Expression of RORγt Marks a Pathogenic Regulatory T Cell Subset in Human Colon Cancer

Abstract

The role of regulatory T cells (T_regs) in human colon cancer (CC) remains controversial: high densities of tumor-infiltrating T_regs can correlate with better or worse clinical outcomes depending on the study. In mouse models of cancer, T_regs have been reported to suppress inflammation and protect the host, suppress T cells and protect the tumor, or even have direct cancer-promoting attributes. These different effects may result from the presence of different T_reg subsets. We report the preferential expansion of a T_reg subset in human CC with potent T cell–suppressive, but compromised anti-inflammatory, properties; these cells are distinguished from T_regs present in healthy donors by their coexpression of Foxp3 and RORγt. T_regs with similar attributes were found to be expanded in mouse models of hereditary polyposis. Indeed, ablation of the RORγt gene in Foxp3⁺ cells in polyp-prone mice stabilized T_reg anti-inflammatory functions, suppressed inflammation, improved polyp-specific immune surveillance, and severely attenuated polyposis. Ablation of interleukin-6 (IL-6), IL-23, IL-17, or tumor necrosis factor–α in polyp-prone mice reduced polyp number but not to the same extent as loss of RORγt. Surprisingly, loss of IL-17A had a dual effect: IL-17A–deficient mice had fewer polyps but continued to have RORγt⁺ T_regs and developed invasive cancer. Thus, we conclude that RORγt has a central role in determining the balance between protective and pathogenic T_regs in CC and that T_reg subtype regulates inflammation, potency of immune surveillance, and severity of disease outcome.

INTRODUCTION

Regulatory T cells (T_regs), which suppress specific cytotoxic T cells (1, 2), have been shown to expand in human colon cancer (CC) and have preferential access to tumors (3). However, the contribution of T_regs to cancer is uncertain. T_reg accumulation in CC tumors has been linked with both poor (4) and favorable clinical outcomes (5-7). T_regs have been reported to be a good prognostic factor in gastric cancer (8), head and neck cancer (9), and breast cancer (10). Protection is speculated to be linked to the ability of T_regs to suppress inflammation (11-14). In addition, suppression of T helper 17 (T_H17) inflammation requires secretion of interleukin-10 (IL-10) by T_regs (15) and IL-10 signaling in T_regs and T cells (16, 17). Secretion of IL-10 by T_regs is also essential for suppression of Helicobacter-induced colitis (18) and gastritis (19), as well as genetically induced polyposis in mice (12).

We recently reported that T_regs in human CC and in mice with polyposis fail to produce IL-10 and instead produce IL-17 (14, 20). In cancer (14, 20) and in inflammatory bowel disease (21), there is detectable expression of IL-17 by T_regs. IL-17 and other T_H17 cytokines are elevated in the polyps and serum of mice with polyposis (14) and in human CC tumors (20), reflecting dysregulation of inflammation. Indeed, expression of IL-17 by tumor-infiltrating lymphocytes has been negatively correlated with patient survival (5). These observations suggest that CD4⁺Foxp3⁺IL-17⁺ cells are a distinct T_reg subset whose anti-inflammatory properties are compromised and expand in both preneoplasia and cancer.

Here, we show that IL-17–expressing T_regs exist within a subset of T_regs that express RORγt. RORγt-expressing T_regs expand in human CC in a cancer stage–dependentmanner and are also abundantly present in mice with polyposis. Our observations establish that expression of RORγt is linked with the inability of T_regs to suppress inflammation and is directly associated with the amount of inflammation and disease progression. Thus, cancer inflammation is controlled by the balance between protective T_regs with anti-inflammatory properties and pathogenic T_regs with proinflammatory properties: These T_reg populations are distinguished by expression of RORγt. We conclude that eliminating pathogenic T_reg subsets is feasible by targeting RORγt; this is a promising strategy for controlling inflammation and tumor growth in CC.

RESULTS

A distinct RORγt⁺Foxp3⁺ T_reg subset expands in human CC patients

We previously reported that CD4⁺Foxp3⁺ T cells from a small cohort of CC patients were defective in IL-10 production but produced IL-17 (20). To confirm these results, we analyzed an independent and larger population of 94 CC patients and 23 healthy donors (HDs) (table S1). To distinguish T_regs from Foxp3⁺ non-T_reg CD4⁺ T cells, we further sub-divided the Foxp3⁺ lymphocytes into three different fractions as defined recently by Sakaguchi (22, 23): Fr.I CD4⁺CD45RA⁺Foxp3^int, Fr.II CD4⁺CD45RA⁻Foxp3^high, and Fr.III CD4⁺CD45RA⁻Foxp3^int (Fig. 1A and fig. S1). Fr.I has naïve characteristics (CD45RA-high, CD45RO-low, CD25-int, HLADR-int), whereas Fr.II has activated characteristics (CD45RA-low, CD45RO-high, CD25-high, HLADR-high) (fig. S2). Fr.III lacks T cell–suppressive properties despite expressing Foxp3 and are, therefore, activated effector or helper T cells (22, 23). In CC, Fr.II preferentially expanded (2.9-fold) in tumors compared with healthy marginal tissue (Fig. 1B) and among peripheral blood mononuclear cells (PBMCs) of CC patients (2.5-fold) compared to PBMCs of HDs (Fig. 1, A and C). In PBMCs, there was a significant and specific increase in Fr.II with cancer stage, which was not evident in the other fractions (Fig. 1D).

Fig. 1.

In human CC, Fr.II T_regs expand in a cancer stage–dependent manner. (A) Representative dot plots depicting flow cytometry gating of five T cell fractions: three of which are Foxp3⁺ and two are Foxp3⁻. The Foxp3⁺ cells are labeled as Fr.I (black, CD45RA⁺Foxp3^int), Fr.II (red, CD45RA⁻Foxp3^high), and Fr.III (blue, CD45RA⁻Foxp3^int). Non-Foxp3 T cells can be divided into naïve CD45RA⁺Foxp3⁻ (green) and memory CD45RA⁻Foxp3⁻ (purple). Cells were pregated on live CD4⁺CD8⁻ cells. (B) Compiled frequencies of T_reg subtypes among CD4 T cells frommarginal or tumor tissue as determined by FACS analysis (n = 13). (C) Compiled frequencies of T_reg subtypes among CD4 T cells from PBMCs of HD or CC patients. HD, n = 23; CC, n = 94. (D) CC PBMC T_reg subtype frequencies according to the cancer stage of each CC patient as defined by the International Union for Cancer Control. Fr.I: white circles, HD; white squares, CC; Fr.II: red circles, HD; red squares, CC; Fr.III: blue circles, HD; blue squares, CC. Statistics are found in table S2. Bar represents median. Each dot in (B) to (D) represents one patient; value was determined from an average of at least two separate FACS stainings.

Because we have previously detected elevated T_H17 cytokines in human CC tumors (20), and Fr.II and Fr.III in healthy individuals were reported to express RORC transcripts (22), we analyzed expression of the T_H17 lineage-specific transcription factor RORγt in these fractions. In CC patients, we observed enrichment of RORγt-expressing cells among Fr.II T_regs not only in tumor compared to marginal tissue but also among PBMCs of CC patients compared to HDs (Fig. 2A and fig. S3). A similar trend was seen for IL-17–expressing T_regs in tumor versus marginal tissue and in PBMCs of CC patients compared to HDs (Fig. 2B and fig. S4). However, T_regs expressing RORγt significantly outnumbered IL-17–expressing T_regs (Fig. 2C). Conversely, we observed significant decreases in expression of IL-10 by T_regs (Fig. 2D). Expression of IL-17 and IL-10 was mutually exclusive, except in the tumor margin where a significant population of T_regs expressed both cytokines (fig. S4). We previously reported that dual expression of these cytokines can indicate progression from IL-10–expressing anti-inflammatory T_regs to IL-17–expressing proinflammatory T_regs (20). To further support this notion, we FACS (fluorescence-activated cell sorting)–sorted (fig. S5) the three Foxp3-expressing fractions and tested the ability of the T_regs to suppressmast cell (MC) degranulation. Fr.I and Fr.II T_regs from CC patients were unable to suppress MC degranulation, whereas HD Fr.I and Fr.II did suppress degranulation (Fig. 3A). These observations are consistent with local as well as systemic expansion of a population of RORγt-expressing T_regs in human CC.

Fig. 2.

CC Fr.II T_regs express a high frequency of RORγt and a minimal amount of anti-inflammatory IL-10. (A) Compiled frequencies of RORγt expression among each Foxp3⁺ fraction isolated from CC tissue (left panel, n = 7) or PBMCs of HD or CC PB (right panel, n = 21 and 74, respectively). (B) Compiled frequencies of IL-17 expression among each Foxp3⁺ fraction isolated from CC tissue (left panel, n = 6) or PBMCs of HD or CC PB (right panel, n = 21 and 71, respectively). (C) Side-by-side comparison of the frequency of IL-17– or RORγt-expressing Foxp3⁺ within each fraction from CC PBMCs. Black squares, Fr.I; red squares, Fr.II; blue squares, Fr.III. (D) Compiled frequencies of IL-10 expression among each Foxp3⁺ fraction isolated from CC tissue (left panel, n = 6) or PBMCs of HD or CC PB (right panel, n = 19 and 69, respectively). For (A), (B), and (D): Fr.I: white circles, HD; white squares, CC; Fr.II: red circles, HD; red squares, CC; Fr.III: blue circles, HD; blue squares, CC. Statistics are found in table S3. Bar represents median.

Fig. 3.

CC Fr.II T_regs are potently T cell–suppressive but lack the ability to suppress MC degranulation. (A) Compiled values of percent LAD2-MC degranulation sensitized with immunoglobulin E (IgE), cross-linked with anti-IgE + antigen (Ag), or incubated with FACS-sorted T_reg population isolated from HD (n = 3) or CC (n = 14) PBMCs. Bar represents median. Statistics are found in table S4. (B) Both Fr.I (black) and Fr.II (red) T_regs from PBMCs of HD (left panel, n = 2) or CC (right panel, n = 3) suppress the proliferation of naïve CD45RA⁺CD4⁺CD25⁻Foxp3⁻ T cells (green); Fr.III T_regs (blue) do not suppress T cell proliferation. T_eff, effector T cells. Each tested in duplicate. Mean and SEM are depicted.

The frequency of RORγt-expressing T_regs in PBMCs of CC patients increased significantly from stage II to stage III (fig. S6), encouraging speculation of its value as a marker for CC progression. In contrast to RORγt-expressing T_regs (fig. S6A, Fr.II), we could not detect consistent cancer stage–dependent expansion of IL-17–expressing T_regs in CC patients (fig. S6B, Fr.II). These observations suggest that expression of RORγt in T_regs does not always correlate with expression of IL-17. We conclude that cancer stage–dependent T_reg expansion can be largely attributed to an increase in the RORγt⁺Foxp3⁺ subset within Fr.II.

Expression of RORγt by T_regs has been associated with T_reg plasticity, eventual loss of T_reg-suppressive properties, and conversion to T_H17 (24, 25). We therefore tested the ability of FACS-sorted T_regs from CC patients’ PBMCs to suppress the proliferation of CD4 T cells. Fr.II was the most potent in suppressing the proliferation of naïve CD4 T cells in vitro, whereas Fr.III was the least effective (Fig. 3B). A similar trend was seen with T cells derived from HD PBMCs. These findings establish that the RORγt-enriched Fr.II T_regs of CC patients were potently T cell–suppressive and, therefore, unlikely to be in transition to T_H17 lineage commitment. To further probe this notion, we compared the cancer stage–dependent expansion of RORγt-expressing T_regs with expansion of T_H17-like CD4⁺ T cells defined by their expression of RORγt or IL-17 but not Foxp3. T_H17-like T cells did not show a consistent cancer stage–dependent increase in numbers [fig. S6, compare T_H17 in (A) with T_H17 in (B)]. Therefore, expansion of the RORγt⁺Foxp3⁺ subset does not coincide with an increase in T_H17-like cells. Hence, our observations thus far do not support T_reg plasticity but indicate that the RORγt⁺Foxp3⁺ cells may be a stable T_reg subset that expands in a cancer stage–dependent manner.

Deficiency in RORγt protects against polyposis, and RORγt⁺ T_regs are a major contributing factor

To evaluate the significance of expression of RORγt by T_regs, we resorted to the previously reported and well-characterized APC^Δ468 mouse model of polyposis (14, 20, 26). This mouse model contains a truncated adenomatous polyposis coli (APC) gene that leads to loss of APC function and development of benign adenomatous polyps, which are initiating events in human CC. First, we established that a significant fraction of T_regs in the intestine of polyp-ridden APC^Δ468 mice were activated and coexpressed Foxp3 and RORγt (fig. S7). As in human CC, CD4⁺RORγt⁺Foxp3⁺ cells disproportionately outnumbered CD4⁺IL-17⁺Foxp3⁺ cells (Fig. 4A and fig. S7A). Additionally, gene expression analysis of FACS-sorted T_regs revealed a T_H17 signature profile in APC^Δ468-derived T_regs compared to those from B6 mice, highlighted in T_regs isolated from the polyp or surrounding tissue of APC^Δ468 mice (Fig. 4B). Because RORγt has been suggested to be vital for T_H17 differentiation, we crossed APC^Δ468 mice to RORγt-deficient mice (27) to generate mice that are genetically prone to polyposis but cannot make RORγt. Surprisingly, serum levels of IL-17 in the RORγt-deficient mice were not significantly altered, although tissue IL-17 had returned to levels comparable to that found in B6 mice (Fig. 4C). Similar trends were observed for IL-1β, IL-6, and IL-23, as well as tumor necrosis factor–α (TNFα) (Fig. 4D), which have been intimately linked with a T_H17 response (14, 28-30). These findings suggest that expression of IL-17 and other T_H17 cytokines does not rely on expression of RORγt. Correlating with decreased levels of cytokines, RORγt deficiency drastically hindered polyposis in a progeny of crosses of RORγt^−/− mice with APC^Δ468 mice (Fig. 5A). In contrast to unmanipulated APC^Δ468 mice that tend to die by 4 months of age, the RORγt-deficient counterparts survived beyond 6 months of age and still had very few polyps (Fig. 5A). Histologic analysis confirmed the benign nature of the polyps (Fig. 5, D to G).

Fig. 4.

Loss of RORγt prevents excessive T_H17 response but does not abolish it. (A) Frequency of IL-17⁺ or RORγt⁺ cells among Foxp3⁺ cells. Comparing APC^Δ468 IL-17⁺Foxp3⁺ to APC^Δ468 RORγt⁺Foxp3⁺, P = 0.0111. (B) Heat map of ExpressCluster v1.3 showing the expression of components of the T_H17 molecular signature in T_regs isolated from the indicated tissues and mice (n = 3). (C) Expression of IL-17 protein in the serum and small bowel (S.B.) lysate as determined by multiplex ELISA. (D) Additional T_H17-associated cytokines were also detected with multiplex ELISA from small bowel lysate. Mean and SEM are depicted. Sera from five mice were each tested in duplicate. Lysates from polyps of threemice were each tested in duplicate. Statistics are found in table S5.

Fig. 5.

Loss of T_H17 cytokines reduces polyp number, but loss of IL-17 promotes invasion. (A) APC^Δ468 mice were crossed to RORγt^−/−, and polyp load was determined at 4 or 6 months of age: APC^Δ468 = 109 ± 19, n = 25; RORγt^−/−xAPC^Δ468 (4 m) = 9 ± 9, P < 0.0001, n = 22; RORγt^−/−xAPC^Δ468 (6 m) = 18 ± 7, P < 0.001, n = 7; comparing RORγt^−/−xAPC^Δ468 4 m to 6 m, P = 0.0152; mean ± SD, P value determined by Wilcoxon rank sum test. (B) APC^Δ468 mice were lethally irradiated at 6 weeks of age, reconstituted with BM progenitor cells from control B6 BM or BM deficient in IL-6, IL-23, IL-17A, TNFα, or RORγt, and allowed to age to 4 months of age. The number of polyps was then counted: B6 = 76 ± 3, n = 10; IL-6^−/− = 67 ± 2, n = 10; IL-23^−/− = 45 ± 6, n = 15; IL-17A^−/− = 27 ± 4, n = 9; RORγt^−/− = 8 ± 2, n = 5; P < 0.0001 using Kruskal-Wallis test comparing all. (C) APC^Δ468 mice were reconstituted at 6 weeks of age with control B6 BM or BM derived from RORγt^floxxFoxp3^Cre mice and allowed to age to 4 months of age. The number of polyps was then counted: B6 BM = 80 ± 24, n = 4; RORγt^floxxFoxp3^Cre BM = 32 ± 20, n = 5; P = 0.032 determined by Wilcoxon rank sum test. (D to L) Representative hematoxylin and eosin of “jelly-roll” paraffin-embedded sections from 4-month-old mice: (D and E) APC^Δ468, (F and G) RORγt^−/−xAPC^Δ468, and (H to L) IL-17^−/− BM in APC^Δ468. Arrows indicate polyps. (E, G, and I) ×200 Magnification of the boxed area in the panel above. (J to L) ×400 Magnification of invasive polyps depicted in (I).

To investigate the etiological link between RORγt, T_H17 response, and polyposis, we genetically ablated RORγt, IL-6, IL-23, IL-17, or TNFα in the bone marrow (BM) of APC^Δ468 mice. Polyposis was differentially affected by cytokine deficiency. Loss of IL-6 or IL-23 caused less than a 50% reduction in polyp load, whereas loss of TNFα or IL-17 significantly attenuated polyposis (Fig. 5B). Surprisingly, and in contrast to an earlier reported protection by IL-17 deficiency (31), all APC^Δ468 mice with IL-17–defective BM had invasive lesions (Fig. 5, J to L, and fig. S8). Loss of RORγt in the BM of APC^Δ468 mice (Fig. 5B) resulted in attenuation of polyposis comparable to that found in the RORγt^−/−xAPC^Δ468 mice (Fig. 5A). Finally, to confirm that this effect was mediated by T_regs, we crossed Foxp3-Cre mice (32) to RORγt^flox/flox mice, ablating the RORγt gene specifically in T_regs. These mice were then used as BM donors to generate BM chimeric APC^Δ468 mice. The resulting mice had significant attenuation of polyposis (Fig. 5C), indicating that expression of RORγt by T_regs is an etiological component of polyposis contributing to polyp growth.

Expression of RORγt, but not IL-17, coincides with the loss of T_reg anti-inflammatory function

Next, we examined the impact of RORγt deficiency versus IL-17 deficiency on expansion of proinflammatory cells during polyposis. Polyp-ridden APC^Δ468 mice have systemic inflammation marked by elevated numbers of proinflammatory F4/80⁺CD11b⁺ macrophages and CD11b⁺Gr1⁺ cells that include neutrophils and myeloid-derived suppressor cells (33). Loss of RORγt prevented an increase in the frequency of these cells in APC^Δ468 mice (Fig. 6, A and B). BM deficient for IL-17 also attenuated expansion of proinflammatory cells but less than in mice with RORγt deficiency (Fig.6, A and B). We have reported an etiological link between focal mastocytosis and polyposis in APC^Δ468 mice (14, 34). Furthermore, we showed earlier that T_regs from healthy mice potently suppress MC differentiation and function, whereas T_regs from polyp-ridden mice stimulate MCs (14, 20). There was a significant decrease in MC density within the polyps of RORγt-deficient mice (Fig. 6, C to F). In contrast, IL-17–deficient mice had a net increase in the density of polyp-infiltrating MCs, which was revealed by staining for MC-specific proteases (Fig. 6, C to F). There was also a qualitative difference; polyp-infiltrating MCs in APC^Δ468 and RORγt^−/−xAPC^Δ468 were intraepithelial and expressed mMCP2 (an MC-specific chymase) but rarely mMCP6 (an MC-specific tryptase) (Fig. 6, G and H). IL-17–deficient mice had elevated levels of MCs that expressed mMCP6, and these were located primarily in the stroma and invasive borders of the lesions (Fig. 6, G and H). These observations suggested that T_regs in RORγt^−/−xAPC^Δ468 mice but not in IL-17–deficient mice had recovered their anti-inflammatory properties.

Fig. 6.

Loss of RORgγt or IL-17 in polyp-pronemice differentially affects systemic and local inflammation. (A) Representative dot plots of Gr1⁺CD11b⁺ (top panel) and F4/80⁺CD11b⁺ (bottom panel) cells in the spleens of 4-month-oldmice. Cells are pregated for live cells. (B) Compiled frequencies of Gr1⁺CD11b⁺ (top) and F4/80⁺CD11b⁺ (bottom) cells. (C to H) Immunohistochemistry of paraffin sections from APC^Δ468 (left columns), IL-17^−/− BM in APC^Δ468 (middle columns), and RORγt^−/−xAPC^Δ468 (right columns) mice stained with chloroacetate esterase (CAE) (C), mMCP2 (E), or mMCP6 (G). Arrows point to stained cells. Frequency of polyp-infiltrating (D) CAE-positive or (F) mMCP2-positive cells among total cells within a 200× field of view. (H) Total number of mMCP6-positive cells found within the stroma of a polyp within a 200× field of view. For all graphs, bar represents median. Statistics are found in table S6.

To test for the anti-inflammatory properties of T_regs, we sorted T_regs from the spleen of wild-type or mutant mice at 4 months of age, confirmed their functional identity by demonstrating that they suppressed T cells in vitro and in vivo (fig. S9), and then measured their ability to suppress MC differentiation and degranulation in ex vivo assays developed by us earlier (14, 20). T_regs from RORγt^−/−xAPC^Δ468 mice were as suppressive as those from wild-type B6 mice, whereas T_regs derived from APC^Δ468 mice and IL-17–deficient APC^Δ468 mice failed to suppress MCs (Fig. 7, A and B). To validate these findings and extend them to suppression of polyposis in vivo, we adoptively transferred T_regs into 2.5-month-old APC^Δ468 mice and measured polyposis 3 weeks later, as previously described (13, 14). T_regs from APC^Δ468 mice failed to suppress polyposis, whereas T_regs from RORγt^−/−xAPC^Δ468 mice or from wildtype B6 mice hindered polyposis (Fig. 7C). Next, we tested the possibility that loss of RORγt rendered T_regs resistant to becoming proinflammatory. Although T_regs are potent inhibitors of MCs (35), their chronic interaction with MCs reverses this property and renders them proinflammatory (20). Upon cocultured with MCs, T_regs derived from wild-type B6 mice readily lost expression of IL-10 and, instead, produced IL-17, whereas RORγt^−/− T_regs were completely stable, failed to produce IL-17, and maintained IL-10 expression (Fig. 7D). These observations indicate that expression of RORγt by T_regs is fundamental to their phenotypic changes and subversion of their anti-inflammatory functions in polyposis.

Fig. 7.

Loss of RORγt, but not IL-17, restores T_reg anti-inflammatory function. (A) Frequency of MC progenitors (MCp) among total mononuclear cells (MNCs) isolated from the intestine of Rag^−/− mice and cultured without (n = 5) or with T_regs at 1:1 ratio from wild-type B6 (n = 6), APC^Δ468 (n = 5), or IL-17^−/− BM in APC^Δ468 (n = 6), or with RORγt^−/−xAPC^Δ468 (n = 5) mice. (B) Compiled values of percent MC degranulation sensitized (IgE) and cross-linked with anti-IgE (IgE + Ag) and incubated without (n = 10) or with T_regs from wild-type B6 (n = 16), APC^Δ468 (n = 15), IL-17^−/− BM in APC^Δ468 (n = 8), or RORγt^−/−xAPC^Δ468 (n = 10) mice. (C) Frequency of polyps after retro-orbital adoptive transfer of 5 × 10⁵ T_regs into 2.5-month-old APC^Δ468 mice. Polyp number was determined 3 weeks later. B6 (n = 5), APC^Δ468 (n = 4), RORγt^−/−xAPC^Δ468 (n = 3). (D) Representative dot plot of intracellular cytokines in T_regs derived from the spleen of wild-type B6 or RORγt^−/− mice. Purified T_regs were cultured without or with MCs for 5 days in the presence of IL-2 and stem cell factor (SCF). Cells were gated on live CD4⁺CD25⁺Foxp3⁺ cells (n = 3). Statistics are found in table S7.

Cancer immune surveillance is improved in RORγt-deficient APC^Δ468 mice

Protective antitumor T cell reactivity has been associated with a T_H17 response, raising the possibility that IL-17 or RORγt deficiencies could impair protective immunity (36, 37). Indeed, using multiplex enzyme-linked immunosorbent assay (ELISA), we found that ablation of IL-17 lowered serum levels of TH1 cytokines (Fig. 8A). However, RORγt^−/−xAPC^Δ468 mice had 34-fold higher levels of serum interferon-γ (IFN-γ) (Fig. 8A) and 2.5-fold higher levels in polyps (fig. S10). There were also increased serum IL-12p70 and IL-27 in the RORγt^−/−xAPC^Δ468 mice compared to APC^Δ468 mice (Fig. 8A). IFN-γ, IL-12, and IL-27 have been shown to enhance production of the chemokines IP10/CXCL10 and MIG/CXCL9, which in turn augment antitumor responses and inhibit angiogenesis (38, 39). IP10 and MIG were up-regulated in the serum and tissue lysates of RORγt^−/−xAPC^Δ468 mice, whereas the angiogenic factor vascular endothelial growth factor (VEGF) was diminished (Fig. 8A and fig. S10). By contrast, in the IL-17^−/− BM-reconstituted APC^Δ468 mice, there was no difference in MIG, but there was a notable drop in IP10 and a significant increase in VEGF (Fig. 8A). Thus, in polyp-prone APC^Δ468 mice, the absence of RORγt changed the quality of the inflammatory response rather than completely abolishing it, whereas loss of IL-17 failed to suppress or improve the quality of inflammation. These findings are consistent with ablation of RORγt, causing a shift to protective immunity in APC^Δ468 mice.

Fig. 8.

Protective inflammation and antitumor immunity is enhanced upon loss of RORγt in polyp-prone mice. (A) Amount of serum cytokines and chemokines from wild-type B6, APC^Δ468, IL-17^−/−BM in APC^Δ468, and RORγt^−/−xAPC^Δ468 mice as determined by multiplex ELISA. Sera from five mice were each tested in duplicate. (B) Improved polyp-specific T cell response assayed by IFN-γ ELISPOT. B6 dendritic cells (DCs) were pulsed with polyp lysate from APC^Δ468 mice as a source of tumor-associated antigens. Loaded DCs were then used to stimulate spleen-derived CD3⁺ T cells of either APC^Δ468, IL-17^−/− BM in APC^Δ468, or RORγt^−/−xAPC^Δ468 mice. T cells from wild-type B6 spleen were used as a control. T cells were also tested on unpulsed DCs to verify that the response was tumor antigen–specific. Black bars, unpulsed DCs; white bars, pulsed DCs. (C) Representative immunofluorescence images of paraffin-embedded sections from APC^Δ468, APC^Δ468 reconstituted with IL-17^−/− BM, or RORγt^−/−xAPC^Δ468 mice stained for perforin (red), granzyme B (green), and 4’,6-diamidino-2-phenylindole (DAPI) (blue); yellow arrow depicts granzyme B⁺perforin⁺ cells. Double-positive cells in healthy tissue of IL-17^−/− BM APC^Δ468 are rare, but inset depicts that they are present. (D) Compiled frequencies of granzyme B⁺perforin⁺ cells (yellow) among total cells within a 200× field of view; n = 3 mice. Mean and SEM are depicted. Statistics are found in table S8.

To test for protective antitumor immunity, we measured the reactivity of T cells to polyp antigens using IFN-γ enzyme-linked immunospot (ELISPOT). Specificity of response was demonstrated by the fact that T cells from APC^Δ468 mice produced IFN-γ when cocultured with polyp extract–pulsed DCs, whereas similarly treated T cells from control B6 mice did not (Fig. 8B and fig. S11). Tumor-specific reactivity of T cells from RORγt^−/−xAPC^Δ468 mice was enhanced 1.7-fold compared to APC^Δ468 (Fig. 8B). By contrast, there was no significant improvement in polyp reactivity by T cells from IL-17–deficient APC^Δ468 mice. To further validate this finding, we performed dual immunofluorescence on paraffin-embedded sections. Coexpression of both granzyme B and perforin is necessary for efficient cell killing and tissue rejection by cytotoxic lymphocytes (40-42). Notably, the frequency of granzyme B⁺perforin⁺ cells infiltrating the tissue adjacent to polyps of RORγt^−/−xAPC^Δ468 mice was increased 6.2-fold compared to that in APC^Δ468 mice and 11.3-fold compared to that in IL-17–deficient APC^Δ468 mice (Fig. 8, C and D). There was a significant decrease in the frequency of polyp-infiltrating granzyme B⁺perforin⁺ cells in IL-17–deficient APC^Δ468 mice compared to APC^Δ468 mice (Fig. 8, C and D). These results demonstrate that inhibition of RORγt in polyposis primarily affects the quality of T_regs and does not impair, and possibly even improves, antitumor immunity.

DISCUSSION

Here, we described different T_reg subsets in human CC with proinflammatory and antiinflammatory phenotypes and functions. Our observations suggested that expression of RORγt marks a distinct subset of activated T_regs that have the potential to express IL-17. These cells expanded preferentially in CC, increased with cancer stage, and were potently T cell–suppressive. Loss of expression of IL-10 and the inability to suppress MC degranulation demonstrated that RORγt⁺ Fr.II T_regs from CC patients had compromised anti-inflammatory properties. Despite the dominant T_H17 characteristic of inflammation in CC tumors (20), we did not observe a parallel increase in RORγt/IL-17–expressing CD4⁺ T cells lacking Foxp3 expression.

Human CD4 T cells that coexpress Foxp3 and IL-17 have been observed by others (22, 43, 44). IL-17–expressing human T_regs expand in Crohn’s disease (21) and in CC (20, 44, 45) patients. Expression of IL-17 is contingent upon activity of the transcriptional factor RORγt (46), RORα (47), or KLF4 (48). Expression of RORγt (49) or RORα (50) was considered to be incompatible with that of the key T_reg transcriptional factor Foxp3. Thus, T_regs that express RORγt together with Foxp3 were suggested to be in the transitional process of down-regulating Foxp3 and differentiating to T_H17 cells (21, 22). Our findings showed a discordance in CC between expansion of RORγt⁺ T_reg and IL-17–expressing CD4 cells, suggesting that the T_reg subset was not differentiating to T_H17 cell.

Our observations in human CC patients were made significant by the findings that T_regs with similar phenotypic and functional characteristics were also present in mice that develop polyposis. Furthermore, ablating RORγt specifically in T_regs severely hindered polyposis. Reduction in polyp load was due to reduced inflammation and enhanced polyp-specific immunity. Loss of RORγt stabilized the anti-inflammatory properties of T_regs, rendering them unable to divert to a proinflammatory phenotype when cultured with MCs. On the basis of these observations, we suggest that inhibitors of RORγt may have cancer-preventive and likely therapeutic action in CC, with much of the benefits of RORγt blockade being attributed to improved T_reg functions and attenuation of inflammation.

It remains to be seen whether recovery of anti-inflammatory properties of T_regs merely reduces the level of inflammation or also improves the quality of inflammation. Importance of the quality of inflammation becomes apparent with IL-17–deficient mice. In contrast to RORγt-deficient mice, IL-17–deficient mice developed invasive cancer. T_regs from IL-17–deficient APC^Δ468 mice not only failed to suppress MC degranulation but also induced changes in the localization and protease expression of MCs that were consistent with the switch from benign to invasive cancer (51). Hence, the primary benefit of ablation of RORγt but not IL-17 is in stabilizing anti-inflammatory properties of T_regs.

Unexpectedly, our observations suggest that loss of IL-17 promotes cancer progression. Ablation of IL-17 in polyposisprone mice decreased the number of benign polyps but drove progression of the remaining benign intestinal lesions to become invasive carcinomas. These observations are in contrast to earlier claims that ablation of IL-17 protects against polyposis (31). Because T_H17 cells are important in host control of gut microbiota (52, 53), this discrepancy may be related to differences in the composition of gut microflora in different mouse colonies. Nevertheless, the point is made that ablation of IL-17 can produce pathogenic outcomes. In addition, IL-17 may be important for eliciting effective tumor rejection, including antitumor T cell responses (36). The potential risks of inhibiting IL-17 have not been fully investigated, and it is questionable whether this can be an effective strategy for cancer therapy.

Our findings have serious implications for immune therapy of CC. Rebalancing T_reg subsets rather than indiscriminate elimination of T_regs is likely to produce the best protection. Special consideration is needed when directly interfering with inflammation, because certain inflammatory responses such as up-regulation of IL-17 may be protective at least for progression from benign to invasive cancer. Targeting RORγt in CC may be a promising strategy for controlling expansion of proinflammatory T_regs, improving the quality of inflammation, and shifting from pathogenic to protective immunity.

MATERIALS AND METHODS

Patient samples

Informed consent was obtained from all participants. Institutional Review Board protocols were approved by the Scientific Review Committee of Northwestern University. All patients had histologic diagnosis of CC and had not received any previous therapy. All tumors were tested for microsatellite instability (MSI) resulting from abnormalities in DNA mismatch repair genes (MLHa, MutL, and MutS).Any MSI⁺ tumor was excluded from this study. PB samples were drawn before surgical removal of the tumor.

Animals

All animal work was approved and conducted under the guidelines of Northwestern University’s Animal Care and Use Committee. C57BL/6J, IL-6^−/−(B6.129S2-Il6^tm1Kopf/J), TNFα^−/−(B6.129S-TNF^tm1Gml/J), Rag^−/−, RORγt^flox(B6-Cg-Rorc^tm3Litt/J), and RORγt^−/−(B6.129P2(Cg)-Rorc^tm2Litt/J) mice were purchased from Jackson Laboratories. IL-23(p19)^−/−, CD4Cre, and IL-17A^−/− mice were previously generated by M.Oft (54), A. Rudensky (32), and Y. Iwakura (55), respectively. APC^Δ468 mice were previously generated by our laboratory (26).

Antibody staining and FACS

Cells were preincubated for 10 min with Fc block (BD Biosciences). For intracellular cytokine stainings, GolgiPlug (BD Biosciences), phorbol 12-myristate 13-acetate (Sigma), and ionomycin (Sigma) were added 4 hours before Fc block. Dead cells were excluded with LIVE/DEAD Violet Dead Cell Stain kit (Invitrogen). FACS acquisition was performed with a BD FACSCanto II or BD Fortessa instrument. Data analysis was conducted with FlowJo software (TreeStar). For cell sorting, DAPI was used to exclude dead cells and sorting was performed on a Dako MoFlow.

Enzyme-linked immunospot

CD11c⁺ DCs and CD3⁺ T cells were tested for polyp-specific reactivity on an IFN-γ ELISPOT plate (Mabtech). Data were acquired on a CTL-ImmunoSpot analyzer and analyzed with CTL ImmunoSpot software. Details are in the Supplementary Materials.

Gene expression analysis

Cells were purified from the respective organs of male Foxp3-GFP (green fluorescent protein) or Foxp3-GFPxAPC^Δ468 reporter mice and subjected to microarray analysis according to ImmGen standard operating procedures (http://immgen.org). Array data were normalized for background. A GenePattern Expression Dataset file was generated from genes in the T_H17 molecular signature (BioCarta). These were further analyzed with the ExpressCluster v1.3 module of the GenePattern package (Broad Institute). Replicates of each sample were grouped to calculate and cluster on the class means.

Histology

Paraffin sections (4 μm) were used throughout. Details are found in the Supplementary Materials.

Inhibition of MC progenitor differentiation

MNCs from Rag^−/− intestine were isolated as described previously (14, 20). MNCs (10,000/100 μl) were serially diluted (1:2) in a 96-well plate containing complete medium, SCF (10 ng/ml), IL-3 (20 ng/ml), IL-2 (50 U/ml), and irradiated Rag^−/− mouse spleen cells (10⁶/ml). T_regs were added 1:1. Cells were incubated for 10 to 15 days at 37°C and 5% CO₂, after which colonies of MCs were scored.

MC degranulation and inhibition by T_regs

MCs were presensitized with dinitrophenyl (DNP)–specific IgE (1 μg/ml), challenged in Tyrode’s buffer with DNP (100 ng/ml, 30 min), and then centrifuged. Supernatants and solubilized cell pellets (0.5% Triton X-100) were incubated with p-nitrophenyl N-acetyl-β-d-glucosaminide in 0.1 M sodium citrate (pH 4.5) for 40 min at 37°C and stopped with 0.2 M glycine (pH 10.7). 4-p-Nitrophenol was detected by absorbance at 405 nm. Degranulation was calculated as [absorbance of the supernatant/(absorbance of supernatant + absorbance of pellet)] × 100. For T_reg inhibition, MCs were preincubated at 1:1 with T_regs for 10 min before the addition of DNP.

Serum, tissue lysates, and multiplex ELISA

Blood was collected into microcentrifuge tubes, allowed to clot, and centrifuged, and the supernatant was collected. Polyps were microdissected from the small and large bowel, minced, homogenized in 1 ml of phosphate-buffered saline, and centrifuged for 20 min at 4°C. Supernatant was filtered (0.22 μm), and protein was determined by the Bradford assay. Multiplex ELISA was conducted according to the manufacturer’s instructions (Millipore). Results were acquired with a Luminex 100 instrument and analyzed with xPONENT software (Luminex Corporation).

T cell proliferation/inhibition

Sorted T cells were cultured in precoated anti-CD3 (1 μg/ml) 96-well plates for 72 hours (37°C, 5% CO₂). [³H]Thymidine (1 μCi) was added, cells were incubated for 18 hours, and uptake was measured with a Perkin-Elmer MicroBeta plate counter. T_reg numbers were kept constant at 40,000. Non-T_reg CD4 cells were added at equal or increasing numbers.

Statistical analysis

Total Foxp3⁺ and individual Fr.I, Fr.II, and Fr.III percentages were compared across HD and cancer groups with Wilcoxon rank sum tests, across tumor stage with Kruskal-Wallis tests, and between margin and tumor samples with Wilcoxon signed-rank tests for paired samples. Wilcoxon rank sum and signed-rank tests were used for all additional nonpaired and paired comparisons, respectively.