Can Biomarkers Guide Oncolytic Virus Immunotherapy?

In this issue of Clinical Cancer Research, Nguyen and colleagues identified defects in the IFNγ-JAK-STAT pathway as biomarkers of response to oncolytic virus (OV) therapy (1). While dysregulation of IFNγ signaling have been reported to confer resistance to immune checkpoint inhibitors in this study, tumor cells derived from an anti-PD-1–resistant melanoma patient were found to have developed a homozygous loss of JAK2. These cells, however, appeared to be more sensitive to infection with an oncolytic vesicular stomatitis virus (VSV) and a herpes simplex virus (HSV-1). The investigators also showed that loss-of-function JAK2 mutations were associated with increased sensitivity to OVs, and this effect was abrogated by replacement of normal JAK2 function. Furthermore, the demonstrated that treatment with ruxolitinib, a selective JAK 1/2 inhibitor, also increased tumor cell sensitivity to OV infection. A review of The Cancer Genome Atlas melanoma data found that 11% of melanomas harbor alterations in IFNγ signaling genes and the authors suggested that patients with such defects might be more responsive to OV therapy.

The use of predictive biomarkers to guide clinical immunotherapy is considered a high priority for cancer research. Indeed, considerable progress has been made in patients treated with immune checkpoint blockade where tumor mutation burden, presence of tumor-infiltrating lymphocytes, induction of an IFNγ gene signature, and in some patient, high programmed cell death ligand 1 (PD-L1) expression, have been associated with improved therapeutic responses. The magnitude of antitumor T-cell responses and associated downstream IFNγ signaling have been cited as the major drivers of immune checkpoint inhibitor activity (2). In contrast, predictive biomarkers of OV activity have not been identified. Yet, the data presented by Nguyen and colleagues (1) add support to the emerging concept that elements of the innate immune sensing pathway, including factors involved in the antiviral machinery, may be important determinants of OV activity.

The ability of OVs to induce therapeutic responses depends on the balance between antiviral immunity which can prematurely clear virus and the antitumor immune responses induced by OV-mediated immunogenic cell death and cross-presentation of tumor-associated antigens. When the antiviral machinery is ineffective at rapidly clearing an intracellular viral infection, the virus can replicate and induce cell lysis. This explains, in part, why tumor cells are more susceptible to OV-mediated cell death and why infection of normal cells can resist OV infection. In support of the findings by Nguyen and colleagues (1) there have been previous reports showing that defects in tumor interferon signaling can promote more effective OV replication and cell lysis (3). Furthermore, loss of other elements of the intracellular innate immune sensing, such as defective cGAS-STING function, has also been associated with increased killing of melanoma cells by oncolytic HSV-1 (4). Because HSV-1 is a DNA virus and the cGAS-STING apparatus is designed to detect foreign DNA, it is tempting to hypothesize that STING loss may be a predictive biomarker for oncolytic DNA viruses. In contrast RNA viruses are usually detected by Toll-like receptors (TLR) and retinoic acid inducible gene I (RIG-I)-like receptors, and thus, we would predict that the status of these receptors may be biomarkers for RNA OVs. This hypothesis awaits experimental validation. Collectively, however, these data suggest that different OVs may utilize distinct pathways to avoid immune clearance. In the study by Nguyen and colleagues (1), loss of JAK2 expression was associated with a 7-fold improvement in lytic activity by HSV-1 but a 22-fold improvement by VSV.

Together, the emerging picture is that the genomic landscape of the primary tumor may be useful in predicting responses to individual OVs. For example, genomic screening of individual tumor biopsy specimens could be used to evaluate specific expression of innate sensing and antiviral machinery elements. As shown in Fig. 1 for JAK1 and JAK2 status, when these are intact patients could be treated with immune checkpoint blockade. However, if these are deficient then patients may be better treated with an OV. Further assessment of additional factors, such as STING, RIG-I, and TLR, could further delineate whether a DNA or RNA OV might be optimal. Another implication of these findings is that combination studies of OVs and immune checkpoint blockade might not be biologically rational because responses to each modality appear to be inversely associated with JAK expression. Thus, alternative combinations in the JAK-deficient setting may be more interesting, or sequential treatment using biomarker detection to guide therapy may be more appropriate.

The work by Nguyen and colleagues (1) also pointed out an interesting opportunity to pharmacologically inhibit JAK expression with ruxolitinib to enhance OV activity. The use of targeted therapy as a mechanism to enhance OV immunotherapy has not yet been clinically tested but is also supported by additional preclinical data. In a murine PD-1–resistant melanoma model, oncolytic HSV-1 therapeutic responses were significantly enhanced in mice treated with MEK I inhibition prior to OV injection (5). In this study, the combination of OV and MEK inhibition also drove local expression of PD-1 and PD-L1, which resulted in highly effective triple combination therapy. Further clinical studies are needed to further explore combinations of OVs and targeted therapy. These studies should also attempt to understand how treatment may alter the sensitivity and/or resistance to immune checkpoint inhibitors.

OVs represent a diverse array of therapeutic agents that have met with limited clinical success, to date. These agents remain attractive for cancer therapy given the specificity of replication in tumor cells and their highly favorable safety profile. However, each virus is unique and is based on a different set of nucleic acid composition, genetic modifications, and arming transgenes. These may influence their biologic interactions in different tumors and under differential gene expression status within infected cells. Further studies are needed to define how changes in specific innate sensing and antiviral machinery elements influence the ability of specific OVs to replicate in individual tumor cells and how these changes influence host antiviral and antitumor immunity. The article by Nguyen and colleagues (1) represents an important first step and while they show a convincing relationship between loss of function in JAK and improved therapeutic potential for OVs, only 11% of patients with melanoma have a baseline JAK mutation. Thus, a more complete survey of all innate sensing elements will likely yield a higher number of patients likely to benefit from OV treatment. A better understanding of how OVs mediate antitumor activity may result in predictive biomarkers that can better guide patient selection and combination strategies.

H.L. Kaufman is an employee of Immuneering Corporation.

The author wishes to thank Gail Iodice for thoughtful review of the article. H.L. Kaufman is supported by the NCI (RO1 CA093696).