Structural Insights into Recognition and Translocation of Oxidized Phospholipid by CD36 Using Mass Spectrometry, Molecular Docking, Dynamics, and Metadynamics Simulations

Abstract

CD36 is a multifunctional receptor widely expressed in immune and nonimmune cells, known for its role in lipid transport and inflammatory signaling. Oxidized phospholipids (oxPLs), a class of prominent lipid oxidation products generated under oxidative stress, bind CD36 with high affinity, contributing to the development of atherogenesis and thrombosis and potentially influencing other CD36-dependent biological events. The molecular basis for the oxPL-CD36 interaction is poorly understood. Here, we used cutting-edge enrichment-mass spectrometry to identify lysine residues of CD36 that directly interact with oxPLs. These residues are located along a putative ligand translocation pathspanning from the apex of the extracellular domain to the entrance, interior, and around the exit of the lipid transport tunnel. Molecular docking revealed two sets of oxPL binding poses: one within a tunnel and the other on a surface loop cluster spanning the top to midsection, including the tallest loop containing oxPL-modified K398/K403. These findings support the selective oxPL binding observed in the LC–MS/MS analysis. Molecular dynamics (MD) simulation demonstrated that the sn-1 chain and headgroup of oxPLs engage distinct CD36 residues through hydrophobic, hydrogen-bonding, and ionic interactions, optimally positioning the reactive sn-2 group for lysine modification. MD and metadynamics simulations further demonstrated oxPL translocation through the tunnel, beginning with sn-1 chain insertion, followed by reorientation at the tunnel midsection, where the sn-2 chain and sn-3 headgroup lead the molecule toward the exit. Together, these studies indicate that CD36 may serve as a transporter of individual oxPL molecules into the cell and outline a translocation pathway, key residues and binding forces involved.

Introduction

CD36 is a multifunctional receptor widely expressed in immune cells such as macrophages, dendritic cells, platelets, T cells, and B cells, as well as in nonimmune cells including myocytes, adipocytes, microvascular endothelial cells, and specialized epithelial cells. CD36 is involved in a variety of biological and pathological processes, including long-chain fatty acid (LCFA) transport, angiogenesis, malaria infection, oxidized low-density lipoprotein (oxLDL) uptake, platelet-dependent thrombosis in dyslipidemia, and others. Correspondingly, CD36 binds a wide variety of ligands, including lipid-related ligands such as LCFA, , oxPLs, and bacterial negatively charged diacylglycerols, as well as protein-related ligands such as proteins containing TSR domains (e.g., thrombospondin-1, TSP-1), beta-amyloid, , PfEMP1 proteins of the malaria parasite, and proinflammatory S100 proteins. It is also a receptor for more complex ligands such as advanced glycation end products (AGEs), , carboxyethylpyrrole (CEP)-modified proteins, oxLDL, and others. However, studies on the binding sites or domains for many of these ligands remain relatively limited. −

CD36 has an oval-shaped extracellular domain flanked by two transmembrane segments, which connect to short intracellular tails at both the N- and C-termini. The extracellular domain consists of an antiparallel β-barrel core capped by a 3-α-helix bundle (α5, α6, and α8) and a large flexible loop connecting two long β-strands (β16 and β17) (Figure S1B, Supporting Information). The β-barrel core forms a LCFA transport tunnel that traverses the length of the extracellular domain and opens next to the cell membrane bilayer. , Studies on the X-ray crystal structure of mammalian CD36, its relatives SR-BI and LIMP2, and CD36 homologues from invertebrates, in combination with mutagenesis and in silico analysis, have led to the consensus that the tunnel facilitates delivery of lipids or related ligands to the cell membrane. − Lipids like LCFA or cholesterol can then cross the membrane by diffusion, while some other ligands may be directly transferred from the tunnel opening to a cognate receptor.

CD36 is a major cellular receptor for oxPLs ,, ligands generated at sites of oxidative stress in vivo. oxPLs are detected in circulation in hyperlipidemia, atherosclerotic lesions, inflamed lungs, and traumatic brain injury, ,,, among many other conditions and tissues. oxPLs possess multiple biological activities, − and their levels in vivo are tightly controlled by the innate immune system. They are mechanistically involved in various oxidative stress- and inflammation-associated pathologies, including atherosclerosis, ,− platelet hyperreactivity thrombosis, , diabetes, neurodegenerative diseases, − ischemia-reperfusion injury, , and others. For example, the interaction of CD36 with oxPLs present in the shell of oxLDL triggers oxLDL uptake by macrophages and foam cell formation. In platelets, oxPLs binding to CD36 induces activation and accelerates thrombosis. The pathological significance of oxPLs has been further demonstrated recently in vivo, where the oxPL-blocking antibody E06 dramatically reduced atherosclerotic lesions, lowered blood cholesterol, reduced hepatic steatosis, attenuated age-associated bone loss, and even increased lifespan in mice. ,

Despite extensive research on CD36’s biological functions, its molecular interaction with oxPLs remains poorly characterized. Phospholipids containing polyunsaturated fatty acids at the sn-2 position are easily oxidized, producing a variety of oxPLs. Among the most potent oxPL ligands for CD36 are those with a terminal γ-hydroxy (or oxo)-α,β-unsaturated carbonyl in the sn-2 acyl group. Mechanistic studies using synthetic analogs have shown that an intact sn-1 hydrophobic chain, a sn-3 headgroup, and a polar sn-2 tail of oxPLs are essential for high-affinity binding of oxPLs to CD36 and SR-BI. , Further studies revealed that a terminal negatively charged carboxylate at the sn-2 position is sufficient to generate high binding affinity to CD36. Additional factorssuch as polarity, rigidity, optimal chain length of the sn-2 groups, the sn-3 headgroup of oxPLs, and negative charge at the sn-3 positionmodulate binding affinity. Initial studies using ligand binding to glutathione S-transferase fusion proteins identified the CD36 region spanning amino acids 157–171 as a binding site for oxPLs, mediated by electrostatic interactions with positively charged lysine residues. Whether CD36 can mediate endocytosis-independent uptake of oxPLs is not yet known.

A subgroup of oxPLs containing an aldehydic group in the sn-2 position can form covalent adducts with protein amino acid residues including Lys, His, and Cys. We recently developed a high-efficiency enrichment protocol for oxPL-peptide adducts, which enabled detailed study of protein covalent modification by endogenous oxPLs in vitro and in vivo using LC–MS/MS analysis. − In this study, we used cutting-edge enrichment-mass spectrometry to identify ten amino acid residues on CD36 directly involved in oxPL binding. These modified residues span the apex domain of CD36 near the entrance to the lipid transport tunnel as well as the interior and the membrane-proximal exit of the tunnel. To independently define oxPL binding domains, we performed extra-precision induced fit docking (IFD). IFD revealed two major noncovalent binding sites for oxPLs on CD36: one at the apex and one inside the lipid transport tunnel. MD simulations further demonstrated that when oxPLs are covalently attached to lysine residues at the CD36 apex, their sn-1 LCFA chains can move into the tunnel entrance. Finally, metadynamics simulations demonstrated the details of the movement of oxPLs through the tunnel.

Together, IFD, covalent docking, and MD and metadynamics simulations revealed that the three chains of oxPLs collaborate to mediate binding and movement of oxPLs across the CD36 surface and tunnel via hydrogen bonding, hydrophobic interactions, and electrostatic forces. These studies, for the first time, systematically define the molecular mechanism by which oxPLs interact with CD36. Our results strongly suggest that CD36 may serve as a transporter of single oxPL molecules into cellsanalogous to its role in LCFA transportand outline a potential translocation pathway from the CD36 apex through its lipid tunnel to the cell membrane.

Results

A High Degree of Structural Similarity between Human and Murine CD36 Was Demonstrated by Protein Sequence Alignments and Superimposition of Their 3D Structures

To study the interaction of oxPL with CD36, we used both human and murine cells; therefore, we first assessed the structural similarity between the two proteins. Sequence alignment via UniProt Align showed 84.11% identity between human and murine CD36 (Figure S1A, Supporting Information). The X-ray crystal structure is available only for the ectodomain of human CD36 (PDB: 5LGD); however, AlphaFold accurately predicted the 3D structures (Figure S1C), allowing comparison between human and murine CD36. Superimposition of the AlphaFold-predicted human and murine CD36 (AF-hCD36 and AF-mCD36) 3D structures using PyMOL revealed remarkable structural similarity with a low root-mean-square deviation (RMSD) of 0.145 Å. Both structures possess 15 alpha helices and 16 beta strands, which align well (Figure S1A–C). The structure of AF-hCD36 also matched the experimental X-ray structure (RMSD 0.375 Å), confirming AlphaFold’s accuracy. All three structures contain interconnected internal cavities forming a Y-shaped tunnel that spans the CD36 ectodomain, with two top openings (likely entrances) and one membrane-proximal opening (likely an exit), which superimposed well (Figure S1C–G).

Lysine Residues along the Lipid Transport Tunnel of CD36 Were Selectively Modified by oxPLs

We previously demonstrated that oxPLs are high-affinity ligands for CD36. To explore the molecular mechanism of oxPL interaction with CD36, we first exposed RAW264.7 cellsa murine macrophage cell line expressing high levels of CD36to oxPLs. We used two representative model aldehydic oxPLs, HODA-PC and PONPC (Figure S2), with γ-hydroxyalkenal and aldehyde groups at their sn-2 positions, respectively. Confluent RAW264.7 cells were incubated with 20 μM HODA-PC (Experiment 1) or a mixture of 20 μM HODA-PC and 10 μM PONPC (Experiment 2) for 40 min at 37 °C. Cells were then washed and treated with 20 mM NaBH₄ to reduce and stabilize the reversible adducts formed between oxPLs and nucleophilic amino acids. Cells were lysed and digested with trypsin, and the oxPL-modified peptides were enriched and subjected to LC–MS/MS analysis using data-dependent acquisition (DDA), as described in Experimental Procedures. MS/MS spectra were searched against a UniProt murine protein database using Proteome Discoverer software.

Six LC–MS/MS data sets from two independent experiments identified oxPL modifications at nine lysine residues on murine CD36 (Figure , residues in magenta; Table S2 and Table S3). In Experiment 1, modification by HODA-PC was detected at eight residues (K52, K56, K337, K398, K403, K426, K431, and K437). In Experiment 2, five residues (K337, K385, K426, K431, and K437) were found to be modified by HODA-PC and/or PONPC. Notably, K385 is located in the midsection of the tunnelproviding the first direct evidence of oxPL presence deep inside the CD36 lipid transport tunnel. In addition to K385, ε-amino groups of six other lysine residues line different sections of the lipid tunnel: K337 at the entrance and K52, K56, K426, K431, and K437 near the membrane-proximal exit (Figure ). For oxPLs to react with these ε-amino groups, localization inside the tunnel is likely required. Two other modified lysine residuesK398 and K403are located on the apex of CD36, close to the tunnel’s wide entrance. They are found on a large flexible loop connecting the β16 and β17 strands (Figure S1B). This elevated and mobile structure may increase the likelihood of capturing ligands approaching the cell surface.

1.

Lysine residues along the lipid transport tunnel of CD36 were selectively modified by oxPLs. RAW264.7 cells were treated with HODA-PC (20 μM) for 40 min at 37 °C, while human platelets were treated with HODA-PC (10 μM) for 20 min at 37 °C. The samples were subjected to cell lysis, tryptic digestion, and LC–MS/MS analysis as described under “Experimental Procedures”. Murine CD36 is shown: (A) in an exterior surface mode with 40% transparency; hydrophilic and hydrophobic regions are shown in cyan and gray, respectively; (B) in cyan cartoon mode for the entire structure, combined with gray surface mode for cavities and tunnels. Modified lysine residues are represented in magenta stick format, while nonmodified lysine residues are shown in blue. For residues where human CD36 differs from murine CD36, the human variants are indicated in black.

MS2 spectra of oxPL-modified peptides containing these lysine residues are shown in Figures and S3. In Figure A,B, side-by-side comparison of spectra for the K337-containing peptide modified by PONPC and HODA-PC reveals identical y-ion fragments lacking the modified K337. In contrast, b-ion fragments, which retain the modified K337, exhibit mass shifts corresponding to the mass difference between the two oxPL adducts. Similarly, in Figure C,D, y-ion mass shifts of 169.14 Da are observed upon PONPC modification of the K385-containing peptide. The comparison between unmodified and modified peptides, as well as between peptides modified by different oxPL species, provides clear spectral evidence supporting the identification of oxPL-modified peptides. No modified residues were observed on the CD36 outer surface far from the apex, supporting a nonrandom spatial distribution of oxPL modifications and suggesting preferential interaction with the apex and tunnel regions. These findings support a model in which a single oxPL molecule translocates from the apex of CD36, through the tunnel, and to the membrane-proximal exit. Supporting this hypothesis, many of the lysine residues that were not modified (shown in blue in Figure ) are located away from the tunnel.

2.

Lysine residues of the CD36 receptor were intensively and selectively modified by chemically reactive oxPLs. RAW264.7 cells were treated with HODA-PC (20 μM) for 40 min at 37 °C. The samples were subjected to cell lysis, tryptic digestion, and LC–MS/MS analysis as described under “Experimental Procedures”. MS/MS spectra of murine CD36 peptide containing K337 modified by PONPC (A) and HODA-PC (B) and the unmodified peptide containing K385 (C) as well as its PONPC-modified counterpart (D). In panels (A) and (B), all y-ion fragments lack the modified K337 residue, resulting in identical m/z values for corresponding y-ions, as indicated by vertical double-headed dashed arrows (blue). In contrast, all b-ion fragments include the modified K337 residue. Due to the 74.03 Da mass difference between HODA-PC and PONPC, the corresponding singly charged b-ions in panels (A) and (B) differ by 74.03 m/z units. Selected b-ions illustrating this difference are highlighted with diagonal double-headed dashed arrows (red). Similarly, in panels (C) and (D), diagonal double-headed arrows (blue) highlight the 169.14 Da mass shift resulting from the adduct of the sn-2 residue of PONPC, which remains covalently attached following aminolysis during the enrichment procedure, while the remainder of PONPC is removed.

K337 (at the tunnel entrance) and K426 (at the tunnel exit) were the most and second-most frequently modified residues based on the number of adduct types detected and the abundance of corresponding peptides (i.e., higher total peptide-spectrum match [PSM] numbers) (Table S2). The γ-hydroxy alkenal group of HODA-PC can form several types of adducts with lysine ε-amino groups. At K337, we detected four types: Michael adduct (MA), cyclic hemiacetal (CH), hydrofuran (HF), and pyrrole (PY) adducts (Figure S2). The diversity and abundance of adducts at K337 and K426 likely indicate longer oxPL residence time at these sites, facilitating conversion from reversible to stable adducts. For the other modified lysines, only reversible MA adducts of HODA-PC were observed, likely reflecting shorter interaction durations. However, conformational constraints at these sites could also prevent conversion to the CH or HF forms. The terminal aldehyde group at the sn-2 position of PONPC can only form Schiff base (SB) adducts with lysine residues. Similar to HODA-PC, PONPC preferentially modified K337 and K426 (Table S2 and S3). It also formed SB adducts with K385located in the midsection of the tunneldemonstrating that oxPLs can access deep regions within the CD36 lipid transport tunnel.

Human CD36 Is Covalently Modified by oxPLs

Human platelets express high levels of CD36 and are activated by oxPLs via CD36. We incubated isolated human platelets with 10 μM HODA-PC at 37 °C for 20 min. A lower concentration of HODA-PC and shorter incubation time were used to avoid platelet aggregation induced by HODA-PC, which could hinder the access of HODA-PC to CD36. In selected experiments, we also used another oxPL CD36 ligand, KODA-PC (Figure S2), with a sn-2 terminal γ-ketoalkenal group. Samples were digested using both trypsin and chymotrypsin. Besides LC–MS/MS analysis in DDA mode, Parallel Reaction Monitoring (PRM) mode, a more targeted and sensitive approach, was employed to detect oxPL-modified peptides. K56, K403, and K426 of human platelet CD36 were modified by oxPLs similar to murine CD36 (Figure S4 and Table S3). We also found modification of K40, which corresponds to R40 in murine CD36. Arginine is significantly less reactive toward oxPLs, making it less likely to form detectable adducts. K40 is located at the tunnel exit, near the cell membrane, and on the opposite side of K437, which was found modified in murine CD36 (Figure ). Compared to murine CD36, we detected fewer modifications in human CD36, likely due to the lower concentration of oxPLs and shorter incubation times used.

Induced Fit Docking Revealed Binding Poses of oxPLs along the Lipid Transport Tunnel of CD36

The identification of modified residues at the CD36 tunnel entrance, inside the tunnel, and at the tunnel exit suggests that oxPLs, similar to fatty acids, can likely traverse the lipid transport tunnel of CD36. Thus, to visualize the oxPL binding poses that may serve as footholds for oxPLs during transition from the apex to the membrane-proximal section of CD36, we performed extra-precision induced fit docking, scanning the entire CD36 extracellular domain. In addition to chemically reactive aldehydic oxPLs, HODA-PC and PONPC, we included three nonreactive carboxylic oxPLs, PGPC, HOdiA-PC, and KDdiA-PC (Figure I). These three oxPLs differ in terminal functional groups and vary in chain lengths at the sn-2 position (Figure I), allowing us to study a wider variety of oxPL species. Only two sets of docked poses were obtained for all five oxPLs: one set on the apex domain of CD36 on the side of a large flexible loop containing K398 and K403 (Figure A–E and Figure S5A) and the other one inside the lipid transport tunnel (Figure F–H and Figure S5B). These findings are consistent with LC–MS/MS data and indicate that oxPLs preferentially bind to the apex and intratunnel regions of CD36, revealing oxPL’s preference for binding near the tunnel entrances at the top domain and inside the lipid transport tunnel.

3.

Induced fit docking experiments revealed that the sn-2 carboxylate of carboxylic oxPLs forms salt bridges with oxPL-modified lysines. The ectodomain of human CD36 is shown in surface mode (A,B,D,E,F) and cartoon mode (C). The lipid transport tunnel is shown in surface mode in F (inset), G, and H. The hydrophilic and hydrophobic residues are shown in cyan and gray colors, respectively. Ionic interaction and hydrogen bonds are shown with dashed red and yellow lines, respectively. Salt bridges between oxPL sn-2 carboxylate and oxPL-modified lysines are highlighted in green dashed circles. (A,B) HOdiA-PC sn-2 carboxylate forms a salt bridge with K403 (A) or K398 (B); meanwhile, the sn-1 LCFA chain partially inserts into the narrow tunnel entrance. (C) Multiple binding forces facilitate the binding (e.g., the HOdiA-PC-K398 complex), including the salt bridge between sn-2 carboxylate and K398, hydrophobic interactions of the sn-1 LCFA chain with surrounding hydrophobic residues, and hydrogen bonds made by oxygen atoms on the glycerol backbone and phospholipid phosphate with surrounding residues. (D,E) PGPC and KDdiA-PC make salt bridges with K398 via their sn-2 carboxylate, similar to HOdiA-PC. (F) HOdiA-PC at the wide tunnel entrance uses its sn-2 carboxylate to form a salt bridge with R/K337 and a hydrogen bond with K334 and extends its sn-1 chain to the intersection of the y-shaped tunnel, making hydrophobic interactions with surrounding hydrophobic residues; meanwhile, its phospholipid phosphate forms hydrogen bonds with K231 and G210. Additionally, the sn-1 acyl and sn-2 γ-hydroxy groups form hydrogen bonds with K334 and C333, respectively. (H) HOdiA-PC sn-2 carboxylate forms a salt bridge with K385, while the sn-1 chain and headgroup of oxPLs are stabilized via multiple binding forces, similar to (F). (G) HOdiA-PC near the tunnel exit uses its sn-2 carboxylate to form a salt bridge with K426 and three hydrogen bonds with T57, G58, and T59; meanwhile, its sn-1 chain extends up to the narrow branch of the y-shaped tunnel, forming hydrophobic interactions. Its phosphate forms three hydrogen bonds with R96, S268, and K385. (I) Structures of the sn-2 chains of oxPLs used in induced fit docking.

The sn-2 Terminal Carboxylate of Carboxylic oxPLs Forms Salt Bridges with Lysine Residues instead of Covalent Bonds Used by Aldehydic oxPLs

Induced fit docking does not take into account covalent binding, making it inaccurate for predicting the binding poses of aldehydic oxPLs such as HODA-PC and PONPC, which can form covalent bonds with lysine residues. On the other hand, oxPLs with a sn-2 terminal carboxylate group form strong salt bridges with the lysine residues, mimicking the covalent binding by aldehydic oxPLs, and thus are well suited for induced fit docking. The docked poses on the top domain of CD36 demonstrate that residues K398 and K403 form salt bridges with the sn-2 terminal carboxylate groups of PGPC, HOdiA-PC, and KDdiA-PC, despite the varying sn-2 chain lengths of 5, 8, and 12 carbons, respectively (Figure A–E). The sn-1 chain of these oxPLs inserts into the narrow tunnel entrance of CD36, while the sn-3 headgroup of oxPLs extends toward the wide tunnel entrance. In addition to the salt bridges of the sn-2 chain, these oxPLs (e.g., HOdiA-PC, Figure C) also employ other binding forces including the following: 1) hydrophobic interactions mainly by the sn-1 chain with CD36 hydrophobic residues such as I317, I318, P356, L392, and I410; 2) hydrogen bonds of the phospholipid phosphate group with K394 and N408, the sn-2 γ-hydroxy group with Q150, Q155, and S396, as well as the sn-2 terminal carboxylate group with E397 and the α-amino group of K398 (Figure C). In addition, another docked pose of HOdiA-PC with the sn-2 terminal carboxylate group binding to K403 shows the sn-1 chain extending toward the wide tunnel entrance (Figure S6A). Similarly, one docked pose of KDdiA-PC binding to K398 also shows its sn-1 chain extending toward the wide tunnel entrance (Figure S6B). Thus, while the sn-2 chain of oxPLs is interacting with K398/K403, the sn-1 chain could either insert into the narrow tunnel entrance or extend toward the wide tunnel entrance.

The docked poses of oxPLs in the lipid transport tunnel of CD36 show the sn-2 terminal carboxylate of carboxylic oxPLs forming salt bridges with three lysine residues that we found are covalently modified by aldehydic oxPLs. These include R_hum/K_mur337 at the wide tunnel entrance (Figure F), K385 at the midsection (Figure H), and K426 at the tunnel exit (Figure G). Multiple additional binding forces between oxPLs and CD36 residues are involved, as well. A docked pose of HOdiA-PC at the wide tunnel entrance (Figure F) shows that the sn-2 terminal carboxylate group forms a salt bridge with R_hum/K_mur337 at the wide tunnel entrance and a hydrogen bond with K334. Meanwhile, the sn-1 chain extends to the intersection of the y-shaped tunnel, making hydrophobic interactions with surrounding hydrophobic residues, including F201, I271, I275, F300, I330, C333, I341, L343, L387, V389, and L416. The sn-3 headgroup of oxPLs lingers around the wide tunnel entrance, forming hydrogen bonds with K231 and G210. Additionally, the sn-1 acyl oxygen atom forms a hydrogen bond with K334, and the sn-2 γ-hydroxy group forms a hydrogen bond with C333. Another docked pose of HOdiA-PC (Figure H) showed that the sn-2 terminal carboxylate group forms a salt bridge with the oxPL-modified K385, a hydrogen bond with S269, and a salt bridge with R96. Similar to the docked pose to R/K337, the sn-1 chain makes hydrophobic interactions with the surrounding residues at the intersection of the y-shaped tunnel. Meanwhile, the sn-3 phosphate and choline groups form salt bridges with K334 and E335, respectively. A third docked pose of HOdiA-PC near the tunnel exit (Figure G) showed that the sn-2 terminal carboxylate group forms a salt bridge with K426 and three hydrogen bonds with T57, G58, and T59. The sn-1 chain extends up to the narrow branch of the y-shaped tunnel, making hydrophobic interactions with surrounding hydrophobic residues, including A251, I271, I275, L295, F300, L328, L343, Y370, L371, F383, and L387. The sn-3 phosphate group forms three hydrogen bonds with residues near the intersection of the y-shaped tunnel, including R96, S268, and K385. Similar docked poses of KDdiA and PGPC making salt bridges with oxPL-modified R/K337, K385, and K426 are shown in Figure S7. Thus, terminal groups of sn-2 chains of carboxylic oxPLs and aldehydic oxPLs interact with the same lysine residues but use different bonds, i.e., salt bridges vs covalent bonds. At the same time, the sn-1 chain and sn-3 headgroup of oxPLs interact with CD36 via multiple additional binding forces.

Induced Fit Docking Demonstrates That Aldehydic oxPLs Position Their sn-2 Terminal Aldehyde Group near the Lysine Residues Identified in the LC–MS/MS Study

Despite the inability of induced fit docking to account for the covalent interactions of chemically reactive oxPLs, it successfully generated several docked poses for aldehydic oxPLs HODA-PC and PONPC, showing that the sn-2 terminal reactive group is positioned near the oxPL-modified lysine residues, and some even form hydrogen bonds with the modified lysine residues, including R/K337 at the wide tunnel entrance, K385 at the midsection, and K426 at the tunnel exit (Figure ). For example, Figure A shows one docked pose of HODA-PC with the sn-2 terminal aldehyde group forming a hydrogen bond with R/K337. Meanwhile, the sn-1 chain, like that of carboxylic HOdiA-PC with a similar docked pose, extends horizontally to the intersection of the y-shaped tunnel, forming hydrophobic interactions with surrounding residues. The sn-3 phosphocholine group forms 3 salt bridges with K231, K233, and E335. In another example (Figure B), the sn-2 terminal aldehyde group of PONPC is close to oxPL-modified K385 while forming a hydrogen bond with R96. The sn-1 chain, similar to carboxylic HOdiA-PC, extends horizontally to the intersection of the y-shaped tunnel, forming hydrophobic interactions with surrounding residues. The sn-3 phosphocholine group forms 2 salt bridges with R/K337 and E335. In addition, K334 makes a hydrogen bond with the oxygen of the sn-1 acyl group. Yet another example (Figure C) shows that HODA-PC positioned its sn-2 terminal aldehyde group close to oxPL-modified K426, forming a hydrogen bond with oxPL-modified K426 via the aldehyde and bifurcated hydrogen bonds with T380/A252 via the γ-hydroxy group. Meanwhile, the sn-1 chain, like that of carboxylic HOdiA-PC with a similar docked pose, extends up to the narrow branch of the y-shaped tunnel, establishing hydrophobic interactions with surrounding hydrophobic residues. The oxPL phosphocholine group forms multiple hydrogen bonds and salt bridges with R96, D209, and K385 (Figure C). Thus, aldehydic oxPLs and carboxylic oxPLs demonstrate similar binding poses, allowing the sn-2 terminal group to interact with selected lysine residues. The docked poses of all oxPLs within the tunnel showed docking scores that are significantly lower than those at the apex of CD36. For instance, the docking scores of HOdiA-PC in the tunnel ranged from −8.6 to −15.9, while those on the top domain ranged from −5.7 to −7.0.

4.

Induced fit docking demonstrates that aldehydic oxPLs position their sn-2 terminal aldehyde group near target lysine residues. The lipid transport tunnel is shown in surface mode, with hydrophilic and hydrophobic residues in cyan and gray, respectively. Ionic interaction and hydrogen bonds are indicated by dashed red and yellow lines, respectively. The oxPL sn-2 aldehyde and target lysine residues are highlighted in green dashed circles. (A) HODA-PC near the wide tunnel entrance extends its sn-1 chain to the intersection of the y-shaped tunnel, making hydrophobic interactions with surrounding hydrophobic residues. It anchors its sn-3 phosphate by forming hydrogen bonds with Y238 and salt bridges with K231/K233. The sn-3 choline group forms a salt bridge with E335, contributing to binding and stabilization. The sn-2 γ-hydroxyl group forms a hydrogen bond with R/K337, while the sn-2 aldehyde group stays close to target R/K337, facilitating the reaction. (B) PONPC near the wide tunnel entrance, like HODA-PC, anchors its sn-1 chain to the tunnel via hydrophobic interactions and its sn-3 headgroup via salt bridges with E335 and R/K337. Its sn-2 aldehyde forms a hydrogen bond with R96 and stays close to target K385, increasing the reaction likelihood. (C) HODA-PC near the tunnel exit extends its sn-1 chain to the narrow branch of the y-shaped tunnel, forming hydrophobic interactions. The sn-3 headgroup of oxPLs is anchored by multiple salt bridges with R96, D209, and K385, as well as hydrogen bonds with R96. The sn-2 aldehyde group protrudes to the tunnel exit, staying close to K426 and increasing the reaction likelihood. Hydrogen bonds of the sn-2 γ-hydroxyl group with A252 and T380 stabilize the sn-2 chain and facilitate the terminal aldehyde reaction with target K426.

These findings indicate that the tunnel provides a more energetically favorable environment for oxPL binding compared with the top domain, suggesting that the binding poses on the top domain act as temporary footholds before oxPLs move into the tunnel. The above induced fit docking unveiled a set of optimal binding poses where the sn-2 terminal groups of oxPLs interact with oxPL-modified residues K398 and K403 on the top of CD36, R/K337 at the tunnel entrance, K385 at the midsection, and K426 at the tunnel exit. The binding of the sn-1 chain and sn-3 headgroup of oxPLs to CD36 through multiple binding forces provides ample time for the sn-2 terminal group to interact and even react (for aldehydic oxPLs) with the oxPL-modified lysine residues. These molecular modeling experiments vividly illustrate how oxPLs bind to CD36 and modify the lysine residues detected by LC–MS/MS analysis.

MD Simulations Show oxPLs inside the Tunnel Moving the sn-2 Reactive Group to Physically Interact with Lysine Residues That Are Susceptible to oxPL Modification

The induced fit experiment produced a docked pose of HODA-PC with its sn-2 terminal group positioned near the oxPL-modified K426 at the tunnel exit (Figure C). The MD simulations initiated from this docked pose demonstrated that the sn-2 terminal alkenal group of HODA-PC exhibited conformational flexibility and maintained dynamic interactions with residue K426 (Supporting Information, Video S1), in agreement with our LC–MS/MS data. HODA-PC maintained its position by anchoring its phosphate/carbonyl groups to R96, S268, and K385 via hydrogen bonds/salt bridges (Video S1) and its sn-1 chain to the narrow tunnel branch via hydrophobic interactions (not shown in the video for simplicity). Similar results were obtained in four independent 300 ns MD simulations using different water models and temperatures (SPCE/310 K, SPCE/300 K, SPC/310 K, and SPC/300 K).

Protein–ligand interaction analyses of the full simulation trajectories demonstrated that residues R96, S268, and K385 maintained persistent contacts with HODA-PC via hydrogen bonds and salt bridges (Figure S8), underscoring their critical roles in anchoring the ligand and facilitating the reaction of the sn-2 terminal group with K426. Interaction frequencies (IF) in Figure S8 quantify the percentage of simulation time during which specific interactions are sustained. Significantly, S268 (IF, 95.9%) and K385 (IF, 95.5%) consistently formed hydrogen bonds with sn-1 carbonyl/phosphate and phosphate groups, respectively, throughout the entire MD simulation (Video S1). Furthermore, R96 (IF, 77.7% for hydrogen bonds and 24.3% for salt bridges) constantly interacted with sn-2 carbonyl/phosphate groups via hydrogen bonds/salt bridges.

Covalent Docking and MD Simulation Uncover a Hydrophobic Patch Formed by oxPLs and a Potential Translocation Pathway for oxPLs Adducted to the Apex of CD36

To investigate the behavior of the sn-1 chain and sn-3 headgroup once the sn-2 chain of HODA-PC is covalently adducted to K398/K403 on the CD36 apex, covalent dockings and subsequent MD simulations using the covalently docked poses were performed as described in Experimental Procedures. The resulting docked poses resemble those of HOdiA-PC forming salt bridges with K398/K403. Remarkably, most docked poses of HODA-PC adducted to K398 have their sn-1 chains close to or inserted into the narrow tunnel entrance (Figure A), while those adducted to K403 have their sn-1 chains extending toward the wide tunnel entrance (Figure B). Like HOdiA-PC, the sn-3 phosphocholine group of HODA-PC forms multiple hydrogen bonds and salt bridges with the surrounding residues. For instance, one docked pose of HODA-PC covalently adducted to K398 forms hydrogen bonds with K316 and P395, as well as salt bridges with E315, while the sn-1 chain is partly inserted into the narrow tunnel entrance (Figure C).

5.

Covalent docking and MD simulation uncover a hydrophobic patch formed by oxPLs and a potential translocation pathway for oxPLs adducted to the apex of CD36. The ectodomain of human CD36 is depicted in cyan cartoon mode, with the lipid transport tunnel shown in light gray surface mode. OxPL-modified lysine residues are illustrated in magenta stick mode. The ligand HODA-PC is represented in stick mode with green carbon atoms, red oxygen atoms, and blue nitrogen atoms. Ionic interactions and hydrogen bonds are indicated by dashed red and yellow lines, respectively. The covalent docking experiments showed that most HODA-PC adducted to K398 have their sn-1 chain partially inserted into the narrow tunnel entrance (A), while those adducted to K403 extend their sn-1 chain toward the wide tunnel entrance (B). In panel (C), the HODA-PC-K398 docked pose demonstrates that HODA-PC covalently binds to K398 via the sn-2 chain, with its sn-1 chain partially entering the narrow tunnel entrance. The sn-3 headgroup of oxPLs forms a hydrogen bond with K316 and a salt bridge with E315. In panels (D) and (E), MD simulations using the HODA-PC-K398 docked pose demonstrated that the sn-1 chain of HODA-PC escapes the narrow tunnel entrance at 104 ns and flips onto the CD36 apex (D), with the sn-1 and sn-2 chains mostly staying parallel, generating a hydrophobic patch on the CD36 apex (D). Shortly after, at 124 ns, the sn-1 chain moves into the wide tunnel entrance and remains there for the rest of the 576 ns MD simulation (E).

MD simulations revealed distinct behaviors of the sn-1 acyl chain in HODA-PC when bound to either K398 or K403 on the CD36 apex. In the simulation of HODA-PC bound to K398 (Video S2), the sn-1 chain exited the narrow tunnel entrance at 104 ns and moved to the top of CD36, aligning parallel to the sn-2 chain (Figure D and Video S2a) and forming a hydrophobic patch on the CD36 apex. Shortly thereafter, at 124 ns, the sn-1 chain entered the wide tunnel entrance (Video S2a) and remained inserted for the rest of the 576 ns MD simulation (Figure E and Video S2b). Similar results were obtained in three independent MD simulations using different water models and force fields (SPCE/OPLS4, SPC/OPLS4, and SPC/OPLS2005). This suggests a potential translocation pathway in which oxPLs initially bind to K398 via the sn-2 chain, followed by the sn-1 chain moving into the wide tunnel entrance.

In contrast, the 750 ns MD simulation of HODA-PC bound to K403 (Video S3) demonstrated a more dynamic behavior of the sn-1 chain. Throughout the simulation, the sn-1 chain alternated between extending downward over the flexible loops below K398 and K403 (360 ns in total, Video S3a) and positioning over the apex of CD36 (390 ns, Video S3b), where sn-1 aligned parallel to the sn-2 chain for a significant amount of time, forming a hydrophobic patch on the CD36 apex. Notably, the sn-1 chain did not insert into the wide tunnel entrance until 610 ns, and even then, the insertion was shallow and transient (Video S3c). Over the full 750 ns simulation, the sn-1 chain occupied the tunnel entrance for only 50 ns, significantly less than that observed in the HODA-PC-K398/CD36 complex.

To further investigate the potential for durable sn-1 chain insertion into the tunnel entrance of the HODA-PC-K403/CD36 complex, we conducted an additional 500 ns MD simulation using a docked pose derived from the late stage of the initial 750 ns trajectory in which the sn-1 chain was partially inserted into the tunnel entrance. During this extended simulation (Video S4), the sn-1 chain alternated among three conformations: inserted into the tunnel entrance (295 ns), floating above the entrance (121 ns), and extending away from the entrance (84 ns). Similar results were obtained in three independent MD simulations using different water models and force fields (SPCE/OPLS4, SPC/OPLS4, and SPC/OPLS2005). Collectively, these results demonstrate that while the sn-1 chain of HODA-PC bound to K403 is capable of tunnel entrance insertion, it requires substantially longer time to initiate this transition (610 ns) compared to the HODA-PC-K398/CD36 complex (124 ns), and the duration of insertion is markedly shorter. Thus, the adduction of oxPLs at K398 is more likely to facilitate translocation of oxPLs into the wide tunnel entrance of CD36.

Metadynamics Simulation of oxPL Translocation through the Lipid Transport Tunnel of CD36

While conventional MD simulations revealed that oxPLs within the tunnel could reposition their sn-2 terminal groups to interact with target lysine residues (e.g., HODA-PC with K426), translocation of the whole oxPL molecule was not observed, likely due to energy barriers within the tunnel. Metadynamics simulations were therefore used to overcome these barriers by applying moderate biasing energy, enabling oxPLs to escape local energy minima and progress through the tunnel. Two initial poses of the PONPC–AF-hCD36 complex were selected based on induced fit docking: Pose A, with the sn-2 chain positioned near K385 at the tunnel midsection, and Pose B, with the sn-2 chain near R337 at the tunnel entrance. In Pose A, the sn-3 headgroup was located at the tunnel entrance, while the sn-1 and sn-2 acyl chains extended into the midsection, flanking K385 (Video S5). Metadynamics simulations at 310 K for Pose A (Video S5) demonstrated a stepwise translocation process: As the sn-2 chain and sn-3 headgroup progressed down the tunnel, the flexible sn-1 chain first bent at its midsection and then flipped its terminal upward toward the CD36 apex, straightening before the entire molecule advanced further along the tunnel. The entire PONPC molecule exited the tunnel at ∼30 ns with the sn-2 and sn-3 moieties leading and pulling the sn-1 chain through. Subsequently, PONPC lingered near the C-terminal region of CD36, close to K426, for ∼7 ns before the sn-1 chain re-entered the tunnel exit. A second exit event resulted in the insertion of the sn-1 chain into the DPPC membrane. Notably, the DPPC membrane model may not fully replicate the lipid raft environment surrounding CD36, potentially affecting the accuracy outside the tunnel. Throughout translocation, the sn-2 terminal aldehyde group of PONPC physically interacted with target lysine residues including K385 (midsection) and K40, K52, and K437 (exit region), corroborating our LC–MS/MS findings of modification of these lysine residues by oxPLs. A parallel metadynamics simulation at a lower temperature of 300 K revealed a marked temperature-dependent slowdown in translocation. The sn-2 chain lingered near the tunnel exit for 88 ns, and the full exit of PONPC occurred after 115 ns. This extended residence time significantly increased the probability of physical interactions between the sn-2 terminal aldehyde group and target lysine residues, as illustrated in Video S6, where interactions were observed at K385, K40, K52, K56, K426, and K437.

Similar simulations were performed for pose B, where the sn-2 chain and oxPL sn-3 headgroup were located at the tunnel entrance and the sn-1 LCFA extended into the midsection above K385 (Figure A). Pose B was slightly farther from the tunnel exit than Pose A. Metadynamics simulations (Figure A–D, Video S7) showed a comparable translocation pattern: As the sn-2 chain and sn-3 headgroup advanced from the tunnel entrance inward, the sn-1 chain bent at its midsection to accommodate them, and then its terminal flipped upward toward the CD36 apex, straightening the chain before the entire molecule progressed further down the tunnel with the sn-2 chain leading the way. Then, PONPC exited the tunnel with the sn-1 chain partially inserted into the DPPC membrane. Across both poses and temperatures, metadynamics simulations consistently demonstrated that PONPC could translocate through the CD36 tunnel, with the reactive sn-2 aldehyde group engaging lysine residues along the pathway. We also obtained similar results for HODA-PC translocation through the tunnel in a pose comparable to Pose A of PONPC (Video S8). These experiments are consistent with our LC–MS findings and strongly support our hypothesis that the oxPLs translocate through the CD36 tunnel and modify the lysine residues lining the tunnel.

6.

Metadynamics simulation of oxPL translocation through the lipid transport tunnel of CD36. Metadynamics simulation was performed for the PONPC-CD36 complex as described in Experimental Procedures. Panels (A–D) show the poses of PONPC at 0, 8.2, 24.0, and 43.0 ns, respectively. The human CD36 is depicted in cyan cartoon mode, with the lipid transport tunnel shown in light gray surface mode at 60% transparency to visualize the ligand inside the tunnel. PONPC is displayed in spheres mode, with carbon atoms colored green for the sn-1 chain, brown for the sn-2 chain, and blue for the phosphocholine group. OxPL-modified lysine residues are highlighted in magenta stick mode. The four palmitoyl chains covalently linked to cysteine residues C3, C7, C464, and C466 within the transmembrane region of CD36 are shown in sphere mode with green carbon atoms. The DPPC membrane is depicted in line mode with green carbon atoms. A selected DPPC molecule used for distance measurements between its sn-2 ester oxygen atom (shown as a red sphere) and target lysine residues K56/K426 is illustrated in stick mode with green carbon atoms. All nitrogen and oxygen atoms are colored blue and red, respectively. Red dashed lines indicate the measured distances.

Metadynamics simulations revealed that several oxPL-modified lysine residues at the tunnel exit, such as K40, K431, and K437, are positioned close to the cell membrane, raising the theoretical possibility that oxPLs may first insert into the membrane independent of CD36 and subsequently attack these residues from the membrane. However, K56 and K426 at the tunnel exit are located too far from the membrane and may not be accessible to oxPLs embedded in the bilayer. In our metadynamics model, oxPLs were not fully incorporated into the membrane. To approximate the spatial relationship between membrane-inserted oxPLs and these target residues, we selected a DPPC molecule from the membrane that was closest to the CD36 tunnel exit and used it as a surrogate for membrane-incorporated oxPLs. We then measured the distances between the sn-2 ester oxygen atom (shown as a red sphere) of this DPPC molecule and the ε-amino group on the side chains of K56 and K426 at four time points: 0, 8.2, 24.0, and 43.0 ns (Figure A–D). The measured distances to K56 ranged from 23.3 to 29.7 Å, and to K426 from 20.6 to 24.2 Å, reflecting dynamic changes during the simulation. Given that the maximum extension of the sn-2 terminal reactive groups of PONPC and HODA-PC from their sn-2 ester oxygen atoms is 8.7 and 10.6 Å, respectively, these distances exceed the reach of oxPLs embedded within the membrane bilayer. These results further supported our hypothesis that oxPLs modified these target residues from within the tunnel during translocation through the tunnel.

Discussion

Despite a substantial body of evidence on the role of oxPLs in human pathology and extensive research into CD36 functions, a detailed understanding of the molecular interactions between CD36 and oxPLs remains limited. In this study, we used a combination of enrichment-mass spectrometry, induced fit and covalent docking, as well as molecular dynamics and metadynamics simulations to dissect the molecular details of CD36 interaction with various oxPLs and to demonstrate the transport of individual molecules of oxPLs by CD36.

Using LC–MS/MS and aldehydic oxPLs, we initially identified ten amino acid residues on CD36 that directly interact with oxPLs. Two of these residues lie on the CD36 apex near the wide tunnel entrance, while others are situated within the tunnelat the entrance, midsection, or exitoutlining a potential translocation path for oxPLs through CD36 to the cell membrane. K398 and K403 are located on the apex of CD36, which features a cluster of flexible loops. Among these, the largest loop extends to the very top of the apex, bridging the β16 and β17 strands and placing K398 and K403 at its highest points (Figure S1B). The elevated position, the increased mobility and ability of this domain to change shape to accommodate large molecules like oxPLs enhance the likelihood of K398/K403 reaching and binding ligands approaching the cell surface. The K398/K403 residues are conserved among mammals and semiconserved among birds, reptiles, amphibians, and fish (Figure S9). They are also found in some insect homologues of CD36, such as sensory neuron membrane protein 1 (SNMP1, Figure S10), indicating that they are functionally and/or structurally important. The structural feature with a large flexible loop is present in CD36 homologues involved in the transport of lipids or pheromones (Figure S10), indicating that it may constitute a common structural solution for lipid binding across species. Remarkably, extra-precision induced fit docking using five different oxPLs identified the cluster of flexible loops harboring K398/K403 as the only binding domain for oxPLs on the CD36 surface (Figure S5A). All other binding poses of these oxPLs were found inside the lipid transport tunnel of CD36 (Figure S5B). These in silico analyses are consistent with the LC–MS/MS analysis, which detected no modified residues on the outer surface of CD36 away from the apex and tunnel openings. These findings clearly indicate the nonrandom nature of oxPL modifications of CD36.

Our initial findings strongly suggested that oxPLs are transported through the CD36 tunnel and that lysine residues around the tunnel exit are modified by oxPLs from inside the tunnel. Indeed, induced fit docking and MD simulation experiments clearly demonstrated that the sn-2 terminal group of oxPLs can reach and interact with the lysine residues near the tunnel exit from within the tunnel. Moreover, no docked poses were observed on the outer surface of CD36 near the tunnel exit, making it unlikely that the sn-2 reactive group would attack lysine residues near the tunnel exit from the cell membrane. Furthermore, the marked distances of K56 and K426 from the cell membrane effectively rule out the modification of these residues around the tunnel exit by oxPLs residing in the lipid bilayer. Together with our observation of modification of K385 deep inside the tunnel, these findings strongly suggest a pathway for translocation of individual molecules of oxPLs, starting from the top of CD36, moving through the tunnel and finally reaching the cell membrane.

Our MD simulations with HODA-PC covalently bound to K398 revealed that when the sn-2 terminal group is linked to K398 on the CD36 apex, the flexible sn-1 LCFA chain readily inserts into the wide tunnel entrance, illustrating how oxPLs can translocate from the CD36 apex into the tunnel. Metadynamics simulations demonstrated that after the insertion of the sn-1 LCFA chain deep inside the tunnel, the polar sn-2 group moves in and pulls the sn-3 headgroup inside the tunnel. Then, the sn-2 group actively moves down the tunnel, pulling behind both the sn-3 headgroup and the sn-1 chain, while the latter bends and follows the lead. Finally, the sn-2 group exits the tunnel first, closely followed by the sn-3 headgroup and then the sn-1 chain. Interestingly, the insertion of oxPLs into the membrane does not happen immediately, probably reflecting the limitations of the DPPC bilayer model used in the simulations, which offers a simplified representation of the native lipid environment, potentially limiting the accuracy for events occurring outside the tunnel. We found that during translocation, the reactive sn-2 aldehyde group of PONPC physically interacted with the target lysine residues previously identified by LC–MS/MS, including K385, K40, K52, K56, K426, and K437. The consistency between the simulation and experimental data strengthens the proposed model of CD36-mediated oxPL transport. This process of transport of individual oxPL molecules by CD36 that we describe is similar to the one described for pheromone translocation in insect sensory neurons. Considering that oxPLs are bulky molecules, they may interfere with a major CD36 functionLCFA transport. These findings also indicate that CD36 may deliver individual molecules of oxPLs into the cell membrane with likely detrimental consequences. Indeed, we detected a number of cellular proteins modified by oxPLs in vitro as well as in vivo in hyperlipidemia and in pathologies associated with chronic inflammation [Gao, Byzova, Podrez, unpublished]. Taken together, the molecular dynamics and metadynamics simulations not only validate our LC–MS/MS results but also provide mechanistic insight into the translocation pathway of oxPLs through CD36.

In this study, K337, at the entrance of the lipid transport tunnel of murine CD36, is the most frequently modified residue. The set of adducts detected at this residue is the most diverse and includes MA, SB, CH, HF, and PY adducts. The MA adduct of HODA-PC with lysine residue is reversible and can convert to a more stable CH adduct after cyclization and to a stable HF adduct after further loss of a molecule of water (Figure S2). The PY adduct arises from a competing pathway, initiated by the formation of a Schiff base intermediate. We interpret the higher abundance and variety of adducts detected at K337 (and also K426 at the tunnel exit) as indicators of longer lingering time for oxPLs at these two residues, which allows the conversion of reversible MA adducts into more stable CH and HF adducts and the formation of the PY adduct. Induced fit docking showed that oxPLs interact with the R_hum337, the counterpart of K_mur337, via the sn-2 terminal functional group. At the same time, the glycerol backbone and oxPL sn-3 headgroup phosphate anchor to CD36 residues through multiple hydrogen bonds, while the sn-1 chain is secured via hydrophobic interactions with surrounding hydrophobic residues (Figure F). These binding interactions allow the sn-2 terminal reactive group to interact with R_hum/K_mur337 and the formation of various stable adducts. Finally, K337 has the highest peptide-spectrum match (PSM) numberthe number of times it was identified during database search. While a high PSM number alone cannot be used as an indicator of peptide abundance, in combination with high adduct diversity and induced fit docking results, it supports our conclusion that CD36 R_hum/K_mur337 is the most prominent target for oxPL modification. It should be noted that positively charged R or K residue at this position of CD36 is highly conserved among vertebrates, suggesting that electrostatic interactions at this position may be functionally important.

Application of reactive aldehydic oxPLs in combination with the recently developed method of peptide enrichment allowed us to initially identify Lys residues of CD36 directly interacting with oxPLs. To expand the selection of oxPLs beyond the aldehydic class, we also studied three carboxylic oxPLs with varying sn-2 chain lengths and different terminal groups in molecular modeling. Extra-precision induced fit docking was used to scan the entire ectodomain of human CD36 to identify oxPL binding sites. This method combines induced fit docking, which models receptor flexibility by iteratively refining the binding site around the ligand, with extra-precision docking, which uses a rigorous scoring system to improve the accuracy of the predicted binding poses. Together, they allow both the ligand and the receptor to adjust their shapes, resulting in more reliable modeling of binding interactions. The docked poses showed oxPL sn-2 carboxylates, regardless of the sn-2 chain length, forming salt bridges with lysine residues previously identified by LC–MS/MS as targets of aldehydic oxPL modification. These include K398 and K403 on the apex and immediately above the tunnel entrance, R/K337 at the tunnel entrance, K385 at the tunnel midsection, and K426 around the tunnel exit. Docking with the aldehydic oxPLs also generates similar docked poses in the tunnel, showing the sn-2 terminal reactive groups interacting with or close to the oxPL-modified residues. These results indicate that CD36, as a pattern recognition receptor, can accommodate multiple types of oxPLs using similar mechanisms that include multiple binding forces.

Our molecular modeling experiments demonstrated that all parts of oxPL molecules are involved in the binding to and movement on CD36 via various binding forces. These include hydrophobic interactions via the sn-1 acyl chain and the midsection of the sn-2 carbon chain; covalent bonding via the sn-2 terminal reactive group; multiple hydrogen bonding via oxPL oxygen atoms, especially the phosphate group with 4 oxygen atoms; and electrostatic interactions with the negative and positive charges in the oxPL headgroup. These remarkable features distinguish oxPLs from LCFA and other short-chain oxidized lipids, such as HNE, which are smaller in size and have fewer options for interacting with proteins. Our finding may explain, at the molecular level, the previous finding that all parts of the oxPL molecules are required for oxPL biological activities. ,, Furthermore, these findings suggest a common structural motif for oxPL binding. OxPLs likely bind at sites where the surrounding domain interacts with the sn-1 chain via hydrophobic forces, anchors the glycerol backbone and headgroup through hydrogen bonds or salt bridges, and positions the sn-2 reactive group for covalent bonding with lysine residues. Stronger anchoring forces would at least partially contribute to more abundant and diverse modifications of residues such as K337 and K426.

Previous studies have shown that CD36 exists as a monomer, dimer, or homo-oligomer within cells and that CD36 dimerization can be triggered by thrombospondin binding. , A recent study reported a crystal structure of the luminal domain dimer of LIMP-2, a CD36 homologue found in lysosomes. The LIMP-2 dimer forms a cavity that is absent in the monomer, allowing it to bind and tether phosphatidylserine liposomes more effectively than the monomer. If the new cavity in the CD36 dimer is formed with some lysine residues located inside it and is used for oxPL transport, we would likely observe modifications of Lys residues located on the side surface of CD36. LC–MS/MS analysis, however, detected no oxPL modifications of such residues, suggesting that the tunnel formed by CD36 dimerization is unlikely to participate in the oxPL interaction with CD36, at least in conditions used in this experiment.

In addition to LCFA, oxPLs, and oxLDL, CD36 also binds many other ligands, including S100 protein, AGEs, , beta-amyloid, , CEP-modified proteins, oxidized phosphatidylserine, and others. While research on the binding sites for many of these ligands is limited, − analysis of their structural properties reveals that many carry net negative charges. The cluster of positively charged lysine residues at the CD36 apex is likely involved in binding many negatively charged ligands via electrostatic interactions. Due to overlapping binding domains, oxPLs may interfere with the binding of other ligands that employ a similar mechanism and, correspondingly, with the biological responses to such ligands. While discovery of the prominent role of CD36 in cancer − and cardiomyopathy − ignited interest in small molecule inhibitors for CD36, our studies may provide valuable guidance for development of such inhibitors.

Taken together, our study identified molecular forces and CD36 residues involved in the interactions of oxPL with CD36. These findings further outline a translocation pathway for oxPLs from the CD36 apex, through the tunnel, and toward the cell membrane. Transport of individual oxPL molecules into the cell via this route may have significant consequences for cell function. This detailed mechanistic insight could aid the development of strategies to disrupt oxPL–CD36 interactions under relevant pathological conditions.

Experimental Procedures

Reagents

Sequencing-grade modified trypsin was purchased from Promega (WI). Oxidized phospholipids (Table S1). HODA-PC (9-hydroxy-12-oxo-10-dodecenoic acid ester of 2-lysophosphocholine) was synthesized as previously described. PONPC (1-Palmitoyl-2-(9-oxo-nonanoyl)-sn-glycero-3-phosphocholine) was purchased from Avanti Polar Lipids, Inc. RAW 264.7 cells were purchased from the American Type Culture Collection (ATCC). All other chemicals were obtained from Sigma or Fisher Scientific Co. unless otherwise specified.

Treatment of RAW 264.7 Cells with HODA-PC/PONPC

RAW 264.7 cells were grown to a 90% confluency in DMEM with 10% fetal bovine serum and then washed once with phosphate-buffered saline (PBS). Next, 6.5 mL of RPMI-1640 medium without cysteine, with 10 mM N-(2-hydroxyethyl)­piperazine-N′-ethanesulfonic acid (HEPES) and 1 μM bafilomycin A1, was added, and cells were incubated for 15 min at 37 °C in a cell culture incubator. Then, HODA-PC (final concentration 20 μM) or a mixture of HODA-PC/PONPC (final concentrations 20 μM/10 μM) was added, and incubation continued for an additional 40 min at 37 °C and then at 4 °C for 30 min. Cells then were washed and incubated in reduction buffer (PBS, 25 mM HEPES, and 25 mM NaBH₄) for 15 min at 4 °C; cells were then collected and stored at −80 °C for subsequent analysis.

Platelet Isolation and Treatment with oxPLs

Platelets were isolated from human venous blood as described previously and resuspended in Tyrode’s buffer with 10 mM HEPES, 0.05% dextrose, and 0.28 μM PGI₂ (pH 7.2) at a concentration of 1.67 × 10^–8/ml. HODA-PC was then added to a final concentration of 10 μM, and the mixture was incubated at 37 °C with gentle shaking for 20 min. Subsequently, 20 mM NaBH₄ was added, followed by a 15 min incubation at room temperature. The platelets were then pelleted by centrifugation at 4000 rpm for 15 min at 22 °C, and the final platelet pellet was frozen at −80 °C for subsequent analysis.

Cell Lysis, Protein Digestion, and Sample Preparation for LC–MS/MS Analysis

The cell pellet was resuspended and lysed in 8 M urea with 150 mM TRIS buffer (pH 7.8). The cell lysate was subjected to reduction using 10 mM dithiothreitol (DTT) and alkylation with 40 mM iodoacetamide (IAA). To remove excess IAA, an additional 20 mM DTT was added. The solution was then diluted with 50 mM TRIS buffer (pH 7.6) to lower the urea concentration to 2 M, followed by overnight tryptic digestion at 37 °C (trypsin: protein ratio = 1:100). The resulting mixture was subjected to an enrichment procedure we developed to enrich oxPL-peptide adducts.

Mass Spectrometry and Data Processing

Chromatographic separation of the peptide samples was performed by the nanoElute 2 high-performance nanoflow LC system equipped with a reversed-phase capillary chromatography column (Bruker 25 cm × 75 μm i.d. C18 ReproSil AQ, 1.9 μm, 120 Å). An elution gradient was used by mixing mobile phase A (0.1% formic acid in water) with solvent B (0.1% formic acid in acetonitrile), as follows: isocratic elution with 2% B from 0 to 5 min; increasing to 40% B from 5 to 76 min; increasing to 70% B from 76 to 78 min; isocratic elution with 70% B from 78 to 85 min; decreasing to 2% B from 85 to 86 min; and isocratic elution with 2% B from 86 to 100 min. Electrospray ionization MS was performed with a Bruker TimsTof Pro2 Q-Tof mass spectrometry system operating in positive ion mode. Five microliters of samples were injected. The peptides eluted from the column at a flow rate of 0.30 μL/min were introduced into the source of the mass spectrometer online. The peptides were analyzed using a parallel accumulation-serial fragmentation (PASF) data-dependent acquisition (DDA) method. This method selected precursor ions for fragmentation through a TIMS-MS scan, followed by 10 PASEF MS/MS scans. The TIMS-MS survey scan was conducted between 0.60 and 1.6 Vs/cm² and 100–1700 m/z, with a ramp time of 166 ms. The total cycle time for the PASEF scans was 1.2 s. The MS/MS experiments were performed with collision energies ranging from 20 eV (0.6 Vs/cm²) to 59 eV (1.6 Vs/cm²). Precursors with 2–5 charges were selected, with a target value of 20,000 au and an intensity threshold of 2500 au Precursors were dynamically excluded for 0.4 min. The MS/MS spectra obtained were searched against a UniProt mouse database (25,365 sequences, 2022) or a UniProt human database (42,279 sequences, 2022) using Thermo Proteome Discoverer (version 1.4). Trypsin was set to cleave at Lys and Arg. The fixed modification included carbamidomethylation on cysteine residues (Cys, 57.021 Da). Dynamic modifications included oxidation on methionine (Met, 15.995 Da) and oxPL modifications in the form of various types of adducts with proteins, including Schiff base adduct (SB), Michael adduct (MA), cyclic hemiacetal adduct (CH), hydrofuran adduct (HF), and pyrrole adduct (PY). The corresponding mass shifts are shown in Table S1. The search parameters used were three missed cleavage sites, a mass tolerance of 20 ppm for the parent ion, and 0.04 Da for the fragment ion. The search results were filtered by a false discovery rate of 1% with a decoy database search.

Protein Preparation

The starting coordinates of human CD36 [PDB ID: 5LGD] were downloaded from the Protein Data Bank (www.rcsb.org) and further modified for Glide docking calculations. The protein was minimized using the Protein Preparation Wizard, employing the OPLS4 (Optimized Potentials for Liquid Simulations) force field. During this process, only non-hydrogen atoms were subjected to progressively weaker restraints. The refinement procedure was performed based on the default protocol of Schrodinger, LLC, NY, USA, as Glide uses the full OPLS4 force field at an intermediate docking stage and is more sensitive to geometrical details than other docking tools. The most probable positions of hydroxyl and thiol hydrogen atoms, protonation states, and tautomers of His residues, as well as Chi “flip” assignments for Asn, Gln, and His residues, were determined. Finally, the minimizations of protein were performed until the average root-mean-square deviation of the non-hydrogen atoms reached 0.3 Å.

Ligand Preparation

OxPLs were drawn using Maestro (Schrodinger, LLC, New York, NY, USA). Appropriate bond orders were assigned to each structure utilizing the LigPrep package from Schrödinger, LLC. The ligands were then converted to the Maestro format (mae) for geometric optimization and partial atomic charge computation. Next, up to 32 poses per ligand, each with distinct steric features, were generated for the subsequent docking study.

Induced Fit Docking (IFD)

Induced fit docking (IFD) (Schrodinger, LLC, NY, USA) was used to dock oxPLs to the CD36 receptor. Initially, the ligand was docked into a rigid receptor model with van der Waals (vdW) radii scaled down by 0.5 for both the protein and ligand nonpolar atoms. A constrained energy minimization was performed on the protein structure, keeping it close to the original crystal structure while eliminating bad steric contacts. Energy minimization was carried out using the OPLS4 force field with an implicit solvation model until default criteria were met. The initial docking was conducted using Glide XP mode, and ligand poses were retained for protein structural refinements. Prime (Schrodinger, LLC, NY, USA) was then used to generate the induced-fit protein–ligand complexes. These complexes underwent side-chain and backbone refinements using the Prime module of Schrodinger, including all residues with at least one atom within 4.0 Å of each corresponding ligand pose. The refined complexes were ranked based on Prime energy, and the top 20 receptor structures within 30 kcal/mol of the minimum energy structure were selected for a final round of Glide docking and scoring. In the final step, each ligand was redocked into the top 20 refined structures using Glide XP.

Covalent Docking

OxPLs were covalently docked against CD36-modified lysine residues via Michael addition or Schiff base reaction using Schrodinger’s CovDock (Schrodinger Release 2020-1: CovDock, Schrodinger, LLC, New York, NY). The workflow begins with a Glide noncovalent docking simulation. To avoid steric clashes, the reactive residue in the protein is temporarily mutated to alanine. The ConfGen module is used to sample ligand conformations prior to docking. Prereaction ligands are then docked noncovalently to the binding sites using Glide, with positional constraints ensuring the ligand’s reactive group is within a 5 Å distance of the target residue. The alanine mutation is subsequently reversed to restore the original reactive residue. The rotamer states of the reactive residue are sampled to facilitate covalent bond formation. CovDock then forms a covalent bond between the ligand and the reactive residue. This step involves sampling various poses to identify the best geometric fit. The covalent complexes are optimized using the Prime VSGB2.0 energy model and finally scored using all-atom molecular mechanics with the OPLS force field and the VSGB2.0 implicit solvent model.

Molecular Dynamics (MD) Simulations

Molecular dynamics (MD) simulations were conducted for oxPL docked poses to the CD36 receptor using the Desmond MD code and the OPLS4 force field for system minimization. A 15 Å buffered cubic system with periodic boundary conditions was constructed using the Desmond system builder, incorporating either SPC or SPC/E explicit water solvent. The overall charge was neutralized with 0.15 mol/L NaCl. Simulations were performed in the NPT ensemble, maintaining a constant number of particles, pressure, and temperature. The temperature was set to 310 K, and the pressure was set to 1.013 bar, controlled by a Berendsen thermostat and barostat, to mimic physiological conditions. An integration time step of 2.0 fs was used. Coulombic interactions were calculated with a cutoff radius of 9.0 Å, and long-range electrostatic interactions were computed using the smooth particle mesh Ewald method.

Metadynamics Simulations

Induced fit docking was performed to generate the initial poses of the PONPC-AF-hCD36 complex. Two docked poses showing potential interactions of the sn-2 terminal aldehyde with target lysine residues were selected for further simulation: Pose A, with the sn-2 acyl chain positioned near target K385 at the tunnel midsection, and Pose B, with the sn-2 chain located near the target K/R337 at the tunnel entrance. In Pose A, the sn-3 phosphohead group was oriented toward the tunnel entrance, while the sn-1 and sn-2 chains extended into the tunnel. In Pose B, the sn-2 and sn-3 regions were positioned at the entrance with the sn-1 chain extending into the midsection. To mimic the native lipidation of CD36, covalent dockings were performed to covalently attach four palmitoyl acyl chains to cysteine residues C3, C7, C464, and C466 of AF-hCD36 in both docking poses. A DPPC bilayer membrane was then constructed around the transmembrane domain using the Desmond System Builder tool to provide a simplified lipid environment.

Metadynamics simulations were carried out using the Metadynamics module in Maestro Desmond (Schrödinger). The collective variable selected was “Distance”, defined between two atom groups: Group 1 included non-hydrogen atoms of the sn-2 and sn-3 regions of PONPC, and Group 2 included non-hydrogen atoms of residues K39, K40, and Q41 of CD36. Simulation parameters included a width of 0.4 and a wall constraint set to 37. The NPγT ensemble was used with a surface tension of 0.04 atm and a pressure of 1.01325 atm. A small hill height of 0.075 kcal/mol was applied to introduce a moderate biasing energy along the selected collective variable, enabling PONPC to escape local energy minima and facilitating its translocation through the tunnel. Each metadynamics simulation was run for 120 ns at either 310 or 300 K to assess the influence of temperature on oxPL mobility and interaction with target lysine residues.

Glossary

Abbreviations

oxPLs

oxidized phospholipids

LCFA

long-chain fatty acid

oxLDL

oxidized low-density lipoprotein

AGEs

advanced glycation end products

CEP

carboxyethylpyrrole

IFD

induced fit docking

MD

molecular dynamics

RMSD

root-mean-square deviation

MA

Michael adduct

SB

Schiff base adduct

HF

2,3-dihydrofuran adduct

CH

cyclic hemiacetal adduct

PY

pyrrole adduct

HODA-PC

9-hydroxy-12-oxo-10-dodecenoic acid ester of 2-lysoPC

KODA-PC

9-keto-12-oxo-10-dodecenoic acid ester of 2-lysoPC

HOdiA-PC

5-hydroxy-8-oxo-6-octenedioic acid ester of 2-lysoPC

KDdiA-PC

9-keto-10-dodecendioic acid ester of 2-lysoPC

PONPC

1-palmitoyl-2-(9-oxo) nonanoyl-sn-glycero-3-phosphocholine

PGPC

1-palmitoyl-2-glutaroyl-sn-glycero-3-phosphocholine

PRM

Parallel Reaction Monitoring

DDA

data-dependent acquisition

PSM

peptide-spectrum match.

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/jacs.5c07761.

• Sequence alignment and structural comparison of CD36; structures and adducts of aldehydic oxPLs; MS2 spectra of oxPL-modified peptides; docking poses and interactions of oxPLs with CD36; MD and metadynamics simulations of CD36–oxPL interactions; structural features of the CD36 flexible loop; mass shifts of oxPL adducts; and CD36 peptides identified by LC–MS/MS (PDF)

• Video S1: MD simulations of oxPL interaction with the target lysine residues (MP4)

• Video S2: MD simulation of HODA-PC bound to K398 (MP4)

• Video S3: MD simulation of HODA-PC bound to K403 for 750 ns (MP4)

• Video S4: MD simulation of HODA-PC bound to K403 for 500 ns (MP4)

• Video S5: Metadynamics simulation for the POPNC-CD36 complex (MP4)

• Video S6: Metadynamics simulations showing dynamic interactions (MP4)

• Video S7: Metadynamics simulation of oxPL translocation (MP4)

• Video S8: Metadynamics simulation for the HODA-PC-CD36 complex (MP4)

All authors have given approval to the final version of the manuscript.

The authors declare no competing financial interest.