Transcription factor IIA mutations show activator-specific defects and reveal a IIA function distinct from stimulation of TBP-DNA binding.

The general transcription factor IIA (TFIIA) binds to the TATA binding protein (TBP) and mediates transcriptional activation by distinct classes of activators. To elucidate the function of TFIIA in transcriptional activation, point mutants were created in the human TFIIA-gamma subunit at positions conserved with the yeast homologue. We have identified a class of TFIIA mutants that stimulate TBP-DNA binding (T-A complex) but fail to support transcriptional activation by several different activators, suggesting that these mutants are defective in their ability to facilitate an activation step subsequent to TBP promoter binding. Point mutations of the hydrophobic core of conserved residues from 65 to 74 resulted in various activation-defective phenotypes. These residues were found to be important for TFIIA gamma-gamma interactions, suggesting that gamma-gamma interactions are critical for TFIIA function as a coactivator. A subset of these TFIIA-gamma mutations disrupted transcriptional activation by all activators tested, except for the Epstein-Barr virus-encoded Zta protein. The gamma Y65F, gamma W72A, and gamma W72F mutants mediate Zta activation, but not GAL4-AH, AP-1, GAL4-CTF, or GAL4-VP16 activation. The gamma W72A mutant failed to stimulate TFIID-DNA binding (D-A complex) but was able to form a complex with TFIID and DNA in the presence of Zta (Z-D-A complex). Thus, the ability of Zta to activate transcription with gamma W72A appears to result from a unique ability to form the stable Z-D-A complex with this mutant. Our results show that different activators utilize the general factor TFIIA in unique ways and that TFIIA contributes transcription activation functions in addition to the facilitation of TBP-DNA binding.

Eukaryotic transcriptional activators stimulate the assembly of general transcription factors into a stable preinitiation complex at the start site of class II promoters (1,2). The binding of the general transcription factor IID (TFIID) 1 to the TATA box nucleates the formation of the preinitiation complex (3,4). TFIID consists of TATA binding protein (TBP) and TBPassociated factors (TAFs), which are essential for the reconsti-tution of activated transcription in vitro (5)(6)(7)(8). Both TBP and TAFs have been shown to interact directly with different classes of transcriptional activation domains (9 -16). Several experimental approaches indicate that binding of TBP to the promoter is rate-limiting in vivo and that activators that interact with TBP stimulate this step in preinitiation complex formation (17)(18)(19)(20)(21)(22).
TBP binding to the TATA box can be stimulated and stabilized by the general transcription factor IIA (TFIIA) (3,23,24). TFIIA has only a modest effect on basal transcription reconstituted with TBP and a minimal set of general transcription factors (25). In contrast, activator-regulated transcription reconstituted with TFIID (TBP plus TAFs) is strongly dependent on TFIIA (23, 26 -29), suggesting that a functional interaction exists between TFIIA, activators, and the TAFs. The assembly of a TFIIA⅐TFIID promoter complex has been shown to be rate-limiting in in vitro transcription reactions, and the acidic activator GAL4-AH was able to stimulate this rate-limiting step (30). The Epstein-Barr virus-encoded lytic activator, Zta, has also been shown to stimulate the formation of a TFIIA⅐TFIID promoter complex that is dependent on the TAFs and the activation domain of Zta (31,32). Thus, it is likely that some activators can stimulate TBP binding by enhancing an interaction between TFIID and TFIIA.
How TFIIA mediates activator function is not entirely clear. TFIIA may function by directly stimulating TBP binding to the TATA box (3,4,24,33). In addition, TFIIA has been shown to compete with repressors for binding to TBP. Thus, TFIIA has been implicated as a derepressor that prevents nonproductive preinitiation complex formation (23, 34 -39). TFIIA induces a conformational change in TBP (33) and interacts with the repeat of lysine residues on helix H2 of yeast TBP (40). This same basic region has been implicated in the TBP interaction with the adenovirus E1A transactivator (10). Mutations in this region of TBP were also defective for transcriptional activation by acidic activators and were reduced for their ability to bind TAF II 250 (41). Thus, TFIIA interacts with a region of TBP critical for multiple activator and TAF functions.
The cDNAs encoding both TFIIA subunits were initially isolated from yeast, and each subunit, TOA1 and TOA2, is required for cell viability (42)(43)(44). Subsequently, human and Drosophila TFIIA were purified as three polypeptides (␣, ␤, and ␥) with molecular weights of 35,19, and 12 kDa, respectively (45)(46)(47). Both the human and Drosophila 35-kDa (␣) and 19-kDa (␤) TFIIA subunits share sequence similarity to the amino and carboxyl-terminal ends of TOA1 (␣␤), respectively, and are encoded by a single gene (called ␣␤) (45-47). The nonconserved spacer domain of TOA1 (␣␤) is dispensable for yeast cell viability (48). Yeast depleted of TFIIA have normal RNA polymerase (pol) I and III activity, yet have reduced pol II transcription in vivo (48), indicating that TFIIA is primarily a * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Isolation of the cDNA encoding the human and Drosophila 12-kDa (␥) TFIIA subunit, a protein 58% homologous to TOA2 (␥), allowed the TFIIA activities to be reconstituted from recombinant components (27-29, 49, 50). Using TFIIA-depleted extracts, both basal and activated transcription in vitro required the addition of recombinant TFIIA (rTFIIA) (27,50). Transfection of both human TFIIA cDNAs in HeLa cells stimulates activator-dependent transcription 4-fold (28). TFIIA is also required to mediate activator-dependent transcription in highly purified transcription systems reconstituted with immunopurified or holo-TFIID (hIID) (27,28). Yeast TFIIA can substitute for human TFIIA in this system, indicating that TFIIA function is conserved in eukaryotes (27). TFIIA-␥ was selected from a yeast TBP mutant screen that had a defective acidic activator response, and the defective phenotype was also rescued by fusion of the mutant TBP to ␥ (51). These data indicate that the binding of TFIIA-␥ to TBP may mediate transcriptional stimulation by acidic activators.
Protein interaction assays reveal that TFIIA contacts multiple components of the preinitiation complex. The Drosophila and human ␣␤ and ␥ subunits both bind to TBP (27)(28)(29), Drosophila ␣␤ binds to dTAF II 110 (47), and human ␥ binds to the VP16 (52) and Zta (27) activators in an activation domaindependent manner. Zta was also shown to bind ␣␤ (52). The interaction of TFIIA with Zta results in the enhanced binding of TFIID as a stable promoter-bound complex, referred to as Z-D-A (32). We have proposed that TFIIA mediates an interaction between Zta and TFIID that induces TAFs to contact promoter sequences downstream of the transcription initiation site (32). Most other activators do not stimulate D-A complex formation as efficiently as does Zta, 2 suggesting that this type of interaction with TFIIA⅐TFIID may be limited to a certain class of activators. While TFIIA appears to be required for all activators to function, it is not clear whether TFIIA is required in the same capacity by all activators. Moreover, it is not clear whether TFIIA has functions in transcriptional activation in addition to the stabilization of TBP binding. To further examine the role of TFIIA in the activation process, we engineered point mutants in conserved TFIIA-␥ residues and assayed these mutants for TFIIA functions in transcription activation and stimulation of TBP binding.

EXPERIMENTAL PROCEDURES
TFIIA-␥ Antibodies (Abs) and Immunoblots-For Ab production, TFIIA-␥ was induced from either the pQE-IIA-␥ or pRSET A-IIA-␥ (see below) constructs in Escherichia coli and purified using Ni-NTA agarose chromatography as described (27). Subsequently, rTFIIA-␥ was gel-purified by 15% SDS-polyacrylamide gel electrophoresis, electroblotted onto nitrocellulose membrane, and detected by staining with Ponceau S. Rabbit Abs raised against purified rTFIIA-␥ were produced by Pocono Rabbit Farm and Laboratory (Canadensis, PA). For the immunoblots, 200 ng of either wild type (WT) or mutant TFIIA protein was used. A 1:500 dilution of rabbit anti-TFIIA-␥ serum was used as a primary Ab, followed by a 1:3000 dilution of horseradish peroxidaselinked goat anti-rabbit serum (Bio-Rad) as a secondary Ab. The immunoblot signal was detected by chemiluminescence (ECL) (Amersham Corp.).
Site-directed Mutagenesis-The site-directed TFIIA-␥ mutants were created using overlap extension PCR methodology as described (53). To produce each partial cDNA fragment for a particular ␥ mutant, first round PCR reactions contained 5 ng of pBS II KS-TFIIA-␥, 0.2 mM dNTPs, 50 mM KCl, 10 mM Tris, pH 8.4, 2.5 mM MgCl 2 , 2.5 units AmpliTaq (Perkin-Elmer), 20 pmol of each primer and were amplified on a DNA Thermal Cycler (Perkin-Elmer) in a 100-l reaction as follows: hot start, 94°C for 1 min, 52°C for 1 min, and 72°C for 1 min. (30 cycles). To create each full-length TFIIA-␥ mutant cDNA, both first round PCR mutant fragments were added together and used in the second round PCR "fusion" reaction. For each TFIIA-␥ mutant, 1 l of each of first round PCR reaction was added to 0.2 mM dNTPs, 10 mM KCl, 10 mM (NH 4 ) 2 SO 4 , 20 mM Tris-HCl, pH 8.8, 2 mM MgSO 4 , 0.1% Triton X-100, 1.0 unit of Vent DNA Polymerase (New England Bio-Labs), 100 pmol flanking primers, which were amplified as above. The ␥Y3A and ␥Y6A mutants had only (5Ј) sense, and the ␥K101A and ␥E109A mutants had only (3Ј) antisense mutant oligonucleotides synthesized and were amplified as the first round PCR reactions, except that 100 pmol primers were used. The PCR-amplified full-length TFIIA-␥ mutant cDNAs were further purified on Microcon-30 DNA concentrators (Amicon).
Plasmid Constructs-The expression construct for human WT TFIIA-␥ (pRSET A-IIA-␥) was created by PCR amplification of the TFIIA-␥ cDNA with a BamHI restriction site immediately preceding the initiation codon and a HindIII restriction site immediately following the termination codon. After digestion, this fragment was cloned into the BamHI and HindIII sites of pRSET A (Invitrogen). The TFIIA-␥ mutant fragments isolated from PCR (see above) were similarly cloned into the BamHI and HindIII sites of pRSET A, except for ␥Y3A, which was cloned into the BamHI and HindIII sites of pQE-9 (QIAGEN). All TFIIA-␥ mutations were confirmed by DNA sequencing in both orientations using an ABI automated 373A DNA sequencer. The resulting pRSET A-IIA-␥ WT or mutant constructs were expressed either in E. coli or in vitro in rabbit reticulocyte lysates (Promega) as indicated. Wild type TFIIA-␥ expressed from either pQE-9 or pRSET A showed no functional differences despite having different amino-terminal tags (data not shown). To create pGST-IIA-␣, the TFIIA-␣ fragment (codons 1-251 from the human ␣␤ gene) was derived from pQE-IIA-␣ (27). pQE-IIA-␣ was digested initially with HindIII, with a subsequent Klenow fill reaction, and followed by digestion with BamHI to isolate the ␣ subunit fragment. The GST vector, pGEX-2T (Pharmacia Biotech Inc.), was digested initially with EcoRI, with a subsequent Klenow fill reaction, and followed by digestion with BamHI. The TFIIA-␣ fragment was then directionally cloned into the digested pGEX-2T vector to create pGST-IIA-␣. The pGST-Zta, pGST-IIA-␥, and pGST-TBP constructs were described (27).
Protein Preparations-The pQE-IIA-␣␤ and pRSET A-IIA-␥ WT or mutant constructs were expressed in M15 and BL21 E. coli strains, respectively. Expressed proteins were purified under denaturing conditions on Ni-NTA agarose columns (Qiagen). The rTFIIA protein was isolated by column fractionation with elution denaturant (8 M urea, 0.1 M NaH 2 PO 4 , 0.01 M Tris, pH 8.0, 7 mM ␤-mercaptoethanol, 1 mM phenylmethylsulfonyl fluoride) of decreasing pH. The purified ␥ mutants were renatured with equal molar amounts of WT human ␣␤ subunit. Dialysis of the ␥ mutants into D100 buffer (20 mM Hepes, pH 7.9 (KOH), 20% glycerol, 0.2 mM EDTA Na 2ϩ , 100 mM KCl, 7 mM ␤-mercaptoethanol, 1 mM phenylmethylsulfonyl fluoride) was performed as described previously (27), except for the ␥E109A mutant, which required a gradual multistep reduction of urea and salt concentration to allow proper renaturation. Generally, the stepwise dialysis protocol yielded 50% higher soluble concentrations of WT or mutant TFIIA compared with Ref. 27 but did not improve the insolubility of the ␥D41A and ␥K42A mutants. rTFIIA polypeptides were used at a final concentration of ϳ0.2 M for transcription reactions. TBP and Zta protein was prepared as described (32). GAL4 fusion proteins were purified as described (54,55).
DNA Binding Reactions-Polyacrylamide EMSA conditions for T-A complex formation were similar to those described (56). Identical polyacrylamide EMSA results were obtained for T-A complex formation at either 22 or 30°C for 30 min. 34 nM of human recombinant TBP was incubated with 16 nM of either WT ␣␤ ϩ ␥ or WT ␣␤ ϩ ␥ mutant to form the T-A complex. For Mg-agarose EMSA analysis of Zta-hIID-IIA complex formation, 0.2 units of holo-TFIID (8), 16 nM Zta, and 50 nM of either WT ␣␤ ϩ ␥ or WT ␣␤ ϩ ␥ mutant was used (32).
In Vitro Transcription Reactions-The promoter templates Z 7 E4TCAT (57), G 5 E1BTCAT (58), pColϪ75/ϩ63CAT (AP-1) (59) were previously described. In vitro transcription reactions contained 100 ng of DNA template, ϳ100 -200 ng of purified recombinant activator protein, and 40 g of HeLa nuclear extract in a 50-l final reaction volume incubated for 1 h at 30°C. Primer extension reactions were described previously (8). The TFIIA-depleted nuclear extracts were prepared as described (27). Purified WT protein expressed from either pQE-IIA-␥ or pRSET A-IIA-␥ has identical in vitro transcription activity (data not shown). The addition of higher than stoichiometric levels of WT ␣␤ subunit failed to stimulate the transcriptional activity of either WT or mutant TFIIA (data not shown). Quantitation of the transcription signals was performed on a PhosphorImager 445 SI (Molecular Dynamics) and visualized by autoradiography on X-Omat-AR film (Kodak). 2 P. M. Lieberman, unpublished results.
GST Fusion Binding Assay-GST proteins were induced in 500-ml E. coli cultures, and soluble extracts were prepared by sonication in 10 ml of lysis buffer (20 mM Tris, pH 8.0, 5 mM EDTA Na 2ϩ , 100 mM NaCl, 0.05% Nonidet P-40 (Sigma), 5 mg of lysozyme (Sigma), 1 mM dithiothreitol, 1 mM phenylmethylsulfonyl fluoride, 10 g of leupeptin, 10 g of pepstatin A). GST extracts were clarified by centrifugation at 27,000 ϫ g for 30 min at 4°C and bound to glutathione-Sepharose 4B beads (Pharmacia) (0.5 ml) for 2 h at 4°C. GST proteins bound to the beads were washed 5 times in wash buffer (135 mM NaCl, 3 mM KCl, 10 mM Na 2 HPO 4 , pH 7.4, 2 mM KH 2 PO 4 , 1 mM dithiothreitol, 1 mM phenylmethylsulfonyl fluoride) and eluted with 10 mM glutathione, 50 mM Tris, pH 8.0, at 22°C with gentle agitation. Labeled WT or mutant ␥ protein (2 ϫ 10 4 cpm of 35 S) was expressed in a rabbit reticulocytecoupled in vitro transcription/translation system (Promega) in 20 l for 1 h at 30°C. Labeled ␥ proteins were used in GST binding reactions, which were performed as described (27). Samples were resolved on either 12 or 15% SDS-polyacrylamide gel electrophoresis gels. Quantitation of the signal was performed on a PhosphorImager 445 SI (Molecular Dynamics) and visualized by autoradiography on X-OMAT-AR film (Kodak).

Mutations of TFIIA-␥ Affect T-A Formation-Human
TFIIA-␥ has an open reading frame of 109 amino acids. Of these, 50 share absolute homology with yeast TFIIA-␥ (58% similarity including conservative substitutions) (Fig. 1A). Single substitution mutations of TFIIA-␥ were engineered by overlap extension PCR (53). Each TFIIA-␥ mutant has just a single absolutely conserved residue converted to encode an alanine residue, which is indicated by a star above the mutated residue in Fig. 1A. The human WT ␣␤ subunit and the mutant ␥ subunits were expressed in E. coli and purified by Ni-NTA agarose affinity chromatography under denaturing conditions. Subsequently, equal molar amounts of WT ␣␤ and a given ␥ mutant were renatured together. Soluble protein concentrations were determined by the Bradford assay and Coomassie staining of SDS-polyacrylamide gels. Purified and renatured TFIIA proteins were estimated to be approximately 70% pure. The concentration of soluble TFIIA-␥ subunits was further verified by Western blot analysis of linearly titrated preparations. Fig. 1B confirms that equal amounts of TFIIA-␥ subunits were present in the normalized panel of TFIIA mutants. The mutants listed in Fig. 1A had similar solubility, with the exception of ␥D41A and ␥K42A (data not shown). These two mutants were later revealed to fail to interact with ␣␤ (see Fig.  2B), consistent with the observation that ␥ does not solubilize in the absence of the ␣␤ subunit (data not shown, Ref. 49).
The TFIIA-␥ mutants were assayed for their ability to stimulate human recombinant TBP binding to a TATA box containing oligonucleotide in polyacrylamide gel assay EMSA ( Fig.  2A). In the absence of TFIIA, TBP binds weakly to the TATA box probe (T complex; Fig. 2A, lane 1). The addition of WT rTFIIA to TBP resulted in the formation of a stable TBP⅐TFIIA DNA complex (T-A complex; Fig. 2A, lane 2). With the E1B TATA oligonucleotide probe, we found that human T-A complex had a faster migration than the T complex, suggesting that TFIIA may alter the conformation of the TBP-DNA complex (33). The addition of the various TFIIA-␥ mutants to TBP produced varying amounts of T-A complex formation. The TFIIA-␥Y65A mutant did not form the T-A complex at all, while the ␥Y3A, ␥D20A, ␥T64A, ␥W72A, and ␥D91A mutants formed weak T-A complexes (less than 35% of WT) ( Fig. 2A). The ␥W72A mutant was unique in its ability to form an altered mobility T-A complex in polyacrylamide-EMSA gels, which was only evident on longer exposures (see below, Fig. 5A).
Mutations of the ␥ Subunit Affect Protein-Protein Interactions-GST fusion proteins have been previously used to demonstrate that the ␥ subunit homodimerized, heterodimerized with the ␣␤ subunit, and bound Zta in an activation domainspecific manner (27). It was also reported that the ␣␤ subunit interaction with ␥ was localized to the ␣ subunit (28), which was confirmed in this study using the GST assay (Fig. 2B, top  panel). 35 S-labeled ␥ point mutants were assayed for their ability to bind to GST, GST-␥, or GST-␣ using similar conditions. The summary of several independent binding assays is presented at the bottom of Fig. 2B. Interactions that were significantly reduced (at least 25% of WT for ␣, and at least 50% of WT for ␥) are indicated by a minus symbol (Fig. 2B). Interestingly, a pattern formed from these GST experiments, indicating that certain conserved domains of ␥ are critical for specific protein interactions. Interactions with GST-␣ were most sensitive to mutations in ␥ residues Asp 41 and Lys 42 , which bound at less than 10% of WT levels. The failure of ␥D41A and ␥K42A to bind ␣ probably explains the poor solubility of these mutants upon renaturation with recombinant ␣␤. Homotypic interactions with GST-␥ were most severely affected by mutations of a pocket of hydrophobic amino acids at positions Tyr 65 , Phe 67 , Cys 68 , Trp 72 , Thr 73 , and Phe 74 (Fig. 2B). Interestingly, we did not find a single point mutation in ␥ that significantly reduced Zta binding (data not shown). Thus, Zta may interact with multiple domains of ␥.
We have previously demonstrated a direct protein-protein interaction between 35 S-labeled ␥ and a GST-TBP fusion protein (27) (Fig. 2C, right panel). Since the TFIIA-␥D20A, ␥T64A, ␥Y65A, ␥W72A, and ␥D91A mutants all formed weak T-A complexes, these 35 S-labeled ␥ mutants were tested for their ability to bind directly to GST-TBP. Both the ␥D20A and ␥Y65A mutants were reduced in GST-TBP binding, 61 and 42% of WT levels, respectively (Fig. 2C, left panel). However, the ␥T64A, Equal amounts of purified human rTFIIA-␥ mutant proteins were solubilized in D100 buffer and resolved on 15% SDS-polyacrylamide gel electrophoresis. The immunoblot was probed with rabbit Abs raised against human rTFIIA-␥. ␥Y3A was expressed in pQE-9, which has a truncated amino-terminal tag. The amino-terminal tags did not affect WT TFIIA function in any assays tested.
TFIIA Mutants Show Activator-specific Defects-TFIIA-de-pleted HeLa cell extracts require rTFIIA to mediate transcriptional activation for all activation domains tested (27). The TFIIA-␥ mutants were assayed for the ability to mediate transcription stimulated by several distinct types of activation domains using TFIIA-depleted HeLa extracts in an in vitro transcription assay. The GAL4 DNA binding domain was fused to distinct classes of activation domains: the synthetic acidic-rich activation domain (GAL4-AH), the herpes simplex virus-encoded VP16 activation domain (GAL4-VP16), or the prolinerich activation domain of the CCAAT-binding factor (GAL4-CTF). These GAL4-activator fusions were examined for their ability to stimulate transcription from the G 5 E1BTCAT promoter (five GAL4 binding sites upstream of the adenovirus E1B TATA element). The Epstein-Barr virus-encoded Zta activator was examined for its ability to stimulate transcription from the Z 7 E4TCAT promoter (seven ZRE binding sites upstream of the adenovirus E4 TATA element). Endogenous levels of AP-1-dependent transcription activation were tested by examining the activity of the WT human collagenase promoter. The G 5 E1BTCAT, Z 7 E4TCAT, and collagenase basal promoters have barely detectable basal levels of transcription in our in vitro transcription system in the absence of added activator or activator binding site (27). For these activators, transcription was reduced to less than 5% in extracts depleted of endogenous TFIIA (27). Activated transcription could be restored to undepleted levels by the addition of 0.2 M WT rTFIIA (27). To determine if the TFIIA-␥-containing mutants (WT ␣␤ ϩ ␥ mutant) coactivated transcription maximally at 0.2 M as did WT rTFIIA in our depleted system, TFIIA-␥ mutants were titrated using the Zta and GAL4-AH activators. We found that, for the panel of TFIIA-␥ mutants, 0.2 M also yielded maximal transcriptional activity (data not shown). For each activator tested in Fig. 3, the addition of 0.2 M WT rTFIIA to TFIIA-depleted extracts was considered a 100% level of transcription, while TFIIA-depleted extracts were considered to be a 0% level of transcription. The levels of activated transcription obtained for a particular TFIIA-␥ mutant and activation domain were plotted as percentage of the WT TFIIA signal (% WT) (Fig. 3).
Only one point mutation of TFIIA-␥ was completely defective for all activators tested. ␥Y65A, which failed to form the T-A complex ( Fig. 2A) and had reduced binding to GST-TBP (Fig.  2C), did not mediate transcription for any activator tested, suggesting that T-A formation is required for transcription activation (Fig. 3A). However, the ␥Y6A and ␥F67A mutants formed T-A complexes at near WT levels but mediated transcription at just 25% or lower values relative to WT for all activators (Fig. 3B). This indicates that T-A formation alone is not sufficient for activator function. Most significantly, the ␥W72A mutant supported activation only for Zta, mediating about 18-fold higher levels of Zta activity compared with GAL4-AH (Fig. 3C). The ␥Y3A and ␥C68A mutants had similar transcription phenotypes to the ␥W72A mutant, although not as dramatic, with at least 2-fold greater activity for Zta compared with all the other activators tested (Fig. 3C). In contrast, the ␥D91A and ␥E109A mutants revealed an opposite phenotype, mediating about 2-fold and 3-4-fold, respectively, higher levels of transcription activation for all the activators compared with Zta (Fig. 3, A and C). These results show that mutations in TFIIA-␥ can affect some activators more dramatically than others, implying that different activators utilize TFIIA in distinct activation pathways.
Zta Overcomes TFIIA-␥ Mutations That Disable Other Activators-TFIIA-␥Y65A and ␥W72A were defective for T-A formation and reduced for transcriptional activation by all activators. While ␥Y65A was completely defective for all activators, ␥W72A mediated just Zta activation but only to 30% of WT. To further investigate the mechanistic basis of these defects, two additional TFIIA-␥ mutants were engineered to more conservative phenylalanine substitutions at the ␥Y65 and ␥W72 residues. The ␥Y65F and ␥W72F mutants were purified and solubilized as described for the other mutants. The ␥Y65F and ␥W72F mutants solubilized as well as WT TFIIA (Fig. 4A) and stimulated formation of the T-A complex as well as WT TFIIA (Fig. 4B). However, when the ␥Y65F and ␥W72F mutants were assayed in transcription using TFIIA-depleted extracts, only Zta was significantly coactivated by both of these mutants (Fig.  4C). The ␥Y65F mutant coactivated Zta at 70% of WT levels, 6.5-fold better than GAL4-AH and 14-fold better than GAL4-CTF (Fig. 4C). Additionally, the ␥W72F mutant only coactivated Zta and not the other activation domains, at levels similar to ␥W72A (Fig. 4C). Since Zta is an Epstein-Barr virus-encoded viral activator, we decided to test whether another commonly studied viral activation domain, herpes simplex VP16, would show a similar phenotype to Zta with these mutants. Interestingly, like Zta, GAL4-VP16 can also form a stable magnesium-agarose EMSA complex with TFIID and TFIIA (60). However, GAL4-VP16 failed to stimulate transcription with any of the ␥Y65 and ␥W72 mutants, thus behaving markedly different from Zta (Fig. 4C). GAL4-VP16 did stimulate transcription with ␥Y6A and ␥F74A to levels similar to that observed with GAL4-AH (data not shown). These results demonstrate that the transcriptional defect of the ␥Y65F, ␥W72A, and ␥W72F mutants is not a result of the failure to form the T-A complex and further support our earlier observation that Zta can uniquely overcome TFIIA deficiencies.
Z-D-A Formation Compensates for the ␥W72A Defect-To investigate the mechanistic basis for the Zta-specific activation T-A complex formation versus the extent of GAL4-AH-activated transcription mediated by the TFIIA-␥ mutants. T-A data are the average values from two independent experiments, and GAL4-AH-activated transcription values are derived from Fig. 3A; both data are presented side by side. C, the same data as in A but graphed to compare the transcription levels of a subset of TFIIA-␥ mutants with different activators. of ␥W72A, we further characterized some biochemical properties of the ␥W72A mutant. In acrylamide gel EMSA, we observed that T-A formation with ␥W72A was reduced to less than 30% of WT levels. A longer autoradiographic exposure showed that ␥W72A produced an additional slower mobility complex, suggesting that it can form an alternative structure with TBP and DNA (Fig. 5A). The structural basis of this altered T-A complex is unclear. While ␥W72F and ␥Y65F formed T-A complexes indistinguishable from WT TFIIA, they had similar transcription-defective phenotypes. Taken together, these results suggest that the transcriptional defect of these TFIIA mutants may result from an aberrant T-A complex and not simply a reduction in the amount of T-A complex that can form.
The ability of Zta to uniquely activate transcription with ␥W72A could be explained if Zta were able to overcome the defect of ␥W72A binding to TBP-DNA. Zta has been shown to stimulate the formation of a stable TFIIA⅐TFIID promoter complex (Z-D-A) (32). We have previously reported that under limiting conditions TFIID does not form a stable magnesiumagarose EMSA complex by itself (Fig. 5B, lane 3) or in the presence of either Zta (lane 4) or TFIIA (lane 5) alone (27,32). To examine if Zta and the TAFs in the TFIID complex are capable of compensating for the ␥W72A mutation, we measured D-A and Z-D-A complex formation in magnesium-agarose EMSA (Fig. 5B). We compared D-A and Z-D-A formation with the ␥W72A mutant and WT TFIIA using higher temperature, longer incubation times, increased amounts of TFIIA protein, or a combination of the above. Under limiting conditions of 22°C and 15-min incubations, ␥W72A forms a much weaker Z-D-A complex (34%) than does WT TFIIA (Fig. 5B, top panel,  compare lanes 8 and 6 or lanes 12 and 10). Quantitation indicated that TFIIA stimulated TFIID binding 2-fold (compare lanes 5 and 9 to lane 3), while ␥W72A did not stimulate TFIID binding (Fig. 5B, top panel; compare lanes 7 and 11 to lane 3). By increasing both the time and temperature of the incubation reactions, the extent of Z-D-A complex formed with ␥W72A increased to levels indistinguishable from WT TFIIA (Fig. 5B, lower panel, compare lanes 10 and 12). However, these conditions failed to improve the binding of ␥W72A to TFIID in the absence of Zta (Fig. 5B, lower panel, compare lanes 11 and 3), while WT TFIIA was improved over 4-fold in D-A formation (lower panel, compare lanes 9 and 3). These results suggest that ␥W72A fails to stimulate TFIID-DNA binding and that Zta can compensate for this defect. DISCUSSION Previous reports identified TFIIA as a critical component in the rate-limiting steps of activated transcription (30 -32, 61). To better elucidate the function of TFIIA in this process, we sought to isolate TFIIA mutants that distinguish the transcription function of several different activation domains. We hypothesized that different TFIIA surfaces would be required to mediate the activation of different activators. Since a previous report determined that nearly all in frame deletions of yeast TFIIA-␥ were nonviable (48), it appeared that subtle alterations of TFIIA would be more likely to yield activator-specific transcriptional defects. To avoid gross alterations of TFIIA structure, alanines were substituted in single conserved residues of TFIIA-␥ in the first panel of mutants. Analysis of these TFIIA mutant proteins in DNA binding and in vitro transcription assays revealed that TFIIA interacts with different activators in distinct ways and mediates at least two mechanistically distinct activator functions.
TFIIA Mutants Reveal Differences Among Transcriptional Activator Mechanisms-Our series of TFIIA mutations reveal differences in the requirements for TFIIA by distinct transcriptional activation domains. TFIIA-␥ mutants Y3A, C68A, W72A, and F74A were all significantly reduced in mediating transcription for all activators compared with Zta (Fig. 3C). In contrast, E109A was reduced for Zta relative to all other activators (Fig. 3C). These results suggest that Zta interacts with TFIIA-␥ in a manner distinct from the other activators tested. Zta has been shown to stimulate the TFIIA⅐TFIID promoter complex, while other activators GAL4-AH and AP-1 (Jun/Fos) fail to stimulate this complex. 2 GAL4-VP16 does stimulate D-A formation, but to a lesser extent than Zta (60). In contrast to Zta, GAL4-VP16 did not activate transcription with ␥W72A, ␥W72F, or ␥Y65F (Fig. 4C), suggesting that Zta and VP16 interact with TFIIA in mechanistically distinct ways. Zta binds directly to both ␥ (27) and ␣␤ (52), which may explain how Zta uniquely compensates for the ␥W72A transcriptional defect.
Transcriptional analysis of the ␥Y6A and ␥F74A mutants also suggests that the activation mechanism of GAL4-CTF may be distinct from the other activators tested. While ␥Y6A and ␥F74A were reduced for most activators relative to WT (the exception being that ␥F74A mediates Zta activation), the effects on GAL4-CTF activity were the most severe (Fig. 3C). ␥Y6A stimulates T-A formation at close to WT levels but is substantially reduced for the formation of Z-D-A complex (data not shown). One likely explanation for these observations is that ␥Y6A, and possibly ␥F74A, fail to interact with a subset of TAFs important for D-A formation and that these TAFs are specifically important for GAL4-CTF-mediated transcription.
Other reports have found mutations of yeast TBP that specifically disrupt activated but not basal transcription in vitro (62). At least one of these mutations was likely to interfere with TFIIA binding. A more recent transfection analysis of human TBP mutations revealed that different activators had different sensitivity to TBP mutations, suggesting they interact with TBP in distinct ways (41). While several activation domains responded differently to various TBP mutants, most of the transcriptional defects could be correlated with the loss of binding to TAF II 250 (41). Similarly, mutagenesis of TFIIB identified a region of TFIIB that disrupted activated transcription for two activators but not basal transcription (63). In contrast, a mutagenesis analysis of the large subunit of TFIIE did not distinguish basal from activated transcription nor differences between two types of activation domains (64). Thus, the extent to which mutations of general transcription factors affect different activators may reflect important differences in activator mechanisms and general transcription factor functions.
TFIIA Functions in Two Mechanistically Distinct Activation Steps-TFIIA stimulates the binding of TBP to DNA, and this has been considered a primary function of TFIIA in transcription activation. However, we have identified mutations of TFIIA-␥ that uncouple the stimulation of TBP-DNA binding from the ability to mediate activator function for several distinct activators (Figs. 2 and 3). The ␥Y6A, ␥F67A, ␥C68A, and ␥W72A mutants all stimulate T-A complex formation significantly better than they mediate transcription activity for GAL4-AH and AP-1, relative to WT TFIIA (Figs. 2 and 3). The lack of correlation between T-A formation and transcriptional activation was most remarkable for the more conservative substitution mutants ␥Y65F and ␥W72F (Fig. 4). Both of these mutants stimulate T-A formation as well as WT but failed to support transcriptional activation by GAL4-AH, GAL4-VP16, and GAL4-CTF (Fig. 4C). The failure of these mutants to mediate activated transcription indicates that TFIIA contributes activation functions distinct from the stimulation of TBP-DNA binding.
Several additional activities have been ascribed to TFIIA besides stimulation of TBP-DNA binding. TFIIA induces a conformational change in TBP (33), and it is conceivable that ␥Y65F and ␥W72F fail to induce a TBP conformational change necessary for transcriptional activation by GAL4-AH, CTF, and VP16. The Drosophila ␣␤ subunit has been shown to bind to dTAF II 110 (47), and mutations in TFIIA-␥ may disrupt these interactions, which are critical for transcriptional activation. The requirement of TFIIA and TAFs in promoter selectivity also supports the model that TFIIA functionally interacts with TAFs (65). Alternatively, coactivators, like PC4 and HMG2, have been shown to interact with D-A complex formation (61,66), and TFIIA-␥W72F and ␥Y65F may fail to interact with these coactivators in the preinitiation complex. TFIIA can also disrupt TBP-specific repressors, like DR1 (36), and although ␥W72F and ␥Y65F form the T-A complex, they may fail to disrupt specific repressor-TBP interactions. TFIIA copurifies with a repressor activity specific for TBP and consensus TATA elements, and our mutations may affect the specificity of this repressor activity (67). Additionally, TFIIA makes direct contact with at least two transcriptional activators (27,52), and it is possible that some of these TFIIA mutations have lost the ability to directly contact specific activators or coactivators necessary for transcription function. While we have not determined which of these possible TFIIA interactions have been disrupted by these mutations, our data strongly suggest that TFIIA interactions subsequent to T-A formation are essential for transcription activation.
The ␥ Subunit Interacts with the ␣ and ␥ Subunits in Distinct Domains-Analysis of GST interaction assays, DNA binding studies and in vitro transcription reactions allow different features of TFIIA to be assigned to distinct subdomains of ␥. Based upon the GST experiments in Fig. 2, the ␥ subunit Phe 40 , Asp 41 , and Lys 42 conserved pocket was shown to be absolutely critical for ␣ subunit interactions. These same residues were important for solubility of ␥, indicating that solubility is largely dependent on the formation of ␣-␥ heterodimers. The ␥ residues Tyr 65 , Phe 67 , Cys 68 , Trp 72 , Thr 73 , and Phe 74 appear to be important for homotypic interactions. While the significance of ␥-␥ interactions are not clear, we (data not shown) and others (43) have found that TFIIA has a native molecular weight consistent with TFIIA being a dimer of both subunits. Thus, ␥-␥ interactions are likely to be important for the oligomerization state of TFIIA. The same amino acid residues important for ␥-␥ interactions are also important for transcriptional activation function and Z-D-A formation. We have analyzed TFIIA-␥Y65F and ␥W72A by gel filtration and found that their molecular mass was indistinguishable from that of WT TFIIA (data not shown). Thus, loss of oligomerization cannot account for the transcriptional defects of TFIIA-␥Y65F and ␥W72A. Nevertheless, we speculate that ␥-␥ interactions are critical for the TFIIA conformation that is necessary for activation functions.
A general model has emerged that suggests that TBP binding to DNA is a rate-limiting step affected by several classes of transcriptional activators (19 -22). TFIIA can stimulate the binding of TBP to DNA, and activators that stimulate TFIIA binding are predicted to enhance transcription (24,32,51,60). The analysis of TFIIA-␥ mutations presented in this study suggests that TFIIA not only enhances TBP-DNA binding but qualitatively changes the preinitiation complex. Our data suggest that TFIIA affects the recruitment of TAFs and/or coactivators into a transcriptionally active conformation. Our analysis also indicates that activators function by distinct mechanisms and that TFIIA plays a central role in distinguishing the mechanism of different activators.