Effects of fecal microbiota transplantation in subjects with irritable bowel syndrome are mirrored by changes in gut microbiome

Abstract

Irritable bowel syndrome (IBS) is a common disorder of the lower gastrointestinal tract. The pathophysiology is far from settled, but a gut microbial dysbiosis is hypothesized to be a contributing factor. We earlier published a randomized double-blind placebo-controlled clinical trial on fecal microbiota transplantation (FMT) for IBS - the REFIT trial. The present data set describes the engraftment and includes participants from the study who received active FMT; 14 participants with effect of FMT (Effect) and 8 without (No effect). Samples were collected at baseline, after 6 and 12 months. Samples from the transplants (Donor) served as a comparator. In total 66 recipient samples and 17 donor samples were subjected to deep metagenomic sequencing, and taxonomic and functional analyses were performed. Alpha diversity measures showed a significantly increased diversity and evenness in the IBS groups compared to the donors. Taxonomic profiles showed higher relative abundance of phylum Firmicutes, and lower relative abundance of phylum Bacteroidetes, compared to donors at baseline. This profile was shifted toward the donor profile following FMT. Imputed growth rates showed that the resulting growth pattern was a conglomerate of donor and recipient activity. Thirty-four functional subclasses showed distinct differences between baseline samples and donors, most of which were shifted toward a donor-like profile after FMT. All of these changes were less pronounced in the No effect group. We conclude that FMT induces long-term changes in gut microbiota, and these changes mirror the clinical effect of the treatment. The study was registered in ClinicalTrials.gov (NCT02154867).

Keywords: Irritable bowel syndrome; diversity; engraftment; fecal microbiota transplantation; functional features; growth rate; metagenomic sequencing; taxonomic profile.

Figures

Figure 1.

Panel A: Box plots with alpha diversity estimates where each sub-panel shows a different type of estimator. The colors group the samples into seven groups, and each point represents the richness estimate per sample. The data was filtered so that taxa occurring Effect group at baseline, after 12 months and the donor samples are visualized separate from the No effect group at baseline, after 12 months and the donor samples. The data were filtered so that taxa with a mean fractional abundance < 10−5 were excluded from the analysis.

Figure 1.

Panel A: Box plots with alpha diversity estimates where each sub-panel shows a different type of estimator. The colors group the samples into seven groups, and each point represents the richness estimate per sample. The data was filtered so that taxa occurring Effect group at baseline, after 12 months and the donor samples are visualized separate from the No effect group at baseline, after 12 months and the donor samples. The data were filtered so that taxa with a mean fractional abundance < 10−5 were excluded from the analysis.

Figure 2.

Panel A: Relative abundance of the most prevalent phyla for the seven groups. The data were filtered so that taxa not seen more than 5 times in at least 20% of the samples in the total dataset were removed. Following this, only taxa with a mean greater than 10−5 (fractional abundance > 0.00001) were kept before agglomeration at phylum level. Panel B: Differential abundance calculated with DESeq2 at phylum and family level, in the Effect group, 12 months vs. baseline. Positive log2FC indicate enriched taxa after FMT, negative log2FC indicate decreased taxa after FMT. Families indicated on the X-axis are colored according to phyla. FDR cut off for inclusion in the plot was < 10−4.

Figure 2.

Panel A: Relative abundance of the most prevalent phyla for the seven groups. The data were filtered so that taxa not seen more than 5 times in at least 20% of the samples in the total dataset were removed. Following this, only taxa with a mean greater than 10−5 (fractional abundance > 0.00001) were kept before agglomeration at phylum level. Panel B: Differential abundance calculated with DESeq2 at phylum and family level, in the Effect group, 12 months vs. baseline. Positive log2FC indicate enriched taxa after FMT, negative log2FC indicate decreased taxa after FMT. Families indicated on the X-axis are colored according to phyla. FDR cut off for inclusion in the plot was < 10−4.

Figure 3.

Multivariate regression analysis of the bacterial species. Each sample was labeled according to the corresponding study group. Panel A: This t1/t2-score plot of the orthogonal partial least squares projection to latent structures (OPLS) model (one predictive component and one orthogonal component) was built from the bacterial species composition in stool samples taken from the Donors, and from the Effect group at 3 time points: baseline, 6, and 12 months after FMT. The performance parameters R2Xcum, R2Ycum and Q2cum were 0.36, 0.61 and 0.32, respectively. Panel B: The top 50 bacterial species ranked by regression coefficients pertaining to the predictive components. To make the coefficients readily comparable, the independent variables for different taxa were scaled and centered prior to the analysis. The error bars indicate the confidence intervals of the coefficients. The coefficient is considered significant (above noise level), when the confidence interval does not include zero. Significant features are color-coded according to phylum.

Figure 3.

Multivariate regression analysis of the bacterial species. Each sample was labeled according to the corresponding study group. Panel A: This t1/t2-score plot of the orthogonal partial least squares projection to latent structures (OPLS) model (one predictive component and one orthogonal component) was built from the bacterial species composition in stool samples taken from the Donors, and from the Effect group at 3 time points: baseline, 6, and 12 months after FMT. The performance parameters R2Xcum, R2Ycum and Q2cum were 0.36, 0.61 and 0.32, respectively. Panel B: The top 50 bacterial species ranked by regression coefficients pertaining to the predictive components. To make the coefficients readily comparable, the independent variables for different taxa were scaled and centered prior to the analysis. The error bars indicate the confidence intervals of the coefficients. The coefficient is considered significant (above noise level), when the confidence interval does not include zero. Significant features are color-coded according to phylum.

Figure 4.

Growth rate score (GRiD) of the most frequent occurring species for samples in: (a) No effect group at baseline (green), after 12 months (purple) and the Donor samples (cyan); and (b) Effect group at baseline (pink), after 12 months (blue) and the Donor samples (cyan). GRiD score > 1 is an indication of bacteria in the growth phase, while GRiD score < 1 indicates bacteria in stationary or lag phase. The data were filtered so that species occurring < 15 times in < 30 samples were excluded from the analysis. FDR cut off for inclusion in the heat map was < 0.00005.

Figure 4.

Growth rate score (GRiD) of the most frequent occurring species for samples in: (a) No effect group at baseline (green), after 12 months (purple) and the Donor samples (cyan); and (b) Effect group at baseline (pink), after 12 months (blue) and the Donor samples (cyan). GRiD score > 1 is an indication of bacteria in the growth phase, while GRiD score < 1 indicates bacteria in stationary or lag phase. The data were filtered so that species occurring < 15 times in < 30 samples were excluded from the analysis. FDR cut off for inclusion in the heat map was < 0.00005.

Figure 5.

Differential abundance of functional subsystems with absolute abundance > 0.01% in the Effect group, 12 months vs. baseline. Positive log2FC indicate enriched subsystems after FMT, negative log2FC indicate decreased subsystems after FMT. Subsystem level 3 indicated on the X-axis are color-coded according to level 1 classification. FDR cut off for inclusion in the plot was < 0.05.

Figure 5.

Differential abundance of functional subsystems with absolute abundance > 0.01% in the Effect group, 12 months vs. baseline. Positive log2FC indicate enriched subsystems after FMT, negative log2FC indicate decreased subsystems after FMT. Subsystem level 3 indicated on the X-axis are color-coded according to level 1 classification. FDR cut off for inclusion in the plot was < 0.05.

Figure 6.

Multivariate regression analysis of the functional groups. Each sample was labeled according to the corresponding study group. Panel A: This t1/t2-score plot of the orthogonal partial least squares projection to latent structures (OPLS) model (one predictive component and one orthogonal component) was built from functional groups in stool samples taken from the Donors, and from the Effect group at 3 time points: baseline, 6, and 12 months after FMT. The performance parameters R2Xcum, R2Ycum and Q2cum were 0.55, 0.52 and 0.30, respectively. Panel B: The top 50 functional groups ranked by regression coefficients pertaining to the predictive components. To make the coefficients readily comparable, the independent variables for different functional groups were scaled and centered prior to the analysis. The error bars indicate the confidence intervals of the coefficients. The coefficient is considered significant (above noise level), when the confidence interval does not include zero. Significant features are color-coded according to level 1 classification.

Figure 6.

Multivariate regression analysis of the functional groups. Each sample was labeled according to the corresponding study group. Panel A: This t1/t2-score plot of the orthogonal partial least squares projection to latent structures (OPLS) model (one predictive component and one orthogonal component) was built from functional groups in stool samples taken from the Donors, and from the Effect group at 3 time points: baseline, 6, and 12 months after FMT. The performance parameters R2Xcum, R2Ycum and Q2cum were 0.55, 0.52 and 0.30, respectively. Panel B: The top 50 functional groups ranked by regression coefficients pertaining to the predictive components. To make the coefficients readily comparable, the independent variables for different functional groups were scaled and centered prior to the analysis. The error bars indicate the confidence intervals of the coefficients. The coefficient is considered significant (above noise level), when the confidence interval does not include zero. Significant features are color-coded according to level 1 classification.