Resveratrol targets PD-L1 glycosylation and dimerization to enhance antitumor T-cell immunity

Abstract

New strategies to block the immune evasion activity of programmed death ligand-1 (PD-L1) are urgently needed. When exploring the PD-L1-targeted effects of mechanistically diverse metabolism-targeting drugs, exposure to the dietary polyphenol resveratrol (RSV) revealed its differential capacity to generate a distinct PD-L1 electrophoretic migration pattern. Using biochemical assays, computer-aided docking/molecular dynamics simulations, and fluorescence microscopy, we found that RSV can operate as a direct inhibitor of glyco-PD-L1-processing enzymes (α-glucosidase/α-mannosidase) that modulate N-linked glycan decoration of PD-L1, thereby promoting the endoplasmic reticulum retention of a mannose-rich, abnormally glycosylated form of PD-L1. RSV was also predicted to interact with the inner surface of PD-L1 involved in the interaction with PD-1, almost perfectly occupying the target space of the small compound BMS-202 that binds to and induces dimerization of PD-L1. The ability of RSV to directly target PD-L1 interferes with its stability and trafficking, ultimately impeding its targeting to the cancer cell plasma membrane. Impedance-based real-time cell analysis (xCELLigence) showed that cytotoxic T-lymphocyte activity was notably exacerbated when cancer cells were previously exposed to RSV. This unforeseen immunomodulating mechanism of RSV might illuminate new approaches to restore T-cell function by targeting the PD-1/PD-L1 immunologic checkpoint with natural polyphenols.

Keywords: PD-L1; T cells; glycosylation; immunotherapy; resveratrol.

Figure 1

PD-L1 is abnormally glycosylated in response to resveratrol. (A) Representative immunoblot of PD-L1 protein in JIMT-1 cells cultured with or without metformin (1a), phenformin (1b), RSV (2), soraphen A (3), C75 (4), PP242 (5), Torin (6), AICAR (7), or compound C (8). (B, D) Representative immunoblots of PD-L1 protein in JIMT-1 cells cultured in the absence of presence of RSV, tunicamycin (TM), and/or cycloheximide (CHX), chloroquine (CLQ), and/or MG-132. (C) PD-L1 glycosylation patterns obtained from cell lysates of RSV-treated or untreated control cells that were further treated with PNGase F and Endo H and analyzed by western blotting (C: untreated control; black arrowhead, glycosylated PD-L1; *orange arrowhead, abnormal glycosylated PD-L1 form; **orange arrowhead, non-glycosylated PD-L1).

Figure 2

Resveratrol alters PD-L1 N-glycosylation in a SIRT1-, AMPK-, and GSK3β-independent manner. (A) Representative immunoblot of SIRT1, acetyl-p53 (Lys382), and PD-L1 in JIMT-1 cells cultured with or without RSV in the absence or presence of the AMPK inhibitor compound C, the SIRT1 inhibitor EX-527, and the GSK3β inhibitors AR-18 and LiCl. (B) Schematic representation of the biosynthesis and processing PD-1 N-linked glycosylation pathway, showing the sites of action of well-known glycoprotein-processing enzymes inhibitors. RSV is proposed to operate as a direct inhibitor of GAA and/or α-Man I enzymatic activities.

Figure 3

Resveratrol is predicted to bind the catalytic site of yeast GAA. Surface and backbone representations of yeast GAA showing the computationally predicted location of acarbose (A) and RSV (B) clusters. “Site A” refers to the pocket containing the proposed catalytic residues of yeast GAA whereas “site B” refers to a second cavity roughly 12 Å away from the active site pocket of yeast GAA [57]. (C) A detailed map of the molecular interactions of RSV in each cluster before (0 ns) and after 100 ns of molecular dynamics simulation. Each inset shows the detailed interactions of each RSV cluster docked to yeast GAA using the PLIP algorithm [124], indicating the participating amino acids involved in the interaction and the type of interaction (hydrogen bonds, hydrophilic interactions, salt bridges, Π-stacking, etc). Figures were prepared using PyMol 2.3 software.

Figure 4

Resveratrol is predicted to bind allosteric sites in the surface of human GAA. Surface and backbone representations of human GAA showing the computationally-predicted location of RSV (top panels) and acarbose (bottom panels) clusters. A detailed map of the molecular interactions of RSV and acarbose in each cluster before (0 ns) and after 100 ns of molecular dynamics simulation. Each inset shows the detailed interactions of each RSV/acarbose cluster docked to human GAA using the PLIP algorithm [124], indicating the participating amino acids involved in the interaction and the type of interaction (hydrogen bonds, hydrophilic interactions, salt bridges, Π-stacking, etc). Figures were prepared using PyMol 2.3 software.

Figure 5

Resveratrol is predicted to bind the catalytic site of human α-mannosidases. Surface and backbone representations of human α-Man I (A) and homology model of human α-Man II (B) showing the computationally predicted location of RSV clusters. A detailed map of the molecular interactions of RSV in each cluster before (0 ns) and after 100 ns of molecular dynamics simulation. Each inset shows the detailed interactions of each RSV cluster docked to human GAA using the PLIP algorithm [124], indicating the participating amino acids involved in the interaction and the type of interaction (hydrogen bonds, hydrophilic interactions, salt bridges, Π-stacking, etc). The white inset in A shows a surface and backbone representations of human α-Man I docked to the α-Man I inhibitor kifunensine. Figures were prepared using PyMol 2.3 software.

Figure 6

Resveratrol alters the sub-cellular expression pattern of PD-L1. Representative immunofluorescence staining of PD-L1 in JIMT-1 (left panels) and MDA-MB-231 (right panels) breast cancer cells cultured in the absence or presence of RSV, using an antibody directed against either an intracellular epitope (top panels) or an extracellular domain epitope (bottom panels) of PD-L1.

Figure 7

Resveratrol enhances the susceptibility to T-cell-mediated tumor cell killing. T-cell-mediated cell death of JIMT-1 cells pre-cultured in the absence of presence of RSV was measured using the xCELLigence system. Shown are the mean (±SD, n=3) of % lysis values calculated from the impedance-based lysis assay at 5 and 10 hours following the addition of T-cells. Also shown are microphotographs of representative T-cell-mediated cancer cell killing assays in which tumor cells were subjected to crystal violet staining. Statistical analysis was performed using GraphPad Prism 7, using two-way ANOVA with Sidak’s multiple comparison post-test, comparing untreated versus RSV-treated per time (* = P < 0.01).

Figure 8

Resveratrol is predicted to bind the PD-1 dimer interface. Top Backbone representation of the PD-L1:PD-L1 dimer showing the computationally-predicted location of RSV (yellow) and BMS-202 (cyan). Chain A shows the location of the four Asn residues that can be glycosylated. The insets show the detailed maps of the molecular interactions of RSV with the amino acids at the hydrophobic pocket accommodating the BMS-202 inhibitor and formed at the PD-L1 dimer surface before (0 ns) and after 100 ns of molecular dynamics (MD) simulation, indicating the participating amino acids involved in the interaction and the type of interaction (hydrogen bonds, hydrophilic interactions, salt bridges, Π-stacking, etc). Bottom. Left. Trajectory of the RSV-forming complex with the PD-L1 dimer. Right. Molecular Mechanics/Poisson-Boltzmann Surface Area free energy analysis of the PD-L1 dimer forming a complex with RSV using YASARA dynamics v19.9.17 software. The best-docked complex as the initial conformation for MD simulation followed by 1000 snapshots (100 ns) obtained from the MD trajectory were employed to calculate the values of free energy binding of RSV. Additionally, the average value calculated for the last 200 snapshots (20 ns) is also displayed. YASARA-calculated binding energy provides positive values when the predicted binding is strong and stable whereas negative values indicate no binding. Figures were prepared using PyMol 2.3 software.

Figure 9

Resveratrol enhances antitumor T cell immunity by promoting abnormal glycosylation and dimerization of PD-L1. Post-translational modifications such as glycosylation, phosphorylation, palmitoylation or ubiquitination are essential for the folding, intracellular transport, and stabilization of the PD-L1 protein [, –114]. PD-L1 is highly glycosylated, and N-linked glycosylation of PD-L1 critically maintains its protein stability and is required for its interaction with PD-1 to efficiently suppress T-cell activity. Recently, Bristol-Myers Squibb (BMS)-developed compounds with a common (2-methyl-4-biphenylyl)methanol scaffold have been reported to block the PD-1/PD-L1 interaction by interacting with the cavity formed by the two PD-L1 monomers and inducing the dimerization of PD-L1 [–77]. Structural studies have revealed a dimeric protein complex with a single small molecule that stabilizes the dimer and thereby occludes the PD-1 interaction surface of PD-L1. We propose that RSV is a naturally occurring, double-strike PD-1/PD-L1 immune checkpoint inhibitor capable of directly blocking the enzymatic machinery in charge of the N-linked glycosylation of the nascent PD-L1 at the endoplasmic reticulum or directly binding to PD-L1 surfaces to induce PD-L1 dimerization and block PD-1 binding. This unforeseen ability of RSV to directly interfere with PD-L1 stability and trafficking impedes the correct targeting of PD-L1 to the cancer cell plasma membrane and ultimately elicits drastically enhanced cytotoxic T-lymphocyte immune-surveillance against tumor cells. These findings might illuminate new approaches to restore T-cell function by targeting the PD-1/PD-L1 immunologic checkpoint with natural polyphenols.