Society: AGA

The microbiome plays a central role in intestinal inflammation in IBD. It is influenced by environmental factors included diet. In turn, the microbiome can incite intestinal inflammation

Background Mounting evidence suggests that food additives, especially emulsifier polysorbate-80 (P80), promote colitis via gut microbiota. Here, we explored the underlying mechanism of P80 and specific species Proteus mirabilis (P. mirabilis), a pathobiont in gut, in pathogenesis and aggravation of intestinal inflammation.

Methods This study includes two independent cohorts comprised of Crohn’s disease (CD) and healthy controls, respectively Hong Kong (92 CD, 21 controls) and Australian (98 CD, 21 controls). Previous and current consumption of P80 was quantified using a validated 26-item food additive questionnaire. Fecal microbiota composition was assessed by shotgun metagenomic sequencing. The interaction between P80 and P. mirabilis was investigated in both in vitro and in vivo models. The effect of P80 on P. mirabilis was determined by bacterial transcriptome sequencing. The interested protein was identified by silver straining followed by mass spectrometry.

Results P. mirabilis was identified as the enriched microbiota in the CD patients with high food additive consumption when compared to controls in both Hong Kong and Australia cohorts. Co-cultivation of P. mirabilis and P80 induced more cell death on human epithelial cells lines, INT 407 and NCM 460, compared to P. mirabilis alone or Escherichia coli control groups (Figure 1A). P80 exposure increased expression of virulence genes in P. mirabilis by RNA-seq. The co-cultured conditional medium of P. mirabilis and P80 (Plus.CM) triggered cell death, which was attenuated by Proteinase K, implying that the secreted protein is the functional participant in this process (Figure 1B). Compared to other fractions, the proteins with molecular weight over 100kDa exhibited the most notable function. By proteomics identification, P. mirabilis-derived enzyme X was uncovered as the potential deleterious protein in arising and aggravating inflammation (Figure 1C). In mouse models, the mice orally challenged with P. mirabilis and fed with P80 diet had shortened colon length (Figure 2A) and higher Lipocalin-2 (LCN2) levels (Figure 2B) compared to P. mirabilis solely and control groups.

Conclusion We have explored that with the addition of P80, P. mirabilis exacerbated the inflammation both in vitro and in vivo. The enzyme X, derived from P. mirabilis, was identified being the possible deleterious participant in this process and may be targeted for future precision therapy.

This work is supported by the Croucher Senior Research Fellowship and The Leona M. and Harry B. Helmsley Charitable Trust.

$<b>Figure 1. (A)</b> Exposure to both <i>P. mirabilis </i>and P80 induced more cell death, in comparison with <i>P. mirabilis</i> alone or other control groups. <b>(B)</b> The Proteinase K treatment attenuated the function of co-cultured conditional medium in triggering cell death. <b>(C) </b>Silver staining of protein fractions above 100kDa indicated the targeted enzyme. Statistical significance was determined by 1-way or 2-way analysis of variance appropriately. ****<i>P </i>< .0001.$

Figure 1. (A) Exposure to both P. mirabilis and P80 induced more cell death, in comparison with P. mirabilis alone or other control groups. (B) The Proteinase K treatment attenuated the function of co-cultured conditional medium in triggering cell death. (C) Silver staining of protein fractions above 100kDa indicated the targeted enzyme. Statistical significance was determined by 1-way or 2-way analysis of variance appropriately. ****P < .0001.

<b>Figure 2. </b>Mice challenged with <i>P. mirabilis</i> and P80 diet <b>(A)</b> shortened colon length and <b>(B)</b> induced higher level of Lcn-2 expression. Statistical significance was determined by 1-way or 2-way analysis of variance appropriately. *<i>P</i> < .05.

Figure 2. Mice challenged with P. mirabilis and P80 diet (A) shortened colon length and (B) induced higher level of Lcn-2 expression. Statistical significance was determined by 1-way or 2-way analysis of variance appropriately. *P < .05.

Increasing evidence suggest that environmental factors, including the gut microbiome, are important in the pathogenesis of inflammatory bowel disease in genetically predisposed individuals. The Winnie mouse is a model of spontaneous chronic colitis due to a point mutation in the Muc2 gene that disrupts the colonic mucus layer, hence, increasing the permeability to microbes in conjunction with dysbiosis. The aim of this study was to test the hypothesis that the environment in which these mice are reared might be a significant modifier of colitis severity when Muc2 function is abnormal. To test this hypothesis, we aseptically rederived Winnie mice from a conventional (Conv) animal facility in Italy into our specific pathogen free (SPF) vivarium at CWRU in Cleveland and performed a time-course of colitis severity in both colonies. In addition, we carried out microbiome and metabolomic analyses to investigate potential mechanistic differences in the two colonies. Surprisingly, Winnie mice housed in our SPF animal facility developed more severe colitis compared to the Conv facility. In addition, SPF Winnie mice developed colon tumors starting at 8 weeks of age in the proximal and medial part of the colon, while Conv housed Winnie mice developed only mild inflammation. Winnie mice raised under germ-free conditions showed no signs of colitis up to 40 weeks of age. We then analyzed the stool microbiome and metabolome of these mice. We found many differentially represented bacterial genera between SPF and conventional Winnie mice. Anaerosporobacter, Dubosiella,Faecalibaculum, Mucispirillum and Prevotellaceae_NK3B31 were more frequent in SPF Winnie mice while Rikenellaand Prevotellaceae_UCG-001 were more represented in Conv mice. Metabolomic analyses showed marked differences between the two colonies. Metabolites related to purine and nitrogen recycling were significantly increased in SPF Winnie mice compared to Conv Winnie mice. Of great interest, analysis of lipidomes revealed an increased presence of sphingomyelins and decreased amounts of ceramides in the SPF group compared to Conv Winnie mice. This class of lipids shares important roles in the onset and development of intestinal inflammation and interact with bacterial species to cause severe dysbiosis, inflammation and tumorigenesis. Further characterization through fecal material transplantation (FMT) experiments and culturomic studies are in progress to better clarify the role of specific bacterial strains and their metabolites in the development of chronic colitis and colitis-associated cancer in this model.

Background: Variation in clinical response to 5-aminosalicylic acid (5-ASA) has been attributed to its inactivation by gut microbes. Recently, in the Inflammatory Bowel Disease (IBD) Multi’omics Database (IBDMDB), a cohort of 105 participants with IBD, we identified 12 gut microbial enzymes from two protein families that convert 5-ASA to N-acetyl 5-ASA, a compound that lacks anti-inflammatory effects. Within the IBDMDB, a subset of these enzymes was cross-sectionally linked with greater risk of treatment failure, defined by corticosteroid use. The aim of this study was to prospectively associate these drug-degrading gut microbial enzymes with treatment failure and to characterize in vitro conversion of 5-ASA by these enzymes.

Methods: We examined the association between gut microbial acetyltransferases with risk of 5-ASA treatment failure in the “Study of a Prospective Adult Research Cohort with IBD” (SPARC IBD). We included 208 participants who at baseline 1) gave a stool sample, 2) were on 5-ASA, and 3) were steroid-free. Fecal metagenomic data was processed by biobakery3. Exposure was defined as metagenomic carriage of 3-4 drug-degrading acetyltransferases compared to 0-2 acetyltransferases, as in the IBDMDB. We calculated odds ratios (ORs) and 95% confidence intervals (CIs) of incident steroid use via multivariable generalized estimating equations, adjusting for age and sex. For biochemical characterization, the 12 enzymes identified in the IBDMDB were heterologously expressed by E. coli. A representative enzyme was selected from each protein family for purification, and then incubated with 5-ASA and acetyl CoA for 6 hr at 37C. Using a custom LC–MS assay, we detected N-acetyl 5-ASA production. Finally, to gain mechanistic insight into how these enzymes convert 5-ASA, we made efforts to crystallize the two purified proteins.

Results: Over a median follow-up of 8 months, we identified 60 cases of corticosteroid use. Consistent with the IBDMDB, we found that metagenomic carriage of 3 or more microbial acetyltransferase genes in SPARC IBD (compared to 2 or fewer) was associated with treatment failure (OR 2.77, 95% CI 1.03-7.43). In a meta-analysis of the two cohorts, we observed a three-fold increased risk of drug failure (OR 3.12, 95% CI 1.41-6.89) (Fig 1). In vitro experiments confirmed the ability of a thiolase and an acyl-CoA N-acetyltransferase to acetylate 5-ASA, with >25% conversion (Fig 2). Finally, we generated an acetylated unliganded crystal structure of the thiolase (1.9Å), and found a catalytic cysteine residue poised to acetylate an incoming substrate.

Conclusions: We characterized two gut microbial protein families that are directly involved in 5-ASA metabolism and, in turn, are prospectively associated with 5-ASA treatment failure. These findings advance the possibility of microbiome-based personalized medicine for patients with IBD.

Introduction
For patients with refractory Crohn’s disease (CD) autologous stem cell transplant (auto-SCT) can induce disease remission in the majority of patients.^1,2 The therapeutic mechanism of auto-SCT is unknown though it is believed to induce an immune “reset”. The impact of auto-SCT on the CD microbiome composition and function has never been studied.

Methods
14 patients with CD were enrolled in a Phase IIa study (NCT03219359). Blood, intestine and stool samples were taken prior to transplant and at intervals post-transplant. Fresh leukocytes were isolated and analyzed by mass cytometry (CyTOF). Microbiome composition was analyzed by 16S sequencing. Patient fecal slurries were inoculated by oral gavage into 6 wk gnotobiotic C57BL/6 mice (M/F). At 4 wks, mouse intestinal leukocytes were analyzed by flow cytometry (Aurora).

Results
Microbial community analysis revealed a significant decrease in diversity post-transplant at hospital discharge followed by a restoration of diversity to baseline by 3 months post-transplant. (Fig 1A) Principal component analysis failed to resolve population differences at any time point during auto-SCT. (Fig 1B) While microbial population statistics did not differ 6 months post-transplant, comparison of microbial composition suggested several changes.(Fig 1C) To understand the magnitude of microbial compositional changes we compared CD patients who underwent auto-SCT to patients treated with ustekinumab.³ This analysis suggested a much more significant change in species composition post auto-SCT compared to treatment with ustekinumab. (Fig 1D) The significant decrease in diversity at discharge suggests long-term changes in microbial composition may be facilitated by damage to microbiome structure during transplant.⁴ Comparing microbial composition at discharge to 6 months confirms new organisms are acquired after discharge.(Fig 1E) Analysis of blood and intestinal immune populations post auto-SCT indicates the most significant changes occur among intestinal CD14⁺ populations.(Fig 2A) Using gnotobiotic mice we studied whether microbial functions post auto-SCT may reinforce changes in immune cell populations. In one patient where there was an increase in intestinal CD14⁺ populations post-transplant we observed a similar increase in Ly6C⁺ populations in mice colonized with stool taken at 6 months post-transplant compared to colonization with stool taken at baseline (Fig 2B, pt CD021). We did not observe a similar increase in mice colonized with patient stool where there was no increase in CD14⁺ populations (Fig 2B, pt CD015).

Conclusion
Our initial analysis of microbiome function and diversity during auto-SCT suggests this treatment modality creates a uniquely fertile ground where microbial composition can be “reset” and the function of these new microbes may reinforce changes in the mucosal immune system.

<b>Fig 1. </b>Bacterial DNA was extracted and the V4 region of the 16S rRNA gene amplified and sequenced (Illumina MiSeq). Amplicon sequence variants (ASV) were assigned and classified using DADA2. <b>A, </b>Mean (95% CI) Shannon diversity indices are plotted. Comparison by ANOVA with Tukey post hoc. <b>B,</b> Principal component analysis. <b>C, </b>Relative abundance of bacterial composition (order) at baseline (B) and 6 months post-transplant (6). Patients ID CDXXX. <b>D, </b>ASV were compared at baseline to follow up in patients undergoing auto-SCT or treatment with ustekinumab. Percent of ASV present at baseline and found in follow up relative to the total number of ASV at follow up are plotted (mean +/- SEM). Samples are compared by Mann Whitney test (p = 0.0005). <b>E, </b>Percent of ASV overlap at different timepoints are plotted demonstrating that the ASV present at discharge post stem cell transplant are not derived from baseline microbiota nor do they influence the post-transplant populations.

Fig 1. Bacterial DNA was extracted and the V4 region of the 16S rRNA gene amplified and sequenced (Illumina MiSeq). Amplicon sequence variants (ASV) were assigned and classified using DADA2. A, Mean (95% CI) Shannon diversity indices are plotted. Comparison by ANOVA with Tukey post hoc. B, Principal component analysis. C, Relative abundance of bacterial composition (order) at baseline (B) and 6 months post-transplant (6). Patients ID CDXXX. D, ASV were compared at baseline to follow up in patients undergoing auto-SCT or treatment with ustekinumab. Percent of ASV present at baseline and found in follow up relative to the total number of ASV at follow up are plotted (mean +/- SEM). Samples are compared by Mann Whitney test (p = 0.0005). E, Percent of ASV overlap at different timepoints are plotted demonstrating that the ASV present at discharge post stem cell transplant are not derived from baseline microbiota nor do they influence the post-transplant populations.

<b>Fig 2</b>. <b>A, </b>Human intestinal leukocytes are analyzed by CyTOF. CD14<sup>+</sup>CD3<sup>-</sup>CD19<sup>-</sup>HLADR<sup>+</sup> cells are plotted as a percent of live singlets (mean +/- SEM). Samples compared by Mann Whitney test (p = 0.0006). <b>B, </b>6 week germ free mice are colonized with stool samples from 4 patients (8 mice per group; 4M, 4F). Mice are colonized with stools from baseline (B) or 6 month follow up (6). Intestinal leukocytes are isolated and analyzed by flow cytometry (Aurora). B220<sup>-</sup>MHCII<sup>-</sup>Ly6C<sup>+</sup> cells are plotted as a percent of live CD45<sup>+</sup> cells. Samples are compared by Mann Whitney test (p = 0.04 for CD021).

Fig 2. A, Human intestinal leukocytes are analyzed by CyTOF. CD14⁺CD3^-CD19^-HLADR⁺ cells are plotted as a percent of live singlets (mean +/- SEM). Samples compared by Mann Whitney test (p = 0.0006). B, 6 week germ free mice are colonized with stool samples from 4 patients (8 mice per group; 4M, 4F). Mice are colonized with stools from baseline (B) or 6 month follow up (6). Intestinal leukocytes are isolated and analyzed by flow cytometry (Aurora). B220^-MHCII^-Ly6C⁺ cells are plotted as a percent of live CD45⁺ cells. Samples are compared by Mann Whitney test (p = 0.04 for CD021).

Background: E. coli produces hydrogen sulfide (H₂S) that degrades mucus and inhibits mitochondrial butyrate oxidation, causing colonic epithelial cell dysfunction. Patients with active IBD, especially Crohn’s disease (CD), have higher prevalence of Ruminococcus gnavus, a mucolytic gut bacteria, and Escherichia coli vs. healthy controls. Functionally aggressive adherent-invasive E. coli (AIEC) are increased in ileal CD. Reciprocal functional interactions between AIEC may enhance inflammation, but interactions between AIEC and R. gnavus remain unclear.
Hypothesis: Substrates released by R. gnavus mucin degradation stimulates growth and H₂S production by an ileal CD patient AIEC isolate, E. coli LF82. H₂S promotes inflammation, LF82 virulence and R. gnavus persistence.
Methods: In vitro: We added porcine gastric mucin (PGM, Sigma), ileal (MIM) or colonic (MCoM) mucus isolated from germ-free (GF) mice to M9 minimal medium. We compared LF82 growth and H₂S production following culture with sterile-filtered supernatants of each mucin/mucus medium precultured with/without R. gnavus and R. gnavus growth in M9 minimal medium supplemented with several mucus-derived LF82 sulfur products. LF82 exposed to Na₂S, a H₂S donor, was cultured with Caco-2 cells to measure LF82 adhesion and invasion and LF82 gene expression.
In vivo: We colonized GF IL-10^-/- 129 mice for 6 wks with either R. gnavus or LF82 alone (mono) or in combination (dual) (N=6/group). We measured weekly fecal lipocalin-2 (fLcn2) by ELISA, cecal H₂S content by silver sulfide colorimetric assay and LF82 spatial ileal and colonic mucus-/tissue-association or invasion relative to luminal colonization in LF82 mono- or dual-colonized mice.
Results: In vitro (Fig. 1): LF82 grew better and produced more H₂S with R. gnavus precultured mucus, particularly ileal mucus, than with mucus alone. All mucus- relevant sulfur substrates supported R. gnavus survival for 24 hours, while Na₂S–produced H₂S sustained R. gnavus for 48 hours. H₂S exposure enhanced LF82 adherence and invasion, and adherent-related gene expression.
In vivo (Fig. 2): fLcn2 values >4 wks colonization, colonic histology scores and H₂S levels were higher in dual-colonized mice than in either mono-colonized group. Histology scores correlated with H₂S concentrations. LF82 ileal mucosal invasion but not mucus-/mucosal-association were higher in dual-colonized mice than in the colon.
Conclusions: R. gnavus supports LF82 growth and H₂S production through mucus degradation substrates. H₂S enhances LF82 adherence/invasion and promotes R. gnavus survival. Our in vitro and in vivo studies show preference LF82 in ileum vs. colon, perhaps explaining why ileal CD patients have highest AIEC frequency. LF82/R. gnavus interactions enhance H₂S production and colitis in dual-associated mice, suggesting that coexistent R. gnavus and AIEC contribute to CD activity.

Figure 1

Figure 2

Background: While strong evidence links the inflammatory bowel diseases (IBD) to gut taxonomic alterations, a longitudinal survey of the interaction between host diet, community-wide functional activity (metatranscription), and disease type has not been conducted.
Methods: We assessed metagenomic (MGX) and metatranscriptomic (MTX) data from 105 densely-phenotyped participants with IBD (n=38 with ulcerative colitis [UC], 67 with Crohn’s disease [CD]) and 27 non-IBD controls from the Integrative Human Microbiome Project. Collectively, they provided 1,246 stool metagenomes and 667 stool metatranscriptomes over one year, each with contemporaneous assessment of short-term diet (i.e., prior week) using an abbreviated food frequency questionnaire. These data, paired with baseline assessment of long-term diet and numerous disease severity indices, were used to construct integrated time-series profiles. Major food groups were assembled into higher order dietary patterns. We employed a novel multivariable framework to longitudinally link diet to changes in gut metatranscription.
Results: In addition to expected alterations in overall gut community structure, we found perturbed expression of microbial pathways in IBD cases compared to controls, ranging from housekeeping pathways, such as nucleotide biosynthesis, to shifts in fermentative and oxidative metabolism (Fig. 1A). Interestingly, while prior studies have reported decreases in pantothenic acid levels and vitamin B-related gene carriage in IBD (Lloyd-Price et al, Nature 2019, Franzosa et al, Nat Microbiol 2019), we observed increases in their expression in both CD and UC vs. non-IBD, indicating a divergence of functional potential (i.e., gene carriage) and microbial transcription of pathways related to gut barrier integrity. Surprisingly, glutaryl-CoA degradation/mevalonate pathways previously linked to colitis differed slightly but significantly between CD and UC, unusual inasmuch as the two diseases share many microbial features (Fig. 1A). Similarly, we also found significant heterogeneity in the relationship between short-term diet and microbially-mediated transcription by disease type with stronger coupling in UC (310 significant diet-by-pathway associations) compared to CD (72) and non-IBD (45; all pFDR<0.05) with consistent changes in microbial RNA along a healthy vs. unhealthy dietary axis (Fig. 1B). Long-term diet in the year preceding enrollment was less tightly coupled to MTX profiles (not shown).
Conclusion: Our results further demonstrate a complex relationship between host diet, gut ecology, and intestinal inflammation, distinguishing between longer-term effects in microbial growth (MGX) and shorter-term changes in transcription (MTX). A deeper exploration of xenobiotic-induced alterations of gut microbial activity may identify novel dietary strategies that may be distinct between CD and UC.

<br /> <b><b>Differentially expressed microbial functions in inflammatory bowel disease (IBD). (A) Differential expression of detected microbial metabolic pathways in IBD patients and non-IBD controls. (B) Heterogeneous microbial transcription by dietary intake. </b></b>In every model, we adjusted for age (years), antibiotic usage (yes/no), microbial dysbiosis (Lloyd-Price <i>et al</i>, Nature 2019), and gene copy number variation (Zhang <i>et al</i>, Bioinformatics 2021) and accounted for repeated measures and sites where data were collected and processed. Symbols indicate <i>p</i>FDR: *** = <i>p</i>FDR ≤ 0.001, ** = 0.001 <<i> p</i>FDR ≤ 0.01, * = 0.01 <<i> p</i>FDR ≤ 0.05, ‘ = 0.05 <<i> p</i>FDR ≤ 0.25 and cells shaded by -log[nominal <i>p</i>-value]*sign[β-coefficient]. Boxplots show only selected pathways with <i>p</i>FDR≤0.05. Heatmaps only display MetaCyc pathways (rows) and dietary factors (columns) with >5 significant associations, clustered by Euclidean distance.

Differentially expressed microbial functions in inflammatory bowel disease (IBD). (A) Differential expression of detected microbial metabolic pathways in IBD patients and non-IBD controls. (B) Heterogeneous microbial transcription by dietary intake. In every model, we adjusted for age (years), antibiotic usage (yes/no), microbial dysbiosis (Lloyd-Price et al, Nature 2019), and gene copy number variation (Zhang et al, Bioinformatics 2021) and accounted for repeated measures and sites where data were collected and processed. Symbols indicate pFDR: *** = pFDR ≤ 0.001, ** = 0.001 < pFDR ≤ 0.01, * = 0.01 < pFDR ≤ 0.05, ‘ = 0.05 < pFDR ≤ 0.25 and cells shaded by -log[nominal p-value]*sign[β-coefficient]. Boxplots show only selected pathways with pFDR≤0.05. Heatmaps only display MetaCyc pathways (rows) and dietary factors (columns) with >5 significant associations, clustered by Euclidean distance.

Ashwin Ananthakrishnan

Massachusetts General Hospital

Appearances