E23K, a common single nucleotide polymorphism in KIR6.2, the pore-forming subunit of pancreatic β-cell ATP-sensitive K+ channels, significantly enhanced open probability of these channels, thus reducing their sensitivity toward inhibitory ATP4− and increasing the threshold concentration for insulin release. Previous association studies and high allelic frequency suggest this effect to critically inhibit secretion and play a major role in pathogenesis of common type 2 diabetes. Based on evidence for functional relevance of E23K in both the heterozygous (E/K; with E in position 23 of KIR6.2 in one allele and K in the other) and homozygous (K/K; with K in position 23 of KIR6.2 in both alleles) genotype, we propose a model in which enhanced susceptibility to type 2 diabetes is associated with evolutionary advantage of the E/K state.
Type 2 diabetes is generally perceived as a polygenic disorder, with disease development being influenced by both hereditary and environmental factors (1). Genes encoding for key components of insulin secretion and glucose metabolism pathways have been widely considered as targets for defects in type 2 diabetes. However, despite intensive investigations, little progress has been made in identifying the genes that impart susceptibility to the common late-onset forms of the disease (2).
In pancreatic β-cells, ATP-sensitive K+ (KATP) channels critically control insulin secretion by coupling metabolism to electrical activity (3). Recent advances resulted in cloning of these channels and elucidation of their subunit composition (4). The β-cell channels are assembled, with tetradimeric stoichiometry, from two structurally distinct subunits: an inwardly rectifying K+ channel subunit (KIR6.2), forming the pore, and the regulatory sulfonylurea receptor subunit-1 (SUR1). While hypoglycemic sulfonylureas (e.g., glibenclamide) exert their effects on channel activity by interaction with SUR1, there is strong evidence that the receptor site for inhibitory ATP4− is formed by KIR6.2.
Three common missense single nucleotide polymorphisms (SNPs) have been observed in KIR6.2 (E23K, L270V, and I337V) (5–9), and their potential impact in type 2 diabetes led us to analyze their functional relevance. Whereas L270V and I337V were without effect on the properties of reconstituted human SUR1/KIR6.2 channels (including expression rate, single channel conductance, spontaneous open probability [PO], and nucleotide and drug sensitivities [results not shown]), E23K markedly affected channel gating, significantly reducing the time spent in long interburst closed states (17 ± 3% for SUR1/the mutant isoform of KIR6.2 with K instead of E in position 23 [KIR6.2E23K] vs. 54 ± 6% for wild-type channels, n = 10 each, P < 0.001) (Fig. 1A and B), thus producing a 1.6-fold increase of PO (PO = 0.66 ± 0.05 for SUR1/KIR6.2E23K vs. 0.41 ± 0.04 for wild-type channels, n = 10 each, P < 0.001). The increase of PO was confirmed by noise analysis (patches with 100–500 channels; PO = 0.85 ± 0.04 for SUR1/KIR6.2E23K vs. 0.58 ± 0.03 for wild-type channels, n = 10 each, P < 0.001) and determination of the maximal increment of the spontaneous patch current (ΔImax; PO = 0.77 ± 0.04 for SUR1/KIR6.2E23K vs. 0.52 ± 0.02 for wild-type channels, n = 10 each, P < 0.001) (Fig. 1C). Importantly, a model for the heterozygous genotype (E/K, with E in position 23 of KIR6.2 in one allele and K in the other; coexpression of KIR6.2E23K and the wild-type isoform of KIR6.2 [KIR6.2wt], with a cDNA ratio of 1:1) (Fig. 1C) resulted in intermediate PO values (0.67 ± 0.02 and 0.61 ± 0.02 from noise analysis and measurement of ΔImax, respectively; n = 10; P < 0.01 for comparison with pure wild-type channels each).
Consistent with the idea that nucleotide-induced channel inhibition results from interaction with the interburst closed states (10,11), E23K decreased sensitivity toward inhibitory ATP4− in the models for the heterozygous (E/K) and homozygous (K/K; with K in position 23 of KIR6.2 in both alleles) genotype by 1.4- and 2.2-fold, respectively (IC50 = 12.4 ± 0.3 μmol/l for coexpression of KIR6.2E23K with KIR6.2wt and 19.7 ± 0.5 μmol/l for pure SUR1/KIR6.2E23K channels vs. 8.9 ± 0.2 μmol/l for pure wild-type channels; n = 15 each; P < 0.01 or <0.001 for comparison with wild-type channels, respectively) (Fig. 2A and B).
Stimulation of insulin secretion requires reduction of the PO of pancreatic β-cell KATP channels to values <0.02 (12). Both increased spontaneous PO values (Fig. 1), and reduced ATP sensitivity (Fig. 2A) contributed to a rightward shift of the corresponding ATP concentration (ICR2, defined as the ATP concentration that suppresses PO to the threshold for insulin secretion [0.02]) through E23K (Fig. 2B). This shift amounted to 1.5- or 3.1-fold in the models for the E/K and K/K state, respectively (ICR2 = 154 ± 6 μmol/l for coexpression of KIR6.2E23K with KIR6.2wt and 303 ± 11 μmol/l for pure SUR1/KIR6.2E23K channels vs. 98 ± 4 μmol/l for pure wild-type channels, n = 15 and P < 0.001 for comparison with wild-type channels each). The effect of E23K on the ICR2 value was not altered by additional introduction of L270V and/or I337V (Fig. 2C).
All results obtained with the human isoforms were confirmed in channels reconstituted from mouse KIR6.2 plus hamster SUR1 (results not shown). Effects of E23K on PO and ATP sensitivity were consistent with previous reports on the role of the KIR6.2 NH2 terminus (10,11,13).
Results indicate that E23K in KIR6.2 significantly affects properties of pancreatic β-cell KATP channels, increasing spontaneous PO and reducing ATP sensitivity in models for the heterozygous (E/K) and homozygous (K/K) state (Figs. 1 and 2). Both effects contribute to a rightward shift of the ATP concentration required to suppress channel activity to the threshold for insulin secretion (ICR2) (Fig. 2B), implying that E23K might induce a critical inhibition of release.
This idea is strongly supported by recent reports that demonstrate that lowering ATP sensitivity by no more than 3.6-fold induces severe neonatal diabetes in transgenic mice (3), and that reveal clear association of homozygous E23K (K/K) with type 2 diabetes in Caucasians (overall χ2 = 21.3 with 1 df, P < 0.000004) (8,9). Particularly, the latter studies suggest that because of its high allelic prevalence (weighted averaged estimate in Caucasians = 34%) (5–9), E23K is of considerable importance in the pathogenesis of common type 2 diabetes.
Partial absence of coupling with E23K led to a lack of detectable association of homozygous I337V with type 2 diabetes (8), arguing against localization of the disease locus downstream (i.e., in the 3′ direction) codon 337 of the KIR6.2 gene.
At 7.5 kb from the 5′ end of the KIR6.2 gene, a common missense SNP in codon 1369 of the SUR1 gene was found to be in linkage disequilibrium with E23K in KIR6.2 (>77%) (6). In another population, however, linkage could not be confirmed (8,16), and thus, similar to the conclusion drawn above, localization of the causative variant upstream (i.e., in the 5′ direction) codon 1369 in the SUR1 gene seems unlikely.
Regions coding for the COOH terminus of SUR1 (exons 33–39) did not enclose any further common SNPs with potential functional relevance (14,16); hence, so far the only alternative candidate chromosomal section left would be that in-between the genes for SUR1 and KIR6.2, encompassing the promoter region of KIR6.2 (17). Modulation of KIR6.2 expression, however, is presumably not involved because isolated targeted overexpression of KIR6.2 does not lead to enhanced formation of functional channels (3).
We conclude that E23K promotes development of type 2 diabetes by increasing the threshold ATP concentration, thus inducing overactivity of pancreatic β-cell KATP channels and inhibiting insulin secretion. This model is consistent with the postulated critical role of an inborn secretory defect in the genesis of type 2 diabetes (1).
The conclusion that E23K may induce a discrete inhibition of insulin secretion is supported by a study in healthy young adults (7). Here, E23K was associated with either direct (slightly reduced glucose-stimulated release) or indirect (elevated insulin sensitivity or glucose effectiveness) (examples in 18,19) evidence for decreased secretion. However, because most of these effects did not reach statistical significance, definite proof of reduced secretion will require additional, more detailed clinical studies.
High allelic frequency of E23K in Caucasians (see above) with similar values in all populations screened (5–9) suggests that E23K represents a balanced polymorphism, conferring selectionary advantage through fine-tuning of insulin secretion in heterozygotes. By reducing glucose uptake in muscle and fat, discrete inhibition of release might result in favorable substrate supply for tissues with insulin-independent uptake, and hence high frequency of the E/K state (∼45%) (5–9) might have evolved as an adaptation to the human brain. Notably, this model implies increased susceptibility to type 2 diabetes as an inherent price for the evolutionary benefit of the heterozygous state, and thus E23K provides evidence in support of the “thrifty genotype” hypothesis (20). However, diverging from this concept, predisposition might have evolved as a response to altered tissue demands rather than periodic famine.
Early studies (5–7) were unable to reveal the impact of E23K because they were not designed to detect a trait with a relative risk as low as 2, based on small quantitative changes of channel properties. Specifically, the number of probands in initial association studies (5–7) were too low to reach statistical significance, and functional analysis was restricted to assessment of end points of metabolic channel control (5). The latter presumably prompted failure to detect relevance of the polymorphism because E23K affects neither maximal nucleoside diphosphate–induced channel activation (i.e., metabolic inhibition through 3 mmol/l azide ) nor maximal ATP-mediated closure (i.e., millimoles per liter of resting ATP levels ) (Fig. 1C).
A recent family-based study failed to replicate association between E23K and type 2 diabetes (21). However, this contradictory result is presumably explained by weak predisposition through the heterozygous state plus specific selection of probands with type 2 diabetes. The study did not focus on the typical type 2 diabetic patient but was mainly grounded on younger probands with prediabetic irregularities of glucose homeostasis. Yet here, discrete inhibition of insulin release is expected to be balanced by compensatory mechanisms (e.g., increased insulin sensitivity or glucose effectiveness; see above).
The present report is the first to identify the subtle variation of KATP channel function that is induced by E23K in KIR6.2. We infer that this functionally relevant SNP plays an important role in the pathogenesis of common type 2 diabetes and propose a model in which predisposition results as an implicit price for advantage associated with the heterozygous state.
RESEARCH DESIGN AND METHODS
All chemicals were obtained from the sources described elsewhere (22).
Point mutations were introduced into human (GenBank Q14654) or mouse (GenBank D50581) KIR6.2 by standard molecular biology techniques, and constructs were sequenced to verify PCR fidelity before transfection. For analysis of combinations of the polymorphisms (Fig. 2C), mutations were sequentially introduced into the same KIR6.2 cDNA.
Transfections were performed as described (22). COS-1 cells were cultured in Dulbecco’s modified Eagle’s medium–high glucose (DMEM-HG) (10 mmol/l glucose) supplemented with 10% FCS, plated at a density of 8 × 104 cells/dish (35 mm), and allowed to attach overnight. For analysis of single-channel kinetics (Fig. 1A and B), 2 μg of pECE-human SUR1 (GenBank NP_000343) complementary DNA and 2 μg of pECE-human KIR6.2 complementary DNA were mixed and used to transfect six 35-mm plates. Amounts of complementary DNA used in transfections for patches with >100 channels (Figs. 1C and 2) were 10-fold higher (20 μg each) than those for measurement of single-channel kinetics. In controls we used pECE-hamster SUR1 (GenBank A56248) and pECE-mouse KIR6.2 instead of the human isoforms. For transfection the cells were incubated for 4 h in a Tris-buffered salt solution containing DNA (5–10 μg/ml) plus DEAE-dextran (1 mg/ml), 2 min in HEPES-buffered salt solution plus DMSO (10%), and 4 h in DMEM-HG plus chloroquine (100 μmol/l). Cells were then returned to DMEM-HG plus 10% FCS. Experiments in the inside-out configuration of the patch-clamp technique were performed 1–2 days after transfection at room temperature. Membrane patches were clamped at −50 mV. The intracellular bath solution contained (in mmol/l): 140 KCl, 2 CaCl2, 10 EDTA, and 5 HEPES (pH 7.15). The pipette solution contained (in mmol/l): 140 KCl, 2.6 CaCl2, 1.2 MgCl2, and 10 HEPES (pH 7.4).
Open and closed time distributions in single-channel registrations (Fig. 1B) were analyzed using pCLAMP Software version 6.0 (Axon Instruments, Union City, CA). TO is the time constant of the open state; TC1, TC2, A1, and A2 are time constants or relative fractions of short (intraburst) or long (interburst) closed states, respectively. The fraction of time spent in long interburst closed states (PC2) was calculated with PC2 = (TC2 × A2)/(TC1 × A1 + TC2 × A2) × (1 − PO), where PO is the mean PO (in the absence of drugs and nucleotides) calculated from single-channel current registrations (see below).
For calculation of PO from single-channel currents, only records without superimposed openings were accepted. PO was calculated by dividing the total time the channel spent in the open state by the sample time (30 s). For assessment of PO by noise analysis (10) or determination of the maximal current increment (ΔImax) (Fig. 1C), patches with multiple channels were used. Density of KATP channels per patch ranged from 100 to 500. Estimates of PO were not affected by this variation. In coexpression experiments (KIR6.2wt plus KIR6.2E23K with a cDNA ratio of 1:1) (Fig. 1C), patches with >150 channels were chosen to attain an acceptable frequency of each channel species (22). PO was calculated with PO = 1 − [ς2/(i × I)] or PO = I/(I + ΔImax), respectively, where I is the mean spontaneous patch current (in the absence of drugs and nucleotides), ς2 is the variance, i is the single channel current, and ΔImax is the increment of the patch current induced by the simultaneous addition of 0.3 mmol/l diazoxide plus 3 mmol/l GDP (Fig. 1C).
For analysis of ATP sensitivity (Fig. 2), patches were chosen with little “run-down” over the measuring period, and effects were corrected for this loss of channel activity by use of linear interpolation. Artifacts caused by incomplete wash-out or slow reversibility were excluded by making sure that experiments with stepwise decrease of the ATP4− concentration yielded EC50 values and slope factors identical to those from experiments with stepwise increases of the concentration. Variation of channel densities (see above) did not affect percentage of sensitive channels, EC50 values, or Hill coefficients (Fig. 2B). Curves shown in Fig. 2B were normalized to PO values calculated from single-channel registrations because we felt these estimates to be more accurate than determinations by noise analysis or measurement of ΔImax (see above). The equivalent value for the 1:1 coexpression, PO-S(1:1), was approximated from the values obtained in patches with multiple channels as follows: PO-S(1:1) = PO-S(wt) + [(PO-S(E23K) − PO-S(wt)) × (PO-M(1:1) − PO-M(wt))/(PO-M(E23K) − PO-M(wt))], where PO-S(wt) and PO-S(E23K) are the PO estimates from single-channel registrations for pure wild-type (0.41) and E23K (0.66) channels, respectively, and PO-M(wt), PO-M(1:1), and PO-M(E23K) are the arithmetic means of corresponding values from noise analysis and determination of ΔImax (0.55, 0.64, and 0.81, respectively; see text).
Data analysis and statistics were performed as described (22). Results are shown as records from representative single experiments (Figs. 1 and 2A) or as the means ± SE (n = 8–16). ICR2 values were estimated by fitting the function PO = 1/[1+([ATP]/IC50)n] to the data of each single experiment, where IC50 is the half-maximally inhibitory ATP-concentration and n is the Hill coefficient. P values were calculated by the Mann-Whitney U test with correction for multiple comparisons or χ2 statistics based on data from Table 2 in ref. 8 and Table 4 in ref. 9.
This work was supported by grants from the Deutsche Forschungsgemeinschaft (to M.S. and C.S.).
We are grateful to Drs. Lydia Aguilar-Bryan and Joseph Bryan (Baylor College of Medicine, Houston, TX) for providing us with the human isoforms of SUR1 and KIR6.2 and to Drs. Heinz dal Ri (University of Göttingen, Göttingen, Germany) and Colin Nichols (Washington University School of Medicine, St. Louis, MO) for helpful discussion. We thank Haide Fürstenberg, Ursula Herbort-Brand, Claudia Ott, Beate Pieper, Carolin Rattunde, and Petra Weber for their excellent technical assistance.
Address correspondence and reprint requests to Dr. M. Schwanstecher, Institut für Pharmakologie und Toxikologie, Universität Braunschweig, Mendelssohnstraβe 1, 38106 Braunschweig, Germany. E-mail:.
Received for publication 18 December 2000 and accepted in revised form 29 November 2001.
DMEM-HG, Dulbecco’s modified Eagle’s medium–high glucose; ICR2, ATP concentration that suppresses spontaneous open probability to the threshold for insulin secretion (0.02); ΔImax, maximal increment of the spontaneous patch current; KATP, ATP-sensitive K+ channel; KIR6.2E23K, mutant isoform of KIR6.2, with K instead of E in position 23; KIR6.2wt, wild-type isoform of KIR6.2; PO, spontaneous open probability; SNP, single nucleotide polymorphism; SUR1, regulatory sulfonylurea receptor subunit-1.