Characterization of Basal Transcriptomes Identifies Potential Metabolic and Virulence-Associated Adaptations Among Diverse Nontyphoidal Salmonella enterica Serovars

Abstract

The zoonotic pathogen Salmonella enterica includes >2,600 serovars, which differ in the range of hosts they infect and the severity of disease they cause. To further elucidate the mechanisms behind these differences, we performed transcriptomic comparisons of nontyphoidal Salmonella (NTS) serovars with the model for NTS pathogenesis, S. Typhimurium. Specifically, we used RNA-seq to characterize the understudied NTS serovars S. Javiana and S. Cerro, representing a serovar frequently attributed to human infection via contact with amphibians and reptiles, and a serovar primarily associated with cattle, respectively. Whole-genome sequence (WGS) data were utilized to ensure that strains characterized with RNA-seq were representative of their respective serovars. RNA extracted from representative strains of each serovar grown to late exponential phase in Luria-Bertani (LB) broth showed that transcript abundances of core genes were significantly higher (p<0.001) than those of accessory genes for all three serovars. Inter-serovar comparisons identified that transcript abundances of genes in Salmonella Pathogenicity Island (SPI) 1 were significantly higher in both S. Javiana and S. Typhimurium compared to S. Cerro. Together, our data highlight potential transcriptional mechanisms that may facilitate S. Cerro and S. Javiana survival in and adaptation to their respective hosts and impact their ability to cause disease in others. Furthermore, our analyses demonstrate the utility of omics approaches in advancing our understanding of the diversity of metabolic and virulence mechanisms of different NTS serovars.

Keywords: Salmonella; genomics; pathogen; transcriptomics; virulence.

Figure 1

Genomic comparisons identify similar pan genome sizes for broad host range serovars Javiana and Typhimurium, but a smaller pan genome for cattle-associated serovar S. Cerro. (A) Maximum likelihood phylogenetic tree constructed from 55,425 core SNPs identified with kSNP3 for isolates representing NTS serovars Javiana (purple), Cerro (blue), Typhimurium (pink) and 18 additional serovars that represent the most common NTS serovars isolated from human clinical infections in the US; one additional assembly of S. Javiana (shown in black font) was included as this serovar is among the top four serovars isolated from human clinical infections in the US, however this isolate’s assembly was not included in the genomic analyses (i.e., core and pan genome size analyses; CDC, 2018b). Clade designations shown are based on those determined previously (Worley et al., 2018) for the serovars included in the tree. RAxML was used to infer the maximum likelihood tree utilizing a general time-reversible model with gamma-distributed sites and Lewis ascertainment bias correction; bootstrap values represent the average of 1,000 bootstrap repetitions and only bootstrap values >70 are shown. S. enterica subsp. indica SRR2585491 was used as an outgroup to root the phylogeny. Strains characterized by RNA-seq are shown in bold font. (B) Comparison of pan genome sizes for each serovar based on gene presence/absence analyses of 16 representative isolates per serovar.

Figure 2

Transcript abundances of core genes are significantly higher than accessory genes in all three serovars. (A) Boxplot of average transcript abundances of core and accessory genes from three independent replicates of S. Cerro FSL R8-2349, Javiana FSL S5-0395, and Typhimurium ATCC 14028S cultured to late exponential phase in LB broth at 37°C. Significance was assessed with ANOVA after fitting a linear mixed-effects model to transcript abundances data and calculating the least square-means. Histograms of average transcript abundances of core (light colors) and accessory genes (dark colors) in (B) S. Cerro FSL R8-2349, (C) S. Javiana FSL S5-0395, and (D) S. Typhimurium ATCC 14028S. Results represent the average of three independent replicates. Core and accessory genes of 16 isolates per serovar were defined using Roary (Page et al., 2015).

Figure 3

Comparison of transcript abundances in strains grown to late exponential phase in LB reveals many genes and orthologous clusters that are differentially expressed in inter-serovar comparisons. Heat map comparing the transcript abundances of the 50 most significantly differentially expressed (FDR<0.05, log₂FC>|2|) orthologous clusters in (A) S. Typhimurium ATCC 14028S vs. S. Cerro FSL R8-2349, (B) S. Typhimurium ATCC 14028S vs. S. Javiana FSL S5-0395, and (C) S. Cerro FSL R8-2349 vs. S. Javiana FSL S5-0395. Differential expression was determined using variance modeling at the observational level (voom method), implemented in the R package limma. Transcript abundances from three biological replicates are shown (labeled as “Rep. 1–3” on the x-axis); z-scores represent the number of standard deviations that each log transformed transcript abundance for a given replicate differs from the average transcript abundances across all repetitions for that orthologous cluster. Clustering was performed using the Ward’s minimum variance method (Ward, 1963) based on Euclidean distance matrices. Genes are shown in bold if transcript abundances are significantly different in multiple comparisons (e.g., sipA shows significantly higher transcript abundances in S. Javiana compared to both S. Cerro and S. Typhimurium).

Figure 4

Inter-serovar comparisons demonstrate that SPI-1 transcript abundances are higher in serovars Javiana and Typhimurium compared to Cerro. Comparison of transcript abundances of genes in the SPI-1 locus. Transcript abundances were averaged from three biological replicates per strain (S. Cerro FSL R8-2349, S. Javiana FSL S5-0395, and S. Typhimurium ATCC 14028S). z-scores represent the number of standard deviations that a given strain’s transcript abundances differ from the transcript abundances across all strains for that gene.

Figure 5

qPCR demonstrates how inherent variability in undefined media can impact transcript abundances of genes involved in metabolic functions. Boxplots comparing the relative expression (2^−∆∆Ct) of cbiF (cob operon; involved in vitamin B12 biosynthesis functions), eutB (eut operon; ethanolamine utilization), pduC (pdu operon; 1,2-propanediol utilization), and sicA (SPI-1 locus; T3SS-mediated invasion) of S. Cerro FSL R8-2349 and S. Typhimurium ATCC 14028S normalized to the expression of rpoB. Results represent three independent replicates performed in technical duplicate. Significance was assessed with ANOVA after fitting a linear mixed effects model to the relative expression data and calculating least square means.