Host origin of microbiota drives functional recovery and Clostridioides difficile clearance in mice

Abstract

Colonization resistance provided by the gut microbiota is essential for resisting both initial Clostridioides difficile infection (CDI) and potential recurrent infection (rCDI). Although fecal microbiota transplantation (FMT) has been successful in treating rCDI by restoring microbial composition and function, mechanisms underlying the efficacy of standardized stool-derived products remain poorly understood. Using a combination of 16S rRNA gene-based and metagenomic sequencing alongside metabolomics, we investigated microbiome recovery following FMT from human and murine donor sources in a mouse model of rCDI. We found that a human-derived microbiota was less effective in clearing C. difficile compared to a mouse-derived microbiota, despite recovery of taxonomic diversity, compositional changes, and bacterial functions typically associated with clearance. Metabolomic analysis revealed deficits in secondary metabolites compared to those that received murine FMT, suggesting a functional remodeling between human microbes in their new host environment. Collectively, our data revealed additional environmental, ecological, or host factors to consider in FMT-based recovery from rCDI.

Importance: Clostridioides difficile is a significant healthcare-associated pathogen, with recurrent infections presenting a major treatment challenge due to further disruption of the microbiota after antibiotic administration. Despite the success of fecal microbiota transplantation (FMT) for the treatment of recurrent infection, the mechanisms mediating its efficacy remain underexplored. This study reveals that the effectiveness of FMT may be compromised by a mismatch between donor microbes and the recipient environment, leading to deficits in key microbial metabolites. These findings highlight additional factors to consider when assessing the efficacy of microbial-based therapeutics for C. difficile infection (CDI) and other conditions.

Keywords: Clostridioides difficile; fecal microbiota transplant; gut microbiome; metabolomics; metagenomics; recurrence.

Fig 1

Human-derived fecal material fails to clear C. difficile in mice with recurrent CDI. (A) Mouse model of recurrent CDI. (B) Percent weight loss (compared to day 0, day of C. difficile infection) over time in mice treated with feces from healthy humans (hFMT; n = 53), colony-matched mice (mFMT; n = 33), mice representative of different genetic or colony backgrounds (mFMT-other; n = 23), or untreated mice (noFMT; n = 24). C. difficile colonization, as assessed using the log₁₀-normalized CFU, was identified in (C) fecal samples over time or (D) cecal samples at day 21 or day 42 post-infection in mice treated with hFMT, mFMT, hFMT, or untreated (noFMT) mice. Statistical significance determined using the Kruskal-Wallis test, with a post hoc Dunn test for panel D (*P < 0.01, **P < 0.001, ***P < 0.0001).

Fig 2

Human-derived fecal bacteria engraft mice with CDI. (A) NMDS of the Bray-Curtis dissimilarity index calculated from OTUs throughout the CDI timeline. Fecal samples throughout the experiment or endpoint cecal samples were collected throughout the experiment (pre-FMT) without treatment (noFMT) and following treatment with various mouse (mFMT1–4) or human (hFMT1–6) feces, with sample designation (including FMT inputs) as indicated in the legend. PERMANOVA, P < 0.001 calculated across groups (mFMT, hFMT, and noFMT) for both cecal and fecal timepoints. (B) Pairwise Bray-Curtis dissimilarity was calculated between treatment groups (mFMT, mFMT-other, hFMT, and noFMT) and their respective inputs, or between and within treatment groups. (C) Average relative abundance of top 98% genera observed in cecal samples prior to any treatment (pre-abx) or at indicated timepoints without treatment (noFMT) or following mFMT, mFMT-other, or hFMT. Shannon diversity index calculated from (C) fecal samples over time or (D) cecal samples at day 21 or day 42 post-infection in mice treated with hFMT, mFMT, hFMT, or untreated (noFMT) mice. (F) Differentially abundant OTUs between mice that did (mFMT, mFMT-other) versus did not clear (hFMT, noFMT) C. difficile (MaAsLin2, linear model with Benjamini-Hochberg (BH) correction, q ≤ 0.001) with the log₁₀-normalized relative abundance (RA+1) per OTU per treatment group. Statistical significance for panels B, D, and E was determined using the Kruskal-Wallis test, with a post hoc Dunn test for panels B and E (*P < 0.01, **P < 0.001, ***P < 0.0001).

Fig 3

FMT source determines species composition post-transplantation, regardless of C. difficile clearance. (A) NMDS of Bray-Curtis dissimilarity (stress = 0.1390) and (B) Bray-Curtis distances based on bacterial species identified in fecal and cecal samples from mice. Asterisks denote significance (****P < 0.0001) determined using the Kruskal-Wallis test, with a post hoc Dunn test. (C) Inverse Simpson diversity of bacterial species identified in cecal samples. Asterisks denote significance (**P < 0.05) determined using the Kruskal-Wallis test, with a post hoc pairwise Wilcoxon rank-sum test. (D) Average relative abundance of bacterial species identified by MetaPhlAn4 from cecal samples of mice. (E) Mean log₁₀-transformed (CPM + 1) of named bacterial genera significantly different in abundance across treatment groups compared to mice treated with mFMT (based on MaAsLin2; linear model with BH correction, q ≤ 0.01).

Fig 4

Genes, pathways, and contributing taxa vary with FMT source, independent of C. difficile clearance. (A) NMDS of Bray-Curtis dissimilarity (stress = 0.1109) and (B) Bray-Curtis distances based on KOs identified in fecal and cecal samples from mice. Asterisks denote significance (**P < 0.05) determined using the Kruskal-Wallis test, with a post hoc Dunn test. (C) KOs identified to be significantly different in abundance between mice that did or did not clear C. difficile (based on MaAsLin2; linear model with BH correction, q ≤ 0.001). (D) Mean log₁₀-transformed (CPM + 1) of amine, amino acid, and carbohydrate metabolism MetaCyc Pathways significantly different across treatment groups compared to mice treated with mFMT (based on MaAsLin2; linear model with BH correction, q ≤ 0.001). (E–K) CPM of genes belonging to the labeled amino acid metabolism MetaCyc Pathway, colored by the genus that encodes it. (L–P) CPM of Uniref90 genes associated with bile salt hydrolase (Bsh) or terminal genes involved in butyrate production (buk, but, 4-Hbt, and Ato). Asterisks denote significance (**P < 0.01, ***P < 0.001, ****P < 0.0001) determined using the Kruskal-Wallis test, with a post hoc pairwise Wilcoxon rank-sum test.

Fig 5

Mice treated with human feces (hFMT) demonstrate differential metabolic recovery. (A) NMDS scaling of the Bray-Curtis dissimilarity index calculated from metabolite abundances from untargeted metabolomics in cecal samples of uninfected mice (healthy), mice with rCDI receiving no treatment (noFMT), or treated with human (hFMT1–3) or mouse feces (mFMT). PERMANOVA, P < 0.001 calculated across groups (healthy, mFMT, hFMT1–3 combined, and noFMT). (B) Pairwise Bray-Curtis dissimilarity was calculated between (healthy, mFMT, hFMT, and noFMT) and within groups. Statistical significance determined using the Kruskal-Wallis test, with a post hoc Dunn test (*P < 0.01, **P < 0.001, ***P < 0.0001). (C) Mean decrease in accuracy of the top 25 important metabolites increased or decreased post-FMT from untargeted metabolomics, as identified by random forest analysis. Mean log₁₀-normalized median-scaled and minimum-imputed abundance of each metabolite. Targeted analysis of (D) acetate, (E) propionate, and (F) butyrate fecal abundances over time in mice with rCDI treated with mFMT, hFMT, or noFMT. Targeted analysis of cecal (G) acetate, (H) propionate, and (I) butyrate abundance at day 21 or 42 post-infection in uninfected mice (healthy; gray) compared to mice with rCDI treated with mFMT, hFMT, or noFMT. (J) Total levels of primary and secondary bile acids from targeted metabolomics in the ceca of uninfected mice (healthy) compared to mice with rCDI treated with mFMT, hFMT, or noFMT. (K) Deoxycholic acid (DCA), (L) w-muricholic acid (MCA), (M) cholic acid (CA), (N) taurocholic acid (TCA), (O) α-MCA, and (P) β-MCA in the same mice. Statistical significance for (D–P) determined using Kruskal-Wallis test, with a post hoc Dunn test for panels G–I and K–P (*P < 0.05, **P < 0.005, ***P < 0.0005).