SCIENCE ADVANCES | RESEARCH ARTICLE BIOCHEMICAL EVOLUTION 2017©TheAuthors, somerightsreserved; Origin of an ancient hormone/receptor couple revealed exclusivelicensee AmericanAssociation by resurrection of an ancestral estrogen fortheAdvancement ofScience.Distributed Gabriel V. Markov,1,2* Juliana Gutierrez-Mazariegos,1 Delphine Pitrat,3 Isabelle M. L. Billas,4,5,6,7 underaCreative François Bonneton,1† Dino Moras,4,5,6,7 Jens Hasserodt,3 Guillaume Lecointre,8 Vincent Laudet1‡§ CommonsAttribution NonCommercial License4.0(CCBY-NC). The originofancientligand/receptorcouplesisoftenanalyzedviareconstructionofancientreceptorsand,when ligandsareproductsofmetabolicpathways,theyarenotsupposedtoevolve.However,becausemetabolicpath- waysareinheritedbydescentwithmodification,theirstructurecanbecomparedusingcladisticanalysis.Usingthis approach, we studied the evolution of steroid hormones. We show that side-chain cleavage is common to most vertebrate steroids, whereas aromatization was co-opted for estrogen synthesis from a more ancient pathway. Theancestralproductsofaromaticactivitywerearomatizedsteroidswithasidechain,whichwenamed“paraestrols.” We synthesized paraestrol A and show that it effectively binds and activates the ancestral steroid receptor. Our studyopensthewaytocomparativestudiesofbiologicallyactivesmallmolecules. D o INTRODUCTION onstrated only in vertebrates, consistent with the origin of CYP11A w n Reconstructingthepasttobetterunderstandevolutionaryprocessesis fromavertebrate-specificgeneduplication(6).Onthecontrary,thearo- lo a animportantgoalofevolutionarysciences.Thisapproachreliesgreat- mataseCYP19Aisspecifictochordates,whichisatoddswithreportsof de d alyreounnaifcoturmalisamcr,otshsetipmrein(c1ip).leUasicncgortdhiinsgprtoincwiphliec,hplaalweosnatnoldogpirsotscehsasvees aturorem,asttoasree,aacntidvitmyeintabmoolilzleusvkesrtaenbdractneisdeaxrisatnerso(i7d,s8,)w.hMicohlluasreksacbaunndcaapn-t from reconstructed models of ancient birds, such as Archaeopteryx, and insewagewater.However,thereisnoconclusiveevidencethattheycan h bsteuednieudsetdhetior rfleicgohntsatrbuilcittieasnc(e2s,tr3a)l.pMrootreeinrsectoenrtelyv,eathlihsisatpoprircoaalcmhuhtaas- elyndporogepnooseudslythsaytntthheesiozebsveerrvteebdraatreo-mtypateizsatetiroonidrheaocrtmioonnsesou(9ts)i.dWeeveprrteevbiroautess- ttp://ad v tionsthathaveshapedtheirfunctionthroughtime(4).Unfortunately,we occuronsteroidsthataredifferentfromvertebrateandrogens(10),which a n donothaveyettheconceptualandtechnicalframeworktoeasilyre- issupportedbytheidentificationofthesearomatizedlong-chainedsteroids ce s constructsmallmoleculesproducedbycomplexmetabolicpathways. inspongesandcnidarians(11,12).However,steroidogenesisismorecom- .s c Thisisproblematicforligand-dependentproteins,suchastheancestral plicatedthanonlythosetwokeyreactions.Unbiasedhypothesistestingof ie n steroidreceptor.Thatiswhymostreconstructionsofhormone/receptor theevolutionofsteroidmetabolicactivitieshastotakeintoaccountall ce m couples assume that only proteins evolve. This assumption is in- theavailableknowledgeonthewell-characterizedsteroidogenicpathways a consistentwiththedistributionofbiochemicalfeaturesrelatedtoestro- inanimals(5,13–17).Thestandardandtransparentprocedureforthis g.o gensynthesisacrossanimals(Fig.1).Thismultisteppathwaystartswith analysisisparsimony,whichiswidelyusedinthecomparativeanatomy org/ side-chaincleavageofcholesteroltopregnenoloneby the CYP11Aen- fieldandhasbeenexportedintobiochemistryfororderingbiochemical n N zymeand,aftermanyadditionalreactions,endswitharomatizationof eventsrelatedtotheuniversalmetabolism(18,19). o v testosteroneintoestradiolbytheCYP19Aenzyme(Fig.1)(5).Todate, e m side-chain cleavage of cholesterol to pregnenolone has been dem- b e RESULTS r 2 1 1MolecularZoologyTeam,InstitutdeGénomiqueFonctionnelledeLyon,Université Finding shared features of metabolic pathways using , 2 parsimony analysis 0 deLyon,UniversitéLyon1,CNRS,InstitutNationaldelaRechercheAgronomique 1 (INRA),ÉcoleNormaleSupérieuredeLyon,46alléed’Italie,69364LyonCedex07, We used standard parsimony methods (cladistics)—in which bio- 8 France.2EvolutiondesRégulationsEndocriniennes,DépartementRégulations,Dé- chemicalpathwaysareconsideredastaxa,andtheindividualreac- veloppementetDiversitéMoléculaire,CNRSUMR7221,SorbonneUniversités,Mu- tions(ortypesofreactions)catalyzedbyidentifiedenzymesshared séumNationald’HistoireNaturelle(MNHN),Paris,France.3LaboratoiredeChimie, UniversitédeLyon,UniversitéLyon1,CNRSUMR5182,ÉcoleNormaleSupérieurede by pathways are the compared character states (Fig. 2)—to inves- Lyon,46alléed’Italie,69364LyonCedex07,France.4CentreforIntegrativeBiology, tigatetheevolutionaryrelationshipsbetweensynthesispathwaysof DepartmentofIntegratedStructuralBiology,InstituteofGeneticsandMolecular vertebrate sex and adrenal steroids, oxysterols and vitamin D, hu- andCellularBiology,Illkirch,France.5CNRSUMR7104,Illkirch,France.6INSERM U964,Illkirch,France.7UniversitédeStrasbourg,Strasbourg,France.8Département man and rodent bile acids, nematode dafachronic acids, and arthro- SystématiqueetEvolution,InstitutdeSystématique,Évolution,Biodiversité,UMR pod ecdysteroids. All these pathways start from cholesterol (except 7205,CNRS–MNHN–UniversitéPierreetMarieCurie(UPMC)–ÉcolePratiquedes vitamin D synthesis, which starts from 7-dehydrocholesterol) and HautesÉtudes(EPHE),SorbonneUniversités,MuséumNationald’HistoireNaturelle, endwithafinalproduct,whichis aligandofanuclearreceptor(NR) CP30,57rueCuvier,75005Paris,France. *Presentaddress:SorbonneUniversités,UPMCUniversitéParis06,CNRS,UMR8227, (figs.S1andS5).Toelucidatetherelativetimingoftheappearanceof IntegrativeBiologyofMarineModels,StationBiologiquedeRoscoff,PlaceGeorges differentmetazoansteroidogenicpathways,wereconstructedtheirre- Teissier,CS90074,29688RoscoffCedex,France. lationshipsbycomparingtheenzymaticreactionsthatareinvolvedin †Presentaddress:Developmentalgenomicsandevolution,UniversitéLyon,Institut deGénomiqueFonctionnelledeLyon,CNRSUMR5242,EcoleNormaleSupérieurede eachofthem.Inourdatasetof72taxa,eachpathwayfromcholesterol Lyon,UniversitéClaudeBernardLyon 1,46alléed’ItalieF-69364Lyon,France. to asteroidis ataxon—for example,cholesterol-to-estrone(C-to-E1; ‡Presentaddress:SorbonneUniversités,ObservatoireocéanologiquedeBanyuls- Fig.2A)orcholesterol-to-estradiol(C-to-E2;Fig.2A)—andenzymat- sur-mer,UPMCUniversitéParis06,CNRS,1avenuePierreFabre,66650Banyuls- icreactionsalongthispathwayarethecharactersthatarecodedfor sur-mer,France. §Correspondingauthor.Email:[email protected] eachtaxon(Fig.2,AandB).Weconstructedadatamatrixinwhich Markovetal.,Sci.Adv.2017;3:e1601778 31March2017 1of13 SCIENCE ADVANCES | RESEARCH ARTICLE D o Fig.1. Estradiolsynthesispathwayinvertebratesanddistributionacrossmetazoansofrelatedbiochemicalfeatures.Vertebrateestradiolsynthesisfromcho- w n lesterolisamultisteppathwayinvolvingside-chaincleavageofcholesteroltopregnenolonebytheCYP11Aenzyme(red),acomplexsuiteofreactionsfrompreg- lo a nenolonetotestosterone(dottedarrow),andaromatizationbytheCYP19Aenzyme(blue).CYP19Aisalreadypresentinchordates(bluedotontree),whereasCYP11Ais de vcaetratelybzreatteh-espaercoimficat(irzeadtiodnotreoancttiroene,).wAhroermeaastizinatmionollhuassksbaenedncdneisdcarirbiaends,inthcisniddeafriinaintesl,ymhoalslutsoksb,eanaddicfefeprheanltoecnhzoyrdmaet,epso(bssluibelyoavaplsa)r.aIlnogceopuhsaCloYcPh(obrlduaetedso,tCsYoPn1m9Aolmluasky d fro m andcnidarianbranchesofthetree).Becausethereisnoconsistentevidenceforside-chaincleavageactivitiesinthoseorganisms,theexactendogenoussubstrateof h thosearomatizingenzymesremainselusive.Itcouldbeasteroid(hencethedottedlinefromcholesterol)butdifferentfromthevertebrateones(hencetheincomplete ttp moleculestructures). ://a d v a n thepathwaysaresystematicallycomparedintermsofenzymaticre- Side-chain cleavage, and not aromatization, as the ce s actions controlling the production of each ligand (Fig. 2, A and B, common shared trait of vertebrate sex and adrenal .s c andfig.S1).The151charactersweredefinedaccordingtotwocriteria steroid synthesis pathways ie n ofhomologies:typeIandtypeII(Fig.2,AandB,andfig.S1).TypeI The unique synapomorphy uniting all bilaterian steroidogenic path- ce m homologies are defined when two reactions in different metabolic ways is hydroxylation of the sterol backbone by a cytochrome P450 a g pathwayssharethesameenzymewiththesamespecificityforasub- (CYP)enzyme(character147atnodeA,greendot;Fig.3).Then,among .o strateandthesameproduct,forexample,side-chaincleavageatcar- bilateriansteroidogenicpathways,threemajorgroupsappear.Afirst rg o/ bon20(seenumberinginFig.2C)inthesynthesisofestrogenandestrone large group gathers synthesis of vertebrate sex and adrenal steroids n (character31inC-to-E1andC-to-E2;Fig.2,AandB)ordehydrogenation (highlightedinlightred),whichshareside-chaincleavageand3b- No ofcholesterolatcarbon7inthesynthesisofecdysoneorD7-dafachronic hydroxylation on a steroid without a lateral side chain (node B, red ve m acid(character 58;Fig.2,AandB). TypeII homologies arecasesin dot). The second group gathers synthesis pathways of bile acids and b e whichpathwaysshareenzymaticfunctionswithoutsharingthespec- D4-dafachronic acid that share 3b-hydroxylation on a steroid with a r 2 1 i(fcihciatryafcoterra1s4u7b;stFriagt.e2i,nAamanodreB()chrealraaxcetderw1a3y4;(Failgso.2s,eAe ManadteBri)aolsralensds lgaattehrearlssisdyencthhaeisnis(cphaathrawcateyrs1o4f1eactdnyostdeeroCid,lsigahntdgrDay7)-.dTafhaechthroirndicgraocuidp , 20 1 Methods),orwhentheyshareanenzymeaddingthesameresidueat that share cholesterol 7,8-dehydrogenation (character 58 at node D, 8 different positions on the carbon skeleton (character 151; fig. S2). darkgray).Therelationshipsbetweenthosethreegroupsremainun- Therearemanywaystogofromcholesteroltosomesteroids;there- resolved and form a seven-branch polytomy with the synthesis of fore,weuseddifferentnumberstorefertothosevariouspathways(for calcitriolandthreeoxysterols.AtnodeB,alargepolytomyof11branches example,C-to-CHENOAC_1andC-to-CHENOAC_2;fig. S2). (withtwosuboptimalnodes)isfoundevenunderverypermissivecon- Treatingthismatrixbyparsimonyanalysis(70ofthe151charac- sensusconditions,whichindicateseitherlackofdataorthattheshared terswereparsimony-informative),weobtainedaconsensusof216equi- common characters are not consistent enough to solve that region of parsimonious trees, each of 260 steps [consistency index (CI), 0.406; thetree.5a-Reductionoftestosteroneinthesynthesisofdihydrotestos- retentionindex(RI),0.554].Tochecktherobustnessofthisconsensus terone (C-to-DHT1-8) has been recruited four times independently tree,weverifiedthatsearchesunderavaryingnumberofiterations (character130atnodesEtoH),whereassteroidaromatizationhasbeen anddifferentheuristicsledtothesametopology(seeMaterialsand recruited seven times independently (character 134, blue dots). This Methods).InFig.3,wepresenta50%consensustree,inwhichwehighlight suggests that enzymes capable of performing 5a-reduction and aro- 12strictlyconservednodes(AtoL)and2suboptimalnodes(S1and matization were already present before the appearance of enzymes S2).ThesuboptimalnodeS1isrecoveredwhentaxaC-to-15OHE2_1 performing side-chain cleavage and that they could later exploit the andC-to-15OHE1_1areexcluded,andthisalsoleadstothegaininthe new possibilities offered by the cleaved steroids. It is therefore likely nodegroupingC-to-E1_1andC-to-E2_1andthebranchleadingtoC- thattheseenzymeswereoriginallyactingonnoncleavedsteroidsand to-E1_5andC-to-E2_5.ThesuboptimalnodeS2andalldownstream were later recruited in the synthesis of steroids with a cleaved side nodesarerecoveredwhenC-to-F2andC-to-CRTSN2areexcluded. chain.Aromatizationoccursconsistentlyinamphioxusovaryhomogenates, Markovetal.,Sci.Adv.2017;3:e1601778 31March2017 2of13 SCIENCE ADVANCES | RESEARCH ARTICLE D o w n lo a d e d fro m h ttp ://a d v a n c e s .s c ie n c e m a g .o rg Fig.2. Parsimonyanalysisofmetazoansteroidogenicpathways.(A)Fiveexamplesofmetazoansteroidogenicpathways.Somereactionsaredetailedtoexplainthe o/ character-codingprocedure.Somepathwaysshareidenticalreactions,definedhereastypeIhomologies(hI,inred),suchasside-chaincleavageofcholesterolto n N pregnenolone(reaction31,inred)inthesynthesisofestradiolandestrone(C-to-E1andC-to-E2pathways)ordehydrogenationofcholesterolto7-dehydroxycholesterol o v (reaction58,inred)intheC-to-ECDYandC-to-DELTA7pathways.Somepathwaysalsosharesimilarreactions,definedastypeIIhomologies(hII),withvariousdegreesof e m similarity.TypeIIahomologies(hIIa)describethesamechemicalmodificationatthesameplaceontwodifferentsubstrates,suchasaromatization(character134,inblue)on b e androstenedioneintheC-to-E1pathway(reaction54)ortestosteroneintheC-to-E2pathway(reaction55).TypeIIchomologies(hIIc)describethesamechemicalmodification r 2 atdifferentplacesontwodifferentsubstrates,suchashydroxylation(character147,ingreen)oncarbon25fromcholesterol(reaction94)intheC-to-CHENOpathwayor 1 hydroxylationoncarbon20fromecdysoneintheC-to-ECDYpathway.C.elegans,Caenorhabditiselegans.(B)Extractofthedatamatrixcorrespondingtothepathwaysand , 20 stepsdescribedabove.Thepresenceorabsenceofcharactersiscodedby1and0,respectively(seecompletematrixinfig.S1).(C)Nomenclatureofcarbonnumbersandrings 18 onthesterolskeleton. in vitro; in contrast, side-chain cleavage was not biochemically de- of 17b-estradiol to these receptors (24, 25) is probably a pharma- tectedunderthesameconditions,butwasinferredfromthegonadal cological effect. expression of a putative CYP11 (20). This expressed gene was later Studies onsteroid receptors fromtheNR3 subfamilyhavesug- shownnottobeaCYP11Abutadistantparalog,whichwasrenamed gestedthattheestrogen-bindingabilityevolvedbeforetheabilityto CYP374(6).TherearesomeCYP11genesinamphioxus(nownamed bindother side-chain cleavedsteroids, such asandrogens (24–26). CYP11DandCYP11E)thatbranchatthebasisofthevertebrate-specific As discussed above, in light of our new results, we infer that the CYP11A/CYP11Bduplication(21);however,todate,wehavenodata ancestral ligand to the first steroid-binding receptor was not an abouttheirexpressionpatternorthebiochemicalactivityoftheiren- 18-carbon estrogen but a molecule with a side chain in addition coded products. Furthermore, the amphioxus genome contains tothearomaticA-ringspecifictovertebrateestrogens(Fig.4A).Ste- an aromatase gene (22, 23) but lacks a one-to-one ortholog of the roidswithanaromaticA-ringandalateralsidechainhavebeeniden- side-chaincleavageenzyme(6).Therefore,thisanalysisstronglysug- tifiedinatropicalsponge(11),andtheverymoleculeweproposeasa gests that steroids with a cleaved side chain are specific to verte- proxyforthisancestral chordatesteroid,19-norcholesta-1,3,5(10)- brates and absent in other animals. This implies that the steroid trien-3-ol,wasrecentlyisolatedfromasoftcoral(12).Thesynthe- receptors present in other bilaterians (particularly the NR3Ds; sisofthesemoleculesbynonbilateriananimalsisprobablyrelatedto Fig.4B)bindsteroidswithasidechain.Ifthisistrue,thenbinding defensive functions rather than to endocrine signaling (27). Because Markovetal.,Sci.Adv.2017;3:e1601778 31March2017 3of13 SCIENCE ADVANCES | RESEARCH ARTICLE D o w n lo a d e d fro m h ttp ://a d v a n c e s .s c ie n c e m a g .o rg o/ n N o v e m b e Fig.3. Relationshipsamonganimalsteroidogenicpathwaysinferredfromstandardparsimonyanalysis.The50%majority-ruleconsensusparsimonytreeis r 21 rootedusingeukaryoticsterolsynthesispathways.S1andS2indicatesuboptimalnodes.Characternumbersonbranchesrefertofig.S1B.Allvertebratesexandadrenal , 2 0 steroidsynthesispathways(highlightedinlightred)aregroupedtogether,sharingaside-chaincleavagereaction(31),whereasaromatizedsteroidsaredispersed 1 8 acrossthetree,indicatingthataromatizationofsteroidswithoutasidechainwasrecruitedmanytimesindependently(blueovals,134).Asharedtraitofallanalyzed animalsteroidsynthesispathwaysistheoccurrenceofhydroxylationonsomecarbonresidues(greendot,147). cholesterol hydroxylation already appeared at node A, the ancestral roidscomprisingacholesterolsidechainandanaromatizedA-ringto steroidmayalsohavehadanadditionalhydroxylgroupataposition avoidconfusionwithothernaturalproductsbearingadditionalmod- thatwecannotexactlydetermine.Theorderofappearanceofthebio- ificationsonthesidechain,suchasethylormethylgroupsoncarbon chemicalactivities(seebelow)indicatesthat24-and25-hydroxylation 24(11,12).Wefurtherreferto19-norcholesta-1,3,5(10)-trien-3-olas were among the first reactions to appear. 25-hydroxylase activity, paraestrolA,beingthefirstmemberofthisclasstobesynthesized which is reported in vertebrates, arthropods, annelids, and mol- denovoandexperimentally characterized. lusks, is the most widespread in bilaterians (28). However, the side-chain cleavage mechanism involves the formation of 20- or 22- Ancestral steroid receptors bound and activated hydroxycholesterol as intermediates (29). Consequently, side-chain by paraestrol A cleavage activity could derive from a 20- or 22-hydroxycholesterol, We synthesized paraestrol A from cholesterol acetate in six steps and the ancestral aromatized steroid might have been hydroxylated (seeFig.4AandtheSupplementaryMaterials).Twodifferentversions at one or many of those four positions. For the sake of simplicity, ofthereceptorwerereconstructed(Fig.4B),takingintoaccountalter- we tested the ligand-binding ability of an aromatized steroid with a nativepossibilitiesconcerningtherelationshipsbetweensteroidre- sidechainandwithnohydroxylation.Wedefine“paraestrols”asste- ceptorsfromchordatesandthosefrommollusksandannelids(6,30). Markovetal.,Sci.Adv.2017;3:e1601778 31March2017 4of13 SCIENCE ADVANCES | RESEARCH ARTICLE D o w n lo a d e d fro m h ttp ://a d v a n c e s .s c ie n c e m a g .o rg o/ n N o v Fig.4. ParaestrolAsynthesisandbindingassays.(A)SyntheticparaestrolAwasproducedinvitrousingasix-stepprotocol,startingfromcholesterolacetate(see em theSupplementaryMaterials).(B)Twoancestralsteroidreceptors(AncSR_ACandAncSR_D)werereconstructed,takingintoaccountphylogeneticuncertaintyregard- be ingthepositionofannelidandmolluskreceptors(seesequencedatasetinfig.S3andtableS1).AR,androgenreceptor;PR,progesterone;MR,mineralocorticoid r 2 1 rtehceelputcoifre;rGaRse,grelupcoorcteorrtgiceoniedwreacsetpetsoter;dSiRn,tsrtaenrsofiedctreedcehputmora;nEReRm,bersytorongicekni-drenleayte(HdErKe)c2e9p3toTrc.e(CllstwoiEth)TinhcereaabsiinlitgycoofnAcnencStrRa_tiAoCnsaonfdpAarnaceSsRtr_oDlAto(1a0c−t6i,v1a0te−5t,haendtra5n×sc1r0ip−5tioMn)oorf , 20 1 estradiol(10−8to10−6M).Humanestrogenreceptora(ERa)wasusedascontrol(Ctrl).Activationfoldsareinarbitraryunits,andtheirabsolutevaluescannotbecompared 8 fromonereceptortotheother.(FtoH)ThebindingabilityofAncSR_ACandAncSR_DforparaestrolAandestradiolwastestedbylimitedproteolyticassays.Lane1, undigestedprotein;lanes2to8,digestedproteinwithethanolasnegativecontrol(lane2),paraestrolAatincreasingconcentrations(lanes3to5at10−5to10−3M,respec- tively),andestradiol(lanes6to8at10−8to10−6M,respectively). Onepossibility(Fig.4B,left)isthatthegroupsNR3AandNR3Care logenyisstillbeingdebated(6,30),wepredictedthesequenceofthe theresultofachordate-specificancestralduplication.Inthiscase,the ligand-bindingdomains(LBDs)foreachofthetwopossibleancestors ancestralsteroidreceptorwouldbethelastcommonancestorpredat- (Fig.5andtableS2)usinganalignmentof142proteins(tableS1).When ingthischordate-specificgeneduplication(AncSR_AC).Consequent- comparedtootherreconstructionsatthesamenode(25,26,31),both ly, members of the NR3D group, classically defined as invertebrate ancestralproteinswerereconstructedwiththesameconfidencelevel, ERs, would be orthologous to the last common ancestor of both es- withameanposteriorprobability(PP)of0.76onthewholeLBD.The trogen and ketosteroid receptors. The alternative possibility (Fig. 4B, confidenceisevenhigherconcerningthe18sitesthatareinvolvedin right)isthatthelastcommonancestorfromthechordateNR3Agroup ligandbinding(PP=0.94forAncSR_DandPP=0.93forAncSR_AC). wasorthologoustothecommonancestorofNR3Dfromannelidsand These values are similar to another reconstruction that used a larger mollusks. This implies that the gene orthologous to the chordate datasetincludinganalysisofinternalvertebratebranches(25).Inthat NR3Cwassecondarilylostinannelidsandmollusks.Inthatcase,the alternative reconstruction, corresponding to our AncSR_AC, the ancestralchordatesteroidreceptorwouldalsobetheancestorforthe mean PP for the LBD is 0.70 and increases to 0.90 for the ligand- annelidandmolluskNR3D(AncSR_DinFig.4B).Becausethisphy- bindingsites.OurreconstructionsandthatofEicketal.(25)areidentical Markovetal.,Sci.Adv.2017;3:e1601778 31March2017 5of13 SCIENCE ADVANCES | RESEARCH ARTICLE D o w n lo a d e d fro m Fig.5. Sequencealignmentsoftheligand-bindingpocketsofancestralreceptorscomparedtohumanERa.a-HelicesthatmakeuptheLBDareboxedand http numberedfrom1to12onthebasisofthecrystalstructurefromERa(81).AminoacidsfromhumanERamakingdirecthydrogenbondswithestradiolarehighlightedin ://a green.Aminoacidsmakinghydrophobicbondswithestradiolarehighlightedinpink.Aminoacidsknowntobeinvolvedincoactivatorinteractionareindicatedwitha d v starontopofeachsite.DifferenceswithhumanERaonligand-bindingorcoactivator-bindingsitesarehighlightedinyellow.SequencesofAncSR_ACandAncSR_D a n havebeeninferredinthisstudy.Theotheronescomefrompreviousanalyses. ce s .s c ie n with regard to the ligand-binding pocket (Fig. 5). There is an 82% limited proteolytic assays (Fig. 4, F to H). This invitro methodtests ce m identityamongthethreereconstructedsequencesforthe230aligned whetherthebindingofaligandcaninduceaconformationalchange a g sites. Therefore, we assume that those small variations will have no inthe receptor LBD, causing a partial protection to trypsin prote- .o significant effectonthebiochemical activityofthereceptor. olysis. With estradiol, ERa and the two ancestral receptors are pro- org/ UsingthesynthesizedAncSR_ACandAncSR_Dsequences,we tected from proteolysis (Fig. 4, F to H). However, with paraestrol A, n N testedtheabilityoftheresultingGAL4-LBDfusionproteintobind onlyAncSR_ACandAncSR_Dareprotected(Fig.4,GandH).These o v paraestrolAandtoactivatetranscriptioninaligand-dependentman- resultsareconsistentwiththetransactivationassaysandconfirmthat e m ner (Fig. 4, C to H). We performed transient transactivation assays paraestrol Aisabonafideligandforthe ancestralsteroidreceptor. b e usingaUAS-tk-luciferasereportersystemandconstructsencodingfu- r 2 1 sionsbetweentheLBDofthereceptorsandtheGAL4DNAbinding Paraestrol A side-chain wobbling in the ancestral , 2 domain (GAL4-LBD). For comparison, we performed assays using receptor pocket 01 humanERa andestradiol.WithparaestrolA,bothancientreceptors To obtain a deeper understanding of the ligand-receptor interaction 8 AncSR_ACandAncSR_Dareabletoactivatetranscriptionofthere- andtocheckwhethertheconformationofparaestrolAiscompatible porter gene in a dose-dependent manner (Fig. 4, D and E). Human withbindingtothereceptor,weperformedliganddockingonathree- ERaismuchmoresensitivetoestradiolthantoparaestrolA(Fig.4C). dimensional (3D) model of the reconstructed ancestral receptors ThisisalsotruefortheAncSR_ACreceptor,althoughtherelativedif- (Fig. 6). Because these reconstructions are known to be sensitive to ferencebetweenbothligandsissmallerwhencomparedtoERa(Fig.4, the initial template choice(35),westartedwithtwo different tem- CandD).Thus,AncSR_ACbehavesasaclassicligand-dependenttran- plates:humanERa[ProteinDataBank(PDB)code,1ERE]andthe scriptionfactor.Bycontrast,AncSR_Ddisplaysaconstitutivetranscrip- ancestralcorticoidreceptor(AncCR;PDBcode,2Q3Y).Consistent tionalactivityintheabsenceofligand(Fig.4E).Thehighconstitutive with the principle of template influence, the inferred 3D structure activityobservedwithAncSR_Disreminiscentoftheconstitutiveactiv- wasfoundtohaveahighlysimilartopologyforbothAncSR_ACand ityoftenobservedwithERb(32,33),incontrasttoERa(34,35).How- AncSR_D[rootmeansquaredeviation(RMSD)=0.002whenmodeled ever, with paraestrol A at the micromolar range or with estradiol, uponAncCRandRMSD=0.006uponERa]modeledfromthesame AncSR_Dactivity can be further increased (Fig. 4E). Together, these template,whereasthedistancewashigherforbothreceptorscomparing resultssuggestthatparaestrolAbehavesasanactivatoroftheances- themselves upon modeling on the two different templates (RMSD = tralsteroidreceptor. 1.124betweenbothAncSR_ACmodelsandRMSD=1.117).Despite Theactivity detected using transactivation assays couldbe due to these differences, the main features of the binding interaction are si- themetabolictransformationofparaestrolAinmammaliancells.To milarinbothcases.ParaestrolA(showningreen)isdockedinside check whether it really binds to ancestral receptors, we performed theligand-bindingpocketofAncSRwithasteroidcoreconformation Markovetal.,Sci.Adv.2017;3:e1601778 31March2017 6of13 SCIENCE ADVANCES | RESEARCH ARTICLE bear hydroxylated carbons that form additional stabilizing interac- tions with other residues in the ligand-binding pocket. This is im- possible with paraestrol A that bears no hydroxyl group on the side chain. Moreover, having a single additional hydroxyl group on the sidechainofparaestrolAwouldnotnecessarilyincreasethebinding affinity.Thegaininbindingenergyiscounteractedbyhigherdesolva- tioncostinthecaseofthe25-OHgroupinecdysteroids,whereasthe presenceofadjacent20-and22-OHmoietieshasalowerdesolvation costbecauseofstabilizingintramolecularinteractionsbetweenthead- jacenthydroxylgroups(37).Thisphenomenoncouldalsoexplainthe observedmicromolaraffinityof27-hydroxycholesteroltohumanERs (41, 42). These docking studies demonstrate that paraestrol A is an excellentproxyforanancestralsteroid,supportingthehypothesisthat asteroidfromtheparaestrolfamilywaslikelythestartingpointfrom which vertebrate-specific sex and adrenal steroids evolved. Further- more,thismodelprovidesastructuralexplanationofhowside-chain cleavage would increase the affinity during the transition from para- estrols to estrogensas ligands forsteroidreceptors. D o w n Fig.6. ParaestrolAdockingintothepocketoftheancestralsteroidrecep- Dating the origins of steroid synthesis pathways loa tor.(A)OverallviewofthehomologymodelofAncSRLBDbasedonthecrystal Aspreviouslydevelopedfortheuniversalmetabolism(18,19),the de (sgtrruecetnu)reisodfoEcRkaedLBinDsiidnecoanmdpbleixndwsittho1t7hbe-eresctreapdtioorl(tPhDroBucgohdteh,e1EreRsEi)d.uPeasraEe5s0troanlAd cdlaatdeistthicemapeptheoardantrcaecionfgbmioectahbeomliiccaplasthtewpasyreevlaotliuvteiotnoaelascohalolothwesr.uWstoe d from R91.Similarresultsareobtainedusingthecrystalstructureoftheancestralcor- usedsynapomorphiesrelatedtofamiliesofreactionstoassigndiffer- h tpiclaotied(rleigchetptoorraningecotmrapclee)x.wBoitthhdterascoexsycaorreticforostmeroAnnecS(PRD_BD,cwoditeh,2thQe3Yt)raacsesaftreomm- ethnatttiemnzeypmeeriospdesctiofictihtyethraeseebvroanlvcehdesfr(ofmig.lSo4w).-sOpuecrifaipciptyroparcohteaisnssumcaets- ttp://ad AncSR_ACbeingidentical.(B)Detailedviewonthebindingpocket.Compared v alyzingawholerangeofactivitiesatlowlevelstoenzymesubfamilies a toofbthinedbiningdHin5g24offr1o7mb-tehsetraredcioelp(toorratnhgreo,uignhseitt)s,1p7abra-heystdrroolxAylalatecdksctahrebopno.s(sCib)iBliety- with potent and highly specialized activities (43–46). If this interpre- nces causeoflackofbindingtoH210,theresiduehomologoustoH524inERa(orange, tationiscorrect,thentheputativecommonancestryofpathwayscan .sc inset),thealiphaticchainofparaestrolAwobblesinsidethepocket.Afewamong bepostulatednotonlyonthebasisofsharedenzymeswithhighspe- ien manypossibleconformationsarerepresentedbydifferentcolors(encircled). cificities(suchasreaction31atnodeBinFig.3)butalsoonthebasis ce m of very similar reactions (type II homologies). Thus, if downstream a g branches that do not bear changes in type II homologies follow a .o similartothatadoptedby17b-estradiol(Fig.6B,orange)intheERa branchofthetree,thenthosedownstreambranchesareofthesame org/ ligand-binding pocket. In particular, the polar residues E50 (in H3) period as the main branch. When a new type II homology occurs n N andR91(inH5)ofAncSR_ACandAnc_Dinteractinanidentical onabranch,itdefinesthenextperiod(forexample,nodeB;fig.S4; o mannertotheERaLBD/17b-estradiolstructure,wheretheseresidues see also Materials and Methods). According to this method, type II ve m anchortheligandthroughinteractionswiththe3-OHhydroxylgroup homologies distinguish between specific time periods during which b e fromthearomatizedA-ring.Incontrast,the17b-OHgroupofestra- innovations—in terms of enzymatic involvement in steroidogenesis r 2 diolisreplacedbyalongaliphaticchaininparaestrolA,whosecon- —haveoccurred.Onthebasisofthis,wedefinedsevenrelativetime 1, 2 formationisshownheretobesimilartothatadoptedbyponasterone periods (Fig. 7Bandfig.S5). 01 8 or 20-hydroxyecdysone bound to the ecdysone receptor (1R1K and Theories on biochemical pathway evolution (Fig. 7A) predict 2R40)(36,37).Theexactconformationofthesidechainfrompara- thatanabolic pathwaysevolvebackward(47),which hasbeen ver- estrolAcannotbepredictedbecauseitisflexibleinnatureandislikely ifiedforglycolysis(48)orforthemandelatepathway(49),whereas toadapttolocalstructuralconstraints(Fig.6,BandC).Inparticular, catabolicpathwaysevolveforward(50),whichhasbeenproven,for thecomparisonbetweenthetwo typesof homology modelssuggests example,fortheureacycle(51).Thedefinitionofperiodsallowedus thattheregionincludinghelixH6andloopsH6-H7andH11(arrows) to decipher the order of appearance of different steroidogenic path- isveryflexibleandcaneasilyaccommodatetheligandtaillocatedin ways (Fig. 7B and fig. S5). Starting from the cholesterol synthesis thisregion.Otherstructuralstudiesindicatethatsyntheticligandside backbone (Fig. 7B, period 1, red), the first pathway to appear was chains that are even bulkier than the side chain of paraestrol A can the synthesis of 24- and 25-hydroxycholesterol (period 2, orange), inducethe formationof anextendedsubpocket in theLBD(38,39). followed by the synthesis of 7-hydroxycholesterol (period 3, brown), Furthermore,thisgivesastructuralexplanationforthesmalleraffinity and later by the synthesis of vitamin D3, D4-dafachronic acid, ecdy- ofparaestrolArelativeto17b-estradiol:Theabsenceof17b-OHgroup sone, progesterone, and 11-deoxycortisol (period 4, green). Cheno- inparaestrolAtakesawaythepossibilityofformingahydrogenbond deoxycholic acid and the bulk of vertebrate sex and adrenal steroids with H210 (homologous to H524) that stabilizes the interaction be- alsoappearedduringthesameperiod(period5,blue).Alaterphaseof tween 17b-estradiol and its receptor. Notably, interaction with H524 diversification led to the synthesis of chenodeoxycholic acid through is not mandatory for ligand binding to E2. Some synthetic agonists, twoadditionalpathways:cortisoneand11-ketotestosterone(period6, suchastamoxifenorothernonsteroidalcompounds,alsobindtoER purple).Thelastpathwaytoappearwasthesynthesisofcholicacid(pe- without interacting with H524 (39, 40). However, those molecules riod 7, black). Because the more upstream reactions inthe pathways Markovetal.,Sci.Adv.2017;3:e1601778 31March2017 7of13 SCIENCE ADVANCES | RESEARCH ARTICLE D o w n lo a d e d fro m h ttp ://a d v a n c e s .s c ie n c e m a g .o rg o/ n N o v e m b e r 2 1 , 2 0 1 8 Fig.7. Sequentialorderingofanimalsteroidmetabolicpathways.(A)Twotheoreticalmodelsaboutmetabolicpathwayevolution:anabolicpathwaysevolving backward,withthemoredownstreamreactionsappearingfirst(47),andcatabolicpathwaysevolvingforward,withthemoreupstreamreactionsappearingfirst (50).(B)Sequentialviewofpathwayappearanceduringsuccessiveperiodsinanimalsteroidogenesis.Colorsindicatesuccessiveperiodscorrespondingtotheappearanceof newtypesofenzymaticactivities.Finalcompoundsappearingateachperiodareindicated.Allcompoundsappearinfig.S5.MyB.P.,millionyearsbeforethepresent. werethefirsttoappear,steroidogenesisthusevolvedinaforwarddi- withindetoxificationenzymesinphylogenetictreesbasedongenomic rection,throughasuccessiveappearanceofstepsinsuccessivegenera- data(6).Accordingtothisview,steroidogenesisderivedfromcatabolic tionsofanimals.Thisresultisconsistentwithadetailedanalysisofthe pathways that are implicatedin xenobiotic detoxification and steroid distribution of bile acids across vertebrates, suggesting that the bile hormonescouldbeviewedasco-optedcholesterolmetabolites(54). acid synthesis pathway evolved from C27 bile alcohols, structurally Differencesinsteroidmetabolismamongmetazoansenableone equivalenttooxysterols,toC24bileacidsthatnecessitatefurthercat- toputtemporallimitstosomeofthedefinedperiods(fig.S4).Onthe alyticstepstobemade(52).Thus,fromanevolutionaryperspective, basisoftheassumptionthatapathwayspecificforagivencladecan- steroidsynthesisshouldbeviewedasacholesteroldegradationpath- not be older than the oldest member of this clade, we identified five wayratherthanasasynthesispathway,assuggestedforbileacids(53). chronologicalanchorpointsthatallowrootingoftherelativeperiods Thisisalsoinlinewiththenestedpositionofsteroidogenicenzymes inanabsolutetimeframe.Afterinitialdivergencefromfungi,around Markovetal.,Sci.Adv.2017;3:e1601778 31March2017 8of13 SCIENCE ADVANCES | RESEARCH ARTICLE 1100to900MyB.P.,thefirstbilaterian-specificsteroidogenicpathway testosterone, or any side-chain intermediate as precursors, and this cannotbeolderthanthedivergencebetweenprotostomesanddeuter- gives the possibility for coexistence of progestagens and paraestrols ostomes,estimatedto733to643MyB.P.(fig.S4,nodeA)(55).The in basal vertebrates before the emergence of modern estrogens (Fig. presence of a clade of vertebrate side-chain steroidogenic pathways 8B).Notably,if17b-estradiolandsimilar aromatizedsteroidsareca- (fig.S4,nodeB)indicatesthatthelastcommonancestorofgnathos- nonical estrogens that act as high-affinity ligands for ERs, then it tomesandlampreyswasalreadyabletosynthesizeatleastonesexor becomesincreasinglyclearthatothernonaromatizedsteroidscouldal- adrenal steroid. This defines the second chronological anchor point so be potential physiological ER ligands (65, 66). For example, 5a- between 500 and 475 My B.P. (56). Then, the elasmobranch-specific androstane-3b,17b-diol(alsocalled3b-Adiol)andD5-androstenediolbind abilitytomake1a-hydroxysteroid(fig.S4,nodeK)appearedafterthe ERa and ERb with nanomolar affinity (67,68). Synthesis of both com- divergence between chondrichtyians and osteichtyians 423 My B.P. poundsdoesnotrequirearomatase,andsynthesisofD5-androstenediol (57).Finally,synthesisoftheteleostean-specific11-ketotestosterone doesnotevenrequire3b-HSDactivity.Morerecently,27-hydroxycholesterol (fig. S4, node L) could not be older than the split between teleosts hasalsobeenfoundtobeanendogenousligandfortheER,bindingit andtetrapods 419MyB.P.(57).Datingtheappearanceofnewen- withmicromolaraffinitythatisphysiologicallyrelevant(41,42).Synthe- zymaticactivitiesandnewsynthesispathwaysalsoenablesonetoin- sis of 27-hydroxycholesterol represents the simplest possible synthesis tegratethesetsofreceptorsandenzymesthatwerepresentatthesame pathways, with just one hydroxylation step from cholesterol. There- timepointonthebasisofgenomicanalyses (6,58,59). fore, it would be extremely interesting to delineate the contribution ofthesealternativeligandsbetter,whichmaycorrespondtomorean- cientmoleculesthatareactivein theestrogenicpathway. D o DISCUSSION Itwaspreviouslysuggestedthatthemostabundantmetabolites w n Beyond molecular actualism: Ancient metabolic pathways can act as helping hands for pathway evolution, providing readily lo a as a helping hand for enzyme recruitment availablesubstrates fornewlyarising catalyticactivities (69). Here, de d Aprlothvoidueghcotnhseisftoernwtaerxdpalanndabtiaocnkswfaorrdtmheoodreilgsinofomfemtaublotilpicleevveorluytaionn- weexpandthisviewwithadetailedcasestudyshowingthatnewly from cientpathwaysrelatedtotheuniversalmetabolism,mostofthelate h mpaettcahbwoolirckpeavtohlwutaiyonevdouleuttioonmiuslttihpoleurgehctrutoitmhaevnetsoocfcuenrrzeydmtehsrofruogmh ttp://ad v preexisting pathways (60). There is an apparent contradiction be- a n tweenthestrictlyforwardpatternthatweobservefortheevolution ce s ofcurrentsteroidsynthesispathwaysandthegenomicandbiochem- .s c ical data. If some enzymatic activities (for example, aromatization) ie n wererecruited,whichisconsistentwithourparsimonyanalysis,then ce m it would fit a patchwork evolution model. This contradiction can be a g resolvedbyintegratingreceptorsandparaestrolsintoanevolutionary .o model(Fig.8).Atthetimewhenoxysterolsynthesisappearedinbasal org/ chordates(correspondingtoperiods2and3),oxysterolscouldprob- n N ably bind to already existing ancestral receptors from the NR1H/I o v group.Weproposethat,duringthisperiod,theligandsfortheAncSR e m were paraestrols synthesized by the ancestral CYP19 aromatase and b e other enzymes, such as an ancestral HSD3B (Fig. 8A). Later, during r 2 1 period4,theappearanceoftheCYP11Aside-chaincleavageenzyme , 2 (Fig. 8B) was concomitant with the appearance of progesterone syn- 01 8 thesis in basal vertebrates (Fig. 8B), which is also consistent with the binding of progestagens to the unduplicated member of the NR3C family in lampreys (61). Finally, during period 5, opportunistic branching of the CYP19A aromatase from C27 sterols to side-chain cleaved steroid substrates led to the synthesis of modern estrogens andsubsequentextinctionoftheparaestrolpathway(Fig.7C).Explain- ingthiscompletereplacement,insteadofmaintainingbothpathwaysin parallel, requires an understanding of which molecular properties allowed estrogens to outcompete paraestrols (62). Our binding assays Fig.8. Originofvertebrate estrogens throughopportunisticconnections and structural modeling data suggest that side-chain cleavage helped fromancientparaestrolsynthesispathways.Coevolutionbetweenmetabolic toincreasethebindingaffinitytoERs.Thisgaininaffinitymayhave pathways,incolorscorrespondingtodifferentperiods(AtoC)andsteroidrecep- facilitatedthetransitionfromapromiscuousfoodsensingsystemtoa torsoftheNR1(gray)andNR3(black)subfamilies.Weinferthat,inadditiontothe sequentialappearanceofactualmetabolitesatdifferenttimesteps(periods2and highly specific coupling between nutritional status and reproduction (63). This model provides an alternative to the “ligand exploitation 3,oxysterols;period4,progestagens;period5,androgens),thehistoryofNRligand synthesispathwaysinvolvedmetabolitescombiningfeaturesthatareseparatedin model,”whichwasbuiltontheassumptionthatestradiolwasthefirst actualmolecules,suchasparaestrolsinperiods2to4.Thoseintermediarymetabo- estrogen, and tried to explain how intermediates in its synthesis-like litesmayhavefacilitatedagradualshiftinsubstratespecificityforancientenzymes, progesterone or testosterone existed long before their receptor suchastheCYP19Aaromatase,andhencethebuildingofnewpathwaysthrough appeared (64). Paraestrols can be synthesized without progesterone, opportunisticconnectionsofoldenzymeswithnewsubstratesduringperiod5. Markovetal.,Sci.Adv.2017;3:e1601778 31March2017 9of13 SCIENCE ADVANCES | RESEARCH ARTICLE appearing metabolites can become substrates for older enzymes, the C-to-ECDY pathway (Fig. 2). Both reactions can be grouped leading to both expansion and extinction of pathways. Thus, just together ascharacter 147(Fig.2,green). asfossilsshareacombinationofcharactersthatareabsentinmodern Type hIIe homologies. organisms, we argue that the actual set of observable metabolites re- We defined here a fifth type of secondary homology (type “hIIe”) presentsonly asubsetofthefull metabolitespace.Our methodpro- forsharingofthesameenzymethatperformstwoslightlydifferent vides the possibility of defining new classes of molecules that can be reactions.Character151providesanexample,wheretheCYP27A1 helpfultoorientscreensinanalyticalchemistryandstudycoevolution enzymehydroxylatedeithercarbon25orcarbon26onthecholes- betweenproteinsandtheir ligands. terolsidechainduringthesynthesisofbileacidsandcalcitriol(fig. S2).Thereisariskassociatedwiththistypeofhomology:Theenzyme couldhavebeenrecruitedbyoneofthetwopathways.However,the MATERIALSAND METHODS abovereactionwassituatedveryearly(firststepofcholesterolhydrox- Taxonomic sampling ylation),favoringthehypothesisofaninitialversatilityoftheenzyme. Thisworkfocusedonmetabolicpathways,leadingtothesynthesis Mostofthecharacterscorrespondedtoasingleenzymaticreaction ofsteroidsthatareallendproductligandsofnuclearhormonere- and a single enzyme. Five characters corresponded to multiple en- ceptors:vertebratesexandadrenalsteroids,bileacids,oxysterolsand zymes (88, 93, 100, 110, and116 in figs. S1 and S5) involved in bile vitamin D, nematode dafachronic acids, and arthropod ecdysteroids. acidb-oxidation,whichisahighlyconservedprocessinwhichallthe Allthesemoleculesweresynthesizedfromacommonprecursor,choles- stepsoccurredinthesameorderforallbileacidprecursors.Therefore, terol,exceptvitaminD,inwhichsynthesisstartsfrom7-deoxycholesterol, codingthesestepsseparatelywouldoverweighthispartofthematrix. D o andthethreeoutgroupsthatarethesynthesispathwaysfromsqua- Some characters referring to type II homologies contain question w n lenetoergosterol,cholesterol,andsitosterol.Somepathwayexamples marksinthematrix.Theseareincludedwhenthepathwaydoesnot lo a aregiveninFig.2A.Acompletelistofthe72pathwaysisshownin exhibittheappropriatesubstrateforthereaction,orwhencodingthe de d fig. S1A. rtieoanctfiroonmisamkeeatonnineginlesbs.oFnocraerxbaomnp1l7e,(creoadcitniognth1e29p)ohssaisbialimtyeoafnainrgedouncly- from Characters and homologies forpathwayswherethereis,atonesteporanother,aketoneonthis h Astacteosmapreleitnedliisvtidoufaclhraeraaccttieornssis(osrhotywpnesinoffirge.aSc1tiBon.Cs)ocmatpaalyrzeeddcbhyariadcetner- ctiaornbopno,inantdofisthmeesaindiengclheasisnfo(rsupcahthawsaiynsewcdhyerseoncearsbyonnth1e7siiss).thejunc- ttp://ad v tified enzymes shared by pathways. They were defined according to a n fourhomologycriteria,asillustratedinFig.2andfig.S2.Thecriteria Phylogenetic reconstruction ce s for formulating a primary homology are as follows: shared specific Tree search. .s c enzymaticactivity(I),sharedenzymaticfunctionwithoutsharedspec- The matrixcontains72taxa(fig.S1A)and151characters(fig.S1C). ie n ificity for a substrate (IIa), shared functional family (IIc), and a new Seventy characters are parsimony-informative. The matrix was ana- ce m kindofhomologythattakesintoaccountthesharingofthesameen- lyzed using TNT version 1.1 (70), made freely available by the Willi a zyme,whichperformstwoslightlydifferentreactions(IIe).Type“hIIb” HennigSociety.Charactersweretreatedasunorderedandunweighted g.o and“hIId”homologiesdefinedpreviously(18)arenotrelevanthere. in the search for the most parsimonious tree. Tree searches were org/ Type hI homologies. carried out using the tree bisection-reconnection algorithm and the n Primary homologies of “type hI” were defined as sharing the same parsimonyratchetforamoreextensiveexplorationofthepossibletree No v reaction, with absolute specificity for the substrate, for two or more space(71).Memorywassettoholdupto1,000,000trees,withagradual e m different pathways. For example, cholesterol side-chain cleavage by increasefrom10,000to150,000iterations,toensurethatthenumberof b e CYP11A1intopregnenolone(character 31)occurredinthepathway most parsimonious trees was decreasing or null. For the parsimony r 2 1 transformingcholesterolintoestrone(C-to-E1,redoval;Fig.2)andin ratchet,theprobabilityofupweightingwassettovaryfromthedefault , 2 thepathwaytransformingcholesterolintoestradiol(C-to-E2,redoval; 4%upto50%,underasearchof10,000iterations,tocheckthatchanges 01 8 Fig.2).ThesetwopathwayssharedtheCYP11A1enzyme,withoutdif- inweightingdonotinfluencethelengthoftheoutputtree.CIandRI ferenceinspecificityforitssubstrate.Inotherwords,thespecificityis were calculated as indicated by Farris (72). Check for rogue taxa was takeninto accountwhendefiningtypeI homologies. carried out using the prunnelsen option, which counts the number Type hII homologies. ofnodesgainedineitherthestrictorsemistrictconsensusandreports Type hII homologies corresponded to a more relaxed homology cri- those prunings,which improve theresults(70). terion,wheretworeactionsdidnotsharethesamemoleculeasasub- Rooting. strate.Threesubtypes (a,c,ande)wereusedhere. Synthesis pathways from squalene to cholesterol, ergosterol, or sito- Type hIIa homologies. sterol (73) were used as outgroups, assuming that the synthesis of Primaryhomologiesof“typehIIa”occurredwhentwopathwaysshared sterol precursors by various eukaryotes is more ancient than that of similarenzymaticfunctionwithoutconsideringtheirrespectivespecific- steroidhormonesfrom thoseprecursors in animals. ity for asubstrate. This was the case for aromatizationof androstene- Defining time span criteria. dione to estrone (character 54 in C-to-E1; Fig. 2) or aromatization of Fromtheroottothetipofbranches,phylogenetictreesprovideare- testosteronetoestradiol(character55inC-to-E2;Fig.2).Bothreactions lative order of transformations (referred to as enzymatic innovations canbegroupedtogetherascharacter134(Fig.2,blue). here)throughtime.Wecalladeep,moreinclusiveinternalbranchor Type hIIc homologies. cladean“upstreamnode”andamoreterminal,lessinclusiveinternal Primaryhomologiesof“typehIIc”areforsharedfunctionalfamilyof branchorcladea“downstreamnode.”Timespansinmetabolismare enzymaticreactions,suchashydroxylationsoneithercarbon25ofthe defined as the time along the tree separating two type II character sterolskeletonintheC-to-CHENOpathway(Fig.2)orcarbon20in changes.Thiscriterionisbasedontheassumptionthattheapparition Markovetal.,Sci.Adv.2017;3:e1601778 31March2017 10of13
Description: