eIF2α phosphorylation: A hallmark of immunogenic cell death

Immunogenic cell death (ICD) was first discovered as a caspase-dependent (and hence apoptotic) phenomenon elicited by exposing cancer cells to anthracyclines, oxaliplatin or radiotherapy.¹ This apoptotic ICD has turned out to determine the long-term effects of chemotherapy beyond treatment discontinuation relying on the induction of an effective anticancer immune response.^2,3 It is coupled to a premortem autophagic response (that facilitates lysosomal ATP secretion), an endoplasmic reticulum (ER) stress response (that triggers the exposure of calreticulin, CALR; at the cell surface),⁴ the stimulation of a type-1 interferon (IFN) response (ignited by the activation of toll like receptor-3 and presumably also a pathway involving the cyclic GMP-AMP synthase [cGAS] and stimulator of interferon genes [STING]),⁵ as well as the release of abundant cytoplasmic proteins (such as annexin A1, ANXA1) and nuclear proteins (such as high mobility group box 1, HMGB1). Each among these signals is required to initiate specific steps in the anticancer immune response, ATP for the attraction of myeloid cells (via an effect on P2Y2 purinergic receptors), ANXA1 for facilitating the juxtaposition of dendritic cells (DC) and dying cancer cells (via an effect on formyl peptid receptor 1, FPR1), CALR for the transfer of tumor-associated antigens into DC rather than macrophages (via an effect on CD91), HMGB1 for stimulating DC maturation (via an effect on TLR4), and products of the type-1 IFN response including chemokine (C-X-C motif) ligand 10 (CXLC10) for stimulating the recruitment of T lymphocytes (via an effect on CXCR3). In addition, tumor cell-derived DNA, in particular mitochondrial DNA, may elicit the cGAS pathway in DC, thus stimulating their antigen cross-priming activity.⁶ This pathway occurs in a non-redundant cascade, meaning that the absence of one single ligand or receptor abolishes the immune response.¹

As mentioned above, the unfolded protein response (UPR) or endoplasmic reticulum (ER) stress response, also called UPR^ER, is required for the exposure of CALR on the surface of cancer cells as they undergo ICD. The UPR^ER usually consists in the simultaneous activation of three pathways (Fig. 1A), (i) the eukaryotic translation initiation factor 2 alpha kinase 3 (EIF2AK3, best known as PERK)-mediated phosphorylation of eukaryotic initiation factor α (eIF2α), resulting in a change in protein translation that favors the expression of activating transcription factor 4 (ATF4), (ii) the translocation of activating transcription factor 6 (ATF6) from the ER to the Golgi and then the nucleus, and (iii) the endoplasmic reticulum to nucleus signaling 1 (ERN, best known as IRE1α)-mediated splicing of X-box binding protein 1 (XBP1) to generate the XBP1s isoform.⁷ ATF4, ATF6 and XBP1 act as transcription factors to activate a gene expression profile that favors homeostatic adaptation of the cell or its apoptosis.

In the context of chemotherapy with a restricted panel of anticancer agents including anthracyclines or oxaliplatin, ER stress manifests with a PERK-dependent eIF2α phosphorylation, yet does not result in the activation of ATF4, ATF6 nor the IRE1α/XBP1 axis, hence resulting in a rather partial ER stress response.⁴ PERK-dependent eIF2α phosphorylation favors the translocation of calreticulin (CALR) from the ER lumen to the plasma membrane surface.⁸ where CALR acts as an ‘eat me’ signal to favor the phagocytosis of portions of cancer cells by inflammatory dendritic cells, thereby igniting anticancer immune responses.¹

How is it possible that prominent anticancer agents only induce one single arm of the UPR^ER? We are currently exploring two possibilities to explain this ‘split’ ER stress response (Fig. 1B). As a first possibility, the activation of PERK (and perhaps that of other eIF2α kinases) occurs in a non-orthodox fashion without the occupation of heat shock protein family A (Hsp70) member 5 (HSPA5, best known as binding immunoglobulin protein, BiP, or 78 kDa glucose-regulated protein, GRP-78) by unfolded proteins that accumulate in the ER lumen, normally assuring the coordinated activation of all three arms of the UPR^ER.⁷ As a second possibility, we examine the hypothesis that ICD-inducing chemotherapeutics actively inhibit the activation of ATF4 downstream of eIF2α phosphorylation as they also prevent the activation of ATF6 and the IRE1α/XBP1 axis. Indeed, anthracyclines have a major inhibitory effect on protein translation,⁴ which might explain their capacity to subvert the activation of ATF4. However, at this point, we do not understand how they may interfere with the ATF6 and IRE1α/XBP1 axis.

Irrespective of these incognita, it appears to be important that only eIF2α phosphorylation but none of the other arms of the UPR^ER contributes to the immunogenicity of cell death elicited by clinically successful chemotherapeutics. Thus, the quantification of eIF2α phosphorylation can be used as a proxy to assess ongoing ICD in patient samples by means of immunohistochemical methods. Indeed eIF2α phosphorylation within cancer tissues correlates with CALR surface exposure on tumor cells as well as prognostically favorable anticancer immune responses as this has been shown for non-small cell lung cancer⁹ as well as for acute myeloid leukemia.¹⁰