Advertisement

Mechanistic insights on novel small molecule allosteric activators of cGMP-dependent protein kinase PKG1α

Open AccessPublished:July 19, 2022DOI:https://doi.org/10.1016/j.jbc.2022.102284
      cGMP-dependent protein kinase (PKG) represents a compelling drug target for treatment of cardiovascular diseases. PKG1 is the major effector of beneficial cGMP signaling which is involved in smooth muscle relaxation and vascular tone, inhibition of platelet aggregation and signaling that leads to cardioprotection. In this study, a novel piperidine series of activators previously identified from an ultrahigh-throughput screen were validated to directly bind partially activated PKG1α and subsequently enhance its kinase activity in a concentration-dependent manner. Compounds from initial optimization efforts showed an ability to activate PKG1α independent of the endogenous activator, cGMP. We demonstrate these small molecule activators mimic the effect of cGMP on the kinetic parameters of PKG1α by positively modulating the KM of the peptide substrate and negatively modulating the apparent KM for ATP with increase in catalytic efficiency, kcat. In addition, these compounds also allosterically modulate the binding affinity of cGMP for PKG1α by increasing the affinity of cGMP for the high-affinity binding site (CNB-A) and decreasing the affinity of cGMP for the low-affinity binding site (CNB-B). We show the mode of action of these activators involves binding to an allosteric site within the regulatory domain, near the CNB-B binding site. To the best of our knowledge, these are the first reported non-cGMP mimetic small molecules shown to directly activate PKG1α. Insights into the mechanism of action of these compounds will enable future development of cardioprotective compounds that function through novel modes of action for the treatment of cardiovascular diseases.

      Keywords

      Abbreviations:

      AI (autoinhibitory), BME (β-mercaptoethanol), BSA (bovine serum albumin), CNB (cyclic nucleotide binding domain), DMSO (dimethyl sulfoxide), FP (fluorescence polarization), MMSA (microfluidic mobility shift assay), MST (microscale thermophoresis), PKG (cGMP-dependent protein kinase), STD (saturation transfer difference), TCEP (Tris(2-carboxyethyl)phosphine), uHTS (ultrahigh-throughput screening)
      The cGMP signaling pathway is impaired in cardiovascular diseases (
      • Dang T.A.
      • Schunkert H.
      • Kessler T.
      cGMP signaling in cardiovascular diseases: linking genotype and phenotype.
      ,
      • Blanton R.M.
      cGMP signaling and modulation in heart failure.
      ,
      • Dunkerly-Eyring B.
      • Kass D.A.
      Myocardial phosphodiesterases and their role in cGMP regulation.
      ,
      • Preedy M.E.J.
      • Baliga R.S.
      • Hobbs A.J.
      Multiplicity of nitric oxide and natriuretic peptide signaling in heart failure.
      ). Targeting this pathway with the aim of restoring the beneficial effects of the signaling cascade represents an important and clinically validated therapeutic strategy. Modulation of the cGMP pathway has been successfully translated in the clinic by use of cGMP-elevating agents that work by enhancing the activity of cGMP-generating enzymes (soluble and particulate guanylate cyclases, nitric oxide enhancers) or by blocking cGMP degradation through inhibition of phosphodiesterase activity (
      • Sandner P.
      • Zimmer D.P.
      • Milne G.T.
      • Follmann M.
      • Hobbs A.
      • Stasch J.P.
      Correction to: soluble guanylate cyclase stimulators and activators.
      ,
      • Park M.
      • Sandner P.
      • Krieg T.
      cGMP at the centre of attention: emerging strategies for activating the cardioprotective PKG pathway.
      ,
      • Kim G.E.
      • Kass D.A.
      Cardiac phosphodiesterases and their modulation for treating heart disease.
      ). Riociguat (soluble guanylate cyclase stimulator) and sildenafil (phosphodiesterase 5 inhibitor), for example, were both approved for the treatment of pulmonary arterial hypertension while Vericiguat (soluble guanylate cyclase stimulator) was recently approved for chronic heart failure.
      cGMP-dependent protein kinase (PKG) is the major effector of cGMP signaling in the cardiovascular system (
      • Dang T.A.
      • Schunkert H.
      • Kessler T.
      cGMP signaling in cardiovascular diseases: linking genotype and phenotype.
      ,
      • Blanton R.M.
      cGMP signaling and modulation in heart failure.
      ,
      • Dunkerly-Eyring B.
      • Kass D.A.
      Myocardial phosphodiesterases and their role in cGMP regulation.
      ,
      • Preedy M.E.J.
      • Baliga R.S.
      • Hobbs A.J.
      Multiplicity of nitric oxide and natriuretic peptide signaling in heart failure.
      ). PKG is involved in regulating physiological processes that includes smooth muscle relaxation and vascular tone (
      • Ignarro L.J.
      • Kadowitz P.J.
      The pharmacological and physiological role of cyclic GMP in vascular smooth muscle relaxation.
      ), inhibition of platelet aggregation (
      • Moro M.A.
      • Russel R.J.
      • Cellek S.
      • Lizasoain I.
      • Su Y.
      • Darley-Usmar V.M.
      • et al.
      cGMP mediates the vascular and platelet actions of nitric oxide: confirmation using an inhibitor of the soluble guanylyl cyclase.
      ), and signaling that lead to cardioprotection (
      • Dang T.A.
      • Schunkert H.
      • Kessler T.
      cGMP signaling in cardiovascular diseases: linking genotype and phenotype.
      ,
      • Blanton R.M.
      cGMP signaling and modulation in heart failure.
      ,
      • Dunkerly-Eyring B.
      • Kass D.A.
      Myocardial phosphodiesterases and their role in cGMP regulation.
      ,
      • Preedy M.E.J.
      • Baliga R.S.
      • Hobbs A.J.
      Multiplicity of nitric oxide and natriuretic peptide signaling in heart failure.
      ). PKG therefore represents a compelling drug target for treatment of cardiovascular diseases. Direct pharmacological activation of PKG, independent of cGMP, may capture the beneficial effects of cGMP-PKG signaling while bypassing the multiple downstream signaling pathways mediated by other cGMP effectors (
      • Feil R.
      • Lohmann S.M.
      • de Jonge H.
      • Walter U.
      • Hofmann F.
      Cyclic GMP-dependent protein kinases and the cardiovascular system: insights from genetically modified mice.
      ).
      PKG1 is a soluble homodimer and is the main PKG isoform in the cardiovascular system (
      • Hofmann F.
      • Bernhard D.
      • Lukowski R.
      • Weinmeister P.
      cGMP regulated protein kinases (cGK).
      ). It exists as two splice variants, PKG1α and PKG1β, that differ in the first 100 amino acids of their sequence, which imparts differential affinity for cGMP as well as protein substrate specificity (
      • Hofmann F.
      • Bernhard D.
      • Lukowski R.
      • Weinmeister P.
      cGMP regulated protein kinases (cGK).
      ,
      • Takimoto E.
      Cyclic GMP-dependent signaling in cardiac myocytes.
      ). The PKG1 polypeptide is divided into two major domains: a regulatory domain and a catalytic domain (Fig. 1). The regulatory domain is comprised of functional subdomains (
      • Pfeifer A.
      • Ruth P.
      • Dostmann W.
      • Sausbier M.
      • Klatt P.
      • Hofmann F.
      Structure and function of cGMP-dependent protein kinases.
      ,
      • Moon T.M.
      • Sheehe J.L.
      • Nukareddy P.
      • Nausch L.W.
      • Wohlfahrt J.
      • Matthews D.E.
      • et al.
      An N-terminally truncated form of cyclic GMP-dependent protein kinase Iα (PKG Iα) is monomeric and autoinhibited and provides a model for activation.
      ,
      • Alverdi V.
      • Mazon H.
      • Versluis C.
      • Hemrika W.
      • Esposito G.
      • van den Heuvel R.
      • et al.
      cGMP-binding prepares PKG for substrate binding by disclosing the C-terminal domain.
      ,
      • Hofmann F.
      The biology of cyclic GMP-dependent protein kinases.
      ,
      • Osborne B.W.
      • Wu J.
      • McFarland C.J.
      • Nickl C.K.
      • Sankaran B.
      • Casteel D.E.
      • et al.
      Crystal structure of cGMP-dependent protein kinase reveals novel site of interchain communication.
      ). At the N terminus is a coiled coil leucine zipper responsible for homodimerization and subcellular localization of the protein. Following the leucine zipper is an autoinhibitory (AI) subdomain that contains an inhibitory motif called the ‘‘pseudo-substrate” motif (
      • Moon T.M.
      • Sheehe J.L.
      • Nukareddy P.
      • Nausch L.W.
      • Wohlfahrt J.
      • Matthews D.E.
      • et al.
      An N-terminally truncated form of cyclic GMP-dependent protein kinase Iα (PKG Iα) is monomeric and autoinhibited and provides a model for activation.
      ,
      • Alverdi V.
      • Mazon H.
      • Versluis C.
      • Hemrika W.
      • Esposito G.
      • van den Heuvel R.
      • et al.
      cGMP-binding prepares PKG for substrate binding by disclosing the C-terminal domain.
      ). The AI subdomain, in the absence of cGMP, interacts with the catalytic domain blocking substrate access and thus locking PKG in the basal, inactive state. Next to the AI subdomain are two tandemly arranged high and low affinity cyclic nucleotide binding domain (CNB) sites (CNB-A and CNB-B, respectively) capable of binding cGMP and cAMP. In PKG1α, the high-affinity CNB-A lies closer to the N terminus (
      • Hofmann F.
      The biology of cyclic GMP-dependent protein kinases.
      ). Following the CNB sites is a helical subdomain (switch helix, S) proposed to be involved in the regulation of kinase activity (
      • Osborne B.W.
      • Wu J.
      • McFarland C.J.
      • Nickl C.K.
      • Sankaran B.
      • Casteel D.E.
      • et al.
      Crystal structure of cGMP-dependent protein kinase reveals novel site of interchain communication.
      ,
      • Moon T.M.
      • Osborne B.W.
      • Dostmann W.R.
      The switch helix: a putative combinatorial relay for interprotomer communication in cGMP-dependent protein kinase.
      ). The C-terminal tail of the switch helix (‘knob’ site) interacts with the hydrophobic (‘nest’) site adjacent to the cGMP-binding site in CNB-B (
      • Osborne B.W.
      • Wu J.
      • McFarland C.J.
      • Nickl C.K.
      • Sankaran B.
      • Casteel D.E.
      • et al.
      Crystal structure of cGMP-dependent protein kinase reveals novel site of interchain communication.
      ). Elements of this ‘knob/nest’ interaction are conserved in PKG1 (
      • Moon T.M.
      • Osborne B.W.
      • Dostmann W.R.
      The switch helix: a putative combinatorial relay for interprotomer communication in cGMP-dependent protein kinase.
      ). Site-directed mutagenesis of key residues in the ‘knob’ site in full-length PKG1α increased the basal activity of the kinase, decreased the activation constant for cGMP, and abolished cGMP-induced enzyme activation cooperativity. All of these suggest the involvement of the switch helix interactions in regulating enzyme activity (
      • Osborne B.W.
      • Wu J.
      • McFarland C.J.
      • Nickl C.K.
      • Sankaran B.
      • Casteel D.E.
      • et al.
      Crystal structure of cGMP-dependent protein kinase reveals novel site of interchain communication.
      ).
      Figure thumbnail gr1
      Figure 1PKG1α domain organization. PKG1α is divided into regulatory (blue) and catalytic (green) regions. The regulatory region is comprised of a leucine zipper (LZ) domain, autoinhibitory domain (AI), cGMP high affinity binding domain (CNB-A), cGMP low affinity binding domain (CNB-B), and the switch helix (S). The catalytic region is subdivided into two domains: the Mg2+/ATP-binding domain (N-lobe) and the protein substrate binding domain (C-lobe). CNB, cyclic nucleotide binding domain; PKG, cGMP-dependent protein kinase.
      The C-terminal catalytic region of PKG1 is comprised of two lobes: the N-terminal (N-) lobe, which contains the Mg2+/ATP-binding site, and the C-terminal (C-) lobe, which contains the protein substrate-binding site (
      • Busch J.L.
      • Bridges T.M.
      • Richie-Jannetta R.
      • Hollett B.P.
      • Francis S.H.
      • Corbin J.D.
      Catalytic site amino acids of PKGI-alpha influence allosteric cGMP binding.
      ). Activation of PKG1 requires binding of cGMP to the CNB sites triggering a significant conformational change that results in switching from a compact ball-like folded form in the basal (cGMP-free) inactive state to an extended and functionally active conformation that releases the catalytic domain from autoinhibition to allow phosphorylation of downstream protein substrates (
      • Hofmann F.
      The biology of cyclic GMP-dependent protein kinases.
      ,
      • Francis S.H.
      • Busch J.L.
      • Corbin J.D.
      • Sibley D.
      cGMP-dependent protein kinases and cGMP phosphodiesterases in nitric oxide and cGMP action.
      ).
      To date, efforts to develop small molecule activators of PKG1 were focused on analogs of cGMP that are susceptible to off-target activity with limited pharmaceutical developability (
      • Butt E.
      cGMP-dependent protein kinase modulators.
      ). Peptide-based activators of PKG1α were recently reported and were shown to inhibit myogenic constriction in smooth muscle cells in isolated, endothelium-denuded cerebral arteries (
      • Moon T.M.
      • Tykocki N.R.
      • Sheehe J.L.
      • Osborne B.W.
      • Tegge W.
      • Brayden J.E.
      • et al.
      Synthetic peptides as cGMP-independent activators of cGMP-dependent protein kinase Iα.
      ). These peptides were derived from the sequence of the switch helix (S) bridging CNB-B and the catalytic domain and were proposed to activate PKG1α by disrupting the switch helix interactions involved in regulating kinase activity (
      • Moon T.M.
      • Tykocki N.R.
      • Sheehe J.L.
      • Osborne B.W.
      • Tegge W.
      • Brayden J.E.
      • et al.
      Synthetic peptides as cGMP-independent activators of cGMP-dependent protein kinase Iα.
      ). Here, we describe the characterization of small molecule activators of PKG1α initially identified from an ultrahigh-throughput screening (uHTS) campaign (
      • Hanisak J.
      • Soriano A.
      • Adam G.C.
      • Basso A.
      • Bauman D.
      • Bell D.
      • et al.
      Discovery of the first non-cGMP mimetic small molecule activators of cGMP-dependent protein kinase 1 α (PKG1α).
      ). Early optimization efforts showed emerging potency to directly activate PKG1α from the basal state. By evaluating the modulation of cGMP binding to the high affinity (CNB-A) and low affinity (CNB-B) sites, we show that these compounds activate PKG1α by binding to a unique site allosteric to the two CNB sites within the regulatory domain of PKG1α.

      Results

      Enhancement of PKG1α kinase activity by small molecule activators

      To search for direct activators of PKG1α, we implemented a uHTS screen that resulted in the discovery of small molecule activators 1 and 2 (Fig. 2). Early optimization efforts led to the synthesis of 21 and 23 (Fig. 2) (
      • Hanisak J.
      • Soriano A.
      • Adam G.C.
      • Basso A.
      • Bauman D.
      • Bell D.
      • et al.
      Discovery of the first non-cGMP mimetic small molecule activators of cGMP-dependent protein kinase 1 α (PKG1α).
      ). All four compounds could increase the enzymatic activity of the partially activated, full-length, recombinant PKG1α (PKG1α2-671) in a concentration-dependent manner (Fig. 3A) with EC50 values between 5 μM and 47 μM (Table 1). These activation potencies were benchmarked against maximal PKG1α activity in the presence of cGMP that has an EC50 ∼ 0.03 μM (Fig. 3A). The screen hits (1 and 2) did not show significant activation of the basal state of PKG1α within the concentration range tested (Fig. 3B), which was limited by compound solubility in aqueous assay buffer (Table 1). Compounds 21 and 23, with improved solubility, enhanced the basal PKG1α activity with an estimated EC50 of ∼ 78 μM for 23 while 32% PKG1α activation was observed in the presence of 200 μM of 21 (Fig. 3B and Table 1). These piperidine series activators also enhanced the activity of PKG1α partially activated by cAMP (1 μM cAMP ∼EC20) with similar potencies (Fig. 3C).
      Figure thumbnail gr2
      Figure 2Structure of small molecule activators of PKG1α. Compounds 1 and 2 were original hits identified from compound library screen; compounds 21 and 23 were synthesized from initial optimization of piperidine 1 (
      • Hanisak J.
      • Soriano A.
      • Adam G.C.
      • Basso A.
      • Bauman D.
      • Bell D.
      • et al.
      Discovery of the first non-cGMP mimetic small molecule activators of cGMP-dependent protein kinase 1 α (PKG1α).
      ). PKG, cGMP-dependent protein kinase.
      Figure thumbnail gr3
      Figure 3Enhancement of PKG1α activity at increasing concentrations of small molecule activators. A, activity of partially activated PKG1α (∼EC20 with 10 nM cGMP). B, basal PKG1α (no cGMP) at increasing concentrations of cGMP, 1, 2, 21, and 23 tracked using ADP-Glo kinase assay. C, activity of partially activated PKG1α (∼EC20 with 1 μM cAMP) at increasing concentrations of cAMP, 1, 2, and 21 tracked using HTRF kinase assay. Error bars indicate SE from 2 to 3 independent trials. PKG, cGMP-dependent protein kinase.
      Table 1Potency (EC50) of small molecule activators
      ActivatorEC50 (μM)
      EC50 values were average of 2 to 3 independent trials (errors represented as ± SE).
      , PKG1α activation
      Solubility
      In assay buffer.
      (μM)
      Kinase assay (ADP-Glo)Kinase assay (HTRF)
      No cGMP (basal)10 nM cGMP (∼EC20)No cGMP (basal)10 nM cGMP (∼EC20)
      1>6747.2 ± 0.3>6745 ± 563
      2>675 ± 1>675 ± 163
      2132 ± 4%
      Percent activation of PKG1α at 200 μM compound concentration.
      7 ± 440 ± 10%
      Percent activation of PKG1α at 200 μM compound concentration.
      6 ± 1200
      2380 ± 1011 ± 1NT
      Not tested.
      NT200
      Abbreviation: HTRF, homogeneous time resolved fluorescence.
      a EC50 values were average of 2 to 3 independent trials (errors represented as ± SE).
      b Percent activation of PKG1α at 200 μM compound concentration.
      c Not tested.
      d In assay buffer.

      Validation of direct target engagement

      Direct binding of these compounds to PKG1α was confirmed using microscale thermophoresis (MST) and saturation transfer difference (STD) NMR spectroscopy. Figure 4A shows changes in thermophoresis of partially activated PKG1α (with 10 nM cGMP) in the presence of increasing concentrations of compound 1, 2, or 21, indicating direct binding by each compound. The concentration–response curves allowed a Kd estimate for compound 21 of ∼16 μM while limitations in the soluble concentration range prevented estimates of Kd values for 1 and 2 (Table 1). 1H STD NMR spectroscopy was used as an orthogonal method to validate direct target engagement of the piperidine series activators. In this technique, 1H NMR spectra are acquired on samples that are mixtures of small molecules and proteins. STD NMR spectra show specific ligand protons that are affected by radio frequency saturation that is transferred from the protein to the ligand during binding. After becoming saturated, weak binding ligands dissociate from the protein and are detected free in solution. Saturation leads to a small decrease in some ligand NMR peaks. Peak intensity changes are identified by collecting reference spectrum (on the same sample with no saturation) that is subtracted from data with saturation, resulting in STD spectra. STD NMR peaks are typically detected for ligand-binding affinities that range from micromolar to millimolar. High-affinity ligands stay bound to the target (slow dissociation) and are not directly detected. Figure 4B shows the results of STD NMR binding studies for 1 and 21 in equilibrium with full-length PKG1α. Weak binding of 100 μM compound 1 to 5 μM apo-PKG1α is confirmed by the presence of ligand STD NMR peaks (blue). Addition of 10 μM cGMP does not affect compound binding (red), indicating that 1 and cGMP bind to different sites. Weak binding of 100 μM compound 21 to 5 μM of apo-PKG1α is confirmed by the presence of ligand STD NMR peaks (blue). Addition of 5 μM and 50 μM cGMP does not affect 21 binding (red and black, respectively), indicating that 21 and cGMP bind to different sites. STD NMR binding studies confirmed 1 and 21 binding to full-length PKG1α. Kd values were not derived from this data.
      Figure thumbnail gr4
      Figure 4Validation of direct target engagement of piperidine series. A, changes in thermophoresis of partially activated PKG1α (with 10 nM cGMP) upon binding of small molecule activators. Concentration–response curves were shown for 1, 2, and 21. Kd for 21 was estimated at ∼16 ± 4 μM (Mean ± SE; n = 2). Kd estimates for 1 and 2 was not possible due to limitations in solubility. B, direct binding of piperidines to PKG1α monitored by 1H STD NMR. Left panel: binding of 1 to apo-PKG1α led to observed peaks in the STD NMR spectrum (blue); compound 1 binding was not affected by the addition of 10 μM cGMP (red). A control spectrum with 1 and no protein shows no peaks (black). Right panel: binding of 21 to apo-PKG1α (blue) was not affected by the addition of 5 and 50 μM cGMP (red and black, respectively). PKG, cGMP-dependent protein kinase; STD, saturation transfer difference.

      Piperidine series activators modulate the steady-state kinetic parameters of partially activated PKG1α similar to the modulatory effects of endogenous cGMP

      To gain insights to the mechanism of activation of the piperidine series, we first looked at the effect of these compounds on the kinetic parameters of partially activated PKG1α using a microfluidic mobility shift assay (MMSA, see Experimental procedures). Compounds 1, 2, and 21 were observed to decrease the KM of peptide substrate (Glasstide) for PKG1α in a concentration-dependent manner without significant modulation of the catalytic efficiency, kcat, (Fig. 5A). The apparent KM and kcat for ATP, on the other hand, increased in the presence of compounds 1, 2, and 21 (Fig. 5B). The same modulations of enzyme kinetic parameters were observed in the cGMP-induced activation of PKG1α. A 5-fold decrease in peptide KM with no significant change in kcat was observed when PKG1α was taken from partially activated state (10 nM cGMP; ∼ EC20) to fully activated state (10 μM cGMP; EC100) (Table S1). In the case of ATP, a 3-fold increase in apparent KM for ATP was observed accompanied by an increase in kcat as PKG1α was taken from partially activated state to the fully activated state (Table S1).
      Figure thumbnail gr5
      Figure 5Modulation of kinetic parameters of partially activated PKG1α by small molecule activators. A, positive modulation of peptide substrate KM (red) with no significant change in kcat (purple) in the presence of (Aa) 1, (Ab) 2, or (Ac) 21. B, negative modulation of apparent KM (red) for ATP with increase in kcat (purple) in the presence of (Ba) 1, (Bb) 2, or (Bc) 21. Error bars indicate 95% confidence interval from nonlinear regression fits used to estimate kinetic parameters. PKG, cGMP-dependent protein kinase.

      Mutual modulation of binding affinities for PKG1α between piperidine series activators and cGMP

      The activity of the original screen hits was observed only for the partially activated PKG1α. This raised the possibility of a binding cooperativity between the piperidine series and cGMP. To evaluate this possibility, we first looked at the effect of the piperidine series on the cGMP-induced activation of PKG1α by tracking the modulation of cGMP EC50 in the presence of increasing concentration of piperidines using MMSA method. As shown in Figure 6A, the EC50 of cGMP was positively modulated in the presence of increasing concentrations of 21 with a maximal increase in cGMP potency (16-fold decrease in EC50) in the presence of saturating concentrations of 21. The concentration–response curves ([cGMP] versus kinase activity) also showed enhancement of basal activity of PKG1α in the presence of increasing concentrations of 21, reflecting an ability to activate PKG1α in the absence of cGMP. The same positive modulation of cGMP EC50 was also observed in the presence of other members of the piperidine series (Fig. S1). We next evaluated the effect of cGMP on the binding affinity of piperidine series for PKG1α using MST. As shown in Figure 6B, an almost 9-fold enhancement in binding affinity of 21 for PKG1α (decrease in apparent Kd from ∼14 μM down to ∼1.6 μM) was observed as cGMP was increased from 10 nM to 100 nM. Overall, our results were consistent with mutual positive allosteric modulation of binding affinities for full-length PKG1α between piperidine series activators and cGMP.
      Figure thumbnail gr6
      Figure 6Mutual positive modulation of affinity for PKG1α between piperidine series activators and cGMP. A, cGMP EC50 estimates in the presence of increasing concentrations of compound 21. The red arrow indicate left shift of the concentration–response curves (decrease in cGMP EC50) in the presence of increasing concentrations of 21; green arrow indicate increase in basal activity (no cGMP) in the presence of increasing concentrations of 21. Inset graph shows cGMP EC50 values for best fit curves plotted against concentration of 21 (error bars indicate SE from two independent trials). PKG1α activity (initial velocities) were measured using microfluidic mobility shift assay. B, increase in binding affinity of 21 for PKG1α in the presence of increasing concentrations of cGMP. Binding affinities (Kd) were estimated by measuring changes in thermophoresis of PKG1α in the presence of varying concentrations of cGMP. Error (±SE) represents the standard error of nonlinear least squares regression. PKG, cGMP-dependent protein kinase; SE, standard error.
      Figure thumbnail gr7
      Figure 7Fluorescence polarization assays measuring effect of piperidines on cGMP binding to CNB-A and CNB-B in full-length PKG1α (PKG1α2-671). A, enhancement of 8-fluo-cGMP binding affinity for CNB-A. Aa, Increase in fraction of 8-fluo-cGMP bound to full-length PKG1α (increase in polarization signal) with increasing concentrations of 21 and 23. In contrast, a decrease in polarization was observed in the presence of increasing concentration of direct competitors (unlabeled cGMP, 8Br-cGMP, and cAMP). Ab, 8-fluo-cGMP affinity estimates. The increase in apparent affinity of 8-fluo-cGMP for CNB-A site with increasing concentration of 23 observed in the left shift trend of the concentration response curves. Error bars represent SE from two replicate measurements. Ac, Curvilinear relationship between affinity of 8-fluo-cGMP for CNB-A and concentration of 23 indicating piperidines bind to a site allosteric to CNB-A (Kd for best fit curves from panel Ab). B, negative modulation of cGMP affinity for CNB-B. Ba, At conditions where both CNB-A and CNB-B were populated by 8-fluo-cGMP, a partial decrease in fraction of 8-fluo-cGMP bound to PKG1α (decrease in polarization signal) was observed in the presence of increasing concentrations of 21 and 23, indicating competition for CNB-B as 8-fluo-cGMP remained bound to CNB-A. In contrast, unlabeled cGMP decreased polarization signal to baseline level indicating direct competition for both CNB-A and CNB-B sites. Bb, Sequential two-site equilibrium binding of 8-fluo-cGMP to PKG1α. A decrease in apparent affinity of 8-fluo-cGMP for CNB-B with increasing concentration of 23 observed as a right shift in the portion of the concentration–response curve corresponding to binding to CNB-B. Error bars represent SE from two replicate measurements. Bc, Curvilinear relationship between apparent 8-fluo-cGMP Kd and concentration of 23 indicating piperidines bind to a site allosteric to CNB-B. CNB, cyclic nucleotide binding domain; PKG, cGMP-dependent protein kinase; SE, standard error.

      Piperidine series are allosteric activators that positively modulate cGMP affinity for CNB-A and negatively modulate cGMP affinity for CNB-B

      To determine the localization of piperidine series binding site relative to CNB-A and CNB-B, we evaluated the modulation of apparent affinity (Kd, app) of cGMP for each site in the presence of increasing concentrations of piperidine series activator. A linear relationship between cGMP Kd, app values with increasing concentration of small molecule activator will indicate direct binding competition for the same CNB site. On the other hand, a curvilinear relationship between cGMP Kd, app with increasing small molecule activator concentration will indicate allosteric modulation of cGMP binding affinity and the existence of a unique piperidine series binding site allosteric from the cGMP-binding site. We used fluorescently labeled cGMP, 8-fluo-cGMP, to probe the modulation of cGMP binding affinity for CNB sites in the presence of small molecule activator. Fluorescence polarization (FP) method was used to monitor binding of 8-fluo-cGMP to PKG1α. In this method, the FP signal of 8-fluo-cGMP increases as a result of a change in its rotation speed in solution: from a fast rotation when free in solution to a much slower solution rotation speed when bound to PKG1α. The increase in FP signal is proportional to the fraction of 8-fluo-cGMP bound to PKG1α.
      To evaluate modulation of cGMP affinity for CNB-A, a FP binding assay was performed to track binding of 8-fluo-cGMP to CNB-A site in full-length PKG1α. At conditions where only CNB-A was populated by 8-fluo-cGMP (10 nM PKG1α and 0.5 nM 8-fluo-cGMP, see Experimental procedures for details), an increase in the fraction of 8-fluo-cGMP bound to PKG1α was observed in the presence of increasing concentrations of 21 or 23 (increase in FP signal, Fig. 7Aa). This indication of enhanced affinity for CNB-A in the presence of piperidines directly contrasted with a decrease in fraction of 8-fluo-cGMP bound to PKG1α (decrease in FP signal) in the presence of increasing concentrations of an orthosteric competitor such as unlabeled cGMP, 8Br-cGMP, or cAMP (Fig. 7Aa). To further clarify the observed positive modulation of cGMP affinity for CNB-A by piperidines, we measured the apparent affinity of 8-fluo-cGMP (FP of 8-fluo-cGMP [0.5 nM] with titration of PKG1α) in the presence of increasing concentrations of piperidines. As shown in Figure 7Ab, an increase in apparent affinity of 8-fluo-cGMP for CNB-A with increasing concentration of 23 was clearly observed in the left shift trend of the concentration–response curves. The apparent Kd values for best fit curves plotted against concentration of 23 revealed a curvilinear relationship, indicating piperidines allosterically enhance the affinity of cGMP for CNB-A (Fig. 7Ac). At saturating concentrations of 23, we observed an almost 4-fold enhancement of affinity of 8-fluo-cGMP for CNB-A.
      We next determined the modulatory effect of piperidines on cGMP affinity for CNB-B. We first set up a FP binding assay at conditions where 8-fluo-cGMP occupied both CNB-A and CNB-B (100 nM PKG1α in equilibrium with 200 nM 8-fluo-cGMP). As shown in Figure 7Ba, a partial decrease in the fraction of 8-fluo-cGMP bound to PKG1α was observed (a partial decrease in FP signal) in the presence of increasing concentrations of either 21 or 23. This indicated piperidines compete with cGMP for the CNB-B site as CNB-A site remained occupied by 8-fluo-cGMP. To further highlight these observations, we used unlabeled cGMP as surrogate for a direct (orthosteric) competitor for both the CNB-A and CNB-B sites. As shown in Figure 7Ba, increasing concentrations of unlabeled cGMP decreased the FP signal down to background levels indicating complete displacement of 8-fluo-cGMP from both CNB-A and CNB-B sites as expected for orthosteric competition.
      The indication that the piperidine series competes for the low affinity cGMP-binding site prompted further studies to elucidate whether these activators bind directly to CNB-B (orthosteric competition) or to a unique site allosteric to CNB-B (allosteric competition). To distinguish between these two modes of competitive binding, we set out to determine the modulation of apparent affinity (Kd, app) of cGMP for CNB-B site in the presence of increasing concentration of piperidine series activator. To enable measurement of cGMP binding affinity for CNB-B, a novel two-site FP binding assay was configured based on a sequential two-site equilibrium binding model where Kd1 and Kd2 are the cGMP binding affinities for the CNB-A and CNB-B site, respectively.
      P+cGMPKd1P(cGMP)+cGMPKd2P(cGMP)2


      Starting at a condition where the two CNB sites are fully occupied at a 1:2 molar ratio of PKG1α and 8-fluo-cGMP (at concentrations well above the Kd values for the two CNB sites), PKG1α and 8-fluo-cGMP concentrations were then varied in a dilution series while maintaining the 1:2 protein to 8-fluo-cGMP ratio (see Experimental procedures for details). The observed FP signal from the resulting solutions tracked the equilibrium titration of 8-fluo-cGMP across the two CNB sites: binding first to the high affinity CNB-A site, then to the low affinity CNB-B site, as the concentrations of both PKG1α and 8-fluo-cGMP increase. The plot of FP (∼fraction of 8-fluo-cGMP bound to PKG1α) versus PKG1α concentration was fitted to a sequential two-site binding equation (Equation 9 in Experimental procedures) to estimate cGMP affinities for both CNB sites (Kd1 and Kd2). The estimated 8-fluo-cGMP Kd for CNB-A and CNB-B using this method closely matched the corresponding cGMP Kd estimates using isothermal titration calorimetry method (Fig. S2C) confirming the robustness of the assay and validate the use of 8-fluo-cGMP as a surrogate for cGMP. We then determined the apparent Kd of 8-fluo-cGMP at both sites in the presence of increasing concentrations of small molecule activator. As shown in Figure 7Bb, although the assay in this format is not sensitive enough to deconvolute the apparent Kd of 8-fluo-cGMP for CNB-A below 2 nM, the positive allosteric modulation of cGMP affinity for CNB-A at increasing concentration of 23 can still be observed as a left shift in the segment of the concentration–response curve corresponding to 8-fluo-cGMP binding to CNB-A. On the other hand, the affinity of 8-fluo-cGMP for CNB-B site was clearly shown to decrease with increasing concentration of 23 as observed in the right shift of the concentration–response curves corresponding to binding to CNB-B (Fig. 7Bb). An almost 6-fold increase in apparent 8-fluo-cGMP Kd (decrease in apparent affinity) for CNB-B was observed in the presence of saturating concentration of 23 (from 280 nM to ∼1700 nM Kd, app; Fig. 7Bc). The curvilinear relationship between apparent 8-fluo-cGMP Kd and concentration of 23 (Fig. 7Bc) indicated allosteric competition for CNB-B. To provide contrast and further highlight this observation, we also determined the relationship between the apparent affinity of 8-fluo-cGMP and concentration of orthosteric competitors (unlabeled cGMP and cAMP) using the same two-site binding assay. The presence of increasing concentrations of each of these direct competitors decreased the apparent affinity of 8-fluo-cGMP for the two binding sites as indicated by rightward shift of the concentration–response curves (Fig. S3, Aa and Ba). A plot of apparent Kd of 8-fluo-cGMP for CNB-A and concentration of cGMP (Fig. S3Ab) or cAMP (Fig. S3Bb) showed a linear relationship, indicating direct competition for the same binding site.

      Piperidine series allosteric binding site is in the regulatory domain

      To confirm the binding localization of the piperidine series and to gain additional insights to mechanism of action, we evaluated the modulation of cGMP binding affinity for the isolated regulatory domain fragment (CNB sites and switch helix; PKG1α79-356) in the presence of piperidine series activator. To evaluate the modulation of cGMP binding to CNB-A, we measured the apparent affinity of 8-fluo-cGMP (polarization of 8-fluo-cGMP [0.5 nM] with titration of PKG1α79-356) in the presence of increasing concentration of 23. As shown in Figure 8, Aa and Bb), no significant shift in the concentration–response curves (and corresponding estimates of apparent 8-fluo-cGMP Kd) was observed at increasing concentration of 23 (up to saturating level ∼200 μM).
      Figure thumbnail gr8
      Figure 8Fluorescence polarization assays measuring effect of piperidines on cGMP binding affinity for CNB-A and CNB-B in regulatory domain, PKG1α79-356. A, effect of piperidines on 8-fluo-cGMP binding to CNB-A. Aa, 8-fluo-cGMP affinity estimates (polarization of 8-fluo-cGMP (0.5 nM) with titration of PKG1α79-356). The presence of compound 23, up to 200 μM, had no effect on the concentration–response curve. Error bars represent SE from two replicate measurements. Ab, No significant modulation of 8-fluo-cGMP affinity with increasing concentration of 23 (Kd values for best fit curves from Panel Aa). B, effect of piperidines on cGMP binding to CNB-B. Ba, Sequential two-site equilibrium binding of 8-fluo-cGMP to PKG1α79-356. A decrease in apparent affinity of 8-fluo-cGMP for CNB-B with increasing concentration of 23 was observed as a right shift in the portion of the dose–response curve corresponding to the low affinity site, CNB-B. Bb, Curvilinear relationship between 8-fluo-cGMP Kd versus compound 23 concentration supported allosteric localization of piperidine series binding site to the regulatory domain, PKG1α79-356. CNB, cyclic nucleotide binding domain; PKG, cGMP-dependent protein kinase.
      To evaluate the effect of piperidines on cGMP binding to CNB-B in PKG1α79-356, the FP two-site binding assay was performed to estimate the apparent Kd of 8-fluo-cGMP for CNB-B, applying similar conditions used for the full-length protein. As shown in Figure 8Ba, a decrease in apparent affinity of 8-fluo-cGMP for CNB-B with increasing concentration of 23 was observed as a right shift in the portion of the concentration–response curve corresponding to binding to the low affinity site indicating a negative modulation of 8-fluo-cGMP affinity. The plot of apparent Kd against the concentration of 23 showed a curvilinear relationship (Fig. 8Bb), confirming the allosteric localization of piperidine series binding site within the regulatory domain.
      A competition binding analysis using 1H STD NMR was also performed to confirm allosteric binding of piperidine series to PKG1α79-356. A solution containing 1 and cAMP in equilibrium with PKG1α79-356 showed signature peaks from both 1 and cAMP in the STD NMR spectrum indicating binding of 1 and cAMP to different sites in PKG1α79-356 (Fig. S4B). In contrast to this observation, addition of a saturating concentration of cGMP to a solution containing cAMP in equilibrium with PKG1α79-356 resulted in the disappearance of the cAMP peaks in the STD NMR spectrum (Fig. S4A), indicating direct competition for the same binding sites. Note that cGMP signal is not observable in STD NMR due to its relatively high affinity to CNB-A and CNB-B (Fig. S2C).

      In silico modeling studies identified a possible piperidine series interaction site near CNB-B

      To further investigate the potential binding site for small molecule activators, an in silico model was built based on the previously published 3SHR (
      • Osborne B.W.
      • Wu J.
      • McFarland C.J.
      • Nickl C.K.
      • Sankaran B.
      • Casteel D.E.
      • et al.
      Crystal structure of cGMP-dependent protein kinase reveals novel site of interchain communication.
      ) structure of the regulatory domain, PKG1α 78-355 (Figs. 9 and 10A). Within this structure, the monomer was probed for potential binding sites. The five largest sites identified were examined. Sites S1 and S2, the two largest sites, were of similar size and corresponded to the cGMP-binding pockets. Site S3 was highly surface exposed and did not yield any significant interactions on binding. Site S5 was eliminated as it occurs at the N terminus of the sequence of 3SHR and is a possible artifact of truncation (Fig. 9). Site S4 appeared to be the most interesting consisting of a mostly hydrophobic pocket within the regulatory domain, proximal to CNB-B. While docking of the small molecule activator, 23, was attempted at all sites, the Site 4 pocket demonstrated excellent shape complementarity with significant arene-π interactions occurring throughout the pocket (Fig. 10, B and C). This pocket is occupied by Phe350,351 and Leu354 of the opposite chain switch helix in the full structure (
      • Osborne B.W.
      • Wu J.
      • McFarland C.J.
      • Nickl C.K.
      • Sankaran B.
      • Casteel D.E.
      • et al.
      Crystal structure of cGMP-dependent protein kinase reveals novel site of interchain communication.
      ). The molecule mimicked the insertion of Phe350,351 and Leu354 (Fig. 11). A search for potential alternate binding sites, including the cGMP-binding pockets themselves, did not identify a region more compatible than that observed at the identified allosteric pocket.
      Figure thumbnail gr9
      Figure 9In silico modeling study to identify possible small molecule binding site based on 3SHR (
      • Osborne B.W.
      • Wu J.
      • McFarland C.J.
      • Nickl C.K.
      • Sankaran B.
      • Casteel D.E.
      • et al.
      Crystal structure of cGMP-dependent protein kinase reveals novel site of interchain communication.
      ) structure of PKG1α 78-355. Largest sites as identified by the site finder algorithm: S1 and S2 correspond to cGMP low affinity and high affinity binding sites, respectively. S3 is a large shallow, mostly hydrophilic site with significant solvent exposure. S5 is located at the N terminus in the truncated region of the sequence. S4 is a large hydrophobic site that was considered to be the most likely binding site for the allosteric interaction site. The sites are rendered with mesh surfaces colored by lipophilicity. PKG, cGMP-dependent protein kinase.
      Figure thumbnail gr10
      Figure 10In silico model of possible allosteric interaction site within PKG1α 78-355 based on 3SHR (
      • Osborne B.W.
      • Wu J.
      • McFarland C.J.
      • Nickl C.K.
      • Sankaran B.
      • Casteel D.E.
      • et al.
      Crystal structure of cGMP-dependent protein kinase reveals novel site of interchain communication.
      ). A, cGMP (gray-spacefill) is bound at both low and high affinity sites with compound 23 (cyan sticks) bound proximal to the low affinity cGMP site at CNB-B. B, a distinct pocket is observed with shape complementarity to the ligand and significant H-π interactions throughout the site as can be observed in the ligand interaction diagram, (C). CNB, cyclic nucleotide binding domain; PKG, cGMP-dependent protein kinase.
      Figure thumbnail gr11
      Figure 11Overlap of possible small molecule–binding site with switch helix interaction site based on 3SHR. A, compound 23 (cyan) as docked into the hydrophobic nest pocket (S4). H-π and H-bonding interactions are observed throughout the pocket and principally occur with Phe257 and Trp288 among others. These interactions were conserved when compared with the switch helix (yellow) (B). When viewed from the side cut-away of the pocket (C), in the overlay, compound 23 is observed to penetrate deeper into the hydrophobic pocket and displays a larger number of interactions, perhaps explaining its ability to activate PKG1α through displacement of the switch helix. PKG, cGMP-dependent protein kinase.

      Discussion

      As the major effector of the beneficial effects of cGMP signaling, PKG1 represents a compelling drug target for treatment of cardiovascular diseases. Current therapeutic agents that modulate the cGMP signaling pathway work by increasing cGMP levels and subsequent activation of PKG1. Direct pharmacological activation of PKG1, independent of cGMP, presents an attractive route to gain additional efficacy as this may capture the benefits of cGMP-PKG signaling while avoiding the possible complicating effects of multiple downstream signaling pathways also activated by cGMP. To search for small molecule activators of PKG1α, we screened a diverse library of ∼2.9 million compounds against the basal and partially active enzyme (
      • Hanisak J.
      • Soriano A.
      • Adam G.C.
      • Basso A.
      • Bauman D.
      • Bell D.
      • et al.
      Discovery of the first non-cGMP mimetic small molecule activators of cGMP-dependent protein kinase 1 α (PKG1α).
      ). The inclusion of the partially activated state in the uHTS screen was based on its likely relevance in vivo due to the known high affinity of cGMP and cAMP for CNB-A (
      • Osborne B.W.
      • Wu J.
      • McFarland C.J.
      • Nickl C.K.
      • Sankaran B.
      • Casteel D.E.
      • et al.
      Crystal structure of cGMP-dependent protein kinase reveals novel site of interchain communication.
      ) and the basal tone of these endogenous activators in the cell (
      • Francis S.H.
      • Blount M.A.
      • Zoraghi R.
      • Corbin J.D.
      Molecular properties of mammalian proteins that interact with cGMP: Protein kinases, cation channels, phosphodiesterases, and multi-drug anion transporters.
      ) that can favor partial activation of PKG1α. Here, we describe the characterization of a piperidine series of activators, initially identified from uHTS screen (Fig. 2), as part of a larger drug discovery effort focused on direct activation of PKG1α.
      The uHTS hits, 1 and 2, and synthesized compounds from initial optimization efforts, 21 and 23, enhance the kinase activity of partially activated PKG1α (∼EC20 with 10 nM cGMP or 1 μM cAMP) in a concentration-dependent manner (Fig. 3, A and C). The observed ability of these compounds to enhance activity of PKG1α partially activated by cAMP may have added significance as cAMP may partially (cross-) activate PKG1α in the background of very low cGMP concentration in the disease state. Initial optimization of the original screen hit, 1, resulted in an immediate 4- to 7-fold improvement in EC50 for analogs 21 and 23 on the partially activated enzyme (Table 1). More importantly, these piperidine analogs have an ability to enhance the activity of PKG1α from the basal state, independent of cGMP (Figs. 3B, 6A, and Table 1). A detailed exploration of structure activity relationships revealed the addition of a hydroxy methyl group in 23 is responsible for improved activity relative to the parent piperidine 1 (
      • Hanisak J.
      • Soriano A.
      • Adam G.C.
      • Basso A.
      • Bauman D.
      • Bell D.
      • et al.
      Discovery of the first non-cGMP mimetic small molecule activators of cGMP-dependent protein kinase 1 α (PKG1α).
      ). Additionally, the combination of a carboxylic acid and the hydroxy methyl group were identified as modifications that enhanced the potency of the N-benzyl analog, 23, relative to the parent piperidine 1. These modifications likely created affinity-enhancing interactions as indicated by the stronger affinity of 21 for the partially activated PKG1α, relative to the parent piperidine 1 (Fig. 4A), with consequent improvement in compound potency. The modifications in 21 and 23 also resulted in relative improvements in solubility (Table 1) and cell permeability (
      • Hanisak J.
      • Soriano A.
      • Adam G.C.
      • Basso A.
      • Bauman D.
      • Bell D.
      • et al.
      Discovery of the first non-cGMP mimetic small molecule activators of cGMP-dependent protein kinase 1 α (PKG1α).
      ) enabling analysis of compound activity in the cell. We demonstrated on-target cell-based activity for these small molecules by measuring the PKG1α phosphorylation of the known endogenous substrate, vasodilator-stimulated phosphoprotein (VASP), which gave EC50 values for 21 and 23 of 9 μM and 15 μM, respectively (
      • Hanisak J.
      • Soriano A.
      • Adam G.C.
      • Basso A.
      • Bauman D.
      • Bell D.
      • et al.
      Discovery of the first non-cGMP mimetic small molecule activators of cGMP-dependent protein kinase 1 α (PKG1α).
      ). The observed cell-based activity added critical validation for the piperidine series and their potential for development.
      To gain insights to the mechanism of PKG1α activation by the piperidine series, we first looked at how these small molecules modulate the kinetic parameters of PKG1α in comparison to cGMP. The observed increase in peptide substrate affinity (decrease in KM of Glasstide) in the presence of small molecule activator suggested an enhanced substrate access to the catalytic site (optimized binding mode) favorable for catalysis (Fig. 5A). This positive modulation of peptide substrate affinity, along with the observed increase in apparent KM for ATP (with increase in kcat) at increasing small molecule concentration (Fig. 5B), demonstrated an overall similarity to modulations in enzyme kinetic parameters observed in the cGMP-induced activation of PKG1α (Table S1 and published reports (
      • Alverdi V.
      • Mazon H.
      • Versluis C.
      • Hemrika W.
      • Esposito G.
      • van den Heuvel R.
      • et al.
      cGMP-binding prepares PKG for substrate binding by disclosing the C-terminal domain.
      ,
      • Døskeland S.O.
      • Vintermyr O.K.
      • Corbin J.D.
      • Ogreid D.
      Studies on the interactions between the cyclic nucleotide-binding sites of cGMP-dependent protein kinase.
      ,
      • McCune R.W.
      • Gill G.N.
      Positive cooperativity in guanosine 3':5'-monophosphate binding to guanosine 3':5'-monophosphate-dependent protein kinase.
      ,
      • Hofmann F.
      • Flockerzi V.
      Characterization of phosphorylated and native cGMP-dependent protein kinase.
      )). These observations indicate that the piperidine series activators elicit the same mechanisms associated with cGMP-induced activation of full-length PKG1α.
      With the observed similarity in modulating enzyme kinetic parameters between piperidine series activators and cGMP, we set out to clarify whether these small molecules are cGMP mimetics and bind to one of the CNB sites. Direct binding to full-length enzyme was demonstrated in the presence and absence of cGMP (Fig. 4, A and B). The binding of these small molecules to PKG1α in the presence of cGMP (Fig. 4B) was an early indication of the existence of a piperidine series binding site allosteric to one or both CNB sites. This aligned with our earlier observation that the uHTS hits (1 and 2) enhanced activity of partially activated PKG1α (Fig. 3, A and C) but not the basal state (Fig. 3B). Indeed, results consistent with mutual positive allosteric modulation of affinities for full-length PKG1α were observed between piperidines and cGMP supporting the existence of an allosteric binding site for these small molecule activators (Figs. 6 and S1).
      To establish the localization of the piperidine series binding site relative to CNB-A and CNB-B, we used FP-based equilibrium binding assays to determine the effect of small molecule activator on the cGMP-binding affinity for each CNB site in full-length PKG1α. A linear relationship between the apparent affinity of labeled cGMP (8-fluo-cGMP) and small molecule concentration would indicate direct (orthosteric) competition for the same binding site, while a curvilinear relationship between 8-fluo-cGMP apparent affinity and small molecule concentration will indicate allosteric binding relative to the cGMP binding site in full-length PKG1α. Because the uHTS hits only enhance the activity of partially activated PKG1α, we hypothesized that the piperidines likely do not bind to CNB-A which should be populated by cGMP in the partially activated enzyme. This was validated by observations of enhanced 8-fluo-cGMP affinity for CNB-A in full-length PKG1α at increasing concentrations of piperidine series activator (Fig. 7A). The curvilinear relationship between apparent 8-fluo-cGMP Kd for CNB-A and small molecule concentration validated the allosteric localization of the piperidine series binding site relative to CNB-A (Fig. 7Ac). The observed allosteric enhancement of 8-fluo-cGMP affinity for CNB-A in the presence of small molecule activator was consistent with earlier observations of mutual positive modulation of binding affinity for PKG1α between cGMP and piperidine series (Figs. 6 and S1).
      In the case of CNB-B site, initial competition binding analysis indicated the piperidines compete with cGMP for the CNB-B site (Fig. 7Ba). We further evaluated the nature of this competition binding (orthosteric versus allosteric competition) to determine the localization of the piperidine series binding site relative to CNB-B. A novel FP-based two-site binding assay allowed the specific measurement of cGMP binding affinity for CNB-A and CNB-B. We observed a negative modulation of 8-fluo-cGMP binding affinity for CNB-B in the presence of increasing concentration of small molecule activator (Fig. 7, Bb and Bc). The curvilinear relationship between the apparent 8-fluo-cGMP Kd and small molecule concentration indicated allosteric competition and the existence of a small molecule binding site distinct from CNB-B. This observation of negative binding cooperativity between small molecule binding site and CNB-B support a mechanism of small molecule–induced PKG1α activation that do not involve binding to CNB-B. We note that in the case of the cGMP-induced PKG1α activation, a negative binding cooperativity between cGMP-binding sites, CNB-A and CNB-B, was demonstrated (
      • Moon T.M.
      • Sheehe J.L.
      • Nukareddy P.
      • Nausch L.W.
      • Wohlfahrt J.
      • Matthews D.E.
      • et al.
      An N-terminally truncated form of cyclic GMP-dependent protein kinase Iα (PKG Iα) is monomeric and autoinhibited and provides a model for activation.
      ). Assessment of the contribution of each of the CNB sites to enzyme activation led to the conclusion that CNB-B site has a nonessential role in cGMP-induced kinase activity (
      • Moon T.M.
      • Sheehe J.L.
      • Nukareddy P.
      • Nausch L.W.
      • Wohlfahrt J.
      • Matthews D.E.
      • et al.
      An N-terminally truncated form of cyclic GMP-dependent protein kinase Iα (PKG Iα) is monomeric and autoinhibited and provides a model for activation.
      ). Using site-directed mutagenesis to disrupt binding of cGMP to each of the CNB site, Moon et al. showed that only the disruption of cGMP binding to CNB-A resulted in abolishing activation of kinase activity (
      • Moon T.M.
      • Sheehe J.L.
      • Nukareddy P.
      • Nausch L.W.
      • Wohlfahrt J.
      • Matthews D.E.
      • et al.
      An N-terminally truncated form of cyclic GMP-dependent protein kinase Iα (PKG Iα) is monomeric and autoinhibited and provides a model for activation.
      ).
      We narrowed down the localization of the piperidine series binding site within the regulatory domain fragment, PKG1α79-356, encompassing the cyclic nucleotide-binding sites and the switch helix (see Fig. 1). A competition binding analysis, monitored by 1H STD NMR, showed signals from both piperidine series activator and cAMP in the STD NMR spectrum supporting the existence of a small molecule–binding site unique from the CNB sites within the isolated PKG1α79-356 (Fig. S4B).
      To gain further insights to the mechanism of action of these small molecule activators, we evaluated the modulation of cGMP affinity for CNB-A and CNB-B by small molecule activator in the context of the isolated regulatory domain fragment (PKG1α79-356) using the same FP-based binding studies performed on full-length enzyme. The positive binding cooperativity observed between CNB-A and piperidine-binding site in the full-length enzyme, PKG1α2-671 (Fig. 7A), was not preserved in the isolated PKG1α79-356 (Fig. 8A). This indicated that the positive cooperativity between CNB-A and small molecule activator binding site was likely facilitated by mechanisms that exist in the environment of the full-length enzyme. The importance of the full-length protein environment was also demonstrated for the observed positive cooperativity in the cGMP-induced activation of full-length PKG1α (
      • Pfeifer A.
      • Ruth P.
      • Dostmann W.
      • Sausbier M.
      • Klatt P.
      • Hofmann F.
      Structure and function of cGMP-dependent protein kinases.
      ,
      • Hofmann F.
      • Flockerzi V.
      Characterization of phosphorylated and native cGMP-dependent protein kinase.
      ). This positive cooperativity was not observed in the truncated enzyme missing the dimerization domain (monomeric PKG1α 78-671) leading to the proposal that the mechanism driving this enzyme activation cooperativity lies at the interface between protomers in the full-length (homodimer) enzyme (
      • Moon T.M.
      • Sheehe J.L.
      • Nukareddy P.
      • Nausch L.W.
      • Wohlfahrt J.
      • Matthews D.E.
      • et al.
      An N-terminally truncated form of cyclic GMP-dependent protein kinase Iα (PKG Iα) is monomeric and autoinhibited and provides a model for activation.
      ,
      • Kim J.J.
      • Lorenz R.
      • Arold S.T.
      • Reger A.S.
      • Sankaran B.
      • Casteel D.E.
      • et al.
      Crystal structure of PKG I:cGMP complex reveals a cGMP-mediated dimeric interface that facilitates cGMP-induced activation.
      ).
      The observed negative cooperativity between CNB-B and piperidine series binding site in full-length PKG1α2-671 (Fig. 7B) was preserved in the isolated regulatory domain fragment, PKG1α79-356 (Fig. 8B). This further validated localization of piperidine series binding site to the regulatory domain and suggests that the small molecule activator binding site is near CNB-B. To further investigate the possible interaction site of piperidines within the regulatory domain, we conducted in silico modeling studies based on the previously published structure of the regulatory domain, PKG1α 78-355 (Fig. 10A) (
      • Osborne B.W.
      • Wu J.
      • McFarland C.J.
      • Nickl C.K.
      • Sankaran B.
      • Casteel D.E.
      • et al.
      Crystal structure of cGMP-dependent protein kinase reveals novel site of interchain communication.
      ). The modeling studies led to the identification of a unique pocket proximal to CNB-B. Docking studies using 23 clearly demonstrated excellent shape complementarity with significant arene-π interactions occurring throughout the pocket (Fig. 10, B and C). The allosteric nature of this potential binding site, its proximity to CNB-B, and the lack of a similar site near CNB-A was consistent with our overall biochemical and biophysical data. Interestingly, this potential small molecule binding pocket is within the ‘nest’ site proposed as the binding site for the recently reported peptide-based activators derived from the sequence of the switch helix. These peptides were thought to mimic the C-terminal switch helix ‘knob’ interaction with the ‘nest’ site adjacent to the cGMP CNB-B site, facilitating a cGMP-independent activation of PKG1α (
      • Moon T.M.
      • Tykocki N.R.
      • Sheehe J.L.
      • Osborne B.W.
      • Tegge W.
      • Brayden J.E.
      • et al.
      Synthetic peptides as cGMP-independent activators of cGMP-dependent protein kinase Iα.
      ). The most potent peptide, S1.5, is a 24-mer (∼2695 Da) truncation derivative of the parent switch helix sequence. The potency of this activator (Ka ∼ 3 μM) has been shown to derive from the preserved peptide helicity and C terminus ‘knob’ motif (Phe350, Phe351, Asp353, and Leu354) (
      • Moon T.M.
      • Tykocki N.R.
      • Sheehe J.L.
      • Osborne B.W.
      • Tegge W.
      • Brayden J.E.
      • et al.
      Synthetic peptides as cGMP-independent activators of cGMP-dependent protein kinase Iα.
      ). Removal or mutation of the ‘knob’ residues, including the critical Phe350 and Phe351, abolished peptide activity, suggesting a mechanism of action involving switch helix interactions observed in the crystal structure of PKG1α regulatory domain (
      • Moon T.M.
      • Tykocki N.R.
      • Sheehe J.L.
      • Osborne B.W.
      • Tegge W.
      • Brayden J.E.
      • et al.
      Synthetic peptides as cGMP-independent activators of cGMP-dependent protein kinase Iα.
      ). In the case of the small molecule activator, 23, the docking model appears to mimic the binding of three ‘knob’ residues in the switch helix (Phe350, Phe351, and Leu354; Fig. 11). In particular, the dichloro-phenyl moiety corresponds to the positioning of Phe350. The small molecule activator penetrates the nest pocket deeper than the switch helix and displays a larger number of interactions (Fig. 11). Modifications on key substituents of the core piperidine ring, including the benzyl ether and dichloro-phenyl ring moieties, rendered compounds inactive, indicating that these substituents are critical for activity (
      • Hanisak J.
      • Soriano A.
      • Adam G.C.
      • Basso A.
      • Bauman D.
      • Bell D.
      • et al.
      Discovery of the first non-cGMP mimetic small molecule activators of cGMP-dependent protein kinase 1 α (PKG1α).
      ) and aligns with their possible involvement in binding site interactions as suggested by the docking model. Crystallographic studies on isolated PKG1α fragment that includes CNB-B and switch helix have since confirmed the binding mode predicted by in silico studies (
      • Mak V.W.
      • Patel A.M.
      • Yen R.
      • Hanisak J.
      • Lim Y.-H.
      • Bao J.
      • et al.
      Optimization and mechanistic investigations of novel allosteric activators of PKG1α.
      ). The overlap of the small molecule–binding site with the switch helix ‘knob’ binding site suggests an activation mechanism involving disruption of switch helix interactions that regulate full-length PKG1α activity.
      In summary, our work focused on elucidating the mechanism of action of small molecule activators in the context of full-length (homodimeric) PKG1α. In the absence of structural information on full-length enzyme, our data clearly demonstrated that the small molecule activator is not a cGMP mimetic. We thus accomplished our main goal of finding a route to directly activate PKG1α that will avoid activation of other signaling pathways dependent on cGMP. The small molecule modulation of enzyme substrate affinities indicates the piperidine compound series elicit similar mechanism of enzyme activation as the endogenous activator, cGMP. The positive modulation of the peptide substrate KM is direct evidence of a compound-induced open conformation allowing substrate access to the catalytic site. We demonstrated that the small molecule activator increases the affinity of cGMP for the high affinity binding site (CNB-A) and decreases the affinity of cGMP for the low affinity binding site (CNB-B) in the regulatory domain. The mutual enhancement of binding affinity between small molecule activator (bound to the ‘nest’ pocket near CNB-B) and cGMP (bound to CNB-A) only exists in the context of the full-length protein. This highlights the importance of the protein environment outside the regulatory domain in facilitating the positive cooperativity between the two sites. This binding cooperativity has direct relevance in the disease state where cGMP is low and PKG1α likely is partially activated by either cGMP or cAMP. We speculate that the compound-induced modulation of PKG1α activity may involve disruption of the switch helix ‘knob’ interactions that regulate kinase activity. How the displacement of the ‘knob’ from the ‘nest’ pocket elicits conformational changes leading to the opening of the catalytic domain remains to be uncovered and will benefit from efforts to solve the structure of the full-length PKG1α.

      Experimental procedures

      Protein purification

      Full-length PKG1α

      The full-length protein (human PKG1α2-671) was cloned into pBAC1 for insect cell expression as N-terminal 8× histidine fusion protein with a TEV protease site for cleaving the tag. The nucleotide sequence was confirmed by DNA sequencing (Genewiz Inc). The protein was expressed in Sf9 cells supplemented with Sf900 II media (Thermofisher) for 72 h at 27 °C using P2 BIICs at 0.5 multiplicity of infection. For the purification, the insect cell pellet was lysed in 50 mM Hepes, pH 7.5, 300 mM NaCl, 1 mM Tris(2-carboxyethyl)phosphine (TCEP), 1 ml/L protease inhibitor cocktail III (EMD Biosciences), and 5% glycerol. After lysis using a microfluidizer, the lysate was clarified by ultracentrifugation (100,000g for 1 h at 4 °C) and loaded onto a Ni-NTA agarose column pre-equilibrated with lysis buffer. The protein was eluted with 0 to 250 mM imidazole gradient. The fractions containing PKG1α2-671 were pooled, concentrated, and further purified using S75 gel filtration column (GE Healthcare). Fractions that showed >95% pure PKG1α2-671 based on SDS-PAGE analyses were pooled for storage. The final storage buffer of the protein was 50 mM Hepes, pH 7.5, 200 mM NaCl, 1 mM TCEP, and 10% glycerol. The concentration of the purified protein was determined using UV spectrophotometry at an extinction coefficient at 280 nm of 75,915 M−1 cm−1 based on the amino acid sequence of the designed protein construct.

      Regulatory domain fragment, PKG1α79-356 (CNB sites and switch helix)

      The isolated cGMP-binding domain fragment, human PKG1α79-356 or bovine PKG1α79-356, was cloned into pET47b for bacterial expression as N-terminal 6× histidine fusion protein. The nucleotide sequence was confirmed by DNA sequencing (Genewiz Inc). The protein was expressed in Escherichia coli BL21 (DE3) Star cells (Novagen) for 18 h at 16 °C with 1 mM IPTG. For the purification, the bacterial pellet was lysed in 50 mM Hepes pH 7.5, 300 mM NaCl, 5 mM MgCl2, 1 mM TCEP, 1 ml/L protease inhibitor cocktail III (EMD Biosciences) and 5% glycerol. After lysis using a microfluidizer, the lysate was clarified by ultracentrifugation (100,000g for 1 h at 4 °C) and loaded onto a Ni-NTA agarose column pre-equilibrated with lysis buffer. The protein was eluted with 0 to 250 mM imidazole gradient. The fractions containing PKG1α79-356 were pooled, concentrated, and further purified using S75 gel filtration column (GE Healthcare). Fractions that showed >95% pure PKG1α79-356 based on SDS-PAGE analyses were pooled for storage. The final storage buffer of the protein was 50 mM Hepes pH 7.0, 300 mM NaCl, 5 mM MgCl2, 1 mM TCEP, and 5% glycerol. The concentration of the purified protein was determined using UV spectrophotometry and an extinction coefficient at 280 nm of 31,775 M−1 cm−1 based on the amino acid sequence of the designed protein construct.

      PKG1α ADP-Glo kinase assay

      Full-length PKG1α (human PKG1α2-671) kinase activity (basal or partially activated by cGMP) in the presence of Glasstide peptide substrate (AnaSpec Inc) and ATP was measured using luminescence-based detection to quantify ADP product formation (ADP-Glo, Promega). Enzyme and substrate working solutions were prepared using the following assay buffer conditions: 50 mM Hepes, 10 mM MgCl2, 150 mM NaCl, 1 mM EDTA, 0.1 mg/ml bovine serum albumin (BSA), 5 mM β-mercaptoethanol (BME), 0.01% Triton X-100, pH 7.3 (BSA, BME, and Triton X-100 were added fresh to stock buffer solution on the day of experiment). Compounds (200 nl in dimethyl sulfoxide [DMSO]) were transferred to 384-well assay plates using Echo acoustic liquid handler (Beckman-Coulter). DMSO (200 nl) and cGMP solution (200 nl, 500 μM) were added to designated control wells on each plate (EC20 and EC100 control wells, respectively). A solution (8 μl) containing 6.25 nM PKG1α, 187.5 nM Glasstide, and 12.5 nM cGMP was added across all wells. After 30 min incubation at room temperature (RT), 2 μl of ATP solution was added to each well to start the reaction (final assay conditions: 5 nM PKG1α, 150 μM Glasstide, 1 mM ATP, 10 nM cGMP, test compound). After 2 h of incubation at RT, 10 μl of ADP-Glo Max Reagent was added to stop the reaction and eliminate excess ATP. Following a 40 min incubation, 20 μl of ADP-Glo Max Detection solution was added to the wells. At the end of a 1 h incubation, luminescence signal was read using a PHERAstar Plus (BMG) or Envision (PerkinElmer) plate reader.

      Data analysis

      Percent PKG1α activation by test compound was calculated relative to baseline (EC20 controls) and fully activated signals (EC100 controls). EC50 values were estimated by fitting % PKG1α activation versus compound concentration to a four-parameter activation dose response equation using Abase (IDBS) or PRISM (GraphPad) software.

      PKG1α homogeneous time resolved fluorescence assay

      Full-length PKG1α (human PKG1α2-671) kinase activity was measured using homogeneous time resolved fluorescence method using the homogeneous time resolved fluorescence KinEASE assay kit (CisBio). The assay uses a biotinylated peptide substrate (STK-2), which when phosphorylated is recognized by a monoclonal antiphosphoresidue antibody, labeled with Eu3+cryptate (donor fluorophore). Product detection occurs by an increase in fluorescence signal from streptavidin-XL665 (acceptor fluorophore) that binds to the biotinylated peptide. Enzyme and substrate working solutions were prepared in 50 mM Hepes, 10 mM MgCl2, 150 mM NaCl, 1 mM EDTA, 0.1 mg/ml BSA, 5 mM BME, and 0.01% Triton X-100, pH 7.3. To assess compound activation of partially activated PKG1α, 0.2 μl of compound (50× concentration in DMSO) or 0.2 μl DMSO were dispensed to designated test and EC20 control wells, respectively, in a 384-well plate (Corning, catalog no. 3676). A solution (5 μl) containing 10 pM PKG1α, 20 nM cGMP, and 2 μM STK2 peptide was then added to test wells and EC20 control wells. The same solution, but with 20 μM cGMP, was added to EC100 control wells. After 30 min incubation, 5 μl of 600 μM ATP was added to all the wells to start the reaction (final assay conditions: 5 pM PKG1α, 1 μM STK2 peptide substrate, 300 μM ATP, 10 nM cGMP (in EC20 control wells and test field wells), 10 μM cGMP (in EC100 control wells), varying concentrations of test compounds and 2% DMSO in assay buffer). The assay plate was then incubated for 30 min at RT. The reactions were quenched and fluorescence signal was developed by adding 10 μl detection solution containing 125 nM streptavidin-XL665 and 2× antibody-cryptate in detection buffer according to manufacturer’s protocol. Following 60 min incubation at RT, the emissions at 620 nm and 665 nm were read after excitation at 337 nm using PHERAstar Plus plate reader (BMG).

      Data analysis

      Percent PKG1α activation by test compound was calculated relative to baseline (EC20 controls) and fully activated signals (EC100 controls). EC50 values were estimated by fitting % PKG1α activation versus compound concentration to a four-parameter activation dose response equation using PRISM software.

      Microfluidic mobility shift assay

      Full-length PKG1α (human PKG1α2-671) kinase activity was directly measured using MMSA with a fluorescence-labeled peptide substrate, FL-Peptide 1 (5-FAM-AKRRRLSSLRA-COOH, catalog no. 760345, PerkinElmer) mixed with label-free peptide (AKRRRLSSLRA-COOH, catalog no. 27191, AnaSpec, Inc). Substrate to product conversion was monitored in a LabChip EZ Reader II (PerkinElmer) using a 12-sipper LabChip. Electrophoretic separation of substrate and product used the following separation conditions: downstream voltage −500 V and upstream voltage −2250 V with a screening pressure of −1.9 psi. The amount of product formed was determined according to the following ratio: product peak/(product + substrate peaks). For real time kinetic analysis, the reaction was monitored as it progressed by sequentially sipping samples onto the chip at various time intervals. To estimate the apparent KM for ATP, initial velocities were measured by assembling reaction solutions in a 384-well plate containing increasing amounts of ATP and 20 nM PKG1α in assay buffer (50 mM Hepes, 10 mM MgCl2, 150 mM NaCl, 0.01%BSA, 1 mM DTT, and 0.01% Triton X-100). The peptide substrate was then added to start the reaction and the assay plate was immediately loaded in the instrument (120 μM final peptide concentration; 1:10 mixture of labeled and unlabeled peptide). The reaction solutions were sampled every 5 minutes for 2 hours at 25 ⁰C. Initial rates were determined from the slopes of the linear regression plots at each ATP concentration and apparent KM was calculated from nonlinear regression analysis using the Michaelis–Menten equation (PRISM software). To determine KM for the peptide substrate, solutions containing 20 nM cGMP and various concentrations of peptide (1:10 mixture of labeled peptide and unlabeled peptide) were loaded in a 384-well plate. Saturating ATP was added (1 mM final concentration) to start the reaction. To estimate kinetic parameters, reaction progression curves at varying peptide concentrations were fitted globally using the following GraphPad Prism–based integrated Michaelis–Menten equation described by Zhang and Wong (
      • Zhang R.
      • Wong K.
      High performance enzyme kinetics of turnover, activation and inhibition for translational drug discovery.
      ) adopted from M. Golicnik (
      • Golicnik M.
      Explicit reformulations of time-dependent solution for a Michaelis-Menten enzyme reaction model.
      ,
      • Goličnik M.
      Explicit analytic approximations for time-dependent solutions of the generalized integrated Michaelis-Menten equation.
      ).
      Z=S0/Kf∗exp((S0-Vf∗X)/Kf)S=Kf∗(1.4586887∗ln(1.2∗Z/ln(2.4∗Z/ln(1 + 2.4∗Z))))0.4586887ln(2Z/ln(1 +2∗Z))Y=background+(S0S)/S0∗100


      As per GraphPad Prism formulation, all subscripts are written without lower offset. Vf and Kf are the maximal steady-state velocities (V) and Michaelis constants (K), respectively. X is the reaction time. Y is the raw turnover kinetics data (% product). S0 is the starting substrate concentration constrained as ‘equal to the column title’ in GraphPad Prism (
      • Zhang R.
      • Wong K.
      High performance enzyme kinetics of turnover, activation and inhibition for translational drug discovery.
      ).

      Microscale thermophoresis

      Full-length PKG1α (human PKG1α2-671) or regulatory domain fragment (bovine PKG1α79-356) was labeled with NT-647-NHS using the Monolith NT Protein Labeling Kit RED-NHS (NanoTemper, L001) following the manufacturer’s instructions. Briefly, PKG1α was exchanged and diluted into labeling buffer supplemented with 0.01% Triton X-100 and 0.1 mM TCEP to a final concentration of 16.2 μM for full-length human PKG1α2-671 and 43 μM for bovine PKG1α79-356 fragment. An equal volume of 43.5 μM NT-647-NHS in the same buffer was added and the resulting solution was incubated at RT for 30 min. Labeled PKG1α was purified by gel filtration into assay buffer (50 mM Hepes, pH 7.3, 150 mM NaCl, 10 mM MgCl2, 1 mM EDTA, 0.01% BSA, 0.01% Triton X-100, and 2 mM TCEP). The final product had a concentration of 3.1 μM or 4.8 μM and was labeled at 1:1.2 or 1:0.94 stoichiometry of PKG1α:NT-647 for (full-length) human PKG1α2-671 and bovine PKG1α79-356, respectively. For measurements of compound affinity to the target protein, 200 nl of serially diluted compound in DMSO was dispensed to 10 μl of 10 nM protein containing varying concentrations of cGMP in assay buffer. After a 30 min equilibration period, samples were loaded into Monolith NT.115 hydrophilic capillaries (NanoTemper, K004) and read on a Monolith NT.115 instrument using an MST power of 35% for human PKG1α2-671 (or 60% for bovine PKG1α79-356) and a LED power of 90%. The fluorescence measured after 5 seconds of thermophoresis was fit to a four-parameter dose-response equation in GraphPad Prism in order to determine Kd values.

      Fluorescence polarization binding assays

      CNB-A and CNB-B competition binding assays using fixed PKG1α and 8-fluo-cGMP concentrations

      FP assays were conducted in 50 mM Hepes, pH 7.3, 150 mM NaCl, 10 mM MgCl2, 1 mM EDTA, 0.01% BSA, 0.01% Triton X-100, and 2 mM TCEP in 384-well plates (Corning 4514) using a final volume of 10 μl. The affinity of the FP probe, 8-fluo-cGMP (Biolog Life Sci., Inc), for CNB-A of PKG1α was first determined by mixing 5 μl of a 2× concentration serial dilution of PKG1α with an equal volume of a 2× fixed concentration of the probe. The final concentration of the probe was 0.5 nM ensuring only binding to CNB-A was measured. Binding reactions were incubated at RT for 1 and 3 h, and the polarization state of the probe was measured on a PHERAstar Plus plate reader (BMG). No difference in the observed polarization was noted between the two time points confirming that equilibrium is reached within 1 h. The Kd of 8-fluo-cGMP was determined by fitting the PKG1α concentration response data in GraphPad Prism using a quadratic equation that accounts for ligand depletion as described by Zhang et al. (
      • Zhang R.
      • Mayhood T.
      • Lipari P.
      • Wang Y.
      • Durkin J.
      • Syto R.
      • et al.
      Fluorescence polarization assay and inhibitor design for MDM2/p53 interaction.
      ). The effect of test compounds on 8-fluo-cGMP binding to CNB-A was determined through equilibrium competition experiments using a fixed concentration of PKG1α (10 nM) and probe (0.5 nM). Briefly, a mixture of PKG1α and probe was added to the assay plate and followed by acoustic dispensing of 200 nl of serially diluted compound in DMSO. The resulting solution was mixed and incubated at RT for 1 h before measuring the polarization of 8-fluo-cGMP using the PheraStar plate reader. Competition experiments using PKG1α loaded with probe at both CNB-A and CNB-B sites were conducted as described above except PKG1α and probe were instead fixed at 100 nM and 200 nM, respectively.

      Sequential two-site binding assay

      To measure the affinity of 8-fluo-cGMP to the low affinity site (CNB-B) of PKG1α by FP, we devised a novel serial titration scheme. Since practical assessment of probe affinity for the target protein requires that (a) a fixed concentration of probe be used, (b) the concentration of the probe be below the Kd for the target, and (c) a significant fraction of the probe be bound to maximize the dynamic range of the mP signal, typical FP methods cannot measure probe binding to secondary low affinity sites because binding would always be superseded by quantitative binding to the higher affinity site. In order to address this problem, we employed a titration scheme (serial dilution) that simultaneously varied both the PKG1α and probe concentrations while maintaining the ratio of PKG1α and probe fixed at 1:2. This novel scheme allows for the sequential engagement of high and low affinity sites at the relevant concentration range for each binding site as the concentration of both PKG1 and the probe are increased while simultaneously preserving the high fractional binding of the probe required to maximize the dynamic range of the assay.
      Experimentally, conditions and procedure were the same as described previously for determination of affinities at CNB-A. Ten microliter of a fixed ratio serial dilution of PKG1α:probe (1:2) was added to the assay plate, followed by acoustic dispensing of 200 nl of compound in DMSO. Each PKG1α:probe concentration condition was read independently and normalized to a well with a matched probe concentration lacking PKG1α. To determine the binding constants for each site from this data, the concentration response curves were fit in GraphPad Prism using a set of custom equations; the derivations are outlined as follows for the target protein (P) and ligand (L):
      [P]+[L][PL]


      [PL]+[L][PLL]


      Kd1=[P][L][PL]
      (1)


      Kd2=[PL][L][PLL]
      (2)


      [L]0=[L]+[PL]+[PLL]
      (3)


      [P]0=[P]+[PL]+[PLL]
      (4)


      From Equations 1, 2, and 3, omitting [PL] and [PLL], we obtain
      [L]0=[L]+[P][L]Kd1+[P][L]2Kd1Kd2
      (5)


      From Equations 3 and 4
      [L]0[P]0=[L][P]
      (6)


      From Equations 5 and 6, omitting [P], we obtain
      [L]3+a[L]2+b[L]+c=0
      (7)


      Where
      a=Kd2+P0L0


      b=Kd1K22+Kd2(P0L0)


      c=L0Kd1Kd2


      Solving the physiologically meaningful root of Equation 7, we obtain
      [L]=a3+23(a23b)cosθ3
      (8)


      Where
      θ=arcos2a3+9ab27c2(a23b)3


      The fluorescence polarization as a function of [L], [PL], and [PLL] is equal to
      Y=FPL[L]+QCQLFPPL[PL]+2QCQLFPPL[PLL][L]+QCQL[PL]+2QCQL[PLL]
      (9)


      Where Y is the total FP value measured, FPL is the FP value of the free ligand (8-fluo-cGMP), FPPL is the FP value of the protein–ligand complex, and QC and QL are the molar emissivity of the protein–ligand and unbound ligand, respectively.
      The equations, as written in GraphPad Prism, are below.
      P0=XL0=nP0C=Kd1Kd2L0A=Kd2+P0L0B=Kd2(Kd1+P0L0)Theta=arccos[(2Aˆ3 +9AB27C)/2/SQRT((AA3B)ˆ3)]L=A/3+2/3SQRT(AA3B)cos(Theta/3)P=L+P0L0C1=PL/Kd1C2=C1L/Kd2Y=(FP_free+ FP_bound1R1C1+FP_bound2R2C2)/(L+R1C1 + R2C2)


      1H STD NMR binding analysis

      Ligand binding was detected by STD NMR spectroscopy. The binding buffer contained 25 mM Hepes-d18 (Cambridge Isotope Laboratories), pH 7.3, 150 mM NaCl, 10 mM MgCl2, 2 mM TCEP, 5 μM full-length PKG1α (human PKG1α2-671) or isolated regulatory domain (human PKG1α79-356) with and without ligands (cGMP, cAMP, or piperidine series activators), and 25 μM TSP in 99.98% D2O. All STD NMR spectra were collected at 298 K on a Bruker 600 MHz Avance spectrometer. Selective saturation of the protein was applied by switching the on- and off-resonance saturation frequency after each scan. A train of Gaussian shape pulses with 50 ms pulse length (corresponding to an excitation width of 100 Hz) separated by a delay of 1 ms was used, with the total length of the selective saturation set to 3 s, and the on- and off-resonance saturation frequencies set to −120 and 20,000 Hz. The total time to acquire one STD NMR dataset, including time to change samples, was 50 min.

      Computational modeling methods

      Force field

      Structures were prepared and refined with the AMBER14:EHT force field (
      • Gerber P.R.
      • Müller K.
      MAB, a generally applicable molecular force field for structure modelling in medicinal chemistry.
      ,
      • Maier J.A.
      • Martinez C.
      • Kasavajhala K.
      • Wickstrom L.
      • Hauser K.E.
      • Simmerling C.
      ff14SB: improving the accuracy of protein side chain and backbone parameters from ff99SB.
      ).

      Site identification

      The PKG1α regulatory domain monomer was processed by the Site Finder tool in MOE 2019.0102 (
      ) and the 5 largest sites were retained for further analysis.

      Docking

      Molecules were docked using the docking application in MOE 2019.0102 (
      ) using the triangle matcher placement algorithm with up to 100 poses retained for placement and 50 poses for the final refinement. Molecules were docked to previously identified sites.

      Data availability

      All data are contained within the manuscript

      Supporting information

      This article contains supporting information.

      Conflict of interest

      The authors declare that they have no conflicts of interest with the contents of this article.

      Acknowledgments

      We thank Dr Nathaniel Elsen for running the ITC analysis on cGMP binding to full-length PKG1α.

      Author cotributions

      W. M. S., D. B., and A. S. conceptualization; P. T., M. A. M., R. Z., and A. S. methodology; P. T., E. M., M. A. M., R. Z., and A. S. data curation; P. T., L. Z., E. M., M. A. M., W. M. S., Y. H., E. F., P. S., J. H., C. S., and A. S. investigation; P. T., L. Z., Y. H., E. F., J. H., C. S., and A. S. formal analysis; P. T., L. Z., Y. H., E. F., J. H., C. S., and A. S. validation; P. T., L. Z., E. M., Y. H., M. A. M., W. M. S., R. Z., E. F., P. S., J. H., C. S., D. B., and A. S. writing–review and editing.

      Supporting information

      Figure thumbnail figs1
      Figure S1Positive modulation of cGMP activation potency by piperidine series activators. cGMP EC50 estimates in the presence of increasing concentrations of 1 (A) and 2 (B). The red arrow indicates left shift of the concentration-response curves (decrease in cGMP EC50) in the presence of increasing concentrations of compounds. Inset graph for each panel shows cGMP EC50 values for best-fit curves plotted against concentration of compound (error bars indicate SE from two independent trials). PKG1a activity (initial velocities) were measured using Microfluidic Mobility Shift assay.
      Figure thumbnail figs2
      Figure S2Validation of sequential two-site–binding assay (FP method) to estimate affinity of 8-fluo-cGMP for CNB-A and CNB-B. A, sequential two-site binding of 8-fluo-cGMP tracked by the measurement of polarization of 8-fluo-cGMP at increasing concentrations of both PKG1α and 8-fluo-cGMP (at 1:2 ratio, [PKG1α]:[8-fluo-cGMP]). The data was fitted using Equation 9 to estimate the apparent Kd of 8-fluo-cGMP for CNB-A and CNB-B. B, ITC titration of cGMP into PKG1α. C, Kd estimates using FP (from panel A) closely match orthogonal estimate using ITC (from panel B). Values are expressed as Mean ± SE from two to three independent measurements (see for details).
      Figure thumbnail figs3
      Figure S3Demonstration of linear relationship between binding affinity of 8-fluo-cGMP and concentration of orthosteric competitor. Affinity estimates of 8-fluo-cGMP for CNB-A and CNB-B at various concentrations of cGMP (Aa) or cAMP (Ba) using sequential two-site equilibrium-binding assay (FP method). The presence of increasing concentrations of direct competitors decreased the apparent affinity of 8-fluo-cGMP for the two binding sites as indicated by rightward shift of the concentration-response curves. Error bars represent SE from two replicate measurements. A linear relationship is shown between apparent affinity of 8-fluo-cGMP for the high affinity CNB-A site with increasing concentrations of cGMP (Panel Ab) or cAMP (Panel Bb) as predicted for orthosteric competitors.
      Figure thumbnail figs4
      Figure S4Demonstration of orthogonal and allosteric binding of ligands to the regulatory domain (PKG1a79-356) tracked by 1H STD NMR. A, direct (orthogonal) competition between cAMP and cGMP for the same binding site. Bottom panel, binding of cAMP to PKG1a79-356 indicated by observed peaks in the STD difference spectrum (blue). Top panel shows loss of cAMP STD signals in the presence of saturating cGMP, indicating direct competition for the same binding site (Note: cGMP STD signal was not detectable due to its higher affinity for PKG1a). B, allosteric binding of 1 and cAMP to regulatory domain. Bottom panel shows observed peaks in the STD spectrum (blue) due to binding of 1 to PKG1a79-356 in the absence of cyclic nucleotide. Middle panel shows peaks in STD spectrum (red) due to binding of cAMP to regulatory domain in the absence of compound. Top panel shows STD spectrum (black) when PKG1a79-356 is in equilibrium with both 1 and cAMP in solution. The appearance of (signature) peak signals for both 1 and cAMP indicate nonoverlapping binding sites within the regulatory domain.

      References

        • Dang T.A.
        • Schunkert H.
        • Kessler T.
        cGMP signaling in cardiovascular diseases: linking genotype and phenotype.
        J. Cardiovasc. Pharmacol. 2020; 75: 516-525
        • Blanton R.M.
        cGMP signaling and modulation in heart failure.
        J. Cardiovasc. Pharmacol. 2020; 75: 385-398
        • Dunkerly-Eyring B.
        • Kass D.A.
        Myocardial phosphodiesterases and their role in cGMP regulation.
        J. Cardiovasc. Pharmacol. 2020; 75: 483-493
        • Preedy M.E.J.
        • Baliga R.S.
        • Hobbs A.J.
        Multiplicity of nitric oxide and natriuretic peptide signaling in heart failure.
        J. Cardiovasc. Pharmacol. 2020; 75: 370-384
        • Sandner P.
        • Zimmer D.P.
        • Milne G.T.
        • Follmann M.
        • Hobbs A.
        • Stasch J.P.
        Correction to: soluble guanylate cyclase stimulators and activators.
        Handb. Exp. Pharmacol. 2021; 264: 425
        • Park M.
        • Sandner P.
        • Krieg T.
        cGMP at the centre of attention: emerging strategies for activating the cardioprotective PKG pathway.
        Basic Res. Cardiol. 2018; 113: 24
        • Kim G.E.
        • Kass D.A.
        Cardiac phosphodiesterases and their modulation for treating heart disease.
        Handb. Exp. Pharmacol. 2017; 243: 249-269
        • Ignarro L.J.
        • Kadowitz P.J.
        The pharmacological and physiological role of cyclic GMP in vascular smooth muscle relaxation.
        Annu. Rev. Pharmacol. Toxicol. 1985; 25: 171-191
        • Moro M.A.
        • Russel R.J.
        • Cellek S.
        • Lizasoain I.
        • Su Y.
        • Darley-Usmar V.M.
        • et al.
        cGMP mediates the vascular and platelet actions of nitric oxide: confirmation using an inhibitor of the soluble guanylyl cyclase.
        Proc. Natl. Acad. Sci. U. S. A. 1996; 93: 1480-1485
        • Feil R.
        • Lohmann S.M.
        • de Jonge H.
        • Walter U.
        • Hofmann F.
        Cyclic GMP-dependent protein kinases and the cardiovascular system: insights from genetically modified mice.
        Circ. Res. 2003; 93: 907-916
        • Hofmann F.
        • Bernhard D.
        • Lukowski R.
        • Weinmeister P.
        cGMP regulated protein kinases (cGK).
        Handb. Exp. Pharmacol. 2009; https://doi.org/10.1007/978-3-540-68964-5_8
        • Takimoto E.
        Cyclic GMP-dependent signaling in cardiac myocytes.
        Circ. J. 2012; 76: 1819-1825
        • Pfeifer A.
        • Ruth P.
        • Dostmann W.
        • Sausbier M.
        • Klatt P.
        • Hofmann F.
        Structure and function of cGMP-dependent protein kinases.
        Rev. Physiol. Biochem. Pharmacol. 1999; 135: 105-149
        • Moon T.M.
        • Sheehe J.L.
        • Nukareddy P.
        • Nausch L.W.
        • Wohlfahrt J.
        • Matthews D.E.
        • et al.
        An N-terminally truncated form of cyclic GMP-dependent protein kinase Iα (PKG Iα) is monomeric and autoinhibited and provides a model for activation.
        J. Biol. Chem. 2018; 293: 7916-7929
        • Alverdi V.
        • Mazon H.
        • Versluis C.
        • Hemrika W.
        • Esposito G.
        • van den Heuvel R.
        • et al.
        cGMP-binding prepares PKG for substrate binding by disclosing the C-terminal domain.
        J. Mol. Biol. 2008; 375: 1380-1393
        • Hofmann F.
        The biology of cyclic GMP-dependent protein kinases.
        J. Biol. Chem. 2005; 280: 1-4
        • Osborne B.W.
        • Wu J.
        • McFarland C.J.
        • Nickl C.K.
        • Sankaran B.
        • Casteel D.E.
        • et al.
        Crystal structure of cGMP-dependent protein kinase reveals novel site of interchain communication.
        Structure. 2011; 19: 1317-1327
        • Moon T.M.
        • Osborne B.W.
        • Dostmann W.R.
        The switch helix: a putative combinatorial relay for interprotomer communication in cGMP-dependent protein kinase.
        Biochim. Biophys. Acta. 2013; 1834: 1346-1351
        • Busch J.L.
        • Bridges T.M.
        • Richie-Jannetta R.
        • Hollett B.P.
        • Francis S.H.
        • Corbin J.D.
        Catalytic site amino acids of PKGI-alpha influence allosteric cGMP binding.
        Front. Biosci. (Schol Ed. 2013; 5: 650-660
        • Francis S.H.
        • Busch J.L.
        • Corbin J.D.
        • Sibley D.
        cGMP-dependent protein kinases and cGMP phosphodiesterases in nitric oxide and cGMP action.
        Pharmacol. Rev. 2010; 62: 525-563
        • Butt E.
        cGMP-dependent protein kinase modulators.
        Handb. Exp. Pharmacol. 2009; https://doi.org/10.1007/978-3-540-68964-5_17
        • Moon T.M.
        • Tykocki N.R.
        • Sheehe J.L.
        • Osborne B.W.
        • Tegge W.
        • Brayden J.E.
        • et al.
        Synthetic peptides as cGMP-independent activators of cGMP-dependent protein kinase Iα.
        Chem. Biol. 2015; 22: 1653-1661
        • Hanisak J.
        • Soriano A.
        • Adam G.C.
        • Basso A.
        • Bauman D.
        • Bell D.
        • et al.
        Discovery of the first non-cGMP mimetic small molecule activators of cGMP-dependent protein kinase 1 α (PKG1α).
        ACS Med. Chem. Lett. 2021; 12: 1275-1282
        • Francis S.H.
        • Blount M.A.
        • Zoraghi R.
        • Corbin J.D.
        Molecular properties of mammalian proteins that interact with cGMP: Protein kinases, cation channels, phosphodiesterases, and multi-drug anion transporters.
        Front. Biosci. 2005; 10: 2097-2117
        • Døskeland S.O.
        • Vintermyr O.K.
        • Corbin J.D.
        • Ogreid D.
        Studies on the interactions between the cyclic nucleotide-binding sites of cGMP-dependent protein kinase.
        J. Biol. Chem. 1987; 262: 3534-3540
        • McCune R.W.
        • Gill G.N.
        Positive cooperativity in guanosine 3':5'-monophosphate binding to guanosine 3':5'-monophosphate-dependent protein kinase.
        J. Biol. Chem. 1979; 254: 5083-5091
        • Hofmann F.
        • Flockerzi V.
        Characterization of phosphorylated and native cGMP-dependent protein kinase.
        Eur. J. Biochem. 1983; 130: 599-603
        • Kim J.J.
        • Lorenz R.
        • Arold S.T.
        • Reger A.S.
        • Sankaran B.
        • Casteel D.E.
        • et al.
        Crystal structure of PKG I:cGMP complex reveals a cGMP-mediated dimeric interface that facilitates cGMP-induced activation.
        Structure. 2016; 24: 710-720
        • Mak V.W.
        • Patel A.M.
        • Yen R.
        • Hanisak J.
        • Lim Y.-H.
        • Bao J.
        • et al.
        Optimization and mechanistic investigations of novel allosteric activators of PKG1α.
        J. Med. Chem. 2022; https://doi.org/10.1021/acs.jmedchem.1c02109
        • Zhang R.
        • Wong K.
        High performance enzyme kinetics of turnover, activation and inhibition for translational drug discovery.
        Expert Opin. Drug Discov. 2017; 12: 17-37
        • Golicnik M.
        Explicit reformulations of time-dependent solution for a Michaelis-Menten enzyme reaction model.
        Anal. Biochem. 2010; 406: 94-96
        • Goličnik M.
        Explicit analytic approximations for time-dependent solutions of the generalized integrated Michaelis-Menten equation.
        Anal. Biochem. 2011; 411: 303-305
        • Zhang R.
        • Mayhood T.
        • Lipari P.
        • Wang Y.
        • Durkin J.
        • Syto R.
        • et al.
        Fluorescence polarization assay and inhibitor design for MDM2/p53 interaction.
        Anal. Biochem. 2004; 331: 138-146
        • Gerber P.R.
        • Müller K.
        MAB, a generally applicable molecular force field for structure modelling in medicinal chemistry.
        J. Comput. Aided Mol. Des. 1995; 9: 251-268
        • Maier J.A.
        • Martinez C.
        • Kasavajhala K.
        • Wickstrom L.
        • Hauser K.E.
        • Simmerling C.
        ff14SB: improving the accuracy of protein side chain and backbone parameters from ff99SB.
        J. Chem. Theor. Comput. 2015; 11: 3696-3713
      1. Molecular Operating Environment (MOE). 2019.01 Ed. Chemical Computing Group ULC, QC, Canada2021