Crystal structure of the FLT3 kinase domain bound to the inhibitor Quizartinib (AC220)

Abstract

More than 30% of acute myeloid leukemia (AML) patients possess activating mutations in the receptor tyrosine kinase FMS-like tyrosine kinase 3 or FLT3. A small-molecule inhibitor of FLT3 (known as quizartinib or AC220) that is currently in clinical trials appears promising for the treatment of AML. Here, we report the co-crystal structure of the kinase domain of FLT3 in complex with quizartinib. FLT3 with quizartinib bound adopts an "Abl-like" inactive conformation with the activation loop stabilized in the "DFG-out" orientation and folded back onto the kinase domain. This conformation is similar to that observed for the uncomplexed intracellular domain of FLT3 as well as for related receptor tyrosine kinases, except for a localized induced fit in the activation loop. The co-crystal structure reveals the interactions between quizartinib and the active site of FLT3 that are key for achieving its high potency against both wild-type FLT3 as well as a FLT3 variant observed in many AML patients. This co-complex further provides a structural rationale for quizartinib-resistance mutations.

Fig 1. FLT3 with quizartinib bound adopts an inactive conformation.

(A) General kinase features are illustrated on the co-crystal structure of FLT3 bound to quizartinib. The kinase is in light blue with the P loop is shown in green, the αC helix is shown in dark blue, and the activation loop is shown in red. (B) Structural changes in kinase domain of FLT3 that occur upon receptor activation are illustrated. The active conformation of FLT3 is modeled from the active c-Kit structure (PDB 1PKG).

Fig 2. An induced fit of the activation loop of FLT3 with quizartinib bound.

(A) An overlay of the kinase domain of FLT3 from the FLT3:quizartinib co-crystal structure (using the color scheme denoted in Fig 1) with the autoinhibited FLT3 (shown in grey, PDB 1RJB). (B) A detailed view of the activation loops on both the FLT3:quizartinib (red) and the autoinhibited FLT3 (grey) structures. The left panel highlights the induced fit of Phe 830 in the DFG motif due to drug binding. Quizartinib (AC220) is included in this zoomed-in view. A 60° rotation of the superposition of the two activation loops is shown in the right panel. (C) A molecular dynamics simulation on the autoinhibited FLT3 with the juxtamembrane segment deleted is illustrated with an overlay of the initial crystal structure in grey and the instantaneous structure at t = 89 ns in light orange. (D) The stability of the activation loop in a molecular dynamics simulation on the FLT3:quizartinib co-complex is illustrated with an overlay of the initial crystal structure in red/ blue and the instantaneous structure at t = 89 ns in light orange. Only the chemical structure of quizartinib (yellow) from the initial structure is shown for clarity. (E) The collapse of the activation loop upon removal of quizartinib in a molecular dynamics simulation (light orange) is illustrated as a comparison to the activation loop in the co-crystal structure (red/ blue).

Fig 3. Quizartinib in the FLT3 co-complex matches the second-ranked docking pose.

(A) An overlay of quizartinib from the co-crystal structure (yellow) with the top-ranked docking pose (white) is shown. (B) An overlay of quizartinib from the co-crystal structure (yellow) with the second-ranked docking pose (white) is shown.

Fig 4. Interactions between FLT3 and quizartinib.

(A) Chemical structure of quizartinib (AC220). (B) An unbiased electron density map (2mF_O-DF_C) of quizartinib (yellow) contoured at 1.0 σ (light blue). A simulated annealing refinement in Phenix on FLT3 with quizartinib deleted resulted in a model that was used to calculate the electron density for quizartinib. The simulated annealing was performed with an initial temperature of 5000 K to a final temperature of 300 K over 50 steps. A superposition of quizartinib with this unbiased electron density map for the compound is shown for clarity. (C) The structure of the FLT3 kinase domain bound to quizartinib (left) with a zoomed-in view of the active site (right). (D) A detailed view of the interactions between FLT3 and quizartinib.

Fig 5. The juxtamembrane segment of FLT3 clashes with quizartinib.

(A) An overlay of the FLT3:quizartinib crystal structure (color scheme depicted in Fig 1A) with the autoinhibited FLT3 (shown in grey, PDB 1RJB). The zoomed-in view of the juxtamembrane segment highlights clashes between the juxtamembrane segment and quizartinib. (B) An overlay of the crystal structure of the VEGFR2 kinase domain bound to sorafenib (shown in red/ blue, PDB 4ASD) and the VEGFR2 kinase domain bound to axitinib (shown in grey, PDB 4AGC). Only sorafenib is illustrated in the active site for clarity. The juxtamembrane segment from the VEGFR2:sorafenib co-crystal structure is shown in orange and is extended away from the active site. The juxtamembrane segment from the VEGFR:axitinib co-crystal structure remains bound to the kinase domain. The detailed view illustrates similar clashes between sorafenib and the VEGFR2 bound conformation of the juxtamembrane segment.

Fig 6. Mutations in the kinase domain of FLT3 in acute myeloid leukemia (AML).

(A) Mutations observed in AML (pink spheres) that show resistance to quizartinib are mapped onto the kinase domain of FLT3. (B) The left panel illustrates the position of a hydrophobic pocket on a homology model of FLT3, which was generated from the active conformation of c-Kit (PDB 1PKG). The zoomed-in view in the middle panel illustrates the detailed interactions between the activation loop and a hydrophobic pocket on the model of FLT3. The right panel highlights an extension of the hydrophobic patch with the disease mutation, D835V. (C) A similar interaction between a hydrophobic pocket on the kinase domain and the activation loop is observed in the crystal structure of the active conformation of LCK (PDB 3LCK).