C57BL/6, Apc^F/F (25), Cdx2-Cre (26) and Foxp3-YFP-Cre mice were obtained from the Jackson Laboratory. Il17ra ^F/F mice (19) were obtained from Dr. Michael Karin’s laboratory at the University of California, San Diego.

To generate the mouse model of sporadic CRC, Cdx2-Cre and Apc^F/F mice were crossed to generate Cdx2-Cre⁺ /Apc ^F/+ mice. These mice were sacrificed at around 5 months of age for tumor analyses. The mouse colon was dissected, and colorectal tumors were measured with a caliper and excised with scissors. Tumor load was represented as the sum of the diameters of all tumors from each animal. Statistical significance of tumor number, size, and load was determined by Student’s t test for pairwise comparison, with p value < 0.05 considered significant.

All mice were maintained in specific-pathogen-free conditions in filter-topped cages on autoclaved food and water at UConn Health. All experiments used co-housed, gender-matched littermates to ensure the consistency of common microflora. Both male and female mice were used for all experiments. All animal experiments were approved by the IACUC of UConn Health.

Six- to eight-week-old recipient mice were irradiated twice in one day to achieve a lethal dose (2 x 600 rad) and intravenously injected with a single-cell suspension of 10⁷ donor bone marrow cells. Recipients were co-housed littermates, which were transplanted with both gene-deficient (Foxp3-YFP-Cre ⁺ /Il17ra^F/F ) and control (Foxp3-YFP-Cre ⁺ /Il17ra^F/WT ) bone marrow for comparison. After transplantation the recipients were placed on sulfamethoxazole and trimethoprim in drinking water for two weeks, followed by regular water. Mice were sacrificed and analyzed for tumor development 3-4 months after transplantation.

Mouse colorectal tumors were minced with scissors and digested with 1mg/kg collagenase IV (Sigma Aldrich, Cat # C5138) for 20 minutes. Cells were filtered with a 70-μm cell sieve, and stained with Live/Dead fixable exclusion dye (Tonbo Bioscience, Cat # 13-0868), followed by fluorochrome-conjugated antibodies in PBS with 2% fetal bovine serum (FBS) and 1mM EDTA. Anti-CD3 (Cat # 100206), anti-CD4 (Cat # 100536), anti-CD45 (Cat # 103138), anti-IFN-γ (Cat # 505808) and anti-IL-17A (Cat # 506904) antibodies were from Biolegend. Anti-Foxp3 (Cat # 11-5773-82) antibody was from eBioscience. Anti-CD8α antibody (Cat # 558106) was from BD Bioscience. For intracellular cytokine staining, cells were stimulated with Cell Stimulation Cocktail (eBioscience, Cat # 00-4975-93) for 4 hours, followed with fixation and staining with Foxp3/transcription factor staining buffer set (eBioscience, Cat # 00-5523-00). Flow cytometry analyses were performed on a BD LSRII flow cytometer. Cell sorting was performed on a BD FACS ARIA II high-speed cell sorter. Data was analyzed with FlowJo software. Statistical significance between groups was determined by Student’s t-test for pairwise comparison (p < 0.05).

Total RNA was extracted with a RNeasy Plus kit (Qiagen, Cat # 74134) and reverse transcribed using an iScript kit (Biorad, Cat # 1708891). qRT-PCR was performed using SsoAdvanced Universal SYBR Green Supermix (Biorad Cat # 1725275) on a Biorad CFX96 machine. Expression data were normalized to Rpl32 mRNA levels. The data were calculated as 2^{(Ct(RPL32-gene of interest))} to compare experimental groups to controls, and are presented in arbitrary units. Students’ t test (p < 0.05) for pairwise comparison was used to determine significance. Primer sequences are listed in Table 1 . Whenever possible, primers were intron-spanning, such that amplification is feasible on complementary DNA.

Naïve T cells were purified from C57BL/6 mouse spleen using EasySep™ Mouse Naïve CD4+ T Cell Isolation Kit (Cat # 19765) from STEMCELL Technologies. 6 x 10⁵ Naïve T cells were cultured with anti-CD3 monoclonal antibodies (2 μg/mL, coated) and anti-CD28 monoclonal antibodies (2 μg/mL, soluble) in a 24-well plate. To induce differentiation of Treg cells, TGF-β (10 ng/mL, PeproTech) and IL-2 (20 ng/mL, PeproTech) were added to the mouse cell culture media with 10% FBS for 72 h, followed by the treatment of 5 ng/mL IL-2 with or without 50 ng/mL IL-6 for 24h. Cells were harvested for q-RT-PCR analysis.

Bone marrow cells from Foxp3-YFP-Cre⁺/Il17ra ^F/F (KO) and Foxp3-YFP-Cre⁺/Il17ra ^F/+ (WT control) mice were adoptively transferred into 6-8-week-old Cdx2-Cre⁺ /Apc ^F/+ mice. Bone marrow recipient mice were then sacrificed at 5 months of age, and their mesenteric lymph nodes (MLN) and tumors were digested with 1 mg/ml collagenase IV for 20 minutes. Cells were filtered with a 70-μm cell sieve and stained with Live/Dead fixable exclusion dye, followed by fluorochrome-conjugated antibodies in PBS with 2% fetal bovine serum (FBS) and 1mM EDTA. Live/CD45⁺/CD3⁺/CD4⁺/YFP⁺ cells were sorted and loaded for capture using the Chromium System with single-cell 3’ reagent kit v2 from 10X Genomics. After capture and lysis, cDNA synthesis and amplification were performed according to manufacturer protocol. Amplified cDNA was used to construct the Illumina sequencing library and sequenced on the NovaSeq 6000. Illumina basecall (.bcl) files were converted to FASTQ using bcl2fastq from the CellRanger pipeline. Fastq files were aligned to the mm39 mouse reference genome and transcriptome and cell-counts matrices were generated using 10x Genomics CellRanger v7.0 with default parameters. Single-cell RNA sequencing cell-counts matrices were analyzed using Seurat v4.3 (27) with RStudio v4.2. A merged Seurat object was generated, and QC was performed; retaining cells with >= 500 genes and <= 30% reads mapping to mitochondrial genes. Gene expression was normalized, scaled and dimensionality reduction was performed. UMAPs of clusters, experimental conditions and IL-17 receptor gene expression were generated using Seurat’s FeaturePlot function and the R package ggplot2. Dotplots, barplots and heatmaps were generated using the R package ggplot2. Differential gene expression was determined using Seurat’s FindMarkers function (adjusted p-value < 0.05, log2 Fold Change >= 0.25, min.pct >= 0.1).

RNA velocity in CRC Tregs from the MLN and tumor were analyzed using Velocyto (28). The output from CellRanger was analyzed using the run10x function from Velocyto with the GenCode mm39 primary assembly transcriptome and mm39 repeat masker file from the UCSC genome browser with all parameters as default. Using the loom file output from Velocyto, the trajectory UMAP of merged data was generated using the scVelo stochastic model (29) with the python modules scanpy v1.9.6 and scvelo v0.2.5.

Genes which were significantly differentially expressed (log2 Fold Change >= 0.25, min.pct >= 0.1, adjusted p-value < 0.05) were subject to pathway enrichment analysis using the R packages gprofiler2 (30), clusterProfiler (31), and enrichplot. Enrichment was conducted using the Gene Ontology database as the pathway source. Pathways with a fold enrichment >= 2 were rank sorted based on their significance (adjusted p-value < 0.05) and the top 20 enriched pathways were plotted as a barplot using ggplot2. The gene network of a subset of pathways was generated using the cnetplot function from enrichplot along with ggplot2.

All human scRNAseq datasets were obtained from the Gene Expression Omnibus (GEO). Pancreatic cancer data (32) was obtained from the GEO accession number GSE155698. Melanoma data (33) was obtained from GSE239750. Triple-negative breast cancer data (34) was obtained from GSE169246. Liver cancer data (35) was obtained from GSE98638. Non-small cell lung cancer data (36) was obtained from GSE99254. Colorectal cancer data (37) was obtained from GSE178341. Counts-matrices were merged, gene expression was normalized and scaled, and Tregs were subset according to the original author specification, where available, with confirmation by expression of Foxp3 in all cells. Conditions for each dataset (i.e. tumor, adjacent normal tissue, peripheral blood, etc.) were determined based on the original manuscripts’ specifications. The tumor Treg signature was generated by calculating the mean and standard error of the mean for the sum value of genes GADD45B, RGS2, NR4A1, TNFRSF9, CCL5, CXCR6, CREM, TNFRSF18, and TNFRSF4. The heatmap of the individual genes comprising the Treg signature was generated using ggplot2. Differential gene expression was determined using Seurat’s FindMarkers function (adjusted p-value < 0.05, log2 Fold Change >= 0.25, min.pct >= 0.1).

Similar to single-cell sequencing described above, MLN and tumor-infiltrating Treg cells were sorted, and their RNA was isolated using Qiagen RNeasy Plus Mini Kit. cDNA libraries were generated for poly-A selected RNA using the Illumina TruSeq mRNA Library Kit. 100 bp paired end reads were sequenced on the Illumina NextSeq 550 v2.5. Paired end reads in fastq format were aligned to the mm39 mouse genome and transcriptome using Hisat2 (38). Transcripts were assembled, gtf files for all conditions were merged, and read coverage tables were generated using StringTie (39) v2.2.1. Read counts were extracted using the prepDE python script from Stringtie, and normalization, calculation of FPKM values, and differential expression analysis were performed using the R package DESeq2 (40) (adjusted p-value < 0.05). Bar plots, heat scatter plots and heatmaps were generated using ggplot2 and coefficient of determination was calculated using the stats R package. Differential isoform usage, and splicing event enrichment were analyzed using the R packages DexSeq and IsoformSwitchAnalyzeR (41) Differential isoform usages were calculated with a dIF absolute value >= 0.1 and an adjusted p-value < 0.05. Splicing event enrichment was calculated using the extractSplicingEnrichmentComparison function from the R package isoformswitchanalyzeR with default parameters (alpha=0.05, dIFcutoff=0.1).

Statistical methods for RNA sequencing data analyses are described in previous sections. Otherwise, data were analyzed by the Student’s t-test for pair-wise comparisons. p values less than 0.05 were considered significant.

IL-17 family cytokines signal to diverse cellular targets, including epithelial cells, fibroblasts, and myeloid cells (15, 18). A direct signaling of IL-17 on Tregs is unknown, but we contend that it may be important for the function of Tregs and development of tumors. To this end, Foxp3-YFP-Cre mice were crossed to Il17ra ^F/F mice to specifically ablate IL-17RA in Tregs (Foxp3-YFP-Cre ⁺/Il17ra^F/F mice, hereby referred to as Il17ra ^ΔTreg mice). YFP transgene under the control of the Foxp3 promoter also serves as a marker of Tregs in vivo and was used to gate these cells for flow cytometry and cell sorting. To investigate how Il17ra KO impacts Tregs in CRC, we adoptively transferred bone marrow cells from Foxp3-YFP-Cre ⁺/Il17ra^F/+ (control) and Il17ra ^ΔTreg mice into recipient mice that develop sporadic CRC (Cdx2-Cre⁺ /Apc ^F/+). These bone marrow recipient mice are herein referred to as “control CRC” and “Il17ra ^ΔTreg CRC” mice, respectively. 3.5 months after bone marrow transfer, control and Il17ra ^ΔTreg CRC mice were sacrificed, and Tregs were purified by cell sorting from the mesenteric lymph node (MLN) and colorectal tumors and subjected to single-cell RNA sequencing (scRNAseq) ( Figure 1A ). Unsupervised clustering of MLN and tumor Tregs revealed a total of 9 clusters within the merged dataset ( Figure 1B ). MLN Tregs were predominantly comprised of naïve, primed, and gut homing/tissue-infiltrating Tregs, while colorectal tumor contains tumor activated Tregs, proliferating Tregs and two tissue-resident immunosuppressive Treg clusters ( Figure 1C ). Treg clusters were named based on analysis of differentially expressed genes ( Figure 1D ): Naïve and primed Tregs in the MLN were marked by the expression of the transcription factor Bach2, which represses effector T cell programs and is required for maintaining naïve T cell state ( Figure 1D ) (42, 43). Following priming, Bach2 was downregulated, along with upregulation of activation markers Lag3 (44), CD44 (45), as well as gut homing signatures Ccr9 (46) and Itgae ( Figure 1D ) (47, 48). In the tumor, Treg clusters displayed a gradual increase in the expression of secondary costimulatory receptor OX40 (Tnfrsf9) (49) and tissue resident Treg markers Areg, Tff1 and Penk ( Figure 1D ) (50–52). The proliferating cluster, comprised of Tregs primed in the MLN and Tregs activated in the TME, was marked by expression of Mki67, Stmn1 and Top2a ( Figure 1D ) (53–55). These data indicate distinct clustering of tumor infiltrating Tregs compared to Tregs in peripheral lymphoid tissues. To validate the progressive transcriptomic shift in Treg identity between clusters, we analyzed RNA velocity in the Treg subpopulations. RNA velocity infers the temporal dynamics of gene expression within a population based on the proportion of spliced versus unspliced transcripts, under the assumption that steady state cell identity will result in a higher proportion of spliced transcripts compared to cells undergoing dynamic shifts in cell identity. For example, differentiation or activation, will display an increased proportion of unspliced transcripts, as transcription precedes mRNA processing (29). RNA velocity revealed a global transcriptomic shift beginning with the naïve and primed Treg clusters in the MLN towards the proliferating and activated Treg subpopulations in the tumor, and finally towards the immunosuppressive tissue resident Treg clusters in the tumor ( Figure 1E ). These data demonstrate that naïve Tregs in the MLN undergo priming and transcriptomic changes allowing for their gut homing and infiltration of colon tumors. Once in the tumor, they are activated, proliferate, and undergo further maturation, allowing for their immunosuppressive function within the tumor microenvironment (TME).

IL-17 mediated tumor-associated inflammation has been shown to promote tumor development in the gut (20, 22, 23). The IL-17 receptor is comprised of 5 currently identified members (14). IL-17RA forms heterodimers with the other IL-17 co-receptors to bind IL-17 ligands. For IL-17 to be able to directly signal to Tregs, these cells should express both IL-17RA and its co-receptors. Thus, we analyzed the expression of IL-17 receptor subunits within the Treg subpopulations. Il17ra was ubiquitously expressed throughout the Treg clusters in both the MLN and tumor ( Figure 2A ). IL-17RA interaction with IL-17RC allows binding of IL-17A, IL-17F, or the IL-17A/F heterodimer (56, 57). IL-17RA/IL-17RB heterodimer binds IL-17B and IL-17E, and IL-17RE interacts with IL-17RA to bind IL-17C (58, 59). Both Il17re and Il17rc were expressed within the proliferating and suppressive 1 clusters of tumor Tregs, whereas Il17rb was expressed within suppressive cluster 2 of the tumor Tregs ( Figure 2A ). IL-17RD, which can heterodimerize with IL-17RA to bind IL-17A, was not reasonably expressed in any clusters of the Treg scRNAseq data ( Figure 2A ) (60, 61). Additionally, the noncanonical IL-17 receptor gene Cd93, whose protein acts as a receptor for IL-17D, was not reasonably expressed in the Treg data ( Figure 2A ) (16). Examining the cells which could form functional IL-17 receptor heterodimers, based on co-expression of Il17ra with either Il17rb/c/e, suggested that while Il17ra was expressed throughout the Treg clusters, the tumor Treg clusters, specifically, suppressive 1, 2, and proliferating clusters, predominantly represented the cells capable of expressing the necessary co-receptors for IL-17 signaling ( Figures 2B, C ). We hypothesized that IL-17 receptors may be upregulated specifically in tumor infiltrating Tregs due to inflammatory signals in the TME. Amongst inflammatory cytokines within the TME, IL-6 is highly expressed in colorectal tumors and has been shown to promote Th17 differentiation and IL-17RE expression (24). To investigate whether IL-6 could contribute to the expression of Il17re on Tregs, we stimulated in vitro differentiated iTregs with IL-6. While Foxp3 mRNA levels were unchanged following stimulation, we observed a significant Il17re upregulation (p < 0.001) as well as a modest, but statistically significant increase in the expression of Rorc (encoding RORγt) (p < 0.05) ( Figure 2D ). Thus, IL-6 secretion by inflammatory cells within the TME, may lead to the upregulation of co-receptors necessary for IL-17 signaling on Tregs.

As the tumor-infiltrating Tregs represented the population capable of responding to IL-17 signaling, we next analyzed the genes differentially expressed between WT vs Il17ra KO Tregs from tumors of control and Il17ra ^ΔTreg CRC mice ( Figure 3A ). Pathway enrichment analysis of genes significantly differentially expressed (log2 fold change >= 0.25, adjusted p-value < 0.05, min.pct >= 0.1) revealed that Il17ra KO tumor Tregs exhibited an increase in pathways related to translation, metabolism, and survival, with a decrease in pathways related to T cell function, namely leukocyte differentiation, integrin binding, and chemokine signaling ( Figure 3B ). Several highlighted pathways in Il17ra KO tumor Tregs ( Figures 3C, D ) show the upregulation of multiple ribosomal subunits, mitochondrial genes related to oxidative phosphorylation, and pro-survival signaling ( Figure 3D ). Oxidative phosphorylation has been shown to be crucial for Treg survival, especially in low glucose, high lactate environments, such as the intestinal tract (62, 63). Il17ra KO tumor Tregs also displayed a downregulation in established IL-17 signaling-related genes which are important for Treg function, including Junb (64, 65), Stat3 (66, 67) and genes related to MAPK cascade (68, 69) ( Figure 3C ). Secondary costimulatory molecule genes upregulated following T cell activation, Tnfrsf9 (4-1BB) (70, 71), Tnfrsf4 (OX40) (49, 72, 73), and Tnfrsf18 (GITR) (74) were also downregulated in Il17ra KO Tregs in the tumor ( Figure 3C ). Also downregulated were genes essential to Treg tissue homing and suppressive function including the αVβ8 integrin receptor genes Itgav and Itgab (75, 76), as well as CD81 (77) and the transcription factor Nr4a3 (78) ( Figure 3C ). A full list of DEGs can be found in Supplementary Table 1 . We also analyzed the subset of Tregs which express IL-17RA co-receptor genes Il17rb/rc/re ( Supplementary Figure 1A ) and found a similar pattern of upregulated and downregulated genes ( Supplementary Figure 1B ). Also included amongst these DEGs were suppression-associated genes Ctla4 (79), Gzmb (80), Ikzf2 (encoding the transcription factor Helios) (81) and Tnfrsf18 (GITR) (82), tissue-resident Treg markers Areg, Penk, Tff1, and Odc1, and activation/maturation-associated genes Tnfrsf4, Tnrfsf9, Mt1, and Got1 ( Supplementary Figures 1B, C ). These data suggest that while Il17ra KO Tregs display an increased capacity for survival and mitochondrial function/metabolism, their function as suppressors of inflammation is reduced.

A small fraction of Tregs in the tumor also upregulated IL-17, indicating plastic functions of these cells ( Figures 4A, D ), although the total number of FOXP3⁺ Tregs did not significantly change upon ablation of IL-17RA ( Figure 4B ). Consistent with our observation that IL-17 promotes Treg suppressive function, ablation of IL-17RA in Tregs leads to an increased proportion of CD4⁺ T cells that express IL-17 ( Figure 4C ). These IL-17 producing cells are either Foxp3⁺ or Foxp3^-, suggesting both enhancement in Th17 population and increased Treg plasticity ( Figures 4D, E ). We did not observe a noticeable change in the proportions of CD4⁺ T cells that produce IFN-γ ( Figures 4F, G ), or changes in the proportion of CD4 or CD8 cells among all CD45⁺ cells in the tumor ( Figures 4H, I ). Importantly, ablation of IL-17 signaling in Tregs resulted in increased tumor number and tumor load in mice, demonstrating a previously unknown tumor-inhibitory role of IL-17R through its function in Tregs ( Figures 4J–L ).

To further investigate how IL-17 signaling impacts Treg function, we analyzed human cancer patient scRNAseq datasets containing FOXP3⁺ Treg populations (32–37) ( Figure 5A ). Analysis of Treg populations within human tumors, normal adjacent tissues, and peripheral blood samples revealed genes associated with activation, differentiation, and suppressive function in the tumor. From these genes we created a tumor Treg signature which identified Tregs from the blood, adjacent tissue, and tumor that can be distinguished based on our gene signature ( Figure 5B ). The gene signature was comprised of genes associated with T cell activation, including RGS2 (83), Gadd45b (84), TNFRSF9 (4-1BB) (70, 71), TNFRSF4 (OX40) (49, 72, 73), and TNFRSF18 (GITR) (74). Also included were the Treg differentiation transcription factor NR4A1 (85), tumor Treg associated gene CCL5 (86), as well as immunosuppression-related genes CREM (87) and CXCR6 (88). A heatmap of the individual genes that comprised the Treg signature, along with unbiased hierarchical clustering of the tissue samples, grouped all tumor, adjacent normal tissue, and peripheral blood samples, irrespective of the cancer/tissue type of origin ( Figure 5C ). An examination of the mouse homolog genes in the scRNAseq datasets of MLN and tumor Tregs from control and Il17ra ^ΔTreg CRC mice revealed a similar pattern, with low expression in both the control and Il17ra ^ΔTreg CRC MLN Tregs, and high expression in the control tumor Tregs. Importantly, the Il17ra KO tumor Tregs had a reduced tumor Treg signature and were shifted towards the lower levels found in the MLN conditions ( Figure 5D ). A heatmap of the individual genes comprising the tumor Treg signature shows each gene is downregulated by Il17ra KO in the tumor Tregs (log2 fold-change >= 0.25, adjusted p-value < 0.05), but were not significantly altered by KO in Tregs in the MLN ( Figure 5E ). These data suggest a Treg gene signature which can unbiasedly distinguish blood, normal tissue, and tumor Tregs, regardless of cancer/tissue type. This gene signature associated with tumor Tregs is conserved between mouse and human and is significantly reduced by ablation of IL-17RA in these cells.

Analysis of RNA velocity in mouse Treg scRNAseq data revealed splicing changes resulting from Treg activation and tumor infiltration ( Figure 1E ). However, conventional techniques employed for scRNAseq limits the capacity for full or deep analysis of splicing due to the high rate of dropout, low sequencing depth and low coverage due to 3’ end sequencing of RNA (89, 90). Therefore, to investigate if Il17ra KO influences alternative splicing in Tregs, we isolated CD4⁺ FOXP3⁺ cells from the MLN and colon tumors and performed bulk RNA sequencing. An analysis of gene expression levels (FPKM) between control and Il17ra KO Tregs in the MLN showed a strong correlation between conditions (R² = 0.976) ( Figure 6A ). However, in the tumor, the correlation between control and Il17ra KO was reduced (R² = 0.809), with visible populations of differentially expressed genes between the two conditions (circled in red) ( Figure 6B ). These findings agree with our observation, through single-cell sequencing, that response to IL-17 signaling is limited to Tregs infiltrating the tumor environment, and that changes in gene expression profiles and Treg functions are tumor-specific. An analysis of differential isoform usage demonstrated that while few significant changes occurred in the MLN (-log10 adjusted p-value) there were robust isoform switches occurring between control and Il17ra KO Tregs in the tumor ( Figures 6C, D ). Consistent with our previous findings from scRNAseq, these data demonstrate that compared to Tregs in the MLN, where Tregs lack the necessary co-receptors of IL-17RA, tumor-infiltrating Tregs exhibit significant changes in isoform usage when Treg-specific Il17ra is knocked out. We next analyzed enrichment of splicing events between Il17ra WT and KO Tregs in the MLN and tumor. We found that while no significant changes were observed between Il17ra WT and KO Tregs in the MLN ( Figure 6E ), in the tumor, Il17ra KO Tregs exhibited significant reductions in alternative 3’ and 5’ donor sites, alternative transcription termination sites, exon/multiple-exon skipping, and an increase in intron retention, relative to control Tregs in the tumor ( Figure 6F ). As RNA binding proteins play an essential role in mRNA processing and generation of alternative isoforms (91), we analyzed the gene expression of known RNA binding proteins (RBPs) which were expressed in the Treg bulk RNAseq data. While in the MLN, RBP expression appears similar, in the tumor, Il17ra KO Tregs showed a substantial number of RBPs which were downregulated, and a smaller proportion which were upregulated, relative to control tumor Tregs ( Figure 6G ). Recently, several studies have identified RBPs which are key regulators of splicing in Tregs and essential for their function and immunosuppressive effects. In particular, the RBPs Srsf1 (92), Wtap (93), and Usp39 (94) were all significantly downregulated in Il17ra KO tumor Tregs ( Figure 6H ). Together, these data suggest that the downregulation of RBPs, in the absence of IL-17 signaling in tumor Tregs, leads to a reduction in alternative splicing that is important for Treg activation and function in the tumor.