Stability by multimer resolution of pJHCMW1 is due to the Tn1331 resolvase and not to the Escherichia coli Xer system

Multimer resolution, active partition, and post-segregational cell killing systems all participate in the stabilization of circular replicons in bacteria (Blakely et al., 1991 ; Hiraga, 1992 ; Jensen & Gerdes, 1995 ; Nordstrom & Austin, 1989 ; Sherratt et al., 1993 ; Summers, 1994 ; Wake & Errington, 1995 ). The formation of plasmid multimers by homologous recombination reduces the number of independently segregating plasmid molecules and increases the chance of producing plasmid-free segregants (Summers et al., 1993 ; Summers & Sherratt, 1984 ). Many plasmids therefore carry multimer-resolution systems to help ensure stable plasmid inheritance. The Escherichia coli Xer site-specific recombination system acts at recombination sites found in many naturally occurring multicopy plasmids. Xer recombination was first identified through its role in multimer resolution of plasmid ColE1, where it acts at a site called cer (Summers & Sherratt, 1984 ). Subsequently, the Xer system has been shown to act at sites on a number of other plasmids, including psi from pSC101 (Cornet et al., 1994 ). Xer site-specific recombination also occurs at the dif site in the terminus region of the E. coli chromosome, allowing proper segregation of chromosomes at cell division (Blakely et al., 1991 ; Colloms et al., 1990 ; Sherratt et al., 1995 ).

The recombination site dif consists of an 11 bp XerC-binding site and an 11 bp XerD-binding site flanking a 6 bp spacer region. Xer recombination at dif is catalysed by two host-encoded recombinases, XerC and XerD, which bind cooperatively to dif and carry out strand exchange (Blakely et al., 1991 ; Kuempel et al., 1991 ). Plasmid recombination sites such as cer and psi require approximately 180 bp of accessory sequences adjacent to a dif-like recombination core (Cornet et al., 1994 ; Stirling et al., 1988 ). In addition to the recombinase proteins, accessory proteins (PepA and ArgR for cer, PepA and ArcA for psi) are required for Xer recombination at plasmid sites (Colloms et al., 1998 , 1996 ). These accessory proteins bind accessory DNA sequences and ensure that recombination is exclusively intramolecular (Stirling et al., 1988 , 1989 ; Summers, 1989 ).

The plasmid pJHCMW1 was isolated from a Klebsiella pneumoniae clinical isolate (Tolmasky et al., 1986 ; Woloj et al., 1986 ). It includes a copy of the transposon Tn1331, which harbours four genes mediating resistance to several aminoglycosides and ß-lactams (Fig. 1) (Dery et al., 1997 ; Tolmasky, 1990 ; Tolmasky & Crosa, 1987 , 1993 ). Another region of pJHCMW1 includes the replication functions, an oriT, and a cer-like site (Fig. 1) (Dery et al., 1997 ; Tolmasky, 1990 ; Tolmasky & Crosa, 1987 , 1993 ). This plasmid replicates using the RNA-regulated general mechanism described for ColE1, p15A and other multicopy plasmids (Helinski et al., 1996 ; Polisky, 1988 ). In this paper we describe the analysis of stabilization by multimer resolution of pJHCMW1. Although there is a site in pJHCMW1 that has high homology to sites acted upon by the E. coli Xer system, it was found that monomerization and stable inheritance of pJHCMW1 are brought about predominantly by the Tn1331-encoded resolvase acting at the Tn1331 res site.

(21K): Fig. 1. Physical and genetic map of pJHCMW1 and its derivatives. At the top is a linearized map of pJHCMW1 indicating Tn1331 and the genes included (Tolmasky, 1990 ). Rep indicates the replication region and the white oval shows the location of pJHCMW1 mwr (cer-like in the figure). The oriT site is also indicated. Restriction endonuclease sites: B, BamHI; EI, EcoRI; EV, EcoRV; H, HincII; S, SacI. The transposon Tn1331 is represented by a white block, with its genes indicated as arrows pointing in the direction of transcription; aac stands for aac(6')-Ib. The thick arrow indicates the tnpR gene and the black oval the res site. Deletion derivatives and recombinant clones are shown below the pJHCMW1 map. In these derivatives, grey bars represent pJHCMW1 DNA, and dashed lines show the DNA portions deleted.

E. coli strains and plasmids.
These are described in Table 1.

Table 1. Strains and plasmids

Bacterial growth media and general DNA procedures.
Bacteria were routinely grown in LuriaBertani broth (L broth) (Sambrook et al., 1989 ). For solid medium, 1·5% agar was added. Transformation was carried out as described before (Cohen et al., 1972 ). Restriction endonuclease and ligase treatments were carried out as recommended by the suppliers. Plasmid DNA preparations were performed using the Wizard Plus kit (Promega). Site-directed mutagenesis of the tnpR gene was carried out using the QuikChange site-directed mutagenesis kit, following the recommendations of the supplier (Stratagene). The mutagenic oligonucleotides were AGCCAGCAGTCCCTCTGAATTCAGATCAGAGCG and CGCTCTGATCTGAATTCAGAGGGACTGCTGGCT. Nucleotide sequencing was performed at the DNA sequencing facility, Department of Biochemistry, University of Oxford, UK. DNA sequence analyses were carried out using the CLUSTAL W program of the Sequencing Analysis Software Package of the University of Wisconsin Genetic Computer Group (GCG) (Devereux et al., 1984 ).

In vivo resolution assays.
To prepare dimers, E. coli JC8679 was transformed with plasmid DNA. The transformed strains were cultured, and plasmid DNA was purified and electrophoresed in a 0·7% agarose gel. DNA of the correct size to be plasmid dimer was purified from agarose gels using the GeneClean kit (Bio101). Since dimers run at the same position as open circular monomer DNA, the isolated samples were used to transform the XerD-deficient E. coli DS9028. In this strain, dimers are not resolved by the Xer system, allowing the isolation of transformants that have obtained a plasmid dimer. Purified plasmid dimers were transformed into E. coli DS941 to determine the efficiency of Xer recombination. To determine the protein requirements for Xer recombination, plasmid dimers were transformed into the appropriate mutant derivatives.

In vitro recombination assays.
These assays were performed as described previously (Colloms et al., 1996 ). The reactions contained pSDC203 DNA and the indicated proteins in a buffer containing 50 mM Tris/HCl pH 8, 50 mM NaCl, 1·25 mM EDTA, 5 mM spermidine, 1 mM L-arginine, 10% (v/v) glycerol, and 25 µg bovine serum albumin ml^-1. The final volume of reaction was 20 µl. The reactions were carried out for 1 h at 37 °C and stopped by phenol extraction followed by ethanol precipitation. The products were resuspended in H₂O, treated with the indicated restriction endonuclease and analysed by agarose gel electrophoresis.

Assay of plasmid stability.
Assays of plasmid stability were carried out basically as described before (Summers & Sherratt, 1984 ). The strain being tested was cultured at 37 °C overnight into 5 ml L broth containing the appropriate antibiotics to ensure that there were no plasmid-free cells at the beginning of the test. The culture was then diluted 10^-6-fold and cultured at 37 °C overnight. This cycle was repeated an appropriate number of times, each time being 20 generations. To test for the number of plasmid-free cells, samples of each culture were diluted and spread onto L agar plates containing no antibiotics. The growing colonies were then tested for presence of plasmid by assaying their resistance. Experiments were done twice, and 100 colonies per generation were assayed.

Gel-mobility-shift assays.
These assays were performed as described previously (Blakely et al., 1993 ). The oligonucleotides used had the following structure (the putative binding regions are underlined): dif, 5'-GATCCTTGGTGCGCATAATGTATATTATGTTAAATGGTACCCTGCA-3' and 3'-CTAGGAACCACGCGTATTACATATA ATACAATTTACCATGGG-5'; pJHCMW1 mwr, 5'-GATC C G G CGGTGCACGCAACAGATGTTATGGTAAAT ACG-3' and 3'-GGGGGCCGCCACGTGCGTTGTCTACAATACCATTTATGCAGCT-5'.

Approximately 10 pmol oligonucleotide was end-labelledwith 50 µCi (1·85 MBq) [³²P]ATP and phage T4 polynucleotide kinase (5 units) in kinase buffer (50 mM Tris/HCl pH 7·5, 10 mM MgCl₂, 5 mM dithiothreitol, 0·1 mM spermidine) in a final volume of 20 µl. The labelled oligonucleotide was purified using a Nuctrap Probe Purification column (Stratagene) followed by ethanol precipitation. The radiolabelled oligonucleotide was dissolved in 15 µl H₂O and then made double-stranded by annealing with 50 pmol of the complementary oligonucleotide. The mixture was heated to 75 °C for a few minutes and then allowed to cool to room temperature overnight. The annealed double-stranded radiolabelled oligonucleotides were purified by electrophoresis on an 8% polyacrylamide gel in Tris/borate buffer (100 mM Tris pH 8, 100 mM boric acid, 2 mM EDTA) as described before (Blakely et al., 1993 ). The radiolabelled oligonucleotides were mixed with 0·1 mg poly(dI-dC) ml^-1 and the appropriate protein(s). The binding reaction was carried out for 10 min at 37 °C and immediately transferred to ice. The samples were analysed by electrophoresis in a polyacrylamide gel as described above. The radioactive complexes were detected by exposure to X-ray film.

Presence of a cer-like (mwr) site in pJHCMW1
Analysis of the nucleotide sequence of the 638 nucleotide SacIEcoRI fragment of pJHCMW1 (Fig. 1) revealed the presence of a cer-like region (Fig. 2). This cer-like region contains a recombination core with good matches to consensus XerC- and XerD-binding sites (Blakely et al., 1993 ; Colloms et al., 1996 ). By analogy to the ColE1-resolution site, we refer to this site as mwr (pJHCMW1 resolution). The pJHCMW1 mwr XerC-binding site is closest to the XerC-binding site from pSC101 psi (10/11 nucleotides identical) whereas the XerD-binding site is 100% identical to that of cer (Fig. 2). The pJHCMW1 mwr site has a 6 nucleotide spacer between the recombinase-binding sites, as is the case for the dif and psi sites, but unlike cer, which has an 8 bp spacer (Fig. 2). Adjacent to the presumptive XerC-binding site of pJHCMW1 mwr is a 182 bp sequence with a high degree of similarity to the accessory sequences of cer (112/182 nucleotides identical) (Fig. 2). This includes an 18 bp sequence that is a good match to the consensus Arg-box found in cer and other cer-like sites, as well as in the promoter regions of ArgR-regulated L-arginine biosynthetic operons (Fig. 2). These homologies suggest that pJHCMW1 contains a multimer-resolution site that has a high degree of sequence similarity to ColE1 cer.

(42K): Fig. 2. Comparison of the pJHCMW1 mwr, cer and psi regions including the XerC- and XerD- and ArgR-binding sites. The first row shows only the comparison of ColE1 and pJHCMW1 DNA because there was not significant homology to pSC101. The binding sites are shown as black boxes and the spacers as a grey box. On top of the sequences, the Arg box consensus sequence as described by Glansdorff (1996) and the most conserved nucleotides among XerC and XerD binding sites as described by Sherrat et al. (1995) are shown. The asterisks indicate nucleotides identical in cer, psi and pJHCMW1; the arrowheads indicate nucleotides identical in ColE1 cer and pJHCMW1. Dashes indicate gaps.

The pJHCMW1 mwr does not confer plasmid stability
To determine whether pJHCMW1 mwr can mediate multimer resolution, we generated deletion derivatives of pJHCMW1 (Fig. 1). Heritable stability of pJHCMW1 and its deletion derivatives was determined in the hyper-recombinogenic E. coli strain JC8679 (recBC sbcA). Due to the high level of intermolecular plasmid recombination, plasmids that do not possess efficient multimer resolution systems are not stably maintained in this strain (Summers & Sherratt, 1984 ). JC8679 is therefore ideally suited to detect deficiencies in plasmid-resolution systems. The removal of pJHCMW1 mwr from pJHCMW1 resulted in a plasmid (pJHCMW1ΔEcoRV; Fig. 1) that was as stable as the parental plasmid after 40 generations in JC8679 (Table 2). Conversely, removal of another region of this plasmid that included the tnpR gene and the res site of Tn1331 (pJHCMW1ΔSacI, Fig. 1) resulted in a substantially unstable plasmid. Removal of a fragment including both pJHCMW1 mwr and the tnpR gene and res site of Tn1331 (pJHCMW1ΔBamHI, Fig. 1) generated a plasmid that was also unstable (Table 2). These results suggest that pJHCMW1 may owe its stability in E. coli JC8679 to the presence of the Tn1331 resolvase and its recombination site. Summers & Sherratt (1984) have shown that a derivative of pACYC184 that includes the Tn1 res site was stabilized when resolvase was supplied in trans. An unexpected result was that pJHCMW1ΔBamHI, lacking both Tn1331 res/resolvase and pJHCMW1 mwr, was consistently more stable than pJHCMW1ΔSacI, which lacks only Tn1331 res/resolvase. The reason for this difference is not yet known. In any case, the results described in this section indicate that pJHCMW1 mwr does not substantially contribute to the stability of pJHCMW1, even in the absence of the Tn1331 resolution system. Another experiment to test the involvement of the tnpR gene product in the stability of pJHCMW1 consisted of substituting the GAT codon (D17) for a TGA (stop), generating a truncation of the resolvase at amino acid 17. This derivative, pJHCMW1-D17Stop, was very unstable when compared to the pJHCMW1 (Table 3).

Table 2. Stability of pJHCMW1 and deletion derivatives

Table 3. Stability of pJHCMW1 and the tnpR deficient derivative

Dimers containing pJHCMW1 mwr are inefficiently resolved by the Xer system in vivo
To analyse whether pJHCMW1 mwr is an efficient substrate for the E. coli Xer system, we generated dimers of the pJHCMW1 deletion derivatives as well as a number of recombinant plasmids containing various fragments of pJHCMW1. These plasmid dimers were tested for their ability to recombine in Xer⁺ and Xer^- E. coli strains. The recombinant clones used for these experiments are shown in Fig. 1. Recombinant plasmid pMETOX1 includes a fragment of pJHCMW1 containing pJHCMW1 mwr; pMETOX2 includes a fragment of pJHCMW1 harbouring tnpR and res (Table 2). Plasmid pMETOX3 contains the EcoRISacI fragment whose nucleotide sequence contains the complete pJHCMW1 mwr sequence. Plasmids pUC18 and pKS492, a pUC18 derivative containing a 280 bp HpaITaqI ColE1 cer fragment (Stirling et al., 1988 ) were used as negative and positive controls. Dimers were generated by transforming the plasmids into E. coli JC8679. For each plasmid, DNA was isolated and the dimeric form was purified from an agarose gel and introduced into the XerD-deficient E. coli strain DS9028. Analysis of E. coli DS9028 transformants showed that no colonies carrying dimers could be isolated for the wild-type pJHCMW1 and pMETOX2 (data not shown). This result can be explained by the fact that both of these plasmids retain an intact Tn1331 res/resolvase system, which can efficiently convert plasmid dimers to monomers in DS9028. Dimers could be isolated in DS9028 from all of the other pJHCMW1 derivatives. These dimers were introduced into Xer⁺ and Xer^- strains DS941 and DS9028 to determine whether they could be resolved by the Xer system. Dimers of pKS492, containing ColE1 cer, were completely resolved to monomers in an Xer-dependent fashion (Fig. 3a, compare lanes D and H), whereas dimers of pMETOX1, containing pJHCMW1 mwr, were resolved poorly by the Xer system (Fig. 3a, compare lanes B and F). This experiment was repeated ten times and in all cases at least 50% of the pJHCMW1 DNA isolated remained as dimer (data not shown). Lanes C and G in Fig. 3(a) show the results of transformation of E. coli DS941 and DS9028, respectively, with control dimers of pUC18. As expected, the pUC18 dimers were not resolved in either of the E. coli strains. In another experiment, pMETOX3 dimers were used to transform E. coli DS941. Dimers of pMETOX3 were poorly resolved in DS941 (Fig. 3b, lane C), whereas dimers of pKS492 were resolved efficiently (Fig. 3b, lane B) and pUC18 dimers were not resolved at all (Fig. 3b, lane A). Thus, although pJHCMW1 mwr is a poor substrate, it is recombined by the E. coli Xer system in vivo. However, the level of recombination is not enough to confer stability in our assays. Fig. 3(b) lane D shows plasmid DNA isolated from E. coli DS941 transformed with monomeric pMETOX3 as control.

(47K): Fig. 3. Resolution of dimers. (a) Dimers or monomers were introduced by transformation into E. coli DS941 (lanes AD) or E. coli DS9028 (lanes EH). The cells were cultured, and plasmid DNA was isolated and subjected to agarose gel electrophoresis. E. coli DS941 was transformed with: lane A, pMETOX1 (monomer); lane B, pMETOX1 (dimer); lane C, pUC18 (dimer); lane D, pKS492 (dimer). E. coli DS9028 was transformed with: lane E, pMETOX1 (monomer); lane F, pMETOX1 (dimer); lane G, pUC18 (dimer); lane H, pKS492 (dimer). To the left, the locations of dimers or monomers of pMETOX1 is shown (d1 and m1). To the right, the positions of dimers or monomers of pUC18 and pKS492, which are very close, are also shown (d2,3 and m2,3). (b) Dimers of pUC18 (lane A), pKS492 (lane B) and pMETOX3 (lane C), or monomers of pMETOX3 (lane D), were introduced by transformation into E. coli DS941. The transformant strains were cultured and plasmid DNA was isolated and subjected to agarose gel electrophoresis; d4 and m4 show the position of pMETOX3 dimers and monomers, respectively.

Requirements for Xer recombination at pJHCMW1 mwr
Xer recombination at ColE1 cer requires ArgR, PepA, XerC and XerD, whereas recombination at psi requires ArcA, PepA, XerC and XerD. To analyse the genetic requirements for Xer resolution of dimers containing pJHCMW1 mwr, E. coli strains deficient in the production of each of these five proteins were transformed with dimers of plasmids containing either ColE1 cer (pKS492) or pJHCMW1 mwr (pMETOX3). Monomerization of pMETOX3 dimers required XerC, ArgR and PepA but not ArcA (Fig. 4). In addition, the results presented in Fig. 3 indicate that XerD is required for resolution of pMETOX3 dimers. Thus recombination at pJHCMW1 mwr site has the same requirements as recombination at ColE1 cer.

(56K): Fig. 4. Requirements for resolution of dimers. Dimers of pMETOX3 (lanes CG) or pKS492 (lanes KO) were introduced by transformation into E. coli DS941 (lanes G and O) or mutant strains deficient in production of XerC (lanes C and K), ArgR (lanes D and L), PepA (lanes E and M) or ArcA (lanes F and N). Plasmid DNA was isolated and subjected to agarose gel electrophoresis. As controls, monomers (lanes A and I) or dimers (lanes B and J) of pMETOX3 and pKS492 were also electrophoresed. Lane H, DNA ladder: 8·1, 7·1, 6·1, 5·1, 4·1, 3·1, 2·0, 1·6 and 1·0 kb.

In vitro formation of Holliday junctions and binding to XerC and XerD
Resolution of plasmid multimers by Xer-mediated site-specific recombination at cer and psi involves a step where a pair of strands is exchanged to form a Holliday junction (Colloms et al., 1996 ; McCulloch et al., 1994 ). At psi, XerC exchanges one pair of strands to produce a Holliday junction; the other pair of strands are then exchanged by XerD to yield a fully recombinant product. Strand exchange at ColE1 cer stops after the first strand-exchange step, and the product contains a Holliday junction. To determine whether pJHCMW1 mwr was able to mediate formation of Holliday junctions, the plasmid pSDC203, containing two pJHCMW1 mwr sites in direct repeat, was incubated with XerC, XerD, ArgR and PepA (Fig. 5). As a control, a similar experiment was performed using pJB43, a plasmid containing directly repeated ColE1 cer sites (not shown). The products of pSDC203 incubation with Xer proteins were digested with NdeI and XhoI and subjected to agarose gel electrophoresis. These enzymes cleave once on either side of each recombination site. A Holliday-junction-containing product, when cleaved in this way, migrates between the linear and open circular forms of the plasmid substrate on an agarose gel (Colloms et al., 1996 ). Incubation of pSDC203 with ArgR, PepA, XerC and XerD yielded Holliday-junction-containing product (Fig. 5). It was of interest that Holliday junctions could be formed in vitro in the absence of ArgR. However, addition of this protein increased the efficiency of formation of Holliday junctions (compare lane B with F, and lane C with G in Fig. 5). These results indicate that pJHCMW1 mwr is a good substrate for Xer recombination in vitro.

(34K): Fig. 5. In vitro recombination at pJHCMW1 mwr. (a) Supercoiled pSDC203 DNA was incubated for 1 h at 37 °C with the proteins indicated and digested with NdeI and XhoI. Lane A, supercoiled DNA (sc). Reactions contained: 1 µM PepA, 0·3 µM ArgR, XerC and XerD (lane B); 0·1 µM PepA, 0·3 µM ArgR, XerC and XerD (lane C); 0·01 µM PepA, 0·3 µM ArgR, XerC and XerD (lane D); 0·3 µM ArgR, XerC and XerD (lane E); 1 µM PepA, XerC and XerD (lane F); 0·1 µM PepA, XerC and XerD (lane G); 0·01 µM PepA, XerC and XerD (lane H). Lane I, no proteins added; lane J, linearized plasmid (pSDC203 digested with NdeI). 1 and 2 represent unrecombined substrate fragment products of restriction endonuclease treatment. HJ, Holliday junction containing recombination product; ln, linear plasmid. (b) Restriction map of pSDC203. The arrowheads represent pJHCMW1 mwr.

The ability of pJHCMW1 mwr core region to bind the Xer site-specific recombinases was analysed using gel-retardation assays. Both XerC and XerD bound to dif separately to form single complexes, and together to form a XerCXerD double complex (Fig. 6, lanes AD). In contrast, the pJHCMW1 mwr core site formed a retarded complex with XerD but was not bound by XerC (Fig. 6, lanes EG). Furthermore, addition of XerC and XerD together did not result in the formation of a XerCXerD double complex (Fig. 6, lane H). This lack of efficient cooperative binding in vitro is in contrast to the efficient in vitro recombination reaction seen at pJHCMW1 mwr. It is possible that the high levels of PepA and ArgR used in the in vitro recombination assay bind and synapse the two recombination sites tightly enough to overcome the poor XerC binding to pJHCMW1 mwr.

(69K): Fig. 6. In vitro proteinDNA binding. Oligonucleotides containing the pJHCMW1 mwr core (lanes EH), or the dif site as control (lanes AD), were end-labelled and incubated with the following additions: lanes A and E, none; lanes B and F, XerC; lanes C and G, XerD; lanes D and H, XerC and XerD. The products of the binding reactions were separated by electrophoresis in an 8% polyacrylamide gel. The nature of the complexes for each signal is shown.

The objective of this work was to identify multimer resolution systems that contribute to the stability of pJHCMW1. The nucleotide sequence of pJHCMW1 includes a region that has high homology to cer; we refer to this region as pJHCMW1 mwr. The presence of this sequence led us to hypothesize that it might be a target for the Xer site-specific recombination system and that it could contribute to the heritable stability of pJHCMW1. However, experiments with pJHCMW1 deletion derivatives indicated that this region does not contribute significantly to plasmid stabilization by multimer resolution. Deletion of a fragment that includes the pJHCMW1 mwr site resulted in a plasmid that was just as stable as pJHCMW1 in a hyper-recombinogenic strain. In contrast, deletion of a DNA fragment containing the Tn1331 tnpR gene res site or introduction of a stop codon at the D17 position of this gene resulted in a plasmid that was unstable in this strain. Transformation of plasmid dimers containing various fragments of pJHCMW1 indicated that pJHCMW1 mwr is a poor substrate for Xer recombination in E. coli, whereas the Tn1331 res/resolvase system efficiently converted plasmid dimers to monomers. Close inspection of the pJHCMW1 mwr sequence revealed that the XerD-binding site is identical to that of mwr, whereas the XerC-binding site is closer to the equivalent site of psi. However, sequences adjacent to the XerC-binding site of pJHCMW1 mwr were highly similar to the accessory sequences of cer (110/182 bp) and contained a reasonable match to a consensus ArgR binding site (7 out of 8 matches to the most highly conserved Arg-box residues) (Glansdorff, 1996 ). This suggests that pJHCMW1 mwr is closer to ColE1 cer than it is to pSC101 psi. In agreement with this, the low level of recombination at pJHCMW1 mwr that occurred in vivo, like recombination at ColE1 cer, required ArgR, PepA, XerC and XerD but not ArcA.

In vitro assays demonstrated that a plasmid carrying two copies of pJHCMW1 mwr site formed Holliday junctions in the presence of PepA, ArgR, XerC and XerD. Although some Holliday junctions could be formed in vitro in the absence of ArgR, addition of this protein increased the efficiency of formation of Holliday junctions. Therefore this reaction had basically the same protein requirements as the in vivo recombination reaction, and was very efficient. Binding experiments showed that XerD bound to the pJHCMW1 mwr core sequence with an affinity similar to that shown for dif. In contrast, binding of XerC to the pJHCMW1 mwr core was very weak, in both the presence and absence of XerD. The low affinity of XerC for this site might account for the poor recombination seen at pJHCMW1 mwr in vivo. However, it is not clear why strand exchange at pJHCMW1 mwr in vitro was so efficient while recombination in vivo was so poor. Perhaps the relatively high concentration of proteins used for the in vitro recombination reaction compensates for the poor XerC-binding site. The suboptimal Arg-box in pJHCMW1 mwr might also contribute to its low activity in E. coli. The Arg-box in pJHCMW1 mwr has a 10 out of 18 nucleotide match to the Arg-box in cer and has only 7 out of 8 of matches to the highly conserved nucleotides in the Arg-box consensus (Glansdorff, 1996 ). Site-directed mutagenesis of key nucleotides in the ArgR and XerC recognition sites may help determine with more precision the causes for the deficient activity of pJHCMW1 mwr.

pJHCMW1 was first isolated from K. pneumoniae, and it may turn out that the pJHCMW1 mwr site has evolved to function in this bacterial species rather than in E. coli. Sequence differences between pJHCMW1 mwr and ColE1 cer may therefore reflect differences between the binding specificities of ArgR and the Xer recombinase proteins in K. pneumoniae and E. coli.

This work was supported by Public Health grants AI39738 and LA Basin Minority International Research Training 5 T37 TW00048-04 from the National Institutes of Health, and by the Wellcome Trust.