Ripk3-induced inflammation by I-MDSCs promotes intestinal tumors

Abstract:
Myeloid derived suppressor cells (MDSC) promote colorectal cancer (CRC) by several mechanisms including suppression of antitumor T cells and production of tumorigenic factors. We previously showed that an intermediate-MDSC subset (I-MDSC) is expanded in an intestinal tumor model (ApcMin/+ mice), but the importance of this subset in promoting tumors is unclear. Here we show that I-MDSCs are a distinct heterogeneous subset due to differential and reduced expression of monocytic marker Ly6C and granulocytic marker Ly6G. Besides causing necroptotic cell death, receptor interacting protein kinase 3 (Ripk3) has an alternate function as a signaling component inducing cytokine synthesis. We evaluated if Ripk3 regulates inflammatory cytokines in I-MDSCs to assess the non-immunosuppressive function of I-MDSCs in promoting tumors. Inhibition of Ripk3 with the commercially available small molecule inhibitor GSK 872 showed that Ripk3-mediated inflammation promoted intestinal tumors in two intestinal tumor models, ApcMin/+ mice and an MC38 transplantable tumor model. Mechanistically, Ripk3 signaling in I-MDSC increased tumor size by expanding IL-17 producing T cells in MC38 tumors. Collectively, these data suggest Ripk3 signaling as a potential therapeutic target in CRC. The specific role of Ripk3 in intestinal tumors and MDSC function sheds light on a key inflammatory mechanism driving tumorigenesis and allows for possible therapeutic intervention.

Introduction
MDSCs in CRC patients correlate with reduced survival and have emerged as a resistance mechanism to existing therapies (1). Targeting MDSC function is beneficial for improving treatment efficacy; however, MDSCs are functionally complex. Success relies on a better understanding of functional differences among MDSC subsets because each subset appears to have a different role in tumorigenesis. In intestinal cancer models, much of the focus has been on how monocytic MDSCs (M-MDSCs, CD11b+Ly6c+Ly6G-) and polymorphonuclear MDSCs (PMN-MDSCs, CD11b+Ly6G+Ly6Clow) promote tumor growth (2,3). We observed a new MDSC subset termed I-MDSCs expressing reduced Ly6C and Ly6G in tumor-bearing ApcMin/+ mice, that has the APC mutation, commonly found in CRC patients (4). Phenotypic heterogeneity is a characteristic feature of MDSCs but less is known about this variation, particularly in I-MDSCs (5). Earlier, we showed that tumor reduction in ApcMin/+ mice deficient in Stat6 resulted in reduction of all MDSC subsets (4). From previous studies, the suppressive function of M-MDSCs and PMN-MDSCs are known to promote tumors; however, the mechanism by which I-MDSCs function in ApcMin/+ mice is not known. Delineating this process would shed light on a targetable tumor promoting mechanism as MDSC depletion or inhibition of MDSC function are attractive therapeutic strategies for CRC patients.Ripk1 and Ripk3 induce necroptotic cell death but also signal through ERK pathway to induce cytokines (6). Upon TNFα activation, Ripk1 in complex 1 activates apoptosis via caspase 8. When caspases are inhibited, the RIP homotypic interacting motif (RHIM) of Ripk1 interacts with Ripk3 RHIM in complex 2 to activate kinase activity and phosphorylation of mixed lineage kinase domain-like (MLKL) protein resulting in necroptosis (7). Hence, Ripk1 activation can lead to either apoptosis or necroptosis. However, deleting Ripk1 in keratinocytes resulted in Ripk3 mediated necroptosis causing excessive inflammation and Ripk1 deficient hematopoietic cells underwent increased apoptosis and necroptosis. (8,9). This was found to be due to the interaction of ZBP-1 RHIM with Ripk3 RHIM, which was prevented by Ripk1 RHIM (10). Other RHIM containing proteins such as Toll/IL-1 receptor domain-containing adaptor protein inducing interferon β (TRIF) or DNA-dependent activator of interferon regulatory factor (DAI) can also activate Ripk3 RHIM to promote its kinase activity (11).

In addition to causing cell death, Ripk3 induced inflammatory cytokines promoted disease severity in inflammatory bowel disease (IBD) and TNFα induced systemic inflammatory response syndrome (SIRS)(12,13). Therefore, Ripk3 is an ideal target for inhibiting inflammation. However, Ripk3 defect produced contradictory responses in murine colitis and intestinal cancer. One study showed that Ripk3 deletion did not affect colitis (12), disease was exacerbated in another study (14), and disease was ameliorated in a different study (15). Deletion of Ripk3 RHIM in CD11c+ cells reduced IL-22 required for tissue repair, exacerbating colitis (16). These Ripk3 RHIM deficient DCs produced less IL-23 and IL-1β due to reduced activation of NFκBp50, suggesting that Ripk3 RHIM is required for its signaling function. Although Ripk3 RHIM has a different function from its kinase domain, interaction of RHIM domains is essential for Ripk3 kinase activity (17). Exacerbation of colitis after deletion of Ripk3 RHIM (16) maynot result in intestinal tumor development. This is possible as shown by increased colitis but reduced AOM-DSS induced tumors after NFκBp50 deletion (18). However, total deletion of Ripk3 increased AOM-DSS induced colonic tumors by promoting tumor cell proliferation and upregulating inflammatory cytokines (19). Conversely, deletion of Ripk3 reduced AOM-DSS induced colon tumors promoted by Hace1 deficiency (15).Moreover, Ripk3 inhibition in bone marrow derived macrophages (BMDMs) reduced inflammatory cytokines and chemokines (20). These reports show that additional evidence is required to define how Ripk3 function impacts intestinal tumors. Considering the importance of inflammatory cytokines in promoting intestinal tumors and the demonstrated role of Ripk3 in inducing these cytokines in myeloid cells, we reasoned that Ripk3 induced cytokines in MDSCs could promote intestinal tumors.

We targeted Ripk3 activity using a commercially available inhibitor GSK 872 that was found to be a highly specific inhibitor of Ripk3 after testing against 300 other protein tyrosine kinases (21). Given its overall tolerability among several cell types at low concentrations, we used it as a tool to elucidate the function of Ripk3 in intestinal cancer and in MDSCs. The correlation between MDSC accumulation and Th17 cells with reduced survival of CRC patients (1) and the importance of MDSC induced IL-17 in inflammation driven tumorigenesis (22,23) suggest that analysis of the signaling mechanism in MDSCs driving IL-17 induction would reveal therapeutic candidates. In this report, we demonstrate that Ripk3 promotes intestinal tumors by upregulating cytokines required for expanding IL-17 producing T cells through I-MDSCs, a distinct MDSC subset.C57BL/6J and ApcMin/+ mice were purchased from Charles River Laboratories and Jackson Laboratory, respectively. Both male and female ApcMin/+ mice at 13 to 24 weeks were used in all experiments. WT littermates were used as controls. ApcMin/+ mice were genotyped to identify ApcMin mutation using previously published primers (24). All mouse protocols were approved by the Yale University Institutional Animal Care and Use Committee.RNA was extracted with Trizol (Invitrogen) from size-matched ApcMin/+ polyps, WT small intestine, ApcMin/+ MDSCs subsets (gated as shown in Supplementary Fig. 1A), or in vitro treated MDSC subsets. Suppression assays for MDSCs are described in supplementary materials and methods. Polyps and small intestine tissue were homogenized in disposable Lysing matrix tubes (MP Biomedicals) before RNA extraction. cDNA was reverse-transcribed from DNase treated RNA using Superscript III reverse transcriptase (Invitrogen). Target gene was amplified by SYBR green PCR master mix (Biorad). Expression was normalized to housekeeping genes RPL32 or GAPDH. The standard curve method was used to quantify relative gene expression.

Primer sequences for all target genes were obtained from Harvard Medical School PrimerBank (25) and checked for specificity by Primer-BLAST on NCBI (26).ApcMin/+ mice were injected every 3 days with GSK 872((N-(6- (Isopropylsulfonyl)quinolin-4-yl)benzo[d]thiazol-5-amine), Millipore) at 0.25 mg/Kg body weight when mice were 13 weeks old. Mice in control group received vehicle (DMSO) dissolved in PBS. GSK 872 was well-tolerated as treated mice showed no signs of lethargy or allergic reaction at 24 hours after treatment with this inhibitor and every 3 days thereafter. At 20 weeks, tumors in the small intestine and colon of ApcMin/+ mice were counted and harvested for qRT-PCR. Splenomegaly was assessed by measuring spleen length. Splenocytes were mechanically dissociated to prepare single cell suspensions for flow cytometry. For MC38 murine colon tumor model, 2 x 104 MC38 cells were implanted subcutaneously in WT female C57BL/6J mice. 0.8 x 106 or 106 splenic MDSCs of each subset from tumor bearing ApcMin/+ mice or BM-MDSCs (purity of sorted MDSCs is shown in Supplementary Fig.1B) were injected directly into similar sized MC38 tumors. After I-MDSC injection, mice were treated with GSK 872 at 0.5 mg/Kg every other day or once with anti-IL-17A or isotype control antibody (75 μg/mouse). Tumor volume was monitored every other day, and calculated by the following formula: short diameter2 x (long diameter)/2. MC38 cells were kindly provided by Dr. Alan Sartorelli at Yale University. MC38 cells were regularly tested for Mycoplasma and found to be contamination-free.Tumor infiltrating lymphocytes (TILs) from MC38 tumors were prepared by digestion of tumor fragments with collagenase (Sigma) and DNase (Roche) at 37οC for 30 minutes;live lymphocytes were separated on a percoll density gradient and stained with described antibodies.

For intracellular cytokine staining, TILs were activated with 50 ng/ml of PMA (Sigma) and 1μg/ml of ionomycin (Sigma) for 5 hours; brefeldin A (BD Biosciences) was added during the last 4 hours of culture before fixation. Cells were surface stained with T cell markers (CD45, TCRβ, CD4, CD8), permeabilized with permeabilization buffer (eBioscience), and stained with anti-IL-17 or isotype control antibody (eBioscience). For sorting MDSC subsets, splenic lymphocytes were depleted of T cells, B cells, and NK cells using biotin labeled CD3, CD19, and CD49d antibodies (Tonbo and eBioscience) and Streptavidin coated magnetic beads (Biolegend) to enrich CD11b+ cells. Enriched CD11b+ cells were stained for MDSC markers including CD11b, Ly6C, Ly6G, and CD11c to sort M-MDSCs, I-MDSCs, and PMN-MDSCs after excluding DCs on BD FACSAria sorter (BD Biosciences). WT monocytes and neutrophils were sorted similarly. Sorting purity for these cells is shown in Supplementary Fig. 1B. Viability was maintained by storing cells in RPMI containing 20% FBS until sorting was completed and assessed by trypan blue exclusion assay.Lamina propria (LP) lymphocytes were prepared from small intestines of ApcMin/+ mice by percoll density gradient and CD11b+ cells were isolated by positive selection with biotin labeled CD11b antibody (BD Biosciences) and streptavidin coated magnetic beads (Biolegend). Prior to LP isolation, intestinal epithelial cells (IECs) were collected from the EDTA-PBS wash in which cleaned small intestine fragments were incubated at 37οC for 30 minutes. Purity of isolated LP CD11b+ cells and IECs are shown in Supplementary Fig. 1C. Refer to supplementary materials and methods for details.FlowJo was used to generate flow cytometry plots and mean fluorescence intensity (MFI) values from FCS files obtained by analyzing cells on the Stratedigm instrument.

Isotype control MFI was subtracted from target MFI to generate the heatmap. ApcMin/+ MDSCs were analyzed by FlowSOM tool in FlowJo to identify subpopulations within MDSC subsets (27). FlowSOM generated a minimal spanning tree (MST) consisting of 4 main clusters based on Ly6C and Ly6G expression. Each sub-population of a MDSC subset included closely arranged nodes, where each node represents a group of similar cells and the size of the node represents the number of cells. Nodes arranged next to each other on the MST are more closely linked than nodes that are further away.Frequencies and MFI values for sub-populations of MDSC subsets were also generated by the FlowSOM tool.Bone marrow cells from WT mice were cultured at 106 per ml in MDSC medium containing RPMI, 10% FBS, 1% Penicillin/Streptomycin, MC38 tumor conditioned medium (TCM), and 40 ng/ml of recombinant murine GM-CSF (Peprotech) for 3 days to generate BM-MDSCs. Single cell suspension of MC38 tumors were cultured in RPMI with 10% FBS at 5 x 105 cells per ml for 24 hours to produce TCM. This protocol was modified from published methods (28,29). Loosely adherent cells were separated into MDSC subsets based on Ly6C and Ly6G expression on CD11b positive cells. 106 M- MDSCs, I-MDSCs, or PMN-MDSCs sorted from in vitro generated BM-MDSCs were stimulated with LPS (Invivogen) at 10 ng/ml with or without zVAD (carbobenzoxy-valyl-alanyl-aspartyl-[O-methyl]- fluoromethylketone) at 50μM; GSK 872 was added at 5μM or 20μM. RNA was extracted after 2 hours and analyzed for cytokine expression by qRT- PCR. Viability of treated cells was assessed with CellTiter-Glo assay (Promega) using a luminometer and ATP units were calculated as percent of untreated cells.Students t-test was used to determine statistical significance. Data are presented as means ± SEM. p value less than 0.05 was considered significant.

Results
The outcome of Ripk3 function in intestinal cancer is unclear due to different results about its role in tumor formation (12,15,16,19,30). We used the ApcMin/+ model that is genetically modified to form intestinal tumors, to assess how Ripk3 affects tumor development. Ripk3 and MLKL were elevated in ApcMin/+ intestinal polyps compared to WT intestinal tissue but Ripk1 expression was unchanged (Fig.1A-C) suggesting that Ripk3 mediated tumor growth. To inhibit Ripk3 induced inflammatory cytokines, we used GSK 872, a Ripk3 inhibitor with an IC50 of 1.3nM (21). 10μM of GSK 872 induced apoptosis in tumor cells but not in BMDMs that were relatively viable. This suggested that GSK 872 could inhibit Ripk3 induced cytokines without inducing excessive apoptosis if used at a low and effective concentration, which is beneficial for controlling inflammation driven CRC. There were significantly fewer tumors in the small intestine of ApcMin/+ mice treated with GSK 872 compared to mice treated with vehicle (Fig. 1D).Tumors in the proximal, middle, and distal segments of the small intestine (SI) of GSK 872 treated ApcMin/+ mice were significantly reduced compared to control (Fig. 1E-G). Colon tumors and splenomegaly were also reduced but did not reach statistical significance (Fig. 1H and I). IL-6, IL-1β, and IL-23 were significantly reduced in polyps from GSK 872 treated ApcMin/+ mice (Fig. 1J-L); IL-17 and COX-2 were also suppressed but there was no change in TNFα (Fig. 1M-O). This shows that Ripk3 inhibition reduced tumor-promoting cytokines. Surprisingly, epithelial stem cell markers including Sox-9 were increased in polyps of GSK 872 treated ApcMin/+ mice (Supplementary Fig. 2A).This could be due to the prosurvival function of Ripk1 when Ripk3 is inhibited or due to other immunoregulatory mechanisms such as PD-1 expression (4). Together these results show that intestinal tumors in ApcMin/+ mice is dependent on Ripk3 enzymatic activity.M-MDSCs and PMN-MDSCs correlate with tumor growth in a transplanted CRC model and in ApcMin/+ mice (2,3). Tumor reduction in ApcMin/+ mice deficient in Stat6 reduced the frequency of splenic M-MDSCs and PMN-MDSCs and strikingly depleted I-MDSCs (4).

In GSK 872 treated ApcMin/+ mice, frequency of M-MDSCs and macrophages were moderately reduced, PMN-MDSCs were unchanged, and I-MDSCs were significantly reduced (Fig. 2A-F). Therefore, I-MDSC accumulation is tumor dependent due to its consistent correlation with intestinal tumor burden (Fig. 1D)(4). I-MDSCs appeared to be less suppressive than M-MDSCs in inhibiting proliferation of non-antigen specific CD4 and CD8 T cells (Fig. 2G and H). Moreover, GSK 872 treatment did not significantlyalter the frequency of T cells (Supplementary Fig. 2B-D). We did not assess the cytotoxic quality of CD8 T cells or the suppressive function of MDSCs in GSK 872 treated ApcMin/+ mice. However, tumor reduction in GSK 872 treated ApcMin/+ mice increased CD45+ cells indicating reversal of lymphodepletion that is typically observed in diseased ApcMin/+ mice (Fig. 2I) (31). Overall, reduction of I-MDSCs and tumors after Ripk3 inhibition suggests that I-MDSCs could be related to Ripk3 mediated tumor progression.We reasoned that I-MDSCs are phenotypically different from other MDSC subsets based on intermediate expression of Ly6C and Ly6G. I-MDSCs accumulated in ApcMin/+ mice but not in MC38 tumor bearing mice (Fig. 3A). Other ApcMin/+ MDSC subsets were similar to MC38 MDSCs due to comparable expression of antigen presenting and co- stimulatory molecules (Supplementary Fig. 3A and B). To better understand the heterogeneity of I-MDSCs, we analyzed each ApcMin/+ MDSC subset for differences in Ly6C and Ly6G expression with the FlowSOM tool. This revealed four sub populations within each MDSC subset (Fig. 3B). There are more P4 M-MDSCs than other subsets (Fig. 3C). All M-MDSC sub-populations (P1- P4) expressed Ly6C and Ly6G, which was unexpected because M-MDSCs are CD11b+Ly6C+Ly6G- (Fig. 3D); however, Ly6C was significantly higher than Ly6G. P1, P3, and P4 have similar Ly6C/Ly6G ratio and are located close to each other in the minimal spanning tree (MST) (Fig. 3B). P2 is phenotypically different from other sub-populations because its Ly6C/Ly6G ratio (2.5) is lower and it is located further from other sub-populations (Fig. 3B).

Frequencies of all I-MDSC sub-populations were similar (Fig. 3E). The difference between Ly6G and Ly6C was significant in P1 and P2 of I-MDSCs but not in P3 and P4 (Fig. 3F). A P1 subset and P2, and P3 and P4 are located close to each other on the MST showing that they are related (Fig. 3B). However, other subsets within P1 are separated on the MST suggesting that they could be different subsets. Therefore, I-MDSCs appear to be heterogeneous. Among PMN-MDSCs, P2 was most frequent with the highest Ly6G to Ly6C ratio (Fig. 3G-H). More than 10 fold increase in Ly6G in all PMN-MDSCs show that they are all granulocytic. MHC class l and ll is reduced in I-MDSCs and PMN- MDSCs compared to M-MDSCs indicating a shared feature between I-MDSCs and PMN-MDSCs (Fig. 3I-J). Notably, M-MDSCs and I-MDSCs express 29 and 15-fold more CSF-1R, and 3 and 2-fold more F4/80, compared to PMN-MDSCs. This suggests that I- MDSCs are similar to M-MDSCs but lack CD80 and IL-4Rα. Therefore, I-MDSCs are a distinct MDSC subset including a spectrum of cells with heterogeneous Ly6C and Ly6G expression and some similarity to other MDSC subsets.CT26 colon tumor cells overexpressing GM-CSF promoted accumulation of I-MDSC like cells with intermediate expression of Ly6C and Ly6G (32). 4T1 breast cancer cells producing high GM-CSF contributed to accumulation of I-MDSC like cells (33). Hence, GM-CSF could selectively favor I-MDSC increase in ApcMin/+ mice, which has high circulating levels of GM-CSF (34) and polyps producing GM-CSF (Fig. 4A). To evaluate if GM-CSF increased I-MDSC accumulation in ApcMin/+ mice, BM cells were treated with MC38 tumor conditioned medium (TCM) and recombinant murine (rm) GM-CSF or rm GM-CSF or TCM. Both I-MDSCs and M-MDSCs accumulated more in either conditionincluding GM-CSF (Fig. 4B-D). BM-MDSCs generated with TCM alone produced more PMN-MDSCs (Fig. 4E).

Furthermore, I-MDSCs accumulated even at low concentration of GM-CSF and M-MDSCs increased with more GM-CSF (Fig. 4F-H). Thus, GM-CSF promotes accumulation of M-MDSCs and I-MDSCs.BM-MDSCs suppress non-antigen specific T cells similar to ApcMin/+ MDSCs (Supplementary Fig. 4). Among BM-MDSCs generated in all conditions, M-MDSCs have the highest Ly6C expression followed by I-MDSCs, and PMN-MDSCs (Fig. 4I-L). TCM appears to promote MHC class ll, CD80, and CD86 expression on M-MDSCs and I- MDSCs (Fig. 4M and Supplementary table 1). MHC ll and CD80 expression on M- MDSCs and MHC l, CD80 and CD86 on I-MDSCs were dependent on GM-CSF. Both GM-CSF and TCM contributed to CSF-1R and F4/80 expression on I-MDSCs whereas only TCM promoted F4/80 expression on M-MDSCs. Therefore, GM-CSF is required for I-MDSC accumulation and both TCM and GM-CSF are essential for their distinct phenotype.I-MDSCs accelerate MC38 tumor growth by Ripk3 mediated IL-17 production Inhibition of Ripk3 activity reduced intestinal tumors, I-MDSCs, and IL-17 in ApcMin/+ mice. This suggests that splenic I-MDSCs could systemically promote Th17 cells by producing cytokines required for generating Th17 cells. ApcMin/+ I-MDSCs produced more IL-1β than M-MDSCs but less TGFβ than M-MDSCs and PMN-MDSCs (Fig. 5A, C). All MDSC subsets expressed IL-23 (Fig 5B). This is consistent with MDSCs from ovarian cancer, which produced IL-23, IL-6, or IL-1β and generated Th17 cells (35). Due to different levels of cytokines produced by MDSC subsets, we evaluated if they differ indriving tumor growth by injecting them into MC38 tumors. The volume of tumors injected with I-MDSCs increased significantly at day 16, 18, and 20 compared to control tumors (Fig. 5D). Notably, IL-17 producing CD4 T cells were significantly increased in MC38 tumors injected with I-MDSCs compared to tumors injected with M-MDSCs or PMN- MDSCs or control tumors (Fig. 5E and F). Tumors receiving M-MDSCs had more CD4+IL-17+ T cells but were not significantly larger than control tumors. Although PMN- MDSCs expressed IL-23, IL-1β, and TGFβ, they did not significantly increase tumor volume or IL-17 producing T cells (Fig. 5).

Fewer T cells in the tumors and spleen of MC38 tumor bearing mice receiving I-MDSCs suggest lymphodepletion in larger tumors compared to smaller tumors in control group (Supplementary Fig. 5A-D). However, there were no differences in CD44 or CD62L expression on CD4 or CD8 T cells, respectively (Supplementary Fig. 5E-L). To show that I-MDSCs promote tumor growth by Ripk3 signaling, MC38 tumors were injected with I-MDSCs and treated with GSK 872. I-MDSC dependent tumor growth and frequency of CD4+IL-17+ cells were significantly reduced after Ripk3 inhibition (Fig. 6A-C). Depleting IL-17 after injecting MC38 tumors with I-MDSCs resulted in smaller tumors showing that I-MDSCs promoted tumor growth by IL-17(Fig. 6D). Together these results show that Ripk3 signaling in I- MDSCs expand IL-17 producing CD4 T cells in tumors to accelerate tumor growth.Ripk3 activity is mediated by TLR4 activated TRIF RHIM independent of Ripk1 (7,36). This could sustain steady production of inflammatory cytokines in MDSCs. Ripk1 and Ripk3 was upregulated in myeloid cells (CD11b+) cells from ApcMin/+ lamina propria (LP)compared to WT cells but MLKL was unchanged (Fig. 7A-C). There was no change in Ripk1, Ripk3 or MLKL between ApcMin/+ and WT IECs. In the spleen, M-MDSCs expressed more Ripk1, Ripk3, and MLKL than I-MDSCs or PMN-MDSCs (Fig. 7D-F) (37). Reduced Ripk3 in I-MDSCs and in intestinal CD11b+ cells from ApcMin/+ mice is more suitable for Ripk3 induced cytokine induction because higher Ripk3 and MLKL expression results in cell death (38). The bacterial ligand LPS promotes IL-23 dependent intestinal inflammation through TLR-Myd88 pathway (23). We used LPS to evaluate if TLR4 activated Ripk3 mediated cytokine induction in MDSCs. TLR4 was similarly expressed on all MDSC subsets but TLR2 and TLR3 were reduced on neutrophils and MDSCs (Supplementary Fig. 6A). To activate TLR4 induced Ripk3 signaling by inhibiting apoptosis, in vitro generated M-MDSCs and I-MDSCs were treated with LPS and zVAD. These cells were treated with GSK 872 to show that LPS induced Ripk3 is critical for cytokine induction. IL-23 and COX-2 were upregulated in I- MDSCs and GSK 872 inhibited IL-23 and IL-1β showing that induction of these cytokines is Ripk3 dependent (Fig. 7G-I). In M-MDSCs, COX-2 was upregulated and GSK 872 significantly inhibited both IL-1β and COX-2. All cells were viable at 2 hours after treatment indicating that Ripk3 signaling did not cause cell death (Supplementary Fig. 6B). As M-MDSCs underwent some cell death with LPS treatment alone and GSK872 did not reverse cell death, it is unlikely to be caused by necroptosis. The specific cause of cell death in this case was not analyzed. Together, these results show that Ripk3 signaling regulates Th17 promoting cytokines in I-MDSCs.

Discussion
Intestinal microbial load is critical for the progression of intestinal tumorigenesis (39). The TLR4 ligand, LPS is a bacterial component and activates Ripk1 and Ripk3 signaling pathway (7). Apart from regulating cell death, Ripk1 and Ripk3 mediate ERK1/2 signaling in myeloid cells indicating their importance in cytokine production (16,20). Demonstrating the pro-tumor role of Ripk3 in MDSCs will identify another strategy for targeting MDSCs to improve cancer treatment. Using a Ripk3 inhibitor GSK 872, we show that Ripk3 activity in I-MDSCs promotes tumors by expanding IL-17 producing CD4 T cells.A rare Ripk1 mutation increases Ripk3 induced inflammatory cytokines causing inflammatory bowel disease, which was reversed by hematopoietic stem cell transplantation (40). Moreover, Ripk3 deletion reduces tumor progression in mice with Hace1 deficiency (15). Therefore, there is a strong premise for Ripk3 induced inflammation in promoting intestinal tumorigenesis; however, there is no evidence supporting this mechanism. In our study, IECs including tumor cells in ApcMin/+ mice in our study expressed more Ripk3 than myeloid cells, suggesting more cell death. However, tumors continue to grow and increase in numbers. Ectopic Ripk3 expression in a lung cancer cell line induced cell death to reduce tumor growth in a xenograft model (41). In contrast, Ripk3 deletion in breast and colon cancer cell lines reduced tumor growth in xenograft models (30). These studies suggest that the role of Ripk3 in tumor cells is unclear. Ripk3 mediated cell death also activates innate immune response by releasing danger associated molecular patterns (DAMPs) (42). However, we assessed if sustained inflammation promoted by Ripk3 signaling in MDSCs propagates tumor promoting T cells. Low Ripk3 expression is more conducive for signaling because high Ripk3 expression generally results in necroptotic cell death (38). Intestinal myeloid cells and splenic I-MDSCs from ApcMin/+ mice expressed less Ripk3 suggesting that Ripk3 could induce cytokine induction in these cells. Although I-MDSCs are not well defined in the LP, they could be present at very low levels and could be included in total myeloid cells from the SI of ApcMin/+ mice (4). By studying how splenic I-MDSCs regulate Th17 cells in MC38 tumors, we show that they could function systemically and possibly migrate to tumors to function locally by this mechanism.

Homozygous Ripk3 kinase dead mutants (D161N) undergo embryonic death due to excessive apoptosis (43). However, heterozygotes are viable suggesting that inhibiting Ripk3 with a low concentration of GSK 872 is possible in vivo and highlights its therapeutic value. At high concentrations, GSK 872 induces apoptosis in cells but sensitivity to apoptosis differed among cell types (21). For example, BMDMs (macrophages) treated with 10μM of GSK 872 retained at least 75% viability up to 18 hours. As BM-MDSCs are different from macrophages, 20μM of GSK 872 was required to inhibit Ripk3 activity but viability was maintained at this concentration. In ApcMin/+ mice, 0.25 mg/Kg of GSK 872 every 3 days inhibited intestinal tumorigenesis. We did not use a higher concentration of GSK 872 to avoid inflammation caused by excessive apoptosis. As tumor cells undergo apoptosis at concentrations greater than 10μM, GSK 872 could promote some tumor cell apoptosis in our study. This is likely limited to tumor cells as frequency of lymphocytes and myeloid cells were unchanged other than reduction of I-MDSCs. Temporal inhibition of Ripk3 could be insufficient for a dramatic change in MDSC frequency as opposed to a genetic deletion of Ripk3. In addition, Ripk3 inhibition could also modify the immunosuppressive function of MDSCs because GSK 872 inhibited both IL-1β and COX-2 in ApcMin/+ polyps and M-MDSCs; these cytokines mediate suppressive function of MDSCs (44). We will evaluate this aspect of Ripk3 function to understand the total effect of inhibiting Ripk3 in MDSCs. This suggests that MDSC subsets could regulate more than one arm of pro-tumor T cell response by promoting pathogenic Th17 cells and suppressing antitumor T cells.

MDSCs from ovarian cancer promoted IL-17 producing T cells (35). Moreover, in a colon tumor model of CPC-APC mice, MDSCs (CD11b+Gr-1+) and not macrophages produced IL-23 to drive tumors by IL-17 (23). However, the role of Ripk3 in MDSC mediated Th17 generation is unknown. IL-23 and IL-1β have tumor-promoting functions in intestinal cancer (4,45). This is consistent with upregulation of IL-23 and IL-1β in ApcMin/+ polyps and MDSC subsets including I-MDSCs. GSK 872 reduced intestinal tumors in ApcMin/+ mice and Th17 differentiating cytokines including IL-6, IL-23, and IL- 1β, suggesting that Ripk3 could be regulating Th17 cells and intestinal tumor growth in ApcMin/+ mice to promote tumors. Using a transplantable CRC tumor model, we show that ApcMin/+ I-MDSCs were more efficient in promoting tumor growth by expanding IL-17 producing CD4 T cells by Ripk3 signaling.
IL-23, IL-1β, and IL-6 generates pathogenic Th17 cells causing EAE disease progression whereas TGFβ, IL-1β, and IL-6 generate conventional Th17 cells (46). T cells treated with TGFβ dramatically reduced IL-17 expression, which is consistent with the role of TGFβ in suppressing IL-17 production by reducing IL-23R expression (46). This would explain why ApcMin/+ M-MDSCs and PMN-MDSCs producing more TGFβ are less efficient in promoting Th17 cells. These Th17 cells could also be conventional Th17 cells that produce IL-10 to suppress inflammation (46) and is likely the reason for reduced tumor growth. In our study, I-MDSCs produced less TGFβ than other subsets and promoted more IL-17 production resulting in increased tumor growth. This supports the possibility that low TGFβ production by I-MDSCs likely promoted pathogenic Th17 cells. A similar mechanism could be occurring in human CRC, by generation of pathogenic Th17 cells by a MDSC subset. In addition to M-MDSCs and PMN-MDSCs, e-MDSCs lacking both CD14 and CD15 have been identified in humans (47); however, their role in Th17 induction is undefined. We will delineate the role of pathogenic versus conventional Th17 cells in intestinal tumorigenesis in future studies.

MDSCs are heterogeneous with relatively few phenotypic markers (monocytic marker Ly6C and granulocytic marker Ly6G) to adequately classify them into different subsets (5). We identified four subpopulations within each ApcMin/+ MDSC subset based on differential expression of Ly6C and Ly6G, which is consistent with the overall heterogeneity of MDSCs. Three M-MDSC sub-populations with high Ly6C expression indicate their monocytic origin. Notably, sub-population P2 within M-MDSC could be related to I-MDSCs due to its location on the MST and low Ly6C/Ly6G ratio. I-MDSCs with an overall reduction of Ly6C and Ly6G are distinct from M-MDSCs and PMN- MDSCs. However, they exhibit some similarity to M-MDSCs due to similar expression of macrophage markers (F4/80 and CSF-1R) and to PMN-MDSCs due to reduction of antigen presenting molecules. GM-CSF promotes I-MDSC accumulation but GM-CSF and TCM were essential for F4/80 and CSF-1R expression on I-MDSCs. Accumulation of I-MDSCs was dependent on GM-CSF, due to which ApcMin/+ mice with high circulating levels of GM-CSF (34) have more I-MDSCs. Similarly, other studies show expansion of I-MDSC like cells in mice with colon tumor cells expressing GM-CSF or breast cancer cells that produce high levels of GM-CSF (32,33). I-MDSCs could migrate to the intestine and eventually converting to tumor associated macrophages or DCs similar to MDSCs in different tumor models (44). However, more evidence is required to delineate this process.
Collectively, we have shown that Ripk3 regulates inflammatory signaling in I-MDSCs to promote IL-17 production in T cells and that it has a tumor promoting function in intestinal cancer. Overall, modulation GSK’872 of CRC promoting cytokines by GSK 872 demonstrates that Ripk3 is targetable in intestinal cancer.