New Modulatory α Subunits for Mammalian ShabK+ Channels*

Two novel K+ channel α subunits, named Kv9.1 and Kv9.2, have been cloned. The Kv9.2 gene is situated in the 8q22 region of the chromosome. mRNAs for these two subunits are highly and selectively expressed in the nervous system. High levels of expressions are found in the olfactory bulb, cerebral cortex, hippocampal formation, habenula, basolateral amygdaloid nuclei, and cerebellum. Interestingly Kv9.1 and Kv9.2 colocalized with Kv2.1 and/or Kv2.2 α subunits in several regions of the brain. Neither Kv9.1 nor Kv9.2 have K+ channel activity by themselves, but both modulate the activity of Kv2.1 and Kv2.2 channels by changing kinetics and levels of expression and by shifting the half-inactivation potential to more polarized values. This report also analyzes the changes in electrophysiological properties of Kv2 subunits induced by Kv5.1 and Kv6.1, two other modulatory subunits. Each modulatory subunit has its own specific properties of regulation of the functional Kv2 subunits, and they can lead to extensive inhibitions, to large changes in kinetics, and/or to large shifts in the voltage dependencies of the inactivation process. The increasing number of modulatory subunits for Kv2.1 and Kv2.2 provides an amazingly new capacity of functional diversity.

Two novel K ؉ channel ␣ subunits, named Kv9.1 and Kv9.2, have been cloned. The Kv9.2 gene is situated in the 8q22 region of the chromosome. mRNAs for these two subunits are highly and selectively expressed in the nervous system. High levels of expressions are found in the olfactory bulb, cerebral cortex, hippocampal formation, habenula, basolateral amygdaloid nuclei, and cerebellum. Interestingly Kv9.1 and Kv9.2 colocalized with Kv2.1 and/or Kv2.2 ␣ subunits in several regions of the brain. Neither Kv9.1 nor Kv9.2 have K ؉ channel activity by themselves, but both modulate the activity of Kv2.1 and Kv2.2 channels by changing kinetics and levels of expression and by shifting the half-inactivation potential to more polarized values. This report also analyzes the changes in electrophysiological properties of Kv2 subunits induced by Kv5.1 and Kv6.1, two other modulatory subunits. Each modulatory subunit has its own specific properties of regulation of the functional Kv2 subunits, and they can lead to extensive inhibitions, to large changes in kinetics, and/or to large shifts in the voltage dependencies of the inactivation process. The increasing number of modulatory subunits for Kv2.1 and Kv2.2 provides an amazingly new capacity of functional diversity.
Voltage-gated potassium channels (Kv) 1 form the largest and most diversified class of ion channels. These proteins are present in both excitable and nonexcitable cells. Their main functions are associated with the regulation of the resting membrane potential and the control of the shape and frequency of action potentials (1,2). K ϩ channel functions are included in very diverse processes such as neuronal integration, cardiac pacemaking, muscle contraction, and hormone secretion in excitable cells (3), as well as in cell proliferation, cell volume regulation, and lymphocyte differentiation (4). Outward rectifiers constitute a large class of voltage-dependent K ϩ channels. They have six transmembrane domains (S1-S6), one very positively charged (S4), and a typical pore region situated between S5 and S6 (5)(6)(7)(8). Sequence similarities between members of the Kv family were initially used to define the different subfamilies of ␣ subunits. The different members within a given subfamily share only a percentage of 30 -50% with members of others subfamilies. To date 20 functional voltage-gated potassium channels ␣ subunits have been described. They belong to six subfamilies designated Kv1 (Shaker), Kv2 (Shab), Kv3 (Shaw), Kv4 (Shal), KvLQT, and EAG. Another family (Kv7) has a single member which has been cloned from Aplysia (9). The diversity of potassium channel functions comes from the diversity of potassium channel genes and is increased by alternate splicing (10,11), regulatory ␤ subunits (12)(13)(14) and heteromultimerization between the different ␣ subunits of the same subfamily (15)(16)(17) or sometimes between different subfamilies (18,19).
This study reports the cloning and brain localization of two new neuronal modulatory ␣ subunits that define a new structural family designated as Kv9. Kv9.1 and Kv9.2 are new modulators of Kv2.␣ subunits but have no functional activity by themselves. The study also shows that Kv5.1 and Kv6.1 are also potent regulators of channels of the Kv2 family. The functional properties of Kv9.1/Kv2.␣ and Kv9.2/Kv2.␣ heteromultimers are compared with those of Kv5.1/Kv2.␣ and Kv6.1/ Kv2.␣ assemblies.

EXPERIMENTAL PROCEDURES
Cloning of Kv9.1 and Kv9.2-A BLAST search of the expressed sequence tag (EST) data base, using the query peptide sequence of Kv8.1, reveals two groups of matching EST coding for two new homologues of Kv channels. Sense and antisense oligonucleotides corresponding to the extremities of these two EST groups were synthesized (Genosys) and used in two polymerase chain reaction amplifications on mouse brain cDNA. The sequencing of cloned polymerase chain reaction products allowed the isolation of Kv9.1 and Kv9.2 probes that were used to screen 5 ϫ 10 5 clones of a mouse brain cDNA library constructed with ZAP II vector (Stratagene). Probes ([␣-32 P]dCTP-labeled Kv9.1 and Kv9.2 fragments) were produced with a random primers kit from Amersham Corp. Filters were hybridized in 50% formamide, 5 ϫ SSC, 4 ϫ Denhardt's solution, 0.1% SDS, and 100 g/ml denatured salmon sperm DNA at 45°C overnight and washed to a final stringency of 0.2 ϫ SSC, 0.1% SDS at 55°C. Three independent clones for Kv9.1 and one for Kv9.2 were isolated. The cDNA inserts were characterized by restriction enzyme analysis and by partial or complete DNA sequencing on both strands by the dideoxy nucleotide method using an automatic sequencer (Applied Biosystems model 373A). Two Kv9.1 clones (2.2-2.3 kb long) and the Kv9.2 insert (5 kb long) were shown to contain a full-length ORF. The third Kv9.1 clone was identical but incomplete in its 5Ј-coding sequence.
Northern Blot Analysis-A mouse multiple tissue Northern blot (CLONTECH) was probed with the 32 P-labeled inserts of pBS Kv9.1 or pBS Kv9.2 according to the manufacturer's protocol.
In Situ Hybridization-Experiments were performed on adult Swiss mice using standard procedures (29). The dissected organs (brains, spinal cords, and retina) were fixed in ice-cold 4% (w/v) paraformaldehyde, 0.1 M sodium phosphate-buffered solution (PBS, pH 7.4) for 8 h and then immersed overnight at 4°C in a 20% sucrose, PBS solution. Frozen sections (10 m) were cut on a Cryostat (Leica) at Ϫ25°C, collected on 3-aminopropylethoxysilane-coated slides, and stored at Ϫ20°C until use. Specific antisense cRNA probes were generated with T7-RNA polymerase (Boehringer Mannheim) by in vitro transcription using D-[ 33 P]UTP (3000 Ci/mmol, ICN Radiochemicals), from EcoRIlinearized plasmid containing a 900-base pair fragment of Kv9.1 cDNA or a 560-base pair fragment of Kv9.2 cDNA in the 3Ј-untranslated sequence and in the cytoplasmic C-terminal coding sequence, inserted into pBluescript SK Ϫ . Sections were treated consecutively with 0.1 M glycine in PBS for 10 min, PBS for 3 min, 5 g/ml proteinase K diluted in 0.1 M Tris, 50 mM EDTA (pH 8.0) for 15 min at 37°C, 4% paraformaldehyde, and PBS (pH 7.2) for 5 min. Slides were then rinsed for 10 min in PBS, acetylated for 10 min in 0.25% acetic anhydride in 0.1 M triethanolamine and dehydrated. Hybridization was carried out overnight at 55°C in hybridization buffer (50% deionized formamide, 10% dextran sulfate, 500 g/ml denatured salmon sperm DNA, 1% Denhardt's solution, 5% Sarcosyl, 250 mg/ml yeast tRNA, 20 mM dithiothreitol, 20 mM NaPO 4 in 2 ϫ SSC, and the radiolabeled probe (at 0.2 ng/ml with specific activities of 8 ϫ 10 8 dpm/g). Following hybridization, sections were washed in 4 ϫ SSC for 15 min and then twice in 1 ϫ SSC for 30 min at 60°C, treated with RNase A (5 g/ml) in 2 ϫ SSC for 15 min at 37°C, and washed twice with 1 ϫ SSC for 30 min, followed by two 15-min washes in 0.1 ϫ SSC at 30°C. Specimens were then dehydrated, air-dried, and exposed to Amersham ␤-max Hyperfilm for 5 days at 4°C. Selected slides were dipped in Amersham LM1 photographic emulsion and exposed for 3 weeks at 4°C and then developed in Kodak D-19 for 4 min. All slides were counterstained with hematoxylin/ eosin. For control experiments, adjacent sections were hybridized with corresponding sense riboprobes or digested with RNase before hybridization.
Xenopus laevis were purchased from C.R.B.M. (Montpellier, France). Pieces of the ovary were surgically removed, and individual oocytes were dissected away in a saline solution (ND96) containing 96 mM NaCl, 2 mM KCl, 1.8 mM CaCl 2 , 2 mM MgCl 2 , and 5 mM HEPES at pH 7.4 with NaOH. Stage V and VI oocytes were treated for 2 h with collagenase (1 mg/ml, type Ia, Sigma) in ND96 to discard follicular cells. cRNA solutions were injected (50 nl/oocyte) using a pressure microinjector (InjectϩMatic, Switzerland). We have previously shown that a high level of expression of some cloned K ϩ channels could lead to high magnitude K ϩ currents with major kinetic and voltage-dependence modifications when compared with currents of lower intensity (32,40). Therefore, particular attention was paid to compare currents of similar and relatively low intensities (under 5 A for a test pulse at ϩ50 mV). cRNAs were injected at 0.5 ng/oocyte for Kv2.1 and 10 ng/oocyte for Kv2.2.
The oocytes were kept at 19°C in the ND96 saline solution supplemented with gentamycin (5 g/ml). Oocytes were studied within 2-4 days following injection of cRNA. In a 0.3-ml perfusion chamber, a single oocyte was impaled with two standard glass microelectrodes (1-2.5-megohm resistance) filled with 3 M KCl and maintained under a voltage clamp using a Dagan TEV 200 amplifier. Stimulation of the preparation, data acquisition, and analysis were performed using pClamp software (Axon Instruments, Burlingame, CA). All experiments were performed at room temperature (21-22°C) in ND96 solutions.
Determination of the Biophysical Parameters-The activation phase of currents elicited by voltage steps to ϩ30 mV, 1 s in duration, were fitted with a single exponential and the time constant act was determined. Inactivation curves (see Figs. 5, E and F, and 7, E and F) were fitted with a Boltzman distribution of the form I ϭ 1/1 ϩ exp(V 1/2 inact / k inact ). V 1/2 inact is the potential of half-inactivation and k inact is the slope factor. The tail currents were recorded at Ϫ40 mV after a 500-ms long prepulse to ϩ50 mV, in standard ND96 solution. These currents were fitted with a single exponential characterized by a time constant tail .
Cell Culture and Transfection-COSm6 cells were grown in Dulbecco's modified Eagle's medium (Life Technologies, Inc.) supplemented with 10% fetal calf serum and antibiotics (60 g/ml penicillin, 50 g/ml streptomycin). Two days before transfection 15 ϫ 10 3 cells were plated on cover glasses onto 15-mm Petri dishes. The cells were transfected by a modification of the DEAE-dextran/chloroquine method (41) using 1 or 2 g of supercoiled DNA.
Indirect Immunofluorescence Microscopy on Transfected COS Cells-Two days after transfection the cells grown on glass coverslips were fixed for 15 min with 4% (v/v) paraformaldehyde in PBS. After rinsing twice with PBS, cells were permeabilized by incubation for 10 min with 0.1% Triton X-100. The sites of nonspecific binding were blocked by 2-h incubation with 5% goat serum, 2% bovine serum albumin in PBS at room temperature. The cells were then incubated for 2 h with a mixture of anti-FLAG M2 monoclonal antibody (1/150 dilution, Kodak) or T7.Tag monoclonal antibody (1/250, NOVAGEN) with a BS solution (2% bovine serum albumin in PBS), followed by washing with PBS and incubation for 1 h with fluorescein isothiocyanate-conjugated goat antimouse Ig (1/200, Sigma) in BS. After washing in PBS then in 10 mM Tris-HCl, pH 7.5, cells were mounted in Vectashield Medium (Vector Laboratories, Inc.) and observed with a Leitz Aristoplan microscope (Wild Leitz) using an interference blue (fluorescein isothiocyanate) filter and a 40ϫ lens.
Amino Acid Sequence Alignments and Phylogenetic Tree-Computer analyses were performed using Genetics Computer Group software. The phylogenetic tree (see Fig. 1C) and the percentage of identity (Table  I) were generated using the PILEUP multiple sequence analysis program with the conserved region corresponding to the amino acids from the position 24 -435 of Kv9.1 (starting from residues NVGG to Y). The members of subfamilies used and their accession numbers are as follows: mKv1.

Cloning of Two New Putative Voltage-gated K ϩ Channels
Which Define the New Kv9 Subfamily Related to the Modulatory ␣ Subunits-The Kv8.1 subunit is a new K ϩ channel subunit which is inactive by itself and which has recently been described to regulate both active subunits of the Kv2 potassium channel subfamily (21,22). This result has suggested the existence of a new class of regulatory ␣ subunits. To identify potential genes coding for parent sequences, the peptide sequence of the Kv8.1 subunit was used to search related sequences in GenBank TM data base using the BLAST sequence alignment program (42). We identified two groups of human EST encoding a portion of a new ␣ subunit (named Kv9.1) from the S3 to the S6 transmembrane domain for the EST group corresponding to the accession numbers H18119, H43697, H49525, H42586, H49759, H2O365, H18164, and H42586, and another new ␣ subunit (named Kv9.2) from the S2 to S4 transmembrane domains for the EST group R19352, H19204, and R34920. It was then postulated that these two groups of EST were partial copies of mRNAs coding for two new subunit homologues of the Kv8.1 K ϩ channel. Two DNA probes derived from these two groups of sequences were used to screen a mouse cDNA brain library. Three clones for Kv9.1 and one clone for Kv9.2 were obtained. Two of the three independent Kv9.1 clones found were full-length and presented the same expected open reading frame (ORF). Nevertheless, these two clones differed in their 5Ј-noncoding sequence upstream of the nucleotide Ϫ21. The 5.0-kb cDNA insert of the Kv9.2 clone also comprises the expected ORF. The methionine initiation codon for the longest ORF is preceded by several in-frame termination triplets. In addition, nucleotide sequences surrounding the initiation codon of both Kv9.1 and Kv9.2 clones correspond to the Kozak consensus sequence (43).
The ORFs of Kv9.1 and Kv9.2 encode proteins of 497 and 477 amino acids, respectively, with a calculated molecular mass of 54.9 and 54.3 kDa (Fig. 1A). Protein sequences reveal that all the structural characteristics of outward rectifier voltage-gated K ϩ channel ␣ subunits (Kv) are conserved in Kv9.1 and Kv9.2, i.e. six putative transmembrane segments (S1 to S6), a transmembrane region (S4) showing five positively charged amino acids and a conserved pore-forming region (named H5 or P domain). As indicated in Fig. 1A, Kv9.1 and Kv9.2 subunits contain several putative phosphorylation sites located in the cytoplasmic regions for protein kinase C, cAMP-dependent protein kinase, Ca 2ϩ -calmodulin kinase II, casein kinase II, and tyrosine kinase. No N-glycosylation sequence was detected.
The identity between these two new ␣ subunits is about 58% and they have only 33-49% identity with other Kv channel subunits (Table I). Thus, they belong to the same subfamily (44). The phylogenetic tree presented in Fig. 1C indicates that the Kv9 ␣ subunits have a common ancestor with other "nonfunctional" ␣ subunits such as Kv5.1, Kv6.1, and Kv8.1, as well as with the Kv2 family.
Distribution of mRNAs Encoding Kv9.1 and Kv9.2 Subunits-Northern blot analysis presented in Fig. 2 shows that Kv9.1 and Kv9.2 mRNAs are expressed only in the brain. Specific probes detected two transcripts for Kv9.1 with an estimated size of 2.2 and 2.7 kb and one for Kv9.2 of approximately 5.3 kb. No expression was observed in heart, spleen, lung, liver, skeletal muscle, kidney, or testis. The two Kv9.1 transcripts probably represent alternatively spliced variants in the 5Ј-noncoding region as shown by sequencing of the different clones.
The regional distribution of Kv9.1 and Kv9.2 mRNA was studied from analysis of x-ray film images and emulsion-coated sections of adult mouse brains cut in sagittal and coronal planes and detected by in situ hybridization. Specific cRNA probes revealed a wide and heterogeneous expression pattern in adult mouse brain (Fig. 3A). The heterogeneous distribution of the hybridization signal and the observation of emulsiondipped sections suggested a neuronal localization of the transcripts. Sense probes did not show significant hybridization (data not shown). Examination of adjacent sections hybridized with Kv9.1 and Kv9.2 cRNA probes indicated that the distribution of Kv9.1 transcripts was strikingly similar to that ob- served for Kv9.2, with highest expression levels in the main olfactory bulb, cerebral cortex, hippocampal formation, habenula, basolateral amygdaloid nuclei, and cerebellum. In the olfactory system (Fig. 3B), cells expressing Kv9.1 and Kv9.2 mRNA were localized in the glomerular cell layer, the densely packed internal granular layer, and the thin mitral and internal plexiform cell layers. Within the glomerular layer, transcripts were restricted to the periglomerular cells whose processes extend into the glomerula. In contrast, light diffuse labeling was observed in the external plexiform layer and the olfactory nerve layer. In addition to the main olfactory bulb, Kv9.1 and Kv9.2 expression was also prominent in the anterior olfactory nuclei and the piriform (primary olfactory) cortex. In the hippocampus (Fig. 3C), Kv9.1 and Kv9.2 mRNAs were strongly expressed in dentate granule cells and hilar neurons as well as in CA1-CA3 pyramidal cells. In addition, intense hybridization signals were observed in large interneurons located in stratum oriens and radiatum of all subfields. Many cells in the subiculum and enthorinal cortex of the hippocampal formation were also found to be positive. In the neocortex (Fig.  3D), Kv9.1 and Kv9.2 gene expression was observed in all cortical areas with a labeling pattern reflecting the laminar structure of cell distribution. These two transcripts were strongly expressed in the large cortical pyramidal neurons as compared with the small cell bodies throughout the cortical layers, which may represent nonpyramidal neurons, oligo-, astro-, or microglia cells and which failed to show hybridization signals. Kv9.1 and Kv9.2 mRNAs were also densely distributed throughout most of the amygdala, including the cortical, lateral and basolateral nuclei. Within the thalamus, the medial and lateral habenula showed particularly high levels of Kv9.1 and Kv9.2. The ventromedial hypothalamic nucleus displayed an intense hybridization signal. A distinct laminar expression pattern was observed in the cerebellum (Fig. 3E). The most intense labeling was detected in the Purkinje and granule cell layers. The molecular layer was weakly labeled except for a few strongly positive cells that are scattered in the cerebellar molecular layer and may be stellate cells and/or basket cells. Large neurons in all deep cerebellar nuclei expressed moderate to high levels of Kv9.1 and Kv9.2 mRNAs. A diffuse expression was observed over most other regions including striatum, globus pallidus of the basal ganglia, substantia nigra, midbrain, and brainstem (pons and medulla). Kv9.1 and Kv9.2 transcripts were also intensely expressed in the rat retina with a distinct stratification pattern. High labeling was apparent in the inner nuclear layer (INL), composed of nuclei of Mü ller cells and amacrine, bipolar, and horizontal neurons, in the retinal ganglion cell layer (RGC) and in the photoreceptor inner segments (IS) (Fig. 3F). The spinal cord showed an intense hybridization signal in several neuronal types. This hybridization pattern was similar from the upper cervical to the sacral region. The majority of spinal cord neurons, including interneurons and preganglionic neurons, in addition to motor neurons and sensory neurons strongly expressed Kv9.1 and Kv9.2 mRNAs (Fig. 3G).  (36) channels. Fig. 4A shows that the mean current amplitudes of Kv2.1 were decreased by the coexpression with Kv5.1 and also, but to a lesser extent, by Kv9.1, Kv9.2, and Kv6.1. To verify the specificity of this inhibition we have coexpressed a similar amount of the FMRF-activated Na ϩ channel (FaNaCh) with the Kv2.1 subunit (45). This particular ligand-gated Na ϩ channel is well expressed in oocytes but it remains completely silent until activation with the FMRF peptide (45). The coexpression of FaNaCh was without any effect on the amplitude of Kv2.1 currents (Fig. 4A). All subsequent experiments included a control for the expression of FaNaCh after the K ϩ currents were recorded. A strong inhibition of Kv2.2, the other functional Shab channel, was also observed after coexpression with Kv9.1 and Kv9.2 (Fig. 4B). Kv5.1 also produced an extensive inhibition of the Kv2.2 current, while Kv6.1 only slightly diminished the current amplitude (as in the FaNaCh control) (Fig. 4B). The inhibition of the Kv2.1 current can be drastically enhanced with higher cRNAs ratio, for example in a 1/10 ratio of Kv2.1 versus Kv9.1 or Kv9.2, Kv2.1 currents were decreased to only 5.4 Ϯ 0.1% (Kv9.1) and 5.3 Ϯ 0.2% (Kv9.2) of the control current (n ϭ 9). The same ratios could not be used for Kv2.2  Kv2.1 and Kv2.2. The potential of half-inactivation of Kv2.1 (V 1/2 inact ϭ Ϫ20.5 Ϯ 0.9 mV, n ϭ 7) is shifted toward negative values in the presence of Kv9.1 or Kv9.2 (Ϫ33.3 Ϯ 0.2 mV, n ϭ 4, and Ϫ35.7 Ϯ 3.2 mV, n ϭ 5, respectively) (Fig. 5E). The potential for half-inactivation of Kv2.2 (V 1/2 inact ϭ Ϫ16.6 Ϯ 1.1 mV, n ϭ 13) is also drastically shifted toward negative values with both Kv9.1 and Kv9.2 (Ϫ32.1 Ϯ 1.6 mV, n ϭ 4, and Ϫ31.5 Ϯ 0.9 mV, n ϭ 4, respectively) (Fig. 5F). The relative currents obtained at potentials positive to 0 mV show a striking increase which does not follow the Boltzman distribution. This behavior has also been observed in COS cells (not shown) and could be due to reactivation by an intracellular mediator, perhaps calcium.
We have recently demonstrated that the Kv8.1 subunit is electrically silent because, when expressed alone, this subunit is unable to reach the plasma membrane (22). We have assumed that the same explanation holds for the Kv9 subunits. To test this hypothesis, Kv9.1 and Kv9.2 subunits have been tagged with an epitope, transfected, and localized by indirect immunofluorescence using Triton-permeabilized COS cells (Fig. 6). N-terminal tagged Kv2.2 subunit-expressing cells show a typical surface labeling, as expected for a subunit that expresses a K ϩ current when transfected in COS cells (Fig. 6A). Conversely, COS cells expressing Kv9.1 (Fig. 6B) and Kv9.2 subunits (Fig. 6C), which show a strong fluorescence localized at the perinuclear region as well as in a fine reticular network extending through the cytoplasm. No surface labeling of Kv9.␣ subunits was observed in these cells, in agreement with the impossibility to record K ϩ currents from these transfected cells.
The effects of Kv5.1 (IK8) and Kv6.1 (K13) on the biophysical properties of Kv2.1 and Kv2.2 expressed in Xenopus oocytes are presented in Fig. 7. Fig. 7A shows that the normalized Kv2.1 currents are strongly modified by Kv6.1 and only slightly altered by Kv5.1. The time constant of activation of Kv2.1 ( act ϭ 15.6 Ϯ 0.9 ms, n ϭ 24) was increased with Kv5.1 (20.4 Ϯ 1.2 ms, n ϭ 5) and a significant increase was observed with Kv6.1 (29.2 Ϯ 2.0 ms, n ϭ 5). The percentage of remaining Kv2.1 current after a 9-s inactivation (33.4 Ϯ 1.2%, n ϭ 19) is higher with Kv5.1 (42.8 Ϯ 1.6%, n ϭ 5) and with Kv6.1 (66.9 Ϯ 1.5 ms, n ϭ 5). Fig. 7B shows that Kv2.2 currents are only slightly modified by Kv5.1 and Kv6.1. The time constant of activation of Kv2.2 ( act ϭ 21.8 Ϯ 0.5 ms, n ϭ 34) was slightly increased by Kv5.1 and Kv6.1 (24.5 Ϯ 2.3 ms, n ϭ 5, and 24.4 Ϯ 0.7 ms, n ϭ 9). The percentage of remaining Kv2.2 current after 9 s (46.7 Ϯ 1.2%, n ϭ 17) was increased with both Kv5.1 and Kv6.1 (60.5 Ϯ 2.7%, n ϭ 5, and 53.9 Ϯ 3.7%, n ϭ 6, respectively). Fig. 7, C and D, shows current-potential relationships for Kv2.1 and Kv2.2 in the presence and absence of the Kv5.1 or Kv6.1 subunits. The inhibition of the Kv2.1 and Kv2.2 K ϩ currents by Kv5.1 and the inhibition of Kv2.1 by Kv6.1 were observed at all potentials. Kv6.1 was without a significant effect on the Kv2.2 current at any potential. The activation thresholds as well as the half-potential for activation were not modified. Fig. 7, E and F, shows the current inactivation curves for Kv2.1 and Kv2.2. The potential for half-inactivation of Kv2.1 (V 1/2 inact ϭ Ϫ20.5 Ϯ 0.9 mV, n ϭ 7) is strongly shifted toward negative values in the presence of Kv5.1 or Kv6.1 (Ϫ60.2 Ϯ 1.1 mV, n ϭ 5, and Ϫ45.2 Ϯ 5.2 mV, n ϭ 5, respectively) (Fig. 7E). The potential for half-inactivation of Kv2.2 (Ϫ16.6 Ϯ 1.1 mV, n ϭ 13) is also strongly shifted toward negative values with both Kv5.1 and Kv6.1 (Ϫ45.7 Ϯ 4.1 mV, n ϭ 4, and Ϫ55.6 Ϯ 4.5 mV, n ϭ 4, respectively) (Fig. 7F). The slowing of inactivation due to the coexpression of Kv6.1 (observed at ϩ30 mV) in Fig.  7A is due to the rebound of the inactivation curve for positive potentials. 2. E, peak inactivated current recorded at a test pulse of ϩ30 mV after prepulses ranging from Ϫ100 to ϩ20 mV in 10-mV steps, 10 s in duration. The relative currents correspond to the ratios of measured currents at a given prepulse potentials to the maximum peak current recorded after the prepulse to Ϫ100 mV. Because the time constants of inactivation are very slow all inactivation curves are 10-s isochronal inactivation curves. Experimental data (mean Ϯ S.E., n ϭ 4) are fitted with a Boltzman distribution from Ϫ100 mV to 0 mV and extrapolated up to ϩ20 mV. The top trace represents the activity of Kv2.1 alone, bottom traces represent the Kv2.1 current when modulated by Kv9.1 or Kv9.2 subunits. F, same as in E with Kv2.2. G, normalized tail current elicited by a voltage pulse to Ϫ40 mV, after a 500-ms long prepulse to ϩ50 mV, in oocytes injected with Kv2.1 alone or coexpressed with Kv9.1 or Kv9.2. H, same as in G with Kv2.2. In E, F, G, and H, for a better analysis of the biophysical properties, the cRNA coding for Kv2.1 and Kv2.2 channels and for Kv9.1 and Kv9.2 modulatory subunits were injected in a channel/subunit ratio of 10:1, which gives a smaller inhibition. In all cases the holding potential was Ϫ80 mV. 1 to obtain sufficient currents. E, peak inactivated current recorded at a test pulse of ϩ30 mV after prepulses ranging from Ϫ100 to ϩ20 mV in 10-mV steps, 10 s in duration. The relative currents correspond to the ratios of measured currents at a given prepulse potentials to the maximum peak current recorded after the prepulse to Ϫ100 mV. Because the time constants of inactivation are very slow all inactivation curves are 10-s isochronal inactivation curves. Experimental data (mean Ϯ S.E., n ϭ 4) are fitted with a Boltzman distribution from Ϫ100 mV to 0 mV and extrapolated up to ϩ20 mV. Kv2.␣ subunits family (21)(22)(23). A newly described subunit in the Kv4.␣ family also has a modulatory role (24).
This study reports the isolation of two new neuronal K ϩ channels ␣ subunits, named Kv9.1 and Kv9.2 (Fig. 1A). They both belong to a new Kv9 K ϩ channel family ( Fig. 1C and Table  I). Like the Kv8.1 subunit, Kv9.1 and Kv9.2 subunits alone are electrically silent when expressed in oocytes and act as specific inhibitors of the Kv2.␣ channels activity when they are expressed at high levels. At lower levels, oocytes coexpressing Kv2.␣ and Kv9.␣ subunits exhibit novel K ϩ currents that display a marked shift in the voltage dependence of inactivation. In addition we have shown that Kv9.1 and Kv9.2, like Kv8.1 (22), are unable to reach the plasma membrane in COS cells (Fig. 6). This is probably why they are silent in these cells as well in Xenopus oocytes when they are expressed alone.
This work also shows that coexpression of Kv2.1 or Kv2.2 subunits with Kv5.1 or Kv6.1 subunits at relatively low concentrations gives rise to residual currents that display very important variations in voltage dependence of inactivation, kinetics of inactivation, and tail currents. At higher concentrations, Kv5.1 inhibits the expression of both Kv2.1 and Kv2.2 while Kv6.1 inhibits only Kv2.1.
Differences in time constants of activation, differences in inactivation kinetics, in voltage-dependencies of the inactivation process, and in the time constants of the tail current of both Kv2.1 and Kv2.2 channels are due to specific interactions between Kv2.␣ and the modulatory subunits as previously demonstrated for Kv8.1 (21) and Kv6.1 (23). Both the reduction of amplitude of expressed Kv2.1 and Kv2.2 currents and the change in their voltage-dependence of inactivation caused by coexpression with Kv9.1 and Kv9.2 subunits occur with minimal changes in inactivation kinetics in contrast to the effects of coexpressing Kv5.1 and Kv6.1 subunits with the same Kv2.␣ subunits in which a marked slowing of inactivation was observed. These functional differences suggest different interaction domains for the various modulatory subunits. The extent of the inhibitory effect by subunits belonging to the Kv9, Kv5, or Kv6 families depends on the level of expression of the modulatory subunit. The mechanism(s) underlying the specific reduction of the Kv2.␣ induced currents is not completely known. However, we have previously proposed for the Kv8.1 subunit that the composition of heteromultimeric channels formed by the assembly of active (Kv2.␣) might play an important role in this inhibition. A high stoichiometry of the Kv8.1 subunit would favor a retention in the endoplasmic reticulum (22).
The S6 region is one of the most conserved region in voltagegated potassium channels with six transmembrane domains. A comparison of the S6 transmembrane domains of the "nonfunctional" ␣ subunits (Kv5.1 (IK8), Kv6.1 (K13), Kv8.1, Kv9.1, and Kv9.2) with those of the functional ␣ subunits (Kv1 to Kv4 and Kv7.1) reveals important differences which are common to all the nonfunctional ␣ subunits class (Fig. 1B). This point is important since the S6 region has been described as the domain responsible for the regulatory function of the Kv8.1 subunit when it associates with members of the Kv2 subfamily (22). This is clearly now a distinctive feature that permits one to identify ␣ subunits (such as Kv5.1, Kv6.1, Kv8.1, Kv9.1, and Kv9.2), which cannot be functional by themselves but which have a modulatory function. Especially the second proline residue (proline 406 in Kv2.1 and proline 414 in Kv2.2) that is conserved in all the functional potassium channels is changed to a threonine residue in Kv6.1, Kv9.1, and Kv9.2, to an histidine residue in Kv5.1, and to an alanine in Kv8.1. Since proline residues are thought to cause a significant bend in ␣-helical structures (46) the replacement of the proline could have a major impact on the conformation of the channel protein. Several other residues conserved in functional subunits are also replaced in the modulatory subunits (Ala 408 in Kv9.1 replaces Gly 390 in Kv2.1) and two isoleucines replace residues that are a cysteine (Cys 394 in Kv2.1) and a valine (Val 398 in Kv2.1) in the Kv2 subfamily. In addition, two very conserved residues, valine 409 and asparagine 411 in Kv2.1, are replaced by various residues in the different modulatory subunits. There are also differences between the S6 domains of the different nonfunctional ␣ subunits, and these differences may be associated with different types of regulatory functions. Another striking common feature to all the nonfunctional ␣ subunits is the presence of a very short C-terminal sequence situated after the S6 segment, which is believed to be cytoplasmic. It is only 56 -89 amino acids long, depending on the regulatory ␣ subunits in comparison with a length of 385-444 amino acids for the two functional Kv2.␣ subunits.
Protein phosphorylation is an important mechanism in the modulation of K ϩ channels (47,48). Kv2.1 is modulated by cAMP-dependent kinase (49). In comparison with the homomultimers of Kv2.1 and Kv2.2, channels composed of heteromultimers comprising Kv2.␣ subunits and modulatory subunits of the Kv5, Kv6, Kv8, or Kv9 families present new potential sites for regulation by protein kinases. For example, Kv5.1 and Kv9.1 introduce a new potential site for protein tyrosine kinase that is situated in the cytoplasmic N terminus. Another example is the introduction of a potential regulatory site for cAMP-dependent kinase that is situated in the S4S5 loop of Kv6.1, in heteromultimers containing this particular subunit. The different potential possibilities of phosphorylation of each modulatory subunit are presented in Table II. In most cases the short modulatory ␣ subunits have a lower number of potential sites of regulation than functional Kv2.␣ subunits. Information regarding regulation of Kv2.1 and Kv2.2 by kinases is still limited.
Kv9.1 and Kv9.2, like Kv8.1 are very highly expressed in the brain (Fig. 2) while they are not detectable in the other tissues. In addition, Kv9.1 and Kv9.2 have very similar distributions in brain, retina, and spinal cord. This distribution coincides with the expression patterns of Kv2.1 and Kv2.2 in several regions of the brain as the olfactory bulb, the cerebral cortex, the hippocampus, the cerebellum, and also in the retina (36). A very similar localization of Kv9.␣ and Kv2.␣ subunits has also been observed at the cellular level in granular cells of olfactory bulb, in Purkinje and granular cells of the cerebellum, in pyramidal cells of the hippocampus, and in granular cells of the dentate gyrus (50) (this work). A particularly intriguing conclusion of this work is that the regulation of the K ϩ channel function of the Kv2.␣ subunits by Kv9.1 and Kv9.2 seems to be required only in the nervous system, since Kv9 mRNAs are not detected in the peripheral tissues while Kv2.1 and Kv2.2 subunits, which are highly expressed in the brain, are also expressed in kidney, heart, retina, and skeletal muscle (36). The EST R19352, corresponding to the Kv9.2 gene, has been mapped to the 8q22 region of the human genome (51). There is a growing number of pathological situations associated with mutations in diverse K ϩ channel genes (52)(53)(54)(55), and it would not be surprising if diseases were also associated with this class of channel subunit. The only unsolved genetic disorder which had been uniquely assigned to 8q22 (56) is the Cohen syndrome. Patients with this syndrome suffer from hypotonia, mental deficiency, as well as from a retinal degeneration. 2 It will be of interest to analyze more closely a possible relation between the Kv9.2 gene and the Cohen syndrome.
In conclusion, it is now clear that the properties of Kv2.␣ potassium currents can change in many different ways in the presence of the different modulatory ␣ subunits. Each type of cells expressing Kv2.1 or/and Kv2.2 could in principle acquire its own electrophysiological characteristics depending on the nature and stoichiometry of the associated ␣ subunits of the Kv5, Kv6, Kv8, and Kv9 family. These subunits could also serve to direct the localization of a particular Kv2.1 or Kv2.2 subunit to specific regions (cell body, dendrites, terminals, and so forth) of a given neuronal cell.