KCR1, a Membrane Protein That Facilitates Functional Expression of Non-inactivating K+ Currents Associates with Rat EAG Voltage-dependent K+Channels*

Cerebellar granule neurons possess a non-inactivating K+ current, which controls resting membrane potentials and modulates the firing rate by means of muscarinic agonists. kcr1 was cloned from the cerebellar cDNA library by suppression cloning. KCR1 is a novel protein with 12 putative transmembrane domains and enhances the functional expression of the cerebellar non-inactivating K+ current inXenopus oocytes. KCR1 also accelerates the activation of rat EAG K+ channels expressed in Xenopusoocytes or in COS-7 cells. Far-Western blotting revealed that KCR1 and EAG proteins interacted with each other by means of their C-terminal regions. These results suggest that KCR1 is the regulatory component of non-inactivating K+ channels.

K ϩ channels are essential components of the plasma membranes of both excitable and non-excitable cells. Three major classes have been described for voltage-gated K ϩ currents in mammalian neurons: A-type currents with rapid inactivation, delayed rectifier currents with slow inactivation, and non-inactivating outward currents (1,2). The first two classes play an important role in shaping action potentials during repolarization and in determining the interval of an action potential. The non-inactivating outward currents can modulate firing frequency upon receptor stimulation by neurotransmitters or produce resting membrane potentials.
Typical K ϩ currents that show no inactivation upon depolarization are M currents (3,4), S currents (5), the standing outward current, I K(SO) , recently reported in cerebellar granule cells (6), and resembling currents in myoblasts (7). These currents have a common feature in that they begin to be activated from deep membrane potentials and are inhibited by neurotransmitters such as muscarinic agonists (4,6), bradykinin (3), and serotonin (5). Except for the many important items of physiological relevance identified for such currents, molecular bases for the responsible channels are poorly understood. The cloned K ϩ channels that exhibit non-inactivating currents are the aKv5.1 (8) and ether à go-go (EAG) 1 K ϩ channels (9,10).
Recently, it has been reported that the expression of Drosophila EAG K ϩ channels in oocytes results in a slow relaxation current resembling M currents (11). However, whether a rat homologue of EAG (r-EAG) contributes to the M channel has been the subject of much debate (12,13).
To characterize low-threshold non-inactivating K ϩ currents, we measured K ϩ currents in Xenopus oocytes injected with the cerebellar poly(A) ϩ RNA, in accordance with the report by Hoger et al. (14). We realized that low-threshold non-inactivating K ϩ currents (I K(ni) ), which are similar to I K(SO) , are expressed in oocytes injected with cerebellar poly(A) ϩ RNA. To clarify the molecular information for I K(ni) , we isolated a cDNA clone required for the functional channel protein for I K(ni) by suppression cloning, which in turn is based on the capacity of antisense single strand DNA to block the expression of I K(ni) in oocytes. This strategy enabled us to obtain a clone that encodes either an integral component or an important regulatory polypeptide (15). We could obtain one clone which specifically suppresses I K(ni) and the full-length cDNA termed kcr1. We describe here the functional characterization of the encoded protein designated KCR1 in oocytes and COS-7 cells and the biochemical interaction with r-EAG K ϩ channels.

EXPERIMENTAL PROCEDURES
cDNA Cloning-A pool of fractionated rat cerebellar poly(A) ϩ RNA (6 -12 kb) was used to construct a randomly primed cDNA library in ZAPII (Stratagene). Excised ssDNA pools were used for the simplified version of suppression cloning with exogenously added ribonuclease H as described elsewhere (15). Briefly, heat-denatured ssDNA and poly(A) ϩ RNA were mixed and incubated at 37°C for 1 h in a solution containing 50 mM sodium acetate (pH 7.8), 2 mM MgCl 2 , 3 mM dithiothreitol, and 1.1 units of ribonuclease H (Takara Shuzo), 2.6 units of ribonuclease inhibitor. The treated samples were directly injected into Xenopus oocytes. When Kv1.2 or Kv3.1 cRNA (2.6 ng/l) was treated with antisense Kv1.2 ssDNA (1.6 ng/l), the expression of Kv1.2 current was suppressed to 20.4 Ϯ 26% (n ϭ 13) and expression of Kv3.1 current to 99 Ϯ 17% (n ϭ 5). In each suppression experiment, expression of I K(ni) was evaluated by using I K(A) as an internal control.
Clones containing the entire coding sequence of kcr1 were obtained by screening a second rat brain library. r-eag cDNA clones were obtained by screening the same library using reverse transcriptase-polymerase chain reaction fragments as probes. The sequences of the oligonucleotides for PCR were 5Ј-CGGCGGATCCACACTCGCGGGTT-GCGCAC (forward) and 5Ј-CGGCAAGCTTTGAAATGATCCTCTCA-GCTTG (reag/rv, reverse). The nucleotide sequence of one clone (pB-SKS/eag3a) carrying nucleotides 47-3164 of the r-eag gene was identical to the published sequence except for nucleotide 1334, where T was replaced by C without changing the amino acid sequence (numerical designations correspond to the published sequence (16)).
Expression Plasmids-The coding sequence of kcr1 was amplified by PCR with following primers containing additional restriction sites * This work was supported by grants from the Ministry of Science, Education and Culture of Japan. 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 nucleotide sequence(s) reported in this paper has been submitted to the GenBank TM /EBI Data Bank with accession number(s) U78090.
Translation in Vitro-kcr1 cRNA was translated in vitro using rabbit reticulocyte lysate (Promega) and [ 35 S]methionine (NEN Life Science Products) in the presence and absence of canine pancreatic microsomes (Promega). Microsomes were then pelleted and treated with 2 milliunits of endoglycosidase H (Boehringer Mannheim) or 5 mM NaOH as described elsewhere (18,19).
Electrophysiological Measurements-Xenopus laevis oocytes were isolated and different RNAs were injected as described previously (20). In the rescue experiment, ssDNA containing the antisense strand of the open reading frame was used to deplete corresponding mRNA species from cerebellar poly(A) ϩ RNA by using the suppression cloning protocol. One day after injecting the depleted RNA (100 ng/oocyte), water or kcr1 cRNA (50 ng/oocyte) was re-injected into half of the oocytes. To express cloned K ϩ channels, each of the cRNAs (50 -500 pg/oocyte) with or without kcr1 cRNA (0.5-5 ng/oocyte) was injected. Oocytes were then tested for expression of currents 2-8 days after injection by using a two-electrode voltage clamp under constant perfusion with the solution (containing in mM: 96 NaCl, 2 KCl, 2 MgCl 2 , 0.3 CaCl 2 , 10 tetraethylammonium, 0.3 niflumic acid, 5 HEPES/NaOH (pH 7.6)). Electrodes filled with 3 M KCl had resistances of 0.4 -1.0 M⍀. COS-7 cells were transiently transfected by the DEAE-dextran method. Two days after transfection, K ϩ currents were recorded at room temperature by using the whole-cell configuration of the discontinuous voltage clamp technique. The external solution at pH 7.4 contained (in mM): 140 NaCl, 5 KCl, 1 CaCl 2 , 2 MgCl 2 , 10 HEPES/NaOH. The pipette solution at pH 7.3 contained (in mM): 140 KCl, 2 MgCl 2 , 10 HEPES/KOH, 2 EGTA. The current traces were fitted by means of the single exponential function of the power of n or by the sum of two exponentials of the first order as described elsewhere (21).
Far-Western Blotting-Purified MBP fusion proteins (2 g) were fractionated in SDS-polyacrylamide gel electrophoresis and electrotransferred onto polyvinylidene difluoride membranes. The proteins were renatured in Tris-buffered saline with 3% skim milk at 37°C for 1 h. The binding reactions were then performed in Tris-buffered saline containing 3% skim milk and GST-KCR1 (2 g/ml) for 1 h at 4°C. The membranes were washed three times with Tris-buffered saline, and binding was detected by anti-GST antibodies using the enhanced chemiluminescence system (Amersham Pharmacia Biotech.).

RESULTS
Suppression Cloning of kcr1 from Cerebellar cDNA Library-Xenopus oocytes injected with poly(A) ϩ RNA isolated from the rat cerebellum exhibited three types of voltage-dependent K ϩ currents. One tetraethylammonium-sensitive delayed rectifier with an IC 50 value of 0.065 Ϯ 0.03 mM (mean Ϯ S.D., n ϭ 12) and two major tetraethylammonium-resistant K ϩ currents upon depolarization: the A-type current (I K(A) ; Ref. 22) and the non-inactivating outward current (I K(ni) ; Refs. 6 and 14). I K(ni) was activated with a potential greater than Ϫ60 mV and did not show apparent inactivation during 30-min depolarization at a membrane potential of ϩ40 mV. The activation and deactivation kinetics of I K(ni) were rapid. When the potential was stepped up, first from Ϫ100 mV to ϩ40 mV and then to Ϫ40 mV, I K(ni) could be fitted by a single exponential function with a time constant of 8.5 Ϯ 2.6 ms (n ϭ 8) for activation and 4.1 Ϯ 1.9 ms (n ϭ 8) for deactivation. It was also found that I K(ni) could be reversibly inhibited by focal applications (10 l) of 100 M serotonin to 84.7 Ϯ 6.4% (n ϭ 11) of the control level. These characteristics of I K(ni) resembled those of I K(SO) in cerebellar granule neurons.
Sucrose gradient fractionation revealed that the cerebellar poly(A) ϩ RNA species responsible for I K(ni) was larger than 6 kb, which made it difficult to apply direct expression cloning strategy to isolate a corresponding cDNA for I K(ni) . Instead, the suppression cloning strategy (15) was employed, because this strategy does not require full-length cDNAs. The simplified version of suppression cloning from a rat cerebellar cDNA library resulted in the isolation of one cDNA clone of 1.5 kb that inhibited the expression of I K(ni) . Screening of a second cDNA library yielded 14 overlapping clones, which covered 7.6 kb. Nucleotide sequencing of the full-length cDNA termed kcr1 (K ϩ channel regulator) revealed that the largest open reading frame encoded 474 amino acid residues preceded by one inframe stop codon, which was designated KCR1.
The deduced amino acid sequence of kcr1 cDNA is shown in Fig.  1A. KCR1 appears to lack the conserved ion pore domain of K ϩ channels (23). A hydropathy plot of the amino acid sequence of KCR1 shows 12 hydrophobic regions (Fig. 1B), which have been allocated to membrane-spanning segments (underlined regions in Fig. 1A). KCR1 has two consensus N-glycosylation sites between S6-S7 and S12-C terminus and one O-glycosylation site between S1-S2 (24). A BLAST search of the National Center for Biotechnology Information gene data bases (25) with KCR1 produced matches with three genes: DIE2 from Saccharomyces cerevisiae (26), SPAC56F8.06C from Schizosaccharomyces pombe, 2 and T24D1.4 from Caenorhabditis elegans. 3 KCR1 shows 31% identity (48% similarity) with SPAC56F8.06C, 26% identity (46% similarity) with DIE2, and 25% identity (43% similarity) with T24D1.4 at the amino acid level. Interestingly, not only the hydrophobic regions but also the three hydrophilic regions (S1-S2, S9-S10, and C terminus) were conserved (Fig. 1A). DIE2 has been shown to relieve the down-regulation of inositol transporter (ITR1) mRNA caused by the transporter substrates (26), but the function of two other gene products has not been identified in any reports.
Northern blot analysis revealed a 7.6-kb mRNA for kcr1. Hybridizable mRNA was found not only in brain but also in many peripheral tissues examined (Fig. 1C), indicating that expression of KCR1 is not restricted to the central nervous system.
In Vitro Translation Product of kcr1-Translation reaction in vitro of kcr1 cRNA yielded a 40-kDa product on a 10% SDS-polyacrylamide gel ( Fig. 2A). Mobility of the translation products varied according to the percentage of gel concentration used; 34 kDa for 7.5%, 44 kDa for 12%, and 47 kDa for 14%. Calibration against the retardation coefficient determined from Ferguson plots (27), in other words, a comparison of the slopes of KCR1 and molecular weight standards in the plot, yielded an apparent molecular weight of 55,500 Ϯ 6,100 which was close to the predicted molecular weight from the deduced amino acid sequence (M r 55,571). When pancreatic microsomes were included, translated proteins predominated in the membrane pellet, indicating that the KCR1 is a membrane protein. In the presence of microsomes, the mobility of the product was also shifted to 42 kDa. This shift of 2 kDa was resistant to endoglycosylase H treatment (19). The ␤-elimination in an mild alkaline condition (18) could partially remove this shift to 40 kDa (Fig. 2B).
Suppression of I K(ni) by kcr1 Antisense and Restoration by kcr1 cRNA in Oocytes-The function of KCR1 in ion channel activity was examined. Xenopus oocytes injected with kcr1 cRNA exhibited small outward currents of 41.3 Ϯ 30.2 nA (n ϭ 40) at ϩ40 mV, which were not significantly different from the endogenous currents in control oocytes (32.4 Ϯ 18.5 nA, n ϭ 49; Fig. 3A, left panel). The inability of KCR1 to permeate K ϩ is not surprising, because the amino acid sequence lacks the conserved ion pore sequence.
We next tested whether KCR1 was necessary for the formation of the functional channels producing I K(ni) in oocytes injected with cerebellar poly(A) ϩ RNA. Depletion of corresponding mRNA from cerebellar poly(A) ϩ RNA with the antisense ssDNA of kcr1 cDNA inhibited expression of I K(ni) (49 Ϯ 22%, for O-glycosylation is indicated by an asterisks, N-glycosylation by a q, and protein kinase C-dependent phosphorylation by a E. B, hydropathy plot of the amino acid sequence of KCR1. The amino acid sequence of KCR1 was subjected to the hydropathy analysis according to Kyte and Doolittle (39) with a window size of 18 amino acids. The 12 hydrophobic peaks are designated S1-S12. C, expression of kcr1 mRNA in different rat tissues. Poly(A) ϩ RNA (4.5 g/lane) was fractionated on a 1% agarose-formaldehyde gel and transferred to a nylon membrane. The 1.1-kb EcoRI-BglII fragment in the coding sequence was labeled with 32 P by random priming and used as a probe. Lanes represent poly(A) ϩ RNA from the tissues indicated. RNA size markers are shown on the left. n ϭ 13; p Ͻ 0.001). This inhibition was restored to 88 Ϯ 41% (n ϭ 6) by re-injecting kcr1 cRNA (Fig. 3A, top). The suppression and rescue effects of KCR1 were restricted to I K(ni) , since identical effects were not observed for I K(A) (Fig. 3A, bottom) or voltage-dependent sodium currents in 4 -9 oocytes (data not shown). The wave shape of the rescued I K(ni) was indistinguishable from that of the original I K(ni) (Fig. 3A), as was K ϩ selectivity (Fig. 3B) and the activation curve (Fig. 3C). Taken altogether, these results show KCR1 is a regulatory protein of the non-inactivating K ϩ channel.
Functional Interaction of KCR1 with Cloned ␣ Subunits of K ϩ Channels in Oocytes-Since we assumed that KCR1 could be a part of channel complexes producing I K(ni) , we examined several ␣ subunits of K ϩ channels, which have been reported to be expressed in cerebellum (16, 28 -30) as a potential target of functional interaction with KCR1. Co-expression of KCR1 with the rKv1.2, rKv2.1, or mKv3.1a channel (28,31,32) in Xenopus oocytes did not have any appreciable effect on the functional expression and activation kinetics of these channels (each coexpression in 7-9 cells; data not shown). These results suggest that such combinations are not responsible for I K(ni). Next, we focused on r-EAG channels, which share basic properties with I K(ni) (10) and are strongly expressed in cerebellar granule cells (16). Furthermore, I K(ni) in oocytes was inhibited to 51 Ϯ 6.6% (n ϭ 8) with antisense cDNA of r-EAG, confirming that r-EAG is a component of the channels producing I K(ni) . The activation kinetics of r-EAG currents expressed in oocytes could be satisfactorily described with two time constants similar to those previously reported (10). The faster time constant was 19.1 Ϯ 1.4 ms (n ϭ 16), and the slower one was 395 Ϯ 26 ms (n ϭ 16), when the currents were evoked by a depolarizing step to 0 mV for 1 s from a holding potential of Ϫ60 mV. The r-EAG currents co-expressed with KCR1 also yielded two time constants: 19.0 Ϯ 1.0 and 377 Ϯ 21 ms (n ϭ 17). Although the time constants for the two types of currents did not change drastically, the r-EAG currents in the co-expressed oocytes showed the faster activation (Fig. 4A). We could not find any changes in the deactivation of r-EAG currents in co-expressed cells (Fig. 4A).
These rapid activations of co-expressed currents were accompanied by an increase in the fraction of fast components over that of slow components (Fig. 4B). The midpoint voltage (V1 ⁄2 ) was Ϫ56.8 Ϯ 2.5 mV (n ϭ 6) for control oocytes with expression of the r-EAG channel alone and Ϫ68.4 Ϯ 5.8 mV (n ϭ 5) with co-expression of KCR1. Thus, the expression of KCR1 caused a ϳ10 mV hyperpolarizing shift (p Ͻ 0.01) in the voltage dependence of the occupancy by the faster component, whereas the steady state activation remained unchanged (Fig. 4C). The half-activation voltages (V1 ⁄2 ) were ϩ2.5 Ϯ 1.7 mV (n ϭ 10) for control oocytes and ϩ2.4 Ϯ 1.2 mV (n ϭ 11) with KCR1. This implies that the KCR1 protein accelerates the rate of transition in r-EAG channels toward the open state.
Fast Activation Kinetics of r-EAG Accelerated by KCR1 in COS-7 Cells-The above data were confirmed in a mammalian   4. Facilitation of activation kinetics of r-EAG channels by KCR1 in oocytes injected with r-eag cRNA alone or with kcr1 cRNA. A, currents elicited by 200-ms pulses from Ϫ40 to ϩ60 mV in 20-mV steps from a holding potential of Ϫ100 mV followed by a constant pulse at Ϫ60 mV. Note that the rapidly rising phase of the outward currents in a co-expressed oocyte (right) is more prominent than that of the r-EAG currents (left). B, the fraction of the total current amplitude at 1 s contributed by the fast component (F f ) plotted against the prepulse potential. The oocytes were held for 1 s at various potentials (V prepulse) between Ϫ140 and Ϫ20 mV, then depolarized with a 1-s test pulse to 0 mV. For each oocyte, the traces were described by two time constants, and the fractions of the fast component were fitted to a Boltzmann function. Values represent the means Ϯ S.E. of five to six oocytes. See text for fitted values. C, normalized conductance-voltage relationship of r-EAG current expressed alone (n ϭ 10) or with KCR1 (n ϭ 11) at 500 ms of the test pulses from a holding potential of Ϫ60 mV to Ϫ40ϳϩ60 mV for two types of oocytes. Smooth curves represent best fits to a Boltzmann function.
cell expression system with much smaller membrane capacitance. In COS-7 cells, both r-EAG and r-EAG ϩ KCR1 currents showed two activating components with fast (12.9 Ϯ 1.8 ms (n ϭ 12) and 10.3 Ϯ 1.8 ms (n ϭ 13)) and slow (146 Ϯ 15 ms and 125 Ϯ 13 ms) time constants, when the currents were evoked by stepped depolarization from a holding potential of Ϫ80 mV to ϩ40 mV for 500 ms. Again, a marked difference was that the co-expressed currents showed rapid activation resulting from the reduction of the slow component.
Since the slow component of the r-EAG current has been reported to be affected by holding potentials and extracellular Mg 2ϩ (21), we measured the voltage and Mg 2ϩ dependence of the slow component in COS-7 cells. The relative amplitudes of the slow component seemed to decrease as the prepulses were being depolarized (Fig. 5, A-C), with the shift from the midpoint voltage of Ϫ65.0 Ϯ 1.2 mV (n ϭ 11) for r-EAG currents to that of Ϫ75.5 Ϯ 1.4 mV (n ϭ 6) for r-EAGϩKCR1 currents (p Ͻ 0.001). Thus, the expression of KCR1 caused a 10 mV hyperpolarizing shift in the voltage dependence of the slow kinetics in the presence of Mg 2ϩ , with the slope factor unchanged: 8.2 Ϯ 0.7 mV for r-EAG currents and 9.1 Ϯ 0.3 mV for r-EAGϩKCR1 currents. This hyperpolarizing shift was eliminated in the absence of Mg 2ϩ . The midpoint voltage in this case was Ϫ137.4 Ϯ 2.6 mV (n ϭ 8) for r-EAG currents and Ϫ139.5 Ϯ 4.2 mV (n ϭ 5) for r-EAGϩKCR1 currents. The modifying effect of KCR1 on the activation kinetics of the r-EAG channels was further evidenced by an estimate of the number of Mg 2ϩ -dependent slow gating particles (Fig. 5D) (21). The power needed for fitting r-EAG currents at a prepulse of Ϫ160 mV was 3.5 Ϯ 0.2 (n ϭ 11) for r-EAG currents in the presence of 2 mM Mg 2ϩ , indicating nearly four gating particles were in the Mg 2ϩ -dependent slow activating mode. However, the corresponding value was 2.5 Ϯ 0.1 (n ϭ 5) for r-EAGϩKCR1 currents, which suggests that at least one particle of the r-EAG channels was released to the Mg 2ϩ -independent fast activating mode by the expression of KCR1.
Molecular Association between KCR1 and r-EAG-The molecular association between KCR1 and r-EAG channels was examined by in vitro binding of recombinant fusion proteins. The GST fusion protein containing the S6-S7 loop and the C-terminal tail of KCR1 carrying 17 amino acids were tested for binding to a MBP fusion protein containing the C-terminal tail of the r-EAG protein and to the MBP alone. Binding of a GST-KCR1 tail fusion protein with a MBP-r-EAG fusion protein could be detected (Fig. 6A, lane 1), but not with a GST fusion protein containing the S6-S7 loop (data not shown). Progressive deletion analysis of the C-terminal tail of r-EAG channels revealed that the 50 amino acids (788 -837th; see Fig.  1 by Ludwig et al. (16)) are sufficient for interaction (Fig. 6,  A-C). However, the fusion proteins lacking amino acid residues 837-962 showed weaker intensity than those containing that region. DISCUSSION I K(ni) , one type of low-threshold non-inactivating K ϩ current derived from cerebellum, which was nearly identical to I K(SO) was characterized in this study. We isolated the kcr1 cDNA from our rat cerebellar cDNA library by suppression cloning and showed that KCR1 facilitates the functional expression of I K(ni) . The amino acid sequence of KCR1 did not contain the K ϩ channel pore sequence (23) or the voltage sensor region seen in voltage-gated K ϩ channels. Structural characteristics of the 12 transmembrane regions in KCR1 appeared to be somewhat similar to those of the transporter superfamily. Since no homology was found between KCR1 and known transporters, it appears that KCR1 must be a new mammalian protein that does not belong to the family of the ␣ subunit of K ϩ channels or transporters.
The translation products in vitro showed the 2-kDa shift, when microsomes were present. Since this shift was partially removed by ␤-elimination (Fig. 2), some ester-linked modification such as phosphorylation, acylation, and O-glycosylation might contribute to the shift (33). However, as O-glycosylation occurs only in the Golgi apparatus, the most likely modification in the presence of microsomes is acylation on serine or threonine residues or phosphorylation on a site exposed when KCR1 translocates into the microsomal membrane. On the contrary, this shift was resistant to endoglycosidase H treatment, showing that the two potential N-glycosylation sites are not glycosylated and suggesting that these sites are located in the cytoplasmic side. On the basis of these interpretations, we propose the membrane topology of KCR1 as shown in Fig. 7.
Since co-expression of cerebellar mRNA with kcr1 increased the outward current in our experiments, KCR1 may function as an up-regulator of the expression of I K(ni) . Similar regulators in ion channel expression have already been reported: the ␤ subunit (Kv␤2) for Shaker-type K ϩ channels (34), the minK (IsK) for K v LQT1 and HERG voltage-gated K ϩ channels (35)(36)(37), and the cysteine string protein for voltage-dependent Ca 2ϩ channels (15,38).
By using r-EAG as a candidate target, we could demonstrate a protein-protein interaction of KCR1. As shown in Figs. 6 and 7, the C-terminal tail of KCR1 peptide of 17 (458 -474th) amino acid residues interacted with the C-terminal region at the 788 -837th amino acid residues of r-EAG. We do not know whether other regions contribute to the interactions between r-EAG and KCR1. However, the presence of r-EAG fusion proteins containing amino acid residues 837-962 results in stronger binding than that of two fusion proteins lacking that region, indicating that other regions of r-EAG may also contribute to the interaction with KCR1. The interaction of KCR1 and r-EAG at the C-terminal region might be specific, because the 788 -837th amino acids of r-EAG are not conserved among other K ϩ channels, and Kv1.2, Kv2.1, or Kv3.1 did in fact not display any functional interaction with KCR1. Furthermore, the shift of the activation kinetics of r-EAG currents caused by the coexpression of KCR1 constitutes functional evidence that molecular interaction of KCR1 with r-EAG channels is taking place. We therefore assume that the association of KCR1 with r-EAG channels leads to faster activation by reducing the Mg 2ϩ -dependent slow component.
The finding that both r-eag and kcr1 antisense ssDNAs could suppress I K(ni) suggests that r-EAG and KCR1 are essential components in the channel complexes producing I K(ni) . On the basis of these observations, we propose the hypothesis that KCR1 combines with r-EAG in forming the K ϩ channels to produce I K(SO) in cerebellar granule neurons. In addition, it would be of interest to examine whether KCR1 can bind to other molecules than EAG in different tissues, because the kcr1 message is also expressed widely in peripheral tissues, where r-eag mRNA is not expressed.
In conclusion, we have isolated the kcr1 gene, which encodes a novel membrane protein, from the cerebellar cDNA library. This KCR1 membrane protein modulates I K(ni) expression and specifically binds to r-EAG at the C-terminal region. Therefore, KCR1 is thought to be a new subunit of voltage-dependent K ϩ channels in rat brain. Further validation is necessary to show that KCR1 and EAG are colocalized in the same single neuron.