Complete sequence of pBFUK1, a carbapenemase-harboring mobilizable plasmid from Bacteroides fragilis, and distribution of pBFUK1-like plasmids among carbapenem-resistant B. fragilis clinical isolates

Bacteroides fragilis, among the most common anaerobic pathogens recovered from various infection sites such as intra-abdominal abscesses, is resistant to a broader range of antimicrobials, compared with most other anaerobes.^{1, 2, 3, 4} In fact, only carbapenems and nitroimidazoles are highly effective for eliminating B. fragilis.⁴ While, to date, carbapenem resistance in B. fragilis still remains rare (approximately 2%),^{3, 4, 5} a class B metallo-β-lactamase (carbapenemase), encoded by the cfiA (also known as ccrA) gene, has been identified in B. fragilis.^{6, 7} The CfiA protein confers high level of resistance (MIC, 25−200 μg ml⁻¹) to carbapenems, other β-lactams, and β-lactamase inhibitor combinations.^{5, 8} In addition, the CfiA expression is determined by the integration, immediately upstream of the cfiA gene, of an insertion sequence (IS) element carrying an efficient promoter (for example, IS613, -615, -942 and -1186).^{9, 10, 11}

Although the cfiA gene is generally located on the B. fragilis chromosome, we previously identified for the first time a clinical isolate, the GAI92082 (10–73) strain, in which CfiA is encoded by a mobilizable plasmid—pBFUK1.¹² Nakano et al.¹³ subsequently identified a second clinical isolate (B. fragilis C68c strain) harboring a CfiA-encoding plasmid in size of approximately 6.4 kb. Despite the risk of carbapenem resistance spreading among B. fragilis strains via these mobile elements, neither the detailed structure of CfiA-encoding plasmids nor their distribution among carbapenem-resistant B. fragilis isolates has been fully investigated.

In this study, we analyzed the complete nucleotide sequence of pBFUK1 to determine whether it contains the cfiA promoter and ISs, as well as its possible origin. We also conducted the first distribution analysis of pBFUK1-like plasmids among carbapenem-resistant B. fragilis isolates.

We sequenced pBFUK1 that was purified from the B. fragilis Tc 30 strain described in our previous report.¹² In total, 101 reads (providing an approximately 3.9-fold coverage across the length of pBFUK1) were sequenced by inverse PCR and primer walking (Supplementary Text S1 and Table S1). Moreover, to investigate the distribution of pBFUK1-like plasmids, we extracted plasmids from two strains (GAI92082 and Tc 30) as positive controls, as well as the following 12 carbapenem-resistant B. fragilis strains isolated from patients in Japan from 1987 to 2006: GAI05079, -05184, -06112, -30079, -30144, -92084, -92085, -92087, -92212, -92213, -92214 and -94199 (Supplementary Table S2 and Text S1).^{9, 12} All strains used in this study are resistant to carbapenem (MIC of imipenem, 16−200 μg ml⁻¹; Supplementary Table S2). Extracted plasmids were separated by electrophoresis in a 0.6% agarose gel. To examine whether these 14 strains contain plasmids with sizes >20 kb, we also performed pulsed-field gel electrophoresis (PFGE; Supplementary Text S1). DNA fragments in the 0.6% mini-gel and PFGE gel were then transferred to a nylon membrane, followed by the Southern blotting analysis to detect seven genes, including cfiA; transposase A and B (tnpA and tnpB); replication protein A (repA); and mobilization protein A, B and C (mobA, mobB and mobC). Probes for Southern blotting were synthesized based on the pBFUK1 sequence (Supplementary Text S1 and Table S1). The pBFUK1 sequence and annotation data have been deposited in the DNA Data Bank of Japan database under the accession number AB646744.

The sequenced pBFUK1 plasmid consists of 12 817 nucleotides (average G+C content, 46.3%) and encodes 16 putative open reading frames (ORFs; Table 1 and Supplementary Figure S1). All ORFs showed 95−100% amino-acid identity to known proteins identified in the genus Bacteroides, B. fragilis in particular (Table 1), suggesting that pBFUK1 was most likely derived from and rearranged in the genus Bacteroides. In general, plasmids can be classified into three categories according to mobility: mobilizable, conjugative and nonmobilizable.¹⁴ pBFUK1 belongs to the mobilizable plasmid group because it encodes mobilization proteins that are sufficient to allow its mobilization when coresident with a chromosome or a helper plasmid encoding the system for mating-pair formation.¹⁴ Three mobilization proteins in pBFUK1, designated as MobA, -B and -C, showed almost 100% amino-acid identity (with the exception of the MobC carboxyl-terminus) to those in the 4.2 kb pLV22a plasmid from the B. fragilis LV22 strain (MbpA, -B and -C, respectively; Supplementary Figure S2).^{15, 16} As Garcillán-Barcia et al.¹⁷ recently reported that pLV22a MbpB (DNA relaxase) belongs to the MOB_P family, we also assigned pBFUK1 MobB to the same family. Like all three-Mbp proteins in pLV22a,¹⁵ three Mob proteins in pBFUK1 may also be essential for mobilization. We previously reported that pBFUK1 is mobilizable into some B. fragilis strains;¹² however, its detailed host range is not yet clear. Moreover, Thomas and Hecht¹⁶ recently reported that pLV22a is mobilizable into Escherichia coli when coresident with the IncP conjugative plasmid RP4, and therefore, pBFUK1 may also be mobilizable into E. coli (or other strains in the genus Bacteroides) with a helper factor (for example, IncP plasmid, conjugative transposon and chromosome) that carries genes encoding the system for mating-pair formation.

Table 1 Details of putative ORFs in pBFUK1 and their closest relationships to previously studied proteins

In the pBFUK1 plasmid, we identified a novel composite transposon (designated as Tn6186) in size of approximately 8.5 kb (Table 1 and Supplementary Figure S1). To the best of our knowledge, Tn6186 in pBFUK1 is the first composite transposon that contains the cfiA gene and its promoter region. In addition, the 5075-bp CfiA−ORF9 region in Tn6186 showed 96.3% nucleotide identity to the corresponding 5170-bp region in the 155-kb contig (GenBank accession no. NZ_ABZX01000055) derived from the B. fragilis 3_1_12 strain (Supplementary Figure S3). As the contig of the 3_1_12 strain contains some housekeeping genes and is probably chromosomal, it is possible that the CfiA−ORF9 region in Tn6186 was also derived from a B. fragilis chromosome (Supplementary Figure S3).

Tn6186 contains two previously identified IS elements, IS613 (1595 bp) and IS615 (1594 bp) (Supplementary Figures S1 and S4a).⁹ We have reported that IS613 and IS615 are also located immediately upstream of cfiA2 and cfiA4 in the B. fragilis GAI92082 and GAI20264 genome, respectively.⁹ Both IS elements in pBFUK1 belong to the IS4 family,⁹ which includes another IS element, IS942, previously found in some drug-resistant B. fragilis strains. The putative transposase genes (designated as tnpA and tnpB, respectively) in IS613 and IS615 of pBFUK1 also show 100% nucleotide identity to those in the GAI92082 and GAI20264 genome, respectively. However, there is only 69.0% nucleotide identity between IS613 and IS615 in pBFUK1. As the homology is not very high, we propose that intramolecular homologous recombination between these two ISs in pBFUK1 does not occur at high frequency. Indeed, we have not yet observed any deletion of the six ORFs (cfiA to ORF9) from pBFUK1 (data not shown). Moreover, the target-site sequence of the transposase in the known cfiA-related IS elements is located at both ends of each IS element,^{9, 11} whereas the target-site sequence (GGGAA) of the transposase(s) in pBFUK1 is present at both ends of Tn6186 (downstream of IS613 IRR and upstream of IS615 IRL; Supplementary Figure S4a). These findings suggest that Tn6186 is a composite transposon. Furthermore, two putative transposases encoded by pBFUK1 exhibit 80.6% identity at the amino-acid level. Further functional studies are needed to determine whether one or both of putative transposases can mediate the Tn6186 transposition.

The cfiA gene in pBFUK1 showed 100% nucleotide identity to cfiA2 in the GAI92082 genome (Supplementary Figure S5). The region upstream of cfiA in pBFUK1 includes a putative cfiA promoter sequence that not only contains the consensus −7 and −33 motifs^{9, 11} (Supplementary Figure S4b) but also shows 100% nucleotide identity to that of IS613 in the GAI92082 genome (Supplementary Figure S5).⁹ Although whether the 1.6-kb amplicon (Supplementary Figure S5) that was PCR amplified from the GAI92082 genome⁹ is derived from the chromosome itself or from the pBFUK1 remains unclear, we suspect the former (Supplementary Figure S5). Moreover, based on the complete nucleotide identity of this cfiA gene and promoter region with those of the GAI92082 genome, we expect that pBFUK1 can express a functionally active CfiA protein. This expectation is also based on our previous study in which not only the carbapenem-sensitive B. fragilis TM4000 strain that was electrotransformed with purified pBFUK1 but also the Tc 30 strain showed sufficient imipenem-hydrolytic activity (Supplementary Table S2).¹²

Next, to investigate the distribution of pBFUK1-like plasmids, plasmid preparations from 14 carbapenem-resistant strains (Supplementary Table S2) were subjected to electrophoresis in a 0.6% agarose gel. Although three strains (GAI05079, -92212 and -92214) contained no plasmids, the remaining strains possessed at least 1–3 plasmids with sizes ranging from approximately 2.8 to 12.8 kb (Supplementary Figure S6a). PFGE also showed that none of these 14 strains contained plasmids >20 kb (Supplementary Figure S6d). One plasmid in size of 35–48 kb was detected in the GAI92082 and Tc 30 strains (Supplementary Figures S7a and S7b), and was further verified as the pBFUK1 plasmid that migrated anomalously¹⁸ during PFGE (Supplementary Figure S7). Subsequently, we performed the Southern blotting analysis using probes derived from pBFUK1 to examine the presence of seven genes (cfiA, tnpA, tnpB, repA, mobA, mobB and mobC) in plasmids of the 14 resistant strains (Supplementary Figure S6a). The Southern blotting analysis on PFGE gel (Supplementary Figure S6d) was also carried out to examine whether homologs of these seven probe targets are present in the chromosomes of these 14 strains. The representative and overall results of Southern blotting are shown in Supplementary Figure S6 and Table 2, respectively. Isolated from five strains (GAI06112, -92084, -92085, -92213 and -94199), the plasmids in sizes of approximately 4.9 kb contained genes encoding RepA, MobA, -B and -C homologs; however, these plasmids had no cfiA or transposase genes. Therefore, these 4.9-kb plasmids, despite being partially similar to pBFUK1, are unlikely to have a role in carbapenem resistance. Overall, we found that CfiA-encoding pBFUK1-like plasmids were not widely distributed among carbapenem-resistant clinical isolates collected in Japan from 1987 to 2006 (Table 2), although we previously reported that pBFUK1 was transferable by mating in vitro between some B. fragilis strains.¹² One explanation for this observation may be that the frequent pBFUK1 transfer among strains in vivo may require a higher density of cell–cell contact. In spite of the low incidence of CfiA-encoding plasmids, continuous surveillance is necessary to monitor their presence, especially in light of dramatically increased tetracycline and clindamycin resistance in Bacteroides species in the past 30 years owing to mobile genetic elements.¹⁹ One should also be careful not to activate a silent cfiA gene, because expression of such a gene can be activated dramatically by exposure to low concentrations of carbapenem.²⁰

Table 2 Location of seven genes encoded by pBFUK1 in different carbapenem-resistant Bacteroides fragilis strains

Furthermore, the GAI92214 and GAI94199 strains had no cfiA, as determined by the lack of a signal in the Southern blotting using the cfiA probe (Supplementary Figure S6e and Table 2). This finding is consistent with the observations that neither PCR amplicons of cfiA and IS nor carbapenemase activity were detected in these strains (Supplementary Table S2). However, these strains did show moderate-level carbapenem resistance (MIC of imipenem, 16−25 μg ml⁻¹; Supplementary Table S2), suggesting that other carbapenem resistance mechanisms, such as efflux pumps,² may exist in these strains.

The GAI92082 chromosome contained cfiA and two transposase genes (tnpA and -B), in addition to those present on the pBFUK1 plasmid (Supplementary Figures S6, S7a, and Table 2). To our knowledge, this strain is the only strain that carries cfiA genes in both its chromosome and plasmid. The tnpA probe (in IS613) and tnpB probe (in IS615) were also hybridized with the GAI92213 chromosome (Table 2); however, further work is needed to determine whether the chromosomes of both strains encode all ORFs in Tn6186. If the GAI92082 chromosome contains the composite transposon, pBFUK1 Tn6186 may have been either replicated from or inserted into the GAI92082 chromosome, which was originally carbapenem sensitive.

We discovered a novel composite transposon, including the cfiA gene and its promoter, in the pBFUK1 sequence. We believe that our first distribution analysis of pBFUK1-like plasmids will also be a valuable first step toward better understanding of the potential spread of carbapenem resistance via CfiA-encoding plasmids.