Endoplasmic Reticulum Stress-responsive Transcription Factor ATF6α Directs Recruitment of the Mediator of RNA Polymerase II Transcription and Multiple Histone Acetyltransferase Complexes*♦

Background: Transcription factor ATF6α is a master regulator of genes induced by endoplasmic reticulum stress. Results: ATF6α can recruit Mediator and histone acetyltransferase complexes to promoter DNA via interactions with overlapping sites in the activation domain of ATF6α. Conclusion: ATF6α sequences essential for gene activation recruit Mediator and histone acetyltransferases. Significance: Learning how coregulators communicate with DNA binding transcription factors is important for understanding gene regulation. The basic leucine zipper transcription factor ATF6α functions as a master regulator of endoplasmic reticulum (ER) stress response genes. Previous studies have established that, in response to ER stress, ATF6α translocates to the nucleus and activates transcription of ER stress response genes upon binding sequence specifically to ER stress response enhancer elements in their promoters. In this study, we investigate the biochemical mechanism by which ATF6α activates transcription. By exploiting a combination of biochemical and multidimensional protein identification technology-based mass spectrometry approaches, we have obtained evidence that ATF6α functions at least in part by recruiting to the ER stress response enhancer elements of ER stress response genes a collection of RNA polymerase II coregulatory complexes, including the Mediator and multiple histone acetyltransferase complexes, among which are the Spt-Ada-Gcn5 acetyltransferase (SAGA) and Ada-Two-A-containing (ATAC) complexes. Our findings shed new light on the mechanism of action of ATF6α, and they outline a straightforward strategy for applying multidimensional protein identification technology mass spectrometry to determine which RNA polymerase II transcription factors and coregulators are recruited to promoters and other regulatory elements to control transcription.

The basic leucine zipper transcription factor ATF6␣ functions as a master regulator of endoplasmic reticulum (ER) stress response genes. Previous studies have established that, in response to ER stress, ATF6␣ translocates to the nucleus and activates transcription of ER stress response genes upon binding sequence specifically to ER stress response enhancer elements in their promoters. In this study, we investigate the biochemical mechanism by which ATF6␣ activates transcription. By exploiting a combination of biochemical and multidimensional protein identification technology-based mass spectrometry approaches, we have obtained evidence that ATF6␣ functions at least in part by recruiting to the ER stress response enhancer elements of ER stress response genes a collection of RNA polymerase II coregulatory complexes, including the Mediator and multiple histone acetyltransferase complexes, among which are the Spt-Ada-Gcn5 acetyltransferase (SAGA) and Ada-Two-Acontaining (ATAC) complexes. Our findings shed new light on the mechanism of action of ATF6␣, and they outline a straightforward strategy for applying multidimensional protein identification technology mass spectrometry to determine which RNA polymerase II transcription factors and coregulators are recruited to promoters and other regulatory elements to control transcription.
Endoplasmic reticulum (ER) 3 stress activates signal transduction pathways implicated in a variety of human diseases, including atherosclerosis, diabetes, and neurodegeneration (1)(2)(3). A major signaling pathway activated by ER stress is the unfolded protein response, which is triggered by accumulation of misfolded proteins in the ER and serves to protect cells from ER stress in part by down-regulating synthesis of some ER-destined proteins and by up-regulating expression of proteins involved in ER protein folding.
A critical downstream event in the unfolded protein response signal transduction pathway is activation of transcription of a collection of ER stress response genes by a mechanism that depends on the basic leucine zipper transcription factor ATF6␣ (1)(2)(3). In unstressed cells, ATF6␣ resides in the ER as a type II transmembrane protein. In response to ER stress, ATF6␣ is transported to the Golgi apparatus, proteolytically processed by the site 1 and site 2 proteases, S1P and S2P (4 -8), and then released into the nucleus, where it binds together with constitutively expressed transcription factors including NF-Y and Yin-Yang 1 (YY1) to ERSEs in the promoters of ER stress response genes (4 -11). Ectopic expression of the proteolytically processed form of ATF6␣ is sufficient to activate the ER stress-induced transcriptional program in unstressed cells, arguing that ATF6␣ acts as a master regulator of ER stress response genes (4,9,12,13).
Although cell biological studies have provided significant insight into the mechanism by which ATF6␣ is proteolytically * This work was supported, in whole or in part, by National Institutes of Health processed and transported to the nucleus, comparatively little is understood about how it activates transcription of ER stress response genes. ATF6␣ has an N-terminal transcription activation domain and a C-terminal DNA binding domain, whose affinity for its binding sites in ERSEs can be enhanced through interactions with NF-Y (11,12,14,15). Upon ER stress, multiple Pol II coregulators including the histone acetyltransferases (HATs) CREB-binding protein (CBP)/p300 and SAGA, the protein arginine methyltransferase PRMT1, and the INO80 chromatin remodeling complex have been shown to be recruited to ERSEs in the promoter of HSPA5 and other stress response genes (10, 16 -18). Results of prior studies suggest that PRMT1 and the INO80 complex are recruited through YY1 (10,17), but it is not known whether the ATF6␣ activation domain also contributes to recruitment of these or other coregulators.
In this study, we applied a combination of biochemical and multidimensional protein identification technology (MudPIT)based mass spectrometry approaches to investigate the role of ATF6␣ in recruitment of Pol II coregulatory proteins to the ERSEs of the HSPA5 gene. Below we present our findings, which are consistent with the model that ATF6␣ activates transcription at least in part by orchestrating the recruitment of a collection of Pol II coregulators, including the Mediator, SAGA, and ATAC complexes, to the ERSEs of its target genes. Dissection of the mechanism by which ATF6␣ recruits Pol II coregulators argues that Mediator and HAT complexes are recruited through interactions with nonidentical, but overlapping, regions of the ATF6␣ transcription activation domain. Taken together, our findings shed new light on the biochemical mechanisms underlying ATF6␣-dependent transcription activation, and they provide a relatively straightforward strategy for exploiting the sensitivity of MudPIT mass spectrometry to determine how particular DNA binding transcription factors recruit Pol II coregulators to genes.
Immobilized HSPA5 Promoter Recruitment Assay-To generate biotinylated DNA containing the wild type HSPA5 promoter, a fragment containing 429 bp of the human HSPA5 promoter (Ϫ282 to ϩ147 relative to the transcriptional start site) was amplified from human genomic DNA by PCR and cloned into pGEM-T Easy Vector (Promega) to generate pGEM-HSPA5P. A biotinylated 673-bp fragment was amplified from pGEM-HSPA5P using a biotinylated primer derived from the T vector and the reverse HSPA5 primer. A biotinylated HSPA5 fragment lacking the ERSEs (HSPA5⌬ERSE) was constructed using a two-step PCR procedure. In the first step, PCR products containing sequences immediately upstream and downstream of the ERSEs were amplified and purified on an agarose gel. In the second step, these PCR products were used as templates for the final PCR reaction. Biotinylated DNA fragments were agarose gel-purified and immobilized on Dynabeads M-280 (Invitrogen) streptavidin magnetic beads (1 pmol of biotinylated DNA:24 l of 1% (w/v) slurry of beads) according to the manufacturer's instructions.
Nuclear extracts were prepared from suspension cultures of HeLa-S3 cells as described (20). 10 -20 l of nuclear extract, with or without 6 pmol of GST-ATF6␣, were incubated for 10 min at 30°C in buffer PB (20 mM Hepes, pH 7.9, 0.1 M NaCl, 15% glycerol, 2 mM MgCl 2 , 0.15 mM EDTA, 1 mM DTT, 0.02% Nonidet P-40) containing 0.1 mg/ml bovine serum albumin (BSA) and 150 g/ml poly(dI⅐dC) in a total volume of 80 l. This mixture was then added to ϳ450 fmol of immobilized HSPA5 promoter fragment on Dynabeads that had been pre-equilibrated with PB containing 0.1 mg/ml BSA. The mixture was incubated a further 30 min at 30°C with occasional mixing. The beads were washed twice with 100 l of PB containing 0.1 mg/ml BSA and once with PB lacking BSA. Bound proteins were eluted with SDS sample buffer (2% SDS, 63 mM Tris-Cl, pH 6.8, 10% glycerol, 0.0025% bromphenol blue, 1.25% ␤-mercaptoethanol) at 99°C for 5 min and analyzed by Western blotting. For MudPIT analysis, 800-l binding reactions contained 100 -200 l of nuclear extracts, 25 pmol of GST-ATF6␣, and ϳ4.5-pmol DNA fragment on Dynabeads. Beads were washed twice with 100 l of PB containing 0.1 mg/ml BSA and once with PB lacking BSA. Bound proteins were then eluted by incubating with 2% SDS, 50 mM Tris-Cl, pH 8.8, 1.25 mM ␤-mercaptoethanol at 70°C for 10 min.
In Vitro Transcription-Biotinylated DNA fragments containing the wild type HSPA5 promoter (Ϫ282 to ϩ147) or HSPA5 promoter lacking the TATA-box sequence TATAAAG were generated by PCR, gel-purified, and immobilized on Dynabeads. HeLa cell nuclear extracts were prepared as described (20) except that nuclei were extracted with 0.24 M KCl and dialyzed into 20 mM Hepes, pH 7.9, 20% glycerol, 0.2 mM EDTA, 100 mM KCl, 1 mM DTT, and 0.5 mM PMSF. Transcription reactions were performed in two stages. First, 25-l binding reactions containing 10 mM Hepes, pH 7.9, 10% glycerol, 0.1 M KCl, 2 mM MgCl 2 , 0.1 mM EDTA, 0.5 mM DTT, 0.25 mM PMSF, 0.1 mg/ml BSA, 12 l of dialyzed nuclear extract, and ϳ450 fmol of immobilized DNA fragment, with or without GST-ATF6␣, were incubated for 30 min at 30°C. Unbound proteins were removed, and beads were equilibrated in transcription buffer by washing twice with 10 mM Hepes, pH 7.9, 10% glycerol, 0.05 M KCl, 6 mM MgCl 2 , 0.1 mM EDTA, 0.5 mM DTT, 0.25 mM PMSF. To initiate transcription, beads were resuspended in 25 l of the same buffer containing 40 units of RNasin (Promega) and 0.4 mM each of ATP, CTP, GTP, and UTP. After 90 min at 30°C, reaction products were analyzed by primer extension as described (21) using an oligonucleotide with the sequence GCT TCC CTC TCA CAC TCG CG.
GST-ATF6␣ Binding Assays-GST-ATF6␣ expression plasmids were introduced into pET41 and transformed into BL-21 DE3 cells. Cells from a single freshly transformed colony were grown to an A 600 of about 0.5, and protein expression was induced by adding isopropyl-1-thio-␤-D-galactopyranoside to 1 mM. Cells were grown overnight at 16°C with shaking at 130 rpm. Cells were harvested at 6000 ϫ g and resuspended in 20 ml of 50 mM Tris-Cl, pH 7.9, 300 mM NaCl, 10% glycerol, 0.5 mM 4-(2-aminoethyl) benzenesulfonyl fluoride hydrochloride, 1 mM DTT. After lysis with a French press, the cell suspension was clarified by centrifugation for 30 min at 100,000 ϫ g at 4°C, and the resulting supernatant was brought to 0.1% Triton X-100. GST-tagged proteins were purified from cell lysates using glutathione-Sepharose TM 4B beads (GE Healthcare) using standard methods. Purified proteins were exchanged into buffer containing 40 mM Hepes-Cl, pH 7.9, 0.05% Triton x-100, 1.5 mM MgCl 2 , 0.1 M NaCl, 1 mM DTT (GB buffer) using Zeba desalt spin columns (Thermo Scientific).
ϳ6 pmol of GST-ATF6␣ or GST-ATF6␣ deletion mutants were mixed with 10 -15 l of HeLa-S3 nuclear extract in GB buffer in a total volume of 80 l, incubated at 30°C for 30 min, and then added to 20 l of glutathione-Sepharose TM 4B equilibrated in GB buffer. The bead-protein slurry was incubated at 4°C for 2 h on an Adams TM nutator mixer (BD Diagnostics), washed three times with 100 l of GB buffer, and eluted with 3 bed volumes of 20 mM glutathione in GB buffer. The same procedure was used to purify proteins for MudPIT analysis, except that binding reactions contained ϳ25 pmol of GST-ATF6␣ and either 100 or 360 l of HeLa nuclear extracts, in total reaction volumes of 420 or 1500 l, respectively.
Mass Spectrometry-Proteins were identified using a modification of the MudPIT procedure (22,23). TCA-precipitated proteins were urea-denatured, reduced, alkylated, and digested with endoproteinase Lys-C (Roche Applied Science) and modified trypsin (Roche Applied Science) as described (22). Peptide mixtures were loaded onto 100-m fused silica microcapillary columns packed with 5-m C 18 reverse phase (Aqua, Phenomenex), strong cation exchange particles (PartiSphere SCX, Whatman), and reverse phase (24). Loaded microcapillary columns were placed in-line with an LTQ ion trap mass spectrometer equipped with a nano-LC electrospray ionization source (Thermo Scientific). Fully automated MudPIT runs were carried out on the electrosprayed peptides as described (23). Tandem mass (MS/MS) spectra were interpreted using SEQUEST (25) against a database of 37,394 human proteins (downloaded from NCBI on July 02, 2009), 4228 proteins from Escherichia coli BL21 strain (downloaded from NCBI on January 10, 2010), and 177 sequences from usual contaminants (human keratins, IgGs, proteolytic enzymes) and complemented with the sequences of each of the GST-ATF6 constructs used in this study. In addition, to estimate false positive discovery rates, each nonredundant sequence was randomized (keeping amino acid composition and length the same), and the resulting "shuffled" sequences were added to the "normal" database (doubling its size) and searched at the same time.
Peptide/spectrum matches were sorted and selected using DTASelect (26) with the following criteria set. Spectra/peptide matches were only retained if they had a DeltaCn of at least 0.08 and a minimum XCorr of 1.8 for singly charged, 2.0 for doubly charged, and 3.0 for triply charged spectra. In addition, peptides had to be fully tryptic and at least 7 amino acids long. Combining all runs, proteins had to be detected by at least two such peptides or one peptide with two independent spectra. Peptide hits from multiple runs were compared using CONTRAST (26). To estimate relative protein levels, distributed normalized spectral abundance factors (dNSAFs) were calculated for each nonredundant protein or protein group using the following equations as described by Zhang et al. (27) in which shared spectral counts (sSpC) are distributed based on spectral counts unique to each protein i (uSpC) divided by the sum of all unique spectral counts for the m protein isoforms that shared peptide j with protein i and where dSAF is the unnormalized distributed spectral abundance factor, and L refers to the length of each protein i in amino acids.

RESULTS AND DISCUSSION
ATF6␣-dependent Recruitment of Mediator and HAT Complexes to the HSPA5 ERSEs in Vitro-ATF6␣ activates transcription of ER stress response genes by binding sequence specifically to its binding sites in the ERSEs of their promoters. The promoter of the HSPA5 gene contains three ERSEs located ϳ60 bp upstream of the transcription start site (9) (Fig. 1A). In an attempt to identify in an unbiased way Pol II transcription regulatory proteins recruited by ATF6␣ to the HSPA5 promoter, we coupled DNA affinity purification with MudPIT mass spectrometry. MudPIT has proven to be a highly sensitive and reproducible means of identifying proteins in complex mixtures. In a MudPIT experiment, a mixture of proteins is first digested into peptides, which are then fractionated by multidimensional HPLC and analyzed by tandem mass spectrometry without first isolating individual proteins from gels (23). Previous studies have shown that for many proteins in a MudPIT dataset, the number of spectra from peptides of that protein is a function of its length and abundance. Consequently, the relative amount of a particular protein in different samples can often be estimated from a normalized spectral abundance factor, or dNSAF (27)(28)(29)(30).
To begin to investigate the function of ATF6␣ in regulation of ER stress response genes, we immobilized on streptavidin-Sepharose beads a biotinylated DNA fragment containing HSPA5 gene sequences from Ϫ282 to ϩ147 of the transcription start site (Fig. 1A); this portion of the HSPA5 gene includes the ERSEs, core promoter, and early transcribed region. In preliminary experiments, we confirmed that the immobilized HSPA5 promoter supports ATF6␣-activated transcription. To do so, we incubated the bead-bound HSPA5 DNA fragment with HeLa cell nuclear extracts, with or without the addition of recombinant ATF6␣, to allow preinitiation complex formation. After washing to remove unbound proteins, beads were incubated with ribonucleoside triphosphates, and transcription was measured using primer extension. As shown in Fig. 1B, a fragment containing the wild type HSPA5 promoter supported ATF6␣-activated transcription, whereas a fragment from which the TATA-box was deleted did not.
To identify proteins recruited to the HSPA5 promoter in an ATF6␣-dependent way, we incubated the bead-bound, wild type HSPA5 DNA fragment with HeLa cell nuclear extracts, with or without the addition of recombinant ATF6␣. After washing, bound proteins were eluted with buffer containing SDS and analyzed by MudPIT. As illustrated in the heat map of Fig. 1C, MudPIT identified subunits of the Mediator and several HAT complexes, including the SAGA, ATAC, and TRRAP-TIP60 complexes, as proteins recruited to the HSPA5 DNA fragment by ATF6␣. The Mediator complex is a very large, multisubunit complex that is required for proper regulation of the majority of Pol II genes and is thought to contribute to transcriptional regulation by acting directly on Pol II and other components of the general transcription apparatus. Although its detailed mechanism(s) of action are poorly understood, it has critical roles in multiple stages of transcription, from assembly and function of the Pol II initiation complex to control of transcript elongation (reviewed in Refs. [31][32][33][34]. The human SAGA, ATAC, and TRRAP-TIP60 HATs are also large multisubunit complexes. SAGA and ATAC share a common catalytic core that includes either of two closely related acetyltransferases, GCN5 or p300/CBP-associated factor (PCAF); each also includes a collection of additional subunits specific to one complex or the other (reviewed in Refs. [35][36][37][38]. Among the SAGA subunits is the transformation/transcription domainassociated protein (TRRAP), which is in turn shared by the TRRAP-TIP60 HAT complex (39 -41). SAGA, ATAC, and TRRAP-TIP60 catalyze acetylation of nucleosomal histones, and all contribute to transcriptional regulation at least in part by regulating histone acetylation in chromatin (35)(36)(37)(38).
The heat map shows the ratio of dNSAFs of Mediator and HAT subunits bound to immobilized HSPA5 promoter DNA in the presence (ϩ) or absence (Ϫ) of ATF6␣. The MudPIT results can be summarized as follows. First, all of the 30 Mediator subunits detected were enriched in the presence of ATF6␣. Twelve Mediator subunits were detected only when binding reactions included exogenous ATF6␣, and an additional eight Mediator subunits were enriched by more than 10-fold. Similarly, several of the Pol II subunits detected were seen only in the presence of ATF6␣. In addition to subunits of Mediator and Pol II, we also observed increased binding to the HSPA5 DNA fragment of subunits of the SAGA, ATAC, and TRRAP-TIP60 complexes. Finally, all three subunits of the ERSE binding transcription factor NF-Y were detected with similar dNSAF values when binding reactions were performed with or without added ATF6␣, consistent with previous studies indicating that NF-Y can be detected by chromatin immunoprecipitation at the HSPA5 ERSEs in the absence of ER stress (10,11).
The ATF6␣-dependent recruitment of Mediator, SAGA, and ATAC to the HSPA5 promoter was investigated further in assays using Western blotting to monitor recruitment of repre- promoter DNA. Preinitiation complexes were assembled with or without ATF6␣ on bead-bound wild type or ⌬TATA HSPA5 promoter fragments, washed, and assayed for their ability to support promoter-specific transcription as described under "Experimental Procedures." C, heat map showing the ratio of dNSAFs of proteins bound to immobilized HSPA5 promoter in the presence (ϩ) or absence (Ϫ) of ATF6␣. Immobilized HSPA5 promoter binding assays were performed with or without the addition of purified ATF6␣, and bound proteins were detected by MudPIT mass spectrometry. See supplemental Table 2 for supporting data.
sentative subunits of each complex. Consistent with the results of MudPIT experiments, recruitment of the Mediator subunit Med6 was almost completely dependent on exogenously added ATF6␣, whereas binding of SAGA subunit ADA2b and ATAC subunit MBIP was enhanced upon the addition of ATF6␣ ( Fig.  2A, lanes 1-4 and 9). Although not detected in our MudPIT datasets, we also observed that subunits of Pol II general initiation factors, including TFIID (TAF6), TFIIB, TFIIE, and TFIIF, were recruited to the HSPA5 promoter in an ATF6␣-dependent manner (Fig. 2B), consistent with previous studies indicating that recruitment to promoters of Mediator (19) and in some cases SAGA (42,43) is accompanied by recruitment of TFIID and other components of the Pol II preinitiation complex.
It was shown previously that ATF6␣-dependent transcription in cells requires a functional ERSE (9,12). Importantly, recruitment of ATF6␣ and Pol II coregulatory proteins to an HSPA5 promoter fragment that lacked the ERSEs (HSPA5⌬ERSE) was substantially reduced (Fig. 2A, compare  lanes 1-4 with lanes 5-8). We note, however, that binding of the ATAC subunit MBIP to HSPA5 was less dependent than MED6 or ADA2b on added ATF6␣ or an intact ERSE, suggesting that it has more nonspecific DNA binding activity and/or that additional factor(s) present in the nuclear extracts also contribute to its binding to the HSPA5 promoter.
ATF6␣-dependent Recruitment of Mediator and HATS to HSPA5 Promoter Depends on Both Its DNA Binding and Its Transcription Activation Domains-Previous studies have shown that the ATF6␣ transcription activation domain resides within its first 150 amino acids (12,15), whereas its basic leucine zipper family DNA binding domain lies between residues 308 and 373 (4). To determine which ATF6␣ regions are required for recruitment of Mediator and HATs to the immobilized HSPA5 promoter, binding reactions were performed using various ATF6␣ deletion mutants (Fig. 2C). Consistent with the earlier findings, deletion of a portion of the ATF6␣ DNA binding domain prevented it from binding the HSPA5 promoter, whereas deletion of the transcription activation domain had no effect on DNA binding. Arguing that the previously mapped ATF6␣ activation domain plays a key role in recruitment of Mediator and SAGA to the HSPA5 promoter in our assays, deletion of either the DNA binding domain or the transcription activation domain led to a dramatic decrease in Med6 and ADA2b recruitment to the promoter. In addition, although there was a significant background of ATAC subunit MBIP even in the absence of ATF6␣, the modest but reproducible ATF6␣-dependent increase in the amount of MBIP recruited to the immobilized promoter required both the ATF6␣ DNA binding and the transcription activation domains. We also observed a modest increase in NF-Y recruitment in the presence of full-length ATF6␣, consistent with previous evidence that ATF6␣ activation can lead to an ϳ2-fold increase in NF-Y detected at the HSPA5 ERSE by chromatin immunoprecipitation (11) and suggesting that co-occupancy of ERSEs by ATF6␣ and NF-Y may lead to an increase in the DNA binding affinity of NF-Y. The ATF6␣ Transcription Activation Domain Is Necessary and Sufficient for Mediator and HAT Binding-To begin to address the mechanism of ATF6␣-dependent recruitment of the Pol II transcription machinery to the HSPA5 promoter in our assays, we sought to determine whether ATF6␣ is capable of interacting with any of these Pol II transcription regulatory proteins in the absence of promoter DNA. To accomplish this, HeLa cell nuclear extracts were incubated with GST-ATF6␣ or one of several GST-ATF6␣ mutants (Fig. 3A). GST-ATF6␣ and associated proteins were enriched using glutathione-Sepharose and analyzed by MudPIT mass spectrometry and Western blotting. As shown in Fig. 3, B and C, these experiments revealed that GST fusion proteins containing the ATF6␣ transcription activation domain can interact with Mediator, Pol II, and each of the various HAT complexes recruited by ATF6␣ to the HSPA5 promoter in the experiments in Figs. 1 and 2, as well as with p300, which has been shown previously to be recruited to the HSPA5 ERSEs in response to ER stress (10). Binding of these Pol II coregulators to ATF6␣ was dependent on the ATF6␣ transcription activation domain because (i) GST-ATF6␣ (1-150) bound as much or more of the coregulators as did GST fused to full-length ATF6␣ and (ii) neither Mediator nor HAT complexes bound to GST-ATF6␣ (151-326), which lacks the transcription activation domain. Finally, although we were able to detect binding of Medi-  Table 2 for supporting data. C, Mediator and HAT subunits bound by GST-ATF6␣ fusion proteins. The indicated proteins were detected by Western blotting.
ator and HATs to the immobilized ATF6␣ transcription activation domain in the absence of HSPA5 promoter DNA, we did not detect any of the Pol II general initiation factors in these experiments (data not shown), suggesting that recruitment of these proteins by ATF6␣ occurs in the context of preinitiation complexes assembled on DNA.
Mediator and HAT Complexes Bind to Nonidentical but Overlapping Regions of ATF6␣ Transcription Activation Domain-We next sought to define in more detail portions of the ATF6␣ activation domain required for its interaction with Mediator and the various HAT complexes using the series of GST-ATF6␣ fusion proteins diagrammed in Fig. 4A. As shown in Fig.  4, B and C, there was little or no interaction between the Mediator complex and GST fusion proteins containing ATF6␣  or ATF6␣ (20 -60). ATF6␣ fragments containing residues 20 -80 and 44 -150 bound less well to Mediator than ATF6␣ (1-150), whereas very similar amounts of Mediator were recovered after glutathione-Sepharose chromatography of proteins that bound GST fusion proteins containing ATF6␣ (20 -100) and the full-length ATF6␣ activation domain (residues 1-150). In contrast, maximal binding of the HAT complexes to ATF6␣ was observed only to GST-ATF6␣ (1-150), suggesting that (i) Mediator and the HAT complexes interact with ATF6␣ through overlapping but nonidentical surfaces and (ii) optimal binding of the HAT complexes to ATF6␣ requires a larger region of the transcription activation domain.
Thureauf et al. (15) previously noted that the first 100 amino acids of the ATF6␣ transcription activation domain exhibit some sequence similarity to the VP16 transcription activation domain. Most similar within this region is an 8-amino acid sequence, DFDLDLMP, that closely resembles a VP16 sequence, DFDLDMLG, referred to as VN8 (15). In VP16, the VN8 sequence is necessary for transcription activation (44,45). In addition, mutations in the VN8 sequence have been reported to abolish interaction of VP16 with the Mediator complex (46), and mutations in the VN8-like domain of ATF6␣ led to a 5-fold reduction in ATF6␣-dependent activation of a luciferase reporter driven by the HSPA5 ERSEs (15).
To explore in more detail the potential functional relationship between the ATF6␣ and VP16 transcription activation domains, we first asked whether the VP16 activation domain competes with ATF6␣ for interaction with Mediator and HAT complexes. Remarkably, adding increasing amounts of a Gal4-VP16 transcription activation domain fusion protein to HeLa cell nuclear extracts prevented binding of GST-ATF6␣ (1-150) to Mediator and Pol II but not to the HAT complexes, as revealed by MudPIT and confirmed by Western blotting (Fig. 5,  A and B); that Pol II was depleted along with Mediator is consistent with the possibility that ATF6␣-dependent recruitment of Pol II is via a Mediator-Pol II holoenzyme complex. Importantly, the observation that the HAT complexes remain associated with ATF6␣ in the presence of Gal4-VP16 indicates that they can associate with ATF6␣ independent of Mediator.
We next considered the possibility that the ATF6␣ VN8-like sequence might contribute to Mediator and/or HAT recruitment. Indeed, we observed that mutating the ATF6␣ VN8-like sequence to DADALLP led to a dramatic decrease in the interaction of ATF6␣ with Mediator (Fig. 5, C and D), reducing it to a level similar to that seen when VP16 was included as competitor. Adding Gal4-VP16 to binding reactions containing the ATF6␣ VN8-like mutant caused little or no further reduction in the amount of Mediator bound. In addition, mutation of the ATF6␣ VN8-like sequence substantially reduced the interaction of ATF6␣ with SAGA and ATAC, although Gal4-VP16 does not prevent binding of ATF6␣ to SAGA and ATAC.  Table 2 for supporting data) and by Western blotting (B). C, wild type and mutant GST-ATF6␣ fusion proteins used in the experiment in panel D. The asterisk corresponds to the VN8-like sequence in the ATF6␣ transcription activation domain. Mutated residues are shown in red. D, glutathione-Sepharose purification of GST-ATF6␣binding proteins was performed with 12 pmol of wild type (WT) or mutant GST-ATF6␣ (1-373) as bait. Where indicated, binding reactions included 12 pmol of Gal4-VP16 as competitor; bound fractions were analyzed by Western blotting.
In complementary experiments, we observed that micromolar concentrations of a 32-amino acid peptide containing four repeats of the ATF6␣ VN8-like sequence blocked binding of GST-ATF6␣ to Mediator, whereas higher concentrations of the peptide were needed to block binding to SAGA (Fig. 6A). Furthermore, we observed that even at millimolar concentrations, the VN8-like peptide had little effect on thyroid hormone (T3)-dependent binding of Mediator to the activation domain of thyroid receptor (TR), supporting the specificity of inhibition by the VN8-like peptide and suggesting that ATF6␣ interacts with a different binding site on Mediator than does TR. Taken together, the observations described thus far suggest that the VN8-like sequence is important for the binding of both Mediator and the HAT complexes to the ATF6␣ transcription activation domain, but that additional sequences outside of the VN8-like region make a more significant contribution to the ATF6␣-HAT interaction.
Finally, we compared the effect of the VN8-like peptide on ATF6␣-dependent recruitment of Mediator, SAGA, and the TFIID component of the Pol II preinitiation complex with the HSPA5 promoter in vitro. As shown in Fig. 6B, concentrations of peptide that had little effect on SAGA binding strongly reduced binding of Mediator to the HSPA5 promoter. Notably, the VN8-like peptide reduced TFIID binding to the HSPA5 promoter in parallel with its reduction of Mediator binding, consistent with the possibility that Mediator contributes to ATF6␣-dependent TFIID recruitment to the promoter.
Summary and Perspectives-In this study, we present findings that shed new light on the mechanism of action of ATF6␣. By exploiting a combination of biochemical and MudPIT-based mass spectrometry approaches, we have obtained evidence that the ATF6␣ transcription activation domain can recruit a collection of Pol II coregulators to the ERSEs in the promoter of the ER stress response gene HSPA5. These Pol II coregulators include the Mediator complex as well as several HATs, including the SAGA, ATAC, and TRRAP-TIP60 complexes and p300. These findings are consistent with previous studies reporting that p300 and the SAGA complex exhibit increased occupancy at the HSPA5 ERSEs following induction of the ER stress response with thapsigargin (10,18). Similarly, in chromatin immunoprecipitation experiments, we have observed increased binding of the Mediator subunit Med26 at the HSPA5 promoter in cells subjected to ER stress (data not shown).
In structure-function experiments dissecting the mechanism by which ATF6␣ recruits coregulators, we observe that its binding to Mediator and HAT complexes depends on nonidentical but overlapping regions of its transcription activation domain. Importantly, the interactions between ATF6␣, Mediator, SAGA, and other HATs described in this study depend strictly on ATF6␣ domains shown previously to be essential for its transcription activity in cells (12,15). Of particular note, mutation of three residues in the ATF6␣ VN8-like sequence, which has been shown to be important for ATF6␣ activation domain function (15), led to a dramatic decrease in binding of ATF6␣ to Mediator, SAGA, and ATAC. Thus, our findings are consistent with the model that ATF6␣ regulates transcription of ER stress response genes at least in part by recruiting these Pol II coregulators to ERSEs.
In the course of these experiments, we observed that the well characterized VP16 transcription activation domain competes with ATF6␣ for binding to the Mediator but not to the HAT complexes, suggesting that the ATF6␣ and VP16 transcription activation domains might bind to the same or overlapping surfaces on Mediator. In light of previous studies indicating that VP16 binds to Mediator through its Med25 subunit (46 -51), we are now investigating the possibility that ATF6␣ might also bind to Mediator through Med25. In the future, these studies should provide a deeper understanding of the mechanism by which ATF6␣ facilitates recruitment of Mediator and other Pol II coregulators to the genes it regulates. Finally, our success in combining DNA affinity chromatography and MudPIT mass spectrometry to identify proteins recruited by ATF6␣ to the HSPA5 promoter suggests that this approach might be generally applicable in future studies aimed at determining how particular DNA binding transcription factors recruit Pol II coregulators to genes to regulate transcription.