Avapritinib-based SAR studies unveil a binding pocket in KIT and PDGFRA

Abstract

Avapritinib is the only potent and selective inhibitor approved for the treatment of D842V-mutant gastrointestinal stromal tumors (GIST), the most common primary mutation of the platelet-derived growth factor receptor α (PDGFRA). The approval was based on the NAVIGATOR trial, which revealed overall response rates of more than 90%. Despite this transformational activity, patients eventually progress, mostly due to acquired resistance mutations or following discontinuation due to neuro-cognitive side effects. These patients have no therapeutic alternative and face a dismal prognosis. Notable, little is known about this drug's binding mode and its medicinal chemistry development, which is instrumental for the development of the next generation of drugs. Against this background, we solve the crystal structures of avapritinib in complex with wild-type and mutant PDGFRA and stem cell factor receptor (KIT), which provide evidence and understanding of inhibitor binding and lead to the identification of a sub-pocket (Gα-pocket). We utilize this information to design, synthesize and characterize avapritinib derivatives for the determination of key pharmacophoric features to overcome drug resistance and limit potential blood-brain barrier penetration.

Fig. 1. Crystal structures of type I and type II inhibitors used in GIST treatment bound to target proteins KIT and PDGFRA.

a Overview of the DFG-out conformation of the kinase domain bound to type II inhibitors. Inhibitors imatinib (PDB-ID: 1T46), sunitinib (PDB-ID: 3G0E) and a ripretinib derivative (PDB-ID: 6MOB) are shown as surfaces, illustrating their binding modes in different pockets of the protein. b Two-dimensional representation of the inhibitors approved for GIST treatment. c, f Schematic representation of the inactive DFG-out (c) and the active DFG-in (f) kinase conformations, reveiling that a D842V-mutation in PDGFRA would shift the equilibrium toward the active DFG-in conformation. d Overview of the obtained co-crystal structure of PDGFRA-T674I bound to avapritinib (1) in the DFG-in conformation (PDB-ID: 8PQH). e Highlighting the Gα-pocket addressed by avapritinib.

Fig. 2. Structural analysis of the gatekeeper mutation occurring during avapritinib treatment.

a Complex crystal structure of 1 bound to KIT-wt (PDB-ID: 8PQ9). Herein a water-mediated interaction between the N4 of the pyrrolotriazine scaffold of 1 and Thr670 is clearly visible in this structure. 1 is shown with its |2F_o-Fc|- electron density map at an r.m.s.d. = 1.0. b Complex crystal structure of 1 bound to KIT-T670I (PDB-ID: 8PQG). It is seen that the isoleucine of Ile670 replaces a water molecule observed in the wt structures, leading to drug resistance. 1 is shown with its |2F_o-Fc|- electron density map at an r.m.s.d. = 1.0. c Apo-crystal structure of PDGFRA-wt showing an auto-inhibited DFG-out kinase conformation with the JMD (green) protruding into the backpocket of the protein. In addition, the structure reveals that a water molecule is bound to the gatekeeper amino acid Thr674 (PDB-ID: 8PQJ). d Apo-crystal structure of PDGFRA-T674I showing an auto-inhibited DFG-out kinase conformation with the JMD (green) protruding into the backpocket of the protein. Furthermore, the interaction between a water molecule and the gatekeeper amino acid Ile674 cannot be observed and the water molecule is displaced (PDB-ID: 8PQK).

Fig. 3. Overview of avapritinib (1) and avapritinib-based inhibitors 2-13 for biological evaluation.

Ligands 1-13 have been characterized both biochemically and cellularly in selected kinases and cell systems in order to evaluate their potencies.

Fig. 4. Structural and biochemical analysis of the most active ligands synthesized during this project.

a Schematic representation of avapritinib’s interaction pattern in PDGFRA and KIT. Shown are the main interactions with the hinge region, the catalytic lysine as well as the DFG motif. Further, a water-mediated interaction between ligand and protein (red dot) and the addressed Gα-pocket (green sphere) are indicated. b Comparison of complex crystal structures of 1 (PDB-ID: 8PQ9) and 4 (PDB-ID: 8PQA) bound to KIT-wt. Structures were aligned to their hinge regions. Without the fluorobenzene moiety (4), Phe600 of the Gly-rich loop is rotated inward the binding pocket so the Gα-pocket is covered by the phenyl side chain of the amino acid. c Graphical representation of the superior toxicity profiles of the synthesized ligands compared to avapritinib (1). Red: less selective on the control cell line than avapritinib (1). White: as selective on the control cell line as avapritinib (1). Blue: more selective on the control cell line than avapritinib (1). Ava: avapritinib (1). d Comparison of complex crystal structures of 1 (PDB-ID: 8PQ9) and 10 (PDB-ID: 8PQD) bound to KIT-wt, where Phe600 of the Gly-rich loop is slightly rotated toward the ligand binding pocket. e Comparison of complex crystal structures of 1 (PDB-ID: 8PQ9) and 11 (PDB-ID: 8PQE) bound to KIT-wt, where an additional interaction between ligand and Gly-rich loop can be observed. The regulatory αC-helix is slightly rotated upward. f Comparison of complex crystal structures of 1 (PDB-ID: 8PQ9) and 12 (PDB-ID: 8PQF) bound to KIT-wt, revealing an additional interaction between the carbamates oxygen and the backbone of Phe600 of the Gly-rich loop, which leads to a slight reorientation of the Gly-rich loop toward the ligand binding pocket. g–i Immunoblots to elucidate dose-dependent downregulation of pPDGFRA-D842V in T1-a-D842V cell lines. Cells were treated with DMSO (control) or inhibitors 1 (g–i), 10 (g), 11 (h) and 12 (i) for 24 h, lysed, blotted and incubated with PDGFRA and pPDGFRA/B specific antibodies as well as downstream protein-specific antibodies for evaluation of the inhibition (n = 1 biologically independent experiment).