Identification of a human CD36 isoform produced by exon skipping. Conservation of exon organization and pre-mRNA splicing patterns with a CD36 gene family member, CLA-1.

During an examination of different cell types for CD36 mRNA splice variants, a partial cDNA from HEL cells was isolated and characterized. This CD36 cDNA had a 309-base pair deletion following the region encoding the first putative transmembrane domain of CD36. The open reading frame of the deleted CD36 cDNA was retained and was predicted to yield a protein lacking 103 amino acid residues. The presence of this variant was confirmed in RNA pools from placental tissue by a reverse transcriptase-coupled polymerase chain reaction assay. Comparison of the HEL CD36 cDNA with the genomic sequence revealed that the mRNA represented by this variant CD36 cDNA was produced by a pre-mRNA splicing reaction that excluded exons 4 and 5. Transient expression of the variant CD36 cDNA in COS-1 cells showed that CD36 immunoreactive protein was expressed on the cell surface but lacked an antigenic epitope defined by amino acid residues 41-143. This cell surface glycoprotein (M(r) approximately 57,000) was of identical molecular weight as a CD36 isoform observed on the surface of HEL cells. The exclusion of exons during CD36 pre-mRNA processing appears to be conserved within one other CD36 gene family member, CLA-1.

CD36 also mediates cytoadherence of erythrocytes infected with Plasmodium falciparum to capillary endothelial cells, a phenomenon that contributes to the morbidity and mortality of malaria in humans. The in vitro interaction of infected red blood cells with endothelial cells is inhibited by addition of the HL46447 (to R. H. L.). The costs of publication of this article were * This work was supported by National Institutes of Health Grant defrayed in part by the payment of page charges. This article must therefore be hereby marked "aduertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. The nucleotide sequence(s) reported in this paper has been submitted to the GenBankmIEMBL Data Bank with accession numberls) L06849.
The role of CD36 as a cell surface receptor has been extended to that of a signal transduction molecule. Proteins that bind to CD36, such as collagen and sequestrin (the P falciparum-encoded receptor for CD36) (Ockenhouse et al., 19911, have been shown to activate platelets and monocytes (Ockenhouse et al., 1989). Huang et al. (1991) have shown that CD36 is closely associated with pp66Yn, pp62ye", and pp54/58'Yn protein tyrosine kinases of the src gene family. The nature of the association between cytoplasmic protein kinases and CD36 remains to be determined.
Human CD36 cDNA and genomic clones have been isolated and characterized (Oquendo et al., 1989;Noguchi et al., 1993;Taylor et al., 1993). Alternative splicing of 5'-untranslated and 3"untranslated exons in CD36 pre-mRNA pools account for some of the heterogeneity in molecular sizes of CD36 mRNAs (Noguchi et al., 1993;Taylor et al., 1993). We wanted to determine if alternatively spliced coding exons could be detected in cell types expressing CD36 and to determine if the functional diversity of CD36 could be explained by alternative splicing of CD36 pre-mRNA.
We have now isolated and characterized a partial CD36 cDNA from HEL cells that has a 309-bp deletion within the coding sequence. Our results demonstrate that a CD36 isoform is expressed by at least one cell type and that the CD36 isoform lacks a domain that may have functional significance. EXPERIMENTAL PROCEDURES Cell Lines and Preparation of Human Tissue-HEL cells (erythroleukemia cell line) and COS-1 (monkey kidney cell line) were cultured as described previously (Taylor et al., 1993). Placental tissue was prepared as described by Taylor et al. (1993).
Monoclonal Antibodies-Anti-CD36 monoclonal antibody OKM5 was the kind gift of Ortho Pharamaceuticals. Anti-CD36 monoclonal antibodies 4Fll and 1E4 were kindly provided by Christian Ockenhouse (Walter Reed Army Medical Center, Washington, DC). The monoclonal antibodies 4Fll and 1E4 were produced using C32 melanoma cell membrane preparations as an enriched source of CD36. Immunizations, hybridoma production, and screening of cultures were described by Harlow and Lane (1988).
Isolation of a Partial CD36 cDNA Lacking 309 bp of Coding Sequence-A partial CD36 cDNA variant was isolated by using reverse transcriptase to synthesize cDNAs that were amplified by the polymerase chain reaction (RTPCR). Briefly, total RNA was isolated by the method of Chomczynski and Sacchi (1987) and poly(A)+ mRNA was enriched according to Sambrook et al. (1989). Amplification conditions have been described previously (Lipsky et al., 1991). The oligonucleotide primers were designed from the original cDNA sequence for CD36 tion consisted of: 5'-GCT TGA AlT CAA ATG GGC TGT GAC CGG (Oquendo et al., 1989). The oligonucleotide primers used for amplifica-AA-3' (sense; nucleotides 209-227) and 5'-CCT GGA A " CCT CAA Identification of a Human CD36 Isoform CAA AAG GTG GAA AT-3' (antisense; nucleotides 969-987). Each primer incorporated an EcoRI restriction endonuclease cleavage site to facilitate cloning of the amplification product. Prior to ligation, the product was centrifuged through a Sephadex G50 spin column to remove unincorporated oligonucleotides and EcoRI ends. The purified PCR product was then ligated into EcoRI-digested pUC18, and the ligated material was used to transform Escherichia coli strain HBlOl using standard procedures (Sambrook et 01.. 1989). Two distinct sizes of insert were obtained from the transformants. Five clones had the expected insert size of approximately 780 bp. However, one clone contained an insert of approximately 450 bp. This clone was given the designation G5.
Isolation of Genomic Clones-A human genomic DNA library was produced and screened a s described by Taylor et al. (1993). The initial hybridization pattern of one of the EMBL3 clones (A261) indicated that it partially overlapped with a previous clone known to contain exons 2a, 2b, 3, and 4 of the human CDJ6 gene (Taylor et al., 1993). A restriction endonuclease cleavage map was produced for the insert and was compared to the DNA hybridization pattern of genomic DNA digested with the same set of restriction enzymes and hybridized with a series of exon-specific probes. The same hybridization patterns for the insert were observed in genomic DNA, which suggested that there were no major rearrangements of the genomic structure within the A261 insert sequences. Subcloning of the 16-kb insert was performed using pBluescript(SK +/-) (Promega, Madison, WI) for DNA sequence determinations.
DNA Sequence Analysis-DNA sequence determination was performed by the dideoxy termination method (Sangeret al., 1977) using a kit obtained from U. S. Biochemical Corp. The sequencing primers used were the "reverse" pUC primer, the "-40" primer, or CD36 cDNA-specific oligonucleotides. Additional sequence information was obtained by digestion of particular subclones using appropriate restriction endonucleases to produce deleted forms of the region to be sequenced. The nucleotide sequence was determined on both strands. The sequence data were merged using the PCGENE program or the IntelliGenetics Suite program (IntelliGenetics, Inc., Mountain View, CA). Exon placement was identified by aligning the cDNA sequence information with the determined genomic sequence.
Detection of a Variant CDS6 mRNA from Placental Tissue hy RTI PCR-Two oligonucleotide primers were constructed to amplify by PCR a CD36 cDNA representing the entire coding region from randomprimed first strand cDNA from placental tissue. The sense oligonucleotide primer corresponded to nucleotides 4 6 6 3 from the 5'-untranslated region of a HEL cell CD36 cDNA (exon 3, 5'-untranslated region) (5'-CCA GAG CTT GTA GAA ACC-3') (Taylor et al., 1993). The second oligonucleotide primer was antisense and corresponded to nucleotides 1522-1540 of an HEL cell CD36 cDNA (Taylor et al., 1993) and represents the last coding exon of the human CDJ6 gene' (5'-TTT TCG ATC TGC ATG CAC-3'). Detection of the variant CD36 PCR product was accomplished using two specific oligonucleotide probes. The first oligonucleotide probe corresponded to nucleotides 529-546 of an HEL cell CDJ6 cDNA (Taylor et al., 1993) (5'-GAC AAC Tl'C ACA GTC TC-3'). The second hybridization probe was based on the junction sequences of the G5 CD36 cDNA (5'-CAA TTAAAAAGG CTG CAT CCC-3'). Hybridization conditions for the oligonucleotide probes were described by Wallace and Miyada (1987). Two final washes were performed, each a t 42 "C in 1 x SSC (SSC: 0.15 M NaCI, 0.015 M trisodium citrate, pH 7.0) for 15 min.
Protein  band in lane I represents the predicted 1200-bp product. The specificity of the hybridization probe was confirmed by its ability to detect a 450-bp EcoRI fragment of authentic G5 CD36 cDNA (indicated by a star, lane 2) but not a 1.8-kb EcoRI fragment of HEL-311 CDd6 cDNA (Taylor et al., 1993) in lane 3 . Panel B , the same blot a s in panel A but stripped of the initial hybridization probe and rehybridized with a coding region probe specific for sequences deleted from the G5 CD36 cDNA. A 1500-bp band is seen in lane 1. This oligonucleotide probe failed to detect G5 CD36 cDNA (lane 2 ) but was able to detect a 1.8-kb EcoRI fragment of HEL-311 CD.36 cDNA (indicated by a star). See "Experimental Procedures" for oligonucleotide hybridization and wash conditions. were rinsed twice with phosphate-buffered saline (PBS; 6.7 m M NapHPO,, 3.3 m M NaH,PO,, 0.14 M NaCI, pH 7.5) and extracted with 1 ml of lysis buffer (2.5% SDS, 10% glycerol, 5 9 2-mercaptoethanol, 0.0625 M 5 s -H C I , pH 6.8) a t approximately 100 "C for 2 min and then stored a t -70 "C. Polyacrylamide gel electrophoresis was performed as described by Lipsky and Silverman (1987) and the fractionated proteins transferred to nylon sheets using an electroblotter a t 500 mA for 1 h (Idea Scientific Co., Minneapolis, MN). The nylon sheets were blocked l h r a t 25 "C with 2% bovine serum albumin (Sigma) in PBS and incubated with the CD36-specific primary monoclonal antibodies 4 F l l and 1E4 (above), each a t a final dilution of 1:2000. After 1 h the blots were washed with 0.05% Tween 20 in PBS and incubated sequentially with goat anti-mouse IgG-conjugated horseradish peroxidase (Pierce Chemical Co.) a t 0.2 pg/ml for 1 h. The blots were washed again a s described above and incubated with ECL reagent (Amersham Corp.) followed by washing and exposure to XAR-5 film for 1-5 min according to the recommendations of the supplier.
Surface Laheling of Proteins-Cell surface proteins were labeled with sulfosuccinimidobiotin according to the method of Ingalls et al. (1986). Briefly, transfected COS-1 cells were grown a t 37 "C in a 5 4 CO, atmosphere in Dulbecco's modified Eagle's medium (Life Technologies, Inc.) containing 10% fetal bovine serum (Life Technologies, Inc.) for 48 h. The cells were suspended in PBS (pH 7.2) and diluted with PBS to a density of 1 x lo7 celldml. For surface labeling, sulfosuccinimidobiotin (Pierce) was added to the cell suspension to 1.0 mM. Biotinylation of surface proteins was camed out a t 37 "C for 10 min. The cells were then washed three times with PBS (pH 7.2) and then lysed using 1% Triton X-100 in PBS containing proteinase inhibitors as described by Taylor et al. (1993). Biotinylated proteins were recovered by centrifugation using avidin-conjugated agarose beads (Pierce) according to the procedure of Lisanti et al. (1989). After washing the immobilized avidin with PBS, the biotinylated proteins were released from the complex by boiling for 10 min in SDS-sample buffer (Lipsky and Silverman, 1987). The recovered proteins were separated on a 7% SDS-polyacrylamide gel and transferred to a nitrocellulose membrane a s described above. Following blocking with 2 8 bovine serum albumin (Sigma), the protein blot was incubated sequentially with anti-CD36 monoclonal antibodies ( 4 F l l and 1E4, above) and a solution of horseradish peroxidase-conjugated IgG a s described above. Detection of biotinylated CD36 was performed using ECL detection reagents according to the supplier's instructions (Amersham Corp.). Deglycosylation of Glycoproteins-For determining the molecular size of the amino acid backbone of CD36 isoforms, we treated cell ex-  Taylor et al. (1993). Exon sequences are shown in uppercase letters. The intervening sequences are shown in lowercase letters. The deduced amino acid sequence is shown in single-letter code, which appears above the first base of the codon to indicate the phase of the exon. The exon sequences not present in HEL CD36 cDNA clone G5 are underlined.

B E E E B
2a 2b   Detection of CD36 on Dansfected COS-1 Cells-Flow cytometric analysis was performed essentially as described by Tandon et al. (1989a), except that transfected COS-1 cells were removed from tissue culture plates with 5 m~ EDTA, 5 m~ EGTA in PBS and the primary monoclonal antibody OKM5, control non-immune mouse IgG, and secondary antibodies were incubated with live transfected COS-1 cells for 30 min each at 25 "C. The cells were pelleted by centrifugation and resuspended in phycoerythrin-conjugated goat anti-mouse F(ab'), (TAGO, Inc., Burlingame, CA) at a dilution of 1:lOO prior to fixation with 5% paraformaldehyde. RESULTS AND DISCUSSION Up to six related transcripts have been detected in cell lines and tissues that express CD36 (Oquendo et al., 1989;Noguchi et al., 1993;Taylor et al., 1993). These CD36 mRNAs are 7.6, 3.1, 2.4, and 2.0 kb in C32 melanoma cells and 5.5 and 3.0 kb in placental tissue and HEL cells. We initially sought to analyze this heterogeneity by PCR amplification and sequencing of partial CD36 cDNAs. Reverse transcriptase-coupled PCR (RT/ PCR) was performed using HEL cell poly(A)+ RNA with nondegenerate primers (see "Experimental Procedures"). The predicted product of 780 bp was visible on ethidium bromide-stained gels (data not shown). No products were detected when reverse transcriptase was omitted. The RT/PCR products were cloned as EcoRI fragments (see "Experimental Procedures"). Four of five randomly selected clones had the 780-bp insert. One clone (G5) had a smaller insert of 450 bp. The nucleotide sequence of the G5 clone was identical to our previously reported HEL CD36 cDNA sequence (Taylor et al., 19931, except that the sequence had a 309-bp deletion. However, the open reading frame for a deleted CD36 isoform was maintained in this cDNA clone. Translation of this sequence was predicted to yield a protein lacking 103 amino acid residues (residues 41-143), with the concomitant loss of three potential N-linked glycosylation sites.

F E P S L S V G T E A D N F T V L N L A V
We previously detected a 450-bp CD36-derived RTPCR product in buffy coat samples that suggested the presence of a variant CD36 mRNA (Lipsky et al., 1991). However, we wanted to use different hybridization probes to examine other normal cell types that expressed CD36 for a deleted CD36 cDNA representative of the G5 clone. To this end, we constructed oligonucleotide probes that represented an internal coding sequence and a junction probe that was specific only for the deleted form of CD36 cDNA (see "Experimental Procedures"). We were able to detect a variant CD36 cDNA lacking -300 bp using a G5 CD36 cDNA junction-specific hybridization probe (Fig. 1, panel  A). We were also able to detect a CD36 cDNA of the predicted size (1500 bp) using an internal hybridization probe specific for sequences deleted from the G5 CD36 cDNA (Fig. 1, panel B ) .
The results of this experiment showed that the variant CD36 mRNA was present in a normal cell type and not restricted to a single cell line.
To better define a plausible mechanism leading to a CD36 mRNA variant lacking 309 nucleotides, we examined the exon organization of the human CD36 gene for structural features consistent with exon loss. The coding region for the 5' end of CD36 mRNA is defined by four exons (Fig. 2, exons 3-6). Examination of the exon coding sequence showed that the 3'-ends of exons 3 and 5 are in the same phase as the 5'-ends of exons 4 and 6, while the 3'-end of exon 4 and 5'-end of exon 5 are in a different phase. Therefore, the splicing of exon 3 to exon 6 maintains the reading frame seen in the HEL CD36 variant cDNA. Due to the splicing phase of exon 3 to exon 6, exactly 103 amino acid residues are deleted without the insertion of a residue not found in the full length molecule. Thus, differential splicing of CD36 pre-mRNA accounted for the 309 bp "missing" from the HEL G5 cDNA sequence. Leung et al. (1992) described two CD36 peptides that block (amino acid residues 93-110) and enhance (residues 139-155) the binding of CD36 to immobilized TSP 1. Furthermore, binding of peptide 139-155 to TSP 1 could be directly inhibited by the monoclonal antibody, OKM5, which specifically blocks CD36-mediated adhesive functions. Thus, the functionally important OKM5 epitope may be partially defined on the CD36 molecule by residues 139-155. The variant CD36 cDNA lacks coding sequence corresponding to residues 41-143. We used this sequence information together with the RTPCR results from placental tissue to construct a CD36 cDNA lacking residues 41-143 but having a normal 3' end (see "Experimental Procedures"). The CD36 cDNA construction was expressed in COS-1 cells for functional studies. We gave the proposed isoform the designation CD36var.l.
Since native CD36 is post-translationally glycosylated, we treated whole cell extracts prepared from COS-1 cells producing either CD36 or CD36var.l with endoglycosidase F to show that their peptide backbones were different. As predicted, the CD36 construct produced a protein with a relative molecular weight of 88,000, which upon removal ofN-linked carbohydrate migrated at -55,000 (Fig. 3,panelA). The CD36var.l construct produced a protein of -57,000 that was reduced to a band of M, -38,000 following treatment with endoglycosidase F (Fig. 3,

panel A).
To demonstrate that a naturally occurring CD36 isoform was present on cells, we performed a protein blot experiment using surface-labeled HEL cells and compared the sizes of the immunoreactive proteins from CD36-and CD36var.l-producing COS-1 cells (see "Experimental Procedures"). We were able to detect an M , 57,000 band in surface-labeled extracts prepared from HEL cells that corresponded to the size of CD36var.l (Fig.  3, panel B ) . We also observed the more prominent M , 88,000 band predicted for HEL cells. We elected to use surface-labeled transfected COS-1 cells and HEL cells, since unglycosylated CD36 molecules present in a whole cell extract ( M , = 55,000) would confound interpretation of the results. We also demonstrated that internal proteins were not labeled with biotin by using a cytoplasmic marker, lactate dehydrogenase (data not shown). It is unlikely that the M , 57,000 band was the result of proteolysis, since we included a control extract prepared from transfected COS-1 cells producing only the M , 85,000 and M , 88,000 forms of CD36 (Fig. 3,panel B ) . Only the M, 85,000 and M , 88,000 products were seen in this sample. We also used proteinase inhibitors in the lysis buffer.
CD36 is a member of a family of proteins, members of which include LIMPII (Vega et al., 1991) (also known as GP85; Fujita et al. (1992)) and CLA (Calvo and Vega, 1993). There is remarkable amino sequence conservation between these three integral membrane proteins. The positions of 5 out of 6 cysteine residues are conserved in the proposed extracellular domain (or the luminal domain in the case of LIMPII) of all three proteins.   Overall, the amino acid similarity of human LIMPII, CLA, and human CD36 is about 34%. Our data clearly show that skipping of coding exons 4 and 5 gives rise to a CD36 isoform. Interestingly, a similar pre-mRNA splicing event can account for a variant CLA-1 cDNA(Ca1vo and Vega, 1993). Therefore, in addition to sharing similar primary amino acid structure, the CLA and CD36 genes also share similar exon organization (for at least the first three coding exons). It will be of great interest to investigate the mechanisms of alternative splicing of CLA and CD36 pre-mRNAs in different cell types and to understand the functional consequences of the differentially spliced products.