Phosphorylation and dephosphorylation of eukaryotic initiation factor 2α (eIF2α) control the initiation of mRNA translation in eukaryotic cells, particularly in response to cellular stress. Early studies on the role of initiation factors in translational control date back to the 1960s. In 1968, Miller and Schweet reported a decrease in reticulocyte ribosomal activity upon salt washing (Miller and Schweet 1968). This first observation led to the isolation of translation initiation factors from rabbit reticulocytes and liver ribosomes (Shafritz et al. 1970). Some years later, it was shown that the eukaryotic translation initiation factor eIF2 consists of three subunits: eIF2α, eIF2β, and eIF2γ (Schreier et al. 1977). In the same year, phosphorylation of eIF2α was shown to impair the ability of eIF2 to initiate translation (Farrell et al. 1977). Two eIF2α kinases were identified: DAI kinase, later renamed double-stranded RNA-dependent protein kinase (PKR), which phosphorylates eIF2α in the presence of dsRNA, and heme-regulated eIF2α kinase (HRI), which phosphorylates eIF2α in the absence of heme (Ernst et al. 1980). In 1987, in vitro studies using purified eIF2 from rabbit reticulocytes or rat liver identified Ser48 and Ser51 as the sites of eIF2α phosphorylation by PKR and HRI kinase (Colthurst et al. 1987). Expression of mutant eIF2α with an Ala substitution at position 51 confirmed that phosphorylation of eIF2α at Ser51 is required for the inhibition of protein synthesis in rabbit reticulocytes (Pathak et al. 1988). Since these early studies, eIF2α phosphorylation on Ser51 has been shown to be ubiquitous in diverse species ranging from yeast to humans indicating that this is a common mechanism used by eukaryotic cells to control mRNA translation.
Structure of eIF2α
Although the structure of human eIF2α is very similar to the yeast eIF2α (Dhaliwal and Hoffman 2003), there are a few major differences that should be noted. Yeast eIF2α lacks Cys69 which in human eIF2α forms a disulfide bridge bond with Cys97 (Dhaliwal and Hoffman 2003). Another major difference between the yeast and human eIF2α structures lies in the visibility of the surface loop in yeast (residues 51–65). In human structure, this loop has not been completely solved. In yeast, it forms part of the helical insert between β3 and β4 of the N-terminal S1 subdomain. The helical insert consists of two single-turn 310 helices (residues 46–50 and 58–61) separated by a short linker containing the Ser51 phospho-regulatory site (Fig. 2b). Ser51 is followed by three arginine residues (Arg52, Arg53, Arg54).
Function of eIF2α in Translational Control
Phosphorylated eIF2α is a direct, competitive inhibitor of a guanine nucleotide exchange factor, eIF2B, and blocks the eIF2B-mediated recycling of eIF2-GDP back to eIF2-GTP (Fig. 3). As a consequence, an increased amount of eIF2-GDP limits the assembly of the ternary complex, thus preventing formation of the 43S PIC and delivery of Met-tRNAi for reinitiating ribosomes (Pakos-Zebrucka et al. 2016).
eIF2α also functions as the key orchestrator of cellular stresses responses in a process called the integrated stress response (ISR). The ISR is an adaptive pathway used to restore cellular homeostasis through a global inhibition of 5′Cap-dependent mRNA translation and the preferential translation of selected mRNAs that contain a series of short upstream open-reading frames (uORFs) in their 5′ untranslated region (5′UTR), particularly ATF4 mRNA. Phosphorylated eIF2α limits the availability of the ternary complex and thus enables ribosomes to bypass the inhibitory uORFs leading to delayed translation reinitiation at the AUG start codon of the ATF4 coding sequence. The resulting induction of ATF4 expression, which is central to eIF2α-dependent stress adaptation and the ISR, transcriptionally activates genes involved in protein folding, amino acid metabolism, redox homeostasis, and cellular defense.
Regulation of eIF2α
Phosphorylation of eIF2α
In mammals, phosphorylation of eIF2α at Ser51 is regulated by four Ser/Thr eIF2α kinases (PERK, GCN2, PKR, and HRI) that are activated by diverse cellular stresses (Fig. 3). These kinases have been extensively reviewed in a review (Donnelly et al. 2013).
PKR-like ER kinase (PERK) is located in the endoplasmic reticulum (ER) membrane where its luminal domain is normally bound by glucose-regulated protein 78 kDa (GRP78). PERK is an important component of the unfolded protein response (UPR), the cellular adaptation program which helps cells to recover from ER stress. It is activated in response to accumulation of unfolded proteins in the ER that can occur as a consequence of protein mutations and perturbations in cellular energy, calcium homeostasis, or redox status. Two models of PERK activation have been proposed: dissociation of GRP78 from PERK or direct binding of unfolded and misfolded proteins to PERK in the ER lumen. These events trigger PERK homodimerization followed by its autophosphorylation. Activated, phospho-PERK phosphorylates eIF2α leading to an overall decrease in protein synthesis, thereby limiting the load of nascent polypeptides arriving to the ER. PERK-deficient cells are more sensitive to conditions that cause ER stress and to ER stress-induced cell death. Furthermore, PERK-deficient mice develop diabetes mellitus and have an impaired bone formation (Donnelly et al. 2013).
General control non-derepressible-2 (GCN2) is composed of a kinase domain and a histidyl-tRNA synthetase (HisRS)-related domain. Limitations in amino acid supply results in an accumulation of deacylated tRNAs that bind to the HisRS domain. Upon tRNA binding, GCN2 undergoes a conformational change that promotes its activation by dimerization and autophosphorylation at Thr882 and Thr887 (Donnelly et al. 2013). Once activated, GCN2 phosphorylates eIF2α at Ser51 resulting in inhibition of translation and the selective induction of genes involved in amino acid biosynthesis. Mice lacking GCN2 are viable, fertile, and display no phenotypic abnormalities; however, they exhibit hypersensitivity to diets deficient in single amino acids such as leucine or tryptophan, displaying increased mortality in prenatal and neonatal mice (Donnelly et al. 2013).
Double-stranded RNA-dependent protein kinase (PKR) is activated by dsRNA during viral infection. PKR is composed of a C-terminal kinase domain and two N-terminal dsRNA binding domains (dsRBD) that regulate its activity. Viral dsRNA molecules bind to PKR through its dsRBDs resulting in PKR dimerization via the kinase domains and autophosphorylation. Fully activated PKR phosphorylates eIF2α leading to inhibition of viral and host protein synthesis. Other stresses such as oxidative and ER stress, growth factor deprivation, cytokine or bacterial infection, ribotoxic stress, stress granules, and heparin have also been shown to stimulate PKR in a dsRNA-independent manner (Donnelly et al. 2013).
Heme-regulated eIF2α kinase (HRI) is expressed in erythroid cells where it couples the translation of globin mRNA with the availability of heme for the production of hemoglobin. HRI consists of two N-terminal heme-binding domains and a kinase insertion domain. Activation of HRI requires dimerization and autophosphorylation of its kinase domain. When heme binds to the kinase insertion domain, it prevents HRI activation by autophosphorylation. HRI can also be activated in a heme-independent manner by other stresses, including arsenite-induced oxidative stress, heat shock, osmotic stress, 26S proteasome inhibition, and nitric oxide (Donnelly et al. 2013). Once HRI is activated, it phosphorylates eIF2α, thus preventing synthesis of more globins than can be used for hemoglobin production. Mice lacking HRI are viable and fertile, with minimal evidence of hematological abnormalities in the absence of stress. However, iron-deficient HRI−/− mice exhibit hypochromic normocytic anemia.
Dephosphorylation of eIF2α
Inhibition of translation through phosphorylated eIF2α is normally transient and is reversed by dephosphorylation of eIF2α. This is mediated by protein phosphatase 1 (PP1), a complex comprised of catalytic subunit (PP1c) and one of two substrate-specific regulatory subunits, growth arrest and DNA damage-inducible protein (GADD34) also known as PPP1R15A, or constitutive repressor of eIF2α phosphorylation (CReP) also known as PPP1R15B (Fig. 3). Both GADD34 and CReP share homology in their C-terminal PP1-binding domains. Cellular levels of GADD34 are normally low under basal conditions, and its expression is induced during conditions of cellular stress downstream of phosphorylated eIF2α. The GADD34-PP1 complex dephosphorylates eIF2α and acts as a negative feedback loop to restore normal protein synthesis following cellular stress. Cells deficient in GADD34 maintain a high level of phospho-eIF2α, and their translational recovery is impeded under conditions of oxidative and ER stress. Ppp1r15a knockout mice embryos are viable but suffer growth retardation and impaired erythropoiesis (Harding et al. 2009). In contrast, CReP is a constitutively expressed factor maintaining low, basal levels of eIF2α phosphorylation under non-stressed conditions (Jousse et al. 2003). CReP stably binds to PP1c to sustain translational homeostasis in unstressed cells by maintaining low levels of eIF2α phosphorylation. A lack of CReP contributes to elevated levels of eIF2α phosphorylation and can protect cells from oxidative, nitrosative, and ER stress (Jousse et al. 2003). Ppp1r15b knockout mice develop well, are fertile, and have an increased tolerance to ER stress-induced tissue damage (Jousse et al. 2003). In contrast to the single knockouts, Ppp1r15a and Ppp1r15b double-knockout mice exhibit early embryonic lethality indicating the importance of eIF2α dephosphorylation for viability (Harding et al. 2009).
Physiological Significance of eIF2α Signaling
Phosphorylation of eIF2α is a master regulator of cellular adaptation to various stress conditions. Depending on the nature, duration, and intensity of the stress, phosphorylated eIF2α can either promote cell survival or induce cell death (Pakos-Zebrucka et al. 2016). There are several physiological conditions that alter cell homeostasis which induce eIF2α phosphorylation, for example, high demand for protein synthesis on secretory cells including pancreatic β cells, hepatocytes, plasma cells, and barrier epithelial cells. Dysregulation of eIF2α signaling, due to reduced phosphorylation resulting from chronic inactivation of one of the four upstream eIF2α kinases, has important pathologic consequences linked to disease, e.g., inflammation and diabetes. In humans, homozygous loss-of-function mutations in EIF2AK3 (the gene which encodes PERK) are associated with Wolcott-Rallison syndrome characterized by early infancy type I diabetes (Donnelly et al. 2013). Genetic studies in mice support a model whereby an inability to phosphorylate eIF2α contributes to diabetes due to the loss of pancreatic β-cells (accompanied by lower insulin levels) and impaired induction of gluconeogenic enzymes in the liver (Scheuner et al. 2001).
Persistent phosphorylation of eIF2α is a major consequence of the accumulation of unfolded and misfolded proteins in cells, which is a distinctive feature of many neurodegenerative diseases. Patient brain tissue from different neurodegenerative diseases exhibits elevated phosphorylation of eIF2α (Bellato and Hajj 2016). These observations are supported by genetic models of neurodegenerative diseases which also have persistent phosphorylation of eIF2α (Bellato and Hajj 2016). Moreover, mouse models with deletion of either PERK or GCN2 display decreased eIF2α phosphorylation which is associated with the absence of memory defects in an AD mouse model, leading to an improvement of the disease symptoms (Ma et al. 2013). Therefore, human and animal studies support the importance of translation regulation in helping neurons to cope with stress (Bellato and Hajj 2016).
Prolonged phosphorylation of eIF2α also supports tumor progression in a harsh tumor microenvironment that is characterized by glucose and oxygen deprivation as well as nutrient limitation. Activation of the PERK and persistent eIF2α phosphorylation contribute to tumor progression, including angiogenesis, tumor growth, and survival under hypoxic stress (Devisscher et al. 2016). Cells that are unable to phosphorylate eIF2α in response to oxygen and glucose deprivation are found to be hypersensitive to hypoxic stress (Koumenis et al. 2002). It is also worth noting that oncogene activation can induce eIF2 phosphorylation. In this context, increased eIF2α phosphorylation has been shown to be associated with Myc-transformed human lymphomas (Hart et al. 2012).
eIF2α displays a highly conserved structural and functional homology from yeast to humans. It is a cytoplasmic ribosome-associated protein that participates in translational initiation. In mammals eIF2α is phosphorylated by four eIF2α kinases, PERK, PKR, GCN2, and HRI, and dephosphorylated by two phosphatases: GADD34-PP1 and CReP-PP1. Various forms of stress activate eIF2α kinases which converge at eIF2α phosphorylation resulting in a global inhibition of protein synthesis along with enhanced translation of selected mRNAs such as ATF4 to restore cellular homeostasis (Pakos-Zebrucka et al. 2016). In addition, as a target of PERK, eIF2α is also an important component of the UPR, a signaling pathway activated in response to ER stress. A reduced level of phosphorylated eIF2α contributes to diabetes, while persistent eIF2α phosphorylation is observed in neurodegenerative diseases and in various types of tumors. A better understanding of the biological significance of modulating eIF2α phosphorylation will allow us determine whether a pharmacological intervention would have any therapeutic value.
The work in our group is funded by Breast Cancer Campaign grant (2010NovPR13), Health Research Board (grant number HRA-POR-2014-643), Belgium Grant (IAP 7/32), a Science Foundation Ireland (SFI) grant co-funded under the European Regional Development Fund (grant Number 13/RC/2073), and EU H2020 MSCA ITN-675448 (TRAINERS). K.M. is funded by an Irish Research Council Fellowship (grant number GOIPD/2014/53).
- Colthurst DR, Campbell DG, Proud CG. Structure and regulation of eukaryotic initiation factor eIF-2. Sequence of the site in the alpha subunit phosphorylated by the haem-controlled repressor and by the double-stranded RNA-activated inhibitor. Eur J Biochem/FEBS. 1987;166:357–63.CrossRefGoogle Scholar
- Dhaliwal S, Hoffman DW. The crystal structure of the N-terminal region of the alpha subunit of translation initiation factor 2 (eIF2alpha) from Saccharomyces cerevisiae provides a view of the loop containing serine 51, the target of the eIF2alpha-specific kinases. J Mol Biol. 2003;334:187–95.PubMedCrossRefGoogle Scholar
- Ernst V, Levin DH, Leroux A, London IM. Site-specific phosphorylation of the alpha subunit of eukaryotic initiation factor eIF-2 by the heme-regulated and double-stranded RNA-activated eIF-2 alpha kinases from rabbit reticulocyte lysates. Proc Natl Acad Sci USA. 1980;77:1286–90.PubMedPubMedCentralCrossRefGoogle Scholar
- Hart LS, Cunningham JT, Datta T, Dey S, Tameire F, Lehman SL, Qiu B, Zhang H, Cerniglia G, Bi M, Li Y, Gao Y, Liu H, Li C, Maity A, Thomas-Tikhonenko A, Perl AE, Koong A, Fuchs SY, Diehl JA, Mills IG, Ruggero D, Koumenis C. ER stress-mediated autophagy promotes Myc-dependent transformation and tumor growth. J Clin Investig. 2012;122:4621–34.PubMedPubMedCentralCrossRefGoogle Scholar
- Koumenis C, Naczki C, Koritzinsky M, Rastani S, Diehl A, Sonenberg N, Koromilas A, Wouters BG. Regulation of protein synthesis by hypoxia via activation of the endoplasmic reticulum kinase PERK and phosphorylation of the translation initiation factor eIF2alpha. Mol Cell Biol. 2002;22:7405–16.PubMedPubMedCentralCrossRefGoogle Scholar
- Pakos-Zebrucka K, Koryga I, Mnich K, Ljuic M, Samali A, Gorman AM. The integrated stress response. EMBO Rep. 2016;17:1374–95. EMBO Reports.Google Scholar