Intratumoral activation of the necroptotic pathway components RIPK1 and RIPK3 potentiates antitumor immunity

See allHide authors and affiliations

Science Immunology  21 Jun 2019:
Vol. 4, Issue 36, eaaw2004
DOI: 10.1126/sciimmunol.aaw2004

Using cell death to resurrect antitumor immunity

Dying cells can trigger activation of the immune system. Here, Snyder et al. have engineered cells that can be induced to undergo necroptotic cell death. By injecting these cells directly into tumors, they have examined the ability of these dying cells to promote antitumor responses in situ and at a distant site harboring syngeneic tumor. In addition to promoting immune response in situ, these injected cells drove a systemic immune response driven by conventional dendritic cells 1 (cDC1s) and CD8+ T cells that promoted regression of tumor at the distant site as well. Most impressively, they found that the dying cells did not have to express tumor-specific antigens to promote antitumor immunity.


Although the signaling events that induce different forms of programmed cell death are well defined, the subsequent immune responses to dying cells in the context of cancer remain relatively unexplored. Necroptosis occurs downstream of the receptor-interacting protein kinases RIPK1 and RIPK3, whose activation leads to lytic cell death accompanied by de novo production of proinflammatory mediators. Here, we show that ectopic introduction of necroptotic cells to the tumor microenvironment promotes BATF3+ cDC1− and CD8+ leukocyte–dependent antitumor immunity accompanied by increased tumor antigen loading by tumor-associated antigen-presenting cells. Furthermore, we report the development of constitutively active forms of the necroptosis-inducing enzyme RIPK3 and show that delivery of a gene encoding this enzyme to tumor cells using adeno-associated viruses induces tumor cell necroptosis, which synergizes with immune checkpoint blockade to promote durable tumor clearance. These findings support a role for RIPK1/RIPK3 activation as a beneficial proximal target in the initiation of tumor immunity. Considering that successful tumor immunotherapy regimens will require the rational application of multiple treatment modalities, we propose that maximizing the immunogenicity of dying cells within the tumor microenvironment through specific activation of the necroptotic pathway represents a beneficial treatment approach that may warrant further clinical development.


Tumor immunotherapy, which boosts the ability of the body’s own immune system to recognize and kill transformed cells, constitutes an immensely promising advance in the modern treatment of cancer. The efficacy of existing T cell–targeted therapies such as immune checkpoint blockade (ICB) can often be boosted upon coadministration of cytotoxic treatments such as irradiation (1, 2). However, the specific forms of programmed cell death (PCD) initiated upon administration of cytotoxic therapies to tumor cells are often not rigorously defined (3). Considering the growing body of evidence supporting differential immune activation or suppression in response to distinct PCD modalities (4), strategies to maximize the immunogenicity of dying tumor cells could potentially function to boost the effects of coadministered treatments including ICB.

Cells can undergo distinct forms of PCD in response to cellular stress, pathogen infection, and organismal development (5, 6). Apoptosis occurs after activation of a family of proteases termed caspases, and the clearance of apoptotic debris is often associated with tolerogenic signaling (7). These immunomodulatory processes include the caspase-directed inactivation of immunostimulatory damage-associated molecular patterns (DAMPs) such as high-mobility group box 1 protein (8), as well as immunosuppressive functions of the Tyro3/Axl/Mertk receptor tyrosine kinases in promoting tissue repair phenotypes in phagocytes that have engulfed apoptotic debris (9). Apoptosis is believed to be the mechanism of PCD in tumor cells after administration of a wide variety of anticancer drugs, including chemotherapeutic agents (10, 11) and specific inducers of apoptosis (1214). Induction of immune tolerance by apoptotic cells may therefore limit synergistic effects when combining these anticancer compounds with ICB or other immunotherapy regimens.

Necroptosis is a form of PCD that occurs downstream of the receptor-interacting protein kinases RIPK1 and RIPK3, which assemble into an oligomeric complex termed the “necrosome” (15, 16). A growing body of evidence supports the idea that necroptosis is a more potently immunogenic form of PCD than apoptosis in certain contexts (4). Necroptotic cells undergo rapid membrane permeabilization via the executioner protein mixed-lineage kinase-like (MLKL), leading to the release of intracellular contents including immunogenic DAMPs that can activate innate immune pattern recognition receptors (PRRs) (1719). Furthermore, death-independent functions of RIPK3 have also been recently defined, including inflammatory chemokine and cytokine production that can promote cross-priming of CD8+ T cell vaccination responses (20) and confer protection during viral infection (21). Therefore, a model emerges in which necroptosis can function as an alternative PCD modality that can eliminate caspase-compromised cells in the event of infection while simultaneously releasing a payload of inflammatory signals to recruit and activate immune cells (22). These findings have not yet been comprehensively applied to the field of tumor immunology, in part due to technical limitations related to the manipulation of PCD programs in vivo. Specific targeting of necroptosis using endogenous signaling components is difficult, because there is extensive regulatory cross-talk between extrinsic apoptotic and necroptotic signaling pathways (16). This is further complicated by the fact that many tumors have mutated or silenced either caspases (23) or the RIPKs (24). Given these obstacles, the specific differential effects of enforced RIPK3 activation versus caspase-8 or caspase-9 activation within the tumor microenvironment (TME) have not been described.

Here, we describe a beneficial role for activation of the necroptotic pathway components RIPK1 and RIPK3 within the TME. Using engineered versions of pro-death enzymes, we present a reductionist system that circumvents endogenous pro-death signaling pathways within tumor cells. Ectopic activation of RIPK3 promotes tumor antigen loading by tumor antigen-presenting cells (APCs) associated with enhanced CD8+ leukocyte–mediated antitumor responses, which leads to systemic tumor control that synergizes robustly with ICB coadministration. These beneficial effects occur specifically after administration of necroptotic cells within solid tumors but not after exposure to apoptotic cells or cells dying via lytic necrosis, indicating that these protective effects are due to signals specifically derived from the RIPK1/RIPK3 necrosome complex. We also present a tractable system for the induction of necroptosis in tumor cells in situ using engineered adeno-associated viruses (AAVs), which successfully recapitulate tumor control effects after necroptosis initiation. Collectively, these findings demonstrate that RIPK1/RIPK3 activation in established solid tumors promotes robust antitumor immunity.


Necroptotic cells confer tumor control across multiple syngeneic flank tumor models

To assess the impact of necroptotic tumor cell death on gross tumor outgrowth responses, we used a model of intratumoral dying cell administration that allowed us to precisely control the timing and number of cells undergoing various cell death pathways within the TME. We used constructs encoding chimeric versions of pro-death proteins fused to activatable (“ac”) FKBPF36V domains, which we have previously shown allow activation of the chimeric protein after incubation with a synthetic bivalent homolog of rapamycin that functions as a nontoxic ligand (25). Tumor cells transduced with activatable versions of either proapoptotic caspase-8 (acCASP8), proapoptotic caspase-9 (acCASP9), or pronecroptotic RIPK3 (acRIPK3) (Fig. 1A) were pulsed with ligand drug in vitro to enforce oligomerization of these pro-death enzymes and then injected intratumorally into pre-established syngeneic flank tumors. In this system, ectopically administered cells are alive at the time of injection but are fated to undergo respective forms of PCD within the TME, accompanied by any signaling activity induced downstream of acCASP8, acCASP9, or acRIPK3. Transduction with proapoptotic acCASP9 was better tolerated in tumor cell lines, whereas acCASP8 was better tolerated in fibroblast cell lines. Using this model, we observed that administration of autologous necroptotic (acRIPK3), but not apoptotic (acCASP9), tumor cells into cell type–matched tumors conferred control of tumor outgrowth and extended survival of animals bearing either B16.F10-ovalbumin (OVA) melanoma flank tumors (Fig. 1B and fig. S1B) or Lewis lung (LL/2)–OVA adenocarcinoma flank tumors (Fig. 1C and fig. S1C).

Fig. 1 Necroptotic cells confer tumor control across multiple syngeneic flank tumor models.

(A) Schematic of pro-death enzyme constructs and respective types of PCD induced downstream after enzyme activation with B/B homodimerizer. (B to F) Tumor growth of B6/J mice bearing (B and D) B16.F10-OVA, (C and E) LL/2-OVA, or (F) E.G7-OVA flank tumors after administration of apoptotic or necroptotic (B and C) autologous or (D to F) unmatched NIH 3T3 fibroblast cells. n = 10 to 16 mice per group. (G and H) Tumor growth of B6/J mice bearing (G) B16.F10 or (H) LL/2 flank tumors after administration of necroptotic NIH 3T3 cells. n = 9 to 14 mice per group. (I) Tumor growth of ipsilateral (“I,” treated) and contralateral (“C,” untreated) B16.F10-OVA tumors after administration of either apoptotic or necroptotic NIH 3T3 cells (left and middle panels) and survival curve of mice from the same experiment (right panel). n = 9 to 11 mice per group. **P < 0.01, ***P < 0.001, and ****P < 0.0001. Black arrows indicate intratumoral dying cell injections. Error bars represent SEM. Data are pooled from three to five independent experiments.

Unexpectedly, the tumor control effects of necroptosis did not require that necroptotic cells themselves carry tumor antigens, because injection of an unrelated fibroblast line, NIH 3T3, similarly led to tumor control and extension of animal survival after necroptotic, but not apoptotic (acCASP8) fibroblast administration in B16.F10-OVA (Fig. 1D and fig. S1D), LL/2-OVA (Fig. 1E and fig. S1E), and E.G7-OVA thymoma (Fig. 1F and fig. S1F) flank tumors (26). These tumor control effects were not an artifact of the immunodominant OVA epitope expressed by tumor cells, because administration of necroptotic fibroblasts also conferred tumor outgrowth and survival extension in mice implanted with non–OVA-expressing B16.F10 (Fig. 1G and fig. S1G) or LL/2 (Fig. 1H and fig. S1H) flank tumors. These data indicate that necroptotic cells delay tumor outgrowth and even confer complete tumor clearance in a small percentage of animals across multiple syngeneic flank tumor models.

Considering these findings, we next tested whether this treatment could act systemically in a bilateral flank tumor model, where a single mouse is implanted with separate B16.F10-OVA tumors on either flank. After injection of necroptotic fibroblasts into a treated (ipsilateral) tumor, we observed control of flank tumor outgrowth in both the ipsilateral and the untreated (contralateral) tumors (Fig. 1I), leading to an extension of animal survival (Fig. 1I). Control of both tumors in this bilateral tumor model indicates an abscopal effect, whereby application of a therapeutic agent to a primary tumor can lead to the control and even elimination of distal, untreated metastases. This effect required tumor antigen matching between the ipsilateral and contralateral tumors, because mice implanted with a B16.F10-OVA ipsilateral tumor and an antigenically disparate LL/2 contralateral tumor failed to exhibit abscopal tumor control after administration of necroptotic fibroblasts (fig. S1I). These data suggest that the introduction of necroptotic cells to the TME initiates a systemic immune response to tumor-derived antigens, irrespective of antigen matching between necroptotic cells and the tumor cells themselves.

Tumor control by necroptotic cells requires BATF3+ cDC1 and CD8+ leukocytes

Given the unexpected finding that introduction of necroptotic cells to the TME promoted systemic tumor control irrespective of antigen matching, we sought to establish that the control we observed was immune mediated. BATF3 is a transcription factor required for the development of conventional dendritic cells 1 (cDC1), which are critical for cross-presentation of exogenous antigens to stimulate CD8-mediated immunity and are required for endogenous antitumor immune responses (27). We found that the tumor control effects of necroptotic fibroblasts required BATF3+ cDC1, because Batf3−/− mice failed to restrict B16.F10-OVA (Fig. 2A) or LL/2-OVA (fig. S2A) tumor growth compared with controls. Consistent with the critical role for cDC1 in mediating antitumor immunity, we also observed that depletion of CD8+ leukocytes (including both CD8α+ cytotoxic T cells and CD8α+ DCs) completely abrogated the therapeutic effect of necroptotic fibroblast administration, whereas depletion of CD4+ leukocytes did not affect tumor control responses (Fig. 2B); depletion of each leukocyte population was confirmed via flow cytometry (fig. S2B). These data indicate that control of tumors after introduction of necroptotic cells is immune mediated and proceeds via activation of CD8+ leukocytes.

Fig. 2 Tumor control by necroptotic cells requires BATF3+cDC1 and CD8+leukocytes.

(A) Day 12 B16.F10-OVA tumor volumes (left) and animal survival (right) of mice with varying Batf3 genotypes after necroptotic fibroblast injections. n = 4 to 12 mice per group. d.11, day 11. (B) B16.F10-OVA tumor growth (left) and animal survival (right) upon coadministration of necroptotic fibroblasts with depleting antibodies against either CD4+ or CD8+ leukocytes. n = 6 to 11 mice per group. *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001. Black arrows indicate intratumoral dying cell injections. Error bars represent SEM. Data are pooled from two to four independent experiments.

Immune-mediated tumor control by necroptotic cells requires NF-κB activation within dying cells but not DAMP release

We next investigated the signals emanating from necroptotic cells within the TME that promoted immune activation and tumor control. Because necroptosis is a lytic form of cell death, necroptotic cells might initiate immune responses through the release of DAMP molecules within the TME. To assess this possibility, we first tested the ability of necroptotic fibroblasts to recapitulate tumor control effects in several knockout mouse strains, whose tumor-infiltrating immune cells lack expression of PRRs capable of recognizing DAMPs or their downstream signaling components. We found that mice deficient in signaling components involved in cytosolic DNA sensing (Tmem173−/−, Mb21d1−/−, and Aim2−/−), cytosolic RNA sensing (Mavs−/−), Toll-like receptor (TLR) signaling (Myd88−/−, Ticam1−/−, and Irf3−/−), or general inflammation (Tnf−/−) all retained the ability to control tumor outgrowth after administration of necroptotic fibroblasts into either B16.F10-OVA (Fig. 3A) or LL/2-OVA (fig. S3A) tumors, indicating that the therapeutic effects of necroptotic cells are not strictly mediated through the singular activity of these innate immune signaling components within tumor-infiltrating leukocytes. Furthermore, antibody-mediated blockade of the necrotic cell sensor CLEC9A (28) did not affect tumor restriction or animal survival extension by necroptotic fibroblasts (Fig. 3B); effective blockade of CLEC9A expression was confirmed on dendritic cell (DC) subsets in both the spleen and tumor (fig. S3B). Together, these results indicate that tumor control by necroptotic cells is not mediated solely through the activation of any of these individual innate immune signaling pathways.

Fig. 3 Immune-mediated tumor control by necroptotic cells requires NF-κB activation within dying cells but not MLKL-mediated cell lysis and DAMP release.

(A) Day 12 volumes of B16.F10-OVA tumors in various innate immune knock-out mice after necroptotic fibroblast injections on days 6, 8, and 10. n = 5 to 15 mice per group. (B) B16.F10-OVA tumor growth (left panel) and animal survival (right panel) upon coadministration of necroptotic fibroblasts with blocking α-CLEC9A antibody. n = 10 mice per group. (C and D) Tumor growth and overall survival after administration of lytic necrotic fibroblasts in (C) single or (D) contralateral B16.F10-OVA flank tumors. Tumor growth and survival curves for PBS, acCASP8, and acRIPK3 as presented in Fig. 1 and are also graphed for comparison. (E) B16.F10-OVA tumor growth curves after injection of necroptotic NIH 3T3 fibroblasts preincubated with the IκBα phosphorylation inhibitor BAY-117085, which prevents NF-κB activation in treated cells. n = 10 mice per group. (F) B16.F10-OVA tumor growth (left) and animal survival (right) after intratumoral injection of PBS, necroptotic fibroblasts, or MLKL−/− necroptotic fibroblasts. n = 7 to 9 mice per group. (G) Assessment of systemic inflammation via Luminex assay for inflammatory serum cytokines and chemokines 48 hours after intratumoral dying NIH 3T3 injection. DMXAA-injected mice were included as a positive control for systemic inflammatory cytokine production. The gray dashed line represents the limit of detection. n = 3 to 5 mice per group. (H) B16.F10-OVA tumor growth curves after intratumoral (IT), intraperitoneal (IP), distal subcutaneous (distal subQ), or intravenous (IV) injection of necroptotic fibroblasts. n = 7 to 9 mice per group. *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001. Black arrows indicate intratumoral dying cell injections. Error bars represent SEM. Data are pooled from two to five independent experiments. ns, not significant.

Our previous findings indicate that RIPK3 activation induces nuclear factor κB (NF-κB)–mediated transcriptional responses in addition to lytic cell death (20). We therefore sought to separate these two potentially immunostimulatory processes to understand the contribution of each to necroptosis-mediated tumor control. To do this, we first administered fibroblasts dying via either necroptosis or lytic necrosis into established B16.F10-OVA tumors, using three different forms of lytic necrotic fibroblasts (fig. S1A): (i) cells expressing a mutated version of activatable RIPK3 lacking the RIP homotypic interaction motif (RHIM) domain (acRIPK3ΔC), which cannot recruit and activate RIPK1 to induce downstream NF-κB–mediated inflammatory gene transcription yet maintains the ability to activate MLKL to induce pore formation and lytic cell death (20); (ii) cells expressing an activatable version of MLKL (acMLKL) to induce pore formation and lytic cell death in the absence of upstream RIPK3 activation (29); and (iii) cells that were mechanically lysed via repeated freeze-thaw cycles immediately before injection. All three forms of lytic necrotic cells similarly release cell-associated DAMPs due to loss of plasma membrane integrity but lack activation of the RIPK1/NF-κB signaling axis that is otherwise observed upon activation of full-length RIPK3 in necroptotic NIH 3T3 cells (20).

We observed that all three treatments of lytic necrotic fibroblasts failed to confer tumor control and extend animal survival compared to fibroblasts dying via acRIPK3-mediated necroptosis, both in single B16.F10-OVA tumors (Fig. 3C) and using acRIPK3ΔC fibroblasts in single LL/2-OVA (fig. S3C, left panels) or E.G7-OVA tumors (fig. S3C, right panels). Consistent with this, administration of acRIPKΔC fibroblasts failed to confer tumor control (Fig. 3D) and extension of survival (fig. S3D) in bilateral B16.F10-OVA tumor–bearing mice. These results revealed that DAMP release is not sufficient for the tumor control effects of necroptotic fibroblasts within the TME, suggesting instead that signaling activities downstream of RIPK1/RIPK3 necrosome complex formation and activation may play a role.

To more directly test whether the immunogenicity of necroptotic fibroblasts in our tumor model depended on intact NF-κB activation in the dying cells, we preincubated NIH 3T3 cells +acRIPK3 with an irreversible inhibitor of nuclear factor κB (NFκB) signaling, BAY-117085, before pulsing with activator drug and injection into B16.F10-OVA tumors. Inhibition of NF-κB activation via BAY-117085 significantly reduced both the tumor control effects (Fig. 3E) and survival advantage (fig. S3E) conferred by necroptotic cells compared to vehicle-treated controls, indicating that NF-κB activity in necroptotic fibroblasts is required for their antitumor effects. To test whether the acRIPK3–NF-κB axis was sufficient to drive the tumor control response in the absence of lytic cell death by these cells, we next generated NIH 3T3 cells +acRIPK3 that lacked expression of the necroptosis executioner MLKL (fig. S3F, left panel). These cells failed to undergo cell death upon exposure to activator drug in vitro, unless expression of the prosurvival protein cellular FADD-like IL-1beta-converting enzyme-inhibitory protein (cFLIP) was concurrently knocked down (fig. S3F, right panel), consistent with reverse signaling by the RIPK3 complex to induce apoptosis (29, 30). Injection of MLKL−/− NIH 3T3 +acRIPK3 cells into B16.F10-OVA tumors conferred similar tumor control (Fig. 3F, left panel) and survival extension (Fig. 3F, right panel) responses as seen in mice that received MLKL-sufficient necroptotic fibroblasts. Together, these results indicate that transcriptional signaling downstream of RIPK1/RIPK3/NF-κB activation is responsible for the immunogenicity of necroptotic fibroblasts in the TME and that this occurs independently of DAMP release by dying cells.

Our data point to the production of NF-kB–dependent cytokines as key to tumor control by necroptotic cells. Because these cytokines can act both locally and systemically, we next tested whether the introduction of necroptotic cells to the TME might drive systemic inflammatory or immune responses. To do this, we measured levels of inflammatory mediators in sera harvested from B16.F10-OVA tumor–bearing mice after necroptotic cell administration. There were no differences between treatment groups with respect to systemic levels of inflammatory chemokines and cytokines relevant for antitumor responses, including interferon-γ (IFN-γ), tumor necrosis factor–α (TNF-α), CCL5, and CXCL10 (Fig. 3G), or chemokines and cytokines known to be produced by necroptotic NIH 3T3 fibroblasts, including interleukin-6 (IL-6), CXCL1, and CCL2 (fig. S3G) (20). Consistent with this, injection of necroptotic fibroblasts into spatially distinct locations distal from the tumor site, including intraperitoneally, intravenously, or subcutaneously on the opposite flank to the tumor, all failed to confer tumor outgrowth control (Fig. 3H) or extend animal survival (fig. S3H) compared with intratumoral injection of necroptotic fibroblasts. These data indicate that administration of necroptotic fibroblasts does not lead to tumor control through nonspecific systemic inflammation, suggesting that the therapeutic effect of this treatment is due to local mechanisms exerted specifically within the TME.

Necroptosis promotes antitumor CD8+ T cell responses and synergizes with ICB

We next sought to understand the nature of the immune response instigated by introduction of necroptotic cells to the TME. To do this, we first assessed the effects of dying cell administration on cytotoxic CD8+ T cells, a critical mediator of antitumor immunity. Using flow cytometric analysis to identify subsets of OVA-specific (SIINFEKL-H2Kb+) CD8+ T cells isolated from B16.F10-OVA tumors (fig. S4A), we observed increased numbers of OVA-specific T cells expressing markers of proliferation (Ki67+), effector function (GranzymeB), and general activation (CD44hi) after necroptotic (acRIPK3), but not apoptotic (acCASP8) or lytic necrotic (acRIPK3ΔC), fibroblast administration (Fig.4A). OVA-specific CD8+ T cells isolated from necroptotic cell–exposed tumor tissue displayed similarly elevated percentages of both CD44hi and programmed cell death protein 1 (PD-1)+ cells (fig. S4B, left and middle panels), indicating that necroptosis correlated with an overall more activated surface phenotype of intratumoral CD8+ T cells, although these CD8+ T cells did not express higher levels of PD-1 on a per cell basis compared with cells exposed to apoptotic or lytic necrotic fibroblasts (fig. S4B, right panel). Furthermore, we observed significant increases in the ratios of both activated (CD44hi) or tumor-specific (SIINFEKL-H2kb+) CD8+ T cells to CD25+Foxp3+ regulatory T cells (Tregs) (Fig. 4B) specifically within tumors that received necroptotic fibroblasts, indicating that the profile of tumor-infiltrating T cells was skewed toward more favorable cytotoxic CD8+ T cells, rather than an immunosuppressive profile dominated by Treg. These data indicate that exposure to necroptotic cells within the TME is associated with increased numbers of tumor-specific CD8+ T cells present in the tumor tissue.

Fig. 4 Necroptosis promotes antitumor CD8+T responses and synergizes with ICB.

(A) Absolute numbers of intratumoral CD8+ T cells with various phenotyping markers for proliferation (Ki67), effector function (GranzymeB, GzmB), and activation (CD44), normalized per gram of tumor tissue. (B) Quantification of the ratio of intratumoral activated (CD44hi, left) or tumor antigen–specific (SIINFEKL-H2Kb+, right) CD8+ T cells to immunosuppressive Foxp3+ CD25+ Treg, normalized per gram of tumor tissue. (C) Sample flow plots (left) and percentages (right) of overall activated CD8+ T cells in the tumor-draining (inguinal) lymph node. (D) Sample flow plots (left) and percentages (right) of OVA-specific and activated CD8+ T cells in the tumor-draining (inguinal) lymph node. (E) B16.F10-OVA tumor growth (left) and animal survival (right) after coadministration of necroptotic fibroblasts with the lymphocyte trafficking inhibitor FTY-720. n = 8 to 12 mice per group. (F) Tumor growth of ipsilateral (treated, left) and contralateral (untreated, right) B16.F10-OVA tumors after administration of necroptotic NIH 3T3 cells with FTY-720. n = 9 to 10 mice per group. (G) Survival curves of B16.F10-OVA tumor–bearing mice after coadministration of necroptotic fibroblasts with the ICB reagent α-PD-1 or IgG2a isotype. n = 8 to 14 mice per group. (H) Left: Schematic of tumor rechallenge experiments in mice from (G) that successfully cleared B16.F10-OVA tumors. Right: Survival of mice rechallenged with B16.F10-OVA cells on the same flank as initial tumor location. n = 10 mice per group. *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001. All flow harvests performed 48 hours after dying cell injection. Error bars represent SEM. Data are pooled from two to four independent experiments.

To characterize the effects of necroptotic cell administration on lymph node priming, we concurrently examined the abundance and quality of CD8+ T cell responses in the tumor-draining lymph node (tdLN) of these mice (fig. S4C). We observed an increased frequency (Fig. 4C) and number (fig. S4D) of overall activated (defined as CD44hi CD62Llo) CD8+ T cells in the tdLN of mice that received intratumoral necroptotic fibroblasts. These were accompanied by increases in the numbers of bulk CD8+ and single-positive CD44hi CD8+ T cells, but not of CD69+ CD8+ T cells (fig. S4D). We also observed similar increases in the frequency (Fig. 4D) and number (fig. S4D) of activated, tumor-specific (defined as CD44hi SIINFEKL-H2kb+) CD8+ T cells in the tdLN of necroptotic cell–treated mice. Therefore, in addition to an expansion of favorable CD8+ T cell phenotypes locally within the TME, necroptotic fibroblast injection also resulted in lymph node priming of tumor-reactive cytotoxic CD8+ T cells.

To test whether recruitment of these newly primed CD8+ T cells into the TME was required for necroptotic cells to exert tumor control, we coadministered necroptotic fibroblasts with the sphingosine-1–phosphate receptor modulator fingolimod (FTY-720) to inhibit egress of lymphocytes from the tdLN. Blockade of lymphocyte trafficking did not affect B16.F10-OVA tumor control by necroptotic fibroblasts, because FTY-720–treated animals still exhibited effective control over tumor outgrowth and extension of animal survival compared with vehicle-treated controls (Fig. 4E). Consistent with a lack of influx of newly primed lymphocytes, enumeration of various tumor-associated lymphocyte populations isolated from B16.F10-OVA tumors 48 hours after dying cell administration revealed similar total numbers of CD19+ B cells, CD4+ T cells, and CD8+ T cells within the TME among tumors that received apoptotic (acCASP8), necroptotic (acRIPK3), or lytic necrotic (acRIPK3ΔC) cell injections (fig. S4E).

Inhibition of lymph node egress by primed leukocytes also did not alter abscopal tumor control in bilateral B16.F10-OVA tumors (Fig. 4F and fig. S4F), suggesting that untreated contralateral tumor control could be mediated by recirculating leukocytes in the periphery. These results show that rapid recruitment of tumor-reactive lymphocytes from the tdLN is not required for growth restriction of treated or distal, untreated tumors responding to necroptotic fibroblast exposure and further implicate local effects of necroptotic cells within the TME.

The efficacy of ICB is often boosted upon coadministration with cytotoxic therapies, including irradiation (2). Because stimuli from necroptotic cells boosted activation of tumor-specific CD8+ T cells both in tumor and in tdLN, we hypothesized that the presence of necroptotic cells within the TME that would synergize with ICB, specifically α-PD-1. To test this, we interleaved injections of necroptotic fibroblasts into B16.F10-OVA flank tumors with administration of α-PD-1 and observed that mice exhibited significantly improved survival outcomes (Fig. 4G) and improved tumor growth restriction (fig. S4G, left panel), because 71.4% of mice successfully cleared their tumors (fig. S4G, right panel) after this coadministration regimen. To determine whether this successful combination therapy conferred protective immune memory, we rechallenged mice ~2 months after they successfully cleared their tumors, injecting identical tumor cells into the same flank that bore the initial B16.F10-OVA tumor (Fig. 4H, left panel). All mice (100%) were protected from tumor rechallenge (fig. S4H) and failed to succumb to tumor outgrowth compared with naïve B6/J controls (Fig. 4H, right panel). Together, these data indicate that necroptosis in the TME can potently synergize with ICB coadministration to promote durable tumor rejection.

Exposure to necroptosis in the TME promotes antigen uptake and activation of tumor-associated APCs

Our data indicate that necroptosis potentiates antitumor CD8+ T cell responses even when necroptotic cells do not contain tumor antigen, implicating broad activation of tumor-associated APCs as the key effect of necroptotic cells within the TME. We therefore aimed to define necroptosis-induced changes to tumor-associated myeloid cell populations that could function to initiate adaptive immunity. Using a previously published gating strategy to identify subsets of tumor-associated innate immune cells (fig. S5C) (31), we enumerated various innate immune cells isolated from B16.F10-OVA tumor tissue after administration of apoptotic (acCASP8), lytic necrotic (acRIPK3ΔC), or necroptotic (acRIdPK3) fibroblasts. We observed a significant increase in the number of CD24+ CD103+ DC1 (Fig. 5A), although there were no significant differences in the number of Ly6Chi monocytes, NK1.1+ natural killer (NK) cells, bulk major histocompatibility complex class II+ (MHCII+) APCs, F4/80+ macrophages, or CD24+ CD11b+ DC2 (Fig. 5A and fig. S5A). This was promising, given that CD103+ DC1 are often viewed as the most functional tumor APC subset with respect to stimulating CD8+ T cell–mediated antitumor immunity (3133). Consistent with this increase in intratumoral CD103+ DC1, we measured significantly elevated levels of the DC chemoattractants CCL3, CCL4, and CCL5 in tumor homogenates after exposure to necroptotic fibroblasts (Fig. 5B). Considering that CD103+ DC1 can be recruited to the TME via NK cell–derived chemokines (34), we tested whether depletion of NK cells (fig. S5B, right panel) abrogated the therapeutic effect of necroptotic fibroblasts. NK cell depletion had no effect on tumor control and survival extension (fig. S5B, left panel) by necroptotic fibroblasts.

Fig. 5 Exposure to necroptosis in the TME promotes antigen uptake and activation of tumor-associated APCs.

(A) Absolute numbers of tumor-associated DC subsets 48 hours after intratumoral dying cell administration, normalized per gram of tumor tissue. (B) Intratumoral concentrations of DC-recruiting chemokines 24 hours after dying NIH 3T3 injection. The gray dashed line represents the limit of detection. n = 3 to 5 mice per group. (C) Experimental schematic of B16.F10-OVA tumor cells expressing zsGreen as a surrogate tumor antigen, allowing for gating on tumor APCs that have phagocytosed tumor antigen. (D) Percent of zsGreen+ tumor APCs after dying cell administration. (E) gMFI of CLEC9A receptor on subsets of tumor DC populations. n = 4 to 6 mice per group. (F) gMFI of PD-L1 on subsets of tumor APC populations and CD45 zsGreen+ tumor cells. n = 4 to 6 mice per group. (G) gMFI of the costimulatory marker CD80 on zsGreen+ subsets of tumor APC populations. n = 3 to 4 mice per group. (H) Quantification of previously activated OT-I T cell proliferation upon coculture with zsGreen+ tumor APC subsets sorted ex vivo from B16.F10-OVA-zsGreen tumors after dying cell injection. n = 3 technical replicates per group, using pooled cells from five mice per treatment group. (I) In vitro characterization of BMDMs cocultured with live or necroptotic B16.F10-zsGreen tumor cells and dextran-fluorophore beads, assessed for phagocytosis via dextran uptake (left) and expression of costimulatory marker CD80 in zsGreen+ BMDMs (middle) or dextran+ beads (right). n = 3 technical replicates per group. *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001. All flow harvests performed 48 hours after dying cell injection. Error bars represent SEM. Data are representative plots from one to three independent experiments (E to I) or pooled from two to three independent experiments (A, B, and D).

We next evaluated the phenotype of phagocytic tumor APCs with respect to tumor antigen loading and their activation status. To do this, we implanted mice with B16.F10-OVA cells that also express the bright and stable fluorophore zsGreen, then gated on zsGreen+ tumor-associated phagocytes to identify tumor APCs that have ingested tumor-derived material (Fig. 5C and fig. S5C). Using this gating strategy, we identified zsGreen+ subsets of six primary tumor APC populations: bulk CD24+ DCs, CD103+ DC1, CD11b+ DC2, bulk F4/80+ tumor-associated macrophages (TAM), CD11b+ TAM1, and CD11c+ TAM2. The proportion of zsGreen+ cells was significantly increased across all tumor APC subsets after administration of non–zsGreen-labeled necroptotic fibroblasts; this increase was particularly pronounced in the DC subsets examined (Fig. 5D). Accordingly, the absolute number of zsGreen+ cells among tumor DC subsets was increased after necroptotic cell exposure (fig. S5D). Because zsGreen expression was restricted to B16.F10-OVA tumor cells in this model, these results show that signals derived from necroptotic fibroblasts act in trans to increase either the rate of phagocytosis or the retention of tumor-associated antigen within tumor APC populations.

To further characterize the phenotype of tumor APCs in our model, we evaluated expression of the necrotic cell uptake marker CLEC9A on DCs and inhibitory PD-L1 on tumor cells and across different APC subsets. Exposure to necroptotic fibroblasts changed neither cell surface expression of either CLEC9A (Fig. 5E) or PD-L1 (Fig. 5F) nor the percentage of CLEC9A+ (fig. S5E) or PD-L1+ (fig. S5F) populations from each cell subset. We also assessed the activation status of zsGreen+ tumor APCs after exposure to dying fibroblasts within the TME. We first observed that zsGreen+ tumor APCs expressed higher levels of the costimulatory marker CD80 on a per cell basis after administration of necroptotic fibroblasts; this increase was consistent across all six tumor APC subsets examined (Fig. 5G). We also observed a significant increase in the geometric mean fluorescence intensity (gMFI) of CD80 after exposure to necroptotic cells when gating on zsGreen (non–tumor antigen-loaded) populations of each tumor APC subset (fig. S5G), revealing that stimuli derived from necroptotic fibroblasts increased CD80 expression across all tumor APCs, regardless of tumor antigen uptake status. This increase in activation marker expression correlated with an improved functional capacity of zsGreen+ tumor APCs after exposure to necroptotic cells in the TME, as zsGreen+ tumor APCs sorted ex vivo were capable of more robustly stimulating proliferation of previously activated transgenic OVA-specific (OT-I) CD8+ T cells in an in vitro coculture system (Fig. 5H). This stimulatory effect was limited to CD8+ T cells with T cell receptor (TCR) specificity for tumor antigen, because ex vivo coculture of zsGreen+ tumor APCs did not induce proliferation of T cells expressing an irrelevant lymphocytic choriomeningitis virus GP33 TCR (P14 transgenic TCR; fig. S5H). Therefore, exposure to necroptotic cells in the TME increases not only the abundance but also the immunostimulatory quality of tumor antigen-loaded APCs.

Consistent with this, we found that necroptotic cells appear to enhance antigen uptake by phagocytes in a tumor-independent setting, because coculturing bone marrow–derived macrophages (BMDMs) with necroptotic B16.F10 tumor cells in vitro resulted in increased uptake of an inert dextran-fluorophore substrate included in the coculture, compared with BMDMs cultured with live B16.F10 cells (Fig. 5I, left panel). Uptake of this bystander substrate was also associated with an increase in CD80 expression on both dextran+ and zsGreen+ BMDMs only after coculture with necroptotic B16.F10 cells (Fig. 5I, middle and right panels), whereas expression of the immunomodulatory markers CD206 and vascular cell adhesion molecule–1 (VCAM1) was decreased on zsGreen+ BMDMs cocultured with necroptotic tumor cells compared with live tumor cell controls (fig. S5I). Collectively, these data indicate that stimuli derived from necroptotic cells increase antigen loading by phagocytic cell subsets and that this effect may constitute a conserved response to necroptotic cell–derived stimuli rather than a tissue-specific effect restricted to the TME.

Engineered AAVs can be used to specifically induce necroptosis of tumor cells in vitro

Intratumoral dying cell injection provides a cleanly controlled model for examining how exposure to stimuli derived from necroptotic cells can influence antitumor immune responses. However, an obvious caveat of this model is that it fails to assess immune responses to tumor cell necroptosis in situ. To address this, we sought to create reagents that would allow direct induction of necroptosis in tumors in vivo. To achieve this, we generated versions of RIPK3 fused to a constitutively oligomerizing (“co”) domain, which consists of a high-affinity 2L6HC3-9 homotrimerizing domain that has been previously synthesized and described (35). These chimeric forms of RIPK3 undergo oligomerization and activation upon their expression in cells, independent of any upstream signaling or the presence of a ligand. To deliver these reagents to tumor cells, we created AAVs containing genes encoding these constructs under control of a synthetic MND (myeloproliferative sarcoma virus enhancer, negative control region deleted, dl587rev primer-binding site substituted) promoter, enabling robust gene expression in target cells. Upon transduction of a target cell by these engineered AAVs, the chimeric pro-death protein of interest is expressed, constitutively oligomerizes, and leads to rapid and specific induction of RIPK3-dependent cell death (Fig. 6A and fig. S6A).

Fig. 6 Engineered AAVs can be used to specifically induce necroptosis of tumor cells in vitro.

(A) Schematic of AAVs used to transduce tumor cells to express engineered pro-death enzymes fused to a constitutively oligomerizing (co) recruitment domain under the control of a synthetic MND promoter, leading to subsequent induction of a corresponding PCD modality. (B and C) Validation and kinetics of AAV2.5 serotype transduction efficiency in B16.F10 cells in vitro. (B) Percent of GFP+ cells transduced with AAV2.5-eGFP control. (C) Percent cell death in cells transduced with various death-inducing constructs. n = 3 technical replicates per group. (D) Heatmap depicting relative expression values of NF-κB–dependent gene targets, chemokines, and cytokines via NanoString analysis of B16.F10 tumor cells compared to eGFP-transduced controls 10 hours after AAV2.5 transduction (1 × 1011 IFU). Data are representative plots from two independent experiments (B and C) or means of three technical replicates from one experiment (D).

AAVs are a flexible tool for primary cell transduction, because several serotypes with varying cellular tropisms have been described. We therefore sought to identify an AAV serotype that would selectively deliver construct expression to B16.F10 melanoma cells. Using a hybrid AAV2.5 serotype (36), we observed robust transduction of cultured B16.F10 tumor cells within 24 hours of enhanced green fluorescent protein (eGFP)–AAV2.5 addition (Fig. 6B). The AAV2.5 serotype also transduced nonleukocytic CD45 cells within B16.F10-OVA tumors in vivo, exhibiting successful eGFP transduction in a higher percentage of CD45 cells compared with AAV5, AAV6, AAV8, or AAV9 serotypes (fig. S6B). eGFP-AAV2.5 also had the lowest percentage of off-target transduction of CD45+ tumor-associated leukocytes in vivo (fig. S6B). We therefore concluded that the hybrid AAV2.5 serotype would maximize tumor cell transduction efficiency while limiting off-target transduction of immune cells when adapted for use in vivo, potentially limiting off-target toxicity effects.

Next, we characterized the kinetics of death induced by AAV2.5 particles that deliver genes encoding chimeric pro-death proteins in vitro. Transduction of B16.F10 tumor cells with necroptosis-targeting AAV2.5 (coRIPK3) or lytic necrosis-targeting AAV2.5 (coRIPK3ΔC) led to 100% cell death within ~15 hours (Fig. 6C). Consistent with induction of necroptosis by these reagents, we found that the pan-caspase inhibitor zVAD-fmk did not affect death induction by coRIPK3 or coRIPK3ΔC (fig. S6C, left panel), whereas addition of the RIPK3 inhibitor GSK-843 eliminated coRIPK3ΔC-induced death while decreasing coRIPK3-induced death; this latter effect was likely due to reverse signaling through the RIPK1/RIPK3 necrosome to induce apoptosis, as previously described (fig. S6C, middle panel) (29, 30). Consistent with this, incubation with both zVAD-fmk and GSK-843 eliminated all cell death associated with coRIPK3 treatment (fig. S6C, right panel). This set of experiments shows that either necroptosis or lytic necrosis can be specifically and rapidly induced in B16.F10 tumor cells in vitro upon AAV-mediated delivery of coRIPK3 or coRIPK3ΔC, respectively.

Because our data using fibroblast injection pointed to activation of NF-κB responses by RIPK3, but not by RIPK3ΔC, as a key mediator of antitumor immune responses, we next assessed the ability of our AAV constructs to activate inflammatory transcription in dying cells. To do this, we infected B16.F10 tumor cells in vitro with AAVs encoding coRIPK3 or coRIPK3ΔC for 10 hours (a time point at which tumor cells have not yet undergone membrane permeabilization, allowing for nucleic acid isolation) and then harvested total RNA for NanoString analysis. Transduction of tumor cells with coRIPK3 yielded a distinct transcriptional signature compared with cells transduced with coRIPK3ΔC (fig. S6D). Further examination of this signature revealed that necroptotic B16.F10 cells exhibited up-regulated expression of numerous NF-κB–dependent gene targets, including Lta, Ltb, Cd40, Cd86, Mef2a, Nod2, and Nos2 in comparison with lytic necrotic tumor cells (Fig. 6D). In addition, necroptotic B16.F10 cells also up-regulated expression of several inflammatory chemokines and cytokines, including Cxcl1, Cxcl3, Ccl2, Ccl3, Ccl4, Ccl21a, Ccl22, Il12b, Il22, and Ifng (Fig. 6D). Up-regulated transcript levels for several of these target genes were independently validated via quantitative reverse transcriptase PCR (qRT-PCR) (fig. S6E). Together, these data indicate that the induction of tumor cell death via coRIPK3 transduction in vitro leads to an inflammatory transcriptional signature consistent with immunogenic necroptosis (20). Furthermore, this gene signature depends on the assembly of the RIPK1/RIPK3 necrosome via RHIM-RHIM interactions, because it is absent in tumor cells transduced with coRIPK3ΔC.

Administration of necroptosis-targeting AAVs in conjunction with α-PD-1 in vivo promotes durable tumor clearance

After validation of our PCD-targeting AAVs in vitro, we applied these tools to study antitumor responses in vivo. Intratumoral administration of coRIPK3 conferred control of B16.F10-OVA tumor outgrowth (Fig. 7A) and extension of animal survival (fig. S7A) in comparison with intratumoral injection of coRIPK3ΔC or control eGFP. Analysis of tumor homogenates revealed increased concentrations of numerous beneficial antitumor cytokines and chemokines after coRIPK3 administration, including IFN-γ, CCL3, CCL5, and CXCL10 (Fig. 7B), whereas levels of IL-6, CXCL1, and CXCL2 were unchanged (fig. S7B). Furthermore, we observed abscopal tumor control effects in a bilateral B16.F10-OVA flank tumor model, as coRIPK3 administration conferred control over tumor outgrowth in both treated (ipsilateral) and untreated (contralateral) tumors (Fig. 7C) and significantly extended animal survival (fig. S7C). These results recapitulate the tumor control effects that we observed in a bilateral tumor model using necroptotic fibroblast administration, showing that enforced RIPK3 activation via AAVs can similarly promote tumor control that is associated with increased intratumoral levels of inflammatory chemokines and cytokines.

Fig. 7 Administration of necroptosis-targeting AAVs in conjunction within vivo promotes durable tumor clearance.

α-PD-1 (A) B16.F10-OVA tumor growth curves after intratumoral administration of 1 × 1011 IFUs of death-inducing AAVs or eGFP control AAV. n = 8 to 14 mice per group. (B) Intratumoral concentrations of inflammatory cytokines and chemokines 48 hours after intratumoral AAV injection. The gray dashed line represents the limit of detection. n = 3 to 4 mice per group. (C) Tumor growth of ipsilateral (I, treated) and contralateral (C, untreated) B16.F10-OVA tumors after AAV administration. n = 10 to 12 mice per group. (D) Survival curves of B16.F10-OVA tumor–bearing mice after coadministration of AAVs with isotype control antibody. n = 14 to 15 mice per group. (E) Survival curves of B16.F10-OVA tumor–bearing mice after coadministration of AAVs with α-PD-1. n = 13 to 16 mice per group. (F) B16.F10-OVA tumor growth upon coadministration of necroptosis-inducing coRIPK3 AAV with α-CD8+ depletion antibody. n = 8 to 10 mice per group. (G) B16.F10-OVA tumor growth in Batf3−/− or wild-type control mice after necroptosis-inducing AAV administration. n = 10 to 13 mice per group. (H) Left: Schematic of tumor rechallenge experiments in mice from (E) successfully clear B16.F10-OVA tumors. (H) Right: Survival of mice rechallenged with B16.F10-OVA cells on the same flank as initial tumor location. n = 8 to 10 mice per group. (I) Kaplan-Meier plot for overall survival of skin cutaneous melanoma patients in TCGA data set. Data are parsed on upper and lower quartiles (25%) of RIPK3 mRNA expression. n = 114 patients per group. *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001. Black arrows indicate intratumoral AAV injections. Error bars represent SEM. Data are pooled from two to four independent experiments (A to H).

Next, we tested whether necroptosis-targeting AAVs could similarly protect mice from single B16.F10-OVA tumor outgrowth upon coadministration with α-PD-1. Not only did administration of coRIPK3 with isotype controls significantly extend animal survival (Fig. 7D) and inhibit tumor growth (fig. S7D) in comparison with eGFP-treated control mice, but also, the coadministration of coRIPK3 with α-PD-1 led to robust responses, with improved overall survival (Fig. 7E), complete tumor clearance in 69.2% of mice (fig. S7E, right panel), and significant control over tumor outgrowth (fig. S7E, left panel). Again, these tumor elimination responses closely paralleled those observed in the intratumoral necroptotic fibroblast injection model.

B16.F10-OVA tumor control after coadministration of coRIPK3 and isotype or α-PD-1 required the presence of CD8+ leukocytes, because depletion of CD8+ cell subsets via antibody injection completely abrogated the protective effects of coRIPK3 and immunoglobulin G2a (IgG2a) or α-PD-1 (Fig. 7F and fig. S7F). In addition, mice lacking BATF3+ cDC1 also failed to control B16.F10-OVA tumors after coRIPK3 and α-PD-1 treatment regimen (Fig. 7G and fig. S7G). Considering that tumor control by necroptotic fibroblasts also necessitated the presence of these immune cell compartments, these experiments revealed similar effector cell subset requirements between both intratumoral dying fibroblast and intratumoral AAV models. With these requirements in mind, we next sought to test whether the mice that had successfully cleared their B16.F10-OVA tumors after dual therapy (Fig. 7E) had developed protective immune memory. To this end, we rechallenged surviving animals with identical tumor cells on the same flank that initially bore the B16.F10-OVA tumors (Fig. 7H, left panel). Most of these animals were protected from mortality due to tumor outgrowth (Fig. 7H, right panel), whereas only 12.5% of mice regrew tumors (fig. S7H) compared with 100% of naïve controls. Overall, these data demonstrate that intratumoral administration of necroptosis-targeting AAVs in conjunction with α-PD-1 confers durable, immune-mediated tumor rejection similar to that observed upon administration of intratumoral necroptotic NIH 3T3 fibroblasts.

AAV-mediated transduction of tumor cells allows for enforced expression of activated RIPK3, regardless of the expression status of endogenous RIPK3. Considering the beneficial effects of enforced RIPK3 activation that we observed in our murine melanoma model, we asked how endogenous levels of RIPK3 correlated with survival outcomes in human cancer patients. Using tumor biopsy RNA sequencing data available through The Cancer Genome Atlas (TCGA) database, we stratified human skin cutaneous melanoma patients based on upper (high) and lower quartiles (low) of RIPK3 transcript expression within the tumor tissue. Patients with high tumor RIPK3 expression exhibited significantly improved survival outcomes compared with low RIPK3-expressing patients (Fig. 7I). Furthermore, multivariate Cox regression modeling revealed a negative coefficient (−0.175), indicating that high expression of RIPK3 is correlated with a better survival outcome (Fig. 7I) (37). Together, these results show that higher levels of RIPK3 expression within melanoma tumors are associated with improved survival in a subset of human patients.


Distinct forms of PCD can differentially instruct subsequent immune responses mounted against antigens derived from dying cells. Here, we describe a role for RIPK1/RIPK3 activation in which necroptotic fibroblasts within the TME drive increased antigen uptake and activation of tumor APCs to potentiate tumor-specific CD8+ T cell immunity, which synergizes with α-PD-1 coadministration to confer durable tumor rejection (Fig. 8). These gross tumor control effects are recapitulated in a model of AAV-mediated induction of necroptosis within melanoma tumor cells in situ, indicating that enforced activation of RIPK3 may lead to beneficial inflammatory signaling that is conserved across multiple cell types. Our data indicate that tumor control by necroptotic cells is primarily mediated via the activation of a RIPK1/RIPK3/NF-κB signaling axis independently of cell lysis, because MLKL deficiency in fibroblasts does not abrogate tumor control (Fig. 3F), whereas NF-κB inhibition (Fig. 3E) or use of a mutant form of RIPK3 that triggers cell lysis without engaging RIPK1-dependent transcription (Fig. 3C) eliminates the therapeutic efficacy of these cells. Thus, although we are engaging necroptotic cell death (as defined by activation of MLKL by RIPK3) in most experiments reported here, activation of transcription that parallels cell death is likely the cause of the beneficial immune stimulation we observe. These findings join a growing number of reports of biological effects of RIPK1/RIPK3 signaling that are independent of MLKL activation (21, 3841).

Fig. 8 Proposed model by which necroptotic cell death within the TME promotes antitumor immunity.

RIPK1/RIPK3 activation in necroptotic cells produces NF-κB–dependent signals that promote CD103+ cDC1− and CD8+ leukocyte–mediated antitumor immunity, which synergizes with α-PD-1 to promote durable tumor clearance.

Existing therapies to target RIPK1/RIPK3 activation in vivo exhibit variable efficacy due to off-target effects of global caspase inhibition (42) and the differential expression status of endogenous RIPK3 in tumor cells (23, 24). AAV-mediated reconstitution of constitutively active RIPK3 within tumor cells represents a previously unidentified strategy to specifically induce this pathway independently of any endogenous signaling requirements. Another group recently reported that intratumoral delivery of mRNA-encoding MLKL to promote cell lysis in situ conferred protection in murine melanoma and colon carcinoma models (43), and although our findings diverge from theirs with regard to a requirement for NF-kB signaling within dying cells, both studies support the idea that reconstituting expression of necroptotic signaling components can promote antitumor immunity. Considering that high levels of RIPK3 expression in human melanoma tumors correlate with improved patient survival (Fig. 7I), we propose that such strategies to restore or increase necroptotic signaling in human tumors represent a promising therapeutic target in future translational research.

Future work will need to define the specific signals derived from necroptotic cells that are responsible for mediating our observed antitumor immune responses. Across both dying cell and AAV administration models, the therapeutic effects of necroptotic cells appear to occur independently of MLKL activation, cell lysis, and subsequent DAMP release, suggesting that NF-κB transcriptional signaling downstream of the RIPK1/RIPK3 necrosome complex is required for therapeutic efficacy of necroptosis in the TME. Our observation of increased tumor-derived antigen within tumor APCs exposed to necroptotic cells is consistent with a previous report of necrotic debris being ingested alongside extracellular contents via macropinocytosis (44). However, the specific signals derived from necroptotic cells that are responsible for driving macropinocytosis to increase sampling of the local extracellular microenvironment remain unknown; our data imply that these signals are defined by cytokines, rather than DAMPs, produced by dying cells. Defining the mechanistic targets of RIPK1/RIPK3 activation and how these targets interact with tumor APCs to drive either increased macropinocytosis or improved retention of tumor antigen to better stimulate cytotoxic CD8+ T cells remains an important area for future study.

The overarching goal of tumor immunotherapy is to engage cytotoxic CD8+ T cells to kill tumor cells. This branch of the immune response evolved to combat intracellular pathogens such as viruses; treatments that activate innate immune pathways associated with viral sensing within tumors have therefore proven effective at provoking cytotoxic antitumor immunity. Such approaches include agonism of nucleic acid sensing via cyclic GMP-AMP synthase (cGAS)/stimulator of interferon genes (STING) (45, 46) and TLR pathways (47). The necroptotic pathway likely evolved to combat viral infection via elimination of the replicative niche and also promotes cross-priming of the CD8+ T cell responses required for viral elimination (20). Activation of this pathway within tumors can therefore be considered an additional strategy to direct antiviral immunity toward tumor elimination. Although much work remains to validate these findings in clinically relevant models, our data suggest that RIPK1/RIPK3 activation within the TME warrants further development as a component of tumor immunotherapy.


Study design

Pilot studies were used to estimate mean differences in tumor growth between treatment groups. Using the Sample Size Calculator resource (Boston University), we calculated biological replicate numbers needed to avoid experimental underpowering by using the percent difference in group means from pilot studies, with a power level of 0.80 and α level = 0.05. Age-matched mice were randomly assigned to treatment groups, and tumor measurements were conducted by a researcher blinded to treatment groups for at least one experimental replicate.

Cell culture

B16.F10-OVA, LL/2-OVA, NIH 3T3, and human embryonic kidney (HEK) 293T cells were maintained in Dulbecco’s modification of Eagle medium (DMEM) supplemented with 10% (v/v) fetal bovine serum (FBS), 2 mM l-glutamine, 10 mM Hepes, and 1 mM sodium pyruvate (complete DMEM). E.G7-OVA cells were maintained in RPMI 1640 supplemented with 10% FBS, 2 mM l-glutamine, 10 mM Hepes, 1 mM sodium pyruvate, 0.05 mM β-mercapthoethanol, geneticin (0.4 mg/ml; G418), and d-glucose (4.5 g/liter). B16.F10 and LL/2 cell lines were transduced with a plasmid (pSLIK) encoding activatable versions of caspase-9 or RIPK3 under thyroid response element control; thus, these cells were cultured in doxycycline (1 μg/ml; Sigma-Aldrich) for 18 hours to induce construct expression before harvesting as described below for dying cell injections. BMDMs were cultured in complete DMEM, penicillin/streptomycin, and recombinant macrophage colony-stimulating factor (20 ng/ml), and differentiated for 7 days before plating for experiments. All cells were cultured at 37°C with 5% CO2.


C57BL6/J (B6/J) mice were purchased (the Jackson Laboratory) and allowed to acclimate up to 1 week before experiment initiation. All other genotypes were bred and housed under specific pathogen–free conditions at the University of Washington. All animals were maintained according to protocols approved by the University of Washington Institutional Animal Care and Use Committee.

Tumor models

Six- to 10-week-old female (B16.F10-OVA, E.G7-OVA) or male (LL/2-OVA) mice were injected subcutaneously on the right flank with 1 × 105 (B16.F10-OVA, E.G7-OVA) or 2 × 105 (LL/2-OVA) tumor cells, mixed in a 1:1 volumetric ratio with the basement membrane matrix Matrigel High Concentration (Corning) for a final injection volume of 100 μl. For bilateral tumor experiments, mice were equivalently implanted with tumor cells on the left flank on the same day of right flank tumor injection. As previously described (45), tumor volume was calculated using the following formula: volume = short axis2 × long axis × 0.523. Mice were euthanized once tumor burden reached a volume of ≥2000 mm3. Mice that developed skin ulceration over the tumor site were excluded from experimental analyses. Complete tumor clearance was determined by the absence of a palpable tumor mass at the site of tumor injection.

Intratumoral dying cell injections

NIH 3T3, B16.F10, or LL/2 cells stably transduced with pro-death constructs were harvested and activated as previously described (20). Briefly, 5 × 106 cells/ml were incubated in complete DMEM and 1 mM B/B homodimerizer (Clontech) for 15 min at 37°C. Cells were then washed with cold phosphate-buffered saline (PBS), resuspended at 20 × 106 cells/ml, and kept on ice before injection. Dying cells (1 × 106) were administered intratumorally in 50 μl. Remaining cells were replated and cultured at 37°C overnight to ensure that >95% of treated cells underwent PCD. Dying cells were administered on days 6, 8, and 10 after initial tumor challenge. For experiments involving IκBα inhibition, NIH 3T3 cells were pretreated with 10 μM BAY-117085 (Cayman Chemical) for 45 min before harvesting for B/B homodimerizer incubation, as described (20).

In vivo antibody administration

A total of 200 μg of α-CD8 (clone 2.43, Bio X Cell), α-CD4 (clone GK1.5, Bio X Cell), α-PD-1 (clone RMP1-14, Bio X Cell), or respective isotype controls were administered to mice via intraperitoneal injection on days 5, 7, 9, and 11 after initial tumor challenge. NK cell depletion experiments followed the same dosing protocol, using 250 μg of α-NK1.1 (clone PK136, Bio X Cell). CLEC9A blocking experiments followed the same dosing protocol, using 400 μg of α-CLEC9A (clone 7H11, Bio X Cell) or isotype control, as described (27).

Recombinant AAV cloning

Design and sequencing analysis of all plasmids was performed using Geneious software v.7.1 (48). The 2L6HC3-13 trimer homo-oligomer domain was a gift from D. Baker (35). Trimerizing RIPK3 constructs were directly cloned into a single-stranded AAV vector using multifragment assembly (In-fusion HD, Takara Biosciences). AAV backbone was linearized using SnaBI digest as previously described (49). Primers for amplification of gene fragments were designed to contain 20–base pair 5′ and 3′ homology to neighboring fusion sequences, and PCR amplification was carried out using Q5 polymerase (New England Biosciences). The shortened 3′ untranslated region woodchuck hepatitis virus posttranscriptional regulatory element (UTR WPRE) and polyA elements were amplified from pAAV-CW3SL-EGFP, a gift from B.-K. Kaang (Addgene no. 61463). Sense (S) and antisense (AS) primer sequences were as follows: fragment 1 (MND promoter), (S) CCGCCATGCTACTTATCTACGGAGTCGTGACCTAGGGAACAGAGAAACAGG and (AS) TTCGAGGAAGTCAAAACAGCGTGG; fragment 2 (RIPK3 and RIPK3ΔC), (S) CGCTGTTTTGACTTCCTCGAACCATGTCTTCTGTCAAGTTATGG, (full-length RIPK3 AS) AGAACCACTCCCTTCTGATCCTTCGGAACCCGTACGCTTGTGGAAGGGCTGCCAGC, and (RIPK3ΔC AS) AGAACCACTCCCTTCTGATCCTTCGGAACCCGTACGTCATTGGATTCGGTGGGGTC; fragment 3 (2L6HC3-13 homotrimer domain), (S) GATCAGAAGGGAGTGGTTCTCATATGGGTACGAAATACG and (AS) CAGAGGTTGATTATGCGGCCTTAGTCACTTTTGGCGTTAATTTTC; and fragment 4 (sWPRE/polyA), (S) GGCCGCATAATCAACCTCTGG and (AS) CCGCCATGCTACTTATCTACAAAAAACCTCCCACATCTCCCCC.

The MND-eGFP self-complementary AAV was a gift from D. Rawlings. The DNA sequence of inserted elements was verified by sequencing, and the integrity of the viral inverted terminal repeat (ITR) within the pAAV backbone was confirmed by restriction digest using AhdI, BglI, or SmaI, before viral production.

AAV production, purification, and quantification

AAVs were produced as described (50, 51). Briefly, AAV stocks were generated in HEK293T cells via polyethylenimine transfection using a vector and a serotype helper (pLTAAV). Cells were harvested 48 hours after transfection, lysed via freeze-thaw cycling, treated with Universal Nuclease (100 U/ml; Thermo Fisher) at 37°C for 30 min, and purified via centrifugation over an iodixanol density-step gradient. Titers of viral stocks were determined via qRT-PCR analysis in conjunction with TaqMan reagents and a ViiA 7 Real-Time PCR apparatus (Applied Biosystems). qRT-PCR for viral titer used primers targeting the conserved ITR, using the following sequences: (forward) GGAACCCCTAGTGATGGAGTT and (reverse) CGGCCTCAGTGAGCGA.

Intratumoral AAV injections

Infectious units (IFU; 1 × 1011) of respective AAVs were administered intratumorally in 50 μl. Virus aliquots used for in vivo experiments were thawed once after initial freezing after purification. AAV injections were administered on days 6, 8, and 10 after initial tumor challenge.

Flow cytometry and cell sorting

Leukocytes were isolated from either the tumor-adjacent inguinal lymph node or spleen by mashing over a 70 μM strainer or from tumor tissue by digesting minced tumors in 1× PBS, collagenase A (2.6 mg/ml; Sigma), and deoxyribonuclease I (23 U/ml; Sigma) at 37°C with agitation for 45 min before mashing tissue over a 70 μM strainer. Cells (1 × 106 to 3 × 106) were blocked with anti-CD16/32 (BD Biosciences) and stained with Zombie viability dye (BioLegend) at room temperature for 30 min. Cells were then incubated with appropriate fluorochrome-conjugated antibodies in 1× PBS, 0.5% FBS, and 2 mM EDTA at 4°C for 1 hour. Permeabilization and intranuclear staining were performed using a Foxp3 Intranuclear Transcription Factor Staining Kit (eBioscience). Data were collected using an LSRII flow cytometer (BD Biosciences) and analyzed using FlowJo software (TreeStar). For sorting of zsGreen+ tumor APC populations, B16.F10-OVA–zsGreen tumors were harvested 48 hours after intratumoral dying cell injection, leukocytes were processed and stained as described above, and subsets were sorted using a FACSAria II (BD Biosciences).

OT-I/P14 proliferation assay

Lymph nodes and spleens from OT-I or P14 TCR transgenic mice were processed and enriched for CD8+ T cells via negative selection using biotinylated antibodies against B220, CD4, CD11b, CD11c, and Ter119 (eBioscience) followed by magnetic separation. Purified transgenic T cells were activated via 6-day coculture with irradiated splenocytes pulsed with SL8 (100 ng/ml; for OT-I) or GP33 (for P14) peptides (InvivoGen). A total of 20,000 previously activated transgenic T cells were labeled with 5 μM CellTrace Violet (Thermo Fisher) and plated with 4000 sorted zsGreen+ tumor APC subsets in 96-well U-bottom plates for 72 hours before analysis of proliferation dye dilution via flow cytometry.

Murine cytokine assessment

To evaluate serum cytokine levels, sera were harvested from mice receiving indicated intratumoral treatments 48 hours after dying cell administration and stored for <2 weeks at −80°C. As a positive control for systemic inflammation, B6/J mice were injected intraperitoneally with the STING agonist DMXAA (40 mg/kg; ApexBio), and sera were harvested 5 hours after injection and then frozen at −80°C. To evaluate intratumoral cytokine levels, tumors were harvested 48 hours after dying cell administration or 72 hours after AAV administration. Tumors were then minced and homogenized using metal beads with vigorous shaking in tubes and then frozen at −80°C. To evaluate cytokine levels in vitro, 3 × 105 B16.F10 cells were infected with 1 × 1011 IFU of AAV for 18 hours, and supernatants were frozen at −80°C. Thawed samples were analyzed using a T helper cell 1 (TH1)/TH2 ProcartaPlex Panel 1 Luminex kit (Thermo Fisher).

NanoString RNA analysis and qRT-PCR

B16.F10 cells (2 × 106) were infected with 1 × 1011 IFU of respective AAVs for 10 hours. Total RNA was isolated using a NucleoSpin RNA Kit (Macherey-Nagel) and run on an nCounter Sprint in conjunction with an nCounter Mouse Inflammation V2 Panel (NanoString). Data were normalized and analyzed using nSolver software (NanoString). For target gene validation, oligo(dT) random hexamers and SuperScript III RT (Life Technologies) were used to synthesize cDNA from the same total RNA samples used for NanoString analysis. Fluorogenic qRT-PCR analysis was performed using previously published oligonucleotide primer sequences using SYBR green reagents and a ViiA 7 Real-Time PCR apparatus (Applied Biosystems). Cycle threshold (CT) values for target genes were normalized to CT values of the housekeeping gene Gapdh (ΔCT = CTTarget – CTGapdh) and subsequently normalized to baseline control values (ΔΔCT = ΔCTExperimental – ΔCTControl).

In vitro cell death assay

B16.F10-OVA cells (1 × 105) were infected with 1 × 1011 IFU of respective AAVs in 24-well plates for 24 hours. Cell viability was evaluated via incorporation of cell viability dye Sytox Green (Molecular Probes) or Yoyo-3 (200 nM; Life Technologies) and quantified using a two-color IncuCyte Zoom bioimaging platform (Essen Biosciences), as described (52). Where indicated, 50 μM zVAD-fmk (SM Biochemicals) or 100 nM GSK-873 (GlaxoSmithKline) were added to inhibit pan-caspase activation or RIPK3 activation, respectively.

CRISPR-Cas9 gene targeting

The following guide RNA (gRNA) sequences were cloned into a pRRL-Cas9-T2A-puromycin CRISPR-Cas9 lentiviral vector [a gift from D. Stetson (53)]: 5′-GCGAGGTATTCGGCTCCGCG-3′ (murine nontargeting gRNA) (54) and 5′-GCACACGGTTTCCTAGACGC-3′ (murine Mlkl gRNA). Constructs were transduced into NIH 3T3 and acRIPK3 cells using standard lentiviral transduction protocols and selected in puromycin (1 μg/ml).

siRNA knockdown

MLKL−/− NIH 3T3 cells (2 × 105) +acRIPK3 were transfected with siGenome SMARTpool small interfering RNA (siRNAs) (Dharmacon) targeting murine RIPK1 (M-040150-01), murine caspase-8 (M-043044-01), murine cFLIP (M-041091-01), or nontargeted “scramble” pool (D-001206-14), using lipofectamine siRNA Max (Life Technologies). Forty-eight hours after transfection, cells were replated and treated with 100 nM B/B homodimerizer (Clontech), and cell death kinetics were characterized as described above.

Western blot

Cell lysates were harvested and quantitated using a BCA Protein Assay (Thermo Fisher). Thirty milligrams of total protein per sample were separated using SDS–polyacrylamide gel electrophoresis gels (Invitrogen) and detected using traditional protocols. The following antibodies were used for protein detection: rat α-MLKL clone 3H1 (EMD Millipore), rabbit α-FKBP12 (Thermo Fisher), mouse α-actin C4 (EMD Millipore), goat α-rat IgG–horseradish peroxidase (HRP) (Santa Cruz Biotechnology), donkey α-rabbit IgG-HRP (Santa Cruz Biotechnology), and goat α-mouse IgG-HRP (Santa Cruz Biotechnology).

In vitro dextran uptake assay

BMDMs were plated in a 1:5 ratio with either live or acRIPK3-expressing B16.F10-zsGreen tumor cells. acRIPK3 cells were induced to die upon 18 hours incubation with doxycycline (1 μg/ml) before coculture with BMDMs and then incubated with 100 nM B/B homodimerizer (Clontech) for 24 hours before adding dextran-phycoerythrin (PE)–TexasRed (1 mg/ml; 10,000 molecular weight, Thermo Scientific). Dextran incubations were performed in triplicate at either 4° or 37°C for 30 min, and plates were tapped every 10 min to mix. Cells were washed three times, stained with fluorochrome-conjugated antibodies, and immediately analyzed on a flow cytometer. Dextran uptake was calculated as follows (31): ΔgMFI = (gMFI dextran binding at 37°C − gMFI dextran binding at 4°C).

TCGA analysis

The OncoLnc package (37) was used to analyze RNASeqV2 and overall survival data generated by TCGA Research Network database (55). OncoLnc was used to conduct survival analyses using multivariate Cox regression modeling, assign log-rank P values and Cox coefficients to assess significance and generate Kaplan-Meier survival curves.


Unless otherwise noted in figure legend, data represent means ± SEM. Survival curves were analyzed via Mantel-Cox log-rank test unless noted otherwise. All other experiments were compared using parametric two-tailed student’s t test, chi-square test, or one-way or two-way analysis of variance, with appropriate corrections for repeated measures of tumor growth curves. All statistical analyses were performed using GraphPad Prism software unless noted otherwise.


Fig. S1. Necroptotic cells extend the survival of tumor-bearing mice.

Fig. S2. BATF3+ and CD8+ leukocyte requirements for tumor control by necroptotic cells.

Fig. S3. DAMP-independent tumor control by necroptotic cells is recapitulated in multiple syngeneic tumor models.

Fig. S4. Gating strategies and quantification of T cell subsets in tumor and tdLN after dying cell administration.

Fig. S5. Gating strategies and effects of dying cell administration on tumor APC subsets.

Fig. S6. Characterization of transduction efficiency and cell death induction by engineered AAVs.

Fig. S7. Tumor growth restriction and survival extension after administration of AAVs targeting tumor cell necroptosis in situ.


Acknowledgments: We thank P. Ralli-Jain for technical assistance. We thank D. Stetson (University of Washington) for Tmem173−/−, Mb21d−/−, and Mavs−/− mice; M. Gale Jr. (University of Washington) for Irf3−/−, Myd88−/−, Ticam1−/−, and P14 TCR transgenic mice; and M. Gerner (University of Washington) for OT-I TCR transgenic mice. We thank N. Subramanian (Institute for Systems Biology) for the pSLIK plasmid and M. Krummel (UCSF) for both the pSIREN-zsGreen plasmid and technical input. We also thank D. Rawlings (Seattle Children’s Research Institute) for AAV-related constructs. Funding: This work was supported by NIH grant R01CA228098 and by a Wade F.B. Thompson CLIP award from the Cancer Research Institute (both to A.O.). A.G.S. was supported by PHS NRSA T32GM007270 and NSF GRFP (DGE-1256082), N.W.H. by T32AI106677-05, M.N.M. by T32CA080416, and B.P.D. by F32 AI129254. Author contributions: A.G.S. and A.O. conceived the study and designed experiments. N.W.H. designed AAV plasmid constructs. D.B. provided oligomerization domain constructs used in AAV experiments. A.G.S., N.W.H., M.N.M., B.P.D., S.B.K., C.E.H., S.L.O., and K.C. performed experiments. A.G.S. and B.P.D. analyzed data. A.G.S. and A.O. wrote the manuscript with editorial input from all authors. Competing interests: The testing of necroptosis-inducing AAVs was supported in part by funds received from FivePrime Therapeutics as part of a sponsored research agreement; A.O. has acted as a paid consultant for FivePrime Therapeutics. A.G.S., N.W.H., and A.O. are inventors on a pending patent held by the University of Washington for the constitutively oligomerizing cell death enzymes described in this manuscript. All other authors declare that they have no competing interests. Data and materials availability: All data needed to evaluate the conclusions in the paper are available in the paper or the Supplementary Materials. Reagents described in this manuscript are available from A.O. for research use under a materials transfer agreement from the University of Washington.

Stay Connected to Science Immunology

Navigate This Article