Identification of potent and selective small molecule TALK-1 inhibitors that promote β-cell function Authors: Arya Y Nakhe, Jordyn R Dobson, Prasanna K Dadi, David Westover, Emily L Days, Paige Vinson, C David Weaver, Jerod S Denton, Spencer J Peachee, Soma Behera, Shannon E Gibson, Valentina A Rodriguez Da Silva, Amit Sela, Matthew T Dickerson, David A Jacobson Published: 2026 ABSTRACT AbstractThe two-pore domain K+ (K2P) channel TALK-1 is a key regulator of β-cell electrical excitability, Ca2+ handling, and glucose-stimulated insulin secretion. KCNK16, the gene encoding TALK-1, is the most abundant β-cell K+ channel transcript and the most islet-restricted ion channel. Genetic evidence also links KCNK16 to diabetic phenotypes, including a gain-of-function polymorphism (rs1535500) that predisposes individuals to type 2 diabetes and a gain-of-function mutation (p.L114P) that causes maturity-onset diabetes of the young. Moreover, ablation of β-cell TALK-1 improves glucose homeostasis in the context of diet-induced obesity. However, it remains unclear if TALK-1 can be pharmacologically targeted to enhance β-cell function and reduce hyperglycemia in type 2 diabetes patients. Here, we identify, to our knowledge, the first selective inhibitors of TALK-1. This was accomplished with a high-throughput fluorescence-based thallium (Tl+) flux assay, which identified 286 TALK-1 inhibitors. Of these, 81 compounds selectively inhibited TALK-1 without affecting other K+ channels (ie, TALK-2, TASK-3, KATP, and hERG). One compound, VU0544194 (K2P16-i7), was selected for further evaluation based on its potency and TALK-1 selectivity. K2P16-i7 inhibited β-cell K2P currents, increased islet Ca2+ influx, enhanced endoplasmic reticulum Ca2+ storage, and augmented glucose-stimulated insulin secretion. We then identified scaffold domains of K2P16-i7 that, with specific modifications, improved potency while retaining the ability to augment endoplasmic reticulum Ca2+ stores. Together, these findings identify, to the best of our knowledge, the first selective small molecule inhibitors of TALK-1 and determine that they improve β-cell function, which could enable an islet-selective therapeutic modality for treating diabetes.Significance StatementTALK-1 channels are involved in regulating glucose homeostasis and weight, making them a promising endocrine-specific target for diabetes and obesity. Here, this study identifies, to our knowledge, the first selective TALK-1 inhibitors that show efficacious regulation of islet Ca2+ handling and insulin secretion; these molecules will enable future investigation of human endocrine TALK-1 function as well as drug design.Key words: KCNK16, TALK-1, Islet, Insulin secretion, Two-pore domain K+ channel, Calcium Introduction Secretagogue-stimulated insulin secretion is dependent on β-cell Ca2+ influx, which is perturbed under diabetic conditions and contributes to hyperglycemia.1 As ion channels regulate β-cell Ca2+ handling, pharmacological ion channel modulators have been used to elevate diabetic β-cell Ca2+ for insulin secretion or reduce Ca2+ to prevent excitotoxicity.2, 3, 4 To date, the only ion channel mutations that cause monogenic diabetes are found in ABCC8 and KCNJ11, the genes encoding the ATP-sensitive K+ channel complex (KATP; associated with maturity-onset diabetes of the young [MODY], permanent and transient neonatal diabetes). Sulfonylureas inhibit KATP channels and are widely used for treating type 2 diabetes (T2D) and monogenic diabetes. However, sulfonylureas cause excessive β-cell Ca2+ entry and lose efficacy over time due to excitotoxicity.5,6 Furthermore, because KATP channels are also expressed in cardiomyocytes, KATP inhibition increases cardiovascular risk.7 Therefore, a barrier to diabetes treatment is a durable β-cell selective medicine that restores Ca2+ handling and insulin secretion without excitotoxicity. KCNK16, the gene encoding TALK-1, is the most highly expressed ion channel in the endocrine pancreas and shows nearly exclusive islet expression specificity index.8, 9, 10 TALK-1 channels are small conductance two-pore domain K+ (K2P) channels that set the β-cell plateau potential from where action potentials fire.11 TALK-1 activity at the plasma membrane reduces β-cell Ca2+ oscillation frequency and, thus, slows the rate of pulsatile insulin secretion.11,12 Interestingly, decreased pulsatile insulin secretion contributes to insulin resistance during T2D pathogenesis.13 Furthermore, a nonsynonymous coding sequence gain-of-function (GOF) polymorphism in KCNK16 (rs1535500, encoding TALK-1 A277E) shows a 1.4-fold increase in TALK-1 activity and is associated with increased risk of T2D.11,14,15 The MODY-associated GOF mutation (L114P) in TALK-1 also results in a drastic >300-fold increase in K+ flux through TALK-1, thereby reducing β-cell Ca2+ influx and limiting insulin secretion.12,16 This mutation results in neonatal lethality in mice, presumably due to severe hyperglycemia, which can be partially rescued with insulin treatment.12 Because TALK-1-mediated changes in β-cell electrical activity are modest compared to KATP inhibition or loss-of-function, the resulting elevation in β-cell Ca2+ following TALK-1 ablation does not result in stress.11 Taken together, this suggests that TALK-1 is an islet-specific target that could be inhibited to normalize Ca2+ handling and pulsatile insulin secretion in T2D. TALK-1 is functional both on the β-cell plasma membrane and the endoplasmic reticulum (ER) membrane.17 ER-localized TALK-1 provides a K+ countercurrent that prevents accumulation of negative charge on the luminal membrane during ER Ca2+ ([Ca2+]ER) leak, thereby enhancing the electrochemical driving force for [Ca2+]ER efflux.17 Thus, TALK-1 GOF polymorphisms or mutations (eg, L114P and A277E) deplete [Ca2+]ER stores by increasing [Ca2+]ER leak, whereas loss of TALK-1 function (eg, global TALK-1 knockout [KO] and β-TALK-1-KO) enhances [Ca2+]ER stores by reducing [Ca2+]ER leak.11,16 TALK-1 control of [Ca2+]ER release regulates cytosolic Ca2+ ([Ca2+]C), which in turn tunes insulin secretion. Importantly, one of the primary roles for [Ca2+]ER is regulating the protein folding activity of Ca2+-dependent ER chaperones.18 Disruption of [Ca2+]ER homeostasis, as observed in T2D, results in ER stress and activation of the unfolded protein response pathway, ultimately reducing insulin production and secretion.19,20 Notably, expression of ER stress markers was reduced in islets from high-fat diet (HFD)-fed TALK-1 KO mice, whereas β-cell ER stress was exacerbated in islets with the T2D-associated GOF TALK-1 polymorphism.17 These findings suggest that TALK-1 inhibition could enhance β-cell [Ca2+]ER storage, thereby mitigating ER stress and preserving β-cell function during T2D. TALK-1 also imparts non-β-cell–mediated effects that could exacerbate T2D pathogenesis. For example, TALK-1 KO mice show reduced glucagon secretion at low glucose concentrations,21 whereas glucagon secretion was increased from TALK-1 GOF islets (L114P) under fasting and euglycemic conditions.12 In T2D, hyperglucagonemia stimulation of hepatic glucose output contributes to hyperglycemia.22,23 Therefore, inhibition of glucagon signaling reduces hyperglycemia in T2D.24 β-TALK-1-KO mice also show protection against HFD-diet-induced weight gain,25 which could be due to the role of TALK-1 in intestinal enteroendocrine cells, where it is also expressed. Taken together, these findings suggest that TALK-1 inhibitors may reduce hyperglucagonemia and obesity associated with T2D. Here, we report, to the best of our knowledge, the first selective small molecule inhibitors of TALK-1. Inhibition of TALK-1 increased β-cell [Ca2+]ER stores, elevated plateau fraction, and enhanced second-phase glucose-stimulated insulin secretion (GSIS). Therefore, small molecule TALK-1 inhibitors replicate changes observed in TALK-1–deficient β-cells and offer powerful tools to explore TALK-1's role in β-cell physiology and dysfunction. This work represents, to the best of our knowledge, the first step in developing TALK-1 inhibitors as potential therapies for diabetes and associated metabolic disorders. Materials and methods Chemicals and reagents All chemicals and reagents were purchased from ThermoFisher or Sigma-Aldrich unless specified otherwise. Generation of T-REx-293 TALK-1 and TALK-2 overexpression cell lines The T-REx-293 TALK-1 human transcript variant 3 stable cell line was generated as previously detailed.11,26 Briefly, early passage T-REx-93 cells were grown to 80% confluency in 75 cm2 flasks in Dulbecco’s modified Eagle’s medium (DMEM) growth medium containing 10% FBS, 50 U/mL penicillin, 50 μg/mL streptomycin, and 5 μg/mL blasticidin S (DMEM-B) and were transfected with pcDNA5/TO-TALK-1 DNA or pcDNA5/TO-TALK-2 DNA utilizing Lipofectamine 3000 according to the manufacturer’s protocol. The transfection medium was replaced with DMEM-B after 5 hours. After 24 hours, the media was replaced again with DMEM-B supplemented with 250 μg/mL hygromycin (DMEM-BH) to begin stable clone selection. Cells were supplemented with fresh DMEM-BH medium every 2 to 3 days up to ∼14 days when small colonies of stable cells developed. These colonies were dissociated using trypsin into a suspension of single cells. The suspension was diluted to a concentration of 0.7 cells/20 μL, and 20 μL of the suspension was plated in each well of a BD PureCoat amine-coated 384-well plate (Corning, Cat # 356719), such that 70% of wells should receive one cell. Monoclonal colonies were subsequently split into two 384-well plates, one to assess thallium (Tl+) flux for selection of the most efficient clone for further studies and the other for further expansion. The monoclonal T-REx-293-TALK-1 colonies were then selected by their tetracycline-induced response to a fluorescent-based Tl+ flux assay (pH 7.3 as detailed below). The monoclonal T-REx-293-TALK-2 cell lines were also selected based on a Tl+ flux assay run at a pH of 7.7. Numerous clones with the strongest tetracycline induction and robust Tl+ flux were expanded and frozen back for future use. The T-REx-293 cell lines expressing Kir4.2 and Kir6.2/SUR1 KATP channels were generated previously.27,28 T-REx-293 hERG expressing cells were provided by Dr David Weaver and were developed as described previously.29 Tl+ flux assay Tl+ flux assays were performed as detailed previously.26 Briefly, stably transfected T-REx-293 cells expressing the indicated K+ channels were cultured overnight in DMEM supplemented with 1% GlutaMAX, 10% FBS, and 1 μg/mL tetracycline at a density of 25,000 cells/well in BD PureCoat amine-coated clear bottom 384-well plates at 37 °C in 5% CO2. On the day of assay, cells were washed with assay buffer (Hanks' balanced salt solution with 20 mM HEPES and pH 7.3) using ELx405 Microplate Washer (Agilent) and then loaded with the Tl+-sensitive dye (Thallos-AM, 2 μM) prepared in assay buffer containing 0.01% (w/v) Pluronic F-127 (Life Technologies) and incubated for 1 hour at 37 °C in 5% CO2. Plates were then washed with assay buffer again and loaded into a whole-plate kinetic imager (Panoptic, Wavefront Bio). A baseline fluorescent recording was obtained at 1 Hz for 10 seconds (excitation 482 ± 32 nm and emission 536 ± 40 nm), followed by the addition of screening compounds (10 μM), DMSO (0.1% v/v), or control (niclosamide) for a 5-minute incubation period after which Tl+ stimulus buffer (5x in mM: 125 D-gluconic acid sodium salt, 5 Tl2SO4, 1 MgSO4, 1.8 CaSO4, 5 D-glucose, 20 HEPES, and pH 7.3) was added and data collection continued for an additional 3 minutes. For T-REx-293-TALK-2 cells, the Tl+ flux assay was performed as described above, but all solutions were at pH 7.7. A 7-point concentration response curve (CRC) was run on selected molecules with a 3-fold dilution series from 50 μM to 400 nM. Western blotting Cell lysates were obtained from T-REx-TALK-1 cells or T-REx-TALK-2 cells treated with or without tetracycline (1 μg/mL) for 24 hours, cultured in 24-well plates by scraping cells into 200 μL ice-cold PBS/well, followed by subsequent sonication for ∼30 seconds. Equal concentrations of protein extracts (5 μg) were obtained from each condition: vehicle (VHL)-treated T-REx-TALK-1, tetracycline-treated T-REx-TALK-1, VHL-treated T-REx-TALK-2, and tetracycline-treated T-REx-TALK-2 (as determined by BCA Protein Assay, Pierce). Protein extracts were run on a 10% SDS-Page Gel (Bio-Rad) followed by a transfer onto a nitrocellulose membrane. The membrane was then blotted with primary antibodies: anti-TALK-1 antibody (APC-170, Alomone Labs) or anti-TALK-2 antibody (NBP1-92041, Novus Biologicals, LLC), followed by the secondary antibody: anti-Rabbit IgG (H+L), HRP conjugate. The membrane was imaged on a Digital ChemiDoc MP using SuperSignal West Pico Chemiluminescent Substrate (ThermoFisher Scientific). Mouse model and islet isolation Mice used for these studies were handled in compliance with guidelines approved by the Vanderbilt University Animal Care and Use Committee protocols (M2200007-00). The generation of whole-body TALK-1 KO (Kcnk16–/–) mice is detailed in Vierra et al.11 Floxed KCNK16 mice were generated as detailed in Graff et al.25 C57Bl/6J mice breeding pairs were obtained from the Jackson Laboratory, and the colonies were expanded and maintained in-house. Islets were isolated from adult male mice (12- to 16-week-old) as previously described.11 Briefly, pancreata were distended and digested using collagenase P (Roche), followed by density gradient centrifugation utilizing Histopaque-1077. Islets were either dispersed into clusters by titration in 0.005% trypsin or cultured as whole islets overnight in RPMI supplemented with 5.5 mM glucose at 37 °C in 5% CO2. Intracellular Ca2+ imaging Islets were washed with RPMI supplemented with 2 mM glucose and loaded with 2 μM Fura-2 AM for 30 minutes. This was followed by 2 washes in modified Krebs-Ringer-HEPES buffer (KRHB) containing (mM) 119 NaCl, 10 HEPES, 4.7 KCl, 2 CaCl2, 1.2 MgSO4, and 1.2 KH2PO4, pH 7.35 adjusted with NaOH and supplemented with the indicated glucose concentrations. Fura-2 AM (ratio 340/380 Ex-535 Em; F340/F380) fluorescence was measured every 5 or 10 seconds as an indicator of intracellular Ca2+ using a Nikon Eclipse Ti2 microscope equipped with a Photometrics Prime 95B 25 mm sCMOS Camera. For cyclopiazonic acid (CPA)-induced [Ca2+]ER release measurements, islets were perifused in KRHB buffer containing 9 mM glucose, 100 μM diazoxide, without extracellular Ca2+. Fura-2 AM fluorescence was monitored to measure changes in [Ca2+]ER through changes in [Ca2+]c. For assessing Ca2+ oscillations, dispersed islet clusters were transduced with or without Ad-mCherry-iCre (Vector Biolabs Cat. # 1771) for 2 days. On the day of imaging, islet clusters were incubated with 16.7 mM glucose KRHB for 30 minutes at room temperature prior to imaging. For transduced islet clusters, only those with mCherry fluorescence were analyzed. In some cases, intracellular Ca2+ was imaged on the Panoptic kinetic plate reader. Cal520 (ATT Bioquest; excitation 482 ± 32 nm and emission 536 ± 40s nm) with 0.01% (w/v) Pluronic F-127 was loaded into the cells for 1 hour prior to imaging. The cells were then washed in assay buffer (Hanks' balanced salt solution without Ca2+ supplemented with 20 mM HEPES). Fluorescent recordings were obtained at 1 Hz for 10 seconds. Baseline measurements in 0 mM Ca2+ were taken for 5 minutes, followed by the addition of CPA (Hello Bio; final concentration 50 μM) for 3 minutes. Electrophysiology Electrophysiology recordings were performed with an Axopatch 200B amplifier with pCLAMP10 software as previously described.12 Islet cells were isolated from mouse pancreata as described above. Islets were then dispersed with Trypsin dissociation reagent. Single islet cells were then plated for electrophysiology. Patch electrodes were pulled to 3–5 megaohm resistance loaded with intracellular solution including (in mM) 140 KCl, 1 MgCl2, 10 EGTA, 10 HEPES, and 4 MgATP (pH with KOH to 7.2). On the day of recordings, the cells were washed and then incubated in Krebs-Ringer buffer containing 11 mM glucose, K+ channel inhibitors tetraethylammonium (10 mM) and tolbutamide (100 μM), and either VHL or K2P16-i7. K2P currents were recorded using the whole-cell voltage clamp configuration with a voltage ramp from –120 mV to 60 mV. The currents were low-pass filtered at 1 kHz and sampled at 10 kHz. Insulin secretion assays C57Bl/6J mouse islets were equilibrated in DMEM (11966025, Gibco) supplemented with 5.6 mM glucose, 0.5 mg/mL bovine serum albumin, 0.5 mM CaCl2, and 10 mM HEPES for 2 hours at 37 °C and 5% CO2. 50 islets were loaded per BioRep perifusion chamber, and Bio-Gel P-4 Media (Bio-Rad) was used to keep islets in place. Islets were perfused at a rate of 100 μL per minute at 37 °C with the indicated solutions. The perfusate was collected at 4 °C in a 96-well plate (PERI-Lite perifusion system, BioRep). A mouse insulin ELISA kit was used to determine the concentration of insulin (Mercodia). Statistical analysis Data acquisition and analysis were performed using Waveguide software at Vanderbilt University High-throughput Screening Core and Microsoft Excel (area under the curve [AUC], percent inhibition, slope, Z′). Only assays with a Z′ >0.5 were used for analysis. Potency (IC50) values were determined by fitting the Hill equation using variable-slope nonlinear regression analyses performed with GraphPad Prism (GraphPad Software). Analysis for Ca2+ imaging was performed with Microsoft Excel and GraphPad Prism. The plateau fraction was calculated by setting the threshold to 50% of the maximum and minimum of the Ca2+ oscillations per islet cluster and measuring the percentage of time above the threshold. All data are represented as the mean ± SEM unless otherwise specified. Electrophysiology experiments utilized Clampfit software. Results A high-throughput screening approach to identify TALK-1 K+ channel inhibitors The pipeline for identifying TALK-1 inhibitors is described in Fig. 1A. First, a stable T-REx-TALK-1 cell line was generated containing the human KCNK16 transcript variant 3 in a tetracycline-inducible promoter construct (pCDNA3.1 5/TO). T-REx-TALK-1 cells displayed strong tetracycline-dependent expression of TALK-1 as determined by a Western blot (Fig. 1B). To evaluate if these cells could be used for high-throughput screening (HTS), a fluorescent Tl+ assay was optimized to assess TALK-1 function (Fig. 1C). A Tl+ concentration response (mM Tl+: 0.5, 1, 2, and 3) identified that the fluorescent response between 1 mM and 2 mM Tl+ would be optimal for monitoring TALK-1 inhibition but also TALK-1 activation; therefore, we tested and utilized 1.5 mM Tl+ for HTS (Fig. 1D). Additionally, a DMSO tolerance test determined that T-REx-TALK-1–mediated Tl+ flux was not impacted at DMSO concentrations reached during addition of small molecules. Utilizing the resulting TALK-1 Tl+ assay, a pilot Tl+ screen was performed with the Spectrum Collection consisting of 2320 molecules and Bioactive Lipid Library (Cayman Chemicals) consisting of 928 molecules. This pilot screen identified 5 TALK-1 channel activators and 8 TALK-1 channel inhibitors, including molecules such as niclosamide, which was utilized as a positive control inhibitor for the Tl+ flux assay (Fig. 1, E and F). A 9-point CRC analysis determined an IC50 = 292 nM for niclosamide (Fig. 1F). The uniformity and reproducibility of the assay were assessed by performing checkerboard analyses of TALK-1 activity in alternative wells with either DMSO (VHL; 0.1% v/v) or 10 μM niclosamide (in a 384-well plate; Fig. 1G). This assay produced an average Z′ score = 0.57, indicating that this assay enables robust hit identification with high confidence (Fig. 1G). High-throughput screening approach to identify TALK-1 inhibitors. (A) Diagram of the HTS pipeline to identify TALK-1 inhibitors. (B) Western blot showing tetracycline (Tet.) induction of TALK-1 ion channel expression in the T-Rex-TALK-1 cell line. (C) Schematic of Tl+ flux assay. Green circles represent thallos dye. (D) Representative Tl+ flux trace of Tet. induced and uninduced T-REx-TALK-1 cells before and after the indicated concentrations of Tl+ addition (n = 12 well/condition). (E) TALK-1 inhibitors identified in the pilot screens. (F) TALK-1 niclosamide CRC (IC50 = 292 nM; 95% confidence interval [CI], 252–337 nM; n = 6 wells/concentration; error bars represent standard deviation). (G) Scatter plot of induced T-REx-TALK-1 cells treated with vehicle (DMSO) or niclosamide (10 μM) in a checkerboard pattern. Z′ calculation for this 384-well plate was 0.664. A high-throughput Tl+ screen identifies 286 TALK-1 inhibitors We next performed a TALK-1 Tl+ HTS utilizing 48,408 structurally diverse compounds from the Vanderbilt Drug Discovery Collection at a concentration of 10 μM (Fig. 2A). TALK-1 inhibitors were selected based on >25% reduction of Tl+ flux slope and area under the Tl+ curve, as well as ≥3 standard deviation difference compared to DMSO control. VU0544194 is a representative TALK-1 inhibitor, which inhibits 55% of TALK-1–mediated Tl+ flux slope normalized to niclosamide (considered 100%; Fig. 2, B and C). The initial TALK-1 HTS identified 788 TALK-1 inhibitors with a 1.62% hit rate. Inhibitors were retested in duplicate with T-Rex-TALK-1 cells and counter-screened against T-Rex-TALK-1 cells without tetracycline induction (uninduced cells: Fig. 2, F and G). Specificity of TALK-1 inhibitors was first assessed with a Tl+ flux assay on T-REx-Kir4.2 cells expressing inward-rectifier Kir4.2 K+ channels (Fig. 2H; representative Tl+ traces Fig. 2, E, G, and I). This resulted in 286 hits that inhibited TALK-1 and did not inhibit Kir4.2. The potencies of these small molecules were measured by 7-point CRCs, which identified many structurally diverse TALK-1 inhibitors with IC50 values <10 μM (representatives shown in Fig. 3). Primary TALK-1 HTS identified potent TALK-1 inhibitors. (A) Cartoon schematic of the workflow used for the full T-REx-TALK-1 HTS. (B) Scatter plot of a 384-well plate showing the percent inhibition of tetracycline (Tet)-induced T-REx-TALK-1 Tl+ flux with TALK-1 inhibitors identified during the primary HTS. (C) Representative Tl+ traces for Tet-induced T-REx-TALK-1 cells treated with either VHL, niclosamide, or VU0544194. (D) Confirmatory screen of the remaining TALK-1 inhibitors identified in the primary HTS: percent inhibition of Tl+ flux slope. (E) Percent inhibition of Tl+ flux AUC in uninduced T-REx-TALK-1 cells. (F) Percent inhibition of Tl+ flux slope in T-REx-Kir4.2 cells. (G) Representative Tl+ flux traces with either VHL, VU0646169, or niclosamide in Tet-induced T-REx-TALK-1 cells. (H) Representative Tl+ flux traces comparing Tl+ flux inhibition of VU0646169 in the uninduced T-REx-TALK-1 cells. (I) Representative Tl+ flux traces comparing Tl+ flux inhibition of VU0646169 in the T-REx-Kir4.2 cells. Potencies of TALK-1 inhibitors were measured using 7-point CRC assays. (A–F) Concentration response curves (CRCs) and structures of 6 molecules tested in triplicate using a 7-point, 3-fold serial dilution series from 50 μM to 400 nM. Percent inhibition of AUC of Tl+ flux by the small molecule compounds was calculated using the formula: ([Compound AUC – VHL AUC]/[Niclosamide AUC – VHL AUC]) × 100, and IC50 values were calculated using a 4-parameter logistic fit (Hill equation). VU0630026: 2.61 μM (95% CI, 0.29–4.171 μM); VU0646163: 5.15 μM (95% CI, 3.62–7.64 μM); VU0621279: 2.29 μM (95% CI, 1.44–3.13 μM); VU0544194: 5.20 μM (95% CI, 3.96–6.92 μM); VU0532253: 5.52 μM (90% CI, 3.5–13.6 μM); VU0520878: 1.27 μM (90% CI, 0.0165–2.01 μM) (n = 3 replicates/concentration; error bars represent standard deviation). Secondary screens identify 81 selective TALK-1 inhibitors Because TALK-1 shares 44% sequence identity with TALK-2, the 286 hits identified in the initial TALK-1 HTS were further tested for selectivity by screening them first on tetracycline-inducible T-REx-TALK-2 cells. The T-REx-TALK-2 cell line was developed in a similar manner as the T-REx-TALK-1 cell line (see Materials and Methods), and tetracycline-dependent expression of TALK-2 was confirmed by a Western blot (Fig. 4A). Due to the alkaline activation of TALK-2, T-REx-TALK-2 Tl+ flux did not occur at pH 7.3 but was present at a pH of 7.7, which was used for the TALK-2 secondary screen. The T-REx-TALK-2 cell line showed optimal Tl+ flux in response to 1 mM Tl+ (Fig. 4, B and C). Tetracycline induction of T-REx-TALK-1 and T-REx-TALK-2 cells increased Tl+ flux compared to uninduced cells (AUC: T-REx-TALK-1 induced 339.5 ± 10.96 vs uninduced 211.1 ± 9.09; T-REx-TALK-2 induced 307.7 ± 0.420 vs uninduced 182.1 ± 1.554). Out of the 286 TALK-1 inhibitors, those that inhibited Tl+ flux in uninduced TALK-1 cells and showed IC50 values >10 μM were removed. Of the remaining 187 hits, 23 were removed because they inhibited >25% Tl+ flux through TALK-2 channels (VU0547061 is a representative TALK-2 inhibitor; Fig. 4, D and E). Additional secondary screens were performed against TASK-3 (K2P channel), KATP, and the voltage-gated K+ channel, Kv11.1/hERG; these identified 33, 9, and 111 nonspecific inhibitors, respectively (Fig. 4F; Supplemental Fig. 1). The secondary screens resulted in 81 TALK-1 selective inhibitors that were grouped into 14 individual cohorts based on structural similarity (cutoff Tanimoto coefficient >60%). Secondary screens for TALK-1 channel inhibitor selectivity. (A) Western blot showing tetracycline-induced TALK-2 expression in the T-REx-TALK-2 cell line. (B) Optimization of Tl+ concentration was performed by testing 0.5, 1, 2, and 3 mM Tl+ in tetracycline-induced or uninduced T-REx-TALK-2 cells (n = 12 wells/condition). (C) AUC of Tl+ flux (1 mM) in tetracycline-induced (+) or uninduced (-) T-REx-TALK-1 and T-REx-TALK-2 cells (AUC: T-REx-TALK-1 induced 339.5 ± 10.96 vs uninduced 211.1 ± 9.09; T-REx-TALK-2 induced 307.7 ± 0.420 vs uninduced 182.1 ± 1.554; n = 3; statistical analysis was performed using Student’s t test; ∗∗∗P < .001, ∗∗∗∗P < .0001). (D) Scatter plot showing Tl+ AUC of tetracycline-induced T-REx-TALK-2 cells treated with TALK-1 inhibitor compounds (n = 3 replicates/compound; error bars represent SEM). (E) Representative Tl+ flux traces for either VHL, VU0547061, or uninduced T-REx-TALK-2. (F) The number of TALK-1 nonselective inhibitors identified in secondary screens against TALK-2, TASK-3, KATP, and hERG channels. Identification of inhibitors that inhibit ER-localized TALK-1 and stimulate [Ca2+]ER leak TALK-1 K+ currents on the ER membrane provide a countercurrent for Ca2+ release and leak from the ER lumen, thereby enhancing Ca2+ efflux from the ER. This was assessed utilizing SERCA inhibition with CPA and monitoring the resulting rise in [Ca2+]C that is representative of [Ca2+]ER leak (Fig. 5A). T-REx-TALK-1 cells without tetracycline induction (uninduced cells) showed typical [Ca2+]ER leak in response to SERCA inhibition, whereas TALK-1 induction reduced [Ca2+]ER leak, likely due to reduced [Ca2+]ER storage (CPA AUC: VHL induced 186.9 ± 1.99 vs VHL uninduced 225.5 ± 2.2; Fig. 5, B and C). Fourteen TALK-1 inhibitors representing all structural similarity groups were assessed for their ability to decrease TALK-1–mediated [Ca2+]ER depletion (Supplemental Fig. 2). Of these, K2P16-i7 showed the greatest recovery of [Ca2+]ER leak (∼66%) following TALK-1 induction compared to VHL (CPA AUC: induced K2P16-i7 treated 211.4 ± 7.27 vs K2P16-i7 uninduced 240.5 ± 6.46; Fig. 5, B and C; K2P16-i7 structure and physicochemical properties Fig. 5, D and E; Supplemental Fig. 3). This suggests that K2P16-i7 inhibits TALK-1 channels localized to the ER membrane in addition to inhibiting Tl+ flux through TALK-1 channels in the plasma membrane. Testing the impact of VU0544194 (K2P16-i7) on Ca2+ handling. (A) Schematic of SERCA inhibition with CPA to evaluate [Ca2+]ER storage. (B) Average [Ca2+]C traces for CPA-induced [Ca2+]ER release measured in T-REx-TALK-1 cells with or without tetracycline induction, pretreated (1.5 hours) with VHL or K2P16-i7 (10 μM). (C) Quantification of AUC for the respective CPA-induced [Ca2+]C responses from (B) (VHL uninduced 225.5 ± 2.2; VHL induced 186.9 ± 1.99; K2P16-i7 uninduced 240.5 ± 6.46; K2P16-i7 induced 211.4 ± 7.27; n = 4 days (avg of 12 wells/d); statistical analysis performed using Tukey’s multiple comparisons test; ∗P < .05, ∗∗P < .01, ∗∗∗P < .001, ∗∗∗∗P < .0001). (D) Structure of K2P16-i7, a TALK-1-selective lead compound. (E) Physicochemical properties of K2P16-i7 indicate alignment with Lipinski’s rule of five for drug-likeness of molecules. K2P16-i7 inhibits primary β-cell TALK-1 currents and increases glucose-stimulated Ca2+ entry We next assessed K2P16-i7 inhibition of β-cell TALK-1 channels. Whole-cell voltage clamp electrophysiology was utilized to measure TALK-1 currents in response to a voltage ramp from –120 mV to 60 mV in mouse β-cells treated with VHL or K2P16-i7 (Fig. 6A). K2P16-i7 resulted in a reduction in β-cell K2P currents at 0 mV, 30 mV, and 60 mV (Fig. 6B). The amount of K2P16-i7 inhibition was determined by setting the VHL to 100% K2P current and subtracting the current densities from the indicated voltages. This showed that on average up to 45% of K2P currents were inhibited across 0 mV, 30 mV, and 60 mV, which is a similar reduction observed in β-cells with TALK-1 ablation (Fig. 6C).11 K2P16-i7 inhibits mouse β-cell K2P currents. (A) Representative K2P current recordings from dispersed mouse β-cells pretreated with either VHL or K2P16-i7 in response to a voltage ramp from –120 to +60 mV. (B) Quantification of the current density (pA/pF) of VHL vs K2P16-i7 pretreated β-cells at the specified voltages. (C) Quantification of the percent K2P currents from β-cells pretreated with K2P16-i7 relative to VHL. Statistical analysis performed using multiple unpaired t test; ∗∗P < .01, ∗∗∗P < .001, n = 16 VHL-treated cells; n = 12 K2P16-i7–treated cells from 2 independent islet preps. As TALK-1 limits the frequency of glucose-stimulated islet Ca2+ oscillations,11 we next tested the impact of K2P16-i7 on islet Ca2+ handling. Islets from KCNK16fl/fl mice were isolated and dispersed into islet clusters. Dispersed islet clusters were transduced with or without Ad-mCherry-iCre to KO KCNK16 (TALK-1 KO). Incubation of these islet clusters with 16.7 mM glucose resulted in Ca2+ oscillations that showed increased frequency with iCre expression compared to controls due to ablation of TALK-1 (Supplemental Fig. 4). Interestingly, KCNK16fl/fl control islet clusters exhibited increased plateau fraction in response to acute treatment with K2P16-i7 (Fig. 7, A and B). Although control islet clusters showed an increase in glucose-stimulated Ca2+ oscillation frequency with pretreatment of K2P16-i7 (1 hour), there was no difference in plateau fraction compared to VHL pretreated islet clusters (Supplemental Fig. 4). The lack of response from TALK-1 KO islet clusters treated with K2P16-i7 confirmed TALK-1 specificity (Fig. 7, C and D). Furthermore, the reversibility of K2P16-i7 was assessed with Ca2+ imaging, which determined that perfusing out K2P16-i7 restores Ca2+ oscillations similar to that as before K2P16-i7 treatment; this indicates that it is a reversible TALK-1 inhibitor (Supplemental Fig. 5). K2P16-i7 increases β-cell Ca2+ in a TALK-1–dependent manner. (A) Representative Ca2+ oscillations from control islet clusters with glucose stimulation (16.7 mM) and K2P16-i7 addition. (B) Quantification of control islet cluster Ca2+ plateau fraction before and after addition of K2P16-i7 (16.7 mM glucose 46.53% ± 1.97%; K2P16-i7 64.67% ± 3.86%; n = 4 mice; paired t test; ∗∗P < .01). (C) Representative Ca2+ oscillations from TALK-1 KO mouse islet clusters at 16.7 mM glucose and K2P16-i7 addition. (D) Quantification of TALK-1 KO islet Ca2+ plateau fraction before and after addition of K2P16-i7 (16.7 mM glucose 48.16% ± 1.51%; K2P16-i7 55.46% ± 4.07%; n = 3 mice; paired t test; P = .113; 95% CI, –4.28 to 18.88). We next assessed if the effects of K2P16-i7 on Ca2+ handling required both plasma membrane and ER membrane TALK-1 inhibition. This was accomplished by using a TALK-1 inhibitor, K2P16-i5 (VU0514204; cmpd 5 in Supplemental Fig. 2), which was found to only inhibit plasma membrane TALK-1 channels (Supplemental Fig. 2). Inhibition of plasma membrane TALK-1 with K2P16-i5 modestly increased glucose-stimulated islet Ca2+ plateau fraction (Supplemental Fig. 6). This suggested that inhibiting both plasma membrane and intracellular TALK-1 resulted in an additive increase in islet Ca2+. Future studies are required to determine how TALK-1 inhibition at the plasma membrane versus the ER membrane affects β-cell Ca2+ handling and health. K2P16-i7 promotes second-phase GSIS Because K2P16-i7 increased [Ca2+]ER release and glucose-stimulated islet Ca2+ oscillations, we next monitored insulin secretion with K2P16-i7. C57Bl/6J islets pretreated with K2P16-i7 or VHL underwent dynamic insulin secretion at 2 mM and 16.7 mM glucose (Fig. 8A). The fold change in GSIS AUC was increased with K2P16-i7 pretreatment compared to controls (Fig. 8B). Moreover, this elevation is largely due to a rise in second-phase insulin secretion, where oscillations in [Ca2+]C are observed. These results recapitulate the observed changes in second-phase insulin secretion with TALK-1-KO islets as well as β-cell–specific TALK-1-KO islets.11,25 K2P16-i7 enhances insulin secretion. (A) Dynamic perifusion insulin secretion traces from mouse islets pretreated with either VHL or K2P16-i7 switching from 2 mM to 16.7 mM glucose. (B) Quantification of the fold change in the AUC of insulin secretion at 16.7 mM glucose (total: VHL 1.00 ± 0.13; K2P16-i7 1.59 ± 0.074; first phase: VHL 1.00 ± 0.31; K2P16-i7 1.19 ± 0.11; second-phase: VHL 1.00 ± 0.10; K2P16-i7 1.68 ± 0.13. The first phase is considered as the first 10 minutes postexposure to high glucose. The second-phase is the remainder of the high glucose response. Statistical analysis performed using one-sample t test; ∗P < .05; n = 3 mice. Identification of structurally similar small molecule scaffolds for improved TALK-1 inhibitor potency Thirty-eight compounds with structural similarity to K2P16-i7 (cutoff Tanimoto coefficient >60%) were utilized to identify scaffold domains that improve potency and selectivity. Preliminary structure-activity relationship (SAR) studies on these molecules identified 2 compounds, K2P16-i7-4 and K2P16-i7-19, which displayed improved IC50 values compared to K2P16-i7 (IC50: K2P16-i7, 6.14 μM; K2P16-i7-4, 1.61 μM; K2P16-i7-19, 4.05 μM) (Fig. 9, A and B). The IC50 of K2P16-i7 was further evaluated by performing a CRC with K2P current recordings from the tetracycline-induced T-REx-TALK-1 cells (IC50 = 4.75 μM; Supplemental Fig. 7). Interestingly, K2P16-i7 showed ∼100% inhibition of TALK-1 currents, which differed from the Tl+ flux assay that found 58% inhibition, which is likely due to Tl+ assay normalization with niclosamide. Similar to the parent compound, K2P16-i7-4 and K2P16-i7-19 enhanced [Ca2+]ER storage in T-REx-TALK-1 cells (∼74% and ∼63% recovery when compared to uninduced cells, respectively; Fig. 9, C and D). We next assessed [Ca2+]ER leak in partially dispersed islets from C57Bl/6J mice pretreated for 1.5 hours with DMSO (VHL) or K2P16-i7-4. Islet pretreatment with K2P16-i7-4 resulted in greater CPA-induced elevations in [Ca2+]C compared with VHL control (AUC: VHL 10.30 ± 0.329; K2P16-i7-4 11.72 ± 0.511; Fig. 9, E and F). Identification of similar molecular scaffolds with improved potency to inhibit TALK-1. (A) Molecular structures of K2P16-i7-4 and K2P16-i7-19 with highlighted regions that are different from K2P16-i7. (B) 11-point CRCs of K2P16-i7 (IC50 = 6.512 μM; 95% CI, 2.96–10 μM), K2P16-i7-19 (IC50 = 4.054 μM: the model failed to generate a reliable confidence interval due to insufficient constraint of the fit), and K2P16-i7-4 (IC50 = 1.609 μM: the model failed to generate a reliable confidence interval due to insufficient constraint of the fit) (n = 6 wells/compound). (C) Average CPA-induced [Ca2+]C traces measured in T-REx-TALK-1 cells with or without tetracycline induction, pretreated with VHL, K2P16-i7-4, or K2P16-i7-19 at 10 μM concentration. (D) Quantification of the AUC for CPA responses from T-REx-TALK-1 cells treated as indicated. Statistical analysis performed using one-way ANOVA (with a Tukey’s multiple comparisons follow-up test) (VHL uninduced 127.5 ± 2.93; VHL induced 36.53 ± 2.99; K2P16-i7-4 83.00 ± 3.32; K2P16-i7-19 73.37 ± 4.44; mean ± SD; n = 11 technical replicates). (E) Representative CPA-induced [Ca2+]C release from the ER measured in dispersed islet clusters from C57Bl/6J mice pretreated with VHL or 10 μM K2P16-i7-4. (F) Quantification of the CPA-induced area under the Ca2+ curve in islet cells pretreated with K2P16-i7-4 or VHL (VHL 10.30 ± 0.329; K2P16-i7-4 11.72 ± 0.511; islet clusters from 2 independent islet preps: n = 50 VHL, n = 54 K2P16-i7-4). Statistical analysis was performed using an unpaired t test. Discussion The lack of TALK-1 pharmacology has limited its functional assessment and evaluation as a target for treating diabetes. Here, we report, to our knowledge, the first potent and selective small molecule inhibitors of TALK-1 K+ channels. Importantly, the TALK-1-selective compound, K2P16-i7, increased islet plateau fraction, elevated [Ca2+]ER stores, and enhanced second-phase GSIS. As excitotoxicity limits the durability of KATP inhibitors for treating diabetes, TALK-1 inhibitors may provide a gentler approach to improving β-cell function in T2D. Moreover, by elevating [Ca2+]ER, TALK-1 inhibition could protect β-cells from ER stress and dysfunction during T2D pathogenesis. β-cell [Ca2+]ER levels are key in maintaining ER homeostasis by regulating the protein folding activity of Ca2+-dependent ER chaperones and reducing ER stress.18 Therefore, TALK-1 activity negatively impacts islet function by increasing [Ca2+]ER leak, and this may be exacerbated under stressful conditions associated with diabetes.17 Indeed, global deletion of TALK-1 protects islets from HFD-induced ER stress, likely through the observed increase in [Ca2+]ER storage in TALK-1 KO islets.17 Furthermore, overactive TALK-1 channels are predicted to further diminish [Ca2+]ER stores. This is exemplified with decreased β-cell [Ca2+]ER and diabetogenic-induced ER stress with GOF SNPs and mutations in KCNK16.16,17 Because the TALK-1 inhibitors improve β-cell [Ca2+]ER storage, this would likely enhance chaperone function and reduce ER stress with elevated metabolic load. Although elevation in [Ca2+]ER by TALK-1 inhibitors is likely due to blocking ER-localized channels, future studies are required to confirm that cell-permeant inhibitors block TALK-1 channels in the ER membrane.17 A subset of TALK-1 inhibitors inhibited plasma membrane currents but did not show an effect on [Ca2+]ER storage, which may indicate plasmalemma impermeability. One of these plasma membrane–specific TALK-1 inhibitors, K2P16-i5, increased islet plateau fraction. This supports that plasma membrane TALK-1 channels alone can increase glucose-stimulated Ca2+ entry. Importantly, a greater difference in plateau fraction was observed in islet clusters treated with K2P16-i7, which demonstrates the contribution of ER-localized TALK-1 channels to islet Ca2+ handling. Further assessment of these cell-impermeable TALK-1 inhibitors will provide a useful tool in defining the roles of β-cell plasma membrane and ER membrane TALK-1 channels. Interestingly, acute K2P16-i7 treatment elevated islet Ca2+ plateau fraction, whereas chronic treatment increased islet Ca2+ oscillation frequency, which phenocopied TALK-1-KO islet responses.11,25 Interestingly, there was no change in plateau fraction following chronic treatment with K2P16-i7, but they did show increased oscillation frequency. This is likely due to K2P16-i7–induced increase in the frequency of silent phases, but with resulting equivalent silent phase duration as VHL-treated islet clusters. The difference in islet Ca2+ response with acute and chronic K2P16-i7 is likely due to adaptation of islets to the high initial Ca2+ resulting from TALK-1 inhibition or ablation. This is supported by a previous study showing that sulfonylurea-mediated plasma membrane depolarization following KATP inhibition initially causes constant islet Ca2+ influx followed by Ca2+ oscillations 30 minutes after treatment.30 Moreover, KATP KO islets also show Ca2+ oscillations, suggesting that prolonged inhibition or ablation of KATP leads to adaptation and restoration of Ca2+ oscillations.31,32 Future studies are required to determine how islets adapt to TALK-1 inhibition or ablation. This study specifically identified TALK-1 homodimer inhibitors; however, TALK-1 can also form heterodimers with other K2P subunits. For example, TALK-1 and TALK-2 form functional heterodimers in a heterologous expression system as well as in human pancreatic β-cells.33 Thus, because TALK-2 is highly expressed in β-cells and associated with increased risk for developing T2D, it will be important to assess if any of the 81 selective TALK-1 homodimer inhibitors can also inhibit TALK-1/TALK-2 heterodimers. Furthermore, the 23 molecules that inhibited both homodimeric TALK-1 and TALK-2 should be assessed for their impact on TALK-1/TALK-2 heterodimers. These dual TALK-1 and TALK-2 inhibitors will also provide useful tools in understanding TALK channel function in human islets, and whether dual inhibitors are more efficacious in augmenting β-cell function and preventing dysfunction. Moreover, future SAR studies on these scaffolds may be useful in identifying the first TALK-2–specific inhibitors. Crystallographic studies will be key in identifying the binding sites for TALK-1 inhibitors and performing extensive SAR studies to optimize the identified scaffolds. TALK-1 channel activity is associated with fasting hyperglycemia upon exposure to HFD feeding, implicating a non-β-cell–specific role of TALK-1. Indeed, mice carrying the MODY-associated GOF mutation in TALK-1 (L114P) exhibit fasting hyperglycemia as well as glucose intolerance on a chow diet.12 The observed fasting hyperglycemia results from elevated glucagon secretion, potentially due to loss of somatostatin secretion in TALK-1 L114P mice. This is supported by the fact that TALK-1 protein is functionally expressed in β-cells and δ-cells, but not in α-cells, and an increase in somatostatin secretion is observed in islets from TALK-1 KO mice.21 The identified TALK-1 inhibitors can, thus, not only be used to enhance GSIS but also reduce hyperglucagonemia observed in T2D by increasing somatostatin secretion under diabetic conditions. A nonislet-specific role of TALK-1 is supported by TALK-1-KO mice being protected from HFD-induced weight gain,34 which we observed as well. This is likely due to TALK-1 regulation of somatostatin-producing D-cells in the upper fundus and duodenum, where it is also expressed.35,36 Whether the identified small molecules inhibit gut TALK-1 channels needs to be tested; however, the possibility exists that TALK-1 inhibitors could be potentially utilized for reducing weight gain in the treatment of obesity. In conclusion, we identified, to the best of our knowledge, the first small molecule inhibitors of TALK-1 channels with optimal potency and efficacy utilizing an HTS approach. The identified compounds are TALK-1 selective and do not modulate the activity of 5 other K+ channels tested. K2P16-i7 increases [Ca2+]ER stores, increases glucose-stimulated plateau fraction, and enhances GSIS. Our data suggest that TALK-1 inhibition could support therapeutic strategies that improve ER function, reduce hyperglucagonemia, and limit obesity in T2D patients. Conflict of interest David A. Jacobson has a patent pending to Vanderbilt University. C. David Weaver reports that equipment, drugs, or supplies were provided by Vanderbilt University School of Medicine. Given his role as co-author, C. David Weaver is an owner of ION Biosciences and WaveFront Biosciences, the makers of Thallos thallium-sensitive dye and the Panoptic plate reader used for thallium flux experiments, respectively. All other authors declare no conflicts of interest. Acknowledgments We would like to thank Vanderbilt’s High-Throughput Screening Core. Figure schematics were created using BioRender.com and PowerPoint. Financial support This work was funded by National Institute for Health grants [DK129340-S1, DK129340, DK136768, and DK097392], training grants [1T32GM139800-01, 5T32DK07563], and by a grant [G-1912-03550] from the Leona M. and Harry B. Helmsley Charitable Trust. Data availability The authors declare that all the data supporting the findings of this study are available within the Paper, Supplemental Material, and are also available upon reasonable request. CRediT authorship contribution statement Arya Y. Nakhe: Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Validation, Visualization, Writing – original draft. Jordyn R. Dobson: Data curation, Formal analysis, Investigation, Methodology, Validation, Visualization, Writing – original draft, Writing – review and editing. Prasanna K. Dadi: Investigation, Methodology, Resources. David Westover: Investigation, Methodology, Writing – review and editing. Emily L. Days: Conceptualization, Data curation, Investigation, Methodology, Validation, Writing – review and editing. Paige Vinson: Conceptualization, Data curation, Methodology. C. David Weaver: Methodology, Writing – review and editing. Jerod S. Denton: Methodology, Writing – review and editing. Spencer J. Peachee: Data curation, Investigation, Writing – review and editing. Soma Behera: Investigation, Methodology, Writing – review and editing. Shannon E. Gibson: Investigation, Writing – review and editing. Valentina A. Rodriguez Da Silva: Data curation, Investigation. Amit Sela: Data curation, Investigation. Matthew T. Dickerson: Writing – review and editing. David A. Jacobson: Conceptualization, Funding acquisition, Methodology, Project Administration, Supervision, Writing – original draft, writing – review and editing. Footnotes A.Y.N. and J.R.D. contributed equally to this work. This article has supplemental material available at molpharm.aspetjournals.org. Supplemental Material Supplementary Figures 1-7 References Associated Data Supplementary Materials Supplementary Figures 1-7 Data Availability Statement The authors declare that all the data supporting the findings of this study are available within the Paper, Supplemental Material, and are also available upon reasonable request.