Next-generation inward rectifier potassium channel modulators: discovery and molecular pharmacology

Abstract

Inward rectifying potassium (Kir) channels play important roles in both excitable and nonexcitable cells of various organ systems and could represent valuable new drug targets for cardiovascular, metabolic, immune, and neurological diseases. In nonexcitable epithelial cells of the kidney tubule, for example, Kir1.1 (KCNJ1) and Kir4.1 (KCNJ10) are linked to sodium reabsorption in the thick ascending limb of Henle’s loop and distal convoluted tubule, respectively, and have been explored as novel-mechanism diuretic targets for managing hypertension and edema. G protein-coupled Kir channels (Kir3) channels expressed in the central nervous system are critical effectors of numerous signal transduction pathways underlying analgesia, addiction, and respiratory-depressive effects of opioids. The historical dearth of pharmacological tool compounds for exploring the therapeutic potential of Kir channels has led to a molecular target-based approach using high-throughput screen (HTS) of small-molecule libraries and medicinal chemistry to develop “next-generation” Kir channel modulators that are both potent and specific for their targets. In this article, we review recent efforts focused specifically on discovery and improvement of target-selective molecular probes. The reader is introduced to fluorescence-based thallium flux assays that have enabled much of this work and then provided with an overview of progress made toward developing modulators of Kir1.1 (VU590, VU591), Kir2.x (ML133), Kir3.X (ML297, GAT1508, GiGA1, VU059331), Kir4.1 (VU0134992), and Kir7.1 (ML418). We discuss what is known about the small molecules’ molecular mechanisms of action, in vitro and in vivo pharmacology, and then close with our view of what critical work remains to be done.

OVERVIEW OF Kir CHANNEL STRUCTURE-FUNCTION RELATIONSHIPS

Most Kir channels exist as tetrameric complexes comprised of either identical (homomeric) or similar (heteromeric) subunits arranged around a centrally located, water-filled pore that serves to conduct potassium ions down their electrochemical gradient. The electrical currents generated by Kir channels are used by cells to carry out important functions in various tissues and organs, making them attractive drug targets for treating certain diseases. The so-called K_ATP channels of the Kir6 subfamily form hetero-octameric complexes with sulfonylurea receptors (i.e., SUR1, SUR2A, SUR2B) that imbue sensitivity of the channels to intracellular nucleotides and magnesium. Given their importance as drug targets for treating type 2 diabetes, the physiology, pathophysiology, and pharmacology of K_ATP channels have been reviewed extensively elsewhere (1–4) and will not be discussed further here.

Kir channel structures are relatively simple compared to members of the voltage-gated potassium channel (Kv) superfamily (5, 6), and this has important implications for the pharmacology of the Kir channel family (discussed below). Each subunit contains only ∼1,000 amino acids, possesses only two transmembrane domains (as opposed to six for Kv channels), and lacks a charged S4 domain that confers voltage-dependent gating to Kv channels (Fig. 1A). The biophysical and namesake property termed “inward rectification” (or “anomalous rectification”) refers to the propensity of Kir channels under voltage-clamp conditions to pass ionic current in the inward direction more readily than in the outward direction (Fig. 1B) (8). The mechanism of limiting outward potassium current flow involves the voltage-dependent blockade of the channel pore by intracellular polyamines, such as spermine and putrescine, and the divalent cation, magnesium (9–17). These cationic molecules interact electrostatically with negatively charged residues lining the membrane-spanning pore. Mutating a single aspartate residue (i.e., D172) in the classical strong inward rectifier, Kir2.1, to an uncharged asparagine residue, changes Kir2.1 into a relatively weak rectifier (13), whereas mutating the equivalent asparagine (i.e., N171) residue into an aspartate residue, turns the weak rectifier, Kir1.1, into a strong rectifier (15). Although there are other residues lining the channel pore that contribute to the rectification properties of a channel, residues at this position seem to play a dominant role in channel rectification and are therefore often called the “rectification controller.” These residues play a key role in determining the small-molecule pharmacology of Kir1.1, Kir2.1, Kir3.1, Kir4.1, and Kir7.1 (18–23).

Figure 1.

Overview of inward rectifier potassium channel structure and function. A: side- (top) and top-down (bottom) views of a Kir2.2 crystal structure model (7). Each of the four subunits is shown with a different color. The top-down view highlights the central location of ion-conduction pore. B: current-to-voltage relationship of Kir2.1 (strong rectifier) and Kir1.1 (weak rectifier) current recorded between -120 mV and +120 mV in the whole cell configuration of the patch-clamp technique. The cartoon figures show the direction of potassium ion flow at voltages more negative (bottom left) or more positive (top) than the Nernst potential for potassium, before (top left) and after (top right) intracellular block by magnesium (blue sphere) present in the pipette solution. The voltage-dependent block of Kir2.1 and positive voltages is responsible for the observed inward rectification. The weaker pore block by magnesium is responsible for weaker rectification observed in Kir1.1.

Unlike most of the other members of the Kir channel family, Kir3 channels (Kir3.1–Kir3.4) are regulated directly through interactions with G proteins of G_i/o-coupled G protein-coupled receptors (GPCRs). It is through this regulation the Kir3 channels derive their more common name, G protein gated, inward rectifying potassium (K⁺) channels (GIRKs). Activation of the Gi_/o-coupled GPCR results in a conformational change in the interaction between GIRK α-subunits and the β/ɣ-subunits of the G protein complex, resulting in increased channel activity. A physiologically important example of this mechanism is the M2 muscarinic acetylcholine receptor-dependent modulation of heart rate. This theme is repeated throughout many types of excitable cells including neurons and hormone-secreting endocrine cells. The details of GPCR regulation of GIRK channels and the physiological and pharmacological consequences of such have been extensively reviewed elsewhere.(5, 24, 25) and will not be dealt with in detail in this review.

CLASSICAL PHARMACOLOGICAL TOOLS ARE WEAK AND NONSPECIFIC

Since Dr. Susumu Hagiwara first demonstrated that anomalously rectifying potassium currents in starfish eggs are inhibited by micromolar concentrations of barium (26), this divalent cation has become a favorite go-to blocker for studying Kir channels in diverse model systems. However, barium is poorly selectivity within the Kir channel family and has pleiotropic effects on numerous ion transport mechanisms (27–29). In this review, we focus on synthetic small molecules that were discovered or improved upon from efforts focused on the development of Kir-specific probes.

THE EVOLUTION OF HIGH-THROUGHPUT SCREENING PLATFORMS FOR POTASSIUM CHANNELS

Early efforts to develop pharmacological modulators of various types of ion channels was hampered by the lack of technologies that enable screening large libraries consisting of thousands or even millions of compounds. Indeed, manual voltage-clamp electrophysiology, still to this day considered the gold-standard method for measuring ion channel activity, is slow, labor intensive, and allows only one compound to be evaluated at a time. Radioligand displacement assays can be useful for high-throughput screening of drug-like chemical libraries, but the technique is limited in that it requires a radioligand and does not allow the measurement of a compound’s effect on channel activity. Some of the first efforts to develop activity-based high-throughput screening assays for potassium channels centered around measurements of radioactive rubidium flux through potassium channels (30). Although this technology, provided a medium-throughput method for screening thousands of compounds-per-day in 96-well plates, it was still too slow to keep pace with the rapidly growing chemical libraries in the late 1990s. Furthermore, the assay used a relatively strongly emitting radioisotope, requiring dedicated laboratory equipment, and they exhibited poor temporal resolution. In the late 1990s, a nonradioactive method of rubidium flux was developed using atomic absorption spectroscopy (31), but it was still limited in both throughput and temporal resolution.

The development of ion- or plasma membrane potential-sensitive, fluorescent sensors and the fluorometric imaging plate reader (FLIPR) (32), which is capable of measuring fluorescence from entire multiwell plates simultaneously, led to a revolution in target-based drug discovery for ion channels. Calcium-sensitive, fluorescent sensors like Fluo-3 and Fluo-4 transformed screening for modulators of GPCRs (33). Fluorescent, intracellular potassium sensors were developed (34) but were not widely used due to problems related to their low sensitivity and spectral properties. Plasma membrane potential-sensitive, fluorescent sensors were more widely used for measuring potassium channel activity (35); however, they also had considerable limitations. For example, the fact that these dyes detect changes in membrane potential resulting from ion flux and do not measure ion flux directly has disadvantages, which affect their utility. Thus, these assays are more susceptible to false negative and false positive effects. Although rubidium flux and plasma membrane potential sensor-based assays were both successfully used for medium- and high-throughput screening for potassium channel modulators, for many years there remained a gap between HTS-capable potassium channel assays and measurement of actual ion flux through potassium channels.

Approximately 15 years ago, the thallium flux technique was introduced (36) to enable a low-cost, fluorescence-based, nonradioactive, FLIPR-compatible assay capable of being scaled to use with the largest compound collections. Unlike plasma membrane potential sensitive sensor-based assays, the thallium-flux assay took advantage of the fact that thallium(I) readily passes through the pores of potassium channels and that fluorescent, thallium-sensitive sensors can be used to detect and quantify its influx into cells in a way analogous to calcium-sensitive fluorescent sensors. We will discuss efforts by us and other investigator to develop potent and specific “next-generation” Kir channel modulators using thallium flux-based HTS of compound libraries and medicinal chemistry for probe optimization.

NEXT-GENERATION Kir CHANNEL MODULATORS

VU590: The First Small-Molecule Inhibitor of Kir1.1 and Kir7.1

The first member of the Kir channel family to be cloned and functionally characterized was Kir1.1, which is encoded by the gene KCNJ1 on chromosome 11 (human) (37). Kir1.1 is commonly referred to as the renal outer medullary K (ROMK) channel due to its prominent and exclusive expression in renal tubules (38–40). The homomeric channel is expressed in the luminal membrane of the thick ascending limb (TAL) of Henle’s loop, the connecting tubule, and the cortical collecting duct (Fig. 2). In the TAL, Kir1.1 recycles potassium into the tubule lumen to energize the Na-K-2Cl co-transporter, NKCC2, which is critically involved in NaCl reabsorption and is the molecular target for the loop diuretics (41).

Figure 2.

Molecular physiology of ion transport in the nephron. A: sodium chloride reabsorption in the thick ascending limb of Henle is mediated by the luminal sodium-potassium-chloride co-transporter, NKCC2. Kir1.1 (ROMK) recycles potassium into the tubule lumen to energize NKCC2. Sodium and chloride exit the basolateral membrane via the sodium pump and ClC-Kb chloride channels, respectively. B: the sodium-chloride co-transporter (NCC) mediates sodium chloride reabsorption in the distal convoluted tubule. Basolateral Kir4.1/5.1. channels hyperpolarize the membrane potential and promote chloride exit through ClC-Kb, which in turn stimulate WNK-SPAK kinase-dependent phosphorylation and activity of NCC. C: in the cortical collecting duct, luminal Kir1.1 channels hyperpolarize the apical membrane potential and create a favorable electrochemical driving force for sodium reabsorption through epithelial sodium channels (ENaC). Basolateral Kir4.1/5.1 and Kir7.1 channels also hyperpolarize the membrane potential and recycle potassium across the membrane to maintain the activity of the sodium pump. NKCC2, sodium-potassium-2-chloride.

Kir1.1 mediates potassium secretion in the connecting tubule and cortical and medullary collecting duct (Fig. 2). In the connecting tubule and collecting duct, Kir1.1 is electrically coupled to a sodium reabsorption pathway mediated by the epithelial sodium channel (ENaC), in which sodium reabsorption depolarizes the luminal membrane potential and thereby increases the electrochemical driving force for Kir1.1-dependent potassium secretion (42).

Kir1.1’s dual role in regulating sodium reabsorption in the thick ascending limb and potassium secretion in the distal nephron led us and others in the pharmaceutical industry to postulate that Kir1.1 might represent a novel diuretic target whose pharmacological blockade could induce natriuresis and diuresis without evoking potassium wasting. This notion was supported by human genetic data from the Framingham cohort study, showing that heterozygous carriers of loss-of-function mutations in KCNJ1 have lower blood pressure and are protected from age-dependent increases in blood pressure (43). However, because the molecular pharmacology of Kir1.1 when we started work in this area consisted of the peptide toxin, Tertiapin-Q (44), the tools for critically evaluating the safety and efficacy of inhibiting Kir1.1 were unavailable. We therefore set out to develop these tools using a molecular target based HTS approach.

We screened ∼225,000 compounds from the NIH Molecular Libraries Screening Center Network (MLSCN) library for small-molecule modulators of Kir1.1 using a fluorescence-based thallium flux assay. The first inhibitor we published, termed VU590 (Fig. 3), was a hit from the primary screen that inhibits Kir1.1 with an IC₅₀ of 240 nM (45). Follow-up experiments revealed that VU590 had no effect on Kir2.1 and Kir4.1, but inhibited Kir7.1 with an IC₅₀ of 8 µM. Both Kir1.1 and Kir7.1 exhibit voltage- and potassium-dependent block in such a way that suggested VU590 is a pore blocker that accesses its binding site through the cytoplasmic pore of the channel. We therefore performed scanning mutagenesis of the membrane-spanning pore, patch-clamp analysis of putative binding mutants, and molecular modeling to identify the functional binding site for VU590 in Kir1.1 and Kir7.1 (19). The only pore-lining residue evaluated that participates in VU590 block of Kir1.1 was asparagine 171 (N171), the so-called “rectification controller” discussed above. Mutation of N171 to either aspartate (N171D) or glutamate (N171E) led to a greater than 70-fold reduction in VU590 sensitivity. Interestingly, mutation of N171 to the structurally related (i.e., one carbon-length difference) amino acid glutamine (N171Q) led to a twofold loss in VU590 sensitivity. This latter observation is consistent with a model in which the amino acid structure and not simply charge at position 171 is an important determinant VU590 block. Surprisingly, we also found that this site could not only accommodate a bulky tyrosine residue at this position, but that this substitution (N171Y) increased VU590 block. Altogether, these data argue that N171 represents a physical binding site for VU590 in the Kir1.1 channel pore (Fig. 4) (19).

Figure 3.

Summary of next generation Kir channel modulators. Kir, inward rectifier potassium. (Reference sources listed in rightmost column in numerical order: 20, 23, 45, 46, 47, 88, 93, 94, 112, 116, 143.)

Figure 4.

VU590 functional binding sites in Kir1.1 and Kir7.1. Site-directed mutagenesis, voltage-clamp electrophysiology, and molecular modeling were used to identify Kir1.1-N171 (A) and Kir7.1-E149 and A150 (B) as essential for optimal block of the channels. These studies also revealed that Kir7.1-T153 creates an energetic barrier to small-molecule inhibitors accessing their deeper binding site near E149 and A150. Kir, inward rectifier potassium. [Reproduced from Kharade et al. (19) with permission.]

We were curious whether VU590 blocked Kir7.1 through a similar molecular mechanism and reasoned that the glutamate residue (Kir7.1-E149) in the equivalent position of Kir1.1-N171 accounted for the lower sensitivity to VU590. We hypothesized that mutation of E149 to an uncharged residue such as asparagine or glutamine would enhance block by VU590. Unfortunately, mutation of E149 to asparagine (Kir7.1-E149N) did not produce functional channels; however, the E149Q mutation led to a dramatic loss, not gain, of VU590 sensitivity. Shortening the side chain length by one carbon while maintaining the negative charge with an aspartate substitution (E149D) led to a significant increase in VU590 sensitivity, suggesting as we observed in Kir1.1 that this site of the channel is a physical binding site for VU590. Consistent with this model, mutating the neighboring alanine residue to a serine (A150S) also led to an abrogation of VU590 sensitivity.

Extending our mutagenesis analysis toward the cytoplasmic end of the channel pore revealed that mutation of threonine 153 to cysteine (T153C) led to a significant increase in block by VU590. Because T153 is predicted from homology modeling to reside only 6 Å and 8 Å away from A150 and E149, respectively, and the molecular dimensions of a low-energy conformer of VU590 (5 Å × 6 Å × 8 Å) approximate these distances, we postulated that there might be a functional interaction between VU590 and the three residues. Indeed, we found that the T153C mutation restored full VU590 potency to Kir7.1 channels also carrying E149Q or A150S mutations, suggesting that in the presence of a compromised binding site, T153 creates an energetic barrier that exacerbates low-affinity block (19).

If this model is correct, we postulated that the T153 barrier might impact other small molecules that bind in the same location but with different affinities. To test this hypothesis, we turned to two structurally related Kir7.1 inhibitors termed VU714 (IC₅₀ = 1.5 µM) (Fig. 3) and ML418 (IC₅₀ = 310 nM) (Fig. 3), the latter of which requires both E149 and A150 for fully efficacious block. Consistent with the model, the T153C mutation had no effect on the higher-affinity blocker ML418 but led to a significant increase in block by VU714. Taken together, these studies revealed a previously unrecognized role of T153 located at the base of the Kir7.1 channel pore in determining the pharmacological properties of intracellular pore blockers. In future high-throughput library screens, employing the Kir7.1-T153C mutant as a surrogate for the wild-type channel might help identify lower-potency pore blockers without affecting the ability to identify higher-affinity blockers (19).

VU591: The First Selective Inhibitor of Kir1.1

Because VU590 exhibited off-target activity toward Kir7.1 and therefore could not be used to interrogate the therapeutic potential of Kir1.1 as a diuretic target, we set out to develop a more selective inhibitors of Kir1.1 that might be used in vivo. One hit from the primary screen was bis-nitro-benzimidazole (BNBI), which exhibited some structural similarities with VU590. Both compounds contain nitro groups flanking an intervening linker region containing oxygen and/or nitrogen atoms, which gave us confidence that BNBI was a bona fide inhibitor of Kir1.1. Unfortunately, BNBI inhibits Kir1.1 with only weak potency (IC₅₀ = 8 µM) compared to VU590. Importantly, however, BNBI had no effects on Kir7.1 at doses up to 100 µM. This indicated that although BNBI is a weak inhibitor of Kir1.1, it might be selective and serve as a good synthetic starting point for the development of more potent and specific analogs with medicinal chemistry. One key difference between BNBI and VU590 was the absence of an ether oxygen in the BNBI linker that might provide structural flexibility and/or polarity present in VU590 needed for high-affinity block. Strikingly, the addition of an ether oxygen in BNBI led to a 30-fold improvement in potency toward Kir1.1 (IC₅₀ = 240 nM) without affecting selectivity over Kir7.1. Evaluating the ancillary pharmacology of this BNBI analog in patch-clamp experiments or radioligand biding assays revealed that it is selective for Kir1.1 over more than 65 potential ion channels, transporters, and receptors. We called this inhibitor VU591 (46) (Fig. 3).

Like VU590, VU591 exhibited voltage- and potassium-dependent block of Kir1.1, suggesting it too is a pore blocker. An unbiased computational approach was used to identify potential binding sites for subsequent interrogation with site-directed mutagenesis and patch-clamp analysis (21). An ensemble of 10 models with distinct conformations was constructed using crystal structures that were available at the time. 99,000 docking simulations of 200 low-energy conformations of VU591 were performed starting from three different locations in the membrane pore. Cluster analysis of the best poses revealed two potential binding sites in the pore: A “lower” site situated near the pore-cytoplasmic domain interface, and an “upper” site located in the center of the membrane (Fig. 5A). Scanning mutagenesis revealed that no pore-facing residues in the lower site are required for potent block of Kir1.1 by VU591. However, mutation of valine 168 (V168) or N171 led to a dramatic loss of VU591 activity. Shortening the side chain length at position 168 to an alanine (V168A) led to a fivefold reduction in potency, whereas lengthening it with leucine (V168L) led to a greater than 168-fold loss of activity. Mutation of N171 to N171D or N171E led to a complete abrogation of VU591 activity, whereas the more conservative mutation, N17Q, led to only a fivefold loss of VU591 sensitivity. The double mutation, V168A/N171Q, led to an additive loss of channel sensitivity, suggesting both residues mediate VU591 block of the channel pore. Closer inspection of a low-energy binding pose of VU591 in the pore (Fig. 5B) raised two intriguing mechanistic possibilities, which are not mutually exclusive. First, the nitro groups flanking the VU591 linker, which were essential for high-affinity block by VU590, are positioned within 5 Å of both residues and could therefore mediate interaction with the channel pore. The second detail is that VU591 must assume a folded “horseshoe” shape to be accommodated at that binding site. It is possible that the ether oxygen that improved potency over BNBI might allow for conformational flexibility needed to bind with high affinity to V168 and N171 (Fig. 5B).

Figure 5.

Identification of the VU591 functional binding site in Kir1.1. A: homology modeling and in silico docking were used to identify energetically favorable VU591 docking sites in putative “upper” and “lower” pore locations that were subsequently tested with mutagenesis and patch-clamp electrophysiology. No mutations in the lower site alter VU591 block; however, mutation of V168 and N171 led to a loss of VU591 potency. B: close-up view of VU591 docked into the upper site interacting with V168 and N171. Kir, inward rectifier potassium. [Reproduced from Swale et al. (21) with permission from Elsevier.]

Merck Kir1.1 Inhibitor Compound A: Therapeutic Proof-of-Concept and Renal Mechanisms of Action

In 2012, the ion channel group at Merck began publishing their work to develop of structurally diverse Kir1.1 inhibitors for use as novel mechanism diuretics (47). Interestingly, the first potent Kir1.1 inhibitor disclosed was a symmetrical small molecule with a nitrogen-containing linker flanked by two nitro groups, two key properties that are shared by VU591. We subsequently confirmed that, like VU590 and VU591, an optimized analog of this Merck compound, termed Compound A (Fig. 3), requires N171 for high-affinity Kir1.1 block (unpublished observations). Importantly, the Merck group demonstrated for the first time that pharmacological inhibition of Kir1.1 induces natriuresis and diuresis without causing potassium wasting, lowers blood pressure, and protects the kidney from hypertension-induced kidney injury (48, 49).

In light of this seminal data, we attributed the failure of VU591 to induce renal excretion to its poor metabolic stability and high serum protein binding. We therefore decided to use Compound A to better understand the renal mechanisms by which Kir1.1 inhibition leads to a potassium-sparing diuresis. We administered Compound A either alone or in combination with diuretics that act on discrete segments of the nephron and evaluated the effects on urine volume and electrolyte composition (50). The NKCC2 inhibitor, bumetanide, was used as an inhibitor of sodium reabsorption in the thick ascending limb, whereas hydrochlorothiazide (HCTZ), an inhibitor of the sodium-chloride co-transporter (NCC), was used to inhibit sodium reabsorption in the distal convoluted tubule. The ENaC inhibitors, amiloride or benzamil, were used to inhibition sodium reabsorption in the collecting duct. We reasoned that an additive or synergistic effect of Compound A with a diuretic would be evidence of the two compounds inhibiting sodium reabsorption in different nephron segments, whereas a lack of synergy would suggest that both compounds act on the same segment. Briefly, we found a synergistic effect between Compound A and HCTZ, which was expected since Kir1.1 is not co-expressed with NCC in the distal convoluted tubule. However, we did not observe synergy between Compound A and bumetanide or amiloride/benzamil, suggesting that Compound A inhibits sodium reabsorption in the thick ascending limb and collecting duct. Interestingly, Compound A was able to block the kaliuresis induced by bumetanide, confirming that at least one source of potassium wasting with loop diuretic administration is unopposed Kir1.1-mediated potassium secretion in the thick ascending limb. Compound A also inhibited potassium wasting induced by HCTZ and the residual potassium secretion in the presence of amiloride. Thus, the combined inhibition of sodium reabsorption in the thick ascending limb and inhibition of potassium secretion in the TAL and collecting duct account for the potassium-sparing diuretic effects of Compound A (50).

ML133: A Selective Inhibitor of Kir2.x Channels

The second member of the inward rectifying channel family to be cloned, Kir2.1 (51), is encoded by the KCNJ2 gene. The structure, function, and regulation of the channel are reviewed in Refs. 5 and 52. The Kir2.1 channel is expressed in numerous tissues including macrophages (51), heart (where it underlies I_K1) (53), vasculature (54), microglia (55), and neurons (56, 57) and neutrophils (58). In these tissues, it plays a wide variety of roles including setting resting membrane potential (59), microglial patterning and development, signaling and migration (60), and smooth muscle proliferation and migration (61). Kir2.1 is known to co-assemble with a number of other members of the Kir2 channel family including Kir2.2, Kir2.3, and Kir2.4 (62, 63). Mutations in Kir2.1 channels have been linked to a variety of disorders including periodic paralysis, Anderson–Tawil, and short-QT syndromes (52, 64, 65). Kir2.1 has also been observed to be upregulated in small cell lung cancer where it may play a role in the development of drug resistance (66).

Although a number of compounds have been reported to be inhibitors (67–73) and in one case an activator of Kir2.x channels (74), they had all been previously described as modulators of other targets of a variety from target classes. However, ML133 (Fig. 3) was developed from a compound discovered through a Kir2.1-focused HTS (23) using a fluorescence-based thallium flux assay. The screen was funded through the Molecular Libraries Screening Center Network (MLSCN).

Initial characterization of ML113 revealed a potency in the low micromolar range. ML133 contains a basic nitrogen (pK_a ∼ 8.8), which led Wang et al. (23) to explore whether pH has any effect on ML133’s ability to inhibit Kir2.1 channels. They observed that ML133 potency was strongly affected by pH, whereas Kir2.1 activity was not sensitive to changes in pH over the range tested. At pH 8.5, ML133 had a measure potency of 290 nM, whereas its potency was 1.8 µM and 10 µM at pH 7.4 and pH 6.5, respectively. These observations together with additional data led the investigators to conclude that it is the neutral form of ML133 that penetrates the membrane and that increasing the extracellular pH results in concentration of ML133 inside the cell.

ML133 was reported to possess a clean ancillary pharmacology profile based primarily on data from a radioligand-binding panel but also specifically including a lack of activity and N- and L-type calcium channels as well as Kv11.1 (hERG). Among other members of the Kir family that were evaluated, ML133 was at least 10-fold more potent at inhibiting Kir2.1 (1.8 µM) compared to Kir1.1 (>300 µM), Kir4.1 (76 µM), and Kir7.1 (33 µM). However, ML113 does inhibit other members of the Kir2 family, Kir2.2, Kir2.3, and Kir2.6, with potencies similar for inhibition of Kir2.1, in the low uM range.

In order to investigate structural determinants of ML133 function and specificity, Wang et al. (23) used Kir1.1 and Kir2.1 chimeric channels to identify D172 and I176 as critical for Kir2.1 inhibition. It is intriguing to note that this critical aspartate residue is in the same position as the asparagine residue found to be critical for VU590 function, namely the “rectification controller.” This finding suggests that this region of Kir channels may be a privileged position for inhibitor function. Although VU590’s effect on ROMK activity required an asparagine at this position, ML133 showed preference for aspartate.

Unfortunately, ML133 was found to have very high plasma protein binding and high intrinsic clearance. Even though these properties limit its utility to that of an in vitro probe, ML133 has seen widespread usage to investigate the broad diversity of roles Kir2 family members play: microglia (60, 75, 76), regulation of neuronal process (57, 77), modulation of autophagy in endothelial progenitor cells (78), sour taste transduction (79), effects on protective effects or ischemic preconditioning in cardiac ventricle (80), and to help validate the expression of Kir2.1 channels in neutrophils (58) and colon (81). Further refinement of ML133 analogs might result in the development of compounds with properties suitable for in vivo administration further expanding our ability to probe the function of Kir2.1 and other members of the Kir2 family and that these studies might reveal potential therapeutic utility of Kir2 inhibition.

ML297, VU0810464, GAT1508, and GiGA1: Selective GIRK1-Containing GIRK Activators

Continuing the theme from Kir1.1 and Kir2.1, until recently the pharmacology of GIRK channels was very poorly developed (82). The molecules published were limited to mostly poorly selective inhibitors with NTC 801 (83), being a notable exception. NTC 801 is a discontinued clinical candidate for the treatment of atrial fibrillation. It shows good specificity for inhibiting GIRK channels with low nanomolar potency. It was demonstrated to be safe in humans but in a limited clinical trial, it did not demonstrate an ability to treat atrial fibrillation.

The GIRK activator landscape was even sparser than the inhibitor landscape. As described above, the activity of GIRK channels is regulated in a positive fashion by Gi/o-coupled GPCRs. GIRK channels are also known to be positively modulated by intracellular Na⁺. However, the only reported exogenous GIRK activators were ethanol (84) the natural product, naringin, which activates GIRK with very low potency (85), and some compounds of undisclosed structure (86).

From an HTS that used GIRK channel activity as a readout for GPCR activity (87), a number of compounds were discovered that activated GIRK channels in the absence of an active Gi/o-couple GPCR. The discovery of these molecules led to the development of ML297 (88) (Fig. 3) and subsequently to a large number of related compounds including closely structurally related GIRK channel inhibitors (89). ML297 exhibited a potency of ∼300 nM and ∼ 5-fold preference for activating heteromeric channels containing the GIRK1 and GIRK2 subunits compared to channels comprised of GIRK1 and GIRK4 or GIRK1 and GIRK3. Surprisingly, ML297 was only able to activate heteromeric GIRK channels, which contain Kir3.1 (GIRK 1). Since GIRK 1 does not form channels on its own, ML297 is able to activate GIRK channels comprised of Kir3.1+ either Kir3.2, Kir3.3, or Kir3.4. A mutagenesis study revealed that two amino acids in GIRK 1 are responsible for ML297’s ability to activate GIRK1 containing channels (90) (Fig. 6). Wydeven et al. (90) demonstrated that transfer of both F137 and D173 residues from GIRK1 to their homologous positions in GIRK2 results in ML297 being able to activate heteromeric GIRK2/GIRK2-double mutant channels. The F137 residue has been previously identified as important for channel gating (91) and critical for preventing GIRK1 from forming functional homomeric channels (92). Among all of the inward rectifier channels, GIRK1 is unique in that it is the only one which contains a phenylalanine at this position. All other Kir channels contain a serine at this position, except for Kir2.4 that contains an alanine. The other key residue is an aspartate, D173. In GIRK2, the analogous amino acid residue is an asparagine. In repetition with a theme previously established with Kir1.1 and Kir2 modulators, this “rectification controller” position appears to also be a “pharmacology controller” position for a number of Kir modulators. Although the effect of ML297 on GIRK1-containing channels has not been investigated in as much mechanistic detail as the effect of flecainide on Kir2.1 (74), like flecainide, ML297 increases the open probability and weakens the inward rectification of the channel.

Figure 6.

Identifications of amino acids critical for activation of GIRK channels by ML297. Shown in A are the aligned amino acid sequences of GIRK1 and GIRK2 highlighting the two positions in the sequence that were determined as critical for ML297 efficacy. Shown in B are the results of thallium flux assays from HEK-293 cells co-expressing wild-type GIRK2 along with either wild-type GIRK1, wild-type GIRK2, mutant GIRK1, or mutant GIRK2 channels. Mutation of either phenylalanine F137 or aspirate D173 in GIRK1 resulted in loss of ML297 efficacy. ML297 efficacy was only observed with co-expression of either wild-type GIRK2 along with wild-type GIRK1 or wild-type GIRK2 along with a GIRK2 mutant containing both residues, F148 and D184 (correspond to residues F137 and F173 found in GIRK1). Shown in C are the positions of GIRK1 F137 and D173 in a computer model of heteromeric GIRK1/GIRK2 based on the crystal structure of homomeric GIRK2. GIRK, G protein-coupled inward rectifier potassium. [Elements of this figure were derived from information previously published by Wydeven et al. (90).]

Although F137 and D173 are required for ML297’s ability to activate GIRK1-containing GIRK channels, it remains to be determined conclusively whether these amino acids interact directly with ML297. Computer modeling, mutagenesis, and medicinal chemistry studies have begun to shed more light on the role of these and other amino acid residues in activation of GIRK by ML297 and the closely structurally related compound, GAT1508 (93). These studies implicated additional residues as important for the preference for some compounds, like GAT1508 GIRK1/GIRK2 heteromers compared to GIRK1/GIRK4 heteromers.

Recently, Zhao et al. (94) reported the discovery of a new GIRK activator, GiGA1 (Fig. 3) discovered using computer modeling and virtual high-throughput screening. They found that, unlike other previously described compounds that require F137 and D173 in the GIRK1 subunit, GiGA1 was not sensitive to mutations of these residues. These data, the fact that the GiGA1 was discovered using a model of GIRK2 alcohol binding pocket, and that residues known to be important for the GIRK activating properties of alcohols also affect the ability of GiGA1 to activate GIRK channels, argue for ML297 and related compounds being functionally distinct from GiGA1.

Since its discovery, ML297 and other more recently discovered GIRK activators have been used by a number of groups to probe the function of GIRK channels in isolated cells, tissues, and in vivo. These studies include demonstration that pharmacologically increasing the activity of GIRK channels can produce efficacy rodent models of numerous disorders: anti-seizure (88, 94,95), anxiolytic (90, 96), anti-nociceptive (97, 98), facilitation of conditioned fear extinction (93), promotion of non-REM sleep (99), and rescue of amyloid-β-evoked deficits in hippocampal function (100, 101). These tools have also been used extensively to probe GIRK’s role in wide variety of systems (102–111). Continued work on these molecules has been focused on improving their drug-like properties and selectivity for the major neuronal GIRK heteromer, GIRK1/2 and away from the primary cardiac form, GIRK1/4: VU0810464 and GAT1508 (93, 96, 112) (Fig. 3).

Although largely overlooked, early medicinal chemistry studies on ML297 identified very closely structurally related compounds, which were GIRK inhibitors, not GIRK activators. Wen et al. (89) reported that the substitution of the methyl at the 3-pyridyl position in ML297 for isopropyl results in inhibition of GIRK channels, whereas the potency remains similar. In this study, the investigators observed that other substitutions at the 3-pyridyl position, but not other positions on the molecule resulted in GIRK inhibition. As is the case with the closely related GIRK channel activators, the efficacy of these ML297-like inhibitors requires the GIRK1 F137 and D173 residues, and thus, these compounds do not inhibit non-GIRK1-containing GIRK channels.

VU0529331 and Ivermectin Activators of Non-GIRK1-Containing GIRK Channels

As described above, ML297 and related compounds are unable to activate GIRK channels that do not contain the GIRK1 subunit. However, there are GIRK channels in the body which do not contain GIRK1 subunits. Although these non-GIRK-1 containing channels exhibit much more restricted expression than the GIRK1-containing GIRK channels, non-GIRK1-containing channels are expressed in some intriguing locations including the ventral tegmental area and the substantia nigra, reviewed in Ref. (24). Recently, compounds capable of activating non-GIRK1-containing GIRKs have been discovered. Using a compound library screening approaches, two groups of investigators discovered that the insecticide ivermectin is able to activate GIRK channels (113, 114). Unfortunately, the utility of this compound is limited because it is known to modulate the activity of a wide variety of ion channels include voltage-gated sodium channels, GABAa and glycine receptors (115). More recently, VU0529331 (Fig. 3) was identified using a thallium flux-based HTS focused specifically on discovering modulators capable of activating non-GIRK1-containing GIRKs (116). Although VU0529331 shows considerably better specificity for activating GIRK channels when compared to ivermectin, relatively low potency (∼ 5 µM) has limited its utility to isolated cell systems. Unfortunately, ivermectin and VU0529331 also activate GIRK1-containing GIRK channels, further limiting their utility as tools to selectively probe non-GIRK1-containing GIRK channels. Unsurprisingly, considering that VU0529331 and Ivermectin can activate GIRK1 and non-GIRK1-containing GIRKs, neither of their ability to activate GIRK channels depends on the presence of the F137 and D173 amino acid residues found in GIRK1 channels. In the future, improvements in VU0529331 may lead to compounds that are useful for probing the functions of non-GiRK1-containing GIRK channels in brain slices or potentially when precisely injected into specific brain regions, but until compounds capable which are selective only for non-GIRK1-containing GIRK channels are developed, the specific roles and therapeutic potential of pharmacological modulation of these channels remains out of reach. As a final note, Zhao et al. (94) report the identification of hits from a virtual screen that have some capacity to activate GIRK2 homomeric channels, but their structures were not revealed and their capacity to activated non-GIRK1-containing GIRKs was not described in detail.

VU0134992: The First Homomeric Kir4.1-Preferring Inhibitor

Kir4.1 is expressed in the kidney tubule epithelium as well as astrocytes and glial cells of the nervous system (91, 117–122). Kir4.1 can assemble into a homotetramer or heterotetramer with Kir5.1 (KCNJ16), with each subtype exhibiting unique functional, regulatory, and pharmacological properties (123–125). One of the best-known functions of homomeric Kir4.1 is its role in “spatial buffering” in the brain and eye, where the channel promotes the uptake of excess extracellular potassium into astroglial cells to maintain normal electrical activity of neurons (119, 126). Changes in the expression levels of Kir4.1 in astrocytes have been associated with treatment-resistant depression (127, 128) and epilepsy (129–131), suggesting Kir4.1 might underlie the etiology of these diseases. Consistent with its known roles in these organ systems, heritable loss-of-function mutations in KCNJ10 lead to SeSAME or EAST syndrome, a mixed neurological and kidney disease characterized by epilepsy, ataxia, sensorineural deafness, and renal salt wasting (132, 133).

The major subtype in the renal tubule is the heteromeric Kir4.1/5.1 form of channel, which is expressed on the basolateral membrane of the thick ascending limb of Henle, distal convoluted tubule, collecting duct (117, 134, 135). A great deal of progress has been made recently in understanding the physiology of Kir4.1/5.1 in the distal convoluted tubule, a tubule segment that plays a critical role in sensing plasma potassium and regulating sodium chloride reabsorption mediated by the thiazide-inhibitable sodium-chloride co-transporter NCC (Fig. 2). Genetic and crude pharmacology data support a model in which Kir4.1/5.1 channels hyperpolarize the basolateral membrane potential, creating a favorable electrochemical gradient that promotes chloride exit via basolateral ClC-Kb channels. Knockout or barium inhibition of Kir4.1/5.1 depolarizes the membrane potential, inhibits chloride exit, and increases intracellular chloride concentrations, which, in turn, inhibits a WNK-SPAK/OSR1 kinase pathway that regulates the phosphorylation state and activity of NCC (136–138). In the collecting duct, there appears to be a relatively minor population of homomeric Kir4.1 channels whose functions have not been clearly elucidated (139) (Fig. 2).

There are a number of neurological drugs (e.g., selective serotonin reuptake inhibitors and tricyclic antidepressants) that inhibit homomeric Kir4.1 channels; however, they block Kir4.1 at concentrations in the tens of micromolar (140–142). As a first step toward developing more potent and specific small-molecule tools for probing the physiology and therapeutic value of Kir4.1, we screened 76,574 compounds from the VICB library for modulators of homomeric Kir4.1 channels using a thallium flux assay (143). Putative inhibitors and potentiators of Kir4.1 were identified in the screen. The most potent confirmed inhibitor was 2-(2-bromo-4-isopropylphenoxy)-N-(2,2,6,6-tetramethylpiperidin-4-yl)acetamide, which we termed VU0134992 (Fig. 3). In whole cell patch-clamp experiments, VU0134992 inhibits Kir4.1 with an IC₅₀ of ∼1 µM and affords 9-fold to 22-fold selectivity over Kir4.1/5.1 at -120 mV and +120 mV, respectively. The voltage dependence of block reflects the location of the VU0134992 functional binding site in the ion-conduction pathway, which we localized to E158 and I159 using site-directed mutagenesis and electrophysiology. It is interesting to note that E158 is the “rectification controller” in Kir4.1 and is analogous to Kir1.1-N171 required for block by VU590 and VU591.

Despite its modest potency, VU0134992 had a favorable pharmacokinetic profile that suggested it might be useful for determining if inhibition of Kir4.1 induces diuresis in a rat model. Rats were administered escalating doses of VU0134992, volume loaded with saline, and then placed in metabolic cages for urine collection over a 4-h period. No effect on urine production was observed at a dose of 10 mg/kg; however, significant increases in urine volume, sodium excretion, and potassium excretion were observed at 50 mg/kg and 100 mg/kg (143). This electrolyte profile is consistent with what is known about the physiology of Kir4.1 in the distal nephron. Inhibition of Kir4.1 and possibly some Kir4.1/5.1 channels should 1) inhibit NCC-mediated sodium chloride reabsorption in the distal convoluted tubule, 2) promote flow-stimulated potassium excretion in the collecting duct, and 3) possibly limit ENaC-mediated sodium reabsorption in the collecting duct. Dissecting the relative contributions of Kir4.1 and Kir4.1/5.1 channels in the diuretic response will require the development of Kir4.1/5.1 as well.

ML418: The First Submicromolar Inhibitor of Kir7.1

Kir7.1 (KCNJ13) is one of the most recently cloned members of the Kir channel family and perhaps the most poorly understood. The channel is broadly expressed as a homotetramer in epithelial cells of the eye, gut, kidney, choroid plexus, and respiratory tract, as well as certain brain neurons where it participates in melanocortin signaling (144–153). Kir7.1’s low sensitivity to barium and unusually small unitary conductance (i.e., 50 femptoSiemens) (152) have made dissecting the physiological functions of the channel in native cells difficult. Adding to this problem, genetic ablation of KCNJ13 leads to lethality 24 h after birth due at least in part to defective tracheal and lung development (154). This scenario highlights the need for potent and specific pharmacological tools for modulating Kir7.1 channel activity in genetically wild type animals.

As noted earlier, VU590 was the first moderately potent (IC₅₀ = 8 µM) and moderately selective (also inhibits Kir1.1 with IC₅₀ = 240 nM) inhibitor of Kir7.1 to be published. Despite these shortcomings, VU590 has been used successfully by several groups to assess the function and physiology of Kir7.1 in various tissues (145, 154–157). In an effort to develop more precise Kir7.1 inhibitors, we performed a high-throughput screen of 5,230 compounds from the VICB library using a mutant of Kir7.1 in which methionine 127 was mutated to an arginine (Kir7.1-M125R) to improve thallium flux through the channel (20). The most potent and selective inhibitor discovered in the screen was a compound termed VU714 (Fig. 3), which inhibited Kir7.1 with an IC₅₀ of 1.5 µM but exhibited off-target activity toward Kir1.1, Kir4.1, and Kir6.2/SUR1. Scanning mutagenesis of pore-lining residues identified E149 and A158 as being essential for potent block of Kir7.1. Exploring the structure-activity relationships with the synthesis of ∼30 VU714 analogs led to the development of a compound termed ML418 (Fig. 3) exhibiting ∼5-fold improved potency (i.e., IC₅₀ = 310 nM) and improved selectivity (i.e., no activity toward Kir1.1, Kir2.1, Kir4.1, but still inhibits Kir6.2/SUR1). The pharmacokinetic properties and CNS penetrance of ML418 should support its use for in vivo studies in the brain and periphery (20).

SUMMARY AND PERSPECTIVES

We have described the current state-of-the-art of Kir channel pharmacology developed using a molecular target-based approach that leverages heterologous channel expression, high-throughput screening of small-molecule libraries, and compound optimization with medicinal chemistry. These efforts have led to first-in-class small-molecule inhibitors of Kir1.1, Kir2.x, Kir4.1, and Kir7.1, as well as activators of Kir3 GIRK channels. Some of these compounds have been used to show proof-of-concept of therapeutic potential of Kir channel targets for managing hypertension (Kir1.1, Kir4.1), epilepsy (Kir3), and pain (Kir3). Yet, there is still a lot of work to be done to more fully advance the molecular pharmacology of the Kir channel family. For example, although ML133 is selective for members of the Kir2.X family, there are currently no inhibitors that can distinguish between homomeric or heteromeric subfamily members, nor have any selective potentiators of Kir2.X been identified. Although VU0134992 shows that selective inhibitors of Kir4.1 can be developed, its moderate potency (IC₅₀ = 1 µM), moderate selectivity over Kir4.1/5.1 channels, and poor CNS penetrance create limitations for fully exploring the therapeutic potential of Kir4.1 in hypertension and CNS disorders. It will be critically important to develop selective inhibitors and activators of heteromeric Kir4.1/5.1 to begin addressing the druggability of this Kir channel subtype in the renal tubule and brain. Finally, although in vivo efficacy of GIRK activators has been attained, much work remains to be done to develop compounds with properties suitable for clinical study as well as a more thorough evaluation of potential side effect liabilities associated with increased systemic GIRK activation. The progress described herein clearly illustrates that modern drug discovery approaches can be harnessed in an academic setting to build a useful toolkit of next-generation Kir channel probes.

GRANTS

This work was supported in part by National Institute of Diabetes and Digestive and Kidney Diseases Grant R01DK120821 (to J.S.D) and funds from Vanderbilt University Department of Pharmacology and Institute of Chemical Biology (to C.D.W.).

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

C.D.W. and J.S.D. prepared figures; drafted manuscript; edited and revised manuscript; and approved final version of manuscript.