The transcriptional co-activator ADA5 is required for HAC1 mRNA processing in vivo.

Accumulation of unfolded proteins in the endoplasmic reticulum (ER) activates signaling pathways to induce transcription of a number of genes encoding ER protein chaperones and-folding catalysts. In Saccharomyces cerevisiae this transcriptional induction is mediated by an increase in the synthesis of the transcription factor Hac1p. The transmembrane receptor Ire1p/Ern1p containing a Ser/Thr protein kinase and endoribonuclease activity transmits the unfolded protein response (UPR) from the ER to the nucleus. Activation of Ire1p kinase induces its endoribonuclease activity to cleave unspliced HAC1 mRNA and generate exon fragments that are subsequently ligated by tRNA ligase (RLG1). Whereas unspliced HAC1 mRNA is poorly translated, spliced HAC1 mRNA is efficiently translated. Subunits of the yeast transcriptional co-activator complex SAGA also play a role in the UPR. Deletion of GCN5, ADA2, or ADA3 reduces, and deletion of ADA5 completely abolishes, the UPR. Although HAC1 mRNA requires only Ire1p and Rlg1p in vitro, we demonstrate that ADA5 is required for the IRE1/RLG1-dependent splicing reaction of HAC1 mRNA in vivo. In addition, Ada5p interacts with Ire1p. These results suggest that subcomponents of transcriptional co-activator complexes may be involved in RNA processing events.

Accumulation of unfolded proteins in the endoplasmic reticulum (ER) activates signaling pathways to induce transcription of a number of genes encoding ER protein chaperones and-folding catalysts. In Saccharomyces cerevisiae this transcriptional induction is mediated by an increase in the synthesis of the transcription factor Hac1p. The transmembrane receptor Ire1p/Ern1p containing a Ser/Thr protein kinase and endoribonuclease activity transmits the unfolded protein response (UPR) from the ER to the nucleus. Activation of Ire1p kinase induces its endoribonuclease activity to cleave unspliced HAC1 mRNA and generate exon fragments that are subsequently ligated by tRNA ligase (RLG1). Whereas unspliced HAC1 mRNA is poorly translated, spliced HAC1 mRNA is efficiently translated. Subunits of the yeast transcriptional co-activator complex SAGA also play a role in the UPR. Deletion of GCN5, ADA2, or ADA3 reduces, and deletion of ADA5 completely abolishes, the UPR. Although HAC1 mRNA requires only Ire1p and Rlg1p in vitro, we demonstrate that ADA5 is required for the IRE1/RLG1-dependent splicing reaction of HAC1 mRNA in vivo. In addition, Ada5p interacts with Ire1p. These results suggest that subcomponents of transcriptional co-activator complexes may be involved in RNA processing events.
The lumen of the endoplasmic reticulum (ER) 1 is a highly specialized compartment in eukaryotic cells. It is the primary site for folding of secretory and transmembrane proteins that constitute about one-third of all cellular proteins. The ER lumen provides an oxidizing environment and contains a number of resident chaperones, such as BiP (GRP78), that facilitate protein folding. These chaperones are proposed to catalyze protein folding and/or prevent aggregation of folding intermediates. Perturbation of the protein folding machinery in the ER leads to an accumulation of unfolded proteins. Cells respond to unfolded protein in the ER by up-regulating the synthesis of resident chaperones, thereby increasing the folding capacity in the ER compartment. This cellular response is termed the unfolded protein response (UPR) and is conserved in all eukaryotic organisms (1).
In the budding yeast Saccharomyces cerevisiae, genes that are transcriptionally activated in response to unfolded proteins in the ER contain a 22-base pair cis-acting promoter element, termed the UPRE, that is necessary and sufficient to mediate the transcriptional induction (2). The UPREs in these genes contain a partially palindromic sequence (CAGCGTG) with a spacer of one nucleotide that is required for the transcriptional induction (3). The trans-acting factor that binds the UPRE is a basic leucine zipper protein (bZIP) called Hac1p (4,5). Recent studies demonstrated that activation of the UPR is dependent upon the cellular levels of Hac1p. HAC1 is constitutively transcribed independently of the protein folding status in the ER. In the absence of unfolded proteins in the ER, HAC1 mRNA is exported to the cytoplasm and engaged with ribosomes, but translation is stalled. A 252-nucleotide fragment near the 3Ј end of HAC1 mRNA acts as a translational attenuator to limit Hac1p synthesis (6,7). However, in the presence of unfolded proteins in the ER, the translational attenuator in HAC1 mRNA is removed by an unconventional splicing reaction that does not utilize the cellular spliceasomal machinery. HAC1 mRNA splicing allows efficient translation of Hac1p and subsequent transcriptional activation of genes regulated by the UPR.
Perhaps the most important component of the UPR pathway is Ire1p/Ern1p (8,9). Ire1p is a type 1 transmembrane protein of the ER that is a bifunctional enzyme. The cytosolic domain contains both a serine/threonine kinase (10, 11) and a sitespecific endoribonuclease activity (12). In response to unfolded proteins in the ER, Ire1p forms oligomers that mediate transautophosphorylation to activate the site-specific endoribonuclease activity that cleaves at the splice site junctions within the 3Ј end of HAC1 mRNA. The cleaved 5Ј and 3Ј exons of HAC1 mRNA are ligated together by the tRNA ligase RLG1, generating a new mRNA species that is efficiently translated (13). Because permanent activation of the UPR pathway is detrimental to cell growth (11,14), the UPR needs to be tightly regulated. Upon phosphorylation, Ire1p recruits a serine/threonine phosphatase, PTC2, that dephosphorylates Ire1p and down-regulates the UPR (15).
The transcriptional co-activator Gcn5p was isolated as a specific interactor with Ire1p in a yeast two-hybrid screen (16). In eukaryotes, transcriptional activation requires functional interaction between the activators that bind upstream activating sequences and the basal factors that occupy the TATA box. It is proposed that functional interaction between these two classes of transcription factors is mediated by transcriptional co-activators. Gcn5p along with other transcriptional co-activators, including Ada1, Ada2p, Ada3p, Spt3p, Spt7p, Spt8p, and Ada5p/Spt20p, constitute a 1.8-mDa SAGA complex that is responsible for histone acetylation during transcriptional activation (17)(18)(19)(20)(21)(22)(23)(24). Mutations in the ADA genes were isolated as suppressors of the lethality induced by over-expression of the herpes simplex virus potent acid transcriptional activator VP-16 (19). The SPT genes were isolated as suppressors of Ty transposon insertions (23). Both sets of genes are required for maximal transcriptional activation from TATA-containing promoters. Yeast strains lacking Gcn5p, Ada2p, and Ada3p are partially defective in transcriptional activation of genes encoding ER protein chaperones in response to unfolded proteins in the ER. In contrast, cells lacking Ada5p are completely defective for the UPR but contain an intact heat shock response, demonstrating a specific requirement for Ada5p in the UPR (16). In the present study, we investigated the role of ADA5 in the UPR pathway. We found that apart from its function as a transcriptional co-activator in the UPR pathway, ADA5 is also required for HAC1 mRNA processing, a hitherto novel function for a transcriptional co-activator. Moreover, our data reveal the molecular basis for the defective UPR in ⌬ada5 cells.

EXPERIMENTAL PROCEDURES
Yeast Strains, General Methods, and Plasmid Constructions-The Escherichia coli strain DH5␣ was used for the propagation of plasmids. The genotype of the S. cerevisiae strain used in this study was BWG1-7a, Mat␣ leu2::3,112 his4 -519 ade1-100 ura3-52 (19). The genetic methods and standard media were previously described (25).
The construction of pAW65 (16), pGEM4Z-ACT1 (15), and the fusion protein containing the Ire1p cytoplasmic (10) domain with glutathione S-transferase (GST) were previously described. To construct GST-HAC1 fusion constructs, PCR amplified fragments from the wild-type HAC1 were subcloned into the BamHI and EcoRI sites of the bacterial expression vector, pGEX1T (Amersham Pharmacia Biotech). The expression GST-Hac1p fusion product was confirmed by Western blotting with anti-GST antibodies. To construct pGEM4Z-HAC1, a 240-base pair fragment from HAC1 was amplified by PCR using primers 5Ј-ccctctagacttcaagtcgactctgcctcc-3Ј and 5Ј-tcaaagcttgcaacaaaagcgtcgtggc-3Ј and subcloned into the XbaI and HindIII sites of pGEM-4Z. Similarly, a 206-base pair fragment from IRE1 was amplified by PCR using primers 5Ј-gagctgggaattcactcttatg-3Ј and 5Ј-aataagcttggggaatgtcggagaaaccc-3Ј and subcloned into the EcoRI and HindIII sites of pGEM-4Z to derive pGEM4Z-IRE1. Construction of plasmid pJC835 containing HAC1 i was described previously (4).
The yeast strain carrying the null allele of HAC1 was created by the method of one step gene disruption (26). BGW1-7a cells were transformed with BamHI digested pHAKO1 (4). Transformants containing the disrupted hac1 were selected for uracil prototrophy and the gene disruption was confirmed by PCR and Southern/Northern blot analyses. The construction of ⌬ire1 and ⌬ada5 strains was described previously (16).
Analysis of Protein Expression-Yeast cell lysates were prepared according to Williams et al. (27). Western blotting was performed by standard procedures (28) using anti-Hac1p primary antibodies (generously provided by Dr. Peter Walter, University of California, San Francisco) and horseradish peroxidase-conjugated goat anti-mouse secondary antibodies (Life Technologies, Inc.). Bands were detected using the enhanced chemiluminescence (ECL) kit (Amersham Pharmacia Biotech) and quantified using NIH image software. Total protein and ␤-galactosidase activity were measured using commercially available kits from Bio-Rad and Promega (Madison, WI), respectively.
In Vitro Pull-down Assays-Plasmids carrying GCN5, ADA2, ADA3, and ADA5 under the T7 promoter (generously provided by Dr. Leonard Guarente, Massachusetts Institute of Technology) were used in coupled in vitro transcription and translation assays to create [ 35 S]methionine labeled products. Equal amounts of each recombinant protein were incubated with glutathione-Sepharose beads containing equal amounts of either GST-Ire1p cytoplasmic domain or GST-Hac1p fusion protein at 4°C for 2 h in the binding buffer (phosphate-buffered saline, 10% glycerol, 2 mM EDTA). As a control, beads containing comparable amounts of GST were used. Beads were recovered, washed sequentially with phosphate-buffered saline containing 10% glycerol and 1, 0.5, and 0.05% Triton X-100, and boiled for 3 min in 1ϫ Laemmli buffer (29). Extracts were electrophoresed on 10% SDS-polyacrylamide gels under reducing conditions, treated with En/Hance (NEN Life Science Prod-ucts), and analyzed by autoradiography and PhosphorImager scanning (Molecular Dynamics, Sunnyvale, CA).
RNase Protection Analysis-Total RNA was isolated according to Schmitt et al. (30) from cells treated with or without Tm for 90 min for analysis using an RNase protection kit (Roche Molecular Biochemicals). pGEM-4ZHAC1, pGEM-4ZIRE1, and pGEM-4ZACT1 were linearized with XbaI, and antisense RNA probes were synthesized using T7 RNA polymerase (Roche Molecular Biochemicals) and [␣ 32 P]CTP (Amersham Pharmacia Biotech). A DNA sequencing ladder from a known template was used as size markers.

Transcription of HAC1 and/or IRE1 Is Not Attenuated in
⌬ada5 Cells-GCN5, ADA2, ADA3, and ADA5 are required for the maximal transcriptional induction of genes encoding ER resident chaperones including KAR2/BiP in response to protein misfolding in the ER (16). In contrast, the SAGA complex is not required for the heat-mediated transcriptional activation of KAR2 (16), suggesting that these co-activators play a specific role in the UPR as opposed to global transcriptional induction. Among these co-activators, Ada5p is the most interesting because cells lacking Ada5p are completely defective in responding to unfolded proteins in the ER and are inositol auxotrophs, the two phenotypes associated with ⌬ire1 and ⌬hac1 cells. To elucidate the mechanistic role of the SAGA complex in the UPR pathway, we asked whether the defective UPR in cells lacking these co-activators results from inefficient HAC1 transcription. HAC1 encodes the bZIP transcription factor Hac1p, which is essential for transcriptional induction of genes responding to the UPR. The levels of HAC1 mRNA were measured in ⌬ada5 cells by an RNase protection assay. In this assay, the presence of HAC1 mRNA (both processed and unprocessed) should protect a 240-base pair nucleotide fragment from the internal labeled probe. ACT1 mRNA served as an internal control for the amount of RNA in the reaction. The results show that HAC1 was transcribed in ⌬ada5 as well as in ⌬ire1 cells. The level of HAC1 transcription in these cells was comparable with that in the wild-type cells with or without tunicamycin (Tm) treatment, a drug that inhibits N-linked glycosylation and therefore disrupts protein folding in the ER (Fig. 1A, lanes 1-4   FIG. 1. Expression of IRE1 and HAC1 in ⌬ada5 cells. Total RNA was isolated for RNase protection assay from wild-type and ⌬ada5 cells grown to early log phase and treated with or without Tm (2 g/ml) for 90 min. The RNA protection assay was performed with HAC1 (A) and IRE1 (B) probes. nt, nucleotide. Shown are results for HAC1 (A) and IRE1 (B). and 6 -7). These results demonstrate that ⌬ada5 cells efficiently transcribe HAC1. Therefore, the ⌬ada5 cells are similar to ⌬ire1 cells where the abrogated UPR does not arise from defective HAC1 transcription.
The cellular level of Hac1p regulates the UPR. Although HAC1 is constitutively transcribed, only the processed transcripts are translated. HAC1 mRNA processing requires the activity of the site-specific endoribonuclease, Ire1p. We therefore asked whether IRE1 transcription is reduced in ⌬ada5 cells. An RNase protection assay demonstrated that both wildtype and ⌬ada5 cells had comparable levels of IRE1 mRNA, indicating that IRE1 is efficiently transcribed in ⌬ada5 cells (Fig. 1B, lanes 1-4). Therefore, the defective UPR in these cells is not due to reduced IRE1 transcription.
Components of the SAGA Complex Directly Interact with Ire1p and Hac1p-Gcn5p was identified as a component in the UPR by its interaction with Ire1p, detected by the yeast twohybrid system and by co-immunoprecipitation from cells coexpressing Gcn5p and Ire1p (16). To identify the component(s) of this complex that interact directly with Ire1p, in vitro affinity adsorbption experiments were performed with [ 35 S]methionine-labeled in vitro transcription-translation products of GCN5, ADA2, ADA3, and ADA5. Hac1p and Ire1p were produced as soluble GST fusion proteins (GST-Hac1p, GST-Ire1p) in E. coli and purified by adsorption to glutathione-Sepharose beads. Beads containing either GST-Hac1p or GST-Ire1p fusion proteins were used to capture the Gcn5p, Ada2p, Ada3p, or Ada5p translation products from reticulocyte lysates. As a control, Sepharose beads bound to equimolar amounts of GST were used. In these in vitro pull-down assays, the amounts of Ada2p, Ada3p, and Ada5p bound to GST-Hac1p were not significantly different from that obtained with control GST beads (Fig. 2,  lanes 5-8 and 10 -12). In contrast, GST-Hac1p brought down a 4.5-fold greater amount of Gcn5p than the GST control (Fig. 2,  lane 5 versus 9), indicating a direct interaction between the two proteins. GST-Ire1p, on the other hand, interacted with Ada2p and Ada5p (4.1-and 3.1-fold more than the GST control, respectively) (Fig. 2, lanes 14 and 16) but not with Gcn5p or Ada3p (Fig. 2, lanes 13 and 15). The specificity for the interaction with Ada5p was further demonstrated by the enrichment of the full-length Ada5p translation products by GST-Ire1p. These results indicate that Ire1p directly interacts with Ada2p and Ada5p and suggest that the original yeast two-hybrid interaction detected between Ire1p and the SAGA complex was mediated through Ada2p and Ada5p but not through Gcn5p. Therefore, the original interaction between Gcn5p and Ire1p detected in yeast was likely indirect and due to endogenous levels of Ada2p and Ada5p that could bridge Gcn5p with Ire1p.
HAC1 mRNA Processing Is Defective in ⌬ada5 Cells-The site-specific endoribonuclease activity of Ire1p is required for the HAC1 mRNA processing event that leads to the generation of a translatable form of HAC1 mRNA and subsequent activation of the UPR. The direct interaction between Ire1p and some components of the SAGA complex suggested that the SAGA complex might play a role in HAC1 mRNA processing. To test this hypothesis, we used a HAC1 probe to perform Northern blot analysis on RNA isolated from different yeast strains. In wild-type cells, upon Tm treatment, the majority of the HAC1 mRNA was processed to a smaller RNA species (HAC1 i , Fig.  3A, lanes 1 and 2). In contrast, upon Tm treatment of ⌬ire1 cells, processed HAC1 mRNA was not generated, indicating a complete absence of UPR-activated Ire1p-dependent HAC1 mRNA processing in these cells (Fig. 3A, lanes 3 and 4). Deletion of GCN5, ADA2, and ADA3 had no effect on the Ire1p-dependent HAC1 mRNA processing, as similar amounts of HAC1 i mRNA were detected after ER stress compared with the wildtype strain (Fig. 3A, lane 2 versus lanes 6, 8, and 10). Like the ⌬ire1 cells, ⌬ada5 cells were completely defective in HAC1 mRNA processing upon activation of the UPR (Fig. 3A, lanes 11  and 12), demonstrating that Ada5p plays a critical role in this mRNA processing reaction.
To substantiate the evidence for defective HAC1 mRNA processing in ⌬ada5 cells, Western blot analysis was performed to measure the steady state levels of Hac1p before and after ER stress. An antibody that specifically recognizes the carboxyl terminus of Hac1 i p can distinguish Hac1 u p from Hac1 i p because they differ in the carboxyl-terminal amino acids due to the Ire1p-mediated splicing reaction. Upon Tm treatment, Hac1 i p was increased 5-fold in wild-type cells (Fig. 3B, lanes 4  and 8) but not in ⌬ada5 cells (Fig. 3B, lanes 1 and 5) or ⌬ire1 cells (Fig. 3B, lanes 3 and 7). In this blot, a co-migrating background band was observed in all strains including ⌬hac1 without Tm treatment (Fig. 3B, lanes 2 and 6). These results are consistent with the observation that ⌬ada5 cells are defec-  9 -12), and GST-Ire1p (lanes [13][14][15][16] were tested for interaction with in vitro translated Gcn5p (lanes 5, 9, and 13), Ada2p (lanes 6, 10, and 14), Ada3p (lanes 7, 11, and 15), and Ada5p (lanes 8, 12, and 16). Proteins bound to the beads were analyzed by SDS-PAGE and fluorography.

FIG. 3. HAC1 mRNA is not processed in ⌬ada5 cells.
A, Northern blot analysis. Yeast cultures were grown in liquid YPD medium (yeast extract, peptone, and dextrose) to early log phase and then incubated for 90 min with (ϩ) or without (Ϫ) Tm (2 g/ml). Total RNA was isolated and probed with an HAC1-specific DNA probe as described under "Experimental Procedures." B, Western blot analysis. Cell lysates were made from yeast cultures grown as described in A and analyzed by Western blotting using an antibody that specifically recognizes the carboxyl terminus of Hac1 i p. tive in HAC1 u mRNA processing and are therefore incapable of generating a translatable form of HAC1 i mRNA.
Expression of Processed HAC1 (HAC1 i ) Partially Restores the UPR Pathway in ⌬ada5 Cells-The data presented thus far suggest that the abrogated UPR in ⌬ada5 cells result from defective HAC1 mRNA processing. We therefore asked whether the UPR could be restored upon expression of HAC1 i in ⌬ada5 cells. To evaluate the UPR, cells were transformed with a centromere-containing reporter plasmid harboring a lacZ reporter gene under the control of a single 22-base pair UPRE. The resultant transformed wild-type cells turned blue on plates containing X-gal and Tm, indicating an intact UPR pathway. In contrast, ⌬ada5 strains harboring the UPRE-lacZ vector remained white, demonstrating a defective UPR pathway. However, after the introduction of HAC1 i , both the wild-type and ⌬ada5 strains turned blue and light blue, respectively, on media containing X-gal plus Tm (data not shown). The amount of HAC1 mRNA detected in ⌬ada5 cells harboring the HAC1 i plasmid was greater than in the wild-type cells (Fig. 1A, lanes  1 and 2 and 9 and 10). This observation further strengthens the notion that the defective UPR in ⌬ada5 cells is not the result of abrogated HAC1 transcription. Quantification of the UPR by liquid assay, revealed that the basal level of ␤-galactosidase expression from the 22-base pair UPRE was reduced 3.5-fold in the ⌬ada5 strain compared with the wild-type strain (Fig. 4,  lanes 1 and 5). Upon Tm treatment, induction of ␤-galactosidase was reduced 20-fold compared with wild-type cells (Fig. 4,  lanes 2 and 6). Expression of HAC1 i increased the basal of ␤-galactosidase activity by 10-and 6-fold in the wild-type and⌬ada5 strains, respectively (Fig. 4, lanes 1, 3, 5, and 7). However, ␤-galactosidase expression in HAC1 i -expressing ⌬ada5 cells was reduced 6-fold compared with HAC1 i -expressing wild-type cells (Fig. 4, lanes 3 and 7). On the other hand, Tm induction of ␤-galactosidase activity was comparable in both wild-type and ⌬ada5 strains expressing HAC1 i , at 6-and 8-fold respectively (Fig. 4, lanes 3, 4, 7, and 8). Therefore, expression of Hac1 i p restored induction, although the basal expression level was reduced. Taken together, these results suggest that both transcriptional co-activator and RNA processing functions of ADA5 are essential for maximal transcriptional activation from the UPRE.

DISCUSSION
Upon accumulation of unfolded proteins in the lumen of the ER, Ire1p initiates a novel mRNA splicing reaction. The endoribonuclease activity of Ire1p cleaves the 5Ј and 3Ј splice site junctions within HAC1 mRNA. Each splice site junction is composed of a simple structure, a stem with a seven-member loop. Only 4 bases within the loop are apparently required for specificity of the cleavage reaction (31). The 5Ј and 3Ј exons of HAC1 mRNA are tethered together by base pairing and joined by tRNA ligase (RLG1). Whereas precursor HAC1 u mRNA is not translated well, the product HAC1 i mRNA is efficiently translated. This splicing reaction was reconstituted in vitro

HAC1 Splicing Requires Co-activator ADA5
with only two components, Ire1p and Rlg1p (12). However, data suggest that the HAC1 u mRNA processing reaction is more complex in vivo. First, the low degree of specificity for the stem-loop structure in HAC1 u mRNA would indicate that other cellular mRNAs may be nonspecifically cleaved by the Ire1p endoribonuclease. This notion is consistent with the observation that oligonucleotides consisting of the stem-loop structures are less efficient substrates than a 600-base pair substrate that contains both 5Ј and 3Ј splice site junctions (12). Second, when the intron within the 3Ј end of HAC1 u mRNA was placed into the 3Ј untranslated region of a green fluorescent protein (GFP) marker gene, GFP was not translated (7). This result is consistent with the proposed role of the intron as a translational attenuator (6). However, upon induction of the UPR, the HAC1 intron within the GFP mRNA was not removed, indicating that Ire1p does not simply recognize the stem-loop and intron sequence within HAC1 u mRNA. Fourth, although human Ire1␣p can cleave the 5Ј splice site of yeast HAC1 mRNA in vitro, yeast HAC1 mRNA was not cleaved when expressed in mammalian cells (32). Finally, using a temperature-sensitive mutant of RNA polymerase, it was shown that only newly transcribed HAC1 mRNA is processed by Ire1p (13). Although the majority of human Ire1p was localized within the ER membrane, a subpopulation preferentially associated with the nuclear envelope and possibly the nuclear pore complex (33), suggesting both cytoplasmic and nucleoplasmic localization of the carboxyl terminus. It was proposed that Ire1p cleaves HAC1 mRNA at the nuclear envelope as it is transported to the cytoplasm for translation. This hypothesis is consistent with detection of Rlg1p at the nuclear pore complex (34).
In this report, we provide data that demonstrate a requirement for ADA5 in the Ire1p-mediated cleavage of HAC1 mRNA in S. cerevisiae. First, whereas the UPR transcriptional induction is reduced severalfold in the ⌬ada2, ⌬ada3, and ⌬gcn5 strains, it is completely absent in the ⌬ada5 strain. Second, upon activation of the UPR, HAC1 i mRNA was produced in the ⌬ada2, ⌬ada3, and ⌬gcn5 strains, but not in the ⌬ada5 strain. As previously demonstrated, cleaved HAC1 mRNA products that are not ligated by the tRNA ligase accumulate in the cytoplasm (35). However, these products were not detected in the ⌬ada5 strain, supporting a role for Ada5p in HAC1 u mRNA cleavage and not in Rlg1p-mediated exon-exon ligation. Third, upon activation of the UPR, the levels of IRE1 mRNA and HAC1 mRNA were not altered, and Hac1 i p was not detected. Finally, an in vitro pull-down assay detected an interaction between Ire1p and both Ada5p and Ada2p, but not Ada3p and Gcn5p, consistent with a role of Ada5p in Ire1p-mediated cleavage. The requirement for ADA5 in the HAC1 mRNA cleavage reaction can explain both the UPR defect and the inositol requirement observed for the ⌬ada5 strain. These two phenotypes are identical to those observed in the ⌬ire1, ⌬hac1, and rlg1-100 strains (36). These results implicate a novel mRNA processing role for Ada5p, a subunit of the transcriptional co-activator SAGA.
Our experimental data support the proposed notion that Gcn5p interacts with the basic leucine zipper protein, Hac1p, to maximally activate the UPR (16). In vitro pull-down assays described here suggest that Gcn5p directly interacts with Hac1p. These findings are not surprising because another bZIP protein, Gcn4p, interacts with Gcn5p for maximal activation of amino acid biosynthetic genes under conditions of amino acid starvation (17,21).
We have incorporated the results of our study into a model for SAGA function in activation of the UPR (Fig. 5). Within the SAGA complex, Gcn5p has histone acetyltransferase activity that is required for opening chromatic structures for transcrip-tional activation. Disruption of ADA2, ADA3, or GCN5 destroys the histone acetyltransferase activity of the complex and reduces transcriptional activation of UPR-responsive genes by severalfold (16). Because ⌬ada2 cells, ⌬ada3 cells, and ⌬gcn5 cells grow without inositol and can splice HAC1 mRNA to activate the UPR, they are not required for the UPR. Gcn5p interacts with the bZIP transcription factor Hac1p to enhance transcriptional induction of ER stress-responsive genes. ADA5/SPT20 and SPT7 are required to maintain the integrity of the SAGA complex (24). Ire1p is localized to the inner leaflet of the nuclear envelope, adjacent to the nuclear pore. Both Ada2p and Ada5p interact with Ire1p, although only Ada5p function is required to promote Ire1p-mediated cleavage of HAC1 mRNA. Although we have not demonstrated that Ada5p acts directly in this reaction, the finding that Ada5p interacts with Ire1p would suggest that Ada5p may function to increase the efficiency of HAC1 mRNA processing by Ire1p.