INTRODUCTION

CD36 is an 88-kDa membrane glycoprotein originally found in human platelets, where it was designated glycoprotein IV (1). It has been identified in a variety of cell types including monocytes (2), epithelial cells (3), endothelial cells (4), and various cultured cell lines (5). A large number of potential functions have been ascribed to CD36, with evidence implicating CD36 as a receptor for thrombospondin (6), collagen (7), oxidized low density lipoprotein (8, 9), fatty acids (10), anionic phospholipids (11), and Plasmodium falciparum malaria parasitized erythrocytes (12). In addition, crosslinking of CD36 activates a signal transduction pathway (13) and CD36 was found to associate with Src family nonreceptor protein tyrosine kinases Fyn, Lyn, and Yes (14). CD36 with its homologs including the scavenger receptor SR-BI (15) (also known as CLA-1; Ref. 16) that was recently identified as a receptor for high density lipoprotein (17), the lysosomal membrane protein LIMP II (18), and the Drosophila epithelial membrane protein Emp (19) have been recognized as a novel gene family.

A recent study showed that the presumed rat homolog of CD36 is palmitoylated (20), although the site of the palmitoylation is not known (these authors suggested it to be exofacial). Palmitoylation involves the covalent, posttranslational attachment of the 16-carbon saturated fatty acid palmitate through a thioester linkage to cysteine residues (21, 22). Unlike proteins modified with the 14-carbon saturated fatty acid myristate, which are found both in the cytoplasm and in the membrane, palmitoylated proteins are almost exclusively associated with the cytoplasmic face of the membrane (21, 22). Palmitoylated proteins include G protein-linked receptors, the -subunits of heterotrimeric G proteins, and the Src family nonreceptor protein tyrosine kinases (see Ref. 23 and references therein). The attachment of palmitate to proteins may play a key role in protein targeting to membranes and in protein-protein and protein-lipid interactions. Of great potential significance, the attachment of palmitate to proteins is reversible, and thus it may play a role in regulation of their subcellular localization and function, as recently demonstrated for the -subunit of G_s (24). We thus initiated our investigation to better understand the molecular structure and the nature of palmitoylation of CD36.

MATERIALS AND METHODS

Human embryonic kidney 293 cells were maintained in Iscove's medium (Life Technologies, Inc.) supplemented with 10% fetal bovine serum, 2 mM L-glutamine, 10 µg/ml streptomycin, and 10 units/ml penicillin in a 5% CO₂, 95% air atmosphere at 37 °C. The mouse monoclonal antibody FA6-152 to CD36 was obtained from Immunotech (Westbrook, ME), and the mouse monoclonal antibody M2 to the FLAG epitope tag was from Eastman Kodak Co. [9,10-³H]Palmitic acid (40-60 Ci/mmol) was from either DuPont NEN or Amersham Corp.

Human CD36 cDNA (12) was provided by Dr. Brian Seed. It was ligated into the expression vector pcDNA3 (Invitrogen, San Diego, CA) after being digested with XhoI and XbaI. Mutations were made by polymerase chain reaction, with all DNA sequences confirmed by DNA sequence analysis. Cysteines were changed to serines using oligonucleotide primers encoding the appropriate mutations and wild-type plasmid as template. N-terminal mutations were made by replacing the fragment covering the start codon through the downstream BamHI site in CD36 cDNA. C-terminal mutations were made by replacing the fragment covering the stop codon through the upstream Bsg I site in CD36 cDNA. Wild-type CD36 is designated [CCCC] based on cysteine residues at positions 3, 7, 464, and 466, and the mutants are labeled based on cysteine to serine mutations at the indicated positions (Table I). The C-terminal truncation mutant (CD36CT) was made by introducing a stop codon just upstream of the C-terminal hydrophobic domain (Table I). The FLAG epitope tag (DYKDDDDK) was introduced at the C terminus of CD36 and CD36CT (after amino acid 472 in CD36 or 439 in CD36CT). DNA transfection was performed with Lipofectamine (Life Technologies, Inc.) for 2 h at 37 °C. The cells were selected 48 h after transfection in medium containing 250 µg/ml G418.

Table I. Amino acid sequences of CD36 and mutants constructs

5 × 10⁶ cells were washed with serum-free Iscove's medium and labeled with 0.3-0.5 mCi/ml [9,10-³H]palmitic acid for 1 h at 37 °C. Cells were lysed in lysis buffer (50 mM Tris-HCl, pH 7.6, 150 mM NaCl, 60 mM octylglucoside, 5 mM EDTA, 1 mM sodium orthovanadate, 10 mM iodoacetamide, 10 µg/ml leupeptin, and 10 µg/ml aprotinin) for 30 min. Lysates were clarified by centrifugation at 16,000 × g for 15 min at 4 °C. The supernatants were precleared with secondary antibody and the immunoabsorbent Pansorbin (Calbiochem, San Diego, CA). CD36 was immunoprecipitated with a mouse monoclonal antibody (FA6-152) and rabbit anti-mouse IgG, followed by Pansorbin. Immunoprecipitates were washed three times in lysis buffer and eluted into Laemmli sample buffer and analyzed by SDS-PAGE¹ on 9% gels. All fluorograph films were exposed for 10-14 days.

Cell lysates were separated by SDS-PAGE, with protein content for loading measured by the bicinchoninic acid assay (Pierce) using bovine serum albumin as a standard. Proteins were transferred to 0.45-µm nitrocellulose membrane (Micron Separations, Westboro, MA) and blocked for 30 min in blocking buffer (Tris-buffered saline, pH 7.6, 0.05% Tween, and 3% nonfat dry milk). After a 60-min incubation with primary antibody diluted in blocking buffer followed by washing, the blot was incubated for 30 min with appropriate secondary anti-IgG-horseradish peroxidase conjugate. The membrane was washed three times for 10 min each in Tris-buffered saline with addition of 0.05% Tween 20 (Sigma), and developed with ECL chemiluminescent substrate (Amersham).

1 × 10⁶ cells were washed twice with 5 ml phosphate-buffered saline (PBS, pH 7.4). The cells were then incubated with primary antibody (FA6-152 at 12 µg/ml) in FACS buffer (PBS supplemented with 1% fetal calf serum, 2% bovine serum albumin, 2 mM EDTA) for 1 h at 4 °C. After two washes in PBS, cells were incubated with goat anti-mouse IgG fluorescein isothiocyanate conjugate (Sigma). After three washes with 0.5 ml of FACS buffer each, the expression level of CD36 and its variants was assessed by flow cytometry carried out on a Becton-Dickinson Facscan using the CellQuest software.

Cell surface proteins were biotinylated with sulfosuccinimidyl-6-(biotinamido)hexanoate from Pierce at a concentration of 100 µg/ml for 2 h at 4 °C. Following cell lysis, immunoprecipitation, gel electrophoresis, and blotting, CD36 was identified by avidin coupled with biotinylated horseradish peroxidase (Vectastain Elite ABC Kit, Vector Laboratories, Burlingame, CA), followed by ECL.

1 × 10⁶ cells were washed twice with 5 ml of PBS. The cell pellet was subjected to hypotonic swelling for 20 min in 10 mM Tris-HCl (pH 7.4) supplemented with 10 µg/ml leupeptin and 10 µg/ml aprotinin, followed by 50 passes in a Dounce homogenizer. Intact cells were removed by centrifugation at 500 × g for 10 min. The resulting supernatant was spun at 100,000 × g for 2 h in a Beckmann ultracentrifuge. The pellet (membrane) was suspended in 1 × SDS-PAGE sample buffer, and the supernatant (cytosol) was mixed with 3 × SDS-PAGE sample buffer. Equal percentages of membrane and cytosol were analyzed by SDS-PAGE and Western blot in order to determine the subcellular distribution of wild-type or mutant CD36.

RESULTS

A recent study by Jochen and Hays showed that rat CD36 is palmitoylated (20). We started our investigation by asking whether or not human CD36 incorporates palmitate. When C32 cells (a human melanoma cell line that expresses CD36) were incubated with [9,10-³H]palmitic acid, an 88-kDa protein recognizable by anti-CD36 antibody (FA6-152) was labeled (data not shown), suggesting that human CD36 is palmitoylated. In order to further investigate the nature of this modification, we transfected 293 cells with an expression vector containing the entire coding region of human CD36 cDNA. The CD36-transfected cells expressed CD36 on the cell surface as assessed by flow cytometry performed using the anti-CD36 monoclonal antibody FA6-152 (Fig. 1A, bottom right panel). Furthermore, Western blotting showed an 88-kDa protein using the same anti-CD36 antibody, as expected (Fig. 1B).

After CD36-transfected 293 cells were incubated with [9,10-³H]palmitic acid for 1 h, an anti-CD36 immunoprecipitate of the cell lysate was analyzed by SDS-PAGE. An 88-kDa radiolabeled band was detected by fluorography (Fig. 2), indicating that CD36 is palmitoylated. Palmitoylation can occur on cysteine residues through an acyl thioester bond, which can be cleaved by treatment with neutral hydroxylamine or mild alkali. Consistent with this thioester linkage, overnight treatment with neutral hydroxylamine removed the incorporated radiolabel from CD36 (Fig. 3). The palmitate can be metabolized during biosynthetic labeling. Therefore, in order to confirm the chemical species incorporated into CD36, the radiolabel was removed by mild alkali and analyzed by thin layer chromatography. This clearly identified the incorporated radiolabel as palmitate based on its identical mobility as compared with authentic palmitate (Fig. 4). Thus, human CD36 is modified by thioester-linked palmitate.

We next wished to identify the specific sites of palmitoylation in CD36. The deduced amino acid sequence of human CD36 contains 10 cysteines (12). Two are located at the N terminus (cysteines 3 and 7), two are at the C terminus (cysteines 464 and 466), while the rest are in the region between the two hydrophobic domains. Although no consensus sequence has emerged for recognition of palmitoylation sites, the N- and the C-terminal cysteines are very close to the potential transmembrane domains, and this is a common location of palmitate attachment. In order to identify the palmitoylated cysteines, we made CD36 double/triple/quadruple mutants having two/three/four cysteines mutated to serines, respectively. These CD36 mutant constructs were transfected into 293 cells and stable cell lines were selected by G418 resistance. All of these cell lines expressed CD36 on the cell surface at comparable levels as determined by flow cytometry (data not shown) and a cell surface biotinylation assay (Fig. 5A, lanes 2-9; lane 1 is the control of non-transfected 293 cells). When metabolically labeled with [³H]palmitate, the quadruple mutant [SSSS] (Fig. 5B, lane 9) did not incorporate any palmitate, indicating that cysteines 243, 272, 311, 313, 322, 333 are not modified by palmitate, i.e. the N- and C-terminal cysteines are the only potential sites of palmitoylation. The double mutants [SSCC] (Fig. 5B, lane 3) and [CCSS] (Fig. 5B, lane 4) both incorporated palmitate, indicating that both the N and C termini of CD36 are palmitoylated. To investigate which specific cysteines are modified by palmitate, mutants with only one of these four N- and C-terminal cysteines were made. All of these triple mutants ([SCSS], [CSSS], [SSSC], [SSCS]; lanes 5-8 in Fig. 5B) incorporated radiolabel, indicating that all four of these cysteines (cysteines 3, 7, 464, and 466) are palmitoylated, although cysteine 464 is not as heavily palmitoylated as the other three cysteines.

As presented above, our study has demonstrated that CD36 is palmitoylated on two pairs of cysteine residues at both the N and C termini. This raises a critical question about the topology of CD36 in the plasma membrane. The sites of palmitoylation of proteins have been identified in a variety of positions along the peptide backbone of the protein, but the palmitoylated residues are located on the cytoplasmic side of the membrane (or at the interface of the cytoplasm and the cytoplasmic leaflet of the membrane) (21, 22). For CD36, the primary structure predicted from the cDNA sequence (12) contains hydrophobic domains of ~25 amino acids at both N and C termini, with a short stretch of ~10 amino acids flanking each domain, and the large extracellular domain (whose extracellular location is established by mapping of antibody epitopes and by the usage of N-linked glycosylation sites) between the two hydrophobic domains. Thus, our results on palmitoylation of CD36 would be most consistent with a topology of the protein that positions both the N and C termini of the protein in the cytoplasm with two transmembrane hydrophobic domains. Indeed, this is one proposed model of the membrane topology of CD36 (5). However, another study by Pearce and colleagues found evidence supporting secretion of a C-terminal truncation mutant of CD36, which the investigators interpret as evidence that only the C-terminal hydrophobic domain is transmembrane, with the N terminus of the protein extracellular (25).

Based on the palmitoylation results, our hypothesis is that there are two transmembrane domains of CD36, and that both the N- and the C termini of the protein are cytoplasmic. To test this hypothesis and definitively establish the membrane topology of CD36, we re-evaluated the C-terminal truncation mutant of CD36, CD36CT, in which the C-terminal hydrophobic domain and cytoplasmic tail have been removed. Transfection of 293 cells with CD36CT resulted in expression of a 70-kDa protein as shown by Western blotting using monoclonal anti-CD36 antibody FA6-152 (Fig. 6, lane 6). This C-terminal truncation mutant remained associated with the cell and was not found in the postculture medium (Fig. 6, lane 3). Quantitative analysis using serial dilutions put an upper limit of 6% for secretion of CD36 or CD36CT. These experiments gave the same results with CD36 and CD36CT (data not shown) and with versions of these two proteins that had a FLAG epitope tag (Fig. 6). Thus, failure to detect CD36 or CD36CT in the postculture medium is not likely to be a result of conformational changes or proteolysis, since anti-CD36 antibody FA6-152 recognizes a domain comprising amino acids 155-183 (26) and anti-FLAG antibody M2 recognizes the linear FLAG epitope placed at the C terminus of the protein.

To further confirm the membrane localization of CD36CT, the distribution of this mutant in the cytoplasmic and membrane cellular fractions was studied by Western blotting following a 100,000 × g spin. The CD36CT truncation mutant (Fig. 7, lane 3) as well as the wild-type CD36 (Fig. 7, lane 2) were recovered exclusively in the membrane fraction. Thus, the N terminus can serve by itself as a transmembrane anchor, consistent with the two transmembrane domain topology of CD36.

DISCUSSION

The major findings of this study are the demonstration of palmitoylation of human CD36, the identification of four cysteine residues at the very N and C termini of the protein as the sites of palmitoylation, and the delineation of the membrane topology of CD36. There are two sites of palmitoylation at both the extreme N terminus and C terminus, flanking the two hydrophobic domains of CD36. This suggests that each of these terminal regions is positioned as a cytoplasmic tail, and a direct analysis of the membrane topology of CD36 confirmed this structural model with two transmembrane domains.

It is well established that the bulk of the segment between the two hydrophobic domains of CD36 is extracellular. This is based on the findings that CD36 is extensively glycosylated (all of the 10 potential N-glycosylation sites are located between the two hydrophobic domains) and the immunodominant functional domain on CD36 recognized by flow cytometry was found between the region of amino acids 155-183 (26). However, whether or not the N terminus is cytosolic has been controversial. Two structural models for CD36 have been proposed. One suggested that CD36 has two transmembrane domains, two short cytoplasmic tails, and an extracellular domain (5). The other suggested that there is only one transmembrane domain at the C terminus of CD36 (25). Our data that both the N and C termini of CD36 are palmitoylated, and that the C-terminal truncation mutant of CD36 is membrane-bound and not secreted suggest that the first model is the correct one. This is consistent with an earlier report on CD36-related lysosomal membrane protein (LIMP II). The C-terminal truncation mutant of LIMP II was found membrane-bound, indicating that the N-terminal hydrophobic domain of LIMP II is a transmembrane domain (18). CD36 shows an extensive homology (34% identity) with LIMP II along the entire sequence except that the C-terminal cytoplasmic tail of LIMP II has additional 14 amino acids including a Leu⁴⁷⁵-Ile⁴⁷⁶ signal, which targets LIMP II to lysosomal membranes (27). One would therefore expect that the truncation mutants made from CD36 and from LIMP II after removing the C-terminal tail and the adjacent transmembrane domain would have the same overall membrane topology. The data on LIMP II (18) is consistent with our results that CD36 is membrane-bound.

The question that must be addressed is why the present results analyzing a C-terminal truncation mutant of CD36 differ markedly from a previous study (25). That study used a CD36 truncation mutant with an additional stretch of 16 amino acids encoded by the 3-untranslated region of the cDNA, but there is no clear reason why that should cause variant results. A notable difference in methodologies, however, is that the current study used Western blotting of equal fractions of postculture medium and cell lysate for analysis, i.e. the same method of detection was used for medium and cells to allow quantitative comparison. In the previous study (25), the proteins in postculture medium were radiolabeled with [¹²⁵I]NaI after purification of CD36, whereas the intact cells were surface-labeled with [¹²⁵I]NaI, and then CD36 was immunoprecipitated. Iodination in solution is more efficient as compared to intact cell surface labeling (28), thus making it impossible to assess the relative amount of the CD36 truncation mutant remaining membrane-bound versus that secreted in postculture medium. Although the method of the earlier study might be more sensitive for detecting a small amount of CD36 in postculture medium, the direct comparison of this medium with cell lysate in the current study (Fig. 6) shows that any secreted CD36 must represent a very small percentage of the total CD36.

CD36 joins a large and growing list of palmitoylated proteins (21, 22). Although no amino acid consensus for sites of palmitoylation has emerged (beyond the cysteine residue itself), the palmitoylated cysteines in CD36 fit a pattern of being located at the junction of a cytoplasmic tail of the protein and the cytoplasmic face of the membrane. Indeed, a recent study of the endoplasmic reticulum protein p63 suggested that the distance of the cysteine residue from the cytoplasmic face of the membrane was the only requirement for palmitoylation of this protein, and moving the cysteine even one position closer or further from the transmembrane domain abrogated palmitoylation (29). This is clearly not the entire story for CD36, as the pair of cysteine residues at the N and C termini are spaced four and two amino acids apart, respectively, yet each supports palmitoylation. In each of the members of the CD36 family of proteins including SR-BI/CLA-1 (15, 16), LIMP II (18), and Emp (19), there are cysteine residues located near the junctions of the cytoplasmic tails and transmembrane domains. Although all four of the palmitoylated cysteines in CD36 are not conserved across the family, we would predict that each of these members of the CD36 family will have a two transmembrane topology and be palmitoylated.

Finally, the question of the role of palmitoylation in the function of CD36 is not addressed by this study. Of particular interest will be the possible role of lipid modification (palmitoylation) in the function of CD36 as a lipid receptor, binding oxidized low density lipoprotein (8, 9), fatty acids (10), and anionic phospholipids (11). The construction of a non-palmitoylated mutant of CD36 as described in this report will allow a direct investigation of this important issue.