Functional Coupling between the Kv1.1 Channel and Aldoketoreductase Kvβ1*♦

The Shaker family voltage-dependent potassium channels (Kv1) assemble with cytosolic β-subunits (Kvβ) to form a stable complex. All Kvβ subunits have a conserved core domain, which in one of them (Kvβ2) is an aldoketoreductase that utilizes NADPH as a cofactor. In addition to this core, Kvβ1 has an N terminus that closes the channel by the N-type inactivation mechanism. Point mutations in the putative catalytic site of Kvβ1 alter the on-rate of inactivation. Whether the core of Kvβ1 functions as an enzyme and whether its enzymatic activity affects N-type inactivation had not been explored. Here, we show that Kvβ1 is a functional aldoketoreductase and that oxidation of the Kvβ1-bound cofactor, either enzymatically by a substrate or non-enzymatically by hydrogen peroxide or NADP+, induces a large increase in open channel current. The modulation is not affected by deletion of the distal C terminus of the channel, which has been suggested in structural studies to interact with Kvβ. The rate of increase in current, which reflects NADPH oxidation, is ∼2-fold faster at 0-mV membrane potential than at -100 mV. Thus, cofactor oxidation by Kvβ1 is regulated by membrane potential, presumably via voltage-dependent structural changes in Kv1.1 channels.

min at 4°C. The supernatant was then loaded onto a Talon Co 2ϩ affinity column (Clontech). Nonspecifically bound protein was washed away with Buffer A supplemented with 20 mM imidazole, and Kv␤1 protein was eluted with Buffer A plus 300 mM imidazole. The 6-histidine tag was removed by incubation with thrombin (Roche Diagnostics) at a Kv␤1/enzyme ratio of 2 mg to 1 unit overnight at 4°C. The protein was then loaded onto a Superdex 200 column (GE Healthcare) for final purification. The column was equilibrated with Buffer B (150 mM KCl and 20 mM Tris (pH 8.0)). The column volume was ϳ25 ml, ϳ7.8 ml of which was the void volume. The Kv␤1 core elutes as a single peak at ϳ11.3 ml (R F Ϸ 0.8). Protein concentration was determined using the BCA kit (Pierce). Compared with the Kv␤2 core, the Kv␤1 core is less stable and tends to aggregate. A large fraction of the overexpressed Kv␤1 core protein was not in the supernatant after the ultracentrifugation (see Fig. 1A, inset, compare lanes 2 and 3).
Single-turnover Enzymatic Reaction Measurement-Singleturnover hydride transfer reaction rates were measured on a FluoroMax-3 spectrofluorometer (HORIBA Jobin Yvon Inc.) at room temperature (20 -24°C). The reaction mixture contained 2 M Kv␤1 protein and different concentrations of substrates in a final volume of 150 l. 4-Cyanobenzaldehyde (4-CY) was first prepared in ethanol as a 0.5 M stock solution and then diluted in Buffer B to the desired concentrations. The excitation wavelength was set at 360 nm with a 1-nm band-pass slit size, and the emission was measured at 454 nm with a 5-nm band-pass slit size at various time points after the reaction was initiated with the addition of 4-CY. The K152M mutant Kv␤1 protein has much lower NADPH fluorescence compared with the wild-type protein, and therefore, an equal molar concentration (2 M) of free NADPH was supplied in the single-turnover experiment.
Channel Expression and Electrophysiology-Full-length cDNA of rat Kv1.1 (NCBI accession number NM_173095) or that of rat Kv␤1 was cloned into a modified pBluescript vector (a gift from Dr. Mark Sonders, Columbia University) between the KpnI and EcoRI sites for in vitro transcription. Kv1.1-inact was generated by splicing the DNA sequence encoding residues 1-70 of Kv␤1 into that of Kv1.1 (encoding residues 2 to 495stop) by the overlapping PCR method, and the PCR product was then inserted into the same modified pBluescript vector. Kv1.1⌬C was generated by amplifying the DNA sequence encoding residues 1 to 435-stop using primers CCGGTAC-CATGTATCCGGAATCAACC and CCGAATTCTCATA-AGTTAGGAGAACTAAC, and the PCR product was then inserted into the same modified pBluescript vector. Several shorter constructs were also tested but failed to generate sufficiently high expression levels for patch clamp studies.
Point mutations were made using the QuikChange kit (Stratagene). The sequences of all constructs were verified by DNA sequencing through the entire coding region. mRNA was prepared by in vitro T7 polymerase transcription after DNA was linearized with NotI. mRNAs were purified using TRIzol reagent (Invitrogen) and injected into Xenopus oocytes for channel expression. For coexpression, mRNAs of Kv1.1 and Kv␤1 were mixed and injected together into oocytes. Different ratios of the two were tested to achieve a sufficiently high level of Kv␤1 expression so that most of the channels assembled with Kv␤1, as indicated by the almost complete inactivation of channel current. Wild-type Kv␤1 has a cysteine residue at position 7 that can be oxidized on inside-out patches to affect channel inactivation, and this can be prevented by the addition of reducing reagents such as dithiothreitol (5). To eliminate concerns that the change in inactivation could be due to cysteine oxidation, we mutated the cysteine to alanine and used Kv␤1(C7A) as the "wild type" throughout this work. In separate experiments, 4-CY was found to induce similar changes in Kv1.1 coexpressed with wild-type Kv␤1 in the presence of 5 mM dithiothreitol (data not shown).
Patch clamp currents were recorded on inside-out patches pulled from oocytes 3-5 days after injection. Electrodes were drawn from patch glass (G85150T-4, Warner Instruments) and polished (MP-803, Narishige Co.) to a resistance of 0.6 -1 megaohms. The pipette solution contained 130 mM KCl, 2 mM MgCl 2 , and 10 mM KH 2 PO4 (pH 7.4). The bath solution contained 80 mM KCl, 5 mM EGTA, and 50 mM KH 2 PO4 (pH 7.4). The pH was adjusted with KOH. K ϩ currents were elicited by holding the patch at Ϫ100 mV for at least 30 s and stepping to ϩ60 mV for 200 ms. The analog signals were amplified by an Axon 200B patch clamp amplifier (Molecular Devices Inc.), filtered at 1 kHz using the built-in Bessel filter, digitized at 100 s by Digidata 1322a (Molecular Devices Inc.), and recorded to a computer hard disk. The volume of the recording chamber (Warner Instruments) was ϳ200 l, and complete exchange of solution was achieved by perfusing at least 1 ml of solution by gravity flow at a flow rate of 2-3 ml/min. Data Analysis-The Kv␤1 subunit confers fast inactivation to otherwise non-inactivating Kv1.1 channels by a "ball-andchain" mechanism (5). Inactivation time constants were obtained by fitting the current decay with exponential functions in Clampfit software (Molecular Devices Inc.). The current decay was best fit with a two-component exponential function: a predominant and fast component and a minor and slower component (see Table 1). The inverse of the fast time constant was defined as the on-rate of the N-type inactivation and is plotted in Fig. 6B. The slower component was likely contributed by the C-type inactivation, and its effect was minimized in the high K ϩ recording solution. Recovery from inactivation was measured with a paired-pulse protocol (21). Briefly, from a holding potential of Ϫ100 mV, inactivation was induced by a first pulse with a depolarization to ϩ60 mV for 200 ms; the patch was then repolarized back to the holding potential; and a second pulse to ϩ60 mV was applied after a delay of 0.01-1 s. The inactivating portion of the current, i.e. peak current subtracted by steady-state current (at the end of the 200-ms pulse), was normalized and is plotted versus the delay time in Fig. 6C. Recovery from inactivation was well fit by a single-component exponential function.
Voltage-dependent channel activation was measured using an instantaneous tail current protocol. Briefly, a patch was held at Ϫ100 mV and stepped to different voltages between Ϫ90 and ϩ60 mV for 200 ms and then stepped back to Ϫ100 mV. When the voltage was reversed to Ϫ100 mV, the high extracellular potassium concentration (ϳ150 mM) produced a pronounced "tail current," shown in Fig. 5C. The peak of the current is a measure of the fraction of channels activated by the preceding 200-ms pulse. The normalized peaks of the tail currents were plotted versus membrane potential and fit with the following Boltzmann function in Clampfit (Equation 1), where R is the universal gas constant; T is the absolute temperature; F is the Faraday constant; and V min and V max are the minimum and maximum voltages used for channel activation, respectively. The two parameters obtained from curve fitting are z␦ and V1 ⁄ 2 , where z␦ is the nominal amount of charges that move across the membrane voltage field, and V1 ⁄ 2 is the voltage at which the relative P open is 0.5. The rate of channel modulation at Ϫ100 mV was measured by first perfusing a patch with either 4-CY or NADP ϩ , holding the patch at Ϫ100 mV, and depolarizing the patch to ϩ60 mV for 200 ms every 60 s. The normalized current amplitudes were then plotted versus the accumulative exposure time and fit with a single-component exponential function. The inverse of the time constant was defined as the rate of modulations.
Because significant channel inactivation occurs at 0 mV, the rate of channel modulation at 0 mV was measured with a slightly different protocol to allow channels to recover from inactivation at Ϫ100 mV. After a stable inside-out patch was obtained, the patch was held at 0 mV, and 4-CY or NADP ϩ was perfused for 2 min. At the end of the 2-min perfusion, 4-CY or NADP ϩ was washed away. The holding potential was then stepped to Ϫ100 mV for 30 s to allow complete channel recovery from inactivation, and current was elicited by stepping the membrane potential to ϩ60 mV. This process was then repeated until the modulation reached steady state. Similar to the measurement at Ϫ100 mV, the normalized current amplitude versus the accumulative exposure time was well fit with a single-component exponential function.
Data Statistics-The Origin 7.5 software package was used for statistical analysis of the data. The results are expressed as means Ϯ S.E. Student's t tests (both paired and unpaired) and one-way analysis of variance (ANOVA) were used to assess changes in a mean value.
Chemical Reagents-Chemical reagents were purchased from Sigma, unless indicated otherwise. 4-CY was first dissolved in ethanol and then diluted to the desired final concentration. NADP ϩ was purchased as a sodium salt, and because sodium ion blocks potassium channels from the intracellular side, we exchanged the sodium with potassium ion using a size exclusion column. Isopropyl ␤-D-thiogalactopyranoside, dithiothreitol, and kanamycin were purchased from LabScientific, Inc.

RESULTS
Kv␤1 Is a Functional AKR-To test the enzymatic activity of Kv␤1, its conserved core (residues 71-401) was expressed and purified to homogeneity (Fig. 1A). The purified protein elutes as a single peak on a Superdex 200 size exclusion column at ϳ11.3 ml (Fig. 1A), identical to Kv␤2, which is known to form a tetramer (18). Also similar to Kv␤2, the purified Kv␤1 protein has a prominent fluorescence emission peak at 454 nm when excited at 360 nm ( Fig. 1B), indicative of a co-purified NADPH cofactor. These properties suggest that the purified Kv␤1 core has the correct structural fold and likely has AKR activities. Because NADP ϩ does not have fluorescence emission, an enzymatic redox reaction, i.e. transfer of a hydride from the Kv␤bound NADPH to a substrate, can be monitored by a reduction in the 454 nm fluorescence.
A known Kv␤2 substrate (4-CY, 5 mM) was then mixed with purified Kv␤1, and the fluorescence intensity at 454 nm was monitored over time. The NADPH fluorescence decreased over time following a single-component exponential function with a time constant of 3.6 Ϯ 0.3 min (n ϭ 3) (Fig. 1, B and C). The single-component exponential kinetics is consistent with the single-step hydride transfer reaction. The inverse of the exponential time constant was defined as the hydride transfer rate, and the rate was measured at different 4-CY concentrations and is plotted in Fig. 1D, along with the same measurement obtained with Kv␤2 (19). Compared with Kv␤2, Kv␤1 has a slightly higher enzymatic rate.
Lysine 152 at the catalytic site was mutated to methionine because the same mutation to the conserved lysine at the equivalent position in Kv␤2 and 7␣-hydroxysteroid dehydrogenase showed significantly reduced enzymatic activity (19,22). Kv␤1(K152M) was expressed, purified, and mixed with 5 mM 4-CY, and the NADPH fluorescence remained essentially the same after 20 min (Fig. 1C). We conclude that Kv␤1 is a functional AKR and that our fluorescence assay measured its activity.
Kv␤1 Substrate Increases Kv1.1 Current-To examine whether the AKR activity of Kv␤1 modulates channel function, Kv1.1 was coexpressed with Kv␤1 in Xenopus oocytes, and channel activity was recorded in inside-out patches. Kv␤1 with cysteine 7 mutated to alanine was used throughout this study as the wild type to eliminate a known cysteine oxidation effect at this position (see "Experimental Procedures" and Ref. 5). The presence of fast inactivation indicates that Kv1.1 and Kv␤1 are co-assembled because Kv1.1 expressed alone has non-inactivating current (Fig. 2D). 4-CY was perfused to the intracellular side of the channel until the current reached steady state and was then replaced with normal inside solution. After the 4-CY perfusion, the on-rate of channel inactivation at ϩ60 mV slowed 3.4-fold from 316 Ϯ 12 to 93 Ϯ 4 s Ϫ1 (n ϭ 30), and as a result, potassium current increased significantly ( Fig. 2A and Table 1). The 4-CY effect was highly reproducible and was observed in all 30 inside-out patches from seven batches of oocytes. To quantify current increase, we defined the parameter ⌬I ss , i.e. increase in steady-state current expressed as a fraction of the initial inactivating current (see "Experimental Procedures"), and the ⌬I ss is 50 Ϯ 4% (n ϭ 30) (Fig. 2G). In a vehicle control of 1% ethanol, a much smaller current increase was observed with a ⌬I ss of 7 Ϯ 0.3% (n ϭ 7) (Fig. 2, C and G). Therefore, 1% ethanol has a small potentiation effect on open channel current, but the large changes in channel inactivation and current level are due mostly to the presence of 4-CY (p Ͻ 0.001 in one-way ANOVA test; see Table 1). Channel modulation reached steady state in ϳ16 min and was only slightly reversed in the presence of 2 mM dithiothreitol (Fig. 2B), indicating that the modulation is not the result of the oxidation of free cysteines.
To test whether modulation of N-type inactivation requires the presence of the conserved core of Kv␤1, we produced a chimeric channel (Kv1.1-inact) in which amino acid residues 1-70 of Kv␤1 were spliced to the N terminus of Kv1.1. Therefore, Kv1.1-inact is a channel that has an N-terminal inactivation gate from Kv␤1 but does not have the conserved AKR core. As expected, Kv1.1-inact produced inactivating K ϩ current similar to that produced by coexpression of Kv1.1 and Kv␤1 (Fig. 2E). When 4-CY was perfused to the intracellular side of Kv1.1-inact, only a small change in current was observed with a ⌬I ss of 6 Ϯ 1% (n ϭ 8) (Fig. 2, E and G, and Table 1). The small ⌬I ss is not significantly different from that of the vehicle control (p Ͼ 0.05 in one-way ANOVA test) (Fig. 2, C and G, and Table  1). Thus, the 4-CY effect is dependent upon the presence of the conserved AKR core.
Next, we examined whether the AKR activity is required for channel modulation. We made the K152M mutation in Kv␤1 and coexpressed the mutant with Kv1.1. Kv␤1(K152M) induced channel inactivation, and the on-rate of channel inactivation was 238 Ϯ 6 s Ϫ1 (n ϭ 7), which is significantly slower than that of the wild type (p Ͻ 0.001 in unpaired t test) ( Fig. 2F and Table 1). The slower inactivation rate is consistent with a previous study that also reported a slower rate when the lysine was mutated to alanine (20). Here, we focused on the 4-CY effect, and we found that 4-CY slightly increased the current with a ⌬I ss of 6 Ϯ 1% (n ϭ 7) (Fig. 2, F and G, and Table 1). The slight change in current is not significantly different compared with the vehicle control and Kv1.1-inact (p Ͼ 0.05 in one-way ANOVA test). This result suggests that the 4-CY effect requires not just the presence but also the catalytic activity of the AKR core.
Oxidation of Kv␤-bound NADPH Induces Channel Modulation-That a substrate induces channel modulation through a functional AKR suggests that NADPH oxidation is required. To find out whether NADPH oxidation is sufficient to induce channel modulation, we next investigated whether reagents (other than substrates) that oxidize Kv␤1-bound NADPH also induce similar changes in channel current. Two non-substrate oxidizing reagents, hydrogen peroxide (H 2 O 2 ) and NADP ϩ , were tested. H 2 O 2 is a physiologically relevant compound because its concentration changes during inflammation and oxidative stress (23). When 10 mM H 2 O 2 was mixed with purified Kv␤1 protein, NADPH fluorescence was eliminated, indicating that the bound cofactor was oxidized (Fig. 3A). In a parallel experiment, H 2 O 2 (100 M) was perfused to inside-out patches expressing both Kv1.1 and Kv␤1, and an increase in channel current (similar to that induced by a substrate) was observed with a ⌬I ss of 40 Ϯ 6% (n ϭ 5) (Fig. 3, B and F, and Table 1). By contrast, as high as 10 mM H 2 O 2 induced only a rather small change in Kv1.1-inact and Kv1.1 (Fig. 3, C, D, and F) with ⌬I ss values of 4 Ϯ 1% (n ϭ 4) and 5 Ϯ 1% (n ϭ 3), respectively. Combined, these results indicate that the H 2 O 2 effect is mediated by the AKR core. Although H 2 O 2 itself can directly oxidize Kv␤-bound NADPH, it can also react with membrane lipids to produce aldehydic metabolic intermediates (24) that could be Kv␤ substrates. Thus, the H 2 O 2 effect could potentially have two pathways. We therefore tested H 2 O 2 on Kv␤1(K152M). We found that H 2 O 2 induced a large increase in current with a ⌬I ss of 35 Ϯ 5% (n ϭ 5) (Fig. 3, E and F,  We next tested NADP ϩ . When purified Kv␤1 protein was mixed with 0.2 mM NADP ϩ , the fluorescence peak at 454 nm decreased to 35 Ϯ 1% of the original intensity (n ϭ 3) (Fig. 4A). The decrease in fluorescence is likely due to the displacement of NADPH from a relatively non-polar binding site environment to an aqueous environment. We confirmed that NADPH had been replaced by NADP ϩ by separating Kv␤1 from the free cofactor after the fluorescence no longer changed (Fig. 4A). Although it is possible that the bound cofactor was oxidized by NADP ϩ instead of being exchanged, it is clear that NADP ϩ converted Kv␤-bound NADPH to NADP ϩ .
It was observed previously that NADP ϩ reduces inactivation in whole-cell patch-clamped cells coexpressing Kv1.5 and Kv␤1 (25). A potential complication of the whole-cell patch clamp experiment is that NADP ϩ could generate metabolic intermediates that react with Kv␤ through other oxidoreductases. To eliminate this concern, we perfused 0.2 mM NADP ϩ directly to inside-out patches expressing both Kv1.1 and Kv␤1, and we observed an increase in channel current with a ⌬I ss of 36 Ϯ 2% (n ϭ 18) (Fig. 4, B and D, and Table 1), which is similar to that induced by H 2 O 2 . The AKR core is required for this effect because when NADP ϩ was perfused to the intracellular side of Kv1.1-inact, the current was almost unchanged with a ⌬I ss of 3 Ϯ 1% (n ϭ 3) (Fig. 4C and Table 1). Combined, the H 2 O 2 and NADP ϩ effects indicate that oxidation of Kv␤-bound NADPH is sufficient to potentiate Kv1 current.
Modulation of Both Activation and Inactivation Gatings by Oxidation-To further dissect the 4-CY effect, we next studied the 4-CY effect on both channel activation and inactivation processes. 4-CY induced a shift in the midpoint of the currentvoltage relationship (V1 ⁄ 2 ) of the Kv1.1-Kv␤1 complex from Ϫ59.1 Ϯ 1.6 to Ϫ65.6 Ϯ 0.9 mV (n ϭ 10) (Fig. 5, A-C). Although the shift in V1 ⁄ 2 is modest, the increase in channel current is large and is ϳ6-fold at Ϫ60 mV (Fig. 5B, right panel). Because many  A, fluorescence spectra of the purified Kv␤1 core before (black trace) and after (red trace) mixing with 0.2 mM NADP ϩ . After the fluorescence no longer changed, Kv␤1 was purified from the reaction mixture, and its spectrum was recorded (blue trace). Shown are K ϩ currents recorded from inside-out patches expressing both Kv1.1 and Kv␤1 (B) and Kv1.1-inact only (C) before (black traces) and after (red traces) perfusion of 0.2 mM NADP ϩ . D, ⌬I ss after NADP ϩ modulation. Error bars are S.E., and n is the number of patches. Unpaired Student's t test was used to assess the NADP ϩ effect on Kv1.1 ϩ Kv␤1 and Kv1.1-inact (⌬⌬⌬, p Ͻ 0.001).
cells have resting membrane potentials at around that voltage, a large change in current level may significantly affect the excitability of a cell.
One prominent feature of the effect of exposure to 4-CY is the almost complete loss of channel inactivation. To further quantify the change, the on-rate of channel inactivation was measured at voltages between Ϫ20 and ϩ60 mV, where significant channel inactivation was observed. After 4-CY perfusion, a reduction in the on-rate was observed at all voltages (Fig. 6B). In contrast to the large change in the on-rate, the rate of recovery from inactivation (measured using a paired-pulsed protocol) was unchanged (Fig. 6, A and C, and Table 1).
Because the V1 ⁄ 2 for Kv1.1 coexpressed with Kv␤1 is approximately Ϫ60 mV, all channels are maximally activated for voltages more positive than Ϫ20 mV (Fig. 5C), and as a result, the large increase in current at positive membrane potentials is due mainly to the slower on-rate of inactivation. In excitable cells, oxidation of NADPH bound to Kv␤1 increases K ϩ current at positive voltages, shortens the action potential, and speeds up repolarization of a cell.
Deletion of the Kv1.1 Distal C Terminus Does Not Affect Current Potentiation-It has been suggested in two structural studies that the C terminus of a Kv1 channel interacts with Kv␤ (13,14), which naturally led us to hypothesize that the C terminus is involved in functional coupling. To test this, we constructed Kv1.1⌬C by deleting the distal C-terminal 60 amino acid residues of the channel. Several other constructs with more C-terminal residues deleted were also generated, but none expressed at sufficiently high levels for patch clamp studies. When aligned with the Kv1.2 channel, the high resolution atomic structure of which was solved by x-ray crystallography (Protein Data Bank code 2A79) (13), Kv1.1⌬C has 15 amino acid residues beyond the last structurally resolved C-terminal residue in the Kv1.2 structure. Because the last structurally resolved residue in Kv1.2 is ϳ51 Å away from the closest residue in Kv␤ in its crystallized complex, the C-terminal 15 amino acids in Kv1.1⌬C are unlikely to interact with Kv␤, even in their most extended conformation.
When paired with Kv␤1, the on-rate of Kv1.1⌬C inactivation at ϩ60 mV is 225 Ϯ 4 s Ϫ1 (n ϭ 32), which is slightly slower than that of the wild type, and the rate of recovery from inactivation is 14 Ϯ 1 s Ϫ1 (n ϭ 9), which is similar to that of the wild type. It seems that deleting the C terminus has only a modest effect on  channel inactivation. 4-CY, NADP ϩ , and H 2 O 2 all induced large potentiations of channel current (Fig. 7, A-D) with ⌬I ss values of 52 Ϯ 3% (n ϭ 13), 44 Ϯ 3% (n ϭ 11), and 46 Ϯ 3% (n ϭ 8), respectively. All three ⌬I ss values are significantly different from that of Kv1.1-inact (p Ͻ 0.001 in unpaired t tests). We conclude that the C terminus of Kv1.1 is not required for its modulation by Kv␤1 observed here.
Modulation of Enzymatic Reaction Rates by Membrane Potentials-The tight association between Kv1 and Kv␤ has inspired the hypothesis that AKR function is modulated by membrane potentials because of the different conformations of a channel, i.e. Kv␤ could be a voltage-dependent enzyme (18). Although the physiological utility of such an AKR is not obvious, an example of a voltage-dependent enzyme has emerged recently. A voltage-dependent phosphatase has a membraneembedded voltage-sensing domain covalently linked to a cytosolic phosphatase, the activity of which is different at different membrane potentials (26).
We do not have a direct measure of enzymatic activity for Kv␤ coexpressed with Kv1.1 on cell membranes. However, because we know that NADPH oxidation induces channel modulation, we could measure the rate of channel modulation as an indirect readout for the rate of NADPH oxidation. After 4-CY was perfused to the intracellular side of Kv1.1 coexpressed with Kv␤1, channel current increased over time and reached a steady state (Fig. 8A). When the normalized current amplitude was plotted against time, the data points could be fit with a single-component exponential function (Fig. 8B). The inverse of the exponential time constant is defined as the rate of channel modulation.
We measured the rate of channel modulation at two different membrane potentials: Ϫ100 mV, at which most of the channels are closed, and 0 mV, at which most of the channels are open and inactivated. Both the 4-CY and NADP ϩ effects occurred ϳ2-fold faster at 0 mV than at Ϫ100 mV (p Ͻ 0.01 in unpaired t test) (Fig. 8B), consistent with the notion that membrane potentials affect NADPH oxidation or exchange.

DISCUSSION
By identifying reagents that oxidize Kv␤1-bound NADPH and studying in parallel the effect of these reagents on channel current, we observed a correlation between NADPH oxidation and potentiation of channel current. These properties suggest that the Kv1-Kv␤ complex is capable of transducing a change in cellular metabolic redox state into a change in channel activity and hence cell excitability. The correlation between cofactor oxidation and channel modulation also provides a measure of enzymatic activity on inside-out patches, and we found that the rates of both hydride transfer and cofactor exchange are significantly faster at 0-mV membrane potential than at Ϫ100 mV, providing the first evidence of a voltage-regulated aldoketoreductase.
The Kv1-Kv␤ complex is widely expressed in excitable cells, such as neurons and muscle cells (27,28), and in non-excitable cells, such as lymphocytes and alveolar cells (29,30). In the central nervous system, Kv1 family channels are preferentially targeted to axons (31) and likely play an important role in controlling the invasion and propagation of an action potential. In humans, mutations of the Kv1 channel have been linked to episodic ataxia (32), and loss of the Kv␤ gene has been associated with seizures (33). In non-excitable T-lymphocytes, the activity of Kv1 channels changes significantly during the T-cell maturation process (34). However, how the redox-sensing capability of the Kv1-Kv␤ complex would serve cellular functions remains to be explored. In a native cell, multiple members of the Kv1 and Kv␤ families express at the same time, so the (Kv1) 4 -(Kv␤) 4 complexes may be formed by heterotetramers of Kv1 and Kv␤ subunits. In such complexes, Kv1 modulation induced by oxidation of Kv␤-bound NADPH would no  Table 1; ***, p Ͻ 0.001). The smooth curves are single-component exponential functions fit to the data points. Error bars are S.E. The time constants and number of patches are as follows: 7.9 Ϯ 0.7 min (Ϫ100 mV for 4-CY, n ϭ 29), 4.9 Ϯ 0.6 min (0 mV for 4-CY, n ϭ 14), 12.2 Ϯ 0.9 min (Ϫ100 mV for NADP ϩ , n ϭ 13), and 5.9 Ϯ 0.9 min (0 mV for NADP ϩ , n ϭ 10).
doubt be more complicated than we have observed in the heterologous oocyte expression system. From the work presented here and a previous study (19), we know that both Kv␤1 and Kv␤2 are functional AKRs and that both mediate substrateinduced channel modulation. Even though Kv␤2 does not have an N-type inactivation gate, it reduces the on-rate of N-type inactivation inherent to Kv1.4 channels (19) and, as a result, potentiates K ϩ current. It is very likely that conditions in a cell that lead to oxidation of bound NADPH would increase Kv1 current, whether it is through the leftward shift of voltage-dependent channel activation or through the reduction of channel inactivation.
Different forms of redox modulation of Kv channels have been reported, including methionine oxidation by methioninesulfoxide reductase or reactive oxygen species (35)(36)(37) and cysteine oxidation by reactive oxygen species (38). This study adds NADPH to the targets of oxidation, and in this case, because it is mediated by an aldoketoreductase, it has the potential of being highly selective for a specific substrate. The challenge now is to identify the physiological substrate of Kv␤.
Because the inactivation gate is on Kv␤1, the change in channel inactivation could be due to the dissociation of Kv␤1 after oxidation of the bound cofactor. This is not the mechanism, however, because the V1 ⁄ 2 measured after oxidation is Ϫ65.6 Ϯ 0.9 mV, which is significantly different from the V1 ⁄ 2 of Kv1.1 expressed alone (Ϫ56.8 Ϯ 0.8 mV, p Ͻ 0.001 in unpaired t test). We speculate that oxidation of the bound cofactor induces a structural change in Kv␤, which may propagate to Kv1.1 to affect channel inactivation. We have ruled out the C terminus of Kv1.1, which is not required for the modulation. A likely prospect is the intracellular T1 domain of Kv1.1, which serves as a docking site for Kv␤.
In conclusion, we have found that Kv␤1 responds to a variety of redox signals and modulates channel function. We have also found that Kv␤1 aldoketoreductase is voltage-dependent: depolarization speeds up the rate of oxidation. Thus, the Kv1.1-Kv␤1 complex exhibits bidirectional allosteric modulation, which we are starting to understand at the molecular level.