Impairment of human ether-à-go-go-related gene (HERG) K+ channel function by hypoglycemia and hyperglycemia. Similar phenotypes but different mechanisms.

Hyperglycemia and hypoglycemia both can cause prolongation of the Q-T interval and ventricular arrhythmias. Here we studied modulation of human ether-à-go-go-related gene (HERG) K(+) channel, the major molecular component of delayed rectifier K(+) current responsible for cardiac repolarization, by glucose in HEK293 cells using whole-cell patch clamp techniques. We found that both hyperglycemia (extracellular glucose concentration [Glu](o) = 10 or 20 mm) and hypoglycemia ([Glu](o) = 2.5, 1, or 0 mm) impaired HERG function by reducing HERG current (I(HERG)) density, as compared with normoglycemia ([Glu](o) = 5 mm). Complete inhibition of glucose metabolism (glycolysis and oxidative phosphorylation) by 2-deoxy-d-glucose mimicked the effects of hypoglycemia, but inhibition of glycolysis or oxidative phosphorylation alone did not cause I(HERG) depression. Depletion of intracellular ATP mimicked the effects of hypoglycemia, and replacement of ATP by GTP or non-hydrolysable ATP failed to prevent the effects. Inhibition of oxidative phosphorylation by NaCN or application of antioxidants vitamin E or superoxide dismutase mimetic (Mn(III) tetrakis(4-benzoic acid) porphyrin chloride) abrogated and incubation with xanthine/xanthine oxidase mimicked the effects of hyperglycemia. Hyperglycemia or xanthine/xanthine oxidase markedly increased intracellular levels of reactive oxygen species, as measured by 5-(and-6)-chloromethyl-2',7'-dichlorodihydrofluorescein diacetate (CM-H(2)DCFDA) fluorescence dye, and this increase was prevented by NaCN, vitamin E, or Mn(III) tetrakis(4-benzoic acid) porphyrin chloride. We conclude that ATP, derived from either glycolysis or oxidative phosphorylation, is critical for normal HERG function; depression of I(HERG) in hypoglycemia results from underproduction of ATP and in hyperglycemia from overproduction of reactive oxygen species. Impairment of HERG function might contribute to Q-T prolongation caused by hypoglycemia and hyperglycemia.

Glucose, the primary end product of the digestion of glycogen, is essential for maintaining life activities in organisms. As a major source of metabolic fuel, degradation of glucose via glycolysis and subsequent oxidative phosphorylation generates high energy phosphates to power the biological processes in the cell. Yet, through an exquisitely complex network of control mechanisms, the rate of glucose metabolism is only as great as needed by the organisms. Moreover, glucose also has other regulatory effects on many cellular functions. Either inadequate or excessive glucose can be harmful to the living system. Therefore, the blood glucose level is dynamically controlled. However, under pathological conditions like diabetes, glucose cannot be efficiently utilized, and the blood glucose level rises. When the blood level of glucose is maintained higher than 7 mM, it is considered as hyperglycemia. Diabetes therapy, on the other hand, can lead to an overly low level of blood glucose, which is referred to as hypoglycemia when the level falls below 3 mM.
Either hypoglycemia or hyperglycemia can have deleterious effects on the cells. One common feature of electrophysiological alterations caused by both hypoglycemia and hyperglycemia in the heart is prolongation of Q-T interval and the associated ventricular arrhythmias that are presumably responsible for sudden cardiac death in diabetic patients (1)(2)(3)(4)(5)(6)(7)(8)(9)(10). However, the ionic mechanisms by which hyperglycemia and hypoglycemia prolong Q-T interval remained unclear, which is at least a part of the reasons why diabetic patients die of mainly cardiac complications.
The human either-à -go-go-related gene (HERG) 1 encodes the rapid component of delayed rectifier K ϩ current in the heart, which is the major repolarizing current in the plateau voltage range of cardiac action potentials. HERG K ϩ channels are susceptible to genetic defects and environmental cues, with the consequence being depression of HERG function in most situations (9). Indeed, most of the cases of long Q-T syndrome are ascribed to dysfunction of HERG channels, particularly that induced by therapeutic drugs (13). It is conceivable that HERG alteration might also be involved in the Q-T prolongation induced by hyperglycemia and hypoglycemia. This thought prompted us to carry out a series of experiments to study the effects of glucose on HERG K ϩ channels and the potential mechanisms.

EXPERIMENTAL PROCEDURES
Cell Culture-HEK293 cells stably expressing HERG (a kind gift from Drs. Zhou and January) (14) were grown in Dulbecco's modified Eagle's medium supplemented with 10% heat-inactivated fetal bovine serum, 200 M G418, 100 units/ml penicillin, and 100 g/ml streptomycin. The cells subcultured to ϳ85% confluency were harvested by trypsinization and stored in the Tyrode's solution containing 0.5% bovine serum albumin at 4°C (12). Electrophysiological recordings were conducted within 10 h of storage.
Pharmacological Probes-D-Glucose (Glu), 2-deoxy-D-glucose (2dG), sodium cyanide (NaCN), pyruvate, ATP, GTP, AMP-PCP (non-hydrolysable analogue of ATP), xanthine (X), xanthine oxidase (XO), and vitamin E (VitE) were all purchased from Sigma. Xanthine was prepared in 2 N NaOH and diluted in the Tyrode's solution 800 times with the pH adjusted to 7.4 with HCl. Xanthine oxidase was added to the xanthine preparation to form the X/XO-reactive oxygen species (ROS) generating system. VitE was dissolved in ethanol and diluted 1000 times to reach the final concentration. Pyruvate in liquid was diluted into the Tyrode's solution, and pH was adjusted to 7.4 with NaOH before use. All other compounds were directly dissolved into the patch clamp recording solutions as specified. Mn(III) tetrakis(4-benzoic acid) porphyrin chloride (MnTBAP) purchased from Calbiochem was dissolved in 1 N NaOH and diluted by 5000 times to reach the desired experimental concentrations. All compounds and reagents were prepared fresh before the experiments.
Intracellular Reactive Oxygen Species (ROS) Measurement-5-(and-6)-chloromethyl-2Ј,7Ј-dichlorodihydrofluorescein diacetate (CM-H 2 DCFDA) (Molecular Probes) is a ROS-sensitive probe that can be used to detect oxidative activity in living cells. It passively diffuses into cells, where its acetate groups are cleaved by intracellular esterases, releasing the corresponding dichlorodihydrofluorescein derivative. Its thiol-reactive chloromethyl group reacts with intracellular glutathione and other thiols. Subsequent oxidation yields a fluorescent adduct that is trapped inside the cell. When it is excited at 480 nm, its emissions at 505-530 nm can be captured. CM-H 2 DCFDA is prepared in dimethyl sulfoxide immediately prior to loading. Glass coverslips were coated with laminin and placed in the wells of a 12-well culture plate before the cells were seeded into the well in a density of 5.0 ϫ 10 4 /well. After overnight incubation, the cells were washed with pre-warmed (37°C) phosphate-buffered saline once and then incubated in the Tyrode's solution containing glucose of varying concentrations or exogenous superoxide-generating system (xanthine/xanthine oxidase) or the reagents as to be otherwise specified, together with the fluorescence dye CM-H 2 DCFDA (10 M). After 30 min of incubation, the coverslips were washed with pre-warmed phosphate-buffered saline twice before being mounted to the glass slides with anti-fading mounting medium and were examined immediately under a laser scanning confocal microscope (Zeiss LSM 510). The percentage of positively stained cells and the fluorescence intensity of staining were determined by densitometric scanning with LSM software (Zeiss).
Data Analysis-Group data are expressed as mean Ϯ S.E. Comparisons among groups were made by analysis of variance (F-test), and Bonferroni-adjusted t tests were used for multiple group comparisons, and paired or unpaired t test was used, as appropriate, for single comparisons. A two-tailed p Ͻ 0.05 was taken to indicate a statistically significant difference. Nonlinear least square curve fitting was performed with CLAMPFIT in pCLAMP 8.0 or GraphPad Prism.

Effects of Glucose on I HERG -
To study the effects of varying concentrations of glucose (Glu, ranging from 0 to 20 mM) on I HERG , our experiments were designed for group comparisons. For each experiment, HERG-expressing HEK293 cells were divided into six groups each superfused with the Tyrode's so-lution containing a given concentration of Glu for 30 min prior to patch clamp recordings. In addition, recordings were performed immediately after formation of whole-cell configuration and adjustments of capacitance and series resistance compensation, and all recordings were made complete within 3 min. In this manner, there was minimal dialysis through the recording pipette and thereby minimal current run-down (time-dependent current decay), and the data best reflect the effects of Glu on I HERG in cells with intact intracellular contents. In addition, such an experimental design also allowed us to study the effect of Glu on I HERG under conditions devoid of influence from exogenous ATP included in the pipette, which is an important issue to be described later. I HERG was elicited by 2.5-s depolarizing steps from Ϫ60 to ϩ40 mV to record the activating current, followed by a repolarizing pulse to Ϫ50 mV for another 2.5 s to observe the deactivating tail current, before being returned to a holding potential of Ϫ80 mV. The results are illustrated in Fig. 1 with both representative raw data and analyzed mean data. Glucose produced two characteristic alterations of HERG channel functions as follows: changes of I HERG amplitude and density and shifts of I-V relationships and activation curves.
Comparison of I HERG recorded at varying extracellular concentrations of Glu ([Glu] o ϭ 0, 1, 2.5, 5, 10, or 20 mM) consistently showed that I HERG density was maximal at a physiological [Glu] o (5 mM), and it was depressed at [Glu] o below or above 5 mM (Fig. 1, A- Several potential mechanisms could explain the observed effects of [Glu] o on I HERG . First, there was a possibility that the effects were a consequence of alterations of extracellular osmolarity with varying [Glu] o . Second, Glu might act directly on HERG proteins to modify the channel function. Finally, Glu metabolism, which generates ATP as well as other metabolic intermediates, also has the potential to modulate I HERG . The following experiments were performed to clarify these issues. To test the first possibility, we performed experiments in which cells were first superfused with a given concentration of Glu (1, 5, or 20 mM) for Ͼ30 min, followed by I HERG recording within 3 min bathing in the Tyrode's solution containing 10 mM Glu. Under such conditions, I HERG demonstrated the same pattern of changes as described above; cells pre-exposed to 1 or 20 mM Glu had markedly smaller I HERG density than those pre-exposed to 5 mM Glu ( Fig. 2A ), respectively. Consistently, the activation curve of I HERG was also shifted to more negative potentials by hypoglycemia (Fig.  2B). A similar difference of I HERG between 5 and 20 mM [Glu] o was also consistently seen when 15 mM cellobiose was added to the Tyrode's solution containing 5 mM Glu (Fig. 2D).
The above data indicate that changes of osmolarity is unlikely the mechanism by which glucose modulates I HERG . Apart from that, the fact that differences of I HERG among the cells pretreated with three different concentrations of Glu (1, 5, and 20 mM) persisted even though I HERG was recorded 10 min after superfusion with the normal Tyrode's solution containing 10 mM [Glu] o suggests that the effects of glucose on I HERG are mediated by some intracellular events, and direct interactions between glucose and HERG channels do not likely play a major role.
Role of Glycolysis and Oxidative Phosphorylation on Glucoseinduced I HERG Enhancement-Glucose, once being taken up into the cell, is metabolized via glycolysis to generate 2 molecules of ATP and 2 molecules of pyruvate (a substrate for oxidative phosphorylation), which is further metabolized via oxidative phosphorylation to produce more ATP molecules. To investigate whether the effects of Glu on I HERG are associated with glucose metabolism, we studied the effects of the glycolysis inhibitor 2-deoxy-D-glucose (2dG). Glucose was replaced by the non-hydrolysable analogue of glucose 2dG to eliminate the glucose metabolism (both glycolysis and subsequent oxidative phosphorylation). Cells were superfused with the Glu-free Tyrode's solution containing 5 mM 2dG for 30 min before patch clamp recordings. I HERG recorded under such a condition was compared with I HERG recorded with the Tyrode's solution containing 5 mM Glu. As shown in Fig. 3, 2dG substitution for Glu reproduced the two characteristic changes of I HERG as observed under hypoglycemia. First, marked depression of I HERG was seen in the presence of 2dG, with I HERG density only ϳ38% that in the presence of 5 mM [Glu] o at 0 mV. Second, similar to the results with 0 mM [Glu] o , the I-V relationship and activation curve were shifted by 10 mV toward hyperpolarizing potentials by 2dG (Ϫ28.2 Ϯ 3.4 mV for 5 mM [Glu] o and Ϫ37.6 Ϯ 4.5 mV for 5 mM 2dG, p Ͻ 0.05, n ϭ 8 for both groups) (Fig. 3B). Because 2dG is a competitive inhibitor of glycolysis, its effects on I HERG in the presence of 5 mM Glu were also investigated. Cells were superfused with the Glu-containing Tyrode's solution with or without 5 mM 2dG for 30 min before patch clamp recordings. The I HERG density was ϳ45% smaller in cells treated with 2dG than in untreated cells (Fig. 3C). For example, the I HERG density at 0 mV was 40.2 Ϯ 4.1 pA/pF for cells treated with 2dG at 5 mM [Glu] o (n ϭ 14) and was 75.6 Ϯ 6.6 pA/pF for cells without 2dG treatment (n ϭ 14, p Ͻ 0.05). These results indicate the importance of glucose metabolism in maintaining the normal HERG function. Intriguingly, when 2dG was added to the hyperglycemic solution containing 20 mM [Glu] o , I HERG was increased as compared with that measured in the hyperglycemic solution without 2dG (Fig. 3E). In other words, 2dG partly reversed the depressed I HERG caused by hyperglycemia toward the normal HERG function seen under normoglycemia.
To dissect further which of the two, glycolysis or oxidative phosphorylation, is truly responsible for HERG modulation, the following experiments were carried out. In the first set of experiments, pyruvate was supplied to the 2dG-containing Glu-free Tyrode's solution. Under such a condition, the glycolysis was inhibited, but the oxidative phosphorylation was maintained. As displayed in Fig. 3A, addition of pyruvate at a concentration of 5 mM restored the depressed I HERG induced by glycolysis inhibition. However, the negative shifts of I-V relationship and activation curve, as seen with hypoglycemia or 2dG substitution for Glu, were still consistently observed with pyruvate ( Fig. 3B). In contrast, elevation of pyruvate to 20 mM weakened the ability to restore the suppressed I HERG caused by glycolysis inhibition with 2dG substitution for Glu. Thus, the I HERG density with 20 mM pyruvate in the 2dG-containing Glu-free Tyrode's solution was considerably smaller than that with normal [Glu] o (5 mM). However, this high concentration of pyruvate still failed to prevent the negative shifts of I HERG I-V relationship and activation curve produced by glycolysis inhibition (Fig. 3B).
In the second set of experiments, the oxidative phosphorylation was inhibited by inclusion of NaCN (2 mM), an uncoupler of oxidative phosphorylation, in the bathing solution, and the glycolysis was kept intact with 5 or 20 mM [Glu] o . Under normoglycemia (5 mM [Glu] o ), inhibition of oxidative phosphoryl-ation by NaCN produced a slight non-significant decrease in I HERG (Fig. 4A). Under hyperglycemia (20 mM [Glu] o ), however, I HERG was markedly diminished, and NaCN restored the depressed I HERG toward the I HERG amplitude seen under normoglycemia (5 mM). For instance, the step and tail I HERG densities in 20 mM glucose were 50.4 Ϯ 5.9 and 56.4 Ϯ 5.3 pA/pF (n ϭ 14) at Ϫ10 mV and were restored by NaCN to 73.9 Ϯ 7.9 and 81.6 Ϯ 6.4 pA/pF (n ϭ 13, p Ͻ 0.05 versus 20 mM [Glu] o ), respectively. Also important is that the oxidative phosphorylation inhibition did not produce any significant voltage shifts of I-V relationships and activation curves, regardless of different [Glu] o values (5 or 20 mM) (Fig. 4B). It appears from the above data that either glycolysis or oxidative phosphorylation was sufficient to sustain the normal function of HERG channels, and significant negative shifts of I HERG I-V relationships and activation curves occurred when glycolysis was inhibited, regardless of whether oxidative phosphorylation was maintained or not.
Role of Intracellular ATP in Maintaining HERG Function-The above experiments indicate that glucose metabolism (gly- colysis and oxidative phosphorylation) is critical for I HERG modulation by Glu. Yet it was unclear whether the I HERG modulation by glucose metabolism is associated with the generation of high energy phosphates (i.e. ATP), and if so whether the ATP generated from the glucose metabolism is glycolysisderived or oxidative phosphorylation-derived. To clarify this issue, we first assessed the influence of intracellular ATP depletion on HERG function in the presence of 5 mM Glu, by omitting ATP from the pipette (internal) solution. I HERG recorded immediately after membrane rupture and capacitance/ resistance compensation was taken as base-line control data, and the same measurement was repeated every 5 min up to 15 min. Under our experimental conditions, 10 min is sufficient to allow complete dialysis, thereby the equilibrium between pipette solution and cytoplasm. As illustrated in Fig. 5, the I HERG recorded with the normal ATP-containing pipette showed only slight run-down over a 15-min period, whereas the I HERG recorded with ATP-free pipette was found significantly reduced with time. There was an ϳ46% decrease in I HERG at 10 min after dialysis at Ϫ10 mV (Fig. 5C), being similar to the reduction of I HERG under hypoglycemia or with the inhibition of glucose metabolism by 2dG. Also consistent with the hypoglycemia and metabolic inhibition was the negative shifts of the I-V relationship (Fig. 5D) and voltage-dependent activation (Fig. 5E) of I HERG with ATP depletion; the V1 ⁄2 was changed by ϳ8 mV from Ϫ31.2 Ϯ 3.6 mV before to Ϫ38.9 Ϯ 8.1 mV (p Ͻ 0.05, n ϭ 7) after [ATP] i depletion.
To investigate whether the requirement of intracellular ATP for HERG function relies on hydrolysis of ATP or is simply because of nucleotide interaction with the nucleotide binding domain of HERG channels (17), we carried out the following series of experiments. We first used the pipette containing the non-hydrolysable AMP-PCP to replace ATP. With 5 mM AMP-PCP in the pipette, the I HERG demonstrated a rapid run-down as observed with the ATP-free internal solution (Fig. 6C). For instance, at Ϫ10 mV the I HERG recorded 10 min after dialysis was 32.2 Ϯ 2.1% smaller than the basal current recorded right after membrane rupture. We then went on to test if substitution of GTP for ATP could prevent I HERG run-down. With 5 mM GTP in the ATP-free pipette solution, the I HERG developed a similar degree of rapid run-down to what was seen with the intracellular ATP depletion alone (Fig. 6F). For example, at Ϫ10 mV, there were 34.4 Ϯ 5.1% decreases in the I HERG amplitude 10 min after dialysis. Moreover, the negative shifts of I-V relationships and activation curves were also seen with AMP-PCP or GTP (Fig. 6, B and E). For instance, V1 ⁄2 was changed from Ϫ33.4 Ϯ 0.8 mV for base line to Ϫ37.3 Ϯ 1.0 mV for 10 min of dialysis with AMP-PCP, and similarly, V1 ⁄2 was shifted from Ϫ31.4 Ϯ 1.5 mV to Ϫ38.9 Ϯ 2.2 mV by GTP (p Ͻ 0.05).
The same effects of ATP depletion on I HERG were consistently reproduced when glyburide (10 M) was included in the superfusate to inhibit ATP-sensitive K ϩ current (K ATP ), if any (data not shown).
Role of ROS on Hyperglycemia-induced I HERG Depression-Collectively from the above experiments, glucose metabolism is necessary for maintaining the HERG channel function, and the ATP produced by either glycolysis or oxidative phosphorylation seems to be a key factor for the regulation; on the other hand, the fact that NaCN restores the depressed I HERG induced by 20 mM [Glu] o or by 20 mM pyruvate suggests that oxidative phosphorylation also produces negative (suppressive) regulation on HERG function. This would imply that the I HERG suppression by high glucose via oxidative phosphorylation is ATP-independent or is the balance between the enhancement by ATP and the suppression by other factors associated with oxidative phosphorylation. It has been well established that mitochondria produce most of the endogenous reactive oxygen species (ROS) through oxidative phosphorylation (18 -23), and hyperglycemia stimulates massive ROS production (19, 24 -29). It is therefore rational to propose that the endogenously produced ROS via oxidative phosphorylation stimulated by hyperglycemia could impair HERG channel function to suppress I HERG .
To test this hypothesis, we performed the following experi-  Fig. 7, A and B, pretreatment of cell with VitE effectively prevented the I HERG suppression by hyperglycemia; the I HERG density in VitE group was virtually identical to that in the normoglycemia group. There was no significant shift of the activation curve along the voltage axis. These data suggest a participation of ROS in the HERG regulation by hyperglycemia. Next, we studied the effects of another antioxidant, superoxide dismutase (SOD) mimetic MnTBAP, on the I HERG depression induced by 20 mM [Glu] o . Because the compound does not readily penetrate the cells, it was intracellularly applied through dialysis of the pipette solution at a concentration of 5 M. Cells were superfused with 20 mM Glu for 30 min prior to formation of the whole-cell membrane patch. To correct for the potential current run-down, the I HERG recorded at various time points after membrane rupture was normalized to the I HERG recorded with the normal Tyrode's solution without MnTBAP at the corresponding time points. As shown in Fig. 7C, MnTBAP caused a time-dependent increase in the I HERG amplitude, indicating a restoration of the depressed HERG function. By comparison, no alterations of I HERG , or virtually a slight decrease (presumably representing run-down of the current), were found with catalase in the pipette (data not shown). These results indicate that the ROS involved in the I HERG suppression by hyperglycemia was of mainly superoxide anion (O 2 Ϫ ). To obtain further evidence for this notion, we assess the effects of exogenously produced O 2 Ϫ by the ROS-generating system xanthine/xanthine oxidase (X/XO) on I HERG . Cells were incubated with or without X/XO (500 M/5 milliunits/ml) in Tyrode's solution containing 5 mM [Glu] o for 40 min before I HERG was recorded under 5 mM [Glu] o . The I HERG density was consistently smaller in X/XO-treated cells than in X/XO nontreated cells (Fig. 8, A and B).
To confirm that ROS production was indeed increased by 20 mM [Glu] o and the hyperglycemia-induced ROS was mainly of O 2 Ϫ , we proceeded to measure the intracellular ROS levels using CM-H 2 DCFDA fluorescence dye. The ROS level was measured in cells preincubated with the Tyrode's solution containing 5 or 20 mM glucose for 30 min. The staining of the cells MnTBAP was applied intracellularly through the pipette. I HERG recorded immediately after whole-cell formation and series resistance compensation was taken as base-line control data (Ctl) and that recorded 10 min after dialysis was used for analysis to reflect the effects of MnTBAP. To correct for potential run-down of the current, the I HERG recorded with MnTBAP was normalized to that recorded with the normal internal solution. *, p Ͻ 0.05 versus control, paired t tests.

FIG. 6. Effects of non-hydrolysable ATP (AMP-PCP) and GTP on I HERG .
A and D, I-V relationships. I HERG was recorded right after membrane rupture and 10 min after dialysis with the ATP-free pipette solution containing AMP-PCP (n ϭ 10) (A) or GTP (n ϭ 7) (D). B and E, activation conductance (G) curves before and after AMP-PCP or GTP in the ATPfree internal solution. The numbers in the legends represent AMP-PCP or GTP concentrations in mM. C and F, time-dependent changes of I HERG at Ϫ10 mV recorded with the pipette containing ATP (control (Ctl)) or AMP-PCP (C) or GTP (F). *, p Ͻ 0.05 versus control, paired t tests; ؉ , p Ͻ 0.05, F test indicating the significance of time dependence. demonstrated two distinct patterns as follows: one localized to the defined rod-shaped structures, and the other one diffused evenly throughout the cytoplasm. The former presumably represents the physiological production of ROS as a by-product of oxidative phosphorylation in mitochondria, and the latter indicates overproduction of ROS as a result of metabolic stress and damage to mitochondria. The cells with diffused staining and with fluorescence intensity Ն5 times the background were defined as positive staining, and the number of cells with positive staining was pooled from 5 fields. The intensity of staining by the fluorescent probe for ROS was analyzed by densitometric scanning using the LSM program, and cells with either localized or diffused staining were taken for analysis, and the data were normalized to the control (5 mM [Glu] o ) values. Under normoglycemia, a majority of cells that was stained by CM-H 2 DCFDA demonstrated the localized pattern, and the diffused staining was sparse. Yet in the cells treated with 20 mM Glu, the number of the cells with positive staining as well as the intensity of staining was consistently higher, as compared with the cells treated with 5 mM Glu (Fig. 9). This high level of ROS production was markedly suppressed in the cells pretreated with NaCN (2 mM, Fig. 9), an uncoupler of oxidative phosphorylation, indicating that the mitochondrion is most likely where the ROS was massively produced. Because we have demonstrated that pyruvate at high concentrations decreased I HERG (Fig. 3A), the ROS level in the cells pretreated with 20 mM pyruvate was also measured. As shown in Fig. 9, like 20 mM Glu, 20 mM pyruvate also significantly increased the ROS production although to a less extent. Moreover, the glycolysis inhibitor 2dG (5 mM) also reduced the ROS level in high glucose (20 mM), which is indicated by fewer positively stained cells and lower intensity of staining (Fig. 9). The results explain why 2dG partly restored the depressed I HERG in 20 mM Glu (Fig. 3E).
We have shown that VitE prevented, and MnTBAP partly reversed, the I HERG depression in hyperglycemia (Fig. 7). To see whether this is indeed attributable to their antioxidant actions, effects of VitE and MnTBAP on hyperglycemia-induced ROS production were also studied. As shown in Fig. 10, A  It has been well documented that the X/XO ROS-generating system stimulates mainly the generation of O 2 Ϫ . To test whether this is also true in our conditions, the ROS that was exogenously generated by X/XO and penetrated cells was measured, and the effect of MnTBAP was studied at 5 mM [Glu] o . As displayed in Fig. 11, the ROS level was significantly higher in the cells treated with X/XO alone, and this increase in ROS level was prevented in the cells pretreated with MnTBAP. DISCUSSION The work described here documents a previously unreported role of glucose in regulating the function of HERG K ϩ channels. Our data revealed that glucose produces two characteristic effects on the HERG channel function: changes of HERG current (I HERG ) amplitude/density and activation voltage. and ATP and ROS are crucial in defining the HERG function with changing extracellular glucose levels.
Depression of HERG Function in Hypoglycemia Likely Results from Underproduction of ATP-In our study, lower glucose levels and inhibition of glucose metabolism both produced similar suppression of HERG function as reflected by the substantial diminishment of HERG current (I HERG ), pointing to a requirement of glucose metabolism for HERG modulation by glucose. Our data allowed us to reach the following conclusions.
Either Glycolysis-or Oxidative Phosphorylation-derived ATP Is Sufficient for Maintaining the Normal HERG Function-The end point of glucose metabolism is the generation of high energy phosphates for maintaining cellular functions, and depletion of ATP could impair cellular processes dependent on high energy phosphates. One of the major findings of this study is that the normal HERG function critically relies on the level of intracellular ATP; depletion of intracellular ATP impairs HERG function to an extent similar to what severe hypoglycemia (0 mM [Glu] o ) does (see Figs. 1, 2, and 5). Complete inhibition of glucose metabolism by 2dG substitution for glucose reproduces the effects of hypoglycemia or ATP depletion on I HERG . Yet, neither inhibition of glycolysis alone by 2dG substitution of Glu with a supply of pyruvate to sustain oxidative phosphorylation nor inhibition of oxidative phosphorylation alone by NaCN in the presence of 5 mM Glu to maintain glycolysis is able to cause depression of HERG function. The results imply that the ATP generated by glucose metabolism plays an important role in maintaining HERG function, and either the glycolytic or oxidative ATP is adequate for the regulation.
ATP synthesis and utilization are subcellularly compartmentalized; glycolysis-derived ATP primarily regulates membrane proteins because the glycolytic pathway is associated with sarcolemma (31), whereas oxidative phosphorylation-derived ATP preferentially supports cytosolic processes because oxidative ATP is generated within the mitochondria and subsequently transported to the cytoplasm (32). Regulation of cardiac ATPsensitive K ϩ channel (K ATP ) (34) and L-type Ca 2ϩ channel (I Ca ) (33) by intracellular ATP has been well documented by some previous studies. It was found that both K ATP and I Ca were preferentially regulated by glycolytic ATP (33,34).
Glycolysis-derived ATP May Be Responsible for Maintaining the Normal Voltage-dependent Activation of I HERG -Although as mentioned above, either glycolytic or oxidative ATP is sufficient for maintaining the normal HERG current amplitude/ density, only glycolysis-derived ATP seems to affect the steadystate voltage-dependent activation property of HERG channels. This notion is supported by several lines of evidence from our experiments. 1) Hypoglycemia (0 or 1 mM [Glu] o ), a situation with inadequate ATP production but not hyperglycemia (10 or 20 mM [Glu] o ), causes negative shifts of I-V relationships and voltage-dependent activation curves. 2) Inhibition of glycolysis (Fig. 3), but not oxidative phosphorylation (Fig. 4), abolishes the negative shifts of the HERG activation. 3) When glycolysis is inhibited by 2dG, preservation of oxidative phosphorylation by addition of pyruvate fails to prevent the negative shift caused by hypoglycemia (see Fig. 3). 4) Depletion of intracellular ATP reproduces negative shifts similar to those seen with hypoglycemia. Similar dependence of glycolytic ATP regulation of K ATP and I Ca has been documented (33)(34).
Role of ATP in Maintaining HERG Function Is Most Likely Due to the Phosphorylation-dependent Mechanisms-Two alternative mechanisms could account for intracellular ATP regulation of ion channels: ATP acts as a substrate for phosphorylation of channel proteins by protein kinase which requires ATP hydrolysis, and ATP interacts with the nucleotide binding domains of channel proteins to produce allosteric regulation not requiring ATP hydrolysis. The latter mechanism has been shown to operate for K ATP and I Ca regulation (33,35). In our case, neither the non-hydrolysable analogue of ATP AMP-PCP nor GTP prevented the I HERG run-down caused by ATP depletion (Fig. 6); instead substitution of AMP-PCP or GTP for ATP in the internal solution produced nearly identical effects as seen with ATP depletion alone. The results suggest that the HERG regulation by ATP under our experimental conditions is phosphorylation-dependent requiring ATP hydrolysis. In other words, ATP serves as a substrate for phosphorylation of HERG channels by protein kinases. Indeed, we have recently found that the normal HERG function requires basal activity of protein kinase B and inhibition of protein kinase B markedly suppresses I HERG and shifts HERG activation along the voltage axis toward more negative potentials (36). These results are in good agreement with the HERG regulation by ATP. Studies are currently undertaken to clarify the link between ATP and protein kinase B modulation of HERG channels.
Depression of HERG Function in Hyperglycemia Results from Overproduction of ROS-The consequence of the physiological role in oxidative phosphorylation is the generation of ROS as by-products of the consumption of molecular oxygen in the electron transport chain (23). Physiologically, these ROS are mostly trapped within mitochondria and rapidly scavenged by endogenous antioxidants like SOD, catalase, glutathione, etc. Yet under metabolic stress, ROS can be overproduced and can cause damages to mitochondria. Consequently, the ROS may diffuse throughout the cytoplasm and cause further deleterious effects on other cellular processes. Abnormally high concentrations of glucose can enhance ROS damage at least in three different ways. First, it has been known that high glucose (25 mM) evoked ROS generation, which was blocked by antioxidants, inhibitors of mitochondrial electron transport chain complex, inhibitors of glycolysis-derived pyruvate transport into mitochondria, uncouplers of oxidative phosphorylation, SOD mimetics, catalase, etc. (19). Superoxide anion (O 2 Ϫ ) is found to be the major ROS produced under hyperglycemia (37)(38)(39)(40)(41)(42), and increases in ROS can be prevented by SOD. Second, glucose itself can auto-oxidize to form ROS including O 2 Ϫ , OH Ϫ , and H 2 O 2 (43). Finally, acute elevations in glucose also depress natural antioxidant defenses. It has been found that incubation of purified bovine CuZn-SOD with 10 to 100 mM glucose reduces the enzyme activity by 60% (44).
Elevated glucose or pyruvate level is expected to enhance oxidative phosphorylation and produce more ATP molecules to support HERG function or increase I HERG . However, our observations are contrary to this expectation. Our results showed that hyperglycemia or excessive pyruvate markedly depressed the HERG function. A reasonable explanation for this is that the ROS produced under hyperglycemia counteract the effects of ATP, and the net outcome is a balance between enhancing effects of ATP and suppressing effects of ROS. Evidently, under our experimental conditions, the effects of increased ROS overwrite the effects of increased ATP, resulting in suppression of I HERG . This notion is supported by the following evidence. First, the depression of I HERG induced by hyperglycemia was prevented or reversed by the antioxidants vitamin E and MnTBAP (SOD mimetic). Second, inhibition of the glycolysis and thereby the subsequent oxidative phosphorylation by 2dG partially reversed the depressed I HERG under hyperglycemia (Fig. 3). Weakened oxidative phosphorylation due to inhibition of the glycolysis would reduce both ATP and ROS productions, but the net result was an increase in I HERG , indicating again that in our experimental conditions ROS overweighs ATP in terms of their effects on I HERG . This is in agreement with the notion that the suppressing effects of ROS overproduction overwhelm the effects of ATP increase. Moreover, 2dG also can compete with glucose for access to glucose transporters and thus decreases glucose uptake which in turn can result in reduction of ROS production in the cells. Finally, our data indeed demonstrated the ability of high glucose to stimulate an overproduction of ROS (see Fig. 9). The fact that a high concentration of pyruvate mimicked, whereas VitE or NaCN abrogated, the ROS overproduction suggests that the ROS were mainly produced via the oxidative phosphorylation in mitochondria in our cells.
It has been reported that the ROS, which generate highly reactive hydroxyl group (OH Ϫ ), such as H 2 O 2 or FeSO 4 /ascorbic acid (an oxidative stimulus analogous to H 2 O 2 ), increased I HERG at negative potentials by shifting the HERG activation to more negative voltages (45)(46). These results are opposite to our observations. One explanation is that the ROS generated under our experimental conditions may be different from the OH Ϫ -generating system (H 2 O 2 or FeSO 4 /ascorbic acid). As already mentioned, previous studies have confirmed that the ROS induced by hyperglycemia is mainly of O 2 Ϫ . Here, we also showed that the SOD mimetic MnTBAP reduced the hyperglycemia-induced ROS overproduction (Fig. 10), and the O 2 Ϫ -generating system X/XO produced ROS which were also abolished by MnTBAP, evidence for O 2 Ϫ as a major ROS generated in our cells. Consistently, depressive effects of hyperglycemia or X/XO on I HERG were significantly weakened by MnTBAP. Indeed, it has been reported that the O 2 Ϫ generated by high glucose (23 mM) or by X/XO in rat small coronary arteries impairs voltagegated K ϩ (K v ) current (39,47); reducing the current density by around 60%, which was partially restored by SOD and catalase. All together, we believe that different ROS might have different effects on I HERG ; OH Ϫ enhances, whereas O 2 Ϫ depresses, I HERG .
Moreover, it has been shown that excessive ROS inhibits glycolysis and the subsequent glycolytic ATP production and even depletes intracellular ATP levels in isolated perfused hearts (48 -50). This fact together with our data suggests that besides the potential direct modulation of I HERG by ROS, ATP reduction potentially caused by ROS may also contribute to the I HERG depression under hyperglycemia. This provides an alternative explanation for the depressive effects of the ROS overproduction overcoming the enhancing effects of the expected ATP increase.
In our study, the effects of pyruvate and X/XO were smaller than those of hyperglycemia. This may be because the O 2 Ϫ generated by hyperglycemia occurs inside the cell but pyruvate does not readily penetrate cells, and the effect of pyruvate observed in the present study may underestimate the true role of oxidative phosphorylation in I HERG modulation. Likewise, the O 2 Ϫ generated by X/XO was primarily extracellular with subsequent entry into the cell which could also underestimate the effects of O 2 Ϫ . It was shown that within minutes of exposure to dihydroxyfumaric acid or xanthine plus xanthine oxidase, both of which produce the superoxide anion, action potential duration was prolonged in canine myocytes, and this effect was followed by the appearance of early after depolarization (51). X/XO caused a 30% increase in action potential duration in superfused papillary muscle or small strips of right ventricular walls of guinea pig hearts (52). However, whether the action potential duration prolongation was associated with inhibition of delayed rectifier K ϩ current (I Kr ) is unknown. Our study provides a potential explanation for these observations. Impairment of HERG Function Might Contribute to Q-T Prolongation Caused by Hypoglycemia and Hyperglycemia-Heart disease is a leading cause of death in diabetic patients. In patients with diabetes a prolongation of the Q-T interval has been associated with an increased risk of sudden cardiac death (2) due to the occurrence of lethal ventricular arrhythmias, particularly Torsade de pointes following bradycardia (1). Several cardiovascular pathological consequences of diabetes such as hypertension and arteriosclerosis affect the heart to varying degrees. Hyperglycemia, as a consequence of diabetes and an independent risk factor, also can directly cause cardiac dam-age. On the other hand, insulin therapy increases the risk of hypoglycemia in type 2 diabetic patients; according to the United Kingdom Prospective Diabetes Study, approximately one-third of the insulin-treated patients reported one or more hypoglycemic episodes per year during the first 3 years (3). Hypoglycemia is presumed to be the cause of death in about 3% of insulin-treated diabetic patients (11). Intriguingly, it is well recognized that both hyperglycemia and hypoglycemia can cause prolongation of Q-T interval. In type 2 diabetes, the prevalence of Q-T prolongation is as high as 26%, and Q-T prolongation during experimentally induced and spontaneously occurring hypoglycemia or diabetic hyperglycemia has also been shown to occur in healthy subjects and in diabetic patients with increased risk of malignant ventricular arrhythmias (4 -11). Yet the potential ionic mechanisms by which hypoglycemia and hyperglycemia cause Q-T prolongation remained poorly understood. Studies on glucose modulation of cardiac ion channels are sparse and have been mostly limited to ATP-sensitive K ϩ current (K ATP ). On the contrary to I HERG , K ATP is closed with increased intracellular ATP levels (34). One study reported by Xu et al. (53) demonstrated that hyperglycemia (18 mM [Glu] o ) decreased the density of transient outward K ϩ current but did not alter the inward rectifier K ϩ current in rat ventricular myocytes that do not express delayed rectifier K ϩ current (I Kr ), the physiological counterpart of I HERG . In arterial smooth muscles, high glucose diminished shaker-type delayed rectifier K ϩ current (39). In rat myelinated nerve fibers, 30 mM glucose increased Ca 2ϩ -activated K ϩ current (30). Whereas none of the data from these studies could fully account for the Q-T prolongation, particularly the Q-T prolongation induced by hypoglycemia, our study provides a plausible, or at least an alternative, explanation; HERG K ϩ channel may be a mechanistic link for the Q-T prolongation induced by both hyperglycemia and hypoglycemia. Yet one should keep it in mind that HERG may be only one of the multiple factors contributing to the Q-T prolongation in hypoglycemia and hyperglycemia.