GPR15 Facilitates Recruitment of Regulatory T Cells to Promote Colorectal Cancer

Ongoing colorectal inflammation as it is seen in patients with ulcerative colitis (UC) is strongly associated with the development and progression of colorectal cancer (1, 2). One hallmark of cancer is the ability to evade the immune system via immunosuppression (3). Therefore, the ratio of pro- and anti-inflammatory immune cells in the tumor microenvironment influences the patients' clinical outcome (4, 5). Although high density of proinflammatory CD8⁺ cytotoxic T cells and CD4⁺ helper T cells type 1 (Th1) is clearly associated with a longer disease-free survival of patients with colorectal cancer (6), the role of FOXP3⁺ regulatory T cells (Treg) during colorectal cancer is still controversial. Tregs are key mediators of immunoescape and tumor progression, as they are potent suppressors of antitumoral immune responses (7). Nevertheless, accumulation of Tregs in the colon of patients with colorectal cancer was shown to correlate with both, better and worse, prognosis (8–11). The origin of tumor-associated Tregs, however, is difficult to distinguish, as both natural occurring Tregs (nTreg)—which originate in the thymus—and peripherally induced Tregs (iTreg) constitutively express the canonical Treg markers FOXP3 and CD25 (12). Many colorectal cancer studies on FOXP3⁺ T-cell infiltration have shown poor prognosis and lower disease-free survival, whereas other demonstrated that FOXP3⁺ T-cell infiltration is also associated with a favorable outcome (8–11). Interestingly, studies revealed that these discrepancies may be due to the heterogeneity of FOXP3 expression and the fact that Tregs can show an inflammatory effector T-helper cell phenotype, expressing FOXP3^loCD45RA⁺, TNFα, TGFβ, and IL12 (13). In mice, transient ablation of FOXP3⁺ Tregs has been successfully used to reduce colorectal cancer progression (7, 14), and targeting Tregs to overcome the suppression of antitumor immune responses is thus widely discussed as colorectal cancer therapy. However, most of the common therapies are unspecific and have many side effects, not least because Tregs compromise different subpopulations with different functions. Given that Tregs are potent suppressors of inflammation, but simultaneously inhibit antitumor immunity, the identification of markers that are exclusively expressed on tumor-associated Tregs would allow to specifically target these cells, without the risk of inducing unwanted side effects.

In this study, we report that the migration of nTregs—rather than the local conversion of naïve T cells into iTregs—underlies the high abundance of tumor-associated Tregs in the colon of colorectal cancer mice. In addition, tumor-associated Tregs were characterized by a unique expression of migration receptors. Specifically, we identified G protein–coupled receptor 15 (GPR15), an unconventional chemokine receptor, preferentially found in the intestinal mucosa, as an important target for the recruitment of Tregs into the colonic tissue during intestinal tumorigenesis. Finally, Gpr15 deficiency reduced the infiltration of tumor-associated Tregs in the tumor microenvironment, which boosted antitumor immunity and diminished colorectal cancer development.

All mice were 8 to 12 weeks old (mixed gender and age matched) when experiments were initiated. Mice were bred and housed in accordance to the guidelines of the Laboratory Animal Facility of the University Hospital Essen. Animal experiments were performed in accordance to the ethical principles and federal guidelines, and approved by the Landesamt für Natur, Umwelt und Verbraucherschutz (LANUV, Germany). BALB/c and C57BL/6 mice were obtained from Envigo RMS GmbH. C.Cg-Foxp3^tm2Tch/J (FOXP3-GFP; BALB/c background), Gpr15^tm1.1Litt/J [Gpr15^+/+ (wild-type, WT); Gpr15^gfp/+ (GPR15-GFP, control littermates); Gpr15^gfp/gfp (knockout, KO); ref. 15], and Foxp3^tm1Flv/J (FOXP3-mRFP) mice (both C57BL/6 background) were obtained from The Jackson Laboratory. Crossing GPR15-GFP mice to FOXP3-mRFP mice resulted in GPR15-GFP x FOXP3-mRFP double-reporter mice.

Blood samples were obtained from 13 healthy volunteers and from 19 patients with colon cancer. Tissues from tumorous and healthy adjacent colon were provided from 11 patients with colorectal cancer. Informed written consent was obtained from all patients. Ethical approval was provided by the Medical Faculty of the University of Duisburg-Essen (AZ 05-2882) and the Cantonal Ethics Committee of Bern (2017-01821 and 2018-01502). Fixed human colorectal cancer tissues and part of the fresh colorectal cancer samples were provided by the Tissue Bank Bern.

Colorectal cancer was induced using the azoxymethane (AOM)/dextran sulfate sodium (DSS) protocol as described previously (7). Tumor distribution from the first flexure to the anus was determined by murine colonoscopy, and tumor sizes were graded from 1 to 5 as described elsewhere (16). Tumor score per mouse was calculated by summing up the tumor sizes of all tumors in a given mouse. Lymphocyte emigration from the secondary lymphoid organs was blocked using FTY720 (Santa Cruz Biotechnology). At week 8, tumor score of colorectal cancer mice was determined by colonoscopy. Mice were separated into two identical groups, and 1 mg/kg FTY720 was injected i.p. twice a week until week 12.

C57BL/6 recipient mice were lethally irradiated with a single dose of 9 Gy, using an Isovolt-320-X-ray machine (Seifert-Pantak). Whole bone marrow (BM) cells (5 × 10⁶) from either GPR15-WT (Gpr15^+/+) or GPR15-KO (Gpr15^gfp/gfp) donor mice were adoptively transfused via i.v. injection, and AOM/DSS treatment of chimeras was initiated 8 weeks after BM reconstitution.

Single-cell suspensions from the spleens and mesenteric lymph nodes (mLN) were prepared by mashing the organs through a 70 μm cell strainer using PBS/2 mmol/L EDTA/2% FCS (PAA Laboratories). Mashed spleens were additionally pretreated with erythrocyte lysis buffer. Murine lamina propria lymphocytes from the intestines were isolated as described previously (7). Blood samples from patients were collected in NH4-Heparin Monovette tubes (Sarstedt). Peripheral blood mononuclear cells (PBMC) were isolated using Biocoll density gradient (Biochrom), washed with PBS/2 mmol/L EDTA/2% FCS, and stored in FCS/10% DMSO (Carl Roth). Informed written consent was obtained from all patients.

Single cells were incubated with fluorochrome-labeled anti-mouse or anti-human antibodies (see Supplementary Table S1). Fixable viability dye was used to stain for dead cells. GPR15 expression was determined using GPR15-GFP x FOXP3-mRFP mice, and after surface staining, cells were fixed with 2% formaldehyde (RotiHistofix; Carl Roth). Intracellular staining of FOXP3, Helios, Ki67, and granzyme B (GZMB) was performed using the eBioscience FOXP3 staining buffer set. To assess IFNγ, IL17, IL10, and TNFα, cells were stimulated for 4 hours with 10 ng/mL PMA and 1 μg/mL ionomycin in the presence of 5 μg/mL Brefeldine A and Monensin (1x; Thermo Fisher Scientific) in complete media (IMDMc; IMDM/10% FCS/2.5 μmol/L β-Mercaptoethanol/100 μg/mL penicillin/streptomycin; all Sigma-Aldrich). After surface staining, cells were fixed with 2% paraformaldehyde, permeabilized with 0.1% NP-40, and stained with antibodies against different cytokines. Flow cytometry analysis of the cells was performed using FACS DIVA software on an LSR II, CANTO, CELESTA, or SYMPHONY instrument; cell sorting was performed on a FACS ARIA (all BD Biosciences).

DNA was isolated from sorted CD4⁺FOXP3⁺ Tregs of healthy and colorectal cancer FOXP3-GFP male mice using QIAamp DNA Mini Kit (QIAGEN). Bisulfite modification of DNA was performed with the BisulFlash DNA Modification Kit (Epigentek). Methylation-sensitive real-time PCR was performed as described elsewhere (17). Primer sequences are listed in Supplementary Table S2. Analysis was run on a Roche LightCycler 480 System using Roche TaqMan Probe Master 480 (Roche Diagnostics).

Gene expression analysis of FACS-sorted FOXP3⁺CD4⁺ Tregs from the colonic lamina propria of healthy and colorectal cancer mice was performed as described elsewhere (18). For DNA microarray analysis of GPR15⁺ and GPR15⁻FOXP3⁺ Tregs, 5,000 cells were directly sorted into single-cell lysis buffer (Thermo Fisher Scientific). Approximately 2 ng of total RNA was used for biotin labeling according to the GeneChip Pico Kit (Affymetrix). Biotinylated cDNAs (5.5 μg) were fragmented and placed in a hybridization cocktail containing four biotinylated hybridization controls (BioB, BioC, BioD, and Cre). Samples were hybridized to an identical lot of Affymetrix Clariom S (400 Format) for 17 hours at 45°C. Hybridization was done for 16 hours at conditions recommended by the manufacturer. Clariom S chips were washed and stained in the Affymetrix Fluidics Station 450. GeneChips were scanned by Affymetrix GCS 3000. Image analysis was done by GeneChip Command Console Software (AGCC) and Expression Console Software (both Affymetrix). Data analysis was performed as described elsewhere (19). Data have been deposited in NCBI's Gene Expression Omnibus and are accessible through GEO Series accession number GSE168744 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE168744).

RNA from murine colonic tissues was isolated using the RNeasy Fibrous Tissue Mini Kit (QIAGEN). RNA isolation from sorted single-cell suspensions was performed using the NucleoSpin XS Kit (Macherey-Nagel). Reverse transcription of RNA was performed using M-MLV reverse transcriptase (Promega). Quantitative real-time PCR analysis was performed with an ABI PRISM cycler (Applied Biosystems), using the Fast SYBR Green Master Mix (Thermo Fisher Scientific) and specific primers (see Supplementary Table S2). Relative mRNA levels were calculated with included standard curves for each gene and normalization to the housekeeping gene RPS9.

A small explant (15–25 μg) from the distal part of the colon was cultured for 6 hours in IMDMc. Cytokine levels in the supernatants were measured by Luminex technology (R&D Systems) on a Luminex 200 instrument using the Luminex IS software (Luminex Corporation). Cytokine concentration was normalized to the respective weight of the colon biopsies.

GPR15 expression was induced on sorted CD4⁺GPR15⁻ T cells as described previously (15). After 5 days, CD4⁺GPR15⁻ and CD4⁺GPR15⁺ T cells were sorted and subjected to a migration assay using Transwell chambers (Corning). SDF-1β (50 ng/mL; PeproTech)/IMDMc was added to the bottom chamber. GPR15⁻ or GPR15⁺ cells (5 × 10⁵) were added to the upper insert and incubated for 4 hours at 37°C, after which, the migrated cells were collected from the lower chamber and counted. Migration index was calculated as the ratio of migrated cells toward SDF-1β in comparison with cell migrated to media alone.

GPR15 expression was induced on sorted CD4⁺ T cells of GPR15-KO mice (Gpr15^gfp/gfp) or control littermates (ctrl, Gpr15^gfp/+) as described previously (15). At day 5, ctrl T cells were stained with cell proliferation dye eFluor 670 (Thermo Fisher Scientific), and KO T cells were stained with cell tracker blue CMAC (Thermo Fisher Scientific), mixed at a 1:1 ratio, and 1 × 10⁷ mixed T cells were transferred i.v. into colorectal cancer–bearing C57BL/6 recipients. Twenty hours later, cells from the blood, spleen, mLN, and the colonic lamina propria were isolated, numbers of KO and ctrl T cells were analyzed via flow cytometry, and the ratio of migrated GPR15-expressing versus GPR15-deficient cells was calculated for each compartment.

Human colorectal cancer tissues were fixed in 4% formaldehyde and embedded in paraffin. Staining reactions were performed by automated staining using a BOND RX autostainer (Leica Biosystems). Sections were first deparaffinized, and antigen was retrieved using 1 mmol/L Tris solution (pH 9.0) for 40 minutes at 95°C. Sections were then stained with anti-human GPR15 (1:20; 60 minutes) and anti-human FOXP3 (1:200; 15 minutes) antibodies. Antibody binding was visualized using horseradish peroxidase and 3,3-diaminobenzidine (brown chromogen), or Fast Red (red chromogen), respectively (all Leica Biosystems). The samples were counterstained with hematoxylin and mounted with Aquatex (Merck). Slides were scanned on whole slide scanners Pannoramic 250 Flash (3DHISTECH) or NanoZoomer S360 (Hamamatsu). Tissue microarray images were dearrayed using ImageJ (Version 1.52p) and the color deconvolution plugin, resulting in individual channels for GPR15 and FOXP3. Specific threshold was set for GPR15 and FOXP3, and images were pseudocolored in green (GPR15) and red (FOXP3), respectively, after which, images were merged.

All analyses were calculated using the GraphPad Prism 7.03 software. Where appropriate, the paired or unpaired Student t test or Mann–Whitney test was used. Differences between means of more than two groups were assessed using one-way ANOVA followed by Dunnett multiple comparison test. Statistical significance between two groups with different factors was calculated using two-way ANOVA followed by Bonferroni posttests. Statistical significance was set at P < 0.05.

Chronic inflammation of the colon is associated with the induction of colorectal cancer (20). Using a murine model of colorectal cancer based on AOM/DSS (Fig. 1A), we determined intestinal dysplastic changes over time in correlation with colonic Treg frequencies. As early as 4 to 5 weeks after AOM administration, first dysplastic changes in the colon were identified by colonic endoscopy, and a significant increase in tumor volumes was detected between weeks 7 and 12, as summarized in the tumor score (Fig. 1B). Concurrently, we observed enhanced frequencies of CD4⁺FOXP3⁺ Tregs at weeks 2 and 5, most likely due to acute inflammation of the colon caused by the DSS administration. From week 7 onward, the frequency of Tregs in the tumorous tissue increased steadily when compared with healthy colonic tissue, whereas the frequencies from healthy control (HC) mice increased only minor from weeks 1 to 12 (Fig. 1C). Further analysis revealed a positive correlation of the tumor score and colonic Treg frequencies (Fig. 1D).

We previously established that Tregs are strongly involved in the progression of colorectal cancer in mice, as the transient ablation of tumor-associated Tregs improved CD8⁺ T-cell–mediated antitumoral immunity (7). Yet, detailed knowledge of the mechanisms underlying Treg accumulation in the colon of colorectal cancer mice remained unclear. To define the origin of tumor-associated Tregs, we next analyzed the methylation status of the foxp3-Treg–specific demethylated region (foxp3-TSDR). Methylation of this genetic region is associated with unstable expression of FOXP3 and commonly found in iTregs or effector T cells, whereas this region is mainly demethylated in thymus-derived nTregs (21). For this, we sort-purified FOXP3-GFP⁺ Tregs from spleen, mLN, and colon of HC and colorectal cancer FOXP3-GFP reporter mice. As expected, sorted Tregs from the spleen and mLN were mainly demethylated in the foxp3-TSDR, confirming the predominantly thymic origin of these cells. This phenotype was independent of colorectal cancer (Fig. 1E). Interestingly, Tregs from the colons of healthy mice were mostly methylated in the foxp3-TSDR, suggesting a high frequency of intestinal iTregs at steady state (Fig. 1E). On the contrary, colorectal cancer–associated Tregs showed a demethylated status (Fig. 1E), hinting at a stable expression of FOXP3 and an nTreg phenotype. The thymic origin of tumor-associated Tregs was further confirmed by an increased expression of the nTreg markers Neuropilin-1 (NRP1) and Helios (22, 23) on tumor-associated Tregs compared with colonic Tregs from control mice (Fig. 1F and G). To get an idea whether the expansion of nTregs in the tumor tissue is based on migration or proliferation, we performed flow cytometry staining for the proliferation markers Ki67 and BrdU. Of note, slightly enhanced expression of Ki67 and BrdU was detected in tumor-associated Tregs compared with Tregs isolated from the colon of HC mice (Fig. 1H; Supplementary Fig. S1A and S1B). However, Ki67 and BrdU expression was also moderately higher in FOXP3⁻ conventional CD4⁺ T cells (cCD4⁺ T cells) of colorectal cancer mice compared with HC mice (Supplementary Fig. S1C and S1D). Interestingly, we found increased expression of Treg-associated survival factors, such as IL2, TNFα, and TGFβ, in the colon of colorectal cancer mice compared with HC mice (Supplementary Fig. S1E), but the overall survival of Tregs in the tumorous tissue was rather decreased (Supplementary Fig. S1F), hinting for a balance in proliferation and apoptosis of tumor-associated Tregs. Altogether, these data indicate that colorectal cancer–associated Tregs have a thymus-derived origin, which suggests migration of nTregs to and/or proliferation of nTregs in the tumor tissue, rather than a local induction of iTregs.

To further address whether Treg migration contributes to Treg accumulation in the colonic tumorous tissue, we next used FTY720 to block the emigration of lymphocytes from secondary lymphoid organs (24) during the time of actual tumor development in AOM/DSS-treated mice (Fig. 2A). Efficacy of FTY720 treatment was verified by analyzing circulating CD4⁺ and CD8⁺ T cells (Fig. 2B). Interestingly, FTY720-treated animals showed reduced tumor scores at time of sacrifice (Fig. 2C). Concomitantly, frequencies and numbers of Tregs in the colons were reduced in FTY720-treated colorectal cancer mice but not in HCs (Fig. 2D). Remarkably, no or only minor alterations in CD8⁺ and cCD4⁺ T-cell frequencies and numbers were detected in the colonic lamina propria of FTY720-treated and nontreated groups (Fig. 2E; Supplementary Fig. S2A and S2B). Nevertheless, in colorectal cancer mice, FTY720 treatment reduced the Treg/CD8⁺ T-cell ratio and promoted the functional activation of cytotoxic CD8⁺ T cells (Fig. 2F and G). Thus, our data indicate that Treg migration into the colon promotes colorectal cancer tumorigenesis, likely via a shift in the Treg/CD8⁺ T-cell ratio, thereby supporting a protumorigenic milieu.

Nonspecific blockade of Treg migration reduced the tumor burden in colorectal cancer mice and enhanced antitumoral immunity. To gain further insights into the molecules supporting nTreg migration to colorectal cancer lesions, we next performed global gene expression profiling. FOXP3-GFP⁺ CD4⁺ Tregs as well as FOXP3-GFP⁻ cCD4 were sort-purified from the colon of HC and colorectal cancer FOXP3-GFP reporter mice and subjected to microarray analysis. Focusing our analysis on molecules associated with migratory properties, we identified a unique expression pattern on Tregs isolated from colonic tumors (Fig. 3A). In particular, tumor-associated Tregs showed upregulated expression of αv integrins (Itgav) as well as the chemokine receptor Ccr5, both on transcript and protein levels (Fig. 3A and B; Supplementary Fig. S3A and S3B). Of note, enhanced αv integrins and CCR5 protein expressions were restricted to cells isolated from the colon of colorectal cancer mice, but the expression was not specific for Tregs (Fig. 3B; Supplementary Fig. S3A and S3B). Surprisingly, GPR15 was one of the molecules specifically upregulated on tumor-associated Tregs (Fig. 3A). In the past few years, GPR15 gained strong interest in the field of mucosal immunology as a homing receptor for Tregs mainly expressed in the lamina propria of the large intestine (15). Flow cytometry analysis using GPR15-GFP x FOXP3-mRFP double-reporter mice revealed that among CD4⁺ T cells, GPR15 was preferentially expressed in FOXP3⁺ cells and more in colonic tumor–associated Tregs than in counterparts from control mice (Fig. 3B and C; Supplementary Fig. S3A). Interestingly, no differences in the expression of GPR15 on a per cell basis were found on GPR15-expressing Tregs (Supplementary Fig. S3B), nor in the expression of GPR15 on colonic CD8⁺ T cells (Supplementary Fig. S3C–S3E) from HC and colorectal cancer mice. In other organs, such as the small intestine, mLNs, and spleen, we did not observe differences in GPR15 expression between control and colorectal cancer mice, and nearly no expression of GPR15 was detected on FOXP3⁻ cCD4⁺ T cells (Supplementary Fig. S4A and S4B).

As GPR15 was recently described as a homing receptor for Tregs to the large intestine (15), we next interrogated the possible role of GPR15 for colorectal cancer. First, we analyzed whether GPR15 expression in general affects T-cell migration. Therefore, GPR15 was induced in vitro by stimulation of CD4⁺ T cells in the presence of IL2, IL21, retinoic acid, and TGFβ (15), and sort-purified GPR15⁺ and GPR15⁻ T cells were subjected to migration assays. Our data revealed that GPR15-expressing cells showed a higher migration capacity than GPR15⁻ cells (Fig. 4A). Next, we performed GPR15 expression profiling on different immune cell subsets isolated from the colonic lamina propria of colorectal cancer mice. In the context of colorectal cancer, GPR15 was predominantly found on FOXP3⁺ Tregs, whereas only very low or no GPR15 expression was observed on CD8⁺ T cells, B cells, macrophages (Mph), dendritic cells (DC), and cCD4⁺ T cells (Fig. 4B). Furthermore, levels of Gpr15 in the colonic mucosa increased over the course of the AOM/DSS treatment compared with naïve conditions (Fig. 4C). Interestingly, Gpr15 expression correlated with both the tumor score and the numbers of Tregs (Fig. 4D).

To gain further insights on the contribution of GPR15 to the phenotype of tumor-associated Tregs, we performed gene expression analysis of sort-purified GPR15⁺ and GPR15⁻ FOXP3⁺ Tregs from the colon of colorectal cancer mice. Compared with GPR15⁻ counterparts, GPR15⁺ colorectal cancer–associated Tregs showed enhanced expression of 84 genes (Fig. 4E and F). Interestingly, expressions of Ccr5 and Sell (L-selectin or CD62L), both of which being genes related to T-cell migration, were upregulated in GPR15⁺ Tregs. Additional genes, such as Rorc (Th17 differentiation), Klrc1 and Lag3 (immune checkpoint), Tnfrsf1a and Il1r1 (cytokine–cytokine receptor interaction), also showed increased expression in GPR15⁺ tumor-associated Tregs, whereas Il1rl1 (IL33 receptor) was downregulated in GPR15⁺ compared with GPR15⁻ tumor–associated Tregs (Fig. 4E and F). In accordance with the gene expression data, we observed enhanced IL17 and TNFα secretion of GPR15⁺ compared with GPR15⁻ Tregs, whereas we found no differences in IFNγ and IL10 production between these cell subsets (Fig. 4G). Taken together, these results indicate that colorectal cancer–derived GPR15⁺ Tregs are phenotypically distinct, suggesting that they have distinct functional properties.

To examine the general relevance of our results from mouse studies, we analyzed GPR15 expression in Tregs from blood and colon tissue samples from patients with colorectal cancer. Frequencies of FOXP3⁺CD4⁺ Tregs and of GPR15-expressing Tregs were enhanced in the blood of patients with colorectal cancer compared with HC donors (Fig. 5A and B), implying that circulating Tregs have a differential expression pattern for GPR15 in patients with colorectal cancer. Unlike our murine data, we also observed enhanced frequencies of GPR15⁺ cCD4⁺ T cells in patients with colorectal cancer, yet the average percentage of GPR15-expressing cells was higher among Tregs (Fig. 5B). Next, we studied immune cell infiltrates in colon cancer tissues. Compared with adjacent tumor-free tissue, FOXP3⁺CD4⁺ Tregs accumulated in colorectal cancer lesions (Fig. 5C). In addition, the numbers of GPR15⁺ Tregs were significantly increased in colorectal cancer tissues, contrarily to GPR15⁺ cCD4⁺ T cells (Fig. 5D). To strengthen these flow cytometry data, we performed double immunohistochemistry stainings of human colorectal cancer tissues. For better distinction between FOXP3⁺ (purple, nuclear) and GPR15⁺ (brown, cytoplasmic) cells, images were dearrayed through pseudocoloring via color deconvolution. Using this approach, we were able to differentiate GPR15⁺ cells (green), FOXP3⁺ (red), and GPR15⁺FOXP3⁺ cells (yellow). In line with the flow cytometry data, GPR15 was detected on both, FOXP3⁺ and FOXP3⁻, cells in colorectal cancer tissues (Fig. 5E). Of note, not all FOXP3⁺ cells expressed GPR15 and vice versa. In accordance with our data from the mouse model, we observed enhanced secretion of IL17 and TNFα in GPR15⁺ Tregs compared with GPR15⁻ Tregs from patients with colorectal cancer, as well as decreased production of IFNγ (Fig. 5F). In summary, these data suggest that GPR15-expressing Tregs modify the tumor microenvironment in human colorectal cancer.

To assess the impact of GPR15 on Treg migration during colorectal cancer progression, we applied AOM/DSS treatment to GPR15-KO mice (Gpr15^gfp/gfp; KO) and compared them with GPR15-GFP control littermates (Gpr15^gfp/+; ctrl; ref. 15). We confirmed Gpr15 deficiency in these mice by comparing Gpr15 expression in the colon of GPR15-KO versus control littermates (Fig. 6A). Endoscopic analysis demonstrated that Gpr15 deficiency results in reduced intestinal tumor burden (Fig. 6B). Importantly, while at steady state, the frequencies of Tregs did not differ in the colon of control versus KO mice, Gpr15 deficiency resulted in a lower proportion of Tregs in colorectal cancer lesions (Fig. 6C and D). Along with the reduced frequencies of Tregs, the Treg/CD8⁺ T-cell ratio was lower in GPR15-KO colorectal cancer mice compared with colorectal cancer control littermates (Fig. 6E). In addition, we noticed enhanced functional capacity of CD8⁺ T cells from GPR15-KO tumors, as illustrated by a larger proportion of GZMB⁺ CD8⁺ T cells (Fig. 6F). Recently, we described a connection between IL33 secretion from colonic explants and extent of colorectal cancer in AOM/DSS-treated mice (18). Interestingly, we found here that the secretion of this protumorigenic cytokine was much lower in cancerous tissues from GPR15-KO than in control colorectal cancer tissues (Fig. 6G).

Although colorectal cancer induction is triggered by inflammatory processes and GPR15 is discussed to be involved in chronic intestinal inflammation (25, 26), during acute DSS colitis no differences were found between GPR15-KO and control littermates, neither in disease progression, nor in Treg frequencies and proliferation (Supplementary Fig. S5A–S5D). These results support the notion that GPR15 is rather important during the later phase of the AOM/DSS model, when tumors develop and Tregs accumulate in the colonic tissue.

To verify that the reduction of Tregs found in GPR15-KO mice was dependent on reduced migration of GPR15-expressing cells into the colon, we performed a short-term in vivo competitive migration experiment. We adoptively transferred equal amounts of fluorescently labeled GPR15-expressing and GPR15-KO cells into C57BL/6 colorectal cancer–bearing mice, and then analyzed the distribution of these cells in the different organs. We detected significant more GPR15-expressing than GPR15-deficient cells in the colon of colorectal cancer mice, whereas only minor discrepancies were observed for the other compartments, such as blood, spleen, and mLN (Fig. 6H). Finally, lethally irradiated C57BL/6 mice were reconstituted with BM from GPR15-KO or GPR15-WT (Gpr15^+/+) littermates, and colorectal cancer was induced via AOM/DSS treatment. Analysis of blood leukocytes revealed that the BM reconstitution was successful, as indicated by a clear reduction of GPR15-expressing CD4⁺ T cells in mice reconstituted with GPR15-KO BM (KO => C57BL/6) compared with control chimeras (WT => C57BL/6; Fig. 6I). Chimeras with Gpr15 deficiency in hematopoietic cells showed reduced tumor growth compared with control BM chimeric mice with normal GPR15 expression, which was accompanied by diminished colonic Treg frequencies and numbers (Fig. 6J and K).

In summary, our data demonstrate that GPR15 expression promotes colorectal cancer by facilitating the recruitment of Tregs into developing colorectal cancer lesions, which in turn impedes the establishment of an effective antitumoral CD8⁺ T-cell response.

Treg frequencies in colonic lesions have strong prognostic properties for patients with colorectal cancer. Targeting Tregs in the colorectal cancer tumor microenvironment might therefore be an effective antitumor therapy. However, understanding the specific features of tumor-associated Tregs is strictly necessary to design targeted tumor-specific Treg therapies. In this study, we observed a positive correlation between colonic tumor burden and Treg frequencies in mice and humans with colorectal cancer. Moreover, we identified that tumor-associated Tregs are of natural origin and belong to a distinct subpopulation of Tregs, characterized by specific expression of GPR15, participating in the overall enhanced migratory capacity of these cells.

In many types of solid cancers, Tregs outnumber effector T cells in the tumor microenvironment (27–30). Although the inflammatory milieu of a tumor may drive the differentiation of CD4⁺ helper T cells into FOXP3⁺ iTregs, it has been clearly shown by Malchow and colleagues that tumors may also drive the recruitment of preexisting, self-specific nTregs reactive to either ubiquitous or tissue-restricted antigens (31, 32). The extent to which this occurs during colorectal cancer remains poorly described. At steady state, it is considered that about 50% of all Tregs in the colon are peripherally induced Tregs with established tolerance for microbial and nutrient antigens (33). This is well in line with our results, as we observed a strong methylation of the foxp3-TSDR locus in colonic Tregs of healthy mice. Contrary, colonic tumor–associated Tregs isolated from AOM/DSS-treated animals were mostly demethylated in the foxp3-TSDR, which strongly hints at a stable expression of FOXP3 and an nTreg phenotype. In patients with colorectal cancer, Zhou and colleagues indeed found high frequencies of Tregs with demethylated foxp3-TSDR in the tumor, but also in the adjacent normal tissues. Interestingly, especially the demethylation of foxp3-TSDR in Tregs of the adjacent normal tissues correlated with worse survival rates (34). These results raise the question, whether nTregs are recruited to the tumor tissue and/or whether nTregs locally proliferate in response to tumor antigens. As blocking the emigration of Tregs from secondary lymphoid organs reduced the frequencies of tumor-associated Tregs, we conclude that during inflammation-induced colorectal cancer, lineage-stable nTregs are recruited into developing tumors. Interestingly, proliferation in tumor-associated Tregs was slightly increased. However, comparing Treg proliferation over the course of colorectal cancer tumorigenesis, this proliferation appears only to be moderate (Supplementary Fig. S1A). Therefore, we conclude that this moderate rise in proliferation cannot solely explain the strong increase in Treg frequencies we observed in colorectal cancer tissues and rather a combination of both recruitment and expansion of highly restricted nTregs contribute to the enhanced Treg pool in colorectal cancer lesions in our model. To further unravel the exact antigen specificity of tumor-associated Tregs, high-throughput analysis on T-cell receptor diversity of colorectal cancer–associated Tregs remains to be elucidated in future experiments.

Treg trafficking is regulated by their ability to cross the tumor endothelium, which is promoted by various combinations of chemoattraction and adhesion signals mediated through the expression of distinct chemotactic receptors (35, 36). For example, the expression of sphingosine 1 phosphate receptor 1 (S1PR1) on Tregs was shown to be crucial for tumor infiltration (37). In accordance, we observed enhanced expression of S1pr1 on tumor-associated Tregs (Supplementary Fig. S6), suggesting that also in our colorectal cancer model, Treg migration into the tumorous tissue might be influenced by S1PR1 signaling. Accordingly, blocking S1PR1 signaling by FTY720 mainly influenced Treg infiltration in our system, highlighting the importance of Treg migration to promote colon cancer. In addition, we found a unique transcriptional pattern of migration molecules for colorectal cancer–associated Tregs, including expression of classical migration molecules such as CCR5, CCR8, αv, and β8 integrins. Nevertheless, enhanced CCR5 and integrin expression was not restricted to Tregs but was also found on effector T cells during colorectal cancer. One molecule, which was specifically expressed on colonic Tregs and especially on tumor-associated Tregs in mice, was GPR15. GPR15 was firstly described as an HIV coreceptor with structural homology to other known chemokine receptors (38). Meanwhile, GPR15 was also identified as a receptor involved in the migration of Tregs to the colon but not to the small intestine in mice (15). Recently, we and others demonstrated that GPR15-expressing human T cells show an enhanced migration capacity and that GPR15 expression is altered on CD4⁺ T cells of patients with UC (25, 26). In agreement with these findings, we here describe a contribution of GPR15-expressing Tregs to the development and progression of colorectal cancer. Gpr15 deficiency in the context of AOM/DSS treatment resulted in decreased colonic Treg counts that changed the Treg/CD8⁺ T-cell ratio and a reduced tumor burden. Our data show that this was mediated by immune cells, as mice with Gpr15 deficiency specifically in the hematopoietic compartment showed a similar phenotype than mice lacking GPR15 in all cells. Of note, the intratumoral ratio of Treg/CD8⁺ T cell is considered as a crucial prognostic factor for many types of cancers (39, 40). Considering that the majority of the immune cells expressing GPR15 in mice are Tregs, we conclude that GPR15 expression directly affects Treg infiltration during colorectal cancer, supporting a protumorigenic microenvironment and tumor growth.

GPR15 expression is influenced by the colonic microenvironment, as TGFβ and microbial short chain fatty acids induce GPR15 expression on Tregs (15). The ligand for GPR15 (GPR15L) is expressed by epithelial cells of the gastrointestinal tract (41, 42). In addition, GPR15L is capable of inducing G₁ arrest and is therefore discussed as a potential inhibitor of colon cancer cell lines in vitro (43). So far, we could not detect any alteration in the expression of Gpr15L in the tumor microenvironment compared with healthy tissue (Supplementary Fig. S7A and S7B), providing evidence that not GPR15–GPR15L interaction alone but an additional mechanism might account for the increased influx of tumor-associated Tregs and their tumor-supportive capacity. Indeed, there are some functional differences between GPR15⁺ and GPR15⁻ cells. Ccr5 and Sell, both genes associated with T-cell migration, were upregulated on GPR15⁺ tumor–associated Tregs, emphasizing that GPR15-expressing Tregs have a preferable migratory capacity. In fact, we demonstrate that Gpr15-deficient cells have a lesser capacity than Gpr15-sufficient cells to migrate from the periphery into the tumorous colonic tissue. Interestingly, enhanced expression of CCR5 on tumor-associated Tregs has been already described, thus CCR5 has been suggested as a potential therapeutic target for Tregs in colorectal cancer (44–47). Ccr5^−/− mice show delayed tumor growth with an associated reduction in tumor Treg infiltration. However, pharmacologic inhibition of CCR5 failed to reduce tumor Treg infiltration in murine tumor models, although it did result in delayed tumor growth in mice and human (46, 48). These complex interactions clearly outline the difficulties of targeting single migration molecules in clinical trials and substantiate the need of unraveling the specific features of tumor-associated Tregs in colorectal cancer. Detailed knowledge on the exact mechanism how GPR15 expression guides Tregs into the tumor microenvironment, e.g., the expression and/or secretion of GPR15-attracting molecules by tumor cells and/or the tumor microenvironment, and the overall functional consequences of GPR15 expression, still remains to be clarified in more detail.

Generally, the suppressive capacity of Tregs does not seem to be affected by the expression of GPR15 (15). It is more likely that reduced frequencies and numbers of Tregs in colorectal cancer lesions account for the enhanced CD8⁺ T-cell–mediated cytotoxic immune response in GPR15-KO compared with control littermates. Still, besides its role in T-cell trafficking, GPR15 expression on T cells is associated with enhanced IL17 secretion (25, 49). Consistently, we found enhanced gene expression of Rorc and production of IL17 in GPR15⁺ tumor–associated Tregs compared with GPR15⁻ counterparts. Of note, RORγt-expressing tumor-infiltrating Tregs were shown to drive tumor growth of colorectal cancer by controlling IL6 secretion in DCs (50). Furthermore, IL17⁺FOXP3⁺ Tregs accumulate in colorectal cancer tissue, express CCR6, TGFβ, and IL6, and significantly suppress CD8⁺ T-cell–mediated immunity (51). Just recently, Xiong and colleagues described an intricate balance between the aryl hydrocarbon receptor (Ahr)–RORγt–Foxp3 axis in controlling Treg intestinal homing by regulating GPR15 expression under the steady state and during inflammation (52). In our study, we conclude that GPR15 expression firstly enables trafficking of Tregs into colonic tumor tissues and secondly identifies a potentially tumor-promoting Th17-like Treg population. Of note, we also found that compared with GPR15⁻ Tregs, GPR15⁺ Tregs secreted more TNFα, which has been reported to support Treg expansion, stability, and functions, thereby supporting tumor growth (53).

Recently, we demonstrated that ST2 expression modulates the phenotype of Tregs to promote intestinal cancer (18). Interestingly, gene expression of Il1rl1, which counts for the IL33 receptor ST2, was downregulated on GPR15⁺ Tregs compared with GPR15⁻ tumor–associated Tregs. Although we found coexpression of GPR15 and ST2 on tumor-associated Tregs in mice and humans, the majority of ST2⁺ Tregs did not express GPR15 (Supplementary Fig. S8A and S8B), providing evidence that GPR15⁺ Tregs belong to a specific migratory subpopulation of Tregs and that these cells, at least partially, might differ from ST2⁺ Tregs. Interestingly, Gpr15 deficiency resulted in reduced tumor-promoting IL33 secretion in the colon of colorectal cancer mice, hinting for an interaction between GPR15 expression and the IL33/ST2 pathway. Therefore, we assume that conjoint blockade of the GPR15-mediated migration of tumor-associated Tregs and of the IL33/ST2 inflammatory pathway might represent a valid therapeutic strategy to specifically tackle a large proportion of Tregs during colorectal cancer.

In conclusion, our findings support the idea that GPR15 presents a promising novel therapeutic target for the treatment of colorectal cancer and provide fresh insights into the complexity of Treg migration and the interaction to other immune cells during intestinal tumorigenesis.

S. Kasper reports grants and personal fees from Bristol-Myers Squibb, Merck Serono, Roche, and Lilly and personal fees from Amgen, Servier, MSD, and Bayer outside the submitted work. M. Schuler reports grants and personal fees from AstraZeneca, Boehringer Ingelheim, and Bristol-Myers Squibb and personal fees from Janssen, GlaxoSmithKline, MorphoSys, Novartis, Roche, Takeda, Amgen, and MSD outside the submitted work. C.M. Lange reports personal fees from Gilead, AbbVie, MSD, Eisai, and Roche outside the submitted work. No disclosures were reported by the other authors.

A. Adamczyk: Conceptualization, data curation, formal analysis, validation, investigation, visualization, methodology, writing–original draft. E. Pastille: Formal analysis, investigation, methodology. J. Kehrmann: Formal analysis, investigation, methodology. V.P. Vu: Formal analysis, investigation. R. Geffers: Software, formal analysis, methodology. M.-H. Wasmer: Formal analysis, investigation. S. Kasper: Resources, formal analysis, methodology. M. Schuler: Resources, writing–review and editing. C.M. Lange: Resources, methodology. B. Muggli: Resources, methodology. T.T. Rau: Formal analysis, validation. D. Klein: Methodology. W. Hansen: Resources, methodology, writing–review and editing. P. Krebs: Formal analysis, supervision, funding acquisition, writing–review and editing. J. Buer: Resources, writing–review and editing. A.M. Westendorf: Conceptualization, data curation, supervision, funding acquisition, writing–original draft, project administration.

The authors acknowledge Christian Fehring, Witold Bartosik Mechthild Hemmler-Roloff, and Christina Liebig for their excellent technical support. They are further grateful to Kristýna Hlavačková for experimental support, as well as to the team of the Translational Research Unit of the Institute of Pathology in Bern. They thank Daniela Catrini for the critical reading of the article. This work was supported by the BIOME-PEP program of the Medical Faculty, University of Duisburg-Essen (to A. Adamczyk), and the Helmut Horten Foundation (to P. Krebs).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.