In the past few years, a major understanding of the regulation of cancer immunity has allowed to develop new immunotherapy approaches that have shown unprecedented efficacy against aggressive bad prognosis cancers.1 2 Natural killer (NK) cells and cytotoxic CD8⁺ T (Tc) cells are the key responsible for the elimination of transformed cells during both cancer immunosurveillance and immunotherapy.3–6 However, meanwhile NK cells eliminate stressed tumor cells and/or cells that have downregulated major histocompatibilty complex (MHC)-I molecules,7 activation of Tc cells strictly depends on the recognition of antigens presented by MHC-I. Thus, the presence of tumor-mutated genes raising new epitopes/antigens (neoantigens) together with proper inflammatory signals leading to efficient antigen cross-presentation by dendritic cells (DCs) are prerequisites for the generation of an effective Tc cell-mediated anticancer response.8 Accordingly, response to immunotherapy treatments that rely on T cell immunity, such as checkpoint antibody therapy, is limited to cancers that express immunodominant mutations that raise an optimal T cell response.

As an alternative to overcome this limitation, in recent years, it has been described that specific chemotherapy drugs and irradiation, in addition to kill cancer cells, have the capacity to prime anticancer immune responses against antigens released from dying cells, a concept known as immunogenic cell death (ICD).9–11 This process enhances the elimination of cancer cells and generates immune memory against the tumor antigens, reducing the chance of cancer recurrence.10 12 This ability is related to the activation of danger signaling pathways evoking emission of damage-associated molecular patterns (DAMPs) leading to inflammation, which favor DC maturation and antigen presentation.10 13–15 ICD increases the immunogenicity of tumor antigens that per se might present a lower antigenic potential and would not induce an efficient T cell response. In recent years, some molecular determinants that dictate the immunogenic characteristics of dying cells have been clarified.16 Apoptosis, ferroptosis, pyroptosis or necroptosis are among the different modalities of programmed cell death that can be immunogenic.17–20

Several molecules involved in the generation of eat-me and danger signals leading to ICD in cancer cells have been found such as exposure of calreticulin on the cell membrane, release of Hsp70 or HMGB1, autophagy and production of interleukin (IL)-1β or type I interferon (IFN).16 At present, the relative importance of each of these signals is not clear and different ICD signals have been described in cells dying under different stimuli. Thus, to find out if cancer cells eliminated by a specific stimulus undergo ICD, the generation of anticancer immunity and tumor development should be analyzed in vivo in animals previously immunized with dead cancer cells.

Despite the extensive studies focused on the mechanisms involved in tumor cell recognition by T cells, the molecular determinants that regulate cell death induced by Tc cells on cancer cells, a key event for successful cancer immunotherapy, are less characterized. Paradoxically, little is known about the immunogenic characteristics of cancer cell death induced by immunotherapy. Specifically, whether cell death induced by Tc cells is immunogenic and, if so, which are the mechanisms responsible for ICD induced by Tc cells?

Tc cells mainly use two pathways to kill cancer cells, death ligands (ie, tumor necrosis factor (TNF)-α, fas ligand and TNF-related apoptosis inducing ligand (TRAIL)) or granule exocytosis.21 22 The later consists of the release of a pore-forming protein perforin (perf) that delivers a family of serine-proteases, the granzymes (gzms) into the cytosol of target cells. Gzms are the ultimate responsible (mainly gzmB) for cancer cell elimination. Apart from regulating target cell death, it was shown that gzms are involved in antigen cross-presentation of dying cells by somehow regulating DC phagocytosis.23 In a recent work, we have shown that Tc cells efficiently eliminate in vivo tumor cells expressing pro-tumorigenic and/or anti-apoptotic mutations.24 These results provided the molecular basis to explain the efficacy of immunotherapy against multidrug-resistant bad prognosis cancer. However, these findings raise a new question to predict refractoriness and/or cancer relapse after immunotherapy: is Tc cell mediated-elimination of cancer cells immunogenic? And if so, what is the impact of mutations in cell death/prosurvival pathways on the immunogenicity of cancer cell death induced by Tc cells?

Here we have employed the EL4 lymphoma mouse model and different cancer vaccination strategies based on the gp33 and OVA tumor Ag models to explore the immunogenic characteristics of cell death induced by Tc cells on cancer cells and the mechanisms involved in this process both in vitro and in vivo. Our results show that T cell cytotoxic activity on tumor cells induces ICD and promotes a protective immune response in vivo, priming the generation of new CD8⁺ Tc cell responses against endogenous tumor antigens. Importantly, this response is able to reduce tumor development in mice challenged with a second tumor. Additionally, our results show that expression of active caspase-3 is key for ICD induced by CD8⁺Tc cells.

To demonstrate whether Ag-specific CD8⁺Tc cell killing in vivo was immunogenic, we employed a clinical relevant immunotherapy protocol consisting of vaccination with DCs loaded with gp33 Ag (DC^gp33) (figure 4A). This protocol has shown good efficacy in vivo n mice using other antigen models.33 A summary of the protocol used is depicted in figure 4A. Immunization with mature DC^gp33 conferred a large protection against EL4.gp33 tumor development compared with mice non-immunized or those immunized with DCs matured but without gp33 (figure 4B). Protection against EL4.gp33 tumor development was completely abrogated when CD8⁺ T cells were depleted or in perf^KO mice (figure 4C). These results indicate that immunization with mature DC^gp33 generates a CD8⁺ T cell-dependent and perf-dependent protection against EL4 tumor development, showing that this model is suitable to analyze immunogenicity of cell death induced by CD8⁺ Tc cells in vivo.

Next, we analyzed if the elimination of EL4.gp33 cells in vivo in mice immunized with DC^gp33 (from now on EL4.gp33^TcDC cells) generated a protective response against parental EL4 tumors. Mice inoculated with mature DC^gp33 developed EL4 tumors and reach the endpoint on day 15 (DC^gp33, figure 5A), similar to control non-immunized mice (figure 3A). In contrast, EL4 tumor development was largely reduced and survival increased in mice immunized following the protocol EL4.gp33^TcDC (figure 5A). This tumor protection was lost on CD8⁺ T cell depletion, showing that elimination of parental EL4 cells was dependent on the generation of new CD8⁺ Tc cells specific against EL4 endogenous antigens. A similar result was found, when the same protocol was performed, but using the OVA antigen instead of gp33. In this case, elimination of EL4^OVA cells in vivo in mice immunized with DC^OVA generated a protective response that significantly delayed the growth of parental EL4 tumors (online supplementary figure 2), indicating that immunization against endogenous EL4 tumors was independent of the antigen model.

To confirm that CD8⁺ Tc cells cross-primed against dying EL4.gp33-derived antigens were the responsible of EL4 tumor elimination, we employed an adoptive transfer protocol using CD8⁺T cells from wt and BATF3^KO mice, previously immunized following the protocol EL4.gp33^TcDC (figure 5B). Only those mice transferred with CD8⁺ Tc cells from EL4.gp33^TcDC immunized wt mice showed protection against EL4 tumor development compared with mice transferred with Tc from wt mice immunized with DC^gp33, or Tc coming from BATF3-deficient mice independently of the immunization stimulus (figure 5B). These results show that in vivo elimination of EL4.gp33 cells by gp33 Ag-specific CD8⁺Tc cells is immunogenic and cross-primes a CD8⁺ Tc cell response against endogenous EL4-derived antigens, suggesting that is a general mechanism of epitope spreading not only applicable to adoptive T cell transfer but also to DC vaccination strategies for cancer immunotherapy.

Finally, we wondered how the mutations on cell death pathways analyzed in figure 1 in vitro would affect the immune response against tumor Ags derived from cancer cells eliminated by Tc cells in vivo. To this aim, we analyzed the role of caspases and the mitochondrial cell death pathway in the generation of immunity against EL4 tumor development after vaccination. The level of protection offered by EL4.gp33 cells killed in vivo after DC^gp33 vaccination was much pronounced as compared with the protocol employing EL4.gp33 killed in vitro (figures 5 and 2). However, in order to reduce the number or mice and to study the mechanism in a more controlled way, we followed the vaccination protocol with EL4.gp33 killed in vitro (EL4.gp33^TcLCMV cells), using EL4.gp33 wt cells as well as mutants thereof overexpressing Bcl-X_L or expressing a dominant negative caspase-3 mutant (DNC3). In addition, we used EL4.gp33 wt cells in which caspases were blocked with the pan-inhibitor Q-VD-OPh.

Mice were immunized employing the cells killed in figure 1A, following the protocol indicated previously (figure 2). As shown in figure 6A, mice immunized with EL4.gp33^TcLCMV cells were protected against tumor development and survived significantly longer than non-immunized control mice. In contrast, protection disappeared in those mice immunized with EL4.gp33^TcLCMV cells that had been killed in the presence of Q-VD, indicating that functional caspases are required to generate protection against parental EL4 tumor development. A similar result was found when EL4DNC3.gp33^TcLCMV cells were used. The protection observed after immunization with EL4.gp33^TcLCMV was lost when cells expressed the caspase-3 mutant (EL4DNC3.gp33^TcLCMV, figure 6B). In contrast, mice immunized with EL4Bcl-X_L.gp33^TcLCMV cells were protected and developed EL4 tumors significantly smaller and survived significantly longer than non-immunized control mice (figure 6B). Although tumors were slightly bigger than mice immunized with EL4.gp33^TcLCMV, neither tumor growth nor mouse survival was significantly different. Therefore, this result suggests that the intrinsic apoptotic pathway is dispensable for CD8⁺ Tc-induced ICD in EL4 tumor model, which critically depends on the presence of active caspase-3.

To confirm the role of caspase-3 in ICD induced by CD8⁺Tc cells, we analyzed the generation of EL4-specific CD8⁺ Tc cells after immunization with EL4.gp33 dead cells. Splenocytes from mice immunized with EL4.gp33^TcLCMV showed increased killing of parental EL4 cells in vitro compared with splenocytes from control mice or those immunized with EL4.gp33^TcLCMV killed in the presence of Q-VD or with EL4DNC3.gp33^TcLCMV cells (figure 6C). Cytotoxic activity of splenocytes from mice immunized with EL4Bcl-X_L.gp33^TcLCMV cells was similar to that from EL4.gp33^TcLCMV immunized mice. These results confirm that caspase-3, but not the intrinsic apoptotic pathway, is required for ICD induced by CD8⁺ Tc cells, including the generation of a protective anti-tumor CD8⁺ Tc cell response.

A major understanding of the regulation of CD8⁺ Tc cell-mediated immunity in cancer has been key to successfully treat mutated bad prognosis cancers with immune checkpoint inhibitors or with modified CAR-T cells. However, the number of patients benefiting from immunotherapy is still relatively low and restricted to a small proportion of some types of cancer. The factors contributing to immunotherapy efficacy and tumor resistance and/or relapse are poorly explored. We have recently shown that antigen-specific Tc cells are able to eliminate tumor cells expressing anti-apoptotic mutations conferring bad prognosis and drug resistance, preventing tumor development in vivo.24 However, it is unclear if these mutations may affect tumor relapse. Here, employing two different models of cancer vaccination in mice and the EL4 lymphoma model, we show that elimination of cancer cells by antigen-specific CD8⁺ Tc cells ex vivo and in vivo is immunogenic and generates a secondary protective immune response against endogenous tumor antigens. Importantly, specific mutation of the cell death executioner caspase-3, although did not affect cell killing and elimination of primary tumors completely abrogated the generation of protective CD8⁺ Tc cell response against secondary tumor development. This finding indicates that ICD induced by CD8⁺ Tc cells and the generation of spread immunity against endogenous tumor antigens rely on caspase-3-dependent apoptosis of EL4 cancer cells.

Although we have mainly used a model of tumor antigen derived from the LCM virus, gp33 antigen, this should not be a major limitation to extrapolate our findings to other immunotherapies, like those employing viral infections (oncolytic viruses) and others.34 First, ICD induced by Tc cells has been confirmed in vivo using a conventional approach consisting of vaccination with peptide (gp33) pulsed bone marrow–derived DCs, which has been previously shown to offer CTL priming and tumor immunity.33 In addition, the affinity of the antigen T cell receptor for viral gp33 is similar to that one for ‘real’ tumor antigens like MelA or gp100.35 36 This affinity is even higher for engineered T cell receptors as in CAR-T cells. Anyway, we have confirmed that our results are not restricted to gp33-antigen since elimination of EL4.OVA cells in mice vaccinated with OVA-pulsed bone marrow–derived DCs significantly delayed the development of parental EL4 tumors. Thus, the use of two different surrogate antigens supports the conclusion that our results are not influenced by the selection of specific surrogate tumor antigens. Moreover, a similar result in other tumor models employing different tumor antigens has been shown by Ignacio Melero’s group in an accompanying paper37 reinforcing that ICD induced by Tc cells is not restricted by specific antigens.

The concept of immunogenic cell death raised from the observation that cancer cells eliminated by specific chemotherapy drugs or radiotherapy induced the generation of protective CD8⁺ Tc cells against antigens released by dying tumor cells.9 10 14 15 20 This response contributed to the elimination of the tumor cells and, in addition, it was shown to prevent cancer recurrence.9 28 Since then, several stimuli have been found to induce ICD on cancer cells including irradiation, hyperthermia, infection or cell starvation.16 However, and paradoxically, studies addressing if cell death induced by CD8⁺ Tc cells is immunogenic are scarce.

Protection correlates with the generation of well-known ICD signals in dying cancer cells: calreticulin membrane exposure and the release of HMGB1 and the proinflammatory cytokine IL-1β.16 However, several signals have been associated to ICD induced by different stimuli. Thus, analyzing the effect of in vivo immunization with dead cells on tumor development is mandatory to analyze if cell death induced by a specific stimulus is immunogenic. Here an important question to genuinely identify ICD on cancer cells is the potential direct in vivo immunomodulatory effects of the stimulus used to kill the cancer cells used for the immunization. For example, different drugs, commonly referred to as ICD inductors, are capable to induce tumor cell death as well as to directly modulate the immune system in vivo making it impossible to attribute the observed biological effects solely to ICD, unless they are removed before immunization. Activated Tc cells also produce several immunomodulatory cytokines that might contribute to the effects observed in vivo in our study. Having in mind this possibility, our study included different controls suggesting that a direct effect of Tc cells is not the responsible of the protection observed during vaccination with dead EL4 tumor cells. First, inoculation of mice with activated Tc cells alone has no effect on tumor development. On top of that, immunization with the mixture of activated Tc cells and dead tumor cells, killed in the absence of caspase-3 activity, did not confer any protection against tumor development. An optimal approximation would be to separate the cell debris from Tc cells after the killing assay. However, this approximation is challenging and would likely change the immunogeneic properties of cell debris like the presence of soluble factors released by dying cells. Thus, being aware of this difficult limitation to overcome the in vivo vaccination experiments strongly support for an immunogeneic phenotype of cell death induced by CD8⁺ Tc cells.

We find herein that general caspase inhibition as well as caspase-3 inhibition impairs immunogenicity and spread immunity against endogenous tumor antigens. However, the role of specific caspases seems to be dependent on the stimuli used29 38 39 and likely the tumor cell type. Our results show that, at least in EL4 lymphoma cells, caspase-3 plays a key role in ICD induced by CD8⁺ Tc cells since vaccination with dying EL4 cells expressing the DN caspase-3 mutant does not stimulate the generation of CD8⁺ Tc cells against endogenous EL4 antigens and does not protect against secondary tumor development. In contrast to caspase-3, the mitochondrial apoptotic pathway does not seem to be involved in ICD induced by CD8⁺ Tc cells since Bcl-X_L overexpression had no effect.

Further confirming that Tc cells induce ICD in tumor cells, we show that the protective effect is loss in mice deficient in TLR4 or in BATF3-dependent DCs. TLR4 is a receptor for HMGB1, a DAMP that interacts with TLR4 in DCs for an efficient processing and cross-presentation of antigens derived from cells killed by chemotherapy or radiotherapy.28 40 Regarding BATF3-dependent cDC1s, they are involved in phagocytosis and cross-presentation of antigens from dying cells, contributing to protection against some but not all viral, bacterial and fungal infections.31 41 A previous study found out that tumor cells killed by OT1-specific Tc cells induced antigen cross-presentation by DCs.23 However, that study did not analyze whether this killing was immunogenic in the tumor context and whether it conferred a protective immune response against tumor development. Our data show that cancer cells killed by Tc cells bear an ICD that promotes cross-presentation by cDC1s, which are specifically required for antitumor immunity following ICD induction by Tc cells. These results concur with the essential role of cDC1s in tumor vaccination42 and its association with improved overall survival in several tumors.43 Indeed, BATF3 is required for an efficient generation of Tc cell immunity against primary immunogenic tumors.31 In addition, BATF3-dependent DCs modulate the efficacy of different immunotherapy protocols including adoptive T cell therapy or mAb against immune checkpoints.32 43 44 Our data show the requirement of cDC1 for generation of an efficient protective immunity against endogenous antigens expressed in dead tumor cells following Tc-induced ICD. Remarkably, cDC1s are not essential for antitumor immunity following other models of ICD.45

As indicated above, our results have been independently confirmed using other cancer immunotherapies and different tumor models, including transgenic T cell receptors and NK cells.37 Recent findings in clinical trials employing T cell-based cancer immunotherapy support the novelty, the conceptual advance and the potential clinical relevance of our results. It has been recently found that during CAR-T cell therapy in gastric cancer, new T cell clones against tumor neoantigens are detected in patients, a concept known as epitope spreading.46 Epitope spreading was also observed during cancer vaccination in humans,47 although the molecular basis for this phenomena remained unexplored. Our results provide an explanation and the molecular mechanism involved in that observation, enhancing our mechanistic knowledge to design rational new vaccines and other T cell-based therapies to overcome antigen loss or the absence of known antigens, a problem commonly observed during cancer vaccination and CAR-T cell therapy. These new protocols may include approaches to overcome potential tumor evasion strategies to counteract epitope spread.

Using a mouse in vivo model of cancer immunotherapy, we have found that CD8⁺ Tc cells induce ICD on cancer cells, generating spread immunity and a protective Tc cell response against endogenous tumor antigens. The mechanism involved depends on the presence of active caspase-3 and is independent of the mitochondrial cell death pathway. These findings indicate that ICD and epitope spreading during cell death induced by CD8⁺ Tc cells contribute to the efficacy of cancer immunotherapy. Moreover, our findings suggest that mutations in caspase-3 or in pathways regulating its activity might increase the risk of tumor refractoriness and/or recurrence after T cell-based cancer immunotherapy. Clinical trials will be required to test whether the presence of caspase-3 mutations and/or inhibitors like XIAP can be used as biomarkers to predict relapse during immunotherapy.