Genomic surveillance of Clostridioides difficile transmission and virulence in a healthcare setting

Abstract

Clostridioides difficile infection (CDI) is a major cause of healthcare-associated diarrhea, despite the widespread implementation of contact precautions for patients with CDI. Here, we investigate strain contamination in a hospital setting and the genomic determinants of disease outcomes. Across two wards over 6 months, we selectively cultured C. difficile from patients (n = 384) and their environments. Whole-genome sequencing (WGS) of 146 isolates revealed that most C. difficile isolates were from clade 1 (131/146, 89.7%), while only one isolate of the hypervirulent ST1 was recovered. Of culture-positive admissions (n = 79), 19 (24%) patients were colonized with toxigenic C. difficile on admission to the hospital. We defined 25 strain networks at ≤2 core gene single nucleotide polymorphisms; two of these networks contain strains from different patients. Strain networks were temporally linked (P < 0.0001). To understand the genomic correlates of the disease, we conducted WGS on an additional cohort of C. difficile (n = 102 isolates) from the same hospital and confirmed that clade 1 isolates are responsible for most CDI cases. We found that while toxigenic C. difficile isolates are associated with the presence of cdtR, nontoxigenic isolates have an increased abundance of prophages. Our pangenomic analysis of clade 1 isolates suggests that while toxin genes (tcdABER and cdtR) were associated with CDI symptoms, they are dispensable for patient colonization. These data indicate that toxigenic and nontoxigenic C. difficile contamination persist in a hospital setting and highlight further investigation into how accessory genomic repertoires contribute to C. difficile colonization and disease.

Importance: Clostridioides difficile infection remains a leading cause of hospital-associated diarrhea, despite increased antibiotic stewardship and transmission prevention strategies. This suggests a changing genomic landscape of C. difficile. Our study provides insight into the nature of prevalent C. difficile strains in a hospital setting and transmission patterns among carriers. Longitudinal sampling of surfaces and patient stool revealed that both toxigenic and nontoxigenic strains of C. difficile clade 1 dominate these two wards. Moreover, quantification of transmission in carriers of these clade 1 isolates underscores the need to revisit infection prevention measures in this patient group. We identified unique genetic signatures associated with virulence in this clade. Our data highlight the complexities of preventing transmission of this pathogen in a hospital setting and the need to investigate the mechanisms of in vivo persistence and virulence of prevalent lineages in the host gut microbiome.

Keywords: C. difficile infection; antibiotic resistance; pathogens; transmission; virulence factors; whole-genome sequencing.

Fig 1

Study sampling and testing overview. (a) We sampled leukemia and hematopoietic stem cell transplant wards at Barnes-Jewish Hospital in St. Louis, USA for 6 and 4 months, respectively. Patients were enrolled and sampled upon admission, and then weekly for their time in the study wards. Surfaces were sampled weekly across the duration of the study. All samples and stool collected as part of routine clinical care were subjected to selective culture and MALDI-TOF MS identification, and isolates were whole-genome sequenced. Results of EIA testing as part of routine care were obtained.

Fig 2

Total samples collected and phylogenetic relationships reveal carriers outnumber CDI patients, and bedrails are the most commonly contaminated surface. Total (a) isolates collected and (b) culture-positive episodes from each source. We found more carriers than CDI patients, and bedrails yielded the most C. difficile isolates. (c) Cladogram of all isolates collected during this study plus references.

Fig 3

Surfaces are a site of environmental contamination and potential for transmission from colonized and CDI patients. (a) Strain networks were defined by ≤2 MLST core gene SNP cutoff. Network 10 includes the non-toxigenic isolates from patient 2245, which are likely not responsible for the CDI. (b) Absolute value of days between isolates within strains and between strains. Isolates within the same strain were significantly temporally linked (P < 2.2e−16, Wilcoxon test). (c) Number of comparisons in each group that fall within strain cutoff. Patient: between two isolates collected from the same patient; bedrail: between a patient isolate and an isolate taken from their bedrail; keyboard: between a patient isolate and an isolate taken from their keyboard. Fisher’s exact test, BH corrected. (d) Strain tracking diagram of transmission networks associated with more than one patient. Colors indicate MLST of network and horizontal lines indicate stay in a room. Patient 2330 shed C. difficile onto the bedrail, and patient 2336 later was identified as a carrier of the same strain.

Fig 4

Clade 1 is responsible for the majority of CDI cases and carries unique correlates to symptom severity. (a) EIA status by clade across this and a previous study (26). Fisher’s exact test, P < 0.01. (b) Differentially abundant genes between toxigenic and nontoxigenic isolates in clade 1 from this study. Genes with a population structure adjusted P-value (LRT P-value) of <0.001 as produced by pyseer. (c) Phylogenetic tree of >1,400 C. difficile isolates from NCBI (Table S3) depicting the presence of binary toxin and PaLoc operons. (d) Presence of full-length cdtR and association with tcdB presence. (e) Filtered results (P-values < 0.01), pyseer analysis evaluating gene association with CDI suspicion in clade 1 isolates using the LRT P-value. Purple color indicates P < 0.001. Positive beta coefficient indicates gene association with CDI suspicion, while negative beta indicates asymptomatic colonization.