To Lump or To Split: Does Strain Lineage for Clostridioides difficile Matter?

Since 2001, numerous descriptive ecological studies of Clostridioides difficile infections (CDI) have identified a single lineage (BI/NAP1/027) associated with the epidemics of CDI, increased severity of CDI, and increased likelihood of incident CDI to become recurrent. Establishing causality between the clinical severity and outcomes for CDI and the lineages of the infecting strains, however, has proved elusive, with many conflicting results in previous observational studies.

ABSTRACT

Since 2001, numerous descriptive ecological studies of Clostridioides difficile infections (CDI) have identified a single lineage (BI/NAP1/027) associated with the epidemics of CDI, increased severity of CDI, and increased likelihood of incident CDI to become recurrent. Establishing causality between the clinical severity and outcomes for CDI and the lineages of the infecting strains, however, has proved elusive, with many conflicting results in previous observational studies. In this issue of the Journal of Clinical Microbiology, J. R. Garneau, C. N. Abou Chakra, L.-C. Fortier, A.-C. Labbé, et al. (J Clin Microbiol 57:e01724-18, 2019, https://doi.org/10.1128/JCM.01724-18) performed multilocus variable-number tandem-repeat analysis (MLVA) on 450 isolates from epidemic strain CDI arising in 10 Canadian centers during a previously well-described epidemic to assess the hypothesis that subpopulations of this lineage are associated with adverse clinical outcomes. The authors’ key finding, however, was that MLVA genotyping grouped infections closely with associated hospital centers; CDI severity was not associated with any particular sublineage by MLVA. While the study does not support any causal inferences about strain-specific virulence of CDI, it does highlight the power of MLVA, a genotyping tool that remains valuable in tracking the geospatial transmission dynamics of CDI.

COMMENTARY

The emergence of Clostridioides difficile epidemics after 2000 was associated with lineages of the organism which had previously been relatively rare (1, 2). In time, numerous ecological descriptions of this “epidemic strain,” named NAP1 by pulsed-field gel electrophoresis (PFGE), 027 by PCR ribotyping, and BI by restriction endonuclease typing and usually referred to as NAP1/BI/027, described this relationship as likely causal (3). This was based in part on the observation that this lineage of C. difficile made more toxin in vitro, had missense and nonsense mutations in the repressor gene in the organism’s pathogenicity locus, and had apparently increased sporulation capacity (4–6).

Establishing a causal link between the “epidemic strain” of C. difficile and infections with clinically distinct outcomes at the level of individual patients has yielded inconsistent results in observational studies. In a Canadian study of 1,005 PFGE-typed C. difficile infection (CDI) cases in 2004 and 2005, NAP1 strains caused 30.8% of cases; NAP1 cases had an observed 12.5% severe outcome compared to 5.9% for patients with non-NAP1 strain infections (P < 0.001) (7). This outcome difference was not consistent across all age deciles, however, and a multivariable analysis for severe clinical infection was not performed (7). In a population-based surveillance study of 2,057 CDI cases during 2009 to 2011 from 6 U.S. states with available strain typing data by PFGE, death, severe outcomes attributable to CDI, and severe clinical CDI occurred in 2.7%, 4.9%, and 17.7% of cases, respectively (8). NAP1 strains (28.4% of all cases) were associated with death, severe outcomes, and severe clinical CDI with adjusted odds ratios of 2.12, 1.66, and 1.74, respectively, after adjusting for other confounding risks (8).

Several studies have failed to replicate any association between the NAP1/BI/027 lineage of C. difficile and either severe clinical CDI or severe CDI-attributable outcomes, however. In a single-center cross-sectional study of 743 CDI patients in Michigan, ribotype 027 was associated with severe CDI in an unadjusted analysis, but this ribotype was not associated with severe CDI when considering 11 other covariates; only leukocytosis and serum albumin level were associated with severe CDI in an adjusted logistic regression model (9). In a similar study of 715 CDI patients in 7 hospitals in Houston, TX, ribotype 027 strain CDI (23.8% of total cases) was an independent predictor of both severe CDI and severe outcome CDI (adjusted odds ratios of 2.24 and 1.71, respectively) when considering a comparison to all other CDI case ribotypes in aggregate (10). No logistic regression model was able to establish any single ribotype as a significant predictor for severe CDI or outcome; in fact, 3 other non-027 ribotypes had point estimates of severe CDI outcome as high as or higher than ribotype 027 in this study (10). A European study noted that ribotype 078 CDI had higher observed mortality than other lineages, including ribotype 027 (11). A case-control study of CDI during 2011 to 2013 in three acute care hospitals in Hong Kong identified no cases of ribotype 027, whereas ribotype 002 patients were observed to have decreased 30-day survival compared to those with other ribotypes. Other observational studies have also failed to observe any consistent relationship between ribotype 027 CDI and severe clinical outcome (9, 12–15).

In an epilogue to the original description of ribotype 027 as hypervirulent, Carlson and colleagues confirmed that isolates of ribotype 027 express higher levels of toxin production than isolates of other ribotype lineages, but this feature did not correlate with the severity of clinical disease, and ribotype 027 strains from this single-center study did not produce more spores than isolates of other ribotypes (16). Enhanced toxin production originally observed for epidemic lineage strains was observed to vary widely across strains of ribotype 027 itself, casting additional doubt on the value of ribotype lineage in assessing the contribution to CDI virulence (17). Taken together, the observed epidemiology and in vitro phenotypic variation of ribotype 027 strain virulence have raised the possibility that the apparent association between ribotype 027 and the changing worldwide epidemiology of CDI represent an ecological fallacy, i.e., the association at the population level does not remain when examined at the level of individual CDI cases.

Garneau and colleagues now present a study of CDI from 10 Canadian hospitals during 2005 to 2008, the purpose of which was to investigate whether the additional strain type discrimination of multilocus variable-number tandem-repeat analysis (MLVA) could refine the relationship between C. difficile strain and various clinical outcomes (18). MLVA is an established genotyping tool that provides fine discrimination of most ribotypes and PFGE pulsovars into hundreds of genotypes defined by concatenating the tandem repeats at 4 to 6 loci in the C. difficile genome (19, 20). Despite the inherent variability of the tandem repeat counts, which change at various rates by locus during slippage and mispairing events during DNA replication, groups of C. difficile strains can be arranged by minimizing the Manhattan distance of the summed tandem repeat difference (STRD) linking them in an unrooted minimum spanning tree. MLVA is thought to perform best with sets of isolates from defined regions of likely related strains, and it has been used to track transmission within single centers, since the tandem repeat loci appear to be stable within patients over short defined time periods (21, 22).

In the authors’ previous study of the same Canadian cohort of 922 ribotyped CDI patients, 52% were infected with ribotype 027 and 8% of patients experienced the outcome of interest, complicated CDI as defined by clinical events attributable to CDI (23). As in many previous studies, a higher proportion of ribotype 027 patients was observed to have complicated CDI (10.9% versus 7.2%, P = 0.008), but this association did not remain significant after adjusting for other variables (23). In the present study, ribotype 027 CDI in this cohort underwent MLVA genotyping and grouping into clusters defined using summed tandem repeat differences of ≤2, such that 450 ribotype 027 CDI infections were discriminated into 371 MLVA types grouped into 14 clonal complexes (clusters) of C. difficile isolates; 95 strains were singleton MLVA genotypes that did not have any relationship by STRD to other strains. No MLVA cluster of ribotype 027 CDI was observed to have a differential rate of severe CDI or recurrent CDI in addition to the complicated CDI outcomes assessed in their prior study. As was the case in their prior study, a multivariate model did not alter this result (18).

While this result might be taken as simply another negative study assessing clinical outcome and C. difficile strain type, the study did reveal a striking correspondence of MLVA genotype complexes not only with the hospital of origin but also with a largely correct spatial clustering of hospitals within Canada on the minimum spanning tree (MST) (Fig. 1 in reference 18). European reference ribotype 027 isolates were positioned at the periphery of the same MST, corresponding with their epidemiologic and geospatial unrelatedness to isolates from this Canadian study cohort. The correct spatial clustering of the MLVA minimum spanning tree using the authors’ isolate collection is particularly surprising given that the study isolates originated from a wide geographic area over a 4-year period; the assumption for any minimum spanning tree built to depict the relationships between MLVA genotypes is that there are a minimum of “missing” genotypes, i.e., genotypes with all potential intermediate states of evolution within tandem repeat regions are represented such that the MLVA accurately groups isolates by STRD. In this study, there were several sources of “missing links” in MLVA strain evolution in the cohort: (i) 7% of ribotype 027 isolates did not undergo MLVA, (ii) at least 60 isolates were not obtained from toxin-positive stools in study centers, (iii) an unknown number of patients diagnosed at each participating center did not provide informed consent, (iv) patients <18 years of age were excluded in all centers, and (v) patients with asymptomatic carriage of C. difficile were not included; such patients with asymptomatic carriage far outnumber patients with CDI in prior ward and hospital-based studies (21, 24, 25). Despite this limitation, the geospatial discrimination of MLVA remained intact, perhaps reflecting the relative stability of the C. difficile genome over space and time due to the long periods that the organism exists in a stable sporulated form.

While associations between epidemic lineage by MLVA and restriction endonuclease typing and recurrence of CDI have been seen in previous settings, such an association was not made within sublineages of isolates within ribotype 027 in this study (26, 27). Due to the small sample size, studies such as this one cannot definitively exclude any association between particular C. difficile lineages with severity of clinical CDI or CDI-attributable severe outcome, but the observed data support the hypothesis that any excess risk conferred by infecting lineage—if present—is not consistently clinically significant, and there remains no proved value to real-time identification of any single lineage of C. difficile for clinicians. As a genotyping tool, however, MLVA appears to have surprising value for tracking transmission patterns for C. difficile isolates collected widely over space and time, and for reasons of cost, throughput, reproducibility, portability, and remarkable discriminatory power, it is likely to remain the favored genotyping tool for CDI molecular epidemiology even in the era of whole-genome sequencing, at least until the latter method can reliably identify C. difficile virulence factors that are clearly associated with clinical adverse outcomes in CDI. Until that time, evidence to assign any particular lineage or pathogen-specific virulence determinant as causally linked to the changing severity, lethality, and recurrence propensity of CDI remains unproved.

The views expressed in this article do not necessarily reflect the views of the journal or of ASM.