Identification of a hidden strain switch provides clues to an ancient structural mechanism in protein kinases

Abstract

The protein kinase catalytic domain contains several conserved residues of unknown functions. Here, using a combination of computational and experimental approaches, we show that the function of some of these residues is to maintain the backbone geometry of the active site in a strained conformation. Specifically, we find that the backbone geometry of the catalytically important HRD motif deviates from ideality in high-resolution structures and the strained geometry results in favorable hydrogen bonds with conserved noncatalytic residues in the active site. In particular, a conserved aspartate in the F-helix hydrogen bonds to the strained HRD backbone in diverse eukaryotic and eukaryotic-like protein kinase crystal structures. Mutations that alter this hydrogen-bonding interaction impair catalytic activity in Aurora kinase. Although the backbone strain is present in most active conformations, several inactive conformations lack the strain because of a peptide flip in the HRD backbone. The peptide flip is correlated with loss of hydrogen bonds with the F-helix aspartate as well as with other interactions associated with kinase regulation. Within protein kinases that are regulated by activation loop phosphorylation, the strained residue is an arginine, which coordinates with the activation loop phosphate. Based on analysis of strain across the protein kinase superfamily, we propose a model in which backbone strain co-evolved with conserved residues for allosteric control of catalytic activity. Our studies provide new clues for the design of allosteric protein kinase inhibitors.

Fig. 1.

Protein kinase active site showing structural interactions associated with catalytic and EPK-ELK component residues. Catalytic residues are colored pink and EPK-ELK component residues are colored orange. The HRD-Arg and phospho-threonine unique to EPKs are shown in green. All nitrogen atoms are colored blue and oxygen atoms are colored red. ATP is shown in a “lines” representation and the substrate serine is shown in yellow. A conserved water molecule seen in all active kinase conformations between HRD and DFG motifs is shown in gray, and the hydrogen bonds made by this water are shown as blue dashed lines. The figure was generated based on a CDK-substrate complex (PDB ID 1QMZ) using PyMOL. The EPK-ELK component residues in Aurora kinase are: E-helix-His (H248), F-helix-Asp (D311), and HRD-His (H254). The corresponding residues in PKA are: E-helix-His (H158), F-helix-Asp (D220), and HRD-His (Y164).

Fig. 2.

Conformational strain in the HRD-Arg backbone. (A) Ramachandran plot showing the torsion angle values (ϕ/ψ) of the HRD-Arg residue in 103 high-resolution structures (<1.7 Å) from diverse families. Contours in the Ramachandran plot are based on the definition of Molprobity (35, 36). The blue faint contour indicates the “favored” regions (enclosing 98% of observed conformations) and the red outer contour indicates the “disfavored but allowed” regions (enclosing 99.8% of observed conformations). Torsion-angle values of the HRD-Arg are indicated by dots. In 89 of the 103 high-resolution structures, the torsion-angle values of the HRD-Arg occur in the disfavored region and are defined as strained. (B) Conformational strain in the HRD-Arg backbone is supported by nonideal bond-angle values. Distribution of the N-Cα-C angle bond-angle values for the HRD-Asp and HRD-Arg in high-resolution structures. Shift in the HRD-Arg bond-angle values (116.33 Å ± 1.6) from the HRD-Asp suggests energetic strain because of bond bending. The average deviation from ideal bond-angle values (111 Å) is estimated to be 2 SDs (111 Å ± 2.8).

Fig. 3.

Frequency of disfavored (strained) and disallowed conformations observed at each residue position in the protein kinase domain. The residues are numbered according to the cAMP dependent kinase (PDB ID 1ATP).

Fig. 4.

Mutational analysis of EPK-ELK component residues. (A) F-helix-Asp mutations. (Left) The controls (with two known kinase dead mutants, D256A and K162A) and the F-helix-Asp single mutants (D311A to D311N). For each mutant, the status of the autophosphorylation, total amount of Aurora protein, substrate phosphorylation status (using phospho-Histone H3 antibody), and total Histone H3 in each lane is given. (Right) The data for double mutants in a background of activating T288E (activation loop phospho-mimic) mutant. (B) HRD-His and E-helix-His in the T288E background.

Fig. 5.

Model of strain-switch in kinase activation. The N and C lobes of the kinase are shown as ellipses, and the F-helix is shown as a cylinder in the C lobe. The zoomed in view of the kinase HRD motif region in active and inactive structures is shown (with side-chain removed for clarity) and the peptide flip is shown on the Ramachandran plot. In addition to the HRD-Arg residue (shown on Ramachandran plot as circles), the DFG-Asp ϕ/ψ values are also shown (as rectangles). The hydrophobic spine is shown as blue blurred circles, with a large blue circle in the smaller N lobe representing the spine residues in the N lobe. The hydrogen-bonding network mediated by the EPK-ELK component residues are shown as dotted orange lines.