Ligand Specificity and Conformational Dependence of the Hepatic Nuclear Factor-4 (cid:1) (HNF-4 (cid:1) )*

Hepatic nuclear factor-4 (cid:1) (HNF-4 (cid:1) ) controls the expression of genes encoding proteins involved in lipid and carbohydrate metabolism. Fatty acyl-CoA thioesters have recently been proposed to be naturally occurring ligands of HNF-4 (cid:1) and to regulate its transcriptional activity as function of their chain length and degree of unsaturation (Hertz, R., Magenheim, J., Berman, I., and Bar-Tana, J. (1998) Nature 392, 512–516). However, the apparent low affinities ( (cid:2) M K d values) obtained with a radiolabeled fatty acyl-CoA ligand binding assay raised questions regarding the physiological significance of this finding. Fur-thermore, it is not known whether interaction with fatty acyl-CoA alters the structure of HNF-4 (cid:1) . These issues were examined using rat recombinant HNF-4 (cid:1) ligand-binding domain (HNF-4 (cid:1) LBD) in conjunction with photon counting fluorescence and circular dichroism. First, fluorescence resonance tion of change highly dependent on the specific type of fatty acyl-CoA.

Hepatic nuclear factor 4 (HNF-4) 1 is a member of the superfamily of nuclear receptors which includes steroid hormone receptors and nonsteroid ligand dependent transcription factors, e.g. thyroid hormone receptor, retinoid X receptor, retinoid acid receptor, and peroxisome proliferator-activated receptors (PPARs) (reviewed in Ref. 1). HNF-4␣ isoforms (␣ 1 -␣ 3 ) have been cloned and characterized and are expressed in mammals in liver, kidney, intestine, and pancreas (reviewed in Ref. 2). Unlike retinoid X receptor ␣, with which it has 40% amino acid sequence homology, HNF-4 does not form heterodimers with any other nuclear receptor, but binds to direct repeat-1 DNA sequences as homodimer (3). Direct repeat-1 motifs are promiscuous binding sites for HNF-4, PPAR, retinoid acid receptor, retinoid X receptor, chicken ovalbumin upstream-promoter transcription factor, and chicken ovalbumin upstream-promoter transcription factor homo-or heterodimers (4).
The first demonstration of putative HNF-4 ligands showed that long chain fatty acyl-CoA (LCFA-CoAs) thioesters (15), as well as CoA-thioesters of hypolipidemic peroxisome prolifera-tors (e.g. fibrate drugs, Medica homologues) (12) bind to HNF-4␣ and, depending on their chain length, degree of saturation, and respective substitutions, are able to activate or inhibit transcription of a reporter gene enhanced by the C3P element of the apoC-III promoter (12,15). However, the radioligand competition assay used to demonstrate acyl-CoA binding in these previous studies yielded K d values in the micromolar range (12,15). Although total tissue levels of LCFA-CoAs (free ϩ bound) are in the range 0.4 -164 M, depending on tissue, cell, nutrition, and pathology (reviewed in Refs. 16 and 17), free (unbound) LCFA-CoA are estimated to be 5-200 nM in cytosol (17) and only 1-10 nM in the nucleus (18). Based on these considerations as well as computer modeling of the HNF-4␣ structure, it has been suggested that HNF-4␣ does not significantly bind LCFA-CoAs (19). However, it must be considered that radioligand competition assays for lipidic ligand-binding proteins typically yield K d values that are 2-3 orders of magnitude higher (i.e. lower affinity) than K d values determined by direct fluorescence or microcalorimetry binding assays (reviewed in Ref. 20). Thus, it seems possible that the radioligand acyl-CoA competition binding results significantly underestimated the affinity of HNF-4␣ for its endogenous ligands.
The objective of the present investigation was to resolve these issues through (i) direct fluorescent ligand binding assays based on quenching of tryptophan emission by putative HNF-4␣ ligands using a recombinant HNF-4␣ ligand-binding domain (HNF-4␣LBD), (ii) fluorescence resonance energy transfer (FRET) between HNF-4␣LBD aromatic amino acids and bound cis-parinaroyl-CoA (naturally occurring fluorescent LCFA-CoA) to calculate the intermolecular distance between ligand and its binding site, and (iii) circular dichroism to characterize the secondary structure of HNF-4␣LBD as well as potential changes induced therein by LCFA-CoA binding.

Purification, SDS-PAGE, and Western Blotting of Recombinant HNF4␣LBD
His-tagged HNF-4␣LBD was purified by affinity chromatography on a nickel-NTA resin (Quiagen, Chatsworth, CA), desalted, lyophilized, and stored at Ϫ70°C. Prior to use, the protein was solubilized in 20 mM Tris-HCl buffer, pH 8.0, containing 0.3 M NaCl, 10% glycerol, and 1 mM 2-mercaptoethanol or dithiothreitol (binding buffer). Protein was determined by BCA Protein Assay (Pierce, Rockford, IL). Purity of recombinant rat His-HNF-4␣LBD (amino acids 132-455 of the wild type rat HNF-4␣1 with 6 His residues attached at the C terminus) was assessed by SDS-PAGE and Western blotting (Fig. 1). SDS-PAGE separated a single protein band of 36 kDa as detected by staining with Coomassie Blue and indicated 97% purity (Fig. 1, panel A). Western blotting with rabbit anti-rat HNF-4␣LBD polyclonal antibody followed by goat antirabbit IgG-alkaline phosphatase conjugate was performed as described (23). A single reacting band at the same molecular weight was detected by Western blot analysis (Fig. 1, panel B).

Synthesis and Purification of Fluorescent cis-Parinaroyl-CoA (cPNA-CoA)
Fluorescent cPNA-CoA was chemically synthesized (22) utilizing a naturally occurring fluorescent fatty acid, cis-parinaric acid (Molecular Probes, Eugene, OR). Unreacted, free cis-parinaric acid was retained into the organic phase of a chloroform:methanol:water mixture (24). Further removal of free CoA from cPNA-CoA in the aqueous phase was performed by HPLC, as described (25). The purity level of the HPLC fraction containing cPNA-CoA was checked by assessing the absorption spectrum with a UV/VIS spectrometer, model Lambda 2 (PerkinElmer Life Sciences, Norwalk, CT). HPLC of the aqueous extract showed two major peaks, 1 and 4 separated by over 32 min, as well as two minor, peaks 2 and 3 with retention times between 3 and 6 min ( Fig. 2A). The absorbance spectrum of the partially purified cPNA-CoA (Fig. 2B) exhibited a 260-nm maximum characteristic for CoA as well as three maxima within 280 -324 nm, characteristic for the conjugated double bonds of cPNA-CoA. The absorbance spectrum of HPLC-peak 1 (Fig. 2C) did not reveal the presence of a conjugated tetraene (peaks between 280 and 324 nm) of cPNA-CoA, but instead showed absorbance maximum near 260 nm, typical of free CoA (not shown). In contrast, the absorbance spectrum of peak 4 taken from the HPLC chromatogram in Fig. 2A exhibited absorbance characteristics of both the conjugated tetraene and thioester linkage present in cPNA-CoA (Fig. 2D). Absorbance spectra of oleoyl-CoA demonstrated only the presence of the thioester linkage but not the conjugated tetraene present only in cPNA-CoA (not shown).

Direct Ligand Binding Assay
Fluorescent Ligands-Direct binding assays not requiring separation of bound from free ligand were performed using cis-parinaroyl-CoA as described (25)(26)(27)(28)(29)(30). To establish specificity for LCFA-CoA, this assay was repeated using fluorescent fatty acids: cis-parinaric acid and NBDstearic acid. Briefly, HNF-4␣LBD was added to phosphate-buffered saline, pH 7.4, or binding buffer as described above to yield a final concentration of 170 nM HNF-4␣LBD. Increasing fluorescent ligand was then added to yield final concentrations of 10 -4000 nM cis-parinaroyl-CoA. Fluorescence emission spectra were obtained using a PC1 photon counting fluorimeter (ISS Inc., Urbana, IL) and maximal intensities measured. The K d and number of binding sites (n) were calculated as previously described (26,27,30). Where it was possible (e.g. NBDstearic acid) the number of binding sites was determined with a higher degree of accuracy from the reverse titration binding curve in which a constant amount of ligand was titrated with increasing concentration of HNF-4ϰLBD.
Quenching of Intrinsic Fluorescence of HNF-4␣LBD Aromatic Amino Acids Tyr/Trp or Trp-To investigate direct binding of nonfluorescent LCFA-CoAs to HNF-4␣LBD, the effect of LCFA-CoA binding on intrinsic fluorescence of HNF-4␣LBD aromatic amino acids was examined. Briefly, 170 nM HNF4␣LBD were titrated with increasing ligand (5-4000 nM). Two types of fluorescence measurements were obtained. First, both Tyr and Trp residues were excited at 280 nm. Second, the contribution of Trp from Tyr residues was resolved by selective excita- tion of Trp at 295 nm. Since fluorescence emission was maximal at 333 nm in both cases, Trp was the main contributor to HNF-4␣LBD fluorescence emission. All steady state fluorescence measurements were performed using a PC1 photon counting fluorimeter (ISS, Champaign, IL) in the L-format with a 300-watt xenon arc lamp light source and 4 nm band-passes in both excitation and emission monochromators. In ligand-induced HNF-4␣LBD fluorescence quenching experiments, the number of binding sites (n) was estimated by fitting the binding curve to a Hill plot according to the equation, y ϭ ax b /(c b ϩ x b ), where a is F max corresponding to B max ϭ nE 0 , b is the number of binding sites (n), and c is K d .

FRET Determination of Intermolecular Distance between the HNF-4␣LBD Trp and Bound cis-Parinaroyl-CoA
Examination of the excitation and emission spectra of HNF4␣-LBD revealed that the emission of Trp overlapped significantly with excitation of cis-parinaroyl-CoA, a condition ideal for Forster FRET (31). Since FRET varies as (intermolecular distance) 1/6 , the donor (HNF-4␣LBD Trp) and acceptor (cis-parinaroyl-CoA) residues must be in very close proximity, generally a few Å, for efficient FRET to occur. To determine the average intermolecular distance between HNF-4␣LBD Trp residues and bound cis-parinaroyl-CoA, HNF-4␣LBD (170 nM) was titrated with increasing cis-parinaroyl-CoA (5-4000 nM final concentration). The HNF-4␣LBD Tyr/Trp or Trp only, were excited at 280 and 295 nm, respectively. Maximal fluorescence emission intensities of HNF-4␣LBD Tyr/Trp or Trp were obtained at 333 nm while maximal emission of cis-parinaroyl-CoA was measured at 430 nm. If bound cis-parinaroyl-CoA acceptor was located within the optimal FRET distance from the HNF-4␣LBD Trp, then efficient energy transfer occurred such that HNF-4␣LBD Tyr/Trp or Trp fluorescence emission was quenched while, concomitantly, sensitized fluorescence emission of cisparinaroy-CoA was at 430 nm. The intermolecular distance between HNF-4␣LBD Trp and bound cis-parinaroyl-CoA was calculated from E ϭ R 0 6 /(R 0 6 ϩ R 2/3 6 ), where E is the FRET efficiency, R 0 is the critical distance for 50% efficiency, and R 2/3 is the actual distance between donor and acceptor (31)(32)(33). The energy transfer efficiency was calculated by quenching of HNF-4␣LBD Trp according to E ϭ 1 Ϫ F DA /F D , where F DA and F D are the fluorescence intensities of HNF-4␣LBD Trp (energy donor) at 333 nm in the presence and absence of cis-parinaroyl-CoA (energy acceptor) upon excitation of HNF-4␣LBD Trp at 295 nm. The transfer efficiency (E) was also calculated from the excitation spectrum of the energy acceptor (i.e. cis-parinaroyl-CoA) as previously described by Stryer (34) where G() is the magnitude of the corrected excitation spectrum of the energy acceptor excited at wavelength 1 ; ⑀ D () and ⑀ A () are the extinction coefficients of the donor and acceptor at wavelength . G is measured at two wavelengths: at 1 , where the donor has low absorption (320 nm), and at 2 , where the extinction coefficient of the donor is large compared with that of the acceptor (280 nm). This second method is especially useful when the local environment of the donor is different in the presence of the acceptor, provided that the absorption spectra of the donor and acceptor moieties are known (34). The critical distance for 50% efficiency (R 0 in Å) was calculated as described (35,36) using R 0 ϭ 0.211[k 2 n Ϫ4 Q D J()] 1/6 with wavelength expressed in nanometers and J(), the overlap integral, expressed as M Ϫ1 cm Ϫ1 nm 4 . Q D , the quantum yield for HNF-4␣LBD Trp and J() were calculated as described earlier (35,36). The orientation factor k 2 and the refractive index n were assumed to be 2/3 and 1.4, respectively, as for proteins in solution.

Circular Dichroism (CD) of HNF-4␣LBD
Far UV CD spectra of HNF-4␣LBD (2 M) were taken in the absence and presence of ligands as described (37), except that 2 mM Tris-HCl, pH 8, containing 30 mM NaCl, 1% glycerol, and 0.1 mM dithiothreitol was used as buffer. The CD measurements were performed with a J-710 Spectropolarimeter (Jasco, Baltimore, MD) using a 1-mm cuvette. Spectra were recorded from 250 to 195 nm at 50 nm/min with a time constant of 1 s and a bandwidth of 2 nm. For each CD profile an average of 10 scans was obtained. Percentages of various secondary structures in HNF-4␣LBD were calculated from the CD spectra by using the CDsstr program (38,39).

RESULTS
Determination of the Binding Parameters of HNF-4␣LBD for a Naturally Occurring Fluorescent Fatty Acyl-CoA: cis-Parinaroyl-CoA Quenching of HNF-4␣LBD Trp-cis-Parinaroyl-CoA was chosen as a fluorescent ligand to examine the fatty acyl-CoA binding properties (K d and B max ) of HNF-4␣LBD because: (i) cis-parinaroyl-CoA is a naturally occurring fluorescent fatty acyl-CoA with structure similar to other fatty acyl-CoAs (Table  I); (ii) cis-parinaroyl-CoA absorbance overlaps with aromatic amino acid fluorescence emission, thereby allowing fluorescence resonance energy transfer; (iii) cis-parinaroyl-CoA has been extensively used in fatty acyl-CoA fluorescence binding assays (25)(26)(27)(28)(29)(30). Furthermore, cis-parinaroyl-CoA fluorescence binding assay does not underestimate fatty acyl-CoA binding as do radioligand binding assays (reviewed in Refs. 20 and 40)). Thus, quenching of HNF-4␣LBD Trp emission at 333 nm was examined upon titration with increasing concentrations of cisparinaroyl-CoA. HNF-4␣LBD was excited either at 280 nm to excite both Trp/Tyr (Fig. 3A) or at 295 nm to selectively excite HNF-4␣LBD Trp (Fig. 3C). In both cases, the fluorescence emission of HNF-4␣LBD was quenched by increasing concentrations of cis-parinaroyl-CoA. Upon excitation of both Trp/Tyr (at 280 nM), the binding curve of cis-parinaroyl-CoA to HNF-4␣LBD demonstrated saturation binding (Fig. 3B). A reciprocal plot of the binding curve (Fig. 3B, inset) was linear, suggesting either a single binding site or two binding sites with the same K d . The number of binding sites was resolved by fitting the binding curve to a Hill plot, as described under "Experimental Procedures." The Hill plot yielded n ϭ 1.08, i.e. one binding site with a K d of 248.5 nM (Table II). The number of binding sites  (Table II). Taken together, measurement of the direct quenching of HNF-4␣LBD aromatic amino acid fluorescence by cis-parinaroyl-CoA indicated that HNF-4␣LBD bound cis-parinaroyl-CoA with nanomolar affinity at one binding site per monomer.
Molecular Interaction of cis-Parinaroyl-CoA with HNF4␣-LBD: FRET-To investigate in more detail the close molecular interaction of HNF4␣ with cis-parinaroyl-CoA, advantage was taken of the spectral properties of cis-parinaroyl-CoA absorbance and HNF4␣LBD Trp emission. FRET provided a sensitive technique for determining if interaction between cis-parinaroyl-CoA and HNF-4␣LBD represented close molecular binding rather than coaggregation. Comparison of a portion of the absorbance characteristics of cis-parinaroyl-CoA (Fig. 4, dashed line) with the emission of HNF-4␣LBD Trp, selectively excited at 295 nm ( Fig. 4, solid line), demonstrated partial but significant overlap. This overlap of the energy acceptor absorbance with the donor fluorescence emission is a required condition for efficient FRET. To determine whether FRET occurred between HNF4-␣LBD and bound cis-parinaroyl-CoA, HNF-4␣LBD was titrated with increasing concentrations of cisparinaroyl-CoA followed by determination of fluorescence emission spectra from 300 to 450 nm upon excitation of Tyr/Trp at 280 nm or of Trp only at 295 nm (Figs. 3 and 5). When HNF-4␣LBD Tyr/Trp were excited at 280 nm, Trp emission at 333 nm was highly quenched with increasing cis-parinaroyl-CoA (Fig. 3A). While this suggested that efficient FRET occurred between HNF-4␣LBD and cis-parinaroyl-CoA, a conformational change in HNF-4␣LBD upon ligand binding may also cause Trp fluorescence quenching. This possibility was resolved by the appearance of cis-parinaroyl-CoA-sensitized emission near 430 nm upon increasing cis-parinaroyl-CoA concentration (Fig. 5A). The increase in the acceptor cis-parinaroyl-CoA-sensitized fluorescence emission was even more apparent when HNF-4␣LBD Trp was selectively excited at 295 nm ( Fig. 5C). Since a conformational change in HNF-4␣LBD cannot elicit sensitized emission of cis-parinaroyl-CoA, these data indicated the close molecular interaction between HNF-4␣LBD and bound cis-parinaroyl-CoA.
Calculation of the Intermolecular Distance between HNF-4␣LBD and Bound cis-Parinaroyl-CoA by FRET-To determine the intermolecular distance between HNF-4␣LBD and bound cis-parinaroyl-CoA it was essential to remove complicating HNF-4␣LBD internal energy transfer from Tyr to Trp residues. This process occurred upon excitation of HNF-4␣LBD at 280 nm (excites both Trp and Tyr residues) since fluorescence emission was detected only from Trp near 333 nm (Fig.  3A). In contrast, preferential excitation of HNF-4␣LBD Trp at 295 nM resulted in maximal emission near 333 nm as expected (Fig. 3C). Since separate emission of Tyr (typically near 308 nm) was not observed upon excitation at 280 nm (Fig. 3A), this was consistent with efficient internal nonradiative transfer of energy from Tyr to Trp within HNF-4␣LBD.
Based on the significant overlap between the absorbance of cis-parinaroyl-CoA (Fig. 4, dashed line) with the emission of HNF-4␣LBD Trp (Fig. 4, solid line), it was possible to calculate R 0 , the critical distance for 50% energy transfer, as described under "Experimental Procedures." The calculated R 0 for the HNF-4␣LBD Trp/cis-parinaroyl-CoA donor/acceptor pair was 30 Å (Table III). Since FRET decreases with (intermolecular distance) 1/6 , efficient FRET will only occur between HNF-4␣LBD Trp and cis-parinaroyl-CoA if these donors/acceptors are close to 30 Å apart.
To determine R 2/3 , the actual intermolecular distance between HNF-4␣LBD and cis-parinaroyl-CoA, HNF-4␣LBD was titrated with increasing concentrations of cis-parinaroyl-CoA followed by selective excitation of HNF-4␣LBD Trp at 295 and determination of fluorescence emission between 300 to 450 nm, so that changes in both HNF-4␣LBD Trp fluorescence at 333 nm and of cis-parinaroyl-CoA emission at 430 nm were observed (Figs. 3C and 5C). The energy transfer efficiency was calculated from these data as described under "Experimental Procedures." Based on the equation that determines energy transfer efficiency with the assumption that no protein conformational change is induced upon ligand binding, a value of 85% energy transfer efficiency between HNF-4␣LBD Trp and cisparinaroyl-CoA was obtained (Table III). Thus, in the absence of a ligand-induced conformational change in HNF-4␣LBD, an intermolecular distance between HNF-4␣LBD and cis-parinaroyl-CoA, i.e. R 2/3 ϭ 22.6 Å, was determined. However, if fatty acyl-CoA binding elicits a conformational change in HNF-4␣LBD then this value may overestimate how close the cisparinaroyl-CoA approaches HNF-4␣LBD. Stryer's equation under "Experimental Procedures" allows calculation of energy transfer efficiency when both FRET and protein conformational changes upon ligand binding contribute to the sensitized emission of the bound ligand (34). For this case a much lower energy transfer efficiency, i.e. 12% was found and the resultant intermolecular distance between HNF-4␣LBD and cis-parinaroyl-CoA, R 2/3 , was 42 Å (Table III). Thus, depending on whether fatty acyl-CoA binding elicits a conformational change in HNF-4␣LBD structure, the intermolecular distance between HNF-4␣LBD Trp and bound cis-parinaroyl-CoA was in the range of 23-42 Å. In either case, these data were consistent with fatty acyl-CoA binding to HNF-4␣LBD being due to close molecular interaction between these molecules. Whether 23 or 42 Å is the correct intermolecular distance was resolved by circular dichroism (see below).
Determination of the Binding Parameters of HNF-4␣ LBD for cis-Parinaroyl-CoA from FRET Data-The appearance of sensitized cis-parinaroyl-CoA fluorescence emission upon FRET  (Table II). Upon selective excitation of HNF-4␣LBD Trp at 295 nm (Fig. 5C), the binding curve based on cis-parinaroyl-CoA sensitized emission again demonstrated binding to saturation (Fig. 5D). A double reciprocal plot of the binding data was linear again, consistent with one binding site or two binding sites of similar affinity (Fig. 5D, inset). A Hill plot yielded n ϭ 1.02, consistent with a single binding site, and a K d of 758.9 nM (Table II).
In summary, the binding curves constructed from the FRETinduced sensitized emission of HNF-4␣LBD bound cis-parinaroyl-CoA indicated that HNF-4␣LBD bound cis-parinaroyl-CoA to one site which, in general, showed a similar affinity for cis-parinaroyl-CoA as based on HNF-4␣LBD Trp emission quenching at 333 nm (Table II).
Binding Specificity of HNF-4␣LBD for cis-Parinaroyl-CoA: Interaction with cis-Parinaric Acid and CoA-The ligand binding specificity of HNF-4␣LBD was determined by titration with increasing amounts CoA or cis-parinaric acid. Upon HNF-   4␣LBD Trp/Tyr excitation at 280 nm, CoA slightly quenched HNF-4␣LBD fluorescence emission at 330 nm (Fig. 6A). The CoA concentration plot was hyperbolic, indicating saturation binding (Fig. 6B). The double reciprocal plot of this curve was linear (Fig. 6B, inset) and a Hill plot resolved one binding site (n ϭ 1.53) with K d of 429.1 nM (Table II). Similarly, excitation of HNF-4␣LBD Trp and determination of fluorescence resulted in quenching of fluorescence upon the presence of increasing concentrations of CoA (Fig. 6C). The binding curve constructed from these data was hyperbolic (Fig. 6D), indicating saturation binding. Both the double reciprocal plot (Fig. 6D, inset) and Hill plot (n ϭ 0.98) indicated a single binding site for CoA. The binding affinity of HNF-4␣LBD to CoA was low, i.e. K d of 680.4 nM (Table II). Thus, quenching of HNF-4␣LBD Trp (regardless of whether both Trp/Tyr or only Trp were excited) fluorescence suggested that (i) HNF-4␣LBD binds CoA at one binding site, and (ii) HNF-4␣LBD has 2-fold lower affinity for CoA at that site as compared with cis-parinaroyl-CoA.
Binding of free cis-parinaric acid to HNF-4␣LBD was assayed similarly as for CoA binding above, except that increasing concentrations of cis-parinaric acid were used. When HNF-4␣LBD Trp/Tyr were excited at 280 nm, increasing cisparinaric acid also quenched HNF-4␣LBD Trp emission (Fig.  7A). However, this binding of cis-parinaric acid did not result in FRET as indicated by the absence of sensitized emission at 430 nm (Fig. 7A, inset). The latter was in contrast to titration with cis-parinaroyl-CoA binding which induced sensitized emission at 430 nm (Fig. 5A). The plot of cis-parinaric acid concentration versus HNF-4␣LBD Trp fluorescence quenching indicated weak binding (Fig. 7B). The double reciprocal plot of the binding curve showed a single binding site (Fig. 7B, inset); this was confirmed by Hill plot which yielded n ϭ 1.09 and a K d of 421.3 nM (Table II). When only HNF-4␣LBD Trp was selectively excited at 295 nm, quenching of HNF-4␣LBD fluorescence was observed ( Fig. 7C) but, again, there was no appearance of cis-parinaric acid sensitized emission at 430 (Fig. 7C, inset). Construction of the cis-parinaric acid binding curve showed a hyperbolic shape (Fig. 7D). The double reciprocal plot again was linear, consistent with a single binding site (Fig. 7D, inset). This was confirmed by a Hill plot which yielded n ϭ 0.82 and K d of 588.9 nM (Table II). Thus, quenching of HNF-4␣LBD Trp (regardless of whether both Trp/Tyr or only Trp were excited) fluorescence suggested that (i) HNF-4␣LBD binds cis-parinaric acid at one binding site, and (ii) HNF-4␣LBD has about 2-fold lower affinity for cis-parinaric acid at that site then for cis-parinaroyl-CoA.
In summary, these data based on quenching of the HNF-4␣LBD aromatic amino acid fluorescence emission indicated that HNF-4␣LBD has a single binding site with K d near 250 nM for cis-parinaroyl-CoA. The individual residues, i.e. CoA and cis-parinaric acid, constituting cis-parinaroyl-CoA, interact with HNF-4␣LBD at lower affinities (i.e. K d values in the range of 420 -680 nM), thus indicating that both portions of the cisparinaroyl-CoA molecule apparently contributed to binding.

TABLE III Forster energy transfer from HNF-4␣LBD Trp to bound cis-parinaroyl-CoA
J is the overlap integral. R 0 is the critical distance that allows 50% energy transfer efficiency and was calculated as described under "Experimental Procedures." E is the energy transfer efficiency. R 2/3 is the actual distance between HNF-4␣LBD Trp and bound cis-parinaroyl-CoA.
In summary, the very high affinity (K d values of 1.7-4.4 nM) of HNF-4␣LBD for non-fluorescent, naturally occurring fatty acyl-CoAs showed that fatty acyl-CoA binding to HNF-4␣LBD was not just a unique property of cis-parinaroyl-CoA. Furthermore, the 2-order of magnitude higher K d values of HNF- 4␣LBD for cis-parinaroyl-CoA (K d s near 250 nM, depending on the type of assay, Table II) suggested that the presence of the conjugated tetraene double bonds in cis-parinaroyl-CoA reduced HNF-4␣LBD's affinity for 18-carbon, naturally occurring fatty acyl-CoA.
Binding Specificity of HNF-4␣LBD: Free Fatty Acids-To further assess the binding specificity of HNF-4␣LBD, the binding of two long-chain free fatty acids was examined: NBDstearic acid (a fluorescently labeled, synthetic 18-carbon saturated fatty acid) and arachidonic acid (a nonfluorescent, naturally occurring 20-carbon polyunsaturated fatty acid). In aqueous buffer, NBD-stearic acid fluoresced weakly with maximal emission near 562 nm (Fig. 9A, spectrum 1). However, when a constant amount of NBD-stearic acid was titrated with increasing concentrations of HNF-4␣LBD (Fig. 9A, spectra  2-8), three types of changes in the NBD-stearic acid fluorescence emission occurred. First, NBD-stearic acid fluorescence emission maximum exhibited a 32-nm blue shift from 562 nm in the absence of HNF-4␣LBD, to 530 nm at saturation with HNF-4␣LBD (Fig. 9A). When compared with a calibration curve of NBD-stearic acid in solvents with increasing dielectric constants (41), the maximal emission wavelength of NBD-stearic acid bound to HNF-4␣LBD corresponded to a dielectric constant near 2, indicating that the HNF-4␣LBD-bound NBDstearic acid was localized in a highly hydrophobic microenvironment. Second, excitation of HNF-4␣LBD protein at 280 nm resulted in a small, but visible increase in the NBD-stearic acid fluorescence emission at 530 nm, due to FRET between HNF-4␣LBD (donor) and NBD-stearate (acceptor) (Fig. 9B). Third, NBD-stearic acid exhibited a saturable increase in fluorescence emission intensity with increasing HNF-4␣LBD (Fig. 9A). Titration of a constant amount of HNF-4␣LBD with increasing concentrations of NBD-stearic acid followed by measurement of the increase in the NBD-stearic acid fluorescence emission intensity at 530 nm (upon excitation at 480 nm), resulted in a hyperbolic, saturable, binding curve (Fig. 9C). The linear plot of this curve demonstrated a single binding site (Fig. 9C, inset) and allowed calculation of a K d of 93 nM for NBD-stearic acid.
Since the affinity of HNF-4␣LBD for NBD-stearic acid (K d of 93 nM) was 5-7-fold higher than for the other fluorescent fatty acid examined, i.e. cis-parinaric acid (K d of 538 -642 nM, Table  II), it was important to determine whether this binding was dependent on the presence of the fluorophores (NBD, conjugated tetraene) present in these free fatty acids. Therefore, the affinity of HNF-4␣LBD for a non-fluorescent fatty acid, arachidonic acid, was determined. Increasing arachidonic acid also decreased the HNF-4␣LBD Trp emission at 333 nm (upon excitation at 280 nm) and resulted in a hyperbolic, saturable binding curve (Fig. 9D). The linear plot of this curve resolved a single binding site with weak affinity as evidenced by K d of 742 nM (Fig. 9D, inset).
In summary, unlike the very high affinity of HNF-4␣LBD for non-fluorescent fatty acyl-CoAs (K d values of 1-5 nM), HNF-4␣LBD only weakly bound non-fluorescent free fatty acid (K d of 742 nM). However, the presence of fluorophores in the free fatty acid increased this affinity by 5-7-fold. These data suggested that HNF-4␣LBD's ligand binding specificity, especially for molecules that were not fatty acyl-CoAs, might be highly dependent on the individual structure of the potential ligand examined.

HNF-4␣LBD Binds Peroxisome Proliferators as CoA-thioesters or Free Fatty
Acids-Because the structures of peroxisome proliferators such as Medica 16 and bezafibrate show some similarities to fatty acids (Table I), the interaction of their CoA-thioesters with HNF-4␣LBD and full-length HNF-4␣ was recently examined (12). However, a radioligand competition assay yielded only low affinities for these drugs, i.e. K d values in the micromolar range (12). Since results with direct fluorescence binding assays presented herein showed that the radioligand binding assay underestimated the affinities of HNF-4␣LBD for fatty acyl-CoAs by 3 orders of magnitude, HNF-4␣LBD binding parameters for Medica 16-CoA, bezafibroyl-CoA, as well as for the free acid forms of these peroxisome proliferators were examined by exciting HNF-4␣LBD Trp/Tyr at 280 nm and determining the effect on HNF-4␣LBD Trp emission at 333 nm.
Medica 16-CoA, unlike all the other ligands tested, induced an increase in fluorescence emission intensity of Tyr/Trp of HNF-4␣LBD. A titration curve of the Medica 16-CoA concentrations versus the Tyr/Trp fluorescence intensities fitted to a hyperbola, illustrating saturation binding (Fig. 10A). A reciprocal plot showed a straight line, consistent with a single binding site (Fig. 10A, inset), which yielded a K d of 2.6 nM (Table IV). Medica 16 in free acid form quenched HNF-4␣LBD Trp fluorescence emission intensity. Again, the titration curve for Medica 16 in free acid form fitted a saturation binding curve (Fig. 10B), which when linearized indicated a single binding site (Fig. 10B, inset) with K d value of 33.6 nM ( Table IV).
Titration of HNF-4␣LBD with bezafibroyl-CoA induced quenching in the intrinsic fluorescence of the protein, resulting in a saturation binding curve (Fig. 10C) which in a double reciprocal plot suggested a single binding site (Fig. 10C, inset) and yielded a K d of 29.3 nM (Table IV). For bezafibrate in free acid form, linear plot of the binding curve (Fig. 10D) yielded a slightly higher K d (57.1 nM) (Table IV).
In summary, HNF-4␣LBD has high affinity (K d values as low as 2.6 nM) for certain classes of peroxisome proliferator drugs, especially in their CoA-thioester form. The order of affinities of the peroxisome proliferator drugs tested was Medica-16-CoA Ͼ Medica-16 Ͼ bezafibroyl-CoA Ͼ bezafibrate.
Secondary Structure of HNF-4␣LBD and Effects of Fatty Acyl-CoAs: Circular Dichroism-One mechanism whereby ligands are thought to affect the function of nuclear binding proteins is to induce a conformational change in the protein to allow further events important for transcription to occur. The observation that non-fluorescent ligands induced significant alterations in HNF-4␣LBD Trp fluorescence emission (regardless of whether Trp/Tyr or Trp alone were excited) (Figs. 8 and 10) suggested that fatty acyl-CoAs altered the secondary structure and conformation of HNF-4␣LBD. This possibility was further tested using circular dichroism.
The far UV circular dichroic spectrum of HNF-4␣LBD exhibited two minima, at 208 and 220 nm and one maximum at 195 nm (Fig. 11, filled circles). Spectrum analysis using the CDsstr program as described under "Experimental Procedures" indicated that the polypeptide structure of HNF-4␣LBD was comprised of 18.6% total helix, 27.3% ␤-strands, 1.5% turns, and 38.7% other types of secondary structures (Table V). Palmitoyl-CoA, a 16-carbon chain length saturated fatty acyl-CoA, significantly changed the circular dichroic spectrum of HNF-4␣LBD such that the minima at 208 and 220 nm significantly decreased to lower molar ellipticity values (Fig. 11, open triangles). Accordingly, a higher total helix percentage of 22% was calculated for HNF-4␣LBD in the presence of palmitoyl-CoA (Table V). Interestingly, arachidonoyl-CoA, a 20-carbon chain length, polyunsaturated fatty acyl-CoA also altered the secondary structure of HNF-4␣LBD, but oppositely to palmitoyl-CoA. The molar ellipticity values of the minima at 208 and 220 nm exhibited by HNF-4␣LBD in the presence of arachidonoyl-CoA were higher (Fig. 11, open squares) than in the absence of the fatty acyl-CoA (Fig. 11, filled circles). This change was reflected in the helical and the ␤-strand structure percentages: nearly 3-fold lower percentages of helical structures, in total only 6.4% (Table V), but higher amount of turns (6.3%) and ␤-strands (35.7%) (Table V) in the presence of arachidonoyl-CoA. In summary, fatty acyl-CoA binding significantly altered the secondary structure of HNF-4␣LBD with the direc- tion of change highly dependent on the specific type of fatty acyl-CoA. DISCUSSION Since hepatic nuclear factor-4␣ (HNF-4␣) regulates expression of liver genes involved in both lipid and carbohydrate metabolism, investigations on its structure and mechanism of action impacts on our understanding of diabetes, obesity, and atherosclerosis. Despite this importance of HNF-4␣, almost nothing is known regarding either its structure, especially that of the ligand-binding domain (i.e. HNF-4␣LBD), or how structure relates to function. In fact, until recently HNF-4␣ was considered an orphan nuclear receptor. However, in 1998 Hertz et al. (15) demonstrated that various LCFA-CoA thioesters bound to recombinant rat HNF-4␣ and affected its transcriptional activity, depending on their chain length and degree of saturation. While this exciting discovery first suggested that fatty acyl-CoAs may be the natural ligand for this nuclear receptor, the radioligand competition assay used to demonstrate binding of fatty acyl-CoAs to HNF-4␣ and HNF-4␣LBD yielded low affinities, i.e. K d values in the micromolar range (15). Consequently, it has been suggested that this low affinity represents nonspecific binding and is insufficient for HNF-4␣ to significantly bind and be modulated by LCFC-CoAs in the nucleus (19). However, a significant potential problem with the radioligand competition assays is that they can seriously underestimate the affinities of binding proteins (20) or transcription factors (e.g. PPAR␣ (42)) for lipidic ligands. The purpose of the present investigation has been to resolve this issue of whether HNF-4␣LBD binds fatty acyl-CoAs with micromolar or nanomolar K d values, to determine the ligand specificity of HNF-4␣LBD, and examine if fatty acyl-CoA binding alters the structure/conformation of HNF-4␣LBD. As shown for the first time herein: (i) Forster fluorescence resonance energy transfer demonstrated close molecular interaction between fatty acyl-CoA and HNF-4␣LBD. (ii) A series of three independent, direct fluorescence binding assays showed that HNF-4␣LBD bound fatty acyl-CoAs with high affinity, K d values as low as 1 nM. (iii) Fatty acyl-CoA binding induced significant changes in HNF-4␣LBD conformation as evidenced by intrinsic fluorescence and circular dichroism. The significance of these findings is detailed as follows.
First, the synthesis of cis-parinaroyl-CoA as ligand with excitation overlapping the emission of HNF-4␣LBD allowed for the first time demonstration of their close molecular interaction and determination of the intermolecular distance between the fluorescent ligand and its binding domain. FRET showed that the average intermolecular distance between HNF-4␣LBD Trp and bound cis-parinaroyl-CoA was in the range of 23 Å (no ligand-induced conformational change in HNF-4␣LBD) to 42 Å (ligand-induced conformational change in HNF-4␣LBD). These are average intermolecular distances since HNF-4␣LBD contains 2 Trp residues (43). As further studies with non-fluorescent fatty acyl-CoAs and circular dichroism supported the conclusion that fatty acyl-CoAs induce conformational changes in HNF-4␣LBD, these data are consistent with the intermolecular distance between bound fatty acyl-CoA and HNF-4␣LBD Trp being Ͻ42 Å. Furthermore, the results demonstrate that fatty acyl-CoA binding with HNF-4␣LBD represent direct, molecular interaction rather than nonspecific coaggregation.
Second, three types of direct fluorescence binding assays showed that previous radioligand competition assays underestimated the binding affinities of HNF-4␣LBD for fatty acyl-CoAs. Instead, the direct fluorescence binding assays showed that HNF-4␣LBD bound fatty acyl-CoAs with high affinities as indicated by K d values as low as 1 nM. This underestimation is characteristic of lipidic radioligand competition binding assays which typically yield K d values that are 2-3 orders of magnitude higher (i.e. lower affinity) than K d values determined by direct fluorescence or titration microcalorimetry binding assays (reviewed in Ref. 20). In contrast, the direct fluorescence binding assays took advantage of three different intrinsic properties of ligand or protein including the fluorescence of cisparinaroyl-CoA, the intrinsic fluorescence of HNF-4␣LBD, or the FRET-sensitized emission of HNF-4␣LBD bound cis-parinaroyl-CoA. None of these direct fluorescence ligand binding assay required separation of bound from free ligand. Taken together, the present study demonstrates that (i) HNF-4␣LBD exhibits saturable fatty acyl-CoA binding, (ii) HNF-4␣LBD has one ligand binding site, (iii) the ligand-binding site of HNF-4␣LBD has K d values for LCFA-CoAs (e.g. palmitoyl-CoA, stearoyl-CoA, linoleoyl-CoA, and arachidonyl-CoA) in the very low nM range (1.6 -4 nM), and (iv) the respective free acids may bind, albeit with a significantly lower affinity (K d values for C20:4-CoA and C20:4 differ by about 200-fold).
Third, fatty acyl-CoA binding was shown to differentially alter the structure of HNF-4␣LBD. This was supported by two types of evidence. (i) A wide variety of fatty acyl-CoAs and peroxisome proliferator-CoAs significantly quenched HNF-4␣LBD Trp fluorescence emission at 333 nm. (ii) Two LCFA-CoAs, a saturated one, C16:0-CoA, and a polyunsaturated one, C20:4-CoA, were tested for their ability to induce changes in the secondary structure of HNF-4␣LBD. Although both ligands quenched HNF-4␣LBD Trp fluorescence emission, they differentially altered circular dichroic spectra of HNF-4␣LBD (Fig.  11). Very significant was the finding that palmitoyl-CoA increased the ␣-helix and turn content, while arachidonoyl-CoA decreased the content of 3 10 -helix and 3 1 -helix while increasing ␤-strands and turns. These very different changes induced by palmitoyl-versus arachidonoyl-CoAs in the secondary structure of HNF-4␣LBD correlated to the opposite effect of saturated and polyunsaturated fatty acyl-CoAs in functional assays (15).
Fourth, the proposed role of LCFA-CoA as natural ligands of HNF-4␣ (15) was challenged by Bogan et al. (19) on the grounds of (i) computer modeling of HNF-4␣LBD structure based on its 22 and 37% overall identity with the progesterone and RXR receptors LBD, respectively, and in particular its homology with helix 1 and the F domain of progesterone receptor LBD; this modeling predicted a ligand binding pocket of around 320 Å 3 as compared with a volume of 850 Å 3 apparently required for the CoA-thioester. (ii) Previously reported micromolar K d values for LCFA-CoA binding to HNF-4␣ as compared with nanomolar K d values for the more abundant cytosolic fatty acyl-CoA-binding proteins, such as ACBP (K d of 7 nM (40)), FABP (K d values of 41-60 nM (26)), and SCP-2 (K d of 4.5 nM (27,29)). (iii) Lack of induced conformational changes in HNF-4␣ as deduced from limited proteolysis (19). It is worth noting, however, that (i) circular dichroic data presented here indicates a higher percentage of ␤-strand (27.3%) than helical secondary structure (3.3% ␣-helix, 6.0% 3 10 -helix, and 9.3% 3 1 -helix) and is inconsistent with the predicted modeling. Furthermore, FRET results indicate direct interaction of LCFA-CoA with the ligand-binding site of HNF-4␣. (ii) While HNF-4␣ is localized in the nucleus (43,44), ACBP and L-FABP are essentially localized outside the nucleus (reviewed in Refs. 16 and 20) and SCP-2 is not detected in the nucleus (reviewed in Ref. 45). More important, the binding data presented here indicates 1-4 nM K d values for LCFA-CoA binding to HNF-4␣, being in the range of other acyl-CoA-binding proteins as well as in the range of unbound free fatty acyl-CoA concentrations estimated in the nucleus (18). It is possible, however, that these extranuclear fatty acyl-CoA-binding proteins may alter the distribution of fatty acyl-CoAs to the nucleus. Alternately, some L-FABP (29,46,47) and ACBP (18) detected within the nucleus may compete with HNF-4␣ for fatty acyl-CoA binding therein. (iii) This study reports direct evidence for ligand-induced conformational change verified by differential quenching of HNF-4␣ LBD Trp fluorescence emission as well as by differential changes in helical content due to ligand binding. This evidence is consistent with previously reported protection from limited proteolysis of HNF-4␣ by C14:0-CoA (12). Taken together, the results presented here support the original proposal by Hertz et al. (15) that LCFA-CoAs may serve as endogenous ligands for HNF-4␣.
This report further substantiates earlier findings by Hertz et al. (15) that CoA thioesters of hypolipidemic peroxisome proliferators such as Medica homologues or fibrate drugs bind to HNF-4␣ or HNF-4LBD and suppress its transcriptional activity. It further confirms that Medica 16-CoA is a more effective ligand than bezafibroyl-CoA (order of magnitude difference in K d values). Similarly to LCFA-CoAs, K d values for the CoAthioesters of hypolipidemic peroxisome proliferators measured by quenching of HNF-4␣ LBD Trp fluorescence emission were in the nanomolar range as compared with micromolar K d values previously reported using radioligand competition assays (12). However, in contrast to the free acid forms of LCFA which show low binding affinity to HNF-4␣ LBD (K d values of 421-742 nM), binding affinities for the free acid forms of hypolipidemic peroxisome proliferators were in the low nanomolar range (34 -57 nM), albeit significantly lower as compared with the respective CoA-thioesters. The capacity of hypolipidemic peroxisome proliferators bound in their free acid form to suppress the transcriptional activity of HNF-4␣ still remains to be investigated. It is worth noting, however, that HNF-4␣ transcriptional activity was not suppressed by amphipathic carboxylates, which due to a structural constraint did not endogenously yield the respective CoA-thioester.
Suppression of HNF-4␣ activity by the CoA-thioesters of hypolipidemic peroxisome proliferators in the nanomolar range may indicate that HNF-4␣ serves as target for their therapeu-  10 , and 3 1 helix. Instead, Total ␣-helix (HJ %) is calculated using a different standard protein structure database with the same program as described under "Experimental Procedures." b C16:0-CoA is palmitoyl-CoA. c C20:4-CoA is arachidonoyl-CoA. tic activity. This is consistent with HNF-4␣ involvement in controlling the expression of genes encoding proteins with a role in production and clearance of plasma lipoproteins (e.g. microsomal triglyceride transfer protein, apoB, apoC-III, and others (7)). Since the human liver is not responsive to PPAR␣ (reviewed in Refs. 48 and 49), the hypolipidemic activity induced by hypolipidemic PPAR␣ agonists in humans must be independent of PPAR␣. In contrast, in rodents, activation of PPAR␣ by hypolipidemic peroxisome proliferators may result in HNF4␣ displacement from direct repeat-1 response elements shared by both transcription factors (50). Suppression of HNF-4␣ function in transactivation by the CoA-thioesters of hypolipidemic peroxisome proliferators offers an hypolipidemic mode of action of peroxisome proliferators in humans independent of liver PPAR␣ and its proliferative carcinogenic activity.
In conclusion, multiple fluorescence spectroscopic and circular dichroism approaches strongly demonstrate that fatty acyl-CoAs and certain peroxisome proliferator-CoAs are specific ligands for HNF-4␣LBD and that fatty acyl-CoA binding induces significant changes in HNF-4␣LBD secondary and overall conformational structure. While naturally occurring fatty acids only weakly bind HNF-4␣LBD, some synthetic fluorescent fatty acids as well as certain peroxisome proliferators also bind as free acids to HNF-4␣LBD but with lower affinities than the respective acyl-CoAs. These data demonstrate for the first time that fatty acyl-CoAs and certain peroxisome proliferator drug thioesters may be natural and pharmacological HNF-4␣LBD ligands at physiological concentrations.