A multisite genomic epidemiology study of Clostridioides difficile infections in the USA supports differential roles of healthcare versus community spread for two common strains

Abstract

Clostridioides difficile is the leading cause of healthcare-associated infectious diarrhoea. However, it is increasingly appreciated that healthcare-associated infections derive from both community and healthcare environments, and that the primary sites of C. difficile transmission may be strain-dependent. We conducted a multisite genomic epidemiology study to assess differential genomic evidence of healthcare vs community spread for two of the most common C. difficile strains in the USA: sequence type (ST) 1 (associated with ribotype 027) and ST2 (associated with ribotype 014/020). We performed whole-genome sequencing and phylogenetic analyses on 382 ST1 and ST2 C. difficile isolates recovered from stool specimens collected during standard clinical care at 3 geographically distinct US medical centres between 2010 and 2017. ST1 and ST2 isolates both displayed some evidence of phylogenetic clustering by study site, but clustering was stronger and more apparent in ST1, consistent with our healthcare-based study more comprehensively sampling local transmission of ST1 compared to ST2 strains. Analyses of pairwise single-nucleotide variant (SNV) distance distributions were also consistent with more evidence of healthcare transmission of ST1 compared to ST2, with 44 % of ST1 isolates being within two SNVs of another isolate from the same geographical collection site compared to 5.5 % of ST2 isolates (P-value=<0.001). Conversely, ST2 isolates were more likely to have close genetic neighbours across disparate geographical sites compared to ST1 isolates, further supporting non-healthcare routes of spread for ST2 and highlighting the potential for misattributing genomic similarity among ST2 isolates to recent healthcare transmission. Finally, we estimated a lower evolutionary rate for the ST2 lineage compared to the ST1 lineage using Bayesian timed phylogenomic analyses, and hypothesize that this may contribute to observed differences in geographical concordance among closely related isolates. Together, these findings suggest that ST1 and ST2, while both common causes of C. difficile infection in hospitals, show differential reliance on community and hospital spread. This conclusion supports the need for strain-specific criteria for interpreting genomic linkages and emphasizes the importance of considering differences in the epidemiology of circulating strains when devising interventions to reduce the burden of C. difficile infections.

Keywords: C. difficile; community; genomic epidemiology; healthcare; transmission; whole-genome sequencing.

Fig. 1.

Maximum-likelihood phylogenetic trees of newly sequenced C. difficile isolates that are ST1 and ST2. Tips are coloured by fluroquinolone-resistant (FQR) vs fluroquinolone-sensitive (FQS) as determined by the presence of previously identified fluroquinolone-resistance-associated gryA and gyrB alleles. Previously identified ST1 lineages (FQS, FQR1 and FQR2) are highlighted and collection site is included in an adjacent heatmap. Tree scales are in single-nucleotide changes per quality- and recombination-filtered site.

Fig. 2.

Pairwise single-nucleotide variant (SNV) distribution between pairs of isolates from the same collection site vs pairs of isolates from geographically distinct collection sites for both ST1 and ST2. The black diamond indicates fifth percentile SNV distances for each category.

Fig. 3.

(a) Bar plot showing the proportion of ST1 and ST2 isolates that are genomically linked to another isolate, either from the same collection site only (green) or from at least one different collection site (orange), at varying SNV thresholds. (b) Scatter plot of days between collection and pairwise SNV distance up to 10 SNVs, where each dot represents 1 pair of isolates. Points are coloured by whether they are collected from the same geographical collection site (green) or different geographical collection sites (orange). Points are jittered to improve clarity.

Fig. 4.

Posterior probability density of the evolutionary rates estimates for C. difficile ST1 and ST2 lineages, with ST1 isolates considered together as well as separated out into FQR-ST1 and FQS-ST1 isolates. Dark shaded areas of the density curves indicate the lower 2.5 % and upper 97.5 % of the distributions; light shaded areas indicate 95 % credible intervals. Evolutionary rates are considered significantly different from one another when the 95 % credible intervals of their posterior probability densities do not overlap.