HOIP limits anti‐tumor immunity by protecting against combined TNF and IFN‐gamma‐induced apoptosis

Abstract

The success of cancer immunotherapy is limited to a subset of patients, highlighting the need to identify the processes by which tumors evade immunity. Using CRISPR/Cas9 screening, we reveal that melanoma cells lacking HOIP, the catalytic subunit of LUBAC, are highly susceptible to both NK and CD8⁺ T‐cell‐mediated killing. We demonstrate that HOIP‐deficient tumor cells exhibit increased sensitivity to the combined effect of the inflammatory cytokines, TNF and IFN‐γ, released by NK and CD8⁺ T cells upon target recognition. Both genetic deletion and pharmacological inhibition of HOIP augment tumor cell sensitivity to combined TNF and IFN‐γ. Together, we unveil a protective regulatory axis, involving HOIP, which limits a transcription‐dependent form of cell death that engages both intrinsic and extrinsic apoptotic machinery upon exposure to TNF and IFN‐γ. Our findings highlight HOIP inhibition as a potential strategy to harness and enhance the killing capacity of TNF and IFN‐γ during immunotherapy.

HOIP protects tumor cells from apoptosis induced by combined TNF and IFN‐γ co‐secreted by NK and CD8⁺ T cells. HOIP inhibition might thus be a strategy to enhance immunotherapy responses.

Introduction

The treatment of several cancer types, particularly metastatic melanoma, has been transformed by immunotherapy. Current forms of immunotherapy include releasing the “brake” on tumor‐specific cytotoxic CD8⁺ T lymphocyte activity through blockade of suppressive checkpoint receptor/ligand interactions (Ribas et al, ²⁰¹⁶), as well as the adoptive transfer of chimeric antigen receptor T cells genetically engineered to recognize and kill target cells expressing a particular tumor antigen (Maude et al, ²⁰¹⁵). Despite the efficacy of these therapies, only a subset of patients benefit due to acquired and intrinsic resistance (Shi et al, ²⁰²⁰), highlighting the need for a better understanding of the proteins and pathways that mediate tumor immune evasion. Identification of these processes is essential for the development of novel therapeutics that may augment anti‐tumor immune responses and lead to more patients achieving durable responses to immunotherapy.

Although CD8⁺ T cells are the primary target for currently approved immunotherapies, natural killer (NK) cells are increasingly being recognized as an additional and essential cell type involved in the anti‐tumor immune response and exhibit promising potential as an additional immunotherapy target, primarily due to their capacity to potently kill major histocompatibility complex I (MHC‐I)‐negative tumor cells (Souza‐Fonseca‐Guimaraes et al, ²⁰¹⁹). CD8⁺ T cells and NK cells can directly destroy malignantly transformed cells through the release of lytic proteins, including perforin and granzymes. Additionally, these cytotoxic lymphocytes produce and secrete pro‐inflammatory cytokines upon target recognition, including tumor necrosis factor (TNF) and interferon‐gamma (IFN‐γ), which elicit pleiotropic effects within the tumor microenvironment following binding to their respective cognate receptors.

We recently conducted a genome‐wide in vitro clustered regularly interspaced short palindromic repeat (CRISPR)/CRISPR‐associated protein 9 (Cas9) screen in B16‐F10 mouse melanoma cells to identify genes, that when deleted, confer sensitivity to killing by NK cells (Freeman et al, ²⁰¹⁹). In addition to known tumor regulators of NK cell killing, this screen also identified genes classically involved in TNF and downstream nuclear factor kappa‐light‐chain‐enhancer of activated B cells (NF‐κB) signaling, including Rnf31. Rnf31 encodes for HOIL‐1‐interacting protein (HOIP), the catalytic subunit of the linear ubiquitin chain assembly complex (LUBAC), which also includes SHARPIN and HOIL‐1L, encoded by Sharpin and Rbck1, respectively (Haas et al, ²⁰⁰⁹; Tokunaga et al, ²⁰¹¹). LUBAC is an E3 ubiquitin ligase capable of forming regulatory N‐terminal methionine‐linked ubiquitin chains, also referred to as linear ubiquitination. LUBAC is required for full TNF/TNF receptor‐1 (TNFR1)‐mediated gene induction (Haas et al, ²⁰⁰⁹) and to prevent TNF/TNFR1‐induced cell death (Gerlach et al, ²⁰¹¹; Tokunaga et al, ²⁰¹¹). Previous work in multiple models supports pro‐tumorigenic and anti‐apoptotic roles for HOIP in other signaling pathways, with recent studies showing positive effects on tumor cell proliferation (Zhu et al, ²⁰¹⁴), viability (Yang et al, ²⁰¹⁴; Zhu et al, ²⁰¹⁶), and resistance to chemotherapy (MacKay et al, ²⁰¹⁴; Ruiz et al, ²⁰¹⁹). In light of these findings, we investigated HOIP in the context of anti‐tumor immunity. Here, we show that tumor HOIP limits the anti‐tumor activity of both CD8⁺ T cells and NK cells through protection against the apoptotic effects of TNF in combination with IFN‐γ.

Results and Discussion

HOIP‐deficient melanoma cells exhibit increased susceptibility to NK and CD8⁺ T‐cell‐mediated killing

Our previous CRISPR/Cas9 screen demonstrated that B16‐F10 tumor cells lacking genes associated with IFN‐γ signaling and antigen processing/presentation were depleted upon NK cell‐mediated killing (Freeman et al, ²⁰¹⁹). Genes associated with TNF and downstream NF‐κB signaling were also significantly depleted (Fig 1A), including two components of LUBAC: Rnf31 and Rbck1 (Fig 1B). To rigorously validate these findings, we performed internally controlled competition assays with purified primary mouse NK cells using an equal mixture of sgControl (mCherry⁺) and sgRnf31 (mCherry^‐) B16‐F10 ovalbumin (ova)‐expressing tumor cells engineered using CRISPR/Cas9 (Fig 1C). sgRnf31 tumor cells, which lack HOIP, were significantly depleted after NK cell co‐culture (Fig 1D). This increased sensitivity to NK cell killing was not a result of altered MHC‐I expression on sgRnf31 cells after IFN‐γ exposure and was not due to differences in basal proliferation (Figs EV1A and B).

Figure 1. HOIP‐deficient melanoma cells exhibit increased susceptibility to NK and CD8⁺ T‐cell‐mediated killing.

1. Summary of top depleted pathways and associated genes from our previous B16‐F10 dropout CRISPR/Cas9 screen performed with NK cells; **P < 0.01, ***P < 0.001, ****P < 0.0001, MAGeCK test.
2. Individual sgRNA count values from (A) for Rnf31 (left) and Rbck1 (right); significance assessed using MAGeCK test.
3. Western blot analysis on whole cell lysates showing HOIP expression for sgControl and sgRnf31 B16‐F10 ova tumor cells created using CRISPR/Cas9; β‐actin was used as a loading control.
4. Representative flow cytometry plots (left) and pooled data (right) of competition assays with sgControl and sgRnf31 B16‐F10 ova tumor cells cultured with NK cells for 72 h; **P < 0.01, Unpaired t‐test for n = 3 biological replicates. Pooled data are represented as mean ± SEM.
5. OT‐I and NK cell B16‐F10 dropout CRISPR/Cas9 screen comparison; significant genes from LUBAC are labeled.
6. Venn diagram showing the shared top significantly depleted genes from (E); LUBAC components are highlighted; indicated genes were filtered by P < 0.01 and for 3/4 affected sgRNA in both screens.
7. Representative flow cytometry plots (left) and pooled data (right) of competition assays with IFN‐γ (1 ng/ml) pre‐treated sgControl and sgRnf31 B16‐F10 ova tumor cells cultured with OT‐I CD8⁺ T cells for 72 h; ****P < 0.0001, Unpaired t‐test for n = 3 biological replicates. Pooled data are represented as mean ± SEM.

Data information: See also Fig EV1.

Source data are available online for this figure.

Figure EV1. Additional validation of HOIP‐deficient tumor cells.

1. Representative flow cytometry histograms (left) and pooled data (right) of H‐2K^b expression on sgControl or sgRnf31 B16‐F10 ova tumor cells before and after overnight exposure to IFN‐γ (5 ng/ml); *P < 0.05, Unpaired t‐test for n = 3 biological replicates. Pooled data are represented as mean ± SEM.
2. In vitro proliferation of sgControl or sgRnf31 B16‐F10 ova tumor cells over 4 days; n = 3 biological replicates. Pooled data are represented as mean ± SEM.

A role for HOIP in tumor immune evasion is further supported by the identification of Rnf31 in previous CRISPR/Cas9 screens investigating tumor cell resistance to cytotoxic T‐cell‐mediated killing (Manguso et al, ²⁰¹⁷; Pan et al, ²⁰¹⁸). A comparison of the depleted genes from a genome‐wide CRISPR/Cas9 screen performed in B16‐F10 ova cells applying ova antigen‐specific OT‐I CD8⁺ T‐cell immune pressure (Pan et al, ²⁰¹⁸), with the depleted genes from our screen applying NK cell pressure (Freeman et al, ²⁰¹⁹), highlighted Rnf31 and Rbck1 as potent protectors of both NK cell‐ and CD8⁺ T‐cell‐mediated killing (Fig 1E and F). Again, we validated this finding by performing internally controlled competition assays with activated OT‐I CD8⁺ T cells using IFN‐γ pre‐treated sgControl and sgRnf31 B16‐F10 ova tumor cells, and observed a significant depletion of sgRnf31 cells (Fig 1G). Together, these data identify, in an unbiased manner, that HOIP is a key regulator of the sensitivity of B16‐F10 melanoma cells to killing by both NK cells and CD8⁺ T cells.

HOIP‐deficient melanoma cells exhibit increased sensitivity to combined TNF and IFN‐γ

Several reports have established a role for HOIP and linear ubiquitination in protection against TNF‐induced cell death (Peltzer et al, ²⁰¹⁴; Tang et al, ²⁰¹⁸); however, the majority of these studies have focused on cells and tissues from non‐tumor origins. In the absence of HOIP, caspase‐8‐mediated apoptosis can be activated by TNF through the formation of the death‐inducing Complex II, which is generally associated with impaired Complex I‐mediated NF‐κB activation (Haas et al, ²⁰⁰⁹). Therefore, we next sought to determine whether TNF secreted by NK and CD8⁺ T cells mediated the depletion of sgRnf31 B16‐F10 ova cells in our competition assays through TNF‐induced cell death. The presence of both TNF and IFN‐γ within co‐culture supernatants prompted us to examine both single and combined effects of TNF and IFN‐γ using recombinant cytokines (Fig EV2A and B). To our surprise, no obvious differences were observed with TNF treatment as a single agent; however, we observed significant depletion of sgRnf31 B16‐F10 ova tumor cells when treated simultaneously with TNF and IFN‐γ (Fig 2A). Intriguingly, this effect was observed only when TNF and IFN‐γ were combined at the same time and could not be reproduced by prior treatment with the opposing cytokine (Fig EV2C). To confirm this result, we performed a genome‐wide CRISPR/Cas9 screen using B16‐F10 tumor cells to identify the genes depleted after extended culture with recombinant TNF and IFN‐γ. Given the previous finding, we hypothesized that multiple LUBAC components would protect against death induced by TNF in combination with IFN‐γ. In support of this, Rnf31 and Rbck1 (HOIL‐1) were among the top significantly depleted genes (Fig 2B and C; Dataset EV1). Accordingly, biological process gene ontology (GO) analysis of the top significantly depleted genes confirmed the involvement of genes implicated in positive regulation of ubiquitin‐protein transferase activity (Fig 2D), suggesting that loss of LUBAC activity sensitizes B16‐F10 tumor cells to TNF and IFN‐γ. Separate comparisons of the depleted genes from this screen with those from our NK cell (Freeman et al, ²⁰¹⁹) and the OT‐I T cell (Pan et al, ²⁰¹⁸) CRISPR/Cas9 screens revealed that Rnf31 is among the top shared significant genes (Fig 2E and F), suggesting that Rnf31 potently protects against the combined effect of TNF and IFN‐γ co‐secreted from both NK and CD8⁺ T cells. In line with this, depletion of sgRnf31 B16‐F10 ova tumor cells in the presence of supernatant derived from a separate OT‐I/B16‐F10 ova co‐culture was rescued when TNF and IFN‐γ were neutralized (Fig EV2D).

Figure EV2. Additional validation of TNF and IFN‐γ in mediating cell death of HOIP‐deficient tumor cells.

1. 24‐h IFN‐γ (left) and TNF (right) secretion in supernatant by NK cells from competition assays performed with sgControl/sgRnf31 B16‐F10 ova tumor cells; n.s.: no significance, Unpaired t‐test for n = 3 biological replicates. Pooled data are represented as mean ± SEM.
2. 24‐h IFN‐γ (left) and TNF (right) secretion in supernatant by OT‐I cells from competition assays performed with sgControl/sgRnf31 B16‐F10 ova tumor cells; **P < 0.01, ****P < 0.0001, Unpaired t‐test for n = 3 biological replicates. Pooled data are represented as mean ± SEM.
3. Outline (top) of experimental setup for cytokine pre‐treatment competition assays and pooled data (bottom) showing proportion of sgControl or sgRnf31 B16‐F10 ova tumor cells after pre‐ and post‐treatments as outlined; all cytokines were used at 10 ng/ml; **P < 0.01; Unpaired t‐test for n = 3 biological replicates. Pooled data are represented as mean ± SEM.
4. Representative flow cytometry histograms (left) and pooled data (right) of competition assays with sgControl and sgRnf31 B16‐F10 ova tumor cells after culture with supernatant (Sup), derived from a previous overnight OT‐I/B16‐F10 ova co‐culture (0.25 E:T ratio), for 24 h in the absence and presence of neutralizing α‐TNF and α‐IFN‐γ antibodies (both 10 μg/ml); n = 2 biological replicates.
5. In vitro proliferation of sgControl or sgRNF31 A375 tumor cells over 4 days; n.s.: no significance, Unpaired t‐test for n = 3 biological replicates. Pooled data are represented as mean ± SEM.

Figure 2. HOIP‐deficient melanoma cells exhibit increased sensitivity to combined TNF and IFN‐γ.

1. Representative flow cytometry plots (left) and pooled data (right) of competition assays with sgControl and sgRnf31 B16‐F10 ova tumor cells after culture with the indicated recombinant cytokine(s) (10 ng/ml) for 48 h; n.s.: no significance, ****P < 0.0001, Unpaired t‐test for n = 3 biological replicates. Pooled data are represented as mean ± SEM.
2. CRISPR/Cas9 screen showing depleted B16‐F10 genes after 7 day culture with TNF + IFN‐γ (both 10 ng/ml); blue genes indicate overall genes with P < 0.001, black genes indicate genes of interest with P < 0.05. Significance assessed using MAGeCK test.
3. Individual sgRNA count values from (B) for Rnf31 (left) and Rbck1 (right); significance assessed using MAGeCK test.
4. GO term analysis showing enriched biological processes for the top depleted genes from (B).
5. TNF/IFN‐γ and NK cell CRISPR/Cas9 screen comparison showing shared depleted genes; indicated genes are P < 0.01 and filtered for 3/4 affected sgRNA in both screens.
6. TNF/IFN‐γ and OT‐I CRISPR/Cas9 screen comparison showing shared depleted genes; indicated genes are P < 0.001 and filtered for 3/4 affected sgRNA in both screens.
7. Western blot analysis on whole cell lysates showing HOIP expression for sgControl and sgRNF31 A375 tumor cells created using CRISPR/Cas9; β‐actin was used as a loading control.
8. Representative flow cytometry plots (left) and pooled data (right) showing PI uptake of sgControl or sgRNF31 A375 tumor cells after culture with the indicated recombinant cytokine(s) (10 ng/ml) for 48 h; n.s.: no significance, *P < 0.05, **P < 0.01, Unpaired t‐test for n = 3 biological replicates. Pooled data are represented as mean ± SEM.

Data information: See also Fig EV2 and Dataset EV1.

Source data are available online for this figure.

We also validated this observation using A375 human melanoma cells, which harbor the common melanoma‐associated BRAF^V600E mutation. We directly assessed recombinant cytokine‐induced cell death using propidium iodide (PI) uptake of sgControl and sgRNF31 A375 cells engineered using CRISPR/Cas9 (Fig 2G). sgRNF31 A375 tumor cells exhibited increased sensitivity to TNF as a single agent; however, this effect was significantly increased when IFN‐γ was also present (Fig 2H). Again, this effect was independent of any effects on basal proliferation (Fig EV2E). Together, these data demonstrate that tumor‐expressed HOIP protects mouse and human melanoma cells from combined TNF and IFN‐γ‐induced cell death. These findings add to the body of evidence that HOIP possesses broader influence on cell death machinery and pathways other than exclusively driving TNF‐induced NF‐κB activity and inhibiting caspase‐8‐mediated apoptosis (MacKay et al, ²⁰¹⁴; Zhu et al, ²⁰¹⁶) and suggests that tumor cells may have evolved to possess additional mechanisms involving HOIP activity to evade cell death. Our findings also challenge the current dogma involving induction of Complex I/II activity solely by TNF and highlight the combinatorial effects with IFN‐γ, which is particularly important during anti‐tumor immune responses that engage NK cells and CD8⁺ T cells, which both co‐secrete these cytokines. Indeed, the physiological relevance of TNF and IFN‐γ synergy has been historically highlighted (Vilcek, 2009), but has been largely overlooked in recent studies investigating anti‐tumor responses involving CD8⁺ T‐cell‐secreted cytokines.

While the cytotoxic effects of TNF have been known for some time (Schulze‐Osthoff et al, ¹⁹⁹⁸), the contribution of TNF to anti‐tumor immunity and response to immunotherapy is only now being appreciated (Kearney et al, ²⁰¹⁸). As a result, strategies to lower tumor sensitivity to TNF during immunotherapy are an active area of research (Kearney et al, ²⁰¹⁷; Michie et al, ²⁰¹⁹; Vredevoogd et al, ²⁰¹⁹). A recent study utilizing similar genetic screening in human melanoma cells identified several components of TNFR1 Complex I/II as important mediators of caspase‐8‐dependent killing by CD8⁺ T‐cell‐derived TNF, including TRAF2 and BIRC2 (Vredevoogd et al, ²⁰¹⁹). Interestingly, RNF31 was not among the top depleted hits, most likely due to this screen being performed in IFN‐γ receptor‐1‐deficient melanoma cells. This finding highlights the importance of IFN‐γ signaling in conjunction with TNF during the apoptotic phenotype observed with HOIP‐deficiency in our study. Taken more broadly, this discrepancy proposes a model where specific components of Complex I may protect tumor cells from TNF‐induced apoptosis, depending on the presence or absence of proficient IFN‐γ signaling.

Although not formally assessed in our study, our genetic screening results thus far also suggest a role for Rbck1 as a regulator of cytotoxic lymphocyte killing through combined TNF and IFN‐γ. Rbck1 encodes for another component of LUBAC, heme‐oxidized iron‐responsive element‐binding protein 2 ubiquitin ligase‐1 (HOIL‐1), which is essential for complex assembly and stability within TNF Complex I (Peltzer et al, ²⁰¹⁸). Similar mechanisms for HOIP and HOIL‐1 in moderating the sensitivity of tumor cells to cytokines may be at play, which involve disruption of proficient LUBAC activity and linear ubiquitin chain formation. Destabilization of other LUBAC complex members has previously been reported upon deletion of single components (Tokunaga et al, ²⁰¹¹; MacKay et al, ²⁰¹⁴). Interestingly, the remaining component of LUBAC, Sharpin, was absent from our significantly depleted screen hits. However, this is consistent with previous work which has shown that HOIP and HOIL‐1 are sufficient to form linear ubiquitin chains (Smit et al, ²⁰¹²), and that mutation of Sharpin has less pronounced effects than loss of HOIL‐1 or HOIP in vivo (Peltzer et al, ^{2014, 2018}; Rickard et al, ²⁰¹⁴).

Combined TNF and IFN‐γ induces transcription‐dependent apoptosis in HOIP‐deficient cells that requires both intrinsic and extrinsic apoptotic machinery

We next assessed the cell death pathway involved in the enhanced death of HOIP‐deficient B16‐F10 tumor cells treated with TNF and IFN‐γ. A significant proportion of overnight TNF and IFN‐γ‐induced cell death of sgRnf31 cells was inhibited in the presence of the pan‐caspase inhibitor, QVD‐OPh (Fig 3A), suggesting caspase‐dependent apoptosis. This apoptotic cell death was confirmed by Western blot analyses, with caspase‐3 cleavage observed in sgRnf31 cells treated with combined TNF and IFN‐γ (Fig 3B, lane 20). sgRnf31 cells treated with TNF alone also exhibited evidence of caspase‐3 cleavage (Fig 3B, lane 8), however, to a lesser extent which additionally failed to translate into confirmed cell death assessed by PI uptake (Fig EV3A). We also observed caspase‐8 cleavage in sgRnf31 cells treated with combined TNF and IFN‐γ (Fig 3B, lane 20). Again, sgRnf31 cells treated with TNF alone exhibited evidence of caspase‐8 cleavage (Fig 3B, lane 8), but to a lesser extent. As expected, caspase‐8/3 cleavage was absent in sgControl cells treated with TNF alone (Fig 3B, lanes 3–5) or combined TNF and IFN‐γ (Fig 3B, lanes 15–17), and in both sgControl and sgRnf31 cells treated with IFN‐γ alone (Fig 3B, lanes 9–14). Together, these data demonstrate that HOIP inhibits caspase‐dependent apoptosis induced by combined TNF and IFN‐γ.

Figure 3. Combined TNF and IFN‐γ induces transcription‐dependent apoptosis in HOIP‐deficient cells that requires both intrinsic and extrinsic apoptotic machinery.

1. PI uptake of sgControl or sgRnf31 B16‐F10 tumor cells treated with combined TNF/IFN‐γ (both 10 ng/ml) for 24 h in the absence and presence of QVD‐OPh (10 μM); n.s.: no significance, ***P < 0.001, ****P < 0.0001, Unpaired t‐test for n = 3 biological replicates. Pooled data are represented as mean ± SEM.
2. Western blot analysis of the indicated proteins on whole cell lysates for sgControl or sgRnf31 B16‐F10 tumor cells cultured with the indicated cytokine(s) (10 ng/ml for all) for 0, 3, 6, or 9 h; β‐actin was used as a loading control.
3. Schematic outlining experimental design of CRISPR/Cas9 screening for sgRnf31 B16‐F10 resistance to killing by combined TNF/IFN‐γ (both 10 ng/ml).
4. CRISPR/Cas9 screen showing enriched genes from sgRnf31 B16‐F10 tumor cells after successive treatment with combined TNF/IFN‐γ, as described in (C); red genes indicate overall genes with P < 0.0001. Significance assessed using MAGeCK test.
5. Individual sgRNA count values from (D) for genes associated with classical IFN‐γ (top) and TNF (bottom) signaling; significance assessed using MAGeCK test.
6. GO term analysis showing enriched biological processes for the top enriched genes from (D); the top 10 terms are displayed.
7. PI uptake of sgControl, sgRnf31, sgRnf31/Stat1, sgRnf31/Casp8, sgRnf31/Bid, or sgRnf31/Rela B16‐F10 tumor cells treated with combined TNF/IFN‐γ for 24 h; *P < 0.05, ****P < 0.0001, asterisk(s) indicated group vs TNF/IFN‐γ‐treated sgRnf31 group, Unpaired t‐test for n ≥ 3 biological replicates. Pooled data are represented as mean ± SEM.

Data information: See also Fig EV3 and Dataset EV2.

Source data are available online for this figure.

Figure EV3. Additional validation of proteins involved in apoptosis of HOIP‐deficient tumor cells induced by combined TNF and IFN‐γ.

1. PI uptake of sgControl or sgRnf31 B16‐F10 tumor cells after treatment with the indicated cytokine(s) (10 ng/ml) for 24 h; n.s.: no significance, **P < 0.01, ****P < 0.0001, Unpaired t‐test for n = 3 biological replicates. Pooled data are represented as mean ± SEM.
2. Individual sgRNA count values from sgRnf31 B16‐F10 resistance screen performed with combined IFN‐γ + TNF for genes associated with DISC assembly. Significance assessed using MAGeCK test.
3. Western blot analysis of whole cell lysates showing expression of Bax and Bak in the indicated B16‐F10 cell line; β‐actin was used as a loading control.
4. PI uptake of sgControl, sgRnf31, or sgRnf31/Bax/Bak1 B16‐F10 tumor cells after treatment with IFN‐γ + TNF (both 10 ng/ml) for 24 h; n.s.: no significance, **P < 0.01, ***P < 0.001, Unpaired t‐test for n = 3 biological replicates. Pooled data are represented as mean ± SEM.
5. Western blot analysis on whole cell lysates showing expression of STAT1, caspase‐8, Bid, and p65 (left to right), in the indicated B16‐F10 cell line created using CRISPR/Cas9; β‐actin was used as a loading control.
6. PI uptake of sgControl or sgRnf31 B16‐F10 tumor cells treated with combined TNF/IFN‐γ (both 10 ng/ml) for 24 h in the absence and presence of 1400W (20 μM); n.s.: no significance, **P < 0.01, ***P < 0.001, Unpaired t‐test for n = 3 biological replicates. Pooled data are represented as mean ± SEM.

Source data are available online for this figure.

To impartially uncover factors involved in driving the enhanced cell death and caspase activity, we next performed a genome‐wide CRISPR/Cas9 screen using sgRnf31 B16‐F10 tumor cells to identify genes that, when lost, conferred resistance to killing by TNF and IFN‐γ (Fig 3C). Consistent with our data showing that a combined TNF and IFN‐γ stimulus is required to induce apoptosis in HOIP‐deficient cells, genes required for classical IFN‐γ (Ifngr1, Ifngr2, Jak1, Jak2, and Stat1) and TNF signaling (Tnfrsf1a, Ripk1, Tradd, Chuk, Ikbkb, and Rela) were both significantly enriched in this screen (Fig 3D and E; Dataset EV2). This was further supported by biological process GO term analysis of the top significantly enriched genes which revealed several terms associated with cytokine‐mediated signaling and signal transduction (Fig 3F). Additionally, the enrichment of Fadd (Fig EV3B), a gene involved in the formation of the death‐inducing signaling complex (DISC)—which is known to regulate activity of the initiator caspase, caspase‐8 (Micheau & Tschopp, 2003), suggested a contribution of classical extrinsic apoptotic machinery. Furthermore, Bid, a pro‐apoptotic member of the Bcl‐2 family of proteins, which is activated by the extrinsic apoptosis effector, caspase‐8 (Li et al, ¹⁹⁹⁸), and which in turn activates mitochondrial Bax/Bak to induce intrinsic apoptosis (Korsmeyer et al, ²⁰⁰⁰), was also significantly enriched in sgRnf31 cells upon combined TNF and IFN‐γ treatment (Fig 3C; Dataset EV2). Indeed, combined TNF and IFN‐γ‐induced apoptosis of sgRnf31 B16‐F10 tumor cells was dependent on the presence of both Bax and Bak (Fig EV3C and D). To validate these findings, we engineered sgRnf31 B16‐F10 tumor cells individually lacking Stat1, Casp8, Bid, and Rela using CRISPR/Cas9 (Fig EV3E). We confirmed dependence on these genes during apoptosis of sgRnf31 B16‐F10 tumor cells treated with TNF and IFN‐γ (Fig 3G), with consistent levels of protection observed relative to those identified in our CRISPR/Cas9 screen. These findings demonstrate that the observed cell death was dependent on both the transcriptional activity of signal transducer and activator of transcription‐1 (STAT1) and p65 (encoded by Rela), which converge on the classical apoptosis‐inducing molecules, caspase‐8, Bid, and Bax/Bak. Taken together, these data demonstrate that HOIP protects tumor cells from a transcription‐dependent form of apoptosis arising from TNF and IFN‐γ synergy that requires components of both intrinsic and extrinsic apoptotic machinery.

Although IFN‐γ itself is not generally associated with cytotoxicity, previous work has highlighted the essential role of IFN‐γ signaling during apoptosis arising from TNF and IFN‐γ synergy in tumor cells (Suk et al, ²⁰⁰¹). This synergy was recently highlighted in the context of tissue damage associated with cytokine shock during severe acute respiratory syndrome coronavirus 2 (SARS‐CoV‐2) infection (Karki et al, ²⁰²¹), where combined TNF and IFN‐γ resulted in induction of nitric oxide (NO) and the simultaneous downstream activation of several apoptotic pathways, coined PANoptosis, in bone‐marrow‐derived macrophages. In contrast to this study, we failed to inhibit the enhanced apoptotic response observed in HOIP‐deficient cells treated with combined TNF and IFN‐γ in the presence of the inducible NO synthase inhibitor, 1400W (Fig EV3F), suggesting that induction of apoptosis in our model is independent of NO activity. Similar to our findings, previous studies investigating TNF and IFN‐γ synergy have demonstrated dependence on STAT1 (Suk et al, ²⁰⁰¹; Karki et al, ²⁰²¹), linking STAT1‐mediated responses to apoptosis. Our study supports this association and suggests that tumors may exploit HOIP activity to inhibit downstream apoptosis linked to STAT1 and NF‐κB activity in the presence of combined TNF and IFN‐γ. STAT1 was recently shown to undergo linear ubiquitination modifications by HOIP, which limited type‐I IFN anti‐viral responses mediated through IFN‐α/β receptor 2 (Zuo et al, ²⁰²⁰). Although our study lacks explicit delineation of the molecular processes involving the genes identified in our cytokine resistance CRISPR/Cas9 screen performed in HOIP‐deficient tumor cells, comparable inhibitory mechanisms by HOIP may be occurring in our models, involving pro‐apoptotic STAT1 responses mediated through IFN‐γ in the presence of TNF. The dependence of classical NF‐κB signaling genes and Rela itself during our apoptotic phenotype also intriguingly mirrors that of previous reports which demonstrate the complex nature of NF‐κB signaling during tumor cell apoptosis (Ryan et al, ²⁰⁰⁰; Perkins & Gilmore, ²⁰⁰⁶).

Pharmacological inhibition of HOIP sensitizes tumor cells to combined TNF and IFN‐γ‐induced apoptosis in vitro

The emergence of data supporting the pro‐tumorigenic and anti‐apoptotic role of LUBAC has resulted in the recent development of small‐molecule inhibitors of HOIP, which reduce linear ubiquitination, canonical NF‐κB signaling, and increase TNF‐induced apoptosis in pre‐clinical studies (Katsuya et al, ²⁰¹⁹; Oikawa et al, ²⁰²⁰). We therefore sought to determine if we could phenocopy genetic deletion of HOIP with pharmacological inhibition using a recently developed HOIP inhibitor (HOIPIN), HOIPIN‐1 (Katsuya et al, ²⁰¹⁸), which has demonstrated superior specificity compared to previously used LUBAC inhibitors (Strickson et al, ²⁰¹³; Sakamoto et al, ²⁰¹⁵; Oikawa et al, ²⁰²⁰). Strikingly, pre‐treatment with HOIPIN‐1, while not intrinsically toxic, significantly sensitized otherwise resistant parental B16‐F10 tumor cells to apoptosis induced by recombinant TNF and IFN‐γ (Fig 4A). To confirm this observation on tumor cell death in a more physiological setting, we repeated this assay using supernatant from OT‐I T‐cell/B16‐F10 ova co‐cultures in place of recombinant cytokine. HOIPIN‐1 pre‐treatment similarly sensitized B16‐F10 tumor cells to co‐culture supernatants derived from various effector to target (E:T) ratios (Fig 4B). To rule out a direct effect of HOIPIN‐1 treatment on T‐cell function upon antigen recognition, we pre‐treated OT‐I T cells with HOIPIN‐1 and then co‐cultured them with B16‐F10 ova cells. We observed no differences in the levels of TNF and IFN‐γ secreted in co‐culture supernatants from HOIPIN‐1 pre‐treated OT‐I T cells compared to vehicle pre‐treated (Fig EV4A), suggesting that HOIP inhibition does not deleteriously affect cytokine secretion from activated CD8⁺ T cells. Taken together, these data demonstrate that pharmacological inhibition of HOIP can sensitize tumor cells to apoptosis induced by combined TNF and IFN‐γ. In addition to mouse melanoma cells, we also demonstrated increased sensitivity to apoptosis induced by TNF and IFN‐γ in parental A375 human melanoma cells following HOIPIN‐1 pre‐treatment (Fig 4C). This observation also extended to HeLa human cervical adenocarcinoma cells (Fig EV4B), suggesting that the effect of HOIP inhibition is not restricted to melanoma cells. Indeed, the tumor immune estimation resource (TIMER; Li et al, ²⁰¹⁷) indicated that RNF31 expression is significantly upregulated in tumors compared to normal tissue for the majority of the Cancer Genome Atlas (TCGA) cancer types where normal tissue data are available (13/17 TCGA cancer types; Fig EV4C). In light of our findings, this suggests that multiple cancer types may upregulate expression of HOIP to facilitate evasion from immune‐mediated apoptosis, which may be overcome by pharmacological HOIP inhibition depending on the level and activity of immune cell infiltrate. Importantly, TNF and IFN‐γ production and release is not limited to innate and adaptive cytotoxic lymphocytes but extends to a myriad of immune cells which can also be present and active within the tumor microenvironment.

Figure 4. Pharmacological inhibition of HOIP sensitizes tumor cells to combined TNF and IFN‐γ‐induced apoptosis in vitro .

1. Representative flow cytometry plots (left) and pooled data (right) showing PI uptake of parental B16‐F10 tumor cells left untreated or treated with recombinant TNF/IFN‐γ (both 10 ng/ml) for 48 h, following a 4‐h pre‐treatment with vehicle or HOIPIN‐1 (50 μM) (the concentration of vehicle/HOIPIN‐1 was halved upon the addition of cytokines); n.s.: no significance, **P < 0.01, Unpaired t‐test for n = 3 biological replicates. Pooled data are represented as mean ± SEM.
2. Representative flow cytometry plots (left) and pooled data (right) showing PI uptake of parental B16‐F10 tumor cells left untreated or treated with supernatant (Sup) (derived from a previous overnight OT‐I/B16‐F10 ova co‐culture at the indicated E:T ratio for 48 h, following a 4‐h pre‐treatment with vehicle or HOIPIN‐1 (50 μM) (the concentration of vehicle/HOIPIN‐1 was halved upon the addition of Sup); *P < 0.05, **P < 0.01, Unpaired t‐test for n = 3 biological replicates. Pooled data are represented as mean ± SEM.
3. Representative flow cytometry plots (left) and pooled data (right) showing PI uptake of parental A375 tumor cells left untreated or treated with TNF/IFN‐γ (both 10 ng/ml) for 48 h, following a 4‐h pre‐treatment with vehicle or HOIPIN‐1 (50 μM) (the concentration of vehicle/HOIPIN‐1 was halved upon the addition of cytokines); *P < 0.05, **P < 0.01, ****P < 0.0001, Unpaired t‐test for n = 3 biological replicates. Pooled data are represented as mean ± SEM.

Data information: See also Fig EV4.

Figure EV4. Additional HOIPIN‐1 experiments and relevance of RNF31 in other tumor types.

1. IFN‐γ (left) and TNF (right) secretion by OT‐I cells after 24‐h co‐culture with B16‐F10 ova tumor cells, following a 4‐h pre‐treatment with vehicle or HOIPIN‐1 (50 µM) (the concentration of vehicle/HOIPIN‐1 was halved upon the addition of IFN‐γ + TNF); n.s.: no significance, Unpaired t‐test for n = 3 biological replicates. Pooled data are represented as mean ± SEM.
2. Representative flow cytometry plots (left) and pooled data (right) showing PI uptake of HeLa tumor cells left untreated or treated with recombinant IFN‐γ + TNF (both 10 ng/ml) for 48 h, following a 4‐h pre‐treatment with vehicle or HOIPIN‐1 (50 µM) (the concentration of vehicle/HOIPIN‐1 was halved upon the addition of cytokines); n.s.: no significance, *P < 0.05, Unpaired t‐test for n = 3 biological replicates. Pooled data are represented as mean ± SEM.
3. RNF31 expression in TCGA tumor types assessed using TIMER v1; gray columns indicate cancer types were normal tissue or metastasis data are available; n.s.: no significance, *P < 0.05, **P < 0.01, ***P < 0.001, Wilcoxon test for n ≥ 9 biological replicates per group. Central band of boxplot represents median, box represents the interquartile range (IQR), and whiskers represent upper and lower limits relative to the IQR. Specific numbers of data points per group are displayed along the x‐axis.

HOIP‐deficient tumors exhibit enhanced OT‐I‐mediated control in vivo

In the absence of a HOIPIN that is established for systemic in vivo treatment of mice, we utilized an adoptive cell therapy model using tumors with genetic deletion of Rnf31 to test whether HOIP inhibition may be clinically feasible (Fig 5A). Consistent with our in vitro data, established sgRnf31 B16‐F10 ova tumors implanted subcutaneously into C57BL/6J mice were controlled significantly better by adoptively transferred pre‐activated OT‐I CD8⁺ T cells compared to wild‐type B16‐F10 ova tumors (Figs 5B and Fig EV5), which importantly translated into significantly extended overall survival of the mice (Fig 5C). These data provide imperative proof‐of‐principle that HOIP influences responses in vivo and suggests that inhibiting HOIP activity within tumors may enhance their response to anti‐tumor immune effector cells, such as CD8⁺ T cells.

Figure 5. HOIP‐deficient tumors exhibit enhanced OT‐I‐mediated control in vivo .

1. Experimental design of in vivo model with subcutaneous B16‐F10 ova tumors treated with adoptive OT‐I therapy.
2. Individual (left) and average (right) in vivo growth of subcutaneous wild‐type or sgRnf31 B16‐F10 ova tumors left untreated or treated with adoptively transferred OT‐I CD8⁺ T cells (on days indicated by black arrows); *P < 0.05, **P < 0.01, Unpaired t‐test for OT‐I‐treated wild‐type vs OT‐I‐treated sgRnf31 with n = 5/6 mice per group. Pooled data are represented as mean ± SEM.
3. Overall survival of mice from (B); *P < 0.05, **P < 0.01, Log‐rank test.

Data information: See also Fig EV5.

Figure EV5. Additional in vivo tumor growth data.

Additional experiment showing individual (left) and average (right) in vivo growth of subcutaneous wild‐type or sgRnf31 B16‐F10 ova tumors treated with adoptively transferred OT‐I CD8+ T cells (on days indicated by black arrows); *P < 0.05, Unpaired t‐test with n = 5/6 mice per group. Pooled data are represented as mean ± SEM.

Our findings provide a mechanistic foundation and rationale for the further investigation of pharmacological HOIP inhibition in the context of immunotherapeutic settings where TNF and IFN‐γ levels are elevated. Our in vivo therapy model demonstrating that genetic deletion of HOIP can enhance tumor control by immune effector cells supports the notion that future exploration of suitable therapeutic approaches to modulate LUBAC activity is warranted. A recently developed derivative of HOIPIN‐1, HOIPIN‐8, exhibited greater potency than HOIPIN‐1 in pre‐clinical studies and may be a suitable lead for future development in this field (Katsuya et al, ²⁰¹⁹; Oikawa et al, ²⁰²⁰). HOIPIN‐8 exhibited accelerated TNF‐induced apoptosis of a human tumor cell line in the presence of protein synthesis inhibition using cycloheximide, compared to HOIPIN‐1 in vitro (Oikawa et al, ²⁰²⁰). Notably, our results utilizing HOIPIN‐1 were performed in the absence of such inhibition and thus represent a feasible strategy to harness a transcription‐dependent form of apoptosis mediated through combined TNF and IFN‐γ. Additionally, our HOIPIN‐1 experiments assessed tumor cell killing by cytokine in the absence of any direct lymphocyte‐tumor conjugation, which may have positive implications for TNF and IFN‐γ‐induced “bystander” killing of antigen‐negative tumor cells that have downregulated essential antigen presentation proteins to evade T‐cell attack during immunotherapy (Restifo et al, ¹⁹⁹⁶; Sade‐Feldman et al, ²⁰¹⁷).

Conclusion

Taken together, we unbiasedly identify a role for tumor‐expressed HOIP, the catalytic component of LUBAC, in limiting anti‐tumor immunity by protecting against the combined apoptotic effects of TNF and IFN‐γ, released by activated NK cells and CD8⁺ T cells. Mechanistically, components of both intrinsic and extrinsic apoptotic machinery are harnessed by TNF and IFN‐γ under conditions of reduced HOIP expression. In addition, pharmacological HOIP inhibition resulted in augmented sensitivity of intrinsically resistant melanoma cells to combined TNF and IFN‐γ‐mediated apoptosis. Our collective findings not only contribute to the growing body of evidence describing the relevance of pro‐inflammatory cytokines during anti‐tumor immune responses, but also highlights HOIP as a target to increase tumor susceptibility to the apoptotic activity of combined TNF and IFN‐γ. Inhibition of tumor HOIP activity may provide a means to ultimately augment the killing capacity of both adaptive and innate cytotoxic lymphocytes during immunotherapeutic interventions.

Materials and Methods

Cell lines and primary cell cultures

B16‐F10 (RRID:CVCL_0159) mouse melanoma cells, originally from American Type Culture Collection, and its ova‐expressing derivative were cultured in DMEM supplemented with 10% FBS and incubated at 37°C with 10% CO₂. A375 (RRID:CVCL_0132) human melanoma cells were provided by the Molecular Oncology Laboratory (Peter MacCallum Cancer Centre Research Division). HeLa (RRID:CVCL_0030) human cervical adenocarcinoma cells were provided by the Organogenesis and Cancer Laboratory (Peter MacCallum Cancer Centre Research Division). A375 and HeLa cells were cultured in a similar fashion to B16‐F10 cells. Cell lines were verified to be Mycoplasma‐negative by external PCR analysis prior to the commencement of any experiments.

Primary mouse NK cells were isolated from splenocytes of 8–12‐week‐old female C57BL/6J mice using the EasySep Mouse NK Cell Isolation Kit (Stemcell Technologies; Cat:19855) according to the manufacturer’s instructions. NK cells were cultured in RPMI medium supplemented with 10% FBS, penicillin/streptomycin, l‐glutamine, non‐essential amino acids, sodium pyruvate, HEPES, β₂‐mercaptoethanol, and 1,000 IU/ml recombinant human interleukin‐2 (IL‐2) (Roche; Cat:R20‐6019) and incubated at 37°C with 5% CO₂. Purified mouse NK cells were used 6 days after culture. Splenocytes of 8–12‐week‐old C57BL/6J.OT‐I mice were activated with 10 ng/ml of the ova_257‐264 peptide, SIINFEKL (Auspep; Cat:2711), for 3 days. OT‐I cells were cultured in the same conditions as NK cells but with 100 IU/ml recombinant human IL‐2. Activated OT‐I cells were used 4 days after culture (in vivo) or 5–7 days after culture (in vitro).

Animals and in vivo experiments

All animal procedures were performed in accordance with the National Health and Medical Research Council (NHMRC) Australian Code for the care and use of animals for scientific purposes 8^th edition (2013) and with approval from the Peter MacCallum Animal Experimentation Ethics Committee. C57BL/6J mice were purchased from the Walter and Eliza Hall Institute (WEHI) of Medical Research, and C57BL/6J.OT‐I transgenic mice were bred in‐house. All experimental animals were housed in the Peter MacCallum Cancer Centre Animal Core Facility, a temperature and light controlled, pathogen‐free environment with food and water provided ad libitum.

For in vivo experiments, female 12‐week‐old C57BL/6J mice were subcutaneously injected with 2 × 10⁵ wild‐type or sgRnf31 B16‐F10 ova cells in the right flank and left to establish for 7 days. Groups containing similar tumor sizes were left untreated or intravenously injected with 2.5 × 10⁶ 4‐day cultured OT‐I T cells on days 9 and 12 post‐tumor inoculation. Tumor area was measured 3 times a week using digital calipers in a non‐blinded fashion. Mice were euthanized if they exhibited symptoms of distress or when tumor size reached ethical endpoint of 200 mm².

Generation of CRISPR/Cas9‐edited cell lines

For validation experiments, 2 × 10⁵ viable B16‐F10 ova or A375 tumor cells were added to a ribonucleoprotein (RNP) complex consisting of recombinant Cas9 nuclease (Integrated DNA Technologies; Cat:1081059) and 2 sgRNA guides (purchased from Synthego) targeting the gene of interest and electroporated using the Amaxa 4D‐Nucleofector System (Lonza). Mouse sgRNA sequences were as follows:

• sgControl: (1) AAGCACAAGAACGGTCCGCC, (2) AAAAAGTCCGCGATTACGTC;

• sgRnf31: (1) GAACTATGAGTTGTTGGACG, (2) GATGGATTGAGTTTCCCCGA;

• sgStat1: (1) GGATAGACGCCCAGCCACTG, (2) TGTGATGTTAGATAAACAGA;

• sgBid: (1) CCACAACATCCAGCCCACAC, (2) GCCAGCCGCTCCTTCAACCA;

• sgRela: (1) TATCAAAAATCGGATGTGAG, (2) GATTCCGCTATAAATGCGAG;

• sgCasp8: (1) ATGATCAGACAGTATCCCCG, (2) CAAGAAGCAGGAGACCATCG;

• sgBax: (1) GGACACGGACTCCCCCCGAG, (2) GTTTCATCCAGGATCGAGCA;

• sgBak1: (1) GGAACTCTGTGTCGTAGCGC, (2) GCAGGAGGCTCTTACCAGAA.

Human sgRNA sequences were as follows:

• sgControl: (1) AAAAAGCTTCCGCCTGATGG, (2) AAAACAGGACGATGTGCGGC;

• sgRNF31: (1) CCACCGTGCTGCGAAAGACA, (2) GGATCATGCTCACTAGCTGG.

Proteins, reagents, and antibodies

The following recombinant cytokines were purchased from PeproTech: mouse TNF (Cat:315‐01A), mouse IFN‐γ (Cat:315‐05), human TNF (Cat:300‐01A), human IFN‐γ (Cat:300‐02). PI was purchased from Thermo Fisher Scientific (Cat:P1304MP). QVD‐OPh was purchased from Merck Millipore‐Calbiochem (Cat:551476). HOIPIN‐1 was purchased from Glixx Laboratories (Cat:GLXC‐22001) and was dissolved in dimethyl sulfoxide. Anti‐mouse neutralizing α‐TNF antibody was purchased from BioLegend (Cat:506325). Anti‐mouse neutralizing α‐IFN‐γ antibody was purchased from BioXCell (Cat:BE0312).

The following primary antibodies used for Western blot analysis were purchased from Abcam, Cell Signaling Technology (CST), Merck Millipore, or Sigma‐Aldrich: HOIP (Abcam; Cat:ab46322), STAT1 (Merck Millipore; Cat:06‐501), p65 (Abcam; Cat:ab7970), Caspase‐3 (CST; Cat:9665), Cleaved caspase‐3 (Asp175) (CST; Cat:9664), β‐Actin (Sigma‐Aldrich; Cat:A2228), Caspase‐8 (CST; Cat:4927), Cleaved caspase‐8 (Asp387) (CST; Cat:8592), Bak (Sigma‐Aldrich; Cat:B5897). Bax and Bid primary antibodies were both gifts from Vivien Sutton, originally generated in‐house at WEHI (Kaufmann et al, ²⁰⁰⁷; Czabotar et al, ²⁰¹³). The following secondary antibodies used for Western blot analysis were purchased from Agilent Dako or Invitrogen: swine anti‐rabbit IgG/HRP (Agilent Dako; Cat:P0217), rabbit anti‐mouse IgG/HRP (Agilent Dako; Cat:P0260), rabbit anti‐rat IgG (H + L) (Invitrogen; Cat:61‐9520).

Combined IFN‐γ/TNF CRISPR/Cas9 screens

For the B16‐F10 dropout screen, Cas9‐expressing B16‐F10 cells were transduced with lentivirus containing the genome‐scale mouse CRISPR library, Brie, as previously described (Freeman et al, ²⁰¹⁹). Brie‐transduced B16‐F10 Cas9 cells were then cultured in the presence of recombinant TNF/IFN‐γ (both 10 ng/ml) for 7 days. An untreated control sample was cultured alongside the treated cells that was used for comparison. For the sgRnf31 resistance screen, sgRnf31 B16‐F10 cells, originally created using RNP‐based CRISPR/Cas9, were transduced to stably express mCherry‐Cas9 using the FUCas9Cherry plasmid (Addgene; Plasmid#70182) and then transduced with lentivirus containing Brie, as previously described (Freeman et al, ²⁰¹⁹). Brie‐transduced sgRnf31 B16‐F10 Cas9 cells were then subjected to 3 successive overnight cultures with recombinant TNF/IFN‐γ (both 10 ng/ml), with washing and recovery after each treatment. An untreated control sample was again used for comparison. Genomic DNA extraction, sgRNA sequence amplification by PCR, next‐generation sequencing, Model‐based Analysis of Genome‐wide CRISPR/Cas9 Knockout (MAGeCK), and visualization were performed for both screens as previously described (Freeman et al, ²⁰¹⁹). The CRISPR/Cas9 screen datasets generated during this study are included as Expanded View material (Datasets EV1 and EV2).

In vitro assays

For competition assays, sgControl or sgRnf31 B16‐F10 ova cells were equally mixed, plated, and allowed to adhere. For mouse NK competition assays, B16‐F10 ova tumor cells were co‐cultured with NK cells at a 1:1 E:T ratio for 72 h. For OT‐I competition assays, B16‐F10 ova cells were pre‐treated overnight with 1 ng/ml IFN‐γ to upregulate MHC‐I expression and then co‐cultured with OT‐I T cells at a 1:4 E:T ratio for 72 h. For recombinant cytokine competition assays, mixed B16‐F10 ova cells were cultured in the presence of TNF (10 ng/ml), IFN‐γ (10 ng/ml), or TNF/IFN‐γ (both 10 ng/ml) for 48 h. Technical triplicate samples were run on a LSRFortessa X20 flow cytometer with FACSDiva software (BD Biosciences), and the proportion of sgControl cells of viable tumor cells was assessed based on mCherry expression using FlowJo software (TreeStar, v10).

For PI assays involving recombinant cytokine and supernatant, percentage of dead cells within technical triplicate wells was assessed by the addition of 1.5 µM PI. Samples were directly run on a LSRII flow cytometer with FACSDiva software (BD Biosciences), and the proportion and gating of dead cells were based on untreated samples using FlowJo software (TreeStar, v10).

To evaluate the induction of MHC‐I on sgControl and sgRnf31 B16‐F10 ova cells, cells were left untreated or cultured in the presence of 5 ng/ml IFN‐γ for 18 h in technical triplicate determinations. Cells were washed in 2% FBS/PBS and stained with diluted anti‐mouse H‐2K^b‐PE (BD Biosciences; Cat:553570) for 30 min at 4°C. Stained cells were washed twice and then run on a LSRII flow cytometer with FACSDiva software (BD Biosciences). MFI fold change was calculated from untreated and IFN‐γ‐treated samples using FlowJo software (TreeStar, v10). To assess proliferation, 1 × 10⁴ sgControl or sgRnf31 B16‐F10 ova cells were plated in separate wells and counted for the following 4 days. For A375 cells, 5 × 10⁴ cells were initially plated.

Cytokine detection

The concentrations of TNF and IFN‐γ in supernatants from NK cell and OT‐I T‐cell co‐culture assays after 24 h were determined using a cytometric bead array Mouse Inflammation Kit (BD Biosciences; Cat:552364), according to the manufacturer’s instructions. Samples were run on a FACSVerse flow cytometer (BD Biosciences), and data were analyzed using FCAP Array software (BD Biosciences, v3).

Western blot analysis

Viable cells were directly lysed in ice‐cold SDS loading buffer (Tris‐Cl, SDS, bromophenol blue, glycerol, and β₂‐mercaptoethanol, pH 6.8), thoroughly mixed, boiled at 95°C, and centrifuged at 1 × 10⁴ g for 10 min at 4°C. Equal amounts of lysates were loaded into Tris‐glycine gels, and proteins were separated using SDS–polyacrylamide gel electrophoresis. Separated proteins were wet transferred onto Immobilon‐P PVDF membranes (Sigma‐Aldrich; Cat:IPVH00010) and blocked with 5% skim milk in 0.1% PBS‐Tween for 1 h at room temperature before overnight incubation with primary antibodies at 4°C. Membranes were then incubated with corresponding secondary antibodies for 1 h at room temperature and visualized on an iBright FL1500 Imaging System (Thermo Fisher Scientific) using Amersham ECL Western Blotting Detection Reagents (Sigma‐Aldrich; Cat:RPN2106) or Chemiluminescent Peroxidase Substrate‐3 (Sigma‐Aldrich; Cat:CPS3500‐KT).

Statistical analysis

Statistical details of experiments can be found in the figure legends, where n represents the number of biological replicates performed. With the exception of CRISPR/Cas9 screen data, GraphPad Prism 8 software was used to determine statistical significance using either an unpaired Student’s t‐test or a log‐rank test. A Wilcoxon test was used to determine statistical significance for TCGA data. Where appropriate, pooled data are represented as mean ± SEM, and differences were considered significant when P < 0.05.

Author contributions

Conceptualization: AJF; Methodology: AJF, SJV, CJK, and JO; Investigation: AJF, JM, and KMR; Visualization: AJF; Writing – Original Draft: AJF; Writing – Review & Editing: AJF, CJK, JS and JO; Funding Acquisition: JO; Supervision: CJK and JO. All authors read and approved the final manuscript.

Conflict of interest

The authors declare that they have no conflict of interest.

Supporting information

Expanded View Figures PDF

Dataset EV1

Dataset EV2

Source Data for Expanded View

Source Data for Figure 1

Source Data for Figure 2

Source Data for Figure 3

EMBO reports (2021) 22: e53391.

Data availability

No primary datasets have been deposited to a public database.