Long-Term Bacterial and Fungal Dynamics following Oral Lyophilized Fecal Microbiota Transplantation in Clostridioides difficile Infection

Abstract

Oral lyophilized fecal microbiota transplantation (FMT) is effective in recurrent Clostridioides difficile infection (CDI); however, limited data exist on its efficacy in primary CDI and long-term microbial engraftment. Patients with primary or recurrent CDI were prospectively enrolled to receive oral FMT. Changes in the bacterial and fungal communities were characterized prior to and up to 6 months following treatment. A total of 37 patients with CDI (15 primary, 22 recurrent) were treated with 6 capsules each containing 0.35-g lyophilized stool extract. A total of 33 patients (89%) had sustained CDI cure, of whom 3 required a second course. There were no safety signals identified. FMT significantly increased bacterial diversity and shifted composition toward donor profiles in responders but not in nonresponders, with robust donor contribution observed to 6 months following FMT (P < 0.001). Responders showed consistent decreases in Enterobacteriaceae and increases in Faecalibacterium sp. to levels seen in donors. Mycobiome profiling revealed an association with FMT failure and increases in one Penicillium taxon, as well as coexclusion relationships between Candida sp. and bacterial taxa enriched in both donors and responders. Primary CDI was associated with more robust changes in the bacterial community than those with recurrent disease. Oral FMT leads to durable microbial engraftment in patients with primary and recurrent CDI, with several microbial taxa being associated with therapy outcome. Novel coexclusion relationships between bacterial and fungal species support the clinical relevance of transkingdom dynamics.IMPORTANCE Clostridioides difficile infection (CDI) is a substantial health concern worldwide, complicated by patterns of increasing antibiotic resistance that may impact primary treatment. Orally administered fecal microbiota transplantation (FMT) is efficacious in the management of recurrent CDI, with specific bacterial species known to influence clinical outcomes. To date, little is known about the efficacy of FMT in primary CDI and the impact of the mycobiome on therapeutic outcomes. We performed matched bacterial and fungal sequencing on longitudinal samples from a cohort of patients treated with oral FMT for primary and recurrent CDI. We validated many bacterial signatures following oral therapy, confirmed engraftment of donor microbiome out to 6 months following therapy, and demonstrated coexclusion relationships between Candida albicans and two bacterial species in the gut microbiota, which has potential significance beyond CDI, including in the control of gut colonization by this fungal species.

Keywords: Clostridioides difficile infection; fecal microbiota transplantation; microbiome; mycobiome.

Copyright © 2021 Haifer et al.

Figures

FIG 1

Changes to the bacterial communities. Both primary and recurrent CDI are included. Two patients did not provide baseline samples. (A) Shannon’s diversity (H´) indices in donors and responders to FMT. Significance was tested using ANOVA with Tukey’s multiple-comparison test, and P-W0 was found to be statistically significantly different from all other groups. No other comparisons were significant. (B) Shannon’s diversity (H´) index in donors and nonresponders to FMT. Patient samples at recurrence were labeled in black and with patient number. P34 had persistent disease. Significance was tested using ANOVA with Tukey’s multiple-comparison test, and P-W4 was found to be significantly different from donors B and C. No other comparisons were significant. (C) Principal-coordinate analysis of responders to FMT and donors. Bray-Curtis resemblance matrix was generated from square-root-transformed relative abundances of bacterial OTUs. All patient subgroups (P-) were significantly different from the donors (DON) when tested using pairwise PERMANOVA (P < 0.005 for all). P-W0 was significantly different from all other patient subgroups (P < 0.006 for all). No other comparisons were significant. ANOSIM confirmed the pairwise PERMANOVA results. (D) Principal-coordinate analysis of nonresponders to FMT and donors. Bray-Curtis resemblance matrix was generated from square-root-transformed relative abundances of bacterial OTUs. Dotted lines indicate samples corresponding to the same patient unless otherwise indicated. All patient subgroups (P-) were significantly different from the donors (DON) when tested using pairwise PERMANOVA (P < 0.004 for all). No other comparisons were significant. ANOSIM confirmed the pairwise PERMANOVA results. (E) Heatmap of mean relative abundance of bacterial OTUs found to be consistently significantly different between responders’ baseline and all post-FMT samples as well as responders’ baseline and donor samples. OTUs were not found to be significantly different in nonresponders but were included for comparison. Tests were performed using LEfSe, and a strict cutoff LDA score of >4 and P value of <0.05 were applied. (F) Heatmap of mean relative abundance of bacterial OTUs found to be consistently significantly different between nonresponders’ baseline and all post-FMT samples as well as nonresponders’ baseline and donor samples. Tests performed using LEfSe and a cutoff LDA score of >3.5 and P value of <0.05 were applied.

FIG 2

Mycobiome diversity and composition. (A) Relative abundance of Penicillium OTU14 which was significantly different between responders and nonresponders at baseline. Testing was performed using LEfSe (LDA score, 3.68; P < 0.05). (B) Species richness (d) at baseline (P-W0) and week 1 of FMT (P-W1) for responders (-Resp) and nonresponders (-Non). Significance was tested using Welch’s t test for each of the two comparisons reported. (C) Coexclusion relationships between Candida OTU2 with similarity to Candida albicans and two bacterial taxa (OTU18 and OTU34). Nonparametric relationships were identified using the MINE framework. OTU18 was classified to Dorea, and OTU34 was classified to Clostridium XVIII. (D) Heatmaps of mean relative abundances of bacterial OTU18 and OTU34 across responders’ baseline and post-FMT samples as well as donor samples.

FIG 3

Donor contribution to the patient bacterial component of the microbiome. (A) Contribution was determined using SourceTracker with donor samples assigned as specific sources (one-to-one) to their matched patient samples (sinks). Significance was tested using ANOVA with Tukey’s multiple-comparison tests. Donor contribution to the baseline sample was significantly lower than the post-FMT samples with the exception of P-W12. (B) Donor contribution was stratified according to response (-Resp) or lack of response (-Non) to FMT. Contribution was determined using SourceTracker with donor samples assigned as specific sources to their matched patient samples (sinks). Patient samples at recurrence were labeled in black and with patient number. Significance was tested using ANOVA with Tukey’s multiple-comparison tests. Only baseline samples of responders (P-W0-Resp) were significantly different from other groups (denoted by asterisks above groups).*, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001.

FIG 4

Bacterial communities in donor samples. (A) Principal-coordinate analysis of Bray-Curtis resemblance matrix generated from square-root-transformed relative abundances of bacterial OTUs. All donors (A, B, C, and D) were found to be significantly different from each other using pairwise PERMANOVA (P < 0.024 for all) except for donors B and C (PERMANOVA: t = 1.46, P = 0.054; ANOSIM: r = 0.288, P = 0.087). (B) Heatmap of mean relative abundance of bacterial OTUs found to be significantly different between donor samples that led to therapy success and those that led to therapy failure. Tests were performed using LEfSe and a cutoff LDA score of >3.5 and P value of <0.05 were applied.

FIG 5

Differences in the bacterial communities between types of C. difficile infection. Only responders were included in this analysis. Severe C. difficile infections or those with persisting disease despite antibiotics were excluded due to low numbers leaving primary (-P) and recurrent (-R) infections. (A) Shannon’s diversity (H´) index across different sample groups. Significance was tested using ANOVA with Tukey’s multiple-comparison test, and only P-W0-P was found to be statistically significantly different from other post-FMT groups. (B) Principal-coordinate analysis of Bray-Curtis resemblance matrix generated from square-root-transformed relative abundances of bacterial OTUs. All patient subgroups (P-) were significantly different from the donors (DON) when tested using pairwise PERMANOVA (P < 0.046 for all). P-W0-P was consistently significantly different from all other post-FMT sample groups in primary CDI (P < 0.015 for all). This result could not be replicated in the patients with recurrent CDI. (C) Heatmap of mean relative abundance of bacterial OTUs found to be significantly different between patients with primary and recurrent CDI at baseline. Tests were performed using LEfSe, and a strict cutoff LDA score of >4 and P value of <0.05 were applied.