Glucose regulates fatty acid binding protein interaction with lipids and peroxisome proliferator-activated receptor α.

Although the pathophysiology of diabetes is characterized by elevated levels of glucose and long-chain fatty acids (LCFA), nuclear mechanisms linking glucose and LCFA metabolism are poorly understood. As the liver fatty acid binding protein (L-FABP) shuttles LCFA to the nucleus, where L-FABP directly interacts with peroxisome proliferator-activated receptor-α (PPARα), the effect of glucose on these processes was examined. In vitro studies showed that L-FABP strongly bound glucose and glucose-1-phosphate (Kd = 103 ± 19 nM and Kd = 20 ± 3 nM, respectively), resulting in altered L-FABP conformation, increased affinity for lipid ligands, and enhanced interaction with PPARα. In living cells, glucose stimulated cellular uptake and nuclear localization of a nonmetabolizable fluorescent fatty acid analog (BODIPY C-16), particularly in the presence of L-FABP. These data suggest for the first time a direct role of glucose in facilitating L-FABP-mediated uptake and distribution of lipidic ligands to the nucleus for regulation of PPARα transcriptional activity.


Fluorescent fatty acid and acyl-CoA binding
The fl uorescent C16 fatty acid analog (BODIPY C-16), the naturally occurring fl uorescent C18 cis -parinaric acid, and fl uorescent C18 trans -parinaric acid were purchased from Molecular Probes (Eugene, OR). The cis -parinaroyl-CoA and BODIPY C16-CoA were synthesized from the respective parent fatty acids and purifi ed by HPLC as previously described ( 9 ). To determine the effect of glucose and glucose metabolites on fl uorescent LCFA or fl uorescent LCFA-CoA binding to L-FABP, 100 nM L-FABP protein was mixed with 50 nM BODIPY C-16, BODIPY C16-CoA, trans -parinaric acid, or cis -parinaroyl-CoA, and the fl uorescence intensity of the fl uorophore was measured in the absence and presence of glucose. Emission spectra were obtained at 24°C with a PC1 photon counting spectrofl uorometer (ISS Inc.). The BODIPY fl uorophore was excited at 460 nm, and emission was measured from 490 nm to 540 nm. The parinaric lipids were excited at 317 nm, and emission measured from 380 nm to 450 nm. Data were corrected for background (buffer, fl uorescent ligands, and solvent effects), and maximal intensities were used to calculate the percent change in LCFA or LCFA-CoA binding as previously described ( 13 ).

Circular dichroism
To determine glucose effects on protein secondary structure, circular dichroic (CD) spectra of 0.8 M PPAR ␣ ⌬ AB, 2.4 M L-FABP, and equal amino acid molar concentrations of the two proteins [0.4 M PPAR ␣ ⌬ AB + 1.2 M L-FABP] were taken in the absence and presence of glucose, maltose, and glucose metabolites with a J-815 spectropolarimeter (JASCO, Easton, MD) as previously described ( 13,20 ). Ten scans per replicate were averaged for percent composition of secondary structures by three different methods (SELCON3, CDSSTR, and CONTIN/LL) with the software package CDPro ( 23 ) as previously described ( 13,20 ). The CD spectrum of the mixed proteins was compared with a theoretical spectrum of combined but noninteracting proteins in the absence and presence of sugar. This spectrum was the average of each protein analyzed separately at a concentration equal to that in the mixture ( 24 ).

Co-immunoprecipitation assays
Briefl y, 20 g of each recombinant protein was mixed and allowed to incubate for 10 min in the absence or presence of glucose. Next, protein mixtures were incubated with anti bodylinked resin (rabbit polyclonal anti-PPAR ␣ [Affi nity BioReagents, Golden, CO] or goat polyclonal anti-L-FABP [Santa Cruz Biotechnology, Santa Cruz, CA]) for 1 1/2 h at room temperature according to the manufacturers' instructions (ProFound TM mammalian co-immunoprecipitation kit, Pierce Biotechnology, Rockford, IL). Eluted proteins were examined by SDS-PAGE analysis as previously described ( 13 ). Values were normalized to the amount of protein detected from the control (e.g., the protein used for co-immunoprecipitation [CoIP]), and samples in absence of glucose were arbitrarily set to 100 percent.

Fluorescence resonance energy transfer
To further confi rm effects of glucose on direct binding of PPAR ␣ ⌬ AB and L-FABP, these recombinant proteins were fl uorescently labeled with Cy3 and Cy5 dyes, respectively, using a Fluorolink-antibody Cy3 and Cy5 labeling kit (Amersham Biosciences, Pittsburgh, PA) according to the manufacturer's recommendations. Absorbance measurements were made of labeled and nonlabeled proteins to determine protein concentrations and dye to protein ratios. Cy3-PPAR ␣ ⌬ AB protein (30 nM) was titrated with increasing concentrations of Cy5-L-FABP, and the endogenous, high-affi nity ligand of PPAR ␣ ( 13, 14 )-thereby suggesting that PPAR ␣ may be responsible for maintaining energy homeostasis through concentration-dependent regulation by both lipids and sugars.
Parallel studies to these developments have begun to elucidate the mechanism whereby highly lipophilic, poorly water soluble ligands, such as LCFA and LCFA-CoA, are transported to the nucleus for interacting with and activating PPAR ␣ (reviewed in Ref. 15 ). Key in this regard is the fi nding that metabolically active tissues, such as liver, express high levels (2-3% of cytosolic protein) of L-FABP, which has high affi nity for both LCFA and LCFA-CoA (16)(17)(18). L-FABP is thought to function as a shuttle that delivers these metabolites to the nucleus for interaction with nuclear receptors, such as PPAR ␣ ( 15,19,20 ). Since glucose potentiates PPAR ␣ activation by LCFA ( 13,14 ), a potential effect of glucose on L-FABP and/or the L-FABP-PPAR ␣ interaction could play an important role in the maintenance of energy homeostasis. To begin to resolve the role L-FABP may play in such regulation, the effect of glucose directly on L-FABP and the PPAR ␣ -L-FABP complex was examined.

Protein expression and purifi cation
Rat recombinant L-FABP protein was produced, purifi ed, and delipidated as described ( 21 ). Mouse recombinant PPAR ␣ ⌬ AB protein (plasmid generously provided by Dr. Noa Noy, Case Western Reserve University) was produced and purifi ed as described ( 8,13 ).

Glucose binding assay
The direct binding of glucose and glucose metabolites (glucose-1-phosphate [G-1-P] and glucose-6-phosphate [G-6-P]) or maltose to L-FABP (500 nM) was determined by quenching of intrinsic L-FABP aromatic amino acid fl uorescence as previously described for PPAR ␣ ( 9, 13 ). Emission spectra from 300-400 nm were obtained at 24°C upon excitation at 280 nm with a PC1 photon counting spectrofl uorometer (ISS, Champaign, IL). Data were corrected for background (buffer, ligands, and solvent effects), and maximal intensities were used to calculate the dissociation constant ( K d ) with the ligand binding function of SigmaPlot (SPSS, Chicago, IL) utilizing the one-site saturation feature ( 22 ).
To determine if binding may also occur without resulting in tyrosine quenching, binding was also determined as quenching of fl uorescently labeled L-FABP in the presence of glucose or glucose metabolites. Recombinant L-FABP protein was chemically labeled with Cy5 dye using a Fluorolink-antibody Cy5 labeling kit (Amersham Biosciences, Pittsburgh, PA) according to the manufacturer's recommendations. Absorbance measurements estimated the labeling effi ciency to be approximately one dye per protein molecule. Cy5-L-FABP (25 nM) was titrated with increasing concentrations of glucose, glucose-1-phosphate, glucose-6phosphate, or maltose at 24°C, and emission spectra from 655-700 nm were obtained upon excitation of the Cy5 dye at 645 nm in a PC1 photon counting spectrofl uorometer (ISS, Champaign, IL). Data were corrected for background (buffer, ligands, and solvent effects) and maximal intensities were used to calculate the dissociation constant ( K d ) with the ligand binding function of SigmaPlot (SPSS, Chicago, IL) utilizing the one-site saturation feature ( 22 ).

L-FABP exhibited high affi nity for glucose and glucose metabolites
Although L-FABP binds LCFA and LCFA-CoA with high affi nity (16)(17)(18), it is not known if L-FABP also binds glucose with high affi nity. Therefore, this possibility was examined by a direct fl uorescence binding assay taking advantage of the change in L-FABP intrinsic aromatic amino acid fl uorescence intensity in response to ligand binding. Plotting the change in L-FABP amino acid fl uorescence as a function of increasing glucose concentration yielded a sharp saturation curve with the maximal change in fl uorescence intensity noted at approximately 400 nM ( Fig. 1A ). Transformation of the binding data into a linear reciprocal plot ( Fig. 1A , inset) indicated high affi nity binding at a single binding site. Multiple replicates (n = 4) showed that L-FABP strongly bound glucose ( K d = 103.1 ± 18.6 nM), in the same range as previously shown for L-FABP binding to LCFA and LCFA-CoA ( 16 ). L-FABP binding to glucose-1-phosphate (G-1-P) was also saturable ( Fig. 1B ) at a single binding site ( Fig. 1B , inset) with even higher affi nity as shown by K d = 20.2 ± 3.5 nM. In contrast, glucose-6-phosphate (G-6-P) was only weakly bound ( Fig.  1C ). To determine the specifi city of this binding, the effect of maltose on L-FABP intrinsic fl uorescence was examined, and no binding was noted ( Fig. 1G ).
As it is possible for binding to occur without resulting in quenching of intrinsic amino acid fl uorescence, the binding of glucose and glucose metabolites to Cy5-labeled L-FABP was examined. The change in Cy5 fl uorescence intensity as a function of glucose ( Fig. 1D ) or G-1-P ( Fig.  1E ) concentration resulted in sharp saturation curves, whereas G-6-P binding resulted in a much shallower curve ( Fig. 1F ). Transformation of these data into linear reciprocal plots ( Fig. 1D-F , insets) established high-affi nity binding of glucose and G-1-P at a single binding site in L-FABP. Multiple replicates (n = 4) yielded binding affi nities for glucose and G-1-P that were similar, although slightly weaker, to those obtained through the direct L-FABP aromatic amino acid fl uorescence binding assay ( K d = 197.6 ± 53.4 nM for glucose and K d = 26.5 ± 4.4 nM for G-1-P, respectively). While the Cy5-L-FABP binding assay was able to detect G-6-P binding to L-FABP, the affi nity was three orders of magnitude weaker ( K d = 26.7 ± 7.9 M), and no binding was noted for maltose ( Fig. 1H ). Taken together with the fact that these binding affi nities of L-FABP for glucose, G-1-P, and G-6-P were in the range of physiological intracellular concentrations of these molecules ( 13 ), these data were consistent with glucose and especially its G-1-P metabolite being potential endogenous ligands of L-FABP.

L-FABP binding to glucose resulted in a glucose concentration-dependent alteration in L-FABP secondary structure
To determine whether glucose binding altered L-FABP conformation, CD was used to examine glucose effects on L-FABP secondary structure. The CD spectrum of L-FABP in the absence of glucose exhibited a large peak in molar emission spectra from 560 nm to 700 nm were measured upon excitation at 550 nm in a PC1 photon counting spectrofl uorometer (ISS) as previously described ( 20,25 ). This experiment was repeated in the presence of varying glucose concentrations. All spectra were corrected for background fl uorescence (buffer and Cy5-L-FABP bleed-through), and maximal fl uorescence intensities were used to calculate the affi nity of the two proteins as well as the intermolecular distance between them as previously described ( 20,26 ).

Cell culture
Murine L-cells (L arpt-tk-), mock-transfected control L-cells, and L-cells transfected with cDNA encoding for L-FABP ( 27 ) were seeded onto Lab-Tek chambered cover glass and cultured as previously described ( 11 ). Culture media was replaced with PBS containing 1.25 M Syto59 DNA-binding dye (Molecular Probes) and incubated at 37°C for 25min. Cells were washed with PBS, followed by the addition of 100 nM BODIPY C-16 (Molecular Probes) in the presence or absence of glucose, maltose, or sodium chloride at 37°C for 15 min, and then images were acquired in PBS at room temperature.

Laser scanning confocal microscopy
The real-time distribution of LCFA to nuclei (localization of PPAR ␣ ) in response to glucose was determined by laser scanning confocal microscopy (LSCM) of BODIPY C-16, a poorly metabolizable fl uorescent LCFA ( 11 ). LSCM studies were performed with a 63× Plan-Fluor oil immersion objective, N.A.1.45, an Axiovert 135 microscope (Zeiss, Carl Zeiss, Thornwood, NY), and MRC-1024 fl uorescence imaging system (Bio-Rad, Hercules, CA) as described ( 11 ). Syto59 and BODIPY C-16 probes were excited at 488/568 lines with a krypton-argon laser (Coherent, Sunnyvale, CA). Emission from BODIPY-labeled fatty acid was recorded by a photomultiplier after passing through a 522/D35 emission fi lter, and emission from Syto59 was collected with a 680/32 emission fi lter, both under manual gain and black level control. The objective was focused to acquire 0.3 m confocal slice images through a median section of cells in the fi eld. Cells were excited for 0.1 s intervals, regulated by a computer-controlled shutter and Laser Sharp software (Bio-Rad). Cells incubated with each component individually were used as background controls and to optimize instrument settings. Images were analyzed using MetaMorph software (Advanced Scientifi c Imaging, Meraux, LA). Syto59 fl uorescence was used to outline the nuclear region, so that the amount of BODIPY C-16 fl uorescence could be measured both in the entire cell and just in the nucleus. The cellular uptake of BODIPY C-16 was calculated by dividing the BODIPY C-16 fl uorescence of the whole cell by the area of the whole cell for each cell; the amount of BODIPY C-16 in the nucleus was calculated by dividing the BODIPY C-16 fl uorescence inside the nucleus by the area of the nucleus for each cell; the percentage of BODIPY C-16 in the nucleus was calculated by dividing the BODIPY C-16 fl uorescence intensity inside the nucleus by the BODIPY C-16 fl uorescence intensity of the whole cell multiplied by 100 percent for each cell. The fi nal data were averages of 50-200 cells after each experiment was replicated three to four times in separate chambers, using several images from different areas within each chamber.

Statistical analysis
All results are expressed as mean ± standard error. Statistical signifi cance between samples in the presence or absence of glucose, glucose metabolites, or maltose was determined by using the Student's t -test or ANOVA with P < 0.05. glucose ( Table 1 ). Although maltose concentrations of 6 mM or lower resulted in only minor spectra changes and no signifi cant changes to the secondary structural composition, maltose concentrations of 20 mM strongly altered the secondary structure ( Fig. 2D ) and percent composition ( Table 1 ). This fi nding suggests that the structural changes resulting from low glucose levels are due to specifi c, highaffi nity binding, while the differing effects seen at high sugar to protein ratios may be due to binding at other sites or changes in osmolarity. The glucose metabolites ( Fig. 2E ; G-1-P, ᭺ ; G-6-P, ᭢ ) elicited weaker conformational changes than glucose itself, and only at higher concentrations. Upon analysis of spectra for percent composition, the addition of physiological concentrations of G-1-P (11 M) or G-6-P (0.2 mM) resulted changes similar to M glucose concentrations, with strongly decreased ␣ -helical content and increased ␤ -sheets ( Table 1 ). These data suggested that glucose-and glucose metabolite-induced L-FABP structural changes might alter L-FABP's interaction with ellipticity at 190 nm and a negative peak at 222 nm ( Fig. 2A -E , ᭹ ). CD spectra obtained over a range of glucose:L-FABP ( Fig. 2A, C ) and maltose:L-FABP ( Fig. 2B, D ) molar ratios demonstrated that glucose signifi cantly altered the relative intensities of these peaks, while maltose had only minor changes on the L-FABP CD spectra. Analysis of these CD spectra for percent composition of individual secondary structural components demonstrated that low glucose ( р 60 M) concentrations elicited strong decreases in the proportion of ␣ -helices and reciprocally strong increases in the proportion of ␤ -sheets, while no signifi cant changes were noted for low concentrations of maltose ( Table 1 ). While the addition of glucose at near-physiological serum levels (6 mM) resulted in only minor alterations to L-FABP (  ) and less so cis -parinaroyl-CoA binding ( Fig. 3B , ᮀ ). These fi ndings suggested that the effect of glucose on enhancing lipidic ligand binding to L-FABP was not just an artifact of the fl uorescent BODIPY moiety but in fact represented the effect one might observe on the binding of naturally occurring LCFA or LCFA-CoA.

Glucose enhanced L-FABP interaction with PPAR ␣ by co-immunoprecipitation
To determine whether the glucose-induced alterations in secondary structures observed by CD would increase not only lipidic ligand binding but also L-FABP interaction with the nuclear receptor (PPAR ␣ ) regulated by these ligands, the effect of glucose on co-immunoprecipitation of L-FABP with PPAR ␣ was examined. Although low levels of glucose had only minor effects, high glucose concentrations (20-30 mM) increased co-immunoprecipitation of L-FABP with PPAR ␣ ( Fig. 4 ). Although the CD results at other ligands or other proteins, a possibility examined in a later section.

Glucose increased L-FABP binding capacity for LCFA and LCFA-CoA
To determine if the high-affi nity binding of glucose altered lipidic ligand binding to L-FABP, the ability of glucose and glucose metabolites to displace an L-FABP-bound fl uorescent LCFA and/or fl uorescent LCFA-CoA from the L-FABP binding pocket was examined. This was accomplished by use of BODIPY C-16 (fl uorescent LCFA analog) and BODIPY C16-CoA (fl uorescent LCFA-CoA analog), both of which were previously shown to be bound by L-FABP ( 12 ). Glucose signifi cantly increased BODIPY C-16 ( Fig. 3A , ) and BODIPY C16-CoA ( Fig. 3A , ᮀ ) binding to L-FABP up to 2-fold. In contrast, glucose metabolites (G-1-P and G-6-P) or phosphate (negative control) did not signifi cantly alter BODIPY-C16 binding; only G-1-P slightly decreased BODIPY C16-CoA binding to L-FABP ( Fig. 3A , ᮀ ).
To determine whether the effect of glucose in enhancing binding of BODIPY-labeled LCFA and LCFA-CoA to L-FABP was due to the presence of BODIPY in these synthetic fl uorescent LCFA and LCFA-CoA derivatives, the (PPAR ␣ + L-FABP) in the presence of 6 M ( Fig. 5A , ᭺ ) and 60 M ( Fig. 5B , ᭺ ) glucose differed from the spectra of (PPAR ␣ + L-FABP) in the absence of glucose ( ᭹ ) but were almost superimposable over the theoretical spectra of (PPAR ␣ + L-FABP) in the presence of glucose if no protein interaction occurs ( ᭢ ). When examined for percent composition, the addition of 0.6 M or 6 M glucose resulted in values that were not signifi cantly different from those calculated for the theoretical values if no protein-protein interaction occurred ( Table 2 ), suggesting that low glucose concentrations prevented the conformational change induced by PPAR ␣ binding with L-FABP. To examine the specifi city of glucose on this interaction, the effect of low concentrations of maltose was examined. No signifi cant dif-these higher sugar concentrations suggested that high sugarto-protein ratios might result in additional effects, the amount of protein in these assays was more than 10-fold greater than in the CD experiments, suggesting that these effects were both relevant and specifi c. These data suggest that although low glucose levels have minor effects on L-FABP interaction with PPAR ␣ , higher glucose concentrations, such as those seen in diabetes, increased L-FABP interaction with PPAR ␣ .

Glucose altered the secondary structures of the L-FABP complex with PPAR ␣
To further confi rm that glucose increased interaction of L-FABP with PPAR ␣ , the effect of glucose on secondary structure was examined by CD. The actual spectra of both proteins Signifi cant differences between L-FABP structural composition in the presence of sugar compared with L-FABP structural composition in the absence of sugar (L-FABP). L-FABP, liver fatty acid binding protein.
a P < 0.001, n = 3-6. b P < 0.01, n = 3-6. c P < 0.05, n = 3-6. complex with L-FABP and that the ratio of sugar to protein may be an important regulatory mechanism in affecting these protein-protein interactions.
To resolve whether glucose metabolites might also affect the PPAR ␣ -L-FABP interaction detected by CD, spectra of (PPAR ␣ +L-FABP) were recorded in the presence of 11 M G-1-P ( Fig. 5G ) and 0.2 mM G-6-P ( Fig. 5H ). These spectra ( ᭺ ) differed strongly from the spectrum of (PPAR ␣ + L-FABP) in the absence of glucose ( ᭹ ), with decreased minima and maxima similar to those observed for 6 M glucose ( Fig. 5A , ᭺ ), and only slight variation was noted between the actual and theoretical spectrum for no protein interaction ( ᭢ ). Both G-1-P and G-6-P decreased the ␣ -helical content and increased the ␤ -sheet content of the PPAR ␣ -L-FABP complex similar to changes obtained with 60 M glucose ( Table 2 ). These data suggested that not only glucose but also both glucose metabolites can elicit conformational changes in the PPAR ␣ complex with L-FABP.

Glucose-induced conformational changes in the PPAR ␣ complex with L-FABP as shown by fl uorescence resonance energy transfer
To determine if the glucose-mediated increases in secondary structure of the PPAR ␣ complex with L-FABP might also extend to changes in tertiary structure, a fl uorescence resonance energy transfer (FRET) study was undertaken using Cy3 and Cy5 dyes, which are small fl uorescent dyes that form an effi cient FRET donor/ acceptor pair ( 24,28 ). As FRET effi ciency varies inversely as the 6th root of intermolecular distance, energy transfer will only occur if Cy3-PPAR ␣ and Cy5-L-FABP are in close proximity (i.e., 0-100Å). The recombinant proteins were chemically labeled with Cy3 (PPAR ␣ ) or Cy5 (L-FABP) dyes, and the labeling ratios of fl uorescent dye to protein were maintained low to ensure an average of one dye per protein molecule. FRET between these fl uorescently labeled proteins was then used to determine the intermolecular distance between the fl uorescent tags in the PPAR ␣ -L-FABP complex. In the absence ( Fig. 6A ) and presence of glucose (6 mM, Fig. 6C ; 20 mM, Fig. 6E ; 30 mM, Fig. 6G ) or glucose metabolites (G-1-P, Fig. 7A ; G-6-P, Fig. 7C ), FRET was observed to quench donor Cy3-PPAR ␣ fl uorescence and to increase the appearance of sensitized emission of acceptor Cy5-L-FABP . In the absence of glucose, the intermolecular distance between Cy3-PPAR ␣ and Cy5-L-FABP was determined to be about 49 Å ( Table 3 ), which is consistent with close molecular interaction. High glucose (20-30 mM), as well as physiological levels of glucose metabolites (G-1-P and G-6-P) all increased the intermolecular distance between Cy3-PPAR ␣ and Cy5-L-FABP by as much as 11Å ( Table 3 ). Taken together, these FRET data suggested that glucose binding altered not only the secondary structure (determined by CD) but also the tertiary conformation of the PPAR ␣ -L-FABP complex.

Glucose increased PPAR ␣ affi nity for L-FABP as shown by FRET
To determine if the glucose-induced alterations (determined by CD and FRET intermolecular distance) in the PPAR ␣ -L-FABP complex were associated with altered ferences in spectra ( Fig. 5C ) or percent composition ( Table  2 ) were noted for maltose at 6 M or 60 M, suggesting that low maltose concentrations had no effect on the PPAR ␣ -L-FABP interaction.
In contrast, higher glucose concentrations resulted in spectra and structural component contributions indicative of signifi cant additional conformational alterations in the PPAR ␣ -L-FABP complex. The spectra of (PPAR ␣ + L-FABP) in the presence of 0.6 mM (spectrum not shown), 6 mM ( Fig. 5D , ᭺ ) and 20 mM ( Fig. 5E , ᭺ ) glucose all exhibited decreased maxima at 190 nm and minima at 222 nm compared with the spectrum of (PPAR ␣ + L-FABP) in the absence of glucose ( ᭹ ). Although the spectra at each of these concentrations also varied from the theoretical spectra for no PPAR ␣ -L-FABP interaction ( ᭢ ), the largest spectral changes were noted for (PPAR ␣ + L-FABP) in the presence of 6 mM glucose. When analyzed for percent composition, decreased ␣ -helices and increased ␤ -sheets were noted compared with the no-glucose controls, and (PPAR ␣ + L-FABP) in the presence of 0.6 mM or 6 mM glucose resulted in even lower ␣ -helical content and increased ␤ -sheets than the theoretical ( Table 2 ). However, while 6 mM maltose resulted in spectra that were superimposable over the spectrum of (PPAR ␣ + L-FABP) in the absence of sugar ( Fig. 5F ) and had no signifi cant alterations to the percent composition, 20 mM maltose strongly altered the CD spectra and resulted in decreased ␣ -helical content and increased ␤ -sheets ( Table 2 ). Taken together, these data suggested that glucose concentrations between 6 M and 6 mM elicited more marked conformational changes in the PPAR ␣

Glucose and L-FABP increase LCFA uptake and nuclear localization in cultured cells
L-FABP is known to enhance transport of bound LCFA to the nucleus ( 11,12,29 ). As the above data showed that glucose increased L-FABP's affi nity for lipidic ligands (LCFA and LCFA-CoA) and for PPAR ␣ , the possibility that glucose might thereby alter the distribution of LCFA to the nucleus was examined using BODIPY C-16, a fl uorescent fatty acid analog that is bound with high affi nity by both PPAR ␣ ( 13 ) and L-FABP ( 12 ). Uptake and nuclear distribution were examined in cells devoid of L-FABP (mock-transfected control L-cells; "control") and transfected cells expressing L-FABP (transfected L-cells overexpressing L-FABP; "L-FABP") ( 11,12 ). These cells were stained with the nuclear SYTO59 stain ( Fig. 8 , red pixels) as a colocalization marker for BODIPY C-16 distribution to nuclei. In control cells not expressing L-FABP, total BODIPY C-16 uptake (Fig. 8A1, green pixels) and the amount of BODIPY C-16 localizing to the nucleus (observed as colocalized pixels with the SYTO59 nuclear stain, Fig. 8A2, yellow pixels) was relatively low in the absence of glucose. However, addition of glucose signifi cantly increased BODIPY C-16 uptake (Fig. 8C1, green pixels; 8E1, green pixels; strength of binding between these proteins, FRET between Cy3-PPAR ␣ and Cy5-L-FABP was used to determine their binding affi nity in the absence and presence of glucose. By plotting the change in fl uorescence intensity of the donor (excitation 550 nm, emission 570 nm) as a function of increasing acceptor concentration, the affi nity of PPAR ␣ for L-FABP was calculated ( Table 3 ). In the absence of glucose ( Fig. 6B ) or in the presence of glucose (6 mM, Fig. 6D ; 20 mM, Fig. 6F ; 30 mM, Fig. 6H ) or glucose metabolites (G-1-P, Fig. 7B ; G-6-P, Fig. 7D ), a saturable binding curve was obtained, suggesting strong affi nity binding. Multiple replicates (n = 4) revealed that PPAR ␣ bound L-FABP with high affi nity in the absence of glucose, K d = 31.3 ± 5.7 nM ( Table 3 ). Increasing concentrations of glucose or G-6-P increased the binding affi nity 3-4-fold ( Table 3 ). In contrast, G-1-P reduced this affi nity by 45% as shown by 1.5-fold weaker K d ( Table 3 ). This suggests that even though glucose, G-1-P, and G-6-P all altered the overall protein conformation, the conformational changes elicited by G-1-P binding were different and resulted in an opposite effect on the PPAR ␣ -L-FABP interaction compared with the effects of glucose and G-6-P. ing cells having decreased uptake compared with control cells, these results varied signifi cantly from those obtained with glucose ( Fig. 9A ). Further, in the presence of L-FABP, the addition of 6 mM maltose or 3 mM sodium chloride had no effect on BODIPY C-16 uptake or distribution to nuclei compared with cells in PBS. This suggested that effects from glucose are not just due to changes in osmolarity.

L-FABP altered the percentage of LCFA localizing to the nucleus
To determine if increased nuclear BODIPY C-16 amounts (concentration) corresponded with increased preferential nuclear BODIPY C-16 distribution (i.e., increased percentage in the nucleus), rather than simply being a consequence of increased overall uptake (total concentration), the proportion of BODIPY C-16 in the nucleus (versus the rest of the cell) was examined and expressed as a percentage of that taken up. In all cases, L-FABP increased the percentage of BODIPY C-16 localizing to the nucleus ( Fig. 9C ). Glucose only seemed to affect the percent distribution at high concentrations and in the presence of L-FABP, while maltose in the absence of L-FABP actually decreased the percent distribution, and sodium chloride had no effect ( Fig. 9C ).
Thus, glucose and L-FABP modifi ed the percentage of BODIPY C-16 in the nucleus in a complex, concentrationdependent manner. For example, although the addition of 20 mM glucose to the control cells resulted in relatively the same amount of nuclear BODIPY C-16 as L-FABPexpressing cells in the absence of glucose ( Fig. 9B ), the distribution of fatty acid within the nucleus was different. The control cells showed a higher percentage of the BODIPY 8G1, green pixels; 8I1, green pixels; 9A, ) and nuclear localization (Fig. 8C2, yellow pixels; 8E2, yellow pixels; 8G2, yellow pixels; 8I2, yellow pixels; 9B, ), with 20 mM glucose eliciting the maximal increase.
Overexpression of L-FABP, in the absence of glucose, increased both total uptake (Fig. 8B1, green pixels) and nuclear localization (Fig. 8B2, yellow pixels) of BODIPY C-16. These increases corresponded to a 1.5-fold increase in the amount of uptake ( Fig. 9A , ᮀ ) and an almost 3-fold increase in nuclear localization ( Fig. 9B , ᮀ ) compared with mock-transfected control cells with no L-FABP expression (fi lled bars). High concentrations of glucose exacerbated the effect of L-FABP overexpression on cellular BODIPY C-16 uptake (Fig. 8F1, green pixels; 8H1, green pixels; 8J1, green pixels; 9A, ᮀ ). Furthermore, both control and L-FABP overexpressing cells showed similar maximal fatty acid uptake at 20 mM, suggesting that at high glucose levels the L-FABP effect on uptake may be diminished. In addition, glucose increased nuclear BODIPY C-16 levels in the L-FABP overexpressing cells (Fig. 8D2, yellow pixels; 8F2, yellow pixels; 8H2, yellow pixels; 8J2, yellow pixels; 9B, ᮀ ) as well as the control cells. These data suggested that both glucose and L-FABP expression functioned to increase LCFA uptake and increase LCFA nuclear localization, with the effect being greatest when both glucose and L-FABP were high.
To determine whether these effects might be due to changes in osmolarity, rather than glucose binding, these experiments were repeated in the presence of 6 mM maltose or 3 mM sodium chloride ( Fig. 9 ). Although the effects of 6 mM maltose and 3 mM sodium chloride on cellular BODIPY C-16 uptake were similar to each other, with L-FABP-express- these ligands to PPAR ␣ for activating transcription of multiple genes involved in LCFA ␤ -oxidation (rev. in Ref. 15 ). However, recent data show that although lipidic ligands, such as LCFA, activate PPAR ␥ (another PPAR isoform), ligand binding is not required for PPAR ␥ functionality ( 30 ). Taken together these fi ndings suggest that cytoplasmic lipid binding proteins, such as L-FABP, may activate PPAR ␣ , either by delivering bound LCFA to the nuclear receptor and/or by directly interacting with the nuclear receptor to change its conformation and coactivator association. Consistent with this possibility, in the absence of lipidic ligands, L-FABP directly interacts with PPAR ␣ to alter conformation and coactivator recruitment ( 20 ). However, in addition to the above mechanisms of lipiddependent and lipid-independent PPAR ␣ activation, recent fi ndings indicate that PPAR ␣ also exhibits high-affi nity binding of and activation by glucose ( 13,14 ). Thus, PPAR ␣ is unique among the ligand-activated nuclear receptors in that its activity is regulated by two very different classes of nutrients/endogenous ligands: hydrophobic lipids (LCFA, LCFA-CoA, and fi brates) ( 8-10 ) and hydrophilic sugars (glucose and glucose metabolites) ( 13,14 ). As L-FABP is the predominant LCFA and LCFA-CoA binding protein in C-16 located at the nuclear envelope membranes (Fig.  8G2), while the L-FABP-expressing cells (L-FABP) showed a more even dispersion of BODIPY C-16 within the nucleoplasm (Fig. 8B2). Addition of glucose to the L-FABPexpressing cells (L-FABP) resulted in a higher level of BODIPY C-16 throughout the nucleoplasm, as well as around the nuclear membrane ( Fig. 8 ; 0.6 mM, D2; 6 mM, F2; 20 mM, H2), while the addition of 30 mM glucose resulted in less BODIPY C-16 at the nuclear membrane (Fig.  8J2). Based on earlier studies showing that L-FABP cotransports bound LCFA into the nucleus ( 29 ), this variation in BODIPY C-16 distribution could be due to BODIPY C-16 entering the nucleus through diffusion versus L-FABP facilitated transport (e.g., free fatty acid versus carrier protein bound).

DISCUSSION
It has been proposed that cytoplasmic lipid binding proteins, such as L-FABP, might function in long-term regulation of hepatic LCFA ␤ -oxidation by transporting bound LCFA or LCFA-CoA into the nucleus, directly interacting with nuclear receptors, such as PPAR ␣ , and transferring this possibility as shown herein, L-FABP bound glucose with high affi nity ( K d = 103.1 ± 18.6 nM), approximately 50-fold weaker than previously reported for PPAR ␣ binding of glucose ( 13 ). Furthermore, glucose enhanced the interaction between PPAR ␣ and L-FABP. Since L-FABP is known to enter nuclei to directly interact with PPAR ␣ ( 12,20 ), these fi ndings suggest L-FABP-bound glucose may be transported to and infl uence transcriptional activity of PPAR ␣ . Based on the glucose binding sites of other glucose binding proteins ( 32 ), it appears that a combination of asparagine (N), aspartic acid (D), or glutamic acid (E) molecules are required for glucose binding to occur, and phenylalanine (F) residues may stabilize such interactions. The X-ray crystal structure of holo-L-FABP reveals not only the expected two LCFA binding sites but also an additional surface site occupied by a small molecule, butyric acid ( 33 ). This site is localized near the end of the fi rst ␣ -helix of L-FABP N-terminus. Furthermore, even with two LCFA molecules bound within the L-FABP binding pocket, there is still space within the binding cavity for several small molecules, and three small subcavities exist ( 33 ). Although the exact binding site for glucose is unknown, based upon the D, E, N, or F amino acid requirement, it is possible that glucose binding occurs through hydrogen bonding at one of two locations: between residues E13, F15, and F18 or near residues E62, F63, E72, and F95.
Second, glucose altered L-FABP conformation to enhance binding of lipidic ligands, as shown by CD and ligand binding assays in the presence of glucose. Furthermore, liver ( 31 ) and directly interacts with PPAR ␣ ( 20 ), it was important to examine the effect of glucose on these processes. The data obtained herein demonstrated for the fi rst time that (i) L-FABP bound glucose with high affi nity, resulting in increased LCFA/LCFA-CoA binding; (ii) Glucose increased PPAR ␣ 's affi nity for L-FABP; (iii) Glucose increased nuclear localization and relative distribution of LCFA in a concentration-dependent manner, both in the presence and absence of L-FABP; and (iv) A complex interaction exists between glucose, lipid, and L-FABP levels in the regulation of PPAR ␣ . These studies yielded several new insights regarding the regulation of PPAR ␣ by L-FABP.
First, L-FABP may function in transporting not only bound lipidic ligands (LCFA and LCFA-CoA) but also bound glucose to PPAR ␣ in the nucleus. Consistent with FRET, fl uorescence resonance energy transfer; G-1-P, glucose-1phosphate; G-6-P, glucose-6-phosphate; L-FABP, liver fatty acid binding protein; PPAR, peroxisome proliferator-activated receptor. affi nity than G-1-P and approximately 30-fold stronger affi nity than G-6-P ( 13 ). In contrast, L-FABP affi nity for G-6-P was extremely weak, almost 250-fold less than for glucose. Further, L-FABP binding of G-1-P was approximately 5-fold higher than glucose binding by L-FABP and about the same as previously reported for PPAR ␣ binding of G-1-P ( 13 ). Additionally, the glucose metabolites differed in their ability to affect L-FABP binding to LCFA/ LCFA-CoA and in their ability to affect the PPAR ␣ -L-FABP glucose-induced L-FABP structural changes were stronger than previously reported for glucose binding to PPAR ␣ ( 13 ). Unlike the glucose-induced PPAR ␣ conformational changes, which did not alter LCFA binding and decreased LCFA-CoA binding, the structural changes resulting from glucose binding by L-FABP increased binding of both LCFA and LCFA-CoA to L-FABP. These data suggest that glucose could function to increase cellular uptake and transport of lipids through its interaction with L-FABP and that L-FABP could simultaneously transport both glucose and LCFA into the nucleus for interaction with and activation of PPAR ␣ . Consistent with this possibility, glucose enhanced cellular uptake of exogenous LCFA uptake and facilitated the delivery of LCFA from the cytoplasm into the nucleus, a process potentiated by L-FABP. Studies in cultured cells have shown that cytoplasmic glucose levels are 2-8-fold higher than nucleoplasmic glucose concentrations, depending on the extracellular glucose concentration ( 13 ). The difference between nucleoplasmic and cytoplasmic glucose levels, together with the higher binding affi nity of PPAR ␣ for glucose, may favor L-FABP offl oading of bound LCFA in the nucleus for binding to and activating PPAR ␣ .
Third, L-FABP and PPAR ␣ differed markedly in their binding of glucose metabolites produced as a result of hexokinase action within cells. Although PPAR ␣ binds not only glucose but also G-1-P and G-6-P with high affi nity, PPAR ␣ binds glucose with approximately 10-fold stronger Fig. 9. L-FABP expression and addition of glucose synergistically increase BODIPY C-16 uptake and nuclear concentrations of BODIPY C-16 in a concentration-dependent manner. (A) Cellular BODIPY C-16 fl uorescence from control ( ᮀ ) and L-FABP-expressing ( ) L-cells was calculated per cell by taking the BODIPY C-16 fl uorescence intensity of the whole cell divided by the area of the whole cell (to compensate for variation due to cell size); these values were then averaged over n = 50-100 cells. (B) Using the Syto59 fl uorescence, the nuclear perimeter was marked, allowing for analysis of the BODIPY C-16 levels in the nucleus. Nuclear BODIPY C-16 fl uorescence from control ( ᮀ ) and L-FABP-expressing ( ) L-cells was calculated per nucleus by taking the BODIPY C-16 fl uorescence intensity of each nucleus divided by the area of the nucleus (to compensate for variation due to cell size); these values were then averaged over n = 50-200 cells. (C) The percentage of BODIPY C-16 localizing to the nucleus, calculated by dividing the total BODIPY C-16 fl uorescence in the nucleus by the total fl uorescence in the whole cell and multiplying by 100%. Values are presented as the mean value ± SEM, from 50-100 cells from two to three separate replicates. Asterisks (*) indicate signifi cant differences from the 0 mM controls; P < 0.05. Pound symbols (#) indicate signifi cant differences due to the presence of L-FABP compared with wild-type cells under the same conditions; P < 0.05. L-FABP, liver fatty acid binding protein. the existence of dynamic interrelationships between glucose, lipidic ligands (LCFA, LCFA-CoA, fi brates, etc.), and lipid binding proteins (PPAR ␣ , L-FABP, etc.) in the nucleus.