Proteomics Reveals a Physical and Functional Link between Hepatocyte Nuclear Factor 4α and Transcription Factor IID*

Proteomic analyses have contributed substantially to our understanding of diverse cellular processes. Improvements in the sensitivity of mass spectrometry approaches are enabling more in-depth analyses of protein-protein networks and, in some cases, are providing surprising new insights into well established, longstanding problems. Here, we describe such a proteomic analysis that exploits MudPIT mass spectrometry and has led to the discovery of a physical and functional link between the orphan nuclear receptor hepatocyte nuclear factor 4α (HNF4α) and transcription factor IID (TFIID). A systematic characterization of the HNF4α-TFIID link revealed that the HNF4α DNA-binding domain binds directly to the TATA box-binding protein (TBP) and, through this interaction, can target TBP or TFIID to promoters containing HNF4α-binding sites in vitro. Supporting the functional significance of this interaction, an HNF4α mutation that blocks binding of TBP to HNF4α interferes with HNF4α transactivation activity in cells. These findings identify an unexpected role for the HNF4α DNA-binding domain in mediating key regulatory interactions and provide new insights into the roles of HNF4α and TFIID in RNA polymerase II transcription.

merase II and the remaining general transcription factors. The TATA box-binding protein (TBP) subunit of TFIID recognizes the TATA box element and can substitute for TFIID in transcription in vitro. Additional TBP-associated factor (TAF) subunits of TFIID recognize the initiator (Inr) and other core promoter elements and contribute to functional interactions between the general transcription machinery and DNA-binding transcription factors and coregulators (1,8,9). TBP has also been shown to be an essential component of SL1 and TFIIIB, which are multisubunit general transcription factors for RNA polymerases I and III, respectively (10 -16).
Because of its central role in transcriptional regulation, there has been considerable interest in defining the repertoire of TBP-interacting proteins. Biochemical studies that led to definition of TFIID, SL1, TFIIIB, and other TBP-associated proteins have typically used conventional chromatography and/or immunoaffinity purification methods to purify TBP-containing complexes (17), followed by protein isolation by reverse-phase chromatography or SDS-PAGE and identification by Edman sequencing or mass spectrometry (e.g. . Although these methods were very successful in identifying stoichiometric components of TBP-containing complexes, they are likely to have missed proteins that interact only weakly or transiently with TBP. Here, we report a comprehensive analysis of TBP-interacting proteins using MudPIT (multidimensional protein identification technology) mass spectrometry. MudPIT is a multidimensional chromatography-based proteomics method in which a mixture of proteins is digested into peptides and analyzed by tandem mass spectrometry without prior isolation of individual proteins. MudPIT has proven to be an unbiased and exquisitely sensitive method for identifying proteins in complex mixtures because it does not suffer from the inevitable losses associated with the identification and elution of proteins from polyacrylamide gels or reverse-phase resins.
The results of these experiments identified as TBP-interacting proteins most subunits of the previously known TBP-containing complexes. In addition, we identified a number of proteins not previously identified as TAFs, including several DNA-binding transcription factors. Among these was the DNA-binding transcription factor and orphan nuclear receptor HNF4␣ (27), which plays an important role in early development and in hepatocyte and intestinal differentiation (28,29). In the adult mammal, HNF4␣ is expressed in liver, intestine, and pancreas, where it is responsible for the expression of genes that control glucose and lipid metabolism (30,31). Mutations in the human HNF4␣ gene cause maturity onset diabetes in the young (type 1), a rare form of non-insulin-dependent diabetes mellitus (32). Like other members of the nuclear receptor superfamily, HNF4␣ possesses a DNA-binding domain with a conserved double zinc finger motif, an N-terminal transactivation domain called AF-1 (activation function 1), and a C-terminal domain called AF-2, which in other nuclear receptors functions as a ligand-dependent activation domain. Although HNF4␣ AF-2 is structurally homologous to other receptors and several reports have indicated that fatty acids or fatty acyl-CoA associates with HNF4␣, no definitive ligand has been identified (33)(34)(35).
In this study, we focused on the newly identified interaction between TBP and HNF4␣, which we demonstrate is mediated via the HNF4␣ DNA-binding domain. Through this interaction, HNF4␣ can target the TFIID complex to promoters containing HNF4␣-binding sites. In vitro, HNF4␣ recruits TFIID or TBP to an immobilized template through direct physical interactions between its DNA-binding domain and TBP. An HNF4␣ mutant that interferes with TBP binding also interferes with recruitment of TBP or TFIID to promoters in vitro and with HNF4␣-dependent gene activation in cells. Taken together, these observations define a novel role for the HNF4␣ DNA-binding domain in mediating key regulatory interactions and are consistent with the model that the interaction of HNF4␣ with TBP contributes to HNF4␣-induced transcription.

EXPERIMENTAL PROCEDURES
Cell Culture-Parental HeLa S3 cells and their derivatives were maintained in Dulbecco's modified Eagle's medium with 5% glucose, 10% fetal bovine serum, 2 mM GlutaMAX, 100 units/ml penicillin, and 100 g/ml streptomycin (Invitrogen). For large scale cultures, HeLa cells were grown in spinner culture in Joklik medium with 5% calf serum. Full-length cDNAs encoding human TAF7 (GenBank TM accession number BC032737) and mouse HNF4␣ (GenBank TM accession number BC039220) were subcloned with FLAG tags into pQCXIH (Clontech). Recombinant retroviruses were generated by transfection into Plat E packaging cells (36) and used to infect a HeLa S3 cell line stably expressing the mouse ecotropic retrovirus receptor (mCAT-1) (37).
Antibodies, Affinity Purification, and Immunoprecipitation-Anti-FLAG (M2) and anti-HA (HA-7) antibodies and anti-FLAG (M2)-agarose were from Sigma; anti-TBP, anti-TAF6, and anti-TAF7 monoclonal antibodies were from Abcam; anti-TAF1 and anti-TAF4 monoclonal antibodies and goat anti-HNF4␣ polyclonal antibody C-19 were obtained from Santa Cruz Biotechnology. Anti-GST monoclonal antibody was obtained from Bethyl Laboratories. Mouse anti-HA monoclonal antibody 12CA5 was a gift from Michael Carey (Department of Biological Chemistry, UCLA School of Medicine). Anti-HA antibody 12CA5 was bound to protein A-Sepharose (Repligen Corp.). Anti-FLAG and anti-HA affinity purifications were performed essentially as described except that proteins were eluted in buffer containing 0.2 mg/ml FLAG or 3ϫHA peptide in Buffer A (20 mM HEPES-KOH (pH 7.9), 0.1 M KCl, 0.1 mM EDTA, and 20% glycerol) plus 0.05% Triton X-100 (38). Mediator complex was purified as described (39).
Mass Spectrometry-Identification of proteins was accomplished using a modification of the MudPIT procedure (40,41). Trichloroacetic acid-precipitated proteins were urea-denatured, reduced, alkylated, and digested with endoproteinase Lys-C (Roche Applied Science) followed by modified trypsin (Roche Applied Science) as described (40). Peptide mixtures were loaded onto 100-m fused silica microcapillary columns packed with 5-m C 18 reverse-phase (Aqua, Phenomenex), strong cation-exchange (PartiSphere SCX, Whatman), and reverse-phase particles (42). Loaded microcapillary columns were placed in-line with an Agilent 1100 series quaternary high pressure liquid chromatography pump and an LTQ ion trap mass spectrometer equipped with a nano-liquid chromatography-electrospray ionization source (Thermo Finnigan). Fully automated 10-step MudPIT runs were carried out on the electrosprayed peptides as described (41). Tandem mass spectra were interpreted using SEQUEST (43) against a data base of 61,427 sequences consisting of 37,742 human proteins (downloaded from NCBI on 03/04/2008), 177 usual contaminants (such as human keratins, IgGs, and proteolytic enzymes), and, to estimate false discovery rates, 30,713 randomized amino acid sequences derived from each non-redundant protein entry. Peptide/spectrum matches were sorted and selected using DTASelect (44) with the following criteria set: peptide/spectrum matches were retained only if they had a DeltaCn of at least 0.08 and minimum XCorr of 1.8 for singly, 2.0 for doubly, and 3.0 for triply charged spectra. In addition, peptides had to be fully tryptic and at least 7 amino acids long. Combining all runs, proteins had to be detected by at least two such peptides or one peptide with two independent spectra. With these criteria, the final false discovery rates at the protein and peptide levels were 0.47 and 0.063 Ϯ0.026%, respectively. Peptide hits from multiple runs were compared using CONTRAST (44). To estimate relative protein levels, normalized spectral abundance factor values were calculated for each detected protein essentially as described (45)(46)(47), with the following exception. To deal with peptides shared between multiple isoforms, distributed normalized spectral abundance factor values were calculated based on distributed spectral counts, in which shared spectral counts are distributed according to the proportion of spectral counts unique to each isoform (48).
Production of Recombinant Proteins in Bacteria-HA-tagged human TBP and FLAG-tagged mouse HNF4␣ (GenBank TM accession number AAH39220) and derivatives were expressed from pET30 (Novagen, Madison, WI) with an N-terminal His 6 tag in Escherichia coli BL21(DE3)-CodonPlus cells (Stratagene). N-terminally GST-tagged human HNF4␣ and derivatives were expressed in E. coli and purified essentially as described (38).
Solution Binding Assays-Anti-FLAG (M2) beads were preequilibrated in binding buffer (150 mM NaCl, 20 mM HEPES-KOH (pH 7.9), 1.5 mM MgCl 2 , and 0.2% Triton X-100). 2 g of purified, bacterially expressed His-FLAG-HNF4␣ or HNF4␣ derivatives were incubated with 20 l of anti-FLAG beads for 30 min at 4°C in 400 l of binding buffer, washed three times, and resuspended in 400 l of binding buffer. Purified recombi-nant HA-TBP was added to the beads and incubated for an additional 30 min at 4°C. After the beads were washed four times with binding buffer, bound proteins were eluted by incubation with 0.2 mg/ml FLAG peptide in binding buffer and subjected to SDS-PAGE and immunoblot analysis with anti-HA and anti-FLAG antibodies. Alternatively, 2 g of GSTtagged HNF4␣ or HNF4␣ derivatives were bound to 20 l of glutathione-Sepharose 4B beads in 400 l of binding buffer plus 1 mM dithiothreitol. The beads were washed three times with binding buffer, resuspended in 400 l of binding buffer, and then incubated with HA-tagged TBP for an additional 30 min at 4°C. The beads were washed four times, and proteins were eluted with 20 mM glutathione in 100 mM Tris-HCl (pH 7.9), 150 mM NaCl, and 0.1% Triton-X-100 and subjected to SDS-PAGE and immunoblot analysis with anti-HA and anti-GST antibodies.
Promoter Binding Assays-Immobilized templates were generated essentially as described (49). To construct pREϫ4-E4T and pREϫ4-MLT, a double-stranded oligonucleotide containing four tandemly repeated copies of the HNF4␣-responsive element from the apolipoprotein A-I enhancer element (TGAACCCTTGACCCCTGC) (50,51) was introduced between the PstI and BamHI sites of pG5-E4T and pG5-MLT (52). The BamHI site is located 13 or 27 bp upstream of the TATA boxes of the adenovirus major late (AdML) or E4 core promoter, respectively. Biotinylated fragments of pREϫ4-E4T or pREϫ4-MLT were amplified using a 27-nucleotide biotinylated primer positioned 204 bp upstream of the HNF4␣-binding sites and a primer positioned 260 bp downstream of the transcription initiation site. Biotinylated TATA mutant DNA fragments were generated from pREϫ4-E4T by two-step PCR mutagenesis. Conversion of TATATATAC to the NotI recognition site GCGGCCGC was confirmed by digestion with NotI. Biotinylated PCR products were fractionated on agarose gels, purified using a QIAquick gel extraction kit (Qiagen), and bound to streptavidin-coupled Dynabeads TM M-280 (Dynal). The beads were resuspended in Buffer A to contain ϳ90 fmol/l DNA fragment.
Promoter binding assays were performed as follows. Various combinations of proteins were incubated with 2.5 l of Dynal bead suspension in 62 mM KCl, 12.5% glycerol, 12.5 mM HEPES (pH 7.9), 20 mM Tris-HCl (pH 7.9), 0.06 mM EDTA (pH 8.0), 7.5 mM MgCl 2 , 0.5 mg/ml bovine serum albumin, 0.3 mM dithiothreitol, and 0.025% Nonidet P-40 in a total volume of 60 l. After 30 min at 30°C, the beads were collected using a magnetic particle concentrator (Dynal), washed three times with the same buffer without bovine serum albumin, and resuspended in SDS-PAGE loading buffer. Bound proteins were identified by Western blotting.
Luciferase Reporter Assays-293T cells were cotransfected with 1 g of pG5-Luc (Promega), which encodes firefly luciferase driven by five Gal4-responsive elements upstream of the AdML core promoter; 100 ng of the control plasmid pRL-tk (Promega), which encodes Renilla luciferase under the control of the thymidine kinase promoter; and varying amounts of effector plasmid expressing wild-type or mutant Gal4-HNF4␣ using FuGENE 6 reagent (Roche Applied Science). Total effector plasmid in each transfection was adjusted to 1 g with empty vector. After 48 h, Gal4-HNF4␣ transactivation activity was determined by measuring firefly and Renilla luciferase activities using the Dual-Luciferase reporter assay kit (Promega) and normalizing firefly to Renilla luciferase.

RESULTS
MudPIT Analysis of TBP-associated Proteins-To define TBP-interacting proteins, HA epitope-tagged TBP and associated proteins were subjected to anti-HA immunopurification from a cell line stably expressing HA-TBP and analyzed by MudPIT mass spectrometry. This analysis identified many of the known TBP-associated proteins (supplemental Table 1). Among these were subunits of TBP-containing complexes needed for transcription by polymerases I, II, and III, respectively (8 -10, 12-16), including all of the known TFIID subunits (TAF II s); the SL1 subunits TAF1A, TAF1B, and TAF1C; and the TFIIIB subunit Brf1. We also detected large amounts of B-TAF1, the mammalian ortholog of Saccharomyces cerevisiae Mot1, an Snf2 family ATPase that uses the energy of ATP to displace TBP from DNA and has been proposed to play a key role in the dynamics of TBP in cells (53)(54)(55)(56)(57).
The proteasomal ATPases, but not other subunits of the 19 S regulatory particle of the proteasome, have been identified in previous proteomic analyses as TBP-interacting proteins (25). Consistent with these observations, all of the proteasomal ATPases copurified with our preparations of HA-TBP. In addition, however, we also identified all other subunits of the 19 S regulatory particle except for the human deubiquitinating enzyme UCH37 and human Rpn13 (also known as ADRM1), which links UCH37 to the 19 S regulatory particle (38,58,59).
Prior proteomic analyses using TBP as a bait have not led to the identification of sequence-specific DNA-binding transcription factors, most likely because only a small fraction of TFIID or other TBP-containing complexes are associated with any one of the thousands of these transcription factors in the cell. Consistent with these prior results, we found very few DNAbinding transcription factors in our MudPIT data sets. Surprisingly, however, we did identify four DNA-binding transcription factors as abundant components of our preparations of TBPassociated proteins. These included a homeobox-containing protein (HMBOX1), the glucocorticoid receptor, a steroid-responsive nuclear receptor, and two closely related orphan nuclear receptors (HNF4␣ and HNF4␥). In the remainder of this work, we explored the functional consequences of the interaction of TBP with HNF4␣.
HNF4␣ Binds TFIID and TBP-To confirm the interaction of HNF4␣ with TBP, nuclear extracts from HeLa cells stably expressing HA-tagged TBP were subjected to immunoprecipitation with anti-HA or control antibodies, and TBP and HNF4␣ were detected by Western blotting. As shown in Fig. 1A, endogenous HNF4␣ specifically coprecipitated with HA-TBP.
In complementary experiments, we used anti-FLAG antibodies to immunoprecipitate proteins from nuclear extracts from parental HeLa cells or a HeLa cell line stably expressing FLAG-tagged HNF4␣. As shown in the Western blots of Fig.  1B, TBP and the TFIID subunits TAF1, TAF4, TAF6, and TAF7 all specifically coprecipitated with FLAG-HNF4␣, indicating that HNF4␣ binds endogenous TBP and the TFIID complex.
Finally, to determine whether HNF4␣ can directly bind TBP or whether the interaction requires one or more of the TAF subunits, we performed GST pulldown assays using bacterially expressed, recombinant GST-tagged HNF4␣ and HA-tagged TBP. HA-tagged TBP interacted with GST-tagged HNF4␣ but not with GST (Fig. 1D). Taken together, these data indicate that TBP can bind directly to HNF4␣ without the assistance of any accessory proteins.
Mutation of the HNF4␣ DNA-binding Domain Interferes with TBP Binding-Nuclear receptors, including HNF4␣, are composed of multiple subregions designated A/B, C, D, E, and F. Regions A/B corresponds to the ligand-independent AF-1 transactivation domain. Regions C and D include the zinc finger and hinge regions (60), which together make up the HNF4␣ DNA-binding domain (61). Region E contains the ligand-binding, dimerization, and ligand-dependent AF-2 transactivation domains, whereas region F is a C-terminal negative regulatory domain that appears to modulate HNF4␣ interaction with coregulators and/or fatty acid ligands ( Fig. 2A) (62).
To identify the region of HNF4␣ responsible for TBP binding, we performed in vitro binding assays with recombinant wild-type FLAG-HNF4␣ or a series of FLAG-tagged HNF4␣ deletion mutants (Fig. 2B). In these assays, the binding of HNF4␣ deletion mutants either lacking the AF-1 transactivation domain (HNF4␣ CDEF) or lacking both AF-1 and C-terminal regulatory domains (HNF4␣ CDE) to TBP was indistinguishable from that of fulllength HNF4␣, whereas a mutant also lacking the zinc fingers (HNF4␣ DE) did not bind well (Fig. 2B). To define further the region of HNF4␣ needed for TBP binding, we produced a series of N-or C-terminal deletion mutants of recombinant GST-tagged HNF4␣ and assayed them for their abilities to bind TBP. GST-HNF4␣ ABCD, which lacks region F and the ligand-binding, AF-2, and dimerization domains, bound TBP as well as did full-length HNF4␣; however, a mutant lacking most of region D (HNF4␣ ABC-(1-115)) exhibited substantially reduced TBP binding, and the TBP-binding activity of a mutant including only regions A-C (residues 1-108) was barely detectable. Finally, HNF4␣ CD-(42-165) was sufficient for binding to TBP (Fig. 2B). Thus, residues 42-165, which fall within the DNA-binding domain, contribute to TBP interaction, whereas residues 108 -115 are particularly important. Residues 108 -115 correspond to conserved ␣-helix III, which is immediately C-terminal to the second zinc finger of the DNA-binding domain (61). Notably, the DNA-binding domains are almost identical in HNF4␣ and HNF4␥, which we also identified as a TBP-interacting protein in our MudPIT analyses, whereas the AF-1 regions of HNF4␣ and HNF4␥ are highly divergent.
We next tested the TBP-binding activity of a series of HNF4␣ mutants in which alanine was substituted for charged residues or serines falling within the region important for TBP interaction. Consistent with our observation that sequences within helix III are particularly important for TBP binding, we observed that mutation of Lys 109 and Lys 110 to Ala or of Arg 104 , Arg 107 , Lys 109 , and Lys 110 to Ala interfered with TBP binding, whereas mutation of residues outside this helix had little or no effect (Fig. 2C).
HNF4␣ Recruits TBP and TFIID to Promoters in Vitro-Our observation that HNF4␣ binds directly to TBP and TFIID raised the possibility that HNF4␣ might activate transcription in part by stabilizing the binding of TBP or TFIID to promoters. We therefore asked whether HNF4␣ could stimulate binding of TBP or TFIID to bead-bound DNA fragments containing four HNF4␣-responsive elements upstream of the core promoters of the AdML (pREϫ4-MLT) or adenovirus E4 (pREϫ4-E4T) gene (Fig. 3A). Although TBP alone can bind both core promoters (Ref. 9 and data not shown), we observed that binding of recombinant TBP to both templates was strongly stimulated by recombinant HNF4␣ when binding assays were performed with limiting concentrations of TBP (Fig. 3B). This stimulation was dependent on sequence-specific interaction of both HNF4␣ and TBP with their cognate binding sites because mutation of the TATA box sequence or deletion of the HNF4␣-  1 and 2) and 1% of input extract (lane 3) were immunoblotted with anti-HNF4␣ or anti-HA antibodies. B, nuclear extracts from parental or FLAG-tagged HNF4␣ (F-HNF4␣)expressing HeLa cells were precipitated with anti-FLAG antibodies. Immunoprecipitates (IP) or 1% of input protein were analyzed by Western blotting (WB). C, HepG2 nuclear extracts were immunoprecipitated with anti-HNF4␣ antibody C-19 or normal goat antibodies conjugated to protein G-Sepharose. Proteins were eluted with 0.25 mg/ml HNF4␣ C-19 antigenic peptide (Santa Cruz Biotechnology) and immunoblotted. Input, 1% of input protein; UB, unbound fraction from immunoprecipitation. D, GST or GST-HNF4␣ immobilized on glutathione-Sepharose 4B was incubated with HA-tagged TBP. After washing, bound proteins were eluted and analyzed by Western blotting. HA-tagged TBP interacted with GST-tagged HNF4␣ but not with GST.
responsive elements resulted in a dramatic decrease in HNF4␣dependent TBP recruitment to immobilized DNA (Fig. 3C).
To determine whether HNF4␣ can also recruit TFIID to a promoter, we performed binding assays with HNF4␣ and TFIID that had been purified by FLAG immunoprecipitation from a HeLa cell line stably expressing FLAG-tagged TAF7. The binding of TFIID subunits TAF1 and TAF6 and TBP to both the adenovirus E4 and AdML core promoters was increased in the presence of HNF4␣ (Fig. 3D).
Mutation of the HNF4␣ DNA-binding Domain Interferes with TBP or TFIID Recruitment by a Gal4-HNF4␣ Fusion Protein-We next wished to determine whether mutations that interfere with the ability of HNF4␣ to bind directly to TBP interfere with its ability to recruit TBP and TFIID to a promoter. Because these mutations fall in a region immediately adjacent to the zinc finger portion of the DNA-binding domain, we expected that they might also interfere with HNF4␣ DNA-binding activity. In preliminary experiments, we found that this was indeed the case (data not shown). For this reason, we generated Gal4-HNF4␣ fusion proteins in which the Gal4 DNA-binding domain was fused either to full-length wild-type HNF4␣ or to the TBP binding-defective HNF4␣ mutant in which Arg 104 , Arg 107 , Lys 109 , and Lys 110 were changed to Ala. We then tested their abilities to recruit TBP or TFIID to an immobilized DNA fragment bearing Gal4 DNA-binding sites upstream of the AdML core promoter (Gal4ϫ5-MLT) (Fig. 4A). As shown in Fig. 4 (B and C), we observed that wild-type Gal4-HNF4␣ increased TBP or TFIID recruitment to the promoter, whereas a Gal4 fusion protein containing an HNF4␣ mutant defective in TBP binding did so less effectively, consistent with the model that direct binding of HNF4␣ to TBP contributes to recruitment of TBP or TFIID to the promoter. We noted that, in some experiments, the mutation appeared to have a smaller effect on recruitment of TAF6 than TBP (Fig. 4C). Whether this reflects additional interaction(s) between one or more of the TAF subunits with HNF4␣ remains to be determined. Residues shown in boldface were mutated to alanine. Residue numbering corresponds to that used previously (61). H, hinge region; LBD, ligand-binding domain; NRD, negative regulatory domain. B, the TBP-binding activities of FLAG-or GST-tagged wild-type (WT) HNF4␣ and derivatives were assayed as described under "Experimental Procedures." In the right panel, the same blot in the left panel was probed sequentially with anti-HA and anti-GST antibodies. The band indicated with the black arrowhead is residual HA-TBP signal; bands indicated with asterisks are proteolytic fragments recognized by anti-GST antibodies. C, the TBP-binding activities of FLAG-tagged wild-type and mutant HNF4␣ proteins were assayed as described under "Experimental Procedures." IP, immunoprecipitation; WB, Western blot; F-HNF4␣, FLAG-tagged HNF4␣. NOVEMBER 20, 2009 • VOLUME 284 • NUMBER 47

Physical and Functional Link between HNF4␣ and TFIID
Mutation of the HNF4␣ DNA-binding Domain Interferes with Gene Activation by a Gal4-HNF4␣ Fusion Protein-To test the possible contribution of the HNF4␣-TBP interaction to HNF4␣ transactivation activity in cells, we tested the ability of the Gal4-HNF4␣ fusion proteins to stimulate luciferase expression driven by five Gal4-responsive elements upstream of the AdML promoter. As expected because both include intact AF-1 and AF-2 transactivation domains, we observed that Gal4-HNF4␣ fusion proteins containing both wild-type and mutant versions of HNF4␣ could stimulate reporter expression. Importantly, however, the fusion protein containing wild-type HNF4␣ had about four times greater transactivation activity compared with the mutant (Fig. 4E).
Activation of transcription by HNF4␣ has been shown previously to depend strongly on an intermediary coregulator referred to as the Mediator complex. Although substantial evidence argues that Mediator functions via interactions with AF-2 (63, 64), we wished to rule out the possibility that DNAbinding region mutations that interfere with interactions of HNF4␣ with TBP also affect its interactions with the Mediator complex. Accordingly, we asked whether the Gal4-HNF4␣ mutant was defective in its ability to recruit Mediator to a promoter in vitro. As shown in Fig. 4D, Gal4-HNF4␣ fusion proteins containing wild-type and mutant HNF4␣ recruited the Mediator complex to the promoter equally well. Taken together, these results are consistent with the model that HNF4␣ activates transcription in part via interaction of its DNA-binding region with TBP.

DISCUSSION
Activation of transcription in vitro by HNF4␣ has been shown to depend strongly on intermediary coactivators  referred to as PC4 (positive cofactor 4) and the Mediator complex, which function via interactions with the ligand-dependent activation domain AF-2 (63,64). The results of a recent study indicate that HNF4␣ can also stimulate basal transcription via a Mediator-independent mechanism in a highly purified, reconstituted transcription system containing RNA polymerase II, general transcription factors, and PC4 (65), a finding we have confirmed in our laboratory (supplemental Fig. 1). However, the mechanism by which it does so has remained obscure. Here, we have presented evidence that supports a possible mechanism for Mediator-independent stimulation of basal transcription by HNF4␣. Our findings indicate that HNF4␣ binds directly to TBP via its DNA-binding domain and, through this interaction, can recruit TBP or the larger TFIID complex to HNF4␣-responsive promoters in vitro. We observed that a mutation in the HNF4␣ DNA-binding domain that blocks recruitment of TBP or TFIID but does not affect interaction with Mediator also blocks HNF4␣-dependent transactivation activity in cells. These observations are consistent with the model that HNF4␣-dependent recruitment of TBP or TFIID contributes to transcription activation.
Our observation that the HNF4␣ DNA-binding domain interacts with TBP was unexpected in light of previous evidence that TBP can interact with several nuclear receptors through the well characterized AF-1 or AF-2 activation domains (66). The AF-1 domains of both the glucocorticoid and estrogen receptors have been shown to bind TBP (67,68). In contrast, retinoid X receptor ␣ is capable of interacting directly with TBP through its ligand-dependent AF-2 transactivation domain. Mutation of a highly conserved lysine residue in retinoid X receptor ␣ AF-2 to glutamic acid reduces its ability to bind TBP and to activate a reporter gene (69). Although the AF-2 domain of HNF4␣ is similar in amino acid sequence to that of retinoid X receptor ␣ (33), we observed that deletion of AF-2 or mutation of the corresponding glutamic acid residue (data not shown) had no effect on the HNF4␣-TBP interaction. Thus, HNF4␣ interacts with TBP by a different mechanism compared with other nuclear receptors. While this work was in progress, Lu et al. (61) reported the crystal structure of the HNF4␣ DNAbinding domain bound to DNA. According to this structure, lysines 109 and 110 are solvent-exposed and therefore would be available for interaction with TBP. In addition, PRMT1 has been reported to interact with the DNA-binding domain of HNF4␣ and regulate the affinity of HNF4␣ for its binding site (70). These studies and our data strongly suggest that the HNF4␣ DNA-binding domain can provide surfaces for binding not only DNA but also other transcriptional regulatory proteins.
In the TFIID complex, TBP directly contacts other TFIID subunits, including TAF1, TAF2, TAF11, TAF12, and TAF13 (71)(72)(73), whereas the other TAFs are thought to associate with TBP indirectly. Furthermore, in the preinitiation complex, TBP also binds directly to DNA as well as the general transcription factors TFIIA and TFIIB (74 -77). In light of all these interactions, it is perhaps surprising that there is room on the TBP surface for functional interactions with HNF4␣. Nevertheless, structural studies indicate that TBP resides not in the middle but rather at the surface of both TFIID and larger transcription intermediates, including the TFIIA⅐TFIID⅐TFIIB complex (78 -80). Thus, in the preinitiation complex, surface(s) of TBP likely remain available for interaction with the HNF4␣ DNA-binding domain as well as other DNA-binding transcription factors and coregulators.
Our observation of a physical and functional link between HNF4␣ and TFIID is intriguing in light of a recent study indicating that TFIID is required for initial activation of HNF4␣dependent transcription during proliferation or differentiation of hepatoblasts but is dispensable for maintenance of transcription in adult hepatocytes (1,81). Thus, it is possible that the HNF4␣-TFIID interaction we have defined makes its greatest contribution to initial activation of transcription in proliferating cells, such as hepatoblasts or the proliferating tissue culture cells used in our studies, whereas mechanisms involving Mediator and other coregulators are sufficient to maintain transcription in non-proliferating adult hepatocytes.
In conclusion, HNF4␣ and other nuclear receptors have been shown previously to function through interactions between their AF-1 or AF-2 activation domains and various general transcription factors or coregulators. Our evidence that interactions of the HNF4␣ DNA-binding domain with TBP can direct recruitment of TFIID to promoters adds yet more complexity to the network of interactions by which this orphan nuclear receptor activates transcription.