TAF8 regions important for TFIID lobe B assembly, or for TAF2 interactions, are required for embryonic stem cell survival

The human general transcription factor TFIID is composed of the TATA-binding protein (TBP) and 13 TBP-associated factors (TAFs). In eukaryotic cells, TFIID is thought to nucleate RNA polymerase II (Pol II) preinitiation complex formation on all protein coding gene promoters and thus, be crucial for Pol II transcription. TFIID is composed of three lobes, named A, B and C. Structural studies showed that TAF8 forms a histone fold pair with TAF10 in lobe B and participates in connecting lobe B to lobe C. In the present study, we have investigated the requirement of the different regions of TAF8 for in vitro TFIID assembly, and the importance of certain TAF8 regions for mouse embryonic stem cell (ESC) viability. We have identified a TAF8 region, different from the histone fold domain of TAF8, important for assembling with the 5TAF core complex in lobe B, and four regions of TAF8 each individually required for interacting with TAF2 in lobe C. Moreover, we show that the 5TAF coreinteracting TAF8 domain, and the proline rich domain of TAF8 that interacts with TAF2, are both required for mouse embryonic stem cell survival. Thus, our study demonstrates that distinct TAF8 regions involved in connecting lobe B to lobe C are crucial for TFIID function and consequent ESC survival.


Introduction
Transcription regulators in eukaryotes can be divided into three functional classes: genespecific transcription regulators, cofactor complexes and the general RNA polymerase transcription machinery. Their collaborative action is necessary to access specific loci in chromatin and allow precise transcription initiation (1). Regulated RNA polymerase II (Pol II) transcription requires a highly concerted, stepwise assembly of transcription factor complexes that form the preinitiation complex (PIC). A functional PIC consists of Pol II and six general transcription factors (GTFs): TFIIA, TFIIB, TFIID, TFIIE, TFIIF and TFIIH (2,3). The evolutionary conserved TFIID complex is the first GTF that binds gene promoters, and along with other GTFs, acts as a platform for PIC formation and consequent transcription initiation (3,4). TFIID is a multisubunit complex of about 1 MDa, composed of the TATA box-binding protein (TBP) and 13 (14 in yeast) TBP-associated factors (TAFs) (5). TFIID not only is essential for the recognition of core promoter sequences and the recruitment of the PIC, but also involved in gene expression via its interactions with cofactors, gene-specific activators and repressors, and chromatin modifications associated with active regions of the genome (4,6,7).
Human TAF8 is a 310 amino acid protein harboring a histone fold domain (HFD) at its N-terminal end, that interacts with the HFD of TAF10, to form a non-canonical histone fold pair arrangement in TFIID (8)(9)(10). TAF8 also interacts with TAF2, and TAF2-TAF8-TAF10 subcomplex assembles in the cytoplasm of human cells (10). Biochemical studies revealed that TFIID is assembled in a step-wise manner, first forming a stable 5TAF core complex, consisting of two copies each of TAF5-TAF6-TAF9-TAF4-TAF12. On one hand, this core is bound by the TAF8-TAF10 heterodimer, forming the 7TAF complex, similar to lobe B (11,12), or by TAF8-TAF10-TAF2 complex, forming the 8TAF complex (10,(12)(13)(14). On the other hand, the TAF5-TAF6-TAF9-TAF4-TAF12 core is bound by TAF11-TAF13 and TAF3-TAF10 HF pairs and TBP to form lobe A (12, 14). Importantly, in vitro the TAF8-TAF10 HF pair does not interact individually with any other HF TAF pair, but it interacts with the 5TAFcore complex, only if all five TAFs of the core complex are simultaneously present and the entire 5TAF core complex is formed (11). In addition, we demonstrated that the building blocks of mammalian TFIID, such as TAF8-TAF10, TAF6-TAF9 and TBP-TAF1, assemble co-translationally in the cytoplasm, in agreement with the stepwise assembly model of TFIID (15). Early electron microscopy (EM) studies established that endogenous human TFIID resembles a horseshoe composed of three main lobes (16,17). Recent human and yeast Komagataella phaffii TFIID cryoEM structures confirmed the three-lobe-structure of TFIID (called lobes A, B and C), and demonstrated evolutionary conservation and high flexibility within TFIID (12, 14,18). The high resolution structures of two TFIID domains indicated that i) lobe B contains the HFD domains of TAF8-TAF10 histone fold pair, together with one copy of the 5TAF core (TAF5-TAF6-TAF9-TAF4-TAF12) complex, ii) TAF8 participates in connecting lobe B and C by interacting with the two HEAT repeats of TAF6 and certain regions of TAF1, and iii) in lobe C the Cterminal half of TAF8 interacts with TAF2 ( Fig. 1A) (12,14,18).
In mice, germ line knock-out of genes encoding several TFIID subunits (Taf7, Taf8, Taf10, and Tbp) results in early embryonic lethality around E4.0 (19-22), suggesting that these mammalian TFIID subunits are absolutely essential for early mouse development. Importantly, a homozygous TAF8 TAF8c.781-1G>A splice site mutation in a human patient causes intellectual disability (23). Patients with this mutation express an unstable TAF8 . CC-BY-NC-ND 4.0 International license made available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprint this version posted July 6, 2021. ; https://doi.org/10.1101/2021.07.06.451281 doi: bioRxiv preprint protein in which the C-terminal 49 wild type amino acids are replaced by a 38 amino acid mutated sequence, caused by the frame shift, leading to partial TFIID dissociation (23).
In the present study, we have investigated the requirement of different TAF8 regions for TFIID assembly, and the importance of some of these TAF8 regions for embryonic stem cell (ESC) viability. We have identified TAF8 regions which are either important for interacting with the 5TAF core complex in lobe B, or for interacting with TAF2 in lobe C, and we show that the 5TAF core-interacting, and that one of the TAF2-interacting TAF8 regions are required for mouse embryonic stem cell survival.

Crosslinking mass spectrometry analysis of human TFIID reveals crosslinking hotspots in TAF8
To gain more insights into the structure/function relation of human TAF8, first we carried out a multiple sequence alignment of TAF8 proteins from several eukaryotic species to better understand the potential domain conservation of TAF8 in addition to its HFD (Fig. S1).
Following our alignment and based on previous publications (9-11), we have subdivided TAF8 in seven regions: N-ter, HFD, ID (intermediary domain), PRD (proline-rich domain), and T2R1, T2R2 and T2R3 (TAF2-interacting regions 1, 2 and 3) (Fig. 1A, 2A and Fig. S1). Prior to the publication of the cryo-EM structure of hTFIID (14), we performed cross-linking mass spectrometry (CXMS) analyses on TFIID to characterize the architecture of the complex. To this end we purified endogenous TBP-containing complexes (TFIID, SL1 and TFIIIB) by an anti-TBP immunoprecipitation from HeLa cell nuclear extracts, (Fig. S2), crosslinked them . CC-BY-NC-ND 4.0 International license made available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprint this version posted July 6, 2021. ; https://doi.org/10.1101/2021.07.06.451281 doi: bioRxiv preprint with the amine-reactive crosslinker BS3 and analyzed the sample by CXMS. The CXMS analyses of the endogenous TFIID complex identified 65 interlinks (crosslinks between different protein molecules) and 43 intralinks (crosslinks involving the same protein) (Table   S1 and Table S2, respectively). We then focused on TAF8 interlinks where TAF8 cross-linked to another TAF peptide and compared our dataset to CXMS data obtained from super core promoter (SCP) bound hTFIID, or SCP bound hTFIID containing PIC (12,14) (Fig. 1B). In all the analyzed endogenous hTFIID complexes TAF8 crosslinked extensively to TAF2, TAF6, TAF9/9b, and with a lesser frequency to TAF1, TAF3, TAF5 and TAF12 with crosslinking hotspots in TAF8 at K:20, K:112, K:132, K:178 and K:205 (Fig. 1B). The K:20 hotspot of TAF8, is absent from the published human TFIID structure (PDB: 6MZM, (14)), and crosslinks to the TAF9/TAF6 HFD, to the TAF6 middle (TAF6M) and TAF6C domains, containing five HEAT repeats (29), and to one position at the C-terminus of TAF5. These crosslinks agree with the potential position of the N-terminal end of TAF8 in the TFIID structure ( Fig. 1A and 1B). K:112 at the C-terminus of the HFD of TAF8 crosslinks to a region downstream of the HFD in TAF9/9b and to TAF3. K:132 in ID of TAF8 crosslinks to TAF3, TAF6 and TAF12b.
K:178 at the C-terminus of the PRD crosslinks to the non-conserved TAF6 linker between TAF6M and TAF6C, the third HEAT repeat of TAF6, to TAF2 and to the N-terminus region of TAF1. K:205 in T2R1 of TAF8 crosslinks in all the three CXMS analyses to K:875 in TAF2.
Surprisingly, between amino acids 205 and 310 of TAF8 (situated in T2R2 and T2R3 regions) no crosslinked peptides were detected in any of the three TFIID preparations (Fig. 1B), suggesting that the C-terminal end of TAF8 does not produce peptides that are amenable to MS analysis, eventually is buried in a non-crosslinkable structure, or is sticking out from the TFIID surface. Note that the visible path of hTAF8 in the cryo-EM structure of the human TFIID is from amino acids 28 to 229 (PDB: 6MZM, (14)) encompassing the HFD, ID, PRD, and T2R1 regions of TAF8 (Fig. 1A). Nevertheless, these crosslinking results in agreement . CC-BY-NC-ND 4.0 International license made available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprint this version posted July 6, 2021. ; https://doi.org/10.1101/2021.07.06.451281 doi: bioRxiv preprint with cryo-EM structures, further indicate that TAF8 plays a triple role in TFIID: i) participates in the assembly of the six HFD-containing 7TAF complex in lobe B via its N-terminus and HFD regions , ii) connects lobe B and C by interacting with the HEAT repeat regions of two TAF6, the TAF1 regions between aa 378-427 and TAF2 via its PRD, and iii) interacts with TAF2 in lobe C via its C-terminal part.
Identification of regions of TAF8 that are required either to stably interact with the 5TAF core complex or with TAF2 To further delineate the regions of TAF8 that are important for lobe B and/or lobe C assembly/interactions in TFIID, based on the above determined domain organization ( Deletions of N-terminal (aa 1-23 and ID (aa 117-140) regions of TAF8 did not influence the formation of the 3TAF (TAF10-TAF8-TAF2) complex ( Fig. 2C and 2D). In our negative control experiments, when TAF10 was absent from the coinfections, TAF8 or TAF2 did not . CC-BY-NC-ND 4.0 International license made available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprint this version posted July 6, 2021. ; https://doi.org/10.1101/2021.07.06.451281 doi: bioRxiv preprint co-purify by the anti-TAF10 IP, or when TAF8 was absent from the coinfections, TAF10 did not interact with TAF2 ( Fig. 2B-2D, lanes 2 and 3).
Next, subunits of the 7TAF complex were co-expressed with either WT TAF8, or with one of the six deletion mutants ( Fig. 2A) followed by anti-TAF10 IPs. When analyzing the 7TAF complex formation, we found that only the deletion the ID region of TAF8 (D117-140) severely reduced the interactions with subunits of the 5TAF core complex, while the TAF8-TAF10 HF pair could still form (Fig. S2A, Fig. 3A, 3B). All other tested TAF8 deletion mutants, including the delta N-terminus mutant deleting the K:20 crosslinking hotspot, which crosslinked extensively to subunits of the 5TAF complex ( Fig. 1B), formed the 7TAF complex ( Fig. 3A, 3B). As described previously (11), TAF10 alone could not interact with the 5TAF core complex when TAF8 was absent from the coinfections, further indicating that the formation of the TAF8-TAF10 HF pair is crucial for forming lobe B complex in TFIID (Fig.   3A, 3B).
To test whether the amino acid sequence of the ID is required for formation of 7TAF complex, we substituted the human TAF8 117-140 amino acid sequences with the S. cerevisiae Taf8 sequences from amino acids 156 to 179. TAF10 co-IP results showed that the amino acid substitution in this region could not functionally replace the human sequences, as the 7TAF complex did not form efficiently under these conditions (Fig. 3C, 3D). The results show that the amino acid sequence of the ID domain is required for efficient formation of the 7TAF complex, and that the ID is not functionally conserved between S. cerevisiae and humans.
Thus, together our recombinant in vitro TFIID sub-assembly results show that i) deletion of the non-conserved N-terminal region of TAF8, which crosslinks to many core TAF subunits in TFIID (Fig. 1B), is not required for formation of the 7TAF core complex, or to interact with TAF2, ii) the ID region of TAF8 (aa 117-140) is required to interact with the 5TAF core complex and it cannot be replaced by the ID region from S. cerevisiae Taf8 Importantly, none of the exogenously expressed TAF8 proteins (WT or mutants) impaired endogenous TFIID assembly when analyzed by an anti-TAF7 IP (Fig. 5E). Thus, these results show that the ID or the PRD regions of TAF8 are required for efficient assembly of 7TAF complexes (or TFIID subcomplexes) in mESCs.

The ID and PRD regions of TAF8 are required for mouse embryonic stem cell survival
To test whether the above identified ID or PRD regions of TAF8 are required for in vivo TFIID assembly and consequent mESC survival, we set out to inactivate the endogenous Taf8 gene using CRISPR/Cas9 genome editing in the ESC lines expressing Flag-TAF8 WT:R, Flag-TAF8D117-140:R (ID), or Flag-TAF8D141-184:R (PRD) ( Figure 6A). Using this strategy in . CC-BY-NC-ND 4.0 International license made available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprint this version posted July 6, 2021. ; https://doi.org/10.1101/2021.07.06.451281 doi: bioRxiv preprint the presence of Dox in the Flag-TAF8 WT:R mESC expressing line, we obtained several viable homozygous Taf8 deletion ESC lines ( Figure 6B), indicating that the CRISPR/Cas9 worked efficiently. By using the same gene editing strategy in the presence of Dox in Flag-TAF8D117-140:R or Flag-TAF8D141-184:R expressing cell lines, we obtained heterozygous clones in the same proportion as for the Flag-TAF8 WT expressing lines, indicating that the CRISPR/Cas9 worked with comparable efficiency at the Taf8 genomic locus in these heterozygous mESC lines, as in the Flag-TAF8 WT:R mESC expressing line (Fig. 6B). In contrast, we could not isolate any homozygous knockout clone when inactivating the Taf8 locus in the Flag-TAF8D117-140:R or in the Flag-TAF8D141-184:R mESC lines, suggesting that the deleted ID or PRD regions of TAF8 have essential roles in endogenous TFIID assembly and/or function, and consequently for mESC survival, and that these functions cannot be compensated by other TAFs or TAF assemblies.

Discussion
Characterization of the structure-function relationship of multisubunit transcription complexes is crucial to understand gene regulation. Whereas three-dimensional models of fully assembled multiprotein complexes derived mainly from single-particle cryo-electron microscopy provide a wealth of information on the architecture of such complexes, the requirement of the precise subunit-subunit and domain-domain interactions to maintain the architecture and function of these gene regulatory complexes, such as TFIID remain largely elusive.
Based on several functional and structural studies (9,10,14,19,31,32), as well as on two independent CXMS experiments carried out with endogenous TFIID (this study and (12,14)), we have subdivided TAF8 into seven regions ( Fig. 2A). Out of these seven regions the HFD of TAF8 has already been characterized (10), thus we have investigated how the deletion of . CC-BY-NC-ND 4.0 International license made available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprint this version posted July 6, 2021. ; https://doi.org/10.1101/2021.07.06.451281 doi: bioRxiv preprint the remaining six regions of TAF8 influence the in vitro assembly of TAF8-containing TFIID building blocks. We identified three types of regions: i) the N-terminus of TAF8 that did not disrupt the tested interactions with either the 5TAFcore complex or with TAF2, ii) the ID region that is required for interactions with the 5TAF core complex, but does not influence the TAF2 interactions, and iii) four successive TAF2 regions which are all required for normal TAF8-TAF2 interactions, but do not influence the interactions with the 5TAF core complex (Figs 2 and 3). Our TAF8-TAF2 interaction results agree with previous surface plasmon resonance (SPR) experiments using immobilized full-length TAF2 as ligand and maltose binding protein (MBP) fusions of TAF8 fragments (10). Thus, our deletion analysis confirms that TAF8 plays a triple role in TFIID: The N-terminus, the HFD and ID regions of TAF8 participate in the assembly of the six HFD-containing 7TAF complex in lobe B. Importantly, when the ID region of human TAF8 was replaced with the non-conserved amino-acids from yeast TAF8, this replacement did not restore the binding of the TAF8-TAF10 HF pair to the 5TAF core complex, further substantiating that this region and its amino sequence is important for forming the 7TAF complex in lobe B. It is possible that residues within the ID make specific contacts with TAF8 or TAF6 or other components of the 5TAF complex that are required for proper interaction of TAF8-TAF10 with the 5TAF complex. In spite of the fact that TAF8 PRD crosslinks to the TAF6 HEAT repeat regions, deletion of TAF8 PRD (aa 141-184) did not influence the formation of the 7TAF complex, suggesting that the TAF8 PRD is involved in the lobe B-C connecting function of TAF8, but not important for lobe B formation. In the cryo-EM structures of TFIID the lobe B-C connecting function of TAF8 was proposed to involve the unstructured PRD of TAF8. Interestingly, our deletion mutant analyses could not separate the lobe B-C connecting function of TAF8 from its TAF2 interacting function. Instead we found that the deletion of either the PRD, or each of the three TAF2 interaction regions (T2R1-T2R3 mutants) individually, all abrogated the interaction of TAF8 with TAF2. It is conceivable . CC-BY-NC-ND 4.0 International license made available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprint this version posted July 6, 2021. ; https://doi.org/10.1101/2021.07.06.451281 doi: bioRxiv preprint that the simple shortening of the 170 amino acid long C-terminal part of TAF8 causes rupture of the TAF8-TAF2 interactions. However, our present deletion experiments, together with the above mentioned SPR experiments using short TAF8 fragments (10), would rather suggest that the entire 170 amino acid TAF8 tail is needed to interact with TAF2. These interactions could happen through several TAF8-TAF2 contact surfaces, which would be all synergistically required for these interactions. In the cryo-EM structures of TFIID the structured TAF8 path stops at amino acid 229 (Fig. 1A) (14), and similarly the unresolved regions of TAF2 could also be involved in TAF8 interactions. Nevertheless, no TAF8-TAF2 crosslinks have been detected in endogenous TFIID complexes either in T2R2 or T2R3 regions. In this respect it is interesting to note that the deletion of the T2R2 or T2R3 regions of TAF8, both abrogate the TAF2 interactions, suggesting that these regions are still crucial for TAF2 interactions.
Intriguingly, the patient with intellectual disability expresses an unstable TAF8 protein in which the C-terminal last 49 wild type amino acids of TAF8 are replaced by a 38 amino acid mutated sequence, caused by the frame shift, leading to potential TFIID dissociation (23). In addition, the reported intellectual disability causing TAF2 mutations (T186R, P416H, or W649R) may also fall in these unmapped TAF2-TAF8 interaction regions (27,28). It is thus conceivable that an important regulatory surveillance mechanism exists in cells to control TAF8-TAF2 interactions and through these interactions stable holo-TFIID assembly.
Nevertheless, pluripotent ESCs with highly active Pol II transcription seem to require fully assembled and functional holo-TFIID as deletion of Tbp, Taf7, Taf8, and Taf10 cause mESC lethality (19-22). Our results show that when the ability of TAF8 to interact with the 5TAF core complex in lobe B (TAF8DID), or with TAF2 in lobe C (TAF8DPRD) are impaired, likely causing defective TFIID assembly, the ESCs cannot survive, further suggesting that ESCs require fully assembled holo-TFIID for function. Future experiments will decide whether such deletions at later stages of ESC differentiation (i.e. embryonic body or neuronal differentiation) . CC-BY-NC-ND 4.0 International license made available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprint this version posted July 6, 2021. ; https://doi.org/10.1101/2021.07.06.451281 doi: bioRxiv preprint would also cause cell death. Here we have identified regions of TAF8 that are required for 1) anchoring TAF8-TAF10 to the core 5TAF complex and 2) for connecting the 7TAF complex in Lobe B to Lobe C. TAF8 plays a critical role in the genesis of TFIID and mutations in TAF8 can lead to human diseases such as ID. As such, defining the role of TAF8 regions in TFIID assembly and function provides key information about how TFIID assembly occurs and how mutations in particular domain can lead to human diseases.
Altogether, we show the significance of precise interactions of TAF8 with other TAFs, establishing TAF8 as a functional bridge between lobes B and C in TFIID. Moreover, our experiments demonstrate a functional role of these interactions in TFIID, required for proper functioning of mESCs.

Plasmids
The baculovirus expression vector for FLAG-tagged human (h) TAF8 WT was PCR amplified from the pPBAC MultiBac vector, expressing TAF8 and TAF10 together, described in (10), and the cDNA was inserted in the pVL1393 baculovirus expression vector digested with Bam HI and Eco RI together with a primer coding for a Flag-tag on the 5' end of the hTAF8 cDNA. The TAF8 deletion series was generated by site directed mutagenesis using pVL1393-Flag-TAF8 WT vector as a template (see Fig. 2A). The deleted Flag-TAF8 cassettes were inserted into the pVL1393 baculovirus expression vector. The constructs were verified by sequencing. All the other baculovirus expression vectors have been described previously (11,33).
Mouse (m) WT Flag-TAF8 cDNA and its deletions, corresponding to the hTAF8 deletions (called Flag-mTAF8D117-140 or Flag-mTAF8D141-184), were cloned into the . CC-BY-NC-ND 4.0 International license made available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprint this version posted July 6, 2021. ; https://doi.org/10.1101/2021.07.06.451281 doi: bioRxiv preprint pUHD10-3 vector (34) in which the expression of WT Flag-mTAF8, Flag-mTAF8D117-140, or Flag-mTAF8D141-184 are under the control of tet-operator.
The pUES-3 plasmid expressing the two Taf8 guide (g) RNAs and co-expressing highfidelity Cas9 (35) fused to EGFP (Cas9-HF-EGFP) were generated by Golden Gate cloning (36). The sequences of the gRNAs are shown in Table S3. The pUES-3 plasmid was verified by sequencing.

Antibodies
Mouse monoclonal (mAb) and rabbit polyclonal (pAb) antibodies raised against the following proteins have been described previously, or were purchased commercially: anti-TBP Cat# 1408) were added freshly to the medium. Cells are grown at 37°C with 5% CO2. For cell collection and/amplification ESCs were trypsinized for 2 to 3 minutes with trypsin-EDTA (Invitrogen, Cat#25200-072) and the digestion was stopped by the addition of pre-warmed 15% . CC-BY-NC-ND 4.0 International license made available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprint this version posted July 6, 2021. ; https://doi.org/10.1101/2021.07.06.451281 doi: bioRxiv preprint FCS. Where indicated doxycycline was added to the medium at a final concentration of 1μg/ml (Sigma Cat#D9891).
Next the rtTA expressing ESC line was transfected using Lipofectamine2000  µ g o f a p U E S -3 plasmid containing the two Taf8 gRNAs ( Figure 6A and Table S3) Table S4.

RT-qPCR
For RT-qPCR, the isolated RNA samples were reverse transcribed to cDNA using superscript II (Transcriptase inverse SuperScript™ II, Invitrogen™ Cat#18064022) following manufacturer's instruction. Then the cDNA samples were amplified using LightCycler® 480 SYBR® Green 2x PCR Master Mix I (Roche, Cat# 04887352001) and 0.6 μM of forward and reverse primers with a LightCycler® 480 (Roche). The primer pairs used for different RT-qPCR reactions are listed in Table 2. For the assessment of mRNA levels, the obtained thresholdvalues were used to calculate the relative gene expression using the 2-ΔΔCT method and considering the individual primer pair efficiencies (46). . CC-BY-NC-ND 4.0 International license made available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprint this version posted July 6, 2021. ; https://doi.org/10.1101/2021.07.06.451281 doi: bioRxiv preprint

Recombinant protein production
Recombinant baculoviruses were generated as described (10) (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprint this version posted July 6, 2021. ; https://doi.org/10.1101/2021.07.06.451281 doi: bioRxiv preprint HCl pH 7.5, 5 mM MgCl2, 10% glycerol, 0.1% NP40, 1 mM DTT and 1x protease inhibitor cocktail) to reach 100 mM KCl in the extracts. Protein inputs were then pre-cleared with 1/10 volume of 100% protein A or G beads for 1 hours at 4°C with overhead agitation. Beads were coupled to the different antibodies (as indicated in the figure legends). Approximately, 1 mg of indicated antibody per ml of protein A or G bead was bound. Beads were incubated for 1h at room temperature with agitation, unbound antibody were removed by washing the beads twice with IP500 buffer (0 buffer containing 500 mM KCl) and twice with IP100 buffer (0 buffer containing 100 mM KCl) before addition of the pre-cleared protein extracts, and further incubated overnight at 4°C with overhead agitation. The following day the beads were collected, and subjected to two rounds of washing for 10 minutes each with 10 volumes of IP500 buffer, followed by 2 x IP100 buffer washes. Proteins IP-ed with an anti-TBP or anti-TAF10 mAb, were eluted by adding 1 bead volume of 2 mg/ml competing epitope peptide for 4 hours, repeated again for 2 hours (42). For anti-Flag and anti-TAF7 co-IPs, proteins were eluted with 0.1 M pH 2.8 glycine, then neutralized with 1.5 M Tris pH 8.8. Eluted proteins were either separated on an 4-12% SDS-PAGE gel, along with the input extract, and were probed with antibodies as indicated in the different figures, silver-stained, or analyzed by mass spectrometry.
. CC-BY-NC-ND 4.0 International license made available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprint this version posted July 6, 2021. ; https://doi.org/10.1101/2021.07.06.451281 doi: bioRxiv preprint The CXMS data have been deposited deposited to the ProteomeXchange Consortium via the PRIDE (48) partner repository with the dataset identifier PXD026575.

LC MS/MS Mass spectrometry analyses
Protein mixtures were precipitated with TCA (Sigma Aldrich, Cat# T0699) overnight at 4°C. Samples were then centrifuged at 14000 g for 30 minutes at 4°C. Pellets were washed twice with 1 ml cold acetone and centrifuged at 14000 g for 10 minutes at 4°C.  (75 µm ID x 2 cm, 3 µm, 100Å, ThermoFisher Scientific) for 3.5 minutes at 5 µL/min with 2% ACN (Sigma Aldrich, Cat# 1207802), 0.1% formic acid (Sigma Aldrich, Cat# 94318) in water and then separated on a C18 Accucore nano-column (75 µm ID x 50 cm, 2.6 µm, 150Å, ThermoFisher Scientific) with a 90 minutes linear gradient from 5% to 35% buffer B (A: 0.1% FA in water/ B: 99% ACN, 0.1% FA in water), then a 20 minutes linear gradient from 35% to 80% buffer B, followed with 5 min at 99% B and 5 minutes of regeneration at 5% B. The total duration was set to 120 minutes at a flow rate of 200 nL/min. The oven temperature was kept constant at 38°C.
The mass spectrometer was operated in positive ionization mode, in data-dependent mode . CC-BY-NC-ND 4.0 International license made available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprint this version posted July 6, 2021. ; https://doi.org/10.1101/2021.07.06.451281 doi: bioRxiv preprint with survey scans from m/z 350-1500 acquired in the Orbitrap at a resolution of 120,000 at m/z 400. The 20 most intense peaks (TOP20) from survey scans were selected for further fragmentation in the Linear Ion Trap with an isolation window of 2.0 Da and were fragmented by CID with normalized collision energy of 35%. Unassigned and single charged states were rejected. The Ion Target Value for the survey scans (in the Orbitrap) and the MS2 mode (in the Linear Ion Trap) were set to 1E6 and 5E3 respectively and the maximum injection time was set to 100 ms for both scan modes. Dynamic exclusion was used. Exclusion duration was set to 20 s, repeat count was set to 1 and exclusion mass width was ± 10 ppm.
Data analysis Proteins were identified by database searching using SequestHT . CC-BY-NC-ND 4.0 International license made available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprint this version posted July 6, 2021.  . CC-BY-NC-ND 4.0 International license made available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprint this version posted July 6, 2021.  . CC-BY-NC-ND 4.0 International license made available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprint this version posted July 6, 2021. ; https://doi.org/10.1101/2021.07.06.451281 doi: bioRxiv preprint  . CC-BY-NC-ND 4.0 International license made available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprint this version posted July 6, 2021. ; https://doi.org/10.1101/2021.07.06.451281 doi: bioRxiv preprint Normalized spectral abundance factor (NSAF) values were calculated and normalized to the bait of the IPs (to Flag-tagged TAF8 or its deletions, in D; and to TAF7 in E). The normalized NSAF results are represented as heat maps with the indicated scales.
. CC-BY-NC-ND 4.0 International license made available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprint this version posted July 6, 2021. ; https://doi.org/10.1101/2021.07.06.451281 doi: bioRxiv preprint . CC-BY-NC-ND 4.0 International license made available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprint this version posted July 6, 2021. ; https://doi.org/10.1101/2021.07.06.451281 doi: bioRxiv preprint