Inwardly Rectifying Potassium Channels: Their Structure, Function, and Physiological Roles

Hiroshi Hibino, Atsushi Inanobe, Kazuharu Furutani, Shingo Murakami, Ian Findlay, Yoshihisa Kurachi


Inwardly rectifying K+ (Kir) channels allow K+ to move more easily into rather than out of the cell. They have diverse physiological functions depending on their type and their location. There are seven Kir channel subfamilies that can be classified into four functional groups: classical Kir channels (Kir2.x) are constitutively active, G protein-gated Kir channels (Kir3.x) are regulated by G protein-coupled receptors, ATP-sensitive K+ channels (Kir6.x) are tightly linked to cellular metabolism, and K+ transport channels (Kir1.x, Kir4.x, Kir5.x, and Kir7.x). Inward rectification results from pore block by intracellular substances such as Mg2+ and polyamines. Kir channel activity can be modulated by ions, phospholipids, and binding proteins. The basic building block of a Kir channel is made up of two transmembrane helices with cytoplasmic NH2 and COOH termini and an extracellular loop which folds back to form the pore-lining ion selectivity filter. In vivo, functional Kir channels are composed of four such subunits which are either homo- or heterotetramers. Gene targeting and genetic analysis have linked Kir channel dysfunction to diverse pathologies. The crystal structure of different Kir channels is opening the way to understanding the structure-function relationships of this simple but diverse ion channel family.

I. Overview of Inwardly Rectifying K+ Channels

A. Introduction

Inwardly rectifying K+ (Kir) currents were first identified in skeletal muscle (363). Instead of outward rectification predicted by the Nernst equation, they showed greater flow into rather than out of the cell (Fig. 1A). Kir currents were therefore originally described as “anomalous” rectifier K+ currents. This characteristic was clearly different from the voltage-gated K+ (Kv) channel current in squid giant axon (273). Their relationship with membrane voltage did not follow Hodgkin-Huxley kinetics (233); instead, their behavior seemed to depend more on the electrochemical gradient for K+ [membrane potential (Em) minus equilibrium potential of K+ (EK)] (Fig. 1A). These defining characteristics of Kir currents result not from a bending of the rules of biological chemistry by the Kir channel but from asymmetric open channel pore block by intracellular divalent cations and other molecules.

Fig. 1.

Functional properties of Kir channels. A: the dependence of inward rectification and conductance of Kir channels on [K+]o. I-V relationships of the starfish egg cell membrane at four different extracellular [K+]o (10, 25, 50, and 100 mM) in Na+-free media. Continuous and broken lines indicate instantaneous and steady-state currents, respectively. [From Hagiwara et al. (236), copyright 1976. Originally published in The Journal of General Physiology.] B: control of excitability in cardiac cells by Kir channels. Schematic representations of action potentials in ventricular myocytes (a and c) and in the sinoatrial node (b) under various conditions are shown. In ventricular myocytes, inhibition of classical Kir and KATP channels causes Eres to depolarize and Em may oscillate (a). On the other hand, activation of KATP channels hyperpolarizes Em, shortens action potential duration, and may suppress action potential generation (c). Cells in the sinoatrial node express a number of KG channels; their activation may result in hyperpolarization of Em and/or bradycardia (b).

Therefore, under physiological conditions, Kir channels generate a large K+ conductance at potentials negative to EK but permit less current flow at potentials positive to EK (237, 537, 575, 679). Cells that express a large Kir conductance are expected to show the resting membrane potential (Eres) close to EK and no spontaneous electrical activity. This, and their essential voltage independence, permits Kir channels to play key roles in the maintenance of Eres and in regulation of the action potential duration in electrically excitable cells such as cardiac muscle (Fig. 1B) (237, 537, 679).

Kir channels have been found in a wide variety of cells: cardiac myocytes (41, 406, 514, 575, 667), neurons (68, 201, 423, 581, 759, 838), blood cells (445, 517), osteoclasts (732), endothelial cells (727), glial cells (400, 563), epithelial cells (217, 247, 475, 480), and oocytes (235237). G protein-gated K+ (KG) channels (405, 678), which are activated via pertussis toxin (PTX)-sensitive G proteins (65, 411, 412, 414, 580, 584, 620), also show inward rectification (see sect. iii). In addition, ATP-sensitive K+ (KATP) channels, which were originally defined as being opened by a decrease in intracellular ATP (ATPi) (577) also belong to the Kir channel family (see sect. iv) (299, 300, 317). Accordingly, Kir channels not only orchestrate the passive and active electrical properties of cells, but they are also involved in G protein-coupled receptor (GPCR) signaling, and they may link cellular metabolic state and membrane excitability in vivo.

In recent years, the molecular make up, basic architecture, physiology, and pathological relevance of different Kir channels have been described. Here we will discuss these topics, with a particular focus on the molecular and functional characters.

B. Molecular Structure and Basic Stoichiometry

In 1993, two Kir channel cDNAs were isolated by expression-cloning techniques. The ATP-dependent Kir channel ROMK1/Kir1.1 (272) and the classical Kir channel IRK1/Kir2.1 (394) were isolated, respectively, from the outer medulla of rat kidney and a mouse macrophage cell line. Their primary structures possessed a common motif of two putative membrane-spanning domains (TM1 and TM2) linked by an extracellular pore-forming region (H5) and cytoplasmic amino (NH2)- and carboxy (COOH)-terminal domains (Fig. 2A). We now recognize this as the basic building block that is common to all types of Kir channel. The H5 region serves as the “ion-selectivity filter” (249) that shares with other K+-selective ion channels the signature sequence T-X-G-Y(F)-G (46). Kir channel structures lack the S4 voltage sensor region that is conserved in voltage-gated Na+, Ca2+, and K+ channels. As a result, Kir channels are insensitive to membrane voltage and, when particular mechanisms regulating channel activity (e.g., ATP closing KATP channels and Gβγ activating KG channels) are absent, would be active at all Em. Their defining characteristic, inward rectification, turns out not to be an intrinsic function of the channel protein but a result of the block of outward K+ flux by intracellular substances such as Mg2+ and polyamines (see sect. iC1). The primary structure of two transmembrane strands is insufficient to form a complete ion channel, and functional Kir channels are made up of four such subunits in a tetrameric complex (see Fig. 4B) (208, 866). This stoichiometry was confirmed by velocity sedimentation with sucrose density gradients, size exclusion column chromatography, and chemical cross-linking (305, 639).

Fig. 2.

Basic structure and Kir channel phylogenetic tree. A: primary structure of the Kir channel subunit (left). Each Kir subunit contains two transmembrane (TM1 and TM2) regions, a pore-forming (H5) loop, and cytosolic NH2 and COOH termini. As a comparison, the structure of voltage-gated K+ (Kv) channel subuit, which possesses six transmembrane (TM1-TM6) regions, is shown on the right. B: amino acid sequence alignment and phylogenetic analysis of the 15 known subunits of human Kir channels. These subunits can be classified into four functional groups.

To date, 15 Kir subunit genes have been identified and classified into seven subfamilies (Kir1.x to Kir7.x) (Fig. 2B). These subfamilies can be categorized into four functional groups: 1) classical Kir channels (Kir2.x), 2) G protein-gated Kir channels (Kir3.x), 3) ATP-sensitive K+ channels (Kir6.x), and 4) K+-transport channels (Kir1.x, Kir4.x, Kir5.x, and Kir7.x) (Fig. 2B).

The simplicity and strong homology of the basic Kir channel subunit allow for both homomeric and heteromeric combinations to form functional Kir channels. Heteromerization generally occurs between members of the same subfamily, for example, Kir2.1 can associate with any one of other Kir2.x subfamily members, namely, Kir2.2, Kir2.3, or Kir2.4 (626, 695) (see sect. ii), and Kir3.1 forms heteromeric complexes with either Kir3.2, Kir3.3, or Kir3.4 (see sect. iii). An exception is where Kir4.1 assembles with Kir5.1 (see sect. vB). Heteromeric assemblies confer distinct properties to their particular channels; they can determine their location on a cell as well as extend the functional range of Kir channels in different cell types.

C. Kir Channel Function and Localization

The physiological activity and functions of Kir channels depend on regulation of pore opening, ion flux, and channel localization in the cell. Major factors that regulate pore opening and ion flux include ions, polyamines, nucleotides, lipids, and a variety of intracellular proteins (Fig. 3A). Many of these interact directly with crucial elements of the Kir channel to modulate properties such as ion flux and channel pore opening kinetics. The localization of the channels in particular regions of a cell, such as apical or basolateral membranes in epithelial cells and pre- or postsynaptic sites in neurons, and that in membrane microdomains where they may be in close proximity with other transport molecules are also important contributors to the functional roles of Kir channels in different cells and tissues (Fig. 3B).

Fig. 3.

Regulators of Kir channel function. A: the functions of Kir channels can be regulated by small substances (top panel) and by proteins (bottom panels). The small substances are ions such as H+, Mg2+, and Na+; polyamines; phosphorylation; and membrane-bound phospholipids. Protein-protein interaction involves sulfonylurea receptors (SUR), G proteins liberated from G protein-coupled receptors (GPCR), and anchoring proteins. B: the localization of Kir channels on a cell may determine their particular function. The channels may be distributed homogeneously in nonpolarized cells or in a specific pattern in membranes of polarized cells such as epithelia and neurons (top panel). Specific localization patterns play a role in unidirectional transport of K+ and in the organization of signal transduction. Particular channels and other transport mechanism may be gathered in microdomains such as detergent-resistant membrane microdomains (DRMs) and caveolae (bottom panel). Such colocalization may be necessary for the physiological coupling of ion-transport mechanisms.

1. Regulation of the Kir channel pore


Inward rectification of K+ flux through Kir channels results from interaction between two intracellular substances, Mg2+ and polyamines, and the lining of the channel pore (Fig. 3A) (see sects. iD and ii). They physically block K+ permeation by binding to residues localized in the transmembrane and cytoplasmic regions of the channels (471, 508).

Early studies led to the conclusion that rectification arises from a combination of intracellular Mg2+ (Mgi2+)-mediated blockage and an intrinsic activation gating process (310, 406, 507, 508). For instance, in the current of cardiac classical Kir channels (IK1), Mgi2+ caused instantaneous inward rectification, which was followed by a time-dependent further enhancement of rectification by the intrinsic gating mechanism (310). But the intrinsic gating property was lost in inside-out patches and restored by submicromolar concentrations of polyamines such as spermine and spermidine applied to the intracellular side of the Kir channels (160, 172, 471, 856). Because polyamines exist in cells at submillimolar concentrations, the putative activation gating behavior is now assumed to result from their slow blocking and unblocking of the Kir channel. Thus, on depolarization, what was previously called “deactivation” corresponds to polyamines causing a time-dependent decrease of outward current. On hyperpolarization, the inward Kir current first increases time-independently due to fast Mg2+ unblocking and then increases in a time-dependent fashion, which was referred to as “activation,” is due to slow polyamine unblocking (472).

Not all types of Kir channels show the same degree of inward rectification. There are “strong” (Kir2.x and Kir3.x), “intermediate” (Kir4.x), and “weak” (Kir1.1 and Kir6.x) rectifiers. A negatively charged residue Asp at position 172 in the TM2 helix in a strong rectifier of Kir2.1 seems to be a critical determinant of inward rectification for various Kir channels. The weakly rectifying Kir1.1 has an uncharged residue Asn (N171) in this position, and switching Asn for Asp (N171D) in Kir1.1 was found to increase the affinity for Mg2+ and cause strong rectification (483, 742, 831, 866). This position is now known as the “D/N site.” Studies of mutated Kir channels have also revealed that Kir channel subunits possess more than one binding site for Mg2+ and polyamines, and further details will be described in section iD and for each Kir channel subfamily. The competitive blockage of Kir channels by Mg2+ and polyamine is crucial for the control of the magnitude of outward current (see an example of Kir2.2 in Ref. 861).


The conductance of all Kir channels except for Kir7.1 (140, 393) increases as extracellular K+ concentration ([K+]o) augments, which depends on the square root of [K+]o (237, 394, 473, 493, 614, 679). This property implies that ion permeations through Kir channels do not follow the Goldman-Hodgkin-Katz permeability theory describing independent movement of ions in the pore but rather agree with the multi-ion pore model (266). In support of this idea, the degree of Cs+-induced block of the Kir current in starfish egg is modified by change of [K+]o (93, 94). Hagiwara and co-workers (93, 94) proposed that Kir channel pore would have at least two binding sites for K+. Moreover, Kir2.1 conductance in the absence of Mg2+ and polyamines also exhibits the square-root dependence on [K+]o, strongly suggesting that this dependence seems to be a property of the open-channel pore (473). Consistently, crystal structure of KcsA K+ channel reveals that the selective filter contains two K+ ions in the condition of 150 mM [K+]o (143).


A membrane-anchored phospholipid, phosphatidylinositol 4,5-bisphosphate [PtdIns(4,5)P2], is essential to sustain the normal function of the majority of Kir channels (263, 264, 287, 760) (Fig. 3A). In membrane patches excised from their parent cells, Kir channel activity gradually declines. This “run-down” activity can be restored by the application of ATP to the intracellular surface of the membrane which replenishes PtdIns(4,5)P2 via the action of lipid kinases (263). Mutation analyses suggest that PtdIns(4,5)P2 is associated with positively charged residues in the COOH termini (287, 474). The structural basis of PtdIns(4,5)P2 regulation can be estimated using the high-resolution structures of Kir subunits (see sects. iD and ii).


Intracellular and/or extracellular pH can regulate Kir channels such as Kir1.1, Kir2.3, and channels which contain the Kir4.1 subunit (Fig. 3A); generally acidic shifts of pH reduce channel activity (see sects. ii and v, A and B). KG channels that contain either Kir3.2 or Kir3.4 can be activated by intracellular Na+ (Nai+) (270, 619) (see sect. iii). KATP channels, which are made up of pore-forming Kir6.x and auxiliary sulfonylurea receptor (SURx) proteins, are inhibited by ATPi and activated by intracellular nucleotide diphosphates (NDPsi) (see sect. iv). ATP controls pore function through direct interaction with cytoplasmic regions of Kir6.2 (15, 111, 344, 450, 632, 649, 790792). On the other hand, NDPsi bind to SURx which then affects the pore function of Kir6.x to increase channel opening (28, 114, 221, 512, 720, 796) (see sect. iv).


Phosphorylation of Kir channel subunits by protein kinases such as protein kinases A and C (PKA and PKC, respectively) can modulate their activity (Fig. 3A). A serum-glucocorticoid-regulated kinase (SGK) phosphorylates Kir1.1 and promotes its surface expression (877) (see sect. vA3). Phosphorylation of Ser residues in Kir1.1 by PKC results in suppression of channel activity (456) (see sect. vA3). PKA phosphorylates both Kir6.1 and SUR2B subunits of the smooth muscle type KATP channel and enhances its activity (637, 713) (see sect. ivB).


Protein-protein interactions are involved in control of Kir channel pore function. They include interaction between βγ subunits of G protein (Gβγ) and Kir3.x, association of SUR with Kir6.x, and binding of anchoring proteins to diverse Kir channels (Fig. 3A). KG channels are activated by direct association between cytoplasmic regions of Kir3.x and Gβγ subunits released from Gi/o-coupled receptors (78, 246, 288, 289, 305, 326, 392, 402, 734, 863) (see sect. iii). Functional KATP channels require the association of SURx subunits with Kir6.x subunits to form a hetero-octameric structure (299, 300, 317, 853). Cytoplasmic regions of SURx interact with Kir6.x to modify channel activity (30, 649) (see sect. iv). Anchoring proteins such as PSD-95, SAP97, and sorting nexin 27 play key roles in determining the localization of some Kir channels in the cell surface (9, 279, 484, 595) (see sects. ii, iii, and v).

2. Regulation of localization of Kir channels


Many cells are polarized. Epithelial cells have apical and basolateral membrane domains; neurons are divided into soma, dendrites, axon, and axonal terminal; and astrocytes harbor somatic, perivascular (end feet), and perisynaptic membrane domains (Fig. 3B) (568, 725). These membrane “macrodomains” may express different Kir channels. The postsynaptic membrane bears KG channels made up of Kir3.1/3.2 (307, 453, 623) (see sect. iii). Renal epithelial cells have Kir1.1 upon their apical membranes (384, 846) and Kir4.1/5.1 on their basolateral membranes (476, 765, 793) (see sect. vB).

This manner of organizing the expression of channels on the cell surface may serve to couple Kir channels to selective apparatus to regulate particular cell functions (197). Thus, in the postsynaptic membrane, Kir3.1/3.2 is associated with Gi/o-coupled receptors such as γ-aminobutyric acid type B (GABAB) to produce ligand-induced inhibitory postsynaptic current (390, 535, 778) (see sect. iii). In the apical membrane of renal epithelia, Kir1.1 is functionally associated with the Na+-K+-2Cl cotransporter to balance salt and water homeostasis (247) (see sect. vA).

A number of mechanisms are involved in the control of localization of membrane proteins to specific macrodomains. They include protein-protein interaction and protein-lipid interaction. Scaffolding proteins containing PSD-95/Dlg/ZO-1 (PDZ) and src-homology-3 (SH3) domains play important roles (197). Sphingolipids are required for axonal delivery of glycosylphosphatidylinositol (GPI)-anchored proteins in neurons (433). Cholesterol is necessary for apical sorting of some membrane proteins in epithelia (365). For certain Kir channels, PDZ domain-containing proteins (PDZ proteins) are involved in their localization in polarized cells (99, 279, 302, 441, 442, 555) (see sects. iiC, iiiC, and vB3).


Membrane microdomains may also contain various functional proteins (730, 731) (Fig. 3B). Microdomains include caveolae, which are flask-shaped membrane invaginations (13). There are also “detergent-resistant microdomains of cell membranes” (DRMs) or “lipid-rafts,” which are enriched with several types of lipids such as cholesterol and sphingolipids and isolated biochemically as nonionic detergent-insoluble components (69, 730, 731) (Fig. 3B). Caveolae contain abundant lipids, and they are now considered as one type of DRMs (730). Distinct microdomains may concentrate particular sets of molecules. For example, caveolae accumulate particular GPCRs, G proteins, and adenylyl cyclases (707). Noncaveolar DRMs also anchor particular receptors and cytosolic proteins such as neurotrophins and mitogen-activated protein (MAP) kinases (731). In the case of Kir channels, some DRMs harbor Kir4.1 in astroglial cells and Kir3.1/3.2 in neurons (127, 259). Such localization would synchronize channel activities with other functional proteins and control various signaling and physiological activities of the cells. Kir4.1 may occur in close vicinity with the water channel AQP4 to mediate K+-driven water transport in astroglial cells (259) (see sect. vB3). Accordingly, the specific localization of Kir channels in membrane macro- and microdomains will contribute to cell and organ function.

D. Structure

X-ray crystallography has provided details of the three-dimensional structure of Kir channels down to the atomic level (306, 404, 573, 574, 611). A generic view of the Kir subunit and the tetrameric Kir channel are shown in Figure 4,A and B (573). The NH2 and COOH termini of Kir subunits are exposed to the cytoplasm and associate with each other to form a cytoplasmic domain that is linked to but distinct from the transmembrane domain (Fig. 4A). The transmembrane domain is mainly responsible for ion selectivity and gating (Fig. 4C). The cytoplasmic domain is thought to act as a gating regulator.

Fig. 4.

Molecular architecture of Kir channels. A: a schematic representation of the structure of a generic Kir channel. The Kir channel is divided into transmembrane and cytoplasmic domains. The NH2 and COOH termini are cytosolic and together contribute to the formation of the cytoplasmic domain. B: tetrameric assembly of Kir channels. The molecular architecture of a tetrameric Kir channel (protein database ID 2QKS: Kir3.1-KirBac3.1 chimera) is represented as a cartoon model. The front subunit has been omitted for clarity. The organization of the tetramer of NH2 and COOH termini leads to an extended pore for ion permeation. C: the transmembrane domain. An enlarged view of the transmembrane domain indicates several important secondary structures for Kir channel function. The transmembrane domain comprises three helices: TM1, pore, and TM2. At the membrane-cytoplasm interface, there is also an amphiphilic slide helix. The residue that is largely responsible for the interaction with polyamines and Mg2+ and thus inward rectification is indicated by the yellow/red spheres (D131 in Kir3.1 which was mutated in the original coordinate). D: the cytoplasmic domain. The opening of Kir channels requires PtdIns(4,5)P2. Those amino acid residues (blue) associated with the interaction with PtdIns(4,5)P2 are distributed on the surface of the cytoplasmic domain towards the plasma membrane. The center of the cytoplasmic domain is a water-filled cavity that contributes to the ion permeation pathway. Some residues associated with inward rectification (red) map along this cavity. Kir3.x channels are activated by direct interaction with a G protein βγ (Gβγ) subunit and channels containing Kir3.2 and Kir3.4 subunits can also be activated by Na+. The amino acids responsible for these stimuli are indicated in green (Gβγ) and pink (Na+). C and D have been derived from B.

1. Transmembrane domain

The architecture of the transmembrane domain of Kir channels has been determined by solving the crystal structures of a bacterial homolog KirBac1.1 (404) and a chimera between bacterial KirBac3.1 and mouse Kir3.1 (573) (Fig. 4C). The domain is composed of outer (TM1) and inner (TM2) membrane spanning helices, with two short additional helical elements, the slide and the pore helices. The channel pore is delimited by one TM2 helix from each of the four Kir subunits. The arrangement of the helices to form the transmembrane domain shares significant similarity with that of other K+ channel families including a bacterial channel KcsA (143, 539, 894), a bacterial calcium-activated channel MthK (336, 337), a bacterial voltage-gated channel KvAP (338), and a mammalian voltage-gated channel Kv1.2 (466, 467).

The ion conduction pore (Fig. 4C) can be functionally divided into three distinct zones that consist of the selectivity filter, a water-filled central cavity, and the internal face of the pore made up of the internal bases of the four inner (TM2) helices. The K+ channel signature sequence [T-X-G-Y(F)-G] is the selectivity filter. This makes a narrow region in the ion conduction pathway which separates the central cavity from the extracellular solution. The central cavity is a 10-Å spherical water-filled space about halfway through the membrane. The bases of four TM2 helices (one from each subunit) come together to make another narrow region at the cytoplasmic face of the channel.

There are two distinct gating mechanisms in Kir channels: slow gating and fast gating. In single-channel recordings, slow gating corresponds to channel openings occurring as bursts separated by long closed periods, and fast gating corresponds to rapid opening and closing within the burst. Mutations introduced around the bundle crossing area of the inner TM2 helix tend to modulate the behavior of bursts (785, 792, 872). On the other hand, mutations around the selectivity filter cause alterations in fast gating (85, 87, 157, 232, 630, 872).

Each TM2 helix can bend around a hinge point (Gly) about halfway along their length. This bending would vary the cross-sectional area of the narrow region at the cytoplasmic face of the transmembrane domain. Analyses of the mechanisms underlying the activation of the constitutively closed KG/Kir3.x channels and comparisons with the constitutively open classical IK1/Kir2.x channels has provided clear evidence for mobility of the TM2 helices contributing to Kir channel gating.

KG/Kir3.x channels are closed in the absence of an external agonist activating a GPCR to release Gβγ. Gβγ causes specific conformation changes in KG channels that allow channel opening. Jin et al. (340) demonstrated that the highly conserved G175 residue in the middle of the TM2 helix was crucially involved in Gβγ-induced activation of Kir3.4. They also found that replacement of certain residues in the TM2 helix below G175 with Pro, which would increase the flexibility of the TM2 helix, formed constitutively active channels (340). These results, together with homology modeling, imply that this Gly may serve as a hinge in the TM2 helix and allow the inner part of the helix to swing away from the permeation pathway upon Gβγ stimulation. This idea is similar to the gating mechanism of the KcsA channel proposed by Jiang et al. (337) and consistent with the result that substitution of Pro for S171 in Kir3.1 disrupted Gβγ-induced activation gating because of loss of the rigidity of the below-hinge part of the TM2 helix (674).

In addition to the importance of Gly as a hinge, another residue in the TM2 helix seems to play a role in activation gating. In Kir3.1 and Kir3.4, the replacement of a bulky hydrophobic residue Phe (F181 in Kir3.1; F187 in Kir3.4) with Ala or Ser, but not with Met, at the bundle-crossing region converted channels from agonist activation to constitutive activity (97, 674). The corresponding residue in KirBac1.1 (F146) is thought to act as a barrier to ion permeation (404) which supports the idea that Gβγ stimulation alters the conformation of this region.

Comparisons between Kir3.x and Kir2.x channels have also been instructive. In the constitutively inactive channel Kir3.2, the mutation V188G, which is situated below Phe (F181) in the bundle-crossing region of TM2, converted activation gating to constitutive activity (872). Homology modeling suggested that, in the closed state, V188 might mediate an interaction between adjacent TM2 helixes and the V118G mutation destabilized the closed conformation by disrupting this interaction (872). The corresponding residue (I176) in the constitutively active Kir2.1 faces the pore cavity and thus provides no interaction between the TM2 helixes in this position (534). These structural differences imply that, in Kir3.2, Gβγ stimulation would induce a clockwise rotation of TM2 to open the channel when it is viewed from the outside of the cell, and this would bring the Kir3.2 channel pore into a configuration equivalent to that of Kir2.1 (872). This model is also supported by the following observation. An overexpression of Gβγ can weaken both the inward rectification and Ba2+ and Cs+ block of KG channels (277). An excess of Gβγ may induce sufficient rotation of TM2 helices to alter the coordination of the binding sites for polyamines, Mg2+, Ba2+, and Cs+ between Kir subunits, because the residues critically involved in block of the pore by these substances are localized near the Gly hinge in TM2 helices (see sect. iiB) (339).

Not only TM2 but also TM1 are critically involved in gating machinery of Kir channels. In Kir3.2, mutation of N94 of TM1 to His results in formation of constitutively active channel, suggesting that this residue is involved in the gating process (872). Homology modeling of Kir1.1 indicates that K80 on TM1, which corresponds to N94 in Kir3.2, forms with A177 of TM2 a hydrogen (H+) bond in the bundle-crossing region (644). Further analyses of Kir1.1 demonstrate that the H+ bonding plays key roles in control of gating property by intracellular substances such as H+ and PtdIns(4,5)P2; mutation of K80 that would disrupt the bonding lowers intracellular pH (pHi) sensitivity and fastens recovery from acidification-induced channel inhibition and reactivation of run-down channel by PtdIns(4,5)P2. The profile of pHi- and PtdIns(4,5)P2-dependent modulation of Kir4.1 that is predicted to harbor H+ bonding resembles the characteristic of Kir1.1, whereas Kir2.1 that is expected to have no H+ bonding mimics the phenotype of the mutated Kir1.1. Therefore, the H+-bond linking between TM1 and TM2 at the bundle-crossing region is a crucial determinant for channel gating in all the Kir channels (644) (also see sect. vA2).

While the pore lining helices in the KcsA structure are arranged in a fourfold symmetry, those in KirBac1.1 are misaligned (404, 894). Furthermore, only three K+ ions could be modeled in place in the selectivity filter of KirBac1.1, and the temperature factor of one of these K+ ions is much smaller than for the other two. This is quite different from the character of the selectivity filter of KcsA and KirBac3.1 where four K+ ions are accommodated with similar temperature factors (573). The unequal ion occupancy among the four K+ binding sites in the selectivity filter of KirBac1.1 suggests that this structure is in the closed state. These structural alterations around the selectivity filter may also serve to gate Kir channels, and they may correspond to fast gating.

The slide helix at the cytoplasmic end of TM1 is a unique structural element of the Kir channel family compared with other K+ channels (143, 336, 338, 404, 466) (Fig. 4C). It is situated outside of the inner (TM2) and the outer (TM1) helices and connects the TM1 domain and the NH2-terminal part of the cytoplasmic domain. Since the slide helix is amphiphilic, it may lie at the interface between the inner leaflet of membrane and the cytoplasm. Loss-of-function mutations in the slide helix of Kir1.1 and Kir2.1 have been identified in human inherited diseases of type II Bartter syndrome (V72E and D74Y) (755) and Andersen syndrome (Y68D, D71V, D71N, T74A, T75A, T75M, T75R, D78G, D78Y) (44, 123, 139, 622, 767) (see sects. iiF and vA6). These mutations led to the impairment of channel activity by either suppressing channel expression in the plasma membrane or inhibiting K+ permeability. Therefore, the slide helix has important roles to play in Kir channel function.

2. Cytoplasmic domain

The crystal structure of the cytoplasmic domains of Kir channels has been obtained in two ways. Stable crystals of the NH2- and COOH-terminal complexes that make up the cytoplasmic domains of Kir3.1 (574), Kir2.1 (611), and Kir3.2 (306) have provided high-resolution information. These results have proven to be comparable to the lesser resolution data obtained from entire Kir channels (404, 573). Therefore, conformational information obtained from the cytoplasmic region of Kir channels should correspond to that of the overall structure of the entire channel protein. But data concerning the region connecting NH2 and COOH termini and the membrane interface should be interpreted with caution. It is worth noting that the amino acid sequence of regions connecting NH2 and COOH termini to the membrane segment is conserved among the Kir channel family (Fig. 4D).

The cytoplasmic domain of Kir channels is made up of the NH2 and COOH termini of four Kir subunits. Each terminus is rich in β-strands (Fig. 4D). These β-strands form three β-sheets. The cytoplasmic domain is mainly formed from the COOH terminus of each subunit, while the NH2 termini are present between adjacent COOH termini and they contribute to formation of the interface between subunits. Each NH2 terminus contains a single β-strand (βA) that forms a β-sheet with two β-strands (βL and βM) in a COOH terminus. This contributes to the stability of the whole cytoplasmic region. The interaction between NH2 and COOH termini has been observed in both Kir3.x and Kir6.x channels. Huang et al. (1995, 1997) demonstrated that NH2- and COOH-terminal domains of Kir3.1 and Kir3.4 subunits could bind to each other and together synergistically enhanced Gβγ binding (288, 289). Tucker and Ashcroft (792) demonstrated a similar association between the COOH terminus of Kir6.2 and a conserved region of the NH2 terminus of either Kir6.2 or Kir2.1, and they suggested that the association of both termini may be common to all Kir channels.

The four groups of associated NH2/COOH termini make up a cylinder that surrounds the so-called cytoplasmic pore. This architecture is a characteristic of the Kir channel family and extends the ion conduction pathway by ∼30 Å (574). Therefore, in Kir channels, K+ has to pass over 60 Å through the pore composed of the transmembrane and cytoplasmic domains. The detailed structure of the cytoplasmic pore and its relevance to channel function and channelopathies (611) will be discussed in reference to Kir2.1 (see sect. ii).

Inward rectification of Kir channel currents results from intracellular Mg2+ and polyamine block of the channel pore when Em is more depolarized than EK (see sect. iC1). Although several residues responsible for this block have been identified on the inner TM2 helix in the transmembrane domain, others have been identified in the cytoplasmic domain (195, 230, 231, 417, 757, 865). These acidic residues have been mapped to the wall of the pore of the cytoplasmic domain (see sect. ii for further details).

PtdIns(4,5)P2 is a molecule that is necessary for the normal functioning of Kir channels (44, 263, 287, 474, 723, 889), as for other ion-transport proteins such as the Na+/Ca2+ exchanger, voltage-gated Ca2+ channels, and transient receptor potential channels (263, 264, 748) (see sect. iC1). In Kir channels, a number of basic and uncharged residues in the cytoplasmic region of the channel are proposed to be involved in the action of PtdIns(4,5)P2 (146, 163, 204, 287, 474, 487, 653, 699, 722, 723, 886, 889). These are mostly to be found upon the surface of the cytoplasmic region which faces the membrane (Fig. 4D). Residues in the transmembrane region may also participate in the association between the channel and PtdIns(4,5)P2; they include basic residues clustered in lower part of the inner TM2 helix and an uncharged residue in the slide helix (474, 660). However, the mechanism of action of PtdIns(4,5)P2 has not been fully elucidated. The crystal structure of the KirBac3.1-Kir3.1 chimera was obtained in the presence of PtdIns(4,5)P2 (573), but the amino acid residues at the membrane-cytoplasm interface as well as the crystal of PtdIns(4,5)P2 itself could not be modeled due to their weak electron density.

The cytoplasmic domain pore abutting to the transmembrane domain is formed by a loop [HI (G) loop] between βH and βI strands (Fig. 4D). The cross-sectional area of the space surrounded by the HI loops is wide in Kir3.2 and narrow in Kir2.1. Mutations in the HI loop disrupted gating and affected inward rectification of Kir2.1 (611) (see sect. iiB for details). In the crystallization of the KirBac3.1-Kir3.1 chimera, two different conformations were obtained (573). A prominent difference between the two is found in the apex of the cytoplasmic pore formed by the HI loop (G-loop). In one conformation the apex is dilated, whereas in the other it is constricted. These two conformations of the chimera channel may represent open and closed states of the cytoplasmic pore, respectively. In support of this idea, several loss-of-function mutations in Kir channelopathies are found in the residues forming the apex (611) (see sect. ii).

The cytoplasmic domains are involved in the control of Kir channel gating by Na+, nucleotides, and G proteins. Structural analyses of Kir3.x strongly suggest that Gβγ and Na+, like PtdIns(4,5)P2, interact with the cytoplasmic region (78, 246, 288, 289, 305, 326, 392, 402, 734, 863) (see sect. iiiB). In the case of KATP channels, homology modeling predicts that one NH2 terminus residue and three COOH terminus residues create the binding pocket for ATP in the Kir6.x subunit (15) (see sect. ivB). Biochemical assays show that, in Kir2.3, acidification strengthens the interaction between NH2 and COOH termini, which is suggested to be involved in channel closure by an unidentified mechanism (634). The modulation of the conformation of NH2 and COOH termini and their interaction clearly play pivotal roles in the regulation of Kir channel gating.

E. Pharmacology

Blockers most commonly used for Kir channels are Ba2+ and Cs+ (189, 234236, 266, 315, 592). Tetraethylammonium (TEA) and 4-aminopyridine (4-AP) are known as the inhibitors of Kv channels but have little effects on Kir channels (236, 598). Ba2+ and Cs+ effectively block the majority of Kir channels. Although high concentrations of Ba2+ can also block Kv channels, at micromolar concentrations this cation is relatively specific to Kir channels (188, 636). Ba2+ and Cs+ are often used to examine the physiological roles of Kir channels in native cells and tissues. Externally applied Ba2+ and Cs+ suppress Kir currents in a voltage-dependent manner. They inhibit Kir channels more strongly as the membrane is hyperpolarized (189, 235, 236, 592). In addition, the blocking effect of Cs+ and Ba2+ decreases substantially as [K+]o increases (235, 236). However, there are different aspects in the behavior of the two blockers. First, following a voltage step, the approach to steady-state block is much faster for Cs+ than Ba2+. Second, the dissociation constant for Cs+ is dependent on extracellular K+ concentration ([K+]o) (236), but that for Ba2+ is independent of this factor (235).

In spite of a limited number of blockers of Kir channels (see above), pharmacological and physiological assays have revealed other compounds that can affect particular types of Kir channels. Tertiapin is a toxin that was isolated from honey bee venom which blocks KG and Kir1.1 channels (342). The oxidation-resistant tertiapinQ (343) blocks KG current, but not KATP current or IK1 in isolated cardiac myocytes (136, 372) (see sect. iiiE). The majority of reagents affecting KATP channel activity react on the auxiliary SURx subunits. KATP channels are the targets of antidiabetic sulfonylurea compounds (219) as well as the class of compounds known as “K+ channel openers” (18) (see sect. ivE). These reagents are also useful tools to examine basic properties of KATP channels.

II. Classical Kir Channels (Kir2.x)

A. Historical View and Molecular Diversity

The inward rectifier K+ channel in skeletal and cardiac muscle belongs to the Kir2.x channel family. The channels of this family are constitutively active and exhibit strong inward rectification. They contribute to the establishment of highly negative Eres and long-lasting action potential plateau in various cells including cardiac myocytes.

The first member of this family to be cloned was from a mouse macrophage cell line and named IRK1/Kir2.1/KCNJ2 (394). Afterward, three more subunits were identified as members of this family. They are Kir2.2(IRK2)/KCNJ12 (389, 758), Kir2.3(IRK3, BIR11)/KCNJ4 (56, 64, 543), and Kir2.4(IRK4)/KCNJ14 (782). The amino acid sequence of mouse brain Kir2.1 shares 70, 61, and 63% identity with Kir2.2, Kir2.3 and Kir2.4, respectively. The sequences are most highly conserved in the TM1, TM2, and H5 regions.

Initially Kir2.x subunits were thought to be made up of only homomeric complexes (780). However, recent studies have revealed that Kir2.x subunits can function as heterotetramers both in vitro and in vivo. Indeed, in vitro electrophysiological experiments have shown that each of Kir2.1, Kir2.2, and Kir2.3 can assemble with any one of the other subunits, and the respective heteromer exhibits different properties from that of their homomers (626) (Table 1). Some of these heteromers have been identified in native tissues. For example, heteromeric assemblies of Kir2.1/2.2 and Kir2.1/2.3 channels are expressed in cardiac myocytes (see sect. iiD), and Kir2.1/2.4 has been found in the brain (695).

View this table:
Table 1.

Biophysical properties of Kir2.x channels

B. Pore Function and Structure Bases

1. Rectification and activation

The basic architecture of Kir channels with transmembrane and cytoplasmic regions and pore structure (see Fig. 4 and sect. iD) is conserved among all Kir2.x subunits. Inward rectification is caused by intracellular ions such as Mg2+ (508, 802) and polyamines (471, 856) (see sect. iC). Studies of mutated Kir2.x channels have revealed that the subunits possess more than one binding site for Mg2+ and polyamines. The strong rectifier Kir2.1 has the negatively charged residue Asp at the D/N site. Another residue in the TM2 helix, S165 in Kir2.1, has been reported to be crucial for block by Mg2+ but not by polyamines (196). The crystal structure of KirBac1.1 suggests that these residues face the transmembrane pore cavity formed by the inner TM2 helixes.

Further site-directed mutagenesis has identified negatively charged amino acids (Glu) at two different positions (E224 and E229 for Kir2.1) in the COOH terminus of the cytoplasmic domain that are critically involved in both Mg2+ and polyamine sensitivity (396, 756, 757, 865). Mutagenesis and substituted cysteine accessibility experiments suggest that these residues directly interact with Mg2+ and polyamines (482, 534). Structural information from KirBac, Kir3.1, and Kir2.1 supports the pore-lining location of these amino acids: the side chains of Glu point to the center of the cytoplasmic domain conduction pathway in each subunit, forming rings of negatively charged residues that create a complimentary electrostatic match for the binding of a positively charged polyamine (404, 574, 611).

The reason why Kir2.1 has residues that bind Mg2+ and polyamines in both the transmembrane and the cytoplasmic domains is still elusive. It is possible that these two types of binding sites play distinct roles: polyamines may plug the pore in the transmembrane domain (plugging site), whereas the cytoplasmic region would serve as an intermediate binding site (nonplugging site) that increases the local concentration of polyamines around the plugging site (396, 434, 472, 844).

Analysis of the crystal structure of the cytoplasmic domain of Kir2.1 has recently identified an intrinsically flexible loop around the membrane face of the cytoplasmic pore (611). The loop constricts the cytoplasmic pore to ∼3 Å and forms a girdle around the central pore axis. The girdle, which consists of a loop between βH and βI strands and is called the “G-loop,” forms the narrowest portion of the ion conduction pathway in the cytoplasmic region (Fig. 5) (see also sect. iD). The narrowest part of the G-loop is made up by A306 and to a lesser extent by E299, G300, M301, and M307. A306 is localized at the apex of the G-loop. The substitution of Glu, Cys, or Thr for A306 abolished Kir2.1 current. Because the side chain of these residues is larger than that of Ala, these substitutions would result in the physical occlusion of the G-loop without changing its backbone conformation. When another constituent of the G-loop M301 was mutated to Ala, an enhancement of inward rectification was observed. Thus the G-loop also contributes to inward rectification in Kir2.1. Charged amino acids, R228, D255, D259, and R260, face the cytoplasmic pore (Fig. 5). Charge reversal at D255 (D255R) or at D259 (D259R) and charge neutralization at D259 (D259A) decreased inward rectification, whereas substitutions of R228 and R260 with Ala had little effect on the rectification profile. Collectively, the diaspartate cluster (D255/D259) that faces the cytoplasmic pore is involved in inward rectification of Kir2.1 channels.

Fig. 5.

The structure of the cytoplasmic pore region of Kir2.1. Side (left and bottom right) and top-down (top right; membrane to cytoplasm) views of the cytoplasmic region (NH2 and COOH termini) of the Kir2.1 structure, highlighting amino acids lining the permeation pathway. Cβ atoms (A306) of the G-loop make up the narrowest ∼3-Å region of the pore and are shown as open circles for clarity. Other residues near the G-loop are labeled. This illustration was constructed from protein database ID 1U4F (611). [From Pegan et al. (611), with permission from Nature Publishing Group.]

Kir2.x are activated by PtdIns(4,5)P2 (264, 287, 760) (see sect. i, C and D). The sequence that was first found to be associated with activation of Kir2.1 by PtdIns(4,5)P2 was the group of positively charged residues, KKR (177–179 in Kir2.1) just beyond the TM2 region (287, 474, 722, 737). Cytoplasmic NH2 and COOH termini also harbor residues involved in PtdIns(4,5)P2 regulation. In Kir2.1, these residues are H53, R67, K187, K188, R189, K219, R228, and R312 (474).

Andersen syndrome is caused by dysfunction of Kir2.1 (see sect. iiF for the phenotypes) (Table 2). Early studies identified a variety of mutations, including R67W, D71V/N, T75R, Δ95–98, S136F, G144S, G146D, P186L, R189I, T192A, G215D, N216H, R218W/Q, G300V/D, V302M, E303K, R312C, and Δ314–315 (11, 139, 281, 622, 787). They generally disrupt Kir2.1 activity via dominant negative interactions in the formation of heteromeric assemblies (44, 281, 622, 626, 787). Of the more than 30 mutations that are currently known to be associated with Anderson syndrome (120, 682), the majority occur in the cytoplasmic domain: some are located around the top region of the cytoplasmic structure (R189, T192, R218, G300, V302, E303, R312, Δ314–315) close to the PtdIns(4,5)P2 binding site (11, 139, 474, 622, 787), and others (G215, N216) are included in βC-βD loop near G-loop. Four locations, G300, V302, E303, and the site of the deletion mutant Δ314–315, are in the G-loop region (Fig. 5), implying a crucial role of this region in the K+-conduction pathway. For example, V302 is located near the apex of the cytoplasmic structure. A loss of function mutation, V302M (787), rendered the channel unable to conduct K+ without affecting subunit assembly or attenuating trafficking, but by impairing PtdIns(4,5)P2 sensitivity (487).

View this table:
Table 2.

Human Kir channelopathies

2. Other regulatory factors

Kir2.x subunits heterologously expressed in Xenopus oocytes and mammalian cells elicit strongly rectifying K+ currents. As summarized in Table 1, the channels made up of various homo- and heteromultimers of Kir2.x subunits exhibit unique single-channel conductance, respectively (394, 461, 543, 626, 758, 782). The steady-state open probability (Po) of Kir2.2 decreases with hyperpolarization, whereas that of Kir2.1 or Kir2.3 remains unchanged. In addition, a remarKABle feature of Kir2.3 is its activation by intracellular as well as extracellular alkalization (pKa = 6.76 and 7.4, respectively) (106, 634, 895). The determinant of this pH sensitivity is a single His residue (H117 in Kir2.3) in the TM1 to H5 linker region (106). Extracellular alkalization also enhances human Kir2.4 channel current with a pKa of 7.14 (290). This subunit has His residue (H130) in the position corresponding to H117 in Kir2.3.

The activity of some Kir2.x channels can be modified by kinases. The K+ current associated with Kir2.3 but not that of Kir2.1 is suppressed by PKC activators, phorbol 12-myristate 13-acetate (PMA) and phorbol 12,13-dibutyrate (PDBu) (251, 357). Apparently, T53 in the NH2-terminal region of Kir2.3 is phosphorylated by PKC (896). Kir2.2 current can also be suppressed by PMA treatment (901). A study analyzing regulation of Kir2.x channels by α1-adrenergic receptors has recently shown that, while Kir2.3 channels are inhibited via activation of PKC, Kir2.2 channels are inhibited via activation of src kinases in a PKC-independent manner (900). Although it was initially reported that PKA activators had little effect on Kir2.1 or Kir2.3 currents (251), Kir2.1 was found to interact with A kinase-anchoring protein 79 (AKAP79). Of interest, Kir2.1 current was increased by cAMP when the channel was coexpressed with AKAP79 and treated with the phosphatase inhibitors okadaic acid or cypermethrin (118). Activation of PKA was also shown to increase Kir2.2 current (901). Physiologically, the phosphorylation of these channels by these kinases may be involved in receptor-dependent modulation of various excitable cells.

A recent study has further suggested that cytoplasmic regulatory factors such as phosphorylation and pH might modulate channel function by affecting the channel-PtdIns(4,5)P2 interaction. Among the Kir2.x subunits, Kir2.1 interacts more strongly with PtdIns(4,5)P2 than Kir2.3 does. Du et al. (146) made two point mutants, Kir2.1(R312Q) and Kir2.3(I213L), which were found to weaken and strengthen channel-PtdIns(4,5)P2 binding, respectively (146). When they compared the function of the mutant channel with their corresponding wild-type channels, they found that inhibition induced by phospholipase C (PLC)-β PLC-γ, protein kinase C (PKC), lipid phosphatases, and protons correlates inversely with the apparent affinity of the channels for PtdIns(4,5)P2.

C. Intracellular Localization

1. Channel trafficking to the membrane

Classical Kir channels possess an amino acid sequence responsible for recruitment of the channel from the endoplasmic reticulum (ER) to the cell surface. This ER export signal is conserved in all subunits of the Kir2.x subfamily and corresponds to the amino acid sequence F-C-Y-E-N-E in the cytoplasmic COOH terminus (488, 745). Disruption of this motif by mutagenesis causes retention of Kir2.1 in the cytosol. It is of note that Kir1.1 has different ER export signals (V-L-S and E-X-D) in its COOH-terminal region (488). It is therefore likely that the surface expression of channels of each Kir subfamily is regulated by different machineries.

In general, after synthesis in ER, membrane proteins are transported to the Golgi apparatus for posttranslational modification, and then carried onto the cell surface. In Kir2.1, particular parts of the NH2-terminal region have been identified to be responsible for Golgi export (744). The double mutation R44A, R46A in the NH2 terminus resulted in disappearance of Kir2.1 from the cell surface and its accumulation in the Golgi complex. In addition, Y-X-X-Φ (position 242–245 in Kir2.1) was found to be an essential element to allow Golgi export (275). Thus two independent motifs seem to regulate trafficking of Kir2.1 from Golgi complex to cell surface.

All the Kir2.x subunits have these ER- and Golgi-export signals (275, 488, 745). However, when Kir2.4 was expressed in cultured cells, it distributed mainly in the Golgi complex with little expression on the cell surface. Mutagenesis assays revealed that the flanking sequence upstream of the Y-X-X-Φ motif in Kir2.4 would interfere with its trafficking (275).

Some of the loss-of-function mutations in Kir2.1 that are linked to Andersen syndrome (see sect. iiF and Table 2) impair trafficking to the cell surface. They are V302M, Δ95–98 in the TM1 region, and Δ314–315 in the COOH-terminal region (622, 787), and they lead to Kir2.1 being retained in the cytoplasm (44; but see Ref. 487 and sect. iiB for the role of V302M in inhibition of currents through the K+-conduction pathway). Therefore, the export signals described above are not alone in determining a channel's trafficking to the cell surface.

Activity of Kir2.1 on the cell surface may be negatively controlled by its internalization, which depends on small GTPase Rho family proteins. Kir2.1 exhibits a high degree of internalization mediated by dynamin, a protein crucial for endocytosis (59, 447). A dominant negative mutant of RacI belonging to Rho family, when coexpressed with Kir2.1 in HEK293T cells, doubly increases K+ current by inhibiting its internalization, and this effect is likely to involve disruption of function of dynamin (59). The RacI-dependent regulation is detectable in neither Kir2.2 nor Kir2.3 and thus specific to Kir2.1. Although a previous study showed that another small GTPase, RhoA, might be crucial for internalization of Kir2.1 in tsA201 cells (345), a recent report has been unable to reproduce the same result (59). This discrepancy would be due to difference of the cell types.

2. Localization in the membrane macrodomains

Kir2.x subunits are identified in specific locations in cells such as neurons and epithelial cells. They occur in somata and dendrites of neurons (302, 633). In mouse olfactory bulb, Kir2.3 is specifically expressed at the postsynaptic membrane of dendritic spines of granule cells, which receive mostly excitatory synaptic inputs (302). The postsynaptic localization of Kir2.x subunits may be determined by mechanisms such as protein-protein interactions. The PDZ-proteins are known to bind various molecules such as glutamate receptors and Kv channels and target them to postsynaptic sites (99, 109, 370, 555, 710). Interaction occurs between the PDZ domains and consensus motifs at the COOH-terminal end of the receptors and channels. Since Kir2.x subunits possess a class I PDZ domain recognition sequence (X-S/T-X-V/I) (736), PDZ-proteins are strong candidates for determining postsynaptic localization of Kir subunits. Several studies support this idea by demonstrating the following: 1) Kir2.1 can interact with PSD-95 and PSD-93/chapsyn110 (99, 447, 555); 2) Kir2.2 can associate with SAP97, CASK, Velis, and Mint-1 (441, 442); and 3) Kir2.3 can bind to PSD-95, PSD-93/chapsyn110, and SAP97 (99, 302, 442, 555). Notably, association of PSD-95 with Kir2.3 causes reduction of its single-channel conductance (555), which could contribute to control of synaptic excitability in the brain. Since these PDZ-proteins simultaneously gather different molecules such as nitric oxide(NO) synthase (66, 67), SynGAP (371), and AKAPs (100), they may not only determine the postsynaptic localization of Kir2.x channels but also make up functional units that mediate efficient signaling. However, much of this regulation via protein-protein interaction remains to be demonstrated in vivo.

When expressed in cultured epithelial cells such as Madin-Darby canine kidney (MDCK) and pig kidney epithelial cell line LLCPK, Kir2.3 is sorted to the basolateral membrane surface (430, 431, 595). In kidney, functional Kir2.3 was found at the basolateral membrane of epithelial cells in collecting ducts (430, 532, 827). At least two processes are thought to be involved in the basolateral targeting of Kir2.3. The first step may be coordinated by PDZ-proteins. In MDCK cells, truncation of the PDZ-binding motif of Kir2.3 (E-S-A-I) resulted in its missorting to intracellular vesicle compartments that contained recycling endosomes (595). Biochemical assays revealed that Kir2.3 interacts with a PDZ-protein, Lin-7, which is expressed at the basolateral membrane and also associates with another PDZ-protein, CASK (9, 595). Collectively, this interaction would retain Kir2.3 at the membrane surface. The second step is mediated by a membrane targeting sequence in the COOH-terminal region of Kir2.3. It was found that deletion of 11 amino acids, which neighbor the PDZ-domain binding motif at the COOH-terminal end, mislocates Kir2.3 from basolateral to apical membrane (431). This sequence arrangement is unique to Kir2.3, and therefore, this channel could utilize a particular mechanism to determine its localization in these epithelial cells. It is still not known why a deletion mutant that lacks both of the PDZ and membrane targeting motifs (15 amino acid truncation) is not internalized and remains at the apical membrane surface (431). Other binding proteins and/or posttranslational modification may be involved in the trafficking process of Kir2.3. Further studies are needed to clarify the mechanism determining the localization of Kir2.3 in epithelial cells.

D. Physiological Functions in Cells and Organs

1. Heart

The classical Kir current IK1 is abundantly expressed in cardiac myocytes, including Purkinje fibers as well as ventricular and atrial tissues (41, 406, 514, 575, 667) but not in nodal cells (578). This conductance dominates the resting conductance of these tissues and is defined as a time-independent background K+ current. It exhibits a large inward current at the Em more negative than EK and a relatively large outward current at the Em slightly more positive than EK. This feature is essential to stabilize the Eres of cardiac myocytes near EK. But, due to inward rectification, as the membrane is further depolarized the IK1 current progressively decreases. This characteristic generates a region of so-called negative slope of the conductance of IK1. The current becomes practically zero at positive Em (Fig. 6A). The lack of outward conductance through IK1 at positive potentials prevents K+ efflux during the action potential plateau, resulting in the maintenance of depolarization (Fig. 6, B and C). When repolarization is initiated by activation of the Kv channels, as Em repolarizes through the negative slope region relatively large outward currents pass through IK1, which accelerates the final stage of repolarization. Hence, IK1 is critically involved in determining the shape of the cardiac action potential, namely, 1) setting the resting potential, 2) permitting the plateau phase, and 3) inducing rapid final stages of repolarization (Fig. 6).

Fig. 6.

Excitability of cardiac myocytes. A: current-voltage relationship and the action potential in the ventricular myocyte. The current-voltage relationship of the instantaneous current is indicated by the solid line and that for the steady-state current at 1 s by the broken line (top panel). The action potential (bottom panel) is drawn sideways to relate it to the current-voltage relationship. Gray area indicates IK1 current. B: regional variation in action potentials in the heart. Their relation to a typical electrocardiogram is indicated at the bottom. C: comparison of action potentials in different heart regions. Eres in sinoatrial (SA) and atrioventricular (AV) nodes is not as hyperpolarized as in other regions. The Em in SA node shows a “phase 4” depolarization that triggers the next action potential. In SA and AV nodes, the rate of rise of the action potential is slow. The SA action potential stimulates atrial myocytes. Atrial myocytes exhibit a hyperpolarized Eres of approximately −90 mV without automatic excitation and a relatively short action potential. Eres of His bundle and Purkinje fibers is also hyperpolarized to approximately −90 mV. Their action potential duration is significantly longer than that of atrial myocytes. Ventricular myocytes show hyperpolarized Eres of about −90 mV and yield action potentials with a sustained plateau phase. The relative amounts of IK1, which is made up of classical Kir2.1 channels, are indicated below. SA and AV nodes express little IK1. IK1 is involved in the formation of deep Eres, the sustained plateau, and rapid repolarization.

In the heart, analyses of mRNA transcripts by in situ hybridization histochemistry, RNase protection assay, and RT-PCR demonstrated that cardiomyocytes express Kir2.1, Kir2.2, and Kir2.3 (37, 63, 134, 461, 638, 821, 830). In contrast, Kir2.4 is restricted to neuronal cells (461).

The precise molecular make up of the cardiac IK1 channel in mouse has recently been identified by the analysis of Kir2.1 and Kir2.2 knockout animals. Whereas Kir2.2 knockout mice showed ∼50% reduction in the IK1 current, no detectable current was observed in Kir2.1 knockout mice (881). It is therefore likely that Kir2.2 subunits assemble with Kir2.1 to form the IK1 current. Also, it is possible that homomeric Kir2.2 channels which form functional K+ channels in heterologous expression systems are nonfunctional in native cardiac myocytes. In rabbit ventricular myocytes, expression of the dominant negative form of either Kir2.1 or Kir2.2 suppressed ∼70% of the IK1 current, supporting the idea that IK1 is generated by heteromeric Kir2.1/2.2 channels (902). These results suggest that among the Kir2.x subfamily, it is Kir2.1 that may be the core subunit generating the IK1 current. The cardiac phenotypes observed in Kir2.1-knockout mice are as follows (see also sect. iiF). The majority of ventricular myocytes isolated from wild-type mice were quiescent with Eres approximately −72 mV. In contrast, most of the cells from Kir2.1 knockout mice showed spontaneous rhythmical action potential firing so that it was difficult to measure a true Eres. In addition, Kir2.1 knockout myocytes exhibited significantly broader action potentials than wild-type myocytes. Such phenotypes were not observed in Kir2.2 knockout mice. On the other hand, the electrocardiograms of Kir2.1 knockout mice showed neither ectopic beats nor reentry arrhythmias, indicating that their cardiac abnormalities did not include alteration of the sinus pacing of the heart. Surprisingly, Kir2.1 knockout mice consistently had a slower heart rate, which was attributed to an indirect effect as the result of disruption of the channel activities in organs other than the heart. Similar phenotypes were detected in animals that overexpressed a dominant negative form of Kir2.1 (518, 526). Notably, all Kir2.1-knockout animals exhibited the complete cleft of the secondary palate. This abnormality prevented the pups from nursing and causes the aspiration of oral secretions. The mice died within 12 h after birth because of the respiratory problems (880).

The molecular constituents of guinea pig IK1 has remained uncertain. The single-channel conductance of IK1 in ventricular cells of guinea pig heart is reported to be 20–40 pS when pipette solution contains 145–150 mM [K+] (406, 507, 679, 716). A detailed study compared the single-channel properties of cloned Kir2.x subunits and IK1 in native cells (461). Guinea pig myocytes display three populations of Kir channels with mean conductance of 34.0, 23.8, and 10.7 pS (461). The cloned guinea pig Kir2.1, Kir2.2, and Kir2.3 channels showed conductances of 30.6, 40.0, and 14.2 pS (461). The Ba2+-block profile of the native 34.0-pS channel was virtually identical to that of the cloned Kir2.2. Consistently, hyperpolarization of Em decreases Po of cardiac myocytes of guinea pig (406), which is the characteristic observed only in cloned mouse Kir2.2 but neither mouse Kir2.1 nor Kir2.3 (315, 758). Another study identified a heteromeric assembly of Kir2.1/2.3 in guinea pig cardiomyocytes by biochemical assays (626). Further studies are needed to clarify the precise subunit components of guinea pig IK1.

2. Blood vessels

The major constituents of vasculature are endothelial and smooth muscle cells. Electrophysiological studies have demonstrated that both cell types express classical Kir channels (2, 570). In vascular endothelial cells, they are considered to be the most prominent channels (570, 571, 811). Functional expression of classical Kir channels sets the Eres of endothelial cells to a negative potential, which provides the driving force for Ca2+ influx through Ca2+-permeable channels. Indeed, block of endothelial Kir channels by Ba2+ inhibits not only flow-induced Ca2+ influx but also vasodilatation caused by Ca2+-dependent production of NO (420, 828). Since an increase in intracellular Ca2+ concentration ([Ca2+]i) activates enzymes such as NO synthase and phospholipase A2 (PLA2) and thus induces the secretion of vasoactive factors, control of Em by classical Kir channels in endothelial cells is one of the key regulatory systems for vascular tone. In support of this idea, shear stress induced by laminar flow evokes classical Kir currents in aortic endothelial cells (591), which would hyperpolarize the cells and relax smooth muscle by secretion of NO (198, 523, 608, 803). The molecular make up of endothelial Kir channels has not been fully elucidated. In aortic endothelial cells, whereas Kir2.1 and Kir2.2 proteins are expressed, Ba2+ and pH sensitivity and single-channel conductance analyzed by patch-clamp techniques suggested Kir2.2 as the dominant channel (166). It would be of interest to examine the endothelial function of Kir2.1 and Kir2.2 knockout mice.

In vascular smooth muscle cells, it has been suggested that classical Kir channels might contribute to vasodilation in response to an increase in [K+]o (152, 374, 515, 516, 556). Although elevated [K+]o usually depolarizes smooth muscle cells and is expected to constrict blood vessels, a mild increase in [K+]o from 6 to 15 mM hyperpolarizes Em and dilates cerebral and coronary arteries (374, 516, 556). The Eres of the arteries' muscle cells is known to be about −45 mV, and an elevation of [K+]o to 15 mM hyperpolarized the Em to about −60 mV by increasing the Kir conductance (374, 556). This hyperpolarization closes voltage-gated Ca2+ channels and therefore reduces [Ca2+]i, which results in vasodilation (373). Ba2+ blocked both K+-induced hyperpolarization and vasodilation of coronary and cerebral arteries (374). In the cerebral arteries, it was reported that [K+]o surrounding the smooth muscle cells was elevated due to K+ secretion from the end feet of astrocytes (174), and disruption of the system by various pathological factors may cause neuronal disease. The local augmentation of [K+]o occurred when neurons were stimulated. This system may therefore couple neuronal activity to control of local blood flow in the brain (174).

In vascular smooth muscle cells, Kir2.1 but neither Kir2.2 nor Kir2.3 was identified (61). Blood vessels in Kir2.2 knockout mice dilated normally in response to high [K+]o stimulation but not vessels from Kir2.1 knockout mice (880). Therefore, Kir2.1 seems to be a main subunit to form the classical Kir current in these cells.

Activity of classical Kir channels is also important for relaxation and contraction of tracheal smooth muscle and is controlled by autacoids. It has been found that neurokinin A, a type of tachykinin, inhibits classical Kir channels in tracheal smooth muscle cells (550), although molecular composition of the channels has remained unknown. This process depolarizes the cells and induces tracheal contraction, which may be involved in pathogenetic process of asthma.

3. Neurons in the brain

Electrophysiological techniques detected Kir currents in hippocampal neurons (68) and neonatal rat spinal motor neurons (759). The currents, which are generated by various tetramers of Kir2.x subunits, are considered to be critically involved in the maintenance of Eres and regulation of the excitability of the neurons. Indeed, Ba2+ block of Kir channels in an isolated neuron caused depolarization and initiated action potential firing (121).

In situ hybridization histochemistry and immunohistochemistry show classical Kir channels to be abundantly and differentially expressed in the brain. Kir2.1 is expressed diffusely and weakly in the whole brain, Kir2.2 moderately throughout the forebrain and strongly in the cerebellum, Kir2.3 mainly in forebrain and olfactory bulb, and Kir2.4 in the cranial nerve motor nuclei in the midbrain, pons, and medulla (280, 302, 360, 633, 782). Expression of classical Kir channels is restricted to neuronal somata and dendrites (302, 633). For example, in mouse olfactory bulb, Kir2.3 is specifically localized at the postsynaptic membrane of dendritic spines of granule cells, which receive mostly excitatory synaptic inputs (302) (and see sect. iiC). The spinal localization of Kir2.3 is also detected in striatopallidal neurons. In this region, the current emerging from Kir2.3-containing channels is significantly suppressed by M1 muscarinic receptor stimulation through depletion of PtdIns(4,5)P2, which results in enhancement of dendritic excitability (708).

Function of brain classical Kir channels may be further modulated by neurotransmitters. The classical Kir channels reconstituted with injection of brain poly(A)+ RNA into Xenopus oocytes are shown to be inhibited by isoproterenol, a β-adrenergic agonist (322). Pharmacological assays imply that this effect is exerted by increase of intracellular cAMP and also cGMP but not by activation of PKA (322). Therefore, β-adrenergic inhibition of brain classical Kir channels seems to be mediated by phosphorylation-independent action of cyclic nucleotides. However, the physiological role of this signaling machinery has not yet been clarified in the brain.

Kir2.1 and Kir2.2 knockout mice show no obvious phenotypes related to neuronal abnormality (880). Since in many cases neurons express multiple types of Kir channel subunits (633), loss of one Kir2.x protein could be compensated by another subunit(s) in the mutant mice.

It is of interest that, in Schwann cells surrounding peripheral nerve fibers, a Kir current has been recorded (842). Immunohistochemical analysis identified that Kir2.1 and Kir2.3 were specifically expressed in the microvilli of Schwann cells at the nodes of Ranvier (525). Since the villi are facing towards the axon, these Kir channels may play a role in maintaining [K+]o by absorbing excess K+ released from excited axons. This function is similar to the “K+-buffering” action of astroglial K+ channels and may therefore be important for sustaining proper function of nerve fibers (see sect. vB).

4. Skeletal muscle

A major determinant of the Eres in skeletal muscle is a Cl conductance (274). Classical Kir channels also participate in setting the Eres and shift it toward the direction of hyperpolarization. All types of Kir2.x subunits are expressed in skeletal muscle at the mRNA level (394, 614, 638, 758, 782). The importance of Kir2.1 in muscle function was highlighted by analysis of Andersen's syndrome (622) (see sect. ii, B and F, and Table 2). This disease is accompanied by periodic paralysis. In this case, reduction of the Kir2.1 conductance should depolarize the Eres, which would inactivate Na+ channels and thus make them unavailable for the initiation and propagation of action potentials.

The functional expression of Kir2.1 seems to be necessary not only for differentiation of myoblasts (387) but also for the fusion of mononucleated myoblasts to form a multinucleated skeletal muscle fiber (182). Both events are Ca2+-dependent processes and essential for skeletal muscle development, growth, and repair. During development, the Eres of myoblasts gradually hyperpolarizes from about −10 to −70 mV (462), due to the developmental increase in Kir2.1 expression (182). This Kir2.1-induced hyperpolarization sets Em in a range where Ca2+ can enter the myoblasts through Ca2+-permeable channels, which promotes the differentiation and fusion of myoblasts. In addition, differentiation is at least in part mediated by expression and activity of myogenic transcription factors, such as myogenin and myocyte enhancer factor-2. Their expression is triggered by Kir2.1-induced hyperpolarization (387). Although abnormality in morphology of skeletal muscles is not obvious in either Andersen's syndrome patients or Kir2.1-null mice, the “slender” build observed in this patient group could be attributed to mild impairment of muscle development caused by dysfunction of the channel (346).


The cortical collecting duct (CCD) is a major site of K+ secretion in kidney. Patch-clamp recording showed that apical and basolateral membranes of CCD express distinct types of inward rectifiers. Whereas apical channels displayed mild inward rectification with an intermediate conductance (20–45 pS) (192, 820), basolateral channels showed strong rectification with a small conductance (27–30 pS) (481). Such an asymmetric profile of Kir channel expression in epithelial membranes is considered to be important for vectorical K+ transport from the basolateral to apical side of the epithelium. Basolateral Kir channels, which are made up of Kir2.3 (430, 827), maintain the potential difference across the basolateral membrane and a favorable electrical driving force for apical K+ exit without significantly recycling K+ back to the interstitium (247, 827). The apical K+ conductance is yielded by Kir1.1 and plays a central role in secreting K+ to urine and sustaining activity of the Na+-K+-2Cl cotransporter which is colocalized with the channel (see sect. vA4) (247).

Another subunit, Kir2.1, is found specifically in juxtaglomerular cells but in neither epithelial cells nor glomeruli (438). Kir2.1 thus seems to be a major determinant for Eres of approximately −60 mV in juxtaglomerular cells (70, 183).

E. Pharmacology

Specific blockers and activators for classical Kir channels are not known. In heterologous expression systems, the homomeric Kir2.x channels differ substantially in their sensitivity to Ba2+ and Cs+ (Table 1). In particular, Kir2.4 was much less sensitive to these blockers than other subunits (Ki values for Ba2+ and Cs+ of 390 μM and ∼8 mM for Kir2.4, respectively; ∼8 μM and 420 μM for Kir2.1) (782) (Table 1). In addition, when K+ currents generated by coexpression of different Kir2.x subunits or expression of concatamers (tandem subunits) were analyzed, it was found that their Ba2+ sensitivity clearly differed from that of homomeric Kir2.x channels (Table 1) (626). These findings suggest that heteromeric assembly of Kir subunits elicits unique electrophysiological properties that may be involved in the functional roles of the channels.

Recent pharmacological and electrophysiological studies have demonstrated that several reagents block Kir2.x channels, although their effects are not specific to them. The first-generation blockers of histamine H1 receptors (mepyramine and diphenhydramine) but neither second- nor third-generation drugs moderately inhibited Kir2.3 current by ∼25 and ∼17% at 100 μM, respectively (460). Kir2.3 current was also partially (∼50% maximum) reduced by genistein, an inhibitor of protein tyrosine kinases, with an IC50 of 16.9 and 19.3 μM when the channels were expressed in Xenopus oocytes and HEK293T cells, respectively (891). These two types of reagents had little effect on Kir2.1 currents (460, 891). Analyses of chimera of Kir2.3 and Kir2.1 suggested that transmembrane regions (TM1 and TM2) and the pore region but not cytoplasmic domains were involved in the action of genistein.

Choloroquine is an important drug for the treatment of malaria. It is a serious problem that this drug induces lethal ventricular arrhythmias mainly by inhibiting cardiac IK1 currents (43, 681). This side effect is attributable to block of Kir2.1 channels. Mutation analyses of Kir2.1 mapped the binding site of chloroquine, unlike most ion channel antagonists, to the cytoplasmic domain rather than the transmembrane pore (658). Molecular modeling implies that chloroquine would plug the cytoplasmic conduction pathway by interacting with E224, E299, and D259 in the cytoplasmic pore region (see Fig. 5) which contains the polyamine binding sites.

A class Ia antiarrhythmic agent, quinidine, is also known to block cardiac IK1 current with IC50 of 4–40 μM (268, 558). This action could be involved in the clinical effects of the drug. The binding site for quinidine in Kir2.x subunits has not been examined.

F. Diseases

1. Andersen's syndrome (LQT type 7)

The finding of the mutations responsible for Andersen's syndrome (Table 2) has had a great impact on electrophysiological and pathological studies of Kir channels and provided a novel insight into channelopathies. Andersen's syndrome is accompanied by cardiac arrhythmias reminiscent of long Q-T syndrome (LQT7), periodic paralysis, and dysmorphic bone structure in the face and fingers (771). The syndrome is an autosomal dominant disorder, and it is caused by mutations in the KCNJ2 gene which encodes the Kir2.1 subunit (622). Some of the mutations elicit dominant negative effects on the K+ current (44, 429, 477, 622, 787) by impairing the interaction between the channels and PtdIns(4,5)P2 (474) or by inhibiting trafficking of Kir2.1 to cell membrane surface (44) (see sect. ii, B and C). The symptoms of Andersen's syndrome in the heart are caused by reduction of Kir2.1 function that prolongs the plateau phase of the action potential and depolarizes Eres. It may unstabilize Em and trigger arrhythmias. The possible mechanism underlying periodic paralysis is discussed in section iiD. Abnormal bone structure in the disease would be caused by provoking dysfunction of osteoclasts. Low extracellular pH in the extracellular matrix maintained by H+ secretion via an ATP-driven proton pump is critical for proper degradation of bone by osteoclasts. Since this H+ secretion is achieved in exchange for K+ transport through Kir channels, disruption of these channels would cause osteoclast dysfunction that could lead to severe bone deformity.

2. Kir2.1 and Kir2.2 knockout mice

Kir2.1 and Kir2.2 knockout mice have been generated and their phenotypes have been studied. These studies have clarified the point that in smooth muscle Kir2.1 rather than Kir2.2 is an important regulator of vascular tone (880) (see sect. iiD2). In addition, the knockout mice study has revealed that Kir2.1 is a major constituent of the IK1 current in heart, although Kir2.2 also contributes to the formation of the current by assembling with Kir2.1 (881) (see sect. iiD). The phenotypes of blood vessels and heart in Kir2.1 knockout mice were described in section iiD.

Kir2.1 knockout mice mimic some of the symptoms of Andersen's syndrome. The mice exhibit a narrow maxilla and complete cleft of the secondary palate (880). These phenotypes may correspond to the facial dysmorphology observed in humans. Although isolated ventricular myocytes of the knockout mice showed longer action potentials and prolonged Q-T interval, no arrhythmia was observed in the mutant mice.

3. Other diseases

A recent study has identified another abnormality of the Kir2.1 gene that causes a cardiac disease. The short Q-T syndrome constitutes a new clinical criterion that is associated with a high incidence of sudden cardiac death, syncope, and/or atrial fibrillation even among young patients and newborns (688). Genetic analysis of a family with this syndrome found a G514A substitution in the KCNJ2 gene that resulted in a change from Asp to Asn at position 172 (D172N) (628) (Table 2). D172, which is close to the distal end of the second transmembrane region (TM2), is a residue crucial for establishment of the strong rectification profile of Kir2.1 (see sect. iiB). Electrophysiological assays in transfected Chinese hamster ovary (CHO) cells demonstrated that the mutant exhibited a larger outward current than the wild type. In silico simulation suggests that the abnormality caused by this Kir2.1 mutation produces an abrupt increase in the rate of final repolarization of the ventricular action potential and shortens its duration (628).

As described in section iiD3, when M1 receptor in striatopallidal neurons of brain is stimulated, activity of Kir2.3-containing channels expressed at these neurons can be significantly suppressed. This event leads to elevation of dendritic excitability and opening of voltage-gated Ca2+ channels, which further seems to cause loss of spines. Several observations imply that this process is involved in abnormality of neuronal network found in Parkinson's disease (519) as follows (708). First, depletion of neuronal dopamine by administration of reserpine, the procedure making model animals of Parkinson's disease, not only induces loss of spines in striatopallidal neurons but also causes increase of mRNA of Kir2.3 and decrease of mRNA of Kir2.1 that is also expressed in these neurons. Second, because Kir2.3 activity is more sensitive to reduction of membrane PtdIns(4,5)P2 content than Kir2.1, the change of expression profile of Kir subunits by reserpine significantly augments the Kir current fraction suppressed by M1 receptor stimulation. Third, in M1 receptor-null mice, the dopamine depletion by reserpine barely affects number of their spines. Fourth, injection of 6-hydroxydopamine into the brain, the procedure which destroys dopamine neurons and thus mimics Parkinsonism, elicits loss of spines in wild-type mice but has little effect on their number in M1 receptor-null mice. Therefore, modulation of function of Kir2.3-containing channels by M1 signaling may be critically involved in neuronal phenotypes of Parkinson's disease (708).

Interestingly, the cholesterol content of a membrane affects the function of classical Kir channels, and this could be involved in some diseases (165, 662, 663). The application of cholesterol reduced the Kir2.1 current density in CHO cells and native classical Kir channels in bovine and human aortic endothelial cells (662, 663). More importantly, analysis of hypercholesterolemic pigs revealed that an increase in plasma cholesterol levels strongly reduced endothelial Kir currents and depolarized the Eres (165). Thus suppression of classical Kir currents may be a factor not only in hypercholesterolemia-induced endothelial dysfunction but also in various vascular diseases.

III. G Protein-Gated Kir Channels (Kir3.x)

A. Historical View and Molecular Diversity

The human genome encodes thousands of GPCRs. They are involved in a wide variety of physiological functions. The activation of a GPCR by its ligand (hormone or neurotransmitter) results in the liberation of two intracellular effector molecules Gα and Gβγ that can influence a large number of cellular processes. KG channels are one of the targets of GPCRs (65, 411, 412, 620). In the heart, muscarinic K+ (KACh) channels, a type of KG channels, are responsible for the effects of ACh and adenosine (Fig. 7, A and B). Stimulation of these GPCRs results in channel opening which hyperpolarizes the cells. Similar signal transduction mechanisms are used by different receptors, including somatostatin, 5-hydroxytryptamine-1, α2-adrenergic, μ- and δ-opioid, D2-dopamine, glutamate, and GABAB receptors in neurons and hormone-secreting cells as well as smooth muscle cells (60, 226, 265, 423, 452, 542, 581, 584, 669, 684, 705, 851).

Fig. 7.

Activation of KG channels by G protein-coupled receptors and G proteins. A: the time-dependent response to 11 μM acethylcholine (ACh) of a whole cell KG channel current in a guinea pig atrial myocyte. In normal extracellular solution which contained 5.4 mM K+, the cell membrane potential was voltage clamped at −53 mV. The pipette contained (in mM) 150 KCl, 2 MgCl2, 5 EGTA, 5 HEPES, and 0.1 GTP (pH 7.4). [From Kurachi et al. (415), with kind permission of Springer Science + Business Media.] B: activation of KG channels by ACh or adenosine (Ado) requires GTPi and is blocked by islet-activating protein (IAP, pertussis toxin). The single-channel recording began in the cell-attached patch configuration with a pipette solution containing either ACh (1.1 μM; top panel) or Ado (10 μM; middle panel). The inside-out patch was formed (heavy arrows) and channel activity declined, then GTP (100 μM) was applied to the intracellular surface. In the bottom panel, KG channels were recorded in an inside-out membrane patch in the presence of GTP (100 μM). The channels were exposed to a mixture of the A protomer of IAP (1 μg/m) and NAD (1 mM). The holding potentials are indicated to the left of each panel. [From Kurachi et al. (412), with kind permission of Springer Science + Business Media.] C: a cartoon representing activation of a KG channel. On the atrial cell membrane, two different GPCRs (P1-purinergic and muscarinic M2 ACh receptors) activate the channel via GTP-binding proteins GK (412).

The KG channels can be activated by intracellular GTP (GTPi) in the presence of agonist or by intracellular GTPγS even in the absence of agonist (Figs. 7B and 8A). This can occur in cell-free, inside-out membrane patches in a “membrane-delimited” manner, suggesting that channel opening is triggered by membrane-bound G proteins (Fig. 7, B and C) (411, 412, 414). After a long controversy, it was finally established that KG channels are activated not by Gα but by Gβγ subunits of PTX-sensitive G proteins (Figs. 7B and 8A) (96, 323, 409, 416, 465, 833, 850, 854, 869, 870).

Fig. 8.

Regulation of KG channels by different G protein subunits and nucleotides. A: activation of KG channels by GTP and Gβγ. KG channels in atrial cardiac myocytes were recorded in inside-out patches (excision at vertical arrows) in the presence of ACh in the pipette solution (see inset). Nucleotides and G protein subunits were applied to the intracellular surface of the excised patches as indicated above each recording. a: Effects of different concentrations of Gβγ, GTPγS-bound Gα subunits, and GTPγS. b: GDP-bound form of the transducin α subunit (Tα-GDP) was applied to channels activated by GTP and ACh. Note: Tβγ opened the channels. c: Tα-GDP was applied to channels preactivated by GTPγS. [From Yamada et al. (850), with permission from Elsevier.] B: concentration-dependent effects of GTPi on KG channel in the absence and the presence of ACh. a: Examples of inside-out patch recordings obtained from guinea pig atrial myocytes. The channel currents were recorded at −80 mV in symmetrical 145 mM K+ solutions. Above, no ACh in the pipette; below, 1 μM ACh. The bars above each trace indicate the periods of application of different concentrations of GTP and 10 μM GTPγS to the internal surface of the patch membrane. b: Relationship between the concentration of GTP and the NPo of KG channels. The data were normalized to the maximum NPo which was induced by 10 μM GTPγS in each patch. Symbols are means. The continuous lines indicate the fit of the relationship between GTP and channel activity in the presence of each concentration of ACh with the following Hill equation: f = Vmax/{1 + (Kd/[GTP])n}, where f is the relative NPo, Vmax is the maximum NPo available in the presence of 10 μM GTPγS, Kd is the apparent dissociation constant of GTP, and n is the Hill coefficient. [From Ito et al. (320), copyright 1991. Originally published in The Journal of General Physiology.] C: relationship between GTP concentration and the fraction of the “available” state in KG channels according to the concerted allosteric model of Monod, Wyman, and Changeux. a: The fraction of the “available” state [A/(A + U)] was estimated from inside-out membrane patch experiments. Symbols represent data concerning the relationship between GTP concentration and the calculated fraction of the “available” state. Lines indicate fits to the data of the Monod, Wyman, and Changeux allosteric model with different assumed values of n. b: A schematic representation of the Monod, Wyman, and Changeux allosteric model. In this scheme, each channel is assumed to be an oligomer composed of four identical subunits (n = 4). Each subunit is in either the tense (T) or the relaxed (R) state represented by squares and circles, respectively. Each subunit in the T or R state binds one dissociated Gβγ subunit (solid circles) independently of each other with a microscopic dissociation constant of either KT or KR. In this model, all subunits in the same oligomer must change their conformations simultaneously. Therefore, the whole channel complex can be either T4 or R4. T4 and R4 are in equilibrium according to the allosteric constant L. [From Hosoya et al. (282), copyright 1996. Originally published in The Journal of General Physiology.]

Molecular cloning techniques have identified four KG channel subunits (Kir3.1, Kir3.2, Kir3.3, and Kir3.4). Physiological assays and studies on knockout mice have revealed that various combinations of subunits from the Kir3 subfamily form different KG channels. The activity and the function of KG channels are precisely tuned by their localization, protein-protein interactions, and a variety of small substances.

1. Molecular characterization

Functional KG channels are tetrameric assemblies of Kir3 family subunits (307, 333, 382, 391, 444, 453) (see Fig. 2B). The assembly can be either homomeric or heteromeric. The subunit composition of KG channels varies among different cells and tissues which allows them to play diverse functional roles.

Kir3.1/GIRK1/KCNJ3 cDNA was isolated from a rat heart cDNA library, and this was the first Kir subunit shown to contribute to formation of functional KG channels (119, 397). Kir3.1 shares 39 and 42% amino acid identity to Kir1.1 and Kir2.1, respectively. Although Kir3.1 may form a homomeric assembly in an overexpression system (104), expression of Kir3.1 alone does not produce functional KG channel current in mammalian cell lines (392, 621, 738, 843, 851) (see also sect. iiiC). Coexpression of Kir3.1 with Kir3.2 (443) or Kir3.4 (392) not only enhanced each of their channel currents but also induced different channel properties (150, 382, 733, 805). In Xenopus oocytes, the current could be elicited by expression of Kir3.1 alone, which was shown to be due to its heteromeric assembly with Kir3.5/XIR, an intrinsic homolog of Kir3.4 (248). Thus Kir3.1 is generally incorporated into the heteromers with other Kir3.x subunits to form KG channels in native cells and tissues.

There are at least four different isoforms of Kir3.2/GIRK2/KCNJ6. They are generated by alternative splicing from a single gene, which is composed of more than eight exons (304, 823, 836). Initially identified as GIRK2A, GIRK2B, and GIRK2C, the first three isoforms are now designated as Kir3.2a-c, respectively (315, 316, 443). Kir3.2b is 87 amino acids shorter than Kir3.2a, and 8 amino acids in the COOH terminus of Kir3.2b are different from those in Kir3.2a. Kir3.2c was initially named GIRK2A (444). This subunit was also called KATP-2, because it was isolated from an insulinoma cell cDNA library and proposed to constitute KATP channels (746). Among the splicing variants, Kir3.2c is the longest isoform: it possesses an additional 11 amino acids in the COOH-terminal end that are not observed in Kir3.2a. Another variant, Kir3.2d, was isolated from a mouse testis cDNA library (304). Kir3.2d is 18 amino acids shorter than Kir3.2c at its NH2-terminal end, which is the sole difference between the two isoforms. Kir3.2 isoforms exhibit differential expression patterns in various tissues such as brain, pituitary, pancreas, and testis, where they usually coincide with the expression of other Kir3.x subunits (307, 360, 375, 453). These histochemical observations and immunoprecipitation assays from native tissues suggest that Kir3.2 isoforms usually form heteromers with other subunits such as Kir3.1 or Kir3.3. Nevertheless, it is also true that Kir3.2 isoforms form homomeric channels in particular areas (see sect. iiiD) (304, 307).

Kir3.3 was isolated from a mouse brain cDNA library in 1994 (443). Afterward, a variant of Kir3.3 was also cloned from mouse brain (334). The lately isolated mouse Kir3.3 possesses additional 17 amino acids at COOH-terminal end and Ser-60 and Ala-77 instead of Arg and Val at the corresponding positions in the first isolated one (334, 443). The amino acid sequences of rat and human Kir3.3 subunits were almost identical to the lately isolated mouse Kir3.3 (334, 694, 843). These Kir3.3 subunits may form heteromers with other Kir3.x subunits, but their functional profile seems to be disputed. As for the first isolated mouse Kir3.3, one study reported that, when this subunit was coexpressed with Kir3.1, Kir3.2, or Kir3.4, negligible current was detectable (444). On the other hand, another study showed that a considerable KG current was observed when the first isolated Kir3.3 and Kir3.1 were expressed together, and coexpression of this Kir3.3 dramatically suppressed Kir3.2 current (382). As for the lately isolated mouse Kir3.3, some works reported that, when this subunit was coexpressed with Kir3.1 (334, 484) or with Kir3.2 (333, 484), a large KG current was detected; in these cases, single-channel property of Kir3.1/3.3 was similar to those of Kir3.1/3.2 and Kir3.1/3.4 (334) and mean open time of single channel of Kir3.2/3.3 was longer than that of Kir3.2 homomer (333). Coexpression of Kir3.1 with rat Kir3.3 whose sequence is similar to the lately isolated mouse Kir3.3 produced a considerable current in one report (843) but displayed little channel activity in another report (489), although both papers demonstrated that the expression of the Kir3.3 reduced heteromeric Kir3.1/3.2 channel activity. In addition, human Kir3.3, even when coexpressed with Kir3.1 or Kir3.2, exhibited negligible current (694). Together, the functional role of Kir3.3 and its relationship with the difference of the sequences have not yet been fully determined, although it seems probable that Kir3.3 makes a complex with other Kir3.x subunits such as Kir3.2 in in vivo brain (333).

Kir3.4/GIRK4/CIR/KCNJ5 was identified in bovine atrial membrane by immunoprecipitation with Kir3.1 antibody (391). The subunit assembles with Kir3.1 in vitro and elicits a current similar to the cardiac K+ current known as IKACh. In support of this idea, Kir3.4-deficient mice lack IKACh in the atria (835). The heterotetrameric assembly of Kir3.4 and Kir3.1 and a stoichiometry of 1:1 have been confirmed by various assays including biochemical experiments (104, 305, 391) and electrophysiological analyses (728, 794). A homotetramer of Kir3.4 was also found in atrial myocytes by biochemical assays, implying that Kir3.4 may form a homomer as well (103).

B. Pore Function and Subunit Structure

1. Activation of KG channels by Gβγ

In the absence of an agonist, Gα binds GDP, the Gαβγ complex is attached to the receptor, and GTPase activity is low (207). When an agonist stimulates a GPCR, the GDP/GTP exchange rate is accelerated, which results in functional separation of Gβγ from Gα. Although it has been reported that both Gα and Gβγ subunits are able to regulate various effectors such as adenylyl cyclases and phopholipases, KG channels are activated only by Gβγ (Fig. 8A) (96, 323, 405, 465, 652, 837).

The direct interaction between Gβγ and KG channel subunits was shown by biochemical pull down assay, yeast two-hybrid assay, and electrophysiological experiments: Gβγ binds to both NH2 and COOH termini of KG channel subunits (78, 246, 288, 289, 305, 326, 392, 402, 734, 863). Hybridization of the G protein-insensitive Kir2.1 by inserting the COOH terminus of Kir3.1 led to channel activation by Gβγ (734).

Key residues that affect Gβγ binding and Kir3.x channel activation are H64 and L262 in Kir3.4 (246, 326) and L344 and G347 in Kir3.2 (181). They correspond to H57, L262, L333, and E336 in Kir3.1 (see Fig. 4D). These residues are localized in the cytoplasmic region. The crystal structure of the cytoplasmic region of Kir3.1 has revealed that two neighboring COOH termini bind to each other and an NH2 terminus is located between them (574). In Kir3.1, H57 is located on the βA strand of the NH2 terminus, and L333 and E336 are on a loop between the βL and βM strands of the COOH terminus. These three β-strands form a β-sheet located on the external surface of the cytoplasmic domain facing the cytoplasm. Interaction of the NH2- and COOH-terminal domains of Kir3.1 and Kir3.4 synergistically enhances Gβγ binding (288, 289), and fluorescence resonance energy transfer studies have detected that Gβγ induces rotation and expansion at the NH2 and COOH termini of the Kir3.1/3.4 heteromer (655). Therefore, the cytoplasmic region of KG channels is the target for Gβγ, and this region seems to be critically involved in channel gating. The importance of the cytoplasmic region for channel function is likely to be common to all Kir channels, which is supported by the observation that Kir1.1 and KirBac3.1 are insensitive to Gβγ, but both show considerable conformational changes in their cytoplasmic regions during gating (403, 645, 697). Recently, it has been suggested that Gβγ regulates the affinity of KG channels for PtdIns(4,5)P2 binding and thus channel activity (287) (see also sect. iiiB3).

The mechanism underlying control of KG channel gating by Gβγ was studied in detail in native KACh channels in inside-out membrane patches from atrial myocytes (Figs. 7B and 8A) (320, 323, 411, 465, 850, 854). The single-channel conductance of KACh channels measured at Em more negative than EK is ∼40 pS with 145 mM [K+]o (411, 412, 414). The mean channel open time in the presence of ACh is ∼1 ms, which is several orders of magnitude shorter than the open time of IK1 channels. In addition to the short open time component, the histogram also reveals a component with a longer mean open time, although its frequency is relatively low. The closed time distribution has at least two distinct components with mean closed times of ∼1 and 100 ms (678). In addition, there are also rare very long closed events (282, 678). The addition of Gβγ but not Gα to the intracellular surface of cell membrane activates the KACh channels (Fig. 8A and see below) (465). We found that intracellular application of GTP, in the presence of extracellular ACh, or that of Gβγ, activates KACh channels in a positively cooperative manner with a Hill coefficient of ∼3 (320, 408, 850), although some studies reported that the coefficient of the relationship between KG channels' activation and Gβγ concentration was 1.5–1.7 (334, 392). Our results suggest that Gβγ binds to multiple sites on a functional channel unit. To elucidate the machinery underlying this positive cooperative activation of KACh channels by Gβγ, Hosoya et al. (282) analyzed channel kinetics at different concentrations of GTP using spectral analysis techniques (Fig. 8, B and C). The results could be accounted for by the Monod, Wyman, and Changeux allosteric model if channels were presumed to be an oligometric protein composed of four or more functionally identical subunits, each of which bound one Gβγ. This observation is consistent with the results of recent molecular biological studies and structural analysis.

Several studies report possible roles of Gα in KG channel function. Inside-out patch-clamp analysis shows that intracellular application of the GDP-bound form of the transducin α-subunit irreversibly inhibits cardiac KG current activated by either GTP with ACh (in pipette solution) or GTPγS (Fig. 8A) (850). This observation indicates that Gα would chelate Gβγ.

In addition, Gα is proposed to be involved in the control of KG channel activity. It has been demonstrated by in vitro biochemical assays that GDP-bound form of Gα can interact not only with the NH2 and COOH terminus of Kir3.1 and Kir3.2 but also with a complex of Gβγ and KG channels (327, 612, 734). These results imply that, even in the resting state (without agonist), KG channels may make a complex with GDP-bound Gα, Gβγ in the cell. This protein complex of three units (GDP-bound Gα, Gβγ, and KG channels) is called “preformed” complex. It is also proposed that, when the ligands bind to the receptor, Gα would be released from the preformed complex and simultaneously Gβγ could become active to stimulate the channels. In this condition, Gβγ may be always in very close vicinity of KG channels even without agonists, and thus could effectively and quickly activate the channels as soon as Gα is dissociated from the preformed complex. Accordingly, formation of the preformed complex is suggested to be involved in rapid activation of the KG channels in response to agonists, which is observed in native tissues (612, 654, 734). In support of this idea, in heterologous expression system, coexpression of Gα fastens activation time course of ACh-induced current of Kir3.1/3.2 channels (327). Moreover, when Gα is coexpressed with KG channels in Xenopus oocytes, the basal current decreases but the total current amplitude including agonist-induced component is unchanged or even enhanced (327, 612). It is therefore considered that, whereas coexpression of Gα may chelate Gβγ as opposed to KG channels and reduce their basal activity in the absence of agonists, it would simultaneously augment the number of the preformed complex and increase the pool of Gβγ available for the channel's activation in response to agonists. More studies will be necessary to determine the precise roles of Gα on KG channels.

2. Regulation of activity and relaxation of KG channels by regulators of G protein signaling proteins

Regulator of G protein signaling (RGS) proteins negatively regulates various G protein-mediated signal pathways by accelerating intrinsic GTPase hydrolysis in the Gα subunit (Fig. 9A) (253, 666). Recently, a role for RGS proteins in controlling KG channel activity has been reported. GPCRs, G proteins, RGS proteins, and KG channels seem to be gathered in close vicinity (95, 576, 654, 890). RGS proteins expressed in Xenopus oocytes and mammalian cell lines accelerate the time course of agonist-induced activation and deactivation of KG currents (142, 194, 676). The acceleration of the “turn-off” response by RGS proteins can be simply explained by the facilitation of the sequestration of Gβγ when RGS proteins enhance the GTPase activity of Gα and thus reduce GTP-bound Gα. On the other hand, the mechanism underlying acceleration of activation time course has not been yet clarified.

Fig. 9.

Mechanism of relaxation behavior in KG channels. A: a schematic representation of the mode of action of RGS proteins. RGS proteins stabilize the transition state (Gα-GTP) of GTP hydrolysis on the Gα subunit, which results in the acceleration of intrinsic GTPase-activity. B: effects of extracellular Ca2+ on KG currents. a: Voltage-clamp protocol (top) and a typical trace of whole cell ACh-induced KG current in an isolated atrial myocyte (bottom). Inward current on stepping membrane voltage to −100 mV first changed instantaneously (Iins) and then slowly increased to a steady state (Imax). b: KG currents evoked by 10−7 M (left) or 10−6 M (right) ACh in control conditions. Currents at −100 mV were recorded after prepulses to between −100 and 40 mV in steps of 20 mV (inset). c: Relationship between prepulse voltage and Iins/Imax ratio for currents elicited by either 10−7 M (open circle) or 10−6 M (solid circle) ACh (n = 10). d and e: KG current evoked by 10−7 M ACh when either extracellular free Ca2+ was chelated by EGTA (d) or intracellular Ca2+ was chelated by BAPTA (e). f: Relationship between prepulse voltage and Iins/Imax ratio with intracellular BAPTA. KG currents were elicited by 10−7 M (open circle) and 10−6 M (solid circle) ACh (n = 8). In each current trace, arrowheads indicate the zero-current level; vertical scale bars represent 500 pA. C: a schematic representation of voltage-dependent relaxation of KG resulting from Ca2+/CaM-dependent facilitation of the action of RGS proteins. In a hyperpolarized state, the action of RGS is inhibited by PtdIns(3,4,5)P3. Once the intracellular Ca2+ concentration is elevated, e.g., upon depolarization, Ca2+/CaM binds to RGS proteins and reverses the inhibitory effect of PtdIns(3,4,5)P3, which results in the negative regulation of the G protein cycle. When the [Ca2+] decreases to the steady-state level again, CaM dissociates from RGS proteins and their action is once again inhibited by PtdIns(3,4,5)P3.x is an unknown Ca2+ transport apparatus. [A and B from Ishii et al. (313).]

We have found that RGS proteins can also be responsible for the relaxation behavior of cardiac KG channels. The cardiac KG current recorded during a hyperpolarizing voltage step is composed of two kinetically distinct processes, instantaneous and time-dependent components (Fig. 9B). The time-dependent alteration of KG currents at a given potential is called “relaxation.” It reflects the gradual increase in Po during the hyperpolarizing voltage step. The instantaneous increase of current on hyperpolarization is seen in almost all Kir channels. It is due to a relief from the blockade of the channel pore by intracellular cations such as Mg2+ and polyamines. The subsequent time-dependent “relaxation” is a characteristic of KG channels. Interestingly, this characteristic depends on the concentration of agonist as well as on Em (851).

Relaxation behavior of KG current was first described in sinoatrial node cells (579), but the mechanism has remained unknown for a long time. We have found that the relaxation of KG currents reconstituted in Xenopus oocytes with Kir3.1/3.4 and the M2 muscarinic acetylcholine receptor emerges when RGS proteins are also expressed (194) and that the interaction of the RGS domain with the Gα subunit mediates the phenomenon (303) (Fig. 9, A and C). Then it was found that the apparent voltage dependence of relaxation is caused by voltage-dependent Ca2+ influx across the cell membrane (Fig. 9B) (313). In the resting conditions, RGS protein activity is inhibited by binding of PtdIns(3,4,5)P3. On depolarization, Ca2+ enters into atrial myocytes and forms a complex with calmodulin (CaM). Ca2+/CaM complex then interacts with RGS to remove PtdIns(3,4,5)P3 in a competitive manner. The activated RGS then accelerates the hydrolysis of GTP on Gα, reduces the amount of free Gβγ, and decreases the activity of KG channels (Fig. 9C) (303, 314). The voltage step to a hyperpolarized voltage stops Ca2+ entry, decreases Ca2+/CaM complex, and allows reassociation of RGS protein, and PtdIns(3,4,5)P3, reduces GTPase activity of Gα, increases the amount of free Gβγ, and enhances KG channel activity. “Relaxation” thus may represent the time course of a series of biochemical reactions and not a gating process intrinsic to the KG channel.

3. Other modulators of KG channel gating


Like other Kir channels, KG channels require PtdIns(4,5)P2 to maintain their activity (287) (see also sects. iD and iiB). When PtdIns(4,5)P2 is removed in inside-out patches, KG current runs down completely (750). The channels in this condition cannot be opened by addition of Gβγ alone. Application of PtdIns(4,5)P2 slowly causes openings of the channels over a time scale of minutes. Unlike most Kir channels, KG channels' activation by PtdIns(4,5)P2 is significantly augmented and accelerated by additional gating molecules such as Gβγ and Nai+, which seem to strengthen the interaction between the channels and PtdIns(4,5)P2 (146, 287, 750). Consistently, the coexpression of Gβγ with Kir3.1/3.4 channels dramatically slows the rate of the loss of channel activity induced by a PtdIns(4,5)P2 antibody (287). The loop between βC- and βD-strands in the cytoplasmic domain is shown to be involved in controlling the affinity of PtdIns(4,5)P2 (889) as well as Nai+ (see below) to KG channels.

Nai+ can activate KG channels which contain Kir3.2 or Kir3.4 subunits (270, 271, 444, 619, 749, 889). Asp in the CD loop between βC-and βD-strands (D228 in Kir3.2 and D223 in Kir3.4) is considered to be the Nai+ sensor (270, 889). The corresponding residue in Kir3.1 is N217 (see Fig. 4D). This supports the notion that Asp in the CD loop plays an important role in Nai-induced activation of KG channels containing Kir3.2 or Kir3.4.

A recent structure-based study indicates an intimate relationship between the effects of PtdIns(4,5)P2 and Nai+ on certain Kir3.x channels (665). Molecular dynamic analysis suggests that, without Nai+, the side chains of D223 and R225 in Kir3.4 are connected via an ionic bond, and this connection may prevent the Arg residue from controlling PtdIns(4,5)P2 sensitivity of the channel. When Nai+ is present, it is expected to be coordinated by side chains of D223 and H228 and free R225 to increase the channel's sensitivity to PtdIns(4,5)P2. Similar results were obtained for Kir3.2 (665).

4. Oxidation-reduction and acidification

Application of the reducing agent dithiothreitol (DTT) causes openings of KG channels (Kir3.1/3.4) without affecting their permeation or rectification properties (883). DTT seems to directly act on the channels because its effect was abolished when Cys in the NH2-terminal cytoplasmic region was mutated. Control of KG channel activity by redox signaling may protect cells and tissues under hypoxic or ischemic insult.

KG channels are also inhibited by intracellular acidification. The pHi sensitivity relies on a few His residues in their NH2 and COOH-terminal cytoplasmic domains (H57, H222, and H346 in Kir3.1; H64, H228, and H352 in Kir3.4) and does not seem to depend on G proteins (496). This modulation may be involved in the control of respiratory activity and CO2 chemoreception in the brain stem (36, 216, 553).

5. Ion permeation, gating, and inward rectification

KG channel currents show strong inward rectification. The mechanisms underlying rectification and its structural basis are similar to those of other Kir channels (see sects. i and ii) (143, 172, 336, 338, 404, 471, 507, 539, 574, 611, 802, 831, 856, 894). One difference is that Kir3.2, Kir3.3, and Kir3.4 have Asn instead of a negatively charged Asp as a polyamine binding site in the TM2 transmembrane helix, although Kir3.1 possesses Asp in this position (611, 733).

In terms of ion selectivity, KG channel subunits share the signature sequence (T-X-G-Y/F-G). The loss of K+ selectivity in the weaver mouse (see sect. iiiF) is caused by the mutation of G156 to Ser in the selectivity filter of Kir3.2 (606), which permits nonselective cation permeation (383, 554, 733). Residues outside the selectivity filter of Kir3.x channels have also been found to contribute to K+ selectivity. First, mutations at E139 and R149 of Kir3.1, E152 of Kir3.2, and E145 and R155 of Kir3.4 cause dramatic loss of K+ selectivity in KG channels (492). Molecular modeling suggests that these residues form a salt bridge behind the selectivity filter and probably act to maintain the structure rigid. Second, substitution of S177 in TM2 of Kir3.2 located at the base of the selectivity filter abolished K+ selectivity (872). Equivalent results were obtained at S166 in Kir3.1 and A172 in Kir3.4 (492).

The roles of the TM2 helices in the gating of Kir3.x channels were described in detail in section iD1. In summary, Gβγ-induced channel openings may depend on a number of processes. Gβγ binding to the cytoplasmic domain induces a configuration change that is transduced to the lower TM2 region and rotates it (340, 404). This might also result in bending of the TM2 helix at the Gly hinge (97, 872). This hypothesis is very convincing, but it needs to be verified in future studies.

C. Channel Localization

1. Membrane trafficking machinery

When Kir3.1 is expressed alone in heterologous expression systems, it is not sorted to the cell membrane surface; instead, it remains in ER compartments (367, 368, 489). On the other hand, when coexpressed with Kir3.2 or Kir3.4, Kir3.1 is efficiently transported to the membrane surface. Homomeric channels composed of Kir3.2 or Kir3.4 do appear on the cell surface (489). Immunolabeling analysis of various mutated Kir subunits suggests that wild-type Kir3.2 and Kir3.4 are expressed on the cell surface due to two forward trafficking signals in their cytoplasmic regions. One is an ER export signal in their NH2 terminus (N-Q-D-M-E-I-G-V in Kir3.2; D-Q-D-V-E-S-P-V in Kir3.4), and the other is a post-ER trafficking signal that is composed of a cluster of acidic residues in the COOH terminus (489). Both of these signals seem to be necessary to traffic the subunits to the plasma membrane.

Of note, some population of Kir3.2, when expressed alone in COS7 cells, is prominently localized in intracellular vesicles (489). In cultured hippocampal neurons, a majority of endogenous Kir3.2 exists in cytoplasmic region, although it is modestly expressed on the cell surface (90). This subunit has a sequence of T-E-S-M-T-N-V-L (amino acid 7–14) in NH2 terminus, which is similar di-Leu endocytosis signal [(D/E)-X-X-X-L-L]. Mutation of V-L to A-A results in disappearance of vesicular localization of Kir3.2 and increase of its surface expression, implying that these amino acids act as an internalization motif (489). Recent studies have shown physiological regulation of intracellular localization of Kir3.2-containing channels in neurons (90) (see sect. iiiD3).

Coexpression with Kir3.3 results in a large decrease in the KG currents normally yielded by homomeric Kir3.2, heteromeric Kir3.1/3.2, and Kir3.1/3.4, which is attributed to retention of the channels in the ER (382, 444, 489, 694). This phenotype depends on the following two mechanisms: 1) Kir3.3 lacks an ER export signal, and 2) Kir3.3 has a lysosomal targeting signal, Y-W-S-I (Y-X-X-Φ), in its COOH terminus. These signals target channels that contain Kir3.3 from the ER to lysosomes without delivering them to the plasma membrane. Even when the membrane-targeting region of Kir2.1 is spliced onto Kir3.3, the hybrid Kir3.3 remains mainly in the cytoplasm (489). In this case, the hybrid channel protein is detected mainly in the ER, probably because the proteins that moved to lysosome may be rapidly degraded. Together, the lysosomal targeting signal seems to have a strong effect on localization of the channels containing Kir3.3.

2. Localization in native tissues

Kir3.1, Kir3.2, and Kir3.3 proteins are expressed throughout the brain (225, 307, 390, 453). But the expression of Kir3.4 is found in only limited areas (297, 834). Studies using light and electron microscopy have revealed the precise localization of Kir3.x subunits in neurons. Kir3.1 and Kir3.2 are localized at both postsynaptic (see Fig. 10C) and presynaptic regions (see Fig. 10D), whereas Kir3.3 has been detected in certain axons (225, 307, 390, 401, 453, 544, 623). Kir3.4 immunoreactivity was observed at the axonal terminus (297, 834). Thus KG channels may be involved not only in slow inhibitory postsynaptic potentials but also in the presynaptic modulation of neuronal activity.

Fig. 10.

Intracellular localization of KG channels and their functional roles. A: Kir3.1 in the rat anterior pituitary lobe. Immunogold electron microscopy shows that Kir3.1 occurs on the intracellular secretory vesicles of the thyrotroph. Kir3.1 signals (15-nm particles) and TSH signals (10-nm particles) were detected on the same vesicles (inset) (a). Scale bar, 0.5 μM (inset, 0.15 μm). After TRH administration (b), a population of Kir3.1-positive vesicles aligns near the plasma membrane, and immunogold particles were detected on the membrane surface (arrows). B, a: simultaneous recordings of membrane capacitance (Cm), conductance (Gm), and membrane current (Im) during stimulation of a thyrotroph cell. The application of 10 nM bromocriptine had no effect on Cm. The addition of 3 μM TRH (arrow) transiently increased Cm. Im was little influenced by bromocriptine. The addition of TRH induced a marked increase of inward Im at the holding potential −100 mV. Arrowhead indicates the zero-current level. Solid circles show calibration signals for Cm [250 femtofarads (fF)]. The scale bar for Cm is 250 fF, which corresponds to 2.35 nS, and the scale bar for Im is 250 pA. Time scale bar = 1 min. b: A schematic illustration of TRH-induced recruitment of KG channels which contain Kir3.1 to the plasma membrane via exocytotic fusion. DA, dopamine; SS, somatostatin. [A and B from Morishige et al. (542), with permission from The American Society for Biochemistry and Molecular Biology.] C: postsynaptic localization of Kir3.2 in the brain. Kir3.2 immunoreactivity was localized on the postsynaptic membranes in the dendrites contacted by several axonal termini in substantia nigra pars reticulata. [From Inanobe et al. (307), with permission from The Journal of Neuroscience.] D: presynaptic localization of Kir3.1 in brain. Kir3.1 immunoreactivity in axon terminals within the paraventricular nucleus of the hypothalamus (a and b). The labeling appears to be condensed in the intracellular region. D, dendrite. [From Morishige et al. (544), with permission from Elsevier.]

Neuronal KG channels are either homomeric or heteromeric complexes of Kir3.x subunits, but the relationship between the specific subunit combination and the subcellular localization is not entirely clear (307, 390, 444, 453). A Kir3.2 homomer composed of Kir3.2a and Kir3.2c was found to occur at the inhibitory postsynaptic region in dopaminergic neurons of the substantia nigra, and this localization may be determined by protein-protein interactions (307). PDZ-proteins are considered to be one of the major determinants of postsynaptic localization of various proteins (711). There is a PDZ-domain binding motif at the distal COOH-terminal region of Kir3.2c but not Kir3.2a. The Kir3.2 complex in substantia nigra would be sorted to the postsynaptic site via interaction between Kir3.2c and PDZ. This type of interaction may further modulate the function of KG channels containing Kir3.2c. Under certain conditions, a KG channel current could be recorded from Xenopus oocytes expressing Kir3.2c and a GPCR only when a PDZ-protein SAP97 was coexpressed (258).

A proteomics approach has recently identified a new regulator for trafficking and function of KG channels. An intracellular PDZ-protein called sorting nexin 27 (SNX27) is reported to directly interact with Kir3.3 (484) (this Kir3.3 corresponds to the lately isolated Kir3.3; see sect. iiiA1). Biochemical assays indicate that the interaction occurs between the PDZ domain of SNX27 and a PDZ-binding motif, ESKV, at the COOH-terminal end of Kir3.3. In HEK293 cells, coexpression of SNX27 largely suppresses the currents due to Kir3.1/3.3 and Kir3.2c/3.3 heteromers. The effect is due to the recruitment of the Kir3.3 subunit from the cell surface to the early endosome. Although SNX27 binds to Kir3.2c via the same PDZ-binding motif, its coexpression affects neither surface expression of Kir3.2c nor its current amplitude. This implies that SNX27 specifically regulates the activity of KG channels containing Kir3.3.

Another mechanism regulating the trafficking of neuronal KG channels involves the neural cell adhesion molecule (NCAM). Cultured hippocampal neurons from mice lacking NCAM exhibit a larger KG current, which is probably composed of Kir3.1/3.2, than that in wild-type neurons. When NCAM140 or NCAM180 was coexpressed with Kir3.1/3.2 in Xenopus oocytes or CHO cells, the surface expression of the channels was decreased. Consistently, whereas in cultured hippocampal neurons of NCAM-null mice the cell surface expression of Kir3.1/3.2 is significantly augmented, a considerable amount of Kir3.1/3.2 in the neurons from wild-type mice is expressed in Golgi apparatus. The effect of NCAM was not observed for Kir3.1/3.4 channels. The regulatory process of Kir3.1/3.2 expression by NCAM is likely to involve the role of DRM (lipid-raft) microdomains. Biochemical assays show that both Kir3.1 subunits and NCAM occur in DRM of transfected CHO cells and the brain. Mutational and pharmacological disruption of the association of NCAM140 in DRMs abolished its effect on the surface delivery of Kir3.1/3.2 channels. Thus, in intracellular compartments such as Golgi apparatus, NCAM incorporated into DRMs would somehow affect Kir3.1/3.2 and prevent its trafficking to the plasma membrane (127).

A KG channel with an identical subunit composition sometimes shows a different localization in different tissues. For example, the KG channels of atrial myocytes and thyrotrophs in the anterior pituitary are both heteromers of Kir3.1/3.4. In atrial myocytes, they are found on the cell membrane (391). On the other hand, in resting thyrotrophs, the channel is found exclusively on intracellular secretory vesicles (Fig. 10A) (542). The mechanism underlying such distinct localization of Kir3.1/3.4 channels in different cell types has not been clarified yet.

D. Physiological Functions in Cells and Organs

1. Heart

In the heart, ACh released from the vagal nerve terminals decelerates the heart beat (see Fig. 1B) (464). This ACh-induced bradycardia was found to be due to an increase in K+ efflux through the membrane (295) and membrane potential hyperpolarization (73, 125). Then, Trautwein colleagues (579, 601, 678, 786) identified KACh channel current in the rabbit sinoatrial node. They proposed that ACh induces activation of a specific population of K+ channels, cardiac KG channels which we now know are composed of Kir3.1 and Kir3.4 subunits (391).

Kir3.4 knockout mice were generated and analyzed to elucidate the role of the channels in the heart (835). As expected, in knockout atrial myocytes, no KG current (IKACh) was detected. Although there was no difference of the resting heart rate between wild-type and knockout mice, the decrease of heart rate in response to parasympathetic stimulation was considerably diminished in knockout mice. Strikingly, the knockout mice lost heart rate variability, i.e., the beat-to-beat fluctuations determined by the balance between sympathetic and parasympathetic influences.

2. Brain

In the brain there are heteromeric complexes of at least Kir3.1/3.2 (307, 453), Kir3.2/3.3 (110, 307, 333), and Kir3.2/3.4 (444) (as for the possibility of Kir3.1/3.3 complex, see below). Also, homomeric expression of Kir3.2 isoforms, i.e., complexes of Kir3.2a and Kir3.2c, were identified in dopaminergic neurons of the substantia nigra (307). A number of GPCRs have been shown to activate KG channels at postsynaptic as well as at presynaptic sites (407, 535, 778, 779, 873). Postsynaptic KG channels produce a slow inhibitory postsynaptic potential (sIPSP) and suppress their excitability. Several studies using the knockdown strategy suggest that Kir3.2 was critically involved in the formation of IPSP in hippocampal and cerebellar neurons (485, 726). The Kir3.2 knockout mice develop spontaneous seizures and are more susceptible to pharmacological maneuvers to induce seizures (726), which is quite different from the case of Kir3.3-gene ablation causing no obvious phenotype (783). Recent analyses of knockout mice have further defined physiological roles of neuronal KG channels in different regions of brain. In locus ceruleus, administration of opioid agonists such as [Met5]-enkephalin (ME) is known to hyperpolarize the neurons (613, 839, 840). KG conductance and/or cAMP-dependent cation conductance has been suggested to cause this effect. Three Kir3.x subunits, i.e., Kir3.1, Kir3.2, and Kir3.3, are expressed in these neurons (82, 360). Compared with the case of wild-type mice, although ME-induced hyperpolarization and whole cell current are not significantly altered in Kir3.3-null mice, they are reduced by 40% in Kir3.2-null mice and by 80% in Kir3.2/3.3 double-null mice (783). These results indicate that the ME-induced hyperpolarization is mainly mediated by KG channels containing at least Kir3.2 but not by cAMP-dependent cation conductance. Kir3.3 also seems to be involved in formation of KG current, although the detailed mechanism is unknown. The phenotype mentioned above would indicate a possibility that both Kir3.1/3.2 and Kir3.1/3.3 exist and function in neurons (the former could be more dominantly expressed than the latter) but, when either Kir3.2 or Kir3.3 is lacked, the remainder would compensate the function. The similar results are also observed in hippocampal CA1 neurons where GABAB-receptor agonists induce KG current to elicit postsynaptic inhibitory effects (390). Although Kir3.4 expression in the hippocampus is low (297, 834), Kir3.4-deficient mice show impaired performance in spatial learning and memory tasks (834).

A few studies have demonstrated an involvement of neuronal KG channels in the mechanism underling drug abuse and addiction. The release of dopamine in response to activation of mesocorticolimbic system, which originates in the ventral tegmental area (VTA), is considered to be critical for induction of compulsive addictive behavior (656). All addictive drugs react on mesocorticolimbic dopamine system (486, 559). Although a low-affinity agonist of GABAB receptor, γ-hydroxybutyric acid (GHB), can cause strong abuse (567), baclofen, which also activates the same receptor, shows anticraving effect (107). This difference seems to depend on a distinct combination of Kir3.x subunits among Kir3.1, Kir3.2c, and Kir3.3 in VTA where dopaminergic neurons are negatively regulated by GABAergic interneurons (110). In brain slices, both dopaminergic and GABAergic neurons postsynaptically express GABAB receptor and KG channels. However, EC50 for baclofen-induced activation of the channels is much higher in the dopaminergic neurons (∼15 μM) than in the GABAergic ones (∼1 μM). Indeed, a low dose of baclofen (0.1 μM) suppresses excitability of GABAergic interneurons by considerably activating KG channels but barely affects GABAB receptor in dopaminergic neurons. This results in a significant increase of firing frequency of action potential in dopaminergic neurons. On the other hand, the high dose (100 μM) of baclofen abolishes the action potential probably by direct activation of KG channels in dopamine neurons. As for expression profile of Kir3.x subunits, dopaminergic neurons have Kir3.2c and Kir3.3, whereas GABAergic neurons harbor Kir3.1, Kir3.2c, and Kir3.3. In heterologous expression system, EC50 for baclofen-induced activation of Kir3.2c/3.3 is significantly higher than that of each Kir3.1/2c, Kir3.1/3.3, and Kir3.2c channels (110; see also Ref. 333). A low-affinity GABAB agonist GHB at concentrations obtained with recreational use (10 mM) activates strongly KG channels in GABAergic neurons but only slightly those in dopaminergic neurons. Therefore, differential sensitivity of KG channels to GABAB agonists, which seems to be elicited by distinct combination of Kir3.x subunits, is at least partially responsible for the opposite pharmacological effect on drug abuse between high-dose baclofen and GHB.

Interestingly, a subsequent study has suggested that modulation of KG channels in dopaminergic neurons by RGS2 is involved in tolerance of GHB-induced addiction observed after a certain period of its chronic administration (421). Inhibition of RGS family proteins by injecting PtdIns(3,4,5)P3 and gene ablation of RGS2 prominently reduces the EC50 value of KG channels' activation by baclofen to ∼5 μM in dopaminergic neurons. This would be a condition that a low-affinity GABAB agonist GHB at concentration of recreational use can decrease excitability of dopaminergic neurons by directly activating Kir3.2c/3.3 and prevent addictive behavior. After animals are chronically exposed to GHB, their dopaminergic neurons decrease not only expression of RGS2 but also the EC50 of KG channels' activation, which coincides with induction of tolerance for the drug. Therefore, regulation of KG channels by RGS may provide a key process in occurrence of drug abuse and its tolerance.

3. Vesicular and intracellular KG channels

A heteromeric complex of Kir3.1 and Kir3.4 has been identified in the secretory vesicles of thyrotrophs of the rat pituitary gland (Fig. 10A) (542). When these thyrotrophs are stimulated with thyrotropin-releasing hormone, the secretory vesicles fuse with the plasma membrane and this enhances dopamine- and somatostatin-induced KG currents (Fig. 10B) (542). This enhancement of KG currents results in hyperpolarization of the thyrotroph Em and may reduce or suppress thyrotrophin secretion by stopping Ca2+ influx via Ca2+ channels. The amplitude of KG currents induced by the inhibitory neurotransmitters is therefore related to the magnitude of thyrotroph stimulation. This would provide an effective negative-feedback loop to control blood plasma hormone concentration (Fig. 10B). We suggest that a similar mechanism could function in the nervous system where KG channels are localized in the cytosolic regions of axonal termini (Fig. 10D) (390, 544, 623).

A recent study actually shows that trafficking of Kir3.2-containing channels may be dynamically controlled by neuronal activity. In cultured hippocampal neurons, the heteromer of Kir3.1 and Kir3.2 forms a functional KG channel, but the majority of these subunits exist in the cytoplasmic region. Activation of NMDA receptor by application of glutamate augments surface expression of Kir3.1 and Kir3.2 in soma, dendrites, and dendritic spines doubly within 15 min (90). This regulation is abolished by mutation of the V-L internalization motif in NH2 terminus of Kir3.2 (position 13 and 14) and accelerated by dephosphorylation of Ser near the motif (position 9) (90), implying that phosphorylation status of Ser modulates functionality of the V-L motif similar to the case of di-Leu-mediated internalization of other proteins (57). Ser is dephosphorylated by protein phosphatase-1, and this enzyme is activated when NMDA receptor is stimulated. Thus excitation of neurons by glutamate would simultaneously prepare an inhibitory machinery by increasing surface KG channels that will cause membrane hyperpolarization in response to neurotransmitters. Interestingly, activation of NMDA receptor increases the population of KG channels that can be activated by stimulation of adenosine A1 but not GABAB receptor (89).

The similar system may also be involved in excitatory synaptic plasticity. In hippocampal neurons, depontentiation of long-term potentiation of field excitatory postsynaptic potential requires activity of NMDA receptor, A1 receptor, and protein phosphatase-1 (285, 286), which are the elements necessary for surface trafficking of Kir3.2-containing channels and their opening (see above). Disruption of Kir3.2 by gene-targeted ablation or application of a KG channel blocker tertiapin inhibits the depontentiation in hippocampal slice (89). These two lines of experimental results suggest that NMDA receptor-dependent increase of Kir3.2 expression on cell surface could be involved in the depontentiation process, one form of synaptic plasticity.

4. Pancreas KG channels

Several neurotransmitters and hormones are known to inhibit secretion of insulin and glucagon from pancreatic islets. Catecholamines and somatostatin suppress insulin secretion from pancreatic β-cells, and somatostatin can also inhibit glucagon secretion from α-cells. These regulators act on the islet cells through multiple signaling mechanisms (706), one of which may involve the hyperpolarization of Em by an increase in K+ conductance (1, 328, 664, 706, 724, 878). In both α- and β-cells, this hyperpolarization seems attributable to activation of KG currents. Kir3.2c and Kir3.4 subunits were found to occur in islet cells (55, 171, 746, 878), and the expression of Kir3.1 and Kir3.3 has also been suggested (328). Homo- or heteromultimers of these subunits may form functional KG channels to regulate hormone secretion from islet cells.

E. Pharmacology

Tertiapin is a potent inhibitor of Kir3.1/3.4 and Kir1.1 channels with nanomolar affinity (342, 372) (see sect. iE). In Kir1.1, tertiapin binds in the outer vestibule of the channel pore. Though since this position is only partially conserved within tertiapin-sensitive Kir channels (343) (see sect. vA5), the toxin may associate with the channels via multiple contacts. Jin and Lu (343) created an oxidation-resistant toxin, tertiapinQ, which is more stable than tertiapin in solution. The modified toxin blocks IKACh, but not other currents in isolated cardiac myocytes (136, 372).

n-Alcohols such as methanol, ethanol, and 1-propanol can activate KG channels made up of different combinations of Kir3.1, Kir3.2, and Kir3.4 subunits, although Kir1.1 and Kir2.1 are insensitive (376, 446). KG channels containing Kir3.2 show the greatest enhancement by ethanol.

KG channels in the heart (KACh channels) are blocked by quinidine and quinine with an IC50 of ∼10 μM (362, 413, 551). Verapamil also inhibits ACh-activated cardiac KG channels with an IC50 of 1 μM, although its effect is partially mediated by block of the M2-muscarinic receptor/G protein system (321). KG channels are rather insensitive to other well-known K+ channel blockers such as TEA, 4-AP, apamine, charybdotoxin, disopyramide, and procainamide (309, 362, 422, 423, 551). The antimuscarinic effect of disopyramide and procainamide is mediated mainly by blocking at the M2-receptor but not at the KG channel itself (551).

Heterologous expression of Kir3.x subunits in Xenopus oocytes has shown that KG channels made up of various combinations of Kir3.x subunits can be modulated by many compounds, including antipsychotic drugs such as haloperidol, thioridazine, pimodine, and clozapine acting on Kir3.1/3.2 and Kir3.1/3.4 channels (377); antidepressants such as imipramine, desipramine, amitriptyline, nortriptyline, clomipramine, maprotiline, citalopram, and fluoxetine acting on Kir3.1/3.2, Kir3.2, and Kir3.1/3.4 channels (378, 380); volatile anesthetics such as halothane, F3 (1-chloro-1,2,2-trifluorocyclobutane), isoflurane, and enflurane (533, 824, 860) acting on Kir3.1/3.2, Kir3.2, and Kir3.1/3.4 channels; and the local anesthetic bupivacaine acting on Kir3.1/3.2, Kir3.2, Kir3.1/3.4, and Kir3.4 channels (893). These compounds block KG channel activity, but they clearly do not show any subunit combination specificity. It is also the case that the selective dopamine D1 receptor antagonist R-(+)-7-chloro-8-hydroxy-3-methyl-1-phenyl-2,3,4,5-tetrahydro-1H-3-benzazepine hydrochloride (419), the NMDA receptor antagonists MK-801 (383) and ifenprodil (379), the inhaled drug of abuse toluene (126), the Ca2+ channel blocker verapamil, and the classical cation channel blocker QX-314 (383) have been shown to inhibit various combinations of Kir3.x subunits. Although the binding sites for most of these compounds have not been examined, those for halothane and bupivacaine are estimated to be in the cytoplasmic regions of the channels (533, 860, 893).

F. Diseases

1. Atrial fibrillation

Atrial fibrillation (AF) is the most common cardiac arrhythmia in clinical practice. AF can become persistent due to remodeling of atrial electrophysiology. The changes in the electrical properties of atrial cardiomyocytes, such as a decrease of L-type Ca2+ current density (800) and an increase of IK1 current density (137), are thought to contribute to the pathology of persistent AF. Recently, IKACh has been found to play an important role in chronic AF (cAF) and thus could form a possible target for AF therapy. Dobrev et al. (136) have reported that electrical remodeling in AF patients causes an increase of a constitutively active component of IKACh but a decrease of its ACh-induced component. This switch from a ligand-gated current to constitutively active behavior would lead KG/Kir3.x channels to shorten atrial action potential duration and refractory period in cAF patients. Although the molecular basis of this abnormality in cAF patients remains elusive, IKACh composed of Kir3.1/3.4 could be a therapeutic target for AF prevention and treatment. Tertiapin has been shown to suppress AF in canine models (77, 244). In this context, it should be remembered that IKACh can be inhibited by classical antiarrhythmic agents such as quinidine (413) and AN-132 [3-(diisopropylaminoethylamino)-2′,6′-dimethylpropionanilide] (410).

2. Weaver mice

A mutation in weaver (wv) mice occurs in the signature amino acid sequence of the selectivity filter of Kir3.2 (606). This results in nonselective cation permeation rather than K+ selectivity (383, 554, 733). The distribution of Kir3.2 clearly overlaps tissues showing defects in wv mice. The mutation results in phenotype abnormalities such as severe locomotor defects, a deficiency in the migration of granule cells of the cerebellum (640), and cellular deficits in the midbrain dopaminergic system (691). Because of abnormal selectivity of KG channels that contain mutant Kir3.2, inhibitory transmission in wild-type turns into excitatory signaling, which results in neuronal death. The failure of sperm production in homozygous male wv mice (243) probably results from the mutation of Kir3.2d which is expressed in the acrosome of spermatids (304).

3. Neuronal disorders

Knockout of Kir3.2 in mice is accompanied by the reduction of Kir3.1 protein expression in the brain, the development of spontaneous seizures, and a range of phenotypes related to sporadic seizures (726). These mice also show an increased susceptibility to convulsant agents (726). These phenotypes would result from the loss of IPSP and the resultant neuronal hyperexcitability (see also sect. iiiD).

IV. KATP Channels (Kir6.x/SURx)

A. Historical View and Molecular Diversity

1. Identification of KATP channels in native tissues

KATP channels were first identified in cardiac myocytes (577, 788). KATP channels are also found in pancreatic β-cells (22, 102), skeletal muscle (739, 740), vascular smooth muscle (40, 349, 353, 741), and neurons (24, 25). In the inside-out membrane patch, KATP channels open spontaneously. Thus the KATP channels should be classified together with the classical Kir2.x channels that have constitutive activities. The openings are inhibited by ATPi (except for in smooth muscle) and activated by NDPsi such as ADP (21, 773). These are the hallmarks of KATP channels and therefore from the onset the channels were proposed to be associated with cellular metabolism and membrane electrophysiology.

The heteromeric complex nature of KATP channels was first suggested by Tung and Kurachi (795) showing that the inhibitory effect by ATP and the stimulatory effect by NDP are mediated by reaction of these nucleotides on distinct sites of the cardiac KATP channel. They proposed that a KATP channel comprises a gate, an ATP-binding unit, and a transducer unit which transduces signals from the ATP-binding unit to the gate (773, 795). KATP channels are now known to be functional octomers composed of four Kir channel subunits forming the ion channel pore and four auxiliary proteins: the sulfonylurea receptors, SURx (Fig. 11A). The Kir subunits are responsible for ATP inhibition and the SUR proteins for NDP activation. A variety of pharmacological agents can either stimulate or inhibit KATP channels by binding to SUR (Fig. 12). The inhibitory agents include sulfonylureas, such as chlorpropamide, tolbutamide, and glibenclamide (144, 154, 832), which are used in the therapeutic treatment of type II diabetes (122, 144, 689, 690, 832) (see sect. ivE). The stimulatory agents are known as K+ channel openers (KCOs), and they include pinacidil, nicorandil, and diazoxide (21, 773) (see sect. ivE).

Fig. 11.

Molecular constituents of KATP channels. A: KATP channels are composed of pore-forming Kir6.x subunits and SUR subunits. SUR subunits contain 17 transmembrane regions grouped into 3 transmembrane domains TMD0, -1, and -2 (top panel). Each SUR unit has two nucleotide-binding domains (NBD, depicted as hemispheres) between TMD1 and TMD2 (NBD1) and in the COOH terminus following TMD2 (NBD2). SUR2A and SUR2B differ only in the COOH-terminal 42 amino acids (C42). A functional KATP channel is a heterooctamer made up of four Kir6.x subunits and four SUR subunits (bottom left). Combinations of different SURx and Kir6.x subunits form distinct types of KATP channels. Bottom right images show top and side views of the entire KATP channel complex analyzed at 18 Å resolution. Blue represents Kir6.x. Yellow represents TMD0 of SUR. Red represents the rest of SUR. [From Mikailov et al. (528), with permission from Nature Publishing Group.] B: homology between SUR subunits. TM, transmembrane region; TMD, transmembrane domain; NBD, nucleotide binding domain. [From Isomoto et al. (315).]

Fig. 12.

Determinants of the characteristics of KATP channels. The single-channel characteristics (unitary conductance and rectification) of KATP channels are determined by the pore-forming Kir6.x subunits. Drugs that modify channel activity, i.e., sulfonylureas and K+ channel openers, act on SURs. Nucleotides can influence both Kir6.x and SUR subunits to regulate channel behavior.

Physiological and pharmacological analysis has characterized distinct types of KATP channels in different tissues (Fig. 11A). Native KATP channels display quite different unitary conductance, for instance, ∼70–90 pS in cardiac muscle, ∼55–75 pS in skeletal muscle, and ∼50–90 pS in pancreatic β-cells (21, 577, 635, 858). In vascular smooth muscle, the initial report of a large-conductance KATP channel (741) has not been reproduced (see also Ref. 40). Instead, small-conductance (∼15 and 25 pS) K+ channels are activated by KCOs (349, 353), and the presence of an NDP is indispensable for KCO-mediated activation of the channel. These are called KNDP channels, and their opening is insensitive to ATPi. The KNDP channels are the major representative of the KATP channel family in vascular smooth muscle cells. KATP channels in different tissues also exhibit considerable variation in their pharmacological properties. The affinity for a radiolabeled analog of glibenclamide classifies SURs into at least two types (202, 690). One type shows a high affinity for sufonylureas and resides in pancreatic β-cells. Another type with a lower affinity is present in the heart, skeletal muscle, and brain. These early findings implied a molecular heterogeneity of KATP channels. Subsequent cDNA cloning identified the molecular constituents of the KATP channels and their diversity (3, 4, 299301, 317, 853).

2. Cloning of the constituents of KATP channels

In 1995, cDNA of uKATP-1 was isolated from rat pancreatic islets (301). This channel possessed the primary structure common to Kir subunits, but its heterogeneous expression did not yield any functional K+ currents.

An SUR was first cloned from cDNA libraries of rat and hamster insulinoma cell lines (4). The deduced amino acid sequence defined the receptor as a member of the ATP-binding cassette (ABC) transporter superfamily with two nucleotide-binding domains (NBD) (Fig. 11B). The receptor possesses 17 TM regions that can be grouped into transmembrane domain 0 (TMD0: TM1-TM5), TMD1 (TM6-TM11), and TMD2 (TM12-TM17) (Fig. 11A). This SUR (SUR1), like other ABC proteins, could bind and hydrolyze ATP (47, 511). Unlike a Cl channel ABCC7 that is also called cystic fibrosis transmembrane conductance regulator (CFTR), however, SUR1 did not show catalysis-coupled ion transport activity (6), and the injection of SUR cRNA alone into Xenopus oocytes generated neither ATP-sensitive nor glibenclamide-sensitive K+ currents (4).

Inagaki et al. (301) cloned another Kir channel subunit (BIR) from a human genome library using uKATP-1 cDNA as the probe. The amino acid sequences of uKATP-1 and BIR were highly homologous with ∼70% shared amino acid identity. These two subunits also shared ∼40–50% identity with other Kir channels. It was therefore suggested that the two clones belonged to the same Kir subfamily, Kir6.x (see Fig. 2B). Today uKATP-1 and BIR are known as Kir6.1/KCNJ8 and Kir6.2/KCNJ11, respectively.

Kir6.2 mRNA is expressed abundantly in pancreatic islets and various pancreatic cell lines and moderately in the heart, skeletal muscle, and brain (301). Kir6.2 and SUR1 mRNAs were found to occur together in a variety of tissues and cell lines, and coexpression of Kir6.2 and SUR1 in COS-1 cells reproduced the main physiological and pharmacological properties of the KATP channels observed in pancreatic β-cells (300).

Several tissues such as heart, skeletal muscle, brain, and smooth muscle that possess functional KATP channels do not express the SUR1 subunit. A homolog of SUR1, SUR2A, was isolated (299) (Fig. 11B). When SUR2A and Kir6.2 were coexpressed, the functional KATP currents were less sensitive to either ATPi or glibenclamide than the currents yielded by the channels composed of SUR1 and Kir6.2 (299). Kir6.2/SUR2A channels were activated by the “cardiac” KCOs, cromakalim and pinacidil, but not by diazoxide, which suggested that Kir6.2/SUR2A may form cardiac and skeletal muscle type channels (299). A variant of SUR2A, SUR2B, was cloned from a mouse heart cDNA library (Fig. 11B) (317). Only the last 42 amino acid residues in the COOH terminus (C42) differ between SUR2B and SUR2A, indicating that they are formed by alternative splicing of a single gene (Fig. 11B) though C42 of SUR2B shares considerable homology with C42 of SUR1. Coexpression of SUR2B and Kir6.2 in HEK 293T cells elicits KATP currents that are activated by pinacidil and diazoxide (317). On the other hand, diazoxide cannot stimulate channels composed of Kir6.2/SUR2A (114). Therefore, under physiological conditions, C42 of SUR2B may be involved in diazoxide activation of KATP channels in smooth muscle cells (317, 853).

The conclusion of the cloning and reconstitution studies is that the KATP channel is a complex of four pore-forming Kir6.x subunits and four auxiliary SURx proteins (98, 721) (Fig. 11A). SUR1, SUR2A, and SUR2B represent pancreatic, cardiac, and vascular smooth muscle types of SUR, respectively, and different combinations of Kir6.1 and Kir6.2 and the SURx determine the properties of each native channel (3). Although it has proven to be possible to construct heteromers of Kir6.x and different SUR subunits which show variant characters (79), the presence of Kir6.1/6.2 heteromeric channels in native tissues is disputed (625, 703).

B. Pore Structure and Function

1. Intrinsic pore function


The single-channel conductance of Kir6.2 is ∼80 pS, whereas that of Kir6.1 is ∼30 pS. Studies using chimera and point mutations of the channel revealed that extracellular links between the two transmembrane segments and the H5 pore region were the critical elements (386, 651). Homology modeling based on the KcsA structure (143) suggests that M148 in Kir6.1 and V138 in Kir6.2 occupy the same position in the extracellular linking region between the H5 pore helix and the second transmembrane segment (TM2) and face the external entrance of the pore (651). The side chain of M148 is larger than that of V138, which may lead to differences in K+ flux through the pore mouth. Also, in the linking region between the H5 pore helix and the first transmembrane helix (TM1), at least three sequential residues (S113-I114-H115 of Kir6.2; N123-V124-R125 of Kir6.1) were found to be involved in the difference of conductance between the two channel types. While homology modeling of Kir6.2 suggests that such triplet residues may be localized far from the channel outer entrance, it also suggests a hydrogen bond interaction between S113 with R136, a residue that is highly conserved among Kir channels and which is considered to play a key role in maintenance of the stability of pore selectivity filter (867). Therefore, via this interaction, the three residues could control the conformation of the permeation pathway and determine the different conductance of Kir6.1 and Kir6.2 pores (651).


KATP channels exhibit weak inward rectification. The Kir6.2 subunit contains an Asn in the TM2 inner helix (N160) at the D/N site (see sects. iC1, iiB, and vA2). Mutation of N160 in Kir6.2 to Asp or Glu resulted in generation of KATP channels showing strong rectification (721).

2. Regulation of KATP channel function

KATP channel activity is regulated by intracellular nucleotides and by various pharmacological agents. The physiological and pharmacological control of KATP channel activity has to be interpreted in terms of conformational changes in the Kir6.x pore through protein-protein interaction with SUR as well as direct action of agents such as ATP on the pore subunit itself (Figs. 12 and 13).

Fig. 13.

Modeling opening and closing of a SUR2x/Kir6.2 channel. A: an allosteric model of SUR2x/Kir6.2 channels. Here, we assume the following: 1) a SUR2/Kir6.2 channel has a hetero-octameric structure composed of four SUR2x and four Kir6.2 subunits; 2) each SUR2x has a receptor (indent) for a ligand (solid circle); 3) SUR2x has two distinct conformations that are able (R conformation) and unable (T conformation) to open the channel pore formed from Kir6.2 subunits; 4) the T and R conformations are in equilibrium determined by the allosteric constant L, where L is the ratio of R to T in the absence of any ligands; 5) the structure of the receptor with regard to the ligand is distinct in the R and T conformations and has higher affinity for the ligand in the R rather than the T conformation; and 6) as a result, the ligand shifts the T-R equilibrium toward the R conformation. The ligand can be regarded as either nicorandil, ATP, or ADP. In the case of nicorandil, the indent represents the drug receptor site. For ATP and ADP, the indent corresponds to either NBD1 and/or NBD2. B: hypothetical conformational change of SUR2x induced by dimerization of NBDs. The illustration demonstrates two opposing Kir6.2 subunits and one SUR2x interacting with one of the Kir6.2 subunits. SUR interacts with Kir6.2 through TMD0 and L0, a cytoplasmic linker between TMD0 and TMD1 (29, 80). In this model, NBD1 and NBD2 face each other. ATPi induces dimerization of these NBDs and thereby a conformational change of TMD1 and TMD2. This conformational change is transferred to Kir6.2 through TMD0-L0 (green arrows), resulting in opening of the channel pore formed by Kir6.2. The conformation of SUR2x with and without NBD dimer could be considered equivalent to the R and T conditions. [A and B from Yamada et al. (852).]

I) Differential action of intracellular nucleotides on KATP channel activity.

Physiologically, regulation of KATP channel activity by intracellular nucleotides is the most significant profile. ATPi is the main regulator and can exert two distinct functions: 1) it closes the channel, and 2) it maintains channel activity in the presence of Mg2+.

The KATP channel stays closed as long as ATP is bound to the channel. In native pancreatic β-cells, the presence of Mg2+ shifted the IC50 value of KATP channels for ATPi from 4 to 26 μM, suggesting that ATP4− in its free acid form is a more potent inhibitor of the channel and that Mg-ATP has little inhibitory effect (19). In smooth muscle cells of portal vein, Mg-ATP is also less effective than ATP4− (349, 557). Cardiac-type KATP channels are inhibited by both ATP4− and Mg-ATP in a similar manner (175, 351, 432).

Cardiac and β-cell KATP channels spontaneously open when their cytoplasmic side is exposed to ATP-free solution. Then, the channel activity declines with a variable time course. This “rundown” can be accelerated by divalent cations such as Mg2+. Treatment of rundown membrane patches with Mg-ATP but not with nonhydrolyzable ATP restores spontaneous channel activity (54, 178, 179, 536, 586, 761, 788). This is a phenomenon that is common to all Kir channels (287); Mg-ATP maintains KATP channel activity through PtdIns(4,5)P2 generation from phosphatidylinositol by ATP-dependent lipid kinase (see sect. i, C1 and D2).

NDPs such as ADP are also essential for the physiological opening of KATP channels, which would otherwise be permanently closed by their overt sensitivity to ATPi. There was evidence that NDPs such as ADP increased KATP channel activity against the ATPi-induced inhibition of channel opening (149, 176, 178, 350, 432, 536). Therefore, it was initially proposed that NDPs would activate KATP channels by competing with ATP-binding to the channels. However, NDPs can also open KATP channels in the absence of ATPi (149, 176, 178, 350, 432, 536). The positive effects of NDPs require the presence of Mg2+, in the absence of Mg2+ they enhance the inhibitory effect of ATPi (149, 176, 178, 350, 432, 536). Of importance, when rundown channels were activated by NDPsi, they could be subsequently inhibited by ATPi with a Ki value similar to that in the absence of the NDPsi (795). Based on these observations, it was proposed that NDPs might not compete with ATP at its binding site but may affect other site(s) to control the function of the channels (795). KATP channels are therefore likely to bear two distinct sites for channel gating, an inhibitory ATP-binding site and a stimulatory NDP-binding site, which could be regulated by NDPs in two different ways, i.e., antagonism and agonism, respectively (775, 795).

II) Modulation of channel kinetics by nucleotides.

In the absence of ATPi, cardiac and pancreatic KATP channels exhibit spontaneous bursts of rapid openings and closings (fast kinetics), which are separated by long closed intervals (slow kinetics) (8) (see also sect. ivA).

The major effect of ATP upon gating is to increase “long” closed lifetimes (7, 8, 145, 156158). ATP may also decrease the duration of the bursts of rapid openings which suggests that ATP destabilizes the channel's open state as well as stabilizes its closed state (156, 162, 448, 566, 785). On the other hand, NDPs eliminate the long closed state between bursts and promote channel opening in sustained bursts without changing channel short open time (7). Therefore, ATPi and NDPsi principally modulate slow-gating but not fast-gating processes. NDPs also counteract the effects of ATPi. A kinetic model has suggested that NDP allows KATP channels to operate in an ATP insensitive state (7), which could explain how NDP allows KATP channel opening in the presence of otherwise inhibitory concentrations of ATPi. KNDP channels are also regulated by ATPi and ADPi, but the mechanism of this regulation remains elusive.

III) Structural basis of channel closure by ATP.

The site for ATPi inhibition of KATP opening locates on the Kir6.2 subunit (791). A number of basic residues and uncharged residues, R50 in the NH2 terminus and I182, K185, R201, and G334 in the COOH terminus, are involved in ATP-mediated inhibition of channel opening (15, 111, 344, 450, 632, 649, 790792). Homology modeling suggests that these five residues form an ATP binding pocket located at the interface of the NH2 and COOH termini (15, 158, 238, 564, 784). The stoichiometry of ATP binding in one Kir6.2 subunit provides one high-affinity ATP-binding site, and therefore, one functional KATP channel bears four sites (501).

This structural information indicates that four ATP molecules could bind to one KATP channel. ATP inhibited the native KATP channel of cardiac myocytes or skeletal muscle, and SUR2A/Kir6.2 as well as SUR2B/Kir6.2 channels, with a Hill coefficient significantly larger than unity (∼1.8–3) (21, 177, 299, 317, 589, 825, 826, 858). On the other hand, in pancreatic β-cell KATP channels, SUR1/Kir6.2, and Kir6.2ΔC, the truncated form of Kir6.2 that can function in absence of SUR, the Hill coefficient was close to 1 (15, 25, 218, 300, 501, 565, 680, 785, 791). These findings suggest that the type of SUR associated with different Kir6.x may influence the effectiveness of ATP binding. In vitro reconstruction of KATP channels with different SUR subunits coexpressed with Kir6.2 showed channels with different opening kinetics in the absence of ATPi (32, 91) which may relate to their different sensitivity to ATPi. But a recent study by Craig et al. has found that, in the case of Kir6.2/SUR1 channel, the binding of one ATP to a single Kir6.2 subunit, by interaction between several of the associated cytoplasmic domains, should suffice to induce channel closure (108). The reason that a Hill coefficient of ATP-induced inhibition of native KATP channels in cardiac myocytes and skeletal muscle is larger than unity remains unknown.

Mutations in Kir6.2 that change intrinsic channel gating may have a secondary effect to reduce ATP sensitivity (145, 157, 158, 719, 785, 792), probably because ATP preferentially binds to the closed channel (8, 157, 720).

PtdIns(4,5)P2 is responsible for sustaining spontaneous KATP channel activity in excised membrane patches (see sect. iC1). PtdIns(4,5)P2 binds to positive charges in the cytoplasmic region of the Kir6.2 subunit (R54, R176, R177, and R206) close to the ATP binding site (39, 722, 723). If PtdIns(4,5)P2 sustains KATP channel activity by stabilizing their open state, it may not be surprising that it can also decrease channel ATP sensitivity. Assays suggest competitive binding between PtdIns(4,5)P2 and ATP to the channel (39, 158, 239, 564, 723).

IV) Mechanisms of activation of KATP channels by NDPs.

The site for NDP activation of KATP channel opening is not located on the Kir6.2 subunit (791). An SUR subunit has two NBDs with Walker-A and Walker-B consensus motifs (G-X-X-X-X-G-K-T/S) (815). NBD1 is located on the cytoplasmic loop between the TM11 and TM12, and NBD2 is situated on the cytoplasmic COOH terminus following the TM17 (Fig. 11A). Mutations on the Walker-A motif of NBD1 prevent nucleotides from binding to both NBDs (796) and interfere with the stimulatory effect of NDPs (28, 114, 221, 512, 720). Thus NDPs activate KATP channels via their binding to the SUR, while ATP inhibits channel activity by binding directly to Kir6.2 (Fig. 12), and there is cooperation between the two NBDs to function as nucleotide receptors (511, 797, 898).

NBDs of other ABC proteins form a dimer when exposed to nucleotides (81, 278, 735). Homology models of the three-dimensional structure of NBDs of SUR2A and SUR2B (855, 857) were made with using the monomeric structure of HisP and the ATP-mediated dimer structure of MJ0796 (292, 735). When mutations equivalent to those that prevent dimerization of MJ0796 were introduced into NBDs of SURs, the NDP-mediated activation of KATP channels was impaired (852). When mutations equivalent to those that permit MJ0796 to be dimerized by Na+-bound ATP were introduced into NBDs of SUR2, Nai+ increased the activity of KATP channels in the presence of ATPi (852). Homology models of the dimerized structure of the NBDs in SUR1 were generated and mutations introduced into the putative interface of the dimerized NBDs altered their affinity for Mg-NDP (75, 503). These results indicate that the two NBDs form a dimer in SUR proteins to contribute to gating of the ion channels. A similar mechanism was reported to control CFTR Cl channel activity where NBDs are dimerized when bound to ATP (806).

In all ABC proteins studied so far, NBD1 and NBD2 sandwich Mg-ATP in the dimer (276). The dimer is needed to generate two catalytic sites for ATP hydrolysis. These sites are formed by the Walker-A motif (or P-loop) of one NBD and the ABC signature sequence motif (L-S-G-G-Q motif or C-loop) of the other NBD. Of the two catalytic ATP-binding sites, that formed by the Walker-A motif of NBD2 and the C-loop of NBD1 shows the most prominent ATPase activity (47, 510). The ATP hydrolysis activity of NBD2 of SUR1 is also higher than that of SUR2A (502) which may contribute to the differences between Kir6.2/SUR1 and Kir6.2/SUR2 gating properties. An analysis of the effects of NDPs on KATP channels which contained either SUR2A or SUR2B with the Monod-Wyman-Changeux (MWC) model (538, 852) (Fig. 13A) suggested KATP channel activation via conformational changes in the SUR subunits transiting between relaxed (R, able to open) and tense (T, unable to open) conformations (Fig. 13) (852, 855, 857). The T-R transition in nucleotide-bound forms of the NBDs was higher in SUR2B than in SUR2A (512, 513, 855). The 42 residues in the COOH terminus of SUR2B (C42) are located at the interface between the NBD dimer (Fig. 13B) where the terminus could interact with the NBD dimer (855, 857). The mechanism of action of NDP added to the internal surface of a membrane-bearing KATP channels may be to lock the NBDs in a “posthydrolytic” configuration that is associated with the open channel (897).

V) Mechanisms underlying SUR modulation of Kir pore function.

It still remains unclear how SUR can modulate the pore function of KATP channels. The unique TMD0 domain of SUR was reported to anchor SUR to the outer TM1 helix and NH2 terminus of Kir6.x (29, 164, 283, 386) (Figs. 11A and 13B). The adjacent L0 linker functionally links Kir6.2/TMD0 with the SUR protein's core (Fig. 13B) (30, 649). This arrangement seems to be involved in both opening and closure of KATP channels. Thus ligand-induced conformation changes in SUR may be transduced via L0 and TMD0 to the Kir6.x channel pore. Recently, the Asp/Glu-rich domain (ED domain), a stretch of 15 negatively charges residues adjacent to NBD1 in SUR2A, was also suggested to have a role transducing conformational rearrangements of the sulfonylurea receptor to the Kir6.2 channel pore (356).

3. Other means of regulation of KATP channel gating


KATP channels can be phosphorylated by Gs-mediated PKA signaling pathways, and importance of this modulation has been highlighted particularly in smooth muscles (42, 458, 557, 635, 637, 713). PKA phosphorylation sites include S385 in Kir6.1 and T633 and S1465 in SUR2B (637), and S1387 in NBD2 of SUR2B (713). Kir6.2 can be phosphorylated at T224 by PKA (458) and at T180 by PKC (455). It is still not clear how phosphorylation of these residues activates the channels.


Gating of the atrial KATP channel is mechanosensitive, and mechanical pressure applied to a cardiac cell leads to an increase in their activity (798, 799). In general, the actin cytoskeleton plays an important role in the detection and/or transduction of mechanical stress to ion channels (203, 810). The disruption of actin filaments with DNase I or cytochalasin B antagonized ATPi-mediated inhibition of KATP channels (774). The same treatment impaired sulfonylurea-induced closure of cardiac KATP channels (62, 874). Although the molecular mechanism remains elusive, this modulation may help to open the cardiac KATP channels during ischemia or hypoxia when mechanical disturbances of the cytoskeleton can occur. The mechanosensitivity of KATP channels is enhanced under ischemic conditions (799).

C. Intracellular Localization

1. Trafficking to the plasma membrane

Only the octameric KATP channel complex can reach the cell plasma membrane (4, 300, 680, 887). However, Kir6.2ΔC26 is transported and forms a functional K+ channel on the membrane without any SUR subunits (791). The truncated region contains an ER-retention sequence, R-K-R (887). SUR1 also contains the R-K-R sequence which prevents its surface expression, and mutation allows the receptor to traffic to the cell surface in the absence of Kir6.x channels (887). Thus the ER retention motifs in Kir6.x and SUR must be mutually masked to permit membrane surface expression (887). These sequences are different from known ER retention signals, i.e., the luminal K-D-E-L and cytoplasmic K-K-X-X motifs (772).

2. Mitochondrial KATP channels

In 1991, Inoue et al. (308) identified KATP channels in rat liver mitochondrial inner membrane by single-channel recording. These mitoKATP channels were inhibited by ATP applied to the luminal matrix face with a Ki of ∼0.8 mM. The channels were also inhibited by the K+ channel blocker 4-AP and by 5 μM glibenclamide. The mitoKATP may be made up of Kir and SUR subunits. Kir6.1 antibody labeled the mitochondria and bound to a 51-kDa protein in the mitochondrial membrane fraction, suggesting that the Kir6.1 subunit is a possible component of mitoKATP (752). Other studies have suggested that Kir6.1, Kir6.2, and SUR2A subunits are present in the mitochondria of the heart and brain (426, 427). However, because neither overexpression of a dominant negative form of either Kir6.1 or Kir6.2 in which the pore signature sequence G-F-G was mutated to A-F-A, nor targeted ablation of the Kir6.1 or the Kir6.2 gene could disrupt mitoKATP function (531, 703, 754), the molecular identity of the mitoKATP channel is still controversial.

D. Physiological Functions in Cells and Organs

1. Pancreas

In pancreatic insulin-secreting β-cells, KATP channels that are made up of Kir6.2 and SUR1 (299, 300) not only set the Eres (about −70 mV) but also serve to couple blood glucose concentration and insulin secretion (19, 20). Insulin secretion is stimulated by high blood glucose levels. At substimulatory blood glucose levels, pancreatic β-cell KATP channels are open and maintain negative Em. As blood glucose levels increase, glucose uptake and metabolism in β-cells are initiated. Then, ATPi concentration is augmented, whereas ADPi concentration is reduced. These changes attenuate KATP channel activity which results in cell depolarization and activation of L-type voltage-gated Ca2+ channels (VGCC). Ca2+ influx via VGCC leads to the fusion of vesicles containing insulin to the membrane and release of the hormone.

Consistent with this predicted function, genetic modification of Kir6.2 or SUR1 results in various phenotypes of glucose homeostasis disorder. Mice lacking either Kir6.2 or SUR1 showed transient neonatal hypoglycemia (530, 702). In humans, a number of polymorphisms in Kir6.2 and SUR1 have been identified, and many of them are associated with hypoglycemia and diabetes. These channelopathies will be discussed in section ivF (see also Table 2).

KATP channels are also expressed in other endocrine cells such as glucagon-secreting α-cells and somatostatin-secreting δ-cells (212, 213). They may be involved in the control of secretion of these hormones.

2. Heart

Early studies assumed that cardiac KATP channels played a cardioprotective role in ischemic conditions (319, 354, 566, 577, 773). KATP channels would shorten the cardiac action potential duration during ischemia (see Fig. 1B) and reduce harmful Ca2+ influx through VGCC. Physiological and pharmacological assays revealed the similarity between native cardiac KATP channels and the current elicited by coexpression of Kir6.2 and SUR2A in heterologous expression systems (27, 299). The myocardium of Kir6.2-knockout mice lacks functional KATP channels (451, 530, 753, 781). Thus, although this strongly suggests that the cardiac KATP channel contains Kir6.2, the SUR partner is less clear. It has been recently reported that Kir6.2 may be expressed with SUR1 in atrial myocytes and with SUR2A in ventricular myocytes of the mouse (184). This observation needs to be extended to other species.

In contrast to pancreatic β-cell KATP channels, cardiac KATP channels are closed under physiological conditions, presumably because of the high ATPi concentration in this tissue (354, 566, 773). Though how this equates with in vitro observations of quite extreme rightward displacement of the ATP channel-closure dose-response curve by NDPs and in particular PtdIns(4,5)P2 (39, 158, 239, 564, 723) remains to be examined. The channels would be opened by metabolic insult such as increased cardiac work load, hypoxia, or ischemia. Elevation of the S-T segment in the electrocardiogram, a hallmark of acute myocardial ischemia, was markedly reduced during ischemia in Kir6.2-knockout mice, implying that opening of KATP channels underlies the S-T elevation (451, 754).

Brief episodes of ischemia result in subsequent protection of the myocardium against later, more severe ischemic insult. This phenomenon is called ischemic preconditioning (547, 871). There is a general consensus that KATP channels play a key role in triggering this event, although whether this involves sarcolemmal KATP channels or mitoKATP is disputed. Gross et al. (224) first proposed that opening of KATP channels was involved in protective effects of preconditioning and KCOs mimic preconditioning, whereas KATP channel blockers eliminate the protective role during ischemia (113, 224, 698, 801). In Kir6.2-knockout mice, preconditioning disappeared (229, 754). A large number of mediators and signal transduction pathways seem to be engaged downstream of the opening of KATP channels (871).

To understand physiological and pathological roles of cardiac KATP channels, analysis of results obtained from Kir6.2-knockout mice are informative. So far, cardiac KATP channels have been implicated in the maintenance of cellular functions and stress adaptation in hyperadrenergic conditions such as physical exertion and decompensated heart failure, which are well-established precipitators of arrhythmia. Kir6.2-knockout mice showed an increased vulnerability to sympathetic stress and vigorous sympathetic challenge caused arrhythmia and sudden death (899). Similar phenotypes were observed in the mice whose cardiac KATP channels were conditionally ablated; they were less tolerant to exercise stress, and their mortality increased compared with control littermates after the age of 4–5 mo (781). In wild-type mice, β-adrenergic-mediated phosphorylation of KATP channels (42, 458) or depletion of ATP under the cell membrane (686) might enhance channel opening and contribute to shortening of cardiac action potentials. Catecholamine challenge shortens action potential duration in wild-type but not Kir6.2-knockout hearts where an inadequate repolarization causes early afterdepolarization and ventricular arrhythmia (463). Under equivalent hemodynamic stress, Kir6.2-knockout mice showed an aberrant prolongation of the action potential with pathological Cai2+ overload compared with wild type, leading to cardiac remodeling, heart failure, and death (355, 859). These findings indicate broad roles for KATP channels in heart disease and suggest possible clinical benefits from KATP channel modulators (see sect. ivF).

3. Smooth muscle

Kir6.1/SUR2B channels resemble native KATP channels observed in vascular smooth muscle in having a unitary conductance of ∼35 pS and no spontaneous opening in the absence of ATPi (315, 317, 853, 858). KATP channels are expressed in vascular smooth muscles throughout the body. Opening of vascular KATP channels hyperpolarizes Em towards EK. It results in closure of VGCC and hence relaxation of the smooth muscle of the blood vessels, especially that of veins (vasodilatation). In smooth muscle cells, the physiological function of KATP channels depends on their sensitivity to cellular metabolic state. Under normal conditions, regulation of smooth muscle contraction may be attributed not only to the direct effects of cellular metabolism itself (e.g., ATP/ADP) on KATP channel activity, but also to the effects on the channels of vasodilators such as prostaglandin, CGRP, and adenosine or vasoconstrictors including endothelin, vasopressin, 5-HT, and histamine secreted from surrounding cells (635). These vasodilators and vasoconstrictors may exert their action at least in part by phosphorylating KATP channels via PKA (42, 458, 635, 637, 713).

KCOs are clinically used as hypotensive agents. Nicorandil, besides its role as a KATP channel opener, can directly supply NO, a potent vasodilator substance. Some KCOs are used to treat angina since they dilate coronary vessels. Uterine myometrial smooth muscles express Kir6.1/SUR2B KATP channels, and their number gradually increases during pregnancy (112, 685). KCOs can inhibit uterine contraction during late pregnancy and might prove to be useful agents for prevention of premature labor (685).

4. Brain

In the hypothalamus there are a variety of “glucose-sensitive” neurons (24, 668). For example, in the mouse lateral hypothalamus, orexin/hypocretin neurons that regulate wakefulness, locomotor activity, and appetite are inhibited by an increase in glucose concentration. Melanin-concentrating hormone neurons that are involved in sleep and energy conservation are excited by an increase in glucose concentration. In the hypothalamus arcuate nucleus, an excitatory action of glucose on the anorexigenic pro-opiomelanocortin (POMC) neuron has been reported, while the appetite-promoting neuropeptide Y (NPY) neuron may be directly inhibited by glucose. These findings emphasize the fundamental importance of the glucose-sensing systems of hypothalamic neurons for orchestrating sleep-wake cycles, energy expenditure, and feeding behavior. Some neurons that are excited by glucose express KATP channels and are thought to employ a glucose-sensing strategy similar to pancreatic β-cells. In a majority of these neurons, elevated ATPi closes KATP channels, which causes membrane depolarization and thus increases excitation (24), and these neurons contain Kir6.2 (359, 529, 882). Some glucose-excited neurons in the ventromedial hypothalamus and glucose-sensitive striatal cholinergic neurons express KATP channels composed of Kir6.1 and SUR1 (435, 436). The mechanism of glucose-induced inhibition of neural activity is less understood. It has been proposed to involve modulation of the Na+-K+-ATPase (597) and activation of a hyperpolarizing Cl current that is presumed to involve CFTR-like Cl channels (668).

KATP channels may play a protective role in neurons under pathological conditions (35, 848). During severe cellular metabolic impairment such as ischemia and hypoxia, most mammalian neurons are known to depolarize and die. In the substantia nigra pars reticulate (SNr) (261, 545), KATP channels suppress neuronal activity during hypoxia by opening postsynaptic KATP channels (335, 546, 849). In contrast, in Kir6.2 knockout mice, these neurons exhibit no such hyperpolarization and are depolarized (849). The SNr acts as a central gating system in the propagation of seizure (128, 296), and the Kir6.2 knockout mice were extremely susceptible to generalized seizure after brief hypoxia (849).

E. Pharmacology

KATP channels are the targets of two major classes of therapeutic compounds, sulfonylureas and KCOs (18, 219). The most clinically important are the sulfonylurea KATP channel blockers that are used to stimulate insulin secretion in patients with type 2 diabetes mellitus (219). KCOs activate KATP channels with a broad range of potential therapeutic applications (18). All of these drugs affect KATP channels via their action on the SUR.

1. Sulfonylureas


Sulfonylureas such as acetohexamide, tolbutamide, glipzide, glibenclamide, and glimepiride contain a S-phenylsulfonylurea structure (144, 154, 832). The antidiabetic action of sulfonylureas was first reported in studies for sulfonamide antibiotics that were originally explored as an effective treatment for typhoid. The compound was unexpectedly found to induce hypoglycemia (607). Sulfonylureas are used exclusively for treatment of type 2 diabetes mellitus.

Sulfonylureas bind to SURx of KATP channels (see Fig. 12). In pancreatic β-cells, this binding inhibits K+ efflux through the KATP channels, thus causing cell membrane depolarization, opening of VDCC, the induction of Ca2+ influx, and insulin secretion.


KATP channels reconstituted by the coexpression of Kir6.2 with SUR1 or SUR2A exhibit different responses to sulfonylureas like native KATP channels in β-cells and cardiac myocytes. Kir6.2/SUR1 is more sensitive to sulfonylureas than Kir6.2/SUR2A (4, 299). The concentration-response curve for tolbutamide strongly suggests that Kir6.2/SUR1 channels harbor two independent binding sites with very different affinities (Ki = 2.0 μM for a high-affinity site and Ki = 1.8 mM for a low-affinity site). The high-affinity tolbutamide binding site S1237 is located in cytoplasmic loop 8 (CL8) between TM15 and TM16 (23, 527). Glibenclamide binding involves two sites on SUR1: the CL8 loop described above and the CL3 loop between TM5 and TM6 (23, 527). Mutation of S1237 in SUR1 to its SUR2 counterpart (Tyr) abolished both high-affinity tolbutamide block and 3H-labeled glibenclamide binding (23). On the other hand, Kir6.2ΔC36, the functional unit of Kir6.2 without SUR, was still blocked by high doses of tolbutamide (Ki = 1.7 mM). Thus the low-affinity block of KATP channels by tolbutamide seems to be due to its direct interaction with the Kir6.2 subunit (222, 223).


Under physiological conditions, a number of additional factors are involved in the action of sulfonylureas. Native KATP channels, analyzed in cell-attached configuration, are almost completely inhibited by sulfonylureas, whereas in excised membrane patches, the drugs can block the same channels by only 50–70% (219). Such a difference may be attributable to the presence of cytoplasmic nucleotides such as Mg-ADP, which itself influences Kir6.2/SUR1 via two mechanisms, strong activation through SUR1 and weak inhibition through the ATP-binding site on Kir6.2, with the former but not the latter counteracted by sulfonylureas. Therefore, the inhibitory effect of sulfonylureas on native β-cell KATP and Kir6.2/SUR1 channels is apparently enhanced by abolishing the stimulatory effect of Mg-NDP (31, 222). The interference of the binding of nucleotides to SUR1 and SUR2 by sulfonylureas has been confirmed in biochemical studies using 3H-labeled glibenclamide and 32P-labeled azido-ATP (241, 569, 701, 797). There may also be reciprocal interactions where Mg-ATP and Mg-ADP reduce binding of glibenclamide to SURs (242). On the other hand, Matsuo et al. (509) showed that glibenclamide abolished the effect but not the binding of ATP and ADP to SURs.

2. KCOs


KCOs include cromakalim, pinacidil, nicorandil, diazoxide, and minoxidil sulfate. In vascular smooth muscle, KCOs such as pinacidil, nicorandil, and diazoxide cause relaxation by opening KATP channels and reduce blood pressure (153, 317). Some KCOs have found a therapeutic role against hypertension. Nicorandil is clinically used to treat angina pectoris in Japan as well as in Europe. KCOs are also suitable for management of a variety of other diseases including acute and chronic myocardial ischemia, congestive heart failure, bronchial asthma, urinary incontinence, and certain skeletal muscle myopathies. Furthermore, openers of β-cell KATP channels such as diazoxide are clinically useful for the treatment of hypersecretion of insulin associated with insulinoma and persistent hyperinsulinemic hypoglycemia of infancy (see sect. ivF).

Native KATP channels in different tissues display distinct sensitivity to KCOs. The pancreatic β-cell KATP channels are readily activated by diazoxide, weakly activated by pinacidil, and unaffected by cromakalim or nicorandil (18, 20). In contrast, cardiac KATP channels are activated by pinacidil, cromakalim, and nicorandil but not by diazoxide (318, 773). The smooth muscle KATP channel is activated by all of these drugs (318, 635). These differences are attributed to the SUR subtypes expressed in each tissue, i.e., pancreatic SUR1, cardiac SUR2A, and smooth muscle SUR2B. Coexpression of Kir6.2 with an appropriate SUR can generate a KATP conductance with the pharmacological properties observed in each native tissue (218, 240, 299, 300, 317, 589, 680, 715).


In SUR2A and SUR2B, the binding sites for KCOs such as cromakalim, nicorandil, and pinacidil are within TMD2, which contains the cytoplasmic loop between TM13 and TM14 and the region encompassing TM16, TM17, and a short segment of NBD2 (28, 114, 797). Two residues in TM17 (L1249 and T1253 in SUR2A, T1286 and M1290 in SUR2B) have been shown to be necessary and sufficient for KCO action (540, 541). In SUR1, the segment including TM6-TM11 and NBD1 may contribute to the sensitivity to diazoxide (28). The binding sites for substrates in other polyspecific ABC proteins, such as multidrug resistance protein MRP1 and MRP3 (117, 324, 361, 497, 590, 650, 888) and multidrug resistance P-glycoprotein (468470), are located in regions homologous to TM17 of SUR. This suggests that KCOs may act as pseudo-substrates in SURs and occupy a substrate-binding pocket which is structurally conserved among SURs and multidrug resistance proteins (540) (see Fig. 12).


The presence of intracellular nucleotides affects the action of KCOs on SURs (Fig. 14) (220, 241, 683, 857). Electrophysiological studies suggest that interaction of Mg-nucleotides with NBDs not only increases the basal activity of KATP channels but also enhances their sensitivity to KCOs. For example, diazoxide cannot induce opening of cardiac-type KATP channels composed of Kir6.2/SUR2A in the absence of ADP but can do so in the presence of ADP (114). In addition, nicorandil requires the presence of ADP to activate cardiac-type channels (330, 709, 857). KCOs may also activate SUR2B/Kir6.1 (KNDP) channels by synergistically interacting with intracellular nucleotides (Fig. 14) (853). KCOs may increase ATP hydrolysis at NBD2 (47), and mutations which prevent ATP or ADP binding in NBDs in SUR2A and SUR2B abolished channel activation by nicorandil (646). Together, these results suggest that the mechanism of action of KCOs involves allosteric interactions between nucleotides and the NBDs and KCO binding sites in SUR (857).

Fig. 14.

Effects of KCOs on KATP and KNDP channels. SUR2B forms KATP channels with Kir6.2 and makes up KNDP channels with Kir6.1. The graphs schematically indicate the action of a KCO on the concentration-dependent effect of ATP in each channel type.

In summary, KCOs decrease the threshold concentrations of intracellular nucleotides for opening of Kir6.1 KNDP channels and also increase the maximum amplitude. On the other hand, KCOs decrease the sensitivity of Kir6.2 KATP channels to ATPi inhibition in a concentration-dependent manner that is mediated at least in part by NDP acting on the NBD dimer of SURs.


It is well known that some, but not all, KATP channel openers have hypertrichosis as a side effect. Minoxidil was first used to treat high blood pressure (147), but currently it is also used to treat baldness (627). Although the mechanism remains unclear, the effect is detected at low concentrations where minoxidil cannot augment the blood flow (71, 829). Minoxidil may have a role in maintaining vascularization of the hair follicle. The hair dermal papilla, which controls hair growth, is characterized in the anagen phase by a highly developed vascular network. In these cells minoxidil upregulates VEGF mRNA (424) whose product may promote vascularization (677).

3. Other drugs


Meglitinide, repaglinide, nateglinide, and mitiglinide are nonsulfonylurea acylaminoalcyl benzoic acid derivatives which represent a new chemical class of drugs for treating type 2 diabetes (51, 428, 495). They inhibit pancreatic β-cell KATP channels (5, 115, 154, 218, 524, 648, 751) and preferentially block Kir6.2/SUR1 channels (115, 587, 648, 751). Mitiglinide inhibits the binding of [3H]glibenclamide to microsomes from insulin-secreting HIT-15 cells and SUR1 (587, 751), suggesting that it interacts with the sulfonylurea binding site in SUR1 but not with the benzamide binding site in SUR2A (Fig. 12). PNU-37883A, PNU-89692, PNU-97025E, and PNU-99963 are blockers of a new generation that have been derived from a cyanoguanidine “KATP opener,” P1075. These compounds antagonize the effects of the vascular KCOs (228, 369, 522).

F. Diseases

To date, the majority of channelopathies involving KATP channels have involved pancreatic β-cells (16, 17, 629). Some mutations that can cause severe dysfunction of reconstituted channels (such as Q52R and V59G) may also disrupt KATP channel function in muscle and brain (629).

1. Loss-of-function mutations

Persistent hyperinsulinemic hypoglycemia of infancy (PHHI), which is also known as congenital hyperinsulinism of infancy (CHI), is characterized by excessive secretion of insulin despite hypoglycemia (148). The hypoglycemia causes irreversible brain damage. PHHI mutations include loss-of-function mutations in SUR1 (ABCC8) (294, 776), which account for ∼50% of the cases (148) and loss-of-function mutations in the gene encoding Kir6.2 (KCNJ11) (16, 252, 293, 560, 647, 777). The pathophysiological process is the loss of control of insulin section by any physiological process following the impairment of KATP channel activity. Patients with mild forms of the pathology may be managed with diazoxide. Some populations of familial PHHI induced by mutations in SUR1 may develop type 2 diabetes in later life (294, 491).

More than 100 mutations in the SUR1 gene and several tens of mutations in the Kir6.2 gene have been identified from the genome of PHHI patients. These mutations distribute throughout the genes. The KATP channel abnormalities associated with these mutations can be roughly classified into two categories: 1) defects of KATP channel trafficking to the surface membrane and 2) disruption of channel gating by nucleotides (16). Because SUR1 is crucial for the trafficking of Kir6.2 to the cell membrane, mutations that impair SUR1 synthesis, maturation, or subunit assembly result in the loss of surface expression of functional KATP channels (16, 148, 605, 769, 862). Mutations that reside in NBDs of SUR1 result in abnormal KATP channels with impaired opening in response to Mg-ADP (16, 148, 294). This dysfunction results in continuously closed KATP channels that no longer respond to alterations in extracellular glucose concentration. Several other mutations in Kir6.2 have been identified to cause PHHI not by impairing the trafficking of the channels to the membrane but by reducing KATP channel activity (16).

Dysfunction of other proteins can also influence KATP channel activity and cause PHHI. Gain-of-function mutations in the glycolytic enzyme glucokinase (GCK) encoded in GCK and the mitochondrial enzyme glutamate dehydrogenase (GDH) encoded in GLOUD1 cause PHHI (16, 148). Their enhanced synthesis of ATPi reduces KATP channel activity and elicits the phenotype.

PHHI patients with the mutations in SUR1 or Kir6.2 are generally insensitive to diazoxide, and their treatment often requires subtotal pancreatectomy. On the other hand, PHHI caused by mutations in the genes such as GCK or GLUD1 respond well to diazoxide (16, 148). The genotyping of PHHI patients is therefore important to determine the effective therapy.

Recently, frame-shift (Fs1524) and missense (A1513T) mutations of SUR2A, which occur at the catalytic ATPase pocket of NBD2 within SUR2A, have been found in patients with dilated cardiomyopathy (48). The abnormal KATP channels (Kir6.2 plus mutated SUR2A) show not only decrease of cell surface expression to some extent but also disruption of their functionality. The mutants have normal capability of ATP binding but show slower kinetics of ATP hydrolysis. As a result, the ATPi sensitivity of the KATP channels is impaired.

2. Gain-of-function mutations

Gain-of-function mutations in either Kir6.2- or SUR1-genes cause neonatal diabetes mellitus with reduced insulin secretion and resultant hyperglycemia (16). Approximately 50% of cases of permanent neonatal diabetes (PNDM) result from mutations in the Kir6.2 gene (151, 209211, 504, 675, 804). Mutations in the SUR1 gene (33, 155) are found in ∼27% of PNDM patients (16). All PNDM mutations reduce the sensitivity of KATP channels to ATPi. Several mutations in Kir6.2 cause extremely severe phenotypes including delayed development of speech and walking and muscle weakness in addition to neonatal diabetes. The most extreme cases show DEND syndrome characterized by marked developmental delay, muscle weakness, epilepsy, dysmorphic feature, and neonatal diabetes (210, 504, 629, 631, 675). The amount of the reduction in ATP sensitivity of the channels correlates with the severity of the phenotype (16).

There are at least two different molecular mechanisms underlying abnormality of KATP channel function in PNDM. First, mutations such as I182V or R201C/H in Kir6.2, which cause neonatal diabetes alone, directly impair ATP binding (16, 210, 211, 344, 629). These mutations are located in close vicinity to the putative ATP-binding site. Second, mutations including Q52R, V59M/G, and I296L in Kir6.2, which cause the severe neonatal diabetes observed in DEND syndrome, affect ATP-binding indirectly (629, 631). These mutations are located within or apposed to the slide helix that may be involved in channel gating and increase channel opening. In consequence, they indirectly reduce the sensitivity of the channels to ATPi (785). Most patients with this type of diabetes mellitus can be treated by oral application of sulfonylureas.

3. KATP channel knockout mice

Studies on genetically modified mice also suggest the involvement of abnormalities in KATP channels in a number of different diseases. Most Kir6.1-knockout mice die between 5 and 6 wk after birth because of a high rate of sudden death syndrome (531). Their electrocardiograms show spontaneous elevation of S-T segments followed by atrioventricular block, indicating that their death is attributable to myocardial ischemia resulting from hypercontractility in coronary arteries (531). This phenotype resembles the symptoms of Prinzmetal's angina, which is characterized by recurrent episodes of chest pain under resting conditions (531). The disruption of SUR2 can also result in transient, repeated episodes of coronary artery spasm similar to the phenotype of Prinzmetal's angina (92). See section ivD for phenotypes of mice lacking Kir6.2 or SUR1 gene.

V. K+ Transport Channels (Kir1.1, Kir4.x, Kir5.x, Kir7.x)

A. Kir1.1

1. Historical view and molecular diversity

Kir1.1/KCNJ1 is the first of 15 members of the Kir family that have been cloned (272). It was initially described as the “rat outer medullary K+ channel” ROMK1. Six alternative splicing isoforms have been identified so far (ROMK1-6) (53, 141, 272, 385, 717, 892). The products translated from mRNA of isoforms ROMK4-6 are the same as the ROMK2 protein. The NH2 termini of ROMK1 (Kir1.1a) and ROMK3 (Kir1.1c) are, respectively, 19 and 26 amino acids longer than the shortest isoform, ROMK2 (Kir1.1b). Like other Kir channels, the Kir1.1 subunit functions as a tetramer (440). Heteromeric assemblies of Kir1.1 with members of other Kir families have not been reported.

2. Pore structure and function

Kir1.1 exhibits weak inward rectification, a single-channel conductance of ∼35-pS, and high Po over a wide range of Em when expressed in Xenopus oocytes (53, 83, 272). The deduced amino acid sequence of Kir1.1 shows that it harbors a Walker type A ATP-binding motif at its COOH terminus. The functional role of this region has not been established (56, 86, 272).


The weak inward rectification of Kir1.1 is attributed to the uncharged N171 at the “D/N site” of the TM2 helix (see sect. iC1).


Intracellular, but not extracellular, pH modulates the activity of Kir1.1. Acidification closes the channels with a pKa of ∼6.5 (88, 138, 161, 520, 789, 816). Mutation analyses and homology modeling suggest that pHi sensitivity of Kir1.1 involves residues in three categories: 1) positively charged residues R41, K80, and R311; 2) a residue (L160) in the “bundle crossing” region; and 3) residues in the TM2 transmembrane region.

Although R41, K80, and R311, the so-called RKR triad, were all originally implicated in the sensitivity of Kir1.1 to pHi (88, 161, 520, 696, 816), homology modeling rather suggests that they are structurally important elements in the general functionality of Kir1.1, and therefore, their mutation only indirectly influences the reactions of the channel to pHi (645). To begin with, R41 and R311 are too far from K80 (>24 Å) to influence its ionization state, and then, K80 seems to be buried within the lipid membrane and inaccessible to solvent. Instead, the NH2 terminus residue R41 forms a stable intrasubunit salt bridge with the COOH-terminal residue E318; the COOH-terminal residue R311 forms a stable intersubunit ion pair with E302 on the COOH terminus of the adjacent subunit in the tetramer; K80 in TM1 and A177 in TM2 occur in the “helix-bundle crossing” region, are <3 Å apart and form a potential intrasubunit interaction. Mutation of any of these six residues alters the pHi sensitivity of Kir1.1 (645). R311W and A177T, which are found to cause Bartter's syndrome, change pKa of Kir1.1 (∼6.5) to 9.2 and 7.8 in Kir1.1, respectively, which results in dysfunction of the channel in physiological pHi range (617, 645, 696) (see also sect. vA6 and Table 2).

The hydrophobic residue L160 may be involved in pHi-dependent gating, replacing L160 by a smaller residue, Gly, abrogated pHi sensitivity of Kir1.1 (672). Since L160 is homologous to Phe in the bundle-crossing region in KirBac1.1 (F146) (404) (see sect. iD), closure of Kir1.1 by acidic pHi could result from steric occlusion of the permeation pathway by the convergence of four Leu residues at the cytoplasmic apex of the inner (TM2) transmembrane helices.

In Kir1.1b, G148 and G157 (G167, G176 in Kir1.1a) are located close to the cytoplasmic end of the inner (TM2) transmembrane helix, and mutation of either G148 or G157 to Ala shifted pKa from 6.6 (wild type) to 7.1 and 7.3, respectively (673). These two Gly residues are conserved in all Kir subunits and may contribute to pHi sensitivity in Kir1.1.

Intracellular acidification causes irreversible collapse of the Kir1.1 channel pore in the absence of extracellular K+. The Kir1.1-A177T mutant channel shows a strong alkaline shift of pHi sensitivity, and the addition of a K80I mutation led to recovery from acid inhibition in the absence of extracellular K+ (645). The rate of increase of the Kir1.1 current on switching of extracellular K+ from 1 mM (almost no current) to 100 mM (large increase of the current) showed that the substitution of L160 for Gly prominently reduced the rate of increase (671). Although how L160 in Kir1.1 is involved in gating of Kir1.1 has remained elusive, a recent study has suggested a mechanism underlying pHi-dependent regulation of the channel's activity via K80 (on TM1) and A177 (on TM2) in bundle-crossing region (644). Homology modeling of Kir1.1 shows that these two residues are bridged by H bonding in the bundle-crossing region. Mutations of K80 to other residues that are predicted to disrupt the H bonding not only lower pHi sensitivity of Kir1.1 but also speeds up recovery of the channel activity by alkalization from acidification-induced inhibition. Furthermore, the H bonding seems to mediate slow kinetics of reactivation of the rundown channel by intracellular PtdIns(4,5)P2 application. Together, the H+ bonding between K80 and A177 at the bundle-crossing region may critically control not only pH-dependent but also PtdIns(4,5)P2-mediated gating of Kir1.1. An additional finding is that this bonding is also required for the process of aforementioned K+-dependent collapse of Kir1.1 pore, a possible phenotype of dysfunction of the selective filter. Therefore, the H bonding at the bundle-crossing region may be critically involved in the conformational change of the pore region.

3. Intracellular localization

Kir1.1 has an ER retention signal, R-X-R (R368-A369-R370), in the COOH terminus (488). Trafficking to the cell surface is regulated by posttranslational modification via several intracellular protein kinases including PKA and SGK. Therefore, a number of physiological signaling pathways modulate the contribution of the Kir1.1 current by promoting or suppressing its surface expression.


Kir1.1 channel activity (Po) requires a PKA-dependent phosphorylation process (521), and Kir1.1 has three PKA phosphorylation sites (S44 in the NH2 terminus, S219 and S313 in the COOH terminus of Kir1.1a). Mutation of any one of these sites reduces the whole cell current expressed in Xenopus oocytes by 35–40% without affecting single-channel conductance. Mutation of any two of the three sites abolishes visible channel activity (847). Phosphorylation of S44 is considered to augment cell surface expression of the Kir1.1 protein (247). PKA-mediated phosphorylation of S219 and S313 increases channel Po (490) possibly by enhancing the interaction between the channel and PtdIns(4,5)P2 (459).


S44 on the NH2 terminus of Kir1.1 can also be phosphorylated by SGK-1. This increases surface expression of Kir1.1 and thus enhances the current (877). Since aldosterone stimulates SGK-1 transcription, the Kir1.1 current, like epithelial Na+ channels, could be regulated by this hormone. However, another study reported that cell surface expression and current density of Kir1.1 are not altered by coexpression of SGK-1 alone but are augmented by the expression of SGK-1 and a scaffolding protein, the Na+-H+-exchanger regulatory factor-2 (NHERF-2) (879). The mechanism of this augmentation is still elusive, because NHERF-2 alone had little effect on Kir1.1 channel current. NHERF-2, which binds multiple ion transport systems, and Kir1.1 are shown to be directly associated with each other by a biochemical assay and colocalized at the apical membrane of renal epithelial cells by an immunocytochemical study (712, 876). Thus cell surface expression of Kir1.1 may be regulated by a combination of SGK-1 and NHERF-2 in vivo.

Two recent studies clarified the mechanism underlying the increase of cell surface expression of Kir1.1 by S44 phosphorylation. S44 phosphorylation suppresses the ER retention signal, R-X-R, which allows delivery of Kir1.1 to the cell surface (582). S44 phosphorylation may also override another ER location signal (K370 and R371) and promote surface expression of Kir1.1 (875).


PKC phosphorylates two Ser residues in Kir1.1a, S4 and S201 (456). Although the phosphorylation promoted surface expression of Kir1.1a in Xenopus oocytes and HEK cells (456), in renal CCD cells it inhibited Kir1.1 channels (819) (see sect. vA4). It has been suggested that the suppression of Kir1.1 by PKC is due to the reduction of membrane PtdIns(4,5)P2, albeit by an unknown mechanism (885).

I) With-no-K (K=Lys) kinases (WNKs).

The localization and function of Kir1.1 are also controlled by other serine-threonine kinases such as WNK1, WNK3, and WNK4. All of these kinases reduce Kir1.1 current by reducing the amount of Kir1.1 in the plasma membrane, although through distinct mechanisms (347, 439, 814). Kir1.1 is retrieved from the cell surface by a clathrin-dependent mechanism via the internalization motif, N-P-X-Y, in the COOH-terminal region (884). The effect of WNK4 is attributed to enhanced clathrin-dependent endocytosis and not its kinase activity (347). Inhibition of Kir1.1 current by WNK3 is also independent of its catalytic activity (439). On the other hand, WNK1 seems to inhibit cell surface expression of Kir1.1 by phosphorylating certain residues within the channel protein (814).

Mutations in either of the WNK1 or WNK4 genes are associated with pseudo-hypoaldosteronism type II (PHAII: also referred to as Gordon's syndrome) which exhibits hypertension with hyperkalemia in an autosomal dominant form (841). Abnormal modulation of Kir1.1 function by the mutated WNKs may be linked with these phenotypes. Mutations of WNK4 that cause PHAII also increase its inhibitory effect on Kir1.1 (884).

II) Ubiquitination.

K22 in the NH2 terminus of Kir1.1 is monoubiquitinated (457). Substitution of K22 by Arg greatly enhanced the Kir1.1 current by increasing its cell surface expression. Therefore, ubiquitination acts as a negative regulator of Kir1.1.

4. Physiological functions in cells and organs


Ion concentrations and the volume of extracellular fluids are finely controlled by the kidney. To perform this task, various types of ion and water transport mechanisms are expressed and functionally coupled together in the apical and basolateral membranes of renal epithelial cells (247, 818). For homeostasis of the urine and blood, K+ channels are crucial in regulating not only [K+] but also the concentrations of other ions such as Na+ and Cl (247). Kir1.1 is considered to play a key role in this regulation. Kir1.1 is expressed in renal epithelial cells of the thick ascending limb of the loop of Henle (TAL), the distal convoluted tubule (DCT), and the cortical collecting duct (CCD) (Fig. 15A) (161, 247, 437). Its isoforms (Kir1.1a-c) are differentially distributed along the nephron (Fig. 15A) (53). Immunohistochemical examinations show that Kir1.1 is specifically localized at the apical membrane of the epithelial cells (Fig. 15A) (384, 600, 846). Patch-clamp single-channel recording and knockout mouse studies showed that Kir1.1 formed 35-pS and 70-pS K+ channels at the apical membrane of TAL cells as well as 35-pS K+ channels in CCD cells (247, 475, 480).

Fig. 15.

Distribution and localization of Kir channels in the kidney and their physiological impact. A: distribution and localization of Kir1.1 channel isoforms and Kir4.1/5.1 heteromeric channels in the kidney. Each Kir1.1 isoform (Kir1.1a, -b, and -c) is expressed in particular parts of the renal epithelia (top panels). The distal straight tubule is also known as the thick ascending limb of the loop of Henle (TAL). Bottom panels indicate the cellular location of each type of channel subunit. Kir1.1 channel isoforms are expressed in the apical membranes of renal epithelia. Kir4.1/5.1 heteromer channels occur in the basolateral membrane. B: ion transport mechanisms associated with Kir channels. In TAL cells (left), Kir1.1 channels in the apical membrane are colocalized with the Na+-K+-2Cl cotransporter. In distal convolute tubule (DCT) cells (right), the heteromeric Kir4.1/5.1 channel is colocalized with the Na+-K+-ATPase in the basolateral membrane. The process known as “K+ recycling” involves these K+ channels supplying K+ to the extracellular K+ site of the transporters.

TAL cells in the kidney reabsorb ∼25% of filtered Na+ from the urine mainly via the Na+-K+-2Cl cotransporter (NKCC) located on the apical membrane. K+ channels made up of Kir1.1 are thought to supply K+ to the extracellular K+ site of NKCC and keep this transporter active (“K+-recycling” action; see sect. vB4) (Fig. 15B) (247, 818). The channels also facilitate Cl absorption via the transporter and thus promote uptake of NaCl. Na+ accumulated in cytoplasmic region would be subsequently carried to the extracellular fluid via Na+-K+-ATPase at the basolateral membrane. Thus functional coupling of Kir1.1 and NKCC at the apical membrane results in driving the unidirectional transport of Na+ from apical to basolateral sides and sustains the activity of the ATPase. Furthermore, Kir1.1 hyperpolarizes the Em of TAL cells, which results in accelerating Cl exit from the basolateral membrane via Cl channels (247). It establishes “the lumen-positive transepithelial potential,” which is the main driving force for paracellular Na+, Ca2+, and Mg2+ transport from lumen to the side of blood vessels (50, 217). These functions were confirmed in Kir1.1 knockout mice, which exhibited impairment of renal NaCl absorption and “Bartter's syndrome”-like phenotypes (475, 480) (see sect. vA6).

The Cl channel CFTR occurs together with Kir1.1 in the apical membrane of TAL cells, and it may regulate the function of Kir1.1 (478). The apical ∼30-pS K+ channels in TAL cells of wild-type mice are largely suppressed by physiological levels of ATPi and blocked by glibenclamide (817). These two properties seem to result from the presence of functional CFTR because they are not observed in mice lacking CFTR or in mice homozygous for the CFTR loss-of-function ΔF508 mutation (478). In addition, when CFTR associates with Kir1.1, the activity of this K+ channel decreases. The effects of CFTR on Kir1.1 are probably achieved by their interactions with PDZ-domain binding protein NHERFs (876), which can be abrogated by activation of PKA (478). The physiological roles of the functional association between CFTR and Kir1.1 may be involved in the effects of the antidiuretic hormone arginine vasopressin (AVP), which increases the apical K+ conductance of TAL (173). When urinary flow is augmented, secretion of AVP from pituitary is suppressed and the apical K+ conductance is reduced by NHERF-mediated interaction between CFTR and Kir1.1. This process would limit K+ secretion from renal tubules and thus urinary K+ loss.

Kir1.1 also seems to be critically involved in ion transport in CCD cells which play a central role in secretion of K+ into the urine (205, 206). K+ secretion occurs in principal cells of the CCD via 35 pS K+ channels (191, 192) which are probably made up of Kir1.1, since principle cells of Kir1.1 knockout mice did not exhibit this K+ current (480).


In situ hybridization showed the Kir1.1 transcript to be expressed in neurons of the cortex and hippocampus but not in other areas of the brain (366). The physiological function of Kir1.1 in the brain remains unknown.

5. Pharmacology

In addition to the nonspecific Kir channel blockers Ba2+ and Cs+, δ-dendrotoxin (Kd 150 nM) and tertiapin (Kd 1 nM) inhibit Kir1.1 channels (298, 342, 494) (see sects. iE and iiiE). Both toxins bind to the external vestibule of the channel pore. The blocking effect of tertiapin is sensitive to pH: alkaline pH diminishes the dissociation constant of the interaction between the toxin and Kir1.1 (642). It seems probable that one tertiapin molecule binds to one functional channel. The binding site is the K+-conduction pore formed by the linker between the first and second transmembrane (TM1-TM2) segments (168, 341, 641). Mutations showed that Met at position 13 in tertiapin interacts with F148 of Kir1.1 (343), a residue located near the external opening of the pore (404, 573). The action of tertiapin is specific to Kir1.1 and Kir3.1/3.4 channels with Kd values of 1–2 nM and 8–10 nM, respectively. Little effect is observed on other types of Kir channel (298, 342, 494). Of interest is a mutated tertiapin where H12 and M13 were, respectively, substituted for Leu and Gln and which showed a greatly reduced affinity for Kir3.1/3.4 but not for Kir1.1 (see sect. iiiE) making the mutated tertiapin a relatively specific blocker for Kir1.1 (Kd for Kir1.1: 1 nM versus Kd for Kir3.1/3.2 and Kir3.1/3.4: 274 and 361 nM, respectively) (643).

6. Diseases

The pathophysiological importance of Kir1.1 is shown by the phenotypes of the Kir1.1 knockout mouse (475, 480) and its “loss of function disease,” Bartter's syndrome. Bartter's syndrome (38) is an autosomal recessive renal tubulopathy that is characterized by hypokalemic metabolic alkalosis, renal salt wasting, hyperreninemia, and hyperaldosteronism (26, 227, 358, 618, 659). This disease can be classified into four types (I–IV) (583). Genetic analysis revealed that type II Bartter's syndrome results from various mutations in the Kir1.1 gene (167, 247, 729, 755, 809) (Table 2). In TAL cells, abrogation of Kir1.1 function greatly diminishes K+ supply to NKCC. In addition, the lumen-positive transepithelial voltage, which drives ∼50% of the paracellular reabsorption of Na+, is lost (50, 217). The mutations in Kir1.1 associated with Bartter's syndrome cause alterations in PKA phosphorylation, pH sensing, channel gating, proteolytic processing, and sorting to the apical membrane (130, 131, 185, 186, 332, 474, 617, 696, 700, 743). One study demonstrated that R311 is crucial for the interaction between Kir1.1 and PtdIns(4,5)P2 and that the Bartter's mutations (R311Q/W) could disrupt this interaction (474), although this idea is disputed (645).

Ablation of the Kir1.1 gene in mice causes a renal Na+, Cl, and K+ wasting phenotype (475, 480), which is consistent with the channel's crucial role in salt absorption in TAL. These mutant mice also mimicked the phenotypes of Bartter's syndrome. In the Kir1.1 knock-out mice, patch-clamp recording in either TAL or CCD cells failed to show either 35- or 70-pS K+ channels in the apical membrane (479, 480). Like Bartter's syndrome patients, knockout mice showed hypokalemia and an excess of K+ in their urine. The mechanism of this phenotype was initially elusive, since Kir1.1 had been proposed to provide the major K+-secretory pathway in renal tubules (190, 191, 215, 247). Dysfunction of the channel was expected to cause hyperkalemia and, indeed, infants with the type II Bartter's genotype often exhibited transient hyperkalemia (84, 180). A recent study clarified the mechanism (34): in mice Kir1.1 knockout reduces K+ absorption from urine in the loop of Henle by impairing function of NKCC2 in TAL cells but sustains K+ secretion by continuous K+ efflux via maxi-K+ (namely, large-conductance Ca2+-activated K+ or BK) channels in the late distal tubule.

B. Kir4.x and Kir5.1

1. Historical view and molecular diversity

Kir4.1/KCNJ10 was initially and independently identified from a brain cDNA library by several groups and assigned different names such as BIR10 (56), KAB-2 (763), BIRK-1 (64), and Kir1.2 (718). In situ hybridization histochemistry showed that Kir4.1 was predominantly expressed in glial cells of the brain (763) (see sect. vB4). Therefore, Kir4.1 has been considered to form a glial K+ conductance that is involved in so-called “K+-buffering” which plays a pivotal role in controlling neuronal function (400, 563) (see sect. vB4). The salmon homolog of mammalian Kir4.1 is known as Kir4.3 (395). Kir5.1/KCNJ16 was first described as BIR9 (56).

The deduced amino acid sequence of Kir4.1 has 53, 43, and 43% identity with Kir1.1, Kir2.1, and Kir3.1, respectively, whereas Kir5.1 shares 39, 50, and 40% identity. When Kir4.1 is expressed alone in a heterologous expression system, it forms a tetramer and elicits a K+ current (616). In native tissues, the channel also exists as a tetramer (254, 348). When Kir5.1 is expressed alone in a heterologous expression system, it is nonfunctional, but coexpression with Kir4.1 forms a functional heteromer whose biophysical properties and pHi sensitivity differ from those of the Kir4.1 homomer (see sect. vB2) (56, 616, 765, 793). The Kir4.1/5.1 heteromer also exists as a tetramer (616), and it has been detected in various tissues including the kidney and the brain. The Kir5.1 homomer is expressed in the cytoplasm of cells such as fibrocytes in the cochlea where it does not seem to form functional channels on the cell membrane surface (see sect. vB3).

Kir4.2/Kir1.3/KCNJ15 was first isolated from a human kidney cDNA library (718). Another group identified the Kir4.2 gene by sequencing the Down syndrome chromosome region 1 on subband q22.2 of chromosome 21 (214). Kir4.2 has 62% identity to human Kir4.1. Kir4.1 has a Walker type A ATP-binding cassette in its COOH terminus (763). Kir4.2 does not possess this motif (609). Kir4.2 forms a functional channel when expressed alone in heterologous expression systems (609, 615). It also forms a functional heteromer when coexpressed with Kir5.1 with a different single-channel conductance and greater cell surface expression resulting in increased whole cell current (609, 615) (see sect. vB2). Kidney, liver, embryonic fibrocytes, and microvascular endothelial cells are reported to express Kir4.2 (124, 425, 609, 718).

2. Pore function and structure

A) KIR4.1 AND KIR5.1.

There have been differing reports of the single-channel conductance of the Kir4.1 homomer and Kir4.1/5.1 heteromer channels, possibly because they can show multiple subconductance states. The conductance of the Kir4.1 homomer ranges between 20 and 40 pS, while that of the Kir4.1/5.1 heteromer ranges between 40 and 60 pS (476, 616, 763, 765, 868). The Kir4.1 homomer exhibits intermediate inward rectification, whereas the rectification profile of the Kir4.1/5.1 heteromer is much stronger. The residues of the D/N site are Glu in Kir4.1 (E158) and Asn in Kir5.1 (N161) (see sect. iC1). In excised membrane patches, even high concentrations of spermine left 10–15% residual conductance of the Kir4.1 homomer. This apparent insensitivity may be due to partial permeation of spermine via the Kir4.1 channel pore because philanthotoxin, a polyamine with a bulky tail, more completely blocked outward current through Kir4.1 (399). The E177 residue in the proximal region of the COOH terminus just under the TM2 region in Kir4.1 is required for heteromer formation with Kir5.1 (388).

An even more remarKABle difference between homomeric Kir4.1 and heteromeric Kir4.1/5.1 channels is their different sensitivity to changes in pHi. In the physiological range of pHi between 6.5 and 8.0, the channel activity of the Kir4.1 homomer is inhibited by acidification with a pKa of ∼6 (56, 616, 765, 868). In contrast, over the same pHi range, the activity of the heteromeric channels is dramatically suppressed by only a slight intracellular acidification and enhanced by alkalization with a pKa of ∼7.5 (765, 868). Several residues in Kir4.1, E158 in the TM2 region (845), K67 in the cytoplasmic NH2-terminal region (868), and H190 in the COOH-terminal region (76) are reported to be involved in the pHi sensitivity of not only homomeric Kir4.1 but also heteromeric Kir4.1/5.1 channels. The mechanism underlying the modification of pHi sensitivity of Kir4.1 by Kir5.1 remains elusive.

Of all Kir channel subunits, Kir4.1 has been found to have the strongest affinity for PtdIns(4,5)P2 (146).

The activity of the heteromeric Kir4.1/5.1 channel is enhanced by Nai+ (665). This property seems to be exerted by D205 in the cytoplasmic COOH terminus of Kir5.1, since the corresponding residue confers Na+ sensitivity to both Kir3.2 and Kir3.4 ion channels (D226 and D223, respectively) (see sect. iiiB). This residue is not conserved in Kir4.1 (N202), and the Kir4.1 homomer channel is not sensitive to Nai+.

B) KIR4.2.

Kir4.2 displays a different biophysical profile. The single-channel conductance of the homomer of Kir4.2 expressed in Xenopus oocytes is ∼25 pS. The amino acid sequence and rectification profile of Kir4.2 are similar to Kir4.1, but Kir4.2 is more sensitive to pHi (Kir4.2: pKa = 7.1, Kir4.1: pKa = 6). Coexpression with Kir5.1 has only minor effects on the pHi sensitivity of Kir4.2 (Kir4.2/5.1: pKa = 7.6). The heteromeric channel exhibits a larger unitary conductance, ∼54 pS (615). The Kir4.2/5.1 heteromer is more sensitive to Ba2+ block than the Kir4.2 homomer (609). Interestingly, mutation K66M in the NH2-terminal region of Kir4.2 almost abrogated its pHi sensitivity (615). This is consistent with the mutation of the corresponding residue (K80), which impairs pHi sensitivity in Kir1.1 (88, 161, 520, 696) (see sect. vA2). It is however impossible to explain the difference between the pHi sensitivity of Kir4.1 and Kir4.2 solely on the role of this Lys residue, because the intracellular COOH-terminal parts of these subunits also appear to be responsible for the difference (615). The crucial residues in this region have not yet been identified.

3. Intracellular localization

I) Localization in cell membranes. A) Epithelial tissues.

When Kir4.1 is expressed in epithelial MDCK cells, it is sorted to the basolateral membrane (764) by association with a PDZ-protein membrane-associated guanylate kinase with inverted domain structure 1 (MAGI-1a long isoform) (766). The interaction occurs between the fifth PDZ domain of MAGI-1a and the PDZ-binding motif in the COOH terminus of Kir4.1. The phosphorylation of S377 in the PDZ-binding motif of Kir4.1 disrupts this interaction and moves Kir4.1 from the basolateral membrane to perinuclear compartments in the transfected MDCK cells (766). When Kir5.1 is expressed alone in MDCK cells, it remains in the intracellular compartment, but coexpression with Kir4.1 targets it to the basolateral membrane (764). Consistent with this pattern, immunohistochemical, biochemical, and patch-clamp studies have demonstrated that Kir4.1 and Kir5.1 form heteromers in the basolateral membrane of some renal epithelial cells (325, 476, 765, 793) (Fig. 15A) (see sect. vB4).

On the other hand, in the stomach, homomeric Kir4.1 is specifically distributed at the apical membrane of parietal cells (193, 364). Careful analysis revealed that the channel was located at the microvilli of the apical membrane surface but not in the tubulovesicles. The mechanism for determining such specific localization of Kir4.1 in parietal cells remains unknown.

The distribution pattern of Kir4.1 and Kir5.1 in the cochlea of the inner ear, which contains a variety of epithelial cells, is unique (260). Each subunit is expressed in the cochlear lateral wall but in different components. Kir4.1 is found in epithelial tissue in the stria vascularis (257). High-resolution immunolabeling shows that this channel is specifically expressed on the apical side of intermediate cells (14, 255). Kir5.1 is expressed in fibrocytes of the spiral ligament and localized in intracellular compartments rather than on the membrane surface (255).

B) Glial cells.

In the central nervous system (CNS) and retina, Kir4.1 and Kir5.1 subunits are predominantly expressed in astroglial cells. They form both Kir4.1 homomers and Kir4.1/5.1 heteromers, each of which exhibits a unique distribution pattern in distinct cell types.

Kir4.1 is widely distributed in astrocytes of the CNS in both brain and spinal cord (262, 348, 449, 561, 594, 624, 763). Kir5.1 is also expressed in brain astrocytes (45, 254, 454). Both subunits are present in Müller cells of the retina (311, 312, 549). In the cortex of the brain, both Kir4.1/5.1-heteromeric and Kir4.1-homomeric channels are detected in astrocytic processes surrounding synapses (perisynaptic processes), whereas only the heteromer is present in processes attached to pia and blood vessels (end feet) (Fig. 16A). Retinal Müller cells express the Kir4.1 homomer at the end feet which face vitreous humor and blood vessels, and the Kir4.1/5.1 heteromer in the perisynaptic processes (311) (Fig. 16B).

Fig. 16.

Localization of astroglial Kir channels. A: brain astrocytes. Astrocytes harbor and localize homomeric Kir4.1 channels and heteromeric Kir4.1/5.1 channels in particular parts of their membrane: both the homomer and the heteromer are present in perisynaptic processes, and the heteromer is situated alone in the end feet. B: retinal Müller cells. These cells express the heteromeric Kir4.1/5.1 channel at perisynaptic sites and the homomeric Kir4.1 channel at perivascular processes and in end feet. In both the astrocytes and Müller cells, the Kir channels coexist with the astroglial water channel aquaporin 4 (AQP4) in the same membrane domains (A and B).

The machinery responsible for sorting each channel to different membrane macrodomains in astroglial cells of either brain or retina have not yet been completely clarified. Several studies suggest that the dystrophin-associated protein complex plays a role in determining localization of Kir4.1 in the end feet. Mutant mice lacking dystrophin (called “mdx3Cv mice” or knockout mice) show reduced Kir4.1 expression in the end feet of Müller cells (101, 116) and display a diffuse localization pattern in isolated cells (187). On the other hand, in mdx3Cv mice, the expression of Kir4.1 at the perisynaptic membrane surface is unaffected, suggesting that other mechanisms traffic Kir4.1 channels to this area.

Satellite cells are another subset of glial cells which also harbor Kir4.1 channels (256, 807, 808, 903). These cells surround the cell bodies of various ganglion neurons with multiple layers of myelin sheath. Kir4.1 is expressed specifically on the sheath membrane surrounding cochlear spiral ganglions, trigeminal ganglions, and superior cervical ganglions (256, 807, 808, 903) (see sect. vB4 for functional roles).

II) Localization in membrane microdomains.

Kir4.1 is concentrated in particular microdomains of the astroglial membrane known as DRMs (see sect. iC2) (259). Biochemical and immunohistochemical assays found that Kir4.1 proteins were abundant in noncaveolar DRMs in the brain as well as in cultured astrocytes and when exogenously expressed in HEK293T cells. In HEK293T cells, depletion of membrane cholesterol by methyl-β-cyclodextrin (MβCD) resulted in loss of the association between Kir4.1 and DRMs and channel activity (Fig. 17A). Therefore, localization of Kir4.1 in MβCD-sensitive noncaveolar DRMs seems to be mandatory for channel function. The astroglial water channel AQP4 also occurs in DRMs, but these microdomains are MβCD insensitive and thus suggested to contain little cholesterol (Fig. 17B). Immunolabeling showed that these different DRMs occurred in close proximity in the astrocytic membrane (Fig. 17, C–E), and such spatial organization may be involved in the coupling of K+ and water transport (see sect. vB4).

Fig. 17.

Kir4.1 and AQP4 exist in different membrane microdomains. A: the subcellular localization of Kir4.1 (top panel) and AQP4 (bottom panel) in the brain. Mouse brains treated with 1% Triton X-100 were fractionized on a discontinuous sucrose gradient. Each fraction was collected from the top of the gradient and immunoblotted with specific antibodies against the proteins indicated on the left. The DRM fractions concentrate Kir4.1 and AQP4. Note that Kir4.1 is also found in non-DRM compartments. Western blotting reveals monomer (x1), dimer (x2), and tetramer (x4) forms of Kir4.1. B: DRM localization of Kir4.1 and AQP4 expressed in HEK293T cells and their cholesterol dependence. The fractions obtained from cells transfected with Kir4.1 (top panels) or AQP4 (bottom panel) cDNA were probed with anti-Kir4.1 or anti-AQP4 antibody. Control, nontreated cells; MβCD, cells treated with MβCD, a reagent that depletes membrane cholesterol (4 mM, 1 h). Two bands show glycosylated (top) and unglycosylated (bottom) Kir4.1. MβCD treatment disrupts DRM association of Kir4.1 but not that of AQP4. C: localization of AQP4 and Kir4.1 in in vivo astrocytes. Cryosections from mouse brain were treated with anti-Kir4.1 antibody, followed by incubation with Alexa488-labeled secondary antibody and Alexa568-labeled anti-AQP4 antibody. a and b show images obtained by focusing at different depths. Solid arrowheads indicate “line-shaped” (a) and arrows (b) indicate “clustered” labeling of AQP4 (red) and Kir4.1 (green). In b, the edge of the wall of a blood vessel is visible (open arrowheads), confirming that the microscope was focused on the surface of the wall. D: a high-magnification image of the boxed region in Cb. AQP4 and Kir4.1 signals are not colocalized but can exist in close vicinity (arrowheads). [A–D from Hibino and Kurachi (259), with permission from Wiley-Blackwell Publishing.] E: a schematic representation of the localization of microdomains which contain Kir4.1 and AQP4 on the astroglial membrane. Kir4.1 and AQP4 are expressed respectively in MβCD-sensitive and MβCD-resistant DRMs, which are occasionally in close proximity.

B) KIR4.2.

When Kir4.2 is expressed alone in heterologous systems, it elicits a relatively small current (131, 609, 718). This seems to be due to the intracellular localization of a large population of the channel. The intracellular localization of Kir4.2 may result from two independent mechanisms, which is suggested by the following observations. First, mutation of a single residue (K110N) in the extracellular domain between the first transmembrane helix and the pore helix/selectivity filter dramatically increases the membrane current (131). Since there is no difference in Po between K110N-Kir4.2 and wild-type Kir4.2 channels (131), this mutation should facilitate the trafficking of the channel protein to the plasma membrane. Second, deletion or mutation of a potential Tyr phosphorylation motif, K-(X)2–3-E-(X)2-Y, in the COOH terminus also augmented the membrane current, which may also be due to facilitation of trafficking of the channel to the membrane surface (610). However, it is not known how K110N and the phosphorylation motif regulate the trafficking of Kir4.2 protein. Of note is the fact that none of the PDZ-domain binding motif (S-X-V), the dileucine targeting motif [D/E-(X)3–5-L-L], and the ER retention signal (R-X-R), all of which are found in the COOH terminus of Kir4.2, seem to be involved in its intracellular localization (610).

Kir4.2 is expressed at the basolateral membrane of renal epithelial cells when it forms a heteromer with Kir5.1 (476).

4. Physiological functions in cells and organs

I) Kidney and stomach.

In the kidney, the epithelium of DCT plays an important role in reabsorption of Na+ (247). This is mediated by epithelial Na+ channels at the apical membrane (105) coupled with the activity of the basolateral Na+-K+-ATPase (Fig. 15B) (see sect. vA4). K+ channels in the basolateral membrane are key players in this process because they supply K+ to the extracellular K+ site of the Na+-K+-ATPase to maintain its activity. This action is called “K+ recycling” (Fig. 15B). Consistently, blocking the basolateral K+ conductance of renal epithelial cells with Ba2+ inhibits vasopressin-stimulated Na+ transport (687).

Kir channels with a unitary conductance of 37–40 pS have been identified in the basolateral membrane of DCT in rabbit (768) and mouse (476). In the latter case, the channels are inhibited by intracellular acidification with a pKa of ∼7.6 (476). It was shown that these channels consisted of heteromers of Kir4.1/5.1 and/or Kir4.2/5.1 (Fig. 15B) (476, 765, 793). Their pHi sensitivity strongly suggests that they act as pHi sensors and are involved in pHi-dependent regulation of ion transport in the kidney.

In the DCT, either Kir4.1 or Kir4.2 may directly interact with a Ca2+-sensing receptor in the basolateral membrane (284). In heterologous expression systems, this interaction significantly reduces the current elicited by either subunit. Although the physiological role of this interaction is unknown, it may be involved in control of salt homeostasis in the kidney.

Recently a type of K+ channel with a unitary conductance of ∼40 pS on the basolateral membrane of principal cells in CCD has been shown to be made up of a Kir4.1/5.1 heteromer (425), which may there play a similar role to the channels in the DCT.

In gastric parietal cells, the Kir4.1 homomer is colocalized with a H+-K+-ATPase at the apical membrane (193, 364). This pump secretes protons in exchange for K+. Thus Kir4.1 may be involved in maintaining the activity of the H+-K+-ATPase by providing K+ for the pump, which is similar to the possible role of this channel in functioning Na+-K+-ATPase in renal epithelial cells (see above). In support of this idea, the application of Ba2+ to the parietal cells diminished their proton secretion (193).

II) Cochlea.

The cochlea of the inner ear contains two extracelluar fluids, perilymph and endolymph. The ionic composition of perilymph is almost identical to usual extracellular fluid. On the other hand, the endolymph contains ∼150 mM K+ and possesses a highly positive potential of ∼+80 mV with reference to either blood or perilymph. This potential is referred to as the “endocochlear potential” (260, 812, 813). The unique ionic and voltage environment of the endolymph is essential for proper audition.

The endocochlear potential is thought to be maintained by K+ circulation from perilymph to endolymph through the cochlear lateral wall. The lateral wall is made up of two components, the stria vascularis and the spiral ligament. The application of Ba2+ to the stria vascularis suppressed the endocochlear potential (499). Kir4.1 is the only Kir channel subunit expressed in the stria vascularis (257). It is likely that Kir4.1, which is specifically expressed at the apical membrane of intermediate cells in the stria (14), elicits a large K+ diffusion potential across this membrane and forms a significant fraction of the endocochlear potential (260, 572, 762). In support of this concept, Kir4.1 knockout mice are deaf and exhibit an endocochlear potential of almost 0 mV with an ∼50% reduction of [K+] in the endolymph (500).

In the spiral ligament, fibrocytes that are bathed in perilymph express Kir5.1 homomeric channels (255). The majority of these channels seem to be localized in intracellular compartments. But since perilymphatic perfusion of Ba2+ slightly increased the endocochlear potential, it is possible that a small population of Kir5.1 homomers on the plasma membrane of the ligament could negatively regulate K+ circulation (255, 498).

III) Glial cells.

Brain astrocytes and retinal Müller cells project their processes not only to synapses and soma of neurons but also to blood vessels, pia matter, and vitreous humor (Fig. 16). These astroglial cells play various roles in the control of synaptic functions. One of the most important tasks of astrocytes is to maintain the ionic and osmotic environment in the extracellular space. They conduct a large Kir current involved in transporting K+ from regions of high [K+]o, which results from synaptic excitation, to those of low [K+]o (400, 563). As an excess [K+]o would interfere with normal signaling of neurons by depolarizing them continuously (563, 599), its rapid clearance by astrocytes is essential for the proper function of synapses. This polarized transport is referred to as “spatial buffering of K+” in the brain, and “K+-siphoning” in retinal Müller cells. To achieve these “K+ buffering” functions, astroglial cells use Kir channels locating at specific membrane domains.

Kir4.1 is abundantly expressed in astroglial cells and forms a major part of their basal K+ conductance in brain, spinal cord, and retina (262, 311, 312, 348, 449, 549, 561, 594, 624, 763). Consistently, the knockout or the knockdown of the Kir4.1 gene causes a large increase of input resistance, decrease of K+ conductance, and/or depolarization of Eres in astroglial cells (135, 381, 561, 594). The phenotypes of Kir4.1 knockout mice highlight the significance of Kir4.1 in astroglial K+ buffering. In the retina of mutant mice, the slow PIII response of the light-evoked electroretinogram (ERG) is absent (381). Since this response is thought to be generated by a light-evoked decrease in [K+]o in the distal portion of the retina, it can be attributed to K+ flux from the distal to the proximal end of the retina in response to hyperpolarization of the photoreceptors in which Kir4.1 plays a role. The hippocampus in Kir4.1 knockout mice also shows an impaired K+ uptake in response to neuronal excitation (135), which suggests that Kir4.1 is also involved in K+ buffering in the brain.

Homomeric Kir4.1 and heteromeric Kir4.1/5.1 channels display distinct distribution patterns in astroglial cells of the brain and retina (254, 311). In cortical astrocytes of the brain, both heteromer and homomer channels are located in the perisynaptic processes, whereas only heteromeric channels are expressed at the end feet (Fig. 16A). This suggests that, in cortical astrocytes, K+ is taken up through heteromeric and homomeric channels and excreted via heteromeric channels. On the other hand, the astrocytes in the thalamus and hippocampus, with abundant synapses, express predominantly the Kir4.1 homomer. In the case of the retina, Müller cells harbor Kir4.1/5.1 at their perisynaptic processes and Kir4.1 at their end feet facing vitreous humor and small blood vessels (Fig. 16B). This suggests that the former takes up K+ and the latter secretes K+.

The physiological significance of these differences in different organs, tissues, and regions is not known. However, it may offer unique characteristics to each cell. The major difference between the Kir4.1 homomer and the Kir4.1/5.1 heteromer is their sensitivity to pHi (see sect. vB2). Astrocytes express various ion transport systems including a Na+-HCO3 cotransporter. The activation of this transporter induces intracellular alkalinization and membrane hyperpolarization because it takes up one Na+ and two or three HCO3 (12, 49). Excess extracellular K+ due to synaptic activation accelerates the transporter by depolarizing the astrocytic membrane. The resulting intracellular alkalinization would enhance the activity of heteromeric Kir4.1/5.1 channels and facilitate the uptake of K+. In contrast, astrocytes in the thalamus and hippocampus, which mainly bear the Kir4.1 homomer, express little Na+-HCO3 cotransporter (692). The physiological role of heterogeneity of Kir channels expressed in the perisynaptic processes in brain astrocytes is unclear.

In the perivascular processes of brain astrocytes and retinal Müller cells, Kir channels are colocalized with the water channel AQP4 (254, 311, 548, 549) (Fig. 16). Deletion of the AQP4 protein slows the rate of astrocytic K+ uptake in mice (602). α-Syntrophin knockout mice, which retain Kir4.1 but lack AQP4 in astroglial perivascular processes, show delayed clearance of excess K+ induced by neuronal stimulation (10). These observations reflect a link between K+ and water transport. Kir4.1 and AQP4 are localized in distinct noncaveolar microdomains as well as MβCD-sensitive and -resistant DRMs, which occur in close vicinity (Fig. 17) (259). This apposition of Kir4.1-DRM and AQP4-DRM may underlie the effective cotransport of K+ and water in the astrocytes.

Kir4.1 may also be functionally coupled to glutamate transporters. Kir4.1 occurs together with the glutamate transporters GLT-1 and GLAST in astroglial cells (135, 593). Knockdown and knockout of the Kir4.1 gene resulted in impairment of glutamate uptake by astrocytes (135, 398). Astrocytic glutamate transport, when actively working, depolarizes the membrane. Since functional expression of Kir4.1 would hyperpolarize the membrane, this could facilitate glutamate influx from the extracellular space via the transporter.

In early postnatal stages, the spinal cord expresses Kir4.1 in both gray and white matter (132, 562). Because the Kir4.1 channel is not expressed in neurons, this observation indicates that Kir4.1 occurs in oligodendrocytes as well as in astrocytes. Kir4.1 labeling becomes weaker in white matter and predominantly confined to gray matter around 10–20 days after birth (132, 348), suggesting that the expression of Kir4.1 shifts from both astrocytes and oligodendrocytes to astrocytes during postnatal development. Kir4.1 expression may therefore be necessary for early development of oligodendrocytes (see also sect. vB6 and phenotypes of Kir4.1-null mice, Ref. 562). A possible role of Kir4.1 during development would be to hyperpolarize the membrane and form the driving force for Ca2+ influx via Ca2+-permeable channels as the case in the linkage between Kir2.1 and differentiation of myoblasts (387) (see sect. iiD4). Kir4.1 has been found in oligodendrocytes of adult animals, in the optic nerve and brain where Kir4.1 is detected in the oligodentrocyte cell bodies and rarely in their processes (myelin sheaths) (352, 624). This pattern of distribution implies that Kir4.1 in adult oligodendrocytes is not involved in K+ buffering (see sect. vB6).

As described in section vB3, the Kir4.1 homomer is expressed on the myelin sheath membrane of satellite cells wrapping ganglion neurons. Like in astroglial cells, Kir4.1 in this region is expected to absorb excess extracellular K+ induced by excitation of ganglions (“K+ buffering”) (256). Silencing of Kir4.1 by RNAi technology clarified the physiological role of Kir4.1 in satellite cells of rat trigeminal ganglions (808) where Kir4.1 disruption decreased the nociceptive threshold and led to spontaneous and mechanically evoked facial painlike behavior in freely moving rats. A possible mechanism is that attenuation of K+ buffering by satellite cells increases K+ concentration in the extracellular space of the ganglions and enhances of their excitability.

B) KIR4.2.

A recent study has demonstrated that Kir4.2 is involved in integrin-dependent migration of mouse embryonic fibrocytes and human microvascular endothelial cells (124). This effect seems to depend on the intracellular polyamine content that is controlled by interaction between α9-integrin and spermidine/spermine acetyltransferase (SSAT). The interaction causes catabolism of spermidine and spermine, which enhances K+ efflux through Kir4.2 by reducing inward rectification. The increase of K+ efflux facilitates the formation of the leading edge of migrating cells by an as yet unidentified mechanism. A mutant α9-integrin which lacks the SSAT binding site, overexpression of a nonfunctional mutant of SSAT, and block of Kir4.2 by Ba2+ or siRNA techniques, all suppress integrin-dependent migration of these cells (124).

Kir4.2 has also been identified in mouse liver (609) where its function is unknown.

5. Pharmacology

Ba2+ and Cs+ block Kir channels that contain Kir4.1 and Kir4.2 (388, 763, 765) (see sect. iE). In addition to these nonselective Kir channel blockers, tricyclic antidepressants (TCAs) such as nortriptyline, amitriptyline, desipramine, and imipramine block Kir4.1 channels (Fig. 18) (747). The inhibition of Kir4.1 by nortriptyline depends on the voltage difference from EK, with greater potency against outwardly flowing currents. Selective serotonin reuptake inhibitors (SSRIs), such as fluoxetine, sertraline, and fluvoxamine, also block Kir4.1 current, in a voltage-independent manner (585) (Fig. 18B). Interestingly, these drugs have little effect on the K+ current carried by the neuronal Kir2.1 channel and by Kir1.1 channel (Figs. 18A). The block of astroglial Kir channels by TCAs and SSRIs may contribute to the therapeutic and/or adverse actions of these compounds. Our recent study has further demonstrated that these antidepressants may interact with T128 and E158 residues facing the central cavity of the pore of Kir4.1 (199).

Fig. 18.

Effects of antidepressants on Kir4.1 currents. A, top panel: response of Kir4.1 currents to nortriptyline (100 μM) and Ba2+ (3 mM). Whole cells currents of HEK293T cells transfected with Kir4.1-cDNA were voltage-clamped at EK (−40 mV) and stepped to EK ± 70 mV for 300 ms each. Arrowheads indicate basal current level at EK. Macroscopic current traces during control (a), application of nortriptyline (b), washout (c), and application of Ba2+ (d) are shown. Bottom panel: response of Kir2.1 currents to nortriptyline (100 μM) and Ba2+ (3 mM) recorded under identical conditions. In both experiments, the cells were bathed in extracellular solution which contained 30 mM [K+]. [From Su et al. (747).] B: a comparison of the effects of various antidepressants on Kir4.1 currents. The response to the drugs (100 μM each, except for 30 μM sertraline) was evaluated as the current ratio (IDrug/IControl) where the current was evoked by a voltage step to −110 mV. Extracellular solution contained 30 mM K+. [Data derived from Ohno et al. (585) and Su et al. (747).]

6. Diseases

The phenotypes of Kir4.1 knockout mice suggest that Kir4.1 channels are relevant to a variety of pathophysiological conditions. First, Kir4.1 knockout mice showed loss of the slow PIII response in the light-evoked ERG (381). Second, targeted disruption of Kir4.1, which is abundantly expressed in oligodendrocytes during early postnatal development (348), induced marked motor impairment in mice (562). The cellular basis of this phenotype appears to be hypomyelination in the spinal cord accompanied by severe spongiform vacuolation, axonal swelling, and degeneration (562). Third, Kir4.1 knockout mice are deaf because of a lack of the endocochlear potential and loss of endolymphatic K+ (500). In addition, anion transporter SLC26A4-null mice, which are a model for Pendred syndrome with thyroid goiter and deafness, have severe hearing loss probably due to loss of Kir4.1 channels from the cochlear stria vascularis (822).

It has been recently suggested that abnormality of Kir4.1 function in the brain could induce epilepsy. Examination of the membranes of glial cells obtained from patients with intractable epilepsies sometimes demonstrates a near-complete lack of Kir conductance (58), reduction of the inward rectification (267), and an impaired ability for K+ clearance (200, 331). Two early studies reported that mutations of the Kir4.1 gene might be involved in epilepsy and seizure. The missense variation (T262S) was found to be a candidate for causing the susceptibility to seizure of DBA/2 mice (170). Linkage analysis identified a mutation in the human Kir4.1 gene R271C that might be associated with generalized seizures in humans (72) (Table 2). But equivalent mutations of Kir4.1 expressed alone or with Kir5.1 in a heterologous system elicited K+ currents with properties almost identical to those of wild-type channels (704). A recent genetic analysis identified several mutations in the Kir4.1 gene of patients afflicted by seizures, sensorineural deafness, ataxia, mental retardation, and electrolyte imbalance (SeSAME syndrome) (52, 693) (Table 2). The organs exhibiting the phenotypes (brain, ear, and kidney) correspond to the distribution of Kir4.1. In vitro electrophysiological assay has revealed that some mutations found in the disease, R65P and G77R, result in large reduction of Kir4.1 current (52). R65P is expected to involve the PtdIns(4,5)P2 binding site and therefore cause loss of channel activity (693). Conditional knockout mice that lack Kir4.1 in astroglial cells display severe ataxia, and stress-induced seizures (135) and conventional ablation of the Kir4.1 gene cause deafness in mice (500), which supports the idea of the pathophysiological relationship of Kir4.1 channel impairment with epilepsy and hearing disorders.

A physiological experiment demonstrated that Kir4.1 could be involved in protecting astroglial cells from severe damage during brain injury (132). In the spinal cord, the application of a 30% hypotonic solution causes swelling of the astroglial cell body, whereas no swelling is detected in their end feet. Hypotonic swelling of astrocyte end feet was detected in Kir4.1-knockout mice and also when Ba2+ was applied. The mechanism by which Kir4.1 prevents the swelling of astrocytic process is unknown. Spinal cord injury sometimes causes a hypotonic condition that results in fluid accumulation and thus edema. Also, astroglial Kir channel expression is downregulated after ischemia of either the rat retina (603), which may occur in the CNS as well. Impaired K+ buffering by Kir4.1 during these pathological conditions may be involved in the formation of edema.

Finally, the Kir4.2 gene is located close to the locus of the Down syndrome chromosome region-1 (DCR1) (see sect. vB1), the chromosome 21 trisomy that causes dysmorphic features, hypotonia, and psychomotor delay (214). This suggests a possible linkage between dysfunction of Kir4.2 and the disease, although the amount of Kir4.2 protein expressed in the brain of Down syndrome patients was comparable to that in the normal human brain (169).

C. Kir7.1

1. Historical view and molecular diversity

Kir7.1 was independently identified by three groups (140, 393, 604). Its sequence is quite different from those of other types of Kir channel and shares only ∼38% homology with its closest relative, Kir4.2. Only one isoform has been isolated so far. Kir7.1 is expected to function as a tetramer. No reports support heteromeric assembly of Kir7.1 with other subunits.

2. Pore function and structure

Kir7.1 expressed in heterologous expression systems elicits inwardly rectifying K+ currents with some unique properties. First, the single-channel conductance of Kir7.1 is extremely small, ∼50 fS (393). Second, the sensitivity of the channel to Ba2+ and Cs+ is very low with IC50 values of ∼1 and ∼10 mM, respectively. These IC50 values are ∼10 times greater than those for other types of Kir channel (393). Third, inward rectification of Kir7.1 is independent of [K+]o (140, 393). The residue responsible for these unusual characteristics in the pore region is M125 (Arg in all other Kir subunits). Replacement of this amino acid with the otherwise conserved Arg dramatically increased the single channel conductance of Kir7.1 ∼20 fold and Ba2+ sensitivity ∼10 fold as well as re-establishing a rectification profile similar to other Kir channels (140, 393). The equivalent residue in Kir2.1 (R148) has been reported to be involved in determining single channel conductance and sensitivity to blockers (670). The residue in this position may be important for the pore structure of Kir channels.

Like other Kir channel subunits, Kir7.1 is activated by PtdIns(4,5)P2. However, its binding affinity is weaker than Kir2.1 and other subunits (661). Kir7.1 can also be activated by PtdIns(3,4,5)P3 but not by PtdIns(3,4)P2 (661).

Kir7.1 expressed in native bovine retinal pigmental epithelia (RPE) (see sect. vC4) and in Xenopus oocytes is sensitive to pHi (291, 864). The response is bell-shaped: maximum activity of the channel is observed at pHi 7.0, and the current is attenuated by either acidification or alkalinization (291). Substitution of His at position 26 in the cytoplasmic NH2 terminus with Ala or Arg results in continuous activation of Kir7.1 at alkaline pHi (>pH 8) and in increase of the sensitivity to proton-induced inhibition. Of note, only Kir2.1 has His at the same position where it is involved in PtdIns(4,5)P2 binding (474); H26 in Kir7.1 does not share this character (291).

3. Intracellular localization

Kir7.1 is expressed on the membrane of various types of epithelial cells (see sect. vC4). However, there is little information concerning the trafficking from intracellular organelle to membrane surface. In MDCK cells, truncation of up to 37 amino acid residues from the COOH terminus of Kir7.1 has no effect on its trafficking to the membrane, but further deletions resulted in retention of the channel in the cytosol; residues 309–323 in the COOH terminus may also be involved in membrane targeting (770).

4. Physiological functions in cells and organs

Kir7.1 protein is found in epithelial cells in the choroid plexus (140, 245, 552) and RPE (418, 714). In these cells Kir7.1 is expressed in their apical membrane (418, 552, 714, 864). On the other hand, Kir7.1 immunoreactivity has been detected in the basolateral membrane of thyroid follicular cells and renal epithelia in the distal convoluted tubule, proximal tubule, and collecting duct (129, 552, 596). The physiological functions of Kir7.1 are largely unknown, although their localization in diverse epithelial cells suggests a role in cellular ion transport mechanisms.

The epithelial cells of the choroid plexus and RPE have an unusual polarity. They express the Na+-K+-ATPase on their apical membrane (74, 133, 159, 505, 506, 588). Thyroid follicular cells and renal epithelial cells are more regular and harbor the pump on their basolateral membrane (247, 552). The pattern of Kir7.1 localization follows the Na+-K+-ATPase in each of these epithelial cells, suggesting a function that resembles the “K+-recycling” action of Kir1.1 and Kir4.1-containing channels (see sect. v, A4 and B4).

An irregular pigment pattern is observed in jaguar/ovelix zebrafish due to mutations in the Kir7.1 gene (329). The mutations occur in either the pore region (T128M and L130F) or in the TM2 helix (F168L). All of these mutations result in the suppression of Kir7.1 currents in heterologous expression systems. Kir7.1 is expressed in melanophores. Cells expressing mutated Kir7.1 could not respond correctly to the melanosome dispersion signal derived from neurons.

5. Pharmacology

There are neither specific blockers nor activators for Kir7.1. Kir7.1 is unusually insensitive to block by Ba2+ and Cs+ (see sect. vC2).

6. Diseases

Snowflake vitreoretinal degeneration (SVD) is an autosomal-dominant pathology characterized as a developmental and progressive eye disease that causes fibrillar degeneration of the vitreous humor, early-onset cataract, minute crystalline deposits in the neurosensory retina, and retinal detachment (269, 657). Genomic analysis of SVD patients revealed a heterozygous mutation (484C > T, R162W) in the Kir7.1 gene, KCNJ13 (250) (Table 2). Modeling of Kir7.1 based on the structure of KirBac1.1 suggested that R162 was localized in the short polypeptide chain between the TM2 α-helix and the COOH terminus composed of β-sheets and which may be involved in channel activation by PtdIns(4,5)P2 (see sect. iD2). The R162W mutant Kir7.1 channel carries a nonselective cation current that depolarized transfected cells and increased their fragility. Functionally, this is the equivalent of the weaver mutation in Kir3.2 (see sect. iii, B5 and F2). Since Kir7.1 is diffusely expressed in various regions of the retina, including the retinal pigmental epithelia and the internal limiting membrane (250, 418), the degeneration of cells in these areas may cause the abnormalities observed in the retina of SVD patients. The relevance of this mutation to disorders of the vitreous humor and cornea remains to be determined.

VI. Conclusion

We have come far, from the identification of an “anomalous” K+ current to the three-dimensional structure of the KATP channel heterooligomer in the plasma membrane. Classical voltage-clamp and then single-cell electrophysiological techniques described Kir channels and suggested roles in the physiological function of a variety of tissues. The development and application of molecular biology techniques isolated 15 Kir channel subunits and provided insights into channel assembly, function, and structure. In particular, Kir channels were found to be made up of homomeric and heteromeric assembly of the subunits which multiplied ion channel characteristics. This divergence expanded the roles of Kir channels in tissues and organs. Biochemical assays identified interactions between Kir subunits and anchoring proteins including PDZ-proteins that are crucial for subcellular localization. Kir channels have been found to be dynamically controlled by posttranslational modification including phosphorylation. The identification of the crucial involvement of PtdIns(4,5)P2 in contributing to the activity of Kir channels has clarified a number of points underlying their functionality and physiological regulation. Gene-targeting techniques and the human genome project have greatly contributed to understanding the contributions of Kir channels to organ physiology and disease. Phenotypes of Kir knockout animals and the identification of loss-of-function and gain-of-function mutations in Kir genes not only reveal the functional significance of Kir channels in cells and tissues but also disclose their pathophysiological relevance. The evidence accumulated by these assays ensures that Kir channels are key elements that set Eres, control cell excitability, regulate hormone release, and drive epithelial transport. The last 10 years have seen enormous advances in structural biology that have clarified not only structure-based mechanisms underlying Kir channel activity but also the relationships between channel architecture, binding of diverse small substances, and regulation of the channel's function. These experimental approaches are now being effectively combined to advance understanding of Kir channels.


H. Hibino and Y. Kurachi are supported by the following research grants and funds: Grant-in-Aid for Scientific Research A20249012 (to Y. Kurachi), Scientific Research on Priority Areas 17081012 (to H. Hibino), the Global COE Program “in silico medicine” at Osaka University (to H. Hibino and Y. Kurachi), and a grant for “Research and Development of Next-Generation Integrated Life Simulation Software” (to Y. Kurachi) from the Ministry of Education, Culture, Sport, Science, and Technology of Japan.


Address for reprint requests and other correspondence: Y. Kurachi, Dept. of Pharmacology, Graduate School of Medicine, Osaka University, 2–2 Yamada-oka, Suita, Osaka 565-0871, Japan (e-mail: ykurachi{at}