Background: Fecal microbiota transplantation (FMT) is an effective treatment for recurrent Clostridium difficile infection (rCDI). The use of freeze-dried, encapsulated donor material for FMT (cap-FMT) allows for an easy route of administration and remains clinically effective in the majority of rCDI patients. We hypothesized that specific shifts in the microbiota in response to cap-FMT could predict clinical outcome. We further evaluated the degree of donor microbiota engraftment to determine the extent that donor transfer contributed to recovery.
Results: In total, 89 patients were treated with 100 separate cap-FMTs, with a success rate (no rCDI 60 days post cap-FMT) of 80%. Among responders, the lower alpha diversity (ANOVA P < 0.05) observed among patient's pre-FMT samples was restored following cap-FMT. At 1 week post-FMT, community composition varied by clinical outcome (ANOSIM P < 0.001), with similar abundances among families (Lachnospiraceae, Ruminococcaceae, and Bacteroidaceae) in responder and donor samples. Families that showed differential abundances by outcome (response vs. recurrence) from samples collected 7 days following cap-FMT were used to construct a regression tree-based model to predict recurrence. Results showed a training accuracy of 100% to predict recurrence and the model was 97% accurate against a test data set of samples collected 8-20 days following cap-FMT. Evaluation of the extent of engraftment using the Bayesian algorithm SourceTracker revealed that approximately 50% of the post-FMT communities of responders were attributable to donor microbiota, while an additional 20-30% of the communities were similar to a composite healthy microbiota consisting of all donor samples.
Conclusions: Regression tree-based analyses of microbial communities identified taxa significantly related to clinical response after 7 days, which can be targeted to improve microbial therapeutics. Furthermore, reinstatement of a healthy assemblage following cap-FMT was only partially attributable to explicit donor engraftment and continued to develop towards an overall healthy assemblage, independent of donor.
Keywords: Clostridium difficile; Encapsulated microbiota; Fecal microbiota transplantation; Machine learning; Microbial community structure; Prediction model.