Portal de Periódicos da CAPES

Integron cassette insertion: a recombination process involving a folded single strand substrate

2005; Springer Nature; Volume: 24; Issue: 24 Linguagem: Inglês

10.1038/sj.emboj.7600898

1460-2075

Marie Bouvier, Gaëlle Demarre, Didier Mazel,

RNA and protein synthesis mechanisms

Article8 December 2005free access Integron cassette insertion: a recombination process involving a folded single strand substrate Marie Bouvier Marie Bouvier Unité Postulante Plasticité du Génome Bactérien, CNRS URA 2171, Institut Pasteur, Paris, France Search for more papers by this author Gaëlle Demarre Gaëlle Demarre Unité Postulante Plasticité du Génome Bactérien, CNRS URA 2171, Institut Pasteur, Paris, France Search for more papers by this author Didier Mazel Corresponding Author Didier Mazel Unité Postulante Plasticité du Génome Bactérien, CNRS URA 2171, Institut Pasteur, Paris, France Search for more papers by this author Marie Bouvier Marie Bouvier Unité Postulante Plasticité du Génome Bactérien, CNRS URA 2171, Institut Pasteur, Paris, France Search for more papers by this author Gaëlle Demarre Gaëlle Demarre Unité Postulante Plasticité du Génome Bactérien, CNRS URA 2171, Institut Pasteur, Paris, France Search for more papers by this author Didier Mazel Corresponding Author Didier Mazel Unité Postulante Plasticité du Génome Bactérien, CNRS URA 2171, Institut Pasteur, Paris, France Search for more papers by this author Author Information Marie Bouvier1,‡, Gaëlle Demarre1,‡ and Didier Mazel 1 1Unité Postulante Plasticité du Génome Bactérien, CNRS URA 2171, Institut Pasteur, Paris, France ‡These authors contributed equally to this work *Corresponding author. Unité Postulante Plasticité du Génome Bactérien, CNRS URA 2171, Institut Pasteur, 25 rue du Dr Roux, 75724, Paris, France. Tel.: +33 1 4061 3284; Fax: +33 1 4568 8834; E-mail: [email protected] The EMBO Journal (2005)24:4356-4367https://doi.org/10.1038/sj.emboj.7600898 PDFDownload PDF of article text and main figures. ToolsAdd to favoritesDownload CitationsTrack CitationsPermissions ShareFacebookTwitterLinked InMendeleyWechatReddit Figures & Info Integrons play a major role in the dissemination of antibiotic resistance genes among Gram-negative pathogens. Integron gene cassettes form circular intermediates carrying a recombination site, attC, and insert into an integron platform at a second site, attI, in a reaction catalyzed by an integron-specific integrase IntI. The IntI1 integron integrase preferentially binds to the ‘bottom strand’ of single-stranded attC. We have addressed the insertion mechanism in vivo using a recombination assay exploiting plasmid conjugation to exclusively deliver either the top or bottom strand of different integrase recombination substrates. Recombination of a single-stranded attC site with an attI site was 1000-fold higher for one strand than for the other. Conversely, following conjugative transfer of either attI strand, recombination with attC is highly unfavorable. These results and those obtained using mutations within a putative attC stem-and-loop strongly support a novel integron cassette insertion model in which the single bottom attC strand adopts a folded structure generating a double strand recombination site. Thus, recombination would insert a single strand cassette, which must be subsequently processed. Introduction Integron cassettes are small DNA units that carry open reading frames generally without promoters. They integrate into an integron platform, consisting of a site-specific recombinase, an associated primary recombination target called the attI site and two appropriately orientated (divergent) promoters, one driving an integrase gene, intI, and the other driving expression of the cassette-associated gene. The integron is the generic name for the integron platform–cassette ensemble. Integrons are key players in the capture and dissemination of antibiotic resistance genes among Gram-negative bacteria (see Hall and Collis, 1998). Their importance has recently been underlined by the discovery of large integrons in the chromosomes of a wide range of bacterial species (Rowe-Magnus et al, 2001). These superintegrons (SI) contain arrays of hundreds of genes for various adaptive functions. The corresponding recombinases, IntI integrases, belong to the phage λ integrase family of tyrosine (Y) recombinases (see Azaro and Landy, 2002). The IntI integrases mediate recombination between their specific attI site and a second type of recombination site carried by a gene cassette, called the attC site (or 59-base element), and which is formed following circularization of the integron cassette. IntI integrases can also catalyze recombination between two attC sites. Although related to λ int, several lines of evidence imply that Int1-mediated recombination may be quite different from that of phage λ. A unique feature of the integron recombination system is the structure of the attI and attC recombination sites. These differ significantly from the canonical Y-recombinase core sites, which are composed of a pair of highly conserved 9- to 13-bp inverted binding sites separated by a 6- to 8-bp central spacer region (see Figure 1A). One of the putative IntI binding sites, within the core of attI, is extremely degenerate and the spacer region differs widely from that of the partner attC sites (see Figure 1A and B). IntI1 recombinase binds to four regions of double-stranded (ds) attI in vitro. Two correspond to the core repeats and two to direct repeats located upstream of the core (Figure 1A) (Collis et al, 1998; Gravel et al, 1998a). The role of the two direct repeats of the attI1 site for the recombination reaction is still unclear (Hansson et al, 1997). The structure of attC is more complex. It consists of two potential core sites, R″–L″ and L′–R′ (also called 1L–2L and 2R–1R (Stokes et al, 1997)), separated by a region that is variable in sequence and length (Figure 1C). A number of these have been demonstrated to be efficiently recombined by IntI1 (Collis et al, 2001; Biskri et al, 2005). While recombination occurs at L′-R′, directed mutagenesis showed that R″-L″ is also essential. All attC sites exhibit extensive potential cruciform structures (Hall et al, 1991; Stokes et al, 1997; Rowe-Magnus et al, 2003). Purified IntI1 binds specifically to the ‘bottom’ strand (bs) of single-stranded (ss) attC site, attCaadA1, DNA but not to a ds attCaadA1 site (Francia et al, 1999). This seminal observation was confirmed, and several key elements that act as recognition determinants for in vitro IntI1 binding were identified in the attCaadA1 sequence. Some appear to play important roles in the potential secondary structure of the attC site (Johansson et al, 2004). Figure 1.Integron recombination sites. (A) Sequence of the ds attI1 site. (B) Sequence of the ds attCaadA7 site. (C) Multiple sequence alignment of the attC sites bs studied in this work. (D) Proposed secondary structure for the attCaadA7 bs. The inverted repeats L, L′ and L″, R, R′ and R″ are indicated with black arrow; the asterisk (*) shows the position of the protruding G present in L″ relative to L′. The attI1 direct repeats bound by InI1 are indicated by horizontal lines with an empty arrowhead (Collis et al, 1998; Gravel et al, 1998a). The putative IntI1 binding domains, as defined by Stokes et al (1997), are marked with gray boxes. Vertical arrows indicate crossover position. The secondary structure was determined using the MFOLD (Walter et al, 1994) online interface at the Pasteur Institute. Download figure Download PowerPoint Integron cassettes are thought to move using an excised circular intermediate (Collis and Hall, 1992). These would have the capacity to form extensive secondary structures if produced as a single strand. For most cassettes, self-pairing on the same single strand can be extended up to the R′ and R″ sequences, which usually show a stretch of 9–11 consecutive complementary nucleotides (Figure 1; Hansson et al, 1997; Rowe-Magnus et al, 2003). Such a self-paired stem could be seen as an almost canonical core site consisting of the L″–L′ duplex and an unpaired central region followed by an R″–R′ duplex (Figure 1D). In the present study, we have addressed the mechanism of integron cassette transfer. We have extended previous observations in vitro (Francia et al, 1999; Johansson et al, 2004) demonstrating specific binding of IntI to the bs of attCaadA1. We show that IntI1 has a similar single strand preference for two additional and structurally distinct attC sites. This demonstrates that strand choice is a general phenomenon and is not associated specifically with attCaadA1. More importantly, we also present evidence strongly suggesting that integration occurs via a single strand intermediate and that a specific single strand of the cassette (that which is bound by IntI) is used. This conclusion is based on recombination frequencies obtained following delivery of one or other single strand by conjugation to a suitable recipient Escherichia coli strain carrying the integron platform and expressing the appropriate integrase. While the attC sites recombine in single strand form, our results suggest that attI must be present in a double strand configuration. However, although the attC recombination intermediate may be single stranded, recombination appears to occur using a ds attC region generated by the secondary structure within the single strand cassette. Thus, while mutations disrupting the potential pairing of non-conserved positions in a putative stem-and-loop structure of the attC bs decreased the recombination frequency, restoration of the complementarity by mutation of the partner sequence restored a high frequency of recombination. We propose an unusual recombination model to explain the insertion of integron cassettes at the attI site. In this model, a first strand exchange occurs using the attC bs folded into a stem-and-loop structure to generate a Holliday junction (HJ), which is then resolved by replication of the recipient replicon. Results IntI1 in vitro binding properties for single- and double-stranded forms of the attI1 site and two attC sites To determine whether bs-specific binding of IntI1 was specific to attCaadA1 (Francia et al, 1999) or is a general feature of attC sites, we tested two additional unrelated sites: attCaadA7 site, which differs only in two positions from the attCaadA1 site, and VCR2/1, the attC site from a Vibrio cholerae SI cassette, which is larger and unrelated to these two sites (Figure 1C). We also repeated IntI1 binding experiments (Francia et al, 1999) using attI1 (68 bp). The attCaadA7 site was carried on a 76 bp DNA fragment and the VCR2/1 on a 149 bp fragment. We used an MBP-IntI1 fusion protein in our in vitro binding experiments, as previous studies had shown that addition of an MBP tag did not disturb IntI1 function in vitro or in vivo (Gravel et al, 1998a, 1998b). As previously observed (Francia et al, 1999), we found that 48 pmol of IntI1 specifically retarded 0.5 pmol of ds attI1 site, but not the corresponding top strands (ts) or bs (Figure 2A and B). In the case of attCaadA7 ds, there are traces of retarded complex visible in Figure 2A. This might be explained by sufficient instability of this 76 bp duplex to leave a fraction of non-paired ss material, which could be bound by IntI1. These complexes likely correspond to attCaadA7 bs–IntI1 complexes, as nuclease S1 treatment led to their elimination (not shown). It is noteworthy that incubation with the larger (149 bp and likely more stable) VCR2/1 ds did not lead to such complexes. Under the same conditions, IntI1 did not alter the mobility of either of the ds attCaadA7 or VCR2/1 sites (Figure 2B). Conversely, we observed specific retardation when 0.5 pmol of either attCaadA7 bs or VCR2/1 bs was incubated with 4.8 pmol of IntI1, while incubation with the ts of either of these attC sites did not lead to any retardation (Figure 2A). Figure 2.Gel retardation of ss or ds attI1, attCaadA7 and VCR2/1 by IntI1. (A) Single strand substrates. A 4.8 pmol portion of IntI1 was incubated with 0.5 pmol of ssDNA containing the ts or the bs of attI1, attCaadA7 or VCR2/1. Lanes 1–4 show the attI1 ts or bs binding study; lanes 5–8 correspond to the attCaadA7 ts or attCaadA7 bs binding study; the last four lanes (9–12) show the VCR2/1 ts or VCR2/1 bs binding study. (B) Double strand substrates. Lanes 1, 2 and 3 correspond to incubation of 0, 24 or 48 pmol of IntI1 with ds attI1, respectively; lanes 4, 5 and 6 correspond to incubation of 0, 24 or 48 pmol of IntI1 with ds attCaadA7, respectively; lanes 7, 8 and 9 correspond to incubation of 0, 24 or 48 pmol of IntI1 with ds VCR2/1, respectively. Download figure Download PowerPoint Recombination of attC sites after conjugative transfer To assess whether an ss structure could be the substrate for recombination in vivo, we used a recombination assay that we developed to compare attC × attI site recombination, which mimics the cassette integration process (Biskri et al, 2005). This assay used conjugation to deliver one of the recombination substrates into a recipient cell expressing the IntI1 integrase and carrying a second recombination target on a pSU38 plasmid derivative (see Figure 3). Conjugative transfer of plasmids occurs by transfer of a single DNA strand (rather than duplex DNA) from donor to recipient. In addition, the orientation of the oriT sequence determines which of the two strands is transferred. The integron recombination site provided by conjugation was carried on an R6K-derived plasmid of the pSW family. Replication of these plasmids relies on the Π protein, provided by a pir gene inserted in the donor genome (Demarre et al, 2005). Transfer functions are also provided by the appropriate plasmid genes inserted into the donor chromosome. Following conjugation, re-circularization of the single transferred strand is catalyzed by the conjugative relaxase enzyme (Pansegrau et al, 1993; Pansegrau and Lanka, 1996). Complete ss transfer and re-circularization precede the complete second strand synthesis. Since the recipient does not supply the Π protein, the transferred plasmid is unable to replicate (Figure 3). This procedure has been called suicide conjugation. Insertion of attC in one orientation or the other in a given pSW derivative would lead to transfer of either attC ts or attC bs. If recombination uses a strand-specific ss substrate, a difference in the recombination rate measured after transfer of either attC ts or attC bs would be expected. On the other hand, if recombination involves a ds substrate, and thus requires second strand synthesis to be effective, no difference in the recombination rate is expected. Figure 3.Schematic representation of the conjugation–recombination assay used for the integron cassette integration reaction. Briefly, the donor cell expresses the Π protein, encoded by pir and required for pAttC replication. This strain also provides the transfer functions necessary for its conjugation. The recipient is devoid of a pir gene and therefore cannot sustain pAttC replication. The recipient also contains a plasmid carrying the attI1 site (pSU38-attI1) and expresses IntI1 (symbolized by green ovals). Core site sequences in the attC and attI1 sites are represented as empty boxes, and correspond to those of Figure 1; red and pink ovals indicate the oriT; de novo synthesized strands are shown in blue. The relaxosome, which cleaves and pumps DNA into the recipient, is shown in yellow, and the donor and recipient cell walls and membranes are shown as gray vertical lines. The donor is represented with a pale yellow background. Download figure Download PowerPoint In a first set of experiments, we compared the recombination of the VCR2/1 bs and VCR2/1 ts after transfer using plasmids pVCR-B and pVCR-T (Table II), which transfer respectively the VCR2/1 bs and VCR2/1 ts in the recipient. pVCR-B and pVCR-T are identical except for the VCR2/1 fragments, which are carried in opposite orientations. We established that both plasmids were transferred at similar rates (2 × 10−1) using strain UB5201-Pi (a UB5201 derivative able to sustain pSW replication) as a recipient. We also determined their IntI1-mediated recombination frequencies with the target attI1 site carried on the compatible plasmid, pSU38-attI1, in the same pSW replication permissive context. This was about 1–5 × 10−2 (Figure 4). We then tested their recombination frequencies following suicide conjugation to a UB5201 recipient (i.e. without Π) carrying the target pSU38-attI1 plasmid and expressing IntI1. An overall rate of 5.5 × 10−3 was obtained for pVCR-B while pVCR-T recombined at a rate of 1 × 10−6. In order to establish that this difference was not due to an unknown contextual difference between plasmids, we inverted the oriT orientation in the two plasmids, leading to plasmids pVCR-TINV (starting from the pVCR-T) and pVCR-BINV (starting from the pVCR-B) (Table II). Again, these two plasmids were transferred at similar rates, 2 × 10−1, yet a 2 × 104-fold higher recombination rate was still obtained when the transferred strand contained the VCR2/1 bs (pVCR-TINV; Figure 4). When recombination of the same constructs was tested in UB5201-Pi, a [pir+] host permitting replication, the ratio of recombination frequencies of the two plasmids obtained with the UB5201 recipient dropped from 2 × 104 to 68. This 68-fold discrepancy may be due to the specific plasmid constructions in some way and is under investigation. Figure 4.Recombination frequencies of the different recombination sites and substrates. For a given substrate, the black bar indicates the recombination frequencies established in the in vivo recombination assay with non-replicative ss substrate and the light gray bar the corresponding value in the recombination control assay in replication permissive conditions, as described in Materials and methods. Recombination frequencies (vertical axis of histogram) correspond to the average of three independent trials. Error bars show standard deviations. For clarity, the recombination site, the strand–bs (bottom) or ts (top)—injected by conjugation, the orientation (+) and (−) of the oriT and the integrase used are indicated below plasmid names. Download figure Download PowerPoint Table 1. Bacterial strains used in this study E. coli strains Description/relevant characteristics Reference DH5α supE44 ΔlacU169 (Φ80lacZ′ΔM15) ΔargF hsdR17 recA1 endA1 gyrA96 thi-1 relA1 Laboratory collection Π1 DH5α ΔthyA::(erm-pir116) [ErmR] Demarre et al (2005) Π1977 Π1 pSU711ΔoriT::aac(3)-IV [GmR KmR ErmR] Demarre et al (2005) ED9 F-ΔlacU169 araD139 rpsL relA flbB ΔmalE444 srl::Tn10 recA1 [TcR] E Dassa (unpublished) β2163 MG1655::ΔdapA::(erm-pir)RP4-2-Tc::Mu [KmR] Demarre et al (2005) UB5201 F-pro met recA56 gyrA [NalR] Martinez and de la Cruz (1990) UB5201-I1 UB5201 pTRC99A::intI1 pSU38-attI1 This study UB5201-Pi UB5201ΔthyA::(erm-pir116) [NalRErmR] This study MG1655 E. coli K12 Laboratory collection MG1657 MG1655 ΔλattB::aadA ΔlacZ recA [SpecR] F Boccard (unpublished) MG1657-PIλ MG1657 pTSA29-CXI-AK pSU38Δ-attP This study We extended our strand recombination analysis to the attCaadA7 site used in the in vitro electrophoretic mobility shift assay (EMSA) (Figure 2). Two plasmids, pAttC-B and -T, were constructed (Table II), allowing the conjugative transfer of either attCaadA7 bs or attCaadA7 ts, respectively. As in the case of pVCR-B and -T, we found that in a [pir+] host, the B and T derivatives were recombined at similar rates (2 × 10−2), whereas in the suicide conjugation assay, attCaadA7 bs recombined at a rate (7.6 × 102) higher than attCaadA7 ts (Figure 4). To eliminate the possibility that these results were specifically linked to the RP4 transfer machinery, we repeated several of these experiments using plasmid R388, which specifies a different transfer system. Using plasmids pSW26 and pSW27, which carry the R388 oriT in opposite orientations (Demarre et al, 2005), we constructed two derivatives of each containing VCR2/1 in either orientation, leading to plasmids p388VCR-B and -T, and p388VCR-BINV and -TINV (Table II). These four plasmids were found to recombine with attI1 at similar rates when tested in UB5201-Pi, a [pir+] host expressing IntI1 (2.7–13 × 10−2; Figure 5). When measured after suicide conjugation from strain Π1977, which expresses the R388 transfer machinery, recombination of VCR2/1 bs occurred at rates 3.4 × 103 (p388VCR-B) and 6.2 × 103 (p388VCR-TINV) higher than VCR2/1 ts, after conjugation from p388VCR-T and p388VCR-BINV, respectively (Figure 5). Figure 5.Recombination frequencies of the VCR2/1 site obtained with the R388-based suicide conjugation assay. For a given substrate, the black bar indicates the recombination frequencies established in the in vivo recombination assay with non-replicative ss substrate and the light gray bar the corresponding value in the recombination control assay in replication permissive conditions, as described in Materials and methods. Recombination frequencies (vertical axis of histogram) correspond to the average of three independent trials. Error bars show standard deviations. For clarity, the recombination site, the strand—bs (bottom) or ts (top)—injected by conjugation and the type and orientation (+) and (−) of the oriT are indicated below plasmid names. Download figure Download PowerPoint Effect of mutations in the potential stem sequence To test our model of recombination involving the attC bs folding into a stem-and-loop structure, we introduced mutations that would disrupt the potential base pairing, and measured their effect on the recombination frequency. As the last positions involved in the potential stem formed by the various ss attC sites are not conserved and cannot be involved directly in the chemistry of the reaction (positions underlined in Figure 1C), we substituted the last 5 nucleotides of the attCaadA7 stem (attCaadA7Mut1; Figure 6). These mutations lead to a 10-fold decrease of the recombination frequency of the attCaadA7 bs, as established after suicide conjugation of plasmid pAttC-B-Mut1 (Figure 6). We then increased the destabilization of the potential secondary structure by the introduction of three additional substitutions further down in the stem structure and covering the last two positions in the L′/L″ potential hybrid (attCaadA7Mut3; Figure 6). These mutations lead to a 90-fold decrease of the recombination frequency of the attCaadA7. We then tested for both mutants the effect of restoration of complementarity, which would stabilize the attCaadA7Mut1 and attCaadA7Mut3 bs folding, on the recombination frequency. In both cases, these secondary mutations restored a level of recombination similar to the one obtained with the wild-type (WT) attCaadA7 site (attCaadA7Mut2 and attCaadA7Mut4; Figure 6), after suicide conjugation of the corresponding bs from pAttC-B-Mut2 and pAttC-B-Mut4. Figure 6.Proposed secondary structure for the attCaadA7 mutants bottom strand (A) and their recombination frequencies (B), as established in the suicide conjugation assay. Red letters indicate mutations introduced in the attCaadA7. Symbols are as in Figure 1. Download figure Download PowerPoint λ phage attB × attP recombination using the suicide conjugative transfer assay To confirm that these results truly reflect a single strand preference, we investigated the related phage λ recombination system that is known to require two strands. Here, orientation should have no effect on recombination. We used the λ phage integration, since its mechanism is known in detail (reviewed by Azaro and Landy, 2002). In this reaction, recombination between the phage attP and chromosomal attB sites requires ds substrates and is catalyzed by the λ integrase, Intλ. The attP site was cloned into pSU38 and introduced into the recipient, which supplied the accessory and necessary host protein IHF. The attB site was cloned in both orientations into pSW23T to create pλattB-1 and pλattB-2 (Table II). When tested in a [pir+] host containing pSU38-λattP and a plasmid expressing IntIλ, a 90 min induction was sufficient to obtain 100% attP × attB cointegrate formation with pλattB-1 and pλattB-2. Recombination of each of the λattB strands was then tested following suicide conjugation of either pλattB-1 or pλattB-2 into MG1657-PIλ, a ΔattB∷aadA E. coli that contained pSU38-λattP as recombination target and expressed Intλ. Conjugation was for 2 h and transconjugants were selected for the pλattB marker. This resulted in cointegrate formation (integration) frequencies of 2 × 10−7 and 0.6 × 10−7 for λattB-1 and λattB-2, respectively (Figure 4). Interestingly, increasing the conjugation time up to 3 h resulted in an ≈103 increase of cointegrate formation for both λattB ss substrates. This suggested that the increased time allowed for an increase in the amount of complementary strand synthesis in the recipient, generating the ds sequences necessary for an efficient attB × attP recombination to be catalyzed. Recombination properties of the attI1 site after suicide conjugative transfer From EMSA assays, neither the ts nor bs DNA of the other recombination partner attI1 appeared to be bound by IntI1, although ds attI1 site was clearly recognized (Figure 2). To determine whether this is also reflected in recombination, we tested recombination proficiency after suicide conjugative transfer. The attI1 site was cloned in both orientations in pSW23T, leading to pAttI1-B and pAttI1-T, and the attCaadA7 was inserted into pSU38, leading to pSU38-attCaadA7 (Table II). When tested in a [pir+] host expressing IntI1, pAttI1-B and pAttI1-T were found to recombine with the target pSU38-attCaadA7 at similar rates of 1.6–2 × 10−3 (Figure 4). These results, which do not significantly differ from those obtained when recombination sites and vector plasmids were reciprocally reversed, showed that under conditions that permit autonomous replication of all plasmids, the properties of the different recombination sites were independent of the backbone plasmid. Conversely, recombination following suicide conjugative transfer of either attI1 strands with the attCaadA7 site on pSU38 in the recipient was found to occur at identical low rates, about 1 × 10−5 (Figure 4), strongly suggesting that ss attI1 are not bona fide substrates. Recombination of attC and λattB sites after transformation using double-stranded non-replicative plasmids To determine whether the ds attC can be used as a recombination substrate, for example by adopting the necessary structure recognized by IntI1, without single strand passage, we transformed the ds circular plasmids pVCR-B and pVCR-T into the [pir−] strain UB5201-I1. This strain expresses IntI1 and carries the target plasmid pSU38-attI1. Competent cells were then transformed with 1, 10 and 50 μg of each of the pVCR derivatives and selected for CmR transformants, as cointegration between the VCR2/1 and the attI1 site carried on the pSU38 would lead to viable CmR transformants. No transformants were obtained for either of the tested plasmids, although the frequency of transformation for a compatible control plasmid, pSC101, was 1.1 × 105 transformants/μg. The maximal recombination frequencies for the ds test plasmids were therefore lower than 5.5 × 10−6. We performed the same type of experiment using plasmids pλattB-1 and pλattB-2, and transforming the [pir−] strain MG1657-PIλ expressing Intλ and containing the target plasmid pSU38-λattP. We obtained CmR transformants at frequencies of 2.1 × 104transformants/μg (pλattB-1) and 3.4 × 104 transformants/μg (pλattB-2), compared to 7.2 × 104 transformants/μg of the control plasmid pSC101. This gives recombination frequencies of 2.9 × 10−1 and 4.7 × 10−1, respectively, for these ds substrates. Discussion We have analyzed the mechanism of IntI1-mediated recombination that occurs during integron cassette acquisition and provide evidence that cassette integration occurs by recombination between ss attC and ds attI. Johansson et al (2004) reported covalent complex formation between IntI1 and the attCaadA1 bs, demonstrating that IntI1 not only recognized the bs (Francia et al, 1999), but was also able to catalytically cleave this substrate without the necessity of a complex cruciform structure formed from both attC strands. We have extended these studies to the related attCaadA7 and the poorly related VCR2/1 sites. Like attCaadA1, IntI1 bound only to the bs of ss attCaadA7 and ss VCR2/1. These observations led us to consider a model in which recombination would only involve a structured attC bs and a canonical ds attI site. The attC bs can potentially adopt a ds DNA-like structure by annealing of L″ to L′ and R″ to R′, which has almost all the structural features of a canonical recombination site (Figure 1D). These regions would be separated by an unpaired central segment. In most circularized cassettes, self-pairing of the same single strand could cover almost the entire attC site and even extend slightly further. Indeed, in most cases, the 7 bp R′ and R″ sequence complementarity is extended on the external part, to form a stretch of 9–11 consecutive complementary nucleotides (Hansson et al, 1997; Rowe-Magnus et al, 2003). In addition, comparison of the secondary structure adopted by the different attC sites, which are efficiently recombined by IntI1, shows that apart from the conserved AAC and GTT in the R′ and R″ boxes, and the flipped out G present in all L′/L″ stem, all the other positions show no conservation (Supplementary Figure 1). This hypothesis was tested in vivo using suicide conjugative transfer of one of the recombination sites, in order to provide an ss substrate (Figure 3). Conjugative transfer of mobile plasmids, such as RP4, occurs by transfer of a single DNA str