The Role of Human TFIIB in Transcription Start Site Selection in Vitro and in Vivo *

The general transcription factor TFIIB plays a crucial role in selecting the transcription initiation site in yeast. We have analyzed the human homologs of TFIIB mutants that have previously been shown to affect transcription start site selection in the yeast Saccharomyces cerevisiae. Despite the distinct mechanisms of transcription start site selection observed in S. cerevisiae and humans, the role of TFIIB in this process is similar. However, unlike their yeast counterparts, the human mutants do not show a severe defect in supporting either basal transcription or transcription stimulated by an acidic activator in vitro. Transient transfection analysis revealed that, in addition to a role in transcription start site selection, human TFIIB residue Arg-66 performs a critical function in vivo that is bypassed in vitro. Furthermore, although correct transcription start site selection is dependent upon an arginine residue at position 66 in human TFIIB, innate function in vivo is determined by the charge of the residue alone. Our observations raise questions as to the evolutionary conservation of TFIIB and uncover an additional function for TFIIB that is required in vivo but can be bypassedin vitro.

The general transcription factor TFIIB plays a crucial role in selecting the transcription initiation site in yeast. We have analyzed the human homologs of TFIIB mutants that have previously been shown to affect transcription start site selection in the yeast Saccharomyces cerevisiae. Despite the distinct mechanisms of transcription start site selection observed in S. cerevisiae and humans, the role of TFIIB in this process is similar. However, unlike their yeast counterparts, the human mutants do not show a severe defect in supporting either basal transcription or transcription stimulated by an acidic activator in vitro. Transient transfection analysis revealed that, in addition to a role in transcription start site selection, human TFIIB residue Arg-66 performs a critical function in vivo that is bypassed in vitro.

Furthermore, although correct transcription start site selection is dependent upon an arginine residue at position 66 in human TFIIB, innate function in vivo is determined by the charge of the residue alone. Our observations raise questions as to the evolutionary conservation of TFIIB and uncover an additional function for TFIIB that is required in vivo but can be bypassed in vitro.
Transcription of a gene by RNA polymerase II (pol II) 1 requires the assembly of the general transcription factors (GTFs) at the promoter to form a preinitiation complex (PIC; reviewed in Ref. 1 and 2). The GTFs assemble in an ordered fashion, beginning with the binding of TFIID to the TATA element present at most promoters. TFIID is itself a multiprotein complex of which one component mediates binding to the TATA element (TATA-binding protein; TBP). TFIIA joins the forming PIC, followed by TFIIB. The complex is then able to incorporate TFIIF/pol II, and assembly is completed by further GTFs, including TFIIE and TFIIH. This pathway of PIC assembly has been challenged by evidence that some GTFs may exist as a partially preformed entity with pol II (a holoenzyme) that is recruited to the promoter in a single step (reviewed in Ref. 3).
TFIIB plays a pivotal role in the PIC, providing a bridge between promoter-bound TFIID and pol II/TFIIF. Consistent with this, TFIIB has been found to interact with several other GTFs including TBP, TFIIF, and pol II (4). TFIIB is a 33-kDa peptide that contains a core C-terminal domain composed of two imperfect direct repeats and an N-terminal region that chelates zinc. Although intact TFIIB has evaded structural analysis, structural data have been obtained for the C-terminal domain in isolation (5,6). The core C-terminal region is mainly ␣-helical, with the end of the first direct repeat forming a basic amphipathic ␣ helix. The second direct repeat contains a surface cluster of hydrophobic amino acids and a helix-turn-helix motif that interacts with a specific DNA sequence upstream of the TATA element (7,8). Structural information has also been obtained for the immediate N terminus of Pyrococcus furiosus TFIIB, revealing a ␤-sheet structure forming a zinc ribbon similar to that present in the elongation factor TFIIS (9). The crystal structure of the TFIIB C-terminal domain in a complex with TBP bound to DNA has been solved (6). This structure suggested that the N terminus of TFIIB would project downstream toward the region of transcriptional initiation. Biochemical and structural evidence suggests that, in native TFIIB, the N-and C-terminal domains are engaged in an intramolecular interaction (10 -12). It is envisaged that to participate in a productive PIC, TFIIB must undergo a conformational change that disturbs this intramolecular interaction, exposing binding sites that are required for further assembly of the PIC. Deletion analysis of human TFIIB demonstrated that TFIIF interacts with the N terminus of TFIIB and pol II with the core C-terminal region (4). Thus, a conformational change in TFIIB may facilitate the interaction between TFIIB and pol II/TFIIF.
Underlying the central role of TFIIB in PIC assembly, TFIIB has been proposed as a target of transcriptional activator proteins (reviewed in Ref. 1 and 2). At the adenovirus E4 promoter, TFIIB assembly is a limiting event that can be facilitated by an activator protein (13). Moreover, several activation domains can interact with TFIIB, and in vitro evidence suggests that this interaction is required for transcriptional activation (14). Significantly, the acidic activation domain of the herpes simplex virus VP16 protein can induce a conformational change in TFIIB that disrupts the intramolecular interaction (12). This provides a possible mechanism by which activators could stimulate further assembly of the PIC.
Once a PIC has assembled, the region around the transcription initiation site melts to provide a template strand for pol II. The actual site at which transcription initiation occurs is approximately 25-30 base pairs downstream of the TATA box in most eukaryotes (reviewed in Ref. 1). Where present, the TATA element is the sole determinant of the transcription initiation site, and initiation will occur at the distance set by TATA regardless of the sequence around the site of initiation (15). One well studied exception to this is the yeast S. cerevisiae, in * This work was funded by the Wellcome Trust (047674/Z/96/Z). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18  which the transcription initiation site can occur anywhere between 40 base pairs and 120 base pairs downstream of the TATA element (reviewed in Ref. 1). Even so, open complex formation occurs at a distance downstream of TATA similar to that seen in other organisms. A scanning polymerase model has been proposed in S. cerevisiae in which the PIC translocates downstream of the melted DNA and initiates transcription at distant sites (16). Schizosaccharomyces pombe exhibits a start site position akin to that seen in human cells, 25-30 base pairs downstream of the TATA element. Factor-swapping experiments between S. pombe and S. cerevisiae found that TFIIB and pol II play a key role in determining the transcription initiation site (17). TFIIB and pol II are not individually interchangeable between the two yeast species. However, when exchanged together, not only does this result in full transcriptional activity in both species, but the start site use reflects the species from which the TFIIB and pol II were derived. Thus, TFIIB and pol II in concert determine the differences in transcription initiation site between S. cerevisiae and other eukaryotes.
A genetic screen in S. cerevisiae recovered mutations in TFIIB that confer a bias toward the use of downstream transcription initiation sites (18). Two critical residues in yeast TFIIB were analyzed; Glu-62 and Arg-78. Substitution of residue Glu-62 with an amino acid of opposing charge (Lys) resulted in a yeast strain that exhibited cold sensitivity and a slow growth phenotype at 30°C (19). Transcription at the CYC1 and ADH1 genes in this strain showed a preference for the use of downstream initiation sites. The mutation of residue Arg-78 to cysteine resulted in a similar phenotype to E62K, but the mutant R78E resulted in a loss of viability. Because a double mutant (E62K/R78E) was phenotypically similar to E62K and not R78E, it was suggested that Glu-62 and Arg-78 may form a salt bridge. Glu-62 and Arg-78 mutants of TFIIB are transcriptionally defective in a yeast in vitro transcription system (10, 20 -22). This can be explained, at least in part, by their failure to interact with pol II in several different assays.
As discussed above, the mechanism of transcription start site selection in S. cerevisiae is distinct from that observed in most other eukaryotes. In this study we analyze human TFIIB mutants homologous to the S. cerevisiae mutants described above both in vitro and in vivo. Surprisingly, we find that all the mutants are able to support both basal transcription and the response to an activator protein. However, the mutants do exhibit a downstream shift in transcription start site selection in a promoter-specific manner. In vivo analysis confirmed the importance of these residues in human TFIIB for correct transcription start site selection. However, we demonstrate that mutation of a particular residue in human TFIIB affects a function in addition to transcription start site selection that manifests itself in vivo but is bypassed in vitro. We discuss these observations in light of current models of the role of TFIIB in both PIC formation and transcription start site selection.

EXPERIMENTAL PROCEDURES
Plasmids-The promoter DNA template G5E4T has been described previously (23). G5ML contains nucleotides Ϫ50 to ϩ22 from the adenovirus major late promoter cloned downstream of 5 GAL4 sites in the vector pGEM3.
Polymerase chain reaction-mediated site-directed mutagenesis was performed to produce the E51R, E51D, E51A, R66E, R66K, R66A, and E51R/R66E mutants. All clones were sequenced to ensure only the intended mutations were present. cDNAs encoding wild type TFIIB and the mutants (E51R, R66E, R66A, R66K, and E51R/R66E) were cloned into the vector pCDNA3 for expression in 293 cells. The in vivo vector expressing GAL4-RII has been described previously (24).
Protein Purification-TFIIB and TFIIB mutants and GAL4-AH were purified as described (4, 23). Polyhistidine-tagged TBP was purified by nickel chelate affinity chromatography as described by the manufacturer (Qiagen). The HeLa fraction containing RNA polymerase II, TFIIF, TFIIE, and TFIIH was purified as described previously (25). Briefly, 20 ml of HeLa nuclear extract (10 mg/ml; Computer Cell Culture Center, Mons Belgium) in buffer A (40 mM Tris, pH 7.9, 20% (v/v) glycerol, 0.5 mM EDTA, 10 mM ␤-mercaptoethanol, and 1 mM phenylmethylsulfonyl fluoride) containing 0.1 M KCl was fractionated over a 15-ml phosphocellulose column. The 0.5 M KCl fraction (20 ml) was dialyzed against buffer A containing 0.1 M KCl. This was applied to a 3-ml DEAE-cellulose column, which was washed extensively with buffer A (0.1 M KCl) and then with buffer A (0.25 M KCl). RNA polymerase II/TFIIF and TFIIE were eluted in 5 ml of buffer A (0.3 M KCl) and dialyzed into buffer D (20 mM HEPES, pH 8, 0.5 mM EDTA, 20% glycerol, 1 mM dithiothreitol, and 1 mM phenylmethylsulfonyl fluoride) containing 0.1 M KCl. This fraction was devoid of significant levels of TFIIB as assessed by immunoblotting, transcription, and electrophoretic mobility shift assay.
In Vitro Transcription-Anti-TFIIB antiserum was produced against full-length recombinant TFIIB (Scottish antibody production unit). HeLa nuclear extract was depleted of TFIIB by immunoaffinity chromatography over a protein-G-Sepharose column containing covalently linked TFIIB antibodies. Transcription assays using HeLa nuclear extract or TFIIB-depleted nuclear extracts were performed as described (13) using the amounts of recombinant TFIIB indicated in the figure legends. The reconstituted transcription assay was performed similarly, except it contained 20 l of the HeLa RNA pol II/TFIIF/TFIIE fraction, 20 ng of recombinant TBP, and the indicated amounts of TFIIB.
Transient Transfection Analysis-Human embryonic kidney 293 cells were cultured as monolayers in Dulbecco's modified Eagle's medium containing 10% fetal calf serum, 5 mM L-glutamine, 100 mg/ml streptomycin, and 100 units/ml penicillin. Cells were transfected in 90-mm dishes at 50% confluency using calcium phosphate. Briefly, 1 g of G5E4CAT (or G5MLCAT), 2 g of RSV-BxGALII, and 2 g of pCDNA3-TFIIB (or mutant) was diluted to 438 l with water, and 61 l of 2 M CaCl 2 was added. This was mixed with 500 l of 2ϫ HBS (54.6 mM HEPES, pH 7.1, 0.274 M NaCl, and 1.5 mM Na 2 HPO 4 ). The mixture was added dropwise to a plate of cells, and the media was changed 24 h later. Cells were harvested 48 h after transfection. For immunoblot, equal volumes of cells were added to SDS-polyacrylamide gel electrophoresis-loading dye, and the proteins were resolved by electrophoresis. After transfer to Immobilon P membrane, immunoblotting was performed with anti-T7 antibody (Novagen), which recognizes a tag at the C terminus of the TFIIB derivatives. Detection was performed by chemiluminescence (ECL, Amersham Pharmacia Biotech). Total RNA was prepared using an Rneasy kit as described by the manufacturer (Qiagen). 50 g of total RNA was annealed to a 32 P-labeled CAT primer, and extension was performed as for the in vitro transcription reactions. All transfections were performed at least three times.

RESULTS
Human TFIIB Mutants E51R, R66E, and E51R/R66E Support Transcription in Vitro-The region of S. cerevisiae TFIIB that is involved in transcription start site selection is very highly conserved between species (Fig. 1A). We have constructed single (E51R, R66E) and a double (E51R/R66E) point mutants of human TFIIB that correspond to the S. cerevisiae TFIIB mutants previously characterized by others to show defects in transcription start site selection (19). The mutants were purified from Escherichia coli and analyzed by SDS-polyacrylamide gel electrophoresis and Coomassie staining, demonstrating that all of the proteins were of equivalent purity (Fig. 1B). Wild type TFIIB and the mutants were then tested in a transcription reaction reconstituted from partially purified/ recombinant factors and dependent upon the addition of TFIIB. The promoter DNA used was the adenovirus major late (AdML) promoter, and transcripts were detected by primer extension. Fig. 2A shows that the system is dependent upon the addition of recombinant TFIIB for the production of accurately initiated transcripts (compare lanes 0 and wt IIB). E51R, R66E, and E51R,R66E were also able to support a level of transcription equivalent to that seen with wild type TFIIB. This is in contrast to the S. cerevisiae TFIIB mutants corresponding to E51R and R66E, which showed a significant defect in transcription in vitro (10, 20 -22). Moreover, all of the mutants utilized the same transcription initiation site as that observed with wild type TFIIB.
In view of our finding that E51R, R66E, and E51R/R66E support transcription in vitro at a level equivalent to wild type TFIIB, we next tested the possibility that the mutants may be defective in supporting the levels of transcription required during transcriptional activation. To monitor the high levels of transcription in the presence of the activator GAL4-AH, we prepared a nuclear extract depleted of TFIIB by chromatogra- phy over a column containing immobilized TFIIB antibodies. The TFIIB-depleted nuclear extract was deficient in supporting significant levels of transcription in the presence of GAL4-AH, but addition of wild type TFIIB restored activated levels of transcription (Fig. 2B). E51R, R66E, and E51R/R66E also restored high levels of transcriptional activation. Thus, the human TFIIB mutants E51R, R66E, and E51R/R66E exhibit wild type transcriptional activity (with respect to level and initiation site) at the AdML promoter both in the absence and presence of an activator protein.
TFIIB Mutants Cause a Downstream Shift in Initiation Site at the Adenovirus E4 Promoter-Transcription at the adenovirus major late promoter largely occurs from a single site. We therefore tested the adenovirus E4 promoter, at which transcription initiation naturally occurs at multiple sites. Fig. 3A shows transcription assays in a TFIIB-depleted extract in the presence of GAL4-AH at the AdE4 promoter. Consistent with the results we obtained with the AdML promoter, E51R, R66E, and E51R/R66E supported a level of transcriptional activation equivalent to wild type TFIIB. However, there was a clear shift toward the use of downstream initiation sites by all three of the mutants compared with wild type TFIIB. Specifically, there was an increase in the use of distal transcription initiation sites (indicated B and C) compared with proximal (A). The same effect was observed when we used GAL4-SP1 as the transcriptional activator (data not shown). To examine this further, we resolved the different transcripts produced by wild type TFIIB and E51R alongside a sequencing ladder of the AdE4 promoter produced using the same radiolabeled primer (Fig. 3B). Comparing the transcripts produced by wild type TFIIB with those produced by E51R shows a clear shift toward the use of downstream initiation sites. The sequence of this region of the AdE4 promoter is shown in Fig. 3C, with arrows indicating the sites of initiation in the presence of wild type TFIIB (shown above the sequence) and E51R (shown below the sequence). The size of the arrows indicates the relative use of each initiation site. Thus, the role of human TFIIB residues Glu-51 and Arg-66 in transcription start site selection is conserved between yeast and mammals. However, unlike in S. cerevisiae, the human mutants support transcription in vitro at a level equivalent to wild type TFIIB.
Determinants of the Transcription Start Site Shift-The TFIIB start site mutations used above involve the substitution of charged amino acids with residues of opposite charge. We next constructed substitutions of Glu-51 and Arg-66 to either alanine (E51A and R66A) or different amino acids of similar charge (E51D and R66K). The recombinant proteins (shown in Fig. 4A) were tested in a TFIIB-depleted nuclear extract at the AdE4 promoter (Fig. 4B). As seen with the original mutants, E51A, E51D, R66A, and R66K all supported transcription in vitro to a level equivalent to that observed with wild type TFIIB. Furthermore, they all caused a downstream shift in the transcription start sites. Thus, the charge per se of these residues is not the critical determinant in correct transcription start site selection in vitro.
Analysis of the Human TFIIB Mutants in Vivo-The human TFIIB start site mutants E51R, R66E, and E51R/R66E support levels of basal and activated transcription in vitro similar to that seen with wild type TFIIB. It would be predicted therefore that the addition of an excess of one of the mutant TFIIB proteins to a native nuclear extract would titrate the HeLa TFIIB and result in a downstream shift in the transcription initiation site. We tested this possibility

FIG. 3. The human TFIIB mutants cause a downstream shift in the transcription initiation site at the adenovirus E4 promoter.
A, wild type (wt) TFIIB or the indicated mutants were tested as in Fig.  2b but with the adenovirus E4 promoter (shown below the autoradiogram). The three major groups of transcripts are indicated (A, B, and C). B, transcripts produced at the AdE4 promoter with wild type TFIIB and E51R as shown in panel A were resolved on a 10% sequencing gel alongside a sequencing ladder of the same promoter. C, sequence of the AdE4 transcription initiation region with arrows above and below, indicating the relative intensity of each transcript produced in the presence of wild type TFIIB and the mutant E51R, respectively.

FIG. 4. Determinants in TFIIB required for correct transcription start site selection.
A, recombinant human TFIIB or the indicated mutants were purified from E. coli and analyzed by SDS-polyacrylamide gel electrophoresis/Coomassie staining. B, wild type (wt) TFIIB or the mutants shown in panel A were tested in a TFIIB-depleted nuclear extract at the AdE4 promoter in the presence of GAL4-AH as described in Fig. 3A. and found that this was indeed the case (data not shown and Fig. 5A). We reasoned that this effect would also occur in a living cell and therefore used a transient transfection assay to determine whether the start site shift effects we observe in vitro could be recapitulated in vivo. Transfection of G5E4T into living cells failed to produce a stable message. We therefore used the same G5E4 core promoter but fused to CAT, which allowed us to detect stable transcripts. Thus, it was first necessary to test G5E4CAT in vitro to ensure that we could make a direct comparison with the in vivo effects. Fig.  5A shows a transcription assay using G5E4CAT and native nuclear extract to which either wild type TFIIB, E51R, R66E, or E51R/R66E had been added. Results comparable with those seen in a TFIIB-depleted extract with G5E4T were obtained. Specifically, the mutants caused a shift toward the use of distal transcription initiation sites (B and C) with a concomitant reduction in the proximal initiation sites (A). Next, wild type TFIIB or the mutants (E51R, R66E, and E51R/R66E) under a cytomegalovirus promoter were transfected into human embryonic kidney 293 cells along with the activator Gal4-RII and G5E4CAT. The cells were harvested 48 h after transfection, total RNA was prepared, then primer extension was performed to analyze the transcripts. A representative assay is shown in Fig. 5B. There are several striking features in this result. First, there is a clear difference in the transcription start site pattern at the AdE4 promoter in vivo compared with that seen in vitro. Essentially, the most distal initiation sites seen in vitro (C) are not used in vivo. This cannot be explained by the difference in cell type as nuclear extracts made from 293 cells produce the same pattern of transcripts at the AdE4 promoter as HeLa cell nuclear extracts (data not shown). Second, in contrast to our in vitro observations, R66E significantly inhibited transcription in the transient transfection assay, but E51R and E51R/R66E did not. This was not because of a difference in the levels of the various TFIIB constructs, as immunoblotting showed that wild type TFIIB and the mutants were of similar abundance (Fig. 5C). Finally, overexpression of E51R and E51R/ R66E caused a shift toward the use of the downstream transcription initiation sites. The intensities of the two sets of transcripts (A and B) were quantified and are expressed as a ratio (B/A) below each lane. Thus, although there are greater constraints on the pattern of transcription initiation at the AdE4 promoter in vivo, there is a clear shift toward the use of the downstream transcription initiation sites in vivo by the TFIIB mutants. Moreover, the low level of E4 transcription remaining in cells transfected with R66E also exhibited the downstream shift. We next performed a set of transfections identical to those above but using G5MLCAT as a reporter (Fig. 5D). As seen with the E4 promoter, R66E significantly inhibited transcription, but E51R and E51R/R66E did not. Consistent with our in vitro data, only a single transcript was produced from the AdML promoter, and the mutants did not cause the use of an alternative site.
In view of the difference we observed in the properties of hTFIIB R66E in vivo versus in vitro, we next compared R66E with the mutants R66A and R66K in a transient transfection assay (Fig. 6A). As before, R66E inhibited transcription, with residual activity showing a downstream shift in transcription initiation site. The mutant R66A was also inhibitory, although we consistently found this mutant to be less inhibitory than R66E. In agreement with our in vitro observations, R66A also caused a downstream shift in the site of transcription initiation. Interestingly, R66K did not significantly inhibit transcription transcription in vivo, although it did cause a downstream shift in the transcription initiation site. All of the mutants were expressed at similar levels in the transfection assay, as assessed by immunoblotting (Fig. 6B). Thus, although the charge of hTFIIB residue 66 is critical in determining the innate transcription function of TFIIB in vivo, it is not sufficient for correct transcription start site utilization either in vivo or in vitro.

DISCUSSION
In this study we have demonstrated that, despite the distinct mechanisms in transcription start site selection observed in S. cerevisiae and mammals, the role of TFIIB in this process is highly conserved. The mutation of two highly conserved charged residues in the N terminus of human TFIIB, homologous to those previously described in the S. cerevisiae, caused a downstream shift in transcription start site selection both in vitro and in vivo. The transcription start site shift caused by FIG. 5. Analysis of the human TFIIB mutants E51R, R66E, and E51R/R66E in vivo. A, recombinant TFIIB or the indicated mutants (250 ng) were added to a native HeLa cell nuclear extract, and transcription assays were performed at the AdE4 promoter as described in Fig. 3. Transcript groups A, B, and C are indicated. B, human embryonic kidney 293 cells were transfected with vectors expressing wild type TFIIB or the indicated mutants under a cytomegalovirus promoter (2 g) alongside the reporter G5E4CAT and a vector driving the expression of the transcriptional activator GAL4-RII. E4 transcripts among the total RNA isolated from the cells was subsequently identified by primer extension. Transcripts (A and B) are indicated, and quantitation is expressed as a ratio B/A is shown below each lane of the autoradiogram. C, immunoblot of total cell lysates prepared from cells transfected as in part B with anti-T7 antibody, recognizing an epitope engineered into the C terminus of the TFIIB (and mutant) constructs. Molecular weight markers (kDa) are shown at the left. D, as in panel B except that the AdMLCAT promoter construct was used.
E51R, R66E, and E51R/R66E was promoter-specific, as it occurred at the AdE4 promoter and not the AdML promoter. Unlike the AdML promoter, the E4 promoter intrinsically exhibits a more flexible transcription initiation site pattern. This observation is similar to previous findings with the S. cerevisiae TFIIB mutants, which caused a transcription start site shift only at promoters that normally exhibit multiple transcription start sites (22,26). It is likely that alterations in transcription initiation site are not tolerated at the AdML promoter. Thus, as the mutants are competent for innate transcription function in vitro, they would appear to function similarly to wild type TFIIB at the AdML promoter.
There are several charged residues in this highly conserved region of TFIIB, suggesting that they may form a functional domain that interacts with a similarly charged surface in another protein. Indeed, alterations to the charge of this region invariably results in TFIIB mutants that confer downstream shifts in the transcription initiation site (10, 22) 2 . Interactions involving complementary electrostatic charge are energetically neutral as each domain is able to form a favorable interaction with water (27). Once removed from water, charged/polar groups need to form ion pairs and hydrogen bonds with a complementary surface. The net result does not generally contribute to the affinity of the interaction but can have effects on the specificity and, therefore, the shape of the resulting complex. This raises the intriguing possibility that the mutants we describe in this study cause a shift in the transcription initiation site not because the interaction with pol II/TFIIF (or other factors) is ablated but because of an alteration in the alignment of the complex. Such a model would be consistent with our finding that the human TFIIB start site mutants support transcription to a level equivalent to wild type TFIIB in vitro. Structural analysis of this region of TFIIB will shed light on the role of the charged cluster domain of TFIIB in transcription start site selection.
The transcription start site shift induced by the various mutants used in this study was the same. Presumably, any alteration in the positioning of pol II at the promoter would be constrained relative to TATA and, therefore, provide a limited number of potential transcription start sites. We also note that the TFIIB mutants are still able to direct transcription initiation at all of the upstream sites seen with wild type TFIIB, albeit to a lesser degree. It is therefore likely that the role of TFIIB in the selection of the transcription initiation site may be restricted to modulating the 3Ј parameter, whereas other factors determine the 5Ј parameter. Indeed, mutants of the S. cerevisiae RBP9 subunit of pol II cause an upstream shift in transcription start site selection (28,29). The conservative TFIIB substitution mutants E51D and R66K also caused a downstream shift in transcription start site selection. Thus, charge per se within this region is not the sole determinant of correct transcription start site selection. Alterations in the size of the side chains of key amino acids may also have effects on the alignment of a protein-protein interaction. Significantly, the residues corresponding to human TFIIB Glu-51 and Arg-66 are totally conserved in all species sequenced to date. Although other charged residues are also totally conserved, some are substituted by residues of similar charge, and still others, where the charge is removed altogether. It is important to note that TFIIB exhibits a high degree of species specificity. Unlike TBP, S. cerevisiae and human TFIIB are not functionally interchangeable, even in basal transcription (30). In addition, it has been reported that human TFIIB can substitute for Drosophila TFIIB only at certain promoters (31). It is possible that the differences in the charge cluster domain of TFIIB contribute to this species specificity.
In contrast to our results with human TFIIB, homologous yeast TFIIB mutants are defective for transcription in vitro (10, 20 -22). This defect of the S. cerevisiae mutants correlates with their inability to stably interact with RNA polymerase II in several different assays. Deletion mapping studies of human TFIIB have shown that the N terminus (including the region studied in this work) is dispensable for interaction with RNA polymerase II (4). However, deletion of the N terminus of S. cerevisiae TFIIB abolishes the interaction with pol II (22). Furthermore, evidence has been presented that the region of S. cerevisiae TFIIB responsible for transcription start site selection and direct interaction with pol II may constitute distinct, but overlapping, domains within the N terminus (22). Thus, it is possible that the nature of the contacts between TFIIB and pol II (and/or other components of the general transcription machinery) have diverged from S. cerevisiae to humans. Perhaps yeast TFIIB requires a more intricate interaction with pol II as part of the scanning mechanism responsible for transcription start site selection in S. cerevisiae. Because S. pombe exhibits a transcription start site pattern similar to that seen in human cells, analysis of the TFIIB-pol II interaction in this species should help resolve this issue.
The human TFIIB mutant R66E was indistinguishable from E51R or E51R/R66E in in vitro transcription assays. However, in transfection analysis R66E was a potent inhibitor of transcription, whereas E51R and E51R/R66E caused effects similar to those seen in vitro. Importantly, the low level of transcription observed in the presence of R66E in vivo exhibited a downstream shift in the transcription initiation site. Thus, TFIIB R66E is severely compromised in its ability to support transcription in vivo in a manner that is bypassed in vitro. Furthermore, this defect can be suppressed by the additional mutation E51R. This compensation effect in vivo is consistent with the previous observation that the S. cerevisiae homolog of human R66E is lethal but can be rescued by a second mutation  Fig. 6B. wt, wild type. corresponding to human E51R (19). The conclusion of the observations in S. cerevisiae was that these two residues of opposite charge form a salt bridge. Our data are consistent with this proposal but also suggest that the salt bridge has a function that is not critical for the innate ability of TFIIB to support transcription of a naked DNA template in vitro. Substitution of Arg-66 with lysine also results in a TFIIB mutant that causes a downstream shift in transcription initiation site both in vitro and in vivo. However, this conservative substitution mutant does not significantly inhibit transcription in vivo. Taken together, these data suggest that Arg-66 plays a role in two functions; first, transcription start site selection that is determined by the specific amino acid side chain, and second, a role that is required in vivo but not in vitro and is dependent upon the charge of the residue only.
A function that is required of TFIIB in vivo, but is bypassed in vitro, has several implications for our understanding of the role of TFIIB in transcription initiation. Moreover, our findings question the minimal requirements for a functional PIC in vivo as compared with those observed in vitro. Several GTFs are dispensable for transcription in vitro using supercoiled DNA templates (reviewed in Ref. 2). TFIIF is one such GTF and is highly relevant in this case, because both subunits (RAP30 and RAP74) can interact with TFIIB and more specifically the N terminus of human TFIIB (4,32). Furthermore, suppressor mutants of the large subunit of TFIIF have been isolated that can reverse the slow growth phenotype and transcription start site defects of the yeast TFIIB mutant E62K (human Glu-51 (33)). However, our preliminary data suggest that neither TFIIB E51R or R66E are defective for interaction with RAP30 or RAP74 in vitro. 2 As discussed above, these mutations in TFIIB may alter the specificity rather than abolish interactions with other factors. It is possible that the RAP74 suppressor mutants restore the alignment of the yTFIIB(E62R)-pol II/ TFIIF interaction, leading to normal transcription start site selection.
DNA topology may affect the ability of R66E to substitute the function of wild type TFIIB in vivo. For example, R66E may be nonfunctional at a nucleosome-assembled promoter. Experiments in vitro using nucleosomal DNA templates could be used to test this hypothesis. It is also possible that the in vivospecific defect of R66E involves the response to upstream transcriptional activators. Indeed, a S. cerevisiae protein (SUB1) similar to the mammalian coactivator PC4 was isolated as a suppressor of the yTFIIB R78H cold-sensitive phenotype (34). The prospect of an additional function for TFIIB underscores the pivotal role that this GTF plays in transcription. Further studies are currently in progress to determine the function of TFIIB that is specifically required in a living cell.