Akkermansia muciniphila exacerbates food allergy in fibre-deprived mice

Abstract

Alterations in the gut microbiome, including diet-driven changes, are linked to the rising prevalence of food allergy. However, little is known about how specific gut bacteria trigger the breakdown of oral tolerance. Here we show that depriving specific-pathogen-free mice of dietary fibre leads to a gut microbiota signature with increases in the mucin-degrading bacterium Akkermansia muciniphila. This signature is associated with intestinal barrier dysfunction, increased expression of type 1 and 2 cytokines and IgE-coated commensals in the colon, which result in an exacerbated allergic reaction to food allergens, ovalbumin and peanut. To demonstrate the causal role of A. muciniphila, we employed a tractable synthetic human gut microbiota in gnotobiotic mice. The presence of A. muciniphila within the microbiota, combined with fibre deprivation, resulted in stronger anti-commensal IgE coating and innate type-2 immune responses, which worsened symptoms of food allergy. Our study provides important insights into how gut microbes can regulate immune pathways of food allergy in a diet-dependent manner.

Fig. 1. Fibre deprivation leads to microbiota-mediated mucosal barrier breakdown.

Mice were fed an FR (green dots) or an FF (red dots) diet for 40 days. a, Differential abundance analysis of taxa in the faeces of FR-fed mice (green bar) versus FF-fed mice (red bar) (n = 10 mice per group, 5 mice per cage, Wald test, P values adjusted using the Benjamini–Hochberg method, features with less than 1 count on average across all samples were excluded). Data are presented as mean ± s.e.m. Taxa in bold depict known/potential mucin-degrading bacteria; for Akkermansia, ~1.78-fold increase in FF- compared with FR-fed mice (change from average relative abundance of 1.7% to 5.7%). b, Representative images of Alcian blue-stained longitudinal colonic sections of mice fed an FR or FF diet for 40 days. Black arrowheads indicate the edges of the mucus layer (n = 5 mice per group) c, Mucus layer thickness measured on colonic section stained with Alcian blue. Each dot is the average of several measurements from one animal (n = 5 mice per group, two-tailed Mann–Whitney test). The average number of measurements per mouse is indicated on the X axis. d, Colonic mucus layer penetrability to 1-μm-sized beads. Each dot is an average of 4–7 measurements from one animal (n = 10 mice per group, two independent experiments). e, Goblet cell counts per crypt. Each dot is the average of multiple measurements from one animal (n = 5 mice per group). f, Intestinal permeability to FITC-dextran (n = 8–9 mice per group, two independent experiments, two-tailed Mann–Whitney test). g–j, Relative transcript levels of Il25 (g), Il33 (h), Ifng (i) and Tnfa (j) mRNA in the colon, caecum and ileum. Expression levels were normalized to the FR group, independently for each tissue (n = 5–10 mice per group, two independent experiments, multiple Mann–Whitney, P values adjusted using the Benjamini–Hochberg method). Each dot represents one mouse. All dot plots are represented with mean ± s.e.m.

Fig. 2. Fibre deprivation induces type-2 immune changes in the intestinal mucosa.

Mice were fed an FR (green dots) or an FF (red dots) diet for 40 d. a, Eosinophil cell frequencies among CD45⁺ single cells in the colonic and ileal lamina propria (n = 7–8 mice per group, two independent experiments, two-way ANOVA, organ effect P = 0.0018, diet effect P = 0.0281). b–e, Frequencies of cytokine-expressing colonic CD4⁺ (b) and CD8⁺ (c) T cells and ileal CD4⁺ (d) and CD8⁺ (e) T cells among CD45⁺ cells (n = 4–5 mice per group, multiple unpaired t-test). f, Serum titres of IgE (n = 9–10 mice per group, two independent experiments, two-way ANOVA). g,h, Frequencies of faecal IgA-coated (g) and IgE-coated (h) bacteria over 5 weeks of feeding on FR or FF diet (n = 1–5 mice per group, two-way ANOVA, P values adjusted using the Benjamini–Hochberg method). i, Frequencies of faecal and ileal IgE-coated bacteria after 40 d of feeding (n = 4–6 per group, multiple Mann–Whitney test, P values adjusted using the Benjamini–Hochberg method). All dot plots are represented with mean ± s.e.m.

Fig. 3. Fibre deprivation worsens food allergic responses.

a, Schematic timeline of OVA sensitization with cholera toxin (CTX) as an adjuvant and control groups. b,c, Blinded symptom scores (b) and core body temperature (c) acquired at OVA challenge (n = 20 mice per group, two independent experiments, Kruskal–Wallis test (b) or two-way ANOVA (c), P values adjusted using the Benjamini–Hochberg method, between FR and FF among OVA-CTX mice). d, Serum titres of OVA-specific IgE and IgG1 (n = 10 mice per group, two-tailed unpaired t-tests). Note that the titres of both OVA-specific IgE and IgG1 for all other groups were below the limit of detection. e, Frequencies of IgE-coated bacteria in the colonic content of mice at the end of the experiment (n = 4 mice per group, two-sided Mann–Whitney test). f,g, Serum titres of total IgE and IgG1 (f) and mouse mast cell protease 1 (g, MCPT1) (n = 10 mice per group, two independent experiments, two-way ANOVA, P values adjusted using the Benjamini–Hochberg method). h,i, Blinded symptom scores (h) and core body temperature (i) acquired at peanut (PN) challenge (n = 20 mice per group, two independent experiments, Kruskal–Wallis test (h) or two-way ANOVA (i), P values adjusted using the Benjamini–Hochberg method, between FR and FF among PN-CTX mice). All dot plots are represented with mean ± s.e.m.

Fig. 4. Fibre deprivation increases colonic inflammation induced by food allergen sensitization.

a, UMAP of the main populations identified using FlowSOM among CD45⁺ cells from the colonic lamina propria of FR- and FF-fed OVA-CTX-sensitized SPF mice 24 h after challenge (n = 5 per group, multiple Mann–Whitney test, *P < 0.05). b–j, Frequencies of indicated cell populations among CD45⁺ cells from the colonic lamina propria of FR-fed (green) and FF-fed (red) mice (n = 5 mice per group, two-way ANOVA, unadjusted P values). k,l, Frequencies of cytokine-expressing CD4⁺ (k) and CD8⁺ (l) T cells among colonic CD45⁺ cells (n = 4–5 mice per group, multiple unpaired t-test, P values adjusted using the Benjamini–Hochberg method). All dot plots are represented with mean ± s.e.m.

Fig. 5. Food allergen sensitization affects the gut microbiota composition in a diet-dependent manner.

a, PCoA plot of microbiome profiles using Bray–Curtis dissimilarity index for FR- or FF-fed mice at the beginning of the feeding period (D1), before sensitization (pre, D40) and after sensitization (post, D96). Inset table: P values from testing for heterogeneity of dispersion (left of diagonal) and distance between group centroids (right of diagonal). Data for D1 and D40 are shared with a previous study. b, Family-level barplots of relative abundance pooled by diet group and time; P < 0.1 indicated with black text colour (n = 4–5, multiple paired t-tests, all comparisons non-significant after adjustment for multiple comparisons using the Benjamini–Hochberg method). c, Heat map of log₂(fold change) of taxonomic features post-challenge (relative to pre-challenge) among FR- or FF-fed mice that were OVA-sensitized or received PBS as a control. Only those taxa with P_adj < 0.05 for at least one group are shown. At the bottom of the heat map, the number of features significantly affected by the challenge are listed for each group. d, Relative abundance of A. muciniphila from the first day of the experiment (D1), from before sensitization (pre) and after sensitization (post) in FR- and FF-fed mice (n = 4–5, two-way ANOVA, individual P < 0.05, non-significant after P values were adjusted using the Benjamini–Hochberg method). e, Concentrations of short- and branched-chain fatty acids in the caecal contents of mice before sensitization (pre) and after sensitization (post) (n = 8–10, two-way ANOVA, P values adjusted using the Benjamini–Hochberg method). All dot plots are represented with mean ± s.e.m.

Fig. 6. A. muciniphila exacerbates type-2 immune responses during food allergy under fibre deprivation in a gnotobiotic mouse model.

a, Relative abundance of microbial strains assessed from pre- and post-sensitization by phylotype-specific qPCR for both 14SM-colonized (left and middle) and 13SM-colonized (right) mice. b,c, Blinded symptom scores (b) and core body temperature (c) acquired after OVA challenge in mice fed an FR (green) or an FF (red) diet (n = 3–9, at least two independent experiments, Kruskal–Wallis test with unadjusted P values, non-significant after adjustment using the Benjamini–Hochberg method (b) or two-way ANOVA (c), with P values adjusted using the Benjamini–Hochberg method between FR vs FF among OVA-CTX mice). d–f, Frequencies of immune cell populations identified with mass cytometry and FlowSOM analysis (d,e, n = 5–10, two independent experiments, two-way ANOVA, non-adjusted individual P values, bolded P values remain below 0.05 after adjustment using the Benjamini–Hochberg method) and of eosinophils as identified by flow cytometry and FlowJo analysis (f, n = 3–7, two-way ANOVA, P values adjusted using the Benjamini–Hochberg method) in the colonic lamina propria of mice fed an FR or an FF diet. g, Frequencies of faecal IgE-coated bacteria at the end of the experiment (n = 4–8, two independent experiments, two-way ANOVA, P values adjusted using the Benjamini–Hochberg method). All dot plots are represented with mean ± s.e.m.

Extended Data Fig. 1

a, Representative immunofluorescence images of distal colon sections stained with α-Muc2 (green) and DAPI (blue). White arrow heads indicate the edges of the mucus layer. b, Mucus thickness measured on colonic section stained with α-Muc2. Each dot is the average of several measurements from one animal (n = 5 mice per group, two-tailed Mann-Whitney test). c, Representative images of distal colonic tissues showing 1-μm-sized beads (red dots) laying on top of the epithelium (green dots) using an ex situ ex vivo method. The space between the beads and the epithelium exemplify the unpenetrable mucus layer. d, Analysis of the images presented in b showing the frequency of beads as a function of the distance of the epithelium for the FR-fed (green line) and the FF-fed (red line) sample. The penetration of the beads into the mucus is represented and quantified as the area under the curve to the peak (colored area under the curve). e, Colonic mucus thickness measured ex vivo as the distance of 1-μm-sized beads peak from the epithelium (as shown in d). Each dot is the average of 4–7 measurements from one animal (n = 10 mice per group, two-tailed Mann-Whitney test). f, Relative transcript levels of Muc2, Tslp, Il5, Il22, Il17a and Il17f in the colon, cecum, and ileum of SPF mice fed a FR (green) or a FF (red) diet. Expression levels were normalized to the FR group, independently for each tissue (n = 5–10 mice per group, two independent experiments, multiple Mann–Whitney test, P values adjusted using the Benjamini-Hochberg method). g, Fecal lipocalin-2 (LCN-2) measured at the beginning (D1) and at the end (D40) of the 40-day feeding period (n = 9–10). h, Relative transcript levels of selected cytokine mRNA in the colon, cecum and ileum of germfree mice. Transcript levels were normalized to the FR group, independently for each tissue. Numbers above dot plots represent the number of samples in which transcripts were not detected (n.d.) (n = 3 mice per group). All dot plots are represented with mean +/− SEM.

Extended Data Fig. 2

a, Gating strategy for eosinophils, defined as Siglec-F⁺CD11b⁺ singlets. b, Gating strategy for cytokine-expressing CD4⁺ and CD8⁺ T lymphocytes. c, Serum titers of IgG1 and mouse mucosal mast cell protease 1 (MCPT1) (n = 2–10, two-way ANOVA, P values adjusted using the Benjamini-Hochberg method). d, Representative plots for IgE-coated bacteria in fiber-rich and fiber-free fed mice. Each sample was stained with an α-IgE antibody (red) or an isotype control (blue).

Extended Data Fig. 3

Weight curves of the mice over the 40-day feeding period and the sensitization period to OVA-CTX (n = 5–10, mixed-effects model with matched values per mice, time effect P < 0.0001).

Extended Data Fig. 4

a, Schematic timeline for peanut protein (PN) sensitization with cholera toxin (CTX) as an adjuvant and control groups. Samples were collected 3 h after the challenge. b, Weight curves of the mice over the 40-day feeding period and the sensitization period to PN-CTX (n = 10, two independent experiments, mixed-effects model with matched values per mice, time effect P < 0.0001). c–g, Serum titers of PN-specific IgE (c, n = 9–10), PN-specific IgG1 (d, n = 9–10), total IgE (e, n = 5–10), total IgG1 (f, n = 5–10), and mouse mast cell protease 1 (g, MCPT1, n = 10). Two-way ANOVA, P values adjusted using the Benjamini-Hochberg method. Fiber-rich (FR, green), fiber-free (FF, red). All dot plots are represented with mean +/− SEM.

Extended Data Fig. 5

Heatmap of the 58 annotated clusters identified by FlowSOM from CyTOF analysis of colonic lamina propria cells of SPF mice 24 h after challenge. Numbers in the heatmap indicate the relative expression of each marker among cell populations. Percentages shown in parantheses represent frequencies of populations among CD45⁺ cells.

Extended Data Fig. 6

Relative abundances of indicated bacterial strains from feces of FR-fed (green) and FF-fed (red) mice, at the end of the experiment (n = 3–6, two-way ANOVA, P values adjusted using the Benjamini-Hochberg method). All dot plots are represented with mean +/− SEM.

Extended Data Fig. 7

a, Weight curves of the mice over the sensitization period to OVA-CTX (n = 3–9, two-way ANOVA with matched values per mice, time effect P < 0.0001). b–f, Serum titers of MCPT1 (b), OVA-specific IgE (c), OVA-specific IgG1 (d), total IgE (e), and total IgG1 (f) (n = 4–10 mice per group, two independent experiments, two-way ANOVA, P values adjusted using the Benjamini-Hochberg method). Note that the serum titers of OVA-specific IgE and OVA-specific IgG1 were below the limit of detection in other groups. g, Uniform manifold approximation and projection (UMAP) of specific populations identified using FlowSOM among CD45⁺ cells from the colonic lamina propria of OVA-CTX sensitized, 14SM- and 13SM-colonized mice 24 h after challenge. Highlighted populations are the ones showing differences between groups. h–k, Frequencies of indicated cell populations among cLP CD45⁺ cells (n = 5–10, two-way ANOVA, unadjusted P values, non-significant after adjustment using the Benjamini-Hochberg method).

Extended Data Fig. 8

Heatmap of the 30 annotated clusters identified by FlowSOM from CyTOF analysis of colonic lamina propria cells of GF and gnotobiotic OVA-sensitized mice 24 h after challenge. Numbers in the heatmap indicate the relative expression of each marker among cell populations. Percentages shown in parentheses represent frequencies of populations among CD45⁺ cells. All dot plots are represented with mean +/− SEM.

Extended Data Fig. 9

a, A figure panel from previously published data by Steimle et al., showing proportions of indicated cell populations among CD45⁺ cells in the colonic lamina propria of mice colonized with 3SM (B. caccae, B. thetaiotaomicron, B. intestinihominis), 4SM (3SM + A. muciniphila), 13SM (3SM + 10 non-mucin-degrading bacteria) or 14SM (13SM + A. muciniphila) (n = 4–5, one-way ANOVA, P values adjusted using the Benjamini-Hochberg method) b, Enzymatic activity of select carbohydrate-active enzymes (CAZymes) and sulfatase measured in the feces of OVA-CTX-sensitized, 14SM- and 13SM-colonized mice (n = 4–7, multiple Mann-Whitney, P values adjusted using the Benjamini-Hochberg method). Fiber-rich (green), fiber-free (red). All dot plots are represented with mean +/− SEM.

Extended Data Fig. 10

Proposed model of colonic immune pathways during an allergic response under a fiber-rich (left) versus a fiber-free (right) diet. Allergic type-2 responses at the intestinal barrier are initiated by the sensing of the allergen and the release of epithelial-derived cytokines IL-25, IL-33 and TSLP that promote the recruitment and activation of type 2 immune cells Th2, M2 macrophages and ILC2, and lead to the production of IgE. Under a fiber-rich diet, Treg cells are maintained in response to microbiota-derived SCFAs, and Akkermansia muciniphila promotes Gata3⁺ Treg cells as well as Th2 cells. In this context, a conventional allergic response is driven by IgE-primed mast cell degranulation. By contrast, fiber deprivation leads to a non-canonical allergy response driven by the altered microbiota composition and the increased intestinal permeability, which altogether predisposes to a mixed inflammatory environment comprising of innate type 2 cells, M2 and ILC2, and type 1 cells, Th1, NK and CD8⁺ T cells. In this context, an allergic response is characterized by the production of IL-5 that is reinforced by CD8⁺ T cells and mediates eosinophilia. This chain of events is paralleled by higher coating of colonic bacteria by IgE, which has an unknown role in the pathology. The fiber deprivation-induced mucus changes and increased intestinal permeability are likely to facilitate the microbiota-driven type 1 responses, but further studies are needed to determine their contribution to the innate type 2 response that is promoted by A. muciniphila. Finally, the role of intestinal IgEs and their antigenic specificity in the eosinophilic reaction remain an important question to adress in the future. Created with BioRender.com.