A Splice Variant of Stress Response Gene ATF3 Counteracts NF-κB-dependent Anti-apoptosis through Inhibiting Recruitment of CREB-binding Protein/p300 Coactivator*

Activating transcription factor (ATF) 3 plays a role in determining cell fate and generates a variety of alternatively spliced isoforms in stress response. We have reported previously that splice variant ATF3ΔZip2, which lacks the leucine zipper region, is induced in response to various stress stimuli. However, its biological function has not been elucidated. By using cells treated with tumor necrosis factor-α and actinomycin D or cells overexpressing ATF3ΔZip2, we showed that ATF3ΔZip2 sensitizes cells to apoptotic cell death in response to tumor necrosis factor-α, at least in part through suppressing nuclear factor (NF)-κB-dependent transcription of anti-apoptotic genes such as cIAP2 and XIAP. ATF3ΔZip2 interacts with a p65 (RelA)-cofactor complex containing CBP/p300 and HDAC1 at NF-κB sites of the proximal promoter region of the cIAP2 gene in vivo and down-regulates the recruitment of CBP/p300. Our study revealed that ATF3ΔZip2 counteracts anti-apoptotic activity of NF-κB, at least in part, by displacing positive cofactor CBP/p300 and provides insight into the mechanism by which ATF3 regulates cell fate through alternative splicing in stress response.

Alternative splicing of pre-mRNAs encoding transcription factors is one of the common mechanisms for generating the complexity and diversity of gene regulation (1). A variety of functionally distinct isoforms is generated from a single gene by use of different combinations of splice junctions, and may play role in coordinating gene regulation in response of cells to various environmental stimuli. However, the altered gene regulatory function and the biological implication of spliced variants still remain elusive.
Activating transcription factor (ATF) 3 3 is a member of the ATF/ CREB family of basic leucine zipper-type transcription factors. It is induced upon exposure of cells to a variety of physiological and pathological stimuli (2), and it is thought to have cell-detrimental effects, such as cell cycle arrest and apoptosis (2)(3)(4)(5). Ectopic expression of ATF3 in heart, liver, and pancreatic ␤-cells causes cardiac enlargement, liver cell dysfunction, and diabetes, respectively (6 -8), supporting cytopathic activity of ATF3. On the other hand, ATF3 is also rapidly induced in the regenerating liver (9) or in cells treated with serum or growth-stimulating factors (10). It can support cell survival of endothelial cells (11) and protect neuronal cells from c-Jun N-terminal kinase-induced cell death (12). Furthermore, we have shown recently that ATF3 is a target of the proto-oncogene c-myc in serum-induced cell proliferation (13). Thus, ATF3 may function differently depending on the cellular context.
ATF3 is composed of 181 amino acids, and the basic region and leucine zipper domain from 88 to 147 amino acids are required for dimer formation and specific DNA binding (14,15). The homodimer of ATF3 represses transcription from various promoters with ATF sites (14 -16), whereas heterodimers with c-Jun or JunB activate transcription (15). In addition to the heteromeric complexity, various spliced isoforms of ATF3 may further generate functional diversity in different cellular context. ATF3⌬Zip has been isolated in serum-stimulated HeLa cells (14), whereas ATF3⌬Zip2c and ⌬Zip3 were identified in amino acid-deprived cells (17). Another isoform, ATF3b, is implicated in mediating cAMP signaling of proglucagon transcription in pancreatic ␣-cells (18). We have isolated previously ATF3⌬Zip2a and -b from cells treated with various stimuli such as A23187, TNF-␣, endoplasmic reticulum stress, or oxidative stress (19). These two isoforms encode the C-terminally truncated protein of 135 amino acids, which shares the N-terminal 116 amino acids with the full-length ATF3 but contains novel 19 amino acids at the C terminus. ATF3⌬Zip2 lacks the leucine zipper domain and thus is capable of DNA binding. It is localized in the nucleus and counteracts the transcriptional regulation by full-length ATF3 in a reporter assay (19). However, the functional role and biological significance of this spliced isoform in stress response are unknown.
Nuclear factor-B (NF-B) is a transcription factor that plays a critical role in the expression of genes involved in immune and inflammatory responses (20,21). It is located in the cytoplasm of nonstimulated cells as a form bound to IB, but it is released by degradation of IB␣ and enters the nucleus in response to stimuli. There are five known members of the mammalian NF-B/Rel family: p65 (RelA), c-Rel, RelB, p50, and p52. p65, RelB, and c-Rel are transcriptionally active, whereas p50 and p52 function as the DNA-binding subunits. Recent analysis of NF-B-deficient mice and cells has further clarified its role in inhibiting apoptosis (22). For instance, RelA-deficient mouse fibroblasts showed increased sensitivity to pro-apoptotic stimuli such as TNF-␣, which by itself is a poor inducer of apoptosis unless accompanied by inhibitors of new RNA or protein synthesis (23,24). NF-B induces the expression of a number of genes including cIAPs, cFLIP, A1, TRAF1, and TRAF2, whose products can inhibit apoptosis (22,25). Among them, cIAP2 contains two functional B sites in the promoter (26), and the cIAP2 protein directly binds and inhibits effector caspases, such as caspase 3 and 7, as well as prevents activation of pro-caspase 6 and 9 (27). Another cIAP, X chromosome-linked IAP (XIAP), also inhibits TNF-induced apoptosis of cells expressing the IB␣ super-repressor mutant (28). NF-B has also been shown to inhibit apoptosis by DNA-damaging agents, including cancer therapeutics (29,30).
It is well known that NF-B-dependent transcription requires the recruitment of multiple coactivator proteins. CBP and its homologue p300 interact with the p65 subunit of NF-B to activate responsive gene promoters (31,32), and this interaction is enhanced by inducible phosphorylation of p65 (33). The histone acetyltransferase activity of p/CAF is also required for the NF-B-activated transcription (34), whereas steroid receptor coactivator-1 interacts with the p50 subunit of NF-B to potentiate the transcription (35). Inversely, HDAC1/2 (36), SNIP1 (37), and PIAS3 (38) interact with p65 to negatively regulate gene expression. More recently, the candidate tumor suppressor gene product ING4 has been reported to interact with p65 and suppress the NF-Bdependent anti-apoptosis (39). Thus, the differential binding of p65 with coactivators or repressors determines the extent of activation or repression of NF-B-dependent gene expression. However, our knowledge is still limited about what molecule(s) modulate NF-B-dependent transcription in stress response.
In this report, we explored the functional role of ATF3⌬Zip2 by using TNF-␣-stimulated cells. It was demonstrated that ATF3⌬Zip2, but not full-length ATF3, sensitizes cells to apoptotic cell death, partly by suppressing the expression of NF-B-dependent anti-apoptotic genes cIAP2 and XIAP. ATF3⌬Zip2 is further shown to bind directly to the p65 subunit of NF-B and down-regulate the CBP/p300 recruitment. Our study provides evidence that ATF3⌬Zip2, which is generated through stress-activated alternative splicing, represses NF-B activity and may play pro-apoptotic role in stress response.
Cell Culture and Transient Expression-TGR-1 is a subclone of the immortalized rat embryonic Rat-1 cell line, and U2OS is a human osteosarcoma cell line, respectively. Cells were cultured in Dulbecco's mod-ified Eagle's medium supplemented with 10% calf serum, 100 units/ml penicillin, and 100 g/ml streptomycin in a 5% CO 2 atmosphere at 37°C. For transient expression, plasmid DNA was vortex-mixed with SuperFect (Qiagen) and transfected into cells according to the manufacturer's instruction.
Adenovirus Preparation and Gene Transfer-An adenovirus vector AdATF3 encoding the full-length ATF3 was as described (11), and a vector AdATF3⌬Zip2 for FLAG-tagged ATF3⌬Zip2 was constructed as in a protocol of an adenovirus expression vector kit from Takara (Otsu, Japan). Briefly, a blunt-ended cDNA fragment of ATF3⌬Zip2 was subcloned into the Swa1 site of the E1-deleted region of a cassette cosmid vector pAxCAwt and was cotransfected into 293 cells with DNA-terminal protein complex (42). Cells expressing recombinant viruses for FLAG-tagged ATF3⌬Zip2 were screened by anti-FLAG immunoblot analysis and cloned by a limiting dilution. An adenovirus vector encoding ␤-galactosidase (AdLacZ) was a generous gift from Dr. Saito of Tokyo University, Japan. These adenoviruses were plaque-purified, and their titers were determined by titration in 293 cells. For adenovirus-mediated gene transfer, cells were exposed to adenoviral vectors at the indicated multiplicity of infection (m.o.i.) and were cultured for the specified times.
Isolation of U2OS Cells Stably Expressing ATF3⌬Zip2 siRNA-An oligonucleotide that targets the C-terminal amino acid regions unique for human ATF3⌬Zip2 from 126 to 132, 5Ј-ACTTGCATGTGGGCT-GTTGGC-3Ј, was subcloned into pMX-puroII-U6 (43). The resulting vector was transfected into Plat E cells, and recombinant retrovirus was generated. U2OS cells stably expressing mouse ecotropic retrovirus receptor were infected with the recombinant retrovirus and were cultured in medium containing 20 g/ml puromycin for 4 days.
Cell Death Assay-TGR1 cells (6 ϫ 10 4 cells) that were infected with AdATF3⌬Zip2, AdATF3, or AdLacZ virus at 25 m.o.i. for 24 h or U2OS cells (1 ϫ 10 5 cells) were treated with the indicated concentration of TNF-␣ for 36 h or 10 ng/ml TNF-␣ and the indicated amount of actinomycin D for 18 h, respectively. Both attached and floating cells were combined and stained with 0.2% trypan blue. Viable cells were counted and expressed as a proportion of total cells as described before (11). Alternatively, cells (6 ϫ 10 4 cells) were transfected with indicated the amounts of pMEFLAGATF3⌬Zip2, pCMVGFP, and effector plasmids. At 16 h post-transfection, cells were treated with 20 ng/ml TNF-␣ and incubated for another 36 h. Cells were then examined for GFP expression and nuclear fragmentation by DAPI staining under a fluorescence microscope (Nikon). Cell death was quantitated as a fraction of cells with fragmented nuclei per GFP-positive cells. For detection of DNA ladder formation, both attached and floating cells were collected 36 h after TNF-␣ treatment. Their chromosomal DNA was extracted using a kit from Wako (Osaka, Japan), separated on a 1.8% agarose gel, and visualized by staining with ethidium bromide.
Cell Extract Preparation and Western Blot Analysis-TGR1 or U2OS cells treated as indicated were harvested, washed in PBS, and resuspended in lysis buffer (50 mM Hepes-KOH, pH 7.5, 150 mM NaCl, 1% Triton X-100, 1.5 mM MgCl 2 , 1 mM EGTA, 0.1 mM PMSF, 10 g/ml each leupeptin and aprotinin, 200 M sodium vanadate, 100 mM NaF, and 10% glycerol). After incubation on ice for 10 min, the cells were centrifuged at 10,000 rpm for 10 min, and the supernatants were taken as whole cell extract. The amounts of protein were measured by the Lowry method using bovine serum albumin as standard (44). Cell extracts (20 g of protein) were separated on an SDS-PAGE, transferred onto a nitrocellulose membrane, and subjected to Western blot using the protocol of ECL kit (Amersham Biosciences).
Nuclear Translocation of NF-B-For nuclear translocation of NF-B, U2OS cells (1 ϫ 10 6 cells) were infected with AdATF3⌬Zip2 or AdLacZ at 25 m.o.i. for 24 h and then treated with 20 ng/ml TNF-␣. At the time points indicated after TNF-␣ treatment, cells were harvested and suspended in buffer of 10 mM Hepes-KOH, pH 7.8, 10 mM KCl, 0.1 mM EDTA, 1 mM dithiothreitol, 1 mM PMSF, and 0.1% Nonidet P-40. Cytoplasmic fraction was obtained by centrifugation of cells at 10,000 rpm for 10 min. The resultant pellet containing nuclei was washed by the above buffer, dissolved in sample buffer for SDS-PAGE, and subjected to Western blot analysis.
Electrophoretic Mobility Shift Assay-U2OS cells (1.5 ϫ 10 7 cells) were infected with 25 m.o.i. of AdATF3⌬Zip2 or AdLacZ for 24 h and then treated with 20 ng/ml TNF-␣ for 3 h. Nuclear extracts were then prepared according to the method of Dignam et al. (45). Nuclear extracts (2 g of protein) were incubated in 20 l of binding buffer (10 mM Hepes-KOH, pH 7.9, 60 mM KCl, 0.5 mM EDTA, 5 mM MgCl 2 , 0.1 mM PMSF, 5 mM ␤-mercaptoethanol) containing 0.5 mg of poly(dI-dC) and 0.5 ng of radiolabeled DNA probe at room temperature for 30 min. For supershift assays, anti-p65 or anti-p50 antibody (0.1 g each) was added and incubated for another 30 min. An oligonucleotide DNA probe for consensus NF-B motif (sc-2505) 5Ј-AGTTGAGGG-GACTTTCCCAGGC-3Ј was obtained from Santa Cruz Biotechnology and radiolabeled with 25 Ci of [␥-32 P]ATP (6000 Ci/mmol) and polynucleotide kinase. Mutant oligonucleotide (sc-2511) used for the competition experiment was also product of Santa Cruz Biotechnology. Binding mixture was applied onto a 5% nondenatured polyacrylamide slab gel in Tris borate-EDTA buffer. After electrophoresis, the gel was dried on a 3MM Whatman paper and visualized by Fuji Bas 2500 image analyzer.
Luciferase Assay-Reporter plasmid, pLuc-cIAP2, was constructed by subcloning a 0.6-kb fragment of the human cIAP2 gene promoter (GenBank TM accession number AF233684) from Ϫ538 to ϩ55 into an XhoI site of pGL3 vector (Promega). NF-B-dependent reporter, pNFB-Luc, which contains four tandem copies of the NF-B consensus motif fused to a TATA-like promoter from the herpes simplex virus thymidine kinase promoter, was obtained from Clontech. U2OS cells (5 ϫ 10 4 cells) were transfected with 0.5 g of pLuc-cIAP2 or pNFB-Luc and the indicated amounts of effector plasmids. At 16 h post-transfection, cells were stimulated with 10 ng/ml TNF-␣ for 24 h. Cells were then harvested, and their extracts were assayed for luciferase activity as described (19). Both firefly and seapansy luciferase activities were measured using a dual luciferase reporter assay system according to the manufacturer's protocol (Promega). pRL-TK (Toyo Ink, Tokyo, Japan) containing the seapansy luciferase gene was used as an internal control of transfection and expression.
Binding Assay of Zip2 Complex-For in vivo binding, U2OS cells (1 ϫ 10 7 cells) were infected with AdATF3⌬Zip2 at 25 m.o.i. for 24 h and then treated with 20 ng/ml TNF-␣ for 3 h. Alternatively, U2OS cells stably expressing ATF3⌬Zip2 siRNA or control GFP siRNA were treated with 10 ng/ml TNF-␣ and 0.05 g/ml actinomycin D for 3 h. Whole cell extracts were prepared and incubated with 0.5 g each of anti-p65 or anti-FLAG antibody at 4°C for 3 h, followed by incubation with 30 l of protein G-Sepharose (Amersham Biosciences) for 2 h. The resulting immunocomplex was washed and subjected to Western blot analysis. For in vitro binding, full-length ATF3 and ATF3⌬Zip2 were expressed as GST fusion as described (19). For deletion mutants of ATF3, cDNA fragments encoding 1-40-, 1-88-, and 1-147-amino acid regions of ATF3 were subcloned into pGEX6P-1 to produce C-terminal deletions dC141, dC96, and dC34, respectively. p65 was expressed as a protein C-terminally tagged with 6 histidines using pET21a, and GST-CBP2N was a GST fusion containing the N-terminal portion of CBP from 117 to 738 amino acids (41). Bacterial extracts containing these recombinant proteins were mixed in a total of 200 l of binding buffer (50 mM Tris-HCl, pH 8.0, 0.1 M NaCl, 1 mM EDTA, 1 mM dithiothreitol, 10% glycerol) at 4°C for 3 h. After further incubation with 30 l of glutathione-Sepharose beads (50% slurry; Amersham Biosciences) or nickel-agarose beads (50% slurry; Qiagen) at 4°C for 1 h, protein complex was centrifuged, washed, and analyzed by Western blot. For binding of CBP and p65, GST-CBP2N and His-tagged p65 were incubated in the absence or presence of cell extracts prepared from U2OS cells infected with either AdLacZ or AdATF3⌬Zip2 and were subjected to pulldown assay as described above.
Statistical Analysis-Multiple comparisons were evaluated by analysis of variance followed by Scheffe's post-hoc test. Data are presented as mean Ϯ S.D. Statistical significance was assigned at the level of p Ͻ 0.05.

Ectopic Expression of ATF3⌬Zip2-sensitized Cells to TNF-␣-induced
Cell Death-Full-length ATF3 and its splice variant ATF3⌬Zip2 are induced in response to various stimuli (19). To investigate the role of ATF3⌬Zip2 in determining cell fate, we first overexpressed it by adenovirus-mediated gene transfer, and we examined cell viability after treatment with TNF-␣. As shown in Fig. 1A, expression of full-length ATF3 or control LacZ had no effect on the viability of TGR1 cells upon TNF-␣ stimulation, consistent with previous observations that TNF-␣ does not induce cell death if not provided with RNA or protein synthesis inhibitors. By contrast, viability of TNF-␣-treated TGR1 cells was significantly decreased when ATF3⌬Zip2 was expressed. However, no change in cell viability was observed when ATF3⌬Zip2 was expressed without TNF-␣ treatment, suggesting that ATF3⌬Zip2 alone is not sufficient to trigger cell death. A microscopic study of the cells expressing ATF3⌬Zip2 revealed an appearance of cells with fragmented nuclei (Fig. 1B, left panel), and the analysis of genomic DNA showed the formation of nucleosomal DNA ladder (Fig. 1B, right panel), providing evidence for apoptotic cell death. Specific cleavage of poly(ADP-ribose) polymerase, which is known to occur in apoptotic cells, was also observed in these cells (data not shown). Next, we further expressed ATF3⌬Zip2 by using plasmid expression vector to minimize nonspecific effects of adenoviral gene transfer. As shown in Fig. 1C, the ATF3⌬Zip2 expression caused the appearance of cells with condensed or fragmented nuclei in transfected GFP-positive cells. In control cells transfected with empty vector, only a basal level of apoptotic cells was observed. These data indicate that ATF3⌬Zip2 expression sensitizes cells to TNF-␣-induced apoptotic cell death.
NF-B Was Normally Activated in ATF3⌬Zip2 Cells after TNF-␣ Stimulation-NF-B is one of the major anti-apoptotic transcription factors that inhibit TNF-␣-induced apoptosis (22)(23)(24). Thus, we next examined the activation of NF-B in ATF3⌬Zip2 cells after TNF-␣ stimulation. Fig. 2A, upper panel, showed that IB␣ was rapidly degraded in ⌬Zip2 cells with kinetics similar to control LacZ cells.  Nuclear import of the p65 subunit of NF-B was then examined. As illustrated in Fig. 2A, lower panel, p65 was efficiently translocated from cytoplasmic to nuclear fraction in ⌬Zip2 cells, and its kinetics showed no difference from control LacZ cells. These data suggested that NF-B is normally activated and translocated into nuclei in cells expressing ATF3⌬Zip2 after TNF-␣ stimulation. In Fig. 2B, DNA binding activity of NF-B was examined. In both control and ⌬Zip2 cells, the binding activity of NF-B significantly increased in response to TNF-␣, and its extent of activation was apparently similar in both cells. Anti-p65 antibody caused a supershift of the NF-B complex, whereas anti-p50 antibody induced a slight supershift. These data indicate that NF-B activation, including IBa degradation, nuclear translocation, and DNA binding of NF-B, is not affected by ATF3⌬Zip2 in TNF-␣-stimulated cells.
Induced Expression of cIAP2 and XIAP Was Down-regulated in ATF3⌬Zip2 Cells after TNF-␣ Treatment-We next performed a microarray analysis of cDNAs to search for gene(s) responsible for increased sensitivity to TNF-␣-induced apoptosis of ATF3⌬Zip2-expressing cells. Among the 21,576 genes examined, 1133 genes (5.2%) and 398 genes (1.8%) were down-and up-regulated in ⌬Zip2 cells with more than a 2-fold difference from control LacZ cells after TNF-␣ treat-ment. The analysis revealed that two anti-apoptotic genes, cIAP2 and XIAP, were significantly down-regulated in ⌬Zip2 cells. As illustrated in Fig. 3A, RT-PCR analysis confirmed that expression of these two genes was remarkably inhibited in ⌬Zip2 cells. On the other hand, the expression of other anti-apoptotic factors, TRAF2 and FLIP, was less affected. To test the possibility that down-regulation of cIAP2 or XIAP is responsible for TNF-␣-induced apoptosis of ⌬Zip2 cells, we examined the effects of cIAP2 and XIAP on cell death. As shown in Fig. 3B, both cIAP2 and XIAP markedly inhibited cell death, suggesting that cell death of ATF3⌬Zip2-expressing cells may be caused by reduced expression of the anti-apoptotic factors cIAP2 and XIAP.

ATF3⌬Zip2 Suppresses cIAP2 Gene and NF-B-dependent Reporter
Activity-Human cIAP2 gene contains at least three NF-B-binding motifs in the proximal promoter region, and two of them are responsible for NF-B-dependent transcription in response to TNF-␣ (26). Therefore, we examined if the cIAP2 induction by TNF-␣ was affected by ATF3⌬Zip2 by using reporter assay. Fig. 4A, upper panel, illustrates the cIAP2 gene promoter from Ϫ500 to ϩ30 containing three NF-B motifs and a TATA sequence. As shown in Fig. 4A, lower left panel, ATF3⌬Zip2 had no effect on the basal activity of the cIAP2 gene promoter in the absence of TNF-␣. In the presence of TNF-␣, the activity was markedly increased, but this activity was significantly inhibited by coexpression of ATF3⌬Zip2 in a dose-dependent manner. Next the assays were performed in cells cotransfected with p65 and p50 expression plasmids. As shown in Fig. 4A, lower right panel, ATF3⌬Zip2 also inhibited the p65/p50-driven reporter activity of the cIAP2 gene, supporting the possibility that ATF3⌬Zip2 specifically inhibited the NF-Bdependent reporter. To address this possibility, we further performed assays using the NF-B-dependent reporter gene in which four tandem repeats of B sites were located upstream of the thymidine kinase gene promoter. As in Fig. 4B, the reporter activity was significantly suppressed by ATF3⌬Zip2 in both TNF-␣-stimulated and p65/p50-cotransfected cells. From these data, it is indicated that ATF3⌬Zip2 downregulates the TNF-␣-induced cIAP2 gene activation, most likely through inhibiting NF-B transactivation.
ATF3⌬Zip2 Binds to cIAP Gene Promoter in TNF-␣-stimulated Cells-Data above suggested that ATF3⌬Zip2 may inhibit cIAP2 gene expression by interacting with the gene promoter region. Thus, we examined whether ATF3⌬Zip2 is recruited to the cIAP2 gene promoter in TNF-␣-stimulated cells by ChIP assay. In Fig. 5, the proximal promoter region containing NF-B sites, but not the distal promoter region, was efficiently immunoprecipitated by anti-p65 antibody in TNF-␣-stimulated cells, consistent with the report that the NF-B element(s) are involved in TNF-␣-mediated induction of cIAP2 (26). When anti-FLAG antibody was used to immunoprecipitate FLAG-tagged ATF3⌬Zip2, the . Input was 10% of total. B, ATF3⌬Zip2 and p65 were expressed as a GST fusion (GST-⌬Zip2) and a fusion protein C-terminally tagged with histidines (His-p65), respectively. In vitro binding of these proteins was assayed by GST pulldown (left panel) or nickel-agarose beads pulldown assay (right panel). Input was 10% of total. C, full-length ATF3 as well as various C-terminal deletions of ATF3 were expressed as GST fusions, and their activity of binding to p65 was assayed by GST pulldown as in B, left panel. Input was 10% of total.
proximal region, but not the distal region, was also efficiently immunoprecipitated. In the absence of TNF-␣ treatment, however, neither anti-p65 nor anti-FLAG antibody did not immunoprecipitate the promoter, suggesting that the recruitment of ATF3⌬Zip2 as well as p65 is TNF-␣dependent. Taken together, it is indicated that ATF3⌬Zip2 is recruited to the proximal region of the cIAP2 gene in response to TNF-␣ treatment.
ATF3⌬Zip2 Is Recruited but CBP/p300 Is Displaced from the p65-Cofactor Complex in TNF-␣-stimulated Cells-NF-B forms the cofactor complex consisting of positive and negative regulators, including CBP/p300 and HDAC, thereby regulating NF-B-dependent transcription. Thus, we next examined whether ATF3⌬Zip2 interacts with this complex. For this purpose, U2OS cells expressing ATF3⌬Zip2 were stimulated with TNF-␣, and their cell extracts were subjected to immunoprecipitation assay. As shown in Fig. 6A, left panel, FLAG-tagged ATF3⌬Zip2 as well as CBP/p300 and HDAC1 were immunoprecipitated specifically with anti-p65 antibody but not with control IgG, indicating that ATF3⌬Zip2 interacts with the p65-cofactor complex. More intriguingly, the amount of CBP/p300 in the complex, but not of HDAC1, was significantly lower in ⌬Zip2 cells than in control LacZ cells, despite the fact that there was no difference in the total amount of CBP/p300 in the cell extracts. When ATF3⌬Zip2 was inversely immunoprecipitated by using anti-FLAG antibody, p65 was specifically immunoprecipitated (Fig. 6A, right panel). These data strongly support that ATF3⌬Zip2 is a component of the p65-cofactor complex, and CBP/ p300 is displaced from this complex in TNF-␣-stimulated cells.
ATF3⌬Zip2 Directly Interacts with p65 in Vitro-We next sought to identify the component(s) of the NF-B-cofactor complex with which ATF3⌬Zip2 directly interacts in TNF-␣-stimulated cells. To this end, an in vitro binding study was performed using recombinant ATF3⌬Zip2 and p65 proteins. As shown in Fig. 6B, GST-ATF3⌬Zip2 and His-tagged p65 directly interacted each other, as detected by GST-pulldown (left panel) or nickel-agarose beads (right panel) assay. We further performed a deletion study of ATF3 to delineate the region responsible for p65 binding. As illustrated in Fig. 6C, p65 was coprecipitated efficiently with ATF3⌬Zip2, whereas dC34 mutant that deletes the C-terminal 34 amino acids but shares the basic region with ATF3⌬Zip2 bound p65 only weakly. By contrast, the full-length ATF3 and the C-terminal deletions dC96 and dC141 showed no significant p65 binding. The data combined strongly indicate that ATF3⌬Zip2 directly binds p65, and the unique C-terminal 19-amino acid sequence of ATF3⌬Zip2 plays a role in the interaction. . Counteraction between ATF3⌬Zip2 and CBP in p65 binding, reporter assay, and cell death. A, GST-CBP2N fusion containing the N-terminal sequence of CBP from 117 to 738 amino acids was mixed with His-p65 and subjected to GST pulldown, and the p65 binding was measured by Western blot (left panel) as under "Experimental Procedures." In the right panel, the mixture of GST-CBP2N and His-p65 was incubated in the absence or presence of cell extracts from AdLacZ-or AdATF3⌬Zip2-infected U2OS cells, and GST pulldown assay was performed for p65 binding (upper panel). In the lower right panel, the mixture was subjected to pulldown of His-p65 using nickel-agarose beads, and the binding of GST-CBP2N and FLAGtagged ATF3⌬Zip2 was measured by Western blot. Input was 10% of total. B, U2OS cells (5 ϫ 10 4 cells) were cotransfected with 0.5 g of pLuc-cIAP2, 0.25 g of p65 expression vector, the indicated amounts of pMEFLAGATF3⌬Zip2, and 0.1 or 0.5 g of CBP expression vector pcDNA3/mCBP-HA, and assayed for luciferase activity. Repressed reporter activity of ATF3⌬Zip2 cells was significantly restored by CBP; *, p Ͻ 0.05. C, U2OS cells (6 ϫ 10 4 cells) were transfected with 2 g of pMEFLAGATF3⌬Zip2, 0.3 g of pCMV-GFP, and 1.5 or 3 g of pcDNA3/mCBP-HA, stimulated with TNF-␣, and cell death was assayed as under "Experimental Procedures." Increased cell death by ATF3⌬Zip2 was significantly suppressed in CBP-expressing cells; *, p Ͻ 0.05.

ATF3⌬Zip2 Counteracts CBP in p65 Binding, Reporter Assay, and
Cell Death-It has been shown that the p65 subunit of NF-B directly interacts with CBP or p300 (31,32), and the data above showed that ATF3⌬Zip2 also binds to p65. By contrast, no significant binding of ATF3⌬Zip2 with CBP was observed in our assay (data not shown). Therefore, we examined whether ATF3⌬Zip2 affects the association of p65 with CBP in vitro. For this purpose, p65 was incubated with GST-CBP2N, a GST fusion with the N-terminal portion of CBP from 117 to 738 amino acids, in the absence or presence of ATF3⌬Zip2. As shown in Fig. 7A, left panel, CBP2N specifically interacted with p65 in the absence of ATF3⌬Zip2 as reported (31). In the presence of ATF3⌬Zip2, however, the binding of p65 to CBP was suppressed in a dose-dependent manner, as detected by GST-pulldown (Fig. 7A, upper right panel) or nickel-agarose beads (lower right panel) assay. Furthermore, under this assay condition, ATF3⌬Zip2 was recruited into the p65 complex. By contrast, the control LacZ protein had no effect. These results indicated that ATF3⌬Zip2 interferes with the association of p65 and CBP, possibly through its direct binding to p65. We next determined whether CBP functionally antagonizes ATF3⌬Zip2 activity in both reporter and cell death assay. Fig. 7B showed that coexpression of CBP alleviated the repression of reporter activity by ATF3⌬Zip2 and restored the p65-dependent reporter activity. In the cell death assay, CBP relieved the cell death induced by ATF3⌬Zip2 in TNF-␣-stimulated cells (Fig. 7C). These results, taken together, strongly indicate that the transcriptional repression by ATF3⌬Zip2 depends on its ability to compete with CBP for p65 binding, and this suppression is responsible for the proapoptotic activity of ATF3⌬Zip2.
Knockdown of Endogenous ATF3⌬Zip2 Restored CBP/p300 Binding and Attenuated Cell Death in TNF-␣/Actinomycin D-treated Cells-To test whether the endogenous level of ATF3⌬Zip2 plays a role in vivo, U2OS cells were treated with TNF-␣ and actinomycin D. As shown in Fig. 8A, the ATF3⌬Zip2 expression was remarkably increased, and cell death was also concomitantly induced in these cells. Next, an RNA interference approach was employed to determine the effects of depletion of endogenous ATF3⌬Zip2. U2OS cells stably expressing ATF3⌬Zip2 siRNA exhibited a significant decrease of adenovirusdriven expression of ATF3⌬Zip2 protein but not full-length ATF3. Control cells for empty vector or GFP siRNA had no effect (Fig. 8B), demonstrating specific knockdown of ATF3⌬Zip2 by RNA interference. In Fig. 8C, cells were treated with TNF-␣ and actinomycin D. ATF3⌬Zip2 expression was specifically reduced in cells harboring ATF3⌬Zip2 siRNA, whereas control empty or GFP siRNA cells induced ATF3⌬Zip2 protein. Under this condition, cell death was partly but significantly suppressed in ATF3⌬Zip2 siRNA cells, compared with control empty or GFP siRNA cells. We further examined the effect of ATF3⌬Zip2 knockdown on the p65-cofactor complex. As shown in Fig.  8D, the addition of actinomycin D reduced the recruitment of CBP/ p300 compared with TNF-␣-treated cells, but it had no effect on the recruited amount of HDAC1. In contrast, the amount of CBP/p300 was significantly restored in cells with ATF3⌬Zip2 siRNA, whereas control empty or GFP siRNA cells showed no such effect. The data combined clearly indicated that the endogenous level of ATF3⌬Zip2 protein plays a role in regulating p65 binding to CBP/p300 and cell death in response to TNF-␣ and actinomycin D.

DISCUSSION
The stress response gene ATF3 is induced by various stress stimuli, and its activation is associated with cell death (2)(3)(4)(5), survival (11,12), and cell proliferation (9,10,13). Thus, ATF3 can function differently depending on cellular context. In this study, we reported that ATF3⌬Zip2, a splice variant, gains a novel activity of suppressing NF-Bdependent transcription, thereby sensitizing cells to apoptosis by TNF-␣ and actinomycin D.
ATF3⌬Zip2 inhibited the p65 binding to coactivator CBP in both overexpressed or endogenous levels, thus attenuating B-dependent transcription. Because ATF3⌬Zip2 is not capable of binding to DNA (19), this inhibitory activity of ATF3⌬Zip2 is considered to be different from the repressive effect of the full-length ATF3 (14 -16). Homo-or heterodimers of full-length ATF3 bind to the ATF/CRE motif of gene promoters and stabilize the repressive cofactors as proposed in the inhibitory cofactor model (14). By contrast, ATF3⌬Zip2 but not fulllength ATF3 directly interacted with the p65 subunit of NF-B, and this  ), respectively, as in A. ␤-Tubulin was a control for protein loading. Cell death was expressed as mean Ϯ S.D. of three independent experiments and significantly suppressed by ATF3⌬Zip2 siRNA; *, p Ͻ 0.05. D, U2OS cells (1 ϫ 10 7 cells) were treated as in C for 6 h. Cell extracts were prepared and subjected to immunoprecipitation (IP) by using anti-p65 antibody or control IgG, and the resulting p65 complex was assayed for p65, CBP, p300, HDAC1, and ATF3⌬Zip2 by Western blot (left panel) as under "Experimental Procedures." In the right panel, input was 10% of total.
interaction suppressed the p65 binding with coactivator CBP (Figs. 6 and 7). ATF3⌬Zip2 lacks the C-terminal 65-amino acid region of the full-length ATF3 but instead has novel sequence of 19 amino acids at the C terminus as shown in Fig. 6C (19). Our deletion study delineated the p65 binding activity of ATF3⌬Zip2 to the C-terminal region, suggesting a role of the C-terminal unique sequence, whereas dC34 mutant bound p65 very weakly (Fig. 6C). It is interesting to note that ATF3⌬Zip2, at the C-terminal junction, creates the LQKLP sequence that resembles the coregulator signature motif LXXLL that is present at the N terminus of PIAS3, another p65-binding repressor (38). It may be possible that ATF3⌬Zip2 interacts with p65 through this novel sequence. Alternatively, ATF3⌬Zip2 lacks the ability of dimer formation because of its loss of the leucine zipper region. This conformational change of ATF3⌬Zip2 may expose a novel structural surface for p65 binding, which is otherwise masked in the dimer structure of the full-length ATF3. Indeed, we found no significant interaction between the fulllength ATF3 and p65 in vitro (Fig. 6C), consistent with a previous report (46). Thus, it is highly likely that ATF3⌬Zip2 gains novel structure for p65 binding through alternative splicing.
The coactivators CBP/p300 interact with a variety of transcriptional activators and repressors to provide a physical link between these factors and the basal transcription machinery. CBP/p300 also recruits other histone acetyltransferases such as p/CAF, whose activity is required for NF-B-dependent transcription (34), and coordinate transcriptional pathways with chromatin remodeling via their intrinsic histone acetyltransferase activity. In this study, we showed that ATF3⌬Zip2 inhibited the association of CBP with p65 both in vivo and in vitro. On the other hand, there was no difference in HDAC1 binding to the p65 complex in ⌬Zip2 cells, compared with control (Fig. 6A). This may represent one of mechanisms by which ATF3⌬Zip2 causes repression of NF-B-dependent transcription, because the balance between CBP/p300 and HDAC1/2 is critical for transcriptional outcome. We speculate that recruitment of ATF3⌬Zip2 into the p65 complex inhibits the association of CBP but not of HDAC1. This imbalance of CBP against HDAC1 may result in dominant histone deacetylase activity and thereby lead to repression of gene transcription. However, at this moment, it is not clear how the binding of ATF3⌬Zip2 to p65 causes impaired interaction of CBP/p300. CBP/p300 interacts with p65 at their N-and C-terminal sequences, and the binding at the N terminus is observed at a higher salt concentration than that at the C terminus (31,32). As in Fig. 7A, ATF3⌬Zip2 efficiently displaced the binding of CBP at the N terminus, suggesting that ATF3⌬Zip2 may competitively bind to the same binding site of p65 as CBP/p300. Other possibility is that ATF3⌬Zip2 binding causes conformational change of p65 and inhibits its interaction with CBP. Detailed molecular mechanism for CBP displacement by ATF3⌬Zip2, however, must await further investigation.
The regulation of NF-B-dependent gene transcription through p65 binding is also reported for several factors. For instance, SNP1 (37) and PIAS3 (38) bind to p65 and down-regulate NF-B-dependent gene transcription. SNIP1, an inhibitor of the transforming growth factor-␤ signaling pathway that depends on CBP/p300, also binds to the C/H domain of CBP/p300 and thereby competes with each other for binding. In contrast, we could not find significant binding of ATF3⌬Zip2 to CBP (data not shown). Therefore, ATF3⌬Zip2 and SNIP1 may recognize p65 in a different mechanism. More recently, candidate tumor suppressor gene protein ING4 (39) and arginine methyltransferase CARM1 (47) are also reported to bind to p65 and repress and activate the B-dependent gene expression, respectively. Our study further supports the crucial role of p65 in regulating the association of multiple cofactors mediating between the B site(s) of gene promoter and transcription machinery. p65 (RelA) is also regulated by its own modification by phosphorylation. NF-B containing phosphorylated p65 associates with CBP/p300 and displaces HDAC1, ensuring that the phosphorylated form of p65 can activate transcription (33). In the present study, however, it was not determined whether the phosphorylation status of p65 was affected by ATF3⌬Zip2. Furthermore, p65 has been shown recently to associate with HDAC1/2 to repress anti-apoptotic gene expression in particular settings such as UVC or daunorubicin stimulation, in contrast to the case of TNF (48). NF-B, especially p65 activation, can promote cell death through the cascade downstream pathway of Egr-1 and GADD45 in UV-irradiated skin fibroblasts (49). ATF3⌬Zip2 is also induced by various stimuli other than TNF-␣, such as UV, genotoxic agents, and endoplasmic reticulum stress (19), under which NF-B is also activated. Therefore, the effect of ATF3⌬Zip2 on the ultimate cell death or survival may depend on the cell type and stimuli, although ATF3⌬Zip2 was remarkably induced by TNF-␣ in the presence of actinomycin D and played a pro-apoptotic role in the present study (Fig. 8). These reports, taken together, further support the possibility that altered activity of p65 by ATF3⌬Zip2 may be implicated in the cell fate determination in response to stress stimuli.
Among several isoforms of ATF3 splice variants that delete the C-terminal leucine zipper region, ATF3⌬Zip2 and ⌬Zip2c share the C-terminal novel sequence of 19 amino acids (17,19), whereas ATF3⌬Zip (14) and ⌬Zip3 (17) contain novel sequence of 3 and 5 amino acids, respectively. It is noteworthy that ⌬Zip2c is induced in response to amino acid or glucose deprivation. We speculate that ⌬Zip2c has similar activity to ATF3⌬Zip2 in p65 binding and plays a role in determining cell fate, although the absence of the N-terminal activation domain of 29 amino acids in ⌬Zip2c might confer the unique activity. The significance of complexity and functional diversity of these ATF3 splice variants in stress response, thus, requires more investigation.
Finally, we demonstrate that ATF3⌬Zip2 represses the NF-B-dependent transcription through its binding to p65 and displacing the CBP/p300 coactivator. This is the first report that illustrates the gain of novel function of stress response gene ATF3 via alternative splicing, and further provides an insight into its role in determining cell fate in stress response.