Cross-Talk between Myeloid-Derived Suppressor Cells and Mast Cells Mediates Tumor-Specific Immunosuppression in Prostate Cancer

Prostate cancer is a leading cause of cancer death worldwide (1). Although surgery and radiotherapy are effective for localized disease, advanced or recurrent disease treated with androgen ablation often develops into castration-resistant and metastatic disease with fatal outcomes (2).

Immunotherapy and immune checkpoint inhibitors have been tested with success in several clinical trials treating melanoma, non–small cell lung cancer, and renal cell carcinoma (3), but with little success in treating prostate cancer (2). Improved understanding of the molecular mechanisms regulating immune suppression in prostate cancer should help in tailoring the immunotherapy for these patients (4).

Mast cells, innate immune cells known for their role in allergy and anaphylaxis, are also proangiogenic (5) and protumorigenic (6, 7). The amount of mast cell infiltration into human prostate cancer correlates with prognosis; mast cells may be a useful therapeutic target (8). We have shown that pharmacologic inhibition of mast cells degranulation in transgenic adenocarcinoma of the mouse prostate (TRAMP) mice reduces incidence and slows progression of prostate adenocarcinoma, but favors prostate tumors with neuroendocrine features (9, 10). Thus, therapeutic targeting of mast cells (i.e., with imatinib; ref. 9) could be effective against prostate adenocarcinoma if done in combination with approaches directed against neuroendocrine differentiation. The molecular mechanisms, however, are only partially explained by the fact that mast cells supply MMP-9 until the tumor becomes able at more advanced stages to produce this protease (10). Mast cells mediate immunologic tolerance in several disease models, including cancer (11), possibly through interaction with other immune suppressive cell populations such as Tregs (12) and myeloid-derived suppressor cells (MDSC; ref. 13). Both Tregs and MDSCs are enriched in prostate cancer patients (14–16). To study the contribution of mast cells to immunosuppression during prostate adenocarcinoma development, we used the TRAMP mouse model, in which prostate epithelium transforms because of the SV40 early genes [small and large T antigens (Tag)] driven by the androgen-responsive rat probasin regulatory element. Selective Tag expression in the prostate epithelium starts at puberty (17), driving progressive development of dysplasia, prostate intraepithelial neoplasia (PIN; weeks 6–16), adenocarcinoma (weeks 16–25), and lymph node and lung metastases (weeks 18–30), resembling the human pathology (18). A small percentage of TRAMP mice also develop neuroendocrine prostate cancer (19), an aggressive form of disease that in patients is associated with resistance to therapies and poor prognosis (20). In the TRAMP mouse model, Tag behaves as a nonmutated tumor-associated antigen: thymic central tolerance deletes high-affinity Tag-specific CD8⁺ T cells (21), whereas low-affinity Tag-specific CD8⁺ T cells leaving the thymus experience peripheral tolerance in correlation with tumor development and Tag expression in the prostate (22).

Here, we crossed TRAMP mice with mast cell–deficient Kit^Wsh mice (23), and tested whether the observed reduction of adenocarcinoma incidence and growth rate was related to altered immunosuppression as a consequence of mast cell deficiency. The reconstitution of Kit^Wsh-TRAMP mice with mast cells from CD40L⁺ or CD40L^– donors revealed CD40L-CD40–mediated cross-talk between mast cells and polymorphonuclear (PMN) MDSCs, which then suppress the tumor-specific T-cell response, thus favoring adenocarcinoma growth. We identified in human prostate cancer patients a signature of mast cell/PMN-MDSC–related genes that correlated with relapse and reduced survival.

TRAMP mice on C57BL6/J background [C57BL/6-tgN (TRAMP)8247Ng] were kindly provided by Dr. Vincenzo Bronte (Verona University Hospital, Italy), under agreement with Dr. Norman Michael Greenberg (formerly at Fred Hutchinson Cancer Research Center, Seattle, USA), and maintained and screened according to (17). Mice deficient in mast cells [C57BL/6-KitW-sh/W-sh (Kit^Wsh; ref. 23)] were purchased from The Jackson Laboratories and intercrossed over 12 generations with TRAMP mice to obtain mast cell–deficient Kit^Wsh-TRAMP mice. CD40L^−/− mice (B6.129S2 – Cd40lg^tm1lmx/J) were a kind gift of Dr. Matteo Iannacone (San Raffaele Scientific Institute, Milan, Italy). Experiments were performed in accordance with Italian law (D.lgs 26/2014). Cromolyn (10 mg/kg in saline; Sigma) was administered intraperitoneally (i.p.) 5 days per week, starting at 8 weeks of age. Mice were sacrificed at 16 or 25 weeks of age.

For T-cell depletion experiments, the thymus of 16-week-old TRAMP and Kit^Wsh-TRAMP mice was surgically removed. The following day, mice were treated i.p. with 300 μg of depleting antibodies to CD4 and CD8 and sacrificed at 25 weeks of age.

Bone marrow precursors from C57BL6/J or CD40L^−/− mice were cultured in vitro in RPMI with 20% FBS, and 20 ng/mL both SCF and IL3 (Peprotech; ref. 10). After 4 weeks, when purity was more than 90%, 5 × 10⁶ BMMCs were injected i.p. into 8-week-old mice. Mast cell degranulation was evaluated by CD107a staining as described in ref. 9.

Dendritic cells (DC) were prepared culturing bone marrow precursors for 7 days with 5 ng/mL of IL4 and 25 ng/mL of GM-CSF (Peprotech). On day 7, DCs were matured for 8 hours with 1 μg/mL LPS (Sigma), pulsed 1 hour with 2 μg/mL of TAG-IV_404-411 peptide and injected (5 × 10⁵ DC/mouse) i.d. into the right flank of mice.

Six days later, mice were injected i.p. with 10⁷ cells containing equal numbers of splenocytes labeled with 1.25 μg/mL (CFSE^hi) or 0.125 μg/mL of CFSE (CFSE^low). CFSE^hi cells were previously pulsed 1 hour with TagIV_404-411 peptide. Mice were sacrificed the following day, and their splenocytes, prostates, and prostate draining lymph nodes (PDLN) analyzed by flow cytometry for the evaluation of the presence of CFSE^hi and CFSE^low cells. Tag-specific cytolytic activity was calculated as: (percentage CFSE^high cells) × 100/(percentage CFSE^low cells) (22). Splenocytes of killed mice were also tested for Tag-specific IFNγ production, as described below.

Cell suspensions were obtained from mechanical disaggregation of spleen and PDLN or from digestion of prostates with collagenase IV (1,600 units/mL) for 1 hour at 37°C. Cells were incubated 10 minutes with FcR blocker and labeled for 15 minutes at 4°C with fluorochrome-conjugated monoclonal antibodies (mAb) or isotype controls. All antibodies used in our study are listed in Supplementary Table S1. For intracellular detection of IFNγ, splenocytes were stimulated 4 hours with TagIV_404-411 peptide (1 μg/mL), adding brefeldin A (10 μg/mL) in the last 3 hours (22). Cells were stained for surface markers, fixed with 2% PFA, permeabilized with saponin (0.5% in PBS) and incubated with anti-IFNγ. Samples were acquired with BD LSRII Fortessa and analyzed with the FlowJo software.

FACS-sorted PMN-MDSCs were seeded in 24-well plates (10⁵ cells/condition), alone or with wild-type or CD40L^–/– BMMCs (1:1 ratio), or with10 μg/mL of agonistic CD40 antibody or of antagonistic CD40L antibody. The following day, cells were stained with anti-CD11b and anti-cKit (Supplementary Table S1). Cells were then fixed, permeabilized, and incubated with specific fluorochrome-labeled nucleotidic probes against Arg1 (Alexa 647), Nos2 (Alexa 488), and Stat3 (Alexa 488), according to the PrimeFlow RNA Assay kit protocol (eBioscience). Positive control samples were stained with Alexa 488 or Alexa 677 probes against Actin. Fluorescence minus one (FMO) control were made by staining samples with all antibodies and RNA probes except one. Levels of RNA expression were measured as—(MFI of the stained samples/MFI of the FMO)—within CD11b⁺ cells.

Cell suspension obtained from murine splenocytes (pools of at least 4 mice per group) was enriched for myeloid cells with the use of CD11b-magnetic microbeads (Miltenyi Biotech; cat. no. 130-049-601), then labeled with anti-mouse CD11b, Ly6G, and Ly6C (Supplementary Table S1), and sorted with a BD FACSAria. For in vitro suppression assays, 10⁵ naïve C57BL6 splenocytes were labeled with CFSE and cultured alone or with MDSCs at different ratios, with anti-CD3 (2 mg/mL; eBioscience, cat. no 16-0031-85) and anti-CD28 (1 mg/mL; eBioscience, cat. no. 16-0281-85) to activate lymphocytes. Proliferation of CD4⁺ and CD8⁺ T cells was assessed 3 days later, evaluating CFSE dilution by flow cytometry. When indicated, BMMCs were added to the coculture in ratios equal to MDSC.

Total RNA was extracted using the Quick RNA micro prep kit (Zymo Research). cDNA was obtained using the MultiScribe- Reverse Transcriptase kit (Applied Biosysyems). Real-time PCR was performed in a total volume of 20 μL using the Taqman Universal PCR Master Mix (Applied Biosystems), 20 ng of cDNA and specific probes for Arg1, iNOS, Stat3, Tgfβ, or IL6 (Applied Biosystems). Values were normalized to internal control (Gapdh) and to MDSCs from control TRAMP mouse using the ΔΔCT method.

Murine urogenital apparata and spleens and human tumor samples were fixed in formalin and embedded in paraffin. Sections (5 μm) were stained with hematoxylin and eosin (H&E; BioOptica) and evaluated by a pathologist. Prostate lesions were scored as dysplasia, prostatic intraepithelial-neoplasia (PIN), adenocarcinoma (ADENO) and neuroendocrine tumor (NE). For evaluation of mast cell infiltration, sections were stained with toluidine blue (BioOptica). Immunohistochemistry was performed using the streptavidin–biotin–peroxidase complex method, and 3,3′-diaminobenzidine tetrahydrochloride as chromogenic substrate. Antibodies used are listed in Supplementary Table S1. Slides were analyzed under an Axioscope A1 microscope equipped with Axiocam 503 Color camera (Zeiss). For tumor burden calculation, three nonoverlapping panoramic microphotographs (at ×50 magnification) of H&E-stained slides were collected from each sample and analyzed using the Zen 2.0 software (Zeiss), by contouring the foci of high-grade dysplasia/PIN, in situ carcinoma, and invasive carcinoma, and quantifying the extension of the relative areas (μm²).

Antibodies used for immunofluorescence are listed in Supplementary Table S1. Imaging was performed using a confocal laser-scanning microscope Leica TCS SP8 X (Leica Microsystems), equipped with a pulsed super-continuum White Light Laser (470–670 nm; 1 nm tuning step size). Laser lines were 495 nm for FITC and Alexa 488, 556 nm for Alexa 546 and 633 nm for APC; detection ranges were 501 to 556 nm, 569 to 630 nm, and 638 to 744 nm, respectively. Images were acquired in the scan format 1024 × 1024 pixel using an HC PL APO 63X/1.40 CS2 oil immersion objective and a pinhole set to 1 Airy unit. Data were analyzed using the software Leica LASX rel.1.1 (Leica Microsystems).

The Chiorino dataset (GEO #GSE60329) is described in ref. (24). For the Taylor (25) and Setlur (26) datasets, data were downloaded from the GEO repository (GSE21032 and GSE8402). Samr was used to identify a gene signature associated with biochemical relapse in the Chiorino dataset (27). Starting from a list of 90 genes specifically expressed by mast cells and PMN-MDSCs retrieved from the literature (28, 29) and from the Cibersort tool (https://cibersort.stanford.edu), we obtained a list of 23 Agilent probes, corresponding to 20 unique genes, high expression of which was associated with biochemical recurrence. The ability of the signature to generate sample groups associated with survival was tested on Kaplan–Meier curves and log-rank test. Curves and P values were generated with the Bioconductor package survival. Samples were divided according to the median of the average expression of the signature genes. sigCheck was used to test random signatures of the same length for each dataset, using 1,000 iterations (30). An empirical P value was computed based on the percentile rank of the performance of the primary signature compared with a null distribution of the performance of the random signatures.

Statistical analyses were performed with the GraphPad Prism software (GraphPad Software); we used Fisher test for comparison of tumor frequencies and the Student t test, or one-way ANOVA followed by Tukey tests for other experiments.

To investigate whether mast cells support spontaneous prostate cancer occurrence, we crossed TRAMP mice with mast cell–deficient Kit^Wsh mice (23). We compared tumor growth and histotype between cohorts of 25-week-old mice. As expected (17, 19), the majority (80%) of observed TRAMP mice showed multifocal invasive prostate adenocarcinoma (Fig. 1A and B), whereas a small percentage developed prostate tumor with neuroendocrine features (13%) and only few had evidence of dysplasia or prostatic intraepithelial neoplasia (PIN; 7%; Fig. 1A). Conversely, Kit^Wsh-TRAMP mice showed impaired adenocarcinoma growth. Adenocarcinoma occurred in only 21.4% of mice; a larger portion of the mice were afflicted only with prostate lesions that were scored as dysplasia or PIN (42.9%; Fig. 1A and B). Thus, mast cells encouraged the growth of invasive prostate adenocarcinoma.

Nevertheless, as previously observed (9, 10), the incidence of prostate neuroendocrine tumors increased in mast cell–deficient mice (35.7% in Kit^Wsh-TRAMP mice versus 13% in TRAMP mice; Fig. 1A), suggesting that mast cells also prevent development of this aggressive variant. Neuroendocrine prostate cancer was identifiable at necropsy as a spherical mass overwhelming the prostate, and can also be detected by ultrasound imaging as early as 15 to 18 weeks of age (31). To retain our focus on the role of mast cells in immunosuppression and adenocarcinoma outgrowth, we excluded from further analysis TRAMP or Kit^Wsh-TRAMP mice bearing neuroendocrine prostate cancer.

We investigated whether mast cell–mediated immunosuppression occurs in prostate cancer and can contribute to adenocarcinoma onset and progression. We analyzed mice at 16 weeks of age, a time when TRAMP mice have already developed peripheral CD8⁺ T-cell tolerance against SV40 Large-T antigen (Tag; ref. 22), which behaves as a nonmutated self-tumor-associated antigen in this model (21). At this time point, both TRAMP and Kit^Wsh-TRAMP mice displayed similar prostatic PIN lesions (Fig. 1C) and Tag expression (Fig. 1D). Age-matched (16-week-old) TRAMP, Kit^Wsh-TRAMP and nononcogenic C57BL6 (B6) and Kit^Wsh littermates were immunized with dendritic cells (DC) pulsed with the CD8-immunodominant Tag epitope (TagIV; ref. 32) and killed 1 week later to test in vivo Tag-specific cytotoxic activity (Supplementary Fig. S1A) and ex vivo IFNγ production by CD8⁺ T cells (Supplementary Fig. S1B). Consistent with previously published data (22), we detected negligible in vivo Tag-specific cytotoxicity in both spleen (30.3% ± 11.70%; Fig. 1E) and prostate-draining lymph nodes (PDLN) of TRAMP mice (29.89% ± 7.81%; Supplementary Fig. S2A). TRAMP mice also showed barely detectable ex vivo Tag-specific IFNγ production by CD8⁺ T cells isolated from spleen (0.19% ± 0.18%; Fig. 1F), PDLN (17.8% ± 1.31%; Supplementary Fig. S2B) or prostates (20. 94% ± 0.18%; Supplementary Fig. S2C). On the other hand, Tag-specific cytotoxic T-cell response and IFNγ production by CD8⁺ T cells from Kit^Wsh-TRAMP mice were almost comparable to that of T cells of B6 mice used as positive control upon immunization with the same vaccine both in spleen (cytotoxicity 67.81% ± 15.17% vs. 86.25% ± 8.33%; Fig. 1E; and IFNγ production 1.06% ± 0.30 vs. 1.82% ± 0.24%; Fig. 1F) and PDLN (cytotoxicity 68.33% ± 14.53% vs. 81.25% ± 17.56%; Supplementary Fig. S2A; and IFNγ production 7.29% ± 0.66% vs. 8.43% ± 2.11%; Supplementary Fig. S2B). CD8⁺ T cells infiltrating the prostates of Kit^Wsh-TRAMP mice released significantly increased amounts of IFNγ compared with CD8⁺ T cells from prostates of TRAMP mice (30.39% ± 9.68% vs. 20.94% ± 3.83%; Supplementary Fig. S2C). These results indicate that in the absence of mast cells, tumor-specific CD8⁺ T cells became manifest, both systemically and within the tumor microenvironment.

To confirm the involvement of mast cells in inhibiting tumor-specific CD8⁺ T cells and in supporting prostate adenocarcinoma growth, we reconstituted 8-week-old Kit^Wsh-TRAMP mice with bone BMMCs. We analyzed the mice 8 weeks later for Tag- specific CD8⁺ T-cell activity. Histologic sections of spleens and prostates stained with toluidine blue showed correct repopulation and distribution of mast cells in reconstituted Kit^Wsh-TRAMP mice (Fig. 2A; Supplementary Fig. S3A). In reconstituted Kit^Wsh-TRAMP mice, both in vivo Tag-specific cytotoxicity and ex-vivo IFNγ production by CD8⁺ T cells was lowered to that of TRAMP mice (0.30% ± 0.32 vs. 0.40% ± 0.33, Fig. 2B, and 61.80% ± 19.40 vs. 62.91% ± 26.94, Fig. 2C, respectively). Consistent with the inverse correlation between effective CD8⁺ T-cell activity and frequency of adenocarcinoma, 60% of Kit^Wsh-TRAMP mice reconstituted with BMMCs developed adenocarcinomas, whereas only 21.4% of nonreconstituted mice developed adenocarcinomas (Fig. 2D). To further confirm the link between effective tumor specific T-cell function and reduced adenocarcinoma growth occurring in Kit^Wsh-TRAMP mice, we depleted in vivo both CD4⁺ and CD8⁺ T lymphocytes, by surgical removal of thymus followed by administration of specific depleting antibodies (Supplementary Fig. S3B). In T cell–depleted Kit^Wsh-TRAMP mice, the frequency of infiltrating adenocarcinoma raised up to levels comparable with TRAMP mice (Fig. 2E and F). On the contrary, adenocarcinoma growth was unaltered in TRAMP mice regardless of whether or not T cells were depleted (Fig. 2E and F), a result consistent with the unresponsiveness of tumor-specific T cells in these mice (Figs. 1E and F; 2B and C and ref. 22). Thus, mast cells impair tumor-specific CD8⁺ T-cell activity, allowing growth of adenocarcinoma in TRAMP mice.

To exclude the possibility that increased antitumor immune response in Kit^Wsh-TRAMP mice could have been due to altered T-cell frequency or distribution, we evaluated by flow cytometry the percentages of T lymphocytes in spleens and prostates of TRAMP, Kit^Wsh-TRAMP, Kit^Wsh-TRAMP mice reconstituted with BMMCs and their nononcogenic littermates (Supplementary Fig. S4). The frequency of CD8⁺ T cells in the spleen was lower in Kit^Wsh-TRAMP or in reconstituted Kit^Wsh−TRAMP mice than in TRAMP mice, excluding an expansion of CD8⁺ T cells in the absence of mast cells. No differences in CD4⁺ and CD8⁺ T cells infiltrating the prostates were noted among the different strains (Supplementary Fig. S4).

Data collected so far indicate that mast cells can hamper tumor-specific CD8⁺ T-cell responses both systemically and at the tumor site, through either direct or indirect interactions between mast cells and T cells. Immunofluorescence analysis excluded direct interaction between mast cells and T cells by revealing that in the spleen of TRAMP mice, mast cells are not present in the splenic white pulp and do not colocalize with CD3-expressing cells (Fig. 3A). On the contrary, we found mast cells within the splenic red pulp, in close contact with Gr1^hi cells (Fig. 3A). Gr1 antibodies (recognizing both Ly6C and Ly6G epitopes) could identify MDSCs, which accumulate in the spleen of tumor bearing mice (33). Mast cells can enhance MDSCs suppressive activity (13, 34). Therefore, we tested whether the increased tumor-specific CD8⁺ T-cell response in mast cell–deficient Kit^Wsh-TRAMP mice correlated with reduced MDSCs activity. The characteristic expansion of immature CD11b⁺Ly6G^intLy6C^int cells that occurs in Kit^Wsh mice (35, 36) was also evident in the spleen of Kit^Wsh-TRAMP mice (Supplementary Fig. S5A and S5B), which also showed more monocytic (M-) MDSC–like cells (CD11b⁺Ly6G^lowLy6C^hi) and polymorphonuclear (PNM-) MDSCs (CD11b⁺Ly6G^hiLy6C^low; Supplementary Fig. S5A and S5B). On the contrary, no differences in the frequencies of M-MDSCs and PMN-MDSCs were detected in the prostates of Kit^Wsh-TRAMP versus TRAMP mice (Supplementary Fig. S5A and S5B).

We sorted M-MDSC–like cells and PMN-MDSCs from the spleen of 16-week-old TRAMP and Kit^Wsh-TRAMP mice with a FACS, and tested them for the ability to suppress CD4⁺ and CD8⁺ T-cell proliferation. No M-MDSC–like cells from any of the strains had suppressive activity (Fig. 3B, top). On the contrary, PMN-MDSCs were suppressive when from TRAMP mice, but less suppressive when they were from Kit^Wsh-TRAMP mice (Fig. 3B, bottom). The reduced function of PMN-MDSCs from Kit^Wsh-TRAMP mice paralleled decreased transcript levels of Arg1, Nos2, and Stat3, which all promote MDSC suppressive activity (Fig. 3C). Conversely, little TGFβ and IL6 were expressed and their RNA level in PMN-MDSCs from both TRAMP and Kit^Wsh-TRAMP mice was similar (Fig. 3C). Thus, mast cells can interact with PMN-MDSCs to increase their suppressive ability.

To investigate the mechanism used by mast cells to enhance PMN-MDSC suppression, we considered their release of soluble mediators (37), and treated TRAMP mice with the mast cells stabilizer Cromolyn, which blocks mast cell degranulation. Treatment had no effect on Tag-specific CD8⁺ T-cell responses (Fig. 4A and B; refs. 9, 10), but partially inhibited adenocarcinoma growth and increased the frequency of neuroendocrine tumors (adenocarcinoma frequency was 50% in cromolyn-treated TRAMP mice, 80% in TRAMP mice, and 21.4% in Kit^Wsh-TRAMP mice; Fig. 4C). We concluded that soluble factors were not necessary for mast cell:PMN-MDSC cross-talk, and that immunosuppression may instead depend on receptor–ligand interaction.

CD40 has been already associated to MDSC suppression (38) and to their interaction with mast cells, at least in vitro (13). We confirmed CD40L expression on BM-derived mast cells cultured in vitro (Supplementary Fig. S6A), and on mast cells isolated from the spleen of TRAMP mice (Supplementary Fig. S6B). PMN-MDSCs isolated from the same spleens showed positivity for CD40 (Supplementary Fig. S6C). The colocalization between mast cells and CD40 in the spleen of TRAMP mice was visualized by triple immunofluorescence with Gr1, CD40, and the mast cell–marker Tryptase (Fig. 5A). No mast cells were found to colocalize with CD40^hi B cells in the splenic follicular zone (Supplementary Fig. S6D).

To further confirm the role of the CD40L–CD40 axis in the interaction between mast cells and PMN-MDSCs, we reconstituted Kit^Wsh-TRAMP mice with BMMCs obtained from either wild-type or CD40L^–/– mice and tested the suppressive activity of PMN-MDSCs (Fig. 5B). Wild-type and CD40L^−/− mast cells differentiated and matured in vitro similarly, were equally responsive to stimuli (Supplementary Figs. S7A and S7B), and were similarly recruited to spleen and prostates in Kit^Wsh-TRAMP mice after in vivo reconstitution (Supplementary Fig. S7C). Their distribution in proximity of Gr1⁺ cells in the spleen of reconstituted Kit^Wsh-TRAMP mice was also comparable (Supplementary Fig. S7D). Reconstitution with wild-type BMMCs restored PMN-MDSC suppressive activity, but reconstitution with CD40L^–/– mast cells did not (Fig. 5B). The expression of Arg1, Nos2, and Stat3 in PMN-MDSCs was recovered in PMN-MDSCs from Kit^Wsh-TRAMP mice that received wild-type BMMCs, but not in mice that received CD40L^–/– BMMCs (Fig. 5C). To confirm the capacity of mast cells to enhance PMN-MDSC function via CD40L–CD40 cross-talk, we performed suppressive assays, coculturing responder T cells, PMN-MDSC, and CD40L⁺ or CD40L^–mast cells, either with or without antagonistic CD40L antibody. T-cell inhibition by PMN-MDSCs isolated from TRAMP mice was not further increased by the presence of mast cells, likely because of in vivo triggering of their CD40 signaling by splenic mast cells. In contrast, the suppressive ability of PMN-MDSCs isolated from Kit^Wsh-TRAMP mice cocultured with wild-type, but not CD40L^−/−, mast cells was comparable to that of TRAMP-derived PMN-MDSCs. This increase in PMN-MDSC suppressive function was partially inhibited by adding a blocking antibody to CD40L (Supplementary Fig. S7E).

To prove that CD40 triggering on PMN-MDSCs modulates Arg1, iNOS, and Stat3 expression, FACS-sorted PMN-MDSCs from the spleen of Kit^Wsh-TRAMP mice were stimulated overnight with an agonistic anti-CD40 or cocultured with wild-type or CD40L^−/− BMMCs, either with or without antagonistic anti-CD40L. Amounts of mRNA encoding Arg1, iNOS, and Stat3 were measured using the PrimeFlow Assay (eBioscience), which couples intracellular RNA detection in single cells with protein measurement by flow cytometry analysis. With this technique, we detected intracellular RNA levels, measured as mean fluorescence intensity, within the gate of PMN-MDSCs from in vitro cultures (Fig. 5D, top). Direct CD40 stimulation increased Arg1, Nos2, and Stat3 transcripts as much or more than did stimulation with IL4 plus IFNγ, which we used as positive control (39). A similar increase was observed when PMN-MDSCs were cocultured in the presence of BMMCs. The response was blunted if antagonistic anti-CD40L was given or if BMMCs were from CD40L^−/− mice (Fig. 5D, bottom). Thus, mast cells partner with PMN-MDSCs to instigate suppressive activity via CD40L–CD40 cross-talk.

These results were consistent with the Tag-specific response observed in Kit^Wsh-TRAMP mice that, although impaired by reconstitution with wild-type, CD40L⁺, mast cells (Fig. 6A and B; Supplementary Fig. S7F–S7H), remained unaltered by mast cells reconstitution with BMMCs from CD40L^–/– mice. According to data on T-cell function, adenocarcinoma burden in Kit^Wsh-TRAMP mice reconstituted with CD40L^–/–- BMMCs was significantly reduced (17.2% in situ and 10.3% infiltrating adenocarcinoma on total tumor burden, respectively) in comparison to TRAMP (17.2% in situ and 78.5% infiltrating adenocarcinoma) and to Kit^Wsh-TRAMP mice reconstituted with wild-type BMMCs (28.1% in situ and 57.2% infiltrating adenocarcinoma; Fig. 6C and D), and it was even lower in comparison to unreconstituted Kit^Wsh-TRAMP mice (40.36% in situ and 30.11% infiltrating adenocarcinoma; Fig. 6C and D). These results indicate that expression of CD40L on mast cells supports their cross-talk with PMN-MDSC, favoring immune escape and adenocarcinoma development, but also suggests that other mechanisms dependent on CD40L expression on mast cells could be necessary for adenocarcinoma growth.

We have previously shown that in human prostate cancer, mast cells accumulate in well-differentiated tumors more than in poorly differentiated areas prone to neuroendocrine differentiation (10). Here we show, with immunohistochemistry on serial sections of human prostate cancer samples, the presence of cells positive for tryptase, CD40L, or CD33 (used as a myeloid cell marker; Supplementary Fig. S8A) in well-differentiated areas. Immunofluorescence confirmed coexpression of tryptase and CD40L on tumor infiltrating mast cells, and proximity of mast cells to CD33⁺ cells (Supplementary Fig. S8B and S8C). To test the translational relevance of our finding, we applied in silico analyses to datasets of prostate cancer patients to look for correlation between the expression of genes characterizing mast cells and PMN-MDSCs (Supplementary Table S2) and clinical outcome. The testing dataset comprised 54 primary prostate cancers associated with patients' follow up for biochemical relapse (Chiorino dataset; ref. 24). A total of 23 probes, corresponding to 20 upregulated genes, selected from the list of Supplementary Table S1 were statistically associated with higher biochemical recurrence (P < 0.001; Fig. 7A). These results were consistent with a publicly available dataset of 140 prostate cancer samples (Taylor dataset; ref. 25), in which the highest expression of genes belonging to the mast cell/PMN-MDSC signature correlated with biochemical relapse (P = 0.00467; Fig. 7B). In both datasets, we found a positive correlation between mast cell genes and genes related to MDSC suppressive function (Supplementary Fig. S9). To investigate the association with overall survival, we explored the Setlur dataset comprising 358 prostate cancer samples (26) and found a statistically significant inverse correlation of this gene signature with survival (P = 0.00512; Fig. 7C). The performance of this signature was tested against 1,000 random gene signatures of the same length for each of the three datasets with sigCheck, leading to statistically significant results for all of them (P < 0.05; Supplementary Fig. S10). Because the three datasets—derived from microarray gene expression analysis—were derived from different platforms, the original 20-gene signature used for the Chiorino dataset was only partially present in the other two datasets. The gene overlap and the signatures tested for each dataset are reported in Fig. 7D. These results indicate that concomitant upregulation of mast cell– and PMN-MDSC–related genes negatively correlates with clinical outcome of prostate cancer patients.

Prostate cancer remains one of the cancers most refractory to immunotherapy, including checkpoint inhibitors (4). A deeper understanding of the immunosuppressive networks established at the time of tumor development is necessary to fight this disease.

We have previously identified mast cells as key stromal accomplices in prostate cancer, capable of influencing the balance between adenocarcinoma and neuroendocrine histotypes (9, 10). Our data showing increase in the frequency of neuroendocrine tumors after pharmacologic inhibition of mast cell degranulation (refs. 9, 10, and this paper) indicate that mast cells can protect against emergence of neuroendocrine prostate cancer, most likely through release of as yet unidentified soluble factors. Mast cells can also foster the growth of prostate adenocarcinoma, at least partially due to their provision of MMP9, until the tumor becomes capable of its own MMP9 production during progression (10). However, adenocarcinoma formation was only moderately restrained in TRAMP mice pharmacologically treated to block mast cell degranulation (refs. 9, 10, and this paper). This suggests that other mechanisms implying interaction of mast cells with other cell types, through surface receptor–ligand pairs, might foster prostate adenocarcinoma.

Thus, our work shows that mast cells contribute to prostate adenocarcinoma through the suppression of antitumor T-cell responses, which in turn control onset and growth extent of adenocarcinoma. Immunosuppression was mediated by direct interactions between mast cells and PMN-MDSCs via CD40L–CD40 engagement, which resulted in enhanced PMN-MDSC suppressive activity.

Mast cells can augment MDSC suppressive functions (13, 34), and these cells can interact by binding of CD40L and CD40, at least in vitro (13). Other than confirming that CD40 expression is required for MDSC activity (38), our in vivo results show that mast cells interact with PMN-MDSCs through CD40L–CD40 signaling in the spleen of tumor-bearing mice, which results in suppression of tumor-specific T cells and unrestrained tumor growth. Depletion of mast cells or interference with the CD40L–CD40 interaction restored the CD8⁺ T-cell response and reduced tumor growth. Once stimulated by mast cells, PMN-MDSCs could mediate T-cell suppression directly in the spleen (40) or through migration to lymph nodes or the tumor site.

Adenocarcinoma burden was lower in Kit^Wsh-TRAMP mice reconstituted with CD40L^–/– BMMCs than in Kit^Wsh-TRAMP mice lacking mast cells, suggesting that other protumor functions of mast cells, perhaps related to cytokines or MMP release, could be fostered by reverse CD40L signaling (13, 41). Although several other myeloid cell types express CD40, mast cells selectively engaged with PMN-MDSCs, an issue that deserves further investigation.

Abnormal myelopoiesis leading to increased accumulation of immature granulocytic precursors (CD11b⁺Ly6G^lowLy6C^low) has been described in Kit^Wsh mice, especially on a mixed (BALB/c × C57BL6)F1 background (36). This CD11b⁺Ly6G^intLy6C^low population is distinct from canonical CD11b⁺Ly6G^hiLy6C^low PMN-MDSCs and CD11b⁺Ly6G^lowLy6C^hi M-MDSCs. CD11b⁺Ly6G^int Ly6C^low cells seem to accumulate regardless of mast cell presence, responding instead to an inversion of the cKit gene promoter (23). Indeed, we found in spleen of Kit^Wsh-TRAMP mice an accumulation of CD11b⁺Ly6G^intLy6C^int cells (2.87% ± 2.06% of CD45⁺ cells), which were very low in TRAMP mice (0.32% ± 0.28% of CD45⁺ cells). The frequency of these immature CD11b⁺Ly6G^int Ly6C^int myeloid cells was unaltered by reconstitution of Kit^Wsh-TRAMP mice with BMMCs. We also observed more CD11b⁺Ly6G^lowLy6C^hi M-MDSC–like cells and CD11b⁺Ly6G^hiLy6C^low PMN-MDSCs. Nevertheless, systemic reconstitution of Kit^Wsh mice with BMMCs corrects the mast cell defect and allows testing of specific mast cell functions in these mice (10, 23). Given that reconstitution of Kit^Wsh−TRAMP mice with wild-type or CD40L^–/–BMMCs induced different outcomes in term of PMN-MDSCs activity, tumor-specific CD8⁺ T-cell suppression and adenocarcinoma growth, we conclude that mast cells do interact with MDSCs in tumor-bearing mice, and that our results are not an artifact due to the intrinsic limitations of this murine model.

In our prostate cancer model, we found that only PMN-MDSCs were endowed with T-cell suppressive ability. In prostate cancer patients, accumulation of both CD14⁺HLA-DR⁻ M-MDSCs and CD15⁺CD33^low PMN-MDSCs has been reported (14, 15). In silico analyses corroborated our results in the TRAMP model showing that the mast cell/PMN-MDSC gene signature correlated with biochemical relapse and reduced survival of prostate cancer patients. Indeed, another study reported that peripheral blood of either metastatic breast or prostate cancer patients shows accumulation of CD33⁺HLA-DR⁻ cells (“early-stage” MDSCs; ref. 33), which expressed CD40 (42). In the same study, increased amounts of soluble CD40L were detected in the serum of the patients (42). Those data raised the question of whether the CD40L–CD40 pathway might be detrimental, inducing immune suppression rather than immune stimulation, in certain clinical settings (42). Consistent with this idea, in a clinical trial evaluating the efficacy of a viral vaccine against prostate cancer (PROSTVAC), higher serum levels of soluble CD40L correlated with reduced patient survival (43).

Our data suggest that CD40L–CD40 signaling may sustain immune suppression in those cancer patients characterized by accumulation of both mast cells and MDSCs. This observation might inform design of immunotherapies. Patients should be screened for the accumulation of mast cells and MDSCs before immunotherapy with agonistic CD40 antibodies (44), which our results indicate could be detrimental. On the other hand, blocking CD40L–CD40 interaction with antagonistic CD40 antibodies, such as lucatumumab (Novartis; ref. 44), may increase the antitumor immune response and might be a therapeutic addition that could be tested in combination with checkpoint inhibitors or other immunotherapies.

No potential conflicts of interest were disclosed.

Conception and design: E. Jachetti, M.P. Colombo

Development of methodology: E. Jachetti, P. Casalini, B. Frossi, C. Chiodoni

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): E. Jachetti, V. Cancila, A. Rigoni, L. Bongiovanni, B. Cappetti, B. Belmonte, C. Enriquez, P. Casalini, S. Sangaletti, C. Chiodoni, C. Tripodo

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): E. Jachetti, V. Cancila, L. Bongiovanni, B. Belmonte, P. Ostano, C. Chiodoni, G. Chiorino, C.E. Pucillo, C. Tripodo, M.P. Colombo

Writing, review, and/or revision of the manuscript: E. Jachetti, C.E. Pucillo, C. Tripodo, M.P. Colombo

Study supervision: E. Jachetti, C.E. Pucillo, M.P. Colombo

Other (provided external gene expression profiling datasets): G. Chiorino

Valeria Cancila, PhD student, is supported, for this research, by the University of Palermo (IT), Doctoral Course of Experimental Oncology and Surgery, Cycle XXXI.

This work was supported by grants from Fondazione Italo Monzino (to M.P. Colombo) and Associazione Italiana per la Ricerca sul Cancro (Investigator Grant n 14194 and Investigator Grant n 18425 to M.P. Colombo). E. Jachetti and A. Rigoni have been awarded by Fellowships from Fondazione Umberto Veronesi.

We thank from Fondazione IRCCS Istituto Nazionale Dei Tumori: Ivano Arioli, Laura Botti, Renata Ferri, and Mariella Parenza for technical assistance; Ester Grande for administrative assistance; and Ivan Muradore and Gabriella Abolafio for cell sorting. We thank Dr. Matteo Iannacone from San Raffaele Scientific Institute for having provided CD40L^−/− mice.

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