Major role of iron uptake systems in the intrinsic extra-intestinal virulence of the genus Escherichia revealed by a genome-wide association study

Abstract

The genus Escherichia is composed of several species and cryptic clades, including E. coli, which behaves as a vertebrate gut commensal, but also as an opportunistic pathogen involved in both diarrheic and extra-intestinal diseases. To characterize the genetic determinants of extra-intestinal virulence within the genus, we carried out an unbiased genome-wide association study (GWAS) on 370 commensal, pathogenic and environmental strains representative of the Escherichia genus phylogenetic diversity and including E. albertii (n = 7), E. fergusonii (n = 5), Escherichia clades (n = 32) and E. coli (n = 326), tested in a mouse model of sepsis. We found that the presence of the high-pathogenicity island (HPI), a ~35 kbp gene island encoding the yersiniabactin siderophore, is highly associated with death in mice, surpassing other associated genetic factors also related to iron uptake, such as the aerobactin and the sitABCD operons. We confirmed the association in vivo by deleting key genes of the HPI in E. coli strains in two phylogenetic backgrounds. We then searched for correlations between virulence, iron capture systems and in vitro growth in a subset of E. coli strains (N = 186) previously phenotyped across growth conditions, including antibiotics and other chemical and physical stressors. We found that virulence and iron capture systems are positively correlated with growth in the presence of numerous antibiotics, probably due to co-selection of virulence and resistance. We also found negative correlations between virulence, iron uptake systems and growth in the presence of specific antibiotics (i.e. cefsulodin and tobramycin), which hints at potential "collateral sensitivities" associated with intrinsic virulence. This study points to the major role of iron capture systems in the extra-intestinal virulence of the genus Escherichia.

Fig 1. The HPI is strongly associated with the extra-intestinal virulence phenotype assessed in the mouse sepsis assay.

A) Core genome phylogenetic tree of the Escherichia strains used in this study rooted on E. albertii strains. Outer ring reports virulence as the number of killed mice over the 10 inoculated per strain, inner ring the phylogroup, clade or species each strain belongs to. B) Results of the unitigs association analysis: for each gene the minimum association p-value and average minimum allele frequency (MAF) across all mapped unitigs is reported. The gene length fraction is computed by dividing the total length of mapped unitigs by the length of the gene. The color of each gene follows the same key as panel C. C) Results of the gene presence/absence association analysis; only those genes with at least one associated unitig mapped to them are represented. D) Scatterplot of gene frequency versus frequency of associated unitigs; points on the diagonal indicate hits where the association is most likely due to a gene’s presence/absence pattern rather than a SNP. The color of each gene follows the same key as panel C. E) The structure of the HPI and of the aerobactin and sitABCD operons in strain IAI39; all associated genes are highlighted.

Fig 2. Phenotypic consequences of HPI deletion.

A) Deletion of HPI leads to a decrease in production of yersiniabactin. Production of yersiniabactin is measured using a luciferase-based reporter (Methods). Strains marked with a “-” and “+” sign indicate a negative and positive control, respectively. The red dashed line indicates an arbitrary threshold for yersiniabactin production, derived from the average signal recorded from the negative controls plus two standard deviations. B-C) Deletion of HPI leads to an increase in survival after infection. Survival curves for wild-type strains and the corresponding irp2 deletion mutant, built after infection of 20 mice for each strain. B) Survival curve for strain NILS46. C) Survival curve for strain NILS9.

Fig 3. Growth profiles can predict virulence and presence of virulence factors.

A-D) Volcano plots for the correlation between the strains’ growth profiles and: A) virulence levels, B) presence of the HPI, C) presence of aerobactin, and D) presence of sitABCD. E-F) Use of the strains’ growth profiles to build a predictor of virulence levels and presence of the three iron uptake systems. E) Receiver operating characteristic (ROC) curves and F) Precision-Recall curve for the four tested predictors. G) Feature importance for the predictors, showing the top 15 conditions contributing to the virulence level predictor.