K
ATP channels are heterooctameric complexes of pore-forming Kir channel subunits together with members of the ATP binding cassette (ABC) family of membrane proteins (). Two genes,
KCNJ8 (Kir6.1) and
KCNJ11 (Kir6.2) encode K
ATP pore-forming subunits (
167,
168), while two SUR genes,
ABCC8
(SUR1) and ABCC9 (SUR2), generate regulatory subunits (
2,
56,
168). Alternative RNA splicing gives rise to several SUR protein variants (e.g. SUR2A and SUR2B) that confer distinct physiological and pharmacological properties on the channel complex (
53,
365). Interestingly (), the genes for Kir6.2 and SUR1 are located next to each other on human chromosome 11p15.1 (
168) suggesting an as yet unconsidered co-regulation at the gene level. In addition, the genes for Kir6.1 and SUR2 are also adjacent to one another on chromosome 12p12.1 (
56,
170), implicating an evolutionary duplication event. In order to recapitulate K
ATP channel activity in a heterologous expression system, both Kir6 and SUR subunits must be co-expressed (
168), and biophysical and biochemical studies show that Kir6.2 and SUR1 subunits combine in a 4:4 stoichiometry () to generate the functional K
ATP channel (
60,
169,
373). Similarly, biochemical studies demonstrate that the SUR2 protein variants, SUR2A and SUR2B, can also coassemble with Kir6 subunits (
20,
167,
311,
442), presumably in a similar octameric arrangement.
The Kir6 subunit Kir6.1 and Kir6.2 are typical inward rectifier channel proteins consisting of two transmembrane helical domains, TM1 and TM2, and cytoplasmic N- and C-termini (). The two transmembrane helices and pore domain of Kir6.1 and Kir6.2 () share significant structural similarity to the pore-forming S5–S6 membrane segment of voltage-gated K
+ channels. Both Kir subunits contain a highly conserved sequence of residues called the K
+ channel signature sequence (TVGY/FG) (
152), that confers K
+ selectivity and is found throughout the K
+ channel family. While the same general pore structure is predicted for both Kir6.1 and Kir6.2, the two subunits exhibit distinct single channel conductances of ~35 and ~80 pS, respectively in 150 mM K
+ solutions (
167,
168,
178,
216,
442), conferred by specific residues in the TM1-TM2 regions (
341).
The signature property of K
ATP is inhibition by intracellular ATP with ATP-binding occurring on the Kir6 subunit (
92,
239,
374,
410,
411). Heterologously expressed K
ATP channels comprised of Kir6.2 plus SUR1 or SUR2A close in the presence of ATP with an average K
1/2 for inhibition of ~10 μM (
298). Although initially reported to be insensitive to ATP (
442), Kir6.1-containing channels have since been shown to exhibit ATP inhibition comparable to Kir6.2 (
18,
356). Both non-hydrolyzable ATP analogues and ATP inhibit K
ATP channels, confirming that inhibition is not a consequence of phosphorylation, but of direct binding to the channel (
110,
197,
237). In the absence of Mg
2+, ADP also blocks channel activity (
197,
237,
305), albeit with lower affinity, suggesting that electrostatic interactions with the phosphate moieties on the ATP molecule, particulary the γ-phosphate, are important for binding (
187).
Following the publication of the structure of the tetrameric cytoplasmic domains of eukaryotic Kir3.1 (
304), and of the full length bacterial homolog KirBac1.1 (
231), modeling of the tetrameric Kir6 pore revealed that three-dimensional folding brings several residues in the cytoplasmic N-and C-termini together (R50, R201, G334, II82) to form a binding pocket for ATP () (
8,
103,
188,
342,
406), for a total of four binding sites per channel (i.e. each Kir subunit in the tetrameric pore binds one molecule of ATP) (
8,
103,
136,
406). In addition, the N-terminus contains an amphipathic “interfacial” or “slide“ helix, predicted to lie parallel to the inner surface of the lipid membrane () (
102). This “slide” helix is physically coupled to the cytoplasmic end of the pore and may provide the physical link between ATP binding and the gate of the channel, which is likely to be located at the crossing point of the inner helix bundle (
230,
231,
321,
343). More detailed homology models of Kir6.2 continue to follow the appearance of more refined bacterial channel structures (KirBac3.1) and eukaryotic channel cytoplasmic domains (Kir3.1, Kir2.1) (
230,
231,
273,
304,
321). Mikhailov et al. have successfully purified a Kir6.2/SUR1 complex and provided a low-resolution structure using single-particle electron microscopy (
273), but direct crystallographic analysis of the K
ATP channel pore structure has thus far been unattainable.
The SUR subunit In the absence of Mg
2+, nucleotides inhibit K
ATP channel activity, but in the presence of Mg
2+, both ATP and ADP stimulate channel activity (
95,
160,
196,
237), through interaction with the SUR subunit (
132,
303,
372). The unifying structural feature of all ABC proteins are the cytoplasmic nucleotide binding folds, NBF1 and NBF2, comprised of Walker A and Walker B nucleotide-binding motifs and other conserved sequences (
154) (). The NBF domains contain the binding sites for Mg
2+-adenosine nucleotides that serve to stimulate K
ATP function (
48,
264). In the ABC superfamily, the transmembrane domains (TMD1,2) each comprised of 6 transmembrane helices typically carry out transmembrane transport function. However, SUR1 and SUR2 are atypical in that no such transport function has been identified, and these subunits contain an additional transmembrane domain at the N-terminus, TMD0, that is joined to TMD1 by a cytoplasmic linker L0 (). Physical interaction between the cytoplasmic N-terminus of Kir6.2 and the TMD0 domain of SUR, including the cytoplasmic L0 linker, is demonstrated to be important in regulation of Kir6.2 gating by SUR (
19).
In bacterial ABC proteins, dimerization of NBFs is favoured by the presence of nucleotides, and ATP hydrolysis requires co-operative interactions of the two NBFs (
233,
380), leading to the hypothesis that the nucleotide-bound dimer is the catalytically active species. By analogy, it is suggested that Mg
2+-dependent ATP hydrolysis at the dimeric SUR NBFs provides the ‘power stroke’ that overcomes the inhibitory effect of ATP on Kir6. The Mg
2+ dependence of α
32P- and γ
32P-azido-ATP labelling of SUR NBFs suggests that SUR NBFs also hydrolyse ATP (
266,
412). Trapping of channels in the activated state by vanadate (a transition-state analogue that mimics a post-hydrolytic, ADP-bound state) and in an inactivated state by beryllium fluoride (a transition-state analogue that mimics the pre-hydrolytic, ATP-bound state) (
460) further support the hypothesis that ATP hydrolysis underlies channel activation in intact cells, where ATP concentrations will always be at millimolar levels, even when metabolism is substantially inhibited (
98). However, direct measurements of nucleotide hydrolysis by SUR NBFs are sparse (
37,
263,
273,
460), and SUR-NBF mutations analogous to those that have drastic effects on ATP hydrolysis in bacterial NBFs have only modest effects on SUR-mediated hydrolysis (
37). In membrane patch-clamp experiments, channel activation by exogenous MgADP is much more effective than activation by MgATP, and the same ‘hydrolysis’ mutations lead to decreased MgADP stimulation of K
ATP channels (
37,
132,
372,
462). Why MgADP should so strongly activate the channel in excised patches if a hydrolytic cycle is required for activation is unclear. This would require that the ‘activated’ state resulting from hydrolysis persists after MgADP dissociation, at least long enough for rebinding of exogenous MgADP to maintain the state. Just how reasonable this notion is has not been addressed and it remains conceivable that hydrolysis, if it occurs, is an epiphenomenon and that preferential binding of MgADP at the second nucleotide binding site is actually the physiological activator of the channel.
Additional subunits of KATP and macromolecular complexes In addition to different gene products, SUR splicing may also be important in determining channel function (
53,
142,
350,
365). Expression of SUR1 splice variants with deletions of TM16/17 (
133), in NBF1 (
142) and in NBF2 (
350) have been reported in the heart. Alternative splicing of SUR2 exons 14 and 17, leading to deletion of segments of NBF1 are also expressed in the heart (
53). The functional significance of these variants is not known, although channels containing SUR2A Δexon17 are reportedly less sensitive to ATP inhibition (
53). In addition, the recent identification of short-form or partial SUR2 gene products (
328) has arisen from studies of the SUR2
−/− mouse. The targeting construct used in the generation of these animals contained a replacement of exons 12–16, coding for regions of NBF1 (
55). Naturally transcribed short-form SUR2 subunits lacking the NBF1 domain have been identified using antibodies raised against distal parts of the protein (
329,
388) and are reported to coassemble with Kir6.1 and Kir6.2 to form glibenclamide-insensitive, ATP-sensitive, currents in ventricular myocytes (
328). A further recent study provides evidence that these short forms are present in mitochondria, and therefore may actually be a component of ‘mitoK
ATP’ (
449).
While the coassembly of Kir6x and SURx are sufficient to reiterate hallmark properties of K
ATP channels, additional proteins may also contribute to a macromolecular channel complex and fine-tune channel function. Recent studies have implicated an elaborate β-cell K
ATP channel macromolecular structure (
318,
368), and there is mounting evidence that metabolic enzymes, including adenylate kinase (AK) (
49), creatine kinase (CK) (
73) and lactate dehydrogenase (LDH) (
71), can physically associate with, and regulate, the cardiac K
ATP channel complex. Formation of a multi-protein complex () that includes both phosphotransfer and glycolytic enzymes (
89,
191) may explain early studies of the native sarcolemmal channel, in which direct application of glycolytic substrates to excised membrane patches inhibited K
ATP channel activity (
426,
427). Recent studies further demonstrate the importance of a local phosphotransfer network in regulating K
ATP channel activity (
363): by amplifying small changes in cytoplasmic ATP concentration, adenylate kinase and creatine kinase may play an integral role in regulating the nucleotide concentration in the localized membrane environment (), thereby regulating the K
ATP response to metabolic events. In the cell, glycolysis can oscillate periodically, driven by feedback loops in regulation of key glycolytic enzymes by free ADP and other metabolites (
155,
309,
444). Several studies have shown that when the capacity to buffer cellular ATP and ADP levels is suppressed by metabolic inhibition, oscillations in glycolysis can cause concurrent oscillations in ventricular action potential duration, due to oscillatory activation of K
ATP (
9,
309,
310,
444), which may promote arrhythmias during acute metabolic stresses, such as myocardial ischemia.
Modulation by cellular lipids, metabolites and protein kinases Nucleotide regulation is the signature property of K
ATP, although other regulatory ligands modulate channel activity through direct interaction with the Kir6 subunit, the SUR subunit or both. Membrane phospholipids, phosphoinositides, in particular (e.g. phosphatidylinosital-4,5-bisphosphate (PIP
2), potently stimulate K
ATP activity by binding the Kir6.2 subunit (
29,
107,
344,
375,
439). Application of PIP
2 directly to the intracellular side of excised membrane patches leads to significant increase in K
ATP channel activity and restores channel activity after channel ‘run-down’ (
107,
156). Conversely, treatment of a patch with phospholipase C (PLC), to hydrolyze PIP
2, reduces channel activity (
107,
156). Similarly, receptor–mediated activation of phospholipase C (PLC) modulates K
ATP activity in recombinant cell systems where receptor and channel proteins are overexpressed (
29,
438), although direct support for a dynamic regulation in native muscle cells is lacking. Importantly, there exists a negative coupling between PIP
2 activation and ATP sensitivity of K
ATP channels such that as PIP
2 increases, channel open probability increases and ATP-sensitivity decreases (
103). Physiologically, this means that ATP sensitivity is not a fixed parameter, and may change dynamically with changes in membrane composition. Residues involved in PIP
2 binding and activation overlap with the ATP binding site on Kir6.2, consistent with their competitive effects observed in binding assays (
103,
258,
344). The activation of channels by PIP
2 requires the presence of both the headgroup and the phospholipid tail, since the cleavage products of PIP
2 hydrolysis (IP
3 and di-acyl glycerol), have little effect on channel activity. The negative charge of the head group is also critical, since phosphatidyl serine (net negative charge) can activate K
ATP but phosphatidyl choline (net positive charge) cannot (
107). Consistent with an electrostatic interaction between the negatively charged head group and a region of the Kir6.2 protein close to the membrane, a number of positively charged amino acid residues in the slide helix region directly interact with PIP
2 in the membrane (
79,
80,
358,
374). While PIP
2 interaction with Kir6.1 has been less extensively studied, Kir6.1 channels are activated by PIP
2 (
335), suggesting conservation of the fundamental determinants of phospholipid binding and gating (
164,
347,
455).
Long chain acyl-coA molecules (LC-CoA), intermediates of β-oxidation of fatty acids, have also been shown to modulate K
ATP channel activity in an analogous manner to membrane phosphoinositides (
249). The same residues on Kir6.2 that mediate PIP
2 activation, also contribute to the stimulatory effect of LC-CoA (
262,
358), suggesting that Kir6.2 is the major site of LC-CoA action. The effect of LC-CoA on ATP sensitivity is greater in ventricular myocytes than in pancreatic β-cells (
43,
44,
117,
235), and the degree of acyl-coA activation increases with chain length for cardiac ventricular channels, while in pancreatic β-cells medium chain fatty acyl moieties seem to be less effective. PIP
2 and lipid modulation of ATP sensitivity is a very powerful modulator of channel activity, and changes of lipids in pathophysiological conditions may be an important regulator. Since cardiac ventricular K
ATP is predominantly SUR2A-dependent, whereas the pancreatic K
ATP is predominantly SUR1-dependent (see below), the distinct effects of acyl-coA on cardiac ventricular and pancreatic channels suggest that there is also a role for differential SUR involvement in determining the activation properties of the lipid metabolites, although this needs further work and the implications remain unclear.
Acidic intracellular pH stimulates K
ATP channel activity (
87,
88,
440), and intracellular acidification concomitant with anaerobic metabolism provides another potential physiological stimulus of K
ATP activity. The mechanistic basis of pH regulation is still not completely established, although there is a general consensus that protons act to decrease sensitivity to inhibitory ATP. Mutagenesis studies of recombinant Kir6.2 subunit have revealed two amino acids (Thr71 and His175) that appear to be critical determinants of pH sensing (
76,
437), but it remains unclear whether these residues are directly protonated and how they regulate the effect of pH on ATP-dependent gating.
Agonist-dependent, protein kinase A (PKA) phosphorylation is an important regulator of smooth muscle K
ATP channels and of pancreatic K
ATP, although the importance of channel phosphorylation is not entirely clear in striated muscle. The Kir6.2 subunit has two consensus PKA phosphorylation sites (
33) which when phosphorylated increase channel open probability (
33,
247). In human SUR1, a unique, constitutively phosphorylated PKA site acts to both increase the surface expression of the channel and decrease channel open probability, although this residue is not conserved in rodent SUR1 (
33). The specific effects of PKA on Kir6.2/SUR2A combinations have not been examined, and it is unknown whether the potential PKA sites of SUR2A are phosphorylated and what effect, if any, this might have on channel activity in striated muscle, where these subunits are predominant. In contrast, the physiological role of PKA-dependent activation of K
ATP channels in vascular smooth muscle is now established, and phosphorylation of both Kir6.1 and SUR2B appear to be critical in regulating channel activity, and hence vascular tone (see section D).
Protein kinase C has mixed actions on native ventricular K
ATP channels, inhibiting at low micromolar ATP concentrations (
242), but activating at high ATP concentrations (
246), through phosphorylation of highly conserved T180 residue in the Kir6.2 subunit (
243). Chronic PKC activation also stimulates the retrieval of Kir6.2 from the surface membrane through a dynamin-dependent mechanism (
162). Kir6.1/SUR2B channels are inhibited (
335) by acute PKC treatment, due to phosphorylation of residues in Kir6.1 (see section D), whereas Kir6.2/SUR2B channel activity is reportedly unaffected by PKC (
366,
402), highlighting the specificity of K
ATP channel subunit combinations in physiological regulation.
Structural basis of KATP pharmacology The SUR subunit also determines the sensitivity of the channel to a huge range of pharmacological K
ATP channel openers (KCOs) and blockers (
2,
17,
21,
22,
37,
82,
139,
282,
292,
413). All SUR isoforms are inhibited by the sulfonylurea glibenclamide, with sulfonylurea inhibition being dependent on nucleotide concentrations, and hence on metabolic state (
223,
339). SUR1 is typically more sensitive to sulfonylureas than SUR2 (
131), a difference that has been exploited with SUR1/2 chimeric constructs to implicate a binding site for tolbutamide in TMD2 of SUR1 (
17,
22). Mutation of residue S1237 in SUR1 to tyrosine, the analogous residue in SUR2, reduces the affinity for tolbutamide and gliblenclamide (
17), while the reverse mutation (Y1206S in SUR2) increases affinity (
141). The SUR subunits confer also sensitivity to KCO’s such as diazoxide, cromakalim, and pinacidil (
298). Diazoxide is an effective activator of SUR1 and SUR2B, but not SUR2A, whereas pinacidil and cromakalim are effective activators of SUR2A and SUR2B, but not SUR1 (
112,
333), highlighting the potential for tissue specificity, which remains an important pharmaceutical strategy (
15). The channel openers typically contain one or more nitrogen-substituted aromatic rings, and include derivatives of cyanogaunidines (e.g. pinacidil), benzopyrans (e.g. cromakalim), but also many others, and a common pharmacophore has proven elusive (
61,
261). Regions important for KCO action are spread throughout the SUR subunits, and include TMD1 and NBD1 for diazoxide (
21), and TMD2 for cromakalim and pinacidil (
140,
282,
413). These openers all require the presence of hydrolysable ATP (
90,
140,
359), which suggests that they act to stabilize or enhance ATP hydrolysis at the NBFs.