Allosteric inhibition explained through conformational ensembles sampling distinct "mixed" states

Abstract

Allosteric modulation provides an effective avenue for selective and potent enzyme inhibition. Here, we summarize and critically discuss recent advances on the mechanisms of allosteric partial agonists for three representative signalling enzymes activated by cyclic nucleotides: the cAMP-dependent protein kinase (PKA), the cGMP-dependent protein kinase (PKG), and the exchange protein activated by cAMP (EPAC). The comparative analysis of partial agonism in PKA, PKG and EPAC reveals a common emerging theme, i.e. the sampling of distinct "mixed" conformational states, either within a single domain or between distinct domains. Here, we show how such "mixed" states play a crucial role in explaining the observed functional response, i.e. partial agonism and allosteric pluripotency, as well as in maximizing inhibition while minimizing potency losses. In addition, by combining Nuclear Magnetic Resonance (NMR), Molecular Dynamics (MD) simulations and Ensemble Allosteric Modeling (EAM), we also show how to map the free-energy landscape of conformational ensembles containing "mixed" states. By discussing selected case studies, we illustrate how MD simulations and EAM complement NMR to quantitatively relate protein dynamics to function. The resulting NMR- and MD-based EAMs are anticipated to inform not only the design of new generations of highly selective allosteric inhibitors, but also the choice of multidrug combinations.

Keywords: Agonism; Allosteric pluripotency; Allostery; Antagonism; EPAC; PKA; PKG; cAMP; cGMP.

Graphical abstract

Fig. 1

Inhibition by allosteric partial agonists through perturbation of conformational equilibria. (A) An example of domain organization in a cyclic-nucleotide activated enzyme, i.e. the cGMP-dependent protein kinase (PKG) of Plasmodium falciparum. In this example, the regulatory and catalytic regions are in the same polypeptide chain, and thus the C-terminal cyclic-nucleotide binding domain (CBD) is directly linked to the catalytic region. (B) An example of the CBD structural architecture in the apo inactive state (red) and the holo active state (green). The grey region indicates the largely invariant β-subdomain. The N-terminal helices (N3A), the C-terminal helices (“B” and “C”), and the phosphate binding cassette (PBC) undergo significant structural changes upon binding to the cGMP effector. (C) When an allosteric domain equilibrates between inactive and active states, binding of its endogenous allosteric effector (activator) typically shifts the equilibrium to the active state. A partial agonist or antagonist can either shift the equilibrium towards the inactive state, or sample an ensemble of conformations where an additional “mixed” intermediate state is sampled, as shown in panel (D). The mixed nature of this intermediate is manifested in its ability to resemble more closely the active state in some regions, and the inactive state in other regions, as depicted by the green/red color pattern. The figure was originally published in the Journal of Biological Chemistry. Byun JA, Van K, Huang J, Henning P, Franz E, Akimoto M, et al. Mechanism of allosteric inhibition in the Plasmodium falciparum cGMP-dependent protein kinase. J. Biol. Chem. 2020;295:8480–91. © the American Society for Biochemistry and Molecular Biology. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

Fig. 2

Sampling mixed intermediate states of isolated CBDs enables partial agonists to maximize inhibition without significantly compromising affinities. (A) CBD-B of human PKG bound to cAMP samples inactive and active conformers as well as an intermediate state (dashed box), where only its C-terminal switch helix (SW) is disengaged and dynamic . The figure was reproduced with permission and was originally published in the Journal of Biological Chemistry. VanSchouwen B, Selvaratnam R, Giri R, Lorenz R, Herberg FW, Kim C, et al. Mechanism of cAMP Partial Agonism in Protein Kinase G (PKG). J. Biol. Chem. 2015; 290:28631–38641. © the American Society for Biochemistry and Molecular Biology. (B) Plasmodium falciparum PKG bound to 8-NBD-cGMP. The mixed intermediate state (dashed box) features an engaged pre-lid region to promote inhibitor binding, but the C-terminal lid remains disengaged to ensure inhibition . The figure was reproduced with permission and was originally published in the Journal of Biological Chemistry. Byun JA, Van K, Huang J, Henning P, Franz E, Akimoto M, et al. Mechanism of allosteric inhibition in the Plasmodium falciparum cGMP-dependent protein kinase. J. Biol. Chem. 2020;295:8480–91. © the American Society for Biochemistry and Molecular Biology. (C) EPAC1 bound to I942. The mixed intermediate (dashed box) includes PBC “in” and C-terminal hinge helix (H) “out” conformations . The out conformation of the hinge helix leads to inhibition of the catalytic region (CR), as it obstructs access to its Rap1 substrate. (D) EPAC1 bound to cAMP and CE3F4R, which uncompetitively binds to EPAC1 and stabilizes the mixed intermediate state (dashed box), where the PBC is “in” and the C-terminal hinge (H) is “out” . The figures are adapted with permission from Boulton S, Selvaratnam R, Blondeau J-P, Lezoualc’h F, Melacini G. Mechanism of Selective Enzyme Inhibition through Uncompetitive Regulation of an Allosteric Agonist. J Am Chem Soc 2018;140:9624–37. Copyright 2018 American Chemical Society.

Fig. 3

The multidomain PKA regulatory subunit bound to an allosteric inhibitor, Rp-cAMPS, samples an excited “mixed” state (A_off-B_on) that drives allosteric pluripotency. (A) Structure of Rp-cAMPS. (B) Domain organization of the regulatory subunit. (C) In the absence of cAMP, the regulatory subunit (R) inhibits the catalytic subunit (C). The inhibitory site linked to CBD-A docks into the active site of the C-subunit, and is further stabilized by MgATP (or MnAMP-PNP, which is a non-hydrolyzable ATP analogue). (D) When cAMP binds to each of the CBDs, conformational changes occur that allow the C-subunit to be released. The inter-CBD interaction through W260 stabilizes the “on” conformations of the CBDs. (E) The free-energy hierarchy of apo R and free energy changes upon binding of Rp. The A_onB_on becomes the ground state, and the inhibition-competent “mixed” state (A_off-B_on) is one of the excited states (red dashed box). (F) When the C-subunit is added to R:Rp₂ in the presence of high [MgATP], the inhibition-competent states (red bars) are stabilized and become the ground states. (G) When the C-subunit is added to R:Rp₂ in the absence of MgATP, the R:C interaction is less stable, and the inhibition-competent states remain excited. Since the ground state conformer exhibits low affinity for PKA C, the kinase function is activated. Figures are adapted from Byun JA, Akimoto M, VanSchouwen B, Lazarou TS, Taylor SS, Melacini G. Allosteric pluripotency as revealed by protein kinase A. Sci Adv 2020;6:eabb1250. Reprinted with permission from AAAS. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

Fig. 4

Effect of 8-NBD-cGMP bound to pfPKG CBD-D on C-terminal helix interactions necessary for kinase activation. (A) Overlay of the bound ligands from cGMP-bound CBD-D, and from representative structures of 8-NBD-cGMP-bound CBD-D generated from MD simulations. The syn orientation of cGMP is preserved in 8-NBD-cGMP. (B) Aligned representative structures of cGMP-bound CBD-D (grey) and 8-NBD-cGMP-bound CBD-D (red) generated from the MD simulations. The dashed box indicates the C-terminal lid region that becomes disordered in the 8-NBD-cGMP-bound CBD-D, as schematically shown in Fig. 2B. (C) Similar to panel (B), but zoomed into the Y480-R528 region. The yellow starburst indicates the steric clash of the 8-NBD substituent with the R528 side chain in the active structure, and the arrow indicates the structural shift of the R528 side chain upon binding of 8-NBD-cGMP. (D) WT-versus-R528K chemical shift differences for cGMP-bound (green) and 8-NBD-cGMP-bound (red) CBD-D. The distance from R528 (as measured in the cGMP-bound structure) is shown as a grey line, and the secondary structure in the cGMP-bound CBD-D is indicated at the top of the plot. The mutation site is indicated by a black star, and the grey highlight indicates residues near Y480 (black arrow), where perturbations induced by the mutation in the cGMP-bound complex are lost when cGMP is replaced by 8-NBD-cGMP. (E) Similar to panel (C), but zoomed into the capping-triad region (i.e. R484, Q532, D533) to highlight the shift of the D533 side chain (arrow) upon binding of 8-NBD-cGMP. (F) Similar to panel (D), but for the R484A mutant of CBD-D. Pre-lid and lid residues are indicated by pink and purple highlights, respectively, and capping-triad residues Q532 and D533 by black arrows. The figures were adapted with permission and were originally published in the Journal of Biological Chemistry. Byun JA, Van K, Huang J, Henning P, Franz E, Akimoto M, et al. Mechanism of allosteric inhibition in the Plasmodium falciparum cGMP-dependent protein kinase. J. Biol. Chem. 2020;295:8480–91. © the American Society for Biochemistry and Molecular Biology. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

Fig. 5

Outline of the hybrid PKA R-subunit initial structures used for MD simulations. (A, B) Fully-inactive structure of the R-subunit (i.e. CBD-A ‘off’/CBD-B ‘off’), without (A) and with (B) bound C-subunit. (C) Fully-active structure of the R-subunit (i.e. CBD-A ‘on’/CBD-B ‘on’). (D-F) Hybrid R-subunit structures, consisting of the inactive CBD-A and active CBD-B (i.e. CBD-A ‘off’/CBD-B ‘on’). (G-I) As in (D-F), but with bound C-subunit. In all structures, the R-subunit and C-subunit residues derived from the inactive structure are shown in red and grey, respectively, while the R-subunit residues derived from the active structure are shown in green. Figures are adapted from Byun JA, Akimoto M, VanSchouwen B, Lazarou TS, Taylor SS, Melacini G. Allosteric pluripotency as revealed by protein kinase A. Sci Adv 2020;6:eabb1250. Reprinted with permission from AAAS. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

Fig. 6

Analysis of the MD simulations of PKA. (A) Inter-center of mass (CM) distances between CBD-A and CBD-B, computed from MD simulations of the PKA R-subunit structures lacking bound C-subunit. (B) RMSDs of CBD-A relative to the fully-active structure, computed from MD simulations of the PKA R-subunit structures lacking bound C-subunit. (C) As in (B), but for the RMSDs of CBD-B relative to the fully-active structure. (D, E) As in (B, C), but for the RMSDs relative to the fully-inactive structure. (F, G) As in (B, D), but for MD simulations of the PKA R-subunit structures with bound C-subunit. (H, I) As in (C, E), but for MD simulations of the PKA R-subunit structures with bound C-subunit. (J) Inter-CM distances between CBD-A and the C-subunit, computed from MD simulations of the PKA R-subunit structures with bound C-subunit. The headers at the top of the figure illustrate the initial structures utilized for the MD simulations (Fig. 5), and the respective color codes in the plots. Notable shifts toward inter-CM distances or RMSD values similar to the fully-active structure that were observed from the simulations are indicated by black arrows. Figures are adapted from Byun JA, Akimoto M, VanSchouwen B, Lazarou TS, Taylor SS, Melacini G. Allosteric pluripotency as revealed by protein kinase A. Sci Adv 2020;6:eabb1250. Reprinted with permission from AAAS.

Fig. 7

Measuring EAM input parameters by NMR. (A) Representative NMR TROSY cross-peaks of apo R-subunit (CBD-A and CBD-B) and the reference cAMP-bound and C-bound R-subunits. (B) Chemical shift correlation plots of CBD-A and CBD-B for the apo sample, where the slope represents the fraction of ‘off’ states in each domain. The closed and open circles represent ¹H and ¹⁵N chemical shifts, respectively. Figures are adapted from Akimoto M, McNicholl ET, Ramkissoon A, Moleschi K, Taylor SS, Melacini G. Mapping the Free Energy Landscape of PKA Inhibition and Activation: A Double-Conformational Selection Model for the Tandem cAMP-Binding Domains of PKA RIα. PLoS Biol. 2015;13:e1002305. (C) Similar to panel (A), but with the addition of the W260A:Rp₂ TROSY spectrum. (D) Similar to panel (B), but for the W260A:Rp₂ sample. (E) Similar to panel (C), but with the addition of the WT:Rp₂ TROSY spectrum. (F) Based on the TROSY cross-peaks in panel (E), the intensities of the minor and major peaks are measured, allowing for the calculation of the open vs. closed population ratios. This population ratio is used to estimate the ΔG_AB, as shown in panel (G). (H) Similar to panel (A) right, but with the addition of the C:R:Rp₂ TROSY spectrum. (I) Similar to panel (D), but for CBD-B of the C:R:Rp₂ complex. (J) The fraction of ‘off’ state of CBD-B in the C:R:Rp₂ sample can be used to estimate the ratio of state-specific association constants of C-subunit for R-subunit (ρ_C). Figures were adapted from Byun JA, Akimoto M, VanSchouwen B, Lazarou TS, Taylor SS, Melacini G. Allosteric pluripotency as revealed by protein kinase A. Sci Adv 2020;6:eabb1250. Reprinted with permission from AAAS.

Fig. 8

The EAM reproduces the functional data for PKA and reveals new drivers of allosteric pluripotency. (A) The kinase activity upon addition of Rp is correctly predicted using EAM. The absence and presence of high [MgATP] is simulated with low and high γ values, respectively. The predicted K_a and Hill coefficients agree with the experimental data within error . Both the inter-domain interaction, and the shifting of CBD-B to the ‘on’ state upon binding of Rp, contribute to the observed agonism. (B) The PKA kinase activity was measured with Rp in the presence of high- and low-affinity substrates (i.e. PKS and PKS2, respectively). The experimentally measured kinase activation values are in good agreement with the predicted values from EAM analysis. Figures are adapted from Byun JA, Akimoto M, VanSchouwen B, Lazarou TS, Taylor SS, Melacini G. Allosteric pluripotency as revealed by protein kinase A. Sci Adv 2020;6:eabb1250. Reprinted with permission from AAAS.

Fig. 9

Abl kinase bound to imatinib adopts a ‘mixed’ conformation. The structures of Abl in the active state (left), the two inactive states (I₁ and I₂; middle), and the imatinib-bound state (right) are shown. The imatinib-bound state exhibits DFG and A-loop conformations resembling the I₂ inactive state, and αC and P-loop conformations resembling the active state . The figure was adapted from Xie T, Saleh T, Rossi P, Kalodimos CG. Conformational states dynamically populated by a kinase determine its function. Science 2020;370:eabc2754. Reprinted with permission from AAAS.

Fig. 10

Different allosteric drivers within the same partial agonist can lead to ‘mixed’ conformational states. The cyclic nucleotide cGMP, which forms interactions with a CBD mainly through its base moiety (allosteric driver 1) and its ribose-phosphate moiety (allosteric driver 2), is shown as an example. The base is stabilized in the binding pocket through interactions with the base-binding region and capping lid of the CBD, whereas the phosphate binding cassette (PBC) of the CBD interacts with the ribose-phosphate of the cyclic nucleotide. When the ribose-phosphate is modified, for example, by replacing the equatorial oxygen with a bulkier sulfur, steric clashes with the PBC lead to the PBC sampling the “out” orientation, typical of the inactive conformation. On the other hand, when the base is modified, for example, by introducing additional aromatic motifs, engagement of the capping lid interaction, typical of the active conformation, may be perturbed. If two distinct allosteric drivers within the same ligand preferentially bind different conformations (e.g. active vs. inactive), mixed intermediate states are stabilized.

Fig. 11

Synergies between NMR, MD and EAM enable quantitative modeling of enzyme function. NMR provides an initial map of the states within the conformational ensemble of a protein:partial-agonist complex, which serves as a basis to build initial structures for MD simulations and an EAM. The MD simulations serve as an effective means to generate refined targeted hypotheses to be tested by NMR. The EAM input parameters can often be measured by NMR. The fully parameterized EAM model enables bridging from protein dynamics to quantitative predictions of enzymatic function (e.g. kinase activity). Figures are adapted from Byun JA, Akimoto M, VanSchouwen B, Lazarou TS, Taylor SS, Melacini G. Allosteric pluripotency as revealed by protein kinase A. Sci Adv 2020;6:eabb1250. Reprinted with permission from AAAS.