Multiple Mobile Promoter Regions for the Rare Carbapenem Resistance Gene of Bacteroides fragilis

ABSTRACT Two novel insertion sequences (IS), IS1187 and IS1188, are described upstream from the carbapenem resistance gene cfiA in strains of Bacteroides fragilis. Mapping, with the RACE procedure, of transcription start sites of cfiA in these and two other previously reported IS showed that transcription of this rarely encountered gene is initiated close to a variety of B. fragilis consensus promoter sequences, as recently defined (D. P. Bayley, E. R. Rocha, and C. J. Smith, FEMS Microbiol. Lett. 193:149–154, 2000). In the cases of IS1186 and IS1188, these sequences overlap with putative Eς70 promoter sequences, while in IS942 and IS1187 such sequences can be observed either upstream or downstream of the B. fragilispromoters.

Bacterial insertion sequences (IS), integral parts of the chromosome of many bacterial species, are mobile DNA elements which typically code only for the transposase which promotes their mobility. These elements are capable of generating mutations and genome rearrangements as a result of translocation, of promoting gene acquisition, and of mobilizing DNA fragments via the formation of compound transposons (13,14). By various mechanisms, the presence or mobility of these elements may affect degradative pathways (9,10), bacterial pathogenicity or virulence (8,35), and resistance to antibiotics (15,17) and may condition gene expression. Genes may be silenced by insertional gene disruption and reactivated by precise excision (8,11,35). Gene activation also may result from the provision of efficient promoters, carried either entirely by the element (10,17) or generated as hybrid structures between IS, upon insertion, and target sequences (5,9,12,15).
Like most bacterial genera in which insertion sequences have been searched for, Bacteroides harbors several such elements. They are IS4351, IS942, IS1186, and IS1224 (or their isoforms), and they belong to the IS30, IS4, IS5, and IS21 families, respectively (13,27). Only IS4351 has also been observed as part of a compound transposon (20). The four elements have been found preferentially in Bacteroides fragilis, the anaerobe species most frequently isolated from human infections (24).
No involvement other than that in the expression of antibiotic resistance genes has, as yet, been observed for the Bacteroides IS elements. While it was speculated that IS1224 may activate the endogenous B. fragilis cephalosporinase gene, cepA, either via inactivation of a repressor function or generation of a hybrid promoter (22,27), it has been shown that the carbapenemase gene cfiA (ccrA) is activated by an IS1186borne promoter (17). It seemed likely that there was a similar contribution to transcription of cfiA by IS942 (19) and IS4351 (I. Podglajen, unpublished data), of the macrolide resistance gene ermF by IS4351 (20), and of the metronidazole resistance genes nimA and nimB by IS1168 (7), nimD by IS1169 and nimC by IS1170 (32).
The carbapenemase CfiA confers resistance to practically all ␤-lactams, the most widely used class of antibiotics. Surprisingly, this conceivably advantageous resistance trait seems to have remained confined to a small, genotypically identified subgroup (or subspecies) of B. fragilis (6,16,23) despite the presence of a variety of genetic elements with the potential for gene mobilization and transfer within the genus Bacteroides (25)(26)(27). In the present study we describe two novel insertion sequences and their capacity to efficiently promote the expression of cfiA.
Identification and characterization of two novel IS elements in B. fragilis. During the systematic analysis of carbapenem resistance in clinical isolates of B. fragilis we observed five imipenem-resistant strains, by using dot blot hybridization (data not shown), which carried cfiA but none of the promoterbearing insertion sequences IS1186 (17), IS942 (19), or IS4351 (20 and I. Podglajen, unpublished) or closely related IS isoforms. Since cfiA has generally been found to be silent unless at least one of these IS is present, we explored the mode of its activation in the five clinical strains which are part of our collection of 65 cfiA-positive strains.
Species identification was carried out using API 20A and API 32A strips (bioMérieux, Marcy-l'Etoile, France). Identification of all strains given in Fig. 1 was confirmed by nucleotide sequence analysis of 16S ribosomal DNA segments amplified with the primers BF16SF (TAACTCCGTGCCAGCAGC) and BF16SR (GTGGACTACCAGGGTATC) and Extra-pol II DNA polymerase (Eurobio, Les Ulis, France) under standard conditions (1). The regions upstream of cfiA were amplified with primers E (CTTCGAATTCGGCGAGGGATACATAA) and G (CGCCAAGCTTTGCCTGCCATTAT), specific for sequences upstream of the known IS insertion sites and cfiA, respectively (17). The nucleotide sequences of the PCR-generated fragments were determined by a commercial provider (Génome Express, Montreuil, France) or by using the Thermo Sequenase [ 33 P] Terminator Cycle Sequencing kit from Amersham. New primers were synthesized as sequence determination proceeded. All sequences were determined in duplicate on both strands.
Two novel IS elements were identified, IS1187 in four of the imipenem-resistant strains and IS1188 in one strain ( Fig. 1). IS1187 was 1,026 bp in length, had 21-bp inverted repeats (with one mismatch), and contained one major open reading frame, spanning most of the element, with a protein-coding capacity of 326 amino acids (aa), a deduced molecular size of 37.5 kDa, and a theoretical pI of 9.3. The stop codon of this reading frame was located within the right inverted repeat. The inverted repeats were flanked by direct repeats of 9 bp in BFr1757 and BFr1761 and 8 bp (with one mismatch) in BFr1758 and BFr1760 (Fig. 2). The second element, IS1188, was 1,691 bp in length, had 17-bp inverted repeats (with three mismatches), and contained two major open reading frames, one with a protein-coding capacity of 448 aa, a deduced molecular size of 52.5 kDa, and a theoretical pI of 9.5; the other, on the opposite strand, with a protein-coding capacity of 140 aa, a molecular size of 16.3 kDa, and a theoretical pI of 9.7. The inverted repeats of this element were flanked by direct repeats of 4 bp (Fig. 2).
Insertion of both elements occurred within ca. 90 bp upstream from the initiation codon of cfiA and in all but one case occurred very close to the sites at which insertion of all IS elements known or suspected to carry cfiA-activating promoters has been observed previously (Fig. 2).   (17); c, e, and f, sequences in BFr908, BFr271R, and BFr930, respectively (I. Podglajen, unpublished); d, sequence in strains TAL2480 (31) and TAL3636 (18,19); g, h, i, and k, sequences in strains BFr902, BFr1757, BFr1758 and BFr1760, and BFr1761, respectively. Nucleotide variations with respect to BFr81 are shown in lower case; Ϫ, no nucleotide at that position.
The copy numbers of IS1187 and IS1188 were estimated in cfiA-positive and cfiA-negative strains (see below) using ISspecific probes after digestion of chromosomal DNA with AvaI and Southern hybridization (17). Probes specific for the respective transposase-coding regions were generated with primers BF1187F (CGTATTGCAGAATGGTAAGTGC) and BF1187R (GTTCCACGTCGTGGTCCTGTTC) amplifying a 725-bp fragment of IS1187 and primers BF1188F (GGCCTGTGCTCAC AACCGAC) and BF1188R (CGGTATGCGGTCACATAT GC) amplifying a 778-bp fragment of IS1188. The probes were randomly labeled with the Megaprime kit and [␣-32 P]dCTP from Amersham. The copy numbers of the two elements varied between one in the two cfiA-negative strains, as for IS1187 in BFr1763 and IS1188 in BFr1765, and three to at least five in the cfiA-positive strains (Fig. 3). Since neither IS contained an AvaI site, each band should represent at least one element. The differences in hybridization intensity might reflect the existence of incomplete copies or, perhaps more likely, of copies with some sequence divergence (IS isoforms). Such an element, with 74% homology to IS1187, has been observed in a clinical B. fragilis isolate in the United Kingdom while this work was under way (3).
When criteria for grouping IS (13,14) were applied, IS1187 and IS1188 appeared to be members of the IS5 family. There was agreement with the overall size of the elements and the size of the inverted and direct repeats. Furthermore, identities in the vicinity of the amino acids of the DDE triad, a conserved motif of the active site of IS transposases (13,14), indicated a relationship of the putative transposase of IS1187 with that of IS5 and of the transposase of IS1188 with that of IS1031, representative of a subgroup of the IS5 family (Fig. 4) (13).
When the amino acid sequences of the putative transposases were compared with sequences in the databases, that of IS1187 was most closely related to a plasmid-coded transposase of ISRa1 from Riemerella anatipestifer (33), with 45% identities and 66% equivalencies over a stretch of 300 aa, to a transposase from Clostridium cellulovorans (30) (34% identities, 50% equivalencies, over a stretch of 251 aa) and to the transposase of IS982 from lactococci (34) (29% identities, 45% equivalencies, over a stretch of 189 aa). The putative transposase of IS1188 was most closely related to similar proteins coded by an IS element from a Sphingomonas sp. (4) (32% identities, 51% equivalencies, over a stretch of 418 aa), by IS1380 from Acetobacter pasteurianus (29) (31% identities, 47% equivalencies, over a stretch of 447 aa), and by the chromosome of Bacillus halodurans (28).
Incidence of IS1187 and IS1188 among B. fragilis isolates. The IS elements reported previously in Bacteroides appear to exist primarily in the small cfiA-positive subgroup of B. fragilis, where they are chromosome borne (16). To test the possibility that this was true also for IS1187 and IS1188, we carried out dot blot hybridization experiments using the IS-specific probes. Seventy-five randomly collected cfiA-negative B. fragilis strains, mostly from our laboratory collection or kindly provided by I. Casin and L. Dubreuil, were assayed for the presence of the two elements. Hybridization was positive with each element in one strain but did not occur with any of the several plasmids in the size range of ca. 3 to 8 kb that were present in both strains (data not shown). This suggested a chromosomal location of the elements in these strains. Except for IS1224 (27), no IS has been reported as being chromosome borne outside of the cfiApositive group. In the cases where IS942 or IS1186 or isoforms thereof have been described outside of this group (7,32), they were carried on small plasmids. There was no coresidence in the IS1187-or IS1188-bearing strains with one of the three known IS elements IS942, IS1186, or IS4351, combinations of which, by contrast, were observed in about one-third of the 50 cfiA-positive strains analyzed previously (16). Whether the absence of such a coresidence reflects a recent introduction of FIG. 3. Copies of IS1187 and IS1188 in the B. fragilis strains studied. AvaI-digested total DNA was hybridized, after Southern blotting, with probes specific for IS1187 (five leftmost lanes) or IS1188 (two rightmost lanes). Strains are given in Fig. 1, except for BFr1763; BFr1763 and BFr1765 were imipenem susceptible and cfiA negative.
FIG. 4. DDE motifs in the putative transposases of IS1187 and IS1188. Data for IS5 and IS1031 were taken from Mahillon and Chandler (13), who also assigned IS1186 (17) to the IS5 family. IS1187 or IS1188 or both into B. fragilis remains a matter of speculation.
Expression of cfiA in IS1187-or IS1188-bearing strains. Expression of cfiA was studied by Northern hybridization using a slot blot procedure with cesium chloride gradient-purified RNA (10 g per slot) and a 32 P-labeled cfiA probe. The relative efficiency of cfiA transcription was estimated after densitometry of the autoradiograph (Fig. 1). There was an apparent correlation between the degree of cfiA transcription and the level of resistance to imipenem, with a somewhat lower relative level of transcription in strain BFr902. Considering that IS1187 and IS1188 had inserted within a ca. 90-bp stretch upstream of a cfiA copy located at the same chromosomal site in all strains (as deduced from the immediate upstream nucleotide se-quences) and close to the insertion sites previously reported for other IS (Fig. 2), it seemed likely that these two IS also provided promoters for cfiA transcription. Close inspection of the IS1187 and IS1188 sequences revealed the existence, as in IS1186 (17) and IS942 (19), of hexamers (see Fig. 6) potentially corresponding to Ϫ35 and Ϫ10 consensus sequences of promoters recognized by the primary sigma factor in Escherichia coli (21). To accurately map the sites of transcription initiation of cfiA in IS1187 and IS1188 and to contribute to the stilllimited knowledge of promoter structures in Bacteroides, we applied the 5Ј RACE System (Rapid Amplification of cDNA Ends, version 2; GibcoBRL-Life Technologies) according to the manufacturer's instructions, except that RNA was prepared as for the Northern blots. RNA from strains BFr1757 (IS1187) and BFr902 (IS1188), and comparatively from strains BFr81R (IS1186) and TAL2480 (IS942), was used. The primers for reverse transcription and for nested PCRs 1 and 2, respectively, were GSP1cfiA (GATAACAATCATCCCGTTG GAAGGTACCATACCCC), GSP2cfiA (CCTTCGATTTCG GCGAGGGATACATAAGTGTACAC), and primer H (17). The fragments generated during the final PCR are shown in Fig. 5. Sequence determination of these fragments indicated that all transcripts started with an adenine, flanked in three out of the four cases by a consensus C(Ϫ1) and T(ϩ2) (21), and located a few nucleotides downstream from an octameric motif, strongly or perfectly resembling the octameric sequence TAnnTTTG (Fig. 6). This "Ϫ7 motif," together with a "Ϫ33 motif," TTTG, has recently been identified 1 to 12 bp upstream from the transcription initiation sites of over 20 B. fragilis genes and shown to be essential, by in vitro mutagenesis, for promoter activity in cepA (2). There were two consecutive Ϫ7 motifs in IS1187. In IS1186 and IS1188, the octameric motifs overlapped with a putative consensus E 70 promoter at the Ϫ10 hexamer, and the TTG triplet of the Ϫ33 motif overlapped with the Ϫ35 hexamer. There was no such overlap in IS942 and IS1187, but putative E 70 promoter sequences were observed at some distance downstream or upstream, respectively (Fig. 6). Whether the Ϫ10 and Ϫ35 hexamers may be functional in 70 homologue-producing species more closely related to the Enterobacteriaceae remains to be tested experimentally. This possibility leads to the further speculation that the multiplicity of consensus sequences might endow Bacteroides IS elements with the potential to provide mobile promoters to distantly related bacterial species. With respect to cfiA, the number of known mobile promoters, currently five, that this rare and normally promoterless gene is able to recruit in order to ensure its transcription is quite remarkable. It is conceivable that the use of carbapenems has somehow favored the accumulation of IS elements in the cfiA-positive B. fragilis population, or alternatively, that this population might possess particular factors enhancing their acquisition or transposition.
Nucleotide sequence accession numbers. The nucleotide sequences of IS1187 and IS1188 have been deposited in the GenBank and EMBL databases under the accession numbers Y18979 and AJ277413, respectively. We gratefully acknowledge I. Casin and L. Dubreuil for the gift of strains and P. Bertin for critically reading the manuscript.