Kir6.2 Variant E23K Increases ATP-Sensitive K⁺ Channel Activity and Is Associated With Impaired Insulin Release and Enhanced Insulin Sensitivity in Adults With Normal Glucose Tolerance

Nondiabetic subjects, aged <65 years in good health and with stable weight for 6 months, were recruited using advertisements. The studies were approved by the Human Research Protective Office.

Nine subjects with the K/K genotype and nine age-, sex-, and BMI-matched subjects with the E/E genotype participated in three separate protocols designed to test insulin secretion, action, and clearance. All subjects were unrelated and had no family history of type 2 diabetes. Insulin secretion was assessed using the 5-h oral glucose tolerance test (OGTT) and graded glucose infusion (GGI). Insulin sensitivity was assessed during the hyperinsulinemic-euglycemic clamp and OGTT. Insulin clearance was assessed during OGTT and GGI.

A total of 461 volunteers who responded to advertisements were genotyped for the E23K variant and underwent 5-h OGTTs for assessments of insulin secretion, action, and clearance.

Subjects adhered to their regular diet and refrained from exercise for 3 days before the studies.

Participants ingested a 75-g glucose load. Blood samples were collected at −15, 0, 10, 20, 30, 60, 90, 120, 150, 180, 240, and 300 min after ingestion to determine plasma glucose, insulin, and C-peptide concentrations. GGI involved the intravenous administration of glucose at increasing rates (1, 2, 3, 4, 6, and 8 mg · kg⁻¹ · min⁻¹) for 40 min each (15,16). This protocol raises the plasma glucose concentration from basal to hyperglycemic levels and defines the dose-response relationships between glucose and insulin secretion.

Studies (17,18) were performed after an overnight fast as described. A primed (22.5 μmol/kg)-constant (0.25 μmol · min⁻¹ · kg⁻¹) infusion of [6,6-²H₂]glucose was started at 0700 h. After 3.5 h of tracer infusion, a hyperinsulinemic-euglycemic clamp was started and continued for 3.5 h. Insulin was infused at 40 mU/m² per min. Dextrose (20%) with [6,6-²H₂]glucose (∼2.5%) was infused to maintain plasma glucose concentration at ∼5.6 mmol/l. The infusion of [²H₂]glucose was decreased by 75% of basal during the clamp for the expected decline in hepatic glucose production. Body composition was assessed using the same dual-energy X-ray absorptiometry instrument in all subjects (Delphi 4500-W; Hologic, Waltham, MA).

The areas under the curve (AUCs) were calculated using the trapezoid method (19). ISR was calculated using parameters for C-peptide kinetics and volume of distribution (20). Static, dynamic, and overall responsivity indexes were calculated as reported (21). The GGI allowed the dose-response relationship between ISR and glucose over the physiological range to be determined. ISRs and glucose concentrations represented the average of the values between 10 and 40 min at each infusion rate. Mean ISR for each glucose infusion rate was plotted against the mean glucose concentration. The slope of the line relating these two variables provided a measure of the sensitivity of the β-cell to glucose. Peripheral insulin sensitivity was calculated from the hyperinsulinemic-euglycemic clamp (17,18,22). Basal endogenous glucose rate of appearance was calculated using Steele's equations (20). Hepatic insulin sensitivity was assessed as the inverse of the product of the basal endogenous glucose production rate (μmol · kg fat-free mass⁻¹ · min⁻¹) and the fasting insulin concentration (pmol/l) × 1,000 (23,24). Whole-body insulin sensitivity was estimated using the oral glucose minimal model (25,26). Composite insulin sensitivity index was calculated using the method of Matsuda (23). Insulin clearance rate was calculated by dividing the 1) ISR AUC by the insulin AUC and the 2) C-peptide AUC by the insulin AUC (15,27).

Plasma glucose was measured using a glucose analyzer (Yellow Springs Instruments, Yellow Springs, OH). Plasma insulin and C-peptide were measured using radioimmunoassays (Linco Research, St.Louis, MO).

DNA was prepared from peripheral blood lymphocytes and the KCNJ11 E23K polymorphism (dbSNP rs5219) typed using a TaqMan SNP Genotyping Assay (Applied Biosystems, Foster City, CA).

Group differences were compared using the Student's t test for unpaired samples for continuous variables and the Fisher's exact test for categorical variables. Where appropriate, we used ANOVA and Tukey's test for post hoc analyses for continuous variables and χ² test for categorical variables. ANCOVA were used to adjust for age, race, and BMI. SPSS version 15.0 (SPSS, Chicago, IL) was used for all analyses. A P value <0.05 was considered to be statistically significant. Results are reported as means ± SE.

Recombinant K_ATP channel Kir6.2 (mouse) + SUR1 (hamster) was transiently expressed in COSm6 cells as described (28). Of the missense variants reported in humans, the Kir6.2 clone contained no variants, while the SUR1 clone contained the A1369S substitution. Patch-clamp experiments were performed as described (28). Bath and pipette solutions (K-INT) contained (in mmol/l): 150 KCl, 10 HEPES, and 1 EGTA (pH 7.4). For experiments with MgATP, the free Mg²⁺ concentration was kept at 2 mmol/l. The ATP dose response was quantified by fitting the raw data with a Hill equation:

where Irel is the current relative to that in the absence of ATP, {ATP} is the ATP concentration, K_1/2,ATP is the half-maximal inhibitory ATP concentration, and H is the Hill coefficient, which was allowed to vary. Dose-response curves for tolbutamide inhibition were described by the product of two Hill components (28), where A represents the high-affinity and B the low-affinity components:

where {Tolb} is the tolbutamide concentration, K_1/2A and K_1/2B are the half-maximal inhibitory tolbutamide concentrations, H_A and H_B are the Hill coefficients, and l is the fraction of current blocked with high affinity.

⁸⁶Rb⁺ efflux experiments were performed as described (28). Transfected cells were preincubated with or without metabolic inhibitors (2.5 μg/ml oligomycin plus 1 mmol/l 2-deoxy-d-glucose). At selected time points, the solution was aspirated and replaced with fresh solution and the aspirated solution was counted in a scintillation counter. The rate constant of ATP-sensitive K⁺–specific (⁸⁶Rb⁺) efflux (k2) was obtained by fitting the data with a single exponential equation:

where the rate constant for nonspecific efflux (k1) was obtained from untransfected cells.

Two approaches were used to estimate P_o,zero, the initial open probability (in zero ATP), of membrane patches containing multiple channels. (PIP₂; method I) PIP₂ was added to the patch until the current reached a saturating level (I_PIP2). This was assumed to represent a maximum P_o,zero of ∼0.97 (29). The fold increase in current was calculated (fold increase = I_PIP2/I_initial) and the P_o,zero was estimated from the following equation:

(NA; method II) Mean P_o,zero was estimated from stationary fluctuation analysis of macroscopic currents (30,31). Mean patch current (I) and variance (α²) in the absence of ATP were obtained by subtraction of the mean current and variance in 5 mmol/l ATP (i.e., all channels closed), respectively. Single-channel current (i) was assumed to be −3.75 pA, corresponding to single-channel conductance of 75 pS. P_o,zero was estimated from the following equation:

Channel density was calculated as I(P_o,max × i) from maximum patch current in zero ATP (I) and assuming a single-channel current (i) = 3.75 pA at −50 mV.

We initially examined insulin secretion in subjects homozygous for the E23K variant, since they might exhibit the greatest difference if E23K genotype affected insulin secretion. The two groups were well matched for age, sex, and BMI (Table 1). Both groups had normal glucose concentrations in the fasting state and following oral and intravenous glucose administration (Table 1) (Fig. 1). The respective AUCs for glucose between E/E and K/K subjects during the OGTT (17.6 ± 0.1 vs. 18.7 ± 0.1 × 10⁴ min · mmol⁻¹ · l⁻¹) and GGI (17.6 ± 0.1 vs. 17.6 ± 0.1 × 10⁴ min · mmol⁻¹ · l⁻¹) were not different (Fig. 1). By contrast, the AUC for insulin in the OGTT (52.6 ± 5.7 vs. 71.0 ± 7.0 × 10³ min · pmol⁻¹ · l⁻¹) and GGI (24.5 ± 2.1 vs. 36.1 ± 4.8 × 10³ min · pmol⁻¹ · l⁻¹) was ∼40% lower (P < 0.05) in the K/K group. The AUC for C-peptide was also reduced (P < 0.05) in the K/K subjects in both the OGTT (30.2 ± 2.7 vs. 58.9 ± 15.1 × 10⁴ min · pmol⁻¹ · l⁻¹) and GGI (26.4 ± 2.1 vs. 36.9 ± 3.3 × 10⁴ min · pmol⁻¹ · l⁻¹). Static, dynamic, and overall β-cell response to glucose were also ∼40% lower (P < 0.05) in the K/K group (Fig. 2,A) based on the OGTT and GGI. In addition, the dose-response curve relating glucose and ISR during the GGI was shifted downward and to the right in the K/K group (Fig. 2 B). Accordingly, the ISR AUC (17.2 ± 1.2 vs. 26.1 ± 0.5 × 10³ pmol/l) and mean ISR (70.6 ± 4.9 vs. 106.7 ± 9.9 pmol · l⁻¹ · min⁻¹) were lower (P < 0.05) in the K/K group. No significant differences were observed in insulin clearance rates in this study (P > 0.05).

To examine the molecular basis for the differences in insulin secretion, we transiently expressed Kir6.2 with residue 23, being either glutamate (E23) or lysine (K23), together with the SUR1 subunit and measured ATP sensitivity in excised membrane patches (Fig. 3,A). Homomeric K23 channels (K/K) exhibit a modest, yet significant, decrease in ATP inhibition compared with homomeric E23 (E/E) channels (K_1/2,ATP = 16 μmol/l [n = 28 patches] and 7.5 μmol/l [n = 24 patches], respectively). A similar relative shift in ATP sensitivities was observed in the presence of physiological (2 mmol/l) Mg²⁺ for E/E and K/K channels (K_1/2,ATP = 19 μmol/l [n = 19 patches] and 26 μmol/l [n = 17 patches], respectively). In contrast, the averaged K_ATP channel density was not different between cells expressing E/E or K/K channels (201 ± 59 and 128 ± 50 channels/patch, respectively; n = 11–19 patches). To recapitulate the heterozygous E23K genotype (E/K), cells were transfected with a 1:1 mixture of E/E and K/K cDNAs. Since four subunits generate the channels, five different ratios of subunits will be present in the resultant channels (1 of 16 channels will be homozygous E/E and homozygous K/K). The ensemble of expressed channels display intermediate ATP sensitivity (K_1/2,ATP = 10.0 μmol/l [n = 10 patches] compared with homomeric K/K and E/E channels) (Fig. 3,B). Our data are similar to those reported by Schwanstecher et al. (5) for recombinant K/K channels. For comparison, the dose-response curves are shown in Fig. 3 B for two Kir6.2 mutations that impair insulin release and underlie NDM. The I182V mutation underlies a transient NDM, whereas the I296L mutation underlies a syndromic form of NDM (32,33). Importantly, the ATP sensitivities of mutant channels correlate with the severity of the disease (K_1/2,ATP = 39 μmol/l for homomeric I182V [n = 17 patches] and 771 μmol/l for homomeric I296L channels [n = 5 patches]).

K_ATP channel activity in metabolically intact cells was screened by ⁸⁶Rb⁺-efflux from transfected COSm6 cells. Efflux from cells transfected with homomeric E/E channels was low under basal conditions and was activated by metabolic inhibition to lower cellular [ATP]/[ADP] (Fig. 4,A). Rb⁺-efflux from cells transfected with recombinant K/K channels was also low under basal conditions and increased with metabolic inhibition. Quantitative estimation of K_ATP channel conductance (see research design and methods) indicates that for K/K channels, basal conductance was a higher fraction of fully activated conductance than E/E channels (Fig. 4 B). In the β-cell, the increased basal flux is expected to impair glucose sensing and account for the association of K/K genotype with reduced insulin secretion.

Kir6.2 mutations can reduce ATP sensitivity by directly reducing ATP binding to the Kir6.2 subunit or indirectly by affecting the intrinsic opening ability (28,29,34). In the latter case, an increase in the open probability (P_o,zero) decreases the frequency with which the channel enters the ATP-accessible closed state, resulting in a decrease in ATP sensitivity. We have modeled this nonlinear relationship between P_o,zero and K_1/2,ATP (29), and this kinetic model describes the diabetes-causing effects of Kir6.2 mutations that underlie NDM (e.g., Q52R, I296L) (Fig. 5,C). To estimate the open probability of K_ATP channels, both nonstationary noise analysis (NA) and phosphatidylinositol biphosphate (PIP₂) application were used to independently examine K_ATP channel gating in isolated membrane patches. As shown in Fig. 5,A, the estimated open probability (P_o,zero) of K/K channels in the absence of ATP (0.68 ± 0.04 [NA method]; 0.67 ± 0.08 [PIP₂ method]) is higher than that of E/E channels (0.49 ± 0.05 [NA method]; 0.46 ± 0.06 [PIP₂ method], P < 0.05), and this increase can fully account for the shifted ATP sensitivity (K_1/2,ATP = 9.4 ± 1.6 μmol/l [E/E] and 21.6 ± 3.3 μmol/l [K/K] from curve fit of individual membrane patches, P < 0.01) (Fig. 5,B). A similar increase in open probability was reported for K/K channels by Schwanstecher et al. (5), using analysis of single-channel records. That multiple methodologies reiterate the same findings strengthens the conclusion that the changes associated with the E23K variant are significant and real. As with more severe NDM, the E23K variant indirectly affects ATP sensitivity by increasing the P_o,zero (Fig. 5 C). The predicted consequence will be reduced excitability of the β-cell, with increasingly severe consequences for insulin secretion (E23K < Q52R < I296L).

Mutations in Kir6.2 that allosterically decrease ATP sensitivity by stabilizing the open state of the channel also reduce high-affinity block by sulfonylureas, a feature of NDM-associated mutations (28). We next examined the effect of the first-generation sulfonyurea, tolbutamide, on channel activity (Fig. 3,C). As shown in Fig. 3 D, homomeric E/E channels exhibit a typical biphasic block by tolbutamide, with both a high-affinity (IC₅₀ = 1.1 μmol/l) and a low-affinity (IC₅₀ = 2 mmol/l) site. The fractional block by the high-affinity therapeutically relevant component is ∼53%. For the K/K channels, the value for the high-affinity block (IC₅₀ = 1.0 μmol/l) is unaltered; however, the fractional block is significantly decreased from 53 to 44%.

The clinical findings described above (of reduced insulin secretion with normal glucose concentrations) suggest a simultaneous change in insulin sensitivity. Hyperinsulinemic-euglycemic clamp experiments revealed that hepatic insulin sensitivity was significantly greater (2.4 ± 0.4 vs. 1.5 ± 0.2 [1,000/μmol · kg fat-free mass⁻¹ · min⁻¹ · pmol⁻¹ · l⁻¹]; P < 0.05) in subjects with the K/K genotype (Fig. 6,A). There was also a strong trend (0.19 ± 0.03 vs. 0.14 ± 0.2 μmol · kg fat-free mass⁻¹ · min⁻¹ · pmol⁻¹ · l⁻¹; P < 0.10) for an increase in peripheral insulin sensitivity, although the differences were not statistically significant. Similar trends were observed for whole-body insulin sensitivity as assessed from the OGTT (Fig. 6 B). Glucose infusion rate, plasma glucose, insulin, glucagon, and tracer-to-tracee ratio are presented in the online appendix (available at http://diabetes.diabetesjournals.org/cgi/content/full/db09-0025/DC1).

The human studies described above were conducted in a small group of E/E or K/K subjects. To confirm the findings on insulin secretion and action in a larger cohort, we carried out 5-h OGTTs in 461 additional subjects (Table 2). Subjects were divided into three groups based on E23K genotype. The three groups had similar glucose concentrations during fasting and 2-h post-glucose challenge. The relative frequencies of E/K (40%) and K/K (13%) in our sample are comparable with those reported (7,8,35), conferring a relative diabetes risk of 1.15–1.65 (3). Consistent with the above findings, the K/K subjects had reduced insulin and C-peptide concentrations, reduced insulin secretory responses, reduced β-cell responsiveness to glucose, and enhanced insulin sensitivity and clearance rates. Interestingly, β-cell responsiveness was also lower in the E/K variant than the E/E.

Type 2 diabetes is a multifactorial disease in which genetic and environmental factors interact to determine the level of predisposition, and, recently, genome-wide association studies have identified common polymorphisms in various genes that are associated with increased diabetes susceptibility (2,36,,–39). These studies often involve the comparison of genetic and phenotypic characteristics of diabetic and control subjects. They have made important contributions to identifying novel genetic loci that determine diabetes risk. However, studies of subjects with diabetes do not provide information on the effects of the respective polymorphisms prior to the onset of glucose intolerance and diabetes. Since hyperglycemia, per se, adversely affects insulin secretion, it is difficult to differentiate between the effects of hyperglycemia and the effects of the polymorphism in studies of established diabetes.

The present study was focused on subjects without a prior history of diabetes who had normal glucose tolerance. Our results demonstrate that the Kir6.2 E23K variant is associated with multiple insulin secretory defects, including a reduction in overall β-cell responsiveness to glucose and a rightward shift in the dose-response curve between glucose and insulin secretion. Overall insulin secretory responses to glucose were reduced by ∼40%, and the extent of the reduction was similar to both oral and intravenous glucose.

We did not anticipate that severe reductions in insulin secretion would be compatible with normal glucose tolerance. The significant increase in insulin sensitivity is presumably the mechanism that makes this possible, and the results of the hyperinsulinemic-euglycemic clamp suggest that the major effects are in the liver. This is the first study to report the coexistence of decreased insulin secretion with increased insulin action in subjects with the E23K polymorphism. Previous studies that evaluated insulin action yielded inconsistent results. One study (11) suggested an increase in insulin sensitivity, while others (8,12,13) found no effect. Moreover, findings from the hyperinsulinemic clamp suggested an increase in peripheral (skeletal muscle) insulin sensitivity, although this did not reach statistical significance (P < 0.10) because of the small sample size. This interpretation is supported by the results obtained in the larger cohort study in which statistically significant increases in insulin sensitivity were observed after oral glucose ingestion. Although BMI was different between groups, measurements of total and regional fat (truncal fat) showed no differences. Nevertheless, because of the differences in BMI, we controlled for BMI using ANCOVA. The differences in insulin sensitivity remained significant (P < 0.001), despite similar and normal glucose tolerance.

The mechanistic basis of this change in insulin sensitivity is not clear. Kir6.2 is expressed in multiple tissues, including skeletal muscle, brain, and heart. The increase in insulin sensitivity could be due to direct effects of the altered K_ATP channel activity in one or more of these tissues, and a link between the K_ATP channel in the hypothalamus and the liver has recently been established (40). An increase in activation of K_ATP channels in the hypothalamus decreases hepatic gluconeogenesis, and this central effect could serve to counterbalance peripheral actions to maintain glucose homeostasis (40,41). It is interesting to note that basal and insulin-stimulated muscle glucose transport is increased in Kir6.2-null mice, which suggests a role for this channel in the regulation of muscle glucose metabolism (10).

Whether the enhanced insulin sensitivity results from direct effects of K_ATP channel activity in muscle or is secondary to effects on insulin secretion is unclear. There is precedent for an increase in insulin sensitivity secondary to reduction in insulin secretion. In the evolution of type 1 diabetes, insulin secretion is reduced before glucose levels rise (42), and increased insulin sensitivity has been demonstrated in normoglycemic carriers of HNF1α mutations with reduced insulin secretion prior to diabetes onset (43). These observations, coupled with those of the present study, suggest that in the evolution of type 2 diabetes, increased insulin sensitivity may compensate for a reduction in insulin secretion, resulting in normal glucose tolerance. Over time, glucose intolerance develops due to progression in the severity of the secretion defect and/or exposure to factors that reduce insulin sensitivity. However, we also cannot completely exclude the likelihood that these E23K variants have not developed diabetes because of their increased insulin sensitivity. In such cases, it might reflect a feature of nondiabetic E23K variants rather than of subjects with E23K variants, per se. Although Kir6.2 levels are high in pancreatic α-cells, we did not find differences in glucagon secretion. An additional new finding is the association of E23K variant with an increase in insulin clearance, as demonstrated in study 2. An increase in insulin clearance may contribute to the reduction in peripheral insulin concentrations that is due largely to the decrease in insulin secretion. The liver is the major site of insulin clearance under physiologic circumstances, and this change in insulin clearance is likely due to a change in hepatic insulin metabolism (44). A link between hepatic insulin receptor binding and action and degradation by the liver has been reported (45).

In heterologous expression studies, reconstituted E23K channels exhibit a mild, yet significant, decrease in ATP sensitivity and a relative increase in basal activity in the intact cell. In the β-cell, decreased ATP inhibition and consequent channel overactivity is predicted to suppress glucose sensing. In contrast to activating mutations in Kir6.2 that underlie NDM (28), the E23K variant has a less radical effect on channel activity and is associated with type 2 diabetes.

The molecular consequence of the E23K variant has been controversial. Given the high K_ATP channel density in β-cells, it is predicted that a change in K_ATP channel activity (<1%) could significantly affect insulin secretion, and, therefore, subtle effects of E23K on channel activity could be physiologically relevant. In support of the conclusion that E23K does not alter channel activity, whole-cell K⁺-currents were similar in Xenpous oocytes expressing E/E or K/K channels (4). A subsequent study (6) reported an increase in ATP sensitivity of K/K channels, relative to E/E, but a subsequent decrease in ATP inhibition upon application of long-chain acyl-CoAs. The most detailed study was carried out by Schwanstecher and colleagues (5,46). Our data parallel their initial findings of an approximately twofold reduction in ATP sensitivity and increase in open probability and reduced sulfonylurea sensitivity (46) associated with the E23K variant. The similarity of our findings, in a completely independent study, utilizing different methodologies to assess open probability, strongly supports the conclusion that the E23K variant decreases ATP inhibition of the K_ATP channel.

Mechanistically, the observed increase in open probability can account for both the reduced insulin secretion and decrease in sulfonylurea inhibition of E23K channels. Whether the E23K variant affects sulfonylurea dosing is unknown, but it is notable that the E23K variant is associated with risk for secondary failure to sulfonylureas in type 2 diabetic patients (47,48), and human islets isolated from E23K donors (E/K and K/K) exhibit a decrease in sulfonylurea-induced insulin secretion (47).

Subjects with the Kir6.2 E23K variant have multiple insulin secretory defects to glucose and decreased responsiveness of the β-cell over a physiologic range of glucose concentrations, and these can be explained by the molecular properties of the E23K channels. These defects in insulin secretion are accompanied by an increase in insulin sensitivity, possibly a compensatory response to reduced insulin secretion. The β-cell secretory defects are present prior to diabetes onset and are likely responsible for the increased risk of type 2 diabetes in subjects with the at-risk E23K genotypes, if superimposed on lifestyle factors causing insulin resistance.

This study was supported by National Institutes of Health (NIH) Grants DK031842 (to K.S.P.) and DK069445 (to C.G.N.), the Washington University Clinical Nutrition Research Unit and Clinical and Translational Science Award (NIH Grants DK56341 and UL1RR0224992, respectively), the Washington University and University of Chicago Diabetes Research and Training Centers (NIH Grants DK20579 and DK20595, respectively), and a gift from the Kovler Family Foundation. No potential conflicts of interest relevant to this article were reported.

We thank the participants for their cooperation and the staff of the intensive research unit of the Institute of Clinical and Translational Sciences for their skilled assistance in the performance of this study. We thank Veronica Paz for her assistance with the genotyping.