Identification of a Novel TATA Element-binding Protein Binding Region at the N Terminus of the Saccharomyces cerevisiae TAF1 Protein*

TFIID, a multiprotein complex composed of TATA el-ement-binding protein (TBP) and 14 TBP-associated factors (TAFs), can directly recognize core promoter elements and mediate transcriptional activation. The TAF N-terminal domain (TAND) of TAF1 may play a significant role in these two principal TFIID functions by reg-ulating the access of TBP to the TATA element. In yeast, TAND consists of two subdomains, TAND1 (10–37 amino acids (aa)) and TAND2 (46–71 aa), which interact with the concave and convex surfaces of TBP, respectively. Here we demonstrate that another region located on the C-terminal side of TAND2 (82–139 aa) can also bind to TBP and induce transcriptional activation when tethered to DNA as a GAL4 fusion protein. As these properties are the same as those of TAND1, we denoted this sequence as TAND3. Detailed mutational analyses revealed that three blocks of hydrophobic amino acid residues located within TAND3 are required not only for TBP binding and transcriptional activation but also for supporting cell growth and the efficient transcription of a subset of genes. We also show that the surface of TBP recognized by TAND3 is broader than that recognized by TAND1, although epitope were purchased from Santa Cruz Biotechnology, Inc. Northern Blot Analysis— Northern blot analysis was performed as described previously (60). To prepare the probes, DNA fragments sur- rounding the initiating methionine were amplified by PCR from yeast genomic DNA, purified, and 32 P-labeled using a random priming method. The PCR primer pairs used were as follows: TK1881/TK1882 HIS4 , CYC3 , , , Fuji BAS 2500 PhosphorImager Lab

In eukaryotes, transcriptional initiation and regulation of class II genes require a plethora of transcription factors including general transcription factors (TFIIA, TFIIB, TFIID, TFIIE, TFIIF, and TFIIH), mediators, cofactors, chromatin modifiers, gene-specific activators or repressors, and RNA polymerase II (reviewed in Refs. [1][2][3][4]. In Saccharomyces cerevisiae, TFIID is composed of TATA-binding protein (TBP) 1 and 14 TBP-associated factors (TAFs) (5,6). TBP binds specifically to the TATA element (7), whereas TAFs bind directly or indirectly to other core promoter elements like the initiator and downstream promoter element (8). TAFs also play a significant role in facilitating transcription in response to various types of activators (2,3). Interestingly, since some of the TAFs are also included in the Spt-Ada-Gcn5-acetyltransferase (SAGA) complex (9) that is centrally involved in remodeling nucleosomal arrays on a subset of promoters (10), they may also be crucial for the recognition of a chromatinized template.
TAF1 is one of the TFIID-specific subunits and is thought to serve as a platform for the assembly of the whole TFIID complex (11). Of the presumptive multiple TBP-binding sites of the TAF1 protein (Taf1p) (12,13), the one located at the N terminus, designated as TAF N-terminal domain (TAND), has been characterized the most extensively (12, 14 -16). TAND consists of two subdomains, TAND1 (10 -37 aa) and TAND2 (46 -71 aa), which bind to the concave and convex surfaces of TBP, respectively (15,17). Importantly, TAND1 can restrict TBP binding to the TATA element, since it overlaps the DNA recognition surface of TBP (15,18,19). In fact, when TBP is mixed together with TAND in vitro, they promptly form a stable complex that is unable to bind to the TATA element (15,16). Moreover, both native and partially reconstituted TFIID were shown to exhibit lower activities for DNA binding and basal transcription than TBP alone (5, 20 -22). On the other hand, since TAND1 and acidic type of activators like VP16 compete for TBP binding (18), it has been proposed that activators may first sever the inhibitory interaction between TAND1 and TBP and subsequently facilitate the transfer of TBP from TAND1 to the TATA element (denoted as the two-step handoff model in Ref. 23). In this model, TFIIA is proposed to also play a pivotal role in dislodging TAND2 from the opposite side of TBP by competitively binding to TBP. In good agreement with this, TFIIA significantly increased the DNA-binding ability of either native TFIID (5,24) or partially reconstituted TFIID containing TAF1, TAF4, and TBP (25). TFIIA also suppressed the temperature-sensitive growth of strains lacking TAND (⌬TAND) (15).
TBP function is also negatively regulated by other factors, such as Mot1 and NC2, or by self-dimerization (reviewed in Ref. 26). Mot1 is a member of the Snf2/Swi2 ATPase family and, in the absence of ATP, forms a complex with TBP in solution or on the TATA element (27,28). However, when ATP is provided to this complex, some conformational change appears to occur in Mot1 or TBP upon ATP hydrolysis, and this results in dissociation of TBP from the TATA element (28). In contrast, NC2 is a histone-folded heterodimer, composed of NC2␣ and NC2␤ (29 -33), that inhibits TATA-dependent transcription as well as facilitating downstream promoter element-dependent transcription (34). Mechanistically, it was shown that NC2 associates preferentially with promoter-bound TBP and thereby precludes further recruitment of TFIIA and TFIIB (29,35). Therefore, the inhibitory mechanism of NC2 differs from those of TAND and Mot1 since it does not entail TBP dissociation from the DNA. Importantly, genome-wide expression analyses demonstrated that all of these TBP inhibitory factors, i.e. TAND, Mot1, and NC2, regulate transcription not only negatively but also positively (36 -40). Although the mechanisms by which these factors exert such opposing functions are poorly understood, recent findings that NC2␣ and NC2␤ are not always associated and can act independently under certain conditions may provide a partial explanation, at least in the case of NC2 (41,42).
Chromatin immunoprecipitation analysis revealed that there are two distinct classes of promoters in S. cerevisiae, namely TAF-dependent and TAF-independent promoters (43,44). The former depends on multiple TAFs for its transcription, whereas the latter does not. In addition, although significant amounts of TAFs are recruited specifically to the former, TBP is recruited to both types, which suggests that there may be other form(s) of TBP besides TFIID. It still remains unclear whether this corresponds to a free form of TBP or form(s) associated with Mot1, NC2, SAGA or other unknown factors (43)(44)(45). In any case, the recruitment of an appropriate form of TBP onto each promoter appears to be directed by upstream activator(s) (46,47). In the case of TFIID, a cooperative interaction with Mediator is also important for the efficient assembly of the productive preinitiation complex (48,49). Intriguingly, in strains lacking TAND1 (⌬TAND1), artificial recruitment of some Mediator components by connection to the Zif DNA binding domain dramatically increased transcription from a template bearing Zif-binding sites (50). In other words, Mediator recruitment by itself is not sufficient to relieve an inhibitory effect of TAND1. Since pre-recruitment of Mediator by an activator, in the absence of TFIID, decreased preinitiation complex assembly and transcription in vitro (49), it appears that the inhibitory effect of TAND1 could be relieved by the concurrent actions of an activator and Mediator.
Accumulating evidence suggests that TAND is in fact involved in transcriptional regulation (19, 40, 50 -54). Despite this important function, the primary sequence of the TAND region is poorly conserved between yeast and Drosophila, with the exception of the small TAND2 segment (17,18). Structural determination of the binary complex composed of Drosophila TAND1 and TBP revealed that TAND1 mimics the minor groove surface of the partially unwound TATA element (55). Since yeast TAND1 (10 -37 aa) is about half the size of Drosophila TAND1 (11-77 aa), it is difficult to imagine that these two TAND1 sequences can form a similar structure. Thus, we suspect that there may be another unidentified TBP-interacting region in close proximity to TAND1 in yeast Taf1p (see details in the first paragraph under "Results"). To test this possibility, we searched for TAND1-like activity in the region adjoining TAND2, and we found that the sequence from 82 to 139 aa meets this criterion. We designated this region as TAND3 and compared its properties with those of TAND1. The relationships of yeast TAND1 and TAND3 to Drosophila TAND1 are also discussed.

EXPERIMENTAL PROCEDURES
Yeast Strains and Culture-Standard techniques were used for yeast growth and transformation (56,57). Yeast strains used in this study are listed in Table I. All were generated from the Y22.1 (⌬taf1) strain (15) using a plasmid shuffle technique (58).
The BamHI-BssHII fragment encoding TAND1 of pM1169 was replaced with the BamHI-BssHII fragment encoding TAND3 that was amplified from pM1169 itself by PCR using the primer pair TK3824 and TK3825. This created pM4171. Similarly, the BamHI-BssHII fragment encoding TAND1 of pM1169 was replaced with the BamHI-BssHII fragment encoding TAND3ABC mut amplified from pM4116 (constructed as described below) by PCR using the primer pair TK5162 and TK5163. This created pM4225.
␤-Galactosidase and 3-AT Sensitivity Assays-Plasmids that express GAL4 fusions, as described above, were transformed into yeast strains SFY526 or HF7c to measure the transcriptional activation of a chromosomally integrated lacZ or HIS3 reporter gene driven by the GAL1 promoter, respectively. Transformants were grown in selective medium and assayed for ␤-galactosidase activity or 3-AT sensitivity, according to the manufacturer's protocol (Clontech).
GST Pulldown Assay-Expression plasmids based on pGEX2T (Amersham Biosciences) or pET28a (Novagen), as described above, were transformed into Escherichia coli DH5␣ or BL21(DE3) codon plus (Stratagene) to express GST-tagged TAND proteins or histidine-tagged yeast TBP, respectively. The preparation of these proteins and the GST pulldown experiments were conducted as described previously (17). For quantitative analysis, gels were stained with SYPRO Ruby Protein Gel Stain (Molecular Probes) and quantitated with a Fuji LAS-1000 Plus image analyzer.

RESULTS
The Region between Amino Acids 82 and 139 of Taf1p Functions as a Strong Activation Domain-Our previous observation that TAND1, but not TAND2, can function as a transcriptional activation domain (AD) prompted us to propose a model in which the transient displacement by an AD of TAND1 from the concave surface of TBP may be important for activation (23). According to this model, any molecule that competes with TAND1 for TBP binding (including TAND1 itself) could function as an AD. Previously, we noticed that the region between 72 and 96 aa of Taf1p could enhance the AD function of TAND1 (6 -41 aa) by about 1.5-fold, although this region itself did not function as an AD (23). Thus, we suspect that another, as yet unidentified, region located at the C-terminal side of TAND2 (46 -71 aa) may bind to some areas that are not occupied with TAND1 on the concave surface of TBP and thereby contribute to the inhibitory effect of TAND on TBP function. This may also provide a plausible explanation for the observation that yeast TAND1 is smaller in size and possesses weaker TBP binding activity than Drosophila TAND1. Furthermore, it is reasonable to postulate that the region between 72 and 96 aa described above constitutes only a part of such a third TBP-interacting domain and can thereby help the AD function of TAND1 without functioning as an independent AD. If this is the case, the complete domain should activate transcription when tethered to DNA.
To explore this possibility, we connected the region between 70 and 170 aa to the GAL4 DNA binding domain, and we tested its ability to activate transcription from the GAL1 promoter containing the GAL4 DNA-binding sites in yeast cells (Fig. 1, construct 1). Two distinct reporter genes, lacZ and HIS3, were used to quantify the AD activity by measuring ␤-galactosidase activity (Fig. 1B, right panel) and plating yeast cells in media containing different concentrations of 3-AT (Fig. 1B, left panel), respectively. Both assays revealed that this region clearly shows AD activity (4th row in Fig. 1B) that was stronger than that of TAND1 (2nd row) but weaker than that of VP16 (3rd row), thus supporting our hypothesis.
Next, to determine the minimal region that functions as an AD, we constructed a series of truncated mutants as described in Fig. 1A. Mutation of the C-terminal region that generated a series of truncated mutants of the region between 70 and 170 aa (constructs 2-9) indicated that the region between 131 and 140 aa must be critical for AD function since a significant difference was observed when the activities of constructs 4 and 5 were compared in both assays. However, the N-terminal mutants lost their activities gradually as this region was truncated up to aa 110, irrespective of whether they carried the C-terminal region between 141 and 170 aa (constructs 10 -13) or not (constructs 18 -21) (compare the results of both assays in Fig. 1B). These results indicate the presence of multiple subdomains that carry AD function within the region between 81 and 140 aa. Importantly, further truncation at or beyond 121 aa eliminated the AD activity (constructs 14 -17), indicating that the region between 111 and 120 aa is essential for AD function. Among these constructs from 1 to 21, construct 18 (81-140 aa) was the shortest one that retained AD activity that was almost equivalent to that of construct 1. Interestingly, the results of several additional mutants, generated by removing methionine clusters at both ends (i.e. Met-82, -83, -85, -137, and -139) of this region (constructs [22][23][24][25][26], revealed that these methionine residues are also critical for AD function. Accordingly, we concluded that the region between 82 and 139 aa (construct 26) is minimally required for AD function. It should be also noted that AD function appears to be specific to this region since the adjoining region between 137 and 190 aa (construct 27) was totally inactive in the same assays (Fig. 1B,  bottom row).
The Region between Amino Acids 82-139 of Taf1p Directly Binds to TBP and Is Required for Cell Growth and the Transcription of a Subset of Genes Under Specific Conditions-The results described above indicate that the region between 82 and 139 aa can function as an AD, just like TAND1. This in turn suggests that this region may be another TBP binding domain of Taf1p. To examine this possibility, we constructed several plasmids that express this region either alone or as a fusion protein with TAND1ϩ2, TAND1, or TAND2 in bacterial cells. GST pulldown assays were conducted to investigate whether these proteins can directly bind to TBP ( Fig. 2A). As described previously, the TAND1ϩ2 proteins could form a stable and stoichiometric complex with TBP ( lanes 6 and 7), whereas TAND1 or TAND2 alone could not (lanes [12][13][14][15]. In parallel with these findings, we found that a fusion protein of this region, i.e. 82-139 aa, with TAND1 or TAND2 could also form a complex with TBP, albeit less stably than TAND1ϩ2 (lanes 8 -11), whereas this region alone could not (lanes 16 and 17). These observations indicate that the region between 82 and 139 aa is a bona fide TBP binding domain that cooperates with either TAND1 or TAND2. We hereafter designate this region as TAND3 since it not only functions as an AD (Fig. 1) but it can also, like TAND1, bind directly to TBP (Fig. 2A). The failure to identify this region earlier as a TBP binding domain most likely reflects the very stable binding of TAND1ϩ2 to TBP (15,17). In fact, there was no discernible stabilizing effect of TAND3 on TBP binding when it was fused to TAND1ϩ2 (compare lanes 5 and 7). However, a stabilizing effect was observed when the concentration of potassium chloride in the binding buffer was increased to 0.3-0.4 M. In this case, the complex of TAND1ϩ2, but not that of TAND1ϩ2ϩ3, with TBP becomes less stable under these conditions (data not shown) (15). Thus, TAND3 may bind to some surfaces of TBP that are not occupied with TAND1 and TAND2, presumably via hydrophobic rather than ionic interactions.
Since TAND1 and TAND2 are required for the normal growth of yeast cells at higher temperatures (17), we wished to determine whether or not TAND3 has a similar function (Fig.  2B). Deletion of TAND3 alone did not affect the growth at either 25 or 37°C, contrary to the effect of deleting TAND1 or TAND2 (Fig. 2B). However, we noticed that the growth rate of the TAND1-deleted (⌬TAND1) strain was reduced slightly, but reproducibly, at 37°C by the additional deletion of TAND3 (compare ⌬1 and ⌬(1ϩ3) in Fig. 2B). Interestingly, this growth retardation effect of ⌬TAND3 was specifically observed in combination with ⌬TAND1 but not with ⌬TAND2. This suggests TAND1 and TAND3 may play a redundant role. We could not detect a clear phenotypic difference between the ⌬TAND(1ϩ2) and ⌬TAND(1ϩ2ϩ3) strains, apparently because the former mutation alone was sufficiently severe to prevent growth at higher temperatures. Immunoblot analysis confirmed that all TAF1 alleles were expressed at comparable levels for at least 24 h after the temperature shift to 37°C (Fig. 2C).
We and others previously identified a subset of genes that are affected in the ⌬TAND(1ϩ2) strain (40,54). 2 To investigate whether ⌬TAND3 can further affect gene expression in this strain, we compared the transcription of several TAND(1ϩ2)dependent (HIS4, CYC3, RPS5) or -independent (YDR119W, MET17, PGK1, ADH1) genes in the wild type, ⌬TAND(1ϩ2), and ⌬TAND(1ϩ2ϩ3) strains before or after shifting the temperature to 37°C (Fig. 2D). Although transcription of the PGK1 and ADH1 genes was not affected in any strain, that of the YDR119W gene was severely impaired in the ⌬TAND(1ϩ2ϩ3) strain at 37°C (see quantified results in the right panel). Similarly, transcription of the MET17 and CYC3 genes was markedly decreased in the ⌬TAND(1ϩ2ϩ3) strain at 37°C. These observations indicate that the function of TAND3 becomes more critical for transcription of these three genes at 37°C, at least when TAND(1ϩ2) is deleted. Intriguingly, TAND3 was not specifically required for transcription at 25°C. In contrast, a similar deleterious effect of ⌬TAND3 was not evident for the transcription of the other two TAND(1ϩ2)-dependent genes, HIS4 and RPS5. Taken together, these results suggest that although TAND3 may not be globally required for gene expression, it seems to be necessary for the efficient transcription of a subset of genes under some specific conditions, e.g. when TAND(1ϩ2) is deleted from Taf1p (see "Discussion").

Hydrophobic Amino Acid Residues Clustered into Three Distinct Subregions of TAND3 Are Important for AD and TBP
Binding Activities-We demonstrated that TAND3 carries multiple functions as described above, i.e. AD and TBP binding activities as well as functions in supporting cell growth and gene expression under specific conditions. To determine whether these activities can be ascribed to the same or differ-2 K. Ohtsuki and T. Kokubo, unpublished observations. We designate it as TAND3, and it is located at the C-terminal side of TAND2 as indicated by brackets above the amino acid sequence. B, GAL4-dependent transcriptional activation in yeast. The expression plasmids described in A were transformed into two yeast strains, HF7c and SFY526, and the transcriptional activities were determined by measuring the HIS3 and lacZ reporter activities, respectively. Expression of the HIS3 gene was monitored by growth on media containing various concentrations of 3-AT (left panel). ␤-Galactosidase activity was calculated as a percentage relative to TAND3 (right panel). As a control, we also tested the GAL4 DNA binding domain alone (none), GAL4-TAND1 (6 -41 aa), and GAL4-VP16 (470 -490 aa). In both panels, rectangles corresponding to TAND3 (construct 26) are shaded .  FIG. 2. In vitro and in vivo characterization of TAND3. A, TAND3 forms a sub-stoichiometric complex with TBP when fused to TAND1 or TAND2. TAND1 (6 -40 aa) (lanes 12 and 13), TAND2 (43-71 aa) (lanes 14 and 15), and TAND3 (82-139 aa) (lanes 16 and 17) alone or in various combinations (lanes 4 -11) were expressed in E. coli as GST-tagged proteins. As a control, the GST portion by itself was also expressed (lanes 2 and ent region(s) of TAND3, we individually substituted 12 blocks of five (or three) consecutive amino acid residues with alanines and tested them for AD (Fig. 3A) and TBP binding activity (Fig.  3B). AD function was measured by ␤-galactosidase activity expressed from a lacZ reporter gene (Fig. 3A) although TBP binding activity was measured by quantifying the amount of TBP precipitated with GST-TAND(2ϩ3) in a GST pulldown assay (Fig. 3B). Generally, the effects of the mutations on both activities correlated closely (compare Fig. 3, A and B), although some inconsistencies existed, e.g. for region 97-101 aa. This is probably due to the lower sensitivity of the pulldown assay. Since residues within regions 97-111 and 127-131 aa were not required for AD function (Fig. 3A), there appears to be three distinct subregions, i.e. 82-96, 112-126, and 132-139 aa, that are involved in mediating these two activities.
No single substitution mutation could abolish either AD or TBP binding activity, even when it was introduced into the three critical subregions defined above. It is thus likely that each of these three subregions alone exhibits functional redundancy. To test this, we substituted a set of hydrophobic amino acid residues that are expressed in each subregion, designated as 3A (82-96 aa), 3B (112-126 aa), and 3C (132-139 aa), with alanines individually or in combination (Fig. 3C) and tested them for AD and TBP binding activity (Fig. 3, C and D, respectively). In these experiments, the hydrophobic residues were selectively mutated because it is well known that they are crucial for the AD function of activators (63, 64) and TAND1 (23). As expected, multiple substitutions of hydrophobic residues within each subregion reduced AD activity to much lower levels than those observed in Fig. 3A (Fig. 3C). Furthermore, any combination of these three multiple substitution mutations completely abolished AD activity (Fig. 3C, 3AB, 3AC, 3BC, and  3ABC). Hence, TAND3 appears to include two layers of functional redundancy, one within each subregion and the other between the subregions. This view is in accordance with the gradual loss of AD activity that we observed in the TAND3 deletion analysis (Fig. 1). Importantly, since AD and TBP binding activities were closely correlated as shown in Fig. 3, A and  B (Fig. 3, C and D), we concluded that these two functions must be carried out by the same or overlapping region(s) of TAND3.
The Same Set of Hydrophobic Residues Is Important Not Only for AD and TBP Binding Activities but Also for Cell Growth and Transcription-Our previous study (23) showed that, in case of TAND1, the same set of amino acid residues was required for AD and TBP binding activities as well as for cell growth. Thus, here we sought to determine whether the same set of hydrophobic residues that are required for the AD and TBP binding activities of TAND3 (Fig. 3, C and D) is also crucial for the other two activities of TAND3, namely supporting cell growth and transcription under specific conditions (Fig.  2, B and D).
To measure TAND3 activity in supporting cell growth, we used the ⌬TAND1 strain as a background (Fig. 2B). Thus, we individually combined each of the substitution mutations tested above (Fig. 3C) with the ⌬TAND1 mutation, and we compared their growth properties with those of the ⌬TAND1 and ⌬TAND(1ϩ3) mutations at 25 and 37°C (Fig. 4A). Unexpectedly, some mutants, such as ⌬TAND1ϩ3AB, 3AC, 3BC, or 3ABC, grew more slowly than the ⌬TAND(1ϩ3) mutant at 37°C. Since immunoblot analysis revealed that all of these TAF1 alleles were expressed at comparable levels (Fig. 4B), this is not simply due to the lowered expression of these Taf1p derivatives. Although the exact reason for the slower growth of these strains in comparison to the ⌬TAND(1, 3) mutant remains unclear, we suspect that it may be due to some dominant negative effects that are specifically enhanced at 37°C. In any case, a good correlation was observed among these mutations for the three activities of TAND3, i.e. AD, TBP binding, and growth at 37°C.
Next, to compare the TAND3 activities of these mutations in transcription, we individually combined each mutation with the ⌬TAND(1ϩ2) mutation and tested them for transcription of the TAND3-dependent CYC3 gene together with two control genes (ADH1 and PGK1) at 37°C (Fig. 4C). Interestingly, although transcription of control genes was not affected, that of the CYC3 gene was affected by these combinatorial mutations in a manner that was in accordance with the other three activities (compare Fig. 4C with Fig. 3, C and D, and 4A). Hence we concluded that all four functions must be carried out by the same or overlapping region(s) of TAND3. Thus, TBP binding appears to be required for mediating the other functions of TAND3.
We reasoned that if the growth and transcriptional defects in the ⌬TAND3 strain were simply due to the weakened interaction between Taf1p and TBP, they would be restored by increasing the expression of TBP. To explore this possibility, we examined whether various TAND-lacking strains that had been transformed with a TBP expression vector or its empty vector could grow at 37°C (Fig. 4D). The results clearly showed that the ectopic TBP expression restored the growth of all of the ⌬TAND strains we tested to levels comparable with those of the wild type strain (Fig. 4D). This is consistent with previous observations that the increased expression of TBP and/or TFIIA could rescue the growth defect of the ⌬TAND(1ϩ2) strain (15,16). We then also tested whether the expression of TBP could rescue the transcription of the TAND3-dependent CYC3 gene in the TAND3-defective mutants, such as the ⌬TAND(1ϩ2)ϩ3ABC and ⌬TAND(1ϩ2ϩ3) strains (Fig. 4E). Unlike its effect on cell growth, TBP expression increased the transcription of the CYC3 gene only slightly in these two mutants. Similar results were obtained for other TAND3-depend- 3). Bacterial cell lysates containing equimolar amounts of GST-tagged proteins or TBP were mixed together and then chromatographed on glutathione-Sepharose resin. Where indicated as "Ϫ," cell lysates expressing TBP were omitted from the reaction to test for specificity (evennumbered lanes). Purified TBP corresponding to 50% of input was run on the gel as a positional marker (lane 1). Complexes bound to the resin were eluted and resolved by SDS-PAGE and then stained with Coomassie Brilliant Blue (upper panel) or transferred to a nitrocellulose membrane and probed with anti-TBP polyclonal antibody (lower panel). The band corresponding to TBP is marked with a closed triangle on the right. Note that GST fusion proteins including the TAND3 portion generated a faint band with faster migration as denoted with a dot. This band is probably generated due to the presence of protease-hypersensitive sites within the region between 85 and 96 aa (23). B, growth comparison of several taf1 mutants at 25 and 37°C. Strains lacking the TAF1 gene but carrying one of several plasmids encoding Taf1p derivatives lacking various TAND subdomains were grown on YPD plates for 3 days at the indicated temperatures. C, immunoblot analysis of cell extracts prepared from the taf1 mutants 24 h after the temperature shift to 37°C. Taf1p derivatives were detected by using a monoclonal antibody against the HA epitope tag. Intact polypeptides of the expected size are marked with a dot. D, transcription analyses of the taf1 mutants. The expression of TAND(1ϩ2)-dependent (HIS4, CYC3, RPS5) and -independent (YDR119W, MET17, PGK1, ADH1) genes was compared in the indicated strains. Cultures were grown in YPD media to log phase at 25°C. A portion of each culture was shifted to 37°C, and incubation was continued for 2 h. Total RNA was isolated from these strains 2 h after the temperature shift to 37°C (lanes 4 -6) or after continuous incubation at 25°C over the same time ( lanes  1-3). The same amounts of total RNA were blotted onto a nylon membrane and hybridized with the probes indicated. The raw Northern blot data are shown in the left panel, whereas the quantified data are summarized in the right panel with total counts corresponding to each band being shown as a percentage relative to the strongest one of the 6 lanes. FIG. 3. Close correlation between the activator function and TBP binding activity of TAND3. A, schematic depiction of alaninescanning substitution mutants of TAND3 and their activator function. These mutants were expressed in yeast as fusions with the GAL4 DNA binding domain and tested for GAL4-dependent transcription, as described in Fig. 1. ␤-Galactosidase activity was calculated as a percentage ent genes, e.g. YDR119W and MET17 (data not shown). Thus, ectopically expressed TBP appears to restore transcription of only a portion of the TAND3-dependent genes whose function might be important for cell growth at 37°C, although it remains possible that greater levels of TBP expression may restore transcription of the CYC3, YDR119W, and MET17 genes. Taken together, these observations suggest that the role of TAND3 may be to regulate TBP function rather than simply to hold it inside TFIID.
TAND3 Can Partly Be Substituted for TAND1 in Cell Growth and Transcription-The experiments described above revealed that TAND3 is functionally similar to TAND1. Thus we next asked whether TAND1 could be replaced with TAND3 in supporting cell growth and transcription at higher temperatures (Fig. 5, A and C). Interestingly, although the ⌬TAND1 mutant was unable to grow at 37.5°C, the strain carrying chimeric TAND in which TAND1 was replaced with TAND3 could grow at 37.5°C, although less well than the wild type (Fig. 5A). Importantly, similar growth recovery was not observed for the strain carrying a mutated version of chimeric TAND in which TAND1 was replaced with inactive TAND3 (3ABC) (Fig. 5A). Immunoblot analysis confirmed that all TAF1 alleles were expressed at comparable levels (Fig. 5B). These results clearly indicate that with regard to its function in cell growth, TAND1 can be functionally replaced with TAND3, at least in part.
We also examined transcription of the CYC3, HIS4, and RPS5 genes along with control genes ADH1 and PGK1 at 37.5°C in the same set of strains used for the growth recovery assay (Fig. 5C). Transcription of the CYC3 gene in the ⌬TAND1 strain (Fig. 5C) was decreased to lower levels than those in the ⌬TAND(1ϩ2) strain (Figs. 2D and 4A). This is apparently due to differences in the experimental conditions used (i.e. incubation for 2.5 h at 37.5°C in Fig. 5C versus for 2 h at 37°C in Figs. 2D and 4A). Substitution of TAND1 with wild type TAND3, but not with inactive TAND3, could sustain the same level of CYC3 transcription, which suggests that TAND1 can be replaced with TAND3, at least for the transcription of this gene. A similar effect was observed with regard to the transcription of the RPS5 gene, although with a small difference being observed when active and inactive TAND3s were substituted. In contrast, TAND3 was a poor substitution for TAND1 in the transcription of the HIS4 gene. Taken together, we conclude that TAND1 and TAND3 are functionally similar but not fully interchangeable.
The Surface of TBP to Which TAND3 Binds Differs from That of TAND1 and Is Broader-Functional similarities between TAND1 and TAND3 indicate that these two molecules may serve as a similar regulatory domain for TFIID activation by binding to the same or overlapping surfaces of TBP. To examine this notion further, the GST-TAND(1ϩ2) and GST-TAND(2ϩ3) fusion proteins were tested for their binding to a set of TBP mutants in GST pulldown assays (Fig. 6A). The amount of mutant TBP precipitated with each GST fusion protein was quantitated and presented as a percentage relative to that of the wild type TBP (Fig. 6B). To define the respective binding surfaces of TBP for TAND1 and TAND3, we needed to compare its binding profiles to GST-TAND(1ϩ2) and GST-TAND(2ϩ3) because TAND1 or TAND3 alone cannot stably bind to TBP (Fig. 2A). We reasoned that residues where mutations reduced binding to TAND(1ϩ2), but not to TAND(2ϩ3), should be located on the surface of TBP recognized by TAND1 but not by TAND3 and vice versa. In this system, however, when mutations reduced the binding to both TAND(1ϩ2) and TAND(2ϩ3), it was not clear whether these residues were on the common surface recognized by both TAND1 and TAND3 or on the specific surface recognized by TAND2. To ensure simple interpretation of the data, we avoided testing TBP derivatives that carried mutations in the vicinity of helix 2 that is supposed to be a binding site for TAND2 (with the exception of Y139A).
The results are schematically summarized in Fig. 6C, where each residue is shown in different color code according to the extent of the damage induced. Whereas only two TBP mutants (i.e. L114K and V161A) were severely affected (i.e. less than 20% of the wild type) in TAND(1ϩ2) binding, eight TBP mutants (i.e. N91E, E93R, L114K, Y139A, F148H, V161A, V198A, and Q219V) were severely affected in TAND(2ϩ3) binding. None of the TBP mutants that we examined were specifically affected in TAND(1ϩ2) binding. L114K and V161A, colored in green, were critically important for binding to both TAND(1ϩ2) and TAND(2ϩ3). They are located on the concave surface of TBP, which is purported to be a binding surface for TAND1 but not for TAND2. This indicates that TAND3 may also bind to the concave surface of TBP. Mutations of the other two residues, i.e. S118L and F155S, colored in cyan, were moderately affected (20 -60% of the wild type) in binding to both TAND(1ϩ2) and TAND(2ϩ3). Since S118L and F155S are located close to, but not exactly on, the concave and convex surfaces of TBP, respectively, it remains unclear whether these mutations impair TAND2 or TAND1/TAND3 binding. In contrast, mutations of six residues, i.e. N91E, E93R, Y139A, F148H, V198A, and Q219V, colored in red, were severely and specifically affected in TAND(2ϩ3) binding, indicating that these residues must be recognized by TAND3 but not by TAND1 or TAND2. Somewhat surprisingly, they are not localized to any small region but rather are dispersed over the surface of TBP. Mutations of five residues, i.e. N69S, T153I, N159D/N159L, E186R, L189E/L189K, colored in magenta, were moderately affected in TAND(2ϩ3) binding but only slightly (60 -100% of the wild type) in TAND(1ϩ2) binding. We propose that all of these residues may also be recognized by TAND3 because they are not located on the convex (i.e. TAND2 binding) surface of TBP (e.g. Asn-69, Asn-159, Glu-186, and Leu-189) and/or because their mutations have little effect on TAND(1ϩ2) binding (e.g. Thr-153, Asn-159, Glu-186, and Leu-189).
Collectively, the residues colored in green, red, and magenta form a possible binding surface for TAND3. It appears to be much broader than and distinct from that for TAND1, which makes contact only throughout the concave surface of TBP (19,55). This view is in good agreement with the observation that relative to wild type TAND3. B, TBP binding activity of alanine-scanning substitution mutants of TAND3. The same set of TAND3 mutants described in A was fused to TAND2 and expressed in E. coli as GST-tagged proteins (lanes 4 -16). As a control, the GST portion by itself (lane 2), TAND(1ϩ2ϩ3) (lane 3), TAND2 (lanes 17), and TAND3 (lane 18) were also expressed in E. coli as GST-tagged proteins. The interaction of these proteins with TBP was examined as described in Fig. 2A, except that the gel was stained with SYPRO Ruby fluorescence dye. The positions of TBP and the processed GST fusion proteins are indicated with a closed triangle and a dot, respectively, as described in Fig. 2A. The amount of precipitated TBP in each lane was normalized against that of GST fusion proteins trapped on the beads and was then calculated as a percentage relative to wild type TAND3 shown in the right panel. C, schematic depiction of alanine substitution mutants of hydrophobic residues clustered into three distinct subregions of TAND3 and their activator function. These mutants were expressed in yeast as fusions with the GAL4 DNA binding domain and tested for GAL4-dependent transcription, as described in A. ␤-Galactosidase activity was calculated as a percentage relative to wild type TAND3. D, TBP binding activity of alanine substitution mutants of the three TAND3 subregions. The same set of TAND3 mutants described in C was fused to TAND2, and these were expressed in E. coli as GST-tagged proteins. The interaction of these proteins and several control proteins with TBP was examined and quantitated as described in B.
FIG. 4. Close correlation between the activity of TAND3 in supporting growth and transcription at higher temperatures and the effect of increased expression of TBP on the growth and/or transcriptional defects of taf1⌬TAND mutants. A, growth comparison of several taf1⌬TAND mutants at 25 and 37°C. Four different dilutions of the strains expressing the indicated Taf1p derivatives were spotted onto the molecular size of TAND3 is almost twice as large as that of TAND1 and that TAND3 is composed of three functionally redundant subregions (denoted here as 3A, 3B, and 3C) (Fig.  3C). Although it still remains elusive how each subregion binds to TBP, it is notable that the concave surface of TBP appears to be bound by both TAND1 and TAND3. Consistent with this hypothesis is that TAND(2ϩ3) can inhibit TBP binding to the TATA element, albeit less strongly than TAND(1ϩ2) (data not shown). The functional significance of these interactions with TBP will be discussed below. DISCUSSION In this study, we identified a novel TBP binding region in the region spanning 82-139 aa of yeast Taf1p and designated it as TAND3. In several functional aspects, TAND3 is similar to TAND1 although they do not show significant sequence similarities. First, TAND1 and TAND3 can stably bind TBP only when fused to TAND2. Second, they both activate transcription when tethered to DNA. Third, at higher temperatures, they are required for supporting normal growth and the transcription of a subset of genes. Importantly, any mutations that reduce TBP binding impaired all other activities of TAND1 and TAND3. Thus, we postulate that regulatory interactions between TAND1/TAND3 and TBP may underlie all the other apparently distinct functions carried out by these two TAND domains.
Interactions between TAND1/TAND3 and TBP-Regardless of these functional similarities, TAND1 and TAND3 are not fully exchangeable. For instance, at higher temperatures, a mutant strain in which TAND1 was replaced by TAND3 grew much faster than the ⌬TAND1 strain, but slower than the wild type (Fig. 5A). Furthermore, TAND(1ϩ2) forms a stoichiometric complex with TBP whereas TAND(2ϩ3) does not ( Fig. 2A). Moreover, although TAND1 appears to bind only to the concave surface of TBP, TAND3 can make contact with a much broader region of TBP that includes the concave surface (Fig. 6). This binding occurs via hydrophobic residues clustered at three distinct subregions, 3A, 3B, and 3C (Fig. 3C). The observation that the same mutations, i.e. L114K and V161A, on the concave surface of TBP severely impaired interactions with both TAND1 and TAND3 may suggest that these two TAND subdomains competitively bind to TBP. However, it is difficult to confirm this notion experimentally since TAND1 and TAND3 alone are unable to stably bind TBP (data not shown). Unfortunately, TAND(1ϩ2) and TAND(2ϩ3) cannot be used for this purpose since both proteins include TAND2.
The surface of TBP to which TAND3 binds seems to overlap with those for several other important factors that regulate TBP function. For example, the Asn-91 and Glu-93 residues, both of which are specifically required for TAND3 binding (Fig.  6), are located on the TFIIA interface (65)(66)(67). The Val-198 residue (corresponding to Ile-296 in human TBP) appears to be crucial for the interaction with NC2␤ (35,68). However, mutation of the Ala-86 and Tyr-195 (human Tyr-293) residues did not affect the interaction with TAND3, although they are crucial for the interaction with TFIIA and NC2␤, respectively (Fig. 6). These observations suggest that the binding site for TAND3 may overlap with but is not the same as those for TFIIA and NC2␤.
The concave surface of TBP can bind to multiple factors, including TAND1 (15,18), TAND3, Mot1 (28,69), and TBP itself (19), in addition to the TATA element. TBP dimerization in vivo is assumed to result in a repressed but potentially active pool of TBP that is in a form that is resistant to proteolytic degradation (19,40). In contrast, TAND1 and Mot1 seem to be more directly involved in transcription. Based on the observation that TAND1 and AD are interchangeable in several assays, we previously proposed a two-step handoff model in which TAND1 may maintain TBP in an inactive state until it encounters AD, which dissociates TAND1 from TBP and allows activated TBP to bind to the TATA element. Our findings that TAND3 shares several fundamental features with TAND1 raise an intriguing possibility that AD may target not only TAND1, but also TAND3 as a first step in the triggering of TFIID activation. In this regard, it is interesting that several residues of TBP (e.g. Asn-91, Glu-93, Tyr-139, Phe-148, Thr-153, Asn-159, Glu-186, and Leu-189) that are postulated to be involved in transcriptional activation (67, 70 -73) exhibited TAND3-specific binding defects (Fig. 6B). Furthermore, the previous observation that two molecules of GAL4 AD bind simultaneously to one molecule of TBP (74) supports our view that AD may target both TAND subdomains, i.e. TAND1 and TAND3, bound on TBP.
Functional Relationships between Yeast TAND1/TAND3 and Drosophila TAND1-Drosophila TAND1 was shown to bind stably to TBP by mimicking the minor groove surface of the partially unwound TATA element (55). Unlike its yeast counterpart, it activates transcription only slightly when tethered to DNA (23). The differences in size and function between yeast TAND1 (10 -37 aa) and Drosophila TAND1 (18 -77 aa) suggest that TAND3 (82-139 aa) may correspond to a missing piece of some functional domain that is specifically included in Drosophila TAND1. This view is consistent with the observation that the Glu-186 and Leu-189 residues on the stirrup region, which provides an interaction surface for TFIIB, are also involved in binding to Drosophila TAND1 (55) and yeast TAND3 ( Fig. 6) but not to yeast TAND1 (Fig. 6) (19). In this regard, it is notable that a small segment of yeast TAND3, which includes part of the subregion 3B, displays significant sequence similarity to a piece of Drosophila TAND1 that interfaces with the stirrup region of TBP (55) (Fig. 6D). Thus, hydrophobic amino acid residues like Ile-112 and/or Leu-116 of TAND3 (Fig.  3C) may interact with Glu-186 and/or Leu-189 of TBP.
On the other hand, NMR studies revealed that Drosophila TAND (1-156 aa), which includes TAND1 (18 -77 aa), TAND2 (118 -143 aa), and the spacer region (78 -117 aa) connecting these two TAND subdomains, interacts not only with the concave surface (i.e. the binding site for TAND1) but also with the TFIIA interface of TBP (75). As TAND2 is supposed to bind to helix 2, the remaining TFIIA interface should be in contact YPD plates and incubated for 3 days at the indicated temperatures. Note that the strains bracketed on the left carry the same set of alanine substitution mutations as described in Fig. 3C in addition to the TAND1 deletion (⌬1). As a control, we also tested the growth of a strain lacking TAND1 and TAND3 (⌬(1ϩ3)). B, immunoblot analysis of several taf1⌬TAND mutants recovered 24 h after the temperature shift to 37°C. Taf1p derivatives were detected and marked as described in Fig. 2C. C, transcription analyses of several taf1⌬TAND mutants at 37°C. The expression of the CYC3, ADH1, and PGK1 genes in the indicated strains was examined and quantitated as described in Fig. 2D. Note that the strains bracketed at the top (lanes 2-9) carry the same set of alanine substitution mutations as described in A in addition to the deletion of both TAND1 and TAND2 (⌬(1ϩ2)). As a control, we also tested the expression of these three genes in the wild type strain (lane 1) and the mutant strain lacking all TAND subdomains (⌬(1ϩ2ϩ3)). D, effect of increased expression of TBP on the growth of several taf1⌬TAND mutants at 25 and 37°C. Strains expressing the indicated Taf1p derivatives that had been transformed with a TBP expression vector (right semicircle) or an empty vector pRS316 as a control (left semicircle) were grown on S.D. plates for 3 days at the indicated temperatures. E, effect of increased expression of TBP on transcription of several taf1⌬TAND mutants at 37°C. The expression of the CYC3 and ADH1 genes in the indicated strains that had been transformed with a TBP expression vector (lanes 5-8) or an empty vector pRS316 as a control (lanes 1-4) was examined and quantitated as described in Fig. 2D. with the spacer region. Thus, it is likely that the spacer region may make contact with the Asn-91 and Phe-148 residues (75), both of which are also essential for TAND3 binding (Fig. 6). Furthermore, the spacer region can activate transcription, albeit less weakly than TAND3, when tethered to DNA. 3 Taken together, it appears that at least some subregions of TAND3 might serve as a functional counterpart of the spacer region of Drosophila TAND. We must conduct further analyses to determine the precise relationships between the three subregions of yeast TAND3 and TAND1/the spacer region of Drosophila TAND.
Functional Redundancy between TAND1 and TAND3-The effects of the TAND1 deletion (⌬TAND1) on various phenotypic characteristics were more obvious than those of the TAND3 deletion (⌬TAND3). For instance, the ⌬TAND1 strain showed a temperature-sensitive growth defect, whereas the ⌬TAND3 3 H. Ohta and T. Kokubo, unpublished observations. FIG. 5. TAND3 can partly substitute TAND1. A, growth comparison of several taf1 mutants at 25 and 37.5°C. Strains expressing the Taf1p derivatives as indicated, e.g. in which TAND1 is replaced with wild type or mutated TAND3, lacking TAND1 or TAND(1ϩ2ϩ3), were grown on YPD plates for 3 days at the indicated temperatures. B, immunoblot analysis of several taf1 mutants recovered 24 h after the temperature shift to 37.5°C. Taf1p derivatives were detected and marked as described in Fig. 2C. C, transcription analyses of several taf1 mutants at 37.5°C. The expression of the CYC3, HIS4, RPS5, ADH1, and PGK1 genes in the same set of strains described in A was examined and quantitated as described in Fig. 2D, except that the cells were shifted to 37.5°C and harvested 2.5 h after the temperature shift. strain did not (Fig. 2B). Moreover, the effects of ⌬TAND3 on growth could only be detected when the ⌬TAND1 and ⌬TAND(1ϩ3) strains were carefully compared (Fig. 2B). Similarly, the effects of ⌬TAND3 on transcription were observed only when the ⌬TAND(1ϩ2) and ⌬TAND(1ϩ2ϩ3) strains were carefully compared (Fig. 2D). These observations indicate that TAND1 might be able to compensate for the loss of TAND3 or that TAND3 is simply less important than TAND1 for the in vivo functioning of TFIID. We favor the former possibility since TAND3 can at least partially suppress the growth and transcriptional defects in the ⌬TAND1 strain (Fig. 5). Moreover, TAND1 could rescue the TAND3-dependent transcription of the CYC3 gene in the ⌬TAND(1ϩ2ϩ3) strain. 4 It is notable that TAND3 is demarcated with two clusters of multiple methionine residues, i.e. Met-82, -83, and -5 and Met-137 and -139. Recently, we found that Taf1p can be produced in vivo in several isoforms by alternative selection of translational initiation sites. 5 In other words, Taf1p lacking TAND(1ϩ2) or TAND(1ϩ2ϩ3) is generated in small amounts by initiating translation at Met-82, -83, and -85 or Met-137 and -139, respectively. Hence the observation that Taf1p⌬TAND(1ϩ2) and Taf1p⌬TAND(1ϩ2ϩ3) are functionally distinct (Fig. 2) should have a physiological relevance to the in vivo mechanisms involved in generating these Taf1p derivatives. However, we failed to see any dominant negative effects of these Taf1p derivatives on cell growth at 25 and 37°C when they were expressed at severalfold higher levels than that of the wild type Taf1p. 4 As Taf1p has been shown to carry multiple binding sites for TBP (12,13), future studies will be needed to determine whether TAND3 corresponds to one of those sites and to reveal the functional role of each contact site. FIG. 6. Interactions between TBP and TAND(2؉3) or TAND(1؉2). A, GST pulldown assays were conducted as described in Fig. 2A for bacterial cell lysates expressing the indicated TBP mutants, GST-tagged TAND(1ϩ2), or TAND(2ϩ3) proteins. SYPRO Ruby stained gels are presented. The migration positions of TBP vary somewhat depending on the chemical features of their mutated side chains. B, the data shown in A were quantitated and presented as described in Fig. 3B. C, the binding properties of mutated residues to the GST-tagged TAND(1ϩ2) and TAND(2ϩ3) proteins are summarized as space filling drawings of yeast TBP that are viewed from four different angles. Mutations that severely reduce TBP binding to both TAND(1ϩ2) and TAND(2ϩ3) proteins (less than 20% of the wild type) are shown in green, whereas those that moderately affect TBP binding to both (20 -60% of the wild type) are shown in cyan. Similarly, those that severely affect TAND(2ϩ3) binding but TAND(1ϩ2) binding only slightly (60 -100% of the wild type) are shown in red, whereas those that moderately affect TAND(2ϩ3) binding but only slightly affect TAND(1ϩ2) binding are shown in magenta. Finally, those that only slightly affect both bindings are shown in yellow. D, a conserved small segment between yeast TAND3 and Drosophila TAND1. Identical residues are shown in red and are connected by broken lines. The region between 52 and 58 aa of Drosophila TAND1 that was shown to form an ␣-helix (55) and that corresponds to the subregion 3B of yeast TAND3 is underlined.