Secreted microbial metabolites modulate gut immunity and inflammatory tone

Evidence is emerging that microbiome–immune system crosstalk regulates the tenor of host intestinal immunity and predisposition to inflammatory bowel disease (IBD). We identified five NF-κB suppressive strains affiliated with Clostridium clusters IV, XIVa and XV that independently suppressed secretion of the chemokine IL-8 by peripheral blood mononuclear cells and gut epithelial organoids from healthy human subjects, as well as patients with the predominant IBD subtypes, Crohn’s disease and ulcerative colitis. The NF-κB suppressive Clostridium bolteae AHG0001, but not C. bolteae BAA-613, suppressed cytokine-driven inflammatory responses and endoplasmic reticulum stress in gut epithelial organoids derived from Winnie mice that develop spontaneous colitis. This predicted in vivo responses thereby validating a precision medicine approach to treat Winnie colitis and suggesting the microbiome may function as an extrinsic regulator of host immunity. Finally, we identified a novel molecule associated with NF-κB suppression indicating gut bacteria could be harnessed to develop new therapeutics.


Introduction 34
The human gut is the largest immune organ of the body and gut epithelial cells play a key role 35 in the establishment and maintenance of gut homeostasis, as well as rapid responses to 36 infection 1 . The gut is colonised by a diverse microbiota that has co-evolved with its host and 37 forms a durable symbiotic relationship through its modulation of innate and adaptive immune 38 and Grp78 in the colon ( Figure 4J). Together, these results showed the feasibility of applying 219 a precision medicine approach using ex vivo organoid cultures to accurately predict treatment 220 response in colitis. 221 Finally, we hypothesised that intraspecies variations in NF-B suppressive capacity, together 222 with the influence of culture media on bioactive production, could facilitate identification of 223 candidate bioactive encoding BGCs and/or bioactive molecules produced by C. bolteae using 224 comparative genomics or metabolomics. Comparative genomic analyses revealed that C. Firmicutes affiliated bacteria are amongst the most abundant gut microbes and these taxa are 243 widely recognised to possess immunomodulatory capacities 15,37,38 . However, they are poorly 244 represented in culture collections and their ability to modulate immune responses remain 245 largely undefined. In this study, we identified five gut bacterial strains affiliated with 246 Clostridium clusters IV, XIVa and XV that are comparable to the well-characterised F. 247 prausnitzii A2-165 strain in their ability to suppress NF-B activation. The NF-B suppressive 248 bioactivities were characterised by significant biochemical and intraspecies variations 249 suggesting there may be extensive functional redundancy and NF-B suppressive capacity may 250 be more prevalent than previously appreciated. This is consistent with Geva-Zatorsky et al., 2 251 who determined that as few as 53 isolates were associated with over 24,000 immune 252 phenotypes that include functionalities relevant to IBD (e.g. Treg induction). Modulating host 253 immune responses may support the ability of gut bacteria to colonise and persist in the gut 254 environment and the ability of the microbiota to act as an extrinsic regulator of host immunity 255 may underpin immune homeostasis and contribute to disease risk in genetically susceptible 12 individuals. 257 IBD is characterised by a dysregulated immune response with select genetic susceptibilities 258 affecting therapeutic responsiveness 30,31 . In order to develop improved precision treatments 259 for IBD we therefore used gut epithelial organoids and immune cells to identify bacteria 260 capable of supressing cytokine mediated inflammatory responses. The heat and proteinase K 261 resilient bioactives showed strong suppression of IL-8 secretion in organoids and immune cells 262 from healthy, CD and UC subjects. Interestingly, the putative peptide bioactives produced by 263 F. prausnitzii A2-165 and C. aldenense AHG0011 were notably less suppressive in UC derived 264 organoids and PBMCs, and CD organoids, when compared to organoids derived from healthy 265 controls; this may be reflective of the increased endogenous protease activity in IBD 39 . Our in 266 vitro and ex vivo data also suggested that functional capacity rather than phylogeny may be the 267 key determinant of biologic effects. To explore this hypothesis, we capitalised on the C. 268 bolteae intraspecies differences and demonstrated that a precision medicine approach could be 269 applied to alleviate established colitis in Winnie mice. Notably, treatment with C. bolteae 270 AHG0001 CS was associated with a rapid onset of action with improvement in diarrhoea, 271 alleviation of inflammation and ER stress, as well as restoration of goblet cell numbers and 272 mucin production. Mucosal and histologic healing are amongst amongst the best predictors of 273 long-term outcomes in IBD and taken together our data suggests a precision medicine approach 274 could be applied to microbiome based IBD treatment. 275 The NF-B suppressive strains carry multiple BGCs, many of whose products remain cryptic, 276 underlining the inherent challenges in applying genomic based approaches to map genotype 277 with phenotype. In addition, the biosynthesis of bioactives by gut bacteria may be principally 278 driven through modest modifications of common primary metabolites that are underpinned by 279 small BGCs 35, 40 . As the medium dependent effects on NF-B suppression may affect the 13 therapeutic efficacy of live biotherapeutics for IBD, we therefore used a bioassay-guided 281 fractionation and a comparative metabolomic approach to identify a novel low molecular 282 weight non-polar molecule that was associated with the NF-B suppressive activity of C. 283 bolteae AHG0001. Critically, this molecule was not associated with medium components 284 suggesting the suppressive activity is unlikely to be due to the biotransformation of medium 285 components 41 . Consistent with other microbial bioactives, the C. bolteae AHG0001 bioactive 286 acts independently of the bacterial cell and suppresses the inflammatory response in animals. 287 In summary, our IBD guided approach provides new opportunities to rationally bioprospect 288 the gut microbiota for precision live biotherapeutic strains and/or bioactives that could be used 289 to expedite the development of safer and more efficacious therapeutics. 290

Materials & Methods 291
Bacterial strains, culture conditions and analyses. Firmicutes  replicates per strain) and following growth as described above, each individual culture was 317 used to inoculate 3 tubes of broth (n=6, consisting of n=2 independent biological replicates per 318 strain with n=3 technical replicate for each biological replicate). The cultures were grown until 319 early stationary phase and then 1.5 ml of each culture was centrifuged at 25,000 x g for 3 320 minutes. Then, 1 ml of the cell-free supernatant fraction was collected and stored at  Table 1). The colonic biopsies were 333 processed and cultured as previously described (Supplementary Information). To assess the 334 ability of the CS to suppress IL-8 secretion the organoids were seeded in a 48 well plate and 335 grown for 48 hours. Then, organoids were treated with 10% v/v of select CS in 50% L-WRN 336 conditioned medium and subsequently stimulated with rhIL-1 (50 ng.ml -1 ) for 24 hours before 337 quantifying IL-8 in the supernatant. Cytotoxicity was assessed using the CytoTox 96® Non-338 Radioactive Cytotoxicity Assay. For the animal experiments, colonic tissues from C57BL/6 339 and Winnie (n=2) mice were segmented and the crypts were isolated and cultured according to 340 previously established protocols (Supplementary Information). For the treatments, the 341 organoids were first seeded in a 24 well plate and grown for 48 hours. The organoids were then 342 pre-treated with 10% v/v of select CS for 30 mins and then stimulated with 50 ng/ml mIL-1 343 for 6 hours. The cells were lysed and used for mRNA expression. 344

Peripheral Blood Mononuclear Cell (PBMC) isolation and immunomodulatory assays. 345
Human peripheral blood was obtained for 6 healthy, CD and UC patients from the Mater 346 Significance was determined using a one-way ANOVA with correction for multiple 394 comparisons with a Dunnett test. Differences were considered significant at p≤0.05. A heat 395 map of the first pass screen data was produced using GraphPad Prism (version 7.0) and the 396 Heatmap tool at the HIV sequence database 397 (https://www.hiv.lanl.gov/content/sequence/HEATMAP/heatmap.html).
The animal 398 experiments were performed twice independently, and the data combined for analysis. The 399 D'Agostino-Pearson omnibus test was used to verify the normal distribution of all data. 400 Significance was determined using t-tests, one-way ANOVA with multiple comparisons 401 (Sidak), two-way ANOVA corrected for multiple comparisons with a Dunnett test using 402 GraphPad Prism (version 7.0). Differences were considered significant at p ≤0.05.   respectively; pink nodes represent compounds common to MCM and BHI extracts or MCM 617 and BHI media. Boxes 5Ca, 5Cb, 5Cc, 5Cd and 5Ce highlight structurally related small 618 molecules clusters (i-vi, vii-x, xi-xii, xiii-xiv and xv-xvi respectively), that are unique to the 619 NF-B suppressive EtOAc extracts of C. bolteae AHG0001 cultured in MCM media. Only 620 the molecules within the 5Ca cluster (red dashed box) are present in semi-preparative HPLC 621 fractions that exhibit NF-B suppressive activity.

Supplementary Results
Bioactive identification. Based on the strain and medium effects on C. bolteae NF-B suppressive activity we applied a comparative metabolomics approach to identify the bioactive.
During purification we noted that direct filtering of CS through a 0.42 µm nylon filter resulted in a loss of NF-B suppressive activity, suggestive of either low water solubility and/or a nonpolar bioactive(s). By contrast, an ethyl acetate (EtOAc) extract derived from the same CS was readily filtered, with the filtrate retaining NF-B suppressive activity. As a next step in the chemical characterisation, EtOAc extracts were prepared from C. bolteae AHG0001 and C. Significantly, NF-B suppressive activity was localised in the non-polar fractions #13 and #14, which UPLC-QTOF analysis using single ion extraction (SIE) monitoring confirmed to colocalise with metabolites i-vi. By contrast, UPLC-QTOF-SIE analysis of other compounds present in media-associated clusters ( Figure 5B), revealed they did not uniquely co-localise into the active fractions.

Bacterial comparative analyses. Phylogenetic trees were constructed by aligning the 16S
rRNA gene sequences using the SILVA database 3 and the alignment was then imported into MEGAX 4 . The alignment was refined, and a maximum-likelihood phylogenetic tree constructed displaying the isolate and select reference sequences. The stability of the maximum-likelihood tree was evaluated by 1000 bootstrap replications and Kimura 2parameter modelling. Where necessary, select isolates were subject to whole cell protein profiling to determine intraspecies variations 5, 6 . High molecular weight DNA was prepared as previously described 7 . The SPAdes assembler v 3.11.0 was used to quality check, filter and then de novo assemble the sequence data 8 . CheckM 9 was used to evaluate the genome sequencing quality by estimating the completeness and contamination based on the phylogenetic assignment of a broad set of marker genes. The C. bolteae AHG0001, C.
Candidate BGC were identified using the antiSMASH webserver 12 14 where a Z-factor 0.5 represents an excellent assay, thereby providing a sensitive and specific approach to assess the NF-B suppressive capacity of the isolates. The Z-factor for each assay was determined and only assays achieving a Z-factor 0.5 were processed for further analysis. The high-throughput assays were performed in 96-well microtiter plates as previously described 2 except that the LS174T reporter cells were stimulated with 50 ng.ml -1 TNF and the Caco-2 cell lines were treated with 7.5% v/v CS in complete DMEM medium. NF-κB driven luciferase expression was assessed using the Pierce TM Firefly Luc One-Step Glow Assay Kit (ThermoFisher Scientific) according to the manufacturer's instructions. The NF-κB suppressive isolates were scored and ranked on their Z-score 15,16 .
Organoid culturing and immunomodulatory assays. The colonic biopsies were processed 4 and cultured as previously described 17 . Briefly, the biopsies were washed with PBS and digested with collagenase type I (2 mg.ml -1 ) supplemented with gentamicin (50 µg.ml -1 ) for 15-20 minutes at 37C. The isolated crypts were washed with DMEM/F12 medium and centrifuged at 50 x g for 5 mins at 4C. The pellets were then suspended in Basement Membrane Extract (BME, Invitrogen) in a 1:1 ratio. Then, 20µl of the mixture was plated in a 24 well tissue culture plate and cultured in 50% L-WRN conditioned medium. The crypts were expanded by serial culture until sufficient numbers were obtained for experimentation. To assess the ability of the CS to suppress IL-8 secretion the organoids were seeded in a 48 well plate and grown for 48 hours. Then, organoids were treated with 10% v/v of select CS in 50% L-WRN conditioned medium for 30 min and subsequently stimulated with rhIL-1 (50 ng.ml -1 ) for 24 hours before quantifying IL-8 in the supernatant. Cytotoxicity was assessed using the CytoTox 96® Non-Radioactive Cytotoxicity Assay.
Colonic tissues from C57BL/6 and Winnie mice (n=2) were segmented and the crypts were isolated and cultured. Briefly, the tissues were segmented and washed with PBS, followed by EDTA (8mM) digestion for 1 hour at 4C and further digested with collagenase type I (2 mg.ml -1 ) (Thermo Fisher Scientific) supplemented with gentamicin (50 µg.ml -1 ) for 15-20 minutes at 37C. The isolated crypts were washed with complete F12 medium (Identical to complete media except DMEM/F12 was used instead of DMEM) and centrifuged at 50 x g for 5 mins at 4C. The pellets were then suspended in BME in a 1:1 ratio. Then, 20µl of the mixture was plated in a 24 well tissue culture plate and cultured in 50% L-WRN conditioned medium. The crypts were expanded by serial culture until sufficient numbers were obtained for experimentation.

Peripheral Blood Mononuclear Cell (PBMC) isolation and immunomodulatory assays.
PBMCs were isolated by Ficoll gradient density centrifugation. Briefly, 20 ml of freshly drawn RNA was extracted using the Bioline RNA extraction kit according to manufacturer's instructions. RNA concentration was measured using a Nanodrop 1000 spectrophotometer, followed by cDNA synthesis using 1 µg of RNA and the iScript cDNA synthesis kit (BioRad). Table 2) were analysed using quantitative real time PCR (qRT-PCR) as previously described 24,25 . Ct values were generated, and relative quantitation was determined by the Ct method.