalone without DEN initiation resulted in similar liver weightchanges as observed for DEN � DCB and DEN � TECBgroups (data not shown). Neither TECB nor DCB (with DENinitiation) caused a significant deviation in total body weight

by the end of the eight-week study (Table 1). The degree ofliver weight change in the DEN � TECB group was much lesscompared to that observed previously in the DEN � HCBgroup, but more than those in the DEN � PECB group (Ou etal., 2001). Histopathological examination of hematoxylin &eosin (H&E)-stained liver sections of TECB-treated animalsshowed hepatocellular hypertrophy indicated by enlargementin cytoplasm, and nucleus in the centrilobular region. Previ-ously, hypertrophy was also observed in the DEN � PECBgroup and even more pronounced in the DEN � HCB group(Ou et al., 2001), while no such abnormality was observed intissues of DCB-treated animals.

Time-course changes in cell division rates. Time-depen-dent changes in cell division rates were determined by theBrdU-labeling index among DEN controls, DEN � TECB, andDEN � DCB groups (Table 2). As shown in Table 2, therewere approximately 60-fold increases in the division rate ofnormal hepatocytes following partial hepatectomy in all treat-ment groups. Neither DCB nor TECB treatments caused anysignificant change in cell division rates compared to DENcontrols at any of the time points examined. The slowestdivision rate of normal hepatocytes determined in this studywas 0.0016 0.0004/day (Table 2). The basal division rate ofhepatocytes was thus approximately 0.0015/day as reportedpreviously (Ou et al., 2001), and was used in our simulationmodel (Table 3).

Time-course development of GST-P foci. Preneoplastic fo-cal lesions in liver were identified by positive staining forGST-P. A regional (centrilobular), nonfocal GST-P stainingpattern outside of the putative preneoplastic live foci was alsoobserved in PECB- or HCB-treated animals (Thomas et al.,1998a, b), but not in TECB or DCB-treated animals. Tissueareas and individual focal transections from each tissue sectionwere recorded and two-dimensional focal transection data werethen converted to three-dimensional foci number and volume

TABLE 1Comparison of 1,2,4,5-Tetrachlorobenzene or 1,4-Dichlorobenzene on Body and Liver Weight of F344 Rats Subjected to an

Initiation/Promotion Protocol

Treatment group

Body weight (g) Final liver weight

Initiala Finalb gb %b

DEN � CO 254.20 12.41 305.18 14.45 10.46 0.12 3.45 0.17DEN � TECB 253.44 4.03 297.86 17.61 14.16 1.67c 4.70 0.3c

DEN � DCB 259.92 13.03 302.72 13.54 10.56 0.40 3.40 0.1

Note. Experimental data on HCB and PECB had been reported earlier (Ou et al., 2001). Animals were administered a dose of DEN (200 mg/kg, ip), and twoweeks later, they were given 0.1 mmol/kg per day, seven days/week of TECB, DCB, or corn-oil (CO) vehicle control for six weeks. A two-thirds partialhepatectomy was performed on all animals on day 21. The data are expressed as mean SD of at least four animals in each group.

aMeasurements on day 0.bMeasurements on day 56.cAnimals in the DEN � TECB group have a significant increase in liver weight and final relative liver weight compared to those in the control DEN group,

p � 0.05.

305STOCHASTIC SIMULATION OF FOCI DEVELOPMENT

by guest on May 25, 2014

http://toxsci.oxfordjournals.org/D

ownloaded from

as described above. As shown in Figure 3, a time-dependentreduction in foci number was accompanied by a concurrentincrease in foci volume from day 23 to day 56 in the DEN �TECB, DEN � DCB, and DEN control groups. At the finaltime point (day 56), DEN � TECB treatment group demon-strated a marked increase in the percentage of total foci volumein the liver (0.369 0.038%) compared to DEN controls0.261 0.024 %). At the same time, the percentage of totalfoci volume in the DEN � DCB group (0.241 0.06%) didnot differ significantly from the DEN control group. The totalfoci number in the DEN � TECB (2123 241) and DEN �DCB group (2394 190) at the final time point (day 56) issignificantly increased compared to DEN controls (1599 126; p � 0.05).

Clonal Growth Simulation Modeling

Model parameters for proliferation of normal hepatocytes.Model parameters for normal hepatocytes prior to DCB orTECB treatment on day 14 (Fig. 4a) were obtained usingpublished values (Kato et al., 1993). A single high dose ofDEN resulted in the loss of cells via cytotoxicity and a corre-sponding decrease in liver weight (Fig. 4b), which was fol-lowed by compensatory proliferation responses (Fig. 4a). Be-cause there were no observed effects of DCB or TECB on thecell division rate of normal hepatocytes (Table 2), DEN �DCB and DEN � TECB groups were modeled as that of DENcontrols from day 14 to the end of the modeling period (Fig.4a). By quantitative morphometric analysis, we observed areduced hepatocyte density (number of hepatocytes per unitvolume) attributable to increases in both cytoplasmic and nu-cleus volumes in PECB, HCB, and TECB treated animals (Ouet al., 2001). Changes in cell division rates and hepatocytedensity were incorporated into the simulation model. Themodel describes the increased liver weight in the DEN �

TECB group (Fig. 4b) and in the DEN � PECB and DEN �PCB groups as shown previously (Ou et al., 2001).

Model parameters for the DEN controls. The model wasfirst parameterized to describe foci development in animalsreceiving DEN initiation but without partial hepatectomy (Fig.5a, DEN panel, solid gray line). The total number of focirapidly increased to its peak level around day 14, and with arapid decline afterwards (Fig. 5a). This profile is not surprisingsince only foci clones larger than two cells in size wererecorded in the simulation model. In comparison to foci data ofDEN controls with partial hepatectomy, both foci volume andtotal foci number were significantly elevated soon after partialhepatectomy (Fig. 5a, DEN panel). Stochastic modeling gen-erates a series of similar outcomes given the same modelparameters, as illustrated by the time course changes in focivolume for five runs (Fig. 5a). Model parameters for describingthe concurrent DEN control data for PECB and HCB studies(Ou et al., 2001) were used to see whether they could alsodescribe concurrent DEN control data in the TECB and DCBstudies. To describe the foci data of the concurrent DENcontrols for the DCB and TECB study (Fig. 5b), no modifica-tions to the division or death rates of A and B cells (�a, �a, �b,�b) were needed and only modifications of the mutation rate toA or B cells were necessary (Table 3). The mutation rate to Aand B cells immediately following DEN treatment (days 0–7)is 9.9 10–5 and 1.5 10–5, respectively, compared to 12.1 10–5 and 3.3 10–5 as estimated for the previous study (Ou etal., 2001). The mutation rate to initiated A cells is approxi-mately sixfold higher than that to initiated B cells, which isconsistent with the reported percentage (5–23%) of resistanceclones (B cells) following DEN initiation (% resistant pheno-type is dependent on the promoter tested; Yusuf et al., 1999).Except for the period of partial hepatectomy, the cell divisionrate of initiated cells is highest immediately following DEN

TABLE 2Determination of Cell Proliferation Rate of Normal Hepatocytes in Liver Tissues of F344 Rats Subjected to an

Initiation/Promotion Protocol

DEN DEN � TECB DEN � DCB

LI (%) Cell division rate (/day) LI (%) Cell division rate (/day) LI (%) Cell division rate (/day1)

Day 23 without PHa 2.50 1.25 0.0042 0.0021 2.78 1.19 0.0047 0.0020 2.67 0.60 0.0045 0.0010Day 23 41.23 2.79 0.2661 0.0241 44.97 5.60 0.3006 0.0509 39.16 6.88 0.2509 0.0576Day 26 14.64 10.68 0.0819 0.0651 13.23 3.89 0.0713 0.0229 6.71 3.0 0.0349 0.0161Day 28 1.45 1.46 0.0073 0.0074 2.85 1.85 0.0145 0.0095 0.62 0.26 0.0031 0.0013Day 47 1.87 0.73 0.0031 0.0012 0.97 0.24 0.0016 0.0004 1.02 0.49 0.0017 0.0008Day 56 1.10 0.74 0.0018 0.0012 4.46 5.41 0.0078 0.0098 2.12 1.23 0.0036 0.0021

Note. The initiation agent, DEN, was administered ip (200 mg/kg) on day 0. A two-third partial hepatectomy was performed on day 21. 1,2,4,5-TECB or 1,4-DCB was delivered by gavage starting on day 14 at the dose of 0.1 mmol/kg/day, seven days/week. Experimental data on HCB and PECB had been reportedearlier (Ou et al., 2001). The data are expressed as mean SD of at least four animals in each group. Labeling index (LI) was determined by the number ofBrdU-labeled hepatocytes divided by the total number of cells counted. LI was converted to cell division rate according to the formula defined by Moolgavkarand Luebeck (1992).

aMeasurements in groups without partial hepatectomy (PH) on day 23.

306 OU ET AL.

by guest on May 25, 2014

http://toxsci.oxfordjournals.org/D

ownloaded from

TABLE 3Model Parameters Used in the Two-Stage Clonal Growth Model

Parameter Notation (unit) Value Parameter estimation

All groups

Cell division rate of normal hepatocytes � (1/day) 0.0015–0.32 Literaturea and measuredCell death rate of normal hepatocytes � (1/day) 0.0015–0.10 Literaturea

Mutation rate to A cells �a (per cell per day) Days 0–7: 9.9 10–5 Literatureb and fittedc

Days 7–14: 1.17 10–5

Days 14–120: 0.17 10–5

Mutation rate to A cells �b (per cell per day) Days 0–7: 1.5 10–5 Literatureb and fittedDays 7–14: 1.17 10–5

Days 14–120: 0.17 10–5

DENHepatocyte density Number of cells/cm3 Days 0–35: 1.68 108 Measured and fitted

Days 35–120: 1.65 108

Cell division rate of initiated hepatocyte (A type) �a (1/day) Days 0–7: 0.42 Literaturea,d and fittedDays 7–14: 0.405Days 14–21: 0.015Days 21–28: 0.13Days 28–35: 0.01Days 35–120: 0.006

Cell death rate of initiated hepatocyte (A type) �a (1/day) Days 0–7: 0.002 Literaturea,d and fittedDays 7–14: 0.002Days 14–21: 0.08Days 21–28: 0.0015Days 28–35: 0.15Days 35–120: 0.10

Cell division rate of initiated hepatocyte (B type) �b (1/day) Days 0–7: 0.42 Literaturea,d and fittedDays 7–14: 0.37Days 14–21: 0.015Days 21–28: 0.35Days 28–35: 0.13Days 35–120: 0.002

Cell death rate of initiated hepatocyte (B type) �b (1/day) Days 0–7: 0.002 Literaturea,d and fittedDays 7–14: 0.002Days 14–21: 0.002Days 21–28: 0.0015Days 28–35: 0.05Days 35–120: 0.004

DEN � TECBHepatocyte density Number of cells /cm3 Days 0–21: 1.68 108 Measured and fitted

Days 21–28: 1.63 108

Days 28–35: 1.39 108

Days 35–120: 1.27 108

Cell division rate of initiated hepatocyte (A type) �a (1/day) Days 0–7: 0.42 Literaturea,d and fittedDays 7–14: 0.405Days 14–21: 0.015Days 21–28: 0.13Days 28–35: 0.01Days 35–120: 0.006

Cell death rate of initiated hepatocyte (A type) �a (1/day) Days 0–7: 0.002 Literaturea,d and fittedDays 7–14: 0.002Days 14–21: 0.08Days 21–28: 0.0015Days 28–35: 0.105 (0.7)e

Days 35–120: 0.09 (0.9)e

Cell division rate of initiated hepatocyte (B type) �b (1/day) Days 0–7: 0.42 Literaturea,d and fittedDays 7–14: 0.37Days 14–21: 0.015Days 21–28: 0.35Days 28–35: 0.13Days 35–120: 0.002

Cell death rate of initiated hepatocyte (B type) �b (1/day) Days 0–7: 0.002 Literaturea,d and fittedDays 7–14: 0.002Days 14–21: 0.002Days 21–28: 0.0015Days 28–35: 0.037 (0.7)e

Days 35–120: 0.002 (0.5)

Note. Parameter values on HCB and PECB had been reported earlier (Ou et al., 2001). Parameters for DCB are identical to those of DEN controls. Daily treatments of TECB or DCB were started on day 14.aKato et al., 1993; Rotstein et al., 1984, 1986; Travis et al., 1991.bDragan et al., 1994; Yusuf et al., 1999.cOptimized by fitted model output to experimental data.dJang et al., 1993; Kato et al., 1993; Satoh et al., 1989; Tiwawech et al., 1991.eThe value in parenthesis indicates the fold increase compared to the DEN controls, the remaining values are identical to the DEN controls.

307STOCHASTIC SIMULATION OF FOCI DEVELOPMENT

by guest on May 25, 2014

http://toxsci.oxfordjournals.org/D

ownloaded from

treatment (0.42/day), and gradually declines to 0.002–0.006/day (Table 3). The growth characteristics of A and B cells donot differ significantly from each other prior to the adminis-tration of chlorobenzenes on day 14. For both A and B cells, itappears that cell death mechanisms are activated in response tothe excess growth of initiated cells (i.e., death rates of A and Bcells are high from days 28–35 and days 35–120, Table 3). Thedeath rate of initiated A cells tends to be higher than that forinitiated B cells. The death rate of initiated A cells is 0.08/dayon days 14–21 compared to 0.002/day for B cells for the sametime interval. The net growth rate of both A and B cells issignificantly elevated immediately following partial hepatec-tomy (days 21–28, Table 3).

Modeling analysis of DEN � TECB and DEN � DCB data.Since the time course of DCB-induced foci number and vol-ume did not differ significantly from those for DEN controls,the parameters for the DEN controls were used to describe the

data for the DEN � DCB group (Table 3, Fig. 5b). As for theDEN � TECB group, we first assumed that TECB was amutagen. With this assumption, the simulation model did notyield satisfactory fit to the time-course dynamic changes of focinumber and foci volume simultaneously. However, when in-corporating the negative selection hypothesis of tumor promo-tion (two-cell hypothesis), the model effectively describedexperimental data in the DEN � TECB group (Fig. 5b). Themodel describes the first increase in foci volume as a result ofTECB treatment during the two weeks following partial hep-atectomy. Peak formation of foci was observed around day 35,regardless of continuous treatment with TECB (Fig. 5b). Theincrease in foci number during partial hepatectomy regenera-tion is followed by a rapid decline (Fig. 5b). These time-coursechanges were similar to those of the DEN � PECB andDEN � HCB group, except that relative increases in the totalfoci volume (%) above the concurrent DEN control followingpartial hepatectomy were much larger in the DEN � PECB andDEN � HCB group than those of the DEN � TECB group(Fig. 5). The parameters for modeling foci growth of theDEN � TECB group are summarized in Table 3. The assump-

FIG. 4. Comparison of the clonal model outputs with experimental mea-surements of cell division rates of normal hepatocytes and liver weights.Time-dependent changes in cell division rates of normal hepatocytes and liverweights were measured in animals subjected to an initiation/promotion proto-col using DEN as an initiator and TECB or DCB as promoting agents. Solidlines represent simulation results from the clonal growth model. (a) Solidcircles shown are hepatocyte division rates determined using the BrdU-label-ing index as described in the Methods. Data points prior to Day 20 in all threegroups are published data from Kato et al., 1993. (b) Solid circles shown areliver weight data expressed as mean SD of at least four animals at each timepoint.

FIG. 3. Time-dependent changes in GST-P foci number and volume inanimals subjected to an initiation/promotion protocol using DCB or TECB asthe promoting agents. Animals were administered a dose of DEN on Day 0,TECB or DCB was delivered by gavage starting on Day 14 at dose of 0.1mmol/kg per day, seven days/week. A two-thirds partial hepatectomy wasperformed on all animals on day 21. Liver tissues were collected 23, 26, 28, 47,56 days (days 23, 26, 28, 47, 56) following initial DEN dosing. Data areexpressed as mean SD of at least four animals at each time point *Signif-icantly different from DEN control group (p � 0.05).

308 OU ET AL.

by guest on May 25, 2014

http://toxsci.oxfordjournals.org/D

ownloaded from

tion of two initiated cell populations (A and B cells) effectivelydescribes the rapid increase in foci volume, with a concomitantdecline of foci number associated with DEN controls, DEN �DCB, and DEN � TECB treatments. The effect of TECB onfoci volume expansion is consistent with the increase in net

growth rate of B cells following partial hepatectomy (day28–35 and day 35–120, Table 4). Despite continuous treat-ments of TECB from day 14 to day 56, the net growth rate ofB cells is gradually reduced between days 21–28 and 35–120.

Sensitivity analysis of model. A summary of the sensitivityanalysis is presented in Figure 6 with the normalized sensitivitycoefficient for each model parameter calculated. As shown inFigure 6, as few as five runs were able to identify sensitiveparameters similar to those derived from larger numbers ofruns (e.g., 160 runs). The most sensitive parameter for deter-mining the final total foci volume at the end of the eight-weekmedium-term bioassay is �b(21–28), the division rate of B cells,immediately following partial hepatectomy (Fig. 6). Thismodel parameter, �b(21–28) has a mean normalized sensitivitycoefficient of about 2.5, implying that a 5% change in param-eter values could result in a 12.5% increase in total focivolume. The mutation rate to B cells (�b) is also among themost sensitive parameters of the model (Fig. 7). Other sensitivedeterminants of total foci volume are �b(28–35) and �b(28–35), thedivision and death rates of B cells one week following partialhepatectomy (days 28–35). A negative sensitivity coefficient,in the case of �b(28–35), implies that a decrease in the parametervalue would result in an increase in the final measurement.Examining time course progression in foci growth as stipulatedin the simulation model provided another means for identifyingcritical determinants of final total foci volume. For example,the simulation model described distinct time courses for num-bers of A and B foci in the DEN � TECB group (Fig. 7). Forthe initial stage up to day 40, there is a significantly highernumber of foci, which are comprised mostly of A cell foci. Astime progresses, total foci number is diminished. B cells witha significant growth advantage following partial hepatectomycontinue to thrive and eventually become the major foci typearound day 60 (Fig. 7). Since these putative B cells are ulti-mately the dominant determinant for final foci volume, it is notsurprising to find that the mutation rate to B cells (�b) is one ofthe most sensitive parameters of the model (Fig. 6).

Comparison of four chlorobenzene congeners. The imple-mentation of clonal growth modeling for four chlorobenzenecongeners allowed us to compare their estimated parametervalues. We have previously shown marked differential focisize distribution between the DEN � PECB and DEN � HCBgroups (Ou et al., 2001). On the final day of the Ito assay (day56), the DEN � HCB group consists of mainly large foci,while the DEN � PECB group contains foci in all size classes(Ou et al., 2001). In order to describe the differential progres-sion profiles of total foci volume and foci number for the fourchlorobenzenes, different sets of parameters were necessary(Table 4). The striking similarity observed for all congeners isthat despite continuous treatment from day 14 to 56, the netgrowth rate of B cells gradually diminishes between days 21 to120 (Table 4). This analysis is consistent with observation infocal hepatocytes during administration of other chemicals

FIG. 5. Comparison of the clonal model outputs with experimental mea-surements of foci growth for four chlorobenzene congeners. Time-dependentchanges in foci growth were measured in animals subjected to an initiation/promotion protocol using DEN as an initiator and DCB, TECB, PECB, or HCBas a promoting agent. To illustrate the stochastic nature of foci growth, solidblack lines show results from five runs of simulation. Solid circles shown areexperimental data expressed as mean SD of at least four animals at eachtime point. (a) Results from PECB or HCB and the concurrent DEN control.For comparison, simulation results for the DEN controls without partial hep-atectomy (lower solid gray lines, including experimental data of Jang et al.(ref. 43; triangle symbols) are shown along with those of the DEN controls. (b)Results from TECB or DCB and the concurrent DEN controls.

309STOCHASTIC SIMULATION OF FOCI DEVELOPMENT

by guest on May 25, 2014

http://toxsci.oxfordjournals.org/D

ownloaded from

such as TCDD (Moolgavkar et al., 1996). Molecular endpointscontributing to foci development were examined on the lasttime point (day 56) of the Ito assay for the four chlorobenzenes(Table 4) and revealed that the estimated net growth rate of Bcells during days 28–35 (�b – �b (28–35)) was the only parameterthat showed correlation with cell and molecular indices. Induc-tion of CYP2B1/2, CYP1A2, c-fos, enlarged liver, and histo-logical changes are correlated with an increasing �b – �b(28–35).Furthermore, larger �b – �b(28–35) values are also associated withhigher total GST-P foci volume. Interestingly, the sensitivityanalysis indicated that the model parameter, �b – �b(28–35) isamong the most sensitive parameters for determining final totalfoci volume and the normalized sensitivity coeffecients of�b(28–35) and �b(28–35) are approximately �0.5 and 1, respectively(Fig. 7). Thus, although the �b – �b (28–35) parameter value forPECB is only 30% higher than that for TECB, this increasecould contribute to a significantly (30 to 45%) elevated finalfoci volume. The effects of different chlorobenzenes on normalhepatocyte proliferation do not correlate with final foci growthand �b – �b(28–35). In addition, altered GSH:GSSG status, or

induction of CYP2E1 and porphyria do not appear to becorrelated with the �b – �b(28–35) parameter.

DISCUSSION

We previously implemented a stochastic simulation modelfor describing foci development following exposure to PECBor HCB in a medium-term bioassay (Ou et al., 2001). In thepresent work, we describe an independent set of TECB andDCB foci data with the same simulation model. The modelconstruction, incorporating two initiated cell populations (re-ferred to as A and B cells), effectively simulated results fromall four chlorobenzene congeners. Comparison of simulationparameters and molecular endpoints of the four congenersshowed correlation between the identified model parametergoverning the net growth rate of B cells during the two-weekregenerative period following partial hepatectomy with induc-tion of c-fos, CYP2B1/2, and CYP1A2 (Gustafson et al.,2000), enlarged liver, and total foci volume. These findings areconsistent with the sensitivity analysis indicating that the net

TABLE 4Summary of Molecular Indices and Simulation Model Parameters for Four Chlorobenzene Congeners

DENa DCB TECB DENb PECB HCB

Net growth rate of A cells�a – �a(21–28) 0.1285 0.1285 0.1285 0.1285 0.1935 0.1935�a – �a (28–35) –0.14 -0.14 –0.095 –0.14 –0.08 –0.215�a – �a (35–120) –0.094 –0.094 –0.084 –0.094 –0.064 –0.144

Net growth rate of B cells�b – �b(21–28) 0.3485 0.3485 0.3485 0.3485 0.3485 0.3485�b – �b(28–35) 0.08 0.08 0.093 0.08 0.128 0.119�b – �b(35–120) –0.002 –0.002 0 –0.002 0 –0.001

Molecular endpoints examined onthe final time point (day 56)

GST-P foci 0 0 � 0 ��(�)c,d ��c,d

Normal Hepatocyte proliferation 0 0 0 0 0c �c

Enlarged liver 0 0 � 0 � ��Histological changes 0 0 � 0 � ��Porphyria 0 0e 0 f 0 �g ��g

c-fos protein 0 0 � 0 �� ��CYP1A2 protein 0 0 � 0 � ��CYP2E1 protein 0 � � 0 � �CYP2B1/2 protein 0 0 � 0 � �GSH:GSSG ratio 0 0 � 0 � 0

Note. Other parameters not listed in this table are those that are not affected by any chlorobenzene treatments. Simulation model parameters for PECB or HCBare obtained from our previous report (Ou et al., 2001). The notations, �a, �a, �b, �b, represent the division and death rates of A, B cells, respectively. Forexample, �b(28–35) represents the death rate of B cells from days 28 to 35, �a – �a (28–35) represents the net growth rate (cell division minus death rate) of A cellsfrom days 28 to 35. 0 � no observable change from DEN controls, � � significantly increased from DEN controls, �� � significantly increased from �. Unlessindicated, results on cell and molecular indices are taken from our previous report (Gustafson et al., 2000).

aConcurrent DEN controls for the DCB and TECB treatments.bConcurrent DEN controls for the PECB and HCB treatments.cOu et al., 2001.dIf only taking into account the large foci, PECB is less potent (�) compared to HCB (��).eFrom Carlson et al., 1977.fFrom Chu et al., 1983.gFrom Thomas et al., 1998b.

310 OU ET AL.

by guest on May 25, 2014

http://toxsci.oxfordjournals.org/D

ownloaded from

growth rate of B cells during the two-week regenerative periodis a sensitive model parameter for determining final total focivolume.

Liver Regeneration and Carcinogenicity

Our quantitative analysis underscores the importance ofassessing effects of chemicals on liver regeneration for under-standing the carcinogenic potential of these compounds. Weshowed that larger increases in total foci volume caused byPECB or HCB were consistently accompanied by larger netgrowth rate of B cells from days 28 to 35, �b – �b(28–35) whencompared to TECB or DCB. A positive correlation betweenincreased foci volume and �b – �b(28–35) may suggest a mech-anism for chemicals to interfere with cellular mechanismsgoverning hepatocytes returning to baseline growth rates afterstimulation of cell proliferation (e.g., 28–35 days). This quan-titative analysis is consistent with several reports showing thatfocal hepatocytes, unlike hepatocytes in the surrounding liver,often show a failure to return to baseline growth rates afterstimulation of cell proliferation by partial hepatectomy (Rot-stein et al., 1986; Tiwawech et al., 1991). Furthermore, mol-ecules involved in maintaining optimal liver mass after partialhepatectomy are shown to overlap extensively with thosewhose perturbations are linked to carcinogenic events. Theseinclude a large number of early response genes such as c-myc,c-fos, and c-jun, those involved in the priming of quiescenthepatocytes, such as injury-related cytokines, TNF and IL-6,and the subsequent activation of NF-kappa B, AP-1, andSTAT3 transcription factors (Fausto and Webber, 1994). Onceprimed, hepatocytes respond to growth factors such as hepa-

tocyte growth factor (HGF) and TGF-� to increase DNAsynthesis (Fausto and Webber, 1994). An important mechanis-tic determinant of liver regeneration after partial hepatectomyis the release of growth suppressing molecules such as TGF-�to slow down or stop the G1/S transition (Fausto and Webber,1994). Dysregulation of TGF-�, and other growth regulatorymolecules such as TGF-� and c-myc, c-fos are often found inchemically induced preneoplastic foci and tumors (Pitot,1996). Thus, TGF-� could potentially represent a candidatemolecular target underlying the failure of return to basalgrowth rates after partial hepatectomy in the presence of pro-moter treatment. The marked similarity between moleculesassociated with hepatic regeneration and carcinogenesis sug-gests that examining interactive effects of chemicals on thewindow of regeneration following partial hepatectomy may beimportant to understanding mechanisms of carcinogenesis.

Correlation between Clonal Growth Model Parameters andOther Molecular Endpoints

Our combined experimental and modeling work with fourchlorobenzene congeners under equimolar dose levels indicatethat the ability of a chemical to induce CYP2B1/2, CYP1A2, isin agreement with concurrent effects on B cell net growth ratefrom day 28 to 35 (�b – �b(28–35)) and increases in foci volume.Theses results indicate that the effect of chlorobenzene admin-istration on the net growth rate of initiated B cell populationduring the regeneration period may be related to the activationof one or more transcription pathways, possibly driven by twoligand activated transcription factors, the aryl-hydrocarbon(Ah) receptor and the putative phenobarbital-type responsereceptor. For example, induction of CYP1A1, an Ah receptor-mediated response is used as an indicator of TCDD-inducedchanges in growth kinetics (Conolly and Andersen, 1997).

FIG. 6. Sensitivity analysis of model parameters for total foci volume.Normalized sensitivity coefficients for total foci volume at the end of 8-weekmedium-term bioassay were calculated based on 5, 20, or 30% changes in theparameter. A 20% change in parameter calculated based on the result of fiveruns gave similar changes with 5% changes in parameters for 160 runs. Thesensitive parameters are �b(21–28), �b(28–35), �b (28–35), and�b(1–7).

FIG. 7. Progression of A and B initiated clones in the model simulation offoci development using TECB as a promoting agent. Model predictions of A,B and total foci number during the course of DEN initiation/TECB promotiontreatment. Partial hepatectomy was performed on day 21.

311STOCHASTIC SIMULATION OF FOCI DEVELOPMENT

by guest on May 25, 2014

http://toxsci.oxfordjournals.org/D

ownloaded from

Induction of CYP2B1/2 and hypertrophy are observed follow-ing exposure to phenobarbital (Staubli et al., 1969) and thecomparative potency of barbiturate-type tumor promoters iscorrelated with their potency as inducers of CYP2B1/2 (Nimset al., 1987). Enlarged livers observed following the chloro-benzene treatments in this study are a result of hypertrophy,and are partly due to increased cytochrome P450 content andenlarged endoplasmic reticulum of hepatocytes (Staubli et al.,1969). Our comparative assessment of four chlorobenzenecongeners provides an approach to the identification of molec-ular candidates that may serves as effective indicators forestimation of foci growth. However, agents with differentmechanisms of action are likely to promote distinct and over-lapping subsets of initiated hepatocytes (Dragan and Pitot,1992), probably in a region-specific pattern within the liver(Chen et al., 1995; Thomas et al., 1998a). Therefore, identifi-cation of a common subset of molecular markers for predictingfinal foci growth will depend on collection of time and regiondependent mechanistic information for many other com-pounds.

Comparison of the Current Model with Previously ReportedModels

Key differences between our simulation model and previ-ously reported models should be noted here. The current sto-chastic model simulates three-dimensional foci data, whereastwo-dimensional data is implemented in the work of Mool-gavkar and others (Luebeck et al., 1991; Portier et al., 1996).Proliferation of normal hepatocytes is described deterministi-cally in the current model, whereas a stochastic mode is usedin previous implementations (Luebeck et al., 1991; Portier etal., 1996). The simplification of our model speeds computa-tion, and provides results equivalent to a fully stochastic cal-culation. The current approach also assumes that with a muchlarger sampling of normal hepatocytes compared to the numberof mutational events, the variability in the proliferation ofnormal hepatocytes is not a significant factor in the predictionof final foci volume. This assumption was confirmed by oursensitivity analysis, showing that the primary determinants offinal foci volume reside in the mutation rate to initiated cellsand growth parameters of initiated cells. The current simula-tion model also uses subjective means (such as visual inspec-tion) rather than objective means of parameter estimationmethods (in most cases maximal likelihood estimation) (Lue-beck et al., 1991; Portier et al., 1996). While subjective meansto approximate the mean behavior might be sufficient fordelivering an understanding of underlying biological pro-cesses, development of rigorous parameter estimation methodswould extend the acceptance of the current simulation modelfor carcinogenesis studies (Luebeck et al., 1991; Portier et al.,1996; Sherman and Portier, 1998). One of our attempts in thecurrent work was to begin to address these issues. By perform-ing sensitivity analysis, we are able to identify key model

parameters that should make future implementation of formaloptimization methods computationally feasible.

Experimental Challenge to Verify the Model Hypothesis

The success of the two-cell hypothesis for describing thecurrent data sets further strengthened previous model predic-tions (Ou et al., 2001) on the heterogeneity of cell kineticsamong initiated cells. In particular, the need to implement thetwo-cell hypothesis for describing the time-course data of DENcontrols in the absence of further treatments is consistent withthe experimental observations that initiated cells with resistantphenotypes (such as B cells) can arise early during carcino-genesis (Yusuf et al., 1999). The need for two types of foci isnot surprising given the known phenotypic diversity of alteredhepatic foci (Dragan and Pitot, 1992). The challenge remainsas to verify the model hypothesis experimentally. Experimentalwork to measure cell division and death rate of initiated focicould be a starting point, however, regional differences in fociformation, cell division rate, and enzyme induction are knownto occur following chemical exposure in the liver (Chen et al.,1995). Effects of chemicals on initiated cell population may beinferred from their effects on normal hepatocytes. WhileTECB, PECB, and HCB all induced significant increase in focivolume, only HCB had effects on the cell division rate ofnormal hepatocytes. Increased cell proliferation in normal tis-sues may not necessarily represent a stimulus for the growth ofpreneoplastic foci, and agents affecting apoptosis could alsocontribute to the growth of preneoplastic foci. Repeated ad-ministration of a mitogen 3,3�,5-triiodo-L-thyronine results inan enhanced proliferation of normal rat liver, but leads to areduction in the number of GST-P lesions with no increase inthe size of the remaining ones (Ledda-Columbano et al., 1999).Thus, effects of chemicals on normal hepatocytes are notnecessarily correlated with their effects on the initiated cellpopulation. Measuring cell division and death rates of enzyme-altered foci along with concurrent analysis of global proteinand gene expression in the foci may help identify candidatemarker genes for the resistant initiated clones (B cell pheno-type). Possible experimental means to distinguish A and Bcells may include immunocytochemistry analysis for bothGST-P and other markers such as TGF-� and mannose 6-phos-phate/insulin-like growth factor II receptor (Mills et al., 1998).The resistant initiated cells (B cells) are likely to stain positivefor GST-P, but with reduced TGF-� receptor levels (Mills etal., 1998).

The successful application of the existing model parameters(Ou et al., 2001) for analyzing new chemical data sets in anindependent experiment indicates the versatility of the currentmodel, and that the current model might also be used toanalyze data for other chemical compounds in medium-termbioassay. The present model may also comprise a flexibleplatform for incorporation of kinetic data from characterizedbiochemical pathways and interactions as well as genome and

312 OU ET AL.

by guest on May 25, 2014

http://toxsci.oxfordjournals.org/D

ownloaded from

proteome expression profiling associated with development ofcarcinogenesis.

ACKNOWLEDGMENTS

The authors want to thank Dr. Peter Brockwell at Dept. of Statistics,Colorado State University, for statistical consultations on the optimizationimplementation of the current model. Many colleagues at the Center forEnvironmental Toxicology and Technology participated in these experimentalstudies, in particular, Drs. Wendy Pott, Charlie Dean, and Mr. Lixin Feng.Their help is gratefully acknowledged. This work was supported in part by aresearch contract from the Air Force Office of Scientific Research (F49620-94-1-0304), an NIEHS Superfund Basic Research Program Project Grant (P42ES05949), and a cooperative agreement (U61/ATU881475) from the Agencyfor Toxic Substance and Disease Registry.

REFERENCES

Andersen, M. E., Mills, J. J., Jirtle, R. L., and Greenlee, W. F. (1995). Negativeselection in hepatic tumor promotion in relation to cancer risk assessment.Toxicology 102, 223–237.

Beck, J. V., and Arnold, K. J. (1977). Parameter estimation in engineering andscience. Wiley and Sons, New York.

Bratley, P., Fox, B. L., and Schroeder, P. (1996). A Guide to Simulation. 2nded. Springer-Verlag, New York.

Cabral, J. R., and Shubik, P. (1986). Carcinogenic activity of hexachloroben-zene in mice and hamsters. IARC Sci. Publ. 77, 411–416.

Cabral, R., Hoshiya, T., Hakoi, K., Hasegawa, R., and Ito, N. (1996). Medium-term bioassay for the hepatocarcinogenicity of hexachlorobenzene. CancerLett. 100, 223–226.

Carey, A. E., Dixon, T. E., and Yang, H. S. (1986). Environmental exposureto hexachlorobenzene in the USA. IARC Sci. Publ. 77, 115–120.

Carlson, G. P. (1977). Chlorinated benzene induction of hepatic porphyria.Experientia 33, 1627–1629.

Chen, Z. Y., White, C. C., He, C. Y., Liu, Y. F., and Eaton, D. L. (1995). Zonaldifferences in DNA synthesis activity and cytochrome P450 gene expressionin livers of male F344 rats treated with five nongenotoxic carcinogens. J.Environ. Pathol. Toxicol. Oncol. 14, 83–99.

Chu, I., Villeneuve, D., Secours, V., and Valli, V. E. (1983). Comparativetoxicity of 1,2,3,4 -, 1,2,4,5-, and 1,2,3,5-tetrachlorobenze in the rat: Resultsof acute and subacute studies. J. Toxicol. Environ. Health 11, 663–677.

Cohen, S. M., and Ellwein, L. B. (1990). Cell proliferation in carcinogenesis[see comments]. Science 249, 1007–1011.

Conolly, R. B., and Andersen, M. E. (1997). Hepatic foci in rats after dieth-ylnitrosamine initiation and 2,3,7,8-tetrachlorodibenzo-p-dioxin promotion:Evaluation of a quantitative two-cell model and of CYP 1A1/1A2 as adosimeter. Toxicol. Appl. Pharmacol. 146, 281–293.

Conolly, R. B., and Kimbell, J. S. (1994). Computer simulation of cell growthgoverned by stochastic processes: Application to clonal growth cancermodels. Toxicol. Appl. Pharmacol. 124, 284–295.

Delesse, M. (1848). Procede mecanique pour determiner la composition desroches. Ann. Mine 13, 379–388.

Dragan, Y. P., Hully, J. R., Nakamura, J., Mass, M. J., Swenberg, J. A., andPitot, H. C. (1994). Biochemical events during initiation of rat hepatocar-cinogenesis. Carcinogenesis 15, 1451–1458.

Dragan, Y. P., and Pitot, H. C. (1992). The role of the stages of initiation andpromotion in phenotypic diversity during hepatocarcinogenesis in the rat.Carcinogenesis 13, 739–750.

Ellwein, L. B., and Cohen, S. M. (1992). Simulation modeling of carcinogen-esis. Toxicol. Appl. Pharmacol. 113, 98–108.

Fausto, N., and Webber, E. M. (1994). Liver regeneration. In The Liver:Biology and Pathology (I. M. Arias, J. L. Boyer, and N. Fausto, Eds.), pp.1059–1084. Raven Press, New York.

Gustafson, D. L., Coulson, A. L., Feng, L., Pott, W. A., Thomas, R. S., Chubb,L. S., Saghir, S. A., Benjamin, S. A., and Yang, R. S. (1998). Use of amedium-term liver focus bioassay to assess the hepatocarcinogenicity of1,2,4,5-tetrachlorobenzene and 1,4-dichlorobenzene. Cancer Lett. 129,39–44.

Gustafson, D. L., Long, M. E., Thomas, R. S., Benjamin, S. A., and Yang,R. S. H. (2000). Comparative hepatocarcinogenicity of hexachlorobenzene,pentachlorobenzene, 1,2,4,5-tetrachlorobenzene, and 1,4-dichlorobenzene:Application of a medium-term liver focus bioassay and molecular andcellular indices. Toxicol. Sci. 53, 245–252.

Hill, R. H., Jr., Ashley, D. L., Head, S. L., Needham, L. L., and Pirkle, J. L.(1995). p-Dichlorobenzene exposure among 1,000 adults in the UnitedStates. Arch. Environ. Health 50, 277–280.

Ito, N., Imaida, K., Hasegawa, R., and Tsuda, H. (1989a). Rapid bioassaymethods for carcinogens and modifiers of hepatocarcinogenesis. Crit. Rev.Toxicol. 19, 385–415.

Ito, N., Tatematsu, M., Hasegawa, R., and Tsuda, H. (1989b). Medium-termbioassay system for detection of carcinogens and modifiers of hepatocarci-nogenesis utilizing the GST-P positive liver cell focus as an endpointmarker. Toxicol. Pathol. 17, 630–641.

Jang, J. J., Henneman, J. R., Kurata, Y., Uno, H., and Ward, J. M. (1993).Alterations in populations of GST-p-immunoreactive single hepatocytes andhepatocellular foci after a single injection of N-nitrosodiethylamine with orwithout phenobarbital promotion in male F344/NCr rats. Cancer Lett. 71,89–95.

Jirtle, R. L., Hankins, G. R., Reisenbichler, H., and Boyer, I. J. (1994).Regulation of mannose 6-phosphate/insulin-like growth factor-II receptorsand transforming growth factor beta during liver tumor promotion withphenobarbital. Carcinogenesis 15, 1473–1478.

Kato, M., Popp, J. A., Conolly, R. B., and Cattley, R. C. (1993). Relationshipbetween hepatocyte necrosis, proliferation, and initiation induced by dieth-ylnitrosamine in the male F344 rat. Fundam. Appl. Toxicol. 20, 155–162.

Ledda-Columbano, G. M., Perra, A., Piga, R., Pibiri, M., Loi, R., Shinozuka,H., and Columbano, A. (1999). Cell proliferation induced by 3,3�,5-triiodo-L-thyronine is associated with a reduction in the number of preneoplastichepatic lesions. Carcinogenesis 20, 2299–2304.

Luebeck, E. G., Moolgavkar, S. H., Buchmann, A., and Schwarz, M. (1991).Effects of polychlorinated biphenyls in rat liver: Quantitative analysis ofenzyme-altered foci. Toxicol. Appl. Pharmacol. 111, 469–484.

Mills, J. J., Falls, J. G., De Souza, A. T., and Jirtle, R. L. (1998). ImprintedM6p/Igf2 receptor is mutated in rat liver tumors. Oncogene 16, 2797–2802.

Moolgavkar, S. H., and Luebeck, E. G. (1992). Interpretation of labelingindices in the presence of cell death [see comments]. Carcinogenesis 13,1007–1010.

Moolgavkar, S. H., and Luebeck, G. (1990). Two-event model for carcino-genesis: Biological, mathematical, and statistical considerations. Risk Anal.10, 323–341.

Moolgavkar, S. H., Luebeck, E. G., Buchmann, A., and Bock, K. W. (1996).Quantitative analysis of enzyme-altered liver foci in rats initiated withdiethylnitrosamine and promoted with 2,3,7,8-tetrachlorodibenzo-p-dioxinor 1,2,3,4,6,7,8-heptachlorodibenzo-p-dioxin. Toxicol. Appl Pharmacol.138, 31–42.

Moolgavkar, S. H., and Venzon, D. J. (2000). Two-event model for carcino-genesis. Math. Biosci. 47, 55–77.

Morita, M. (1977). Chlorinated benzenes in the environment. Ecotoxicol.Environ. Saf. 1, 1–6.

Nims, R. W., Devor, D. E., Henneman, J. R., and Lubet, R. A. (1987).Induction of alkoxyresorufin O-dealkylases, epoxide hydrolase, and liver

313STOCHASTIC SIMULATION OF FOCI DEVELOPMENT

by guest on May 25, 2014

http://toxsci.oxfordjournals.org/D

ownloaded from

weight gain: Correlation with liver tumor-promoting potential in a series ofbarbiturates. Carcinogenesis 8, 67–71.

NTP (U.S. National Toxicology Program). (1985a). Toxicology and carcino-genesis studies of 1,2 di-chlorobenzene (O-dichlorobenzene) in F344/N ratsand B6C3F1 mice. TR255/ NTP, Research Triangle Park, NC.

NTP (U.S. National Toxicology Program) (1985b). Toxicology and carcino-genesis studies of chlorobenzene in F344/N rats and B6C3F1 mice. TR261/NTP, Research Triangle Park, NC.

NTP (U.S. National Toxicology Program) (1987). Toxicology and carcinogen-esis studies of 1,4 di-chlorobenzene (O-dichlorobenzene) in F344/N rats andB6C3F1 mice. TR319/ NTP, Research Triangle Park, NC.

Ogiso, T., Tatematsu, M., Tamano, S., Hasegawa, R., and Ito, N. (1990).Correlation between medium-term liver bioassay system data and results oflong-term testing in rats. Carcinogenesis 11, 561–566.

Ou, Y. C., Conolly, R. B., Thomas, R. S., Xu, Y., Andersen, M. E., Chubb,L. S., Pitot, H. C., and Yang, R. S. H. (2001). A clonal growth model:Time-course simulations of liver foci growth following penta- or hexa-chlorobenzene treatment in a medium-term bioassay. Cancer Res. 61, 1879–1889.

Peattie, M. E., Lindsay, D. G., and Hoodless, R. A. (1984). Dietary exposureof man to chlorinated benzenes in the United Kingdom. Sci. Total. Environ.34, 73–86.

Pitot, H. C. (1996). Stage-specific gene expression during hepatocarcinogen-esis in the rat. J. Cancer Res. Clin. Oncol. 122, 257–265.

Portier, C. J., Sherman, C. D., Kohn, M., Edler, L., Kopp-Schneider, A.,Maronpot, R. M., and Lucier, G. (1996). Modeling the number and size ofhepatic focal lesions following exposure to 2,3,7,8-TCDD. Toxicol. Appl.Pharmacol. 138, 20–30.

Rotstein, J., Macdonald, P. D., Rabes, H. M., and Farber, E. (1984). Cell cyclekinetics of rat hepatocytes in early putative preneoplastic lesions in hepa-tocarcinogenesis. Cancer Res. 44, 2913–2917.

Rotstein, J., Sarma, D. S., and Farber, E. (1986). Sequential alterations ingrowth control and cell dynamics of rat hepatocytes in early precanceroussteps in hepatocarcinogenesis. Cancer Res. 46, 2377–2385.

Saltykov, S. A. (1967). The determination of the size distribution of particlesin an opaque material from a measurement of the size distribution of theirsections. In Proceedings of the Second International Congress for Stereol-ogy, Chicago, IL, April 8–13 (H. Elias, Ed.), pp. 163–173. Springer-Verlag,Berlin.

Satoh, K., Hatayama, I., Tateoka, N., Tamai, K., Shimizu, T., Tatematsu, M.,

Ito, N., and Sato, K. (1989). Transient induction of single GST-P positivehepatocytes by DEN. Carcinogenesis 10, 2107–2111.

Sherman, C. D., and Portier, C. J. (1998). Eyes closed: Simple, intuitive,statistically sound, and efficient methods for estimating parameters of clonalgrowth cancer models [letter; comment]. Risk Anal. 18, 529–534.

Staubli, W., Hess, R., and Weibel, E. R. (1969). Correlated morphometric andbiochemical studies on the liver cell. II. Effects of phenobarbital on rathepatocytes. J. Cell Biol. 42, 92–112.

Stewart, F. P., Manson, M. M., Cabral, J. R., and Smith, A. G. (1989).Hexachlorobenzene as a promoter of diethylnitrosamine-initiated hepatocar-cinogenesis in rats and comparison with induction of porphyria. Carcino-genesis 10, 1225–1230.

Thomas, R. S., Conolly, R. B., Long, M. E., Benjamin, S. A., and Yang,R. S. H. (2000). A physiologically-based pharmacodynamic analysis ofhepatic-foci within a medium-term liver bioassay using pentachlorobenzeneas a promoter and diethylnitrosamine as an initiator. Toxicol. Appl. Phar-macol. 166, 128–137.

Thomas, R. S., Gustafson, D. L., Pott, W. A., Long, M. E., Benjamin, S. A.,and Yang, R. S. H. (1998a). Evidence for hepatocarcinogenic activity ofpentachlorobenzene with intralobular variation in foci incidence. Carcino-genesis 19, 1855–1862.

Thomas, R. S., Gustafson, D. L., Ramsdell, H. S., el-Masri, H. A., Benjamin,S. A., and Yang, R. S. (1998b). Enhanced regional expression of glutathioneS-transferase P1–1 with colocalized AP-1 and CYP 1A2 induction in chlo-robenzene-induced porphyria. Toxicol. Appl. Pharmacol. 150, 22–31.

Tiwawech, D., Hasegawa, R., Kurata, Y., Tatematsu, M., Shibata, M. A.,Thamavit, W., and Ito, N. (1991). Dose-dependent effects of 2-acetylamin-ofluorene on hepatic foci development and cell proliferation in rats. Carci-nogenesis 12, 985–990.

Tobin, P. (1986). Known and potential sources of hexachlorobenzene. IARCSci. Publ. 77, 3–11.

Travis, C. C., McClain, T. W., and Birkner, P. D. (1991). Diethylnitrosamine-induced hepatocarcinogenesis in rats: A theoretical study. Toxicol. Appl.Pharmacol. 109, 289–304.

Xu, Y. H., Dragan, Y. P., Campbell, H. A., and Pitot, H. C. (1998). STEREO:A program on a PC-Windows 95 platform for recording and evaluatingquantitative stereologic investigations of multistage hepatocarcinogenesis inrodents. Comput. Methods Programs Biomed. 56, 49–63.

Yusuf, A., Rao, P. M., Rajalakshmi, S., and Sarma, D. S. (1999). Developmentof resistance during the early stages of experimental liver carcinogenesis.Carcinogenesis 20, 1641–1644.

314 OU ET AL.

by guest on May 25, 2014

http://toxsci.oxfordjournals.org/D

ownloaded from