Methods and compositions for treating ulcerative colitis

Abstract

The present disclosure provides methods and pharmaceutical compositions for treating ulcerative colitis (UC) in a subject in need thereof. In particular, the compositions described here comprise or are designed based on fecal bacteria associated with FMT-based UC treatment success or failure. Also provided are methods for screening patients for their suitability for a fecal bacteria-based UC treatment. Further provided are methods for screening fecal donors for optimized source materials for producing a fecal bacteria-based pharmaceutical composition.

Claims

1. A method of manufacturing a fecal microbe preparation comprising a bacterial strain for incorporation into a pharmaceutical composition, the method comprising: a) selecting a source material from a human donor based on a higher relative abundance of a metabolite in the source material, wherein the metabolite is selected from the group consisting of a secondary bile acid, a short chain fatty acid (SCFA), and a combination thereof, wherein the human donor is a member of a plurality of donors, and the higher relative abundance is relative to the average relative abundance of the metabolite in source material of the plurality of donors; and b) harvesting the bacterial strain from the selected source material to produce the fecal microbe preparation.

2. The method of claim 1, wherein the pharmaceutical composition is formulated so as to be suitable for treating ulcerative colitis (UC).

3. The method of claim 1, wherein the metabolite is the secondary bile acid.

4. The method of claim 1, wherein a plurality of samples of source material from the human donor are selected.

5. The method of claim 1, wherein the harvesting comprises culturing the bacterial strain.

6. The method of claim 5, wherein the cultured bacterial strain is from a genus selected from the group consisting of Eubacterium, Bacteroides, Faecalibacterium and Roseburia.

7. The method of claim 6, wherein the genus is Eubacterium and the bacterial species is Eubacterium rectale.

8. The method of claim 6, wherein the genus is Faecalibacterium and the bacterial species is Faecalibacterium prausnitzii.

9. The method of claim 5, wherein the method further comprises supplementing the fecal microbe preparation with viable cultured bacteria.

10. The method of claim 9, wherein the viable cultured bacteria comprise lyophilized bacteria.

11. The method of claim 1, wherein the pharmaceutical composition comprises a capsule, and the method further comprises incorporating the fecal microbe preparation into the capsule.

12. The method of claim 1, wherein the metabolite is a SCFA.

13. The method of claim 12, wherein the SCFA is butyrate.

14. The method of claim 1, wherein the metabolite comprises both the SCFA and the secondary bile acid.

15. The method of claim 1, further comprising lyophilizing the fecal microbe preparation in the presence of a cryoprotectant.

16. The method of claim 1, further comprising spray-drying the fecal microbe preparation.

17. The method of claim 15, wherein the cryoprotectant is selected from the group consisting of polyethylene glycol, skim milk, erythritol, arabitol, sorbitol, glucose, fructose, alanine, glycine, proline, sucrose, lactose, ribose, trehalose, dimethyl sulfoxide (DMSO), glycerol, or a combination thereof.

18. The method of claim 17, wherein the cryoprotectant is trehalose.

19. The method of claim 1, wherein the fecal microbe preparation comprises a community of microbes harvested from the source material.

20. The method of claim 3, wherein the secondary bile acid is selected from the group consisting of lithocholic acid, deoxycholic acid, ursodeoxycholic acid, and a combination thereof.

Description

EXAMPLES

Example 1

Trial Design and Microbiome Characterization of a FMT-Based Therapy of UC

(1) Patients with active UC are randomised in a double-blind controlled trial to intensive multi-donor FMT or placebo enemas 5 days per week for 8 weeks. Patients randomised to placebo are eligible to receive open-label FMT after the double-blind study period. FMT infusions are constituted from the blended homogenised stool of 3 to 7 unrelated donors, to increase microbial heterogeneity. Each patient receives all their FMT infusions from the same donor batch. See ClinicalTrials.gov at NCT01896635 and Paramsothy et al., Lancet 2017; 389: 1218-28, both of which are incorporated by reference in their entirety.

(2) Fecal samples are collected from individual donors, multi-donor FMT batches and study patients for molecular microbiological analyses and gastrointestinal microbial community profiling. Donor fecal samples (n=105) are collected from the 14 individual donors (n=55) and the 21 multi-donor FMT batches (n=50); eight samples are also taken from 4 placebo batches to serve as control.

(3) Seventy study patients provide a total of 314 fecal samples at screening, then every 4 weeks during treatment (blinded, and open label if applicable) and eight weeks after completing blinded or open-label FMT therapy. These patients also contribute 160 colonoscopic large bowel biopsies at study entry prior to treatment, after eight weeks of active or placebo treatment (the primary study endpoint), and where relevant after a further eight weeks of open-label treatment.

(4) All samples are stored at −80° C. immediately after collection until nucleic acid extraction. Fecal samples are homogenised and both DNA and RNA extracted using the MOBIO PowerViral RNA/DNA Isolation kit. Fecal RNA is then isolated from DNA using the MOBIO On-Spin Column DNase kit and Bioline Isolate II RNA micro clean-up kit. Colonic biopsy samples are homogenised and bacterial DNA and RNA extracted using the Macherey-Nagel RNA Isolation Kit. Colonic RNA is then isolated from DNA using the MOBIO On-Spin Column DNase kit and Macherey-Nagel RNA clean-up kit. Fecal and colonic RNA is then converted to cDNA using the SensiFAST cDNA Synthesis Kit (Bioline).

(5) The 16S rRNA gene fragment of the extracted DNA and RNA converted to cDNA is amplified using the Immolase DNA polymerase (95° C. for 10 min, 35 cycles of 94° C. for 30 s 55° C. for 10 s, 72° C. for 45 s, followed by a final step of 72° C. for 10 min) and the primers F27-519R. Sample indices and Illumina sequencing adapters are attached using the Nextera XT Index Kit according to the manufacturer's instructions. Amplicon sequencing is performed with the Illumina MiSeq Reagent kit v3 (2x300 bp) at the Ramaciotti Centre for Genomics. Shotgun metagenomics is performed on DNA extracted from 285 donor and patient fecal samples using Nextera XT DNA library prep kit and 2×250 bp HiSeq 2500 chemistry. This results in five datasets including fecal 16S DNA, fecal 16S cDNA, colonic biopsy 16S DNA, colonic biopsy 16S cDNA, and fecal shotgun DNA.

(6) Quality filtering of 16S rRNA sequences is conducted using software package mothur (Schloss (2009). “Introducing mothur: open-source, platform-independent, community-supported software for describing and comparing microbial communities.” Applied and Environmental Microbiology 75(23): 7537-7541) and follows the mothur MiSeq SOP (Kozich (2013). “Development of a Dual-Index Sequencing Strategy and Curation Pipeline for Analyzing Amplicon Sequence Data on the MiSeq Illumina Sequencing Platform.” Applied and Environmental Microbiology 79(17): 5112-5120). Paired-end sequences are merged into contigs, and poor quality contigs removed based on alignment quality and ambiguous base calls. A multiple sequence alignment is constructed using the SILVA SEED 16S rRNA reference alignment (Quast The SILVA ribosomal RNA gene database project: improved data processing and web-based tools. Nucleic Acids Res 2013; 41: D590-6), and poorly aligned sequences removed. To eliminate artefacts of sequencing at high frequency, rare sequences with high similarity to abundant sequences are clustered together. Chimeric sequences are removed using uchime (Edgar Bioinformatics, 2011; 27: 2194-200). Sequences are taxonomically classified using the Ribosomal Database project taxonomic outline (Wang Appl Environ Microbiol, 2007; 73: 5261-7) and those without classification at the kingdom level (unknown) or classified as mitochondrial or chloroplast are removed. Quality filtered sequences are then clustered into operational taxonomic units (OTUs) at 97% similarity using the opti-dust average neighbour algorithm (Westcott OptiClust, an Improved Method for Assigning Amplicon-Based Sequence Data to Operational Taxonomic Units. mSphere 2017; 2), and consensus taxonomies of the OTUs obtained using the classifications of sequences within each OTU. The resulting OTU count by sample data matrix is used for data analysis.

(7) Shotgun metagenomic DNA sequence reads are first analysed with DeconSeq (Schmieder Fast identification and removal of sequence contamination from genomic and metagenomic datasets. PLoS One 2011; 6: e17288) for identification and filtering of human DNA sequences. Sequencing reads are assessed for quality using FastQC (version 0.11.2). SolexaQA is then applied to calculate sequence quality statistics and perform quality filtering of the Illumina reads. Paired-end raw reads are trimmed with the BWA trimming mode at a threshold of Q13 (P=0.05) using the read trimmer module DynamicTrim. Filtered reads that are less than 50 bp in length are then discarded using LengthSort. The average microbial read counts per sample are 4,590,171±119,145 reads. MetaPhlAn2 (Truong et al. MetaPhlAn2 for enhanced metagenomic taxonomic profiling. Nat Methods 2015; 12: 902-3) is employed to generate taxonomic profiles from the shotgun reads, while HUMAnN2 (HMP Unified Metabolic Analysis Network) (Hall et al. A novel Ruminococcus gnavus clade enriched in inflammatory bowel disease patients. Genome medicine 2017; 9: 103) is used to determine the metabolic contributions within the samples. The HUMAnN2 pipeline involved mapping of the metagenomic reads against Uniref orthologous gene family, MetCyc UniPathway, and KEGG.

(8) During the initial double-blind FMT trial patients are allocated to active FMT treatment or placebo groups (FMT: two levels—Placebo or FMT, factor type—fixed) and each patient is sampled at three time points over eight weeks (Time: three levels—0, 4 and 8 weeks, factor type—fixed). Each patient is included in the experimental design as a random factor. After 8 weeks, the initial placebo group receive active FMT and are sampled further at weeks 4 and 8 of the open-label (non-blind) period. All patients receiving active FMT are also sampled at 8 weeks after completing active FMT therapy (blinded or open-label).

(9) To examine which microbial taxa differed between patients showing remission, data are combined from the blinded and open-label study periods, and then groups are created based on remission (Remission, two levels—Yes or No, factor type—fixed) and treatment group (FMT: three levels—Placebo, FMTblind or FMTopen). We then examined the effect of different categories of remission, including remission within the blinded trial (Remission among Placebo and FMTblind), and regardless of study phase (Remission among Placebo, FMTblind and FMTopen). Analyses are made using four different endpoint classifications—primary endpoint, clinical remission, endoscopic response, and endoscopic remission (steroid free endoscopic Mayo score of 0) (Paramsothy et al., Lancet 2017; 389: 1218-28).

(10) The effect of donor batch on remission is examined by allocating donor batches into two groups based on the number of patient remissions observed for each batch (DonorRemission, two levels—Yes or No, factor type—fixed). If more than 50% of the patients receiving a particular donor batch showed remission, the donor batch is allocated to the DonorRemission=Yes group, while all other donors are allocated to the DonorRemission=No group.

(11) Microbial communities are examined with respect to the above analyses in terms of alpha-diversity and beta-diversity, as well as comparing each taxon individually. Prior to diversity comparisons, the OTU counts are rarefied to account for uneven sequencing depths among samples (35,371,968 total clean reads; rarefied to 6447 clean reads/sample). Linear mixed models (LMMs) are used to examine the effects of the various predictors mentioned above. Models are created using the R packages LME4 (Bates et al., (2015). “Fitting Linear Mixed-Effects Models Using lme4.” Journal of Statistical Software 2015 67(1): 48) and lmerTest (Kuznetsova et al., (2017). “lmerTest Package: Tests in Linear Mixed Effects Models.” Journal of Statistical Software 82(13): 1-26).

(12) The Bray-Curtis (BC) dissimilarity coefficient for beta-diversity comparisons is employed, and prior to calculation of BC dissimilarities, OTU counts are transformed into square-root relative abundances. The BC distance matrix is visualized using non-metric multidimensional scaling (nMDS). PERMANOVA is used to examine the effects of the various predictors mentioned above. Dissimilarities, Figures and models are created using the R package ‘vegan’ (Oksanen, (2017) Vegan: Community Ecology Package). For per taxon comparisons, un-rarefied OTU counts are used in negative binomial generalised linear models (GLMs) with the sample totals used as an offset term. Contrasts are employed to examine the comparisons of interest within each analysis. Models are created using the R package DESeq2 (Love et al., Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2.” Genome Biology 15(12): 550). For confirmation, Linear Discriminant Analysis Effect Size (LEfSe) analyses are also performed (Segata et al. Metagenomic biomarker discovery and explanation. Genome biology 2011; 12: R60).

Example 2

Specific Bacterial Taxa Associated with Therapeutic Outcome

(13) Despite FMT therapy, five patients do not appear to have a major change in overall microbial structure, with their baseline samples clustering tightly with their samples during and post-FMT. Surprisingly, one of these patients achieves the primary outcome, and on further analysis, their overall microbiota structure is similar at baseline and during and post-FMT except for replacement of key species Bacteroides clarus (11.5% to 0.06%) and Akkermansia muciniphila (11.1% to 0%) with Faecalibacterium prausnitzii (4.9% to 11.1%), Eubacterium rectale (0.19% to 9.9%), and Eubacterium siraeum (0.96% to 14.2%).

(14) To identify specific microbial taxa significantly associated with achieving or not achieving the primary outcome across all patients, the abundances of each OTU is modeled in each dataset using negative binomial GLMs with remission as a predictor and presented the most discriminating taxa as potential biomarkers. A range of microbial taxa associated with lack of remission including Fusobacterium, Sutterella, Haemophilus, Escherichia, Megamonas, Clostridium cluster XIVa, Prevotella, Dialister, Veillonella and Bilophila, and these associations are in some datasets clearer when blinded and open label patients are stratified. The most consistent association with lack of achieving primary outcome is with Fusobacterium gonidiaformans, with this taxon identified in fecal 16S rRNA gene, mucosal 16S rRNA gene and transcript, and shotgun sequencing data. Of interest, Prevotella OTU2 (Prevotella copri in shotgun data) appears to flourish in several patients with FMT; however, this OTU is associated with lack of remission and patients who achieve remission tend to be those who resist dominance by Prevotella, having lower levels relative to patients who do not achieve remission.

(15) There is less consistency in taxa associated with remission across the datasets—these most commonly involved members of Firmicutes e.g. Clostridium cluster XVIII, Ruminococcus, Lachnospiraceae, Roseburia inulinivorans, and Eubacterium hallii. The associations among a range of these microbial taxa and primary outcome are confirmed using LEfSe.

(16) To further examine the consistency of these associations, GLMs is used with three other therapeutic outcomes including the stricter endpoint of complete endoscopic remission (steroid free endoscopic Mayo 0), endoscopic response, and clinical remission. Fusobacterium gonidiaformans, Sutterella wadsworthensis, Haemophilus, Escherichia, Megamonas, Clostridium cluster XIVa, Prevotella, Dialister, Veillonella and Bilophila are all consistently associated with lack of endoscopic remission. Analyses of endoscopic response and clinical remission are less consistent, likely due to the less strict nature of these endpoints; however, a range of the above taxa (including Fusobacterium, Haemophilus, Escherichia, Dialister and Veillonella) are still associated with negative outcomes.

Example 3

FMT Results in Functional Changes Associated with Therapeutic Outcome

(17) Microbial functional changes across FMT therapy and therapeutic outcome are characterised, with analysis focusing on outputs from KEGG and MetaCyc pathways. FMT, but not placebo, resulted in significant changes in microbial KEGG (F.sub.1,42=2.5, P=0.027) and MetaCyc pathways (F.sub.1,43=2.3, P=0.010). Despite intense FMT, patient microbial functional profiles remain significantly different to that of the donors (475)=2.0, P=0.001, Permutations=999). Similar to the taxonomic profiles, FMT increases homogeneity (reduced dispersion) in the functional profiles across patient samples, but not to the level of individual donors or donor batches. Due to the significant patient variability that is observed in the data, a constraint on the factor ‘patient’ is applied, which shows a clearer delineation between FMT and placebo.

(18) Specific pathways associated with primary outcome are then identified using GLMs. Pathways such as benzoate degradation, glycerophospholipid metabolism, secondary bile acid biosynthesis, ppGpp biosynthesis, pyruvate fermentation to acetate and lactate (short chain fatty acid biosynthesis), biosynthesis of ansamycins, and starch degradation are all associated with positive primary outcome. Taxa contributing to beneficial pathways included Eubacterium, Ruminococcus, Lachnospiraceae, Roseburia, Dorea and Coprococcus, consistent with the taxonomic analysis associating these species with positive therapeutic outcome. Furthermore, the relationship between short chain fatty acid biosynthesis and positive primary outcome is confirmed in the predicted metagenome (PICRUSt) of the mucosal microbiome. In contrast, heme biosynthesis, lipopolysaccharide biosynthesis, ubiquinone and other terpenoid quinine biosynthesis, lysine biosynthesis and oxidative phosphorylation pathways are all associated with negative primary outcome. The relationships between a range of these pathways and primary outcome are confirmed using LEfSe.

(19) Similar to the taxonomic analysis, the analyses are replicated against the three other therapeutic outcomes of complete endoscopic remission, endoscopic response, and clinical remission. The results from these endpoints show consistent associations as those observed for the primary study outcome.

Example 4

Donor Taxonomic and Functional Profiles Associated with Therapeutic Outcome

(20) Despite FMT therapy over 8 weeks, patient taxonomic and functional profiles remain different to those of individual donors and donor batches. Thus, specific factors associated with donor suitability are evaluated by analysis of the donor fecal samples (16S rRNA gene and transcript, as well as shotgun metagenomic datasets) relative to the four different therapeutic outcomes.

(21) Donor batches are categorised based on the total number of samples and number of patients that achieved a positive primary outcome, with donor batches leading to >50% remission classified as effective and the rest as ineffective. α-diversity and β-diversity within effective and ineffective batches are compared in all datasets, with no clear patterns emerging between the two groups. While some differences in global β-diversity are observed (F.sub.1,18=1.7, P=0.071), this is likely due to the high inter-donor variability.

(22) Specific taxonomic differences between effective and ineffective batches are then analysed using GLMs. Bacteroides OTU187 is in higher abundance in effective batches, and consistently Bacteroides fragilis, as well as Bacteroides finegoldii, are identified as taxonomic markers in these batches. In contrast, Bacteroides uniformis and Bacteroides coprocola are associated with ineffective batches. Other donor microbial taxa associated with ineffective batches included Clostridium cluster XIVa (OTU173), a taxon that is associated with negative primary outcomes, and Streptococcus (OTU56), which is found in both the 16S rRNA gene and transcript datasets.

(23) No clear differences in global pathway compositions are found between effective and ineffective batches. However, a range of pathways are identified to be in higher abundance in either effective or ineffective batches. Specifically, pathways such as fatty acid biosynthesis and propanoate metabolism are higher in effective batches while terpenoid backbone biosynthesis and bacterial chemotaxis are higher in ineffective batches.

(24) Similar analyses are conducted for the three other therapeutic endpoints. The strict endpoint of endoscopic remission and the less strict endpoint of endoscopic response showed similar outcomes to the primary study endpoint. One notable difference is the clustering of effective batches at the higher end of α-diversity when shotgun taxonomic data is classified by endoscopic response. This higher level of α-diversity is also identified for effective batches classified by the clinical remission endpoint. In fact, classification of donor batch effectiveness based on the clinical remission endpoint showed the strongest signs of consistency with the results of the patient analysis. Sutterella wadsworthensis, previously associated with lack of remission in patients, is associated with ineffective batches in clinical remission. Further, pathways such as secondary bile acid biosynthesis, glycerophospholipid metabolism and biosynthesis of ansamycins are all associated with positive patient outcomes and are associated with effective batches. Moreover, heme biosynthesis is a strong marker for negative primary outcome in patients and is higher in ineffective batches.