Structural and catalytic motifs of PKA-C.

(A) Backbone representation of the ternary complex of PKA-C bound to ATP and the PKI5-24 peptide (not depicted), PDB Code 4WB5. Highlighted are key motifs including the αA, αC, and αE helices, C-terminal tail, activation loop, peptide-positioning loop, Gly-rich loop, and the DFG and APE motifs. (B) The hydrophobic core of PKA-C features the regulatory spine (R-spine, gold), catalytic spine (C-spine, blue), shell residues (cyan), and the αC-β4 loop (hot pink), which locks the αE helix and couples the two lobes of PKA-C.

Free energy landscape of PKA-C obtained from replica-averaged metadynamics (RAM) simulations.

(A) Convergence of the bias deposition along the first three collective variables (CVs). The free energy (expressed in kcal/mol) of the different CVs was averaged over the last 100 ns of RAM simulations. The standard deviations are reported as red error bars. (B-D) Free energy landscape along the first two principal components (PC1 and PC2) of PKA-C in the apo, ATP-bound, and ATP/PKI bound forms. PC1 and PC2 are projected from the first three CVs. The vertices represent conformational states. In the apo form, multiple states have comparable free energy with ΔG < 5 kcal/mol, whereas in the binary form, fewer states have ΔG < 5 kcal/mol. For the ternary form only a major ground state is populated.

Free energy surfaces and dynamic transitions determined by a Markov State Model (MSM) for apo, ATP-bound, and ATP/PKI-bound PKA-C.

(A) Free energy landscape projected along the first two time-lagged independent components (tICs) of apo PKA-C, featuring three basins, GS, ES1, and ES2. The transition from GS to ES1 (arrow) highlights the changes around the αB-αC loop, with the disruption of the K72-E91 salt bridge and the PIF pocket (V80-I85-F347) hydrophobic interactions. The GS to ES2 transition (arrow) displays the rearrangement of the hydrophobic packing around the αC-β4 loop. (B, C) Free energy surfaces projected along the first two tICs for the ATP-bound and ATP/PKI-bound PKA-C, respectively. Known crystal structures for the three forms are indicated by small white triangles.

Conformational transitions of apo and ATP-bound PKA-C from GS to ES1 and ES2 states as shown by the kinetic Monte Carlo trajectories.

(A, B) Time course of the structural transitions from GS to ES1 GS to ES2 for apo and ATP-bound PKA-C, respectively. The GS to ES1 transition is characterized by the disruption of the K72-E91 salt bridge and occurs frequently for both PKA-C forms. In contrast, the structural transition from GS to ES2 occurs only for the apo PKA-C, and it features the interactions between F100 and V104 that cause allosteric changes between D166-N171, K168-T201, and W222-A206-P207. The dark color traces indicate the moving averages calculated every 10 frames. (C) Structural snapshot of the GS conformation showing that the key catalytic motifs are poised for phosphoryl transfer. (D) Structural snapshot of the ES2 conformation with a disrupted configuration of key catalytic motifs typical of inactive kinase.

Comparison of experimental versus calculated 13C chemical shifts of methyl groups of PKA-C.

(A) Experimental and calculated CS for Val104-Cγ1 of apo PKA-C. The GS is in blue and the ES in magenta. (B) Corresponding CS profiles for Ile150-Cδ1. The experimental CS is shown in dotted lines for GS (black) and ES (red). (C) Correlation between predicted | ΔωPred| and experimental | ΔωExp| CS differences for methyl groups near the αC-β4 loop. The fitted linear correlation has a slope of 0.86 and R2 of 0.82.

F100A mutation increases the dynamics of the αC-β4 loop, perturbing the local hydrophobic packing and its anchoring to the αE helix.

(A) Time series of the αC-β4 loop dynamics, H-bond occurrence for the β− and γ-turns, F102 χ1 angle, and N99 and Y156 for WT (black) and F100A (red) in the ATP-bound state. (B) Representative structural snapshots showing the β-turn conformation for PKA-CWT (green) and γ-turn for PKA-CF100A (magenta).

Structural responses to ATP binding of PKA-CF100A mutant.

(A) Superposition of the hydrophobic cores (C spine, R spine, and shell residues) for PKA-CWT (lime) and PKA-CF100A (hot pink), highlighting the structural perturbations of the R spine and shell residues. (B) Structural perturbation upon ATP binding for the hydrophobic core residues of PKA-CWT and PKA-CF100A shown as changes in the population densities vs. rmsd (C, D) First (PC1) and second (PC2) principal components describing breathing and shearing motions of the two lobes. (E) 2D projections and distributions of PC1 and PC2 for PKA-CWT and PKA-CF100A.

Mutual information analysis of backbone and side chain rotamers of ATP-bound PKA-CWT and PKA-CF100A.

(A) Mutual information matrix for PKA-CWT showing well-organized clusters of interactions within each lobe and distinct inter-lobe communication typical of an active kinase. (B) Mutual information matrix for PKA-CF100A revealing an overall reorganization of the allosteric network caused by the disruption of the hydrophobic core.

NMR map of the structural response of PKA-CF100A to nucleotide and PKI binding.

(A) Comparison of the chemical shift perturbation (CSP) of the amide resonances for PKA-CF100A (black) and PKA-CWT (cyan) upon ATPγN binding. The dashed line indicates one standard deviation from the average CSP. (B) CSPs of PKA-CF100A/ATPγN amide resonances mapped onto the crystal structure (PDB: 4WB5). (C) Comparison of the CSPs of the amide resonances for PKA-CF100A and PKA-CWT upon binding ATPγN and PKI5-24 (black). (D) CSPs for the F100A/ATPγN/PKI complex mapped onto the crystal structure (PDB: 4WB5). To define the allosteric network of the kinase upon binding nucleotides and substrate, we examined the CS using CHEmical Shift Covariance Analysis (CHESCA),22,23,27 a statistical method that identifies correlated responses of residue pairs to a specific perturbation (i.e., ligand binding, mutations, etc.). CHESCA works under the assumption that pairwise correlated CS changes of residues identify possible intramolecular allosteric networks.52,53 For PKA-C, we found that coordinated structural rearrangements, as identified by CHESCA, are directly related to the extent of binding cooperativity.22,23,27 Therefore, we compared the CHESCA maps for PKA-CWT and PKACF100A for four different states: apo, ATPγN-, ADP-, and ATPγN/PKI5–24-bound. For PKA-CF100A, the CHESCA matrix exhibits sparser and more attenuated correlations (i.e., lower correlation co-efficient value) relative to PKA-CWT (Figure 10A). Although many inter-lobe correlations are still present for F100A, several other correlations in specific structural domains such as the αG-, αH-, and αI-helices are absent or attenuated. For instance, the F100A mutation does not display cor-relations between the αA-helix and the C-terminal tail that constitute a critical “complement to the kinase core”.54 We also utilized CHESCA to assess the allosteric communication among the PKA-C communities as defined by McClendon et al.55 The CHESCA community map for PKA-CWT shows strong correlations across the enzyme, especially for structurally adjacent communities and at the interface between the two lobes (see for instance the correlations among ComA, ComB, ComC, ComE, and ComH) (Figure 10B-C). For F100A, the CHESCA community map shows that the cross-talk between the nucleotide-binding (ComA) and positioning of αC-helix (ComB) communities, as well as the R-spine assembly (ComC) and the activation loop (ComF) communities are preserved (Figure 10B-C). However, the correlations between ComE, responsible for stabilizing the C spine, and ComC, involved in the assembly of the R spine, are absent. Similarly, the long-range correlations between the C and R spines (i.e., ComD with ComC) are missing. Finally, several correlations between ComF1, ComG, and ComH are no longer present. These communities orchestrate substrate recognition and R subunits binding. Overall, the CHESCA analysis for PKA-CF100A suggests that the reduced degree of cooperativity we determined thermodynamically corresponds to a decrease in coordinated structural changes upon ligand binding. The latter is apparent from the loss of correlated structural changes among the structural communities, including the hydrophobic spines, substrate binding cleft, and the docking surface for PKA interactions with other binding partners.

Correlated chemical shift changes reveal the uncoupling of the intramolecular allosteric network in PKA-CF100A.

(A) Comparison of the CHESCA matrices obtained from the analysis of the amide CS of PKA-CWT (blue correlations) and PKA-CF100A (black correlations). The correlations coefficients (Rij) were calculated using the apo, ADP-bound, ATPγN-bound, and ATPγN/PKI5-24-bound states. For clarity, only correlation with Rij > 0.98 are displayed. For the enlarged CHESCA map of F100A see Figure 10 – figure supplement 1. The data for the PKA-CWT matrix were taken from Walker et al.22. (B) Community CHESCA analysis of PKA-CWT (blue correlations) and PKA-CF100A (black correlations). Only correlations with RA,B > 0.98 are shown. Spider plot showing the extent of intramolecular correlations identified by the community CHESCA analysis for PKA-CF100A mapped onto the crystal structure (PDB: 4WB5). The thickness of each line in the spider plot indicates the extent of coupling between the communities.

Illustration of the collective variables (CVs) used in the RAM simulations.

(A) Backbone ψ angles of loops not in contact with ATP (Back-far). The Cα atoms are depicted as blue spheres. (B) Backbone ψ angles of loops in contact with ATP (Back-close). The Cα atoms are colored in magenta. (C) Side chains χ1 angles of loops in contact with ATP (Side-close). The side chains are represented in sticks colored in magenta. (D) The radius of gyration is calculated over the rigid part of the protein (Rgss), where the residues involved are colored in cyan.

Distribution of the Root-Mean-Square-Error (RMSE) of the chemical shifts for the different simulation schemes.

(A) RMSE of CS for apo PKA-C calculated from standard MD (left), REX (middle), and RAM (right). (B) RMSE of CS for PKA-C/ATP calculated from standard MD (left), REX (middle), and RAM (right). (C) RMSE of CS for PKA-C/ATP/PKI5-24 from standard MD (left), REX (middle), and RAM (right). Color codes for different backbone atoms (C, Cα, CO, H, and N) are indicated in the left figures.

Replica-averaged metadynamics (RAM) simulations explore a larger conformational space than standard MD and replica exchange (REX) simulations.

(A) Comparison of conformational space (CV1 vs. CV2) sampled by RAM Replica 1, standard MD, and REX Replica 1 for PKA-C. (B) Comparison of conformational space (CV3 vs. CV2) sampled by RAM Replica 1, standard MD, and REX Replica 1 for apo PKA-C.

Accumulative deposition of history-dependent biases along the first three CVs for the RAM simulations of the apo PKA-C.

The accumulative biases converged after ∼300 ns in the three CVs.

R spine and shell residues selected for two time-lagged independent components (tICA) and Markov State Model (MSM) analysis.

(A) Atom motions of key residues that define tIC1 of apo PKA-C colored according to the superposition deviations. Backbone atoms of Val104 show the largest change in tIC1. (B) Atom motions of key residues that define tIC2 of apo PKA-C colored according to the superposition deviations. Backbone atoms of Phe185 and Val104 show the largest change in tIC2.

Structural features of αC-out transition (ES1) in various inactive kinases.

(A) Crystal structures of PKA-C in active and inactive states highlight the αB-αC loop alterations. Crystal structures of PKA-C in the active (1ATP) and inactive conformations (3AG9, 1SZM, 4DFY), highlighting the electrostatic interactions between K72 in β3 and E91 in theαC. (B) Disruption of the K72 - E91salt bridge in the inactive structures of Abl (1OPJ), Src (1FMK), and CDK2 (4EK3). (C) The αC helix orientation of active PKA-C compared to the orientation in inactive kinases. (D) Structural transition from the GS state to ES1 state characterized by the outward movement of the αC helix (i.e., kinase inactivation).

Distinct hydrophobic packing for residues around the αC-β4 loop in the GS and ES states of apo PKA-C.

(A) Projections of randomly selected conformations for the GS (blue) and ES (magenta) onto the conformational landscape of apo PKA-C. Snapshots with tIC1 < 1.2 were clustered to separate the ES and GS, whereas those confomers with tIC1 > 0.2 were clustered as GS. (B,C) Close up of the hydrophobic packing in the ES and GS (C) states, highlighting Leu103, Val104, Ile150, Leu172, and Ile180 that show slow exchange in the CPMG experiments.

Distribution of predicted and experimental 13C CS of selected methyl groups.

(A-C) ES (magenta) and GS (blue) of apo PKA-C for Leu103-Cδ2 (A), Leu172-Cδ1 (B), and Ile180-Cδ1 (C). The experimental CSs are shown as dotted lines for GS (black) and ES (red).

Time series of the distance between F102 and R308 for PKA-CWT (black) and the PKA-CF100A mutant (magenta).

The π-cation interactions between the aromatic ring of F102 and the guanidine group of R308 are more persistent in the WT enzyme than in the F100A mutant.

NMR fingerprints of PKA-CF100A.

(A) [1H,15N]-WADE-TROSY spectra of apo, ADP-bound, ATPγN-bound, and ATPγN/PKI5-24 bound PKA-CF100A. (B) Changes in the chemical shift perturbation (CSP) between PKA-CWT and PKA-CF100A bound to ATPγN. (C) Changes in CSP (ΔδWT - ΔδF100A) upon binding ATPγN and PKI5-24.

CONCISE plot showing the shifts of the probability distribution of the amide resonances as a function of nucleotides and substrate binding.

The per-residue chemical shift information is averaged into the average principal component (PC) score indicative of the position of each conformational state of the kinase along the open-to-closed equilibrium.

Intermolecular allosteric network of F100A mapped using CHESCA and community CHESCA.

(A) CHESCA matrix obtained from the amide chemical shift trajectories of PKA-CF100A in the apo, ADP-bound, ATPγN-bound, and ATPγN/PKI5-24-bound states. Only correlations with Rij > 0.98 are displayed. (B) Plot of the correlation scores vs. residue calculated for PKA-CWT (blue) and PKA-CF100A (black). (C) Community CHESCA analysis of PKA-CF100A. Only correlations with RA,B > 0.98 are shown. (D) Spider plots indicating the correlated structural communities of PKA-CF100A and PKA-CWT plotted on their corresponding structures. The size of each node is independent of the number of residues it encompasses, and the weight of each line indicates the strength of coupling between the individual communities.

ΔG (kcal/mol) and relative population of the ground state and the first 6 excited states in different forms of PKA-C obtained from the RAM simulations.

Kinetic parameters of Kemptide phosphorylation by PKA-CWT and PKA-CF100A obtained from coupled assays.

The KM and Vmax values were obtained from a non-linear least squares analysis of the concentration-dependent initial phosphorylation rates. Errors in the kcat/KM ratios were propagated from the individual errors in KM and kcat.

Changes in enthalpy, entropy, free energy, and dissociation constants for nucleotide binding to PKA-CWT and PKA-CF100A.

All errors were calculated from triplicate measurements. Values for PKA-CWT are taken from Walker et al. 22

Changes in enthalpy, entropy, free energy, and dissociation constants for PKI5-24 binding to apo and ATPγN - saturated PKA-CWT and PKA-CF100A.

All errors were derived from triplicate measurements. The error for the cooperativity coefficient (σ) was propagated from the errors in Kd. Values for PKA-CWT were originally published in Walker et al. 22.