Personalized Therapeutics for K_ATP-Dependent Pathologies

Abstract

Ubiquitously expressed throughout the body, ATP-sensitive potassium (K_ATP) channels couple cellular metabolism to electrical activity in multiple tissues; their unique assembly as four Kir6 pore-forming subunits and four sulfonylurea receptor (SUR) subunits has resulted in a large armory of selective channel opener and inhibitor drugs. The spectrum of monogenic pathologies that result from gain- or loss-of-function mutations in these channels, and the potential for therapeutic correction of these pathologies, is now clear. However, while available drugs can be effective treatments for specific pathologies, cross-reactivity with the other Kir6 or SUR subfamily members can result in drug-induced versions of each pathology and may limit therapeutic usefulness. This review discusses the background to K_ATP channel physiology, pathology, and pharmacology and consider the potential for more specific or effective therapeutic agents.

K_ATP CHANNEL STRUCTURE AND PHYSIOLOGICAL REGULATION BY ADENINE NUCLEOTIDES

Present in multiple tissues throughout the body, ATP-sensitive potassium (K_ATP) channels are hetero-octameric complexes of four pore-forming inward rectifier potassium channel subunits (Kir6.1 or Kir6.2, encoded by KCNJ8 and KCNJ11, respectively) and four regulatory sulfonylurea receptor (SUR1 or SUR2, encoded by ABCC8 and ABCC9, respectively) subunits (1) (Figure 1). Alternative splicing of the 3′ exons of ABCC9 gives rise to distinct SUR2A and SUR2B subunits that differ in the carboxyl terminal 42 amino acids. Intriguingly, KCNJ8 is immediately downstream of ABCC9 on human chromosome 12, and KCNJ11 is immediately downstream of ABCC8 on chromosome 11. This indicates that one pair of genes is a duplication of the other and hints at each pair of genes being co-regulated.

Figure 1.

Molecular regulation of ATP-sensitive potassium (K_ATP) channels (a) K_ATP channels are octameric complexes in which four pore-forming inward rectifier potassium channel (Kir6) subunits generate the channel pore and four sulfonylurea receptor (SUR) subunits serve the regulatory role. Each Kir6 subunit consists of two transmembrane helices (M1, M2) and a reentrant pore loop that forms the K selectivity filter. Each SUR subunit consists of three major domains, the unique 5 helix TM0 and the ATP-binding cassette (ABC) core 6 helix TM1 and TM2 domains, each of which are followed by linker between TM0 and TM1 (L0), nucleotide-binding fold 1 (NBF1), and NBF2, respectively. (b) The major physiological regulation is via a gate at the cytoplasmic end of the inner cavity. ATP binding to the Kir6 cytoplasmic region provides the energetic push to close channels. Magnesium-bound ATP (MgATP) binding at the ATP-binding site (ABS1) formed at the NBF1-NBF2 interface, together with ATP hydrolysis or magnesium-bound ADP (MgADP) binding at ABS2, results in a conformational activated state that is transduced to override ATP inhibition. Phosphatidyl inositol 4,5 bisphosphate (PIP₂) interaction at a site near the ATP inhibitory site also provides an energetic pull to open channels. Pharmacologically, sulfonylureas (SUs) or K channel openers (KCOs), interacting with sites at the TM1/2 interface, respectively, promote channel closure or opening. (c) Model of the Kir6.2/SUR1 channel constructed from a cryo-electron microscopy density map, viewed from the side (left) and from the extracellular side (right). Key features are indicated, including transmembrane domains (TMDs), intracellular NBFs [referred to as nucleotide-binding domains (NBDs)], and the C-terminal domain (CTD). Panel c adapted with permission from Reference 2.

As in all Kir channel subunits, Kir6 subunits comprise two membrane-spanning domains flanking a re-entrant pore loop that generates the K⁺-selective pore and a large cytoplasmic domain formed from the intracellular amino and carboxyl termini. SUR1 and SUR2 are members of the ATP-binding cassette (ABC) superfamily (subfamily C) of membrane proteins, with each SUR subunit containing 17 transmembrane domains and two intracellular nucleotide-binding folds (NBF1, NBF2) or domains (NBD1, NBD2) (Figure 1) that come together to form two adenine nucleotide-binding sites (ABS1, ABS2) at the interface between them. Recent cryo-electron microscopy (cryo-EM) structures of the pancreatic (Kir6.2/SUR1) K_ATP complex (2–5) confirm overall octameric assembly of the channel (6) and demonstrate the regulatory nucleotide-binding sites on Kir6.2 and SUR1 (Figure 1c). Physiologically, K_ATP channels are inhibited by nonhydrolytic binding of ATP at sites on the Kir6 subunits. By incorporating mutations in Kir6.2 that disrupt the ATP-binding site, enhance PIP₂ activation, or stabilize the channel open state, potentially activated structures of the Kir6.2/SUR1 complex have recently been obtained (7, 8). In one such structure (7), the ATP-binding site is distorted, which may help to explain why ATP preferentially binds to the closed state, underlying the ATP inhibitory effect (9). K_ATP channels are also inactivated by magnesium-bound ADP and ATP (MgADP and MgATP) interactions with the NBDs of SUR1 (1) (Figure 1b). In each subunit, the interface between transmembrane domain 1 (TMD1) and TMD2 is open to the cytoplasm, and the two NBDs are separated in the absence of Mg nucleotides (2). The NBDs come together to form two adenine nucleotide-binding sites (ABSs) at the dimer interface in the presence of Mg nucleotides, with MgATP being bound at a closed, catalytically inactive ABS1 site and MgADP being bound in the potentially catalytically competent ABS2 site (4) (Figure 1b). Rigid body rotations of SUR1 TMD1 and TMD2 underlie distinct propeller and quatrefoil conformations of the channel (4). The latter disrupts the interaction between Kir6.2 and the linker between TM0 and TM1 (L0) that underlies increased ATP sensitivity of Kir6 subunits when coexpressed with SURs (10). Very recent cryo-EM structures of Kir6.1/SUR2B channels, determined in the presence of ATP and glibenclamide, also include distinct propeller and quatrefoil conformations (11), showing that both can be obtained in both NDB-dimerized (i.e., activatory) and glibenclamide-bound (i.e., inhibitory) conditions, leaving the functional relevance of the two conformations unclear.

Finally, phosphatidyl inositol 4,5 bisphosphate (PIP₂) is a ubiquitous activator of Kir channels, binding to sites that overlap and functionally compete with inhibitory ATP binding on Kir6.2 (9). Intriguingly, the Kir6.1 cytoplasmic domain is translocated downward compared to Kir6.2 structures, similar to that in apo-Kir2.1 channel structures (3JYC), in which the PIP₂-binding site is disrupted. The functional relevance is not yet clear but may be consistent with the apparently minimal effects of PIP₂ depletion on SUR2B/Kir6.1 activity (12), which suggests that Kir6.1 channels may not require PIP₂ for activity.

MONOGENIC K_ATP DISEASES

Functional K_ATP channels can be reconstituted by expressing any combination of Kir6.1 or Kir6.2 with SUR1, SUR2A, or SUR2B (13, 14), and more than one SUR or Kir6 isoform can coexist in a single channel (15–17). Native K_ATP channel composition could thus vary continuously across tissues (18), and each subunit gene could, in principle, have its own associated pathologic syndrome. However, as the monogenic pathologies associated with each subunit have become clear, it has also become clear that these are tightly matched. Despite the fact that certain tissues appear to express noncanonical pairings, notably striated cardiac and skeletal muscle (in which Kir6.2 and SUR2 predominate), the monogenic Kir6.2 and SUR1 pathologies are largely indistinguishable from one another, and thus far the same appears to be the case for Kir6.1 and SUR2 pathologies.

Kir6.2 and SUR1 Loss-of-Function Mutations Cause Congenital Hyperinsulinism

Loss-of-function (LOF) mutations in KCNJ11 and ABCC8 were first causally associated with congenital hyperinsulinism (CHI; MIM 256450) over 25 years ago (19–21). CHI is typically identified in the neonatal period by insulin hypersecretion despite low blood glucose levels that, in the absence of treatment, can lead to severe mental retardation and epilepsy. CHI may be morphologically diffuse or focal: In the diffuse form, mutations can be recessively or dominantly inherited (22, 23), and all pancreatic β cells are affected, whereas in the focal form, the affected cell population is in a localized region of the pancreas (24). The focal form is typically a result of inheriting one copy of the mutated, inactive gene from the unaffected father. During embryonic development, a mutation occurs in the other, active copy of the gene and is then found within only a subset of the pancreatic islets. Many recessive ABCC8 mutations are intronic. These lead to splicing defects or exon skipping (25). Small insertions and deletions and a large number of missense mutations typically either (a) reduce channel biosynthesis, trafficking, or assembly, or (b) decrease intrinsic channel activity (26). Many of the latter localize to the NBFs and result in a reduced stimulatory effect of MgADP (26). In a subset of NBF mutations, residual channel activity is still present due to a partial MgADP response (23), and, as a consequence, the associated hyperinsulinism phenotype is generally milder than that associated with complete loss of function (22) and can be treated by K_ATP channel-opening (KCO) drugs that increase the residual K_ATP activity (see below).

Kir6.2 and SUR1 Loss-of-Function Mutations Cause Neonatal Diabetes and DEND Syndrome

Conversely, more than 70 gain-of-function (GOF) mutations in KCNJ11 and ABCC8 have now been associated with the corollary disorder, neonatal diabetes mellitus (NDM; MIM 606176) (27–30). Typically diagnosed soon after birth, with low birth weight, marked hyperglycemia, and ketoacidosis, NDM is a rare disorder with incidence of up to approximately 1 in 90,000 live births (31). It can be permanent (PNDM), requiring antidiabetic treatment for life, or a milder transient form (TNDM) in which hyperglycemia usually resolves within the first two years but can later relapse. The exact mechanisms underlying TNDM remission are unclear but may reflect a secondary increase in insulin sensitivity, compensatory increase in β cell function, or both (see below). In the most extreme K_ATP-dependent PNDM cases, patients also exhibit developmental delay in motor and intellectual skills, epilepsy, and neonatal diabetes (DEND) syndrome or intermediate DEND without epilepsy (32, 33), reflecting K_ATP overactivity outside the pancreas (34).

In the pancreas, elevated glucose raises the [ATP]:[ADP] ratio, normally closing Kir6.2/SUR1-dependent K_ATP channels and triggering action potential firing and insulin secretion. Overactive K_ATP channels suppress electrical activity, and as a result, insulin secretion is inhibited and glucose rises (35–37). Kir6.2 and SUR1 are also widely expressed in the hippocampus and cerebellum as well as the midbrain and brainstem (38). Transgenic mice overexpressing the Kir6.2[V59M] DEND mutation in either muscle or nerve indicate that motor impairments originate in the central nervous system (39). Hippocampus-specific K_ATP-GOF transgenic mice exhibit mostly learning and memory deficiencies (40), pointing to a specific hippocampal involvement, but new models and approaches will be needed to fully define the origins of nonpancreatic DEND features.

All functionally assessed mutations result in enhanced channel activity at any given [ATP]:[ADP] ratio by either reducing ATP inhibition (41–43) or increasing sensitivity to stimulatory Mg²⁺ adenosine (28, 42, 44). Residues located in or near the inhibitory sites on Kir6.2 reduce ATP binding directly (2). Mutations outside the ATP-binding site, or in the SUR1 TMD0 domain, increase channel open state stability, allosterically reducing ATP sensitivity (41, 45). Since ATP interacts predominantly with the closed state of the channel (9), access to the binding site is reduced when the open state is favored, and as a result, ATP sensitivity is decreased. Regardless of the molecular mechanism, disease severity generally correlates with the magnitude of channel overactivity (42, 45, 46). The association of DEND with more severe activating mutations suggests that extrapancreatic tissues are less sensitive to changes in K_ATP activity than the pancreatic β cell and that a threshold of channel overactivity must be reached before neuronal firing rates are suppressed and neurological features become evident.

Interestingly, phenotypic heterogeneity has been reported within families carrying a single NDM-causing K_ATP mutation. In a pedigree with the activating Kir6.2[C42R] mutation, cross-generational phenotypes of affected family members ranged from PNDM and gestational diabetes to late-onset type 2 diabetes mellitus (T2DM) (47). Similarly, in several pedigrees with SUR1 mutations, carriers were variously unaffected, type 2 diabetic, or neonatally diabetic (28, 48, 49). Such findings suggest that the severity of the β cell defect in K_ATP-induced NDM can be modulated by additional, unknown genetic and/or environmental factors that integrate to determine the clinical presentation.

Kir6.1 and SUR2 Gain-of-Function Mutations Cause Cantú Syndrome

First described as a distinct syndrome in 1982 (50), Cantú syndrome (CS; MIM 239850) is a multiorgan disorder characterized by congenital hypertrichosis, distinctive facial features, osteochondrodysplasia, multiple cardiovascular features (51), and potentially lethal pulmonary hypertension and bronchopulmonary dysplasia (52). Genetically associated with ABCC9 in 2012 (53, 54) and with KCNJ8 in 2014 (55, 56), over 30 mutations have since been identified in more than 100 patients (51, 52, 57–62). All assessed mutations lead to recombinant K_ATP channel GOF (56, 63, 64). Similar to NDM mutations in SUR1 and Kir6.2, the underlying mechanisms include decreased ATP inhibition and enhanced MgADP activation (63–65).

The recognition that essentially identical CS features are seen with KCNJ8 or ABCC9 mutations (56, 64) definitively ties the syndrome to Kir6.1/SUR2 channel overactivity but does not immediately explain the pathophysiologic consequences. Smooth muscle (SM) K_ATP is predominantly formed of SUR2B/Kir6.1 subunits (66), and recent animal experiments have led to a mechanistic framework for understanding many CS features as a consequence of SM dysfunction. Hypotension, vascular dilation, and twisted blood vessels, as well as cardiomegaly, are reiterated in Cantú mice, in which CS-associated SUR2[A478V] and Kir6.1[V65M] mutations were introduced to the equivalent mouse loci (67). Enhanced basal K_ATP conductance in vascular smooth muscle (VSM) provides a ready explanation for vasodilation and hence reduced systemic vascular resistance (SVR), leading to lower blood pressure (BP).

Striated muscle K_ATP channels may have various subunit combinations, with the noncanonical SUR2A/Kir6.2 being prominent in both heart and skeletal muscle. Consistent with cardiac channels being formed of SUR2 and Kir6.2 (but not Kir6.1), the ATP sensitivity of the ventricular myocyte K_ATP channel is reduced in SUR2[A478V] but normal in Kir6.1[V65M] mice. However, cardiac enlargement is even greater in the latter (67) and must therefore arise as a secondary consequence of K_ATP GOF in other cells. Decreased SVR in these animals is accompanied by elevated renin-angiotensin signaling (RAS), and cardiac enlargement can be reversed by blockade of RAS signaling (68) or by dominant-negative suppression of SM K_ATP (69), indicating that cardiac remodeling is a secondary consequence of the vascular GOF. It is accompanied by expression of genes associated with pathological hypertrophy and by decreased exercise tolerance, suggestive of reduced cardiac reserve (68). Longitudinal studies of CS patients (G.K. Singh, C. McClenaghan, M. Aggarwal, H. Gu, M.S. Remedi, et al., unpublished data) suggest that the long-term consequence of this reduced reserve is the gradual development of heart failure, a potentially important therapeutic issue.

Muscular fatigue is a common experience in CS patients (51). Consistent with skeletal muscle K_ATP being primarily composed of Kir6.2/SUR1, there is also a K_ATP channel GOF in flexor digitorum brevis myocytes from SUR2[A478V] mice (70) but not from Kir6.1[V65M] mice (71). However, reduced forelimb muscle strength, with evidence of atrophy and noninflammatory edema, is observed in both genotypes (70, 71), paralleling the finding of reduced channel ATP sensitivity in cardiac myocytes from A478V but not V65M, yet pathologic enlargement and hypercontractility in both (67). RAS is also a known mediator of atrophy and sarcopenia in skeletal muscle (72) and is potentially a major driver of the muscle pathology, but further study is needed.

Reduced SM contractility might underlie other CS features, including hypertrichosis. By opening surface blood vessels, more oxygen and nutrients are directed to hair follicles, potentially causing follicles in the telogen phase to shed and be replaced by thicker hairs in a new anagen phase (73). Kir6.1/SUR2 subunits underlie K_ATP channels in SM throughout the gastrointestinal (GI) tract (74), and GI dysmotility is also common in CS (51). Cantú mice exhibit reduced intrinsic contractility throughout the intestine, resulting in death in the most severely affected homozygous Kir6.1[V65M] animals when weaned onto solid food (74). Death is avoided by weaning onto a liquid gel diet, and GI transit is normalized by treatment with the K_ATP inhibitor glibenclamide (74). Kir6.1 and SUR2B are also expressed in mouse lymphatic SM (75–77), but Kir6.1 alone is expressed in lymphatic endothelium (75). Spontaneous popliteal lymphatic vessel contraction is inhibited by SM-specific expression of a Kir6.1 GOF subunit, but it is essentially unaffected by expression in lymphatic endothelium (75), suggesting that K_ATP GOF in lymphatic SM also underlies lymphedema in CS.

Potential GOF mutations in KCNJ8 and ABCC9 have been reported in association with other pathologies. We recently examined a de novo case of idiopathic lymphedema associated with a SUR2 mutation. Patch-clamp analysis showed this mutation to generate a marked GOF in recombinant K_ATP channels, and retrospective analysis led to a clear rediagnosis of the patient as suffering from CS (J. Gao, C. McClenaghan, I. Christiaans, M. Alders, K. van Duinen, et al., unpublished data). SUR2[V734I], reported to reduce MgATP sensitivity in mutant Kir6.2/SUR2B channels but not in Kir6.2/SUR2A or Kir6.1/SUR2B channels, was detected in patients with myocardial infarction (78, 79) and in patients with cardiac early repolarization syndrome (ERS) (80). Whether these patients might also actually represent a mild CS genotype is unclear. Kir6.1[S422L] has also been reported to underlie ERS (81), and some studies suggest that a recombinant K_ATP channel GOF may also exist (82–84). However, we could not demonstrate any such effect (56), and the recognition that this variant is common in Ashkenazim (85) suggests no causal association with early repolarization, highlighting the need for careful functional characterization and replication of potential pathological associations.

SUR2 Loss-of-Function Mutation Causes AIMS

LOF mutations in ABCC9 and KCNJ8 have been reported in association with various pathologies, including idiopathic dilated cardiomyopathy and rhythm disturbances (86), predisposition to adrenergic atrial fibrillation (87), and sudden unexplained nocturnal death syndrome (88, 89). However, there is as yet no fully established pathology arising from Kir6.1 or SUR2 LOF. We recently identified six patients from two families with a consistent phenotype of mild intellectual disability, similar facial appearance, myopathy, and cerebral white matter hyperintensities and with cardiac systolic dysfunction in the oldest patients (90). The patients are homozygous for an ABCC9 variant that causes in-frame deletion of SUR2 exon 8, which generates nonfunctional K_ATP channels in recombinant assays (90). There is similar fatigability and cardiac dysfunction in SUR2^−/− mice and zebrafish. We termed the condition ABCC9-related intellectual disability myopathy syndrome [AIMS; also referred to as intellectual disability and myopathy syndrome (IDMYS); MIM 619719]. Further studies are needed, but we suggest that AIMS represents the primary ABCC9 LOF syndrome.

Association of K_ATP Genes with More Common Diseases: Type 2 Diabetes and Pulmonary Hypertension

With a high allelic frequency (>30%), the common Kir6.2[E23K] variant is a prominent risk allele for T2DM (91, 92). Recombinant functional analyses consistently report slightly enhanced channel activity due to Kir6.2[E23K] (91, 93) or to SUR1[S1369A], which exists in linkage dis-equilibrium (94). As in NDM, variants that increase K_ATP activity are predicted to impair β cell response and decrease serum insulin levels. The variant is associated with reduced insulin secretion in humans (93, 95), and in the presence of diet-induced insulin resistance, this may be expected to exacerbate hyperglycemia. Homozygous Kir6.2[E23K] knock-in mice (96) show no effect on body weight or blood glucose, and insignificant effects on insulin secretion, on a normal diet. On a high-fat diet, there is a trend toward greater weight gain and worsened glucose tolerance in E23K females yet enhancement of insulin secretion (96). Thus, it is still not clear exactly how the variant predisposes individuals to T2DM nor, interestingly, why E23K carriers with T2DM appear to be at increased risk of secondary failure to sulfonylurea (SU) treatment (97).

SUR1 is also a regulatory subunit of K_ATP channels in mouse atria (98) and of ventricular K_ATP in humans (99). It may also be prominent in pulmonary arteries (100), and we recently reported multiple SUR1 LOF variants, some previously associated with CHI, in patients with pulmonary arterial hypertension (101). Whether these variants are causal, and how pulmonary hypertension might be linked to SUR1 LOF, but also to SUR2 GOF in CS, is not clear. Potentially, the precise K_ATP subunit composition in any given cell type may vary subtly or be more labile than is currently perceived, making it critical to focus on precise subunit distributions, and further detailed comparative analysis of cardiovascular phenotypes and outcomes in SUR1^−/− and SUR2^−/− mice may be warranted.

K_ATP CHANNEL PHARMACOLOGY

K_ATP Channel Inhibitors

The unique involvement of SUR subunits in K_ATP channel complexes underlies a rich pharmacology specific to this class of channels. This includes the SU class of inhibitors, the first-generation tolbutamide being introduced clinically in 1952 before the existence of K_ATP channels was even recognized (102). Increasingly potent and longer-acting SUs [gliclazide, glipizide, glibenclamide (glyburide), glimepiride], as well as a new structural class [glinides (metiglinide, nateglinide, and repaglinide)], have since been introduced (103).

SU inhibition of K_ATP channel activity depends in a complex way on cellular nucleotide levels (65, 104). Inhibition is also typically biphasic, reflecting two inhibitory binding sites (104). High-affinity binding at the SUR, which underlies clinical action, is transduced allosterically to closure at the Kir6 subunit, and the fraction of high- versus low-affinity inhibition depends critically on the open state stability of the channel, that is, the stability of the open versus closed state, which is in turn affected by the membrane PIP₂ level (9). Open state stability can also be markedly affected by mutations throughout Kir6 subunits and the SUR TMD0 domains. As a consequence, SU inhibition of Kir6.2/SUR1-dependent NDM mutations (41) and Kir6.1/SUR2-dependent CS mutations (64, 65) can be markedly reduced (Figure 2a), an important consideration for their use in these conditions and in the search for novel agents.

Figure 2.

Sulfonylurea inhibition of K_ATP channels. (a) Kir6.2/sulfonylurea receptor subunit 1 (SUR1)-dependent K_ATP channel activity as a function of [tolbutamide], recorded in inside-out patches, shows biphasic inhibition. Dashed lines and blue circles show wild-type (WT) channel activity. Tolbutamide sensitivity is unaffected by transient neonatal diabetes mellitus (NDM) mutant Kir6.2[I182V] but is reduced by permanent NDM mutants V59M and Q52R and abolished by the developmental delay, epilepsy, and neonatal diabetes (DEND) syndrome mutation Kir6.2[I296L] (significance indicated: *P < 0.05 and **P < 0.01). Even in the heterozygous condition (mixed wild type and I296L), high-affinity tolbutamide sensitivity is markedly reduced. Panel a adapted from Reference 41. (b) A structural model of the repaglinide (RPG)-bound SUR1 ATP-binding cassette transporter core module viewed from the side shows the slice viewed from the top and side (indicated by the two gray lines) at higher magnification to the right. The pharmacochaperone pocket is shown from the top and the side of the channel in the states indicated. Ligand density corresponding to RPG is shown in magenta, carbamazepine (CBZ) in red, and glibenclamide (GBC) in blue. (c) The Kir6.2 N-terminal cryo-electron microscopy density (pink mesh) is superposed with the polyalanine model shown in the RPG-bound (green) SUR1 structural model. The piperidino moiety of RPG is highlighted to show close proximity to the N-terminal methionine of the modeled Kir6.2 N-terminal peptide. The close proximity of SUR1 residue C1142 and Kir6.2 residue L2 is also highlighted. Panels b and c adapted with permission from Reference 105.

Cryo-EM studies reveal a common binding site for SUs and other K_ATP channel inhibitors (3, 105–107) between TMD1 and TMD2 of SUR1 (Figure 2b). Glibenclamide binding, involving two overlapping, so-called A and B sites that bridge TMD1 and TMD2, likely prevents NBF dimerization, explaining its inhibitory action (105, 108). The chlorobenzamidoethyl group interacts with the B site formed by residues on transmembrane helices 6–8 and 11, while the cyclo-hexyl moiety binds to the hydrophobic A site pocket formed by residues on transmembrane helices 16 and 17 (Figure 2b), in close proximity to S1238 in helix 16. SUR1-dependent channels are generally more sensitive to SUs than are SUR2-dependent channels, and interestingly, the equivalent residue in SUR2 is a more bulky Tyr. Substituting these residues converts high-affinity inhibition of SUR1-dependent channels to a lower SUR2-like sensitivity (109) and increases the affinity of binding and potency of inhibition of SUR2B-dependent channels (110). Deletion of up to 30 amino acids from the Kir6.2 N terminus intrinsically stabilizes the open channel, thereby reducing SU action (104, 111). Potentially explaining this finding, additional cryo-EM structures have revealed a previously unimagined insertion of the full Kir6.2 N terminus into the central cavity of SUR1, between TM1 and TMD2, adjacent to the SU-binding pocket (Figure 2c). In so doing, this potentially stabilizes bound SUs and increases inhibitory potency (105–107).

K_ATP Channel Openers

Conversely, structurally diverse benzopyrans, benzothiadiazines, cyanoguanidines, pyridyl nitrates, and thioformamides and second-generation cyclobutenediones, dihydropyridines, and tertiary carbinols (112), collectively referred to as KCOs, activate K_ATP channels. Mutagenesis studies suggest a common KCO binding site (113–115), with two regions (between Thr1059-Leu1087 and Arg1218-Asn1320 in TMD2 of SUR2) being critical for binding of the KCO P1075 (114). The first cryo-EM structure with a KCO (NN414) bound to Kir6.2/SUR1 channels (8) (Figure 3a) reveals the ligand in the cleft between TMD1 and TMD2, in contact with residues I1030, D1031, C1072, and T1286 of TMD2 as well as residues P551-V555, H584, and L580 in TMD1, confirming key findings of earlier studies. A recent paper places both the P1075 and levcromakalim activators in essentially the same site (116). As with inhibitory SUs, KCO potency and selectivity are modulated by channel open state stability and by the interactions of Mg nucleotides at NBFs (104, 117), and as discussed below, it is likely that KCO sensitivities are also strongly affected by disease mutations.

Figure 3.

KCO action and crossover drug-induced pathologies in K_ATP channels. (a, left) Cryo- electron microscopy density map shows SUR1 in complex with Mg nucleotides and NN414 (viewed from the side). TMD1-NBD1, TMD2-NBD2, and NN414 are colored in pink, blue, and red, respectively. (Right) Close-up views of the NN414-binding site. TMD1 and TMD2 are colored in pink and blue, respectively. NN414 (orange) and residues that interact with NN414 are shown as sticks. Panel a adapted with permission from Reference 8. (b) Canonical K_ATP channels are formed by the products of the ABCC8/KCNJ11 and ABCC9/KCNJ8 gene pairs. LOF mutations in either subunit of each pair (causing CHI and AIMS, respectively) and GOF mutations (causing NDM and Cantú syndrome, respectively) may be treatable with channel-activating KCOs or inhibitory SUs. However, crossover effects on the unaffected channels encoded by the other gene pair cause inevitable side effects. (c) Hypertrichosis and fluid retention characterize both Cantú syndrome, caused by a Kir6.1/SUR2 GOF mutation, and CHI when treated with diazoxide. Images in panel c reproduced with permission from References 53 and 179. Abbreviations: AIMS, ABCC9-related intellectual disability myopathy syndrome; CHI, congenital hyperinsulinism; GI, gastrointestinal; GOF, gain-of-function; K_ATP, ATP-sensitive potassium; KCO, K_ATP channel opener; Kir, pore-forming inward rectifier potassium channel subunit; LOF, loss-of-function; NBD, nucleotide-binding domain; NDM, neonatal diabetes mellitus; SU, sulfonylurea; SUR, sulfonylurea receptor subunit; SVR, systemic vascular resistance; TM, transmembrane; TMD, transmembrane domain.

MONOGENIC K_ATP DISEASE THERAPIES

Current Pharmacological Therapies for Congenital Hyperinsulinism

First sold as Hyperstat, for emergent hypertension, diazoxide was recognized 60 years ago as inducing VSM relaxation and hence blood vessel vasodilation (118). Almost as long ago, it was realized that it also inhibited insulin secretion from pancreatic β cells (119). Some CHI patients with K_ATP LOF mutations must maintain some K_ATP channel activity since they are responsive to diazoxide (120), which (sold as Proglycem) is now the preferred treatment for responsive CHI patients (121, 122). Other K_ATP-dependent CHI cases completely lack K_ATP channels. In these and other cases that are unresponsive to diazoxide (or other drugs, including chlorothiazide, nifedipine, glucagon, and somatostatin, all of which aim to suppress insulin secretion), partial or total pancreatectomy is generally required (121). Surgical resection is also effective in patients with focal hyperinsulinism once the lesion has been preoperatively located (123). Importantly, in nonsurgically treated K_ATP-dependent and non-K_ATP-dependent CHI, hyperinsulinism symptoms typically lessen with age and diazoxide may no longer be needed (124–127). One study found that in adult carriers of dominant K_ATP mutations, all of whom would be expected to have suffered hyperinsulinism as children, essentially half were symptomless (potentially indicating a preceding remission), and several were diabetic (22). Multiple other studies report CHI patients with dominant ABCC8 LOF mutations who remitted and then progressed to diabetes later in life (23, 125, 128–130).

Partial loss of K_ATP in mice causes hyperinsulinism (131, 132), but mice with complete loss of K_ATP cross over to undersecretion and glucose intolerance (133–135), and overt diabetes after being fed a high-fat diet (136). Potentially this crossover reflects the same progression as that seen in human CHI. Strikingly, SUR1[E1506K] knock-in mice, carrying the same mutation as a diabetes-prone Finnish CHI pedigree (23), show hyperinsulinism in the neonatal period but again rapidly progress to undersecretion (137). Also seen in patients with CHI caused by GOF mutations in glucokinase (GCK), which lead to an elevated ATP:ADP ratio and consequent suppression of K_ATP activity (138, 139), this progression from hyperinsulinism to diabetes is unexplained but needs to be understood for appropriate long-term treatment and management of CHI.

Current Pharmacological Therapies for Neonatal Diabetes Mellitus

The realization that K_ATP channel GOF mutations cause NDM rapidly shifted therapy from insulin injections to oral SUs. By directly inhibiting overactive K_ATP channels, these drugs can provide successful control of blood glucose levels, removing (or markedly reducing) insulin requirements and significantly improving glycemic control and overall lifestyle for most K_ATP-induced NDM patients (140, 141). Unlike injected insulin, which indiscriminately lowers blood glucose by promoting peripheral uptake, SUs directly target K_ATP overactivity, and insulin secretion becomes quasi-physiological, with much better glucose control than when insulin treated (142).

However, the many mutations that reduce both ATP and SU sensitivity (41) contribute to SU dose requirements being generally higher than those for treatment of T2DM (140, 143), and the success rate for SU treatment is lower in patients with the more severe DEND symptoms (140, 144–147). Moreover, while SUs can resolve diabetes and improve neurological function in some DEND cases (32, 142, 148), neurologic features are unresponsive in most. SUs are transported out of the brain by the ABC protein P-glycoprotein, potentially lowering central nervous system (CNS) concentrations (149, 150). Together with the marked loss of SU sensitivity in many DEND mutations (41, 151) (Figure 2a), this concentration lowering in the CNS provides an explanation for variable SU response in DEND. As discussed below, new agents with better brain-penetrating abilities, avoiding P-glycoprotein export, as well as new mechanisms of action, are needed to resolve this issue.

There is a correlation between the age at which SU transfer is attempted in human NDM and overall β cell responsiveness: The younger the patient, the greater the chance for successful SU therapy (140, 143). One case report showed that the number of β cells in an insulin-treated 2-year-old female with a KCNJ11 mutation was significantly lower than in normal age-matched children (152), and loss of β cell mass is also consistently observed in diabetic pancreatic K_ATP GOF mice (36, 37). This loss of β-cell mass is associated with dedifferentiation to progenitor cells (153), but importantly, these dedifferentiated cells can redifferentiate to mature insulin-secreting β cells after normalization of blood glucose by intensive insulin therapy (153). If initial hyperglycemia is avoided, either by early SU therapy or by syngeneic islet transplantation, loss of β cell mass is prevented (36), and acute SU treatment at the onset of transgene induction in a subset of pancreatic K_ATP GOF mice actually led to persistently near-normal glucose levels once the SU therapy was terminated (154). Conceivably, this normalization could reflect the mechanistic progression of human TNDM, wherein early avoidance of hyperglycemia might allow a long-term compensatory mechanism to provide remission. While unexplained, the finding predicts that early intervention with SUs in K_ATP-dependent NDM may help preserve β cell function. Retrospective study shows that initiation of SU therapy before genetic testing results are available is generally safe and often successful in NDM patients (155), supporting aggressive early SU intervention.

One potential caveat in the undoubted success of SU therapy for NDM is that, by analogy to the progressive loss of SU responsiveness seen in T2DM (156), SU treatment might eventually fail, in which case insulin therapy may be needed. Experience to date has not borne this out (140, 157, 158), suggesting that SU treatment for NDM—which is intended to normalize insulin secretion, not hyperstimulate it—may be safe and effective in the long-term.

Therapies for Cantú Syndrome and AIMS

Reduced vascular contractility probably underlies the common finding in CS of persistent patent ductus arteriosus (51), often requiring surgical closure. Pericardial effusion sometimes requires pericardiocentesis and, ultimately, pericardial stripping to prevent reaccumulation. High incidence of cerebrovascular abnormalities in CS (159) may lead to magnetic resonance angiograms to evaluate persistent headaches or other neurological symptoms becoming standard of care. Treatments for hypertrichosis include shaving, use of depilatories, or laser hair removal. CS patients also require monitoring for edema, which often develops in adolescence or early adulthood and may benefit from compression stockings. Other than these generally palliative treatments, there is currently no specific therapy for CS. Some treatments are essential and effective, others are incompletely effective, and some currently prescribed cardiac agents (51) such as angiotensin-converting enzyme inhibitors or β-blockers may even be contraindicated: Captopril causes extreme BP lowering in CS mice (68), and propranolol causes significant cardiac contractile inhibition (160).

Clearly, the use of appropriate K_ATP channel inhibitors that could reverse the underlying molecular defect is a potential magic-bullet therapy, as SUs have proven to be for NDM. In SUR2[A478V] mice, chronic glibenclamide treatment (1-month slow-release pellets), initiated at 2 months of age, does substantially reverse cardiac enlargement and normalize BP, although it fails to reverse aortic enlargement, and normalization of cardiac size and BP is incomplete in the more severely affected V65M mice (69). In one case study, glibenclamide did appear to be beneficial in a CS child with SUR2[R1116H] (161). At the time of writing, the child is 5 years old and still on the same glibenclamide dose, without overt cardiac or other complications (A. Ma, personal communication) but with no resolution of hypertrichosis. Additional unpublished efforts in infants thus far have not shown obvious symptom improvement. Nevertheless, the US Food and Drug Administration (FDA)-approved status of SUs, the availability of validated animals in which to assess their action (67), and the availability of a cohort of motivated CS patients in whom to assess clinical outcomes (51) justify an urgent need to examine the clinical potential for SU use in CS, and to this end, we are currently carrying out initial glibenclamide tolerance tests in older CS patients.

Unfortunately, all commercially approved and available K_ATP channel inhibitors are also active on Kir6.2/SUR1 channels, and while it seems likely that cardiovascular, cerebrovascular, and lymphatic dysfunction in human CS will all respond to SUs, there is a risk of hypoglycemia, as discussed below.

Crossover drug-induced pathologies.

As discussed above, current therapies for K_ATP pathologies rely on reversing channel GOF or LOF mutations by using SU inhibitors or KCOs, respectively. In each case, the target is the mutated Kir6.2/SUR1 or Kir6.1/SUR2 channel pair, but the poor selectivity of currently available agents will inevitably have an undesired action on the unaffected channel pair (Figure 3b).

K_ATP channel openers.

This crossover action is clearly a problem for diazoxide treatment of CHI, with extensive documentation of complications, most notably hypertrichosis but also the development of abnormal facial features, low BP, and edema, as well as potentially lethal respiratory failure and pulmonary hypertension (162), as seen in CS (52), prompting a black box warning by the FDA in 2015. Suspected necrotizing enterocolitis with abdominal distension (which resolved after discontinuation of diazoxide) has also been reported in both diazoxide treatment (162) and CS (52).

Marketed as Rogaine in the United States, and Regaine in Europe, the KCO minoxidil is a popular over-the-counter treatment for hair loss. The mechanism of action remains unclear (73), but since it is an effective opener of Kir6.1/SUR2B channels, the association of Kir6.1/SUR2 GOF mutations with hypertrichosis in CS definitively ties hair growth to K_ATP channel activation. Since minoxidil preferentially acts on SUR2-dependent channels, blood-glucose lowering is not generally reported as a side effect, but other drug-induced Cantú features similar to those seen with diazoxide therapy are common and have led to very severe problems in people using the drug for hair growth (73) and even in their pets after accidental ingestion (163).

K_ATP channel inhibitors.

Since Kir6.2/SUR1 channels are generally more sensitive to glibenclamide than are Kir6.1/SUR2 channels, side effects of glibenclamide therapy for NDM are generally minimal, although hypersensitivity of the skin, stomach upset, and diarrhea can be common (164). It is tempting to suggest that the latter results from inhibition of K_ATP channels in the GI SM, given that glibenclamide markedly increases motility in Cantú animals (74). More significantly, the higher sensitivity of Kir6.2/SUR1 channels to SUs presents the potential for severe blood-glucose lowering if SUs are used to treat CS. As discussed above, the glucose-lowering effect of high-dose SU is only transient in mice, and then the animals rapidly lose glucose-sensitive insulin secretion, resulting in slight glucose intolerance, as in K_ATP knockout animals (165). Whether the same occurs in humans is unclear. The evidence from treatment of NDM, in which the SU dose is adjusted to provide pseudonormal excitability and insulin secretion, indicates that there is a sustained effect. However, in T2DM, where SU treatment promotes supraphysiological insulin secretion by inducing supranormal excitability, the drugs gradually become ineffective. SU sensitivity can be restored, however, by so-called β cell rest with intensive insulin therapy, conceivably reversing the same desensitization that occurs in mice with high-dose SU treatment. If so, then the same desensitization of the otherwise normal pancreas might occur if SU is used at high doses to reverse CS features. Intriguingly, in the single reported human CS case treated with glibenclamide, transient hypoglycemia was observed with each dose escalation (161), suggesting the concern may not be preclusive.

POTENTIAL FOR NEW THERAPEUTICS

Crossover drug-induced pathologies, compounded by reduced sensitivity to channel inhibitors with GOF mutations, highlight a need for new inhibitors and new activators for K_ATP and especially a need for sharper therapeutics (103) with greater isoform specificity.

New Small Molecules

High-throughput screening may identify novel or more potent inhibitors or KCOs acting through the identified binding sites in the SUR cores (8), but the structural similarity of these sites in SUR1 and SUR2 may ultimately preclude true isoform specificity, making crossover drug-induced complications unavoidable. One potential lead is the non-SU inhibitor U37883A, which has prohibitive off-target effects (166) but does have a novel site of action on Kir6.1 (167). To overcome the potential that new inhibitors may still be ineffective because of open state stabilization with GOF channel mutations, Kharade et al. (103) proposed to identify K_ATP channel correctors by screening chemical libraries using functional assays with disease-causing mutant channels. With the use of rapid cloning techniques, including landing-pad cells, in which a stably integrated cloning site has been genomically incorporated (168), it is now feasible to rapidly develop the multiple necessary cell lines (169). The use of tetracycline induction of the active channel minimizes cell line degeneration and loss of assay performance associated with constitutive K_ATP channel overexpression, making it possible to repeatedly assay the same lines (170).

As initially implemented, fluorescent assays are performed in multiwell plates using a standard kinetic imaging plate reader to measure the flux of thallium (Tl⁺) as a K⁺ surrogate with a Tl⁺-sensitive dye (171). Both Kir6.2/SUR1 and Kir6.1/SUR2B channels are mostly closed under basal physiologic conditions due to the relatively high cytoplasmic [ATP]:[ADP] ratio in HEK-293 cells, and channels must first be opened with metabolic inhibitors or a KCO. However, as we have recently shown (169), voltage-sensitive DiBAC₄(3) [bis-(1,3-dibutylbarbituric acid)trimethine oxonol] fluorescence provides a very sensitive assay for the hyperpolarizing effect of even a very low basal K_ATP conductance. Such an approach makes high-throughput screening for viable inhibitors of Kir6.2/SUR1 or Kir6.1/SUR2B channels with GOF mutations possible under basal physiological conditions. Screening for activators can be achieved by simply leaving the channels closed and looking for compounds that induce Tl⁺ flux. A novel-scaffold xanthine derivative (VU063), discovered using this approach, is an effective KCO that acts on Kir6.2/SUR1 channels (171), is approximately 10 times more potent than diazoxide, and is an effective inhibitor of glucose-stimulated calcium entry in primary mouse pancreatic β cells.

New Biological Therapeutics

Advances in gene editing may achieve the ultimate treatment for monogenic pathologies—correction of the underlying mutations (172). In the meantime, new biological agents, particularly targeted antibodies and nanobodies, are revolutionizing disease therapeutics. Facilitated by unique cryo-EM-identified structural motifs, such agents may be ideal for achieving subunit specificity. For instance, the unique structure of the Kir6.1 extracellular turret region has now been resolved (11) and might represent a targetable Kir6.1-specific domain. Conceptually, an inhibitor acting on the channel from the extracellular side might also avoid the issue of drug interactions with nucleotide modulation.

A natural venom screen recently identified a novel 54-residue protein toxin (SpTx-1) that inhibits both wild-type and NDM mutant Kir6.2/SUR1 channels with low nanomolar potency from the extracellular side (173). Intriguingly, toxin sensitivity was dependent on human-specific residue E108 (174), and in K_ATP GOF mice expressing humanized Kir6.2[ΔN30, K185Q, V108E], the toxin was effective at controlling blood sugar levels (174), a promising indicator that such an approach may provide another avenue to identifying novel channel inhibitors. High-resolution structures also open up the possibility of developing new modulators using in silico approaches. Molecular dynamics simulations of DEND syndrome Kir6.2[Q52R] and [L164P] mutations recently identified betaxolol and levobetaxolol as effective pore blockers, as well as tavoprost, a SUR-binding prostaglandin with potencies in the micromolar range (175).

Finally, both SUs and KCOs can rescue surface expression of mistrafficked mutant channels in vitro (176–178), raising the possibility of molecular chaperone drugs as a novel therapy to treat trafficking defects in HI. Since SUs present a Catch-22 situation by inhibiting channels even as they increase trafficking, translation to a viable therapy may be difficult, but it does illustrate the concept that there may be additional routes to modifying channel activity.