Interferon-γ: teammate or opponent in the tumour microenvironment?

Abstract

Cancer immunotherapy offers substantive benefit to patients with various tumour types, in some cases leading to complete tumour clearance. However, many patients do not respond to immunotherapy, galvanizing the field to define the mechanisms of pre-existing and acquired resistance. Interferon-γ (IFNγ) is a cytokine that has both protumour and antitumour activities, suggesting that it may serve as a nexus for responsiveness to immunotherapy. Many cancer immunotherapies and chemotherapies induce IFNγ production by various cell types, including activated T cells and natural killer cells. Patients resistant to these therapies commonly have molecular aberrations in the IFNγ signalling pathway or express resistance molecules driven by IFNγ. Given that all nucleated cells can respond to IFNγ, the functional consequences of IFNγ production need to be carefully dissected on a cell-by-cell basis. Here, we review the cells that produce IFNγ and the different effects of IFNγ in the tumour microenvironment, highlighting the pleiotropic nature of this multifunctional and abundant cytokine.

In 1965, leukocytes were found to produce an antiviral molecule in response to phytohaemagglutinin that was different to the previously described type I interferons (interferon-α (IFNα) and IFNβ)¹. It was not until 1980 that this type II interferon was formally designated IFNγ. In addition to its role in microbial infections, IFNγ has prominent roles in other diseases, such as cancer. Initially, it was thought that IFNγ has only antitumour effects: rejection of a transplanted fibrosarcoma in mice by treatment with a bacterial endotoxin was prevented when the mice were given an IFNγ-neutralizing antibody and IFNγ-neutralized tumours grew faster². Similarly, studies of mice lacking IFNγ receptor (IFNGR) and signal transducer and activator of transcription 1 (STAT1) showed that endogenous IFNγ prevents the development of carcinogen-induced sarcomas. Additionally, the antitumour effects of the IFNγ-inducing cytokine interleukin-12 (IL-12) were ablated by neutralization of IFNγ³. These early studies uncovered the cytotoxic effects of IFNγ on tumour cells. Since then, protumour effects of IFNγ have emerged. The discovery that IFNγ promotes expression of the inhibitory molecules programmed cell death 1 ligand 1 (PDL1), PDL2, indoleamine 2,3-dioxygenase 1 (IDO1), inducible nitric oxide synthase (iNOS), FAS and FAS ligand (FASL), all of which limit antitumour immunity, has increased caution with the use of IFNγ-modulating cancer immunotherapies.

Programmed cell death 1 ligand 1.

(PDL1). A ligand that binds to programmed cell death 1 (PD1) on T cells to inhibit their activation, proliferation and cytokine production. PDL1 is also known as CD274 and B7-H1, and PDL2 is also known as CD273 or B7-C.

In this Review, we discuss IFNγ in the context of cancer and the challenges of targeting IFNγ therapeutically. Although IFNγ has direct cytotoxic effects on tumour cells, its therapeutic application is currently not possible owing to the broad expression of IFNGR and thus potential cytotoxic effects on antitumour immune cells. Therefore, IFNγ-inducing cancer immunotherapies have the undesirable potential to exacerbate tumour burden. We propose that IFNγ may serve different functions when produced, or responded to, by different immune cells, which thereby act as teammates (with immunostimulating, antitumour functions) or opponents (with immunosuppressing, protumour functions) in the tumour microenvironment (TME). By considering IFNγ production and responses within the TME, it may be possible to develop more mechanistically tailored approaches to bias IFNγ-based cancer immunotherapy towards solely antitumour effects.

Regulation of IFNγ expression

IFNγ expression is tightly regulated by epigenetic, transcriptional, post-transcriptional and post-translational modifications. These mechanisms prevent IFNγ expression by non-immune cells, naive T cells and even some activated immune cells⁴. Understanding how IFNγ expression is regulated may uncover mechanisms of immune exploitation by tumours to escape immunosurveillance and novel IFNγ-inducing pathways for therapeutic intervention.

Epigenetic and transcriptional regulation of IFNG.

IFNG is actively silenced in naive T cells via methylation and hypoacetylation. Conserved non-coding sequences within the proximal regulatory regions of IFNG allow epigenetic regulation, transcription factor binding and cellspecific expression⁵. Acetylation of histones H3 and H4 is important for T helper 1 (T_H1) cell differentiation, which is abrogated in STAT4-deficient cells, indicating that STAT4 promotes Ifng expression via acetylation and binding to the Ifng promoter⁶. Interestingly, hyperacetylation of IFNγ-encoding chromatin differs among T cells and can extend outside the IFNG locus, allowing cell-specific epigenetic regulation. For example, epigenetic regulation of IFNG by the long non-coding RNA NeST (also known as Tmevpg1 or Ifng-AS1) and WD repeat-containing protein 5 (WDR5) is critical for T_H1 cell expression of IFNγ via the transcription factor T-bet⁷. Acetylation of H3 and H4 is similar in T_H1 cells and CD8⁺ T cells; however, in CD8⁺ T cells it occurs independently of T-bet and is greater in memory T cells, in which it requires CD4⁺ T cell help. Ifng histone modifications induced by T-bet overexpression in CD4⁺ T cells are sufficient to drive Ifng expression under T_H2-polarizing conditions, uncovering epigenetic mechanisms that ensure cell lineage-specific expression of IFNγ⁸.

Epigenetic regulation.

Control of gene expression through phenotypic changes that do not alter the DNA sequence. Examples include DNA methylation, histone modifications, microRNAs, long non-coding RNAs and nucleosome positioning.

Similarly to the case for other cytokines, IFNG transcription is promoted by various stimuli, including T cell receptor (TCR) engagement, and transcription factors, such as activator protein 1 (AP-1), T-bet, eomesodermin (EOMES), nuclear factor of activated T cells (NFAT) and nuclear factor-κB (NF-κB)^9–12. Interestingly, a positive-feedback loop is formed by which activated immune cell products, such as IL-2, hydrogen peroxide and leukotrienes, further induce IFNγ production via second messengers (protein kinase C and cyclic GMP)^13–15. In addition, IL-12 and IL-18 promote IFNG transcription¹⁶; IL-18 acts as a cofactor by signalling via NF-κB and AP-1, and IL-12 promotes transcription of IL-18 receptor, establishing a positive-feedback loop¹⁷. These parallel signalling pathways allow synergistic regulation of IFNγ expression. IL-12 activates STAT4, and IL-18 activates the AP-1 subunit JUN, forming a STAT4–AP-1 complex which enhances STAT4 binding to the IFNG promoter^18,19. Consequently, prevention of IL-12 production may result in the equivalent of IFNG deletion by blocking IFNγ production in the TME.

Post-transcriptional regulation of IFNG.

Preformed IFNG mRNA in both the nuclear compartment and the cytoplasmic compartment allows fast translation and secretion upon appropriate stimulation. The mitogen-activated protein kinase p38 induced by IL-12 and IL-18 binds to an AU-rich element in the 3′ untranslated region of IFNG mRNA and stabilizes the mRNA^20,21. IFNG mRNA is also negatively regulated indirectly by the microRNAs miR-29, miR-146a and miR-142–3P^22–24. However, tumours can hijack these suppressive mechanisms to limit IFNG expression. Various tumour types have been shown to secrete miR-29 and miR-146, which remodel the TME in favour of tumour growth and metastasis^25,26.

Secretion of IFNγ.

The protein structure of IFNγ comprises mainly α-helices, allowing two molecules of IFNγ to dimerize in an antiparallel manner through noncovalent bonds²⁷. The functional IFNγ homodimer is immediately secreted and can be detected extracellularly as early as 6 h after TCR activation, and its level peaks at 12–24 h. Once secreted, mouse IFNγ persists in the blood for much longer (0.94 h) than other cytokines such as IL-2 (0.2 h)^28,29. This long half-life is thought to be due to expression of IFNGR by platelets, which may bind IFNγ to facilitate systemic transport³⁰. This systemic stabilization of IFNγ may explain the undesired systemic cytotoxic effects that are seen with IFNγ-modulating cancer immunotherapy³¹.

Producers of IFNγ

Numerous immune cell subsets, including T cells, natural killer (NK) cells, invariant NK T cells (iNKT cells), regulatory T (T_reg) cells, γδ T cells and B cells, produce IFNγ in the TME. Although IFNγ has pleiotropic effects, IFNγ production by immune cells is generally antitumorigenic rather than protumorigenic. However, the secretion of other cytotoxic, proinflammatory and anti-inflammatory cytokines together with IFNγ may modulate its activity in the TME. Thus, IFNγ produced by different cell types can have unique and distinct effects on its intended targets and bystanders within the TME. In addition, the way in which IFNγ is secreted by these cells, such as synaptic, leaky synaptic or multidirectional secretion, can affect the outcome of IFNγ production³² (FIG. 1).

Fig. 1 |. Classical IFNγ producers in the tumour microenvironment.

The spatial pattern of interferon-γ (IFNγ) release by T helper 1 (T_H1) cells, cytotoxic T lymphocytes (CTLs) and natural killer (NK) cells. T_H1 cells release IFNγ, interleukin-2 (IL-2), granulocyte–monocyte colony-stimulating factor (GM-CSF) and lymphotoxin-α (LTα) in a concentrated manner within the antigen-presenting cell (APC)–T_H1 cell synapse (synaptic release). Tumour necrosis factor (TNF) is released in many directions, both towards and away from the synapse (multidirectional release). CTLs release IFNγ towards the synapse but the release is not well directed, allowing IFNγ to exert effects on cells beyond, but near, the synapse (leaky synaptic release). The release of TNF and perforin is synaptic. NK cells release IFNγ in a multidirectional manner.

Invariant NK T cells.

(iNKT cells). Innate-like T cells that express a T cell receptor α-chain that recognizes lipid antigens presented by the non-classical MHC molecule CD1d expressed on dendritic cells.

γδ T cells.

T cells that express T cell receptor γ and δ chains and represent 1–4% of the T cell population. They produce interferon-γ (IFNγ) rapidly following activation in a non-MHC-restricted manner by tumour-derived lipids, glycoproteins and phosphorus-containing compounds.

Effector T cells.

CD8⁺ cytotoxic T lymphocytes (CTLs) are well-known key producers of IFNγ and are crucial for antitumour immunity. Unlike for T_H1 cells, expression of IFNγ by CTLs, as well as their production of perforin, granzymes, tumour necrosis factor (TNF) and IL-2, is independent of T-bet and requires the T-bet paralogue EOMES^33,34. TCR activation induces EOMES expression, which promotes IFNγ release from CTLs in a leaky synaptic manner; this ensures that the target cell receives a concentrated IFNγ signal but IFNγ also reaches neighbouring cells³⁵ (FIG. 1). This mode of IFNγ secretion differs from that for the cytotoxic molecules TNF and perforin, which are concentrated at the immunological synapse to mediate targeted cell killing. IFNγ⁺ T cells form a synapse with neighbouring non-antigen-specific CD8⁺ T cells, which is crucial to promote expansion and differentiation of bystander T cells, which in turn become IFNγ producers^36–38. Leaky synaptic release of IFNγ combined with its increased stability and widespread distribution compared with other cytokines allows hours of IFNγ exposure, which is required for the transcriptional effects that mediate effective tumour cell killing. Importantly, IFNγ production by CD8⁺ T cells is required for a response to therapy using antibody against the immune checkpoint molecule programmed cell death 1 (PD1)³⁹. In addition, the presence of proliferating CD8⁺IFNγ⁺ T cells in the TME is a biomarker for therapeutic response to the kinase inhibitor sorafenib in hepatocellular carcinoma⁴⁰. These findings illustrate the importance of IFNγ production by CD8⁺ T cells in mediating beneficial widespread and durable cytotoxic tumour killing and proinflammatory effects in the TME.

IFNγ is also the signature cytokine of T_H1 cells, which also produce TNF, IL-2, lymphotoxin-α (LTα) and granulocyte–monocyte colony-stimulating factor (GM-CSF). The release of IFNγ, IL-2, LTα and GM-CSF from T_H1 cells is synaptic, which directs prosurvival signals to antigen-presenting cells (APCs) (FIG. 1). By contrast, TNF is released from T_H1 cells in a multidirectional manner to mediate cytotoxicity in the TME, while promoting dendritic cell (DC) maturation and macrophage activation at the synapse⁴¹. The importance of T_H1 cells in the TME was shown in a mouse lung carcinoma model in which a switch from a T_H2 cell-dominated TME to a T_H1 cell-dominated TME promoted tumour clearance with combinatorial immunotherapies, and IFNγ production negatively correlated with tumour size⁴².

Human T_H17 cells, which produce IL-17 and resemble terminally differentiated memory T cells, have been shown to produce other effector cytokines, including IFNγ. Tumour-infiltrating T_H17 cells showed a positive correlation with CD4⁺IFNγ⁺ T cells and CD8⁺IFNγ⁺ T cells, and produced a potent antitumour response in an ovarian cancer model⁴³. Interestingly, IFNγ and IL-17 synergize to induce production of the T_H1-type chemokines CXC-chemokine ligand 9 (CXCL9) and CXCL10 by tumour cells to facilitate effector cell recruitment to the TME⁴⁴.

NK cells.

NK cells are innate cytotoxic cells that provide the first line of defence against tumour growth. NK cells recognize non-self targets, such as tumour cells, and mediate cytotoxic effects through the production of IFNγ, which is induced by IL-2 and IL-12, and is potentiated by TNF^45,46. IFNγ and TNF are stored in recycled endosomes, which deliver these cytokines to localized areas away from the target cell for multidirectional release⁴⁷. This makes IFNγ available to promote the activation of inflammatory cells and their recruitment to the TME (FIG. 1). NK cell tumour infiltration positively correlates with better cancer prognosis, and cancer stage negatively correlates with NK cell activity, specifically IFNγ production^48–50. A diagnostic test that measures blood NK cell activity for IFNγ production in patients with gastric cancer may be a promising non-invasive test for monitoring disease progression⁵¹.

iNKT cells.

When iNKT cells form a synapse with a DC presenting iNKT cell ligand, IL-12 from the DC is released and binds to IL-12 receptor on iNKT cells to induce IFNγ production⁵². The pattern of IFNγ release by iNKT cells may be synaptic owing to a iNKT cell–DC positive-feedback loop in which IFNγ produced by iNKT cells promotes DC maturation via the upregulation of co-stimulatory molecules^53,54. However, iNKT cells also generate IFNγ in response to DC-derived IL-12 in the absence of CD1d-presented antigen⁵⁵.

Activation of iNKT cells with the synthetic ligand α-galactosylceramide (α-GalCer) promotes antitumour activity in the clinic, which is dependent on an increase in the number of IFNγ⁺ cells in peripheral blood^56,57. Unfortunately, α-GalCer has been shown to produce iNKT cell anergy, contributing to the suboptimal therapeutic effects. Similar to T cells, iNKT cell anergy can be reinvigorated with anti-PD1 or anti-PDL1 therapy, which when combined with use of α-GalCer in mice prolonged the antitumour effect⁵⁸.

T_reg cells.

T_reg cells are a canonically immunosuppressive subset of CD4⁺ T cells characterized by expression of the transcription factor forkhead box P3 (FOXP3) and production of the inhibitory cytokines IL-10, IL-35 and transforming growth factor-β (TGFβ)^59,60. FOXP3 maintains T_reg cell suppressive identity through AKT inhibition via nuclear sequestration of forkhead box O1 (FOXO1), which represses IFNG transcription^61,62. Despite these mechanisms to suppress IFNγ production, FOXP3⁺IFNγ⁺ T_reg cells are present in various autoimmune diseases and in the TME⁶². IFNγ⁺ T_reg cells found in autoimmunity and bacterial infections express a T_H1-like transcriptional programme that includes expression of CXC-chemokine receptor 3 (CXCR3), which mediates their recruitment to sites of inflammation and suppression of T_H1 cells^63,64. However, IFNγ⁺ T_reg cells in the TME have impaired suppressive function, which allows greater antitumour immunity and contributes to decreased tumour growth⁶⁵.

In various mouse models of inflammatory disease, IFNγ and the alarmin IL-33 form a regulatory loop: IL-33 promotes IFNγ production by NK cells and γδ T cells, and IFNγ promotes IL-33 production by keratinocytes and fibroblasts^66,67. However, in the TME, IL-33 has a unique role in maintaining stability and function of T_reg cells, as IL-33-deficient T_reg cells produce IFNγ and exhibit a loss of suppressive function, which potentiates the antitumour effects of immune checkpoint blockade therapy⁶⁸. Although a direct mechanism of IFNγ repression by IL-33 in T_reg cells has not been elucidated, IL-33 supported the induction of a T_H2-like environment during chronic inflammation that promoted T_reg cell stability and was protumorigenic. Additionally, T_reg cell-specific deletion of the IL-33 receptor ST2 blocked tumour development in models of inflammation-induced skin and colon cancer⁶⁹. Collectively, the impact of this IFNγ–IL-33 axis on T_reg cells differs from its impact on other immune cells but supports the antitumorigenic role of IFNγ⁺ fragile T_reg cells in the TME.

γδ T cells.

γδ T cells are another T cell population that produces IFNγ rapidly following activation via T-bet and EOMES. Mice have two subsets of γδ T cells: IL-17⁺ γδ T cells, which promote tumour growth (owing to increased PDL1 expression and recruitment of immunosuppressive neutrophils and macrophages) and IFNγ⁺ γδ T cells, which have antitumour effects (owing to increased production of IFNγ, TNF, perforin and granzymes). IL-17⁺ γδ T cells are rare in humans, whereas IFNγ⁺ γδ T cells are more common in the TME of human cancers. γδ T cells are recruited to the TME before αβ T cells. Interestingly, γδ T cell-deficient mice have increased incidence of tumour development and growth. Mice with IFNγ-deficient γδ T cells fail to control tumour initiation and growth and also have impaired IFNγ production by αβ T cells⁷⁰. Similarly to the case for conventional αβ T cells, PD1 blockade in patients with leukaemia reinstated IFNγ production by tumour-infiltrating γδ T cells⁷¹. Interestingly, γδ T cell prevalence in the epithelia negatively correlates with epithelial malignancies, whereas γδ T cell prevalence in the TME correlates with good prognosis in various human cancers^72,73.

B cells.

In addition to their ability to produce antibodies, B cells mediate antibody-independent functions via cytokine secretion⁷⁴. A subset of innate CD11a^hiCD16/CD32^hi B cells produce IFNγ during early stages of bacterial infection, similarly to NK cells⁷⁵. Interestingly, IFNγ production by CD11a^hiCD16/CD32^hi B cells requires IFNGR and T-bet expression, as well as Bruton’s tyrosine kinase (BTK; via NF-κB), IL-1β and CD40–CD40L signalling⁷⁶. However, the role of IFNγ⁺CD11a^hiCD16/CD32^hi B cells in the TME and the manner in which IFNγ is released are unknown. It is possible that the antitumour effects induced by immunotherapy with CD40 agonists, which activate NF-κB in DCs and B cells, involve the induction of IFNγ by CD11a^hiCD16/CD32^hi B cells, but this has yet to be proven^77,78.

IFNγ responders in the TME

All nucleated cells constitutively express IFNGR1 and can respond to IFNγ, and therefore its pleiotropic effects in the TME are complex and the overall impact on tumour growth depends on the balance of antitumour IFNγ signalling (tumour cell killing, effector function, cell migration, immune cell proliferation and antigen presentation) acting as a teammate for the immune system and protumour IFNγ signalling (immunosuppression, angiogenesis and tumour cell proliferation) acting as an opponent of the immune system. Details of IFNγ signalling via its receptor are described in BOX 1, and an abbreviated list of notable IFNγ-regulated genes and their known roles in the TME are given in TABLE 1 (see REFS^79,80 for comprehensive reviews of IFNγ-induced genes in various diseases). The complex network of IFNγ responders in the TME is delineated here, focusing on T cells, NK cells, APCs, tumour cells, the vasculature and lymphatics (FIG. 2).

Box 1 |. IFNGR and signalling.

Structure and expression of IFNGR

Interferon-γ (IFNγ) receptor (IFNGR) consists of IFNGR1 (α-subunit) and IFNGR2 (β-subunit)¹⁷³ (see the figure). IFNGR1 is constitutively expressed by all nucleated cells at 200–25,000 molecules per cell. IFNGR1 expression is highest in non-lymphoid tissues such as the skin, nerves, placenta and syncytiotrophoblasts, suggesting a role for IFNγ in embryonic development, tissue homeostasis and tolerance¹⁷⁴. the levels of IFNGR1 mrNa expression by immune cells differ, with monocytes expressing the highest levels, then B cells, NK cells and lastly T cells^175,176. Conversely, the inducible expression of IFNGR2 by transcription factor SP1, activating protein 2 (AP-2) and nuclear factor-κB (NF-κB) allows regulation of IFNγ-induced signalling¹⁷⁷. IFNGR1 has a major role in ligand binding, whereas IFNGR2 has a predominant role in signalling via Janus kinases (JAKs) and signal transducers and activators of transcription (STATs), albeit both receptor subunits are required¹⁷⁸. interestingly, only IFNGR1 has a nuclear localization signal for translocation with STAT1 (REF.¹⁷⁹).

IFNGR signalling

IFNGR lacks intrinsic kinase activity and requires the adaptors JAK1 and JAK2 and the transcription factor STAT1 to mediate downstream signalling¹⁷³. IFNγ binding and receptor association triggers JAK2 autophosphorylation, transphosphorylation and activation, which uncovers docking sites for SH2 domain-containing signal transducers¹⁸⁰.

Canonical IFNGR signalling.

Each molecule of the STAT1–STAT1 antiparallel homodimer is phosphorylated at Tyr701 by JAK1 and/or JAK2, resulting in immediate receptor dissociation. STAT1 is phosphorylated at Ser727 via a PI3K- and AKT-dependent mechanism that is required for maximum transcriptional activity. Phosphorylated STAT1–STAT1 dimers undergo nuclear translocation via importin α5 (REF.¹⁸¹). STAT1 then binds to IFNγ-activating sites (GASs) containing the consensus sequence TTCN_2–4GAA within the promoters of interferon-responsive genes (IRGs)¹⁸². IFNγ promotes the transcription of interferon regulatory factor 1 (IRF1) and IRF9, thus further amplifying IFNγ-induced gene transcription¹⁸².

Non-canonical IFNGR signalling.

Studies of Stat1-knockout cells reveal STAT1–STAT1-independent IFNγ signalling via STAT3 (REFS^183,184). STAT3 and STAT1 are structurally similar and compete for binding to the phosphorylated Tyr419 of IFNGR, although STAT1 has higher affinity¹⁸⁵. STAT3 forms either a homodimer (STAT3–STAT3) or a heterodimer (STAT1–STAT3), which regulate GAS3 and GAS2, respectively. As interleukin-6 (IL-6) and IL-10 both signal via STAT3, the expression of cytokine receptors dictates which signalling pathway dominates¹⁸⁶. Differing levels of expression of STAT1 and STAT3 by specific cell types results in unique IFNγ-induced transcriptional effects in the tumour microenvironment^185,187.

Suppressor of cytokine signalling proteins.

Prolonged IFNγ-induced signalling promotes lethal autoimmune diseases in mice; therefore, intrinsic feedback inhibition mediated by suppressor of cytokine signalling (SOCS) proteins is imperative¹⁸⁸. SOCS1 transcription is promoted by SP1 and IRF1, which are induced by T cell receptor activation and IFNγ signalling, respectively¹⁸⁹. SOCS1 binds to JAK1 and JAK2 via its kinase inhibitory domain to target them for proteasomal degradation¹⁹⁰.

Table 1 |.

IFNγ-responsive genes and their role in the tumour microenvironment

Function in the TME	Gene symbol	Protein	Effect on tumour	Mechanism	Transcription factors
Antigen presentation	HLA-A, HLA-B, HHL-C	MHC class I molecules	Antitumorigenic	Antigen presentation to T cells	STAT1, IRF1, NF-κB
			Protumorigenic	Inhibitory receptor engagement on NK cells
	CD80	CD80 (also known as B7–1)	Antitumorigenic	Co-stimulatory molecule for CD28 on T cells	STAT1, STAT3, IRF1
	CD86	CD86 (also known as B7–2)	Antitumorigenic	Co-stimulatory molecule for CD28 on T cells	STAT3, IRF1
	CTSS	Cathepsin S	Antitumorigenic	Processing of tumour-specific peptides for antigen presentation	STAT1, STAT3, IRF1
Proliferation	CDKN1A	Cyclin-dependent kinase inhibitor 1A (also known as p21 or CIP1)	Antitumorigenic	Cell cycle arrest of tumour cells (G1 → S phase)	STAT1, STAT3, IRF1
	CDKN1B	Cyclin-dependent kinase inhibitor 1B (also known as p27 or KIP1)	Antitumorigenic	Cell cycle arrest of tumour cells (G1 → S phase)	STAT1, IRF1
	RB1	Retinoblastoma-associated protein	Protumorigenic	Cell cycle progression in tumour cells (G1 → S phase)	STAT1, STAT3, IRF1, NF-κB
	MYC	MYC	Protumorigenic	Cell cycle progression in tumour cells (G1 → S phase); promotes cellular metabolism	STAT1, STAT3, IRF1, NF-κB
Apoptosis	FAS	FAS	Antitumorigenic	Apoptosis of tumour cells	STAT1, IRF1,
			Protumorigenic	Apoptosis of B cells, T cells and endothelial cells	NF-κB
	FASL	FAS ligand	Antitumorigenic	Apoptosis of tumour cells	STAT1, IRF1,
			Protumorigenic	Apoptosis of B cells, T cells and endothelial cells	NF-κB
	CTSD	Cathepsin D	Antitumorigenic	Apoptosis of tumour cells	STAT3
Immunosuppression	CD274	Programmed cell death 1 ligand 1	Protumorigenic	T cell exhaustion; Treg cell, M2 macrophage and TH17 cell differentiation; inhibition of phagocytosis	STAT3, NF-κB
	PDCD1LG2	Programmed cell death 1 ligand 2	Protumorigenic	T cell exhaustion and cell death	STAT1, STAT3, IRF1, NF-κB
	IDO1	Indoleamine 2,3-dioxygenase 1	Protumorigenic	T cell anergy and death; Treg cell expansion; angiogenesis; TGFp production	STAT1, STAT3, IRF1, NF-κB
Effector	TBX21	T-bet	Antitumorigenic	TH1 cell differentiation; IFNγ production	STAT3, NF-κB
	SERPINB9	Serpin B9	Antitumorigenic	Inactivation of granzyme B to protect lymphocytes	IRF1
			Protumorigenic	Inactivation of granzyme B to protect tumours
	ITK	IL-2-inducible T cell kinase	Antitumorigenic	T cell activation	STAT1, STAT3,
			Protumorigenic	TH2 cell differentiation	IRF1, NF-κB
Migration	CXCR3	CXC-chemokine receptor 3	Antitumorigenic	Migration of TH1 cells, CTLs, NK cells and NKT cells	STAT1, STAT3, IRF1, NF-κB
	CXCL9	CXC-chemokine ligand 9	Antitumorigenic	Lymphocyte migration; TH1 and TH17 cell differentiation; inhibition of angiogenesis	STAT1, STAT3, IRF1, NF-κB
	CXCL10	CXC-chemokine ligand 10	Antitumorigenic	Lymphocyte migration; TH1 and TH17 cell differentiation; inhibition of angiogenesis	STAT1, STAT3, NF-κB
	CXCL11	CXC-chemokine ligand 11	Antitumorigenic	Lymphocyte migration	STAT3
			Protumorigenic	TH2 cell differentiation; angiogenesis; IL-10 production
	CCL2	CC-chemokine ligand 2 (also known as MCP1)	Protumorigenic	Monocyte migration; tumour-associated macrophage and TH2 cell differentiation; IL-4 production	STAT1, STAT3
	CCL5	CC-chemokine ligand 5 (also known as RANTES)	Antitumorigenic	T cell migration	STAT3, IRF1, NF-κB
			Protumorigenic	Monocyte and Treg cell migration; angiogenesis
	ICAM1	Intracellular adhesion molecule 1	Antitumorigenic	Transendothelial leukocyte migration; inhibition of M2 macrophage differentiation	STAT1, NF-κB
Transcription	IRF1	Interferon regulatory factor 1	Antitumorigenic	Apoptosis; inhibition of proliferation; antigen presentation	STAT1, IRF1, NF-κB
	BCL6	B cell lymphoma 6	Antitumorigenic	Germinal centre formation; T follicular helper cell formation	STAT1
			Protumorigenic	DNA damage; proliferation; tumour cell migration and invasion
	NFKB1	Nuclear factor-κB	Antitumorigenic	Lymphocyte survival and proliferation; inflammatory cytokine production	STAT1, STAT3, NF-κB
			Protumorigenic	Angiogenesis; tumour proliferation and invasion
	CREB1	cAMP-responsive element-binding protein 1	Antitumorigenic	Leukocyte survival and proliferation; B and T cell activation; TH1 cell differentiation; antibody production; IFNγ production	STAT3, IRF1, NF-κB
			Protumorigenic	Tumour cell proliferation, migration and invasion; TH2 cell differentiation; Treg cell stability; IL-10 production
	JUN	Activator protein 1 subunit JUN	Antitumorigenic	T cell activation, proliferation and differentiation; production of proinflammatory cytokines; inhibition of T cell exhaustion	STAT1, STAT3, NF-κB
			Protumorigenic	Neoplastic transformation; tumour cell proliferation
	FOS	Cellular oncogene FOS	Antitumorigenic	Tumour cell apoptosis; T cell activation and differentiation	STAT1, STAT3, IRF1, NF-κB
			Protumorigenic	Apoptosis; neoplastic transformation; tumour cell proliferation; IL-10 production

cAMP, cyclic AMP; CTL, cytotoxic T lymphocyte; IFNγ, interferon-γ; IL, interleukin; NF-κB, nuclear factor-κB; NK cell, natural killer cell; NKT cell, natural killer T cell; STAT, signal transducer and activator of transcription; TGFβ, transforming growth factor-β; T_H, T helper; TME, tumour microenvironment; T_reg cell, regulatory T cell.

Fig. 2 |. IFNγ responders in the tumour microenvironment.

Interferon-γ (IFNγ)-induced signalling occurs in many types of immune cells (such as T helper 1 (T_H1) cells, cytotoxic T lymphocytes (CTLs), antigen-presenting cells (APCs), regulatory T (T_reg) cells and natural killer (NK) cells) and non-immune cells (vasculature and tumour cells) within the tumour microenvironment. Key proteins upregulated by IFNγ and the interacting ligands and receptors are shown. The biological consequences of IFNγ-induced signalling in each cell type are summarized in boxes. IFNγ responses that make these cell types teammates are in green, whereas responses that make them opponents are in red. CXCR3, CXC-chemokine receptor 3; cDC1, conventional type 1 dendritic cell; FASL, FAS ligand; GC, germinal centre; IDO1, indoleamine 2,3-dioxygenase 1; IL-12R, interleukin-12 receptor; IFNGR, interferon-γ receptor; NO, nitric oxide; PD1, programmed cell death 1; PDL1, programmed cell death 1 ligand 1; pSTAT1, phosphorylated signal transducer and activator of transcription 1; TCR, T cell receptor.

Cytotoxic T lymphocytes.

As well as being important producers of IFNγ, CTLs express IFNGR and respond to IFNγ in the TME (FIG. 2). In the context of an infection, IFNγ regulates the contraction phase of CTL responses via FAS–FASL-mediated and BIM-mediated apoptosis, both of which are induced by STAT1 signalling⁸¹. High levels of IFNγ during the CTL expansion phase limit the size of the memory population by inhibiting IL-7Rα expression, thus reducing signalling by the prosurvival cytokine IL-7 (REFS^82,83). The mechanism by which IFNγ regulates IL-7Rα expression remains elusive, but it may involve an AKT–FOXO1 pathway^84,85. Treatment of patients with IFNγ-inducing immunotherapies induces effector and memory CD8⁺ T cell expansion, but whether these expanding cells have low IFNGR expression protecting them from apoptosis and promoting IL-7Rα expression is unclear^86,87. However, in vitro, activated IFNγ⁺ CTLs do not express IFNGR2 nor do they upregulate IFNγ-inducible genes in response to IFNγ⁸⁸. Conversely, CTLs in mouse models of low tumour burden express more IFNGR than naive T cells do, and IFNγ induction by treatment with antibodies against cytotoxic T lymphocyte antigen 4 (CTLA4) or PD1 resulted in activation-induced cell death, which limited effector memory formation and resulted in tumour growth⁸⁹. Due to this dichotomy, assessment of tumour burden, CTL infiltration, IFNGR expression and IFNγ levels before checkpoint blockade may help to identify treatment-responsive patients.

IFNγ promotes the recruitment of immune cells to the TME through transcriptional regulation of CXCL9, CXCL10 and CXCL11, and their cognate receptor CXCR3 on T cells, NK cells, monocytes, DCs and cancer cells⁹⁰. Increased chemotaxis of activated CTLs to the TME enhances cytotoxic effects and limits tumour growth. A tumour-selective oncolytic vaccinia virus engineered to express CXCL11 induced CXCR3⁺ CTL recruitment into the TME of a mouse mesothelioma model and elicited profound antitumour effects⁹¹. IFNγ also promotes CTL motility via chemokine-independent mechanisms⁹².

IFNγ has other protumorigenic effects on CTLs besides apoptosis induction. Overexpression of Ifngr2 on CTLs did not affect their development or proliferation but limited their cytotoxic activity in response to antigenic stimulation by an unknown mechanism⁸⁸. IFNγ also upregulates the expression of PDL1 and/or PDL2 on many cell types. Owing to the expression of both PD1 and PDL1 by T cells, self-inhibition may occur in trans in the TME⁹³ (FIG. 2).

CD4⁺ effector T cells.

Similarly to IFNγ⁺ CTLs, IFNγ-producing T_H1 cells decrease IFNGR2 expression following differentiation, enhancing their survival and thus antitumour effects in the TME⁹⁴. T_H1 cells actively repress T_H17 cell polarization via T-bet and inhibition of RUNX1, which provides reinforced T_H1 cell commitment and expression of a gene transcriptional programme that is most effective for tumour clearance^95,96 (FIG. 2). Additionally, IFNγ prevents T_H2 cell polarization via suppressor of cytokine signalling 1 (SOCS1) and T-bet, which inhibit IL-4 receptor (IL-4R) signalling and GATA3 expression and function, respectively^97,98. Interestingly, upon TCR stimulation in CD4⁺ T_H cell precursors, IFNGR1 becomes localized at the immunological synapse along with STAT1, a process that is inhibited by IL-4R expression in T_H2 cells. This co-recruitment of IFNGR1 and STAT1 to the immunological synapse creates a ‘T_H1 cell readiness’ upon TCR stimulation and ‘primes’ the cells to quickly polarize and mediate T_H1 cell signalling. Ultimately, IFNGR2 expression is downregulated in T_H1 cells; therefore, the continuation of dominant T_H2 cell/T_H17 cell antagonism may be reinforced by T-bet, rather than STAT1.

IFNγ also has protumorigenic effects on T_H1 cells. PDL1 expression on tumour-infiltrating effector T cells prevents T_H1 cell differentiation, providing an additional negative-feedback loop in the TME to limit IFNγ production⁹⁹. However, it is unclear which PD1⁺ cells interact with PDL1⁺CD4⁺ effector T cells in the TME and/or whether there is self-inhibition among T_H1 cells. IFNγ also promotes apoptosis via reduction of BCL-2 expression, upregulation of Fas and Fasl, and production of a detrimental oxidative environment^100–102.

T_reg cells.

The effects of IFNγ on T_reg cells have been studied in various disease states and are a topic of ongoing debate. As mentioned earlier, T_reg cells have been shown to adopt a T_H-like (T-bet⁺IFNγ⁺) effector phenotype to better suppress the appropriate effector responses in models of autoimmune disease and bacterial infection⁶³ (FIG. 2). In the TME however, IFNγ drives a ‘fragile’ T_reg cell phenotype, in which T_reg cells lose suppressive activity yet maintain FOXP3 expression, to undermine their protumour activity^65,103. Strikingly, mice with Ifngr1-knockout T_reg cells are resistant to anti-PD1 therapy in tumour models. A potential mechanism is that IFNγ-resistant T_reg cells change from a T_H1-like state to a T_H2-like state, the latter of which exhibits the highest viability and activation potential of T_reg cells (protumorigenic) and is enriched in melanoma and colorectal cancer¹⁰⁴. These findings suggest that in the TME, IFNγ-induced T_reg cell dysfunction allows full reinvigoration of CTL-mediated antitumour effects unleashed by anti-PD1 therapy. IFNγ also induces PDL1 expression on T_reg cells, and high numbers of PDL1⁺ T_reg cells in non-small-cell lung cancer correlate with better responses to PD1 and/or PDL1 blockade¹⁰⁵. These data suggest that PDL1⁺ T_reg cells create a barrier to antitumour immunity that can be disrupted only with PD1 and/or PDL1 blockade. Supporting these seemingly paradoxical findings, it has been shown that an IFNγ–STAT1-induced T_H1-like T_reg cell programme promotes suppression of T_H1 cells (protumour) but maintenance of T_reg cell stability through the delayed induction of IL-12 receptor, thereby protecting T_reg cells from STAT4-dependent dysfunction¹⁰⁶.

IFNγ also promotes antitumorigenic effects by inducing IDO1 expression, which catalyses the breakdown of tryptophan into kynurenines, which induce T cell apoptosis via caspase 8 activation and mitochondrial cytochrome c release¹⁰⁷ (FIG. 2). Interestingly, expression of IDO1, PDL1 and CTLA4 on T_reg cells is interconnected in peripheral blood of patients with melanoma, and strongly correlates with advanced disease and negative outcome¹⁰⁸. Several small-molecule IDO1 inhibitors are being investigated in the clinic as ‘immunometabolic adjuvants’ to widen the therapeutic window and limit autoimmune side effects of current cancer therapies¹⁰⁹. However, a recent phase I/II study with pembrolizumab (anti-PD1) plus the IDO1 inhibitor epacadostat did not show clinical benefit in patients with solid tumours¹¹⁰.

NK cells.

Data suggest that antitumorigenic functions of NK cells are activated by IFNγ. Phosphorylation of STAT1 on Tyr701 in NK cells occurs following transactivation by IFNγ⁺ iNKT cells in response to IL-12 from DCs¹¹¹ (FIG. 2). Studies using Stat1^Y701F-knock-in mice revealed that Tyr701-phosphorylated STAT1 is required for NK cell maturation, suggesting that STAT1 activation promotes antitumour immunity¹¹². Indeed, NK cell tumour infiltration is largely dependent on IFNγ-induced CXCR3 expression, as Ifngr1-knockout mice and Cxcr3-knockout mice have fewer tumourinfiltrating NK cells¹¹³ (FIG. 2). Additionally, IFNγ produced by bystander T cells acts on NK cells to promote maturation and tumour killing via TNF-related apoptosis-inducing ligand (TRAIL), expression of which is enhanced by IFNγ-induced interferon regulatory factor 1 (IRF1)¹¹⁴.

Conversely, phosphorylation of Ser727 on STAT1 in resting NK cells by cyclin-dependent kinase 8 results in decreased production of granzyme B and perforin, thus decreasing NK cell cytotoxicity (protumorigenic). Stat1^S727A-knock-in mice are more resistant to leukaemia and melanoma than controls, and are completely resistant to breast cancer metastasis¹¹⁵.

Antigen-presenting cells.

A key antitumorigenic function of IFNγ is the induced expression of MHC class I and class II molecules by APCs, such as DCs, macrophages and B cells, for presentation of tumour antigens to T cells¹¹⁶ (FIG. 2). IFNγ induces STAT1 and IRF1 binding to promoter IV of MHC class II transactivator (CIITA), which is the non-DNA-binding master regulator of MHC class II transcription¹¹⁷. IFNγ induces MHC class I expression via IRF1 binding to the promoter of NLRC5, which is a transcriptional regulator of MHC class I¹¹⁸. IFNγ also induces expression of the co-stimulatory molecules CD80 and CD86 by APCs, which promote T cell activation via CD28 engagement (FIG. 2). The induction of CD80 and CD86 expression offsets the immunosuppression induced by competitive binding of CTLA4 to CD28 (REFS^119,120).

IFNγ also drives antitumorigenic effects via DC differentiation into conventional type 1 DCs (cDC1s) through the expression of CD80, CD86, MHC class I, CD40, CD54 and CC-chemokine receptor 7 (CCR7), and the production of IL-1β and IL-12, which promote T_H1 cell differentiation and activation of CD8⁺ T cells¹²¹. Indeed, for full therapeutic efficacy, cDC1s are required to respond to IFNγ produced by CD8⁺ T cells during anti-PD1 therapy¹²².

The impact of IFNγ on B cells has only recently been described in models of autoimmunity and implies that IFNγ–STAT1 signalling is required for spontaneous development of germinal centres and T follicular helper cells, suggesting a potential antitumour effect. Specifically, IFNγ, in combination with B cell receptor and CD40 activating signals, induces expression of the germinal centre master transcription factor B cell lymphoma 6 (BCL-6)¹²³. IFNγ also works in concert with IL-12 to promote antibody class switching from IgM to IgG2a, therefore generating higher-affinity and specialized antibodies with antibody-dependent cytotoxicity, thus potentially promoting tumour antigen processing and presentation¹²⁴. Although a direct role for IFNγ in B cells and germinal centre formation in cancer is not known, recent studies described a patient survival advantage with the presence of B cells and tertiary lymphoid structures in the TME^125,126.

IFNγ was initially named ‘macrophage activation factor’ due to its role in driving classical activation of macrophages, commonly referred to as ‘IFNγ priming’. IFNγ signalling prepares macrophages for activation by Toll-like receptor (TLR)-induced inflammatory responses. IFNγ priming drives macrophages towards a proinflammatory and antitumorigenic, M1-like phenotype via the downregulation of miR-3473b, thereby suppressing an M2-like phenotype¹²⁷. In addition, IFNγ blocks sterol regulatory element-binding protein 1-dependent fatty acid synthesis in M2-like tumour-associated macrophages (TAMs), an effect that is required for responses to anti-PD1 therapy in a mouse model of melanoma³⁹. M1-like macrophages have increased phagocytic and tumoricidal activity compared with M2-like macrophages that is important for tumour surveillance¹²⁸. TAMs produce CXCL9 and CXCL10 in response to IFNγ, which not only promotes immune cell infiltration of the TME but may also inhibit angiogenesis¹²⁹.

In terms of protumorigenic activities, APCs are the predominant PDL1-expressing immune cell population in the TME. In addition to cancer cells and cancer-associated stroma, DCs and TAMs express IDO1 in response to IFNγ, which promotes immunosuppression through metabolic disruption and angiogenesis¹³⁰ (FIG. 2). Kynurenines produced by IDO1 induce TGFβ production in DCs, which promotes T_reg cell differentiation and immunosuppression¹³¹.

IFNγ also promotes iNOS expression in myeloid cells, which catabolizes the essential amino acid l-arginine to the free radical nitric oxide (NO). The role of NO in the TME is paradoxical and complex due to its effect on stromal cells, immune cells and tumour cells. The overall impact of NO largely depends on its expression level, the duration of exposure and the genetic makeup of the tumour¹³². NO promotes antitumour effects by inducing apoptosis, chromosome condensation and DNA fragmentation of immune cells, but it has protumour effects by inducing genomic instability of tumour cells via p53 and promoting angiogenesis. The T_H2-type cytokine-inducible enzyme arginase antagonizes NO production via competition with iNOS for l-arginine. Arginase has immunosuppressive effects through inhibition of immune cell proliferation, cytokine production, TCR activation and promotion of apoptosis¹³³. The overall balance of T_H1-type and T_H2-type cytokines dictates the impact of l-arginine catabolism on tumour growth.

Tumour cells.

Cancer cells are key responders to IFNγ in the TME and, like for immune cells, IFNγ drives both immunoactivating (teammate) and immunosuppressing (opponent) effects. The immunoactivating activity of IFNγ on tumour cells is largely attributed to induced tumour cell expression of MHC class I and secretion of CXCL9, CXCL10 and CXCL11 by tumour cells, monocytes, endothelial cells and fibroblasts to promote lymphocyte migration and inhibit angiogenesis (antitumorigenic)¹³⁴. Conversely, CXCL11 has pleiotropic activities due to binding to CXCR7, which promotes angiogenesis and tumour growth. CXCL9 and CXCL10 promote T_H1 and T_H17 effector cell function, whereas CXCL11 promotes a T_H2 cell response and regulatory function via IL-10 (REFS^135,136). Pharmacological approaches are in development to create biased synthetic ligands that favour CXCL9 or CXCL10 T cell signalling via CXCR3, rather than CXCL11-induced signalling, to promote antitumour immunity.

Similarly to APCs, tumours present antigens on their surface to T cells via MHC class I (antitumour); however, MHC class I molecules can also serve as a marker of ‘self ‘ which engages inhibitory receptors on NK cells to prevent killing (protumour)¹³⁷ (FIG. 2). MHC class I expression on tumour cells is variable, with immunosuppressive tumours often downregulating MHC class I expression, thus escaping immunosurveillance.

IFNγ regulates many survival and apoptotic pathways in tumour cells. For example, IFNγ induces apoptosis through IFNGR on tumour cells (antitumour). Knockdown of Ifngr1 in B16 melanoma cells results in impaired tumour rejection with anti-CTLA4 therapy, suggesting that IFNγ produced by the revigorated effector response must act directly on the tumour cells to elicit antitumour effects¹³⁸. Conversely, protumour effects of IFNγ on tumours are mediated through induction of PDL1, IDO1, iNOS, FAS and FASL expression (FIG. 2). Prevention of IFNγ signalling decreased PDL1 expression by tumour cells and increased IFNγ-responsive gene expression by immune cells, including exhausted T cells¹³⁹. Tumour cells are the main source of IDO1 in the TME and a major source of NO; however, expression can differ among tumour types¹⁴⁰. Tumour-derived iNOS promotes angiogenesis, which allows increased vascularization and tumour growth¹⁴¹. Tumour cells express FAS and FASL, with the former mediating antitumorigenic effects (tumour cell apoptosis by cytolytic effector cells) and the latter mediating protumorigenic effects (apoptosis of immune effector cells).

Vasculature and lymphatics.

The vasculature and lymphatics within the TME are underappreciated IFNγ responders, with both protumorigenic and antitumorigenic effects. Angiogenesis within the TME has been targeted by therapeutics for years but has produced mixed clinical results¹⁴². The lymphatics have recently been shown to serve not only as passive conduits for immune cell exchange in the TME but also as important regulators of inflammation and immunity.

IFNγ directly promotes protumorigenic effects on the lymphatics; T cell-derived IFNγ inhibits lymphangiogenesis via downregulation of lymphatic vessel endothelial hyaluronan receptor 1, podoplanin and prospero homeobox protein 1 on lymphatic endothelial cells, the last of which is a key transcription factor required for the growth, proliferation and invasion of lymphatics¹⁴³. IFNγ does not affect the initiation of lymphangiogenesis but instead inhibits the continuation of lymphatic vessel formation, resulting in reduced lymphatic density¹⁴⁴. Similarly to its effects on other cells, IFNγ induces PDL1 expression on lymphatics, which limits CD8⁺ T cell accumulation in the TME and prevents tumour control¹⁴⁵. IFNγ also promotes neovascularization in the TME indirectly through the induction of CXCL9, CXCL10 and IDO1 expression¹⁴⁶.

IFNγ also has indirect antitumorigenic effects to limit angiogenesis via the polarization of TAMs to an M1-like phenotype, which limits the amount of vascular endothelial growth factor that will be secreted by M2-like TAMs¹⁴⁷. Additionally, IFNγ production induced by IL-12 or pulse IL-2 therapy led to endothelial cell apoptosis via FAS–FASL, promoting tumour regression.

IFNγ and cancer immunotherapy

Almost all cancer immunotherapies, such as recombinant cytokines, vaccines, checkpoint inhibitors, chimeric antigen receptor T cell therapy and TLR agonists, modulate IFNγ^148–151 (BOX 2). These therapies aim to induce inflammation to aid tumour clearance; however, IFNγ-driven adaptive immune resistance can precipitate therapeutic resistance or disease exacerbation. Many preclinical studies of IFNγ-modulating immunotherapies over the past decade have aimed to exploit the antitumour effects and block the protumour effects of IFNγ in the TME.

Box 2 |. Promising IFNγ-modulating cancer immunotherapies in clinical trials.

Checkpoint blockade

Interferon-γ (IFNγ)-induced expression of programmed cell death 1 ligand 1 (PDL1) creates an anergic immunological state. Various antibodies blocking programmed cell death 1 (PD1; nivolumab, pembrolizumab, toripalimab and tislelizumab), PDL1 (atezolizumab, durvalumab and avelumab) and cytotoxic T lymphocyte antigen 4 (CTLA4; ipilimumab) are approved by the US Food and Drug Administration, with many more in clinical trials. interestingly, IFNG expression in the tumour microenvironment is a strong predictor of therapeutic response^138,164.

CAR T cell therapy

Autologous CD8⁺ T cells from patients with cancer are engineered to express a chimeric antigen receptor (CAR) specific for a tumour antigen; they are then expanded and transferred back to the patient¹⁹¹. First-generation CD19-targeted CAR T cells led to complete cures in patients with B cell acute lymphoblastic leukaemia; however, antigen loss and T cell exhaustion precipitated resistance¹⁹². second-, third- and fourth-generation CAR T cell therapies have been developed targeting various different antigens (including CD22, CD30, CD147, carcinoembryonic antigen, mesothelin, NKG2D, prostate-specific membrane antigen, tumour-associated mucin 1, B cell maturation antigen and disialoganglioside) and have been modified to increase antitumour efficacy and overcome immunosuppression in the tumour microenvironment.

IFNγ1b

Intratumoural injection of non-glycosylated recombinant IFNγ (IFNγ1b) was performed on the basis that it would increase intratumoural concentrations of interferon-inducible chemokines and increase immune cell infiltration of melanoma. However, the results were disappointing owing to upregulation of the immunosuppressive IFNγ-induced molecules indoleamine 2,3-dioxygenase 1 (IDO1) and PDL1 (REF.¹⁹³). IFNγ1b in conjunction with the anti-PD1 antibody pembrolizumab is currently in a phase ii clinical trial.

Recombinant IL-12

Human recombinant interleukin-12 (IL-12; edodekin alfa and NM-IL12) is currently being tested in phase I/II clinical trials with or without chemotherapy. it has been shown to result in stimulation of natural killer cells, T cells and natural killer T cells and in production of IFNγ and chemokines^194,195.

Cancer vaccines

Cancer vaccines present tumour-associated neoantigens, and sometimes patient-specific tumour antigens, to T cells and B cells, resulting in activation, maturation, proliferation and antibody production¹⁹⁶. Current vaccines in development use antigen-loaded antigen-presenting cells, adenoviruses and DNA or RNA approaches¹⁹⁷.

TLR agonists

Toll-like receptors (TLRs), expressed by innate immune cells, are activated by microbial products during infection and induce cytokine production and co-stimulatory molecules for T cell activation¹⁹⁸. the TLR7 and TLR8 agonist resiquimod (R848) promotes a strong T helper 1-type antitumour response via IFNγ and IL-12 production by natural killer cells and dendritic cells, respectively^199–201. However, resiquimod has poor pharmacokinetic profiles, and when administered systemically it promotes widespread immune activation, leading to autoimmune responses. Novel delivery methods for resiquimod using nanoparticles or thermosensitive liposomes are being tested to limit these toxic effects, and have produced promising results in combination with anti-PD1 in preclinical models²⁰².

Adaptive immune resistance.

The upregulation of immunosuppressive mechanisms in response to chronic proinflammatory stimuli.

Therapeutic approaches to deliver IFNγ to the TME.

The first clinical use of recombinant IFNγ was more than 30 years ago for the treatment of cancer and viral infections. Treatment with modified recombinant human IFNγ1b however generated disappointing results in the clinic, production was costly and the protein had a short half-life^152,153. IFNγ fusion proteins have since been engineered with longer half-lives and tissue-specific homing to enhance therapeutic effects and limit adverse effects¹⁵⁴. However, the toxicity of these IFNγ fusion proteins remains a challenge owing to widespread expression of IFNGR and receptor trapping of IFNγ that prevents effective tumour targeting¹⁵⁵.

The induction of IFNγ expression in the TME through alternative, gene-based approaches, such as viral transduction, also has technical limitations, such as transgene size, selective integration and expression efficiency¹⁵⁶. Clinical use of a replication-defective adenovirus encoding human IFNγ showed beneficial responses in most patients with cutaneous T cell lymphoma in a phase II clinical trial³¹. Oncolytic viruses encoding IFNγ allow concentrated cytokine release in the TME, which activates DCs and enhances T cell-mediated antitumour effects, prolonging survival of tumour-bearing mice¹⁵⁷. Delivery of an oncolytic adenovirus encoding the IFNγ-inducing cytokine IL-12, in conjunction with the collagen-associated extracellular matrix proteoglycan decorin to limit T_reg cell expansion, produced a potent antitumour response in a mouse model of breast cancer¹⁵⁸.

Non-viral genetic approaches include the delivery of IFNG gene therapy via plasmids, vectors and liposomes. The best route of delivery has been shown to be a promoter and plasmid backbone that results in constant and steady IFNγ production and lacks an initial burst, which is responsible for adverse effects¹⁵⁹. Recently, an Ifng-loaded lipoplex and an antigen-loaded liposome had synergistic effects of targeting DCs to present antigens and produce IFNγ in mice. This lipoplex–liposome combination resulted in tumour clearance and enhanced mouse survival that was dependent on CTL activation¹⁶⁰. Despite early optimism, approaches targeting IFNγ to the TME have largely failed to provide any clinical benefit. IFNγ-induced adaptive immune resistance highlights the importance of improving the delivery of and increasing the specificity and duration of IFNγ-induced immunotherapies and simultaneously limiting or blocking the expression or activity of IDO1, PDL1, NO, FAS and FASL.

Antitumour-biased IFNγ agonists.

A recent novel receptor engineering approach resulted in increased affinity of IFNGR2 for the IFNγ–IFNGR1 complex, and crystallization of this hexameric complex (in a 2:2:2 ratio) has revealed numerous targets for biochemical intervention to decouple the protumour and antitumour effects of IFNγ signalling¹⁶¹. Recently developed antitumour-biased IFNγ agonists for IFNGR retain IFNγ-induced upregulation of MHC class I expression but have impaired upregulation of PDL1 expression¹⁶¹. These biased agonists dimerize with one molecule of endogenous IFNγ to prevent full assembly of the hexameric ligand–receptor complex. These findings illustrate that the second IFNGR2 molecule of the hexameric complex may be redundant, and loss of this single IFNGR2 molecule maintains MHC class I induction, yet limits PDL1 induction. The therapeutic potential of these antitumour-biased agonists is promising, but lessons from IFNγ1b suggest that the addition of fusion proteins to induce favourable pharmacokinetics may be required.

Immune checkpoint blockade.

As mentioned earlier, IFNγ has a prominent role in immune checkpoint blockade with anti-PD1 or anti-PDL1. IFNγ was found to be localized to regions of high PDL1 expression on the surface of melanomas, implying that CTLs may trigger autoinhibition through IFNγ-driven PDL1 expression¹⁶². This mechanism of adaptive immune resistance may explain tumour escape from immunosurveillance. As well as driving upregulation of PDL1 expression, IFNγ production by CTLs is required to mediate their therapeutic effects^39,139,163. The requirement for CD8⁺ T cells and CD4⁺ T cells in mediating the response to anti-PD1 therapy in mouse models is now well established. Anti-PD1-induced IFNγ production by CTLs acts on cDC1s to produce IL-12, all of which are required to elicit a therapeutic response¹²². Although the cellular target (or targets) of IL-12 remain unclear, it is possible that IL-12-mediated downregulation of IFNGR2 may protect CTLs from IFNγ-induced apoptosis.

Resistance to cancer immunotherapy

Despite the potential for immunotherapy to transform the cell context and cytokine milieu of the TME, only a subset of patients have a complete response. Various clinical studies have been conducted at the genetic, epigenetic and metabolic levels to better understand adaptive immune resistance mechanisms of IFNγ-modulating cancer immunotherapies.

Genetic mutations.

Genomic and transcriptomic studies of responders and non-responders to checkpoint blockade therapy identified IFNγ-stimulated genes as key mediators of the therapeutic response¹⁶⁴. Specifically, non-responders to anti-CTLA4 therapy had defects in IFNγ signalling within the tumour, showing downregulation of ten genes (IFNG, STAT1, CCR5, CXCL9, CXCL10, CXCL11, IDO1, PRF1, GZMA and HLA-DRA) that constitute an IFNγ signature^138,. Additionally, JAK1 or JAK2 loss-of-function tumour mutations resulted in the lack of response to IFNγ and anti-PD1 therapy¹⁶⁵. Interestingly, JAK1 or JAK2 mutations promoted both adaptive and primary resistance to anti-PD1 therapy, with primary resistance evident in JAK1-mutated or JAK2-mutated, PDL1-negative tumours¹⁶⁶.

Epigenetic modulations.

Immune evasion through epigenetic silencing of CXCL9 and CXCL10 in tumour cells was associated with poor patient outcome, and pharmacological removal of these epigenetic marks increased effector T cell infiltration and efficacy of anti-PDL1 therapy¹⁶⁷. CXCL9 and CXCL10 are also silenced by enhancer of zeste homologue 2 (EZH2), which itself is antagonized by binding of AT-rich interaction domain-containing protein 1A (ARID1A). Interestingly, ARID1A mutations are highly prevalent in various cancers, especially ovarian cancer (50%), resulting in a decreased T_H1 cell signature and poor clinical benefit following checkpoint blockade¹⁶⁸.

Metabolic disruptions.

IFNγ can disrupt tumour cell metabolism and promote antitumour effects. IFNγ-induced downregulation of cystine–glutamate antiporter (X_c⁻) on tumour cells impairs cystine uptake and promotes lipid peroxidation and ferroptosis. Accordingly, a favourable outcome with anti-PD1 therapy was seen in patients with reduced expression of X_c⁻; however, the mechanism of IFNγ-mediated suppression of X_c⁻ expression is unclear¹⁶⁹. As mentioned, IFNγ-mediated induction of IDO1 expression with checkpoint blockade therapy has been postulated as a mechanism of adaptive immune resistance. Although monotherapy with IDO1 modulators was disappointing, they are safe and well tolerated. Preclinical studies have shown synergy with IDO1 inhibition and CTLA4 or PD1 blockade, and clinical studies have produced a wide range of response rates¹⁷⁰. The complex role of IDO1 in various tumours needs further investigation to ensure optimal therapeutic targeting.

Finally, by profiling IFNγ-regulated genes within the TME of patients before and after therapy, it may be possible to predict the proinflammatory and anti-inflammatory influence of IFNγ with specific treatments to better predict therapeutic responses¹⁶⁴.

Concluding remarks

Although IFNγ was discovered more than 50 years ago, the complex nature of this pleotropic cytokine in the TME is continuously being unravelled. The principles learned in infectious disease, graft-versus-host disease and autoimmune disease have provided insight into the role of IFNγ in the TME. However, the immune context of tumours can differ greatly from that of other disease states, and IFNγ induced with novel immunotherapies creates conflicting IFNγ-induced antitumour or protumour signalling events.

The categorization of IFNγ-induced signalling as antitumour or protumour is proposed to depend largely on the duration (acute versus chronic) and magnitude of IFNγ signalling. Interestingly, IFNγ is captured by phosphatidylserine residues on the surface of cells and slowly released to mediate autocrine and paracrine signalling (‘catch and release’) and contributes to preserving or delaying IFNγ signalling¹⁷¹. In the TME, the duration and magnitude of IFNγ signalling are also largely dictated by tumour burden and the state of immune cell infiltrate, respectively.

Initial IFNγ exposure recruits teammates (via CXCL9, CXCL10, CXCL11 and CXCR3) to promote antigen presentation (MHC class I and class II), T cell priming and activation (CD80, CD86 and CD40) and tumour cell killing (FAS and FASL). However, prolonged IFNγ exposure converts teammates into opponents, promoting protumorigenic effects via immunosuppression (PDL1, IDO1, FAS and FASL), angiogenesis (CXCL9, CXCL10, CXCL11, IDO1 and iNOS) and tumour cell proliferation¹⁷². In addition, opponents are likely to be present in the TME initially; however, teammates may dominate to promote overall antitumorigenic effects. Thus, studies that clearly measure and map IFNγ production and response over time in the TME are warranted.

There are several other key questions that warrant further investigation, including whether there are as-yet-unidentified IFNγ-induced genes, which IFNγ-producing cells are most important for antitumour effects, which IFNγ-expressing cells in the TME mediate resistance to immunotherapy and how IFNγ could be introduced to turn ‘cold’ tumours ‘hot’ (BOX 3). Future studies must use novel approaches to tease apart the proinflammatory effects from the anti-inflammatory effects of IFNγ to design better therapeutics to bias its antitumour capabilities and prevent immune escape.

Box 3 |. Outstanding questions on role of IFNγ in tumours.

• Are there novel genes induced by interferon-γ (IFNγ) yet to be identified and characterized? With the recent advances in single-cell transcriptomics, additional proinflammatory and anti-inflammatory genes induced by IFNγ may be identified. Novel IFNγ-induced proinflammatory molecules may be induced therapeutically, whereas novel IFNγ-induced anti-inflammatory molecules could serve as druggable targets to combat resistance to IFNγ-inducing cancer immunotherapies.

• Are all immune cells that can produce IFNγ required to do so to elicit an antitumour response? It is possible that a specific subset of IFNγ⁺ cells in the tumour microenvironment (TME) switch from acting as teammates to acting as opponents during disease or therapy. Alternatively, do different cell populations drive protumour effects versus antitumour effects? IFNγ production by cytotoxic T lymphocytes, T helper 1 cells and natural killer cells in the TME has been extensively studied, and they all appear to produce profound antitumour effects. However, the role of IFNγ⁺ regulatory T cells and B cells in the TME remains unclear, and they may also need to produce IFNγ to promote protumour or antitumour effects. For instance, tumour-infiltrating IFNγ⁺ regulatory T cells may have other effector functions besides suppressive ability, a phenomenon that could be required to unleash the antitumour effects of T helper 1 cells and cytotoxic T lymphocytes. Furthermore, IFNγ⁺CD11a^hiCD16/CD32^hi B cells in the TME may be an understudied population mediating the therapeutic effects of CD40 ligands in the clinic. Preclinical studies that selectively delete IFNγ expression from individual cell populations would aid in addressing these key questions.

• Do the location and identity of IFNγ⁺ and IFNGR2⁺ cells in the TME explain why some patients respond to therapy whereas others are resistant? Furthermore, is it possible to determine which IFNγ⁺ cells are located close to cells expressing indoleamine 2,3-dioxygenase 1 (IDO1) and programmed cell death 1 ligand 1 (PDL1), thus inducing protumour effects? PDL1 is highly expressed by antigen-presenting cells and tumour cells, which are most commonly surveyed by CD4⁺ effector T cells, cytotoxic T lymphocytes and natural killer cells. Through identification of these protumour IFNGR2⁺ cell types, bispecific antibodies composed of an IFNγ receptor 2 (IFNGR2) blocking antibody conjugated to a cell-specific homing antibody could prove advantageous.

• How can IFNγ be introduced into the TME of ‘cold’ immune-excluded solid tumours? Intratumoural injection of recombinant IFNγ is possible for easily accessible tumours such as melanoma, but not all tumours. This is an important challenge for many cell-based immunotherapies, and therefore direct IFNγ delivery through chimeric antigen receptor T cells, adenovirus vectors and vaccines may be viable and may negate the pharmacological and toxicity issues associated with systemic administration of IFNγ agents. Alternatively, indirect IFNγ-inducing therapeutic strategies such as pattern recognition receptor agonists could be used and have been shown to overcome resistance to checkpoint blockade^203,204.