Microbial shifts and signatures of long-term remission in ulcerative colitis after faecal microbiota transplantation

Abstract

Faecal microbiota transplantation (FMT) may contribute towards disease remission in ulcerative colitis (UC), but it is unknown which factors determine long-term effect of treatment. Here, we aimed to identify bacterial signatures associated with sustained remission. To this end, samples from healthy donors and UC patients-grouped into responders and non-responders at a primary end point (week 12) and further stratified by sustained clinical remission and relapse assessed at ⩾1-year follow-up were analysed, comparing the efficacy of FMT from either a healthy donor or autologous faeces. Microbiota composition was determined with a 16S rRNA gene-based phylogenetic microarray on faecal and mucosal samples, and functional profiles were predicted using PICRUSt with quantitative PCR verification of the butyrate production capacity; short-chain fatty acids were measured in faecal samples. At baseline, UC patients showed reduced amounts of bacterial groups from the Clostridium cluster XIVa, and significantly higher levels of Bacteroidetes as compared with donors. These differences were reduced after FMT mostly in responders. Sustained remission was associated with known butyrate producers and overall increased butyrate production capacity, while relapse was associated with Proteobacteria and Bacteroidetes. Ruminococcus gnavus was found at high levels in donors of failed FMT. A microbial ecosystem rich in Bacteroidetes and Proteobacteria and low in Clostridium clusters IV and XIVa observed in UC patients after FMT was predictive of poor sustained response, unless modified with a donor microbiota rich in specific members from the Clostridium clusters IV and XIVa. Additionally, sustained response was associated with restoration of the butyrate production capacity.

Figure 1

Kruskal–Wallis comparison of the 130 genus-like groups included in the HITChip between donors, non-responders and responders (the latter subdivided into patients who relapse (Rel) or sustain remission (SustRem) at ⩾1-year FU). Abundance of the different bacterial groups range from less (blue) to more (red) abundant. Red and blue outlines highlight groups either enriched or at lower levels in patients, respectively. *Genus-like groups of known butyrate-producing bacteria. Significant different groups at (a) baseline (t0, P<0.05; FDR<0.3) and (b) t12w (P<0.05; FDR<0.2).

Figure 2

Redundancy analysis (RDA) of samples at t0, t12w and t1 year. Grey arrows represent the 65 best-fitting genus-like bacterial groups, shown grouped into their corresponding phyla (Bacteroides and Proteobacteria) and orders belonging to the Firmicutes phylum (Bacilli, Clostridium clusters IV and XIVa) (detailed information of bacterial groups associated with quadrants I–IV in Supplementary Table 3). Blue arrow in quadrant IV indicates R. gnavus. Twenty per cent of the variation in the data set is explained in the first two axes. (A) Samples are divided into FMT-A with sustained remission (a) and relapse (b) and FMT-D with sustained remission (c) and relapse (d) and their corresponding donors (of first or second FMT, shown as I or II). (B) Red crosses show association of explanatory variable ‘Sustained Remission’ (which includes group relapse, sustained remission and donors) with samples and bacterial groups. *Patient 8 includes two samples after 1 year. **⩾1- year sample from Patient 5 did not fulfil the quality control criteria of the HITChip pipeline.

Figure 3

Principal coordinate analysis (PCoA) for prediction of (sustained) response based on baseline microbiota composition for the FMT-A group (a) and the FMT-D group (c). (a) In the FMT-A group, patients with microbiota profiles at baseline more similar to donor samples (‘healthy’-like profiles, on the negative section on the x axis, highlighted in yellow) were more likely to respond (highlighted in green). This section is associated with known butyrate-producing bacteria related to Coprococcus eutactus. (b) Responders and non-responders showed significantly different baseline microbiota composition (P=0.012). Distances in baseline microbiota profiles to donor samples (score on the first principal coordinate) could also predict long-term response (FU ⩾1 year). (c) In the FMT-D group, baseline microbiota composition of patients was not a determining factor on the prediction of (sustained) response. NR, non-responders; R-Rel and R-SustRem, responders that relapse or sustain remission respectively at the ⩾1-year FU.

Figure 4

(a) qPCR of butyryl CoA (normalized to 16S rRNA gene copies) of donor samples and patients samples at baseline (t0) and 12 weeks after treatment (t12w). Patients are divided into non-responders (NRs) and responders, and were subdivided into those that relapse (R-Rel) or sustain remission (R-SustRem) at ⩾1-year FU. Levels of butyryl CoA were significantly lower in patients (NRs and responders) at t0 and in NRs at t12. Levels of ButCoA in responder samples with sustained remission at 1 year increase by 6.7-fold from t0 to t12w. (b and c) Relative abundance (%) of genus-like groups that include known butyrate-producing bacteria in (b) faecal samples of donors and patients at baseline (t0) and 12 weeks after treatment (t12w) and (c) in biopsy samples of patients at baseline (t0) and 12 weeks after treatment (t12w).

Figure 5

Proposed model of microbiota signatures in UC and healthy donors, associated with short- and long-term success of FMT.