Variation in virulence among clades of Escherichia coli O157:H7 associated with disease outbreaks

Abstract

Escherichia coli O157:H7, a toxin-producing food and waterborne bacterial pathogen, has been linked to large outbreaks of gastrointestinal illness for more than two decades. E. coli O157 causes a wide range of clinical illness that varies by outbreak, although factors that contribute to variation in disease severity are poorly understood. Several recent outbreaks involving O157 contamination of fresh produce (e.g., spinach) were associated with more severe disease, as defined by higher hemolytic uremic syndrome and hospitalization frequencies, suggesting that increased virulence has evolved. To test this hypothesis, we developed a system that detects SNPs in 96 loci and applied it to >500 E. coli O157 clinical strains. Phylogenetic analyses identified 39 SNP genotypes that differ at 20% of SNP loci and are separated into nine distinct clades. Differences were observed between clades in the frequency and distribution of Shiga toxin genes and in the type of clinical disease reported. Patients with hemolytic uremic syndrome were significantly more likely to be infected with clade 8 strains, which have increased in frequency over the past 5 years. Genome sequencing of a spinach outbreak strain, a member of clade 8, also revealed substantial genomic differences. These findings suggest that an emergent subpopulation of the clade 8 lineage has acquired critical factors that contribute to more severe disease. The ability to detect and rapidly genotype O157 strains belonging to such lineages is important and will have a significant impact on both disease diagnosis and treatment guidelines.

Fig. 1.

Genetic relatedness of E. coli O157 among 403 O157 and closely related O55:H7 strains based on 96 SNPs. (A) The location of 83 genes within 96 SNP loci on the E. coli O157:H7 genomic map of the Sakai strain. Real-time PCR assays detected 52 loci with nonsynonymous polymorphisms (black circles), 43 loci with synonymous polymorphisms (white circles), and one locus (uidA-686) with a GG insertion (open triangle). (B) The distribution of nucleotide diversity (π) across 96 SNP loci. Diversity ranges from 0 for two monomorphic SNP loci to a maximum between 0.45 and 0.50 for 26 loci. The average π for the 96 loci is 0.212 ± 0.199. (C) Phylogenetic relationships among SNP genotypes (SGs) using the minimum evolution algorithm based on the distance matrix of pairwise differences between SGs. The consensus tree is shown with the percentages at the nodes of the >70% bootstrap confidence values based on 1,000 replicates. Both the GUD⁺ and Sor⁺, which occur in clade 9, are negative (GUD⁻ and Sor⁻) in the derived clades 1–8.

Fig. 2.

The phylogenetic network applied to 48 parsimoniously informative (PI) sites using the Neighbor-net algorithm for 528 E. coli O157 strains. The colored ellipses mark clades supported in the minimum evolution phylogeny. The numbers at the nodes denote the SNP genotypes (SGs) 1–39, and the white circle nodes contain two SGs that match at the 48 PI sites. The seven SGs found among multiple continents are marked with squares.

Fig. 3.

Distribution of Shiga toxin (Stx) genes in E. coli O157 clades. (A) The frequency of 528 O157 strains that were classified into one of nine clades based on SNP genotyping, ranked from left to right in the histogram by decreasing frequency. The four most common clades were clades 2 (47.6%), 8 (25.4%), 3 (10.6%), and 7 (7.3%). (B) Distribution of Shiga toxin variants (stx1, stx2, and stx2c) among 519 of the 528 O157 strains organized into nine clades. The percentage of PCR assay positive strains overall is given in parentheses.

Fig. 4.

Odd ratios with 95% confidence intervals (dotted lines) highlighting the association between patient characteristics and infection with specific clades. Logistic regression models were adjusted for age, gender, bloody diarrhea, diarrhea, abdominal pain, chills, HUS, hospitalization, and body aches. Dark circles show significant associations.