Primary and Acquired Resistance to Immune Checkpoint Inhibitors in Metastatic Melanoma

Immune checkpoint inhibitors have revolutionized the treatment of advanced melanoma (1–5) and have significant clinical activity across an increasing range of many other solid malignancies, including non–small cell lung cancer (6, 7), renal cell carcinoma (8), head and neck cancer (9), Merkel cell carcinoma (10), and bladder cancer (11, 12). Understanding the biology behind response and resistance to immune checkpoint blockade is critical to further improving outcomes of patients with metastatic melanoma.

The first immune checkpoint to be clinically targeted, the cytotoxic T-lymphocyte antigen 4 (CTLA-4), is expressed on the surface of activated T cells and binds to its ligands, B7-1 and B7-2, on antigen-presenting cells (APC), resulting in the transmission of inhibitory signals to T cells. In patients with metastatic melanoma, phase III clinical trials of ipilimumab, a fully human IgG1 monoclonal antibody inhibiting CTLA-4, demonstrated a significant improvement in progression-free survival (PFS) and overall survival (OS) when compared with a gp100 vaccine (13) or dacarbazine chemotherapy (4).

Drugs targeting the programmed cell death receptor 1 (PD-1, PDCD1) showed a further increase in response rates, PFS (2), and OS (14–16) compared with anti–CTLA-4 blockade. PD-1 is also expressed on the surface of activated T cells and binds to the programmed cell death ligand 1 (PD-L1, CD274) to negatively regulate T-cell activation and differentiation. PD-L1 is constitutively expressed by T cells, macrophages, and dendritic cells (DC), as well as by some tumor cells including melanoma (17). Follow-up data from phase I clinical trials of the fully human IgG4 monoclonal antibody, nivolumab, showed a median OS of 17.3 months, with a 5 year OS rate of 34% (18). In a phase III study of nivolumab versus dacarbazine in patients with BRAF wild-type metastatic melanoma, the median OS was not reached for nivolumab at the most recent analysis, versus 11.2 months for dacarbazine [hazard ratio (HR), 0.43, P < 0.001], and the 1- and 2-year OS rates were 73% and 58%, respectively, for nivolumab (1, 14). Pembrolizumab, a humanized IgG4 monoclonal antibody against PD-1, also showed 1-, 2-, and 3-year OS rates of 67%, 50%, and 40%, respectively, in a phase I trial of ipilimumab-treated and ipilimumab-naïve patients with advanced melanoma (3). Furthermore, in a phase III trial of pembrolizumab versus ipilimumab, the 2-year OS rates were 55% versus 43%, respectively (5, 15).

More recently, combined anti–CTLA-4 and anti–PD-1 immunotherapies have shown improved response rates and clinical outcomes in comparison to ipilimumab monotherapy (the study was not powered to compare the two nivolumab treating arms: nivolumab plus ipilimumab and nivolumab alone). A phase III study showed an increase in the median PFS of patients treated with nivolumab and ipilimumab (11.5 months; HR, 0.42, P < 0.001) and nivolumab alone (6.9 months; HR, 0.57, P < 0.001) compared with ipilimumab alone (2.9 months; ref. 2). At a minimum follow-up of 28 months, the median OS had not been reached in the combination or nivolumab-alone groups and was 20 months for ipilimumab alone [HR: combination vs. ipilimumab, 0.55 (P < 0.0001); nivolumab vs. ipilimumab, 0.63 (P < 0.0001); ref. 16]. The two-year OS rates were 64%, 59%, and 45% in the combination, nivolumab, and ipilimumab groups, respectively (16).

The results of these clinical trials highlight the significant impact immunotherapies have had on the clinical management of patients with advanced-stage metastatic melanoma. However, although approximately 35% to 60% of patients have a RECIST response (10%–12% a complete response) to anti–PD-1-based immunotherapy (2, 14, 15), 40% to 65% have shown minimal or no RECIST response at the outset, and 43% of responders develop acquired resistance by 3 years (3). The underlying mechanisms driving these variations in response are not yet well understood. For an immunotherapy to elicit an efficient antitumor immune response, the cancer immune cycle must be initiated and the subsequent steps successfully completed. This involves efficient (i) antigen presentation and T-cell activation, (ii) T-cell trafficking and tumor infiltration, and (iii) T-cell killing activity within the tumor microenvironment (Fig. 1). Studies examining possible predictive biomarkers of response to immunotherapy have reported a higher density of preexisting cytotoxic T lymphocytes in tumor biopsies of patients who displayed a greater response to anti–PD-1/PD-L1 immunotherapy (19–21), and more significantly, an increased influx of T cells and PD-L1⁺ macrophages early during treatment (22). In this review, we discuss the different forms of immunotherapy resistance, the mechanisms by which tumors evade the immune system, and strategies to overcome or prevent resistance in the future.

Primary resistance to immune checkpoint blockade occurs in approximately 40% to 65% of patients with melanoma treated with anti–PD-1 based therapy (Fig. 2), depending on whether anti–PD-1 is given upfront or after progression on other therapies (2, 14, 15), and >70% of those treated with anti–CTLA-4 therapy (4, 13). This key unsolved clinical problem occurs when there is failure to induce an effective antitumor immune response at any of the three stages of the cancer immune cycle (Fig. 1). To date, the clinicopathologic factors that have been associated with primary resistance are elevated levels of baseline serum LDH (23), increased baseline tumor burden (24), lack of PD-L1 expression in baseline melanoma tissue samples (Fig. 3; ref. 25), lack of T-cell infiltration (Fig. 3; ref. 21), the absence of PD-1 T cells and PD-L1 macrophages in melanoma biopsies taken early during treatment (22), insufficient neoantigens and low mutational burden (26), the presence of an innate anti–PD-1 resistance signature (IPRES) transcriptional signature (27), or absence of an interferon signature (28). It is currently unknown whether these measures are surrogates for resistance or have a direct mechanistic role in preventing response. Here, we discuss the immune escape mechanisms that can occur at each stage of the cancer immune cycle (Fig. 1), thereby promoting both the growth and metastasis of tumors and resistance to immune checkpoint inhibitor therapies.

Upon encountering and engulfing foreign antigens, such as that of cancer cells, DCs migrate from the tumor to regional lymph nodes where they present the antigens on major histocompatibility complex (MHC) class I molecules to CD8⁺ T cells, resulting in activation of the latter. Barriers at this stage of the cancer immune cycle prevent optimal T-cell priming and activation, hence resulting in evasion of the immune system by the tumor (Table 1).

Some tumors lack sufficient antigen presentation by the immune system (29, 30) or do not present antigens that can be recognized as foreign (31, 32). The process of distinguishing tumor cells from normal cells depends on T-cell recognition of tumor-specific or tumor-associated antigens (TAA; ref. 33). Tumor immune evasion by TAA-negative cells was reported in patients with melanoma who relapsed after responding to peptide vaccinations (34). Recognition of tumor neoantigens by T cells has been associated with increased and durable response to immunotherapies and increased tumor regression, indicating a significant role for neoantigens in improving the outcome of patients with metastatic melanoma (26, 35). Circulating CD8⁺PD-1⁺ lymphocytes in peripheral blood of patients with melanoma can target patient-specific neoantigens, and neoantigen-specific T cells can in turn recognize autologous tumors (36). The immunogenicity of neoantigens can be predicted by combining exome sequencing and mass spectrometry data, thereby facilitating the identification of antigens that can be used to generate active T-cell responses (37).

Analysis of The Cancer Genome Atlas (TCGA) data from melanoma cases revealed that cutaneous melanoma displays a high mutational burden and UV signature (38). In addition to neoantigen recognition, a high mutational load was also found to correlate with clinical benefit to immune checkpoint blockade (26). Similarly, a positive correlation was observed between a higher mutational load and increased CD8⁺ T-cell infiltration (39). Furthermore, an increased mutational burden is associated with elevated PD-L1 expression in advanced melanoma (40). In a pooled analysis of 832 patients with melanoma, an increased PFS was observed in PD-L1–positive patients treated with nivolumab and ipilimumab combined immunotherapy, as well as in patients treated with nivolumab alone, compared with PD-L1–negative patients (2, 41). Similarly, PD-L1–positive patients treated with pembrolizumab had increased PFS, OS, and overall response rate (ORR), highlighting PD-L1 as a potential biomarker of response (42). As PD-L1 positivity is associated with improved response in patients with melanoma, a lack of PD-L1 correlates with primary resistance (Fig. 3). Nevertheless, some patients with PD-L1–positive tumors do not respond to PD-1 blockade, and conversely, some patients with PD-L1–negative tumors respond. For these reasons, intratumoral PD-L1 expression is a suboptimal predictive biomarker (2, 41). Together, the aforementioned data indicate that PD-L1 expression is a possible surrogate for lack of immunogenicity, as well as other failures further down the immune cycle.

In order to efficiently activate T cells, DCs must undergo a process called maturation, where they increase their capacity to stimulate T cells by displaying increased expression of various costimulatory molecules required for T-cell activation, such as MHC class I/II, CD80, CD86, and CD40 (43). The density of DCs strongly correlates with activated T cells in melanoma (44). The function of DCs can be impaired via multiple pathways. Interleukin (IL) 37b, a protein with a critical role in the inhibition of the innate immune response, suppresses DC maturation and function by decreasing CD80 and CD86 expression via the ERK/S6K/NF-κB signaling pathways (45). Furthermore, DC maturation and tumor infiltration increased significantly in melanoma following the inhibition of signal transducer and activator of transcription 3 (STAT3), a transcription factor that is required for tumor growth and metastasis (46). STAT3 is also involved in the cross-talk between melanoma cells and immune cells, resulting in the induction of other immunosuppressive factors such as the vascular endothelial growth factor (VEGF), IL10, regulatory T cells (Treg), and transforming growth factor β (TGF-β), all of which have inhibitory effects on DC maturation (47–49).

Tumors can use a number of immune evasive mechanisms to prevent T-cell trafficking and infiltration into tumors. Assuming that the T cells were successfully activated in the previous steps, disruption during this stage is a likely cause for lack of response to immunotherapy.

The differential expression of chemokine receptors is required for effective T-cell homing and recruitment in cancer. In particular, CXCR3 has been identified as an important chemokine receptor critical for T-cell infiltration. In mouse melanoma models, CXCR3 was highly expressed on a number of T-cell subsets, and transfection with its ligand, CXCL9, resulted in a significant increase in both CD4⁺ and CD8⁺ T-cell infiltration (50). Similarly, human melanoma samples with high CD8⁺ T-cell expression were associated with increased levels of CXCL9 and CXCL10 (51). Interferon gamma (IFN-γ, IFNG) has previously been shown to mediate trafficking of Tregs, T helper cells, and cytotoxic T cells (52, 53). STAT3 inhibits CXCL10 production by tumor-associated myeloid cells and T-cell recruitment into tumors by downregulating IFN-γ production by CD8⁺ T cells (54). Conversely, Stat3 ablation increases CXCR3 expression on CD8⁺ T cells, allowing T-cell tumor infiltration (54).

Epigenetic alterations including DNA methylation and histone modifications have also been identified as important mechanisms of chemokine repression and tumor progression. Epigenetic modifications are heritable modifications to DNA that result in changes to the gene expression profiles of tumor cells, thereby allowing them to evade the immune system (55). Epigenetic silencing resulted in the suppression of CXCL9 and CXCL10 in ovarian cancer, and treatment with epigenetic modulators increased chemokine expression and T-cell infiltration into tumors (56). Increased expression of chemokines, T-cell recruitment, and tumor regression was also observed in lung cancer cell lines and mice treated with the histone deacetylase (HDAC) inhibitor romidepsin (57).

T-cell trafficking through the tumor and lymph nodes is controlled by a number of endothelial signals, regulating T-cell homing, adhesion, and migration (58). The interaction between endothelin 1 (ET-1, EDN1) and the endothelin A receptor (ETAR, EDNRA) promotes tumorigenesis through various pathways including cell proliferation, invasion, angiogenesis, bone remodeling, and inhibition of apoptosis (59). The endothelin B receptor (ETBR, EDNRB) counterregulates ET-1/ETAR activity via the increased production of nitric oxide, activation of apoptotic pathways, and clearance of ET-1 (60). The endothelin system has been implicated in the pathogenesis of a number of cancers, including ovarian cancer, prostate cancer, and colon cancer. Interestingly, ETBR is upregulated in melanoma and has also been proposed as a marker of melanoma progression, suggesting a role for ETBR in melanoma tumorigenesis (61). ETBR inhibition in 10 human melanoma cell lines using the ETBR antagonist BQ788 resulted in an increase in apoptosis and cell death, as well as an increase in angiogenesis in the tumors (62). In human ovarian cancers, ETBR was found to correlate with an absence of tumor-infiltrating lymphocytes (TIL) as well as decreased patient survival time. Administration of BQ788 increased T-cell homing to tumors and improved the efficacy of immunotherapy (58), highlighting ETBR as a potential immune escape mechanism and future target in patients who fail to respond to immunotherapy.

Increased levels of the proangiogenic factor VEGF in plasma and tissue samples have also been associated with the growth and progression of melanoma (63, 64). VEGF-A downregulates T-cell adhesion to the endothelium via the suppression of intercellular adhesion molecule 1 (ICAM-1) and vascular cell adhesion molecule 1 (VCAM-1) on endothelial cells (65). Increased expression of VEGF in tumor was associated with an absence of intratumoral TILs and a shorter OS time in patients with ovarian carcinoma (66). Inhibition of VEGF resulted in an increase in T-cell infiltration into B16 melanoma tumors via the upregulation of CXCL10 and CXCL11 chemokines (67). Additionally, VEGF-A, along with prostaglandin E2 (PGE2) and IL10, upregulates the Fas ligand, resulting in CD8⁺ T-cell death, and subsequent inhibition of VEGF and PGE2 increased CD8⁺ T-cell infiltration (68). Corresponding with these data, in melanoma tumor biopsies, increased levels of VEGFA were observed in nonresponders to anti–CTLA-4 and anti–PD-1 immunotherapies in comparison to responders (19).

Primary resistance to immunotherapy also occurs when T cells become successfully activated and infiltrate the tumor; however, their function is hindered by the presence of immunosuppressive molecules within the tumor microenvironment (31).

Primary resistance can be driven by the constitutive expression of PD-L1 through oncogenic signaling (69, 70). The increased expression of PD-L1 by cells in the tumor microenvironment results in decreased function of cytotoxic T cells and apoptosis, hence providing an immune escape mechanism for tumor cells.

Several studies have revealed a correlation between loss of the phosphatase and tensin homolog (PTEN) in cancer and the upregulation of PD-L1, implicating PD-L1 in tumor immune evasion. PTEN is a tumor suppressor that negatively regulates the P13K/AKT pathway. This pathway is responsible for the regulation of cellular processes such as proliferation and survival. The loss of PTEN and activation of the P13K/AKT pathway in human glioma cell lines resulted in an increase in posttranscriptional CD274 expression (71). PD-L1 expression was also upregulated in lung squamous cell carcinoma following the simultaneous depletion of Pten and Lkb1 [also known as Stk11 (serine–threonine kinase 11); ref. 72]. In melanoma, loss of PTEN led to a decrease in T-cell trafficking, infiltration, and T-cell activity (73). However, silencing PTEN did not significantly alter the expression of PD-L1 in melanoma cell lines in vitro or in xenograft models in vivo, indicating that PD-L1 regulation may not be the principal mechanism of immune resistance resulting from a loss of PTEN (73).

Other mechanisms that have been shown to have a role in the constitutive upregulation of PD-L1 include the transcription factor interferon regulatory factor 1 (IRF-1) and mutations in the epidermal growth factor receptor (EGFR). IRF-1 is responsible for the regulation of cell proliferation, apoptosis, and immunity (74). The knockdown of IRF-1 using siRNA resulted in the decrease in transcription and translation of PD-L1 in a lung carcinoma cell line (75). Similarly, activation of the EGFR pathway resulted in the increased expression of PD-L1 in lung cancer cell lines (76, 77) and tissue (76). Increased expression of markers of T-cell exhaustion, such as PD-1 and FOXP3, was also observed in the tumor microenvironment (76). PD-1 blockade increased cytotoxic T-cell numbers as well as effector T-cell function (76), highlighting the role of the PD-1/PD-L1 axis in immune evasion and its manipulation as a therapeutic strategy.

Another molecule proposed to play a critical role in the negative regulation of T-cell function is indoleamine 2,3-dioxygenase (IDO, IDO1). IDO is expressed in a wide range of human cancers and is the rate-limiting enzyme responsible for the degradation of tryptophan into kynurenine (78, 79). T lymphocytes undergo arrest in response to this tryptophan depletion, resulting in the suppression of T-cell proliferation and activity (80). To understand the mechanism of immunosuppression induced by IDO, Holmgaard and colleagues (81) developed a B16 melanoma tumor model overexpressing IDO and revealed a correlation between IDO expression and increased tumor-infiltrating myeloid-derived suppressor cells (MDSC), as well as CD4⁺FOXP3⁺ Tregs. This association demonstrated that IDO suppresses T-cell activity through the recruitment and activation of MDSCs in a Treg-dependent fashion (81). Systemic inhibition of IDO in mice using a tryptophan analogue, 1-methyl-L-tryptophan (1MT), reduced tumor progression in a T-cell–dependent manner (79). Similarly, administration of 1MT in combination with anti–CTLA-4 immunotherapy in B16F10 mouse melanoma models resulted in a significant delay in tumor growth and an increase in OS (78). These findings provide a strong rationale for targeting IDO to improve the efficacy of immunotherapies in patients with melanoma.

The upregulation of FOXP3-expressing Tregs has been observed in a number of melanoma studies, revealing a possible role for Tregs in melanoma tumorigenesis (82–84). Tregs promote tumor growth by inhibiting the activity of T-cell subsets, either through direct cell-to-cell contact or indirectly through the secretion of anti-inflammatory cytokines, such as IL10 and TGF-β (85, 86). The presence of CD4⁺CD25⁺ FOXP3 Tregs was observed amongst TILs in metastatic melanoma (87) and the transfer of CD25 (IL2RA)-depleted splenic T cells into B16 mouse melanoma models resulted in the suppression of tumor growth in vivo (82). Populations of tumor-infiltrating Tregs also significantly correlated with increased tumor growth in B16BL6 mice (88). Furthermore, a decrease in FOXP3⁺ Tregs significantly correlated with increased tumor control and survival in patients with melanoma treated with ipilimumab (89), highlighting the immunosuppressive function of Tregs in melanoma.

Elevated levels of extracellular adenosine and CD73 (NT5E) have also been implicated in immune suppression and tumor progression. Adenosine is produced via the conversion of extracellular ATP by the ectoenzymes CD39 (ENTPD1) and CD73 and binds to the adenosine A2A receptor (A2AR, ADORA2A) to inhibit effector T-cell function (90). Increased CD73 expression correlated with advanced-stage disease in melanoma (91), and an upregulation in CD73 was observed in patients who progressed following treatment with anti–PD-1 immunotherapy (92). A2AR inhibition increased CD8⁺ T-cell infiltration and significantly reduced tumor growth in mouse melanoma models (91), suggesting a role for the CD73/adenosine axis in promoting immune escape.

The expression of IPRES has recently been identified as a mechanism of primary resistance to immunotherapy. Transcriptomal analyses of responding and nonresponding pretreatment melanoma biopsies from patients treated with anti–PD-1 immunotherapy revealed the coenrichment of genes associated with mesenchymal transition, wound healing, and angiogenesis in nonresponding patient samples (27). This was observed not only in metastatic melanoma but also in other major cancer types such as pancreatic cancer (27).

Mutations in the janus kinases 1 and 2 (JAK1/2) have also been shown to be involved in primary resistance to anti–PD-1 immunotherapy. JAK1/2 loss-of-function mutations identified in one of 23 melanoma tumor biopsies, and two of 48 human melanoma cell lines via whole-exome sequencing, resulted in a lack of PD-L1 expression due to an inability to respond to IFN-γ signaling (93). Furthermore, the recent development of a two cell type-CRISPR (2CT-CRISPR) screening assay revealed an important role for apelin receptor (APLNR) loss-of-function mutations in disturbing effector T-cell function (94). Retroviral overexpression of APLNR correlated with an increase in JAK1 as well as tumor sensitivity to effector T-cell function (94). Conversely, APLNR-knockout cells demonstrated decreased activation of the JAK/STAT pathway following IFN-γ treatment, and Aplnr knockout in mouse melanoma in vivo resulted in a decrease in the efficacy of anti–CTLA-4 immunotherapy (94). These findings provide a strong rationale for further investigating APLNR as a potential target to prevent immune evasion by tumors.

Recent studies have highlighted a possible role for the gut microbiome in patient response to immunotherapy. Dysbiosis (an imbalance of the microbiota) involves decreases in the diversity and stability of the microbiome, thereby promoting tumorigenesis (95). Sequencing of the oral and gut microbiome of patients with metastatic melanoma revealed a correlation between higher gut microbiome diversity and response to anti–PD-1 monotherapy (96). Responders also had a significantly different gut microbiome composition in comparison with nonresponders, and this correlated with differences in PFS (96). The increased abundance of specific bacteria in the gut microbiome also correlated with a higher CD8⁺ T-cell density in responders to anti–PD-1 immunotherapy (96). Similarly, the composition of the baseline gut microbiome in patients with metastatic melanoma was associated with response to ipilimumab, and improved PFS and OS were associated with specific groups of bacteria such as Faecalibacterium and other Firmicutes (97). Additionally, a significant decrease in TILs and lack of response to CTLA-4 blockade was observed in tumors of mice housed in germ-free conditions (98). The anticancer therapeutic effects of the anti–CTLA-4 antibody were restored upon oral feeding of the germ-free mice with Bacteroides fragilis (B. fragilis), a Bacteroides isolate, as well as with the adoptive transfer of B. fragilis–specific memory T cells (98). The mechanisms through which the gut microbiome influences the immune response are currently being investigated.

Acquired resistance occurs in patients who relapse after exhibiting an initial response to immunotherapy (Fig. 4). Currently, little is understood about the mechanisms that give rise to acquired resistance, and many are likely to be similar to those underlying primary resistance (Table 1).

Acquired resistance to immunotherapy can develop when there is Darwinian selection of subpopulations of tumor cells with genetic and epigenetic traits that allow them to evade the immune system (32). An example of one such trait is beta-2-microglobulin (B2M), a component of MHC class I molecules that is necessary for their functional expression. The loss of B2M expression was reported in melanoma cell lines from five patients who had been treated with immunotherapy and cytokine–gene therapy (99). This resulted in a loss of MHC class I expression and, therefore, a subsequent decrease in recognition by CD8⁺ T cells. Archival tissues taken prior to immunotherapy from three of these patients were B2M positive, suggesting loss of B2M expression as a mechanism of acquired resistance (99). Similarly, the loss of B2M has been observed in sequential lesions obtained from a patient with metastatic melanoma following immunotherapy treatment with DCs transfected with autologous tumor mRNA (100).

JAK1/2 mutations have also recently been identified as genetic markers of acquired resistance to immunotherapy in melanoma. These mutations in tumor cells lead to decreased sensitivity to IFN-γ, ultimately preventing IFN-γ–induced cell growth arrest (101). Upon exposure to IFN-γ (produced by activated T cells), JAK1/2 become activated and subsequently phosphorylate a tyrosine residue present on STATs (102). This JAK/STAT signaling pathway is responsible for cell proliferation, differentiation, cell migration, and apoptosis (102). However, IFN-γ also results in the upregulation of PD-L1 on tumor cells, thus inactivating antitumor T cells (70). Loss-of-function mutations in the genes encoding JAK1 or JAK2 were found in relapsed tumors in two of four patients following whole-exome sequencing of baseline and progression biopsies; all patients had an objective response to treatment with pembrolizumab and then progressed after a median of 1.8 years (101). The anti–PD-1-resistant cells harboring JAK mutations were derived from cells clonally selected from the baseline tumor. These findings demonstrate the role of the JAK/STAT pathway in promoting acquired resistance to immunotherapy.

In addition, acquired resistance can also occur on the level of the individual cells, whereby tumor cells alter their gene expression in response to immune molecules within the tumor microenvironment (32). For example, PD-L1 can be upregulated by tumor cells in response to immune cytokines, such as IFN-γ released by T cells, hence limiting T-cell function (70), and can occur in both primary and acquired resistance (32, 103).

Other immune checkpoint markers such as lymphocyte activation gene 3 (LAG-3) and T-cell immunoglobulin and mucin domain 3 (TIM-3, HAVCR2) have also been revealed to interfere with the activity of T cells (70, 104), resulting in acquired resistance to immunotherapy. In a recent study, TIM-3 upregulation was observed in patients who developed adaptive immune resistance to anti–PD-1 immunotherapy (105). Furthermore, TIM-3 blockade in mice resulted in a significant increase in survival time, as well as increased production of IFN-γ and proliferation of CD8⁺ T cells (105). LAG-3 is also overexpressed in PD-L1–positive melanoma, suggesting LAG-3 upregulation as a potential immune evasion mechanism (106).

From the above, it is clear that there exist multiple immune evasive mechanisms that can be utilized by tumors at each of the different stages of the cancer immune cycle that may induce either primary or acquired resistance to immunotherapy. Determining the specific mechanisms underlying resistance to immunotherapy in these patients is a crucial step toward effective treatment and ultimately producing durable responses for them.

The immune escape mechanisms discussed above do not act in isolation. Together, the overlap between various signaling pathways and the interactions between several of the immunosuppressive molecules leads to resistance. It is likely that combinations of therapy will be more effective than single-agent therapies for a given patient. However, the challenge remains to determine which of these combinations are most effective and in which patient given interpatient heterogeneity. Currently, multiple clinical trials are underway examining the activity and toxicity of combined immunotherapies, particularly using an anti–PD-1 drug in combination with an agent targeting a complimentary part of the immune system (Supplementary Table S1).

Dual blockade of PD-1 and LAG-3 in mouse cancer models resulted in the regression of tumors in most mice, as well as an increased survival rate (107). In a recent study, all mice that were treated with a triple combination of LAG-3 blockade, PD-1 blockade, and poxvirus-based immunotherapy demonstrated complete tumor regression (108). Phase I clinical trials involving LAG-3 blockade with and without PD-1 blockade in solid tumors are currently underway, and in patients who had progressed on anti–PD-1 monotherapy, a response rate of 16% (any tumor reduction in 45%) was observed in those patients whose tumors expressed LAG-3 (109). Similarly, in mouse models, combined TIM-3 and PD-L1 blockade significantly reduced tumor growth in comparison with single-agent immunotherapy (110). Furthermore, dual blockade of TIM-3 and PD-L1 increased the ability of CD8⁺ T cells to produce IFN-γ, thereby restoring their function (110). Consistent with this, anti–TIM-3 and anti–PD-1/anti–CTLA-4 combined immunotherapy resulted in a significant decrease in the tumor sizes of multiple cancers (111).

Phase I/II clinical trials testing the combination of an IDO inhibitor with PD-1/PD-L1 inhibitors in metastatic melanoma are also currently underway. These include the administration of nivolumab and a PD-L1/IDO peptide vaccine (NCT03047928), and indoximod (IDO inhibitor) in combination with CTLA-4 or PD-1 inhibitors (NCT02073123). Furthermore, the combination of epacadostat (an IDO1 inhibitor) plus pembrolizumab was generally well tolerated and correlated with improved response in various cancers (112–114), leading to the initiation of phase III studies such as that in patients with treatment-naïve advanced melanoma (NCT02752074).

Many studies have also demonstrated the increased efficacy of combined radiotherapy and immune checkpoint blockade. Radiotherapy activates the inflammatory pathways and induces DNA damage, resulting in the release of antigens from irradiated cells and an increase in tumor sensitization to T-cell immune responses (115). Radiotherapy in patients with advanced melanoma who had progressed following treatment with ipilimumab led to an abscopal effect in some patients, whereby irradiation of the primary tumor induced regression in other nonirradiated metastases, which correlated with longer OS (116). Significant tumor regression was also observed in patients with metastatic melanoma who received combined radiotherapy and anti–CTLA-4 immunotherapy, and experiments in mice showed that a combination of anti–CTLA-4 immunotherapy, anti–PD-L1 immunotherapy, and radiotherapy was required to achieve optimum response (117).

The combination of epigenetic modulators and immune checkpoint blockade has recently been shown to be more effective than single-agent immunotherapy in patients with melanoma. The inhibition of HDAC using LBH589 in combination with PD-1 blockade in B16F10 mouse melanoma models resulted in delayed tumor growth and increased survival compared with the control group and PD-1 blockade alone (118). Similarly, a reduction in tumor size and Tregs was observed in B16F10 mouse melanoma models treated with a combination of anti–CTLA-4 immunotherapy and epigenetic modulation of trimethylation of lysine 27 on histone H3 (H3K27me3) compared with CTLA-4 blockade alone (119). The efficacy of combined epigenetic and immune checkpoint therapies is currently being tested in various human cancers (Supplementary Table S1).

These findings provide a strong rationale for the development of combinations of different treatment strategies in clinical trials as an effective means of melanoma treatment.

Checkpoint inhibitor immunotherapy has revolutionized the treatment of multiple cancer types. However, responses in patients with metastatic melanoma are diverse, with many patients displaying primary or acquired resistance. An important ongoing, major unmet clinical need remains to identify predictors and causes of this resistance and strategies to overcome them. A key element of effective immunotherapy is identifying the various mechanisms by which the tumors evade the immune system on an individual basis. Improving our understanding of these mechanisms and determining which immune markers to target in each patient will then allow for the administration of the most appropriate form of therapy to achieve optimal response and improve the overall outcomes of patients with metastatic melanoma.

G.V. Long reports receiving speakers bureau honoraria from Bristol-Myers Squibb, Incyte, Merck, Novartis, and Roche, and is a consultant/advisory board member for Amgen, Array, Bristol-Myers Squibb, Incyte, Merck, Novartis, Pierre Fabre, and Roche. No potential conflicts of interest were disclosed by the other authors.

This work was supported by Melanoma Institute Australia, the New South Wales Ministry of Health, NSW Health Pathology, National Health and Medical Research Council of Australia (NHMRC), and Cancer Institute NSW. J.S. Wilmott, R.A. Scolyer, and G.V. Long are supported by NHMRC Fellowships. J.S. Wilmott is also supported by a CINSW Fellowship. G.V. Long is also supported by the University of Sydney Medical Foundation. T.N. Gide is supported by The University of Sydney and Melanoma Institute Australia Scholarships.

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.