A robust high-throughput multilocus sequence typing (MLST) scheme for Clostridium difficile was developed and validated using a diverse collection of 50 reference isolates representing 45 different PCR ribotypes and 102 isolates from recent clinical samples. A total of 49 PCR ribotypes were represented overall. All isolates were typed by MLST and yielded 40 sequence types (STs). A web-accessible database was set up (http://pubmlst .org/cdifficile/) to facilitate the dissemination and comparison of C. difficile MLST genotyping data among laboratories. MLST and PCR ribotyping were similar in discriminatory abilities, having indices of discrimination of 0.90 and 0.92, respectively. Some STs corresponded to a single PCR ribotype (32/40), other STs corresponded to multiple PCR ribotypes (8/40), and, conversely, the PCR ribotype was not always predictive of the ST. The total number of variable nucleotide sites in the concatenated MLST sequences was 103/3,501 (2.9%). Concatenated MLST sequences were used to construct a neighbor-joining tree which identified four phylogenetic groups of STs and one outlier (ST-11; PCR ribotype 078). These groups apparently correlate with clades identified previously by comparative genomics. The MLST scheme was sufficiently robust to allow direct genotyping of C. difficile in total stool DNA extracts without isolate culture. The direct (nonculture) MLST approach may prove useful as a rapid genotyping method, potentially benefiting individual patients and informing hospital infection control.

Clostridium difficile infection (CDI) is an important cause of mortality and morbidity in healthcare settings. The major virulence determinants are large clostridial toxins, toxin A (tcdA) and toxin B (tcdB), encoded within the pathogenicity locus (PaLoc). Isolates vary in pathogenicity from hypervirulent PCR-ribotypes 027 and 078 with high mortality, to benign non-toxigenic strains carried asymptomatically. The relative pathogenicity of most toxigenic genotypes is still unclear, but may be influenced by PaLoc genetic variant. This is the largest study of C. difficile molecular epidemiology performed to date, in which a representative collection of recent isolates (n = 1290) from patients with CDI in Oxfordshire, UK, was genotyped by multilocus sequence typing. The population structure was described using NeighborNet and ClonalFrame. Sequence variation within toxin B (tcdB) and its negative regulator (tcdC), was mapped onto the population structure. The 69 Sequence Types (ST) showed evidence for homologous recombination with an effect on genetic diversification four times lower than mutation. Five previously recognised genetic groups or clades persisted, designated 1 to 5, each having a strikingly congruent association with tcdB and tcdC variants. Hypervirulent ST-11 (078) was the only member of clade 5, which was divergent from the other four clades within the MLST loci. However, it was closely related to the other clades within the tcdB and tcdC loci. ST-11 (078) may represent a divergent formerly non-toxigenic strain that acquired the PaLoc (at least) by genetic recombination. This study focused on human clinical isolates collected from a single geographic location, to achieve a uniquely high density of sampling. It sets a baseline of MLST data for future comparative studies investigating genotype virulence potential (using clinical severity data for these isolates), possible reservoirs of human CDI, and the evolutionary origins of hypervirulent strains.