Mobile Insertion Cassette Elements Found in Small Non-Transmissible Plasmids in Proteeae May Explain qnrD Mobilization

qnrD is a plasmid mediated quinolone resistance gene from unknown origin, recently described in Enterobacteriaceae. It encodes a pentapeptide repeat protein 36–60% different from the other Qnr (A, B, C, S and VC). Since most qnrD-positive strains were described as strains belonging to Proteus or Providencia genera, we hypothesized that qnrD originated in Proteeae before disseminating to other enterobacterial species. We screened 317 strains of Proteeae for qnrD and its genetic support by PCR. For all the seven qnrD-positive strains (4 Proteus mirabilis, 1 Proteus vulgaris and 2 Providencia rettgeri) the gene was carried onto a small non-transmissible plasmid, contrarily to other qnr genes that are usually carried onto large multi-resistant plasmids. Nucleotide sequences of the qnrD-bearing plasmids were 96% identical. Plasmids contained 3 ORFs apart from qnrD and belonged to an undescribed incompatibility group. Only one plasmid, in P. vulgaris, was slightly different with a 1,568-bp insertion between qnrD and its promoter, leading to absence of quinolone resistance. We sought for similar plasmids in 15 reference strains of Proteeae, but which were tested negative for qnrD, and found a 48% identical plasmid (pVERM) in Providencia vermicola. In order to explain how qnrD could have been inserted into such native plasmid, we sought for gene mobilization structures. qnrD was found to be located within a mobile insertion cassette (mic) element which sequences are similar to one mic also found in pVERM. Our conclusions are that (i) the small non-transmissible qnrD-plasmids described here may result from the recombination between an as-yet-unknown progenitor of qnrD and pVERM, (ii) these plasmids are maintained in Proteeae being a qnrD reservoir (iii) the mic element may explain qnrD mobilization from non-transmissible plasmids to mobilizable or conjugative plasmids from other Enterobacteriaceae, (iv) they can recombined with larger multiresistant plasmids conjugated in Proteeae.


Introduction
Quinolones inhibit replication and transcription by inhibiting the bacterial type II topoisomerases, DNA gyrase and topoisomerase IV [1]. Quinolones attach to the DNA-topoisomerase complex, which becomes irreversible, leading to immobilization of the enzymes resulting in bacteriostasis and to the release of DNA double-strand breaks that activate the SOS system and produce the ''poison'' effect responsible for the intense bactericidal action of quinolones [2]. Fluoroquinolones (the main subgroup of quinolones) are currently among the most heavily prescribed antimicrobials in the world because of their pharmacodynamic and pharmacokinetic properties [1]. They are very potent, especially for treatment of urinary tract infections due to Enterobacteriaceae [1,2], and as a consequence of their intense use, quinolone resistance rate has increased much for the last years [1,3,4]. Quinolone resistance mechanisms are multiple such as those reducing permeability of the bacterial wall, increasing efflux, reducing target affinity, producing inactivating enzymes and target protection proteins. Clinical resistance mostly results from the combination of several mechanisms [1]. Most of these mechanisms are chromosome-mediated [1,2] but plasmid-mediated genes have been described for a decade [5]. qnr genes were the first plasmidmediated quinolone resistance genes described in 1998 [6].
Qnr proteins are pentapeptide repeat proteins that protect DNA gyrase and topoisomerase IV from quinolone binding. Six families of plasmid-mediated qnr gene were described: qnrA, qnrB, qnrC, qnrD and qnrS [7][8][9][10]. Several alleles have been reported for each gene: 7 alleles of qnrA, 73 of qnrB, 1 of qnrC, 2 of qnrD, 9 of qnrS and 6 qnrVC. Alleles are numbered consecutively in a qnr library (http://lahey.org/qnrStudies, last update December 08, 2013). Most of qnr genes were detected in Enterobacteriaceae, where they located on large conjugative multi-resistance plasmids, such as pMG252 described in the original Klebsiella pneumoniae qnrA1-positive isolate [6]. The qnrB gene is the most prevalent in clinical bacterial strains, whereas qnrS gene is prevalent in environmental strains. Since qnrA alleles were found also as chromosome-borne in Shewanella spp. [11], this species was proposed to be a qnrA progenitor. Similarly, a qnrB allele was found in the chromosome of Citrobacter spp. [12], a qnrS and qnrVC allele in Vibrionaceae [10,13]. qnr genetic environment usually showed mobile elements that can explain their spread from their progenitor, although they confer only a low level resistance to quinolones. Most of qnrA and qnrB genes were reported on large conjugative multi-drug resistant plasmids including aminoglycoside and beta-lactam resistance genes such as those encoding extended spectrum beta-lactamases, and in the vicinity of complex type 1 integrons [14,15]. On the contrary, qnrS and qnrD genes were described on small nonconjugative plasmids, that may harbor mob genes allowing their mobilization [16,17].
qnrD was first reported on a 4,270-bp (p2007057) plasmid in four Salmonella enterica isolates [7]. We subsequently described two Providencia rettgeri clinical isolates carrying qnrD on a 2,683-bp plasmid [18]. We hypothesized that qnrD can be originated from Proteeae or present in these bacteria as a reservoir. Our study determined how much qnrD is prevalent in recent clinical isolates of Proteeae, and investigated how qnrD could have disseminated to other bacterial species. We found seven nearly identical small qnrD-bearing plasmids, where qnrD is surrounded by sequences of mobile insertion cassette (mic) that may explain qnrD dissemination. Since the half part of the plasmid sequence excluding qnrD, was found in a native plasmid of Providencia vermicola, we think that qnrD might be present on the chromosome of a close, as yet unknown, Proteeae species and that its DNA recombination with P. vermicola native plasmid led to the small non-transmissible qnrpositive plasmid found in clinical isolates.
Detection of qnr genes was performed by multiplex real time PCR as previously described [19]. Briefly, after DNA extraction using the NucliSens easyMAG (bioMérieux, Marcy l9Etoile, France), multiplex real-time PCR was carried out using Light-CyclerH 480 (LC480) with HRM Supermix (Roche Molecular Diagnostics, Germany). The melt-curve analysis showed characteristic curves for each qnr gene family. All strains detected were confirmed to be qnr-positives by conventional PCR-sequencing.
The qnrD-positive isolates were compared for genome fingerprinting by Random Amplification of Polymorphic DNA (RAPD) using the Ready to Go RAPD analysis kit according to the manufacturer's instructions (Amersham Biosciences).

Characterization of qnrD-bearing Plasmids
Plasmid extraction was performed by the Kieser method from the qnrD-positive strains [20]. Plasmids harboring qnrD were sequenced on both strands using a Walk DS strategy [18].
Nucleotide sequences were aligned using the program BioEdit version 7.09.0 and BLAST searches using the National Center for Biotechnology Information website (http://www.ncbi.nlm.nih. gov) and the blastn algorithms. Putative promoters, ribosome binding site, and transcription start site were searched using BPROM (http://linux1.softberry.com) and Promoter Prediction by Neural Network [21].

Search for qnrD and Small Non-transmissible Plasmids in Reference Strains of Proteeae
Plasmid extraction was performed by the Kieser method from the 15 reference strains and from 50 qnrD-negative strains randomly selected [20]. Based on the nucleotide sequences of the qnrD-plasmids characterized in the clinical isolates, six pairs of primers were designed corresponding to overlapping fragments (Table S1). Amplification products were visualized on 2% agar gel after electrophoresis and sequenced.

Investigation of qnrD Transfer
Conjugation experiments were carried out using liquid and filter mating methods. Liquid mating experiments were performed in brain-heart infusion (BHI) broth with E. coli J53 azide-resistant [15], P. mirabilis ATCC 29906 Rif R [23] and E. coli C600 Rif R [23] as recipient strains. Donor and recipient cells in the logarithmic phase of growth were mixed (1 ml each) and incubated at 37uC for 3 hours without shaking. For filter conjugation, cultures (logarithmic phase of growth) of the donor and the different recipient cells were collected on a MF-type filter (Millipore, Molsheim, France). The hydrophobic edge membrane was incubated at 37uC on the surface of a BHI agar plate overnight. Transconjugants were selected on BHI agar plates containing sodium azide (100 mg/ml) or rifampin (250 mg/ml), with regard to the recipient strain, and nalidixic acid (50 mg/ml) or ciprofloxacin (0.015 mg/ml, 0.03 mg/ ml, 0.06 mg/ml and 0.12 mg/ml) or ofloxacin (0.03 mg/ml, 0.06 mg/ml, 0.12 mg/ml, and 0.25 mg/ml).
The main capture systems (intl1, ISCR1, IS26, IS10, and ISEcp1) that have been associated with plasmid-mediated quinolone resistance determinants were detected using specific PCR (Table  S1). In silico analysis was conducted for searching insertion sequence (IS) using IS Finder (http://www-is.biotoul.fr/is.html) and inverted repeats using REPFIND [24].

Quinolone Resistance Testing
Since the qnrD-positive clinical isolates may have additional mechanisms conferring quinolone resistance, including chromosomal mutations, we determined the quinolone resistance of both the original Proteeae strains and the transformants obtained after transfer of qnrD-bearing plasmids into competent E. coli DH10B cells (Invitrogen, Cergy-Pontoise, France). Plasmid DNA was extracted from the qnrD-positive strains using the QIamp DNA mini Kit (Qiagen, Courtaboeuf, France) and introduced by electroporation (Gene Pulser, Biorad, Marnes la Coquette, France). Transformants were selected on BHI agar plates containing ciprofloxacin (0.015 mg/ml, 0.03 mg/ml, 0.06 mg/ml and 0.12 mg/ml) or ofloxacin (0.03 mg/ml, 0.06 mg/ml, 0.12 mg/ ml, and 0.25 mg/ml). In order to distinguish transformants and quinolone resistant mutants growing on quinolone containing media, we sequenced the QRDRs in gyrA and parC, in addition to qnrD detection by PCR.
Minimal inhibitory concentrations (MICs) of nalidixic acid, ciprofloxacin, levofloxacin, moxifloxacin, norfloxacin, and ofloxacin were determined by the agar dilution method using Mueller-Hinton agar plates containing serially twofold-diluted antibiotics. Plates were inoculated with a Steers-type multiprong device with ca. 10 4 CFU per spot, and were read after incubation for 18 hours at 37uC.
Conjugation assays, performed extensively, failed in transferring qnrD from clinical isolates to any of the recipient strains. From the seven qnrD-positive isolates, DNA extraction by the Kieser method showed several plasmids but a similar small plasmid of ca. 2.6 Kb in six isolates and one ca. 4 Kb in the remaining isolate. qnrDplasmids were negative for known incompatibility groups.
Quinolone MICs are presented in Table 3. Quinolone resistance conferred by the qnrD-bearing plasmids was similar for all plasmids, with a 60-fold increase in ciprofloxacin MIC. For the plasmid pRS12-78, transformation experiments on ofloxacin or ciprofloxacin, even at low concentration, selected S83L gyrA chromosomal mutants harboring pRS12-78. Consequently we cannot measure the exact level of resistance conferred by the plasmid. However, since the ciprofloxacin MIC was similar to that on a gyrA mutant [25], we concluded that the plasmid did not confer quinolone resistance, which is explained by the insertion between the putative promoter and the qnrD start codon (see below). No mutation in the QRDR of gyrA and parC were found in the other transformants.
Plasmid pRS12-78 (GenBank accession number KF364954) was identical to the 2.6-Kb plasmids but with an additional 1,568 bp insert between qnrD and its promoter ( Figure 1A). BLAST comparison highlighted a putative ORF5 exhibiting 76% identity with a hypothetical protein (GenBank accession number ADR29581) close to the ATPase domain of the E. coli ParA protein. Moreover, at 147 nucleotides downstream the stop codon of ORF5, the presence of inverted repeats followed by a short string of A's suggested a factor-independent termination site. Interestingly, among the randomly selected qnrD-negative strains for plasmid characterization using Kieser extraction, none of them harbored small plasmids (data not shown).

qnrD is Included in a Mobile Insertion Cassette
In silico analysis did not show any gene involved in mobilization and specific PCRs for intl1, ISCR1, IS26 and ISEcp1 were negative. We also looked for mobile insertion cassette (mic) elements that are similar to transposons. A mic element is bracketed by two inverted repeats without any transposaseencoding gene found nearby and duplication of the target site is usually observed indicating the transposition. In the 2,683-bp long plasmids, qnrD was localized within a 1,519-bp ''mic'' element, bracketed by two 24-bp imperfect inverted repeats (IR) with six mismatches ( Figure 1B). No transposase-encoding gene was found. Acquisition of the mic-qnrD element was likely the result of a transposition process because it is bracketed by a 2-bp duplication of the target site (CA). The mic-qnrD element described here carries the necessary elements for expression of the qnrD gene: (i) promoter sequences including a -35 box (TTTTATA) and a -10 box (TCTTAAAAT) separated by 11-bps, (ii) and a Shine-Dalgarno sequence (AAAAGGG) upstream on the qnrD initiation codon (ATG).
In the 4,686-bp long plasmid pRS12-78, qnrD is a part of a 3,072-bp ''mic'' element given that the ORF5 described above was inserted between qnrD and a putative promoter ( Figure 1A). Although this second mic-qnrD element is slightly different in length, it has the same characteristics.
Reference Strains of Proteeae Carried Native Plasmids but no qnrD Gene qnrD was not detected by multiplex or simplex real-time PCR in the fifteen reference strains of Proteeae but plasmids were found in most of the reference strains (Table 1). These plasmids did not belong to any known incompatibility group detected using the PBRT method [22].
Overlapping PCRs designed on the sequence of pDIJ09-518a and performed using template DNA of the Proteeae reference strains showed that one small plasmid of P. vermicola, pVERM, carried a fragment exhibiting 89% identity with the sequence of pDIJ09-518a. pVERM was linearized using AclI restriction enzyme. After gel extraction, the sequence was determined using a walk DS strategy on the amplification product obtained with primers pDIJ-2558F and orf2-39R. pVERM is a 3,682-bp long  plasmid containing a 1,794-bp region sharing more than 98% identity with the qnrD-plasmids. In this region, we identified three ORFs similar to ORF2, ORF3 and ORF4 described in pDIJ09-518a. BLAST of the 1,888-bp remaining region showed a 267-bp ORF1 (named pVERM_ORF1 in the following) covering 76% of a putative protein described in Vibrio brasiliensis (accession number ZP_080099857.1) with 31% similarities. Two 24-bp long imperfect IR, with eight mismatches, and CA duplications were found to bracket pVERM_ORF1 and pVERM_ORF2 leading to identify another mic element. The 2,664-bp long mic element harbored by pVERM is different from mic-qnrD for 1,145 bps containing pVERM_ORF1 and lacking qnrD ( Figure 1B). Nonetheless, IRR and IRL are similar to those bracketing mic-qnrD on the qnrD plasmids such as pDIJ09-518a with 2 and 1 single nucleotide differences for IRLs and IRRs respectively ( Figure 1B). No known mobilization structure such as IS were identified in this 2,664-bp long mic element.

Discussion
We investigated the emergence of the new plasmid-borne quinolone resistance gene, qnrD in Proteeae clinical bacterial isolates. The qnr genes are encoding a group of pentapeptide repeat proteins recently discovered but certainly present on bacterial chromosomes long before quinolones were used clinically. They emerged 10 years ago, following the 1990-2010 worldwide increase in fluoroquinolone usage [3]. qnrD was initially described in Salmonella enterica Kentucky and Bovismorbificans isolates [7], our investigation showed that qnrD was the qnr gene mainly present in isolates belonging to the Proteeae tribe (Proteus, Providencia and Morganella genera) since the screening of all qnr genes in 317 Proteeae clinical isolates resulted in 8 qnr-positive strains but seven (87.5%) positive for qnrD. This unexpected result converges with data in GenBank where among 38 sequences of qnrD half were reported in Proteeae (Table S2, last update October 30, 2013). We consequently hypothesized that Proteeae are the reservoir of qnrD genes. For the termination site of ORF5, two black arrows indicate inverted repeats, and the string of ''A's'' is boxed. The inverted repeats (IRL and IRR) bracketing the mic element are boxed as arrows shaded in grey. In pRS12-78, qnrD is far from its own promoter due to the insertion of ORF5. B. Comparison of the mic elements found in the plasmid, pDIJ09-518, pRS12-78, pVERM and in p2007057. Open reading frames (ORFs) are indicated by horizontal arrows. White arrows are identical ORFs harbored by the different plasmids beyond the mic. Red arrows are identical ORFs harbored by the different plasmids within the mic. The right and left inverted repeats (IRR and IRL) are indicated as black triangles with the duplication sites (CA). Nucleotide sequences of the IRL and IRR are outlined and the nucleotides which differed between them are in bold type and underlined. Asterisks show nucleotide differences of IRs with regard to those in pDIJ09-518a. doi:10.1371/journal.pone.0087801.g001 In Proteeae clinical isolates, we found qnrD on small ca. 2.6-kb non-transmissible plasmids, in contrast with what was mainly described for other qnr genes carried by large conjugative plasmids [5]. Although qnrD was also reported to on such large conjugative plasmids, none of these plasmids was well characterized [26][27][28][29]. All the 15 qnrD-bearing plasmids found in Proteeae (the 7 from our study and 8 plasmids from GenBank sequences, Table S2) have the same size (2,683-bp) and have nearly identical sequences, which is in favor of a common origin or ancestor [30][31][32]. The only exception found during our study is the plasmid pRS12-78 that presented a plasmid scaffold identical to the other characterized plasmids but with a 1,568-bp insertion corresponding to a putative additional ORF. By contrast, the plasmids bearing qnrD found in strains of Enterobacteriaceae other than Proteeae are 4,270 bp long with sequences nearly identical to p2007057 described in S. enterica. These latter plasmids usually harbor a mob gene that may explain their transfer. The two types of plasmids carrying qnrD show a highly conserved backbone, but the smallest are not transmissible. This was previously described for qnrS, as it was described on a 10,066-bp non-conjugative but mobilizable plasmid in S. enterica Typhimurium [16]. It was suggested that this plasmid results from the recombination of a qnrS1 conjugative plasmid from S. enterica Infantis (pINF5) and a small mobilizable plasmid from E. coli (pEC278). The qnrD-2,683-bp plasmids were stable in E. coli DH10B after electroporation, indicating a host range including other species of Enterobacteriaceae. However, the gene encoding the replicase is probably different to what is usually found since the incompatibility group cannot be assessed. Non classic replication systems were recently reported for some small plasmids in Proteeae [18,33] and we share the view of Galac et al. that cryptic plasmids in Proteeae, including the small qnrD-plasmids, may use a not yet characterized replication method [33]. Since reference type strains of Proteeae did not harbor qnrD, we sought for similar plasmid backbone in their constitutive DNA. We found, in a reference strain of P. vermicola, a native plasmid that shares 1,794 bps with more than 98% identity with our qnrD-bearing plasmids.
We described here qnrD as located in a mobile insertion cassette (mic) element. Mic elements were first described in Bacillus cereus (MIC231A1) as a DNA cassette able to be mobilized in trans by the IS231A transposase [34]. MIC231A1 contains an active Dstereospecific endopeptidase (adp) gene instead of a transposase. Further studies conducted on B. cereus showed the presence of antibiotic resistance determinant such as a fosfomycin resistance gene in MIC231D, described as a novel active composite cassette [35]. It was then proposed mic element could be also a vehicle for antibiotic resistance genes beside class 1 integrons where multiple antibiotic cassettes can be inserted. With regard to plasmid mediated quinolone resistance genes, qnrS2 was described in a mic element in a strain of Aeromonas punctata isolated in water sampled from the River Seine in Paris [36]. Unfortunately, as reported by Cattoir et al., no transposases were found, on the basis of the IR nucleotide sequences, suitable for demonstrating trans-transposition in an in vitro assay [36]. qnrA genes were mainly associated with ISCR1 at the vicinity of complex class 1 integrons which include various other resistance determinants (e.g. bla genes or aac(69)-Ib-cr). qnrB genes were also described beside class 1 integrons, but associated with ISEcp1 and ORF1005, another type of inserted sequences [37,38]. Unlike qnrA and qnrB, qnrS genes were mainly associated with structure such as ISEcl2, IS26 and mic structure [16,36,39], and were not reported as part of complex integrons. These gene-capture systems explain the mobilization of qnr genes from their progenitors to self-transmissible plasmids leading to their widespread in Enterobacteriaceae.
We think that qnrD was imported from a Proteeae species and inserted in pVERM through mic mobilization (Figure 2). Further recombination, insertion or deletion events may have occurred after the mic-qnrD mobilization including qnrD deletion, resulting in the plasmids pVERM and pDIJ09-518a reported here. Indeed, the IRs found in the mic-qnrD plasmids from Proteeae clinical strains are more similar to those found in pVERM than those found in plasmids in S. enterica. pVERM differed from qnrD-plasmids by 1,888 bps located inside the very same lacking qnrD mic. This qnrD-lacking-mic element found in pVERM is 1,145 bp longer than mic-qnrD and carried a pVERM_ORF1 present neither on pDIJ09-518 nor on p2007057. While mic-qnrD IRR and IRL showed seven unmatched nucleotides for pVERM vs six for mic-qnrD, more importantly a comparison of IRLs between mic-qnrD and pVERM found only 2 SNPs and only 1 SNPs was found comparing both IRRs. Finally, IRs in both qnrD-plasmid and pVERM were 24-bp long. ( Figure 1A). Moreover, none of the 50 qnr-negative clinical strains, randomly selected, harbored such small plasmids. Therefore, our current hypothesis is that pVERM could be the ancestor plasmid where qnrD was later deleted. Alternatively, pVERM could be a qnrD-plasmid resulting from another mic-qnrD harbored by an unknown qnrD progenitor, from which qnrD was subsequently lost.
The qnrD progenitor is not yet identified. For the other qnr genes, progenitors are mainly bacteria living in the environment (Shewanella spp., Citrobacter spp., and Vibrio spp) where Proteeae are also present [13,36,40,41]. The Proteeae could have acquired qnrD from its progenitor by recombination between an as-yet-unknown bacterial source of qnrD and the cryptic plasmid pVERM harbored by P. vermicola leading to small non-mobilizable 2,683-bp long plasmids carrying qnrD embedded in a mic-qnrD element ( Figure 2). Given that all plasmids of about 4,270 bps such as p2007057 (e.g. P. mirabilis isolates CGP180 and CGH40 described by Zhang et al.) also harbored a mic-qnrD with a conserved 24-bp IRL and a modified 26-bp IRR showing 1 and 7 nucleotide differences, respectively, with regard to 2,683-bp qnrD-plasmids, we think that the mic-qnrD element could have been later inserted by trans transposition in a mobilizable plasmid such as the 4,270-bp qnrDplasmid p2007057. It is also possible that the mic-qnrD was transferred from a common progenitor to both Proteeae and Salmonella spp. in parallel onto two different types of plasmids in which the IRs sequences evolved and diverged especially for the IRR.
Finally, we investigate the quinolone resistance conferred by the qnrD-bearing plasmids. The MICs of fluoroquinolones were similar and comparable to MICs previously reported for p2007057, except for the plasmid pRS12-78. In this plasmid, the insertion of a 1,568-bp long sequence, can hamper qnrD expression from a putative promoter then located far from qnrD. Although the level of resistance is rather low for classical quinolones such as nalidixic acid, it can reach 60-fold increase in the MIC of ciprofloxacin. Such an increase can cause failure in experimental model of pyelonephritis and in pulmonary infections [42][43][44]. Further studies are needed to investigate the fitness cost of qnrD harbored by these small plasmids, but as described for qnrA3 [45], qnrD could contribute to a fitness gain leading to the emergence of qnrD-bearing plasmids even in the absence of quinolones.