The Human SNF5/INI1 Protein Facilitates the Function of the Growth Arrest and DNA Damage-inducible Protein (GADD34) and Modulates GADD34-bound Protein Phosphatase-1 Activity*

The growth arrest andDNA damage-inducible protein (GADD34) mediates growth arrest and apoptosis in response to DNA damage, negative growth signals, and protein malfolding. GADD34 binds to protein phosphatase-1 (PP1) and can attenuate translational elongation of key transcriptional factors through dephosphorylation of eukaryotic initiation factor-2α. We reported previously that the human trithorax leukemia fusion protein (HRX) can bind to GADD34 and abrogate GADD34-mediated apoptosis in response to UV irradiation. We found that hSNF5/INI1, a component of the hSWI/SNF chromatin remodeling complex, also binds to GADD34 and can coexist with GADD34 and HRX fusion proteins as a trimolecular complexesin vivo. In the present report, we demonstrate that hSNF5/INI1 binds to GADD34 in part through the PP1 docking site within a domain homologous to herpes simplex virus-1 ICP34.5. We found that hSNF5/INI1 can bind PP1 independently and weakly stimulate its phosphatase activity in solution and in complex with GADD34. hSNF5/INI1 and PP1 do not compete for binding to GADD34 but rather form a stable heterotrimeric complex with GADD34. We also show that Epstein-Barr nuclear protein 2, which binds hSNF5/INI1, can disrupt hSNF5/INI1 binding to GADD34 and partially reverse the GADD34-mediated growth suppression function in Ha-ras expressing HIH-3T3 (3T3-ras) cells. These results implicate hSNF5/INI1 in the function of GADD34 and suggest that hSNF5/INI1 may regulate PP1 activity in vivo.

GADD34 1 is a growth arrest and DNA damage-inducible gene (1), whose level of transcript is increased in response to a variety of agents that elicit genomic damage and apoptosis (2,3) as well as by amino acid deprivation and by agents that lead to protein malfolding in the endoplasmic reticulum (4). The transcriptional regulation of GADD34 has been shown to be independent of functional p53 (2). At its carboxyl terminus, GADD34 harbors a highly conserved domain homologous to ICP34.5 of HSV1, a virulence factor that blocks the premature shut off of protein synthesis in HSV1-infected neuroblastoma cells and may interfere with apoptosis (5,6). Although the homologous domain of murine GADD34 (MyD116) can supply the anti-apoptotic functions of viral ICP34.5 (7), mammalian GADD34 likely functions to facilitate both growth arrest and apoptosis in response to DNA damage and other cellular stresses (3,8,9).
Recently both viral ICP34.5 and mammalian GADD34 proteins have been shown to regulate the activity of protein phosphatase 1 (PP1) in vitro (10,11). ICP34.5 coexists in a high molecular weight complex with PP1 in cellular extracts derived from HSV1-infected HeLa cells (10). Mutations in either ICP34.5 or GADD34 which disrupt their binding to PP1 also impair PP1 activity and the protein malfolding response (4,12). Recent evidence suggests that the translational elongation factor eIF2␣ is a likely biological substrate for PP1 (10 -12). eIF2␣ can be phosphorylated by several serine/threonine kinases, including the double-stranded RNA-activated kinase that is induced in response to interferon (13). It is also phosphorylated by the stress-induced kinases PERK (PKR-related ER-kinase) (14,15) and GCN2 (general control nonderepressible) (16) in cells undergoing stress from protein malfolding in the endoplasmic reticulum or from amino acid deprivation. Phosphorylation of eIF2␣ interferes with the formation of the translational preinitiation complex by binding to and inhibiting eIF2B (17); however, phosphorylated eIF2␣ paradoxically enhances the translation of the activating transcription factor-4 (ATF4), which, in turn, is necessary for the induction of stress-inducible gene CHOP (also known as GADD153) and other stress-responsive genes (15,18). GADD34-mediated activation of PP1 has been proposed to attenuate eIF2␣-mediated ATF4 function and thus complete a feedback loop initiated by the protein malfolding response (4).
Recently we reported the association of human SNF5 protein (hSNF5/INI1) and GADD34 and showed that both proteins can coexist in a trimeric complex with chimeric leukemic HRX fusion proteins, generated as a result of chromosomal translocations involving the 11q23 locus targeted in acute leukemia (9). The overexpression of HRX fusion proteins inhibited GADD34-mediated apoptosis in UV-treated SW480 cells. hSNF5/IN1 is an invariant member of the hSWI/SNF chromatin remodeling complex (19), which may have both transcriptional activation and repression functions (20 -22). A genetic search for a locus responsible for the aggressive malignant rhabdoid tumor occurring in children revealed the bi-allelic inactivation of the hSNF5/INI1 gene in a high percentage of these tumors (23). Moreover, knockout mice haploinsufficient for the murine hSNF5 gene have been shown to be predisposed to the early occurrence of a solid tumor resembling malignant rhabdoid tumor (24). Thus, hSNF5/INI1 appears to be a classic tumor suppressor gene whose protein function may extend beyond chromatin remodeling. In fact, Brm and BRG-1 (two isoforms of the mammalian SWI2/SNF2 protein) have been found to be phosphorylated differentially at the G 2 /M transition (25). This modification leads to the exclusion of these proteins from the condensed chromosome during mitosis. Phosphorylated Brm is then targeted for degredation, and Brm accumulates in cells arrested in G 0 , apparently as a result of resynthesis upon exiting the cell cycle (22).
The observed association of GADD34 with hSNF5/INI1 prompted us to ask whether hSNF5/INI1 might participate in GADD34 function and whether the interaction with GADD34 might play a role in hSNF5/INI1 tumor suppression. In this report, we characterize further the interaction of hSNF5/INI1 with GADD34. We have mapped the hSNF5/INI1 interaction domain to a small region in GADD34 which encompasses the ICP34.5 homologous domain on GADD34. Alanine replacement of residues 555-558 (VRFS) results in a nearly complete disruption of the hSNF5/INI1-GADD34 interaction. We show here that hSNF5/INI1 can bind independently to the PP1 catalytic subunit (PP1c) and that GADD34, PP1c, and hSNF5/INI1 can coexist in a heterotrimeric complex in 293T cells. We further demonstrate that affinity-purified hSNF5/INI1 protein can stimulate PP1c activity in vitro. We show that competitive disruption of the GADD34-hSNF5/INI1 interaction by enforced expression of Epstein-Barr nuclear protein 2 (EBNA2) inhibits GADD34-mediated growth suppression of Ha-ras-expressing NIH-3T3 (3T3-ras) cells. Taken together, these results suggest that GADD34 mediates growth suppression at least in part through its interaction with hSNF5/INI1 and that hSNF5/INI1 may function as a regulatory subunit of PP1 either independently or as a part of the GADD34 complex.

Generation of GADD34 and hSNF5 Mutants and Subcloning into
Plasmids-External and internal deletion mutants of GADD34 and hSNF5/INI1 were generated by PCR with appropriate external primers at the indicated positions and ligated into either pSG5-FL (9) at the EcoRI/BamHI sites or pCS2/MT (26) vectors at EcoRI/XhoI sites to generate in-frame expression cassettes. The internal deletions introduce two junctional amino acids (GS) into the GADD34 internal deletion mutants and two junctional amino acids (YY) into hSNF5 internal deletion mutants. The alanine and serine GADD34 mutants were generated with PCR replacing the indicated sequences with either AAAA or SSS sequences. pSG5-PP1c and pSG5-EBNA2 have been described previously (9,27).
Cell Lines and Transfections-Human 293T cells and 3T3-ras cells (a gift from R. Bruce Montgomery) were maintained in Dulbecco's modified Eagle's medium (Bio-Whittaker) supplemented with 10% fetal bovine serum (Bio-Whittaker) and 1% penicillin/streptomycin. For coimmunoprecipitation analysis, 10 g of the indicated expression plasmids was transfected by the calcium phosphate method into 293T cells following standard protocols (28). The total amount of plasmid DNA in each transfection was constant. Blank expression vectors were used to make up for the difference in the amount of specified vectors indicated in each condition. The transfection media were replaced at 24 h, and the cells were allowed to recover for an additional 24 h. For survival and clonogenic assays, 5-12.5 g of the indicated plasmids along with the marker plasmids was transfected into 3T3-ras cells with Superfect reagent (Quiagen) following the manufacturer's recommendations. In transient survival assays, the cells were assayed for ␤-galactosidase expression at 48 -72 h post-transfection. In the clonogenic assays, selective antibiotics, either 500 g/ml G418 or 1 g/ml puromycin, were added 24 h after transfection.
Transient Survival and Clonogenic Assays-In transient survival assays, 10 5 cells were seeded onto 60-mm plates 16 h before transfection. 48 h after transfection of 3T3-ras cells, the cells were rinsed gently with phosphate-buffered saline and fixed with a solution consisting of 2% paraformaldehyde, 0.2% glutaraldehyde in 0.1 M sodium phosphate, pH 7.3, for 10 min at 4°C. The in situ ␤-galactosidase expression was assayed in X-gal stain (100 mM sodium phosphate, pH 7.3, 1.3 mM MgCl 2 , 3 mM K 4 Fe(CN) 6 , 3 mM K 3 Fe(CN) 6 , and 1 mg/ml X-gal). 16 -24 h after X-gal stain, the blue cells were counted under a dissecting microscope. For clonogenic assays, 2-5 ϫ 10 4 cells were seeded onto 60-mm plates 16 -24 h before transfection. 24 h after plasmid transfection, the medium from each plate was changed, and G418 or puromycin was added at the indicated concentration. The cells were maintained for 14 days with fresh medium and drug replacement every 3-4 days. The number of colonies formed/plate was determined quantitatively under a dissecting microscope. Triplicate transfections were performed for each condition for the quantitative analysis.
Generation of FLAG-tagged hSNF5/INI1-The recombinant proteins were isolated from Sf9 cells infected with recombinant baculovirus harboring an amino-terminal FLAG-tagged full-length hSNF5/INI1 sequence according to the manufacturer's protocol (BD PharMingen). Briefly, the baculovirus transfer vector pVL1393 was modified by the insertion of a double-stranded nucleotide containing a consensus Kozak sequence (GCCACC), an initiating ATG followed by a FLAG encoding sequence, and multiple cloning sites (BamHI, EcoRI, PstI, NheI, and RsrII). The hSNF5/INI1 cDNA sequence encoding the full-length protein (29) was removed by digestion with EcoRI and cloned into this modified transfer vector. Sf9 cells were transfected by calcium phosphate transfection, and recombinant viruses were identified and plaque purified following the vendor's protocols. The recombinant proteins were extracted from cell pellets (containing ϳ10 8 cells) in 1 ml of cold lysis buffer (50 mM Tris-HCl, pH 7.5, 0.5% Nonidet P-40, 100 mM NaCl, 5 mM EDTA, and protease inhibitor mixture) on ice for 15 min. After centrifugation, the FLAG-tagged proteins were adsorbed onto 400 l of anti-FLAG M2 affinity gel (resuspended 1:1 in Tris-buffered saline, Sigma) and eluted with 400 l of FLAG-peptide (150 ng/l). The hSNF5/INI1 concentration was determined by Western analysis with known control FLAG-tagged proteins, and the purity of the preparations was estimated by Coomassie staining (60). The eluent was dialyzed against Tris-buffered saline supplemented with 0.1 mM phenylmethylsulfonyl fluoride.
Protein Phosphatase Assays-The substrate 32 P-labeled histone (type III-SS, Sigma) was generated as described by Li et al. (30) with the substitution of 5 units/ml porcine heart protein kinase catalytic subunit (Sigma), precipitated twice with 20% trichloroacetic acid, and washed extensively with cold acetone. The pellet was resuspended in 1 ml of 50 mM Tris-HCl, pH 7.5. In vitro protein phosphatase assays were carried out in phosphatase buffer (20 mM Tris-HCl, pH 7.5, 1 mg/ml bovine serum albumin, 0.1 mM EGTA, 0.1% ␤-mercaptoethanol) with 20 g of 32 P-labeled histone/reaction in the presence of the indicated amount of recombinant human PP1c␥ (Calbiochem) and affinity-purified FLAGtagged hSNF5/INI1, and in 50-l reaction volumes for 1-4 h at 30°C. Okadaic acid (Calbiochem) or nuclear inhibitor of PP1 (Calbiochem) was included in indicated samples. In the experiments in which phosphatase activities were measured in immunoprecipitates, the immune complexes immobilized on protein A-Sepharose were washed twice with 1 ϫ lysis buffer and once with the phosphatase buffer before the phos-phatase assay (30). The reactions were terminated with 100 l of 20% trichloroacetic acid, centrifuged for 5 min, and 100 l of the supernatant was counted by Cerenkov counting.

GADD34
Interacts with hSNF5/INI1 through the HSV1 ICP34.5 Homologous Domain-Previously we demonstrated that full-length GADD34 interacts with hSNF5/INI1 in vitro and in vivo (9). To identify the hSNF5/INI1 binding site on GADD34, we performed both external and internal deletion analyses ( Fig. 1) by examining the interaction of mutant GADD34 with hSNF5/INI1 in coimmunoprecipitation assays (Fig. 1A). The results of this deletion analysis are summarized in Fig. 1B and show that the hSNF5/INI1 interaction domain on GADD34 lies between amino acids 536 and 583 is encompassed within a region homologous to the HSV1 ICP34.5. The deletion of these amino acids (Fig. 1A, lane 4) results in the complete disruption of hSNF5/INI1 binding. Of note, both fulllength and carboxyl-terminal deletion (⌬C67) forms of hSNF5/ INI1 bind equally well to GADD34 (see Fig. 6). In subsequent coimmunoprecipitation analyses the truncated form of hSNF5/ INI1 was used because it can be distinguished readily from the immunoglobulin heavy chain on Western analysis.
Alanine Substitution Mutation of GADD34 at Positions 555-558 Weakens the Interaction with hSNF5/INI1 and Disrupts GADD34 Function-A comparison of aligned mouse, hamster, and human GADD34 sequences with that of HSV1 ICP34.5 previously revealed several short stretches of conserved residues (depicted in bold letters in Fig. 2A) within the 48-amino acid hSNF5/INI1 binding region (2). We made substitution mutations to determine which of these residues may be crucial

hSNF5/INI1 in GADD34 Function
to the GADD34-hSNF5/INI1 interaction and asked whether disruption of these sequences would result in a mutant GADD34 protein that is functionally defective. In the substitution mutation analysis, we generated either serine or alanine substitutions that respectively replaced the sequence within each of these neutral or charged amino acid stretch. We analyzed the interaction of these GADD34 mutants with hSNF5/ INI1 by coimmunoprecipitation analysis (Fig. 2B). The observed strength of the hSNF5/INI1-GADD34 interaction is summarized in the top panel and shows that specific replacement of residues 555-558 (VRFS) with alanine residues significantly reduced the binding of GADD34 to hSNF5/INI1 compared with serine or alanine substitutions at other positions (Fig. 2B, bottom panel). To test the effect of these mutations on the overall GADD34 function, we assayed the growth suppression function of these mutants in clonogenic assays using 3T3ras cells. 3T3-ras is a subline of NIH-3T3 cells which stably expresses and is transformed by the activated G12V Ha-ras (31). In this assay, 3T3-ras cells were transfected with expression vectors encoding either wild-type or mutant GADD34 along with a selection vector (for either neomycin or puromycin). The number of colonies appearing in duplicate 60-mm dishes was scored 10 days after antibiotic selection. The results from three independent experiments are depicted in Fig. 2C and show that the alanine substitution mutation at positions 555-558 uniquely disrupted the inhibition of 3T3-ras colony formation by GADD34. Other serine and alanine substitutions did not affect the growth inhibitory function of GADD34. This result correlates with the GADD34 binding to hSNF5/INI1 ( Fig. 2A) and supports the hypothesis that the association with hSNF5/INI1 is pertinent for some aspects of GADD34 function.
GADD34, hSNF5, and PP1c Form a Stable Trimolecular Complex-GADD34 forms a complex with PP1c and promotes the PP1c-mediated dephosphorylation of eIF2␣ in vitro. This binding is mediated through a PP1 docking motif (KVRF) in GADD34, where substitutions, either Val 3 Glu in MyD116 or Val 3 Glu and Phe 3 Leu in ICP34.5, have been shown to abrogate this interaction (4,12). This PP1 docking motif is disrupted by the alanine mutation at positions 555-558 shown above to abrogate the hSNF5/INI1 binding partially, raising the possibility that the observed functional defect of the GADD34 alanine mutant at 555-558 may be the result of its inability to bind PP1 (see Fig. 2C). We attempted to separate the PP1 and hSNF5/INI1 binding sites genetically within this four-amino acid region by analyzing the effect of single alanine substitutions, but we could not isolate a mutant that is specifically defective in its interaction with either hSNF5/INI1 or PP1 (data not shown). These observations suggest that hSNF5/ INI1 may compete with PP1c for binding to GADD34. To test this hypothesis, we performed coimmunoprecipitation analysis from lysates of 293T cells after cotransfection of expression vectors for Myc-GADD34 and FLAG-PP1c in the presence of increasing amounts of plasmid for FLAG-hSNF5/INI1 (Fig.  3A). In this experiment, the Myc-GADD34 expression is driven from pCS2, a vector lacking the SV40 origin of replication. For this reason the level of Myc-GADD34 expression (top panel) is predictably lower than that of the FLAG-tagged proteins that are expressed from the pSG5 vector that contains the SV40 origin of replication and therefore results in elevated expression level in SV40 T antigen-expressing 293T cells (middle panel). Under this condition of limiting and invariant Myc-GADD34 expression, increasing the level of the hSNF5/INI1 protein (Fig. 3A, middle panel, lanes 3-5)

hSNF5/INI1 in GADD34 Function
shows that increasing the expression of hSNF5/INI1 resulted in increased levels of hSNF5/INI1 in the immunoprecipitated complex without displacing PP1c. This result demonstrates that all three proteins can coexist in a stable trimolecular complex under these conditions.
To determine whether hSNF5/INI1 can bind directly to PP1c, we tested the binding in vitro. In the experiment shown in Fig. 3B, mixtures of affinity-purified FLAG-hSNF5/INI1 with either FLAG-peptide or recombinant human PP1c␥ were precipitated with anti-FLAG antibody and captured onto protein A-Sepharose beads. The immune complexes were assayed for phosphatase activity by 32 P i release from phosphohistone as described under "Materials and Methods" (Fig. 3B). The results show that PP1 activity was coprecipitated selectively with FLAG-hSNF5/INI1 but not with FLAG-peptide nor by PP1 itself. The addition of the PP1 inhibitor NIPP1 reduced the phosphatase activity by 80%, confirming the presence of PP1 in the hSNF5/INI1 immunoprecipitates. To determine whether endogenous PP1 can bind hSNF5/INI1, we transfected 293T cells with either the blank vector or the expression plasmid for FLAG-hSNF5/INI1 and carried out immunoprecipitation of the cell lysates with either anti-FLAG or EBNA2-hemagglutinin antibodies (Fig. 3C). The immunoprecipitates were then probed against anti-PP1 antibodies to determine the coprecipitated PP1 (bottom panel). The results show that endogenous PP1 is selectively coprecipitated with FLAG-hSF5/INI1 by anti-FLAG antibodies and not by EBNA2-hemagglutinin antibodies. The results here confirm the in vitro binding of PP1 to hSNF5/INI1 and suggest that the heterotrimeric complex of GADD34, hSNF5/INI1, and PP1 is stabilized by the intermolecular interactions of all three proteins.
GADD34-bound PP1c Activity Is Stimulated by hSNF5/ INI1-We tested the possibility that hSNF5/INI1 may be a substrate for PP1c but could not find sufficient evidence for hSNF5/INI1 phosphorylation in vivo. Despite the presence of several possible phosphorylation sites on hSNF5/INI1 for protein kinase A, protein kinase C, and casein kinase II, none of these kinases could phosphorylate hSNF5/INI1 in vitro (data not shown). We next tested the hypothesis that hSNF5/INI1 may regulate PP1 activity. We tested this initially by examining the phosphatase activity of PP1c in the presence of affinitypurified hSNF5/INI1 in vitro (Fig. 4A). For this experiment, FLAG-hSNF5/INI1 was affinity purified on an anti-FLAG antibody column and eluted with FLAG-peptide. To account for the possible residual effects of FLAG-peptide in the eluate on phosphatase activity, we included a series of control assays in which proportional quantities of FLAG-peptide were added to the reaction. The results show that hSNF5/INI1 can modestly enhance the phosphatase activity of PP1c (by ϳ2-fold). This up-regulation of PP1 activity is unrelated to the possible residual presence of FLAG-peptide which inhibited the PP1 activity for reasons that are unclear. The hSNF5/INI1-enhanced phosphatase activity is completely inhibited by either the presence of 50 nM okadaic acid (OA) or 10 pM NIPP1 (32)(33)(34).
We proceeded to determine whether hSNF5/INI1 can also modulate the phosphatase activity of endogenous PP1 that is bound to GADD34. In the experiments shown in Fig. 4, B and C, we isolated GADD34 complexes immunoprecipitated with anti-Myc antibodies from 293T cells transfected with a Myc-GADD34 expression plasmid. These GADD34 complexes are found to be associated with phosphatase activities that can be differentiated based on their sensitivity or resistance to OA (or to NIPP1). As shown in Fig. 4B, the PP1-specific activity can only be found in the immune complexes precipitated with the anti-Myc antibodies and not in the immunoprecipitates with a monoclonal antibody against an unrelated epitope (R3 against the carboxyl terminus of EBNA2), suggesting that endogenous PP1 is specifically in complex with GADD34 (also shown in the immunoblots of Fig. 4C). We then cotransfected a hSNF5/INI1 (full-length) expression construct to determine whether the hSNF5/INI1 association will influence GADD34-bound PP1 activity. In the experiment, shown in Fig. 4C, the PP1-specific phosphatase activity associated with the GADD34 immune complex is determined by measuring the fraction of the total phosphatase activity which can be inhibited with 50 nM OA. Both OA-inhibited and -uninhibited phosphatase activity are reported, and the results also show that hSNF5/INI1 modestly enhances GADD34-associated PP1 activities (OA-inhibited phosphatase activity). In contrast, the OA-uninhibited phosphatase activities remain unaffected by hSNF5/INI1. Similar results were obtained when NIPP1 was used to identify the associated PP1 activity (data not shown). The immunoblots shown in Fig. 4C depict the levels of hSNF5/INI1 and GADD34 expression and the level of coprecipitated endogenous PP1 and hSNF5/INI1 proteins. We confirm the above observation that hSNF5/INI1 does not competitively displace the endogenous PP1 bound to GADD34 because the coprecipitated PP1 level was found to be unaltered by the coexpression of hSNF5/INI1. The increased GADD34-bound PP1 activity is therefore a result of modulation by the hSNF5/INI1 protein and not a reflection of increased levels of PP1 bound to GADD34. Of note, because the presence of DNA has been shown to stimulate both hSWI/SNF complex ATPase activity and helicase function (35,36), we also examined the effect of free plasmid DNA on the stimulation of PP1 activity by hSNF5/INI1. Free DNA did not affect PP1specific phosphatase activity either in vitro or in GADD34associated protein complex.
EBNA2 Competitively Disrupts GADD34-hSNF5/INI1 Interaction and Can Partially Reverse GADD34-mediated Growth Suppression in 3T3-ras Cells-We have reported previously that EBNA2 interacts with hSNF5/INI1 (27) and demonstrated the in vivo recruitment of the hSWI/SNF complex to EBNA2responsive promoter elements (37). Deletion mapping of the hSNF5/INI1 protein narrowed its GADD34 interaction to a conserved 15-amino acid region encoded in exon 7 (residues 304 -318) of the hSNF5/INI1 gene (see Fig. 6). This region on hSNF5/INI1 is close to the EBNA2 interaction domain (27). Because EBNA2 does not bind to GADD34 or PP1 (data not shown) and does not affect growth of epithelial cells (Fig. 5B) (38), we used EBNA2 to perturb the hSNF5/INI1-GADD34 interaction to assess GADD34 growth suppression function in 3T3-ras cells under conditions that disrupt the GADD34-hSNF5/INI1 interaction.
To determine whether EBNA2 could competitively disrupt the interaction between hSNF5/INI1 and GADD34, we performed coimmunoprecipitation of hSNF5 and GADD34 in the presence of increasing amounts of EBNA2. In the experiment shown in Fig. 5A, 293T cells were transfected with expression vectors for FL-hSNF5/INI1 and Myc-GADD34 along with increasing amounts of EBNA2 expression vector. Immunoblots of the lysates (middle and bottom panels) show that the expression levels of GADD34 and hSNF5/INI1 remained constant, unaffected by increasing amounts of EBNA2 expression (lanes 3-6). At an EBNA2:GADD34 vector ratio of 6:1 and 8:1, EBNA2 disrupted the GADD34-hSNF5/INI1 interaction as shown by the decreasing amounts of hSNF5/INI1 immunoprecipitated with GADD34 (bottom panel, lanes 4 -6). This result demonstrates that GADD34 and EBNA2 can compete for hSNF5/INI1 binding in vivo. We next determined whether the hSNF5/INI1 interaction is essential to some aspect of GADD34 growth suppression function. Because EBNA2 can partially disrupt the GADD34-hSNF5/INI1 interaction we tested its ability to block GADD34-mediated growth suppression in transiently transfected 3T3-ras cells (Fig. 5B). In this experiment, 3T3-ras cells were transfected with the indicated amounts of plasmids encoding FLAG-GADD34 and EBNA2 along with a fixed amount of the marker plasmid pCMV-LacZ. The empty vector, pSG5-FL, was added to each transfection such that the total amount of the transfected plasmid was invariant. To measure cell survival at 48 h, we determined the number of LacZ-expressing cells by an in situ ␤-galactosidase assay. Fig.  5B shows that increasing EBNA2 partially reversed the GADD34-mediated reduction in the number of surviving 3T3ras cells in a concentration-dependent manner (left panel, columns [3][4][5][6]. At an EBNA2:GADD34 transfected plasmid ratio of 8:1 (column 6), the number of ␤-galactosidase-expressing cells was ϳ80% of the control (column 1) compared with 20% of control for GADD34 transfection without EBNA2 (column 2). EBNA2 expression alone did not result in increased ␤-galactosidase-expressing 3T3 cells (right panel). These results suggest that the disruption of GADD34 function by EBNA2 is at least in part mediated by its competitive disruption of the GADD34-hSNF5/INI1 interaction. These results further support a functional role of hSNF5INI1 in GADD34-mediated growth suppression.
GADD34 Binds to hSNF5 through a Conserved Region Encoded by Exon 7-We have reported previously that the binding of hSNF5/INI1 to GADD34 in vitro can be disrupted with the non-ionic detergent Nonidet P-40, suggesting that intermolecular hydrophobic interactions are important to the association of these two proteins. We undertook a mutational analysis to map the GADD34 binding domain on the hSNF5/INI1 protein.
We constructed a series of carboxyl-terminal and internal deletion mutants and tested their interaction with the full-length GADD34 (Fig. 6) and found that a small region between residues 305 and 318 was important to the overall binding. This region, which is encoded by exon 7, overlaps with a stretch of hydrophobic residues within the second of two conserved im- perfect repeats reported previously (39). To confirm that this stretch of amino acids in hSNF5/INI1 contributes to its interaction with GADD34, we generated amino acid substitutions replacing LGLGGEFVTTIA (residues 303-314) with LGEDEEYYTTIA, thus decreasing the hydrophobicity of this region. This hSNF5/INI1 substitution mutant was found to bind weakly to GADD34. An analogous replacement of a hydrophobic sequence (FVPAIASAI) at positions 233-241 encoded by exon 6 did not affect GADD34 binding. These results suggest that GADD34 binds to the carboxyl-terminal end of hSNF5/INI1 in part through its interaction with a small hydrophobic region contained within a conserved repeat encoded by exon 7 of the hSNF5/INI1 gene. This finding is consistent with mutational data of hSNF5/INI1 found in malignant rhabdoid tumor samples, which suggests that events of exon 7 are likely important to the overall hSNF5/INI1 tumor suppression function (40). DISCUSSION GADD34 is a cellular stress response protein whose transcript is regulated independently of functional cellular p53. Because ϳ50% of all cancers have been estimated to harbor a mutation in p53 (41), GADD34-mediated genotoxic response may well play an important role in cellular response to cancer therapy and perhaps to clinical outcome. Characterization of the components of the GADD34 functional pathway may lead to the identification of new targets for cancer drug development. In the present work, we present data supporting a role for hSNF5/INI1, a component of the hSWI/SNF complex, in the functions of GADD34. We show that a segment on GADD34 mediates hSNF5/INI1 binding, and this short stretch of amino acids resides in the region homologous to the functional domain in HSV1 ICP34.5, and the hSNF5/INI1 binding site overlaps the PP1 docking motif on GADD34. To our surprise, hSNF5/ INI1 does not compete with PP1 for GADD34 binding; instead, all three proteins form a stable heterotrimeric complex in 293T cells overexpressing these proteins. hSNF5/INI1 binds to PP1 in vitro, and it modestly stimulates the PP1-specific phosphatase activity in solution and in the GADD34-bound protein complex. Thus hSNF5/INI1 fulfills the definition of a regulatory subunit of PP1c. The Epstein-Barr virus immortalization protein EBNA2 competes with GADD34 for binding to hSNF5/ INI1 and can partially reverse growth suppression that is mediated by GADD34 when it is coexpressed with GADD34 in 3T3-ras cells.
The type I serine/threonine phosphatase, PP1, regulates diverse cellular processes such as cell cycle progression, protein synthesis, muscle contraction, carbohydrate metabolism, transcription, and neuronal signaling (42). PP1 acquires specificity through its association with targeting regulatory subunits that direct the enzyme to specific cellular compartments and confer substrate specificity, and PP1 is also regulated by inhibitory subunits that control enzyme activity (43). In general, the targeting regulatory subunits themselves do not serve as PP1 substrates (44). Consistent with these findings, hSNF5/INI1 appears to function as a targeting regulatory subunit; however, unlike the classical targeting proteins of PP1 hSNF5/INI1 can modulate PP1c activity (Fig. 4). Recently, GADD34 was shown to alter the substrate specificity of PP1c, maintaining the PP1 activity for phosphorylated eIF2␣ but reducing its activity for phosphorylated phosphorylase B (11). Although it remains to be determined whether hSNF5/INI1 would have a similar affect on PP1, an attractive hypothesis would involve the heterotrimeric GADD34⅐hSNF5/INI1⅐PP1 complex conferring substrate specificity to PP1 and targeting the enzyme to potential protein substrates associated with hSWI/SNF complex.
The hSWI/SNF complex is composed of at least 12 protein subunits (19), and the purified complex can remodel chromatin in vitro (45,46). In yeast, the SWI and SNF proteins are required for regulated transcription of a set of genes involved in metabolism and mating-type switching (47) and act both to activate and repress the transcription of a restricted set of genes (48). Yet, the mammalian SWI/SNF complex is clearly involved in functions in addition to transcription activation. Both mammalian SWI2 homologs BRG-1 and hBrm can bind to RB. hBrm in complex with retinoblastoma protein suppresses transcription factor E2F1 activity and can mediate G 1 cell cycle arrest (49). Cells deficient in BRG-1 were found to be resistant to retinoblastoma protein-mediated cell cycle inhibition but could regain sensitivity through conditional expression of BRG-1 (50). Both BRG-1 and hBrm have been found to be differentially phosphorylated in the cell cycle such that phosphorylation occurs during mitosis, and hBrm is excluded from the condensed chromatin (25). Although phosphorylated hBrm is targeted for degredation, the level of BRG-1 appears to be invariant. Purified BRG-1 containing hSWI/SNF complexes can be inactivated in vitro through phosphorylation by extracellular signal-regulated kinase 1 and rendered incapable of nucleosome remodeling. The complex can then be reactivated by PP2A (46). BRG1, hBrm, and hSWI3, also known to be phosphorylated, are all possible substrates for PP1 within the hSWI/SNF complex. Both PP1 and hSNF5/INI1 associate with GADD34 at its ICP34.5 homologous domain. Previously the Src kinase Lyn and proliferating cell nuclear antigen have both been shown to interact with GADD34 through this domain (51,52). Lyn binds to a SH3-like motif 50 amino acids apart from the PP1 and hSNF5/INI1 binding site (52). GADD34 was found to be phosphorylated by Lyn in response to DNA damage. Although both GADD34 and Lyn have been implicated in the DNA damage response, the coexpression of Lyn curiously interferes with GADD34-mediated apoptosis. The proliferating cell nuclear antigen binding domain within the ICP34.5 homologous region has not been delineated further. Additionally, GADD34 has been demonstrated to interact with the kinesin family protein KIF3A, Translin, and a new member of the Hsp40 family of heat shock proteins in regions outside of the ICP34.5 domain (53)(54)(55). The role of these proteins and that of proliferating cell nuclear antigen in GADD34 function remains to be clarified.
In the present report, the growth suppression function of GADD34 is assayed in the G12V Ha-ras transformed NIH-3T3 cells. The activation of ras pathways has been reported recently to down-regulate the expression of Brm, and the reintroduction of Brm into ras-transformed cells leads to partial reversion of the transformed phenotype (56). We have found that the stable expression of wild-type GADD34 in 3T3-ras cells is possible (data not shown), suggesting that both transfection reagents and selection antibiotics may introduce cellular stress and lead to the observed GADD34 effects in clonogenic and short term survival assays. Our data also suggest that ras and GADD34 activation likely exert opposing effect on the hSWI/SNF complex function.
EBNA2 is one of the six viral nuclear proteins expressed in latently infected B lymphocytes and has been found to be essential for the immortalization of B cells by Epstein-Barr virus. EBNA2 promotes transcriptional transactivation of viral and cellular genes by acting as an adaptor molecule that binds to cellular sequence-specific DNA-binding proteins (57,58) and transcription factors, including the hSWI/SNF complex (27). We have demonstrated recently the targeting of the hSWI/SNF complex to both episomal and endogenous cellular chromatin at EBNA2-responsive DNA sites (37). This targeting is mediated through the interaction of EBNA2 with hSNF5/INI1. In the work presented here, we showed that EBNA2 can competitively disrupt the GADD34-hSNF5/INI1 interaction (Fig. 5A) and can partially reverse the GADD34 growth suppression function when coexpressed with GADD34 (Fig. 6). Kempkes et al. (59) have shown that EBNA2 expression is essential for both establishment and maintenance of Epstein-Barr virus-induced transformation of primary normal B cells. EBNA2 expression also overcomes cell cycle arrest that normally occurs at both G 1 and G 2 checkpoints. Because EBNA2 is invariably expressed in proliferating Epstein-Barr virus-transformed B cells, the results presented here also suggest that it can impair the GADD34-mediated cellular stress response pathway by sequestering hSNF5/INI1. In so doing, EBNA2 may have the dual roles of altering the normal cellular transcription program and of imparting a survival advantage to latently infected B lymphocytes undergoing cellular stress.