Research Article

Structural Basis of Transcription: An RNA Polymerase II-TFIIB Cocrystal at 4.5 Angstroms

See allHide authors and affiliations

Science  13 Feb 2004:
Vol. 303, Issue 5660, pp. 983-988
DOI: 10.1126/science.1090838

Abstract

The structure of the general transcription factor IIB (TFIIB) in a complex with RNA polymerase II reveals three features crucial for transcription initiation: an N-terminal zinc ribbon domain of TFIIB that contacts the “dock” domain of the polymerase, near the path of RNA exit from a transcribing enzyme; a “finger” domain of TFIIB that is inserted into the polymerase active center; and a C-terminal domain, whose interaction with both the polymerase and with a TATA box–binding protein (TBP)–promoter DNA complex orients the DNA for unwinding and transcription. TFIIB stabilizes an early initiation complex, containing an incomplete RNA-DNA hybrid region. It may interact with the template strand, which sets the location of the transcription start site, and may interfere with RNA exit, which leads to abortive initiation or promoter escape. The trajectory of promoter DNA determined by the C-terminal domain of TFIIB traverses sites of interaction with TFIIE, TFIIF, and TFIIH, serving to define their roles in the transcription initiation process.

RNA polymerase II (pol II) assembles with five general transcription factors and Mediator at every promoter before the initiation of transcription (Table 1). This giant complex of 50 or more polypeptides, with a total mass in excess of 2.5 million daltons, recognizes promoter DNA, responds to regulatory information, and synthesizes the first dozen residues of the RNA transcript. Elucidation of the structure and mechanism of the initiation complex is necessary for understanding transcriptional regulation. We took a first step in this direction with the x-ray structure determination of pol II, itself a complex of 12 polypeptides and mass of 0.5 MD (14). We also determined the structure of pol II in the act of transcription, revealing the template unwound in a “transcription bubble” and the transcript in an RNA-DNA hybrid (5). We now report the structure of a cocrystal of pol II with general transcription factor IIB (TFIIB), the critical component for both assembly and disassembly of the transcription initiation complex.

Table 1.

RNA polymerase II transcription machines. Mass data are for yeast proteins. Details about function are in the text. TAF, TBP-associated factor.

Component Subunits Mass (kD) Function
Pol II 12 520 RNA synthesis
TFIIB 1 38 Start site determination
TFIID (TBP) 1 27 Bending TATA box DNA around TFIIB and pol II
    (TAFs) 14 749 Promoter recognition
TFIIE 2 92 Coupling pol II—promoter interaction to recruitment of TFIIH
TFIIF 3 156 Interaction with nontemplate DNA strand
TFIIH 9 525 Promoter opening, pol II phosphorylation
Mediator 20 1003 Regulatory signal transduction

TFIIB and TFIID are responsible for promoter recognition and interaction with pol II; together with pol II, they form a minimal initiation complex, capable of transcription under certain conditions (see below). The TATA box of a pol II promoter is bound in the initiation complex by the TBP subunit of TFIID, which bends the DNA around the C-terminal domain of TFIIB (TFIIBC). The N-terminal domain of TFIIB (TFIIBN) interacts with pol II (6, 7). The first evidence for a role of TFIIB in “bridging” between promoter DNA and pol II came from binding studies, which showed a requirement for TFIIB for interaction of a TBP-promoter DNA complex with pol II (8). The functional relevance of this interaction was demonstrated by transcription with purified proteins from budding and fission yeast. Neither pol II nor most general transcription factors could be exchanged individually between budding and fission yeast systems. Some pairs of proteins could, however, be exchanged. In particular, pol II and TFIIB from fission yeast could be substituted for their counterparts in the budding yeast system without loss of transcription activity. Evidently interaction between pol II and TFIIB is essential for the initiation of transcription (9).

Genetic studies point to a role of TFIIB in transcription start site determination. Mutations of residues in a subdomain of TFIIBN alter the start site in a manner dependent on the DNA sequence around the start site (1015). Mutation of a residue in the same subdomain allows the assembly of a transcription initiation complex but diminishes the level of transcription (16, 17). Structural studies of TFIIB have not so far been able to explain the biochemical or genetic results. Nuclear magnetic resonance (NMR) studies of TFIIBN in solution revealed only a zinc ribbon motif in a part of the protein (18, 19). The x-ray crystal structure of a complex of TFIIBC with TBP and a TATA box fragment has shown non-sequence-specific contacts of TFIIBC with the DNA upstream and downstream of the TATA box, as well as sequence-specific interactions with an element (BRE) immediately upstream (2022). These structural results shed no light on the critical pol II–TFIIB interaction nor on the mechanism of transcription start site determination. The structure of pol II associated with the entire TFIIB, reported here, illuminates these aspects.

Conformational change of pol II. A 10-subunit form of budding yeast (Saccharomyces cerevisiae) pol II, whose structure we have previously reported, was crystallized in the presence of a sixfold molar excess of recombinant, S. cerevisiae TFIIB. We collected diffraction data complete to 4.5 Å resolution, and we determined the structure by molecular replacement with the structure of pol II alone (Table 2). Three forms of pol II were previously solved, differing in the position of the “clamp” (Fig. 1A), one of four mobile modules that make up the entire structure (4, 5). The clamp swings from an “open” position in the absence of DNA and RNA to a “closed” position in their presence. All three forms were used in molecular replacement, and that with the clamp in a closed state (5) gave the best initial solution. Rigid body refinement, allowing all four mobile modules to move, improved the fit to the data. The “jaw-lobe” module moved inward toward the DNA-binding cleft (arrow in Fig. 1A), which produced a 6.5 Å shift of an associated zinc ion, revealed by an anomalous difference map. The average movement of the entire jaw-lobe module, pivoting about the rest of the structure, was 3.5 Å.

Fig. 1.

TFIIBN in the pol II–TFIIB complex. Top (A and B) and side (C and D) views of a backbone model of the pol II–TFIIB complex. These views are related by a 90° rotation about a horizontal axis, as indicated, and are the same as those defined previously (3). Dashed boxes in (A) and (C) are magnified in (B) and (D), with the inclusion of electron density for TFIIBN (green net, a 2FobsFcalc SIGMAA-weighted map, with phases from pol II alone, contoured at 1.2σ) in (B) and (D). Color code for domains of pol II in close proximity to TFIIBN and for TFIIBN is shown below. The arrow in (A) indicates the direction of movement of the jaw-lobe module between transcribing complex and pol II–TFIIB structures. The location of the pol II saddle is indicated by a circle in (B).

Table 2.

Crystallographic data for pol II–TFIIB complex. Values in parentheses correspond to the highest resolution shell. Rsym = Σi,h|I(i,h) – 〈I(h)〉 |/Σi,h|I(i,h)| where 〈I(h)〉 is the mean of the i observations of reflection h. Rsym was calculated with anomalous pairs merged; no sigma cut-off was applied.

Parameter Value
Space group I222
Unit cell dimensions (Å) 212.4 by 217.2 by 422.9
Wavelength (Å) 0.98
Mosaicity (°) 0.80
Resolution (Å) 50-4.5 (4.7-4.5)
Completeness (%) 99.4 (99.2)
Redundancy 4.4 (3.4)
Unique reflections 57906 (8951)
II 14 (3.4)
Rsym (%) 9.1 (27.7)

TFIIB zinc ribbon domain interacts with the pol II dock domain. There were nine peaks in the anomalous difference map, ranging in intensity from 5σ to 10σ. Eight of these peaks corresponded to the positions of the eight zinc ions previously identified in the pol II structure (3). The additional peak, with an intensity of 7σ (Fig. 1D, Zn indicated by arrow, anomalous map in white), could only be due to TFIIB, and it was located above the previously described “dock” domain (Fig. 1 and fig. S1) of Rpb1, the largest subunit of pol II (4). A difference electron density map (2FobsFcalc, pol II–TFIIB minus pol II) showed additional density, attributable to TFIIB, surrounding the additional zinc peak (Fig. 1D). There was a good fit of the zinc ribbon motif from the NMR structure of human TFIIBN to this additional density (19). The fit encompassed amino acid residues 20 to 54 (yeast numbering, Fig. 2), which form three antiparallel β strands surrounding the zinc ion. Only minor changes in the human NMR zinc ribbon in this region were required: the addition of residues 30 and 31 into extra density seen in the map, corresponding to an insertion in the yeast TFIIB amino acid sequence; and minor movements of the polypeptide backbone to better fit the observed density. Three additional residues, 17 to 19, could also be built into density seen at the N terminus of the ribbon domain.

Fig. 2.

Sequence alignment and domain structure of TFIIBN. Side view (see Fig. 1) of a back-bone model of TFIIBN. Domain names and col- or code are shown in the bar directly below. Amino acid sequences aligned at the bottom are from Y, yeast; H, human; and P, archaebacteria (Pyrococcus furiosus). Sequence alignment was performed with ClustalW.

The TFIIB zinc ribbon domain clearly contacts Rpb1 between residues 409 and 419 of the dock domain. This contact explains genetic findings that deletion of TFIIBN or mutation of the cysteine residues that chelate the zinc ion abolish TFIIB–pol II interaction (6, 7, 2325). It is, furthermore, consistent with recent evidence from photo-cross-linking and hydroxyl radical cleavage for interaction of the TFIIB zinc ribbon with the pol II dock domain (26).

TFIIB finger domain projects into the active center of pol II. The difference electron density (2FobsFcalc) map showed clearly connected TFIIB density beyond residue 54 of the zinc ribbon domain, passing across the “saddle” between the pol II clamp and “wall” and then down into the enzyme active center (Fig. 1). The density reaches down almost to the base of the cleft, before turning upward and exiting the way it came. A polyalanine model of this “finger” domain could be built into the entire connected density, extending from residue 54, where it departs from the zinc ribbon domain, to residue 88, where the connected density ends, back at the zinc ribbon domain. Even at the resolution of our analysis, some bulky side chains were clearly visible, for example, Trp63 and Arg64, which formed a distinctive pair and which confirmed the registration of the model to this point in the sequence.

TFIIB C-terminal domain orients promoter DNA in an initiation complex. Beyond the end of the finger domain, there were two additional regions of connected electron density due to TFIIB. One region, corresponding to about 15 amino acid residues, appeared to interact with the zinc ribbon domain. This region could be attributed to the linker (residues 89 to 120) between the finger domain and TFIIBC (Fig. 2), because of its proximity to the end of the finger domain, and because it could not be fit by any part of the NMR structure of TFIIBN or the x-ray structure of TFIIBC.

The second region of connected density lay within an area of otherwise choppy density, unconnected but above background, adjacent to the linker. This second region was clearly helical, and it could be fit by the first α helix of the first of two repeat domains comprising TFIIBC (Fig. 3, A and B). This fit was preferred over other helices of the first repeat because of the close proximity to the linker, and because the other helices are associated in a bundle; if one helix gave rise to a region of connected density, then the rest of the bundle should have done so as well. The occurrence in the difference map of density due to the linker and first α helix of TFIIBC, but not the rest of TFIIBC, is indicative of a rocking motion, due to mobility between the finger domain and linker and between the linker and TFIIBC. Comparison of NMR solution and x-ray crystal structures reveals additional mobility between the first and second repeat domains of TFIIBC (27).

Fig. 3.

TFIIBC in the pol II–TFIIB complex and the inferred locations of TBP and promoter DNA in a transcription initiation complex. (A) Side view (see Fig. 1) of the pol II–TFIIB structure, with an α helix from TFIIBc indicated by an arrow. The color code, the same as that in Fig. 1, is shown at the bottom. The dashed box is magnified in (B), with the inclusion of electron density for TFIIBC (green net, a 2FobsFcalc SIGMAA-weighted map, with phases from pol II alone, contoured at 1.2σ). (C) Same as (A) with a published model of a TFIIBc–TBP–TATA box DNA complex docked by least-squares alignment to the observed TFIIBc helix and with the TATA box DNA extended by the addition of 20 bp of B-form DNA at both ends. (D) Top view (see Fig. 1) of the model in (C).

On the basis of the location of the first α helix of TFIIBC, we could dock the published cocrystal structure of a TATA-box fragment with TBP and TFIIBC to our cocrystal structure of pol II with TFIIB. We extended the TATA box fragment at both ends with straight B-form DNA to produce a model of a minimal initiation complex (Fig. 3, C and D). In consequence of the bend imparted by TBP to the TATA box and of the orientation due to TFIIBC, the DNA wraps around pol II and TFIIB in the model. It hugs the protein surface, both upstream and downstream of the TATA box. The downstream DNA is directed toward the pol II active center. It passes above the pol II saddle at a point about 15 base pairs (bp) from the TATA box, which is consistent with two previous observations: the structure of a pol II transcribing complex (5) places the beginning of DNA unwinding at the upstream end of the transcription bubble, above the saddle; and permanganate cleavage experiments have shown that unwinding begins in an initiation complex about 12 bp downstream of the TATA box (28).

Once unwound above the saddle, the template DNA strand must pass to the active center. The transcribing complex structure and biochemical evidence both indicate that about 12 residues extend from the point of unwinding to the active center. A total of about 15 bp from the TATA box to the saddle plus 12 residues from the saddle to the active center corresponds well with the conserved distance of 25 to 30 bp from the TATA box to the transcription start site in pol II promoters [except in S. cerevisiae and related fungi (29)].

The minimal initiation complex model thus explains how pol II–TFIIB interaction determines the location of the transcription start site: TFIIB brings a TBP–TATA box complex to a point on the pol II surface from which the DNA need only follow a straight path and, after unwinding, the transcription start site in the DNA will be juxtaposed with the active center. The DNA region between the TATA box and transcription start site serves as a spacer, whose length of about 25 bp is important, and therefore conserved, but whose sequence is not. Indeed, the high variability of the sequence in this region not only supports a role as a spacer but indicates a function as an essentially unbent element, because a requirement for curvature would likely lead to the selection of a preferred sequence (30).

TFIIB finger domain performs multiple roles in transcription initiation. The template DNA strand passing from the pol II saddle to the active center of a transcribing complex follows a path similar to that of the TFIIB finger domain projecting from the saddle into the active center cleft. Superimposition of the RNA-DNA hybrid from the pol II transcribing complex structure upon the TFIIB finger domain shows a close proximity of the template DNA strand to the finger and also a steric clash of the RNA strand beyond about the fifth residue with the finger (Fig. 4). We investigated the possible interaction of the template DNA strand with the finger by surface plasmon resonance (SPR). These experiments took advantage of a report that a DNA strand and complementary RNA oligonucleotide bind to pol II to form a stable transcribing complex. Optimal stability was obtained with an eight-residue RNA (31). The x-ray crystal structure of a transcribing complex formed in this way confirmed its identity with a complex whose RNA component was produced by pol II transcription (32). We performed SPR measurements with a DNA strand attached by one end to a sensor surface. Pol II alone, TFIIB alone, or pol II with a complementary five-residue RNA oligonucleotide showed no detectable affinity for the immobilized DNA (Fig. 5A). Pol II with the five-residue RNA and TFIIB, however, did bind the DNA. Evidently, TFIIB stabilizes a complex of pol II with a short RNA-DNA hybrid. Pol II with a nine-residue complementary RNA bound the DNA in the absence of TFIIB, as expected, and the addition of TFIIB had little effect (Fig. 5B). We presume the finger domain shifts slightly in position to accommodate a full-length RNA-DNA hybrid in the active center cleft. Perhaps related to the requirement for such a shift, a barrier to the synthesis of RNA beyond 5 residues in length has been noted for RNA polymerase III, whose initiation factors include a counterpart of TFIIB (33).

Fig. 4.

Interaction of the B finger domain of TFIIBN with DNA template and RNA transcript. A stereo pair is shown, including the B finger domain from the pol II–TFIIBN structure, RNA-DNA hybrid helix from the pol II transcribing complex structure (5), and active site Mg ion. Color code is shown below.

Fig. 5.

Stabilization of a transcription initiation complex, containing a short transcript, by TFIIB. (A) A single strand of DNA was bound to the surface of a Biacore Biosensor chip. Combinations of pol II, TFIIB, and a five-residue RNA complementary to the DNA were applied as indicated, and the change in refractive index near the chip surface, in resonance units (ru), was measured as a function of time. (B) Same as (A) but with nine-residue RNA.

The enhancement of pol II binding to a short RNA-DNA hybrid by TFIIB is most likely due to interaction of the DNA with conserved residues 59 to 80 of the finger domain (Fig. 2). There appear to be two points of close contact: between nucleotides –6 to –8 (relative to the nucleotide addition site at +1) and finger residues 62 to 66 (side of the finger); and between nucleotides +1, –1, –2 and finger residues 69 to 74 (the “fingertip”). Residues 62 to 66 on the side of the finger have been identified by genetic analysis as important for start site selection. Mutations of residues 62, 63, 64, and 66 have all been shown to cause start site shifts or changes in the relative use of multiple start sites (1013, 15, 17, 34). Mutation of residue 78 (Arg to Leu) has a similar effect, which has been interpreted in terms of a possible salt bridge with Glu62. A double mutant, swapping charged residues between the two positions, gave partial restoration of function (15). In the model of the finger domain, Glu62 is on the opposite side of the finger from Arg78, which is compatible with a salt bridge and which may be important for maintaining the conformation of the finger in the absence of significant secondary structure.

The region around the start site contacted by the fingertip sometimes contains a seven-base sequence element termed the initiator, originally identified in a mammalian promoter, and which has no reported counterpart in S. cerevisiae (35). The initiator contributes to promoter strength and can suffice for transcription in the absence of a TATA box. It may be significant that the human fingertip contains two lysine residues, favorable for tight binding to DNA, whereas the S. cerevisiae fingertip contains only a histidine for possible DNA interaction, with one lysine replaced by aspartic acid and the other by glycine and aspartic acid. A lack of interaction or even repulsion in S. cerevisiae may account for a variable location of the start site (40 to 120 bp from the TATA box) compared with humans and most other eukaryotes (25 to 30 bp from the TATA box) (29).

Mutation of residue 62 in the human finger affects the start site location differently for different initiators (14). This reveals a functional interplay between the side of the finger and the fingertip. A preferred DNA contact at one site may augment a contact at the other. Or if preferred contacts at the two sites are out of register, then multiple transcription start sites may arise, or promoter strength may be diminished. Details of these DNA contacts and their interplay await structure determination of pol II with both TFIIB and DNA, and at higher resolution.

Thus, transcription start site determination by TFIIB may be regarded as a two-step process: coarse setting of the start site, due to TFIIBC binding to the TBP–TATA box complex and TFIIB zinc ribbon binding to pol II, and fine setting, after DNA unwinding, due to finger domain interaction with the template strand around the start site and the region immediately upstream. Our results also indicate a role of TFIIB in facilitating the transition from an open promoter complex (DNA unwound, before transcription initiation) to a transcribing complex (containing a 9-bp RNA-DNA hybrid), by the stabilization of intermediates containing very short RNA-DNA hybrids.

The finger domain may contribute to the initiation process in additional ways. For example, it may facilitate DNA's unwinding by capture of the template single strand in a transiently unwound duplex. And it may help position the initiating nucleoside triphosphate in the –1 site (also known as the i site, which is occupied by the 3′-end of the RNA during transcription elongation) for phosphodiester bond formation with the incoming nucleoside triphosphate in the +1 site (or i + 1 site), as has been suggested for the bacterial sigma factor (36).

TFIIB finger domain may play an inhibitory role. Beyond its multiple roles in facilitating transcription initiation, TFIIB may also interfere with the process, by clashing with the RNA in the transcribing complex structure on the pol II saddle (32) and, as already noted, in the RNA-DNA hybrid region. The clash on the saddle occurs near the junction of the finger and zinc ribbon domains of TFIIB and at about residues 10 to 12 of the RNA. Competition between TFIIB and RNA for position on the saddle could account for abortive initiation (failure of many transcripts to grow past about 10 residues) and for promoter escape (disruption of pol II–TFIIB interaction and departure of the transcribing enzyme from the promoter after synthesis of RNA longer than about 10 residues). These possible effects of TFIIB on transcription represent further points of similarity with proposals for bacterial sigma factor (37).

Architecture of the transcription initiation complex. Structural information on the remaining general transcription factors, TFIIE, TFIIF, and TFIIH places all three in the path of promoter DNA as defined by the minimal initiation complex model. The structure of TFIIF has been determined in a complex with pol II at about 18 Å resolution by electron microscopy (EM) (38). Tfg2, the second largest subunit of TFIIF, is divided in multiple, discrete domains, which lie in close proximity to promoter DNA in the minimal initiation complex model (Fig. 6). Tfg2 is homologous to bacterial sigma factor, which occurs in essentially the same location in a complex with bacterial RNA polymerase, where it contacts the “–10” element of the bacterial promoter and interacts with the nontemplate strand after unwinding of the promoter DNA (39).

Fig. 6.

Model of the pol II transcription initiation complex. Same as Fig. 3D, with the addition of electron microscope structures of TFIIE, TFIIF, and TFIIH. An exploded view is shown at the left, to reveal the individual structures and their relationship to the pol II structure more clearly. Regions of the second largest subunit of TFIIF corresponding to domains 2 and 3 of bacterial sigma factor are indicated (“2” and “3”).

The structure of TFIIE has also been determined in a complex with pol II by EM, in this case in projection at 15 Å resolution (40). TFIIE appears dumbbell-shaped, interacting with the pol II “jaws” at the downstream end of the active center cleft, in position to contact promoter DNA about 25 bp downstream of the transcription start site (Fig. 6). TFIIE recruits TFIIH to the transcription initiation complex (9, 4143), and TFIIH has been cross-linked to promoter DNA at the +25 position, as well as further upstream around the start site (44). The structure of TFIIH has been determined at 13 Å resolution by electron crystallography (45), but docking to the initiation complex model is not possible with currently available information. However, TFIIH can be placed approximately (Fig. 6), on the basis of interaction between its Tfb1 subunit and the Tfa1 subunit of TFIIE (43), whose locations within the structures revealed by EM have been inferred from other evidence (40, 45).

Although the placement of TFIIH in the initiation complex model is approximate, it is nonetheless informative with regard to TFIIH function. TFIIH contains two ATPase/helicases and is responsible for opening the promoter DNA double helix to a single-stranded form at the transcription start site. Two mechanisms of promoter opening have been proposed: direct action of the helicases on the DNA around the start site; and action at a distance, with the helicases situated downstream and rotating the DNA against a fixed point upstream (the TBP–TFIIB–pol II complex) until promoter opening occurs. To these two possibilities we would add a third, action at a distance which does not cause immediate promoter opening, but which induces negative supercoiling, enhancing the transient thermal opening (“breathing”) of the double helix; a transiently open region would occasionally be captured through interaction with pol II and, as explained below, also TFIIF. The initiation complex model, with TFIIH bound at a downstream site but capable of reaching upstream (Fig. 6), can accommodate both direct and indirect modes of TFIIH action and, indeed, both may occur.

A mechanism involving negative supercoiling might explain why eukaryotes require helicases to open the double helix for transcription, whereas bacteria do not. Small DNA circles isolated from living cells invariably contain about one negative supercoil for every 200 bp (46). In bacteria, this supercoiling is established and maintained by DNA gyrase (47). Eukaryotes, on the other hand, lack gyrase; supercoiling results instead from the wrapping of the DNA around histones in the nucleosome, with the DNA between nucleosomes maintained in a relaxed state by abundant DNA untwisting enzymes. Bacterial promoters therefore enjoy a significant frequency of transient thermal opening due to negatively supercoiling, whereas eukaryotic promoters require helicase activity to achieve a comparable, transiently open state. Support for this idea comes from the use of the immunoglobulin IgH heavy-chain promoter in a negatively supercoiled DNA circle for transcription initiation by a minimal system of TBP, TFIIB, and pol II. Similar results were obtained for 11 other mammalian promoters, except that TFIIF was also required, perhaps because of the involvement of its Tfg2 subunit in promoter opening, as described below. All 12 promoters were dependent on TFIIE and TFIIH for transcription in the absence of negative supercoiling (48). Some promoters, such as those in linear bacteriophage genomes, and those for eukaryotic pol I and pol III, are exceptional and require no negative supercoiling or helicase action, presumably because their sequences are especially conducive to transient thermal opening.

A bacterial promoter is captured in the transiently open state through interaction of its “–10” region with domain 2 of sigma factor. In eukaryotes, the corresponding domain of the Tfg2 subunit of TFIIF (labeled “2” in Fig. 6) may perform a similar role, but in the initiation complex model, this domain of Tfg2 contacts promoter DNA around position +5 to +10 (Fig. 6). The discrepancy may be resolved by movement of the DNA after promoter opening because of a number of factors, such as the expansion in length of DNA on conversion to the single-stranded form and the movement of TFIIBC across the pol II surface, with which it interacts only very weakly in the cocrystal structure.

Conclusions: The mechanism of pol II transcription initiation. The structure of the pol II–TFIIB complex provides the key for unraveling the mechanism of transcription initiation. It leads to an integrated picture, encompassing all general factors in the transcription initiation process. Our current understanding may be summarized as follows:

TBP bends TATA box DNA, creating a context for binding TFIIBC and a curvature of the DNA appropriate for wrapping around pol II. TFIIBC binds sequences immediately upstream and downstream of the TATA box, and TFIIBN binds to the dock domain, saddle, and active center cleft of pol II. In consequence of TFIIB binding, the promoter DNA is directed toward and above the active center region of pol II, running straight across the face of the initiation complex, passing the Tfg2 subunit of TFIIF above the cleft and also passing TFIIE and TFIIH at the down-stream end of the cleft. This path of the DNA is defined entirely by interactions with general transcription factors; there appears to be no contact of promoter DNA with pol II before DNA strand separation.

TFIIE interacts with the pol II jaws, possibly in consequence of a conformational change caused by pol II–TFIIB interaction. TFIIH, bound to TFIIE, causes transient promoter opening, through direct helicase action, or indirectly by the introduction of negative superhelical strain. A transiently open region is trapped by interaction of the nontemplate single strand with Tfg2 and the template strand with both the TFIIB finger domain and pol II in the active center cleft. It is at this point that promoter DNA first bends, because of the flexibility of the single-stranded region, and descends into the cleft. It is also at this point that the DNA first interacts with pol II, in the active center cleft, and with protein loops that define the boundaries of the transcription bubble: the “rudder” and “zipper” domains at the upstream end and fork loop 2 at the downstream end (4, 5).

TFIIB determines the location of the transcription start site at two levels, coarse setting by TFIIBC binding at the TATA box, and fine setting by the finger domain of TFIIBN interacting with the template DNA single strand. Coarse setting relies on the straightness and rigidity of B-form DNA, with the region downstream of the TATA box serving as a spacer. Fine setting may involve interaction of the fingertip with the initiator region of some promoters. Such interactions of the TFIIB finger domain stabilize initiation complexes containing RNA-DNA hybrid regions of less than 9 bp. TFIIBN competes with RNA of length greater than about 10 residues for occupancy of the pol II saddle, likely accounting for abortive initiation and the release of TFIIB during promoter escape.

Supporting Online Material

www.sciencemag.org/cgi/content/full/303/5660/983/DC1

Materials and Methods

Fig. S1

References

References and Notes

View Abstract

Navigate This Article