Effects of MAPK and PI3K Pathways on PD-L1 Expression in Melanoma

This article is featured in Highlights of This Issue, p. 3359

Recent developments in the field of targeted therapy and immunotherapy have increased the life expectancy of patients with advanced melanoma. These therapies include BRAF inhibitors (BRAFi) such as vemurafenib and dabrafenib (1, 2), the MEK inhibitor (MEKi) trametinib (3), and the anticytotoxic T-lymphocyte-4 (CTLA4) antibody ipilimumab (4, 5). Further advances are being made with other therapies including antiprogrammed cell death 1 (PD-1) and programmed death ligand-1 (PD-L1, B7-H1, or CD274) antibodies (6, 7, 8), and tumor-infiltrating lymphocytes (TIL) or T-cell receptor (TCR)-engineered adoptive cell transfer therapy (9).

BRAF mutation at position V600 is detected in approximately half of melanomas. This mutation causes overactivity of downstream signaling in the MAPK pathway (10). Growth of melanoma cells with the BRAF^V600 mutation can be effectively blocked by BRAFi (11, 12). Interestingly, contrary to BRAF-mutant cancer cells, in other cells with wild-type BRAF gene, BRAFi paradoxically activates the MAPK pathway (13), which is also evident in activated lymphocytes (14). Despite the initial response of BRAF-mutant tumors to the BRAFi, acquired resistance develops frequently and most patients will relapse within months (15). To prevent the emergence of resistance through the reactivation of the MAPK pathway, treatment regiments with the combinations of BRAF and MEKi have been investigated in clinical trials indicating higher beneficiary effects and lower side effects (3). Moreover, to block the resistance through alternative signaling pathways (16), clinical trials with the combinations of BRAF and PI3K/AKT inhibitors have been initiated (U.S. cooperative group clinical trial S1221, NCT 01902173).

Melanoma has been considered an immunogenic malignancy. However, the endogenous antitumor immune response is not sufficient to control the cancer in the great majority of cases. Therefore, different strategies have been devised to augment the immune reaction against the melanoma cells (17, 18, 19). Ipilimumab is an anti-CTLA4 antibody that blocks the CTLA4-induced T-cell inhibition in the activation phase of an antitumor immune response. In two randomized clinical trials, this anti-CTLA4 antibody has shown improvement in overall survival of patients with metastatic melanoma (4, 20).

The coinhibitory receptor/ligand pair, PD-1/PD-L1, is one of the main peripheral regulatory mechanisms for induction of anergy in immune cells and maintenance of peripheral tolerance (21). PD-1 expression is induced on the cell surface upon T-cell activation (22). The cytoplasmic domain of PD-1 contains a tyrosine-based inhibitory motif (ITIM) and an immunoreceptor tyrosine-based switch motif (ITSM). Interaction with PD-L1 causes phosphorylation of a tyrosine in the ITSM motif of PD-1 receptor. Consequently, SH2-domain containing tyrosine phosphatase 2 (SHP-2), and possibly SHP-1, are recruited to this motif of PD-1 resulting in downregulation of the PI3K/AKT and MAPK signaling pathways downstream of the TCR and blockade of cell-cycle progression in the immune cells (23, 24).

Constitutive PD-L1 expression has been detected in different tumors including melanoma (21). Moreover, on tumor cells, expression of PD-L1 can be induced by INFγ and other cytokines that are produced by the activated lymphocytes (25). In other words, tumor-reacting immune cells inadvertently trigger an inhibitory mechanism, which has been termed adaptive immune resistance (26). Therefore, therapeutic attempts to block PD-L1 and PD-1 interaction should shift the balance toward higher activity of antitumor immune cells. Indeed, in a recent clinical trial with the anti-PD-1 antibody, nivolumab objective responses in 18% to 28% of patients with different cancers were observed (7), and another anti-PD-1 antibody, MK-3475 (lambrolizumab) had a response rate of 38% in patients with advanced melanoma (6).

Both targeted therapy and immunotherapy have shown to be effective in melanoma, while they function through different mechanisms. Therefore, the idea of combining these two types of therapies has been very appealing. However, the effects of targeted therapy drugs on the effectors and targets of immune system have not been clearly understood (27). Moreover, the associations or the cross-talk between the oncogenic driver pathways in cancer cells and the immunoregulatory pathways, such as PD-L1, have not been fully elucidated. In this study, we investigated the expression level of PD-L1 in a panel of 51 melanoma cell lines, including vemurafenib-sensitive and -resistant BRAF mutants, NRAS mutants, and NRAS/BRAF wild-types. Some of the cell lines also contained additional mutations in signaling molecules involved in the PI3K/AKT pathway. Moreover, in the presence or absence of INFγ, or in a coculture of melanoma cells and lymphocytes, we studied the effects of blocking the main oncogenic driver pathways on expression of PD-L1. The paradoxical MAPK activating effect of vemurafenib on signaling and cytokine production of lymphocytes that are exposed to the inhibitory effects of PD-1/PD-L1 interaction was studied as well.

Vemurafenib and trametinib were purchased from Selleck chemicals and the pan-PI3K inhibitor (PI3Ki) GSK2126458 was obtained from GlaxoSmithKline under a material transfer agreement. INFγ was obtained from Sigma-Aldrich. Human melanoma cell lines (M series) were established from patient's biopsies under UCLA IRB 11-003254 as previously described (16). WM1366 and SBCL2 were obtained from the ATCC. Growing media, maintenance, and mycoplasma testing of the cell lines were as described before (16). BRAF-mutant cell lines with the in vitro acquired vemurafenib resistance are indicated by the AR suffixes. Except M249AR4 that contains a secondary NRAS mutation, the rest of acquired resistant cell lines were maintained in the growing media containing 1 μmol/L of vemurafenib.

Mutations in the cell lines were determined as it has been described before (12). In brief, the genomic DNA isolated from each cell line was subjected to analysis by OncoMap 3 or Iontrone, and the mutations were confirmed by direct Sanger sequencing.

The growth and viability of the cells treated with serial dilutions of vemurafenib were determined by the bioluminescence assay (Promega) as it was described previously (16). Growth inhibition rates were used to calculate the IC₅₀ for each cell line. Each assay was performed at least twice in duplicates.

Cell lines were transfected with the PD-L1 specific or no target control siRNAs (Dharmacon) as it has been described before (16). Transfected cells were harvested after 72 hours for detection of PD-L1 levels by flow cytometry assays.

To induce production of PD-1, frozen peripheral blood mononuclear cells (PBMC; obtained under UCLA IRB 10-001598) were defrosted and after treatment with 227 U/mL of DNase (Sigma) for 30 minutes, transferred to RPMI-1640 containing 5 μg/mL phytohemagglutinin (Sigma), 5% heat inactivated human AB serum, 1% penicillin, streptomycin, and amphotericin (Omega Scientific), and incubated for 48 hours (24). This PBMC culturing method was used to induce proliferation of activated T lymphocytes by mitogen activation and precondition them to express PD-1. Then cells were rested overnight in the same growth condition minus the phytohemagglutinin (Fig. 4A). These cells then were used for the coculture with the melanoma cells or for the exposure to the recombinant PD-L1 (SinoBio Biotech).

Melanoma cells were plated in two sets in 12-well plates and the following day, they were treated either with DMSO, BRAFi (vemurafenib), MEKi (trametinib), or the PI3Ki. After another 24 hours, one set was treated with media and the other set with INFγ. The final concentrations of the drugs were 1.5 μmol/L of vemurafenib, 25 nmol/L of trametinib, 100 nmol/L of PI3Ki, and 200 U/mL of INFγ. In the case of cocultures, all the steps and conditions were the same except at the start of drug treatment, the media were switched to the media containing 5% of human AB serum instead of FBS, and the next day a suspension of lymphocytes (primed as described in the above) was added to each well. The final concentration of lymphocytes was 550,000 cells/mL. Each assay was repeated at least twice.

Nontissue culture 24-well plates were coated overnight with a mixture of 60 μg/mL of Fc-PD-L1 recombinant protein (SinoBio Biotech), 15.4 μg/mL of anti-CD3, and 19.4 μg/mL of anti-CD28 (both from BD biosciences) in PBS or the same mixture containing Fc instead of Fc-PD-L1 as the control. The next day, plates were blocked with 2.5% human serum albumin (Octapharma) in PBS for one hour. After removing the blocking solution, at a concentration of 700,000 cells/mL, lymphocytes (prepared as described in the above) were added to the wells in the presence of DMSO or 1.5 μmol/L of vemurafenib in RPMI-1640 + 5% of human AB serum and 1% of antibiotics. After 24 hours, each supernatant was collected for performing cytokine ELISA assays and the cells were harvested for the Western blot analysis. Concentrations of IL-1, IL-6, TNFα, and INFγ were determined by eBioscience ELISA Kits. The assays were performed in duplicates according to the instruction of the manufacturer after diluting the samples 1:100 for TNFα and INFγ assays and 1:10 for IL-1 and IL-6 assays.

Lymphocytes were primed and treated as it was mentioned in the above and after harvest were lysed and subjected to Western blotting. Primary antibodies included p-ERK Thr204/205, ERK, p-MEK Ser217/221, MEK, GAPDH (all from Cell Signaling Technology) and PD-1 from BioLegend. Western blotting was performed as previously described (16). Densitometry of bands was performed by using ImageJ program.

Melanoma cells were probed with anti-PD-L1 (CD274) in APC (catalog # 329708; BioLegend). In the coculture experiments to distinguish melanoma cells from the immune cells, samples were stained with both anti-CD274 and anti-CD45 in Brilliant Violet 605 (catalog # 304042; BioLegend) (Fig. 5A). All live/dead discrimination was performed with 7-Aminoactinomycin D (7AAD) (A07704, Beckman Coulter). Lymphocytes were stained with CD45, CD8a in Brilliant Violet 650(301041; BioLegend), and CD4 in Brilliant Violet 510 (317444; BioLegend). All samples were run on a Beckton Dickinson LSRII flow cytometer. The optimal staining amounts of antibodies were determined by internal laboratory experiments, and applied to the cells at 15 minutes at room temperature, while protected from light. Cells were gated according to the following schema: morphology was determined by using the area of the forward scatter emission peak (FSC-A) versus the area of the side scatter emission peak (SSC-A). Segregation of single cells was determined using SSC-A versus the width of the side scatter emission (SSC-W). Comparing the 7AAD with the APC emission peak allowed analysis of PD-L1 for the melanoma cells. For coculture assays, live/dead and lymphocyte discrimination was determined by comparing area of the 7AAD emission peak with the area of the CD45 emission peak and then 7AAD with APC for melanoma cell. Median fluorescent intensity (MFI) of PD-L1 was taken from plots of PD-L1.

Analyses were performed in the R environment (R Core Team; 2013) and MS Excel.

To investigate the levels and prevalence of PD-L1 expression, a panel of 51 melanoma cell lines was analyzed by flow cytometry. For this purpose, at first, the validity of the anti-PD-L1 antibody was evaluated by its ability to detect the increase in expression of PD-L1 upon treatment of two melanoma cell lines (M233 and M257) with INFγ and the decrease of PD-L1 level after siRNA knockdown of this ligand. By using the optimal titration of the antibody, flow cytometry assay indicated around 2-fold induction of MFI by INFγ and around 90% reduction of MFI by the PD-L1 siRNA pool in comparison with the nontargeted control siRNA pool (Fig. 1A–C).

The panel of melanoma cell lines used in this study contained 27 BRAF mutants, 15 RAS mutants, three BRAF/NRAS double mutants, and six BRAF/RAS wild-types (Table 1). Expression level of PD-L1 was variable among melanoma cell lines showing a range of more than 2 log difference (Fig. 1D). Of note, in this analysis we evaluated the unstimulated (constitutive) expression of PD-L1 without exposure to IFNs. In this panel of cell lines, the measured MFI levels of PD-L1 showed random distributions (Kolmogorov–Smirnov test) independent of the driver mutation in the melanoma cells (Fig. 1E). The majority of cell lines showed very low levels of PD-L1 expression (MFI<100). The highest PD-L1 expression was detected in four of the cell lines (M296, M233, M257, and WM1366), which were 4- to 22-fold higher than the average expression of PD-L1 on all the cell lines in the panel. M296 and WM1366 are NRAS mutants, M233 is a BRAF mutant, and M257 is a NRAS/BRAF wild-type. These results indicate that very high expression of PD-L1 is not specific to any particular group of driver oncogene-mutated melanoma cell lines.

It had been previously proposed that PTEN deletion and increased PI3K/AKT signaling resulted in increased PD-L1 expression in glioblastoma (28). To investigate a possible association between overactivity of the PI3K/AKT pathway and the expression of PD-L1, the same panel of melanoma cell lines was analyzed according to the absence or presence of known mutations, deletions, or amplification of genes involved in activity of the PI3K/AKT pathway (Table 1). Cell lines with PI3KCA and heterozygous PTEN mutations, PTEN null expression, and amplification or mutation in any of the AKT isoforms showed a distribution pattern of PD-L1 expression similar to the other cell lines (Kolmogorov–Smirnov test) ranging from very low to very high levels (Fig. 1F). Therefore, PD-L1 expression in this large panel of melanoma cell lines with defined oncogenic alterations was independent of the oncogenic events in the MAPK and PI3K/AKT pathways.

It has been reported that vemurafenib-resistant melanoma cell lines exhibit induction of PD-L1 expression (29). In the current study, after performing growth inhibition assays, any cell line with an IC₅₀>1,000 nmol/L of vemurafenib was considered to be resistant to this drug. Expression of PD-L1 on BRAF-mutated cell lines and their levels of sensitivity to vemurafenib were graphed side by side (Fig. 2A). Cell lines were arranged from the highest to the lowest expression of PD1-L. As compared with the expression levels of PD-L1, cell lines with different IC₅₀ were randomly distributed. Although no association between the level of PD-L1 and the response to the BRAFi could be seen, all the sensitive cell lines (IC₅₀<1,000 nmol/L) showed relatively low levels of PD-L1 expression (MFI<275) with the median MFIs of 106. Cell lines with an average resistance to vemurafenib (1,000<IC₅₀<3,000 nmol/L) also had low levels of PD-L1 expression with the median of MFIs equal to 65. However, very resistant cell lines (IC₅₀>10,000 nmol/L) showed a bimodal distribution of PD-L1 expression with an overall median of MFIs equal to 229. Out of eight very resistant cell lines, four exhibited low levels of PD-L1 (MFI<139) and four showed high levels between 320 and 9,000 PD-L1 MFI. It should be mentioned that the BRAF-mutant resistant cell lines, M376 and M398, contain secondary NRAS mutations (16, 30) which is the cause of resistance to vemurafenib. Both of these cell lines showed very low levels of PD-L1 expression.

The same analysis was performed with five pairs of sensitive parental and their in vitro acquired vemurafenib-resistant subcell lines. Similar to the innately resistant cell lines, in vitro acquired vemurafenib-resistant subcell lines also showed a bimodal pattern of PD-L1 expression. As shown in Fig. 2B, two of the resistant cell lines (M229AR9 and M238AR2) exhibited more than 9-fold induction of the ligand in comparison with their parental sensitive cell lines. Expression of PD-L1 in M249AR4 was still at low levels (MFI = 216), although it was twice of PD-L1 expression in the M249 parental cell line. Two of these resistant lines (M397AR and M409AR) showed the same or slightly lower levels of PD-L1 expression compared with their parental lines. M229AR9 and M238AR2, the two resistant cell lines with induction of PD-L1, also show induction of PDGFRβ expression and their mechanism of resistance to vemurafenib is through the activation of alternative signaling pathways other than the MAPK (30). M249AR4 is a PTEN null cell line with a secondary NRAS mutation that causes reactivation of the MAPK pathway and resistance to vemurafenib (16, 30). Mechanism of resistance in M397AR is alternative splicing of BRAF mRNA and expression of truncated BRAF, which renders vemurafenib ineffective (31). The mechanism of resistance in M409AR is still under investigation. Therefore, it seems that the resistance due to the activation of alternative signaling pathways is accompanied with a high induction of PD-L1, whereas the resistance due to the reactivation of MAPK has no or slight inducing effect on the expression of this ligand.

The expression of PD-L1 is mainly controlled by INFγ through the activation of the JAK/STAT pathway (25), whereas overactivity of the MAPK pathway is the main oncogenic driver in the majority of melanomas. In addition, the role of the PI3K/AKT pathway and its cross-talk with the MAPK pathway in survival and resistance to inhibitors of the MAPK pathway has been described (16). To investigate the potential effects of these pathways on expression of PD-L1, in the presence and absence of INFγ, cell lines representative of BRAF-mutant, NRAS-mutant, and BRAF/NRAS wild-type melanoma cell lines were treated with solvent, vemurafenib, the MEKi trametinib, or a PI3Ki and analyzed for the MFI of PD-L1 surface expression (Fig. 3A).

In the majority of cell lines, vemurafenib at 1.5 μmol/L level had no effect on the expression of PD-L1. However, in one NRAS-mutant cell line (WM1366), vemurafenib caused a slight induction of the ligand (Fig. 3B), which may be related to paradoxical MAPK activation. Also, in the presence of INFγ, vemurafenib caused induction of PD-L1 in one of the vemurafenib-resistant cell lines (M233). On the other hand, M229AR9, despite being resistant to the growth inhibitory effect of vemurafenib, was the only BRAF-mutant cell line that showed a reduction of PD-L1 upon the treatment with vemurafenib or combination of vemurafenib and INFγ (Fig. 3B). For these studies, the control cultures of this resistant cell lines were off the drug for 24 hours, and treated sample had the reintroduction of the drug. The difference between the expression of PD-L1 in the control and treated samples can be interpreted in two ways: (i) the induction of PD-L1 in the control in the absence of vemurafenib or (ii) the reduction of PD-L1 upon reintroduction of vemurafenib. Either way it seems that vemurafenib causes the decrease in PD-L1 because either removing it from the control is accompanied with the induction of this protein in the control cells or vemurafenib causing the decrease in expression of this protein upon addition of the drug to the treated cells.

In two of the tested cell lines (M229AR9 and M296), the MEKi trametinib, at 25 nmol/L concentration, caused a significant reduction in the expression of PD-L1 (Fig. 3C). Note that M229AR9 showed a similar pattern in the presence of vemurafenib, whereas M296, a NRAS-mutant cell line, responded only to MEKi. These results indicate that at least in some cases, expression of PD-L1 can be modulated by inhibitors of the MAPK pathway. On the contrary, in some of the tested cell lines, combination of the MEKi with INFγ caused a slight (M230) or significant induction (M238AR2) of PD-L1.

The effect of blocking the PI3K/AKT pathway on the expression of PD-L1 was also variable from cell line to cell line. The PI3Ki at 100 nmol/L concentration decreased expression of PD-L1 significantly in three cell lines. However, this effect was eliminated in the presence of INFγ, except for one cell line (Fig. 3D). These results indicate that, similar to the MAPK pathway, inhibition of the PI3K/AKT pathway can, in principle, also affect the expression of PD-L1. However, these effects are variable among different cell lines and perhaps are regulated by the specific signaling context of each cell line.

Interaction of PD-L1 with its receptor PD-1 on the surface of lymphocytes triggers the inhibitory signaling and decreases the activity of signaling pathways such as PI3K/AKT and MAPK (23, 24). On the contrary, vemurafenib can paradoxically activate the MAPK pathway in lymphocytes, which do not carry BRAF gene mutations. We investigated the possibility of restoring the MAPK pathway activity and enhancement of cytokine production by vemurafenib in PBMC cultures in conditions that expand T lymphocytes. These PBMC cultures were preconditioned to express PD-1 by mitogen activation leading to proliferation of activated T lymphocytes. The analysis of these primed cells indicated that out of all the live CD45⁺ T cells, 43.9% were CD4⁺ and 38.6% were CD8⁺ T lymphocytes (a total of 82.5%). The primed cells were then cultured in plates coated with a mixture of anti-CD3 and anti-CD28 and either Fc (control) or recombinant Fc-PD-L1, in the presence or absence of vemurafenib (Fig. 4A).In the presence of vemurafenib alone, p-MEK and p-ERK were induced because of paradoxical activation of the MAPK pathway in lymphocytes. Vemurafenib alone also induced higher levels of four tested cytokines (IL-1, IL-6, TNFα, and INFγ) as compared with the control (Fig. 4B–D). On the contrary, exposure of lymphocytes to PD-L1 for 24 hours caused a reduction in the activity of the MAPK pathway as indicated by the decreases in p-MEK and p-ERK (Fig. 4B and C). PD-L1 also caused lower levels of cytokines IL-6, TNFα, and INFγ but higher levels of IL-1 in the supernatant of lymphocytes. Meanwhile, in the PD-L1–exposed lymphocytes, vemurafenib restored the phosphorylation of p-MEK and p-ERK to levels similar to the untreated control sample (Fig. 4B). Except for INFγ, levels of all other tested cytokines were induced in the presence of vemurafenib and PD-L1 compared with PD1-L alone. However, induction of IL-6 and TNFα was not sufficient to restore these cytokines to the control levels (Fig. 4D). In unprimed resting PBMCs, the MAPK pathway is not active even in the presence of vemurafenib (Supplementary Fig. S1). Moreover, in unprimed resting PBMCs, expression of cytokines was at the background level even in the presence of vemurafenib (data not shown)

In patients, it is common that lymphocytes infiltrate tumors and therefore tumor cells are directly exposed to the cytokines produced by the TILs. Our results indicated that expression of PD-L1 was reduced in the BRAF-mutant M229AR9 melanoma cell line by vemurafenib and MEKi. The NRAS-mutated cell line M296, showed decrease in PD-L1 expression by MEKi but not vemurafenib (Fig. 3B). Meanwhile, vemurafenib induced the production of some cytokines by lymphocytes (Fig. 4C), which potentially can induce higher levels of PD-L1 expression in the adjacent tumor cells. However, MEKi does not cause induction of cytokines in lymphocytes (data not shown). To investigate the net effects of vemurafenib and MEKi on PD-L1 expression in a mixture of melanoma cells and lymphocytes, an in vitro coculture experiment was set up where these cells were treated with either one of these drugs. Similar to what was observed and interpreted before in Fig. 3B, vemurafenib treatment resulted in a significant reduction of PD-L1 expression in the M229AR9 cell line (ANOVA P<0.05), but not in the M296 cell line (Fig. 5B and C). M229AR9 alone or in coculture showed a 35% PD-L1 reduction upon vemurafenib treatment. MEKi treatment led to a significant PD-L1 reduction in both cell lines (ANOVA P<0.05).

Prior studies have reported several mechanisms by which PD-L1 can be expressed by cancer cells, including the well-documented inducible PD-L1 expression in response to IFNs, and reports on expression of PD-L1 downstream of oncogenic signaling through the MAPK or the PI3K/AKT pathways. In a study analyzing anaplastic large cell lymphoma and Hodgkin lymphoma, PD-L1 expression was shown to be controlled by MAPK pathway signaling (32). Furthermore, in a study using glioblastoma cell lines, Parsa and colleagues found that higher activity of AKT pathway, particularly upon the mutation and loss of PTEN, was associated with the increased PD-L1 expression (28). However, in our study we did not identify an association between the level of PD-L1 expression and alterations in MAPK signaling, mutations in PTEN or other activating mutations or gene amplifications in the PI3K/AKT pathway. The disagreement in the findings of these studies with our series may be due to the differences in signaling context of melanoma and other cancers. However, similar to what has been shown in another study in melanoma cell lines (29), we also found that, in the presence of a PI3Ki, expression of PD-L1 was decreased moderately only in some cell lines. Interestingly, in some cell lines, the reducing effect of PI3Ki on expression of PD-L1 was abrogated in the presence of INFγ.

In a recent study using melanoma cell lines, Jiang and colleagues found that the development of resistance to vemurafenib was associated with higher levels of c-Jun transcription factor which in turn caused higher PD-L1 expression (29). The apparent discrepancies with our results are related to the fact that resistance to vemurafenib can occur through a variety of mechanisms. These resistant mechanisms include activation of alternative signaling pathways such as PI3K/AKT (16) or reactivation of the MAPK pathway through a secondary mutation in NRAS (30), amplification of BRAF gene (33), activating mutations in MEK (34), overexpression of MAP3K8 (COT/Tp12), and continuous dimerization of truncated BRAF isoforms (35, 31). Resistance due to the activation of alternative signaling pathways is usually accompanied with high expression of receptor tyrosine kinases such as PDGFRβ, higher expression and dependency on c-jun, and phenotypic switch of melanoma cells toward a more mesenchymal phenotype (30; Titz and colleagues; sumbitted for publication). In our panel of cell lines, resistant sublines that were categorized under the activation of alternative signaling pathway-resistant mechanism also showed significantly higher expression of PD-L1 in comparison with their parental sensitive lines (M229AR9 and M238AR2). This is in agreement with the findings of Jiang and colleagues about the higher expression of c-Jun and PD-L1 in the resistant cell lines (29). However, this effect does not occur in all the mechanisms of acquired resistance to BRAFi because no or only a slight increase in expression of PD-L1 could be detected in the resistant cell lines with reactivation of the MAPK pathway.

The effects of targeted therapy drugs on the signaling and function of immune cells and the possibility of combining targeted therapy with different immunotherapy strategies have been the focal points of several studies (36, 37, 14). Our findings in the current study indicate that the effects of MAPK-targeted therapy agents on expression of PD-L1 are complex and variable among the cell lines. The dual effects of BRAFi such as vemurafenib on melanoma and immune cells are of particular interest (27). Aside from their inhibitory effect on mutated BRAF in cancer cells, these drugs can paradoxically activate the MAPK pathway in the cells with normal BRAF. This paradoxical effect can be particularly important in the context of PD-L1/PD-1 inhibitory therapy. In this study, we showed that the exposure of lymphocytes to PD-L1 caused a decrease in the activity of the MAPK pathway and reduced production of some of the cytokines. However, in such conditions vemurafenib restored the MAPK activity and induced positive effects on cytokine secretion.

In conclusion, our studies analyzing PD-L1 expression in a large panel of melanoma cell lines with or without exposure to MAPK inhibitors show that there is no straightforward cell-intrinsic explanation of PD-L1 expression as it had been previously hypothesized. The functional significance of constitutive and targeted therapy-modulated PD-L1 expression in vivo needs to be evaluated in comparison with the expression of PD-L1 induced by T-cell infiltration. It is possible that only when PD-L1 is expressed upon T-cell infiltration, in response to IFNs and other immune stimulating cytokines, will it be important as a biomarker for response to PD-1 or PD-L1 blocking antibodies. Constitutive PD-L1 expression or PD-L1 expression upon alterations of signaling pathways, in the absence of a T-cell infiltration, may not serve as a biomarker. Encouragingly, the MAPK paradoxical activation effects of BRAFi on T cells can serve to offset the inhibition of TCR signaling through the MAPK pathway induced by PD-L1 expression, providing hope for synergistic activity of BRAF inhibition and PD-1/PD-L1 blockade.

No potential conflicts of interest were disclosed.

Conception and design: M. Atefi, D. Foulad, T. Chodon, B. Comin-Anduix, A. Ribas

Development of methodology: M. Atefi, D. Foulad, T. Chodon, B. Comin-Anduix, A. Ribas

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): M. Atefi, E. Avramis, D.J.L. Wong, L. Robert, T. Chodon, A. Ribas

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): M. Atefi, E. Avramis, D. Foulad, M. Cerniglia, B.R. Titz, T.G. Graeber, B. Comin-Anduix, A. Ribas

Writing, review, and or revision of the manuscript: M. Atefi, E. Avramis, A. Lassen, L. Robert, D. Foulad, M. Cerniglia, T.G. Graeber, B. Comin-Anduix, A. Ribas

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): E. Avramis, A. Lassen, A. Ribas

Study supervision: M. Atefi, A. Ribas

The authors thank Jennifer Tsoi for her help in revising some of the graphs in this article.

This work was funded by NIH grant P01 CA168585, The Seaver Institute, the Dr. Robert Vigen memorial fund, the Ressler Family Foundation, the Wesley Coyle Memorial Fund, the Louise Belley and Richard Schnarr Fund, the Garcia-Corsini Family Fund, the Bila Alon Hacker Memorial Fund, the Fred L. Hartley Family Foundation, the Ruby Family Foundation, the Jonsson Cancer Center Foundation, and the Caltech-UCLA Joint Center for Translational Medicine (to Ribas. M. Atefi was partially supported by P50 CA086306 In Vivo Imaging in Cancer Biology (ICMIC) career development award. L. Robert was supported by the V Foundation-Gil Nickel Family Endowed Fellowship in Melanoma Research and a grant from the Spanish Society of Medical Oncology (SEOM) for Translational Research in Reference Centers. T.G. Graeber is the recipient of a Research Scholar Award from the American Cancer Society (RSG-12-257-01-TBE) and an Established Investigator Award from the Melanoma Research Alliance (20120279).

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.