Dynamics of the fecal microbiome in patients with recurrent and nonrecurrent Clostridium difficile infection

Abstract

Background: Recurrent Clostridium difficile infection (CDI) remains problematic, with up to 30 % of individuals diagnosed with primary CDI experiencing at least one episode of recurrence. The success of microbial-based therapeutics, such as fecal microbiota transplantation, for the treatment of recurrent CDI underscores the importance of restoring the microbiota. However, few studies have looked at the microbial factors that contribute to the development of recurrent disease. Here we compare microbial changes over time in patients with or without recurrence to identify microbial signatures associated with the development of recurrence.

Methods: We used 16S rRNA-encoding gene sequence analysis to compare the fecal microbiota of 93 patients with recurrent and nonrecurrent CDI, sampled longitudinally. Cross-group and intra-individual differences in microbial community diversity and similarity were compared prior to the development of recurrent disease and over time.

Results: Samples from these patient groups exhibited variable community profiles, clustering into four distinct community groups. Cross-group comparison of the index sample collected from patients that did or did not develop recurrence revealed differences in diversity and community structure (analysis of molecular variance, p < 0.05). Intra-individual comparisons of the microbiota were more informative and samples from recurrent patients were less likely to recover in diversity (Chi-square test, p < 0.005), exhibiting less community similarity overall (Kruskal-Wallis test, p < 0.05). Interestingly, patients with severe disease harbored a significantly less diverse community, a trend that was observed across both nonrecurrent and recurrent patient groups (Wilcoxon test, p < 0.05).

Conclusions: To date, this study represents one of the largest studies focused on the relationship between predictive signals from the gut microbiota and the development of recurrent CDI. Our data demonstrate that specific microbiota-derived characteristics associate with disease severity and recurrence and that future studies could incorporate these characteristics into predictive models.

Fig. 1

Study design and sample collection timeline. Relative timeline (days) of samples collected from patients diagnosed with initial Clostridium difficile infection (CDI) (index sample = 0 days) categorized into three patient groups (nonrecurrent, recurrent, and reinfected). Patients who did not develop recurrence (n = 42) remained free of a subsequent CDI diagnosis. Patients with recurrent disease (n = 32) were diagnosed with CDI (positive clinical lab result) 14–56 days following index sample collection. Patients diagnosed with another CDI index outside of the recurrence window (>56 days) were considered reinfected (n = 19) NA = test not available

Fig. 2

Samples clustered into four major community profiles. The relative abundance of the 40 most abundant operational taxonomic units (OTUs), with classification to the genus level and organized by bacterial phylum, is shown in columns. Samples were binned into four major clusters using the Partitioning Around Medoids (PAM) algorithm based on the Jensen–Shannon divergence. The mean inverse Simpson index (λ) per cluster is shown on the left axis (samples). Sample categorization on the left axis is based on the following classifications: patient group category (nonrecurrent, recurrent, or reinfected); clinical lab results (Quik Chek, positive or negative); C. difficile cultivation results (positive or negative); and disease severity (severe or non-severe) at sample collection during a CDI diagnosis NA = text result not available

Fig. 3

Differentially abundant members of the microbiota in patients with C. difficile infection. The mean relative abundance plus standard error (se) of differentially abundant operational taxonomic units (OTUs) identified by linear discriminant analysis (LDA) effect size (LEfSe) in (a) samples that tested positive or negative for C. difficile by the clinical laboratory (Quik Chek) or (b) severe or non-severe samples. OTUs that were overrepresented in the specified groups are color-coded by the respective group in each panel

Fig. 4

Fecal microbial diversity during initial C. difficile infection. The inverse Simpson index (λ) of the microbiota in (a) index samples collected at initial C. difficile infection (CDI) diagnosis in nonrecurrent (n = 42), recurrent (n = 32), and reinfected (n = 19) patients (Kruskal–Wallis, not significant (ns)) and (b) index samples from patients diagnosed with severe (n = 36) or non-severe (n = 50) CDI (Wilcoxon test, p = 0.022)

Fig. 5

Community structure of patients with or without recurrent C. difficile infection. Principal coordinates analysis (PCoA) was used to plot the Yue and Clayton dissimilarity index (θ _YC). a The community structure of the microbiota in samples from nonrecurrent, recurrent, and reinfected patients (analysis of molecular variance (AMOVA), p = 0.016). b The community structure of samples positive or negative for C. difficile as determined by the clinical lab using Quik Chek (AMOVA, p = 0.015)

Fig. 6

Intra-individual similarity of the microbiota in patients with or without recurrent C. difficile infection. The microbial community similarity within patients was compared using the Yue and Clayton dissimilarity index (θ _YC). a Intra-individual similarity was lower in patients with recurrence compared with patients without recurrence or reinfected with C. difficile (Kruskal–Wallis test, p = 0.025). b Microbial community similarity of the index sample from a patient was compared with different stages of clinical diagnosis in nonrecurrent, recurrent, and reinfected patients: to recovery (nonrecurrent, non-reinfected samples >14 days of a positive sample), to recurrence (subsequent positive sample within 14–56 days), reinfection (subsequent positive sample >56 days), and during treatment (sample collected within 14 days of a positive sample) (Kruskal–Wallis, not significant). ns not significant