When Cancer Cells Become the Enablers of an Antitumor Immune Response

Response to immune-checkpoint blockade (ICB) therapy is primarily mediated by antitumor cytotoxic cells that recognize tumor antigens presented by major histocompatibility complexes (MHC) on the surface of cancer cells. Most cancers express MHC class I molecules, which present antigenic peptides to CD8 cytotoxic T cells. As carcinogen-induced cancers and cancers induced by viruses are the indications with higher response rates to ICB, neoantigens resultant from somatic mutations or viral integration are the most likely targets of cytotoxic T cells (1, 2). To achieve antigen recognition on cancer cells, CD8 T cells establish a series of one-on-one interactions with the cancer cells that they are physically in contact with, creating a cytotoxic T-cell immunologic synapse. In this synapse, T-cell receptor (TCR)–MHC–antigenic peptide pairs are surrounded by multiple adhesion molecules, such as the leukocyte function–associated molecule 1 (LFA1) and the intercellular adhesion molecule 1 (ICAM1), creating a supramolecular activation complex (SMAC) centered around the TCR (3). TCR signaling results in actin and microtubule reorganization to polarize the cytotoxic T cells to allow lytic granules to traffic to the synapse area with the cancer cell. In this tight context, cytotoxic T cells release lytic granules that induce apoptotic death of that specific target cancer cell through the local release of perforin and granzymes, as well as the surface expression of Fas ligand (CD95L or CD178) and TNF-related apoptosis-inducing ligand (TRAIL; ref. 4).

However, tumor-specific T cells exist at a substantial numerical disadvantage to the cancer cells they target in the tumor site. If direct cell-mediated cytotoxicity were the sole mechanism of action of an antitumor immune response, then the relatively small number of tumor antigen–specific cytotoxic T cells would face an uphill battle to kill each individual cancer cell sequentially to induce a durable antitumor immune response. Next to this numerical disadvantage, the immune system would be at a major strategic disadvantage, as solid tumors are an unwelcoming environment for T cells to carry out their job. In the tumor microenvironment (TME), there is competition for metabolites that are required for T cells to expand, as well as immune-suppressive factors produced by cancer cells and professional immune suppressor cells such as M2 macrophages, myeloid-derived suppressor cells, and T regulatory cells (Treg; Fig. 1A; refs. 5, 6). Because of these limitations, the immune system may in fact require means to alter the TME once an initial nucleus of immune activity has been induced by ICB therapy in order to shift the balance in favor of the function of antitumor cytotoxic T cells (Fig. 1B).

After a first nucleus of antigen recognition is established, an increase in the susceptibility of surrounding cancer cells could lead to successive waves of amplification of the antitumor immune response. Such increased susceptibility could be achieved by the secretion of cytokines that induce the expression of genes that promote antigen presentation and immune sensitivity of cancer cells far away from the site of the first cancer cell encounter, paving the way for effective immune infiltration and immune attack. The primary suspect to achieve these effects is IFNγ, which is integrally involved in clinical responses to PD-1 blockade therapy (7, 8). Indeed, it would not quite make sense if the action of IFNγ was restricted to those target cells that are being recognized by T cells; these cancer cells are destined to be killed anyway through the cytotoxic effects of CD8 T cells, without a further benefit of the expression of most of the IFNγ-stimulated genes by these cells. The ability of IFNγ to diffuse beyond the immediate environment of the T cell secreting it serves to modulate the behavior of remote cancer cells. The spreading of the IFNγ response can reach distances of >800 μm in preclinical models, which amounts to 30 to 40 cell layers (9–11). Therefore, T-cell recognition of a minor fraction of cancer cells leads to sensing by a large part of the tumor mass through IFNγ secretion. Based on this logic, it is proposed that a dominant effect of IFNγ in the TME is to remodel the tumor tissue environment through IFNγ sensing at larger distances, thereby rendering it more permissive for immune attack (Fig. 1C).

IFNγ has a broad array of functions, changing the expression of over 500 genes within 6 hours and thousands of additional genes in the following 24 hours (8, 12). This IFNγ-induced alteration of the tumor cell transcriptome markedly changes a large number of processes in cancer cells that are associated with increased antitumor immune responses, including (i) increased expression of multiple molecules in the antigen presentation pathway, such as proteasome sub­units, Tap transporters, beta-2 macroglobulin (B2M), and MHC class I and II molecules and their pathway transactivators (NLRC5 and CIITA), that together increase the “visibility” of cancer cells to CD4 and CD8 T cells; (ii) increased expression of IFNγ receptor pathway molecules, including the IFNγ receptor subunits (IFNGR1 and 2), Janus kinases 1 and 2 (JAK1 and JAK2), signal transduction and activators of transcription (STAT), and interferon-response factors (IRF), which serve to amplify the effects of IFNγ in cancer cells, creating a positive forward loop; and (iii) expression of chemokines, such as CXCL9, 10, and 11, leading to the attraction of multiple immune cell types, including additional antitumor T cells. In addition, IFNγ can in certain settings induce cellular growth arrest or senescence (13), a phenomenon that was first noted for interferons in 1957, when they were described as soluble molecules that interfere with the growth of virally infected cells (8, 12, 14). This latter effect may be of particular importance to potentially extend immune control to antigen-negative cancer cells (9). By the same token, IFNγ-induced chemokines can have angiostatic and antiangiogenic effects through the shared CXC chemokine receptor 3 (CXCR3; ref. 15), which may further mediate antitumor activity beyond antigen-expressing cells. This combination of immune-stimulating and additional effects places IFNγ at the center of cancer immune surveillance, in line with the observation that the absence of IFNγ results in increased cancer development in classic studies of murine carcinogenesis (16). Next to the series of effects that enhance tumor control, IFNγ also has well-recognized negative immune regulatory functions, best exemplified by the induction of PD-L1 expression that effectively inhibits antitumor immune responses in steady-state conditions (as the single targeting of PD-1 or PD-L1 can result in clinical responses in a subset of patients), as well as the expression of genes that influence immune control through other mechanisms, such as the tryptophan converting enzyme indoleamine 2,3 dioxygenase (IDO) or the negative regulator of the IFNγ receptor pathway suppressor of cytokine signaling 1 (SOCS1; ref. 12).

Very few cells in the human body can release IFNγ, with dominant roles for activated CD4 T helper and CD8 cytotoxic T cells upon cognate antigen recognition, and natural killer (NK) cells when recognizing their activating ligands in the absence of inhibitory ligands (12). In contrast, IFNγ receptors are expressed by most somatic cells as well as their malignant counterparts. Therefore, upon recognition by activated immune cells, the great majority of cancer cells express the full array of IFNγ response genes (8, 12). As described above, among these induced genes is PD-L1, which—in the absence of therapeutic intervention—allows cancer cells to inhibit an antitumor response by binding to the PD-1 receptor on T cells that are trying to attack them. Specifically, PD-L1 triggering leads to the docking of the SHP2 phosphatase to the immunoreceptor tyrosine-based inhibition motif in the intracellular domain of PD-1 and subsequent inhibition of T-cell activation (Fig. 1A). Such negative feedback regulation of an immune response by IFNγ can explain the improved control of IFNγ receptor signaling–incompetent tumors under some conditions in preclinical implantable tumor models (17–19). However, upon the administration of antibodies that block the PD-L1–PD-1 interaction, this negative feedback loop is disrupted (20) and IFNγ levels are expected to increase, thereby boosting tumor antigen presentation, further amplifying IFNγ receptor signaling and attracting additional immune cells (Fig. 1B). In this setting, IFNγ activity in the TME may be viewed as a moving front in the tumor, in which, after a first nucleus of antigen recognition, the increased visibility of surrounding cancer cells and favorable changes to noncancer cells lead to a wave of immune infiltration deep into the tumor to result in a clinical response (Fig. 1C). Whether either the induction of—for example, increased antigen presentation—the chemokine-mediated attraction of a novel pool of T cells or the direct effect of IFNγ on tumor cell proliferation forms the dominant force in tumor control in this situation is presently unknown and may well be context-dependent. For example, the long-distance bystander effects of IFNγ, with its ability to induce direct cytotoxicity, may further enhance the antitumor activity against antigen-negative cancer cells mediated by Fas ligand (21). However, it seems likely that PD-1 blockade therapy works at least in part by shifting the balance of IFNγ effects on cancer cells—no longer allowing cancer cells to hide from the cytotoxic T cells through the expression of PD-L1—and turning the cancer cells into enablers of an IFNγ-mediated antitumor response.

The initial cellular response to IFNγ seems to be an all-or-nothing event, with the set of transcriptional changes arising from exposure to IFNγ being conserved in most cancer cells (12), at least when tested in a large panel of melanoma cell lines (8). The exception is formed by cancer cells that have lost the expression of a key component of the IFNγ receptor signaling pathway, most frequently the JAK1 or JAK2 kinases (22–25). Loss of either JAK1 or JAK2 is sufficient to prevent signaling through the IFNγ receptor (12), and the JAK kinases are thus a bottleneck of the IFNγ receptor pathway. However, less than 1% of treatment-naïve cancers have such loss-of-function mutations in JAK1 or JAK2, including tumor types with frequent tumor-specific T-cell reactivity (23), suggesting that at steady state, cancer cells can thrive with an intact IFNγ receptor signaling pathway. As noted above, IFNγ has the potential to increase the hostility of the TME to the immune response by increasing inhibitory ligands such as PD-L1, and even the upregulation of MHC class I, an effect that renders tumor cells more visible to CD8 T cells, has the potential to inhibit NK cell function. Thus, although the diffusion of IFNγ across the tumor bed can serve as a force multiplier for outnumbered T cells, the specific context of the tumor probably dictates whether IFNγ serves, on balance, to either augment or inhibit the immune response (18). Indeed, cancers with the highest baseline frequency of JAK1 or JAK2 loss-of-function mutations are highly immunogenic, microsatellite instability–high endometrial cancers (26), in which the balance may be shifted to favor the loss of IFNγ signaling due to their high inherent immunogenicity.

Clinically, loss of JAK1 or JAK2 is recognized as a mechanism of resistance to immunotherapies, in which the selective pressure of an antitumor immune response changes the benefit ratio of signaling through the IFNγ receptor in cancer cells. In CRISPR screens, loss of JAK1 or JAK2 is often (but not always) among the top hits of cancer cell–intrinsic resistance screens to immunotherapies (25, 27–31). Likewise, loss-of-function mutations in JAK1 or JAK2, and in other molecules of the IFNγ receptor pathway, are recurrently reported in progressive cancers in a subset of biopsies for patients who developed acquired resistance to immunotherapies, including patients with transient clinical responses to ICB, cytokine therapy, and TCR-engineered adoptive cell transfer (ACT) therapies (22, 24, 32, 33). The critical role of a functional IFNγ receptor signaling pathway to respond to both ICB- and ACT-based immunotherapies is highlighted in mouse model systems in which loss of JAK1 or JAK2 can result in resistance to immunotherapies (34–37). In the setting of JAK1 or JAK2 loss, cancer cells can have baseline levels of MHC class I expression and antigen presentation, still being recognized by T cells and being targets of their cytotoxic effects (22, 35). However, the loss of the cancer cell–induced signal amplification through the production of IFNγ-stimulated genes results in the cancer escaping from a previously effective antitumor immune response. Collectively, these data highlight the importance of IFNγ signaling and amplification as key components of how the immune system can result in clinical responses in patients with metastatic cancers.

PD-1 blockade therapy can lead to clinical responses in patients with multiple cancers by enlisting rare T cells with a TCR that recognizes tumor antigens whose function is limited by the PD-1:PD-L1 axis (38). In a matter of days to weeks, antitumor T cells need to expand in a hostile TME to be able to kill enough cancer cells at multiple metastatic sites to result in a clinical response to therapy. To do this, they enlist the cancer cells and other cells within the tumor to help in their pursuit by changing their gene expression profile to produce hundreds of IFNγ-stimulated genes. The ability of IFNγ to change gene expression in cancer cells’ multiple layers away from where the CD8 T cell recognized its cognate tumor antigen (9, 10) is proposed to result in the needed change in the cancer milieu to facilitate the antitumor effects of cytotoxic T cells. This view explains why IFNγ induces so many effects that are less relevant to the target cell in the immediate vicinity of the T cell. Understanding the factors that influence IFNγ spreading and sensing as well as understanding the contexts in which the different immune-stimulatory and immune-inhibitory effects of IFNγ dominate will help refine how immunotherapies are used for the greatest clinical benefit.

A. Ribas reports personal fees from Amgen, Chugai, Genentech, Merck, Novartis, Roche, Sanofi, Vedanta, 4C Biomed, Appia, Apricity, Arcus, Highlight, Compugen, ImaginAb, Kalthera/ImmPACT Bio, MapKure, Merus, Rgenix, Lutris, Nextech, PACT Pharma, Synthekine, Tango, Advaxis, CytomX, Five Prime, RAPT, Isoplexis, and Kite/Gilead and grants from Agilent and Bristol Myers Squibb outside the submitted work. W.N. Haining reports equity in and is employed by Arsenal Biosciences. T.N.M. Schumacher reports personal fees from Allogene Therapeutics, Asher Bio, Celsius, Cell Control, Merus, Neogene Therapeutics, Scenic Biotech, and Third Rock Ventures outside the submitted work.