Identifying and Assessing Putative Allosteric Sites and Modulators for CXCR4 Predicted through Network Modeling and Site Identification by Ligand Competitive Saturation

Abstract

The chemokine receptor CXCR4 is a critical target for the treatment of several cancer types and HIV-1 infections. While orthosteric and allosteric modulators have been developed targeting its extracellular or transmembrane regions, the intramembrane region of CXCR4 may also include allosteric binding sites suitable for the development of allosteric drugs. To investigate this, we apply the Gaussian Network Model (GNM) to the monomeric and dimeric forms of CXCR4 to identify residues essential for its local and global motions located in the hinge regions of the protein. Residue interaction network (RIN) analysis suggests hub residues that participate in allosteric communication throughout the receptor. Mutual residues from the network models reside in regions with a high capacity to alter receptor dynamics upon ligand binding. We then investigate the druggability of these potential allosteric regions using the site identification by ligand competitive saturation (SILCS) approach, revealing two putative allosteric sites on the monomer and three on the homodimer. Two screening campaigns with Glide and SILCS-Monte Carlo docking using FDA-approved drugs suggest 20 putative hit compounds including antifungal drugs, anticancer agents, HIV protease inhibitors, and antimalarial drugs. In vitro assays considering mAB 12G5 and CXCL12 demonstrate both positive and negative allosteric activities of these compounds, supporting our computational approach. However, in vivo functional assays based on the recruitment of β-arrestin to CXCR4 do not show significant agonism and antagonism at a single compound concentration. The present computational pipeline brings a new perspective to computer-aided drug design by combining conformational dynamics based on network analysis and cosolvent analysis based on the SILCS technology to identify putative allosteric binding sites using CXCR4 as a showcase.

Figure 1

Low-frequency global motions (dynamic domains colored in wheat and purple) and the high-frequency local motions in the 10 fastest modes (shown as green surface) of CXCR4 (a) monomer and (b) homodimer predicted by GNM. The orthosteric site consists of major and minor binding pockets, shown in red and cyan, respectively.

Figure 2

Hub residues with high betweenness predicted by RIN for monomer (a) and homodimer (b) are shown in blue. High-frequency fluctuating residues at the hinge regions determined with GNM are in green, and residues with known critical functions are colored in orange.

Figure 3

(a) Orthosteric site of CXCR4 and its inhibitor 1Tit, potential allosteric sites for monomer, MA1, and MA2 and (b) potential allosteric sites for the homodimer CXCR4, DA1, DA2, and DA3. Both in (a) and (b), the Hotspots are shown in vdW spheres colored based on mean LGFE score (from red—most favorable to blue—least favorable); SILCS FragMaps are visualized as, positive (cyan, −1.2 kcal/mol), negative (orange, −1.2 kcal/mol), apolar (green, −1.2 kcal/mol), H-bond donor (blue, −0.9 kcal/mol), and H-bond acceptor (red, −0.9 kcal/mol).

Figure 4

Identification of hit compound clusters by hierarchical clustering analysis. (a) Bivariate plot of the clustering data shown in the dendrogram in (b). The Roman numerals correspond to the clusters in panels (a) and (b). Effects of hit compounds on binding of (c) 12G5 and (d) CXCL12 to CXCR4 on the CXCR4+Cf2Th cell line. (e) Tango CXCR4-bla agonism assay using CXCL12 (10 nM) as the positive control. (f) Tango CXCR4-bla antagonism assay using CXCL12 (10 nM) as the positive control and CXCL12 (10 nM) plus AMD1300. The compound indices (1–20) are given in Table 1.