Induction of Grp78/BiP by Translational Block

Mammalian cells respond to endoplasmic reticulum (ER) stress by attenuation of protein translation mediated through the PERK-eIF2α pathway and transcriptional activation of genes such as Grp78/BiP encoding ER chaperone proteins. The disruption of PERK function or the blocking of eIF2α Ser51 phosphorylation fails to attenuate translation after ER stress and also results in substantial impairment of Grp78/BiP induction by ER stress. While the activation of the Grp78 promoter by the ATF6 pathway through the endoplasmic reticulum stress elements (ERSEs) is well documented, the molecular mechanism linking PERK activation to Grp78 stress induction is unknown. We report here that ATF4, a transcription factor whose translation is up-regulated by the PERK-eIF2α pathway, can activate the Grp78 promoter independent of the ERSE. The ATF4-activating site is localized to an ATF/CRE sequence upstream of the ERSEs and is distinct from the C/EBP-ATF composite site previously identified as the ATF4 binding site in the ER stress-inducible chop promoter. In vitro translated ATF4 binding to the ATF/CRE site requires other nuclear co-factors from non-stressed cells, forming a complex that exhibits identical electrophoretic mobility as a thapsigargin-stress induced complex. Here we have identified the closely related ATF1 and CREB1 as nuclear co-factors that form in vivo complexes with endogenous ATF4. ER stress induces CREB1 phosphorylation and ATF1/CREB1 binding to the Grp78 promoter. Through the use of adenoviral vector expression systems, we provide evidence that when ATF4 function is suppressed and its binding partners are not able to compensate for its function, Grp78 induction by Tg and Tu is partially inhibited. Our studies resolve a mechanism responsible for inhibition of Grp78 mRNA induction by ER stress in cells that are functionally null for PERK or devoid of eIF2α phosphorylation.

The endoplasmic reticulum (ER) 1 is an organelle where secretory proteins and cellular membrane proteins are made or processed as well as a major intracellular site for calcium storage. Eukaryotic cells have evolved multiple regulatory mechanisms to maintain ER homeostasis. Under physiologically or pharmacologically adverse conditions that result in the accumulation of malfolded proteins or depletion of the ER calcium store, the cells trigger a series of protective measures generally referred to as the unfolded protein response (UPR) (1)(2)(3). An immediate response is attenuation of translation initiation (4). Presumably, by reducing protein synthesis rates during ER stress, the load of protein substrates subjected to the ER folding machinery will be reduced. The second arm of the UPR is activation of genes encoding ER chaperones, folding enzymes and components of the ER protein degradation apparatus (5). Thus, the activation of the gene encoding GRP78/BiP, a 78 kDa glucose-regulated protein which functions as a ER molecular chaperone and a calcium-binding protein, has been used extensively as a biomarker for the onset of the UPR (6). In addition, ER-stressed cells undergo a process, referred to as ER-associated protein degradation, that targets malfolded proteins to proteasome degradation (7). ER stress can also trigger the apoptotic program, as exemplified by the induction of the gene encoding C/EBP homologous transcription factor (CHOP/ GADD153) (8 -10) and activation of ER-associated caspases (11,12). However, how cells balance between the survival and death responses with the respective induction of anti-apoptotic GRP78 and pro-apoptotic CHOP proteins remains to be determined.
Upon ER stress, the attenuation of protein synthesis in mammalian cells is mediated through the PERK-eIF2␣ pathway (13). PERK is an ER resident kinase serving as one of the ER stress sensors. Under normal culture conditions, the activity of PERK is inhibited through its binding to GRP78 (14). The release from GRP78 under ER stress conditions allows PERK to oligomerize, resulting in activation. Activated PERK phosphorylates eIF2 on the ␣ subunit at Ser 51 , preventing the GDP/GTP exchange and thus reduces protein translation (15). In mouse embryo fibroblasts derived from Perk Ϫ/Ϫ mice, the attenuation of protein translation following ER stress was abolished; further, chop activation was completely blocked and Grp78 mRNA induction was suppressed (16). In embryo fibroblasts derived from the eIF2␣ phosphorylation homozygous mutant mice, Grp78 mRNA induction by tunicamycin (Tu) was reduced to 1.5-fold from the 6-fold observed in wild type (17). These results provide the first evidence that translational control by the PERK-eIF2␣ signaling pathway directly regulates the transcriptional arm of the UPR, affecting both the pro-and anti-apoptotic components.
The abolishment of chop induction in cells devoid of PERK or with eIF2␣ mutated at Ser 51 can be explained by the dependence of ER stress-induced chop transcription on ATF4. The transcription factor ATF4, which belongs to the ATF/CREB protein family, is absent or present in very low amounts in non-stressed cells but is efficiently translated only when Ser 51 in eIF2␣ is phosphorylated (16,18). By binding to the C/EBP-ATF composite site (5Ј-ATTGCATCA-3Ј) of the human chop promoter, ATF4 activates chop transcription (9,10). A similar site (5Ј-GTTTCATCA-3Ј) is identified on the promoter of the human asparagine synthetase gene, which also binds ATF4 (19). However, no such site is present on the Grp78 promoter. Further, recent studies suggest that transcriptional activation of the Grp78 promoter by ER stress is strictly dependent on S2P-mediated proteolytic cleavage of the transcription factor ATF6, which specifically targets the ERSE (20,21). All Grp promoters contain multiple copies of highly conserved ERSEs with a consensus sequence 5Ј-CCAAT(N 9 )CCACG-3Ј that bears no resemblance to the above ATF4 binding sites (22,23). Other transcription factors that regulate the Grp78 promoter upon ER stress, including NF-Y, YY1, and TFII-I (24 -27), all act through the ERSE. This raises the question of how attenuation of translation through eIF2␣ phosphorylation could enhance Grp78 transcription in response to ER stress. To resolve this, it is necessary to identify the effector molecule that mediates this regulation.
We report here that ATF4 is a new activator of the Grp78 promoter. ATF4, also referred to as CREB2, is a member of the ATF/CREB family that can bind as homodimers or heterodimers to the cAMP responsive element (CRE) within the ATF/CREB family and with members of the AP-1 transcription factor family (18). We have discovered a conserved ATF4 binding site (5Ј-TGACGTGA-3Ј) upstream of the ERSEs in the mammalian Grp78 promoters distinct from the C/EBP-ATF composite site previously described for the chop and asparagine synthetase promoters. This ATF/CRE-like site was first identified as a binding site for a CREB-related protein, however, its role in ER stress induction of the Grp78 promoter was undefined (28). As ATF4 protein level is up-regulated upon ER stress, it binds to the ATF/CRE-like site of the Grp78 promoter in the presence of other transcription co-factors. Here we identify the closely related ATF1 and CREB1 as the nuclear cofactors that form in vivo complex with ATF4 and ER stress induces phosphorylation of CREB1 as well as ATF1/CREB1 binding to the Grp78 promoter. Through mutational analysis of this ATF/CRE-like site with exogenously expressed ATF4, we provide evidence that ATF4 is a new activator of the Grp78 promoter independent of the ERSEs. Further, through the use of adenoviral vector expression systems, we provide evidence that when ATF4 function is suppressed and its binding partners are not able to compensate for its function, Grp78 induction by Tg and Tu is partially inhibited. These results confirm that ATF4 can contribute to the ER stress induction of Grp78.

EXPERIMENTAL PROCEDURES
Cell Culture Conditions and Drug Treatment-NIH3T3 and 293 cells were grown in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum. The HeLa S3 cells were grown as described (28). For stress treatment, the cells were treated with 300 nM of thapsigargin (Tg) or 1.5 g/ml tunicamycin (Tu) for the time indicated.
Plasmids-The 3kb/LacZ reporter construct was constructed by subcloning the 3-kb rat Grp78 promoter into the HindIII site of pSV/LacZ. To construct the D170/LacZ and D300/LacZ, the 3-kb rat Grp78 promoter fragment was subcloned into the HindIII site of pUC8 plasmid. The deletions were carried out using the ExSitePCR site-directed mutagenesis kit (Stratagene, La Jolla, CA). The deletions (Ϫ170 to Ϫ70 and Ϫ300 to Ϫ70) were confirmed by DNA sequencing and subcloned back to the pSV/LacZ vector. The plasmid D170m was generated by site-directed mutagenesis using the QuikChange kit (Stratagene) and the mutagenic oligonucleotide primer, 5Ј-CCTGGGGGGCGTACCctgtc-gactcGTTGCGCTCCGAGGA-3Ј. The sequence of the mutated construct was confirmed by DNA sequencing. The construction of Ϫ169/LUC has been described (29). For the construction of Ϫ457/LUC and Ϫ457(mut)/ LUC, the StuI/BamHI fragment of Ϫ457CAT and LS(183-192)CAT, which includes the CAT gene and part of the Grp78 promoter (28) was replaced by the counterpart from the luciferase reporter Ϫ169/LUC. The plasmid ␤-actin/LacZ, in which the LacZ gene is driven by a chicken ␤-actin promoter, was a gift from Dr. Ebrahim Zandi (University of Southern California). The SV40/LacZ plasmid has been described (30). The human ATF4 cDNA was cloned by RT-PCR of mRNA isolated from 293 cells, confirmed by DNA sequencing, and subcloned in-frame into the XbaI and BamHI site of pCMV5Јmyc to obtain the Myc-tagged ATF4 expression plasmid, pMyc-ATF4.
Transfection and Reporter Assays-NIH3T3 or 293 cells were grown to 80% confluence in 6-well plates and co-transfected with pMyc-ATF4 or the empty vector pCMV5Јmyc with different reporter genes using superfect or polyfect, in accordance with the manufacturer's protocol (Qiagen, Valencia, CA). For each well of a 6-well plate of NIH3T3 cells, 0.5 g of Ϫ169/LUC, and 0.5 g of LacZ reporter (3kb/LacZ, D170/LacZ or D300/LacZ) were co-transfected with either 0.5 g of empty vector or 0.5 g of pMyc-ATF4 using 10 l of superfect or polyfect. For 293 cells, 0.7 g of each plasmid was used instead. Twenty-four hours after transfection, the cells were harvested and assayed for luciferase activity according to the manufacturer's instructions (Promega, Madison, WI) or ␤-galactosidase activity as described (31). The experiments were performed in duplicates or triplicates and repeated 3-4 times. The activity levels were analyzed by Student's t test (*, p Ͻ 0.05; **, p Ͻ 0.01).
In Vitro Protein Translation-The full-length ATF4 cDNA was subcloned into XbaI and BamHI of the pBlueScript KS(ϩ)vector. In vitro translation was carried out using a TNT T7-coupled reticulocyte lysate system (Promega) according to the manufacturer's instructions, either in the absence or presence of [ 35 S]methionine. The translated products were separated on 12% SDS-PAGE. The non-radiolabeled products were subjected to Western blot and the radiolabeled products were subjected to autoradiography.
Western Blot-Cell extracts or in vitro translated ATF4 were separated on 12% SDS-PAGE and transferred to nitrocellulose membranes. The ATF4 protein was detected using a rabbit polyclonal antibody (Santa Cruz Biotechnology Inc., Santa Cruz, CA) at a dilution of 1:1000. For detection of the closely related ATF1/CREB1, an anti-ATF1 mouse monoclonal antibody (Santa Cruz Biotechnology Inc.) that recognizes both proteins was used at 1:500 dilution. For detection of the phosphorylated forms, an anti-phosphorylated ATF1/CREB1 goat polyclonal antibody (Santa Cruz Biotechnology Inc.) was used at 1:200 dilution. The same filter was Western-blotted with a monoclonal antibody against ␤-actin (Sigma) at a dilution of 1:3000 to monitor for protein loading.
Electromobility Shift Assay-The gel shift and supershift assays were performed as described (28). The oligonucleotides used for probe were 5Ј-ctcgagGGCTGGGGGGCGCGTACCA GTGACGTGAGTTGCG-GAGGAG-3Ј and 5Ј-gtcgacCTCCTCCGCAACTCACGTCACTGGT ACG-CGCCCCCCAGCC-3Ј. The CRE-like site is underlined. For assays with in vitro translated ATF4, 5 l of in vitro translated ATF4 was added into the reaction mixture. The CRE-mutant competitor fragment spanning Ϫ300 to Ϫ170 was generated by PCR using the CRE-mutant/ LUC plasmid as template. All antibodies for supershift assays were from Santa Cruz Biotechnology Inc.
Chromatin Immunoprecipitation Assays-NIH 3T3 cells were grown in 15-cm plates under normal cell culture conditions to 80% confluence and treated with 300 nM Tg for 2 or 4 h or 10 mM dithiothreitol for 4 h and then fixed by addition of formaldehyde to the growth media to a final concentration of 1%. To harvest NIH 3T3 cells, plates were rinsed with cold PBS, covered with 10 ml of 5% fetal bovine serum in PBS, and then scraped. Chromatin was prepared using a kit from Upstate Biotech Inc. (Lake Placid, NY) according to the recommendations of the manufacturer, with twenty-four 5-s sonication pulses at 10-s intervals, which yielded chromatin fragments of an apparent size of 100 -400 bp. An aliquot from each sample representing 5% of the total volume was removed to be used as the input fraction and was processed with the eluted immunoprecipitates beginning at the cross-link reversal step. Equal amounts of chromatin from each sample were incubated at 4°C overnight with 5 l of anti-ATF1/CREB1 antibody (Santa Cruz Biotechnology Inc.) or 5 l of anti-H3 antibody (Santa Cruz Biotechnology Inc.). Formaldehyde-induced cross-linking was reversed (4 h at 65°C) and the DNA was purified by phenol-chloroform extraction and ethanol precipitation. Purified DNA from the input and IP samples was subjected to 35 cycles of PCR and the products were run on a 1.5% agarose gel and visualized with EtBr staining. The primers used for the endogenous Grp78 promoter were: 5Ј-CATTGGTGGCCGTTAAGAATGAC-CAG (forward) and 5Ј-AGTATCGAGCGCGCCGTCGC (reverse), yielding a 223-bp product.
Co-immunoprecipitation-NIH3T3 cells were grown in 15-cm plates to about 80 -90% confluency. The cells were either non-treated or treated with Tg or Tu for 6 h and then harvested in Triton X-100 lysis buffer and the immunoprecipitation with an anti-ATF1 mouse monoclonal antibody (Santa Cruz Biotechnology Inc.) was carried out as described (29). The immunoprecipitated complexes were then resolved by 12% SDS-polyacrylamide gel electrophoresis and subjected to Western blot with an anti-ATF4 rabbit polyclonal antibody. The same membrane was re-blotted with anti-ATF1 antibody to control for immunoprecipitation of ATF-1 in each of the samples.
ATF4 Adenoviral Vector Construction-Wild-type and DN mutant ATF4 cDNAs were expressed using the AdEasy adenoviral vector system (32) provided by Bert Vogelstein (Howard Hughes Institute, Johns Hopkins University). Expression plasmids pEFmATF4myc and pEF/ mATF4 M were kindly provided by Jawed Alam (Department of Biochemistry and Molecular Biology, Louisiana State University Health Sciences Center). These plasmids contain wild-type murine ATF4 cDNA and a DN mutant (ATF4⌬RK) encoding murine ATF4 protein with a seven amino acid substitutions within the DNA-binding domain ( 292 RYRQKKR 298 to 292 GYLEAAA 298 ) (33). To clone into the AdEasy system, the wt and ⌬RK coding sequences were PCR amplified from these vectors using a 5Ј-NotI-containing (underlined) forward primer (5Ј-AACAACAACGCGGCCGCTGTCGTGAACACCATGACCGAG-3Ј) and a 5Ј-HindIII-containing (underlined) reverse primer (5Ј-GTTGTT-GTTAAGCTTTAGACTATGCG GCCCCATTCAG-3Ј). The PCR reaction was carried out with Taq polymerase (Applied Biosystems, Foster City, CA) under the following conditions: 95°C 1 min, 55°C 2 min, 72°C 1 min, for 35 cycles followed by a 7 min 72°C incubation. The PCR products were cloned into the pAdTrack-CMV adenoviral shuttle vector encoding kanamycin resistance and containing a second expression cassette encoding green fluorescent protein (GFP). The shuttle vectors were then linearized with the restriction enzyme PmeI and electroporated into DY329 electro-competent cells along with the adenoviral backbone plasmid pAdEasy-1. Clones containing recombinant plasmids pAE-ATF4 and pAE-ATF4⌬RK, formed by homologous recombination, were subsequently selected for kanamycin resistance and identified by plasmid size in conjunction with endonuclease analysis. The coding sequences and insertion points of pAE-ATF4 and pAE-ATF4⌬RK inserts were confirmed by DNA sequencing. Recombinant adenoviral vectors were generated by transfecting the recombinant plasmid into the mammalian 293 packaging cell line. Transfected cultures were maintained until the percentage of cells exhibiting green fluorescence approached 100%. Virus was obtained by freeze-thaw lysis of the cells in PBS, followed by clarification of the lysate by centrifugation at 6000 ϫ g at 4°C for 10 min. Lysate was used to infect large-scale cultures of 293 cells for production of viral stock, again by freeze-thaw lysis of the cells in PBS, followed by clarification of the lysate by centrifugation at 6000 ϫ g at 4°C for 10 min. Expression of ATF4 and ATF4⌬RK proteins by infected mammalian cells was confirmed by Western blotting analysis.
RNA Blot-NIH3T3 cells were grown in 60 mm dishes until 60 -70% confluent. The cells were infected either with the empty virus, or virus overexpressing the wild-type ATF4 or the mutant ATF4⌬RK in DMEM for 6 h. The medium was subsequently changed to DMEM with 10% fetal bovine serum. The cells were further cultured for 24 h before they were either non-treated or treated with Tg or Tu for 6 h. The conditions for the extraction of total RNA and the Northern blot analysis have been described (29).

RESULTS
Exogenous ATF4 Activates the Grp78 Promoter but Not through the ERSEs-To investigate the functional role of ATF4 in the induction of the Grp78 promoter under ER stress conditions, we first co-transfected into NIH3T3 cells a plasmid expressing the Myc-tagged human ATF4 (pMyc-ATF4) with two Grp78 promoter reporters (Fig. 1A). The 3kb/LacZ reporter contains a 3-kb rat Grp78 promoter driving the expression of the LacZ reporter gene. In addition to the TATA element and three tandem copies of the ERSE, the 3-kb Grp78 promoter fragment also contains a CRE-like site, additional CCAAT motifs and several putative Sp1 sites. In Ϫ169/LUC, the regulatory elements upstream of the ERSEs and the TATA element were eliminated and this shorter Grp78 promoter fragment drives the expression of the luciferase reporter gene. We observed that the exogenously expressed ATF4 activated the 3kb/LacZ by about 2.5-fold (Fig. 1B). In contrast, overexpression of ATF4 had no effect on the activity of Ϫ169/LUC (Fig.  1C). To further confirm that the differences between the two reporter gene responses to ATF4 were not due to differences arising from separate transfection experiments, we co-transfected ATF4 together with both 3kb/LacZ and Ϫ169/LUC into the same cells and measured the LacZ and luciferase activities respectively. Consistent with our previous results, exogenous ATF4 only activated 3kb/LacZ but not Ϫ169/LUC (Fig. 1D). The same co-transfection experiments were also performed in 293 cells, and similar results were observed (data not shown). Collectively, these results suggest that ATF4 can act as a transcription activator for the Grp78 promoter, but the transactivation is not mediated through the ERSEs.
ATF4 Activates Grp78 through a Cis-element Located between Ϫ300 and Ϫ170 of the Promoter-To locate the ATF4 target site on the Grp78 promoter and to confirm that ATF4 activation of the Grp78 promoter does not require the ERSEs, we constructed D170/LacZ, D300/LacZ and D170m/LacZ ( Fig.  2A). D170/LacZ, with a deletion spanning Ϫ170 to Ϫ70, is identical to 3kb/LacZ with the only difference being the internal deletion eliminating all three ERSEs. D300/LacZ extended the 5Ј-border of the internal deletion to Ϫ300, resulting in the elimination not only of the three ERSEs but also of the CRElike site, three CCAAT elements and two SpI sites. D170m/ LacZ is identical to D170/LacZ with the additional mutation of the CRE site through site-directed mutagenesis.
First, we determined whether the Grp78 promoter absolutely requires the ERSEs for ER stress induction. For these experiments, either 3kb/LacZ, D170/LacZ, D170m/LacZ, or SV40/LacZ was transfected separately into NIH3T3 cells. The SV40/LacZ reporter was used for normalization of cell viability upon stress treatment. The transfected cells were treated with Tg, which depletes the ER calcium store and is a potent ERstress inducer of the Grp78 promoter (34). In the context of the 3-kb Grp78 promoter, elimination of all three ERSEs resulted in a substantial drop in fold induction (from 4.7-1.8-fold) by Tg, nonetheless there is residual stress inducibility which was largely eliminated when the CRE element upstream of the ERSEs was destroyed (Fig. 2B). While these results reaffirm the importance of the ERSEs toward stress induction of Grp78, they also reveal a potential role of the CRE element in allowing the Grp78 promoter to respond to ER stress.
We next examined the effect of exogenously transfected ATF4 on the wild-type and the two internal deletion mutations of the Grp78 promoter by performing co-transfections in NIH3T3 cells. We noted that the internal deletions of the promoter affected the basal level promoter activity, with D170 and D300 reduced to 60 and 20% respectively of that observed for 3kb/LacZ (Fig. 2C). However, while the exogenously expressed ATF4 could still activate D170/LacZ, it had no effect on the D300/LacZ (Fig. 2C). Further calculation showed that the fold induction of 3kb/LacZ and D170/LacZ by ATF4 were comparable, such that both promoters were activated by about 3-fold (Fig. 2D). Thus, these results suggest that the ATF4 responsive element might be localized in the region between Ϫ300 and Ϫ170 of the rat Grp78 promoter.
ATF4 Activates the Grp78 Promoter through a ATF/CRElike Cis-element Upstream of the ERSE-Computer prediction using the TFSEARCH program (35) showed that between Ϫ300 and Ϫ170 of the rat Grp78 promoter, there is an ATF/CRE-like site. This sequence, TGACGTGA, spanning Ϫ190 to Ϫ183, is identical to the consensus CRE sequence TGACGTCA with the exception of one base. Since ATF4 belongs to the ATF/CREB family, and within the Ϫ300 to Ϫ170 region we did not find any other ATF/CRE-related site, one possibility is that ATF4 activates the Grp78 promoter through this CRE site. To test this, we constructed two luciferase reporter genes, Ϫ457/LUC and CRE-mut/LUC (Fig. 3A). In Ϫ457/LUC, the expression of the luciferase gene is driven by a Grp78 promoter fragment spanning from Ϫ457 to Ϫ29, which includes the TATA element, three ERSEs, the ATF/CRE site and upstream sequences up to Ϫ457. In the CRE-mut/LUC, the ATF/CRE-containing sequence was specifically mutated while the three ERSEs remained intact (Fig. 3B). The CRE-mut/LUC showed a lower induction level by Tg and to a lesser extent a lower basal level (Fig. 3C). By co-transfection with the ATF4 expression plasmid, we observed that Ϫ457/LUC is activated by ATF4 in a dosagedependent manner, whereas the activation by ATF4 was greatly impaired by mutation at the ATF/CRE site (Fig. 3D). Thus, these results show that ATF4 activates the Grp78 promoter through the ATF/CRE element.
ATF4 Binds to the ATF/CRE Site Following Up-regulation by ER Stress-To determine whether ATF4 regulates the Grp78 promoter directly or through induction of other factors, we examined whether ATF4 can bind to the ATF/CRE site on the Grp78 promoter. First, we prepared ATF4 by in vitro translation using rabbit reticulocyte lysate. A major protein band of the expected 47 kDa size was observed by [ 35 S]methionine labeling (Fig. 4A, lane 1) and its identity as ATF4 was confirmed by Western blot using antibody directed against ATF4 in the presence of a mock control (Fig. 4A, lanes 2 and 3). However, in gel shift assays, in vitro translated ATF4 by itself did not bind the reannealed synthetic oligonucleotides spanning Ϫ210 to Ϫ171 containing the ATF/CRE site, as no new complex was observed besides the nonspecific band also present with probe alone in the presence of reaction buffer containing bovine serum albumin (Fig. 4D, lanes 1 and 2).
We next determined the ATF4 protein level in NIH3T3 cells following Tg treatment. In agreement with an earlier report (10), we detected a very low basal level of ATF4 in non-stressed cells, and after 4 -8 h of Tg treatment, the ATF4 protein level was greatly elevated, as compared with the relatively even ␤-actin levels (Fig. 4B). The ATF4 level in the non-stressed HeLa nuclear extract was also very low and increased substantially upon Tg treatment (Fig. 4C). Gel shift assays using the control and Tg-treated HeLa nuclear extracts showed that while two major complexes (I and III) were present in both control and Tg-treated extracts (Fig. 4D, lanes 3 and 5), there was a third complex (II) present only in Tg-treated extract. Complex III has been previously shown by supershift assays to contain the DNA-binding protein Ku that has an affinity for DNA termini (28). While in vitro translated ATF4 by itself was unable to bind the probe (Fig. 4D, lane 2), when it was added to the control nuclear extract, a new complex was detected that exhibited identical electrophoretic mobility as the Tg stressinducible complex II in the Tg-treated nuclear extract lane (Fig. 4D, lane 4). Formation of the new complex II was further enhanced over the other complexes when the amount of the control nuclear extract was decreased to 0.2-fold and the amount of in vitro translated ATF4 was increased 2-fold (Fig.  4D, lane 6). Thus, in vitro translated ATF4 can bind to the ATF/CRE site of the Grp78 promoter in the presence of nuclear FIG. 1. Exogenous ATF4 activates a 3-kb Grp78 promoter but not a short Grp78 promoter containing three ERSEs. A, schematic representation of the rat Grp78 promoter reporter genes 3kb/LacZ and Ϫ169/LUC. The locations of the TATA box, ERSE, and the CRE element are indicated. Additional CCAAT sequences and their orientation with respect to the TATA element are represented by arrows. The lollipop symbols indicate the occurrence of GC-rich sequences similar to Sp1 binding sites. B, ATF4 activates the 3-kb LacZ reporter gene. NIH3T3 cells were co-transfected with 3kb/LacZ and either pMyc-ATF4 or control pCMVmyc plasmids. Twenty-four hours after transfection, the cells were harvested and assayed for ␤-gal activity. The relative promoter activities are plotted with standard deviations and the promoter activity in cells transfected with control plasmid was set as 1. The stars ** indicate p Ͻ 0.01. C, Ϫ169/LUC were co-transfected with either pMyc-ATF4 or pCMVmyc plasmids into NIH3T3 cells. In addition, ␤-actin/LacZ was co-transfected to control for transfection efficiency. Twenty-four h after transfection, the cells were harvested and assayed for luciferase and ␤-gal activity. The luciferase activity in each group is plotted with ␤-gal activity normalization. D, 3kb/LacZ and Ϫ169/LUC were transfected into the same cells with either pMyc-ATF4 or control pCMVmyc plasmids. Twenty-four hours after transfection, the cells were harvested and assayed for ␤-gal or luciferase activity. The ␤-gal and luciferase activities in each sample are plotted respectively with the reporter activities in the control group set as 1. The stars ** indicate p Ͻ 0.01. extract from non-stressed cells. The new complex has identical electrophoretic mobility as the Tg-stress inducible complex observed with nuclear extract prepared from Tg-stressed cells.
To further confirm that complex II binds specifically to the ATF/CRE site, gel shift assays using Tg-treated HeLa nuclear extract were performed in the presence of different competitors. An CRE-mutant competitor was generated by PCR using an upstream primer starting at Ϫ300 and a downstream primer at Ϫ29 of rat Grp78 promoter using Ϫ457(mut)/LUC as a template. We also noted that a longer electrophoresis time separated complex II into a doublet band, referred to below as IIA and IIB. Using the unlabeled probe (Ϫ210/Ϫ171 of rat Grp78 promoter) as a competitor, as was expected, all three complexes (I, II, and III) were competed away in a dosage dependent manner (Fig. 4E, lanes 2 to 4). Complex III consisting of the abundant Ku protein was most resistant to the competition. In contrast, the CRE-mutant competitor spanning Ϫ300 to Ϫ29 was able to compete away complexes I and III but not the complex II doublet (Fig. 4E, lanes 5 and 6). These results indicate that complex II specifically binds to the CREsite and no other site between Ϫ300 and Ϫ29 can bind ATF4. In addition, it shows that complexes I and III bind sequences outside the ATF/CRE site and therefore can be titrated away by the CRE-mutant competitor. Collectively, our results show that ATF4 is either absent or present in minimal amounts in nonstressed NIH3T3 cells; following ER stress, its level increases greatly and can bind to the Grp78 promoter at the ATF/CRE site in the presence of other nuclear factors to act as a transcription activator.
Identification of ATF1 and CREB1 as the Nuclear Co-factors of ATF4 -ATF4 has been demonstrated to form heterodimers with a variety of bZiP proteins including Fos, Jun, and C/EBP (18,36). To determine the components of the Tg-stress inducible complex II that binds specifically to the CRE site, antibodies against ATF4, Fos, and Jun were added to the Tg-nuclear extract prior to addition of the probe in gel shift assays. We also tested for the presence of ATF1 and CREB1, which are closely related members of the ATF/CREB family, using an antibody that can recognize both proteins. Confirming that ATF4 is a component of complex II, addition of increasing amounts of anti-ATF4 antibody correlated with greater suppression of complex II formation and inverse enhancement of complex III formation (Fig. 5A, lanes 3 and 4). Strikingly, addition of anti-ATF1/CREB1 antibody specifically supershifted complex II with no effect on complex I or III (Fig. 5A, lanes 5 and 6). In contrast, addition of anti-c-Fos or anti-c-Jun antibody showed no effect on any of the complexes and addition of bovine serum albumin further enhanced the formation of complex II (Fig. 5A,  lanes 2, 7, and 8). Further, whereas the anti-ATF1/CREB1 antibody was highly effective in eliminating complex II, addition of either anti-C/EBP(C19) antibody which is specific for C/EBP␤, or anti-C/EBP(⌬198) antibody, which can react with C/EBP␤, C/EBP␣, C/EBP␦, and C/EBP⑀ were without effect (Fig. 5B).
Complex II consists of a doublet with a faster migrating IIA and a slower migrating IIB band (Fig. 5, A and B). Addition of the anti-ATF4 antibody suppressed formation of both bands and addition of the anti-ATF1/CREB1 antibody supershifted both bands. Western blots with whole cell extracts confirmed that the anti-ATF1/CREB1 antibody can specifically recognize the faster migrating ATF1 (35 kDa) and the slower migrating CREB1 (43 kDa) (Fig. 5C). Thus, these results are consistent with band IIA representing an ATF4/ATF1 complex and band IIB representing an ATF4/CREB1 complex, and in agreement with our earlier observation that recombinant CREB1 can bind the ATF/CRE site (28). Unlike ATF4, the level of ATF1/CREB1 remained relatively constant after Tg treatment (Fig. 5C). Further, using an antibody that recognizes the phosphorylated forms of both ATF1 and CREB1, we observed strong CREB1 phosphorylation following 4 h of Tg treatment while ATF1 phosphorylation was below detection limit (Fig. 5D, lanes 1-4). As a positive control, both CREB1 and ATF1 were phosphorylated in UV-treated cells (Fig. 5D, lane 5).
To test whether ER-stress induced binding detected in vitro can be observed in vivo, chromatin immunoprecipitation (ChIP) assays were performed with NIH3T3 cells subjected to Tg or DTT treatment. The latter creates protein with disrupted disulphide bonds, leading to malfolded protein formation in the ER. Using the anti-ATF1/CREB1 antibody as the immunoprecipitating antibody and equal amounts of input DNA, we detected minimal ATF1/CREB1 binding to the Grp78 promoter in non-stressed cells; upon 2-4 h of Tg treatment, there was gradual increase in ATF1/CREB1 binding (Fig. 6A, lanes 1-7). High level binding of ATF1/CREB1 to the Grp78 promoter was also detected after 4 h of dithiothreitol of treatment (Fig. 6A,  lane 8). In contrast, core histone H3 binding to the same promoter region was similar before and after Tg treatment (Fig.  6A, lanes 9 -11). Serial dilution of the PCR reactions confirmed that the band intensities shown in Fig. 6A were within the linear range of the PCR reactions (Fig. 6B).

ER-stress Induced Complex Formation Between Endogenous
ATF4 and ATF1-Western blot analysis of whole cell lysates showed that ATF1 and CREB1 were constitutively expressed and their levels were relatively constant in control and Tgtreated cells (Fig. 5C). This is in contrast with the dramatic up-regulation of ATF4 by either Tg (Fig. 4B) or Tu treatment (Fig. 7A, lanes 5 and 6). Immunoprecipitation assays followed by immunoblot analysis showed in non-stressed cells, we did not detect ATF4 in complex with ATF1/CREB1 (Fig. 7A, lane  3). Upon Tu or Tg stress, ATF4 was readily detected in complex with ATF1/CREB1 (Fig. 7A, lanes 4 and 7). We note that the anti-ATF1/CREB1 antibody immunoprecipitated ATF1 more efficiently than CREB1 (Fig. 7A, lanes 3, 4, and 7), as compared with immunoblots of whole cell extracts using the same antibody (Fig. 7A, lanes 5, 6, and 8). Collectively, these results show that ER stress induces the complex formation of endogenous ATF4 with ATF1/CREB1 in vivo, thus providing a possible mechanism for ER-stressed induced ATF4 binding to the Grp78 promoter through complex formation with ATF1 and CREB1.
Dominant Negative Mutant of ATF4 Suppresses ER-stress Induction of Grp78 mRNA-Since ATF4 belongs to a multiprotein family and can form heterodimers with partner proteins with compensatory function, we tested the requirement of ATF4 in ER-stress induction of endogenous Grp78 by overexpressing a dominant negative mutant ATF4⌬RK. This mutant protein has been previously shown to sequester ATF4 as well as its binding partners into nonfunctional heteromeric complexes and effectively block their function as transcription activators (19,33). As a control, we also overexpressed the wildtype ATF4 protein that will sequester the same binding partners into functional complexes. However, like the mutant ATF4⌬RK, wild-type ATF4 will also compete away the same binding partners from forming complexes that do not contain ATF4. Thus, if the effect is due to blocking of ATF4 function, we would only observe suppression with the mutant protein. If it is due to sequestration of a binding partner that binds to the promoter as part of a complex that does not include ATF4, both wild-type and mutant proteins will block expression.
We first determined through green fluorescence protein expression that over 90% of the cells were infected (data not shown). Immunoblot of whole cell extracts confirmed expression of ATF4 wild-type and mutant proteins in the infected cells (Fig. 7B). The infected cells were either non-treated or subjected to Tg and Tu treatment and RNA was extracted for determination of Grp78 mRNA level. Our results revealed a 30and 40-fold induction of Grp78 mRNA by Tg and Tu, respectively in cells infected with the empty vector (Fig. 7C, lanes   1-3). Overexpression of the dominant negative mutant ATF4⌬RK slightly reduced the Grp78 mRNA basal level (from 1.0 to 0.85) but substantially suppressed the induced level by Tg and Tu (Fig. 7C, lanes 4 -6). In contrast, overexpression of the wild-type protein slightly increased the Grp78 mRNA basal level (from 1.0 to 1.35) and the overall induction level by Tg and Tu (Fig. 7C, lanes 7-9). These results show that when ATF4 function is blocked by the dominant negative mutant and its binding partners are also sequestered such that they cannot compensate for its function, ER stress induction of endogenous Grp78 mRNA is attenuated. DISCUSSION The Grp78 gene provides a model system to study how the ER chaperones are regulated by the UPR at a transcriptional level. Recent studies have primarily focused on the key transcription factor ATF6 and its proteolytic cleavage by the S1P/ S2P system upon ER stress (20,26,29,(37)(38)(39). Another recent discovery is the Ire1p-mediated XBP-1 pathway that in turn is regulated by ATF6 (21, 40, 41). Both the ATF6 and IreIp-XBP-1 pathways activate Grp78 through the ERSE elements, which we confirm here as important cis-elements for Grp78 stress induction. However, these pathways cannot explain why transcription of Grp78 induced by Tu is substantially suppressed in cells with homozygous mutation at an eIF2␣ phosphorylation site that resulted in blockage of PERK-mediated translation arrest in response to ER stress. We made several novel observations while resolving this puzzle.
Our discovery of the evolutionarily conserved ATF4 target site on the Grp78 promoter and the ability of exogenously expressed ATF4 to transactivate the Grp78 promoter provide evidence that ATF4 directly interacts with the Grp78 promoter and is a new activator for Grp78, which is a major target of the UPR. The ATF4 pathway is independent of ATF6 processing and does not require the ERSE on the Grp78 promoter. In support of this conclusion, deletion of all three ERSEs has no effect on activation of the Grp78 promoter by ATF4. Furthermore, the nuclear form of ATF6, while able to induce the 3 kb/LacZ reporter gene by about 20-fold over control, is without effect on D170/LacZ (data not shown).
The ATF4 binding site that we have identified on the Grp78 promoter is distinct from the C/EBP-ATF composite site of the chop and the asparagine synthetase gene promoter that also binds ATF4. The ATF4 binding site in the Grp78 promoter is nearly identical to the ATF/CRE consensus binding site and its conservation between the human and rat Grp78 promoter is consistent with it being a functional regulatory element of Grp78 promoter activity (28). Previously, we reported that this ATF/CRE-like site not only serves as a basal regulatory element but is also required to maintain high promoter activity under ER stress conditions (28). Using the luciferase reporter systems described here, we confirmed the previous observation that CRE-site mutation decreased both the basal level and the activity induced by Tg. These results raise the question of how this element can affect both the basal Grp78 promoter activity and ER-stress inducibility. Since ATF4 is either absent or present in very low levels in non-stressed cells, it is likely that other transcription factors bind to this site under normal culture conditions and contribute to the basal activity of the Grp78 promoter, thus explaining the drop in basal promoter activity when the ATF/CRE site is mutated. In support of this, in vivo footprinting using HeLa nuclear extract showed constitutive occupancy of this site (42). Gel mobility shift assays using HeLa nuclear extract from non-stressed cells further revealed multiple factors binding to this site, one of which is antigenicallyrelated to CREB1 and another containing the Ku protein (28). We propose here that as ATF4 level is highly elevated after ER stress, it complexes with other nuclear co-factors and/or undergoes post-translational modification, allowing it to bind the ATF/CRE site and contribute toward Grp78 promoter activation. Thus, when the ATF/CRE site is mutated, ATF4 binding is lost, resulting in a drop of ER stress inducibility of the Grp78 promoter. Nonetheless, Grp78 promoter can still be ER-stress induced through the ERSEs utilizing ATF6, YY1, NF-Y, TFII-I and possibly XBP-1. This scenario is not unique to the Grp78 promoter since mutation of the composite C/EBP-ATF site of the chop promoter also suppresses both the basal and ER stress induced promoter activity (10). Both observations could be attributed to the ATF site being involved in the control of basal expression activity through binding of basal transcription factors while ER stress replaces or augments these factors with ATF4 and its nuclear co-factors. Further, our finding that in vitro-translated ATF4 by itself is unable to bind to the Grp78 promoter and that binding requires specific components from the nuclear extracts such as co-factors or post-translational modifiers provides a plausible explanation why ATF4 increase under specific stress conditions may be sufficient to induce other promoters but not Grp78 (19).
In search of ATF4 interactive partners relevant to the ER stress response, we discovered that ER stress induces complex formation between endogenous ATF1/CREB1 and ATF4, correlating with the appearance of novel ATF4 complex binding to the ATF/CRE site of the Grp78 promoter in gel shift assays. Using such assays, we were able to resolve two ATF4 containing complexes in Tg-treated nuclear extracts. According to the relative molecular size of ATF1 and CREB1, our results are consistent with the faster migrating complex containing ATF4 and ATF1, and the slower complex containing ATF4 and CREB1. Further, we have previously observed that recombinant CREB1 can bind the ATF/CRE site of the Grp78 promoter (28). While we were able to detect in vivo binding of ATF1/ CREB1 to the Grp78 promoter, attempts to immunoprecipitate ATF4 using anti-ATF4 antibody were unsuccessful in both ChIP and immunoprecipitation assays (data not shown). This could be due to either a weak antibody titer or the unavailability of the ATF4 epitope in the complex.
In contrast to ATF4, which undergoes dramatic increase in protein level after ER stress, the level of both ATF1 and CREB1 are relatively constant. While ATF1 is not phosphorylated in NIH3T3 cells subjected to 300 nM Tg treatment, Tg induces strong phosphorylation of CREB with kinetics parallel to Grp78 induction (34). This new result provides the physiologi-cal relevance for our previous observation that the Grp78 promoter can transactivated by CREB that has been phosphorylated and therefore activated (28). Interestingly, in human melanoma cells a higher dose of Tg (1 M) can cause transient phosphorylation of both CREB1 and ATF1 within 30 min of the drug treatment (43), and ATF1 is phosphorylated in response to EGF stimulation (44). Thus, in different cell types and under different conditions, post-translational modification of ATF1/ CREB1 such as stress-induced phosphorylation can play an important roles in modifying the ability as transcription cofactors. Future studies are required to dissect whether such modifications are required for Grp78 activation.
Since ATF1 and CREB1 can compensate for each other functionally (45), it is likely that both ATF1 and CREB1 can both act as a cofactor for ATF4. Likewise, it is also possible that other members of the ATF4 protein family can substitute for its function. To circumvent the complications due to functional redundancy, we utilized adenoviral vector to overexpress a dominant negative mutant of ATF4 that can sequester ATF4 and its binding partners. This mutant has been used previously to inhibit basal and cadmium-dependent activation of the mouse heme oxygenase-1 gene distal enhancer by blocking the formation of functional ATF4/Nrf-2 complex and to inhibit the activation of the human asparagine synthetase proximal promoter by blocking the formation of functional ATF4/C/EBP complex (19,33). Using the asparagine synthetase system, we confirmed that the wild-type and the mutant ATF4 proteins expressed by the adenoviral vectors functioned as expected (data not shown). In the infected cells, we observed that only overexpression of the dominant negative mutant ATF4 but not the wild-type protein resulted in suppression of ER-stress induced Grp78 mRNA. This provides direct evidence that when ATF4 function is suppressed and its binding partners are not able to compensate for its function, Grp78 induction by ER stress is suppressed. Similarly, the dominant negative ATF4 mutant but not the wild-type protein was able to suppress endogenous Grp78 mRNA induction by homocysteine, glutamine-starvation and anoxia in retinal pigmented epithelial cells. 2 Thus, the contribution of ATF4 to Grp78 induction applies to multiple forms of stress. While this manuscript is in FIG. 6. ATF1/CREB1 is recruited to the Grp78 promoter following ER stress in vivo. A, chromatin immunoprecipitation assays to detect ATF1/CREB1 binding to the Grp78 promoter in vivo. NIH3T3 cells were treated with 300 nM Tg or 10 mM dithiothreitol for the time (in hours) indicated, formaldehyde cross-linked, and chromatin was immunoprecipitated with the antibodies indicated on top. Purified, immunoprecipitated or input DNA were analyzed by PCR using primers recognizing the mouse Grp78 promoter (the region amplified includes from Ϫ239 to Ϫ16 relative to the transcriptional start site). PCR products of input (lanes 1-4) or immunoprecipitated DNA with anti-ATF1/CREB1 antibody (lanes 5-8) or anti-histone H3 antibody (lanes 9 -11) were electrophoresed on 1.5% agarose gel and stained with ethidium bromide. B, serial dilutions (1; 1:3, 1:9) of DNA from the same samples of B (lanes 5-8) were analyzed by PCR for the confirmation of results within the linear range.
preparation, it was reported that in MEF derived from ATF4 Ϫ/Ϫ mouse, Grp78 induction by Tg treatment was largely intact (46). As discussed above, ATF4 belongs to a multiprotein family with functional redundancy. Thus, one possibility is that the ATF4 dominant negative mutant was able to block the functions of the compensating factors and that is why the repressive effect could be detected. Another possibility for this is that in the ATF4 Ϫ/Ϫ MEF, there are some other ATF4 compensating factor(s) for the Grp78 stress induction, and these factor(s) are no longer available in stable cell lines like NIH3T3 or retinal cells. Since the unfolded protein response as exemplified by Grp78 induction is of central importance for maintenance of cellular homeostasis, it is not surprising that the cells have evolved multiple compensatory pathways to ensure stress induction of Grp78.
ATF1/CREB1 and its related protein family members are known to mediate transcriptional responses to various extra-cellular signals, including that of calcium efflux from the intracellular stores such as the ER and is important for cell survival (45). GRP78 exhibits anti-apoptotic properties and its up-regulation has also been reported in a variety of human cancers, correlating with tumorigenicity and malignancy (6). Specific overexpression of GRP78 can confer drug resistance in cancer cells through inhibition of caspase activation (47). Thus, GRP78 as a potential novel target gene for ATF1/CREB1 can mediate its survival effect and contribute to cancer growth and resistance to therapy. In support of the CRE regulatory site in Grp78 expression, preliminary data observed with Grp transgenic mouse models suggested that the in vivo Grp78 transcription may not be entirely dependent on the ERSE, suggesting the involvement of novel pathways independent of ATF6 and other factors acting exclusively through the ERSE (48, and data not shown). Further, mouse models with disrupted PERK and ATF4 function results in increased tissue-specific apopto- FIG. 7. ER stress-induced formation of endogenous ATF4 and ATF1/CREB1 complex and functional contribution of ATF4. A, endogenous ATF1/CREB1 and ATF4 associate with each other following ER stress. NIH3T3 cells were either non-treated or treated with Tu or Tg for 6 h. The cell lysates were subjected to immunoprecipitation with normal mouse IgG (lanes 1 and 2) or antibody against ATF1/CREB1 (lanes 3, 4, and 7). Whole cell extract (WCE) from the non-treated (Ϫ) or treated cells were loaded as a positive control for the Western blot (lanes 5, 6, and 8). The immunoprecipitated complexes were resolved by SDS-PAGE and subjected to Western blot against ATF4 (upper panel) or ATF1/CREB1 (lower panel). The positions of ATF1, CREB1, and ␤-actin are indicated. B, cell lysates prepared from NIH3T3 cells which were infected with empty vector (v) or infected with adenoviral vector expressing the dominant negative form (ATF4⌬RK) (lane 2) or wild-type ATF4 (wt) (lane 3) were subjected to immunoblot with anti-ATF4 and anti-␤-actin antibodies. C, RNA blot analysis of Grp78 mRNA levels in mock-infected cells (lanes 1-3), cells infected with adenoviral vector expressing wild-type ATF4 (Wt) (lanes 4 -6) or the dominant negative form (ATF4⌬RK) (lanes 7-9). The cells were either non-treated (Ϫ), or treated with Tg or Tu as indicated on top. The level of GAPDH mRNA was also determined and used as loading control for the RNA samples. The RNA band intensities were quantitated by phosphorimaging, normalized against the GAPDH levels and plotted below. The level of Grp78 mRNA in the non-treated, empty vector-infected cells was set as 1.
sis (49,50). This raises the question of what are the relevant critical physiologic targets mediating this effect. Our studies suggest that in addition to general translational arrest, upon PERK activation the transcription induction of anti-apoptotic proteins such as GRP78 by ATF4 may in part contribute to the survival response in specific tissues. In conclusion, mammalian cells have evolved highly versatile mechanisms to respond to ER stress and ATF4 represents a novel pathway linking PERK/ eIF2␣ to transcriptional activation of a major chaperone protein in the ER that may contribute to cell survival.