An evaluation of the species and subspecies of the genus Salmonella with whole genome sequence data: Proposal of type strains and epithets for novel S. enterica subspecies VII, VIII, IX, X and XI

Abstract

Species and subspecies within the Salmonella genus have been defined for public health purposes by biochemical properties; however, reference laboratories have increasingly adopted sequence-based, and especially whole genome sequence (WGS), methods for surveillance and routine identification. This leads to potential disparities in subspecies definitions, routine typing, and the ability to detect novel subspecies. A large-scale analysis of WGS data from the routine sequencing of clinical isolates was employed to define and characterise Salmonella subspecies population structure, demonstrating that the Salmonella species and subspecies were genetically distinct, including those previously identified through phylogenetic approaches, namely: S. enterica subspecies londinensis (VII), subspecies brasiliensis (VIII), subspecies hibernicus (IX) and subspecies essexiensis (X). The analysis also identified an additional novel subspecies, reptilium (XI). Further, these analyses indicated that S. enterica subspecies arizonae (IIIa) isolates were divergent from the other S. enterica subspecies, which clustered together and, on the basis of ANI analysis, subspecies IIIa was sufficiently distinct to be classified as a separate species, S. arizonae. Multiple phylogenetic and statistical approaches generated congruent results, suggesting that the proposed species and subspecies structure was sufficiently biologically robust for routine application. Biochemical analyses demonstrated that not all subspecies were distinguishable by these means and that biochemical approaches did not capture the genomic diversity of the genus. We recommend the adoption of standardised genomic definitions of species and subspecies and a genome sequence-based approach to routine typing for the identification and definition of novel subspecies.

Keywords: Brasiliensis; Essexiensis; Genomics; Hibernicus; Londinensis; Reptilium; Salmonella.

Fig. 1

A and B: Neighbour-Joining Tree based on the rMLST loci of 569 representative isolates (A) Neighbour-Joining Tree based on the cgMLST loci of 569 representative isolates (B). Neighbour-Joining Trees of 569 isolates rooted using S. bongori. These trees were constructed using the BIGSdb Genome Comparator tool and SplitsTree software and edited using the ITOL online tool. Each isolate was randomly selected from the 1801 initial isolates to represent up to 100 different rST profiles per species or subspecies. A: Neighbour-Joining Tree of 569 isolates, based on rMLST. The rMLST tree clustered all of the species and subspecies into distinct clades. The only isolate that didn't cluster as expected was SAL_BA7507AA, whose subspecies name is highlighted in yellow. B: Neighbour-Joining Tree of 569 isolates, based on cgMLST. Within the cgMLST tree SAL_BA7507AA (highlighted in yellow) clustered with subspecies II but represented a sister clade to the other subspecies II isolates included in the analysis. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Fig. 2

a and b: structure analysis (A) of 569 isolates using rMLST (B) of 57 isolates using cgMLST. structure analyses of representative isolates of the species and subspecies. The structure algorithm was run until no further higher-level taxonomical groups were observed. Both of the structure outputs were edited using the distruct software. A: The structure algorithm was performed on the rMLST profiles of dataset A (Table S2) isolates with a K of 12. S. bongori, S. arizonae and subspecies I, II, IIIb, IV, VI, VII, IX, and X showed very little admixture and were all distinct populations. In comparison subspecies VIII and XI were composed of multiple rMLST populations. Subspecies VIII was an admixture of subspecies VII, IV and I, all of which were present in similar proportions in each isolate. Subspecies XI was predominantly a novel population, however there was a high level of admixture from subspecies II. The isolate SAL_BA7507AA grouped with subspecies II. B: The structure algorithm was performed on the cgMLST profiles of Dataset B (Table S2) with a K of 18. Species S. bongori and S. arizonae and subspecies II, IIIb, IV, VI, VII, IX, X, and XI were all distinct populations with little admixture. Subspecies I serovars Paratyphi A and Typhi also formed a distinct population (red), but showed admixture with the rest of subspecies I. One isolate of subspecies VIII was admixed with subspecies IV. Isolate SAL_BA7507AA was highly admixed with five novel populations and some of subspecies II.

Fig. 3

A, B and C: ANI analysis of (A) type strains and recommended type strains using WGS (B) ANI analysis of type strains and recommended type strains using cgMLSTv1.0 (C) ANI analysis of type strains and recommended type strains using rMLST. ANI analyses of isolates from dataset C (Table S2) where each isolate was compared to all of the other species or subspecies isolates and with a secondary isolate that belonged to the same species or subspecies using the OrthoANI algorithm. A: ANI analysis of type strains and recommended type strains using WGS. Within this analysis all comparisons of species and subspecies against S. bongori and S. arizonae scored below 95%. Comparisons between the other subspecies ranged between 92.93 and 97.13%, with an average of 95.15%. Comparisons within the subspecies ranged from 97.89 to 99.99%. B: ANI analysis of type strains and recommended type strains using cgMLST. S. bongori and S. arizonae scored below 95%. Comparisons between the other subspecies were between 94.3 and 97.71%, with an average of 96.10% and comparisons within subspecies were between 98.79 and 100%. C: ANI analysis of type strains and recommended type strains using the rMLST loci. The patterns observed within the rMLST analysis was very similar to that observed within the cgMLST and WGS analyses. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Fig. 4

Dendogram based on metabolic phenotyping. Pairs of isolates from each (sub)species underwent metabolic phenotyping using the biolog GenIII MicroPlate Assays. Runs were conducted three times and an average of the three runs was used to create signal values. These values were used to construct a dendrogram based on metabolic respiration values.