Development of a Genomics-Based Approach To Identify Putative Hypervirulent Nontyphoidal Salmonella Isolates: Salmonella enterica Serovar Saintpaul as a Model

Abstract

While differences in human virulence have been reported across nontyphoidal Salmonella (NTS) serovars and associated subtypes, a rational and scalable approach to identify Salmonella subtypes with differential ability to cause human diseases is not available. Here, we used NTS serovar Saintpaul (S. Saintpaul) as a model to determine if metadata and associated whole-genome sequence (WGS) data in the NCBI Pathogen Detection (PD) database can be used to identify (i) subtypes with differential likelihoods of causing human diseases and (ii) genes and single nucleotide polymorphisms (SNPs) potentially responsible for such differences. S. Saintpaul SNP clusters (n = 211) were assigned different epidemiology types (epi-types) based on statistically significant over- or underrepresentation of human clinical isolates, including human associated (HA; n = 29), non-human associated (NHA; n = 23), and other (n = 159). Comparative genomic analyses identified 384 and 619 genes overrepresented among isolates in 5 HA and 4 NHA SNP clusters most significantly associated with the respective isolation source. These genes included 5 HA-associated virulence genes previously reported to be present on Gifsy-1/Gifsy-2 prophages. Additionally, premature stop codons in 3 and 7 genes were overrepresented among the selected HA and NHA SNP clusters, respectively. Tissue culture experiments with strains representing 4 HA and 3 NHA SNP clusters did not reveal evidence for enhanced invasion or intracellular survival for HA strains. However, the presence of sodCI (encoding a superoxide dismutase), found in 4 HA and 1 NHA SNP clusters, was positively correlated with intracellular survival in macrophage-like cells. Post hoc analyses also suggested a possible difference in intracellular survival among S. Saintpaul lineages. IMPORTANCE Not all Salmonella isolates are equally likely to cause human disease, and Salmonella control strategies may unintentionally focus on serovars and subtypes with high prevalence in source populations but are rarely associated with human clinical illness. We describe a framework leveraging WGS data in the NCBI PD database to identify Salmonella subtypes over- and underrepresented among human clinical cases. While we identified genomic signatures associated with HA/NHA SNP clusters, tissue culture experiments failed to identify consistent phenotypic characteristics indicative of enhanced human virulence of HA strains. Our findings illustrate the challenges of defining hypo- and hypervirulent S. Saintpaul and potential limitations of phenotypic assays when evaluating human virulence, for which in vivo experiments are essential. Identification of sodCI, an HA-associated virulence gene associated with enhanced intracellular survival, however, illustrates the potential of the framework and is consistent with prior work identifying specific genomic features responsible for enhanced or reduced virulence of nontyphoidal Salmonella.

Keywords: SNP clusters; comparative genomic analyses; human virulence; intracellular survival; invasion; nontyphoidal Salmonella; pathogen detection; phenotypic characterization; regulatory policy; serovar Saintpaul.

FIG 1

Phylogenetic analyses reveal a polyphyletic structure for S. Saintpaul. (A) Maximum likelihood phylogenetic tree inferred from core SNPs among representative isolates for 211 S. Saintpaul SNP clusters and reference isolates for (i) 313 unique Salmonella enterica subsp. enterica serovars and (ii) 5 additional Salmonella enterica subspecies. Clustering confidence was assessed using 1,000 bootstrap repetitions. The tree is rooted using the reference isolate (assembly accession no. GCA_000018625) of Salmonella enterica subsp. arizonae as the outgroup. Branch lengths represent the average pairwise number of nucleotide substitutions per site. Internal nodes indicating the most recent common ancestors (MRCAs) of Salmonella enterica subsp. enterica clades are labeled with the corresponding bootstrap values. All S. Saintpaul SNP clusters fall within clade A of Salmonella enterica subsp. enterica and belong to one of four phylogenetic groups (designated S. Saintpaul groups I to IV; gray sections). The ancestral node of each S. Saintpaul phylogenetic group is marked by a red star, and the exact bootstrap value of the node is shown. (B) Maximum likelihood phylogenetic tree constructed based on core SNPs for all representative isolates and reference isolates under the S. Saintpaul group I ancestral node shown in panel A. Clustering confidence was assessed using 1,000 bootstrap repetitions. The tree is rooted using two S. Saintpaul representative isolates (assembly accession no. GCA_011384505.1 and GCA_008999795.1) as outgroups. Branch lengths represent the average pairwise number of nucleotide substitutions per site. Most S. Saintpaul SNP clusters within group I belong to one of the four monophyletic lineages (designated lineages IA to ID). The ancestral node of each group I lineage is marked by a red star, and the exact bootstrap value of the node is shown. The 5 HA and 5 NHA SNP clusters that had the most significant association with the corresponding isolation sources are all positioned within group I and marked in red and blue at the corresponding tree leaves, respectively.

FIG 2

Genes that are overrepresented among HA or NHA isolates may be attributed to plasmids and prophages. Maximum likelihood phylogeny is inferred from core SNPs among 90 S. Saintpaul isolates representing 5 HA and 4 NHA SNP clusters, respectively. Clustering confidence was assessed using 1,000 bootstrap repetitions. The tree is rooted using the reference isolate of S. Haifa (assembly accession no. GCA_006378355) and S. Coeln (assembly accession no. GCA_008488945) as outgroups. Branch lengths represent the average pairwise number of nucleotide substitutions per site. The ancestral node of each lineage is marked by a red star, and the exact bootstrap value of the node is shown. Bar graphs to the right of the phylogeny show, for each isolate, the number of total genes (dark gray, HA; light gray, NHA), plasmid-borne genes (dark green, HA; light green, NHA), and prophage-borne genes (dark purple, HA; light purple, NHA) present in the genomes that are overrepresented among HA or NHA isolates.

FIG 3

Selected genomic signatures may be responsible for the association of HA or NHA SNP clusters with their corresponding isolation sources. The inner section shows the maximum likelihood phylogeny inferred from core SNPs among 90 S. Saintpaul isolates representing 5 and 4 HA and NHA SNP clusters, respectively. Clustering confidence is assessed using 1,000 bootstrap repetitions. The tree is rooted using the reference isolates of S. Haifa (assembly accession no. GCA_006378355) and S. Coeln (assembly accession no. GCA_008488945) as outgroups. Branch lengths represent the average pairwise number of nucleotide substitutions per site. The ancestral node of each lineage is marked by a red star, and the exact bootstrap value of the node is shown. The outer section shows the genotypic patterns across the isolates with respect to the selected genomic signatures; these include (i) virulence-associated genes carried on prophages Gifsy-1/Gifsy-2 (denoted by circles), (ii) core-SNP-affected genes that tend to be intact in HA isolates while disrupted by a PMSC in NHA isolates (denoted by red stars), and (iii) core-SNP-affected genes that tend to be intact in NHA isolates while disrupted by a PMSC in HA isolates (denoted by blue stars).

FIG 4

Lineage and genotype of bglA/bepF, but not epi-type, are associated with the differences in the ability of S. Saintpaul strains to invade HIEC-6 cells. HIEC-6 cells were infected by representative strains of S. Saintpaul (MOI, 100 per cell), followed by incubation at 37°C with 5% CO₂ for 1 h. Extracellular salmonellae were killed by treating the infected HIEC-6 cells with tissue culture medium containing 20 μg/mL gentamicin, followed by a 1-h incubation at 37°C with 5% CO₂. The invasion efficiency is represented by the recovery rate, defined as the ratio of the number of bacterial cells that successfully invaded the HIEC-6 cells to the number of bacterial cells used for infection. All data points of recovery rates are shown, in addition to the means (diamonds) ± standard errors (error bars) across biological replicates and strains for each epi-type (A), lineage (B), and genotype of bglA/bepF (C). The estimated marginal means of recovery rate (RREMM) based on the linear regression model (the INV model) (Table 4) are shown as red dots. Post hoc comparisons of the RREMM were performed between strains (i) from different lineages (Tukey’s honestly significant difference test), (ii) with different genotypes of bglA/bepF (t test), and (iii) from SNP clusters classified as different epi-types (i.e., HA or NHA; t test). P values for each of the comparisons are presented (red, P < 0.05; black, P ≥ 0.05).

FIG 5

Factors associated with the survival of S. Saintpaul strains differ across multiple time periods after their entry into human macrophage-like cells. (A) Effect of lineage on the intracellular survival of S. Saintpaul strains from 0 to 2 hpi. (B) Effect of lineage on the intracellular survival of S. Saintpaul strains from 2 to 6 hpi. (C) Effect of presence/absence of sodCI on the intracellular survival of S. Saintpaul strains from 6 to 24 hpi. THP-1 cells were differentiated into macrophage-like cells in the presence of 20 ng/mL PMA for 3 days. Macrophage-like cells were then infected by representative strains of S. Saintpaul (MOI, 10 per cell), followed by incubation at 37°C with 5% CO₂ for up to 24 h. Extracellular salmonellae were killed by treating the infected macrophage-like cells with tissue culture medium containing 20 μg/mL gentamicin, followed by incubation at 37°C with 5% CO₂ for the indicated time periods. The intracellular survival within a given time period is represented by the difference in log₁₀ CFU/mL, calculated by subtracting the log₁₀ CFU/mL at the beginning from the log₁₀ CFU/mL at the end of the time period. All data points of the difference in log₁₀ CFU/mL are shown, in addition to the means (diamonds) ± standard errors (error bars) across biological replicates and strains for different lineages between 0 and 2 hpi (A) and 2 and 6 hpi (B) and for different genotypes of sodCI between 6 and 24 hpi (C). The estimated marginal means of the difference in log₁₀ CFU/mL (LDEMM) based on the linear regression models (the ICS-1, ICS-2, and ICS-3 models) (Table 4) are shown as red dots. Post hoc comparisons of the LDEMM were performed between strains from different lineages (Tukey’s honestly significant difference test and t test) and with different genotypes (t test). P values for each of the comparisons are presented (red, P < 0.05; black, P ≥ 0.05).