AP-5258

CD36 is a transmembrane glycoprotein involved in lipid uptake and signal transduction, playing a crucial role in various physiological and pathological processes. Structurally, it is composed primarily of an ectodomain (residues 30-439), which is essential for ligand binding and facilitating Long-Chain Fatty Acid (LCFA) uptake. Two lysine residues (K164 and K166) have been proposed to contribute significantly to LCFA internalization. Elucidating the ligand-binding cavities within CD36 can provide structural insights that may serve as a foundation for drug design.

Search for probable CD36 ligands A rational search was conducted using scientific databases including PubMed (https://pubmed.ncbi.nlm.nih.gov/), Scopus (https://www.scopus.com/search/form.uri?display=basic#basic), SciELO (https://scielo.org/es/), and Google Academics (https://scholar.google.es/schhp?hl=es). The keywords used were: CD36 activity, CD36 ligands, docking, and MD simulations. From the search results, only studies that did not involve in silico simulations (which could introduce bias into our analysis) were selected. Instead, preference was given to sources containing robust experimental biological data supporting CD36 modulation. This approach identified three derivatives from N'-(2-hydroxy-4,6-dimethoxybenzylidene)acetohydrazide scaffold (AP5055, AP5156, and AP5258), two polyphenolic compounds (Puerarin and Salvianolic acid), and two well-known inhibitors of this receptor (SSO and MTN) as potential ligands for CD36. Binding pocket and LCFA cavity detection The CD36 structure (PDB entry: 5LGD) was obtained from RCSB-PDB (https://www.rcsb.org/). After detailed visualization, all non-essential molecules unrelated to the receptor’s inherent activity (such as water molecules, cofactors, metals, and sugars) were removed. Particularly, the two co-crystallized PLM molecules were eliminated to avoid interference with the analysis, as well as the CIDRa domain from MCvar1 PfEMP1. For this purpose, the protein structure was used without further processing. Detection of the probable LCFA (long-chain fatty acids) acceptor cavities and small-molecules acceptor cavity in CD36 was carried out using two specialized web servers, CavityPlus (http://www.pkumdl.cn:8000/cavityplus/index.php#/) and POCASA (https://g6altair.sci.hokudai.ac.jp/g6/service/pocasa/). CavityPlus is a geometry-based method that uses a probe-sphere model to scan the protein surface, mimicking the kinetic behavior of small molecules attempting to bind the protein. Hence, all the search parameters were kept at their default settings. Briefly, the algorithm first creates 3D grid points encompassing the entire protein, with a spacing of 0.5 Å in each grid point. A 10 Å radius probe sphere then identifies grid points it contacts, labeling them as "outside grid points," while the remaining points are marked as "vacant grid points," representing putative cavity spaces. Refinement of these cavities is performed using the “layer and depth” concepts combined with a “shrink-and-expand” algorithm37. The POCASA algorithm was applied according to our previous methodology38. Ligand preparation Ligand structures were initially generated in 3D conformations using ChemSketch software (https://www.acdlabs.com/resources/free-chemistry-software-apps/chemsketch-freeware/). These structures were then converted to Z-matrix format using GaussView 5.039 to delimit a geometrical and energetic minimization and correct proton assignment, using Gaussian 0940, under a semi-empirical AM1 method. The resulting output files were converted to .pdb format for subsequent docking calculations in Autodock (v4.2.6)41. For docking simulations with MOE42, ligands were prepared using the same protocol but employed in .mol2 format. Molecular docking protocols Two methodological strategies were employed to validate and strengthen our analysis. Specifically, MOE and AutoDock (v4.2.6) programs were used to evaluate and characterize the potential binding interactions between the tested ligands and CD36. For both software programs, the docking simulations were designed to model two receptor states: CD36-unoccupied (receptor without bound PLM molecules) and CD36-occupied (receptor bound to two PLM molecules), based on the crystal structure of CD36 available in the RCSB-PDB (PDB entry: 5LGD). MOE program was selected as the initial docking tool, considering the Triangle Matcher algorithm for initial ligand placement and the Rigid Receptor for refinement. The scoring functions used were the Affinity dG for initial pose generation (computing 30 poses), and GBVI/WSA dG for final refinement (computing 5 poses). As a complementary approach, the AutoDock (v4.2.6) program was employed, using parameters previously reported by our research group38, with a minor modification: the grid box was set at 126 Å3 across the X, Y, and Z dimensions (blind docking procedure), thereby encompassing nearly the entire ectodomain of the protein. Ligand-receptor interactions were visualized and analyzed using the Ligand interactions tool in MOE (for 2D representations, data not shown) and Pymol v0.9943 (for 3D structural interpretation). Molecular dynamics simulation with MMGBSA approach Receptor-ligand complexes were prepared using antechamber and tleap modules of the AMBER22 software package (Case et al., 2005). Systems were solvated using the TIP3P water model and neutralized with counterions by submerging them in a truncated octahedral water at 12 Å. The protein, water, and ions were defined by the ff14SB AMBER force field44. The General Amber force field (GAFF) was taken into consideration when defining the force field characteristics of ligands using AM1-BCC atomic charges45. Four thousand conjugate gradient cycles and five thousand steepest descent cycles made up minimization. Under an NVT ensemble and considering a Berendsen thermostat, the temperature was raised for 200 ps from 0 to 310 K46. A Langevin thermostat47-48 and a Berendsen barostat46 were used to equilibrate the density for 200 ps under an NTP ensemble. Using a Langevin thermostat and a Berendsen barostat, 100 ns long molecular dynamics (MD) simulations with a time step of 2 fs were run at 310 K under an NPT ensemble. The Particle Mesh Ewald approach was used to define the electrostatic term, while a cut-off of 10 Å was taken into consideration for the non-bonded interactions49. The SHAKE algorithm50 was used to limit bond lengths at specific equilibrium. The cpptraj function in Amber22 software was used to analyze the root means squared deviation (RMSD), radius of gyration (Rg), root means squared fluctuations (RMSF), and clustering analysis. To find the binding free energy (ΔGbind), the Molecular Mechanics Generalized Born Surface Area (MMGBSA) method51 was used. Using implicit solvent models52, a 0.10 M salt concentration, and a total of 2000 snapshots, this analysis was conducted over the final 20 ns of equilibrated simulation time. As previously mentioned, the ΔGbind values were calculated53.

In this study, we employed a combination of molecular docking, molecular dynamics simulations, and cavity-detection algorithms to investigate the structural basis of CD36 interactions with small molecules and fatty acids.

We identified three main ligand-binding regions: a primary LCFA-binding cavity, a secondary LCFA-binding cavity, and a Small-Molecule Acceptor Cavity (SMAC). We evaluated three derivatives of the N-(2-hydroxy-4,6-dimethoxybenzylidene)acetohydrazide scaffold (AP5055, AP5156, and AP5258), two polyphenolic compounds (Puerarin and Salvianolic acid), and two well-known CD36 inhibitors (SSO and MTN).

Our findings implicate residues K334, E335, and R337 in disrupting LCFA uptake, while residues T195, L200, F201, Y202, P203, T207, A208, D209, Y230, and K231 delineate the initial segment of the SMAC.

This study highlights critical residues and structural features involved in ligand recognition by CD36. The identification of the SMAC and its role in modulating LCFA uptake provides promising insights for the development of therapeutic agents targeting CD36-related pathologies, including cancer, diabetes, and metabolic disorders.