Anti-PD-L1 Efficacy Can Be Enhanced by Inhibition of Myeloid-Derived Suppressor Cells with a Selective Inhibitor of PI3Kδ/γ

Most head and neck squamous cell carcinomas (HNSCC) arise secondary to carcinogen exposure that contributes to genomic instability and production of neoantigens capable of being targeted by adaptive immunity (1–3). Accordingly, significant subsets of HNSCCs are T-cell inflamed (4, 5). Immune checkpoints, such a programmed cell death protein 1 (PD-1) and its ligand (PD-L1), play a critical role in adaptive immune resistance within the microenvironment of these tumors (6). PD-based checkpoint inhibition, recently FDA-approved for recurrent/metastatic HNSCC, results in clinically significant and durable responses in a subset of HNSCC patients (7). However, not all HNSCC patients with T-cell–inflamed tumors benefit from checkpoint inhibitor therapy, due at least to immunosuppression within the HNSCC tumor microenvironment. Myeloid-derived suppressor cells (MDSC) suppress effector immune cells and are a major driver of local immune suppression within HNSCCs (8–16).

Following recruitment into the tumor microenvironment through tumor cell–dependent chemokine expression (17), MDSCs mediate T-lymphocyte immunosuppression through STAT3-dependent mechanisms that include expression of arginase and inducible nitric oxide synthetase (iNOS; ref. 9). Cumulatively, MDSCs deplete the tumor microenvironment of nutrients critical for T-lymphocyte function and directly suppress immunity through the production of reactive oxygen species (ROS) and immunosuppressive cytokines (8, 15).

PI3K signaling links environmental cues to alterations in cellular metabolism, growth, and survival. Closely related class I PI3K isoforms include p110α and p110β, which are ubiquitously present in most cell types and frequently overexpressed or constitutively active in HNSCC (3), and p110δ and p110γ, which are expressed primarily in hematopoietic cells (18). Recent reports have demonstrated that functional inhibition or polarization of tumor-infiltrating myeloid cells with a p110γ-specific inhibitor (IPI-549) sensitized HPV⁺ and HPV⁻ models of HNSCC and B16F10 tumors engineered to recruit myeloid cells through constitutive GM-CSF production to PD-based checkpoint inhibition (19, 20). However, both the p110δ and p110γ isoforms of PI3K have been implicated in immunosuppression mediated by myeloid cells in solid tumors (21). IPI-145 is a selective p110δ/γ inhibitor that has been evaluated clinically for the treatment of hematopoietic malignancies (22, 23). We hypothesized that dual p110δ/γ inhibition with IPI-145 would reverse MDSC-mediated local immunosuppression and sensitize carcinogen-induced, HPV⁻ HNSCC tumors with robust myeloid cell recruitment to PD-based checkpoint inhibition. We demonstrated that selective p110δ/γ inhibition with low-dose IPI-145 partially abrogated local immunosuppression mediated by granulocytic MDSCs (gMDSC) and enhanced CD8-dependent responses to PD-L1 mAb therapy. IPI-145 suppressed gMDSC production of arginase and iNOS in a dose-dependent fashion both ex vivo and in vivo, leading to enhanced antigen-specific T-lymphocyte responses. However, IPI-145 also appeared to suppress tumor-infiltrating lymphocyte (TIL) function in a dose-dependent fashion, as high-dose IPI-145 treatment in vivo reversed tumor growth control observed with low-dose IPI-145 and PD-L1 mAb combination therapy through TIL suppression. These findings further validate the approach of targeting immunosuppressive myeloid cells with selective PI3K inhibition, but suggest that a therapeutic window may exist with dual p110δ/γ inhibition where greater suppression of MDSCs than TILs leads to enhanced responses to checkpoint inhibition.

Murine oral cancer (MOC) cells were provided by Dr. R. Uppaluri (Washington University School of Medicine, St. Louis, MO) to our laboratory in 2014 and have been cultured as described previously (24). MOC cells were validated to be of epithelial origin (25), and routinely tested for mycoplasma. All experiments were approved by the NIDCD Animal Care and Use Committee. To establish MOC tumors, 5 × 10⁶ MOC1 or 1 × 10⁵ MOC2 cells were injected subcutaneously into the flank of wild-type (WT) C57BL/6 (B6) mice (Charles River). IPI-145 (Active Biochem) was administered via oral gavage daily for 14 days. Control mice received oral gavage of vehicle (0.5% carboxymethylcellulose, 0.05% Tween 80 in ultra-pure water) alone. PD-L1 mAb (clone 10F.9G2, BioXCell), CD8 mAb (clone YTS 169.4), Ly6G mAb (clone 1A8), or isotype control antibody (clone LFT-2) treatments were performed via intraperitoneal injection (200 μg/injection).

All tissues were used fresh. Spleen and lymph nodes were processed by mechanical dissociation between frosted slides followed by RBC lysis. Dissected normal oral mucosa from WT B6 mice or tumor tissues were processed into single-cell suspensions by mincing, chemical (Murine Tumor Dissociation Kit, Miltenyi Biotec) and mechanical (gentleMACS, Miltenyi Biotec) dissociation per the manufacturer's protocol. Suspensions were filtered through a 100 μmol/L filter and washed with 1% BSA in PBS prior to blocking nonspecific staining with anti-CD16/32 (Biolegend) antibody. Cell surface staining was performed using fluorophore-conjugated anti-mouse CD45.2 clone 104, CD3 clone 145-2C11, CD4 clone GK1.5, CD8 clone 53-6.7, CD31 clone 390, PDGFR clone APA5, PD-L1 clone 10F.9G2, H2-K^b clone AF6-88.5, CD107a clone 1D4B, CD69 clone H1.2F3, PD-1 clone RMP1-30, CD11b clone M1/70, Ly6G clone 1A8, Ly6C clone HK1.4, and CD44 clone IM7 antibodies from Biolegend, and 41BB clone 17B5 and OX40 clone OX-86 were from eBioscience. FoxP3⁺ regulatory T-cell staining performed with the mouse regulatory T-Cell Staining Kit #1 (eBioscience) as per manufacturer's protocol. For intracellular phosphoprotein staining, cells were fixed and permeabilized using the Fixation and Permeabilization Buffer Set (eBioscience) as per manufacturer's protocol and stained with pAKT (S473) and pS6 (S240/244) antibodies (Cell Signaling Technology) or isotype (rabbit IgG) followed by goat anti-rabbit secondary antibody conjugated to APC (Biolegend). Dead cells were excluded via 7AAD (Biolegend) negativity for cell surface staining or Live/Dead cell viability dye (Thermo Scientific) negativity for intracellular staining. Isotype control antibodies and a “fluorescence minus one” method of antibody combination were used for specific staining validation. Data were acquired on a FACSCanto using FACSDiva software (BD Biosciences) and analyzed on FlowJo software vX10.0.7r2.

For ex vivo expression or functional T-lymphocyte analysis, splenic or lymph node suspensions were sorted on an autoMACS Pro Separator (Miltenyi Biotec) using the Pan T-Cell Kit (Miltenyi Biotec, negative selection) to select T lymphocytes or the Anti-Ly6G Microbead Kit (Miltenyi Biotec, positive selection) to select gMDSCs per manufacturer's protocol. To enrich draining lymph node T lymphocytes, tissues were processed into single-cell suspensions and subjected to negative T-cell magnetic selection alone. To enrich TILs, digested tumor single-cell suspensions were first enriched for lymphocytes using a 40/80% isotonic Percoll (Sigma) gradient (centrifuged at 325 × g for 23 minutes at room temperature), followed by positive selection of T lymphocytes using the CD3ϵ Microbead Kit (Miltenyi Biotec). To enrich tumor-infiltrating gMDSCs, a similar Percoll gradient was followed by gMDSC selection using the anti-Ly6G Microbead Kit. Purity of cells enriched from spleen and lymph node populations was consistently >90% and purity of cells enriched from tumor was consistently >95% as assessed by flow cytometry.

Whole-cell lysates were obtained using NP40 lysis buffer, mixed with NuPAGE LDS sample buffer and NuPAGE sample reducing agent (Life Technologies), heated at 95°C for 5 minutes and subjected to electrophoresis using 4%–12% Bis-Tris precast gels (Life Technologies) at 150 V for 100 minutes. The Invitrogen iBlot Dry Blotting System was used to transfer proteins onto a PVDF membrane. Primary antibodies were diluted in 5% BSA prepared from Tween 20-TBS: rabbit monoclonal anti-pAKT (Ser473) antibody, 1:2,000 (Cell Signaling Technology); rabbit monoclonal anti-pS6 (Ser240/244) antibody, 1:1,000 (Cell Signaling Technology); mouse monoclonal anti-β-actin, 1:5,000 (Calbiochem). Each blot was incubated with Chemiluminescent HRP Antibody Detection Reagent (Denville Scientific Inc.) and imaged using Image Studio software (LI-COR Biosciences).

In vitro cell viability was quantified via XTT assay (Trevigen) per manufacturer's instructions or via dual acridine orange and propidium iodine staining per manufacturer's instructions, quantified on a Cellometer Auto 2000 (Nexcelcom).

T cells isolated from naïve B6 spleens were stained with 5 μmol/L carboxyfluorescein succinimidyl ester (CFSE; Sigma) and stimulated using plate-bound CD3 (clone 145-2C11, eBioscience) and CD28 (clone 37.51, eBioscience) antibodies (26). For antigen-specific experiments, sorted OT-1 splenic T lymphocytes were exposed to irradiated (20 Gy) naïve splenocytes pulsed with OVA_257-264 (SIINFEKL; 1 μg/mL; InVivoGen). Where indicated, T cells were cocultured with MDSCs, IPI-145, nor-NOHA, and/or L-NMMA (Cayman Chemicals, 300 μmol/L each) for 4 hours prior to stimulation, and flow cytometry was used to quantify 72-hour CFSE dilution. Proliferation was quantified as the average number of divisions for all cells in the culture (division index) using FlowJo software (27). Media for all functional immune assays consisted of RPMI1640 supplemented with 10% FCS, 2 μmol/L β-ME, HEPES, nonessential amino acids, glutamine, and antibiotics.

Splenocytes from OT-1 mice were cultured in the presence of SIINFEKL (2 μg/mL) with daily 2:1 splitting. After 72 hours in culture, >80% of remaining cells are CD8⁺Vα2⁺ cells (data not shown). OT-1 CTLs were exposed to SIINFEKL-pulsed EL4 cells labeled with indium¹¹¹ with or without MDSCs and indicated inhibitors. Four-hour supernatants were analyzed for γ radiation counts on a WIZARD2 Automatic Gamma Counter (PerkinElmer).

T lymphocytes sorted from naïve spleen were stimulated using plate-bound CD3 and CD28 antibodies with or without MDSCs and indicated inhibitors. In other experiments, sorted splenic, lymph node, or tumor T cells from tumor-bearing mice were cocultured with IFNγ-pretreated (20 ng/mL, 24 hours) and irradiated (50 Gy) MOC1 cells at a 10:1 T-lymphocyte:MOC1 cell ratio for 48 hours in flat-bottom 96-well plates. Supernatant IFNγ levels were quantified by ELISA (eBioscience) per manufacturer's protocol.

RNA was extracted from MOC cells using the RNEasy Mini Kit (Qiagen) or from sorted immune cell subsets using the PicoPure RNA Isolation Kit (Thermo Scientific). cDNA was synthesized using high-capacity reverse transcription. Gene expression was determined relative to GAPDH using the indicated primers (Life Technologies) on a 7900HT Sequence Detection System (Applied Biosystems).

Tests of significance between pairs of data are reported as P values, derived using a Student t test with a two-tailed distribution and calculated at 95% confidence. Comparison of multiple sets of data was achieved with ANOVA with Tukey multiple comparisons. Survival analysis was determined by log-rank (Mantel–Cox) analysis. All error bars indicate SD. Statistical significant was set to P < 0.05. All analysis was performed using GraphPad Prism v7.

When injected into wild-type B6 mice, MOC1 cells generate T-cell–inflamed tumors capable of inducing immunologic memory and MOC2 cells generate non-T-cell–inflamed tumors that do not generate immunology memory (24, 28). Compared with naïve spleen and normal oral mucosa, spleens and tumors in MOC1 tumor-bearing mice demonstrated robust accumulation of myeloid cells (Fig. 1A, left). The majority of these CD11b⁺ cells displayed surface markers, indicating a granulocytic (Ly6G^hi, Ly6C^int) as opposed to a monocytic (Ly6G^low, high Ly6C^hi) phenotype. Analysis of immune cell distributions revealed expansion of these myeloid cell populations within spleen and tumors of MOC1 tumor-bearing mice (day 30) but not FoxP3⁺ regulatory T lymphocytes (Tregs; Fig. 1A, right charts). Using magnetic separation strategies, these granulocytic myeloid cells could be enriched for functional analysis (Supplementary Fig. S1). When evaluated in a T-lymphocyte proliferation assay, these cells inhibited CD4 and CD8 T-lymphocyte proliferation in a dose-dependent fashion (Fig. 1B). Splenic and tumor myeloid cells were evaluated for expression of immunosuppressive enzymes compared with nontumor-bearing spleen, and were found to have increased expression of arginase and iNOS (Fig. 1C) but not indoleamine 2,3-dioxygenase or NADPH oxidase 2 (Supplementary Fig. S2). Inhibition of CD4 and CD8 T-lymphocyte proliferation by these cells was partially reversed in the presence of norNOHA (arginase inhibitor) and L-NMMA (iNOS inhibitor) alone or in combination (Fig. 1D). Splenic and tumor myeloid cells were also evaluated for their ability to suppress OT-1 antigen-specific CTL lysis of SIINFEKL-pulsed target cells. Near complete suppression of 4-hour CTL lysis in the presence of splenic or tumor myeloid cells was significantly reversed in the presence of arginase and iNOS inhibition (Fig. 1E). Similar inhibition of T-lymphocyte proliferation and CTL function were observed with gMDSCs sorted from MOC2 tumors. Thus, granulocytic myeloid cells accumulate in the periphery and tumor microenvironment of MOC tumor–bearing mice and possess the ability to significantly suppress T-lymphocyte function in a process partially dependent upon arginase and iNOS, functionally validating them as gMDSCs.

To determine differential expression of PI3K class I isoforms in different cell types, we measured transcript levels in MOC1 and MOC2 cells as well as sorted TILs and gMDSCs from MOC1 tumors. Using the α isoform from MOC1 as a reference, PI3K δ and γ isoforms were expressed at higher levels in sorted TILs and gMDSCs compared with MOC cells (Fig. 2A). MDSCs express significantly more Pik3cg levels than T lymphocytes. Treatment with increasing doses of the selective PI3K δ and γ isoform inhibitor IPI-145 had little effect on MOC cell viability as measured by XTT viability assay (IC₅₀>10 μmol/L) and exclusion of propidium iodide (Fig. 2B). IPI-145 reduced phosphorylation of AKT and S6 downstream of PI3K in gMDSCs, but not MOC1 cells (Fig. 2C), consistent with immune cell dependence upon PI3K δ and γ isoforms for downstream signaling. Similar results were observed for MOC2 treated with IPI-145. To assess functional alteration, sorted splenic and tumor gMDSCs were exposed to IPI-145 (10 nmol/L) ex vivo for 4 hours. Although treatment at this concentration did not induce gMDSC loss of viability, expression of Arg1 and Nos2 transcript levels were significantly reduced in tumor MDSCs, whereas only Arg1 levels were reduced in splenic MDSC (Fig. 2D). This correlated with significant reversal of the ability of gMDSCs to suppress the proliferation of T lymphocytes, to a greater degree for tumor than splenic gMDSCs. Cumulatively, these data suggested that ex vivo PI3K δ and γ inhibition with IPI-145 treatment has little direct effect on MOC cell viability but can partially reverse the immunosuppressive phenotype of peripheral and tumor-infiltrating gMDSCs through alteration of Arg1 and Nos2 expression.

One potential drawback of PI3K δ/γ inhibition is direct suppression of T-lymphocyte function. Exposure of T lymphocytes sorted from naïve B6 mice stimulated with plate-bound CD3 and CD28 antibodies to IPI-145 resulted in dose-dependent suppression of proliferation (Fig. 3A). To validate these findings in an antigen-specific system, CD8 T lymphocytes were sorted from OT-1 spleens and exposed to SIINFEKL-pulsed target cells. Treatment with IPI-145 suppressed OT-1 CD8 T-lymphocyte IFNγ production and activation marker expression in a dose-dependent fashion (Fig. 3B) as well as proliferation (Fig. 3C). Similarly, the presence of IPI-145 inhibited the ability of antigen-specific CTLs to lyse target cells (Fig. 3D). Taken together, these data suggested that although PI3K δ/γ inhibition with IPI-145 appears to partially reverse the suppressive capacity of gMDSCs, it also appears to inhibit priming and effector function in antigen-specific T lymphocytes in a dose-dependent fashion.

We assessed the effect of IPI-145 monotherapy on MOC primary tumor growth and immune correlates. Given data that an in vivo dose of 10 mg/kg daily significantly altered neutrophil chemotaxis in rats (22), an initial dose of 15 mg/kg was administered. Daily IPI-145 treatment at 15 mg/kg did not alter MOC1 primary tumor growth to a statistically significant degree (Fig. 4A). However, analysis of tumor single-cell suspensions revealed enhanced infiltration of CD8 TILs and expression of the activation marker CD107a and exhaustion marker PD1 (Fig. 4B). CD4 TIL infiltration and OX40 expression were not significantly affected (Fig. 4C). Tumor cell specific MHC class I (H2-K^b) and PD-L1 expression on both tumor and endothelial cells was enhanced (Fig. 4D). To assess the effect of IPI-145 monotherapy on MDSC function, gMDSCs were sorted from treated MOC1 tumors and evaluated in a T-lymphocyte suppression assay. Although accumulation of MDSCs into MOC1 tumors was not altered and PD-L1 expression on gMDSCs appeared to be modestly enhanced by IPI-145 treatment (Fig. 4E), the ability of gMDSCs to directly suppress T lymphocytes proliferation was reduced (Fig. 4F). Tumor microenvironment PD-L1 levels were likely increased due to overall immune activation indicative of adaptive immune resistance (6), and these findings suggest that the immunoactivating effects of IPI-145 are independent of PD-L1 expression on tumor or infiltrating immune cells. T lymphocytes from tumor draining lymph nodes were sorted and stimulated with IFNγ pretreated and irradiated MOC1 cells. Suppression of gMDSC function with IPI-145 correlated with a significant increase in T-lymphocyte IFNγ production when exposed to MOC1 antigen (Fig. 4G), suggesting that although IPI-145 monotherapy at 15 mg/kg does not significantly affect MOC1 primary tumor growth, it does alter the immunosuppressive tumor microenvironment and enhance T-lymphocyte activation potential. Evaluation of spleens from the same animals revealed similar trends with modestly but significantly increased CD8 concentrations (Supplementary Fig. S3A), no change in MDSC concentrations, and reduced gMDSC suppression potential (Supplementary Fig. S3B), albeit to a lesser degree, than that observed in tumor. Treatment of non-T-cell–inflamed MOC2 tumor-bearing mice with IPI-145 15 mg/kg resulted in no measurable tumor growth inhibition, change in TIL infiltration or enhancement of MHC class I or PD-L1 expression indicative of modulation of the tumor microenvironment (Supplementary Fig. S4A–S4D).

Given the enhanced T-lymphocyte responses and expression of PD1 and PD-L1 within the tumor microenvironment following treatment of MOC1 tumor-bearing mice, we hypothesized that IPI-145 could prime MOC1 tumors to immune activation with PD-L1 mAb. Combination IPI-145 at 15 mg/kg and PD-L1 mAb resulted in significantly enhanced primary tumor growth control (Fig. 5A) and survival of MOC1 tumor-bearing mice (Fig. 5B) over either treatment alone. Individual tumor growth curves are shown in Figure 5C. Analysis of treated tumor single-cell suspensions revealed enhanced infiltration of CD8 TILs (Fig. 6A) and expression of CD8–T-lymphocyte activation and exhaustion markers (Fig. 6B). Tumor cell–specific MHC class I expression was also significantly enhanced (Fig. 6C). Similar to IPI-145 alone, combination treatment did not significantly alter MDSC accumulation within the tumor microenvironment (Supplementary Fig. S5). Tumor draining lymph node T lymphocytes were isolated and activated using nonspecific (CD3/28 antibodies) or antigen-specific (exposed to IFNγ pretreated and irradiated MOC1 cells) stimuli and found to have significantly enhanced activation potential with combination treatment over any treatment alone (Fig. 6D). To validate a CD8 T-lymphocyte–dependent mechanism of response to IPI-145 and PD-L1 mAb treatment, MOC1 tumor-bearing mice were treated with or without CD8 depleting antibody. Depletion of CD8 T lymphocytes resulted in near-complete abrogation of all treatment responses (Fig. 6E). The addition of IPI-145 treatment to mice treated with a gMDSC-depleting antibody alone or in combination with PD-L1 mAb resulted in similar levels of tumor control (Fig. 6F), suggesting that IPI-145 primarily exerts is effects through gMDSC suppression and not suppression of other cell types. Treatment of non-T-cell–inflamed MOC2 tumor-bearing mice with IPI-145 15 mg/kg plus PD-L1 mAb resulted in no measurable tumor growth inhibition (Supplementary Fig. S6). Thus, IPI-145 and PD-L1 mAb treatment can enhance antitumor responses over either treatment alone in T-cell–inflamed MOC1 tumors, in a CD8 T-lymphocyte–dependent fashion at least in part through IPI-145 induced partial reversal of gMDSC-suppressive capacity.

Given subtotal reversal of gMDSC-suppressive capacity with IPI-145 administered at 15 mg/kg, we hypothesized that an increased dose could further suppress gMDSC-immunosuppressive activity in MOC1 tumor-bearing mice. Here, treatment with low-dose (15 mg/kg) but not high-dose (50 mg/kg) IPI-145 alone resulted in a modest but statistically significant enhancement of survival (Fig. 7A, left). When combined with PD-L1 mAb, treatment with low-dose IPI-145 significantly enhanced primary tumor control and prolonged survival consistent with prior experiments; however, combining high-dose IPI-145 with PD-L1 mAb failed to demonstrate such responses (Fig. 7A, right). When evaluated functionally, splenic and tumor gMDSCs isolated from treated mice demonstrated significant reversal of immunosuppressive capacity in a dose-dependent fashion with greater reversal of suppression observed in tumors treated with high-dose IPI-145 (Fig. 7B). To evaluate the cumulative effect of increasing in vivo IPI-145 doses on adaptive immunity, T lymphocytes were isolated from the spleens, tumor draining lymph nodes and tumors of treated mice and stimulated in an antigen-specific fashion with IFNγ pretreated and irradiated MOC1 cells (Fig. 7C). Although peripheral T-lymphocyte responses were suppressed with both low- and high-dose IPI-145 in a dose-dependent fashion, enhanced responses observed with low-dose IPI-145 in both draining lymph node T lymphocytes and TILs were significantly attenuated with high-dose IPI-145. Thus, abrogated tumor control and survival observed with high-dose IPI-145 and PD-L1 mAb appears to be due to attenuation of T-lymphocyte responses within the tumor microenvironment despite enhanced reversal of gMDSC-suppressive capacity.

Demonstration of durable responses following checkpoint inhibition in patients with solid tumors has led to FDA approval for multiple cancer types including recurrent/metastatic HNSCC (7). However, only about 20% of patients demonstrate clinical responses (7). Similar to human HNSCC, MOC1-derived tumors lack significant responses to PD-based checkpoint inhibition alone, despite the presence of known tumor rejection antigen and being T-cell inflamed (24), and provide a model to study resistance to PD-1/PD-L1 blockade. MOC2-derived tumors model non-T-cell–inflamed tumors with few genetic alterations that are less likely to respond to immunotherapy. Here we demonstrated that functional inhibition of gMDSCs through pharmacologic p110δ/γ inhibition partially reversed suppression of T-lymphocyte proliferation and cytolytic capacity ex vivo and enhanced responses to PD-L1 mAb in T-cell–inflamed MOC1 but not non-T-cell–inflamed MOC2 tumors in vivo. These data suggest that, similar to responses to checkpoint inhibitors alone (29, 30), patients with identifiable antigens and T-cell–inflamed tumors are more likely to demonstrate responses to alteration of the tumor microenvironment through MDSC inhibition. The lack of responses observed in non-T-cell–inflamed MOC2 tumors could have many explanations, including lack of a tumor rejection antigen or other cell-intrinsic mechanisms of immune escape.

Mechanistically, the effect of IPI-145 was MOC tumor cell independent, and p110δ/γ inhibition suppressed tumor gMDSC arginase and iNOS expression and suppressive capacity but not MDSC viability or tumor infiltration. Furthermore, IPI-145 treatment did not block PD-L1 expression within the tumor microenvironment, but rather enhanced it in an effect consistent with adaptive immune resistance. An IPI-145 dose-dependent therapeutic window exists as high-dose IPI-145 reversed the sensitization of MOC1 tumors to PD-L1 blockade. Though high-dose IPI-145 reversed MDSC suppressive capacity more than low-dose in vivo, it also appeared to directly inhibit antigen-specific T-lymphocyte activity. Many of these alterations were significantly different between the peripheral and tumor microenvironment immune compartments, highlighting the pitfall of relying on peripheral immune alterations as a surrogate measure of changes within a complex tumor microenvironment.

The role of PI3K signaling in the regulation of arginase and iNOS expression in myeloid cells is established (31, 32), with contributory regulation by STAT6 and suppressor of cytokine signaling 1 (SOCS1; refs. 33, 34). Effectively modeling how best to alter PI3K signaling within myeloid cells to abrogate suppressive capacity while limiting undesirable effects on effector immune cells has important clinical implications as trials combining PI3K inhibitors and immunotherapies are underway. Recent reports have demonstrated the therapeutic role of a PI3Kγ selective inhibitor, IPI-549, in abrogating tumor-infiltrating myeloid cell suppression in multiple preclinical models (19, 20). However, both p110γ (35) and p110δ (21) have been shown to contribute to MDSC-mediated immunosuppression in genetic deletion models. Although myeloid cells express both the δ and γ isoforms, T-lymphocytes express δ but very little γ, suggesting that a dual p110δ/γ could lead to undesirable T-lymphocyte suppression (36). Here we demonstrate undesirable suppression of antigen-specific T-lymphocyte function with high-dose but not low-dose IPI-145. Ali and colleagues also demonstrated a degree of CTL inhibition with pharmacologic inhibition of p110δ, but concluded that more robust inhibition of tumor-infiltrating Tregs and myeloid cells shifted the overall balance toward antitumor immunity (21). Consistent trends among these studies and ours indicate that functional inhibition of immunosuppressive cell subsets within the tumor microenvironment with selective PI3K isoform inhibition represents a rational approach to enhancing baseline or checkpoint inhibitor induced antitumor immunity. However, therapeutic windows likely exist due to undesirable suppression of effector T-lymphocyte function with inhibition of p110δ. Clearly, the use of selective PI3Kδ/γ inhibitors to treat solid tumors would require careful study of different scheduling regimens and doses in the clinical setting.

HNSCCs demonstrate robust accumulation of MDSCs with immunosuppressive capacity (9, 15, 37), but Tregs also contribute to immunosuppression within the HNSCC tumor microenvironment (10, 38). Treg function appears to rely upon p110δ signaling (21). Time course analysis demonstrates robust accumulation of potently immunosuppressive gMDSCs but not Tregs in MOC tumors, suggesting that gMDSCs are the predominant driver of local immunosuppression with tumor progression in these models. Our inability to assess the effect of dual PI3Kδ/γ inhibition on Tregs due to their paucity within these models is a limitation of our study. Early infiltration of Tregs into solid tumors may play a critical role in establishing local immunosuppression (39, 40), and IPI-145 could modulate their function. The lack of additional growth suppression with IPI-145 treatment in the setting of gMDSC depletion strongly suggests that the mechanism of IPI-145 was gMDSC-dependent in our experiments, but evaluation of the role of IPI-145 in preclinical models with Treg-mediated immunosuppression would add clarity to this issue. Although survival of MOC1 tumor–bearing mice was significantly enhanced with combination low-dose IPI-145 and PD-L1 mAb, no established tumors were rejected. Our study does not clearly define whether this is due to subtotal reversal of MDSC-suppressive capacity or some other immunosuppressive or tumor cell–intrinsic mechanism.

In conclusion, dual PI3Kδ/γ inhibition with low-dose IPI-145 sensitizes T-cell–inflamed MOC1 oral cavity cancers to PD-L1 checkpoint inhibition through at least modulation of gMDSC arginine and iNOS expression and T-lymphocyte–suppressive capacity. High-dose IPI-145 reversed this enhanced response to PD-L1 checkpoint inhibition due to direct inhibition of T-lymphocyte responses. Such demonstration of a therapeutic window where gMDSC function is modulated to a greater degree than T-lymphocyte function critically informs the design of clinical trials utilizing selective PI3K inhibitors and immunotherapy.

No potential conflicts of interest were disclosed.

Conception and design: R.J. Davis, H. Cash, Z. Chen, C. Van Waes, C. Allen

Development of methodology: R.J. Davis, P.E. Clavijo, H. Cash, C. Allen

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): R.J. Davis, E.C. Moore, P.E. Clavijo, H. Cash, C. Silvin, C. Allen

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): R.J. Davis, P.E. Clavijo, H. Cash, Z. Chen, C. Van Waes, C. Allen

Writing, review, and/or revision of the manuscript: R.J. Davis, H. Cash, Z. Chen, C. Silvin, C. Van Waes, C. Allen

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): E.C. Moore, P.E. Clavijo, H. Cash, Z. Chen, C. Allen

Study supervision: C. Allen

Other (assisted in generating scientific data for this paper): J. Friedman

We thank Ravindra Uppaluri for mentorship, Sophie Carlson for technical assistance, and Jeffrey Schlom and James Hodge for collaborative resource sharing.

C. Allen was supported by the Intramural Research Program of the NIH, NIDCD, project number ZIA-DC000087 and the American Academy of Otolaryngology/American Head and Neck Society Duane Sewell Young Investigators Combined Award. R.J. Davis was supported by through the NIH Medical Research Scholars Program, a public–private partnership supported jointly by the NIH and generous contributions to the Foundation for the NIH from Pfizer Inc., The Doris Duke Charitable Foundation, The Newport Foundation, The American Association for Dental Research, The Howard Hughes Medical Institute, and the Colgate-Palmolive Company, as well as other private donors.

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.