Distinct amino acid residues may be involved in coactivator and ligand interactions in hepatocyte nuclear factor-4alpha.

Hepatocyte nuclear factor-4 (HNF-4) is a transcription factor of the nuclear hormone receptor superfamily that is constitutively active without the addition of exogenous ligand. Crystallographic analysis of the HNF-4alpha and HNF-4gamma ligand binding domains (LBDs) demonstrated the presence of endogenous ligands that may act as structural cofactors for HNF-4. It was also proposed by crystallographic studies that a combination of ligand and coactivator might be required to lock the receptor in its active state. We previously showed that mutations in amino acid residues Ser-181 and Met-182 in H3, Leu-219 and Leu-220 and Arg-226 in H5, Ileu-338 in H10, and Ileu-346 in H11, which line the LBD pocket in HNF-4alpha and come in contact with the ligand, impair its transactivation potential. In the present study, physical and functional interaction assays were utilized with two different coactivators, PGC-1 and SRC-3, to address the role of coactivators in HNF-4 function. We show that the integrity of the hinge (D) domain of HNF-4alpha and the activation function (AF)-2 activation domain region are critical for coactivation. Surprisingly, a different mode of coactivation is observed among the LBD point mutants that lack transcriptional activity. In particular, coactivation is maintained in mutants Ser-181, Arg-226, and Ile-346 but is abolished in mutants Met-182, Leu-219, and Ile-338. Physical interactions confirm this pattern of activation, implying that distinct amino acid residues may be involved in coactivator and ligand interactions, although some residues may be critical for both functions. Our results provide evidence and expand predictions based on the crystallographic data as to the role of coactivators in HNF-4alpha constitutive transcriptional activity.

Ileu-346 in H11, which line the LBD pocket in HNF-4␣ and come in contact with the ligand, impair its transactivation potential. In the present study, physical and functional interaction assays were utilized with two different coactivators, PGC-1 and SRC-3, to address the role of coactivators in HNF-4 function. We show that the integrity of the hinge (D) domain of HNF-4␣ and the activation function (AF)-2 activation domain region are critical for coactivation. Surprisingly, a different mode of coactivation is observed among the LBD point mutants that lack transcriptional activity. In particular, coactivation is maintained in mutants Ser-181, Arg-226, and Ile-346 but is abolished in mutants Met-182, Leu-219, and Ile-338. Physical interactions confirm this pattern of activation, implying that distinct amino acid residues may be involved in coactivator and ligand interactions, although some residues may be critical for both functions. Our results provide evidence and expand predictions based on the crystallographic data as to the role of coactivators in HNF-4␣ constitutive transcriptional activity.
Nuclear receptors are ligand-activated transcription factors, which regulate a large number of developmental and physiological processes in response to small lipophilic molecules (1). However, certain members of the superfamily are classified as orphan receptors since no exogenous ligand has been identified for them, among which, until recently, was HNF-4. 1 HNF-4 is a crucial regulator of several metabolic pathways that are important for lipid and glucose homeostasis and plays a master role in the differentiation of hepatocytes and in maintaining the adult liver phenotype (2)(3)(4). There have been conflicting reports in the literature concerning molecules that could act as potential ligands for this receptor (5,6). However, the recently solved crystal structures of two family members, HNF-4␣ and -␥, have revealed the existence of constitutively bound fatty acids in their ligand binding pockets (7,8), defining fatty acids as the endogenous ligands of HNF-4.
A combination of crystallographic and functional studies of nuclear receptors has shown that ligand binding triggers a conformational change in the LBD through a translocation of several of the helices forming the nuclear receptor ligand binding pocket, with most prominent being that of H12, which exposes the solvent residues of the AF-2 core motif (AF-2 AD). This motif was shown to be critical for coactivator interaction.
However, the recent crystallographic data on HNF-4 suggested a different model of transactivation. In the crystal of HNF-4␣, the receptor forms homodimers, and the two molecules in each homodimer adopt distinct conformations, the so-called "open" and "closed" forms, in which H12 is either fully extended into the solvent and collinear with H10 or packed against the body of the receptor, respectively (7). The ligand binding pocket forms a narrow channel that is lined almost exclusively with side chains of hydrophobic residues present in H3, H5, H7, and H10 (7). It was proposed that the inability to displace the fatty acids from the LBDs of both HNF-4␣ and HNF-4␥ might reflect the fact that these molecules are acquired as the nascent protein is folded. In this respect, they may act as structural cofactors, helping the protein to assume its correct conformation. These molecules do not appear to act at a subsequent level of activation, as traditional ligands do for other nuclear hormone receptors (8). This is a revolutionary view that could, however, explain the constitutive transcriptional activity of HNF-4, which is well established from functional studies (9). In fact, it was proposed that a combination of ligand and coactivator might be required to lock this receptor in its closed and active state (7). This hypothesis was recently confirmed by the crystallization of the LBD of HNF-4␣ bound to both fatty acid ligand and a coactivator sequence derived from SRC-1 (10).
Various studies have shown that HNF-4 interacts strongly with the p160 coactivators (11)(12)(13), and HNF-4 activity can also be enhanced by the action of CREB-binding protein (14), probably through acetylation, which is crucial for its proper nuclear retention (15). In addition, HNF-4 interacts with PGC-1, having an active role in controlling hepatic gluconeogenesis in response to starvation (16). Although PGC-1 contains an LXXLL motif, similar to that of other coactivators, which has been shown to mediate ligand-dependent interactions of the coactivator with several members of the nuclear hormone receptor superfamily (17), this motif exhibits distinct binding characteristics from those found in the p160 family members, as proposed from its surrounding amino acids and from transcriptional analysis studies in other receptors (18 -20).
Based on the fact that PGC-1 forms a complex with HNF-4␣ enhancing transcriptional activation (16) and that HNF-4␣ activity can be enhanced by the actions of CREB-binding protein, SRC-1, GRIP-1, and SRC-3 (11)(12)(13)(14)16), we sought to identify the key regions of HNF-4␣ required for enhancement by coactivators. We found that the integrity of AF-2 AD and hinge (D) regions of HNF-4␣ (particularly amino acids 160 -175) are critical for coactivation by PGC-1. Mutation in the critical residue E363K in the conserved AF-2 AD motif resulted in complete loss of transactivation, and this was maintained in the presence of coactivator. However, mutagenesis in the positively charged amino acid residues R168G/K170N/K171N that lie in the hinge (D) region had a small effect in HNF-4␣mediated transactivation, which was enhanced in the presence of coactivator. Furthermore, in a recent study, we have shown that point mutations in residues that are located in helices H3, H5, H10, and H11, which come in contact with the ligand, cause a dramatic decrease in HNF-4␣ activity (21). Structural modeling showed that these residues play a significant role in maintaining the structural integrity of the HNF-4␣ LBD. To investigate further the involvement of these amino acid residues in coactivator function, we used transient transfection analysis and interaction assays. We demonstrate that distinct residues in the LBD pocket, previously shown to control transcriptional activity (21), may be involved in coactivator and ligand interactions, although some residues may be critical for both functions.
Constructs pG 5 CAT, containing five GAL4 binding sites upstream of the ␤-globin promoter and the chloramphenicol acetyltransferase (CAT) gene, and p(BA1) 5 CAT, containing five copies of the HNF-4␣ binding element from the apoB gene in front of the minimal apoB promoter and the CAT gene, were used as the reporters to assay the degree of transactivation (22). Plasmid apoCIII(Ϫ890/ϩ24)CAT, containing the natural apolipoprotein CIII promoter, was also used in transcriptional analysis of the pcDNA3.1-LBD point mutants (23).
The pEV-BirA and pcDNA3.1-Bio plasmids, encoding the bacterial biotin ligase BirA and the 23-amino-acid biotinylation epitope, were generous gifts of Drs. John Strouboulis (Erasmus University Medical Center, Rotterdam, the Netherlands) and Dr. Dimitris Kardassis (University of Crete), respectively. The Bio-PGC-1 plasmid was created by amplifying PGC-1 from the HA-PGC-1 vector (forward primer 5Ј-ATT GCC GAT ATC GC ATG GCG TGG GAC ATG TCC A-3Ј and reverse primer 5Ј-GGC AAT GCG GCC GC TTA CCT GCG CAA GCT TCT CT-3Ј). The underlined sequences show the EcoRV and NotI restriction sites engineered by the primers into the 5Ј and 3Ј end of the PGC-1 sequence, respectively, with the addition of nucleotides where appropriate (in bold), to clone the sequence into frame with the Bio-epitope sequence in the pcDNA3.1-Bio vector. The resulting Bio-PGC-1-expressing plasmid was used in cell transfections for in vivo copurification assays. Cell Transfections and CAT Assays-HepG2, COS-7, and HEK-293 cells were maintained as stocks in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum. 50 -60% confluent 30-mm dishes were transfected using the calcium phosphate coprecipitation method. Plasmids were transfected into COS-7 and/or HepG2 cells and assayed for their ability to promote transcription of the CAT gene. The transfection mixture contained 3 g of the CAT reporter plasmid, 200 ng of the GAL-HNF-4␣, pCMXGAL-LBD or pcDNA3.1-LBD wild type or mutant plasmids, 1 g of cytomegalovirus ␤-galactosidase plasmid and, in coactivator interaction experiments, 1.25 g of PGC-1 or SRC-3 plasmid. In each case, vector DNA was added as necessary to achieve a constant amount of transfected DNA (5.45 g). 40 h after transfection, the cells were washed with PBS and collected in TEN solution (0.04 M Tris-HCl, pH 7.8, 1 mM EDTA, pH 8.0, 0.15 M NaCl). Whole cell extracts were prepared in 0.25 M Tris-HCl, pH 7.8, by three sequential freezethaw cycles. The ␤-galactosidase activity of cell lysates was determined as described previously (22), and the values obtained were used to normalize variability in the efficiency of transfection. CAT activities were determined using [ 14 C]chloramphenicol and acetyl-CoA as described previously (22). CAT enzyme levels that exhibited more than 60% conversion of acetylated product were diluted and re-assayed for CAT activity in the linear range. The results represent the mean of at least three independent transfection experiments, each carried out in duplicate.
Electrophoretic Mobility Shift Assay (EMSA)-The various pcDNA3.1-LBD mutants were transfected into HEK-293 cells, and nuclear extracts were isolated and used to evaluate the dimerization and DNA binding properties of all mutants as described previously (24). A doublestranded oligonucleotide corresponding to the B regulatory element of the apoCIII promoter (CIIIB), which is a high affinity binding site for HNF-4␣, was used as a probe (23). The double-stranded oligonucleotide is composed of CIIIBfor, 5Ј-GGTCAGCAGGTGACCTTTGCCCAGCG-3Ј, and the complementary CIIIBrev, 5Ј-CGCTGGGCAAAGGTCACCT-GCTGACC-3Ј. Nuclear extracts were incubated with the 32 P-labeled double-stranded oligonucleotide probe for 30 min at 4°C in the presence of 25 mM HEPES, pH 7.6, 40 mM KCl, 1 mM dithiothreitol, 5 mM MgCl 2 , and 0.6 g of poly(dI-dC). Protein-DNA complexes were analyzed by electrophoresis in a 5% nondenaturing gel followed by autoradiography. In dimerization experiments, extracts from cells transfected with the point mutants were incubated with extracts from CD1b transfectants for 15 min prior to the addition of the probe. CD1b is a truncated form of HNF-4␣ that has been previously shown to retain the wild type DNA binding and dimerization properties (9).
GST Pull-down Assays-Cells of the BL21 Escherichia coli strain were transformed with the pGEX-4T2 and pGEX-PGC-1 (aa 91-408) expression plasmids, and isolated colonies were grown in 200-500 ml of appropriate selective medium. The cultures were induced with 0.5 mM isopropyl-1-thio-␤-D-galactopyranoside and, 3 h later, cells were harvested and lysed in 1 volume of cold lysis buffer (100 mM EDTA, 2 mM phenylmethylsulfonyl fluoride, 2 g/ml aprotinin, 2 mg/ml lysozyme, 10 mM ␤-mercaptoethanol, 1% Triton X-100 in PBS), with sonication. The cell debris was pelleted, and supernatant was incubated for 1 h at 4°C with 0.1 volumes of glutathione-agarose beads (Sigma) that were equilibrated in PBS. After binding, the beads were washed four times in cold PBS and resuspended in an equal volume of cold washing buffer (20 mM Hepes, pH 7.7, 75 mM KCl, 2.5 mM MgCl 2 , 0.1 mM EDTA, 2 mM dithiothreitol, 0.05% Nonidet P-40). Nuclear extracts from HEK-293 transfected cells were prepared. Cells were resuspended in hypotonic buffer A (10 mM Hepes, pH 7.9, 10 mM KCl, 1.5 mM MgCl 2 , 0.5 mM dithiothreitol) supplemented with protease inhibitors (300 M phenylmethylsulfonyl fluoride and 200 M leupeptin) and incubated on ice for 10 min. Samples were homogenized and centrifuged (10,000 ϫ g, 5 min, 4°C). Pellets were resuspended in nuclear extraction buffer B ϩ (20 mM Hepes, pH 7.9, 1.5 mM MgCl 2 , 20% (v/v) glycerol, 0.5 mM dithiothreitol, 0.3 M KCl) supplemented with protease inhibitors (300 M phenylmethylsulfonyl fluoride and 200 M leupep-tin) and extracted on ice for 30 min with occasional mixing. Samples were centrifuged (10,000 ϫ g, 5 min, 4°C), and 2 volumes of buffer B Ϫ (20 mM Hepes, pH 7.9, 1.5 mM MgCl 2 , 20% (v/v) glycerol, 0.5 mM dithiothreitol, 0.1 M KCl) were added to the supernatant. GST interactions were carried out for 1 h at 4°C in reaction volumes of 500 l with cold interaction buffer (same as washing buffer plus 10% glycerol) by using 2 g of GST or GST fusion protein and 20 -60 l of the various nuclear extracts per reaction. The precise amounts of nuclear extracts used in each interaction were normalized to mutant expression, as assessed by Western blot analysis. The beads were extensively washed in cold washing buffer after each interaction. Proteins were eluted in 30 l of 2ϫ loading buffer (120 mM Tris-HCl, pH 6.8, 4% SDS, 20% glycerol, 10% ␤-mercaptoethanol, and 0.02% bromphenol blue) by boiling for 5 min, and they were subsequently analyzed on a 10% SDS-polyacrylamide gel followed by Western blot analysis, as described previously (21). In detail, proteins were transferred to an Immobilon P membrane (Millipore, Bedford, MA) by electroblotting, and the membranes were preincubated in PBS containing 0.1% Tween 20 (PBST), 5% nonfat dry milk, and 0.5% bovine serum albumin for 1 h at 25°C. Subsequently, they were incubated with the primary goat polyclonal anti-HNF-4␣ C-19 antibody (Santa Cruz Biotechnology) at a dilution of 1:5,000 in PBST for 1 h at 25°C. Membranes were washed with PBS containing 0.1% Tween 20 and incubated with the secondary antibody, rabbit anti-goat IgG, conjugated to horseradish peroxidase (Santa Cruz Biotechnology) at a dilution of 1:10,000 in PBST for 1 h at 25°C. Membranes were washed in PBS containing 0.1% Tween 20 and once in PBS, and proteins were visualized by exposure to ECL Plus reagent (Amersham Biosciences) according to the manufacturer's specifications.
Copurification of Biotin-tagged PGC-1 and HNF-4␣ from Cell Extracts-50 -60% confluent 60-mm dishes of HEK-293 cells were cotransfected with 5 g each of the following plasmid vectors: pEV-BirA, Bio-PGC-1, and the pcDNA3.1-LBD wild type or mutant plasmids. Empty vector DNA was added as appropriate to achieve a constant amount of transfected DNA. Cells were harvested 48 h after transfection in 1 ml of lysis buffer (20 mM Tris-HCl, pH 7.5, 150 mM NaCl, 10% glycerol, 1% Triton X-100, plus protease inhibitors) by gentle rocking for 20 min at 4°C. Depending on protein expression, 100 -200 l of the cell extracts were added to 50 l of 50% streptavidin beads (Sigma), in lysis buffer, up to a final volume of 500 l/reaction, and reactions were incubated by rotation at 4°C overnight. Following extensive washing of the beads, the bound proteins were eluted by boiling for 5 min in 20 l of 2ϫ loading buffer, as described previously (25), and they were electrophoretically separated on 8% SDS-polyacrylamide gel. The proteins were then transferred to an Immobilon P membrane (Millipore) as described above, and membranes were incubated with a horseradish peroxidase-conjugated streptavidin polymer (Sigma) at a 1:7,500 dilution for 1 h at room temperature following incubation in Tris Buffered Saline ϩ0.1% Tween-20 containing 5% nonfat dry milk and 0.5% bovine serum albumin for 1 h at room temperature. Biotinylated PGC-1 was visualized by exposure to ECL Plus reagent (Amersham Biosciences), according to the manufacturer's instructions. Following stripping of the membranes in 25 mM glycine, 1% Triton X-100, 1% SDS stripping solution, pH 2.5, membranes were reincubated with the anti-HNF-4␣ C19 primary antibody (Santa Cruz Biotechnology) and horseradish peroxidase-conjugated anti-goat secondary antibody (Santa Cruz Biotechnology) at the dilutions described above. Proteins were again visualized with the aid of the ECL Plus reagent (Amersham Biosciences).
Having identified the importance of the AF-2 AD core motif (⌽⌽X⌭⌽⌽) for PGC-1 coactivation, we wished to examine the effect of a point mutation in the conserved amino acid residue Glu-363 in the context of the full-length HNF-4 molecule. Furthermore, it was interesting to map potential residues of the D domain important for PGC-1-mediated transactivation by introducing point mutations. It was previously proposed that Arg-168 of the D domain could be important for HNF-4␣ activity since the homologous residue in retinoic acid receptor-␥ and retinoid X receptor-␣ was shown by crystallographic analysis to form part of the -loop that switches position upon ligand binding and plays a role in the repositioning of H12 (11). For this reason, we constructed mutants pcDNA3.1 E363K and pcDNA3.1 R168G/K170N/K171N, a triple mutation that eliminates positive charge around Arg-168. Before studying their effect on HNF-4␣-mediated transcription, we tested whether they retained the ability to bind to DNA either as homodimers or as heterodimers with WT HNF-4. Fig. 2A shows the results of EMSA analysis, in which it was verified that both mutants bind the CIIIB probe with equal intensity as the wild type. To test for dimerization, CD1b, a truncated HNF-4␣ protein that was previously shown to retain binding and dimerization properties (9), was included in the binding reactions. The presence of an intermediate band seen between the complexes of the point mutant and the CD1b homodimers confirmed the ability of both E363K and R168G/K170N/K171N to heterodimerize with the wild type. In Fig. 2A (right panel), Western blot analysis in HEK-293 transfected cells revealed equal expression of WT and mutant proteins that were used in the EMSA analysis.
When mutant R168G/K170N/K171N was tested by transient transfection experiments for its ability to activate transcription from the apoCIII promoter, it was noticed that it retained a significant degree of activity, rising up to 60% of the activation attained by the wild type (Fig. 2B). The addition of PGC-1 further increased activity by 2-fold, versus 4-fold of the wild type, indicating that the triple mutation in the D region did not reduce coactivation potential too drastically.
On the contrary, a set of similar experiments for mutant pcDNA3.1 E363K showed that this mutation had a strong effect on the transcriptional activation of the apoCIII promoter causing a 90% reduction of the wild type HNF-4␣ activity (Fig.  2C). The presence of PGC-1 was not able to overcome the effect of the mutation and restore transcription from the native apo-CIII promoter. Similar results were obtained with the HNF-4 homopolymeric promoter construct (BA1) 5 CAT, where again the coactivator failed to restore the compromised activity of the E363K mutant (Fig. 2C).
The drastic reduction in transcriptional activity observed in the presence of mutant E363K led us to test whether this mutation could act as dominant negative. When COS-7 cells were cotransfected with a constant amount of wild type HNF-4␣ and an increasing concentration of the mutant expression plasmid, there was a suppression of HNF-4␣ activity (Fig.  2D), which further confirmed the ability of this mutant to heterodimerize with wild type HNF-4␣ and proved the severity of the mutation.
The R168G/K170N/K171N and E363K mutants studied above are part of the AF-2 domain, but they are located outside the ligand binding pocket. To further explore the significance of specific residues of the HNF-4␣ ligand binding pocket for coactivation potential, we turned to residues present in helices H3, H5, H10, and H11 of the HNF-4␣ LBD that were shown previously to be critical for its transcriptional activity (21). We sought to examine whether these mutations, apart from suppressing HNF-4␣ transcriptional activation potential, also influence the ability of coactivators to enhance transcription from synthetic reporter constructs or natural HNF-4␣ target promoters. The pCMXGAL-LBD constructs of HNF-4␣ harboring mutations S181Y, M182K, R212G, L219Q, L220Q, R226G, I338F, and I346F were used in cotransfection experiments with the pG 5 CAT reporter plasmid and the PGC-1 coactivator or the p160 member SRC-3. We found that S181Y, R212G, R226G, and I346F retain the ability to be coactivated by both PGC-1 and SRC-3, whereas M182K, L219Q, L220Q, and I338F are not responsive to coactivators (Fig. 3, A and B). It is noteworthy that although mutations S181Y and R226G exhibited a 40 -60fold increase in activity in the presence of PGC-1 and a 10 -15fold increase in the presence of SRC-3, versus a 4-and 2.5-fold increase in wild type activity, respectively, these mutations were unable to stimulate transcription in the absence of coactivators (Fig. 3, A and B). On the other hand, mutations in residues 182, 219, 220, and 338 were shown to be inactive both in the absence and in the presence of coactivators. Interestingly, the increased transactivation observed in pCMXGAL-LBD mutants S181Y, R226G, and I346F in the presence of SRC-3 was less pronounced when compared with the one observed with PGC-1 (10 -15 versus 40-fold enhancement), and a FIG. 1. Coactivation of GAL-HNF-4␣ and its deletion mutants by PGC-1 (A). The reporter plasmid pG 5 CAT was cotransfected into HepG2 cells with pCMV-␤-galactosidase and effector plasmids expressing GAL-HNF-4␣ or the indicated deletion mutants, in the absence or presence of PGC-1. The relative CAT activity (ϮS.E.) of three independent experiments is shown as the percentage of the activity obtained with the pG 5 CAT reporter construct cotransfected with GAL-HNF-4␣ wild type. A schematic representation of GAL-HNF-4␣ and its deletion mutants is shown on the left. B, expression of the GAL-HNF-4␣ deletion mutants indicated in A. All mutants are expressed in comparable levels, which cannot account for the differences observed in transcriptional activity.

FIG. 2. DNA binding, dimerization and transcriptional properties of HNF-4␣ and its point mutants R168G/K170N/K171N and E363K.
A, an EMSA analysis of DNA binding was performed using the 32 P-labeled double stranded oligonucleotide CIIIB as a probe. Protein-DNA complexes were analyzed by electrophoresis in 5% nondenaturing gels followed by autoradiography. The lower arrow indicates homodimers of the HNF-4␣ deletion mutant CD1b, whereas the upper arrow indicates the presence of heterodimers between CD1b and point mutants. The expression of proteins used in the binding reactions is shown with Western blot analysis on the right. B, the effect of mutant R168G/K170N/K171N on the enhancement of HNF-4␣ transcriptional activity by PGC-1 using the apoCIII-890CAT reporter construct. COS-7 cells were transiently transfected with 0.2 g of wild type or mutated HNF-4␣ expression vector and 1.25 g of empty vector or PGC-1. The relative CAT activity (ϮS.E.) of three independent experiments is shown as the percentage of the activity obtained with the apoCIII-890CAT reporter construct cotransfected with pcDNA3.1-HNF-4␣ wild type LBD. C, the effect of mutant E363K on the enhancement of HNF-4␣ transcriptional activity by PGC-1 using the similar pattern was observed in all mutants tested, indicating that overall, PGC-1 is a much stronger coactivator for HNF-4␣ than SRC-3.
Next we went on to assess the ability of the point mutations to affect PGC-1 and SRC-3 stimulation of HNF-4␣-mediated transcription when the mutations were expressed in the context of the intact HNF-4 molecule. For this purpose, the apo-CIII promoter was used in cotransfection experiments with the pcDNA3.1 HNF-4␣ wild type or LBD point mutant plasmids in COS-7 cells, in the presence of PGC-1 (Fig. 4). In the context of the intact molecule, point mutations S181Y, R226G, and I346F behaved exactly the same way as in the pCMXGAL-LBD chimeric constructs, enhancing HNF-4␣-mediated transcription by 8-, 33-, and 2.2-fold, respectively, whereas mutants M182K, L219Q, and I338F remained inactive despite the addition of coactivators, as it was also observed with the respective GAL-HNF-4 chimeric constructs.
Physical interactions of the pcDNA3.1-LBD point mutants with the coactivator molecules were studied in an effort to explain the above functional data. We reasoned that point mutations, which abolished PGC-1 and SRC-3 stimulation, should exhibit a reduced, if any, interaction with the coactivator, whereas point mutations that retained a degree of stimulation by PGC-1 and SRC-3 should be able to show a substantial physical interaction in in vitro and in vivo experiments. We thus proceeded to the expression, in E. coli BL21 bacterial cells, of GST-PGC-1 (aa 91-408), a truncated form of the PGC-1 molecule that contains the intact LXXLL motif required for interaction with nuclear receptors (16). The fusion protein was purified and immobilized on glutathione-agarose beads, and the appropriate amounts were incubated with nuclear extracts from HEK-293 cells that were transfected with the pcDNA3.1 wild type or various LBD point mutants. As seen in Fig. 5A, mutant S181Y, as well as mutant I346F, retained a substantial apoCIII-890CAT and (BA1) 5  degree of interaction with GST-PGC-1 (aa 91-408) when compared with the wild type, whereas I338F showed a much weaker interaction and, in the case of mutant M182K, no interaction was visible. Naturally, the mutations do not interact with GST protein alone in control experiments.
Mutant L219Q also showed no detectable interaction with GST-PGC-1 (aa 91-408), but the expected interaction of mutation R226G with GST-PGC-1 (aa 91-408) could not be observed in any of the repeated in vitro experiments, although the proteins were equally expressed (Fig. 5A, upper panel), which led us to seek an alternative method to investigate whether the two molecules physically interacted in the natural context of the cell. For this reason, HEK-293 cells were transfected with the pcDNA3.1 HNF-4␣ wild type or LBD point mutants and a tagged protein (Bio-PGC-1). The tag consists of a 23-amino-acid biotinylation epitope that was cloned in the amino terminus of PGC-1 and, by the use of streptavidin beads that bind strongly to biotinylated templates (25), we were able to pull down the tagged PGC-1 protein, as well as putative proteins in complex with PGC-1. In the case of WT and mutant R226G, Western blot analysis of the coprecipitated proteins revealed the retention of HNF-4␣ in addition to Bio-PGC-1, which confirmed the in vivo interaction of PGC-1 with the mutant R226G (Fig. 5B). On the contrary, this was not observed with mutant L219Q, where no HNF-4␣ could be detected (Fig. 5B). Thus, the protein-protein interaction experiments corroborate our functional data showing that specific LBD residues are involved in coactivator interactions despite their inability to activate transcription alone. The results arising from this study are summarized in Fig. 6. In the presence of coactivator, mutant R226G regains activity, similar to the wild type, whereas mutants M182K, L219Q, and I338F remain inactive. DISCUSSION In contrast to other members of the nuclear receptor superfamily, HNF-4␣ has been known to activate transcription as a constitutively active transcription factor. However, when HNF-4␣ and -␥ were recently crystallized in the absence of added ligands (7,8), the ligand binding pocket of both receptors contained a mixture of endogenous saturated and monounsaturated fatty acids in the crystals. The fatty acids could not be displaced from the pockets even after prolonged dialysis or partial denaturation of the proteins (7,8), leading to the proposal that these endogenous molecules/ligands may be constitutively bound to the LBD of HNF-4, acquired soon after or even during the folding of the protein. This hypothesis offered an explanation for the constitutive transcriptional activity of HNF-4, and it was proposed that additional modifications or cofactor recruitment could be the critical factors for modulating HNF-4 activity in cells (8). In this context, we found it challenging to conduct a study in which we would examine in detail the role of coactivators in HNF-4␣-mediated transcription by mapping the regions of HNF-4␣ that are important for functional interactions with coactivators and examining whether A, wild type HNF-4␣ and its point mutants were transfected in HEK-293 cells and were incubated with GST and GST-PGC-1 (91-408 aa) fusion protein immobilized on glutathione-agarose beads. After extensive washing of the beads, the bound proteins were eluted, separated by SDS-PAGE, and detected by ECL using anti-HNF-4␣ antibody. In the upper part, HNF-4␣ protein input is shown. B, the expression of HNF-4 WT and its mutants transfected in HEK-293 cells was monitored by Western blot (WB) analysis using anti-HNF-4␣ antibody. In the lower part, nuclear extracts expressing HNF-4, Bio-PGC-1, and pEV-BirA ligase were incubated with streptavidin beads. The bound proteins were eluted and examined by SDS-PAGE and Western blot analysis using streptavidin-horseradish peroxidase to detect bound biotinylated PGC-1. The same membrane was stripped and reprobed with anti-HNF-4␣ antibody, and physical interactions of wild type HNF-4␣ or its point mutants with Bio-PGC-1 were detected. residues that we previously showed to be critical for constitutive HNF-4␣ activity (21) would also affect stimulation by coactivators.
Transient transfection experiments were first carried out to analyze the importance of specific regions for functional interactions with the coactivator PGC-1. In a recent study, in vitro binding experiments were conducted to investigate the HNF-4␣/PGC-1 physical interactions, and it was demonstrated that the AF-2 AD core motif of the receptor and the LXXLL motif of the coactivator are required for a strong interaction (16). However, the physical interaction was not entirely abolished either by a point mutation in the LXXLL core motif of PGC-1 or by deletion of the AF-2 AD core motif of HNF-4␣, implying that there may be additional domains mediating physical interaction of the two proteins both in vitro and in vivo (16). We thus went on to conduct a detailed functional analysis of HNF-4␣/ PGC-1 interactions by using a set of GAL-HNF-4␣ deletion constructs (9). Indeed, we found that the D domain (particularly amino acids 160 -175), in addition to the intact AF-2 AD core motif, were indispensable for the observed synergism between HNF-4␣ and PGC-1 (Fig. 1). This finding about the requirement of the H1-containing D region is not unique among nuclear receptor-coactivator interaction studies. In fact, similar interactions of PGC-1 with helix 1 of the thyroid hormone receptor TR␤1 and the hinge domain of estrogen receptor ␣ (ER␣) have been reported to be required for interaction of both receptors with the coactivator (18,19).
The observations rising from the study with the deletion constructs led us to investigate in more detail the effect of mutagenesis of particular residues involved in those two important regions of activation. In a previous study, it was proposed that Arg-168 of the D domain could be important for HNF-4␣ activity since the homologous residue in retinoic acid receptor-␥ and retinoid X receptor-␣ was shown by crystallo-graphic analysis to form part of the -loop that switches position upon ligand binding and plays a role in the relative position of H12 (11). We constructed a mutant in which the positive charge of three neighboring residues of the D domain was removed. Mutation pcDNA3.1 R168G/K170N/K171N retained 60% of wild type activity in transfected cells, and this activity was further stimulated 2-fold by PGC-1. Since H1 has been proposed to play a central role in stabilizing the global structure of the LBD (26), we expected that removing the charge of residues immediately adjacent to and within H1 would destabilize the overall structure, resulting in a pronounced reduction in activity. However, the effect of the triple mutation was subtler than expected, not only in retaining wild type DNA binding and dimerization properties ( Fig. 2A) but also in the degree of reporter gene activation that it induced (Fig. 2B). On the contrary, predictions as to the effect of mutation of the conserved glutamate at position 363 (10) were confirmed in the context of the full-length receptor with the inversion of charge (E 3 K), resulting in a drastic reduction of activity and abolishment of coactivation by PGC-1 (Fig. 2C), although it did not affect DNA binding ( Fig. 2A).
The interactions between coactivators and the AF-2 regions of nuclear receptors have been shown, by crystallographic and mutagenesis studies, to occur through an AF-2-coactivator interaction interface formed by both charged and hydrophobic residues that are located mainly in helices H3, H5, and H12 of receptor LBDs and the LXXLL motif(s) of coactivators (27). In the case of estrogen receptor ␣, mutagenesis identified a cluster of residues from helices H3, H5, H6, and H12 as being of critical importance for driving the coactivator reaction (27), and in the case of the holo-retinoid X receptor-␣ crystal, observations were similar. H12 in the active conformation was packed against helices H3, H4, H5, and H11 mainly by a set of tight hydrophobic interactions, although hydrogen bonds between residues of H3, H5, and H12 also stabilized H12 packing (28). According to the authors, this pattern was similar to the one observed in the holo-LBDs of other nuclear receptors, with residues of the H12 interaction groove stabilizing H12 onto the main body of the LBD, and residues of the H12 core motif forming the hydrophobic cleft that corresponds to the coactivator binding site (28).
In keeping with the above observations, in the HNF-4␣ crystal and in the closed state of the receptor, residues Leu-360, Leu-361, Met-364, and Leu-365 of H12 were found to form a hydrophobic face, which interacted with a corresponding hydrophobic surface on the body of the domain created by residues from H3 (Met-182, Lys-183, Leu-186, Leu-187, Leu-189, Val-190) and H4 (Leu-211, Ala-215, Gly-216, Leu-219) (7), thus packing H12 onto the main body of the receptor. The authors proposed that since the fatty acid ligands similarly occupied the ligand binding pockets of molecules in the open and closed conformations, ligand binding in itself does not seem to serve as the structural "switch" that triggers transition to the active state (7). In fact, very recently, crystallization of the HNF-4␣ LBD in the presence of both fatty acid and a coactivator peptide sequence derived from SCR-1 strengthened this hypothesis since both HNF-4␣ molecules in the ternary complex were found in the closed state (10). In contrast, in the crystal of HNF-4␣ LBD without coactivator, one molecule in each homodimer was found to be in the open state, and the other was found to be in the closed state. It can be thus argued from crystallographic studies that coactivator binding indeed locks the LBD in a closed and compact shape.
In our previous functional analysis of LBD amino acid residues that line the ligand binding pocket in HNF-4␣ (21), in an effort to map residues that are critical for HNF-4␣ constitutive FIG. 6. Schematic presentation of the mode of action of point mutations introduced in the HNF-4␣ LBD region. A, wild type HNF-4␣ is active in the absence of coactivator, and its action is enhanced by its presence. B, LBD point mutants M182K, L219Q, R226G, and I338F, which are all inactive in the absence of coactivator, behave differently in its presence, indicating that distinct amino acid residues may be involved in ligand and coactivator interactions. activity, we had mutated residues in H3, H5, H10, and H11 and highlighted their importance in HNF-4␣-mediated transactivation. Among them were Met-182 of H3 and Leu-219 of H5, which were shown from the crystal structure to be involved in H12-stabilizing interactions in the closed state of the receptor. Indeed, we found that mutation of residues Met-182 and Leu-219, but also Ser-181 in H3, Leu-220 and Arg-226 in H5, Ile-338 in H10 and, to a lesser degree, Ile-346 in H11, severely affected transactivation when mutated, and we employed structural modeling to reveal the potential disturbance that these mutations caused on ligand binding and the local microenvironment. Importantly, with the exception of Ile-338, all of the above residues were found from crystallographic analysis to come in direct contact with the ligand. The severe reduction almost all of the mutations inflicted on HNF-4␣ transactivation potential could be explained not only by perturbations they caused in direct ligand binding and in correct orientation of the ligand but also by local changes they produced in interactions with residues of their immediate vicinity (21).
On this basis, we postulated that some of the mutations could also affect H12-stabilizing interactions, which create a favorable conformation for interaction with coactivators. In this study, transient transfection experiments were employed in which the pCMXGAL-LBD or pcDNA3.1-LBD wild type and mutant plasmids were transfected into cells together with expression vectors for the PGC-1 or SRC-3 coactivators. The results of these experiments are presented in Figs. 3 and 4. In particular, an elimination of transcriptional enhancement by PGC-1 and SRC-3 was observed in mutants M182K, L219Q, L220Q, and I338F, which were also inactive in the absence of coactivator. Furthermore, coactivation potential was maintained in mutant I346F, which retains a significant degree of transcriptional activity even in the absence of coactivator. However, and most unexpectedly, point mutant S181Y, which retained a low level of activity, and especially R226G, which was inactive in the absence of coactivator, were shown to regain transactivation potential in the presence of coactivators (Figs. 3, 4, and 6).
In an attempt to interpret the functional rescue of these mutations, we addressed the issue of their physical interactions with the coactivator PGC-1. As seen in Fig. 5A, the S181Y and I346F point mutants retained the ability to physically interact with the LXXLL-containing PGC-1 fragment, much like the wild type and in agreement with the functional data, whereas a weaker interaction was observed in the case of mutant I338F, which could not be transcriptionally enhanced by PGC-1, and no interaction could be seen for mutant M182K, as expected. However, under the conditions employed, it was not possible to detect an interaction in the case of mutant R226G, which, according to the functional data, should exist. This led us to adopt an alternative strategy. As seen in Fig. 5B, mutation R226G interacted with the biotinylated coactivator (Bio-PGC-1) in HEK-293 cotransfected cells, in contrast to mutation L219Q, which could not be enhanced by the action of PGC-1 in transient transfection experiments (Figs. 3 and 4) nor bound to it in vivo (Fig. 5B). Thus, our functional data can be interpreted in terms of the physical interactions observed between the HNF-4␣ point mutants and PGC-1.
Based on the results described above, LBD residues such as Met-182 and Leu-219 seem to play a key role in all aspects of HNF-4␣ activity by affecting both constitutive and coactivatorstimulated activity, whereas residues such as Arg-212 and Ile-346 may not greatly influence either type of interaction. Interestingly, Met-182 and Leu-219 participate in the H12stabilizing hydrophobic interactions that are observed in the closed state of the receptor (7). In contrast to Met-182 and Leu-219, mutation of residue Arg-212, which is also involved in an H12-stabilizing interaction by formation of a hydrogen bond with residue Glu-363 of H12 (7), does not appear to affect activation or coactivation potential of the HNF-4␣ protein. On the other hand, the role of amino acids such as Arg-226 is perplexing; although we show Arg-226 to be dispensable for stimulation by coactivators, it was revealed by crystallographic data to be the key residue involved in stabilizing the ligand in the LBD pocket. This finding reveals the existence of residues with a dual function in the LBD pocket of HNF-4␣. It can be thus argued that the interface involved in nuclear receptorcoactivator interactions is distinct from the ligand binding interface. Indeed, mutagenesis of R226G, which has a dominant negative effect in the absence of coactivator by suppressing endogenous HNF-4␣ activity in HepG2 cells (21), still permits the receptor to assume its activated closed state when excess amounts of coactivator are present. Rather than ligand, it is coactivator interactions that seem to be the "key" in determining activation of transcription by HNF-4␣. Since coactivators are expressed in restricted amounts in vivo, it is logical to assume that their availability in cells is what plays the major role in deciding the state of activation of a given nuclear receptor, such as HNF-4␣.
Our findings indicate the distinct roles that ligand and coactivator may play in transcriptional activation through binding to separate, although overlapping, surfaces to stabilize the same LBD structure. We propose that distinct residues may be important for coactivator and ligand interactions in nuclear receptor LBD regions, although some residues may be critical for both functions.