<%BANNER%>
PRIVATE ITEM
Digitization of this item is currently in progress.
A Computational model of the hypothalamic - pituitary - gonadal axis in female fathead minnows (Pimephales promelas) exp...
CITATION PDF VIEWER
Full Citation
STANDARD VIEW MARC VIEW
Permanent Link: http://ufdc.ufl.edu/AA00009672/00001
 Material Information
Title: A Computational model of the hypothalamic - pituitary - gonadal axis in female fathead minnows (Pimephales promelas) exposed to 17a-ethynylestradiol and 17b-trenbolone
Abbreviated Title: BMC Systems Biology
Physical Description: Book
Language: English
Creator: Li, Zhenhong
Kroll, Kevin J.
Jensen, Kathleen
Villeneauve, Daniel L.
Ankley, Gerald T.
Brian, Jayne V.
Sepúlveda, María S.
Orlando, Edward F.
Lazorchak, James M.
Kostich, Mitchell
Armstrong, Brandon
Denslow, Nancy D.
Watanabe, Karen H.
Publisher: BioMed Central
Publication Date: 2011
 Notes
Abstract: Abstract Background: Endocrine disrupting chemicals (e.g., estrogens, androgens and their mimics) are known to affect reproduction in fish. 17a-ethynylestradiol is a synthetic estrogen used in birth control pills. 17b-trenbolone is a relatively stable metabolite of trenbolone acetate, a synthetic androgen used as a growth promoter in livestock. Both 17a-ethynylestradiol and 17b-trenbolone have been found in the aquatic environment and affect fish reproduction. In this study, we developed a physiologically-based computational model for female fathead minnows (FHM, Pimephales promelas), a small fish species used in ecotoxicology, to simulate how estrogens (i.e., 17a-ethynylestradiol) or androgens (i.e., 17b-trenbolone) affect reproductive endpoints such as plasma concentrations of steroid hormones (e.g., 17b-estradiol and testosterone) and vitellogenin (a precursor to egg yolk proteins). Results: Using Markov Chain Monte Carlo simulations, the model was calibrated with data from unexposed, 17aethynylestradiol- exposed, and 17b-trenbolone-exposed FHMs. Four Markov chains were simulated, and the chains for each calibrated model parameter (26 in total) converged within 20,000 iterations. With the converged parameter values, we evaluated the model’s predictive ability by simulating a variety of independent experimental data. The model predictions agreed with the experimental data well. Conclusions: The physiologically-based computational model represents the hypothalamic-pituitary-gonadal axis in adult female FHM robustly. The model is useful to estimate how estrogens (e.g., 17a-ethynylestradiol) or androgens (e.g., 17b-trenbolone) affect plasma concentrations of 17b-estradiol, testosterone and vitellogenin, which are important determinants of fecundity in fish.
 Record Information
Source Institution: University of Florida
Holding Location: University of Florida
Rights Management: All rights reserved by the source institution.
System ID: AA00009672:00001

Downloads

This item is only available as the following downloads:

( PDF )


Full Text

PAGE 1

RESEARCHARTICLEOpenAccess Acomputationalmodelofthehypothalamicpituitary-gonadalaxisinfemalefathead minnows( Pimephalespromelas )exposedto 17 a -ethynylestradioland17 b -trenbolone ZhenhongLi 1 ,KevinJKroll 2 ,KathleenMJensen 3 ,DanielLVilleneuve 3 ,GeraldTAnkley 3 ,JayneVBrian 4 MaraSSeplveda 5 ,EdwardFOrlando 6 ,JamesMLazorchak 7 ,MitchellKostich 7 ,BrandonArmstrong 8 NancyDDenslow 2 andKarenHWatanabe 1* Abstract Background: Endocrinedisruptingchemicals(e.g.,estrogens,androgensandtheirmimics)areknowntoaffect reproductioninfish.17 a -ethynylestradiolisasyntheticestrogenusedinbirthcontrolpills.17 b -trenboloneisa relativelystablemetaboliteoftrenboloneacetate,asyntheticandrogenusedasagrowthpromoterinlivestock. Both17 a -ethynylestradioland17 b -trenbolonehavebeenfoundintheaquaticenvironmentandaffectfish reproduction.Inthisstudy,wedevelopedaphysiologically-basedcomputationalmodelforfemalefathead minnows(FHM, Pimephalespromelas ),asmallfishspeciesusedinecotoxicology,tosimulatehowestrogens(i.e., 17 a -ethynylestradiol)orandrogens(i.e.,17 b -trenbolone)affectreproductiveendpointssuchasplasma concentrationsofsteroidhormones(e.g.,17 b -estradiolandtestosterone)andvitellogenin(aprecursortoeggyolk proteins). Results: UsingMarkovChainMonteCarlosimulations,themodelwascalibratedwithdatafromunexposed,17 a ethynylestradiol-exposed,and17 b -trenbolone-exposedFHMs.FourMarkovchainsweresimulated,andthechains foreachcalibratedmodelparameter(26intotal)convergedwithin20,000iterations.Withtheconverged parametervalues,weevaluatedthemodel ’ spredictiveabilitybysimulatingavarietyofindependentexperimental data.Themodelpredictionsagreedwiththeexperimentaldatawell. Conclusions: Thephysiologically-basedcomputationalmodelrepresentsthehypothalamic-pituitary-gonadalaxisin adultfemaleFHMrobustly.Themodelisusefultoestimatehowestrogens(e.g.,17 a -ethynylestradiol)orandrogens (e.g.,17 b -trenbolone)affectplasmaconcentrationsof17 b -estradiol,testosteroneandvitellogenin,whichare importantdeterminantsoffecundityinfish. Background Invertebrates,suchasfish,thehypothalamic-pituitarygonadal(HPG)axiscontrolsreproductiveprocesses throughavarietyofhormoneswhichactontargettissuesdirectlyorindirectly[1,2].TheHPGaxiscanbe alteredbyendocrinedisruptingchemicals(EDCs)in theaquaticenvironmentwhichmimicendogenous hormones,altertheirconcentrations,orblocktheir actions[3]. Inrecentyears,manyscientificstudieshavebeenconductedtostudyreproductiveeffectsofEDCsinfathead minnow(FHM, Pimephalespromelas ),amodelsmall fishspeciesusedinecotoxicology[4-6].TwoEDCs, 17 a -ethynylestradioland17 b -trenbolone,havebeen widelystudiedasmodelestrogensandandrogens, respectively[7-11].Bothcompoundsalsoareenvironmentallyrelevantcontaminants. *Correspondence:watanabe@ebs.ogi.edu 1 DivisionofEnvironmentalandBiomolecularSystems,OregonHealth& ScienceUniversity,Beaverton,OR,97006,USA Fulllistofauthorinformationisavailableattheendofthearticle Li etal BMCSystemsBiology 2011, 5 :63 http://www.biomedcentral.com/1752-0509/5/63 2011Lietal;licenseeBioMedCentralLtd.ThisisanOpenAccessarticledistributedunderthetermsoftheCreativeCommons AttributionLicense(http://creativecommons.org/licenses/by/2.0),whichpermitsunrestricteduse,distribution,andreproductionin anymedium,providedtheoriginalworkisproperlycited.

PAGE 2

17 a -ethynylestradiol(EE2),asyntheticestrogenused inbirthcontrolpills,enterstheenvironmentmainly througheffluentsfromwastewa tertreatmentfacilities. ThereportedmedianEE2concentrationintheaquatic environmentvariesfrom<0.5to15ng/L[12].Duein parttoitshighbindingaffinityforestrogenreceptor (ER)[13-15],EE2affectstheHPGaxisinFHMatenvironmentallyrelevantconcentrations.ExposuretoEE2has beenshowntoresultinalteredhormoneprofiles,and increasedvitellogenin(VTG,aprecursorofeggyolk proteins)levelsinbothmaleandfemaleFHMs[16].In addition,aseven-year,whole-lakeexperimentconducted inCanada[17]showedthatchronicexposureofFHMs to5-6ngEE2/Lledtonear-extinctionofthisspecies fromthelake. 17 b -trenbolone(TB)isarelativelystablemetabolic productoftrenboloneaceta te,asyntheticandrogen usedasagrowthpromoterinlivestock(e.g.,cattle).TB enterstheenvironmentmainlyasrunofffromlivestock feedlots.Schifferetal.[18]reportedthattheTBconcentrationineffluentsofsolidcattledungwasaround19 ng/L.Durhanetal.[19]studiedacattlefeedlotlocated insouthwestcentralOhio,andreportedthattheTB concentrationinfeedlotdischargewasbetween10and 20ng/L.TBhasahighbindingaffinityfortheandrogen receptor(AR).WaterexposuretoTBatconcentrations similartothosefoundintheenvironmentdecreasesegg productioninFHMinconjunctionwithchangesin plasmaconcentrationsof17 b -estradiol(E2),testosterone (T),andVTGinfemales[7].Interestingly,relationships betweenTBwaterexposureconcentrationsandplasma E2,TandVTGconcentrationswerenotmonotonic,but were “ U-shaped ” [7]. TobetterunderstandthedynamicsoftheHPGaxisin femaleFHMsandtofacilitatetheevaluationofadverse outcomesonreproductionfrombothestrogenicand androgenicEDCexposure,wedevelopedaphysiologicallybasedcomputationalmodeltosimulatekeyreproductiveendpoints,suchasplasmaconcentrationsofE2, T,andVTG,inadultfemaleFHMs.Themodelsimulatesabsorption,distribution,andeliminationofTBand EE2byincorporatingsalientphysiologicalcharacteristics ofFHMsandmodellingbiochemicalpathwaysandreactionsmathematically.Thismodelisafirststeptoward predictingadverseoutcomesonreproduction,whichis animportantcomponentofec ologicalriskassessment. ItrobustlylinksTBandEE2exposuretoplasmasteroid hormoneandVTGconcentrations,whichcanthenbe usedtopredicteffectsonfecundity.Thoughthismodel doesnotsimulateoocytegrowthdynamicstopredict fecundity,itcanbeintegratedwithanoocytegrowth dynamicsmodeltodoso.Toourknowledge,itisthe firstphysiologicallybasedm odelcapableofsimulating exposuretoamixtureofanestrogenandanandrogen.MethodsModelFormulationWedevelopedtheHPGaxismodelforfemaleFHMby modifyingacomputationalmodelformaleFHM describedbyWatanabeetal.[20].Themodelsimulates timecontinuously,butitdoesnothaveaseasonalcomponent.Inthefollowing,wemainlyfocusontheunique formulationsand/orassumptionsinthismodelfor femaleFHMs. ThemodelforfemaleFHMscontainssixtissuecompartmentswhichrepresentorgansortissuesimportant forabsorption,distribution,metabolism,andelimination ofexogenousandendogenouschemicalsofinterest (Figure1).Thesixcompartmentsaregill,brain,gonad, liver,venousbloodand “ other ” .Inthearterialblood, theconcentrationsofbothfreeandboundchemicals areequaltothoseinthevenousbloodcompartment, unlessachemical(s)entersthebodythroughawater exposure.Asaresult,wedidnotcountarterialbloodas anindependentcompartment.Baseduponamassbalanceforeachchemicalofinterest,asetofcoupled ordinarydifferentialequationswereformulatedineach compartmentfollowingtheprinciplesofphysiologically basedpharmacokineticmodeling.Adetaileddescription ofthedifferentialequationscanbefoundinAdditional File1:DifferentialequationsusedintheHPGaxis model. Inthebrain,gonad,andlivercompartments,wesimulatedbothERandARdynamics.TheARcomponent wasnotincludedinthemodelformaleFHMpublished byWatanabeetal.[20].ERbindsestrogens(e.g.,E2and EE2),andboundERaffectstheproductionofVTG.AR bindsandrogens(e.g.,TandTB),andsubsequentlyregulatesbiochemicalprocessessuchastheproductionof gonadotropins[21].Ageneralmathematicalformulation ofligand-receptorbindingisshowninEquation1. d ( CiR, jVj) dt = k1 iR, jCi jCR jVj Kd iR, jk1 iR, jCiR, jV j (1) where, CiR, j(nmol/L)istheconcentrationofcompound i (e.g.T,TB,E2andEE2)boundtoitsreceptor incompartment j (e.g.brain,liver,gonad,andvenous blood); Vj(L)isthevolumeofcompartment j ; k1_iR, j(L/nmol/hr)istheassociationrateconstantofcompound i withitsreceptorincompartment j ; Ci,j(nmol/ L)istheconcentrationoffreecompound i incompartment j ; CR,j(nmol/L)istheconcentrationofunbound receptorofcompound i incompartment j ; Kd_iR, j(nmol/L)istheequilibriumdissociationconstantof compound i withitsreceptorincompartment j .GillInthegillcompartment,wedidnotsimulateanyproductionofproteins(e.g.,VTG),hormones(e.g.,Li etal BMCSystemsBiology 2011, 5 :63 http://www.biomedcentral.com/1752-0509/5/63 Page2of22

PAGE 3

Figure1 ConceptualmodeloftheHPGaxisinadultfemaleFHMs .TissuesinadultfemaleFHMsarecategorizedintosixcompartments:gill, brain,gonad,liver,venousblood,andother.Eachcompartmentisdefinedbyvolume,bloodflow,andpartitioncoefficient,andperforms multiplephysiologicalfunctions. Li etal BMCSystemsBiology 2011, 5 :63 http://www.biomedcentral.com/1752-0509/5/63 Page3of22

PAGE 4

luteinizinghormone,LH),orhormonereceptors(e.g,ER andAR).ERmRNAispresentinFHMgills,however, wedidnotsimulateERinthegillcompartmentbecause thegillexpressionofERisverylowcomparedtoother tissues[22].WesimulatedtheexposureoffemaleFHMs toTBand/orEE2inwater,andthegillcompartmentis wheretheexogenouschemicalsareabsorbed.Theconcentrationofeachchemicalinexposurewaterwas representedasafunction oftime.Then,equilibrium partitioningwasassumed, andtheFHMarterialblood concentrationwascalculatedfromthewaterconcentrationusinganequilibriumpartitioncoefficientassigned foreachchemical(Equation2).Inaddition,weassumed thatthegillcompartmentdidnotaccumulateanychemical(s). CArti= FWgil C i ,H20+ Fcar C Veni FWgil i bld+ Fcar (2) where,CArti(nmol/L)istheconcentrationofexogenouschemical i inarterialblood;FWgil(L/hr)isthevolumetricflowrateofwaterthroughthegills; Ci ,H20(nmol/L)istheconcentrationofexogenouschemical i (e.g.TBandEE2)inexposurewaterasafunctionof time; Fcar(L/hr)iscardiacoutput;andCVeni(nmol/L) istheconcentrationofexogenouschemical i invenous blood; li ,bldisthepartitioncoefficientforexogenous chemical i betweenbloodandwater.Partitioncoefficientsareoftendeterminedexperimentally,butwhena measuredormodel-estimatedvalueisunavailable,we calibrateittofittheexperimentaldata.BrainInthebraincompartment,threekeyassumptionswere made:(i)thedown-regulationofLH(gonadotropinII) synthesisbyboundAR[23,24];(ii)theup-regulationof LHsynthesisbyboundER[25];and(iii)thedown-regulationofARsynthesisbyfreeandrogens[26,27]. Inthebrain,androgenshaveanegativefeedbackon thesynthesisandreleaseofgonadotropinreleasinghormone(GnRH)[23],whichinturncontrolsthesynthesis ofgonadotropins.Toinvestigatehowandrogensmay regulateGnRH,wesearchedforanandrogenresponse element(ARE)inthepromoterregionsof gnrh genes.Duetoalackofinformationongenepromoter sequencesinFHM,weconductedthesearchinzebrafish( Daniorerio ),acyprinidfishcloselyrelatedto FHM.Wefoundthat gnrh promoterscontainseveral AREhalfsites(tgttct)[24].Thus,wepostulatedthat androgenshaveanegativecontrolonGnRHsynthesis mainlythroughboundAR.However,wedidnothave anymeasurementsofGnRHinFHMandGnRHwas notincludedinthemodel,soweformulatedadownregulationofLHsynthesisbyboundARinthemodel. Second,weassumedanup-regulationofLHsynthesis byboundERinthebraincompartment.Thisassumptionwasbaseduponobservationsofestrogenresponse elements(EREs)inthepromoterregionofthe lh gene andreportsofestrogen-stimulatedLHproductionin fish[25].Equation3describestheLHproductionrate inthebraincompartmentasafunctionofboundAR andER.Intheequation, PLH,brn(nmol/hr)istheproductionrateofLHinbrain; Pb_LH,brn(nmol/hr)isthe backgroundproductionrateofLHinbrain,whichwas formulatedasadiurnalcycle; CER_bd,brn(nmol/L)isthe totalconcentrationofboundERinbrain,whichequals thesumofE2-andEE2-boundERconcentrations; ru_LH,brn(nmol/L)isaninductionfactorforLHproductionbyboundER; CAR_bd,brn(nmol/L)isthetotal concentrationofboundARinbrain,whichequalsthe sumofT-andTB-boundARconcentrations; rd_LH, brn(nmol/L)isafactorforinhibitionofLHproduction byboundAR. PLH,brn= Pb LH,brn 1+ C ER bd,brn u LH,brn 1+ CAR bd,brn d LH brn (3) Thebraincompartmentisalsoveryimportantforthe regulationofARproduction[26,27].Inmammals(e.g. rats,mice,andhuman),ARmRNAinbrainisdownregulatedbyandrogens,suchasTanddihydrotestosterone[26,27],thoughlittleisknownaboutthe correspondingmechanisms .WesearchedforAREsin thepromoterregionofthe ar geneinzebrafish,butdid notfindanymatch.Hence,wepostulatedthatthe down-regulationofARmRNAbyandrogensisassociatedwithanon-genomicpathway[28],orassociated withcellfactorsotherthanthesolubleARsimulatedin ourmodel[29].Thus,weassumedadown-regulationof ARproductionbyfreeandrogensinthebraincompartment.WhenARsareproduced,somebindandrogens, someremainunbound,andothersdegrade.Basedupon amassbalanceforfreeAR,Equation4describesthe processesofARproduction,associationanddissociation withTorTB,anddegradation. d ( CAR free,brn) dt = PbgAR,brn 1+ ( CT ,brn+ CTB,brn) KAR,brn ( k1 TAR,brn CT ,brn CAR free,brn Kd TAR,brn k1 TAR,brn CTAR,brn) ( k1 TBAR,brn CTB,brn CAR free,br n Kd TBAR brn k1 TBAR brn CTBAR brn ) ke AR brn CAR free brn (4) where, CAR_free,brn(nmol/L)isthefreeARconcentrationinbrain;PbgAR,brn(nmol/L/hr)isthebackground productionrateofARinbrain; CT ,brn(nmol/L)isthe freeTconcentrationinbrain; CTB,brn(nmol/L)isthe freeTBconcentrationinbrain; KAR,brn(nmol/L)isan inhibitionrateconstantforARproductionbyfree TandTB; k1_TAR,brn(L/nmol/hr)istheassociationrateLi etal BMCSystemsBiology 2011, 5 :63 http://www.biomedcentral.com/1752-0509/5/63 Page4of22

PAGE 5

constantforTandAR;Kd_TAR,brn(nmol/L)isthe dissociationrateconstantforTboundtoAR; CTAR,brn(nmol/L)istheconcentrationoftheT-ARcomplexin brain; k1_TBAR,brn(L/nmol/hr)istheassociationrateof TBtoAR;Kd_TBAR,brn(nmol/L)isthedissociationrate constantforTBboundtoAR; CTBAR,brn(nmol/L)isthe concentrationoftheTB-ARcomplexinbrain;and ke_AR,brn(1/hr)istheeliminationrateforfreeAR.We includedtheinhibitionofLHbyboundARandthe inhibitionofARbyfreeandrogenstoaccountforthe U-shapeddose-responsecurvesforplasmaE2,Tand VTGconcentrationsobservedinfemaleFHMsexposed toTB[7].Theseassumptionsandmathematicalformulationsprovidedarobustfittotheavailabledata[7].An alternateformulationbaseduponbrainARandgonad ARwithdifferentbindingaffinitieswastriedfirst.However,becauseofalackofparameterinformationand evidenceforthebiologicalmechanism,weabandoned thisapproachforthepresentversion.Thepresent modelformulationmakesmoresensebiologically,andis simpler.GonadInthegonadcompartment,modificationstothemodel formulationsformaleFHMsinclude(i)absorptionof VTGintooocytes;and(ii)up-regulationofE2productionbyboundLH.TheabsorptionofVTGintooocytes wasformulatedasafirstorderkineticprocess.VTGis synthesizedintheliver[30],andcirculatestothe gonadswhereitistakenupviareceptor-mediatedendocytosisintooocytes,andthenprocessedintoyolkproteins[31].AlthoughthemolecularmechanismofVTG uptakeisknown,wedidnothavedatatodescribethis processquantitatively.Asaresult,afirstorderkinetic equationwithanassumedfirstorderrateconstantwas formulatedtorepresenttheprocess(Equation5). RVTG, g on= kVTG, g on CVTG, g on V g o n (5) where, RVTG,gon(nmol/hr)istheabsorptionrateof VTGintooocytesinthegonadcompartment; kVTG,gon(1/hr)istheabsorptionrateconstantforVTGinto oocytesinthegonadcompartment; CVTG,gon(nmol/L) istheconcentrationofVTGinthegonadcompartment; and Vgon(1/L)isthevolumeofthegonadcompartment. Secondly,wesimulatedanup-regulationofE2productionbyboundLHinthegonadcompartment.Itwas observedthatLHstimulatestheactivityandgene expressionofaromataseinthegonadsofteleosts[32]. Inourmodel,weformulatedtheregulationofE2productionasbeingproportionaltotheconcentrationof boundLHinthegonadcompartment(Equation6). PE2,gon= E2 LHLR,gon CLHLR,gon Vmaxaro,gon CT ,gon Kmaro, g on+ CT g on (6) where, PE2,gon(nmol/hr)istherateofE2production; rE2_LHLR,gon(L/nmol)isaninductionfactorofE2productionbyboundLH; CLHLR,gon(nmol/L)isthe concentrationofboundLH;Vmaxaro,gon(nmol/hr)is themaximumrateofE2productionbygonadaromatase;Kmaro,gon(nmol/L)istheMichaelis-Mentenconstantforgonadaromatase; CT ,gon(nmol/L)isthe concentrationofT.LiverInthelivercompartment,formulationsincludingER auto-regulationandbound-ER-stimulatedVTGproductionarethesameasthosedescribedbyWatanabeetal. [20],exceptthatweaddedligand-receptorbindingofT andTBtotheAR.VenousbloodBesidesE2andT,wesimulatedtheassociationanddissociationprocessesofEE2tosteroid-bindingproteins (SBPs)inthevenousbloodcompartment.Thereiscontradictoryinformationaboutthebindingaffinitiesof EE2toSBPsinfish.ComparedtoE2,somefishspecies suchaschannelcatfish( Ictaluruspunctatus )andzebrafish( Daniorerio )havehighbindingaffinityofEE2to SBPs[33,34],whileother fishspeciessuchasArctic charr( Salvelinusalpinus )havealowbindingaffinity [35].Todate,bindingaffinitymeasurementsofEE2to SBPsinFHMhavenotbeenmade.Watanabeetal.[20] didnotincludethebindingprocessofEE2toSBPsin blood.IntheirmodellingworkformaleFHMs,thetotal concentrationofSBPswasassumedtobe20nmol/L baseduponameasurementinhumanmales[36].Such alowvaluehaslittleeffectonfreeplasmaEE2concentrationormodelperformance.However,inourmodel forfemaleFHMs,weassumedthetotalconcentrationof SBPstobe400nmol/L[36,37]baseduponSBPmeasurementsinfemalespottedseatrout( Cynoscionnebulosus )[37]andinhumanfemales[36].Consequently,a largeamountofEE2couldbeboundbySBPsinblood, whichwouldaffectthetotalconcentrationofEE2inthe plasma.Therefore,weincludedthebindingprocessof EE2toSBPsinthismodel,andformulateditusing Equation1.OtherInourmodel,the ‘ Other ’ compartmentiswhereeliminationofexogenousandendogenouschemicalsand proteinsoccur.BesidesE2,EE2,T,VTG,andLH (includedinWatanabeetal.[20]),weaddedafirst orderkineticequationtodescribetheeliminationofTB, andthefirstordereliminationrateconstantwas assumedtobethesameasthatofEE2(Equation7). RTB oth= ke TB oth CTB oth Vot h (7) where RTB,oth(nmol/hr)istheeliminationrateofTB intheOthercompartment; ke_TB,oth(1/hr)istheLi etal BMCSystemsBiology 2011, 5 :63 http://www.biomedcentral.com/1752-0509/5/63 Page5of22

PAGE 6

eliminationrateconstantforTBintheOthercompartment; CTB,oth(nmol/L)istheconcentrationofTBin theOthercompartment;and Voth(1/L)isthevolumeof theOthercompartment.ExperimentalDataTocalibratemodelparametersandtoevaluatemodel predictions,weuseddatafromunexposed,TB-exposed, andEE2-exposedadultfemaleFHMsfrom18different studies.Allstudieswereconductedwithsexuallymature (fivetosevenmonthold)femaleFHMs.Chemicalexposureswereconductedinthelaboratoryunderoptimal conditionsforFHMreproduction.Forexample,the temperaturewas25C,photoperiodwas16/8hr(lightto dark),andfoodwasnotlimited.Undersuchconditions, FHMscanremaininreproductiveconditionandspawn yeararound.Foreachfish,physiologicalparameters, includingbodyweight(BW),gonadosomaticindex (GSI),andhepatosomaticindex(HSI),wereinputinto themodel.Forallexperimentaldatausedinmodelcalibrationorvalidation,whenanymeasurementsofBW, GSI,orHSIweremissing,weusedthemediansofmeasuredBW,GSI,orHSI,respectively[38].Ideally,ifall experimentaldatahadbeenavailablewhenwestartedto developthemodelin2006,wewouldhaverandomly selectedasubsetofdatafromeachexperimentfor modelcalibrationandusedtheremainderformodel evaluation.However,severaloftheexperimentalstudies wereconductedwhilethem odelwasbeingdeveloped. Thus,weuseddataastheybecameavailable.Thefollowingsummarizestheexperimentaldataandhowthe datawereused. Thereproductiveendpointdatainunexposed(control)adultfemaleFHMswereobtainedfromanearlier paperbyWatanabeetal.[38].Inatotalof170female FHMs,thedataincludemeasurementsofplasmaE2,T, andVTGconcentrations;allmeasurementsweremade insomefish,andinothersonlyasubsetofendpoints (e.g.,plasmaE2andVTGconcentrations)weremeasured.Werandomlysplitthedata;thefirst75records wereusedtocalibrateourmodel;theremaining95 recordswereusedinmodelvalidation. ExperimentaldatafromTB-exposedadultfemale FHMswereobtainedfromthreestudies:(i)aflowthroughwaterexposuretonominalconcentrationsof 0.005,0.05,0.5,5.0,and50 gTB/Lfor21daysby Ankleyetal.[7];(ii)astaticexposuretonominalconcentrationsof0.05,0.5,5 gTB/Lfor48hoursbyGarcia-Reyeroetal.[39];and (iii)aflow-throughwater exposuretonominalconcentrationsof0.05and0.5 g TB/LinadultfemaleFHMsforeightdays,followedby aneight-daydepurationdes cribedbyEkmanetal.[40]. InAnkleyetal.[7],12femaleFHMswereexposedin eachtreatmentgroup.Onthe21stdayofexposure,all FHMsweresacrificed;plasmaconcentrationsofE2,T, andVTGweremeasuredineachfish.InGarcia-Reyero etal.[39]eightFHMswereexposedineachtreatment group.Aftera48-hourexposure,thefishweresacrificed.Foreachtreatmentgroup,plasmaE2concentrationsweremeasuredineachoffourfish,andplasma VTGconcentrationsweremeasuredineachofthefour remainingfish.TheconcentrationsofVTGandE2were notmeasuredinthesamefishbecauseDr.Orlando ’ s laboratorymeasuredE2andDr.Denslow ’ slaboratory measuredVTG.InEkmanetal.[40],64FHMswere exposedtoTBineachtreatmentgroup.Onthe1st,2nd, 4th,and8thdayofexposureandthe1st,2nd,4th,and8thdayofdepuration(testdays9,10,12,and16),foreach treatmentgroup,eightFHMsweresacrificedtomeasure plasmaE2andVTGconcentrationsineachfish.Data fromAnkleyetal.[7]wereusedtocalibrateourmodel, anddatafromGarcia-Reyeroetal.[39]andEkmanet al.[40]wereusedtoevaluateourmodelpredictions. VTGplasmaconcentrationsinadultfemaleFHMs exposedtoEE2wereobtainedfromthreestudies:(i)a flow-throughwaterexposuretonominalconcentrations of10or100ngEE2/LinadultfemaleFHMsforeight days,followedbyaneight-daydepuration[41];(ii)a flow-throughwaterexposuretoanominalconcentration of0.5,1.5,and4.5ngEE2/LinadultfemaleFHMsfor 21daysbyLazorchaketal.[42];and(iii)aflow-through waterexposuretoanominalconcentrationof1.5ng EE2/LinadultfemaleFHMsfor21daysbyBrianetal. [43].InEkmanetal.[41],foreachtreatmentgroupand eachsamplingtime,eightFHMsweresacrificedtomeasureplasmaVTGconcentrationineachfish.Sampling occurredonthe1st,4th,and8thdayofexposuretoEE2, andthe8thdayofEE2depuration(testday16).In Lazorchaketal.[42],28FHMsineachofthetreatment groups(0.5,1.5,and4.5ngEE2/L)weresacrificedto measureplasmaVTGconcentrationineachfishonthe 21stday.InBrianetal.[43],fourFHMsweresacrificed tomeasureplasmaVTGconcentrationineachfishon the21stdayafterexposureto1.5ngEE2/L.Asopposed tothethreeTBandtwoEE2studieswhichdidnotuse carriersolvents,Brianetal.[39]usedN,N-dimethylformamide,DMF,asachemicalcarrierforEE2.Datafrom Ekmanetal.[41]wereusedtocalibrateourmodel,and datafromLazorchaketal.[42],andBrianetal.[43] wereusedtoevaluateourmodelpredictions.ModelCalibrationIntotal,ourmodelcontains123inputparameters,such asvolumeandbloodflowratesofeachcompartment, chemicalequilibriumpartitioncoefficients,ligand-receptorassociationanddissociationrateconstants,and kineticrateconstantsforeachbiochemicalreaction. Theparameterswerefixedwithknownvalues,orLi etal BMCSystemsBiology 2011, 5 :63 http://www.biomedcentral.com/1752-0509/5/63 Page6of22

PAGE 7

calibratedusingexperimentaldatacollectedinadult femaleFHMs.Intotal,97modelparameterswerefixed withvaluesobtainedfrompublishedliteratureormeasuredforthisstudy(Table1).Theremaining26model parameterswerecalibratedusingMarkovChainMonte Carlosimulation[44-47],whi chrequiresthedefinition ofpriordistributionsforeachparameterbeing calibrated. Ofthe26calibratedmodelparameters,17weresensitivemodelparameterswit hlittleornoinformation availableintheopenliterature(Table2).Vagueprior distributionswereusedfor these17modelparameters. Forexample,wecouldnotfindapublishedvaluefor thedissociatio nconstantofE2bindingtoERinFHM brainspecifically( Kd_E2ER,brn).Dennyetal.[15] reportedthatthedissociationconstantofE2bindingin femaleFHMlivercytosolis8.6nmol/L.Asaresult,we assignedalognormaldistributionwithageometric meanof8.6andageometricstandarddeviationof three,whichcorrespondstoacoefficientofvariation equalto1.5.Whennodatawereavailableintheopen literature,weassignedauniformorlog-uniformprior distributionwithalargerangeboundedbybiological plausibility.Forexample,weknowthattheEE2partition coefficientforbloodtowaterisaround300[20],and thusfixedtheparametervalueat300.However,there werenopublisheddataforthebloodtowaterTBpartitioncoefficient( lTB,bld).Therefore,weassignedavague priordistributionfor lTB,bld,whichisalog-uniform distributionwithalowerboundofoneandanupper boundof1000. Theremainingnineparame terswereerrorvariances forplasmaE2,T,andVTGconcentrationsinunexposed,TB-exposed,andEE2-exposedFHMs.We assumedthattheerrorsfoll owedalognormaldistributionwithgeometricmeansequaltothemodel-predicted concentrationsofplasmaE2,T,andVTG,respectively. Thevariancewasestimatedbydividingtheexperimental dataintothreedifferentgroups:unexposed,TB-exposed, andEE2-exposedFHM,respectively;andtheerrorvariancesofthethreereproductiveendpointsforeach groupwereestimated[45,47].Foreachofthenineerror variances,weassignedanInverseGammapriordistributionbaseduponanaturallogarithmtransformationof themeasuredplasmaE2,TandVTGconcentrations [20].AnInverseGammapriordistributionistheconjugateofanormaldistribution[46],whichsimplifiesthe modelcomputations. ToperformtheMarkovChainMonteCarlosimulations,weusedMCSim[48],asoftwarepackagefreely availableonlinehttp://direc tory.fsf.org/math/mcsim. html.FourindependentMarkovchainswithrandom seedswererunfor20,000iterations.Foreachofthe fourchains,wesavedthelast10,000iterations,and extractedonesetofmodelparametersoutofevery10. Foreachcalibratedmodelparameter,convergencewas evaluatedusingthe1,000iterationsfromeachchainand apotentialscalereductioncriterion(Rhat)[46].AcceptablevaluesofRhatrangedfrom1.0to1.2;thisisessentiallyaratioofthecalibratedmodelparametervariance betweenthefourMarkovchainstothevariancewithin achain.ModelEvaluationWeevaluatedthepredictiveabilityofourmodelby simulatingreproductiveendpoints(i.e.,plasmaconcentrationsofE2,T,orVTG)fromindependentstudies. The1,000iterationsobtainedfromeachMarkovchain werepooled,andthe4,000setsofparametervalues weretreatedasapoolofadultfemaleFHMs.Werandomlysampled n (numberoffishinastudy)parameter setstorepresentthe n fishusedinthestudy,andsimulatedthereproductiveendpointsmeasuredforeachfish. Thedetailedsimulationpro ceduresfollowedthemethodsdescribedbyWatanabeetal.[20]. Aftercompleting n simulationsforastudy,wepredicted eachreproductiveendpointbaseduponourlognormal errormodel.AsdescribedintheModelCalibrationsection,errorvarianceswereestimatedduringmodelcalibration.Usingthemodelpredictionandtheestimated varianceastwoparameters,werandomlysampledfrom thelognormaldistributionforeachendpointineachfish. Thesampledvalueswerecomparedwithexperimental measurements.PredictionofUnmeasuredReproductiveEndpointsToobserveEDCeffectsonunmeasuredcomponentsof theHPGaxis(e.g.,ER,AR,andLH),andtoobservethe effectsonreproductiveendpointsbyamixtureofTB andEE2,wedidthreeextrasimulations.Wesimulated liverERconcentration,brainARconcentration,and plasmaE2,T,VTG,andLHconcentrationsasafunctionoftimeinadultfemaleFHMsexposedto15ng TB/L,10ngEE2/L,oramixtureof15ngTB/Land 10ngEE2/Lfor48hours,respectively.TheconcentrationsofTBandEE2werechosenbecausetheyare environmentallyrelevant[12,18,19].Inallthreesimulations,weusedthereported[38]medianbodyweight, GSI,andHSIvaluesinadultfemaleFHMsasinput parameters.ResultsandDiscussionModelCalibrationAgoodfitoftheexperimentaldatawasobtainedby runningfourMarkovchainsusingtheMarkovChain MonteCarlosimulations.Fo rthe26calibratedmodel parameters,thefourMarkovchainsconvergedwithin 20,000iterations.ThemodelcalibrationspeedisaroundLi etal BMCSystemsBiology 2011, 5 :63 http://www.biomedcentral.com/1752-0509/5/63 Page7of22

PAGE 8

Table1Modelparameterstreatedasconstants( n =97)ParameterdescriptionSymbolsValueReference BodyweightaBodyWt0.0016(kg)Watanabeetal.[38] VolumetricwaterflowingthroughgillsFWgil10.6BodyWt0.75(L/hr)Nicholsetal.[57] CardiacoutputFcar2.06BodyWt0.75(L/hr)Nicholsetal.[57] Percentageofbraintobodyweight(BSI)Pbrn1.18MeasuredbyD.Villeneuve Percentageofgonadstobodyweight(GSI)bPgon11Watanabeetal.[38] Percentageoflivertobodyweight(HSI)cPliv3.0Watanabeetal.[38] PercentageofgillstobodyweightPgil1.67Nicholsetal.[58] PercentageofvenousbloodtobodyweightPven2.59Robinsonetal.[59] Nicholsetal.[58] Percentageof “ other ” tobodyweightPoth=100-Pbrn-Pgon-Pliv-Pgil-PvenWatanabeetal.[20] Fractionofbloodflowinbraintocardiacoutput F brn F car 0.036 P brn 0.036 Pbrn+0.036 Pgon+0.024 Pliv+0.007 Poth Nicholsetal.[58] Fractionofbloodflowingonadtocardiacoutput F gon Fcar 0.036 Pgon 0.036 Pbrn+0.036 Pgon+0.024 Pliv+0.007 Poth Nicholsetal.[58] Fractionofbloodflowinlivertocardiacoutput F liv Fcar 0.024 P liv 0.036 Pbrn+0.036 P g on+0.024 Pliv+0.007 Poth Nicholsetal.[58] Fractionofbloodflowin “ other ” tocardiacoutput F oth F car 0.007 Poth 0.036 Pbrn+0.036 P g on+0.024 Pliv+0.007 Poth Nicholsetal.[58] FractionofplasmainvenousbloodFplasma,ven0.45MeasuredbyK.Kroll TotalconcentrationofestrogenreceptorsinbrainCER,brn14.3(nmol/Ltissue)PlowchalkandTeaguarden [60] TotalconcentrationofestrogenreceptorsingonadCER,gon29(nmol/Ltissue)PlowchalkandTeaguarden [60] TotalconcentrationofLHreceptorsingonadCLR,gon2.0(nmol/Ltissue)Miwaetal.[61] TotalconcentrationofSBPinbloodCSBP,ven400(nmol/Lblood)LaidleyandThomas[37] TeeguardenandBarton[36] TotalconcentrationofARingonadCAR,gon1.05(nmol/Ltissue)SperryandThomas[62] TotalconcentrationofARinliverCAR,liv=CAR,gonassumed AssociationrateofE2toestrogenreceptorinbrain k1_E2ER,brn0.743Murphyetal.[63] DissociationconstantofE2toestrogenreceptoringonad Kd_E2ER,gon= Kd _E2ER,brnassumed AssociationrateofE2toestrogenreceptoringonad k1_E2ER,gon= k1 _E2ER,brnassumed DissociationconstantofE2toestrogenreceptorinliver Kd_E2ER,liv= Kd _E2ER,brnassumed AssociationrateofE2toestrogenreceptorinliver K1_E2ER,liv= k1 _E2ER,brnassumed DissociationconstantofEE2toestrogenreceptorinbrain Kd_EE2ER,brn= Kd _E2ER,brn/RBAEE2_E2Dennyetal.[15] AssociationrateofEE2toestrogenreceptorinbrain k1_EE2ER,brn= k1 _E2ER,brnassumed DissociationconstantofEE2toestrogenreceptoringonad Kd_EE2ER,gon= Kd _EE2ER,brnassumedLi etal BMCSystemsBiology 2011, 5 :63 http://www.biomedcentral.com/1752-0509/5/63 Page8of22

PAGE 9

Table1Modelparameterstreatedasconstants( n =97) (Continued)AssociationrateofEE2toestrogenreceptoringonad k1_EE2ER,gon= k1 _EE2ER,brnassumed DissociationconstantofEE2toestrogenreceptorinliver Kd_EE2ER,liv= Kd _EE2ER,brnassumed AssociationrateofEE2toestrogenreceptorinliver k1_EE2ER,liv= k1 _EE2ER,brnassumed DissociationconstantofTtoandrogenreceptorinbrain Kd_TAR,brn3(nmol/L)SperryandThomas[62] AssociationrateofTtoandrogenreceptorinbrain k1_TAR,brn0.08(L/nmol/hr)SperryandThomas[62] DissociationconstantofTtoandrogenreceptoringonad Kd_TAR,gon= Kd _TAR,brnassumed AssociationrateofTtoandrogenreceptoringonad k1_TAR,gon= k1 _TAR,brnassumed DissociationconstantofTtoandrogenreceptorinliver Kd_TAR,liv= Kd _TAR,brnassumed AssociationrateofTtoandrogenreceptorinliver k1_TAR,liv= k1 _TAR,brnassumed DissociationconstantofTBtoandrogenreceptorinbrain Kd_TBAR,brn= Kd _TAR,brn/RBATB_TWilsonetal.[64] AssociationrateofTBtoandrogenreceptorinbrain k1_TBAR,brn= k1 _TAR,brnassumed DissociationconstantofTBtoandrogenreceptoringonad Kd_TBAR,gon= Kd _TBAR,brnassumed AssociationrateofTBtoandrogenreceptoringonad k1_TBAR,gon= k1 _TBAR,brnassumed DissociationconstantofTBtoandrogenreceptorinliver Kd_TBAR,liv= Kd _TBAR,brnassumed AssociationrateofTBtoandrogenreceptorinliver k1_TBAR,liv= k1 _TBAR,brnassumed DissociationconstantofE2toSBPinblood Kd_E2SBP,ven3.13(nmol/L)Murphyetal.[63] AssociationrateofE2toSBPinblood k1_E2SBP,ven5.6687(L/nmol/hr)Murphyetal.[63] DissociationconstantofTtoSBPinblood Kd_TSBP,ven4.89(nmol/L)Murphyetal.[63] AssociationrateofTtoSBPinblood K1_TSBP,ven5.6687(L/nmol/hr)Murphyetal.[63] DissociationconstantofEE2toSBPinblood Kd_EE2SBP,ven0.58(nmol/L)Miguel-Queraltand Hammond[34] AssociationrateofEE2toSBPinblood k1_EE2SBP,ven5.6687(L/nmol/hr)Murphyetal.[63] DissociationconstantofLHtoLHreceptoringonad Kd_LHLR,gon2.9(nmol/L)Crimetal.[65] AssociationrateofLHtoLHreceptoringonad k1_LHLR,gon0.2(L/nmol/hr)Watanabeetal.[20] ScalingcoefficientofVmaxofTproductioningonad(=Vmax/bodyweight0.75)sc_VmaxScc, gon1.1e+05(nmol/hr/kgbodyweight)Kashiwagietal.[66]; ShikitaandHall[67] K0.5ofTproductioningonadK0.5Scc,gon190(nmol/L)ShikitaandHall[67] InhibitionconstantofTproductionbyboundERKT0.016Watanabeetal.[20] KmofE2productioningonadKmaro,gon9.6(nmol/L)Zhaoetal.[68] ConcentrationofmicrosomalproteiningonadsDmp,gon3100(mg/L)MeasuredbyD.Villeneuve RatiobetweentheconcentrationsofmicrosoamlproteiningonadsandbrainRhomp0.174MeasuredbyD.Villeneuve ScalingcoefficientofVmaxofE2productioninbrain(=Vmax/massofmicrosomalprotein inbrain) sc_Vmaxaro, brn=4.6sc_Vmaxaro,gonZhaoetal.[68] KmofE2productioninbrainKmaro,brn9.6(nmol/L)Zhaoetal.[68] ConcentrationofmicrosomalproteininbrainDmp,brn=Dmp,gon/RhompMeasuredbyD.Villeneuve RatiobetweenconcentrationsofSTARandboundLRingonadsRhoSTAR,gon1assumedLi etal BMCSystemsBiology 2011, 5 :63 http://www.biomedcentral.com/1752-0509/5/63 Page9of22

PAGE 10

Table1Modelparameterstreatedasconstants( n =97) (Continued)RateconstantforVtguptakeintooocytes kvtg,gon0.05assumed K0.5ofVtgproductioninliverproductionK0.5Vtg,liv1.0(nmol/L)Watanabeetal.[20] EliminationrateconstantforER inthelivercompartment ke_ER,liv0.01(1/hr)Murphyetal.[63] EliminationrateconstantforAR inthebraincompartment ke_AR,brn0.01(1/hr)Assumed EliminationrateconstantforLH inthe “ other ” compartment ke_LH,oth0.1(1/hr)TeeguardenandBarton[36] EliminationrateconstantforE2inthe “ other ” compartment ke_E2,oth0.1(1/hr)TeeguardenandBarton[36] EliminationrateconstantforT inthe “ other ” compartment ke_T,oth0.1(1/hr)TeeguardenandBarton[36] EliminationrateconstantforEE2inthe “ other ” compartment ke_EE2,oth0.1(1/hr)TeeguardenandBarton[36] EliminationrateconstantforTB inthe “ other ” compartment ke_TB,oth0.1(1/hr)TeeguardenandBarton[36] EliminationrateconstantforVtg inthe “ other ” compartment ke_Vtg,oth0.001(1/hr)TeeguardenandBarton[36] PartitioncoefficientofLH (braintoblood) lLH,brn1TeeguardenandBarton[36] PartitioncoefficientofLH (gonadtoblood) lLH,gon1TeeguardenandBarton[36] PartitioncoefficientofLH (livertoblood) lLH,liv1TeeguardenandBarton[36] PartitioncoefficientofLH ("other ” toblood) lLH,oth1TeeguardenandBarton[36] PartitioncoefficientofVTG (braintoblood) lVTG,brn1TeeguardenandBarton[36] PartitioncoefficientofVTG (gonadtoblood) lVTG,gon1TeeguardenandBarton[36] PartitioncoefficientofVTG (livertoblood) lVTG,liv1TeeguardenandBarton[36] PartitioncoefficientofVTG ("other ” toblood) lVTG,oth1TeeguardenandBarton[36] PartitioncoefficientofEE2(bloodtowater) lEE2,bld300Watanabeetal.[20] PartitioncoefficientofEE2(braintoblood) lEE2,brn1TeeguardenandBarton[36] PartitioncoefficientofEE2(gonadtoblood) lEE2,gon1TeeguardenandBarton[36] PartitioncoefficientofEE2(livertoblood) lEE2,liv3Watanabeetal.[20]Li etal BMCSystemsBiology 2011, 5 :63 http://www.biomedcentral.com/1752-0509/5/63 Page10of22

PAGE 11

Table1Modelparameterstreatedasconstants( n =97) (Continued)PartitioncoefficientofEE2("other ” toblood) lEE2,oth1TeeguardenandBarton[36] PartitioncoefficientofE2(bloodtowater) lE2,bld300Watanabeetal.[20] PartitioncoefficientofE2(braintoblood) lE2,brn1TeeguardenandBarton[36] PartitioncoefficientofE2(gonadtoblood) lE2,gon1PlowchalkandTeeguarden [60] PartitioncoefficientofE2(livertoblood) lE2,liv3Watanabeetal.[20] PartitioncoefficientofE2("other ” toblood) lE2,oth1PlowchalkandTeeguarden [60] PartitioncoefficientofT (braintoblood) lT,brn1BartonandAndersen[69] PartitioncoefficientofT (gonadtoblood) lT,gon1BartonandAndersen[69] PartitioncoefficientofT (livertoblood) lT,liv1BartonandAndersen[69] PartitioncoefficientofT ("other ” toblood) lT,oth1BartonandAndersen[69] PartitioncoefficientofTB (braintoblood) lTB ,brn1BartonandAndersen[69] PartitioncoefficientofTB (gonadtoblood) lTB,gon1BartonandAndersen[69] PartitioncoefficientofTB (livertoblood) lTB,liv1BartonandAndersen[69] PartitioncoefficientofTB ("other ” toblood) lTB,oth1BartonandAndersen[69]a,b,andcwereassignedwithmeasuredvaluesineachfish;thedefaultvalueswereusedonlywhenmeasureddataweremissing.Li etal BMCSystemsBiology 2011, 5 :63 http://www.biomedcentral.com/1752-0509/5/63 Page11of22

PAGE 12

Table2Summarystatisticsforpriorandposteriordistributionsofcalibratedmodelparameters( n =26)ParameterdescriptionSymbolsPrior Distribution (P1,P2)aReferenceMean(Posterior Distribution) Median(Posterior Distribution) 95%ConfidenceInterval (PosteriorDistribution) PartitioncoefficientofTB(bloodtowater) lTB,bldLoguniform (1,1.0E+3) Assumed7.477.47(5.96,8.93) DissociationconstantofE2bindingtoERin brain(nmol/L) Kd_E2ER,brnLognormal (8.6,3) Dennyetal.[15]1.121.08(0.71,1.87) RelativebindingaffinityofEE2toE2forERbindingRBAEE2_E2Lognormal (1.66,3) Dennyetal.[15] Galeetal.[14] 3.241.64(0.030,16.79) RelativebindingaffinityofTBtoTforARbindingRBATB_TLognormal (6.03,3) Wilsonetal.[64]5.254.84(2.29,10.76) InhibitionfactorforLHproductionbybound AR(nmol/L) rd_LH,brnLogUniform (0.01,1.0E+3) Assumed0.110.10(0.042,0.21) InductionfactorforLHproductionbybound ER(nmol/L) ru_LH,brnLogUniform (0.01,1.0E+3) Assumed238138(4.23,864) HillcoefficientforTproduction nTLognormal (1.8,3) Murphyetal.[63]1.031.01(0.93,1.19) ProportionalityconstantrelatingcholesteroltoStAR rChol,go nLoguniform (1,5.0E+3) Artemenketal.[70]2.371.83(1.04,6.69) ScalingcoefficientofVmaxforE2productionin gonad(nmol/hr/mgmicro-protein) sc_Vmaxaro,gonLoguniform (2.3E-5,0.23) Zhaoetal.[71]1.56E-31.53E-3(1.15E-3,2.12E-3) InductionfactorofE2productionbybound LH(L/nmol) rE2_LHLR,gonLoguniform (0.1,100) assumed79.8482.79(42.61,99.15) ScalingcoefficientofVmaxforVtgproductionin liver(=Vmax/BodyWeight0.75)(nmol/hr/kg0.75) sc_VmaxVtg,livLoguniform (1,1.0E+4) Watanabeetal.[20]175169(121,271) HillcoefficientofVtgproductioninliver nVTGUniform (1,10) Assumed2.882.87(1.97,3.87) ERbackgroundproductionrateinliver(nmol/L/hr)PbgER,livLoguniform (5.0E-5,0.5) assumed0.120.12(0.084,0.17) InductionrateconstantforERproductionin liver(1/hr) kER,livLognormal (0.08,3) Watanabeetal.[20]0.0270.025(5.73E-3,0.065) ARbackgroundproductionrateinbrain (nmol/L/hr) PbgAR,brnLoguniform (5.0E-5,0.5) assumed0.0120.012(9.1E-3,0.015) InhibitionfactorofARproductionbyfree androgens(nmol/L) KAR,brnLoguniform (5E-4,5) assumed3.954.08(2.15,4.95) MagnitudeofLHproduction(nmol/hr)MagLHLoguniform (2.7E-7,2.7E-3) Schulzetal.[72]8.86E-68.75E-6(6.29E-6,1.20E-5) ErrorvarianceofplasmaE2concentrationinnatural logspaceforunexposedfemaleFHMs Var_Ln_CE2tot_pla_ngmlInverseGamma (2,1.19) Boisetal.[73]0.520.51(0.38,0.73) ErrorvarianceofplasmaTconcentrationinnatural logspaceforunexposedfemaleFHMs Var_Ln_CTtot_pla_ngmlInverseGamma (2,0.53) Boisetal.[73]0.480.47(0.34,0.69)Li etal BMCSystemsBiology 2011, 5 :63 http://www.biomedcentral.com/1752-0509/5/63 Page12of22

PAGE 13

Table2Summarystatisticsforpriorandposteriordistributionsofcalibratedmodelparameters( n =26) (Continued)ErrorvarianceofplasmaVTGconcentrationin naturallogspaceforunexposedfemaleFHMs Var_Ln_CVTG_pla_mgmlInverseGamma (2,5.31) Boisetal.[73]0.490.48(0.35,0.68) ErrorvarianceofplasmaE2concentrationinnatural logspaceforTB-exposedfemaleFHMs Var_Ln_CE2tot_pla_ngmlInverseGamma (2,1.19) Boisetal.[73]0.700.69(0.48,1.03) ErrorvarianceofplasmaTconcentrationinnatural logspaceforTB-exposedfemaleFHMs Var_Ln_CTtot_pla_ngmlInverseGamma (2,0.53) Boisetal.[73]0.400.39(0.27,0.60) ErrorvarianceofplasmaVTGconcentrationin naturallogspaceforTB-exposedfemaleFHMs Var_Ln_CVTG_pla_mgmlInverseGamma (2,5.31) Boisetal.[73]5.865.72(3.98,8.60) ErrorvarianceofplasmaE2concentrationinnatural logspaceforEE2-exposedfemaleFHMs Var_Ln_CE2tot_pla_ngmlInverseGamma (2,1.19) Boisetal.[73]1.430.81(0.22,6.31) ErrorvarianceofplasmaTconcentrationinnatural logspaceforEE2-exposedfemaleFHMs Var_Ln_CTtot_pla_ngmlInverseGamma (2,0.53) Boisetal.[73]0.590.34(0.10,2.76) ErrorvarianceofplasmaVTGconcentrationin naturallogspaceforEE2-exposedfemaleFHMs Var_Ln_CVTG_pla_mgmlInverseGamma (2,5.31) Boisetal.[73]0.730.71(0.51,1.03)aDefinitionofP1andP2ofpriordistributions.Loguniform:P1=minimumofthesamplingrangeinnaturalspace;P2=maximumofthesamplingrangeinnat uralspace.Lognormal:P1=geometricmean (exponentialofthemeaninlog-space);P2=geometricstandarddeviation(exponentialofthestandarddeviationinlog-space,strictlysuperiorto 1).Uniform:P1=minimumofthesamplingrangeinnaturalspace; P2=maximumofthesamplingrangeinnaturalspace.Inversegamma:P1=shape;P2=scale(bothoftheparametersarestrictlypositive).Li etal BMCSystemsBiology 2011, 5 :63 http://www.biomedcentral.com/1752-0509/5/63 Page13of22

PAGE 14

12hoursper100iterations.TheRhatvaluesofthe26 parameterswerealllessthan1.2,indicatingacceptable convergence.Figure2plotsthetrajectoriesofthefour MarkovchainsfortherelativebindingaffinityofTBto T(RBATB_T),whichisoneofthe26calibratedmodel parameters.Thefourchainsforthisparametermixed wellandconvergedwithin20,000iterations. Table2includesthesummarystatisticsofposterior distributionsforthe26calibratedparameters.Theposteriordistributionsummarystatisticsarebasedonthe 4,000iterations,1,000iterationsfromeachofthefour chains.Inbrief,ourmodelimprovedestimatesof23 modelparameters.Ofthe26parameters,21had95% confidenceintervals(CIs)narrowerthanthoseoftheir priordistributions;threeparameters(i.e.,RBAEE2_E2, errorvariancesofE2andTforEE2-exposedFHMs)had 95%CIssimilartotheirpriordistributionCIs;andtwo parameters(i.e.,errorvariancesofVTGinunexposed andEE2-exposedFHMs)had95%CIsslightlydifferent fromtheirpriordistributionCIs.Fortheerrorvariance ofVTGinunexposedFHMs,theupper95%confidence limitoftheposteriordistributionwas72%ofthe2.5thpercentileofitspriordist ribution.FortheerrorvariancesofVTGinEE2-exposedFHMs,the95%CIsof thepriorandposteriordistributionsoverlappedwith eachother.Buttheupper95%confidencelimitofthe posteriordistributionwasonly5%ofthe97.5thpercentileofitspriordistribution.Theselargedifferences occurredmainlybecausetheassignedpriordistributions fortheerrorvarianceswerebaseduponexperimental datavariances,whichdonotrepresenttheerrors exactly,butweregoodstartingpointsforthemodel calibration. Itisimportanttonotethattheposteriordistributions listedinTable2areconditionaluponfixedmodelparameters(Table1),priordistributionsofcalibratedparameters(Table2),andthedatasetsusedincalibration. Anychangeinthesecomponentsmayleadtodifferent posteriordistributionsofthecalibratedparameters.In thisstudy,wecarefullysearchedtheliteraturetoassign ourmodelparameterswithme aningfulandphysiologicallybasedvaluesorpriordistributions.Asadditional databecomeavailable,ourmodelcouldbere-calibrated tobetterdefineparameterposteriordistributions.ModelEvaluationInthisstudy,ourmodelwasusedtosimulateexperimentsranginginlengthfrom48hrsto21days.The modeliscapableofsimulatinglongerperiodsoftime, butitdoesnotincludeaseasonalcomponent.Thatis, theFHMssimulatedinourstudywereheldunder laboratoryconditionsoptimalforreproductionand spawnyearround.Themodelcouldbemodifiedto accountfortheeffectofseasonsuponreproductionin ordertosimulateconditionsexperiencedbywildfish.PredictionsforPlasmaE2,T,andVTGConcentrationsin UnexposedFHMsWiththecalibratedmodelparameters,wesimulated plasmaconcentrationsofE2,T,andVTGin95unexposed adultfemaleFHMs[38].Figure3showsacomparisonof modelpredictionsandexperimentaldata.Forallthree endpoints,themeanandmedianmodelpredictionswere within80to150%ofthemeasuredmeansandmedians, respectively.Model-predic ted95%CIsencompassedthe meanandmedianmeasurements,andmodel-predicted meansandmedianswerewithinthe95%CIsofthemeasureddata.Thus,inunexposedadultfemaleFHMs,our modelsuccessfullypredictedallthreeendpoints(Figure 3).Thisisanimprovement comparedtothemodelfor maleFHMs[20],whichpredictedthemediansofthemeasureddata,butunder-predictedthevariancesforallthree endpoints.Includinginformationfromthelognormal errormodelenabledbetterpredictionsofbothmedians andvariancesofthemeasureddata.PredictionsforplasmaE2andVTGconcentrationsinTBexposedFHMsFigure4comparesthemeasuredandmodel-predicted plasmaVTG(Figure4A)andE2(Figure4B)concentrationsinfemaleFHMsexposedto0,0.05,0.5,and5 g TB/Lfor48hours[39].Ourmodelpredictionsfollowedthegeneraltrendofthemeasureddata,andthe modelpredictionrangeoverlappedwiththemeasured datarangeforbothendpointsateachTBconcentration.ForplasmaVTGconcentrations,themedian modelpredictionswerewithin96%to579%ofthe Figure2 FourMarkovchains .Androgenreceptorrelativebinding affinity(RBA)forTBrelativetoT(RBATB_T).Thisisoneofthe26 calibratedmodelparametersillustratingwell-mixedMarkovchain trajectories. Li etal BMCSystemsBiology 2011, 5 :63 http://www.biomedcentral.com/1752-0509/5/63 Page14of22

PAGE 15

medianmeasurements.The579%differenceseems high,andasaresult,welookedintodetailsparticularly forthismodelprediction.WefoundthatthispredictionhappenedwhenTBconcentrationequalto0.5 g/ L.AtthisTBconcentration,wecollectedplasmaVTG concentrationsineachof4adultfemaleFHMs,which were4.61,26.21,1.99,and0.06mg/ml.Thelastmeasurement(0.06mgVTG/ml)ismorethan30-fold lowerthanthesecondlowestmeasurement(1.99mg VTG/ml).Asaresult,thisdatapointisanoutlier,and ourmodeldidnotcaptureit.Ifweexcludethisdata point,ourmodelpredictions(17.63,20.54,and5.07 mg/ml)wouldmatchtheexperimentaldatawell.For plasmaE2concentrations,themedianmodelpredictionswerewithin44to113%ofthemedianmeasurements.Kolmogorov-Smirnovtests( a =0.05)showed thatthemodelpredictionswerenotsignificantlydifferentfromthemeasureddataforbothplasmaVTGand E2concentrations. Tofurtherevaluatethemodel ’ spredictiveabilityfor TB-exposedFHMs,wesimulatedplasmaE2andVTG concentrationsinFHMsexposedto0,0.05,and0.5 g TB/Lfor8daysfollowedbyan8-daydepuration[40]. ForplasmaE2concentrations(Figure5A,B,and5C),the 95%CIsofmodelpredictionsencompassedthemedians ofthemeasureddatafor16outof24experimentalconditions(eightsamplingtimesandthreedifferentTBconcentrations).Generally,ourmodelpredictedtheplasma E2concentrationsbetterdu ringtheTBexposurephase thanduringthedepurationphase.Thisisnotsurprising sinceweonlycalibratedthemodelwithexperimental datafromaTBexposure[7].Inaddition,itisinteresting toseethatthemeasuredplasmaE2concentrations declinedfromthet=P48toP192hoursforbothcontrol FHMsandFHMsexposedtodifferentconcentrationsof TB.However,themodelpredictionsshowedadifferent trend;thatis,forcontrolFHMs,thepredictedplasmaE2concentrationsremainedrelativelystablethroughoutthe experimentalperiod(Figure5A);forTB-exposedFHMs, aftertheexposure,theplasmaE2concentrations increasedandrecoveredtoconcentrationsseeninunexposedFHMs.SincethemeasuredplasmaE2concentrationsdecreasedinbothcontrolFHMsandFHMs exposedtoTB,wesuspectthattheremightbesome experimentalfactorsthatwehavenotaccountedforin themodelduringthedepurationphase. Figure5D,E,and5Fcomparemodel-predictedplasma VTGconcentrationswiththemeasureddata.Themedian modelpredictionswerewithin0.2to3.6foldofthemeasuredmedian,andthe95%CIsofmodelpredictions encompassedallthemeasuredmediansateachsampling time.Theseresultsshowthatthemodelworkedwellfor Figure3 Comparisonofmodelpredictionswithmeasureddata inunexposedfemaleFHMs n =95.Whiteboxesrepresentmodel predictions,andgreyboxesrepresentmeasureddata[38].Thesolid linewithintheboxmarksthemedian;theboundaryofthebox farthestfromzeroindicatesthe75thpercentile;theboundaryofthe boxclosesttozeroindicatesthe25thpercentile;thewhisker(error bar)farthestfromzeromarksthe90thpercentile;whisker(errorbar) closesttozeromarksthe10thpercentile;thecirclefarthestfrom zeromarksthe95thpercentile;andthecircleclosesttozeromarks the5thpercentile. Li etal BMCSystemsBiology 2011, 5 :63 http://www.biomedcentral.com/1752-0509/5/63 Page15of22

PAGE 16

predictingtheplasmaE2andVTGconcentrationsin femaleFHMsexposedto0,0.05,and0.5 gTB/Lfor eightdays. Afterbeingcalibratedwiththeexperimentaldatafrom Ankleyetal.[7],ourmodelaccuratelypredictedthe plasmaE2,T,andVTGconcentrationsinadultfemale FHMsexposedtoTB.Thiswasachievedbysimulating AR-relatedligand-receptorbindingprocesses,andby assumingtwogeneregulationmechanisms:i)downregulationofARproductionbyfreeandrogens;andii)down regulationofLHproductionbyboundAR.Itisnoteworthythatthemodelwasabletoaccuratelyfitnotonly thecalibrationdata(seeAdditionalfile2),butalsothe VTGandE2datafromindependentstudiesbyGarciaReyeroetal.[39]andEkmanetal.[40].Theseresults indicatethatourAR-basedmodellingframeworkisplausible,andcouldbeusedinstudiesfocusedonregulatory aspectsoftheARonHPGfunction.Inarecentstudy, Shoemakeretal.[49]develo pedacomputationalmodel tosimulatemoredetailedbio chemicalreactionsinthe FHMsteroidogenicpathway.However,theirmodeldid notincorporateanyAR-rela tedsignallingpathways.As ARplaysanessentialroleforandrogenresponsesand subsequentregulationofsteroidogenesis,ourmodel advancestheworkofShoemakeretal.[49]bysimulating AR-relatedsignallingpathways. Thecalibrationandevaluationresultsshowedthatthe modelwasabletopredictthethreereproductiveendpointsfromdifferentstudieswithdifferentexperimental conditions.Althoughthedatasetsusedtocalibrateand validatethemodelwerefr omstudieswithdifferent experimentaldesignsandanalyticmethods,themodel accountedforthedifferencesandpredictedtheendpointswell.Forexample,thecalibrationdataweremeasuredinFHMsexposedtoTBfor21dayswithaflowthroughwaterexposurede sign,andtheplasmaVTG concentrationsweremeas uredbyapolyclonalFHMbasedELISA[7].Incontrast,onevalidationdatasetwas fromFHMsexposedtoTBfor48hourswithastatic waterexposuredesign,andplasmaVTGconcentrations measuredusingamonoclonalcarp-basedELISA[39], whiletheothervalidationdatasetwasfromFHMs exposedtoTBfor8daysfollowedbyan8daydepuration inaflow-throughsystem,withplasmaVTGconcentrationsmeasuredusingthepolyclonalFHM-basedELISA. Withtheparametersetcalibratedwiththedatafromone study,ourmodelpredictedplasmaE2andVTGconcentrationscomparabletothemeasurementsfromtheother twostudies.Thisindicatesthatthemodelnotonlyfitthe dataempirically,butalsocapturedmajorfeaturesofthe HPGaxisinfemaleFHMsexposedtoTB.Inaddition, thetwomodelevaluationsalsosupportedthepointby Watanabeetal.[20]thattheVTGmeasurementsbya polyclonalFHM-basedELISAandbyamonoclonalcarpbasedELISAareconsistent.PredictionsforplasmaVTGconcentrationsinEE2-exposed FHMsFigure6comparesmodel-predictedandmeasured plasmaVTGconcentrationsinfemaleFHMsexposedto Figure4 Comparisonofmodelpredictionswithmeasur eddatainfemaleFHMsexposedtoTBfor48hours n =32.Whitecircles representmodelpredictions,andgreycirclesrepresentmeasureddata[39].Eachcirclerepresentsonemeasurementinonefish.(A)plasmaVTG concentrations,and(B)plasmaE2concentrations.Thex-axisrepresentsTBconcentrationsin g/L.Note:forpanelB,at0.5 gTB/L,thereare only3measureddatapoints. Li etal BMCSystemsBiology 2011, 5 :63 http://www.biomedcentral.com/1752-0509/5/63 Page16of22

PAGE 17

threedifferentconcentrationsofEE2for21days[42]. Forthe0.5,1.5and4.5ng/Lexposures,respectively,the 90%,80%,and50%CIsofmodel-predictedVTGconcentrationsencompassedthemediansofthemeasured data.Thistrendsuggeststhatthemodelpredictsthe endpointbetterwhentheEE2exposureconcentrationis highandclosertotheexposureconcentrationsusedto calibratethemodel(i.e.,10and100ngEE2/L).For FHMsexposedto4.5ngEE2/L,themedianofour modelpredictionswasaround2timeshigherthanthe measureddata,andallmeasureddatawerewithinthe 95%CIsofthemodelpredictions.Consideringthat Figure5 ComparisonofmodelpredictionswithmeasureddatainfemaleFHMsexposedtoTBforeightdaysfollowedbyaneightdaydepuration n =8ateachsamplingtime.Whiteboxesrepresentmodelpredictions,andgreyboxesrepresentmeasureddata[40].The solidlinewithintheboxmarksthemedian;theboundaryoftheboxfarthestfromzeroindicatesthe75thpercentile;theboundaryofthebox closesttozeroindicatesthe25thpercentile.Becauseofthesmalldatasize( n =8),theplotsonlyshowthe50%confidenceintervals.(A)plasma E2concentrationsincontrolFHMs,(B)plasmaE2concentrationsinFHMsexposedto0.05 gTB/L,(C)plasmaE2concentrationsinFHMs exposedto0.5 gTB/L,(D)plasmaVTGconcentrationsincontrolFHMs,(E)plasmaVTGconcentrationsinFHMsexposedto0.05 gTB/L,(F) plasmaVTGconcentrationsinFHMsexposedto0.5 gTB/L.Thex-axisrepresentstimeinhours.P24,P48,P96,andP192represent24,48,96, and192hourspost-exposure,respectively. Li etal BMCSystemsBiology 2011, 5 :63 http://www.biomedcentral.com/1752-0509/5/63 Page17of22

PAGE 18

exposureconcentrationslessthan10ngEE2/Land exposuredurationslongerthan8daysareanextrapolationofthemodel,modelpredictionsofplasmaVTG concentrationsforthe21-day4.5ngEE2/Lexposure wereasurprisinglygoodfit .Thelowexposureconcentration,longertimeframeexposureismoreenvironmentallyrelevantbecauseEE2concentrationsrange from0.5to15ngEE2/Lintheaquaticenvironment [50-53],andaquaticanimalsmaybeexposedtothechemicalthroughouttheirlifetime. Additionally,wesimulatedplasmaVTGconcentrations inFHMsexposedto1.5ngEE2/Lfor21daysasreported byBrianetal.[43].Intotal,fourcontrolFHMsandfour FHMsexposedtoEE2weresimulated.Brianetal.measuredtheVTGconcentrationswithapolyclonalcarp VTGELISA,whichusespolyclonalantibodiesprepared fromcarpVTG.Incontrast,VTGdatausedtocalibrate themodelweremeasuredwithahomologousFHMVTG ELISA,whichusespolyclonalantibodiespreparedfrom FHMVTG.Directcomparisonofthetwomethodshave shownthatmeasurementsofFHMplasmaVTGconcentrationscandifferbyseveralordersofmagnitude[54].As aresult,insteadofcomparingthemodelpredictionswith themeasureddatadirectly,wecomparedtherelative changesofplasmaVTGconcentrations.Theresults showedthattherangeofmodel-predictedrelativechange was0.44to4.93,whiletherangeofthemeasureddata relativechangewas0.78to0.82,allwithintherangeof modelpredictions.PredictionsforreproductiveendpointsinamixtureofTB andEE2Inthenextphaseofouranalysis,wepredictedliverER concentration,brainARconcentration,andplasmaE2, T,VTG,andluteinizingho rmone(LH)concentrations infemaleFHMsexposedto15ngTB/L,10ngEE2/L, andamixtureof15ngTB/Land10ngEE2/Lfor48 hours,respectively.Forallendpoints,therewasa changeafterthechemicalexposurebeganfollowedby arecoverytobaselinevaluesaftertheexposureended. InpanelsA,B,andC,afterexposuretoTB,the plasmaE2,T,andVTGconcentrationsfolloweda trendconsistentwiththedatausedinthemodelcalibrationandevaluation.AfterexposuretoEE2,plasma E2andTconcentrationsdecreasedmoredramatically thanthatproducedbyTBexposure.Wedidnotfind anyreportsofplasmaE2orTconcentrationsinfemale FHMsexposedtoEE2.However,infemalezebrafish,it wasobservedthatbothplasmaE2andTconcentrationsdecreasedafterexposureto15ngEE2/Lfor48 hours[55],whichagreeswithourmodelpredictions. Inaddition,plasmaVTGconcentrationsincreased afterexposuretoEE2,consistentwiththedatausedto calibrateandevaluateourmo del.Interestingly,after exposuretoamixtureofTBandEE2,ourmodelpredictedthattheplasmaE2andTconcentrations decreasedinanadditivemanner.Incontrast,the plasmaVTGconcentrationincreasedandfollowedthe trendofanEE2exposure. InpanelsD,E,andFofFigure7,weplottedliver ER,brainAR,andplasmaLHconcentrations,respectively,asafunctionoftimeunderthethreedifferent exposureconditions.LiverERconcentrationswere predictedtodecreaseslig htlyafterexposuretoTB, andincreasedramaticallyafterexposuretoEE2andin responsetothemixture.PredictedliverERconcentrationsafterEE2exposureareconsistentwiththegene expressiondatainfemaleFHMsexposedto10ng EE2/L[56].BrainARconcentrationswerepredictedto increaseafterexposureto TBandthemixture,and decreaseslightlyafterexposuretoEE2.PlasmaLH concentrationswerepredictedtodecreaseafterexposuretoTBandthemixture,andincreaseslightlyafter exposuretoEE2(consistentwithobservationsinteleostsexposedtoEE2[25]).Todate,wedonothave publisheddatatoevaluatemodel-predictedeffectsfor amixtureofTBandEE2.Inaddition,threeofthepredictedendpoints(liverER,brainAR,andplasmaLH concentrations)havenotbeenmeasuredinFHMata proteinlevelbecauseofexperimentallimitations. However,thepredictionscanbeusedtogenerate hypothesesandhelpexplorepossiblemechanismsand pathways,whichmightbetestedinthefuture. Figure6 Comparisonofmodelpredictionswithmeasureddata infemaleFHMsexposedtoEE2. n =28ateachsamplingtime. Whiteboxesrepresentmodelpredictions,andgreyboxesrepresent measureddata[42].Thex-axisrepresentsEE2concentrationsinng/ L.Thesolidlinewithintheboxmarksthemedian;theboundaryof theboxfarthestfromzeroindicatesthe75thpercentile;the boundaryoftheboxclosesttozeroindicatesthe25thpercentile; thewhisker(errorbar)farthestfromzeromarksthe90thpercentile; whisker(errorbar)closesttozeromarksthe10thpercentile;the circlefarthestfromzeromarksthe95thpercentile;andthecircle closesttozeromarksthe5thpercentile. Li etal BMCSystemsBiology 2011, 5 :63 http://www.biomedcentral.com/1752-0509/5/63 Page18of22

PAGE 19

ConclusionsThemodelrepresentstheHPGaxisinadultfemale FHMrobustly,andpredictsplasmaE2,TandVTGconcentrationsinfemaleFHMsexposedtoTB,EE2,ora mixtureofTBandEE2.Thismodellinksenvironmental estrogenandandrogenexposuretochangesinapical reproductiveendpoints,andservesasafoundationthat canbeextendedtosimulateoocytegrowthdynamics andotheraspectsofreproduction.Inthisstudy,the modelpredictedreproductiveendpointsfromindependentstudieswell.Formoret han85%ofthesimulation results,the95%CIsofmodelpredictionsencompassed themedianoftheexperimentaldata.Tofurtherevaluate themodel ’ spredictiveability,moreexperimentaldata areneeded,especiallyfortheendpointsinFHMs exposedtoamixtureofTBandEE2. Importantnewfeaturesofthismodelinclude:(i)the simulationofARinmultipletissuecompartments(i.e., brain,liver,andgonad);(ii)ARbindinganditseffects upontheHPGaxis;and(iii)freeandrogeneffectson brainARconcentration.Asaresult,thismodelprovides acomputationalframeworkforendocrineresponsesof EDCsfunctioningthroughbothERandAR. Themodelcanbeusedtogeneratehypothesesto facilitatestudiesofendocrineresponsesinfemale FHMsexposedtootherestrogenicEDCsinadditionto EE2,orotherandrogenicEDCsinadditiontoTB.The applicationcanbeachievedbydefiningchemical-specificparameters,suchaspartitioncoefficients(e.g., bloodtowater,ortissuetoblood),andbindingaffinitiestoERandAR.Furthermore,theendpointssimulatedinthisstudy(i.e.plasmaE2,TandVTG concentrations)areimportantdeterminantsaffecting eggproductioninFHMs.Inthefuture,thismodel couldbelinkedtoanoocytegrowthdynamicsmodel developedbyLietal.(ac cepted).Linkingthesetwo modelswouldbuildaconnectionbetweenEDCeffects atamolecularlevelwitheffectsuponanorganism, andthusapopulation,whichisanurgentneedinecologicalriskassessment.AdditionalmaterialAdditionalfile1:DifferentialequationsusedintheHPGaxismodel ThefilewascreatedinMicrosoftOfficeWord2003.Thefilecontainsalist ofthedifferentialequationsusedintheHPGaxismodelforfemale fatheadminnows. Plasma E 2 Conc. (ng/ml) 0.1 1 10 Time (hr) Plasma T Conc. (ng/ml) 0.01 0.1 1 Plasma VTG Conc. (mg/ml) 15 20 25 30 35 Time (hr) Liver ER Conc. (nmol/L) 10 100 Brain AR Conc. (nmol/L) 1.08 1.12 1.16 Time (hr) Plasma LH Conc. (nmol/L) 1 2 3 AC BD E F600900 300 0600900 300 0600900 300 0 15 ng/L TB; 0 ng/L EE2 0 ng/L TB; 10 ng/L EE2 15 ng/L TB; 10 ng/L EE2 Figure7 ModelpredictionsforunmeasuredreproductiveendpointsinfemaleFHMs .ThepredictionsareforfemaleFHMsexposedto15 ng/LTB,10ng/LEE2,oramixtureof15ng/LTBand10ng/LEE2for48hours,respectively:(A)plasmaE2concentration,(B)plasmaT concentration,(C)plasmaVTGconcentration,(D)liverERconcentration,(E)brainARconcentration,and(F)plasmaLHconcentration. Li etal BMCSystemsBiology 2011, 5 :63 http://www.biomedcentral.com/1752-0509/5/63 Page19of22

PAGE 20

Additionalfile2:U-shapeddose-responsecurvesbetweenTBwater exposureconcentrationsandplasmaE2,T,andVTGconcentrations inadultfemaleFHMs .ThefilewascreatedinMicrosoftOfficeWord 2003.Thefilecontainsthreeplotsforthenon-monotonicrelationship betweenTBwaterexposureconcentrationsandplasmaE2,T,andVTG concentrationsinadultfemaleFHMs[7]. Acknowledgements ThisresearchwassupportedinpartbygrantsfromtheMedicalResearch FoundationofOregon[Grant#0634toK.W.];U.S.EnvironmentalProtection Agency(EPA)SciencetoAchieveResults(STAR)program[Cooperative Agreement#RD-83184801-0toN.D.,K.W.,M.S.,andE.O.];andbyresources fromtheNationalCenterforComputationalToxicologyoftheEPAOfficeof ResearchandDevelopment,awardedtotheEcologicalExposureResearch andEcosystemResearchDivisions(NationalExposureResearchLaboratory) inCincinnati,OHandAthens,GA,respectively,andtheMid-Continent EcologyDivision(NationalHealthandEnvironmentalEffectsResearch Laboratory)inDuluth,MN. Althoughtheresearchdescribedinthisarticlehasbeenfundedinpartby theU.S.EnvironmentalProtectionAgency,ithasnotbeensubjectedtoany formalEPAreviewanddoesnotnecessarilyreflecttheviewsoftheAgency. Noofficialendorsementshouldbeinferred. Theauthorsgreatlyappreciatetheinsightandinputfromseveralindividuals: Drs.DrewEkmanandTimColletteattheU.S.EPAEcosystemResearch DivisioninAthens,GA,USA;Dr.FredericBois(INERIS,FrenchNational InstituteforIndustrialEnvironmentandRisks)providedadviceonhowto improvemodelpredictionsofvariance;andtworeviewerswhosecomments improvedthequalityofthismanuscript.Theauthorswouldalsoliketo thanktheDivisionofEnvironmentalandBiomolecularSytemsatOHSUfor theirsupport. Authordetails1DivisionofEnvironmentalandBiomolecularSystems,OregonHealth& ScienceUniversity,Beaverton,OR,97006,USA.2DepartmentofPhysiological SciencesandCenterforEnvironmentalandHumanToxicology,Universityof Florida,Gainesville,FL,32611,USA.3U.S.EPA,Mid-ContinentEcology Division,Duluth,MN,55804,USA.4InstitutefortheEnvironment,Brunel University,Uxbridge,Middlesex,UB83PH,UK.5DepartmentofForestryand NaturalResources,PurdueUniversity,Lafayette,IN,47907,USA.6Department ofAnimal&AvianSciences,UniversityofMaryland,CollegePark,MD20742 USA.7U.S.EPA,MolecularIndicatorResearchBranch,Cincinnati,OH,45268, USA.8TheMcConnellGroupc/oU.S.EPANERLEERD,USA. Authors ’ contributions ZLcontributedtomodeldevelopment,simulation,andresultanalysis.DLV andNDDcontributedtomodeldesign,andprovidedexperimentaldata. KJK,KMJ,GTA,JVB,MSS,EFO,JML,MK,andBAprovidedexperimentaldata. KHWdirectedtheresearch,andcontributedtomodeldevelopment.All authorscontributedto,haveread,andapprovedthefinalversionofthe manuscript. Received:20December2010Accepted:5May2011 Published:5May2011 References1.NagahamaY: Endocrineregulationofgametogenesisinfish. IntJDevBiol 1994, 38 :217-229. 2.WeltzienFA,AnderssonE,AndersenO,Shalchian-TabriziK,NorbergB: The brain-pituitary-gonadaxisinmaleteleosts,withspecialemphasison flatfish( Pleuronectiformes ). CompBiochemPhysiolAMolIntegrPhysiol 2004, 137 :447-477. 3.AnkleyGT,BrooksBW,HuggettDB,SumpterJP: Repeatinghistory: pharmaceuticalsintheenvironment. EnvironSciTechnol 2007, 41 :8211-8217. 4.AnkleyGT,JensenKM,KahlMD,KorteJJ,MakynenEA: Descriptionand evaluationofashort-termreproductiontestwiththefatheadminnow ( Pimephalespromelas ). EnvironToxicolChem 2001, 20 :1276-1290. 5.PanterGH,HutchinsonTH,HurdKS,BamforthJ,StanleyRD,DuffellS, HargreavesA,GimenoS,TylerCR: Developmentofchronictestsfor endocrineactivechemicals.Part1.Anextendedfishearly-lifestagetest foroestrogenicactivechemicalsinthefatheadminnow( Pimephales promelas ). AquatToxicol 2006, 77 :279-290. 6.ThorpeKL,BensteadR,HutchinsonTH,TylerCR: Anoptimised experimentaltestprocedureformeasuringchemicaleffectson reproductioninthefatheadminnow, Pimephalespromelas AquatToxicol 2007, 81 :90-98. 7.AnkleyGT,JensenKM,MakynenEA,KahlMD,KorteJJ,HornungMW, HenryTR,DennyJS,LeinoRL,WilsonVS, etal : Effectsoftheandrogenic growthpromoter17-beta-trenboloneonfecundityandreproductive endocrinologyofthefatheadminnow. EnvironToxicolChem 2003, 22 :1350-1360. 8.OrlandoEF,KolokAS,BinzcikGA,GatesJL,HortonMK,LambrightCS, GrayLEJr,SotoAM,GuilletteLJJr: Endocrine-disruptingeffectsofcattle feedloteffluentonanaquaticsentinelspecies,thefatheadminnow. EnvironHealthPerspect 2004, 112 :353-358. 9.PawlowskiS,vanAerleR,TylerCR,BraunbeckT: Effectsof17alphaethinylestradiolinafatheadminnow( Pimephalespromelas )gonadal recrudescenceassay. EcotoxicolEnvironSaf 2004, 57 :330-345. 10.SekiM,FujishimaS,NozakaT,MaedaM,KobayashiK: Comparisonof responseto17beta-estradioland17beta-trenboloneamongthree smallfishspecies. EnvironToxicolChem 2006, 25 :2742-2752. 11.LngeR,HutchinsonTH,CroudaceCP,SiegmundF,SchweinfurthH, HampeP,PanterGH,SumpterJP: Effectsofthesyntheticestrogen17 alpha-ethinylestradiolonthelife-cycleofthefatheadminnow ( Pimephalespromelas ). EnvironToxicolChem 2001, 20 :1216-1227. 12.EricsonJF,LaengeR,SullivanDE: Commenton “ Pharmaceuticals,hormones,andotherorganicwastewatercontaminantsinU.S.streams, 1999-2000:anationalreconnaissance ” EnvironSciTechnol 2002, 36 :4005-4006,authorreply4007-4008. 13.PoirierD,LabrieC,MerandY,LabrieF: Derivativesofethynylestradiolwith oxygenated17alpha-alkylsidechain:synthesisandbiologicalactivity. J SteroidBiochem 1990, 36 :133-142. 14.GaleWL,PatinoR,MauleAG: Interactionofxenobioticswithestrogen receptorsalphaandbetaandaputativeplasmasexhormone-binding globulinfromchannelcatfish( Ictaluruspunctatus ). GenCompEndocrinol 2004, 136 :338-345. 15.DennyJS,TapperMA,SchmiederPK,HornungMW,JensenKM,AnkleyGT, HenryTR: Comparisonofrelativebindingaffinitiesofendocrineactive compoundstofatheadminnowandrainbowtroutestrogenreceptors. EnvironToxicolChem 2005, 24 :2948. 16.ParrottJL,BluntBR: Life-cycleexposureoffatheadminnows( Pimephales promelas )toanethinylestradiolconcentrationbelow1ng/Lreduces eggfertilizationsuccessanddemasculinizesmales. EnvironToxicol 2005, 20 :131-141. 17.KiddKA,BlanchfieldPJ,MillsKH,PalaceVP,EvansRE,LazorchakJM, FlickRW: Collapseofafishpopulationafterexposuretoasynthetic estrogen. ProcNatlAcadSciUSA 2007, 104 :8897-8901. 18.SchifferB,DaxenbergerA,MeyerK,MeyerHH: Thefateoftrenboloneacetate andmelengestrolacetateafterapplicationasgrowthpromotersincattle: environmentalstudies. EnvironHealthPerspect 2001, 109 :1145-1151. 19.DurhanEJ,LambrightCS,MakynenEA,LazorchakJ,HartigPC,WilsonVS, GrayLE,AnkleyGT: Identificationofmetabolitesoftrenboloneacetatein androgenicrunofffromabeeffeedlot. EnvironHealthPerspect 2006, 114(Suppl1) :65-68. 20.WatanabeKH,LiZ,KrollKJ,VilleneuveDL,Garcia-ReyeroN,OrlandoEF, SepulvedaMS,ColletteTW,EkmanDR,AnkleyGT,DenslowND: A computationalmodelofthehypothalamic-pituitary-gonadalaxisinmale fatheadminnowsexposedto17alpha-ethinylestradioland17betaestradiol. ToxicolSci 2009, 109 :180-192. 21.HeinleinCA,ChangC: Androgenreceptor(AR)coregulators:anoverview. EndocrRev 2002, 23 :175-200. 22.FilbyAL,TylerCR: MolecularCharacterizationofEstrogenReceptors1,2a, and2bandTheirTissueandOntogenicExpressionProfilesinFathead Minnow(Pimephalespromelas). BiolReprod 2005, 73 :648-662. 23.VilleneuveDL,LarkinP,KnoeblI,MiracleAL,KahlMD,JensenKM, MakynenEA,DurhanEJ,CarterBJ,DenslowND,AnkleyGT: Agraphical systemsmodeltofacilitatehypothesis-drivenecotoxicogenomicsLi etal BMCSystemsBiology 2011, 5 :63 http://www.biomedcentral.com/1752-0509/5/63 Page20of22

PAGE 21

researchontheteleostbrain-pituitary-gonadalaxis. EnvironSciTechnol 2007, 41 :321-330. 24.VerrijdtG,HaelensA,ClaessensF: SelectiveDNArecognitionbythe androgenreceptorasamechanismforhormone-specificregulationof geneexpression. MolGenetMetabol 2003, 78 :175-185. 25.YaronZ,GurG,MelamedP,RosenfeldH,ElizurA,Levavi-SivanB: Regulationoffishgonadotropins. IntRevCytol 2003, 225 :131-185. 26.KumarRC,ThakurMK: AndrogenreceptormRNAisinverselyregulatedby testosteroneandestradiolinadultmousebrain. NeurobiolAging 2004, 25 :925-933. 27.BurgessLH,HandaRJ: HormonalregulationofandrogenreceptormRNA inthebrainandanteriorpituitaryglandofthemalerat. BrainResMol BrainRes 1993, 19 :31-38. 28.ForadoriCD,WeiserMJ,HandaRJ: Non-genomicactionsofandrogens. FrontNeuroendocrinol 2008, 29 :169-181. 29.BurnsteinKL: Regulationofandrogenreceptorlevels:implicationsfor prostatecancerprogressionandtherapy. JCellBiochem 2005, 95 :657-669. 30.FlouriotG,PakdelF,DucouretB,LedreanY,ValotaireY: Differential regulationoftwogenesimplicatedinfishreproduction:vitellogenin andestrogenreceptorgenes. MolReprodDev 1997, 48 :317-323. 31.NagahamaY,YoshikuniM,YamashitaM,TokumotoT,KatsuY: Regulation ofoocytegrowthandmaturationinfish. CurrTopDevBiol 1995, 30 :103-145. 32.KagawaH,GenK,OkuzawaK,TanakaH: Effectsofluteinizinghormone andfollicle-stimulatinghormoneandinsulin-likegrowthfactor-Ion aromataseactivityandP450aromatasegeneexpressionintheovarian folliclesofredseabream,Pagrusmajor. BiolReprod 2003, 68 :1562-1568. 33.GaleWL,PatinoR,MauleAG: Interactionofxenobioticswithestrogen receptors[alpha]and[beta]andaputativeplasmasexhormonebindingglobulinfromchannelcatfish( Ictaluruspunctatus ). GenComp Endocrinol 2004, 136 :338-345. 34.Miguel-QueraltS,HammondGL: Sexhormone-bindingglobulininfish gillsisaportalforsexsteroidsbreachedbyxenobiotics. Endocrinology 2008, 149 :4269-4275. 35.TollefsenKE,OvrevikJ,StenersenJ: Bindingofxenoestrogenstothesex steroid-bindingproteininplasmafromArcticcharr( SalvelinusalpinusL .). CompBiochemPhysiolCToxicolPharmacol 2004, 139 :127-133. 36.TeeguardenJG,BartonHA: Computationalmodelingofserum-binding proteinsandclearanceinextrapolationsacrosslifestagesandspecies forendocrineactivecompounds. RiskAnal 2004, 24 :751-770.37.LaidleyCW,ThomasP: Partialcharacterizationofasex-steroidbindingproteininthespottedsea-trout( Cynoscionnebulosus ). BiolReprod 1994, 51 :982-992. 38.WatanabeKH,JensenKM,OrlandoEF,AnkleyGT: Whatisnormal? Biologicalvariabilityinreproductiveparametersofunexposedfathead minnows( Pimephalespromelas ). CompBiochemPhysiolCToxicol Pharmacol 2007, 146 :348-356. 39.Garcia-ReyeroN,VilleneuveDL,KrollKJ,LiuL,OrlandoEF,WatanabeKH, SepulvedaMS,AnkleyGT,DenslowND: Expressionsignaturesforamodel androgenandantiandrogeninthefatheadminnow( Pimephales promelas )ovary. EnvironSciTechnol 2009, 43 :2614-2619. 40.EkmanDR,VilleneuveDL,TengQ,Ralston-HooperKJ,Martinovi -WeigeltD, KahlMD,JensenKM,DurhanEJ,MakynenEA,AnkleyGT,ColletteTW: Use ofGeneExpression,BiochemicalandMetaboliteProfilestoEnhance ExposureandEffectsAssessmentoftheModelAndrogen17bTrenboloneinFish. EnvironToxicolChem 41.EkmanDR,TengQ,VilleneuveDL,KahlMD,JensenKM,DurhanEJ, AnkleyGT,ColletteTW: Investigatingcompensationandrecoveryof fatheadminnow( Pimephalespromelas )exposedto17 -ethynylestradiol withmetaboliteprofiling. EnvironSciTechnol 2008, 42 :4188-4194. 42.LazorchakJ,ArmstrongB,CraneM,AnkleyG,BencicD,BialesA,ColletteT, EkmanD,FlickR,JensenK, etal : Resultsofa21-DayFatheadMinnow ( PimphalesPromelas )FecundityStudyFollowingExposureto Ethinylestradiol. ProceedingsoftheSocietyofEnvironmentalToxicologyand ChemistryNorthAmerica31stAnnualMeeting;7-11November2010;Portland Oregon 2010,180. 43.BrianJV,HarrisCA,ScholzeM,KortenkampA,BooyP,LamoreeM,PojanaG, JonkersN,MarcominiA,SumpterJP: Evidenceofestrogenicmixtureeffects onthereproductiveperformanceoffish. EnvironSciTechnol 2007, 41 :337-344. 44.WatanabeKH,LinH,BartHLJr,MartinatP,MeansJC,KunasML,GrimmDA: Bayesianestimationofkineticrateconstantsinafoodwebmodelof polycyclicaromatichydrocarbonbioaccumulation. EcolModel 2005, 181 :229-246. 45.LinH,BerzinsDW,MyersL,GeorgeWJ,AbdelghaniA,WatanabeKH: A Bayesianbioaccumulationmodelofpolycyclicaromatichydrocarbonsin crayfish( Procambarusspp .). EnvironToxicolChem 2004, 23 :2259-2266. 46.GelmanAJ,CarlinJB,SternHS,RubinDB: BayesianDataAnalysis Boca Raton,FL:Chapman&Hall/CRC;1995. 47.BoisFY,JacksonET,PekariK,SmithMT: Populationtoxicokineticsof benzene. EnvironHealthPerspect 1996, 104 :1405-1411. 48.BoisFY,MaszleDR: MCSim:AMonteCarloSimulationProgram. JStatSoftw 1997, 2 :1-56. 49.ShoemakerJ,GayenK,Garcia-ReyeroN,PerkinsE,VilleneuveD,LiuL, DoyleF: Fatheadminnowsteroidogenesis:insilicoanalysesreveals tradeoffsbetweennominaltargetefficacyandrobustnesstocross-talk. BMCSystBiol 2010, 4 :89. 50.EricsonJF,LaengeR,SullivanDE: Commenton “ Pharmaceuticals, Hormones,andOtherOrganicWastewaterContaminantsinU.S. Streams,1999-2000:ANationalReconnaissance ” EnvironSciTechnol 2002, 36 :4005-4006. 51.YingGG,KookanaRS,KumarA: Fateofestrogensandxenoestrogensin foursewagetreatmentplantswithdifferenttechnologies. EnvironToxicol Chem 2008, 27 :87-94. 52.DesbrowC,RoutledgeEJ,BrightyGC,SumpterJP,WaldockM: IdentificationofestrogenicchemicalsinSTWeffluent.1.Chemical fractionationandinvitrobiologicalscreening. EnvironSciTechnol 1998, 32 :1549-1558. 53.RibeiroC,PardalMA,MartinhoF,MargalhoR,TiritanME,RochaE, RochaMJ: DistributionofendocrinedisruptorsintheMondegoRiver estuary,Portugal. EnvironMonitAssess 2008. 54.MylchreestE,SnajdrS,KorteJJ,AnkleyGT: ComparisonofELISAsfor detectingvitellogenininthefatheadminnow( Pimephalespromelas ). CompBiochemPhysiolCToxicolPharmacol 2003, 134 :251-257. 55.HoffmannJL,TorontaliSP,ThomasonRG,LeeDM,BrillJL,PriceBB,CarrGJ, VersteegDJ: HepaticgeneexpressionprofilingusingGenechipsin zebrafishexposedto17alpha-ethynylestradiol. AquatToxicol 2006, 79 :233-246. 56.FilbyAL,ThorpeKL,MaackG,TylerCR: Geneexpressionprofilesrevealing themechanismsofanti-androgen-andestrogen-inducedfeminizationin fish. AquatToxicol 2007, 81 :219-231. 57.NicholsJW,FitzsimmonsPN,WhitemanFW,DawsonTD,BabeuL, JuenemannJ: Aphysiologicallybasedtoxicokineticmodelfordietary uptakeofhydrophobicorganiccompoundsbyfish:I.Feedingstudies with2,2 ’ ,5,5 ’ -tetrachlorobiphenyl. ToxicolSci 2004, 77 :206-218. 58.NicholsJW,McKimJM,LienGJ,HoffmanAD,BertelsenSL,ElonenCM: A physiologicallybasedtoxicokineticmodelfordermalabsorptionof organicchemicalsbyfish. FundamApplToxicol 1996, 31 :229-242. 59.RobinsonDE,BalterNJ,SchwartzSL: Aphysiologicallybased pharmacokineticmodelfornicotineandcotinineinman. J PharmacokinetBiopharm 1992, 20 :591-609. 60.PlowchalkDR,TeeguardenJ: Developmentofaphysiologicallybased pharmacokineticmodelforestradiolinratsandhumans:abiologically motivatedquantitativeframeworkforevaluatingresponsestoestradiol andotherendocrine-activecompounds. ToxicolSci2002, 69 :60-78. 61.MiwaS,YanL,SwansonP: Localizationoftwogonadotropinreceptorsin thesalmongonadbyinvitroligandautoradiography. BiolReprod 1994, 50 :629-642. 62.SperryTS,ThomasP: Characterizationoftwonuclearandrogenreceptors inAtlanticcroaker:comparisonoftheirbiochemicalpropertiesand bindingspecificities. Endocrinology 1999, 140 :1602-1611. 63.MurphyCA,RoseKA,ThomasP: Modelingvitellogenesisinfemalefish exposedtoenvironmentalstressors:predictingtheeffectsofendocrine disturbanceduetoexposuretoaPCBmixtureandcadmium. Reprod Toxicol 2005, 19 :395-409. 64.WilsonVS,CardonMC,GrayLE,HartigPC: Competitivebinding comparisonofendocrine-disruptingcompoundstorecombinant androgenreceptorfromfatheadminnow,rainbowtrout,andhuman. EnvironToxicolChem 2007, 26 :1793-1802. 65.CrimLW,ArnaudRS,LavoieM,LabrieF: AstudyofLH-RHreceptorsinthe pituitaryglandofthewinterflounder( Pseudopleuronectesamericanus Walbaum ). GenCompEndocrinol 1988, 69 :372-377.Li etal BMCSystemsBiology 2011, 5 :63 http://www.biomedcentral.com/1752-0509/5/63 Page21of22

PAGE 22

66.KashiwagiK,DafeldeckerW,SalhanickH: Purificationandcharacterization ofmitochondrialcytochromeP-450associatedwithcholesterolside chaincleavagefrombovinecorpusluteum. JBiolChem 1980, 255 :2606-2611. 67.ShikitaM,HallPF: CytochromeP-450frombovineadrenocortical mitochondria:anenzymeforthesidechaincleavageofcholesterol.I. Purificationandproperties. JBiolChem 1973, 248 :5598-5604. 68.ZhaoJ,MakP,TchoudakovaA,CallardG,ChenS: Differentcatalytic propertiesandinhibitorresponsesofthegoldfishbrainandovary aromataseisozymes. GenCompEndocrinol 2001, 123 :180-191. 69.BartonHA,AndersenME: Amodelforpharmacokineticsand physiologicalfeedbackamonghormonesofthetesticular-pituitaryaxis inadultmalerats:Aframeworkforevaluatingeffectsofendocrine activecompounds. ToxicolSci 1998, 45 :174-187. 70.ArtemenkoIP,ZhaoD,HalesDB,HalesKH,JefcoateCR: Mitochondrial processingofnewlysynthesizedsteroidogenicacuteregulatoryprotein (StAR),butnottotalStAR,mediatescholesteroltransfertocytochrome P450sidechaincleavageenzymeinadrenalcells. JBiolChem 2001, 276 :46583-46596. 71.GeR-S,GaoH-B,NacharajuVL,GunsalusGL,HardyMP: Identificationofa kineticallydistinctactivityof11 -hydroxysteroiddehydrogenaseinrat Leydigcells. Endocrinology 1997, 138 :2435-2442. 72.SchulzR,BosmaP,ZandbergenM,VanderSandenM,VanDijkW,PeuteJ, BogerdJ,GoosH: Twogonadotropin-releasinghormonesintheAfrican catfish, Clariasgariepinus :localization,pituitaryreceptorbinding,and gonadotropinreleaseactivity. Endocrinology 1993, 133 :1569-1577. 73.BoisFY,GelmanA,JiangJ,MaszleDR,ZeiseL,AlexeefG: Population toxicokineticsoftetrachloroethylene. ArchToxicol 1996, 70 :347-355. doi:10.1186/1752-0509-5-63 Citethisarticleas: Li etal .: Acomputationalmodelofthe hypothalamic-pituitary-gonadalaxisinfemalefatheadminnows ( Pimephalespromelas )exposedto17 a -ethynylestradioland17 b trenbolone. BMCSystemsBiology 2011 5 :63. Submit your next manuscript to BioMed Central and take full advantage of: Convenient online submission Thorough peer review No space constraints or color gure charges Immediate publication on acceptance Inclusion in PubMed, CAS, Scopus and Google Scholar Research which is freely available for redistribution Submit your manuscript at www.biomedcentral.com/submit Li etal BMCSystemsBiology 2011, 5 :63 http://www.biomedcentral.com/1752-0509/5/63 Page22of22

PAGE 23

1 Additional F ile 1 : Differential equations used in the HPG axis model In this file, we include all differential equations used in the HPG model for female FHMs. At the end of each model equation where there is a corresponding equation in the main manuscri pt, we include the main manuscript equation number in parentheses as a cross reference. Brain (Equation 1) (Equation 1) (Equation 1) (Equation 1) (Equation 4 ) (Equation 5)

PAGE 24

2 Gonad (Equation 1) (Equation 1) (Equation 1) (Equation 1) (Equation 1) (Equation 6 ) (Equati on 7 ) Liver (Equation 1)

PAGE 25

3 (Equation 1) (Equation 1) (Equation 1) Other (Equation 8 ) Venous blood (Equation 1)

PAGE 26

4 (Equation 1) (Equation 1)

PAGE 27

1 Additional file 2 : U shaped dose response curves between TB water exposure concentrations and plasma E 2 T, and VTG conc entrations in adult female FHMs Figure AF1: Comparison of model predicted plasma E 2 concentrations with experimental data in adult fema le FHMs exposed to TB for 21 days [ 1 ] Gr a y bars represent model predictions, black bars represent experimental data, and error bars represent one standard deviation. Figure AF2: Comparison of model predicted plasma T concentrations with experimental data in adult female FHMs exposed to TB for 21 days [ 1 ] Gray bars represent model predictions, black bars represent experimental data, and error bars represent one standard deviation.

PAGE 28

2 Figure AF 3 : Comparison of model predicted plasma VTG concentrations with experimental data in adult female FHMs exposed to TB for 21 days [ 1 ] Gra y bars represent model predictions, black bars represent experimental data, and error bars represent one standard deviation. References cited: 1. Ankley GT, Jensen KM, Makynen EA, Kahl MD, Korte JJ, Hornung MW, Henry TR, Denny JS, Leino RL, Wilson VS, et al: Effects of the androgenic growth promoter 17 trenbolone on fecundity and reproductive endocrinology of the fathead minnow. Environ Toxicol Chem 2003, 22: 1350 1360.