Effects of the Isoform-specific Characteristics of ATF6α and ATF6β on Endoplasmic Reticulum Stress Response Gene Expression and Cell Viability*

The endoplasmic reticulum (ER)-transmembrane proteins, ATF6α and ATF6β, are cleaved during the ER stress response (ERSR). The resulting N-terminal fragments (N-ATF6α and N-ATF6β) have conserved DNA-binding domains and divergent transcriptional activation domains. N-ATF6α and N-ATF6β translocate to the nucleus, bind to specific regulatory elements, and influence expression of ERSR genes, such as glucose-regulated protein 78 (GRP78), that contribute to resolving the ERSR, thus, enhancing cell viability. We previously showed that N-ATF6α is a rapidly degraded, strong transcriptional activator, whereas β is a slowly degraded, weak activator. In this study we explored the molecular basis and functional impact of these isoform-specific characteristics in HeLa cells. Mutants in the transcriptional activation domain or DNA-binding domain of N-ATF6α exhibited loss of function and increased expression, the latter of which suggested decreased rates of degradation. Fusing N-ATF6α to the mutant estrogen receptor generated N-ATF6α-MER, which, without tamoxifen exhibited loss-of-function and high expression, but in the presence of tamoxifen N-ATF6α-MER exhibited gain-of-function and low expression. N-ATF6β conferred loss-of-function and high expression to N-ATF6α, suggesting that ATF6β is an endogenous inhibitor of ATF6α. In vitro DNA binding experiments showed that recombinant N-ATF6β inhibited the binding of recombinant N-ATF6α to an ERSR element from the GRP78 promoter. Moreover, siRNA-mediated knock-down of endogenous ATF6β increased GRP78 promoter activity and GRP78 gene expression, as well as augmenting cell viability. Thus, the relative levels of ATF6α and -β, may contribute to regulating the strength and duration of ATF6-dependent ERSR gene induction and cell viability.

Stresses that alter the rough ER 3 environment can impair folding of proteins synthesized by this organelle (1)(2)(3)(4). Numer-ous proteins induced under such conditions are targeted to the ER, where they aid in nascent protein folding and thus, counteract the stress; this ER-initiated signaling process is known as the ER stress response (ERSR). ERSR elements (ERSEs) are located in the regulatory regions of many ERSR genes. One of the transcription factors that mediates ERSR gene induction via ERSEs is ATF6␣, a 670-aa ER trans-membrane protein (5, 6) ( Fig. 1A, ATF6␣). ER stress activates the proteolytic cleavage of ϳ400 aa from the N terminus of ATF6␣ (N-ATF6␣) (7), which translocates to the nucleus and activates numerous ERSR genes (8,9). The transcriptional activation domain (TAD) of N-ATF6␣ resides in the N-terminal portion of the protein, whereas the basic leucine zipper (b-Zip) and nuclear localization domains reside in the C terminus (Fig. 1B, N-ATF6␣) (8,10). N-ATF6␣ can bind directly to ATF6 binding sites (9), or it can combine with several other proteins to form a complex that binds to ERSEs and augments the induction of numerous ERSGs, such as the ER chaperone, glucose-regulated protein 78 kDa (GRP78) (8,9,(11)(12)(13). N-ATF6␣ exhibits potent transcriptional activity, however, it is susceptible to proteasome-mediated degradation, and mutations in the TAD that reduce N-ATF6␣ transcriptional activity decrease degradation (14). Several other potent transcription factors that exert rapid, transient effects exhibit similar coupling of transcriptional activation and degradation (15), including the virally encoded protein, VP16 (16). An 8-aa domain in VP16, called VN8, confers strong transcriptional activity and susceptibility to degradation, and mutations in VN8 that reduce VP16 activity decrease degradation (17,18). The TAD of ATF6␣ possesses a VN8-like sequence, and mutating it in ways known to decrease VP16 activity decrease ATF6␣ activity and degradation (14). To the best of our knowledge, the VN8 domain has not been found in any other mammalian transcription factor, including a second isoform of ATF6, ATF6␤.
Like ATF6␣, ATF6␤ is an ER-transmembrane protein (Fig.  1A, ATF6␤), and during ER stress proteolysis generates an N-terminal fragment of ϳ400 aa (19). N-ATF6␣ and N-ATF6␤ possess highly conserved b-Zip domains, which allow them to bind to ERSEs as homo-or heterodimers (20); however, the N-terminal regions are divergent. For example, the region of N-ATF6␤ corresponding to the VN8 of N-ATF6␣ differs in 5 of 8 aa in ways predicted from studies with VP16 to diminish transcriptional activity (21) (Fig. 1B, N-ATF6␤). In support of this prediction were findings that ectopically expressed N-ATF6␤ is a poor ERSR gene inducer (6) that exhibits much greater stability than N-ATF6␣ (14). Accordingly, although they can bind to the same regulatory elements, N-ATF6␣ and -␤ exhibit isoform-specific transcriptional activation and stability characteristics. Thus, N-ATF6␣ and -␤ might function in a combinatorial fashion to fine-tune the strength of ERSR gene activation.
In the present study, we examined the molecular mechanism and function of the isoform-specific characteristics of N-ATF6␣ and -␤, addressing the following hypotheses: 1) the isoform-specific characteristics of N-ATF6␣ and -␤ are conferred by their divergent N-terminal TADs; 2) N-ATF6␣-mediated transcriptional activation and rapid degradation are coordinated processes, and 3) the relative levels of N-ATF6␣ and -␤ impact ERSR gene induction and cell viability in ways consistent with roles of N-ATF6␤ as a transcriptional repressor of N-ATF6␣.

Methods
Replicates and Statistical Analysis-Unless otherwise stated in the legends or figures, each treatment was performed on three identical cultures, and each experiment was repeated at least three times. Representative experiments are shown. Statistical analyses were performed using a one-way ANOVA followed by the Student-Newman-Keul post-hoc analysis. *, #, or § ϭ p Ͻ 0.05; **, ##, § §, or ᮍᮍ ϭ p Ͻ 0.01.
Cell Culture-HeLa Cells were maintained in Dulbecco's modified Eagle's medium containing 10% fetal calf serum. For transfection experiments, HeLa cells were resuspended at 5-9 ϫ 10 6 cells per 400 l of cold Dulbecco's phosphate-buffered saline and electroporated in a 0.4-cm gap electroporation cuvette at 250 V and 950 microfarads using a GenePulser II Electroporator (Bio-Rad). The cells were then plated at a density of 0.5 ϫ 10 6 per 24-mm well for experiments involving luciferase and ␤-galactosidase enzyme assays, or 1.5 ϫ 10 6 per 35-mm well for experiments involving immunoblotting. Reporter assays and immunoblotting were carried out as previously described (22).

Plasmids
CMV-Galactosidase-CMV-␤galactosidase, which codes for a ␤-galactosidase reporter driven by the CMV promoter, was used to normalize for transfection efficiency.
N-ATF6␣-MER Fusion Protein-A PCR product composed of the nucleotides encoding aa 281-599 of the mouse mutated (G525R) estrogen receptor (MER, a gift from Dr. Michael Reth, Max-Plank-Institute, Freiburg, Germany) was cloned into the NotI/EcoR1 site of pCDNA3.1-3ϫ FLAG vector to create 3ϫ FLAG-MER. Subsequently, a PCR product of N-ATF6␣ was generated, which introduced a NotI site and removed the termination site after aa 373, was cloned into the XhoI/NotI site of 3ϫ FLAG-MER to create 3ϫ FLAG-N-ATF6␣-MER.

Small Interfering RNAs
The use of small interfering (si) RNA targeted against human ATF6␣ and -␤ has been described elsewhere (22). Briefly, HeLa cells were plated on 6-well plates at ϳ400 K cells per well, then transfected with 50 ng of the relevant dicer siRNAs, using Lipofectamine 2000 TM . After 24 h, cells were treated with or without tunicamycin (2 g/ml), and then examined by real-time quantitative PCR (see below), or, before tunicamycin treatment, they were removed from the plate with TripLE (Invitrogen), and re-plated in 96-well plates at ϳ10 K cells per well in preparation for viability assays. To examine viability, cells in 96-well plates were treated with or without tunicamycin (2 g/ml) and 2-deoxyglucose (3 mM) in serum-free media for 32 h. Cell viability was then assessed using an MTT Cell Proliferation Kit according to the manufacturer's protocol (Roche Applied Science). Samples were read at 570 nm in a VersaMax microplate reader (Molecular Devices, Downingtown, PA).

Isoform-specific Characteristics of N-ATF6␣ and -␤ Are Conferred by Their Divergent N-terminal TADs-Because we
previously showed the importance of the VN8 sequence for Constructs 1 and 2 are native N-ATF6␣ and N-ATF6␤, respectively. In the diagrams of constructs 3-5, the white boxes represent sequences from ATF6␣ and gray boxes represent sequences from ATF6␤. The locations of the boundaries from each ATF6 isoform are shown; for example, construct 3 is composed of ATF6␣-(1-114) fused to ATF6␤-(116 -392). The boundary amino acids were selected based on sequence homology. B, GRP78 promoter activity: HeLa cells were co-transfected with either an empty vector control (Con), or the ATF6 expression constructs shown, and GRP78-luciferase or pGL2P and CMV-␤-galactosidase, as in Fig. 2. After 48 h in culture, extracts were assayed for reporter enzyme activities, as   114). B, GRP78 luciferase or Gal4 luciferase: HeLa cells were transfected with the constructs shown (Con ϭ empty vector), and either GRP78-luciferase, or Gal4-luciferase, and after 48 h in culture, extracts were assayed for reporter enzyme activities, as described under "Methods." Rel Luciferase ϭ GRP78-luciferase/␤-galactosidase, or Gal4luciferase/␤-galactosidase. Shown are mean Rel Luciferase values Ϯ S.E. (n ϭ 3 cultures). In the GRP78-Luc panel, values for Con and construct 2 are in one group, and in the Gal4-Luc panel, constructs 3, 4, and 6 are in one group. For both panels, **, p Ͻ 0.01 are different from all other values in that panel, as determined using ANOVA followed by Newman-Keuls post hoc analysis. C, FLAG and Gal4 DBD immunoblots (IB): extracts from the cultures described in B were analyzed by SDS-PAGE and immunoblotting for FLAG or Gal4, as shown. Constructs 1-6 refer to the same constructs shown in A. All lanes were loaded with 30 g of protein, except those for construct 3, which were loaded with 3 g. The mean relative expression levels Ϯ S.E. are shown at the top of each gel. the transcriptional activity and rapid degradation of ATF6␣ (22), we assessed whether the lack of a VN8 sequence in ATF6␤ is responsible for its low transcriptional activity and high stability. Upon sequence alignment of N-ATF6␣ and -␤, we found that residues 64 -71 of ␤ correspond to residues 61-68 of ␣, the VN8-like region (Fig. 1B). Accordingly, residues 64 -67 of N-ATF6␤ (i.e. ATF6␤-(1-392)) were mutated to the same residues found in the VN8-like region of ATF6␣, which possess the Phe and Leu known to be required for optimal activity ( Fig. 2A, construct 3, ATF6␤-VN8-M1). We also prepared a mutation that converted the entire 64 -71 region N-ATF6␤ to be identical to the VN8 in ATF6␣ ( Fig.  2A, construct 4, ATF6␤-VN8-M2). The abilities of native N-ATF6␤ or the VN8 mutations to activate the promoter of the prototypical ERSR gene, GRP78, in HeLa cells were compared with N-ATF6␣ (i.e. ATF6␣-(1-373)). As previously seen (22), N-ATF6␣ exerted strong GRP78 promoter activation, whereas native N-ATF6␤ exhibited weak effects (Fig. 2B, constructs 1  versus 2). Among the ATF6␤ VN8 mutations, only ATF6␤-VN8-M2 exhibited detectible GRP78 promoter activation, although it amounted to only ϳ8% of N-ATF6␣ (Fig. 2B, construct 4). Previous studies showed that the relative levels of ectopically expressed ATF6␣ and -␤ are proportional to their half-lives (22). N-ATF6␣ was expressed in very low quantities (Fig. 2C, construct 1), consistent with its known short half-life, whereas all forms of N-ATF6␤ were expressed at much higher levels (Fig. 2C, constructs  2-4), suggesting that they exhibited relatively long half-lives, as previously shown for native N-ATF6␤ (22). Quantification demonstrated that N-ATF6␤, N-ATF6␤-VN8-M1, and N-ATF6␤-VN8-M2 were expressed at ϳ40-, 29-, and 20-fold higher levels than N-ATF6␣, respectively (Fig. 2C). Thus, although their expression levels and apparent half-lives decreased in coordination with the minor increases in activity, it was apparent that the low transcriptional activity and high stability of N-ATF6␤ were not due entirely to the lack of a consensus VN8 sequence, but that larger portions of ATF6␤ must be required to confer these isoform-specific characteristics.
To examine the effects of mutating larger portions of ATF6␣ and -␤, a series of domain-swap mutations were generated where the N terminus of N-ATF6␤ was replaced with progressively larger portions of the corresponding sequences from N-ATF6␣ (Fig. 3A). As expected, native N-ATF6␣ was a strong activator of the GRP78 promoter, whereas native N-ATF6␤ was much weaker (Fig. 3B, constructs 1 versus 2). However, when the N-terminal 115 or 190 aa of N-ATF6␤ were replaced with corresponding sequences from ␣, transcriptional activity increased progressively (Fig.  3B, constructs 3 and 4). Finally, when the N-terminal 321 aa of N-ATF6␤, representing all but the 71-aa b-Zip domain, were replaced by corresponding sequences from N-ATF6␣, GRP78 promoter activity increased to about the same level as that observed using native N-ATF6␣ (Fig. 3B, construct 5), suggesting that the b-Zip domains of N-ATF6␣ and -␤ were interchangeable. As expected, the level of expression of N-ATF6␤ was ϳ80-fold greater than N-ATF6␣ (Fig. 3C, constructs 1 and 2); moreover, expression levels of the chimeras declined coordinately as more sequences from N-ATF6␣ replaced corresponding sequences in ␤ (Fig. 3C, constructs [3][4][5], consistent with the hypothesis that the degradation rate of ATF6 coordinates with its transcriptional activity. When GRP78 promoter activity was normalized to the levels of ectopic N-ATF6␣ or -␤ protein expression, the only domain-swap mutant exhibiting activity approximating that of native N-ATF6␣ was construct 5 (Fig. 3D). Accordingly, these data suggested that, although the b-Zip domain of ATF6␤ can substitute for the b-Zip domain of ATF6␣ without much loss of function, most of the sequences lying to the N terminus of the b-Zip domain of ATF6␣ are necessary to confer the full transcriptional activity and rapid degradation characteristic of this ATF6 isoform.
N-ATF6␣-mediated Transcriptional Activation and Rapid Degradation Are Coordinated Processes-It is not known whether it is the sequences in the TAD of ATF6␣ that confer strong transcriptional activation and rapid degradation, or whether rapid degradation is a function of the engagement of ATF6␣ in a productive transcription complex. If the latter is true, then mutating the basic region of the b-Zip domain to disrupt binding of N-ATF6␣ to ERSEs should decrease transcriptional activation and decrease degradation. Consistent with this hypothesis was our finding that mutating the basic region of N-ATF6␣ (Fig. 4A, construct 2) to disrupt the binding of N-ATF6␣ to ERSEs (9) resulted in decreased GRP78 promoter activation (Fig. 4B, constructs 1 and 2) and increased N-ATF6␣ expression of Ͼ3-fold (Fig. 4C, FLAG   blot, constructs 1 and 2). To test this hypothesis in a heterologous gene expression system, we used a truncated form of the yeast transcription factor Gal4, Gal4-(1-147), composed of the Gal4 DBD, which does not possess a TAD. The binding of the Gal4 DBD to appropriate DNA sequences was assessed using a luciferase reporter driven by a neutral promoter flanked by tandem repeats of the Gal4 binding element. A mutation known to block the binding of Gal4-(1-147) to the Gal4 binding element (26) was introduced into a construct featuring the TAD of ATF6␣ without the ATF6 DBD, i.e. ATF6␣- (1-114), fused to the Gal4 DBD (Fig. 4A, constructs 5  and 6). As expected, the ATF6␣(1-114)/Gal4 DBD fusion protein without the DBD mutation exhibited robust transcriptional activation, compared with the Gal4 DBD alone (Fig. 4B, constructs 3 and 5); however, the ATF6␣(1-114)/ Gal4 DBD fusion protein harboring the DBD mutation exhibited no transcriptional activation (Fig. 4B, construct 6). Moreover, the level of expression of ATF6␣/Gal4 DBD-M was ϳ2-fold greater than that of ATF6␣/Gal4 DBD (Fig. 4C, Gal4 blot, constructs 5 and 6), whereas the level of expression of Gal4 DBD-M was actually somewhat lower than that of Gal4 DBD (Fig. 4C, Gal4 blot, constructs 3 and 4). These results are consistent with the hypothesis that rapid degradation of N-ATF6␣ requires its engagement in transcriptional activation.
To examine the relationship between transcriptional engagement and ATF6 degradation in a different model system, we designed a method for conditionally activating N-ATF6␣ in a ligand-dependent manner. For this purpose, we generated a construct encoding N-ATF6␣ fused to a fragment of the MER, which has no TAD or DBD, but features a tamoxifen-ligand-binding domain replacing the estrogenbinding domain. By analogy to the way MER affects other proteins to which is has been fused (27), we reasoned that, in the absence of tamoxifen, the MER would attract other cellular components, e.g. HSP90, which would block functional domains of ATF6, but that, upon tamoxifen binding, release of HSP90, among others, would reveal functional domains and allow full engagement of ATF6 in transcription (Fig. 5A). Accordingly, constructs encoding FLAG-MER or FLAG-N-ATF6␣-MER, where MER is fused to the C terminus of FLAG-N-ATF6␣, were prepared (Fig. 5B), and the abilities of each to activate the GRP78 promoter were examined. As expected, FLAG-MER exhibited essentially no activity (Fig.  5C, construct 1), whereas FLAG-N-ATF6␣ exhibited high activity that was affected very little by tamoxifen (Fig. 5C,  construct 3). In contrast, FLAG-N-ATF6␣-MER exhibited little activity in the absence of tamoxifen, but, upon tamoxifen addition, activity increased 3-fold, nearly equal to that of FLAG-N-ATF6␣ (Fig. 5C, construct 2). The protein levels of FLAG-MER were relatively high in the absence of tamoxifen (Fig. 5D, lanes 1 and 2), and actually increased by 1.6-fold in the presence of tamoxifen (Fig. 5D, lanes 3 and 4), which was somewhat expected, because tamoxifen stabilizes MER. The levels of FLAG-N-ATF6␣ were low and unchanged by tamoxifen (Fig. 5D, lanes 9 -12). However, the levels of FLAG-N-ATF6␣-MER were decreased by ϳ5-fold in tamoxifen-treated cells (Fig. 5D, lanes 5-8), suggesting coordination of tamoxifen-activated transcription and rapid degradation of FLAG-ATF6␣-MER.

Relative Levels of N-ATF6␣ and -␤ Impact ERSR Gene Induction and Cell Viability in Ways Consistent with Roles of N-ATF6␤ as a Transcriptional Repressor of N-ATF6␣-The
ATF6␣ loss-of-function mutations in this and previous studies exhibit decreases in degradation. We previously showed that N-ATF6␤ mimicked ATF6␣ loss-of-function mutations in terms of inhibiting N-ATF6␣-mediated transcription (22), but its effect on ATF6␣ expression level and degradation is not known. Accordingly, a construct encoding FLAG-N-ATF6␣ was used to distinguish it from HA-N-ATF6␤ on immunoblots, and the ratios of ectopically expressed FLAG-N-ATF6␣ and HA-N-ATF6␤ were varied by transfecting HeLa cells with different amounts of the appropriate plasmids.
In the first series of experiments, the level of FLAG-N-ATF6␣-encoding plasmid was held constant, while the level of the HA-N-ATF6␤-encoding plasmid was varied. As expected, GRP78 promoter activation by FLAG-N-ATF6␣ was inhibited as the level of HA-N-ATF6␤ was increased (Fig. 6A, transfections 1-3). FLAG and HA immunoblots showed that the quantity of HA-N-ATF6␤ increased as a function of increased plasmid, as expected (Fig. 6B, HA-ATF6␤); interestingly, the levels of FLAG-N-ATF6␣ also increased, even though each culture had been transfected with the same quantity of the FLAG-N-ATF6␣ plasmid (Fig. 6B, FLAG-ATF6␣). These results suggested that HA-N-ATF6␤ not only inhibited the ability of FLAG-N-ATF6␣ to activate the GRP78 promoter, but also increased its half-life. We examined degradation of FLAG-N-ATF6␣ using cycloheximide (CHX) to inhibit new protein synthesis, as previously described (28). The apparent degradation of FLAG-N-ATF6␣ was extremely rapid when no HA-N-ATF6␤ was co-expressed. Within 9 min of CHX addition, only 12% of the FLAG-N-ATF6␣ originally present remained (Fig. 6C, transfection 1, Blot A, versus transfection 1, Blot B). In contrast, the degradation rate of FLAG-N-ATF6␣ was reduced in the presence of HA-N-ATF6␤; moreover, as the level of HA-N-ATF6␤ was increased, degradation rate of FLAG-N-ATF6␣ decreased. For example, at intermediate or high levels of HA-FLAG-N-ATF6␤, 66 and 82% of the original FLAG-N-ATF6␣ remained after 9 min of CHX treatment (Fig. 6C, transfections 2 and 3

, Blot A versus Blot B).
In the second series of experiments, the FLAG-N-ATF6␣encoding plasmid was varied, while the HA-N-ATF6␤-encoding plasmid was held constant. As expected, HA-N-ATF6␤ alone conferred very little GRP78 promoter activation (Fig. 6D,  transfection 4), whereas, increasing the levels of FLAG-N-ATF6␣ increased GRP78 promoter activity (Fig. 6D, transfections 5 and 6). FLAG and HA immunoblots showed that the level of FLAG-N-ATF6␣ increased as more plasmid was transfected, as expected (Fig. 6E, FLAG-ATF6␣, transfections 4 -6); however, surprisingly, the levels of HA-N-ATF6␤ decreased, even though each culture had been transfected with the same quantity of that plasmid (Fig. 6E, HA-ATF6␤, transfections 5  and 6). These results suggested that FLAG-N-ATF6␣ can increase the degradation rate of HA-N-ATF6␤. Consistent with this hypothesis was the finding that, in the absence of FLAG-N-ATF6␣, ϳ58% of the original HA-N-ATF6␤ was still present following 17 min of CHX treatment (Fig. 6F, transfection 4, Blot  A versus B). In contrast, the degradation rate of HA-N-ATF6␤ was increased in the presence of FLAG-N-ATF6␣; moreover, as FLAG-N-ATF6␣ was increased, the degradation rate of HA-N-ATF6␤ increased. For example, at intermediate and high levels of FLAG-N-ATF6␣, 32 and 26% of the original HA-N-ATF6␤ was still present 17 min after CHX treatment (Fig. 6F, transfections 5 and 6, Blot A versus B). Taken together, the results of the experiments shown in Fig. 6 indicate that ATF6␣ and -␤ can influence each other, such that the isoform-specific transcriptional and degradation characteristics of each are dependent upon their relative levels. This finding is consistent with a mechanism whereby N-ATF6␣ and -␤ can regulate ERSR gene expression and cellular function in a combinatorial fashion.
Because ATF6␣ and -␤ can both bind to ERSEs (20), EMSAs were performed to assess the abilities of recombinant ATF6␣ and -␤ to compete for binding to an ERSE in the GRP78 gene. Incubation of nuclear extract from untreated HeLa cells with a labeled oligonucleotide that replicates ERSE-1 in the GRP78 gene resulted in the formation of complex 1 (Fig. 7A, lane 1). Formation of complex 1 has previously been shown to be due to binding of other nuclear proteins (e.g. NF-Y A, B, and C) to the ERSE in the absence of ATF6 (25). Adding recombinant ATF6␣-(1-373) or ATF6␤-(1-392) to the nuclear extract resulted in the formation of complexes 3 and 4, respectively, which migrated with relative mobilities consistent with the sizes of each form of ATF6 that was added (Fig. 7A, lanes 2 and  3). Adding a shortened form of ATF6, ATF6␣-(115-373), which should retain ERSE-binding ability, also exhibited a complex, complex 2, the mobility of which was consistent with the size of ATF6␣-(115-373) relative to the other forms of ATF6 used in this analysis (Fig. 7A, lane 4). When an excess of unlabeled wild-type GRP78 ERSE-1 oligonucleotide was added to the incubation, all of the complexes disappeared (Fig. 7A, lanes  5-8), as expected. However, excess unlabeled mutated GRP78 ERSE-1 was unable to compete for labeled oligonucleotide binding (Fig. 7A, lanes 9 -12). These results demonstrate the dependence of each complex on the presence of the native GRP78 ERSE-1.
To verify the presence of ATF6 isoforms in complexes 2-4, supershift EMSA experiments were carried out. Addition of preimmune antiserum to the nuclear extract did not alter the EMSA profile (Fig. 7B, lanes 1-4). However, addition of an antiserum specific for ATF6␣ altered the mobility of com-plexes 2 and 3, only (Fig. 7B, compare lanes 6 and 8 to lanes 2 and 4, respectively), whereas addition of an antiserum specific for ATF6␤ altered the mobility of complex 4, only (Fig. 7B To examine the cellular effects of altering the relative levels of endogenous ATF6␣ and -␤, we used an siRNA approach that was previously shown by immunoblotting to selectively reduce the quantity of each ATF6 isoform in HeLa cells (22). The selectivity of the siRNA reagents was verified here by quantitative reverse transcription-PCR assessment of ATF6␣ and -␤ mRNA in extracts from cells treated with siRNA targeted to green fluorescent protein (control), ATF6␣, ATF6␤, or another ERSR gene that is not the focus of this study, XBP1. Validating the specificity of the siRNAs was the finding that the ATF6␣-targeted siRNA reagent decreased the level of ATF6␣ and not ATF6␤ or XBP1 mRNA (Fig. 8A), whereas the ATF6␤-targeted siRNA decreased the level of ATF6␤ and not ATF6␣ or XBP1 mRNA (Fig. 8B). Knocking down ATF6␣ decreased basal and tunicamycin (TM)-stimulated GRP78 promoter activity by ϳ2to 3-fold (Fig. 9A, bars 1  versus 2 and 4 versus 5). In contrast, knocking down endogenous ATF6␤ had little effect on basal GRP78 promoter activity, but it increased TM-induced GRP78 luciferase by 2-fold (Fig. 9A, bars 4 versus 6). Coordinate with these results were the findings that knockdown of ATF6␣ decreased TM-induced GRP78 mRNA by ϳ2-fold (Fig. 9B,  bar 4 versus 5), whereas knockdown of ATF6␤ increased TM-induced GRP78 mRNA by ϳ1.5-fold (Fig. 9B, bar 6). Because many ERSR genes, including GRP78, encode proteins that foster protection, we examined the effects of knocking down endogenous ATF6␣ or -␤ on HeLa cell viability. We found that, although 32 h of TM treatment conferred no change in viability in cells treated with control siRNA (Fig. 9C; bars 1 versus 4), knockdown of ATF6␣ significantly decreased viability with or without TM ( Fig. 9C; bars 1 versus 2, and 4 versus 5), although knockdown of ATF6␤ significantly increased viability with or without TM ( Fig. 9C; bars 1 versus 3 and 4 versus 6). These findings indicate that the isoform-specific characteristics of ATF6␣ and -␤ can influence TM-stimulated GRP78 expression, as well as viability of HeLa cells in ways that are consistent with the protective aspects of N-ATF6␣ and the putative abilities of N-ATF6␤ to serve as an endogenous repressor of ATF6␣. Moreover, because knocking down ATF6␣ or ␤ altered viability-TM, it is apparent that even in the absence of TMmediated ER stress, ERSR genes, such as GRP78, must contribute to cell viability.

DISCUSSION
In this study we examined the structural features underlying the isoform-specific characteristics of ATF6␣ and -␤, the coordination and mechanism of ATF6␣ transcriptional activation and rapid degradation, and whether the relative levels of ATF6␣ and -␤ affect their binding to ERSEs and regulate ERSR gene induction and cell viability. Our findings showed that there is structural information spanning most of the N-terminal 300 aa of ATF6␣ and -␤ that is required for isoform-specific characteristics. We also found that the rapid degradation of ATF6␣ is coordinate with its engagement in an active transcription complex, the latter of which can evidently be modulated by ATF6␤. Lastly, we determined the ratio of ATF6␣ and -␤ that modulates ERSR gene induction, as well as cell viability, in a manner consistent with the hypothesis that ATF6␣ is a strong but labile transcriptional activator, whereas ATF6␤ is a weak, stable transcriptional activator.
In addition to transcriptional activity and the rate of degradation, the timing of ATF6␣ and -␤ activation following ER stress is likely to be another important, albeit, not thor-  2-4, 6 -8, and 10 -12, respectively, were prepared by in vitro transcription/translation and then added to HeLa cell nuclear extracts. The 32 P-labeled GRP78 ERSE-1 probe was added to initiate the binding reactions. Complex 1 has been shown to be to the direct binding of nuclear extract-derived proteins, e.g. NF-Y, YY1, etc., to the ERSE. Under these experimental conditions, the binding of these accessory proteins is required before ATF6 will bind. Unlabeled double-stranded oligonucleotides representing the native GRP78 EMSA-1, or the mutated GRP78 EMSA-1 (MM) were added to lanes 5-8 and 9 -12, respectively. B, supershift: EMSA was carried out as described in A, lanes 1-4, except for the addition of either pre-immune, ATF6␣, or ATF6␤ antisera to lanes 1-4, 5-8, or 9 -12, respectively. C, ATF6␣/␤ competition: EMSA was carried out as described in A, except for the addition of recombinant ATF6␣ and -␤ together with nuclear extracts in lanes 4, 7, 11, and 14  A, ATF6␣ mRNA: cultures were transfected with siRNA targeted against green fluorescent protein (GFP), ATF6␣, ATF6␤, or XBP1; 48 h later, cell extracts were analyzed for ATF6␣ mRNA by quantitative reverse transcription-PCR, as described under "Methods." Mean mRNA levels (% of maximum) Ϯ S.E., n ϭ 3 cultures are shown. *, p Ͻ 0.05 are different from all other values, as determined using ANOVA followed by Newman-Keuls post hoc analysis. B, ATF6␤ mRNA: cultures were transfected with siRNA and extracted as described in A, except ATF6␤ mRNA levels were assessed by quantitative reverse transcription-PCR, as described under "Methods." Mean mRNA levels (% of maximum) Ϯ S.E., n ϭ 3 cultures are shown. *, p Ͻ 0.05 are different from all other values, as determined using ANOVA followed by Newman-Keuls post hoc analysis. GFP-targeted siRNA was used as a control for an siRNA targeted to a non-expressed protein, and XBP1-targeted siRNA was used as a control to show that targeting a non-ATF6related transcript did not affect the levels of ATF6␣ or -␤ mRNA.
oughly studied isoform-specific characteristic. Although it is well known that both ATF6 isoforms are cleaved upon ER stress, to our knowledge, only one study showed that, depending on the stress, activation of ATF6␣ can occur earlier than that of ATF6␤ (19). Combined with their isoformspecific characteristics, sequential activation of the ATF6 isoforms (Fig. 10A) is consistent with the possibility that their relative levels could change as a function of time after ER stress, such that there is an initial, strong activation of ATF6-mediated ERSR gene induction, followed by modulation toward weak activation (Fig. 10B). One potential mechanism by which ATF6␣ and -␤ could regulate the strength of ER stress involves how these isoforms bind to ERSR genes. ATF6␣ and -␤ bind to ERSEs, and possibly other elements, as dimers, which, interact with the C subunit of the NF-Y A, B, and C trimer (19), as well as with other proteins, e.g. SRF (5), TFII-I (12), and perhaps YY1 (13). Together, these proteins evidently facilitate ERSR gene induction. Thus, it is conceivable that, as a result of isoform-specific rates of generation and degradation, the relative levels of ATF6␣ and -␤ in transcriptional complexes change during progression of the ERSR and that, as a result of differences in their transcriptional activities, ERSR gene induction is finely tuned, as shown in Fig. 10C. The results of the gel shift experiments in this study showed that ATF6␣ and -␤ can compete with each other for binding to the GRP78 ERSE (Fig. 7C), which lends further support to this hypothesis.
The mechanism governing the rate of degradation of ATF6 during transcription is not known. However, a great deal is known about the coupling of transcriptional activation and rapid degradation of other labile transcription factors, e.g. c-Myc, Gal4, VP16, SMAD2, STAT, and Hac1p (29). In those cases, the most active transcription factors are also very susceptible to ubiquitination and proteasome-mediated degradation, both of which evidently take place during engagement in transcription (15,30). This "unstable when active" phenomenon is thought to allow tight control over transcription by ensuring that the activation of target genes is linked to the ongoing synthesis of the transcriptional regulator (30). Accordingly, such transcription factors are usually potent target gene activators, which must effect their function in a transient manner, although, rapid degradation of transcription factors has been linked to modulation as well as augmentation of activity (31,32). Because ubiquitinligase/proteasome machinery exists in the nucleus and is FIGURE 9. Effect of ATF6␣ or -␤ siRNA on GRP78 induction and cell viability. A, GRP78-Luciferase: HeLa cells were transfected with GRP78-luciferase, CMV-␤-galactosidase, and the siRNA shown, and then treated with or without tunicamycin (2 g/ml) for 32 h. Cell extracts were then analyzed for reporter enzyme activities, as described under "Methods." Values shown are mean Ϯ S.E. (n ϭ 3). All-TM values are in one group, whereas **, ##, and § § ϭ p Ͻ 0.01 are different from all other values, as determined using ANOVA followed by Newman-Keuls post hoc analysis. B, GRP78 mRNA: HeLa cells were treated as in A, except they were not transfected with reporter enzymes, and RNA was extracted. Values shown are mean Ϯ S.E. (n ϭ 3). All -TM are in one group, while **, ##, and § § ϭ p Ͻ 0.01 are different from all other values, as determined using ANOVA followed by Newman-Keuls post hoc analysis. C, Viability: Hela cells were treated as in B, except after siRNA transfection, they were transferred into 96-well plates, treated with or without TM (2 g/ml) in 3 mM 2-doxyglucose (2-DG) for 32 h, then cell viability was determined, as described under "Methods." Values shown are mean Ϯ S.E. (n ϭ 3). **, ##, and § § ϭ p Ͻ 0.01 are different from all other values, as determined using ANOVA followed by Newman-Keuls post hoc analysis.