Stimulation of Protein Synthesis in COS Cells Transfected with Variants of the a-Subunit of Initiation Factor eIF-2*

The role of eukaryotic initiation factor 2 (eIF-2) phosphorylation in translational control has been demonstrated in vivo by overexpressing variant forms of eIF-2 alpha that are not phosphorylated. COS-1 cells transiently transfected with expression vectors for human eIF-2 alpha contain 10-20-fold more eIF-2 alpha subunit than the endogenous COS cell eIF-2 trimeric complex. Expression of the variant form of eIF-2 alpha, Ser51Asp, where Asp replaces Ser51, causes inhibition of protein synthesis, whereas the Ser48Asp variant does not. When either Ser48 or Ser51 is replaced by Ala, the variants stimulate dihydrofolate reductase synthesis when the eIF-2 alpha kinase, DAI, is activated. In order to elucidate these mechanisms, we have separated eIF-2 trimeric complexes from free overexpressed eIF-2 alpha subunits by fast protein liquid chromatography Superose chromatography. Pulse-labeled cells transfected with wild-type or variant DNAs produced eIF-2 preparations with greater than 10-fold higher specific radioactivity in the alpha-subunit compared to the gamma-subunit, thus demonstrating that the human eIF-2 alpha produced from the plasmids readily exchanges into COS cell eIF-2 complexes. Both wild-type and Ser48Ala variant forms of the free 2 alpha-subunit, further purified by MonoQ chromatography, are poor substrates for the heme-regulated eIF-2 alpha kinase, HRI, but are good substrates for double-stranded RNA-activated inhibitor in vitro; the Ser51Ala variant subunit is not phosphorylated by either kinase. None of the purified free eIF-2 alpha subunits inhibits phosphorylation of eIF-2 in vitro, even at up to 8-fold molar excess. Examination of the extent of eIF-2 alpha phosphorylation in the COS cell eIF-2 complexes by two-dimensional polyacrylamide gel electrophoresis shows that the stimulation of dihydrofolate reductase synthesis by the Ser51Ala variant is most readily explained by failure of eIF-2 to be phosphorylated. Stimulation by the Ser48Ala variant appears to occur by mitigation of the effect of phosphorylation at Ser51 since the double variant, Ser48Ala-Ser51Asp, inhibits protein synthesis less than the single variant Ser51Asp. The evidence argues strongly against there being a second site of phosphorylation involved in translational repression.


Stimulation of Protein Synthesis in COS Cells Transfected with Variants of the a-Subunit of Initiation Factor eIF-2*
(Received for publication, April 25, 1991) Sang-Yun Choi, Bradley J. SchererS The role of eukaryotic initiation factor 2 (eIF-2) phosphorylation in translational control has been demonstrated in vivo by overexpressing variant forms of eIF-2a that are not phosphorylated. COS-1 cells transiently transfected with expression vectors for human eIF-2a contain 10-20-fold more eIF-20 subunit than the endogenous COS cell eIF-2 trimeric complex. Expression of the variant form of eIF-a~t, Ser'lAsp, where Asp replaces Ser", causes inhibition of protein synthesis, whereas the Ser4'Asp variant does not. When either Ser4' or Ser" is replaced by Ala, the variants stimulate dihydrofolate reductase synthesis when the eIF-2cu kinase, DAI, is activated. In order to elucidate these mechanisms, we have separated eIF-2 trimeric complexes from free overexpressed eIF-2a subunits by fast protein liquid chromatography Superose chromatography. Pulse-labeled cells transfected with wild-type or variant DNAs produced eIF-2 preparations with greater than 10-fold higher specific radioactivity in the a-subunit compared to the y-subunit, thus demonstrating that the human eIF-2a produced from the plasmids readily exchanges into COS cell eIF-2 complexes. Both wild-type and Ser4'Ala variant forms of the free 2a-subunit, further purified by MonoQ chromatography, are poor substrates for the heme-regulated eIF-2cu kinase, HRI, but are good substrates for double-stranded RNA-activated inhibitor in vitro; the Ser61Ala variant subunit is not phosphorylated by either kinase. None of the purified free eIF-2a subunits inhibits phosphorylation of eIF-2 in vitro, even at up to 8-fold molar excess. Examination of the extent of eIF-2cu phosphorylation in the COS cell eIF-2 complexes by two-dimensional polyacrylamide gel electrophoresis shows that the stimulation of dihydrofolate reductase synthesis by the Ser51Ala variant is most readily explained by failure of eIF-2 to be phosphorylated. Stimulation by the Ser4'Ala variant appears to occur by mitigation of the effect of phosphorylation at S e P since the double variant, Ser4'Ala-Ser'lAsp, inhibits protein synthesis less than the single variant Ser51Asp. The evidence argues strongly against there being a second site of phosphorylation involved in translational repression.
* This work was supported in part by National Institutes of Health Grant GM22135 from the United States Public Health Service. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "aduertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
$ Supported by National Insitutes of Health Training Grant GM07377.
ll To whom correspondence should he addressed.
Eukaryotic initiation factor e1F'-2 is identified as one of the factors involved in regulating protein synthesis (for recent reviews, see Refs. 1, 2). The factor forms a ternary complex with the initiator Met-tRNAj and GTP, and promotes the binding of the initiator tRNA to ribosomes. eIF-2 comprises three dissimilar subunits named a, /3, and y, whose molecular masses are 36, 38, and 52 kDa, respectively. Phosphorylation of the a-subunit of eIF-2 correlates with an inhibition of initiation of protein synthesis and has been implicated in numerous translational control mechanisms (3). Two protein kinases have been characterized that phosphorylate eIF-2a: the hemin-regulated inhibitor (HRI) and the double-stranded RNA-activated inhibitor (DAI). A correlation of eIF-2a phosphorylation and inhibition of protein synthesis has been made in cells subjected to serum deprivation (4), heat shock ( 5 ) , interferon treatment followed by virus infection (6), transfection with certain plasmid DNAs ( 7 ) , inter al. It is believed that limited phosphorylation of eIF-2a results in effectively sequestering the less abundant guanine nucleotide exchange factor (eIF-2B, aka GEF), thereby preventing the recycling of eIF-2 (8,9).
In previous work we asked whether or not eIF-2a phosphorylation is the primary cause or a sufficient condition for translation inhibition. The putative phosphorylation sites on eIF-Za, Ser51 a n d S e P , each were altered by site-directed mutagenesis (10, 11) to alanine residues which are expected to prevent phosphorylation. The Ser48Ala variant' (2a-AS) was phosphorylated i n vitro by HRI and DAI, whereas the Ser"A1a variant (2a-SA) was not, suggesting that the site of phosphorylation for these kinases is Ser" (10). The 2 Ala variants as well as the Ser4'Asp (201-DS) and Ser'lAsp (201-SD) variants (designed to mimic phosphorylation at either site) were tested for possible effects on translation in intact cells. Wild-type and mutated cDNAs were individually subcloned in a mammalian expression vector and cotransfected into COS-1 cells with a dihydrofolate reductase (DHFR) expression vector, pD61 (11). DHFR translation from mRNA expressed from pD61 is inefficient due to activation of DAI kinase in the transfected cells (7). Cotransfection with the 2a-SA variant stimulates DHFR synthesis, apparently by preventing phosphorylation or otherwise interfering with the effects of an activated DAI kinase. In contrast, cotransfection with the 2a-SD, but not the 2a-DS, variant completely inhibits DHFR translation, indicating that phosphorylation at Ser61 but not Ser48 is sufficient for translational repression in these cells. Unexpectedly, the 2a-AS variant also stimulates DHFR synthesis, even though the 2a-AS protein can be phosphorylated (10,11). This surprising result suggests that the mechanism of action of the overproduced variant proteins is more complicated than initially envisaged. It also has been argued that phosphorylation at both Ser5' and Ser48 may be required for translation inhibition (12).
In order to understand better the mechanism of variant eIF-2a action in the transiently transfected COS-1 cells, we asked whether or not the over-expressed free eIF-2a proteins serve as HRI and DAI substrates or directly inhibit these kinases in vitro, whether or not they exchange into the endogenous COS cell eIF-2 heterotrimeric complexes, and to what extent the eIF-20 subunit is phosphorylated, either free or in the eIF-2 complexes. In addition, we have examined the expression of double variants of eIF-2a to ascertain the physiological importance of both Ser48 and Ser'l, thereby defining the molecular basis for the action of this initiation factor in translational regulation.
Oligonucleotide-directed mutagenesis was performed to generate Ser"Ala-Ser"Ala (2a-AA) and Ser4"Ala-Ser51Asp (2a-AD) double variants of eIF-2a by the gapped duplex DNA method (14) using pMa/c plasmid vectors. Human eIF-2n cDNA encoding 2n-AS (11) in pUC18 was used for the target DNA fragment. A 23-mer oligonucleotide (5'-CG-CCT-TCT-G(G/T)C-CAA-TTC-AGC-AAG-3') was synthesized and used to mutate T T A -E into TTG-GCC or TTG-GAC resulting in the substitution of an Ala or Asp f o r m ' , respectively (the nucleotides corresponding to codon 51 are underlined). The 2a-AA variant was designed to create a BulI restriction site. Mutated cDNAs were screened and selected by sequencing with an oligonucleotide primer (5'-CT-GCC-AAT-TCG-GAT-GAG-3') by the dideoxynucleotide termination method (15). The EcoRI-HincII fragments of 2a-AA and 20-AD cDNA from the pMc vector were blunt ended with the Klenow fragment of Escherichia coli DNA polymerase, ligated with EcoRI linkers, and digested with EcoRI. The fragments were purified by agarose gel electrophoresis and subcloned into the unique EcoRI site of pMT2VA-to obtain pMT2-2a-AA or pMT2-Pa-AD. The correct orientation and structure of both clones were confirmed by restriction enzyme digestion and gel electrophoresis.
Cell Culture and Transfection-COS-1 cells grown on 100-mm plates were transfected by the DEAE-dextran method (11). Briefly, approximately 70% confluent plates of cells were fed 4 ml of Dulbec-CO'S minimal essential medium (DME) containing 1 mg of DEAEdextran (Pharmacia LKB Biotechnology Inc.) and 8 pg of eIF-201 expression plasmid and/or 8 pg of pD6l/plate for 6-12 h. After washing plates with DME or phosphate-buffered saline, the monolayer cultures were treated with 10% dimethyl sulfoxide for 2 min followed by 0.1 mM chloroquin in DME with 10% fetal bovine serum for 2 h. Plates were then incubated with fresh DME plus serum. To monitor DHFR synthesis and exchange of eIF-2a into eIF-2 complex (see below), cells were pulse-labeled with [3sS]methionine (100 pCi/ ml; >lo00 Ci/mmol, Amersham Corp.) for either 30 min or 8 h beginning at 40-h post-transfection. After 48-or 72-b post-transfection, cells were harvested for IEF-PAGE gel analysis of total cell proteins by direct addition of an ampholyse solution (9.8 M urea, 2% Nonidet P-40, 2% pH 3-10 Ampholytes, 25 mM NaF, and 1% 2mercaptoethanol) as described (16). For analyses under non-denaturing conditions, cells were collected with cold phosphate-buffered saline, pelleted, and stored immediately at -70 "C until further use.
Separation of the eIF-2 Complex from Free e l F -2~~ Subunits by Gel Filtration-Gel filtration column chromatography was performed using a Pharmacia FPLC system to separate the COS cell eIF-2 trimeric complex from free overproduced human eIF-2cu subunit. Cell pellets from one 100-mm plate (-2 X lo6 cells) were suspended in 200 p1 of non-denaturing lysis buffer (20 mM HEPES, pH 7.5,50 mM KCl, 1 mM EDTA, 25 mM NaF, 7 mM 2-mercaptoethanol, 5% glycerol, 0.5% Nonidet P-40), and subsequently lysed by 25 strokes in a Dounce homogenizer. Bovine serum albumin (BSA) (30 pg) was added as carrier protein. The cell lysate was centrifuged at 10,000 X g for 10 min to obtain a post-mitochondrial supernatant. The supernatant was transferred into a siliconized Eppendorf tube with 30 pg of lysozyme added as carrier protein, adjusted to 500 mM KCl, and loaded on a Pharmacia HR 10/30 Superose-12 FPLC column. The column was previously washed with buffer H(500) (20 mM HEPES, pH 7.5, 500 mM KCl, 2 mM EDTA, 50 mM NaF, 7 mM 2-mercaptoethanol, 0.5 mM phenylmethylsulfonyl fluoride, 10% glycerol) and eluted at 0.3 ml/min. Fractions (0.5 ml) were collected in siliconized Eppendorf tubes supplemented with 100 pg of lysozyme, quickly frozen, and stored at -70 "C. The fractions containing eIF-2 complex (fractions [22][23][24] or free eIF-Pa subunit (fractions 27-28) were determined in separate experiments by fractionation of the COS-1 cell lysate together with HeLa cell lysate (2 X 10' cells) and immunoblotting with affinity-purified antibodies to eIF-201 and eIF-2y (17). Such an elution profile of eIF-2 and free eIF-fa subunits from the Superose-12 column is illustrated in Fig. 1. For the routine preparation of eIF-2 complexes and free eIF-2a subunits, transfected cells from 10 plates (-2 X 10' cells total) were used, and no HeLa cell lysate was added.
Purificution of Overexpressed eZF-2%-The initial steps of purification of eIF-2a subunits were identical to those described above. Peak fractions from the Superose-12 column of free eIF-Pa were pooled, supplemented with BSA to a final concentration of 0.3 mg/ ml, and the salt concentration was reduced to 100 mM by dilut,ion with H(0) buffer (identical to H(500) but lacking KC1). The diluted sample was then immediately applied to a Pharmacia HR 5/5 MonoQ FPLC column previously equilibrated with H(100) buffer. The column was washed with H(100) buffer and then developed with a linear gradient of 100-500 mM KC1 in the same buffer. eIF-2a eluted at 260 mM KC1 as determined by immunoblotting with affinity-purified anti-eIF-2n antibody (17). The eIF-2a subunit was >90% pure as determined by SDS-PAGE and staining with Coomassie Blue (results not shown). Its protein concentration was determined by comparison of stained gel band intensities with those generated by known amounts of purified eIF-2 (18) and by immunoblotting.
Zn Vitro Phosphorylation of Overexpressed eIF-2a"Purified free eIF-2cu subunits from the MonoQ column were phosphorylated by HRI or DAI kinase in the presence or absence of purified HeLa eIF-2 (18). HRI was highly purified through the last step of the procedure described by Jackson and Hunt followed by passage through a MonoS column to remove casein kinase I1 activity (19); DAI was highly purified through the MonoS step as described (20) and was kindly provided by Dr. M. Matthews (Cold Spring Harbor Laboratory). The concentration of both kinases was limited to that just sufficient to phosphorylate half of the HeLa eIF-2 complex (1 pmol) under the conditions used. Phosphorylation by HRI was carried out in 25 pl of reaction mixtures containing 25 mM HEPES, pH 7.4, 100 mM KC1, 1 mM MgCl,, 0.3 mg/ml BSA, 50 p~ [y-:"P]ATP (>1 X lo4 cpm/ pmol), and purified eIF-2a and/or HeLa eIF-2 as indicated in Fig. 2. After 15 min at 30 "C, the reaction was quenched by the addition of 4 X SDS buffer and the extent of phosphorylation was monitored by 10% SDS-PAGE (21) and autoradiography.
Gel Electrophoresis and Zmmunoblotting-SDS-PAGE was performed according to Laemmli (21). Two-dimensional gel electrophoresis using either IEF/SDS-PAGE or NEPHGE-SDS-PAGE followed essentially the procedures described (16,22). Minor changes are as indicated in the figures. The identification of spots corresponding to eIF-2 subunits was by immunoblotting as described below.

RESULTS
Interaction of Free eIF-2a Subunits with Kinases-Overexpression of variant forms of e1F-2~1, where Ala is substituted for Ser4' (2a-AS) or Ser5' (20-SA), stimulates translation of plasmid-derived mRNAs after transfection of COS-1 monkey cells (11). A possible mechanism whereby these alanine variant forms stimulate protein synthesis is through direct inhibition of DAI kinase. Peptide analogues of phosphorylation sequences where Ser is replaced by Ala often function as specific inhibitors of their kinases (26). Since eIF-2a is overexpressed in transiently transfected COS-1 cells and accumulates to levels 10-20-fold greater than that of the endogenous eIF-2 trimeric complex (11), the excess free subunit of the factor may function as an inhibitor of DAI or other eIF-2 a kinases. In order to test directly this possibility, free 2a-SS, 2a-AS, and 2a-SA subunits were purified from transiently transfected cells. Cell proteins were fractionated as described under "Materials and Methods" by FPLC chromatography on a Superose-12 column to separate endogenous COS cell eIF-2 from the free human eIF-2a subunit based on their difference in size. As shown in Fig. 1, the eIF-2 complex and free eIF-2a subunit were identified in different fractions by immunoblotting with affinity-purified antibodies against eIF-2a and eIF-27.
The wild-type and variant eIF-2a subunits were used to determine whether the free subunit is a substrate for eIF-20 kinases and whether it directly inhibits phosphorylation of the eIF-2 complex. We performed in vitro phosphorylation reactions as described under "Materials and Methods," by addition of HRI to reaction mixtures containing the partially purified free eIF-2a subunit prepared by Superose-12 chromatography. Reaction conditions were optimized for HeLa eIF-2, but an unknown potent HRI inhibitor in the eIF-2a fractions inhibited both the autophosphorylation of HRI and the phosphorylation of eIF-2. Comparable Superose-12 fractions from control lysates of COS-1 cells transfected with a vector (pD61) lacking an eIF-Pa insert also inhibited the kinase (data not shown), suggesting that the eIF-Pa subunit is not responsible for the inhibition. For this reason, the free eIF-2a subunit from Superose-12 was further purified by using a MonoQ column as described under "Materials and Methods.'' This procedure removes the HRI inhibitor and generates eIF-2a preparations that are >90% pure (results not shown). cell lysate (no HeLa lysate carrier), immunoblotted with affinity-purified eIF-2n antibodies and visualized with a second antibody conjugated with alkaline phosphatase as described under "Materials and Methods." ated a t detectable levels, consistent with prior results (10).
Next, we tested whether the free eIF-Pa forms might inhibit HRI phosphorylation of the eIF-2 complex. The effects of the wild-type and variant forms were monitored in reaction mixtures containing HeLa eIF-2 substrate, a limiting amount of HRI, and varying amounts of the free eIF-2a subunits. The relative amounts of HeLa eIF-2 and the purified eIF-2a subunits were carefully determined by comparison of Coomassie Blue-stained band intensities and immunoblot intensities. As shown in Fig. 2A (lanes 10, 11, 14, 15, 18, and 19), none of the free 2a-subunits inhibits the phosphorylation of HeLa eIF-2 or the autophosphorylation of HRI, even at up to 8-fold molar excess. The intensity of radioactivity in the eIF-2a band approximates the sum of the intensities of eIF-2 alone (lane 3 ) plus the free subunit alone.
In another set of experiments, the same fractions of free eIF-2a subunits were tested as substrates or inhibitors of DAI activated with poly(I).poly(C) (Fig. 2B). In this set of phosphorylation reactions, the HeLa eIF-2 purified on MonoQ and Monos columns was contaminated with a potent inhibitor of DAI activation. When HeLa eIF-2 was added to DAI prior to its activation with dsRNA, the autokinase and eIF-2a kinase activities of DAI were completely inhibited (lane 18). However, when dsRNA is added before the eIF-2 substrate, no inhibition was observed, suggesting that the inhibitor prevents the activation of DAI by interfering with the   Overexpressed eIF-2a Subunits Exchange into Endogenous eIF-2 Complexes-An alternative mechanism to explain the stimulation of protein synthesis by 2a-AS and 2cu-SA is the incorporation of the variant forms into the cell's eIF-2 complexes by de novo synthesis and assembly and by exchange into preexisting complexes. Since transiently transfected cells are no longer growing exponentially when overproduction of eIF-2a occurs, the contribution of de novo assembly may be relatively minor compared to exchange into endogenous eIF-2 complexes. A strategy to determine if exchange occurs is to pulse-label transfected cells with ["S]methionine and then measure the specific radioactivities of the a-and y-subunits in the endogenous eIF-2 complex previously separated from free eIF-Za subunits. Radioactivity in the y-subunit is a measure of de novo synthesis and assembly into eIF-2 during the pulse, and a comparable amount of radioactivity is expected for the other subunits if exchange does not occur. The specific activity of the free, overexpressed eIF-2a subunit will be much higher since its rate of synthesis is greater and there is little or no endogenous free subunit to dilute its specific radioactivity. Therefore, if exchange occurs, the specific activity of the a-subunit in the eIF-2 complex should be higher than that of the other eIF-2 subunits. To determine whether or not exchange occurs, pulse-labeled COS-1 cells transiently expressing eIF-2a were lysed and the eIF-2 complex was partially purified by FPLC Superose-12 chromatography as described previously for the preparation of the free eIF-2cu subunit. Three fractions containing the trimeric eIF-2 complex were pooled and analyzed by NEPHGE-SDS-PAGE which separates the a-and y-subunits from other COS-1 proteins (Fig. 3). In control cells transfected with pD61 which lacks eIF-2a cDNA, the intensities of the spots for eIF-2y and eIF-2a are both very weak, as expected. However, in analyses of fractions containing the eIF-2 complex from 2a-SS, 2a-AS, and 2a-SA transfected cells, the specific radioactivity of the a-subunit is 10-20-fold higher than that of the y-subunit as determined by visual inspection. We conclude that very extensive exchange of the eIF-2a subunit occurs between the complex and free forms. Thus, in the case where a variant form of eIF-2a is overexpressed, essentially all of the endogenous eIF-2 complex in the transfected cells would contain the variant form of the subunit. With the 2a-SA mutant, it seems highly likely that such eIF-2 complexes cannot be phosphorylated by HRI or DAI, since the free form is not phosphorylated, and that protein synthesis would continue uninhibited even in the presence of activated eIF-2a kinases. However, with the 2a-AS variant, it is possible that such eIF-2 complexes are readily phosphorylated. If so, why does overexpression of the 2a-AS also reverse the inhibition of protein synthesis caused by the action of DAI?
Analysis of Double Variant Forms of eIF-2a in COS-1 Cells-In order to shed more light on the effects of mutating the Ser codons at positions 48 and 51 of eIF-Za, two double mutations of the cDNA were constructed. The eIF-2a cDNA was altered at both codons 48 and 51 by site-directed mutagenesis as described under "Materials and Methods," so that it expresses either an Ser4RAla-Ser"Ala form (2a-AA) or an Ser4'Ala-SerS1Asp form (2a-AD). These double variants serve to evaluate whether or not eIF-Pa is phosphorylated a t both Ser4R and Ser", as proposed recently (12), and to assess their ability to stimulate or inhibit protein synthesis.
The effect of expression of the various variant eIF-20 forms was determined by cotransfection with a reporter gene which expresses DHFR (pD61). Expression of DHFR from mRNA transcribed from pD61 is inefficient due to DAI kinase activation in transfected cells (7). pD61 was cotransfected with  (lanes 1-7). To determine the amount of eIF-2a expressed from the various eIF-2a expression plasmids, equivalent amounts of trichloroacetic acid-precipitable radioactivity (1 X lo6 cpm for lanes 8-12; 10 X lo6 cpm for lanes 13-1 5 ) were immunoprecipitated with a rabbit antiserum specific for HeLa eIF-2a as described (27). Parallel plates were harvested and total RNA was isolated and analyzed by RNA blot hybridization to a DHFR-specific probe, as well as to an actin probe. Band intensities on autoradiograms were quantitated by using an Ultroscan laser densitometer (model 2202; CKB Instruments, Inc., Rockville, MD).  12) is equally well synthesized. Northern blot analysis using a DHFR-specific probe which hybridizes to the 3' end of the eIF-Pa mRNA shows equal amounts of mRNA in each of these transfected cell populations (Fig. 4, bottom panel), demonstrating that these variants are translated with equal efficiency. In contrast, expression of the 2a-SD (lune 14) and 2a-AD (lune 15) is significantly reduced. For these immunoprecipitation reactions (lunes [13][14][15], 10 times the amount of trichloroacetic acid-precipitable counts were immunoprecipitated in order to detect eIF-2a synthesis. Quantitation of the band intensities after subtraction of the endogenous level of eIF-2a synthesis (lane 13) demonstrates that Pa-SD synthesis is reduced 200-fold compared to wild-type eIF-Za, consistent with results reported previously (ll), and 2a-AD synthesis is reduced 35-fold. Quantitation of eIF-2a mRNA in the transfected cells shows a 4-fold reduction for 2a-SD and a 2.5-fold reduction for 2a-AD compared to 2a-SS. This reduction in mRNA level was reproducible in separate experiments and may result from a secondary effect of inhibition of protein synthesis in cells expressing 2a-SD or 2a-AD. Corrected for the cellular levels of eIF-2a mRNA, the translation of 2a-SD late in transfection is reduced 50-fold compared to wild-type. Significantly, translation of 2a-AD is reduced only 14-fold compared to wild-type. Thus, mutation of the Ser to Ala codon for residue 48 decreases approximately 3.5-fold the inhibitory effect of the Asp" residue on the translation rate.
The effect of wild-type and variant eIF-Pa expression on DHFR translation was examined by analysis of DHFR synthesis detected in the cell extracts. DHFR is detected as a 20-kDa protein by SDS-PAGE of lysates from pD61-transfected cells (Fig. 4, lanes 2-5) compared to cells that did not receive DNA (lane 1). Cotransfection with either variant 2a-AS or Pa -SA (lanes 3 and 4 ) stimulates DHFR synthesis compared to the wild-type 2a-SS (lane 2 ) , as previously observed (11). Cotransfection with the 2a-AA variant also stimulates DHFR synthesis to a similar degree as that observed with either of the single Ser to Ala variants (lane 5). Since similar levels of DHFR mRNA are present in all of these cotransfected cells (Fig. 4, bottom punel), the differences in DHFR synthesis result from differences in translational efficiency. These results show that prevention of phosphorylation at both residues 48 and 51 does not cause additional stimulation of protein synthesis over that observed with either Ala replacements alone. It was not possible to detect DHFR synthesis in cells cotransfected with either the 2a-SD (lune 7) or the 2a-AD (lane 6 ) variants, suggesting that phosphorylation at Ser" I (mimicked by the Asp residue) is sufficient to inhibit protein synthesis.
The phosphorylation states of the overexpressed eIF-Pa subunit forms were evaluated by immunoblotting to assess the effects of the various substitutions at Ser. The free eIF-2a subunits were partially purified by FPLC Superose-12 chromatography, then subjected to VS-IEF-PAGE and immunoblotting. Overexpression of either the wild-type or Ala variant cDNAs results in the accummulation of large amounts of free Sa-subunits compared to that in pD61-transfected cells lacking eIF-2a cDNA (Fig. 5). The ratio of phosphorylated (right band) to non-phosphorylated (left band) protein is comparable for the wild-type and all Ala variant forms, and lies in the range of 0.1-0.15. The presence of low amounts of phosphorylated eIF-Pa in the transfection with the 2a-AA cDNA requires explanation, since phosphorylation of this protein is prevented at both residues 48 and 51. It is unlikely that non-transfected cells contribute significant amounts of free phosphorylated 2a-subunit since very little free eIF-2a is detected in cells that receive pD61 alone. It is more likely that the phosphorylated protein results from wild-type eIF-Za that exchanges out of the endogenous COS cell eIF-2 complex in cells that overexpress the 2a-subunit. Evidence for such exchange has been presented in Fig. 3, and also is seen in Fig.  5, lanes 6 and 7, where moderately intense non-phosphorylated eIF-Sa bands are seen that must be due to exchanged endogenous COS cell eIF-Za subunits. A third, as yet unidentified, phosphorylation site also might contribute to the overall phosphorylation pattern of the Ala mutants. Since the phosphorylation pattern is comparable for cells transfected with both 2a-SA and 2a-AA, it is apparent that the Za-SA subunit is not appreciably phosphorylated on Ser4'.
Analysis of the Pa-SD and 2a-AD forms shows some nonphosphorylated eIF-2a and a new band with an isoelectric point slightly less acidic than phosphorylated eIF-2a (Fig. 5), consistent with an Asp replacement of Ser. The intensity of the Pa-AD band is much greater than the corresponding band (lane 15) in Fig. 4. This is due to the fact that Fig. 5 measures the total accumulation of free eIF-2a over about 2 days, whereas Fig. 4 measures the rate of eIF-2a synthesis late in transfection when the effects of accumulated variants are most apparent. The lower abundant protein corresponding to non-phosphorylated eIF-Za must be due to COS cell eIF-2a subunits, by an exchange mechanism and possibly from nontransfected cells. No band more acidic than 2a-AD or 2a-SD is detected for either variant form. This result shows that neither 2a-SD nor 2a-AD is phosphorylated to a significant extent. The importance of this observation in terms of the number of phosphorylations required for translational repression is discussed below. Analysis of the in Viuo Phosphorylation State of eIF-2-Since the free eIF-Za variant forms do not inhibit HRI or DAI activities in vitro, and since they freely exchange into endogenous COS cell eIF-2 complexes, it is highly likely that the Ala and Asp variants affect translation by converting the endogenous eIF-2 complex into a variant form. It is therefore important to evaluate the extent of phosphorylation of the Za-subunit in the eIF-2 complex. This is not a trivial matter because only up to 30% of the cells are actually transfected, which results in about 70% of the eIF-2 in a cell preparation coming from unaltered (wild-type) eIF-2. However, since the specific activity of the Pa-subunit is high because of overexpression, analysis of radioactivity rather than mass will reflect the status of eIF-Za in essentially only the transfected cell population. Cells were cotransfected with eIF-Za wild-type or mutated cDNAs and with pD61 (to activate DAI in order to elicit phosphorylation) and labeled with [3sS]methionine for 8 h at 40-h post-transfection. Subsequently, the eIF-2 complexes were separated from free, overproduced Za-subunits as described above. The fractions containing the eIF-2 complexes were then subjected to two-dimensional IEF/SDS-PAGE to separate and identify phosphorylated and nonphosphorylated 20-subunits. As shown in Fig. 6, the more intense Pa-subunit for 2a-SS, 2a-AS, and 2a-SA confirms the overproduction and exchange of these proteins into the complex. Moderate phosphorylation of 2a-SS and Za-AS is observed. This suggests that replacement of Ser for Ala at position 48 does not significantly influence the extent of phosphorylation of Ser at position 51. In contrast, no distinct spot of enhanced intensity corresponding to the phosphorylated form of Za-SA or 2a-AA is detected, as expected.
In the case of 2a-SD and 2a-AD, a new spot, slightly less acidic than phosphorylated eIF-Ba, is seen which corresponds to the variant protein. It is apparent that the variant eIF-Za subunits have exchanged into endogenous eIF-2. Since the variant form of the eIF-2 complex mimics the structure of the phosphorylated factor, it is expected to inhibit all of the eIF-2B in the cell since eIF-2B is thought to be present at much lower cellular levels than eIF-2 (28). This is indeed observed in Fig. 4, where strong repression of DHFR synthesis occurs with these variants.

DISCUSSION
The inhibition of protein synthesis by the phosphorylation of eIF-2 is the best characterized translational control mechanism operating at the level of covalent modification of translational components. Phosphorylation of the a-subunit correlates with repression in numerous physiological conditions; a convincing change in factor activity has been demonstrated in vitro; and two highly specific protein kinases have been identified and their control has been characterized (reviewed in Ref. 3). Evidence that eIF-2a phosphorylation is the actual cause of translation inhibition in vivo has been more difficult to obtain. The fact that the activation of two different protein kinases with high specificity for eIF-2a both lead to inhibition of initiation is strongly suggestive, but there is always uncertainty that other as yet unidentified substrates may be critically important for the repression. To resolve this ambiguity, variant forms of eIF-2a were expressed in transiently or long term transfected cells in order to alter the extent of translation inhibition upon activation of the eIF-2a kinase, DAI (11,27). The findings that expression of the 2a-SD variant inhibits total protein synthesis whereas expression of the 2a-SA variant mitigates the inhibition provide the best evidence that eIF-2a phosphorylation truly causes translational repression i n vivo. However, an unexplained observation is that whereas the 2a-DS variant causes no inhibition (suggesting that phosphorylation at Ser4' does not lead to inhibition), the 201-AS variant is as active in decreasing translation inhibition as is the 2a-SA variant in cells where DAI is activated (11,27). Therefore, the work described here was undertaken to increase our understanding of how the overexpressed eIF-2a variant proteins actually function to alter translation rates.
An important first step was to determine whether or not the free eIF-2a subunit is itself active in these cells. We obtained highly purified, overexpressed eIF-2a subunits from transiently transfected COS-1 cells by employing rapid FPLC chromatographic methods. The 2a-SS wild-type and 2a-AS variant subunits serve as substrates for HRI and DAI, although in the case of HRI they are considerably less active than the purified HeLa eIF-2 complex. That these proteins are phosphorylated at all is of interest, since the 2a-subunit obtained by dissociation of eIF-2 with denaturing buffer is not a substrate. Apparently, the conformation of the protein is important for its recognition by the protein kinases, suggesting that the accumulated, overexpressed subunits possess a conformation comparable to that in the eIF-2 complex. Although 2a-SS and Sa-AS are poor substrates for HRI and 2a-SA is not phosphorylated appreciably, none of these proteins inhibits the phosphorylation of the eIF-2 trimeric complex, even when tested in 81 molar excess. It therefore is very likely that the free mutant eIF-2a subunit, which accumulates up to 10-20 mol/mol of endogenous COS cell eIF-2, does not inhibit activated DAI in transfected cells (11,27). Our biochemical results are consistent with the finding that DAI assayed i n vitro in transfected cell lysates is active on exogenous eIF-2 substrate (27). In effect, no regulatory role for the free eIF-2a subunits was demonstrated.
The possibility remained that the active form of the variant proteins involves the endogenous eIF-2 trimeric complexes, generated by an exchange mechanism that replaces the endogenous 2a-subunit with the variant form. Our results indicate that an efficient exchange process occurs i n vivo for 2a-SS and all of the variants. The precise rate of exchange is not known, but greater than 90% exchange must occur within the 8 h (and sometimes 5 h) labeling periods employed in the experiments. An active exchange implies that eIF-2 may be in equilibrium with all three of its subunits. However, the failure to readily detect free eIF-2a subunits in cells transfected only with pD61 and not with the eIF-2a cDNA expression vector suggests that significant levels of free subunits do not occur in normal cells. Exchange of eIF-2a into the eIF-2 complex was unexpected because biochemical characterization of eIF-2 indicates a very strong complex that resists dissociation into subunits. The exchange phenomenon is consistent with, but does not prove, a mechanism of action of eIF-2 involving a transiently free form for the eIF-Sa subunit, as suggested by Gupta and co-workers (29).
A likely mechanism of action of the variant forms is to alter the extent of phosphorylation of the eIF-2 complex. eIF-2 containing the 2a-SA variant is expected to resist phosphorylation and thereby evade the inhibition potentially caused by DAI. To explain the similar effects of the 2a-AS variant, we hypothesized that although the 2a-AS protein can be phosphorylated on Sersl, the relative activities of the DAI and phosphoprotein phosphatases are such that the equilibrium with the variant would be shifted to less phosphorylation, thereby resulting in an eIF-2 that does not repress translation. However, examination of the extent of 2a-AS subunit phosphorylation in the eIF-2 trimeric complex by IEF/SDS-PAGE (Fig. 6) indicates an amount of phosphorylation comparable to that of the wild-type subunit, which does not reverse the effects of DAI so effectively. Two features of this experiment complicate the interpretation of the results. First, it is necessary to separate the free 2a-subunits from the eIF-2 complex, a step involving non-denaturing conditions during which the extent of phosphorylation can change. The lysate preparation and Superose-12 chromatography were performed in buffers designed to inhibit both kinase and phosphatase action, but it is difficult to rule out some changes in the extent of phosphorylation of the partially purified eIF-2 complex during purification and analysis. Second, only 20-30% of the cell population is transfected, resulting in only a small fraction of eIF-2 complexes containing the variant form. However, since transfected cells contain high specific radioactivity in eIF-2a, and thus contribute the major part of the radioactivity in the eIF-2a spots, the radioactivity patterns reflect phosphorylation primarily in transfected cells. The results strongly suggest that the 2a-AS mutant stimulates translation by a mechanism other than altering the extent of phosphorylation.
A possibility is that phosphorylation of eIF-2a occurs on two sites, namely Ser4' and Ser", both essential for inhibition as argued recently by Kramer (12). If this were true, then both the 2a-AS and Sa-SA variants would resist the double phosphorylation and thereby stimulate protein synthesis, as is observed (11). However, the data reported here do not support the view that simultaneous phosphorylation on both Ser4' and Sersl occurs. Analysis by VS-IEF-PAGE of the free 2a-subunit from cells transfected with wild-type 2a-SS shows extensive monophosphorylation but no discernable band corresponding to a second phosphate (Fig. 5 ) . A comparable amount of monophosphorylated eIF-2a is seen when the variant 2a-AA is expressed, even though the variant subunit cannot be phosphorylated at positions 48 and 51. The phosphorylated 2a-subunit seen with 2a-AA is very likely due to the wild-type COS cell %-subunit that has exchanged out of the endogenous eIF-2 complex; a small contribution by phosphorylation at a third site cannot be ruled out, however. The single Ala variant, Pa-SA, shows a phosphorylation pattern essentially identical to the 2a-AA variant, indicating that a negligible amount of phosphorylation a t Ser4' occurs on this protein. Furthermore, the variants 2a-SA, 2a-AS, and 2a-AA prevent the inhibitory action of DAI equally effectively. Even clearer evidence comes from the analyses of the Asp variants. The 2a-SD variant strongly inhibits DHFR synthesis (Fig.  4), yet VS-IEF-PAGE gels (Fig. 5) show no evidence for phosphorylation of the variant subunit. The 201-AD variant, which accumulates to a higher level, also shows no sign of phosphorylation (Fig. 5). Variant 2a-AD inhibits DHFR synthesis about 14-fold compared to the wild-type 2a-subunit at late times post-transfection (Fig. 4), yet the former cannot be phosphorylated at position 48. It is therefore highly likely that Ser4' phosphorylation is not involved in the repression of translation by DAI or HRI.
The variant Sa-AD is somewhat more highly expressed than 2a-SD (but expressed much lower than the other eIF-2a forms) and is a 3.5-fold weaker inhibitor of DHFR synthesis than is 2a-SD. Since neither 2a-AD nor 2a-SD is phosphorylated, it appears that the presence of Ala a t position 48 decreases the inhibitory action of Asp at position 51. This suggests that 2a-AS phosphorylated on Ser51 might not repress protein synthesis as effectively as the phosphorylated wild-type protein. We speculate that Ala at residue 48 either reduces the affinity of eIF-2.GDP for eIF-2B, resulting in a failure to sequester eIF-2B, or that the variant eIF-2 has an enhanced rate of guanine nucleotide exchange that may not require catalysis by eIF-2B. To test these postulates, nearly pure variant forms of eIF-2 are required, but are not readily obtained from transiently transfected cell populations since only a minor fraction of the cells are altered. In order to test the biochemical behavior of variant forms of eIF-2, it will be necessary to purify variant eIF-2 from long term transfected cell lines overexpressing variant eIF-2a subunits, or alternatively to purify the various subunits expressed in bacteria and reconstitute variant eIF-2 complexes.