YAP-Induced PD-L1 Expression Drives Immune Evasion in BRAFi-Resistant Melanoma

Blockade of the immune checkpoint receptor PD-1 by monoclonal antibodies (mAbs) leads to remarkable clinical responses in advanced melanoma (1, 2). The interaction of PD-1 with PD-L1 relays inhibitory signals to T cells, resulting in T-cell exhaustion (3, 4). Aberrant expression of PD-L1 on the surface of cancer cells is a central mechanism by which many cancers escape the antitumor immune response (5, 6), and PD-L1 expression serves as a predictive biomarker of a favorable response to immune checkpoint blockade (7, 8). Although previous studies have reported increased expression of PD-L1 in 13% to 72% of tumors in melanoma, gastric cancer, and non–small cell lung cancer patients (9–11), PD-L1 expression patterns are spatially heterogeneous and temporally dynamic (11, 12). PD-L1 expression can be induced adaptively by continuous exposure to interferon-γ (IFNγ), which is secreted by tumor-infiltrating T cells (13), and is also driven intrinsically by the activation of several oncogenic signaling molecules including EGFR, AKT, and ALK (14–16). However, factors that modulate dynamic PD-L1 expression during tumorigenesis and progression are not yet fully understood.

Activating BRAF mutations occur in roughly half of all melanomas, and treatment with BRAF or MEK inhibitors efficiently induces tumor shrinkage (17, 18). The antitumor efficacy of BRAF/MEK inhibitors is dependent, in part, on their immunosensitizing effects (19). MAPK pathway inhibition results in increased antigen presentation (20, 21), decreased immunosuppressive cytokine production (20), and increased CD8⁺ T-cell infiltration into melanoma tissue (20, 22, 23). Resistance to BRAF inhibitor (BRAFi) invariably arises after a median of 6 to 8 months in most patients. Unlike tumors on initial BRAFi therapy, tumors that progress after BRAFi therapy exhibit a decrease in both tumor-infiltrating T cells and tumor antigen expression (20, 24), as well as increased T-cell exhaustion markers, including PD-L1 (20, 22, 24–26). These results imply that the acquisition of BRAFi resistance in melanoma is linked with suppression of T-cell immune responses on multiple fronts, and understanding the evasion mechanism of T-cell antitumor immune response is essential for the improvement of BRAFi therapy efficacy.

YAP is a transcriptional coactivator that functions as an effector of the Hippo pathway and plays key roles in controlling tissue growth, regeneration, and stem cell homeostasis (27). YAP becomes activated in cancer cells by deletion of Hippo pathway components (28), GNAQ mutations (29), actin cytoskeleton rearrangement (30), and oncogenic mutations in APC and RAS genes (31, 32). Aberrant YAP activation promotes tumorigenesis and stem cell-like features of malignant cells (33) and also induces resistance to anticancer agents, including BRAF/MEK inhibitors (30, 34, 35). Accumulating evidence suggests an immunomodulatory effect of YAP in malignant tumors. Increased YAP activity leads to changes in the cytokine repertoire secreted by cancer cells, resulting in the recruitment of myeloid-derived suppressive cells (MDSC) and type II macrophages that suppress antitumor immune responses (36–38). These results suggest an interesting link between oncogenic YAP activation and immune evasion processes, but the direct influence of YAP activation on cytotoxic T-cell immune responses has been largely unaddressed.

Here, we show that YAP promotes PD-L1 expression and YAP-induced PD-L1 drives immune evasion in BRAFi-resistant melanoma. Increased YAP activity in melanoma cells potently suppressed both cytotoxic function and cytokine production of tumor antigen–specific CD8⁺ T cells. Anti–PD-1 and anti–PD-L1 blockade reversed YAP-mediated T-cell suppression. Our results demonstrate that the direct inhibitory effect of YAP on cytotoxic T cells serves as a key mechanism of YAP-mediated immune evasion. We also suggest that targeting YAP-mediated T-cell suppression would be a candidate strategy to improve prognosis of BRAF-mutant melanoma patients.

SKMEL28 and WM3248 cells were purchased from ATCC and Coriell Institute, respectively. Parental and BRAFi-resistant SKMEL28 and WM3248 cells were generated and maintained as described previously (30). A375SM cells were acquired from the Korean Cell Line Bank and grown in DMEM (Welgene) supplemented with 10% FBS (Welgene). MCF7, MDA-MB231, and A172 cells were acquired from ATCC and grown in DMEM supplemented with 10% FBS. HT29, KM12, and A549 cells were acquired from the Korean Cell Line Bank and grown in RPMI1640 (Welgene) supplemented with 10% FBS. Melanoma cell lines were confirmed for BRAFV600E mutation at the time of purchase by Sanger sequencing. Re-authentication of cell lines after purchase was not performed. BRAFi-resistant A375SM cells were generated by continuous treatment of 2 μmol/L PLX4032 (vemurafenib; Selleckchem) for 2 months. Established BRAFi-resistant cell lines were maintained in culture media containing 2 μmol/L PLX4032. BRAFV600E mutation of resistant cell lines was confirmed by Sanger sequencing. Cell lines were routinely tested for mycoplasma infection. All cell lines were stored in liquid nitrogen and cultured for no longer than 6 months before use.

Erlotinib and MK-2206 were purchased from Selleckchem, PD0325901 from Calbiochem, verteporfin from Sigma-Aldrich, and recombinant human IFNγ from R&D Systems. Retroviral vectors were used to generate melanoma cells and HEK293T cells that stably express YAP and its mutants. FLAG-YAP wild-type, FLAG-YAP-5SA, and FLAG-YAP-5SA-S94A cDNAs that were cloned into pMSCV-puro vector were provided by Dr. Dae-Sik Lim (KAIST). Retrovirus particle assembly, transfection, and puromycin selection were performed as described previously (30). We transfected cells with siRNAs (Supplementary Table S1) using Lipofectamine RNAiMAX (Invitrogen) according to the manufacturer's instructions.

Immunofluorescence staining and microscopy were performed as described previously (30). The primary antibodies used in this study are described in Supplementary Table S2. ImageJ software was used to analyze acquired images. YAP localization patterns were classified by inspecting at least 150 to 200 cells for each condition.

Immunoblotting and quantitative RT-PCR were performed as described previously (30). iQ SYBR Green Supermix (Bio-Rad) and CFX96 system (Bio-Rad) were used for real-time PCR. Target gene primers used in this study are described in Supplementary Table S3. Relative target gene expression levels normalized to GAPDH were determined by the ΔΔC(t) method using CFX Manager software (Bio-Rad).

Melanoma cells were plated on 96-well cell culture plates (SPL) at 2,500 cells/well and incubated for 24 hours after plating. Cells were treated with variable doses of PLX4032 (1 nmol/L to 100 μmol/L) for 72 hours, and cell viability was determined by Cell Counting Kit-8 reagent (Dojindo) according to the manufacturer's instructions. Cell viability data were fitted to sigmoidal dose–response curves and IC₅₀ values were calculated using GraphPad Prism (GraphPad Software).

Candidate YAP–TEAD binding sites near the PD-L1 transcription start site were identified by reviewing the ChIP-Seq data deposited by Zanconato and colleagues (GSE66081; ref. 39). A primer pair was designed for a ChIP assay targeting a narrow peak of TEAD4 binding at 13-kb upstream of the PD-L1 transcription start site (hg19, chr9:5,437,232-5,437,617): CATCGGGATTACCACGCTGA (Forward) and TTCGTTCCATTAGAGCGCGT (Reverse). ChIP assay was performed in YAP-5SA–expressing SKMEL28 and WM3248 cells using Magna ChIP A/G kit (Millipore) according to the manufacturer's instructions. Sheared chromatin was immunoprecipitated using control mouse IgG or anti-FLAG antibody (Sigma-Aldrich, F1804) overnight at 4°C. YAP interaction with the 13-kb upstream site was measured by quantitative PCR using iQ SYBR Green Supermix (Bio-Rad) and CFX96 system (Bio-Rad). The ChIP-qPCR signals from samples treated with control mouse IgG or anti-FLAG antibodies were normalized to the signals obtained from input samples.

A 800-bp fragment of the human genomic DNA sequence around the YAP–TEAD binding site in the 13-kb upstream from the PD-L1 gene was cloned into pGL4.26 luciferase vector (Promega). For luciferase assay, mock and YAP-5SA–expressing HEK293T cells were plated on 96-well cell culture plates (SPL) at 5,000 cells/well. The cells were cotransfected with pcDNA3.1-His-lacZ and pGL4.26:YAP–TEAD binding site for 24 hours using Lipofectamine LTX. Luciferase activity was measured using a luciferase assay kit (Promega) according to the manufacturer's instructions. Luciferase activity was normalized to β-galactosidase activity measured by a β-galactosidase assay kit (Promega).

Culture media containing 10% FBS were applied to mock or YAP-5SA–expressing melanoma cells, and culture supernatants were collected after 72-hour incubation for cytokine analyses and culture supernatant transfer experiments. IL2, IL4, IL7, GM-CSF, IFNα, and IFNγ levels in the culture supernatants were measured by CBA kit (BD Biosciences). IFNβ was measured by Verikine ELISA kit (PBL Assay Science), and IL15 was measured by Quantikine ELISA kit (R&D Systems) according to the manufacturer's instructions.

Cells were trypsinized, harvested, and stained using LIVE/DEAD Fixable Red Dead Cell Stain kit (Thermo Fisher Scientific) to exclude dead cells before incubation with fluorochrome-conjugated antibodies. For intracellular cytokine staining, brefeldin A (GolgiPlug, BD Biosciences), monensin (GolgiSTOP, BD Biosciences), and anti-CD107a antibody were added to the culture and maintained for 6 hours. After surface staining for 15 min at room temperature, the cells were processed using a fixation/permeabilization solution kit (BD Biosciences) and further stained with fluorochrome-conjugated anti-IFNγ and anti-TNFα antibodies. All flow cytometry analyses were performed using an LSR II instrument (BD Biosciences) and data were analyzed using FlowJo software (TreeStar). The antibodies used for flow cytometry are described in Supplementary Table S2.

Peripheral blood mononuclear cells (PBMC) were obtained from HLA-A2 positive healthy donors and CD8⁺ T cells were negatively isolated using a magnetic bead separation (MACS) kit (Miltenyi Biotec). Subsequently, Melan-A_26-35–specific CD8⁺ T cells were positively selected by MACS with phycoerythrin-conjugated HLA-A2-Melan-A_26-35 (ELAGIGILTV) pentamers (Proimmune Ltd.) and antiphycoerythrin microbeads (Miltenyi Biotec). Selected CD8⁺ T cells were expanded in RPMI 1640 media containing anti-CD3 (50 ng/mL), IL2 (200 IU/mL), IL7 (10 ng/mL), and IL15 (100 ng/mL) for 4 weeks using irradiated autologous PBMCs as feeder cells. The purity of Melan-A_26-35–specific CD8⁺ T cells was >95% based on flow cytometry using HLA-A2-Melan-A_26-35 pentamers.

Target melanoma cells were labeled with PKH26 dye (Sigma-Aldrich) according to the manufacturer's instructions and pulsed with 10 μg/mL Melan-A_26-35 peptide (ELAGIGILTV; PeproTech) for 1 hours at 37°C in a 5% CO₂ incubator. The target cells were cocultured with Melan-A-specific CD8⁺ T cells in 12 × 75 mm FACS tubes (Falcon) at variable effector:target ratios in the presence of anti–PD-1 (clone EH12.2H7, BioLegend, Inc.), anti–PD-L1 (clone 29E.2A3, BioLegend, Inc.), anti–PD-L2 (clone MIH18, BioLegend, Inc.) blocking antibodies, or IgG isotype control (clone MOPC-21, BioLegend, Inc.). After 6 hours, cells were harvested and stained with TO-PRO-3 dye (Thermo Fisher Scientific) to detect dead cells with disintegrative cytoplasmic membrane via flow cytometry. Specific killing was calculated by subtracting the baseline death rate of target cells from the death rate under variable conditions.

Melanoma tumor tissues were retrieved from 65 melanoma patients diagnosed at the Department of Pathology of Severance Hospital, South Korea (Supplementary Table S4). This study was reviewed and approved by the institutional review board at Severance Hospital. Formalin-fixed, paraffin-embedded tumor samples were used to obtain 5-μm-thick sections. IHC was performed with anti-YAP (sc-10119; Santa Cruz Biotechnology; 1:100 dilution) and anti–PD-L1 (E1L3N; Cell Signaling Technology; 1:100 dilution) antibodies in an automated immunohistochemical staining instrument (Ventana BenchMark XT; Ventana Medical System) according to the manufacturer's instructions. Ultraview Universal Alkaline Phosphatase Red Detection Kit (Roche Diagnostics) was used for detection. IHC staining of all slides was interpreted by an expert dermatology pathologist (Sang-Kyum Kim). YAP staining was classified as nuclear, nucleocytoplasmic (NC), or cytoplasmic according to the subcellular YAP staining pattern, and PD-L1 staining as − (negative), + (one-positive), or ++ (two-positive) depending on the proportion of stained cells and staining intensity.

The RNA-seq data for melanoma samples published by Hugo and colleagues (GSE65815; ref. 24) were obtained from the GEO database (40). Log₂(FPKM+1) values for CD274 (PD-L1) and IFNG (IFNγ) were plotted and linear regression analysis was performed for both pre-MAPK inhibitor treatment tumors (n = 17) and post-MAPK inhibitor treatment tumors (n = 43). YAP signature enrichment results were retrieved from the reference (24) and compared with the CD274 FPKM values. YAP-enriched tumors were defined as tumors enriched for either CORDENONSI_YAP_CONSERVED or YAP_HIPPO_KD_MOHSENI signature (Molecular Signature database). TCGA RNA-seq data for 472 melanoma tumors from 469 patients were downloaded through Firebrowse (Broad Institute). The normalized count calculated by RNA-Seq by expectation maximization (RSEM) analysis (41) was inserted into the gene set variation analysis (GSVA; ref. 42). The enrichment score for the YAP target gene signature (CORDENONSI_YAP_CONSERVED_SIGNATURE in Molecular Signature database) was determined by GSVA for each tumor sample, and the TCGA melanoma tumor samples were classified into YAP^low (samples with a negative enrichment score), YAP^mid (samples with enrichment score below median), and YAP^high (samples with enrichment score above median). CD274 Log₂(RSEM normalized count+1) values of the three groups were compared. The GSVA was performed using R version 3.3.1.

GraphPad Prism (GraphPad Software) was used for analyzing data and creating graphs. Significance was set at 0.05 in the Mann–Whitney U test or unpaired t test.

To access the influence of YAP activation on immune checkpoint pathways, we first examined the surface expression of T-cell coinhibitory/costimulatory ligands (PD-L1, PD-L2, B7-1, B7-2, Galectin-9, and CD155) in melanoma cells expressing constitutively active YAP. We established BRAF^V600E-mutant melanoma cell lines (SKMEL28, WM3248, and A375SM) harboring FLAG-tagged YAP-5SA vector. Previous studies have shown that 5SA mutations cause constitutive nuclear localization and transcriptional activation of YAP by preventing inhibitory phosphorylation of YAP by the Hippo pathway kinases LATS1 and LATS2. Interestingly, melanoma cell lines expressing YAP-5SA displayed greater PD-L1 and PD-L2 surface expression when compared with parental cells (Supplementary Fig. S1A and S1B). In addition, overexpression of YAP-5SA, but not wild-type YAP, increased PD-L1 mRNA expression as well as its surface protein expression in the three melanoma cell lines (Fig. 1A and B; Supplementary Fig. S1C). These results indicate that YAP activation induced PD-L1 expression in melanoma cells. We also observed YAP-5SA–induced PD-L1 upregulation in breast cancer (MCF7, MDA-MB231), non–small cell lung cancer (A549), colon cancer (HT29), and glioblastoma (A172) cells (Supplementary Fig. S2). However, YAP-5SA expression did not significantly increase PD-L1 levels in KM12 colon cancer cells, suggesting that YAP-mediated PD-L1 induction was dependent on cellular context.

Our group and others have reported robust YAP activation in melanoma cells resistant to BRAFi (30, 34). Therefore, we examined PD-L1 mRNA and protein expression in BRAFi (vemurafenib)-resistant melanoma cells. BRAFi-resistant melanoma cell lines were generated as previously described (30), and we confirmed the acquisition of BRAFi resistance (Supplementary Fig. S3A). As expected, SKMEL28, WM3248, and A375SM cells resistant to BRAFi expressed more PD-L1 mRNA and cell surface protein compared with parental cells (Fig. 1C and D). Moreover, RNAi-mediated knockdown of YAP and its paralog TAZ suppressed both PD-L1 transcription and cell surface expression in BRAFi-resistant cells, demonstrating that enhanced YAP/TAZ activity is associated with increased PD-L1 expression (Fig. 1E and F; Supplementary Fig. S3B). Our previous microarray data (GSE68599; ref. 30) that compared parental and BRAFi-resistant melanoma cells, as well as YAP/TAZ-depleted resistant cells, also confirm both the upregulation of PD-L1 in resistant cells among immune checkpoint ligands and the decline of PD-L1 expression after YAP/TAZ knockdown (Supplementary Fig. S3C). Collectively, these results indicate that YAP activation upregulated PD-L1 expression in melanoma cells.

Previous studies have reported that oncogenic activation of the EGFR, AKT, and MAPK pathways can induce PD-L1 expression in cancer cells (14, 15, 25). To test whether these pathways are involved in YAP-mediated PD-L1 expression, we treated YAP-5SA–expressing melanoma cells with erlotinib (EGFR inhibitor), MK-2206 (AKT inhibitor), or PD0325901 (MEK inhibitor) and measured PD-L1 expression. Elevated PD-L1 expression was largely unaffected by suppression of the oncogenic signaling pathways, indicating that YAP-mediated PD-L1 expression does not require these pathways (Fig. 2A and B). PD-L1 expression is also induced by IFNγ exposure during prolonged T-cell immune response, eliciting adaptive immune evasion, so we tested whether YAP was involved in IFNγ-mediated PD-L1 upregulation. Short-term (48 hours) and long-term (7 days) IFNγ treatment did not promote YAP nuclear localization, which was an indication of YAP activation (Fig. 2C). In addition, YAP/TAZ depletion did not affect PD-L1 induction by IFNγ (Fig. 2D). We also examined the effect of YAP activation on the JAK/STAT pathway, the major downstream effector of IFNγ signaling. There was no phospho-STAT1 activity both in mock and YAP-5SA–expressing melanoma cells, but IFNγ-treated mock A375SM cells showed high phospho-STAT1 (Supplementary Fig. S4A). These results suggested that the IFNγ pathway and YAP independently regulate PD-L1 expression.

We next examined possible involvement of secretory cytokines in YAP-mediated PD-L1 upregulation. We measured the concentrations of IL2, IL4, IL7, IL15, GM-CSF, IFNα, IFNβ, and IFNγ, which were previously reported to regulate PD-L1 expression (43), in the culture supernatants of mock and YAP-5SA–expressing melanoma cells. However, no changes in the concentrations of the examined cytokines were detectable (Supplementary Fig. S4B). Moreover, addition of the culture supernatants derived from YAP-5SA–expressing melanoma cells did not upregulate PD-L1 expression in parental melanoma cells (Supplementary Fig. S4C). These results excluded the possibility that activation of an autocrine cytokine signaling is a central mechanism by which YAP promotes PD-L1 expression.

YAP promotes target gene transcription mainly through its interaction with TEAD family transcription factors that bind to promoters or enhancers (44). The introduction of a mutation at the TEAD binding site (S94A) significantly dampened PD-L1 upregulation by YAP-5SA in SKMEL28 and WM3248 cells (Fig. 3A and B). Moreover, treatment of YAP-5SA–expressing cells with verteporfin, an inhibitor of YAP–TEAD interaction, decreased PD-L1 mRNA and surface expression (Fig. 3C and D). Thus, an intact YAP–TEAD interaction was essential for the upregulation of PD-L1 transcription by YAP activity. Next, we searched for YAP–TEAD binding sites in potential PD-L1 gene enhancer regions by reviewing publicly available ChIP-seq data (39). We found a narrow peak of TEAD4 binding at 13-kb upstream of the PD-L1 transcription start site that may provide YAP–TEAD binding sites. To confirm YAP binding to this region, we performed ChIP analysis. We detected specific binding activity of YAP-5SA to the 13-kb upstream region (Fig. 3E). To validate YAP-driven PD-L1 transcriptional activation by this candidate enhancer region, we cloned the 13-kb upstream sequence (800-bp fragment) into a luciferase vector with minimal promoter. The luciferase activity driven by the enhancer sequence was significantly higher in YAP-5SA–expressing HEK293T cells than in mock-transfected HEK293T cells (Fig. 3F). These results suggest that YAP–TEAD binding to the 13-kb upstream enhancer directly induced PD-L1 transcription.

Our finding that YAP activation promotes both BRAFi resistance (30) and PD-L1 expression suggested that YAP served as a molecular link between BRAFi resistance and immune evasion processes. Therefore, we investigated the influence of YAP activation in BRAFi-resistant melanoma cells on the effector functions of tumor antigen–specific CD8⁺ T cells. We performed cytotoxicity assays using A375SM cells, which express HLA-A2. Neither YAP-5SA expression nor BRAFi resistance acquisition affected HLA-A2 expression (Supplementary Fig. S5A). We established CD8⁺ T-cell lines specific for an HLA-A2–restricted Melan-A peptide (Melan-A_26-35, ELAGIGILTV) from PBMCs of HLA-A2⁺ healthy donors. The resulting cell cultures contained 95.9% Melan-A_26-35–specific CD8⁺ T cells and exhibited minimal PD-1 expression (Fig. 4A and B). To mimic T-cell exhaustion in the tumor microenvironment, we cultured the T-cell lines with Melan-A_26-35 peptide-pulsed PD-L1⁺ A375SM melanoma cells harboring YAP-5SA. After 7 days of coculture, we observed a strong induction of PD-1 expression on the T cells (Fig. 4B). Next, we examined the cytotoxicity of PD-1⁺ Melan-A_26-35–specific CD8⁺ T cells against cognate antigen-pulsed parental and BRAFi-resistant A375SM cells. CD8⁺ T cells successfully killed parental A375SM cells, showing a proportional increase in cytotoxicity as the effector-to-target ratio increased (Supplementary Fig. S5B). BRAFi-resistant A375SM cells exhibited resistance to CD8⁺ T-cell cytotoxicity (Fig. 4C), which was restored by PD-1 blockade by anti–PD-1 antibody, whereas anti–PD-1 antibody did not affect the killing of parental cells. We also examined other effector functions of T cells cocultured with melanoma cells. We observed significant decreases in the production of IFNγ, tumor necrosis factor-α (TNFα), and the expression of CD107a, a marker of cytotoxic degranulation activity, in CD8⁺ T cells cocultured with BRAFi-resistant melanoma cells (Fig. 4D). This confirms T-cell exhaustion in BRAFi-resistant melanoma cells. In agreement with the cytotoxicity recovery, anti–PD-1 antibody treatment restored T-cell effector functions (Fig. 4D). These results demonstrate that BRAFi-resistant melanoma cells induced immune evasion via PD-1–dependent exhaustion of CD8⁺ T cells.

We next examined the impact of YAP activation on PD-1⁺ Melan-A_26-35–specific CD8⁺ T-cell immune responses by assaying cytotoxicity against mock and YAP-5SA–expressing A375SM cells. Melanoma cells expressing YAP-5SA were more resistant than mock cells, and PD-1 blockade reversed the T-cell cytotoxicity toward YAP-5SA–expressing A375SM cells (Fig. 5A). Coculture of Melan-A_26-35–specific CD8⁺ T cells with YAP-5SA–expressing cells decreased their production of IFNγ and TNFα, as well as decreased expression of CD107a (Fig. 5B). Decreases in IFNγ, TNFα, and CD107a expression were restored by PD-1 blockade (Fig. 5B). Because YAP-5SA induced the expression of PD-L2 as well as PD-L1 (Supplementary Fig. S1A and S1B), we next tested whether both PD-L1 and PD-L2 contribute to YAP-mediated T-cell suppression. Similar to anti–PD-1, PD-L1 blocking antibody restored T-cell cytotoxic function (Supplementary Fig. S6A), IFNγ and TNFα production, and CD107a expression (Supplementary Fig. S6B and S6C). In contrast, PD-L2 blocking antibody treatment did not affect cytotoxicity and cytokine production of CD8⁺ T cells against YAP-5SA-expressing melanoma cells, suggesting that PD-L2 upregulation is dispensable for YAP-mediated T-cell suppression. Taken together, these results show that YAP activation in melanoma cells was responsible for promoting PD-1/PD-L1–dependent evasion of cytotoxic T cells.

PD-1 blockade resulted in incomplete restorations of cytotoxic function of CD8⁺ T cells cocultured with YAP-5SA–expressing melanoma cells (Fig. 5A). One possibility is that PD-1/PD-L1 interaction may not be completely blocked by the addition of anti–PD-1 to the culture. Alternatively, it is also possible that YAP can elicit PD-L1–independent mechanisms for immune evasion. We observed suppression of cytotoxic function and cytokine production in nonexhausted PD-1–negative Melan-A_26-35–specific CD8⁺ T cells after coculture with YAP-5SA–expressing melanoma cells (Fig. 5C and D). The suppression was not affected by PD-1 blockade. This observation indicates that YAP could promote melanoma cell's immune evasion in both PD-L1–dependent and PD-L1-independent manner.

To validate the clinical significance of YAP-mediated PD-L1 expression, we investigated the association between YAP activity and PD-L1 expression in human melanoma tumor samples. We first analyzed RNA-seq data for 472 melanoma tumors from TCGA database. We stratified the tumors based on the YAP signature enrichment score calculated by GSVA (42). Tumors with high YAP enrichment scores showed significantly higher PD-L1 expression (Fig. 6A) in line with the above findings. In addition, we performed immunohistochemical staining of YAP and PD-L1 in tumor specimens from 65 melanoma patients (Supplementary Table S4). As expected, tumors with nuclear or NC YAP staining (defined as high YAP activity) had significantly higher PD-L1 expression than tumors with cytoplasmic YAP staining (defined as low YAP activity; Fig. 6B). We also explored the linkage between YAP and PD-L1 in previously published data on BRAFi-resistant melanoma. Hugo and colleagues comprehensively analyzed paired melanoma samples collected before and after (at progression) BRAFi/MEKi therapy (GSE65185; ref. 24). They found YAP signature activation in a subset of resistant tumors. PD-L1 expression correlated with IFNγ expression in the pretreatment tumors (Fig. 6C). However, the correlation weakens after acquiring resistance, suggesting that factors other than IFNγ play a key role in controlling PD-L1 expression in BRAFi/MEKi-resistant tumors. PD-L1 was also highly upregulated in a subset (4 out of 17) of YAP signature–enriched BRAFi/MEKi-resistant tumors (Fig. 6D). Taken together, these data provided in vivo evidence for YAP-mediated PD-L1 upregulation in human melanoma tumors.

The PD-1/PD-L1 axis is a main target for immune checkpoint blockade. Previous studies have shown that PD-L1 expression in tumor cells is induced by IFNγ and oncogenic signals (13–16). In the present study, we have demonstrated a mechanism by which acquisition of BRAFi resistance evokes PD-L1 upregulation. It has been shown that aberrant activation of YAP is an important mechanism of BRAFi resistance in melanoma (30, 34). Our results demonstrated that enhanced YAP activity in BRAFi-resistant melanoma cells directly inhibited cytotoxic T-cell immune responses by PD-1/PD-L1 immune checkpoint pathways. Thus, YAP-mediated BRAFi resistance acquisition not only indicates resistance to apoptotic and antiproliferative effects of BRAFi, but also evasion from antitumor T-cell responses in melanoma.

YAP is a versatile player in malignant processes, receiving inputs from the tumor microenvironment and interacting with several oncogenic pathways, including WNT, GPCR, and KRAS (29, 45, 46). YAP activates potent transcriptional programs for cancer cell survival, metastasis, stem cell–like properties, and drug resistance (33). The deletion of Hippo pathway components or constitutive YAP activation sufficiently induces tumorigenesis in mouse models by increasing cell proliferation and stemness (47, 48). Our finding that YAP-mediated direct suppression of cytotoxic T-cell immune responses adds another layer to the complexity of the role of YAP in cancer pathogenesis. Because cytotoxic T-cell immune response is a primary mechanism of immune surveillance of tumor cells and a vital target for tumor immunotherapy, the current study provides potential evidence of YAP involvement in antitumor immune responses and susceptibility to immunotherapy. In addition, YAP can recruit and accumulate MDSCs in prostate and pancreatic cancers by upregulating multiple chemokines (37, 38). Therefore, suppression of the tumor immune response is one of the key features of YAP-driven cancer pathogenesis. YAP in tumor cells receives diverse regulatory inputs, including cell polarity, actin cytoskeleton dynamics, extracellular matrix stiffness, and cell metabolism (27, 33). Thus, we speculate that tumor-infiltrating lymphocytes are likely under the control of the complex factors that modulate YAP activity in tumor cells. Because a considerable proportion of solid cancers are reported to express YAP (49), the influence of YAP on the antitumor immune response and T-cell exhaustion needs to be further investigated in various clinical and molecular contexts.

Our group and others previously identified YAP as an important player in BRAFi-resistance establishment in melanoma cells. YAP has been reported to counteract anticancer effects of BRAF inhibition by promoting E2F-related cell cycle progression and also by upregulating antiapoptotic protein Bcl-xL (30, 34). The current study proposes another role of YAP in BRAFi resistance that promotes PD-L1 upregulation and evasion from T-cell immune responses. Previous studies have reported significant changes in immunological properties of melanoma tumors progressed on BRAFi, showing decreased T-cell infiltration and increased T-cell exhaustion markers. Our findings suggest that activation of YAP in melanoma cells is a key mechanism that mediates the interplay between BRAFi resistance and immune microenvironment, providing a clue to understand dynamic immunological changes of melanoma tumors upon BRAFi treatment.

BRAFi-resistant and YAP-5SA–expressing cells become susceptible to cytotoxic T-cell attack after anti–PD-1 and anti–PD-L1 antibody treatment, with YAP-mediated T-cell suppression reversible by PD-1 and PD-L1 blockade, which suggests that YAP-mediated immune evasion can be targeted by PD-1/PD-L1 blockade. In contrast, PD-L2 blocking antibody did not affect YAP-mediated T-cell suppression. Preclinical studies using a melanoma mouse model have reported an increase in antitumor activity with the BRAFi/anti–PD-1 combination compared with single agent therapy (23, 50), and our study further supports the rationale for combination or sequential therapy. However, it should be also noted that YAP can induce resistance to cytotoxicity of nonexhausted PD-1–negative CD8⁺ T cells (Fig. 5C). Although the PD-1/PD-L1–mediated immune checkpoint plays a major role in melanoma's evasion from CD8⁺ T-cell immune responses, PD-L1–independent mechanisms promoted by YAP may also contribute to the inhibition of CD8⁺ T-cell functions. Previous studies have reported a decrease in T-cell infiltration in YAP-activated cancer tissues (37, 38). These findings suggest that YAP activity inhibits adaptive immune response against tumors in a complex manner, and YAP–TEAD targeting agents with or without combination with immune checkpoint blockade need to be tested to inhibit YAP-mediated immune evasion processes.

In summary, our present work demonstrates an interplay between drug resistance in molecular targeted therapy and tumor immune evasion. YAP confers an immune evasion mechanism to BRAFi-resistant melanoma cells, via contributing to PD-L1 expression and evasion of T-cell immune responses, which can be targeted by anti–PD-1/PD-L1 blockade. We expect that targeting of YAP-mediated immunological changes will improve BRAFi therapy efficacy and melanoma patient survival.

No potential conflicts of interest were disclosed.

Conception and design: M.H. Kim, C.G. Kim, S.-H. Park, E.-C. Shin, J. Kim

Development of methodology: M.H. Kim, C.G. Kim, S.-H. Park, E.-C. Shin

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): M.H. Kim, C.G. Kim, S.-K. Kim, S.J. Shin, E.A. Choe, S.-H. Park

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): M.H. Kim, C.G. Kim, S.-K. Kim, S.J. Shin, E.A. Choe, S.-H. Park, J. Kim

Writing, review, and/or revision of the manuscript: M.H. Kim, C.G. Kim, S.-K. Kim, S.J. Shin, E.A. Choe, S.-H. Park, E.-C. Shin, J. Kim

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): C.G. Kim, S.J. Shin, E.A. Choe, J. Kim

Study supervision: S.-H. Park, E.-C. Shin, J. Kim

This study was supported by research grants through the National Research Foundation of Korea (2014R1A2A1A10053662 and 2016M3A9B4915821), and was also supported by a grant of the Korea Health Industry Development Institute (KHIDI) funded by the Ministry of Health and Welfare, Republic of Korea (HI15C2817). We thank Professor Dae-Sik Lim (KAIST) for reagents and helpful comments on the manuscript.

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.