K_ATP channel subunit constructs used are Kir6.2-myc, Flag-SUR1, Flag-SUR2A, and Kir6.2Δ36. Kir6.2-F55L, Kir6.2-T62M, Kir6.2-R301C, SUR1-V715G, SUR1-S1386P, and SUR1-L1390P were provided by Dr. Show-Ling Shyng (Oregon Health and Science University, Portland, United States).

HEK-293 cells were cultured in DMEM with 10% fetal bovine serum; 100 U/mL penicillin; 100 μg/mL streptomycin at 37°C in a humidified 5% CO₂ atmosphere. Kir6.2/SUR2A and Kir6.1/SUR2B stable HEK-293 cell lines were from Dr. A. Tinker (University of London, United Kingdom). For transient transfections, cells were plated in a 60 mm dish and cultured for 24 h. The next day cells were co-transfected with 0.2 µg of Kir6.2-myc, Kir6.2 Δ36, or mutant Kir6.2 with 1.8 µg of Flag-SUR1 or Flag-SUR2A utilizing Lipofectamine 2000 (Thermo Fisher Scientific). Twenty-4 h after transfection, cells were re-plated onto glass coverslips or 96-well plates for experimentation.

The cells were plated at a density of 3.0–4.0 × 10⁵ cells/well in black, flat-bottom, 96-well plates and cultured overnight until confluent. The cells were then incubated for 1 h with FluxOR II reagent dye in loading buffer, followed by washing and 10 min of exposure to 80 µL of assay buffer containing DMSO (vehicle control), pinacidil (K_ATP channel activation control), the K_ATP channel inhibitor glibenclamide or test compounds. All compounds were dissolved in DMSO. Fluorescence was recorded at excitation/emission wavelengths of 495 nm/510 nm using a microplate reader equipped with a dispenser unit for adding compounds during the experiment (FlexStation 3, Molecular Devices). Fluorescence intensities were recorded every 2 s for the duration of the experiment (∼2 min). An averaged baseline fluorescence intensity value was calculated from at least 8 repeated measurements, after which a stimulus buffer containing Tl₂SO₄ (20 µL) was added to each well. The fluorescence intensity of each well was normalized with an average baseline fluorescence intensity value from 8 repeated measurements prior to stimulus buffer injection. Given that the Tl⁺ flux assay is cumulative, the maximum fluorescence attained during a prolonged experiment have low sensitivity and reproducibility. By contrast, the initial rate is a more sensitive index of activity, as also used by others (Carmosino et al., 2013). Moreover, the initial rate of Tl⁺ flux is more comparable to data obtained with patch clamping (Du et al., 2015; Philippaert et al., 2018). For this reason, we used the initial rate of Tl⁺ flux as an index of channel activity.

Cells, plated on coverslips, were subjected to patch clamping using a Multiclamp 700 A amplifier (Molecular Devices, San Jose, CA, United States). Borosilicate glass electrodes were pulled with a DMZ Universal Puller (Zeitz, Germany). For inside-out K_ATP current recordings, pipettes (3–5 MΩ resistance) were filled with a pipette solution containing (in mM): 110 K-gluconate, 30 KCl, 2 CaCl₂, 1 MgCl₂, 10 HEPES, at pH 7.4. The bath solution contained (in mM): 110 K-gluconate, 30 KCl, 1 EGTA, 1 MgCl₂, 10 HEPES, at pH 7.2. Following patch excision, the pipette potential was held at +80 mV, which corresponds to a membrane potential of −80 mV. Currents were recorded immediately after patch excision. Recordings with significant rundown were discarded. Compounds were applied using a rapid solution changer (RSC160, BioLogic SAS, Seyssinet-Pariset, France). For whole-cell K_ATP current recordings, the pipettes (2∼3 MΩ) were filled with (in mM): 110 K-gluconate, 10 KCl, 5 NaCl, 1 MgCl₂, 5 EGTA, 10 HEPES, at pH 7.2. The bath solution contained (in mM): 110 K-gluconate, 10 KCl, 2 CaCl₂, 1 MgCl₂, 10 HEPES, and 10 glucose at pH 7.4. Whole-cell current–voltage relationships were obtained using a 2 s ramp protocol from 0 mV to −80 mV and back to 0 mV. Compound application was performed using the rapid solution changer. Data were filtered at 1 kHz (8-pole Bessel response), digitized (Digidata 1550A; Molecular Devices), acquired at 10 kHz and analyzed using pClamp 10 (Molecular Devices).

For Kir2.1 current recordings, the pipettes (2∼3 MΩ) were filled with (in mM): 110 mM KCl, 5 NaCl, 1 MgCl₂, 5 EGTA, 10 HEPES at pH 7.2. The bath solution contained (in mM): 140 NaCl, 4 KCl, 1 MgCl₂, 10 HEPES, 2 CaCl₂ at pH 7.4. The Kir2.1 current was recorded from a holding potential of 0 mV using 200 ms voltage steps −140 to +40 mV in 10 mV increments. Currents were normalized by cell size and reported as current density (pA/pF).

Voltage-gated sodium currents were recorded using a QPatch II (Sophion Bioscience, Denmark) with 48 channel QPlates, with a one patch hole per channel. The QPlates had a hole diameter of ∼1 μm and a resistance of ∼2 MΩ. The QPlates were primed with an internal solution containing (in mM): 110 CsF, 10 CsCl, 10 NaCl, 10 EGTA, 10 HEPES at pH 7.2 (osmolarity ∼286 mOsm). TE671 cells were washed and re-suspended with extracellular solution containing (in mM): 140 NaCl 4 KCl, 2 CaCl₂, 1 MgCl₂, 5 glucose, 10 HEPES at pH 7.4 (osmolarity ∼298 mOsm) to ensure complete removal of media before being loaded into the QPatch II. Whole-cell configuration was achieved by the QPatch II utilizing a GΩ seal formation and rupture protocol optimized for TE671 cells (the protocol is available upon request). After establishing a successful whole-cell configuration, cells were held at −120 mV and a 3 ms voltage step to 0 mV was repetitively applied at a rate of 0.03 Hz for 2 min to assess the stability of the recordings. Current-voltage relationships were measured from a holding potential of −120 mV, the membrane was depolarized to voltages between −80 mV and 60 mV for 300 ms. Series resistances were compensated by ∼85%. Currents were low-pass filtered at 10 kHz with an 8-pole Bessel filter response and recorded at a sampling rate of 25 kHz. Currents were normalized to cell size and reported as current density (pA/pF).

Oocytes were isolated from female Xenopus laevis as previously described (Chen et al., 2022). Frogs were purchased from Watanabe Zoushoku (Hyogo, Japan) and housed at 18°C in the animal facility of Wakayama Medical University. After anesthetization of the frog with 0.15% tricaine (Tokyo Chemical Industry, Tokyo, Japan) through skin absorption for 30 min, oocytes were dissociated on ice after making a small abdominal incision. Follicle cells were removed with 2 mg mL⁻¹ collagenase type I (Sigma-Aldrich, St. Louis, MO, Unites States) for 6 h in a solution containing (in mM) 88 NaCl, 1 KCl, 2.4 NaHCO₃, 0.3 Ca(NO₃)₂, 0.41 CaCl₂, 0.82 MgSO₄ and 15 HEPES at pH 7.6. The cRNA of mouse Kir3.1 and rat Kir3.4 were transcribed with the mMessage mMachine T3 Transcription Kit (Invitrogen, Carlsbad, CA, Unites States) from their linearized cDNA in pBluescript II and then were injected into each oocyte (25 ng each for Kir3.1 and Kir3.4). After 3–5 days of the injection, oocytes were used for two-electrode voltage-clamp recordings.

GIRK currents were recorded from oocytes at room temperature using pClamp 10 (Molecular Devices), a two-electrode voltage clamp amplifier OC-725C (Warner Instruments, Holliston, MA, Unites States) and an analog-to-digital converter Digidata 1550 B (Molecular Devices). Recoding electrodes (0.2–0.5 MΩ) were filled with a pipette solution containing 3 MK-acetate and 10 mM KCl. CL-705G was applied to an extracellular high K⁺ solution that contained (in mM): 96 KCl, 3 MgCl₂, 10 HEPES, at pH 7.3 adjusted with n-methyl-d-glucamine. The background leakage current was determined by replacing the extracellular K⁺ with the impermeable Na⁺ and data with significant leak were excluded from the analysis.

Mouse ventricular myocytes were isolated using a Langendorff-free method (Ackers-Johnson et al., 2016). Following calcium reintroduction, cardiomyocytes were resuspended in culture medium (M199, 0.1% BSA, CD Lipid, Insulin-Transferrin-Selenium) and plated in 96-well plates coated with 1% Geltrex in M199 media. Cells were allowed to attach in an O₂/CO₂ incubator containing a humidified atmosphere of 5% CO₂ at 37°C for 2 h. Cells were then treated with phenylephrine (5 µM) in the presence or absence of glibenclamide (1 µM), or with CL-705G (100 µM) for 24 h. Hypoxia was induced by placing cells in an oxygen controlled chamber (Biospherix, Lacona, NY) containing 1% O₂, 5% CO, and 94% N₂ at 37°C for 2 h. Cells were reoxygenated by changing media and placing them in normoxic incubator for 18 h. At the end of the sequence of hypoxia/reoxygenation, cardiomyocyte survival was measured by determining total cellular ATP content (CellTiter Glo 2.0 Cell Viability assay, Promega).

All molecular modeling, alignments and computational simulations were performed with the Internal Coordinates Mechanics software, (ICM-Pro, Molsoft LLC, La Jolla CA). For the open conformation K_ATP channel, two 3D models were evaluated, the 3.7 Å, near-atomic resolution cryo-EM structure of Kir6.2 in an open conformation (PDB 7S5X) and, for atomic resolution, a homology model of Kir6.2 in the conformation observed in the PIP₂-bound, presumably open conformation, of Kir3.2 (PDB: 3SYQ; 3.44 Å). To generate the latter, zero-end gap global sequence alignment between the human Kir6.2 sequence (Uniprot: Q14654) and the Kir3.2 sequence produced a p-value of 10⁻³² for 3D structural similarity. In prior work, precise homology modeling produces 3D structural models of equivalent accuracy to the experimental template structure (3.44 Å) when the p-value is so low (Abagyan and Batalov, 1997). A homology model of the human Kir6.2 channel was built as previously described (Cardozo et al., 1995), by using 3SYQ as a structural template. For the closed conformation of K_ATP channel, we used PDB: 6C3P, which was resolved by cryo-EM at 5.6 Å, as the starting point for a 3D model. For both 7S5X and 6C3P, missing side-chains were built into the model, followed by Biased Probability Monte Carlo energy minimization in order to optimize the energy of the structure and correct any modeling errors from the density (Abagyan and Totrov, 1994). For point mutations, individual models exhibiting palmitoylation or mutations of cysteine-166 (Cys¹⁶⁶) were generated using the “modify” command in ICM-Pro, which replaces, in the open conformation homology model or the closed conformation refined model, a specified side chain atom group and bonding pattern (in this case Cys¹⁶⁶) with the side chain atoms and bonding pattern of a different specified amino acid. The newly introduced side-chain, neighboring sidechains and the local backbone are then all subjected to Monte Carlo energy minimization as a single ensemble. Docking was performed with three orthogonal programs: ICM-Pro, AutoDock (Morris et al., 2008) and CB-Dock2 (Liu et al., 2022). Virtual library screening was performed with ICM-Pro: briefly, grid potentials for van der Waals and electrostatics were generated from the Kir6.2 model coordinates for the region of the pocket in the structures and full-atom models of the compounds were docked into these grids. The Chembridge Core chemical library (Chembridge Corp. La Jolla CA) was screened as mol format files for each compound, and the CL-705G compound structure is encoded as a mol formatted data structure within this library. Compounds were considered hits with an ICM energy score of < −32. The compounds in Table 1 are all the compounds achieving this docking score without common substructures.

When comparing two groups, we used the Student’s t-test. A one-way or two-way ANOVA was used for comparison of multiple groups, followed by the Dunnett’s t-test for comparisons to a single control. A value of p < 0.05 was considered significant. The statistical model underlying the homology modeling alignments has previously been described (Abagyan and Batalov, 1997).

All relevant material is contained within the article or the Supplementary Material.

There are now several cryo-EM structures available for Kir6.2/SUR1 and Kir6.1/SUR2B channels. Currently, most are in the closed state (e.g., 6C3P) (Lee et al., 2017), but an atomic-resolution X-ray crystallographic structure of mouse PIP₂-bound Kir3.2 (PDB: 3SYQ, 3.44 Å) (Whorton and MacKinnon, 2011), a few high-resolution X-ray structures of fowl Kir2.2 in open and closed conformations (e.g., PDBs: 6M84, 2.81 Å; 5KUK 2 Å) (Lee et al., 2016; Zangerl-Plessl et al., 2020), and a recent near-atomic-resolution ensemble of cryo-EM structures of open Kir6.2/SUR1 (e.g., PDB 7S5V, 7S5X) are available. We used our previously published methods (Martinez-Ortiz and Cardozo, 2018) to refine these existing structures and to generate maximally realistic 3D models of the Kir6.2/SUR homotetramers in both the open and closed states (Yang et al., 2020). We previously found that S-palmitoylation of a specific Cys-166 residue in the M2 region of Kir6.2 promotes K_ATP channel opening by allosterically connecting M2 to the PIP₂ binding site (Yang et al., 2020). The cavity occupied by the palmitoyl moiety and PIP₂ in this model represents a potentially druggable pocket influencing channel opening (Figure 1).

Given the novelty of the identification of a potential drug-binding pocket that could mediate channel opening, with all its implications for K_ATP channel pharmacology and therapeutic possibilities, we conducted a virtual library screen (Cardozo and Abagyan, 2005) of the Chembridge Core chemical library of 492,000 drug-like compounds for their biophysical fit to this pocket, as it appears in each of the 3D models of Kir6.2 we obtained, both with and without PIP₂ present at the location it appears in the PDB 3SYQ structure. The 15 diverse (without common substructures) compounds achieving the threshold docking scores, indicating excellent fit to the shape and electrostatics of the pocket, were selected for in vitro testing (Table 1).

We initially employed a thallium-sensitive fluorescence assay to assess K_ATP channel activity. Experiments were performed with HEK293 cells transiently transfected with Kir6. x/SURx cDNAs, or with HEK293 cell stably expressing Kir6.2/SUR2A or Kir6.1/SUR2B. Cells, grown in 96-well plates, were loaded with a Tl⁺-sensitive dye and Tl⁺ flux was initiated by adding Tl⁺ to the extracellular solution. Addition of Tl⁺ led to a large increase of fluorescence in cells expressing Kir6.2/SUR2A when compared to untransfected cells (Figure 2A). The Tl⁺ flux rate was strongly increased by pinacidil, and inhibited by glibenclamide, demonstrating that the Tl⁺ flux occurred through K_ATP channels. We utilized the initial rate of Tl⁺ flux (Figure 2B) as an index of K_ATP channel activity. We next tested the effects of 15 of the top ranking compounds identified as potential Kir6.2 binders by virtual library screening (Table 1). Pinacidil and glibenclamide were included as positive controls, respectively to open and block the K_ATP channel. The majority of the compounds had little effect on the Tl⁺ influx rate. By contrast, compound CL-705G strongly activated Tl⁺ flux (Figure 2C).

To confirm that CL-705G activates K_ATP channels, we performed whole-cell voltage clamping of HEK293 cells expressing Kir6.2/SUR2A. Ramp voltage pulses were applied between 0 and −80 mV. Application of 100 µM CL-705G led to a rapid and reversible increase in K_ATP currents (Figures 2D,E). Moreover, the current was blocked by glibenclamide (10 µM), demonstrating the veracity of the recordings. The K_ATP channel activation by 100 µM CL-705G was similar to that induced by pinacidil (Figures 2F,G). These data demonstrate that CL-705G activates Kir6.2/SUR2A K_ATP channels.

We identified nine analogs of compound CL-705G in the Chembridge Core chemical library with 2D structural similarities ranging from 86%–93%. These compounds were docked to our Kir6.2 model as previously described. Of the nine analogs, five docked to the same pocket as CL-705G (Table 2). However, none of the five compounds (100 µM each) resulted in an increased Tl⁺ flux rate comparable to CL-705G or pinacidil (Figure 3). The remainder of the studies therefore focused on CL-705G.

Using the Tl⁺ flux assay we next determined dose-response relationships of CL-705G and pinacidil to open K_ATP channels. Kir6.2/SUR2A expressing HEK293 cells were preincubated with various concentrations (0.03–100 µM) of CL-705G or pinacidil before determining the initial Tl⁺ flux rates. Similar to published studies (Shindo et al., 1998), pinacidil opened the Kir6.2/SUR2A channel with an EC₅₀ of 11 µM (Figure 4). The CL-705G compound had an EC₅₀ of 9 μM, demonstrating its potency as a K_ATP channel opener similar to pinacidil.

To determine if compound CL-705G influences the smooth muscle K_ATP channel subtype, the Tl⁺ flux assay was repeated using HEK293 cells stably expressing Kir6.1/SUR2B. By contrast to Kir6.2/SUR2A, CL-705G had no effect on Kir6.1/SUR2B at 10 µM (not shown). Even when tested at 100 μM (more than 10x the EC₅₀ for Kir6.2), the effect of CL-705G was negligible compared to the vehicle-only control group (Figures 5A,B), despite the high sequence similarity between Kir6.1 and Kir6.2. We further tested each of the 15 identified compounds against Kir6.1/SUR2B channels and found that none of these activated the smooth muscle K_ATP channel subtype (Supplementary Figure S1). By contrast, some of these structural homologs, such as 44516625 and 41380967, inhibited TI⁺ flux rates through Kir6.1/SUR2B channels. We chose not to explore this observation further since there are already Kir6.1 blockers available commercially (e.g., PNU-37883 A). Based on these studies, we conclude that CL-705G preferentially open Kir6.2 channels and not Kir6.1 channels. A caveat to these studies is that we have not performed experiments with combinations of Kir6.2/SUR2B, Kir6.1/SUR1 or Kir6.1/SUR2A since these are not thought to occur naturally. Our focus has been to study well-described K_ATP subtypes Kir6.2/SUR2A (heart), Kir6.2/SUR1 (pancreas) and Kir6.1/SUR2B (vascular).

We next examined whether compound CL-705G affects other ion channels. We focused on the major Kir channels expressed in the heart to test for cardiotoxicity, including Kir2.1 and Kir3.1/Kir3.4 channels. We conducted whole-cell patch clamping with HEK293 cells transiently transfected with Kir2.1, which produced strongly inward rectifying K⁺ currents. Compound CL-705G (10–100 µM) had no effect on these Kir2.1 current (Figures 5C–E). To determine whether CL-705G affects Kir3 channels, we studied Kir3.1/Kir3.4 heterotetrameric channel heterologously expressed in Xenopus oocytyes. At 15 μM, compound CL-705G had no effect on Kir3.1/Kir3.4 currents (Figures 5F–H), and reduced Kir3.1/Kir3.4 currents by 17% at 100 μM, which is 10x higher than the EC₅₀ for K_ATP channels (data not shown). We also tested the effects of CL-705G against Nav1.7 given its involvement in pain responses (Hameed, 2019). For this experiment, we used cerebellar medulloblastoma TE671 cells, which endogenously express Nav1.7 (Ngum et al., 2022). However, CL-705G had no effect on these Na⁺ channels at 15 μM CL-705G (Figures 5I–K), while a slight inhibition was seen at 100 µM (not shown). Thus, amongst the targets tested, CL-705G preferentially activated Kir6.2-containing K_ATP channels. A more comprehensive screen would be needed against other channels before this compound can be considered specific to Kir6.2.

CL-705G can activate Kir6.2 when co-expressed with SUR2A (Figure 2) or with SUR1 (see later). We designed an experiment to determine whether CL-705G can activate Kir6.2 in the absence of a SURx subunit. Normally, only fully assembled Kir6.2/SURx channels effectively traffic to the surface membrane, whereas subunits expressed individually are retained in the endoplasmic reticulum. Removal of the ER retention signal (KRK), for example, by truncation of 36 amino acids of the Kir6.2 C terminus (Kir6.2Δ36), allows a degree of surface expression (Zerangue et al., 1999). We therefore transfected HEK293 cells with Kir6.2Δ36 in the absence of SURx. Whole-cell patch clamping was performed and currents were recorded in response to voltage ramps from 0 to −80 mV. Compound CL-705G did not activate Kir6.2Δ36 currents, but had a slight inhibitory effect, which was reversed upon washout (Figures 6A–C). Similarly, CL-705G had negligible effects on Tl⁺ fluxes of Kir6.2Δ36 when expressed by itself in HEK-293 cells (Figure 6D). By contrast, CL-705G strongly activated Tl⁺ fluxes of Kir6.2Δ36 when co-expressed with SUR2A (Figure 6E). We conclude, therefore, that the SUR subunit is needed for CL-705G to activate Kir6.2-containing K_ATP channels.

Our structural model docks CL-705G in the vicinity of the PIP₂ binding site (Figure 1). We therefore next examined whether the effect of the compound depends on PIP₂. In a first approach, we recorded Kir6.2/SUR2A currents in excised membrane patches and employed a method previously described to deplete PIP₂ at the cytosolic face of the membrane by applying a ∼5s pulse of 100 μM Ca²⁺ (Xie et al., 1999). Consistent with a key role for PIP₂ in K_ATP channel opening (Shyng et al., 2000), this intervention led to a rapid and sustained inhibition of K_ATP channel activity (Ca²⁺-induced “rundown”). Prolonged (∼20 s) application of 5 mM ATP activates phosphatidylinositol kinases, leading to de novo PIP₂ synthesis and restoration of PIP₂ levels (Xie et al., 1999). As a result, K_ATP channels are reactivated from Ca²⁺-induced “rundown”, as evidenced by robust K_ATP channel current upon ATP washout (Figure 7). Compound CL-705G strongly and rapidly activated K_ATP channels even after Ca²⁺-induced “rundown” and PIP₂ depletion (Figure 7). This result demonstrates that CL-705G can activate K_ATP channels in the absence of PIP₂.

To test whether CL-705G competes with PIP₂ for a common binding site, we designed an experiment to assess whether interventions that change intracellular PIP₂ levels would affect the response of the K_ATP channel to CL-705G. At nanomolar concentrations, wortmannin specifically blocks phosphoinositide 3-kinases (PI 3-kinases) which degrade PIP₂, whereas micromolar concentrations also blocks PI 4-kinases, which are required to generate PIP₂ (Nakanishi et al., 1995). We therefore treated Kir6.2/SUR2A-expressing HEK293 cells with 10 nM wortmannin to elevate PIP₂ levels (Huang et al., 2011). In another group, we used 10 µM wortmannin, together with 100 µM carbacol to activate muscarinic receptors and phospholipase C, which depletes PIP₂ (Trebak et al., 2009). Consistent with expectations, low wortmannin concentrations (elevated PIP₂) led to an increased Tl⁺ flux rate of Kir6.2/SUR2A-expressing HEK293 cells, whereas the higher wortmannin concentration plus carbacol (low PIP₂) mitigated K_ATP channel activity (Figures 7C,D). Neither of these two interventions, however, had any effect on the dose-dependence of CL-705G to activate Kir6.2/SUR2A channels (Figure 7E). This result suggests that the binding sites of CL-705G and PIP₂ are independent and non-overlapping.

Given the cardioprotective effects of K_ATP channel opening during hypoxia and ischemic preconditioning, we tested whether CL-705G would have cardioprotective actions. We utilized a model of ischemia and pharmacological preconditioning with isolated mouse cardiac myocytes, in which a protective role of K_ATP channels has previously been established (Storey et al., 2013; Yang et al., 2018). In brief, isolated mouse ventricular myocytes were rendered hypoxic by incubation in 1% O₂ followed by reoxygenation, and cell viability was determined. As expected, pharmacologically preconditioning with phenylephrine improved cell survival in a K_ATP dependent manner (Storey et al., 2013). Pre-incubation of the myocytes with 100 µM CL-705G led to a cell survival benefit of the same magnitude (Figure 8). Future experiments can be directed to determine effects of CL-705G on insulin secretion (c.f. diazoxide, to test effects on physiologically relevant Kir6.2/SUR1 in pancreatic β-cells) or to examine effects on blood pressure or blood flow (c.f. pinacidil, to test effects on physiologically relevant Kir6.1/SUR2B in the vasculature).

Loss-of-function pathogenic variants in KCNJ11 (the gene that encodes Kir6.2) can lead to congenital hyperinsulinism (CHI), a disorder characterized by persistent β-cell depolarization which results in excessive insulin secretion even in the presence of severe hypoglycemia. We tested whether CL-705G can rescue the gating defects of three such variants using the Tl⁺ flux assay; Kir6.2-F55L, Kir6.2-T72M and Kir6.2-R301C (Lin et al., 2006; Lin et al., 2008). As expected, when co-expressed with SUR1 to recapitulate the pancreatic β-cell K_ATP channel, the Tl⁺ flux rates were strongly diminished in all of the variants when compared to wild-type Kir6.2/SUR1. Compound CL-705G did not rescue Tl⁺ flux of Kir6.2-F55L/SUR1 or Kir6.2-T72M/SUR1 (not shown). By contrast, Kir6.2-R301C/SUR1 was minimally activated by diazoxide, but strongly activated by compound CL-705G (Figure 9). We have also tested three loss-of-function gating-defective SUR1 mutants (V715G, S1386P, and L1390P). Based our data, we did not expect CL-705G to activate any of these SUR1 mutants. Indeed, CL-705G did not rescue Tl⁺ flux deficits in any of these SUR1 mutants (data not shown).

Diverse compounds exist that can either inhibit (block) or activate (open) K_ATP channel activity (Tinker et al., 2018). Sulfonylureas are a class of K_ATP channel blockers and include compounds such as tolbutamide, glibenclamide, glipizide, and gliclazide. Nonsulfonyurea derivatives of glibenclamide (e.g., “glinides”) have emerged as a new generation of K_ATP channel blockers, represented by compounds such as meglitinide and repaglinide. Other nonsulfonylurea K_ATP channel inhibitors include PNU99963 and PNU37883A. Other compounds with very different structures activate K_ATP channel opening. These include benzothiadiazines (diazoxide), pyrimidine sulfates (minoxidil), pyridyl nitrates (nicorandil), benzopyrans (cromakalim), carbothiamides (aprikalim), and cyanoguanidines (pinacidil).

A therapeutic challenge is to find compounds that demonstrate selectivity against a specific K_ATP channel subtype. A K_ATP channel opener such as diazoxide is a hypotensive agent, but also promotes hyperglycemia, demonstrating significant cross-reactivity between smooth muscle and pancreatic β-cell K_ATP channels (Coetzee, 2013). Nicorandil and cromakalim activate both smooth muscle and cardiac muscle K_ATP channels. High affinity K_ATP channel blockers such as glibenclamide are being used therapeutically to block pancreatic β-cell K_ATP channels, but the tissue specificity is marginal since other K_ATP channels are effectively blocked at higher concentrations of the drug. Our data show that compound CL-705G shows a degree of selectivity by targeting only Kir6.2-containing channels, such as those in the cardiac ventricle and pancreatic β-cells.

The vast majority of K_ATP channel activators and blockers act by binding to the SURx subunit. Cryo-EM studies have demonstrated, for example, that glibenclamide binds in a pocket located within the transmembrane (TM) region above NBD1, which is framed by TM helices from TMD1 and TMD2 and further stabilized by a hairpin loop in L0 (Driggers and Shyng, 2023). Two amino acid residues in SUR1 (S1238 and T1242) are responsible for increased binding affinity and explains why glibenclamide blocks SUR1 channels better than SUR2x-containing channels. Other K_ATP channel blockers, such as repaglinide and carbamazepine bind to a similar region of the channel, but binding is coordinated by different amino acid residues. K_ATP channel openers, including NN414, P1075 and levcromakalim also bind to SUR1 and SUR2 within a transmembrane region, where binding is coordinated by specific and overlapping transmembrane helices (Driggers and Shyng, 2023). An important exception is a compound named PNU37886A, which appears to have significant specificity at low micromolar concentrations to block smooth muscle K_ATP channels without affecting K_ATP currents in cardiac and skeletal myocytes (Wellman et al., 1999; Kovalev et al., 2004). PNU37886A is one of the few compounds that act as a pore-blocker by interacting with the Kir6.1 subunit, and does not act via SURx. Studies to identify new K_ATP channel modulators that are specific to a subclass of K_ATP channels are therefore needed.

Our virtual library screen was designed to identify compounds that potentially bind to a pocket identified lined by adjacent Kir6.2 subunits in the heteromeric channel complex (Figure 1). The amino acid residues predicted to interact with Kir6.2 are shown in Supplementary Figure S2. Of the 15 compounds initially tested (Table 1) and the five 2D structural homologs (Table 2), CL-705G was the most potent of these compounds to activate K_ATP channels, with a potency akin to pinacidil. Pinacidil binds to SUR, with a higher affinity to SUR2A and SUR2B when compared to SUR1 (Moreau et al., 2000). Compound CL-705G, by contrast, might mediate its K_ATP channel opening properties by binding to Kir6.2. The presence of SURx is required, however, since when Kir6.2 is expressed by itself (as a C-terminal truncated subunit Kir6.2Δ36), it was not activated by CL-705G. Co-expression of SUR2A with Kir6.2Δ36, by contrast, led to strong activation of K_ATP channel activity by CL-705G. We deem it unlikely that CL-705G mediates its effect by binding to SURx since our docking attempts did not support the possibility that CL-705G binds to a conserved pocket of SUR1 or SUR2 (not shown). Moreover, if CL-705G activated K_ATP channels by binding to SUR2, then we would expect Kir6.1/SUR2B channels to be activated at least partially. By contrast, the Kir6.1 channels were unaffected by CL-705G even at 10x the effective concentration needed to activate Kir6.2. A more likely explanation is that the CL-705G binding pocket in Kir6.2 is disrupted in the absence of SURx, or alternatively that SUR participates in CL-705G binding to Kir6.2. The latter possibility is supported by the proximity of SUR Loop 0 to the pocket to which CL-705G docks (not shown).

Since CL-705G is predicted to bind in close proximity of the PIP₂ binding site (Figure 1), we tested whether they interact. Following Ca²⁺-induced rundown, which inactivates K_ATP channels by depleting PIP₂ levels (Xie et al., 1999), CL-705G led to a rapid and robust K_ATP channel activation. This result might be interpreted that CL-705G rescues the channel from an inactivated state. Certain Kir6.2 mutations inhibit gating by promoting an inactivated state, from which channel activity can be rescued by ATP (in the absence of Mg²⁺) or by the non-hydrolysable AMP-PNP (Lin et al., 2008). Indeed, the finding that the compound rescued the activity of one of these gating-defective Kir6.2 mutants (Kir6.2-R301C) is consistent with this interpretation. If so, however, it would not be clear why CL-705G specifically rescued R301C, and no other similar Kir6.2 mutants that enter the inactivated state (F55L and T72M) or any of the SUR1 mutants tested (V715G, S1386P, and L1390P). Moreover, there is little evidence that wild-type Kir6.2 channels enter the inactivated state and the data in Figure 9 cannot easily be explained by this mechanism. Another possibility is that CL-705G interacts with, or substitutes for, PIP₂ to activate the K_ATP channel. We tested this possibility by determining whether PIP₂ changes the EC₅₀ of CL-705G. Even though our interventions to manipulate PIP₂ (low vs. high wortmannin concentrations) had the expected effects in baseline K_ATP channel activity, we not observe an obvious change to the apparent EC₅₀ of CL-705G, suggesting that PIP₂ and CL-705G bind to distinct sites. A possible mechanism is presented in Supplementary Figure S2, which demonstrates that CL-705G may bind to residues in the M2 transmembrane helix, as well as residues in the Kir6.2 C-terminus. Thus, CL-705G may bridge these functional domains to cause a conformational change that favors the channel open state. Some of these amino acid residues are conserved in Kir6.1. However, the Kir6.1 C-terminus is offset from the membrane interface by ∼5.8 Å relative to Kir6.2 (Sung et al., 2021), which would preclude CL-705G from having a similar effect, which may partially explain the specificity of the compound against Kir6.2. Further studies are needed to test these ideas.