|Home | About | Journals | Submit | Contact Us | Français|
The electron density maps of the hydrated TFIID, the TFIID-TFIIA-DNA complex and the TFIID-TFIIA-Rap1-DNA Complex I and Complex II were deposited in the EM Database under accession codes EM-5175, EM-5178, EM-5176 and EM-5177, respectively.
Users may view, print, copy, download and text and data- mine the content in such documents, for the purposes of academic research, subject always to the full Conditions of use: http://www.nature.com/authors/editorial_policies/license.html#terms
Transcription of eukaryotic mRNA encoding genes by RNA polymerase II (Pol II) is triggered by the binding of transactivating proteins to enhancer DNA, which stimulates the recruitment of general transcription factors (GTFs; TFIIA, B, D, E, F, H) and Pol II on the cis-linked promoter leading to preinitiation complex (PIC) formation and transcription1. In TFIID-dependent activation pathways, this TATA box Binding Protein (TBP)-containing GTF is first recruited on the promoter through interaction with activators1-3 and cooperates with TFIIA to form a committed PIC4. However, neither the mechanisms by which activation signals are communicated between these factors, nor the structural organization of the activated PIC are known. Here we used cryo-electron microscopy to determine the architecture of nucleoprotein complexes composed of TFIID, TFIIA, the transcriptional activator Rap1 and yeast enhancer-promoter DNA. These structures revealed the mode of binding of Rap1 and TFIIA to TFIID, as well as a reorganization of TFIIA induced by its interaction with Rap1. We propose that this change in position increases the exposure of TBP within TFIID, consequently enhancing its ability to interact with the promoter. A large Rap1-dependent DNA loop forms between the activator binding site and the proximal promoter region, and this loop is topologically locked by a TFIIA-Rap1 protein bridge that folds over the DNA. These results highlight the role of TFIIA in transcriptional activation, define a molecular mechanism for enhancer-promoter communication and provide important new structural insights into the pathways of intramolecular communication that convey transcription activation signals through the TFIID complex.
Three-dimensional (3-D) electron microscopy (EM) has provided structural models for yeast TFIID5,6 that bears close similarities with its human ortholog7,8. TFIID can be divided into five modules5,7, lobes A, B, C1, C2 and D that adopt a clamp-like structure in which lobes A and B form the jaws (Fig. 1a). TFIID serves as a coactivator for yeast transcription factor Rap1 through direct interactions with TFIID subunits Taf4, 5 and 12. The Rap1 transactivator was used here as a model to study the architecture of a DNA-bound activator-TFIID-TFIIA complex9. Rap1 is a multifunctional protein that plays important roles in gene transcription and telomere length regulation, acts both as a repressor or activator on different genes10, drives transcription of over 40% of Pol II transcription in yeast, and is essential for transcription of the ribosomal protein genes used here as a model9,11. In order to elucidate the architecture of an activator-TFIID-promoter DNA complex we assembled three TFIID-containing complexes and solved their molecular structure to a resolution of 18.6 - 35 , mostly by utilizing cryo-EM methods (supplementary Table 1).
To visualize the Rap1 binding site on TFIID we first analyzed the structure of the TFIID-Rap1 complex formed as described in Methods with five-fold molar excess of Rap1. The recorded image dataset was analyzed and ultimately separated into two species, Rap1-TFIID complex, and holo-TFIID (supplemental material). A comparison of the two resulting models revealed that while the TFIID-Rap1 complex contained an additional density, the binding of Rap1 did not induce major conformational changes in TFIID, in agreement with a recent report describing the interaction of human TFIID with activators12. The additional Rap1-dependent protein density is found in TFIID lobe B (Fig. 1b and Supplementary Fig. 5a) where it colocalizes with Taf5 and the Histone Fold Domain containing heterodimer Taf4/Taf12, consistent with our immunolabelling experiments8,13 and with Rap1-Taf binding studies9,14, thus we conclude that the extra mass in the TFIID-Rap1 complex corresponds to Rap1 binding with Tafs 4, 5 and 12.
In order to position TFIIA and TFIID-DNA interactions in the absence of activator we recorded cryo-EM images of a TFIID-TFIIA-DNA complex reconstituted on the Adenovirus 2 Major Late Promoter (Ad2 MLP)15 with forty-fold molar excess of TFIIA and four-fold molar excess of DNA. Upon refinement, the image dataset was sorted to remove free, non-DNA bound TFIID molecules thus yielding 3-D models for the TFIID-TFIIA-DNA complex and holo-TFIID (Fig. 1c and Supplementary Fig. 4a and 5b). In the TFIID-TFIIA-DNA complex an additional globular density was bound to lobe C1, which showed a rod-shaped extension towards lobe D. The globular density is compatible with the size of TFIIA and positions close to the mapped location of TBP13 consistent with the known TBP-TFIIA interaction16,17. The rod-shaped density connects to lobe D and runs along its surface where we have mapped the C-terminus of Taf2 (see Supplementary Fig. 7). Since Taf2 is documented to interact with the Inr sequence at the transcription start site in the Ad2 MLP18,19, we propose that the extra density depicted in green in Fig. 1c corresponds to promoter DNA. The additional mass present in the TFIIA-TFIID-DNA complex thus contains TFIIA and a promoter DNA fragment connecting TBP to Taf2.
We next incubated TFIID, TFIIA and Rap1 with a chimeric yeast enhancer-promoter DNA fragment composed of the 41 bp Rap1 Upstream Activating Sequence of the ribosomal protein gene RPS8A (UASRAP1) fused to the PGK1 core promoter20; UASRAP1-PGK1, Supplementary Fig. 8), by using a ten-fold molar excess of the DNA fragment and five-fold excess of both TFIIA and Rap1. Importantly, this minimal construct which contains two Rap1 binding sites fused to the PGK1 promoter, has been well characterized. Transcription of this construct is both TFIID- and Rap1-dependent in vivo and in vitro9,20. A large dataset of 110,000 images was collected, aligned and classified according to particle orientation. Substantial heterogeneity was observed for similarly oriented particles indicating that the dataset contained a mixed population of complexes. Therefore 3-D multivariate statistical analysis was applied to separate different TFIID complexes21. This method reveals three dominant structures termed Complexes I, II and III. Complex III corresponds to free, unbound holo-TFIID, and serves here as an internal reference structure (Supplementary Fig. 4c and 5c,d).
Complex I contains TFIID, Rap1 and enhancer promoter DNA and is characterized by two additional densities bound to both faces of Lobe B (red, Fig 2a) positioned similarly to Rap1 within the TFIID-Rap1 complex. Importantly, the shape of the inner density, closest to lobe C1, accommodates the crystal structure of the Rap1 DNA Binding Domain (DBD)22 (Fig. 2b). This observation, indicating that Rap1 DBD binds directly to Lobe B, is consistent with biochemical and genetic evidence demonstrating that this central domain of Rap1 (residues 361 to 596) interacts with TFIID9. Moreover, this density shows a rod-shaped protuberance running towards and contacting lobe C2 (Lobe C2, green in Fig. 2a). When the atomic structure of the Rap1-DBD-DNA complex22 is fitted into the inner density the orientation and position of DNA overlaps the protruding rod suggesting that the narrow density connecting Rap1 on lobe B to lobe C2 corresponds to promoter DNA downstream of the Rap1-binding site. The shape of the density bound to the external face of lobe B, opposite to lobe D, is different from that of the Rap1-DBD and there is no protruding rod that could account for bound DNA, suggesting that it corresponds to other domains of Rap1. The 92 kDa Rap1 consists of DNA bending, BRCT, DBD, toxicity (Tox), AD and silencing (SD) domains23. Among these, both the DBD and the C-terminal portions of Rap1 (containing Tox, AD and SD domains) interact directly with the holo-TFIID complex, indicating that the C-terminal domain of Rap1 likely contributes to the observed external density9. In a control experiment we examined the structure of isolated Rap1 molecules and found that they adopt a two-lobed horseshoe shaped structure, consistent in size and shape with the two densities found in lobe B (Supplementary Fig. 10). We can however not completely rule out the possibility that this second density corresponds to a second Rap1 molecule. In addition to Rap1, Complex I contains an extra density associated with TFIID domain D near the Taf2 site (Lobe D, green in Fig. 2a). A similar density was detected at the same location in the TFIID-TFIIA-DNA complex suggesting that this density corresponds to DNA bound to Taf2. Thus, in Complex I, TFIID appears to contact DNA through Rap1 and Taf2, but apparently not through TBP and TFIIA, since no extra density was detected in lobe C1. Taken together these data suggest that complex I may correspond to an initial “recruitment mode” of binding, where TFIID interacts with the promoter-bound activator in the absence of TFIIA.
Complex II, a quaternary complex containing TFIID, Rap1, TFIIA and enhancer promoter DNA, is characterized by a continuous density between TBP and the Rap1 binding sites that bridges lobes C1 and B (Fig. 3a). This bridge is too large to be composed of DNA alone and therefore must also contain protein. We propose that the mass next to lobe C1 corresponds to TFIIA, as TFIIA binds to a similar site of TFIID in the TFIID-TFIIA-DNA complex, and because this extra mass accommodates the atomic structure of the TFIIA-TBP-DNA complex24 (Fig. 3b). Notably though, comparison of Complex II with the TFIID-TFIIA-DNA complex, indicates that the position of TFIIA is rotated 130° around its TFIID interaction site (Fig. 3c). The repositioning of TFIIA within Complex II brings it close enough to directly interact with the DBD of Rap1, which is located on the inner face of lobe B as in Complex I. The proposed position of the Rap1 DBD is supported by the docking of its atomic structure into the remaining part of the protein bridge (Fig. 3b). Interactions between activators and TFIIA have been described25-28 and in one case TFIIA was shown to be required to release TBP from DNA binding autoinhibition mediated by the N-terminal TAND domain of Taf129, whereas in the other instances TFIIA was described to stimulate activator dependent transcription by interacting with the activator, the situation we observe here in Complex II. This reorganization of TFIIA within the TFIID-activator-promoter complex likely affects the position and therefore the accessibility (and functionality) of TBP, contributing to both increased TFIID-promoter interaction, PIC formation, and ultimately initiation efficiency. The exact path of promoter DNA cannot be traced in Complex II, though several extra densities signal its position. A patch of density is found on the inner wall of the clamp between lobes C1 and D, and a second more robust density is located in lobe D as observed in both Complex I and the TFIID-TFIIA-DNA complex (Lobe D, green in Fig. 3a). We interpret these stretches of density to correspond to the bound promoter DNA between the TATA box that interacts with TBP and the DNA close to the transcription start site bound by Taf2.
The DNA between the TATA-box and the Rap1 binding sites was not detected, most likely because of the flexibility of this segment. The proposed arrangement of the Rap1 DBD and TBP/TFIIA complexes implies that the intervening DNA loops out away from TFIID. To test this hypothesis we formed complexes on the yeast enhancer-promoter DNA fragment with TFIID and TFIIA in the presence and absence of Rap1. The resulting complexes were visualized after platinum shadowing and clearly revealed loops of DNA protruding from TFIID (Fig. 3d). Rap1 plays an essential role in loop formation since in the absence of Rap1, loops were observed in less than 2% of the DNA-bound TFIID complexes, while when Rap1 was added, 35% of the DNA bound TFIID molecules had DNA loops. Similar Rap1-dependent loop formation is observed on the natural RPS1A gene (Supplementary Fig. 9). Moreover TFIIA mutants unaffected in TBP binding but showing impaired transcription show a reduced ability to form DNA loops (Supplementary Fig. 11 and 12). These observations demonstrate that the Rap1 activator favours the formation of DNA loops allowing communication with distant TFIID-bound promoter sequences.
We propose that Complexes I and II represent functional intermediates in the pathway leading to PIC formation. From the molecular snapshots we have captured we infer a possible mechanism for activated TFIID-TFIIA-DNA complex formation (Fig.4). In a first step, Rap1 interacts with its recognition element (Fig. 4a) and this Rap1-enhancer complex can simultaneously bind TFIID (Fig. 4b). In all complexes analyzed, the Taf2–containing D lobe was found to interact with the DNA template indicating that in the absence of detectable interaction with TBP or TFIIA, the DNA fragment is already looped out between the Rap1-binding sites and the proximal promoter (Fig. 4c and Supplementary Fig. 12). Whereas the role of TFIIA in releasing TBP autoinhibition in basal transcription is well established, its contribution to activated transcription at the molecular level is less well understood. The discovery of a class of TFIIA mutants that stimulate TBP-DNA binding but fail to support activation favours a model in which TFIIA acts in two mechanistically distinct activation steps27 consistent with the structures reported here (Fig. 4c,d,e), where we propose that the action of the activator is mediated by a protein bridge between lobes B and C1 of TFIID through an interaction between TFIIA and the Rap1 DBD (Fig. 4d). Thus activator-induced repositioning of TFIIA, and probably of its interaction partner TBP, may affect the accessibility of the DNA binding surface of TBP thereby facilitating functional PIC formation and activation of transcription (Complex II, Fig. 4d). In addition our model predicts that the bridge closes over the TBP-bound DNA and topologically locks the proximal promoter DNA in the resulting clamp (Complex I to Complex II; Fig. 4c,d). Such a trapping process could result in an increased residence time of the promoter DNA within TFIID and participate in the activation mechanism. Collectively our results support a role for TFIID as an assembly platform that plays an active and important role in PIC formation and transcription. Our data provide new structural insights into how, Rap1 collaborating with TFIIA, transduces activating intramolecular signals within the TFIID coactivator complex that ultimately can lead to PIC formation. Further, the structures reported here highlight the complex network of protein-protein and protein-DNA interactions regulating activated transcription.
HA- and TAP-TFIID were purified from Saccharomyces cerevisiae as described5,9,15. The TFIID-Rap1 complex was assembled from TAP-tagged TFIID and a 5-fold molar excess of Rap1. The sample was brought to a TFIID concentration of 30 μg/ml, crosslinked with 0.1% glutaraldehyde for 5 sec and mounted as described5. A transmission electron microscope (TEM, Philips CM120) equipped with a LaB6 cathode and operating at 120 kV was used to collect images with a CCD camera (Model 794, Gatan, Pleasanton). The TFIID-TFIIA-DNA complex was assembled from TAP-tagged TFIID with a 40-fold molar excess of TFIIA and 4-fold molar excess of a 400 bp long Ad2 MLP fragment (Supplementary Fig. 4). The specimen was crosslinked and adsorbed on a thin carbon film sustained by a holey carbon grid and plunge-frozen into liquid ethane. Images were recorded on a cryo-TEM (FEI Tecnai F20) and were digitized as described5. To form the TFIID-TFIIA-Rap1-DNA complex, HA-tagged TFIID was incubated with a 10-fold molar excess of a 282 bp UASRAP1-PGK1 DNA fragment (Supplementary Fig. 4)9 and a 5-fold excess of both TFIIA and Rap1. The specimen was prepared as for the TFIID-TFIIA-DNA complex and data was collected on a cryo-TEM (FEI Tecnai Polara) operating at 300 kV. Images were selected and processed as described earlier5. Resolution was estimated according to the half bit criterion30 and the final reconstructions were filtered accordingly. During refinement images of the TFIID-Rap1 and TFIID-TFIIA-DNA complexes were split into holo-TFIID and multicomponent complexes. Analysis of the TFIID-TFIIA-Rap1-DNA complex was similar except for a 3-D multivariate statistical analysis and clustering step that was performed to separate distinct conformational states according to ref 24.
This work was supported by grants from the Institut National de la Santé et de la Recherche Médicale, the Centre National pour la Recherche Scientifique, the Association pour la Recherche sur le Cancer (ARC), the Fondation pour la recherche médicale (FRM), the european SPINE program, QLG2-CT-00988 (G.P., C.R., P.S.), and the National Institutes of Health (NIH GM52461; M.K.T., J.H.L. and P.A.W.).
HA1-Taf1- and TAP-Taf1-tagged TFIID were purified from Saccharomyces cerevisiae as described previously5,15,31. Rap1 and TFIIA were expressed in E. coli and purified as described9,15,16. Two promoter-containing DNA fragments were used to form complexes. The TFIID-TFIIA-DNA complex was assembled on a 400 bp long fragment of the Ad2 MLP, while a 282 bp long UASRAP1-PGK1 chimeric yeast promoter (Supplementary Fig. 4) was used for the assembly of the TFIID-TFIIA-Rap1-DNA complex. This promoter fragment contains the 41 bp UASRAP1 element, derived from the ribosomal protein-encoding RPS8A gene (from −252 to −212), which contains two binding sites for Rap1 (CTTTACATCCATACACCCTCTTTAACACCCTTACACTTTTA; Rap binding sites bold, underlined) fused to the PGK1 core promoter (−211 to +30)9,20,32.
The final concentration of TFIID used in the negative stain and cryo-EM experiments was 30 μg/ml and of 50 μg/ml respectively. The crosslinked samples were treated with 0.1% glutaraldehyde for 5 sec prior to adsorption on a thin carbon film. To assemble the TFIID-Rap1 complex Tap-tagged TFIID was incubated with a 5-fold molar excess of Rap1 for 30 min on ice in 20 mM HEPES pH 7.9, 250 mM NaCl, 1 mM DTT, 0.2 mM PMSF and 10% glycerol. The sample was adsorbed onto air glow-discharged grids covered with a 10 nm thick carbon film and sandwiched with a second carbon film after negative staining with 2% uranyl acetate. A transmission electron microscope (TEM, Philips CM120) equipped with a LaB6 cathode and operating at 120 kV was used to collect images at 45,000X magnification on a Pelletier-cooled slow scan CCD camera (Model 794, Gatan, Pleasanton, CA) resulting in a pixel spacing of 0.37 nm on the object. To assemble the TFIID-TFIIA-DNA complex, TAP-tagged TFIID was incubated for 20 min at 20°C with a 40-fold molar excess of TFIIA and 4-fold molar excess of the Ad2 MLP in 10 mM HEPES pH 7.4, 60 mM KCl, 6 mM MgCl2 and 2 mM DTT. The specimen was adsorbed on a holey carbon grid covered with a 3-4 nm thick carbon film. Images were recorded on a cryo-TEM (FEI Tecnai F20) equipped with a field emission gun (FEG) and operating at 200 kV. Images of well dispersed individual complexes were recorded at liquid nitrogen temperature on Kodak SO-163 films at 40,000 X magnification and in low dose conditions (15-20 e-/Å2). Negatives were digitized with a 5 μm raster size using a drum scanner (Primescan D7100, Heidelberg) and were coarsened twice to obtain a pixel spacing of 0.254 nm on the object. For the TFIID-TFIIA-Rap1-DNA complex, HA-tagged TFIID was incubated 30 min at 20 °C with a 10-fold molar excess of UASRAP1-PGK1 enhancer-promoter DNA, a 5-fold excess of TFIIA and a 5-fold excess of Rap1 in 10 mM Tris-HCl pH 7.9, 170 mM KOAc and 5 mM MgCl2. The specimen prepared as detailed above for the TFIID-TFIIA-DNA complex, was vitrified in Vitrobot (FEI) and observed with a cryo-TEM (FEI Tecnai Polara) equipped with a FEG (field emission gun) operating at 300 kV. Images were collected at liquid nitrogen temperature under low-dose condition (15-20 e/Å2), at a magnification of 39,000 X on Kodak SO-163 films. The pixel spacing of the digitized negatives was of 0.26 nm.
DNA loop formation was visualized after absorption of the DNA-protein complexes onto air glow-discharged grids that were positively stained with uranyl-acetate and platinum shadowed after air-drying.
Boxing of the images of the TFIID-Rap1 and TFIID-TFIIA-DNA complexes was performed with the EMAN software33 package whereas the images of TFIID-TFIIA-Rap1-DNA complex were selected with the boxer2 option of EMAN234. The contrast transfer function (CTF) of the microscope was estimated using Bsoft:Bshow35 and the images were corrected by phase flipping. Image processing was performed using the IMAGIC36 (Image Science Software, Berlin, Germany) and Spider37 software packages as described earlier5. The resolutions of the final reconstructions were estimated according to the intersection point of the half bit curve with the FSC curve (half bit criterion)30. The final reconstructions were filtered to the estimated resolution.
In order to improve specimen homogeneity, TFIID-containing complexes that did not contain Taf2 were removed from the dataset5. During refinement, images of the TFIID-Rap1 and TFIID-TFIIA-DNA complex were split into holo-TFIID and complex. To do so, images were sorted for their best cross-correlation with reprojections of either the mixed (holo-TFIID + complex) model or the holo-TFIID reference model. This separation was iterated several times and resulted in a progressive enrichment of the mixed model in TFIID-Rap1 or TFIID-TFIIA-DNA complexes. The analysis of TFIID-TFIIA-Rap1-DNA complex was similar except for a 3-D statistical analysis and clustering step that was performed to separate distinct conformational states according to ref 24. Briefly, a large number of 3-D models were reconstructed from a few randomly selected and pre-aligned class average images. This repertoire of 3-D models was subjected to multivariate statistical analysis and was clustered into groups corresponding to different conformations of the complex. The class-sum volumes, characteristic for each conformation, were used as references for subsequent refinement rounds. Fitting of atomic coordinates into EM density maps were performed using UCSF Chimera’s fit in map tool38 and Sculptor39. Images were created with UCSF Chimera.