Physical and Functional Interaction of Acyl-CoA-binding Protein with Hepatocyte Nuclear Factor-4α*

Although acyl-CoA-binding protein (ACBP) has been detected in the nucleus, the physiological significance of this observation is unknown. As shown herein for the first time, ACBP in the nucleus physically and functionally interacted with hepatocyte nuclear factor-4α (HNF-4α), a nuclear binding protein that regulates transcription of genes involved in both lipid and glucose metabolism. Five lines of evidence showed that ACBP bound HNF-4α in vitro and in the nucleus of intact cells. (i) ACBP interaction with HNF-4α elicited significant changes in secondary structure. (ii) ACBP and HNF-4α were coimmunoprecipitated by antibodies to each protein. (iii) Double immunolabeling and laser scanning confocal microscopy (LSCM) of rat hepatoma cells and transfected COS-7 cells significantly colocalized ACBP and HNF-4α within the nucleus and in the perinuclear region close to the nuclear membrane. (iv) LSCM fluorescence resonance energy transfer determined an intermolecular distance of 53 Å between ACBP and HNF-4α in rat hepatoma cell nuclei. (v) Immunogold electron microscopy detected ACBP within 43 Å of HNF-4α. These interactions were specific since ACBP did not interact with Sp1 or glucocorticoid receptor in these assays. The functional significance of ACBP interaction with HNF-4α was evidenced by mammalian two-hybrid and transactivation assays. ACBP overexpression in COS-7 or rat hepatoma cells enhanced transactivation of an HNF-4α-dependent luciferase reporter plasmid by 3.2- and 1.6-fold, respectively. In contrast, cotransfection with antisense ACBP expression vector inhibited transactivation. LSCM of the individual triple fluorescent-labeled (HNF-4α, ACBP, and luciferase) rat hepatoma cells showed a high correlation (r2, 0.936) between the level of luciferase and the level of ACBP expression. In summary, ACBP physically interacted with HNF-4α in vitro and in intact cells, although ACBP expression level directly correlated with HNF-4α-mediated transactivation in individual cells.

Although ACBP and HNF-4␣ both bind LCFA-CoAs with nanomolar affinities and have been reported to modulate lipid metabolism, it is not known if (i) ACBP interacts directly with HNF-4␣ at the molecular level, (ii) ACBP interacts directly with HNF-4␣ at the cellular level, and (iii) ACBP expression influences HNF-4␣ transcriptional functions.

EXPERIMENTAL PROCEDURES
Materials-Recombinant mouse ACBP was produced and purified as described (6). HNF-␣-LBD was obtained as described earlier (16,17). Fetal bovine serum, bovine serum albumin, protein A-Sepharose 4CL, FITC-goat anti-rat IgG, and protease inhibitor mixture were purchased from Sigma. Lab-Tek coverglass slides were purchased from Fisher. Mouse anti-firefly luciferase monoclonal antibody was from Novus Biologicals (Littleton, CO). Rabbit polyclonal antibodies against glucocorticoid receptor ␣ (GR) and Sp1 protein were purchased from Santa Cruz Biotechnology (Santa Cruz, CA). Gold-labeled polyclonal antibodies against rabbit IgG (Aurion, Wageningen, The Netherlands) were purchased through Electron Microscopy Sciences (Fort Washington, PA). Texas Red goat anti-rabbit IgG, TOTO-3 (a DNA staining dimeric cyanine fluorophore), and SlowFade kit were from Molecular Probes (Eugene, OR). FluoroLink Cy5-labeled goat anti-mouse IgG was from Amersham Biosciences. The mammalian expression vector pCi-neo, Renilla luciferase expression plasmid pRL-CMV, dual luciferase reporter assay system, and CheckMate mammalian two-hybrid system were from Promega (Madison, WI). LipofectAMINE 2000, used for DNA transfections, was purchased from Invitrogen. All reagents and solvents used were of the highest grade available and were cell culture tested as necessary.
CD of ACBP and HNF-4␣-LBD-Far-UV circular dichroic spectra of HNF-4␣-LBD and ACBP in 2 mM Tri-HCl, pH 8, containing 0.5% glycerol and 0.05 mM dithiothreitol were measured separately and in mixture. 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, ACBP, and (HNF-4␣-LBD ϩ ACBP) were calculated from CD spectra by using the CDstr program (26,27). The CD spectrum of (ACBP ϩ HNF-4␣-LBD) was compared with a theoretical spectrum obtained by summing the spectra of the HNF-4␣-LBD and ACBP recorded separately for each protein in a concentration equal to that in the mixture. Because of the molecular weight difference between ACBP (10,000 Da) and HNF-4␣-LBD (amino acids 132-455; 36,172 Da), 4 M ACBP and 1.11 M HNF-4␣-LBD solutions were used, respectively, such that the molar concentrations of amino acids were equivalent. For mixtures of the two proteins, 2 M ACBP and 0.56 M HNF-4␣-LBD were used in order to (i) maintain equivalent concentrations of amino acids from the two proteins and (ii) maintain the same total concentration of amino acids as was used for CD of the individual proteins.
Coimmunoprecipitation-The ability of polyclonal anti-ACBP or polyclonal anti-HNF-4␣-LBD antiserum to coimmunoprecipitate both proteins from mouse liver homogenate and rat hepatoma cells was tested. Mouse liver was diced in RIPA buffer containing protease inhibitor mixture (1 ml of buffer per 200 mg of liver tissue) and subjected to 30 strokes within a Dounce homogenizer. RIPA buffer consisted of 50 mM Tris, pH 8, 150 mM NaCl, 1% Nonidet P-40, 0.5% deoxycholate, 0.1% SDS (31). Unbroken cells and debris were removed by centrifugation at 600 ϫ g for 5 min. Homogenates of mouse liver and hepatoma cells were incubated with antibodies against rat ACBP, rat-HNF-4␣-LBD, mouse GR, and mouse Sp1 protein, respectively, at 4°C overnight. Protein A-Sepharose 4CL was blocked for nonspecific binding with 2% bovine serum albumin in RIPA buffer, added to the overnight immunoprecipitated complexes, and incubated for 2 h at 4°C. After extensive washings with 500 mM NaCl in RIPA buffer, the beads were solubilized in 2ϫ sample buffer for further SDS-PAGE separation and Western blotting. The immunoprecipitates were tested for the presence of ACBP, HNF-4␣, Sp1 protein, and GR by Western blotting as follows.
Western Blotting-Protein concentration was determined by BCA Protein Assay (Pierce). SDS-PAGE and protein transfer on nitrocellulose membranes were performed as described (17,32). Primary antibodies used in Western blotting were rabbit anti-rat ACBP and antirat-HNF-4␣-LBD polyclonal antibodies prepared as described above. Specific proteins were visualized either by a colorimetric method utilizing alkaline phosphatase-conjugated goat anti-rabbit IgG and 5-bromo-4-chloro-3-indolyl phosphate/nitro blue tetrazolium substrate (Sigma) or by a chemiluminescent method employing horseradish peroxidase-conjugated goat anti-rabbit IgG and luminol/hydrogen peroxide substrates (Amersham Biosciences).
Detection of ACBP⅐HNF-4␣ Complexes by Immunoelectron Microscopy in Mouse Liver-Four-month-old mice were anesthetized with tribromoethanol and perfused through the left heart ventricle with 3% formaldehyde, 0.05% glutaraldehyde in 0.1 M sodium phosphate buffer, pH 7.3, with 3.2% sucrose. After 5 min of perfusion at 24°C, the liver was collected, minced, and immersed in the same fixative for 1 h at 4°C and then washed with 0.1 M sodium phosphate buffer containing 3.5% sucrose. The tissue was treated with 0.25% tannic acid for 1 h at 4°C and with 2% uranyl acetate in 70% ethanol cooled at Ϫ35°C for 40 min, and then dehydrated in an ethanol series and embedded in Lowicryl K4M resin. Ultrathin tissue sections (50 -60 nm) were placed on Formvar-coated nickel grids and immunogold stained with anti-ACBP antiserum raised in rat (diluted 1:50 -1:80) alone or in mixture with either anti-HNF-4␣ or anti-glucocorticoid receptor, raised in rabbit (diluted 1:150). These sections were washed and incubated with a mixture of goat anti-rat conjugated to 6-nm gold particles and goat anti-rabbit conjugated to 15-nm gold particles. All sections were post-stained very briefly with aqueous uranyl acetate and Reynold's lead citrate and examined with a Zeiss 10c transmission electron microscope.
Cell Culture-COS-7 cells were obtained from the American Type Culture Collection (ATCC, Manassas, VA). Rat hepatoma cells T-7 were generously provided by Dr. Charles Baum (University of Chicago, Chicago). The cells were grown in high glucose/Dulbecco's modified Eagle's culture medium containing 10% fetal bovine serum. 4-Well LabTek chamber slides were used for the immunofluorescence confocal microscopy experiments, and 6-well plates and 10-cm cell culture dishes for all the other purposes (e.g. transfection experiments).
Preparation of Cells for Indirect Immunofluorescence Microscopy-COS-7 and rat hepatoma cells grown in 4-well LabTek chamber slides were washed with Hanks' solution, fixed with 70:30 acetone/ethanol (v/v) for 30 min at Ϫ20°C, and blocked against nonspecific binding with 10% fetal bovine serum in Hanks' solution for 1 h at room temperature. The cells were then incubated with primary antibodies for 1 h at room temperature, followed by incubation with fluorescently labeled secondary antibodies for another 1 h at room temperature. After extensive washing of unbound antibodies with Hanks' solution, the cells were mounted by the use of Slow Fade kit. Rabbit anti-rat HNF-4␣ polyclonal, rat anti-mouse ACBP polyclonal, and mouse anti-firefly luciferase monoclonal antisera were used as primary antibodies. Texas Red goat anti-rabbit IgG, FITC goat anti-rat-IgG, and FluoroLink Cy5labeled goat anti-mouse IgG were used as secondary antibodies. In some experiments the nuclear areas of cells were defined by staining the DNA with TOTO-3, a dimeric cyanine fluorophore. Specificity of immunostaining was determined by deleting the primary antibody and using either one or both secondary antibodies. Several dilutions (1:20 -1:200) of each primary and secondary antibody were tested and optimized. This minimized nonspecific adsorption of fluorescent antibodies, ensured separation of the fluorescent signals, and optimized fluorophore concentration to preclude self-quenching (33).
LSCM and Image Analysis-LSCM was performed on an MRC-1024 fluorescence imaging system (Bio-Rad) consisting of an Axiovert 135 microscope (Zeiss, New York) equipped with three independent low noise photomultiplier tube channels. The excitation light ( ϭ 488, 568, and 647 nm) from a 15-milliwatt krypton-argon laser was delivered to the sample through ϫ63 Zeiss Plan-Fluor oil immersion objective, numerical aperture 1.45. Texas Red and Cy3 were excited at 568 nm, and their emission was detected through an HQ 585/40 bandpass filter. FITC was excited with 488 nm light, and its emission was selected through a 540/30 filter. TOTO-3 (DNA stain) and Cy5 were excited with 647 nm light and their emission selected through a 680/32 bandpass filter. The conventional colors of emitted light were red (for Texas Red and Cy3), green (for FITC), and blue (for Cy5 and TOTO-3). Images were acquired and analyzed using LaserSharp (from Bio-Rad) and MetaMorph Image Analysis (from Advanced Scientific Imaging, Meraux, LA) software. Colocalization analysis was performed on an average of 20 cells from four different wells, and the results obtained are shown for a representative set of cells. LaserSharp version 3.0 software was used for colocalization analysis, and two values (colocalization coefficients) were calculated based on Pearson's correlation coefficient, which is a well established means of describing the degree of overlap between patterns or images (34). The software calculated the colocalization (coloc) coefficients for two different fluorophores, designated red (R) and green (G) herein, according to the following equations: C red ϭ ⌺ R i,coloc /⌺ R i and C green ϭ ⌺ G i,coloc /⌺ G i . In these equations ⌺ R i,coloc represents the sum of intensities of all red pixels that also have a green component; ⌺ R i is the sum of all the red pixels in the image; ⌺ G i,coloc represents the sum of intensities of all green pixels that have a red component; ⌺ G i is the sum of intensities of all green pixels in the image. The coefficients generated by LaserSharp have values between 0 and 1; a value of 0 indicates that there is no colocalization, whereas a value of 1.0 means there is complete colocalization. Similarly, for three channel images (designated red, green, and blue, respectively) analyses were performed not only to compare red and green colocalization but also green with blue and red with blue. These three channel analyses yielded three different colocalization coefficients, one for each pair of color combinations. The software allowed for independent background selection and subtraction before calculation of the coefficients. Fluorograms were generated from the entire image, displaying the intensity and distribution of different colored pixels within the merged image as a scattergram.
FRET Microscopy-In order to estimate the intermolecular distance between ACBP and HNF-4␣ molecules that were detected colocalized in confocal microscopy images, FRET between Cy3-labeled anti-HNF-4␣ IgG and Cy5-labeled anti-ACBP IgG was determined as described earlier (29,30). Rat hepatoma cells were processed as for immunofluorescence microscopy, except that only primary antibodies that had been Cy3-and Cy5-labeled were used. Qualitatively, FRET from Cy3 to Cy5 was demonstrated by detection of sensitized emission of Cy5 (through the 680/32 bandpass filter) upon excitation of Cy3 at 568 nm, after correction of some bleed through of Cy3 emission through the same filter as follows. Fluorescence emission through 680/32 filter was recorded before and after photobleaching Cy5 at 647 nm for 3 min (the duration of photobleaching was optimized such that a good decrease in Cy5 fluorescence was obtained without affecting the emission intensity of Cy3). The image recorded after photobleaching Cy5 was the bleed through of Cy3 emission through PMT3 (with 680/32 bandpass filter). Fluorescence intensity of emission through this 680/32 filter before photobleaching Cy5 (FRET acceptor), being the sum resulted from sensitized emission of Cy5 and the bleed through of Cy5, was always higher than the one taken after Cy5 photobleaching, clearly indicating FRET from Cy3 to Cy5. Quantitative measurements for FRET efficiency estimation were carried out as described (29). FRET efficiency was calculated according to E ϭ (I post Ϫ I pre )/I post , where E is FRET efficiency; I post is the acceptor (Cy3) fluorescence intensity after Cy5 photobleach, and I pre is the acceptor (Cy3) fluorescence intensity before Cy5 photobleach. An average value of E was calculated from measurements of Cy3 fluorescence emission increase after Cy5 photobleaching in 50 different cells. The calculated E was further used to estimate R, the intermolecular distance between Cy3 and Cy5 (distance ultimately dictated in this experiment by the proximity between ACBP and HNF-4␣), according to E ϭ 1/(1 ϩ (R/R 0 ) 6 ), where R 0 , the distance for which a 50% FRET efficiency occurs, is known to be 50 Å for the Cy3/Cy5 FRET pair.
DNA Constructs and Plasmids Used in Transfection Assays-COS-7 and rat hepatoma cells were transfected with the following expression and reporter plasmids. A set of sense and antisense mouse ACBP expression vectors was constructed by inserting the 276-bp ACBP cDNA into an XhoI cloning site of pCI-neo mammalian expression vector purchased from Promega (Madison, WI). For an efficient repression of ACBP expression, an antisense ACBP vector was constructed using a pUC plasmid that contained the mouse pgk-1 gene promoter, as described (35). The HNF-4␣ expression vector pLEN4S, with the 3-kb rat HNF-4␣1 cDNA under a human metallothionein-II promoter and human growth hormone 3Ј-untranslated region, was a gift from Dr. F. Sladek (University of California, Riverside). The ApoBLuc reporter vector, containing 4 inserts of the 85-47 HNF-4␣-responsive elements in the apoB gene promoter in front of the firefly luciferase cDNA was obtained from the same laboratory. For monitoring the transfection efficiencies, pRL-CMV (Promega, Madison, WI), the expression vector for Renilla luciferase was cotransfected with the other expression vectors.
Functional Interaction of ACBP with HNF-4␣ in Intact Cells, Transactivation Assay-COS-7 and rat hepatoma cells grown in 6-well culture plates were transfected with 1 g each of ApoBLuc (reporter plasmid) and pLEN4S (HNF-4␣ expression vector), pCI-sACBP (sense ACBP expression vector), or pgk-aACBP (antisense ACBP RNA expression vector) per well. Renilla luciferase plasmid pRL-CMV (0.05 g per well) was always cotransfected as internal standard for estimating the transfection efficiency. Transfections were performed with Lipofectamine 2000 reagent from Invitrogen according to the manufacturer's instructions.
The firefly luciferase activity, normalized to Renilla luciferase (as internal control for transfection efficiency), was determined by using the dual-luciferase reporter assay system from Promega (Madison, WI). Luminescence was read and measured with a Microlite ML3000 Microtiter plate luminometer equipped with BioLinx assay management software (Dynatech Laboratories, Inc., Chantilly, VA).
Appropriate controls were included as follows. (i) Controls for luciferase activity in cells transfected with ApoBLuc alone or ApoBLuc in the presence of empty vectors pLEN4, pCI-neo, and pgk, separate and together, were run to ensure that luciferase activity was an accurate reporter for HNF-4␣-driven expression. (ii) In experiments assessing the effect of ACBP on HNF-4␣-mediated transactivation, luciferase activity levels in cells transfected with sense or antisense ACBP expression vectors (i.e. pCI-sACBP or pgk-aACBP) were compared with cells transfected with empty vectors pCI-neo and pgk as negative controls.
Functional Interaction of ACBP with HNF-4␣ in Intact Cells, Mammalian Two-hybrid Assay-The interaction of ACBP with HNF-4␣ in intact cells was also examined by a mammalian two-hybrid system CheckMate Mammalian Two-Hybrid System from Promega (Madison, WI). The rat recombinant HNF-4␣ cDNA was from Dr. J. Bar-Tana (Hebrew University Medical School, Jerusalem, Israel). The mouse recombinant ACBP cDNA was prepared as described (6). Full-length HNF-4␣ and ACBP cDNA fragments were amplified by PCR using primer sets containing specific restriction cloning sites as follows: (i) for ACBP, 29-mer 5Ј-CTGGATCCGTATGTCTCAGGCTGAATTTG-3Ј and 33-mer 5Ј-GGTCTAGATTATATTCCGTATTTCTTCTTTAGC-3Ј; (ii) for HNF-4␣, 30-mer 5Ј-CAGGATCCACATGGACATGGCTGACTACAG-3Ј and 28-mer 5Ј-GCTCTAGACTAGATGGCTTCCTGCTTGG-3Ј. PCR fragments were restriction-digested with BamHI/XbaI and ligated to pACT and pBIND vectors to produce pACT/HNF-4␣ and pBIND/ACBP, pACT/ACBP and pBIND/HNF-4␣. The four constructs were purified and sequenced at their 5Ј junctions to verify that the introduced cDNAs were in the correct reading frame. Two separate sets of experiments were performed to ensure there was no vector preference for either cDNA. COS-7 cells, were cotransfected with pACT/HNF-4␣, pBIND/ ACBP in a first set of experiments, and with pACT/ACBP, pBIND/ HNF-4␣ in a second set, together with the firefly luciferase reporter plasmid pG5luc, using LipofectAMINE 2000 (from Invitrogen). Renilla luciferase plasmid, pRL-CMV (Promega, Madison, WI), was used as internal control to monitor transfection efficiency at all times. Transfected COS-7 cells were cultured for 24 h, harvested, and lysed. Renilla and firefly luciferase activities were measured by using Dual Reporter Assay System (Promega, Madison, WI) and an ML3000 Microtiter Plate luminometer equipped with BioLinx assay management software (Dynatech Laboratories Inc., Chantilly, VA).

ACBP⅐HNF-4␣ Complex Formation in Vitro, Circular
Dichroism-To determine whether ACBP and HNF-4␣-LBD physically interact to alter conformation, CD was used to determine the secondary structure of these proteins obtained individually and in combination. The shape of the far-UV CD spectrum of ACBP indicated the presence of high amounts of ␣-helix ( Fig. 1, solid triangles). Quantitative analysis of multiple CD spectra of ACBP showed that ACBP contains nearly half of its polypeptide chain as ␣-helix (48%) structure, with very little ␤-sheet (2%) in its secondary structure. In contrast, CD spectra of HNF-4␣-LBD in buffered aqueous solution indicated that ␣-helix structures represented only a minor component of this protein (Fig. 1, solid circles). This was confirmed by quantitative analysis of multiple CD spectra that showed HNF-4␣-LBD as predominantly ␤-strand structure (27.3%) with low content of ␣-helices (3.3%) ( Table I).
The significant differences in CD spectra of pure ACBP versus pure HNF-4␣-LBD provided a means for detecting direct interaction between ACBP and HNF-4␣-LBD measured as conformational change. Theoretically, if ACBP did not interact with HNF-4␣-LBD, then a mixture of the two proteins should show a CD spectrum equally intermediate between pure ACBP and pure HNF-4␣-LBD. On the contrary, the experimental data showed that (ACBP ϩ HNF-4␣-LBD) exhibited a CD spectrum that was not equally intermediate between pure ACBP and pure HNF-4␣-LBD (Fig. 1, open circles). The CD spectrum of (ACBP ϩ HNF-4␣-LBD) was much closer to that of pure HNF-4␣-LBD, suggesting a significant conformational change. Quantitative analysis of multiple CD spectra of (ACBP ϩ HNF-4␣-LBD) showed that the amount of ␣-helix (14.7%) in the (ACBP ϩ HNF-4␣-LBD) mixture was much lower than predicted from the theoretical non-interactive CD spectrum (25.7%) ( Table I). In contrast, (ACBP ϩ HNF-4␣-LBD) exhibited much higher ␤-strand structures (21.3%) than predicted from the theoretical non-interactive CD spectrum (11.7%) ( Table I). These data were consistent with ACBP binding directly with HNF-4␣-LBD to form an ACBP⅐HNF-4␣-LBD complex in vitro. This interaction was detectable as a change in protein conformation resulting in altered protein secondary structure. In this experiment, the molecular weight difference between ACBP (10,000 Da) and HNF-4␣-LBD (36,172 Da) was taken into account and corrected for by using appropriate concentrations of each protein (see "Experimental Procedures"). This allowed direct comparison of changes in the overall CD spectrum of the protein mixture versus that estimated from the calculated sum of CD spectra of the individual proteins taken separately. These data showed that the CD spectral changes in the mixture were caused mainly by the formation of a complex with altered conformation, different from that of ACBP or HNF-4␣-LBD alone.
Detection of ACBP/HNF-4␣ Interaction in Rat Hepatoma Cells and Mouse Liver, Coimmunoprecipitation-To investi-gate further the possibility that ACBP binds HNF-4␣-LBD to form an ACBP⅐HNF-4␣ complex, the homogenate of mouse liver tissue was treated with polyclonal antisera to ACBP or HNF-4␣-LBD in a coimmunoprecipitation assay as described under "Experimental Procedures." Anti-ACBP serum immunoprecipitated a high amount of ACBP from mouse liver homogenate, as expected (Fig. 2, lane 1, row a) and also a lower but significant amount of HNF-4␣ (Fig. 2, lane 1, row b). The coimmunoprecipitation of ACBP and HNF-4␣ was also demonstrated with anti-HNF-4␣ serum which produced high amounts of HNF-4␣ (Fig. 2, lane 2, row b) but also detectable amounts of ACBP (Fig. 2, lane 2, row a) from the liver homogenate. As controls for the specificity of ACBP/HNF-4␣ coimmunoprecipitation, polyclonal antibodies against two other nuclear proteins with roles in transcription regulation, i.e. Sp1 protein and glucocorticoid receptor (a nuclear receptor with modular structure, similar to HNF-4␣), were used to test coimmunoprecipitation with ACBP (Fig. 2). Sp1 protein was not detected in immunoprecipitates generated with anti-ACBP (Fig. 2, lane 1, row c), even though it was found present in liver homogenate Although a trace amount of ACBP was detected after anti-GR immunoprecipitation ( Fig. 2 lane 4, row a), the fact that anti-ACBP did not immunoprecipitate GR suggested that this was nonspecific. These results were consistent with ACBP directly and specifically interacted with HNF-4␣ in mouse liver as well as in hepatoma cells in culture (data not shown for hepatoma cells).
Detection of ACBP⅐HNF-4␣ Complexes in Mouse Liver Cells, Immunoelectron Microscopy-In mouse liver ultrathin sections, ACBP (immunolabeled with 6-nm gold particles) and HNF-4␣ (immunolabeled with 15-nm gold particles) were detected, often located in close proximity, in the peripheral nucleoplasm, as indicated by the arrowhead in Fig. 3A. The inset in the same panel exhibits a ϫ4 magnification of the area illustrated by the arrow demonstrating an average distance of 43 Å (limit resolution) between the 6-and 15-nm gold particles, i.e. between ACBP and HNF-4␣. In order to estimate the specificity of ACBP/HNF-4␣ association, the intermolecular distances between ACBP and GR, a nuclear receptor superfamily member like HNF-4␣, were determined similarly in liver ultrathin sections. Immunogold labeling of ACBP and GR (Fig. 3B) demonstrated that the 6-and 15-nm gold particles were totally separated or at distances Ͼ400 Å from each other, indicating no association between ACBP and GR. These data in situ taken together with the coimmunoprecipitation from cell homogenates (preceding section) strongly suggest that ACBP specifically interacts with HNF-4␣.
Colocalization of ACBP and HNF-4␣ within Nuclei of Fixed Rat Hepatoma Cells-To determine whether ACBP and HNF-4␣-LBD interacted not only in vitro but could potentially interact in intact cells, rat hepatoma cells expressing both proteins were double immunolabeled and examined by LSCM. Rat hepatoma cells were fixed, coimmunolabeled with rabbit anti-HNF-4␣ and rat anti-ACBP primary antisera, followed by treatment with Texas Red goat anti-rabbit (to detect HNF-4␣) and FITC goat anti-rat (to detect ACBP) IgG secondary antibodies, as described under "Experimental Procedures." HNF-4␣ was distributed strongly throughout rat hepatoma nuclei with more intense staining near the nuclear envelope (Fig. 4A). Outside nuclei, HNF-4␣ was only weakly detected diffusely (Fig. 4A). In contrast, ACBP was stained most intensely in the perinuclear region and throughout the cytoplasm ( Fig. 4B). Smaller amounts of ACBP were also detected as more diffuse punctate regions within the nucleus (Fig. 4B). Superposition of simultaneously acquired red and green fluorescence images indicated the regions where the two labels were most colocalized, the yellow pixels (Fig. 4C). The yellow pixels with the highest degree of colocalization (i.e. the pixels with high fluorescence intensity of both red and green fluorophores) are also shown as a separate panel (Fig. 4D). This indicated that both ACBP and HNF-4␣ were strongly colocalized in the perinuclear region and as punctate structures within the nuclei of rat hepatoma cells (Fig. 4D). Although the limit of resolution of LSCM is about 0.22 m, the significant colocalization of the two proteins is consistent with the possibility that ACBP and HNF-4␣ are distributed to potentially interact physically in intact cells.
Determination of ACBP/HNF-4␣ Intermolecular Distance in Fixed Cells, FRET Microscopy-To increase the resolution of LSCM, the intermolecular distance between ACBP and HNF-4␣ within hepatoma cell nuclei was determined by FRET. FRET efficiency between Cy3-labeled anti-HNF-4␣ (Fig. 5A) and Cy5-labeled anti-ACBP (Fig. 5B) was measured as described under "Experimental Procedures." Detection of Cy5 (FRET acceptor)-sensitized emission through a 680/32 bandpass filter upon excitation of Cy3 (FRET donor) at 568 nm was the first indication of FRET transfer (Fig. 5, C and H). To determine the energy transfer efficiency, the Cy5 (FRET acceptor) was photobleached by excitation at 647 nm for 3 min until no fluorescence could be detected through the 680/32 filter with direct excitation of Cy5 at 647 nm (Fig. 5E). The Cy3 (FRET donor) fluorescence emission was then detected through a HQ598/40 bandpass filter with excitation at 568 nm (Fig. 5, D and I). As indicated by the larger amount of red/yellow pixels in Fig. 4I, Cy3 (FRET donor) fluorescence emission at 568 nm after photobleaching at 647 nm (Fig. 5C) was significantly greater than fluorescence emission of Cy3 (FRET donor) at 568 nm before photobleaching (Fig. 5G). Quantitative analyses of these intensities as described under "Experimental Procedures" allowed calculation of an intermolecular distance of 53 Å between ACBP and HNF-4␣. This indicated a close molecular association of ACBP and HNF-4␣, especially within the peripheral region of hepatoma cell nuclei.
ACBP/HNF-4␣ Interaction in Living Cells, Mammalian Twohybrid Assay-To examine whether ACBP interacted with  HNF-4␣ in living cells, a mammalian two-hybrid assay was performed. ACBP and HNF-4␣ cDNAs were ligated to pACT (DNA binding component) and pBIND (transcription complex component) and assayed for transactivation. In both combination sets, i.e. when either ACBP was on the DNA binding component and HNF-4␣ on the transcription side (experiment A) or vice versa (experiment B), the transactivation was higher than negative controls (Fig. 6, groups 2 and 3) and not as great as the positive control (Fig. 6, group 4). These data suggested that there was a low but significant level of direct interaction between ACBP and HNF-4␣ in the nuclei of living cells.
Expression of ACBP and HNF-4␣ in COS-7 Cells and Rat Hepatoma-To begin to establish the functional significance of ACBP interaction with HNF-4␣, it was necessary to utilize transfected cells in order to vary the content of ACBP and HNF-4␣. Two types of cells (COS-7 and rat hepatoma) were used. COS-7 cells do not express a detectable amount of HNF-4␣ as indicated by Western blotting (Fig. 7A, lane 1; HNF-4␣ band at 55 kDa was not detected in COS-7 cell homogenate). Only low levels of ACBP were found in COS-7 cells, as detected by a band at 10 kDa in Western blots of COS-7 cell homogenate (Fig. 7A, lane 1). Comparison with standards re-vealed that 0.08% of protein in COS-7 cells was ACBP. The expression level of ACBP was increased 11-fold in COS-7 cells transfected with sense ACBP expression vector, pCI-sACBP (Fig. 7A, lane 2). Conversely, ACBP expression level was decreased up to 90% by transfection of COS-7 cells with antisense ACBP vector, pgk-aACBP (Fig. 7A, lane 5). Both ACBP overexpression and antisense-induced underexpression were maintained in COS-7 cells cotransfected with HNF-4␣ expression vector, pLEN4s (Fig. 7A, lanes 3-5). Comparison of ACBP levels in cells cotransfected with HNF-4␣ expression vector (pLEN4s) revealed that HNF-4␣ was highly expressed in cells cotransfected with the antisense ACBP vector (Fig. 7A, lane 5), but only at a lower level in cells cotransfected with sense ACBP expression vector (Fig. 7A, lane 4). This may indicate that the expression of ACBP in COS-7 cells has a negative effect on HNF-4␣ expression level.
Western blots of ACBP and HNF-4␣ in rat hepatoma cells showed a high level of these proteins in nontransfected cells (0.6% ACBP, 0.5% HNF-4␣; Fig. 7B, lane 1). Transfection of rat hepatoma cells with the sense ACBP expression vector increased the ACBP protein expression by 2-fold, although the transfection of antisense ACBP vector, pgk-aACBP, did not  significantly decrease of ACBP expression (Fig. 7B, lanes 2 and  3). This different effect of an expression vector in hepatoma cells as compared with COS-7 cells is not uncommon (36 -38) and might be explained by different pathways of ACBP transcription regulation, involving different coactivators or corepressors specific for the two types of cells. Another possible explanation is that the transfection efficiency is very different for the two types of cells, i.e. hepatoma cells are more difficult to transfect than COS-7 cells.
These Western blotting data were confirmed by indirect immunofluorescence microscopy. COS-7 cells either nontransfected (Fig. 8A) or transfected with HNF-4␣ expression vector, pLEN4s (Fig. 8B), sense ACBP expression vector, pCI-sACBP (Fig. 8, D-F), and antisense ACBP vector, pgk-aACBP (Fig.  8C), were fixed and labeled with Texas Red (red fluorescence) for HNF-4␣ and with FITC (green fluorescence) for ACBP. The ACBP level was detectable in nontransfected COS-7 cells (Fig.  8A), increased in numerous cells (80%) upon transfection with pCI-sACBP alone (Fig. 8D) or together with HNF-4␣ expression vector (Fig. 8, E and F), and decreased in most of the cells (90%) cotransfected with antisense ACBP vector, pgk-aACBP (Fig. 8C). HNF-4␣ expression in COS-7 cells assessed by confocal microscopy demonstrated that nontransfected cells did not exhibit any HNF-4␣ (Fig. 8A), although pLEN4S (HNF-4␣) vector-transfected COS-7 cells expressed HNF-4␣ in 20% of cells (Fig. 8, B, C, E, and F) in contrast to the ectopic expression of ACBP, in 80% of the cells. This difference in expression efficiency was apparently determined not only by the transfection efficiency but also by the promoter strength. When cotransfected with HNF-4␣ and sense ACBP expression vectors, some cells expressed both proteins (Fig. 8E) and some of them expressed only one of them (Fig. 8, E and F). Thus, the overexpression of ACBP in COS-7 as determined by Western blotting represents the total amount of ACBP expressed in transfected cells but not necessarily ACBP simultaneously overexpressed with HNF-4␣ in the same cells.

Intracellular Distribution of ACBP and HNF-4␣ in Transfected COS-7 and Rat Hepatoma Cells Overexpressing ACBP
and HNF-4␣-In order to determine the intracellular localization of the two proteins (ACBP and HNF-4␣) ectopically expressed, COS-7 cells cotransfected with HNF-4␣ and ACBP expression vectors were processed for LSCM as described under "Experimental Procedures." Triple fluorescent labeling was performed by the use of Texas Red (for HNF-4␣), FITC (for ACBP), and TOTO-3 (DNA stain). Three fluorescence images (each dye through a separate photomultiplier) were simultaneously acquired and analyzed for colocalization. In COS-7 cells expressing high levels of both ACBP and HNF-4␣, ACBP was spread throughout the cell, although HNF-4␣ was localized mostly within the nucleus (Fig. 9, A-C). Quantitative analysis of colocalization analysis showed 61% of ACBP inside the nucleus (overlapping with the DNA stain TOTO-3) and 96% of HNF-4␣ within the nucleus (Fig. 9, D and E). ACBP and HNF-4␣ were significantly codistributed since 57% of ACBP and 60% of HNF-4␣ within a single cell colocalized (Fig. 9F). When the pixels with highest green and red fluorescence intensities were selected (the boxed area in Fig. 9F), the region where ACBP and HNF-4␣ colocalized the most was observed (Fig. 9I). This region was located inside the nucleus, at the peripheral zone of the nucleus. The ACBP pixels that colocalized the most with DNA (i.e. TOTO-3) pixels were also located at the periphery within the nucleus (Fig. 9G).
Hepatoma cells transfected with ACBP expression vector were also analyzed for colocalization of the three labels (conventional fluorescent colors: red for HNF-4␣, green for ACBP, and blue for DNA) (Fig. 10, A-C). Interestingly, a higher overlapping of ACBP (80%) and DNA (67%) was determined (Fig. 10, D and G) than in COS-7 cells. In rat hepatoma cells, most of the HNF-4␣ (94%) was colocalized with the DNA stain TOTO-3 (93%) as expected (Fig. 10, E and H). A higher colocalization was also found for ACBP (94%) and HNF-4␣ (93%) in rat hepatoma cells than in COS-7 cells (Fig. 10, F and I). The most colocalized ACBP/HNF-4␣ (green/red generating yellow) pixels were located within the nucleus, at the peripheral zone of the nuclei (Fig. 10H). A few distinct colocalized ACBP and HNF-4␣ pixels were also found within the central zone of the nucleus as well as outside nuclei, within the cytoplasm (Fig.  10H).
Functional Significance of ACBP Expression on HNF-4␣ Transcriptional Activity, Transactivation Assays-From molecular studies in vitro, in fixed cells, and in intact cells, it was evident that ACBP and HNF-4␣ can form a complex by direct physical association. In cultured cells frequent ACBP⅐HNF-4␣ complexes were detected inside the nuclei, at the peripheral zone, close to the nuclear membrane. These findings suggested that this association between ACBP and HNF-4␣ might be functionally significant. To test this hypothesis, the influence of ACBP overexpression and underexpression upon HNF-4␣mediated transactivation of a reporter vector, consisting of firefly luciferase gene under HNF-4␣-response elements from apoB gene promoter, was examined. Since COS-7 cells express no HNF-4␣, they were cotransfected with pLEN4S (HNF-4␣ expression vector), in addition to pCI-sACBP (ACBP expression vector) and ApoBLuc (reporter plasmid) as described under "Experimental Procedures." COS-7 cells overexpressing ACBP exhibited a 3.2-fold increase in transcription of luciferase under HNF-4␣-response element apoB promoter compared with cells transfected with pCI empty vector (Fig. 11A, bars 6 and 7) in the presence of an equal amount of HNF-4␣ expressing vector. COS-7 cells transfected with empty pLEN, empty pCI, or pCI-ACBP in the absence of HNF-4␣ did not show luciferase expression (Fig. 11A, bars 1 and 3-5), confirming the conclusion that only ACBP overexpression was responsible for the enhancement of luciferase transactivation in the presence of a constant amount of HNF-4␣.
Rat hepatoma, a cell line that in contrast to COS-7 cells expresses naturally significant amounts of HNF-4␣ and ACBP, was tested in transactivation assays. Transfection of rat hepatoma cells to overexpress ACBP about 2-fold resulted in a low but significant 1.6-fold increase in luciferase transcription (Fig.  11A, bars 8 -10). Even though comparing the two cell lines may have little relevance due to many different factors that could contribute to the difference (such as larger diversity and amounts of coactivators and other transcription factors in hepatoma than in COS-7 cells), it was interesting to note that in hepatoma cells the level of HNF-4␣-luciferase transactivation was 10.4-fold higher (Fig. 11A, bar 8) compared with COS-7 cells expressing HNF-4␣ but not ACBP (Fig. 11A, bars 2 and 6), and only 3.3-fold higher than in COS-7 cells expressing both HNF-4␣ and ACBP proteins (Fig. 11A, bars 7 and 10). Regardless, overexpression of ACBP in either cell type resulted in HNF-4␣ transactivation. A dose response of the ACBP expression versus luciferase activity was determined in COS-7 cells cotransfected with various amount of ACBP sense or antisense vector and constant amounts of HNF-4␣ expression plasmid and ApoBLuc reporter vector (Fig. 11B). Luciferase transactivation in COS-7 cells was directly proportional to transfection with increasing amounts of pCI-sACBP, the overexpression vector for ACBP (Fig. 11B, sense ACBP). Conversely, luciferase transactivation in COS-7 cells was inversely proportional to transfection with increasing amounts of antisense ACBP vector, pgk-aACBP (Fig. 11B, antisense-ACBP). Thus, ACBP stimulated HNF-4␣-mediated transactivation proportional to the level of ACBP expression in the COS-7 and rat hepatoma cells. The expression of either sense or antisense mRNAs for ACBP and HNF-4␣ proteins was strongly dependent on the cell type. Thus, pgk-aACBP promoter was a good antisense expression vector in COS-7 but not in rat hepatoma cells; in contrast, pLEN4S-HNF-4␣ was a good sense expression vector in COS-7 but not in hepatoma cells (data not shown). This differential efficiency of the same expression vector in different cell lines has been reported previously (36 -38) for numerous promoters, in addition to the difference in transfection efficiency between COS-7 and hepatoma cells.
Functional Significance of ACBP Expression on HNF-4␣ Transcriptional Activity, LSCM Imaging of Luciferase Transactivation in Individual Cells-To assess more accurately the effect of ACBP expression on HNF-4␣-mediated luciferase transcription, a new technical approach was used. Transfected rat hepatoma cells were fixed and triple fluorescent-labeled to simultaneously detect HNF-4␣, ACBP, and luciferase. Thus, cells transfected with pCI-sACBP (ACBP overexpression vector) and ApoBLuc (luciferase reporter vector) were labeled with Texas Red for HNF-4␣, FITC for ACBP, and Cy5 for luciferase (Fig. 12, A-C). By image analysis, many individual cells were analyzed for their content in HNF-4␣, ACBP, and luciferase and then the correlation curves for the three protein expression levels were studied (Fig. 12, D and E). A plot of ACBP versus luciferase expression (Fig. 12D) revealed a very high correlation coefficient (r 2 , 0.936) suggesting that cells with higher amounts of ACBP exhibited higher levels of luciferase (and implicitly had a higher HNF-4␣-mediated transactivation). As both luciferase versus HNF-4␣ and ACBP versus HNF-4␣ showed very good correlation (r 2 , 0.899 and 0.856, respectively; plots not shown), luciferase and ACBP were normalized to HNF-4␣, and then the degree of correlation was again determined. Interestingly, for ratios of ACBP/HNF-4␣ lower than 0.7, a good correlation coefficient was found, i.e. r 2 of 0.853 (Fig.  12E). For ratios of ACBP/HNF-4␣ higher than 0.7, the correla- tion did not hold. As both ACBP and HNF-4␣ had been reported to bind fatty acyl-CoAs with very high affinities (in the low nanomolar range) (1,17), the ratio between the two proteins may be a very important factor in the way ACBP influences HNF-4␣ activity. DISCUSSION ACBP was previously thought to be a cytoplasmic protein involved in the metabolism of LCFA-CoA as well as influencing LCFA-CoA-mediated signaling pathways (reviewed in Refs. 1  and 12). However, the recent discovery that ACBP is also localized to the nuclei of CV-1 and 3T3-L1 cells suggested additional potential gene-regulatory roles (15). Two mechanisms may be suggested.
Alternatively, ACBP may interact directly with nuclear transcription factors to influence transcriptional activity. Support for this possibility comes from the observation that ACBP expression significantly decreases PPAR-␥-mediated transactivation (15). Since LCFA-CoAs antagonize the effects of peroxisome proliferators on PPAR-␣ but not -␥ or -␦ (20), it is unlikely that the effects of ACBP on PPAR-␥ are mediated through the ligand LCFA-CoAs. It has been shown that L-FABP interacts with PPAR-␣ and -␥ and stimulates fatty acid and hypolipidemic drug activation of PPAR-dependent transcription (38). Interestingly, Hertz et al. (39) demonstrated that overexpression of acyl-CoA synthase decreased the capacity of amphipathic carboxylic peroxisomal proliferators to induce PPARdependent transactivation, but it increased the effect of long chain fatty acids on HNF-4␣-dependent transactivation (16). This may suggest that free fatty acids interact readily with L-FABP⅐PPAR complexes, although acyl-CoAs affect primarily ACBP⅐HNF-4␣ complexes. The present work showed for the first time that there is a direct, physical, and functional interaction between ACBP and HNF-4␣.
First, in vitro studies by circular dichroism demonstrated that recombinant ACBP and HNF-4␣-ligand binding domain when mixed together interact to form a complex with altered conformation (Fig. 1). Altered conformation of nuclear receptors can modulate cofactor recruitment and thereby influence transcriptional activity (21). Circular dichroism has been used previously to detect profound changes in secondary structures of two proteins upon interaction to form a complex (40,41). Thus, for example, by comparing the theoretical sum of individual far-UV circular dichroic spectra of estrogen receptor-␣ and TATA box-binding protein to the actual spectrum of an equimolecular mixture of the two proteins, significant differences were obtained indicating that interaction between estrogen receptor-␣ and TATA box-binding protein resulted in a conformational change in either or both proteins (40). Similarly, the present CD spectroscopy study demonstrates that a mixture of HNF-4␣-LBD and ACBP recombinant proteins exhibited a CD spectrum that was different from the theoretical sum of individual protein CD spectra, indicating a structural change upon HNF-4␣-LBD complex formation with ACBP ( Fig.   FIG. 11. Reporter assays indicate that ACBP stimulates HNF-4␣-mediated transactivation. Cells were transfected with reporter plasmid ApoBLuc and reference Renilla luciferase in all experiments. In addition, ACBP and HNF-4␣ expression vectors or empty vectors (for control) were cotransfected as indicated. A, luciferase transcription activity (relative light units) in COS-7 cells (bars 1-7) and hepatoma cells (bars 8 -10) transfected with expression vectors as indicated. B, luciferase activity was determined in COS-7 cells transfected with 0.5 and 1.0 g of sense ACBP expression vector (bars 5 and 6, respectively) or antisense ACBP plasmid (bars 9 and 10, respectively) in the presence of a constant amount of HNF-4␣ expression vector. Controls (bars 1-4 and 7 and 8) were run for cells transfected with no ACBP vector or empty vectors. s/a, sense/antisense. Table I). It is well known that nuclear receptors when purified as recombinant proteins exhibit a partially unstructured or random-coiled conformation in buffered aqueous solutions in contrast with their conformation in x-ray crystals (42). Hydrophobic solvents like trifluoroethanol or specific proteins acting as coactivators can increase the helical structure of nuclear receptor molecules in buffered aqueous solutions (40,42). Under our experimental conditions, in aqueous buffer HNF-4␣-LBD had a low helical structure; in contrast, ACBP, a very soluble protein in aqueous buffer, had a high content of ␣-helix, but in the presence of HNF-4␣-LBD it formed a complex with a CD spectrum indicating similarity to the individual HNF-4␣-LBD but with a higher content of ␣-helix than HNF-4␣-LBD alone ( Fig. 1 and Table I).

and
Second, ACBP directly interacted with HNF-4␣ in hepatoma cells in vitro and liver tissue in vivo as demonstrated by coimmunoprecipitation, immunogold electron microscopy, and confocal microscopy. HNF-4␣ was detected in anti-ACBP immunoprecipitates of liver and hepatoma cell homogenates (Fig. 2) and also by immunogold electron microscopy in nuclei of mouse liver cells (Fig. 3). Control experiments for the specificity of ACBP/HNF-4␣ association were run for both techniques used. In coimmunoprecipitation experiments, ACBP was found in precipitates generated with anti-HNF-4␣ but not with anti-Sp1 antibody. Immunoprecipitates produced with anti-ACBP antibody contained HNF-4␣ but not Sp-1 protein or GR nuclear receptor. In the literature, direct physical interaction of Sp1 protein with numerous nuclear transcription factors such as RAR/RXR (43), TR, vitamin D 3 receptor, and PPAR have been reported (44). Regarding HNF-4␣, however, the literature is controversial. Even though several reports (45)(46)(47)(48) demonstrated close proximity of response elements recognized by HNF-4 and Sp1 within regulatory sequences of several genes, only one article (49) suggested a direct interaction of HNF-4 with the Sp1 protein. The latter results are different from our immunoprecipitation data that showed no association of ACBP and Sp1 protein in mouse liver and rat hepatoma cells and may be explained by the use of different experimental conditions. For example, the pull-down assays described in this article (49) were run with in vitro synthesized 35 S-labeled HNF-4 and GST-Sp1 fusion proteins, which might have favored finding protein-protein association because of the following reasons. (i) In vitro incubation of large amounts of purified proteins may conduct nonspecific binding between them. (ii) 35 S-Labeled protein detection is severalfold more sensitive than the colorimetric (alkaline-phosphatase/colorimetric substrate) or chemiluminescent (horseradish peroxidase/chemiluminescent substrate) methods that were used in our experiments. In the same report (49), coimmunoprecipitation of FLAG-HNF-4 with Sp1 was demonstrated only in HepG2 cells that had been transfected to overexpress FLAG-tagged HNF-4 but not in nontransfected cells. This experimental approach does not exclude a nonspecific association between overexpressed FLAG-HNF-4 and Sp1. The fact that in our experiments, ACBP was found associated with HNF-4␣ but not with Sp1 suggested that either HNF-4␣ was not physically connected to Sp1 at all times under the experimental conditions employed or that ACBP was associated with HNF-4␣ that was not involved with Sp1. Using immunogold electron microscopy as a technique that provides accurate detection of protein-protein interactions, ACBP was found closely associated with HNF-4␣ (within 43 Å) but not with GR. Confocal microscopy experiments confirmed that ACBP and HNF-4␣ significantly colocalized in the nuclei of rat hepatoma (nontransfected cells) and COS-7 cells transfected with expression vectors for ACBP and HNF-4␣ (Figs. 4, 9, and 10). The intermolecular distance between ACBP and HNF-4␣ was determined to be 53 Å by FRET microscopy (Fig. 5). Additional support for close molecular interaction between these proteins was provided by mammalian two-hybrid assay with ACBP and HNF-4␣ cDNAs ligated to pACT and pBIND (Fig. 6).
Third, transactivation assays demonstrated that ACBP interaction with HNF-4␣ was functionally significant. Functional studies in COS-7 cells cotransfected with HNF-4␣, ACBP expression vectors, and ApoBLuc reporter plasmid indicated that ACBP had a stimulatory effect on HNF-4␣-mediated transactivation (Fig. 11A). An overexpression of ACBP in rat hepatoma cells resulted in a significant increase in HNF-4␣-mediated transactivation (Fig. 11A). In COS-7 cells, a dose response in HNF-4␣-mediated transactivation over a range of ACBP sense and antisense expression vectors was obtained, demonstrating that the more ACBP was expressed in the cells the higher the HNF-4␣-mediated transactivation was observed (Fig. 11B). When individual cells were examined for the expression of ACBP, HNF-4␣, and luciferase reporter, correlation curves demonstrated that individual cells with higher amounts of ACBP expressed more luciferase (Fig. 12). These data suggest that the stimulatory effect of ACBP on HNF-4␣-mediated transactivation could be explained by the ability of ACBP to interact directly with HNF-4␣ to alter its conformation. Interestingly, the stimulatory effect of ACBP on HNF-4␣ transactivation function is in agreement with ACBP inhibitory effects on PPAR␣ (15,20). Cross-talk between HNF-4␣ and PPAR␣ has been demonstrated to regulate apoC-III gene through common DR-1 consensus elements in its promoter (50). Thus, the overall process may include apoC-III transcription stimulation when ACBP (or ACBP/LCFA-CoA) activates HNF-4␣ to bind to DR-1-response elements, in balance with the opposite effect, i.e. apoC-III transcription down-regulation when free fatty acid/L-FABP stimulates PPAR␣/retinoid X receptor heterodimers to displace HNF-4␣ homodimers from DR-1 cis-elements.
In summary, the data presented herein demonstrate for the first time that ACBP directly interacts with HNF-4␣ in vitro, in intact fixed cells, and in living cells. Furthermore, ACBP expression levels directly correlated with HNF-4␣-mediated transactivation in transfected cells. These data were consistent with ACBP directly affecting HNF-4␣ transcriptional activity and/or altering the nuclear distribution of LCFA-CoA ligands that activate or inhibit HNF-4␣ (16 -18). Based on the fact that LCFA is cotransported to the nucleus as an LCFA⅐L-FABP complex and not as free LCFA (37), a possible scenario is that LCFA-CoA⅐ACBP complexes may also pass through the nuclear membrane pores.
Finally, the fact that ACBP directly interacts with HNF-4␣ to alter its structure provides for the first time a novel mechanism, possibly independent of specific ligands, whereby a primarily cytoplasmic protein involved in LCFA-CoA metabolism may affect the transcriptional activity of a nuclear regulatory protein. Regardless of whether HNF-4␣ binds no ligand, binds LCFA-CoA, and/or binds LCFA, the present data demonstrate the physical and functional interaction of ACBP with HNF-4␣. This suggests for the first time that ACBP may act as a coregulator involved in the transcription of genes targeted by HNF-4␣. How this interaction relates to LCFA-CoA interaction with HNF-4␣ is beyond the scope of the present investigation but is an interesting future possibility to be examined.