JAK Mutations as Escape Mechanisms to Anti–PD-1 Therapy

Immune-targeted therapies blocking the PD-1/PD-L1 pathway are currently revolutionizing oncology for two main reasons. First, the widths of their spectrum of activity in more than 20 different cancer types demonstrate the validity of targeting immune cells to treat cancers. Second, as opposed to tumor-targeted therapies, they generate durable tumor responses and disease stabilizations which translate into overall survival benefits for some patients when compared with standard of care.

This paradigm shift from tumor-targeted to immune-targeted therapies has been inaugurated in melanoma, first with anti–CTLA-4 antibodies, then with anti–PD-1/PD-L1 antibodies. These drugs have been designed as antagonistic antibodies to block T-cell coinhibitory receptors. However, their precise in vivo mechanism of action remains unclear.

A feature often reported in tumor lesions responding to these treatments is a high level of CD8⁺ T-cell infiltrates. Therefore, it is believed that anti–CTLA-4 and anti–PD-1/PD-L1 efficacy relies on the development of an adaptive tumor-specific CD8⁺ T-cell response. The corollary assumption is that in order to generate tumor responses, T cells need to recognize cognate tumor-specific epitopes presented by MHC-I molecules on melanoma cells.

The activation of antitumor CD8⁺ T cells via MHC-I molecules is usually followed by T-cell IFNγ release and upregulation of membrane PD-1 receptor. Upon exposure to IFNγ, cancer cells, like other self-cells, can enter into cell-cycle arrest and upregulate both MHC-I and PD-L1 to their membrane, therefore preventing, through the PD-1/PD-L1 interaction, a T-cell–mediated cell cytotoxicity. Therefore, it is believed that anti–PD-1/PD-L1 therapy works by blocking this negative feedback loop and allows for T cell–mediated cancer cell death (Fig. 1A).

This rationale is supported by results showing a correlation between MHC-I binding neoepitope signatures and survival in patients with metastatic melanoma or non–small cell lung cancer treated with anti–CTLA-4 and anti–PD-1, respectively (1, 2). This rationale is also supported by the correlation between the level of tumor-infiltrating CD8⁺ T cells, the level of PD-L1 expression in the tumor, and the efficacy of anti–PD-1 therapy in metastatic melanoma (3).

However, the majority of patients with melanoma do not respond to anti–PD-1 monotherapy, including some patients with high levels of PD-L1 expression and CD8⁺ T-cell infiltrates in their tumor. Moreover, although tumor responses upon anti–PD-1 therapy are long lasting, some patients can eventually relapse from their disease. The mechanisms behind these primary and secondary refractory diseases to anti–PD-1 therapy remain mostly unknown.

The team of Antoni Ribas and colleagues has recently shown that an escape mechanism found in secondary refractory disease can be attributed to mutations of genes involved in the IFNγ pathway, either via loss-of-function mutations in the genes encoding for JAK1 and JAK2 (with concurrent deletion of the wild-type allele) or via truncating mutations of the MHC-I sub-unit beta-2-microglobulin (B2M; ref. 4). In this issue of Cancer Discovery, the same team reports that loss-of-function JAK mutations can also be found in patients with primary resistance to anti–PD-1 therapy (5).

Inactivating JAK1 and JAK2 mutations might in theory impair the downstream signaling of IFNγ receptors and therefore the ability of IFNγ to induce a cell-cycle arrest of tumor cells and to upregulate PD-L1 and MHC-I expression on their outer membrane (Fig. 1B). Shin and colleagues show here in a melanoma cell line generated from a patient with primary resistance to anti–PD-1 that a missense mutation of JAK1 associated with an amplification of its gene locus resulted in a 4:1 mutant:wild-type allele ratio. This high variant allele frequency seemed to correlate with a loss of function of JAK signaling because of an inability of these melanoma cells to upregulate PD-L1 membrane expression upon IFNγ exposure. Other mutations of JAK1/2 and IFNγ receptors (IFNGR) were found in both responders and nonresponders, but these were associated with a nonmutated allele counterpart which should in theory compensate for the downstream IFNγ signaling. In their series of melanoma cell lines, they showed that monoallelic loss-of-function JAK1 or JAK2 mutations are sufficient to impair PD-L1 upregulation upon IFNγ exposure when these cancer cell lines have a loss of the wild-type allele. These JAK1 and JAK2 mutations with loss of function were also found in other types of cancers, including one case of mismatch repair–deficient (MMRD) colon cancer with primary resistance to anti–PD-1 therapy.

These results are significant because they suggest that JAK mutations could be used as a genetic biomarker to identify patients who might not benefit from an anti–PD-1/PD-L1 therapy. However, the overall incidence of primary inactivating JAK1/2 mutations reported here remains low at around 5%: 1/23 patients with melanoma; 2/48 melanoma cell lines; 1/16 MMRD colon cancers. But, as suggested by the authors, epigenetic silencing of the JAK/STAT pathway could also contribute to anti–PD-1 resistance beyond somatic gene loss of function. Interestingly, patients bearing tumors with inactivating mutations of JAK1/2 or B2M could benefit from other types of immunotherapies, such as bispecific T-cell engaging antibodies or CAR-T cells, which should be active in patients without tumor MHC-I expression or preexisting antitumor immunity and especially in the absence of PD-L1 tumor expression. Moreover, pattern recognition receptor agonists such as Toll-like receptors and STING agonists could directly activate proinflammatory pathways in a JAK-independent fashion and could therefore overcome the above-mentioned immune escape mechanisms (6).

The link between JAK mutation and resistance to IFNγ-induced PD-L1 upregulation has been nicely demonstrated by Shin and colleagues in patients with primary anti–PD-1 refractory disease. However, it is not clear here if these inactivating mutations are indeed associated with an inhibition of cancer cell proliferation or MHC-I upregulation on tumor cells upon IFNγ exposure, as demonstrated before in secondary refractory tumors (4).

Also, it is not clear if MHC-I expression is an absolute requirement for anti–PD-1 efficacy. Indeed, about 70% of cases of Hodgkin disease present with B2M-inactivating mutations and therefore no functional MHC-I expression (7). However, these patients show dramatic response rates to anti–PD-1 therapy (8). One hypothesis could be that CD4⁺ T cells could also contribute to the antitumor T-cell response via MHC-II molecules. Interestingly, a recent report has demonstrated that MHC-II–positive melanomas are indeed showing good sensitivity to anti–PD-1/PD-L1 therapy (9).

The absence of inflammation in melanoma tumors does not seem to be related to an absence of immunogenic epitopes (10). Therefore, beyond MHC molecule expression and IFNγ signaling, other factors contribute to the overall immunogenicity of cancer cells and sensitivity to immune checkpoint–targeted therapies. The ongoing clinical trials testing numerous anti–PD-1/PD-L1 combinations shall hopefully identify synergistic combinations which will circumvent the predominant escape mechanisms to single immune checkpoint blockade therapy.

A. Marabelle is a consultant/advisory board member for AstraZeneca, BMS, Merck, Merck Serono, Pfizer, and Roche/Genentech. J.-C. Soria is a consultant/advisory board member for AstraZeneca, BMS, Merck, Pfizer/Merck Serono, and Roche. No potential conflicts of interest were disclosed by the other authors.

This work has benefited from the support of the grant SIRIC SOCRATE from the Institut National du Cancer (INCa).