E. coli O157:H7 isolates selected for the study and assembled.

Three sets of isolates, all originating from cattle or humans, were included in the study.

Maximum likelihood core SNP tree of the 1,215 E. coli O157:H7 isolates referenced in the study.

This includes 659 isolates from Alberta, Canada, from 2007 through 2019, 494 isolates from the U.S. from 1996 through 2019, and 62 isolates from elsewhere around the globe from 2007 to 2016. The tree was rooted at clade A. (A) shows all clades, with tips colored according to clade, geographic origin shown on the inner ring, and species of origin on the outer ring. (B) shows clade G(vi), which constituted 73.6% of all isolates and 88.3% of Alberta isolates. Tips are colored by geographic origin, and the ring indicates species of origin.

Distribution of study isolates by geographic source, clade, and Shiga toxin gene (stx) profile

SNP distance between E. coli O157:H7 strains and their nearest relative, by species, Alberta, Canada, 2007-2015.

Distances for isolates from 121 reported human cases and 108 beef cattle are shown. Cattle isolates were highly related with 56.5% of cattle isolates within 5 SNPs of another cattle isolate and 94.4% within 25 SNPs. Human isolates showed a bimodal distribution in their relationship to cattle isolates, with 86.0% within 50 SNPs of a cattle isolate and the remainder 185-396 SNPs apart. Nineteen human isolates (15.7%) were within 5 SNPs of a cattle isolate.

Maximum clade credibility (MCC) tree of structured coalescent analysis of E. coli O157:H7 strains isolated from 115 reported human cases and 84 beef cattle in Alberta, Canada, 2007-2015.

Isolates were down-sampled prior to phylodynamic analysis to remove isolates that were highly similar. The structured coalescent analysis estimated migration and state transitions between humans and cattle. The MCC tree was colored by inferred host, cattle (blue) or human (orange). The minimum SNP distance between all isolates in the LPL, and the number of human and cattle isolates after down-sampling are shown for each LPL. The majority of ancestral nodes inferred as cattle suggests cattle as the primary reservoir. The root was estimated at 1802 (95% HPD 1731, 1861). Eleven locally persistent lineages (LPLs) were identified, all in the G(vi) clade and labeled LPL 1 through 11. For each LPL, the minimum number of SNPs all isolates are within (30 was set as the maximum) and the number of human and cattle isolates within the LPL, not including down-sampled isolates, are shown. With down-sampled isolates reincorporated, LPLs accounted for 46 human (38.0%) and 71 cattle (65.7%) isolates. The structured coalescent model estimated 108 cattle-to-human state transitions between branches, compared to only 14 human-to-human transitions, inferring cattle as the origin of 88.5% of human lineages.

Extension of Alberta, Canada E. coli O157:H7 analysis to include 229 randomly selected study isolates and 430 additional public health isolates available from 2009 to 2019.

The maximum clade credibility (MCC) tree was constructed from a coalescent analysis with constant population size after down-sampling. Six locally persistent lineages (LPLs) in clade G(vi) continued to be associated with disease after the initial study period. LPLs are colored and labeled as in Figure 4. After re-incorporating the down-sampled isolates, 74.7% of reported cases in 2018 and 2019 were associated with an LPL.

Shiga toxin gene (stx) profile by locally persistent lineage (LPL) status of extended analysis of Alberta, Canada E. coli O157:H7 isolated from cattle and humans, 2007 to 2019.

The stx profile across all clades shifted from the initial study period (2007-2015) to the later study period (2016-2019), with more of the virulent stx2a-only profile observed in 2018 and 2019 than in previous years. In 2018 and 2019, 51.2% of LPL isolates carried only stx2a, compared to 10.9% of non-LPL isolates. The peak in sequences in 2014 is due to two outbreaks; routine sequencing began in 2018 and 2019, accounting for the rise in sequenced cases during those years.

Comparison of Alberta, U.S., and global E. coli O157:H7 isolates.

Tips are colored based on isolate origin, and LPLs from Figure 4 are highlighted. A total of nine U.S. isolates arose from Alberta LPLs 2, 4, 7, and 11, all of which had Alberta isolates predating the U.S. isolates. No global isolates were associated with Alberta LPLs. Clade A was excluded from the analysis due to its high level of divergence from the rest of the E. coli O157:H7 population.

Genomic clustering of 690 Alberta E. coli O157 isolates.

Clustering performed from raw reads using PopPUNK v2.5.0 with 10,146 E. coli reference genomes.23 From the 246 isolates selected for sequencing for the study and 445 additional Alberta Health isolates included for contextualization, one isolate was removed prior to clustering analysis, because it was identified through metadata review as an environmental (non-human, non-cattle) isolate. Cluster 1 included the Sakai and EDL933 reference strains. Clusters 826 and 827 were novel clusters. Isolates outside of Cluster 1 were excluded from all subsequent analyses.

Key parameters drawn from the posterior distributions of four Markov chain Monte Carlo chains using different starting seeds compared to draws from the prior distribution.

The posterior distributions of the four chains for the tree height, clock rate, kappa, cattle effective population size, human effective population size, and backward migration rate all differed substantially from the prior distribution, shown on the far right in green for each graph.

Maximum clade credibility (MCC) tree of structured coalescent analysis of 168 subsampled isolates from humans and cattle from Alberta, 2007-2015.

LPLs are labeled as in the primary analysis (Figure 4). From an initial 121 human and 108 cattle isolates, our primary analysis contained 115 human and 84 cattle isolates after down-sampling. In this sensitivity analysis, we repeated our primary analysis with a subsample of 84 randomly selected isolates from humans and the 84 cattle isolates remaining after down-sampling. As in the primary analysis, cattle were inferred as the host of the majority of ancestral nodes. The root was estimated at 1796 (95% HPD 1722, 1859), very close to the root at 1802 estimated in the primary analysis. The most recent common ancestor (MRCA) of clade G(vi) strains in Alberta was inferred to be a cattle strain, dated to 1968 (95% HPD 1956, 1979), compared to 1969 in the primary analysis. We estimated 82 (95% HPD 79, 84) human lineages arose from cattle lineages, and 5 (95% HPD 0, 11) arose from other human lineages, meaning we inferred that 94.3% of human lineages arose from cattle lineages, compared to 88.5% in the primary analysis. The LPLs identified in this sensitivity analysis were mostly identical to those identified in the primary analysis. Differences in the sensitivity analysis were that G(vi)-AB LPL 8 expanded to a larger set of isolates; G(vi)-AB LPL 9, which included only 5 isolates in the primary analysis, was no longer identified as an LPL, as it no longer met the 5-isolate criterion; and there were minor topological changes.

Sampled trees from four independent chains of the Alberta 2007-2015 structured coalescent analysis.

Across the four chains, 1,070,460,000 trees were sampled. Depicted here are 963,414,000 post-burn-in trees, with panels A-D each showing the trees from one chain. Lineages inferred with cattle ancestry are shown in blue, lineages inferred with human ancestry are shown in green. Reflecting the strong support for the lineages identified as LPLs, the tree topology is well resolved.

LPLs from the primary analysis defined using alternate SNP thresholds of 50 (left) and 75 (right) SNPs.

LPL 50.1 includes G(vi)-AB LPLs 1, 2, and 3 from the primary analysis using a SNP threshold of 30. LPLs 50.2 and 50.3 correspond to G(vi)-AB LPLs 4 and 5, respectively. LPL 50.4 includes G(vi)-AB LPLs 6, 7, and 8. LPL 75.1 includes all of the above LPLs at the lower thresholds. All 50- and 75-threshold LPLs in this section of the tree include isolates not included at the lower SNP threshold(s). No part of LPLs 50.5 or 50.6 was identified as an LPL using a threshold of 30, but LPLs 50.7 and 50.8 correspond exactly to G(vi)-AB LPLs 9 and 10. LPL 75.2 includes LPLs 50.5-50.8 and G(vi)-AB LPLs 9 and 10. G(vi)-AB LPL 11, LPL 50.9, and LPL 75.3 are all identical. At the 75-SNP threshold, two additional LPLs are identified with no corresponding LPLs at the 30- or 50-SNP thresholds. One of these is outside the G(vi) clade.

Analyses conducted and model priors.

Estimated migrations from structured coalescent analysis of Alberta, U.S., and global isolates, excluding clade A.