Selective inhibition of the K_ir2 family of inward rectifier potassium channels by a small molecule probe: the discovery, SAR and pharmacological characterization of ML133

Abstract

The K_ir inward rectifying potassium channels have a broad tissue distribution and are implicated in a variety of functional roles. At least seven classes (K_ir1 – K_ir7) of structurally related inward rectifier potassium channels are known, and there are no selective small molecule tools to study their function. In an effort to develop selective K_ir2.1 inhibitors, we performed a high-throughput screen (HTS) of more than 300,000 small molecules within the MLPCN for modulators of K_ir2.1 function. Here we report one potent K_ir2.1 inhibitor, ML133, which inhibits K_ir2.1 with IC₅₀ of 1.8 μM at pH 7.4 and 290 nM at pH 8.5, but exhibits little selectivity against other members of Kir2.x family channels. However, ML133 has no effect on K_ir1.1 (IC₅₀ > 300 μM), and displays weak activity for K_ir4.1 (76 μM) and K_ir7.1 (33 μM), making ML133 the most selective small molecule inhibitor of the K_ir family reported to date. Due to the high homology within the K_ir family, the channels share a common design of a pore region flanked by two transmembrane domains, identification of site(s) critical for isoform specificity would be an important basis for future development of more specific and potent K_ir inhibitors. Using chimeric channels between K_ir2.1 and K_ir1.1 and site-directed mutagenesis, we have identified D172 and I176 within M2 segment of K_ir2.1 as molecular determinants critical for the potency of ML133 mediated inhibition. Double mutation of the corresponding residues of K_ir1.1 to those of K_ir2.1 (N171D and C175I) transplants ML133 inhibition to K_ir1.1. Together, the combination of a potent, K_ir2 family selective inhibitor and identification of molecular determinants for the specificity provides both a tool and a model system to enable further mechanistic studies of modulation of K_ir2 inward rectifier potassium channels.

INTRODUCTION

Potassium (K⁺) channels are well recognized as targets for treatment of cardiovascular, neurological, renal and metabolic disorders [1–5]. The inward rectifier potassium (K_ir or IRK) channels, a subset of potassium selective ion channels, have garnered a great deal of interest as potential therapeutic targets for multiple cardiovascular disorders as well as pain [6–10]. Inwardly rectifying K_ir channels allow positively charged K⁺ ions to more readily flow in the inward, rather than outward direction, and serve to regulate the resting membrane potential of cells. Structurally, K_ir channels are tetramers composed of four membrane-spanning α-helical subunits that surround a water-filled pore through which K⁺ ions selectively permeate (Figure 1). Each channel subunit is comprised of two α-helical membrane-spanning domains (M1 and M2) that are separated by an extracellular loop and a region that forms the K⁺-selectivity filter. K_ir channels also possess a helix bundle crossing motif at the membrane-cytoplasm interface and a large cytoplasmic domain. Seven subfamilies of K_ir channels are known (K_ir1–K_ir7) and most subfamilies have multiple members (Figure 1) with high sequence conservation in the pore forming domains [6–10]. K_ir channels are expressed in variety of tissues with distinct, but overlapping patterns. For example, both K_ir2.1 and K_ir4.1 are expressed in astrocytes and glial cells, where K_ir4.1 is associated with epilepsy in both causative and protective roles [11,12], but the role of K_ir2.1 in astrocytes and glial cells is not yet clear. K_ir1.1, K_ir2.3, K_ir4.1 and K_ir7.1 coexist and regulate K⁺ ion handling in the kidney tubules [6]. Pharmacological modulation with isoform specificity, or preference, would be of considerable utility in determining the functional roles of these channels, particularly in tissues where multiple isoforms or subtypes may be present. With the recent exception of MLPCN probes for K_ir1.1 (ROMK) [13,14], small molecule tools to study K_ir function are limited to nonselective neurologic and cardiovascular drugs (Figure 2) that happen to possess weak activity toward the K_ir channels [10, 15–21]; therefore, the pharmacology of specific channels and their druggability remain in question. However, in 2010, a patent application from Merck described a novel series of K_ir1.1 inhibitors, represented by 10, with potent K_ir1.1 inhibition (IC₅₀s <100 nM), but no selectivity data was reported [22].

Figure 1.

Structural model of an inward rectifier potassium channel based on the K_ir2.2 crystal structure, highlighting the tetrameric structure. Each K_ir channel subunit possesses two membrane-spanning α-helical domains (M1 and M2) that are separated by an extracellular loop which forms the K⁺ ion selectivity filter. The TMs then lead to the helix bundle crossing and the large cytoplasmic domain. Also shown is the K_ir family of ion channels highlighting both the gene and aliases commonly used in the literature.

Figure 2.

Structures and activities of representative K_ir small molecule inhibitors.

The Johns Hopkins Ion Channel Center and the Vanderbilt Specialized Chemistry Center were established as members of the Molecular Libraries Production Center Network (MLPCN) initiated and supported by the NIH Molecular Libraries Roadmap [23]. The MLPCN is a nationwide consortium of facilities that provide high-throughput small molecule screening and medicinal chemistry expertise for the development of chemical probes for use as tools to explore biological targets/pathways for which small molecule tools are unavailable. One such target which lacks the appropriate small molecule tools are the K_ir channels, the K_ir2.x subfamily and K_ir2.1, in particular [10]. The inward rectifying potassium channel K_ir2.1, also known as IRK1 or KCNJ2, is expressed in the heart, brain, vasculature and other tissues [6]. The K_ir2.1 channel mediates a small hyperpolarizing K⁺ current at negative membrane potentials, which plays a major role to establish resting membrane potential and also contributes to the terminal phase of action potential repolarization in diverse cell types [6]. Gain- or loss- of function mutations in the K_ir2.1 gene lead to cardiac, developmental and other pathologies, including Andersen-Tawil syndrome [24], short QT syndrome (SQT3) [25] and long QT syndrome (LQT7) [26]. Recently, viral introduction of K_ir2.1 channels has been used as a tool to study the effects of membrane potential on neuronal cell activity, sensory neuron modulation, as well as cardiac arrhythmias [27,28]. Thus, our MLPCN Centers initiated a campaign to identify and develop, utilizing high throughput Tl⁺ flux screening methods combined with automated and conventional patch clamp techniques, Kir2.x selective small molecule tools. This effort ultimately afforded ML133, a potent inhibitor of K_ir2.x family channels and the most selective K_ir2.x ligand reported to date.

RESULTS AND DISCUSSION

High Throughput Screening to Identify K_ir2.1 Lead Compounds

Development of isoform specific inward rectifier potassium channel modulators would be of considerable utility in determining their physiological roles and providing a rationale for developing novel therapeutics [6–10]. By combining a Tl⁺ flux assay, IonWorks assay, and conventional patch clamp experiments, we were able to identify novel K_ir2.1 and K_ir2.x family inhibitors. The MLSMR library containing 305,616 small-molecules was screened at a concentration of 10 μM employing a thalium (Tl⁺) flux assay in 384 well format in a HEK293 cell line stably expressing K_ir2.1 channels (see Materials and Methods). The assay performed well (Z′ = 0.74) and afforded 2,545 putative K_ir2.1 inhibitors, which decreased the Tl⁺ responses by more than 3 standard deviation (SD) values from the average of the DMSO control values. Confirmation screening in duplicate then provided 927 active compounds, which were further triaged by counter-screening against the parental HEK293 cell line eliminating another 426 compounds. The remaining 320 confirmed hits with were further counter-screened against hERG and KCNQ9 using Tl⁺ flux assays. Based on the chemical structures of the remaining hits, the K_ir2.1 potency and K⁺ ion channel selectivity, a single small molecule remained: CID17367817 (11), N-(4-methoxybenzyl)-1-(naphthalene-1-yl)methanamine (Figure 3A). Importantly, the PubChem database [29] indicated that 11 had been assayed in 218 MLPCN HTS campaigns, and was active in only two assays not based on K_ir2.1; thus, 11 had a relatively clean ancillary pharmacology and represented an attractive lead.

Figure 3.

Characterization of a novel K_ir2.1 inhibitor, 11. A) Molecular structure of 11. B) Representative traces for 11 inhibition of K_ir2.1 channel at 3μM. Current traces shown were recorded before (black trace) and after (red trace) applying 3μM 11 to the bath for 10 minutes. C) Representative time-course curve of K_ir2.1 inhibition by 3μM 11 at pH 8.5 (triangle), pH7.4 (circle), and pH6.5 (square). The time courses were representative results from 3–6 repeats. D) 11 dose-response curves for K_ir2.1 at pH 8.5 (triangle), pH 7.4 (circle) and pH6.5 (square). At pH 6.5, IC₅₀ is 10.0 μM and Hill coefficient (N_H) is 1.8; at pH 7.4, IC₅₀ is 1.9 μM and N_H is 1.7; at pH 8.5, IC₅₀ is 290 nM and N_H is 4.0; n=6–12.

Evaluation of Kir2.1 block in electrophysiological experiments

Blocking effects of 11 on Kir2.1 channels were examined in whole cell patch clamp recordings. As shown in Figure 3B, the voltage protocol consisted of a step to − 100 mV to record K_ir2.1 currents, and then a ramp from −100 mV to 100 mV to monitor the quality of the recording. The protocol was repeated every 10 seconds to determine changes of K_ir2.1 currents after compound application. Representative traces of the K_ir2.1 currents before and after 20 minute incubation with 3 μM 11 at pH 7.4 are also shown in Figure 3B. CID17367817 has two hydrophobic benzylic moieties bridged by a linker containing a basic 2° nitrogen, thus, 11 exists in protonated (charged) or non-protonated (neutral) forms, depending on the pH. We therefore examined the effects of pH on inhibition by 11 (predicted pK_a=8.79, from SciFinder) (Figure 3C & D). Increases in bath solution pH enhanced the potency of 11 in inhibiting K_ir2.1. As shown in Figure 3D, the IC₅₀ of 11 was 290 nM at pH 8.5, 1.8 μM at pH 7.4, and 10.0 μM at pH 6.5, respectively. Figure 3C shows representative time courses of 11 inhibition of K_ir2.1. At pH 8.5, 3 μM of 11 caused complete inhibition of K_ir2.1 within 5 minutes, while the wash out of compound effects took approximately 10 minutes; at pH 7.4, block by 3 μM required approximately 20 minutes to reach the steady state, and wash out of 11 effects took only 5 minutes; at pH 6.5, 3 μM of 11 had no significant effect on K_ir2.1 currents over the 20 minute incubation period. Importantly, K_ir2.1 channel activity itself was not affected by extracellular buffer of pH 6.5 or pH 8.5 (not shown).

Electrophysiological experiments in which block of K_ir2.1 by 11 was modified by permeant ions along with mapping studies (see below) suggest an intracellular site of action. In support of an intracellular site of action, 11 inhibited K_ir2.1 channels when directly applied from intracellular side in inside-out patch experiments: at pH 7.4, the IC₅₀ for 11 inhibition of K_ir2.1 was 4.3 μM, close to the IC₅₀ (1.9 μM) obtained in whole cell patch clamp recordings, in which 11 is added to the external side at the same pH. Additional inside-out patch clamp experiments suggest that the protonated form of 11 may inhibit K_ir2.1 channels. At pH 6.5, approximately 99% of 11 molecules would likely be protonated and 11 inhibited Kir2.1 currents more effectively than at pH 7.4 or pH 8.5, at which fewer molecules of 11 are protonated (pH 6.5, IC₅₀=2.6 μM; pH 7.4, IC₅₀=4.3 μM; pH 8.5, IC₅₀=12.5 μM). The pH-dependence of block by internally applied 11 was less pronounced, and was the inverse of the pH-dependence of extracellularly applied 11 as discussed above. These results suggest a mechanism in which extracellular pH regulates inhibition of K_ir2.1 by 11, possibly by changing the ratio of protonated and non-protonated forms of 11 and causing differential membrane permeation.

Small molecules that are more water-soluble when ionized and more lipid soluble when un-ionized. 11 would likely permeate cell membranes in the neutral form, which would be more prevalent at high pH values. As such, 11 would likely enter cells more readily in the de-protonated form and would then be re-protonated at the intracellular side (pH 7.4 maintained by HEPES in pipette solution) leading to intracellular accumulation. Some simple calculations support this explanation. At steady state, the concentrations of the non-protonated form of 11 should be equivalent inside and outside of the cell. The relative concentrations of protonated and unprotonated forms of 10 can be calculated based on the predicted pKa (8.79) of 11. At pH 6.5 on the external side, the ratio of total concentration of extracellular 11 versus intracellular 11 is (195+1):(24.5+1)=1:0.13. At pH 7.4, the intracellular and extracellular pH values are identical as will be total concentrations of 11. At external pH 8.5, the ratio of total concentration of extracellular 11 versus intracellular 11 is (1.9+1):(24.5+1)=1:9.09. This yields an expected 70-fold increase in intracellular concentration of 11 when extracellular pH is elevated from 6.5 to 8.5, which approximates the observed 34-fold change in IC₅₀ for channel block. These estimates of external pH effects on intracellular concentration of 11 can be used to calculate expected IC₅₀ values for channel block at different external pH values, assuming that extracellular pH changes do not directly change channel sensitivity to block by 11. Using the observed IC₅₀ of 1.8 μM at external pH 7.4 as a reference point, at external pH 6.5, the predicted IC₅₀ value of 14.6 μM agrees with the observed value (10.0 μM), and at external pH 8.5, the predicted IC₅₀ value of 0.21 μM again agrees with the observed value (0.29 μM).

To further validate our model, we set out to test compound 14k (Figure 5), a derivative with a different pKa due to methylation of the central nitrogen. The measured IC₅₀ of 14k is 31 μM at pH 7.4 and 9.2 μM at pH 8.5, a 3.37 fold decrease of IC₅₀ at pH 8.5 compared with pH 7.4. The predicted pKa of compound 14k is 7.93. Using calculations as described above, the expected intracellular 14k will increase 3.45-fold at pH 8.5 compared with pH 7.4, which can account for the 3.37-fold decrease of IC₅₀ over this pH range. Thus, this model is consistent with the observed pH-dependent changes in IC₅₀ values for block and the kinetics of inhibition, but other models may also be consistent with the present data. Thus, focus now shifted towards developing structure activity relationships (SAR) for 11.

Figure 5.

SAR and lead optimization summary for CID17367817, 11. A) SAR quickly established the necessity of the 4-OMebenzyl moiety, and all active analogs possessed this group; B) summary of observed SAR of over 50 analogs synthesized and evaluated examining all three regions of 11.

Chemical Lead Optimization

For the chemical lead optimization of CID17367817 (11), we divided the lead compound into three sections denoted western aryl, linker and eastern aryl (Figure 4A). First generation libraries held either the naphthyl moiety or the 4-methoxybenzyl moiety constant and surveyed alternative aryl moieties (2 × 20 members). The chemistry for these first generation libraries was straightforward [30] and involved a reductive amination sequence with an aryl aldehyde 12 and a benzyl amine 13 with NaB(OAc)₃H and conversion of the analogs 14 to the HCl salt prior to screening in manual patch clamp at pH 8.5 (Figure 4B). In addition, the original hit was re-synthesized as the HCl salt, and assigned CID781301 (Pubchem SID85281105/SID 87457855). As shown in Figure 5, the re-synthesized 11 confirmed with a K_ir2.1 IC₅₀ of 290 nM in manual patch clamp experiments. Clear SAR was established, and we quickly discovered that the 4-methoxybenzyl moiety was critical for K_ir2.1 activity. Libraries holding the naphthyl moiety constant and surveying alternate aryl groups for the 4-methoxybenzyl group lost all activity. However, modest K_ir2.1 activity (K_ir2.1 IC₅₀ <5 μM) could be retained when the 4-methoxybenzyl group was maintained as in 14a-k. The 4-chloro and 3-chlorobenzyl analogs (14a, CID 44483168; 14b, CID 44483172) were equipotent (K_ir2.1 IC₅₀ = 0.77 μM and 1.06 μM, respectively). The 4-tert-butylbenzyl analog (14d, CID 44483167) was comparable to 14b (K_ir2.1 IC₅₀ = 1.39 μM). Substitution in the 2-position, in the form of either a 2-methylbenzyl (14e, CID 44483176), 2,4-dichlorobenzyl (13c, CID 44483171), or 2-chlorobenzyl (14j, CID 44483175) analogs led to diminished potency, 5.15 μM, 3.89 μM, and 13.88 μM respectively. All other analogs led to either inactive compounds, or much reduced potency compared to re-synthesized 11 (CID 781301). Figure 5B provides an overview of the observed SAR. In addition to evaluating the western and eastern aryl groups, we also explored alternative linker moieties. Replacement of the NH with an oxygen led to a complete loss of K_ir2.1 activity as did replacement of either benzylic CH₂ with a carbonyl to provide the analogous amide congeners. Introduction of alkyl groups at the benzylic positions also led to a significant diminution in K_ir2.1 potency. Thus, the re-synthesized HTS hit 11 proved to be the the most potent K_ir2.1 inhibitor reported in this study, a rare instance where the HTS lead could not be improved upon. At this point, 11 was declared an MLPCN probe and assigned the probe identifier ML133.

Figure 4.

Design and chemical lead optimization strategy for CID17367817, 11/ML133. A) diversity-oriented modular approach to survey three regions of 11; B) general synthetic approach employed for the iterative library synthesis of analogs 14.

Selectivity of ML133 inhibition of K_ir family channels

To test whether ML133 exhibits selectivity over other inward rectifiers, we measured its effects on other K_ir2.x channels, including K_ir2.2 [41], K_ir2.3 and K_ir2.6, and all showed similar IC₅₀ values for ML133 block compared with K_ir2.1 (Table 1). In contrast, K_ir4.1 and K_ir7.1 were much less sensitive displaying higher IC₅₀ values, 76 μM and 33 μM respectively for ML133 block. Notably, K_ir1.1, also known as ROMK, is essentially insensitive to ML133 (Table 1). K_ir2.1 and K_ir1.1 are two highly conserved channel proteins with 54% similarity over the entire amino acid sequence and 67.7% within the transmembrane and pore domains (M1-P-M2) [10]. In addition to the relatively clean profile in over 218 MLPCN assays, ML133 was evaluated in Ricerca’s Lead Profiling Screen [32], which employs 68 radioligand binding assays across numerous GPCRs, ion channels and transporters. Overall, ML133 displayed clean ancillary pharmacology and was inactive at both L- and N-type calcium channels as well as several other potassium channels including hERG. Thus, ML133 is the most potent and selective K_ir2.x family selective small molecule reported to date.

Table 1.

The selectivity of ML133 inhibition and K_ir2.1 D172/I176 motif

Channel	Sequence alignment of M2 transmembrane domain																									IC50 (μM) pH 7.4
																172				176
mKir2.1	A	V	F	M	V	V	F	Q	S	I	V	G	C	I	I	D	A	F	I	I	G	A	V	M	A	1.9
rKir1.1	A	I	F	L	L	I	F	Q	S	I	L	G	V	I	I	N	S	F	M	C	G	A	L	L	A	>300
hKir2.2	A	I	F	M	V	V	A	Q	S	I	V	G	C	I	I	D	S	F	M	I	G	A	I	M	A	2.9
hKir2.3	A	V	I	A	V	V	V	Q	S	I	V	G	C	V	I	D	S	F	M	I	G	T	I	M	A	4.0
hKir2.6	A	V	F	M	V	V	A	Q	S	I	V	G	C	I	I	D	S	F	M	I	G	A	I	M	A	2.8
rKir4.1	A	I	E	L	L	I	A	Q	L	V	L	T	T	I	L	E	I	F	I	T	G	T	F	L	A	76.0
rKir6.2	A	I	L	I	L	I	V	Q	N	I	V	G	L	M	L	N	A	I	M	L	G	C	I	F	M	7.7
hKir7.1	A	I	A	L	L	A	I	Q	N	L	L	G	L	M	L	E	A	F	I	T	G	A	F	V	A	32.9

ML133 was also evaluated in our tier 1 in vitro DMPK battery to further establish its utility as a small molecule probe. In CYP₄₅₀ assays, ML133 was clean against 3A4 and 2C9 (IC₅₀ >30 μM), displayed moderate inhibition of 1A2 (IC₅₀ = 3.3 μM) but proved to be a potent inhibitor of 2D6 (IC₅₀ = 0.13 μM). In addition, ML133 was highly protein bound (>99%) in both human and rat and also displayed high intrinsic clearance in both species. Thus, the utility of ML133 is limited to in vitro studies of the K_ir2.x family.

Possible pore blocking mechanism for ML133

Ion channel pore blockers may display a phenomenon termed “knock-off”, where, in some cases, an intracellular pore blocker can be displaced from its binding site by inward K⁺ ion flux [13,14]. Typically, such an effect is more readily observed with a weaker blocker [21]. The interaction between M133 and wild-type K_ir2.1 may be too strong to be significantly influenced by these mechanisms in the current experimental settings (Figure 6). In order to further explore the mechanism of inhibition of this class of K_ir channel blockers, we took advantage of a bis-4-methoxybenzylamine analog of ML133, 14g (CID 44483166), with an IC₅₀ for K_ir2.1 block of 35 μM at pH 7.4. At 30 μM, 14g showed a dramatic “knock-off” profile in the wild-type K_ir2.1 channel (pH 7.4, 140 mM K_outside). The apparent knock-off of 14g by K⁺ ions diminished when the extracellular K⁺ was reduced to 4 mM (Figure 6D) in agreement with a mechanism in which blockers bind to a site in the permeation pathway and interact with permeant ions. In this experiment (Figure 6D), the membrane potential was stepped from −80 mV to −180 mV to maintain a similar absolute change in membrane potential and a similar driving force for K⁺ ions relative to the potassium reversal potential, although the channel conductance is decreased relative to 140 mM potassium bath solution. These data argue against a direct effect of membrane potential causing the time-dependent changes in channel block.

Figure 6.

ML133 blocked the ion conducting pathway of K_ir2.1 channel. A) 10 μM ML133 was not displaced by K⁺ influx from the wild-type K_ir2.1. Cell was bathed in a pH 7.4 solution containing 140 mM K_o ⁺. B) K⁺ influx (140 mM K_o ⁺) displaces 100 μM ML133 from the K_ir2.1 D172N/I176C mutant due to decreased compound affinity with channel. C) K⁺ influx displaces compound 30 μM 14g from the K_ir2.1 WT channel. D) Decrease of K_o ⁺ from 140 mM to 4 mM prevented 30 μM 14g from being displaced from K_ir2.1 WT by K⁺ influx. Note: Current traces shown above were recorded before (black trace) and after (grey trace) compounds (ML133 and 14g) reached maximal inhibition at pH 7.4.

Chimeric channels reveal critical domain(s) in K_ir2.1 mediating ML133 inhibition

To carefully examine the effect of ML133 on K_ir1.1, we tested ML133 up to 300 μM after a prolonged incubation of more than 10 minutes (Figure 7A). The currents with or without drug essentially overlapped. The leak current was negligible since barium inhibited over 85% of the K_ir1.1 current at the end of the recording. At pH 8.5, ML133 did show enhanced potency for K_ir1.1 block, with estimated IC₅₀ of 85.5 μM (Figure 7B).

Figure 7.

Chimeric channels reveal critical domain(s) in K_ir2.1 mediating ML133 inhibition. A) K_ir1.1 is not sensitive to ML133 inhibition. 30μM, 100μM, or 300 μM ML133 did not inhibit K_ir1.1 (n=3–5). Inset: K_ir1.1 currents were not inhibited by 300 μM ML133 at pH 7.4 (black trace), while 5 mM Ba²⁺ completely inhibited K_ir1.1 (grey trace) B) ML133 dose-response curves for K_ir1.1 at pH 7.4 and pH 8.5. At pH 7.4, K_ir1.1 is not inhibited by up to 300 μM ML133. At pH 8.5, ML133 inhibited K_ir1.1 with IC₅₀ of 85.5 μM. C) Chimeras between K_ir1.1 and K_ir2.1 exhibited differential sensitivity towards 30 μM ML133 at pH 7.4. Left panel: Chimeras were made by gradually displacing K_ir2.1 sequence by corresponding K_ir1.1 sequence as indicated. N-terminus: K_ir2.1 M1-L85; M1: K_ir2.1 V86-L109; MH: K_ir2.1 H110-T130; H5: K_ir2.1 A131-R148; HM: K_ir2.1 C149-A157; M2: K_ir2.1 V158-A181; C-terminus: K_ir2.1 K182-I428. Right panel: 30 μM ML133 inhibited over 80% of C8 and C108 current, but less than 20% of C25 and C13 currents.

In order to identify the molecular region(s) conferring high affinity inhibition of K_ir2.1, we evaluated a series of chimeric channels between K_ir2.1 and K_ir1.1 as shown in Figure 7C [33]. These chimeric channels are fully functional and display gating properties similar to that of K_ir2.1 or K_ir1.1 depending on their amino acid attribution. At a test concentration of 30 μM of ML133 (pH 7.4), both chimeras C8 and C108 were strongly inhibited (>90%), similar to wild-type K_ir2.1. However, chimeras C25 and C13 remained resistant to 30 μM ML133, comparable to that of wild-type K_ir1.1. These results suggest that M1 and/or M2 transmembrane domains contain the critical molecular determinant(s) for ML133 inhibition of K_ir2.1.

Identification of critical residues for ML133 inhibition of Kir2.1

There are two potential interpretations one may draw from results of chimeric studies. First, the M1 and/or M2 segments in K_ir2.1 contain key residues that confer the binding to ML133 and subsequent sensitivity to inhibition. Alternatively, The M1 and/or M2 regions of K_ir1.1, in fact, can interact with the compound, but have certain “toxic” residues/effects that prevent ML133 from causing inhibition. To address this, we first compared the amino acid sequences among all tested K_ir channels (Table 1). Recent advances in understanding of potassium channel structure, based on the 3.2 Å crystal structure of K_ir2.2, suggest that the M2 transmembrane domain forms the inner lining of the K_ir2.x ion conducting pathway [33]. Intriguingly, within M2 region there are several residues that appear to be conserved in correlation with the ML133 sensitivity. Considering that the inhibitory site appears to be accessible from the intracellular side, we began to focus our mutagenesis on those residues proximal to the cytoplasmic face of the pore domain. Sequence comparison analysis is consistent with a critical role for several residues, particularly in positions corresponding to D172/I176 in Kir2.1 channels (Table 1).

In order to evaluate their relevance, we converted these residues to K_ir1.1 residues in otherwise wild-type K_ir2.1 background (Figure 8). The Kir2.1 D172N single mutation greatly reduced the potency of ML133 inhibition, increasing the IC₅₀ to 15 μM, while K_ir2.1 D172N/I176C double mutation further reduced ML133 potency with an IC₅₀ of 45 μM. Conversely, if these residues are key mediators for the ML133 sensitivity, one would expect the changes of K_ir1.1 residues to those of K_ir2.1 would confer sensitivity to ML133. The C25 chimera has the transmembrane M1-Pore-M2 domain of K_ir1.1 but K_ir2.1 cytoplasmic domain, hence being used for these experiments. As shown in Figure 8C, with merely about 30% inhibition by 100 μM ML133, wild-type C25 chimera exhibited little sensitivity toward ML133 at pH 7.4. The C25 N172D single mutant shifted the IC₅₀ to 75 μM, suggesting K_ir2.1 D172 may be involved in ML133 inhibition. The C25 N172D/M175I/C176I triple mutant almost completely restored sensitivity to ML133 inhibition, with IC₅₀ value of 5.8 μM, which is comparable to the IC₅₀ of 1.9 μM for K_ir2.1. Since K_ir2.1 I175 varies among K_ir2.x family (K_ir2.2, K_ir2.3 and K_ir2.6 have the same methionine residual as K_ir1.1 and are sensitive to ML133), K_ir2.1 I176 rather than Kir2.1 I175 may serve as a more important determinant of ML133 binding. We therefore constructed N171D single or N171D/C175I double mutants in K_ir1.1 background in attempt to engineer the ML133 sensitivity (Figure 8D). Indeed, IC₅₀ of K_ir1.1 N171D was 83.6 μM and IC₅₀ of K_ir1.1 N171D/C175I double mutant was 64.0 μM, both of which are significantly sensitized compared to wild-type K_ir1.1. Therefore, these data suggests that K_ir2.1 D172 and I176 are necessary for ML133 inhibition, and can partially transfer ML133 sensitivity to insensitive K_ir1.1 channels.

Figure 8.

Identification of the critical residuals for ML133 inhibition of K_ir2.1. A) Structural layout of K_ir2.1 transmembrane domains, and the sequence alignment of K_ir2.1 and K_ir1.1 inner transmembrane domain (M2). Non-conserved residues (e.g. Kir2.1 D172 and I176) facing the central cavity were highlighted in grey. B) K_ir2.1 N172D (IC₅₀=15.4 μM, N_H=1.3, circle) and K_ir2.1 N172D/I176C (IC₅₀=43.6μM, N_H=1.8, square) mutant is less sensitive to ML133 compared with K_ir2.1 WT (IC₅₀= 1.9μM, N_H=1.7, triangle) K_ir2.1 wild-type. C) ML133 dose-dependent block of C25 and its mutants at pH 7.4. C25 is slightly inhibited at 100 μM (square). C25 N172D is more sensitive to ML133 (IC₅₀=75.2μM, N_H=2.2, circle). C25 N172D/M175I/I176C restored high-affinity ML133 block (IC₅₀=5.8μM, N_H=1.1, triangle). K_ir 2.1 wild-type IC₅₀=1.9μM (N_H=1.7, grey hollow circle).D. K_ir1.1 N171D (IC₅₀=83.6 μM, N_H=1.6, circle) and K_ir1.1 N171D/C175I (IC₅₀=64.0μM, N_H=2.3, triangle) partially restored sensitivity to ML133 compared with K_ir1.1 (square).

The side-chain groups corresponding to K_ir2.1 D172 and I176 in K_ir2.1 are positioned to face the central region of the pore in the crystal structure of K_ir2.2 (Figure 9A). The distance from K_ir2.1 D172 to I176 of the same subunit is estimated to be 6.24 Å based on the crystal structure of K_ir2.2 channel [33]. In comparison, the distance between the central protonable nitrogen to one of the aromatic rings is approximately 5.08 Å, sufficient for M133 to bridge K_ir2.1 D172 and I176 as shown in a homology model of K_ir2.1 constructed [34,35] from the K_ir2.2 crystal structure (Figure 9B). Based on the K_ir2.2 crystal structure [33], K_ir2.1 I176 likely forms hydrophobic seals beneath the central cavity to close off the pore section leading to the cytoplasm. This signature structure is unique for some eukaryotic ML133 sensitive K_ir channels such as K_ir2.1 (I176) and K_ir6.2 (L164). In contrast, K_ir channels lacking this motif (K_ir1.1, K_ir3, K_ir4.1, and K_ir7.1) displayed weaker block by ML133.

Figure 9.

Crystal structure of K_ir2.2 and our K_ir2.1 homology model. A) View of K_ir2.2 TM2 where ML133 is proposed to bind in K_ir2.1. B) Kir2.1 homology model based on crystal structure of K_ir2.2 with ML133 docked based on mutagenesis studies.

Structurally, the hydrophobic girdle at K_ir2.1 I176 is narrow enough to prevent the passage of water and hydrated K⁺ ions, suggesting a role in gating for K_ir2.1 I176. Functionally, K_ir2.1 I176C is readily inhibited by MTSET using substituted-cysteine accessibility method (SCAM). The SCAM inhibition kinetics at his site are the fastest among all pore lining residues [36], and far faster than for non-pore lining residues [37]. These results agree with structural data and indicate that K_ir2.1 I176 is accessible from cytoplasma side, and possibly having a role in Kir channel gating. Indeed, K_ir2.1 I176 analogous mutation in K_ir3.2 V188G (A) is constitutively active in the absence of active G protein subunits, suggesting the participation of K_ir3.2 V188 in channel gating [38]. Analogous mutation in K_ir6.2 (L164P) is also constitutively active and causes permanent neonatal diabetes mellitus (PNDM). The mutant channel can no longer be inhibited by intracellular ATP and displays altered SUR1 modulation [39]. These data support a role for K_ir2.1 I176 and analogous residues in K_ir channel gating.

ML133 can be used as a probe not only for K_ir2.1 but also for K_ir2.x family. It may be particular useful to parse out the physiological role of K_ir2.x current in distal tubulule of kidney, where K_ir2.3 are expressed at the basolateral side of the cell and Kir1.1 are expressed in the apical side of the cell. It has been speculated that the role of K_ir2.3 is to recycle K⁺ back into the cell in order to maintain the function of Na⁺/K⁺ pump, but this hypothesis has never been tested due to lack of probe differentiating K_ir2.3 from K_ir1.1.

Besides abundant expression in pancreatic cells, K_ir6.2 channel is only moderately expressed in the heart, skeletal muscle, and brain, together with different SUR subunits, the complex is also known as K_ATP channel. K_ir2.x expression is usually much higher in these tissues[6]. In addition, K_ATP channel is highly inhibited by endogenous ATP and is not constitutively active as K_ir2.x does, thus even when K_ATP channel co-expressed with K_ir2.x, they will probably remain silent and not contribute to the K⁺ conductance. Under certain circumstances when K_ATP channel activity become prominent, like in smooth muscle[6], one can use specific K_ATP channel inhibitors like glibenclamide (IC₅₀=3 nM) to abolish K_ATP channel current before using ML133 to study K_ir2.x function.

CONCLUSION

An HTS of more than 300,000 small molecules within the MLPCN was performed to identify modulators of K_ir2.1 function employing Tl⁺ flux assay. Here we report one potent K_ir2.1 inhibitor, ML133, which inhibits K_ir2.1 with IC₅₀ of 290 nM at pH 8.5 and 1.8 μM at pH 7.4, but exhibits little selectivity against other member of Kir2.x family channels. However, ML133 has no effect on K_ir1.1 (IC₅₀ > 300 μM), and displays weak activity for K_ir4.1 (76 μM) and K_ir7.1 (33 μM). In addition, ML133 was inactive in all but two of 213 MLPCN HTS assays not directed to K_ir channels as well as displaying a clean profile in a radioligand binding panel of 68 GPCRs, ion channels and transporters. Therefore, ML133 is the most selective small molecule inhibitor of the K_ir2.x family reported to date. Moreover, a multidimensional library effort failed to improve upon the potency of the HTS hit – a rare case in our experience. Detailed characterization, employing chimeras, mutagenesis and knock-off experiments, suggests a mechanism for inhibition of K_ir2.x channels in which ML133 binds to a site in the ion conducting pore formed by M2 transmembrane domains. These lines of evidence are consistent with a pore blocking mechanism for ML133 and suggest a putative interaction site. Based on the crystal structure of K_ir2.2, we constructed a K_ir2.1 homology model, and developed a manual docked hypothesis for energy minimized ML13 that is based on the mutagenesis studies. In addition, comparison of the K_ir2.1 model with a model for K_ir1.1 illustrate a more open conformation in the floor of the putative ML133 binding site. Further optimization efforts may employ these new models to guide ligand refinement.

MATERIALS/METHODS

Cell line, transfection reagent and chemicals

For conventional patch clamp experiments, the K_ir2.1 stable cell line and HEK 293 cell were cultured in DMEM/F12 medium, supplemented with 10% Fetal Bovine Serum (FBS), 50 IU/mL penicillin, 50 ug/mL streptomycin. Transient transfection on HEK293 cell was performed 48 hour before experiment using Lipofectamine LTX kit (Invitrogen, Carlsbad, CA). Usually, a total of 1 μg of DNA/3 μl Lipofectamine was used for transfection in one 35 mm dish. Compound ML133 was originally from Molecular Library Small Molecule Repository (MLSMR) and synthesized as a dry powder by the Vanderbilt Specialized Chemistry Center. Other compounds for SAR analysis were also provided by the Vanderbilt Specialized Chemistry Center.

Molecular Biology

Mouse K_ir2.1 pcDNA3 is as previously described [17]. Rat K_ir1.1 pEGFP-N3, human K_ir2.6 pcDNA3, rat K_ir4.1 in pCruz-myc vector and Human K_ir7.1 in pcDNA3 vector were gifts from Dr. Chou-Long Huang (University of Texas Southwestern Medical Center at Dallas). Chimeras between mK_ir2.1 and rK_ir1.1, namely C108, C8, C13, C25, were kindly made by Dr. Henry Sackin (Rosalind Franklin University of Medicine and Science, North Chicago, IL) [31]. The chimeras were originally in pSPORT vector. The inserts were then amplified by PCR and subcloned into pIRES2-EGFP vector by XhoI and PstI sites. Human K_ir2.2 (pCXN2 vector) and human K_ir2.3 (pcDNA3 vector) were gift from Dr. Anthony Collins (University of Belfast, Northern Ireland, UK). Rat K_ir6.2 was cloned into pcDNA1 vector. Hamster-SUR1 was cloned into pECE vector and co-transfected with K_ir6.2 during experiments. Point mutations were generated by site-directed mutagenesis (QuikChange kit; Stratagene) and confirmed by sequencing.

Conventional patch clamp

The electrodes were pulled from borosilicate glass capillaries (World Precision Instruments, Sarasota, FL). Pipette resistance was around 3–4 MΩ. Whole-cell currents of K_ir2.1 were recorded using an Axopatch 200B amplifier. The bath and pipette solutions contained 140 mM KCl, 2 mM MgCl₂, 2 mM CaCl₂, 10 mM HEPES (pH 7.4) and 140 mM KCl, 2 mM EDTA, 10 mM HEPES (pH 7.4), respectively. To apply the compound, 7–10 mL of compound solution was introduced by a 10 mL syringe into recording chamber containing 0.5 mL bath solution. Excessive solution was removed by suction. To record currents, voltage was stepped from 0 mV holding potential to −100 mV (500 ms) with an interval of 30 seconds. Capacitance compensation and 75% series resistance compensation were applied. Currents were filtered at 1 kHz, and data were acquired at 5 kHz with a Digidata 1322A computer interface and pClamp 9.2 software (Molecular Devices Corp.). To check the quality of seal during recording, each voltage step was followed by a ramp protocol (500 ms, from −100 mV to 100 mV). Cells losing seal during recording were identified by a sudden increase of outward current in the ramp protocol, and were not used.

Thallium-based fluorescence assay

Activity of potassium channels was monitored by the influx of thallium (Tl⁺), a surrogate ion for potassium. Thallium influx was detected through the use of a thallium-sensitive fluorescent dye, FluxOR (Invitrogen, Carlsbad, CA, Cat #F10017). Cells were seeded at 15,000 cells per well into BD Biocoat poly-D-Lysine coated 384-well-plates (BD, Cat# (35)6663 and Lot #8163495) using a Multidrop (Thermo scientific, Hudson, New Hampshire) and incubated overnight at 37 °C and 5% CO₂. The thallium-based fluorescence assay for K_ir2.1 was performed according to the manufacture’s recommended protocol. Briefly, medium was removed; cells were incubated with 1x FluxOR solution, 25 μL/well, for 90 minutes at room temperature (RT) in the dark; the 1x FluxOR solution was replaced by assay buffer (Hanks Balanced Salt Solution containing 5.8 mM potassium; product # 14065, Gibco, NY), 20 μL/well; test compounds in 7.5x stock of final concentration (10 μM) from a 10 mM stock supplied by the NIH Molecular Libraries Small Molecule Repository (BioFocus DPI, San Francisco, CA) or controls in assay buffer were then added to cells, 4 μL/well; 20 minutes later, cell plates were loaded to Hamamatsu FDSS 6000 kinetic imaging plate reader; after establishing fluorescence baseline by 1 Hz scanning for 10 seconds, Tl⁺ influx was triggered by addition of 6 μL/well of 5x stimulus buffer (containing 25 mM K₂SO₄ and 7 mM Tl₂SO₄) and fluorescence measurement was continued at 1 Hz for another 110 seconds. Controls wells contained assay buffer (IC₀), or IC₁₀₀ of chlorpromazine (all with DMSO concentrations matched to that of test compounds). The fluorescence ratio readouts were calculated by F(max−min)/F0, as well as Z prime factors from the average and standard deviation for negative and positive controls in each plate. Hit selection was based on the B scores [40] for test compounds calculated from the fluorescence ratios. If the B score of the test compound was less than minus 3 times the standard deviation (SD) of the B scores of ratios of the library compounds (<=-3*SD), and the B score of initial fluorescence intensity is within 2 times the standard deviation of the B scores of the library compounds, the compound is designated as an inhibitor/blocker of K_ir2.1 channels. Otherwise, it is designated as inactive.

Automated patch clamp

Automated patch-clamp experiments were performed using population patch clamp mode of automated voltage clamp recording with an IonWorks Quattro (Molecular Devices, Sunnyvale, CA). Briefly, compound effects on the K_ir2.1 channel were tested at 25 μM. The HEK-293 cells stably expressing K_ir2.1 channels were dislodged using 0.25 % Trypsin/EDTA and re-suspended at 2×10⁶ cells/mL in external buffer (140 mM K-gluconate, 1.8 mM CaCl₂, 1 mM MgCl₂, 10 mM HEPES, 10 mM Glucose, pH 7.4 or otherwise specified). Cells were then dispensed to the 384-well population patch clamp (PPC) plate. After dispensing, seal resistance of cells was measured for each well and cells were perforated on the internal solution side by incubation with 0.1 mg/mL amphotericin B in internal buffer (40 mM KCl, 100 mM K-gluconate, 1 mM MgCl₂, 5 mM HEPES, 2 mM CaCl₂ pH 7.2). Activity of K_ir2.1 was then measured with the following recording protocol: hold cells at +10 mV, step to −100 mV for 800 ms, back to +10 mV for 600 ms and run a ramp from −100 mV to +100 mV in 800 ms followed by a step back to +10 mV. Aliquots of 3.5 μl 3X compounds (final concentration 25 μM) prepared using external buffer, as well as controls with DMSO concentrations matched to that of test compounds, were dispensed to individual wells in the patch plate and incubated for 4 minutes. Currents were measured after compounds addition with the same protocol. Percentage of inhibition was calculated as (Ipre −Ipost)/Ipre*100% for each compound. Compared to negative controls, if a compound caused 3SD or more decrease in current amplitude, it was classified as an inhibitor.

Statistics

Manual electrophysiology data were analyzed using pClamp 9.2 followed by Origin 6 (OriginLab Corporation, Northampton, MA). The fluorescence data of the thallium-based flux assay were analyzed by manufacturer’s software package provided by Hamamatsu Photonics (Hamamatsu, Japan). Automated patch clamp data were analyzed in IonWorks 2.0.4.4 (Moledular Devices Corp., Sunnyvale, CA) and then exported for further analysis by Excel (Microsoft, CA) and Origin 6.

Structure-based Visualization and Putative ML133 Binding Site Models

Molecular models for the putative ML133 binding site for K_ir2.1 and K_ir1.1 tetramers were generated using the template structure of chicken K_ir2.2 (PDB_ID: 3JYC) and the sequence alignment of Tao, et al. [32]. Briefly, the backbone coordinates of the putative binding site region of the tetrameric K_ir2.2 structure were copied and side chains replaced for human K_ir2.1 and K_ir1.1 according to the alignment of Tao, et al. using the most probable amino acid rotamers for each residue from a backbone dependent rotamer library [37] as implemented in UCSF Chimera [38]. Low energy conformers for the ML133 probe molecule were generated using a stochastic search (10,000 cycles) of all rotatable bonds as implemented in MOE (Chemical Computing Group; Montreal, Canada). The lowest energy structure returned from this search was subjected to 500 steps of gradient energy minimization using the Merck force field (MMFF94) partial charges and then manually docked into a putative binding site (containing the identified mutations) just below the pore region of the K_ir2.1 model. The manually docked model was utilized only for visualization and comparison with the K_ir1.1 model to facilitate hypothesis generation for evaluation of potential intermolecular structure activity relationships. Molecular graphics were generated and rendered using the UCSF Chimera package from the Resource for Biocomputing, Visualization, and Informatics at the University of California, San Francisco (supported by NIH P41 RR001081).

Compound Synthesis and Characterization

General Experimental

All reagents were purchased from Sigma-Aldrich Corp., TCI America, and Rieke Metals, Inc. and were used without purification. All polymer supported reagents were purchased from Biotage, Inc. Analytical thin-layer chromatography (TLC) was performed on 250 μm silica gel plates from Sorbent Technologies. Visualization was accomplished via UV light, and/or the use of ninhydrin and potassium permanganate solutions followed by application of heat. Chromatography was performed using Silica Gel 60 (230–400 mesh) from Sorbent Technologies or Silica RediSep R_f flash columns on a CombiFlash R_f automated flash chromatography system. All ¹H and ¹³C NMR spectra were recorded on a Bruker AV-400 (400 MHz) instrument. Chemical shifts are reported in ppm relative to residual solvent peaks as an internal standard set to δ 7.26 and δ 77.16 (CDCl₃). Data are reported as follows: chemical shift, multiplicity (s = singlet, d = doublet, t =triplet, q = quartet, br = broad, dd=doublet of doublets, dq=doublet of quartets, td =triplet of doublets, pd = pentet of doublets, m = multiplet), coupling constant (Hz), integration. Low resolution mass spectra (LCMS) were obtained on an Agilent 1200 LCMS with electrospray ionization. High resolution mass spectra (HRMS) were recorded on a Waters Qtof-API-US plus Acquity system with ES as the ion source. Analytical high pressure liquid chromatography (HPLC) was performed on an Agilent 1200 analytical LCMS with UV detection at 214 nm and 254 nm along with ELSD detection.

General Library Synthesis of 14a–k

To a 4 mL vial was placed 0.5 mmol of amine, 0.5 mmol of aldehyde and diluted with 2 mL of DCE. MP-NaB(OAc)₃H (3 equiv. 1.5 mmol) was added and the vial placed on a rotator overnight. Vials were then filtered to remove the resin, concentrated and purified by mass-0directed preparative HLPC to analytical purity (>98% by 214 nm, 254 nm and ELSD).

ML133 (11). N-(4-methoxybenzyl)-1-(naphthalen-1-yl)methanamine (Figure 4B)

To a solution of (4-methoxyphenyl)methanamine (1.15 mL, 8.83 mmol) in DCE (25 mL) was added 1-naphthaldehyde (0.69 mL, 5.08 mmol). After 15 min, NaB(OAc)₃H (2.80 g, 13.23 mmol) was added portion wise over 5 min. After 12 h, water (25 mL) was added to the rxn and after another additional 1h, transferred to DCM:water (1:1; 250 mL). The organic layer was separated, washed with water (2 × 50 mL), and passed through a Phase separator. After concentration, the desired product was purified by preparative HPLC to afford an off-white solid (0.70 g, 50%): Analytical LCMS: single peak (214 nm), 1.123 min; ¹H NMR (400 MHz, d6-DMSO): δ 9.25 (br s, 1H), 8.07 (d, J = 8.0 Hz, 1H), 8.00 (d, J = 8.0 Hz, 2H), 7.68-7.48 (m, 6H), 7.01 (d, J = 8.4 Hz, 2H), 4.59 (s, 2H), 4.26 (s, 2H), 3.77 (s, 3H); HRMS, calculated for C₁₉H₂₀NO (M+ H+), 278.1545; found 278.1544. HCl Salt Formation: To a solution of the amine in DCM (0.2 M) at 0 °C was added 4 M HCl in 1,4-dioxane (5 equiv.) dropwise. After 15 min, the ice bath was removed. The solvent was removed after additional 30 min at RT to provide a pure HCl salt of the appropriate amine.

ABBREVIATIONS

K_ir

Inward rectifying potassium channel

HTS

High throughput screening

IC₅₀

Half maximal inhibitory concentration

N_H

Hill coefficient

SAR

Structure activity relationship

PNDM

Permanent neonatal diabetes mellitus