A Multikinase and DNA-PK Inhibitor Combination Immunomodulates Melanomas, Suppresses Tumor Progression, and Enhances Immunotherapies

New treatment options have improved prognoses for melanoma patients over the last half decade. These improvements stem from the development of two treatments: targeted therapies and immunotherapies. However, both classes pose unique challenges that limit their efficacy and application.

Targeted therapies for melanoma include BRAF and MEK inhibitors. Currently, targeted therapies are only available for the roughly 50% of melanoma patients who harbor BRAFV600 mutations. Of these patients, only 50% and 20% respond to BRAF or MEK inhibitors, respectively, and responses are typically short-lived (1–3). Indeed, acquired resistance develops rapidly in around half of patients treated with BRAF inhibitors with a median time of 6 to 8 months (1, 2). BRAF and MEK inhibitor combinations improve response rates and durations, though only marginally (4, 5). Moreover, targeted therapies are not available for patients without BRAF mutations. Thus, there is a need to identify new targeted therapies in order to expand application to a larger patient population and prolong response duration.

Immunotherapies, particularly antibodies blocking checkpoint receptors such as cytotoxic T-lymphocyte antigen-4 (CTLA-4) and programmed death-1 (PD-1), have revolutionized melanoma treatment. Durable and often complete tumor regressions in the setting of widespread metastatic disease have been observed. Yet, objective responses to PD-1 antibodies only occur in roughly one third of patients and complete regressions occur in approximately 5% of patients (6, 7). Thus, strategies that increase response rates are necessary to improve the overall efficacy of these therapies.

Efforts have been made to combine targeted therapies and immunotherapies. Toxicities such as hepatotoxicity and gastrointestinal complications, however, have led to the termination of such clinical trials (8, 9). Trials targeting PD-1 or its primary ligand, programmed death-ligand 1 (PD-L1), in combination with BRAF and/or MEK inhibitors, appear more promising, but are still in progress (10). Other trials have shifted to establishing an optimal sequential application of the two treatments, which may not be as effective as simultaneous administration.

We sought to identify and characterize targeted therapies that would enhance immunotherapy efficacy. To achieve this goal, we performed flow cytometry-based high-throughput screening (HTS) on more than 1,500 small molecule inhibitors, chemotherapies, medications, and natural compounds, examining their ability to alter cell surface molecules associated with immunosuppressive or immunostimulatory processes. Many of the screened compounds are already FDA approved for various diseases. Using HTS, we identified two compounds, regorafenib (Reg) and NU7441 (NU), for further characterization.

Reg is a multikinase inhibitor that targets various tyrosine kinases, including VEGFR, c-KIT, PDGFR, and others (11). It is FDA approved for pretreated colorectal carcinomas, gastrointestinal stromal tumors (GIST), and hepatocellular carcinomas. NU is a small molecule DNA-PK inhibitor (12). Reg and NU altered a variety of immunogenic and oncogenic properties of melanomas while favorably influencing T cells. In melanoma animal models, Reg decreased tumor growth alone and in combination with immunotherapies.

A375*, C32*, G-361*, MNT-1^‡, SK-MEL-5*, SK-MEL-24*, SK-MEL-28*, and B16-F1* were maintained in DMEM with 10% FBS (12306C, Sigma Aldrich) and 1% antibiotics (15140122, Gibco). Malme-3M* were maintained in IMDM with 20% FBS and 1% antibiotics. All remaining cell lines (624-Mel^‡, COLO-829^†, LOX-IMVI^†, RPMI-7951*, UACC-62^†, UACC-257^†, WM-115^†, WM-266-4^†, C8161^‡, Mel-Juso^†, MM415^†, MM485^†, SK-MEL-2*, SK-MEL-30^†, SK-MEL-103^†, SK-MEL-119^†, SK-MEL-147^†, SK-MEL-173^†, and UCD-Mel-N^†) were maintained in RPMI-1940 with 10% FBS and 1% antibiotics. Cell lines were obtained from ATCC (*) or generously provided by either Dr. Thomas Hornyak (†; University of Maryland, Baltimore) or Dr. Suzanne Ostrand-Rosenberg (‡; University of Maryland, Baltimore County) between 2009 and 2014. After thawing, cells were cultured for no longer than 6 weeks and tested for mycoplasma monthly. Cell lines were authenticated using quantitative PCR (qPCR) for melanoma transcripts (gp100, MART-1, and TYRP1). Human PBMCs were purified by Ficoll gradient (17-1440-02, GE Healthcare) from buffy coats purchased from New York Blood Center. Purified PBMCs were cryopreserved, thawed, and cultured in AIM-V media (Gibco, 12055091) with 5% human serum (H4522, Sigma Aldrich), 1% nonessential amino acids (11140050, Gibco), and 1% antibiotics.

Antibodies from BioLegend for the following human molecules were used: PD-L1 (329718), CD155 (337610), HLA-A/B/C (311404), CD73 (344012), CD271 (345112), CD55 (311312), CD4 (317408), CD8 (300914), IFNγ (502509), TNFα (502912), and IL2 (500324). Antibodies from BioLegend for the following mouse molecules were used: CD45.2 (109828), CD8 (100725), CD4 (100406), Lag3 (125128), Tim3 (119718), PD-1 (135216), CD25 (102012), Foxp3 (126404), CD11b (101206), CD11c (117308), F4/80 (123124), Gr1 (108412), NK1.1 (108728), PD-L1 (124314), CD73 (127212), and CD155 (131508). Anti-H2-D^b was from Abcam (ab112492). Anti-mouse IFNγ (12-7311-81, eBioscience), IL2 (560544, BD Biosciences), and TNFα (560658, BD Biosciences) were used for T-cell cytokine experiments. LEGENDScreen human and mouse cell screening kits (700001 and 700005) from BioLegend were used for flow cytometry protein arrays.

The SC200 drug library was generously provided by Dr. Paul Wilder (University of Maryland, Baltimore). The L1100 drug library was from SelleckChem. Reg (BAY73-4506, S1178, SelleckChem) was dissolved in dimethyl sulfoxide (DMSO, 0231, Quality Biological) or 5% Kolliphor EL (C5135, Sigma Aldrich) for in vitro and in vivo studies, respectively. NU7441 (S2638, SelleckChem) and NU7026 (S2893, SelleckChem) were dissolved in DMSO or 10% DMSO for in vitro and in vivo studies, respectively.

For HTS, C8161 cells were treated with the indicated compounds for 48 hours. Treated cells were analyzed by flow cytometry for expression of indicated molecules. Viable cells were gated using a fixable viability dye (423101, BioLegend) or using light scatter. For IFNγ experiments, cells were pretreated for 24 hours with 20 U/mL human recombinant IFNγ (14-8311-63, eBioscience), and IFNγ was maintained in the media throughout the experiment. For assessing drug effects on T-cell phenotype, PBMCs were stimulated with 20 ng/mL anti-CD3 (OKT3, 16-0037-85, eBioscience), 50 U/mL human recombinant IL2 (589106, BioLegend), and drug for five days and analyzed using CD4 and CD8 antibodies to distinguish between T-cell subsets. For T-cell cytokines, PBMCs were stimulated with anti-CD3 (100 ng/mL) for 72 hours. Some cells were restimulated with phorbol myristate acetate (PMA, 10 ng/mL) and ionomycin (0.5 μg/mL) for 6 hours. During the final 6 hours, cells were treated with Brefeldin A (GolgiPlug, 555029, BD Biosciences). A cell fixation and permeabilization kit (554714, BD Biosciences) was used.

Molecule expression was measured in cells treated with six concentrations of Reg, NU, or the combination and MFIs were compared with vehicle-treated cells to calculate fold change. Using the Chou-Talalay method, combination index (CI) values were calculated with CompuSyn software (ComboSyn, Inc). Synergy was defined as at least four of six concentrations yielding CI values below one, additive interactions as at least four CI values within 0.5 to 1.5, and antagonism as at least four of six CI values above two. Cell lines that met none of these classifications were defined as not determined. The fractionated product analysis method was used to calculate in vivo synergy. A ratio greater than one was considered synergistic, equal to one as additive, and less than one as antagonistic.

RNA was collected from cells treated with Reg (2 μmol/L) and/or NU (1 μmol/L) for 48 hours using Qiagen RNeasy Mini kits (74104) following the manufacturer's protocol. cDNA was prepared using a high-capacity reverse transcription kit (4368814, Applied Biosystems). qPCR was performed with iTaq Universal SYBR Green Supermix (1725121, Bio-Rad) and primers for gp100 (Fwd: CTGCCTCAATGTGTCTCTGGCT, Rev: CAAGGACCACAGCCATCAACAC), MART-1 (Fwd: GGACAGCAAAGTGTCTCTTCAAG, Rev: TCAGGTGTCTCGCTGGCTCTTA), TYRP1 (Fwd: TCTCAATGGCGAGTGGTCTGTG, Rev: CCTGTGGTTCAGGAAGACGTTG), and beta-actin (Fwd: CACCATTGGCAATGAGCGGTTC, Rev: AGGTCTTTGCGGATGTCCACGT). Relative fold changes were calculated using the ΔΔCt method normalizing to beta-actin.

Melanoma cell lines were treated with varying concentrations of Reg, NU, or vemurafenib for 48 hours. PBMCs were cultured with 20–50 ng/mL anti-CD3 and 50–100 U/mL human recombinant IL2 for five days. All cells were cultured with ³H-thymidine (0.1 μCi/ml, NET027E005MC, Perkin Elmer) for the final 16 hours of drug treatment to assess thymidine incorporation.

Cells were treated with varying concentrations of Reg and NU for 24 to 48 hours and lysed with RIPA buffer (R0278, Sigma Aldrich) containing protease and phosphatase inhibitors (78440, Thermo Scientific). Phospho-MEK1/2 (Clone 41G9, 9154S), MEK1/2 (Clone 47E6, 9126S), phospho-Akt (Clone D9E, 4060S), Akt (Clone 11E7, 4685S), β-Actin (Clone D6A8, 8457S), and GAPDH (Clone D16H11, 5174S) antibodies from Cell Signaling Technology were used. gp100 (ab137078) was from Abcam. Polyclonal Tyrp1 antibody was generously provided by Dr. Thomas Hornyak (University of Maryland, Baltimore). Densitometry was performed using ImageJ.

Studies were approved by the UMB Institutional Animal Care and Use Committee. C57BL/6J and pmel (B6.Cg-Thy1/Cy Tg(TcraTcrb)8Rest/J) mice were purchased from The Jackson Laboratory. Animal experiments contained 5 to 7 animals per group for tumor growth and survival with 3 to 5 for ex vivo analyses. Six- to 8-week-old C57BL/6J mice were injected with 2 × 10⁵ B16-F1 melanoma cells subcutaneously on the right flank. Tumors were allowed to establish and reached approximately 50 to 100 mm² before treatment. Animal weights and tumor sizes were monitored every 2 to 3 days. Tumor volumes were calculated using the following formula: volume = (height × width²)/2. For studies without immunotherapies, mice were treated with Reg (8 mg/kg) and/or NU7026 (6 mg/kg) on the following days post-tumor inoculation: 9–10, 12–15, 17–20, 22–25, 27–30, and 32–35. Drugs were delivered intraperitoneally (i.p.) once per day. For studies using anti-CD40 and c-di-GMP, mice were treated with Reg (4–5 mg/kg) on days 8–9, 11–14, 16–19, 21–24, 26–29, 31–34, and 36–39. Some mice received 100 μg of anti-CD40 (clone FGK4.5, BE0016-2, BioXCell) i.p. and 50 μg c-di-GMP (tlrl-nacdg, InvivoGen) peritumorally on days 10, 15, 20, and 25. For CD8⁺ T-cell depletion studies, mice were treated with 400 μg anti-CD8 (clone 53-6.7, BE0004-1, BioXCell) 1 day prior to tumor inoculation. Subsequently, mice were treated twice per week with 100 μg anti-CD8. For adoptive transfer experiments, mice were treated with Reg (5 mg/kg) on days 8–12 post-tumor inoculation, irradiated with 650 radians of γ-irradiation on day 12, and received intravenous delivery of 3 × 10⁶ pmel T cells (from pmel splenocytes and activated 6 days prior with hgp100_25-33) on day 13. Adoptive transfer included i.p. injections of 200 μg anti–4-1BB (clone 3H3, BE0239, BioXCell) on days 14 and 19. For all ex vivo experiments, tumors were processed through 100 μm filters, cells were stained for flow cytometry, and fixed in 4% paraformaldehyde (19943, Affymetrix). Regulatory T cells were assessed using a Foxp3/transcription factor staining buffer set (00-5523-00, eBioscience) following the manufacturer's protocol. For ex vivo cytokines, intratumoral leukocytes were isolated by Ficoll gradient and activated with PMA (20 ng/mL) and ionomycin (1 μg/mL) for 6 hours.

For all flow cytometry–based experiments, data were analyzed using FlowJo software (FlowJo, LLC). Viable populations were assessed using a fixable viability dye and/or using light scatter. Fold changes were determined using median fluorescent intensities (MFI). Specifically, fold change = (experimental MFI − isotype MFI)/(vehicle MFI − isotype MFI). For all molecules, MFI values below 200 in vehicle-treated cells were considered not expressed. For ex vivo analyses normalized to tumor cell number, tumor cells were defined as CD45⁻ cells. Mann–Whitney U tests were used to assess differences in responses between cells with different mutations. Kruskal–Wallis tests were used to evaluate overall differences in cell lines treated with different drugs. If P values for Kruskal–Wallis tests were less than 0.05, Mann–Whitney U tests were used to compare two drugs at a time. Paired t tests were used to examine differences in T-cell markers and cytokines between vehicle- and drug-treated T cells. One-way ANOVA with Tukey posttests were used to determine differences in tumor growth and cell populations. Unpaired t tests were used to determine differences in T-cell levels and markers between ACT and ACT+Reg groups. Significant differences in survival were assessed using the log-rank test. P values below 0.05 were considered statistically significant.

To identify compounds that have immunomodulatory effects on melanomas, we examined the ability for compounds to alter PD-L1 and human leukocyte antigen class I (HLA-I) expression in C8161, a human melanoma cell line that lacks BRAFV600 or NRAS mutations. PD-L1 was selected because therapies directed at the PD-1-PD-L1 axis have generated clinical success. HLA-I is essential for target recognition by T cells and a central component of antitumor immunity. Two large drug libraries, SC200 and L1100, were evaluated. In the SC200 library, only 20 of the 1,120 FDA-approved drugs, chemicals, and natural compounds reduced PD-L1 by at least 10% (Supplementary Fig. S1). Of these, 17 were glucocorticoid or anti-inflammatory drugs that are unfit for immunotherapeutic application. Thus, compounds from the SC200 library were excluded from further studies. However, 38 compounds decreased PD-L1 by at least 20% and 44 compounds increased HLA-I by at least 50% in the 484-compound L1100 library, and 44 candidates were selected for further evaluation (Fig. 1A, representative examples in Fig. 1B and C).

The candidate inhibitors targeted proteins that act in the MAPK, PI3K/Akt/mTOR, and Src oncogenic signaling pathways, suggesting that these pathways may contribute to the regulation of both PD-L1 and HLA-I. To verify that the candidates were active in different melanomas, their effects on PD-L1 and HLA-I expression were tested in several BRAF- and NRAS-mutant melanomas (Supplementary Figs. S2 and S3). In parallel, the effects of the candidate therapeutics on T-cell proliferation were examined (Supplementary Figs. S2 and S3, right). EGFR inhibitors reduced PD-L1 in only a subset of cell lines and did not alter HLA-I. VEGFR inhibitors decreased PD-L1 by less than 25% and increased HLA-I. p38 inhibitors had no effect. MEK, PI3K, and Src inhibitors, which often reduced PD-L1 and increased HLA-I on melanomas, also potently inhibited T-cell proliferation. A selective BRAFV600 inhibitor, PLX-4720, did not alter PD-L1 or HLA-I. Therefore, despite positive results from HTS, these compounds were not studied further.

Two compounds, Reg and NU7441, altered PD-L1 and HLA-I and were chosen for additional evaluation. Reg is a multikinase inhibitor with antiangiogenic properties due to its inhibition of VEGFR and TIE2. However, Reg also blocks the activity of RET, PDGFR, BRAF, and c-KIT, many of which are activated in melanomas (13–16). Reg decreased PD-L1 and induced HLA-I in C8161 (Fig. 1D). NU7441 (NU), a DNA-PK inhibitor that reduced PD-L1 and increased HLA-I expression in a dose-dependent manner, was also selected (Fig. 1E). Reg and NU combination treatment further reduced PD-L1 and increased HLA-I at certain concentrations (Fig. 1F and G). Therefore, inhibition of commonly activated kinases and DNA-PK by Reg and NU, respectively, altered PD-L1 and HLA-I at a range of concentrations.

To characterize the effects that Reg and NU had on the “immunophenotype” of melanomas, a flow cytometry–based array was used to examine expression of 332 cell surface proteins on four melanoma cell lines treated with combination Reg (2 μmol/L) and NU (1 μmol/L; Supplementary Fig. S4A). Sixty-four cell surface markers were expressed by at least three of four cell lines. Of these, several molecules involved in complement regulation, including CD46, CD55, and CD97 were downregulated by Reg and NU (Supplementary Fig. S4B). Other molecules involved in immunosuppressive processes, including CD73 (ecto-5′-nucleotidase, ecto-5′-NT), CD155 (poliovirus receptor, PVR), and CD271 (Nerve Growth Factor Receptor, NGFR) were also downregulated by drug treatment (Supplementary Fig. S4C).

The immunophenotypic effects of Reg (4 μmol/L) and NU (2 μmol/L) were evaluated in a panel of heterogeneous melanoma cell lines (Supplementary Table S1). Twenty-seven melanoma lines were examined including 16 harboring BRAFV600 mutations, 10 with NRAS mutations, and one BRAF and NRAS wild-type line. Reg and NU concentrations used correlated to physiologically achievable plasma concentrations (17–19). This approach allowed for analysis of molecule expression levels and changes induced by Reg and NU in melanomas with different mutations. The effects of a BRAFV600 inhibitor, vemurafenib (Vem, 1 μmol/L), were also examined. BRAF inhibitors have been reported to increase melanoma antigen expression, decrease immunosuppressive cytokine secretion, and increase intratumoral T-cell infiltration in both preclinical and clinical studies (20–23). Therefore, the effects of Vem were compared with Reg and NU. A reduction was defined as a decrease in expression of at least 30% compared with vehicle-treated cells unless otherwise specified. Statistical analyses comparing responses in cell lines with different mutations and responses to different drugs are shown in Supplementary Tables S2–S8.

PD-L1 expression was measured in melanoma cell lines treated with Reg, NU, or Vem. The regulation of PD-L1 is not well characterized, and clinically, few melanomas express consistently high PD-L1 (24). All 27 cell lines expressed PD-L1, with 44.4% (12/27) expressing low PD-L1 (Supplementary Fig. S5A). Reg treatment reduced PD-L1 in 37% (10/27) of the cell lines (Fig. 2A, representative histograms in Supplementary Fig. S5B). NU reduced PD-L1 in 74.1% (20/27) of the lines. Combination Reg plus NU reduced PD-L1 in all but one of the melanomas (96.3%). Moreover, reductions in PD-L1 were augmented with combination treatment (Supplementary Fig. S5C). Responses to Reg and NU treatment were similar between BRAF- and NRAS-mutant cell lines. PD-L1 reductions were greater in NRAS-mutant compared with BRAF-mutant lines, though not significantly (P = 0.07). In line with previous studies, BRAFV600 inhibition with Vem did not consistently affect PD-L1 expression in melanomas (25). BRAF inhibitors can paradoxically activate MAPK signaling in BRAF wild-type cells (26). Accordingly, we observed an upregulation of PD-L1 in Vem-treated NRAS-mutant cell lines. Similar trends were observed for all of the examined molecules. Finally, Reg and NU synergistically or additively reduced PD-L1 in 59.3% (16/27) of cell lines, suggesting synergistic modulation of PD-L1 by the two therapies is common in melanomas (Fig. 2B; examples of synergistic, additive, and antagonistic effects in Supplementary Fig. S6).

CD155 is one of two major ligands for the checkpoint receptor TIGIT. Like PD-1, TIGIT has been associated with T-cell exhaustion—a phenomenon in which T cells progressively lose functional capacity and eventually undergo cell death (27). CD155 was expressed at high levels in all 27 melanoma lines (Supplementary Fig. S7A). Reg reduced CD155 in roughly half of the melanomas (48.1%, 13/27), whereas NU or combination treatment reduced expression in most melanomas (81.5%, 22/27 and 85.2%, 23/27, respectively; Fig. 2C, Supplementary Fig. S7B). Despite high constitutive expression, combination treatment led to greater than 70% reductions in 14.8% (4/27) of the lines. Vem reduced CD155 in only 31.3% (5/16) of BRAF-mutant melanomas, but decreases of at least 10% occurred in 87.5% (14/16) of lines. Therefore, although factors regulating CD155 in melanoma are largely unknown, these data suggest that MAPK signaling may regulate CD155 to a limited degree. Like PD-L1, Reg and NU additively or synergistically reduced CD155 in a majority of melanomas (59.3%, 16/27), including similar proportions of BRAF- and NRAS-mutant lines (Fig. 2D).

HLA-I is crucial for target recognition by CD8⁺ T cells. HLA-I loss has been observed in a variety of malignancies, although complete loss appears rare in melanoma (28). Only one of 27 lines exhibited undetectable levels of HLA-I, and most remaining lines expressed medium or high levels (Supplementary Fig. S8A). Reg increased HLA-I expression by at least 50% in 18.5% (5/27) of melanomas, whereas NU did not increase HLA-I (Fig. 2E, Supplementary Fig. S8B). MAPK inhibition induces HLA-I expression (29). We also observed increased HLA-I expression following Vem treatment, but upregulation of at least 50% was uncommon, occurring in 25% (4/16) of BRAF-mutant lines. Taken together, these data suggest that Reg and NU reduce ligands of immunosuppressive checkpoint receptors while maintaining or increasing HLA-I, and collectively may enhance antitumor T-cell immunity.

PD-L1, CD155, and HLA-I alter TCR signaling or costimulation in T cells. Beyond these molecules, we also focused on others that modulate antimelanoma immunity. CD73 catalyzes the conversion of extracellular AMP to adenosine. Adenosine suppresses immune cells, including various effector T cells, and enhances regulatory T-cell (Treg) function (30). CD73 is often overexpressed on tumors, including melanoma (30). CD73 was expressed at relatively low levels in a subset of melanoma cell lines, including 62.5% (10/16) of BRAF-mutant and all NRAS-mutant cell lines (Supplementary Fig. S9A). CD73 was downregulated in 66.7% (14/21) of lines treated with Reg or NU monotherapies (Fig. 2F, Supplementary Fig. S9B). Combination treatment was particularly effective where CD73 was reduced by greater than 50% in 81% (17/21) of lines. Vem downregulated CD73 in 40% (4/10) of the CD73-expressing BRAF-mutant lines. Finally, synergistic or additive combinatorial effects were observed in most lines (71.4%, 15/21) without evidence of antagonism (Fig. 2G).

NGFR is expressed on a variety of cell types and is often upregulated in cancer (31). In melanoma, NGFR expression has been associated with properties of cancer stem cells and immunosuppressive functions (32–35). NGFR expression was detected in 59.3% (16/27) of melanomas (Supplementary Fig. S10A). A higher proportion of NRAS-mutant cell lines expressed NGFR compared with BRAF-mutant lines (80% versus 50%). NGFR expression varied widely in melanomas. Although Reg or NU typically reduced NGFR, Reg also induced NGFR by more than 20% in 25% (4/16) of the cell lines (Fig. 2H, Supplementary Fig. S10B). NGFR was reduced in a synergistic manner by Reg and NU in only 31.3% (5/16) of cell lines (Fig. 2I). Thus, NGFR expression and regulation in melanoma is heterogeneous and complex, but NGFR can often be downregulated by Reg and NU treatment.

CD55, also known as decay accelerating factor (DAF), is expressed on most cells to prevent activation of the complement system and is frequently overexpressed on cancer cells (36). The functions of humoral immunity, cancer-specific antibodies, and the complement system remain unclear in cancer, but evidence for their roles in tumor immunology has increased (36–38). Detectable, but typically low CD55 expression was observed in 92.6% (25/27) of lines (Supplementary Fig. S11A). Reg and NU treatment consistently decreased CD55 expression and did so with equal effectiveness (Fig. 2J; Supplementary Fig. S11B). Vem also reduced CD55 expression in one third (5/15) of the BRAF-mutant melanomas suggesting a role for the MAPK pathway in CD55 maintenance. Combination therapy resulted in synergistic or additive reduction of CD55 in over half of the melanomas examined (56%, 14/25; Fig. 2K). Thus, Reg and NU can regulate a variety of molecules with varying immunomodulatory functions.

IFNγ induces PD-L1 and HLA-I expression. Indeed, PD-L1 expression in tumors is often observed at the interface of IFNγ-producing T cells (39). We confirmed that PD-L1 and HLA-I were upregulated by IFNγ treatment (Fig. 3A and B; Supplementary Figs. S5A and S8A). The magnitudes of PD-L1 reduction or HLA-I induction by Reg, NU, or combination treatment were often enhanced in the setting of IFNγ pretreatment. Thus, Reg and NU suppress PD-L1 and increase HLA-I in diverse melanomas and in the setting of IFNγ induction.

Studies with BRAF and MEK inhibitors have shown that MAPK pathway inhibition can upregulate various melanoma antigens (20, 22, 23). We examined three melanoma antigens—gp100, MART-1, and TYRP1—in melanoma cell lines treated with Reg (2 μmol/L) and/or NU (1 μmol/L) by qPCR (Fig. 3C). All three melanoma antigens were upregulated by Reg and combination treatment. Increased protein levels were confirmed for gp100 and Tyrp1 (Supplementary Fig. S12). Antigens were upregulated in both BRAF- and NRAS-mutant cell lines.

We also measured melanoma proliferation after Reg, NU, or Vem treatment (Supplementary Fig. S13). Reg and NU suppressed melanoma proliferation with average 50% inhibitory concentrations around 3 μmol/L. Vem reduced proliferation at lower concentrations in BRAF-mutant lines without affecting NRAS-mutant lines.

To understand the mechanisms underlying these findings, we examined activation of oncogenic signaling pathways in drug-treated melanoma cell lines (Supplementary Fig. S14). MEK1/2 phosphorylation was decreased in all cell lines, and Akt phosphorylation was reduced in 63.3% (7/11) of melanomas assessed. Phosphorylation was decreased in both BRAF- and NRAS-mutant lines. Therefore, Reg and NU can induce antigen expression, inhibit proliferation, and reduce oncogenic signaling pathways in heterogeneous melanomas.

We sought to identify therapies that would enhance the efficacy of cancer immunotherapies. Consequently, the effects of Reg and NU on T-cell proliferation, phenotype, and function were assessed. Activation of T cells from peripheral blood mononuclear cells (PBMC) with CD3 stimulation in the presence of Reg and/or NU reduced T-cell proliferation by a maximum of 25% (Supplementary Fig. S15). In contrast, antiproliferative effects were observed when T cells were treated with a variety of other targeted therapies (Supplementary Figs. S2 and S3).

We also characterized the phenotypic impact Reg and NU had on T cells (Fig. 4A–C; Supplementary Fig. S16). Although CD44 was decreased, several other T-cell activation or costimulatory markers including CD25, CD28, and inducible T-cell costimulator (ICOS) were upregulated in CD4⁺ and CD8⁺ T cells (Fig. 4A and B). Additionally, coinhibitory checkpoint receptors, including CD244 (2B4), PD-1, and Tim-3 were reduced on CD4⁺ and/or CD8⁺ T cells (Fig. 4C).

As a surrogate for T-cell function, production of IFNγ, TNFα, and IL2 was examined. Cytokine production was increased when T cells were treated with combination Reg and NU after CD3 stimulation with or without PMA and ionomycin reactivation (Fig. 4D; Supplementary Figs. S17 and S18). Collectively, these studies suggest that Reg and NU may be suitable for enhancing T-cell cancer immunotherapies.

In vitro studies cannot recapitulate the complex and diverse immunosuppressive mechanisms of the tumor microenvironment. Consequently, we examined the efficacy of Reg and NU in the B16 melanoma model with and without various immunotherapies. In vitro, B16 cells treated with Reg and/or NU expressed decreased PD-L1, without changes to CD155 or MHC-I (Supplementary Fig. S19). Mice with established B16 melanomas were treated with Reg, NU7026, or the combination. In vivo Reg doses were lower than those used in murine studies and in humans (11, 18). NU7026 is a DNA-PK inhibitor related to NU7441 that was used for animal trials due to improved solubility. No evidence of toxicity was observed throughout animal experiments.

Daily Reg treatment delayed tumor progression and prolonged survival (Fig. 5A and B). NU7026 had no effect alone and did not enhance the efficacy of Reg. Responses in Reg- or combination-treated mice were associated with trends toward increased intratumoral CD45⁺ leukocytes and CD8⁺ T cells (Fig. 5C and D). CD8⁺ T cells were increased within the CD45⁺ population in animals treated with combination therapy (Fig. 5E). No changes in the levels of intratumoral dendritic cells, macrophages, myeloid-derived suppressor cells, or natural killer cells were observed (Supplementary Fig. S20). B16 tumor cells from treated mice expressed lower CD155 and trends toward increased MHC-I (Supplementary Fig. S21). PD-L1 was increased on B16 cells but decreased on leukocytes in treated mice. However, a variety of potentially immunosuppressive molecules, in vitro and/or in vivo, were altered on B16 (Supplemental Fig. S22). Specifically, semaphorin 4D, CEACAM1, and galectin-9 have immunosuppressive functions and were reduced in Reg-treated B16 cells (39–41). Thus, Reg alters B16 phenotype, suppresses tumor progression, and increases CD8⁺ T-cell accumulation in the tumor microenvironment.

We next studied how low doses of Reg (4–5 mg/kg) affected the efficacy of two immunotherapeutic regimens. First, agonistic anti-CD40 was used in combination with c-di-GMP, a stimulator of interferon genes (STING) agonist. These therapies promote myeloid maturation and antigen presenting cell function and have shown promise in preclinical studies and clinical trials (42–44). We predicted that enhancing antigen presentation would further augment intratumoral CD8⁺ T-cell populations as seen with Reg alone. Anti-CD40 with STING agonist treatment resulted in tumor growth inhibition (Fig. 6A). Reg combined with anti-CD40 and STING agonist additively reduced tumor growth and prolonged animal survival (Fig. 6A and B; Supplementary Table S9). Combination therapy also increased intratumoral CD45⁺ leukocytes, CD4⁺ T cells, and CD8⁺ T cells, but did not increase CD4⁺ T cells within the CD45⁺ population (Fig. 6C–G). Thus, promoting antigen presentation in mice treated with Reg can reduce tumor progression.

In order to determine whether Reg efficacy was dependent on increases in CD8⁺ T cells, studies examining the function of intratumoral T cells were performed alongside CD8⁺ T-cell depletion experiments. IFNγ, TNFα, and IL2 production were increased in CD4⁺ and CD8⁺ T cells of treated mice (Fig. 6H and I; Supplementary Fig. S23). Additionally, CD8⁺ T-cell depletion resulted in near complete loss of efficacy of Reg alone and when administered with anti-CD40 and STING agonist (Fig. 6J and K). Thus, Reg treatment improves CD8⁺ T-cell function and these cells are necessary for therapeutic efficacy.

Finally, tumor-bearing mice were treated with Reg in combination with adoptive cell therapy (ACT) using tumor-specific T cells (Fig. 7). Pmel T cells, which recognize the gp100_25–33 antigen expressed on B16 cells, were administered in mice treated with a 5-day course of Reg. Animals receiving both ACT and Reg exhibited synergistically reduced tumor growth and prolonged survival compared with monotherapies (Fig. 7A and B, Supplementary Table S10). These responses were associated with decreased PD-L1 and CD73 expression on intratumoral leukocytes, and decreased CD155 on both leukocytes and tumor cells (Supplementary Fig. S24). No changes were detected in PD-L1 or MHC-I expression on tumor cells. Similarly to previous findings, the CD45⁺ population was increased in mice treated with ACT plus Reg compared with mice treated with ACT alone (Fig. 7C). Additionally, antitumor responses were associated with increased intratumoral CD4⁺ and CD8⁺ T cell and reduced T_reg densities (Fig. 7D–H). Reductions in exhaustion marker (Lag-3, Tim3, and PD-1) coexpression on both CD4⁺ and CD8⁺ T cells were also observed (Fig. 7I and J).

Therefore, Reg cooperates with various immunotherapies to suppress melanoma progression and alter T cells in a manner associated with greater antitumor potential. These data serve as a proof-of-concept for using Reg as an immunotherapeutic targeted inhibitor along with additional established immunotherapies.

Despite considerable progress, immunotherapeutics only benefit a small subset of melanoma patients. Attempts to address this by combining checkpoint blockade with targeted therapies have thus far proven unsuccessful. One major limitation is that trials have focused on BRAF inhibitors and checkpoint blockade. There remains potential for the identification and development of new combination therapies that rely on the ability for the immune system to eliminate melanoma. The findings herein demonstrate a translational approach to repurpose targeted therapies for use in combination with immunotherapies for melanoma.

Although our primary goal was to identify lead compounds and study their ability to alter a variety of molecules, our drug screening approach also allowed for the examination of PD-L1 regulation. PI3K, VEGFR, MEK, and Src inhibitors led to decreases in PD-L1 expression in a variety of cell lines, suggesting that many pathways may contribute to PD-L1 regulation.

The approach also delineated expression levels of several immunomodulatory molecules in a panel of melanomas. Expression of PD-L1, CD73, NGFR, and CD55 varied, and detectable expression of CD73 and NGFR was more frequently associated with NRAS mutations. These data contribute to gaining a more thorough understanding of the immunobiology of melanoma, and may become more valuable as therapies targeting these molecules are developed.

Animal models demonstrated that Reg is capable of suppressing melanoma growth alone and in combination with emerging immunotherapies. Although Reg reduced B16 PD-L1 expression in vitro, PD-L1 was increased in vivo. This may be a reflection of increased IFNγ derived from infiltrating lymphocytes. Despite this, Reg may still function by immunomodulating B16 tumors, as several potentially immunosuppressive molecules were decreased by Reg treatment. Low doses of Reg also enhanced the efficacy of immunotherapies. As a therapy already approved for the treatment of other cancers, future clinical trials using Reg could be accelerated. It remains difficult to predict whether immunotherapy combinations with Reg would be tolerable. Nevertheless, no evidence of toxicity was observed in animal trials, and Reg is well tolerated in humans (45, 46). Moreover, dosing for combination with immunotherapies could be lower compared with standard Reg dosing. Finally, Reg has potential for further development in melanoma patients without BRAF mutations for which targeted therapies are unavailable. Thus, Reg is a candidate for further studies in melanoma with and without additional immunotherapies.

DNA-PK inhibitors failed to elicit changes in animal studies despite promising in vitro findings. One reason for these findings may be related to limited solubility and poor pharmacokinetic properties of DNA-PK inhibitors. NU7441 is poorly absorbed and rapidly metabolized in mice, and these factors have hampered the clinical application of DNA-PK inhibitors (19, 47). Yet targeting DNA-PK remains a promising approach to immunosensitize melanomas. Orally available DNA-PK inhibitors with improved pharmacokinetics are currently in development and clinical trials (48, 49).

Combination therapies can improve treatment efficacy in melanoma. Our findings demonstrate that repurposing existing therapies, including Reg, in combination with T cell-based immunotherapies, is a viable strategy for controlling melanoma progression.

No potential conflicts of interest were disclosed.

Conception and design: A.K. Tsai, E. Davila

Development of methodology: A.K. Tsai

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): A.K. Tsai, C.E. Worgo, L.L. Wang

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): A.K. Tsai, A.Y. Khan, C.E. Worgo, Y. Liang

Writing, review, and/or revision of the manuscript: A.K. Tsai, L.L. Wang, Y. Liang, E. Davila

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): A.K. Tsai, L.L. Wang

Study supervision: A.K. Tsai, E. Davila

This research was supported by the University of Maryland Marlene and Stewart Greenebaum Comprehensive Cancer Center, Melanoma Research Foundation, National Cancer Institute (R01CA140917 and T32CA154274), and VA Merit Award BX002142.

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.