DNA repair by the cryptic endonuclease activity ofMu transposaseWonyoung Choi and Rasika M. Harshey1

Section of Molecular Genetics and Microbiology and Institute of Cellular and Molecular Biology, University of Texas at Austin, Austin, TX 78712

Edited by Charles M. Radding, Yale University School of Medicine, New Haven, CT 06510, and approved December 23, 2009 (received for reviewNovember 2, 2009)

Phage Mu transposes by two distinct pathways depending on thespecific stage of its life cycle. A common θ strand transfer inter-mediate is resolved differentially in the two pathways. During lyticgrowth, the θ intermediate is resolved by replication of Muinitiated within the flanking target DNA; during integration ofinfecting Mu, it is resolved without replication, by removal andrepair of DNA from a previous host that is still attached to the endsof the incoming Mu genome. We have discovered that the crypticendonuclease activity reported for the isolated C-terminal domainof the transposase MuA [Wu Z, Chaconas G (1995) A novel DNAbinding and nuclease activity in domain III of Mu transposase: Evi-dence for a catalytic region involved in donor cleavage. EMBO J14:3835–3843], which is not observed in the full-length proteinor in the assembled transpososome in vitro, is required in vivofor removal of the attached host DNA or “5′flap” after theinfecting Mu genome has integrated into the E. coli chromosome.Efficient flap removal also requires the host protein ClpX, which isknown to interact with the C-terminus of MuA to remodel thetranspososome for replication. We hypothesize that ClpX consti-tutes part of a highly regulated mechanism that unmasks the cryp-tic nuclease activity of MuA specifically in the repair pathway.

MuA transposase ∣ Mu DNA transposition ∣ ClpX ∣ retroviral integration ∣post-integration repair

Mu transposition has two outcomes, depending on the phaseof its life cycle (Fig. S1) (1). InfectingMu integrates without

replication (2–4). Every subsequent transposition event duringthe lytic growth that follows is accompanied by replication (5).The replicative pathway is well studied in vitro. These studieshave established that Mu transposes by a “cointegrate” mechan-ism where water-mediated single-stranded cleavages at Mu endsgenerate 3′OHs, which subsequently attack target DNA to createa θ strand transfer intermediate (Fig. 1) (6). (DNA strands thatparticipate in reaction chemistry are “transferred strands”whereas those that do not are “nontransferred strands.”) TheMu-target joint leaves 3′OHs on target ends, which are usedas primers for Mu replication (7, 8). These reactions take placewithin a highly stable transpososome, built frommultiple subunitsof the transposase MuA (9). The transition between transpositionand replication is enabled by disassembly of the transpososomeand assembly of the replisome on Mu ends. Disassembly requiresthe E. coli unfolding protein ClpX, which interacts with theC-terminal domain of MuA (10–12). MuA is exchanged with aseries of host proteins and finally with the replication restart com-plex that includes PriA (8, 13).

In contrast to the replicative pathway, the nonreplicativepathway is unexplored. The donor substrate in this pathway isthe infecting Mu genome, which is linear in the phage head,and includes host sequences flanking both ends of its original in-sertion site (Fig. S1). An injected phage protein N binds to theseends, protecting them from degradation and converting linearMuinto a noncovalently closed supercoiled circle (14–16). As ob-served in vitro with miniMu plasmids, a purified Mu-N proteincomplex also generates the θ strand transfer intermediate upontransposition into a plasmid target (17). In vivo evidence for a co-

integratemechanismofMu integration came from the observationthat the host DNA sequences flanking both ends of infecting Muare found integrated along with Mu in the chromosome (18).These sequences are removed fairly rapidly. Thus, integrationof infectingMuoccurs by a variation of the cointegratemechanismin which, instead of the “nick-join-replicate” pathway, Mu followsa “nick-join-repair” pathway (18) (Fig. 1). Removal of flankinghost DNA is only the first step in repair, which must be completedwith help of polymerase and ligase activities to fill the 5 bp gaps leftat the Mu-host junction by the transposase.

The present study was initiated to identify the function(s)responsible for the initial removal of 5′ flanking host DNA afterintegration of Mu. We refer to these sequences as “flaps.” Wereasoned that such a function was likely phage-encoded, becauseefficient integration is vital to phage development. A prime can-didate is the MuA transposase itself, because an isolated C-term-inal domain of MuA was found to have a strong endonucleaseactivity (19). This activity is not manifest in the full-lengthprotein, suggesting that it is normally masked. Other phage can-didates could be one or more of the many small ORFs in the earlyor “semiessential” region of the phage genome (20). Differentphenotypic effects have been traced to this region includingDNA end protection (gam/sot) and ligase (lig) activities(21, 22). We also considered host involvement. Whereas severalhost genes (in addition to those required for DNA replication)are known to specifically affect Mu replicative transposition(eg. clpX, gyrA, gyrB, and priA) (23–25), almost nothing is knownabout host functions that specifically affect nonreplicative trans-position. However, we were interested in the influence of ClpX,because the transpososome remains strongly bound toMu ends inits absence (8, 10–12, 26). Does the transpososome have to bedisassembled before the flaps can be removed?

We show that MuA mutants defective in the C-terminal endo-nuclease activity are defective in flap removal, whereas deletionof the semiessential region of Mu does not affect this process. Wefind that ClpX, previously thought to participate only in replica-tive transposition, is also necessary for efficient removal of theflap. We propose that ClpX participates in stimulating the nucle-ase activity of MuA in a highly regulated reaction that cleavesthe nontransferred strands of Mu, severing the flaps prior toor concomitant with disassembly of the transpososome.

ResultsEndonuclease, DNA-Binding and Transposition Activities of Mutants inthe Nonspecific DNA-Binding and Endonuclease (BAN) Region of MuA.Isolated domain III displays an activity designated as BAN (19).A basic amino acid cluster RRRQK at residues 575–579 in this

Author contributions: W.C. and R.M.H. designed research; W.C. performed research; W.C.contributed new reagents/analytic tools; W.C. and R.M.H. analyzed data; and R.M.H. wrotethe paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.1To whom correspondence should be addressed. E-mail: rasika@uts.cc.utexas.edu.

This article contains supporting information online at www.pnas.org/cgi/content/full/0912615107/DCSupplemental.

10014–10019 ∣ PNAS ∣ June 1, 2010 ∣ vol. 107 ∣ no. 22 www.pnas.org/cgi/doi/10.1073/pnas.0912615107

domain was shown to be important for both activities. To testwhether the BAN region is involved in flap removal in vivo,we generated three variants in this region (Fig. 2A). The RRRQKresidues were altered to give RQRQQ, LQLQQ, and KKKQR.The first two mutants were reported to be partially or completelydefective, respectively, in both DNA-binding and endonucleaseactivities of isolated domain III (19). In the KKKQR mutant,the basic character of the residues was retained but their identitywas changed, with the thought that DNA-binding activity mightbe retained but nuclease activity might be lost, thus separatingthe two activities. The endonuclease activities of wild-type andmutant domain IIIs are shown in Fig. 2B Left. Complete degra-dation of the input pUC plasmid was seen with wild-type domainIII. The KKKQR mutant did not degrade the DNA like thewild-type domain, but showed significant nicking activity. TheLQLQQ and RQRQQ mutants showed negligible nicking.

DNA-binding was measured by a gel-shift assay. The KKKQRmutant was proficient in DNA-binding whereas the LQLQQand RQRQQ mutants were defective (19) (Fig. 2B Right). Thegel-shift seen with the KKKQR mutant was not eliminated evenin the presence of SDS, and required addition of Proteinase K(see Methods), suggesting that the strong DNA-binding by thismutant may prevent multiple rounds of endonucleolytic cleavage,accounting for the difference in the nuclease activities of the wild-type versus the KKKQR mutant (Fig. 2B Left).

Full-length MuAðLQLQQÞ575–579 [referred to as MuABAN byWu and Chaconas (19)], was reported to be defective in transpo-sosome assembly, but this defect could be suppressed by MuBprotein to give approximately 20% of wild-type strand trans-fer activity. The strand transfer activity of full-length MuA con-taining all three sets of mutations in domain III was examined inthe presence of MuB, along with the ability of the transposo-

somes to be disassembled by ClpX (Fig. 2C). The Type II trans-pososome, formed by strand transfer of the donor miniMuplasmid into a target pUC plasmid, was assembled efficiently bytwo of the three mutants; MuAðLQLQQÞ575–579 showedpoor Type II activity as reported earlier. Addition of ClpX(Type II-2 reaction) gave rise to a series of θ DNA topoisomers,diagnostic of disassembly. Efficiency of disassembly ofMuAðKKKQRÞ575–579 and MuAðRQRQQÞ575–579 complexeswas similar to that of wild-type. These mutants should thereforesupport wild-type transposition activities in vivo. If the nucleasedomain participates in flap removal after integration, the mutantswould be expected to differ only in the efficiency of this reaction.

Flap Removal In Vivo Is Dependent on the Endonuclease Activity ofMuA. All three sets of domain III mutations were moved intothe A gene on a Mu prophage inserted in lacZ (18). We hadshown earlier that approximately 4% of the induced phage fromthis strain package the original prophage insertion, i.e., have lacZsequences linked to both ends.Wild-type andmutantMuAphageswere used to infect a Δlac host; cell lysis profiles with two ofthe three mutant phages were similar to wild-type (Fig. S2A), inkeeping with the expectation from in vitro results that thesemutants would not be affected for those MuA functions neededfor replication (Fig. 2C). Total cellular DNA was isolated atvarious times up to the onset of lysis at 50 min; infection withMuAðLQLQQÞ575–579 phagewas followed longer because the cellsdid not lyse. TheDNAwas subjected to pulse field gel electrophor-esis (PFGE) to separate integratedMu in chromosomal fragmentsfrom free Mu (18). Genomic DNA bands were excised and sub-jected to PCR with primers specific to Mu (within the B gene)or to the lacZ flap reporter (spanning the MuR-lacZ junction)as described in ref. 18 and Fig. 3A). Using real-time PCR, Auet al. (18) had established that the majority of infecting phagesintegrate with the flap still attached.

In cells infected with phage carrying wild-type MuA, Mu se-quences were detected in the chromosome by 15 min, concomi-tant with detection of flap sequences (Fig. 3B). Whereas the Musignal increased at 30 and 50 min, the flap signal showed the op-posite trend and was undetectable by 50 min. A similar flap DNAprofile was observed during infection with MuAðKKKQRÞ575–579phage; recall that this MuA variant showed significant nicking ac-tivity in its domain III in vitro (Fig. 2B). In infections with thenuclease-deficient MuAðRQRQQÞ575–579 phage, flap sequenceswere observed up to 50 min. MuAðLQLQQÞ575–579 phage showeda 15 min delay in integration compared to wild-type, but theflap signal in these nuclease-deficient phage was still detectableuntil 120 min.

To better assess the longevity of the flaps in mutant phages thatshowed delayed flap removal, we repeated these infections usinga himA host, which does not support Mu replication. The himAlocus encodes one of the two subunits of the integration hostfactor, involved in the regulation of early Mu gene expression.The efficiency of Mu integration is decreased by a factor of 2to 3 in this mutant (18, 27). As shown earlier, flap sequences weremaximally detected at 30 min in the himA host (compared to15 min in a wild-type host), but the kinetics of flap removalwas similar to that observed in a wild-type host when the infectingphages were wild-type (Fig. 3C). However, phages with eitherMuAðRQRQQÞ575–579 or MuAðLQLQQÞ575–579 failed to signifi-cantly process the flap even up to 120 min.With both mutants, theflap signal was consistently higher at all time points compared towild-type MuA (Fig. 3D).

In summary, flap removal in vivo is strongly correlated withthe endonuclease activity observed for domain III of MuAin vitro. The longevity of the flaps in the nuclease-deficient phageinfections suggests that the strand transfer joint is protected.

Fig. 1. Two outcomes of Mu DNA transposition. A common θ intermediate isresolved differentially during replicative vs nonreplicative Mu transposition.

Choi and Harshey PNAS ∣ June 1, 2010 ∣ vol. 107 ∣ no. 22 ∣ 10015

BIOCH

EMISTR

Y

Flap Removal Is Independent of the Semiessential Region of Mu.A single large early transcript that includes A and B also includesthe semiessential or SE region with several ORFs of unknownfunction (20, 28) (Fig. S3A). Although called semiessentialbecause insertions and deletions in this region reduce but donot abolish phage growth, it is possible that these genes areimportant, but bacterial genes can substitute for them. To testif the SE region is involved in flap removal, we deleted this regionin a prophage and used the resulting phage in infection as before.The kinetics of flap removal was indistinguishable between wild-type and ΔSE Mu phage infections (Fig. S3B). We conclude thatthe SE region is not involved in flap removal.

Flap Removal Is Dependent on ClpX but Independent of PriA. The twopathways of Mu transposition are thought to diverge at the TypeII-2 stage, when ClpX disassembles the transpososome in ahighly choreographed transition to replication (Fig. 4A).Whereas Mu replication is blocked, Mu lysogens are reportedto be readily recovered in a clpX mutant (23). This finding wasinterpreted to mean that ClpX functions only in the replicativepathway, and is not required for the nonreplicative pathway.

In the absence of ClpX, the assembled transpososome is notexpected to disassemble rapidly. To test the fate of the flap in sucha situation, we infected a clpX mutant host with wild-type Mu.

Because replicative growth is blocked in this mutant, we followedflap sequences up to 120 min as we had with himA infections. Thekinetics and efficiency of Mu integration were similar in both thewild-type and clpX mutant strains (the time points start at 30 minfor clpX because this strain was followed longer). Unlike the wild-type host where the flap sequences diminish significantly at30 min, the clpX mutant retained these until 60 min, decreasingthereafter, but still detectable by 120 min (Fig. 4B and C). Thesedata suggest that ClpX is required for efficient flap removal.Growth of the infected clpX host stalled until 60 min, but resumedsubsequently (Fig. S2B), suggesting that other host mechanismseventually disassemble the transpososome to repair the integrant.Infection of the clpX host with MuAðRQRQQÞ575–579 phageshowed a flap removal profile similar to that of infection of thismutant with wild-type phage (Fig. S4).

Although we would have liked to have examined flap removalat each of the known stages of the Mu replication pathway (seeFig. 4A), the identity of MRFα, which functions after ClpX, is notknown, and IF2, the next protein in the replication pathway, isessential for growth. However, PriA, which is required for the laststep prior to assembling the replisome at Mu ends (29), can bedeleted, although this mutant is defective in double-strand breakrepair and grows slowly (30). When Mu infection was performedin the priA mutant, flap removal was similar to that seen duringinfection of a wild-type host (Fig. 4B and C).

We conclude that in addition to the MuA nuclease activity,ClpX is also required for efficient flap removal, but PriA is not.

DiscussionWe report a physiologically relevant function for the cryptic en-donuclease activity discovered within the C-terminal domain ofthe transposase MuA (19). We call this activity “flap endonu-clease” for removing the 5′ flaps attached to the integratedMu genome upon Mu infection. Whereas this is the first reportof a DNA repair function encoded within a transposase, we notethat other proteins involved in repair of DNA forks and flapsalso have endonuclease domains (31). We propose a model fora spatial location of the nuclease domain within the transposo-some, and discuss how this activity might be regulated in vivo.We also discuss the relevance of these findings to the cellularrepair of transposition intermediates.

Model for Endonuclease Activity in the Context of the Transpososome.MuA is organized in three domains (9, 32) (Fig. 2A). Domain Ibinds the att and enhancer sites (Fig. S3A), domain II has thecatalytic DDE residues for transposition, and domain III hasthe BAN region as well as a regulatory region at its C-terminusthat interacts with both MuB and ClpX (Fig. 2A). The Mu trans-pososome is built from six MuA subunits, whose assembly at theMu ends is directed by an ordered interaction between att andenhancer sites (33, 34). The two MuA subunits bound at the Land R termini are responsible for catalysis, each acting in transto carry out successive cleavage and strand transfer throughthe DDE active sites (35–37). A specific function has not beenassigned to the other subunits in the transpososome.

In a transpososome in which the ends have not yet undergonecleavage, an 8–10 bp region around the cleavage sites is melted(38, 39), likely promoting engagement of the transferred strandsin the DDE active sites in preparation for single-strand cleavage(40). Based on the inefficient assembly of MuAðLQLQQÞ575–579,as well as the inefficient assembly of other domain IIIα mutants(41), Wu and Chaconas (19) had speculated that the BAN regionbinds to the Mu-host junction to stabilize the transpososome.We extend this proposal to suggest that this region specificallyengages the nontransferred strands (Fig. 5). Thus, the two MuDNA strands are held in two different active sites. However,the nuclease active site is more fastidious and is not “on” untilcertain conditions are met, one of these being interaction of

Fig. 2. Endonuclease, DNA-binding, and transposition assays. (A) Domainorganization of MuA showing the various functions mapped to individualdomains. Residues targeted for mutagenesis in the BAN domain are boxedand in gray letters; the multiple changes made in three mutants in this regionare indicated underneath. (B) Endonuclease and DNA-binding activities ofisolated domain III (His-tagged MuA residues 575–663) carrying indicatedmutations in residues 575–579. We surmise that addition of SDStriggered the nuclease, accounting for presence of DNA in the shiftedcomplex in WT DNA-binding (no-SDS) lane. OC and SC, open circular andsupercoiled forms of pUC19. (C) Type II (strand transfer) and Type II-2(ClpX-mediated disassembly) reactions using full-length MuA and its BANvariants. D, donor pSP104; T, target pUC19; θ, disassembled θ topoisomers.

10016 ∣ www.pnas.org/cgi/doi/10.1073/pnas.0912615107 Choi and Harshey

domain III with ClpX (Fig. 4). This interaction is not a sufficientcondition, because flap removal is not observed in vitro despitethe presence of ClpX (Fig. 2C). When conditions that activate thenuclease are met (see below), we propose that a specific phospho-diester bond in the nontransferred strand undergoes endonucleo-lytic cleavage, releasing the flap. The released flaps are thendegraded by cellular nucleases, monitored in our assays as disap-pearance of the flap DNA signal. The longevity of the flap in boththe MuA BAN domain mutants (Fig. 3) and in the clpX host(Fig. 4), shows that the Mu-host joint is not accessible to hostnucleases and its efficient removal is a highly regulated process.

To Repair or Replicate.What determines the alternate biochemicalfates of a common θ intermediate in the same cellular milieu? It isapparent from our studies that ClpX is required for both replica-tion and repair, so the two pathways must diverge soon after ClpXaction, prior to recruitment of PriA. ClpX alone is clearly notsufficient for flap resection in vitro, so there must be additionalplayers that promote MuA nuclease activation. The phage tailprotein N, which is injected along with phage DNA (16), and pro-tects the ends from degradation (14, 15) (Fig. S1), has long beentouted as playing a decisive role. In the absence of N, when MuDNA is injected in Salmonella by phage P22, Mu transposition islethal unless the strain is Recþ (42). This finding was explainedby the suggestion that the N-less Mu cannot repair and thereforeproceeds to replicate, thereby linearing the chromosome (seeFig. 1), which would then require recombination between the

two copies of Mu to recover lysogens. A biochemical activityfor the N protein remains to be demonstrated in vitro. Anothercandidate promoting repair might be the enhancer DNA seg-ment, which was observed in vitro to remain associated withthe transpososome even after completion of strand transfer(33). This finding is curious because the enhancer plays an earlyrole in assembly and is not required for the subsequent chemistryof transposition (43). The enhancer is part of the operator regioncontrolling the Mu lysis/lysogeny decision (Fig. S3A). Immedi-ately after integration of infecting Mu, the early region must at-tract RNA polymerase for transcription from a set of divergentpromoters within the enhancer. Interaction of the transcriptioncomplex with the transpososome could alter it in some mannerthat promotes ClpX stimulation of the nuclease domains in a sub-set of the MuA subunits. Finally, like replication factors MRF-αand IF-2 that were identified only in biochemical assays (Fig. 4A),there are likely as yet unidentified host proteins that function inthe repair pathway.

Cellular Repair of Stalled Mu Intermediates. The longevity of theflaps in nuclease-deficient, ClpX disassembly-proficient mutantsof MuA (Figs. 2 and 3) suggests that the Mu-host joints areprotected in vivo even after disassembly of MuA. Very likely,MuA is exchanged for host factors specific to the repair pathway,as it is for those specific to the replication pathway (Fig. 4A). Inthe absence of Mu-specific flap removal, cellular functions muststep in, as evidenced by isolation of lysogens after infection with

Fig. 3. Flap removal from integrated Mu following infection with MuA BAN domain mutants in vivo. (A) Experimental strategy for monitoring flap removal.(B) Wild-type host BU1384 was infected with phages carrying wild-type or indicated mutants in theMuA BAN region. Mu (B gene) or flap sequences (MuR-lacZ)were amplified from isolated genomic DNA at indicated times after infection. 50 ng of DNA was used in all reactions with 30 cycles for detecting Mu and 40cycles for detecting flaps. C is a control where genomic DNA from uninfected cells was mixed with Mu DNA and isolated by PFGE, to gauge the extent ofcontamination of the genomic DNA band with free Mu in subsequent PCRs. DNA loading was standardized to a chromosomal marker tsr. N, no templatecontrol, M, inputMu. (C) As in B, except infection was in a himA host (BU1382). Sixty nanograms DNA (30 cycles) was used for detectingMu and 75 ng (40 cycles)for detecting the flap. The results in each panel were verified in at least three independent repeats. (D) The flap DNA bands in panel C were quantified asdescribed in Methods and normalized to the signal from a chromosomal gene marker tsr.

Choi and Harshey PNAS ∣ June 1, 2010 ∣ vol. 107 ∣ no. 22 ∣ 10017

BIOCH

EMISTR

Y

MuAðLQLQQÞ575–579 mutant phage, which had correctly re-paired the Mu insertions (Table S1). We imagine that in the ab-sence of the Mu endonuclease, the flap endonculease of DNApolymerase I might substitute. In the absence of ClpX, whenMu θ repair is also stalled, one expects that the undisassembledtranspososome would additionally block passage of the replica-tion machinery, leading to cessation of growth or cell death.In keeping with this expectation, the infected clpX mutant ceasesgrowth initially (Fig. S2B). However, it recovers after 60 min sug-gesting that cellular proteins eventually disassemble the transpo-sosome and repair the block. Lysogens recovered from infectionof a clpX mutant had also repaired the Mu insertions correctly(Table S1). Thus, cellular repair of Mu integrants is robust. Thismay explain why flap removal was not more delayed in a clpXmutant infected with wild-type phage compared to a clpXmutantinfected with MuAðRQRQQÞ575–579 phage (Fig. S4).

We had expected that failure of repair after Mu integrationwould trigger a checkpoint that would block replication, becausethe consequences of replicating a linear donor integrant would beto linearize the chromosome. However, this appears not to be thecase as evidenced by the normal lysis and phage titres obtainedupon infection with MuAðRQRQQÞ575–579 phage (Fig. S2A),which are defective in flap removal (Fig. 3). We interpret thisto mean that there is a narrow window of opportunity for com-pletion of repair, in the absence of which replication can proceed.The free linear ends so generated would be protected by the Nprotein so that replicative amplification of Mu is unaffected.

Cellular Repair of Other Mu-Like Transposition Events. The DNAjoints created by Mu transposition resemble those made by retro-viral DNA, Tn7, and some IS elements (44, 45). The Mu flap isunusual in being several hundred to several thousand base pairslong; in other elements the flap is only a few nucleotides long. It isnot known whether dedicated enzymes are involved in flapremoval in these elements. It is assumed that the gap left inthe target after strand transfer is repaired by polymerases and

ligases. In retroviruses, a strong link between integration andcellular nonhomologous end-joining mechanisms involved inDNA repair has been observed (46). These proteins are normallyinvolved in double-strand DNA break repair, but are alsorecruited to single-strand breaks, stalled replication forks, orbulky lesions. The exact mechanism by which retroviral integra-tion intermediates are repaired remains to be demonstrated.

Similarities abound between retroviral integrases and theMuAtransposase (45). The structures of their DDE domains are super-imposable, as are their reaction mechanisms. Interestingly,the basic amino acid motif found in the BAN region of MuAis also found in the C-terminal domain of retroviral integrasesthat exhibit nonspecific DNA-binding (47). A peptide containingan RRKAK sequence derived from HIV integrase exhibitedDNA-binding but not nuclease activity (48). It would be interest-

Fig. 4. Flap removal is delayed in a clpX but not a priA mutant. (A) Posttransposition events leading to Mu replication. The transpososome has six MuAsubunits, four of which are tightly associated and two more loosely held. ClpX disassembles the complex, exchanging MuA with a series of protein factorsin preparation for replication. The Mu-host joints after transpososome disassembly by ClpX are drawn according to ref. 13. [Reproduced with permission fromref. 13 (Copyright 2007, Molecular Microbiology).] See text for details. (B) Wild-type Mu was used to infect wild-type, clpX or priA mutant hosts. Reactionconditions as in Fig. 3B. (C) The flap DNA bands in indicated infections were analyzed as in Fig. 3D.

Fig. 5. Model for domain III contribution to cleavage of nontransferredstrands. Two catalytic subunits positioned at Mu termini, act in trans to carryout successive cleavage and strand transfer. The transferred strand isengaged in the DDE active site (domain II), whereas the nontransferredstrand is engaged in the BAN active site (domain III). Domain IIIs could becontributed by noncatalytic subunits (19), indicated by unconnected hingeregions. Arrowheads, specific cleavage by BAN endonuclease. Mu (ThinLines), target (Dotted Lines), and flap (Thick Lines).

10018 ∣ www.pnas.org/cgi/doi/10.1073/pnas.0912615107 Choi and Harshey

ing to see if the flap removal function is unique to the Mu trans-posase, or if retroviral integrases also cleave their own flap, anduse host proteins only for gap repair.

MethodsStrains, Plasmids, Phage, Genomic DNA, and Protein Purification. See SI Text andTables S2 and S3.

Nuclease and DNA-Binding Activity of Domain III. 0.3–3 nmole of domain IIIproteins were incubated with supercoiled plasmid pUC19 (5–15 μg∕ml;2.6–8 nmole) in 20 μl of 25 mM Hepes-KOH (pH 7.6), 140 mM NaCl, and10 mMMgCl2 at 37 °C for 1 h. Nuclease reactions were stopped by incubationwith 0.5% SDS and 100 μg∕ml of Proteinase K (Sigma Aldrich) for 1 h at 50 °C,and electrophoresed on a 0.7% agarose gel. Proteinase K addition wasnecessary because when the reactions were stopped with SDS alone, theKKKQR mutant appeared to still be bound to DNA. DNA-binding reactionswere loaded directly on a 1% agarose gel.

Transposition and Transpososome Disassembly Assays. Type II strand transferreactions contains supercoiled miniMu donor plasmid pSP104 (26 μg∕ml),target plasmid pUC19 (23 μg∕ml), E. coli HU protein (10 μg∕ml), and MuAand its variants (25 μg∕ml), MuB (0.3 μg), and ATP (2 mM), 20 mM Hepes-KOH (pH 7.6), 10 mMMgCl2, and 140 mM NaCl in 20 μl. Reactions wereincubated for 30 min at 30 °C. For Type II-2 reactions, 5 mM ATP and145 μg∕ml of His6-ClpX were added to Type II reactions at 30 °C for 1 h.Reaction products were electrophoresed on 1% agarose gels and viewedby staining with ethidium bromide.

Detection of Integrated Mu and Flap Sequences. See SI Text.

ACKNOWLEDGMENTS. We thank Sooin Jang for sequencing most of the Muinsertions listed in Table S1, and James Sawitzke for strain SIMD30. This workwas supported by National Institutes of Health Grant GM 33247 and in partby the Robert Welch Foundation Grant F-1351.

1. Symonds N, Toussaint A, Van de Putte P, Howe MM (1987) Phage Mu (Cold SpringHarbor Laboratory Press, Plainview, NY).

2. Akroyd JE, Symonds N (1983) Evidence for a conservative pathway of transposition ofbacteriophage Mu. Nature 303:84–86.

3. Harshey RM (1984) Transposition without duplication of infecting bacteriophage MuDNA. Nature 311:580–581.

4. Liebart JC, Ghelardini P, Paolozzi L (1982) Conservative integration of bacteriophageMu DNA into pBR322 plasmid. Proc Natl Acad Sci USA 79:4362–4366.

5. Chaconas G, Harshey RM, Sarvetnick N, Bukhari AI (1981) Predominant end products ofprophage Mu DNA transposition during the lytic cycle are replicon fusions. J Mol Biol150:341–359.

6. Mizuuchi K (1992) Transpositional recombination: Mechanistic insights from studies ofMu and other elements. Annu Rev Biochem 61:1011–1051.

7. Mizuuchi K (1984) Mechanism of transposition of bacteriophage Mu: Polarity of thestrand transfer reaction at the initiation of transposition. Cell 39:395–404.

8. Nakai H, Doseeva V, Jones JM (2001) Handoff from recombinase to replisome: Insightsfrom transposition. Proc Natl Acad Sci USA 98:8247–8254.

9. Chaconas G, Harshey RM (2002) Transposition of phage Mu DNA. Mobile DNA II, edCraig NL (ASM Press, Washington DC), pp 384–402.

10. Levchenko I, Luo L, Baker TA (1995) Disassembly of theMu transposase tetramer by theClpx chaperone. Genes Dev 9:2399–2408.

11. Nakai H, Kruklitis R (1995) Disassembly of the bacteriophage Mu transposase for theinitiation of Mu DNA replication. J Biol Chem 270:19591–19598.

12. Abdelhakim AH, Oakes EC, Sauer RT, Baker TA (2008) Unique contacts direct high-priority recognition of the tetrameric Mu transposase-DNA complex by the AAAþunfoldase ClpX. Mol Cell 30:39–50.

13. North SH, Kirtland SE, Nakai H (2007) Translation factor IF2 at the interface of trans-position and replication by the PriA-PriC pathway. Mol Microbiol 66:1566–1578.

14. Harshey RM, Bukhari AI (1983) Infecting bacteriophage Mu DNA forms a circularDNA-protein complex. J Mol Biol 167:427–441.

15. Puspurs AH, Trun NJ, Reeve JN (1983) Bacteriophage Mu DNA circularizes followinginfection of Escherichia coli. EMBO J 2:345–352.

16. Gloor G, Chaconas G (1986) The bacteriophage Mu N gene encodes the 64-kDa virionprotein which is injected with, and circularizes, infecting Mu DNA. J Biol Chem261:16682–16688.

17. Gloor GB (1988) Characterization of the integrative precursor protein-DNA complex ofbacteriophage Mu. Ph.D thesis (University of Western Ontario, London, Ontario,Canada).

18. Au TK, Agrawal P, Harshey RM (2006) Chromosomal integration mechanism ofinfecting Mu virion DNA. J Bacteriol 188:1829–1834.

19. Wu Z, Chaconas G (1995) A novel DNA binding and nuclease activity in domain III ofMu transposase: Evidence for a catalytic region involved in donor cleavage. EMBO J14:3835–3843.

20. Morgan GJ, Hatfull GF, Casjens S, Hendrix RW (2002) Bacteriophage Mu genomesequence: Analysis and comparison with Mu-like prophages, Haemophilus,Neisseria,and Deinococcus. J Mol Biol 317:337–359.

21. Paolozzi L, Symonds N (1987) The SE region. Phage Mu, ed Symonds N (Cold SpringHarbor Laboratory Press, Plainview, NY), pp 53–62.

22. d’Adda di Fagagna F, Weller GR, Doherty AJ, Jackson SP (2003) The Gam protein ofbacteriophage Mu is an orthologue of eukaryotic Ku. EMBO Rep 4:47–52.

23. Mhammedi-Alaoui A, Pato M, Gama MJ, Toussaint A (1994) A new component ofbacteriophage Mu replicative transposition machinery: The Escherichia coli ClpXprotein. Mol Microbiol 11:1109–1116.

24. Sokolsky TD, Baker TA (2003) DNA gyrase requirements distinguish the alternatepathways of Mu transposition. Mol Microbiol 47:397–409.

25. North SH, Nakai H (2005) Host factors that promote transpososome disassembly andthe PriA-PriC pathway for restart primosome assembly. Mol Microbiol 56:1601–1616.

26. Surette MG, Buch SJ, Chaconas G (1987) Transpososomes: Stable protein-DNAcomplexes involved in the in vitro transposition of bacteriophage Mu DNA. Cell49:253–262.

27. Chaconas G, et al. (1984) Transposition of bacteriophage Mu DNA: Expression of the Aand B proteins from lambda pL and analysis of infecting Mu DNA. Cold Spring HarbSymp Quant Biol 49:279–284.

28. Goosen N, van de Putte P (1987) Regulation of transcription. Phage Mu, ed Symonds N(Cold Spring Harbor Laboratory Press, Plainview, NY), pp 41–52.

29. Jones JM, Nakai H (1999) Duplex opening by primosome protein PriA for replisomeassembly on a recombination intermediate. J Mol Biol 289:503–516.

30. Kogoma T, Cadwell GW, Barnard KG, Asai T (1996) The DNA replication primingprotein, PriA, is required for homologous recombination and double strand breakrepair. J Bacteriol 178:1258–1264.

31. Van Duyne GD (2005) Bending and cutting forks and flaps. Structure 13:1092–1093.32. Nakayama C, Teplow DB, Harshey RM (1987) Structural domains in phage Mu trans-

posase: Identification of the site-specific DNA-binding domain. Proc Natl Acad Sci USA84:1809–1813.

33. Pathania S, JayaramM, Harshey RM (2003) A unique right end-enhancer complex pre-cedes synapsis of Mu ends: The enhancer is sequestered within the transpososomethroughout transposition. EMBO J 22:3725–3736.

34. Yin Z, Suzuki A, Lou Z, Jayaram M, Harshey RM (2007) Interactions of phageMu enhancer and termini that specify the assembly of a topologically unique inter-wrapped transpososome. J Mol Biol 372:382–396.

35. Namgoong SY, Harshey RM (1998) The same two monomers within a MuA tetramerprovide the DDE domains for the strand cleavage and strand transfer steps oftransposition. EMBO J 17:3775–3785.

36. Savilahti H, Mizuuchi K (1996) Mu transpositional recombination—Donor DNAcleavage and strand transfer in trans by the Mu transposase. Cell 85:271–280.

37. Aldaz H, Schuster E, Baker TA (1996) The interwoven architecture of the Mu transpo-sase couples DNA synapsis to catalysis. Cell 85:257–269.

38. Wang Z, Namgoong S-Y, Zhang X, Harshey RM (1996) Kinetic and structural probing ofthe pre-cleavage synaptic complex (type 0) formed during phage Mu transposition:Action of metal ions and reagents specific to single stranded DNA. J Biol Chem271:9619–9626.

39. Kobryn K, Watson MA, Allison RG, Chaconas G (2002) The Mu three-site synapse:A strained assembly platform in which delivery of the L1 transposase binding sitetriggers catalytic commitment. Mol Cell 10:659–669.

40. Lee I, Harshey RM (2003) The conserved CA/TG motif at Mu termini: T specifies stabletranspososome assembly. J Mol Biol 330:261–275.

41. Naigamwalla DZ, Coros CJ, Wu Z, Chaconas G (1998) Mutations in domain IIIα of theMu transposase: Evidence suggesting an active site component which interacts withthe Mu-host junction. J Mol Biol 282:265–274.

42. Sonti RV, Keating DH, Roth JR (1993) Lethal transposition ofMud phages in Rec- strainsof Salmonella typhimurium. Genetics 133:17–28.

43. SuretteMG, Chaconas G (1992) TheMu transpositional enhancer can function in trans:Requirement of the enhancer for synapsis but not strand cleavage. Cell 68:1101–1108.

44. Hindmarsh P, Leis J (1999) Retroviral DNA integration. Microbiol Mol Biol Rev63:836–843.

45. Haren L, Ton-Hoang B, Chandler M (1999) Integrating DNA: Transposases and retro-viral integrases. Annu Rev Microbiol 53:245–281.

46. Smith JA, Daniel R (2006) Following the path of the virus: The exploitation of host DNArepair mechanisms by retroviruses. ACS Chem Biol 1:217–226.

47. Andrake MD, Skalka AM (1996) Retroviral integrase, putting the pieces together.J Biol Chem 271:19633–19636.

48. Puras Lutzke RA, Vink C, Plasterk RH (1994) Characterization of the minimal DNA-binding domain of the HIV integrase protein. Nucleic Acids Res 22:4125–4131.

Choi and Harshey PNAS ∣ June 1, 2010 ∣ vol. 107 ∣ no. 22 ∣ 10019

BIOCH

EMISTR

Y