Carteretal.BiologyDirect2014,9:11 http://www.biologydirect.com/content/9/1/11 RESEARCH Open Access The Rodin-Ohno hypothesis that two enzyme superfamilies descended from one ancestral gene: an unlikely scenario for the origins of translation that will not be dismissed Charles W Carter Jr1*, Li Li1, Violetta Weinreb1, Martha Collier1, Katiria Gonzalez-Rivera1, Mariel Jimenez-Rodriguez1, Ozgün Erdogan1, Brian Kuhlman1, Xavier Ambroggio2, Tishan Williams1 and S Niranj Chandrasekharan1 Abstract Background: Because amino acidactivationis rate-limiting for uncatalyzed protein synthesis, it is a key puzzle in understanding theorigin of the genetic code. Two unrelated classes (Iand II)of contemporary aminoacyl-tRNA synthetases (aaRS) now translate the code. Observingthat codons for the most highly conserved, Class Icatalytic peptides, when read in thereverse direction, are very nearly anticodons for ClassII defining catalytic peptides, Rodinand Ohnoproposed that thetwosuperfamiliesdescendedfrom opposite strands ofthesame ancestralgene. Thisunusualhypothesis languished for a decade, perhaps because it appearedto be unfalsifiable. Results: The proposed sense/antisense alignment makes important predictions. Fragments that align inantiparallel orientations, and containthe respective active sites,should catalyze thesame two reactions catalyzed by contemporary synthetases. Recent experiments confirmed that prediction. Invariant cores from both classes, called Urzymes after Ur=primitive, authentic, plus enzyme and representing~20% ofthecontemporary structures, can be expressed and exhibit high, proportionate rate accelerations for both amino-acid activationand tRNA acylation. A major fraction (60%)of the catalytic rate acceleration by contemporary synthetases resides in segments that align sense/antisense. Bioinformatic evidence for sense/antisense ancestryextendsto codons specifyingtheinvariant secondary and tertiary structures outside the active sites of thetwo synthetase classes. Peptides from a designed, 46-residue gene constrainedby Rosetta to encode Class I and II ATP binding sites with fully complementary sequences both accelerateamino acid activation by ATP ~400 fold. Conclusions: Biochemical and bioinformatic results substantially enhance the posterior probability that ancestors ofthe two synthetase classes arose from opposite strands of thesame ancestral gene. The remarkable acceleration byshort peptides ofthe rate-limiting step in uncatalyzed protein synthesis,together with thesynergy ofsynthetase Urzymes and theircognate tRNAs, introduce a new paradigm for theorigin ofprotein catalysts, emphasizethe potential relevance of anoperationalRNA code embedded inthe tRNA acceptor stems, and challenge the RNA-World hypothesis. Reviewers: This article was reviewed byDr. Paul Schimmel(nominatedby Laura Landweber), Dr. Eugene Koonin and Professor David Ardell. Keywords: Aminoacyl-tRNA synthetases, Urzymes, Genetic code, Origin ofTranslation, RNA World hypothesis, Amino acid activation, Structural homology, Ancestral genes, Sense/antisense coding *Correspondence:[email protected] 1DepartmentofBiochemistryandBiophysics,CB7260UniversityofNorth CarolinaatChapelHill,ChapelHill,NC27599-7260,USA Fulllistofauthorinformationisavailableattheendofthearticle ©2014Carteretal.;licenseeBioMedCentralLtd.ThisisanOpenAccessarticledistributedunderthetermsoftheCreative CommonsAttributionLicense(http://creativecommons.org/licenses/by/4.0),whichpermitsunrestricteduse,distribution,and reproductioninanymedium,providedtheoriginalworkisproperlycredited.TheCreativeCommonsPublicDomain Dedicationwaiver(http://creativecommons.org/publicdomain/zero/1.0/)appliestothedatamadeavailableinthisarticle, unlessotherwisestated. Carteretal.BiologyDirect2014,9:11 Page2of23 http://www.biologydirect.com/content/9/1/11 Open peer review Reviewed by Dr. Paul Schimmel (nominated by Laura Landweber), Dr. Eugene Koonin and Professor David Ardell. For the full reviews, please go to the Reviewers' Reportssection. Dedication “…thereisnosinglepathtocreativity. Weare constrained notbythenecessarydisciplineofrigor butbythelimitsofourown imaginationsand intellectual courage. Inthe wordsofJazzmusician FatsWaller,Daretobewrongoryou mayneverbe Figure1Aminoacyl-tRNAsynthetaseshavethreeimportant right.” functionsinthecell.TheyuseATPtoactivatethealphacarboxyl group,makingtheaminoacidmorereactive(Activation).This - J.MichaelBishop[1] activationstepaddstheadenosinemoietyasan“affinitytag”tothe aminoacid,increasingitsbindingaffinityby~1000-fold(Retention) Thisslowsthereleaseofaveryreactivespeciesandenables subsequentstepsthatenhancethefidelityofthefinalstep.Finally, “Howoften haveIsaid toyouthatwhen youhave theycatalyzetransferoftheactivatedcarboxylgrouptothe3’CCA eliminatedtheimpossible, whateverremains,however terminusoftRNA(Acylation),completingthetranslationofthe improbable, mustbethetruth?” geneticcodebythecovalentlinkagetothetRNAanticodonthatis interpretedbythe30Sribosomalsubunitinresponsetocodonsin mRNA.Theapproximaterateaccelerationsachievedbycontemporary -SirArthurConanDoyle [2] enzymesindicatedarebasedoncomparisonsofkcat/K valuestothe M uncatalyzedrates,estimatedfromexperimentalratesofmodel Sergei Rodin (1947-2011) was both a mentor and a reactions,asdescribedin[38,39]. collaborator. When the paper that launched this work [3] was challenged [4], Sergei was so incensed that he andhisson,Andrei,wroteabrilliant,rebuttalonourbe- NTP hydrolysis into the highly unfavorable bond forma- half [5].Thus,Ialsoconsideredhimmyfriend. tion leading to mixed anhydride aminoacyl-5’adenylates, The origin of the universal genetic code is one of the (ii) addition of a covalent tag to the amino acid, i.e., the most important, fascinating, and vexing questions facing adenosine moiety, greatly extending the residence time contemporary biologists. Sergei devoted much of his oftheactivatedaminoacidwithintheactivesitetoallow professionallifepursuingthisquestion usingseveralper- dissociation and/or hydrolysis of incorrectly activated spectives from “outside the box” [5-11]. One of his more amino acids [17,18], and (iii) specific acylation of tRNA unlikely hypotheses was the possible ancestry of Class I with cognate amino acid. Translation occurs in the third and Class II aminoacyl-tRNA synthetases as sense- and step[19-26]. antisense- gene products expressed from the same prim- Absent catalysts, aminoacyl-5’ adenylates form both ordial gene [11]. That hypothesis was an elegant very slowly and in very low equilibrium yield. The acti- realization of thoughts expressed by both Bishop and vation step proceeds 103-104 times more slowly than the Doyle. This review considers recent experimental sup- second in aqueous solutiona, and therefore requires a portfor,andimplicationsof,theirhypothesis. correspondingly more potent catalyst. Release and sub- sequent hydrolysis of the pyrophosphate leaving group Background are both necessary to ensure that activated amino acids The origin of life has challenged scientific investigation are formed in high yield. Further, once formed, the for a century or more, generating diverse opinions relat- aminoacyl-5’ adenylate is exceeded in reactivity only by ing to several disjoint questions about life’s definition in acyl-halides [27]. In fact, of all reactions involved in terms of compartmentation, metabolism, and informa- ribosomal protein synthesis and in both kinetic and tion transfer/storage [12-16]. A central issue linking thermodynamic terms,aminoacidactivation is,mechan- these questions is a transition from chemistry to biology. istically,byfarthemostchallenging. Our focus is the origin of codon-dependent translation. Just how far back in time the three functions can be Contemporary Aminoacyl-tRNA synthetases (aaRS) ac- tracedliesclosetotheheartofthecode’sorigins.Consen- celerate two successive chemical steps: amino acid acti- susholdsthataaRSenzymeshadessentiallyassumedtheir vation and tRNA acylation. These coupled tasks involve modernconfigurationsinthelastuniversalcellularances- (Figure 1):(i) transductionofthe chemicalfree energy of tor, LUCA [28-30]. It is certainly not idle speculation, Carteretal.BiologyDirect2014,9:11 Page3of23 http://www.biologydirect.com/content/9/1/11 therefore, that simpler ancestral aaRS preceded LUCA by binary codeoftwoaminoacid TYPESthat specify eons. Nevertheless, because LUCA represents a localized “inside”and “outside”seems torepresentalmost “Big Bang” [31,32], it was associated with intense genetic sufficientinformation(turnsexcepted)toencode exchange [33]. Thus, it becomes much more difficult to globularobjectswith selectablefunctionsand hence, trace phylogenetic lineages for either activity much be- tolaunchnaturalselection. Combinatoriallibraries yondthathypotheticallandmark.Onepossibleavenuelies ofpolypeptides basedonabinary,middle-base code in the identification and functional annotation of broadly thatdifferentiates onlybetweencoreand surface conserved tertiary packing motifs, illustrated pointedly aminoacidscontainhighproportionsofproducts [34] by a nearly invariant core packing motif belonging with thebiophysical characteristicsofmoltenglob- to~125 families in the Rossmannoid superfamily. That ules[45],and giverisetosignificantfunctionalities motif is associated with a discrete supersecondary struc- [46,47].Twodifferentkindsofsynthetaseswithrudi- turethatbindsATPandnucleotidesingeneral,inkeeping mentaryspecificitiesreflectedinmedian hydropa- with its possible role in primordial chemical free energy thies ofthe contemporaryClassIand IIaaRS conversion, and has been identified as a “protoallosteric substratesmight thushavebeensufficient tolaunch site” [35]. A related effort is the expression and engineer- codon-directedprotein synthesis. ing of invariant cores from enzyme superfamilies [36-40]. ii. Physicalchemistry.Amino acidsubstrates ofthetwo We call these constructs Urzymes, from Ur=primitive, classessortintojust suchadistinction.The original;theyareourcentralfocushere(seealso[41]). apparent symmetryrelatingthe three subclasses and theinordinatewaterpreference ofClassIarginine Results [48]maskan overwhelmingdifferencebetweenthe Aminoacyl-tRNAsynthetases:whytwofamilies? hydrophobiccharacter ofClassIAand ClassIIA The activation and acylation reactions are so important, aminoacids.Despiteexceptions,ClassIIamino and the twenty canonical amino acids so diverse, that na- acids generallyprefertheaqueous phase,ClassI tureinventedtwodifferentaaRSfamiliestoactivateall20 aminoacidsthe hydrocarbonphase.Theirmedian canonical amino acids. The two enzyme families differ freeenergiesoftransferof betweenwaterand markedly in primary, secondary and tertiary structure cyclohexanediffer by-4.6 kcal/mole[40].ClassI [42]. Amino acids activated by Class I and II aaRS divide (larger)andClassII(smaller) amino acidsarealso intoapparentlysymmetricalclassesof10eachb,withthree distinguished bysize. Solvent transferfreeenergy comparably sized subclasses, I,IIA(6), I,IIB(2-3), and I,IIC (P<10−7)and massdifference (P<10−4)contribute (1-2). synergistically(P<10−2)tothe solvent accessible Class I and II active sites exhibit several relevant surface areainfoldedproteins(Carter,CWJr& distinctions (Figure 2). First, the amino acid-binding Wolfenden, RtRNA Acceptor-Stemand Anticodon pocketsofClassIenzymesliedeepwithintheRossmann BasesFormIndependentCodesRelatedtoProtein dinucleotide-binding fold, whereasamino acids bind to a Folding,inpreparation); ClassIIamino acidside shallow crevice close tothesurfaceofClassIIenzymes. chainsare,onaverage,54%exposedwhereasClassI The most highly-conserved aaRS active-site amino aminoacidsare 32%exposed(P~0.03). acids occur in three sets of signature sequences. We’ll iii.Geneticlinkage.Apre-cellularworldpopulatedby focus on Class I HIGH and KMSKS and Class II Motif 1 quasispeciesmay haveplacedapremiumonefficient and 2.Withrare exceptions, conserved amino acids with informationstorage[49].Wefurtherbelievethat a a direct, catalytic role in Class I active sites are drawn substantialselectiveadvantagewouldarise if both from amino acid substrates activated by Class II en- requiredkindsofsynthetaseswerepresentatthe zymes, and conversely (Figure 2). It is hard to imagine same timeand place.Coding ClassIandIIon how this came about unless the evolutionary ancestors oppositestrandswouldlinkgenesforthe twoclasses appeared simultaneously, rather than sequentially, as is genetically, assuring thatwhen onewaspresent,so often argued[43,44]. wasthe other. In any case, sharply contrasting architectures render it inconceivable that the two classes had a common ances- Sense/antisenseancestry tor.Why,then,didnatureinventtwowaystodothesame Rodin and Ohno observed a statistically significant com- job? Three related aspects of the class division seem rele- plementarity between consensus coding sequences for vanttoansweringthisquestion: class-I defining PxxxxHIGH and KMSKS peptide signa- tures and Class II Motif 2 and Motif 1 sequences, and i. Coding simplicity.Theearliestproteins probably conversely [11]. They inferred from this that ancestral wereencodedbyamuch simpler geneticcode than Class I and II aaRS descended from opposite strands of the code of20aminoacids wehavetoday.Infact,a the same gene, a proposal we call the Rodin-Ohno (RO) Carteretal.BiologyDirect2014,9:11 Page4of23 http://www.biologydirect.com/content/9/1/11 Figure2DetailedcomparisonbetweenactivesitesofClassIandClassIIaminoacyl-tRNAsynthetases.SubstratebindingsitesforATPand aminoacidareburiedinClassI,andmuchclosertothesurfaceinClassIIenzymes.Thecoloroftheactive-sitespheresillustratesthatconserved active-siteresiduesthatmotivatedRodinandOhnotoadvancetheirsense/antisensecodinghypothesisaredrawnexclusivelyfromthesetof substratesactivatedbytheotherclass.ThemediantransferfreeenergyfromwatertocyclohexaneisfavorableforClassIsubstratesandunfavorable forClassIIsubstrates. hypothesis. Despite the strength of their statistical tests to perform similar analyses. Contemporary aaRS are mod- (vide infra), it was not obvious when the idea first ap- erately large enzymes, in keeping with their sophisticated peared that experiments could either falsify or confirm tasks. Their structures also exhibit considerable variation thisextraordinaryproposal.Intheinterim, however,spe- withinthetwoclasses(Figure3).Fromtheoutsidethefour cifying the hypothesis more precisely in terms of the re- ClassIandIImonomerssuperimposedineachpartofthe spective tertiary structures has clarified its more important figure look quite different. Inside, however, a much implications, opening experimental and bioinformatic ave- smaller, invariant core of ~120-130 amino acids is nearly nuestoassessitsvalidity. identicalinall10examplesofeachclass. Key to these new developments is the notion that con- Not surprisingly, cores for each class encompasses the temporary enzymes grew from ancestral forms similar to active sites for amino acid activation and acyl-transfer, invariant cores that can be identified within superim- and hence also the most highly conservedsecondary and posed members of protein superfamilies (Figure 3). tertiary structures. Urzymology consists of the methods These cores invariably comprise the active sites. They wehaveintroducedtoidentify,re-design,express,purify, can be expressed in soluble form, sometimes after re- characterize, authenticate, and exploit the properties of design to modify hydrophobic side chains at newly these cores (Table1). solvent-exposed surfaces [39]. They exhibit substantial Whereas invariant cores within each class are very rate accelerations for the reactions catalyzed by the en- similar to each other, they retain sufficient identity that zymes from which theyderive [37-40]. sub-class relationships are preserved. Four examples In the following, we first summarize the salient features (two from subclass A, and one each from subclasses B of Urzymology, the study of Urzymes. Then we describe and C) were superimposed by partial order structure howUrzymologyfacilitatesthemodulardeconstructionof alignment(POSA;http://fatcat.burnham.org/POSA;[52])). Class I TrpRS and Class II HisRS and recapitulation of Urzymes deduced for members of the same subclass (IA, their evolution. Finally, we summarize how these two IIA) were more similar to each other than to members of UrzymeshelpvalidatetheRodin-Ohnohypothesis. other subclasses, and the subclassesICand IIC (Figure4) weremoststructurallydistinct. Urzymology:structuralbiologyyieldsinsightsaboutthe We examine implications from structural biology for invariantcores the RO hypothesis in Figure 5. Class I and II invariant Comparativeanatomyhasalwaysbeenthesinequanonof cores themselves actually can be aligned sense/antisense phylogenic inference. Because our interest here concerns as noted by Rodin and Ohno [11]. Contemporary aaRS, molecules far earlier than LUCA, our approach begins however, violate this alignment in two ways. (i) Large and withstructuralbiologyand3Dsuperposition,whoseappli- variable insertion domains (in Class I these are called cation to aaRS has been reviewed [50]. We used a variety ConnectingPeptides1and2[53,54]),interrupttheactive- of manual least squares and automated algorithms [51,52] site alignment. (ii) Anticodon-binding domains in both Carteretal.BiologyDirect2014,9:11 Page5of23 http://www.biologydirect.com/content/9/1/11 Figure3SuperpositionoffourClassIandfourClassIIaaRS.Specificenzymesarecoloreddifferentlyandlabeled.Full-lengthcontemporary monomersareshownassurfacesthatare65%transparent,torevealtheinvariantcores,shownascartoonsinsidethesurfaces. Classes lie outside the range where antiparallel alignment AARSUrzymesbothactivate,andacylatetRNAwith, isevenpossible. cognateaminoacids. On the other hand, removing all insertion and Aminoacidactivation anticodon-bindingdomainsfrombothclassesleavesthe Two Class I (TrpRS and LeuRS) and one Class II potential sense/antisense alignment intact (Figure 5B). (HisRS) Urzymes accelerate the rate-limiting amino acid Remarkably, both invariant cores include complete ATP- activation reaction ~108-109-fold. These rates are within and amino acid-binding sites, together with rudimentary 10−5 (Figure 6; Figure 5 in [37]), so transition-state binding sites for the 3’ CCA termini of tRNA (Figure 5C; stabilization free energies of both Class IandII Urzymes [40]). aretherefore ~60%of,those offull-lengthaaRS. Figure 5C illustrates the chief experimental prediction ofthe RO hypothesis:parts ofeithergene that cannotbe tRNAaminoacylation relatedsense/antisense—bothanticodon-bindingdomains, Catalysis of amino acid activation by aaRS Urzymes left theinsertionsandClassIIMotif3—appearinsomesense a key question unanswered: do these peptide catalysts functionally superfluous. Removing them leaves precisely also accelerateaminoacylation oftRNA?Acentralimpli- the invariant cores we had identified for both enzymes, cation of the RO hypothesis is that sense/antisense and these alignquite closely, sense/antisense.Thekey ex- encoded fragmentsshould beexhibitboth activities.Pre- perimentalquestion is:how activeare ClassI and IIaaRS cedent and structural arguments led us to expect recog- Urzymes? Too little catalytic activity to produce activated nition of tRNA by aaRS Urzymes, even without the aminoacidsata sufficientratetosupport uncatalyzedas- anticodon-binding domain, which is often considered a sembly into peptides would effectively falsify the RO hy- late addition [23]. A considerable literature describes the pothesis.Urzymes are catalytically very much more active acylation of isolated tRNA modules containing the ac- thannecessary. ceptor stem [19,20,55,56]. Comparable experiments have Table1ProceduresusedinUrzymology Process Remarks Superimposehomologs POSAserver:http://fatcat.burnham.org/POSA/ IDsharedinvariantsubset Usually10-30%ofmonomericforms Re-designexposedsurface Proteindesign:Rosettahttp://www.rosettacommons.org/ Clone&Express MaltoseBindingProtein(MBP)fusions;TEVcleavage Characterize Rateaccelerations,substratespecificities Authenticate Singleturnoveractive-sitetitration,Steady-statekinetics,geneticmutationandmanipulation Evolutionaryrecapitulation Multi-scalemodularde-andreconstruction Carteretal.BiologyDirect2014,9:11 Page6of23 http://www.biologydirect.com/content/9/1/11 be sure that they do not result from a tiny amount–1 in 100,000–of contaminating wild-type enzyme? Seven dif- ferent tests (Table 2) argue from multiple points of view that these activities are authentic. The last four func- tional assays uniquely distinguish Urzymes from native aaRS,andarediscussedindetail: 1. Empty vector controls showessentially noactivity. WeexpressallUrzymesasmaltose-bindingprotein (MBP)fusions toimprovesolubility. Nounfused MBPexpressed andpurifiedinthe same manner on anamyloseresin exceededbackgroundwhenassayed at12mg/mlwith32PPiexchangemixesfortrypto- phan,histidine,and leucine. 2. ActiveTrpRSUrzymecan berenaturedfrom inclusionbodies[40].Itisunlikely thatnative full Figure4PartialorderstructurealignmentoffourClassIand length enzyme wouldcontaminate inclusion body ClassIIUrzymes.Despitethestrongsimilaritiesbetweenthefour preparations. Urzymesfromeachclass,POSAidentifiesappropriatesub-classification. 3. Cleavage ofMBP fusionproteins releasescryptic activity.MBPfusions inhibitbothTrpRSand HisRS until now not been performed with modular fragments Urzymes~50-fold [38,39].Crypticactivity released of aaRS, owing to the greater difficulty of constructing byTobaccoEtchVirusprotease cleavageofpurified and purifying proteins, compared to RNA. Further, sim- fusion proteinsimplicates boththepurifiedfragment ultaneous appearance of a fully-developed genetic code, and protease cleavage(seealso 5below). depending heavily on binding the anticodon loop of 4. Active-sitetitrationsshowsignificant pre-steady state most tRNAs [56] is difficult to envision. Accordingly, burstsinsingle turnoverassays,amounting to10- Giegé, Schimmel, and others proposed that an earlier, 90%ofthe totalnumberofmolecules.Active-siteti- “operational RNA code” in the tRNA acceptor stem was trations measuresingleturnovertime courses.If aforerunnerofthe present daycode[23]. productreleaseisrate-limiting,thenturnover will The crystal structure of human cytosolic TrpRS com- beslowerthanthe firstroundofcatalysis,andex- plexed with the acceptor stem of tRNATrp [57] affords a trapolationofthesteady-staterate totheorigin can model for the corresponding interaction with theTrpRS beused toestimate the“burst”orthe amplitudeof Urzyme (from H. Hu’s MD simulations; [39]) Figure 7A. thefirst-orderportion ofthereaction.Burst size Buried surface area calculation with a probe radius of thereforeestimatestheproportionofactive mole- 1.5 Å shows that 522 Å2 of the Urzyme is potentially in cules.Contaminatingactivity from atiny amount of contact with tRNATrp. Further, complementary tRNA- wild typefull-length enzyme 105-fold moreactive binding surface in the α-helix of the second crossover thanthe Urzymeswouldexhibitaninsignificant connection of the Urzyme includes an invariant, E152, burst,andthe entiretime course wouldrepresentits that specifically recognizes the discriminator base A73. steady-state rate.Activefractionsalsoprovide more Similar considerations apply to the HisRS Urzyme(cid:129)tR- accuratek values. cat NAHiscomplex(Figure 7B). BothUrzymesshowsubstantialbursts,whichrange Data shown in Figure 7A, 7B for 32P-3’ adenosine- between~10and90%—much biggerthanthose labeled tRNATrp and tRNAHis [58] demonstrate that expectedfromarare,veryactive contaminant.Full TrpRS [39] and HisRS [38] Urzymes catalyze tRNA ami- length aaRSbindtightlytothe aminoacyl-adenylate noacylation. TrpRS and HisRS Urzymes therefore retain toprotectthe cellfrom ahighlyreactiveadenylating afullfunctionalrepertoire.Inparticular,catalyticactivities reagent andto preservethespecificity achievedby required for protein synthesis are as finely tuned as the theactivationstep [17].Itisremarkablethat the contemporary enzymes, between activation and acylation Urzymesalsosequester theintermediate. Pre-steady and between Classes, [37]. They therefore represent con- stateburstsarethus athirdkey function offull- vincingmodelsforancestralClassIandClassIIaaRS. length aaRS(Figure1)retained byUrzymesfrom bothClasses. AreUrzymeactivitiesauthentic? 5. Active-sitemutantsandmodular variantshave Urzyme catalysis of amino acid activation is ~105-fold altered activities.Molecular biologistsrecognizethat weakerthanthatofcontemporaryenzymes. Howcanwe manipulating thegeneofa suspectedsourceof Carteretal.BiologyDirect2014,9:11 Page7of23 http://www.biologydirect.com/content/9/1/11 Figure5TheRodin-OhnohypothesisholdsthatancestralformsofClassIandIIaminoacyl-tRNAsynthetases(aaRS)hadfully complementarynucleicacidcodingsequencesandthatcontemporaryaaRSdescendedfromoppositestrandsofasinglegene.The schematicinAillustrateshowthishypothesisleadsdirectlytotheconceptofaaRSUrzymes.Antiparallelalignmentofaminoacidsequencesfor theClass-definingmotifs(HIGHandKMSKSfromClassI;Motifs1and2fromClassII)revealsthatneitheranticodon-bindingdomain,noralong insertionineachClasscanphysicallybeapartofsuchanancestralgene.Asaresult,theonlyportionsofthetwoClassesconsistentwiththe hypothesis(B)areabout120-130residueslong.Thesefragmentscoincidewithinvarianttertiarystructuralcoressharedbyallsuperfamilymembers. Moreover,thesesegmentstogethercomposeaminimalactivesite,containingbindingsitesforallthreesubstrates(C).AminoacidandATP determinantsarereflectedacrossthegenesequence,whiletRNAbindingdeterminantsarerelatedbytwo-foldrotation[40]. activity canimplicatethatgene product inthe significantamount(Figure8;[38]).WePCR observedactivity.Wethereforetestedactive-site amplifiedthe122residue fragmentcontainingonly mutationsand modular variantsintheTrpRSand Motifs 1and 2,addingeithera six-residueN- HisRSUrzymes.Allsuchexperimentssignificantly terminalextension(red),orMotif 3(yellow),or alteredactivity. Oneresult—mutationD146Ain both,giving usabalanced assayforthe effectsof TrpRSUrzymeactuallyincreased, rather thanredu- bothfactors.Theintrinsiccatalyticenhancementof cing activity asitdoesintheWT enzyme—was Motif 3totransition-statestabilizationbyHisRS counter-intuitive.However,the catalyticfunction of Urzyme,-0.85 kcal/Mole,isessentiallyidenticalto D146 infull-lengthTrpRSlikelyrequiresallosteric thatofthemuch shorterandlessobviousN- couplingofmissingCP1 and ABDmodules,and its terminalextension;their synergisticeffectisnearly presenceintheUrzyme likelystabilizes theground twicethat.Figure 8emphasizes thatthetwomod- state, ratherthanthe transitionstate [59]. ulesstabilize theATP bindingsitefrom opposite Ofparticularinterest arethermodynamiccyclesthat faces ofthe molecule. involve anaaRSUrzymeand complementing FullTrpRSspecificityand tRNATrpaminoacylation segments.Veryshort(6-20aa) modules accelerate activity both require essentially complete HisRSUrzymeaminoacid activationbya smallbut interdomainsynergy(Figure9;[36]). TheTrpRS Carteretal.BiologyDirect2014,9:11 Page8of23 http://www.biologydirect.com/content/9/1/11 Figure6QuantitativeframeworkinwhichtoassessthecatalyticsignificanceofUrzymesandvariousotherputativestagesofaaRS evolution.A.Rateaccelerationsestimatedfromexperimentaldataforsinglesubstrate(red)andbi-substrate(Black,Bold)reactionsadaptedfrom [75]toincludeuncatalyzedandcatalyzedratesofbi-substratereactionsoftheribosome[74],aminoacidactivation[39]andkinases[106].Second orderrateconstants(blackbars)wereconvertedintocomparableunitsbymultiplyingby0.002M,whichistheATPconcentrationusedtoassay thecatalystsshowninB.B.Experimentalrateaccelerationsestimatedfromsteadystatekineticsaskcat/K foraseriesofcatalystsderivedfrom M ClassIandClassIIaminoacyl-tRNAsynthetases([38,39]anddataofV.Weinreb,L.Li,M.Collier,andK.Gonzalez-Riverapresentedhereinasubsequent section).VerticalscalesinAandBarethesame,andtheoriginofthehistograminBhasbeensetequaltotheuncatalyzedrateofaminoacid activationin(AAact)inA.RedbarsdenoteClassITryptophanyl-andLeucyl-tRNAsynthetaseconstructs,bluebarsdenoteClassIIHistidyl-tRNA synthetaseconstructs,andgreendenotesaribozymalcatalyst[97],includedforcomparison.ResearchpresentedinA,Bwasoriginallypublishedin [37].©TheAmericanSocietyforBiochemistryandMolecularBiology.C.ClassILeuRSandClassIIHisRSUrzymeaminoacidspecificities.Aminoacids withmorenegativeΔGk /K valuesindicatehigheractivities.By~1kcal/mole(lightbands)or~five-fold,eachUrzymepreferssubstratesfromthe cat M classtowhichitbelongs(darkbands).Nonetheless,bothactivatearangeofnon-cognateaminoacids,andarepromiscuous. Urzyme favors tryptophanactivationby~10-fold peptide segmentsthatcanalignantisense tothe overcompeting tyrosine and is~400timesless HisRSUrzyme, isactuallybetter atthe twotasks— specificthanfull-lengthTrpRS.CP1 andthe amino acidrecognition and tRNA aminoacylation— anticodon-bindingdomain (ABD) mustaccount for requiredofaaRS,and hencelie closer totheactual theincreasedspecificity ofnativeTrpRS.Wead- pathofaaRSevolutionthaneitherintermediate, dressed thisquestionbycomparing specificitiesof potentiallymoreadvancedconstruct.Evolutionary intermediate constructsinwhich theCP1and ABD developmentofcontemporaryenzymes mustbe modules wereadded backindividually tothe moresubtlethansimplyaccumulating onemodule Urzyme [36]toformacompletefactorialexperi- atatime. ment inthosetwovariables(Figure9A). 6. Steady stateK valuesdiffer fromthose oftheWT M Quitesurprisingly(Figure9B),although addingback enzymes.Enzymologistsrecognize thatthesteady- eitherCP1 ortheABD doesenhance tryptophan stateK valueisan independent signature. M binding,thiseffectisnon-specific.Additionofeither Contaminatingwild-type enzymes,irrespectiveof domain alsoreduced K fortyrosine,suchthat the concentration, wouldsaturateatthesame amino M logofthespecificity ratio(k /K ) /(k /K ) , acid concentrations.AlteredK values arethus cat M Trp cat M Tyr M wasactually~0.0 (Figure9B).The ~400-foldin- strong evidenceagainstcontaminating wild-type en- creaseinspecificityobservedforfull-lengthTrpRS, zyme activity. @ATP-and amino acid-dependent relative totheUrzymedependsentirely oncoopera- Michaelis-MentendatafortheTrpRSand HisRS tiveinteractions(alsocalled epistasis[60-62])be- Urzymesshow thatATPbindingaffinity iseitherthe tween thetwodomains[36]. same ortighter toUrzymesthan tocontemporary tRNATrpaminoacylationrequirescomparable aaRS[38,39].Thek values arenearlycomparable cat interdomainsynergy(Figure9B).These experiments tothose ofthenative enzymes. Urzymeaminoacid withTrpRSUrzymesupporttheunexpected K s,however,are quitedifferent fromthose offull- M conclusionthattheUrzymeitself,consistingonlyof length,nativeenzymes.TheTrpRSUrzymeK for M Carteretal.BiologyDirect2014,9:11 Page9of23 http://www.biologydirect.com/content/9/1/11 Figure7tRNATrpAcylationbyTrpRSUrzyme.A.ModeloftheputativeinteractionbetweenTrpRSUrzymeandtRNATrp,derivedfromthe crystalstructureofthecomplexbetweenhumanTrpRSandtRNATrp[57].AutoradiogramsdocumentingtheacylationoftRNATrpbyWildType TrpRS,TrpRSUrzyme,andtwointermediatemodularconstructs,containingeitherCP1ortheanticodon-bindingdomain.B.Modelfor interactionoftRNAHiswithHisRS4UrzymeandautoradiogramsshowingacylationbyHisRS1,2,and4asinA.SpheresshowTrp-5’AMP,and His-5’sulfoamyladenylate.Thesedatawerepublishedoriginallyin[37]©TheAmericanSocietyforBiochemistryandMolecularBiology. Table2CriteriafortheauthenticityofUrzymecatalyticactivities Criterion Implementation Remarks Emptyvectorcontrols Purify,assayMBP DeRigueur,butunconvincing Renaturationfrominclusionbodies TaggedUrzymespurifiedfrom WTEnzymesdonotsegregatewithinclusionbodies pellet MBPfusionsreleasecrypticactivityonTEV Assayfusionproteins±TEV Inhibitioninfusionproteinsiswidespread,notuniversal. cleavage. cleavage Active-sitetitrationsUrzyme Singleturnovertime-courses Akeycriterion,thisisalsoessentialforcomparingk /K . cat M preparationshavesignificantbursts. Mutations,modularalterationsinduce Determineeffectofactive-site Active-sitemutationsgenerallyaffectUrzymeactivitiesdifferently predictablechangesinactivity. mutations,geneticmanipulations andcanactuallyenhanceactivitybecausemechanismsaredifferent. Urzymes,WTenzymeshavedifferent Measure:k ,K ,k /K ContaminationbyWTenzymewouldsaturateatWTK . cat M cat M M Steady-stateK values. M Aminoacidspecificityisdifferent Compare:(k /K ) /(k /K ) Urzymesaregenerallylowspecificity,highk catalysts. cat MW cat MY cat fromfull-length Carteretal.BiologyDirect2014,9:11 Page10of23 http://www.biologydirect.com/content/9/1/11 tryptophanis ~1mM,500 timeshigherthan thatof wild typeTrpRS[39].That forHisRS-3,containing Motifs 1,2,and3,butlackingthesix-aminoacid N- terminalextensiontoMotif 1,is120μM, compared to30 μMforwild-typeHisRSand45 μMforthe N catalyticdomain [38]. cat 7. HisRSandTrpRSUrzymeshave reduced,but Class- dependent specificities.Weaker aminoacid affinities imply thattheUrzymeslikelyhavereduced specificity fortheircognateaminoacids. TheTrpRS Urzymeretainsa10-fold preferencefortryptophan vstyrosine[36,39].Second-order rate constant free energies,-RTln(k /K ),foraminoacid substrates cat M activatedbyClassILeuRSand ClassIIHisRS Urzymes(Figure6C)showthatbothUrzymesare promiscuous.However,they bothpreferentiallyacti- vate aminoacids similar totheoriginalsubstrate (i.e., LeuandHis). Byafactor of~5-fold,they both Figure8Quantitativeanalysisofthecatalyticcontributionsof preferaminoacid substratesfromtheClasstowhich Motif3andashortN-terminalextensionoftheMotif1helixto theaccelerationofhistidineactivationbyHisRSUrzyme. they belong.Thismodest,class-dependent substrate GraphicsincludetheHistidyl-5‘adenylateasspheres.Detailsofthe specificity alsorulesoutadventitiousactivitiesunre- constructionsaredescribedin[38]. lated toaaRS-derivedconstructs.Aswithactive-site titrationand steady-statekinetic parameters,con- taminatingwild-typeaaRSwouldhavea specificity of~4000-fold. Figure9FactorialanalysisofTrpRSinter-domaineffects.A.ThecatalyticactivityoftheUrzymefacilitatesafull-factorialanalysisofthebenefits ofaddingeithertheCP1oranticodon-bindingdomain,jointlywiththeirsynergisticeffectinthefull-lengthenzyme.B.Freeenergyhistogramsforthe factorialdesigninA,showingnearlyidenticalpatternsinwhichtheCP1andABDactuallydiminishbothspecificityfortryptophanvstyrosineand acylationoftRNATrp.TheentiredifferencebetweenUrzymeandfull-lengthenzymeisachievedonlyviathesynergisticparticipationofbothdomains missingintheUrzyme.