Direct TFIIA-TFIID Protein Contacts Drive Budding Yeast Ribosomal Protein Gene Transcription*

Background: Enhancer-bound transactivators communicate with promoter-bound transcription machinery through specific interactions. Results: We identified and dissected the interacting surfaces of TFIIA, TFIID, and the transactivator Rap1. Conclusion: Rap1 and TFIIA make specific contacts with the Taf4 subunit of TFIID, leading to Rap1-controlled transcription of ribosomal protein coding genes. Significance: Knowledge of the mechanisms of mRNA gene transcription activation is crucial for understanding gene control. We have previously shown that yeast TFIID provides coactivator function on the promoters of ribosomal protein-encoding genes (RPGs) by making direct contact with the transactivator repressor activator protein 1 (Rap1). Further, our structural studies of assemblies generated with purified Rap1, TFIID, and TFIIA on RPG enhancer-promoter DNA indicate that Rap1-TFIID interaction induces dramatic conformational rearrangements of enhancer-promoter DNA and TFIID-bound TFIIA. These data indicate a previously unknown yet critical role for yeast TFIIA in the integration of activator-TFIID contacts with promoter conformation and downstream preinitiation complex formation and/or function. Here we describe the use of systematic mutagenesis to define how specific TFIIA contacts contribute to these processes. We have verified that TFIIA is required for RPG transcription in vivo and in vitro, consistent with the existence of a critical Rap1-TFIIA-TFIID interaction network. We also identified essential points of contact for TFIIA and Rap1 within the Rap1 binding domain of the Taf4 subunit of TFIID. These data suggest a mechanism for how interactions between TFIID, TFIIA, and Rap1 contribute to the high rate of transcription initiation seen on RPGs in vivo.

The eukaryotic general transcription factors TFIIA, -B, -D, -E, -F, and -H contribute importantly to the process of accurate initiation of mRNA synthesis by RNA polymerase II (1,2). However, depending on in vitro transcription assay conditions, there is not always a strict requirement for all six general transcription factors (3), and consistent with this observation, genome-wide approaches have documented variable contributions of general transcription factors, particularly TFIID, to gene-specific transcriptional programs (4 -8). Interestingly, the canonical TFIID complex can be altered, diminished, or abolished during the process of cellular differentiation (9 -11). Collectively, these observations point to a context-specific role for TFIID and are also consistent with models suggesting a coregulatory role for TFIID, where this multisubunit complex directly engages activator proteins to drive dedicated transcriptional programs (12)(13)(14). Consequently, the exact relationships between gene promoter architecture, chromatin modifications and conformation, activator-TFIID interactions, dependence on TFIID function, and mechanisms of TFIID-containing PIC formation and function remain elusive.
Consistent with the overall similarities of TFIID subunit composition, sequence conservation, and function, electron microscopy (EM) of yeast and human complexes reveal a simi-lar trilobed architecture (38 -40). Biochemistry, genetics (15), and EM difference mapping with subunit-specific antibodies have been used to determine the stoichiometry and location of Tafs within budding yeast TFIID. These data have greatly facilitated the structural characterization of promoter-, activator-, and TFIIA-TFIID interactions and conformational rearrangements of complexes containing these molecules (40 -43). Recent high resolution mapping of a 7-Taf recombinant human TFIID subcomplex indicates similar patterns of Taf location with the TFIID holocomplex (44). Collectively, these structural studies allow for the connection of biochemical data with molecular genetic characterization of conserved Taf/TFIID protein-protein interactions (e.g. the binding of activators with TFIID and TFIIA). Such experiments can be accomplished with ease and high precision in the yeast system due to the many powerful molecular genetic tools available in Saccharomyces cerevisiae.
Further, we have recently expanded the use of RPGs to investigate TFIID function through a series of EM studies that defined the structure of several minimal pre-PIC assemblies containing TFIID, TFIIA, Rap1, and RPG enhancer-promoter DNA (42). Intriguingly, we found that the intervening DNA between Rap1 bound to the UAS RAP1 enhancer site and TFIID bound to the promoter forms an extended loop and, concomitantly, a conformational change in TFIIA within the quarternary Rap1-TFIID-TFIIA-DNA complex. The resulting DNA "locked" features of this complex probably contribute critically to the high levels of RPG transcription seen in vivo. Here we report the extension of these analyses, where we examine the genetic and biochemical relationships between Rap1, TFIIA, and TFIID. In this study, we have systematically probed the TFIIA and Taf4 amino acid residues required for RPG transcription in order to identify the surface(s) that participate in protein-protein interactions between TFIIA, Rap1, and TFIID while simultaneously generating key reagents for further dissection of these regulatory protein-protein and protein-DNA interactions.
We have found that TFIIA makes critical contributions to RPG transcription through a variety of contact points in the N-terminal domain of the Toa2 subunit. We identified direct TFIIA-TFIID interactions through the RBD of Taf4, the domain that we had previously characterized as a conserved, high affinity Rap1 binding site (24). Finally, we defined taf4 RBD mutants that compromise growth, TFIID-TFIIA binding, Rap1-TFIID binding, and RPG transcription in vivo and in vitro. These phenotypes can be partially alleviated by overexpression of TFIIA. Collectively, our data indicate that Taf4 is vital for TFIID-Rap1 and TFIID-TFIIA interactions. Hence, this TFIID-specific subunit represents a critical functional nexus for RPG transcription regulation and, consequently, cellular proliferation.
PCR-mediated site-directed mutagenesis was performed using the gene splicing by overlap extension (SOEing) method (53) with recombinant Pfu DNA polymerase and TOA1 or TOA2 templates subcloned as BamHI/XhoI fragments into the pBG101 bacterial T7 expression vector (see the Vanderbilt University Center for Structural Biology Web site), including sequence for an N-terminal His 6 -GST tag. An oligonucleotide corresponding to the GST coding sequence was used as a universal 5Ј mutagenesis primer, and the T7 terminator reverse primer was used as a universal 3Ј primer. Final mutated PCR products were reintroduced as BamHI/XhoI fragments into p413HA 1 3C TOA1 or p415 HA 3 NLS TOA2 that lacked the TOA1 or TOA2 coding sequence.
Plasmid Shuffle Assays-Shuffling strains were transformed with the appropriate covering plasmids encoding WT or variant proteins. Transformants were plated on SC selective media, either SCϪHis (TOA1 and TAF4 genetic/growth assays), SCϪLeu (TOA2 assays), or SCϪHisϪLeu (toa2-and taf4-TFIIA high copy suppression genetic/growth assays). Single colonies were picked to a liquid culture of the appropriate selective medium and incubated at room temperature for 7-10 days to reach stationary phase. A portion of each culture was transferred to one well of a 96-well Bacti-plate (NUNC) and diluted serially 1:4. A 96-well pinning tool was used to transfer cells from each well to plates containing appropriate selective medium supplemented with 5-FOA. Replicates were plated and incubated at room temperature (20 -23º C) or 30, 34, or 37°C. Plates were photographed after 3 days of incubation except for the room temperature plate, which was photographed after 7 days. Single colonies from the room temperature plate were inoculated into yeast extract peptone dextrose liquid cultures and used as a source of preinoculum for temperature shift, immunoblots, and immunoprecipitations.
Antibodies-Anti-Taf and anti-Rap1 rabbit polyclonal antibodies were described previously and were used as total IgG prepared by ammonium sulfate precipitation of antiserum and subsequent Protein A affinity chromatography (35,57). Secondary detection was with goat anti-rabbit Fc-horseradish peroxidase (HRP) conjugate (Pierce). Anti-HA used for immunoblotting was the 3F10-HRP conjugate (Roche Applied Science). Anti-HA used for immunoprecipitation was Protein A-purified and originated from the 12CA5 hybridoma (Vanderbilt Antibody/Protein Resource). Anti-FLAG for immunoblotting was the M2-HRP conjugate (Sigma), whereas standard M2 was used for immunoprecipitation. Polyclonal anti-TFIIA antisera were generated (Bethyl Laboratories) by immunization of rabbits with purified TFIIA complex renatured from recombinant denatured subunits. IgG was prepared as above. Immunoprecipitation and immunoblotting were performed as described previously (57).
Temperature Shift, RNA Isolation, Primer Extension, and RNA Sequence (RNA-Seq) Analyses-Paired yeast extract peptone dextrose cultures of each strain were grown at room temperature to mid-log phase, and one was subjected to temperature shift at 37°C (24,35,58). After 2 h of additional incubation, cultures were harvested by filtration, and RNA was isolated immediately by extraction of cells on the filters with hot acidic phenol, essentially as described (59,60). Transcript abundance measurement by multiplex 5Ј primer extension was performed as described previously (24). RNA samples used for deep sequencing were treated with DNase I to remove residual genomic DNA. The quality and quantity of each RNA preparation were determined by measuring A 260 /A 280 on a Nanodrop ND-1000 spectrophotometer and then validated with the Qubit fluorometer RNA assay. RNA integrity was determined by capillary electrophoresis using an RNA Nano chip on an Agilent BioAnalyzer. Sequencing libraries were generated using the Illumina TruSeq RNA Sample Prep kit with modifications to the manufacturer's protocol. Briefly, poly(A) ϩ mRNA was isolated from 500 ng of total RNA on oligo(dT) magnetic beads. Purified RNA was fragmented, annealed with random hexamer primers, and applied as template for first strand cDNA synthesis by SuperScript II reverse transcriptase (Invitrogen). Residual dNTPs were removed using the RNA Clean and Concentrator-5 kit (Zymo Research) and eluted in 20 l of water. Second strand cDNA was synthesized using the NEBNext mRNA second strand synthesis module (New England Biolabs) using the dNTP-free reaction buffer and a dTTP-free dNTP mix containing dUTP (U.S. Biochemical). Double-stranded cDNA was purified with AMPure XP beads and applied to the standard Illumina library protocol including end repair, dA tailing, and ligation with Illumina sequencing adapters. Prior to PCR, dsDNA libraries were treated with USER enzyme to introduce single strand breaks (New England Biolabs). Following PCR enrichment, libraries underwent fragment size confirmation using an HS DNA chip on an Agilent BioAnalyzer and were quantified using the Illumina library quantification kit (KAPA). Library pools underwent cluster generation on an Illumina cBot HiSeq 2000 instrument at the Vanderbilt Technologies for Advanced Genomics Core Facility (VANTAGE, Nashville, TN). Raw sequencing reads were processed through CASAVA-1.8.2 for FASTQ conversion and demultiplexing. The CASAVA quality filter was used, and pass filter reads were provided. Transcript-specific reads/kilobase/million reads (RKPM) values were defined using the QSeq routine of ArrayStar3 software.
Recombinant Protein Expression and Purification-A TOA1 fragment was digested with BamHI, blunted with the Klenow fragment of Escherichia coli DNA polymerase I, and then cut with XhoI and inserted into pBG100 cut with NdeI, blunted with Klenow, and cut with XhoI. The resultant vector includes the T7 promoter and ribosome binding site element immediately upstream of TOA1 without any additional encoded N-or C-terminal amino acids. TOA2 bacterial expression vectors included BamHI/XhoI fragments from TOA2 or toa2 yeast vec-tors inserted into the equivalent sites of pBG100, resulting in encoded Toa2 with an N-terminal His 6 tag. Expression of Toa1, Toa2, and Toa2 variants was performed in parallel with room temperature cultures using the Rosetta E. coli DE3 strain (EMD Biosciences) and an autoinduction medium system (24,61). Induced cell pellets from 1-liter cultures were harvested by centrifugation and immediately resuspended in buffer containing 30 mM Tris, pH 8.0, 45 mM NaCl, 5 mM NH 4 Cl, 1 mM EDTA, 10% glycerol, and 250 g/ml egg white lysozyme. After 30 min at room temperature, cell suspensions were rapidly frozen at Ϫ80°C. Material was thawed, and Triton X-100 was added to 1% final concentration. Lysate viscosity was reduced by sonication with a model 250 Branson sonicator equipped with the large tip. Insoluble material was pelleted by centrifugation, supernatant was discarded, and the pellet was resuspended in buffer containing 30 mM Tris, pH 8.0, 450 mM NaCl, 50 mM NH 4 Cl, 1 mM EDTA, 1% Triton X-100, and 10% glycerol. Insoluble material was again pelleted, and this wash step was repeated twice more. Two more wash steps were performed with buffer containing 30 mM Tris, pH 8.0, 450 mM NaCl, 50 mM NH 4 Cl, 1 mM EDTA, and 10% glycerol, followed by two final wash steps with 30 mM Tris, pH 8.0, 450 mM NaCl, 50 mM NH 4 Cl, 1 mM EDTA, 5 M urea, and 10% glycerol. Inclusion body proteins were solubilized in 30 mM Tris, pH 8.0, 450 mM NaCl, 50 mM NH 4 Cl, 1 mM EDTA, 7 M urea, 1 mM DTT, and 10% glycerol. Denatured Toa1 was mixed with Toa2 or Toa2 variants at a 2:1 Toa1/Toa2 molar ratio and diluted with solubilization buffer to 45 ml at 0.25 mg/ml final protein concentrations and then dialyzed over 24 h against 4 liters of 25 mM HEPES, pH 7.6, 450 mM NaCl, 50 mM NH 4 Cl, 20 mM imidazole, and 10% glycerol. Renaturation of all TFIIA forms was performed in parallel. Dialyzed material was centrifuged to remove precipitated protein, and cleared supernatant was applied to nickel-nitrilotriacetic acid columns equilibrated and washed with 25 mM HEPES, pH 7.6, 100 mM sodium acetate, 20 mM imidazole, 10% glycerol and eluted with the same buffer containing 200 mM imidazole. TFIIA variants were concentrated using centrifugal filtration devices (Amicon; 10,000 molecular weight cut-off) and subjected to size exclusion chromatography on a calibrated 3 ϫ 30-cm Superose 12 column in 25 mM HEPES, pH 7.6, 300 mM sodium acetate, and 10% glycerol. Peak Toa1/Toa2-containing fractions corresponding to the expected molecular mass of TFIIA (ϳ48 kDa) were combined and stored at Ϫ80°C.
BamHI/SalI fragments from p413 HA 3 NLS TAF4 or taf4 were transferred to pLM302, resulting in T7 expression vectors expressing Taf4 with N-terminal His 6 -maltose-binding protein (MBP) tags. pLM302 TAF4 vectors were mixed with pACYC184 11b TAF12 (62) and transformed into E. coli BL21 DE3 cells containing the pSC101 RIL vector (additional arginine, isoleucine, leucine tRNA genes, compatible pSC101 origin, streptomycin resistance gene). 3 Coexpression of His 6 -MBP-Taf4 with Taf12 occurred in 250-ml cultures grown in autoinduction medium at room temperature. Cells were harvested and resuspended in buffer containing 25 mM HEPES, pH 7.6, 250 sodium acetate, 20 mM imidazole, 0.1% Triton X-100, and 10% glycerol along with a panel of protease inhibitors. Resuspended cells were lysed by sonication, and insoluble material was removed by centrifugation. The soluble supernatant was subjected to nickel-nitrilotriacetic acid chromatography, washed with the lysis buffer, and eluted with the lysis buffer containing 200 mM imidazole. Eluted material was applied to SP-Sepharose columns (GE Healthcare); washed with 25 mM HEPES, pH 7.6, 250 mM sodium acetate, 0.1% Triton X-100, 10% glycerol; and eluted with the same buffer containing 1 M sodium acetate. Peak MBP-Taf4/Taf12-containing fractions were combined and stored at Ϫ80°C. His 6 -Rap1 from pET28A RAP1 or His 6 -GST-Rap1 expressed from pBG28A RAP1 was purified essentially as described above for Taf4/ Taf12 heterodimers. TBP was prepared according to established procedures (63,64).
In Vitro Transcription Assays-Whole cell extract (WCE; final protein concentration 35 mg/ml) in 20 mM HEPES, pH 7.9, 50 mM potassium acetate, 10 mM MgSO 4 , 10 mM EGTA, 20% glycerol, and 5 mM DTT) was prepared, antibody-depleted of Rap1 and TFIIA, and assayed by G-free template transcription as described previously (15,35) with a few modifications to optimize signals; 30-l assays involved a 10-min, 20º C preincubation of WCE (525 g) and DNAs in the absence of NTPs to allow PIC formation and 30 min of transcription at 20º C in the presence of 400 M ATP, 400 M CTP, 2 M cold UTP, 100 M 3Ј-O-methyl-GTP, and 5 Ci of [␣-32 P]UTP (ϳ800 Ci/mmol). Final solution conditions were 30 mM HEPES, pH 7.9, 75 mM potassium acetate, 12 mM magnesium acetate, 0.75% polyethylene glycol, 13.3% glycerol, 2.5 mM DTT, 5 ng/l creatine kinase, 25 M phosphocreatine, 20 units of RNAsin (Promega), 150 units of ribonuclease T1, 500 ng of UAS RAP1 -G-free template, 18.75 ng of UAS GAL -G-free template, and 50 ng of purified recombinant Gal4-VP16. The content of Rap1 and TFIIA in WCE, as determined by quantitative immunoblotting using purified recombinant Rap1 and TFIIA as standards was 45 and 1.5 ng/l, respectively. Where add-back experiments were per-formed, the amounts of purified recombinant WT and mutant TFIIA and WT Rap1 added to transcription assays were 1ϫ for Rap1 and 10ϫ for TFIIA, relative to the amount of these proteins present in the original control WCE; these amounts were empirically determined to be the minimal amount of protein that generated the maximal signals. The Rap1-driven transcription template contained two copies of the Rap1 enhancer upstream of the PGK1 core promoter (35,42,65) fused to the 378 bp G-free cassette derived from pML⌬53 (66), whereas the Gal4-driven template contained two copies of UAS GAL upstream of the CYC1 promoter fused to the 278-bp G-free cassette of pJJ470 (67). Transcription was stopped by the addition of 270 l of 150 mM sodium acetate, pH 5.5, 0.1% SDS, and 25 g of E. coli tRNA. RNA was purified by sequential phenol/ CHCl 3 and CHCl 3 extraction and ethanol precipitation. RNA samples were vacuum-dried and resuspended in 20 l of 90% (v/v) deionized formamide, 50 mM EDTA, pH 7.9, 0.1% (w/v) SDS, 0.05% xylene cyanol, and 0.05% bromphenol blue. Dissolved samples were heated for 2 min at 100º C and held at this temperature until loaded on denaturing 6% polyacrylamide gels (prepared from a 30:0.8% acrylamide/bisacrylamide stock) containing 7 M urea and buffered with 1ϫ TBE; gels were run at a constant power sufficient to maintain gel temperatures of 50 -55ºC. Gels were soaked in H 2 O for 10 min on a gyro shaker to remove urea and vacuum-dried, and 32 P-labeled RNAs were visualized by exposure to an Eastman Kodak Co. K-screen and imaged with a Bio-Rad PharosFX imager. Data were quantified using QuantityOne software.
In Vitro Pull-down Assays-GST pull-down assays were conducted as 100-l reactions containing 20 mM HEPES, pH 7.6, 150 mM sodium acetate, 0.05% Triton X-100, 10% glycerol, 1 mM DTT, 25 ng/l BSA, 5 l of glutathione-Sepharose Fast Flow (GE Healthcare), 7.5 pmol of GST or 2.5 pmol of GST-Rap1, and between 2.1 and 42 pmol of His 6 -MBP-Taf4/Taf12 variants as input. Binding proceeded for 1 h at room temperature, followed by three washes with GST pull-down buffer. Bead-bound material was eluted with 1ϫ NuPAGE sample buffer, resolved on 4 -12% NuPAGE gels, stained with Sypro Ruby (Invitrogen), and imaged with a Bio-Rad PharosFx imager, and binding data were quantified with QuantityOne Software. MBP pull-down reactions were similar and used 5 l of amylose-agarose (New England Biolabs), 8 pmol of MBP or 4 pmol of MBP-Taf4/Taf12, and input of between 4 and 70 pmol of TFIIA. Binding data were analyzed and plotted using GraphPad Prism.

Identification and Characterization of Novel toa1 and toa2
Mutants-In our previous cryo-EM studies of Rap1-TFIIA-TFIID-RPG DNA complexes, we generated two particular yeast TFIIA variants that contained alanine (Ala) substitutions in the Toa2 subunit (42). Yeast Toa2 Y10A and yeast Toa2 F71A were chosen based on equivalent Y6A and F67A substitutions in human TFIIA␥ that cause defects in activator-dependent but not basal transcription in vitro, without apparent effects on either TFIIA stability or the formation of TFIIA-TBP-DNA ternary complexes (68). These published results suggest that Y6A and F67A substitutions in human TFIIA␥ cause defects in pro-tein-protein interactions that probably involve activators and/or coactivators, such as TFIID (68). When we compared Toa1-Toa2 heterodimeric yeast TFIIA preparations carrying the human equivalent Toa2 Ala substitutions (i.e. yeast Y10A and/or F71A) with WT TFIIA, we observed reduced Rap1-, TFIIA-, and TFIID-dependent looping of RPG DNA (42), results indicating that TFIIA interacts directly with TFIID possibly with Rap1 to drive RPG transcription. We thus reasoned that identifying additional mutant variants of TFIIA (i.e. mutations within Toa1 and Toa2 subunits) and mutations within a/the (putative) TFIIA TFIID-Taf subunit interaction partner, we would both gain greater insight into transcriptional transactivation mechanisms and generate valuable tools for the molecular dissection of the coactivator functions of TFIIA and TFIID.
Before embarking on a systematic molecular genetic dissection of TFIIA-encoding subunits, however, we needed to demonstrate that the biochemical/structural defects we had observed with the Y10A and F71A Toa2/TFIIA variants in our EM studies correlated with in vivo growth deficiencies, as would be predicted if these interactions contributed to RPG transcription, which is tightly coupled to growth rates. Because TOA2 (and TOA1) is a single copy essential yeast gene (69), we constructed a pseudodiploid TOA2 wild type (WT) strain for plasmid shuffle assays in order to test for any growth defects associated with either (or both) of our Toa2 alanine substitutions. We observed that, indeed, both Y10A and F71A Toa2 variants cause temperature-sensitive growth (Ts ϩ ), whereas the Y10A/F71A double mutant was inviable; all proteins were as stable as WT Toa2 (Fig. 1A) and assemble to form TFIIA. Hence, steady state protein levels do not explain the observed growth deficiencies of these strains. Rather, collectively, these data indicate that Toa2 amino acid residues Tyr 10 and Phe 71 do indeed contribute importantly to TFIIA function in vivo and motivated us to systematically analyze both TOA1 and TOA2 by alanine-scanning mutagenesis. A number of Ts ϩ toa1 and toa2 mutants have been described previously; these include toa1-25 (K255A/R257A/K259A) and toa2-3 (D21A/D24A) (56), toa2 I27A or I27K (26), and toa2 L23A, D24A, D24L, I27A, E33L, or A34L (27). However, these mutants were not identified through a comprehensive alanine-scanning mutagenesis approach.
The TFIIA-TBP-TATA co-crystal structure (Protein Data Bank structure 1RM1) reveals a two-domain TFIIA Toa1-Toa2 heterodimeric assembly. One domain is composed of interacting hydrophobic C-terminal Toa1 and Toa2 ␤-sheets that form an extensive TBP interaction surface as well as a small region that binds TATA DNA (70 -72). A second domain is generated by interactions between two N-terminal Toa1 ␣-helices with two Toa2 N-terminal ␣-helices to generate a four-helix bundle. It has been proposed that this four-helix bundle is an extended scaffold that makes important interactions with coregulatory factors, including NC2 (negative cofactor 2), Tafs, and SAGA (26,27,70,(72)(73)(74). To elucidate which TFIIA amino acids contribute to transcription, particularly to RPG transcription, we created non-overlapping three-codon Ala block mutations that span the length of the TOA1 gene as well as similar two-codon Ala substitutions spanning TOA2. AUGUST 9, 2013 • VOLUME 288 • NUMBER 32

TFIID-TFIIA-Rap1 Regulatory Interactions
A number of the toa1 Ala scan mutants caused overt growth deficiencies. Fourteen toa1 block mutants caused lethality or slow/Ts ϩ growth ( Table 1). All such toa1 mutants involve amino acids within either the N or C termini of Toa1, results consistent with a previous deletion analysis of TOA1 performed by Hahn and colleagues (56). Steady state protein levels were lower than WT for many of these toa1 mutants (not shown).
Similarly, a number of toa2 2-Ala block mutations caused growth defects. Seventeen toa2 mutants resulted in lethality, whereas nine were slow growing and/or Ts ϩ . The mutated residues responsible spanned the length of Toa2 (Table 1). All toa2 mutants were readily detectable by immunoblotting (data not shown).
We followed these Ala block mutagenesis experiments by generating and testing the growth of yeast strains carrying a single Ala substitution of several Toa2 amino acids compared with the corresponding block mutants (Table 1). Finally, we identified those toa1 and toa2 mutants that compromised pro- )) were spotted onto selective medium plates and grown at room temperature (20 -23º C). The same dilutions of cells were inoculated onto 5-FOA-containing plates to effect plasmid shuffle (i.e. selection for cells that have lost the covering URA3-marked TOA2 plasmid). These plates were incubated at 23 or 37º C (top) and photographed after 2 (37º C) or 3 (room temperature) days. Bottom, extracts were prepared from these yeast strains, and proteins were fractionated by SDS-PAGE and immunoblotted with anti-HA IgG to score HA-Toa2; Rap1 was similarly probed using anti-Rap1 IgG and served as loading control. The relevant strain genotypes (i.e. preshuffle (toa2⌬, TOA2, toa2 X ), where X represents vector, WT, Y10A, F71A, or Y10A/F71A (left) or postshuffle (toa2⌬, -, toa2 x )) are shown at the top of the cell spot images. B, toa1 mutants conferring Ts ϩ growth generated by alanine-scanning mutagenesis (see text and Table 1). All analyses were performed as in A. C, toa2 Ala mutants conferring Ts ϩ growth generated by alanine scanning (see text and Table 1); plasmid shuffle and steady state protein level assays were performed as in A.

TFIID-TFIIA-Rap1 Regulatory Interactions
tein stability; these were excluded from further analysis, leaving five toa1 Ts ϩ mutants and 16 toa2 Ts ϩ mutants. The growth properties and steady state protein levels of these 21 toa1 and toa2 variants are presented in Fig. 1, B and C. Essentially all of these Toa1 and Toa2 variants accumulate to levels that are similar to WT.
In aggregate, the data of Fig. 1 and Table 1 indicate that multiple domains of both TOA1 and TOA2 are sensitive to Ala substitution mutation, although TOA1 is relatively less sensitive than TOA2. Moreover, toa1 mutants have a generally greater effect on protein stability than toa2 mutants. Mutants exhibiting recessive, altered/conditional growth phenotypes, without apparent affect on protein stability, were of special interest to us as potential tools for investigating the mechanisms of TFIIA coregulatory protein-protein interactions. Finally, these mutant screens confirm and extend the array of variant TFIIA-specific reagents available for mechanistic studies.
RPG Transcript Abundance Decreases in toa1 and toa2 Ts ϩ Conditional Mutants after Shift to Non-permissive Temperature-The tight Ts ϩ growth phenotypes of the toa1 and toa2 mutants allowed us to test whether the observed temperature-conditional growth properties correlated with reduced RPG tran-script levels after a shift to non-permissive temperature. We have previously used this approach to examine the effect of amino acid substitutions within the Rap1 binding domains of Taf4 and Taf5 on RPG transcription (24). RPG mRNA levels are a reliable proxy of initiation rates because RPG mRNA halflives are, on average, shorter than those of most yeast mRNAs. Moreover, RPG transcription initiation rates are among the highest of all RNA polymerase II-transcribed genes; it has been estimated that the 130 ϩ Rap1-driven RPG promoters account for as much as 50% of RNA polymerase II transcription initiation events in rapidly dividing cells (75,76). RPS2, RPS3, and RPS5 RPG mRNA levels were scored in total RNA prepared from cells grown at 23º C and 2 h after a shift to 37º C. Transcript abundance was normalized to U3 snRNA content and compared with WT. PGK1 mRNA was included as an abundant TFIID-independent yet Rap1-dependent transcript. RDN1 encodes the rRNA primary transcript and is transcribed by RNA polymerase I. RPG transcription decreases at elevated temperature in WT cells as a general stress response; thus, the differential between high and low temperatures must be scored for each strain and compared with WT (65). Qualitatively, all toa1 and toa2 Ts ϩ mutants exhibit a more drastic reduction in RPS2, RPS3, and RPS5 transcripts than WT ( Fig. 2A). This effect is quantified for RPS5 mRNA in Fig. 2B. RPS5 mRNA levels dropped at non-permissive temperature (RPS2 and RPS3 mRNAs behave similarly; data not shown), whereas PGK1 mRNA levels increased slightly at 37º C (as metabolic rates generally increase as temperature is increased from 23 to 37º C). To confirm and extend these observations, we performed RNA-Seq analysis on RNA from the toa2 Ts ϩ strains as compared with WT. As expected from the data above and the known coordinate regulation of RPG regulon transcription, all toa2 mutants exhibit a general deficiency in total RPG transcript levels (i.e. sum of RKPM counts for all 137 RPG mRNAs) at 37º C compared with WT ( Fig. 2C and supplemental Tables 1  and 2). Notably, validating the data scoring RPS5 mRNA gene transcription in the two mRNA assays of Fig. 2, A-C, direct comparisons of quantitation of RPS5 mRNA levels via primer extension and RNA-Seq (supplemental Table 1) show excellent agreement (data not shown). The data presented in Fig. 2 indicate a general TFIIA requirement for RPG transcription and show that TFIIA, via Toa1 and particularly Toa2, contributes critically and specifically to high level transcription of RPGs.
Biochemical Characterization of TFIIA Variants-We next chose the subset of Toa2 variants expected to minimize potential global affects of toa2 mutation by focusing on Ala mutations that alter Toa2 amino acids within the N-terminal domain four-helix bundle of the Toa1-Toa2 TFIIA heterodimer: amino acid residues Arg 11 , Ser 18 -Leu 19 , Asp 24 , Leu 26 , Ile 27 , Ile 32 , Met 38 -Arg 39 , Leu 41 , Phe 44 , and Asp 45 . All of these evolutionarily conserved, solvent-accessible residues are distinct from the Toa1-Toa2 dimerization domain and located far from the TBP-TATA DNA-binding portion of TFIIA (Fig. 3); consequently, we expected that all functions of TFIIA containing these Toa2 variants would be WT for "basal" TFIIA functions.
We first showed both that recombinant Toa2 variants could efficiently assemble in vitro with recombinant WT Toa1 to form TFIIA and that the resulting TFIIA could be readily puri- fied. The purity and concentration of the resulting purified, recombinant E. coli-expressed TFIIA isoforms was scored by denaturing SDS-PAGE (Fig. 4A). These preparations of purified TFIIA were then tested for their ability to stabilize a TBP-TATA DNA complex in vitro and their ability to support activated transcription in vitro. TBP-TATA complex stabilization was scored by EMSA in the absence of Mg 2ϩ , conditions where only ternary TFIIA-TBP-TATA DNA complexes are stable (33,68). With the exception of TFIIA containing the R11A variant of Toa2, all other TFIIA variants efficiently form a stable ternary TFIIA-TBP-TATA complex, indicating that both the Toa2 TBP-and Toa2 TATA DNA binding domains (cf. Fig. 3) of the Toa2 variants were functional (Fig. 4B).
We next used a two-template G-free cassette transcription assay to assess the ability of the TFIIA forms to support transcription. Using a large G-free cassette fused to the UAS RAP1 enhancer and the smaller G-free cassette fused to the UAS GAL enhancer as templates, transcription assays were performed using WCE that had been doubly immunodepleted (35) of Rap1 and TFIIA (Ͼ95% and Ͼ80% depletion, respectively; data not shown). All assays were supplemented with Gal4 DNA binding domain-VP16 activation domain fusion protein as control. VP16 targets SAGA but not the TFIID coactivator (77,78), and VP16 activation domain-driven transcription is TFIIA-dependent (68). Control experiments showed that the transcription of both templates is TFIIA-dependent, as was expected (Fig.  4C). Moreover, this system supports Rap1-mediated transactivation (Fig. 4C), consistent with our previous in vitro transcription studies (35).
Having established the Rap1 and TFIIA dependence of transcription in vitro, the double-depleted WCE was used to score the ability of the TFIIA forms to support activated transcription. Each variant was scored in duplicate. The different TFIIA preparations variably supported activated transcription (Fig.  4D). As in the TBP-DNA stabilization assay, the R11A variant displayed lower activity, whereas the D24A variant exhibited the lowest ability to support activated transcription of both templates, although all of the Toa2-mutated TFIIA forms were less active than WT. These data indicate that the mutant TFIIA forms are variably defective in Rap1-activated transcription in vitro, results consistent with our RNA-Seq analyses of RPG regulon transcription in vivo. Indeed, with the exception of the D24A Toa2 variant, the overall patterns of Rap1-driven transcription activity seen in vitro and in vivo are quite similar (see direct comparison in Fig. 4E). Collectively, these data demonstrate that although Toa1-Toa2 heterodimer assembly and TBP-TATA DNA complex stabilization activities of the Toa2 variant TFIIA forms are nearly all WT, these proteins display defects in Rap1-activated transcription.
Toa2 Residues Contribute to TFIIA-TFIID Interaction through the RBD of Taf4-Having identified RPG transcriptional defects in a variety of TFIIA mutants, particularly mutations in TOA2, and knowing that mutations in both TAF4 and TAF5 RBDencoding sequences also cause defects in RPG transcription (24), we reasoned that compromised interaction between TFIIA and TFIID could account, at least in part, for this transcriptional phenotype. Further, because we had previously mapped the Taf4 Rap1 binding domain to Taf4 amino acids ϳ250 -350 (24), we speculated that binding of both Rap1 and TFIIA to the RBD would provide a potent and efficient regulatory node for RPG transcription-specific protein-protein interactions. Consequently, we checked whether TFIID and TFIIA interacted, and if so, whether this interaction depended upon Taf4 RBD sequences. To test these hypotheses, we cultured pseudodiploid cells expressing TAF4 and the lethal variant taf4⌬284 -326. The taf4⌬284 -326 allele carries a deletion of the portion of TAF4 encoding most of the Taf4 RBD, amino acids 284 -326. Both the WT and Taf4⌬284 -326 forms of Taf4 incorporate into TFIID (Fig. 5A), results consistent with the fact that there are 2 mol of the Taf4-Taf12 heterodimer per mol of TFIID (15,44). Moreover, TFIID integrity, as scored by HA-Taf4 and HA-Taf4⌬284 -326 anti-HA IgG co-immunoprecipitation (co-IP) assays, is roughly equivalent regardless of whether WT or WT ϩ ⌬284 -326 forms of Taf4 were coexpressed (Fig. 5B). Importantly, anti-FLAG co-IP experiments using strains expressing the HA-tagged Taf4 (WT or ⌬284 -326) and FLAG-tagged Toa2 show that TFIIA-TFIID coprecipitation is reduced by ϳ90% in the RBD-deleted Taf4-expressing strain (Fig. 5C). These data suggest both that TFIIA and TFIID interact, perhaps via direct binding, and second, that this interaction is dependent upon the RBD of Taf4, as we had speculated.
In order to test whether TFIIA Toa2 sequences contribute to TFIIA-TFIID interaction, we prepared WCEs from toa2 Y10A-, toa2 F71A-, and toa2 Y10A/F71A-expressing cells and again scored TFIIA-TFIID interaction via co-IP. The WCE IP input shows essentially equal amounts of Toa2 and Taf4 in all extracts (Fig. 5D, Lysate). Anti-HA-Toa2 TFIIA immunoprecipitates were formed using bead-bound anti-HA IgG, and the resulting immunoprecipitates were probed with either anti-HA IgG to score IP efficiency or with polyclonal rabbit anti-Taf4 IgG to monitor the amounts of TFIIA-TFIID complex in the different strains. TFIIA containing the Ala-substituted Toa2 proteins showed reduced TFIID-TFIIA coprecipitation (from ϳ5 to ϳ50% of WT Toa2; see legend to Fig. 5). These data indicate that the TFIIA-TFIID interaction is dependent upon specific N-terminal Toa2 residues, a result consistent with binding of these two multisubunit complexes via specific contacts that require unique surfaces within subunits of both coactivators. FIGURE 2. toa1 and toa2 Ts ؉ mutants exhibit reduced RPG transcript levels upon a shift to non-permissive temperature. A, steady state RNA levels in WT and toa1 and toa2 Ts ϩ mutants grown at 23º C or for 2 h at 37º C; see bottom of the gel image. Shown is a representative K-screen image of gel-fractionated multiplex 5Ј primer extension products (24,35); U3 RNA was the loading control. RNA from two independent TOA1 cultures and four independent WT TOA2 cultures were analyzed as controls. B, quantification of the RPS5 signals in A, adjusted for U3 content and normalized to the leftmost TOA1 or TOA2 RPS5/U3 values; WT values ϭ 100%. C, RNA-seq measurements of total RPG regulon RKPM from WT TOA2 (performed in duplicate and averaged; WT TOA2 total average RPG RKPM ϭ 278,970 Ϯ 3101; error bar, S.E.) and selected toa2 Ts ϩ mutants grown at 37º C; no error bars because n ϭ 1. Data are normalized to the WT (TOA2 value ϭ 100%). The primary RKPM data for the individual RPGs and selected transcription proteins are presented in supplemental Tables 1 and 2. AUGUST 9, 2013 • VOLUME 288 • NUMBER 32

JOURNAL OF BIOLOGICAL CHEMISTRY 23281
To more fully explore RBD/Toa2-mediated TFIID-TFIIA interactions, we tested whether overexpression of Taf4 or several other related transcription proteins might rescue the Ts ϩ growth of the toa2 conditional mutants. Such a genetic, high copy suppression interaction(s) would support the notion of direct, physical TFIID-TFIIA Toa2 interaction mediated by Taf4. Synthetic gain-and loss-of-function genetic experiments have been widely used to support the possibility of direct interactions between proteins.
We performed growth assays with yeast expressing a single toa2 Ala mutant transformed with a multicopy plasmid encoding Toa2, Toa1, TBP (Spt15), Taf4, Taf11, or Rap1. All of these

TFIID-TFIIA-Rap1 Regulatory Interactions
genes encode proteins known to interact or suspected of interacting with Toa2/TFIIA. As expected, growth of all of the mutants was restored to WT levels by high copy TOA2 (and TOA1). By contrast, high copy SPT15 and RAP1 slightly inhibit growth, even of the WT strain (Fig. 6). Growth of the toa2 D24A-expressing mutant at elevated temperatures was modestly improved by high copy TAF4, whereas growth of the L41A mutant was significantly increased by high copy TAF11 (Fig. 6). Toa2 mutants R11A, I27A, I32A, and F44A behaved similarly to the D24A toa2 allele, whereas D45A toa2 growth patterns

TFIID-TFIIA-Rap1 Regulatory Interactions
resembled those of L41A in these overexpression plate growth tests (data not shown). Collectively, these synthetic, allele-specific positive and negative genetic interactions indicate likely links between TFIIA, Taf4, Taf11, TBP, and Rap1 and further support the hypothesis that specific toa2 mutants are defective in making direct, specific coregulatory interactions with TFIID. Additionally, these data again suggest a likely Taf target within the context of holo-TFIID-Taf4 (and perhaps Taf11; see Refs. 26 and 74).
Mapping Critical Domains within TAF4 RBD-encoding Sequences via Alanine-scanning Mutagenesis-To pursue the ideas that direct TFIIA-TFIID binding is mediated through direct Toa2-Taf4 RBD interactions and that these interactions are critical for RPG transcription, we subjected essential Taf4 RBD amino acids 250 -350 (24) to alanine block scanning mutation. 6-Ala, 2-Ala, 1-Ala, and ultimately codon randomization mutations within DNA encoding Taf4 amino acids 251-346 were sequentially generated and screened. Each screen was guided by the data generated from the preceding screen (complete data set not shown). Only one 6-Ala block mutant, RBD amino acids 317-322, conferred a Ts ϩ growth phenotype. This Ts ϩ phenotype (Fig. 7) was at least as severe as any of our three previously described taf4 Ts ϩ mutants (taf4-116, taf4-141, taf4-219) (24). 2-Ala block mutagenesis of the 317-322 interval identified mutants G316A/W317A and W317A/L318A as conferring Ts ϩ phenotypes, pointing to Trp 317 as the responsible residue. Consequently, Trp 317 was subjected to codon randomization; both W317D and W317E substitution conferred Ts ϩ growth phenotypes. All variant proteins (6-Ala, 2-Ala, 1-Ala, and codon-randomized) were expressed at roughly WT levels (data not shown, but see Fig. 8). We concluded that RBD amino acid Trp 317 contributes critically to Taf4 function.
Synthetic Sick, Lethal (SSL) Mutants of TAF4 RBD-encoding Sequences Identify Amino Acids Contributing to Taf4-TFIIA Binding-To reconcile how a single Taf4 residue might compromise multiple RBD functions, such as TFIIA-TFIID and   AUGUST 9, 2013 • VOLUME 288 • NUMBER 32

JOURNAL OF BIOLOGICAL CHEMISTRY 23285
Rap1-TFIID interactions, we speculated that additional Taf4 residues might contribute to these binding events. To address this hypothesis, we tested whether combining Taf4 Ala block mutants with the W317D variant would result in SSL growth phenotypes. Interestingly, every 6-Ala block mutation tested within the 251-286 interval, as well as the 306 -310 block, FIGURE 8. SSL/Ts ؉ growth phenotypes of taf4 mutants bearing the W317D substitution plus 6-Ala block mutations. A, top, plasmid shuffle growth assays of strains carrying the indicated form of TAF4 (left): empty vector, WT TAF4, W317D, or 6-Ala X ϩW317D, where X represents the coordinates of the 6 amino acid residues mutated to Ala. B, five viable HA-tagged, shuffled 6-Ala X ϩW317D Taf4 mutants from A were used for co-IP assays to score TFIID integrity and Rap1-TFIID association (middle) and TFIIA-TFIID association (right); immunoblots of the lysates (left) indicate equivalent Rap1, Taf4, Taf5, Taf11, Taf12, and Toa1 content. As noted in the legend of Fig. 5B, we were not able to detect TFIIA in anti-Taf immunoprecipitates (not shown), presumably due to the small amount of TFIIA bound to TFIID. C, effect of W317D substitution on direct interaction between Rap1 and Taf4/Taf12 heterodimers. Purified GST (7500 fmol) or GST-Rap1 (2500 fmol) proteins were immobilized on glutathione-Sepharose beads and incubated with increasing amounts of MBP-Taf4 fusion proteins (coexpressed and copurified with Taf12) or the equivalent Taf4 W317D variant. Bead-bound material was analyzed on SyproRuby-stained NuPAGE gels; gels were scanned, and a quantification of gel data is shown (right); background binding to GST alone was subtracted from all binding reactions. Under these binding conditions, we failed to observe Rap1-TFIIA binding (not shown).
Cells carrying five of the taf4 mutants displaying SSL growth phenotypes grew in liquid culture when the variant protein was the only form of Taf4 in the cell. These five yeast strains were utilized for co-IP experiments (Fig. 8B) to determine whether TFIID-TFIIA and/or TFIID-Rap1 binding was compromised in the synthetic variants. Although TFIID integrity was similar among the taf4 mutants (Fig. 8B, middle), the W317D substitution severely compromised Rap1-TFIID binding (Fig. 8B,  middle); all Taf4 variant proteins were as stable as WT. Subsequent GST pull-down assays conducted with purified recombinant Rap1 and purified recombinant WT and mutant Taf4 proteins showed both that the binding of Rap1 with the TFIID subunit Taf4 is direct and that binding is sensitive to mutation of Taf4 tryptophan 317 to aspartic acid (Fig. 8C).
By contrast with Rap1-TFIID binding, TFIIA-TFIID binding was only modestly affected by the W317D substitution. However, when the W317D Taf4 mutation was combined with 269 -274, 281-286, or 306 -310 6-Ala block Taf4 mutations, a severe decrease in TFIID-TFIIA complex formation was observed (Fig. 8B, right). Collectively, these data indicate that multiple Taf4 RBD residues contribute to interaction with TFIIA. Moreover, the correlation between taf4 mutant slow growth and deficit in Rap1-TFIID and TFIIA-TFIID binding indicates that both of these events are functionally important.
To more precisely define the residues contributing to Taf4/ TFIID-TFIIA binding, we combined W317D with the 2-Ala block and 1-Ala substitutions. The 2-Ala-W317D synthetic screen identified residues Leu 259 -Arg 260 , Glu 271 -Arg 272 , Gly 281 -Leu 282 , and Arg 308 , as causing SSL growth (data not shown). Upon combining 1-Ala substitutions with W317D, we observed variably severe SSL phenotypes when W317D was combined with L259A, R260A, G281A, L282A, and R308A (Fig.  9A). The allele specificity of these genetic interactions suggests that the scored interactions between TFIIA and Taf4/TFIID are also probably direct. To test this hypothesis, we again performed pull-down protein-protein binding assays, although here we used purified recombinant MBP-Taf4 (WT, W317D, and R308A/W317D Taf4) variants and purified recombinant TFIIA. As shown in Fig. 9B, which presents the results of a series of single concentration point titration assays, we found that all three forms of Taf4/12 heterodimer bound TFIIA, with apparent relative affinities of WT Ͼ W317D Ͼ R308A/W317D (Fig. 9B). To more rigorously test the apparent large difference in TFIIA binding between the WT and R308A/W317D Taf4 forms, we performed the pull-down titration assay with quintuplicate replicates of each concentration of TFIIA tested. In this assay, we found that the R308A/W317D Taf4 double mutant variant reproducibly bound TFIIA 2-fold less well than WT (Fig. 9C). In aggregate, these data indicate that (a) TFIIA interacts directly with the Taf4/Taf12 heterodimer; (b) this binding is mutationally sensitive; and (c) TFIIA appears to interact with multiple surfaces/regions of the Taf4 RBD.
Overexpression of TFIIA Rescues the Ts ϩ Growth Deficiencies and RPG Transcription Defects of taf4 SSL Mutants-Considering that toa2 mutants exhibit Ts ϩ growth phenotypes that generally correlate with defective TFIIA-TFIID interaction and that specific taf4 mutants compromise both TFIIA-TFIID interaction in vivo and Taf4-TFIIA binding in vitro, we reasoned overexpression of TFIIA might suppress taf4 SSL/Ts ϩ growth phenotypes. Such approaches have been used to investigate TFIIA-TBP interaction (55,79), and our high copy suppression of toa2 mutants by TAF overexpression also suggested that this might be a useful strategy. To test this hypothesis, a multicopy TOA1 ϩ TOA2 expression plasmid was transformed into a subset of our taf4 yeast strains. These strains were then analyzed (post-plasmid shuffle) for growth at different temperatures. TFIIA overexpression had no effect on the lethality of the taf4⌬284 -326 mutant that removes most of the Taf4 RBD and hence the TFIIA and Rap1 binding site(s). By contrast, the slow/Ts ϩ growth phenotypes of the W317D and, indeed, all of the synthetic single Ala ϩ W317D mutants were alleviated at all temperatures tested (Fig. 10A). Strikingly, even the extreme slow growth and temperature sensitivity of the R308A/W317D double mutant was improved. These data support the existence of a TFIIA binding domain within the Taf4 RBD that involves multiple residues in the Leu 259 -Trp 317 interval. This function is required for growth.
To assess the relevance of the Taf4 TFIIA interaction domain to RPG transcription, specific mRNA abundance was scored in taf4 mutants with or without high copy TOA1 ϩ TOA2. Specific RPG mRNAs were less abundant in each of the taf4 mutants lacking high copy TFIIA compared with WT (Fig.  10B). TFIIA overexpression had no positive effect on RPG mRNA levels in WT cells, whereas RPG transcripts were increased in each of the taf4 mutants with high copy TOA1 ϩ TOA2. Quantification of RPS2 mRNA showed a roughly 2-fold increase in transcript levels in many of the taf4 strains when TFIIA was overexpressed (Fig. 10C); RPS3 and RPS5 mRNAs were similar (Fig. 10B) (data not shown). Once more, these data support the notion that TFIIA and TFIID directly interact through specific Toa2 and Taf4 RBD domains. Note that the TFIID-independent PGK1 mRNA was also increased when TFIIA was overexpressed; PGK1 is a known TFIIA-dependent gene (26,74). We speculate that the effect of TFIIA overexpression on PGK1 transcription could be due to a general acceleration of cellular metabolic activity when TFIIA levels rise.

DISCUSSION
Activator-stimulated conformational rearrangements of Taf-containing complexes have been proposed based upon biochemical and functional studies of TFIID binding to enhancerpromoter DNA in the presence and absence of specific transactivators (33, 34, 80 -84). Moreover, single molecule studies of the effects of activators and TFIIA on TFIID-promoter DNA binding are consistent with such models, whereas recent EM studies confirm that structural transitions can be stimulated by TFIID-promoter DNA binding in the presence of TFIIA (85,86). Although several laboratories have reported that TFIIA is not required in vitro for basal transcription catalyzed by purified RNA polymerase II plus TFIIB, -E, -F, and -H and TBP (3), it is notable that two very recent studies (87,88) have shown that TFIIA is essential for the formation of cryo-EM-visualizable, functional PICs formed from purified components. Finally, quaternary complexes of Rap1 enhancer-promoter DNA, TFIIA, TFIID, and Rap1 undergo induced conformational rearrangements (42). Thus, several groups agree that these key transcription proteins can interact dynamically upon engaging enhancer-promoter DNA. However, neither the physiological relevance of such activator-induced TFIID and TFIIA conformational rearrangements toward PIC formation in vivo nor the exact TFIIA or Taf domain(s) involved are known. This lack of physiological data was the driving force for our embarking on the series of studies described in this report. We believe that analysis of yeast RPG transcription affords FIGURE 9. SSL growth tests identify multiple Taf4 residues that collaborate with Trp 317 to support growth and facilitate Taf4-TFIIA interaction. A, plasmid shuffle growth assays of strains carrying the indicated form of TAF4 (left): empty vector, WT TAF4, W317D, or 1-Ala X ϩW317D combinations. B, left, effect of Taf4 amino acid substitutions (W317D and R308A/W317D) on direct Taf4-TFIIA interaction. Purified MBP (8000 fmol) or MBP-Taf4/12 heterodimer variants (4000 fmol) were bound to amylose-Sepharose beads and incubated with increasing amounts of purified TFIIA. Binding (white arrows; Toa1 (top) and Toa2 (bottom)) was scored by SDS-PAGE as in Fig. 8 (left); background binding to MBP alone was subtracted from all binding reactions. Right, quantification of binding data. C, binding assays were repeated for WT and R308A/W317D Taf4 variants as in B, except that the binding reactions were performed with five replicate binding reactions for each TFIIA concentration. The data were quantified and plotted (open circles, WT; filled triangles, R308A/W317D); background binding to MBP alone was subtracted from all binding reactions; the 95% confidence intervals for the plots are shown for each curve (dashed lines), as are the mean and S.E. (error bars) for each individual measurement. t tests indicated that the R308A/W317D Taf4 variant bound 2-fold less TFIIA than WT Taf4 at all concentrations tested (p Ͻ 0.05).

TFIID-TFIIA-Rap1 Regulatory Interactions
unique opportunities for systematically addressing these gaps in our understanding of the mechanism of action of the TFIIA and TFIID coactivators.
The vigorously transcribed RPGs are TFIID-dependent, and transcription of this gene family is driven in large part through direct Rap1-Taf interactions (24,35). Moreover, mutations that weaken Rap1-TFIID (Taf4/5) RPG transcription-activating contacts exhibit overt growth deficiencies (24). This is due to the fact that yeast cell growth is so tightly linked to expression of the genes encoding the components of the translation machinery (i.e. RPGs). We capitalized upon this relationship between RPG transcription and cellular growth to systematically probe the importance of individual TFIIA residues to RPG transcription, initially by monitoring yeast cell viability and temperature-conditional growth.
Consistent with earlier reports (26,27,55,56,79), we found the TOA1 and TOA2 are sensitive to Ala substitutions ( Fig. 1 and Table 1). This was particularly true for TOA2. Toa1 apparently makes more significant contributions to TFIIA complex formation/stability because a large portion of our toa1 Ala mutants destabilized Toa1 and hence TFIIA levels. Consistent with the idea that Toa2 N-terminal residues make critical protein-protein interactions with coregulators like TFIID, we identified multiple toa2 Ala substitutions with WT protein stability that induced slow growth (Fig. 1). Several of our mutants are consistent with previously reported toa2 mutants that affect N-terminal residues Asp 24 and Ile 27 (26,27,56). The clustering of many of these Ala-substituted residues within the TFIIA/ Toa2 structure suggests shared function(s) (Fig. 3).
To assess the possible contributions of Ala-mutated Toa1 and Toa2 residues toward ribosomal protein gene expression, we tested RPG transcription at permissive and non-permissive temperatures in a subset of our mutant collection. These analyses confirmed the importance of TFIIA function to transcription of both individually analyzed RPG mRNA-encoding genes (Fig. 2, A and B) and the entire RPG regulon in vivo (Fig. 2C). Results from the use of these same mutants for in vitro studies, such as TBP-TATA DNA complex stabilization and transcription (Fig. 4, B-D), suggest quite similar allele-specific patterns of Toa2 residue function with RPG regulon transcription (Fig.  4E), although some differences were evident between in vivo RPG mRNA results and the biochemical data. For example, toa2 I32A exhibited the most severe RPG transcript phenotype of all of the mutants in vivo but not in vitro (Figs. 2 and 4). Similarly, the toa2 R11A in vivo data did not correlate perfectly with the biochemical data. Overall, however, considering that these are quite different assays (i.e. one assay scored expression of all RPGs (Fig. 2C), another scored the in vivo transcription of a handful of RPGs (e.g. Fig. 2, A and B; RPS2, RPS3, and RPS5), and the third scored the in vitro transcription of a simple chimeric TFIID-dependent, Rap1-driven synthetic reporter gene (24,35) (Fig. 4D)), we believe it is highly significant that these results are so similar regarding the relative contribution of individual TFIIA residues to RPG/Rap1-driven transcription. We also observed toa2 allele-specific effects upon TFIIA-TFIID interaction (Fig. 5). Ala mutation of N-terminal sequences located within and abutting the TFIIA four-helix bundle (Fig.  3B) affected TFIIA-TFIID binding.
Consistent with previous reports of direct binding of purified TFIIA and Taf4 (28,29), we obtained genetic evidence suggesting TFIIA-TFIID interaction through Taf4 and Toa2 D24 (Fig.  6). Notably, the toa2 D24A mutant exhibits a strong Ts ϩ growth phenotype, normal protein abundance, and severely compromised RPG transcription in vivo (Figs. 1 and 2). TFIIA containing Toa2 D24A efficiently binds and stabilizes the TBP-TATA DNA complex yet has the weakest ability to reconstitute in vitro transcription, whether Rap1/TFIID-or Gal4-VP16/ SAGA-dependent (Fig. 4). Hahn and colleagues (27) have previously described cross-linking of TFIIA to Taf4 and Spt8 in vitro, results suggesting that TFIIA-TFIID and TFIIA-SAGA interactions occur through similar TFIIA N-terminal residues, including those proximal to Toa2 Asp 24 . These workers also observed conditional growth of toa2 D24A mutants and SSL genetic interactions between toa2 D24L and both spt8 and spt3 null alleles (27). Thus, it is likely that TFIIA provides coactivator functions on both TFIID-dependent and SAGA-dependent genes by using Toa2 Asp 24 as a component of distinct proteinprotein interaction pathways. RPG transcription in vivo is TFIID-dependent but SAGA-independent (4,24,35,89). Notably, in addition to our reported binary TFIIA-Taf4 interaction, Stargell and colleagues (26,74) have identified and analyzed TFIIA-Taf11 binding through Toa2 Ile 27 . We too observed genetic interactions between TOA2 and TAF11, but through Toa2 residue Leu 41 rather than Ile 27 (Fig. 6). The reason(s) for this discrepancy is unclear at present. As with most genetic interaction studies, overexpression complementation of the type we observed here with TAF4 and/or TAF11 can only support the idea of direct protein-protein interactions. In this report, these experiments represent but one form of data supporting direct Toa2/TFIIA-Taf4/TFIID binding.
Given the apparent evolutionary conservation of the putative TFIIA-TFIID interaction via Taf4, we pursued characterization of the TFIIA binding domain in Taf4. Preliminary analyses (Fig.  5, A-C) implicated Taf4 RBD residues (24) as a/the domain of Taf4 that formed the binding site for TFIIA. The possibility that the Taf4 RBD, the direct binding site for Rap1, was also a binding site for TFIIA was intriguing. If multiple regulatory protein interactions targeted the Taf4 RBD, such multivalent interactions would provide important new insights into the physiological relevance and mechanism underlying the formation of the stable, conformationally rearranged Rap1-TFIIA-TFIID-DNA quarternary complex that we have observed previously (42).
We performed a series of genetic screens of the RBD-encoding sequences of TAF4 and found that Taf4 amino acid residue FIGURE 10. Overexpression of TFIIA rescues the synthetic growth and RPG transcription deficiencies of taf4 1-Ala X ؉W317D mutants. A, plasmid shuffle assays tested the effect of high copy TOA1 ϩ TOA2 on the growth phenotypes of TAF4, lethal taf4⌬284 -326, W317D, or the indicated Taf4 1-Ala X ϩW317D strains. Only the TOA1-TOA2 3Ј to 3Ј gene orientation efficiently rescued growth (data not shown). B, total RNA was prepared from room temperature cultures of shuffled mutant strains with or without high copy TOA1/TOA2 and subjected to multiplex 5Ј primer extension analysis as in Fig. 2. A representative K-screen gel scan image is shown. C, quantification of RPS2 mRNA levels.

TFIID-TFIIA-Rap1 Regulatory Interactions
Trp 317 was the most important single residue within the mapped TFIIA-RBD interaction network (Fig. 7). However, additional genetic screens of the Taf4 RBD demonstrate that there are clearly additional amino acid residues that collaborate with Trp 317 (Fig. 8). Combining Taf4 RBD Ala block mutations with W317D resulted in SSL and Ts ϩ growth defects that correlated with weaker TFIIA-TFIID interaction than that seen with Taf4 W317D alone (Fig. 8B). W317D reduced but did not eliminate apparent Rap1-Taf4 binding, consistent with polyvalent Rap1-TFIID interactions occurring through Taf4, Taf5, and Taf12 (24); this was evident in direct Rap1-Taf4 binding assays with purified recombinant proteins (Fig. 8C). Moreover, the loss of function in both TFIIA-Taf4 and Rap1-Taf4 binding that resulted from W317D indicates likely overlapping functions of Taf4 residues that direct each binding event. Our additional Ala substitution mutagenesis and SSL screens with TAF4 identified the responsible Taf4 residues as Leu 259 , Arg 260 , Glu 271 , Arg 272 , Arg 277 , Ile 278 , Ala 279 , Leu 280 , Gly 281 , Leu 282 , Arg 308 , Ala 309 , and Ser 311 . All of these amino acid residues abut Trp 317 (Fig. 9A). Structural studies could determine whether these residues actually reside within close proximity in the tertiary structure of Taf4. Consistent with the overlapping function model, the R308A/W317D variant protein displays markedly reduced direct binding to TFIIA in vitro (Fig. 9, B and C). Thus, it is likely that these Taf4 RBD residues contribute to both TFIIA-TFIID binding and Rap1-TFIID binding, although the determination of the extent of functional overlap must await future fine structure mapping.
Given these observations, we hypothesized that overexpression of TFIIA might rescue both the SSL/Ts ϩ growth phenotypes of the taf4 AlaϩW317D mutants and the observed RPG transcription defects if slow growth was indeed caused by deficiencies in TFIIA-TFIID and Rap1-TFIID interactions. As a control, we tested the taf4⌬284 -326-expressing strain in this assay. These cells carry a mutation that removes much of the RBD and, consequently, the TFIIA/Rap1 binding residues mapped herein. The taf4⌬284 -326-derived strains are inviable, despite the fact that the expressed, truncated Taf4 protein is stable (24) and incorporates efficiently into TFIID (Fig. 5A). Not surprisingly, overexpression of TFIIA does not rescue growth of this strain (Fig. 10A), indicating that TFIIA overexpression per se cannot bypass the requirement for TFIIAand/or Rap1-TFIID interactions. By contrast, overexpression of TFIIA corrected the SSL/Ts ϩ growth of both the single W317D mutant and the synthetic AlaϩW317D mutants (Fig. 10A); notably, even the nearly inviable R308A/W317D mutant grew well when TFIIA was overexpressed. Finally, and most importantly, analysis of RPG mRNA gene transcription in these same strains showed that overexpression of TFIIA restored mRNAs to nearly WT levels (Fig. 10, B and C), providing strong support for our hypotheses regarding TFIIA-/Rap1-TFIID binding and RPG expression. Presumably, elevated TFIIA increases binding and/or stabilizes the interaction of both TFIIA and Rap1 to the RBD of Taf4/TFIID, although it remains for these hypotheses to be tested directly.
The implications of our findings are that TFIIA and Rap1 both bind to TFIID through a small set of amino acids within the Taf4 RBD (as well as to Taf5 and Taf12 RBDs) (24, 35). We do not yet know the exact extent of overlap of the Taf4 residues used for Rap1 or TFIIA binding; nor do we know if binding occurs (see Fig. 11) simultaneously, cooperatively, or competitively with each other or with a possible regulatory repressor protein(s) (see below). Cryo-EM has shown (42) that in quarternary pre-PIC subcomplexes, Rap1 and TFIIA co-occupy a region of TFIID that contains one of the two asymmetrically exposed Taf4/Taf12 heterodimers present in the complex (40 -44). Should competitive binding to TFIID occur, Rap1 might actively displace or restrict TFIIA to a different location on TFIID. Such a scenario is similar to the reported competition between the Spt8 SAGA subunit and DNA for TFIIA/TBP binding (27,90) while also reminiscent of the model proposed by Olave et al. (29), wherein the mammalian RBP repressor displaces human TFIIA by competition for a common binding site on human Taf4. In such models, a TFIIA-Rap1 competitive dynamic for binding to the Taf4 RBD might help to establish alternate modes of TFIID-promoter DNA binding involving TFIIA-TBP interactions and/or Rap1-TFIID-DNA binding, thereby initiating key conformational alterations of the Rap1-TFIIA-TFIID-DNA complex that are important for stabilizing this assembly.
Final tests of these ideas will require the generation of true separation-of-function mutant alleles of TAF4, RAP1, and TOA2. Preferably, such separation-of-function mutants will be tested in the context of intact, purified TFIIA, Rap1, and most importantly TFIID, as was recently reported for the analysis of the contributions of the human Taf3 plant homeodomain finger domain in human TFIID transcriptional functions (91). The FIGURE 11. Schematic depicting possible modes of binding of TFIIA and Rap1 to the RBD of the Taf4 subunit(s) of TFIID. The 388-residue Taf4 protein contains an unusual C-terminal "split" histone fold domain (HFD) that forms a heterodimer with the histone fold domain of Taf12 (92, 93) (Taf12 not shown). The Taf4 RBD (i) comprises amino acid residues 253-344 (24). The 13 key Taf4 RBD amino acid residues identified in this study are indicated, 259 -309 (red type) and Trp 317 (W317; blue type). Similarly, the 11 critical N-terminal Toa2 residues important for mediating protein-protein binding that were identified by this study (and others) are shown by the black oval within the red Toa2 protein. TFIIA, Rap1, and Taf4 could either directly form a ternary complex (in ordered or random fashion) (ii), form exclusive binary Rap1-Taf4 complexes (iii), or form exclusive binary TFIIA-Taf4 complexes (iv). Presumably, in order for Rap1-dependent TFIIA rearrangement to take place within the TFIID-TFIIA-Rap1/UAS RAP1 enhancer-promoter DNA complex (42), either the complex in ii or the complex in iii or iv (after the addition of the missing component) serves as substrate for conformational change(s). See text for additional discussion. AUGUST 9, 2013 • VOLUME 288 • NUMBER 32 tools and methods we have developed in this study, along with our efficient TFIID isolation protocols (15,35), will greatly facilitate this endeavor. The utilization of such mutant TFIID for combined biochemical and structural studies will help to precisely define the mechanism(s) underlying PIC structural transitions and their contribution to transcription regulation.