Cooperative Immune-Mediated Mechanisms of the HDAC Inhibitor Entinostat, an IL15 Superagonist, and a Cancer Vaccine Effectively Synergize as a Novel Cancer Therapy

This article is featured in Highlights of This Issue, p. 521

Immunotherapy has induced unprecedented long-lasting clinical responses across multiple carcinoma types (1). However, the majority of patients fail to respond or develop resistance. Clinical benefit may be achieved by rationally combining immunotherapies to simultaneously target multiple components of the immune system and tumor microenvironment (TME), shifting the balance of immune suppression and activation to promote a more inflamed TME. This encompasses generating and expanding tumor antigen–specific immune responses, increasing their cytolytic potential, and promoting their infiltration into the TME. To this end, we utilized the combination of a therapeutic cancer vaccine against a tumor-associated antigen (TAA), the immunostimulatory cytokine N-803, and the histone deacetylase inhibitor (HDACi) entinostat.

To promote antigen-specific immune responses, we utilized adenoviral vaccines targeting the TAAs human carcinoembryonic antigen (CEA) or murine Twist1. This novel recombinant adenoviral vaccine platform has previously been described (2, 3). CEA has been shown to be highly expressed in multiple human carcinomas, but not murine (4). In order to target human CEA in mice, we implanted MC38 cells stably transfected with human CEA into CEA-transgenic mice (CEA-Tg) and vaccinated against CEA to overcome immune tolerance to the now murine self-antigen CEA (5). Vaccines targeting CEA have been shown to induce CEA-specific T cells and delay tumor growth in the absence of autoimmunity (3, 6–8). Alternatively, we vaccinated against Twist1, a human and murine transcription factor associated with the metastatic process and poor patient prognosis (9). We previously demonstrated that vaccinating against Twist1 reduces tumor growth (10).

To expand antigen-specific T cells generated from vaccination and/or cascade responses, we utilized N-803, comprising an IL15 superagonist (IL-15N72D) bound to an IL15 receptor alpha/IgG1 Fc fusion protein (formerly ALT-803). N-803 has enhanced biological activity and a substantially longer half-life than native IL15 (11). Additionally, N-803 increases CD8⁺ T-cell and natural killer (NK) cell numbers and function to promote antitumor efficacy in multiple murine tumor models (11–13). Currently, N-803 is being examined clinically in combination with multiple therapies, including vaccine.

Mounting preclinical and clinical evidence suggests that combining HDAC inhibitors with immunotherapy generates potent antitumor efficacy (14, 15). In particular, the class I HDACi entinostat reduces the number and/or function of immunosuppressive regulatory T cells (Treg) or myeloid-derived suppressor cells (MDSC; refs. 16–20) and alters the tumor phenotype to be more amenable to immune killing (21–23). Therefore, we hypothesized that entinostat would increase the antitumor efficacy of N-803 combined with a vaccine. Here, we demonstrate that this novel combination promotes significant antitumor efficacy in 4T1 triple-negative breast (TNB) and MC38-CEA colon carcinoma models. We provide a comprehensive analysis of immune subsets in the spleen and TME, identifying how the effects of each therapy cooperate to produce significant antitumor efficacy. Associated pathways were investigated by gene-expression analysis of the TME. Notably, the combination therapy improved CD8⁺ T-cell function, activation, and tumor infiltration. Overall, these findings provide a rationale for combining entinostat, N-803, and a therapeutic cancer vaccine in the clinical setting.

4T1 murine TNB cancer cells from the American Type Culture Collection were maintained according to supplier recommendations. MC38 murine colon carcinoma cells expressing human CEA (MC38-CEA) were generated and maintained as previously described (24). All cell lines were low passage and free of Mycoplasma as determined by MycoAlert Mycoplasma Detection Kit (Lonza).

The adenoviral Ad5 [E1-, E2b-]-CEA (Ad-CEA) vaccine was constructed as previously described (3). Under Cooperative Research and Development Agreements with the NCI, entinostat was provided by Syndax, and N-803, Ad-CEA and Ad-Twist vaccines were provided by ImmunityBio. Entinostat was formulated in a low-fat diet with 35% sucrose for target murine doses of 3, 6, 9, and 12 mg/kg/day (Research Diets). A dose of 12 mg/kg/day was used with CEA-Tg and nod-scid-gamma (NSG) mice and 3 mg/kg/day for Balb/c mice. Entinostat was quantified in plasma as previously described (25).

Six- to 8-week-old female Balb/c, NSG, and C57BL/6 CEA-Tg mice (5) were obtained from the NCI Frederick Cancer Research Facility (Frederick, MD). Mice were housed in microisolator cages under pathogen-free conditions, in accordance with Association for Assessment and Accreditation of Laboratory Animal Care guidelines.

MC38-CEA cells (3 × 10⁵) were implanted subcutaneously (s.c.) into the right flank of CEA-Tg or NSG mice. 4T1 cells (5 × 10⁴) were orthotopically implanted into the mammary fat pad of Balb/c mice. Once tumors reached 25 mm³ on average, mice were randomized to receive control (Cntrl) or entinostat (ENT) diet alone, or in combination with Ad-CEA or Ad-Twist vaccine (V, 10¹⁰ virus particles, s.c.). Vaccinated mice received vaccine boosts plus N-803 (1 μg, s.c.) 14 and 21 days after tumor implant. Tumor volume was measured twice weekly and calculated by (length² × width)/2. For tumor rechallenge, tumor cells were implanted as before into cured mice and naïve controls at least 1 month after tumor rejection. Quantification of 4T1 lung metastasis was performed as previously described (26).

Additional materials and methods are described in the Supplementary Data, including Supplementary Table S1, listing the antibodies used for flow cytometry.

Statistical analyses were performed in GraphPad Prism 7 (GraphPad Software) and listed in figure legends. Statistical significance was set at P < 0.05. *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001.

All animal studies were conducted under approval of the NIH Intramural Animal Care and Use Committee.

As monotherapies, entinostat, N-803, and therapeutic cancer vaccines targeting TAAs have demonstrated significant, but not curative, antitumor efficacy in murine models of solid carcinomas (3, 7, 10, 12, 16–22, 27–30). In the MC38-CEA colon cancer model, administration of N-803 with Ad-CEA vaccination produced significant, although modest, antitumor efficacy (Supplementary Fig. S1A). Using antigen-specific 9-mer peptides in an ELISPOT, we examined T-cell responses against the vaccination antigen CEA and the cascade antigens, janus kinase 1 (Jak1), prostaglandin F receptor (ptgfr), and the retroviral protein p15e. Jak1 and ptgfr are two recently identified neoepitopes in this model (31). There was a trend toward N-803 plus vaccine generating superior antigen-specific T-cell responses against the vaccination and cascade antigens relative to either agent alone. More importantly, vaccine promoted greater responses to neoantigens compared with N-803 alone (Supplementary Fig. S1B). Given the immune altering effects of entinostat (16–23), we hypothesized that its addition to N-803 plus vaccine would synergistically enhance tumor control. Given the short half-life of entinostat in mice, entinostat was administered continuously in the diet to better mimic the longer half-life observed in cancer patients (32, 33). MC38-CEA tumor–bearing mice fed an entinostat diet had plasma levels of entinostat within the therapeutic window achieved in patients (Supplementary Fig. S1C; ref. 33).

Next, the antitumor efficacy of entinostat combined with N-803 and Ad-Twist vaccination was examined in the 4T1 TNB cancer murine model. Entinostat, N-803 plus vaccine, and the combination of the three therapies minimally, but significantly, reduced primary tumor growth versus control (Fig. 1A). However, the combination generated a greater reduction in the number of lung metastases (73%) than either therapy alone. Entinostat or N-803 plus vaccine reduced lung metastasis by 42% and 46%, respectively (Fig. 1B and C).

The efficacy of this combination was assessed in a second tumor model using MC38-CEA. Although N-803 plus vaccine had no significant antitumor effects, entinostat alone significantly delayed tumor growth and enhanced median overall survival (OS) from 25 days (control) to 55 days, curing 5% of mice (Fig. 1D and E). This effect was immune mediated as entinostat did not affect tumor growth in immune-deficient NSG mice (Supplementary Fig. S1D). Survival was further improved to 73 days with the combination, curing 28% of mice (Fig. 1E). Upon tumor rechallenge, 67% of cured mice had a protective memory response (Fig. 1F). Overall, our data suggest that combining entinostat, N-803, and a tumor-specific vaccine has significant antitumor efficacy in two distinct murine models of solid carcinomas.

Because combining multiple therapies can lead to adverse effects, we monitored toxicity of the combination therapy in the MC38-CEA model. Mice showed no significant weight change during the course of treatment (Supplementary Fig. S2A). Via organ histopathology, fatty changes to the liver were observed in mice treated with entinostat alone or in combination, with no other treatment abnormalities identified (Supplementary Fig. S2B). Serum chemistry analysis identified a modest, albeit significant elevation in chloride and albumin in mice treated with entinostat alone, with no significant effects observed with combination therapy (Supplementary Fig. S2C). Of note, splenic cellularity was reduced with entinostat alone or in combination, which was associated with a large decrease in B cells (Supplementary Fig. S3A and S3B). This effect is consistent with transient lymphopenia reported in patients treated with entinostat (32, 33). Overall, the combination of entinostat, N-803, and vaccine was well tolerated.

To deduce the mechanism of antitumor activity, we monitored the systemic effects of treatment via serum cytokines. For all mechanistic investigations in the MC38-CEA model, mice were sacrificed on day 24 after tumor implant. Combination treatment significantly elevated the proinflammatory cytokines IFNγ and TNFα 3-fold and 2-fold, respectively, whereas entinostat or N-803 plus vaccine alone had no significant effect (Fig. 1G). The anti-inflammatory cytokine IL10 was not altered relative to control mice, whereas the immunosuppressive cytokine TGFβ1 was decreased 66% with entinostat alone or in combination (Fig. 1G). Entinostat alone or in combination reduced CXCL1 (KC/GRO) serum levels to that of a naïve mouse (Fig. 1G). Recently, ablation of CXCL1 was implicated in enhanced T-cell tumor infiltration and response to immunotherapy (34). These data suggest that entinostat is synergizing with N-803 plus vaccine to promote a proinflammatory cytokine environment.

To determine which immune effector subset was required for antitumor efficacy of the combination therapy, the effect of depleting each cell type on MC38-CEA tumor growth and survival was examined. Depletion efficacy was confirmed in the blood (Fig. 2A–C). CD8⁺ T cells were required for the antitumor efficacy of the combination therapy as OS decreased from 58.5 days with combination therapy to 29 and 28 days with CD8 or CD8+NK depletion, respectively. Surprisingly, CD4 depletion cured 80% of combination-treated mice. NK cell depletion had no effect on tumor growth and survival. Notably, even with CD8+NK depletion, tumor growth was still significantly reduced compared with control mice, suggesting additional components may be providing minor antitumor benefits with combination treatment (Fig. 2D and E). These data demonstrate that CD8⁺ T cells are essential for the antitumor efficacy of the combination therapy.

To further elucidate the mechanism of action of the combination therapy, we examined the immune subsets in the periphery (spleen and tumor-draining lymph node). Because results were extremely similar in both tissues, only data from the spleen are shown. Concordant with previous reports (16, 29), entinostat treatment alone or in combination significantly reduced splenic Tregs and FoxP3 expression (Fig. 3A–C). Despite reduced Treg numbers, their proliferative capacity (Ki67) was increased by entinostat (Fig. 3C). Notably, treatment with entinostat alone or in combination reduced V-domain Ig suppressor of T-cell activation (VISTA) levels on Tregs (Fig. 3C). VISTA is an immune checkpoint expressed on human and murine lymphocytes that negatively regulates T-cell function and promotes conversion of CD4⁺ T cells into Tregs (35–37). Also, VISTA has been implicated as an immune evasion mechanism for patients on immunotherapy (38, 39). Together, these data suggest that combination therapy is reducing splenic Treg numbers and potentially decreasing their immunosuppressive capacity.

Entinostat has also been reported to reduce the number and function of MDSCs (17–20). Entinostat treatment reduced the number of splenic monocytic-MDSCs (M-MDSC) and granulocytic-MDSCs (G-MDSC) due to the decreased splenic cellularity. However, the frequency of G-MDSCs, but not M-MDSCs, increased with entinostat treatment (Supplementary Fig. S4A and S4B). Combination treatment also elevated programmed death-ligand 1 (PD-L1) and decreased VISTA on both MDSC subsets (Supplementary Fig. S4A and S4B).

Because entinostat reduced VISTA on both Tregs and MDSCs, this effect was further interrogated by treating non–tumor-bearing CEA-Tg and Balb/c mice with different doses of entinostat. In both strains, entinostat significantly decreased VISTA on CD4⁺ and CD8⁺ T cells, G-MDSCs, and M-MDSCs (Supplementary Fig. S5). These data suggest a generalized mechanism by which entinostat reduces VISTA levels on immune cells. Overall, these data suggest that entinostat alone or in combination is reducing the immune suppression in the periphery by reducing Tregs and VISTA expression.

Because the decrease in immune suppression suggests that T cells may be more active, we examined the effects on T cells. We observed that entinostat alone or in combination reduced CD4 numbers due to decreased splenic cellularity. The proportion of splenic CD4⁺ T cells increased with entinostat (Supplementary Fig. S6A). Combination therapy induced a more active phenotype on CD4⁺ T cells by increasing the activation marker CD25 and decreasing VISTA (Supplementary Fig. S6A and S6B). Combination therapy produced similar effects in the blood of 4T1 tumor–bearing mice, reducing VISTA levels, while the frequency of activated (CD44^hi) or total CD4⁺ T cells remained unchanged (Supplementary Fig. S6C).

Combination therapy did not significantly alter the frequency of CD8⁺ T cells, but a reduction in cell number was observed due to decreased spleen cellularity (Fig. 3D). Both N-803 plus vaccine and combination therapy promoted a central memory phenotype in splenic CD8⁺ T cells, an effect previously reported with N-803 (Fig. 3E; refs. 12, 13, 40). Combination therapy promoted an active phenotype on CD8⁺ T cells by increasing CD44^hi, CD25, and Ki67 and decreasing VISTA (Fig. 3F and G). With the exception of CD25, the activation phenotype was driven by N-803 plus vaccine as these effects were not observed with entinostat alone. In addition, the ratio of splenic CD8⁺ T cells to Tregs was significantly increased with entinostat alone or in combination (Fig. 3H). In the blood of 4T1 tumor–bearing mice, similar effects were observed in the frequency of total and CD44^hi CD8⁺ T cells and levels of VISTA with combination therapy (Supplementary Fig. S7). Altogether, these data suggest that in the spleen, combination therapy increases CD8⁺ T-cell activation, while reducing Treg numbers and their suppressive phenotype, thus shifting the immune balance toward a more activated, less suppressed peripheral environment.

Next, we investigated the effect on immune subsets in the TME. Although the reduction of Tregs in the spleen was driven by entinostat, in the TME, both N-803 plus vaccine and combination therapy reduced the number of infiltrated Tregs (Fig. 4A). The proliferative capacity (Ki67) of tumor-infiltrating Tregs was augmented with entinostat and combination therapy (Fig. 4B). Of note, VISTA expression was reduced, whereas granzyme B levels were significantly elevated with combination therapy in tumor-infiltrated Tregs (Fig. 4C and D). A previous study suggested that Tregs utilize granzyme B and perforin to inhibit CD8⁺ T cells in the TME (41). Therefore, although the enhanced granzyme B levels suggest Tregs may be more suppressive on a per cell basis, overall immune suppression by Tregs may be reduced due to markedly decreased tumor infiltration.

Tumor infiltration of either G-MDSCs or M-MDSCs was unaffected by any treatment (Supplementary Fig. S8A and S8C). Combination therapy decreased VISTA expression and enhanced PD-L1 levels on M-MDSCs with no effect on G-MDSCs (Supplementary Fig. S8B and S8D). Altogether, these data suggest that the combination therapy has mixed effects on the MDSC inhibitory phenotype.

We were particularly interested in investigating the impact of combination therapy on tumor-infiltrating T lymphocytes (TIL) given the depletion results. Combination therapy significantly enhanced CD8⁺ TILs 4.8-fold over control (Fig. 4E and F). These infiltrated cells had primarily an effector (Teff) or effector memory (Tem) phenotype, driven by entinostat (Fig. 4G). Additionally, CD8⁺ TILs had an activated phenotype denoted by increased CD25 and Ki67 expression (Fig. 4H). Overall, combination therapy promoted an 18-fold increase in the ratio of CD8⁺ T cells to Tregs in the TME (Fig. 4I).

Contrary to the CD8 results, we observed no change in CD4⁺ TILs (Supplementary Fig. S8E). CD4⁺ T cells displayed a mixed activation profile with increased Ki67 and reduced VISTA (inhibitory checkpoint), but decreased CD25 (activation; Supplementary Fig. S8F and S8G). Altogether, these data suggest that combination therapy is promoting the infiltration and activation of CD8⁺ T cells with a reduction in Treg tumor infiltration, resulting in a significant increase in the CD8/Treg ratio in the TME.

The striking increase in CD8⁺ TILs induced by combination therapy led us to interrogate possible mechanisms of tumor infiltration such as chemokines and chemokine receptors. In particular, CXCR3 is one of the main chemokine receptors involved in trafficking of CD8⁺ T cells into the TME (42, 43). Combination therapy or N-803 plus vaccine alone significantly increased the frequency of CXCR3⁺ CD8⁺ T cells in the spleen. CXCR3 was minimally expressed on CD8⁺ TILs (Supplementary Fig. S9A).

Both combination therapy and N-803 plus vaccine alone increased splenic CXCR3⁺ CD8⁺ T cells, yet the superior TME infiltration was observed only with combination therapy, suggesting that the addition of entinostat induces CD8 tumor infiltration. Thus, we postulated that entinostat could be modulating chemokines and/or chemokine receptors in the TME, as previously reported (44). NanoString analysis of total tumor RNA revealed a significant upregulation in cytokine and chemokine signaling with entinostat alone or in combination, including many receptor and ligand combinations implicated in CD8 tumor infiltration (refs. 42, 43, 45–47; Fig. 4J and K). Specifically, entinostat elevated both the receptor CXCR3 and its ligands CXCL9, CXCL10, and CXCL11, as well as CCR5 and its ligands CCL3, CCL4, and CCL5 (Fig. 4K). Chemokine protein levels were also examined in the tumor lysate using a bead-based immunoassay. Despite the transcripts for CXCL9 (MIG) and CXCL10 (IP-10) being elevated with combination therapy, protein levels in the tumor lysate were not enhanced, with CXCL10 being significantly reduced (Supplementary Fig. S9B). However, consistent with the transcript levels, CXCL13 and CCL4 (MIP-1B) proteins were significantly elevated with entinostat alone or in combination. CCL2 (MCP-1), CCL3 (MIP-1A), and CCL5 (RANTES) proteins were increased 1.6- to 2-fold with combination therapy versus control, albeit not significantly (Fig. 4L). Of note, chemokines were examined on day 24 when significant CD8 tumor infiltration was already present. Nonetheless, these data suggest a possible mechanism for the enhanced CD8 tumor infiltration whereby entinostat is modulating tumor chemokines to attract T cells, whereas N-803 plus vaccine is elevating the expression of the receptor CXCR3 on CD8⁺ T cells in the periphery.

HDAC inhibitors have been shown to modulate MHC class I and antigen-processing machinery in vitro in human carcinoma cells, increasing their recognition and lysis by antigen-specific cytotoxic CD8⁺ T cells (21–23). Interestingly, NanoString analysis of TME showed that entinostat alone or in combination elevated genes involved in antigen processing and presentation, including several MHC genes (Supplementary Fig. S10A). This elevation in MHC gene expression was also observed at the protein level in the nonimmune, CD45-negative cells, with entinostat treatment alone or in combination increasing MHC class I (H-2K^b, H-2D^b) expression on a per cell basis (Supplementary Fig. S10B and S10C).

To assess the direct effect of entinostat on tumor cells, we exposed MC38-CEA and 4T1 cells in vitro to a clinically relevant concentration (C_max = 0.5 μmol/L) of entinostat. Entinostat exposure increased the expression of MHC class I on MC38-CEA cells and 4T1 cells, suggesting these tumor cells would be more amenable to T-cell recognition and killing (Supplementary Fig. S10E, S10F, S10H, and S10I). Of note, both in vivo and in vitro exposure to entinostat elevated the expression of PD-L1 on MC38-CEA or 4T1 cells (Supplementary Fig. S10D, S10G, and S10J). Although increased PD-L1 expression may be decreasing the antitumor efficacy of the combination therapy by impeding T cells, the combination therapy is still producing significant tumor control.

Given the necessity of CD8⁺ T cells for the antitumor efficacy of combination therapy, the effect on T-cell function was investigated. Previously, we demonstrated that the combination of vaccine plus N-803 enhanced antigen-specific T-cell responses (Supplementary Fig. S1B). The addition of entinostat to the duet further improved the number of IFNγ-producing cells specific for CEA, p15e, and neoepitopes relative to either therapy alone (Fig. 5A and B). These data indicate that the combination of all three therapies is required for significant induction of CEA and cascade immune responses.

Next, we investigated the ability of T cells to produce IFNγ and TNFα upon T-cell receptor (TCR) stimulation. The treatment effects on CD4 cytokine production were substantially larger in the TME versus spleen, with combination therapy significantly enhancing CD4 IFNγ and TNFα production in the TME (Supplementary Fig. S11A–S11F). Moreover, the percentage of bifunctional cells in the TME that produce both cytokines was increased with combination therapy (Supplementary Fig. S11F). These data suggest that CD4⁺ T cells provide cytokine support in the TME.

Combination therapy enhanced the frequency of splenic CD8⁺ T cells that produce IFNγ or TNFα and the amount produced per cell, as well as the frequency of bifunctional cells (Fig. 5C–E). This effect was driven by N-803 plus vaccine as entinostat did not elevate cytokine production. In the TME, combination therapy increased production of TNFα [% and geometric mean fluorescence intensity (gMFI)], but not IFNγ (Fig. 5F and G). Only N-803 plus vaccine alone increased the frequency of bifunctional cells (Fig. 5H). Overall, these data suggest that combination therapy is enhancing T-cell function in the spleen and TME. This is further supported by NanoString analysis of the TME where entinostat alone or in combination enhanced genes involved in costimulatory and interferon signaling (Fig. 5I and J).

Although N-803 plus vaccine generated more bifunctional CD8⁺ TILs, that did not translate into significant antitumor efficacy. Thus, cytokine production may be not the sole indicator of the ability of CD8⁺ T cells to exert tumor control. The granular exocytosis pathway is one of the major components of the cytolytic ability of a T cell, which utilizes perforin and granzymes A and B to induce cell death (48). Our NanoString analysis of MC38-CEA tumors indicated entinostat alone or in combination increased cytotoxicity genes (Fig. 6A). Specifically, genes for granzymes and perforin were elevated, with granzyme B being increased the most (Fig. 6B). Therefore, we measured granzyme B levels by flow cytometry. Combination therapy significantly elevated granzyme B levels in splenic CD8⁺ and CD4⁺ T cells versus all other therapies (Supplementary Fig. S12A). In the TME, N-803 plus vaccine enhanced the percentage of CD8⁺ TILs containing granzyme B relative to control but not the amount of granzyme B per cell (gMFI; Fig. 6C and D). Notably, combination therapy increased granzyme B levels 2.8-fold versus control (% and gMFI), an effect driven by entinostat (Fig. 6C and D). Entinostat alone or in combination also elevated granzyme B levels in CD4⁺ TILs (% and gMFI; Supplementary Fig. S12B). The promotion of granzyme B in the TME by entinostat alone or in combination correlated with augmented tumor necrosis by hematoxylin and eosin (H&E) staining (Supplementary Fig. S13A–S13C).

CD8⁺ TILs and granzyme B levels were also examined in the TME via immunofluorescence (Fig. 6E–G). Combination therapy increased CD8⁺ TILs 7-fold and total granzyme B puncta 7.9-fold in the TME versus control (Fig. 6F and G). These data are similar to what was determined by flow cytometry (Figs. 4E and 6C). Of note, the quantification of total granzyme B puncta is a summation of granzyme in cytotoxic cells and any that has been recently released. These data are comparable to monitoring granzyme B levels in all CD45⁺ cells by flow cytometry (Supplementary Fig. S12C).

The effect of entinostat and N-803 on granzyme B levels was assessed in healthy donor (HD) peripheral blood mononuclear cells (PBMC) exposed in vitro to entinostat followed by N-803 and then TCR stimulated. Granzyme B levels were elevated in T cells upon stimulation regardless of treatment. Furthermore, exposure to both N-803 and entinostat elevated granzyme B levels in human CD8⁺ and CD4⁺ T cells regardless of stimulation (Fig. 6H; Supplementary Fig. S12D). Overall, combination therapy elevates the cytolytic potential of T cells in the periphery and TME, which was recapitulated in HD PBMCs in vitro.

Combining therapies to alleviate immunosuppression while promoting immune effector function may improve clinical responses to immunotherapy. Here we provide the rationale for combining the HDACi entinostat with the IL15 superagonist N-803 plus a therapeutic cancer vaccine. Although the combination therapy was well tolerated, entinostat-induced lymphopenia was observed. This is concordant with reports of transient, non–dose-limiting leukopenia observed in some patients treated with entinostat (32, 33). Importantly, despite lymphopenia, the combination therapy had significant antitumor effects in both murine models.

Comprehensive studies in the laboratory have demonstrated that N-803 in combination with vaccine has superior antitumor effects relative to either agent alone (K.P. Fabian; manuscript in preparation). Ergo, in this study, we utilized N-803 plus vaccine as the “vaccine component.” Nonetheless, the contribution of the vaccine is evident in the ELISPOT data where the addition of a vaccine to N-803 increases antigen cascade to p15e and neoepitope antigens, an effect not observed with N-803 alone. Antigen cascade has been repeatedly observed both preclinically and clinically with therapeutic cancer vaccines and linked to the generation of durable long-term antitumor activity and superior clinical outcomes (49, 50). Importantly, we also demonstrated that the addition of entinostat to the vaccine/N-803 duet further enhances antigen cascade, which is not observed with entinostat alone. Given the long duration of tumor control in the MC38-CEA model by the combination therapy, one could speculate this is due to the epitope spreading we observed in CD8⁺ T-cell responses, including to neoantigens.

Although we have not identified antigen-specific T cells in the TME, it appears that we are inducing an antigen cascade as evident from the ELISPOT data. In addition, the loss of antitumor activity with the CD8 depletion demonstrates the role of antigen-specific CD8⁺ T cells in the antitumor effect. Interestingly, we observed that depletion of CD4⁺ T cells cured 80% of the combination-treated mice. Others have reported similar findings, with one study demonstrating that CD8⁺ T-cell responses were enhanced in CD4-depleted mice resulting in tumor control (51, 52). Hence, it is possible that depletion of CD4⁺ T cells may have enhanced CD8⁺ T-cell responses elicited by combination therapy. In addition, the inherent deletion of Tregs encompassed by CD4⁺ T-cell depletion could further contribute to CD8-mediated tumor resolution. Although our data indicated that after 17 days of entinostat treatment Tregs were reduced, we have never determined how rapidly this effect occurs. It is feasible that entinostat does not alter Tregs at the initiation of treatment. Therefore, Treg depletion in the early stages of tumor growth may be sufficient to resolve tumors in conjunction with combination therapy. It is unclear if depletion of CD4⁺ T cells later in tumor growth would have produced a different result, especially given that combination therapy increased cytokine production by CD4⁺ T cells in the TME.

Our comprehensive immune analysis in the spleen and TME unveiled several instances where N-803 plus vaccine synergized with entinostat. For example, entinostat increased genes associated with antigen processing and presentation as well as protein and transcript levels of MHC on tumor cells. This would result in increased CD8 recognition of tumor cells. To this end, we observed an increase in antigen-specific T cells only in combination-treated mice. This could be a consequence of entinostat enhancing the expression of transcripts involved in antigen presentation, costimulatory signaling, and interferon, in combination with CD8 activation induced by N-803 and vaccine (Supplementary Fig. S14, A's). Moreover, entinostat alone or in combination promoted effector memory CD8⁺ TILs. Because memory T cells have an enhanced proliferative capacity and better longevity than effector T cells (53), one could speculate that entinostat promoting effector memory formation in the tumor contributes to the antitumor efficacy of the combination therapy.

The superior CD8 tumor infiltration elicited by combination therapy further illustrates the three therapies synergizing. We demonstrated that N-803 plus vaccine increased the expression of CXCR3 on peripheral CD8⁺ T cells, a receptor shown to be required for efficient T-cell tumor infiltration (42, 43). NanoString analysis revealed an entinostat-induced increase in cytokine and chemokine signaling, including augmented transcripts for the CXCR3 ligands, CXCL9, CXCL10, and CXCL11. This is concordant with previous reports showing that HDAC inhibitors, including entinostat, elevate these chemokines (44, 54). The protein levels of these chemokines at day 24 did not follow this trend. Nonetheless, one could speculate that the entinostat-induced increase in CXCL9, CXCL10, and CXCL11, in combination with N-803 plus vaccine-induced elevation in CXCR3⁺ CD8⁺ T cells, would lead to increased CD8 tumor infiltration (Supplementary Fig. S14, B's). Additionally, due to the pleiotropic and promiscuous nature of chemokines and chemokine receptors, it is possible that other chemokines upregulated with combination therapy (CXCL13, CCL2, CCL3, CCL4, and CCL5) may be playing a role given their implication in increasing CD8⁺ TILs (45–47). One study showed there was synergy between CXCR3 chemokines and CCL5 in attracting T cells into melanoma tumors (45). Although we did not investigate the expression of other chemokine receptors, this study suggests that CCL3, CCL4, and CCL5 may be contributing to CD8 tumor infiltration.

A key finding of our studies and a final instance of N-803 and vaccine synergizing with entinostat was the enhanced cytolytic potential of CD8⁺ TILs observed at both the gene and protein levels. For the first time, we demonstrate that entinostat synergizes with N-803 plus vaccine to increase granzyme B levels in CD4⁺ and CD8⁺ T cells in the periphery and TME. This elevation in granzyme B in T cells was recapitulated in human PBMCs after in vitro exposure to N-803 and/or entinostat. Previous research has shown that granzyme B levels are elevated in human PBMCs treated with N-803 (13). Here, we demonstrated that exposure to entinostat and N-803 elevated granzyme B levels in T cells to a greater extent than either agent alone, suggesting a cooperative mechanism that further enhances the cytolytic potential of T cells.

For the first time, we show that a therapy reduces the inhibitory checkpoint VISTA on CD8, CD4, Tregs, and MDSCs in tumor-bearing and non–tumor-bearing mice. Although the precise role of VISTA in antitumor immune responses is still under investigation, VISTA has been shown to negatively affect T-cell function (35, 36). Thus, it is possible that the decrease in VISTA is contributing to the improved antitumor activity observed with combination therapy. Additionally, it has been suggested that inhibiting/reducing multiple checkpoints may improve the antitumor efficacy of checkpoint blockade. Therefore, it is tempting to speculate that the reduction of VISTA via entinostat could, in part, explain why entinostat enhances the antitumor effects of an antiprogrammed cell death-1 (αPD-1) treatment, although VISTA levels were not reported in those studies (17, 18). It has been reported that entinostat reduces the suppressive ability of MDSCs through multiple mechanisms (17, 20). The reduction of VISTA on MDSCs may provide an additional mechanism as to how entinostat reduces the suppressive capacity of MDSCs.

To conclude, the antitumor efficacy of the combination therapy stems from simultaneously targeting multiple aspects of the immune system and the tumor, thereby shifting the balance of immune suppression and activation in the TME toward a more activated and inflamed microenvironment. Combination therapy enhanced tumor infiltration of CD8 TILs with increased cytolytic potential able to efficiently target tumor cells whose phenotype had been altered by entinostat to render them more amenable to recognition and killing (Supplementary Fig. S14, A's and B's). Further, these CD8 TILs would also display increased functionality given the reduction in immune suppression arising from a decrease in tumor Tregs and VISTA levels (Supplementary Fig. S14, C's).

Although combination therapy tripled survival of MC38-CEA tumor–bearing mice and induced various immune effects associated with antitumor efficacy, it did not resolve all the tumors. It is feasible that T-cell exhaustion or immune regulatory elements were induced at a later time point. Alternatively, through our comprehensive studies, we observed that PD-L1 was elevated on both tumor cells and MDSCs after entinostat treatment, suggesting that PD-1/PD-L1 blockade could further enhance tumor control. To this end, ongoing studies have been initiated combining entinostat with M7824, a bifunctional molecule comprising an anti–PD-L1 fused to a TGFβ trap. Importantly, all of the agents used in this study are currently being examined in numerous clinical trials. Our proof-of-concept studies indicate that the clinical combination of entinostat, N-803, and therapeutic cancer vaccines may increase response rates to immunotherapy in patients with solid malignancies.

P. Ordentlich is an employee/paid consultant for and holds ownership interest (including patents) in Syndax Pharmaceuticals. F.R. Jones is an employee/paid consultant for and holds ownership interest (including patents) in Etubics Corporation. S. Rabizadeh is an employee/paid consultant for ImmunityBio, and is an advisory board member/unpaid consultant for NantOmics and NantWorks. P. Soon-Shiong is an employee/paid consultant for and holds ownership interest (including patents) in NantWorks. No potential conflicts of interest were disclosed by the other authors.

Conception and design: K.C. Hicks, K.M. Knudson, K.L. Lee, D.H. Hamilton, W.D. Figg, P. Soon-Shiong, J. Schlom, S.R. Gameiro

Development of methodology: K.C. Hicks, K.M. Knudson, J.W. Hodge, W.D. Figg, P. Soon-Shiong, S.R. Gameiro

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): K.C. Hicks, K.M. Knudson, K.L. Lee, J.W. Hodge, W.D. Figg

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): K.C. Hicks, K.L. Lee, D.H. Hamilton, W.D. Figg, P. Ordentlich, F.R. Jones, S. Rabizadeh, P. Soon-Shiong, J. Schlom, S.R. Gameiro

Writing, review, and/or revision of the manuscript: K.C. Hicks, W.D. Figg, F.R. Jones, S. Rabizadeh, P. Soon-Shiong, J. Schlom, S.R. Gameiro

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): J. Schlom

Study supervision: J. Schlom, S.R. Gameiro

The authors thank Curtis Randolph for excellent technical assistance, Debra Weingarten for her excellent assistance in the preparation of this manuscript, and the NCI Collaborative Protein Technology Resource Laboratory and Clinical Support Laboratory/Frederick National Laboratory for Cancer Research for their excellent technical support. This research was supported by the Intramural Research Program of the Center for Cancer Research, NCI, NIH, and by Cooperative Research and Development Agreements (CRADA) between the NCI and Syndax Pharmaceuticals, Inc., and the NCI and NantBioScience, Inc.

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.