The Conjugative Transposon Tn5397 Has a Strong Preference for Integration into Its Clostridium difficile Target Site

Abstract

Tn5397 is a conjugative transposon, originally isolated from Clostridium difficile. The Tn5397 transposase TndX is related to the phage-encoded serine integrases and the Clostridium perfringens Tn4451 transposase TnpX. TndX is required for the insertion and excision of the transposon. Tn5397 inserts at one locus, attB_Cd, in C. difficile but at multiple sites in Bacillus subtilis. Apart from a conserved 5′ GA dinucleotide at the recombination site, there appears to be little sequence conservation between the known target sites. To test the target site preference of Tn5397, attB_Cd was introduced into the B. subtilis genome. When Tn5397 was transferred into this strain, 100% of the 50 independent transconjugants tested had Tn5397 inserted into attB_Cd. This experiment was repeated using a 50-bp attB_Cd with no loss of target preference. The mutation of the 5′ GA to 5′ TC in the attB_Cd target site caused a switch in the polarity of insertion of Tn5397, which is consistent with this dinucleotide being at the crossover site and in keeping with the mechanism of other serine recombinases. Tn5397 could also transpose into 50-bp sequences encoding the end joints attL and attR but, surprisingly, could not recombine into the circular joint of Tn5397, attTn. Purified TndX was shown to bind specifically to 50-bp attB_Cd, attL, attR, attTn, and attB_Bs3 with relative binding affinities attTn ≈ attR > attL > attB_Cd > attB_Bs3. We conclude that TndX has a strong preference for attB_Cd over other potential recombination sites in the B. subtilis genome and therefore behaves as a site-specific recombinase.

Conjugative transposons are genetic elements that can mobilize via transposition and conjugation from the genome of a donor to that of a recipient, sometimes across large phylogenetic distances. As they commonly encode antibiotic resistances, they are clinically important (7, 22, 24, 27, 29). Tn5397 is a 21-kb tetracycline resistance-encoding conjugative transposon originally found in Clostridium difficile (13, 20, 21). It can transfer from C. difficile to Bacillus subtilis and back to C. difficile (20). It can also transfer in a model oral biofilm community, indicating that the element is likely able to transfer in natural environments (26).

Tn5397 has been completely sequenced, revealing that it is very closely related to the extensively studied, conjugative transposon Tn916 in the regions concerned with transfer and resistance to tetracycline (25). However, the regions required for transposition in Tn916 and Tn5397 are completely different. The insertion and excision of Tn5397 are dependent on the large serine recombinase TndX, the only Tn5397-encoded protein required for these functions (33, 34). Tn916, on the other hand, requires the tyrosine recombinase (Int) for integration and Int and the accessory factor Xis for excision (15). Although Tn916 can insert into multiple sites in most hosts, it does have preferred integration sites and, in some strains of C. difficile, it has one highly preferred site (33). The clostridia also contain mobilizable transposons such as Tn4451 and its close relatives that integrate and excise via the transposase TnpX (1, 2, 9, 17-19). TnpX and TndX share about 30% amino acid identity. Because Tn4451 transposes at a low frequency, only a limited number of insertion sites have been sequenced. However, this analysis indicates that there is a consensus target for TnpX which resembles the ends of the transposon (9).

Recombination by TndX is believed to occur using a mechanism similar to that described for the resolvase/invertases (22). In the model for excision, the ends of the transposon, attL and attR, synapse with the transposase, which is followed by concerted cleavage of all four DNA strands, forming 2-bp staggered breaks and transient covalent linkages to protein via the recessed 5′ ends. By analogy with the resolvase/invertase mechanism, a 180° rotation of one pair of half sites then occurs to configure the DNA sites into the recombinant format and the DNA backbone is relegated. Thus, the products of excision are an intact (donor) chromosome having suffered a precise deletion of the transposon and a circular form of Tn5397 that is the transposition and conjugational intermediate (22). The recombination site in the circular form of Tn5397, attTn, is the joint where the two ends of the transposon have been ligated (Fig. 1A). When the element transfers to a new host, attTn recombines with the target site attB to establish the transposon in the genome of the recipient. Transposition of Tn5397 has been observed in C. difficile, B. subtilis, Enterococcus faecalis, and Escherichia coli (in a genetically engineered system) (20, 35), and the sequences of the ends of the insertions suggest that the crossover has occurred between 5′ GA in the genome target and 5′ GA in attTn.

FIG. 1.

Sequences of recombination substrates for TndX. (A) Comparison of the sequences of attTn formed within the circular joined intermediate of Tn5397 and the preferred target site in C. difficile, attB_Cd. (B) Sequences of the known target sites for Tn5397 in C. difficile (attB_Cd) and B. subtilis (att_Bs2, att_Bs3, att_Bs4, and att_Bs5a). Black shading indicates identity, dark-gray shading indicates 80% conservation, and light-gray shading indicates 60% conservation. The two asterisks above the sequences 5′ GA indicate the dinucleotides that are inferred to take part in the DNA cleavage and rejoining.

In order to understand the mechanism of transposition by Tn5397, we have investigated the nature of the target sites into which Tn5397 inserts. There is a “hot spot” in C. difficile CD37 where Tn5397 is always found if present, but in B. subtilis CU2189, the element enters the genome at multiple sites and there is no obvious consensus target site in this host apart from the 5′ GA dinucleotide at the center of the target (Fig. 1B) (33). In this work, we wished to resolve this apparent contradiction and further investigate the properties of TndX. We show, using an in vivo assay, that TndX does indeed have a favored target site, the original insertion site from C. difficile (attB_Cd), and that it binds in vitro to this target in a sequence-specific manner. This is the first demonstration of a conjugative transposon that has a preferred insertion site that it will use in two very different hosts while still being able to use other sites if the preferred site is not present.

MATERIALS AND METHODS

Bacterial strains, plasmids, and in vivo methods.

The bacterial strains and plasmids used are listed in Table 1. C. difficile 630 was used as a donor in conjugation. B. subtilis CU2189 and its derivatives, containing the Tn5397 att sites inserted via the vector pSWEET (3), were used as recipients. Plasmids were linearized by cutting with PstI and introduced into competent B. subtilis (14). Conjugations were performed using a filter-mating procedure as described previously (33).

TABLE 1.

Plasmids and strains used in this study

Plasmid or strain	Commentsa	Reference or source
Plasmids
pSWEET	B. subtilis suicide vector for integration at amyE	3
pHWattB-CD	pSWEET containing the 436-bp attB-CD cloned in the BamHI site	This work
pHWattB-CD2	pSWEET containing two copies of attB-CD in direct repeat	This work
pHWattB-CDTC	pSWEET containing attBcd(TC)	This work
pHWattB50	pSWEET containing the 50-bp attB-CD	This work
pHWattL	pSWEET containing attL	This work
pHWattR	pSWEET containing attR	This work
pHWattTn	pSWEET containing attTn	This work
Strains
C. difficile CD630	Tcr EmrC. difficile strain containing Tn5397 and Tn5398	12
B. subtilis CU2189	Tcs recipient strain	6

^aTc, tetracycline; Em, erythromycin.

DNA manipulations.

DNA manipulations were performed according to standard procedures (28).

To construct pHWattB_Cd, the insertion site of Tn5397 in C. difficile (attB_Cd) was amplified by PCR with primers P1 (5397RGI/Bam) (5′-GCGGGATCCGAAAACTGCTTGGATTCAGA-3′) and P2 (5397flank/Bam) (5′-GCGGGATCCGCATATTACGCATCTCATTA-3′), using CD37 genomic DNA as a template. Underlining in the sequences indicates restriction enzyme recognition sites. This fragment was digested with BamHI and cloned into the vector pSWEET (3) which was digested with the same enzyme. pHWattB_Cd₂ was obtained by ligating two copies of the above-mentioned fragment in direct repeat and then ligating it to the vector. To construct pHWattB_Cd(TC) by changing the central GA to TC (top strand) and TC to GA (bottom strand), two rounds of PCRs were carried out. In the first round, primers P1 and P3 (IS/tc-ga) (5′-GTTCTTCCATTACCAGAACTAAAAGGATGAAC-3′) were used to amplify the left part of the target, while primers P4 (IS/ga-tc) (5′-GTTCATCCTTTTAGTTCTGGTAATGGAAGAAC-3′) and P2 were used to amplify the right part of the target in two independent reactions. The products were purified and used as templates in the second-round PCR with primers P1 and P4. The final product was cloned into a pSWEET vector in the same way as that for pHWattB_Cd. To generate the 50-bp C. difficile (attB50) target, two oligonucleotides, IS₅₀-top (5′-GATCCTTTGTATATGTTCATCCTTTTAGTGATGGTAATGGAAGAACATCAAGAG-3′) and IS₅₀-bottom (5′-GATCCTCTTGATGTTCTTCCATTACCATCACTAAAAGGATGAACATATACAAAG-3′), were designed. These were annealed to a double-stranded fragment with BamHI sites at both ends. This was cloned into the pSWEET vector via the BamHI site as described before to generate pHWattB50. To generate attL, two oligonucleotides, attL-top (5′-GATCCTTTGTATATGTTCATCCTTTTAGTGATGGAAATGTACCATCAAGACACCT-3′) and attL-bottom (5′-GATCCAGGTGTCTTGATGGTACATTTCCATCACTAAAAGGATGAACATATACAAA-3′), were designed. These were annealed and cloned into the pSWEET vector via the BamHI site as described before to generate pHWattL. To generate attR, two oligonucleotides, attR-top (5′-GATCCAGTGTCTCATTGATACATTCTCTGATGGTAATGGAAGAACATCAAGAGC-3′) and attR-bottom (5′-GATCCGCTCTTGATGTTCTTCCATTACCATCAGAGAATGTATCAATGAGACACTG-3′), were designed. These were annealed and cloned into the pSWEET vector via the BamHI site as described before to generate pHWattR. To generate attTn, two oligonucleotides, attTn-top (5′-GATCCAGTGTCTCATTGATACATTCTCTGA TGGAAATGTACCATCAAGACACCT-3′) and attTn-bottom (5′-GATCCAGGTGTCTTGATGGTACATTTCCATCAGAGAATGTATCAATGAGACACTG-3′), were designed. These were annealed and cloned into the pSWEET vector via the BamHI site as described before to generate pHWattTn. All constructs were confirmed by PCR and DNA sequencing using primers 5′Bam/SW (5′ GATGTAGCAGTGTTAAGAGAGC-3′) and 3′Bam/SW (5′CGGGCAGACATGGCCTGCCCGG-3′). To generate attB_Bs2, oligonucleotides attBBS2-top (5′-GATCAGGATGTTCATGCACCCATTTCGGGAAGAAAATAATGCCATGCATGCGTT-3′) and attBS2-bottom (5′-GATCAACGCATGCATGGCATTATTTTCTTCCCGAAATGGGTGCATGAACATCCT-3′) were annealed. To generate attB_Bs3, oligonucleotides attBS3-top (5′-GATCCCAGCTCCTGGATATTTGTTGTATGATGGAAACGGGGAAACCCATACAGC-3′) and attBS3-bottom (5′-GATCGCTGTATGGGTTTCCCCGTTTCCATCATACAACAAATATCCAGGAGCTGG-3′) were annealed.

Genomic DNA from C. difficile and B. subtilis was prepared by using the gram-positive DNA isolation kit (Puregene). Plasmid DNA was prepared by using the QIAGEN plasmid mini-prep kit (QIAGEN). The Southern blotting experiments were performed using an ECL direct nucleic acid labeling and detection system (Amersham Pharmacia Biotech) according to the manufacturer's instructions.

To carry out band shift assays, the instruction manual of the digoxigenin (DIG) gel shift kit (Roche) was followed with some modifications. To label the probe, 3.85 pmol of double-stranded DNA was dissolved in distilled water to make the final volume of 10 μl and then the following reagents were added: 4 μl of 5× labeling buffer (1 M potassium cacodylate, 0.125 M Tris-HCl, 1.25 mg/ml bovine serum albumin, pH 6.6 [25°C]), 4 μl of 25 mM CoCl₂, 1 μl of 1 mM DIG-11-ddUTP (DIG-ddUTP), and 1 μl of terminal transferase (50 units/μl). The mixture was incubated at 37°C for 15 min before being placed on ice. The reaction was stopped by adding 2 μl of 0.2 M EDTA (pH 8.0). In the band shift reaction, purified TndX was added to a mixture containing 200 mM potassium cacodylate, 25 mM Tris-HCl, 0.250 mg/ml bovine serum albumin, 1 μg of poly(dI-dC), 0.1 μg of poly l-lysine, 0.310 pmol of DIG-labeled probe and water to make up to 20 μl. Up to 100× molar ratio of specific competitor was added when required. The reaction mixture was incubated at 4°C for 20 min before being mixed with 5 μl of loading buffer (0.25× Tris-borate-EDTA buffer, 60%; glycerol, 40%; bromophenol blue, 0.2% [wt/vol]) and analyzed on a 6% acrylamide gel. After the electrophoresis, the DNA was transferred to a nylon membrane by contact blotting. The DNA was fixed to the membrane by UV cross-linking and then detected by chemiluminescent detection.

The binding affinity of TndX protein for each of the above-mentioned DNA fragments was determined quantitatively with Scion Image analyzing software (Scion Co.).

Purification of TndX.

TndX was expressed in E. coli as a fusion to the C terminus of a 10-His tag as described before (35). The cell pellets were dissolved, and the soluble fraction was obtained by using Bug Buster master mix (Novagen). TndX was purified by using the Ni-resin (Clontech Laboratories, Inc.). The manufacturer's protocols were followed. Protease inhibitor cocktails (for use in the purification of poly-His-tagged proteins [Sigma]) were added to the supernatant. The amount used was 1 ml cocktail per 20 g of E. coli cell extract.

RESULTS

The target from C. difficile is a hot spot for Tn5397 insertion in B. subtilis.

Our previous work showed that Tn5397 inserts in only a single target site in C. difficile but in multiple sites in B. subtilis (20). All of the targets analyzed have a central 5′ GA dinucleotide in common. In this work, we have employed a vector, pSWEET (3), to integrate target sites into the amyE locus of the B. subtilis genome. The B. subtilis strains containing the targets were then used as recipients in filter-mating experiments with C. difficile 630 as a Tn5397 donor. The Tn5397-containing, tetracycline-resistant transconjugants were then analyzed for the presence of Tn5397 in the target sequence. In the first experiment, a 436-bp fragment containing the C. difficile target attB_Cd was amplified by PCR and inserted into pSWEET to form pHWattB_Cd, which was then introduced into B. subtilis to create B. subtilis::attB_Cd. This strain was used as a recipient in a mating experiment with C. difficile 630.

Analysis of the genomic DNA of 28 independent, tetracycline-resistant transconjugants by Southern blotting and DNA sequencing showed that, in every case, Tn5397 had inserted into attB_Cd (Fig. 2A and E, lane 1). No insertions into any other B. subtilis targets were found. Another construct that contained two copies of the C. difficile target in direct repeat, B. subtilis::attB_Cd2 (derived by transformation of B. subtilis with pHWattB_Cd2), was also made. When B. subtilis::attB_Cd2 was used as a recipient, Tn5397 inserted into only one of the two sites (Fig. 2B and E, lanes 2 and 3). Analysis of 22 independent transconjugants showed that the element has almost equal chances of insertion into one site or the other. In total, of the 50 independent transconjugants obtained from mating experiments with C. difficile and B. subtilis containing attB_Cd, all had Tn5397 inserted in the C. difficile target. As a control, pSWEET lacking any insert was introduced into the B. subtilis target. When this strain was used as a recipient, we found Tn5397 inserted in eight different sites when nine independent transconjugants were analyzed (Fig. 2F). This result is similar to those obtained when B. subtilis cells without the integrated vector were used as recipients (33).

FIG. 2.

Analysis of the target site choices of Tn5397 in B. subtilis. (A to D) Schematic showing the insertion of Tn5397 into different target sites used here. The annotations are as follows: the light-blue shaded boxes represent the target inserted into the B. subtilis chromosome and the 5′GA dinucleotide that is thought to be the crossover region. The thin bands represent the flanking vector sequence. The expected product(s) after the conjugation and insertion of Tn5397 (dark blue), flanked by attL and attR, is shown underneath. The tndX gene was used as a probe in Southern blotting experiments with HindIII-digested chromosomal DNA. HindIII has one recognition site in tndX. If Tn5397 had inserted into any of these targets, two bands would hybridize to the probe (as shown). The different target sites used were attB_Cd (A), attB_Cd2 (B), attB_Cd(TC) (C), and the 50-bp attB target, attB_Cd50 (D). Panels E and F show results from Southern blotting of transconjugants containing various insertions of Tn5397 at different target sites. The DNA was digested with HindIII and probed with tndX. For clarity, only one clone from each cross is shown here. (E) Lane 1 shows DNA from an insertion into attB_Cd (see panel A); lanes 2 and 3 show DNA from an insertion into attB_Cd (see panel B); lane 4 shows DNA from an insertion into attB_Cd(CT) (lane 5) DNA in which the target site had been reduced to 50 bp (see panel D). (F) Southern blot analysis of B. subtilis DNA (containing unmodified pSWEET integrated into the chromosome) from six transconjugants containing insertions of Tn5397 isolated from independent mating experiments. If the element were integrating into the genome at random, one would expect to see two hybridizing fragments, one of 5,425 bp internal to Tn5397 and one of variable size representing the junction region.

The 5′ GA dinucleotide in the C. difficile target determines the polarity of Tn5397 insertion.

We have previously shown that Tn5397 insertions are always flanked by 5′ GA. Moreover, attTn and the target sites always contain 5′ GA. It is proposed that by analogy to other serine recombinases, this dinucleotide forms the crossover site. It has recently been shown for two phage-encoded serine integrases that this dinucleotide is the sole determinant of the polarity of the recombination site (10, 11, 31). To determine whether this is also the case for TndX and as a test of the putative mechanism of this transposase, the 5′ GA in the top strand of attB_Cd was changed to 5′ TC to form attB_Cd(TC). When B. subtilis cells containing attB_Cd(TC) were used as recipients, 18 of the 19 independent transconjugants analyzed had Tn5397 inserted in this site. Further analysis by PCR and sequencing showed that in the 18 transconjugants where Tn5397 was inserted into attB_Cd(TC), all 18 were in the opposite orientation relative to Tn5397 inserted into attB_Cd (Fig. 2C and E, lane 4). Only one transconjugant contained an unoccupied attB_Cd(TC), and we presume that Tn5397 inserted elsewhere in the B. subtilis genome. These results indicate that the central 5′ GA dinucleotide is responsible solely for the polarity of attB_Cd and that attB_Cd(TC) is still a preferred target.

In order to determine whether Tn5397 was capable of excising from the att_Cd(TC) target, PCR for the circular form of Tn5397 in these strains was performed (33). A PCR product of the appropriate size was produced in all of the transconjugants tested (results not shown), indicating that the element is capable of excision from this site. As excision appears to be a requirement for the transfer of Tn5397, (34) it is very likely that these strains will be able to act as conjugal donors of the element.

attB_Cd could be reduced to 50 bp without loss of activity.

The recombination sites employed by the phage-encoded serine integrases are approximately 50 bp in length (5, 10, 23, 32). The protected sites by TnpX are approximately 68 to 93 bp (1). We decided to test whether a 50-bp attB_Cd was still a preferred target site for Tn5397. Oligonucleotides encoding attB_Cd with the crossover sequence, 5′ GA, located at the center were annealed and then ligated into pSWEET to generate pHWattB_Cd50, which was then transformed into B. subtilis. B. subtilis::attB_Cd50 transconjugants containing Tn5397 were analyzed by Southern blotting, PCR, and DNA sequencing, which showed that all 15 transconjugants analyzed had Tn5397 inserted into attB_Cd50 via the 5′ GA dinucleotide at the center of the 50-bp target, resulting in an element flanked by 5′ GA dinucleotides (Fig. 2D and E, lane 5). Thus, the 50-bp attB_Cd has retained its activity as a preferred target for insertion of Tn5397.

Tn5397 can integrate into attL and attR but not into attTn.

The left (attL) or the right (attR) junction of Tn5397 inserted into attB_Cd each contain one-half of the attB_Cd preferred target. TndX can cause excision by recombination between attL and attR in the absence of any further Tn5397 gene products (35). This contrasts with the phage-encoded integrases that are highly directional, being able to cause only attP-attB recombination in the absence of any other proteins (32). It therefore seemed likely that TndX could recombine any combination of att sites, including attTn-attL and attTn-attR. To test whether attL and attR could be used as targets for Tn5397 insertion and whether they were preferred over other B. subtilis target sequences, pHWattL and pHWattR were constructed and the att sites were introduced into B. subtilis to form B. subtilis::attL and B. subtilis::attR, respectively. These 50-bp sites comprised, for attL, 25 bp of the left side of attB_Cd and 25 bp of the right side of attTn and, for attR, 25 bp of the left side of attTn and 25 bp of the right side of attB_Cd. Transconjugants of B. subtilis::attL and B. subtilis::attR containing Tn5397 were analyzed by Southern blotting and PCR and showed that Tn5397 inserted into the attL in four out of five transconjugants and into attR in three out of five transconjugants (Fig. 3A). DNA sequencing of the PCR products generated from transconjugants containing Tn5397 inserted at attL or attR demonstrated that insertion had occurred at the 5′ GA crossover dinucleotide. A PCR was also performed for the empty target site; a product was obtained only when the transposon had inserted into a site other than attL or attR (Fig. 3B). The insertion of Tn5397 into attL or attR was unlikely to be due to homologous recombination, as there was only 25 bp of identity between attL or attR and Tn5397. Moreover, homologous recombination was not observed in experiments with two attTn sites (see below). Thus, attL and attR sites were recognized as preferred targets by the transposon.

FIG. 3.

Southern blot and PCR analyses of B. subtilis transconjugants containing either an attL or an attR site. DNA was prepared from independent transconjugants, digested with HindIII, blotted, and probed with tndX. If the element were integrating into the genome at random, one would expect to see two hybridizing fragments, one of 5,425 bp internal to Tn5397 and one of variable size representing the junction region, whereas if Tn5397 had integrated specifically into either attL or attR, bands of 5,425 bp and 2,387 bp should be observed. (A) Southern blots of five transconjugants from B. subtilis::attL (lanes 1 to 5) and B. subtilis::attR (lanes 6 to 10) (see the text for more details). If the targets are empty, they will be amplified by PCR using primers 5′Bam/SW and 3′Bam/SW, but if they are occupied by Tn5397, the 22-kb element will not generate a product under the conditions used. DNAs corresponding to the transconjugants used in panel A were subjected to PCR and run in agarose gels. The lane marked “+” is a positive control, where pHWattL or pHWattR was used as a template for PCR, and the lane marked “−” was a negative control with no template.

All of the recombination reactions described here use attTn as one of the substrates, as this is the circular joint formed after the excision of Tn5397. attB_Cd and attTn share some sequence identity (Fig. 1A). If TndX is merely seeking out attachment sites that resemble the sequence of attTn, then attTn should be a good target sequence for integration. Oligonucleotides encoding a 50-bp attTn sequence containing the joined ends of Tn5397 as occurs in the circular form were ligated into pSWEET to form pHWattTn, and this was introduced into B. subtilis. Tn5397 was transferred into BS::attTn, and the transconjugants were analyzed as described before. Southern blot analysis of 19 independent transconjugants showed that Tn5397 had inserted into different sites; none of the Tn5397 transconjugants contained a simple insertion into attTn (Fig. 4), as judged by PCR analysis for the empty target site (results not shown).

FIG. 4.

Southern blot analysis of B. subtilis transconjugants containing attTn. B. subtilis cells containing attTn were used as recipients in matings with C. difficile containing Tn5397. DNA was prepared from transconjugants and digested with HindIII and probed with tndX. If the element were integrating into the attTn site, one would expect to see two hybridizing fragments, one of 5,425 bp internal to Tn5397 (Fig. 1 and 2) and one of 2,387 bp (or 603 bp, depending the orientation of the transposon in the genome), whereas insertion elsewhere in the genome would yield the 5,425-bp band and one of variable size representing the junction region. The results of Southern blotting of eight transconjugants are shown (see the text for more details).

When a B. subtilis strain containing a 1.1-kb attTn was used as a recipient, no transconjugants were ever obtained in 30 independent filter-mating experiments. This indicates not only that this region cannot be used as a target but also that it appears to prevent other target sites in the B. subtilis genome from being used.

TndX binds sequence specifically to 50-bp attB_Cd, attL, attR, attTn, and attB_Bs, but with different affinities.

The data presented above strongly suggest that TndX is acting as a site-specific recombinase, targeting Tn5397 to a specific site in the C. difficile genome. If this is the case, then TndX will bind to attB_Cd in a sequence-specific manner and with an affinity not too dissimilar to those for its other substrates (attTn, attL, and attRs). A C-terminal His-tagged derivative of TndX was purified from E. coli and used in gel shift assays (Fig. 5) (35). The DNA probes were DIG labeled. Under the conditions used, TndX bound with the highest affinities to attTn and attR. In each case, even with the lowest concentration of protein used (0.275 μM), nearly the entire probe was shifted to form a complex with slower mobility (Fig. 5). With the attL probe, approximately half of the probe was bound at 0.275 μM TndX, suggesting a slightly lower relative affinity, and with the attB_Cd probe, about 50% of the probe was bound in the presence of 1.1 μM TndX. TndX bound to all of these probes in the presence of competitor DNA, and binding was greatly diminished if an excess of unlabeled probe was added. In contrast, TndX bound very poorly to attB_Bs3, with only a very small amount of bound probe observed in the presence of 1.35 μM TndX. These data confirm that TndX binds to all its substrate recombination sites, attTn, attL, attR, and attB_Cd, in a sequence-specific manner and that there is a binding preference, attTn ≈ attR > attL > attB_Cd > attB_Bs3.

FIG. 5.

TndX binding to its recombination sites. Purified TndX with a C-terminal His tag was used in gel shift assays to determine relative binding affinities to attL (A), attR (B), attTn (C), and attB_Cd (D) and the pseudo-attB site from B. subtilis, attB_Bs3 (E). In all panels, the white arrowhead indicates the position of the free probe and the black arrowhead is the position of the major TndX-probe complex. (A to D) Increasing concentrations of TndX were added to the binding reactions to generate final concentrations of 0, 0.28, 0.55, 0.83, and 1.1 μM (lanes 1 to 5, respectively). Lane 6 contains the same binding reaction as does lane 5 except that 10-fold more unlabeled probe was added as a specific competitor. (E) The concentration of TndX added to each binding reaction was 0.17, 0.34, 0.68, and 1.35 μM (lanes 1 to 5, respectively).

DISCUSSION

In this paper, we set out to determine whether Tn5397 has a specific target or whether it integrates into random targets when it transfers from host to host. The element does indeed have a preferred target site, attB_Cd, a sequence originating from C. difficile. In the absence of this site, however, Tn5397 can insert into alternative target sites, all of which have the crossover dinucleotide sequence, 5′ GA, and some of which have limited sequence similarity with attB_Cd (Fig. 1). The transfer of Tn5397 from C. difficile to B. subtilis is a rare event, and in each filter-mating experiment, only a few transconjugants were obtained. Thus, the relative frequency of use of these targets could not be determined. However, we observed that attB_Cd is the preferred insertion site for at least 50 independent mating experiments, which is indicative of site-specific insertion. Although we have not demonstrated the minimum sequence required for the use of attB_Cd by TndX, 50 bp was sufficient to maintain site-specific insertion by Tn5397 in B. subtilis. This is comparable to the attachment sites used by the phage-encoded serine integrases, such as φC31, Bxb1, and φRV1 (4, 11, 16, 30, 31, 32). The ability to switch the polarity of insertion of Tn5397 by changing the crossover sequence in attB_Cd to 5′ TC is also strongly reminiscent of the phage integrases and implies that the same mechanism of recombination is employed by TndX. These data support the idea that during integration, the 5′ GA dinucleotide in the attB_Cd is cleaved at the 3′ end and exchanged with the similarly cleaved attTn site to generate the attL and attR sequences. The reverse occurs for excision. For the preferred site attB_Cd, the sequences that flank the 5′ GA dinucleotide are probably recognized in a sequence-specific manner by TndX.

The experiments described here indicate similarities between the properties of the phage-encoded serine integrases and the transposase TndX. These similarities extend to the use of a preferred target site for integration. In vitro binding experiments with the TndX recombination sites indicated an order of preference, attTn ≈ attR > attL > attB_Cd, and that binding was sequence specific. Although the apparent affinity for attB_Cd was the lowest out of these four substrates, the affinity was only about fourfold less than that for attL. When this specific target site is not present in the genome, however, there is sufficient flexibility in target site recognition such that other, pseudo-attB sites can be used, as is the case for B. subtilis, e.g., attB_Bs3. The use of pseudo-attB sites by the phage-encoded serine integrases has also been shown to occur (8). Binding assays with TndX indicated very low affinity for one of its pseudo-attB sites, attB_Bs3 (50% binding requires greater than 1.35 μM TndX), and this is consistent with it being occupied only in the absence of attB_Cd. These data are in contrast to the results obtained with the related enzyme TnpX, responsible for the integration and excision of the mobilizable clostridial transposon Tn4451 that had at least a 40-fold-higher affinity for the ends of the element than it did for its targets (1). It is possible that Tn4451 also has a preferred target that has not yet been identified.

Despite the similarities between the properties of TndX with the phage integrases, TndX is fundamentally different from the phage integrases as it alone can catalyze both excision and integration. We therefore expected TndX to be less selective in its use of different combinations of att sites for recombination. In fact, TndX could utilize attL and attR as targets for Tn5397 insertion. In these assays, insertions into other B. subtilis target sites were observed (one out of five for attL and two out of five for attR), suggesting that attL and attR are not as highly preferred as attB_Cd but are still preferred. The use of attL and attR by Tn5397 would imply that tandem insertions of Tn5397 may be obtained occasionally. In support of this idea, we observed that when Tn5397 is transferred to C. difficile CD37, two copies of the element are indeed found at specific sites (20). To our surprise, we observed that attTn was not used as a target for Tn5397 integration and appeared to be avoided. TndX bound to attTn and attR with the highest affinities, yet no recombination was observed between attTn and attTn and recombination between attTn and attR was arguably less preferred than that between attTn and attB_Cd. Therefore, it is not the strength of binding that determines the frequency of recombination. Instead, we propose that it is the conformation adopted by TndX when bound to its recombination sites that determines whether recombination occurs. This inability to recombine attTn-attTn is reminiscent of the phage integrases that also do not recombine attP-attP (or attB-attB, etc.) and this property may reflect a fundamental feature of the mechanism of recombination by the large serine recombinases.