|Home | About | Journals | Submit | Contact Us | Français|
In this study, we report on the transposition behavior of the mercury(II) resistance transposons Tn502 and Tn512, which are members of the Tn5053 family. These transposons exhibit targeted and oriented insertion in the par region of plasmid RP1, since par-encoded components, namely, the ParA resolvase and its cognate res region, are essential for such transposition. Tn502 and, under some circumstances, Tn512 can transpose when par is absent, providing evidence for an alternative, par-independent pathway of transposition. We show that the alternative pathway proceeds by a two-step replicative process involving random target selection and orientation of insertion, leading to the formation of cointegrates as the predominant product of the first stage of transposition. Cointegrates remain unresolved because the transposon-encoded (TniR) recombination system is relatively inefficient, as is the host-encoded (RecA) system. In the presence of the res-ParA recombination system, TniR-mediated (and RecA-mediated) cointegrate resolution is highly efficient, enabling resolution both of cointegrates involving functional transposons (Tn502 and Tn512) and of defective elements (In0 and In2). These findings implicate the target-encoded accessory functions in the second stage of transposition as well as in the first. We also show that the par-independent pathway enables the formation of deletions in the target molecule.
It is widely recognized that mobile genetic elements contribute to genome plasticity and have been a driving force in the emergence and spread of resistance determinants within and between bacterial species; their impact is ongoing (10, 51). Significant among these elements are various classes of plasmids, transposons, and integrons which may lack resistance determinants or carry one or multiple determinants. Resistance determinants that have become globally dispersed in environmental and clinically significant bacteria include mercury(II) resistance (2, 17), evident even in ancient bacteria (27), and antibiotic resistance, which has increased in dominance since the advent of the antibiotic era (23, 40).
This paper concerns the mercury resistance (mer) transposons Tn502 and Tn512, whose sequence organization and transpositional behavior show that they are new members of a family of elements exemplified by the mer transposon Tn5053 (22). These elements are closely related to those in the Tn402 family, which contain an integron (intI) recombination system (14, 36). Members of the two families differ in the positions of the mer or intI determinants (modules) near one end of the transposition (tni) module. The latter module contains four genes (tniABQR), and the entire transposon is bounded by 25-bp inverted-repeat termini (IRi and IRt). TniA, TniB, and TniQ are required to form the transpositional cointegrate, which is then resolved by the action of TniR (a serine resolvase) on a resolution (res) sequence located between tniR and tniQ (22). The transposon in its new location is flanked by 5-bp direct repeats (DRs) (20, 22). TniA, which contains a D,D(35)E transposase catalytic motif, is thought to function cooperatively with TniB, a putative nucleotide-binding protein, as the active TniAB transposase (21, 36). Studies of TniA conducted in vitro show binding to the IRs and to additional 19-bp repeat sequences that make up the complex termini of the transposon (21). The precise role of TniQ is unknown.
An unexpected and unique feature of Tn5053 and Tn402 is that they depend on externally coded accessory functions for efficient transposition, namely, a res site served by a cognate resolvase (25). As a consequence, these transposons exhibit a strong transpositional bias for some target res sites (20, 25, 26) and have aptly been described as “res site hunters” (25). One such efficient interaction involves the res-ParA multimer resolution system of plasmid RP1 (IncPα); other plasmid- or transposon-encoded systems are less efficient or are refractory. Although the role of the external resolvase remains obscure, its capacity to bind to its cognate res is an essential requirement whereas its catalytic activity is not (20). For each interaction system, the target sites typically cluster in a single part of res but not necessarily within the same subregion and, on occasion, can lie in the vicinity of res. Typically, the transposon is in a single orientation with IRi closest to the resolvase gene. In one study, Tn402 clustered at two target sites, one within res and one nearby, and the orientations were different at the two sites (20).
The experimentally observed target preference described above also occurs in natural associations of Tn5053/Tn402-like elements and became evident on sequencing class 1 integrons, which were often found positioned close to different res-resolvase gene regions (6, 20, 25). Most Tn402 family elements are comprised of an intI module that is flanked on the left by IRi and on the right by a 3′ conserved sequence (3′-CS) (13). In others, a remnant tni gene cluster may be present instead of the 3′-CS, and IRt occurs at the right flank. The structure of the latter category of integrons strongly indicated that they are defective transposons that were presumably capable of relocation provided that tni functions were supplied in trans (6, 32). The movement of In33 (Tn2521) from a chromosomal to a plasmid location appears to have been such an in trans event (30, 42), and others involving In0 and In2 are demonstrated in this study. In contrast, the integrons that lack the IRt end appear to be nonmobile remnants of Tn402-like transposons; they belong to several lineages, including those in which the incurred deletions are attributable to acquired insertion sequences (6). More recently, intact Tn5053/Tn402-like transposons and class 1 integrons have increasingly been detected in the res-parA region of IncP plasmids (39), which are arguably the most promiscuous of known plasmids (50). These various experimental and natural interactions provide insight into the dispersal pathways possible for Tn5053/Tn402-like elements.
The res-hunting attribute is a striking feature that is experimentally supported by studies of four family members (namely, Tn5053 [22, 25], Tn402 [20, 26], and in this study, Tn502  and Tn512). Another facet of the transposition of Tn502 is explored here. It concerns the observation that loss of the preferred par target region in RP1 does not abolish transposition of Tn502 (48), contrary to the finding with Tn5053 (25, 26) and, in this study, Tn512. The continued, low-frequency transposition of Tn502 involved at least three dispersed locations (48); however, nothing is known about the nature of these sites or about the features and requirements of the transposition process. Here we address these issues and uncover the existence of an alternative, par-independent pathway that is employed by Tn502 and is available to Tn512 under some circumstances. The study also provides information on the roles of the TniR and host (RecA) recombination systems in the resolution of transpositional cointegrates and on the ability of the par-independent transposition pathway to generate plasmid deletions.
The Escherichia coli K-12 derivatives used were DH5α (recA1 Res− Mod+ Nalr) (15), JC3272 (Strr auxotroph) (1), and the spontaneous rifampin-resistant mutants LT111 and LT112 obtained from UB281 (pro met Nalr) (16) and its recA56 derivative UB5201 (38), respectively. The auxotrophic rifampin-resistant Pseudomonas aeruginosa strains PAO9501 (46), its derivative PAO9529 (which has Tn502 in the chromosome) (48), and PAO9505 (46) were also used, as was the IncP-specific phage PR4 (45). Nutrient agar (NA), nutrient broth (NB), and diagnostic sensitivity agar (DST) have been described previously (28). For selection of E. coli strains (or of P. aeruginosa, as indicated), antimicrobials were used in NA at the following final concentrations (μg/ml): ampicillin (Ap) at 100, carbenicillin (Cb) at 250 (for P. aeruginosa), chloramphenicol (Cm) at 10, kanamycin sulfate (Km) at 10 or 350 (for P. aeruginosa), mercuric chloride (Hg) at 10 (for both E. coli and P. aeruginosa), nalidixic acid (Nal) at 8, rifampin (Rif) at 100 (for both E. coli and P. aeruginosa), streptomycin sulfate (Sm) at 100 (for P. aeruginosa and for chromosome-borne resistance in E. coli) or 5 (for plasmid-borne resistance in E. coli), and tetracycline hydrochloride (Tc) at 10 or 80 (for P. aeruginosa). Sulfonamide (Su) at 80 or 350 (for P. aeruginosa) μg/ml was used in DST supplemented with lysed horse blood (2% [vol/vol]).
The newly sequenced mer transposons used in this study are Tn502 (9,647 bp) (48) and Tn512 (8,450 bp) (34), both from clinical isolates of P. aeruginosa strains. The plasmids used are listed in Table Table1,1, except for additional derivatives of pUB307 and pUB1601, which are described in Fig. Fig.11 and and3.3. The ΔtniQ allele in pVS1703 was generated by removing the ClaI site in tet of pVS983 (by AatII-EcoRV digestion and religation to form pVS1702) and then treating the ClaI site in tniQ of pVS1702 with mung bean nuclease. The religated plasmid, pVS1703, lacks a 12-nucleotide (nt) sequence at the ClaI site. The predicted ΔTniQ protein is identical to the wild-type TniQ protein of Tn502 (TniQ502) (405 amino acids) except for the loss of amino acids 333 to 336. The Tn21-associated tnpA gene in pUB2401 was inactivated by KpnI digestion of the plasmid, treatment with mung bean nuclease, and religation to produce pVS1697. An amplicon of tniM (0.6 kb) was generated using the forward primer SP23 (5′-GCGGAATTCGGTATTCCTGCCTGCAAC) and the reverse primer SP15 (5′-GAATTCGGGACGAAACTCTCAAC) and cloned into pDHC29 to produce pVS1715.
Cloning and DNA manipulation were performed using standard methods (37). Plasmid DNA from various E. coli strains and derivatives of P. aeruginosa PAO9505 was isolated from 10-ml overnight NB cultures by use of an alkaline lysis miniprep method applied to 1.5 ml or 10 ml of culture, respectively. DH5α served as the recipient in transformation experiments designed to isolate or purify recombinant plasmids arising from in vitro or in vivo events. To map Tn502 and Tn512 insertion sites, DNA was amplified by PCR using tniA- and parE-specific primers SP11 (5′-GATGATTTCCGCCCGTTG) and SP12 (5′-CTCATGTCCTTAAACGGG), respectively, in reaction mixtures containing 10% (vol/vol) dimethyl sulfoxide and Taq DNA polymerase (Invitrogen). The mixtures were overlaid with sterile mineral oil and subjected to 36 thermal cycles as follows: 92°C, 3 min (first cycle only); 92°C, 1 min; 60°C, 1 min; 72°C, 2 min; and 72°C, 5 min (last cycle only). The reaction products were detected using agarose gel electrophoresis and staining with ethidium bromide. DNA sequencing was performed at the VABC Aggenomics sequencing facility (La Trobe University) to precisely identify Tn502 and Tn512 insertion sites (using SP11 and merR-specific primer SP21 [5′-CACTGCGAGGAAGCCAGCAGCTTG]) and In0 In2 insertion sites (using SP11 and intI1-specific primer SP18 [5′-CCGGATCAGAACGTATAC]).
All conjugation experiments, including transposition assays, were conducted by the quantitative filter method (48) and involved equal volumes (0.5 ml) of donor and recipient bacteria from NB-grown cultures. The donor strains (usually derivatives of DH5α) contained two or more compatible plasmids that were sequentially introduced by transformation (for nonconjugative plasmids) or conjugation (for self-transferable plasmids). In each step of the construction, selection was imposed for the distinctive resistance marker associated with the incoming plasmid, followed by purification of a mixture of the colonies on NA selective for the incoming and resident plasmid(s). Nonconjugative plasmids were introduced first, and the conjugative plasmid was introduced last. When the full complement of plasmids was present, a mixture of colonies was subcultured three times on the selective medium, and a donor culture was prepared in 10 ml NB and grown to late exponential phase. P. aeruginosa PAO9505, grown overnight at 42°C to induce a temporary restriction-deficient phenotype (24), was used as the recipient. This strain is highly sensitive to Hg(II) and, unlike E. coli K-12, produces no “background” growth when high cell numbers are plated on Hg(II)-containing selective NA. The use of PAO9505 thus enabled the detection of low-frequency Hgr transfer events. All assays were performed on three independent occasions, and the average transfer or transposition frequency (transconjugants per donor cell) was calculated.
Transconjugants from all transposition assays involving pUB307 were subjected to PCR analysis (using SP11 and SP12) to identify insertion (and orientation) of the elements within the res region. In addition, Hgr transconjugants from all transposition assays involving Tn502 were screened for resistance phenotype to determine plasmid content. Coinheritance of the distinctive Apr marker of the transposon donor (pVS983) indicated the presence of a plasmid cointegrate (i.e., pUB307- or pUB1601-pVS983) whereas an Aps phenotype indicated the presence of a resolved plasmid (i.e., pUB307::Tn502 or pUB1601::Tn502). Among the latter category, the occurrence of a Tcs phenotype indicated insertion of Tn502 into the tet gene of the target plasmid. In matings involving Tn512, the transposon donor (pVS1718) lacked a distinctive marker. Consequently, plasmid DNA was isolated from Hgr transconjugants and analyzed following SalI digestion (for pUB307 derivatives) or SstII digestion (for pUB1601 derivatives). Thus, pUB307::Tn512 resolved plasmids yielded two large fragments (≈20 kb and ≈44 kb), and pUB307-pVS1718 cointegrates yielded two additional fragments of about 2 kb (from the pBR322 component of pVS1718) and ≈12 kb (one of the Tn512-pUB307 junction fragments); only pUB1601-pVS1718 cointegrates were detected, and the restriction profile varied depending on which SstII fragment of pUB1601 (i.e., 21.5 kb, 14.5 kb, 10.0 kb, or 2.0 kb) had the Tn512 insert.
Transposition of In2 from pVS1697 to the target plasmid involved the selection of Sur transconjugants of PAO9505 or Smr transconjugants of LT111 or LT112. Transconjugants with resolved pUB307::In2 plasmids were detectable based on resistance phenotype. Transconjugants that displayed the full complement of pUB307 and pVS1697 resistance markers were outcrossed by conjugation. Linkage analysis of the transconjugants from the outcross enabled the inheritance of pUB307::In2 plasmids to be distinguished from that of pUB307-pVS1697 cointegrates. The single pUB1601::pVS1697 cointegrate that was detected in a transconjugant from PAO9505 (Table (Table2)2) was confirmed by transformational outcross experiments to DH5α and SalI restriction analysis of the recovered plasmid. The profile obtained (i.e., fragments of 8 kb, an ≈11.5-kb doublet, and one fragment of >23 kb) was consistent with the unique SalI sites in pUB1601 and the components (pACYC184 and In2) of pVS1697 and with the expected duplication of In2 in the cointegrate. The exact location of In2 in pUB1601 could not be established; however, the 8-kb fragment that was generated indicated that it is either ≈1.1 kb or 4 kb (depending on the orientation) from the SalI site in pUB1601.
The nucleotide sequences for Tn502 and Tn512 have been deposited in GenBank under accession numbers EU306743 and EU306744, respectively.
Based on the recovery of Hgr transconjugants from matings involving E. coli donors and P. aeruginosa PAO9505 as the recipient (see Materials and Methods), transposition of Tn502 and Tn512 to the par+ conjugative plasmid pUB307 (3) occurred efficiently, although the frequency was typically higher (≈10-fold) for Tn502 (Table (Table2,2, lines 2 and 4). All of the pUB307::Tn512 and pUB307::Tn502 plasmids that were studied (10 from each of five experiments for the Tn502+ and the Tn512+ donor) generated an ≈800-bp PCR product when tniA- and parE-specific primers (SP11 and SP12, respectively) were used, with the exception of two pUB307::Tn502 plasmids which produced no PCR product. This outcome suggested that Tn502 and Tn512 have a preferred target site(s) within the res sequence of par and that both transposons insert in a single orientation (i.e., with the tniA end of the transposon closest to parD). Ten independently derived plasmids were studied further to determine the exact insertion sites (Fig. (Fig.1B).1B). In five Tn512+ plasmids and three Tn502+ plasmids, the insertions were within an AATTT sequence in res (corresponding to nt 35014 to 35021 of RP1); in the two exceptional Tn502+ plasmids mentioned above, the transposon was nearby (at nt 35030) in the nonpreferred orientation or in parE (at nt 35664). As expected for members of the Tn5053 family, Tn502 and Tn512 in each plasmid were flanked by 5-bp DRs of the target DNA sequence.
In contrast, the outcome was significantly different when the par-deleted (Δpar) plasmid pUB1601 (53) served as the target plasmid (Table (Table2,2, lines 3 and 5); transposition of Tn512 was undetectable (<2.0 × 10−9), and that of Tn502 was markedly reduced (>104-fold). The latter confirmed our earlier observation (48) that Tn502 can transpose when par is absent, albeit inefficiently, an attribute that is not displayed by Tn512. To gain insight into the process of par-independent transposition of Tn502, we further analyzed the Hgr transconjugants that were isolated from donors carrying pUB1601 (Δpar) (Tcr Kmr) and pVS983 (Tn502+) (Apr Hgr) (typically 60 to 100 transconjugants per mating experiment). The vast majority contained pUB1601-pVS983 cointegrate plasmids, as determined from their resistance phenotype and pattern of marker coinheritance on outcrossing (see Materials and Methods). Some of the transconjugants (12/488) had the phenotype expected of pUB1601::Tn502 resolved products (i.e., Tcr Kmr Hgr Aps). When the respective plasmids were isolated and the sequence of the junction regions was determined (using the tniA502- and merR502-specific primers SP11 and SP21, respectively), the plasmids were found to be bona fide transpositional derivatives in that each Tn502 insertion was flanked by 5-bp DRs of the target DNA sequence. Moreover, the transposon was inserted at different sites in pUB1601, including widely separated sites that shared no obvious sequence similarity, suggestive of a preferred secondary site (Fig. (Fig.1A).1A). In addition, the insertions occurred in the two possible orientations. These various findings showed that par-independent transposition involves random target selection and orientation of insertion and generates unresolved cointegrates as the major product.
The role of tniQ in transposition was assessed using a Tn502 mutant whose TniQ protein, ΔTniQ, lacks a four-amino-acid sequence close to the C terminus (see Materials and Methods). The mutated allele caused a reduction of ≈100-fold in the efficiency of Tn502 transposition to the par+ target plasmid (Table (Table2,2, lines 2 and 6), a finding consistent with the essential role ascribed to tniQ in a study of a tniQ truncation mutant of Tn5053 (22). Only pUB307::Tn502ΔtniQ resolved products were obtained, and three that were sequenced had insertions in the par region (Fig. (Fig.1B).1B). In contrast, the Tn502ΔtniQ element continued to transpose to the par-deleted pUB1601 (Table (Table2,2, line 7), and this time, unresolved plasmid cointegrates were recovered from the transconjugants (60 tested). These findings suggested that tniQ is not required for par-independent transposition of Tn502 but has a role in the par-dependent process.
Tn402 is an exceptional class 1 integron (32, 36) since it is also a functional transposon (20, 26, 41) whose tni module is similar to that of Tn502 and Tn512. A few other class 1 integrons, such as In0, In2, and In5, have IRi and IRt as well as tniA and part of tniB (6). Although the latter elements are not self mobile, we considered that they might transpose if tni502 genes were provided in trans. We therefore tested the transposition of In2 from its native site in Tn21 by first blocking Tn21 transposition (via mutation of tnpA21 in pUB2401) and providing pVS1696 (tni502+) in trans. Transposition occurred with high efficiency to pUB307 (par+) in E. coli, and the plasmids were resolved cointegrates (Table (Table2,2, lines 8 and 9). Using a different experimental system, we also tested transposition of In0 (Sur) from its native plasmid pVS1 (which also carries the functional Tn501 mer transposon) to pUB307. When Tn502 was present in the P. aeruginosa donor, the majority (84%) of Sur transconjugants were Hgs, consistent with the transposition of In0 to pUB307 to form resolved cointegrates (Table (Table3,3, column 4). Further analysis of three of the pUB307::In plasmids from each experiment showed that In2 and In0 were present in the res region and were appropriately oriented and flanked by 5-bp DRs (Fig. (Fig.1B1B).
These findings fulfilled our original expectations; however, on further reflection they were surprising, since In2 and In0 lack the tni-encoded resolution system, that is, they have no res site at which TniR supplied from tni502 can act. Consequently, unresolved cointegrates would have been expected. Only one such cointegrate was obtained, and it represented the sole example of In2 transposition to pUB1601 (Table (Table2,2, line 11) (cointegrate structure was judged by SalI restriction profile and linkage analysis on retransfer; see Materials and Methods). Overall, these findings demonstrated that tni-assisted transposition is possible for some class 1 integrons and that the products are either resolved or unresolved cointegrates depending on whether the target plasmid is par+ (pUB307) or Δpar (pUB1601), an outcome similar to that observed during transposition of the fully functional Tn502.
Since In0 and In2 both lack the res sequence at which TniR can act, it was presumed that the resolution of In-containing cointegrates with pUB307 had devolved to the host recombination (RecA) system. This explanation is not entirely satisfactory, since the high incidence of unresolved pUB1601 cointegrates involving Tn502 (Table (Table2,2, line 3) suggested that RecA-mediated resolution and, more significantly, TniR-mediated resolution are inefficient processes. To gain further insight into the cointegrate formation and cointegrate resolution phases of transposition to pUB307, we monitored the former via the incidence of conductional transfer events (by selecting transconjugants that inherited the Cmr [pACYC184] or Apr [pBR322] donor plasmid) and the latter via the incidence of transposition events (by selecting transconjugants that inherited the Hgr [Tn502] or Smr [In2] element). The experiments were conducted using E. coli and involved a RecA− donor mated with RecA+ and RecA− recipients (Table (Table44).
In the two matings involving Tn502, similar numbers of Hgr transconjugants were obtained, and the transferred plasmids, based on phenotype (Kmr Aps), were resolved products that had formed in the RecA− donor and been transferred as such to the recipients. The transfer of cointegrates (present in Apr transconjugants that were also Kmr Hgr) was ≈100-fold less frequent, and the plasmids were rapidly resolved, irrespective of the RecA status of the recipient (i.e., 12 transconjugants, when outcrossed, yielded pUB307::Tn502 plasmids). This demonstrated that the TniR502 resolution system is efficient and operates independently of the host RecA system. In contrast, the plasmids transferred from the In2+ donor were all cointegrates, as indicated by the similar numbers and full resistance profile of transconjugants from the two recipients irrespective of the selection imposed (Smr or Cmr). This was as expected, since In2 is resolution deficient. The cointegrates in the RecA+ recipient underwent rapid resolution (i.e., 12 Cmr transconjugants, when outcrossed, yielded pUB307::In2 plasmids) whereas those in the RecA− recipient were stable until transferred to a RecA+ host (12 Cmr transconjugants were outcrossed to RecA+ and RecA− hosts). These findings showed that RecA-mediated recombination can efficiently resolve In2-containing cointegrates in the absence of the TniR system.
The single In2 cointegrate with pUB1601 was also passaged in a RecA+ strain (both P. aeruginosa and E. coli) without yielding resolution products detectable by outcross experiments (>1%); this behavior was similar to that of the unresolved pUB1601 cointegrate with Tn502.
Taken together, these various comparisons clearly established that RecA- and TniR-mediated resolutions are both highly efficient when the par locus is present (pUB307) but are not efficient in its absence (pUB1601). We propose that resolution depends on the formation of a res-ParA complex, which can form on pUB307, enabling the two In or Tn copies in the cointegrate to juxtapose (Fig. (Fig.2).2). Such alignment would facilitate RecA-mediated resolution (in the case of In2 and In0) and augment TniR-mediated resolution (in the case of Tn502). The RecA and TniR recombination systems are individually relatively inefficient, since in the absence of the proposed res-ParA complex (the situation that applies to pUB1601) cointegrates remain stable.
The formation of 5-bp flanking DRs is a feature of the Tn and In elements studied here and of Tn5053 (22) and Tn402 (20). However, exceptions exist among naturally occurring elements, for example, Tn402 in R751 (36) and Tn4672 in pUO1 (43) (both IncPβ plasmids) and In5 in pSCH884 (6, 13) (Inc type unknown). In the IncPβ plasmids, the transposon occurs in a region that corresponds to the par locus of RP1 (44), suggesting that insertion occurred originally via the par-dependent process. Our observations that Tn502 can sometimes insert in the vicinity of the res target region in pUB307, rather than within it (Fig. (Fig.1B)1B) (a feature also noted with Tn5053  and Tn402 ), and that it can also transpose to random target sites in pUB1601 (Fig. (Fig.1A)1A) raised the possibility that the absence of 5-bp DRs may be due to transposon-mediated deletion events such as might arise following intramolecular transposition (55) and subsequent loss of DNA via TniR-mediated (or RecA-mediated) resolution.
To test this possibility, we devised a screening procedure to detect rare Tn502-mediated deletion events that involved the par region and the adjacent Tra1 or Tra2 region, each of which encodes genes required for sensitivity to PR4, a donor-specific phage (29, 45). Cultures of PR4-sensitive P. aeruginosa PAO9505 strains carrying either pUB1601traA::Tn502 or pUB307parE::Tn502 were exposed to PR4 (1010 PFU) on selective NA, and the colonies obtained (Hgr, PR4 resistant) were tested for sensitivity to kanamycin. Kms mutants arising from bacteria carrying pUB1601traA::Tn502 were expected to have deletions that encompassed the Kmr locus, IS21, and some trb (Tra2) genes (Fig. (Fig.3);3); two such mutants were isolated, ΩTn502trbL and ΩTn502trbF, and were shown by sequence analysis to have deletions that extend from the IRi-pUB1601 junction into the respective trb genes. In contrast, Kms mutants arising from bacteria carrying pUB307parE::Tn502 included those with deletions that extend from the IRt-pUB307 junction into traG or traI (mutants ΩTn502traG1, ΩTn502traG2, and ΩTn502traI) (Fig. (Fig.3)3) or, alternatively, that extend from the IR border of IS21 into the tra genes. Three mutants of the latter type were detected (data not shown) and were similar to IS21-induced deletions described previously (11).
All five of the Tn502-induced deletion mutants that were studied (Fig. (Fig.3)3) retain on one flank the 5-bp DR formed on initial insertion into traA or parE, whereas on the other flank they have an unrelated 5-bp sequence (i.e., that formed during the subsequent transposition event). These findings provide the first experimental evidence of deletion formation induced by a Tn5053/Tn402-like element and explain the absence of 5-bp DRs at some insertion sites detected in nature. They also provide further evidence that par-independent transposition involves random target selection, since the sites in Tra1 and Tra2 are additional to those shown in Fig. Fig.1A1A.
The ability of Tn502 to transpose to pUB1601 (Δpar) is a distinctive ability not exhibited by Tn512 (Table (Table2)2) or by Tn5053 when it was tested in a comparable experimental system (25, 26). Although these three transposons differ at the sequence level, their TniABQR products are identical or nearly so (95% identity) and hence are not likely to account for the exceptional behavior of Tn502. We suspected that tniM may be involved in transposition to pUB1601 since this gene is present in Tn502 and not in Tn512 (or Tn5053). It is located between the last gene of the mer module, merE, and the tniR gene and is related to a transposition modulator gene, tnpM (83% identity), in Tn21 and Tn501 (18). We therefore studied the effects of a cloned tniM502 gene on the transposition of Tn512 when the former was provided in trans. In the presence of pVS1715 (tniM502+), transposition of Tn512 to pUB307 (par+) was increased about 10-fold (Table (Table2,2, lines 4 and 13). More significantly, transposition to pUB1601 (Δpar) was detected, although it was not as efficient as that observed with Tn502 (Table (Table2,2, lines 5 and 15). The transconjugants all carried pUB1601-pVS1718 unresolved cointegrates (60 tested). SacII restriction profiles of 15 of the plasmids showed that Tn512 could insert into any of the three large SacII fragments of pUB1601; however, whether the two orientations were represented could not be determined (data not shown; see Materials and Methods). These findings implicated tniM in the par-independent transposition process and further reinforced the consistent observation that the products of this process are unresolved cointegrates; tniM may also marginally influence par-dependent transposition.
Tn5053/Tn402-like transposons, including Tn502 and Tn512 studied here, exhibit high-frequency targeted insertion in the par locus of RP1 since the par-encoded components (res-ParA) are essential accessory functions in transposition (25). Tn502 is exceptional in that transposition continues when par is absent (Table (Table2).2). The data presented here establish the existence of a par-independent transposition pathway and define its features. First, par-independent transposition operates at low frequency and involves randomly selected target sites; 14 such dispersed sites were detected in the par deletant pUB1601 (Fig. (Fig.1A1A and and3),3), and Tn502 occurred in either orientation. This contrasts with the high-frequency, targeted, and oriented transposition in the case of par-dependent transposition. Second, the products of par-independent transposition are mainly (>95%) stable, unresolved cointegrates; the resolved cointegrates all contain Tn502 flanked by 5-bp DRs. In par-dependent transposition, resolved cointegrates predominate (Table (Table22).
We attribute the ability of Tn502 to transpose to pUB1601 to its carriage of tniM, since Tn512 also transposed to pUB1601 when tniM502 was provided in trans (Table (Table2).2). The latter events involved at least three target sites and produced unresolved cointegrates, consistent with the par-independent process. Apart from tniM, Tn502 and Tn512 encode nearly identical TniABQR products but have different mer modules; tniM502 occurs between merE and tniR (separated by 633 and 11 bp, respectively) and is apparently part of the mer operon. The effect of TniM, a 12.5-kDa cytoplasmic protein (34), is therefore intriguing and perhaps incidental. It has no distinctive in silico features that might indicate its role; however, a homologue, TnpM (83% identity), has subtle effects (2.5- to 22-fold) on the transposition of Tn21 and Tn501 (18). These elements, like those of the Tn5053 and Tn402 families, move by a two-stage replicative process so that the observed effects of TniM and TnpM on transposition seem significant. Elucidation of their role(s) in transposition awaits further study. Of the tniABQ genes that are essential for stage 1 of transposition (22), we tested only tniQ, confirming its involvement in par-dependent transposition and showing that it is not essential for the par-independent process (Table (Table2).2). The role of TniQ is unknown.
A significant finding of the study concerned cointegrate resolution (stage 2 of transposition) and its management in the two transposition pathways. In par-independent transposition of Tn502 and Tn512, unresolved cointegrates were the predominant product, yet when the res-ParA system was present, they occurred in only <0.5% of transconjugants (Table (Table2,2, lines 2 and 4, and Table Table4).4). The implication of this finding is that the transposon-encoded resolution system res-TniR, hitherto deemed highly efficient (22), is in fact efficient only when the plasmid-determined res-ParA targeting system is present. The epithet “res site hunters” aptly describes the transposons for two reasons: first, because the res-ParA accessory components are crucial in initiating transposition to res (22), and second, because they also play a key role in efficient resolution.
The host (RecA)-mediated recombination system also contributes to cointegrate resolution but is comparatively inefficient since, even with a functional res-TniR system, only ≈1% of transconjugants carried resolved pUB1601::Tn502 plasmids. In contrast, RecA-mediated resolution was paramount when it operated in concert with res-ParA. This was evident in the efficient resolution of In0 and In2 cointegrates with pUB307; both integrons lack the res-TniR system, so that resolution, as shown with In2 (Table (Table4),4), occurs exclusively via RecA. We propose that the binding of ParA to its cognate res regions in the cointegrate is a critical event; it facilitates DNA synapsis, thereby enhancing the likelihood of resolution via either the TniR- or the RecA-mediated system (Fig. (Fig.2).2). In the absence of the res-ParA complex (as is the situation for cointegrates involving pUB1601), such resolution events occur but with reduced efficiency.
In light of the observations made in this study, a model for par-independent transposition can be proposed. The TniAB transposase is presumably essential for the process and initiates transposition at a random target site, though with poor efficiency. TniM either enhances or is essential for transposition, whereas TniQ is dispensable (Table (Table2).2). TniM may serve a general role, such as that of a scaffold or stabilizing protein, in the transposition complex. The distinctive requirements of the par-independent (TniABM) and par-dependent (TniABQ) pathways are reminiscent of Tn7 transposition, which involves essential components (TnsABC) and those dedicated to the targeted insertion (TnsD) and random insertion (TnsE) pathways (33). The striking orientation specificity observed in par-dependent, but not par-independent, transposition possibly implicates TniQ and/or ParA in the former process. TniQ and/or ParA in the transposition complex may occlude one of the possible ligation events, thereby favoring insertion in a single orientation; a similar model has been proposed to account for the oriented insertion of Tn7 (33). In the absence of TniQ and/or ParA, the two possible ligation events can occur. Lastly, because of the inherent inefficiency of the transposon resolution system, par-independent transposition results mainly in unresolved cointegrates.
One of the further consequences of par-independent transposition was the ability of Tn502 to cause deletions in the carrier plasmid. Such deletions extended from the initial target site to random sites in the adjoining DNA so that the transposon lacks recognizable flanking 5-bp DRs, a situation also found among Tn5053/Tn402-like elements located in the res-parA region of most IncPβ plasmids (Fig. (Fig.4).4). Sequence comparisons with the intact res-parA backbone region of pBP136, which lacks any inserts (44), suggest that inheritance of the Tn5053/Tn402-like elements resulted in deletions in the IncPβ plasmids of 0.4 to 2.2 kb (Fig. (Fig.4).4). These naturally arising deletions are small relative to those caused by Tn502 (all >11 kb because of the selection criteria employed); however, all presumably originate from intramolecular transposition or acquisition of a second element. Indeed, a surprising size range of deletions (0.5 to 9.4 kb) can be generated even in a small (13-kb) plasmid, as was found during transposition of an engineered Tn1000 (γδ) transposon (54). IncPβ plasmids with large deletions that extend into the Tra regions, like those induced by Tn502 (Fig. (Fig.3),3), most likely occur in nature; they are not represented among the studied plasmids, because most of these were selected as conjugation-proficient plasmids (39).
Finally, in addition to the par-independent transposition events observed with pUB1601, other rare instances are known, i.e., Tn502 insertions into the tet gene of pBR322 (48) and into the P. aeruginosa PAO chromosome (48) as well as Tn5053 insertions into the E. coli K-12 chromosome (25) and Tn402 insertions into phage λ (41). In the last case, Tn402 transposition may have been aided by a tniM homologue (orf2 of Tn4321) in the R751 carrier plasmid (31, 52). We envisage that tniM homologues in the widespread Tn21/Tn501-like transposons may facilitate nonconventional movement of Tn5053/Tn402-like elements during natural encounters. Such movements can generate variability at random loci or have multifactorial effects, such as DNA deletions and the formation of stable cointegrates. Such events are separate from the dissemination of Tn5053/Tn402-like elements afforded by their affinity for some res target regions, such as that in the par locus of the promiscuous IncP plasmids (12, 49).
We thank two anonymous reviewers for their generous and perspicacious comments on the manuscript, and we thank Ruth M. Hall for providing us with the plasmid pUB2401.
Steve Petrovski was the recipient of a La Trobe University postgraduate award.
Published ahead of print on 29 January 2010.