Weekly Digests
‹ Back to August

The interferon seesaw between tumor cells and immune cells determines response to immune checkpoint blockade

August 14, 2019

Interferon is full of paradoxes. Some patients have cancers with a mutated IFN pathway, but still respond to immune checkpoint blockade (ICB). Other patients have high serum levels of IFNγ, but their cancer progresses with ICB treatment. IFNγ can boost immune function, but also promotes T cell exhaustion via upregulation of PD-L1. In a paper recently published in Cell, Benci and Johnson et al. set out to explore how the opposing effects of IFN affect the therapeutic response to ICB.

The researchers began by analyzing the single-cell RNAseq data of IFN-stimulated genes (ISGs) across various cell populations in human melanoma samples in TCGA. They found that the IFN-stimulated genes resistance signature (ISG.RS; initially associated with resistance to radiation and chemotherapy and later associated with resistance to immune checkpoint blockade) is mostly expressed by cancer cells, while a subset of the IFNγ hallmark gene signature chosen to not overlap with ISG.RS (IFNG.GS) is mostly expressed by the immune cells within the tumor, including T cells, NK cells, and macrophages. Low IFNG.GS/ISG.RS expression ratio correlated with resistance to anti-PD-1, while high IFNG.GS/ISG.RS ratio correlated with increased CD8+ T cells, activated NK cells, and response to anti-PD-1; this ratio predicted response independent of tumor mutational burden (TMB). These results indicated that IFN signaling in the tumor cells opposed the effects of IFN signaling in adaptive and innate immune cells to limit immune response.

In order to understand the mechanism behind the effect of the IFNG.GS/ISG.RS ratio, Benci and Johnson et al. utilized numerous tumor models that varied in baseline MHC-I expression, TMB, and neoantigen immunoediting, and prepared knockouts of either or both of the IFNα (type I) or IFNγ (type II) receptors in these tumors. They found that how the ratio of IFN signaling in immune cells (IFNG.GS) versus tumor cells (ISG.RS) affects various cell populations that participate in the antitumor response depends on context – particularly the MHC-I status and neoantigen availability (related to TMB) within the tumor.

In mice with CT26 colorectal tumors – which have high constitutive MHC-I expression and high TMB – knockout of type I and II IFN signaling in tumors slowed tumor growth and led to spontaneous regressions in some mice, and anti-PD-1 further boosted survival; complete responders were protected from rechallenge. The antitumor effect in this model was dependent on CD8+ T cells, which were able to be activated due to the maintenance of high levels of tumor MHC-I expression even in the absence of IFN signaling.

In contrast, Res 499 melanoma tumors, which are derived from B16F10 tumors, require IFN signaling for high MHC-I expression, and have a lower number of neoantigens than their parental tumors. In such tumors, knockout of the IFNγ receptor IFNGR improved response to anti-CTLA-4, and the antitumor response was dependent on NK1.1+ innate immune cells (including NK cells and innate lymphoid cells [ILC1s]), was independent of perforin, and did not protect against rechallenge. In this model, knocking out IFNGR in the tumors decreased tumor MHC-I expression and therefore, CD8+ T cells were not involved in direct killing of the tumors. Surprisingly, however, depletion of CD8+ (but not CD4+) T cells negatively affected IFNGR-knockout Res 499 tumors’ response to anti-CTLA-4, suggesting that CD8+ T cells play a supporting role in this tumor model.

Exploring the indirect role of CD8+ T cells in the IFNGR-knockout Res 499 tumor model, Benci and Johnson et al. found that IFNGR knockout in the tumor increased intratumoral terminally exhausted CD8+ T cells, and anti-CTLA-4 increased the CD8+ T cell production of IFNγ. The IFNγ in turn enhanced the IFNG.GS signature in myeloid cells and DCs, including the Cxcl9 and Cxcl10 genes that encode chemokines that help recruit, activate, and promote the maturation of NK cells. Furthermore, IFNGR knockout increased the proportion of TRAIL+ (TNF-Related Apoptosis Inducing Ligand) PD-1+ ILC1s. While IFNGR knockout in tumors decreased tumor PD-L1 expression, it increased the tumor expression of the TRAIL receptor TRAILR2. Decreased PD-L1 and increased TRAILR2 expression allowed for activation of NK/ILC1 cells, which mediated tumor killing in this model. Together, these results demonstrate that IFNγ produced by terminally exhausted CD8+ T cells is necessary for the NK/ILC1-dependent anti-CTLA-4 response in IFNGR-knockout Res 499 tumors. Interestingly, MHC-I expression on tumors was not required to activate CD8+ T cells, as cross-priming of tumor-specific T cells or activation of bystander T cells was sufficient to support NK/ILC1 function.

Having discovered the role of NK cells in the antitumor response to ICB, Benci and Johnson et al. once again looked at the human melanoma data and found that the proportion of activated NK cells inversely correlated with the proportion of Tregs in the tumor. In mice, the researchers found that Treg depletion played a role in the CT26 tumor model, and failure to deplete Tregs abrogated the response to anti-CTLA-4 in IFNGR-knockout Res 499 tumors. These observations suggest that Treg depletion may be required for optimal engagement of both innate and adaptive immune responses in the context of inhibited tumor IFNγ signaling.

Armed with the knowledge of the opposing mechanisms of IFN, the researchers analyzed human exome data of patients with non-small cell lung cancer (NSCLC) to explore whether mutations that decrease tumor IFN signaling may affect clinical response to ICB. They found that patients treated with anti-PD-1 plus anti-CTLA-4 who had tumor mutations that led to a decrease in ISG.RS gene expression experienced significantly longer progression-free survival compared to patients without such mutations. Consistent with preclinical data, tumors with such IFN variants had decreased PD-L1 expression, and the IFN pathway variants predicted response to ICB independent of TMB and PD-L1 expression.

Overall, Benci and Johnson et al. have begun to unravel the pro- and anti-tumor effects of IFN signaling by distinguishing signaling in immune cells (IFNG.GS) versus tumor cells (ISG.RS). Chronic IFN signaling in tumors coupled with immunosuppressive Treg cells leads to upregulation of PD-L1 on tumors, which contributes to the inability of exhausted T cells to directly kill tumor cells and to produce sufficient IFNγ to support NK/ILC1-driven tumor control. Immune checkpoint blockade with anti-PD-1 (stimulating exhausted effector T cells) or anti-CTLA-4 (depleting Tregs) unlocks the adaptive and innate immune cells, which can both contribute to tumor control dependent on the MHC-I status and neoantigen availability within the tumor.

by Anna Scherer


This week, we asked first co-author Lexus Johnson a few questions about his research.

What prompted you to tackle this research question?

A number of groups have come to conflicting conclusions about how interferon affects the tumor microenvironment, so we were interested in figuring out how these opposite effects might fit together. Beyond that, we were curious how immune-edited tumors respond to immune checkpoint blockade despite a lack of T cell targets.

What was the most surprising finding of this study for you?

I think our most surprising finding in this paper is that NK cells can play such a large role in immune-edited solid tumors. Previously, they've received much more focus in liquid cancers, and it was exciting to find a major role for their function in this context.

What was the coolest thing you’ve learned (about) recently outside of work?

I recently took a trip to Alaska and learned some new ocean fishing techniques. The biodiversity in the southeast part of the state is truly remarkable.


Benci J.L., Johnson L.R., Choa R., Xu Y., Qiu J., Zhou Z., Xu B., Ye D., Nathanson K.L., June C.H., Wherry E.J., Zhang N.R., Ishwaran H., Hellmann M.D., Wolchok J.D., Kambayashi T., Minn A.J. Opposing Functions of Interferon Coordinate Adaptive and Innate Immune Responses to Cancer Immune Checkpoint Blockade. Cell. 2019 Aug 8.

In the Spotlight...

Tumor cell-intrinsic EPHA2 suppresses anti-tumor immunity by regulating PTGS2 (COX-2)

Markosyan and Li et al. found that EPHA2 expression negatively correlated with CD8+ T cell infiltration in human and mouse pancreatic ductal adenocarcinoma (PDA). Murine EPHA2-/- tumors increased T cell infiltration, MHC-I and inflammatory gene expression and chemo- immunotherapy treatment sensitivity. PTGS2 was highly expressed in T cell-excluded PDA tumors and PTGS2-/- PDA had greater T cell numbers, reduced immune-suppressive markers, and extended mouse survival; overexpression had the opposite effects. PTGS2 appears to be regulated by EPHA2 via the TGFβ/SMAD3 axis, and inhibition improved treatment outcome.

Contributed by Alex Najibi

Sustained delivery and molecular targeting of a therapeutic monoclonal antibody to metastases in the central nervous system of mice

Wen and Wu et al. show rituximab (anti-CD20) encapsulated in crosslinked, biodegradable, polymeric nanocapsules sustained drug kinetics in the blood, penetrated rodents’ brains, and was not neurotoxic. Compared to free antibody, nanocapsules improved tumor control in an immunodeficient mouse model of CD20+ non-Hodgkin’s lymphoma (NHL) with brain metastases, which was likely mediated via direct apoptosis. Arming nanocapsules with CXCL13 to target CXCR5 on NHL cells increased efficacy, showing systemic and brain tumor elimination in a xenograft humanized NHL mouse model, likely enhanced by human NK/macrophage ADCC.

Contributed by Paula Hochman

REVIEW: Mechanisms of immunotherapy resistance: lessons from glioblastoma

In this review, Jackson and Lim describe glioblastoma (GBM) as a paradigm for immune resistance mechanisms. GBM resides in an environment where immune cells are largely excluded and where tolerogenic pressure is prominent; it exhibits molecular heterogeneity, which makes it challenging to target and prone to immune escape; it often lacks obvious targetable antigens; and it expresses molecules that engage immune checkpoints. Recent strategies to target resistance in GBM include targeting multiple checkpoints, using oncolytic viruses to trigger in situ immunization, and reprogramming myeloid cells in the TME.

Natural killer cell recruitment and activation are regulated by CD47 expression in the tumor microenvironment

TCGA gene expression data revealed that high NK cell CD47 expression correlated with improved survival, specifically in melanoma. In vitro, treatment of mouse splenic NK cells with anti-CD47 enhanced CD69 and IFNγ. In vivo, intratumoral injection of anti-CD47 slowed growth of B16F10 melanoma, potentially related to blockade of thrombospondin-1 negative signaling. In contrast, B16F10 melanoma grew faster in CD47-/- mice, which had fewer immune-competent splenic NK cells than wild type controls, suggesting additional positive regulatory functions for NK cell CD47 in mouse melanoma, consistent with the human expression data.

Contributed by Samuel Goldman

CLINICAL TRIAL: Immunotherapy of Relapsed and Refractory Solid Tumors With Ex Vivo Expanded Multiantigen-Associated Specific Cytotoxic T Lymphocytes: A Phase I Study

In this phase I trial, autologous tumor-associated antigen cytotoxic T cells (TAA-Ts) were induced ex vivo with three target antigens (WT-1, PRAME, or survivin). Treatment of patients with relapsed/refractory solid tumors expressing one of the three antigens was well tolerated, with no dose-limiting toxicities. Of 15 evaluable patients, 11 responded (stable disease or better at 45 days post-infusion). The highest dose showed the best clinical outcomes and trended toward improved PFS at 6 months. Products that induced responses had more CD8+ than CD4+ T cells. Most responders showed reduced circulating TAA DNA and evidence of antigen spreading.

Removal of N-Linked Glycosylation Enhances PD-L1 Detection and Predicts Anti-PD-1/PD-L1 Therapeutic Efficacy

Lee and Wang et al. demonstrated that enzymatic removal of N-linked glycans from PD-L1 on the surface of tumor cells in patient samples significantly improved PD-L1 binding affinity and signal intensity in a variety of assays commonly used to stratify patients for cancer immunotherapy. After sample deglycosylation, improved PD-L1 detection was observed in different types of cancers and positively correlated with PFS and OS after anti-PD-1/PD-L1 therapy. The enhanced PD-L1 signal decreased the false negatives and underdetection of PD-L1, significantly increasing the fraction of patients eligible for anti-PD-1/PD-L1 therapy.

Contributed by Katherine Turner

Everything New this Week In...

Close Modal