Architecture of TAF11/TAF13/TBP complex suggests novel regulatory state in General Transcription Factor TFIID function

General transcription factor TFIID is a key component of RNA polymerase II transcription initiation. Human TFIID is a megadalton-sized complex comprising TATA-binding protein (TBP) and 13 TBP-associated factors (TAFs). TBP binds to core promoter DNA, recognizing the TATA-box. We identified a ternary complex formed by TBP and the histone fold (HF) domain-containing TFIID subunits TAF11 and TAF13. We demonstrate that TAF11/TAF13 competes for TBP binding with TATA-box DNA, and also with the N-terminal domain of TAF1 previously implicated in TATA-box mimicry. In an integrative approach combining crystal coordinates, biochemical analyses and data from cross-linking mass-spectrometry (CLMS), we determine the architecture of the TAF11/TAF13/TBP complex, revealing TAF11/TAF13 interaction with the DNA binding surface of TBP. We identify a highly conserved C-terminal TBP-binding domain (CTID) in TAF13 which is essential for supporting cell growth. Our results thus have implications for cellular TFIID assembly and suggest a novel regulatory state for TFIID function.


46
Eukaryotic gene expression is a highly regulated process which is controlled by a plethora of 47 proteins, arranged in multiprotein complexes including the general transcription factors 48 (GTFs), Mediator and RNA polymerase II (Pol II) (Gupta et al. 2016; Thomas and Chiang 49 2006). Regulated class II gene transcription is initiated by sequential nucleation of GTFs and 50 Mediator on core promoter DNA (Rhee and Pugh 2012). The GTF TFIID is a cornerstone in 51 this process and links cellular signaling events with regulatory DNA elements and the 52 components of the transcription machinery (Albright and Tjian 2000). A basal transcription 53 system which supports initiation can be reconstituted with TBP and the GTFs TFIIA, TFIIB, 54 TFIIE, TFIIF and TFIIH in vitro, however, TFIID is required to respond to activators 55 (Hampsey and Reinberg 1999). In mammalian cells, the promoters of virtually all protein-56  (Kim et al. 1993;Nikolov et al. 1996;Nikolov et al. 1992). 86 The DNA-binding activity of TBP/TFIID is tightly regulated by gene-specific co-87 factors that can activate or inhibit transcription (Koster et al. 2015; Tora and Timmers 2010). 88 The mechanism of a number of these regulatory factors has been described in molecular 89 detail. The TFIID component TAF1 was found to associate with the concave DNA-binding 90 surface of TBP via its N-terminal domain (TAF1-TAND), exhibiting TATA-box mimicry 91 (Anandapadamanaban et al. 2013). TAF1-TAND, unstructured in isolation, was found to 92 adopt a three-dimensional structure closely resembling the TATA-element is shape and 93 charge distribution when bound to TBP (Burley and Roeder 1998;Liu et al. 1998). This 94 interaction is conserved in yeast, Drosophila and human (Anandapadamanaban et al. 2013; 95 Burley and Roeder 1998;Liu et al. 1998; Mal et al. 2004). The recent high-resolution 96 structure of TBP bound to yeast TAF1-TAND revealed anchoring patterns in transcriptional 97 regulation shared by TBP interactors, providing insight into the competitive 98 multiprotein TBP interplay critical to transcriptional regulation (Anandapadamanaban et al. 99 2013). Mot1 is an ATP dependent inhibitor of TBP/TATA-DNA complex formation (Auble 100 and Hahn 1993). Mot1 regulates the genomic distribution of TBP and was shown to influence 101 transcription levels both positively and negatively (Pereira et al. 2003). Recent structural 102 analysis revealed the molecular mechanism of Mot1 wrapping around TBP resembling a 103 bottle opener, with a 'latch helix' blocking the concave DNA-binding surface of TBP and 104 acting as a chaperone to prevent DNA re-binding to ensure promoter clearance (Wollmann et 105 al. 2011). Mot1 and negative co-factor 2 (NC2) are thought to cooperate in gene-specific 106 repression of TBP activity (Hsu et al. 2008). The GTF TFIIA, on the other hand, competes 107 with NC2 for TBP (Kamada et al. 2001; Xie et al. 2000) and promotes TBP/DNA interactions 108 in a ternary TFIIA/TBP/DNA complex, facilitating formation of and stabilizing the 109 preinitiation complex (PIC). Interaction of TFIIA with TBP results in the exclusion of 110 negative factors that would interfere with PIC formation, and TFIIA acts as a coactivator 111 assisting transcriptional activators in increasing transcription levels (Bleichenbacher et al. 112 2003). 113 Genetic and biochemical experiments suggested that the TFIIA/TBP/DNA complex is 114 further stabilized by the histone-fold containing TFIID subunits TAF11 and TAF13 115 conveying the formation of a TAF11/TAF13/TFIIA/TBP/DNA assembly (Kraemer et al. 116 2001; Lavigne et al. 1999; Robinson et al. 2005). We therefore set out to investigate this 117 putative pentameric complex in detail. Unexpectedly, we did not observe a stabilization of 118 TFIIA/TBP/DNA complex by TAF11/TAF13 but found a marked destabilization of the 119 TFIIA/TBP/DNA complex by TAF11/TAF13, resulting in the release of free DNA and the 120 formation of a stable ternary complex formed by TAF11/TAF13 and TBP. We analysed the 121 TAF11/TAF13/TBP complex biochemically and structurally utilizing a comprehensive, 122 integrative approach. We report a novel interaction of the TAF11/TAF13 dimer binding to 123 the concave DNA-binding groove of TBP, thus excluding TATA-box containing DNA. 124 Using pull-down experiments with immobilized TAF1-TAND, we demonstrate competition 125 between TAF11/TAF13 and TAF1-TAND for TBP binding. We identify a novel C-terminal 126 TBP-binding domain (CTID) within TAF13 which is highly conserved from yeast to man. 127 We probe key residues within this TAF13 CTID by mutagenesis in vitro and in vivo in cell 128 growth experiments, revealing a key role of this domain for viability. We contrast the 129 TAF11/TAF13 interaction with other TBP DNA-binding groove interactors including Mot1 130 and discuss the implications of our findings in the context of TFIID assembly. Based on our 131 results, we propose a novel, functional state of TFIID in transcription regulation. 132

Identification of a novel TAF11/TAF13/TBP ternary complex 135
We set out to analyze the structure of a putative pentameric TAF11/TAF13/TFIIA/TBP/DNA 136 complex (Kraemer et al. 2001;Lavigne et al. 1999;Robinson et al. 2005), with the objective 137 to better understand the possible roles of TAF11/TAF13 in TFIID function. First we purified 138 human TAF11/TAF13 complex and TBP to homogeneity ( Figure 1A). TFIIA in human cells 139 is made from two precursor polypeptides, TFIIA and TFIIA, with TFIIA processed in 140 vivo into two separate polypeptides,  and , by proteolytic cleavage mediated by the 141 protease Taspase1 (Hoiby et al. 2007). Recombinant human TFIIA is typically produced in E. 142 coli by refolding from three separate polypeptides expressed in inclusion bodies, which 143 correspond to the native and chains (Bleichenbacher et al. 2003). To facilitate 144 recombinant human TFIIA production, we designed a single-chain construct (TFIIA s-c ) by 145 connecting ,  and by flexible linkers, based on atomic coordinates taken from the crystal 146 structure of human TFIIA/TBP/DNA complex (Bleichenbacher et al. 2003). TFIIA s-c could 147 be produced in high amounts in soluble form in E. coli and purified to homogeneity without 148 any need for refolding steps (see Methods section). We stored highly purified TFIIA s-c at 4⁰C, 149 and observed the formation of needle-shaped crystals in the storage buffer after several 150 weeks. We improved the crystals manually and determined the 2.4Å crystal structure of 151  Table 1). The crystal structure revealed a 152 virtually identical conformation of unliganded TFIIA s-c as compared to TFIIA in the 153 TFIIA/TBP/DNA complex. Moreover, the crystal structure highlighted the importance of the 154 connecting loops we had introduced in stabilizing the three-dimensional crystal lattice 155 ( Figure 1-Figure Supplement 1). TFIIA s-c was active in a band-shift assay with TBP and 156 adenovirus major late promoter (AdMLP) TATA-DNA ( Figure 1A), similar to purified 157 TFIIA using the classical refolding protocol (Bleichenbacher et al. 2003). 158 Next, we attempted to reconstitute the TAF11/TAF13/TFIIA/TBP/DNA complex 159 following published procedures (Kraemer et al. 2001;Robinson et al. 2005). Titrating 160 TAF11/TAF13 dimer to a preformed TFIIA/TBP/DNA complex had been reported to 161 stabilize TFIIA/TBP/DNA in band-shift assays using AdMLP TATA-DNA (Robinson et al. 162 2005). Surprisingly, in our titration experiments, TAF11/TAF13 did not stabilize the 163 preformed TFIIA/TBP/DNA complex but, in marked contrast, resulted in TAF11/TAF13-164 dependent release of free promoter-containing DNA in band-shift assays ( Figure 1B 1C). We concluded that human TAF11/TAF13 did not further stabilize the preformed 171 TFIIA/TBP/DNA complex, but rather sequestered TBP from this complex giving rise to a 172 novel assembly comprising TAF11, TAF13 and TBP. 173 TATA-DNA and TAF1-TAND compete with TAF11/TAF13 for TBP binding  174 We tested competition between TAF11/TAF13/TBP formation and TBP binding to AdMLP 175 DNA and showed that the ternary TAF11/TAF13/TBP complex remained stable in the 176 presence of excess AdMLP DNA ( Figure 1D). Thus, our results indicate that TAF11/TAF13 177 and TATA-DNA compete for at least parts of the same binding interface within TBP, and 178 that once TAF11/TAF13 is bound to TBP, TATA-DNA binding is precluded. 179 TAF1 had been shown previously to bind to the DNA-binding surface of TBP via its 180 TAND domain (Anandapadamanaban et al. 2013;Burley and Roeder 1998;Liu et al. 1998). 181 We produced human TAF1-TAND fused to maltose-binding protein (MBP) and immobilized 182 highly purified fusion protein to an amylose column (Methods). We added preformed, 183 purified TAF11/TAF13/TBP complex in a pull-down assay using immobilized TAF1-TAND. 184 We found that TAF1-TAND effectively sequestered TBP from the TAF11/TAF13/TBP  185   complex, evidenced by TAF11/TAF13 eluting in the flow-through fraction. Elution by  186 maltose, in contrast, revealed a TAF1-TAND/TBP complex ( Figure 1E). Together, these 187 findings substantiate our view that TAF11/TAF13, TAF1-TAND and AdMLP TATA-DNA 188 all interact with the concave DNA-binding surface of TBP, and that the interactions are 189 mutually exclusive. The extent of protection within TBP further indicates that the binding of TAF11/TAF13 223 engages both symmetric pseudo-repeats in TBP, thus spanning the entire concave interface 224 ( Figure 2D). Interestingly, we identified one peptide (AA residues 157-167) in TBP which 225 evidenced an increased level of deuteration upon TAF11/TAF13/TBP complex formation in 226 the HDX-MS experiments ( Figure 2D). This peptide is located at the dyad relating the two 227 pseudo-symmetric repeats in TBP. We interpret this result as an indication that this particular 228 region within TBP is more protected in a presumed TBP dimer which dissociates when 229 TAF11/TAF13 is binding and the 1:1:1 complex is formed. Our HDX-MS experiments 230 provide direct evidence that TAF11/TAF13 engage to the concave DNA-binding surface of 231 TBP, in excellent agreement with our above described biochemical experiments involving 232 TAF11/TAF13, TBP, AdMLP DNA and TAF1-TAND. 233

Architecture of the TAF11/TAF13/TBP complex 234
We proceeded to determine the architecture of the TAF11/TAF13/TBP complex by using a 235 comprehensive, integrative multi-parameter approach. We utilized the available crystal 236 structure of TBP (Nikolov et al. 1992) as well as the crystal structure of the globular histone-237 fold containing domains of the TAF11/TAF13 dimer (Birck et al. 1998) in our approach, and 238 combined these atomic coordinates with our native MS, SAXS, AUC and HDX-MS results. 239 We acquired distance constraints to define our structural model by carrying out cross-240 linking/mass-spectrometry (CLMS) experiments using two different approaches. We first 241 approach, we carried out calculations using alternative starting models. For instance, we 261 rotated TBP by 180⁰ around its axes to artificially expose the convex surface to 262 TAF11/TAF13, or, alternatively, to reverse the location of the N-and C-terminal stir-ups of 263 TBP (data now shown). These alternative calculations were far inferior in accommodating 264 experimental spatial and distance restraints, in addition to being inconsistent with our 265 biochemical data, thus substantiating our TAF11/TAF13/TBP structural model. In case of Mutant B, on the other hand, TBP interaction was likewise diminished, however, 284 residual TAF11/TAF13/TBP complex formation was clearly observed ( Figure 4B). These 285 resultsprovide evidence for a C-terminal TBP interaction domain (CTID) in TAF13, which is 286 highly conserved throughout evolution. As the human TAF13 CTID is very well conserved, 287 we generated two mutants in Saccharomyces cerevisiae (sc) Taf13 CTID that were mutating 288 the same amino acid residues that were deleterious in the human TAF11/TAF13/TBP 289  Figure 4D). Thus, our observations consistently suggest that the amino 302 acids in TAF13 CTID, which when mutated destroy TAF11/TAF13 interactions with TBP, 303 are required for functional TFIID formation. 304

Co-immunoprecipitation experiments reveal cytoplasmic TAF11/TAF13 and TBP 305 dynamics in nuclear holo-TFIID 306
We recently demonstrated that human TFIID assembly involves preformed cytosolic and 307 nuclear submodules (Trowitzsch et al. 2015), and we now asked whether the human 308 TAF11/TAF13/TBP complex would likewise represent such a sub-assembly. To this end we 309 performed co-immunoprecipitations (co-IPs) from HeLa cell cytosolic and nuclear extracts 310 using an anti-TAF11 antibody ( Figure 5). We found dimeric TAF11/TAF13 complex in the 311 cytosol representing the complete HF pair. We could not detect TBP in cytosolic co-IPs, 312 however, our experiments evidenced TAF7 association with cytoplasmic TAF11/TAF13. The 313 anti-TAF11 co-IP from nuclear extract, in contrast, contained all TFIID components. given TFIID complex, the TAF1-dependent and the TAF11/TAF13 HF pair-dependent TBP 374 blocking activities are mutually exclusive, but they may compete with each other to ascertain 375 full blocking activity. Interestingly however, it appears that this TAF11/TAF13 HF pair-376 dependent TBP binding/blocking activity is essential/required for normal TFIID function, 377 because when we interfered in the TBP binding through mutating the CTID, yeast growth 378 was compromised at the non-permissive conditions. Thus, it is not clear at the moment 379 whether or not the TAF1-dependent and the TAF11/TAF13 HF pair-dependent TBP blocking 380 activities are really competing with each other, or would be simply part of a step-wise TFIID 381 conformational change, or "activation", process that would allow TFIID to bind to DNA only 382 when open promoter structures would become available. Further experiments will be needed 383 to answer these exciting questions. Transcription activators and chromatin remodeling factors 384 may direct inhibited TFIID to specific promoters, which could be poised to be transcribed by 385 histone H3K4 trimethylation, and alleviate the TBP-blocking through TAF-interactions or by 386 TAF-chromatin mark interactions. Alternatively, it is conceivable that once TFIID is brought 387 to a promoter by interactions with transcription activators and positive chromatin marks (i.e. 388 histone H3K4me3), DNA and TFIIA together may synergize to liberate the TATA-box 389 binding surface of TBP from the inhibitory TAF-interactions. 390 The general roles of individual TAFs and the holo-TFIID complex are increasingly 391 better understood, the mechanisms by which the cell assembles this essential multiprotein 392 complex however remains largely enigmatic. The existence of discrete TFIID subassemblies 393 containing a subset of TAFs, such as nuclear core-TFIID and the TAF2/TAF8/TAF10 394 complex present in the cytoplasm, provides evidence that holo-TFIID may be assembled in a 395 regulated manner in the nucleus from preformed submodules (Bieniossek et  integrate into a core-TFIID and TAF2/TAF8/TAF10 containing "8TAF" assembly 408 (Trowitzsch et al. 2015) in the formation pathway to the complete nuclear holo-TFIID 409 complex. In this TAF1/TAF7/TAF11/TAF13/TBP module, TBP would be tightly bound to 410 either TAF1 or TAF11/TAF13, which could serve to ascertain that this putative TFIID 411 submodule is efficiently blocked from any potentially detrimental interactions with DNA 412 until holo-TFIID formation is completed. 413 In the nucleus, IPs utilizing antibodies against several different TFIID specific TAFs 414 co-precipitated all known TFIID subunits, although with variable stoichiometry. Strikingly, 415 stoichiometry analyses carried out by NSAF calculations of our nuclear anti-TAF IPs 416 indicated that TBP was only present in less than half of the TFIID specimens, when 417 compared to TAF1 or TAF7 for example, suggesting that TFIID-type complexes may exist 418 which do not contain TBP. TBP is thought to be highly mobile structurally in the context of  (Table 2) and SAXS Sample Buffer (25mM Tris pH 8.0, 300 mM NaCl, 1mM 569 EDTA, 1mM DTT and complete protease inhibitor) were exposed to X-rays and scattering 570 data collected using the robotic sample handling available at the beamline. Ten individual 571 frames were collected for every exposure, each 2 seconds in duration, using the Pilatus 1M 572 detector (Dectris AG). Data were processed with the ATSAS software package (Petoukhov et  software was used for data acquisition. The HD Examiner software (Sierra Analytics) was 664 used for HDX-MS data processing. Identification of peptides generated by the digestion was 665 done as described previously (Giladi et al. 2016). Different proteases (pepsin, nepenthesin-1, 666 nepenthesin-2, rhizopuspepsin) or their combinations were tested for protein digestion with 667 pepsin-nepenthesin-1 pair providing the best digestion parameters and sequence coverage. 668 Integrative multiparameter-based model building and refinement. Initial models of the 669 two component structures (TAF11/TAF13, TBP) were taken from the PDB (1BH8 and 670 1CDW) (Birck et al. 1998;Nikolov et al. 1996). 1BH8 was extended to include a helix 671 structure missing from the complete histone-fold domain as described before (Birck et al. 672 1998). The structure of the complex was constructed in a two-stage workflow. Initially, a 673 model of the structured core of the complex was constructed by rigid body docking using the 674 HADDOCK webserver (de Vries et al. 2010). The resulting complex structures and their 675 scores were visually analysed against the SAXS data to select the highest scoring structure 676 that fit within the SAXS envelopes. 677 The selected complex with the highest scores was then refined integrating the cross-678 linking data. The HADDOCK complex was used as an input to MODELLER 9.14 (Webb and 679 Sali 2014) with the complete sequences (including loop structures). Observed cross-links 680 were included as restraints in the refinement with a mean distance of 11.4 Å. Refinement was 681 performed iteratively until more than 90% of all distance constraints could be accommodated 682 while maintaining the fit to the SAXS envelope. 683

Cell growth experiments 684
Yeast Taf13 wild-type (WT), as well as Mutant A and Mutant B, were cloned along with 685 native promoters into the LEU2 (auxotrophic marker) containing plasmid pRS415 (Genscript 686 Corp., Piscataway, NJ, USA) by using the BamHI and NotI restriction enzyme sites. 687 Constructs thus generated were transformed into yeast strain BY4741 (comprising 688 endogenous wild-type Taf13) as well as the temperature sensitive (ts) yeast strains TSA797 689 (ts taf13) and TSA636 (ts taf13) (EuroSCARF, SRD GmbH, Germany). Transformed yeast 690 containing the plasmids were restreaked onto selective media and grown at permissive (30°C) 691 or non-permissive (37°C) temperatures, and plates imaged. To determine growth rates, ts 692 strains transformed with empty vector or Taf13 expression plasmids were grown in liquid 693 media at 37°C. In a separate experiment, the above constructs were transformed into a Taf13  The inset shows the native MS spectrum recorded for separately purified TBP alone, 877 evidencing a TBP dimer. 878  Table  886 2. 887 the cross-linked TAF11/TAF13/TBP sample. Bands were excised, reduced, alkylated, trypsin 895 digested and desalted followed by mass spectrometric analysis. Cross-links were identified 896 between TAF11 and TAF13, TAF11 and TBP as well as TAF13 and TBP. Cross-links 897 between TAF11-TAF13, TAF11-TBP and TAF13-TBP are shown using 5% FDR cutoff data 898 (right). The N-terminal region of TAF11 is lacking amino acids R and K, and consequently, 899 no cross-links were observed with BS3. Observed cross-links are listed in Table 4. 900 which was then exposed to UV light. A cartoon representation of TAF/TAF13/TBP is shown 906 (left) with a zoom-in on the site of ncAA introduction (right). K34 of TAF13 was chosen as it 907 appears to be located at/near the interfaces between TAF13, TAF11 and TBP in our model of 908 the ternary complex. Cross-linked sample then was separated by SDS-PAGE, excised from 909 gel, reduced, alkylated, trypsin digested and desalted followed by mass spectrometry 910 analysis. Specific cross-linking patterns were obtained upon TAF11/TAF13 complex 911 formation with TBP. Example of mass spectrum for a cross-linked peptide is shown below. 912 ('TATA-box mimicry'), but is less pronounced in TAF1-TAND from yeast. Yeast TAF1-939 TAND binding is more extensive involving also the convex surface of TBP (Table 5).