The Stability of the TFIIA-TBP-DNA Complex Is Dependent on the Sequence of the TATAAA Element*

To determine the mechanistic differences between canonical and noncanonical TATA elements, we compared the functional activity of two sequences: TATAAA (ca-nonical) and CATAAA (noncanonical). The TATAAA element can support high levels of transcription in vivo , whereas the CATAAA element is severely defective for this function. This dramatic functional difference is not likely to be due to a difference in TBP (TATA-binding protein) binding efficiency because protein-DNA complex studies in vitro indicate little difference between the two DNA sequences in the formation and stability of the TBP-DNA complex. In addition, the binding and stability of the TFIIB-TBP-DNA complex is similar for the two elements. In striking contrast, the TFIIA-TBP-DNA complex is significantly less stable on the CATAAA element when compared with the TATAAA element. A role for TFIIA in distinguishing between TATAAA and CATAAA in vivo was tested by fusing a subunit of TFIIA to TBP. We found that fusion of TFIIA to TBP dramatically increases transcription from CATAAA in yeast cells. Taken together, these results indicate that the stability of the TFIIA-TBP complex depends strongly on the sequence of the core promoter element and that the TFIIA-TBP complex plays an important function in rec-ognizing optimal promoters in vivo .

Initiation of mRNA synthesis by RNA polymerase II is the major step of regulation of eukaryotic gene expression and occurs at core promoters that typically consist of a TATA box and initiator element (1). The first step in promoter recognition is binding of TBP (TATA-binding protein) to the TATAAA element (reviewed in Ref. 2). Recruitment of TBP is a ratelimiting step in transcriptional initiation in vivo at a majority of promoters (3)(4)(5), and promoter occupancy of TBP correlates very well with transcriptional activity (6,7). The TBP-TATA interaction sets the stage for the nucleation of the remainder of the preinitiation complex, which includes TFIIA, TFIIB, TFIIE, TFIIF, TFIIH, TFIIJ, and RNA polymerase II (reviewed in Ref. 8). TFIIA and TFIIB have each been shown to make direct contacts with TBP and DNA and, in so doing, stabilize the TBP-DNA complex (reviewed in Ref. 9). In addition, TFIIA and TFIIB promoter occupancy is coordinated with that of TBP (10); thus interactions between the DNA, TBP, TFIIA, and TFIIB are likely to play an important role in a majority of promoters in vivo.
Despite the apparent requirement of the TATAAA sequence in the initiation of RNA polymerase II transcription, many natural eukaryotic core promoters lack recognizable TATA elements but still require TBP for transcription initiation (11)(12)(13). This noncanonical or "TATA-less" subclass includes the promoters of several different types of genes, including the ubiquitously expressed "housekeeping" genes, developmentally regulated genes (growth factors and growth factor receptors), and many oncogenes (14 -16). In addition, several instances of multiple, yet functionally distinct elements driving the expression of a single gene have been described in yeast (17)(18)(19). For example, the yeast HIS3 gene contains two distinct promoter elements: a canonical TATAAA element (T R ) that initiates from ϩ13 and a second element (T C ) that lacks the canonical TATAAA sequence, extends over ϳ30 base pairs, and initiates from ϩ1 (20). The two tandem yeast HIS3 promoter elements are utilized differentially depending on the overall levels of HIS3 production (19,21). At low levels of overall transcription, both T C and T R are utilized with approximately equal efficiencies. However, at high levels of transcription, expression from T C increases only slightly, whereas expression from T R shows a large and dramatic increase. Thus, the noncanonical T C element saturates at much lower levels of expression than does the canonical TATA element.
The in vivo regulation observed for T C can be mimicked by substituting T C with CATAAA, a sequence with a single base pair change of the first thymine in the TATAAA element to a cytosine (22). Because CATAAA can functionally replace T C and CATAAA serves as a simpler model than T C for noncanonical core promoters, we utilized TATAAA and CATAAA to study the mechanistic differences between canonical and noncanonical elements in vitro and in vivo. We find that the TBP-DNA and the TFIIB-TBP-DNA complexes behave very similarly on TATAAA and CATAAA. In contrast, the TFIIA-TBP complex is much less stable on the CATAAA element compared with the TATAAA element. Transcriptional analysis indicates that the CATAAA element is severely defective for supporting high levels of transcription in vivo, but a TFIIA-TBP fusion molecule can restore activity from this element. These results suggest that the stability of the TFIIA-TBP-DNA complex is sensitive to the sequence of the core promoter element and that the TFIIA-TBP complex plays an important role in core promoter recognition and the level of transcriptional expression in vivo.
The DNA elements used for the in vivo transcription analysis and the in vitro protein DNA interaction studies were 23-base pair oligonucleotides designed from sequences at the promoter of the yeast HIS3 gene. The TATAAA oligo contains a core promoter element identical to the HIS3 T R with the sequence 5Ј-AATTCCTATAAAGTAATGTGGAG-3Ј. The CATAAA oligo is identical to TATAAA except that the first thymine in the core element is substituted with a cytosine (5Ј-AATTCCCATA-AAGTAATGTGGAG-3Ј). To make double stranded probe, a 19-base pair complementary oligonucleotide was synthesized and annealed to the 23-base pair oligo described above leaving an AT-rich overhang on the 5Ј end of the probe. Klenow enzyme was used to fill in the singlestranded overhang with 25 Ci of [␣-32 P]dATP in the presence of 2.0 M dTTP.
The plasmids used for the in vivo transcription analysis were derivatives of YCp86 containing the 2-kilobase fragment from pUC18 including the bla gene and the origin of replication, a 1.9-kilobase fragment of yeast DNA containing cen3 and ars1, a 1.1-kilobase fragment containing the URA3 gene, and a polylinker. The hybrid HIS3 promoterYCp86-SC3801 (24,25) has the wild type initiation and amino-terminal region of the HIS3 gene fused in frame with a functional Escherichia coli LacZ gene and is used to detect levels of transcription in vivo. The promoter region contains a 365-base pair GAL1,10 fragment containing four GAL-4 binding sites fused upstream of the EcoRI-SacI restriction endonuclease sites, between which the TATAAA and CATAAA oligonucleotides (described above) could be inserted. When the two oligos were cloned into this molecule it was renamed pJS3801 (TATAAA) and pJS3803 (CATAAA), respectively.
Transcriptional Analysis-For the experiments measuring activated transcription in vivo, the pJS3801 and pJS3803 constructs (described above) were transformed into yeast strain yJS156. The resulting transformants were used in liquid assays for ␤-galactosidase activity. The cells were grown in selective medium containing either 2% raffinose or 2% galactose to mid-exponential growth phase (A 600 Х 1-2). Enzyme activities were determined in triplicate and normalized to the A 600 of the cultures.
Quantitative RNA analysis was done by S1 nuclease digestion of ϳ50 g of RNA prepared from cells grown in synthetic complete medium to an A 600 of 0.6 (26,27). Total RNA was prepared by hot phenol extraction and hybridized with 100-fold excess of radiolabeled single-stranded HIS3, DED1, and tRNA w probes and subjected to S1 nuclease digestion. The RNA amounts used in each reaction were normalized to the RNA levels obtained from a probe to the intron of the tryptophan tRNA gene (tRNA W ).
Protein Expression and Purification-Untagged, full-length yeast TBP was expressed in E. coli strain BL21 (DE3) with a pET11a vector (called pYTBP) as described (28). The soluble fraction was purified over Q, SP, and Heparin HiTrap columns (Amersham Pharmacia Biotech) equilibrated in buffer A (10 mM sodium phosphate, 100 mM KCl, 5 mM MgCl 2 , 10% (v/v) glycerol, 1 mM dithiothreitol, pH adjusted to 7.5). TBP flowed through the Q column and was eluted from the SP and heparin columns using linear NaCl gradients in buffer A. Final purification of TBP was by gel filtration with Sephacryl S-100 equilibrated in 100 mM KCl, 40 mM HEPES (pH 7.9), 20 mM Tris (pH 7.5), 10% glycerol, 1 mM dithiothreitol, and 0.5 mM phenylmethylsulfonyl fluoride.
TFIIB was cloned into pET-15b for expression and purification of His-tagged protein from E. coli BL21DE3/PlysS (Novagen). The cells were grown in LB medium at 37°C to an optical density of 0.7. Isopropyl-␤-D-thiogalactopyranoside (final concentration, 0.1 mM) was added, and the cells were incubated for 2 h at 30°C. The cells were harvested by centrifugation and washed with a 20 mM Tris, 50 mM NaCl buffer. Following sonication, the lysate was incubated with shaking at 4°C with 1 ml of His-bind resin (Novagen). After two consecutive wash steps with buffers containing 20 mM imidizole and 80 mM imidizole, respectively, the His-tagged protein was eluted with buffer containing 200 mM imidizole. The eluate was dialyzed against 100 mM KCl, 40 mM HEPES (pH 7.9), 20 mM Tris (pH 7.5), 10% glycerol, 1 mM dithiothreitol, and 0.5 mM phenylmethylsulfonyl flouride (EMSA binding buffer without MgCl 2 ). 1 The protein was shown to be 80% pure upon staining with Coomassie Blue.
Recombinant yeast TFIIA was purified as described (29). This procedure involves expressing each subunit, Toa1 and Toa2, in separate strains of E. coli BL21DE3. Isopropyl-␤-D-thiogalactopyranoside (final concentration, 0.1 mM) was added at an optical density of 0.6, and after the cells were ruptured by sonication, insoluble material was collected by centrifugation. The proteins were then denatured in a buffer containing 8 M urea and renatured in the presence of the other subunit. The renatured proteins were dialyzed against 100 mM KCl, 40 mM HEPES (pH 7.9), 20 mM Tris (pH 7.5), 10% glycerol, 1 mM dithiothreitol, and 0.5 mM phenylmethylsulfonyl flouride (EMSA binding buffer without MgCl 2 ). The TFIIA was ϳ60% pure as determined by Coomassie staining.
DNA Binding Studies-Protein-DNA interactions in vitro were studied by incubation of purified proteins with 32 P-labeled TATAAA and CATAAA probes. Binding reactions contained 10 M poly(dG-dC) nonspecific competitor, 100 mM KCl, 40 mM HEPES (pH 7.9), 20 mM Tris (pH 7.5), 5 mM MgCl 2 , 10% glycerol, 1 mM dithiothreitol, and 0.5 mM phenylmethylsulfonyl flouride. 32 P-Labeled TATAAA and CATAAA probes (2.4 pmol) were incubated with 13.8 nM TBP at 25°C for the indicated amount of time. For EMSA the complexes were separated on 5% acrylamide gel containing 0.5ϫ Tris borate-EDTA and 2 mM MgCl 2 in both the gel and running buffer. Recombinant yeast TFIIA (5.0 nM) and His-tagged TFIIB (8.5 nM) were isolated and incubated with TBP and probe DNA and treated as described above, except that MgCl 2 was omitted from the gel and running buffer in the experiments shown. Omission of the divalent cations from the binding reaction allows for the stable formation of only the ternary complex, which makes quantitation simpler because there is only a single shifted complex. Assays of higher order complex formation were also performed in the presence of MgCl 2 (data not shown). Inclusion of MgCl 2 did not alter the results of the experiments. The gels were transferred to Whatman 3MM paper, dried under vacuum at 80°C, and subjected to PhosphorImager screen or autoradiography at Ϫ70°C with an intensifying screen.
Quantitative Analysis of DNA Binding-The concentration of active protein was determined by DNA titration in which the reaction mixtures contained known concentrations of radiolabeled probe and a fixed amount of protein (30,31), and the fraction of DNA bound was determined. By this analysis, the TBP preparation used for most experiments shown in this work was about 6% active, consistent with 10% activity reported by others for yeast TBP preparations (30). All TBP concentrations given in the text refer to the concentration of active protein.
For measuring association rates, the reactions contained 2.4 pmol of probe DNA and 13.8 nM TBP (or 13.8 nM TBP and 5.0 nM TFIIA) and were incubated at 25°C in a final reaction volume of 360 l. At the indicated times, the extent of complex formation was analyzed by EMSA by loading 10 l of the binding reaction on to the gel. For filter binding assays, the samples were applied to a pretreated nitrocellulose membrane via a dot blot apparatus exactly as described (32). After sample application, each well was washed with 100 l of binding buffer (100 mM KCl, 40 mM HEPES (pH 7.9), 20 mM Tris (pH 7.5), 5 mM MgCl 2 , and 10% glycerol), and the filters were air dried and subjected to a PhosphorImager screen for quantitation. For both the EMSA and filter binding assays, the data are presented as the fraction bound at time (t). Fraction bound is calculated from the ratio B/B 30

Mechanistic Differences between Canonical and Noncanonical Elements May
Depend on TFIIA Activity-Recent work characterizing a TBP allele defective for interaction with TFIIA (the N2-1 derivative of TBP) has shown that the TFIIA-TBP interaction is important for HIS3 core element utilization (33). When the amount of RNA initiating from ϩ1 (which is driven by the noncanonical element, T C ) and ϩ13 (driven by TATA) start sites were compared, the N2-1 strain showed a preferential increase in transcription initiating from ϩ13, even when the overall level of transcription was low (33). This is in stark contrast to the utilization of the HIS3 promoter elements occurring in wild type cells, which shows similar levels of output (22). These results suggest that the loss of the TFIIA-TBP  a The parent yeast strain yJS156 has the chromosomal copy of Spt15 deleted. TBP functions are provided by a URA3-marked plasmid containing the TBP gene. Plasmid shuffling on 5-fluoroorotic acid was used to create the TRP1-marked wild type TBP or the Toa2-TBP and TFIIB-TBP fusion strains assayed.
b Strains were transformed with a URA3-marked HIS3/Lac Z fusion plasmid containing either the canonical TATAAA element (pJS3801) or the noncanonical CATAAA element (pJS3803), driven by the GAL1 UAS. c ␤-Galactosidase activities were performed with 10 7 cells cultured in a medium containing either 2% raffinose or 2% galactose for the indicated amount of time.
d Incubation in the presence of raffinose was the base line because this carbon source does not stimulate activation in this system but rather derepresses the effect of glucose.
interaction is having a more dramatic effect on the noncanonical element (T C ) than on the canonical element (33). To further test the importance of the TFIIA-TBP interaction at the HIS3 promoter, we examined the effect of introducing a fusion construct of a subunit of TFIIA (Toa2) to TBP (34). The idea is that fusion of TFIIA to TBP increases the effective concentration of TFIIA at the promoter. We compared the amount of RNA initiating from the ϩ1 start site (T C ) and the ϩ13 start site (TATA) in strains containing either Toa2-TBP or wild type TBP. Interestingly, the amount of transcription initiating at ϩ1 was 10-fold higher in the Toa2-TBP strain compared with the TBP strain (Fig. 1). In contrast, the amount of transcription initiating at ϩ13 is essentially the same in the two strains. Thus, fusion of TFIIA to TBP causes a preferential increase in transcription from a natural, noncanonical element, indicating that transcription from this class of elements is more sensitive to perturbations in the TFIIA-TBP interaction than is a consensus TATA element.
The CATAAA Element Is Unable to Support a High Level of Transcription in Vivo-Because the CATAAA sequence can functionally replace T C (22) and CATAAA is a simpler model for noncanonical promoters, we compared the functional activities of CATAAA and TATAAA. The TATAAA and CATAAA elements were cloned individually into the core promoter region of a reporter plasmid containing the GAL1,10 UAS upstream of the core promoter and the wild type HIS3 initiation region driving the expression of a HIS3-LACZ fusion (24,25).
Thus, the only difference between the two constructs is in the first position of the TATAAA box sequence where TATAAA contains a thymine and CATAAA contains a cytosine. These constructs were used to measure the response to Gal4, a potent acidic activator that stimulates transcription in the presence of galactose.
Under noninducing conditions (raffinose) both TATAAA and CATAAA exhibited similar levels of ␤-galactosidase activity (Table I). In contrast, incubation in galactose medium resulted in a 50-fold increase in activity from the TATAAA-driven reporter, whereas the CATAAA-driven reporter did not show a significant increase in activity over that of cells grown under noninducing conditions. Thus, like T C , the CATAAA element is greatly diminished in its ability to support high levels of transcription in vivo.
Fusion of TBP and TFIIA Results in an Increase in Expression from CATAAA in Vivo-If CATAAA serves as a good model for T C (a noncanonical TATA element), then one would predict that the sensitivity to changes in TFIIA activity would be similar. To test this hypothesis, we again utilized the Toa2-TBP fusion construct. The response to Gal4 from TATAAA or CATAAA in the strain containing the Toa2-TBP fusion as the sole source of TBP was measured using the ␤-galactosidase assay. Incubation in galactose medium results in a 50-fold increase in activity in either the wild type TBP or the Toa2-TBP fusion strain harboring the TATAAA driven reporter (Table I). Whereas activity from the TATAAA element was unaffected by the Toa2-TBP fusion, the CATAAA-driven reporter showed a dramatic increase (15-fold) in activity in the fusion strain over the TBP strain containing the CATAAA element. In addition, this increase in activity from CATAAA is specific for TFIIA because no such increase is observed in a TFIIB-TBP fusion strain. Thus, CATAAA and T C are up-regulated similarly in the Toa2-TBP background.
TBP Binding Affinity and Stability Are Similar on the TATAAA and CATAAA Elements-Because TBP is the sequence-specific factor that recognizes the TATA element, it seemed likely that differences in TATAAA and CATAAA element activity might be due to insufficient binding of TBP or instability of the TBP-CATAAA complex. To test this hypothesis, the relative binding and stability of TBP on the TATAAA or the CATAAA element was measured. The two 33-base pair sequences used in this assay differ only in position 11, where TATAAA contains a thymine and CATAAA contains a cytosine.
The rate of formation of the TBP-DNA complex on TATAAA and CATAAA was measured by monitoring complex formation over increasing time using the electrophoretic mobility shift assay (Fig. 2, A-C) and filter binding assays (Fig. 2D). The observed kinetics of TBP binding to the TATAAA and CATAAA sequences appears to follow first order kinetics because the data were fitted by single-exponential curves (Fig. 2, C and D). The curves fit to the data obtained from the shift assay corresponded to a half-time association (t1 ⁄2 ) of 1.2 min for TATAAA and 3.0 min for CATAAA. Similar rates were obtained using the filter binding assay: 0.9 min for TATA and 2.1 min for CATA. The rates obtained for the TBP-TATA association are consistent with previously obtained rates for the formation of this complex (30). Thus, both assays reveal a slightly faster formation of the TBP-TATA complex as compared with the TBP-CATA complex at early time points. However, the absolute amount of complex formed at 30 min is similar for both TATAAA and CATAAA.
We next examined the stability of the TBP-DNA complex by determining the dissociation kinetics of the two complexes (Fig.  3). TBP-DNA complexes were formed and then challenged with a large excess of an adenine-thymine rich sequence, poly(dA-dT). Poly(dA-dT) is used in these competition assays as specific competitor because the sequence of alternating adenine and thymine closely resembles a TATAAA element. TBP was incubated with the DNA for 30 min, because this results in a similar amount of TBP-DNA complex formed on TATAAA and CATAAA (Figs. 2 and 3), followed by the addition of 1000-fold molar excess of poly(dA-dT). TBP behaves similarly on both elements with a t1 ⁄2 of 40 min. DNase I footprinting experiments confirmed that TBP was specifically protecting the TATAAA and CATAAA elements (data not shown). Thus, TBP binds with equal stability on both the TATAAA and CATAAA elements. Could the 100-fold difference in activity in vivo be due to the slight difference in early binding kinetics of TBP to TATAAA and CATAAA? Possibly, but we next examined higher order complex formation to determine whether more dramatic differences could be observed for the two elements.
The TFIIA-TBP Complex Is More Stable on the TATAAA Element Compared with CATAAA-Because the binding and stability of TBP was not dramatically different on the two elements, we compared the two sequences for their ability to form TFIIA-TBP-DNA complexes. Because the fusion of TFIIA to TBP resulted in elevated levels of transcription from a noncanonical element in vivo ( Fig. 1 and Table I), either the formation or stability of the TFIIA-TBP-DNA complex may be sensitive to the sequence of the TATAAA element.
The rate of formation of the TFIIA-TBP complex was measured on both elements and found to be the same, within experimental error (Fig. 4). This shows that TFIIA can neutralize the subtle kinetic rate difference found on CATAAA versus TATAAA when TBP alone is measured (Fig. 2). In addition, the total amount of TFIIA-TBP-DNA complex formed is equivalent for both elements.
In contrast, the stability of the TFIIA-TBP-TATAAA complex differs significantly from the complex formed on the CATAAA element (Fig. 5). The TFIIA-TBP complex was extremely stable on TATAAA, with a t1 ⁄2 of 14 h under the conditions assayed (determined by extrapolation from the exponential decay graph). This half-life is in good accord with previous measurements of the stability of the TFIIA-TBP-TATAAA complex (35). In contrast, the TFIIA-TBP complex on CATAAA shows significant loss of complex over the course of the experiment with a short comparative half-life of only 1.4 h. Thus, although the TFIIA-TBP complex is rapidly formed to the same extent on both TATAAA and CATAAA, these complexes differ in that the TFIIA-TBP-TATAAA complex is significantly more stable than the complex formed on CATAAA.
The TFIIB-TBP Complex Behaves Similarly on the Two Elements Indicating the Proper Formation of Certain Higher Order Complexes-Because the stability of the TFIIA-TBP-DNA complex is compromised on the CATAAA element, we wished to determine whether this was specific to the TFIIA-TBP complex or whether a loss of stability is also observed for other higher order complexes. Thus, we compared the binding and stability of the TFIIB-TBP-DNA complex on TATAAA and CATAAA. As is the case for TBP and TFIIA, the relative amount of initial binding of the TFIIB-TBP complex is very similar on both elements (Fig. 6). In contrast to the results obtained with TFIIA, there is no observable difference in the stability of the TFIIB-TBP-DNA complex on TATAAA versus CATAAA. Both the TFIIB-TBP-TATAAA and TFIIB-TBP-CATAAA complexes have a t1 ⁄2 of 120 min under the conditions assayed. We conclude that the formation of the TFIIB-TBP higher order complex behaves similarly on the two elements, indicating that the difference in stability of the TFIIA-TBP-DNA complex on CATAAA is specific to the complex containing TFIIA. DISCUSSION The more similar a core promoter element is to the consensus sequence TATAAA, the greater the output from the promoter (24, 25, 36 -38). In addition, binding of TBP to the TATA box is a rate-limiting step in transcription (3)(4)(5)39), and TBP occupancy at the promoter correlates very well with transcriptional initiation at a majority of promoters (6,7). One would therefore hypothesize that the recognition of the TATA element by TBP is a critical and sequence-specific event in transcription initiation, and yet a comparison of the crystal structures of TBP bound to a number of TATA variants revealed no differences in TBP conformation or the trajectory of the DNA (40). Likewise, we find that TBP recognizes and binds to TATAAA or CATAAA with similar affinities and that the stability of TBP bound to the elements is identical. On a structural basis, the substitution of a cytosine for the thymine in position 1 of the TATAAA element should allow for TBP binding because no steric clashes exist in the minor groove between the exocyclic NH 2 group from the guanine and the adjacent side chains of TBP (40,41). Because we observe a 50-fold difference in expression level from these two elements in vivo, we examined higher ordered interactions to determine which factors play a critical role in modulating the output of gene expression on noncanonical elements.
The crystal structures of both the TFIIA-TBP-TATAAA and the TFIIB-TBP-TATAAA complexes and a large body of biochemical and genetic studies show that these transcription factors play a role in binding and stabilizing the TBP-DNA interaction (42)(43)(44)(45)(46)(47)(48). We found that the TFIIA-TBP-DNA complex formed on CATAAA was significantly less stable than the complex formed on TATAAA in vitro. Moreover, a fusion of TFIIA and TBP resulted in a 15-fold increase in expression from a promoter containing the CATAAA element in vivo. Thus, increasing the effective concentration of TFIIA at the promoter via a TBP fusion can increase the transcriptional output from CATAAA in vivo. This effect is specific to the TFIIA-TBP complex, because the stability of the TFIIB-TBP complex was identical on the two elements in vitro, and fusion of TFIIB to TBP had no effect in vivo.
How can a loss of stability of the TFIIA-TBP complex account for such dramatic differences in expression in vivo? A high level of gene output not only depends on initial induction of the gene but also on the number of times the gene is continuously transcribed or reinitiated (49 -51). Analysis of the fate of several general transcription factors during the transition from initiation to elongation has shown that many factors are released from the DNA template, whereas TBP, TFIIA, and some activators remain associated (49,52). The sequence of the TATA element has also been shown to be critically important to the rate at which reinitiation occurs, and TATA-less promoters fail to show rapid reinitiation (38,53). Because TBP and TFIIA remain associated with the core promoter after polymerase escape, and these factors have been implicated in directing rapid reinitiation, it seems likely that a difference in stability of this complex could result in significant changes in the rate of reinitiation. The hypothesis that CATAAA is defective for reinitiation can also explain the overall low amounts of transcription measured for this element in vivo.
TBP binds relatively promiscuously to both TATA-containing and non-TATA-containing DNA (16,30,31,35,54). This apparent lack of specificity of TBP for its cognate site is at least partially due to the fact that TBP binds to the minor groove of DNA (55)(56)(57)(58), which provides few functional groups to direct specificity of binding (41). Genetic and biochemical studies suggest that the cell has devised several means of increasing the specificity of TBP for TATA-containing promoters. Certain TBP-associated factors (TAFs) (59), Mot1 (60, 61), NC2 (62), and the Not complex (27), each may play a role in regulating transcription from noncanonical TATA elements. Interestingly, TFIIA can counteract the negative effects on TBP-DNA complex formation of the amino terminus of TAF145 (63), of Mot1 (64), and there are genetic interactions between TFIIA and NC2 (65). Thus, loss of a stable TFIIA-TBP-DNA interaction on noncanonical elements could tip the balance to favor the effects of these negative factors. In addition, transcriptional output could be effected on a number of other levels because TFIIA has also been implicated in core promoter functions (33,66,67) and activated transcription (68 -73). Taken together with the results presented here, this suggests that the stability of the TFIIA-TBP-DNA complex plays a critical functional role in promoter selectivity by increasing the specificity for consensus TATA elements in vivo.