Conservation and Divergence of the Yeast and Mammalian Unfolded Protein Response

Yeast Hac1 (yHac1), the transcription factor that binds and activates the unfolded protein response element of endoplasmic reticulum (ER)-chaperone gene promoters, only accumulates in stressed cells after an unconventional splicesosome-free mRNA processing step and escape from translation block. In determining whether the novel regulatory mechanisms for yHac1 are conserved in mammalian cells, we discovered a unique unfolded protein response element-like sequence within the endoplasmic reticulum stress element 163, one of the three ER stress elements recently identified in the ratgrp78 promoter. The unspliced form of yHac1 is stably expressed in nonstressed mammalian cells and is as active as the spliced form in stimulating the promoter activities of grpgenes. Further, the yHac1 mRNA is not processed in response to ER stress in mammalian cells. We identified a CCAGC motif as the yHac1 binding site, which is contained within a YY1 binding site previously shown to be important for mammalian UPR. Dissection of the yHac1 and the YY1 binding sites uncovered specific contact points for an activator protein predicted to be the mammalian homolog of yHac1, the activity of which can be stimulated by YY1. A model of the conserved and unique features of the yeast and mammalian unfolded protein response transcription machinery is proposed.

The endoplasmic reticulum (ER) 1 is an essential organelle for cell viability. In addition to serving as a major intracellular store for calcium and the production site for lipids and sterols, ER is the cellular organelle for the synthesis, assembly, and glycosylation of proteins that are destined for secretion or transport to the cell surface. Perturbations of ER result in stress signaling from the ER to the nucleus, leading to the activation of specific sets of genes (1,2). Conserved from yeast to human is a mechanism referred to as the unfolded protein response (UPR) (3,4). In Saccharomyces cerevisiae, several key components of the UPR have been identified through genetic analysis and biochemical characterizations. Thus, the yeast UPR is mediated by a novel signaling pathway that requires ER stress-inducible splicing of the primary transcript of Hac1, also referred to as Ern4p (5,6). The open reading frame of the Hac1 gene encodes a protein of 230 amino acids (aa), whereas the active transcription factor is a 238-aa protein. In nonstressed yeast cells, the unspliced Hac1 mRNA is prevented from translation into the 230-aa protein by the presence of a translation attenuator in a 252-bp intron (6,7). Upon induction of the UPR, unconventional splicing mediated by Ire1p results in the elimination of the intronic sequence, which contains the codons encoding the last 10 aa of the 230-aa Hac1 and part of the 3Ј-untranslated region. Although the last 10 aa of 230-aa Hac1 is deleted through this process, the newly spliced Hac1 mRNA contains an additional exon of 18 aa, resulting in the translation of the 238-aa Hac1. Finally, in yeast cells, 230-aa Hac1 and 238-aa Hac1 produced from exogenous vectors are stable and can bind UPRE; however, the transactivating activity of the 230-aa Hac1 is lower than that of the 238-aa Hac1 (6).
The recognition sequence for yHac1 is a 22-bp sequence referred to as the unfolded protein response element (UPRE), which is found in the ER stress-inducible promoter of yeast ER resident protein genes such as KAR2 and PDI1. The yeast UPRE contains an imperfect palindrome with a spacer of one nucleotide (CAGcGTG). The steric configuration of the UPRE appears to be critical because insertion or deletion of a single nucleotide within the spacer region results in the complete loss of its ability to mediate the UPR (8). One of the best characterized ER stress-inducible genes is grp78, the mammalian homolog of KAR2 (1). Contrary to yeast, the rat grp78 promoter is functionally redundant and contains three ER stress elements termed ERSE (9,10). The consensus ERSE derived from comparison of a variety of ER stress-inducible mammalian promoters consists of a tripartite structure CCAAT(N 9 )CCACG, with N 9 being a strikingly GC-rich region of 9-bp. The ERSE physically and functionally interacts with multiple mammalian transcription factors including that of YY1, Y-box proteins, YB-1, dbpA, NF-Y/CBF, and a newly discovered ER stress-inducible complex termed ERSF (10)(11)(12)(13). Although the various ERSEs share similar sequence motifs and can independently confer the mammalian ER stress response to heterologous promoters (10,14,15), there are subtle differences that are evolutionarily conserved and could have functional implications for their regulation. For example, ERSE-163 contained within the grp78 core (16) bears the sequence CCAGC instead of the consensus CCACG found in ERSE-98 contained within the C1 element, which is most proximal to the TATA element (10). The CCAGC sequence within ERSE-163 of the grp78 core is of particular significance because in vivo genomic footprinting revealed that it is the major region on the grp78 promoter where ER stress-induced factor occupancy changes are detected (16), correlating with its functional importance based on 5Ј deletion analysis (17).
We have previously identified the multifunctional transcription factor YY1 as a major DNA-binding protein to ERSE-163 (12). The unconventional YY1 binding site encompasses the CCAGC motif and its flanking sequence. In co-transfection studies using NIH3T3 cells, YY1 had little effect on the basal expression of the grp78 promoter but specifically enhanced induction of the grp78 promoter in a DNA binding-dependent manner by a variety of ER stress conditions. Consistent with divergent transcriptional regulatory mechanisms for the different ERSEs, YY1 stimulates the grp78 core but the same effect was not observed for the C1 element containing ERSE-98 (10). Interestingly, YY1 has not been found in yeast, and the mammalian homolog of Hac1 has not yet been identified.
Based on the similarity between the mammalian and yeast UPR, we seek to determine whether the unique regulatory mechanisms for Hac1 discovered in S. cerevisiae are conserved in mammalian cells. Our studies revealed several novel observations suggesting both conservation and divergence of the mammalian and yeast UPR. We discovered that among the three ERSEs of the rat grp78 promoter, only ERSE-163 contained within the grp78 core shares sequence identity with the partial palindromic sequence of the yeast UPRE. This is the only ERSE that can bind and can be transactivated by yHac1. We further showed that yHac1 mRNA is not processed in response to ER stress in mammalian cells. Both the unspliced and spliced form of yHac1 can be stably expressed in nonstressed mammalian cells and exhibit similar stimulatory activity toward the grp78 promoter. The grp94 promoter, which shares sequence identity with the 78core (17), is also strongly activated by yHac1. The binding site of yHac1 is contained within the YY1 binding site, and binding of yHac1 to ERSE-163 is strictly dependent on the sequence integrity of the CCAGC motif. Through base mutations that selectively destroy either the yHac1 or YY1 binding site, we defined critical contact points for an activator protein predicted to be the mammalian homolog of yHac1, the activity of which can be stimulated by YY1. Based on these observations, we propose a model of the conserved and unique features of the yeast and mammalian UPR transcription machinery.

MATERIALS AND METHODS
Plasmid Constructions-To construct the mammalian expression vector for 230-aa Hac1, Hac1 was released from the glutathione Stransferase-Hac1 construct by BamHI and EcoRI digestion, and the 700-bp fragment was subcloned into the same sites of the mammalian expression vector pMCX-PL1 (gift of R. Buettner, University of Regensburg, Germany) to generate pMCX-Hac1. The HA epitope-tagged, spliced version of Hac1 (HAC1 i ) was constructed by PCR amplification using pJC835 (5) as template DNA and primers N2 and C2. The PCR product was digested with XbaI and subcloned into the XbaI site of pED (18). Overlap extension PCR was performed to construct HA epitopetagged, unspliced HAC1 (HAC1 u ). This was amplified as two fragments with primers N2, C3 and N3, C2 using pJS835 (5), and yeast genomic DNA as templates, respectively. The sequences for the new primers are: N2, acatctagaaaccgccaccatggaaatgact; N3, aaatctagaagagtcgacgacgctacctg; C2, aaaaaatctagataccctcttgcgattgtcttc; and C3, caaagggtagactgtttcccgc. The PCR products were mixed and reamplified with primers N1 and C1, digested with XbaI, and subcloned into the XbaI site of pED.
The construction of the 78core(M5)CAT is identical to the construction of 78core/CAT [(formerly referred to as (Ϫ169/Ϫ135)MCAT), which has been described (14), with the exception of a single base mutation from G to T at Ϫ146 as shown in Fig. 7. For the construction of Ϫ154(M6)CAT, PCR-directed site mutagenesis was used to create the base mutations at Ϫ140, Ϫ143, and Ϫ144 as shown in Fig. 8. All mutations were verified by DNA sequencing. Expression vectors CMV-YY1 and the DNA-binding deletion mutant CMV-YY1⌬ were gifts from Y. Shi (Harvard Medical School), and their construction has been described (19).
Transfection Conditions-NIH3T3 cells, a murine embryonic fibroblast cell line, was maintained in high glucose Dulbecco's modified Eagle's medium containing 10% fetal bovine serum at 35°C. The cells were transfected using either the calcium phosphate method as described previously (12) or with SuperFect reagent (Qiagen). The grp78 promoter/CAT fusion genes Ϫ154CAT, 78coreCAT (formerly referred to as (Ϫ169/Ϫ135)MCAT), and 78C1CAT (formerly referred to as (Ϫ109/ Ϫ74)MCAT) have been described (14). The grp94 promoter/CAT fusion gene Ϫ357CAT has been described (20). The construction of the 78core(M5)CAT and the Ϫ154(M6)CAT was described above. For the calcium phosphate transfections, NIH3T3 cells, seeded in duplicate sets in 10-cm-diameter dishes, were co-transfected with 5 g of the CAT reporter gene, 5 g of the expression vector for Hac1, or the vector alone. 5 g of the pSV40 ␤-galactosidase plasmid were also co-transfected to monitor efficiency of transfection. Each transfection experiment was repeated at least three times. For transfections using Super-Fect, NIH3T3 cells were seeded in duplicate in 10-cm-diameter plates, treated for 2 h with a mixture of DNA/SuperFect 1:5 containing 5 g of CAT reporter gene, 2 g of Hac1 and/or YY1 expression plasmid, 1 g of pSV40 ␤-galactosidase, and pMCX-L1 in the amount requested to keep the total DNA concentration to 10 g/plate. For stress induction, 24 h after transfection the cells were treated with 1.5 g/ml of tunicamycin or 300 nM thapsigargin (Tg) for 16 h prior to harvesting. Preparation of the cell lysates for CAT assays and the quantitation of the CAT assays have been described (12). Each transfection was repeated three to five times. COS-1 cells were transiently transfected with pED, pED-HA-Hac i , pED-HA-HAC u , pED-hIRE1 (21), and pED-hIRE1 K599A (21) using a calcium phosphate procedure (22). At 48 h post-transfection, cells were treated with tunicamycin (10 g/ml) for 6 h and harvested in Nonidet P-40 lysis buffer. Cell extracts were analyzed by SDS-polyacrylamide gel electrophoresis and Western blotting.
In Vitro Transcription and Translation-In vitro translated Hac1 and YY1 were obtained with the TNT coupled reticulocyte system (Promega, Madison, WI) according to the manufacturer's instructions. For this purpose, pCMX-Hac1 was linearized with XbaI, whereas pCMV-YY1 (gift of Y. Shi, Harvard Medical School) was used undigested.
Gel Mobility Shift Assays-For EMSAs with in vitro translated proteins, 2 l of the TNT in vitro coupled transcription/translation reaction mixture were used in the binding reactions. In some reactions, 5 l of cell lysate from transfected cells were used. Hac1 binding conditions were accomplished in 20 l of 20 mM Hepes, 50 mM KCl, 2.5 mM EDTA, 0.5 mM dithiothreitol, and 1 g of poly(dI⅐dC). YY1 binding conditions have been previously described (12). Samples were preincubated for 5-10 min at room temperature in binding buffer, to which competitor oligos were added whenever specified. After addition of 1-5 ng of probe, the reactions were further incubated at room temperature for 15 min and loaded onto a nondenaturing 4% polyacrylamide gel containing 2.5% glycerol.
Western Blotting-Western blotting was performed by standard procedures (23). The blot was sequentially probed with anti-influenza HA epitope (Roche Molecular Biochemicals Corp.) and anti-Hac1 (raised against the last 10 aa of processed yHac1 and kindly provided by P. Walter (5)) primary antibodies and horseradish peroxidase-conjugated goat anti-mouse secondary antibodies. Bands were visualized using the ECL kit (Amersham Pharmacia Biotech).

RESULTS
Unique Occurrence of a UPRE-like Sequence within the Conserved grp78 Core-The rat grp78 promoter contains three ERSEs (ERSE-98, ERSE-131, and ERSE-163) upstream of the TATA element (Fig. 1A). In slight variation to the consensus ERSE sequence of CCAAT(N 9 )CCACG, ERSE-163 contains a CGAAT motif separated by 9-bp from a CCAGC motif. Further, in contrast to other ERSEs, ERSE-163 contained within the grp78 core region spanning Ϫ170 to Ϫ135 exhibits two distinct features, suggesting its functional importance in the in vivo regulation of the endogenous grp78 promoter. First, in vivo genomic footprinting revealed that stress-inducible footprint changes occurred primarily within the GGCCAGCTTG se-quence of ERSE-163 (Fig. 1A). Second, with the exception of 2-bp, this sequence is identical to the core of the yeast UPRE sequence GGACAGCGTG, which is the DNA-binding site for yHac1. The binding of Hac1 to the UPRE is strictly dependent on the sequence integrity of the UPRE, because insertion of only 1 bp within the UPRE as in the case of UPRE m (Fig. 1A) completely eliminates Hac1 binding. Scanning of the rat 78 promoter sequence reveals no other substantial match with the yeast UPRE, and no Hac1 binding was detectable with the other ERSEs (data not shown).
Functional Conservation of yHac1 as a Transcription Regulator of the Mammalian grp78 Promoter-With the identification of a UPRE-like sequence within the grp78 core, we next tested whether yHac1 can be stably expressed in mammalian cells and what effect, if any, it has on the grp78 promoter activity under stressed and nonstressed conditions. First, 230-aa Hac1 was subcloned into a mammalian expression vector ( Fig. 1B) and transiently transfected into NIH3T3 cells. To test for the ability of the mammalian cells to stably express yHac1, cell lysates were prepared from cells either transfected with the empty vector or the Hac1 expression vector. The cell extracts were used in EMSAs with the UPRE as probe (Fig.  1C). In the control cells transfected with empty vector, no UPRE binding activity was detected. In cells transfected with the Hac1 expression vector, a strong UPRE binding complex was readily detected. The specificity of UPRE binding was confirmed by efficient competition with the wild type but not the mutated UPRE.
To test the effect of overexpression of 230-aa Hac1 on grp78 promoter activity, the expression vector for Hac1 was co-transfected into NIH3T3 cells with Ϫ154CAT, which retains all the ER stress-inducible properties of the rat grp78 promoter (14). As expected, in cells transfected with the empty vector, an 8-fold induction of the promoter activity was detected with the ER stress inducer Tg, which depletes the ER calcium store and tunicamycin, which blocks protein glycosylation (Fig. 2). In nonstressed cells, co-transfection of 230-aa Hac1 resulted in a 7-fold induction of the basal grp78 promoter activity. Upon treatment of the cells with Tg or tunicamycin, the overall promoter activity further increased to about 25-fold (Fig. 2). Similarly, the grp94 promoter, which shares sequence identity with the 78core (17), was also strongly induced by yHac1. These results indicate that yeast 230-aa Hac1, when overexpressed in mammalian cells, is highly functional as a transcription factor and acts as a potent stimulator of the grp promoter activities in both nonstressed and stressed cells.
Stable Expression of Unspliced yHac1 in Mammalian Cells-Because yHac1 can exist as a 230-aa protein if the translation is terminated at the first termination codon or as a 238-aa protein if the primary transcript has undergone splicing, we compared the potency of the various forms of yHac1 to enhance grp78 promoter activity. Prior to these experiments, it is of interest to determine whether transfection of the intact yHac1 gene into nonstressed mammalian cells would result in translation block as observed in S. cerevisiae or whether translation would proceed resulting in a 230-aa protein that could be stably expressed. To facilitate detection of the yHac1, the HA epitope was inserted into the amino end of the Hac1 coding sequence, which was subcloned into the mammalian expression vector, pED (Fig. 3A). Similarly, an HA-tagged Hac1 gene encoding the induced form of 238-aa Hac1 was also constructed. Upon tran-FIG. 1. Identification of a UPRE-like site within the grp78 core sequence. A, localization of the ERSEs in the rat grp78 promoter and comparison of the grp78 core sequence, the yeast UPRE, as well as the one nucleotide insertion mutant UPRE m , that were used in EMSAs to investigate binding ability and specificity of yHac1. Boxed nucleotides in the 78core sequence represent the two conserved motifs separated by 9-bp contained within the consensus ERSE (10). Open stars and arrows denote inducible changes detected by in vivo genomic footprinting (16). In the UPRE and UPRE m sequences, the yHac1 binding site is boxed, with the single base addition within UPRE m indicated by an italicized lowercase letter. Other lowercase letters represent linker sequences. sient transfection into COS cells, yHac1 protein expression level was determined by Western blotting with antibodies against the HA-epitope (Fig. 3B). The unusual electrophoretic mobility of the 230-aa Hac1 and 238-aa Hac1 was in agreement with that previously reported for Hac1 (6). As expected, 238-aa Hac1 was produced from the vector expressing the induced form of Hac1. However, with the expression vector containing the unspliced form of the Hac1 gene, predominantly 230-aa Hac1 was produced. These results revealed that in contrast to S. cerevisiae, the unspliced form of yHac1 can be translated and stably expressed in nonstressed mammalian cells.
To investigate further whether yHac1 mRNA is processed to produce the 238-aa Hac1 in response to ER stress in mammalian cells, COS-1 cells transiently transfected with the expression vectors for HAC1 I and HAC1 u were treated with tunicamycin and cell lysates prepared for analysis by Western blotting. In contrast to the anti-HA antibody, which detected the 238-and 230-aa forms in the HAC1 I and HAC1 u transfected cells, respectively, probing the same blot with anti-Hac1 antibody that specifically recognizes 238-aa yHac1 indicated that the 238-aa form was present only in cells transfected with HAC1 i but not with HAC1 u (Fig. 3C, lanes 2 and 3). Both tunicamycin treatment and co-transfection with the human Ire1 (21) failed to produce the 238-aa form in the HAC1 u transfectants (Fig. 3C, lanes 4 and 5). The inclusion of the K599A mutant form of Ire1 (21) also showed no effect (Fig. 3C, lane 6). Further, reverse transcription-PCR failed to detect the predicted spliced form of Hac1 mRNA (data not shown). These combined results show that yHac1 mRNA is not processed in COS-1 cells.
Having established that yHac1 encoded by various DNA templates can be stably expressed in mammalian cells, their effect on the Ϫ154CAT reporter gene in both control and Tgtreated NIH3T3 cells were determined in co-transfection experiments. Our results showed that all three forms of yHac1 (230-aa Hac1, HA-tagged 230-aa Hac1 from unspliced template, and HA-tagged 238-aa Hac1 from spliced template) were able to stimulate the basal grp78 promoter activity by about 7-fold, after Tg induction, up to an overall promoter activity of 20 -30-fold, as compared with the 8-fold induction observed in Tg-treated transfection with the empty vector (Fig. 4). Collectively, these experiments suggest that a Hac1-like activating activity is rate-limiting in mammalian cells toward transcription induction of the grp78 promoter, and this activity can be mimicked by overexpression of the yHac1, either in the form of a 230-or 238-aa protein.
Binding of yHac1 to ERSE-163 within the grp78 Core Sequence-With the identification of a UPRE-like site within the rat grp78 core sequence, we first determined whether yHac1 can bind to the grp78 promoter sequence. For this purpose, 230-aa Hac1 was translated in vitro and used in EMSAs with either the grp78 core sequence or the yeast UPRE as radiolabeled probes (Fig. 5). Yeast Hac1 was able to bind to the grp78 core sequence, and its binding to the UPRE was stronger. The binding activity to the 78core sequence is attributed to yHac1 because it could be specifically competed by molar excess of the wild type but not the mutant UPRE. Further, as reported for yHac1 (8), this binding activity was enhanced by the addition of 10 mM magnesium in the binding reaction (data not shown).
To map more precisely the critical DNA contact points for yHac1 within the 78core, the ability of a set of mutants of the 78core (M1 through M4) with clustered 4-bp mutations around the UPRE-like site to bind in vitro translated yHac1 was examined (Fig. 6A). The sequence of the mutated bases is shown in Fig. 6B. Each of the core sequences was individually labeled and used as probes in EMSAs. The wild-type core sequence and the UPRE were included as positive controls. As expected, the strongest binding of yHac1 was observed with the UPRE probe. Among the 78core sequence, yHac1 bound with similar affinity to M3 and M4 as with the wild-type sequence. In contrast, binding of Hac1 to M2 was severely diminished. In M2, the CAGC motif of ERSE-163, which shares complete sequence identity with the UPRE was mutated to ACTA. The result obtained for M3, although unexpected, suggested that the exact TTGG sequence motif at 3Ј half of the UPRE-like sequence may not be critical for yHac1 binding. For M4, which lies outside the boundary of the UPRE-like sequence, this mutation was without effect on yHac1 binding. Interestingly, for M1, yHac1 binding was slightly stronger. This could be due to the fact that the last mutated base in M1 changed a C residue to A, resulting in more similarity with the UPRE core sequence ACAGCTTG. As summarized in Fig. 6B, the most critical yHac1 binding domain to the CAGC motif spans Ϫ148 to Ϫ145, which occurs within the larger YY1 binding site previously mapped by this same set of mutations to Ϫ152 to Ϫ141 (16). For YY1 binding to the 78core, M1 and M3 mutation severely diminished YY1 binding, whereas M2 retained weak binding affinity for YY1 (16). Lastly, in vitro translated YY1 while exhibiting strong binding to the 78core does not bind the UPRE (Fig. 6B).
Hac1 Stimulation of Specific Mammalian ERSE-If yHac1 stimulation is dependent on its binding to the grp78 promoter, the prediction is that yHac1 should be able to stimulate the promoter activity mediated by ERSE-163 contained within the grp78 core sequence but not the C1 element, which does not contain a UPRE-like sequence and does not bind yHac1 (Fig.  7A). Further, mutation of the UPRE-like sequence to eliminate specifically yHac1 binding within the 78core should diminish Hac1-mediated stimulation. Because the M2 mutation resulted in the loss of Hac1 binding and reduction in YY1 binding activity (Fig. 6), we created a new 78core mutant M5, which contains a single base mutation that changes the critical CAGC sequence to CATC. This mutation resulted in even stronger binding for YY1 (Fig. 7A), because the CATC motif corresponds to the consensus YY1 binding site (24). However, the ability to bind yHac1 was severely diminished for M5 as compared with the wild-type core (Fig. 7A). For 78C1, although it does not bind yHac1, YY1 binding can be detected and is mediated by the CCACG motif (11).
To test the promoter sequence requirements for yHac1 stimulation in vivo, CAT reporter genes 78coreCAT, 78core(M5)CAT, and 78C1CAT containing tandem copies of the synthetic oligomers shown in Fig. 7A linked to the minimal promoter of mouse mammary tumor virus were co-transfected with the Hac1 expression vector or empty vector into NIH3T3 cells. In nonstressed cells, overexpression of yHac1 was able to stimulate the 78core promoter activity by about 4-fold (Fig. 7B). For M5, there was minimal increase of 1.2-fold, and for 78C1, no increase was observed. In cells treated with Tg, yHac1 further stimulated the overall 78core promoter activity about 6-fold. For M5, Tg stress induction was lost, and in the presence of yHac1, only 1.5-fold increase was detected. In the case of 78C1, its promoter activity increased to about 5-fold in Tg-treated cells and was not affected by overexpression of yHac1. Thus, the stimulation of the grp78 promoter activity by yHac1 is promoter sequence-specific and strictly dependent on the sequence integrity of the CAGC sequence motif. Further, results from the 78core(M5) mutation shows that YY1 binding is not sufficient to confer stress inducibility, suggesting the involvement of an as yet unidentified mammalian factor with CAGC as its critical binding site.
Optimal Hac1 Stimulation of the grp78 Promoter Requires an Intact YY1 Binding Site Adjacent to the UPRE-like Sequence-To test whether yHac1 binding alone is sufficient for its stimulative activity, mutation M6 was created within the context of the native grp78 promoter. In M6, the sequence from Ϫ144 to Ϫ140 containing the 3Ј half of the UPRE-like sequence within ERSE-163 was mutated from TTGGT to ACGGA, whereas the critical CAGC sequence motif was preserved (Fig.  8A). Consistent with the boundaries previously defined for YY1 binding site (16), YY1 was not able to bind M6 in EMSA experiments, whereas yHac1 binding was fully retained (Fig.  8B). Thus, through M5 and M6 mutations of ERSE-163, we discovered that yHac1 binding to the UPRE-like sequence of the grp78 promoter is primarily mediated through the CAGC sequence, and YY1 binding requires the sequence immediately 3Ј to the CAGC motif. Despite the ability of yHac1 to bind to the promoter sequence bearing the M6 mutation, when the construct Ϫ154(M6)CAT was tested in co-transfection experiments, yHac1-mediated stimulation of the grp78 promoter was reduced to half of that observed with the wild-type promoter under both nonstressed and Tg stressed conditions (Fig. 8C). These results showed that the YY1 binding site is required for the optimal stimulation of the rat grp78 promoter by yHac1. Further, it is noted that in the absence of exogenously expressed yHac1, the stress-inducibility of Ϫ154(M6)CAT was similar to Ϫ154CAT. This is consistent with the functional redundancy of the grp78 promoter because of multiple ERSEs (25).
Enhancement of yHac1 Stimulation of the grp78 Core Sequence by YY1-The requirement of the YY1 binding site for optimal Hac1 stimulating activity on ERSE-163 suggests that YY1 may be a co-activator for yHac1. To test this, expression vectors for YY1 and yHac1 were co-transfected into NIH3T3 cells using either 78coreCAT or 78core(M5)CAT as the reporter genes. Our results confirmed that yHac1 stimulates both basal and Tg-induced promoter activity of the 78core, whereas YY1 only exerts a stimulative effect under Tg stressed conditions (Fig. 9A). When both proteins were overexpressed, the 78core promoter activity was enhanced over the level achieved by yHac1 alone. The stimulation of yHac1 activity by YY1, although only in the range of 1.5-2-fold, is strictly dependent on the sequence integrity of the yHac1 binding domain CAGC because no stimulation was observed with the M5 mutant that lost yHac1 binding (Fig. 9B). Further, whereas yHac1 stimulation of the 78core increased with increasing amount of yHac1, optimal stimulation by YY1 was observed at low dosage, and increasing the dosage of YY1 beyond the optimal amount resulted in mild suppression instead (Fig. 9C and data not  shown). The suppressive effect observed with high level expression of YY1 could be associated with squelching of the transcription machinery (26). Collectively, these results suggest that yHac1 activity is rate-limiting, but there may already be a saturating amount of endogenous YY1 dampening the effect of exogenous YY1. In agreement with the requirement of the YY1 binding site for optimal yHac1 stimulation of the rat grp78 promoter (Fig. 8), deletion of the DNA-binding domain of YY1 resulted in minimal stimulation of the yHac1 activity on the 78core activity (Fig. 9C). DISCUSSION The recent discovery of the tripartite structure of the mammalian ER stress response element ERSE provides important information on the regulatory components mediating the mammalian UPR (10). Thus, within the rat grp78 promoter are three ERSEs, named ERSE-98, ERSE-131, and ERSE-163, with highly similar but distinct sequences that are evolutionarily conserved among promoter elements of ER chaperone genes. The fact that each ERSE is able to confer ER stress inducibility to a heterologous promoter (14) suggests that the grp78 promoter contains functionally redundant ERSEs. Our studies reveal several new findings leading to the proposal of a more complete model for mammalian stress induction mediated by the ERSE-163 (Fig. 10). Importantly, our model suggests that despite major differences, some key features of the mammalian UPR are conserved with the yeast UPR.
The yeast promoter contains two simple regulatory sequences, UPRE and heat shock element, that allow the yeast KAR2 gene to respond to the accumulation of malfolded proteins in the ER and in the cytoplasm. In analyzing transcription factors that interact with various ERSEs, we discovered that ERSE-163 is the only ERSE of the grp78 promoter that shares near sequence identity with the yeast UPRE. We provide here evidence that yHac1 selectively binds the ERSE-163, and the binding occurs within the sequence motif GGC-CAGCTG, which distinctively undergoes stress-inducible changes (16). ERSE-163 is also strongly trans-activated by overexpression of yHac1. Because other ERSEs such as ERSE-98, although unable to bind Hac1 are nonetheless strong mediators of the UPR (10,14,15), one explanation is that the organization of the mammalian grp78 promoter has diverged from the yeast gene to ensure the induction of mammalian UPR by multiple pathways. Alternatively, the mammalian homolog of yHac1 may differ from yHac1 such that it is able to interact with all the ERSEs, which might be regulated through one common mechanism. The resolution of this awaits the isolation of mammalian Hac1.
Unexpectedly, mutational analysis revealed that the core binding site required for yHac1 binding on the mammalian grp78 promoter is considerably less stringent than the yeast UPRE. In S. cerevisiae, Hac1 is the only trans-acting factor demonstrated to bind the UPRE so far, whereas the more complex mammalian ERSEs have the ability to bind multiple transcription factors, as the CCAAT binding factor NF-Y, also referred to as CBF (11,15,27,28), YY1 (12,29), and the recently identified novel transcription complex termed ERSF (Ref. 10 and data not shown). Another transcription factor, ATF-6, is a potent activator of the grp promoters, but its direct binding to ERSE has not been demonstrated (9). In yeast, Hac1 mRNA is the substrate of the serine-threonine kinase Ire1p with a dual endoribonuclease function, and Hac1 expression is uniquely controlled at the level of translation. Here we showed that mammalian cells are not able to process yHac1 mRNA to produce the 238-aa form of Hac1 in response to ER stress. Two mammalian homologs hIre1p(␣) and hIre1p(␤) of the yeast IRE1p kinase have been recently identified (21,30). The human hIre1p(␣) is able to cleave the yHac1 mRNA substrate at the identical 5Ј splice site as yeast Ire1p but not at the 3Ј splice site. So far, ribonuclease activity has not been demonstrated for hIre1p(␤). Thus, the mammalian homolog of yHac1, if one exists, may have a divergent 3Ј splice site, or another yet unidentified homolog of hIre1p with the necessary splicing capability may be involved. Alternatively, the putative mammalian Hac1-like activity upon stress induction may be regulated entirely differently from yHac1. Another divergent feature of Hac1 regulation in yeast and mammalian cells is that the unspliced form of yHac1 can be translated and stably expressed in nonstressed mammalian cells. Thus, the unprecedented translation attenuation mechanism to suppress expres-sion of Hac1 in nonstressed yeast cells is apparently not conserved in mammalian cells.
It has been demonstrated that yHac1 acts in concert with co-factors to activate transcription. In S. cerevisiae, the transcriptional co-activator complex having histone acetylase activity composed by Ada2, Ada3, Ada5, and Gcn5 is required for maximal transcriptional induction of the yeast grp78 promoter (31). Interaction between Gcn5 and Ire1p is also demonstrated. The direct association of these co-activator complexes with Ire1p, which is localized to subcompartments within the ER, with particular concentration around the nuclear envelope, led to the suggestion that the nucleoplasmic domain of Ire1p may serve as a nucleation site for assembly of a multisubunit transcriptional complex (21). Our studies here demonstrated that one such co-activator of the putative mammalian Hac1 activity as part of the multi-protein complex in mammalian cells may be YY1. When co-transfected, YY1 enhances the stimulating activity of yHac1, and this requires the DNA-binding domain of YY1 for full activity. In support for the requirement of YY1 binding for maximal Hac1 activity, a selective mutation of the YY1 site adjacent to the UPRE-like sequence within ERSE-163, which eliminates YY1 binding while retaining yHac1 binding, results in the inability of Hac1 to transactivate the promoter at the same magnitude as in the presence of a functional YY1 binding site. YY1 is able to interact with members of the ATF/CREB family proteins, which like yHac1 are bZIP proteins (32). Detailed structure-function analyses in vitro and in vivo revealed that the interaction is mediated through the C-terminal zinc finger domain of YY1 and the basic-leucine zipper region of ATFa2. This study suggests that the C-terminal domain of YY1 is able to interact with yHac1 through its bZIP domain and that YY1 devoid of its C-terminal domain would not be able to stimulate Hac1 activity. In support of this hypothesis, in vitro glutathione S-transferase pull-down assays confirmed that YY1 is able to interact with Hac1, and laser confocal microscopy revealed that transfected yHac1 and YY1 co-localized in the nuclei (data not shown). Further, enhancement of yHac1 activity by YY1 requires its C-terminal domain, which confirms our previous observation that stimulation of grp78 promoter by YY1 in stressed mammalian cells requires its C-terminal domain (12). The lack of stimulation by the truncated YY1 could be attributed to loss of DNA binding and/or protein-protein interaction. This issue is now resolved in part by mutating the core YY1 site from CAG to CAT within the ERSE-163. This mutation results in even stronger binding of YY1 to the site because CAT is the core of the YY1 consensus binding site (24). Strikingly, this mutation resulted in the loss of YY1 stimulation of the ERSE-163-mediated promoter activity under ER stress induction. This result showed that YY1 binding alone is not sufficient for its stimulation effect and further suggests that this mutation could have disrupted another regulatory factor required for YY1 activity. Indeed, this one base mutation severely diminishes yHac1 binding, correlating with the loss of its stimulatory activity. Thus, if YY1 requires mammalian Hac1 binding to ERSE-163 to act and because the production of mammalian Hac1 is expected to be regulated by ER stress, this could explain why the constitutively expressed YY1 can stimulate the grp78 promoter only under ER stress conditions. The same mutation raises the question of how is it possible that two co-factors that interact at the protein-protein level and have the same DNA-binding sequence requirement (CAGC) may co-occupy the site, if binding of both is a prerequisite. A possible explanation is that when the complex is formed between yHac1 and YY1, the latter binds the AC-rich sequence on the noncoding strand immediately 3Ј to the CAGC motif. In . Both promoters contain a UPRE or UPRElike functional element that has the ability to bind Hac1, a bZIP protein. The Ada family of proteins serve as adapter proteins between Hac1 and Gcn5 in S. cerevisiae, whereas YY1 as an interactive partner of p300 and its associating protein PCAF could serve as the bridge protein in mammalian cells. Both Gcn5 and PCAF possess histone acetylase activity that could activate transcription through modification of chromatin structure. There are also unique features of each promoter system. The yeast promoter contains a heat shock element that is absent from the rat grp78 promoter, whereas the grp78 promoter contains two other functional ERSEs downstream of ERSE-163. In addition, the mammalian ERSE contains a tripartite structure, with the CCAAT or CCAAT-like motif binding NF-Y/CBF, and the 9-bp region required for the stress-inducible ERSF binding (10). The model predicts that a Hac1-like protein yet unidentified in mammalian cells binds to the UPRE-like sequence within ERSE-163 and, in cooperation with YY1 and other co-factors, activates grp78 stress induction. support, the M2 mutation of the grp78 core, which has mutated the CAGC sequence but contains intact the adjacent AC-rich sequence and the rest of the larger YY1 binding site, was able to compete partially with the wild-type grp78 core for YY1 binding when compared with other mutated sequences (16). Several properties of YY1 suggest that it could play a role similar to that of the co-activator complexes that act in concert with Hac1 and Ire1p in yeast. YY1 and p300/CBF are known to form a physical complex in vivo and in vitro. Binding assays further mapped the p300-interacting domain to the C-terminal half of YY1 (33). By recruiting p300/CBF and its associating factors such as PCAF with intrinsic histone acetylase activity (34), YY1 binding to ERSE-163 could lead to targeted histone acetylation, resulting in change of chromatin structure and activation of gene transcription (Fig. 10). Preliminary analysis suggests that histone acetylation is part of the regulatory machinery of grp expression in mammalian cells. 2 Despite the strong prediction of the existence of a mammalian Hac1 based on the isolation of yHac1 in 1996, the mammalian homologs of yeast Ire1p have only been isolated recently. The identification of a mammalian Hac1 activator protein defined as a direct substrate of the Ire1p kinase and the ability to bind a UPRE-like sequence in mammalian chaperone promoters remains elusive. Because of the lack of a genetic screen in mammalian cells, the isolation of mammalian Hac1, if one exists, relies on precise definition of its binding site and transactivating properties on the mammalian ERSE. Here we identified the boundaries of a unique binding site for yHac1 within ERSE-163 and have created mutants that can dissect its binding from the overlapping YY1 binding site. This will greatly facilitate the isolation and characterization of a mammalian equivalent of yHac1, which will lead to a direct test of the model proposed for the ER stress induction of grp78 and other coordinately regulated ER-chaperone gene promoters in mammalian cells.