The γ134.5 Protein of Herpes Simplex Virus 1 Has the Structural and Functional Attributes of a Protein Phosphatase 1 Regulatory Subunit and Is Present in a High Molecular Weight Complex with the Enzyme in Infected Cells*

The carboxyl-terminal domain of the γ134.5 protein of the herpes simplex virus 1 binds to protein phosphatase 1α (PP1) and is required to prevent the shut-off of protein synthesis resulting from phosphorylation of the α subunit of eIF-2 by the double-stranded RNA-activated protein kinase. The corresponding domain of the conserved GADD34 protein homologous to γ134.5 functionally substitutes for γ134.5. This report shows that γ134.5 and PP1 form a complex in the infected cells, that fractions containing this complex specifically dephosphorylate eIF-2α, and that both γ134.5 and GADD34 proteins contain the amino acid sequence motif common to subunits of PP1 that is required for binding to the PP1 catalytic subunit. An oligopeptide containing this motif competes with γ134.5 for binding to PP1. Substitution of Val193 and Phe195 in the PP1-binding motif abolished activity. These results suggest that the carboxyl-terminal domain of γ134.5 protein has the structural and functional attributes of a subunit of PP1 specific for eIF-2α, that it has evolved to preclude shut-off of protein synthesis, and that GADD34 may have a similar function.

The ␥ 1 34.5 gene of herpes simplex virus 1 (HSV-1) 1 encodes two functions. The first enables the virus to replicate in vivo and particularly to multiply and spread in the central nervous system of experimental animal systems (1,2). This function appears to map throughout the coding domain of the gene (3,4). The second blocks the shut-off of protein synthesis resulting from phosphorylation of the ␣ subunit of the translation initiation factor eIF-2 by the double-stranded RNA-activated protein kinase (PKR). This function maps in the 3Ј-terminal domain of the 263-codon gene (5,6). Earlier studies have shown that in HSV-1-infected cells PKR is activated but that in cells infected with wild-type virus or virus carrying in-frame deletions of the 5Ј-terminal coding domain of the gene eIF-2␣ was not phosphorylated (7). In subsequent studies (8) we have shown that the ␥ 1 34.5 protein interacts with the protein phos-phatase 1␣ (PP1). Indeed, infected cells contain a phosphatase activity that specifically dephosphorylates eIF-2␣ at a rate 3000-fold greater than that measured in uninfected cells. This phosphatase activity is inhibited by inhibitors of PP1. The hypothesis that emerged from these studies is that transcription of complementary sequences of the HSV-1 DNA results in the formation of double-stranded RNA, that PKR is activated in cells infected with both wild-type and mutant viruses, and that a domain of the ␥ 1 34.5 protein binds PP1 and redirects its activity to dephosphorylate eIF-2␣. This report centers on one aspect of this hypothesis: we show that in infected cells ␥ 1 34.5 protein is a component of a multi-protein, cytoplasmic complex containing PP1 and that the interacting domain of the ␥ 1 34.5 protein is near the carboxyl terminus of the protein and has an amino acid motif shared with accessory proteins or subunits interacting with the catalytic subunit of PP1.
Relevant to this report are the following: (i) The HSV-1 ␥ 1 34.5 gene encodes a protein consisting of a 159-amino acid amino-terminal domain, the amino acids Ala-Thr-Pro repeated 5-10 times depending on virus strain, and a 74-amino acid carboxyl-terminal domain (9 -11). The gene maps in the inverted repeat sequence ab and bЈaЈ flanking the unique long (U L ) sequence, between the terminal a sequence and the gene encoding the regulatory infected cell protein number 0. Earlier studies described a series of mutants in which in-frame deletions truncated the amino-terminal domain or prematurely terminated the translation of the carboxyl-terminal domains.
The key observation was that protein synthesis was shut off after the onset of synthesis of viral DNA in cells infected with mutants either lacking the entire ␥ 1 34.5 coding domain or that are unable to express the carboxyl-terminal domain (6). (ii) The ␥ 1 34.5 gene is conserved in very few herpesviruses, suggesting that other herpesviruses have evolved different methods for blocking the consequences of double-stranded RNA accumulation in infected cells. Homologs of the ␥ 1 34.5 carboxyl-terminal domain have been found, however, in the African swine fever virus (12), a DNA virus belonging to the unrelated, Iridovirus family, and in the corresponding domain of a highly conserved mammalian protein known as GADD34 (growth arrest and DNA damage protein 34) (13)(14)(15)(16). The GADD proteins are induced in cells subjected to growth arrest as a consequence of serum deprivation or damage to their DNA or in the course of differentiation (14 -16). It has been reported that overexpression of GADD34 results in apoptosis (17). It is of interest, however, that the carboxyl-terminal domain of the mouse GADD34 protein substituted for the HSV-1 ␥ 1 34.5 carboxylterminal domain in blocking the shut-off of protein synthesis (18). This observation suggests that the shared sequences are sufficient to block the shut-off of protein synthesis and that the corresponding GADD34 protein domain may perform upon induction a similar function. The homologous domains of known GADD34 proteins and of the HSV-1 and HSV-2 and African swine fever virus homologs are shown in Fig. 1. (iii) The catalytic subunit of PP1 is a highly conserved protein with a M r of 38,000 (19). PP1 exists in holoenzyme complexes with noncatalytic or regulatory components that modulate catalytic activity or restrict the subcellular localization of the catalytic subunit. In essence, PP1 is regulated by different cellular proteins. Many of such regulatory proteins have been described (20 -25), and these complexes have diverse functions within the cell (19,26). While this work was in progress, it has been reported that these subunits share an amino acid sequence required for binding to the catalytic subunit of PP1 (27,28).

MATERIALS AND METHODS
Cells and Viruses-The HeLa and SK-N-SH cell lines were obtained from American Type Culture Collection and propagated in Dulbecco's modified Eagle's medium supplemented with 5% (HeLa) or 10% (SK-N-SH cells) fetal bovine serum, respectively. The rabbit skin cell line originally obtained from J. McClaren was propagated in Dulbecco's modified Eagle's medium supplemented with 5% fetal bovine serum. HSV-1(F) is the prototype HSV-1 strain used in this laboratory (29). Recombinant virus R3616 lacks 1000 base pairs from the coding region of the HSV-1(F) ␥ 1 34.5 gene (1). In recombinant R8301, the 3Ј sequence encoding amino acids 206 -263 was deleted from the carboxyl-terminal domain of the ␥ 1 34.5 gene (18). In recombinant viruses R4002, R931, R908, and R909, the sequences of the ␥ 1 34.5 encoding amino acids 1-30, 31-72, 72-106, and 107-146, respectively, were deleted (6). As previously reported, all recombinant viruses listed above except R3616 and R8301 preclude the shut-off of protein synthesis.
For infection, cells were exposed to the indicated virus for 2 h. at 37°C. The inoculum was then replaced with medium 199V consisting of mixture 199 supplemented with 1% calf serum.
Plasmids-In plasmid pRB4892 the coding domain of glutathione S-transferase (GST) is fused to the entire coding domain of PP1 except for the initiator methionine codon (8). In plasmid pRB4893 the coding domain of GST is fused to codons 146 -263 of the ␥ 1 34.5 gene (8). To construct plasmid pRB4895, an oligonucleotide linker, AATTCCCAGC-ACGTGTACGTTTCTCGCCTCACGTCCGAGTACGTCACG, and its complement, TCGACGTGACGTACTCGGACGTGAGGCGAGAAACG-TACACGTGCTGGG, were cloned into the EcoRI and SalI sites of pGEX4T-1. In this plasmid, codons 190 -203 of the ␥ 1 34.5 gene were fused in frame to the coding sequences of GST. To construct plasmid pRB4894, an oligonucleotide linker, AATTCCACCTGGTGTCCGGA, and its complement, TCGACTCCGGACACCAGGTGG, were inserted into the EcoRI and SalI sites of pGEX 4T-1, yielding plasmid pRB4896. A 244-base pair DraIII-BspEI fragment encoding amino acids 205-263 was then isolated from plasmid pRB3207 (18) and ligated into the DraIII and BspEI sites of pRB4896.
To construct pRB4897, a BstEII-DraIII fragment encoding codons 28 -205 of ␥ 1 34.5 was amplified by polymerase chain reaction from plasmid pRB143 with primers CCACCCCGGCACGCTCTCTGT and CAGACCACCAGGTGGCGCACCCGGACGTGGGGCGATAAGCGCTC-CCGCGCGGGGGTC. The amplified polymerase chain reaction fragment with nucleotide changes (underlined) incorporated in the 3Ј end primer was inserted into the BstEII and DraIII sites of pRB143, resulting in plasmid pRB4897, which was sequenced to verify that the plasmid contained the desired mutations. In this plasmid, the codons encoding Val 193 and Phe 195 of ␥ 1 34.5 were replaced with those encoding Glu and Leu, respectively.
Expression of GST fusion proteins was induced by the addition of isopropyl ␤-D-thiogalactoside to cultures of Escherichia coli BL21 cells transformed with plasmid pRB4892, pRB4893, pRB4894, or pRB4895, followed by affinity purification of the fusion proteins from bacterial lysates on agarose beads conjugated with glutathione. Purified PP1 was obtained by cleavage of GST-PP1 fusion protein with thrombin (Sigma).
Construction of the Recombinant Virus R8321-Recombinant virus R8321 was constructed by cotransfection of the intact viral DNA of R3659 (30) with the plasmid pRB4897 on rabbit skin cells. In the parent virus R3659 (31), a 1-kilobase fragment from the coding sequences of ␥ 1 34.5 was replaced with the chimeric ␣27-tk gene. The progeny of the recombinant was selected and plaque-purified on 143 TK mutant cells in medium consisting of mixture 199V supplement with 100 g of bromodeoxyuridine/ml and 2% fetal calf serum. Preparation of viral stocks and titrations of infectivity were done with Vero cells.
Cell Lysates-HeLa cells either mock-infected or infected with viruses were harvested at 18 h after infection, rinsed with phosphatebuffered saline, resuspended in lysis buffer containing 10 mM Hepes (pH 7.6), 250 mM NaCl, 10 mM MgCl 2 , 1% Triton X-100, 0.5 mM phenylmethylsulfonyl fluoride, and 2 mM benzamidine, stored on wet ice for 30 min, and subjected to low speed centrifugation to remove nuclei. The supernatant fluids (S10 fractions) were saved for analysis.

FIG. 2.
A, schematic representation of the genome structure and sequence arrangements of HSV-1(F) and of the ␥ 1 34.5 deletion mutants. The top line represents the two covalently linked components of HSV-1 DNA, L and S, each consisting of unique sequences (U L and U S , respectively) flanked by inverted repeats (42,43). The reiterated sequences flanking U L , designated ab and bЈaЈ, are each 9 kilobase pairs in size, whereas the repeats flanking U S , designated aЈcЈ and ca, are 6.3 kilobase pairs in size (43). The location of the ␥ 1 34.5 gene is shown in an expanded portion of the inverted repeat sequences b and bЈ. The shaded bar and the arrow indicate the coding region and the direction of transcription of ␥ 1 34.5, respectively. Because the b sequence is repeated in an inverted orientation, the HSV-1 genome contains two copies of the SDS-polyacrylamide gels, transferred to a nitrocellulose sheet, blocked with 5% nonfat milk, reacted with either anti-␥ 1 34.5 antibody (9) or anti-PP1 antibody (Upstate Biotechnology Inc. Lake Placid, NY), rinsed, and reacted with goat anti-rabbit immunoglobulin coupled to alkaline phosphatase as recommended by the manufacturer (Bio-Rad).
[ 35 S]Methionine Labeling-Replicate cultures of SK-N-SH cells were either mock-infected or exposed to 10 plaque-forming units (pfu) of HSV-1(F) or one of the ␥ 1 34.5 mutants per cell at 37°C. At 14 h after infection, cells were overlaid with 1 ml of medium lacking methionine but supplemented with 50 Ci of [ 35 S]methionine (specific activity, Ͼ1000 Ci/mmol; Amersham Pharmacia Biotech) and reincubated for 1 h. The cells were then harvested, solubilized, subjected to electrophoresis in denaturing polyacrylamide gels, and transferred to a nitrocellulose sheet and subjected to autoradiography as described previously (5).
Determination of eIF-2 Phosphatase Activity-Rabbit reticulocyte eIF-2, purified as described previously (32), was phosphorylated with partially purified, hemin-controlled translational repressor as described elsewhere (33,34) to yield phosphorylated eIF-2␣ and eIF-2␤ (1.0 and 0.7 mol/mol of eIF-2, respectively). This preparation also contains a phosphoprotein with a M r of 39,000. S10 fractions were prepared from lysates of HeLa cells infected with HSV-1(F) (7) and fractionated on a Superdex 200 column as described above. Aliquots (3.0 l) of the fractions were incubated with 1.2 pmol of phosphorylated eIF-2 and a final concentration of 0.8 mM ATP in a final volume of 4.0 l for 1.0 min at 34°C. Reactions were terminated with SDS and subjected to electrophoresis on 7% denaturing, polyacrylamide gels, followed by silver staining, drying, and autoradiography as described (8). The percentage of eIF-2(␣ 32 P) and eIF-2(␤ 32 P) that was dephosphorylated was determined by excising these bands and comparing their Cerenkov radiation to that of equivalent eIF-2(␣,␤ 32 P) that was not further incubated and was subjected to electrophoresis in parallel. 34.5 Protein-Earlier studies have shown that the carboxylterminal domain of the ␥ 1 34.5 protein (Fig. 1) is required to prevent the shut-off of protein synthesis, that it interacts with PP1, and that it is highly conserved among several viral and GADD34 proteins. The objective of this series of experiments was to determine the PP1-binding sites in the ␥ 1 34.5 protein.

Localization of Protein Phosphatase 1-binding Site(s) in
In the first series of experiments, purified GST-PP1 protein bound to beads was incubated with cell extracts prepared from HeLa cells infected with 20 pfu of wild-type virus or mutants carrying ␥ 1 34.5 genes from which various domains of the gene had been deleted (Fig. 2A). The proteins bound to GST-PP1 were solubilized, electrophoretically separated in denaturing gels, transferred to nitrocellulose, and reacted with antibody to the Ala-Thr-Pro repeat of the ␥ 1 34.5 protein (9). The results in FIG. 3. Association of PP1 with ␥ 1 34.5 protein in vitro. A, replicate HeLa cell cultures were harvested in lysis buffer containing 10 mM Hepes (pH 7.6), 250 mM NaCl, 10 mM MgCl 2 , 1% Triton X-100, 0.5 mM phenylmethylsulfonyl fluoride, and 2 mM benzamidine 18 h after mock infection or infection with 10 pfu of the indicated viruses. After 30 min on ice and low speed centrifugation to remove nuclei, the supernatant fluids were precleared with GST beads and then reacted with GST-PP1 beads at 4°C for 14 h. The proteins bound to beads were rinsed extensively, solubilized by boiling in disruption buffer containing 50 mM Tris-Cl (pH 7.0), 5% 2-mercaptoethanol, 2% SDS, and 2.75% sucrose, electrophoretically separated on denaturing 12% polyacrylamide gels, transferred to a nitrocellulose sheet, and reacted with rabbit polyclonal anti-␥ 1 34.5 serum (9). The positions of the full-length and truncated ␥ 1 34.5 are shown. B, an aliquot of GST-PP1 fusion protein bound to beads was reacted with 25 units of thrombin (Sigma) in phosphatebuffered saline at room temperature. After 12 h, the mixture was centrifuged in a tabletop centrifuge, and the supernatant fluid containing PP1 was then dialyzed against lysis buffer, reacted with GST or GST-␥ 1 34.5c chimeric proteins containing the amino acids (aa) 190 -203, 205-263, or 146 -263 bound to beads, and processed as described above. PP1 was detected with anti-PP1 antibody (Upstate Biotechnology, Inc.).  Fig. 3. The cell lysates were reacted with GST-PP1 bound to beads in the absence or presence of increasing amounts of the peptide for 14 h. The beads were then collected and rinsed extensively. The proteins bound to the beads were solubilized in disruption buffer, separated on a denaturing 12% polyacrylamide gel, and transferred to a nitrocellulose sheet (5). The ␥ 1 34.5 protein was detected with the antibody described by Ackermann et al.  In the second set of experiments, purified PP1 was reacted with beads carrying GST or GST fused to the sequence of amino acids 146 -263, 190 -203, or 205-263 of the ␥ 1 34.5 protein (Fig.  2B). The proteins bound to GST fusion proteins were solubilized, subjected to electrophoresis on a denaturing gel, and reacted with anti-PP1 antibody. The results in Fig. 3B indicate that PP1 bound strongly to the ␥ 1 34.5 peptide sequence encompassing amino acids 190 -203. In addition, PP1 bound less efficiently to the carboxyl-terminal domain of the ␥ 1 34.5 protein encompassing amino acids 205-263. PP1 did not bind to GST (Fig. 3B).
In the third series of experiments, a competition assay was done to test whether a synthetic test peptide containing amino acids 185-211 could block the interaction between intact ␥ 1 34.5 protein and PP1. To control for the specificity, we synthesized a second peptide containing the same amino acids but in a scrambled order. GST-PP1 bound to beads was mixed with cell lysates prepared from HeLa cells infected with HSV-1(F) in the presence or absence of the synthetic peptides. The protein complex was separated in a denaturing protein gel and processed for immunoblotting with the anti-␥ 1 34.5 antibody. As shown in Fig. 4, the test peptide competed with the ␥ 1 34.5 protein for binding to PP1, whereas the control peptide was not.
We conclude from these experiments that a PP1-binding site of ␥ 1 34.5 is located between amino acids 190 and 203. PP1 also interacts with the carboxyl-terminal domain of the ␥ 1 34.5 protein (amino acids 205-263) but with a lower affinity.  virus R8321 in which the conserved amino acids Val 193 and Phe 195 in the PP1-binding motif were mutated to Glu and Leu, respectively, and created the novel restriction site Eco47III. The inserted mutations were verified by sequencing of the plasmid and the presence of the restriction site in the viral DNA (data not shown).
Replicate cultures of human neuroblastma SK-N-SH cells were either mock-infected or infected with HSV-1(F), R3616, or R8321 at 10 pfu/cell. At 14 h after infection, the cells were labeled for 1 h with [ 35 S]methionine and then harvested, solubilized, subjected to electrophoresis in denaturing polyacrylamide gel, and transferred to a nitrocellulose sheet. The autoradiogram shown in Fig. 6A indicates premature shut-off of protein synthesis in cells infected with R8321 or R3616 but not in mock-infected cells or those infected with HSV-1(F).
To determine whether R8321 expressed the mutant ␥ 1 34.5 in SK-N-SH cells, the nitrocelluse sheet described above was reacted with the anti-␥ 1 34.5 antibody. As shown in Fig. 6B, the lysates of cells infected with HSV-1(F) or R8321 expressed full-length ␥ 1 34.5 protein. The decrease in the amount of the ␥ 1 34.5 protein made in cells infected with the R8321 mutant was expected because ␥ 1 34.5, as its name indicates, is made mostly after the onset of DNA synthesis, and its accumulation would be affected by the shut-off of protein synthesis. The antibody did not react with lysates of mock-infected cells or cells infected with the R3616. On the basis of these experiments, we conclude that Val 193 and Phe 195 in the PP1 biding motif of ␥ 1 34.5 are essential for the function of ␥ 1 34.5.
Cytoplasmic PP1 and ␥ 1 34.5 Protein Cofractionate in a Complex with an Apparent Molecular Weight of 340,000-The purpose of this experiment was to determine whether PP1 forms a complex with ␥ 1 34.5 protein in cells infected with wild-type virus. Replicate 150-cm 2 flask cultures of HeLa cells were exposed to 20 pfu of HSV-1(F)/cell. At 15 h after infection, the cells were harvested and lysed, and the S10 fraction was chromatographed on a Superdex 200 column. The fractions were then assayed for their ability to dephosphorylate eIF-2(␣P). Aliquots of these fraction were also solubilized, subjected to electrophoresis in denaturing gels, transferred to nitrocellulose sheets, and reacted with anti-PP1 and ␥ 1 34.5 antibodies, respectively, as described under "Materials and Methods." The results (Fig. 7) indicate that the eIF-2(␣P) phosphatase eluted as a single discrete component in fractions 22-24. This activity was relatively specific for eIF-2␣, because it had little or no effect on eIF-2(␤P) or on the 39-kDa phosphoprotein (Figs. 7 and 8B). A separate activity that dephosphorylated eIF-2(␤P) eluted in fractions 25-26. Comparison of the elution of eIF-2(␣ 32 P) phosphatase with that of known proteins (Fig. 7) yielded an estimated molecular weight of 340,000 for the eIF-2␣-specific phosphatase activity.
The results of assays of the fractions for PP1 and ␥ 1 34.5 proteins shown in Fig. 8A indicate that virtually all of ␥ 1 34.5 protein and a majority of PP1␣ cofractionated in fractions 22-24, exactly paralleling the elution of eIF-2(␣P) phosphatase (Fig. 7). These results were reproduced in a separate experiment in which smaller aliquots (30 l) of fractions 15-36 of Superdex 200 fractions were subjected to electrophoresis and reacted with the antibody to the ␥ 1 34.5 protein. The results were identical to those shown in Fig. 8A and showed no detectable ␥ 1 34.5 protein in fractions 27-36 (data not shown). Because fractions 22-24 constitute a minimum of the protein eluting from the column, as estimated by the A 280 profile (Fig.   FIG. 7.

Gel filtration analysis of cytoplasmic extracts from HSV-1(F)-infected cells on a Superdex 200 column.
HeLa cells were exposed to 20 pfu of HSV-1(F)/cell and harvested 15 h after incubation at 37°C. The S10 fraction (0.5 ml) prepared as described under "Materials and Methods" was chromatographed on a Superdex 200 column (1.0 ϫ 30 cm), and the fractions were assayed for their ability to dephosphorylate eIF-2(␣ 32 P) and eIF-2(␤ 32 P) (8,32). The smooth tracing represents the absorbance at 280 nm. The Superdex 200 column was calibrated by chromatographing 200 g of molecular size marker proteins individually using conditions that were identical to the chromatography of HeLa cell S10 fraction on this column, except the absorbance was monitored at 10-fold greater sensitivity. The molecular weight size markers and their M r s were: horse spleen apoferritin (465,000); aldolase (150,000); bovine serum albumin (69,000); ovalbumin (45,000); and rabbit reticulocyte thioredoxin (11, 600). 7), the results strongly suggest that the activated eIF-2(␣P) phosphatase in HSV-1-infected HeLa cells consists of a complex with a molecular weight of approximately 340,000 that contains the ␥ 1 34.5 protein and PP1 either alone but in multiple copies or in association with other proteins.
To characterize this complex further, we chromatographed crude HSV-1(F)-infected HeLa cell lysate or the purified complex obtained from gel filtration on Superdex 200 (Figs. 7 and 8) on a Mono Q column. To our surprise, none of the fractions from either chromatographic procedures yielded any phosphatase activity specific for eIF-2␣. Assays of the aliquots of fractions 2-40 from the Mono Q chromatography indicated that the ␥ 1 34.5 protein eluted in fractions 28 -34 with peaks at 29 and 32-33 (data not shown). Because the KCl in these fractions is approximately 0.53-0.68 M, our inability to recover activated eIF-2(␣P) phosphatase was probably the result of the dissociation of the complex by the high KCl concentrations required to elute the ␥ 1 34.5 protein.

DISCUSSION
The salient features of this report are as follows. First, in an earlier article we reported that the ␥ 1 34.5 protein binds to PP1 and that the cells infected with wild-type virus but not the cells infected with the ␥ 1 34.5 Ϫ mutant contain a highly potent, phosphatase activity specific for the ␣ subunit of the translation initiation factor eIF-2. In this report we link the two observations by showing that the cytoplasm of wild-type virus-infected cells contained a high molecular weight complex that contained both ␥ 1 34.5 protein and PP1 and that this fraction specifically dephosphorylated eIF-2(␣P). The approximate molecular weight of the complex (340,000) is significantly higher than the combined molecular weight of single copies of ␥ 1 34.5 protein and PP1. The results suggest that either one or both interacting proteins are present in several copies or that the complex has additional as yet unknown polypeptides.
Although we have not specifically addressed the issue, the results of our studies do not support the hypothesis that the amount of PP1 increased after infection. In several experiments we have not observed significant change in the level of protein phosphatase 1, as measured by immunoblotting with anti-PP1 antibody in HeLa cells infected with HSV-1 when compared with HeLa cells that were mock-infected or infected with the ␥ 1 34.5 deletion mutant (data not shown). Rather, the results of this and preceding studies are consistent with the hypothesis that the ␥ 1 34.5 protein binds to PP1 and that the activity is enhanced and redirected to dephosphorylate eIF-2␣.
Second, we have identified the domain of the ␥ 1 34.5 protein that binds to PP1. Specifically, the sequence of amino acids 190 -203 of the ␥ 1 34.5 protein fused to GST strongly bound and pulled down PP1. Moreover, a peptide containing amino acids 185-211 of the ␥ 1 34.5 protein competed with the protein for binding to PP1 and substitution of Val 193 and Phe 195 with Glu and Leu, respectively, abolished the activity of the ␥ 1 34.5 protein. The significance of these observations stems from the fact that this sequence is highly conserved among homologs of GADD34 and is also present in the NL protein of African swine fever virus (Fig. 1). More important, this sequence, exemplified in greater detail in Table I, contains a motif ((Arg/Lys)-(Val/ Ile)-Xaa-Phe) present in all of the subunit, accessory, or regulatory proteins that bind directly to the catalytic subunit of PP1 (27,28). These proteins interact with the catalytic subunit of PP1 (PP1c) in a mutually exclusive manner. Among the interacting proteins studied to date are the G-subunit (G M ), which targets the PP1 to glycogen particles in muscle (35), DARPP-32 (dopamine and cAMP-regulated phosphoprotein) (36), inhibitor 1 (37), splicing factor polypyrimidine tract-binding proteinassociated splicing factor (38), p53BP2 (39), and NIPP-1 (40).  1 and 2) and 250-l portions of fractions 15-26 (numbered at top and in lanes [3][4][5][6][7][8][9][10][11][12][13][14], to which were added 40 g of sperm whale apomyoglobin as carrier, were precipitated by mixing with 2 volumes of acetone at Ϫ20°C (44). The solubilized precipitates and 20 l of S10 fraction that was not precipitated (lane 15) were subjected to electrophoresis in denaturing gels, transferred electrically to nitrocellulose sheets, and reacted with antibody to PP1 or ␥ 1 34.5 protein as described under "Materials and Methods." B, autoradiographic image of purified, in vitro phosphorylated eIF-2 reacted with fractions 17-32 described in the legend to Fig. 7. Aliquots of the indicated Superdex 200 fractions from HSV-1-infected HeLa cell lysates were tested for their ability to dephosphorylate 32 P-labeled eIF-2(␣,␤) as described under "Materials and Methods." Lane 8 contains untreated, 32 P-labeled eIF-2, and the mixture in lane 9 received 0.5 l of the initial HeLa cell S10 fraction. The arrows marked ␣ and ␤ indicate the positions of the ␣ and ␤ subunits of eIF-2. The arrow marked 39 K indicates the position of a M r 39,000 protein unrelated to eIF-2␣ (8).  (27,28). In this report, we also showed that the ␥ 1 34.5 peptide containing this motif disrupted the binding of the ␥ 1 34.5 protein to PP1.
The current definition of PP1 regulatory or subunit proteins is that they bind to the catalytic domain of the PP1 and that they regulate or direct PP1 activity to specific substrates. Inasmuch as (i) ␥ 1 34.5 protein binds PP1, (ii) the binding site has the motif common to PP1 subunits that bind the catalytic subunit, and (iii) the function of PP1 is redirected to dephosphorylate eIF-2(␣P), the carboxyl-terminal domain of the ␥ 1 34.5 protein has the structural and functional attributes of a PP1 subunit.
Third, as shown in Fig. 5, the recombinant virus carrying a truncated ␥ 1 34.5 protein that contains the PP1-binding motif did not block the shut-off of protein synthesis associated with activated PKR. This result indicates that the binding site is necessary to enable binding but not sufficient to redirect PP1 to dephosphorylate eIF-2(␣P). We have also noted that the domain of the ␥ 1 34.5 protein carboxyl-terminal to the binding motif (amino acids 205-263) also binds PP1 but weakly. Relevant to the interpretation of these data are two observations: (i) subunits of PP1 may contain catalytic subunit-binding sites in addition to the (Arg/Lys)-(Val/Ile)-Xaa-Phe binding motif (28) and (ii) as noted in the Introduction, the carboxyl terminus of the GADD34 protein can substitute for the carboxyl terminus of the ␥ 1 34.5 protein in blocking the shut-off of protein synthesis and also binds PP1. GADD34 and ␥ 1 34.5 proteins share significant homology in their amino acid sequences carboxylterminal to the (Arg/Lys)-(Val/Ile)-Xaa-Phe binding motif. We interpret these results to suggest that like other subunits of the PP1, ␥ 1 34.5 protein contains sequences that interact with the PP1 catalytic domain and also, possibly, with other proteins to bring about the direct interaction of PP1 with eIF-2␣. As is the case with many other subunits of the PP1 protein, this aspect of the three-body interaction (the catalytic PP1 domain, the regulatory subunit, and the substrate) remains to be elucidated.
Fourth, the question remains as to the function of GADD34. Of the various GADD proteins, GADD34 is the least well understood. Our studies suggest that at least the carboxyl-terminal domain of the protein has functions consistent with those of a PP1 subunit similar to that of the ␥ 1 34.5 protein. The conclusion that GADD34 can perform this function does not exclude the possibility that this protein has other functions as well.
Lastly, it seems appropriate to stress that activation of PKR is a common obstacle facing most viruses studied to date. In turn, viruses have evolved a variety of mechanisms to block the shut-off of protein synthesis resulting from phosphorylation of the eIF-2␣. The herpes simplex viruses and possibly, African Swine fever virus, have a mechanism very different from any studied to date in that they have expropriated a piece of a cellular protein to cause the dephosphorylation of eIF-2␣ rather than to block the activation of PKR or the phosphorylation of eIF-2␣. It is also important to note that this mechanism is not universally used by all herpesviruses because the ␥ 1 34.5 gene is conserved among very few herpesviruses sequenced to date. These observations are consistent with the hypothesis that GADD34 does indeed enable the dephosphorylation of eIF-2␣ in response to stress and that the transient expression of the gene may explain why so few viruses captured the essential domain of the GADD34 gene to nullify the effect of activated PKR.