Identification of clinical and ecological determinants of strain engraftment after fecal microbiota transplantation using metagenomics

Abstract

Fecal microbiota transplantation (FMT) is a promising therapeutic approach for microbiota-associated pathologies, but our understanding of the post-FMT microbiome assembly process and its ecological and clinical determinants is incomplete. Here we perform a comprehensive fecal metagenome analysis of 14 FMT trials, involving five pathologies and >250 individuals, and determine the origins of strains in patients after FMT. Independently of the underlying clinical condition, conspecific coexistence of donor and recipient strains after FMT is uncommon and donor strain engraftment is strongly positively correlated with pre-FMT recipient microbiota dysbiosis. Donor strain engraftment was enhanced through antibiotic pretreatment and bowel lavage and dependent on donor and recipient ɑ-diversity; strains from relatively abundant species were more likely and from predicted oral, oxygen-tolerant, and gram-positive species less likely to engraft. We introduce a general mechanistic framework for post-FMT microbiome assembly in alignment with ecological theory, which can guide development of optimized, more targeted, and personalized FMT therapies.

Keywords: donor strain engraftment; dysbiosis; ecological theory; fecal microbiota transplantation; metagenomics; microbiota depletion; personalized FMT; post-FMT microbiome assembly; shared strain analysis.

Conflict of interest statement

Declaration of interests The authors declare no competing interests.

Copyright © 2022 The Author(s). Published by Elsevier Inc. All rights reserved.

Figures

Graphical abstract

Figure 1

Overview and taxonomic microbiota composition of the FMT meta-cohort (A) Overview of treatment modalities, number and distribution of cases, and samples for 13 studies (Table S1) from five conditions included in the FMT meta-cohort. (B) Taxonomic microbiota compositions based on principal-component analysis (PCA) of centered log-ratio-transformed relative species abundance. Samples in the main plot are categorized and color-coded as pre-FMTABx− (blue), antibiotically pretreated pre-FMTABx+ (purple), and post-FMT (yellow) patient and donor (red) samples. Ridgeline density plots show sample distributions along the two first principal components based on the same categories. In the small plots (right side) samples from the same PCA are color-coded based on (from top to bottom) scaled cumulative relative abundances of Bacteroidetes, Firmicutes, dysbiosis score, and ɑ-diversity (Shannon index). (C) Ridgeline density plots showing only pre-FMT (untreated and antibiotically pretreated) patient samples, colored by disease category. The MDR studies included patients with a history of antibiotic treatments, even before the beginning of the FMT trials.

Figure 2

Taxonomic and functional microbiota comparison of FMT recipients, with or without antibiotic treatment, post-FMT patients, and donors in the different studies Generalized linear mixed-effects models (GLMM, see STAR Methods) highlight differences in taxonomic and functional microbiota metrics between pre-FMTABx− (blue), pretreated pre-FMTABx+ (purple), Post-FMT (yellow), and donor (red) samples relative to a reference cohort of 739 healthy adults (gray line and area denote mean ± SD, respectively), based on the average distance to healthy control samples (β-diversity, Aitchison distance), the dysbiosis score, α-Diversity (Shannon index), and the cumulative relative abundance of oxygen-tolerant or oral bacterial species. Metadata abbreviations indicate pretreatment with ABx and lavage (+/−); single (+), two (++), or multiple (+++) FMTs; colonoscopic (C), nasogastric (NG), or gastroduodenal (GD) FMT route, enema (E), and pill/capsule (P) administration. Sample sizes (see STAR Methods for the reference cohort) are as shown in Figure 1A, excluding post-FMT samples from patients who have been exposed to antibiotics after FMT (OSU = 9, ZAR = 5, PAU = 4). Significant differences of the different sample types from healthy controls were determined for each metric and study separately with GLMMs. Asterisks denote significance thresholds: ˙p ≤ 0.1, ∗p ≤ 0.01, ∗∗p ≤ 0.001, ∗∗∗p ≤ 0.0001.

Figure 3

Strain profiling in FMT recipients showed substantial variation in donor strain engraftment between studies that are linked to FMT treatment modalities (A) Post-FMT microbiota relative abundance fractions contributed from donor (red), patient (blue), new (yellow), or coexisting (gray) strains, as detected in the last available post-FMT sample per patient. Darker colors refer to species-level relative abundance fractions if strains could not be resolved. Left: Circle chart showing average cumulative relative abundances of shared species in pre- and post-FMT patient and donor samples (gray) and between pre- and post-FMT patient (blue) and post-FMT patient and donor (red) samples. (B) Validation of SameStr’s specificity to infer donor strain engraftment from shared strain detection. Very few shared strains were identified between pre-FMT patient and donor (gray) and between unrelated post-FMT patient and donor (white) sample pairs, whereas strain sharing is frequent between pre- and post-FMT patient (blue) and between post-FMT patient and corresponding donor (red) sample pairs. The microbiota compositions of all sample pairs overlapped widely at higher taxonomic levels (family, genus, species). (C) Comparison of donor-derived (red), recipient-derived (blue), and coexisting (gray) strain fractions in post-FMT patient samples (of the sum of donor and recipient-derived strains) between disease groups and individual studies from the meta-cohort. Symbols denote the mean value of the latest available post-FMT sample per patient and across all cases of a study. Study metadata are shown as follows: Antibiotic (ABx) and bowel lavage patient pretreatment: Yes (+), No (−); Number of FMTs: single (+), two (++), or multiple (+++) rounds; FMT application: by colon (C), nasogastric (NG), or gastroduodenal (GD) endoscopy, enema (E), or pill/capsule (P). (D) Longitudinal comparison of donor-derived (red), patient-derived (blue), and coexisting (white) strain fractions in post-FMT samples from antibiotically pretreated (BOU) and non-pretreated (ZAR) ICI-refractory melanoma patients.

Figure 4

For individual patients, donor microbiota engraftment after FMT is dependent on patient and donor microbiota characteristics and clinical modalities of the FMT treatment (A) Forest plot showing the relevance of microbiota and clinical parameters for donor-derived strain fractions in post-FMT patients in the FMT meta-cohort, as determined with a generalized linear mixed model. The model is based on data from 254 clinical FMT cases, including samples from post-FMT patients (samples n = 801; FRI = 12, ALM = 24, ERE = 51, CHA = 105, BOR = 15, OSU = 112, ZAR = 146, BOU = 30, GOR = 58, KAa = 95, KAb = 63, XAV = 33, PAU = 13, HUT = 44), their respective donors (n = 140), and earliest available pre-FMT samples (n = 254). (B) Simulations with this model to determine the marginal effects of α-diversity on donor strain engraftment, i.e., using real values in combination with the minimum or maximum Shannon index detected in any donor in the FMT cohort (min/max within 95% confidence intervals), indicate a disproportionate impact of high-α-diversity donors on low-α-diversity FMT recipients. (C) Similar simulations predict independent marginal effects of ABx and lavage pretreatment on donor strain engraftment. Shaded areas and bars denote the 95% confidence intervals (Wald). Asterisks denote significance thresholds: ∗p ≤ 0.01, ∗∗p ≤ 0.001, ∗∗∗p ≤ 0.0001.

Figure 5

Donor strain engraftment probabilities for individual strains (A) GLMM-based estimated donor strain engraftment probabilities in relation to phylogeny and microbial species features (Gram-staining, spore formation, oxygen tolerance, oral habitat). (B) Median donor strain engraftment probabilities for different phyla (top) and genera (bottom), together with the estimated minimum and maximum probabilities, when using the lowest and highest species relative abundances that were detected in the meta-cohort as alternative input variables to the model. Crosses denote the donor strain engraftment probabilities without GLMM adjustments for the estimated influence of microbial, microbiota, and clinical FMT variables (see Figure 5C). (C) Species features (patient and donor), microbiota parameters, and clinical FMT variables with relevance for the engraftment of individual donor strains based on the GLMM. Bars denote 95% confidence intervals (Wald). Asterisks denote significance thresholds: ∗p ≤ 0.01, ∗∗p ≤ 0.001, ∗∗∗p ≤ 0.0001. (D) Total numbers of donor-derived strains in post-FMT patients, based on GLMM predictions for individual donor strains, vary substantially depending on recipient/donor pairing. Top left: They deviate ± 5-fold (log2) for the worst (blue) and best (yellow) simulated donor pair, relative to the actual recipient/donor pairs. Bottom left: Variations between different recipient/donor pairs for one donor (y axis range) generally exceed variations between different donors (x axis range). Black dots and bars: mean values ±SD. Top right, bottom middle, and right: Using patient 54C as an example, the predicted engraftment varies between 40 strains for different donors, but the worst (blue) and best (yellow) donors for this patient are predicted to also result in a broad range of engrafted donor strains in pairings with other patients. Dark lines indicate observations from actual recipient/donor pairings.

Figure 6

Donor microbiota engraftment and clinical response to FMT Responders (R, yellow) and non-responders (NR, blue) from two FMT trials to overcome resistance to anti-PD-1 therapy in ICI-refractory melanoma patients and two FMT trials to induce remission in IBD patients were compared based on donor-derived strain fractions (of recipient and donor-derived strains), cumulative relative abundances of species represented by donor-derived strains and total numbers of donor-derived strains in post-FMT samples (last available sample within ≤4 months after FMT). See STAR Methods for a description of R/NR numbers and definitions. IBD patients from the two study branches of Paramsothy et al., which applied different FMT protocols resulting in different levels of donor microbiota engraftment (Figure 3C), were compared separately; p-values based on the Wilcoxon test with false discovery rate (Benjamini-Hochberg) correction for multiple hypothesis testing within studies.