Immune checkpoint inhibitor-induced colitis is mediated by polyfunctional lymphocytes and is dependent on an IL23/IFNγ axis

Abstract

Immune checkpoint inhibitors (CPIs) are a relatively newly licenced cancer treatment, which make a once previously untreatable disease now amenable to a potential cure. Combination regimens of anti-CTLA4 and anti-PD-1 show enhanced efficacy but are prone to off-target immune-mediated tissue injury, particularly at the barrier surfaces. To probe the impact of immune checkpoints on intestinal homoeostasis, mice are challenged with anti-CTLA4 and anti-PD-1 immunotherapy and manipulation of the intestinal microbiota. The immune profile of the colon of these mice with CPI-colitis is analysed using bulk RNA sequencing, single-cell RNA sequencing and flow cytometry. CPI-colitis in mice is dependent on the composition of the intestinal microbiota and by the induction of lymphocytes expressing interferon-γ (IFNγ), cytotoxicity molecules and other pro-inflammatory cytokines/chemokines. This pre-clinical model of CPI-colitis could be attenuated following blockade of the IL23/IFNγ axis. Therapeutic targeting of IFNγ-producing lymphocytes or regulatory networks, may hold the key to reversing CPI-colitis.

Fig. 1. Intestinal microbiota regulates susceptibility to CPI-induced colitis.

a Colon mass in wildtype female 6-week-old Balb/C mice at 1 week, 3 weeks and 6 weeks without treatment (Control: n = 56 at 1 week, n = 79 at 3 weeks, n = 4 at 6 weeks), treatment with combination anti-CTLA4/anti-PD-1 (CPI: n = 10 at 1 week, n = 25 at 3 weeks, n = 6 at 6 weeks), treatment with fecal microbiota (FMT: n = 6 at 1 week, n = 35 at 3 weeks, n = 24 at 6 weeks) and mice treated with both CPI and FMT (CPI + FMT: n = 68 at 1 week, n = 182 at 3 weeks, n = 23 at 6 weeks). Two-way ANOVA with Tukey’s multiple comparison test used ****P < 0.0001. CPI injections were given once a week intraperitoneally. b Representative colon sections cut at 3 µm thick FFPE from control, CPI, FMT and CPI + FMT treated wildtype female 6-week-old Balb/C mice stained with H&E (Leica). c Summary histological colitis score from mice treated with CPI-C (n = 11), plus controls (Control: n = 6, CPI: n = 7 and FMT: n = 7). Average colitis scored based on apoptosis (0–3), infiltrating lymphocytes (0–3), crypt abscess formation (0–3). Bar graph showing mean with SEM and Two-way ANOVA with Tukey’s multiple comparison test used whereby *P = 0.0106 **P = 0.0051 and ****P < 0.0001. d Representative flow cytometry plot showing the CD4⁺ and CD8⁺ T cell gating. e Number of CD4⁺ and CD8⁺ T cells from the colonic lamina propria of Control (n = 29), CPI (n = 12), FMT (n = 30) and CPI + FMT (n = 79) treated wildtype female 6-week-old Balb/C mice. Two-way ANOVA with Tukey’s multiple comparison test used ****P < 0.0001. f Representative flow cytometry plot showing Gr-1^hi neutrophils from Control, CPI, FMT and CPI + FMT treated wildtype female 6-week-old Balb/C mice. g Number of Gr-1^hi neutrophils from the colonic lamina propria of Control (n = 70), CPI (n = 18), FMT (n = 33) and CPI + FMT (n = 95) treated wildtype female 6-week-old Balb/C mice. Two-way ANOVA with Tukey’s multiple comparison test used **P<0.01 ****P < 0.0001.

Fig. 2. Longitudinal 16S analysis of microbiota shows early and persistent changes in composition of microbiota after FMT.

a Alpha-diversity (Shannon diversity index) of the microbiota between untreated control Balb/C wildtype female 6-week-old mice (n = 6), Balb/C wildtype mice treated with FMT (n = 6) and Balb/C wildtype female 6-week-old mice treated with CPI + FMT (n = 6) longitudinally over a 3 week period. A liner mixed model was used, based on ANCOM-BC, to account for the repeated measures of the mice. FDR Bonferroni corrected Padj. Box plot is defined by: Week 0 – Untreated = “Min.:3.775” “1st Qu.:3.787” “Median :3.828” “Mean:3.856” “3rd Qu.:3.867” “Max.:4.051” Week 1 – Untreated = “Min. :3.640” “1st Qu.:3.670” “Median :3.819” “Mean :3.909” “3rd Qu.:3.929” “Max. :4.586” Week 2 – Untreated =“Min. :3.659” “1st Qu.:3.823” “Median :3.871” “Mean :3.911” “3rd Qu.:3.968” “Max. :4.255“Week 3 – Untreated =“Min. :3.148” “1st Qu.:4.105” “Median :4.236” “Mean :4.049” “3rd Qu.:4.331” “Max. :4.427” Week 0 – FMT-only = “Min.:3.603” “1st Qu.:3.823” “Median:3.852” “Mean :3.890” “3rd Qu.:4.033” “Max. :4.137” Week 1 – FMT-only = “Min. :3.610” “1st Qu.:3.901” “Median :3.990” “Mean :4.007” “3rd Qu.:4.100” “Max. :4.445” Week 2 – FMT-only = “Min. :3.859” “1st Qu.:3.927” “Median :4.008” “Mean :4.021” “3rd Qu.:4.114” “Max. :4.206” Week 3 – FMT-only = “Min.:3.832” “1st Qu.:4.016” “Median :4.037” “Mean :4.069” “3rd Qu.:4.193” “Max. :4.250” Week 0 – CPI = “Min. :3.591” “1st Qu.:3.734” “Median :3.801” “Mean :3.772” “3rd Qu.:3.836” “Max. :3.878” Week 1 – CPI= “Min. :3.080” “1st Qu.:3.265” “Median :3.515” “Mean :3.557” “3rd Qu.:3.844” “Max. :4.097” Week 2 – CPI= “Min. :3.770” “1st Qu.:3.952” “Median :4.030” “Mean :4.084” “3rd Qu.:4.280” “Max. :4.384” Week 3 – CPI = “Min. :3.902” “1st Qu.:3.944” “Median :4.461” “Mean :4.348” “3rd Qu.:4.592” “Max. :4.839” b Non-metric dimensional scaling plot showing the beta diversity of the microbiota from untreated control Balb/C wildtype female 6-week-old mice (n = 6), Balb/C wildtype female 6-week-old mice treated with FMT (n = 6) and Balb/C wildtype female 6-week-old mice treated with CPI + FMT (n = 6) longitudinally over a 3 week period. c Phylum level relative abundance profiles from untreated control Balb/C wildtype female 6-week-old mice (n = 6), Balb/C wildtype female 6-week-old mice treated with FMT (n = 6) and Balb/C wildtype female 6-week-old mice treated with CPI + FMT (n = 6) longitudinally over a 3 week period. d Family level relative abundance profiles from untreated control Balb/C wildtype female 6-week-old mice (n = 6), Balb/C wildtype female 6-week-old mice treated with FMT (n = 6) and Balb/C wildtype female 6-week-old mice treated with CPI + FMT (n = 6) longitudinally over a 3 week period. Genus level coefficient changes comparing Balb/C wildtype female 6-week-old mice treated with FMT (n = 6) and Balb/C wildtype female 6-week-old mice treated with CPI + FMT (n = 6) at e 1 week, f 2 weeks and g 3 weeks of treatment. A liner mixed model was used, based on ANCOM-BC, to account for the repeated measures of the mice. FDR Bonferroni corrected Padj.

Fig. 3. Transcriptome analysis shows colonic epithelial dysfunction and activation of T cells, cytotoxicity, and immune compartments in mice with CPI-induced colitis.

a Venn diagram of the DEGs (FDR < 0.05) identified by comparing the expression profiles of whole colon biopsies from wildtype female 6-week-old mice with only CPI treatment (n = 4), only FMT (n = 3) or with both CPI + FMT (n = 3) against control wildtype female 6-week-old mice (n = 4). b Volcano plot highlighting the DEGs (FDR < 0.05) from the comparison of RNA samples from wildtype female 6-week-old mice treated with both CPI + FMT (n = 3) versus control wildtype female 6-week-old mice (n = 4). Positive log fold changes indicate over-expression in treated mice, while negative log fold changes indicate upregulation in wildtype mice. c The topmost significantly upregulated differentially expressed genes involved in epithelial function, protease/protease inhibitors, immune activation, interferon-stimulated genes and antigen processing/presentation in the distal colon of wildtype female 6-week-old mice with CPI-induced colitis (n = 3) in comparison with control wildtype female 6-week-old mice (n = 4). d Gene expression changes in the distal colon of wildtype female 6-week-old mice with CPI-induced colitis (n = 3) in comparison with control wildtype female 6-week-old mice (n = 4) were used to identify which biological pathways were significantly impacted, using IPA Canonical Pathways. P value adjustment for multiple comparisons was applied with the Benjamini and Hochberg method. e Hallmark pathway analysis using GSEA in the gene expression changes in the distal colon of wildtype female 6-week-old mice with CPI-induced colitis (n = 3) in comparison with control wildtype female 6-week-old mice (n = 4). P value adjustment for multiple comparisons was applied with the Benjamini and Hochberg method.

Fig. 4. Comparison between anti-CTLA4 and anti-PD-1 monotherapy treated mice showed shared and distinct patterns of mucosal immune activation.

a Venn diagram of the DEGs (FDR < 0.05) identified by comparing the expression profiles of whole colon biopsies from wildtype female 6-week-old mice with only anti-CTLA4 treatment (n = 3) or anti-PD-1 treatment (n = 3) in comparison to control wildtype female 6-week-old mice (n = 3). b Chair plots showing differentially expressed transcripts ranked by decreasing log fold change and coloured by estimated FDR in wildtype female 6-week-old mice with only anti-CTLA4 treatment (n = 3) or anti-PD-1 treatment (n = 3) in comparison to control wildtype female 6-week-old mice (n = 3). c Circular plot showing activation score of biological pathways which were significantly impacted, using IPA Canonical Pathways, in wildtype female 6-week-old mice with only anti-CTLA4 treatment (n = 3) or anti-PD-1 treatment (n = 3) in comparison to control wildtype female 6-week-old mice (n = 3). d Hallmark pathway analysis using GSEA in the gene expression changes in the distal colon of wildtype female 6-week-old mice with only anti-CTLA4 treatment (n = 3) or anti-PD-1 treatment (n = 3) in comparison to control wildtype female 6-week-old mice (n = 3). P value adjustment for multiple comparisons was applied with the Benjamini and Hochberg method. e Radar plot showing log2 fold change for the top 20 genes associated to the IFNγ responsive pathway in the gene expression changes in the distal colon of wildtype female 6-week-old mice with only anti-CTLA4 treatment (n = 3), anti-PD-1 treatment (n = 3) or combination treatment (n = 3), in comparison to control wildtype female 6-week-old mice (n = 3). f GSEA running score for the IFNγ responsive pathway in the gene expression changes in the distal colon of wildtype female 6-week-old mice with only anti-CTLA4 treatment (n = 3), anti-PD-1 treatment (n = 3) or combination treatment (n = 3), in comparison to control wildtype female 6-week-old mice (n = 3).

Fig. 5. Transcriptomic landscape of colonic lymphocytes in CPI-induced colitis at single-cell resolution.

a UMAP plots of the 17 lymphocyte populations labelled according to Louvain clustering, coloured based on SingleR cell-type assignments and split by the two conditions under consideration. b Dot plot of the key genes used to identify each of the 17 clusters as shown by average expression and percentage expressed in each cluster. c Average proportional changes shown across all clusters in wildtype female 6-week-old mice with CPI-induced colitis (n = 3) vs control wildtype female 6-week-old mice (n = 3). Bar graph showing mean with SEM and Two-way ANOVA with Sidak’s multiple comparison test used whereby ***P = 0.0004 and ****P < 0.0001. d Chair plots showing differentially expressed transcripts ranked by decreasing log fold change and coloured by estimated false discovery rate (FDR) across all CD4⁺ and CD8⁺ T cell clusters in wildtype female 6-week-old mice with CPI-induced colitis (n = 3) vs control wildtype female 6-week-old mice (n = 3). e Pathways, identified using GSEA Hallmark, upregulated in across all CD4⁺ and CD8⁺ T cell clusters in wildtype female 6-week-old mice with CPI-induced colitis (n = 3) vs control wildtype female 6-week-old mice (n = 3). P value adjustment for multiple comparisons was applied with the Benjamini and Hochberg method. f Violin plots showing the expression levels of cytokines across CD4⁺ and CD8⁺ T cell and B cell clusters in wildtype female 6-week-old mice with CPI-induced colitis (n = 3) vs control wildtype female 6-week-old mice (n = 3).

Fig. 6. CPI-induced colitis exhibits a profound and increased cytotoxic phenotype.

a UMAP plot and b the Pearson correlation showing the increased occurrence of Ifng and Gzmb expressing cells in CPI-induced colitis (n = 3) vs control samples (n = 4). P value adjustment for multiple comparisons was applied with the Benjamini and Hochberg method. c Heatmap of cytokine and chemokine expression shown by log₂ fold changes between CPI-induced colitis (n = 3) and control samples (n = 4) in colonic CD4⁺ and CD8⁺ T cell clusters. d Representative flow cytometry histograms and bubble plots showing the percentage of granzyme B, IFNγ and perforin expressing CD3⁺ CD4⁺ T cells and e CD3⁺ CD8⁺ T cells in wildtype female 6-week-old mice with CPI-induced colitis (n = 16) and control wildtype female 6-week-old mice (n = 16). f Concentrations of IFNγ and granzyme B were determined via Luminex of supernatants from cultured colonic lamina propria cells at 2 × 10⁶/ml for 24 h with plate bound anti-CD3 (2 µg/ml) for untreated Balb/C wildtype female 6-week-old mice (Control: n = 4) and wildtype female 6-week-old mice treated with CPI colitis model (CPI + FMT: n = 4). Mean and standard deviation are shown. *P < 0.05 two-sided Mann–Whitney U T test performed. g Colon mass (Control: n = 21) or wildtype female 6-week-old CPI colitis mice treated with an isotype control (CPI colitis: n = 21) or anti-IFNγ (Anti-IFNγ: n = 24), Kruskal-Wallis Test with Dunn’s multiple corrections performed where **P = 0.0014 and ****P < 0.0001. h Average colitis score, using the same previous colitis score parameters (Control: n = 7) or wildtype female 6-week-old CPI colitis mice treated with an isotype control (CPI colitis: n = 17) or anti-IFNγ (Anti-IFNγ: n = 8). Bar graph showing mean with SEM and Two-way ANOVA with Tukey’s multiple comparison test used whereby ***P = 0.0003 and ****P < 0.0001. i Number of Gr-1^hi neutrophils of untreated wildtype female 6-week-old Balb/C mice (Control: n = 25) or wildtype female 6-week-old CPI colitis mice treated with an isotype control (CPI colitis: n = 43) or anti-IFNγ (Anti-IFNγ: n = 24) Kruskal–Wallis Test with Dunn’s multiple corrections performed where **P = 0.0025 and ****P < 0.0001.

Fig. 7. CPI-induced colitis is mediated by CXCR6 expressing tissue-resident T cells.

a Heatmap of co-stimulatory and co-inhibitory expression from Ifng⁺ expressing CD4⁺ and CD8⁺ T cells in comparison to Ifng^--expressing CD4⁺ and CD8⁺ T cells. b Heatmap of chemokines and gut homing gene expression from Ifng⁺ expressing CD4⁺ and CD8⁺ T cells in comparison to Ifng^--expressing CD4⁺ and CD8⁺ T cells. c Percentage of CD3⁺ α4β7⁺ in the colon and mesenteric lymph node (mLN), d colon mass and e number of Gr-1^hi neutrophils in wildtype female 6-week-old CPI colitis mice treated with an isotype control (n = 8) or anti-α4β7 depleting antibody (n = 8). f Number of CD4⁺ and CD8⁺ T cells from the colonic lamina propria and g intraepithelial layer of untreated wildtype female 6-week-old Balb/C mice (Control: n = 10) and CPI + FMT (CPI colitis: n = 10). For both f, g, a two-tailed Mann–Whitney U Test was used where *P = 0.0172 **P = 0.0039 ***P = 0.0007 and ****P < 0.0001.

Fig. 8. IL23 blockade attenuates the development of CPI colitis.

a Network analysis, using Enhanced Causal Network using Qiagen IPA, of predicted upstream regulators of Ifng-expressing CD4⁺ T cell in CPI colitis. Multiple cytokines, transmembrane receptors and transcriptional regulators were predicted to significantly (FDR < 0.05) regulate Ifng-expressing CD4⁺ T cell. b Representative flow cytometry plot and c percentage of IFNγ⁺/TNFα⁺ CD4⁺ T cells from wildtype female 6-week-old CPI colitis mice treated with an isotype control (n = 7) or an IL-23 blocking antibody (n = 8). Two-tailed Mann–Whitney Test performed where *P = 0.019. d Colon mass, in either control mice (n = 12) or wildtype female 6-week-old CPI colitis mice treated with an isotype control (n = 23) or an IL-23 blocking antibody (n = 24). Kruskal-Wallis test with Dunn’s multiple comparison used where ****P < 0.0001. e Gr-1^hi neutrophil counts in either wildtype female 6-week-old CPI colitis mice treated with an isotype control (n = 24) or an IL-23 blocking antibody (n = 26), using a two-tailed Mann–Whitney Test to show ***P = 0.0006 and f Average colitis score, using the same previous colitis score parameters, in either control mice (n = 7) or wildtype female 6-week-old CPI colitis mice treated with an isotype control (n = 17) or an IL-23 blocking antibody (n = 8). Bar graph showing mean with SEM and Two-way ANOVA with Tukey’s multiple comparison test used whereby ****P < 0.0001. g Colon mass, using a two-tailed Mann–Whitney Test to show ***P = 0.0003, and h Gr-1^hi neutrophil counts in either wildtype female 6-week-old mice (n = 17) or female 6-week-old Il23^−/− mice (n = 20) treated with CPI colitis. A two-tailed Mann–Whitney Test to show **P = 0.005, i Percentage of IFNγ⁺/TNFα⁺ CD4⁺ T cells from CPI colitis treated wildtype female 6-week-old mice (n = 8) or female 6-week-old Il23^−/− mice (n = 12). A two-tailed Mann–Whitney Test to show *P = 0.0, j PathLinker network analysis of Ifng-expressing CD4⁺ T cell in CPI colitis showing the highest-ranking shortest paths involving differentially expressed genes linking Il23r to Ifng through regulatory and protein-protein interactions queried from OmniPath. Paths with wider edges rank higher with upregulated genes shown in orange, downregulated in blue.