Characterization of the initiation factor eIF2B and its regulation in Drosophila melanogaster.

Eukaryotic initiation factor (eIF) 2B catalyzes a key regulatory step in the initiation of mRNA translation. eIF2B is well characterized in mammals and in yeast, although little is known about it in other eukaryotes. eIF2B is a hetropentamer which mediates the exchange of GDP for GTP on eIF2. In mammals and yeast, its activity is regulated by phosphorylation of eIF2alpha. Here we have cloned Drosophila melanogaster cDNAs encoding polypeptides showing substantial similarity to eIF2B subunits from yeast and mammals. They also exhibit the other conserved features of these proteins. D. melanogaster eIF2Balpha confers regulation of eIF2B function in yeast, while eIF2Bepsilon shows guanine nucleotide exchange activity. In common with mammalian eIF2Bepsilon, D. melanogaster eIF2Bepsilon is phosphorylated by glycogen synthase kinase-3 and casein kinase II. Phosphorylation of partially purified D. melanogaster eIF2B by glycogen synthase kinase-3 inhibits its activity. Extracts of D. melanogaster S2 Schneider cells display eIF2B activity, which is inhibited by phosphorylation of eIF2alpha, showing the insect factor is regulated similarly to eIF2B from other species. In S2 cells, serum starvation increases eIF2alpha phosphorylation, which correlates with inhibition of eIF2B, and both effects are reversed by serum treatment. This shows that eIF2alpha phosphorylation and eIF2B activity are under dynamic regulation by serum. eIF2alpha phosphorylation is also increased by endoplasmic reticulum stress in S2 cells. These are the first data concerning the structure, function or control of eIF2B from D. melanogaster.

Eukaryotic initiation factor (eIF) 2B catalyzes a key regulatory step in the initiation of mRNA translation. eIF2B is well characterized in mammals and in yeast, although little is known about it in other eukaryotes. eIF2B is a hetropentamer which mediates the exchange of GDP for GTP on eIF2. In mammals and yeast, its activity is regulated by phosphorylation of eIF2␣. Here we have cloned Drosophila melanogaster cDNAs encoding polypeptides showing substantial similarity to eIF2B subunits from yeast and mammals. They also exhibit the other conserved features of these proteins. D. melanogaster eIF2B␣ confers regulation of eIF2B function in yeast, while eIF2B⑀ shows guanine nucleotide exchange activity. In common with mammalian eIF2B⑀, D. melanogaster eIF2B⑀ is phosphorylated by glycogen synthase kinase-3 and casein kinase II. Phosphorylation of partially purified D. melanogaster eIF2B by glycogen synthase kinase-3 inhibits its activity. Extracts of D. melanogaster S2 Schneider cells display eIF2B activity, which is inhibited by phosphorylation of eIF2␣, showing the insect factor is regulated similarly to eIF2B from other species. In S2 cells, serum starvation increases eIF2␣ phosphorylation, which correlates with inhibition of eIF2B, and both effects are reversed by serum treatment. This shows that eIF2␣ phosphorylation and eIF2B activity are under dynamic regulation by serum. eIF2␣ phosphorylation is also increased by endoplasmic reticulum stress in S2 cells. These are the first data concerning the structure, function or control of eIF2B from D. melanogaster.
The binding of the initiator Met-tRNA i to the 40 S ribosomal subunit is a key control point in the initiation of mRNA translation in both Saccharomyces cerevisiae and mammals (1)(2)(3). This step is mediated by eukaryotic initiation factor (eIF) 1 2, a heterotrimeric GTP-binding protein, which, when liganded with GTP, is able to bind the initiator Met-tRNA to form a ternary complex (eIF2⅐GTP⅐Met-tRNA i ). This complex then binds to the 40S ribosomal subunit, forming the 43S pre-initiation complex that then interacts with other initiation factors on the mRNA to allow selection of the translation start codon. The GTP molecule is hydrolyzed late in the initiation process releasing eIF2 as a relatively stable and inactive binary complex (eIF2⅐GDP). At physiological magnesium concentrations, the affinity of eIF2 for GDP is high and recycling of eIF2 to the active GTP-bound state requires a further protein factor, eIF2B, a guanine nucleotide-exchange factor, which in yeast and mammals is composed of five nonidentical subunits (␣-⑀) (2).
The activity of eIF2B is limiting for peptide chain initiation and is regulated under a variety of conditions. The best characterized mechanism of regulation, which is known to operate both in yeast and in mammalian cells, is the inhibition of eIF2B by the phosphorylation of its substrate, eIF2, at a highly conserved serine residue (Ser 51 in mammals) in its ␣-subunit (3). Phosphorylated eIF2␣ acts as a potent competitive inhibitor of eIF2B, and, as cellular levels of eIF2 generally exceed those of eIF2B, low levels of eIF2␣ phosphorylation can cause substantial inhibition of eIF2B. Recent data suggest that the ␣-, ␤-, and ␦-subunits of eIF2B interact with eIF2 and that these subunits are required to sensitize eIF2B to inhibition by this mechanism (4,5).
To date, a number of eIF2␣ kinases have been studied in yeast and mammals. In S. cerevisiae, general control nonderepressible 2 (GCN2), is activated under conditions of amino acid deprivation, leading to increased phosphorylation of eIF2␣ and partial inhibition of global protein synthesis. However, translation of the mRNA for the transcriptional activator GCN4 is actually enhanced by virtue of a set of short upstream open reading frames in its 5Ј-leader region (1). GCN2 homologs have been described in mammals and in Drosophila melanogaster (6 -8), suggesting analogous regulatory mechanisms may operate in these species. Recently, a new eIF2␣ kinase, pancreatic eukaryotic initiation factor-2␣ kinase (PEK), was identified in mammalian pancreatic cells and characterized as a membranebound protein localized in the lumen of the endoplasmic reticulum (ER). This kinase has been implicated in the control of translation in response to ER stresses such as improper protein folding. In total, mammalian cells possess at least four eIF2␣ kinases (heme-regulated inhibitor (HRI; Ref. 9), interferoninduced double-stranded RNA-activated protein kinase (PKR; Ref. 10), general control nonderepressible 2 (GCN2; Ref. 11), and pancreatic eukaryotic initiation factor 2␣ kinase (PEK, also termed PERK; Refs. 6 and 7)), which are generally activated under conditions of cellular stress (e.g. viral infection, disruption of endoplasmic reticulum function, and heme deprivation (Ref. 3)).
The initiation phase of translation and the role of eIF2 and its regulation have been studied intensively in S. cerevisiae and also in mammalian systems (12). However, relatively little work has focused on other metazoans. In D. melanogaster the evidence that a mechanism of guanine nucleotide exchange exists remains equivocal. Previous studies indicated that D. melanogaster eIF2 can be purified as a stable binary complex with GDP. The affinity of eIF2 for GDP at physiological magnesium concentrations was such that a mechanism of catalysis would be required to form the active eIF2⅐GTP complex (13,14). However, Mateu and colleagues (13,15) reported that nucleotide exchange on D. melanogaster eIF2 was independent of an exchange factor under several conditions. Moreover, no guanine nucleotide exchange activity for eIF2 in D. melanogaster was detected in embryos (14).
The sequence of eIF2␣ from D. melanogaster does, however, contain a seryl residue (Ser 50 ) in the position corresponding to Ser 51 in mammals and in a very similar sequence context. Indeed, this residue and the surrounding 19 amino acids are conserved with those found at the phosphorylation site (Ser 51 ) in mammalian and S. cerevisiae eIF2␣ (16). D. melanogaster eIF2␣ can be phosphorylated in vitro by reticulocyte HRI, and this phosphorylation was shown to inhibit guanine nucleotide exchange by mammalian eIF2B (13,14). Phosphorylation of the ␣-subunit of D. melanogaster eIF2 at residue Ser 50 in vivo has never been examined. However, two eIF2␣ kinases have been identified in D. melanogaster. A D. melanogaster ortholog of the S. cerevisiae GCN2p kinase has been identified and characterized (17,18). Complementation experiments in gcn2-deleted strains of S. cerevisiae have confirmed that D. melanogaster GCN2 (dGCN2) is a functional homolog of GCN2p. Expression studies show that dGCN2 is expressed in a developmentally regulated manner and is restricted to the central nervous system during later stages of development. More recently, a D. melanogaster ortholog of the mammalian PEK has been identified through sequence homology (8). However, the physiological conditions and mechanism by which these kinases function in vivo in D. melanogaster remain unclear, especially in the absence of information about eIF2B and its regulation in this species. Thus, it was unclear whether D. melanogaster possessed or required a factor equivalent to eIF2B or whether this process was truly regulated by eIF2␣ phosphorylation in this organism.
Another mechanism by which eIF2B can be regulated in mammalian systems is the phosphorylation of its ⑀-subunit (eIF2B⑀) by glycogen synthase kinase-3␤ (GSK-3␤). The activity of GSK-3␤ is known to be modulated in response to insulin, which induces the phosphorylation and inactivation of GSK-3␤ (19 -21). This response occurs concomitantly with the dephosphorylation of the ⑀-subunit of mammalian eIF2B, at the site of phosphorylation by GSK-3␤, causing the activation of eIF2B (22). D. melanogaster has a homolog of GSK-3␤, Shaggy (23), however, although its role in insulin signaling has not been elucidated. Genetic evidence has indicated that Shaggy acts downstream of Dishevelled in the Wingless pathway inactivating Armadillo (the D. melanogaster homolog of ␤-catenin) (24), and biochemical evidence suggests Shaggy is downstream of protein kinase C (25). A putative phosphorylation site for GSK-3␤ has been identified in D. melanogaster eIF2B⑀, based on sequence homology (26). However, the role of phosphorylation of this site has not been studied.
Together, these data suggest that a mechanism of guanine nucleotide exchange and its regulation by eIF2␣ and eIF2B⑀ phosphorylation probably exist in D. melanogaster. However, neither nucleotide exchange (eIF2B) activity nor the phosphorylation of eIF2␣ has been demonstrated in vivo in D. melanogaster. Given the recent discoveries of eIF2␣ kinases in this species, establishing that eIF2␣ phosphorylation is a regulatory mechanism in the initiation of translation, in fruit flies, was an important priority. In this study we report the presence of guanine nucleotide exchange activity in D. melanogaster S2 cell lines and its regulation by eIF2␣ phosphorylation in vitro and in vivo. We identify cDNAs encoding all five subunits of D. melanogaster eIF2B (␣, ␤, ␥, ␦, and ⑀), and have cloned cDNAs encoding the ␣-, ␤-, ␥-, and ⑀-subunits. We have also characterized the functions of the ␣and ⑀-subunits. We also report that eIF2␣ phosphorylation occurs in a regulated manner in vivo, that GSK-3␤ phosphorylates eIF2B⑀ in vitro, and that eIF2B activity can be inhibited in vitro by GSK-3␤. eIF2B and its regulation in this species appear to be similar to other eukaryotic organisms that have so far been studied. Cell Culture, Treatment, and Lysis-D. melanogaster Schneider (S2) cells were grown in 75-ml tissue culture flasks in DES expression medium with L-glutamine (Invitrogen, Netherlands), containing 10% heat-inactivated fetal calf serum (Life Technologies, Paisley, UK) (27). S2 cell used in guanine nucleotide exchange assays were harvested (at a density of 8 ϫ 10 6 cells/ml) in lysis buffer containing 100 mM Tris, pH 7.6, 50 mM ␤-glycerophosphate, 0.5 mM sodium orthovanadate (Na 3 VO 4 ), 1.5 mM EGTA, 0.1% Triton, 0.1 mM dithiothreitol, 1 g/ml microcystin, protease inhibitors (leupeptin, aprotinin, pepstatin, and benzamidine; all 1 g/ml), and 0.1 mM phenylmethylsulfonyl fluoride. For starvation experiments, cells were starved by placing them into medium containing no serum when they reached a density of 3 ϫ 10 6 cells/ml and lysed or recovered at the times indicated. For experiments involving ER stresses, thapsigargin (1 g/ml) in dimethyl sulfoxide or tunicamycin (1 M) also in dimethyl sulfoxide was added to the cells 12 h prior to lysis; control cells were incubated for the same time in the same concentration of dimethyl sulfoxide.

Chemicals and Biochemicals-Chemicals
Gel Electrophoresis and Immunoblotting-SDS-polyacrylamide gel electrophoresis was performed using gels containing 12.5% acrylamide and 0.1% N,NЈ-methylene-bis-acrylamide (28). Gels were either stained with Coomassie Blue and dried or transferred to polyvinylidene difluoride membranes (Immobilon, Millipore) and subjected to immunoblotting. For Western blotting, samples of total extracts from D. melanogaster S2 cells were subjected to SDS-polyacrylamide gel electrophoresis and probed with either an antibody raised against the peptide CQFDPE-KEFNHKGSGAGR corresponding to residues 313-330 of D. melanogaster eIF2␣ (␣DeIF2B␣) or an antibody against a peptide with the sequence GMILLSELSpRRRIRIN (where Sp denotes a phosphoseryl residue) corresponding to the phosphorylation site in eIF2␣ (New England Biolabs). Anti-His and anti-Myc antibodies (both Sigma-Aldrich) were used as indicated in the figure legends. Antibody-antigen complexes were detected using ECL (Amersham Pharmacia Biotech) and horseradish peroxidase-conjugated sheep, rabbit, or mouse secondary antisera.
Assays for Translation Factors-Guanine nucleotide exchange (eIF2B) activity was determined by measuring the loss of [ 3 H]GDP from pre-formed mammalian eIF2⅐[ 3 H]GDP binary complexes, in the pres-ence of GTP, in an assay similar to that described (29,30). More specifically, formation of the binary complex was achieved by incubating 570 nM purified eIF2 with 7.2 M [ 3 H]GDP in 20 mM Tris-HCl, pH 7.6, 100 mM KCl, 1 mg/ml bovine serum albumin, and 1 mM dithiothreitol for 20 min at 30°C (ϳ1 pmol of of eIF2 binds 1 pmol of [ 3 H]GDP). Assays were carried out, following the addition of 20 mM Tris-HCl, pH 7.6, 100 mM KCl, 1 mM MgCl 2 , and 200 M GTP, at 30°C. D. melanogaster S2 cell extract (45 g of protein) was used in each assay, unless otherwise stated. Following a given time, a sample was removed and diluted in 1 ml of ice-cold 50 mM Tris-HCl, pH 7.6, 90 mM KCl, and 5 mM Mg(CH 3 CO 2 ) 2 and filtered through nitrocellulose. Filters were washed in the same buffer and dried, and associated radioactivity was determined by scintillation counting.
Isolation of eIF2 and eIF2B-Purification of eIF2, as a substrate for eIF2B assays, was carried out as described (32) except that HeLa cell extracts were used as the source instead of rabbit reticulocyte lysate. The partial purification of D. melanogaster eIF2⅐eIF2B complex was also performed in a similar manner. 5 liters of S2 cells were grown to a density of 6 ϫ 10 6 cells/ml and harvested by centrifugation at 480 ϫ g. The cells were then lysed mechanically in lysis buffer containing 20 mM HEPES/KOH, pH 7.6, 0.5% glycerol, 100 mM KCl, 0.1 mM EDTA, 1 mM dithiothreitol, protease inhibitors (leupeptin, aprotinin, pepstatin, and benzamidine; all 1 g/ml), and 0.1 mM phenylmethylsulfonyl fluoride. Extract was then loaded on to a Fast Flow Q-Sepharose column and eluted using a continuous gradient of KCl (from 100 mM TO 1 M in a total of 20 ml). Fractions (0.5 ml) were collected.
DNA Cloning and Sequence Analysis and Transfections-Sequence homology searches were performed using the BLAST program (33) in the D. melanogaster data base Flybase (available via the World Wide Web) to obtain sequence data for the eIF2B subunits. These searches revealed either expressed sequence-tagged cDNA encoding a partial sequence (HL01112) for the ␣-subunit, or genomic sequence (from the D. melanogaster sequencing project for the others). The ␤-subunit sequence is encoded within locus DMC100G10.3 (accession no. AL023874), and the ⑀-subunit is encoded within the cosmid clone 86E4. Unfortunately, these sequences were either partial sequences or contained introns. To obtain full-length sequence, oligonucleotide primers were designed to the 5Ј end of the known sequence and used with a T7 primer (at the 3Ј end of the cDNA library used) to amplify cDNAs encoding the full-length sequence. These were then cloned into the vector pGEMTeasy by ligation of the A and T nucleotides on the 3Ј and 5Ј ends of the PCR product and the vector. The sequence for the ␥-subunit was also identified using BLAST searches, which revealed an expressed sequence tag encoding what initially seemed to be a partial sequence (GM07434, accession no. AA696313). However, when sequenced, the full-length coding region of this protein was revealed. The sequences encoding the putative ␦-subunit of D. melanogaster eIF2B was found by searching the recently completed D. melanogaster genomic sequence; however, cDNA encoding this subunit has not been acquired. This has revealed a full-length sequence encoding 626 amino acid residues with homology to mammalian and yeast eIF2B␦ (see "Results").
Where appropriate, oligonucleotide primers were then designed and used to PCR amplify cDNAs from the Nicholas Brown cDNA library (34). PCR reaction products were then gel purified and subcloned into a pGEMTeasy vector using a TA cloning kit (Promega). cDNAs were sequenced on both strands by dideoxy chain termination method using the ABI PRISM dye terminator cycle sequence ready reaction kit with AmpliTaq DNA polymerase FS and the Automatic Sequencer system 373A (Applied Biosystems). For bacterial expression the cDNA sequence for the ␣-subunit was the cloned (in frame) into pET28c(ϩ) (Novagen), to produce a His-tagged eIF2B␣, using NdeI and BamHI (sites for cloning were introduced on the oligonucleotide primers). The cDNA sequence encoding the ⑀-subunit was cloned into pGEX-HA (in frame) using NdeI and XhoI for bacterial expression and pcDNA3.1(Ϫ)/Myc-His (Invitrogen) using EcoRI and HindIII for expression in human embryonic kidney (HEK) 293 cells. HEK 293 cells were transfected using the calcium phosphate method as described previously (35). Cells were harvested 3 days after transfection and lysed in the buffer used for lysing S2 cells (described above).
Transcription and Translation of cDNA Sequences-In vitro transcription and translation reactions were performed using the T'n'T reticulocyte lysate system (Promega).
Expression of Recombinant Proteins-Escherichia coli (BL21 DE3 or JM109) transformed with the appropriate vector was grown at 37°C overnight in LB containing 100 g/ml ampicillin. They were then di-luted 1/10 and grown to an A 600 of 1. Cultures were then cooled on ice for 15 min and induced with 0.5 mM isopropyl-1-thio-␤-D-galactopyranoside for 5 h. Cells were harvested by centrifugation at 3500 ϫ g and lysed in 20 mM Tris-HCl, pH 7.6, 200 mM NaCl, 10% glycerol, 0.5% Nonidet P-40, protease inhibitors (leupeptin, aprotinin, pepstatin, and benzamidine; all 1 g/ml), 0.1 mM phenylmethylsulfonyl fluoride, and 0.2 g/ml lysozyme for 30 min on ice. To ensure lysis and shear any DNA, the cells were sonicated. Purification of recombinant proteins was performed using either nickel-nitrilotriacetic acid-agarose (Qiagen) or glutathione-Sepharose 4B (Amersham Pharmacia Biotech). Proteins were used the same day.

Schneider Cell Extracts Display eIF2B Activity, Which Is
Inhibited by Phosphorylation of eIF2␣-Previous data concerning the existence of eIF2B in D. melanogaster were equivocal. Thus, to assess whether D. melanogaster cells contain a protein with eIF2B activity, we assayed extracts of S2 Schneider cells for their ability to catalyze guanine nucleotide exchange on eIF2 using complexes containing mammalian eIF2 and [ 3 H]GDP as substrate. The data (Fig. 1A) clearly show that S2 cell extracts effectively mediate nucleotide exchange on eIF2, allowing bound [ 3 H]GDP to be replaced by GTP.
Pretreatment of S2 cell extracts with the eIF2␣ kinase HRI led to increased phosphorylation of eIF2 as assessed using an antibody specific for the phosphorylated form of eIF2␣ (Fig.  1B). The identity of the band as eIF2␣ was confirmed by comparison with the positions of mammalian and D. melanogaster eIF2␣ (probed with antibodies specific for the respective proteins). This confirms the earlier finding that D. melanogaster eIF2␣ is a substrate for HRI (15). PKR was also able to phosphorylate eIF2␣ in extracts of Schneider cells (data not shown).
When eIF2B assays were performed with cell extract that had been pretreated with HRI, eIF2B activity was markedly reduced (Fig. 1A). Inhibition of eIF2B activity by eIF2␣ phosphorylation is a property common to both yeast and mammalian eIF2B.
Isolation of eIF2 and eIF2B from Schneider Cells-When isolated from mammalian cells, eIF2 and eIF2B tend to copurify with one another through a number of ion-exchange steps (32,40,41). To characterize further the corresponding factors from D. melanogaster, Schneider cell extracts were subjected to ion-exchange chromatography on an Mono-Q column, which was developed with a salt gradient from 0.1 to 1.0 M KCl. Fractions were subjected to immunoblotting with the antibody to D. melanogaster eIF2␣. They were also assayed both for eIF2 activity (measured as formation of ternary complexes) with [ 35 S]Met-tRNA i in the presence of GTP and for eIF2B activity using the mammalian eIF2⅐[ 3 H]GDP complex as substrate.
Western blotting revealed a strong signal with the anti-eIF2␣ antiserum in the region of the gradient corresponding to 350 -450 mM KCl at an apparent molecular mass of 38 kDa (Fig. 2A). These fractions also showed eIF2 activity (Fig. 2B). When fractions in this region of the gradient were assayed for eIF2B activity, nucleotide-exchange activity was observed in fraction 11 (ϵ 490 mM KCl), i.e. just after the peak of eIF2 protein detected by immunoblotting (Fig. 2C). This behavior is similar to that of mammalian eIF2 and eIF2B on Mono-Q as purification produces two pools of eIF2 (due to the excess of eIF2 over eIF2B) consisting of eIF2 (eluted first) and eIF2⅐eIF2B as a complex (eluted slightly later) (32). These data suggest that the chromatographic behavior of D. melanogaster eIF2 and eIF2B is similar to that of their mammalian counterparts and provides evidence that, as in mammals and yeast, these two proteins copurify through ion-exchange chromatography.
Identification and Cloning of D. melanogaster cDNAs with Homology to Subunits of eIF2B from Other Eukaryotes-The above data strongly suggested that D. melanogaster cells contain a factor equivalent to eIF2B from other eukaryotes. To identify sequences encoding potential eIF2B subunits from this species, we searched D. melanogaster nucleotide sequence data bases with the sequences corresponding to subunits of yeast and mammalian eIF2B. This revealed the presence of genomic or expressed sequence tag sequences with high levels of identity to all five subunits of eIF2B (␣-⑀) (Tables I and II, and supplementary data available on-line). This strongly supports the initial conclusion that D. melanogaster does possess a protein homologous to the eIF2B complex in yeast and mammals.
The D. melanogaster subunits are generally similar in size to the corresponding subunits from yeast and from mammals (Table I). Their sequences show high identity to those of the mammalian proteins (about 50% in each case, Table II) and slightly lower identity (but still high similarity) to the yeast polypeptides (identity 29 -37%, similarity around 50%). It has already been noted that the sequences of these five subunits show mutual sequence similarity in yeast (42-44) and mam-mals (45). This is also the case in D. melanogaster (Fig. 3). In addition to the D. melanogaster sequences for eIF2B, reported here, other putative sequences have become available for a range of other species. When D. melanogaster eIF2B sequences are aligned with these other sequences, it is very striking that certain residues are completely conserved, or only very conservatively replaced, across plants (Arabidopsis thaliana), budding and fission yeast, lower animals (Caenorhabditis elegans), insects (D. melanogaster), and mammals (only one sequence included here of the several known to avoid artificially "biasing" the appearance of the alignment).
In the case of the ␣-, ␤-, and ␦-subunits, there are 11 positions in which all 12 sequences have the same residue, and another 11 where the residue is always a hydrophobic one (Leu, Ile, Val, Phe, or Met). This feature of the sequences of these subunits of eIF2B has been noted before (45) but is shown here for more species. These residues do show the rough periodicity of seven, which one would expect for residues involved in interactions between helical regions of the proteins, which might, e.g., be involved in complex assembly. The most highly conserved residues in these sequences fall into three clusters. One is near the N terminus, where the ␣ sequences show similarity to one another not that is shared with ␤ or ␦. The second is a large region (about 130 residues) toward the center of each sequence, which contains many of the highly conserved residues, especially the hydrophobic ones. Not immediately evident from this alignment is the high similarity between the different sequences for the individual subunits (␣, ␤, and ␦) within this region. Finally, there is a highly conserved region almost at the C terminus of the protein (28 residues), which shows marked identities/similarities across all three sequences from all spe-

FIG. 2. Copurification of eIF2 and eIF2B from D. melanogaster
Schneider cell extracts. Protein from Schneider cell extracts was partially purified on a Mono-Q-Sepharose column, which was developed with a KCl gradient ranging from 100 mM to 1 M (as described under "Materials and Methods"). A, fractions were then subjected to Western blotting using an antibody directed against D. melanogaster eIF2␣. B, to determine the activity of eIF2 in each fraction, ternary complex formation was assayed by measuring the binding of [ 35 S]methionyl-tRNA i to eIF2 (20 min at 30°C) as described under "Materials and Methods." C, the eIF2B activity of each fraction was determined using the guanine nucleotide exchange assay as described under "Materials and Methods," except each assay was performed for 10 min at 30°C. All fractions are labeled according to the KCl concentration at which they were eluted. The above data represent two separate purifications.
cies. Within this region, the sequences for any one subunit also show a high degree of identity. For eIF2B␣, 18/28 residues are identical/conservatively replaced in this region; for eIF2B␤, the figure is 20/28; and, for eIF2B␦, 24/28.
The sequences of the ␥and ⑀-subunits of eIF2B also show mutual similarities within a given species. Koonin (46) identified within eIF2B⑀ three motifs similar to those found in many nucleotidyl transferases: (i) a variant of the phosphate-binding loop (P-loop), (ii) a version of the magnesium-binding site of such proteins, and (iii) more C-terminal than the first two, a region containing imperfectly repeated units usually with a hydrophobic residue in the first position. All three motifs are well conserved in the ␥and ⑀-subunits of eIF2B from D. melanogaster ( Fig. 3 and supplementary data available on-line). Within the third of these regions, the most C-terminal one, the positions of the hydrophobic residues in other species are also occupied by hydrophobic residues in D. melanogaster eIF2B⑀. The observation (made for the mammalian and yeast sequences of eIF2B⑀) that the Ile residues within this region are often followed by Gly is not, however, a consistent feature of the D. melanogaster sequence, where only two such pairs are found. The conserved triplet NFD (residues 249 -251 of yeast eIF2B⑀, noted by Gomez and Pavitt (Ref. 36)) is also found in D. melanogaster eIF2B⑀. This therefore remains as the only completely conserved triplet sequence in eIF2B⑀ sequences. Mutations at the Asn and Phe residues within this triplet impair the activity of yeast eIF2B (36). eIF2B⑀ from mammals contains a conserved phosphorylation site for glycogen synthase kinase-3 (GSK-3) located at Ser 540 in the rabbit sequence (22) (shown bold here, ELDSRAGSPQL). Four residues C-terminal to this is a second seryl residue (underlined), which undergoes phosphorylation and probably serves as a "priming" site for phosphorylation by GSK-3 (47,48). The D. melanogaster sequence contains a seryl residue in the position corresponding to the GSK-3 site in mammals, but the priming site is occupied by Thr in D. melanogaster (EDAS-RAVTPLP). We have shown previously that phospho-threonyl residues can efficiently prime GSK-3 mediated phosphorylation, at least when peptides are used as substrates (26).
Characterization of Proteins Encoded by the D. melanogaster cDNAs for eIF2B Homologs-cDNAs encoding the putative subunits of D. melanogaster eIF2B were cloned into pGEM3Z (␣, ␤, and ⑀) or pBluescript (␥), and the polypeptides encoded were studied by in vitro coupled transcription/translation, using [ 35 S]methionine to label the synthesized polypeptides, which were then resolved on SDS-polyacrylamide gel electrophoresis. As shown in Fig. 3C, each cDNA gave rise to a single major translation product, and no product was observed for the empty vector control (data not shown). The cDNA clones for eIF2B␣, -␤, -␥, and -⑀ produced polypeptides of ϳ34, 39, 50, and 72 kDa, which compare well with the molecular masses expected for these polypeptides (Table I). Since no clone was available for the ␦-subunit, we were unable to perform a similar analysis for this polypeptide.
eIF2B⑀ is thought to be the (principal) catalytic subunit of the eIF2B complex (4,36,49). To test whether the D. melanogaster homolog of eIF2B⑀ actually possessed guanine nucleotide exchange activity, we expressed it both as a hexahistidineand Myc-tagged protein in HEK 293 cells or as a glutathione S-transferase (GST) fusion protein in E. coli. eIF2B⑀ and, as a control, eIF2B␣ were each expressed separately with Myc-His tags in 293 cells (Fig. 4A). To determine whether the ⑀-subunit of D. melanogaster eIF2B possessed guanine nucleotide-exchange activity, samples of extracts from HEK 293 cells expressing eIF2B⑀ and corresponding controls, transfected with empty vector or a vector encoding the noncatalytic ␣-subunit of eIF2B, were analyzed in our standard exchange assay (Fig.  4B). Samples from cells expressing eIF2B⑀ showed substantially enhanced exchange activity relative to the control (empty vector). Cells expressing eIF2B␣ showed no increase in nucleotide-exchange activity relative to cells transfected with the empty vector, confirming that eIF2B␣ itself is inactive in nucleotide exchange.
D. melanogaster eIF2B⑀ was also expressed in E. coli as a GST fusion protein, and purified using glutathione-Sepharose 4B (see Fig. 4C). Samples were subjected to our standard assays for GDP/GTP exchange, and substantial activity was observed, while none was seen for cells transformed with empty vector, i.e. expressing only GST (Fig. 4D). This confirms that the ⑀-subunit of eIF2B from D. melanogaster itself has exchange activity as concluded previously.
As noted above, the sequence of D. melanogaster eIF2B⑀ contains a seryl residue in a similar to position of the GSK-3 site in mammalian eIF2B⑀ and this is followed by a threonyl residue at the ϩ4 position which, when phosphorylated can "prime" the phosphorylation of the more N-terminal residue by GSK-3␤ at least when synthetic peptides based on this sequence are studied. The priming site in both mammals and D. melanogaster eIF2B⑀ is followed by a prolyl residue, and it seems likely that it is a target for a proline-directed protein kinase. To date, however, we have not been able to identify the kinase responsible for phosphorylating the priming site in mammalian eIF2B⑀; a number of proline-directed kinases such as members of the mitogen-activated protein kinase and cdc2 families have been tested, but none was able to phosphorylate this site. However, preparations of GSK-3␤ made in Spodoptera frugiperda cells appear to be contaminated with a kinase which can catalyze phosphorylation of this site. 2 We therefore used such preparations of GSK-3␤ to test whether they could phosphorylate D. melanogaster eIF2B⑀ which had been expressed in E. coli. As shown in Fig. 5A, GSK-3␤ catalyzed phosphorylation of the D. melanogaster eIF2B⑀. No phosphorylation was observed when GSK-3␤ was omitted from the reactions. However, phosphorylation by GSK-3 was substoichiometric, even when extended incubation periods were used, possibly because the amount of the priming kinase present is too low to permit more efficient phosphorylation. To test whether this phosphorylation by GSK-3␤ inhibited the activity of D. melanogaster eIF2B, as is the case with mammalian eIF2B, D. melanogaster eIF2B was purified using a Mono-Q column to remove any endogenous Shaggy and incubated with GSK-3␤ prior to using it in a guanine nucleotide exchange assay. Following treatment with GSK-3␤, the activity of this partially purified D. melanogaster eIF2B was inhibited by ϳ50% (Fig. 6) indicating that, in vitro, D. melanogaster eIF2B activity is impaired by GSK-3␤-mediated phosphorylation. Mammalian eIF2B⑀ is also phosphorylated by casein kinase II (CK-II, Refs. 32 and 50). We therefore tested whether CK-II could also phosphorylate the recombinant D. melanogaster eIF2B⑀ made in E. coli. Recombinant CK-II catalyzed phosphorylation of fly eIF2B⑀ in vitro (Fig. 5B). Phosphorylation of mammalian eIF2B⑀ by CK-II appears to occur at seryl residues within an acidic region at the C terminus of the protein. 3 D. melanogaster eIF2B⑀ contains a similar acidic region at its C terminus (DDQSSEEDDDEEDDD), and it is likely that CK-II phosphorylates one or both seryl residues within this region.
Complementation of Yeast GCN3 by D. melanogaster eIF2B␣-In yeast cells, amino acid deprivation leads to the activation of the eIF2␣ kinase GCN2. In addition to reducing overall translation initiation, phosphorylation of eIF2␣ induces translation of GCN4 mRNA (51). GCN4 is a transcriptional activator of multiple amino acid biosynthetic genes; therefore, induction of GCN4 translation is required for growth in the presence of 3-AT, an inhibitor of histidine biosynthesis. Thus, deletion of GCN2 makes yeast cells sensitive to 3-AT (3-AT S ). Deletion of GCN3, the yeast eIF2B␣ subunit, also causes a 3-AT S phenotype as it apparently renders yeast eIF2B insensitive to the normally inhibitory effects of eIF2␣ phosphorylation on eIF2B-catalyzed guanine-nucleotide exchange (4). It has been shown previously that mammalian (rat) eIF2B␣ can substitute for GCN3 in vivo in yeast (52). We therefore asked whether expression of the D. melanogaster protein could complement the GCN4-dependent defect in histidine biosynthesis caused by deletion of GCN3.
The D. melanogaster eIF2B␣ cDNA was cloned downstream of a yeast galactose inducible promoter on a high copy number plasmid and introduced into yeast strains deleted for GCN3 or GCN2. In parallel, plasmids bearing yeast GCN3 or GCN2 were also transformed into these two strains. In the gcn3⌬ strain, D. melanogaster eIF2B␣ conferred resistance to 3-AT equivalent to that shown by GCN3. The control plasmid bearing GCN2 remained 3-AT S (Fig. 7, gcn3⌬ panel labeled SGalϩ3-AT). All transformed strains grew equivalently in the absence of starvation (Fig. 7, panels labeled SGal). These results indicate that D. melanogaster eIF2B␣ can substitute for GCN3 in yeast eIF2B and restore the normal response to amino 3 X. Wang, unpublished data. In vitro transcription and translation reactions were carried out as described under "Materials and Methods." Translation products labeled with [ 35 S]methionine were resolved on a 12.5% SDS-polyacrylamide gel, which was treated with Amplify (Amersham Pharmacia Biotech) and exposed to the autoradiographic film overnight. Migration of the subunits was compared with broad range protein markers from Bio-Rad, their positions being shown on the left. acid starvation. DeIF2B␣, like GCN3, did not confer 3-AT resistance in the gcn2⌬ strain (Fig. 7, right panel). This shows that 3-AT resistance conferred by D. melanogaster eIF2B␣ is dependent on phosphorylation of eIF2␣ by GCN2 rather than the result of a bypass caused by interfering with the activity of eIF2 or eIF2B.
The Phosphorylation of eIF2␣ and eIF2B Activity Are Regulated in S2 Schneider Cells-When extracts from 72-h serumstarved cells were assayed for eIF2B activity, none was detected (Fig. 8B), whereas substantial activity was seen in extracts from control cells analyzed in parallel. When S2 cells were starved of serum for 72 h, the level of phosphorylation of eIF2␣ increased markedly as revealed using the antibody against eIF2␣ phosphorylated at Ser 50 (Fig. 8A, lane 5 as  compared with lane 4). Taken together, these results suggest that the removal of serum causes the inhibition of eIF2B activity through the phosphorylation of its substrate eIF2 and also that there may be an eIF2␣ kinase in S2 cells that is activated under conditions of serum deprivation. To study this further, we subjected S2 cells to differing periods of serum withdrawal up to 72 h, and then used the anti-phosphorylated eIF2␣ antibody to assess the level of eIF2␣ phosphorylation. The data clearly showed that, even after 12 h without serum, the level of eIF2␣ phosphorylation was substantially increased compared with the serum-fed control (Fig. 8C), taking into account the signal from a loading control blot performed using an antiserum that detects eIF2␣ irrespective of its state of phosphorylation (Fig. 8C). Consistent with this increase in eIF2␣ phosphorylation, we observed a marked decrease in eIF2B activity at all times of serum withdrawal (Fig. 8D). Readdition of fresh serum for 1 h caused substantial reactivation of eIF2B (Fig. 8E). Serum treatment for 1 h clearly caused a marked reduction in the level of phosphorylation of eIF2␣, which may account for the observed increase in eIF2B activity.
The Phosphorylation of eIF2␣ Increases in Response to ER Stress in D. melanogaster S2 Cells-The recent identification by sequence homology of the eIF2␣ kinase, PEK, in D. melanogaster (8) and the characterization of its homolog in mammals (7) prompted us to investigate whether eIF2␣ phosphorylation occurred during endoplasmic reticulum stress in D. melanogaster. D. melanogaster S2 cells were incubated for 12 h in the presence and absence of thapsigargin and tunicamycin (agents that interfere with ER function) and then subjected to immunoblotting using the anti-phospho-eIF2␣ antibody. Fig. 9 clearly shows that eIF2␣ phosphorylation is increased upon addition of these agents when compared with control cells. DISCUSSION The data presented here clearly demonstrate that D. melanogaster cells possess eIF2B activity and that this organism contains genes encoding proteins very similar to all five subunits of eIF2B previously identified in other eukaryotes such as yeast and mammals. In particular, we show that the putative ␣-subunit of D. melanogaster eIF2B can functionally replace the endogenous yeast protein in budding yeast and confer regulation on the yeast eIF2B complex by phosphorylation of eIF2␣, and that the D. melanogaster eIF2B⑀ protein, when expressed in bacteria or in mammalian cells, catalyzes guanine nucleotide exchange on eIF2. We can therefore conclude that these genes do encode authentic subunits of eIF2B. Furthermore, treatment of D. melanogaster S2 cell extracts with an eIF2␣ kinase to phosphorylate the endogenous eIF2␣ caused near complete inhibition of nucleotide exchange on exogenous eIF2. This shows that phosphorylation of D. melanogaster eIF2 on its ␣-subunit inhibits eIF2B activity as is also the case for eIF2B from budding yeast and from mammals. In addition, when eIF2␣ phosphorylation was increased within S2 cells as a consequence of serum starvation, inhibition of eIF2B activity was again observed. Thus, D. melanogaster possesses an eIF2B complex, which can be regulated by eIF2␣ phosphorylation in vivo. The sequence data also strongly indicate that the subunits of the complex are closely related to the corresponding orthologs from other eukaryotes. Chefalo et al. (53) have previously reported that the eIF2B activity detected in extracts of Sf9 (S. frugiperda) cells was inhibited when the eIF2 kinase HRI was expressed in those cells. This suggested that eIF2B in insect cells was sensitive to inhibition by phosphorylation of eIF2, consistent with our present data.
The earlier data (15) suggested that D. melanogaster did not possess an eIF2B-like factor. This was based partly on the fact that, under some conditions (e.g. at elevated temperature or using partially purified eIF2 preparations), GDP/GTP exchange occurred freely on D. melanogaster eIF2, without a requirement for eIF2B. Consistent with this, under these conditions, exchange was not affected by raising the concentration of Mg-ions (15). It is not clear why temperature should have this effect of rendering exchange independent of eIF2B, but in the case of the partially purified material it is very likely that these effects were due to the presence in the eIF2 fractions of eIF2B, which copurifies with eIF2 from D. melanogaster cells (this work) and those of other species (see, e.g., Ref. 32). The presence of eIF2B in the substrate eIF2 would overcome the inhibitory effect of magnesium ions and eliminate the need for added eIF2B. The presence of sufficient (e.g. stoichiometric) amounts of eIF2B could also explain the observation that phosphorylation of eIF2␣ did not inhibit exchange under the specific conditions mentioned.
The ⑀-subunit of D. melanogaster eIF2B contains a consensus sequence for phosphorylation by GSK-3 (as is also the case for the factor from all mammals for which eIF2B⑀ has so far been sequenced). We show that D. melanogaster eIF2B⑀ is indeed phosphorylated by GSK-3 in vitro, and that the activity of partially purified D. melanogaster eIF2B is inhibited by in vitro phosphorylation by GSK-3␤. The ability of GSK-3 to phosphorylate eIF2B⑀ that has been expressed in E. coli seems surprising given that it should not be phosphorylated in the priming site, which is normally a prerequisite for phosphorylation of eIF2B⑀ (47,48) and several other substrates by GSK-3 (54 -56). However, this is also seen for the mammalian factor expressed in this system and reflects the presence in our GSK-3 preparations (expressed in the baculovirus system) of a kinase that allowing growth of yeast cells on histidine starvation medium. Plasmids expressing DeIF2B␣, GCN3, or GCN2 were transformed into yeast strains deleted for gcn3⌬ (GP3153, left panels) and gcn2⌬(GP3140, right panels). Cells were grown to confluence on SGal medium and replica plated to SGal medium and to SGal supplemented with 25 mM 3-AT (as indicated). Plates were incubated at 30°C for 3 days.
can phosphorylate the priming site of the mammalian protein. 4 The D. melanogaster GSK-3 homolog Shaggy has a similar substrate specificity to mammalian GSK-3 (26); in particular, it can efficiently phosphorylate peptides based on the sequence around the putative GSK-3 site in D. melanogaster eIF2B⑀.
Shaggy is known to play key roles during development in D. melanogaster and other metazoans, and it is possible that some of its effects are exerted via phosphorylation of eIF2B and the regulation of the translation of specific mRNAs. This regulation would be analogous to the control of GCN4 expression in yeast, which involves modulation of the translation of its mRNA mediated by changes in eIF2B activity. It is known that there are upstream open reading frames in some mRNAs in D. melanogaster (see, e.g., Ref. 37).
Our data also show that the activity of eIF2B is subject to regulation in response to serum in S2 cells. Serum withdrawal causes a marked decline in its activity concomitant with an increase in the phosphorylation of eIF2␣. Resupplying serum led to an increase in eIF2B activity and a fall in eIF2␣ phosphorylation. The data suggest that the changes in eIF2B activity are due to alterations in eIF2␣ phosphorylation, although, as serum restoration did not completely restore eIF2B activity, other regulatory events may also be involved. It is also a possibility that cells starved for long periods of time contain less eIF2B. This would also explain why complete restoration was not achieved. These are the first data showing regulation of eIF2B activity or eIF2␣ phosphorylation in invertebrates.
These findings suggest that serum may regulate the activity of an eIF2␣ kinase in S2 cells. Two eIF2␣ kinases have been described from D. melanogaster, dGCN2 (18) and dPEK (8), and the recently published genome sequence from this organism suggests that they are the only ones. The data of Berlanga et al. (17) indicated that mammalian GCN2 may be activated by serum withdrawal, at least when it is overexpressed in human embryonic kidney 293 cells, suggesting that dGCN2 activity might also be controlled by serum in S2 cells.
We have also reported here that ER stresses induced by thapsigargin, which causes the release of calcium from the ER, and tunicamycin, an inhibitor of protein glycosylation, do cause an increase in eIF2␣ phosphorylation. These data suggest a role for dPEK in D. melanogaster in inhibiting translation in response to ER stress. However, the possibility that only a single eIF2␣ kinase is present and activated by two separate mechanisms cannot be dismissed. Further experimentation will be required to characterize the regulatory roles of these kinases in vivo.
These data demonstrate, using a variety of approaches, that D. melanogaster possesses an eIF2B factor, and regulatory mechanisms to control it, which are similar to those which operate in higher eukaryotes. Subsequent studies will focus on the role of these mechanisms in regulating protein synthesis and thus gene expression in D. melanogaster.