Defining a new nomenclature for the structures of active and inactive kinases

Abstract

Targeting protein kinases is an important strategy for intervention in cancer. Inhibitors are directed at the active conformation or a variety of inactive conformations. While attempts have been made to classify these conformations, a structurally rigorous catalog of states has not been achieved. The kinase activation loop is crucial for catalysis and begins with the conserved DFGmotif. This motif is observed in two major classes of conformations, DFGin-a set of active and inactive conformations where the Phe residue is in contact with the C-helix of the N-terminal lobe-and DFGout-an inactive form where Phe occupies the ATP site exposing the C-helix pocket. We have developed a clustering of kinase conformations based on the location of the Phe side chain (DFGin, DFGout, and DFGinter or intermediate) and the backbone dihedral angles of the sequence X-D-F, where X is the residue before the DFGmotif, and the DFG-Phe side-chain rotamer, utilizing a density-based clustering algorithm. We have identified eight distinct conformations and labeled them based on the Ramachandran regions (A, alpha; B, beta; L, left) of the XDF motif and the Phe rotamer (minus, plus, trans). Our clustering divides the DFGin group into six clusters including BLAminus, which contains active structures, and two common inactive forms, BLBplus and ABAminus. DFGout structures are predominantly in the BBAminus conformation, which is essentially required for binding type II inhibitors. The inactive conformations have specific features that make them unable to bind ATP, magnesium, and/or substrates. Our structurally intuitive nomenclature will aid in understanding the conformational dynamics of kinases and structure-based development of kinase drugs.

Keywords: cell signaling; protein kinases; structural bioinformatics.

Fig. 1.

Structure of a typical protein kinase domain displaying ATP binding site and conserved elements around it (INSR kinase, PDB ID code 1GAG).

Fig. 2.

Analysis of catalytically primed structures. (A) Catalytically primed structure of insulin receptor kinase (PDB ID code 3BU5) with bound ATP and Mg²⁺ ion. Important interactions are marked with dashed lines. (B) Ramachandran plot of catalytically primed structures (those with bound ATP and Mg²⁺, a phosphorylated activation loop, and resolution ≤2.25 Å): X-DFG (residue before the DFGmotif, blue), DFG-Asp (green), DFG-Phe (orange), and DFG-Gly (red). The Ramachandran regions are marked A (alpha), B (beta), L (left), and E (epsilon). (C) Scatterplot of side-chain dihedral angles χ₁ and χ₂ for DFG-Asp (green) and DFG-Phe (orange) of catalytically primed structures. Asp is in a trans rotamer (χ₁ ∼ 180°) and Phe is in a g⁻ rotamer (χ₁ ∼ 300° or −60°).

Fig. 3.

Locations of Phe side chain in DFGin, DFGinter, and DFGout structures of kinases. (A) Positions of DFG-Phe side chain in DFGin (cyan), DFGinter (orange), and DFGout (purple) of EGFR. (B) Distances, D1 and D2, used to identify the location of the Phe side chain of the DFGmotif. (C) Scatterplot of D1 and D2. DFGin (cyan), DFGinter (orange), and DFGout (purple) structures identified with hierarchical clustering; catalytically primed structures (cyan triangles) and structures with bound type II inhibitors (purple triangles) are marked.

Fig. 4.

DBSCAN clustering of dihedral angles that position the DFG-Phe side chain. Clustering was performed on the DFGin, DFGinter, and DFGout groups separately with an angular metric of the backbone dihedral angles of the X-DFG, DFG-Asp, and DFG-Phe residues and the χ₁ of DFG-Phe. Clusters are named by the Ramachandran regions of the X-DFG, DFG-Asp, and DFG-Phe residues and the χ₁ rotamer of DFG-Phe: g⁻ (χ₁ ∼ −60°, minus, magenta); g⁺ (χ₁ ∼ +60°, plus, blue); trans rotamer (χ₁ ∼ 180°, trans, green).

Fig. 5.

Structure examples of each cluster. (A) BLAminus; (B) BLAplus; (C) ABAminus; (D) BLBminus; (E) BLBplus; (F) BLBtrans; (G) BBAminus (DFGout); (H) BABtrans (DFGinter).

Fig. 6.

Electron densities indicate incorrectly modeled ABAminus structures. (A) Electron density of ABAminus structures consistent with BLAminus structure. 2Fo-Fc (gray); Fo-Fc (green, positive density, indicating density not represented by an atom; red, negative density, indicating density where an atom has been placed without electron density support). PDB ID code 1K3A_A (IGF1R); 4BL1_A (MELK); 3IEC_D (MARK2). (B) Kernel density estimates of the EDIA scores of the backbone carbonyl atom (X_O) of the X-DFG residue for ABAminus (orange) and BLAminus (black) structures. (C) Scatterplot and von Mises kernel regressions of EDIA of X_O vs. χ₂ of DFG-Asp of ABAminus (orange) and BLAminus (black) structures. BLAminus structures mostly have χ₂ near 0°. ABAminus structures with χ₂ near 0° have poor electron density and may be mismodeled BLAminus structures.

Fig. 7.

Inhibitors bound to structures in multiple clusters: BLAminus (sky blue), BLAplus (dark green), BLBminus (light gray), BLBplus (light green), BBAminus (purple), BABtrans (orange), DFGout-noise (light purple), and DFGinter-noise (light orange). (A) Sunitinib bound to PAK6, PHKG2, and STK24 in BLAminus; ITK in BLAplus; CDK2 in BLBminus; KIT in BBAminus; KIT in DFGout-noise. (B) Dasatinib bound to ABL1, STK10, and STK24 in BLAminus; BTK in BLAplus; EPHA2 in BLBminus; PTK6 in BLBplus; BMX and BTK in BABtrans; ABL1 in DFGinter-noise; DDR1 in DFGout-noise. (C) Bosutinib bound to EPHA2, PMYT1, SRC in BLAminus; ERBB3 in BLBplus; STK10 in DFGin-noise; ABL1 and KCC2A in DFGinter-noise. (D) Imatinib bound to ABL1, ABL2, CSF1R, DDR1, KIT, and LCK in BBAminus.

Fig. 8.

The intact regulatory spine of active and inactive structures. (A) Active BLAminus (3BU5_A, INSR); (B) inactive BLAplus (4F64_A, FGFR1); (C) inactive BLBminus (5NK7_A, EPHA2); (D) inactive BLBplus (4UY9_B, M3K9). We consider the spine intact if there is at least one atom–atom contact (≤4.5 Å) between the side chains of each residue of the spine.