Systematic Variation of Both the Aromatic Cage and Dialkyllysine via GCE-SAR Reveal Mechanistic Insights in CBX5 Reader Protein Binding

Abstract

Development of inhibitors for histone methyllysine reader proteins is an active area of research due to the importance of reader protein-methyllysine interactions in transcriptional regulation and disease. Optimized peptide-based chemical probes targeting methyllysine readers favor larger alkyllysine residues in place of methyllysine. However, the mechanism by which these larger substituents drive tighter binding is not well understood. This study describes the development of a two-pronged approach combining genetic code expansion (GCE) and structure-activity relationships (SAR) through systematic variation of both the aromatic binding pocket in the protein and the alkyllysine residues in the peptide to probe inhibitor recognition in the CBX5 chromodomain. We demonstrate a novel change in driving force for larger alkyllysines, which weaken cation-π interactions but increases dispersion forces, resulting in tighter binding. This GCE-SAR approach establishes discrete energetic contributions to binding from both ligand and protein, providing a powerful tool to gain mechanistic understanding of SAR trends.

Figure 1.

A. Structures of UNC6212 (Kme2, orange), UNC6349 ( Ket2, pink), and UNC6864 (Kei, purple) peptide-based ligands. All ligands share the same scaffold and vary only in the alkylation on lysine. B. Aromatic residues (Y20, W41, and F44) in the CBX5 chromodomain aromatic cage bound to a H3K9me3 histone peptide (PDB: 3FDT). Y20 (green) and F44 (blue) were targeted in this study.

Figure 2.

The effect of R-groups on electrostatic surface potential (ESP) maps and calculated cation-π binding energy (CπBE, kcal/mol) of amino acids tested within the aromatic cage. ESP maps were calculated in Spartan at the ωB97X-D/6–31G(d) level of theory. CπBE are calculated for substituted benzenes and Na⁺. Measured K_D values (μM) for CBX5 variants at positions Y20 and F44 with peptide-based ligands, UNC6212 (Kme2), UNC6349 (Ket2), and UNC6864 (Kei). Affinities corresponding to wild-type protein are shown in bold. Binding affinity of the F44pCF₃Phe variant was not determined. Errors given for K_D are the standard deviation for 3 experimental replicates or the highest individual error from an individual experiment among the replicates, whichever is greater.

Figure 3.

Linear free energy relationship (LFER) plots analyzing the correlation between ΔG_binding with calculated cation-π binding energies (CπBE) for a range of Y20 (green) and F44 (blue) variants binding peptidomimetic ligands UNC6212 (Kme2), UNC6349 (Ket2), and UNC6864 (Kei) from left to right. Error bars reflect the standard deviation for replicates or the highest individual error from an individual experiment among the replicates, whichever is greater. The error in the slopes is ±0.01 for both Y20 and F44 with UNC6212 (Kme2), ±0.01 for Y20 and ±0.02 for F44 with UNC6349 (Ket2), and ±0.01 for Y20 and ±0.02 for F44 with UNC6864 (Kei).

Figure 4.. Structural analysis and computational modeling of CBX5.

A. Measured distances (Å) between the center of Y20 (green) and F44 (blue) with Kme3 substituents from H3K9me3. B and C. Contact surface of Kme3 with Y20 (B) and F44 (C) viewed normal to the plane of the ring. Interaction energies (E_int) for Kme3 and each aromatic residue are shown and were calculated at the M06–2X/6–311+G(d,p) level of theory. In all panels, the structure of CBX5 bound to H3K9me3 is used (PDB: 3FDT).

Figure 5.

Structural comparison of CBX5 bound to Kme3 with CBX7 bound to UNC3866 containing Ket2. A. The overlay of the aromatic cages of CBX5 (PDB: 3FDT; with Y20 in green and F44 in blue) bound to H3K9me3 (yellow) and CBX7 bound to UNC3866 (containing Ket2; PDB: 5EPJ, salmon) suggest how dialkylated substrates would bind in the aromatic cage of CBX5. A conserved water-bridged hydrogen bond between di-alkylated substrate and the carbonyl of Y39 in CBX7 (H48 in CBX5) is shown with distances (Å). B. Position of Ket2 from UNC3866 bound to CBX7 as a model for binding in CBX5. Distances (Å) from ligand substituents with respect to the center of the ring of Y20 and F44 (CBX5, grey dashed lines) and F11 and W35 (CBX7, salmon dashed lines) are shown. View rotated ~90° with respect to A. W41/W32 and H48/Y39 are removed for clarity.

Figure 6.

Partial charges calculated for Kme2, Ket2, and Kei lysine substituents using CM5 (charge model 5) charges calculated at the M06–2X/6–311+G(d,p) level of theory.

Figure 7.

Relationship between ΔG_binding and substituent characteristics for pCNPhe variant. Plots analyzing the correlation between ΔG_binding and electrostatic potential, Log P, and polarizability for the peptidomimetic ligands UNC6212 (Kme2), UNC6349 (Ket2), and UNC6864 (Kei) for Y20 (green) and F44 (blue) positions with respect to the ΔG_binding for the pCNPhe variant are shown. Electrostatic potential, Log P, and polarizability were calculated in Spartan at the ωB97X-D/6–31G(d) level of theory. ΔG_binding error bars reflect the standard deviation for replicates or the highest individual error from an individual experiment among the replicates, whichever is greater. The same trends are observed for the pClPhe variant (Figure S10).