caviae clade (Additional file 3: Figure S3 a). Distance and ML trees were reconstructed for each of the 7 genes and compared to the concatenated sequence-based click here trees. For all genes and phylogenetic methods, single locus phylogenies (SLPA) displayed lower bootstrap values than MLPA trees (data not shown). Moreover, differences in branching order were observed in SLPA, suggesting the occurrence of recombination events (data not shown). In detail, phylogenetic discordance was observed for 11 strains based on single-gene phylogenetic analysis:

all of these strains grouped in a robust cluster that was different from the cluster defined based on the 6 other genes or the concatenated sequence (shown in bold text in Table 1). Identical alleles were observed in strains belonging to different MLPA clusters, i.e., gyrB allele 83, common to the two environmental strains A. veronii strain AK250 and A. hydrophila strain AK218; zipA allele 97, common to the A. media and A. enteropelogenes type strains; and zipA allele 94, which was identical in the A. caviae type strain AZD1390 solubility dmso and A. salmonicida strain CIP104001 (Table 1). In addition, strain BVH53 belonged to the A. veronii clade in the MLPA, while it was robustly grouped with the A. jandaei type strain in the gyrB-based phylogeny (bootstrap value of 100% in both the ML and distance-based trees) (data not

shown). Similarly, among the isolated strains, the A. fluvialis type strain showed a divergent phylogenetic position

between the gltA-based tree, where it robustly grouped with the A. schubertii type strain, and other gene-based phylogenies or the MLPA. Finally, old strain BVH39 grouped within the A. salmonicida clade in the multilocus tree, while it was excluded from the corresponding clade defined in the dnaK-based tree. These phylogenetic incongruities revealed a total of 12 recombination events (0.9% of the sequences), which occurred in 11 strains (4, 3 and 4 strains of human, animal and environmental origin, respectively) (5.8% of the total strains) and concerned 5 out of the 7 genes addressed in our MLSA scheme, i.e., dnaK (1 strain), gltA (1 strain), gyrB (4 strains), tsf (3 strains), and zipA (3 strains) (Table 1). Multilocus phylogenetic trees reconstructed excluding the strains subjected to recombination showed increased bootstrap values for the A. veronii clade (90 to 100%) as well as for most interclade nodes, confirming that recombination distorted the MLPA (data not shown). Despite its relatively low frequency of occurrence in the genus Aeromonas, recombination may account, at least in part, for some controversial taxonomic issues. For example, strain CCM 1271 is closely related to A. bestiarum in the gyrB-based phylogenetic tree (data not shown), whereas it is clearly individualized from this species in the MLPA. Discussion In this study, we investigated the genetic diversity and population structure linked with strain origin using MLSA.