Unmasking the ancestral activity of integron integrases reveals a smooth evolutionary transition

13 Pages·2017·1.41 MB·English

Unmasking the ancestral activity of integron integrases
reveals a smooth evolutionary transition during
functional innovation
Jose Antonio Escudero, Celine Loot, Vincent Parissi, Aleksandra Nivina,
Christiane Bouchier, Didier Mazel
To cite this version:
Jose Antonio Escudero, Celine Loot, Vincent Parissi, Aleksandra Nivina, Christiane Bouchier, et
al.. Unmasking the ancestral activity of integron integrases reveals a smooth evolutionary tran-
sition during functional innovation. Nature Communications, 2016, 7, pp.Article number 10937.
￿10.1038/ncomms10937￿. ￿pasteur-01292328￿
HAL Id: pasteur-01292328
https://hal-pasteur.archives-ouvertes.fr/pasteur-01292328
Submitted on 22 Mar 2016
HAL is a multi-disciplinary open access L’archive ouverte pluridisciplinaire HAL, est
archive for the deposit and dissemination of sci- destinée au dépôt et à la diffusion de documents
entific research documents, whether they are pub- scientifiques de niveau recherche, publiés ou non,
lished or not. The documents may come from émanant des établissements d’enseignement et de
teaching and research institutions in France or recherche français ou étrangers, des laboratoires
abroad, or from public or private research centers. publics ou privés.
Distributed under a Creative Commons Attribution| 4.0 International License
ARTICLE
Received5Oct 2015|Accepted 3Feb2016 |Published 10Mar2016 DOI:10.1038/ncomms10937 OPEN
Unmasking the ancestral activity of integron
integrases reveals a smooth evolutionary
transition during functional innovation
Jose Antonio Escudero1,2, Celine Loot1,2, Vincent Parissi3, Aleksandra Nivina1,2,4, Christiane Bouchier5
& Didier Mazel1,2
Tyrosine (Y)-recombinases have evolved to deliver mechanistically different reactions on a
varietyofsubstrates,buttheseevolutionarytransitionsarepoorlyunderstood.Amongthem,
integron integrases are hybrid systems recombining single- and double-stranded DNA
partners. These reactions are asymmetric and need a replicative resolution pathway, an
exception to the canonical second strand exchange model of Y-recombinases. Integron
integrasespossessaspecificdomainforthisspecializedpathway.Hereweshowthatdespite
this,integrasesarestillcapableofefficientlyoperatingtheancestralsecondstrandexchange
in symmetrical reactions between double-stranded substrates. During these reactions, both
strandsarereactiveandHollidayjunctionresolutioncanfolloweitherpathway.Anoveldeep-
sequencingapproachallowsmappingofthecrossoverpointforthesecondstrandexchange.
Thepersistenceoftheancestralactivityinintegrasesillustratestheirrobustness andshows
that innovation towards new recombination substrates and resolution pathways was a
smooth evolutionary process.
1InstitutPasteur,Unite´dePlasticite´duGe´nomeBacte´rien,De´partementGe´nomesetGe´ne´tique,28RueduDrRoux,75015Paris,France.2CNRS,UMR3525,
28RueduDrRoux,75015Paris,France.3LaboratoiredeMicrobiologieFondamentaleetPathoge´nicite´UMR-CNRS5234.Bat3A3e`mee´tage,146rueLe´o
Saignat,33076Bordeaux,France.4Universite´ParisDescartes,SorbonneParisCite´,75006Paris,France.5InstitutPasteur,GenomicPlatform,28Ruedu
DrRoux,75015Paris,France.CorrespondenceandrequestsformaterialsshouldbeaddressedtoD.M.(email:[email protected]).
NATURECOMMUNICATIONS|7:10937|DOI:10.1038/ncomms10937|www.nature.com/naturecommunications 1
ARTICLE
NATURECOMMUNICATIONS|DOI:10.1038/ncomms10937
T
yrosine (Y)-recombinases mediate recombination between and the aHJ is then resolved through replication21 (Fig. 1a). To
two specific DNA sites to deliver a variety of biological accommodateEHBs,integrasespossessa20-residue-longdomain
functions(forareviewsee(ref.1)).Theyarepresentinthe dubbedI2(ref.22;Fig.1c)thathasnotbeenidentifiedelsewhere,
three domains of life but are most widespread among suggesting that it could have emerged recently23 among
prokaryotes, where they play fundamental roles such as the Y-recombinases. Such domain insertions are well-characterized
integration/excision of viral genomes2, the alternation of gene phenomena in protein evolution24, allowing for a modular
expression3andtheresolutionofdeleteriouschromosomedimers organization of protein function.
arising during replication4. Members of this family show a ssDNA is central to the integron, also controlling integrase
conserved fold of the catalytic domain including a three- expression through the SOS response25. Integrons are, therefore,
dimensional clustering of highly conserved RKHRH residues. unique recombination platforms streamlined to recombine
Y-recombinasestypicallyrecognizespecificDNAsitescomprising ssDNA through a distinct resolution pathway. They represent a
twoinvertedbindingdomainsseparatedbya6–8bpspacer.The molecular example of functional innovation26.
recombination reaction is a stepwise process that starts with the While attC sites have been extensively studied17,19,20,27,28,
cleavage and transfer of one strand, leading to the transient knowledge is lacking on the recognition and processing of attI
formation of a Holliday junction (HJ) that is further resolved sites. These are more canonical than attC sites, since they are
through a second exchange of strands. This sequence of steps, recognized as dsDNA16. They have two binding domains, the
whichwewillrefertoastheclassicalpathway(Fig.1a),ishighly L and R boxes, although one is degenerate (Fig. 1d), and are
conserved among Y-recombinases. Type IB topoisomerases and generally recognized specifically by their cognate integrases29.
telomere resolvases share the fold and catalytic chemistry, and Some attI sites possess accessory binding domains30.
can therefore be considered as distant relatives within the same Integrasesiterecognitionunderscoresaremarkableconforma-
superfamily, despite sequence divergence1,5–7. This implies that tionaldynamism31.Identicalmonomersrecognizeds-attIsitesby
recombinases have evolved to deliver mechanistically distinct sequenceandss-attCsitesthroughstructurallandmarks.Froman
reactions (for example, the cleavage of only one strand by evolutionary standpoint, this duality could reflect a single
topoisomerases),andtorecognizestructurallydifferentsubstrates specialized activity (the recombination of ss/ds DNA partners)
(for example, hairpins in telomeres). Such evolvability is likely or two distinct ones: the canonical Y-recombinase activity of
facilitated by the robustness of the fold6 (its capacity to integraseancestorsonattI,andthefunctionalinnovation,thatis,
accommodate mutations while preserving its tertiary structure), the recombination of ssDNA structures, on attC sites. For the
as well as the modularity of protein function through the latter to be true, both classical and innovative activities must be
combination of different functional domains. Nevertheless, complete:integrasesshouldbecapableofresolvingHJsthrougha
information on evolutionary pathways that enable functional second strand exchange.
innovation among these enzymes is scarce. Tounmasktheancestralactivityofintegrases,wehavestudied
Integrons are recombination platforms that allow bacteria to thereactionbetweentwoattIsites.Thisisasymmetricalreaction
evolveandadaptrapidlythroughtheacquisition,stockpilingand involving exclusively ds partners for which resolution through a
reorderingofgenesembeddedinmobilegeneticelementsnamed secondstrandexchangeistheoreticallypossible,sinceitisneither
integron cassettes8–10 (Fig. 1b). They are ancient structures abortive nor impeded by EHBs. We show that the top strand of
that have driven bacterial evolution for hundreds of millions attI1 is reactive and that HJ resolution can take place through a
of years11, and, through their mobilization onto conjugative secondstrandexchange.Wehavedeterminedthecrossoverpoint
plasmids,theyhaveplayedacriticalroleintheriseofmultidrug through a novel deep-sequencing approach that is applicable
resistance among clinically relevant bacteria12. Recombination to other recombination systems, and validated this approach
in integrons is governed by integron integrases13 (hereafter invitro.Finally,weprovethatattCsitesgovernthedirectionality
integrases), a distinct group of Y-recombinases closely related to ofthereaction.Toourknowledge,integrasesarethefirstexample
XerCDrecombinases8.Integrasesshowtheconservedfoldofthe amongrecombinasesofasmoothevolutionarytransitionbetween
family,aswellastheRKHRHpentadofresiduesintheactivesite substrates and processing pathways. We discuss the possible
(Fig. 1c). Incorporation of cassettes into the integron platform origins and the evolutionary implications of the dual activity of
occurs through site-specific recombination between the attC site integrases, as well as the importance of attI1(cid:2)attI1 reactions in
in the cassette and the attI site in the integron14. Genes in antibiotic resistance.
cassettes are then expressed from the Pc promoter within the
integron platform (Fig. 1b). Cassette recombination represents
Results
an exception to the archetypical reaction mediated by
For this work we chose to study attI1, the integration site of the
Y-recombinases, as a consequence of the structural peculiarities
mostwidespreadand best-studied integronplatform, theClass1
of attC sites. These sites show no sequence conservation, except
integron.
for the crossover point on the bottom strand (bs) (50-AAC-30),
thatprovidesmicrohomologywithattIsites,necessaryforstrand
transfer15. Instead, attC sites share a conserved palindromic Dispensability of known HJ-processing host factors. The attI1
structure,andareonlyrecognizedasafoldedsinglestrand(ss)16 site possesses two direct repeats upstream the L–R boxes that
(Fig.1d).Recombinationbetweenass-attCandthedoublestrand make the site B65-bp long30. It is therefore possible that
(ds)-attIsiteformsanasymmetricHJ(aHJ)(Fig.1a)thatcannot host factors specialized in homologous recombination, branch
be resolved through a second exchange of strands since it would migration and resolution of HJ-like structures (such as
linearizetherepliconandbeabortive17.DuringattCsitefolding, stalled replication forks), could influence the resolution of the
the imperfect pairing of both arms of the bs allows for the attI1(cid:2)attI1 HJ, biasing our results. To assess the impact of
protrusion of a set of extrahelical bases (EHBs) (Fig. 1d) that such host machinery, we have tested the influence of RecA
are key in strand selectivity, favouring the recombination of and two well-characterized HJ-processing complexes, RecG32
the bs18,19. EHBs also induce conformational changes in one and RuvABC33,34, using a classic suicide conjugation assay35.
integrasemonomerduringthereaction,pullingapartthetyrosine Recombination rates of mutants of these complexes were not
residue and avoiding the nucleophilic attack on the L box of the statistically different from those of the wild-type (WT) strain,
attC site20 (Fig. 1d). This prevents the second strand exchange, suggesting that known host HJ-processing systems do not
2 NATURECOMMUNICATIONS|7:10937|DOI:10.1038/ncomms10937|www.nature.com/naturecommunications
ARTICLE
NATURECOMMUNICATIONS|DOI:10.1038/ncomms10937
a b
Y-recombinase-mediated site-specific recombination Pc
attI attC attC attC attC
Classical pathway Replicative pathway
Canonical reaction Cassette integration Integrase Pint Cassette 1 Cassette 2 Cassette 3 Cassette n
attI-containing
substrate Integration Excision
attI × attC attC × attC
Cassette ds attI
Shuffling
ds substrates
ss attC
First strand cleavage
n
o
ati
J form Y OHYYOH Y Y OYHYOH Y
H
First strand transfer c
VchIntIA
HJ aHJ
HJ isomerization HJ replication
Second strand cleavage
Replication fork
on progression
uti Y
sol YOHOHY
e
J r Y
H Second strand transfer
Cre
/attI
attC
Cointegrates ds attI
attC/
attI
Integrated cassette Substrate
d
attI1 site attC site
aadA7
Crossover
DR2 DR1 L R point
ts 5′ 3′ 5′ R′ L′
bs 3′ 5′ Folded bs
Crossover point 3′ R″ L″
* *
Core site
Recombinogenic form (ds) R L
Recombinogenic form (ss)
Integrase binding domains R″ L″ L′ R′
ts 5′ 3′
* Extrahelical bases (EHBs) G16 and T23 bs3′ * * 5′
Crossover point
Non-recombinogenic form (ds)
Figure1|Recombinationintheintegron.(a)Diagramshowingthetworecombinationmodelsamongtyrosinerecombinases.Theleftcolumnshowsthe
recombinationpathwaysharedbyamajorityofY-recombinases,excepttheintegronintegraseforwhichthepathwayisrepresentedintherightcolumn.
(b)Schematicrepresentationoftheintegronarchitectureandrecombinationreactions.Openreadingframesarerepresentedasboxes,witharrowheads
showingthedirectionoftranscription.ArrowsmarkedasPcandPintrepresentthepromotersofthecassettearrayandtheintegrase,respectively.
AcquisitionofnewcassettesinvolvestherecombinationoftheintegronattIsiteandthecassetteattCsite,whilecassetteexcisioninvolvestwoattCsites.
Sequentialcouplingofexcisionandintegrationreactionsshufflestheorderofcassetteswithinthearray.(c)CrystalstructureofV.choleraeintegrase20
(blue)andCre60(green).RKHRHresiduesaredepictedaspurplesticks.Thecatalytictyrosineisshowninorange.RedresiduesrepresenttheI2domainin
VchIntIAandtheresiduesinCrethatalignwiththosedelimitingI2inVchIntIA.ResiduesinVchIntIAthatinteractwithEHBsarerepresentedingreen.
(d)SequenceoftheClass1integronattIsite(attI1),andtheattC site.aHJ,asymmetricHJ;bs,bottomstrand;DR,directrepeats;ds,doublestranded;
aadA7
HJ,Hollidayjunction;ss,singlestranded;ts,topstrand.
NATURECOMMUNICATIONS|7:10937|DOI:10.1038/ncomms10937|www.nature.com/naturecommunications 3
ARTICLE
NATURECOMMUNICATIONS|DOI:10.1038/ncomms10937
significantly influence the attI1(cid:2)attI1 reaction (Supplementary respectively, to a SacIIþ/NarI(cid:3) and SacII(cid:3)/NarIþ restriction
Fig. 1). pattern. The classical pathway of double-strand exchange would
yield cointegrates containing both mismatches that, once
replicated, would produce a mixed population of plasmids and
Both strands of attI1 are reactive. In cassette integration hence partial restriction profiles with both enzymes (Fig. 2b).
reactions, only the bottom strands of attC and attI sites are WeverifiedthecorrectassemblyoftheDNAcircles(asinref.21
exchanged,withthecrossoverpointlocatedwithintheRboxesof (Supplementary Fig. 2)) and transformed them into a mutS(cid:3)
both sites. To test the hypothesis that the attI1(cid:2)attI1 reaction MG1655 Escherichia coli (E. coli) strain containing an attI-
can take place through both the bottom and the top strand, we bearing pSU plasmid (p929) as well as a plasmid with the
used an attI1 carried on a mismatched covalent circle21. These integrase under the control of a P promoter (pBAD-intI1).
BAD
syntheticcirclesmimicnon-replicativeplasmidsassembledusing The mutS(cid:3) background ensures that the mismatches are not
complementary strands produced separately and hybridized repaired. By selecting for the circle resistance marker, we
together in vitro. This allowed us to specifically tag each strand recovered clones in which these molecules had recombined
with a mutation in a restriction site (SacII or NarI) so that the with the resident pSU plasmid through an integrase-mediated
restriction pattern of cointegrates reveals which strand is attI1(cid:2)attI1 reaction. The mismatch-containing region was
transferred during recombination (Fig. 2a). In our setting, the PCR amplified from 100 recombinant clones and used for the
transfer and replication of the bottom and the top strands leads, restrictionpatternanalysis(Fig.2c),revealingthatbothstrandsof
a
SacII+
oriP15a attI1
SacII+
pSU-attI1
SacII–
SacII–
oriR6K bs
IntI1 ts
NarI+
E. coli MG1655
mutS NarI–
Single-strand Double-strand
pathway pathway
b
Top and bottom strand exchange
Top strand exchange Bottom strand exchange
+ replication + replication
NarI+
SacII+
NarI+
SacII+
PCR
amplification
NarI+ SacII+ SacII+
NarI+
M Clone 17 Clone 18 Clone 36
c (bp) NarI SacII NarI SacII NarI SacII d
Clone 36
NarI site SacII site
G G T G C C C C A C G G
1,000
500
200
100
Figure2|ThetopstrandofattI1isreactive.(a)Experimentalset-upusedformismatched-covalent-circlesrecombination.Mismatchedcirclesbearingan
attI1sitearetransformedinanE.coliMG1655mutS(cid:3) strainbearingapSU-attI1plasmidandexpressingtheintegraseintrans.(b)Representationofthe
recombinationproductsresultingfromthethreepossibleHJresolutionpathways.Themismatch-bearingregionswerePCRamplifiedandsubjectedtoNarI
andSacIIrestriction.(c)Restrictionpatternanalysisofthreeclonesobtained,representingthethreeresolutionpathways.(d)Sequencingofclone36PCR
productrevealedtheexpectedmismatches.
4 NATURECOMMUNICATIONS|7:10937|DOI:10.1038/ncomms10937|www.nature.com/naturecommunications
ARTICLE
NATURECOMMUNICATIONS|DOI:10.1038/ncomms10937
the attI site are reactive and can be transferred independently, orientationbothresolutionpathwaysresultintheexcisionofthe
withoutaclearbiasonstrandspecificity.Thisiscontrarytowhat region between the sites and are hence productive, serving as a
isobservedfortheattI1(cid:2)attCreactionwhereonlythebsofattI1 control for the experiment. Our results show that integrase
isreactive17.Someclonesshowedpartialrestrictionpatternsand expressionincreasedsignificantlytheyieldofrecombinantclones
mixedplasmidpopulations(Fig.2d),suggestingthepossibilityof for sites in direct orientation (44-fold) but also for sites in
HJ resolution through second strand exchange. Nevertheless, it inverted orientation (72-fold) (Fig. 3b) (verified by PCR). This
could not be ruled out that this pattern was the consequence of demonstrates that the attI1(cid:2)attI1 HJ can be resolved through
twoindependentrecombinationreactionsinvolvingtwodifferent a second strand exchange, that is, through the classical
plasmids. Y-recombinase pathway. Furthermore, comparison of the
increases in recombination frequencies for both orientations
provesthattheclassicalresolutionpathwayisextremelyefficient.
The integrase resolves HJs through a second strand exchange. To avoid expression of dapA before recombination, one attI1
To further assess if both strands of the attI1 site could be site was modified 1bp upstream of the L box to include a stop
transferredduringthesamereaction,weusedamodifiedversion codon (attI ) (Fig. 3c). As a consequence, resolution through
STOP
of the dapA recombination reporter36. In this system, branch migration is not possible, suggesting again that host
recombination between attI1 sites leads to reconstitution of the factors were not necessary in the reaction. It implies too that
essential gene dapA, allowing recombinants to grow on media resolution of the HJ had taken place not further than 1bp
lacking diaminopimelic acid (DAP, the reticulating agent of upstreamoftheLbox,asconfirmedbythechangeoflocationof
peptidoglycan in E. coli). Comparison of the number of clones the stop codon after recombination.
growingwithandwithoutDAPyieldsarecombinationfrequency A similar experiment was performed using two attC sites,
for a given reaction. We produced two versions of this reporter attC and attC (Supplementary Fig. 3). When sites were
aadA7 ereA2
system, one with two attI1 sites in direct orientation, and the in direct orientation, we observed an increase in recombination
other with sites in inverted orientation. Both structures were frequenciesoffiveordersofmagnitudeduetothepresenceofthe
insertedectopicallyinsinglecopyinthechromosomeofanE.coli integrase. Contrary to our observations for the attI1(cid:2)attI1
MG1655 DdapA strain. The rationale of this experiment is that, reaction, no clear increase in recombination frequency was
when both sites are in inverted orientation, viable recombinants observedforattCsitesininvertedorientation.Sequencingofthe
can only be recovered if HJ resolution takes place through a few events (n¼8) in which the central structure was found
secondstrandexchangethatinvertstheregionbetweenbothsites. inverted showed that the site had not been recombined in a
In this orientation the replicative resolution pathway linearizes double-stranded form. Indeed, the resulting sites were largely
the chromosome and is hence abortive (Fig. 3a; Supplementary conserved,withthepresenceofmoietiesoftheoriginalsitesonly
Movies 1 and 2). On the other hand, when sites are in direct visible after the canonical 50-AAC-30 cleavage point. Although
a b
Direct orientation Inverted orientation
Direct orientation Inverted orientation
* dapA * dapA
NS
*
First strand exchange *
* dapA cy 10–5 * **
* dapA n
HJ resolution pathways ue
q
e
HJ replication Second strand exchange HJ replication Second strand exchange on fr 10–6
ati
DapA DapA DapA n
dapA dapA *dapA * dapA mbi
+ * + * eco 10–7
R
+ Original substrate
Productive Productive Abortive Productive
10–8
c
DR2 DR1 L R
Top5′ 3′attI1 attI1WT x attI1STOP + + – + + –
Bottom3′ 5′ WT attI1AAA x attI1STOP – – + – – +
BotTtoomp53′′ 35′′attI1STOP pBAD::intI1 – + + – + +
Top5′ 3′attI1
Bottom3′ x 5′ AAA
Figure3|TheintegrasecanresolveHJsthroughsecondstrandexchange.(a)Graphicrepresentationofthestructuralset-upofourexperimentinwhich
anattI1(cid:2)attI1reactionreconstitutesthedapAgene,withattI1sitesinadirectorientation(leftcolumn)andinaninvertedorientation(rightcolumn).
AsterisksindicateattIsitesmodifiedtoincludeaSTOPcodon(attI1 ).ThetopologyofHJsresultingfromthefirststrandexchangeisshownforboth
STOP
orientations.RecombinationproductsfollowingthetwopossibleHJresolutionpathwaysareshownbelow.(b)Recombinationfrequencyofallset-ups.
Columnsrepresentthemeananderrorbarsthes.d.(nZ3).SignificancetestingwasperformedwithKolmogorov–Smirnovtest(*Pvalueo0.05;**Pvalue
o0.01;NS,notsignificant).Forthedirectorientationset-up,strainsusedareB36,B82andC162(first,secondandthirdcolumn,respectively).Forthe
invertedorientationset-up,strainsusedareB37,B83andC163(fourth,fifthandsixthcolumn,respectively);theirrelevantgenotypesaredescribedinthe
tablebeneaththegraph.(c)SequenceoftheattIsitesusedinthisexperiment:attI1 (thecleavagepointonthebottomstrandismarkedwithanarrow),
WT
attI ,avariantnotencodinganORF,thatallowstoreducebackgroundnoiseintheexperimentwhileremainingrecombinogenic;andattI ,avariantin
STOP AAA
whichthebottomstrandisinactive.
NATURECOMMUNICATIONS|7:10937|DOI:10.1038/ncomms10937|www.nature.com/naturecommunications 5
ARTICLE
NATURECOMMUNICATIONS|DOI:10.1038/ncomms10937
theseinversioneventsrequireasecondcleavagethatcouldnotbe MBP-IntI1 protein, which is known to be fully active in vivo39,
elucidated, and therefore seem to be the consequence of andattI1fragmentsbearingamismatchbetweenthetopandthe
illegitimate recombination reactions (possibly between sites on bottom strand in the adenine that is transferred during the
different replicating chromosomes), they are strong proof that reaction (ts sequence 50-ATC-30, bs sequence 50-GTT-30)
attCsitesareexclusivelyrecombinedintheirstructuredformand (Fig. 5a). Such a mismatch eliminates the microhomology
confirms previous observations17. necessary for docking of the transferred strand, and should
Tofurtherprovethatthesecondstrandexchangeoccursandis abolishthetransferofthetopstrand.Whilethereactionbetween
mediatedbytheintegrase,weexploredthepossibilityofimpeding twoWTattI1sitesispoorlydetectedinvitro40(likelybecauseHJ
the bottom strand recombination. To do so, we mutated the intermediates are quickly resolved), the reaction between
cleavagepointintheRboxofoneattI1siteindirectandinverted L-box-mismatched substrates led to a significant accumulation
orientationfrom50-AAC-30 to50-AAA-30 (seeattI inFig.3c). of HJs (Fig. 5b), proving the role of the L-box cleavage in the
AAA
It has been observed that a C to A modification drops resolution of these structures, and the dispensability of host
recombination rates for attI1 with both attC and attI1 factors in the reaction. The complementary effect (that is, the
partners15,37. Our results show a complete abolition of accumulation of HJs due to an R-box mismatch) was not
recombination when both sites are in the inverted orientation observed, probably because mismatches in the R box led to a
(thatis,whenthesecondstrandexchangeisneededtoresolvethe significantdecreaseinintegrasebinding(SupplementaryFig.4a).
HJ), proving that the bs is no longer reactive. In contrast, The fact that binding is only affected by R-box mismatches is in
recombinationratesremainconstantwhenthesitesareindirect accordance with the sequential and cooperative binding of
orientation, proving that integrase-mediated recombination monomers, starting on the R box38. As a control, all reactions
through the ts is still possible for the attI site, and were performed in parallel with the catalytically inactive version
AAA
supporting the previously observed lack of clear bias in strand of the integrase, IntI1 , for which no transfer activity was
Y312F
selection. observed.
attC sites govern the directionality of the reaction. During
Mapping the ts crossover point using deep sequencing.
attC site folding, the inverted repeats of the dsDNA molecule
Recombinationbetweensitesondifferentrepliconsresultsinco-
hybridize to form the R and L boxes in the hairpin (Fig. 1d).
integrate molecules bearing two sites. Each of them is composed
These boxes are present in hairpins arising from the folding of
of two moieties of the initial sites, but when the latter are
boththetopandbottomstrands.Despiteaclearbiastowardsbs
identical, moieties are not easily identified and the crossover
recombination, top strands are also recombined, although
point of the reaction remains cryptic. Such is the case of the
at much lower frequencies. In all cases, the crossover in the
attI1(cid:2)attI1 reaction. Hence, to map the crossover point on the
attI(cid:2)attC reaction takes place in the R boxes of both sites.
top strand of attI1, we built a library of attI1 sites in which we
This means that top-strand recombination integrates cassettes
randomizedallbasepairsintheLbox,thetwoadjacentbasepairs
opposite to the Pc promoter and are likely non-functional.
withinthespacerregionandtheCinthecrossoverpointoftheR
Therefore,thesimilarfrequencyatwhichLandRboxesofattI1
box(Fig.4a).UsingN-containingprimers,10attI1 sitesetswere
N recombine in the attI1(cid:2)attI1 reaction is in clear opposition to
produced,eachbearingrandombasesinoneofthetenpositions.
that found for the attI1(cid:2)attC reaction. An event that has never
These sites were cloned in pir-dependent pSW plasmids and
been reported is the recombination of attC sites through the L
mixedtoformasmalllibrary(n¼40)ofpSW-attI1 plasmidsto
N box. We sought to detect such recombination events with attI1,
be used in our suicide conjugation assay. The rationale of this
sincetheywouldpresumablyoccurthroughtheLboxofattI1ts.
experiment is that the pattern of segregation of random bases
We reasoned that, since bottom strands of both sites recombine
between the two resulting sites in the cointegrate molecule will
veryefficientlythroughtheRboxes,itwouldbeeasiertoobserve
reveal the precise crossover point in the attI top strand (L box).
such events using the ts of attC. Using our conjugation assay,
Furthermore, since base composition of the L box can influence
we delivered the ts of attC and checked a total of 607
IntI1 binding, some attI mutants might be poorly recognized. aadA7
recombinants. Amongthem,anestimateof 251hadrecombined
A skew in the proportion of bases in a given position after
through the ts (the rest having recombined through the
recombination would underscore the importance of a specific
resynthesized complementary bs) and only 1 (0.16%) had
base in the binding process (Fig. 4b). The shortcoming of
recombined through the L box with the top strand of attI1
this experimental set-up is that, as seen before (Fig. 2), the
(SupplementaryFig.5).SincetheEHBT onthebottomstrand
reaction can take place exclusively through the bottom strand 23
impedes the nucleophilic attack on the L box, we assessed if the
(bs-exchange plus replication), without splitting the randomized
EHB complementary to T on the top strand plays a role in
L box. To overcome this, having confirmed the lack of bs 23
this bias, using a DT mutant site. We found three L-box
exchange for attI1 , we used a pSU-attI1 plasmid as 23
AAA AAA recombinantsoutof642coloniestested(0.47%)suggestingarole
recombination partner for the library, forcing recombination to
(although minor) for the complementary A base in avoiding
occur at the randomized L box. Deep sequencing of pSW-attI1
N L-boxrecombinationoftsattCsites.Thecrossoverpointinthese
before recombination confirmed the expected distribution of
recombinants could not be determined precisely due to an
7.5% mutated bases in the randomized positions (explained in
adenine stretch, but it was not coherent with the cleavage site
Methods)(Fig.4c).Sequencingoftheresultingleftandrightsites
determinedpreviouslyfortheattI1Lbox.Thisisprobablyrelated
in the cointegrate revealed a complementary distribution of
to conformational differences between attI and attC sites that
random bases among sites, allowing location of the crossover
force the crossover to shift. This need for a different crossover
pointintheLboxbetweenbothadeninesofthe50-AAC-30triplet
point could be itself at the basis of the low frequency of L-box
(Fig. 4c). It also allowed us to infer the importance of each
events in the attI(cid:2)attC reaction.
base for protein–DNA interactions, revealing lower sequence
stringency in the central 50-CT-30 base pairs of the L box
(50-CCCTAAA-30),inaccordancewithpreviousinterferencedata Discussion
for the central T base38. To confirm the crossover position, In this work we show that integron integrases can perform
we performed the attI1(cid:2)attI1 reaction in vitro using purified the second strand exchange in reactions involving exclusively
6 NATURECOMMUNICATIONS|7:10937|DOI:10.1038/ncomms10937|www.nature.com/naturecommunications
ARTICLE
NATURECOMMUNICATIONS|DOI:10.1038/ncomms10937
a c
DR2 DR1 L R
5′ 3′attI1 WT base Non-WT base
3′ 5′ N
b LibraNNry (n=4S0u)icidal conjugation of the library oriP15a LibraryBase frequency at each position1998877660505050500G C A G C A G G G C A T A G G C C C T A A A A C A A A G T T A C G
DR1 L R
N N N N N N N N N N
pSW-attI1N oriT pSU-attI1AAA attIN site
oriR6K
E.P Dciorolni oβr2163 InEt.I 1Rcoerlcie iMcpAiGe–n1t655 Right attI siteBase frequency at each position 1998877660505050500G C A G C A G G G C A G T C G C C C T A A A A C A A A G T T A CG
Deep-sequencing analysis of recombinants attI moiety
N
Right Main crossover point
attI moiety
N
N Left attI siteBase frequency at each position1998877660505050500G C A G C A G GG C A T A GG C C C T A A A A C A A A T T T A G G
Left
Location of the N depends Low sequence
on the crossover point stringency
Figure4|Mappingofthecrossoverpointonthetopstrand.(a)PositionsrandomizedintheattISTOPsitetoproducetheattINlibrary(n¼40).
Thetheoreticalfrequencyofbasesineveryrandomizedpositionisof92.5%WTbaseand2.5%ofeveryotherbase.(b)Diagramoftheexperimental
set-upandrecombinationreaction.RecombinationbetweenattI andattI takesplacethroughthetopstrand,splittingtherandomizedbases
N AAA
betweenmoietiesintheresultingsitesofthecointegrate(rightandleft).(c)GraphicalrepresentationofbasefrequencywithintheattI librarybefore
N
recombination,andtherightandleftsitesafterrecombination.Green,WTbase;Purple,non-WTbase.
a F F F F F b
2 2 2 2 2
31 31 31 31 31 Transfer activity
Y Y Y Y Y
w/o IntI1IntI1 w/o IntI1IntI1 w/o IntI1IntI1 w/o IntI1IntI1 w/o IntI1IntI1 25 *
20
HJ HJ 15
of
% 10
Sub * *
5 **
ND
L R L R L R L R L R 0
53′′ attI1WT 35′′ 53′′ attI1ttt 35′′ 53′′ attI1Mismatch L 35′′ 53′′ attI1Mismatch R 35′′35′′attI1Mismatch L+R53′′ attI1 WT attI1TTMTismatch MiLsmatcMih sRmatch L+R
Substrate
Figure5|MismatchesinthetopstrandabolishHJresolutioninvitro.InvitroactivityofIntI1ondifferentattIsubstrates.(a)GelStrandtransfertest
allowingtheobservationofHJintermediates.Resolutionofthesestructuresthroughasecondcleavageconvertsthembacktotheoriginalsubstrate(sub).
AdiagramoftheLandRboxesofthesiteisshownbelowitscognategel(therestofthesiteisidenticaltothewildtypeinallcases),wherered
basesrepresentchangeswithrespecttothewild-typesequence.Crossoverpointsinbothstrandsareindicatedwithblackarrows.(b)Transferactivity
quantifiedastheamountofsignalretainedinthefilter(HJs)overthetotalamountofsignal.Columnsrepresentthemeananderrorbarsthes.d.(n¼3).
TransferactivityforIntI1 wasundetectableinallassays.StatisticalanalysiswasperformedusingDunnett’stestwiththeactivityontheWTsiteasthe
Y312F
control(*significant(a¼0.05);**,significant(a¼0.01);ND,notdetected).
double-stranded sites. We found that integrases recombine the during the reaction. The crossover point in the top strand of
top strand of attI sites at frequencies matching those of the the attI site has been elucidated using a novel technique that
bottom strand and that there is no order in the use of strands involvesdeepsequencing.Thisapproachhasallowedovercoming
NATURECOMMUNICATIONS|7:10937|DOI:10.1038/ncomms10937|www.nature.com/naturecommunications 7
ARTICLE
NATURECOMMUNICATIONS|DOI:10.1038/ncomms10937
technicaldifficultiesfound withclassicalinvitromethodsdue to This work shows that despite millions of years of evolution
the low recombination activity observed. With deep sequencing towardsasystemspecializedinstructuredssDNArecombination,
becoming more accessible, this technique could be useful for and the acquisition of a 20-residue-long domain within the
others working with recombinases that show low activity or for catalytic core, integron integrases have conserved the activity of
which optimal conditions in vitro have not been elucidated. their canonical Y-recombinase ancestors. Many examples of
Our work reveals surprising aspects of the dynamics of DNA breakage and rejoining enzymes acting on dsDNA (like
attI(cid:2)attI reactions, with broad evolutionary implications that Y-recombinases) or ssDNA (like HUH endonucleases45) can be
arediscussedbelow.Still,otheraspectsoftheattI–IntIinteraction found, but this is the first report of a recombinase with full dual
remain unclear. attI sites are recognized for their sequence, as activity on both types of sites, as well as a substrate-dependent
proven by cognate IntI–attI recognition29. We (and others switch in recombination pathways. Integrases are capable of
previously38) have observed in our in vitro experiments a recognizingdistinctsubstrates(dsversussshairpins)indifferent
cooperative binding starting at the R box. It is also known that ways (sequence versus structure specific) and processing the
secondaryrecombinationsitescontainthe50-AAC-30signatureof reaction through different pathways (double strand exchange
theRboxofallattIsites37.GiventhattherestoftheRboxofthe versus replicative resolution). This duality has deep biological
site changes its sequence at every cassette integration event, one implications, since one pathway is semiconservative and
could presume that this triplet is fundamental for IntI binding. dependent on host machinery, while the other is not. Integrons
The almost complete loss of binding activity in our experiments could function, at least theoretically, using only the classical
whenamismatchislocatedinthistriplet(SupplementaryFig.4) recombinationpathwayondsDNAtodelivercassetteintegration
supportsthisidea.Nevertheless,thistripletisuniversaltoallattI and excision reactions. It is therefore tempting to speculate that
sites and would not allow integrase discrimination of cognate the force driving the evolution of integrons towards a mixed
sites. Also, the sequence of the attI1 bears a 50-AAC-30 triplet at ss/ds-DNA system derives from the benefits of semiconservative
the border of the L box, but the crossover point occurs between recombination. By producing recombined and non-recombined
different bases (A/A) compared with that of the bottom strand offspring, it allows testing the adaptive value of incoming DNA,
(A/C),suggestingthattopologyismoreimportantthansequence while minimizing the deleterious effect of capturing maladaptive
inthecleavageofthetopstrand.Therefore,therelativeinfluence genes. Furthermore, it also represents a mechanism for gene
of the universal crossover triplet and of the rest of the R box in duplication, as observed in the Vibrio cholerae superintegron10
thebindingremainsasubjectofinterestthatwouldneedfurther and possibly in some mobile integrons46.
studying. Fromanevolutionaryperspective,thefunctionalinnovationof
Thisworkshowsthattheavoidanceofthesecondcleavageon integrases towards the recombination of folded ss-attC sites,
attCsitesisnotduetothemerepresenceoftheintegrase-specific posedimportantmechanisticconstraintsintherecognitionofthe
I2 a-helix, but rather to its interaction with attC sites. The siteandtheresolutionofaHJtodeliverproductivereactions.The
extremelylowfrequenciesatwhichattC recombinesthrough solution in integrases comes in the form of the additional I2
aadA7
theLboxhighlightsthestrongconstraintsthatattCsitesimpose domainthatrecognizestheEHBsinattCsites.Thisimpedesthe
onthereaction.ItisnoteworthythatinnaturalconditionsL-box secondcleavage,allowingforthereplicativeresolutionoftheHJ.
reactions between attC and attI sites would place the attI site Interestingly, EHBs and the integrase I2 domain are a good
downstream of the cassette. Hence, such tight control on example of coevolution. These structures were probably the
directionality likely serves to ensure the correct functioning of first step in the evolutionary transition from the ancestral
theintegronbykeepingtheattIsitenexttothePcpromoterafter recombinationsystemtothess/ds-DNArecombinationplatform
consecutive cassette insertions. integrons are today. Nevertheless, the acquisition of the I2
From a functional perspective, the reaction between attI sites domain raises a chicken-egg paradox: how could the domain be
has an unclear adaptive meaning. It could be argued that in selected for in the absence of attC sites to provide a gain of
chromosomal integrons these reactions have deleterious effects function? And how could attC sites be selected for before the
on bacterial fitness, since they lead to the formation of recognitiondomainexisted?Suchaconundrumisfoundinmany
chromosomal dimers that have to be further resolved before evolutionary processes and can easily be understood with the
cell division. This is probably the reason for the 3–4 orders discovery of functional intermediary forms that allow to infer
of magnitude lower recombination frequency for attI(cid:2)attI smooth evolutionary transitions between states. We believe
reactionscomparedwiththoseinvolvingattCsites.Nevertheless, integron integrases are representatives of such a bi-functional
inthemulticopyworldof mobileintegrons, onecellcancontain state. In the light of our results, it now seems possible that an
more than one integron on the same or different replicons41–43. I2-like domain was acquired at a given time by the ancestor of
Given the possibility of having Pc’s of different strengths, an today’s integron integrases and, to some degree, permitted the
attI(cid:2)attI reaction can lead to the en bloc transfer of cassette recognition of new ss-substrates while retaining the original
arraysbetweenintegrons.Thiswouldaltertheexpressionlevelsof activity(orpartofit)ontheinitialsite.Fromthisstartingpoint,
multiple antimicrobial resistance genes at once, and could have the streamlining through natural selection, to reach what
importanteffectsonchemotherapyoutcomes.Itcouldbeargued integrons are today, seems a more straightforward process.
that the importance of such rearrangements is probably minor, To date, the only recombination reaction that shares some
due to the low recombination frequencies for attI(cid:2)attI in similarities with cassette recombination is the integration of
laboratory conditions. Nevertheless, one aspect of this reaction phage CTXf of Vibrio cholerae. Nevertheless, CTXf cannot be
that has been constantly overlooked is that, while attC sites considered a complete system since it does not encode an
are normally in their non-recombinogenic form, attI sites are integrasebutinsteadhijacksthehostXerCDrecombinases47.Asa
recombinogenicduringalmosttheentirecellcycle,withthebrief consequence, no specialization of the system towards the
exception of the passage of the replication fork. Despite this low recombination of ss-DNA is observed: XerCD lack structural
recombination frequency, it is plausible that attI(cid:2)attI reactions adaptations for the recognition and processing of ss-substrates
are more prevalent among mobile integrons than previously (I2-like domains), and the phage site stem does not have
envisaged, especially under transient, low-level (or even EHBs but rather mimics a ds-DNA molecule. Nevertheless,
stochastic) expression of the integrase. These expression pulses the resemblances between cassettes and CTXf integration, the
typically occur during horizontal gene transfer events36,44. close phylogenetic relation between integrases and XerCD
8 NATURECOMMUNICATIONS|7:10937|DOI:10.1038/ncomms10937|www.nature.com/naturecommunications
ARTICLE
NATURECOMMUNICATIONS|DOI:10.1038/ncomms10937
recombinases and the fact that both systems can coexist in the accordingtothemanufacturer’sinstructions.Agarose(1%)electrophoresisgels
same bacterial species, makes it tempting to speculate about a wereusedtovisualizeDNA.DNApurificationfromPCRproductsandgels,aswell
viral origin for integrons. Interestingly, the link to the SOS asplasmidextractions,wereperformedusingQiagenkits.Whennecessary,DNA
sequencewasverifiedusinganABIBigDyeTerminatorv.3.1sequencingkitandan
response found in all integrons is also present in many phages,
ABIPrism3100CapillaryGeneticAnalyzer(AppliedBiosystem).GATCand
including CTXf (ref. 48). To date, no chromosome-dimer EUROFINSsequencingserviceswerealsoused.
resolution recombinase of the XerCD family is known to
possess a domain similar to I2 and directionality is ensured Insilicowork. ProteinmodellingwasperformedusingPhyre2(sbg.bio.ic.ac.uk/
throughtheirinteractionwithotherhostproteinssuchasFtsK49. phyre2)andMacPymolontemplatesc2a3vA(VchIntIA)andc1ma7A(Cre).
Nevertheless, one could imagine that acquisition of an I2-like SequenceanalysiswasperformedusingGeneious7.1.
domain by a Xer-like ancestor could have provided the function
ofavoidingthesecondstrandexchangeduringphageintegration. InfluenceofhostmachineryintheattI(cid:2)attIreaction. TotestwhetherRecA,
This would protect the integrity of the chromosome while RecGandRuvABCimpacttheattI(cid:2)attIreaction,weperformedasuicide
conjugationassaybasedonthatofBiskrietal.35andpreviouslyimplementedin
favouring correct integration events, and would hence be of Bouvieretal.17.Briefly,theattIsiteprovidedbyconjugationiscarriedonasuicide
adaptivevaluetothecellandthephage.Wecannowimaginethat vectorfromtheR6K-basedpSWfamilythatisknowntousethePirproteinto
the chromosome dimer resolution activity of the hybrid initiateitsownreplication.ThisplasmidalsocontainsanRP4originoftransfer
recombinase would still be conserved. Phages, or any other (oriTRP4).Thedonorstrainb2163carriesthetransferfunctionsinits
integrative mobile element exploiting Xer (IMEX49) could have chromosome,requiresDAPtogrowinrichmediumandcansustainpSW
replicationthroughtheexpressionofachromosomallyintegratedpirgene.The
then acted as a vehicle for new adaptive genes, just as they MG1655recipientstrains,whichcontainthepBAD::intI1[ApR](expressingthe
currently do50,51, representing an ancestor of integron cassettes. IntI1integrase)andthepSU38D::attI1[SpR](carryingtheattI1site),lacksthepir
Previous work on evolutionary aspects of Y-recombinases geneandthereforecannotsustainreplicationoftheincomingattI-containingpSW
vector.Theonlywayforthisvectortobemaintainedintherecipientcellistoform
focused on the properties that allow related phage integrases to
acointegratebyattI(cid:2)attIrecombination.Weusedtheo10001strainasthedonor
target specific sequences in the bacterial genome and the andasrecipientstrainstheMG1655E.colioA197(control),anditsderivatives
possibility of adapting these proteins to recognize non-cognate o9987(recA-),o9988(recA-,recG-),o9989(recA-,ruvABC-)ando9990(recA-,
sites52–54. This work sheds light on the evolutionary potential of recG-,ruvABC-)(SupplementaryTables1and2).Therecombinationfrequencyis
the Y-recombinase protein fold towards evolutionary innovation calculatedastheratiooftransconjugantsexpressingthepSWmarker[CmR]
tothetotalnumberofrecipientclones[ApR,SpR].Recombinationfrequencies
(not adaptation). Our results suggest that Y-recombinase-
correspondtotheaverageofatleastthreeindependenttrials.TheattI(cid:2)attI
mediated reactions are a more dynamic process than previously co-integrateformationwascheckedbyPCRwiththeappropriateSWbegand
assumed. They underscore the robustness of the protein fold, a MFDprimers(oneightrandomlychosenclonesperexperiment).
characteristic that likely confers a high evolvability to the
members of this superfamily6. It is arguable that the duality in Mismatchedcovalentcirclespreparation.Constructionofthep8669andp8670
therecognitionandbindingofintegronintegrasestothesiteshas phagemidvectorstoproducemismatchedcovalentcircleswasperformedasinref.
21withsomemodifications.Briefly,weexchangedattCsitesinplasmidsp7770and
stronglybeenconservedbecausecassetteshavetobeintegratedin
p7771forattI1sites,givingrisetoplasmidsp8669andp8670(plasmids,carrying
the attI site. Nevertheless, the classical recombination pathway
theoriFDinbothorientationsensuringtheproductionofrespectivelythebottom
could have been lost through the specialization of integrases (oriFd1)andtopstrand(oriFd2)).SinglestrandDNAproductionwasperformed
towards the single strand exchange pathway, without affecting usingM13K07HelperPhageasinref.21.Bothphagemidvectors(p8669and
any of the known aspects of integron dynamics. Therefore, p8670)areintroducedbytransformationintoF0carryingpirstraincells(b2150,
preservationofthispathwayaftereonsofevolution11canbeseen o4446)torespectivelyobtaintheo8675ando8676strains.M13infectionand
purificationofssDNAwereperformedaccordingtothemanufacturer(M13K07
as the consequence of a lack of negative selection55, or as HelperPhageandQIAprepSpinM13kit).ThecomplementarysinglestrandDNA
reflectinganimportantbiologicalfunction.Thefactthattopand moleculeswereannealed,digestedbybothEcoRIandMfeIrestrictionenzymesto
bottom strand cleavage of attI sites occur at similar frequencies eliminateoriFd(whichdoesnotannealduetotheinvertedorientationineach
molecule),andself-ligated.
(Fig. 3b) suggests that top strand cleavage is not a promiscuous
activity of the integrase and favours the biological function
Non-replicativerecombinationassay.Thisassaywasdevelopedpreviouslyinthe
hypothesis. The flexibility of the integrase might not be
laboratory58.Wetransformwith1–2mgmismatchedcovalentcirclescontaining
completely unveiled yet, and functions dependent on some theattIsiteintoMG1655mutS215straincontainingthepBAD::intI1(p3938)and
aspects of this plasticity might remain unknown. The second thepSU38D::attI1(p929)plasmids(straino7994).Circlescannotreplicateinthis
strand exchange resolution pathway opens new avenues for geneticbackground(o7994).NotethatthisassaywasfirstenvisagedinamutS-,
recA-strainbuttransformationefficiencywastoolowduetoafilamenting(sick)
possible mechanisms of cassette genesis, a subject for which no
reliable model is yet available10. phenotypeinthedoublemutant.WehencestudiedtheinfluenceofrecAandfound
ittoberesponsibleforo0.5%ofallrecombinantsinourassay(seebelow
recombinationfrequencieswithandwithoutpBAD::intI1).Thisjustifiedtheuse
Methods ofarecAþ strainneglectingtheinfluenceofhomologousrecombinationin
therecombinationprocess.Competentcellswerepreparedinthepresenceof
Bacterialmediaandconditions.E.colistrainsweregrowninLuriaBertanibroth
0.2%arabinosetoallowintegraseexpression.Selectiononchloramphenicol
(LB)at37(cid:2)C.Growthat42(cid:2)Cwasusedtoinducetheexpressionandforcetheloss
(Cm)-containingplates(themismatchedcirclesmarker)yielded,almost
oftheplasmidbearinglambdaphageintegrase.Antibioticswereusedatthe
exclusively,attI(cid:2)attIrecombinationevents.TheattI(cid:2)attIcointegrateformation
followingconcentrations:carbenicillin(Ap),100mgml(cid:3)1,chloramphenicol(Cm),
wascheckedon101clonesbyPCRwithSeqattIandSeqNarprimers.ThesePCR
25mgml(cid:3)1,kanamycin(Km),25mgml(cid:3)1,spectinomycin(Sp),40mgml(cid:3)1,
productsincludethemismatch-containingregion,andwereanalysedfortheir
Thymidine(Thy)andDAPweresupplementedwhennecessarytoafinal
resistanceorsensitivitytobothSacIIandNarIrestrictionenzymes.
concentrationof0.3mM.GlucoseandL-arabinosewereaddedat10and Asacontrol,weperformedasimilarexperimentusinganMG1655mutS215
2mgml(cid:3)1finalconcentration,respectively.Chemicalswereobtainedfrom
straincontainingthepSU38D::attI1butlackingpBAD::intI1(strainoA266).
Sigma-Aldrich(France).
Toestablisharecombinationfrequency,wetransformedinparallelthesamemutS
strainlackingboththepSU38D::attI1andpBAD::intI1plasmidsbutcontaininga
Bacterialstrains,plasmidsandprimers.BacterialstrainsinthisstudyareE.coli Pir-expressingplasmid(pSB118::pir116,p1177)ensuringthereplicationofthe
DH5a,MG1655(laboratorycollection),&1,andb2163(ref.56).Detailsonstrains, incomingcircle(straino7120).Recombinationactivitycorrespondstotheratioof
plasmidsandprimersaredescribed,respectively,inSupplementaryTables1–3. CmRclonesobtainedinpir(cid:3) conditions(withandwithoutintegrase)tothose
obtainedinpirþ conditions.Notethattheefficiencyoftransformationofeach
strainwasdeterminedbeforehandandusedtoadjustthefinalratioandnormalize
DNAprocedures. StandardtechniqueswereusedforDNAmanipulationand theresults.Weobtainedarecombinationfrequencyof1.17(cid:2)10(cid:3)3inpresenceof
cloning57.RestrictionandDNA-modifyingenzymeswerepurchasedfromNew integrase,andof4.24(cid:2)10(cid:3)6intheabsenceofintegrase.
EnglandBiolabsandFermentas(ThermoScientific).PCRswereperformedwith Asasupplementarycontrol,weanalysedtheCmRclonesobtainedinpirþ
DreamtaqDNApolymerase,andPhusionpolymerase(ThermoScientific) context(SupplementaryFig.2).Afterpoolingof319clones,plasmidswere
NATURECOMMUNICATIONS|7:10937|DOI:10.1038/ncomms10937|www.nature.com/naturecommunications 9