The Unfolded Protein Response in Immunogenic Cell Death and Cancer Immunotherapy

Trends

The ER is a key organelle in cell physiology, and it evolved as an elaborated signaling pathway to cope with life-threatening perturbations of its homeostatic state. This process, called the UPR, is exploited by cancer cells to survive in their microenvironment and to promote tumor progression.

Figure [1]

Many approaches have been investigated to inhibit or exacerbate UPR to kill cancer cells. Among these, a defined set of agents or procedures can induce a form of ER stress-mediated cancer cell death that is immunogenic (the so-called ICD), generating an effective antitumor immunity.

ICD is characterized by a spatiotemporally defined emission of DAMPs that establish a productive interface with immune cells.

Combinatorial strategies based on ICD and immune checkpoint blockade are currently being investigated.

The unfolded protein response (UPR) is a conserved pathway that is stimulated when endoplasmic reticulum (ER) proteostasis is disturbed or lost. Accumulating evidence indicates that chronic activation of the UPR supports the main hallmarks of cancer by favoring cancer cell-autonomous and nonautonomous processes, which ultimately foster the immunosuppressive and protumorigenic microenvironment. However, certain forms of therapy-induced ER stress can elicit immunogenic cancer cell death (ICD), which enables the release of key immunostimulatory or danger signals, eventually driving efficient antitumor immunity. In this review, after a brief discussion of the interplay between ER stress and protumorigenic inflammation, we review the relevance of therapy-mediated ER stress pathways in evoking ICD and how they could be used to optimize current immunotherapy approaches against cancer.

Endoplasmic Reticulum Stress-Induced Unfolded Protein Response

The endoplasmic reticulum (ER) is a crucial organelle that extends throughout the volume of the cytoplasm, mainly serving a threefold purpose: (i) calcium (Ca²⁺) storage and buffering, (ii) lipid biosynthesis, and (iii) productive folding and assembly of secretory and transmembrane proteins (assisted by the agency of a variety of molecular chaperones). Indeed, the ER represents the first compartment of the secretory pathway and about one-third of the polypeptides synthesized by a cell enter the ER lumen, where they undergo different maturation steps (e.g., folding, glycosylation, and disulfide bond formation), thus ensuring that only properly folded proteins exit the ER and reach their final destination [1]. If ER quality control mechanisms fail, misfolded proteins exit the ER and are degraded through the ER-associated degradation pathway or autophagy. Since a single organelle orchestrates these processes pivotal for cellular survival and homeostasis, it is not surprising that ER-associated perturbations have critical consequences for the overall biochemistry and fate of intracellular and extracellular milieus of a cell. Such perturbations include (i) ER protein overloading either due to increased protein translation or inhibition of degradation of misfolded proteins; (ii) hypoxia and/or glucose/nutrient deprivation; (iii) disruption of the ER–Ca²⁺ pool or buffering capacity; and (iv) altered redox regulation of the highly oxidative ER lumen, which impairs the formation of disulfide bonds [2]. Deregulation of ER homeostasis (caused by both intracellular and extracellular stressors) underlies several diseases, including (but not limited to) metabolic and inflammatory diseases, neurodegeneration, retinal dystrophies, cystic fibrosis, and cancer [3].

Considering the relevance of maintaining ER homeostasis, cells have developed a major, evolutionarily conserved adaptation mechanism to rapidly sense and adequately respond to perturbations in the ER folding machinery, called the unfolded protein response (UPR). The UPR aims to restore the ER-associated protein folding capacity by increasing ER volume and expression of ER-associated chaperones, as well as (transiently) attenuating global protein translation. In mammalian cells, the UPR is governed by three ER stress sensors, namely, inositol-requiring enzyme 1 alpha (IRE1α), protein kinase R-like ER kinase (PERK), and activating transcription factor 6 (ATF6) [4]. In physiological conditions, these proteins are kept in an inactive state by the master regulator of the UPR, that is, the glucose-regulated protein (GRP)-78 (also known as BiP) [5]. However, in case of loss of ER proteostasis or Ca²⁺ dysregulation, BiP disassociates from these ER stress sensors, thereby activating the UPR signaling pathway (Figure 1). Overall, when the ER stress is of mild intensity, the UPR facilitates cell-autonomous re-establishment of cellular homeostasis and survival, as well as inflammatory responses aimed at tissue turnover or repair. However, when ER stress intensity is too severe (thereby rendering a repair and rescue program obsolete), the UPR engages signaling pathways culminating into cell death, which are driven by the coordinated (transcriptionally regulated) actions of ATF4 and C/EBP homologous protein (CHOP), downstream the PERK pathway (Figure 1) [6, 7]. In specific contexts, the activation of inflammatory transcriptional programs and/or danger signaling pathways may accompany lethal UPR signaling, aiming to alert the immune system about diseased or damaged cells (discussed later). Although cell death mediated by ER stress is mainly executed through the apoptotic program [2], recent studies have linked loss of ER homeostasis also to various forms of regulated necrosis, such as necroptosis [8] and ferroptosis [9]. An in-depth discussion of the molecular mechanisms underlying these emerging regulated cell death modalities and their intersection with the immune system is beyond the scope of this review. In addition, how these cell death pathways modulate ER stress-mediated extrinsic responses, such as inflammation and immunity, has not been elucidated yet.

This coordinated regulation of cell fate and communication with extracellular entities is of great importance for cancer progression and anticancer therapy response, especially immunotherapy. In this review, we briefly examine the ER stress-associated signaling pathways regulating tumor-associated inflammation and antitumor immunity. We primarily focus on the relevance of therapeutic modulation of ER stress [especially in the context of immunogenic cell death (ICD)] in the cancer cells and how this may influence the overall susceptibility or resistance to cancer immunotherapy. The role of UPR in immune cells regulating crucial immunomodulatory and inflammatory responses within the tumor microenvironment (TME) will not be discussed here but has been addressed in excellent recent reviews [10, 11].

ER Stress-Driven Protumorigenic Inflammation

Tumor cells, to grow and proliferate, need to survive in a deeply hostile (micro)environment. Several recent reviews highlight that this peculiar condition evokes chronic ER stress, and that UPR is exploited by cancer cells not only for survival, but also to promote tumor growth and progression by supporting nearly all the (established) hallmarks of cancer [12, 13, 14, 15]. Adaptation to the increasingly stressed TME requires a plastic process enabling cancer cells to coevolve with the tumor stroma while maintaining an enhanced ability to survive and outgrow. Overactivation of the UPR provides a clear survival advantage to cancer cells since beyond supporting their oncogene-driven, increased synthetic and secretory demand, it fosters crucial remodeling of the TME and proinflammatory pathways with profound autocrine and paracrine impact on stromal and immune cells [4, 11, 16, 17]. Indeed, several pieces of evidence have disclosed the ability of cancer cell-associated UPR [in particular, the PERK–ATF4 and IRE1α–X-box binding protein 1 (XBP1) pathways] to release vascular endothelial growth factor and other proangiogenic cytokines, thus interfacing with the endothelial cells, leading to dysfunctional and tumor-supporting angiogenesis [13, 18].

However, among the hallmarks of cancer, evasion from immune destruction and facilitation of protumorigenic inflammation are two of the most emergent themes of research addressing the role of ER stress or UPR in tumorigenesis over the last decade [19].

Almost all neoplasms undergo the process of cancer immunosurveillance before reaching an equilibrium marked by infiltration of mostly tumor-supporting immune cells, which favors the clonal expansion of cancer cells that have evolved the capability to escape immune control [20]. Such a TME is characterized by T-cell exhaustion [resulting from chronic but suboptimal exposure to tumor-associated antigens (TAAs) in absence of costimulation and metabolic stress] or poor infiltration of cytotoxic T cells, and presence of immunosuppressive immune cells (i.e., Type 2 polarized cells like T_H2 CD4⁺ T cells, M2 macrophages, N2 neutrophils, or autoregulatory cells like T_reg and myeloid-derived suppressor cells) [20, 21]. This immune microenvironment fuels tumor growth because on the one hand it is incapable of killing the cancer cells (due to T-cell anergy), yet on the other it encourages either tolerance toward the tumor (through autoregulatory immune cells) or actively contributes to tumor progression via protumorigenic cytokines/chemokines (through Type 2 polarized or immunosuppressive immune cells).

In general, UPR activation in different cell types has been linked to inflammatory pathways through, for example, PERK–eukaryotic initiation factor 2 (eIF2α)-mediated inhibition of de novo synthesis, which by reducing the levels of the short-lived IκB results in the activation of nuclear factor-κB (NF-κB) [22], the major regulator of inflammatory responses [23]. Likewise, the scaffolding function of IRE1α can cause the tumor necrosis factor receptor-associated factor 2 (TRAF2)-mediated activation of IκB kinase and apoptosis signal-regulating kinase 1–c-Jun N-terminal kinase (ASK1–JNK) signaling [24, 25], while ATF6 may activate NF-κB through an AKT-dependent pathway [26] (Figure 2). ER stress can trigger inflammatory responses also through the activation of the NLRP3 inflammasome that culminates in the production of interleukin-1β (IL-1β) and IL-18, through both UPR-dependent and UPR-independent pathways [27, 28]. Which of these UPR-modulated mechanisms prevalently operate in the cancer cells will likely depend on various intrinsic (oncogene expression, metabolic program, oxidative stress, etc.) and extrinsic TME factors (hypoxia, nutrient availability, etc.). However, once overactivated, these UPR-regulated pathways ultimately converge in the secretion of particular proinflammatory cytokines (e.g., IL-11, IL-1β, IL-6, IL-23, and tumor necrosis factor [23, 29]) in the extracellular milieu. On the one hand, these protumorigenic cytokines tend to promote cellular survival and engage the proliferation signaling cascade(s) in an autocrine/paracrine fashion, by binding their cognate receptors on cancer cell surface (which may also show upregulation, e.g., IL-6-based engagement of IL6R/gp130 on cancer cells [30]). On the other hand, these chemokines and cytokines can also recruit and/or sustain immunosuppressive immune cells to further fuel production of protumorigenic factors in a feed-forward loop [23]. While attracting regulatory or Type 2 polarized immune cells, cancer cells may also exploit UPR to blunt any de novo generation of antitumor immunity. For example, IL-23 secreted by tumor cells negatively influences the activity of CD8⁺cytotoxic T cells (cytotoxic T lymphocytes) and supports the expansion of protumorigenic T_H17 cells [31]. Through the activation of the NF-κB and p38 mitogen-activated protein kinases signaling pathways, ER stress can contribute to the upregulation of cyclooxygenase-2 in the cancer cells and the release of prostaglandin E₂, a negative regulator of antitumor immunity responses [32].

Moreover, activation of the UPR in cancer cells can propagate ER stress in myeloid cells, through a process termed ‘transmissible ER stress’. This transmissible ER stress causes a proinflammatory/suppressive phenotype in myeloid cells, hallmarked by the upregulation of arginase and secretion of IL-6 and IL-23, thereby further compromising antitumor immunity responses [33]. Although the identity of the cancer cell-associated factors responsible for this crosstalk is still elusive, the enhanced proinflammatory responses induced in the receiver myeloid cells required a cancer cells-secreted Toll-like receptor 4 (TLR4) agonist [34].

The UPR can also modulate the presentation of antigens via the MHC molecule Class I, which is loaded with the antigenic peptide within the ER before being transported to the cell surface. However, how exactly ER stress affects this process in a tumoral context is still debated [35].

Altogether, these studies indicate that the initial aim of ER stress response is to repair the tumor-inflicted tissue ‘wounding’ by attraction of innate immune cells (e.g., neutrophils, monocytes, mast cells) that exert potent antimicrobial (and in this case tumor cytotoxic) functions by producing high levels of reactive oxygen species (ROS) and inflammatory cytokines and favor new tissue formation by releasing proangiogenic factors and metalloproteinases. Yet, chronic continuation of such responses ultimately fuels autoregulatory inflammatory loops and degradation of tumor stroma supporting tumor migration, eventually favoring (rather than impede) tumorigenesis [36].

UPR and ER Stress-Driven Immunogenic Cancer Cell Death

The chronic persistence of mild ER stress modulates the extracellular microenvironment to support tumorigenesis. However, the overall situation changes when ER stress is exceedingly elevated and UPR is skewed toward the proapoptotic module. Indeed, in this situation, the ER may communicate the danger state of a stressed cell to the microenvironment eventually orchestrating an antitumor immune response. Apoptosis has been traditionally documented to be an immunologically silent or tolerogenic process (and hence referred to as ‘physiological apoptosis’), while necrosis has been described as a proinflammatory cell death pathway that facilitates protumorigenic processes [37]. Therefore, since most of the anticancer therapies currently used in the clinic primarily induce a mixture of physiological apoptosis and necrosis (depending on the dosage and number of cycles administered), they were all thought to facilitate either tolerogenicity toward dying cancer cells or protumorigenic inflammation [38]. However, during the last decade, this dogmatic view has changed. Indeed, several independent studies have identified that a particular, yet diverse set of commonly used anticancer therapies (such as anthracyclines, radiotherapy, oncolytic viruses, photodynamic therapies) can accentuate the immunogenic potential of dying cancer cells. A distinctive structure–function relationship for different ICD inducers is virtually nonexistent, as exemplified by the different immunogenic potential of two structurally related compounds like oxaliplatin and cisplatin. The key to ICD induction for all these inducers is the concomitant and sustained induction of ROS and ER stress [39]. In fact, ER stress is so crucial for the paradigm that different ICD inducers may be classified based on the quality of ER stress they induce (e.g., Type I or Type II inducers of ICD, discussed later). Interestingly, during ICD, while all three branches of the UPR are activated, yet only the PERK branch is so far mandatory for ICD. ICD-inducing anticancer treatments elicit a form of ER stress that is capable of warning the immune cells of a state of danger, through the release (or surface exposure) of immunostimulatory factors or damage-associated molecular patterns (DAMPs) acting as danger signals. DAMPs are endogenous molecules with various housekeeping functions, which acquire immunomodulatory functions when released or surface exposed by the dying cancer cells. In analogy to the pathogen-associated molecular patterns, DAMPs bind to pattern recognition receptors (including various TLRs) on the innate immune cells [e.g., professional antigen-presenting cells, like dendritic cells (DCs)] and favor the establishment of a productive interface between the dying cancer cells and the immune system (Figure 3). This, in turn, results in the priming of the host immune system for the TAAs and subsequent elicitation of TAA-specific T-cell-mediated immune responses, leading to the elimination of the residual cancer cells as well as the establishment of an immunological memory [40].

There are specific DAMPs associated with ICD: (i) early surface exposure (ecto-) of the ER chaperone calreticulin (CRT) during the preapoptotic phase [41], wherein ecto-CRT acts as an ‘eat-me’ signal and is recognized by the CD91 receptor on phagocytes, stimulating the engulfment of dying cancer cells. Of note, other ER or cytosolic chaperones like heat shock protein (HSP)-70/90 can also be surface exposed and may act, redundantly with ecto-CRT, as ‘eat me’ signals [42]; (ii) active or passive secretion of ATP [43] during the preapoptotic or early/midapoptotic phases that could act either as a potent short-range ‘find-me’ signal for attracting monocytes by binding the P₂Y₂ purinoceptors, or as a proinflammatory (inflammasome-activating) molecule by binding the P₂X₇ receptor on innate immune cells; (iii) passive, late-apoptotic release of high mobility group box 1 (HMGB1) that binds TLR2 and TLR4 on DCs, stimulating production of proinflammatory cytokines and assisting in proper antigen presentation [44]; and (iv) passive release of nucleic acids, including double-stranded DNA, that signal via TLR-7/8/9 on innate immune cells (like neutrophils), thereby regulating their activation and anticancer activity [45]. Of note, cells undergoing ICD can also release Annexin A1 that binds to formyl peptide receptor-1/2 (FPR1 or FPR2) on innate immune cells. However, the immunological impact of Annexin A1 is controversial since one study documented a proimmunogenic effect centered on DCs [46], while another documented an immunosuppressive effect centered on neutrophils [45].

Beyond emission of DAMPs/danger signals, the cancer cells undergoing ICD can also activate the immune system by performing ‘altered-self mimicry’, that is, mimicking host defense response typically triggered by pathogens or viral infection, which stimulates strong immunological responses against the ‘altered-self’ cell (e.g., cells undergoing ICD). For instance, doxorubicin-induced ICD has been shown to be associated with the release of dsRNA that binds to TLR3 on other cancer cells (paracrine signaling), thereby inducing release of Type I interferon (IFN) cytokines [47], which mediate anticancer immunity. By contrast, cancer cells dying via ICD can also orchestrate a cancer cell-autonomous release of chemokine (C–X–C motif) ligand 1 (CXCL1), chemokine (C–C motif) ligand 2 (CCL2), and CXCL10 that attract neutrophils and eventually pave the way for their ATP- and nucleic acids-based activation, in a manner reminiscent of response to bacterial or viral infections. These neutrophils eventually exert direct anticancer cytotoxic activity against residual cancer cells via respiratory burst [45]. However, there is a significant inducer- and context-dependent plasticity in the amount, diversity, kinetics, and immunological activity of DAMPs, danger signals, and cytokines/chemokines emitted upon treatment (Table 1) [48, 49]. At the same time, many non-ICD inducers have been observed to elicit mobilization of one or more DAMPs [49], although without promoting effective DC maturation and antitumor immunity.

Table 1Characteristics of Clinically Relevant ICD Inducers
ICD inducer	Indications	On-target effect	ER effect	DAMPs	Note
Type I ICD inducers
Bleomycin	Multiple hematological and solid tumors	DNA-damaging agent	ER stress induced through an unknown mechanism	CRT, ATP, IFN, HMGB1	Ambivalent immunological outcome due to expansion of Tregcells
Bortezomib	Mantle cell lymphoma Multiple myeloma	Proteasomal inhibitor	ER stress induced indirectly by protein overload due to impaired ER-associated degradation machinery	CRT, HSP70, HSP90, IFN, HMGB1
Cyclophosphamide	Multiple hematological and solid tumors	Alkylating agent	ER stress induced through an unknown mechanism	CRT, ATP, IFN, HMGB1	Immunosuppressive at high doses, decreases the percentages of Tregat lower or metronomic doses
Anthracyclines (doxorubicin, epirubicin, idarubicin, mitoxantrone)	Multiple hematological and solid tumors	DNA-intercalating agent	Associate with cellular membranes (including ER) and cause ROS production resulting in ROS-based ER stress	CRT, ATP, IFN, HSP70, HMGB1
Oxaliplatin	Colorectal carcinoma	Alkylating agent	Associates with cellular membranes (including ER) and causes ROS production resulting in ROS-based ER stress	CRT, ATP, IFN, HSP70, HMGB1
Type II ICD inducers
Hypericin-photodynamic therapy	Nonmelanoma skin cancer, cutaneous T-cell lymphoma, mesothelioma, basal or squamous cell carcinoma	Induction of oxidative stress with damage to ER membranes	ER membrane-associated photosensitive drug causes massive production of ROS and ROS-based ER stress upon light activation	CRT, ATP, HSP70, HSP90, HMGB1
Oncolytic viruses	In clinical trials for multiple solid cancers	Selective lysis of malignant cells	ER stress induced through viral protein overload

View Table in HTML

Either mRNA and/or protein level of DAMPs (together with markers of ER stress) has been proposed as possible biomarkers for ICD induction in patients. However, their prognostic value shows significant variability according to different studies performed in various cancer types, possibly due also to technical and conceptual limits (Table 1).

However, while ROS-mediated ER stress is a central process conferring to apoptotic cell death an immunogenic character, ER stress may be dispensable for ICD signaling and antitumor immunity elicited by other forms of regulated cell death, such as necroptosis. Indeed, a recent study showed that cancer cells induced to undergo necroptosis, via Fas-associated protein with death domain (FADD) dimerization-mediated receptor-interacting serine/threonine-protein kinase 3 (RIPK3) activation, released some of ICD-related DAMPs (i.e., ATP and HMGB1) and triggered adaptive immune responses in vivo without an apparent contribution of ER stress [50]. In the context of chemotherapy-induced ICD, the necroptotic mediators RIPK3 and mixed lineage kinase domain-like protein (MLKL) were shown to contribute to immunogenicity and DAMPs emission, but whether chemotherapy-induced ER stress was required has not been evaluated [51]. Finally, it would be important to evaluate immunogenic potential of other forms of therapy-induced regulated necrosis, such as ferroptosis, distinguished by elevated ER stress and ROS levels.

Thus, while cancer cell-constitutive activation of the prosurvival program of the UPR facilitates many aspects of the protumorigenic microenvironment, harnessing the lethal arm of the UPR through oxidative stress enables the release of cancer cell death-associated danger signals, which are decoded by the immune system and escalate antitumor immune responses.

ER Stress-Driven Pathways during ICD in Cancer Cells

Considering the compelling reliance of ICD on the induction of a sustained ER stress, it is not surprising that the different intensity and the kinetics of ER stress may have a strong impact on the quality and quantity of danger signaling. This led to the subdivision of ICD inducers into two broad categories (Table 1): Type I and Type II ICD inducers [40]. ‘Type I ICD inducers’ encompass all the drugs that trigger ICD-associated immunogenicity through secondary ‘off-target’ (mostly mild) ER stress in parallel with the main ‘on-target’ effect that drives apoptosis via non-ER targets. Most clinically employed ICD inducers, such as anthracyclines, oxaliplatin, bortezomib, cyclophosphamide, and radiotherapy, fall within this category [48]. ‘Type II ICD inducers’, instead, selectively target the ER and orchestrate both danger and apoptotic signaling through ‘focused/on-target’ (ROS-based) ER stress. The first treatment identified as Type II ICD inducer was hypericin (Hyp)-based photodynamic therapy [52], followed by identification of other treatments, such as some oncolytic viruses (e.g., Newcastle disease virus) [53, 54] and platinum^II–N-heterocyclic carbene complex [55]. The quality and quantity of ER stress elicited by the Type II ICD inducers are superior to Type I ICD inducers in terms of both overall amounts and speed of the emission of danger signals/DAMPs exposed during the preapoptotic stage [40]. Irrespective of this, although the signaling pathways leading to DAMP exposure/secretion show a certain degree of dissimilarity between Type I and Type II ICD inducers (Figure 4), they both share key apical and proximal mediators and require an intact secretory pathway and the actin cytoskeleton.

As presented in Figure 4, PERK is at the ‘core’ of ICD and the apical coordinator of DAMP trafficking mechanisms. Intriguingly, the function of PERK seems to differ between Type I (activating classical UPR pathway through eIF2α phosphorylation) and Type II (regulating expansion of the secretory pathway through an as-yet-unknown interactor) ICD inducers. Recently, a peculiar ICD inducer at the interface between Type I and Type II, high hydrostatic pressure, has also been shown to engage a PERK-modulated pathway for ecto-CRT exposure involving caspase-2 [56].

The reason behind the reliance of the danger signaling on PERK rather than on other UPR sensors (e.g., IRE1α) remains enigmatic. Likewise, the mechanism, shared by Type I and Type II ICD inducers, linking PERK to intracellular Ca²⁺elevation and the actin cytoskeleton, in the path to mobilize DAMPs, remains unclear. This discrepancy could be explained by a newly discovered, UPR-independent function of PERK in modulating the dynamics of the actin cytoskeleton, through its interaction with the actin-binding protein Filamin A (FLNA) [57]. A recent study from our laboratory has provided compelling evidence showing that PERK is able to sense and rapidly respond to cytosolic Ca²⁺ elevations through its cytosolic domain, by enabling the formation of ER–plasma membrane (PM) appositions, resulting in opening of the STIM1-ORA1 sensor stromal interaction molecule 1 (STIM1)-​ORAI calcium release-activated calcium modulator 1 (ORAI1) channel and Ca²⁺ influx, through its interaction with FLNA [57]. This newly identified function of PERK in forming ER–PM contact sites could be relevant for the mechanisms of trafficking of DAMPs via the secretory pathway and SNAP (soluble N-ethylmaleimide-sensitive factor attachment protein) receptor (SNARE)-mediated exocytosis, house-keeping processes that have been shown to be required for Type I and Type II ICD inducers [52, 58]. A PERK–FLNA axis could in fact sustain, through the rapid formation of ER–PM junctions, intracellular Ca²⁺ levels and Ca²⁺-modulated actin cytoskeleton remodeling, which are key mediators of ER-to-Golgi trafficking and vesicle exocytosis [59].

Overall, the key function of the UPR and ER stress sensors in modulating ICD immunogenicity by guiding the surface exposure and release of DAMPs and the master role of PERK in this process have been thoroughly demonstrated. However, more studies are still needed to unravel key mechanisms underlying ICD induction that are common to, if possible, all ICD inducers.

Clinical Implications of ER Stress-Driven Immunogenicity for Cancer Immunotherapy

Most of the currently identified ICD inducers have been intensively used in clinicalpractice and often (but not always) constitute the first line of therapy (especially when employed through ‘on-label’ use, Table 1) [60]. However, the current (direct) evidence for ICD is largely derived from rodent studies (Box 1), whereas clinicalevidence is still limited. Indeed, highly reliable biomarkers for monitoring ICD induction in situ are still missing (Table 2). Nevertheless, retrospective analyses support (to a certain extent) the existence of ICD in the clinical settings, based on the correlation between the expression levels of key ICD-related genes and a more favorable prognosis selectively in patients treated with ICD inducers [61, 62]. Similarly, mutations or defects in the molecular machinery mediating ICD are usually associated with worse prognosis in patients treated with ICD inducers [43, 63, 64, 65].