Sequencing of gyrA indicated that this locus was informative in several different ways for characterizing Campylobacter isolates. First, the alleles of the 496 nucleotide fragments were suitably different in sequence identity between C. jejuni and C. coli to be assigned to one or the other of these species. The distribution of these alleles confirmed that recombination events between species occur rather infrequently and in an asymmetric gene flow : one C. jejuni had a typical C. coli allele whereas 4 C. coli had a typical C. jejuni allele. Two other studies using PCR and sequencing data targeting gyrA also identified a C. jejuni segment within a C. coli isolate [34,35], supporting previous findings that gene flow is rather unidirectional from C. jejuni to C. coli [33,36].
Sequencing of gyrA revealed a similar population structure as that obtained by MLST or rMLST (Ribosomal Multilocus Sequence Typing, ). In particular, the phylogenetic analysis clearly organized C. coli into 3 distinct clades as previously described by Sheppard et al. [33,36] (Figure 1). Furthermore, peptide groups 301A and 302 in our study (Table 2) contain alleles commonly found in domestic animals, and they correspond to the agricultural C. coli lineage of the evolutionary scenario proposed by Sheppard et al. . In addition, peptide groups 301B and 301C (Table 2) match with the clades 2 and 3 observed by Sheppard et al.  including only alleles recovered from environmental isolates, i.e. from surface waters in our study. In contrast to C. jejuni, the C. coli assigned alleles are predominated by synonymous mutations. As a result, the peptide group 301C is characterized by alleles with a higher GC content (Figure 2A) generated by nucleotide changes only located in the third positions of codons. This trend was also reflected in genotypes linked to this peptide group 301C i.e. by compiling GC content from the internal fragments of the 7 MLST housekeeping genes with the gyrA alleles (a total of 3,805 bp in length, see Additional file 1). This kind of GC rich version of genes, independent of adaptive codon usage was significantly associated with effects on bacterial fitness, which could be explained by higher stability of mRNAs . The study of Foerstner et al.  linked the genomic GC pattern of bacterial populations to environmental factors like ultraviolet irradiation as an example. Thus, the difference in synonymous GC contents found in the gyrA alleles from the peptide groups 301B and 301C, suggests that these lineages originated from two distinct but not yet identified ecological niches. By using concatenated nucleotide sequences from MLST data, isolates from our gyrA peptide group 301B would be classified in the clade 2 from the study of Colles et al.  (see Additional file 2) including the majority of the STs identified from wild Mallard ducks. Among our collection of surface water isolates, we similarly observed three clades: one associated with domestic animals and the other two of wildlife origin, one of which potentially linked to waterfowl. Nevertheless, with a more discriminative approach based on genotypes defined by combining the 7 housekeeping genes from MLST with the gyrA, the populations of C. coli displayed a high specificity in their distribution by sources (Figure 3). None of the 194 genotypes identified was found in all three collections (SW, DM and P) and F
STs values calculated by pair comparisons were about 4 times higher than those computed from C. jejuni pairs. The fact that domesticated mammal isolates were poorly represented in our environmental samples could have resulted from a temporal and geographic sampling bias. Half of the collection was mainly isolated in 2006  and the other half was collected from distant geographic locations. As to the isolates originating from poultry, it must be emphasized here that domestic production of broilers is negligible and there is no poultry hatchery in the country. Thus, direct contamination of environmental waters by local poultry farms is largely restricted.
Regarding the C. jejuni gyrA sequences, two lineages were clearly distinguished (Figure 1). One branch is represented by the peptide group #14, encoded by the alleles #54 and #55 recovered from surface waters isolates only. These nucleotide sequences are again mainly differentiated by their GC content, but this time, below the mean of each of the other groups (Figure 2). The two STs associated with these strains are newly described (ST 5841 and ST 6171) and correspond to variants of a C. jejuni clone associated with bank voles . Interestingly, these strains also displayed atypical profiles with the duplex-real time PCR implemented in this study for identifying isolates at the species level. An extra PCR was needed to confirm the presence of the hipO gene (see the Methods section). In summary, this phylogenetic lineage originated from a wildlife source, whereas the other one is composed of gyrA alleles mainly shared by domesticated mammals. However, the peptide group #1 from the main branch which is encoded by the largest number of alleles (N = 23), could be subdivided into two sets of sub-clusters: one set harboring strains isolated from domestic mammals (N = 9) and the other set being highly specific to environmental samples (N = 14). From this last set, five sequences (#19, 40, 74, 76 and 79) display a slightly higher GC content (Figure 2B) as a potential “trace signature” of different ecological niches. In addition, within this same peptide group #1, the nucleotide alleles with the synonymous substitution G408A (#11, 39, 40, 41, 56, 66 and 79) were never recovered from poultry strains. This change is also present in alleles from peptide group #14 previously discussed and linked to small mammals . The most obvious host signature established in our study is the non-synonymous substitution A64G corresponding to the change Ser22Gly in the amino acid sequence. This point mutation was previously observed by Ge et al.  in a study on antimicrobial resistance of strains isolated from poultry meat in which 76.2% ciprofloxacin-resistant C. jejuni harbored this particular substitution in their gyrA sequence (N = 42). Jesse et al.  also noticed this mutation in isolates from chicken and turkeys and suggested that it does not contribute to quinolone resistance but may be indicative of gyrA alleles predominantly found in poultry. Our results confirm this finding: 11 isolates with the Ser22Gly but without the Thr86Ile substitution were classified as susceptible to quinolones according to the cut off values recommended by the European commission  (see Additional file 3). Also, peptide groups #3, 4, 5 and 8 with this particular change on codon 22, are significantly associated with poultry source (P = 0.001). This host signature could be used as a specific molecular marker of domestic birds.
Our study also found that quinolone resistance was higher in isolates originating from poultry than from other sources. Recently, Han et al.  demonstrated that this particular mutation generates a fitness advantage for Campylobacter in chicken through a reduced supercoiling activity of the GyrA enzyme. As DNA supercoiling is directly involved in gene expression, their findings suggested that the altered function of the enzyme modulates the fitness of resistant strains whose prevalence persists in poultry production even in the absence of fluoroquinolone use. The European report on antimicrobial resistance in zoonotic bacteria  reported very high fluoroquinolone resistance levels in Campylobacter isolated from broilers (76%) and broiler meat (58%). Our results concur with the report in that resistance levels vary substantially in different hosts.