cGAS restricts colon cancer development by protecting intestinal barrier integrity

To assess the function of cGAS in the intestine, we fed the cGas^−/− and WT mice with 3% DSS in drinking water for 5 d to induce acute colitis (Fig. 1A). This chemical-induced colitis model has been shown to recapitulate many of the clinical features associated with ulcerative colitis (22). Colitis susceptibility was evaluated daily by monitoring body weight changes and clinical features of colitis, such as diarrhea and occult stool bleeding. cGas^−/− mice lost significantly more body weight and exhibited increased diarrhea and rectal bleeding compared to WT mice (Fig. 1 B–D). Mice were killed on day 8 and examined for histopathological changes. cGas^−/− mouse colons were significantly shorter than those in the WT mice (Fig. 1 E and F). The severity of colitis in cGas^−/− mice was further supported by histological staining of whole colon tissue sections (Fig. 1G). Compared to WT mice, much of the distal colon in cGas^−/− mice showed increased severity of disease with complete crypt destruction, epithelium ulceration, and submucosal edema (Fig. 1G). Histopathological analysis showed significantly higher scores for inflammation, ulceration, and crypt distortion in cGas^−/− mouse colons (Fig. 1H). These data suggest a protective role for cGAS in acute colitis induced by DSS.

Patients with IBD are at a higher risk of colorectal cancer (2). We further explored the role of cGAS in the development of IBD-associated colon tumorigenesis by using azoxymethane (AOM) as a carcinogen in the DSS-induced inflammation-driven colon cancer model (Fig. 2A) (23, 24). Following initial injection with AOM and three cycles of 1.5% DSS treatments, cGas^−/− mice developed significantly increased numbers of macroscopic colon tumors throughout the whole colon, including proximal, middle, and distal colon (Fig. 2 B and C), while only several scattered tumors appeared mainly in the distal colon of WT mice. The sizes of macroscopic tumors (maximal cross-sectional area of tumors >2 mm) were greatly increased in cGas^−/− mice compared to WT controls (Fig. 2D). However, colon lengths were similar between WT and cGas^−/− mice (SI Appendix, Fig. S1A). In addition, we found that around 50% cGas^−/− mice died early, whereas WT mice survived when given 2.5% DSS water (Fig. 2E). cGas^−/− mice also showed splenomegaly and heavier spleen weight compared to WT mice after CAC development (SI Appendix, Fig. S1 B and C), while both WT and cGas^−/− spleens in untreated animals had similar weights (SI Appendix, Fig. S1D).

Histopathological analyses showed that cGas^−/− mice exhibited increased hyperplasia in the proximal, middle, and distal colon, while WT mice showed none or much less hyperplasia in the distal colon (Fig. 2F). Low-grade colonic mucosal dysplasia was observed in tumor-bearing WT mice, whereas low-grade to high-grade dysplasia was found in cGas^−/− mice (Fig. 2 F and G). Histological analysis also showed more adenocarcinomas in cGAS-deficient mice than in WT mice (Fig. 2F), suggesting the absence of cGAS accelerates tumor progression. However, we did not observe submucosal invasion of tumor cells in cGas^−/− mice, indicating that cGAS deficiency did not impact metastasis in the CAC model. Interestingly, mice lacking STING (Sting^gt/gt and Sting^−/−) or type I IFN receptor (IFNAR^−/−) did not show a significant increase of colon tumors in this DSS-AOM model, suggesting that cGAS restricts colon cancer development through a mechanism that is independent of STING or type I IFN signaling (Fig. 2H).

The microbiota is widely reported to regulate the pathogenesis and severity of gastrointestinal diseases, including CAC. 16S rRNA analysis showed similar levels of several known mouse gut commensal and colitogenic bacteria, such as Bacteroides, Lactobacillus, segmented filamentous bacteria, Staphylococcus, Streptococcus, Escherichia coli, Enterococcus, and Bifidobacteria, between WT and cGas^−/− mice (SI Appendix, Fig. S1E). To further explore the role of microbiota, WT and cGas^−/− mice were cohoused for 4 wk before inducing colon cancer. Similar to single-housed mice, cohoused cGas^−/− mice also had higher tumor numbers than cohoused WT mice (SI Appendix, Fig. S1F). In addition, examination of the littermates from breeding of cGAS heterozygous mice showed cGas^−/− mice also had significantly more tumor numbers than WT mice (SI Appendix, Fig. S1G). Importantly, cGas^−/− mice pretreated with antibiotics developed similar tumor numbers to those in the cGas^−/− mice without antibiotic treatment; but the tumor numbers in both cGas^−/− mice were more than those in the WT mice that were pretreated with antibiotics (SI Appendix, Fig. S1H). These results indicate that the microbiota is not a key determinant for accelerated tumor development in the cGas^−/− mice.

Colorectal cancers develop either spontaneously or as a complication of IBD. As cGAS prevents colitis-associated colon tumorigenesis, we further explored whether cGAS is also involved in sporadic colon cancer using the Apc^min/+ mouse, a noncolitis-driven colorectal cancer model (25). Mutations of the APC gene, which normally inhibits the Wnt/β-catenin pathway, accelerates colorectal tumorigenesis in mice and humans. We analyzed sex- and age-matched cohorts of Apc^min/+ mice on cGAS-sufficient and -deficient backgrounds. cGas^−/−Apc^min/+ mice had similar numbers of colon tumors as cGas^+/+Apc^min/+ mice (SI Appendix, Fig. S1I). Although cGas^+/+Apc^min/+ mice had high numbers of tumors in the small intestine, cGas^−/−Apc^min/+ mice did not show obvious changes in overall tumor numbers compared to cGas^+/+Apc^min/+ mice (SI Appendix, Fig. S1J). Altogether, these results suggest that cGAS does not restrict sporadic colon tumorigenesis, but inhibits CAC.

To identify the cellular source responsible for the enhanced tumorigenesis in the absence of cGAS, we performed bone marrow transplantation experiments to examine the contribution of the radio-resistant nonhematopoietic compartment relative to that of the radio-sensitive hematopoietic compartment by generating bone marrow-chimeric mice. WT and cGas^−/− mice were exposed to whole-body irradiation to deplete hematopoietic stem cells and progenitor compartments. Irradiated recipients were then injected with either WT or cGas^−/− bone marrow. After 2-mo reconstitution with an efficiency around 90%, CAC was induced with AOM and DSS, as described above. WT mice reconstituted with cGas^−/− bone marrow (cGas^−/−→WT) exhibited increased tumor sizes and number as compared to WT control mice (WT→WT) (Fig. 3 A–C). In contrast, cGas^−/− recipient mice reconstituted with WT bone marrow (WT→cGas^−/−) had significantly lower tumor burden than mice lacking cGAS in both compartments (cGas^−/−→cGas^−/−), suggesting that cGAS expression in the hematopoietic compartment plays a major role in the restriction of tumorigenesis during CAC (Fig. 3 A–C). However, the WT→cGas^−/− group still had increased colon tumor loads as compared to the WT→WT controls (Fig. 3A), and the cGas^−/−→WT group exhibited significantly smaller tumor sizes and lower numbers compared to the cGas^−/−→ cGas^−/− group (Fig. 3 A–C), indicating that cGAS expression in radio-resistant, nonhematopoietic cells also limits tumorigenesis during CAC. Colon lengths did not show significant statistical differences among different groups (Fig. 3D). Histopathology analyses further confirmed that cGAS expression in both hematopoietic and nonhematopoietic compartments contribute to CAC development (Fig. 3E). Altogether, these results suggest that cGAS protects colon cancer development in both hematopoietic and nonhematopoietic compartments.

To assess cGAS function in CAC, we investigated cGAS expression in the intestinal crypts that contain stem cells and differentiated intestinal epithelial cells (IECs), lamina propria cells, colons, and spleens. Interestingly, cGAS expression was much higher in the crypts than in the lamina propria cells and splenocytes. In contrast, STING expression was much higher in splenocytes and lamina propria cells than in crypts (Fig. 4A). These results suggest a cGAS-dependent but STING-independent function in crypts. By analyzing published single-cell RNA-sequencing (scRNA-seq) data by Haber et al. (26), we found that cGAS (Mb21d1) expression was the highest in the intestinal stem cell (Lgr5⁺) population. The expression of cGAS was gradually down-regulated during differentiation with medium expression levels in early differentiated cells, such as transient amplifying cells and enterocyte progenitors, and low expression level in terminal differentiated cells, such as enterocytes, paneth cells, and goblet cells (Fig. 4B).

Consistent with these results, analysis of the published scRNA-seq data from Ayyaz et al. (27) showed that the intestinal stem cells expressed much higher levels of cGAS RNA than myeloid cells and lymphocytes (SI Appendix, Fig. S2A). However, STING (Tmem173) was highly expressed in immune cells but not in IECs (Fig. 4B and SI Appendix, Fig. S2A). Importantly, by analyzing cells from Lgr5-EGFP reporter mice, we found that cGAS expression was higher in the Lgr5⁺ populations compared to the Lgr5dim and spleen cell population, whereas STING expression was higher in the spleen but lower in the Lgr5⁺ and Lgr5dim populations (Fig. 4C). Together, these results suggest that cGAS may regulate intestinal stem cells through a STING-independent mechanism. While Lgr5 expression was initially comparable between WT and cGas^−/− mice, AOM and DSS treatment decreased Lgr5 expression in the crypts of cGas^−/− mice, whereas the levels of Lgr5 expression were largely unchanged in WT and Sting^gt/gt mice (Fig. 4 D and E). The results suggest that cGAS, but not STING, plays a critical role in regulating intestinal stem cells in the crypts. As the epithelial barrier controls intestinal homeostasis, we also investigated intestinal barrier integrity during the acute inflammation phase after AOM/DSS treatment. Compared to WT and Sting^gt/gt mice, significantly more FITC-dextran was found in the serum of cGas^−/− mice after DSS-induced injury (Fig. 4F and SI Appendix, Fig. S2B). No difference in FITC-dextran concentrations was observed between cGas^−/−Apc^min/+ and cGas^+/+Apc^min/+ mice (SI Appendix, Fig. S1K). These results indicate that cGAS protects gastrointestinal barrier integrity to prevent CAC.

To further investigate the mechanism by which the loss of cGAS exacerbates CAC development, we examined whether cGAS regulates cellular proliferation following exposure to AOM and DSS. Immunohistochemical analyses at 80 d after injection of AOM revealed significantly higher numbers of K_i67⁺ cells per crypt in the intestinal epithelium of cGas^−/− mice than in WT mice (Fig. 4G), whereas no difference was observed in untreated WT and cGas^−/− mice (SI Appendix, Fig. S2C). These results suggest that cGAS deficiency led to faster proliferation of IECs during cancer development. Consistently, we also found a significantly higher number of BrdU⁺ cells per crypt in the colons of cGas^−/− mice at 80 d after injection of AOM (Fig. 4H), with similar low incorporation numbers for the untreated WT and cGas^−/− mice (SI Appendix, Fig. S2D). The tumor areas of cGas^−/− mice showed even more K_i67⁺ and BrdU⁺ cells (SI Appendix, Fig. S2E). S100a9 was reported to be an important proinflammatory mediator in colonic inflammation and contributed to tumorigenic processes by regulating cell proliferation (28). Real-time PCR analysis found that the levels of S100a9 RNA in the colon of cGas^−/− mice at 15 and 80 d after injection of AOM were higher than those in WT mice (Fig. 4I).

Increased intestinal cell proliferation may be related to reduced apoptosis. However, the RNA levels of the antiapoptotic gene Bcl-2 in the colon did not show a significant difference between WT and cGas^−/− mice on days 15 and 80 after injection of AOM, while the apoptotic gene Bax in the colon of cGas^−/− mice increased on day 80 (SI Appendix, Fig. S3A). Moreover, apoptotic caspase-7 and caspase-3 cleavage in the colon of cGas^−/− mice was stronger than that in WT mice on day 80 (SI Appendix, Fig. S3B), suggesting more cell death in cGas^−/− mice during CAC development. Terminal deoxynucleotidyl transferase dUTP nick end-labeling (TUNEL) staining of colon sections also revealed accelerated apoptosis in cGas^−/− mice, especially within the tumor (Fig. 4J and SI Appendix, Fig. S3C). There was also more TUNEL⁺ staining in cGas^−/− mice during acute colitis (SI Appendix, Fig. S3 D and E). It’s possible that the enhanced apoptosis of IECs in cGas^−/− mice led to compensatory hyperproliferation of these cells, which contributes to colon tumor development in cGas^−/− mice.

To examine possible intrinsic defects in the mucus layer, we analyzed whether cGAS regulates the expression of genes encoding mucins in the colon. The expression of MUC2, MUC3, and MUC4 in the colon of WT and cGas^−/− mice were comparable in untreated animals. After the AOM/DSS treatment, although only Muc2 RNA levels showed a lower expression in the cGas^−/− mice than in the WT mice on day 15, the RNA levels of Muc2, Muc3, and Muc4 all decreased on day 80 in the colon of cGas^−/− mice compared to those of WT mice (SI Appendix, Fig. S3F). In contrast to mucin genes, the expression levels of tight junction genes ZO-1, ZO-2, and occludin were similar in the crypts between WT and cGas^−/− mice on day 80 after the AOM/DSS treatment (SI Appendix, Fig. S3G). The decreased expression of the mucin genes in the cGas^−/− mice may contribute to the leaky intestinal barrier in these mice.

To investigate the molecular mechanism underpinning cGAS-dependent protection against colorectal tumorigenesis, we examined whether there was a defect in type I IFN signaling in cGas^−/− mice during colon cancer development, as the function of cGAS in infection and autoimmunity depends on the production of type I IFN. Similarly low levels of IFN-α and IFN-β in the colon were detected in untreated WT and cGas^−/− mice (SI Appendix, Fig. S4A). We then analyzed IFN-α and IFN-β production in the whole colon of WT and cGas^−/− mice at 15 d after injection of AOM and found similar levels of type I IFN production in the early stage of CAC (SI Appendix, Fig. S4B). The production of type I IFN in the whole colon also showed no overt differences between WT and cGas^−/− mice on day 80 when the colon cancer had reached the advanced stage (SI Appendix, Fig. S4B). IFN-β levels in the serum on days 15 and 80 were also comparable between WT and cGas^−/− mice (SI Appendix, Fig. S4C). IFNAR1 expression in the CD11b⁺ and CD11b⁺F4/80⁺ populations in the colon also showed similar levels between WT and cGas^−/− mice (SI Appendix, Fig. S4D). Similar gene-expression levels of IFN-stimulated genes (ISGs)—including CXCL10, ISG15, and IFIT3—were observed in the colon tissues of WT and cGas^−/− mice at 15 d after injection of AOM, but the expression of CXCL10 and ISG15 was even higher in cGas^−/− mice at 80 d after injection of AOM (SI Appendix, Fig. S4E). cGAS expression level in the whole colon was also not altered at days 15 and 80 after injection of AOM (SI Appendix, Fig. S4F). We further analyzed colon STAT1 phosphorylation, a hallmark of type I IFN pathway activation. cGas^−/− mice showed higher levels of STAT1 phosphorylation in the colon compared to WT mice on days 15 and 80 (Fig. 5A and SI Appendix, Fig. S5A). Altogether, these results suggest no apparent defect in type I IFN signaling in the colons of cGas^−/− mice compared to WT mice during CAC; instead, type I IFN signaling appeared to be enhanced at the later stage of CAC in the cGas^−/− mice.

STAT3 activation has been implicated in driving colon cancer development in both CRC patients and mouse colon tumor models (29, 30). We found by Western blotting that on days 15 and 80 the phosphorylation of STAT3 in the colon was stronger in cGas^−/− mice than in WT mice (Fig. 5A and SI Appendix, Fig. S5A). Immunohistology staining further confirmed higher phosphorylation of STAT3 in the colon of cGas^−/− mice on day 80 (Fig. 5B). Even in the acute colitis model induced by DSS, the colon of cGas^−/− mice showed stronger phosphorylation of STAT3 than WT mice on day 8 after treatment (SI Appendix, Fig. S5A). Apc^min/+ displayed high levels of STAT3 phosphorylation in the colon (31). However, cGAS deficiency did not further affect STAT3 phosphorylation in the Apc^min/+ mice (SI Appendix, Fig. S5B). In addition, we found that mRNA levels of interleukin (IL)-6, IL-1β, and MIP2 in the whole colon were significantly higher in the cGas^−/− mice bearing colon tumors (Fig. 5C). Higher levels of IL-6 protein in the colon of cGas^−/− mice were further detected by ELISA, suggesting that IL-6–STAT3 signaling may be important for the CAC in the cGas^−/− mice (Fig. 5D). Using MC38 colon cancer cell lines, we found IL-6–induced STAT3 phosphorylation was not impaired by cGAS deficiency, but was blocked by an anti–IL-6 antibody or the STAT3 inhibitor Cpd188 (SI Appendix, Fig. S5C) (32). These results indicate that the enhanced phosphorylation of STAT3 observed in the colon tissues of cGas^−/− mice was not due to increased sensitivity to IL-6 stimulation but increased production of IL-6 and other inflammatory cytokines.

WT and cGAS-deficient bone marrow-derived macrophages (BMDMs) showed similar levels of phosphorylation of IκBα and ERK1/2 after treatment with tumor necrosis factor-α (TNF-α) or the ligands for Toll-like receptor (TLR)4 and TLR2 (SI Appendix, Fig. S5D). Both AIM2- and NLRP3-inflammasome activation played a protective role in regulating colon inflammation and cancer development (33–35). Compared to WT BMDMs, cGas^−/− BMDMs showed similar activation of AIM2 and NLRP3 inflammasomes upon stimulation with double-stranded DNA and nigericin, respectively (SI Appendix, Fig. S5E). The DNA from mouse stool could also induce similar inflammasome activation in WT and cGas^−/− BMDMs (SI Appendix, Fig. S5E). These results suggest that the enhanced CAC development in cGas^−/− mice was not due to aberrant inflammasome activation.

To further explore whether STAT3 activation promotes colon cancer development in cGas^−/− mice, we treated colon cancer-bearing mice with the STAT3 inhibitor Cpd188 intraperitoneally (32). At the end of the treatment period, tumor numbers in the Cpd188-treated group decreased as compared with vehicle-treated group (SI Appendix, Fig. S5E). The sizes of tumors in the Cpd188 treated group were also smaller than those of vehicle-treated mice (Fig. 5F). No gross toxicities were evident as determined by the change of body weight. Histological analysis showed that low inflammation and dysplasia was observed for the Cpd188-treated group, while severe inflammation and high dysplasia were found in the vehicle-treated group (Fig. 5G). These results suggest that STAT3 activation in the cGas^−/− mice promoted colon cancer development.

The cGAS product cGAMP has been shown to exert strong antitumor effects in several mouse models bearing transplanted tumor cell lines (12). However, the effect of cGAMP on tumors arising from endogenous tissues has not been reported. The aggressive development of colon cancers induced by DSS and AOM in the cGas^−/− mice offers an opportunity to test the effect of cGAMP on endogenous tumors. In vitro stimulation with cGAMP could induce IFN-β and CXCL10 expression in both WT and cGAS-deficient BMDM (SI Appendix, Fig. S6A). Intraperitoneal injection of cGAMP could also induce IFN-α and IFN-β production in the crypts and spleen of cGas^−/− mice, indicating that cGAMP intraperitoneal administration induces not only local activation, but also systemic responses (SI Appendix, Fig. S6B). These results suggest that cGAMP might have potential therapeutic effects for colon cancer treatment of cGas^−/− mice. cGAMP was injected to mice intraperitoneally during the recovery phase of the DSS treatment cycle (SI Appendix, Fig. S6C). This treatment partially but significantly reduced the number of tumors in the colon when compared to untreated mice (Fig. 5H), providing evidence that cGAMP has therapeutic potential in tumors arising from an internal organ. The sizes of tumors were smaller and histological study showed lower dysplasia in cGAMP-treated cGas^−/− mice than PBS-treated mice (SI Appendix, Fig. S6 D and E).

T cell dysfunction is implicated in regulating colonic cancer development. Although there was a similar percentage of CD3⁺CD4⁺ and CD3⁺CD8⁺ T cells in the spleen of untreated WT and cGas^−/− mice (SI Appendix, Fig. S7A), cGas^−/− mice showed splenomegaly on day 80 when they developed advanced colon cancer (SI Appendix, Fig. S1 B and C). However, the percentage of CD3⁺CD4⁺ and CD3⁺CD8⁺ T cell subsets in the spleen did not show a significant difference between WT and cGas^−/− mice (SI Appendix, Fig. S7B). The populations of CD3⁺CD4⁺ and CD3⁺CD8⁺ T cells in the lamina propria were markedly higher in cGas^−/− mice compared to those in WT mice on day 80 after AOM injection (Fig. 6A). Both CD3⁺CD4⁺ and CD3⁺CD8⁺ T cells in the colon of cGas^−/− mice expressed higher levels of CD44, CD69, and PD-1 (Fig. 6 B and D), indicating more activated T cells in cGas^−/− mice during CAC. However, more CD4⁺CD25⁺ T cells were also found in the colon of cGas^−/− mice, suggesting an adaptive increase of regulatory T cells in response to colon inflammation. Further analysis of T cell population showed that significantly more memory CD4⁺ and CD8⁺ T cells but less naïve CD4⁺ and CD8⁺ T cells were present in the colon of cGas^−/− mice (Fig. 6 C and E). CD8⁺ T cells in tumor-bearing cGas^−/− mice showed higher production of Granzyme B, but not IFN-γ (SI Appendix, Fig. S7C).

To more clearly define which cell types infiltrate into the tumors, we compared immune cells within the tumor and adjacent normal tissues. As the tumor numbers and sizes in WT mice were too small, we only performed this comparison for cGas^−/− mice. Results showed that there were more CD4⁺ and CD8⁺ T cells infiltrating into tumors compared to adjacent normal tissues (SI Appendix, Fig. S7D). The CD4⁺ T and CD8⁺ T cells were mainly of the memory type (SI Appendix, Fig. S7D). However, both CD4⁺ and CD8⁺ T cells in tumors showed much higher PD-1 expression (SI Appendix, Fig. S7E), indicating T cell exhaustion may occur within tumors. These findings suggest that cGas^−/− mice have more inflammation in the colon and maintain an immune suppressive microenvironment that may promote colon cancer growth and development.

Higher numbers and percentages of myeloid-derived suppressor cells (MDSCs), which are known to suppress antitumor immunity, were detected in both the colon and spleen of cGas^−/− mice with CAC (Fig. 7A). In the tumors there were significantly higher numbers of MDSCs compared to adjacent normal tissues (SI Appendix, Fig. S7F). MDSC recruitment is regulated by chemokines, such as CXCL1 and CCL2 (36). Indeed, the expression levels of both CXCL1 and CCL2 were higher in the colon of cGas^−/− mice (Fig. 7 B and C).

Recent studies have shown that the Th17 cell population promotes colonic inflammation and cancer development (37). To explore whether cGAS deficiency promotes colon inflammation and cancer development by regulating T cell activation and differentiation, we isolated colon lamina propria cells, and measured cytokine production after stimulation with Phorbol 12-myristate 13-acetate and ionomycin. Compared to WT mice, cGas^−/− mice showed significantly more IL-17–producing CD4⁺ T cells and less IL-10–producing CD4⁺ T cells in the lamina propria during colon cancer development (Fig. 7D). However, IFN-γ–producing CD4⁺ T cells in the lamina propria were similar between WT and cGas^−/− mice (Fig. 7D). Similar populations of IL-17– and IL-10–producing T cells were present in the colons of WT and cGas^−/− mice in the untreated groups (SI Appendix, Fig. S7G). Further analysis showed that IL-17A mRNA levels were elevated in the colons of cGas^−/− mice compared to WT mice on day 80 after the AOM treatment, while serum IL-17A concentrations in cGas^−/− mice increased on day 15 and day 80 after the treatment (Fig. 7 E and F). Retinoid acid receptor-related orphan receptor-γ t (RORγt) is a Th17 cell differentiation lineage-specific transcription factor. Consistently, a higher percentage of RORγt⁺ population in CD3⁺CD4⁺ T cells from the colon of cGas^−/− mice was detected by flow cytometry on day 80 (Fig. 7G). The populations of IFN-γ–, IL-17–, and IL-10–producing CD4⁺ T cells in the spleen were similar between WT and cGas^−/− mice on day 80 (SI Appendix, Fig. S7H), suggesting that cGAS deficiency promotes Th17 cell development within the colonic local environment. The cGas^−/− mice developed splenomegaly, and the absolute numbers of total spleen cells, CD4⁺ T cells, CD8⁺ T cells, Th17⁺ cells and B cells in the spleen of cGas^−/− mice were higher than those in the WT mice (SI Appendix, Fig. S7I). However, for DSS-induced acute colitis, although there were higher percentages of CD3⁺CD4⁺ and CD3⁺CD8⁺ T cells in the colonic lamina propria of cGas^−/− mice compared to those of WT mice on day 8 (SI Appendix, Fig. S7J), the production of IL-10, IL-17, and IFN-γ in T cells was similar between WT and cGas^−/− mice (SI Appendix, Fig. S7K), suggesting that Th17 cell differentiation occurred at the later stage of the colon cancer development.

cGas^−/− mice were generated as previously reported (48). WT C57BL/6J, B6.SJL-Ptprc^a Pepc^b/BoyJ, B6(Cg)-Sting1tm1.2Camb/J, C57BL/6J-Sting1gt/J and Lgr5-EGFP-IRES-creERT2 mice were purchased from the Jackson Laboratory. The Apc(min/+) and IFNAR^−/− mice were kindly provided, respectively, by Joshua Mendell and David Farrar, University of Texas Southwestern, Dallas, TX. All of these strains were maintained on C57BL6/J background and bred in a specific pathogen-free facility. To rule out the contribution of microbiota in influencing colon inflammation and cancer development, WT and cGas^−/− mice were cohoused for at least 2 wk for all experiments; in some cases, where indicated, the mice were cohoused throughout the duration of the experiments. All study protocols were approved by the Institutional Care and Use Committee of the University of Texas Southwestern Medical Center and were conducted in accordance with the Institutional Animal Care and Use Committee guidelines and the NIH Guide for the Care and Use of Laboratory Animals (49). All the experiments were conducted with sex and age-matched mice and both male and female mice were included.

To induce CAC, mice were injected intraperitoneally with 8 mg of the mutagen AOM (Sigma-Aldrich) per kilogram body weight followed by three cycles of 1.5% or indicated concentration of DSS in drinking water for 5 d, with regular drinking water for 14 d between each cycle. Mice were euthanized on day 80 or when moribund. For day 15 samples, mice were injected with AOM, and after 5 d, they were fed with DSS for 5 d. Mice were then fed with regular water for 5 d before experimental analyses.

The epithelial cell dataset (GSE92332) was loaded to Seurat v3 (26, 50). Data were transformed to log₂(TPM+1) (TPM, transcripts per million) and scaled in Seurat. Principal components analysis was calculated based on top 2,000 variable features. Cells were plotted by t-distributed stochastic neighbor embedding (t-SNE) method and cell types were labeled based on the original annotations of ref. 26. scRNA-seq datasets from whole epithelium and enriched crypts (GSE123516) were read into Seurat and filtered based on the proportion of mitochondrial genes (percent.mt < 50) (27). Data were log-normalized and scaled in the Seurat. The resolution for cell clustering is based on the original paper. Cell types were labeled based on the gene-expression pattern shown in the Extended Data of ref. 27.

Data were analyzed by Prism7 (GraphPad Software) and statistical significance was determined by two-tailed Student t test unless specifically defined otherwise. P < 0.05 was considered statistically significant.