Microbial bile salt hydrolases mediate the efficacy of faecal microbiota transplant in the treatment of recurrent Clostridioides difficile infection

Objective Faecal microbiota transplant (FMT) effectively treats recurrent Clostridioides difficile infection (rCDI), but its mechanisms of action remain poorly defined. Certain bile acids affect C. difficile germination or vegetative growth. We hypothesised that loss of gut microbiota-derived bile salt hydrolases (BSHs) predisposes to CDI by perturbing gut bile metabolism, and that BSH restitution is a key mediator of FMT’s efficacy in treating the condition. Design Using stool collected from patients and donors pre-FMT/post-FMT for rCDI, we performed 16S rRNA gene sequencing, ultra performance liquid chromatography mass spectrometry (UPLC-MS) bile acid profiling, BSH activity measurement, and qPCR of bsh/baiCD genes involved in bile metabolism. Human data were validated in C. difficile batch cultures and a C57BL/6 mouse model of rCDI. Results From metataxonomics, pre-FMT stool demonstrated a reduced proportion of BSH-producing bacterial species compared with donors/post-FMT. Pre-FMT stool was enriched in taurocholic acid (TCA, a potent C. difficile germinant); TCA levels negatively correlated with key bacterial genera containing BSH-producing organisms. Post-FMT samples demonstrated recovered BSH activity and bsh/baiCD gene copy number compared with pretreatment (p<0.05). In batch cultures, supernatant from engineered bsh-expressing E. coli and naturally BSH-producing organisms (Bacteroides ovatus, Collinsella aerofaciens, Bacteroides vulgatus and Blautia obeum) reduced TCA-mediated C. difficile germination relative to culture supernatant of wild-type (BSH-negative) E. coli. C. difficile total viable counts were ~70% reduced in an rCDI mouse model after administration of E. coli expressing highly active BSH relative to mice administered BSH-negative E. coli (p<0.05). Conclusion Restoration of gut BSH functionality contributes to the efficacy of FMT in treating rCDI.


Supplementary Material
16S rRNA gene copy number between donor and FMT samples, and Wilcoxon signed-rank test to compare changes between pre-and post-FMT.

Metataxonomic analysis:
The output data was analysed using the Mothur package (v1.35.1) following the MiSeq SOP Pipeline [4]. Sequence alignments were performed using the Silva bacterial database (www.arb-silva.de/), and the RDP database reference sequence files were used for sequence classification using the Wang method [5]. Operational Taxonomic unit (OTU) taxonomies (phylum to genus) were established using the RDP MultiClassifier Script. Data was resampled and normalised to the lowest read count in Mothur (11604 reads per sample), which resulted in >99.5% coverage within each sample. Where possible, species were identified from OTU data using a standard nucleotide BLAST of the 16S rRNA sequences (NCBI) with strict criteria (query cover 100% and ≥97% identity, with no other candidate species above ≥97% identity). Genus-level annotation was made where query cover was 100% and ≥94% identity.
The non-metric multidimensional scaling (NMDS) plot and PERMANOVA p-values were generated using the UniFrac weighted distance matrix generated from Mothur, and analysed using the Vegan library within the R statistical package [6]. Family-level extended error bar plots were generated using the Statistical Analysis of Metagenomic Profiles (STAMP) software package using White's nonparametric t-test with Benjamini-Hochberg FDR [7]. The α-diversity (Shannon diversity index, H') and richness (total number of bacterial taxa observed, Sobs) were calculated within Mothur and statistical tests (see Supplementary Methods 1.10) were performed using GraphPad Prism v7.03. A p-value of 0.05 and a q-value of 0.05 was considered significant.
Changes in microbial composition were also assessed at the OTU level. Differences in mean out relative proportions > 1% were measured between donor and pre-FMT samples, and between pre-FMT and post-FMT samples, using White's non-parametric test and Benjamini-Hochberg FDR. From these data, OTUs were analysed that were enriched in donors in comparison to pre-FMT samples, and those enriched post-FMT in comparison to pre-FMT samples.
Sequencing data from this study (in fastq-format) are publicly available for download at the European Nucleotide Archive (ENA) database using study accession number PRJEB30298 (http://www.ebi.ac.uk/ena/data/view/PRJEB30298).

Supplementary Material
Bile acid extraction was performed by taking 75µl of supernatant and adding 225µl of cold methanol, followed by incubation at -30ᵒC for 2 hours; tubes were centrifuged at 9500 x g and 4ᵒC for 20 min and 120µl of supernatant was loaded into vials. UPLC MS analysis was otherwise as described above.

Integration of metataxonomic and UPLC-MS bile acid data:
Regularised Canonical Correlation Analysis (rCCA) was used to correlate metataxonomic data (familylevel) with UPLC-MS bile acid profiling data from the same samples using the mixOmics library within R [14,15]. This technique maximises the correlation between the two data sets X and Y. The shrinkage method was applied to determine regularisation parameters. Unit representation plots were generated using the plotIndiv function, where each sample is represented as a single point on the

Bacteria used as standards for bsh gene qPCR and in batch cultures:
Bacteroides plebius, Bacteroides ovatus, Bacteroides vulgatus, Collinsella aerofaciens and Blautia obeum were previously isolated from the stool of a healthy male donor in his 20's. Bacteroides plebius was isolated from fastidious anaerobe agar plates (Acumedia, USA) with 5% horse blood (VWR, USA).
DNA extraction was performed on the isolates using the EZNA Bacterial DNA Kit (Omega, USA) with the addition of a bead beating using the Bullet Blender Storm (speed 8 for 3 min). A ~900 bp region of the 16S gene was amplified using previously published primers[16] and DNA was sequenced at Macrogen Europe. Isolates were identified by performing a standard nucleotide BLAST of the 16S rRNA sequences (NCBI).
For qPCR, one bacterial strain from the relevant reference group was selected as a standard for each primer set (bsh group 1A -Bacteroides plebius; bsh group 1B -Bacteroides ovatus; bsh group 3C -Blautia obeum; baiCD -Clostridium scindens (DSMZ 5676, Braunschweig, Germany)). Serial dilutions of each isolate were used to create a standard curve. Whilst bsh primers were degenerate, each primer set used was specific for an individual BSH group. The protocol used for qPCR thermocycling and gene copy number calculation was as previously-outlined [17]. For batch cultures, naturally-BSHproducing organisms that were used were: Bacteroides ovatus (BSH group 1B), Collinsella aerofaciens (group 2), Bacteroides vulgatus (group 3C) and Blautia obeum (group 3C) (two organisms were picked from group 3C given that this is a large group). The E. coli used that had been engineered to constitutively express bsh genes were: E. coli expressing a bsh gene with low activity ('E. coli BSHlow', with bsh gene cloned from Lactobacillus salivarius, with narrow substrate range against conjugated bile acids), and E. coli expressing a bsh gene with high activity ('E. coli BSHhigh', with bsh gene cloned from Bifidobacterium adolescentis, containing BSH with high glycine-and taurine-deconjugating activity).

Batch cultures -Clostridioides difficile spore preparation and enumeration, and further methodology:
C. difficile spores were prepared using previously-described methods [18]. Specifically, C. difficile 010, 012 and 027 were grown anaerobically on fastidious anaerobe agar plates supplemented with 5% defibrinated horse blood (VWR, Radnor, USA) and incubated at 37ᵒC for 7 days. The growth was removed from the plates using a sterile loop and resuspended in 1ml of sterile water. Next, 1ml of 95% ethanol was mixed with the cell suspension and was incubated for 1 hour at room temperature.
The cell suspension was then centrifuged at 3000 x g and resuspended in 1ml sterile water. Spores were stored at -80ᵒC until use.

Enumeration of C. difficile counts from mouse stool samples:
Mouse stool samples were collected into Carey-Blair medium and immediately homogenised.

Statistical analysis:
Multivariate UPLC-MS bile acid profiling data analysis is described in Supplementary Methods 1.5.1.
Univariate statistics were performed using GraphPad Prism, v7.03; Mann-Whitney test was used to compare donor with pre-FMT or post-FMT, whilst Wilcoxon rank sum or Friedman test was used as appropriate to compare pre-FMT with post-FMT samples (all statistics were two-tailed tests).

Correlation of metataxonomic and metabolomics data was undertaken via regularised Canonical
Correlation Analysis (rCCA), using the mixOmics library within R [21].

Supplementary Results:
Stool from patients with rCDI demonstrated a significantly reduced α-diversity (as assessed by Shannon diversity index, p<0.001, Mann-Whitney, Supplementary Figure 3A), significantly reduced richness (Sobs, p<0.0001, Supplementary Figure 3B), and profoundly altered microbial community structure (as measured by NMDS, p<0.001, PERMANOVA, Supplementary Figure 3C) as compared to Supplementary Material contribute to this finding. One such explanation is that the restitution of gut BSH functionality post-FMT creates a larger pool of deconjugated primary bile acids, the substrate for further gut bacterial enzyme degradation and conversion of primary into secondary bile acids within the colon. An additional explanation is that FMT may also be associated with the restitution of microorganisms with 7--dehydroxylase activity that convert primary to secondary bile acids. We identified the enrichment in stool of unconjugated primary bile acids (chenodeoxycholic acid (CDCA) and cholic acid (CA)) and reduction in baiCD operon copy number pre-FMT in comparison to healthy donors, and that baiCD copy number/ predicted 7--dehydroxylase functionality was restored by FMT. It is interesting to note that C. scindens abundance in the gut microbiota was reduced pre-FMT in comparison with donors (albeit with <1% change in mean abundance), but there was no increase in abundance of this species post-FMT, suggesting that the increase in baiCD post-FMT was by yet unidentified bacteria with 7--dehydroxylation activity.
One drawback with interpreting our human sample data is (as is standard with most human studies of FMT and rCDI), our patients were taking vancomycin at the time of collection of pre-FMT samples, meaning that we are unable to definitively state that the restored gut BSH functionality that we observed post-FMT represents transfer of BSH-producing organisms from the donor, rather than the recovery of species in the gut microbiota of recipients that were being suppressed by vancomycin. It was for this reason that went on to perform batch culture and mouse experiments that enabled us to assess further the direct impact of BSH activity upon C. difficile in the rCDI setting.
There are other intuitive reasons that would support restoration of BSH-producing organisms to be a key contributor to FMT's success in the treatment of rCDI. Firstly, BSH-producing organisms have a relatively high prevalence in healthy stool, and the enzyme itself has relative insensitivity to oxygen, and robust activity over a wide pH range [25,26]. It has been demonstrated that however unrefined the preparation process of the slurry is for FMT is (i.e. regardless of ambient air, diluent or route of administration used, etc), that FMT remains highly-effective at treating rCDI. Of particular note, even a sterile faecal filtrate -prepared in a commercial blender, and not in anaerobic conditions -has been demonstrated to be of comparable efficacy to conventional FMT at treating rCDI [27]. It is more feasible that BSH (rather than 7--dehydroxylase) is able to reach the distal gut functionally intact after FMT (particularly in the scenario of sterile faecal filtrate, where no spores are present in the administered material, given that 7--dehydroxylase is produced by spore-forming Clostridia).