Phosphorylation of Translation Initiation Factor eIF2α at Ser51 Depends on Site- and Context-specific Information

Abstract

Protein kinases phosphorylate specific amino acid residues of substrate proteins and regulate many cellular processes. Specificity for phosphorylation depends on the accessibility of these residues, and more importantly, kinases have preferences for certain residues flanking the phospho-acceptor site. Translation initiation factor 2α (eIF2α) kinase phosphorylates Serine51 (Ser51) of eIF2α and downregulates cellular protein synthesis. Structural information on eIF2α reveals that Ser51 is located within a flexible loop, referred to as the Ser51 loop. Recently, we have shown that conformational change of the Ser51 loop increases the accessibility of Ser51 to the kinase active site for phosphorylation. Here, we show that the specificity of Ser51 phosphorylation depends largely on its relative position in the Ser51 loop and minimally on the flanking residues.

INTRODUCTION

Protein kinases phosphorylate protein substrates and regulate a myriad of cellular processes. Each phosphorylation reaction is highly specific, with each individual kinase recognizing and phosphorylating a restricted set of substrates. Substrate specificity is determined by multiple factors, including the factors determining the intracellular co-localization of kinase-substrate and the specificity determining elements of kinase, substrate, or both(1). Classic examples of kinases showing the localized specificity are G-protein coupled receptor kinases (GRKs). Out of six families, members of GRK 1–3 families are mostly cytosolic, whereas members of GRK 4–6 families are trans-membrane proteins (2). The elements that determine specificity include a small subset of amino acids present at the local, proximal or distal region of the phosphorylation site(1) and/or at the kinase active site(3). For examples, several members of the AGC (cAMP-dependent, cGMP-dependent and protein kinase C) family of protein kinases preferentially recognize and phosphorylate Arg-X-Arg-X-X-Ser/Thr motifs (RXRXXS/T, X is any residue and S/T the phosphorylation site)(4), whereas members of the eIF2α (eukaryotic translation initiation factor 2α) family of protein kinases recognize a Lys-Gly-Tyr-Ile-Asp (KGYID) motif, which is remote from the phosphorylation site(5).

In mammals, four eIF2α kinases have been identified: PKR (protein kinase R), Gcn2 (general control non-depressible 2), HRI (heme-regulated inhibitor) and PERK (PRK-like endoplasmic reticulum kinase)(6). In normal conditions, eIF2α kinases remain inactive, whereas each kinase is uniquely and independently activated under a particular cellular condition as follows. Production of double stranded RNA (dsRNA) during viral infections activates the PKR, whereas acute heme-deficient states activate the HRI. An excessive amount of uncharged tRNA molecules during nutrient starvation activates the Gcn2, whereas an over-accumulation of unfolded proteins in the lumen of endoplasmic reticulum (ER) activates the PERK. Once activated, eIF2α kinases specifically phosphorylate eIF2α at Ser51. Phosphorylated eIF2α inhibits the function of a guanine nucleotide exchange factor eIF2B, thus down-regulates global protein synthesis (7).

Structural features of eIF2α and PKR have provided important insights into the mechanisms by which Ser51 is phosphorylated by eIF2α kinases. The N-terminal half of yeast eIF2α (residues 1–175) has been shown to fold into an oligonucleotide (OB)-fold domain and a helical domain (8). The OB-fold domain is composed of five β strands in which Ser51 resides within a flexible loop (residues Leu47 to Gly65, hereafter referred to as the Ser51 loop) connecting the strands β3 and β4 (Fig 1A and Fig 1C). Residues making up the Ser51 loop have been shown to form two random coils and two short 3₁₀-helices (residues Ser48-Glu59-Leu50 and Ser58-Ile59-Gln60) with Ser51 residing at the C-terminus of one 3₁₀-helix (Fig 1A and Fig 1C). Like yeast eIF2α, the N-terminal half of human eIF2α comprises of an OB-fold domain and a helical domain, whereas the C-terminal half adopts an αβ domain architecture similar to the alpha subunit of eukaryotic translation elongation factor 2(9). In several human eIF2α proteins, the Ser51 loops are structurally variable with the side chain of Ser51 buried inside the protein.

Figure 1. Analysis of sequence and structure of PKR and eIF2α proteins.

(A) Alignment of eIF2α protein sequences. The eIF2α protein sequences from yeast (UniProtKB-P20459), fruit fly (UniProtKB-P41374), human (UniProtKB-P05198) and zebra fish (UniProtKB-Q919E9) were aligned using Clustal W. Alignment of the N-terminal half of sequences with an OB-fold domain is shown. The conserved residues are indicated by the symbol “*”, and the secondary structural elements are indicated on the top of the alignment.

(B) The residue Ser51 is 19.6 Å away from the PKR catalytic base D414. Free eIF2α (PDB code = 1Q46, green ribbon) is super-positioned on the eIF2α bound to PKR (PDB code 2A1A, grey ribbon). The ADP molecule bound to the kinase active site is colored blue, whereas the catalytic base D414 in pink and the Ser51 in red.

(C) A cluster of four hydrophobic residues on the Ser-51 loop. The conformation of the Ser51 loop with indicated residues Ser51 (red), L47 (orange), L50 (blue), I58 (grey) and I62 (green) are indicated.

The co-crystal structure of eIF2α bound to the PKR kinase domain (PKR-KD) shows that PKR-KD folds like a typical protein kinase domain with a smaller N-terminal lobe (N-lobe) and larger C-terminal lobe (C-lobe) (10) (Fig 1B). The C-lobe of PKR-KD makes several inter-molecular contacts with residues of the eIF2α OB-fold, including residues Met44, Lys79, Tyr32 Tyr81 and Asp83. In the PKR-eIF2α complex, the Ser51 position has not been accurately assigned because 11 residues connecting Ser49 to Lys60 are substantially disordered. Superposition of a free eIF2α on the eIF2α bound to PKR-KD shows that Ser51 is distal to the catalytic base Asp414 (D414), positioned 19.6 Å away from the carboxylate oxygen of D414 (Fig 1B). The distal Ser51 site and the variable Ser51 loop structures suggest that Ser51 must re-orient to access the active site during the phosphate transfer reaction. Supporting the notion, we have shown that the specificity in the eIF2α kinase-substrate recognition and Ser51 phosphorylation was abolished when the hydrophobic pocket surrounding the Ser51 residue was disrupted(11). Previously, we have shown that phosphorylation of Ser51 was impaired when several remote residues, including residues Met44 and Asp83, were individually mutated, establishing the importance of remote residues in eIF2α kinase-substrate recognition and Ser51 phosphorylation events (5). Taken together, we propose a bipartite substrate recognition model for Ser51 phosphorylation, where binding of PKR to the OB-fold segment of eIF2α likely induces a conformational change in the phosphorylation site sequences, resulting in optimal positioning of Ser51 to the kinase active site(11). However, the direct experimental evidence for induced conformational change of the phosphorylation site residues needs to be determined.

In this article, we provide additional evidence for conformational plasticity of the Ser51 loop during its phosphorylation events. By mutational analysis, we show that the local phosphorylation site residues do not play a major role in Ser51 phosphorylation. These results suggest that the specificity of Ser51 phosphorylation depends largely on the context of the target residue and minimally on what flanks it. We also engineered the Ser51 phosphorylation site by inserting an alanine in between residues Leu46 and Leu47. In this engineered eIF2α protein, we show that the residue at position 51, with respect to the parent protein, is a major phosphate accessible residue and a translational regulator in yeast Saccharomyces cerevisiae containing the sole eIF2α kinase Gcn2.

RESULTS

Specificity of Ser51 phosphorylation is not based on its flanking residues

Three upstream and downstream residues of the Ser51 phosphorylation site are Ser-Glu-Leu (SEL_48–50) and Arg-Arg-Arg (RRR_52–53), respectively (Fig 1A). We have reported that a single mutation at residues E49, L50 and R52 did not affect Ser51 phosphorylation in vivo or in vitro(5). To test the combined mutational effect of six neighboring residues on Ser51 phosphorylation, we mutated these immediate residues (SEL_48–50 and RRR_52–54) to alanine in a single protein, generating a glutathione S-transferase (GST)-fused eIF2α-SEL_48–50AAA-S₅₁-RRR_52–54AAA mutant (here referred to as GST-eIF2α-AAA-S₅₁-AAA). We then substituted Ser51 of the GST-eIF2α-AAA-S₅₁-AAA protein with alanine, generating a GST-eIF2α-SEL_48–50AAA-A₅₁-RRR_52–54AAA mutant (here referred to as GST-eIF2α-AAA-A₅₁-AAA). Both GST-eIF2α-AAA-S₅₁-AAA and GST-eIF2α-AAA-A₅₁-AAA proteins were expressed in bacteria and the recombinant proteins were purified. Purified proteins were then used as substrates for in vitro kinase assays in the presence of γ−³³P-ATP and eIF2α kinase PKR (Fig 2).

Figure 2. Phosphorylation of Ser51 depends largely on its relative position in the Ser51 loop and minimally on flanking residues.

Indicated GST-fused eIF2α proteins (residues 1–180, GST-eIF2α) were mixed with the purified PKR in a kinase buffer containing γ−³³P-ATP for 20 minutes. The reaction mixture was then quenched by 6X-SDS dye and loaded on an SDS-PAGE to separate PKR and eIF2α proteins. The gel was stained, dried and auto-radiographed to detect phosphate (³³P) incorporation in PKR and eIF2α proteins.

As shown in the Fig 2, PKR was able to incorporate phosphate (³³P) in the wild-type eIF2α protein (referred to as WT or SEL-S₅₁-RRR, eIF2α-P, lane 1), but not in the GST-eIF2α-S₅₁A mutant protein (referred to as S₅₁A or SEL-S₅₁A-RRR, eIF2α-P, lane 2). These results are consistent with our previous report that PKR phosphorylates specifically at Ser51(12). Interestingly, we observed that PKR robustly phosphorylated the recombinant GST-eIF2α-AAA-S₅₁-AAA protein, whereas very moderately the GST-eIF2α-AAA-A₅₁-AAA mutant protein (Fig 2, eIF2α-P, compare lanes 3 and 4). The modest phosphorylation of the GST-eIF2α-AAA-A₅₁-AAA protein was likely due to nonspecific Ser57 phosphorylation within the Ser51 loop. Collectively, these results suggest that the specificity of Ser51 phosphorylation depends on the target residue, but not based on specific flanking residues.

Protein kinase Cα (PKCα) can phosphorylate Ser51 of an eIF2α mutant

To test further the flexibility of Ser51 loop and its site-specific phosphorylation, we replaced the residues SEL_48–50 by Arg-Ala-Ala (RAA), thus generating a GST-eIF2α-SEL_48–50RAA mutant (referred to as RAA-S₅₁-RRR). We also substituted the Ser51 of GST-eIF2α-RAA-S₅₁-RRR mutant with alanine, creating another GST-eIF2α- RAA-S₅₁A-RRR mutant. We observed that PKR was able to efficiently phosphorylate Ser51 of GST-eIF2α-RAA-S₅₁-RRR mutant protein with almost same efficiently as wild-type protein (Fig 3A, compare lanes 1 and 3). These data further confirm that the specificity of Ser51 phosphorylation is not based on specific flanking residues. Interestingly, we observed that human protein kinase Cα (PKCα), which is known to phosphorylate protein substrates containing an RXXS/T motif (X is any residue and S/T is the phosphorylation site)(4), was able to phosphorylate GST-eIF2α-RAA-S₅₁-RRR protein (Fig 3A, eIF2α-P, lane 7), but was unable to phosphorylate the wild-type eIF2α protein (Fig 3A, eIF2α-P, lane 5). Substantial reduction in phosphorylation of the GST-eIF2α-RAA-S₅₁A-RRR mutant protein (Fig 3A, eIF2α-P, lane 8) suggested that the majority of phosphorylation occurred at the residue Ser51 (Fig 1A). A low level of phosphorylation at the GST-eIF2α-SEL_48–50RAA-A₅₁ mutant protein suggest that PKCα likely phosphorylated the residue Ser57 (see Fig 1A) within the Ser51 loop

Figure 3. PKCα phosphorylates Ser51 an eIF2α mutant protein.

(A) Analysis of eIF2α phosphorylation by PKR and PKCα. Indicated GST-eIF2α proteins (residues 1–180) were mixed with PKR in a kinase buffer containing γ−³³P-ATP as described in the Fig 2. The incorporations of phosphate in protein kinases PKR and PKCα are indicated as PK~P.

(B) Analysis of heat-denatured eIF2α phosphorylation by PKR. Heat-denatured recombinant eIF2α protein was mixed with PKR in a kinase buffer containing γ−³³P-ATP as described in the Fig 2. Incorporations of phosphate in PKR and eIF2α proteins are indicated as PKR~P and eIF2α~P.

(C) Analysis of yeast growth. The yeast strain (eIF2αΔ) containing the indicated eIF2α alleles were serially diluted, spotted and grown at 30°C for 48 hours. Whole cell extracts were prepared from the indicated yeast strains and subjected to Western blot analysis using antibodies against phospho-Ser51, eIF2α and Gcd6 (house keeping protein).

We also heat-denatured the recombinant eIF2α protein at 80°C for 5 minutes and then mixed with the wild-type (WT) PKR or catalytically inactive K296R mutant protein in a kinase buffer. The wild-type PKR, like its catalytically inactive K296R mutant, was unable to phosphorylate the heat denatured GST-eIF2α-RAA-S₅₁A-RRR protein (Fig 3B, lanes 1 and 2). These data suggest that an RAAS motif in the context of the folded protein was important for phosphorylation by PKR. Collectively, our data demonstrate that the parent protein structure, but not the nearest neighbors of phosphorylation site, plays a critical role in substrate selection for Ser51 phosphorylation by PKR.

To test whether eIF2α-AAA-S₅₁-RRR and eIF2α-RAA-S₅₁-RRR proteins were substrates of Gcn2 or PKCα-like kinase in yeast cells, we expressed those proteins in an eIF2αΔ deletion yeast strain along with wild-type eIF2α and its phosphorylation site mutants eIF2α-S₅₁A and eIF2α-RAA-S₅₁A-RRR (Materials and Methods). We observed that eIF2α-AAA-S₅₁-RRR, eIF2α_-RAA-S₅₁-RRR and eIF2α-RAA-S₅₁A-RAA proteins were expressed like wild type protein (Fig 3C, Westerns, lanes 1–5), and complemented the essential function of eIF2α and allowed yeast cells to grow on the normal synthetic dextrose medium (Fig 3C). Western blot analysis showed that eIF2α protein was phosphorylated at Ser51 in wild type cells (Fig 3C, lane 1). However, we were unable to detect any phosphorylation at Ser51 in yeast cells expressing eIF2α- AAA-S₅₁-RRR and eIF2α-RAA-S₅₁-RRR proteins (Fig 3C, Westerns, lanes 3 and 4). We reason that either yeast cells lack the specific PKCα-like kinase or the phospho-Ser51-antibody is specific to the wild type protein.

Identification of site-specific phosphorylation sites in eIF2α protein

The crystal structure of eIF2α shows that the side chains of Leu47, Leu50, Ile58 and Ile62 form a hydrophobic network, consequently burying the adjacent Ser51 inside the protein (Fig 1C). Thus, it is possible that the SEL_48–50 motif unwinds and elongates during the kinase-recognition, thereby exposing Ser51 to the protein surface and consequently to the kinase catalytic base. With this notion, we hypothesized that position 50 would be accessible for phosphorylation if a single amino acid residue was inserted to increase the length of the linker preceding the SEL-3₁₀-helix and if the position 50 occupied a phospho-acceptor residue. To directly test this hypothesis, we engineered the eIF2α gene by inserting a GCA codon encoding an alanine in between codons of residues L₄₆ and L₄₇. The engineered eIF2α allele was referred to as the eIF2α^Ala, in which the new position for the residues Leu50 and Ser51 became 51 and 52, respectively. We further mutated the positions 51 and/or 52 of eIF2α^Ala mutant to serine and/or alanine, generating five new mutants eIF2α^Ala-S₅₂A (only the 52 position occupied a non-phosphorylatable alanine), eIF2α^Ala-S₅₁S₅₂ (both 51 and 52 positions occupied phospho-acceptor serine), eIF2α^Ala-A₅₁S₅₂ (51 position with a non-phosphorylatable alanine and 52 position with a phospho-acceptor serine), eIF2α^Ala-S₅₁A₅₂ (51 position with a phospho-acceptor serine and 52 position with a non-phosphorylatable alanine) and eIF2α^Ala-A₅₁A₅₂ (both positions occupied non-phosphorylatable alanine) (Fig 4A). Then, we expressed these mutants in bacteria and purified the recombinant proteins for in vitro kinase assays using either PKR (see Fig 4B) or Gcn2 (see Fig 4C).

Figure 4. PKR and Gcn2 phosphorylate serine at positions 51 or 52 of eIF2α^Ala protein.

(A) An alanine insertion before the Ser51 loop. (Left panel) Protein sequence of the Ser51 loop in wild type (WT) eIF2α is shown on the top. Protein sequences of the Ser51 loop of various eIF2α^Ala mutant proteins are shown below. (Right panel) The insertion of alanine in between the stand β3 and 3₁₀-helix is indicated by an arrow.

(B) (C) & (D) Analysis of eIF2α phosphorylation by protein kinases PKR and Gcn2. Indicated GST-eIF2α proteins (residues 1–180) were mixed with PKR or Gcn2 in a kinase buffer containing γ−³³P-ATP as described in the Fig 2.

As shown in the Fig 4B, PKR was able to phosphorylate eIF2α^Ala protein with less efficiently than did to wild-type eIF2α protein (eIF2α-P, compare lanes 1 and 3). No phosphorylation was observed when Ser51 of eIF2α or Ser52 of eIF2α^Ala was mutated to alanine (Fig 4B, eIF2α-P compare lanes 2 and 4). These data suggested that the residue position 52 of eIF2α might be phosphorylated if occupied a phospho-acceptor residue. We also observed that PKR was able to phosphorylate eIF2α^Ala-S₅₁S₅₂, eIF2α^Ala-A₅₁S₅₂, eIF2α^Ala-S₅₁A₅₂ proteins (eIF2α-P lanes 2–4), albeit less efficiently than wild-type eIF2α protein (Fig 4C, eIF2α-P, lane 1). But, PKR was unable to phosphorylate the eIF2α^Ala-A₅₁A₅₂ mutant protein (Fig 4C, eIF2α-P, lane 5), suggesting that PKR phosphorylated residues occupying the positions 51 and 52. Like PKR, purified Gcn2 phosphorylated eIF2α^Ala-S₅₁S₅₂, eIF2α^Ala-A₅₁S₅₂, eIF2α^Ala-S₅₁A_52, but not eIF2α^Ala-A₅₁A₅₂ protein (Fig 4D, eIF2α-P). To be noted that Gcn2 phosphorylated all mutant versions of eIF2α^Ala with same efficiency in contrast to PKR (compare eIF2α-P in Fig 4C and Fig 4D). It is not still fully clear to us why PKR and GCN2 behaved differently in phosphorylating the eIF2α^Ala protein. However, these results suggest that the SEL_48–50 motif of eIF2α^Ala protein unwinds and elongates, resulting in phosphorylation of both positions 51 and 52 by PKR or Gcn2.

The above results indicated that residues at positions 51 and 52 (i.e., residues Ser51 and Arg52) of the wild-type eIF2α protein should reach at the catalytic center of the kinase during the phospho-transfer reaction. To test that possibility, we performed in vitro kinase assays with eIF2α mutant proteins containing a serine residue at the position 50 or 52. Consistent with our previous report(11), we observed that PKR efficiently phosphorylated the eIF2α-L₅₀S mutant protein (Fig 5, lane 3), suggesting that phosphorylation might occur at Ser51 as well as at Ser50 created by mutating the Leu50. Complete absence of phosphorylation in eIF2α-L₅₀S,S₅₁A mutant protein (Fig 5, lane 4) suggested that serine at position 50 is not a phospho-acceptor residue. Then, we made an eIF2α mutant in which Arg52 was replaced with a serine, generating an eIF2α-R₅₂S mutant. We also substituted the residue Ser51 of eIF2α-R₅₂S mutant with alanine, generating an eIF2α-S₅₁A,R₅₂S, mutant. We found that PKR efficiently phosphorylated the recombinant eIF2α-R₅₂S protein (Fig 5, lane 7), suggesting that phosphorylation might occur at Ser51 as well as at Ser52 created by mutating the residue Arg52. Interestingly, we also found that PKR phosphorylated the eIF2α-S₅₁A,R₅₂S, mutant protein, albeit at a low efficiency (Fig 5, lane 8). The difference in the auto-phosphorylation efficiencies in PKR molecules was due to varying activities of PKR in reaction buffers used at different times; however, the net result of eIF2α phosphorylation suggested that serine at the position 52 was a phospho-acceptor residue. Taken together, it appears that the residue at the position 52, but not at the position 50, of the native eIF2α protein is oriented very close to the catalytic site of PKR and could be partially phosphorylated if occupied a phospho-acceptor residue.

Figure 5.

PKR can phosphorylate serine at the position 52 of eIF2α. Indicated GST-eIF2α proteins (residues 1–180) were mixed with PKR in a kinase buffer containing γ−³³P-ATP as described in the Fig 2.

Phosphorylation of position 51 of eIF2α^Ala regulates GCN4 mRNA translation

We examined the gene-specific translational control of the yeast GCN4 mRNA by phosphorylation of engineered proteins eIF2α^Ala, eIF2α^Ala-S₅₁S₅₂, eIF2α^Ala-A₅₁S₅₂, eIF2α^Ala-S₅₁A₅₂ and eIF2α^Ala-A₅₁A₅₂. The eIF2α^Ala and its derivatives were introduced in the yeast strain H1643 (GCN2) and its isogenic strain H1945 (gcn2Δ). The resulting strains were then tested for their growth on a synthetic dextrose (SD) medium and an SD medium containing 3-AT (3-amino-triazole), an inhibitor of histidine biosynthesis(13). Yeast cells expressing wild type eIF2α and its derivatives eIF2α-S₅₁A, eIF2α^Ala-S₅₁S₅₂, eIF2α^Ala-A₅₁S₅₂, eIF2α^Ala-S₅₁A₅₂ and eIF2α^Ala-A₅₁A₅₂ grew normally on the SD medium (Fig 6A, SD medium), suggesting that each one produced functional protein. As expected, yeast cells grew on the 3-AT medium when expressing both Gcn2 and eIF2α proteins (Fig 6A, left two panels, row 1), but did not grow when expressing eIF2α-S₅₁A mutant protein (Fig 6A, left two panels, row 2) or lacking Gcn2 protein (Fig 6A, right two panels, rows 1 and 2). A model illustrates the control cascade (Fig 6B)(5). Amino acid starvation conditions using 3-AT activate the Gcn2 kinase that phosphorylates eIF2α at Ser51. Phosphorylated eIF2α on Ser51 forms a stable eIF2-GDP-eIF2B (GTP-exchange factor) complex and prevents GDP-GTP exchange reaction, resulting in inhibition of total cellular protein synthesis(7). Accompanying this global translational regulation, phosphorylated eIF2α paradoxically activates the translation of Gcn4 transcription factor that, in turn, activates expressions of several amino acid biosynthesis enzyme genes, resulting in a 3-AT resistant phenotype(14).

Figure 6. Phosphorylation at the position 51 regulates gene specific translational control.

(A) Yeast growth analysis. Yeast strains H1643 (GCN2) or its isogenic gcn2Δ strain H1925 (gcn2Δ) was transformed with a low-copy number LEU2 plasmid expressing the indicated eIF2α or eIF2α^Αλα allele. Transformants were replica-printed to a 5-fluoroorotic (5-FOA) medium in order to evict the URA3 plasmid carrying the wild-type eIF2α gene. The resulting strains were then serially diluted and spotted on SD and 3-AT media and then grown at 30°C for 48 hours.

(B) Scheme of Gcn2 mediated translational regulation. The flow chart shows that protein kinase Gcn2 phosphorylates eIF2α (eIF2α~P) that inhibits eIF2B, leading to 3-AT resistant phenotype (3-AT^r).

Interestingly, we observed that yeast cells expressing the eIF2α^Ala-S₅₁A₅₂ protein grew on the 3-AT medium in the presence of chromosomal GCN2 gene (Fig 6A, left two panels, row 5), but did not grow when chromosomal GCN2 gene was deleted (Fig 6A, right two panels, row 5) or the Ser51 phosphorylation site in eIF2α protein was mutated (Fig 6A, left two panels, row 6). Although the growth phenotype was not as comparable as wild type eIF2α (Fig 6A, compare rows 1 and 5), these data suggested that growth was dependent on Gcn2 kinase function. In contrast, yeast cells expressing the eIF2α^Ala-S₅₁S₅₂ protein did not grow on the 3-AT medium in the presence or absence of chromosomal GCN2 gene (Fig 6, row 3). The sensitivity to 3-AT was likely due to the fact that phosphorylated eIF2α^Ala-S₅₁S₅₂ protein was unable to inhibit the eIF2B function, similar to what we reported previously for eIF2α-E₄₉R mutant(5). Collectively, these results provide genetic evidence for the position-specific phosphorylation of eIF2α and its impact on GCN4 mRNA translational control.

DISCUSSIONS

Conformational plasticity of eIF2α phosphorylation site sequences

We show that the specificity of Ser51 phosphorylation depends largely on the context of target residue and barely on residues that flank it (Fig 2). Then, we show that an alanine insertion in between residues Leu46 and Leu47 leads to phosphorylation of the serine residue at the position 50 (the relative position is 51 in the context of engineered eIF2α^Ala protein) (Fig 4). Consistent with these results, we have observed that the recombinant eIF2α-S₅₁A,R₅₂S, protein in which the position 52 occupies a serine can be phosphorylated by PKR (Fig 4A), whereas recombinant eIF2α-L₅₀S,S₅₁A protein is not a substrate of PKR (Fig 5). Previously, it has been shown that PKR recognizes a large contiguous surface on the eIF2α protein(5) and can phosphorylate Ser/Thr/Tyr at the position 51(15). Collectively, these results suggest that a specific linker length connecting the strand β3 and the SEL_48–50-3₁₀-helix (Fig 1C) is required to project the phospho-acceptor Ser/Thr/Tyr site to the active site of kinase. Together, these observations demonstrate that allosteric binding of kinase at the remote surface likely induces a conformational change of the phosphorylation site to project Ser51 to the catalytic center of eIF2α kinase.

Site specificity in eIF2α phosphorylation

An intriguing question is why did cells evolve such control mechanism for Ser51 phosphorylation? A straightforward answer is that phosphorylation of eIF2α at Ser51 is a conserved regulatory mechanism of translation initiation to down-regulate cellular protein synthesis, which is an adaptive response to various physiological changes including cellular stresses(16), pathogen infections(17), synaptic plasticity and long-term memory (18). On the other hand, excessive phosphorylation of Ser51 of eIF2α is toxic in yeast(19) and maybe, by extension, deleterious to other organisms. Therefore, it makes sense to assume that both the phosphorylation of eIF2α by its kinase at sub-lethal level and the subsequent inhibition of eIF2B by phospho-eIF2 are controlled at multiple levels. For examples, the eIF2α kinase domains are auto-regulated by various cis-acting regulatory domains, such as PKR kinase domain by two double stranded RNA binding domains, Gcn2 KD by HisRS like domain, HRI KD by heme-regulated domains, PERK KD by IRE1 like element(6). In addition, we reported previously that Ser51 phosphorylation by eIF2α kinases was achieved by a bipartite substrate recognition process(5,11). Here, we describe that conformational changes expose Ser51 for phosphorylation. Overall, such cascade control mechanisms must have evolved for controlling gene regulation that is essential for accommodating cellular stresses.

MATERIALS AND METHODS

Yeast strains

Standard methods were used for culturing and transforming yeast cells. A yeast strain H1643 (MATa ura3–52 leu2–3 lu2–112 trp1-Δ63 sui2Δ p[SUI2,URA3]<GCN4-LacZ,TRP1>TRP1) or its isogenic gcn2Δ strain H1925 (5) was transformed with LEU2 plasmids containing the eIF2α mutants. Transformants were then transferred to a medium containing 5-fluoroorotic acid (5-FOA) to evict the URA3 plasmid containing the wild type eIF2α.

Purification of PKR, Gcn2 and eIF2α proteins

The yeast strain J223 (MATα ura3-52 leu2-3 leu2-112 gcn2, SUI2-S51A) was transformed with the plasmids pC2436, pC1685 and pB2828 and to express the Flag-Gcn2, Flag-PKR and Flag-PKR-296R proteins, respectively, from a galactose inducible GAL1-CYC1 promoter. Transformants were grown in a complete medium containing 10% galactose and 2% raffinose for 24 hours and harvested. The recombinant Flag-Gcn2 and Flag-PKR proteins were purified using anti-FLAG M2 agarose (Sigma). Wild type GST-eIF2α and its mutant proteins were expressed in BL21 (DE3) cells and purified by glutathione cross-linked agarose column (Themo Scientific). The standard protocols were followed to purify the recombinant PKR, PKR-K296R, Gcn2 and eIF2α proteins as described earlier(19).

Western blot analysis

Yeast cells were grown overnight in a synthetic complete (SC) medium without histidine (SC-His), diluted to a fresh SC-His medium to OD₆₀₀ ~ 0.1 and grown further to OD₆₀₀ ~ 0.6. Then, 3-AT (30 mM) was added to the medium. Cells were grown for another one hour and harvested. Whole cell extracts (WCEs) were prepared from yeast cells as described earlier(5). The WCEs were separated by SDS-PAGE and subjected to Western blot analysis by Ser51-phospho-specific, anti-GCD6 and polyclonal antibodies of eIF2α protein. All antibodies are obtained from Dr. Dever, National Institutes of Health.

In vitro kinase assay

Purified Flag-PKR or Flag-Gcn2 and GST-eIF2α proteins were mixed in a reaction buffer (20mM Tris-HCl pH8.0, 50mM KCl, 25mM MgCl₂ and 1mM phenylmethylsulfonyl fluoride) containing γ−³³P-ATP as descried earlier(5). The recombinant human PKCα was purchased from Invitrogen (catalog number: P2232) and used for in vitro assay in a reaction buffer without lipid mixtures (20mM HEPES pH7.4, 10mM MgCl₂, 100mM CaCl₂, 50mM ATP and 5 μCi γ−³³P-ATP).

Structural models and figures

The coordinates of eIF2α (PDB codes 1Q46) and PKR (PDB code 2A1A) were used to generate structural model using PyMol software (https://pymol.org).

Table 1:

Plasmids used in this study are listed in the following Table 1

	Gene	Mutation	Plasmid name	Ref
1	eIF2α in pRS315	WT	pC171	Dever et al. (1992)
2	-	S51A	pB1098	Dever et al. (1992)
3	-	L46AL47SELS51	pC3131	This study
4	-	L46AL47SELA51	pC3132	This study
5	-	L46AL47SESS51	pC3133	This study
6	-	L46AL47SESA51	pC3134	This study
7	-	L46AL47SEAS51	pC3135	This study
8	-	L46AL47SEAA51	pC3136	This study
9	-	SEL48-50AAA	pC3145	This study
10	-	SEL48-50RAA	pC3146	This study
11	-	SELS48-51RAAA	pC3147	This study
12	-	R52S	pC3148	This study
13	PKR in pEMBLyex4	S51A,R52S	pC3149	This study
14	K296R in pEMBLyex4	WT	pC1685	Dey et al. (2005)
15	GCN2 in pEMBLyex4	K296R	P2828	Dey et al. (2005)
16	GST-eIF2α(1–180)	WT	pC2436	Dey et al. (2007)
17	-	WT	pC1638	Dey et al. (2005)
18	-	S51A	pC1613	Dey et al. (2005)
19	-	L46AL47SELS51	pC2968	This study
20	-	L46AL47SELA51	pC2969	This study
21	-	L46AL47SESS51	pC2970	This study
22	-	L46AL47SEAS51	pC2971	This study
23	-	L46AL47SESA51	pC3037	This study
24	-	L46AL47SEAA51	pC3038	This study
25	-	RAA-S51	pC3076	This study
26	-	RAA-S51A	pC3077	This study
27	-	R52S	pC3150	This study
28	-	S51A,R52S	pC3151	This study
29	-	AAA-S51-AAA	D483	This study
30		AAA-S51A-AAA	D819	This study