Insertion Sequence IS26 Reorganizes Plasmids in Clinically Isolated Multidrug-Resistant Bacteria by Replicative Transposition.

UNLABELLED: Carbapenemase-producing Enterobacteriaceae (CPE), which are resistant to most or all known antibiotics, constitute a global threat to public health. Transposable elements are often associated with antibiotic resistance determinants, suggesting a role in the emergence of resistance. One insertion sequence, IS26, is frequently associated with resistance determinants, but its role remains unclear. We have analyzed the genomic contexts of 70 IS26 copies in several clinical and surveillance CPE isolates from the National Institutes of Health Clinical Center. We used target site duplications and their patterns as guides and found that a large fraction of plasmid reorganizations result from IS26 replicative transpositions, including replicon fusions, DNA inversions, and deletions. Replicative transposition could also be inferred for transposon Tn4401, which harbors the carbapenemase blaKPC gene. Thus, replicative transposition is important in the ongoing reorganization of plasmids carrying multidrug-resistant determinants, an observation that carries substantial clinical and epidemiological implications for understanding how such extreme drug resistance phenotypes evolve. IMPORTANCE: Although IS26 is frequently reported to reside in resistance plasmids of clinical isolates, the characteristic hallmark of transposition, target site duplication (TSD), is generally not observed, raising questions about the mode of transposition for IS26. The previous observation of cointegrate formation during transposition implies that IS26 transposes via a replicative mechanism. The other possible outcome of replicative transposition is DNA inversion or deletion, when transposition occurs intramolecularly, and this would also generate a specific TSD pattern that might also serve as supporting evidence for the transposition mechanism. The numerous examples we present here demonstrate that replicative transposition, used by many mobile elements (including IS26 and Tn4401), is prevalent in the plasmids of clinical isolates and results in significant plasmid reorganization. This study also provides a method to trace the evolution of resistance plasmids based on TSD patterns.

INTRODUCTION

Members of the family Enterobacteriaceae are major causes of severe and often lethal hospital-acquired infections. In the past two decades, carbapenemase-producing Enterobacteriaceae (CPE) have appeared that are resistant to most or all clinically available antibiotics, including carbapenems, which are often considered the antibiotics of last resort (1). It has been known for over 40 years that antibiotic resistance determinants are often associated with mobile genetic elements (2), a phenomenon that has been reinforced by genome sequencing of clinical isolates. For instance, blaCTX, the coding gene of an extended-spectrum β-lactamase (ESBL), is often located downstream of insertion sequence (IS) ISEcp1 (3–6), and among carbapenemase-coding genes, blaKPC is typically carried by the Tn3 family transposon Tn4401 (7, 8), whereas blaNDM is frequently located downstream of ISAba125 (9, 10).

Mobile elements are also associated with formation of resistance gene clusters in which different determinants that give rise to a multidrug-resistant phenotype are found in close proximity. For example, specific recombination platforms, integrons (11), can recruit cassettes carrying additional resistance genes via site-specific recombination (12–14). These are, in turn, often associated with transposons and ISCRs (insertion sequences with a common region), a class of mobile elements that has been implicated in capturing genes or gene fragments (15). Mobile elements can also integrate control elements such as promoters upstream of resistance determinants, leading to increased expression (16–18). The horizontal dissemination of resistance between bacteria can occur via conjugative plasmids and integrative conjugative elements (ICEs) (19, 20).

ISs are the simplest autonomous mobile elements in bacterial genomes. They are typically composed of short terminal sequences, often arranged as inverted repeats (terminal inverted repeats [TIRs]) at their ends and at least one open reading frame (ORF) that encodes the transposase, the only protein essential for mobility (21). The TIRs serve as specific recognition sites for the transposase, the enzyme that carries out the strand cleavage and transfer reactions which confer mobility. While the average IS DNA content of bacterial chromosomes is below 3%, in plasmids larger than 20 kb it is between 5 and 15%, with one extreme case exceeding 40% (22). The 4,240 different currently identified ISs can be divided into 30 families based on their overall genetic organization and specific genetic signatures (https://www-is.biotoul.fr/) (23). ISs within one family are assumed to use the same or a similar transposition mechanism to one another, although only a subset of IS families has been investigated experimentally, and the mechanisms of many remain poorly understood.

A vast majority of presently identified ISs encode transposases whose catalytic domain is homologous to those of RNase H and members of the retroviral integrase superfamily (24, 25), the so-called DDE transposases. Some ISs transpose by a cut-and-paste mechanism by excising from one genetic location and integrating into another (21). Others transpose by using a copy-in (26) or replicative transposition (21) pathway that requires extensive DNA replication and typically generates products known as cointegrates, which have a second copy of the IS at the target site while the original copy is left intact (27, 28).

During replicative transposition, a branched or forked “Shapiro intermediate” is formed between the donor IS and the target DNA (Fig. 1a) (28). When the IS and the target site are in two different replicons (intermolecular transposition) (Fig. 1a), subsequent DNA replication at the intermediate branch fuses the two replicons, duplicating both the transposable element and a short nucleotide sequence (target site duplication [TSD]) flanking the insertion site. The TSD becomes two new direct repeats (Fig. 1a, red arrows). Cointegrates can be subsequently resolved by recombination between the two directly repeated ISs, either by a dedicated resolvase or a host-mediated process, to regenerate the two replicons, each with a single IS copy, and the TSDs will flank the copy in the target replicon.

FIG 1 

Schema of replicative transposition by an insertion sequence. (a) Intermolecular transposition. Cleavage at both TIRs of the IS results in nicks on both strands, generating 3′-OH groups which attack the target site, leading to the formation of a Shapiro intermediate. DNA replication generates the cointegrate containing a duplication of the IS and the target site. The cointegrate can be subsequently resolved into a plasmid identical to the original donor plasmid and a modified target plasmid carrying a copy of the IS flanked by TSDs arranged as direct repeats. (b) Intramolecular transposition. When an IS targets a target site in the same replicon, cleavages at both TIRs generate 3′-OH groups that can either attack the target site on the same strand (cis) or the opposite strand (trans). In the cis pathway, DNA between the IS and target site (dashed lines) becomes circularized and contains one IS copy and target site. In the trans pathway, DNA between IS and target site is instead inverted (“a b” becomes “b a”), bracketed by the original IS and a new copy in an inverted orientation. The target site is also duplicated but in inverted orientation, and each TSD is associated with one IS copy. (c) Intermolecular transposition of a composite transposon. A composite transposon consisting of two ISs (i.e., the DNA between TSDs 1 and 4 in the donor plasmid; purple) forms a cointegrate when either of the two flanking IS copies (left or right) transposes into a target plasmid (dashed lines), in which both the IS and the target site are duplicated. Resolution via homologous recombination yields a target plasmid carrying the whole composite transposon. IS, dark yellow rectangle; left and right TIRs, open and filled triangles; red arrows, targets for transposition; black arrows, potential TSDs from previous transposition events; gray oval or rectangle, origin of replication; filled green circle, 3′-OH. Arrowheads indicate orientation, and different numbers represent different sequences.

When the donor IS and target sites are present in the same replicon (intramolecular transposition) (Fig. 1b), there are two possible and very different outcomes. After the initial DNA cleavage steps, the liberated 3′-OH groups at the two IS ends can attack the target DNA in two different orientations, i.e., the 3′-OH of the “top” strand of the IS can attack either the “top” or “bottom” strand at the target site (27).

Attack in one orientation (Fig. 1b, cis) results in the deletion of the DNA between the IS and the target site, leaving only one copy of the IS. A circular product is also created that contains a second IS copy, but this will survive only if it contains an origin of replication. Neither of the resulting IS copies is flanked by TSDs: these are distributed between the two products of deletion.

Attack in the other orientation (Fig. 1b, trans) results in IS duplication in an inverted configuration, and the DNA segment between the original IS and the target site is inverted; in addition, the target site is duplicated. However, as the TSD is at one end of the original IS and the opposite end of the copy, again, neither of the two resulting IS copies is flanked by TSDs. In principle, cis and trans attacks are equally likely, given the pseudo twofold symmetry of double-stranded DNA (dsDNA) and the general lack of target sequence specificity of ISs.

Two identical IS copies within one replicon can generate a composite transposon in which cleavage at two outer ends can mobilize both ISs and the DNA segment between them. If the IS uses a replicative transposition pathway, an alternative parsimonious way could be used to mobilize the composite transposon. Transposition of one of the ISs can generate a cointegrate with three IS copies (cointegrate, Fig. 1c). Resolution of this cointegrate between two IS copies (arrows, Fig. 1c) will result in the transfer of the composite transposon into the target replicon (21). This pathway has been shown experimentally in Escherichia coli for IS26 (38).

In 2011, the National Institutes of Health Clinical Center (NIHCC) experienced an outbreak of carbapenem-resistant Klebsiella pneumoniae infections initially involving 18 patients with a high fatality rate. Sequencing of all isolates recovered during the 2011 outbreak demonstrated that the outbreak was clonal (29). The NIHCC Hospital Epidemiology and Microbiology Services now routinely performs surveillance cultures of patients and the hospital environment, and other CPE isolates have since been recovered.

Here, we have analyzed CPE genomes from the NIHCC collection isolated between 2011 and the end of 2013, focusing largely on IS content and particularly on replicative transposons. Although the possible outcomes of replicative transposition have been described (27, 28), no systematic investigation on the genomic level has been undertaken to understand its impact in clinically relevant circumstances. Our analysis of the distinct patterns of sequences that flank these mobile elements allows us to propose mechanisms and pathways that explain several changes due to transposition events, and we demonstrate that IS26 in particular actively remodels resistance plasmids by both inter- and intramolecular replicative transposition.

RESULTS

IS annotations of the genomes of carbapenem-resistant Klebsiella pneumoniae isolates from the 2011 NIH outbreak.

Since the 2011 outbreak, the NIHCC has routinely cultured CPE isolates from patients and from the clinical environment. In the subsequent 2 years, a number of CPE isolates, including Klebsiella pneumoniae, Klebsiella oxytoca, Enterobacter cloacae complex, Citrobacter freundii complex, Escherichia coli, and Pantoea spp. have been isolated from these clinical and surveillance cultures. The genomes of some of these strains have been sequenced at a very high coverage using Pacific Biosciences RS II single-molecule real-time (SMRT) technology which generates 14- to 40-kb-long reads (30). We therefore took advantage of these sequence data to investigate the diversity of ISs present in clinical isolates by first performing IS annotation of one representative isolate, KPNIH1 (29), from the clonal NIHCC K. pneumoniae outbreak. The assembled genome of this isolate contains one chromosome and three plasmids of >15 kb. Carbapenemase resistance is conferred by blaKPC which is carried by Tn4401 in plasmid pKpQIL, as previously observed in isolates from other hospitals (31, 32).

Whole-genome IS annotation of KPNIH1 using ISsaga (http://issaga.biotoul.fr/) (33) revealed 29 different ISs belonging to 12 different IS families (Fig. 2). Some of these are present either as a single copy or only in a partial form. In contrast, the ISKpn1, ISKpn26, IS903B, and IS26 representatives are present in numerous copies, suggesting that they are currently active. The locations of these ISs vary: all ISKpn1 and most ISKpn26 copies are located in the chromosome, whereas most IS26 copies are located in plasmids.

FIG 2 

IS distribution in Klebsiella pneumoniae strain KPNIH1. IS copy numbers in the three plasmids and single chromosome are shown in red and blue, respectively. Pale color, partial IS; dark color, complete IS. IS family classifications (in green) were determined using ISfinder (https://www-is.biotoul.fr/).

The four IS families with multiple copies present in KPNIH1 are all reported to generate TSDs (ISfinder tool, at https://www-is.biotoul.fr/). TSDs are genomic signatures of transposition events carried out by DDE transposases, and their length is a characteristic property of the transposon, reflecting the particular structure of the transpososome (the molecular assembly composed of transposase and transposon ends that catalyzes transposition) (34). Among the IS families with multiple representatives, most chromosome-located ISs had directly abutted TSDs consistent with ISfinder records, yet those located on plasmids rarely had them (see Table S1 in the supplemental material). Interestingly, none of the six full IS26 copies, including two found on the chromosome, had directly flanking TSDs. Lack of TSDs has also been recently reported in another U.S. isolate of K. pneumoniae, ATCC BAA-2146, which carries a blaNDM carbapenemase gene and in which none of the 11 genomic IS26 copies has TSDs (35). Although the lack of TSDs was attributed by those authors to homologous recombination following IS26 transposition—and this is certainly possible—we also suspect that, since IS26 transposes via a cointegrating mechanism, its movement might in itself generate copies without flanking TSDs.

Transposition mechanism of IS26.

The evidence that IS26 transposes by using replicative transposition and cointegrate formation is compelling. IS26, a member of the IS6 insertion sequence family, is 820 bp long and has 14 bp TIRs (36). It is a component of Tn2680, in which two IS26 copies bracket a 5-kb DNA segment, including a kanamycin resistance cassette on the Proteus vulgaris R plasmid Rtsl (36). Tn2680 transposition was observed from Rtsl into the bacteriophage P1 genome in Escherichia coli, indicating that the two IS26 copies were responsible for mobility (37). Movement of IS26 from bacteriophage P1 to a pBR322 plasmid resulted in a cointegrate between the entire circular prophage genome and the target plasmid and generated an additional IS26 copy. An 8-bp TSD was observed at the junction of IS26 and the pBR322 insertion site. RecA was required to resolve the cointegrate to generate pBR322 containing the insertion (36, 38). Another study involving IS15, the product of insertion of an IS26-like element into itself, also demonstrated that a cointegrate is the transposition product and that resolution requires RecA-initiated homologous recombination (39).

IS26 in the genomes of Enterobacteriaceae strains isolated in the NIHCC from 2011 to 2013.

When we extended our ISsaga analysis of NIHCC clinical strains by annotating eight additional fully assembled genomes (for a total of 36 replicons [see Table S2 in the supplemental material]), IS26 was observed to be the most predominant IS in the plasmids of these strains without any apparent species specificity. Although these eight genomes contained about 70 IS26 copies, only 6 were chromosomal. Furthermore, not a single IS26 copy exhibited flanking 8-bp TSDs (Table 1). On the other hand, we observed several identical flanking sequences at a single end of two different IS26 copies (highlighted with the same color label in Table 1). These were identified not only in IS copies within the same replicon but also in different replicons. These identical sequences are therefore “tracers” and allow us to follow the replicative transposition of IS26 and its consequences in the form of DNA rearrangements within these plasmids.

Table 1 

The 8-bp flanking sequences of IS26 in nine genomes of Enterobacteriaceae

Identical sequences are shown in the same color pattern. A blank implies a partial IS26 copy.

Intermolecular transposition events in the analyzed NIH Clinical Center strains.

When we examined the pairs of identical 8-bp flanking sequences of different IS26 copies in their genomic contexts, it became clear that these sequences provide evidence for replicative transposition. For example, we detected the hallmark of replicative transposition—cointegrate formation—when we compared pAAC154-a50 from the KPNIH1 strain and pKPN-819 from strain KPNIH24 from a patient hospitalized in NIHCC in July 2012. pAAC154-a50 is a small 15-kb plasmid that carries an ISSwi1 transposon derivative of the Tn3 family, ISSwi1-m2, which appears to be a descendant of ISSwi1 (Fig. 3a). ISSwi1 encodes a transposase (tnpA), a resolvase (tnpR), and a TEM beta-lactamase (blaTEM-1) and is present in plasmids pZM-3 from Salmonella enterica serovar Wien (40) and pNJST258N4 from K. pneumoniae NJST258_1 (41). ISSwi1-m2 most likely arose from another ISSwi1 derivative, ISSwi1-m1 (also called Tn1331 [42-44]), which is present in plasmid pNJST258N5 of K. pneumoniae NJST258_1 due to an IS26 insertion into blaTEM-1 of parental ISSwi1 and adjacent DNA deletion.

FIG 3 

Examples of intermolecular transposition events observed within plasmids from NIHCC isolates. (a) Alignment of three ISSwi1 derivatives from different plasmids. ISSwi1 is shown in light beige, and IS26 is shown in dark yellow. (b) Intermolecular transposition of IS26, leading to replicon fusion and composite transposition. The composite IS26-IS26 transposon is highlighted with a purple background. A second composite transposon derived from ancestor 2 is highlighted with a green background. (c) IS26-mediated plasmid rearrangements. Transposition of a composite IS26-IS26 transposon (background in blue) into an ancestor plasmid to generate pKEC-dc3 is shown. pKOX-86d is derived from an intramolecular IS26 transposition (cis refers to the insertion orientation in Fig. 1b) in this ancestor. The sequences of the different flanking sequences 0 to 7 are listed in Fig. S2 in the supplemental material. Plasmid names are shown in black. Hypothetical ancestor plasmids are labeled in red. ISs are shown as rectangles (different colors represent different ISs), with triangles representing the TIRs. Full-length genes are labeled in black, and truncated genes are labeled in blue. Left TIR, open triangle; right TIR, closed triangle; integron, gray rectangle; transposase or resolvase (tnpA and tnpR), dark red arrows; resistance genes, dark orange arrows. The different flanking 8-bp sequences of IS26 are labeled with different numbers, and their sequences are listed in figures in the supplemental material. Arrowheads indicate the orientation. These symbols and color patterns are applied in all subsequent figures.

We noted that pAAC154-a50 is highly similar to a portion of pKPN-819 (Fig. 3b; see also Fig. S1a and b for further details). The indication that IS26 is involved is that the sequence in pKPN-819 homologous to pAAC154-a50 is bordered by two IS26 copies in the same orientation and is flanked by the same 8 bp at one side of each of the copies (labeled “0” in Fig. 3b). The most parsimonious explanation for the homology is that an IS26 in the ancestor of pKPN-819 inserted into the target sequence “0” in pAAC154-a50 to generate an additional IS26 copy and TSDs (Fig. 3b). Therefore, pKPN-819 appears to be an IS26-mediated cointegrate of two plasmids.

There are three IS26 copies in pKPN-819, and it appears that two of these form an active composite transposon (Fig. 3b, highlighted in purple) able to insert into the K. pneumoniae genome. The IS26 copies bracket a 5-kb DNA segment, and the proposed composite transposon can be perfectly aligned to a chromosomal locus in KPNIH24 (see Fig. S1c in the supplemental material). Furthermore, the 5-kb DNA segment is duplicated at the chromosomal locus as tandem repeats separated by an IS26 and bracketed by two additional IS26 copies. This remarkable DNA structure suggests that the chromosomal locus is likely the result of IS26 composite transposition (Fig. 1c) followed by nonreciprocal recombination at the IS26 copies (unequal crossover [45, 46]) (see Fig. S1d). Further supporting evidence for this proposed pathway is the absence of the tandem repeat in the chromosomal locus of KPNIH1. Instead, we found an IS26 flanked by octanucleotides 6 and 7 (Fig. 3b) that flank the tandem repeats in KPNIH24. Therefore, one possibility (Fig. 3b) is that the KPNIH1 and KPNIH24 chromosomes are two segregants of an unequal crossover event that occurred at the chromosomal IS26 composite transposon and resulted in a duplication and a deletion segregant. Alternatively, recombination between two IS26s in KPNIH24 could have given rise to the pattern of IS26s seen in KPNIH1.

If the composite transposon in KPNIH24 arose via the pathway in Fig. 1c, TSDs would be expected at the target replicon, whereas the flanking sequences 6 and 7 in KPNIH24 are different. However, sequence 7 is observed elsewhere in the KPNIH24 and KPNIH1 chromosomes as the flanking 8-bp sequence of another IS26 upstream of an integron. This sequence pattern thus suggests a second composite transposon, embracing the original composite transposon much like nested Russian dolls, transposed prior to unequal crossover.

There have also been several IS26-mediated plasmid rearrangements in the pKEC plasmid series from the NIHCC CPE isolates (Fig. 3c). Three isolates—K. pneumoniae KPNIH27, C. freundii CFNIH1, and E. cloacae ECNIH2—carry highly similar pKEC series plasmids (see Fig. S2a in the supplemental material). In all three, we found a 27-kb DNA segment bracketed by two IS26 copies in the same orientation (Fig. 3c, highlighted in blue; for simplicity, only pKEC-dc3 from KPNIH27 is shown) and with identical 8-bp TSDs arranged as direct repeats at one side of each copy (“0”). This 27-kb DNA segment encodes many conjugative transfer and plasmid stability proteins, the highly mutagenic polymerase UmuDC, as well as anti-restriction-modification proteins. This entire region is absent in the otherwise-highly homologous pKOX-86d from K. oxytocia strain KONH1 (Fig. 3c, bottom left), but the “0” 8-bp sequence is present. This suggests either intermolecular IS26 transposition resulting in replicon fusion (pathway in Fig. 1a, if the 27-kb region contained a replication origin) or transposition by an IS26-IS26 composite transposon followed by resolution (pathway in Fig. 1c). Another difference between these two plasmids is a 74-kb region in pKEC-dc3—missing in pKOX-86d—that is next to an IS26 copy (Fig. 3c; see also Fig. S2b in the supplemental material). This is probably due to IS26-mediated deletion in the ancestral plasmid upon intramolecular transposition in cis.

Intramolecular transposition leading to DNA inversions.

Matched pairs of 8-bp sequences flanking different IS26 copies within the same plasmid provide evidence for intramolecular replicative transposition in trans, leading to DNA inversion; four examples are shown in Fig. 4. Plasmid pKPN-068 from KPNIH27 (Fig. 4a, right) carries multiple antibiotic resistance genes for two beta-lactamases (blaTEM-1 and blaSHV-12), two aminoglycoside modification enzymes (ant2 and cata1), and a quinolone resistance gene (qnrB). This 80-kb plasmid has seven ISs, including four IS26 copies, one copy of ISEcp1 (an IS1380 family member), an ISAcsp1-like IS (a Tn3 family member), and one ISSwi1 copy that is disrupted into three segments (pA, pB, and pC). This disruption pattern is likely the result of ISAcsp1 insertion and IS26-mediated intramolecular transposition giving rise to inversion. The first disruption is clearly made by insertion of the ISAcsp1-like IS (shown in green) into the ISSwi1 transposase gene (tnpA) (see Fig. S3a in the supplemental material) as indicated by the flanking 5-bp TSDs characteristic of ISAcsp1 (shown as unfilled arrows in pKPN-068). The second disruption is an inversion of pA relative to the rest of ISSwi1 and its separation from pB by ~30 kb. In pKPN-068, two copies of IS26 adjacent to the pA and pB segments are in opposite orientation and are flanked by identical 8-bp repeats inverted relative to each other (“0” in red). This arrangement is consistent with intramolecular transposition in trans of a single IS26 into a “0” target site in an assumed hypothetical ancestor at the junction of the pB and pA segments of ISSwi1, resulting in the inversion of the intervening DNA.

FIG 4 

Intramolecular transposition (trans) resulting in DNA inversion. Cases of deduced intramolecular transposition events within hospital isolated strains are shown. Inverted regions are highlighted with a pink background. The cis and trans notations indicate the insertion orientation, as shown in Fig. 1b. Different ISs are shown in different colors. (a) Intramolecular transposition leading to pKPN-068. pA, pB, and pC represent the disrupted segments of ISSwi1. Further details, including sequences of the target and the flanking sequences 1 to 5 can be found in Fig. S3a in the supplemental material. (b) Intramolecular transposition leading to pKPC-727. pA and pB represent the partial DNAs constituting the conjugal transfer operon; further details, including the sequences of the target and the flanking 1 and 2 sequences can be found in Fig. S3b. (c) Intramolecular transposition leading to pENT-576. The sequences of target and flanking sequences 1 and 2 can be found in Fig. S3c. (d) Intramolecular transposition events among a series of pKpQIL plasmids isolated from New York and New Jersey hospitals. Further details, including sequences 1 and 2, can be found in Fig. S4 in the supplemental material.

The hypothetical ancestor plasmid shown on the left in Fig. 4a would have carried three copies of IS26, two with identical 8-bp flanking DNAs (“1”) and in the same orientation. This was probably a consequence of intermolecular transposition of an IS26-IS26 composite transposon (grey background) carrying the blaSHV-2, qnrB, and cata1 resistance genes, followed by resolution (pathway in Fig. 1c). These steps would have brought two antibiotic resistance determinants, blaTEM-1 and ant2, previously 30 kb apart, into the immediate neighborhood of blaSHV-12, thereby augmenting an existing resistance cluster. This illustrates an IS26-mediated mechanism of resistance cluster formation different from and independent of integron- or ISCR-mediated rearrangements.

We also found a pattern of diagnostic repeats within the plasmid pKPC-727 from K. oxytoca KONIH1 (Fig. 4b), where two IS26 copies located next to a split conjugal transfer operon include two 8-bp flanking sequences (“0” in red) in an inverted configuration. This is consistent with IS26-mediated transposition resulting in intramolecular inversion, leading to the remodeling of the operon. We found a similar pattern of IS26 elements and flanking repeats in plasmid pENT-576 of E. cloacae ECNIH3 (Fig. 4c), where a 16-kb DNA region that included ABC transporter and urease genes was probably inverted due to intramolecular IS26 transposition into an intergenic region.

Examples of IS26-mediated intramolecular plasmid rearrangements extend beyond the NIHCC isolates. We detected similar IS26-related patterns in a series of pKpQIL-like carbapenemase blaKPC3-carrying plasmids embedded in strains of K. pneumoniae, E. coli, and Enterobacter aerogenes isolates from New York and New Jersey hospitals (47). Those authors noted an inversion bracketed by two IS26 copies in pKpQIL-234, but only one of these IS26 copies is present in pKpQIL-10 (Fig. 4d; see also Fig. S4 in the supplemental material). This was interpreted as an insertion of a second IS26 followed by subsequent DNA inversion between two IS26 elements. However, it seems more likely that pKpQIL-234 is a product of a one-step intramolecular transposition in pKpQIL-10 by IS26 into the target “0” octanucleotide. Those authors also noted another rearrangement in pKpQIL-Ec (Fig. 4d, bottom), a deletion of 8 kb of DNA starting from the right flank of ISKpn14 (an IS1 family element). We suggest that this is consistent with intramolecular replicative transposition in cis of ISKpn14 resulting in DNA deletion, as IS1 family members also appear to use replicative transposition (48).

ISSwi1 is a hot spot for transposition.

As shown in Fig. 1b, one possible outcome of intramolecular transposition is DNA segment deletion without generating telltale abutted TSDs. Deletion events are difficult to identify unless replicons both before and after the transposition event have been detected. Although this is rare, deletion can be inferred when a partial gene is left behind. As over half the 8-bp flanking sequences in Table 1 are unique, we believe some of these are due to transposition-mediated deletion.

Figure 5a depicts the many ways that the 16 copies of the ISSwi1 locus in different plasmids in the analyzed NIHCC genomes have been disrupted. ISSwi1 is also found in hospital-isolated strains ATCC BAA-2146 (35), NJST258_1, and NJST258_2 (41) and, again, most of the copies are disrupted. The disruption pattern is usually different among the different copies but, remarkably, it is always caused by other transposable elements, such as IS26 and Tn4401 (interestingly, the three isoforms of Tn4401 we observed [a, b, and d] give rise to the highest bla expression among five reported Tn4401 isoforms [49]). Thus, it appears that ISSwi1 might be a hot spot targeted by these elements. The mechanism responsible for this targeting is not clear, but as ISSwi1 carries one blaTEM-1 gene and Tn4401 carries carbapenemase gene blaKPC, the result is often the clustering of these genes. When the disruptions are caused by IS26, deletions are often observed at the antibiotic resistance gene locus. We suspect that these observations indicate ongoing plasmid streamlining that continues even after the acquisition of resistance determinants.

FIG 5 

ISSwi1 derivatives and IS26-mediated gene deletions. (a) ISSwi1 serves as an insertion hot spot for IS26 and other ISs. Plasmids isolated from hospitals other than the NIHCC are boxed by rectangles in gray, and the detailed genetic context of different ISSwi1 derivatives in these plasmids is shown in Fig. S5 in the supplemental material. Three ISSwi1 derivatives and the corresponding plasmid names are shown. (Top) Intact ISSwi1; (middle) ISSwi1-m1; (bottom) ISSwi1-m2. Inserted DNA to generate ISSwi1-m1 is highlighted in light orange. Deleted DNA by IS26 insertion to generate ISSwi1-m2 is highlighted in gray. Arrows show the other disruptions, by either IS26 or other ISs in the indicated plasmids; note that the IS26 insertions often disrupt the antibiotic resistance gene locus (aminoglycoside modification genes or blaTEM-1). Complete filled arrows indicate simple insertions, and half arrows indicate that one part of ISSwi1 was deleted after IS insertion. Red empty arrows indicate disruption was caused by intramolecular transposition in trans leading to DNA inversion. (b) Tn4401 truncations by IS26s and other ISs.

Tn4401 is partially deleted by IS26 in pKPC-47e and the pKEC series plasmids (Fig. 5b) and is flanked by two IS26 copies, suggesting the possibility of a functional composite transposon. If this were the case, this composite transposon could be a new vehicle for blaKPC dissemination. As there are many more copies of IS26 than of Tn4401, we speculate that this composite element may have higher transposition activity than the original Tn4401 and could be an example of IS26-mediated mobility enhancement of a critical resistance determinant.

Intramolecular transposition by Tn4401 creates an antibiotic resistance gene cluster.

Within the NIHCC strains, replicative transposition is not restricted to IS26; Tn4401 also appears to play a role in plasmid reorganization. Although Tn3 family elements such as Tn4401 are known to use replicative transposition (27), they are believed to be constrained, as they often display transposition immunity (50), a regulatory mechanism preventing the insertion of a second copy into a replicon already containing a copy of the same transposable element. Consistent with this, most replicons analyzed here carry only a single copy of Tn4401. pKPC-727 from K. oxytoca KONIH1 is an exception (Fig. 6a, right). It carries two inverted Tn4401 copies, 73 kb apart. Neither copy has a characteristic 5-bp abutted TSD (51). Interestingly, the two Tn4401 copies are located next to two partial ISSwi1 elements (pA and pB) which are inverted and flanked by the same 5-bp sequence (“0” in red), an expected length for Tn3 family transposons and present in an inverted orientation. The DNA structures in pKPC-727 are thus consistent with a single step of Tn4401 intramolecular transposition in trans. Tn4401 in the hypothetical ancestor would have had perfect pentanucleotide TSDs (“1”), indicating a conventional intermolecular transposition event as shown in Fig. 1a. Although this intramolecular transposition event disrupted blaTEM-1, it duplicated blaKPC-2 and simultaneously brought several other antibiotic resistance determinants close to one blaKPC-2 copy, thereby forming a resistance cluster.

FIG 6 

Examples of intramolecular transposition of Tn4401. (a) Deduced intramolecular transposition event leading to pKPC-727. pA and pB represent DNA segments which partially reconstitute ISSwi1-m1. Reversed regions are highlighted with a pink background. The notation trans indicates the insertion orientation shown in Fig. 1b. Further details, including the sequences 0 and 1 are shown in Fig. S6 in the supplemental material. (b) Deduced intramolecular transposition event leading to a plasmid from K. pneumoniae S9. Details, including the sequences 0, 1, and 2 are shown in Fig. S7 in the supplemental material.

Tn4401 also appears to have carried out intramolecular transposition within a plasmid from a K. pneumoniae isolate S9 from a New York City Hospital (Fig. 6b) (8). S9 carries two Tn4401 elements in inverted orientation and without detectable abutted TSDs. As group II intron fragments were detected at both ends of Tn4401, those authors hypothesized that Tn4401 duplication was due to self-splicing by the intron. However, we suggest that the duplication is much more likely due to the intramolecular transposition of Tn4401 into a target site within the group II intron.

DISCUSSION

We have carried out a global analysis of bacterial insertion sequences in the genomes of nine CPE isolates from the NIH Clinical Center collected from the 2011 outbreak to the end of 2013. As these genomes have been fully sequenced with long reads and with high coverage (30), we were able to precisely analyze their mobile elements using ISsaga. Analyzing multiple isolates was key to facilitating the identification of DNA rearrangements within the bacterial genomes. Although the chromosomes of these strains appear relatively stable, the plasmids where most of the resistance determinants reside are dynamic and display a large degree of interstrain variation. The presence of a highly active IS26 is consistent with previous reports on other clinical isolates (52–54), but here we have been able to trace its impact by taking into account its replicative transposition mechanism and using its expected TSD patterns as guides. In several cases, our findings allow us to trace not only precisely how resistance plasmids are reorganized by this mobile element, but also the temporal relationship between related plasmids.

High activity of IS26 in the analyzed strains is suggested by its high copy number. Although target site duplication is usually considered a fingerprint of transposition, within the genomes analyzed we did not find a single IS26 copy flanked by a matching pair of TSDs. However, the lack of flanking TSDs cannot, by itself, be interpreted as lack of recent IS26 activity or be arbitrarily assumed to indicate an unconventional transposition mechanism, as intermolecular replicative transposition without resolution does not generate abutted TSDs flanking a single IS copy. Also, the various possible consequences of intramolecular transposition must be taken into account. Nevertheless, the lack of flanking TSDs does imply that either intermolecular transposition by IS26 occurs at a very low frequency or that the resolution step of this process is very inefficient. As resolution is believed to be dependent on RecA-mediated homologous recombination (39), it seems unlikely that this host machinery is limiting.

It is clear that intramolecular transposition in trans that results in DNA inversion occurs fairly often, leading to large DNA rearrangements. This pathway generates not only additional IS26 copies but also an inverted DNA segment bracketed between two copies of IS26 which in principle can act as a composite transposon, mobilizing the DNA segment between them in subsequent transposition events.

Recently, Hall and coworkers investigated IS26 transposition in RecA-deficient E. coli (55) and reported that the presence of an IS26 in a target plasmid greatly increased the frequency of additional IS26 insertions into that plasmid. Using a PCR-based assay, they found replicon fusions between IS26-containing donor and target plasmids, but the resulting cointegrate did not contain an additional copy of IS26 as would be expected for replicative transposition. Despite the absence of RecA, the observed cointegrate was structurally equivalent to the recombination product between the two IS26 copies in the donor and target plasmids. Although we cannot exclude RecA-independent mechanisms of template switching (56), the transposase of IS26 might be responsible for this homology searching and pairing. One possibility is that two IS ends from different IS copies in separate replicons form a synapse in the same transpososome (see Fig. S8 in the supplemental material), a process that resembles transposition of a composite transposon which involves TIRs from two different IS copies located in the same replicon. Strand exchange followed by replication would thus result in the observed replicon fusion. Such in trans synapsis thus allows the possibility of “self-targeted” insertion. A similar mechanism has been invoked to explain “targeted” insertion of IS3 and IS30 family members into TIRs (57–59). This process would be expected to be independent of RecA but, in the case of the IS3 family member IS911, it is known to depend on the RecG helicase (60). However, in these cases, the strand exchange reaction is expected to produce tandem copies of the element, due to their different transposition mechanisms. IS911 is known to use a copy-and-paste mechanism (21), whereas IS30 has been proposed to use both direct insertion and cointegrate formation (59). It would be certainly interesting to investigate experimentally whether such events occur and how often, or if IS26 in some circumstances transposes via mechanisms different from replicative transposition.

As we observed, intramolecular transposition can lead to an increase in the IS26 copy number within a genome (trans [Fig. 1b]), but if the opposite strand of the target site is used (cis [Fig. 1b]), the outcome is DNA deletion. More than half the 70 IS26 copies in the nine genomes we analyzed have flanking 8-bp sequences not associated with any of the other IS26 copies. These may well be the result of intramolecular transposition, implying that it is relatively common, as might be expected if there is no strand preference during the strand transfer step of transposition. Intramolecular transposition accompanied by deletion is clearly a useful mechanism for keeping bacterial genome size under control during the onslaught of exogenous DNA acquired, for instance, through horizontal transfer. As the horizontal acquisition of resistance determinants is clearly important, it is possible that IS26-mediated genome trimming plays a significant role in this process, enabling the maintenance of a reasonable genome size and therefore viability (of course, we do not exclude the possibility that other ISs or transposons can accomplish the same thing). Deletions could also eliminate certain antigens to allow escape from the host immunity system, as seen in the Bordetellae, where IS activity resulted in a genome size reduction concomitant with gene inactivation (61). We have also seen that cis-mediated disruptions and deletions can occur within antibiotic resistance genes (Fig. 5). Most of these involve aminoglycoside modification enzymes or narrow-spectrum beta-lactamases. Perhaps this is yet another sign of genome optimization to increase fitness by eliminating “narrow-spectrum” resistance genes no longer needed once a broader-spectrum resistance gene in the same class has been acquired (for example, elimination of an ESBL gene once a carbapenemase gene is acquired).

In addition to playing a role in the organization of resistance genes, we observed that transposition of IS26 remodeled some plasmid-borne genes involved in conjugational transfer and plasmid stability, for instance. As this may have conflicting effects on plasmid maintenance and stability, some mechanisms would be needed to rectify conflicts. Furthermore, replicative transposition could lead to plasmid fusion or plasmid integration into the chromosome in the absence of homologous sequences. For example, this has been observed in Yersinia pseudotuberculosis due to the activity of the IS6 family member ISYps1 (62).

Replicative transposition is not restricted to the IS6 family, and other presumed replicative mobile elements, such as Tn4401 (Tn3 family), IS903B (IS5 family), and ISKpn14 (IS1 family), also exist (27, 48, 63). In contrast to the IS26 examples identified here, we found several copies of these family members flanked by direct repeats of the expected lengths, suggesting intermolecular transposition followed by resolution. While Tn3 family members such as Tn4401 often encode their own resolvase (64), IS903B and ISKpn14 do not (see the ISfinder database). Perhaps resolution of cointegrates generated by these elements relies on RecA-mediated homologous recombination, as suggested for IS26, or these IS elements may be able to transpose by simple insertion (65, 66).

Results of our analysis suggest that IS26 prefers to transpose within plasmids rather than into the chromosome. This is not a general feature of ISs, as there are many other ISs present at very high copy number in the chromosomes analyzed here. For example, up to 17 copies of ISKpn26 were observed in certain chromosomes (e.g., KONIH1 [data not shown]). As the chromosomal IS copies tend to have conserved flanking TSDs more frequently than those in the plasmids, it seems that most chromosomal transposition events employ simple transposition mechanisms, such as cut-and-paste or copy out-and-paste (26), which might be better tolerated, as they cause less overall disruption than replicative transposition. Although there are several chromosomal IS26 copies within the nine analyzed NIHCC strains, these are exclusively located in a prophage-carrying region and may have transposed into the phage genome before its integration into the host chromosome. We also believe that selection bias might be involved, which would constrain the IS26 copy number in the chromosome so as to avoid potentially harmful deletions. However, we cannot exclude the possibility of replicative transposition within chromosomes: IS903B may have deleted a 15-kb region from the KPNIH1 chromosome (region 2716602 to 2731997 in CP008827) that is rich in transporter genes and whose loss results in the absence of this region in NJST258_1 (data not shown). This is consistent with the reported intramolecular replicative transposition activity of IS903B (63).

The availability of whole-genome sequencing carried out with new tools, such as Pacific Biosciences RS II SMRT technology, makes possible the unambiguous identification of mobile elements, and we expect many more clinical isolates to be sequenced. Combining these sequence data with IS annotation tools like ISsaga provides robust and reliable data from which to derive mechanistic information about the activity of mobile elements in their native hosts and therefore to understand the evolution of their host genomes. This in turn may open up a new avenue to trace IS dissemination and host adaptation. In particular, following transposition events of mobile elements such as IS26 is a way to understand the evolution of resistance plasmids, revealing changes that may optimize the viability of the pathogen in a hospital environment. The ability to follow these events clearly establishes directionality as the rearrangements generated by replicative transposition events are not reversible.

It is not obvious why IS26, among the large variety of mobile genetic elements, gained such prevalence in the resistance plasmids in the examined enterobacterial strains. One possibility is a founder effect, in which the first IS26 element inserted in proximity to resistance genes, which allowed it a foothold with these particular genes. Alternatively, there might be some currently unknown host factor involved in IS26 mobility, making it highly active and perhaps influencing the choice toward intramolecular pathways. By extending these investigations to other clinically isolated pathogens, we hope to uncover common features of mobile genetic element-driven resistance plasmid reorganizations and to understand their effects in a mechanistic framework.

MATERIALS AND METHODS

Genomic sequences of the nine NIHCC isolates as well as other clinical isolates analyzed in this paper are from the NCBI public database (see Table S2 in the supplemental material for a list of the analyzed genomes). IS annotation for these genomes was carried out with the ISsaga program (insertion sequence semi-automatic genome annotation; at http://issaga.biotoul.fr/ISsaga/issaga_index.php) (33). SnapGene 2.5 was used to visualize the annotation results. Mauve 2.3.1 was used to perform comparative genome alignments.

(a) Plasmid alignment between pAAC154-a50 and pKPN-819. Mauve 2.3.1 was used to perform comparative genome alignment for these two plasmids. The full sequence of KPNIH1_pAAC154-a50 is shown as two blocks in red and green and are homologous to the sequence found in KPNIH24_pKPN-819. The rest of the sequence in KPNIH24_pKPN-819 is not colored, but ORFs are presented as small rectangles. IS26 copies are shown as dark yellow rectangles, with two triangles representing the TIRs (open, left TIR; closed, right TIR). The 8-bp flanking sequences are shown as black arrows, indicating their orientation. Different numbers represent different sequences as listed at the bottom right. (b) A potential IS26-IS26 composite transposon is generated after replicon fusion. The potential composite transposon is highlighted with a purple background. The blue arrow represents the 5-kb DNA between two IS26s in a composite transposon which is homologous to a piece of chromosomal DNA in tandem repeats in KPNIH24 but absent in the KPNIH1 chromosome (shown in panel c). Integron, gray rectangle; attC site, blue narrow rectangle; attI site, green narrow rectangle. A second composite transposon is highlighted with a green background. (d) Schema of unequal crossover leading to tandem repeats or deletion. Homologous recombination (nonreciprocal recombination) occurs between the right IS26 from one sister replicon and the left IS26 from the other sister replicon of an IS26-IS26 composite transposon. DNA crossover leads to one daughter replicon containing tandem repeats of DNA bracketed within the composite transposon and the other with this DNA deleted and retaining a single IS26. Download

Figure S1, EPS file, 2 MB

(a) Alignments within three pKEC plasmids from KPNIH27, CFNIH1, and ECNIH2. Homologous regions are in red. (b) Plasmid alignment of pKEC-dc3 and pKOX-86d. The composite transposon in pKEC-dc3 is highlighted with a blue background. Two small gaps present in the pKEC series plasmids are due to the insertion of two other IS elements. One is an IS3 family member with a low sequence identity to any ISs in the ISfinder database but with 5-bp TSDs, suggesting a transposition event. The other is an ISApu2 family member which generated a 10-bp TSD as expected (see the ISfinder database). One major gap in pKOX-86d is the presence of an additional integron carrying blaPSE-1. This integron is adjacent to the 5′ end of another one, which might be due to homologous recombination mediated at the integrase gene. We are unable to explain the other gap. TSDs are shown by black arrows. Different numbers indicates different sequences, as shown at the bottom right. Download

Figure S2, PDF file, 0.8 MB

Intramolecular transposition (trans) resulting in DNA inversion. Three cases are illustrated of the deduced intramolecular transposition events within pKPN-068 (a), pKPC-727 (b), and pENT-576 (c). pA, pB, and pC in panel a represent the disrupted segments of ISSwi1. Reversed regions are highlighted in pink. Target sites for intramolecular transposition are indicated by red arrows, and TSDs are shown by black arrows. Different numbers indicate different sequences, which are shown at the bottom. Download

Figure S3, EPS file, 1.5 MB

Alignment of pKpQIL-series plasmids isolated from New York and New Jersey hospitals (adapted from Fig. 1 in reference 31 with permission). (The IS26 in pKpQIL-234 that is labeled with a red asterisk was mislabeled as ΔIS26 in the original paper.) Target sites for IS26 intramolecular transposition are shown with red arrows, and other TSDs are shown as black arrows. The 8-bp target of ISKpn14 is shown as the black empty arrow. This TSD length was as expected for IS1 family elements. Different numbers shown at the bottom of each plasmid scheme indicate different TSD sequences. Download

Figure S4, EPS file, 2.3 MB

ISSwi1 derivatives. DNA regions within a series of plasmids demonstrated the variability associated with the presence of ISSwi1 and IS26. The plasmids are aligned either to the tnpR and tnpA of ISSwi1 or, in the absence of both genes, to blaTEM-1. Download

Figure S5, EPS file, 1.7 MB

Detailed pattern of the Tn4401 intramolecular transposition leading to DNA inversion in pKPC-727. Target sites for Tn4401 intramolecular transposition are shown as red arrows, and other TSDs are shown as black arrows. Download

Figure S6, EPS file, 1.3 MB

Genetic context of two copies of Tn4401 in a plasmid from K. pneumoniae S9 (adapted from Fig. 4 in reference 8 with permission) and its underlying intramolecular transposition event. Target sites for Tn4401 intramolecular transposition are indicated by red arrows, and the other TSDs are shown by black arrows. Download

Figure S7, EPS file, 1.8 MB

Schema of IS26 synapsis in trans and hypothetical consequences in cointegrate formation. Green oval, IS26 transposase. Synapsis between IS26s in two plasmids results in cleavage at a single TIR of each IS26 copy. We assume that targeting is less favored in this case. The repair of the nick is carried out either by host ligase (which could generate original individual plasmids or cross-ligate, as shown) or by host replication. For the latter, the peeled-off single-strand DNA is degraded by exonucleases and ligation takes place. Replication generates a copy of the original plasmid and a cointegrate. The second case of replication initiated by the other replication origin (red vertical rectangle) generating the red original plasmid and the same cointegrate are not shown. Download

Figure S8, EPS file, 0.9 MB

Flanking sequences of the four most abundant ISs in the KPNIH1 genome (TSDs are in red, and nonapplicable refers to partial ISs)

Table S1, DOCX file, 0.02 MB

List of analyzed genomes and their NCBI accession numbers (strains listed in the upper part above the gray line are from the NIHCC, and those below the gray line are from other hospitals)

Table S2, DOCX file, 0.02 MB