Multiple protein sequence alignments were carried out with the program ClustalW2 http://www.ebi.ac.uk/Tools/clustalw2/index.html56. Similarity search was performed with the program BLAST (NCBI) http://www.ncbi.nlm.nih.gov/blast/Blast.cgi57 and genomicBlast http://www.ncbi.nlm.nih.gov/sutils/genom_table.cgi. Identification of conserved motifs 58 (59 was done by using Conserved Domain Database and Search Service at http://www.ncbi.nlm.nih.gov/Structure/cdd/cdd.shtml. The Qnr proteins alignments and homology trees were obtained using CLUSTALW2 and Jalview alignment editor with default parameters 60. Homology tree was calculated using the method "Average distance using percentage of identity" http://www.ebi.ac.uk/Tools/clustalw2/index.html.

The predicted Qnr proteins were named following the rules proposed in 61.

The bacterial strains and plasmids used are shown in Table 1. The strains were grown in Luria-Bertani (LB) broth 62 at 37°C unless otherwise specified.

The susceptibility assays for the different antibiotics were performed in Mueller Hinton broth (Pronadisa) plus Isopropyl-thio-ß-D-galactopyranoside 0.5 mM, using the twofold dilution method in 96-well microtiter plates. The results were recorded after 48 h of incubation at 37°C. To ensure that the observed changes were consistent, all Minimal Inhibitory Concentrations (MICs) were determined in three independent assays, using different bacterial cultures on different days. In all cases, a control strain containing the plasmid pGEM-T without any insert was included in MICs determinations. In most cases, there were not inter-assay differences in the MIC values. In a few cases, there were one-dilution differences. For the latter, the assay was repeated one more time to further assure assay reliability.

The quinolones used were ciprofloxacin, enoxacin, garenoxacin, grepafloxacin, levofloxacin, moxifloxacin, nalidixic acid, norfloxacin, trovafloxacin and sparfloxacin.

The genomic DNA was extracted using the GNOME^® DNA Kit (Q-BIOgene). The PCR Master Mix (Promega) was used to amplify full-length Smqnr genes from the different S. maltophilia strains without their promoter sequences. The reaction contained 100 ng of genomic DNA of each S. maltophilia isolate as template, and 1 μM of two different sets of specific Smqnr-primers, QnrM+ (5'-CTTGGCATGGAATCCCTGAT-3')/QnrM-(5'-TGATGCCTACGGCACCAC-3') and QnrMR55+ (5'-CATGGCATGGAATCCCCGAT-3')/QnrMR55-(5'-TGATGTCTACGGCACCAC-3'). We used two sets of primers because the regions around qnr are slightly different in the sequenced S. maltophilia strains K279a and R551-3. The reactions had one denaturation step at 94°C for 5 minutes, followed by 35 amplification cycles: 94°C for 30 seconds, 61°C for 45 seconds for annealing, and 72°C for 1 minute for elongation, with a final extension step of 72°C for 5 minutes. The PCR products (660 bp) obtained from the different S. maltophilia strains (Table 1) were electrophoresed in 1% agarose gels with TAE, purified from the gel with GFX™ PCR DNA and Gel Band Purification Kit (GE Healthcare) and cloned in pGEM-T plasmid (Promega) generating the corresponding recombinant plasmids (Table 1). The transformation of E. coli KZM120 was made as described 63. We chose this particular E. coli strain because it lacks the major E. coli MDR pump and it has been shown previously that the utilization of MDR-defective strains facilitates the characterization of low-level mechanisms of resistance 39. The transformants were selected in LB agar with 100 μg/ml carbenicillin. The recombinant plasmids were purified using Wizard^® Plus SV Miniprep Kit (Promega) and the insert of each plasmid was sequenced using the universal primers M13 forward and M13 reverse by Secugen S.L. http://www.secugen.es/ to confirm the identity of the sequence and establish the orientation of the qnr gene. The sequences of the different Smqnr genes have been deposited at GenBank with numbers from EU681371 to EU681385. Only those plasmids containing the Smqnr gene in the right orientation to allow expression from the pGEM-T lac promoter were used in the functional assays.

The potential presence of Smqnr in both clinical and environmental S. maltophilia strains was evaluated by PCR using the primers qnrI1 (5'-AGAAAGTGGTCGACCAGCAG-3')/qnrI2 (5'-GCAGGTTCGACTTCTTGATG-3') and qnrI3 (5'-CAACGCCAGCTTCATGAACC-3')/qnrI4 (5'-AGTTGGCGCTGTTCCAGTCG-3'), which amplify 312 bp and 220 bp fragments of two internal regions in the Smqnr gene of S. maltophilia respectively.

The PCR was made as described above, with the following program, one denaturation cycle at 94°C for 5 minutes, followed by 35 amplification cycles: 94°C for 30 seconds, 55°C for 45 seconds for annealing, and 72°C for 30 seconds for polymerization, with a final extension of 72°C for 5 minutes. The PCR products were analyzed in 1% agarose gels with TAE.

The amount of SmQnr protein was estimated by SDS-PAGE 63 and Coomassie blue staining, using cell extracts from bacteria grown in LB at 37°C overnight. The amount of proteins loaded in each lane was normalized to around 4 × 10⁷ cells. In all cases, the global amount of proteins was equal in all lanes of a given gel as can be seen after staining. This serves as a supplementary loading control 64.

To ensure that the protein with the predicted molecular size was indeed SmQnr, the corresponding band was excised and identified, as described 6539 by Peptide Mass Fingerprinting using MS-MALDI TOF (Proteored: http://www.proteored.org/).

This investigation did not require ethical clearance.

To address the presence of putative qnr genes in the chromosomes of bacteria that could serve as reservoirs of this family of quinolone resistance determinants, a bioinformatic iterative search by sequence homology to several alleles of QnrA, QnrB and QnrS was conducted against genomic or Whole Genome Shotgun Sequence (WGS) databases at the NCBI home page http://www.ncbi.nlm.nih.gov/. Most of the putative qnr genes found in this search have been previously annotated as hypothetical proteins containing the COG1357 motif. As shown in Figure 1, chromosomally-encoded Qnr proteins present a large degree of homology.

Putative qnr genes were found in 22 out of the 960 genomes tested. The species containing putative qnr genes belong to eight different genera (Figure 2). Most of them inhabit an aquatic environment (in some instances the deep sea) suggesting that plasmidic qnr may have been originated from aquatic bacteria 171918.

Since the large majority of bacterial species have not been cultured so far, metagenomic studies allow a more powerful search of relevant genes in different environments by means of non-culture based methodologies. Thus, besides analysing sequenced genomes, available metagenomic sequence databases http://www.ebi.ac.uk/fasta33/wgs.html were also searched for the presence of qnr genes. Our study showed that the metagenomes from sea-water 47 contain genes homologous to qnr (hereafter named as Mtgqnr), further supporting the notion that qnr genes are originated from aquatic microorganisms. One of the qnr sequences found in this metagenomic search was highly homologous to plasmid-encoded QnrB 15 proteins (Figure 3). It has been shown that S. algae is likely the origin of plasmid-encoded qnrA genes 17. Our analyses, strongly suggest that plasmid-encoded qnrB has also originated from a currently unknown, marine microorganism.

The regions surrounding the chromosomal qnr genes found in our bioinformatics analysis were analysed to ascertain whether, based in the level of sinteny, the origin of these qnr determinants is likely monophyletic or polyphyletic. As shown in Figure 2, and in Additional file 1, the regions are very similar for some of the analysed species belonging to the same genus, although clear differences are also observed for other species.

For instance, the structure surrounding qnr in Vibrio is very similar for three species, whereas the structure in five other species and in Photobacterium profundum is different. The high level of observed sinteny in these regions, suggests that qnr was acquired by Vibrio parahaemolyticus, Vibrio alginolyticus and Vibrio harveyi before their divergence, and was acquired by the other Vibrionaceae after their divergence. This idea fits well with the relationship of Vibrio Qnr proteins (Figure 4).

Sequence data were inspected for the presence of elements potentially involved in the transfer of qnr genes. As shown in Figure 2, putative transposases or integrases were found in the regions surrounding the qnr determinants in some Shewanella species, and one putative recombinase was found as well in Moritella sp. Whether those elements are remnant of former gene transfer events or constitute a risk for the transference of those qnr genes through HGT to a new host, remains to be established.

In any case, the analyses of available sequences of bacterial genomes suggest that qnr genes are ancient elements in the bacterial chromosomes of a specific subset of bacterial species. Noteworthy, those genes are frequently flanked by genes coding for putative efflux pumps (Figure 2), a situation that suggests that Qnr may have functional roles in detoxification processes in water-dwelling bacteria. Although this linkage has not been so far observed for plasmid-encoded qnr genes, the cooperation of two different mechanisms of resistance might be an important element in current resistance to quinolones.

Once the bioinformatic analysis was made, we focused our efforts in the study of the organism with highest clinical relevance. Among the organisms carrying putative uncharacterized qnr genes we chose S. maltophilia, which is an intrinsically resistant opportunistic pathogen responsible for 4.3% of Gram-negative infections among intensive care units patients in the USA 48.

Noteworthy, in S. maltophilia resistance to fluoroquinolones (particularly to less active ones, such as norfloxacin) is not always accompanied by high level resistance to nalidixic acid, which is the usual and classical association in Enterobacteria. Interestingly, resistance to fluoroquinolones and susceptibility to nalidixic acid in S. maltophilia resemble the phenotype recently observed by Cano ME et al (unpublished results) in some Enterobacteriaceae containing qnr genes. A putative qnr gene (Smqnr) was identified in the two available full-genome sequences of S. maltophilia: the clinical isolate K279a http://www.sanger.ac.uk/Projects/S_maltophilia/49, and the environmental strain R551-3, http://genome.jgi-psf.org/draft_microbes/stema/stema.home.html. As shown in Figure 2, the genomic structure around Smqnr was the same in both strains, in spite their different (clinical and environmental) source of isolation.

This high degree of sinteny, together with the lack of elements involved in HGT near Smqnr suggests that this gene has an ancient origin in S. maltophilia. Further confirmation of this likely monophyletic origin has been obtained from the analyses of the Smqnr sequence from different S. maltophilia isolates in this study. The predicted translation of Smqnr sequence is a 219 amino acid protein (Figure 5), containing tandem repetitions of the pattern (A/C/S/T/V)(D/N)(L/F)(S/T/R)(G/R) typical of the pentapeptide family 201822, with two domains of pentapeptide repeats separated by a single glycine, that corresponds to a COG1357 motif.

Since Smqnr might be involved in quinolone resistance, we wanted to ascertain the prevalence of this gene in S. maltophilia populations. To that goal, and since the sequences of the qnr genes of the strains K279a and R551-3 were slightly different, two pairs of oligonucleotides QnrM+/- and QnrMR55+/- were designed to amplify Smqnr genes from different S. maltophilia strains. The set of analysed strains comprised both clinical and environmental S. maltophilia isolates (Table 1).

The SmQnr sequences from the different strains display a high amino acid identity, ranking from 94 to 100% (Figure 5). However the amino acid identity of the SmQnr protein encoded in the chromosome of the sequenced S. maltophilia strain K279a, is low compared with other known Qnr proteins, sharing 38%, 59%, 38% and 41% with QnrA1 (AY070235), QnrB2 (DQ351241), QnrS2 (AB187515) and VvQnr (DQ889870) respectively. For this reason we consider that Smqnr represents a possible new subtype of qnr genes. Clustal analysis (Figure 4) indicates that the SmQnr protein clusters with plasmidic QnrB proteins, although forms a separate branch. Therefore, our results suggest that SmQnr is the first characterized chromosomally encoded QnrB-like protein.

We were unable to amplify the full qnr gene from the isolates E729, F227, E301, F861 and e-p5. Since qnr sequences where slightly different in K279a and R551-3, as well as in the isolates analysed in the present study, two new pairs of oligonucleotides (qnrI1/qnrI2 and qnrI3/qnrI4) were used to amplify an internal conserved region of Smqnr. Positive amplification was obtained in all cases (not shown) indicating that all S. maltophilia isolates contained the qnr gene.

To ascertain whether SmQnr might contribute to quinolone resistance in a heterologous host, the different alleles of the qnr gene obtained from different S. maltophilia strains were cloned into the plasmid pGEM-T. Corresponding recombinant plasmids, containing the Smqnr gene, were expressed in E. coli, and the quinolone susceptibility of strains expressing SmQnr was compared to that of the isogenic E. coli strain. We found that in trans expression of SmQnr from different recombinant plasmid results in a 2 to 32-fols decrease in the quinolone susceptibility of E. coli (Table 2).

Noteworthy, different recombinant strains presented slight, but consistent, differences in their susceptibility to several quinolones (Table 2). It has been previously stated that transconjugants of plasmids containing the qnr gene present disparate levels of quinolone resistance, likely due to different levels of expression of the Qnr protein in the different transconjugants 50. To ascertain whether this could be the situation with the different SmQnr-expressing plasmids, the expression of the SmQnr protein was estimated in the different E. coli transformants. As shown in Figure 6A, transformants containing either pBS3.25 or pBS3.13 expressed higher levels of SmQnr than the other transformants.

During our work, we detected that some Qnr-expressing plasmids were lost after subculturing even in the presence of ampicillin, although they could be maintained in the presence of carbenicillin. Plasmid loss after the bacteria has degraded the selective antibiotic is an indication of physiological burden 51. Our results suggest that high-level expression of SmQnr is likely harmful for E. coli. It is possible that, as described for other systems 52, a fast adaptation between the host cell and the plasmid occurred that results in reduced levels of SmQnr and that this allowed some of the clones to overcome the putative SmQnr toxicity. To address this possibility, E. coli KZM120 was retransformed with the plasmids pBS3.8, pBS3.25 and pBS3.13. Two different clones from each retransformation were chosen and their susceptibility to quinolones tested. As shown in Table 3, an overall reduction in MICs as well as in the level of SmQnr expression (Figure 6B) was observed for the new transformants containing the plasmids pBS3.8, pBS3.25 and pBS3.13. However, the phenotype of reduced susceptibility to quinolones was maintained. Increased resistance due to higher expression of antibiotic resistance genes has been described mainly associated with plasmidic beta-lactamases 535455. Our results together with already published data indicate that gene-dosage might be relevant for qnr-mediated quinolone resistance.