Targeted Conservative Cointegrate Formation Mediated by IS26 Family Members Requires Sequence Identity at the Reacting End

The IS26 family includes the ISs that are commonly found associated with antibiotic resistance genes in multiply resistant Gram-negative and Gram-positive bacteria. IS26 is most prevalent in Gram-negative species and can generate the clusters of antibiotic resistance genes interspersed with directly oriented IS26 seen in multiply resistant pathogens. ABSTRACT IS26 forms cointegrates using two distinct routes, a copy-in mechanism involving one insertion sequence (IS) and a target and a targeted conservative mechanism involving two ISs in different DNA molecules. In this study, the ability of IS26 and some close relatives, IS1006, IS1008, and a natural hybrid, IS1006/IS1008, which are found predominantly in Acinetobacter spp., to interact was examined. IS1006/1008 consists of 175 bp from IS1006 at the left end, with the remainder from IS1008. These ISs all have the same 14-bp terminal inverted repeats, and the Tnp26, Tnp1006, and Tnp1008 transposases, with pairwise identities of 83.7% to 93.1%, should be able to recognize each other’s ends. In a recA-negative Escherichia coli strain, IS1006, IS1008, and IS1006/1008 each formed cointegrates via the copy-in route and via the targeted conservative route, albeit at frequencies for the targeted reaction at least 10-fold lower than for IS26. However, using mixed pairs, targeted cointegration was detected only when IS1008 was paired with the IS1006/1008 hybrid, which also encodes Tnp1008, and the targeted cointegrates formed all arose from a reaction occurring at the end where the DNA sequences are identical. The reaction also occurred at the end with extended DNA identity using IS26 paired with IS26::catA1, an artificially constructed IS26 derivative that includes the catA1 gene. Thus, both identical transposases and identical DNA sequences at the reacting end were required. These features indicate that the targeted conservative pathway proceeds via a single transposase-catalyzed strand transfer, followed by migration and resolution of the Holliday junction formed. IMPORTANCE The IS26 family includes the ISs that are commonly found associated with antibiotic resistance genes in multiply resistant Gram-negative and Gram-positive bacteria. IS26 is most prevalent in Gram-negative species and can generate the clusters of antibiotic resistance genes interspersed with directly oriented IS26 seen in multiply resistant pathogens. This ability relies on the novel dual mechanistic capabilities of IS26 family members. However, the mechanism underlying the recently discovered targeted conservative mode of cointegrate formation mediated by IS26, IS257/IS431, and IS1216, which is unlike any previously studied IS movement mechanism, is not well understood. An important question is what features of the IS and the transposase are key to allowing IS26 family members to undertake targeted conservative reaction. In this study, this question was addressed using mixed-partner crosses involving IS26 and naturally occurring close relatives of IS26 that are found near resistance genes in Acinetobacter baumannii and are widespread in Acinetobacter species.

though the frequency of cointegrate formation was not reported. However, IS1006mediated cointegrates were not detected (15). These IS have subsequently been found in several different Acinetobacter plasmids often associated with complex antibiotic resistance regions (16)(17)(18)(19).
In this study, IS1006, IS1008, and an IS1006-IS1008 hybrid designated IS1006/1008 were first shown to perform both the untargeted copy-in reaction and the targeted conservative reaction. Then, to gain insight into the requirements for targeted conservative cointegrate formation, mixed pairs with IS26 or with one another were tested for their ability to act together in this mode.

RESULTS
IS1006, IS1008, and IS1006/1008. IS26 shares 75.4% nucleotide identity with IS1006 and 72.2% nucleotide identity with IS1008, and IS1006 and IS1008 share 87.4% nucleotide identity. The IS26 transposase Tnp26 shares 85.2% and 83.7% amino acid identity with Tnp1008 and Tnp1006, respectively (Fig. 1A), and Tnp1008 and Tnp1006 are even more closely related, sharing 93.1% amino acid identity. The majority of the amino acid differences occur in the DDE catalytic domain (Fig. 1A). There are no differences between Tnp1006 and Tnp1008 in the predicted helix-helix-turn-helix (H-HTH) DNA binding domain and only three differences relative to Tnp26 in this domain (Fig. 1A).
We have also identified an additional related IS, originally named IS1008-like (16) but henceforth called IS1006/1008, that is a hybrid formed between IS1006 and IS1008 (bases 42337 to 43155 in Acinetobacter baumannii strain J9 plasmid pJ9-3 [GenBank accession number CP041590]). The crossover occurs between bases 172 and 175, with the first 175 bases identical to IS1006 and the remaining bases identical to IS1008 (Fig. 2). The transposase encoded by IS1006/1008 is identical to the IS1008 transposase, as the nucleotide differences before the crossover do not result in any amino acid substitutions.
Importantly, the 16-bp terminal inverted repeats (TIR) of IS1006, IS1008, and IS1006/ 1008 are identical to one another and to the IS26 sequence at the left end. At the right end, the first 14 bp are identical to the 14-bp TIR of IS26 (Fig. 1B). In the case of IS1006 and IS1008, identity at their right ends extends beyond the 16-bp TIR for a total of 33 bp. As transposases generally recognize and bind to the TIR, we predict that the transposases of IS26, IS1006, IS1008, and IS1006/1008 should be able to recognize each other's TIR, opening the possibility that these related transposases can function together to perform the targeted conservative reaction.
IS1006, IS1008, and IS1006/1008 are active in copy-in and targeted conservative cointegrate formation. Untargeted cointegrate formation in a recA-negative E. coli was examined using a mating-out assay to detect cointegrates formed between the FIG 2 Schematic representations of IS1006, IS1008, the hybrid IS1006/1008, and IS26::catA1. (A) Nucleotide sequences belonging to IS1006 and IS1008 are shaded in orange and blue, respectively. The crossover between the two sequences in IS1006/1008 is marked by a vertical line, indicating the first base belonging exclusively to IS1008. The extent and orientation of the transposase open reading frame are indicated by a black arrow. Amino acids that differ between the transposases are marked by red and black letters. The H-HTH putative DNA binding domain and the DDE catalytic domain are marked at the top. The positions of the conserved DDE catalytic triad are marked by bold letters. In IS1006/1008, the nucleotide identity between IS1006 and IS1008 is indicated in the two segments. Drawn to scale from GenBank accession numbers CP012956, CP041590, and KU744946 for IS1006, IS1006/1008, and IS1008, respectively. (B) The synthetic IS26::catA1 construct is shown with IS26 in green and the catA1 gene fragment in orange. The extent and orientation of the tnp26 and catA1 genes are indicated by a black arrow. The position of the inserted catA1 fragment in IS26 is marked.
To confirm that target selection was not sequence specific, the locations of the junctions between the IS1006-, IS1008-, and IS1006/1008-containing plasmids and R388 in the cointegrate were initially determined using the restriction mapping strategy with BglII and BsiWI described in Materials and Methods. Analysis of the fragments showed that in each cointegrate examined, pRMH1011, pRM1012, or pRMH1013 had incorporated at a different position in the R388 backbone (see Fig. S1 in the supplemental material). Then, a series of PCR primers was designed (Table S1) to map and sequence the precise boundaries and relative orientation of the two plasmids in six  cointegrates from each experiment. In each case, both potential orientations of the plasmids relative to one another were observed (1 and 2 in Fig. S1). Targeted cointegration was tested using pRMH1011 (IS1006 Ap r ) and R388::IS1006 (Tp r ), pRMH1012 (IS1008 Ap r ) and R388::IS1008 (Tp r ), or pRMH1013 (IS1006/1008 Ap r ) and R388::IS1006/1008 in a recA-negative background to ensure that all events detected were catalyzed by the available transposase. Cointegrates were formed between identical pairs of IS at frequencies ranging from 3.1 Â 10 25 to 5.1 Â 10 25 per transconjugant, averaged from three independent determinations (Table 2). These frequencies are approximately 100-fold higher than for the copy-in values. However, they are approximately 10-fold lower than the values obtained in this study (Table 2) and previously for the reaction between two wild-type IS26 under the same conditions (5,6,8,9), which average 4.3 Â 10 24 cointegrates/transconjugant from 16 independent determinations.
Ten Ap r Tp r cointegrates from each of the three independent experiments for IS1006, IS1008, and IS1006/1008 were screened by PCR using primer pairs (RH2563 with RH2703 and RH2702 with RH2735) that amplify across each of the two IS separating the two replicons in targeted cointegrates when the two IS in are the same orientation (Fig. 3A). This confirmed that in all instances pRMH1011, pRMH1012, or pRMH1013 had been incorporated adjacent to the existing IS in R388::IS1006, R388::IS1008, or R388:: IS1006/1008, respectively. Hence, the IS26 relatives IS1006, IS1008, and IS1006/1008 are able to perform the targeted conservative cointegrate formation reaction previously demonstrated for IS26, IS257, and IS1216.
Is cross-recognition possible between related IS26 family members? Granted their identical TIR and the high similarity between their transposases, the possibility that IS1006, IS1008, and/or IS1006/1008 may be able to perform the targeted conservative reaction with IS26 was examined. Both combinations where one molecule contained IS1006, IS1008, or IS1006/1008 and the other molecule contained IS26 were tested. E. coli UB5201 (Nx r ) containing R388::IS26 (Tp r ) and either pRMH1011 (IS1006 Ap r ), pRMH1012 (IS1008 Ap r ), or pRMH1013 (IS1006/1008 Ap r ) were mated with E. coli UB1637 (Sm r ), and in all three cases, Ap r Tp r cointegrates were detected (Table 3). However, the frequency of cointegrate formation, ranging from 3.7 Â 10 27 to 9.1 Â 10 27 cointegrates per transconjugant (Table 3), was significantly lower than that expected if the cointegrates were being formed via the targeted conservative route but similar to those obtained for the copy-in route ( Table 2). In the reciprocal experiment, performed using E. coli UB5201 containing R388::IS1006, R388::IS1008, or R388::IS1006/1008 and pRMH977 (IS26 Ap r ) as the donor, cointegrates formed at average frequencies of between 6.6 Â 10 27 and 8.1 Â 10 27 per transconjugant (Table 3). Again, these frequencies are similar to those obtained via the untargeted copy-in route for each IS (Table 2).
PCR screening of DNA from three Ap r Tp r cointegrates from each experiment (9 cointegrates per combination) did not generate the products expected if the reaction was targeted (Table 4). This indicates that the cointegrates that were detected had not formed via a targeted reaction between the mixed pairs of IS and had likely been formed by one of the IS present using the copy-in cointegration route at a random target site (Fig. 3B). To determine which IS formed the cointegrate, pairs of primers in the plasmid backbones that amplify across the original single IS were used to identify which IS remained in its original position, indicating that it had not mediated the cointegration event. As shown in Fig. 3B, primer P1 with P2 detects the IS in the pUC19 constructs (pRMH977, pRMH1011, pRMH1012, and pRMH1013), and primer P3 with P4 detects the IS in the R388 constructs (R388::IS26, R388::IS1006, R388::IS1008, and R388:: IS1006/1008). DNA from the nine cointegrates from each IS combination screened as described above was examined using this strategy, and either the IS in pUC19 or the IS in R388 was still in the original position in each of the cointegrates tested (Table 4). Cointegrates were formed at approximately equal frequencies by the IS in pUC19 and the IS in R388, despite the difference in copy numbers of the two plasmids. Hence, the cointegrates had all been formed via the copy-in route and IS1006, IS1008, and IS1006/ 1008 were not able to combine with IS26 to perform the targeted conservative reaction. Can IS1006 and IS1008 act together? Even though IS1006 and IS1008 were unable to react with IS26, it remained possible that they may be able to react with one another, as Tnp1006 and Tnp1008 share 93.1% amino acid identity and 96.6% amino acid similarity (226/234 residues), with no differences in the putative H-HTH DNA binding domain, and they have identical 16-bp TIR. However, Ap r Tp r cointegrates that formed between R388::IS1006 and pRMH1012 (IS1008) or between R388::IS1008 and pRMH1011 (IS1006) were detected at mean frequencies of 5.9 Â 10 27 or 4.0 Â 10 27 per transconjugant (averaged from three independent determinations) ( Table 3), suggesting the copy-in route. Using the PCR screening strategy illustrated in Fig. 3B with DNA from three cointegrates from each independent cross (18 in total), it was shown that none had been formed by the targeted reaction. Rather, all 18 had been formed by either IS1006 or IS1008 performing an untargeted copy-in reaction at a naive site in the second molecule (Table 4), and again, no preference for either IS was observed. Hence, despite their close relationship, IS1006 and IS1008 were unable to act together.
IS1008 and IS1006/1008 can act together. IS1008 and IS1006/1008 both encode Tnp1008, and they share 649 bp of nucleotide sequence identity at their right ends   (Fig. 2). When pRMH1013 (IS1006/1008) was tested in combination with R388::IS1008, cointegrates formed at an average frequency of 4.2 Â 10 25 per transconjugant and the reciprocal experiment, pRMH1012 (IS1008) in combination with R388::IS1006/1008, yielded a similar result (shaded rows in Table 3). These values are similar to the frequencies observed for the targeted cointegration reactions between identical IS ( Table 2). PCR screening of a total of 30 cointegrates (5 Ap r Tp r cointegrates from each of the three independent experiments per pair) confirmed that in all instances, pRMH1012 or pRMH1013 had been incorporated adjacent to the existing IS in R388:: IS1006/1008 or R388::IS1008, with the two IS in the same orientation. Hence, IS1006/ 1008 and IS1008 are able to act together in the targeted conservative cointegration reaction.
The targeted cointegration reaction requires extended identical DNA sequences. We considered the possibility that nucleotide identity could be critical if branch migration is needed to process the strand transfer intermediate, predicted to be a nicked Holliday junction (HJ) formed via a single-strand transfer event, formed by the transposase. Hence, we examined whether the reaction had occurred only (or preferentially) at the right end, where IS1008 and IS1006/1008 share significant length of nucleotide identity (649 bp), and not at the left end, where only the first 16 bp are identical and the 175-bp segment derived from IS1006 differs from the IS1008 sequence at 32 positions. The PCR products generated as described above ( Table 4) that spanned the junctions of the 15 Ap r Tp r cointegrates formed between pRMH1013 (IS1006/1008) and R388::IS1008 were sequenced. In all 15 instances, IS1006/1008 had been distributed to the right-hand junction of the cointegrate, demonstrating that the reaction must have occurred between the right ends of IS1008 and IS1006/1008 (Fig. 4A). Conversely, in the 15 Ap r Tp r cointegrates formed via the reaction between pRMH1012 (IS1008) and R388::IS1006/1008, IS1006/1008 had been distributed to the left-hand junction (Fig. 4C). The absence of the possible alternative configurations (Fig. 4B and D) indicates that the reaction could not occur via the left ends, where only 16 bp of sequence is shared. Taken together, these results indicate that the targeted conservative reaction can occur between two IS that share identical transposases and share nucleotide identity extending for a significant distance inward from one end.
To confirm this observation, we used a chemically synthesized IS26 derivative designated IS26::catA1, which includes the catA1 chloramphenicol resistance gene and an upstream promoter inserted downstream of the tnp26 gene (Fig. 2B). In this construct, 775 bp at the left end and 44 bp at the right end are identical to IS26. Hence, as the transposase produced is Tnp26, it should be able to pair productively with IS26, but in this case, the reaction should occur predominantly at the left end. Cointegrates formed between pRMH1015 containing IS26::catA1 (Ap r Cm r ) and R388::IS26 were detected at a mean frequency of 5.5 Â 10 24 Ap r Cm r Tp r cointegrates/Tp r transconjugant (Table 3, last row), equivalent to the frequency for the reaction of IS26 with IS26 ( Table 3). As determined by using the primers RH1451 with RH1472 and RH1452 with RH1471 described previously (6), which are specific for the position of IS26 in R388 (but equivalent to those shown in Fig. 3A), 30 of 30 cointegrates examined (10 from each independent cross) had formed via a targeted reaction ( Table 4). The sizes of the PCR products produced using the mapping primers revealed that, as predicted, in all cases the reaction occurred at the left end. The simplest explanation for these findings is that the targeted reaction requires progression of an HJ formed by the transposase (Tnp1008 or Tnp26).
IS1006 and IS1006/1008 do not act together. Although IS1006 and the hybrid IS1006/1008 also share an extended region of 175 bp of sequence identity, when pRMH1013 (IS1006/1008) was tested in combination with R388::IS1006 or pRMH1011 (IS1006) with R388::IS1006/1008, Ap r Tp r cointegrates were formed at an averaged frequency of 5.6 Â 10 27 or 6.9 Â 10 27 per transconjugant (Table 3). PCR screening of 18 cointegrates showed that each had been formed via the untargeted route (Table 4), mediated by either IS1006/1008 or IS1006. Hence, if the 175 bp of identical DNA sequence at the left ends of IS1006 and IS1006/1008 is sufficient to enable the targeted conservative reaction, then Tnp1006 and Tnp1008, after binding in cis to one end of the IS they were produced from, are unable to form a productive synapse.
To examine the if 175 bp of DNA identity is sufficient, we used IS26-FS constructs, which include a frameshifting alteration, a duplication of 4 bp located at 114 to 117 from the left end in IS26-FS-L or a deletion of 13 bp located 184 bp from the right end in IS26-FS-R. These differences should severely impede progress of an HJ formed between IS26 and either IS26-FS-L or IS26-FS-R. We have previously shown that when IS26-FS-L or IS26-FS-R was paired with IS26, cointegrates were formed at low frequency but always by the targeted route (6). The relatively even distribution of the IS26-FS to the left-hand or right-hand position in 10 cointegrates for each IS (6) indicates that 117 or 184 bp constituted sufficient stretches of identity to allow that IS end to be used. In this study, a further 10 cointegrates formed by pRMH962 containing IS26-FS-L with R388::IS26 or by pRMH762 containing IS26 with R388::IS26-FS-R were isolated and FIG 4 Outcome of targeted conservative cointegrate formation between IS1006/IS1008 and IS1008 or IS1006. Shown are reactions occurring between the right ends (A and C) and the left ends (B and D) of IS1008 and IS1006/1008 or between the left (E) or right (F) ends of IS1006 and IS1006/1008. The predicted cointegrate configurations are depicted below the vertical arrow. The red cross indicates where a viable cointegrate was not detected. IS1008-and IS1006derived sequences are in blue and orange, respectively. An arrow indicates the position and orientation of the transposase gene. The two DNA molecules are identified by either solid lines (R388) or dashed lines (pUC19 derivatives). For clarity, the specific sequences next to each IS are marked with abc, def, UVW, or XYZ. The reacting ends are indicated by a red double-ended arrow. examined as described previously (12). Again, approximately equal numbers of targeted cointegrates had arisen via a reaction at the left or via the right end. Totals from the two data sets were 13 on the left and 7 on the right for IS26-FS-R and 11 on the left and 9 on the right for IS26-FS-L.
As 117 bp was sufficient for the targeted reaction to occur, we conclude that the stretch of identity between IS1006 and IS1006/1008 of 175 bp should be sufficient. Hence, we conclude that the reason that IS1006 did not work with IS1006/1008 ( Fig. 4E and F) arises from the differences in the catalytic domain of the Tnp1006 and Tnp1008 transposases. Though relatively few in number, 1 or more of these 16 differences appear to have been sufficient to impede formation of a productive synapse.

DISCUSSION
The work reported here increases to six the number of IS belonging to the IS26 family, as recently defined (1), that have been shown experimentally to be able to form cointegrates using both copy-in and targeted conservative mechanisms. Investigations of the possible interactions between several pairs of closely related IS revealed that cointegrates formed via the targeted conservative route were recovered only when the same transposase was produced by both of the participating IS. However, there was an additional requirement for DNA identity at the IS end (left or right) involved in formation of a cointegrate. To explain the need for sequence identity at the reacting end, we propose that HJ branch migration is required to ensure that the single-end transfer catalyzed by the transposase leads to a cointegrate molecule. The model proposed is shown in Fig. 5A. Based on previous observations, we expect the transposase to bind preferentially in cis to one end of the IS it is produced from (7). We also expect that after the ends are brought together, presumably via formation of a transposase dimer, only a single-strand cleavage and strand transfer event occurs and this would form a nicked HJ. This is because when one strand transfer occurs it will cleave the phosphodiester bond that would be the target of a potential second strand transfer (Fig. 5B). As the ends involved in a conservative reaction are identical (left with left or right with right), strand cleavage and transfer can be initiated by one end of either participating IS and attack the equivalent end of the other IS. As the reaction can occur at either the left or the right end, there will be four initial intermediates in total (two are shown in Fig. 5B).
We have suggested previously that replication could be required to complete formation of the cointegrate (8). However, the observed requirement for extended DNA identity points to migration of the HJ. Supporting this conclusion, we have observed gene conversion occurring in the targeted conservative mode when one wild-type IS26 has reacted with an IS26 containing a mutation causing a single amino acid substitution located near the left end of the IS (C. H. Pong and R. M. Hall, unpublished observations), and this is consistent with the repair of mismatches arising from branch migration. When HJ migration is followed by HJ resolution involving the nontransferred strands, this would produce the cointegrate products observed (Fig. 5A). To the best of our knowledge, this mechanism has not been observed for any scenario involving a DDE transposase (12,20) and hence represents a novel completely conservative reaction route.
Investigation of further details of this process is now needed. For example, if mutations were introduced into IS1006/1008 in order to convert Tnp1008 produced to Tnp1006, pairing of this construct with IS1006 would be predicted to move the reacting end to that derived from IS1006 and pairing with IS1008 would not produce products. Information on the extent of the DNA identity needed for an efficient reaction and what length of sequence identity is the minimum that can sustain the conservative reaction warrants further investigation. Whether the sequence identity must be within the IS or can also be adjacent to it also warrants investigation. The involvement of proteins that process the HJ, such as RecG or RuvABC, should shed further light on the mechanism.
Branch migration would stall when there are multiple mismatched nucleotides and particularly at clusters of mismatches, and this may be one reason why the closely related IS, such as IS1006 and IS1008 or IS1006/1008, were unable to act together in the targeted conservative reaction. However, IS1006 and IS1006/1008 could not interact to support conservative cointegrate formation even when extended sequence identity was available. This indicates that 1 or more of the modest number of amino acid differences between Tnp1008 and Tnp1006 (16 total, 6 conservative) are sufficient to prevent the reaction from occurring at all. Further work is needed to investigate this aspect. For example, the 16 mutants that each cause a single amino acid differences will need to be constructed and tested to begin to identify some of the key residues involved in multimerization and synapse formation.

MATERIALS AND METHODS
Bacterial strains and media. E. coli DH5a (supE44 DlacU169 [f 80 lacZDM15] hsdR17 recA1 endA1 gyrA96 thi-1 relA1) was used to propagate plasmids. E. coli UB5201 (pro met recA Nx r ) was used as a donor in mating-out experiments, and E. coli UB1637 (lys his trp lac recA Sm r ) was used as a recipient. Antibiotics (Sigma) were added at the indicated concentrations to either Mueller-Hinton broth or Mueller-Hinton agar, as appropriate: ampicillin (Ap), 100 mg/ml; nalidixic acid (Nx), 25 mg/ml; streptomycin (Sm), 25 mg/ml; and trimethoprim (Tp), 25 mg/ml. DNA manipulation. Plasmid DNA was isolated by alkaline lysis as described previously (6). DNA was digested with restriction enzymes according to the manufacturer's instructions, and fragments were separated through 0.7% agarose in 1Â Tris-acetate-EDTA (TAE) buffer. The size standards were a 1-kb ladder and l-HindIII (New England BioLabs). PCRs were performed using conditions previously described (6), and routine sequencing of PCR products was performed as previously described (6). The sequences for all primers used in this study are listed in Table S1.