Three-Dimensional Structure of the Human TFIID-IIA-IIB Complex

See allHide authors and affiliations

Science  10 Dec 1999:
Vol. 286, Issue 5447, pp. 2153-2156
DOI: 10.1126/science.286.5447.2153


The multisubunit transcription factor IID (TFIID) is an essential component of the eukaryotic RNA polymerase II machinery that works in concert with TFIIA (IIA) and TFIIB (IIB) to assemble initiation complexes at core eukaryotic promoters. Here the structures of human TFIID and the TFIID-IIA-IIB complex that were obtained by electron microscopy and image analysis to 35 angstrom resolution are presented. TFIID is a trilobed, horseshoe-shaped structure, with TFIIA and TFIIB bound on opposite lobes and flanking a central cavity. Antibody studies locate the TATA-binding protein (TBP) between TFIIA and TFIIB at the top of the cavity that most likely encompasses the TATA DNA binding region of the supramolecular complex.

The accurate and regulated transcription of protein coding genes in all eukaryotic organisms requires the assembly at specific promoter elements of a complex molecular machine that includes general transcription factors in association with RNA polymerase II (RNA pol II) (1). Recognition of core promoter DNA sequences by TFIID (2) is followed by the assembly of a fully activated preinitiation complex that contains TFIIA, TFIIB, TFIIE, TFIIF, TFIIH, and RNA pol II (3). In the absence of activators bound to enhancer elements, this core transcription complex can accurately initiate basal levels of RNA synthesis. In the presence of gene-selective enhancer and promoter binding activators, significantly elevated levels of transcription initiation can be achieved. A key step in the multistep process of gene activation is the recruitment and assembly of the TFIID-IIA-IIB complex at the TATA DNA region that is found in many core promoters of eukaryotic genes. The TBP subunit and some of the TBP-associated factors (TAFII) subunits (TAFII250, TAFII150, and TAFII70) make specific contacts with the TATA box and other core promoter elements, including initiator (INR) and downstream promoter elements (DPE) (4). Other TAFII's of the multisubunit TFIID complex such as TAFII130, -100, -55, -32, -30, and -28 likely serve as targets of activation domains involved in the recruitment and stabilization of TFIID at core promoters by upstream enhancer binding factors (5–7).

The binding of TFIID to the core promoter is coordinated with the assembly of an active preinitiation complex that includes IIA and IIB. X-ray diffraction and nuclear magnetic resonance (NMR) studies have revealed the structures of various subdomains and truncated fragments of IIA and IIB and of subunits contained within TFIID. For example, the high-resolution structures of TBP bound to TATA DNA and to domains of either IIA or IIB have been determined (8). However, both the size of the complex and the inherent difficulties in obtaining large quantities of purified holo-TFIID, -IIA, and -IIB have precluded conventional x-ray diffraction studies of the full complex. Consequently, the overall shape and relative position of the components within the TFIID-IIA-IIB complex remain unknown. As a first step toward determining the structure of the intact native human TFIID, we used electron microscopy (EM) and single-particle image analysis to obtain the structure of TFIID and its complex with full-length IIA and IIB at 35 Å resolution.

Homogeneous preparations of TFIID suitable for EM studies were purified as described (9). For preparations of TFIID bound to IIB and IIA, recombinant IIB and IIA subunits were purified and reconstituted with antibody affinity-purified TFIID in vitro (10–12). The activity of purified TFIID, IIA, and IIB was confirmed by in vitro transcription assays (Fig. 1A). Preparations of holo-TFIID containing the full complement of TAFs were highly active in mediating Sp1-dependent transcription (Fig. 1B) when reconstituted with all the necessary general factors and the coactivator CRSP (11, 13). These functionally active preparations of TFIID, IIA, and IIB were used for EM analysis.

Figure 1

(A) Silver-stained SDS–polyacrylamide gel electrophoresis gels of recombinant IIA, IIB, and immunopurified native TFIID. The apparent molecular weight of each polypeptide is indicated to the left of each panel. (B) Reconstituted transcription reactions with the (GC)3-BCAT template, supplemented with TFIID or Sp1, or both, as indicated. The final products were analyzed by primer extension.

Electron microscopy grids of TFIID (∼45 ng/μl) were prepared and stained with uranyl acetate (4%). To obtain an initial three-dimensional (3D) reconstruction, we collected image pairs of tilted (32°) and untilted (0°) samples with the use of the random conical tilt method (14). Images were obtained under low-dose conditions, with the use of a JEOL 4000EX electron microscope at ×30,000 magnification with a defocus of 1 μm. Micrographs showed that the native human TFIID preparation was homogenous (Fig. 2A). Each micrograph contained ∼300 particles. The micrographs were digitized with a scan step of 21 μm (corresponding to a pixel size in the sample of 7 Å). Image analysis was carried out with the WEB and SPIDER software packages (15). A total of 4418 particle image pairs were windowed (64 pixels by 64 pixels) and were used for further processing. Untilted images were subjected to reference-free alignment and merged by performing in-plane shifts and rotations into 25 classes defined by K-means clustering (16). The large number of classes chosen reflects the fact that TFIID orients almost randomly on the carbon film. 3D structures for each of the 25 classes were calculated by back projection using the paired tilted images. Twenty-four of the classes showed excellent correlation (R > 0.85), thus a single 3D structure was obtained with the single particle–tilted images from these 24 classes. This initial structure served as a reference for angular refinement (17). Eighty-three reference projections were computed from the crude volume, with a difference in view angle of 15°. The untilted particle images were matched to the reference projections on the basis of highest cross-correlation by performing in-plane rotations and shifts. A refined volume was then calculated with the newly identified Euler angles. This procedure was repeated twice. Final angular refinement was performed by generating 5088 reference projections from the newly obtained structure with an angular step of 2°. The projection images of TFIID cover the full range of angular distribution (Fig. 2B). The final refined structure is depicted from six viewing angles (Fig. 3). The structure has been rendered to correspond to the molecular mass estimated for TFIID (775 kD). The reconstruction has a resolution of 35 Å, based on the 0.5 Fourier shell correlation criteria (18).

Figure 2

(A) Electron micrograph of negatively stained human TFIID. TFIID particles appear in a variety of orientations on the carbon support. (B) Angular distribution of the different projections (views) of TFIID used to generate the 3D structure shown in Fig. 3. Some of the different views are shown for reference. The diagram shows that the reconstruction of TFIID is free of missing cone artifacts. The same applies to the reconstructions obtained from TFIID-IIB and TFIID-IIA-IIB complexes (25).

Figure 3

3D reconstruction of human TFIID at 35 Å resolution. (A) and (B) are front and back views, respectively, showing the overall horseshoe structure of TFIID. The three lobes have been designated A, B, and C. (C) through (F) are side views obtained by turning the front view in (A) out of the paper by 90°, then rotating around a horizontal axis through the center of the cavity. The V-shaped channels between lobes are indicated by arrows.

The 3D reconstruction of TFIID reveals a horseshoe-shaped structure (roughly 200 Å by 135 Å by 110 Å) consisting of three main lobes, A, B, and C. These lobes are connected by narrower bridges (∼20 Å wide) and arranged around a cavity 65 Å in diameter (Fig. 3). The lobes are roughly equal in size (∼60 Å in diameter) but differ in structural detail. Above the connecting bridges and between the lobes are two V-shaped channels, one ∼25 Å across (large arrow in Fig. 3, D and E) and the other ∼15 Å across (small arrow in Fig. 3, D and F), that separate the lobes on the front face. The central cavity and its 40 Å open channel at the bottom of the horseshoe could easily accommodate a strand of double-helical DNA 20 Å in diameter.

The 3D structure of the binary complex, TFIID-IIB, was obtained by mixing the two factors in equimolar concentrations and incubating them for 1 hour at 4°C. Sample preparation, data collection, and image analysis were as described for holo-TFIID. As for TFIID alone, the dimeric TFIID-IIB complexes were randomly oriented on the carbon support of the EM grid. Four thousand individual particle images were aligned three-dimensionally by reference to the refined 35 Å structure of naked TFIID (the molecular weight of IIB is 4% that of holo-TFIID). Once Euler angles were assigned, the new particle images were back-projected to generate a reconstruction of the TFIID-IIB complex that was used as reference for further angular refinement to a resolution of ∼35 Å. The structures of TFIID and the TFIID-IIB complex were then normalized according to the TFIID protein density, scaled, and subtracted to generate a difference map corresponding to the density of IIB (green in Fig. 4A). Part of the density attributable to IIB lies within the TFIID density. This may reflect a structural change in TFIID upon binding of IIB, or a differential staining in the two samples. The difference density is positioned mainly on lobe B, close to the ∼15 Å bridge connecting it with lobe C.

Figure 4

Position of IIB and IIA on the TFIID structure and mapping of the TBP. The blue mesh corresponds to the holo-TFIID, with the A, B, and C lobes indicated. (A) The green mesh corresponds to the density difference between the holo-TFIID and the TFIID-IIB complex. (B) The magenta and green meshes show the density difference between the holo-TFIID and the trimeric complex TFIID-IIA-IIB. The density depicted in light green can be attributed to TFIIB by comparison with (A), and the magenta density therefore corresponds to IIA. (C) The yellow mesh shows the density difference between the holo-TFIID and TFIID that is bound to the TBP antibody.

The 3D structure of the ternary complex TFIID-IIA-IIB was obtained with the use of the same procedure as for the TFIID-IIB complex. Equimolar concentrations of the three components were incubated. A total of 5000 particle images were used to obtain a reconstruction at ∼35 Å resolution. The difference map obtained between TFIID and TFIID-IIA-IIB (Fig. 4B) contains two distinct regions of density: one corresponding to the location of the difference obtained in the presence of IIB alone (green) and a second one in lobe A consisting of a cluster of differentiated densities (magenta). This result confirms the position of IIB and identifies the location of IIA. Three main densities in the difference map can be attributed either to the binding of IIA or to conformational changes in TFIID upon binding of IIA. One of these densities constitutes a new bridge between lobes A and C, which occupies part of the central cavity. If this cavity is involved in binding the core promoter, then this bridge of density might affect the stability of the TFIID-DNA complex, in agreement with the enhanced DNA binding of TFIID by IIA (19). The subunits of IIA bind to TFIID on the opposite side of the central cavity relative to the IIB binding site, similar to the binding of IIB and IIA to the opposite “stirrups” of TBP (8). The location of IIA and IIB on different regions of the TFIID structure suggests that these factors could influence the overall shape of the preinitiation complex and, in particular, the shape of the putative DNA binding cavity, from two different ends of the complex.

The crystal structure of TBP bound to IIA and IIB suggested that TBP should bind in the central domain C, between these two general factors (8). To test this hypothesis, we employed a monoclonal antibody raised against TBP (anti-TBP) (20). TFIID–anti-TBP binding and preparation for EM were performed as described for the TFIID-IIB complex, using a 5:1 anti-TBP:TFIID ratio. More than 9000 particle images of the TFIID-anti-TBP complex were used to obtain a reconstruction at ∼30 Å resolution. Figure 4C indicates the position of the bound anti-TBP obtained in the difference map applying the same methodology we used to find the IIA and IIB binding positions. The binding site of anti-TBP indicates that TBP resides at the top of the cavity within the central domain C and faces the cavity in a position between IIA and IIB. We propose that TFIID binds the core promoter within the central cavity at the central lobe C through TBP.

The structure of the TFIID-IIA-IIB complex, together with the existing knowledge of the interaction between different TAFs and other components of the initiation complex, suggest a hypothetical distribution of TAFs on the three lobes of the TFIID structure. Lobe A is likely to contain TAF130, which is known to interact with IIA (21), whereas TAF32 might form part of lobe B because of its interaction with IIB (22). Because IIB is known to contact RNA pol II (23), we suspect that lobe B may be a major interaction region for polymerase. TAF250, which interacts with TBP and contributes multiple contacts for the assembly of the complex (7,24), is expected to form a major part of lobe C. Finally, TAF18, -20, and -28 may also be located in lobe C, because they have been reported to interact directly with TBP (7).

The EM studies presented here reveal a model for the structure of TFIID complexed to both IIA and IIB in the absence of DNA. Our findings provide a 35 Å structure of TFIID and the binding locations of IIA and IIB in relation to the larger TFIID. Antibody mapping of TBP within TFIID strongly suggests that the binding position of DNA is at the top of the central cavity within the TFIID complex. The horseshoe shape of TFIID and its trilobal structure suggest conformational flexibility around the narrow contact regions. The orientation between the lobes is probably regulated by transcriptional factors, and this would affect the shape of the central cavity and the binding to DNA. IIA and IIB reside on opposite sides of TFIID and are likely to stiffen the structure of TFIID by bridging different lobes. IIA binds across the 25 Å channel near the surface of the central cavity. Thus, IIA could increase the surface available for DNA binding and affect the affinity of TFIID-IIA for DNA that has been observed in DNA binding studies.

  • * To whom correspondence should be addressed. E-mail: enogales{at}


View Abstract

Navigate This Article