Ligand deconstruction: Why some fragment binding positions are conserved and others are not

Abstract

Fragment-based drug discovery (FBDD) relies on the premise that the fragment binding mode will be conserved on subsequent expansion to a larger ligand. However, no general condition has been established to explain when fragment binding modes will be conserved. We show that a remarkably simple condition can be developed in terms of how fragments coincide with binding energy hot spots--regions of the protein where interactions with a ligand contribute substantial binding free energy--the locations of which can easily be determined computationally. Because a substantial fraction of the free energy of ligand binding comes from interacting with the residues in the energetically most important hot spot, a ligand moiety that sufficiently overlaps with this region will retain its location even when other parts of the ligand are removed. This hypothesis is supported by eight case studies. The condition helps identify whether a protein is suitable for FBDD, predicts the size of fragments required for screening, and determines whether a fragment hit can be extended into a higher affinity ligand. Our results show that ligand binding sites can usefully be thought of in terms of an anchor site, which is the top-ranked hot spot and dominates the free energy of binding, surrounded by a number of weaker satellite sites that confer improved affinity and selectivity for a particular ligand and that it is the intrinsic binding potential of the protein surface that determines whether it can serve as a robust binding site for a suitably optimized ligand.

Keywords: binding hot spot; druggability; fragment library; fragment-based drug discovery; protein–ligand interaction.

Fig. 1.

Bound conformation of ligands and fragments. Compounds are shown in stick representation, superimposed on the main hot spots of their target proteins. Each main hot spot is shown as a transparent surface spanned by the representative probe molecules in the consensus cluster. (A) Chitinase inhibitor argifin (compound 1). (B) Superposition of argifin (green, 1), dimethylguanylurea (cyan, 2), and the monopeptide fragment 3 of argifin (yellow). (C) Phenylmalonate-based inhibitor 4 of human ^pp60Src SH2. (D) Phenylmalonate 5, binding at the main hot spot of ^pp60Src SH2. (E) Inhibitor 6 of the interaction between VHL protein and HIF-1α. (F) N-acetyl-Hyp-N-methyl 7, binding at the main hot spot of the VHL protein. (G) Oxadiazol-based (8, green) and biarylphenylalanine amide-based (10, yellow) inhibitors of DPP-4. (H) Val-Pyr fragment 9, common to both oxadiazol-based and biarylphenylalanine amide-based inhibitors, binding at the main hot spot of DPP-4.

Fig. 2.

Bound conformation of ligands and fragments. Secondary hot spots are shown as clusters of probe cluster representatives (thin sticks). (A) Thrombin inhibitor 11, based on fragment screening and fragment linking. Representative probes of second strongest hot spot (16 probe clusters) are shown as magenta lines. (B) Fragment hits 12 and 13 used for the discovery of thrombin inhibitor 11. The second hot spot supports the binding of fragment 13 (yellow sticks) that also protrudes into the main hot spot. (C) AmpC β-lactamase inhibitor 14. Representative probes of the fourth ranked hot spot (11 probe clusters) are shown as blue lines. (D) AmpC β-lactamase inhibitor fragment 15 (cyan), 16 (yellow), and 17 (magenta). (E) AmpC β-lactamase inhibitor fragment 7-N-formyl-cephalosporanic acid 19. (F) AmpC β-lactamase inhibitor fragment 20, derived from fragment 19. (G) Nutlin-3 (compound 22), an inhibitor of the MDM2:p53 interaction. The second strongest hot spot of MDM2 (22 probe clusters) is shown as magenta lines. (H) Fragment 23, the smallest Nutlin-3 fragment capable of binding to MDM2.

Fig. 3.

Deconstructing an inhibitor of the interaction between IL-2 and its receptor IL-2Rα. (A) Inhibitor 24 of the IL-2:IL-2Rα interaction. Molecules representing the probe clusters in fourth ranked hot spot (10 probe clusters) of ligand-free IL-2 are shown as magenta lines. (B) Fragment 25 was tethered to and cocrystallized with IL-2. (C) Because the binding of fragment 26 was shown by chemical shift perturbations observed in ¹⁵N/¹H HSQC NMR spectra, and no X-ray structure is available, the bound position shown is based on the binding mode of fragment 25. (D) Hypothetical position of fragment 27, which does not bind to IL-2, based on the binding mode of fragment 25.