CD36 Binds Oxidized Low Density Lipoprotein (LDL) in a Mechanism Dependent upon Fatty Acid Binding*

Background: CD36 is expressed in many cell types and binds diverse ligands including oxidized LDL (oxLDL) and unesterified fatty acid (FA). Results: FA bind to CD36, which leads to oxLDL binding and uptake independent of CD36 disulfide bonds. Conclusion: Typical dietary FA enhance uptake of oxLDL. Significance: This study provides a possible mechanism for oxLDL binding CD36 that is dependent upon specific FA types. The association of unesterified fatty acid (FA) with the scavenger receptor CD36 has been actively researched, with focuses on FA and oxidized low density lipoprotein (oxLDL) uptake. CD36 has been shown to bind FA, but this interaction has been poorly characterized to date. To gain new insights into the physiological relevance of binding of FA to CD36, we characterized FA binding to the ectodomain of CD36 by the biophysical method surface plasmon resonance. Five structurally distinct FAs (saturated, monounsaturated (cis and trans), polyunsaturated, and oxidized) were pulsed across surface plasmon resonance channels, generating association and dissociation binding curves. Except for the oxidized FA HODE, all FAs bound to CD36, with rapid association and dissociation kinetics similar to HSA. Next, to elucidate the role that each FA might play in CD36-mediated oxLDL uptake, we used a fluorescent oxLDL (Dii-oxLDL) live cell assay with confocal microscopy imaging. CD36-mediated uptake in serum-free medium was very low but greatly increased when serum was present. The addition of exogenous FA in serum-free medium increased oxLDL binding and uptake to levels found with serum and affected CD36 plasma membrane distribution. Binding/uptake of oxLDL was dependent upon the FA dose, except for docosahexaenoic acid, which exhibited binding to CD36 but did not activate the uptake of oxLDL. HODE also did not affect oxLDL uptake. High affinity FA binding to CD36 and the effects of each FA on oxLDL uptake have important implications for protein conformation, binding of other ligands, functional properties of CD36, and high plasma FA levels in obesity and type 2 diabetes.

The scavenger receptor family, which includes CD36, has expanded to encompass eight different subclasses of structurally unrelated receptors that, under the classical definition, share the defining feature of binding modified forms of LDL (1,2). The uptake of modified lipoproteins by scavenger receptors is thought to be central to foam cell formation and is widely believed to represent one of the major activation events stimu-lating the proinflammatory phenotype of lesional macrophages (1,3,4). Furthermore, scavenger receptors initiate signaling cascades that regulate macrophage activation, lipid metabolism, and inflammatory programs that may influence the development and stability of atherosclerotic plaques (5,6). CD36 is a prominent member of the scavenger receptor family in part because of its robust catalysis of oxLDL 2 uptake and the CD36specific atheroinflammatory signaling connected to these events (4). Furthermore, its expression in cells is associated with cardiovascular disease, which includes surface expression on platelets, adipocytes, skeletal and cardiac myocytes, capillary endothelial cells, most professional phagocytes, and epithelial cells in gut, kidney, and breast (7). The physiological relevance of oxLDL has been questioned and also debated. Despite mixed results from various randomized controlled trials using antioxidant therapy to treat oxLDL, 14 clinical cohort studies have demonstrated an association between oxLDL measurement and cardiovascular events (8). Furthermore, specifically CD36 and oxLDL levels have recently been shown to be associated with cardiovascular risk factors in young subjects (4).
CD36 binds oxLDL with an affinity as high as or higher than any other currently examined scavenger receptor family members (9,10). Early studies illustrated the significant role of CD36 in binding/uptake of oxLDL based upon a genetic polymorphism in the CD36 gene identified in Japanese subjects shown to result in deficient expression of CD36 (the NAK aϪ phenotype) (11). Monocyte-derived macrophages isolated from these patients were seen to bind ϳ40% less oxLDL and accumulate ϳ40% less cholesteryl ester from oxLDL than cells derived from normal controls (12). The site of this oxLDL binding to CD36 has been identified and recently mapped to amino acids 157-171, with critical lysines at positions 164 and 166 (13)(14)(15). This is consistent with the pH profile and salt concentration dependence, which suggest an electrostatic contribution to the CD36-oxLDL binding (14). Additionally, CD36 is part of the conserved scavenger receptor cysteine-rich superfamily characterized by the presence of one or several repeats of an ancient and highly conserved protein module (16). The three characterized disulfide bonds in the ectodomain of CD36 (17) are active in CD36-mediated cytoadherence of Plasmodium falciparum (malaria)-infected red blood cells (18), indicating potential oxLDL binding to CD36 through disulfide bonds and treatment strategies using reducing strategies such as targeted nanodrugs.
Although the best characterized function of CD36 is to bind and internalize oxLDL, another widely studied role of CD36 is in direct binding of FA (19) and in augmenting FA uptake (20 -22). We have recently shown that CD36 increases FA uptake through intracellular esterification but does not increase mass transport of FA across the plasma membrane, transport that occurs within seconds in cells with or without CD36 (21). This raises questions regarding the mechanism and role of FA binding to CD36.
Initial studies, conducted using the probe sulfo-N-hydroxysuccinimidyl ester of oleate (SSO) and the FA derivative isopropyl myristate, showed apparent inhibition of adipocyte FA uptake, and a subsequent study showed that radiolabeled SSO bound covalently to CD36 (23,24). CD36 was shown to have a significant homology to M-FABP throughout 73% of its sequence, suggesting that a FA binding region was present in the extracellular domain, yet the only reported study of natural FA showed very weak binding to CD36 with apparent dissociation constants of ϳ1 M (19). However, the approaches used had several technical limitations and showed FA to have similar low affinity for albumin, which has well characterized binding constants in the nanomolar range (25). Using SSO as a probe of FA binding, point mutations of CD36 appear to disrupt oxLDL binding to CD36 and inhibit FA uptake in CHO cells (23,26). Furthermore, new studies have indicated that certain FAs inhibit oxLDL binding to synthetic CD36 peptide fragments (27,28). Along with our recent findings, these new studies provide a strong rationale for applying new methods to better characterize FA binding to CD36.
In this study, we applied and optimized a new methodology to study CD36 binding of various label-free FA from five structural groups (oleic acid (OA), elaidic acid (trans-oleic acid; EA), 9S-hydroxy-10E,12Z-octadecadienoic acid (oxidized linoleic acid commonly found in oxLDL (29); HODE), palmitic acid (PA), and docosahexaenoic acid (DHA)). Our chosen method, surface plasmon resonance (SPR), provided quantitative information about affinity, specificity, and kinetics. Next, we sought to provide insight into molecular mechanisms for the binding of CD36 with oxLDL together with each of the above FA, by investigating Dii-oxLDL uptake using a fluoresence binding/ uptake assay and confocal microscopy. Our findings support the proposed electrostatic mechanism of CD36 binding of oxLDL, along with FA acyl chain contributions. Finally, we have shown that binding of CD36 to oxLDL is independent of the CD36 disulfide bonds using DTT and tris(2-carboxyethyl) phosphine (TCEP) reductions.

EXPERIMENTAL PROCEDURES
Cell Culture and DNA Transfection-For all cell culture experiments, low passage number cells were used, purchased from ATCC (Manassas, VA). The cells were grown at 37°C and 5% CO 2 in phenol-free, low glucose (5 mM) DMEM (Invitrogen) supplemented with 10% heat-inactivated FBS (Invitrogen) and 1% penicillin/streptomycin. DNA transfections were done using Lipofectamine2000 (Invitrogen) diluted into unsupplemented Opti-MEM (Invitrogen) using either HeLa cells or human embryonic kidney cells (HEK293) grown to ϳ80% confluence in 60-mm dishes. All DNA used had the same vector backbone (pCI-neo-empty or pCI-neo-CD36), and expression levels and localization were previously verified using Western blotting, microscopy, and fractionation as published (21), except pcDNA3.1-EGFP-CD36, in which CD36 was cloned into a pcDNA3.1-EGFP (pcDNA-3.1 enhanced green fluorescent protein) plasmid (Invitrogen). DNA used in transfections had a final concentration at 0.8 g each DNA per 60-mm dish. All experiments using transfected cells were performed within a 5-day period from day 1 of transfection to maximize protein expression.
Immunofluorescence Staining with Dii-oxLDL and Confocal Microscopy-For Dii-oxLDL uptake experiments, HEK293 cells were grown in 8-well BD Falcon TM culture slides (Franklin Lakes, NJ) to ϳ60% confluence. Dii-oxLDL (1,1Ј-dioctadecyl-3,3,3Ј,3Ј-tetramethylindocarbocyanine perchlorate) from LDL isolated from human plasma and oxidized via copper sulfate oxidation and then analyzed for the degree of oxidation using the thiobarbituric acid reactive substances (TBARS) assay and migration versus native LDL on agarose gel electrophoresis was purchased from Biomedical Technologies Inc. (Stoughton, MA). Dii-oxLDL was also stored at 4°C unfrozen and used fresh within 2 weeks of receiving, per manufacturer's instructions, and was premixed at 5 g/ml into DMEM with 10% FBS, or DMEM with 10% charcoal-stripped FBS (Sigma, catalog no. F6765), or serum-free DMEM with added FA (OA, EA, HODE, PA, and DHA). For metabolic studies, the acyl CoA synthetase inhibitor triacsin C (Enzo Life Sciences, Farmingdale, NY) was additionally added at a final concentration of 24 M as our lab has done previously (30). Prior to being added, all FA was solubilized into Dulbecco's PBS (DPBS) with methyl-␤-cyclodextrin (M␤CD) at molar ratios of 1 FA:6 M␤CD for OA, EA, and HODE and 1 PA:12 M␤CD and 1 DHA:1 M␤CD. Into each cell culture well, 200 l/well of media was added for 90 min at 37°C. All steps after this point were completed under darkness. After Dii-oxLDL incubation, the cells were washed twice in warm DPBS ϩ 0.1% Tween 20 and 1% BSA and then fixed with 4% paraformaldehyde in DPBS at room temperature for 10 min. The cells were washed again with DPBS once, permeabilized with 0.1% Triton X-100 in DPBS, washed twice in DPBS, and blocked in 10% BSA in DPBS for 1 h at room temperature. This blocking was followed by overnight 4°C incubation with monoclonal rabbit anti-CD36 primary antibody (EPR6573, Abcam) or monoclonal mouse anti-CD36 primary antibody (JC63.1, Abcam) at a dilution of 1:100. The cells were then washed three times for 5 min in DPBS buffer and then incubated with Alexa Fluor 488 goat anti-rabbit IgG antibody (Invitrogen, catalog no. A-11008) or Alexa Fluor 488 goat anti-mouse IgG antibody (Invitrogen, catalog no. A-11001) for 1 h at room temperature. After washing with DPBS three times for 5 min, the cells were mounted with ProLong Gold antifade reagent containing DAPI (Invitrogen, catalog no. P-36931) and a #1 glass cover slide. The stained cells, including three-dimensional z-stack images, were photographed using an Olympus DSU Spinning Disk Confocal microscope (Center Valley, PA) using CellSens Acquisition Software, maintaining the gain and exposure settings between each channel for all images. Red Dii-ox-LDL fluorescence was quantified by opening only the red channel images in ImageJ (National Institutes of Health, Bethesda, MD), subtracting background, and using Analyze/Measure to get a total red fluorescence value. This number was normalized by dividing it by the total cell number, obtained by converting the DAPI nuclear channel image to binary format and using the Image-based Tool for Counting Nuclei plugin (31).
Surface Plasmon Resonance Analysis-Using the ProteOn XPR36 protein interaction array system (Bio-Rad) (32), all experiments were performed in high salt PBS (200 mM, pH 7.4) running buffer to decrease nonspecific binding, with or without 5 mM M␤CD to dissociate FA, at a flow rate of 30 l/min at 25°C. ProteOn GLM chips were chosen for the experiment after our first initial trials with GLC chips did not give sufficient signal response. All six channels (GLM sensor chip), including the empty well used to normalize nonspecific binding, were activated for 5 min with a mixture of 0.2 M 1-ethyl-3-(3-dimethylaminopropyl)carbodiimide hydrochloride and 0.05 M sulfo-N-hydroxysuccinimide. Immediately after activation of the surfaces, proteins were added for covalent linking. Channel 1 was used as a reference. For the negative control, normal rabbit IgG (Cell Signaling Inc., Danvers, MA, catalog no. 2729) was diluted into 10 mM sodium acetate buffer, pH 4.5, at 300 l/well. Using 0.25 mg/ml normal rabbit IgG stock solution, 5 l/well yielded ϳ1,300 resonance units (RU), injected across channel 2. Recombinant ectodomain of human CD36 (rCD36) from amino acids 30 -439, i.e. lacking both the transmembrane and the small intracellular domains (Sino Biologicals Inc., Daxing, China), was diluted into 4:1 PBS:10 mM sodium acetate buffer, pH 4.5 (Bio-Rad) at 300 l per/well. PBS was used because the theoretical isoelectric point of CD36 is 8.19 (ExPASy). Experimentally it was found that using 0.25 mg/ml rCD36 stock solution at 3 l/well yielded ϳ1,300 RU, and 12 l/well yielded ϳ5,000 RU in channels 3 and 4, respectively. For a positive control, human serum albumin (HSA, Sigma-Aldrich, catalog no. A3782) was diluted into acetate buffer, pH 4.5, at 300 l/well. Using 0.25 g/ml HSA stock solution, experimentally adding 5 l/well yielded ϳ1,300 RU and 10 l/well yielded ϳ5,000 RU in channels 5 and 6, respectively. Finally, after protein linkage, all six channels were blocked with 1 M ethanolamine HCl (pH 8.5).
Various FA were then injected into channels perpendicular to ligands, at six different concentrations, using a 2-fold dilution series within a variable range, dependent upon experimental FA R max , which differed for each individual FA tested. In additional experiments, Dii-oxLDL was injected at doses of 1, 4, and 8 nM, with or without 150 M OA, using an assigned and approximate molecular mass value for Dii-oxLDL at 3,000,000 g/mol. In these assays, Dii-oxLDL molecular mass was assigned at 3 million daltons (34) for the purpose of Langmuir modeling. The six concentrations were injected simultaneously at a flow rate of 200 l/min at 25°C, unless otherwise noted. The GLM sensor chip was regenerated with short injection of 1 M NaCl between consecutive FA measurements. For Dii-oxLDL regeneration, three separate injections of glycine buffer (pH 3.0), 0.5% SDS, and 1 M NaCl, respectively, were used. The results are expressed in arbitrary RU with subtraction of RU values from channel 1. The data were analyzed with ProteOn Manager TM software, using the Langmuir model (A ϩ B [dharrow] AB) for fitting kinetic data. For each FA, as well as for all Dii-oxLDL SPR experiments, a set of five separate experiments were performed and analyzed. Error bars were not included for visual clarity but were ϽϮ25 RU for each experiment set.
In Vitro Solid Phase CD36 and Dii-oxLDL Binding Assay-CD36 and Dii-oxLDL binding was measured using slight modifications to previously published work (35,36), such as the use of n-dodecyl-␤-D-maltoside (ThermoScientific) in place of octyl ␤-D-glucoside. In brief, rCD36-His 6 purified from HEK293 cells lacking both the transmembrane and the small intracellular domains (Sino Biologicals Inc., Daxing, China) was added at 1 g/well of a 96-well Ni-NTA-coated HisSorb TM plate (Qiagen) as a 100-l aliquot in protein binding buffer (20 mM Tris-HCl, pH 6.8, 150 mM NaCl, 1.5 mM MgCl 2 , 5% (v/v) glycerol, 0.5% (w/v) n-dodecyl-␤-D-maltoside) and bound overnight at 4°C with gentle rocking. Unbound protein was aspirated, and each well was washed with 2 ϫ 200 l of room temperature ligand binding buffer (PBS, 1 mM MgCl 2 , 1 mM CaCl 2 , 0.5% (w/v) n-dodecyl-␤-D-maltoside, 0.2% (w/v) fatty acid-free BSA). After this wash, 10 mM DTT (ThermoScientific) or 10 mM TCEP (ThermoScientific) were added in ligand binding buffer, to select wells, for 5 min at room temperature with gentle rocking and then removed by aspiration. Dii-oxLDL was then added in ligand binding buffer (Ϯ 10 mM DTT or 10 mM TCEP) in a total assay volume of 100 l. The binding reaction was incubated for 90 min at room temperature, with gentle rocking. Unbound ligand was removed, and the wells were washed three times with 150 l of ice-cold wash buffer (PBS, 1 mM MgCl 2 , 0.05% BSA). PBS (50 l) was added per well prior to determining the amount of bound fluorescence using a fluorescence plate reader (Infinite M1000 PRO, Tecan; excitation, 514 nm; emission, 550 nm). Specific binding of Dii-oxLDL was calculated after subtraction of the level of Dii-oxLDL bound nonspecifically to empty wells.
The data from all binding assays were analyzed using Origin-Lab version 8.6 software. Saturation binding data that showed saturable binding of ligand to CD36 over a range of ligand concentrations were best fitted by the Langmuir adsorption isotherm, shown in Equation 1, which describes binding of ligand to a single class of binding site, where B is bound ligand (relative fluorescence units), B max is maximal binding (relative fluorescence units), [L] is concentrations of ligand (g/ml), and K d is the concentration of ligand (g/ml) giving half-maximal binding and is a measure of the affinity of the receptor-ligand interaction.

CD36 Binds Oleic Acid with Comparable Kinetic On/Off Rates to Albumin but with Decreased Affinity and/or Binding
Sites-To study the binding of CD36 to natural FA without potentially perturbing labels, we applied a new methodology (SPR) and a novel approach to solubilize FA for this binding assay. Fully glycosylated recombinant CD36, purified from HEK293 cells without the TM domains, was linked to an SPR sensor chip surface, and various OA doses were pulsed across the chip surface to investigate the potential binding affinity and kinetics of the OA interaction with CD36 ( Fig. 1, Schematic). Because this was a new experimental approach for CD36, we included extensive controls ( Fig. 1, B, C, E, and G). In addition, the design of the experiment necessitated maintaining FA in a soluble form in high concentrations. To overcome solubility limitations and artifacts, we used M␤CD as a delivery vehicle. Our lab has characterized the interactions of several FA with M␤CD and demonstrated that in aqueous buffer, FAs dissociate from M␤CD and partition into lipid bilayers within ms (37,38). Additionally, FA:M␤CD in PBS buffer is more physiologically relevant than typical organic solvents used to solubilize FA Although these numbers may not represent accurate K D values, the relative value is important. Note that when higher amounts of OA were added with CD36, no higher signal was observed (data not shown). G, no significant binding of OA was observed with IgG bound to the chip. The bottom left panel illustrates an N-hydroxysuccinimidyl-activated SPR chip surface (left part) with CD36 mobilized by covalent linkage (middle part) and FA being pulsed through the flow channel containing CD36 (right part). The covalent linkage of CD36 is at amino acids 30 and 439 (full-length protein is 1-472), which is adjacent to the lipid bilayer in the full-length protein. FEBRUARY 20, 2015 • VOLUME 290 • NUMBER 8 (e.g. 100% DMSO). Organic solvents are also incompatible with SPR at high volume percentages because of their high refractive index leading to maximal SPR resonance units. In our SPR experiments, we tested up to 5 mM M␤CD with CD36 linked to the SPR chip surface and saw no significant binding or dose dependence in these controls (Fig. 1, A-C).

Fatty Acid Structure Affects CD36-mediated Uptake of oxLDL
For comparison with our CD36 results, we performed parallel experiments with HSA as a positive control. HSA was covalently linked to the SPR chip surface with the same RU amount bound as CD36 (experimentally determined, as described under "Experimental Procedures"; Fig. 1, B and E). We found that HSA bound and released OA rapidly and with a high RU response (for example, Fig. 1E), which was expected from extensive previous studies of FA binding, including those in our lab with albumin (39,40) and FABP1 (41,42). As an additional negative control, we bound IgG to the chip surface and saw no OA binding, as expected (Fig. 1G). Fig. 1F shows our SPR data for OA binding to CD36 binding with a dose-responsive RU response. There was a large RU increase between 100 and 200 M, relative to the other OA doses, which was also seen in the positive control HSA. The upper range of RU saturation for CD36 was found at ϳ200 M OA, and higher OA doses were found to not significantly increase RU signal beyond 200 M OA concentrations (R max ).
Serum Is Required for Large Quantity CD36-mediated oxLDL Binding-Although our SPR experiment does not directly quantify the number of OA binding sites, the comparison of data for HSA (with nine known sites) and CD36 provides the first biophysical evidence for the presence of several binding sites on CD36. Most of the focus on interactions of CD36 with FA has been on binding of a single FA molecule to a site in the large ectodomain with structural homology FABP and how such binding could promote mass transport of FA through the plasma membrane (20,43). FA binding (either one or several molecules) would not likely be sufficient binding for mass FA transport but could affect binding of LDL, especially with evidence that site-specific CD36 inhibitors have indicated that FA and oxLDL bind CD36 in the same peptide region (26). Consequently, we investigated the physiological significance of this FA binding using an established fluorescence-labeled oxLDL (Dii-oxLDL) binding/uptake assay. We first examined whether FA binding was a prerequisite for oxLDL binding by adding Dii-oxLDL to HEK293 cells expressing CD36 (pCI-neo-CD36) in either DMEM with 10% FBS, which contains a mixture of biological FA, or DMEM without serum (Fig. 2). HEK293 cells expressing empty vector control (pCI-neo-empty) were used as a negative control (Fig. 2, u-x and C).
As shown in Fig. 2 (ad), robust binding of Dii-oxLDL occurred in the presence of serum, but very little binding was detected when serum was not present in the assay (Fig. 2, e-h). Additionally, when 10% FBS was used that had been charcoalstripped of hydrophobic molecules (i.e. FA, hormones, etc.), the Dii-oxLDL binding was again nearly completely eliminated (Fig. 2, i-l). Because oxLDL purified from human plasma contains enough FA to affect LDL electronegativity (44), it is possible that the observed small amount of basal Dii-oxLDL binding may be accomplished through FA already present on oxLDL.
Oleic Acid Enhances CD36-mediated oxLDL Binding-FBS has been shown to contain a ratio of saturated, monounsaturated, and polyunsaturated FA (34.0, 34.3, and 27.8%, respectively), with almost 30% of the total FA composition being comprised of OA (45). Because SPR demonstrated that CD36 binds OA, we added OA to Dii-oxLDL in serum-free DMEM to HEK293 cells expressing CD36 (Fig. 2, m-t) or to empty control vector (Fig. 2, u-x). Not only was Dii-oxLDL binding rescued with OA added to the serum-free medium, the surface expression of CD36 was altered at higher doses of OA, and binding became even greater than with serum (Fig. 2, q-t, A, and B).
Having demonstrated binding of oxLDL to the surface receptor CD36 by microscopy, it is important to determine whether Dii-oxLDL was being internalized by the cells expressing CD36 (uptake) or whether Dii-oxLDL was only binding CD36 at the plasma membrane, where CD36 is predominantly localized in our HEK293 cell system (21). To discriminate these two possibilities, we expressed GFP-CD36 and used three-dimensional confocal microscopy to visualize the Dii-oxLDL location in the z-plane relative to CD36 localized at the cell surface and relative to the cell nuclei (Fig. 3, A-C). The merged image panel shows that most of the Dii-oxLDL bound by cells expressing GFP-CD36 (Fig. 3C, yellow) was internalized, indicating that binding to CD36 was followed by oxLDL uptake when OA is present, within the 90-min incubation. In addition, because we have recently shown that CD36 is involved in activating and therefore traps FA inside cells by enhancing esterification (21), we inhibited the first step of FA activation by using the acyl-CoA synthetase inhibitor, triacsin C, for the 90-min incubation of Dii-oxLDL and OA and found no significant difference with or without this inhibitor (see Fig. 5, A-D).
Next, to focus directly on FA effects of oxLDL binding to CD36, we performed additional in vitro SPR experiments in which Dii-oxLDL was pulsed across the SRP chip surface containing bound CD36 in the presence and absence of OA. Here we observed significantly more binding of Dii-oxLDL to CD36 when 150 M OA was present along with Dii-oxLDL than with Dii-oxLDL alone and very little binding of oxLDL to the negative control, IgG (Fig. 3, D-F). For example, when 4 nM oxLDL by itself is pulsed across CD36, the RU of binding is ϳ130 (Fig.  3E, green line). When 4 nM oxLDL with 150 M OA:M␤CD is pulsed across CD36, ϳ400 RU of binding is observed (Fig. 3F, green line), even though OA:M␤CD by itself only adds ϳ100 RU (expected RU of binding: 130 RU ϩ 100 RU ϭ 230 RU; observed ϭ ϳ400 RU). This indicates that OA may be opening/ facilitating an oxLDL binding site on CD36, to increase oxLDL binding. Furthermore, because human source oxLDL contains FA, it is possible that the FA present on oxLDL is enabling the binding to CD36 seen when additional OA is not added (Figs. 3E and 2, e-l). Experiments with FA-free oxLDL would need to be performed in future studies to determine whether this is the case.
The trans-Isomer of Oleic Acid (Elaidic Acid) Binds to CD36 and Increases Dii-oxLDL Uptake-The OA trans-isomer, EA, commonly formed by the partial hydrogenation of vegetable oils and commonly consumed by humans, also bound both CD36 and HSA but with slower and possibly biphasic kinetics of adsorption (Fig. 4, A and E). Along with each experiment using HSA and CD36, we included IgG, but under no condition was IgG seen to bind any FA above Ϯ 10 RU, indicating that FA binding was specific only to HSA and CD36 (Fig. 1G shown as representative, other control data not shown).
We then tested whether EA by itself increased Dii-oxLDL uptake and found uptake to be even more pronounced than OA at much lower doses, with doses higher than ϳ100 M showing no additional uptake (ϳR max of Dii-oxLDL uptake). Interestingly, the presence of EA even at lower doses appears to drastically affect the CD36 distribution on the cell surface, especially when compared alongside other FA types (Fig. 5, a- (29), we used SPR to test whether CD36 could bind HODE. Unexpectedly, we observed dose-responsive binding of HODE with HSA but no binding to CD36 (Fig. 4, B and F). When HODE was included in the serum-free Dii-oxLDL uptake assay, no increase in Dii-oxLDL uptake was observed in doses up to 100 M (Fig. 5, panels e-h). Doses of HODE above 100 M were tested but led to a dramatic loss of cytoadherence and no noticeable Dii-ox-LDL uptake difference (high dose data not shown; low n per microscopy field).

Saturated Fatty Acid Palmitic Acid Binds CD36 and Increases Dii-oxLDL Uptake, Whereas the Polyunsaturated DHA Binds CD36 but Does Not Increase Dii-oxLDL Uptake-
To investigate other physiologically important FA classes, we selected saturated PA and polyunsaturated DHA for SPR and Dii-oxLDL uptake studies. We found that CD36 bound both DHA and PA, with particularly high binding responses with DHA before reaching saturating doses (Fig. 4H).
Despite the high binding response of DHA with CD36, DHA did not increase Dii-oxLDL uptake, even at the high dose of 500 M DHA in serum-free medium (Fig. 5, panels m-p). This suggests that DHA is binding CD36 while inhibiting oxLDL uptake, FIGURE 2. Panels a-x, merged 20ϫ confocal micrographs of HEK293 cells, transfected with pCI-neo-CD36 (panels a-t) or pCI-neo-empty vector control (panels u-x), with 5 g/ml Dii-oxLDL (shown in red in the merged image) added for 90 min at 37°C in DMEM with 10% FBS (panels a-d), DMEM without serum (panels e-h), DMEM with 10% charcoal-stripped FBS (panels i-l), or serum-free DMEM with only OA added back (panels m-x). After Dii-oxLDL incubation, the cells were fixed and stained for CD36 (shown in green in the merged image) and the nuclear stain DAPI (shown in blue in the merged image). Under different conditions, CD36 surface distribution appears to change (panels a, e, i, m, and q) especially with or without serum (panel a versus panel e) or when increasing doses of OA were added (panel m versus panel q). White arrows in each merged image (panels d, h, l, p, t, and x) indicate representative enlarged cells in the original color scheme, shown on the small right panels. All OA added was complexed to M␤CD (1 OA:6 M␤CD), as in SPR experiments. A and B show quantification of (Dii-oxLDL) from Fig. 2 (panels a-l), after background subtractions, in a 20ϫ magnified, 200-cell field (A) and quantification of red fluorescence pixels where doses of OA were added (panels m-x, 1 mM OA addition confocal images not shown) (B), where R max is defined as ϳ500 M. C, additional control merged confocal micrograph of HEK293 cells transfected with pCI-neo-empty control vector, with 5 g/ml Dii-oxLDL (red) added for 90 min at 37°C in DMEM with 10% FBS. The cells were then fixed and stained for CD36 (green) and DAPI (blue). FEBRUARY 20, 2015 • VOLUME 290 • NUMBER 8

JOURNAL OF BIOLOGICAL CHEMISTRY 4595
which is consistent with other studies showing DHA decreases oxLDL uptake (46). Furthermore, whereas PA showed binding saturation to CD36 at relatively low doses on SPR compared with other FA classes that were investigated (Fig. 4G), PA increased Dii-ox-LDL uptake with dose saturation comparable with OA (saturation at ϳ500 M PA), while again appearing to change the CD36 surface distribution at higher doses (Fig. 5, panels i-l, and data not shown).
CD36 Binds oxLDL Independently of Its Disulfide Bonds-In extended studies to gain insight into localizing FA and oxLDL binding on CD36, we performed a recombinant CD36 Dii-ox-LDL binding assay, previously applied to study CD36 binding to various modified forms of LDL, including oxLDL (35). However, we included a heretofore uncharacterized disulfide bond reduction of CD36, employing both the commonly used reducing agent DTT, as well as TCEP. TCEP has the advantage of being a more powerful reducing agent, is considered irreversible by leaving two free cysteines, and, most importantly, TCEP does not reduce metals such as Ni-NTA, used here to immobilize CD36-His 6 .
Even under strong reducing conditions, no difference in Dii-oxLDL binding to CD36 was observed (Fig. 6), indicating that the CD36 disulfides are not required for oxLDL binding and may not be a useful drug target for potentially inhibiting oxLDL uptake by CD36. Additionally, our K D (K D ϭ K a /K d ) values for each of the three conditions (ϳ25 g/ml Ϯ ϳ8) were similar to previously published values for oxLDL binding CD36-His 12 (K D ϭ 10.44 Ϯ 0.08 g/ml (35)), despite our use of dodecyl-␤maltoside in place of octylglucoside in the ligand binding buffer In the merged image, arrow 1 illustrates CD36 (green) localized on the cell surface, and arrow 2 indicates Dii-oxLDL and CD36 (yellow/red) taken into the cell at the nuclear level, indicating uptake rather than cell surface binding. D is negative control IgG bound to a GLM SPR chip, whereas E-G show CD36 bound to a GLM chip with comparable RU amounts to IgG bound. For E and F, doses of 0, 1, 4, and 8 nM of oxLDL were pulsed for 60s in PBS, followed by a PBS wash without Dii-oxLDL. In F, 150 M OA:M␤CD (1 OA:6 M␤CD) was added to each dose of Dii-oxLDL. G is from Fig. 1F, to contrast the effects of OA:M␤CD added without oxLDL. Similarly, in F, the arrow highlights the 0 nM oxLDL dose and a binding response caused by the presence of 150 M OA:M␤CD. In each panel, note the difference in scale. Also note the RU change from oxLDL alone in E is ϳ30, ϳ130, and ϳ250 for 1, 4, and 8 nM oxLDL, respectively. With OA:M␤CD, an additional ϳ100 RU is expected, but instead RU change in F is ϳ200, ϳ400, and ϳ500 for 1, 4, and 8 nM, respectively. D, only slight binding of oxLDL was observed with IgG bound to the chip. Significant binding of oxLDL is seen with CD36 (B); K a ϭ 1.0 e 3 1/ms, K d ϭ 1.7 e Ϫ4 1/s, and K D ϭ 1.7 e Ϫ7 M. Increased binding of oxLDL is seen with CD36 with 150 M OA present (C); K a ϭ 1.0 e 3 1/ms, K d ϭ 4.4 e Ϫ4 1/s, and K D ϭ 4.4 e Ϫ7 m. As with SPR and FA, these numbers may not represent accurate K D values, but the relative value is important.  FEBRUARY 20, 2015 • VOLUME 290 • NUMBER 8 and the wash buffer. Homology modeling of CD36, based on the crystal structure of family member LIMP2, indicates that the three disulfide bonds of CD36 reside on one "face" of the CD36 protein (15), so we infer that FA and oxLDL bind to CD36 on the other face of its tertiary structure (Fig. 7).

Fatty Acid Binding: Novel Findings from Surface Plasmon
Resonance-In very recent studies, biophysical approaches have yielded valuable new insights into the structure of CD36 and the molecular interactions of FA analogues in the extracellular region with homology to FABP (15,47). Throughout the long history of CD36 research, the hypothesis of a single FA binding site has been proposed as a potential mechanism for transport of FA across the plasma membrane into cells (48), as a taste bud sensor of FA (49,50), and/or as an important site for uptake of oxLDL (51). Our recent study combining biophysics, cell biology, and metabolism in HEK293 cells with and without CD36 led to the finding that although CD36 does not enhance FA transport across the plasma membrane, its presence enhances intracellular triglyceride accumulation (21). New work by other investigators has led to additional evidence of the involvement of a single FA binding site in oxLDL uptake (26).
Here we show that a new approach to study FA binding to CD36 (SPR) yielded the novel finding that the ectodomain of CD36 likely binds several molecules of FA, which could have important ramifications beyond the binding of a single FA molecule for CD36 function. The experimental design (Fig. 1, Schematic), including the use of cyclodextrin delivery, permitted quantitative and direct evaluation of the binding of natural FA without the use of high concentrations (up to 1 mM (52)) of nonphysiological probes such as SSO and/or without separation procedures such as column chromatography (e.g. Lipidx resin (19)). SPR showed that CD36 binds structurally different long chain FA, including OA, DHA, EA, and PA (monounsatu-  -p), all complexed to M␤CD as in SPR experiments. Higher doses of HODE were found to abolish cytoadherence, and higher EA doses did not increase Dii-oxLDL uptake (data not shown). PA was found to increase Dii-oxLDL up to ϳ500 M (data not shown), similar to the Dii-oxLDL R max when OA was present (Fig. 2m-2t). After Dii-oxLDL incubation, the cells were fixed and stained for CD36 (shown in green in the merged images) and DAPI (shown in blue in the merged images). Significant Dii-oxLDL binding is only seen with added EA (panels a-d) and PA (panels i-l) but not with HODE (panels e-h) or DHA (panels m-p). Additionally, CD36 surface distribution changes with each FA (panels a, e, i, and m), especially with EA (panel a). Furthermore, Dii-oxLDL uptake was substantial and highly diffuse at the cell surface when EA was added (panel j and enlarged cell example). White arrows in each merged image (panels d, h, l, and p) indicate representative enlarged cells in the original color scheme, shown on the small right panels. In A-D, a similar experiment was performed using serum-free DMEM with 100 M OA added and 5 g/ml Dii-oxLDL, along with 24 M triacsin C for the duration of this 90-min Dii-oxLDL and OA incubation.  concentrations of 0, 1, 2. 5, 5, 10, 15, 20, 25, 30, 35, or 40 g/ml for 90 min at room temperature with no disulfide reduction (black squares), 10 mM DTT added (red circles), or 10 mM TCEP added (blue triangles). Both DTT and TCEP were added in oxLDL binding buffer for 5 min and then aspirated just prior to adding doses of Dii-oxLDL in oxLDL binding buffer. Binding curves were depicted with relative fluorescence units for ligand binding plotted as a function of ligand concentration. Curve fitting was conducted as described under "Experimental Procedures," from which values for K D and R max were determined. CD36 with no reduction bound oxLDL with a K D ϭ 27.9 Ϯ 2.1 g/ml; R max ϭ 8.2 Ϯ 0.3 g/ml. CD36 with added DTT bound oxLDL with a K D ϭ 29.2 Ϯ 8.6 g/ml; R max ϭ 8.7 Ϯ 1.2 g/ml. CD36 with added TCEP bound oxLDL with a K D ϭ 22.8 Ϯ 4.6 g/ml; R max ϭ 8.2 Ϯ 0 g/ml. Three independent experiments were performed for each data set.  (Table 1).
In comparison with CD36, FA interactions with albumin have been characterized in detail by many biophysical methods, including NMR spectroscopy and x-ray crystallography approaches (40). Nine individual sites on HSA have been recently been detected by two-dimensional NMR, and several NMR peaks were correlated with crystallographic sites well characterized for long chain FA (39). Binding constants measured by two-dimensional NMR on a site-specific basis varied only by a factor of 10 between the highest and lowest affinity sites, with affinity constant (K D ) values similar to those reported by earlier studies (note: The high affinity FA binding to CD36 has the important implication that binding at some or all sites could affect protein conformation. The similarity of the kinetics of dissociation of FA from albumin and CD36 observed in our experiments carries the implication that FA have a long enough residency time to have effects on protein structure but also a short enough residency time to allow rapid equilibration between the surrounding aqueous environment, the plasma membrane, and albumin in the plasma (Fig. 7A). By contrast, low affinity binding with weak interactions, as suggested by some previous studies (19), would not likely affect protein structure and function or equilibration among the different pools of FA. Furthermore, the amount of FA bound (mol/mol of CD36) could alter protein conformation (Fig. 7B), as has been observed for HSA (40), and the apparent bi-phasic binding curve, seen with EA especially, indicates that this is the case. Thus, the existence of these multiple sites could provide a stronger sensor for extracellular and membrane-bound FA concentrations than would a single site.
Effects of Fatty Acid on oxLDL Uptake by CD36 -To study effects of FA on CD36-mediated uptake of oxLDL, we utilized our HEK293 cell line designed to express CD36 in the glycosylated form at levels comparable with its endogenous expression in adipocytes (21). As with our previous studies of CD36-mediated uptake of FA, HEK293 cells offered advantages of low or absent endogenous expression of other purported FA transporters (including CD36) (21) and slow metabolism of exogenous FA (53), which permitted focus on mechanisms related to CD36. For the present study, these HEK293 cells are also advantageous because they lack multiple scavenger receptors found on adipocytes and most other cell types, as recognized in other studies using HEK293 cells (54,55). Our rationale is further FIGURE 7. A, dissociation of long chain FA from albumin and phospholipid bilayers occurs in a time of milliseconds to seconds (33), and we expect generally similar kinetics for the CD36 binding sites, especially with our SPR data. The high abundance of the plasma membrane must be taken in account to determine the partitioning of FA between proteins such as CD36 and the surrounding membrane. B, using homology modeling of CD36 from the recently published crystal structure of CD36 family member LIMP-2 (15), we designed this schematic showing a possible mechanism for oxLDL binding, after FA opens the binding site, based on combining our FA binding studies (SPR) and Dii-oxLDL binding/uptake data (microscopy). The site of FA and oxLDL binding is presumably on the opposite face of the CD36 protein structure (left side, because the CD36 protein is oriented in the schematic) from the disulfide bond-rich region (right side), because we found disulfide bond reduction does not alter oxLDL binding. LDL is shown in the upper background with a single apolipoprotein B-100 wrapping the particle, and white dots on the LDL indicate oxidation. One of the implications of the newly characterized FA sites on CD36 is that as the concentration of FA in the plasma and/or plasma membrane increases, the sites on CD36 will have high occupancy, further increasing oxLDL binding and uptake. The figure is not drawn to scale.

TABLE 1 A summary of the FA we tested that bound CD36
Plus signs indicate binding of the FA to CD36. Similarly, uptake of Dii-oxLDL in live HEK293 cells is indicated with plus sign. In general, when a FA bound CD36, uptake of Dii-oxLDL ensued, except for DHA. HODE showed no binding (Ϫ) and no enhancement of oxLDL uptake (Ϫ).

Fatty acid CD36 binding Increased oxLDL uptake
Oleic acid ϩ ϩ 9(S)-HODE Ϫ Ϫ Elaidic acid ϩ ϩ Palmitic acid ϩ ϩ DHA ϩ Ϫ Fatty Acid Structure Affects CD36-mediated Uptake of oxLDL FEBRUARY 20, 2015 • VOLUME 290 • NUMBER 8 supported by the complication that uptake of modified lipoproteins by scavenger receptors is thought to be central to foam cell formation (56), which has previously been observed to occur over a relatively long time period (1-3 h) (55) and could confound our observation of binding and internalization of oxLDL. Studies of cells are also complicated because FA bind to the lipid bilayer, and most of the added FA may bind (partition) to the plasma membrane surrounding CD36 (Fig. 7A). The presence of the cell membrane must be considered in interpreting the effects of added FA on CD36-mediated uptake of oxLDL. The rapid dissociation of FA from the lipid bilayer and plasma albumin will permit binding to sites on CD36 by equilibration, as illustrated in Fig. 7A. We can assume that FA entering the membrane (even from an intracellular origin) or in the proximity of the membrane will bind to sites on CD36, and we can predict that higher concentrations of FA will lead to greater occupancy of CD36 sites. Furthermore, from our cell-free binding studies with SPR, we can predict that effects of FA on CD36 might be differentiated from nonspecific effects by dependence on the FA structure.
Our major findings in our HEK293 cells are that: (i) oxLDL uptake was limited in serum-free medium without added FA and in charcoal-stripped FA-depleted serum; (ii) addition of exogenous OA to serum-free medium enhanced oxLDL uptake to a level observed with serum; (iii) PA and EA also enhanced oxLDL in a dose-dependent manner; (iv) DHA did not enhance CD36-mediated uptake of oxLDL, although it bound to CD36 in our SPR assay; and (v) HODE did not enhance uptake of oxLDL and did not bind to CD36 in our SPR assay. These findings are summarized in Table 1.
The work presented is the first systematic study of the effects of FA together with oxLDL on CD36. Although published results to date from other groups have used similar assays with labeled oxLDL in cells expressing CD36, they have not shown dependence of oxLDL binding upon serum and/or FA. This is likely due to different experimental conditions and because the studies were not comprehensive. For example, other studies have used monocyte-like or macrophage-like cell lines that express other compensating oxLDL scavenger receptors that likely bind oxLDL in a FA-independent manner (e.g. THP-1 cells (57)). Additionally, compared with our study that used high doses of FA for in vitro experiments (up to 500 M to observe the R max of Dii-oxLDL uptake), physiological FA concentrations range between 250 and 3,000 M, dependent on factors such as fasting/nutrition state and exercise (58). Lower levels of oxLDL uptake can be seen with highly sensitive oxLDL detection techniques such as radiolabeling ( 125 I-oxLDL) or high magnification/high exposure microscopy to observe oxLDL binding/uptake, coupled with serum-free medium delivery of oxLDL (15,59,60). We quantified the differences between Dii-oxLDL uptake using serum-free medium (small levels of uptake), serumcontaining medium (large levels of uptake), and increasing doses of FA in serum-free medium (increasing levels of Dii-oxLDL uptake, rescued to serum-containing levels), and we did not observe any unusual cell phenotype changes, in the short time frame of our observation (Fig. 2). Similarly, although OA and PA have been found to increase CD36 surface expression (35), our experimental time scale likely does not allow enough time for increases in surface expression of CD36 to drastically alter oxLDL uptake. However, our finding that CD36 surface distribution was changed using different doses of FA is of interest and an aim in our future work. We additionally utilized ethanol-free delivery of all FA in these assays, namely using M␤CD and water, to eliminate possible solvent perturbations in the binding and uptake of Dii-oxLDL.
Implications of Our Uptake Studies for Protein Structure: Disulfide Bonds of CD36 -CD36 mediated cytoadherence of P. falciparum (malaria)-infected red blood cells is lost following treatment of CD36-expressing cells with the reducing agent DTT (18). This finding indicates therapeutic potential in reducing agents, such as the nontoxic drug N-acetylcysteine, to prevent or treat malaria complications caused by infected red blood cell cytoadhesion. However, these same treatment strategies may not be useful in preventing oxLDL binding and internalization through a CD36-dependent mechanism, because we observed that disulfide reduction did not affect Dii-oxLDL binding to CD36 (Fig. 6). This also suggests that the FA and oxLDL binding site on CD36 is not located in the cysteine-rich region that interacts with P. falciparum-infected red blood cells, but possibly on the other face of CD36, as represented by the bound OA in Fig. 7B.
Evidence for Direct Mechanistic Effects of FA on oxLDL Uptake by CD36 -Because CD36 binds and initiates uptake of oxLDL and we found FA to be required for this CD36-dependent mass-transport of oxLDL, we speculate that FA binding alters the protein conformation with functional consequences. The structure LIMP-2 (15) provides a very plausible mechanism for homologous family member CD36 uptake of cholesteryl esters through a hydrophobic groove. oxLDL fits into a large central cavity homologous to the LIMP-2 cavity for glucocerebrosidase. We speculate that CD36 maintains a "closed" conformation to oxLDL binding until a specific FA (i.e. OA, EA, and PA) is present. FA binding could shift CD36 into an "open" conformation that is favorable to oxLDL binding (Fig. 7B). The one well characterized FA binding site and the oxLDL binding site appear to be in the same region, specifically K164 (26). Binding to this site and other sites could influence oxLDL binding and contribute to opening or closing the binding pocket.
The contribution of direct FA binding enhancement of Dii-oxLDL docking to CD36 is supported by our in vitro SPR results, pulsing oxLDL together with OA (Fig. 3, D-F). This interpretation is consistent with recent observations that fragments of the CD36 protein on multiwall plates had up to 50% less binding of oxLDL when certain FA (including OA) were present at extremely high (ϳ1 mM) concentrations, levels that outcompete the oxLDL at this (single) oxLDL binding/docking site. Other studies of CD36 binding or "scavenging" diverse lipid ligands also indicate a pH profile and salt concentration dependence, suggestive of an electrostatic mechanism of lipid binding (14). Because the FA in these assays is mainly deprotonated in the extracellular environment (i.e. OA pK a ϭ 9.85, etc. (61)), FA could facilitate oxLDL docking through anionic interactions with CD36. Hydrophobic interactions with the FA acyl chain appear to play a contribution, because the two least hydrophobic FA studied either do not bind to CD36 (HODE) or do not enhance uptake of oxLDL (DHA). Because DHA binds avidly to CD36 (SPR shows higher binding than to HSA; Fig. 4, D and H), it could promote an open oxLDL binding conformation on CD36 unfavorable for the binding of oxLDL. This finding agrees with previous data showing that DHA lowers oxLDL uptake (62). Conversely, it is interesting to note that EA is the most abundant trans-FA found in atheromatous plaque (63,64), and we found that EA bound CD36 on SPR and aggressively increased Dii-oxLDL.
Conclusions-Overall, our study provides a novel mechanism of CD36 binding oxLDL, through first binding FA. Our data also indicate that multiple FA binding sites likely exist on CD36 but did not identify specific sites within the protein. Future studies using hydrogen exchange mass spectroscopy with recombinant, full-length CD36 embedded into nanodiscs with added FA are being pursued (65,66), as well as NMR studies using the soluble CD36 ectodomain. In addition, we recognize that because all FA species studied can partition into the lipid bilayer of the plasma membrane, FA could exert effects on membrane organization, which could also influence oxLDL uptake. We have initiated systematic studies with membrane fluidity probes in our HEK293 cells, with and without CD36, to investigate changes in membrane order with different FA.
The presence of multiple binding sites for FA on CD36 could enhance the sensitivity of CD36 to the concentration of plasma FA. Because obesity and type 2 diabetes are characterized by high plasmas levels of long chain FA (67,68), the demonstrated enhancement of oxLDL uptake by increases in common dietary FA may contribute to the pathophysiology of these diseases. Our results also support the hypothesis that CD36 may be a link between insulin resistance states from studies showing that increased plasma FA levels were associated with enhanced CD36 expression and increased internalization of oxLDL (69).
The anomalous and intriguing results for DHA and HODE may have implications for cardiovascular disease because oxLDL uptake triggers a signaling response that is pro-inflammatory and pro-atherogenic (7). DHA might inhibit inflammatory responses by binding to CD36 in a manner that inhibits CD36-oxLDL binding. And although human oxLDL often contains significant amounts of HODE (29,70), HODE did not enhance CD36-mediated uptake of oxLDL, possibly because it does not bind to the ectodomain of CD36. Therefore, other proteins could be affected by HODE, because HODE consumption, in association with a high fat diet, has been shown to increase aortic lesion areas by Ͼ100% (62).