Functional Expression of Two KvLQT1-related Potassium Channels Responsible for an Inherited Idiopathic Epilepsy*

Benign familial neonatal convulsions (BFNC), a class of idiopathic generalized epilepsy, is an autosomal dominantly inherited disorder of newborns. BFNC has been linked to mutations in two putative K+ channel genes, KCNQ2 andKCNQ3. Amino acid sequence comparison reveals that both genes share strong homology to KvLQT1, the potassium channel encoded byKCNQ1, which is responsible for over 50% of inherited long QT syndrome. Here we describe the cloning, functional expression, and characterization of K+ channels encoded byKCNQ2 and KCNQ3 cDNAs. Individually, expression of KCNQ2 or KCNQ3 in Xenopus oocytes elicits voltage-gated, rapidly activating K+-selective currents similar to KCNQ1. However, unlike KCNQ1, KCNQ2 and KCNQ3 currents are not augmented by coexpression with the KCNQ1 β subunit, KCNE1 (minK, IsK). Northern blot analyses reveal that KCNQ2 andKCNQ3 exhibit similar expression patterns in different regions within the brain. Interestingly, coexpression of KCNQ2 and KCNQ3 results in a substantial synergistic increase in current amplitude. Coexpression of KCNE1 with the two channels strongly suppressed current amplitude and slowed kinetics of activation. The pharmacological and biophysical properties of the K+currents observed in the coinjected oocytes differ somewhat from those observed after injecting either KCNQ2 or KCNQ3 by itself. The functional interaction between KCNQ2 and KCNQ3 provides a framework for understanding how mutations in either channel can cause a form of idiopathic generalized epilepsy.

bination with the KCNE1 subunit, encodes the slow component of the cardiac delayed rectifier K ϩ current (2)(3)(4), and mutations in KCNQ1, which occur in LQTS patients, partially or completely inhibit the channel in a dominant-negative fashion (5,6). In an attempt to identify additional members of the KCNQ1 K ϩ channel gene family, the KCNQ1 sequence was used to search DNA and protein sequence data banks. Two additional KCNQ1-related genes, KCNQ2 and KCNQ3, were identified.
Recent publications indicate that mutations in KCNQ2 or KCNQ3 are associated with BFNC, an autosomal dominantly inherited epilepsy in newborns (7)(8)(9). Preliminary functional characterization of KCNQ2 confirmed that this gene encodes a voltage-activated K ϩ channel (9). Here we describe the cloning, tissue distribution, and functional expression of both KCNQ2 and KCNQ3. More importantly, we demonstrate that these two channels interact functionally with each other and with KCNE1.

EXPERIMENTAL PROCEDURES
Molecular Cloning and Expression of KCNQ2 and KCNQ3-5Ј Rapid amplification of cDNA ends polymerase chain reaction was performed by amplifying human brain or fetal brain cDNA libraries or Marathon-Ready cDNAs (CLONTECH) using primers derived from the KvLQT1related EST sequences (EST numbers yn72g11, yo31c08, and ys93a07). Polymerase chain reaction products were gel-purified, subcloned, and sequenced. The Gene Trapper experiment was performed using the protocol as described in the manufacturer's manual (Life Technologies Inc.). Random-primed 32 P-labeled DNA probes containing specific regions of KCNQ2 or KCNQ3 sequence were used for screening of cDNA libraries and Northern blot analysis using standard protocols. The composite full-length KCNQ2 and KCNQ3 cDNA clones were obtained by restriction enzyme digestion and ligation of overlapping cDNA clones. The full-length cDNAs were subcloned into a Xenopus expression vector, derived from pSP64T plasmid. Capped cRNA for microinjection was synthesized using mMessage mMachine Kit (Ambion) as described (4,6).
Electrophysiology-Stages V and VI Xenopus laevis oocytes were defolliculated with collagenase treatment and injected with cRNAs as described previously (4). Currents were recorded at room temperature using conventional two-microelectrode voltage clamp (Dagan TEV-200) 3-4 days after injecting KCNQ2 (15 ng), KCNQ3 (15 ng), or KCNE1 (2 ng) cRNA alone or in combination. Microelectrodes (0.8 to 1.5 megaohms) were filled with 3 M KCl. Bath solution contained (in mM): 96 NaCl, 2 KCl, 0.4 CaCl 2 , 2 MgCl 2 , and 5 HEPES (pH 7.5). K ϩ selectivity was assessed by determining the dependence of tail current reversal potential on the external K ϩ concentration. Tail currents were elicited at potentials of Ϫ110 to ϩ10 mV following a voltage step to ϩ20 mV while the external K ϩ concentration was varied between 2, 10, 40, and 98 mM. Current reversal potential under each condition was determined for each oocyte by measuring the zero intercept from a plot of tail current amplitude versus test potential. KCl was varied in selectivity experiments by equimolar substitution with NaCl. PCLAMP 6.0 software (Axon Instruments) was used for data acquisition and analysis. All data were sampled at rates at least two times the low pass filter rate. Clofilium was obtained from RBI Biochemicals and 4-aminopyridine (4-AP), tetraeth-* This work was supported by the Bristol-Myers Squibb Pharmaceutical Research Institute. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. ‡ The first two authors contributed equally to this work.

RESULTS AND DISCUSSION
Cloning and Tissue Distribution of KCNQ2 and KCNQ3-KCNQ1-related expressed sequence tags (ESTs) were identified in a GCG BLAST search of the GenBank TM data base with KCNQ1 sequence. Primers, derived from the consensus sequences of EST clones, were used to amplify human brainderived cDNA, and 877-base pair and 325-base pair fragments were isolated for KCNQ2 and KCNQ3, respectively. To obtain full-length cDNA sequences of both genes, we employed 5Ј rapid amplification of cDNA ends polymerase chain reaction, screening of cDNA libraries, and Gene Trapper techniques. The composite full-length cDNAs of KCNQ2 and KCNQ3 contain an open reading frame (ORF) encoding an 871-and 854-amino acid polypeptide, respectively. DNA sequence analysis and conceptual translation of both cDNAs reveals that they encode proteins with the structural features of a voltage-gated potassium channel and are most closely related to KCNQ1 (3,4). KCNQ2 exhibits a high degree of sequence similarity with KCNQ3 (Ϸ70%), indicating that they belong to the same subfamily. Both proteins have a longer C-terminal domain (ϳ200 amino acids) than KCNQ1.
Unlike KCNQ1, which is expressed strongly in human heart and pancreas (1,4), Northern blot analysis revealed that KCNQ2-and KCNQ3-specific transcripts are detectable only in human brain (Fig. 1). The sizes of the major transcripts for KCNQ2 and KCNQ3 are 8.5 kilobases and 10.5 kilobases, respectively. Expression of human KCNQ2 is high in the hippocampus, caudate nucleus, and amygdala; moderate in the thalamus, and weak in the subthalamic nucleus, substantia nigra, and corpus callosum (Fig. 1, top panel). A separate Northern blot demonstrates that expression of human KCNQ2 is high in the cerebral cortex, is moderate in the putamen, temporal lobe, frontal lobe, occipital pole, and cer-ebellum, and is barely detectable in the medulla and spinal cord (Fig. 1, top panel). A similar pattern of expression was observed previously for KCNQ2 (7). Interestingly, KCNQ3 exhibits a nearly identical expression pattern in the brain (Fig. 1, bottom panel).
Functional Expression and Characterization of KCNQ2 and KCNQ3-The full-length KCNQ2 and KCNQ3 cDNAs were subcloned into a Xenopus expression vector, and cRNA was generated by in vitro transcription. The properties of the channels encoded by KCNQ2 and KCNQ3 were investigated by expressing the transcribed cRNAs in Xenopus oocytes. Depolarizing voltage steps elicited outward currents in oocytes injected with KCNQ2 ( Fig. 2A). The currents activated at potentials positive to Ϫ60 mV and showed slight inward rectification at the more positive potentials. Similar currents never were observed in water-injected control oocytes. KCNQ2 currents exhibited a rapidly activating delayed rectifier current phenotype similar to KCNQ1 current (2)(3)(4). Fig. 2B shows the current-voltage (I-V) relationship for KCNQ2 currents recorded at the end of the 1-s voltage steps. The K ϩ selectivity of the expressed current was examined by investigation of tail current reversal potentials in bath solutions containing 2, 10, 40, and 98 mM K ϩ . Reversal potentials closely followed the Nernst potential for K ϩ revealing a predominantly K ϩ -selective channel (Fig. 2C). The reversal potential for KCNQ2 current shifted by 51 mV per 10-fold change in external K ϩ .
Inhibitors of K ϩ channels were used to investigate the pharmacology of KCNQ2. The effects of 4-AP, E-4031, clofilium, charybdotoxin (CTX), and TEA on KCNQ2 currents recorded from a single oocyte are shown in Fig. 2D. Each of these compounds also was tested alone in individual oocytes, and the effects of each agent were consistent with the data shown in Fig. 2. CTX (100 nM), a protein from scorpion venom that inhibits a variety of Ca 2ϩ -activated and voltage-dependent K ϩ channels (10, 11), did not inhibit KCNQ2 current. CTX also had no effect on KCNQ1 current (not shown). E-4031 (10 M), a selective inhibitor of the HERG K ϩ channel (12), and 4-AP (2 mM), an inhibitor of Shaker-type K ϩ channels (13), also had no significant effects on KCNQ2 current. Similarly, neither agent inhibits KCNQ1 channels (4). Clofilium (10 M), a compound that inhibits KCNQ1 (4) with an IC 50 Ͻ10 M, had little effect on KCNQ2 current. TEA (1 mM), a nonselective K ϩ channel inhibitor and weak KCNQ1 antagonist (4), reduced KCNQ2 current by 90%.
A family of currents elicited by depolarizing voltage steps in an oocyte injected with KCNQ3 cRNA are shown in Fig.  2E. The currents activate at potentials positive to Ϫ70 mV and rectify inwardly at potentials greater than 0 mV, as is obvious from the I-V relationship (Fig. 2F). The KCNQ3 reversal potential shifted 49 mV per 10-fold change in external K ϩ (Fig. 2G). Thus, although still predominantly selective for K ϩ , KCNQ3 is slightly less K ϩ -selective than KCNQ2. The pharmacology of KCNQ3 was significantly different from that of KCNQ2 (Fig. 2H). Clofilium (10 M) reduced KCNQ3 current by 30% from control but had little effect on KCNQ2. TEA, which strongly inhibited KCNQ2 at 1 mM, produced little inhibition of KCNQ3 at 5 mM. CTX (100 nM; not shown), 4-AP (2 mM), and E-4031 (10 M) also had no effect on KCNQ3 current. Thus, although both KCNQ2 and KCNQ3 encode related voltage-activated K ϩ channels, significant differences include: (a) degree of rectification at positive voltages, (b) minimum activation voltage, (c) selectivity for K ϩ , and (d) pharmacology.
KCNQ2 and KCNQ3 Functionally Interact-The overlapping expression pattern of KCNQ2 and KCNQ3 in different brain regions (Fig. 1), together with the fact that mutations in either KCNQ2 or KCNQ3 cause the same inherited epilepsy (BFNC; 7-9), prompted us to test for functional interaction between the two channels. Families of currents elicited by depolarizing voltage steps in oocytes injected with KCNQ2 and KCNQ3 alone and together are shown in Fig. 3A. Current amplitudes recorded from oocytes coexpressing the two channels were 15-fold greater than in oocytes injected with each of the channels individually. Peak current amplitudes at ϩ30 mV for KCNQ2, KCNQ3, and KCNQ2ϩKCNQ3 were 0.98 Ϯ 0.09 (n ϭ 6), 0.98 Ϯ 0.06 (n ϭ 5), and 14.2 Ϯ 0.62 M (n ϭ 6), respectively. Quantitatively similar results were obtained in three separate batches of oocytes. The I-V relationship shows that KCNQ2ϩKCNQ3 currents activated at potentials positive to Ϫ60 mV and did not rectify, unlike KCNQ2 and particularly KCNQ3, at positive voltages (Fig. 3B). The reversal potential of tail currents shifted by 57 mV per 10-fold change in external K ϩ indicating that KCNQ2ϩKCNQ3 is nearly perfectly selective for K ϩ (Fig. 3C). KCNQ2ϩKCNQ3 current is weakly sensitive to inhibition by 5 mM TEA and 10 M clofilium but not to 100 nM CTX or 2 mM 4-AP (Fig. 3D). E-4031 (10 M) also did not inhibit KCNQ2ϩKCNQ3 current (not shown). These results suggest strongly that KCNQ2ϩKCNQ3 interact to form a channel with properties distinct from either KCNQ2 or KCNQ3 channels.
minK Interacts with KCNQ2ϩKCNQ3 Channels-The ␤ subunit KCNE1 dramatically alters the amplitude and gating kinetics of the KCNQ1 channel (2-4, 14). Because KCNQ2 and KCNQ3 are members of the same K ϩ channel subfamily, we   FIG. 2. Functional and pharmacologic characterization of KCNQ2 and KCNQ3 currents. A, families of currents from KCNQ2 cRNA-injected oocytes elicited by 1-s voltage steps, from a holding potential of Ϫ80 mV, to test potentials ranging from Ϫ70 to ϩ50 mV in 10-mV increments. B, current-voltage (I-V) relationship for oocytes expressing KCNQ2 (n ϭ 6). Currents were recorded using the protocol in A. C, dependence of tail current reversal potential (E rev ) on the external K ϩ concentration. The dashed line has a slope of 58 mV and is drawn according to the Nernst equation for a perfectly selective K ϩ channel. Each value is the mean Ϯ S.E. from 6 oocytes. D, effects of E-4031, 4-AP, TEA, charybdotoxin, and clofilium on KCNQ2 current. Superimposed currents were recorded during 1-s steps to ϩ20 mV, from Ϫ80 mV, during the same experiment. Compounds were applied via bath perfusion in order from top to bottom. The bath was perfused with control solution for 5 min or until effects reversed completely, between compounds. Similar results were obtained in three additional oocytes. E, families of currents from KCNQ3 cRNA-injected oocytes elicited using the protocol in A. F, I-V relationship for oocytes expressing KCNQ3 (n ϭ 6). G, dependence of tail current E rev on the external K ϩ calculated using the protocol in C (n ϭ 6). H, effects of E-4031, 4-AP, TEA, and clofilium on KCNQ3 current. Similar results were obtained in three additional oocytes. tested for an interaction between KCNE1 and KCNQ2ϩKCNQ3 channels. Fig. 4 shows currents elicited by 1-s depolarizing voltage steps in oocytes expressing KCNE1 alone, KCNQ2ϩKCNQ3, and KCNQ2ϩKCNQ3ϩKCNE1. KCNE1 significantly attenuated KCNQ2ϩKCNQ3 current amplitude and slowed gating kinetics. Peak current amplitude at ϩ30 mV was reduced by 62 Ϯ 6.0% (n ϭ 6) in oocytes coexpressing KCNE1. Activating currents were fitted to a bi-exponential function to determine fast and slow time constants of activation. Fast and slow time constants for activation of KCNQ2ϩKCNQ3 current at ϩ10 mV were 50.1 Ϯ 3.4 (n ϭ 6) and 239.3 Ϯ 17.5 ms (n ϭ 6), respectively; these were shifted to 124.7 Ϯ 8.8 (n ϭ 5) and 680.7 Ϯ 71.4 ms (n ϭ 6) when KNCE1 was injected together with KCNQ2ϩKCNQ3. Similar results were obtained in more than 15 oocytes from each group in this and two additional batches of oocytes. KCNE1 currents appear absent because of the duration (1 s) of the voltage steps used and the scale at which the currents are shown; however, as shown clearly in the inset, 5-s voltage steps elicited typical KCNE1 currents in the same oocyte. Whether regions of the FIG. 3. Coexpression of KCNQ2 and KCNQ3. A, families of currents from KCNQ2, KCNQ3, and KCNQ2ϩKCNQ3 cRNA-injected oocytes elicited by 1-s voltage steps, from a holding potential of Ϫ80 mV, to test potentials ranging from Ϫ70 to ϩ50 mV (10-mV increments). B, current-voltage (I-V) relationship for oocytes expressing KCNQ2ϩKCNQ3 (n ϭ 6). Currents were recorded using the protocol in A. C, dependence of tail current reversal potential (E rev ) on the external K ϩ concentration. The dashed line has a slope predicted by the Nernst equation for a perfectly selective K ϩ channel. Each value is the mean Ϯ S.E. from 6 oocytes. D, effects of 4-AP, TEA, charybdotoxin, and clofilium on KCNQ2ϩKCNQ3 current. Superimposed currents were recorded during 1-s steps to ϩ20 mV, from Ϫ80 mV, during the same experiment. Compounds were applied via bath perfusion in order from top to bottom. Similar results were obtained in 4 additional oocytes.
FIG. 4. Interaction of KCNE1 with KCNQ2؉KCNQ3 currents. Families of currents from KCNE1, KCNQ2ϩKCNQ3, and KCNQ2ϩKCNQ3ϩKCNE1 cRNA-injected oocytes elicited by 1-s voltage steps, from a holding potential of Ϫ80 mV, to test potentials ranging from Ϫ70 to ϩ50 mV (10 mV increments). Inset shows KCNE1 currents elicited by 5-s voltage steps from Ϫ80 mV to potentials ranging from Ϫ30 to ϩ50 mV (20-mV increments) in the same oocyte brain which coexpress KCNQ2 and KCNQ3 also express KCNE1 remains to be determined. The effect of KCNE1 on gating kinetics is similar for KCNQ1 and KCNQ2ϩKCNQ3 channels. In contrast, KCNE1 augments KCNQ1 current but inhibits KCNQ2ϩKCNQ3. Mutations in KCNE1 cause LQTS and produce dominant-negative suppression of KCNQ1 current (15). Although KCNE1 has opposite effects on KCNQ1 and KCNQ2ϩKCNQ3 channels, it will be interesting to determine whether mutations in KCNE1 account for altered neuronal excitability.
The results explain why mutations in either of two unlinked K ϩ -channel encoding genes yield the same phenotype. BFNCassociated mutations in either KCNQ2 or KCNQ3 could cause a profound reduction in KCNQ2ϩKCNQ3 current amplitude. Interestingly, a BFNC-causing mutation resulting in an nonfunctional, truncated KCNQ2 protein, failed to produce a dominant-negative inhibition of wild-type KCNQ2 channels expressed in oocytes (7). Our results, demonstrating a synergistic interaction between KCNQ2 and KCNQ3, may provide a likely explanation for this finding. That is, mutations in KCNQ2 may only produce dominant-negative effects when coexpressed with wild-type KCNQ3 channels and vice versa. This supports the suggestion from the previous study (7) that dominant-negative effects of KCNQ2 mutants may require a ␤-subunit or second protein. This information will prove important for the evaluation of functional effects of channel mutations that cause BFNC and perhaps other disorders of neuronal excitability.