Novel cell-free high-throughput screening method for pharmacological tools targeting K⁺ channels

Assay Description.

The principle of the assay is illustrated in Fig. 1B. Purified K⁺ channels are reconstituted into lipid vesicles in the presence of KCl ranging in concentration from 150 to 300 mM. The channel-containing vesicles are usually frozen for storage at this stage. To assay, thawed vesicles are diluted into a NaCl solution, which creates a strong gradient for the efflux of K⁺. Potassium efflux is initiated by the addition of the H⁺ ionophore carbonyl cyanide m-chlorophenylhydrazone (CCCP), which allows influx of H⁺ to counterbalance the efflux of K⁺. The H⁺ influx is monitored by the H⁺-dependent quenching of a fluorescent dye, 9-amino-6-chloro-2-methoxyacridine (ACMA). In the example shown in Fig. 1C, K⁺ efflux leads to dye quenching when K⁺ channels are active (green trace), but not when they are inhibited (red trace). The K⁺ ionophore valinomycin is finally added to allow K⁺ channel-independent efflux. For many K⁺ channels, assay conditions under which the channels are inactive can be found to screen for activators that will initiate K⁺ efflux. Fig. 1D illustrates the sequence of additions used in a multiwell plate format and Fig. 1C shows an example of the data readout.

GIRK2.

GIRK2 is a K⁺ channel that mediates neural inhibition in response to stimulation of Gα_i-coupled GPCRs. On neurotransmitter stimulation, G protein subunits (Gβγ) are released from the receptor, bind to GIRK2, and cause it to open (Fig. 2A). The open K⁺ channel drives the cell membrane voltage toward the K⁺ reversal potential and thus inhibits cellular excitation. In addition to Gβγ, GIRK2 activation also requires PIP₂, a signaling lipid, and activation is also modulated by Na⁺ (Fig. 2 A and B). Because GPCR stimulation and multiple intracellular ligands are required to activate GIRK2, cell-based assays will naturally identify many off-target effectors. LFA permits strong biochemical control and is more directed at the channel rather than upstream components of the signaling pathway. Fig. 2 illustrates the biochemical control. When reconstituted in vesicles, GIRK2 channels will incorporate in both directions, inside out and outside out (Fig. 2B). By applying Gβγ subunits and a soluble version of PIP₂ outside the vesicles, only inside-out channels become activated. LFA replicates electrophysiological data showing that both Gβγ and PIP₂ are required to robustly activate GIRK2 (Fig. 2C). The rate of optical signal response can be controlled to an extent through the GIRK2 protein-to-lipid ratio of vesicles (Fig. S1A). This level of fine control enables the use of rate (i.e., slope of the optical signal) as a measure of relative channel activity during inhibitor or activator titrations. Furthermore, titrations of natural ligands such as Gβγ and PIP₂ can be performed with LFA (Fig. S1 B and C).

To test the ability of LFA to identify small molecules that modify GIRK2 activity, we screened a 100,000 member compound library under two different conditions. First, in the presence of Gβγ and PIP₂ we sought to identify inhibitors. Second, in the absence of Gβγ, we sought to identify compounds that could activate GIRK2 by possibly short-circuiting the G protein pathway. Four inhibitors and one activator were identified (Fig. 2D and Figs. S2 and S3).

The first inhibitor, RU-GIRK-1, is shown to inhibit GIRK2-mediated flux with an IC50 of about 0.35 μM (Fig. 2D). The activity of RU-GIRK-1 was confirmed in an electrophysiological assay in which GIRK2 channels were expressed in cultured cells (Fig. 2E). The specificity of this compound for GIRK2 among several K⁺ channels, including several other members of the K_ir channel subfamily, was also examined (Fig. 2F). The compound is also active in a more native context. RU-GIRK-1 inhibits baclofen-stimulated K⁺ currents in cultured mouse hippocampal neurons. The compound thus directly confirms the importance of GIRK2 to GPCR-mediated firing of these neurons (Fig. 2G).

Three additional inhibitors were identified and characterized (Fig. S2 A, C, and E). Graphs of the slope of the optical signal (relative to the slope in the absence of inhibitor) as a function of compound concentration generated inhibition curves that follow a rectangular hyperbola form. Each of these compounds also inhibited GIRK2 channels in an electrophysiological assay (Fig. S2 B, D, F, and G). LFA was also used to assess specificity of these compounds for the GIRK2 channel (Fig. S2H).

In the absence of Gβγ, LFA identified the antiparasitic agent ivermectin as an apparently G protein-independent activator of GIRK2 (Fig. S3A). This activity was confirmed in a whole-cell electrophysiology assay (Fig. S3 B–D). Although a cell assay alone could not exclude the possibility that ivermectin enhances the action of small quantities of Gβγ, LFA was carried out in the complete absence of Gβγ. Thus, ivermectin truly appears to short-circuit the G protein signaling pathway.

G protein-independent activation of GIRK2 by ivermectin is interesting because this drug is used in the treatment of parasitic infections, and its major side effect is CNS depression (24, 25). The mechanism of depression is thought to occur through activation of mammalian GABA and glycine receptor Cl⁻ channels (24, 26–28). However, direct activation of GIRK2 channels in the nervous system could equally contribute to this side effect.

TRAAK.

K2P channels are the most extreme structural outliers in the K⁺ channel family. Instead of containing four subunits they contain only two, but each has an internal repeat with two K⁺ channel signature sequences, providing the required complement of four such sequences to build a K⁺ selectivity filter (Figs. 1A and 3A). K2P channels as a subfamily are the least understood in terms of their biological function. This lack of understanding is due in large part to the paucity of specific K2P channel inhibitors. Our laboratory is working to understand in particular a K2P channel called TRAAK, a mechanosensitive K⁺ channel for which no inhibitors have been identified. We identified conditions under which TRAAK yields a robust LFA signal (Fig. S4A). We screened 300,000 compounds and identified two inhibitors: RU-TRAAK-1 (Fig. 3B) and RU-TRAAK-2 (Fig. S4B). These compounds also inhibit TRAAK in electrophysiological assays (Fig. 3 C–E and Fig. S4 C and D). RU-TRAAK-1 is poorly reversible, whereas RU-TRAAK-2 is completely reversible. Neither compound inhibits non-K2P channels tested (Kv1.2, Slo1 and GIRK2), but both show some cross-reactivity with certain other members of the K2P subfamily (Fig. 3E and Fig. S4D). TRAAK’s relative resistance to inhibition (only two inhibitors of 300,000 compounds tested) might reflect its unusual structure. Compared with most other eukaryotic K⁺ channels, TRAAK is relatively small and does not have structured regulatory domains outside the membrane that would create opportunity for binding small inhibitory molecules. We anticipate that these inhibitors, applied in well-controlled experiments, will offer new information on the biological roles of TRAAK channels.

High Conductance Ca²⁺-Activated K⁺ Channel.

High conductance Ca²⁺-activated K⁺ channel, here referred to as Slo1, regulates smooth muscle tone and several other cellular processes related to feedback inhibition by intracellular Ca²⁺. Human mutations that cause asthma and mouse genetic and pharmacological studies suggest that Slo1 activators might provide benefit in the treatment of asthma, hypertension and over-reactive bladder (9–15). We thus applied LFA to the Slo1 channel to identify molecules that would open the channel. A control experiment shows that the natural intracellular ligand, Ca²⁺, opens the channel and induces K⁺ flux in LFA (Fig. 4A). A channel opener (i.e., activator) screen was run at subthreshold Ca²⁺ concentrations (1.0 mM EDTA); 300,000 compounds were assayed and seven activators belonging to five distinct chemical categories were identified (Fig. 4B and Fig. S5 A, C, E, G, I, and K). One of these—RU-Slo1-2—was previously identified as a Slo1 activator (29). Each of these compounds were further examined in an electrophysiology assay for their ability to activate Slo1 at very low Ca²⁺ concentrations (Fig. 4 C and D and Fig. S5 B, D, F, H, J, and L). Because Slo1 is also a voltage-dependent K⁺ channel, it can open at low Ca²⁺ concentrations as long as that the membrane is sufficiently depolarized (i.e., voltage on the inside is made very positive relative to outside; Fig. 4C). We found that all activators promote channel opening at less depolarizing voltages (Fig. 4 C and D and Fig. S5 B, D, F, H, J, and L). Ca²⁺ is known to have precisely this effect (30). Therefore, the activators appear to mimic the effect of Ca²⁺ on the Slo1 channel. Such compounds have the potential to induce smooth muscle relaxation by opening Slo1 channels and hyperpolarizing the membrane.

hERG.

The hERG channel is present in cardiac cells where it participates in repolarization of the action potential. When hERG function is impeded, a characteristic lengthening of the Q-T interval is detectable on the electrocardiogram of mammals. This phenomenon is associated in humans with a lethal arrhythmia called torsade de pointes. For a reason that is not well understood, the hERG channel is highly susceptible to block by many structurally diverse small molecules. Our ability to predict which small molecules will block hERG is poor and therefore all compounds under consideration for drug development are tested in a hERG activity assay. Several assays exist: cell-based functional assays, such as thallium and/or rubidium flux, nonfunctional assays, such as ligand binding and/or fluorescence polarization, and electrophysiology (22, 23, 31, 32). The gold standard is electrophysiology, but it can be slow and expensive when the need exists for evaluating many compounds.

We applied LFA to the hERG channel. We first had to solve how to produce sufficient quantities of a stable version of this human K⁺ channel because the native, full-length channel aggregates when purified. Through trial and error we discovered that if we modified the amino acid sequence by deleting two different intracellular loops we could purify a biochemically well-behaved hERG channel with good function and correct pharmacology (Fig. 5). The behavior in LFA of two well-known hERG channel blockers, dofetilide and astemizole, is shown (Fig. 5 C–E). These drugs inhibit flux as predicted on the basis of electrophysiology assays (31, 33).

To test the sensitivity of an LFA-based hERG safety assay, we examined 50 known inhibitors at a concentration of 10⁻⁵ M and obtained zero false negatives (Fig. S6 and Table S1). A complete dose–response relation was determined for all 50 hERG channel blockers (Fig. S6 and Table S1). A plot showing IC50 values determined using LFA against IC50 values using electrophysiology reported in the literature shows that LFA correlates well with electrophysiology-determined values (Fig. 5F). We note that the IC50 determined by LFA for hERG blockers is, on average, approximately 10 times higher (i.e., lower affinity) than the IC50 determined by electrophysiology. The origin of this offset is due mainly to the following property of the LFA assay: K⁺ efflux is coupled to H⁺ influx and for some channels such as hERG the latter is rate limiting until K⁺ channels are mostly inhibited. Therefore, IC50 in LFA corresponds to a percentage of channel inhibition much higher than 50%, which leads to the IC50 offset observed (Fig. 5 F and G. Also see Methods for a detailed explanation). Thus, we consider LFA a semiquantitative high throughput assay for rapid identification of hERG blockers. Once an inhibitor is identified, the IC50 can then be confirmed using an electrophysiology assay.

To test the assay’s specificity we examined 50 known noninhibitors (34) at the same concentration and obtained a zero false-positive rate (Fig. S7 and Table S2). On repeat at 10 times higher concentration (approaching the solubility of many drugs), less than 25% inhibition was observed compared with a DMSO control (Table S2).

The data shown in this study were carried out using 384-well plates; however, an example using a 1,536-well plate with a similarly high signal demonstrates the ease with which the assay can be scaled (Fig. S8).

Protein Purifications.

Mouse GIRK2 was expressed and purified from Pichia pastori yeast cells. Human Gβγ complex was expressed and purified from High Five insect cells. Mouse TRAAK was expressed and purified from P. pastori yeast cells. Human Slo1 was expressed and purified from Sf9 (Spodoptera frugiperda) insect cells. hERG was expressed and purified from HEK293S GnTI⁻ mammalian cells.

Proteoliposome Reconstitution.

Purified proteins were reconstituted into POPE (1-palmitoyl-2-oleoyl-sn-glycero-3-phosphoethanolamine) and POPG [1-palmitoyl-2-oleoyl-sn-glycero-3-phospho-(1'-rac-glycerol)] vesicles [with a ratio of 3:1 (wt:wt)] by a dialysis method.

High-Throughput LFA Screens.

A 300,000 compound library was constructed from Enamine, Chembridge, and Rockefeller chemical library sources. The primary screen was carried out with a FDSS6000 (Hamamatsu) plate reader with automatic pipetting in 384-well format. The FDSS6000 pipetted and mixed 12 µL compound solution, 6 µL ACMA solution, and 6 µL vesicle solution into an assay plate, and the baseline fluorescence was recorded. Then, 6 µL CCCP was added and mixed to initiate the K⁺ flux, and the fluorescent signal was monitored every 5 s for 55 cycles. The final concentrations for the components in the LFA reaction were 1 µM compound, 1.3 µM ACMA, and 3.2 µM CCCP.

Hit Confirmation Using LFA.

Once the primary screens were finished and primary hits were selected using offline data analysis, the hits were confirmed using a fluorescent plate reader (Tecan, Infinite M1000 with excitation wavelength 410 nm and emission wavelength 490 nm) with manual pipetting following the same protocol as the primary screen in 384-well plates. Confirmed hits were purchased freshly from the same vendors that had supplied the primary screening compounds and drug titrations were conducted. The flux data were normalized, plotted, and fitted as detailed in SI Methods.

Animals.

All animal protocols were approved by the IACUC (Institutional Animal Care and Use Committee) of The Rockefeller University. Newborn C57BL/6J mice (The Jackson Laboratory) were killed for the dissection of hippocampal neurons.

Hippocampal Neuron Cultures.

Hippocampus tissues were dissected from P0 neonatal mice in cold Hank's balanced salt solution without Ca²⁺ and Mg²⁺. Dissected tissues were digested by trypsin, gently triturated, and plated on poly-d-lysine– and laminin-coated coverslips. Neurons matured after 1 wk after plating and were patched between 1 and 2 wk.

Plasmid Transfection.

HEK293 or CHO cells were cultured to 80% confluency and transfected with Lipofectamine 2000. Cells were treated with trypsin and plated out on coverslips 12–24 h after transfection and patched 6–48 h after plating.

Electrophysiology.

Different channel constructs were expressed in HEK293 or CHO cells and patched using standard whole-cell or excised configuration.

Data Analysis.

All data are presented as mean ± SEM.

Constructs.

DNA fragments encoding different constructs were amplified by PCR using Phusion DNA polymerase (New England Biolabs) and cloned into expression vectors modified by the MacKinnon laboratory for yeast, insect, and mammalian cells. All constructs were confirmed by sequencing (Genewiz).

Constructs for Protein Purification.

Mouse GIRK2 (mGIRK2) truncation mutant (residues 52–380) was cloned into RML1 vector and was expressed in Pichia pastoris strain SMD1163 (Invitrogen). RML1 vector is modified from pPICZ (Invitrogen) by adding a PreScission protease recognition sequence (LEVLFQ/GP) followed by an EGFP sequence and then a 1D4 tag sequence (TETSQVAPA) to the C terminus of cloned genes.

Human Gβ1 and Gγ2 were individually cloned into pFastbac vector (Invitrogen), and the Gβγ complex was expressed in High Five insect cells (Invitrogen) by virus coinfection.

Mouse TRAAK (mTRAAK) truncation mutant (residues 1–275 with N81Q and N84Q glycosylation mutations) was fused to an N-terminal 27 residue peptide from human TRAAK to increase expression. Chimeric mouse TRAAK construct was cloned into RML2 vector and was expressed in Pichia cells. RML2 vector is modified from pPICZ by adding a PreScission protease recognition sequence followed by an EGFP sequence and then a His10 tag sequence to the C terminus of cloned genes.

Human Slo1 (hSlo1) was cloned into RML11 vector and was expressed in Sf9 (Spodoptera frugiperda) cells (ATCC). RML11 vector is modified from pFastBac vector by adding a PreScission protease recognition sequence followed by an EGFP sequence and then a 1D4 tag sequence to the C terminus of cloned genes.

hERG truncation mutant with internal deletions of unstructured cytoplasmic loops (residues 141–380 and 871–1005) was cloned into RML13 BacMam vector and was expressed in HEK293S GnTI⁻ cells (ATCC). RML13 vector is modified from pEG (39) (a generous gift from Eric Gouaux, The Vollum Institute, Portland, OR), by adding a PreScission protease recognition sequence followed by an EGFP sequence and then a 1D4 tag sequence to the C terminus of cloned genes.

Three of the above constructs (for GIRK2, TRAAK, and hERG) represent channels that have been modified from their full-length form to gain biochemical stability. Although it is possible that disordered loops could form small molecule binding sites, the function of GIRK2 (response to Na⁺, PIP₂, and Gβγ, as well as to toxin inhibitors) and TRAAK (mechanical sensitivity) are indistinguishable from the full-length channels. For the case of the hERG channel, the N-terminal truncation causes increased rates of activation and deactivation. However, the correlation between drug sensitivity between published electrophysiological assays using the full-length channel and the LFA assay using the truncated channel lead us to believe that the truncated channel reports drug sensitivity with good fidelity.

Constructs for Electrophysiology.

mGIRK2 truncation mutant (residues 1–414, for reference the full-length protein is 1–425) was cloned into RML6 vector and was expressed in HEK293 cells (ATCC). RML6 vector is a mammalian cell expression vector adding a thrombin protease recognition sequence (LVPRGS) followed by an EGFP sequence and then a His8 tag sequence to the C terminus of cloned genes.

Rat Kv1.2 was cloned into RML5 vector and was expressed in CHO (ATCC) cells. RML5 vector is a mammalian cell expression vector adding a His8 tag sequence followed by an EGFP sequence then a thrombin protease recognition sequence to the N terminus of cloned genes.

hSlo1 was cloned into RML6 vector and was expressed in CHO cells.

mTRAAK chimeric construct used in protein purification was cloned into RML6 vector and was expressed in CHO cells.

Human ROMK truncation mutant (residues 35–367), mouse Kir2.1 truncation mutant (residues 41–368), and human Kir7.1 truncation mutant (residues 13–360) were cloned into RML6 vector and were expressed in HEK293 cells.

Mouse THIK1, mouse TRESK, rat TWIK2, mouse TALK1, and mouse TASK3 genes were cloned into RML6 vector and were expressed in CHO cells.

Protein Purification.

Proteoliposome Reconstitution.

Lipids in organic solvent were handled with glass pipettes and vials; 10 mg/mL POPE and POPG [1-palmitoyl-2-oleoyl-sn-glycero-3-phospho-(1'-rac-glycerol)] from Avanti in chloroform were mixed [with a ratio of 3:1 (wt:wt)] and aliquoted in 4 mL per glass vial (Kimble Chase; 16 × 125 mm); 1% brain PIP₂ lipid (Avanti) was added for the mGIRK2 inhibitor screen and the mGIRK2 PIP2-based activator screen. Lipid aliquots were air-dried under argon, washed with pentane, and dried again under argon in a laboratory chemical hood. Lipids were protected from light by wrapping with aluminum foil and protected from oxidation by topping with argon in all subsequent steps. Lipids were further dried overnight in a vacuum chamber with loose caps on the vials. Each dried aliquot was rehydrated in 2 mL reconstitution buffer (20 mM Hepes-KOH, pH7.4, 1 mM EDTA, 150 mM KCl, except for hERG 300 mM KCl were used). Lipid solutions were sonicated until translucent in a bath sonicator (Laboratory Supplies; model G112SP1T) with multiple cycles of 1 min on and 1 min off to prevent overheating. Sonicated lipid solutions were mixed with an equal volume of solubilization buffer (reconstitution buffer with 20 mM DM (for mGIRK2, hERG) or 160 mM n-octyl-β-d-maltopyranoside [OM; Anatrace; for the other proteins) and 20 mM DTT (no DTT for mTRAAK)] and incubated for 30 min by rotating. Solubilized lipid-detergent solution was mixed with purified protein at desired ratios and incubated for 1 h at room temperature by rotation. After incubation, protein-lipid-detergent solution was immediately transferred to 50 kDa MWCO dialysis bags (Spectrum Labs) in cold reconstitution buffer with fresh 3 mM DTT (no DTT for mTRAAK) and dialyzed at 4 °C (4 L buffer/20 mL protein-lipid-detergent solution). Reconstitution buffer was changed every 12 h for six cycles. After visible vesicle formation, bags were massaged manually each day to break up large lipid aggregates. Prewetted Bio-Beads SM2 (Bio-Rad, first wetted in methanol, subsequently washed five times with ddH₂O and one time with reconstitution buffer) were added to reconstitution buffer (5 mL/4 L buffer) for the last three cycles to remove residual detergents; 50-μL aliquots of the proteoliposomes were flash-frozen in liquid nitrogen and stored at −80 °C.

High-Throughput LFA Screens.

Explanation of the IC50 Offset in the hERG Assay.

In the LFA, efflux of K⁺ is coupled to ionophore-mediated influx of H⁺. Through analysis of efflux rates as a function of protein-to-lipid ratio with various K⁺ channels, we know that under many circumstances (specific channels such as hERG and at high protein-to-lipid ratios), when K⁺ channels are not blocked, the influx of H⁺ limits the efflux of K⁺. Consequently, reduced K⁺ efflux is observed only when K⁺ channels are sufficiently blocked to allow K⁺ efflux to become rate limiting. This effect causes an offset in the estimate of the IC50 of a K⁺ channel blocker. To demonstrate with a simple model how this occurs, we model the processes of K⁺ efflux and H⁺ influx as conductors in series: $G_K$ and $G_H$. Therefore, recorded conductance is

If a blocker at concentration x inhibits K⁺ conductance according to

where $G_{K0}$ is the K⁺ conductance in the absence of blocker, then

where K_d is the equilibrium dissociation constant for the blocker binding to its inhibitory site on the channel. Fig. 5G shows that a titration curve is shifted ∼10-fold when $G_{K0}$ = 10 $G_H$. By setting

we find that the IC50 for the LFA curve is

This model explains in simple terms why LFA can contain an offset in the estimated affinity of a blocker. As shown by the absence of false-negative and false-positive determinations in Tables S1 and S2, respectively, the offset does not render the assay inaccurate in its ability to identify channel blocking agents.

Animals.

All animal protocols were approved by IACUC of The Rockefeller University. Newborn C57BL/6J mice (The Jackson Laboratory) were killed for the dissection of hippocampal neurons.

Hippocampal Neuron Cultures.

Twelve-millimeter round poly-d-lysine–coated glass coverslips (neuVitro) were further coated with 10 μg/mL laminin (Invitrogen) in Dulbecco's PBS buffer (DPBS; Life Technologies) at 37 °C for 3 h in 24-well plates (Falcon). Before dissection, coated coverslips were washed four times with DPBS and incubated with plating medium (Neurobasal-A medium; Life Technologies) supplemented with 1× B27 supplement, 2% (vol/vol) horse serum, 2 mM GlutaMAX supplement, and 25 μM glutamate at 37 °C with moisture and 5% CO₂. Hippocampus tissues from P0 mice were quickly dissected out in cold Hank's balanced salt solution (HBSS; Life Technologies) without Ca²⁺ and Mg²⁺. Hippocampus tissues were digested with 0.25% trypsin (Sigma; 1:250 tryptic activity) at 37 °C for 15 min. Digested tissues were washed four times with warm HBSS without Ca²⁺ and Mg²⁺ and gently triturated by a P200 pipette. Viable cells were counted with Trypan Blue (Lonza) and plated on coated coverslips in plating medium at a density of 0.5–1 × 10⁵ cells per well. Four days after plating, plating medium was gradually changed to culturing medium (plating medium without glutamate). Culturing medium was renewed every 3 d. Neurons matured after 1 wk and were patched between 1 and 2 wk after plating.

Cell Line Cultures.

HEK293 cells were cultured in DMEM (Life Technologies) supplemented with 10% (vol/vol) FBS, 2 mM l-glutamine, and 1% penicillin/streptomycin in an incubator (Thermo Forma; Series II 3110) at 37 °C with moisture and 5% CO₂. CHO-K1 cells were cultured in DMEM: Nutrient Mixture F-12 (DMEM/F12; Life Technologies) supplemented with 10% (vol/vol) FBS, 2 mM l-glutamine, and 1% penicillin/streptomycin in an incubator at 37 °C with moisture and 5% CO₂. Cells were cultured to 80% confluency and transfected with Lipofectamine 2000 (Invitrogen) following the manufacturer’s protocol and passaged 12–24 h after transfection onto 12-mm poly-d-lysine– and laminin-coated glass coverslips (BioCoat) at low density. Patch-clamp recordings were performed 6–48 h after plating.

Electrophysiology.

All recordings were performed at room temperature. Pipettes of borosilicate glass (Sutter Instruments; BF150-86-10) were pulled to 2- to 4-MΩ resistance (except the pipettes used to patch BK were 1–2 MΩ) with a micropipette puller (Sutter Instruments; P-97) and polished with a microforge (Narishige; MF-83). Recordings were made with an Axopatch 200B amplifier (Molecular Devices) using standard whole-cell or excised patch-clamp techniques. Current-clamp experiments were performed on the I-fast mode. Recordings were acquired at 10 kHz and filtered at 1 kHz (Molecular Devices; 1440A). Pressure stimulation in inside-out patches for TRAAK was performed with a high-speed pressure clamp (ALA Scientific; HSPC) controlled through the Clampex 10 software (Molecular Devices). Pressure application velocity was set to the maximum rate of 8.3 mmHg/ms.

For whole-cell recordings of the mGIRK2 channel for inhibitor characterizations, the pipette solution was 10 mM Hepes-KOH, 20 mM NaCl, 130 mM KCl, 5 mM EGTA, and 2 mM MgCl₂, pH 7.4 (adjusted with KOH), and the bath solution was 10 mM Hepes-KOH, 150 mM KCl, 5 mM EGTA, and 2 mM MgCl₂, pH 7.4 (adjusted with KOH). For whole-cell recordings of the mGIRK2 channel with ivermectin, symmetrical pipette and bath solutions were 10 mM Hepes-KOH, 150 mM KCl, 5 mM EGTA, and 2 mM MgCl₂, pH 7.4 (adjusted with KOH). For whole-cell recordings of the Kv1.2, hK_ir1.1, mK_ir2.1, hK_ir7.1, or K2P channels, the pipette solution was 10 mM Hepes-KOH, 150 mM KCl, 5 mM EGTA, and 2 mM MgCl₂, pH 7.4 (adjusted with KOH), and the bath solution was 10mM Hepe-NaOH, 15 mM KCl, 140 mM NaCl, 2 mM MgCl₂, and 1 mM CaCl₂, pH 7.4 (adjusted with NaOH). For inside-out recordings of the mTRAAK channel, the pipette solution was 10 mM Hepe-NaOH, 15 mM KCl, 140 mM NaCl, 2 mM MgCl₂, and 1 mM CaCl₂, pH 7.4 (adjusted with NaOH), and the bath solution was 10 mM Hepes-KOH, 150 mM KCl, 5 mM EGTA, and 2 mM MgCl₂, pH 7.4 (adjusted with KOH). For whole-cell and inside-out recordings of the hSlo1 channel, symmetrical pipette and bath solutions were 10 mM Hepes-KOH, 150 mM KCl, 5 mM EGTA, and 2 mM MgCl₂, pH 7.4 (adjusted with KOH). For whole-cell hippocampal neuron recordings, the pipette solution was 10 mM Hepes-KOH, 150 mM K-gluconate, 5 mM NaCl, 5 mM EGTA, 2 mM MgCl₂, 4 mM Mg-ATP, 0.4 mM Na-GTP, and 10 mM phosphocreatine (Tris salt), pH 7.4 (adjusted with KOH), and the bath solution was 10 mM Hepes-NaOH, 140 mM NaCl, 5 mM KCl, 2 mM MgCl₂, 2 mM CaCl₂, and 10 mM glucose, pH 7.4 (adjusted with NaOH).

All drug perfusions were performed using home-made bath perfusion chambers except for the hippocampal neuron recordings, which used a fast pressurized microperfusion system (ALA Scientific; ALAVC3X8PP).

Data Analysis.

All data are presented as mean ± SEM.

Flux titration data were first normalized to eliminate baseline fluorescence fluctuations (due to ACMA pipetting variance and intrinsic fluorescence of testing compounds) using the following equation:

where F_normalized is the normalized fluorescence plotted in the flux titration figures, F is measured fluorescence in arbitrary units, F_start is the average of measured fluorescence before addition of CCCP, and F_end is the measured end point fluorescence after addition of valinomycin. Normalizations were performed with Excel (Microsoft). Plots were made using Prism software (GraphPad).

The initial slopes of flux after CCCP addition (the average slopes of first 10 s after CCCP addition) were extracted from normalized flux titration data. The slope values were normalized to the maximum slope to obtain normalized flux activity plotted in the dose–response curves of flux titrations. Of note, in an inhibitor titration, the maximum slope is the slope of DMSO control whereas in an activator titration the maximum slope is the slope at highest activator concentration tested. The dose–response curves of flux titrations were fitted with the Hill equation with OriginPro software (OriginLab)

where y is the normalized flux activity (the normalized initial slope), START is the normalized flux activity of DMSO control, END is the normalized flux activity at the highest drug concentration tested, x is drug concentration, and n is the Hill coefficient. The Hill coefficient ranged between 0.8 and 3.2.

Current-voltage curves of Slo1 recordings (Fig. 4D and Fig. S5) were fitted with the Boltzmann equation in Prism

where I is the measured current, I_bottom is the measure current at the lowest voltage, I_top is the measure current at the highest voltage, V is the command depolarization voltage to open the channels, V_mid is the command voltage at which the channels have reached halfway between I_bottom and I_top, F is the Faraday constant, R is the gas constant, T is the absolute temperature, and Z is the apparent valence of the voltage dependence.

The scatter plot of hERG positive control drugs (Fig. 5F) was fitted with linear regression in OriginPro (OriginLab).