GBE The Molecular Chaperone DnaK Is a Source of Mutational Robustness Jose´ Aguilar-Rodr´ıguez1,2,y, Beatriz Sabater-Mun˜oz3,4,y, Roser Montagud-Mart´ınez3, Vı´ctor Berlanga3, David Alvarez-Ponce5, Andreas Wagner1,2,6,*, and Mario A. Fares3,4,* 1DepartmentofEvolutionaryBiologyandEnvironmentalStudies,UniversityofZurich,Zurich,Switzerland 2SwissInstituteofBioinformatics,Lausanne,Switzerland 3DepartmentofAbioticStress,InstitutodeBiolog´ıaMolecularyCelulardePlantas(CSIC-UPV),Valencia,Spain D o 4DepartmentofGenetics,SmurfitInstituteofGenetics,UniversityofDublinTrinityCollegeDublin,Dublin,Ireland wn lo 5DepartmentofBiology,UniversityofNevada,Reno,Reno,Nevada,USA a d e 6SantaFeInstitute,SantaFe,NewMexico,USA d fro m *Correspondingauthors:E-mail:[email protected];[email protected],[email protected]. h yTheseauthorscontributedequallytothiswork. ttps ://a Accepted:July 22, 2016 c a d Datadeposition:NewgenomicsequenceshavebeendepositedintheSequenceReadArchiveundertheaccessionnumberSRP074414. e m ic .o u p .c Abstract o m /g Molecularchaperones,alsoknownasheat-shockproteins,refoldmisfoldedproteinsandhelpotherproteinsreachtheirnative be /a conformation. Thankstotheseabilities,some chaperones, suchastheHsp90proteinor thechaperoninGroEL,canbufferthe rtic deleteriousphenotypiceffectsofmutationsthatalterproteinstructureandfunction.Hsp70chaperonesuseachaperoningmech- le -a anismdifferentfromthatofHsp90andGroEL,anditisnotknownwhethertheycanalsobuffermutations.Here,weshowthatthey b s can.Tothisend,weperformedamutationaccumulationexperimentinEscherichiacoli,followedbywhole-genomeresequencing. tra c OverexpressionoftheHsp70chaperoneDnaKhelpscellscopewithmutationalloadandcompletelyavoidtheextinctionsweobserve t/8 inlineagesevolvingwithoutchaperoneoverproduction.Additionally,oursequencedatashowthatDnaKoverexpressionincreases /9/2 mutationalrobustness,thetoleranceofitsclientstononsynonymousnucleotidesubstitutions.Wealsoshowthatthiselevated 97 9 mutationalbufferingtranslatesintodifferencesinevolutionaryratesonintermediateandlongevolutionarytimescales.Specifically, /2 2 westudiedtheevolutionaryratesofDnaKclientsusingthegenomesofE.coli,Salmonellaenterica,and83othergamma-proteo- 36 2 bacteria.WefindthatclientsthatinteractstronglywithDnaKevolvefasterthanweaklyinteractingclients.Ourresultsimplythatall 2 5 threemajorchaperoneclassescanbuffermutationsandaffectproteinevolution.Theyillustratehowanindividualproteinlikea by chaperonecanhaveadisproportionateeffectontheevolutionofaproteome. gu e s Key words: molecular chaperones, mutational robustness, experimental evolution, protein evolution, Escherichia coli, DnaK. t o n 3 1 M Introduction mutationalorgeneticrobustness.Molecularchaperones(Ellis arch Robustnessisoneofthefundamentalpropertiesoflivingsys- 1987) are one of the best-known sources of both types of 20 1 tems(deVisseretal.2003;Wagner2005;MaselandSiegal robustness (Fares 2015). Chaperones, also called heat-shock 9 2009; Fares 2015). This property describes the ability of a proteins,assistproteinsinreachingtheirnativeconformations, biologicalsystemtopreserveitsphenotypeinaparticularen- prevent protein aggregation, and refold misfolded proteins vironment despite perturbations that it encounters. The ro- (Young et al. 2004; Hartl and Hayer-Hartl 2009; Hartl et al. bustness of a system against perturbations that are 2011). Thanks to these roles, chaperones can restore the environmental (e.g., a change in temperature) is referred to nativeconformationofproteinsdestabilizedbyenvironmental asenvironmentalrobustness,whereasrobustnessagainstper- perturbations,thusprovidingenvironmentalrobustnesstoor- turbationscausedbygeneticmutationsreceivesthenameof ganisms coping with stressful conditions. Because some (cid:2)TheAuthor2016.PublishedbyOxfordUniversityPressonbehalfoftheSocietyforMolecularBiologyandEvolution. ThisisanOpenAccessarticledistributedunderthetermsoftheCreativeCommonsAttributionNon-CommercialLicense(http://creativecommons.org/licenses/by-nc/4.0/),whichpermits non-commercialre-use,distribution,andreproductioninanymedium,providedtheoriginalworkisproperlycited.Forcommercialre-use,[email protected] GenomeBiol.Evol.8(9):2979–2991. doi:10.1093/gbe/evw176 AdvanceAccesspublicationAugust6,2016 2979 GBE Aguilar-Rodr´ıguezetal. chaperones can buffer the deleterious effects of mutations 2010; Warnecke and Hurst 2010; Williams and Fares 2010; thataffectproteinfolding,theyarealsoasourceofmutational PechmannandFrydman2014). robustness. In the context of protein evolution, chaperones WhilenoHsp70chaperonehasbeendirectlyimplicatedin areabletoincreaseaprotein’smutationalrobustnessbecause mutationalbuffering,pertinentcircumstantialevidenceexists. they alter the mapping from protein genotypes into protein Forexample,DnaK—themajorbacterialHsp70chaperone— phenotypes, that is, into the structures that proteins form is overexpressed together with GroEL in S. enterica lineages (Rutherford 2003). Specifically, they increase the number of withreducedfitnesscausedbytheaccumulationofdeleteri- amino acid sequences that fold into the same structure and ous mutations (Maisnier-Patin et al. 2005). In addition, D. thatcanperformthefunctionassociatedwiththisstructure. melanogaster populations showing inbreeding depression, There are three main chaperone systems, which are the where increased homozygosity exposes recessive deleterious D o Hsp90system,theHsp70system,andtheHsp60system(or mutations, significantly up-regulate the expression of Hsp70 w n chaperonins), of which the bacterial GroEL is a prominent comparedwithoutbredpopulations(Pedersenetal.2005). lo a d member (Hartl et al. 2011). Overwhelming evidence shows ThechaperonesfromtheHsp70systemareveryconserved e d that Hsp90 and GroEL can buffer mutations (Bogumil and frombacteriatohumans(PowersandBalch2013).Theyplaya fro Dagan 2012), but whether the same holds for any major centralroleinproteomeintegrity,andareinvolvedbothinco- m h chaperone from the Hsp70 system is to our knowledge un- andpost-translational folding (Hartl et al. 2011). In bacteria, ttp s known. A recent study has shown that RNA chaperones— the Hsp70 chaperone DnaK (together with GroEL and the ://a they help RNA molecules to fold properly, and comprise a Trigger Factor) is one of the main molecular chaperones, ca d classofchaperonesdifferentfromthesethreesystems—can whereitisthecentralhubinthechaperonenetworkofthe em also buffer deleterious mutations in Escherichia coli (Rudan cytosol(BukauandWalker1989;Callonietal.2012).Itinter- ic .o etal.2015). actswithatleast~700mostlycytosolicproteins(Callonietal. up .c Pioneering work carried out by Rutherford and Lindquist 2012).TheDnaKinteractomewascharacterizedbytheisola- o m (1998) showed that inhibition of the chaperone Hsp90 can tionofDnaKinteractorsusingimmobilizedmetalaffinitychro- /g b unveil cryptic genetic variation—genotypic variation without matography, followed by liquid chromatography mass e /a phenotypicvariation—inthefruitflyDrosophilamelanogaster. spectrometry.TheseregularclientsofDnaKareenrichedfor rtic Subsequently, similar observations have been made in the proteinswithlowintrinsicsolubility,proteinsthattendtobe le-a plant Arabidopsis thaliana (Queitsch et al. 2002), the yeast membersofhetero-oligomericcomplexesand/orproteinsthat bs Saccharomyces cerevisae (Cowen and Lindquist 2005) and showahighdensityofhydrophobicpatchesflankedbypos- trac thefishAstyanaxmexicanus(Rohneretal.2013).Furthersup- itive residues (Calloni et al. 2012). DnaK is highly expressed t/8 /9 portwasprovidedbyBurgaetal.(2011),whofoundthathigh constitutivelyandessentialat42(cid:2)C(BukauandWalker1989; /2 9 7 induction of Hsp90 during development of the nematode Callonietal.2012).DuringitsATP-dependentreactioncycle, 9 /2 Caenorhabditis elegans reduced the penetrance of certain DnaK interacts with the Hsp40 co-chaperone DnaJ, which 2 3 mutations. Additionally, Lachowiec et al. (2013) found that determines the client binding specificity of DnaK (Straus 62 2 paralogs of duplicated kinase-coding genes that encode a et al. 1990; Hoffmann et al. 1992), and the nucleotide ex- 5 b substrateofHsp90(i.e.,aHsp90“client”)inSaccharomyces changefactorGrpE(Hartletal.2011).Thechaperonesystem y g u cerevisiaeoftenevolvefasterthanparalogsencodingnoncli- formedbythesethreeproteinscanbothfoldnascentproteins e s ents. In general, the rate at which nonconservative substitu- and refold denatured proteins. It does so by binding to ex- t o n tions—those that alter physicochemical properties of amino posed hydrophobic patches in unfolded or partially folded 3 1 acids—accumulate is especially accelerated in Hsp90 clients protein substrates, thus preventing detrimental interactions M a (PechmannandFrydman2014). withotherpolypeptidesinthecrowdedcellularmilieu.Bysuc- rc h Multiplestudiesalsodemonstratemutationalbufferingme- cessivelybindingandreleasingaproteinsubstrateinacyclic 20 1 diatedbythebacterialchaperoninGroEL.Forexample,Fares process that consumes ATP, the chaperone system DnaK– 9 etal.(2002)showedthatoverexpressingGroELconsiderably DnaJ–GrpEallowsthesubstratetograduallyexploreitscom- improvedthefitnessofE.colistrainswithahighloadofdel- plex folding energy landscape (Hartl and Hayer-Hartl 2009; eteriousmutations,apatternthatwasalsoobservedlaterin Hartletal.2011).Forsomeproteins(~20%ofthetotalpro- Salmonella enterica (Maisnier-Patin et al. 2005). Moreover, teome), several of these bind-release cycles are enough to GroEL overexpressionin E. coli increases the ability of GroEL achieve the native conformation. However, other proteins client proteins to tolerate mutations (Tokuriki and Tawfik (~10% of the total proteome) still require the downstream 2009; Bershtein et al. 2013; Wyganowski et al. 2013; chaperone system GroEL/ES (Hartl et al. 2011). The impor- Sabater-Mun˜ozetal.2015),aswellastheirabilitytoundergo tanceoftheDnaK–DnaJ–GrpEsysteminthebacterialchaper- adaptive evolution (Tokuriki and Tawfik 2009; Wyganowski one network is obvious from its strong conservation across et al. 2013). Buffering of destabilizing mutations accelerates bacteria, except for two species from the order Aquificales the evolutionary rates of GroEL clients (Bogumil and Dagan that have lost the entire system, and individual losses of 2980 GenomeBiol.Evol.8(9):2979–2991. doi:10.1093/gbe/evw176 AdvanceAccesspublicationAugust6,2016 GBE TheChaperoneDnaKIsaSourceofMutationalRobustness dnaJ and grpE in obligate endosymbionts that have experi- dnaK–dnaJ–grpE is deleted (fig. 1B). Cells of all 16 control encedconsiderablegenomereductions(Warnecke2012). lines therefore cannot overexpress DnaK, even though their Most mutations affecting proteins are neutral or deleteri- growth medium contains L-arabinose (DnaK(cid:3) lines). At each ous(Eyre-WalkerandKeightley2007),andfunctionallyimpor- temperature, half of the lines evolved in the presence of L- tantmutationsoftendestabilizeproteins(Tokurikietal.2008; arabinose,whereastheotherhalfevolvedinmediumdevoid Wyganowskietal.2013).IfDnaKbuffersdestabilizingmuta- of this expression inducer. In total, we evolved 86 bacterial tions,thenthedeleteriouseffectsofmutationsinhighlyinter- populations:68DnaK+linesand16DnaK(cid:3)lines.Westopped acting(strong)clientsshouldbelowerthaninsporadic(weak) the evolution experiment after 85 single-cell bottlenecks, or clients, where they should be lower than in nonclients. In ~1,870generations(assumingconservatively~22generations otherwords,themorestronglyaprotein’sintegritydepends perdailygrowthcycle). D o onDnaK,thehighershouldbeitstolerancetomutations,and w n the lower the signature of purifying selection that purges EvolvingLineagesTendtoGoExtinctintheAbsenceof lo a those mutations. With this reasoning in mind, we here use DnaKOverexpression de d laboratoryexperimentstoevaluatetheeffectofDnaKbuffer- OneofthefirstindicationsthatDnaKoverexpressioncouldbe fro ingontheevolutionofitsclientproteomeonshortevolution- m bufferingdeleteriousmutationsaccumulatedduringtheevo- h ary time scales. We complement our experimental lution experiment is the observed pattern of extinctions ttp s observations with sequence analyses to study the effect of (fig.1).Someevolvinglineswentextinct,presumablydueto ://a DnaKonintermediateandlongevolutionarytimescales. c highlevelsofmutationalload,andremarkably,allextinctions a d e occurred in lines that were not overexpressing DnaK. They m Results wereeitherDnaK(cid:3) linesorDnaK+evolvedintheabsenceof ic.o theinducer.Morespecifically,75%oftheDnaK(cid:3)lines(12of up ExperimentalEvolutionofE.coliunderDnaK 16)wentextinctbeforetheendoftheevolutionexperiment .co m Overexpression (fig.1B).AmongtheDnaK+lines,62.5%ofthelines(5of8) /g b TostudytheeffectofDnaKoverexpressiononproteinevolu- evolvingintheabsenceoftheinducerwentextinct,whereas e/a tionexperimentally,weperformedmutationaccumulationex- noneofthe60linesevolvinginthepresenceoftheinducer rtic periments similar to those we reported recently for the experiencedanyextinction(fig.1A).Thisobservationstrongly le-a Mchuanp˜oerzoentinal.G2r0o1E5L,).bBuriteflfyo,rwDenianKitiaotevdere6x8ppreasrsaiollenla(nSadbiantdeer-- scuregagseesdtsththeartobouvestrnexepssreossfintghethceellschtaopethroenaeccDunmauKlahtiaosnino-f bstrac deleterious mutations, helping them cope with mutational t/8 pendentclonallinesofevolution,allofwhichderivedfromthe /9 same hypermutable clone (E. coli K12 MG1655 (cid:2)mutS) load. /29 7 (Sabater-Mun˜oz et al. 2015) (fig. 1A). Cells of the 68 lines 9 /2 all harbored the plasmid pKJE7, which contains the operon OverexpressingDnaKIncreasestheRobustnessto 23 dnaK–dnaJ–grpEunder the control ofthe L-arabinose-induc- NonsynonymousMutationsofDnaKClients 622 5 ible araB promoter PBAD (Nishihara et al. 1998). We refer to In order to study the effect of DnaK buffering on genome by this strain as DnaK+. We evolved 60 of the 68 DnaK+lines evolution, we sequenced the genomes of some lines at the g u through repeated single-cell bottlenecks in the presence of endoftheevolutionexperiment,after85passages,andcom- es theinducer,toensureoverexpressionofDnaK,aswellasof pared them to the ancestral genome, which we had se- t on the cochaperone DnaJ and the nucleotide exchange factor quenced in a previous study (Sabater-Mun˜oz et al. 2015). 31 M GrpE. All evolving lineages were passaged after 24hours of Amongtheclonallinesevolvedinthepresenceoftheinducer a incubation.Becauseofthebottleneckstowhichweexposed L-arabinose, we randomly selected for sequencing 3 rch thepopulations,geneticdriftwasstrongandtheefficiencyof DnaK+lines evolved at 37(cid:2)C, and another 3 at 42(cid:2)C. We 20 1 selectionwasweakduringthe experiment, suchthatnonle- also sequenced the only two surviving control DnaK(cid:3) lines 9 thal mutations are free to accumulate (Barrick and Lenski evolved with L-arabinose in the medium, each at a different 2013). We evolved 30 of the 60 clonal lines at 37(cid:2)C, and temperature.Althoughallsequencedlinesevolvedinthepres- the other 30 at 42(cid:2)C. The higher temperature serves to in- enceoftheinducer,onlyDnaK+linesareabletooverexpress creasethedeleteriouseffectofdestabilizingmutationsinthe thechaperone. bacterialproteome(BukauandWalker1989).Finally,there- Inordertoevaluateifasignificantdifferenceexistedinthe maining8DnaK+lineswereevolvedintheabsenceofinducer, mutation rate (or generation time) between the sequenced andthereforewithoutDnaKoverexpression(4linesat37(cid:2)C DnaK+and DnaK(cid:3) lines, we compared the number of accu- andtheother4at42(cid:2)C). mulatedsynonymousmutationsbetweenthem (supplemen- Ateachofthetwotemperatures,weadditionallyevolved8 tary table S1, Supplementary Material online). We observed control clonal lines founded from the same parental strain, an average number of 78 synonymous mutations per but carrying a pKJE7-derived plasmid where the operon DnaK+lineand66synonymoussubstitutionsperDnaK(cid:3)line, GenomeBiol.Evol.8(9):2979–2991. doi:10.1093/gbe/evw176 AdvanceAccesspublicationAugust6,2016 2981 GBE Aguilar-Rodr´ıguezetal. D o w n lo a d e d fro m h ttp s ://a c a d e m ic .o u p .c o m /g b e /a rtic le -a b s tra c t/8 /9 /2 9 7 9 /2 2 3 FIG.1.—Mutationaccumulationexperiment.Evolutionaryhistoryofthepopulationsevolvedinthisstudyfromthefirstdailytransferorsingle-cell 6 2 bottleneck(T)untiltheendoftheevolutionexperiment(T ).WeconstructedtwostrainsderivedfromanancestralEscherichiacoliK-12MG1655strain 2 0 85 5 lackingthemismatchrepairgenemutS.TheDnaK+strainharboursthe~15-copyplasmidpKJE7thatcontainstheDnaK/DnaJ/GrpEchaperonesystemunder b y thecontrolofthepromoterP induciblebyL-arabinose(Nishiharaetal.1998).TheDnaK(cid:3)straincontainsacontrolpKJE7-derivedplasmidwherethe g BAD u e operondnaK–dnaJ–grpEhasbeendeleted.Weevolvedinparallelmultipleindependentpopulationsofbothstrainsthroughsingle-cellbottlenecksunderthe s effectofstronggeneticdriftattwodifferenttemperatures(37(cid:2)Cand42(cid:2)C).Ateachtemperatureweevolvedsomepopulationsinthepresenceof t o n L-arabinose(L-ara+),andsomeintheabsenceofthisexpressioninducer(L-ara–).(A)Duringtheevolutionof 68DnaK+populations,fiveoutofeightlines 3 1 evolvingintheabsenceofinducerwentextinct(indicatedbyacross).Noneofthe60linesevolvingunderDnaKoverexpressionexperiencedanyextinction. M a (B)Ofthe16independentDnaK(cid:3)populations,12populationswentextinct.Wefinishedtheevolutionexperimentafter85single-cellbottlenecks(T85),or rch ~1,870generations. 2 0 1 9 which is not significantly different (binomial test, P=0.359; TheoverexpressionofDnaKmaybeenergeticallycostly,just supplementarytableS1,SupplementaryMaterialonline).We as is the case for the chaperonin GroEL (Fares et al. 2002; alsodidnotobserveanysignificantdifferenceinthenumber Sabater-Mun˜oz et al. 2015). In principle, this cost could of accumulated nonsynonymous mutations (binomial test, favortheaccumulationofmutationsthatleadtoadecrease P=0.646; table 1), the number of indels (binomial test, in the expression of DnaK during the evolution experiment, P=0.332; supplementary table S1, Supplementary Material especially if the energetic cost of overproducing the chaper- online), or the ratio of transitions to transversions ((cid:2)2 test, oneisgreaterthanthebenefitsderivedfrommutationalbuff- P=0.273; supplementary table S1, Supplementary Material ering (Sabater-Mun˜oz et al. 2015). However, we observed online). that for the sequenced lines the overexpression of DnaK We also verified that DnaK was still overexpressed at was maintained through the mutation accumulation experi- the end of the experiment in the 8 sequenced lines. mentatboth37(cid:2)Cand42(cid:2)C(fig.2;supplementaryfig.S1, 2982 GenomeBiol.Evol.8(9):2979–2991. doi:10.1093/gbe/evw176 AdvanceAccesspublicationAugust6,2016 GBE TheChaperoneDnaKIsaSourceofMutationalRobustness Table1 comparedwiththeDnaK(cid:3)lineevolvedat42(cid:2)C,where~13% Distributionof NonsynonymousNucleotideSubstitutionsamong DnaK of all mutationsaffected DnaK clients (15 out of 114 muta- Client andNonclient ProteinsAfter ~1,870Generationsof Evolution tions),theDnaK+linesshowedsignificantlymoremutationsin inMutation Accumulation ExperimentsConducted at 37(cid:2)Cand42(cid:2)C clients(71outof384totalmutations,~18%;binomialtest: Temperature Linea Numberofmutations P=0.003;fig.3A).Theseresultssuggestthatoverexpressing DnaK does indeed increase the robustness of its clients to Clients Nonclients amino acid replacements. Temperature itself had no signifi- Strongb Weakc Total cant effect on the fraction of all mutations affecting DnaK 37(cid:2)C DnaK+#1 3 0 12 50 clients in DnaK+lines (Fisher’s exact test: odds ratio F=1.06, DnaK+#2 7 2 19 91 P=0.77) and DnaK(cid:3) lines (Fisher’s exact test: odds ratio D DnaK+#3 11 2 25 91 F=1.04,P=0.999). ow DnaK(cid:3) 12 1 17 108 nlo 42(cid:2)C DnaK+#1 11 1 22 95 StrongDnaKClientsAccumulateMoreNonsynonymous ade DDnnaaKK++##23 139 32 2272 110135 MutationsthanWeakClients d fro m DnaK(cid:3) 11 1 15 99 Next,westudiedifstronglyinteractingDnaKclientsaremore h robust to mutations than weakly interacting clients, as evi- ttp aExperimentalevolutionlinessequencedinthisstudy.Foreachtemperature, s wesequencedthreelinesoverexpressingtheDnaK–DnaJ–GrpEchaperonesystem denced by the pattern of mutations fixed in the mutation ://a (DnaK+lines)andacontrollinewherethissystemisexpressedatwildtypelevels accumulation experiment under DnaK overexpression. To ca (DnaK(cid:3)line). d e bStrong clients are those with a high relative enrichment factor on DnaK assess how strongly aproteindepends on DnaK forfolding, m withicnWtehaektchliierdntqsuaarertitlheooseftwhiethdaistlroibwurteiolant.iveenrichmentfactoronDnaKwithin we used recent experimental proteomic data which deter- ic.ou thefirstquartileofthedistribution. mined how strongly 668 DnaK-interacting proteins interact p.c with DnaK by measuring the fraction of cellular protein o m boundtoDnaKat37(cid:2)C,apropertythatcorrelateswithchap- /g b SLt-huaeprapsblteianmrotesoenf,taatrhlyleMDenavatoeKlru+iatliliononenslienoxevp)ee.rrIenimxtpherenetsp,sreebdsuetDnacnelasoKofantthoetthionendleuyncaedtr, Eeti.rmocenoeloisdftetrhapieenn,pdsrtoerntoecniyngfocolnriefDnotnldsaianKrge(Camnaodllroemnpiaerinottnaeeln.ta2on0fc1oe2ram).nIdangragesr(cid:2)eidgdeanntaceKes e/article-a except for one of the DnaK+lines evolved at 42(cid:2)C. tphraontewineoankcDlinenaKts,isinadicgaotoindgpthroaxtythfoerrethlaetivdeepenenridchemnceentupoofna bstrac However, this loss of overexpression occurred towards the DnaKforfolding(Callonietal.2012).Weconsiderasstrong t/8 end of the experiment and even then DnaK was still over- /9 clientsthosewitharelativeenrichmentfactoronDnaKwithin /2 expressedformostofthedailygrowthcycleofthisline(sup- 9 the third quartile of the distribution of DnaK dependency 79 plementaryfig.S2,SupplementaryMaterialonline).Innoline /2 (N=167), and weak clients those within the first quartile 2 didweobserveoverexpressionintheabsenceoftheinducer. 3 ThecontrolDnaK(cid:3)linesalwaysexhibitedwild-typeexpression (N=167).Therefore,strongclientsincludethoseclientswith 622 thehighestDnaKdependency,whereasweakclientsinclude 5 levelsofDnaK. b InthegenomesoftheevolvedDnaK+lines,wefirststudied clientswiththelowestchaperonedependency. y g IntheDnaK+linesevolvedat37(cid:2)C,wefoundmorenon- ue the incidence of nonsynonymous nucleotide substitutions s synonymoussubstitutionsinstrongclients(21mutations)than t o amongDnaKclientsandnonclients(table1).Inthisanalysis, n in weak clients (4 mutations) (table 1). Considering the 3 weconsideredasnonclientsallproteinsfromtheE.colipro- 1 number of nonsynonymous sites in strong clients (43,731 M teome that are not part of a set of 674DnaK clients deter- a sites)andweakclients(41,238sites),thisdifferencewassig- rc minedbyCallonietal.(2012),andanalyzedthelinesevolved h at 37(cid:2)C and 42(cid:2)C independently. To improve statistical nificant (Fisher’s exact test: F=4.95, P=0.001; fig. 3B). At 20 power, we combined mutations across DnaK+lines evolved 42(cid:2)C, the results were similar, with 33 mutations in strong 19 clients and 6 in weak clients (Fisher’s exact test: F=5.12, at the same temperature after classifyingthem accordingto P=1.9(cid:4)10(cid:3)5;fig.3B;table1).Inconclusion,atbothtem- whether they affect DnaK clients or nonclients. If DnaK is peratures, client proteins that are more dependent upon buffering deleterious mutations, we would expect a higher DnaK for folding accumulate significantly more mutations proportionofmutationsaffectingclientsinthelinesevolved thanlessdependentclients. underDnaKoverexpression. In the DnaK(cid:3) line evolved at 37(cid:2)C, ~14% of nonsynon- DnaKAcceleratesProteinEvolutionOnIntermediateand ymous mutations (17 out of 125) affected DnaK clients. LongEvolutionaryTimeScales Compared with this proportion when DnaK is not overex- pressed, the proportion of mutations in clients in the WewantedtofindoutiftheDnaK-mediatedmutationalbuff- DnaK+lines was significantly higher (56 out of the total 288 eringweobservedontheshorttimescalesoflaboratoryevo- mutations,~19%;binomialtest:P=0.006;fig.3A).Similarly, lution has also left signatures on longer evolutionary time GenomeBiol.Evol.8(9):2979–2991. doi:10.1093/gbe/evw176 AdvanceAccesspublicationAugust6,2016 2983 GBE Aguilar-Rodr´ıguezetal. D o w n lo a d e d fro m h ttp s ://a c a d e m ic .o u p FIG.2.—DnaKabundanceatthebeginningandtheendofthemutationaccumulationexperiment.Wemeasuredtheabundanceofthechaperone .c o DnaKforthe8sequencedlinesevolvedtrough85single-cellbottlenecks(~1,870generations)at37(cid:2)Cor42(cid:2)C.Forcomparison,wealsomeasuredthe m abundanceofthechaperoneintheancestralDnaK+andDnaK(cid:3)strainsatbothtemperatures.WedeterminedDnaKlevelsinthepresenceandabsenceofthe /gb e inducerL-arabinose(L-ara+andL-ara–,respectively),asdescribedinMaterialsandMethods(“VerificationofDnaKoverexpression”),viatheintensityofthe /a DnaKbandinaWesternblot.TheevolvedlinesdidnotlosetheabilitytooverexpressDnaKinthepresenceoftheinducerL-arabinoseexceptforaDnaK+line rtic evolvedat42(cid:2)C(line#2),whichexplainsthedecreaseintheaverageDnaKabundanceattheendoftheevolutionexperiment.However,thislossof le-a b overexpressionoccurredlateintheevolutionexperiment,anditisnotevencompleteformostofthedailygrowthcycleofthisline(supplementaryfig.S2, s SupplementaryMaterialonline).TheheightofthebarsindicatesmeanDnaKabundanceacrosstwoexperimentalreplicatesperstrainandcondition.Error tra c barsrepresent1SDofthemean. t/8 /9 /2 9 7 9 /2 scales. To this end, we determined two measures of evolu- correlation coefficient, r=0.367, N=627, P<2.2(cid:4)10(cid:3)16; 2 3 6 tionary rates for protein-coding genes from gamma-proteo- fig. 4A). This indicates that the stronger the interaction of a 2 2 bacteria. The first, nonsynonymous divergence among one- proteinwith DnaK, the faster the protein evolves. The same 5 b y to-one orthologs of E. coli andS. enterica, isrelevant forin- pattern is obtained at the larger time scales of protein dis- g u termediate evolutionary time scales. The second, protein tances for 85 gamma-proteobacterial genomes (r=0.257, e s (amino acid) distance among orthologous proteins found in N=311, P=4.4(cid:4)10(cid:3)6; fig. 4B). Gene expression level, t o n 85gamma-proteobacterialgenomes(includingE.coliandS. whichisthemostimportantdeterminantofproteinevolution- 3 1 enterica),isrelevantforlongtimescales.Weemployprotein ary rates, at least in unicellular organisms (Pa´l et al. 2001; M a distanceinsteadofnonsynonymousdistancebecauseamino Drummondetal.2005;ZhangandYang2015),isapossible rc h acidreplacementsarelesssensitivethannucleotidesubstitu- confoundingfactorinthisanalysis.Forexample,usingcodon 20 1 tionstotheexpectedlossofphylogeneticsignalbetweense- usagebias(CUB)asaproxyforgeneexpression,weobserve 9 quences of distantly related taxa. To assess how strongly a thatgeneswithhigherCUBshowlowernonsynonymousdi- proteindependsonDnaKforfolding,weusetherelativeen- vergence (r=(cid:3)0.558, N=1014, P<2.2(cid:4)10(cid:3)16), protein richment of the protein on DnaK as a proxy for the depen- distance(r=(cid:3)0.255,N=3159,P<2.2(cid:4)10(cid:3)16)andDnaK dence of the protein upon DnaK for folding (Calloni et al. dependency (r=(cid:3)0.262, N=627, P=2.5(cid:4)10(cid:3)11). 2012). We note that this interaction strength is more likely However, the association between DnaK dependency and tohaveremainedunchangedduringthedivergenceofE.coli evolutionaryratecannotbesolelyexplainedbythisconfound- andS.enterica,thanduringthedivergenceofalltheother83 ingfactor:Apartialcorrelationanalysisshowsthattheasso- gamma-proteobacterialspeciesweanalyzed. ciation still holds after controlling for CUB, both on Wefindastrongandhighlysignificantpositiveassociation intermediate time scales (r=0.295, N=627, P= betweenDnaKdependencyandtherateofnonsynonymous 1.2(cid:4)10(cid:3)14) and long time scales (r=0.229, N=311, substitutions for S. enterica and E. coli (Spearman’s rank P=3.8(cid:4)10(cid:3)5). We use CUB instead of gene expression 2984 GenomeBiol.Evol.8(9):2979–2991. doi:10.1093/gbe/evw176 AdvanceAccesspublicationAugust6,2016 GBE TheChaperoneDnaKIsaSourceofMutationalRobustness D o w n lo a d e d fro m h ttp s ://a c a d e m ic .o u p .c o m FIG.3.—NonsynonymousmutationsaccumulatedinDnaKclients.(A)TheproportionofnonsynonymousmutationsthataffectDnaKclientsissignif- /g b icantlyhigherinDnaK+linesthatoverexpressDnaKthaninthecontrolDnaK(cid:3)linesthatdonotexpressthechaperoneatsuchhighlevels.Thisisobserved e /a bothforlinesevolvedat37(cid:2)Cand42(cid:2)C.WecombinedmutationsacrossDnaK+linesevolvedatthesametemperature.Thesignificanceofthedifferencein rtic the proportions was evaluated using a binomial test. (B) At both temperatures, strong clients have accumulated significantly more nonsynonymous le substitutionsthanweakclientsinDnaK+lines.StrongclientsincludethoseclientswiththehighestDnaKdependency,whereasweakclientsincludeclients -ab s withthelowestchaperonedependency.StatisticalsignificancewasevaluatedusingFisher’stest. tra c t/8 /9 /2 dataherefortwomainreasons.First,wecancomputeCUB online). The reason for this observation couldbe that clients 9 7 9 for all 631 DnaK clients in our data set, whereas expression areintrinsicallylessrobusttomutationsthannonclientsdueto /2 2 dataisonlyavailablefor457clients.Second,geneexpression some general physicochemical difference. For example, 3 6 2 data has been measured in just one environment and one Calloni et al. (2012) found that DnaK clients have generally 2 5 strainofE.coli,whereasCUBistheresultofselectivepressures low solubility, often belong to heterooligomeric complexes, b y imposedbymanydifferentenvironmentsoverlongperiodsof and are prone to misfolding. However, in accordance with gu e time.Nonetheless,theassociationbetweenevolutionaryrate the mutational buffering hypothesis we observe that strong s t o and DnaK dependency still holds after correcting for gene clientsevolvefasterthanweakclients(fig.5;supplementary n 3 expression directly (supplementary information section 1.1, informationsection1.3,SupplementaryMaterialonline).The 1 M Supplementary Material online). Together, these results indi- accelerated evolution of strong clients compared with weak a rc cate that the chaperone DnaK affects protein evolution in clientsexactlymirrorsthegreateraccumulationofnonsynon- h 2 accordance with the mutational buffering hypothesis. ymousmutationsinstrongclientsduringtheevolutionexper- 0 1 Importantly, this effect is not only independent of CUB and iment(fig.4B). 9 geneexpression,butalsoofotherbiologicalfactors,suchas essentiality and number of protein–protein interactions DnaK-MediatedAccelerationofProteinEvolutionIs (supplementary information section 1.2 and table S2, IndependentofGroELBuffering SupplementaryMaterialonline). Inasubsequentanalysis,wefindthatclientsevolvemore TheabilityofDnaKtofacilitatetheaccumulationofnonsynon- slowlythannonclients(supplementaryfig.S3andsupplemen- ymous mutations in DnaK clients resembles the well-studied taryinformationsection1.3,SupplementaryMaterialonline). mutational buffering by the chaperonin GroEL (Fares et al. This last difference cannot be explained by the number of 2002; Tokuriki and Tawfik 2009; Bershtein et al. 2013; protein–proteininteractions,byessentiality,orbyCUBascon- Wyganowski et al. 2013; Sabater-Mun˜oz et al. 2015). foundingfactors(supplementaryinformationsection1.3and Additionally,theobservedcorrelationbetweenDnaKdepen- supplementary tables S3 and S4, Supplementary Material dencyandproteinevolutionaryratesissimilartothepreviously GenomeBiol.Evol.8(9):2979–2991. doi:10.1093/gbe/evw176 AdvanceAccesspublicationAugust6,2016 2985 GBE Aguilar-Rodr´ıguezetal. D o w n lo a d e d fro m h ttp s ://a c a d e m ic FIG.4.—DnaKacceleratesproteinevolutiononintermediateandlongevolutionarytimescales.Scatter-plotsshowingtherelationshipbetweenDnaK .ou p dependency(calculatedasarelativeenrichmentfactorthatindicatesthefractionofcellularproteinboundtoDnaKat37(cid:2)C,horizontalaxis)andthedegree .c o of divergence over (A) intermediatetimescales,measured asnonsynonymous divergence (Spearman rankcorrelation coefficient,r=0.367,N=627, m P<2.2(cid:4)10(cid:3)16), and (B) long timescales, measured as protein (amino acid) distance (r=0.257, N=311, P=4.4(cid:4)10(cid:3)6) (vertical axes). Solid lines /gb e representthebestfittothepoints.Notethelogarithmicscaleonbothaxes. /a rtic le -a b s tra c t/8 /9 /2 9 7 9 /2 2 3 6 2 2 5 b y g u e s t o n 3 1 M a rc h 2 0 1 9 FIG.5.—Strongclientsevolvefasterthanweakclients.(A)Wefindthatstrongclientsevolvefasterthanweakclientsonintermediateevolutionarytime scales,measuredastherateofnonsynonymoussubstitutions(Wilcoxonrank-sumtest,P<2.2(cid:4)10(cid:3)16).(B)Onlongevolutionarytimescales,wealsofind thatstrongclientsevolvefasterthanweakclients(Wilcoxonrank-sumtest,P=2.3(cid:4)10(cid:3)3).Thethickhorizontallineinthemiddleofeachboxrepresentsthe medianofthedata,whereasthebottomandtopofeachboxrepresentthe25thand75thpercentiles,respectively.Notethelogarithmicscaleonthey-axis in(A). reportedaccelerationofproteinevolutionbyGroEL(Bogumil mutation accumulation and evolutionary rates. We defined and Dagan 2010; Williams and Fares 2010). We therefore theGroELinteractomeinE.coliastheunionoftwopreviously removedknownGroELclientsfromourdatasettoinvestigate reportedsetsofDnaKinteractors(Kerneretal.2005;Fujiwara ifourobservationsareindependentoftheeffectofGroELon etal.2010).Ofthe253GroELclientsthatcomprisetheGroEL 2986 GenomeBiol.Evol.8(9):2979–2991. doi:10.1093/gbe/evw176 AdvanceAccesspublicationAugust6,2016 GBE TheChaperoneDnaKIsaSourceofMutationalRobustness interactome, there are 122 proteins that are also clients of Forexample,clientsarepronetoaggregationandmisfolding DnaK. (Calloni et al. 2012), which may make them intrinsically less The observation that DnaK overexpression increases the robusttodestabilizingmutations. proportion of nonsynonymous substitutions affecting DnaK Despitethegreatdifferencesinthemechanismofchaper- clients is still significant after removing GroEL clients. oneactionbetweenthethreemajorchaperonefamilies—cha- CombiningmutationsfromtheDnaK+linesevolvedat37(cid:2)C peronins,Hsp90chaperonesandHsp70chaperones—(Young and42(cid:2)Cwefindthat~16%ofmutations(97outof 621) etal.2004;Hartletal.2011;BogumilandDagan2012;Kim affectedDnaKclients,whichissignificantlyhigherthanwhat et al. 2013), at least some of their members seem to have wefindintheDnaK(cid:3)lines(28outof230mutations,~12%; qualitativelycomparableeffectsonproteinevolution.Butpro- binomialtest: P=0.01). Similarly,consideringthenumber of teinchaperonesarenottheonlychaperonesthatcanincrease D nonsynonymoussitesinstrongclients(36,026sites)andweak themutationalrobustnessoftheirsubstrates:Arecentstudy ow n clients(33,047sites)afterremovingGroELclients,westillfind hasfoundthatsomeRNAchaperonescanbufferdeleterious lo a that strong DnaK clients accumulate more nonsynonymous mutationsinE.coliandthereforeaffectRNAevolution(Rudan de d srautbiostiFtu=ti4o.n7s, Pth=an2(cid:4)w1e0ak(cid:3)5c).lieFnintsall(yF,isthheer’spoesxitaivcet atesssot:ciaotdiodns leatteadl.t2o01p5ro).teTihnecsehacphearpoenreosn,easr,ewRhNicAh-bairnedicnogmpprloetteeliynsutnhraet- from h betweenDnaKdependencyandevolutionaryratesstillholds facilitatetheproperfoldingofRNAmolecules.Elucidatingto ttp afterremovingGroELclientsandcontrollingforCUBinapar- whatextentthebufferingmechanismsofallthesechaperones s://a tial correlation analysis, both on intermediate time scales differisanimportantfuturedirectionofenquiry. c a (r=0.318, N=511, P=3.9(cid:4)10(cid:3)14) and long time scales Thankstotheirfosteringofmutationalrobustness,chaper- de m (r=0.226,N=240,P=3.6(cid:4)10(cid:3)4). onescanfacilitateevolutionaryinnovations(Rutherford2003), ic .o eventhoughwedonotstudysuchinnovationshere.Thein- u p Discussion crease in the mutational robustness of a protein caused by .co m chaperoneinteractionsreducestheefficiencyofpurifyingse- /g We show how the overexpression of the DnaK–DnaJ–GrpE b lectioninpurgingmutationsintheprotein. Thanks tochap- e chaperonesystemoverthecourseofamutationaccumulation /a experimentincreasestheproportionofnonsynonymoussub- erone-mediated buffering, many such mutations are neutral rtic andcanpersistinapopulation.Importantly,thesecrypticge- le stitutions affecting DnaK clients. In addition, strong clients -a netic variants may include preadaptive mutations that can b s accumulate more nonsynonymous mutations than weak cli- generate evolutionary innovations in new environments tra ents.AdditionalevidenceofmutationalbufferingbyDnaKis (TokurikiandTawfik2009;Wyganowskietal.2013).Toillu- ct/8 providedbytheobservationthatevolvinglinesoverproducing /9 minate if and how DnaK can increase the ability to evolve /2 thischaperoneavoidextinctionafter experiencing85single- 9 functional innovations of its client proteome will also be an 7 cellbottlenecks.Recently,weobtainedsimilarresultsinhyper- interestingsubjectforfuturework. 9/2 2 mutable E. coli cells evolving in identical conditions but 3 Insummary,weanalyzedtheevolutionofproteinsthatare 6 overproducing the GroEL-GroES chaperonin system subject to DnaK-assisted folding on short, intermediate, and 225 (Sabater-Mun˜oz et al. 2015). There, we observed that lines b longevolutionarytimescalesthroughacombinationofexper- y evolvingwithhighlevelsofGroELwerenotonlylessproneto g imental and comparative approaches. Most of our evidence u e extinction under strong genetic drift than control lines, but s also that they were accumulating significantly more indels itniodnicsaitnesittshacltietnhtepbraoctteeirnias,lcahnadpethroantethDenseaKpcroatneibnusfftehremreufotare- t on 3 andreplacementsbetweenaminoacidsbelongingtodifferent 1 evolvefasterthanintheabsenceofDnaK-mediatedfolding. M phyWsiceoaclhsoemfinicdalthcaatteDgnoariKe-sm. ediatedmutationalbufferinghas This is, to our knowledge, the first demonstration that a arch member of the Hsp70family can buffer the effect of muta- 2 leftatraceinDnaKclientsduringthedivergenceof85differ- 0 tions, with long-term consequences on protein evolution 1 ent gamma-proteobacterial species over much longer evolu- 9 (BogumilandDagan2012).Throughitsroleinproteinfolding, tionary time scales than those explored in our laboratory anindividualchaperonesuchasDnaKcanhaveadispropor- evolution experiment. We find that clients that depend tionate effect on proteome evolution, and thus on genome moreonDnaKforfoldingtendtoevolvefasterthanlessin- evolution. teractingclients.Similarchaperone-mediatedaccelerationsof protein evolution have been observed in GroEL clients (Bogumil and Dagan 2010; Williams and Fares 2010) and Materials and Methods Hsp90 clients (Lachowiec et al. 2013; Pechmann and BacterialStrainsandPlasmids Frydman 2014). However, we notice that DnaK clients evolveslowerthanproteinsnotknowntobeDnaKinteractors WeobtainedE.coliK-12substr.MG1655(cid:2)mutS::FRTfrom (Callonietal.2012).Thisislikelytheresultofimportantphys- Ivan Matic (Universite´ Paris Descartes, INSERM U1001, Paris, icochemical differences between clients and nonclients. France) through Jesu´s Bla´zquez (Centro Nacional de GenomeBiol.Evol.8(9):2979–2991. doi:10.1093/gbe/evw176 AdvanceAccesspublicationAugust6,2016 2987 GBE Aguilar-Rodr´ıguezetal. Biotecnolog´ıa, CSIC, Madrid, Spain) (Sabater-Mun˜oz et al. debrisbycentrifugation,wequantifiedsolubleproteinsusing 2015). In this E. coli strain, the gene encoding the protein the Bradford method (Bradford 1976). We loaded 1 mg of MutS has been deleted. This protein is a component of the total protein for each sample in SDS–PAGE gels (12.5% re- mismatchrepairsystem that recognizesandbindsmispaired solvinggel).Inaddition,weloadedontoallgelssamplesfrom nucleotides so that the mispairing can be corrected by two theancestralDnaK(cid:3)andDnaK+strainsgrowninthepresence furtherrepairproteins,MutLandMutH.ThestrainMG1655 ofinducerat37(cid:2)C,ascontrolstofacilitateinter-gelcompar- (cid:2)mutShasapredictedmutationratethatis1000-foldhigher isons.WedetectedDnaKproteinbyWesternblottingusingas thanthewildtype(Turrientesetal.2013),whichensuresthat primaryantibodyamousemonoclonalantibodyspecifictoE. asufficient numberofmutations occur duringthemutation coliDnaK(Abcam#ab69617)ata1:10,000dilution,andas accumulationexperiment.Wetransformedthisstrainwiththe secondary antibody a goat polyclonal (alkaline phosphatase- D plasmid pKJE7 (Takara, Cat. #3340), which contains an conjugated) antibody specific to mouse IgG1 (Abcam ow n operon encoding DnaK, and its co-chaperones DnaJ and #ab97237).Wescannedmembranesaftercolorimetricdetec- lo a GrpEundertheregulationofasinglepromoterinducibleby tion of conjugated antibodies with the BCIP(cid:3)/NBT-Purple de d L-arabinose (Nishihara et al. 1998). We generated a control liquid substrate system (Sigma-Aldrich #BP3679), and used fro strainbytransformingthesame(cid:2)mutSstrainwithaplasmid ImageJ to quantify the intensity of DnaK bands on the m h thatlackstheoperondnaK–dnaJ–grpEbutisotherwiseiden- Western blots (Schneider et al. 2012). We used the control ttp ticaltopKJE7.WerefertothisplasmidaspKJE7-DEL(dnaK– samplestonormalizeabundances,whichallowthecompari- s://a dnaJ–grpE).Thiscontrolplasmidwasderivedfromtheplasmid sonofDnaKlevelsacrossexperiments. c a pKJE7byremovaloftheoperondnaK–dnaJ–grpEwithare- WeexaminedthechangeinDnaKlevelsalongadailycycle de m strictiondigestusingBamHIandSpeI,followedbyreligation, ofgrowthforaDnaK+lineevolvedat42(cid:2)C(line#2,supple- ic after obtaining permission for plasmid modification from mentaryfig.S2,SupplementaryMaterialonline)thatshowed .ou p Takara. alowedDnaKlevelafter~1,870generationsofmutationac- .c o cumulation. After 24 h of exponential growth at 42(cid:2)C in m/g EvolutionExperiment liquid LB medium supplemented with chloramphenicol, we be /a Weevolved68clonallinesofthehypermutableE.coli(cid:2)mutS dilutedtheculturetoOD~0.3,andinducedDnaKexpression rtic straincontainingpKJE7(DnaK+lines)and16linescontaining byadding10mMofL-arabinose.Weallowedthecultureto le-a the control plasmid pKJE7-DEL(dnaK–dnaJ–grpE) (DnaK(cid:3) growforanother24hinthepresenceofthisexpressionin- bs lines)bydailypassagingthemthroughsingle-cellbottlenecks ducer.Eachhour,1mlofculturewasremovedandtheDnaK trac onsolidLBmedium(agarplates;Pronadisa#1551and#1800) levelfollowingtheprotocoldescribedearlierwasmeasured. t/8 /9 supplemented with 20 mg/ml of chloramphenicol (Sigma- /2 9 Aldrich #C0378) (fig. 1). Except for 8 DnaK+lines and 8 Whole-GenomeResequencing 79 DnaK(cid:3)lines,alltheremaininglineswereevolvedinthepres- We sequenced the genomes of 2 DnaK(cid:3) and 6 DnaK+lines /223 ence of 0.2% (w/v) of L-arabinose (Sigma-Aldrich #A3256), after85single-cellbottlenecks.Alloftheselinesevolvedinthe 622 which induces the expression of DnaK/DnaJ/GrpE from the 5 presence of L-arabinose in the medium, although only b plasmid pKJE7 but not from the control plasmid pKJE7- DnaK+cells are able to overexpress DnaK. Half of the se- y g DEL(dnaK–dnaJ–grpE). We passaged both the DnaK(cid:3) and quenced DnaK(cid:3) and DnaK+lines evolved at 37(cid:2)C, whereas ues DnaK+linesduring85daysor~1,870generations(conserva- the other lines evolved at 42(cid:2)C. We used the genome se- t on tively assuming ~22 generations per daily growth cycle), quence of the ancestral (cid:2)mutS strain from which both the 31 eexncdepotfftohretheoxpseerliimneesntth.aWteweenvotlevextdinhctalbfeofofrtehereaDcnhainKg+atnhde D(SnaabKat+earn-MduDn˜noazKe(cid:3)tlianl.es20w1e5re).derivedfromourpreviousstudy March DnaK(cid:3) lines under mild heat-stress (42(cid:2)C) whereas the 2 Specifically,fortheevolvedlinesweperformedpaired-end 0 otherhalfremainedat37(cid:2)C. Illuminawhole-genomesequencing.ForDNAextraction,we 19 used the QIAmp DNA mini kit (Qiagen, Venlo [Pays Bas], VerificationofDnaKOverexpression Germany)inaQiaCubeautomaticDNAextractorusingbac- Wegrewtheancestralandevolvedstrains(DnaK+andDnaK–, terialpelletsobtainedfrom~10mlcultures.Weconstructed at37(cid:2)Cand42(cid:2)C)fromglycerolstocksinliquidLBmedium multiplexed DNAseq libraries from each clonal evolution line supplementedwith20mg/mlofchloramphenicolinthepres- using the TrueSeq DNA polymerase chain reaction-free HT enceorabsenceoftheinducerL-arabinose(0.2%).After24h sample preparation kit (Illumina). We performed paired-end ofgrowth,wepelletedcellsbycentrifugationat12,000rpm. sequencing on an Illumina HiSeq2000 platform, using a Weresuspendedthepelletedcellsin100mllysisbuffer(con- 2(cid:4)100cyclesconfiguration. taining200mMTris–HClpH6.8,10mMDTT,5%SDS,50% We converted sequencing reads from Illumina quality glycerol). To prepare a crude extract, we first boiled resus- scoresintoSangerqualityscores.Subsequently,weusedthe pended cells at 95(cid:2)C for 15 min. After the removal of cell breseqv0.24rc4(version4)pipeline(DeatherageandBarrick 2988 GenomeBiol.Evol.8(9):2979–2991. doi:10.1093/gbe/evw176 AdvanceAccesspublicationAugust6,2016
Description: