Defining and Evaluating a Core Genome Multilocus Sequence Typing Scheme for Genome-Wide Typing of Clostridium difficile

Abstract

Clostridium difficile, recently renamed Clostridioides difficile, is the most common cause of antibiotic-associated nosocomial gastrointestinal infections worldwide. To differentiate endogenous infections and transmission events, highly discriminatory subtyping is necessary. Today, methods based on whole-genome sequencing data are increasingly used to subtype bacterial pathogens; however, frequently a standardized methodology and typing nomenclature are missing. Here we report a core genome multilocus sequence typing (cgMLST) approach developed for C. difficile Initially, we determined the breadth of the C. difficile population based on all available MLST sequence types with Bayesian inference (BAPS). The resulting BAPS partitions were used in combination with C. difficile clade information to select representative isolates that were subsequently used to define cgMLST target genes. Finally, we evaluated the novel cgMLST scheme with genomes from 3,025 isolates. BAPS grouping (n = 6 groups) together with the clade information led to a total of 11 representative isolates that were included for cgMLST definition and resulted in 2,270 cgMLST genes that were present in all isolates. Overall, 2,184 to 2,268 cgMLST targets were detected in the genome sequences of 70 outbreak-associated and reference strains, and on average 99.3% cgMLST targets (1,116 to 2,270 targets) were present in 2,954 genomes downloaded from the NCBI database, underlining the representativeness of the cgMLST scheme. Moreover, reanalyzing different cluster scenarios with cgMLST were concordant to published single nucleotide variant analyses. In conclusion, the novel cgMLST is representative for the whole C. difficile population, is highly discriminatory in outbreak situations, and provides a unique nomenclature facilitating interlaboratory exchange.

Keywords: Clostridium difficile; cgMLST; typing; whole-genome sequencing.

FIG 1

Neighbor-joining tree of the 11 C. difficile isolates used for cgMLST target definition based on cgMLST target genes with pairwise ignore missing values. In addition to the sample name, the clade is given and the BAPS partitions are colored. The distance is given as the number of cgMLST genes.

FIG 2

Minimum-spanning tree of two spatiotemporal clusters (9). Each node represents a unique cgMLST allele profile. The numbers on connecting lines display the number of differing alleles between the genotypes (line length not to scale). The different nodes are colored by the MLST ST, and closely related genotypes (≤6 different cgMLST alleles) are shaded. (A) Short-term cluster of four cases, where one transmission event was epidemiologically confirmed (2,244 to 2,265 cgMLST target genes [mean, 99.5%] were analyzed). (B) Short-term cluster of four cases, where a clonal transmission was ruled out (2,237 to 2,265 cgMLST target genes [mean, 99.1%] were analyzed).

FIG 3

Minimum-spanning tree and epidemiological curve illustrating a long-term spatiotemporal C. difficile cluster with two identified peaks (15). The 22 cluster isolates are colored according to their peaks, and the two reference strains are marked in green. (A) Minimum-spanning tree of the reanalyzed sequences based on cgMLST targets. Each node represents a unique allelic profile, and the size of the nodes represents the number of isolates. The numbers on connecting lines are the numbers of differing alleles between the genotypes (not to scale), and closely related genotypes (≤6 different cgMLST alleles) are shaded; 2,243 to 2,267 cgMLST target genes (mean of 99.6% of all cgMLST targets) were analyzed. All isolates of peaks 1 and 2 belonged to ST1. (B) Epidemiological curve. Each box represents one isolate, and boxes are colored according to their peak affiliation.