Coordination of ATF6-mediated Transcription and ATF6 Degradation by a Domain That Is Shared with the Viral Transcription Factor, VP16*

ATF6 is a 670-amino acid endoplasmic reticulum (ER) transmembrane protein that is cleaved in response to ER stress. The resulting N-terminal fragment of ∼400 amino acids translocates to the nucleus and activates selected ER stress-inducible genes, such asGRP-78 and sarco/endoplasmic reticulum ATPase, which are required for cell survival. In studying the mechanism of ATF6-activated transcription, we found that when HeLa cells were transfected with a plasmid encoding ATF6-(1–373), ER stress-inducible reporter gene activation was high, but ATF6-(1–373) expression was low, unless a proteasome inhibitor was added. In contrast, transfection with a plasmid encoding ATF6-(94–373) resulted in low reporter activation and high expression of ATF6-(94–373), which was independent of the proteasome inhibitor. Thus, the information responsible for transcriptional activation and proteasomal degradation must lie within the N-terminal 93 amino acids of ATF6. This portion of ATF6 was found to be homologous to the herpes simplex viral protein, VP16. One 8-amino acid domain of particular interest in this region of ATF6 is 75% identical to the VN8 region in VP16. VN8 is required for VP16-mediated transcription, as well as rapid degradation of VP16 by proteasomes. Point mutations in the VN8-like region of ATF6 caused a loss of transcription, increased expression levels, and an increase in half-life. Thus, the potent transcriptional activities and rapid degradation of ATF6 and VP16 require the VN8 domains in each protein. Homology searches indicate that ATF6 is the only eukaryotic protein known that possesses an active VN8 domain, raising questions about how this domain evolved and the functional importance underlying its appearance in only these two transcription factors.

Stresses leading to alterations in the ER 1 environment can result in the incorrect folding of nascent proteins in the ER (1). A number of genes induced under such conditions are targeted to the ER, where they aid in folding and/or counteract the stress (2)(3)(4)(5)(6). A regulatory element located in many of these genes is the ER stress-response element (ERSE), which is required for transcriptional induction of many ER stress-response genes (7). Activating Transcription Factor 6 (ATF6), a member of the ATF/cAMP-response element-binding protein family of transcription factors (8), is required for induction of numerous ER stress-response genes that possess the ERSE sequence (9), such as GRP-78 and SERCA2. Along with other proteins, ATF6 binds to the ERSE sequences in these genes, an event that is required for transcriptional induction.
ATF6 is composed of 670 amino acids and resides in the ER membrane, probably as a result of the hydrophobic sequence between amino acids 378 and 398 ( Fig. 1A) (10). Upon ER stress, the cytosolic N-terminal portion of ATF6 is released as a result of regulated intramembrane proteolysis (10,11). The proteases responsible for this cleavage are apparently the same as those required for sterol regulatory element-binding protein maturation and probably cleave ATF6 in and/or near the intra-ER membrane region (12). Following proteolytic cleavage, the N terminus of ATF6, which possesses several putative nuclear localization signals and a basic leucine zipper (b-Zip) domain between residues 308 and 369, translocates to the nucleus where it combines with several other proteins to form an ERSE-binding complex that is responsible for the induction of ER stress-responsive genes (7,10).
Although the ATF6 transcriptional activation domain (TAD) is known to lie in the N-terminal half of the protein, which is released from the ER by regulated intramembrane proteolysis (13,14), little is understood about how ATF6 activates transcription. During the course of experiments designed to address this problem, we found that many of the forms of ATF6 that displayed potent transcriptional activity were expressed at extremely low levels. This led us to the present study where we examined the hypothesis that transcriptionally active forms of ATF6 are susceptible to rapid degradation.

Cell Culture
HeLa Cells were maintained in Dulbecco's modified Eagle's medium containing 10% fetal calf serum. 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 luciferase and ␤-galactosidase assays, 1.5 ϫ 10 6 per 35-mm well for Western blots, or 3 ϫ 10 6 per 60-mm dish for metabolic labeling experiments. by five GAL4-binding sites (15), was obtained from R. Davis (University of Massachusetts Medical School).
Luciferase-After cell lysis and centrifugation, as described above, 100-l samples of cell lysate were combined with 100 l of luciferase buffer (the above described lysis buffer containing 0.3 mM D-luciferin and 3 mM ATP). An Optocompt II luminometer (MGM Instruments, Inc.) was used to measure light emission of each sample for 10 s. Relative luciferase activities were determined by dividing luciferase values by ␤-galactosidase values. All values shown are the mean of three cultures Ϯ S.E.

Western Analyses
Cells were extracted in 2ϫ Laemmli buffer containing 1 mM dithiothreitol, 1 mM phenylmethylsulfonyl fluoride, and 15 l of 2-mercaptoethanol, boiled for 5 min, fractionated by 10% SDS-PAGE, and then transferred to a polyvinylidene difluoride membrane. Membranes were then probed with either a Gal4 dbd antiserum (Santa Cruz Biotechnology, sc-510) or a FLAG antiserum, M2 antibody (Sigma, F-3165).

Pulse-Chase Experiments
HeLa cells were transfected with 30 g of the test expression vector, as described above, and 3 ϫ 10 6 cells were plated on 60-mm dishes. Following a 14 -18-h incubation in serum-containing medium, the cells were rinsed three times with warm Hanks' buffer (Invitrogen) and incubated for 2 h with 250 Ci of [ 35 S]methionine/cysteine (Easytag Express Protein Labeling Mix, PerkinElmer Life Sciences) diluted in 2 ml of methionine/cystine-free Dulbecco's modified Eagle's medium (Invitrogen) supplemented with 5% dialyzed fetal calf serum (Invitrogen). Following a 2-h incubation, this medium was removed, and the plates were washed twice and then incubated with 3 ml of chase medium (Dulbecco's modified Eagle's medium containing 10% fetal calf serum, 2 mM unlabeled methionine, and 2 mM unlabeled cysteine). At the indi-cated times, the cells were rinsed three times with phosphate-buffered saline and then scraped into 100 l of lysis buffer (50 mM Tris, pH 7.5, 1 mM EDTA, 1 mM EGTA, 1% Triton X-100, 5 mM sodium pyrophosphate, 10 mM sodium glycerophosphate, 50 mM NaF, 150 mM NaCl) containing 0.1% SDS. The lysate was then diluted with 400 l of lysis buffer containing no SDS so that the final concentration of SDS was 0.2%. The lysate was cleared by centrifugation, and the supernatant was incubated 12-18 h at 4°C with 2 g of FLAG antibody (Sigma, F-3165) followed by incubation with protein G-Sepharose beads and elution with SDS-PAGE sample buffer. Immunoprecipitated material was then resolved by SDS-PAGE (10% SDS gel), dried down, and exposed on a PhosphorImaging screen.

RESULTS
Previous studies (13,14) demonstrated that the region responsible for transcriptional activation lies within the N-terminal 273 amino acids of ATF6. To resolve better the location of the ATF6 transcriptional activation domain (TAD), expression constructs were prepared that encode proteins composed of the yeast Gal4 DNA binding domain (Gal4 dbd; Gal4-(1-147) (17)) fused to the N terminus of ATF6-(1-273) and to other C-terminally truncated forms of ATF6 (Fig. 1A). When tested in HeLa cells, the transcriptional activities of the fusion proteins were apparently high, even after removal of all but the Nterminal 38 amino acids of ATF6 (Fig. 1B). Gal4/ATF6-(1-28) did not support significant transcriptional activity, indicating a minimal requirement for the N-terminal 38 amino acids.
When expression of the fusion proteins was assessed, Gal4/ ATF6-(1-273) and Gal4/ATF6-(1-93) were undetectable ( Fig  1C, lanes 1 and 3), unless the cells had been incubated with the proteasome inhibitor, ALLN (N-acetyl-Leu-Leu-Nle-CHO) (Fig.  1C, lanes 2 and 4). In contrast, the more truncated fusion proteins were expressed at high levels in a manner that was generally independent of whether ALLN was present (Fig. 1C, lanes [5][6][7][8]. Thus, when the protein expression level was taken into account, the specific transcriptional activity of Gal4/ATF6-(1-93) was much greater than that of Gal4/ATF6-(1-38), indicating that a major transcriptional activation domain resides between amino acids 38 and 93 of ATF6. 2 These results also indicated that the fusion proteins displaying the greatest transcriptional activities were expressed at the lowest levels. Because expression of transcriptionally active fusion proteins was enhanced considerably by ALLN, these forms of ATF6 must be rapidly degraded by proteasomes in a manner that coordinates with their transcriptional activities.
To evaluate these interesting properties in a context relevant to the ER stress response, the abilities of various forms of ATF6 ( Fig. 2A) to activate a reporter gene flanked by an ERSE 3 were assessed. ATF6-(1-373) maximally activated ERSE-inducible transcription 4 ; however, ATF6-(39 -373) and -(94 -373) displayed only 10 and 2% of maximal activity, respectively (Fig. 2 In Fig. 1B, the observed luciferase levels for Gal4/ATF6-(1-93) and -(1-38) are fortuitously similar, which may give the appearance that they possess similar transcriptional potencies. However, as can be seen in Fig. 1C, the expression level of Gal4/ATF6-(1-38) is very high, while the expression level of 1-93 is very low. Thus, the ability of each molecule of Gal4/ATF6-(1-38) to stimulate transcription is a great deal lower than the transcriptional induction ability of each molecule of Gal4/ATF6-(1-93). This is due in part to the removal of the sequences in 52-93, which is addressed further in Fig. 2. 3 One of the genes induced during the ER stress response is glucose response protein, GRP-78, which is targeted to the ER lumen and serves a chaperone function (4 -6). A regulatory element in the GRP-78 gene, the ER stress response element (ERSE), is required for transcriptional induction during the ER stress response. This element has been found in other ER stress-responsive genes (7). Accordingly, it has been used in the present study as an indicator of ATF6-inducible transcription. 4 Forms of ATF6 that include the b-Zip region (308 -369), but which lack the ER transmembrane domain (380 -400), localize to the nucleus (10) and activate ATF6-dependent transcription of genes possessing ERSEs. 2B), emphasizing the importance of the N-terminal 93 amino acids. To resolve better the location of critical domains in this region of ATF6, several constructs harboring an internal deletion were prepared. ATF6-(1-51/94 -373) and -(39 -51/94 -373) displayed only 12 and 3% of maximal activity, respectively (Fig.  2B). Thus, the rank order of transcriptional activity for these forms of ATF6 was 1-373 Ͼ Ͼ 39 -373 ϭ 1-51/94 -373 Ͼ 94 -373 ϭ 39 -51/94 -373, indicating that domains in both 1-38 and 52-93 were indispensable for full activity. In contrast to the relative activities, however, the rank order of expression and the dependence on ALLN was reversed: 39 -51/94 -373 ϭ 94 -373 Ͼ 1-51/94 -373 ϭ 39 -373 Ͼ Ͼ 1-373 (Fig. 2C, lanes 1-6), supporting the hypothesis that the forms of ATF6 with the greatest transcriptional activity are the most susceptible to degradation. Because removal of either 1-38 or 52-93 resulted in a dramatic 70-fold loss of specific transcriptional activity (Fig. 2D, compare column 1 to 2 and 4), it is apparent that domains within both of these regions of ATF6 are indispensable for maximal activity.
Given the results to this point, it appears that forms of ATF6 that are actively engaged in transcription are highly susceptible to rapid degradation. Moreover, forms of ATF6 harboring mutations that reduce transcriptional potency display coordi- B, ATF6-dependent luciferase induction. HeLa cells were co-transfected with N-terminally FLAG-tagged versions of the constructs diagrammed in A, GRP78-ERSE-luciferase, 3 and CMV-␤-gal. After 48 h in serumcontaining media, culture extracts were submitted to luciferase and ␤-galactosidase enzyme assays, as described under "Materials and Methods." Each value represents the average and S.E. of triplicate cultures. C, expression of ATF6-related proteins. HeLa cells, transfected as described in B, were extracted and analyzed for expression of FLAG-ATF6-related proteins by Western blotting using a FLAG-specific antiserum. In some cases (lanes 2, 4, 6, 8, 10, and 12), ALLN (25 g/ml) was added to the cultures ϳ14 h before extraction. The arrows designate the expected migration locations of the ATF6-related proteins based on predicted molecular mass. D, specific transcriptional activities. The relative luciferase values from B were divided by densitometrically acquired digital quantification values from the bands shown in C, lanes 1, 3, 5, 7, and 9, to determine specific transcriptional activity of each ATF6-related protein. The value for ATF6-(1-373) was set to 100%.
nately reduced susceptibility to degradation. To test this hypothesis further, we assessed the expression level of a form of ATF6 that is known to be disengaged from transcription, not through mutation in the transcriptional activation domain but through mutation in the DNA binding domain. For this purpose we used ATF6-(1-373)-dbd-M, a form of ATF6 known to be inactive as a result of mutations in the b-Zip domain (16, 18) ( Fig. 2A). ATF6-(1-373)-dbd-M displayed very low activity (Fig.  2B), as expected, and it was expressed at relatively high levels, with little dependence on ALLN (Fig. 2C, lanes 11 and 12); the specific transcriptional activity was reduced by about 200-fold, compared with 1-373 (Fig. 2D, column 6). Thus, four different constructs with mutations within the N-terminal 93 amino acids, and one with a mutation in the b-Zip domain, displayed low levels of transcriptional activation and high levels of expression that were relatively insensitive to ALLN, consistent with the hypothesis that proteasomal degradation of ATF6 is linked to ATF6-mediated transcription.
To begin to understand the features of ATF6 that engender these properties, a search was carried out to identify ATF6-like proteins for which transcription and degradation are linked. Although a number of potent eukaryotic transcription factors (e.g. Myc, ␤-catenin, and Rel) exhibit such properties (19), only VP16, the virion protein 16 transcription factor from herpes simplex virus type I, possesses significant homology to ATF6. VP16 is a potent transcriptional activator of herpes simplex immediate early genes. Our search revealed that the C-terminal TAD of VP16 bears striking homology to the N-terminal TAD of ATF6 (Fig. 3). ClustalW alignment analysis (20) showed that 35% of the residues in ATF6-(1-100) were conserved or strongly conserved with those in VP16-(356 -489). Of particular interest is a region of ATF6 between residues 61 and 68, which is 75% identical to a region of VP16 between residues 441 and 448 (Fig. 3, boxed  sequence). In VP16, these 8 amino acids are known as VN8 (21) and are required for transcriptional activation and the rapid proteasomal degradation of VP16 (22). The relationship between transcriptional potency and degradation was demonstrated by Salghetti et al. (19), who showed that as the numbers of VN8 repeats fused to the Gal4 dbd were increased, transcriptional potency increased and fusion protein expression levels decreased. Additionally, Phe-442 and Leu-444 have been shown to be critical for VN8 function, such that substitution of alanine at these positions abolishes transcriptional activation and stabilizes the protein (19,21,22). Accordingly, a construct encoding ATF6-(1-373) harboring F62A and L64A substitutions in the VN8-like region was prepared (Fig. 4A, 1-373-VN8-M). When tested in HeLa cells, the expression level of 1-373-VN8-M was ϳ7-fold higher than 1-373 (Fig 4B), supporting the hypothesis that Phe-62 and Leu-64 are important for low level expression of ATF6. In contrast to the effects on expression, compared with 1-373, the 1-373-VN8-M exhibited an approximate 5-fold reduction of transcription (Fig 4C), which translates to a 27-fold reduction of specific transcriptional activity. Thus, Phe-62 and Leu-64 are required for both efficient transcription and low level expression of ATF6.
Experiments were carried out to determine whether the low level of ATF6 expression is the result of a short half-life and, if so, whether the F62A and L64A substitutions increased ATF6 expression by increasing its half-life. Pulse-chase labeling analyses showed that ATF6-(1-373) was, indeed, relatively short lived, displaying a half-life of about 40 min (Fig. 5, 1-373). Moreover, mutating Phe-62 and Leu-64 to alanine resulted in an approximate 2-fold increase in the half-life (Fig. 5, 1-373-VN8-M). Accordingly, the increased levels of ATF6-(1-373)-VN8-M compared with 1-373 are partly the result of a decrease in the rate of degradation. 5 The finding that two amino acid substitutions in a transcription factor nearly 400 amino acids in length can make a 2-fold difference in half-life and a 5-fold reduction of transcriptional activity is, to our knowledge, unprecedented. Taken together with the results shown in Fig. 4, these findings indicate that in addition to being required for optimal transcription, Phe-62 and Leu-64 also contribute to establishing the rate at which ATF6 is degraded, constituting another demonstration that ATF6 transcriptional activity and ATF6 degradation are coupled. DISCUSSION ATF6 is not highly homologous to other potent unstable eukaryotic transcription factors. However, a region near the N terminus of ATF6 is homologous to the C-terminal TAD of the viral transcription factor, VP16 ( Fig. 3 and Fig. 6A). Of particular interest is the area of VP16 between residues 441 and 448, the VN8 region. ATF6 is the only eukaryotic protein that we were able to find with a sequence that would be predicted to behave like VN8 does in VP16. The N terminus of G13, the closest known relative of ATF6 (23,24), exhibits a great deal of homology to ATF6; however, it does not appear to possess an active VN8-like sequence (Fig. 6B). This is because there is an absolute requirement for phenylalanine at position 2 of the VP16 VN8, such that substitution of valine at this position, as naturally occurs in G13, results in a 90% loss of VP16 transcriptional activity (25). Accordingly, because G13 apparently lacks a functional VN8 sequence, compared with ATF6, it would be predicted to possess reduced transcriptional potency. Consistent with this view is a recent study showing that G13 displays reduced transcriptional activity compared with ATF6 and that it is expressed at higher levels (24). Accordingly, although G13 and ATF6 display many of the same structural and functional features, they most likely employ very different transcriptional activation mechanisms. Future studies of these mechanistic differences and how they contribute to the roles of ATF6 and G13 during the stress response will be of interest.
VP16 possesses several acid-rich regions between residues 380 and 425 (Fig. 3), which play important roles in transcriptional activation. ATF6 also possesses acid-rich domains in this region of the TAD, the most notable of which is between residues 25 and 35 (Fig. 6B). Evidence from the present study 5 In preliminary experiments we have found that transcriptionally active forms of ATF6 reduce the expression rates of many proteins, including those expressed from transgenes, mostly likely through a global translational inhibition, which is a hallmark of the ER stress response. Accordingly, it is most likely that ATF6-(1-373)-VN8-M is translated at a higher rate than 1-373 and that this, coupled with a 2-fold reduction of half-life, accounts for the 5-fold increase in the expression level of ATF6-(1-373)-VN8-M compared with 1-373.  6. Alignment of VP16, ATF6,  and G13. A, diagram of transcriptional activation domains. Shown are diagrams of the transcriptional activation domains of VP16, ATF6, and G13. Also shown are the locations of the VN8 or VN8-like regions in each. The degrees of identity for the VN8-like regions in ATF6 and G13, compared with VP16, are 75 and 25%, respectively. B, alignment of the N termini of ATF6 and G13. The N-terminal regions of human ATF6, GenBank TM accession number P18850, was aligned with the same region of human G13, GenBank TM accession number Q99941, using ClustalW (20) (npsapbil.ibcp.fr/cgi-bin/npsa_automat.pl? page ϭ npsa_clustalw.html). The identical strongly conserved and weakly conserved residues are indicated in red, green, and blue, respectively. Also shown is the hypothetical functional domain between residues 25-35 of ATF6 and 25-38 of G13 (see box), and the VN8-like region of ATF6-(61-68) (see box). The VN8-like region of G13 is not included in the box, because it is predicted to be inactive (see text). supports a role for this region in ATF6-mediated transcription (see Fig. 1 and Fig. 2), and disruption of this 11 amino acid stretch reduces transcription. 6 A similar acid-rich region can be found in G13 (Fig. 6B); however, it is naturally disrupted by a 3 amino acid insert, which is also consistent with the reduced transcriptional activity of G13, compared with ATF6 (24). Thus, based on sequence homology and on the results of the present study, this N-terminal acid-rich region is predicted to serve an important role in ATF6-mediated transcription, although it may be a minor role compared with the VN8-like sequence within ATF6- (61-68).
The processes governing the coordination of ATF6-mediated transcription and ATF6 degradation are not yet resolved. Several other potent transcription factors that are very unstable display similar properties, and recent studies have shown that in many of those cases the sequences that signal their ubiquitination, the degron, overlaps closely with the TAD (19,22,26). Additionally, recruitment of the unstable transcriptional activators to DNA is necessary for proficient degradation (22), and the 19 S proteasome subunit has been shown recently to play a critical role in transcriptional elongation (27). RNA polymerase II co-immunoprecipitates with ubiquitin-protein ligases, ubiquitin-hydrolases, and other proteasome proteins in the nucleus (19,22,28), supporting the hypothesis that proteasomal machinery actually composes a portion of the transcription complex (29). In fact, it is possible that ubiquitination itself, in addition to serving as a molecular tag for protein degradation, can also serve as a post-translational modification required for optimal activity of some proteins in the transcription complex. The demonstration that ubiquitination of RNA polymerase II is required for optimal activity supports this view (30). Also, it has been postulated that ubiquitination of transcription factors serves to recruit the 19 S proteasomal subunit to promoters where it serves a chaperone role to promote transcription (31).
In summary, ATF6 is a potent, unstable transcription factor that displays the property of coordinate transcriptional activation and degradation. Limiting the activities, locations, and quantities of strong, transiently active transcription factors is likely to be critical for proper cellular regulation. Such a mechanism might be required to enable the fine regulation of the activities of transcription factors that could, if left unchecked, have deleterious effects on long term cell survival. Certain genes induced by long term activation of ATF6, such as CHOP/ GADD-153, could potentially lead to cell death, indicating a need for finely tuned transient induction of ATF6. Future studies oriented toward understanding more about the domains of ATF6 that are responsible for linking transcription and degradation will reveal more about the role of ATF6 in the ER stress response. It will also be interesting to determine whether G13, a close relative of ATF6, also displays coordinate transcription and degradation; if it does, then deciphering the mechanism should provide clues as to how and why two such similar transcription factors might have evolved so they target enhanced expression of the same ER-stress response genes but through different transcriptional/degradation regulatory mechanisms. Finally, the high degree of homology between the ATF6 and VP16 TADs implies some overlap in transcriptional mechanisms, suggesting that during viral infection, VP16 might in some ways mimic some of the actions of ATF6 during the ER stress response.