ISCR Elements: Novel Gene-Capturing Systems of the 21st Century?

INTRODUCTION

Bacterial antibiotic resistance is, once again, high on the clinical and scientific agenda. The ability of bacteria to evolve antibiotic-resistant forms is impressive, as attested to by a continuing concern that untreatable “superbugs” will continue to emerge in clinical medicine. Bacteria are noteworthy for their remarkable ability to adapt to changes in their environments, not least for their adaptability in reassorting and redistributing genetic information. Bacteria possess an impressive set of tools with which to adjust the blueprint of the cell, the bacterial genome, enabling the cell to alter, add, or lose genetic information. Nowhere is this potential for change better illustrated than the bacterial riposte to antibiotics, antiseptics, disinfectants, and heavy metals used in clinical medicine. Although mutation has an important role to play in the evolution of antibiotic resistance, the predominant factor for the escalation of antibiotic resistance in more than half a century is the acquisition of antibiotic resistance genes.

The acquisition and spread of antibiotic resistance genes among bacteria that are intimately associated with humans and their domesticated animals are well documented. Data from studies examining this horizontal gene transfer indicate that it is driven by multiple systems that involve both cell-to-cell transfer and gene transfer from one DNA molecule to another. The latter can take place regardless of similarity between the donating and recipient molecules (38, 73) and includes systems such as transposons and site-specific recombination systems termed integrons (8, 9, 19). Recently it has become clear that while these systems can account for much of the movement of resistance genes between DNA molecules, they fail to explain the spread of a substantial and growing subset of resistance genes. These resistance genes are linked to sequences termed “common regions” (CRs), which are often found beyond but close to the 3′ conserved sequences of class 1 integrons. A comparative analysis of these CRs has revealed that they are related to each other and resemble an atypical class of insertion sequences (ISs), designated IS91-like. IS91 and like elements differ from the insertion sequence paradigm (59) in that they lack terminal inverted repeats (IRs) and are thought to transpose by a mechanism termed rolling-circle (RC) transposition (because the mechanism involves rolling-circle replication). As a consequence, they can transpose adjacent DNA sequences, mediated by a single copy of the element (96). This is unlike most IS elements, where two copies (one of which must be intact) need to flank the mobilized gene (6), as in the mobilization of the bla_CTX-M genes by ISEcp1 (75, 76). We will argue that CRs are members of an extended family of IS91-like elements that transpose by RC transposition and are responsible for the mobilization of virtually every class of antibiotic resistance genes, including those encoding extended-spectrum β-lactamases (ESBLs), carbapenemases, and enzymes conferring broad-spectrum aminoglycoside resistance, florfenicol/chloramphenicol resistance, and resistance to trimethoprim and quinolones.

CRs AND THEIR DISCOVERY

Common regions were first discovered and reported in the early 1990s as a DNA sequence of 2,154 bp that was found in two complex class 1 integrons, In6 and In7 (93). These arrays possess partial duplications of the 3′ conserved sequence of the class 1 integrons (Fig. 1), between which are located the CR and antibiotic resistance genes. The sequence was initially described as a common region to distinguish it from the 5′ and 3′ conserved sequences (CS) of class 1 integrons and was found immediately downstream of a shortened 3′ CS (Fig. 1A).

Class 1 integrons are genetic elements that possess a specific recombination site, attI, into which resistance genes, in the form of gene cassettes, can be inserted by site-specific recombination. Gene cassettes are small circular DNA molecules, approximately 1 kb in size, comprising a single gene together with a recombination site termed a 59-base element (59be) (6, 19). Each 59be has imperfect terminal IRs of approximately 20 bp, and the different sizes of 59bes reflect the fact that the central sequences differ markedly from one element to another. Each terminal IR comprises oppositely oriented putative integrase binding sites of 7 or 8 bp, designated 1L and 2L at the 5′ end and 1R and 2R at the 3′ end, separated by 5 bp and 5 or 6 bp, respectively (94). Cassette integration usually occurs at the attI site of the integron, and the original cassette 59be is reformed whenever the gene cassette is excised from the integron. Thus, resistance genes found on gene cassettes are potentially highly mobile (7-9, 19).

However, the gene arrays of In6 and In7 are therefore somewhat different from the more common form of class 1 integrons, which contain only one copy of the 3′ CS and no CR, and are often called complex class 1 integrons in the literature (9). These common regions were also interesting in that they are intimately associated with different resistance genes carried on each integron. In the case of In6 the CR is found immediately upstream of a chloramphenicol resistance gene (catAII) (Fig. 1A), while the CR in In7 is associated with a novel trimethoprim resistance gene (dfrA10) (Fig. 1A). The particularly striking aspect of these two CR-associated resistance genes is the fact that neither is linked to 59bes, as is typical for the overwhelming majority of resistance genes embedded in class 1 integrons. Integrons generally evolve by integration and excision of a variety of gene cassettes. Accordingly, it was unexpected to find noncassette resistance genes closely associated with integron structures. Searches of available databases in 1993 detected only one other sequence with high identity to the CR, i.e., a 349-bp section of the small nonconjugative plasmid, RSF1010 (93). This plasmid is a broad-host-range (incQ) plasmid that confers resistance to sulfonamide and streptomycin and has been the basis of numerous broad-host-range vectors for recombinant DNA manipulation. This fragment of the CR is the result of truncation at both ends relative to the sequence found in In6 and In7. The sequence is related but not identical to the CR and is now known to be a degenerate copy of a closely related variant of the original CR sequence (now designated CR2 and CR1, respectively). The truncated CR in RSF1010 is also associated with a gene (sul2) encoding resistance to sulfonamide (Fig. 2) (accession no. D37825). Integrons In6 and In7 were discovered on plasmids pSa and pDG0100, respectively, with pDG0100 being a 120-kb incC group plasmid and pSa a 39-kb incW group plasmid (93). Recently, a retrospective study found an integron that is virtually identical to In6 on the incU group plasmid pAR-32, which confers resistance to sulfonamide, chloramphenicol, and streptomycin. This plasmid, which is globally distributed and identical to the incU reference plasmid RA3, was found in a resistant Aeromonas salmonicida isolate that came from Japan in 1970 (92) (Fig. 1A).

The CRs in integrons In6 and In7 and the homologous CR segment of RSF1010 comprised, therefore, the sum of the information known about these elements in 1993, and for several years accumulation of information concerning them was slow (90, 93). Recently, however, there has been a major proliferation of information regarding these important genetic elements and the various antibiotic resistance genes that are intimately associated with them. Most of this information has been reported in the last 2 to 3 years, and it probably represents just a small portion of these elements associated with resistance genes. CR elements have now been found in numerous gram-negative organisms of clinical importance, as well as in a few gram-positive bacteria (2) (Fig. 1 to 4). Various studies have shown CRs to be present on both plasmids and chromosomes, and they have been found in bacteria worldwide. CRs are now known to be closely associated with many antibiotic resistance genes, including those encoding extended-spectrum class A β-lactamases, extended-spectrum class C β-lactamases, the metallo-β-lactamase (MBL) bla_SPM-1, the extended-spectrum class D β-lactamase bla_OXA-45, and plasmid-mediated quinolone resistance (qnr), as well as various trimethoprim resistance genes, the chloramphenicol resistance gene catAII, broad-spectrum aminoglycoside resistance genes, tetracycline resistance genes, the florfenicol resistance gene floR, the macrolide resistance gene ereB, and others (Fig. 1 to 4).

There is substantial evidence linking the origin of a subset of these resistance genes to the chromosomes of environmental bacteria (1, 86, 103), and there is considerable circumstantial evidence that CR elements are responsible for the mobility and dissemination of many genes (23, 107). CRs have also been implicated in the construction of the multidrug resistance (MDR) regions of the integrative conjugative element SXT and the Salmonella genomic island 1 (SGI1) and their variants (4, 15). If this is true, as we will argue, then the potential of CRs to mobilize adjacent antibiotic resistance genes may pose a considerable threat to already beleaguered anti-infective regimens, particularly for gram-negative bacteria. We contend that CRs represent a family of related mobile genetic elements that are similar to the rather unusual IS element IS91 (34). This review will itemize the available information concerning these elements and the resistance genes associated with them and will attempt to give some insight into their mode of movement and their origins.

CHARACTERIZATION OF CRs

CRs comprise a set of related sequences. Each CR identified so far accommodates a single open reading frame (ORF). The predicted products of these putative genes, all approximately 500 amino acids long, are closely related to each other, differing by 4 to 88% in amino acid identity (Fig. 5 and 6). However, CR identity extends beyond the ORFs to include nucleotide sequences in both directions. To the left-hand end of the CR ORFs, the sequences show some homology for 240 to 250 bp before ending in a short section of conserved sequence (Fig. 7). This end of a CR sequence will be referred to as the 3′ end of the CR because it is 3′ to the ORF. The other end, therefore, is the 5′ end.

Initial screening of the EMBL prokaryote nucleotide sequence databases with the CR sequence of In6 identified many sequences with different levels of identity to the search sequence. These were found to fall into a number of subgroups. Many sequences have been found to display identity throughout the single gene accommodated on each CR, while others show a significant degree of identity only to the sequence downstream, in some cases encompassing just the 29 bp at the 3′ end of the probe sequence.

CRs ARE IS91-LIKE TRANSPOSABLE ELEMENTS

The CR embedded in the complex class 1 integrons In6 and In7 is 2,154 bp long and accommodates an ORF, Orf513, encoding a putative product of 513 amino acids. This was originally misidentified as an ORF of 341 codons and accordingly designated Orf341. The mistake was corrected when it was discovered (71). When the CR sequence is used to interrogate the nucleotide databases, the only sequences that show significant degrees of homology are CRs. The search provides no convincing clues as to CR identity. This is also the case when the amino acid sequence of Orf513 is used to interrogate the protein databases. No convincing identity is revealed, although Cloeckaert et al. noted that Orf513 displays some identity with the transposition protein of IS801 but that the identity is insufficient to be confident of the match in the absence of other data (17). The putative identity of CRs was also suggested by other studies after the publication of that by Cloeckaert et al. (17, 29, 60, 70, 71). Significant identity is seen only between Orf513 and the putative products of other CR ORFs. Given that CRs are frequently associated with antibiotic resistance genes that clearly are mobile, and given their sizes, it is possible that CRs are atypical transposable elements. However, there is a trio of somewhat unusual IS elements, typified by IS91 and including IS801, that employ transposition proteins and lack the DDE motifs of standard transposases (34). Instead, the activities of these proteins are tyrosine based, and they are related to the Rep proteins of Staphylococcus aureus plasmids and to the replication protein of coliphage φX174 (64). When the amino acid sequences of the putative products of the CR ORFs are aligned with those of the transposition proteins of IS91 and its relatives, IS801 and IS1294, although overall amino acid identity between the CR-encoded proteins and the transposition proteins of IS91, IS801, and IS1294 is very low, key amino acid motifs of the transposition proteins can be seen to be present in the CR gene products (Fig. 8), indicating they also may indeed be transposition proteins, consistent with the idea that CRs are transposable elements (34, 96).

IS91 and the related elements IS801 and IS1294 are unusual in that they lack the terminal IRs typical of most IS elements (10, 33). Instead, their termini are distinctive and have different functions (34). This is because their transposition mechanism differs from that of most transposable elements in that it involves RC transposition; accordingly, it has been termed RC transposition (34, 65). The single protein encoded by each element is the cognate transposase, which is responsible for initiating replication of the element and is also believed to be involved in terminating replication prior to the final recombination step of the transposition event (34, 96).

With these elements, transposition is a processive system in which one end of the element, oriIS, serves as an origin of replication while the other, terIS, is a replication terminator (34, 96). The oriIS sequence is 5′GxTTTTxAAATTCCTATxCAT3′ and is located approximately 300 bp downstream from the transposase gene in IS91 and IS1294 but only 179 bp downstream from the transposase gene in IS801. Sequences in a similar position relative to CR ORFs, i.e., 240 to 250 bp downstream and at the 3′ end of the CR sequence, have the consensus sequence 5′GCGTTTGAACTTCCTATACxx3′ (Fig. 7). This sequence is strikingly similar to the oriIS consensus sequence (identity between the nucleotide sequence of the consensus of the 3′ end of CRs and that of the oriIS of IS91-like elements is indicated in boldface), to the extent that, given the likely identity of CR-encoded proteins, it is not unreasonable to conclude that this end of a CR sequence is also an origin for RC replication (34, 65, 96). Hence, CR sequences have two key features of IS91-like elements that are similarly located. We conclude, therefore, that these sequences are IS91-like transposable elements. Accordingly, we propose that they be designated ISCR elements, and will so refer to them hereafter, so as to highlight both their identities as insertion sequences and the history of their discovery. Further, the numbers presently used to distinguish CRs are retained to distinguish different ISCRs. Therefore, CR1 becomes ISCR1, CR2 becomes ISCR2, etc.

The third feature of IS91 that is strongly conserved both in IS801 and in IS1294 is terIS (96), located approximately 300 bp upstream from the start of the transposase gene in IS91 and IS1294 but only 101 bp from the start of the transposase gene in IS801 (33, 34, 65, 96). The replication terminator sequence, terIS, is a 25- or 26-bp sequence that contains a perfect 6-nucleotide punctuated inverted repeat and a final terminating tetranucleotide, 5′GTTC3′ (96). Unfortunately, in the cases of ISCR elements there is, in general, too little information to determine with confidence the 5′ ends of the elements. In the case of ISCR1, although there are many examples of this element (Fig. 1), the sequence at the 5′ end is identical in virtually every case. Therefore, it is not possible to determine precisely where the junction between ISCR1 and the carrier sequence is. However, there are four instances where ISCR2 has different sequences on the 5′ side of its ORF (Fig. 9). As with the 3′ end, the marked differences between these sequences can be used to determine the likely 5′ end of ISCR2, which is approximately 120 bp upstream from the start of the transposase gene (Fig. 9). A comparison between this end of ISCR2 and the terIS sequence consensus for IS91, IS801, and IS1294 shows differences but also similarities (34, 96). The terIS of ISCR2, like the terIS of IS91 group elements, has a short punctuated inverted repeat, but it is only 4 bp, in comparison with 6 bp for IS91 (Fig. 10). The overall sequence identity is about 50% and is accompanied by the nucleotide pair GA instead of the tetranucleotide GAAC for IS91 and IS1294 (Fig. 9 and 10). Whether the difference represents sequence drift or replacement with a sequence similar to terIS that can perform the same function is difficult to judge, given the paucity of relevant data.

The IS91-family element IS1294 has been studied in detail (96), and it has been shown that it can move adjacent sequences when the rolling-circle replication mechanism misidentifies terIS and proceeds to replicate the DNA adjacent to terIS. This has been calculated to occur in between 1% and 10% of normal transposition events for IS1294 (96). In particular, IS1294 was shown to mobilize the kanamycin resistance gene of pUB2380 to another plasmid, R388, at a frequency of approximately 10⁻⁵ to 10⁻⁶. IS1294 is unusual in that a single intact copy can promote transposition of sequences adjacent to one end of the element, specifically to those linked to the 3′ terminus of the element, which has also been shown for IS91 (65). This is very different from the case for most other insertion sequences, which can only mobilize additional DNA found between two copies of the insertion element (6-8, 38). IS1294 therefore gives us a precedent for a single copy of an insertion element moving adjacent DNA sequences. Thus, we can hypothesize that ISCR elements mobilize chromosomal genes by first transposing into a position adjacent to them, which is followed by a second but aberrant transposition event that mobilizes the adjacent sequence onto a conjugative plasmid and then, via conjugation, into other species/genera of bacteria. We hypothesize that, like IS1294, ISCR elements present a very powerful mobilization system which can, in theory, mobilize any section of DNA. In this case, they are likely to be more important than the integron system, which can mobilize DNA only in the form of gene cassettes.

ISCR ELEMENTS AND ASSOCIATED GENES

ISCR elements are notable for their close association with a wide variety of antibiotic resistance genes.

ISCRs MOBILIZE AND CONSTRUCT COMPLEX CLASS 1 INTEGRONS

ISCR1 appears to be unusual in the sense that it is almost always found in a complex class 1 integron structure. The sequence upstream of ISCR1 is always the same: i.e., the junction between ISCR1 and the sequence adjacent to the 3′ CS of the class 1 integrons is invariant. This suggests that a single event has been responsible for this feature of all complex class 1 integrons. Given that ISCR1 is significantly shorter than other ISCR and IS91-like elements, it is also likely that the terIS end of the element has been deleted following transposition (Fig. 11). This would fuse the remainder of the ISCR1 element (the transposase gene, orf513, and oriIS) to the 3′ CS end of the class 1 integron and provide the integron with the basis for genetic mobilization, since the oriIS and transposase gene remain intact. This ISCR1-class 1 integron fusion would also require that in the event of transposition being initiated, an alternative termination site would have to be utilized due to the fact that the original termination site, terIS-1, has now been deleted (Fig. 11). One consequence of the generation of ISCR1 by deletion of its terIS is that it would be expected to cotranspose the adjacent sequence, i.e., class 1 integron sequences, more often. This, then, allows the construction of a model in which ISCR1 has mediated the assembly of a variety of complex class 1 integrons (Fig. 12).

At some time in the past, possibly in the first half of the 20th century, the progenitor of ISCR1 transposed into a site close to the 3′ CS of a class 1 integron (Fig. 12). This transposition was followed by a deletion that removed the terIS terminus of the element and some of the 3′ CS normally associated with class 1 integrons, i.e., orf5 and orf6 (Fig. 12). This variant, ISCR1, then mediated a series of secondary transposition events that translocated ISCR1 and various sections of the adjacent class 1 integron into sites next to other, noncassette resistance genes, such as catA2, dfrA, qnr, and various CMY genes (Fig. 12). Once in these locations, the ability of IS91-like elements to generate free circular forms was manifest. Circular entities carrying ISCR1 and sequences adjacent to it and distal to oriIS, including at least the truncated 3′ CS and the linked resistance gene, were generated. These were then, in turn, rescued by homologous recombination into either the 3′ CS of conventional class 1 integrons or that of an integron-ISCR1 variant. The difference in the arrangements that would arise from these recombination events is that in the former case the array would carry only a single copy of ISCR1 adjacent to the class 1 integron between direct repeats of the 3′ CS of the class 1 integron, while in the latter case there would be a direct duplication of the ISCR1 element following the duplication of the 3′ CS, a particular arrangement that has been reported by Mammeri et al. (60). In that study, two unusual sequences were reported, where two copies of “Orf513” flanking a qnrA/qacΔE/sul region were present on different plasmids (pMG252 and pQR1) from the same E. coli strain (pMG252 and pQR1) (60). It is therefore possible that two copies of ISCR1 have been transposed independently onto these plasmids via an RC mechanism. Furthermore, given that the genetic contexts of ISCR1 on pMG252 and pQR1 are virtually identical, it is possible that ISCR1 has mobilized itself and adjacent DNA from pMG252 to pQR1 or vice versa. According to our model, the terminal full-length copy of the 3′ CS on a complex class 1 integron that lacks an ISCR1 duplication represents that originally linked to the int gene cassette array depicted on the left of the complex integron in Fig. 12. The internal, shortened version of the 3′ CS, together with the adjacent copy of ISCR1 and any other gene(s), has been acquired by rescue, via homologous recombination, of an ISCR1-generated circularized DNA fragment. The reverse event, i.e., deletion of sequences flanked by extensive direct repeats, would also be expected on occasion, and indeed it has been demonstrated experimentally that this can happen (71).

It is possible that ISCR3 and ISCR4 mobilize adjacent DNA sequences which are subsequently rescued by homologous recombination via flanking groEL sequences rather than the 3′ CS of class 1 integrons as seen for ISCR1. ISCR4 also appears to have lost its original terIS sequence, possibly by a deletion event linking groEL to the 5′ end of ISCR4. This has caused the groEL termination sequence to be in a position similar to where the original terIS should be, and this may function as an alternative terIS (Fig. 10). A similar event also appears to have happened with ISCR3, and groEL is now found immediately upstream of ISCR3 in SGI1 and variants as well as upstream of ISCR3 associated with the aminoglycoside resistance rmtB gene (Fig. 3).

SIMILARITIES AND DIFFERENCES BETWEEN ISCR AND IS91-LIKE TRANSPOSASES

An alignment of the predicted amino acid sequences of the transposase proteins of ISCR1 to -5 (Fig. 6) reveals that they differ in amino acid identity from 53% to 95% (Fig. 5). However, within the principal motifs, the alignment reveals key conserved residues that have been implicated as being essential for transposition of IS91 and like elements. The most notable exception to the match is Y249, which in ISCR1, ISCR2, and ISCR4 is replaced by arginine and in ISCR3 is replaced by lysine. In the case of IS91, it has been proposed that RC transposition involves two conserved tyrosine residues (Y249 and Y253), one of which serves as a nucleophile to attack the oriIS sequence of the IS element while the other attacks the _target site, both effecting strand cleavage and the formation of a covalent bond between the amino acid residue and the 5′ phosphate group released by DNA strand cleavage (34, 34). The nucleophilic centers provided by the hydroxyl groups of the tyrosine residues are clearly critical to the proposed reactions. When Y249 or/and Y253 in the IS91 transposase were mutated to phenylalanine or serine, transposition activity was no longer detected, consistent with the view that both residues are essential to the transposase activity (33). However, it is yet to be proved that both tyrosine residues do participate directly in DNA strand cleavage, as proposed by Garcillan-Barcia et al. (33, 34), in that altering the Y249 amino acid residue could simply adversely affect the overall active-site architecture. The ISCR substitution of Y249 brings into question the proposed role of Y249 in the IS91 transposase; however, it is possible that ISCRs have found alternative residues to facilitate RC transposition. Interestingly, the Rep proteins encoded by plasmid pT181 possess the equivalent of Y253 but not Y249 and act as dimers, with the second molecule nicking the displaced old-to-new leading strand junction to initiate termination of replication (32, 49). Other Rep proteins, such as those mediated by plasmid pC194, require a glutamate residue as well as tyrosine for enzyme activity; the former is thought to catalyze the release of the protein from the intermediate at the termination step (69). Recently, the crystal structure of a TnpA, an RC replication protein, has been solved (84). TnpA also appears to possess the single tyrosine and, as demonstrated by the cocrystal structure with DNA stem-loops, acts as a dimer. Clearly, more research into the active sites of IS91-like molecules is needed. The other notable exception in terms of conserved amino acid residues that is evident from the alignment of ISCR transposases with those of IS91 and IS1294 is H149Q, where glutamine has replaced histidine in all the ISCR-encoded proteins (Fig. 6). This may be a change enforced by the Y249K/L substitution needed to retain transposase activity.

SPECIFICITY OF INSERTION OF IS91 AND ISCR ELEMENTS

IS91 and the related family members IS1294 and IS801 _target and insert at a specific tetranucleotide sequence (GTTC or minor variants thereof), and similar sequences are also recognized at terIS. These elements insert into the _target DNA 5′ to this tetranucleotide so that the _target site is then found immediately adjacent to the oriIS sequence of the element and is necessary for further transposition events (63). This sequence is also precisely the cleavage site for the PC194 plasmid family and for φX174-like phages (34).

Interestingly, in all examples of ISCR1 and ISCR2 available in the sequence databases, there appears to be no obvious common tetranucleotide found adjacent to the oriIS. It is therefore possible that the ISCR transposase no longer relies on the junction between _target site and oriIS for RC transposition origin or that the majority of transposition events are dead ends in the sense that retransposition is prevented. In support of the former suggestion, the tetranucleotide sequences that are thought to be involved in _target site recognition in IS91 are the same amino acids that are changed in ISCR elements (63). Perhaps, therefore, ISCR elements are no longer _target site dependent. However, nonfunctional insertions may be rescued if a second copy of the element with the correct oriIS/tetranucleotide junction is present in the cell, as has been shown for IS91 (63).

ISCR ELEMENTS AND PROMOTER SEQUENCES

Insertion elements are known, on occasion, to provide promoters for expression of otherwise silent genes by transposing into a site immediately upstream of the promoterless gene: e.g., the cfiA MBL gene of Bacteroides fragilis is normally silent, but ceftazidime-resistant isolates are occasionally found (106). When these are genetically analyzed, they are found to have an insertion element upstream of the cfiA gene that provides a promoter sequence for its expression and, as a result, acquisition of clinical resistance. Promoter sequences have been found at the 5′ end of the ISCR1 element (60, 107). A hybrid promoter was also suggested to have been formed between a −10 sequence that is found naturally upstream of bla_SPM-1 and a −35 sequence found within the right-hand boundary of ISCR4. The spacing between these sequences is 17 bp. This combination of mobility and expression via ISCR elements is loosely comparable to the integron structure, where promoterless genes are expressed by virtue of proximity to promoters in the 5′ end of the integrase gene. However, the ISCR elements do not have to provide a promoter for the genes they associate with because many of these resistance genes have their own promoters.

ORIGINS OF CR ELEMENTS

A number of observations suggest that the ISCR1 element may have originated from a waterborne bacterium. Both SXT and the SGI1 elements are found in V. cholerae and S. enterica serovar Typhimurium, which often inhabit aquatic environments. ISCR1 is found on plasmids closely associated with the waterborne organism A. salmonicida (82). Moreover, the ciprofloxacin resistance gene qnrA originates from a waterborne organism, S. algae, and ISCR1 has been found adjacent to several members of the bla_CMY-1 group genes, which are suggested to have come from an Aeromonas sp. However, they have not been found associated with the bla_CMY-2 group members, which are thought to have originated from enteric organisms such as C. freundii. bla_PER-3 is closely associated with ISCR1 and was isolated from A. punctata in France (accession no. AY740681). GenBank screens of incomplete microbial genome sequence databases with ISCR elements have also revealed many positive results (>50) in the A. hydrophila genome sequence that are ∼54% identical to ISCR5. This organism may be the source of another ISCR family member, provisionally named ISCR13. It is also noteworthy that the first florfenicol resistance gene, floR, was detected on a plasmid from the fish pathogen P. damsalae subsp. piscida (formerly known as P. piscida), which may also suggest a marine environmental source of ISCR2.

The various ISCR elements vary in their G+C contents from 54% for ISCR1 to 69% G+C for ISCR5 (Table 1), which suggests that the different ISCR family members have originated from different source organisms. ISCR1 and ISCR2 have the lowest percent G+C of all ISCR family members isolated so far, with 54% and 59.5% G+C, respectively. The rest of the ISCR family members have G+C contents of above 60%. ISCR3 and ISCR4 are both found adjacent to partial groEL genes in their hosts S. enterica serovar Typhimurium/S. marcescens and P. aeruginosa, respectively. These partial groEL genes both display highest identity with the groEL genes of Xanthomonas campestris and S. maltophilia. Both of these organisms are found in the soil and have high G+C content, which is consistent with the hypothesis that the groEL genes and ISCR3/ISCR4 were acquired at the same time, possibly from a soil/plant rhizosphere-associated organism. The significant differences in G+C content shown for the various ISCR elements may indicate that their sequences have been diverging for some considerable time.

IS91 AND VIRULENCE FACTORS

IS91, the prototype element of this expanding family of IS91-like insertion elements, was first discovered in plasmids encoding the α-hemolytic function in E. coli in 1982 (113) and was subsequently identified in all plasmids and chromosomal pathogenicity islands harboring the α-hemolysin operon (Hly) (34, 37, 53), which codes for synthesis, activation, and export of α-hemolysin. The Hly plasmids belonged to different incompatibility groups, which suggested that IS91 was involved in the dissemination of these pathogenicity determinants (21, 37, 53, 114). IS91 or very closely related isoforms have also been found adjacent to various other virulence genes in enteropathogenic, enterohemolytic, and enterotoxigenic strains of E. coli (16, 42, 109), including the eltAB toxin-encoding genes of the heat-labile enterotoxin (89). Furthermore, direct movement of eltAB genes by IS91 has been demonstrated (89). Similarly, IS1294 has been associated with virulence genes in E. coli (96), and IS801 has been discovered in close association with virulence genes in the plant pathogen Pseudomonas syringae (66, 83, 95).

CONCLUSIONS

Not less than 10 years ago there were numerous examples of IS91 found adjacent to virulence genes but not found adjacent to antibiotic resistance genes. This phenomenon has prompted Garcillan-Barcia et al. (34) to speculate that IS91-like elements are best suited to movement of virulence genes and that transposons and integrons are more suited to the dissemination of antibiotic resistance genes. However, there is no reason why these interesting elements cannot mobilize any piece of adjacent DNA. Therefore, it is not surprising that this powerful gene mobilization mechanism has connected with such easily selectable markers as antibiotic resistance genes. In fact, the presence of ISCR1 as part of the complex class 1 integrons In6 and In7 on the plasmids pSa and pDG1000 indicates that they were already associated with chloramphenicol and trimethoprim resistance genes in the late 1950s and early 1960s, shortly after these antibiotics were first used. Clinically, the most worrying aspect of ISCR elements is that they are increasingly being linked with more potent examples of resistance, i.e., MBL in P. aeruginosa and co-trimoxazole resistance in S. maltophilia. Furthermore, if ISCR elements do move via “unchecked RC transposition,” as has been speculated for ISCR1, then this mechanism provides antibiotic resistance genes with a highly mobile genetic vehicle that could eclipse the effects of previously reported mobile genetic mechanisms.