Epithelial IFNγ signalling and compartmentalized antigen presentation orchestrate gut immunity

Abstract

All nucleated cells express major histocompatibility complex I and interferon-γ (IFNγ) receptor¹, but an epithelial cell-specific function of IFNγ signalling or antigen presentation by means of major histocompatibility complex I has not been explored. We show here that on sensing IFNγ, colonic epithelial cells productively present pathogen and self-derived antigens to cognate intra-epithelial T cells, which are critically located at the epithelial barrier. Antigen presentation by the epithelial cells confers extracellular ATPase expression in cognate intra-epithelial T cells, which limits the accumulation of extracellular adenosine triphosphate and consequent activation of the NLRP3 inflammasome in tissue macrophages. By contrast, antigen presentation by the tissue macrophages alongside inflammasome-associated interleukin-1α and interleukin-1β production promotes a pathogenic transformation of CD4⁺ T cells into granulocyte-macrophage colony-stimulating-factor (GM-CSF)-producing T cells in vivo, which promotes colitis and colorectal cancer. Taken together, our study unravels critical checkpoints requiring IFNγ sensing and antigen presentation by epithelial cells that control the development of pathogenic CD4⁺ T cell responses in vivo.

Extended Data Fig. 1 |. Progressive increase in IFN-γ signaling and antigen presentation by IECs during colitis.

C57BL/6 J mice were infected with Citrobacter rodentium. (A) Citrobacter colonization in the colon (n = 5 per group) and (B) quantification of Ifng transcript in colon tissue by RT-PCR (n = 4 for d0 and d21, n = 6 for d6 and n = 5 for d12). STAT1 pTyr701 analyses in IECs by (C) immunoblotting, representative of two independent experiments and (D) flow cytometry (n = 3 for d0, n = 6 for d6 and n = 5 for d12 and n = 4 for d21). (E) Quantification of Irf1 transcript in purified IECs (EpCAM⁺) by RT-PCR (n = 4 for d0, d6 and d21, n = 5 for d12). Flow cytometric analyses for (F and G) MHCI (n = 4 for d0, n = 5 for d6 and d12) and (H and I) MHCII (n = 4 for d0, n = 5 for d6 and d12) in the IECs from colons of mice at indicated day post-infection. (J) Colon histology score in Ifngr1^fl/fl and Ifngr1^fl/fl Vil^Cre mice. The horizontal bar represents the median, and each symbol and each lane in the immunoblot represents an individual mouse. Data were analyzed by (A, B, D, E, G and I) Brown-Forsthye and Welsch ANOVA tests followed by the Dunnet’s post-hoc test.

Extended Data Fig. 2 |. Epithelial sensing of IFN-γ orchestrates T cell responses during colitis.

(A-I) Ifngr1^fl/fl and Ifngr1^fl/fl Vil^Cre mice were infected with Citrobacter rodentium (n = 5 per group). (A-B) Enumeration of I.E. CD4⁺ T cells and (C-G) evaluation of production of indicated factors by them at indicated days post-infection. Enumeration of (H-L) indicated cells in the epithelium. (M) Comparison of GM-CSF production from ILCs and CD4⁺ T cells (n = 7 per group) and (N) GM-CSF MFI in ILCs and CD4⁺ T cells in epithelium of Ifngr1^fl/fl Vil^Cre mice at day 12 p.i. (n = 7 per group). (O) The indicated factors were gated globally on the live cell gate and then analyzed for the contribution of CD8αβ⁺ or CD4⁺ colonic IE-T cells in their production. Each symbol represents an individual mouse, and the bar indicates the median. Data were analyzed by (A-L) ANOVA followed by the Holm–Sidak post-hoc test or (M-N) Mann-Whitney U test (two-sided). MFI, median fluorescence intensity.

Extended Data Fig. 3 |. Generation of the ovalbumin-expressing strain of C. rodentium and characterization of ova-specific IE-T cells.

Ovalbumin coding sequence was cloned into the NocI site of pOBX18 plasmid containing the artificial and constitutive Enterobacteriaceae-specific promoter and terminator sequence into the temperature-sensitive transposase 7 attachment site (attTn7) targeting vector (vector described in) (A) plasmid map and (B) gel picture of the transformed vector. Then, this vector was electroporated into C. rodentium. Growth at permissive temperature along with constitutive expression of the recombination machinery led to the integration of the promoter-ova-terminator sequence into the attTn7 site. Next, the plasmid was cured by subsequent growth at non-permissive temperatures. Site-specific integration was confirmed by (C) PCR and sequencing from the neighboring glmS region of the attTntn7 locus and the expression of ova was confirmed by (D) RT-PCR, representative of two independent clones. (E) Ifngr1^fl/fl and Ifngr1 ^fl/fl Vil^Cre mice were infected with ova⁺ Citrobacter and (F) Ifngr ^fl/fl Vil^creERT2 Ova^Tg and Ifngr ^fll+ Vil^creERT2 Ova^Tg mice were infected with wild-type Citrobacter and administered tamoxifen on days 3–6 p.i and evaluated for IFN-γ production from the ova-specific CD4⁺ IE-T cells (n = 4 per group). (G) Flow cytometric phenotyping of ova-specific CD4⁺ IE-T cells at day 12 post-infection with ova⁺ Citrobacter. (H-K) Ifngr1^fl/fl, Ifngr1^fl/fl Vil^CreERT2, and Ifngr1^fl/fl Vil^CreERT2 R26^{LSL Irf1} mice were infected with ova⁺ Citrobacter rodentium on day 0, injected with tamoxifen at days 3–6 post-infection, and analyzed on day 12 for CD4⁺ IE-T cells and (I) Citrobacter colonization in the colon (n = 5 per group). (L) Ifngr1^fl/fl and Ifngr1 ^fl/fl Vil^Cre mice were infected with Citrobacter and treated with isotype or anti-CD4 antibody every 2 days starting at day 7 analyzed for CD4⁺ T cells on day 12 p.i. (n = 8 per group). The horizontal bar represents the median, and each symbol represents an individual mouse. Data were analyzed by (E and F) Mann-Whitney U test (two-sided) or (H-L) Kruskal-Wallis test followed by Dunn’s post-test.

Extended Data Fig. 4 |. Epithelial sensing of IFN-γ restrains NLRP3 inflammasome activation in macrophages.

(A-K) Ifngr1^fl/fl and Ifngr1^fl/fl Vil^Cre mice were infected with Citrobacter rodentium. (A and B) Evaluation of indicated cytokines in colon explants (n = 4 for d0, d6 and n = 8 for d12, per group) and (C) immunoblotting for caspase-1 p10 in colon lysates, representative of two independent experiments. (D) Evaluation of IL-1α in colon explants, (E-F) Colon histology analyses and (G) Citrobacter colonization after isotype (n = 8), αβ-ATP (n = 8), apyrase (n = 6), MCC950 (n = 6) or anti-IL1R (n = 6) treatment. (H-K) Ifngr1^fl/fl and Ifngr1^fl/fl Vil^Cre mice were infected and given 500ug of anti-CSF1R or isotype control antibody on days 7, 9 and 11 p.i and sacrificed on day 12 p.i. (n = 8 per group). (H) Flow cytometric plot for colon macrophages from Ifngr1^fl/fl Vil^Cre mice treated with anti-CSF1R or isotype control antibody. (I) Evaluation of indicated cytokines in colon explants. (J) Flow-cytometric analyses of CD4⁺ IE-T cells and GM-CSF production from them and (K) colon length at day 12 p.i. Data were analyzed by Brown-Forsythe and Welch’s ANOVA followed by Dunnett’s post-test. Scale bar = 100 um.

Extended Data Fig. 5 |. Antigen presentation by CX3CR1 + APCs promote T_H1 immunity.

H2ab ^fl/fl and H2ab^ΔiCX3CR1 mice were infected with Citrobacter, injected with tamoxifen on days 3–6 and sacrificed on day 12 p.i. (A) Flow cytometric analyses of MHCII expression in CD172a⁺ cells from the MLNs of H2ab ^fl/fl and H2ab^ΔiCX3CR1 mice (n = 9 per group). (B) enumeration of GM-CSF⁻ IL-17A⁺ CD4⁺ IE-T cells from the MLNs of H2ab ^fl/fl»Ifngr ^ΔiIEC and H2ab^ΔiCX3CR1»Ifngr ^ΔiIEC chimeric mice after treatments described in Fig. 3g at day 12 p.i. (n = 7 per group). (C) H2ab ^fl/fl and H2ab^ΔiCX3RC1 mice were treated with tamoxifen for 4 consecutive days, followed by infection with Citro^ova 3 days after the last injection. Mice were injected with CTV-labelled 1 × 10⁶ Rag-OTII cells on day 5 p.i., treated with 20ug of FTY720 i.p. on days 6–8 and sacrificed on day 9 (n = 6 per group). (C) Experimental scheme (D) OTII cells from MLNs were analyzed for T-bet induction. Data were analyzed by (A and D) by Mann-Whitney U test (two-sided) or (B) Kruskal-Wallis test followed by Dunn’s post-test.

Extended Data Fig. 6 |. Epithelial sensing of IFN-γ promotes eATPase production from IE-T cells.

(A-I) Ifngr1^fl/fl and Ifngr ^fl/fl Vil^Cre mice were infected with Citrobacter rodentium. (A and B) RT-PCR analyses for indicated factors in colon epithelial tissue fraction (n = 4 for d0, d6 per group and n = 8 for Ifngr1^fl/fl and n = 6 for Ifngr1^fl/fl Vil^Cre for d12). (C, E-I) flow cytometric (n = 3 for d0, n = 5 for d6 and d12 per group) and (D) immunofluorescence evaluation of CD39 expression, representative of two independent experiments and (G) RT-PCR analyses of Entpd5 expression from the purified IE-T cells and IECs (n = 6 per group). (I) Ifngr1^fl/fl and Ifngr1^fl/fl Vil^Cre mice were infected with ova⁺ Citrobacter and the ova-specific CD8αβ⁺ and CD4⁺ IE-T cells were evaluated for CD39 expression and (J) comparison of GM-CSF and CD39 producing CD4⁺ colonic IE-T cells in Ifngr1^fl/fl mice at day 12 p.i. (K-M) Ifngr1^fl/fl and Ifngr1^fl/fl Vil^Cre mice were infected with Citrobacter rodentium and treated with 250ug isotype or CD8β-depleting antibodies every 2 days after the infection and sacrificed on day 12 p.i. (n = 8 per group). (K) FACS plot for CD8⁺ IE-T cells from mice given isotype or CD8β-depleting antibodies. (L) Colon explants were analyzed for indicated analytes and (M) colon length measurement. (N) Ifngr1^fl/fl, Ifngr1^fl/fl Vil^CreERT2 and Ifngr1^fl/fl Vil^CreERT2 R26^LSLIrf1 mice were infected with Citrobacter rodentium on day 0, injected with tamoxifen at days 3–6 post-infection and analyzed on day 12 for CD39 expression from CD8αβ⁺ and CD4⁺ IE-T cells (n = 5 per group). The horizontal bar represents the median, and each symbol represents an individual mouse. Data were analyzed by (A, B) by Brown-Forsythe and Welch’s ANOVA followed by Dunnett’s post-test or (E-I) Mann-Whitney U test (two-sided) or (L-N) Kruskal-Wallis test followed by Dunn’s post-test. Scale bar = 100 um.

Extended Data Fig. 7 |. Epithelial IFN-γ signaling controls antigen presentation by IECs.

Flow cytometric enumeration of MHCII (A and C) or MHCI (B) or SIINFEKL-MHCI complex (D-E) on the surface of the colonic IECs after infection with Citro^WT (A and E) or Citro^Ova (B-D) at indicated day p.i. (A) or on day 12 p.i. (n = 5 per group). (B-E). Tamoxifen treatment was applied to induce the expression of the transgenes (B-E) at days 3–6 p.i. (F) Ifngr1^fl/fl and Ifngr1^fl/fl Vil^Cre mice were infected with Citrobacter and treated with anti-CD4 (n = 6) or isotype antibodies (n = 4) every 2 days starting at day 6 and sacrificed on day 12 p.i. The horizontal bar represents the median, and each symbol represents an individual mouse. Data were analyzed by (A) by ANOVA followed by Holm-Sidak post-test or (B-D and F) Kruskal-Wallis test followed by Dunn’s post-test or (E) Mann-Whitney U test (two-sided).

Extended Data Fig. 8 |. Antigen presentation via MHCI by IECs promotes eATPase production from CD8ab⁺ IE-T cells.

(A) Design and genotyping of B2m ^fl/fl mouse, representative of two independent lines. (B-K) B2m ^fl/fl and B2m ^fl/flVil^CreERT2 mice were infected with ova-expressing Citrobacter on day 0, treated with tamoxifen on days 3–6 and sacrificed on day 12. Colonic IECs on day 12 were analyzed for (B) MHCI expression (n = 5 for B2m ^fl/fl and n = 7 B2m ^fl/fl Vil^CreERT2) and (C) IE-T cells for CD39 expression (n = 5 per group) (D-E) (n = 3 for d0 and n = 5 for d6, d12 per group) by flow cytometry. (F-I) quantification of indicated analytes from colon explants (n = 8 per group) and (J) GM-CSF production from the ova-specific CD4⁺ IE-T cells and (K) Citrobacter colonization in the colon at day 12 p.i. (n = 5 per group). (L-N) B2m ^fl/fl and B2m ^fl/flVil^CreERT2 mice were infected with ova-expressing Citrobacter on day 0, treated with tamoxifen on days 3–6, injected with anti-NK1.1 or isotype antibodies on days 2, 4, 6, 8 and 10 and sacrificed on day 12 (n = 5 per group). (L) IFN-γ production from NK1.1⁺ TCRαβ⁻ TCRγδ⁻ cells (M) CD45⁺CD49b⁺ cells and (N) colon length at day 12. The horizontal bar represents the median, and each symbol represents an individual mouse. Data were analyzed by (B, D, E and N) ANOVA test followed by Sidak’s post-test or (C, F-L) Kolmogorov-Smirnov test.

Extended Data Fig. 9 |. Antigen presentation by IECs promotes eATPase production from CD4⁺ IE-T cells.

H2ab ^fl/fl and H2ab ^fl/flVil^CreERT2 mice were infected with ova-expressing Citrobacter on day 0, treated with tamoxifen on days 3–6, and sacrificed on day 12 (n = 5–8 per group). (A) Colonic IECs and myeloid cells were analyzed for MHCII expression on day 12 and (C-F) IE-T cells for CD39 expression by flow cytometry (n = 5 per group). (G) Entpd5 expression by RT-PCR after purification (n = 6 per group). (H-K) quantification of indicated analytes from colon explants (n = 8 pe group). (L-M) GM-CSF production from the CD4⁺ IE-T cells (n = 5 per group). (N) colon length and (O) Citrobacter colonization in the colon at day 12 p.i. (n = 8 per group). The horizontal bar represents the median, and each symbol represents an individual mouse. Data were analyzed by (A-O) Mann-Whitney U test (two-sided).

Extended Data Fig. 10 |. Antigen presentation by IECs restrains pathogenic CD4 T cells that promote chemical colitis.

(A-I) Ifngr1^fl/fl and Ifngr1^fl/flVil^Cre were treated with 1% DSS for 9 days followed by regular drinking water and sacrificed on day 12. (A) Expression of MHCI and (B) MHCII on the epithelial cell surface (n = 6 per group). Expression of CD39 on (C) CD8αβ⁺ and (D) CD4⁺ IE-T cells (n = 5 per group). (E) Quantification of ATP in the colon explants (n = 8 per group) and (F) production of GM-CSF from the CD4⁺ IE-T cells at day 12 (n = 5 per group). (G) Enumeration of neutrophils in the epithelium and (H) Bodyweight change and (I) colon length (n = 6 per group) and representative picture at day 12. (J-M) B2m ^fl/fl and B2m ^fl/flVil^CreERT2 mice (n = 5 per group) were treated with 1% DSS for 9 days followed by regular drinking water and sacrificed on day 12. Mice were also treated with tamoxifen on days 3–6. (N) Colon lengths from Ifngr1^fl/fl and Ifngr1^fl/flVil^Cre mice treated isotype (n = 10), α-CD4 (n = 8), α-GM-CSF (n = 8), αβ-ATP (n = 10), apyrase (n = 10), MCC950 (n = 10) or α-IL1R (n = 10) during DSS administration and sacrificed on day 12. Each symbol represents an individual mouse, and the bar represents the median except for (H) where each dot indicated the mean and bar indicates the SEM. Data were analyzed by (A-G, I-M) Mann-Whitney U test (two-sided) or (H) ANOVA followed by Sidak post-test or (N) Kruskal-Wallis test followed by Dunn’s post-test.

Extended Data Fig. 11 |. Model.

Upon sensing IFN-γ, colonic epithelial cells productively present pathogen-derived and self-antigens to the cognate intraepithelial T cells, critically situated at the epithelial barrier. Antigen presentation by the epithelial cells induces extracellular adenosine triphosphatase (ATPase) expression in the cognate intraepithelial T cells, which limits accumulation of extracellular ATP and resultant activation of the NLRP3 inflammasome in macrophages resident in the tissue. In contrast, macrophage antigen presentation alongside inflammasome-associated interleukin 1 alpha (IL-1α) and IL-1β production elicits pathogenic transformation of CD4-positive T cells into granulocyte-macrophage colony stimulating factor (GM-CSF)-producing T cells in vivo, promoting colitis and colorectal carcinogenesis. The figure was created using BioRender (https://biorender.com).

Fig. 1 |. Epithelial sensing of IFNγ protects from colitis by means of upregulation of IRF1.

Ifngr1^fl/fl and Ifngr1^fl/flVil^Cre mice were infected with C. rodentium. a–c, Citrobacter colonization in colon (a), colon length (b) and enumeration (c) of the CD11b⁺Ly6G⁺ cells in the epithelium at the indicated day p.i. and representative flow cytometry plot gated on CD45⁺Lin⁻ (EpCAM, Ter119, CD19, TCR⁻) cells at day 12 (n = 5 per group). d, Representative pictures of H&E-stained colon sections at ×100 magnification and histology scores. e, STAT1 pTyr701 analyses in EpCAM⁺ cells by flow cytometry and representative flow cytometry plot gated on EpCAM⁺CD45⁻ cells at day 12 (n = 5 per group). f, Quantification of Irf1 transcript in purified EpCAM⁺ IECs (n = 4 Ifngr1^fl/fl and n = 6 Ifngr1^fl/flVil^Cre). g, Ifngr1^fl/fl, Ifngr1^fl/flVil^CreERT2 and Ifngr1^fl/flVil^CreERT2R26^LSLIrf1 mice were infected with C. rodentium on day 0, injected with tamoxifen at days 3–6 p.i., and analysed on day 12 (n = 5 per group). h, Immunoblotting for IRF1 in the purified IECs (representative of two independent experiments). i, Flow cytometric analyses of IFNγR1 in IECs. j, Colon length. k, Enumeration of the proportion of CD11b⁺Ly6G⁺ cells in CD45⁺Lin⁻ (EpCAM, Ter119, CD19, TCR⁻) cells in the epithelial compartment. l, Histology score and representative H&E-stained colon section at ×100 magnification (n = 5 per group). The horizontal bar represents the median, and each symbol and lane in the immunoblot represents an individual mouse. Data were analysed by analysis of variance (ANOVA) tests (a–f,k) followed by the Dunnet’s or Sidak post hoc test or Kruskal–Walis test followed by Dunn’s post-test (j,l). Scale bar, 100 μm. CFU, colony-forming unit; MFI, median fluorescence intensity.

Fig. 2 |. Epithelial sensing of IFNγ restrains pathogenic CD4 GM-CSF response.

a, Flow cytometric analysis of ova expression by Citrobacter. b, Representative flow cytometry plot and enumeration of ova-specific CD4⁺ IE-T cells. c, Intracellular staining in ova-specific CD4⁺ IE-T cells obtained from the colons of Ifngr1^fl/fl and Ifngr1^fl/flVil^Cre mice at day 12 p.i. with the ova⁺ Citrobacter (n = 5 per group). d, R26^LSLOva and Vil^CreERT2R26^LSLOva mice were injected with tamoxifen for four consecutive days, and purified IECs were analysed for the expression of ova by immunoblotting (n = 4 per group). e,f, Ifngr1^fll+Vil^CreERT2R26^LSLOva and Ifngr1^fl/flVil^CreERT2R26^LSLOva mice were infected with C. rodentium on day 0, injected with tamoxifen on days 3–6 p.i. and sacrificed on day 12. Enumeration of ova-specific CD4⁺ IE-T cells (e) and intracellular staining in them (f) (n = 4 per group). Intracellular staining for IL-17A, IFNγ and GM-CSF was performed after brief restimulation. g–i, Ifngr1^fl/fl and Ifngr1^fl/flVil^Cre mice were infected with Citrobacter; treated with anti-CD4, anti-GM-CSF or anti-TNF antibodies every two days starting at day 7; and sacrificed on day 12 p.i. (n = 8 per group) GM-CSF in colon explant (g), colon length (h) and histology score and representative H&E-stained colon section at ×100 magnification (i). Each symbol represents an individual mouse, and the bar indicates the median. Data were analysed by Mann–Whitney U test (two-sided; b,c,e,f) or Kruskal–Wallis test followed by Dunn’s post-test (g–i). Scale bar, 100 μm.

Fig. 3 |. Epithelial sensing of IFNγ promotes eATPase expression in IE-T cells to restrain the colitogenic NLRP3 inflammasome.

Ifngr1^fl/fl and Ifngr1^fl/flVil^Cre mice were infected with C. rodentium. a, Quantification of ATP secreted by the colon explants (n = 8 per group). Ifngr1^fl/fl and Ifngr1^fl/flVil^Cre mice were infected with Citrobacter and treated with an isotype (n = 8), αβ-ATP (n = 8), apyrase (n = 6), MCC950 (n = 6) or anti-IL-1R (n = 6) starting on day 7 and sacrificed on day 12 p.i. b, Immunoblotting for cleaved caspase-1 in the colon lysates. c,d, ELISA for IL-1β (c) and GM-CSF (d) in the colon explant supernatants. e, Colon length. f, Flow cytometric analysis MHCII expression by CX3CR1⁺ cells in WT colon at day 12 p.i. g, Experimental strategy. h, GM-CSF and IL-17A production in CD4⁺ IE-T cells. i, Colon length at day 72 (n = 7 per group). j, CD39 expression in CD8αβ⁺ and CD4⁺ IE-T cells at indicated days p.i. (n = 3 for day 0 and n = 5 for rest of the groups). Each symbol represents an individual mouse, and the bar indicates the median. Data were analysed by ANOVA tests followed by the Dunnet’s post hoc test. P value is two-sided. α, anti.

Fig. 4 |. IFNγ-mediated antigen presentation by IECs promotes eATPase production by cognate IE-T cells.

Ifngr1^fl/fl and Ifngr1^fl/flVil^Cre mice were infected with C. rodentium. a, Quantification of MHCI expression on the surface of IECs (n = 5 per group). b, Ifngr1^fl/fl and Ifngr1^fl/flVil^Cre mice were infected with Citro^Ova, and IECs were analysed for the presentation of H2Kb:Ova complex on the surface by flow cytometry (n = 5 per group). c–g, B2m ^fl/fl and B2m ^fl/flVil^CreERT2 mice were infected with ova-expressing Citrobacter on day 0, treated with tamoxifen on days 3–6, and sacrificed on day 12 p.i. Expression of CD39 by CD8αβ⁺ (c) and CD4⁺ (d) IE-T cells (n = 5 per group). Entpd5 expression in purified CD8αβ⁺ and CD4⁺ IE-T cells (e) (n = 6 per group), GM-CSF production from ova-specific CD4⁺ IE-T cells (f) and colon length (g) (n = 5 per group). Each symbol represents an individual mouse, and the bar represents the median. h, Structural prediction by motif analyses. i–k, Flow cytometric staining for IFNγR2 in WT and Ifngr2^P32L cell lines (four biological replicates) (i). These cells were stimulated with IFNγ and evaluated for STAT1 pTyr701 and IRF1 induction (representative of two independent experiments) (j) and MHCI expression on the cell surface by flow cytometry (k). Each symbol represents an individual human patient, and the bar represents the median. Data were analysed by ANOVA tests (a,b) followed by the Dunnet’s post hoc test, Mann–Whitney U test (two-sided) (c–g) or Kruskal–Wallis test followed by Dunn’s post-test (h,l). Scale bar, 100 μm. MFI, median fluorescence intensity.

Fig. 5 |. IFNγ-mediated antigen presentation restricts colon cancer.

Ifngr1^fl/fl (n = 8) Ifngr1^fl/flVil^Cre (n = 6) mice were injected with AOM on day 0, followed by administration of 1.5% DSS in drinking water on days 6–11, 25–31 and 45–51; they were sacrificed on day 80. a, AOM–DSS treatment model. b, Representative pictures of the colon. c, Enumeration of tumours in the proximal and distal half of the colon from Ifngr1^fl/fl (n = 8) and Ifngr1^fl/flVil^Cre (n = 6) mice at day 80. d, Flow cytometric enumeration of infiltrating CD11b⁺Ly6G⁺ PMNs. e, CD39 expression by CD8αβ⁺ IE-T cells from proximal and distal half of the colon at day 80. f,g, Secretion of ATP (f) and production of GM-CSF (g) from CD4⁺ IE-T cells from proximal and distal half of the colon from Ifngr1^fl/fl (n = 8) and Ifngr1^fl/flVil^Cre (n = 6) mice at day 80. h–k, B2m ^fl/fl and B2m ^fl/flVil^CreERT2 mice were administered tamoxifen for five consecutive days and seven days later subjected to AOM–DSS treatment. Enumeration of tumours in the colon (h) and flow cytometric analyses of PMN infiltration (i), CD39 expression by CD8αβ⁺ IE-T cells (j) and the proportion of GM-CSF-positive CD4⁺ IE-T cells (k) in the distal colon at day 80 (n = 5 per group). l, Enumeration of tumours in the colons of Ifngr1^fl/fl and Ifngr1^fl/flVil^Cre mice that were treated with AOM–DSS; administered isotype (n = 10), anti-CD8β (n = 9), αβ-ATP (n = 10), apyrase (n = 10), MCC950 (n = 10), anti-IL-1R (n = 10), anti-CD4 (n = 8) or anti-GM-CSF (n = 8) on days 9, 11, 13, 26, 28, 30, 32 and 46, 48, 50 and 52 post-AOM; and sacrificed on day 80. Each symbol represents an individual mouse, and the bar represents the median. Data were analysed by ANOVA tests (c–g,l) followed by the Dunnet’s post hoc test or Kolmogorov–Smirnov test (h–k). α, anti.