Cohort-specific determinants of donor strain engraftment following multi-donor faecal microbiota transplantation in two randomised clinical trials

Abstract

Disrupted human gut microbiota have been associated with the development of certain disease states, including obesity and ulcerative colitis (UC). Faecal microbiota transplantation (FMT) from healthy donors is a promising avenue to shift the microbiome profile of the recipient towards that of the donor, potentially ameliorating related symptoms. Several recent meta-analyses have investigated the clinical and microbial determinants that influence the retention of transplanted donor microbial strains within the recipient gut microbiome following FMT (i.e. engraftment). However, the specific factors that affect donor strain engraftment in different disease states require further exploration. Here, we perform a strain engraftment analysis on data from two multi-donor FMT clinical trials: the Gut Bugs Trial for obesity and the FOCUS Trial for UC. Using donor strain matching, the donor-recipient pairings of the FOCUS Trial were first predicted in a blinded manner. The subsequent, unblinded, strain engraftment analysis of both datasets highlighted a differential effect of donor-recipient microbiome complementarity on engraftment across the two disease cohorts; greater engraftment efficiency was associated with increased donor-recipient microbial similarity in the FOCUS Trial, and decreased similarity in the Gut Bugs Trial, suggesting that the factors influencing engraftment may differ across disease cohorts.

Keywords: Faecal microbiota transplantation; human gut microbiome; obesity; strain engraftment; ulcerative colitis.

Figure 1.

Predicted donor profile for each FOCUS Trial FMT recipient with donor-matching strains. (a): Novel donor-matching strains (i.e. not present at baseline) identified from a stringent normalised DNA distance threshold of 0.2 are shown for FMT recipients (above; dark red identified rows) and placebo recipients (below; dark blue identified rows) at 8 weeks post-intervention. (b): Ambiguous donor strains identified in placebo recipients were removed from the FMT data, and the number of donor-matching strains at week 8 was re-plotted. The heatmap scale represents the number of predicted engraftment events from each donor (columns) to each recipient (rows) involving distinct species strains. Only recipients with donor-matching strains are shown (FMT n = 28/32, placebo n = 9/20). (c): The density of DNA distance (median normalised) values for strains from inter- (grey) and intra-subject (pink) pairwise comparisons is plotted. Donor batch samples were excluded from pairwise comparisons due to their ambiguity as intra- or inter-subject comparisons. Total pairwise comparisons considered = 40,052. Pairwise comparisons ≥5 were not plotted due to their low density. Strains with ≤0.2 normalised DNA distance (dashed line) are considered a strain match. (d): Novel donor-matching strains unique to the FMT recipient cohort were identified in 28/32 FMT recipients at week 8. The distribution of donors with matching strains was plotted to assign a tentative donor profile for each of these FMT recipients (range: 1−7 donors per recipient). Bars are coloured by FMT donors. Count data and bar height correspond to the number of distinct species strains matching each donor in each FMT recipient sample at week 8. FMT, faecal microbiota transplantation.

Figure 2.

FMT recipient microbiome profiles shifted towards D043 following treatment. (a): Classical (metric) multidimensional scaling of species relative abundance in FMT and placebo recipient samples at baseline and 8 weeks post-intervention using the Bray-Curtis dissimilarity index, compared to individual donor and donor batch samples. Each point represents a sample. Where a single donor (individual or batch) had multiple samples, the mean PC1 and PC2 coordinates were plotted. (b): D043 species strain matches in FMT recipients at week 8. (c): Prevotella/Bacteroides ratio and (d) Shannon’s diversity index values for individual donor samples. Boxes represent the interquartile range (IQR) split by the median, with whiskers extending up to 1.5x the IQR. FMT, faecal microbiota transplantation; PC1, principal coordinate 1; PC2, principal coordinate 2; *p < 0.05.

Figure 3.

Optimised strain matching thresholds have greater F1 scores than initial thresholds. The F1 score was calculated for DNA distance thresholds between 0.001−3, at intervals of 0.001 for (a) the FOCUS Trial and (b) the Gut Bugs Trial, by comparing the respective predicted FMT donor-recipient pairings with the true (reference) pairings. By default, donor strain matches in FMT recipients at the first post-intervention timepoint (FOCUS Trial: week 8; Gut Bugs Trial: week 6) were only considered to be novel if the strain was absent from the recipient at baseline (baseline subtraction). An additional step to remove donor strains with matches in the placebo cohort was also performed (placebo + baseline subtraction). In (a), thresholds ≥0.002 are plotted, to remove the initial long tail. The confusion matrices for the FOCUS Trial (c) optimal (0.267) and (d) initial (0.2), and the Gut Bugs Trial (e) optimal (0.123) and (f) initial (0.2) strain matching thresholds used for donor strain engraftment analysis are also shown. FMT, faecal microbiota transplantation; ref, reference; pred, predicted.

Figure 4.

Engraftment efficiency varied across donors within each FMT trial. Mean strain engraftment efficiency was calculated as the mean percentage of distinct donor strains engrafted in their true recipients, for each FMT donor in (a) the FOCUS Trial and (b) the Gut Bugs Trial. Only significant pairwise comparisons where p < 0.01 following post-hoc testing (Dunn test) are shown. FMT, faecal microbiota transplantation; **p < 0.01; ***p < 0.001.

Figure 5.

The correlation between donor-recipient species- or functional β-diversity (Bray-Curtis dissimilarity index) and engraftment efficiency differed between the trials. The correlation between strain engraftment efficiency and mean donor-recipient (baseline) species Bray-Curtis β-diversity was plotted using a linear model for (a) the FOCUS Trial and (c) the Gut Bugs Trial. Species profiles were obtained from MetaPhlAn3. Points are coloured by FMT donors. The grey shaded region represents the 95% confidence interval. FOCUS Trial degrees of freedom = 171; Gut Bugs Trial degrees of freedom = 154. The correlation between strain engraftment efficiency and mean donor-recipient (baseline) functional Bray-Curtis β-diversity was plotted using a linear model for (b) the FOCUS Trial and (d) the Gut Bugs Trial. Genes on high-quality MAGs (high-quality genes) were used for functional analysis. COG functional annotations for high-quality genes in each sample were obtained using eggNOG. The relative abundance of each COG functional category was calculated as a proportion of the total number of high-quality genes for each sample. Relative abundance data was renormalised for each sample as some genes had multiple functional annotations. Samples with no high-quality genes were omitted. Points are coloured by FMT donors. The grey shaded region represents the 95% confidence interval. FOCUS Trial degrees of freedom = 110; Gut Bugs Trial degrees of freedom = 154. FMT, faecal microbiota transplantation; MAG, metagenome-assembled genome; COG, clusters of orthologous groups.

Figure 6.

The influence of biological metrics on recipient-specific donor engraftment efficiency differs across FMT trials. Engraftment efficiency was compared for each true donor-recipient pairing using a linear mixed model in (a) the FOCUS Trial and (b) the Gut Bugs Trial. Donor and recipient α-diversity (Shannon index), donor and recipient Prevotella/Bacteroides ratio, species β-diversity (Bray-Curtis dissimilarity index and Jaccard dissimilarity index), functional β-diversity (Bray-Curtis dissimilarity index) and batch size (FOCUS Trial only) were considered fixed effects. Within-donor and within-recipient variation were considered random effects. For each dataset, the fixed effect estimates are plotted for both the full model (interactions between donor and recipient α-diversity, and between donor and recipient Prevotella/Bacteroides ratio; dark grey bars) and reduced model (no interactions between the fixed effects; light grey bars). Fixed effects are grouped as recipient-specific, donor-recipient interaction, donor-specific, or procedural factors (represented by coloured vertical bars to the left of each plot). FMT, faecal microbiota transplantation; LMM, linear mixed model; *p < 0.05; **p < 0.01.