The deduced amino acid sequences for cow (Bos taurus), opossum (Monodelphis domestica), chicken (Gallus gallus), frog (Xenopus tropicalis) and zebrafish (Danio rerio) CD36 are shown in Figure 1 together with previously reported sequences for human and mouse CD36 (Table 1) [45,46]. Alignments of human with other vertebrate CD36 sequences examined were 53–100% identical, suggesting that these are products of the same family of genes, whereas comparisons of sequence identities of vertebrate CD36 proteins with human SCARB1 and SCARB2 proteins exhibited lower levels of sequence identities (30–32%), indicating that these are members of distinct CD36-like gene families (Supplementary Table 1)

The amino acid sequences for eutherian mammalian CD36 contained 472 residues, whereas opossum (Monodelphis domestica), platypus (Ornithorhynchus anatinus) and chicken (Gallus gallus) CD36 sequences contained 471 residues, while frog (Xenopus tropicalis) and zebrafish (Danio rerio) CD36 sequences contained 470 and 465 amino acids, respectively (Table 1; Figure 1). Previous studies have reported several key regions and residues for human and mouse CD36 proteins (human CD36 amino acid residues were identified in each case). These included cytoplasmic N-terminal and C-terminal residues: residues 2-6 and 462-472; N-terminal and C-terminal trans-membrane helical regions: residues 7-28 and 440-461 [32,45]; palmitoylated cysteine residues (Cys3; Cys7; Cys464; and Cys466) in the N- and C-terminal CD36 cytoplasmic tails [47]; exoplasmic Thr92, which is phosphorylated by protein kinase C alpha and contributes to the suppression of thrombospondin-1 binding in vitro [48]; His242 which contributes to the interaction of CD36-dependent endothelial cell adherence with Plasmodium falcurum [4]; and six exoplasmic disulfide bond forming residues: Cys243, Cys272, Cys311, Cys313, Cys322 and Cys333 [49].

Ten exoplasmic N-glycosylation sites for human CD36 have been previously identified for this protein (Figure 1; Table 2) [50]. One of these sites (site 2) contained a proline residue at the second position and may not function as an N-glycosylation site due to proline-induced inaccessibility [51]. Eight of these sites were predominantly retained among the 19 vertebrate CD36 sequences examined (sites 4, 5, 10, 15, 19, 23 and 25) (Figure 1; Table 2). The sequence conservation observed for these residues among the vertebrate CD36 sequences examined suggests that they contribute significantly to the structure and function of vertebrate CD36 as a glycoprotein. The multiple N-glycosylation sites observed for vertebrate CD36 sequences suggest a role for N-proteoglycan residues exposed on the external surface of plasma membranes in the performance of CD36 functions in binding various lipid molecules, including long chain fatty acids. This is also supported by recent animal model studies examining the impacts of reduced N-glycosylation upon cardiac long chain fatty metabolism, which demonstrated a key role for N-glycosylation in the recruitment of CD36 into cardiac membranes [52].

The N-terminal region for vertebrate CD36 sequences (residues 1-29 for human CD36) contained cytoplasmic (residues 2-7) and trans-membrane (residues 8-29) motifs which underwent changes in amino acid sequence but retained predicted cytoplasmic and trans-membrane properties in each case, respectively (Figure 1). Vertebrate N-terminal trans-membrane sequences, in particular, were predominantly conserved, especially for CD36 Gly12, Gly16 and Gly24/Gly25 residues, which were observed among the vertebrate CD36 sequences examined (Figure 1). Site directed mutagenesis studies of the related human SCARB1 sequence have demonstrated key roles for N-terminus trans-membrane sequence glycine residues, by facilitating oligomerisation and selective lipid uptake by SCARB1 conserved glycine residues [53] and similar roles may apply to the conserved C-terminal domain CD36 glycine residues. A recent report has shown, however, that CD36 is capable of binding acetylated and oxidized low-density lipoproteins as a monomer, even though multiple homo- and hetero-protein interactions are formed in the plasma membrane [8]. A conserved glycine residue was also observed for the vertebrate C-terminal trans-membrane sequences (human CD36 Gly452) (Figure 1), however the role for this residue has not been investigated.

Ten cysteine residues of the vertebrate CD36 sequences were conserved, including two within each of the N- (Cys3 and Cys7) and C-terminal (Cys464 and Cys466) cytoplasmic sequences, and six within the vertebrate exoplasmic sequences (Cys243; Cys272; Cys311; Cys313; Cys322; and Cys333) (Figure 1). The CD36 N- and C-terminal conserved cytoplasmic cysteine residues have been shown to be palmitoylated [47], which may contribute to protein-protein interactions, protein trafficking and membrane localization [54]. Comparative studies of vertebrate SCARB1 sequences have shown that N- and C-terminal cytoplasmic sequences lacked any conserved cysteine residues in this region [55]. The six conserved exoplasmic vertebrate CD36 cysteine residues participate in disulfide bridge formation for bovine CD36 (Cys243-Cys311; Cys272-Cys333; and Cys313-Cys322), resulting in a 1-3, 2-6 and 4-5 arrangement of the disulfide bridges [49]. In contrast, vertebrate SCARB1 exoplasmic sequences contain only four conserved cysteine residues forming disulfide bridges (Cys281; Cys321; Cys323; and Cys334); a fifth cysteine (Cys251) was not conserved among vertebrate SCARB1 sequences [55]; and a conserved sixth cysteine (not observed in the CD36 sequence) (human SCARB1 Cys384) which functions in lipid transfer activity [56,57].

Predicted secondary structures for vertebrate CD36 sequences were examined (Figure 1), particularly for the exoplasmic sequences. α-Helix and β-sheet structures were similar in each case, with a α-helix extending beyond the N-terminal and C-terminal trans-membrane regions, forming α1 and α7, respectively. A consistent sequence of predicted secondary structure was observed for each of the vertebrate CD36 sequences: N-terminal cytoplasmic sequence--N-terminal transmembrane sequence--α1--β1--α2--β2--β3--β4--β5--β6--α3--β7--α4--β8--β10--β11--β12--β13--β14--β15--α5--C-terminal trans-membrane sequence--C-terminal cytoplasmic sequence. Further description of the secondary and tertiary structures for CD36 must await the determination of the three dimensional structure for this protein, particularly for the exoplasmic region which directly binds oxidized LDL lipids and a wide range of other lipid-like structures, including long chain fatty acids [1,2,3,4,5,6,7,8,9,10].

Supplementary Figure 1 shows the alignment of 7 vertebrate CD36 amino acid sequences for the exoplasmic domain with colors depicting the properties of individual amino acids and conservation observed for some of these protein sequences. In addition to the key vertebrate CD36 amino acids detailed previously, others were also conserved, including 17 proline residues. A human CD36 genetic deficiency of one of these conserved prolines (Pro90→Ser) confirmed the significance of this residue, which lacked platelet CD36 [56]. Human CD36 deficiency has been shown to cause systemic metabolic changes in glucose and long chain fatty acid metabolism [59]. Prolines play a major role in protein folding and protein-protein interactions, involving the cyclic pyrrolidine amino acid side chain, which may introduce turns (or kinks) in the polypeptide chain as well as having destabilizing effects on α-helix and β-strand conformations [60]. In addition, the presence of sequential prolines within a protein sequence may confer further restriction in folding conformation and create a distinctive structure, such as that reported for the mammalian Na⁺/H⁺ exchanger, which plays a major role in cation transport [61]. Sequential prolines (Pro258-Pro259) were conserved for 6 of 7 vertebrate CD36 sequences examined and these may confer a distinctive conformation in this region supporting the lipid receptor functions for this protein. Moreover, regions of water exposed proteins with high levels of proline residues are often sites for protein-protein interactions [62] and these residues may significantly contribute to the binding of lipoproteins by the exoplasmic region of CD36. Similar results have been recently reported for the vertebrate SCARB1 exoplasmic region, however in this case, 30 conserved proline residues were observed [55].

Supplementary Figure 1 also shows conservation of 14 glycine residues for vertebrate CD36 exoplasmic domains, which due to their small size, may be essential for static turns, bends or close packing in the domain, or required for conformational dynamics during long chain fatty acid receptor on-off switching, as in the case of the aspartate receptor protein [63]. Both proline and glycine residues are frequently found in turn and loop structures of proteins, and usually influence short loop formation within proteins containing between 2 and 10 amino acids [61]. Evidence for these short loop structures within vertebrate CD36 exoplasmic sequences was evident from the predicted secondary structures for vertebrate CD36 (Figure 1), with proline and/or glycine residues found at the start of the following structures: α1 (Pro28; Gly30), β1 (Gly58), α2 (Pro73), β3 (Gly89-Pro90), β8 (Gly210), β12 (Gly287) and α5 (Gly420; Gly423). Moreover, CD36 sequential proline residues (Pro255-Pro256) were located in a region with no predicted secondary structure (between β9 and β10) but with disulfide bonds, which suggests that this is a region of conformational significance for CD36.

In addition to the prolines and glycine residues for the vertebrate exoplasmic CD36 sequences, there are several conserved charged amino acid residue positions, including positively charged Lys40/Lys41 located within the first predicted exoplasmic helix (α1); Arg/Lys89, Arg95; Arg97 and Lys101 within or near the predicted strand-β3/strand β4 THP-binding domain region; Lys233/Lys235/Arg236 near the PE-binding domain; Lys263 located near the β10 strand; Arg276 within the β11 strand and adjacent to a disulphide bond; Lys288 which lies between predicted β11 and β12 strands; Lys337 and Arg/Lys340 near a disulphide bond; Lys388/Arg389 near the predicted β15 strand; and Lys401/Lys409 within the last exoplasmic helix (α5). Two domains of the exoplasmic CD36 sequence have been potentially implicated in the binding and endocytosis of apoptotic neutrophils: residues 155-183; and 93-120 (see [7]) The latter domain is called CLESH (for CD36 LIMP-II Emp [erythrocyte membrane protein] sequence homology) which is predominantly conserved, particularly near Thr92, which is phosphorylated by protein kinase C alpha and contributes to the suppression of thrombospondin-1 binding in vitro [48]. One or more of these positively charged CD36 exoplasmic regions may contribute to long chain fatty acid binding prior to the translocation of fatty acids inside the cell membrane. There are also several conserved acidic amino acid regions, particularly a sequence of three acidic amino acids (367Asp/68Asp/369Asp) near the β13 predicted strand. The conserved nature of these CD36 charged residues suggests that play key functional roles for this cell membrane protein, which may include serving as the long chain fatty acid CD36 receptor site.

The amino acid sequences for human CD36, SCARB1 and SCARB2 (see Table 1) are aligned in Figure 2. The sequences were 30-33% identical and showed similarities in several key features and residues, including cytoplasmic N-terminal and C-terminal residues; N-terminal and C-terminal trans-membrane helical regions; exoplasmic disulfide bond forming residues, previously identified for bovine CD36: Cys243-Cys311; Cys272-Cys333; and Cys313-Cys322 [47]; several predicted N-glycosylation sites for human CD36 (10 sites), SCARB1 (9 sites) and SCARB2 (9 sites), of which only two are shared between these sequences (N-glycosylation sites 15 and 21 (Table 2); and similar predicted secondary structures previously identified for SCARB1 [55] (Figure 1). The Cys384 residue, for which the free-SH group plays a major role in SCARB1-mediated lipid transport [57], was unique to SCARB1, being replaced by other residues for the corresponding CD36 and SCARB2 proteins (Phe383 and Ala379, respectively). N-terminal trans-membrane glycine residues, which play a role in the formation of SCARB1 oligomers [53], were also observed for the human CD36 sequence, with twin-glycines (Gly23-Gly24) conserved for the vertebrate CD36 sequences (Figure 1). In contrast, only one of these glycines (Gly10) was observed for the human SCARB2 sequence. These results suggest that human CD36, SCARB1 and SCARB2 proteins share several important properties, features and conserved residues, including being membrane-bound with cytoplasmic and transmembrane regions, having similar secondary structures, but being significantly different to serve distinct functions.

Alignments were also prepared for the predicted lancelet (Branchiostoma floridae) and sea squirt (Ciona intestinalis) CD36-like sequences and a major epithelial membrane protein (EMP) from fruit fly (Drosophila melanogaster) (FBpp0072309) with the human CD36, SCARB1 and SCARB2 sequences (Figure 2). The lancelet, sea squirt and fruit fly sequences examined shared many features with the CD36-like human sequences, including the N- and C-terminal cytoplasmic and transmembrane sequences; similarities in predicted secondary structures; positional identities for five conserved cysteine residues, indicating conservation of at least 2 disulfide bridges for these proteins; predicted N-glycosylation sites, including one which is shared across all 6 CD-like sequences (site 15 in Table 2); and trans-membrane glycine residues, which were observed in both the N- and C-terminal sequences.

Table 1 summarizes the predicted locations for vertebrate CD36 genes based upon BLAT interrogations of several vertebrate genomes using the reported human CD36 sequence [45] and the predicted sequences for other vertebrate CD36 genes and the UC Santa Cruz genome browser [64]. Vertebrate CD36 genes were transcribed on either the positive strand (e.g., human, chimpanzee, gibbon, rhesus, rat and dog genomes) or the negative strand (e.g., mouse, cow, pig, opossum, chicken, frog and zebrafish genomes). Figure 1 summarizes the predicted exonic start sites for human, mouse, cow, opossum, chicken, frog and zebrafish CD36 genes with each having 12 coding exons, in identical or similar positions to those reported for the human CD36 gene [28].

Figure 3 shows the predicted structures of mRNAs for two major human CD36 transcripts and the major Cd36 transcripts for mouse and rat Cd36 genes [46,65,66]. The human transcripts were ~2 kbs in length with 14 (isoform c) or 15 (isoform e) introns present for these CD36 mRNA transcripts and in each case, a 3’-untranslated region (UTR) was observed. The human CD36 genome sequence contained a number of predicted transcription factor binding sites (TFBS), including the dual promoter structure of PPARA (peroxisome proliferator-activated receptor-α) and PPARG (peroxisome proliferator-activated receptor-γ) sites [67,68]. Moreover, the mouse Cd36 gene is regulated in a tissue specific manner by PPARA in liver and by PPARG in adipose tissues [69]. Other TFBS sites predicted for the human CD36 5’ promoter region included RSRFC4, a myocyte enhancer factor 2A found in muscle-specific and ‘immediate early’ genes [70]; CART1, a paired-class homeodomain transcription factor [71]; FOXJ2, a fork head transcriptional activator which is active during early development [72]; XBP1, a transcription factor which is critical for cell fate determination in response to endoplasmic reticulum stress [73]; and CDC5, a transcription activator and cell cycle regulator [74]. Hepatic upregulation of CD36 transcription in human patients has been recently shown to be significantly associated with insulin resistance, hyperinsulinaemia and increased steatosis in non-alcoholic steatohepatitis and chronic hepatitis C [43].

Figure 4 presents ‘heat maps’ showing comparative gene expression for various human and mouse tissues obtained from GNF Expression Atlas Data using the U133A and GNF1H (human) and GNF1M (mouse) chips (http://genome.ucsc.edu; http://biogps.gnf.org) [75]. These data supported a broad and high level of tissue expression for human and mouse CD36, particularly for adipose tissue, heart, skeletal muscle and liver, which is consistent with previous reports for these genes [11,32,66]. Overall, human and mouse CD36 tissue expressions levels were 4-6 times the average level of gene expression which supports the key role played by this enzyme in fatty acid metabolism, especially in liver, muscle and adipose tissue.

The broad tissue and high level of gene expression reported for human and mouse CD36 reflects key roles for this major cell membrane and muscle outer mitochondrial membrane glycoprotein in fatty acyl translocation and as a multiple ligand cell surface receptor of oxidized LDL lipoproteins (ox-LDL) and long chain fatty acids [7,11,33,66]. CD36 has also been described as a lipid ‘sensor’ playing a lipid receptor role for cells and tissues of the body [8,40]. Moreover, CD36 upregulation is associated with insulin resistance and hyperinsulinaemia, leading to liver pathology and increased steatosis [43]. In addition, cardiomyocyte CD36 cell surface recruitment is induced by insulin, AMP-dependent protein kinase (AMPK) activity or contraction, and is regulated in its vesicular trafficking by the RabGAP-AS160 substrate and AS160-Rab8a GTPase activating protein (GAP) [76,77,78]. These features provide a link between cell membrane CD36 and the reported insulin-stimulated phosphorylation of AS160 involved with the translocation of the glucose transporter GLUT4 to the plasma membrane [79,80]. It is also relevant to report that plasma levels of soluble CD36 are increased in type 2 diabetic patients [81].

Significant levels of CD36 expression have also been described in brain tissues, where CD36 contributes to cerebrovascular oxidative stress and neurovascular dysfunction induced by amyloid-beta in Alzeheimer’s dementia [12,13], and in transporting long chain fatty acids across the blood-brain barrier [82].

A phylogenetic tree (Figure 5) was calculated by the progressive alignment of 21 vertebrate CD36 amino acid sequences with human, mouse, chicken and zebrafish SCARB1 and SCARB2 sequences. The phylogenetic tree was ‘rooted’ with the lancelet (Branchiostoma floridae) CD36 sequence (see Table 1). The phylogenetic tree showed clustering of the CD36 sequences into groups which were consistent with their evolutionary relatedness as well as groups for human, mouse, chicken and zebrafish SCARB1 and SCARB2 sequences, which were distinct from the lancelet CD36 sequence. These groups were significantly different from each other (with bootstrap values of ~100/100) with the clustering observed supporting a closer phylogenetic relationship between CD36 and SCARB2, with the SCARB1 gene being more distantly related. This is suggestive of a sequence of CD36-like gene duplication events: ancestral CD36 gene duplication → SCARB1 and CD36 genes; followed by a further CD36 duplication, generating the SCARB2 and CD36 genes found in all vertebrate species examined (Figure 5). It is apparent from this study of vertebrate CD36-like genes and proteins that this is an ancient protein for which a proposed common ancestor for the CD36, SCARB1 and SCARB2 genes may have predated the appearance of fish > 500 million years ago [83]. In parallel with the evolution of CD36 and other CD36-like proteins (SCARB1 and SCARB2), thrombospondins (TSPs) are also undergoing evolutionary changes in their structures and functions [84], with gene duplication events proposed at the origin of deuterostomes.

BLAST (Basic Local Alignment Search Tool) studies were undertaken using web tools from the National Center for Biotechnology Information (NCBI) (http://blast.ncbi.nlm.nih.gov/Blast.cgi) [85]. Protein BLAST analyses used vertebrate CD36 amino acid sequences previously described [8,45] (Table 1). Non-redundant protein sequence databases for several vertebrate genomes were examined using the blastp algorithm from sources previously described [55]. This procedure produced multiple BLAST ‘hits’ for each of the protein databases which were individually examined and retained in FASTA format, and a record kept of the sequences for predicted mRNAs and encoded CD36-like proteins. Predicted CD36-like protein sequences were obtained in each case and subjected to analyses of predicted protein and gene structures.

BLAT (Blast-like Alignment Tool) analyses were subsequently undertaken for each of the predicted CD36 amino acid sequences using the UC Santa Cruz Genome Browser [64] with the default settings to obtain the predicted locations for each of the vertebrate CD36 genes, including predicted exon boundary locations and gene sizes. BLAT analyses were similarly undertaken for vertebrate SCARB1 and SCARB2 genes using previously reported sequences in each case (see Table 1). Structures for human and mouse isoforms (splicing variants) for human CD36, mouse Cd36 and rat Cd36 were obtained using the AceView website to examine predicted gene and protein structures [66].

Predicted secondary structures for vertebrate CD36 proteins, human SCARB1 and SCARB2, lancelet (Branchiostoma floridae) CD36, sea squirt (Ciona intestinalis) CD36 and a fruit fly (Drosophila melanogaster) epithelial membrane protein (FBpp0072309) were obtained using the PSIPRED v2.5 web site tools provided by Brunel University [86]. Molecular weights, N-glycosylation sites [49] and predicted trans-membrane, cytosolic and exocellular sequences for vertebrate SCARB1 proteins were obtained using Expasy web tools (http://au.expasy.org/tools/pi_tool.html).

The genome browser (http://genome.ucsc.edu) [62] was used to examine GNF Expression Atlas 2 data using various expression chips for human and mouse CD36 genes (http://biogps.gnf.org) [74]. Gene array expression ‘heat maps’ were examined for comparative gene expression levels among human and mouse tissues showing high (red); intermediate (black); and low (green) expression levels.

Alignments of vertebrate CD36, SCARB1 and SCARB2 sequences were assembled using BioEdit v.5.0.1 and the default settings [87]. Alignment ambiguous regions, including the amino and carboxyl termini, were excluded prior to phylogenetic analysis yielding alignments of 431 residues for comparisons of vertebrate CD36 sequences with human, mouse, chicken and zebra-fish SCARB1 and SCARB2 sequences with the lancelet (Branchiostoma floridae) CD36 sequence (Table 1). Evolutionary distances and phylogenetic trees were calculated as previously described [85]. Tree topology was reexamined by the boot-strap method (100 bootstraps were applied) of resampling and only values that were highly significant (≥95) are shown [88].