PERK is a stress-sensing kinase with roles in antitumor immunity

Mammalian cells regularly encounter exogenous cell stress such as nutrient and amino acid deprivation, hypoxia, and reactive oxygen species (ROS). Such cell stressors lead to a burden of misfolded proteins in the endoplasmic reticulum (ER) lumen that engage ER stress sensors to trigger the integrated stress response (ISR) in an attempt to return the cell to homeostatic conditions [1, 2]. Three stress sensors; inositol-requiring enzyme-1α (IRE1α), activating transcription factor 6 (ATF6), and PKR-like ER kinase (PERK) are activated by the unfolded protein response (UPR) [3,4,5]. Each of the aforementioned sensors lead to the activation of the regulatory transcription factors XBP1, ATF6p50, and activating transcription factor 4 (ATF4), respectively [6, 7]. The transcriptional programs induced by ER stress and the UPR lead to increased expression of genes required for protein folding, the inhibition of new protein synthesis through regulated IRE1-dependent mRNA decay (RIDD) and the induction of autophagy programs that help to eliminate unfolded proteins in the ER [8, 9].

A key step in activation of the UPR is the direct inhibition of protein translation following PERK activation. Sensing of a burden of unfolded/misfolded proteins catalyzes the chaperone protein Bip/GRP78 to dissociate from binding the luminal domain of PERK enabling activation of PERK through dimerization and autophosphorylation [5]. Activated PERK inhibits eukaryotic translation initiation factor 2α (eIF2α) through phosphorylation on serine residue 51 which prevents the formation of Met-tRNAiMet complexes and prohibits initiation of translation of cap-dependent mRNAs [10]. Following this discovery, eIF2α was shown to be required for cell survival and organism viability in response to unfolded protein accumulation in the ER. Mice bearing homozygous serine to alanine mutations at residue 51 of eIF2α die shortly after birth due to hypoglycemia and defects in the gluconeogenesis pathway [11]. The kinases PERK and general control nonderepressible 2 (GCN2) were found to play a central role in this process and were required for gluconeogenesis to take place due to their roles in phosphorylating eIF2α. In addition to temporarily restricting protein synthesis as a protective mechanism to the cell, PERK subsequently allows for resumption of protein synthesis after ER stress through a mechanism involving the stimulation of the protein growth arrest and DNA damage-inducible protein, GADD34 [12, 13]. GADD34 is a phosphatase that dephosphorylates p-eIF2α in a PERK-dependent manner, highlighting an important negative feedback loop of the stress response.

The acute role of PERK–p-eIF2α to inhibit protein translation in antitumor immunity remains elusive. Activation of the stress response is intimately connected to T cell differentiation. Primed T helper 2 (Th2) cells exhibit enhanced p-eIF2α and increased features of the ISR. Upon restimulation, Th2 cells exhibit dephosphorylation of p-eIF2α followed by cytokine secretion and cycloheximide treatment extinguished the effect, tying regulation of protein translation to T cell function [14]. We were the first group to show that effector CD8 T cells exposed to the tumor microenvironment (TME) or in tumors experience inhibition of protein synthesis that is associated with poor tumor control [15] but the role of PERK–p-eIF2α to orchestrate this process remains untested and the role of PERK in antitumor immunity is ill-defined. The TME is replete with elements such as lactic acid and ROS that likely present an acute stress to tumor infiltrating T cells (TILs) [16, 17]. However, it was demonstrated that mice with conditional deletion of PERK in the T cell compartment experience superior tumor control of B16 melanomas relative to WT counterparts [18] and we found that PERK-deficient T cells confer powerful tumor control in the adoptive cellular therapy setting [19]. In both instances, the deleterious role of PERK in antitumor immunity was attributed to activation of the chronic arm of the stress response. In contrast to the acutely protective role of PERK–p-eIF2α, a persistent burden of unfolded/misfolded proteins in the ER lumen induces the chronic stress response comprised of ATF4/C/EBP homologous protein (CHOP) and ER oxidoreductase 1α (ERO1α) activity that can induce cell death [20, 21].

In line with PERK’s role to restrict protein translation as a means to protect cells from ER stress, PERK mediates cell cycle. Cell cycle progression is tightly controlled by the temporal expression of cyclins and cyclin-dependent kinases (CDKs) [22]. In cancer immunotherapy, cell cycle inhibitors have been combined with checkpoint therapies and have shown tremendous promise as a treatment modality for cancer patients. The effect occurs by enhancing cytotoxic T cell function toward tumor clearance and by suppression of regulatory T cells [23]. In a breast cancer model, the CDK4/CDK6 inhibitor Abemaciclib led to synergy when administered in combination with programmed death ligand 1 (PD-L1) checkpoint blockade, enabling tumor clearance and development of long-lived memory T cells [24]. We found that anti-programmed cell death protein 1 (a-PD-1) therapy combined with PERK inhibition resulted in complete and durable tumor responses in a mouse model of sarcoma [19]. Following ER stress, activation of PERK results in the loss of cyclin D1 and subsequent cell cycle arrest. Brewer et al. demonstrated that inhibition of PERK-induced p-eIF2α restricted accumulation of cyclin D1; whereas, overexpression of PERK blocked cyclin D1 expression [25]. Loss of PERK kinase activity abrogated the aforementioned effect, directly implicating PERK’s role in phosphorylation as central to regulation of cell cycle arrest [25]. Thereafter, it was discovered that ER stress and activation of PERK also led to the induction of p53 and cell cycle arrest through multiple mechanisms [26, 27]. Based on the robust known roles of PERK to orchestrate elements of biology that are proving to be cornerstones of successful cancer immunotherapy, study of regulation by the ER kinase in tumor immunity is prudent.

Potential impacts of PERK on antitumor metabolism and immunity

Cell stress is intimately linked to bioenergetics. The potential impact of the stress response to program cell metabolism as a reaction to the stress of the TME could present a unifying target to the cancer immunotherapy field. The metabolism of T cells is closely linked to their environment. T cells within the TME of solid tumors face unique stresses that impact their metabolic profiles. In the TME, T cells must overcome cell stressors that negatively impact their metabolism and subsequent function such as hypoxia, high lactate, glucose deprivation, and competition for nutrients [28]. In addition to hypoxia and nutrient deprivation, chronic antigen exposure can lead to exhausted T cell phenotypes with impaired mitochondrial function [29]. Scharping et al. demonstrated that T cells infiltrating murine and human tumors exhibited depletion of mitochondrial biomass and impaired oxidative phosphorylation (OXPHOS) due to loss of PPARG coactivator 1α (PGC1α) expression. PGC1α is a key regulator of mitochondrial biogenesis and forced PGC1α expression restored biomass and improved T cell antitumor function [30]. While suppression of OXPHOS in the TME limits T cell function, general competition for nutrients between tumor and T cells dramatically impacts T cell function. As a consequence of enhanced glycolysis within antigenic solid tumors, restriction of glucose access to effector cells within solid tumors leads to T cell dysfunction [31]. Based on the projected protective role of PERK–p-eIF2α in T cells exposed to the stress of tumors it is surprising that PERK-deficient T cells appear to control tumor growth better than littermate controls [18, 19] and the data suggest that the specific metabolic role of PERK in T cells in tumors needs to be elucidated.

Given the competition for glucose that persists in the TME, reprogramming T cell metabolism away from glucose dependence has proven efficacious to promote tumor control and a metabolic shift is accompanied by a change in lineage fate [32, 33]. It is established that T cells that exhibit potent antitumor function display both proficient effector function and the capacity to persist as long-lived memory cells. Enhanced spare respiratory capacity (SRC) and OXPHOS are metabolic traits associated with memory T cells and expression of the enzyme carnitine palmitoyltransferase (Cpt1a) that catalyzes β-oxidation of fatty acids (FAO) generates such traits resulting in T cells with productive tumor control [34]. Reprogramming T cells through transient glucose restriction enables heightened glycolytic capacity measured by increased extracellular acidification rates in response to glucose and oligomycin treatments. The enhanced glycolytic capacity endowed by transient glucose deprivation leads to superior antitumor function of T cells infused into solid tumors [35]. Acute nutrient deprivation such as transient glucose restriction may allow reprogramming of antitumor immunity that ensures T cells can resist the metabolic stress of the TME. For example, inhibition of glycolysis was previously shown to reprogram T cells to depend on FAO that supports antitumor immunity through a robust store of mitochondrial ATP associated with increased effector function in tumors and prolonged survival [32]. Given that acute glucose deprivation increased the capacity for antitumor cytokine secretion [35], it is possible that metabolic conditioning reshapes the ISR that was previously shown to be integral for cytokine secretion upon restimulation [14].

Recent literature contests the importance of FAO for memory T cell development [36] and other metabolic programs are essential to program superior antitumor immunity. Of critical importance, the pathways that regulate glycogen synthesis (glycogenesis) and breakdown (glycogenesis) impact T cell differentiation and function. Ma et al. found that gene deletion and pharmacological inhibition of gluconeogenesis enzyme phosphoenolpyruvate carboxykinase 1 (PCK1) impaired memory T cell development and damaged tumor control [37]. Inhibition of glycogenolysis enzyme glycogen phosphorylase proved critical for memory T cell development in vivo and in vitro, propelling the breakdown of glycogen to the forefront of T cell lineage fate. T cells deficient in gluconeogenesis or impaired in glycogenolysis were unable to control tumor growth and could not retain in vivo development to memory [37]. In our 2019 study, we found that T cells devoid of PERK exhibited memory T cell metabolic traits such as enhanced oxidative metabolism, elevated SRC, and reduced accumulation of ROS in mitochondria [19]. Gene array data of the UPR between cytokine conditioned effector and memory-like T cells in vitro yielded the finding that the PERK axis consisting of Atf4, Ddit3 (Chop), and Ero1l was the sole UPR axis upregulated in cytokine-conditioned effector T cells relative to memory-like cells [38] and proteomic analysis of these populations confirmed these results [15]. These data suggest a potential role of the stress response in T cell immunometabolism and lineage fate.

Though it is established that T cells with memory-like bioenergetics promote superior tumor control, it is not understood whether alternative metabolism enables stress resistance in the TME. The energy sensor mammalian target of rapamycin (mTOR) guides energy production in nutrient replete conditions through glycolysis. mTORC1 and mTORC2 play distinct roles in influencing CD8 T cell differentiation to effector and memory phenotypes. Forced mTORC1 activation enhances glycolytic metabolism and effector maturation generating better antitumor immunity as a result of hyper effector function [39]. Genetic ablation of mTORC1 activation reduced effector T cell function, but retained memory characteristics. Tumor control was lost due to abrogation of effector function. Alternatively, deletion of mTORC2 led to metabolic changes that promoted both memory and effector T cell traits, but antitumor immunity was not tested [39]. PERK exhibits lipid kinase activity acting on the substrate diacylglyceride (DAG), which produces phosphatidic acid and leads to mTOR activation [40]. The concept has not been tested, but PERK-mediated control of mTOR could explain memory-like metabolic traits in PERK−/− T cells (Fig. 1). In contrast to mTOR signaling in nutrient-rich conditions, AMPK is a metabolic energy sensor that responds to nutrient stress [41]. The role of AMPK has not been widely studied in tumor immunity, but several papers showed that pharmacological stimulation of AMPK promotes tumor control [15, 42]. The data suggest that priming an energy sensor designed to control cell function under stress supports antitumor function. Collectively, these findings highlight the unique potential to modulate the ER stress response and UPR to influence metabolic programs and cell fate decisions in T cell-based immunotherapy.

Fig. 1
figure 1

Metabolic avenues of PERK activity. In response to ER stress, the stress-sensing kinase PERK phosphorylates the transcription initiation factor eIF2α to prevent cell cycle progression through loss of cyclin D1, canonical protein translation, and prohibit generation of the active nuclear form SREBP1 (SREBP1c) that induces lipogenesis. Additionally, PERK phosphorylates diacylglyceride (DAG) resulting in the production of phosphatidic acid (PA), and subsequent activation of mTOR signaling

Full size image

In nutrient-rich conditions, mTORC1 promotes lipid storage through activation of the transcription factor sterol regulatory element binding protein 1 (SREBP1) [43]. Inhibition of mTOR with Rapamycin prevented nuclear accumulation of SREBP1, and RNA silencing of SREBP1 blocked lipogenesis and impaired increases in cell size. The effects were shown to be dependent on Akt [43]. In the context of tumor immunity, Akt is deleterious to antitumor T cells based on hindrance of FAO metabolism and memory T cell formation [33]. Together, the mTOR/Akt axis that regulates SREBP1 appears to be important for T cell-mediated tumor control though the effects need to be further elucidated in the endogenous settings. While PERK has been shown to regulate mTOR, PERK is also required for SREBP1 activation. Gene deletion of PERK in mammary epithelial cells leads to loss of cleaved SREBP1 (SREBP1c), the active nuclear form of SREBP1. The effect was shown to be dependent on p-eIF2α and resulted in abrogation of key lipogenic enzymes, thus deletion of PERK in mouse embryonic fibroblasts (MEFs) led to impaired differentiation into adipocytes [44]. The data highlight the importance of the PERK–SREBP1 axis in lipid metabolism. A consequence of SREBP1 activation is the induction of fatty acid synthesis (FAS) [45], a process that cannot occur simultaneously to FAO. While mTOR/Akt and PERK activate FAS through regulation of SREBP1, under nutrient stress AMPK directly phosphorylates and inhibits SREBP1 [46] to restrict FAS and promote FAO [47]. Collectively, these data suggest that PERK activity in response to cell stress contains the unique capacity to program master regulators of antitumor metabolism.

ATF4 is a stress-induced transcription factor with potential roles in antitumor immunity

While PERK has intrinsic kinase activity that allows for regulation of multiple metabolic players, PERK-mediated phosphorylation of eIF2α has potent consequences for cell biology and fate. PERK phosphorylation of eIF2α results in the selective translation of the transcription factor ATF4 [6]. In this way, ATF4 acts as a critical downstream effector of the PERK arm of the UPR to induce expression of genes involved in processes crucial to the ISR, such as amino acid metabolism and the oxidative stress response [48]. Globally, phosphorylation of eIF2α is a key process in numerous stress-signaling pathways as the post-translational modification can be catalyzed through multiple stress-induced kinases that include (1) PERK, in response to ER stress [21, 49, 50], (2) GCN2, in response to amino acid deprivation [51, 52], (3) heme-regulated inhibitor (HRI) responding to oxidative stress [53] or (4) protein kinase RNA-activated (PKR), in the face of stress response to viral infection [54]. Upon phosphorylation of eIF2α by one of the aforementioned stress kinases, global protein synthesis is suppressed to temporarily alleviate ER, amino acid, oxidative, or viral stressors [6, 55]. Thus, ATF4 activation induced by p-EIF2α is a converging point for multiple ER kinases responding to diverse cell stressors [21, 48, 55, 56] that ensures the capacity to tune cell survival in the face of acute stress [48] and to induce cell death in response to a burden of persistent/chronic misfolded proteins (Fig. 2) [20]. In the context of antitumor immunity, it is likely that differential TME conditions engage the unique ER kinases that unilaterally converge on ATF4. Treatment with ovarian cancer ascites induces phosphorylation of PERK in CD8 T cells, indicating that the TME alone generates unfolded protein stress that activates the PERK response [18]. We found that the TME restricts protein synthesis in T cells in tumors, suggesting that a critical downstream function of the stress response mediated by p-eIF2α may be enacted to thwart antitumor immunity [15]. Multiple studies have shown that ROS secreted by tumor cells in the TME suppress T cell activation and infiltration to tumors [57, 58]. Though activation of HRI in T cells by TME ROS has yet to be tested, it is plausible that the stress kinase could play a role in controlling antitumor outcomes. Finally, in murine gliomas T cell-specific GCN2 was required for proliferation and function in tumors [59], suggesting activation of another stress kinase pathway that leads to induction of ATF4.

Fig. 2
figure 2

ATF4 is a converging point for cell stress pathways. The induction of ER stress kinases PERK, HRI, GCN2 or PKR through the URP, oxidative stress, nutrient deprivation or viral stress, respectively, results in the phosphorylation of eIF2α. To temporarily alleviate cell stress p-eIF2α suppresses global protein translation while access to ATF4 mRNA is increased. Under acute oxidative stress, ATF4 serves a protective role by increasing GSH production resulting in a decrease in ROS and ultimately promoting cell survival. Conversely, when faced with persistent cell stress, ATF4 cooperatively binds with CHOP. Together the two transcription factors act upon genes that increase protein translation that leads to indirect cell death through generation of ROS

Full size image

ATF4 is a basic leucine-zipper (bZIP) class transcription factor that drives the expression of genes involved in multiple metabolic pathways. ATF4 has been shown to enhance metabolic processes such as glycolysis, glutaminolysis, and OXPHOS by regulating a network of genes controlling amino acid intake, mTORC1 activation, and protein translation [60, 61]. To regulate target gene expression ATF4 forms transcription factor complexes by binding DNA and other transcription factors. ATF4 binds to DNA at CREs (cAMP responsive elements) or C/EBP-ATF response elements (CAREs) [62, 63]. In addition to binding DNA, ATF4 forms transcription factor complexes to collectively regulate target gene expression. Alone, AFT4 homodimers are unstable, and stability must be increased to effectively regulate gene expression. Increasing stability is accomplished when ATF4 forms heterodimers with other transcription factors such as FOS/JUN or CHOP [48, 62,63,64]. Together, FOS and JUN regulate genes known to protect cells from apoptosis [65,66,67], while ATF4 and CHOP synergize to increase protein synthesis, in turn accumulating ROS and leading to cell death (Fig. 2) [20]. As indicated above, the role of ATF4 varies greatly and whether ATF4 acts in a protective or deleterious manner is determined by several variables, including its relationship with binding partners.

Antioxidant defense is a well-studied acute role of ATF4. ATF4 regulates the expression of genes involved in protection from oxidative stress such as those involved in the biosynthesis of glutathione (GSH), a key component for reducing endogenous peroxides [21]. For example, endogenous peroxides in Atf4−/− MEF cells are elevated relative to WT levels [21] and Atf4−/− MEFs experience elevated cell death. Treatment of Atf4−/− MEFs with reducing agents sustains cell survival, indicative of intrinsic ROS that promote cell death in the absence of ATF4 [21]. Along these lines, in Hepa1-6 cells, mRNA levels of Cth, a critical gene in the GSH pathway, and Sod2, a gene involved in clearing ROS from the mitochondria, are reduced in the absence of ATF4 leading to increased cell death [68]. Of note, Sod2 was determined to be a gene associated with better tumor control in T cells selected for low mitochondrial membrane potential [69]. In vivo ATF4 is required for protecting cells from oxidative stress. Mice with liver-specific ATF4 deletion exhibit increased cell death and heightened levels of malondialdehyde, a highly reactive compound that serves as a marker for oxidative stress [68]. Together, the studies suggest that ATF4 is critical for cellular protection from oxidative stress.

While ATF4 acutely protects cells from oxidative stress, in the face of chronic stress the induction of ATF4 promotes cell death through accumulation of intracellular ROS [20]. ATF4 binds with CHOP and RNA-seq of genes targeted by these transcription factors suggests that together they direct deleterious protein synthesis and the UPR [20]. In vitro and in vivo studies found that together the two transcription factors induce cell death. Forced expression of CHOP did not reduce MEF viability but forced expression of ATF4 promoted low levels of death. When ATF4 and CHOP were co-expressed in MEFs, apoptosis was increased beyond ATF4 expression alone [20]. In vivo results amplified the in vitro findings as adenovirus-mediated overexpression of ATF4 and CHOP in the liver of mice proved that ATF4/CHOP led to cell death [20]. ERO1α is an oxidoreductase enzyme that catalyzes protein folding in the ER lumen. Both in vivo and in vitro, ATF4- and CHOP-dependent cell autonomous apoptosis was attributed to an increase in protein synthesis that led to ERO1α-mediated generation of ROS that was the direct mechanism to promote cell death [20].

Regarding global tumor immunity, it has been shown that the suppressive nature of myeloid-derived suppressor cells (MDSCs) is catalyzed by the oxidative stress of the TME and alleviation of MDSC-intrinsic ATF4-CHOP response allays their protumorigenic nature [70]. These data indicate that ATF4-CHOP plays a deleterious role in the immune response to promote tumor growth. In multiple studies that established a detrimental role for PERK in tumor control [18, 38], the antitumor efficacy of PERK−/− T cells was attributed to abrogation of the chronic arm of the stress response. The oxidative stress of the TME was found to enhance PERK–ATF4–CHOP in CD8 TILs and led to inhibition of antitumor capability in a CHOP-dependent manner [18]. We demonstrated that the ATF4–CHOP–ERO1α axis is significantly increased in effector T cells relative to memory-like T cells [15, 19, 38]. These data suggest that ATF4 could have a contribution to T cell lineage fate and play a defining role in antitumor immunity. In CD4 T cells, the oxidative stress response activates ATF4 to promote redox homeostasis, proliferation, and cytokine production [61]. These data point to a protective role of ATF4 in CD4 T cells. It is likely that ATF4’s role in the face of acute stress is protective to antitumor immunity through buffering antioxidant capacity. In contrast, chronic activation and expression of ATF4 likely has a deleterious role in antitumor immunity through ATF4’s hyperactivating role in concert with CHOP that drives protein translation that results in cell death through generation of intrinsic cell ROS [20].

Taken together a picture has emerged in T cell biology of a delicate balance for intracellular ROS, likely moderated by the stress response, as a positive regulator of cell growth and signaling contrasted to accumulation of pathological levels of ROS that trigger death. A seminal paper by Sena et al. determined that ROS generated by mitochondria (mitoROS) are required for nuclear factor of activated T cells (NFAT) activation and IL-2 production as well as antigen-specific inflammatory responses in CD4 T cell in vivo [71]. Along these lines, CD8 T cells proliferating in tumor-draining lymph nodes of mice treated with a-PD-L1 therapy expressed elevated mitoROS relative to T cells with diminished proliferation and pharmacological promotion of ROS-activated responses to a-PD-L1 therapy in tumor-bearing mice [72]. We discovered that PD-1 CD8 TILs in mice and patients bearing sarcomas are replete with mitoROS, but mitoROShigh T cells exhibit reduced SRC, diminished IFN-γ production, and signs of early cell death [19]. Together the data illustrate the necessity for mitoROS generation in T cells for successful immunotherapy, but suggest that a constant cycle of ROS detoxification must also take place. GSH is a key mediator of antioxidant capability in T cells necessary for mitoROS detoxification. GSH was shown to be critical to support NFAT activation and subsequent metabolic reprogramming needed for T cell antiviral defense [73]. In successful antitumor T cells enriched for gluconeogenesis, the cycle was critical to produce NADPH to reduce oxidized glutathione (GSSG) back to GSH. Inhibition of gluconeogenesis through multiple substrates diminished NADPH/NADP+ ratios leading to accumulation of GSSG and elevated intracellular ROS [37]. The data show the essential role of GSH to detoxify cell ROS to support antitumor T cells.

ATF4 and amino acid stress in tumor immunity

A common stress presented to immune cells in the TME is deficiency of amino acids and ATF4 is likely central to the stress response of antitumor TILs. Amino acids are precursors for protein and nucleic acid synthesis [74] and the metabolic shift that occurs in response to T cell activation requires robust increases in amino acid intake and production for effector function and development of memory [74, 75]. A seminal paper by Harding et al. identified the essential role of ATF4 in transcribing genes necessary for import of amino acids. Under cell stress induced by ER stress enhancer tunicamycin ATF4−/− MEFs were unable to generate amino acids, glycine transporter 1 (Glyt), and Asparagine Synthetase (ASNS), an enzyme that catalyzes the biosynthesis of asparagine and glutamate from aspartate and glutamine, respectively. Thus, ATF4−/− MEFs were unable to proliferate in the absence of non-essential amino acids (NEAA), indicating an intrinsic defect in NEAA production [21]. In CD4 T cells, oxidizing environments and amino acid deprivation were shown to induce ATF4 expression that subsequently was imperative for activation of mTORC1 and synthesis of amino acids. Based on deficiency of mTORC1 in ATF4−/− CD4 T cells, ATF4 was required to generate glycolysis, OXPHOS, and effector function in CD4 T cells [61]. The intrinsic role of ATF4 in antitumor TILs remains to be elucidated.

GCN2 is the ER stress sensor primed to sense and respond amino acid stress by binding uncharged tRNAs [76, 77]. Upon amino acid starvation, GCN2 phosphorylates eIF2α and enables the increased translation of ATF4. This event serves as the converging point between nutrient stress and ATF4 signaling [61, 78]. A seminal paper by Munn et al. indicated that GCN2 acts as a sensor enabling T cells to detect and react to amino acid deprivation in their environment. Dendritic cell (DC) expression of indoleamine 2,3 dioxygenase (IDO), an enzyme that degrades the essential amino acid tryptophan, induced GCN2-mediated stress sensing and response in T cells. GCN2-deficient antigen-specific CD8 T cells primed in the presence of IDO-producing plasmacytoid DCs (pDCs) exhibited enhanced proliferation relative to WT controls. The effect in WT T cells was rescued by administration of tryptophan or inhibition of IDO, leading to the conclusion that GCN2 senses tryptophan deprivation causing inhibition of T cell proliferation [79]. The findings were later challenged by the demonstration that GCN2−/− T cells failed to proliferate in the presence of IDO or low tryptophan by employing a titration assay to determine the limiting amount of tryptophan required for cell proliferation. Of note, the latter study commented that limiting amounts of amino acids may impact the capacity of antigen presenting cells, thus potentially accounting for low proliferative activity of GCN2−/− T cells. However, in vivo a proliferation defect was also noted in GCN2-deficient T cells [77]. Together the data suggest that GCN2 impacts T cell proliferation and homeostasis.

Regarding tumor immunity, recent works established that GCN2 is required for proliferation, survival, and function of CD8 T cells in murine gliomas. In vitro, GCN2−/− T cells were less activated compared to WT counterparts. In GCN2−/− mice bearing glioblastoma, the number and percentage of CD8 T cells was significantly reduced in tumors, but no significant changes were noted in peripheral lymphoid organs [59]. These data indicate that the TME may be responsible for driving the reduction of CD8 T cells in tumor. Upon examination of the effects of amino acid deprivation on GCN2−/− T cells in vivo, WT T cells survived nutrient deplete conditions better than GCN2−/− T cells. Thus, GCN2 may be required for CD8 T cell survival in the TME, a plausible explanation for the reduction in CD8 T cells in gliomas [59].

GCN2 has also been proven critical for myeloid cells and their suppressive function in tumors. In mice with B16F10 melanomas, GCN2−/− myeloid cells generated reduced tumor volume relative to controls. Analysis suggested that tumors from mice harboring GCN2−/− myeloid cells produced tenfold less immunosuppressive IL-10 [80]. ATF4 is a critical downstream mediator of GCN2 and was shown to drive the immunosuppressive nature of MDSCs in a GCN2-dependent manner. Specifically, ATF4 activation did not occur in myeloid cells devoid of GCN2 and siRNA-mediated knockdown of ATF4 in macrophages in vivo resulted in decreased tumor growth. These data support reports that ATF4-CHOP axis promotes suppression in MDSC populations that enable tumor growth [70], and identify GCN2 as the potential upstream mediator to promote the protumorigenic effects.

Summary

The family of ER kinases that includes PERK, GCN2, HRI, and PKR primed to activate eIF2α and generate ATF4 expression, comprise a hub of stress sensors poised to respond to a variety of environmental insults. PERK and GCN2 have already been implicated in responding to stressors presented by the TME, but the specific acute and chronic responses imparted in pro- and antitumor immune populations remain poorly understood. It is important to note that the family of ER kinases represents a group of druggable targets that could serve to promote the efficacy of standard immune checkpoint blockade (ICB) therapies. We found that PERK inhibition with first-generation inhibitor GSK26064141 [81] generated complete responses in mice bearing sarcomas when combined with ICB therapy [19]. However, toxicity associated with pancreatic islet cell degradation is a concern for applications with first- and second-generation [82] inhibitors that directly target PERK. If PERK proves to tune T cell antitumor metabolism, it is possible that such therapies could be administered on a schedule that alleviates in vivo toxicity. Along these lines, non-toxic therapeutics that affect downstream targets of PERK are in development and could prove useful in the space of immune oncology. Integrated stress response inhibitor (ISRIB) is a small molecule that selectively inhibits p-eIF2α-mediated attenuation of translation [83, 84] downstream of PERK and early reports suggest that ISRIB treatment circumvents pancreatic toxicity [85] associated with targeting PERK. Given that the ER is a hub for cell metabolism and fate, study of master ER stress sensors to refine metabolic efficacy and tune stress relief in tumor immunity holds immense potential to generate novel therapeutic strategies to improve the efficacy of cancer immunotherapies.