Computational Identification of Novel Kir6 Channel Inhibitors

KATP channels consist of four Kir6.x pore–forming subunits and four regulatory sulfonylurea receptor (SUR) subunits. These channels couple the metabolic state of the cell to membrane excitability and play a key role in physiological processes such as insulin secretion in the pancreas, protection of cardiac muscle during ischemia and hypoxic vasodilation of arterial smooth muscle cells. Abnormal channel function resulting from inherited gain or loss-of-function mutations in either the Kir6.x and/or SUR subunits are associated with severe diseases such as neonatal diabetes, congenital hyperinsulinism, or Cantú syndrome (CS). CS is an ultra-rare genetic autosomal dominant disorder, caused by dominant gain-of-function mutations in SUR2A or Kir6.1 subunits. No specific pharmacotherapeutic treatment options are currently available for CS. Kir6 specific inhibitors could be beneficial for the development of novel drug therapies for CS, particular for mutations, which lack high affinity for sulfonylurea inhibitor glibenclamide. By applying a combination of computational methods including atomistic MD simulations, free energy calculations and pharmacophore modeling, we identified several novel Kir6.1 inhibitors, which might be possible candidates for drug repurposing. The in silico predictions were confirmed using inside/out patch-clamp analysis. Importantly, Cantú mutation C166S in Kir6.2 (equivalent to C176S in Kir6.1) and S1020P in SUR2A, retained high affinity toward the novel inhibitors. Summarizing, the inhibitors identified in this study might provide a starting point toward developing novel therapies for Cantú disease.

Introduction

Cantú syndrome (CS) is a rare genetic autosomal dominant disorder caused by dominant gain-of-function mutations in the ATP-dependent potassium (KATP) channel genes ABCC9 (Harakalova et al., 2012; Van Bon et al., 2012) and KCNJ8 (Brownstein et al., 2013; Cooper et al., 2014, 2017), encoding sulfonylurea receptor 2 SUR2 and inward rectifier potassium channel 6.1 Kir6.1, respectively. Cantú patients are chronically ill; they suffer from congenital hypertrichosis, distinctive facial features and cardiac defects (Cantú et al., 1982; Scurr et al., 2011; Nichols et al., 2013) and have a decreased life expectancy. Currently, no specific pharmacotherapeutic options are available to treat the disease (Kharade et al., 2016).

Recent breakthroughs in solving atomic and near-atomic resolution structures of eukaryotic inward rectifier potassium (Kir) channels provide an excellent opportunity to investigate the structural basis of CS mutations and for developing novel therapies for KATP channelopathies. Starting from January 2017, the first (near-)atomic resolution structures (resolution ranges from 3.63 to 6.3 Å) of these hetero-octameric complexes have been solved by cryo-EM microscopy by three independent labs (Lee et al., 2017; Li et al., 2017; Martin et al., 2017a,b; Wu et al., 2018). These structures confirm that KATP channels are formed by four Kir6.x pore–forming subunits and four regulatory sulfonylurea receptor (SUR) subunits.

KATP channels couple the metabolic state of the cell to membrane excitability and play a key role in physiological processes such as insulin secretion in the pancreas (Ashcroft, 2005), protection of cardiac muscle during ischemia (Nichols and Lederer, 1991; Crawford et al., 2002; Zingman et al., 2007), and hypoxic vasodilation of arterial smooth muscle cells (Dart and Standen, 1995).

Channel activity is regulated by voltage and ligands. While inhibitory adenosine-triphosphate (ATP) binds to the Kir6.x subunit, magnesium-adenosine-triphosphate-(MgATP), and adenosine-diphosphate (ADP) activate the channel via interacting with the SUR subunits (Matsuo et al., 1999, 2000; Tanabe et al., 1999; Ueda et al., 1999; MacGregor et al., 2002; Vanoye et al., 2002). Phospholipid phosphatidylinositol-4,5-bisphosphate (PIP₂), is necessary for channel opening of all inward rectifying potassium channels and binds to the Kir6.x subunit (Huang et al., 1998; Shyng and Nichols, 1998; Zhang et al., 1999). Abnormal channel function, resulting from inherited gain or loss-of-function mutations in either the Kir6.x and/or SUR subunits are associated with severe diseases such as neonatal diabetes, congenital hyperinsulinism and CS (Remedi and Nichols, 2009; Harakalova et al., 2012). Further, SUR2 subunits have been shown to play a role in human neurological disease, including prevalent diseases of the aged brain (Nelson et al., 2015).

Pharmaceutical interventions in KATP channels include sulfonylurea-like inhibitors such as glibenclamide and channel openers, such as diazoxide, which are clinically used to treat neonatal diabetes and hypertension and target the sulfonylurea subunits (Gribble and Reimann, 2003; Pearson et al., 2006). More recently, side effects due to inhibition of KATP channels have been reported as well. For example, in Yu et al. (2011) reported that all isoforms of KATP channels are blocked by rosiglitazone (RSG) at micro molar concentrations, which could be harmful due to promotion of adverse cardiovascular effects. RSG is a high-affinity agonist of the peroxisome proliferator-activated receptor γ, which was introduced in 1999 for the treatment of type II diabetes mellitus. The drug increases insulin sensitivity in fat cells by regulating genes involved in glucose and lipid metabolism and might have additional beneficial effects including anti-atherosclerotic, anti-inflammatory, and anticancer effects (Brown and Plutzky, 2007). Due to reports of increased risk of myocardial infarction, RSG was withdrawn from the market in Europe in 2010 (Agency European Medicines, 2010) and had its access restricted in the United States by the FDA in 2011 (U.S. Food and Drug Administration, 2011). Recently it was shown that this increased cardiovascular risk might be due to modification of different ion channels including Kv4.3, L-type calcium channels and KATP channels (Hancox, 2011; Jeong et al., 2011; Szentandrássy et al., 2011; Yu et al., 2011, 2012). Studies on pigs demonstrate that RSG and other thiazolidinedione drugs can block cardiac sarcolemmal KATP channels in vivo at clinically relevant doses (Lu et al., 2008). The reported IC₅₀ of this drug is 45 μM for Kir6.2/SURx (pancreatic and heart) channels and 10 μM for vascular Kir6.1/SUR2B. Interestingly, potency has been shown to be even higher in the presence of therapeutic concentrations of sulfonylureas reducing the IC₅₀ to 2 μM. Since plasma concentrations of RSG used to treat type II diabetes mellitus are in the range of 3 μM (Cox et al., 2000), block of KATP channels could be a serious problem. Experiments performed on Kir6.2ΔC36 constructs revealed that the drug acts predominantly on the pore-forming Kir6.x subunits and not on the SUR subunits. Further analysis of single KATP channels suggests that the drug suppresses channel activity by extending long-lasting channel closures, most likely via modulating the gating mechanism (Yu et al., 2012).

Kir6 inhibitors such as RSG, which block channels at clinically relevant doses, could provide a good starting point toward development of novel, specific inhibitors, suitable for developing drugs toward treatment of CS. Thus, in this study, we applied a novel combination of computational methods to investigate the structural mechanism of block of RSG in Kir6 channels. We carried out extensive unbiased full atomistic simulations of drug binding to the closed channel state. Based on the thereby identified binding site, we postulate a structural mechanism by which the drug might prolong the long-closed state of the channel. Further, structure-based pharmacophore models were constructed to enable identification of novel Kir6 inhibitors, which might be useful for future drug development to treat Cantú disease.

Results and Discussion

Based on the experimental finding that RSG predominantly acts on the long-closed state of Kir6.1 channels (Yu et al., 2012), a homology model of the Kir6.1 pore model was constructed using the closed state crystal structure of the Kir3.2 channel [Protein Data Bank (PDB) code: 3SYA (Whorton and MacKinnon, 2011)] as template. A sequence alignment is shown in Supplementary Figure S1A. The root mean square deviation (RMSD) of the Kir6.1 model converged to ∼4 Å at around 100 ns, indicating that the simulated systems are stable and at equilibrium (see Supplementary Figure S1C). Starting in October 2017, the first atomic resolution structures of the KATP channel formed by Kir6.2 and SUR1 were published (Lee et al., 2017; Li et al., 2017; Martin et al., 2017a, 2017b; Wu et al., 2018). Thus, we compared the structural differences of our Kir3.2 based homology model with the new Kir6.2 templates. Due to the low RMSD of the structural alignments (<1.5 Å, see Supplementary Figure S1B) we continued to use the Kir6.1 homology model based on the Kir3.2 template in subsequent molecular dynamics (MD) simulations.

Unbiased μs Time Scale MD Simulations Identify Putative RSG Binding Sites

In an effort to identify the putative binding site of RSG (5-[[4-[2-[methyl(pyridin-2-yl)amino]ethoxy]phenyl]methyl]-1,3-thiazolidine-2,4-dione), and its main metabolite N-desmethyl Rosiglitazone (5-[[4-[2-(pyridin-2-ylamino) ethoxy]phenyl]methyl]-1,3-thiazolidine-2,4-dione), N-RSG for short, full atomistic MD simulations were performed. Specifically, the binding was probed by adding 20 molecules (10 × S conformer, 10 × R conformer, since the prescribed drug is a racemic mixture) randomly into the solvent, leading to an effective drug concentration of ∼170 mM. As seen in Figure 1, 13 ligands partition into the lipid membrane within 1.5 μs. Three out of 20 molecules occupied sites close to the protein for 1.3 – 1.5 μs (see Table 1). The three major binding sites (Figure 2) identified are: close to the PIP₂ binding site, denoted site A; the interface between two cytoplasmic domains (CTDs), denoted site B, and between the β-sheet βD and the βG-αG loop, denoted site C.

FIGURE 1

Figure 1. Simulation setups: The first and last frame of free MD simulation. Protein (gray cartoon) is embedded in the lipid bilayer (cyan lines with red head groups). As a starting simulation setup (t = 0), 20 ligands (shown as spheres) were randomly distributed in the solvent. After a 1.5 μs MD simulation, the ligands either are bound to the protein or entered the lipid phase.

TABLE 1

Table 1. Three binding sites observed during unbiased MD simulations and their corresponding occupation time.

FIGURE 2

Figure 2. Three binding sites observed after 1.5 μs unbiased MD simulation. The protein is represented as cartoon; adjacent subunits are colored in green and gray, respectively. PIP₂ molecules are shown as orange sticks. Ligands are represented as purple spheres. (A) Side view of the whole protein highlighting the three ligands bound to binding sites A – C. Site A: in close proximity to the PIP₂ binding site; Site B: at the interface between 2 CTDs; and Site C: between the β-sheet βD and the βG-αG loop. (B) Bottom view of Site B and Site C. (C) Detail view of site A: residues within 6 Å of the ligand are labeled and presented as gray sticks. The ligand mainly forms hydrophobic interaction with LEU57 and ILE60; additionally, it forms hydrophobic and hydrogen bond (HB) interactions with lipid molecules (transparent sticks in light gray).

Binding site A is of particular interest since the ligand binds close to the transmembrane gate and the PIP₂ binding site, a plausible location for extending long-lasting channel closures, as suggested by experiments (Yu et al., 2012). Binding includes mainly hydrophobic interactions with residue LEU57 and ILE60. Additionally, hydrophobic and hydrogen bond (HB) interactions with POPC lipid molecules are observed. Even though PIP₂ is within 5 Å of ligand no specific interactions with RSG are observed. All residues within 6 Å of ligand are shown in Figure 2C. Given the importance of this area for channel gating (Zhang et al., 2015) and the surface exposure of the binding site, we decided to perform a more exhaustive sampling of this region. We therefore used computationally less demanding docking simulations with both compounds, followed by 250 ns MD simulations of the best-scored pose for RSG, and 200 ns of the best-scored pose for N-RSG. This way, strengthened binding interactions (defined as increased accessible surface area: 0.9 nm² for site A vs. 3 nm² when using the best docking pose) were obtained. We refer to this new site A as site A_ref. Since binding sites B is in a more constrained pocket (see Supplementary Figure S6) and site C is in a region not shown to be relevant for channel gating, we did not further attempt to refine these two sites.

To further characterize the three binding sites, the potential of mean force (PMF) of the ligands at site A_ref, B and C, were calculated using umbrella sampling (US) simulations. As can be seen in Figure 3, the PMFs reveal clear differences between the three sites. Binding of RSG to site A_ref is most favorable. Since PMF calculations reveal shallow binding for ligands bound at sites B and C, we did not further investigate these two binding sites. Nevertheless, ligand binding site B, might be potentially interesting, since intersubunit interactions in this region have been shown previously to be important for the inactivation process in Kir6.2 channels (Lin et al., 2003; Borschel et al., 2017) and could be further investigated in future studies.

FIGURE 3

Figure 3. Potential of Mean Forces (PMFs) derived from US for both RSG and N-RSG at the three binding sites. The energy profiles are depicted including their standard deviations. The WHAM histograms are shown in Supplementary Figure S4. Site A_ref shows higher binding affinity for both, RSG and N-RSG, compared to the other binding sites.

Detailed Characterization of Binding Site A_ref

The interactions of RSG and N-RSG at site A_ref over 200 ns of the MD simulations were quantified using interaction matrices as described in the methods section. Ligands were decomposed as five parts (Figure 4A and Supplementary Figure S2): the pyridine (Ring A), the benzene (Ring B), the thiazolidinedione (Ring C), the linker connecting pyridine and benzene (Linker D), and the linker connecting benzene and thiazolidinedione (Linker E). RSG mainly formed hydrophobic interactions with the binding site throughout the trajectory. More than 75% of the frames contain hydrophobic interactions between the Ring A and residue PHE76, VAL172, and MET173, and between the Linker E and TRP69. Residues LEU64, ILE75, PHE80, ILE168, and ILE169 also frequently formed hydrophobic interactions with RSG (Figures 4B,D). The interaction map of N-RSG molecule reveals a similar trend indicating mainly hydrophobic interactions with the binding site (see Supplementary Figure S2). Both ligands also form HBs with POPC lipid molecules, which are not included in the interaction map calculations (RMSD plots of ligands are shown in Figure 4C and Supplementary Figure S2C). For relative location of binding site A_ref with respect to the SUR subunit, see Supplementary Figure S7.

FIGURE 4

Figure 4. RSG interactions with the protein and PIP₂ at binding site A_ref. (A) Molecular structure of RSG including the denotation corresponding to the interaction map. (B) Interaction map of RSG with protein and PIP₂ during 200 ns MD simulation. The matrix is colored and numbered by the percentage of frames, in which interactions were observed: aromatic (AR), hydrophobic (H), and hydrogen bond (HB). The residues in the Kir6.1 are named by the corresponding amino acid with its residue number and chain ID (A or D). (C) RMSD plot of RSG at binding site A_ref during a 250 ns MD simulation. (D) Best PMF energy pose: Kir6.1 is represented as cartoon with two neighboring subunits colored in gray and green, respectively. RSG (purple), the surrounding residues within 3.5 Å, and the PIP₂ (orange) are shown as stick.

Structure Based Dynamic Pharmacophore Models of RSG Binding to Site A_ref

Understanding the inhibition mechanisms of the pore-forming Kir6.1 subunit could be a first step to develop novel blockers for the treatment of rare disease causing mutations (e.g., Cantú mutations V64M), which are not amenable for sulfonylurea therapy (Cooper et al., 2017). In line with this reasoning we constructed dynamic structure-based pharmacophore models and screened for hits in DrugBank (Law et al., 2014), which contains all marketed drugs, by using the common hits approach (CHA) (Wieder et al., 2017). Structure-based pharmacophore models were generated using 5,000 frames from the MD simulation, which included the lipid bilayer but omitted solvent molecules. Pharmacophore models, which contain common pharmacophore features and identical involved ligand atoms, are considered as one representative pharmacophore model. Five representative pharmacophore models (Model 1 – 5) were observed from the frames (shown in Supplementary Figure S3). Model 1 was observed most frequently (>95%, 4,776 times out of 5,000 of frames). The other four models appeared in less than 5% of the frames. All models share two hydrophobic features formed with pyridine and benzene in the ligand. Model 1 contains one additional HB donor with the NH moiety of thiazolidinedione. Model 2 only contains the shared hydrophobic features. Model 3 and Model 4 also share the same hydrophobic features plus one HB donor and one HB acceptor. Model 5 comprises the shared hydrophobic features and one HB acceptor. All five pharmacophore models were used to screen DrugBank. The top ranked hitlist for binding site A_ref is shown in Table 2.

TABLE 2

Table 2. Top ranked hit-list for binding site A_ref.

After careful assessment of the main-targets and diseases intended to treat with the compounds and the side effects (listed in Table 3), we selected Travoprost, Betaxolol, and Ritodrine (Docking poses are shown in Supplementary Figure S5) for further testing, since these compounds are not known to cause cardiac side effects or hypoglycemia.

TABLE 3

Table 3. Indication, description, and relevant side effect of the top ranked hit-list compounds at binding site A_ref.

Inhibition of Kir6.2/SUR2A by Travoprost, Betaxolol, and Ritodrine

From earlier work it was established that in the absence of pharmacological activation, Kir6.1/SUR2A channels yield very low current amplitude which hampers efficacy assessment of blockers (Cooper et al., 2014). Therefore, Travoprost, Betaxolol, and Ritodrine [three top ranked hits (Table 2)] were tested for Kir6.2/SUR2A inhibition instead, using the inside/out mode on HEK293T cell derived excised membrane patches. Travoprost (IC₅₀ outward = 2.46 ± 0.52 μM; Hill coefficient 0.71; IC₅₀ inward = 2.30 ± 1.26 μM; Hill coefficient 0.65) dose-dependently inhibited inward and outward current components of I_Kir6.2/SUR2A whereas Betaxolol (IC₅₀ outward = 22.06 ± 2.47 μM; Hill coefficient 0.89) and Ritodrine (IC₅₀ outward = 7.09 ± 0.45 μM; Hill coefficient 0.86) markedly and dose-dependently inhibited the outward component (Figure 5). Betaxolol and Ritodrine induced rectification behavior of the channel, i.e., outward current was more strongly inhibited than inward current (Figure 5B). This induction of rectification is in contrast with inhibition characteristics of Travoprost and RSG. Similar findings have been made with Pentamidine and Pentamidine-Analog-6 (De Boer et al., 2010; Takanari et al., 2013). These structurally related compounds bind to the same site in the Kir2.1 channel, but whereas Pentamidine induces channel rectification (De Boer et al., 2010), Pentamidine-Analog-6 inhibited both inward and outward current with similar efficacy (Takanari et al., 2013). Despite the clear block of Kir6.2 – KATP channels by Travoprost, Betaxolol, and Ritodrine, these drugs do not show KATP channel related clinical effects, however, it has to be kept in mind that the former two drugs are not applied systemically.

FIGURE 5

Figure 5. Inhibition of Kir6.2/SUR2A carried I_KATP by Travoprost, Betaxolol, and Ritodrine. (A) Current traces of Kir6.2/SUR2A channels in the inside-out orientation exposed to Travoprost, Betaxolol, or Ritodrine at the indicated drug concentrations (0, 3 and 30 μM). Dotted horizontal line at 0 pA. (B) Normalized block of inward (black bars, at –80 mV) and outward (open bars, at +50 mV) currents with 30 μM of the indicated drug. *P < 0.001 (paired T-test, inward vs. outward; n = 8, n = 11, and n = 9 for Travoprost, Betaxolol, and Ritodrine, respectively. (C) IC₅₀ curves of outward components of Kir6.2/SUR2A in response to different concentrations of Travoprost (open squares; n = 8), Betaxolol [black triangles; n = 11 or n = 7 (100 μM)] and Ritodrine [black circles; n = 9 (control and 1 μM), n = 8 (0.3 μM), or n = 7 (3, 10, 30 μM)]. Data were fitted with Hill equation to estimate the IC₅₀ values. Data in panels b and c are shown as mean ± SEM. The data in panel B are extracted from the ramp recordings similar to traces as shown in panel A.

Further Support of RSG Binding to Site A_ref

Experimental data from Yu et al. (2012) revealed that RSG binding to inward rectifier potassium channels is Kir6.x specific. Electrophysiology measurements on Kir1.1, Kir2.1, and Kir4.1 channels showed that these channels are insensitive to RSG. Thus, we performed a multiple sequence alignment of these channels and evaluated the conservation of the predicted binding site residues. As shown in Figure 6, differences between Kir6.x and members of other Kir channels at site A_ref can be found at positions 64 (LEU vs. VAL/CYS/PHE), 79 (SER vs. ALA/THR), 172 (VAL/ILE vs. PHE), and 176 (CYS vs. ALA/THR). Of particular interest is position 172, which contains a bulky phenylalanine side chain in non-Kir6.x channels, which would prevent RSG from binding in a similar mode in these channels. Experimental mapping of the binding site has not been performed, possibly due to challenges measuring Kir6.1 subunits (Cooper et al., 2014). Nevertheless, mutational data on the closely related Kir6.2 channel supports that residues in the binding area are critical for normal channel gating (Zhang et al., 2015). Interestingly, two Cantú disease causing gain-of-function mutations (V65M, C176S) (Brownstein et al., 2013; Cooper et al., 2014) are in close proximity to binding site A_ref, further supporting the importance of this region for gating. A further, indirect validation of the binding site was gained by correctly identifying novel hits in DrugBank, based on the in silico predicted structure-based drug-protein interactions.

FIGURE 6

Figure 6. Sequence alignment of Kir6.x, Kir1.1, Kir2.1, and Kir4.1 from binding site A_ref. The alignment of residues on chain A and chain D within 3.5 Å of RSG at the lowest PMF energy pose. Coloring of the alignment was performed using the BLOSUM62 algorithm. SH: slide helix; TM1: transmembrane helix 1; and TM2: transmembrane helix 2.

Suggestions for Experimental Validation of the Binding Site

Our modeling predictions suggest that mutating position VAL172 to PHE should decrease or prevent binding of RSG and its main metabolite N-RSG. Unfortunately, mutating the equivalent position in Kir6.2 (ILE162) to PHE does not produce functional channels (Piao et al., 2001), preventing experimental validation of this prediction. Since the binding site is very close to the PIP₂ molecule, interference of PIP₂ interactions with the channel are likely. Previous studies on other Kir channels, support drug PIP₂ interference for drugs such as carvedilol and ivermectin (Kikuta et al., 2006; Ferrer et al., 2011; Chen et al., 2017) or the anti-cancer agent gamboic acid (Scherer et al., 2017). For a recent review see (Heyden et al., 2013).

Other means of validation could include Rb⁺ flux measurements or transfer of the binding pocket to another inward rectifier channel such as Kir2.1, which has been shown to be insensitive to RSG. A real advancement toward understanding inhibition of Kir6 channels would, however, require obtaining co-crystals of KATP channels bound to RSG or other inhibitors. Further, Kir6.1 containing electrophysiology-based screening assays need to be developed to progress CS research.

Proposed Mechanism of Drug Action

RSG binds at the interface between two subunits, very close to PIP₂, an essential gating modulator of inward rectifier channels. It is conceivable that the drug interferes with normal channel activation, possibly via “blocking” the activation gate and/or via hindering normal lipid modulation of channels. In line with this reasoning, we observed frequent HBs to lipid molecules in our simulations. Further simulations, using different lipid types will be necessary in the future to investigate this possibility. In addition, the predicted binding site A_ref is in close proximity of two previously identified gain-of-function mutations causing Cantú disease [V65M and C176S (Brownstein et al., 2013; Cooper et al., 2014)]. So far 100 ns MD simulations on these two mutations have been performed (Cooper et al., 2017), revealing no changes in this region. Thus, in order to investigate if RSG or the newly identified inhibitors might be able to counterbalance the gating disturbance effects of Cantú mutations, we determined dose-response effects of currents mediated by C166S Kir6.2, which is homologs to C176S in Kir6.1.

Cantú Mutations C166S (Kir6.2) and S1020P (SUR2A) Are Inhibited by RSG, Ritodrine, Travoprost, and Betaxolol

Since CS C176S mutant is in close proximity of the predicted binding site A_ref, we performed inside/out measurements on Kir6.2 C166S, the homologs mutation of Kir6.1 C176S, with RSG and the newly identified drugs Travoprost, Betaxolol, and Ritodrine. As shown in Figure 7 all drugs dose-dependent inhibit outward current, having IC₅₀ similar as WT channels (RSG: WT 25.98 ± 1.49 μM vs. C166S 34.88 ± 2.34 μM n.s.; Ritodrine: WT 7.09 ± 0.45 μM vs. C166S 10.42 ± 0.87 μM n.s.; Betaxolol: WT 22.06 ± 2.47 μM vs. 41.16 ± 2.89 μM n.s.) except for Travoprost (WT 2.46 ± 0.52 μM vs. 14.82 ± 2.16 μM, p < 0.05).

FIGURE 7

Figure 7. Inhibition of C166S (Kir6.2) and S1020P (SUR2A) by Rosiglitazone, Travoprost, Betaxolol, and Ritodrine. IC₅₀ curves of outward components of Kir6.2/SUR2A (WT, C166S Kir6.2, and S1020P SUR2A) in response to different concentrations of Rosiglitazone, Travoprost, Betaxolol, and Ritodrine. N-values are: WT, C166S, and S1020P, respectively: Rosiglitazone n = 8, 8, 6; Travoprost n = 8, 7, 11; Betaxolol n = 11, 7, 7; and Ritodrine n = 9, 8, 7. Data were fitted with Hill equation to estimate the IC₅₀ values. Data are shown as mean ± SEM.

Since the majority of Cantú mutations have been identified in the SUR2A subunits, we also tested the inhibitors on currents mediated by the S1020P SUR2A mutation. Again, mutant channels were sensitive for all for compounds with outward current IC₅₀ values (Rosiglitazone: 25.38 ± 4.24 μM; Ritodrine: 6.77 ± 0.49 μM; Betaxolol: 29.51 ± 3.78 μM) not significantly different from WT channels except for Travoprost (10.99 ± 1.28 μM, p < 0.05 vs. WT).

Interestingly, previous studies on Kir6.2-C166S revealed increased open probability, reduced ATP sensitivity of Kir6.2ΔC and reduced tolbutamide sensitivity when expressed with SUR1 (Trapp et al., 1998). Since the three identified inhibitors are active on mutant channels with increased open probability, it would be interesting to test if the newly identified inhibitors would also display reduced sensitivity toward SUR1, or if the difference originates from the binding to the different subunits (Kir6 vs. SUR subunit). This is of particular interest, since mutation C166Y also displays reduced sensitivity for tolbutamide (Babiker et al., 2016) and DEND (developmental delay and neonatal diabetes) patients with this mutation are glibenclamide unresponsive (Della Manna et al., 2008). Thus, the compounds studied here could be of great interest to these patients, if efficacious on C166Y mutant channels co-expressed with SUR1.

Whereas this proof-of-concept study reveals the general feasibility of inhibiting CS mutants with these drugs, the efficacy of the identified drugs for glibenclamide-insensitive mutant channels remains to be tested in future studies.

Conclusion and Limitations

Understanding the molecular mechanisms of inhibition of Kir6.x channels is critical in paving the way to develop novel blockers, useful for the treatment of channelopathies such as neonatal diabetes (Remedi et al., 2017) or CS (Brownstein et al., 2013; Cooper et al., 2014). Further, our study provides insights into how RSG might exert its cardiovascular side effects, via interfering with its gating mechanism. We performed unbiased MD simulations on the microsecond time scale of the pore forming Kir6.1 model with RSG randomly placed in the solvent. After identification of a putative RSG binding site, we constructed dynamic pharmacophore models using the recently introduced CHA and screened for hits in DrugBank, which contains all drugs available on the market. Functional testing confirmed three new high affinity blockers, with different chemical scaffolds (see Figure 5). The identified compounds (see Table 2) provide an important first starting point for developing novel therapies for rare diseases such as CS.

Although this proof-of-concept study demonstrations the general feasibility of our computational approach to identify novel Kir6.x inhibitors, clinical efficacy and safety remain to be tested in the future. In particular, even though the side-effect profile of approved drugs is in principal well defined, the suitability for repurposing such drugs to pharmacologically correct CS associated gain-of-function mutations awaits further studies. Further, it is currently not foreseeable, which clinical features of CS could be treated/corrected by inhibiting KATP channels.

Another limitation is the inability to measure large enough Kir6.1 currents (in the absence of activators such as pinacidil) to reliably assess drug affinity. This prevents addressing the specificity of Kir6.1 vs. Kir6.2 inhibitors. In our opinion, it will require joint efforts from multiple institutes to obtain favorable cell lines, expression constructs, forward trafficking enhancing factors and patch-clamp protocols, in order to move the field forward.

Nevertheless, this study provides novel insights into the structural basis of Kir6.x channel block, could provide a solid basis for future studies and may have broader implications for the molecular pharmacology of Kir6 channels in general.

Finally, this study provides interesting insights into the polypharmacology of betaxolol, ritodrine and travoprost, since none of these drugs have been linked to KATP channel inhibition before.

Materials and Methods

Homology Modeling

At the beginning of this study, no atomic resolution structure of a KATP channel was available. Thus, a Kir6.1 homology model in the closed state was built using the crystal structure of Kir3.2 (PDB code: 3SYA (Whorton and MacKinnon, 2011), 2.98 Å resolution) as template with the program Modeller9.11 (Martí-Renom et al., 2000). The sequence identity between Kir6.1 and Kir3.2 is 48.36%. The sequence alignments can be found in Supplementary Figure S1A. Comparison of the Kir3.2 template with recent available Kir6.2 structures reveal that the structures are highly similar with RMSD values below 1 Å (comparing the transmembrane domains). The structural alignments, generated with the Swiss-PDB-viewer (Guex and Peitsch, 1997) are shown in Supplementary Figure S1B.

MD Simulations

MD simulations were performed using Gromacs5.1 (Abraham et al., 2015) and the Amber99sb force field (Hornak et al., 2006). The Kir6.1 protein was embedded into the palmitoyloleoylphosphatidylcholine (POPC) lipid bilayer with four PIP₂ molecules bound to the channel, as described previously (Lee et al., 2016). PIP₂ was parameterized using the Hartree-Fock geometry optimization with the 6–31G* basis set (Frisch et al., 2013). POPC parameters were taken from Berger lipids parameters (Berger et al., 1997). The system was solvated using the SPCE water model (Berendsen et al., 1987; Kusalik and Svishchev, 1994) and 150 mM KCl were added to the solvent. To keep the selectivity filter stable, five K⁺ ions were placed at sites S0 to S4. The force field parameters of the ligand were generated and optimized with Gaussian09 (HF/6–31G* basis set) and antechamber (Wang et al., 2004, 2006). Ten R-form and ten S-form ligands were randomly placed in the solvent of the system. The algorithm to integrate Newton’s equation of motion was leap-frog, with a time step of 2 fs. The LINCS algorithm (Hess et al., 1997) was used to constrain all bonds. The cutoff-scheme for neighbor searching used Verlet (Verlet, 1967) within 1 nm and updated the list every 10 fs. The electrostatics and VdW interactions were measured with the particle-mesh Ewald (PME) method (Darden et al., 1993), using a cut-off of 1 nm and Fourier spacing of 0.16 nm. Temperature coupling used the V-rescale method (Bussi et al., 2007) at a reference temperature of 310 K and time constant 0.1 ps. The pressure was kept constant at 1 bar by using the Parrinello-Rahman barostat algorithm (Parrinello and Rahman, 1981) with a coupling constant of 2 ps. The system was minimized with the steepest descent algorithm, followed by a 6 ns equilibration simulation. 1.5 μs unbiased MD simulations were performed to detect the ligand binding sites. Additionally, 250 ns and 200 ns MD simulations were run from the best docking poses of RSG and N-RSG, respectively.

Docking

RSG was docked at the putative binding site identified from unbiased MD simulations, using the program Gold4.0.1 (Jones et al., 1997). The binding sites identified in the 1.5 μs free MD simulations were used as starting point and the radius was set to 20 Å. 100,000 operations of the GOLD genetic algorithm were used to dock the compounds with the ChemPLP scoring function.

Umbrella Sampling (US)

In order to estimate the ligand binding affinity, we performed US at each binding site. Ligands were firstly pulled into the solvent using the pull code in GROMACS by applying a harmonic biasing force between the center of mass (COM) of ligand and the COM of binding site (defined by residues within 5 Å of the ligand). The initial systems were taken from the last frames of the MD simulations for binding sites A_ref, B and C. To ensure that the ligands were pulled along the reaction coordinates fully into the solvent area, a harmonic force of 1,000 kJ/(mol⋅nm²) was applied for most of the pulling simulations. In cases, where the ligand displayed high mobility during the pulling trajectories, the harmonic force was increased to 2,000 kJ/(mol⋅nm²). Starting configurations for US were chosen from the pulling trajectories by taking steps every 0.1 nm along the reaction coordinates. Several intermediate windows were added if the adjacent US windows did not overlap sufficiently. Harmonic forces of 500, 1,000, 2,000, or 3,000 kJ/(mol⋅nm²) were applied to restrict the ligands during US sampling. For each window, a 10 ns simulation was performed, excluding the first 1 ns as equilibration. In total, 242 windows were simulated. Thus, in total, 2.42 μs simulations were performed to obtain good US window overlaps (Supplementary Figure S4). The PMF were calculated by using weighted histogram analysis method (WHAM) (Hub et al., 2010) and the statistical errors were estimated by 100 times bootstrap analysis (Efron, 1979). A more detailed description about the US method can be found in the Supplementary Methods.

Pharmacophore Modeling

The recent published CHA (Wieder et al., 2017) was applied to construct dynamic pharmacophore models and to generate a hit-list by virtual screening in DrugBank (Law et al., 2014). The CHA is implemented by LigandScout 4.10 (Wolber and Langer, 2005).

5,000 snapshots were extracted from the last 100 ns MD simulation of RSG at binding site A_ref and used as input for the CHA. For each snapshot, a pharmacophore model was built by considering the ligand interactions with protein and lipids. Water molecules were discarded during the pharmacophore generation. Pharmacophore features (mainly including hydrophobic interactions, HB donor/acceptor, aromatic ring, ionizable area, etc.) and constrains were defined as described in detail in the LigandScout user manual (LigandScout User Manual, 2010).

Representative pharmacophore models were obtained by merging all identical features, extracted from the 5,000 frames. In the end, five representative pharmacophore models were used for virtual screening against DrugBank4.0 (Law et al., 2014; see Supplementary Figure S3). The molecules in DrugBank were prepared as libraries for virtual screening using the LigandScout command line tool idbgen. Conformers for each molecule in the database were generated using the icon best option in idbgen; this option produces a maximum number of 200 conformations for each molecule processed. The CHA produced a ranked hit-list for the binding site. The approved drugs that fits at least two of the five representative pharmacophore models were proposed in the final hit-list shown in Table 2.

Interaction Map

The interactions of RSG and N-RSG at site A_ref (only protein and PIP₂ were considered) over the 200 ns MD simulations were analyzed and quantified by interaction maps, which were generated by the python package matplotlib (Hunter, 2007) and the chemoinformatics toolkit CDPkit (Seidel and Langer, 2017). Interactions were analyzed by generating a structure-based pharmacophore model at every saved frame of the MD trajectories and subsequently analyzing the frequency of the individual features. The interaction types were defined and described as pharmacophore features in the LigandScout user manual, including hydrophobic (H), hydrogen bond (HB) acceptor/donor, positive ionizable (PI) and aromatic (AR) features. The ligands were decomposed into five areas (Figure 4A and Supplementary Figure S2): the pyridine (Ring A), the benzene (Ring B), the thiazolidinedione (Ring C), the linker connecting pyridine and benzene (Linker D), and the linker connecting benzene and thiazolidinedione (Linker E). The frequencies of interactions observed were numbered and colored in the interaction map.

Electrophysiology

Inside-out patch clamp electrophysiology was performed as described previously (Harakalova et al., 2012). In short, HEK293T cells were cultured on 10 mm glass coverslips and transfected with 0.16 μg of rat pCMV6-Kir6.2, 0.16 μg of rat pCMV6-SUR2A and 0.08 μg of pEGFP1 expression constructs. Measurements were performed using an AxoPatch 200B amplifier controlled by pClamp 9 software (Molecular Devices) at 22°C using a ramp protocol ranging from −100 to +100 mV in 5 s from a holding potential of −40 mV. The sampling rate was 50 kHz, filter frequency was 2 kHz. Bath solution contained 131 mM KCl, 1 mM EGTA, 7.2 mM K₂HPO₄, 2.8 mM KH₂PO₄, 1 mM MgCl₂, and pH 7.20/KOH. The pipette solution contained 145 mM KCl, 1 mM CaCl₂, 1 mM MgCl₂, 5 mM HEPES, and pH 7.40/KOH. Pipette resistance was 1.5–3 MΩ. Data were not corrected for rundown, which was less than 10% at 10 min as shown earlier (Houtman et al., 2019). All measurements were performed within a timeframe of 8–10 min. Fractional block at −80 and +50 mV was determined by dividing current levels obtained with test compound containing solutions by current levels of control traces recorded in the absence of test compound. Statistics (unpaired Students t-tests (two tailed)) were performed using Kaleidagraph 4.0 software (Synergy Software, Reading, PA, United States).

Betaxolol (Sigma-Aldrich, St. Louis MO, United States) and Ritodrine (Sigma-Aldrich) were dissolved in H₂O at 100 mM. Travoprost (MedChemExpress, Monmouth Junction, NJ, United States) was dissolved in DMSO at 10 mM. Test compounds were diluted in bath solution at the indicated concentrations before the start of measurements.

Author Contributions

XC, MW, AG, and MH performed the research. AS-W and MvdH designed the study. XC, MW, E-MZ-P, AG, MH, TL, MvdH, and AS-W analyzed the data. AS-W, MW, MvdH, and XC wrote the manuscript. All authors reviewed the manuscript.

Funding

This work was supported by the Austrian Science Fund (FWF; http://www.fwf.ac.at). AS-W and XC are supported by FWF grant I2101 (E-RARE 2). AS-W, XC, and E-MZ-P are supported by the doctoral program “Molecular drug targets” W1232 (FWF). XC is supported by the dissertation completion fellowship from the University of Vienna. MvdH and MH are supported by the E-Rare 2 Joint Transnational CantuTreat program.

Conflict of Interest Statement

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Acknowledgments

The computational results presented have been achieved in part using the Vienna Scientific Cluster (VSC). This manuscript has been released as a Pre-Print at BioRxiv (Chen et al., 2019).

Supplementary Material

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fphar.2019.00549/full#supplementary-material

References

Abraham, M. J., Murtola, T., Schulz, R., Páll, S., Smith, J. C., Hess, B., et al. (2015). Gromacs: high performance molecular simulations through multi-level parallelism from laptops to supercomputers. SoftwareX 1–2, 19–25. doi: 10.1016/j.softx.2015.06.001

CrossRef Full Text | Google Scholar

Agency European Medicines (2010). European Medicines Agency: EMA/585784/2010 - European Medicines Agency Recommends Suspension of Avandia, Avandamet and Avaglim. Available at: http://www.ema.europa.eu/docs/en_GB/document_library/Press_release/2010/09/WC500096996.pdf (accessed May 16, 2018).

Google Scholar

Ashcroft, F. M. (2005). ATP-sensitive potassium channelopathies: focus on insulin secretion. J. Clin. Invest. 115, 2047–2058. doi: 10.1172/JCI25495

PubMed Abstract | CrossRef Full Text | Google Scholar

Babiker, T., Vedovato, N., Patel, K., Thomas, N., Finn, R., Männikkö, R., et al. (2016). Successful transfer to sulfonylureas in KCNJ11 neonatal diabetes is determined by the mutation and duration of diabetes. Diabetologia 59, 1162–1166. doi: 10.1007/s00125-016-3921-8

PubMed Abstract | CrossRef Full Text | Google Scholar

Berendsen, H. J. C., Grigera, J. R., and Straatsma, T. P. (1987). The missing term in effective pair potentials. J. Phys. Chem. 91, 6269–6271. doi: 10.1021/j100308a038

PubMed Abstract | CrossRef Full Text | Google Scholar

Berger, O., Edholm, O., and Jähnig, F. (1997). Molecular dynamics simulations of a fluid bilayer of dipalmitoylphosphatidylcholine at full hydration, constant pressure, and constant temperature. Biophys. J. 72, 2002–2013. doi: 10.1016/S0006-3495(97)78845-3

PubMed Abstract | CrossRef Full Text | Google Scholar

Borschel, W. F., Wang, S., Lee, S., and Nichols, C. G. (2017). Control of kir channel gating by cytoplasmic domain interface interactions. J. Gen. Physiol. 149, 561–576. doi: 10.1085/jgp.201611719

PubMed Abstract | CrossRef Full Text | Google Scholar

Brown, J. D., and Plutzky, J. (2007). Peroxisome proliferator-activated receptors as transcriptional nodal points and therapeutic targets. Circulation 115, 518–533. doi: 10.1161/CIRCULATIONAHA.104.475673

PubMed Abstract | CrossRef Full Text | Google Scholar

Brownstein, C. A., Towne, M. C., Luquette, L. J., Harris, D. J., Marinakis, N. S., Meinecke, P., et al. (2013). Mutation of KCNJ8 in a patient with Cantú syndrome with unique vascular abnormalities - support for the role of K(ATP) channels in this condition. Eur. J. Med. Genet. 56, 678–682. doi: 10.1016/j.ejmg.2013.09.009

PubMed Abstract | CrossRef Full Text | Google Scholar

Bussi, G., Donadio, D., and Parrinello, M. (2007). Canonical sampling through velocity rescaling. J. Chem. Phys. 126:014101. doi: 10.1063/1.2408420

PubMed Abstract | CrossRef Full Text | Google Scholar

Cantú, J. M., Garcia-Cruz, D., Sánchez-Corona, J., Hernández, A., and Nazará, Z. (1982). A distinct osteochondrodysplasia with hypertrichosis-Individualization of a probable autosomal recessive entity. Hum. Genet. 60, 36–41. doi: 10.1007/BF00281261

CrossRef Full Text | Google Scholar

Chen, I. S., Tateyama, M., Fukata, Y., Uesugi, M., and Kubo, Y. (2017). Ivermectin activates GIRK channels in a PIP2-dependent, Gβγ-independent manner and an amino acid residue at the slide helix governs the activation. J. Physiol. 595, 5895–5912. doi: 10.1113/JP274871

PubMed Abstract | CrossRef Full Text | Google Scholar

Chen, X., Garon, A., Wieder, M., Houtman, M. J. C., Zangerl-Plessl, E. M., Langer, T., et al. (2019). Computational identification of novel Kir6 channel inhibitors. BioRxiv [preprint]. Available at: https://doi.org/10.1101/539460 (accessed May 10, 2019).

Google Scholar

Cooper, P. E., McClenaghan, C., Chen, X., Stary-Weinzinger, A., and Nichols, C. G. (2017). Conserved functional consequences of disease-associated mutations in the slide helix of Kir6.1 and Kir6.2 subunits of the ATP-sensitive potassium channel. J. Biol. Chem. 292, 17387–17398. doi: 10.1074/jbc.M117.804971

PubMed Abstract | CrossRef Full Text | Google Scholar

Cooper, P. E., Reutter, H., Woelfle, J., Engels, H., Grange, D. K., van Haaften, G., et al. (2014). Cantú syndrome resulting from activating mutation in the KCNJ8 gene. Hum. Mutat. 35, 809–813. doi: 10.1002/humu.22555

PubMed Abstract | CrossRef Full Text | Google Scholar

Cox, P. J., Ryan, D. A., Hollis, F. J., Harris, A. M., Miller, A. K., Vousden, M., et al. (2000). Absorption, disposition, and metabolism of rosiglitazone, a potent thiazolidinedione insulin sensitizer, in humans. Drug Metab. Dispos. 28, 772–780.

PubMed Abstract | Google Scholar

Crawford, R. M., Budas, G. R., Jovanović, S., Ranki, H. J., Wilson, T. J., Davies, A. M., et al. (2002). M-LDH serves as a sarcolemmal KATP channel subunit essential for cell protection against ischemia. EMBO J. 21, 3936–3948. doi: 10.1093/emboj/cdf388

PubMed Abstract | CrossRef Full Text | Google Scholar

Darden, T., York, D., and Pedersen, L. (1993). Particle mesh ewald: an N⋅log(N) method for ewald sums in large systems. J. Chem. Phys. 98, 10089–10092. doi: 10.1063/1.464397

CrossRef Full Text | Google Scholar

Dart, C., and Standen, N. B. (1995). Activation of ATP-dependent K+ channels by hypoxia in smooth muscle cells isolated from the pig coronary artery. J. Physiol. 483, 29–39. doi: 10.1113/jphysiol.1995.sp020565

PubMed Abstract | CrossRef Full Text | Google Scholar

De Boer, T., Nalos, L., Stary, A., Kok, B., Houtman, M., Antoons, G., et al. (2010). The anti-protozoal drug pentamidine blocks KIR2.x-mediated inward rectifier current by entering the cytoplasmic pore region of the channel. Br. J. Pharmacol. 159, 1532–1541. doi: 10.1111/j.1476-5381.2010.00658.x

PubMed Abstract | CrossRef Full Text | Google Scholar

Della Manna, T., Battistim, C., Radonsky, V., Savoldelli, R. D., Damiani, D., Kok, F., et al. (2008). Glibenclamide unresponsiveness in a Brazilian child with permanent neonatal diabetes mellitus and DEND syndrome due to a C166Y mutation in KCNJ11 (Kir6.2) gene. Arq. Bras. Endocrinol. Metab. 52, 1350–1355. doi: 10.1590/S0004-27302008000800024

PubMed Abstract | CrossRef Full Text | Google Scholar

Efron, B. (1979). Bootstrap methods: another look at the Jackknife. Ann. Stat. 7, 1–26. doi: 10.1214/aos/1176344552

CrossRef Full Text | Google Scholar

Ferrer, T., Ponce-Balbuena, D., López-Izquierdo, A., Aréchiga-Figueroa, I. A., De Boer, T. P., Van Der Heyden, M. A. G., et al. (2011). Carvedilol inhibits Kir2.3 channels by interference with PIP2-channel interaction. Eur. J. Pharmacol. 668, 72–77. doi: 10.1016/j.ejphar.2011.05.067

PubMed Abstract | CrossRef Full Text | Google Scholar

Frisch, M. J., Trucks, G. W., Schlegel, H. B., Scuseria, G. E., Robb, M. A., Cheeseman, J. R., et al. (2013). Gaussian 09, Revision D.01. Wallingford CT: Gaussian. Inc.

Google Scholar

Gribble, F. M., and Reimann, F. (2003). Sulphonylurea action revisited: the post-cloning era. Diabetologia 46, 875–891. doi: 10.1007/s00125-003-1143-3

PubMed Abstract | CrossRef Full Text | Google Scholar

Guex, N., and Peitsch, M. C. (1997). SWISS-MODEL and the swiss-Pdb viewer: an environment for comparative protein modeling. Electrophoresis 18, 2714–2723. doi: 10.1002/elps.1150181505

PubMed Abstract | CrossRef Full Text | Google Scholar

Hancox, J. C. (2011). Cardiac ion channel modulation by the hypoglycaemic agent rosiglitazone. Br. J. Pharmacol. 163, 496–498. doi: 10.1111/j.1476-5381.2011.01281.x

PubMed Abstract | CrossRef Full Text | Google Scholar

Harakalova, M., Van Harssel, J. J. T., Terhal, P. A., Van Lieshout, S., Duran, K., Renkens, I., et al. (2012). Dominant missense mutations in ABCC9 cause Cantú syndrome. Nat. Genet. 44, 793–796. doi: 10.1038/ng.2324

PubMed Abstract | CrossRef Full Text | Google Scholar

Hess, B., Bekker, H., Berendsen, H. J. C., and Fraaije, J. G. E. M. (1997). LINCS: a linear constraint solver for molecular simulations. J. Comput. Chem. 18, 1463–1472.

Google Scholar

Heyden, M., Stary-Weinzinger, A., and Sanchez-Chapula, J. (2013). Inhibition of cardiac inward rectifier currents by cationic amphiphilic drugs. Curr. Mol. Med. 13, 1284–1298. doi: 10.2174/15665240113139990043

PubMed Abstract | CrossRef Full Text | Google Scholar

Hornak, V., Abel, R., Okur, A., Strockbine, B., Roitberg, A., and Simmerling, C. (2006). Comparison of multiple amber force fields and development of improved protein backbone parameters. Proteins 65, 712–725. doi: 10.1002/prot.21123

PubMed Abstract | CrossRef Full Text | Google Scholar

Houtman, M. J. C., Chen, X., Qile, M., Duran, K., Van Haaften, G., Stary-Weinzinger, A., et al. (2019). Glibenclamide and HMR1098 normalize Cantú syndrome-associated gain-of-function currents. J. Cell. Mol. Med. (in press). doi: 10.1111/jcmm.14329

CrossRef Full Text | Google Scholar

Huang, C. L., Feng, S. Y., and Hilgemann, D. W. (1998). Direct activation of inward rectifier potassium channels by PIP2 and its stabilization by G beta gamma. Nature 391, 803–806. doi: 10.1038/35882

PubMed Abstract | CrossRef Full Text | Google Scholar

Hub, J. S., De Groot, B. L., and Van Der Spoel, D. (2010). g_whams-a free weighted histogram analysis implementation including robust error and autocorrelation estimates. J. Chem. Theory Comput. 6, 3713–3720. doi: 10.1021/ct100494z

CrossRef Full Text | Google Scholar

Hunter, J. D. (2007). Matplotlib: a 2D graphics environment. Comput. Sci. Eng. 9, 99–104. doi: 10.1109/MCSE.2007.55

CrossRef Full Text | Google Scholar

Jeong, I., Choi, B., and Hahn, S. (2011). Rosiglitazone inhibits Kv4.3 potassium channels by open-channel block and acceleration of closed-state inactivation. Br. J. Pharmacol. 163, 510–520. doi: 10.1111/j.1476-5381.2011.01210.x

PubMed Abstract | CrossRef Full Text | Google Scholar

Jones, G., Willett, P., Glen, R. C., Leach, A. R., and Taylor, R. (1997). Development and validation of a genetic algorithm for flexible docking. J. Mol. Biol. 267, 727–748. doi: 10.1006/jmbi.1996.0897

PubMed Abstract | CrossRef Full Text | Google Scholar

Kharade, S. V., Nichols, C., and Denton, J. S. (2016). The shifting landscape of KATP channelopathies and the need for “sharper” therapeutics. Future Med. Chem. 8, 789–802. doi: 10.4155/fmc-2016-0005

PubMed Abstract | CrossRef Full Text | Google Scholar

Kikuta, J. I., Ishii, M., Kishimoto, K., and Kurachi, Y. (2006). Carvedilol blocks cardiac KATP and KG but not IK1 channels by acting at the bundle-crossing regions. Eur. J. Pharmacol. 529, 47–54. doi: 10.1016/j.ejphar.2005.10.059

PubMed Abstract | CrossRef Full Text | Google Scholar

Kuhn, M., Campillos, M., Letunic, I., Jensen, L. J., and Bork, P. (2010). A side effect resource to capture phenotypic effects of drugs. Mol. Syst. Biol. 6:343. doi: 10.1038/msb.2009.98

PubMed Abstract | CrossRef Full Text | Google Scholar

Kuhn, M., Letunic, I., Jensen, L. J., and Bork, P. (2016). The SIDER database of drugs and side effects. Nucleic Acids Res. 44, D1075–D1079. doi: 10.1093/nar/gkv1075

PubMed Abstract | CrossRef Full Text | Google Scholar

Kusalik, P. G., and Svishchev, I. M. (1994). The spatial structure in liquid water. Science 265, 1219–1221. doi: 10.1126/science.265.5176.1219

PubMed Abstract | CrossRef Full Text | Google Scholar

Law, V., Knox, C., Djoumbou, Y., Jewison, T., Guo, A. C., Liu, Y., et al. (2014). DrugBank 4.0: shedding new light on drug metabolism. Nucleic Acids Res. 42, D1091–D1097. doi: 10.1093/nar/gkt1068

PubMed Abstract | CrossRef Full Text | Google Scholar

Lee, K. P. K., Chen, J., and MacKinnon, R. (2017). Molecular structure of human KATP in complex with ATP and ADP. eLife 6:e32481. doi: 10.7554/eLife.32481

PubMed Abstract | CrossRef Full Text | Google Scholar

Lee, S.-J., Ren, F., Zangerl-Plessl, E.-M., Heyman, S., Stary-Weinzinger, A., Yuan, P., et al. (2016). Structural basis of control of inward rectifier Kir2 channel gating by bulk anionic phospholipids. J. Gen. Physiol. 148, 227–237. doi: 10.1085/jgp.201611616

PubMed Abstract | CrossRef Full Text | Google Scholar

Li, N., Wu, J. X., Ding, D., Cheng, J., Gao, N., and Chen, L. (2017). Structure of a pancreatic ATP-sensitive potassium channel. Cell 168, 101–110.e10. doi: 10.1016/j.cell.2016.12.028

PubMed Abstract | CrossRef Full Text | Google Scholar

LigandScout User Manual (2010). Available at: http://www.inteligand.com/ligandscout3/downloads/ligandscout-manual-2010-04-15.pdf (accessed May 16, 2018).

Google Scholar

Lin, Y.-W., Jia, T., Weinsoft, A. M., and Shyng, S.-L. (2003). Stabilization of the activity of ATP-sensitive potassium channels by ion pairs formed between adjacent Kir6.2 subunits. J. Gen. Physiol. 122, 225–237. doi: 10.1085/jgp.200308822

PubMed Abstract | CrossRef Full Text | Google Scholar

Lu, L., Reiter, M. J., Xu, Y., Chicco, A., Greyson, C. R., and Schwartz, G. G. (2008). Thiazolidinedione drugs block cardiac KATP channels and may increase propensity for ischaemic ventricular fibrillation in pigs. Diabetologia 51, 675–685. doi: 10.1007/s00125-008-0924-0

PubMed Abstract | CrossRef Full Text | Google Scholar

MacGregor, G. G., Dong, K., Vanoye, C. G., Tang, L., Giebisch, G., and Hebert, S. C. (2002). Nucleotides and phospholipids compete for binding to the C terminus of KATP channels. Proc. Natl. Acad. Sci. U.S.A. 99, 2726–2731. doi: 10.1073/pnas.042688899

PubMed Abstract | CrossRef Full Text | Google Scholar

Martin, G. M., Kandasamy, B., DiMaio, F., Yoshioka, C., and Shyng, S.-L. (2017a). Anti-diabetic drug binding site in a mammalian KATP channel revealed by Cryo-EM. eLife 6:e31054. doi: 10.7554/eLife.31054

PubMed Abstract | CrossRef Full Text | Google Scholar

Martin, G. M., Yoshioka, C., Rex, E. A., Fay, J. F., Xie, Q., Whorton, M. R., et al. (2017b). Cryo-EM structure of the ATP-sensitive potassium channel illuminates mechanisms of assembly and gating. eLife 6:e24149. doi: 10.7554/eLife.24149

PubMed Abstract | CrossRef Full Text | Google Scholar

Martí-Renom, M. A., Stuart, A. C., Fiser, A., Sánchez, R., Melo, F., and Šali, A. (2000). Comparative protein structure modeling of genes and genomes. Annu. Rev. Biophys. Biomol. Struct. 29, 291–325. doi: 10.1146/annurev.biophys.29.1.291

PubMed Abstract | CrossRef Full Text | Google Scholar

Matsuo, M., Kioka, N., Amachi, T., and Ueda, K. (1999). ATP binding properties of the nucleotide-binding folds of SUR1. J. Biol. Chem. 274, 37479–37482. doi: 10.1074/jbc.274.52.37479

PubMed Abstract | CrossRef Full Text | Google Scholar

Matsuo, M., Tanabe, K., Kioka, N., Amachi, T., and Ueda, K. (2000). Different binding properties and affinities for ATP and ADP among sulfonylurea receptor subtypes, SUR1, SUR2A, and SUR2B. J. Biol. Chem. 275, 28757–28763. doi: 10.1074/jbc.M004818200

PubMed Abstract | CrossRef Full Text | Google Scholar

Nelson, P. T., Wang, W. X., Wilfred, B. R., Wei, A., Dimayuga, J., Huang, Q., et al. (2015). Novel human ABCC9/SUR2 brain-expressed transcripts and an eQTL relevant to hippocampal sclerosis of aging. J. Neurochem. 134, 1026–1039. doi: 10.1111/jnc.13202

PubMed Abstract | CrossRef Full Text | Google Scholar

Nichols, C. G., and Lederer, W. J. (1991). Adenosine triphosphate-sensitive potassium channels in the cardiovascular system. Am. J. Physiol. Hear. Circ. Physiol. 261, 1675–1686. doi: 10.1152/ajpheart.1991.261.6.H1675

PubMed Abstract | CrossRef Full Text | Google Scholar

Nichols, C. G., Singh, G. K., and Grange, D. K. (2013). KATP channels and cardiovascular disease: suddenly a syndrome. Circ. Res. 112, 1059–1072. doi: 10.1161/CIRCRESAHA.112.300514

PubMed Abstract | CrossRef Full Text | Google Scholar

Parrinello, M., and Rahman, A. (1981). Polymorphic transitions in single crystals: a new molecular dynamics method. J. Appl. Phys. 52, 7182–7190. doi: 10.1063/1.328693

CrossRef Full Text | Google Scholar

Pearson, E. R., Flechtner, I., Njølstad, P. R., Malecki, M. T., Flanagan, S. E., Larkin, B., et al. (2006). Switching from insulin to oral sulfonylureas in patients with diabetes due to Kir6.2 mutations. N. Engl. J. Med. 355, 467–477. doi: 10.1056/NEJMoa061759

PubMed Abstract | CrossRef Full Text | Google Scholar

Piao, H., Cui, N., Xu, H., Mao, J., Rojas, A., Wang, R., et al. (2001). Requirement of multiple protein domains and residues for gating K(ATP) channels by intracellular pH. J. Biol. Chem. 276, 36673–36680. doi: 10.1074/jbc.M106123200

PubMed Abstract | CrossRef Full Text | Google Scholar

Remedi, M. S., Friedman, J. B., and Nichols, C. G. (2017). Diabetes induced by gain-of-function mutations in the Kir6.1 subunit of the KATP channel. J. Gen. Physiol. 149, 75–84. doi: 10.1085/jgp.201611653

PubMed Abstract | CrossRef Full Text | Google Scholar

Remedi, M. S., and Nichols, C. G. (2009). Hyperinsulinism and diabetes: genetic dissection of β cell metabolism-excitation coupling in mice. Cell Metab. 10, 442–453. doi: 10.1016/j.cmet.2009.10.011

PubMed Abstract | CrossRef Full Text | Google Scholar

Scherer, D., Schworm, B., Seyler, C., Xynogalos, P., Scholz, E. P., Thomas, D., et al. (2017). Inhibition of inwardly rectifying Kir2.x channels by the novel anti-cancer agent gambogic acid depends on both pore block and PIP2 interference. Naunyn. Schmiedebergs. Arch. Pharmacol. 390, 701–710. doi: 10.1007/s00210-017-1372-5

PubMed Abstract | CrossRef Full Text | Google Scholar

Scurr, I., Wilson, L., Lees, M., Robertson, S., Kirk, E., Turner, A., et al. (2011). Cantú syndrome: report of nine new cases and expansion of the clinical phenotype. Am. J. Med. Genet. Part A 155, 508–518. doi: 10.1002/ajmg.a.33885

PubMed Abstract | CrossRef Full Text | Google Scholar

Seidel, T., and Langer, T. (2017). CDPkit, the Chemical Data Processing Toolkit. Available at: https://github.com/aglanger/CDPKit (accessed May 16, 2018).

Google Scholar

Shyng, S. L., and Nichols, C. G. (1998). Membrane phospholipid control of nucleotide sensitivity of K(ATP) channels. Science 282, 1138–1141. doi: 10.1126/science.282.5391.1138

CrossRef Full Text | Google Scholar

Szentandrássy, N., Harmati, G., Bárándi, L., Simkõ, J., Horváth, B., Magyar, J., et al. (2011). Effects of rosiglitazone on the configuration of action potentials and ion currents in canine ventricular cells. Br. J. Pharmacol. 163, 499–509. doi: 10.1111/j.1476-5381.2011.01215.x

PubMed Abstract | CrossRef Full Text | Google Scholar

Takanari, H., Nalos, L., Stary-Weinzinger, A., De Git, K. C. G., Varkevisser, R., Linder, T., et al. (2013). Efficient and specific cardiac IK1 inhibition by a new pentamidine analogue. Cardiovasc. Res. 99, 203–214. doi: 10.1093/cvr/cvt103

PubMed Abstract | CrossRef Full Text | Google Scholar

Tanabe, K., Tucker, S. J., Matsuo, M., Proks, P., Ashcroft, F. M., Seino, S., et al. (1999). Direct photoaffinity labeling of the Kir6.2 subunit of the ATP-sensitive K+ channel by 8-azido-ATP. J. Biol. Chem. 274, 3931–3933. doi: 10.1074/jbc.274.7.3931

PubMed Abstract | CrossRef Full Text | Google Scholar

Trapp, S., Proks, P., Tucker, S. J., and Ashcroft, F. M. (1998). Molecular analysis of ATP-sensitive K channel gating and implications for channel inhibition by ATP. J. Gen. Physiol. 112:333. doi: 10.1085/jgp.112.3.333

PubMed Abstract | CrossRef Full Text | Google Scholar

U.S. Food and Drug Administration (2011). Reminder to Healthcare Providers and Patients to Enroll in the Avandia-Rosiglitazone Medicines Access Program. FDA Drug Saf. Commun. Available at: https://www.fda.gov/Drugs/DrugSafety/ucm277629.htm (accessed May 16, 2018).

Google Scholar

Ueda, K., Komine, J., Matsuo, M., Seino, S., and Amachi, T. (1999). Cooperative binding of ATP and MgADP in the sulfonylurea receptor is modulated by Figure. Proc. Natl. Acad. Sci. U.S.A. 96, 1268–1272. doi: 10.1073/pnas.96.4.1268

PubMed Abstract | CrossRef Full Text | Google Scholar

Van Bon, B. W. M., Gilissen, C., Grange, D. K., Hennekam, R. C. M., Kayserili, H., Engels, H., et al. (2012). Cantú syndrome is caused by mutations in ABCC9. Am. J. Hum. Genet. 90, 1094–1101. doi: 10.1016/j.ajhg.2012.04.014

PubMed Abstract | CrossRef Full Text | Google Scholar

Vanoye, C. G., MacGregor, G. G., Dong, K., Tang, L., Buschmann, A. S., Hall, A. E., et al. (2002). The carboxyl termini of KATP channels bind nucleotides. J. Biol. Chem. 277, 23260–23270. doi: 10.1074/jbc.M112004200

PubMed Abstract | CrossRef Full Text | Google Scholar

Verlet, L. (1967). Computer “experiments” on classical fluids. I. Thermodynamical properties of lennard-jones molecules. Phys. Rev. 159, 98–103. doi: 10.1103/PhysRev.159.98

CrossRef Full Text | Google Scholar

Wang, J., Wang, W., Kollman, P. A., and Case, D. A. (2006). Automatic atom type and bond type perception in molecular mechanical calculations. J. Mol. Graph. Model. 25, 247–260. doi: 10.1016/j.jmgm.2005.12.005

PubMed Abstract | CrossRef Full Text | Google Scholar

Wang, J., Wolf, R. M., Caldwell, J. W., Kollman, P. A., and Case, D. A. (2004). Development and testing of a general Amber force field. J. Comput. Chem. 25, 1157–1174. doi: 10.1002/jcc.20035

PubMed Abstract | CrossRef Full Text | Google Scholar

Whorton, M. R., and MacKinnon, R. (2011). Crystal structure of the mammalian GIRK2 K+ channel and gating regulation by G proteins, PIP2, and sodium. Cell 147, 199–208. doi: 10.1016/j.cell.2011.07.046

PubMed Abstract | CrossRef Full Text | Google Scholar

Wieder, M., Garon, A., Perricone, U., Boresch, S., Seidel, T., Almerico, A. M., et al. (2017). Common hits approach: combining pharmacophore modeling and molecular dynamics simulations. J. Chem. Inf. Model. 57, 365–385. doi: 10.1021/acs.jcim.6b00674

PubMed Abstract | CrossRef Full Text | Google Scholar

Wolber, G., and Langer, T. (2005). LigandScout: 3-D pharmacophores derived from protein-bound ligands and their use as virtual screening filters. J. Chem. Inf. Model. 45, 160–169. doi: 10.1021/ci049885e

PubMed Abstract | CrossRef Full Text | Google Scholar

Wu, J. X., Ding, D., Wang, M., Kang, Y., Zeng, X., and Chen, L. (2018). Ligand binding and conformational changes of SUR1 subunit in pancreatic ATP-sensitive potassium channels. Protein Cell 9, 553–567. doi: 10.1007/s13238-018-0530-y

PubMed Abstract | CrossRef Full Text | Google Scholar

Yu, L., Jin, X., Cui, N., Wu, Y., Shi, Z., Zhu, D., et al. (2012). Rosiglitazone selectively inhibits KATP channels by acting on the KIR6 subunit. Br. J. Pharmacol. 167, 26–36. doi: 10.1111/j.1476-5381.2012.01934.x

PubMed Abstract | CrossRef Full Text | Google Scholar

Yu, L., Jin, X., Yang, Y., Cui, N., and Jiang, C. (2011). Rosiglitazone inhibits vascular KATP channels and coronary vasodilation produced by isoprenaline. Br. J. Pharmacol. 164, 2064–2072. doi: 10.1111/j.1476-5381.2011.01539.x

PubMed Abstract | CrossRef Full Text | Google Scholar

Zhang, H., He, C., Yan, X., Mirshahi, T., and Logothetis, D. E. (1999). Activation of inwardly rectifying K+ channels by distinct PtdIns(4,5)P2 interactions. Nat. Cell Biol. 1, 183–188. doi: 10.1038/11103

PubMed Abstract | CrossRef Full Text | Google Scholar

Zhang, R. S., Wright, J. D., Pless, S. A., Nunez, J. J., Kim, R. Y., Li, J. B. W., et al. (2015). A conserved residue cluster that governs kinetics of ATP-dependent gating of Kir6.2 potassium channels. J. Biol. Chem. 290, 15450–15461. doi: 10.1074/jbc.M114.631960

PubMed Abstract | CrossRef Full Text | Google Scholar

Zingman, L. V., Alekseev, A. E., Hodgson-Zingman, D. M., and Terzic, A. (2007). ATP-sensitive potassium channels: metabolic sensing and cardioprotection. J. Appl. Physiol. 103, 1888–1893. doi: 10.1152/japplphysiol.00747.2007

PubMed Abstract | CrossRef Full Text | Google Scholar