Cloning of a New Mouse Two-P Domain Channel Subunit and a Human Homologue with a Unique Pore Structure*

Mouse KCNK6 is a new subunit belonging to the TWIK channel family. This 335-amino acid polypeptide has four transmembrane segments, two pore-forming domains, and a Ca2+-binding EF-hand motif. Expression of KCNK6 transcripts is principally observed in eyes, lung, stomach and embryo. In the eyes, immunohistochemistry reveals protein expression only in some of the retina neurons. Although KCNK6 is able to dimerize as other functional two-P domain K+ channels when it is expressed in COS-7 cells, it remains in the endoplasmic reticulum and is unable to generate ionic channel activity. Deletions, mutations, and chimera constructions suggest that KCNK6 is not an intracellular channel but rather a subunit that needs to associate with a partner, which remains to be discovered, in order to reach the plasma membrane. A closely related human KCNK7-A subunit has been cloned. KCNK7 displays an intriguing GLE sequence in its filter region instead of the G(Y/F/L)G sequence, which is considered to be the K+ channel signature. This subunit is alternatively spliced and gives rise to the shorter forms KCNK7-B and -C. None of the KCNK7 structures can generate channel activity by itself. The KCNK7 gene is situated on chromosome 11, in the q13 region, where several candidate diseases have been identified.

Ion channels are present in excitable and nonexcitable eukaryotic cells, where they control the electrical potential across the cell membrane, secretion, and signal transduction (1). Electrophysiological studies have allowed identification and characterization of a great variety of ion channels, which are differentiated first by their selectivity and then by their other biophysical properties, their pharmacology, and their regulation (2). Recent cloning efforts and analysis of structure-function relationships have now provided a molecular basis for many of the biophysical properties. Thus, the mechanism that determines ion channel selectivity is now fairly well understood for K ϩ channels (3). K ϩ selectivity is carried by a structural element called P domain (P for pore-forming). This domain is highly conserved among the three main structural classes of K ϩ channel subunits with six-transmembrane, fourtransmembrane, or two-transmembrane domains (4).
The most recently described family of K ϩ channels has four transmembrane domains (5)(6)(7)(8)(9)(10). Members of this family have two pore-forming domains (P1 and P2), an extended extracellular loop between the M1 and P1 domains with a conserved cysteine residue (except for TASK-1), and a short cytoplasmic N-terminal extremity. TWIK-1 has been shown to form covalent homodimers (8,11). Despite a relatively low sequence similarity and different functional properties, these two-P domain K ϩ channels all produce quasi-instantaneous and noninactivating currents, although the TASK-2 currents displays relatively slow activation kinetics. These new K ϩ channels are presently classified into three distinct functional subfamilies. TASK-1 and TASK-2 are background K ϩ channels sensitive to small external pH variations near physiological pH (9,10,12). TREK-1 (13) and TRAAK (6,14) are arachidonic acid-activated mechanosensitive K ϩ channels; however, TREK-1 is inhibited by both protein kinase A and protein kinase C (5), while TRAAK is not. TWIK-1 is a weakly inward rectifying K ϩ channel that is stimulated by PKC and inhibited by internal acidification (7,8). The different types of regulations indicate that these K ϩ channels are probably involved in a great diversity of physiological and pathophysiological roles.
More than 40 genes potentially encoding two-P domain K ϩ channel subunits have now been identified in the Caenorhabditis elegans genome (15). This observation suggests that a large number of two-P domain channels might also exist in mammals. The present paper describes the cloning of two novel mammalian two-P domain channels related to TWIK-1, one of mouse and the other of human origin.
Analysis of KCNK6 and KCNK7 mRNA Distribution-For Northern blot analysis, mouse and human multiple-tissue Northern blots (CLON-TECH) were probed at 65 or 68°C in ExpressHyb solution with the 32 P-labeled coding sequences of KCNK6 or KCNK7 following the manufacturer's protocol. For reverse transcription-PCR, total RNAs were extracted from mouse tissues with the SNAP total RNA isolation kit (Invitrogen) and treated with DNase I. Fifteen g of total RNA were reverse-transcribed following the manufacturer's instructions (Life Technologies), and 1 ⁄40 of each sample was used as template for PCR amplification (Taq DNA polymerase, Life Technologies) using KCNK6 (nucleotides 12-31, 5Ј-GAAACCATGGGCCCGATACC-3Ј, and 366 -385, 5Ј-CAAGGGCTGCATAGACCACA-3Ј) or glyceraldehyde-3phosphate dehydrogenase (CLONTECH) primers. KCNK6-amplified fragments were transferred onto nylon membranes and then probed with a 32 P-labeled internal primer (bases 302-323, 5Ј-GCATCCTCAC-CACCACCGGTTA-3Ј).
In Situ Hybridization-Adult Swiss mice were killed by a transcardial perfusion with 0.9% NaCl followed by ice-cold 4% paraformaldehyde/phosphate-buffered saline (PBS, pH 7.4). The brain and retina were postfixed in an ice-cold 4% (w/v) paraformaldehyde/PBS solution for 3 h and then immersed overnight at 4°C in a 20% sucrose/PBS solution. Frozen sections (10 m) were collected on 3-aminopropylethoxysilane-coated slides and stored at Ϫ20°C. Specific antisense cRNA probes were generated with T7-RNA polymerase (Boehringer), by in vitro transcription using [ 33 P]D-UTP (3000 Ci/mmol, ICN Radiochemicals), from a BamHI-linearized pBluescript SK-plasmid containing an 800-base pair fragment from the 3Ј-untranslated sequence of KCNK6. Sections were treated using standard procedures (17). For control experiments, adjacent sections were hybridized with a corresponding sense riboprobe or digested with RNase before hybridization.
Production and Characterization of Anti-KCNK6 Antibodies-DNA fragments coding for regions extending from Gln 29 to Asn 89 and from Glu 254 to Pro 335 were amplified by PCR and subcloned behind the glutathione S-transferase (GST) coding sequence into the pGEX3x plasmid. The resulting GST-KCNK6 fusion proteins were produced in the BL21 (DE3 pLysS) Escherichia coli strain and purified on glutathione-Sepharose according to the manufacturer's protocol (Amersham Pharmacia Biotech). Antibodies were then raised in New Zealand female rabbits immunized with a mixture containing 300 g of purified fusion protein in the presence of complete Freund's adjuvant and then boosted every 4 weeks first with 150 g of each protein in incomplete Freund's adjuvant. For purification, the antisera was first depleted of the anti-GST antibodies by absorption to nitrocellulose strips saturated with the GST, followed by affinity purification of the anti-KCNK6 polyclonal antibodies against nitrocellulose strips saturated with the immunogen as described previously (18).
Western Blot-For the preparation of microsomes, transfected COS-7 cells (see below for the transfection) were washed, sonicated in lysis buffer (140 mM NaCl, 20 mM Tris, pH 7.4, 1 mM phenylmethylsulfonyl fluoride, 10 mM iodoacetamide, 2 mM EDTA), and the lysate was cleared by centrifugation at 1000 ϫ g for 10 min. The supernatant was then centrifuged at 100,000 ϫ g for 20 min to obtain microsomes. 0.5-1 g were then loaded on a SDS-polyacrylamide gel electrophoresis (9% gel) and transferred onto nitrocellulose membrane (Hybond-C extra, Amersham Pharmacia Biotech). Blots were saturated with 4% low fat dry milk in PBS and incubated 1 h with affinity-purified anti-KCNK6 antibodies or M2 anti-tag monoclonal antibodies (Eastman Kodak Co.) diluted 500-fold, followed by an additional 1-h incubation with horseradish peroxidase-conjugated goat anti-rabbit IgG or goat anti-mouse IgG (Jackson) with extensive washing steps after antibody incubation, and finally revealed with Super Signal (Pierce). Indirect Immunofluorescence on Transfected COS-7 Cells-Transfected COS-7 cells were plated on glass coverslips onto 15-mm Petri dishes. 24 h after plating, the cells were fixed for 15 min with a 4% (v/v) paraformaldehyde/PBS solution, rinsed with PBS, and then permeabilized by incubation for 10 min with 0.1% Triton X-100 (without Triton for nonpermeabilized conditions). Nonspecific binding was eliminated by a 2-h incubation with 5% goat serum, 2% bovine serum albumin in PBS at room temperature. The cells were then incubated for 1 h with affinity-purified anti-KCNK6 antibody (1:400) or anti-TRAAK antibody (1:400) 2 in 2% bovine serum albumin/PBS, followed by washing with PBS and incubation for 1 h with tetramethylrhodamine isothiocyanateconjugated goat anti-rabbit Ig (1:200; Sigma) or fluorescein isothiocyanate-conjugated goat anti-rabbit Ig (1:200, Sigma). After washing in PBS and finally 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) or green (rhodamine) filter and a ϫ 40 lens.
Immunohistochemistry on Mouse Retina-The eyes were enucleated and immediately fixed with ice-cold 4% paraformaldehyde/PBS for 4 h. The lens and the vitreous were removed, and the eye-cup was cryoprotected in PBS containing 20% sucrose. Frozen sections (7 m) were cut on a Leica cryostat and stored at Ϫ70°C. For each retina, three sections were placed on 3-aminopropylethoxysilane-coated slides. Frozen sections were permeabilized with 0.3% Tween 20 in PBS for 3.5 min at room temperature followed by two rinses in PBS. Sections were then incubated with a blocking solution (1% goat serum/PBS) for 5 h, and then the primary antibody (1:400) was added overnight. After three rinses in PBS, the second goat anti-rabbit IgG biotinylated antibody (1:330) was incubated for 1 h and amplified with the Vectastain Elite ABC kit from Vector Laboratories. The antigen-antibody complexes were visualized by 3,3Ј-diaminobenzidine staining using the Vector Laboratories Kit. Control for the specificity of KCNK6 immunostaining was done by preabsorption of the primary antibodies (1:400) with the GST-KCNK6 fusion protein. All sections were washed three times in PBS and one time in distilled water and then mounted with Entellan (Merck).
Construction of N-terminally Tagged KCNK6 and KCNK6 Mutants-The coding sequence of KCNK6 was subcloned into the HpaI-NotI restriction sites of Flag-pRc/CMV vector to create an N-terminally tagged KCNK6 protein as described previously for the N-terminally tagged Kv2.2 protein (19). Deletion channels and chimeras between KCNK6, mTREK1, and hTASK1 were constructed by PCR as described (20). PCRs were performed on Flag-pRc/CMVKCNK6 to keep the Flag peptide sequence at the amino extremity of the constructions. Products were subcloned into EcoRI-BamHI site of pIRES-CD8 vector and verified by sequencing.
Cell Culture and Transfection-COS-7 cells were grown in Dulbecco's modified Eagle's medium (Life Technologies, Inc.) supplemented with 10% fetal calf serum and antibiotics (60 mg/ml penicillin, 50 mg/ml streptomycin). One day before transfection, 20 ϫ 10 3 cells for electrophysiology experiment or 70 ϫ 10 3 cells for immunofluorescence microscopy were plated on cover glasses onto 35-mm Petri dishes. The cells were transfected by the classical DEAE-dextran/chloroquine method using 1 or 2 g of supercoiled DNA.
Human Chromosomal Mapping of KCNK7-Gene mapping was performed by PCR on the Genebridge 4 radiation mapping panel (Research Genetics, Inc.) with the following primers: sense, 5Ј-CATGGCCCCAC-TATCGC-3Ј; antisense, 5Ј-CTGCAGTCGCCCTGAAGG-3Ј. The results were analyzed by using the RHMAPPER program at the Whitehead Institute with a logarithm of odds score of 21.
Amino Acid Sequence Alignments and Dendrogram-The alignment was generated with the ClustalW multiple sequence analysis program. The dendrogram and the percentage of identity were deduced from the conserved region extending from the M1 to the M4 transmembrane domains. Accession numbers of the channel aligned are as follows: mTWIK, AF033017; mTRAAK, AF056492; hTASK-1, AF006823; hTASK-2, AF084830; mTREK1, U73488.

Cloning of KCNK6, a New Ion Channel Subunit-The
TWIK-1 sequence (7) was used to search related sequences in the Genbank TM data base by using the Blast alignment program. A mouse EST (accession number W18545) was identified encoding a portion of a new ␣-subunit (named KCNK6). The corresponding full-length cDNA was obtained by RACE and sequenced. The coding sequence was then reamplified from both mouse brain and lung cDNAs to verify the sequence and to exclude all possible PCR errors. The KCNK6 cDNA contains an open reading frame of 1008 base pairs and codes for a protein of 335 amino acids (Fig. 1A) with a calculated molecular mass of 35 kDa. KCNK6 is more related to TWIK-1 than to the other two-P domain K ϩ channels (40% of amino acid identity (Fig.  1C)). Like all of these channels, KCNK6 displays four potential transmembrane segments (M1-M4) and two P domains (P1 and P2, P for pore-forming domain) (Fig. 1, A and B). Another important characteristic is the presence of an extended extracellular loop between M1 and P1 that contains a cysteine residue at a position (Cys 57 ) analogous to the cysteine residue (Cys 69 ) involved in the disulfide-bridged homodimerization of TWIK-1 (11). The protein sequence of KCNK6 contains consensus sites for N-linked glycosylation (residue 82) and phosphorylation by protein kinase C (residue 3) and casein kinase II (residues 255, 274, 276, 308, and 327). Three potential Src homology 3 binding motifs are present in the C-terminal part of the protein (minimal consensus sequence PXXP (21,22)). Computer-based analysis also indicates the very interesting presence of a potential Ca 2ϩ -binding site (EF-hand motif) at its C terminus (Fig. 1D). This region includes the complete EF-hand loop as well as the first residue that follows the loop, which is always hydrophobic. Moreover, as in a classical EF-hand domain (23)(24)(25), the 12 residues loop of this region of KCNK6 are flanked on both sides by ␣-helical structures (nnPredict program; data not shown).
Tissue Distribution of KCNK6 -In order to investigate KCNK6 mRNA expression, a Northern blot analysis was carried out using a mouse multiple-tissue Northern blot from CLONTECH. No KCNK6 transcripts were detected, indicating that KCNK6 is not expressed or is only expressed at very low levels in the tested tissues (heart, brain, spleen, lung, liver, skeletal muscle, kidney, and testis). The expression of KCNK6 was further studied on a larger panel of tissues with a more sensitive technique, reverse transcription-PCR ( Fig. 2A). KCNK6 mRNA was detected in embryo (16 -18 days), in eye, stomach, and lung. A weak expression was found in colon and testis. In the other tissues, the expression of KCNK6 mRNA was very low (atria, kidney, intestine, bladder, uterus, ovary, salivary gland, thymus, and brain stem) or not detectable (brain, cerebellum, spinal cord, heart, ventricle, skeletal muscle, liver, placenta, pancreas). In situ hybridization was performed on a vertical section of mouse retina. An antisense KCNK6 cRNA probe revealed an intense expression with a distinct stratification pattern. High labeling was apparent in the retinal ganglion cell layer (RGC) and in the inner nuclear layer (INL) composed of nuclei of Mü ller cells and amacrine, bipolar, and horizontal neurons (Fig. 2B). The distribution of the hybridization signal suggests a neuronal localization of transcripts. The control sense probe did not show significant hybridization (data not shown). To confirm these results, im-munohistochemistry was performed on an equivalent section. The KCNK6 protein was immunodetected in the same neurons. A strong immunoreactivity was observed in the retinal ganglion cell layer. In addition, regular punctual staining was detected in a specific neuron population of the distal area of the inner nuclear layer (Fig. 2C). To demonstrate the specificity of anti-KCNK6 antibodies, an equivalent section was incubated with anti-KCNK6 antiserum, which was previously absorbed with the GST-KCNK6 fusion proteins. As expected, the staining was prevented (data not shown). In situ hybridization and immunohistochemistry experiments did not show significant detection in the brain (data not shown).
Heterologous Expression of KCNK6 Subunits-Electrophysiological attempts to record KCNK6 channel activity in various cell types like Xenopus oocytes, COS-7, HEK293, and Sf9 cells were unsuccessful. In all of these expression systems, attempts to activate channel activity with 1 M Ca 2ϩ at the intracellular side (because of the presence of the EF-hand domain), with external pH variations from pH 7.0 to 4.0, or with application of a phorbol ester (phorbol 12-myristate 13-acetate) (there is one potential protein kinase C site (Fig. 1A)) were without any success. We also tried a coexpression with the ␤ 1 ␥ 2 subunits of G proteins, which lead to activation of G-protein-gated inward rectifier (26), as well as coexpression with the closely related two-P domain subunit TWIK-1, but again without success. Constructions with pCI and pRc/CMV expression vectors were also tested with the same negative results.
Western blot analysis was then carried out to verify KCNK6 expression in transfected cells. Under nonreducing conditions, anti-tag antibodies (M2) detected a major band at a molecular mass of 83 kDa from COS-7 cells transfected with a tagged form of KCNK6. Under reducing conditions (i.e. in the presence of ␤-mercaptoethanol), two bands were detected with molecular mass of 44 and 40 kDa. A similar blot profile was revealed using affinity-purified anti-KCNK6 antibodies (Fig. 3B). In order to demonstrate the specificity of the anti-KCNK6 antiserum and to justify its use to analyze the expression of the KCNK6 protein in transfected cells, we checked that no signal was obtained from control COS-7 cells expressing the TWIK-1 protein. This expression of the human TWIK-1 protein was verified in the same experiment with the affinity-purified rabbit antibodies directed against the C-terminal part of TWIK-1 (11) (data not shown). The masses of 37 and 40 kDa are in good agreement with the theoretical values of 35 and 37 kDa for KCNK6 and its tagged form, respectively. As for the TWIK-1 subunit (11), the observation that the band moving at 77 kDa gives rise to a band moving at 37 kDa in the presence of a reducing agent strongly suggests that KCNK6 can self-associate via a disulfide bond to form a homodimer. The masses of 41 and 44 kDa for KCNK6 and its tagged form suggest, as previously demonstrated for TWIK-1 (11), the existence of glycosylated forms. All of these results were reproduced in Sf9 cells (data not shown).
Intracellular Retention of KCNK6 -One possible reason for the lack of channel activity could of course be an intracellular localization of the expressed KCNK6 protein. To test this point, the KCNK6 subunit was localized by indirect immunofluorescence with the anti-KCNK6 antibodies (Fig. 3C) on Tritonpermeabilized COS-7 cells. A strong fluorescence staining was detected at the perinuclear region as well as in a fine reticular network extending through the cytoplasm. This pattern was observed in all of the KCNK6-expressing cells from multiple independent transfections. This pattern is similar to that seen with known ER markers (27), suggesting that KCNK6 is specifically sequestered in the ER and cannot reach the plasma membrane. This pattern of labeling is very different from that obtained with the TRAAK channel. As shown in Fig. 3D, TRAAK-expressing cells, which produce strong K ϩ currents (6), show a typical surface labeling.
Why Is KCNK6 Unable to Reach the Plasma Membrane?-The first reason why KCNK6 cannot reach the surface membrane could reside in the peptide sequence itself. Protein retention in intracellular compartments is usually determined either by the presence of short transmembrane domains (27)(28)(29)(30) or by the presence of specific cytoplasmic sequences. These are the ER retrieval signals (KDEL, KK, or RR (31)), the peptide signals targeting to post-Golgi compartments (LL or YXXO, where O is an amino acid with a bulky hydrophobic group (32)), the peptide signals targeting to the trans-Golgi network (SXYQRL (33), AYRV (34), or SDSEEDE (35,36)), the signals involved in endosomal trafficking (YKGL (35,36)), or the endocytosis signal (DAKTI or DAKSS (37)). The transmembrane domains of the KCNK6 protein present the requisite length to reach the plasma membrane as for the other functional two-P domain K ϩ channels. Then a possible retention signal was searched on the cytoplasmic regions of the KCNK6 subunit. A particular sequence was found in the cytoplasmic tail between Ser 274 and Asp 282 . This sequence S CKII RT CKII D-EDQD is similar to a signal required for trans-Golgi network localization, S CKII DS CKII EEDE (CKII indicates amino acids phosphorylated by casein kinase II) (35). This specific localization requires a cluster of acidic amino acids and a pair of amino acids phosphorylated by casein kinase II (38). Interestingly, this KCNK6 region overlaps with the EF-hand domain and in particular the residues that are expected to be involved in the coordination of Ca 2ϩ . One possibility would then be that the function of the putative retention signal could be eliminated by Ca 2ϩ fixation. To test this point, an ionomycin/Ca 2ϩ (1 M) treatment of 5, 10, or 15 min at 37°C was performed on transfected COS-7 cells, as described previously (39). Unfortunately, indirect immunofluorescence microscopy with anti-KCNK6 antibodies did not show any plasma membrane staining on cells having undergone this treatment (not shown). In addition, suppression of this potential retention signal in the KCNK6⌬C274/296 deletion mutant (Fig. 4) also failed to create access of KCNK6 to the plasma membrane. The immunolocal-ization of this mutant displays a pattern identical to that of the wild type subunit, and no current could be recorded in electrophysiological experiments. Other deletion mutants (KCNK6⌬C296, KCNK6⌬C282, KCNK6⌬C273) corresponding to the C terminus domain, where retention signals are generally localized, were constructed and expressed (Fig. 4). Again, the intracellular localization of these mutants was the same as for native KCNK6, and these mutants were electrophysiologically inactive. In order to suppress all possible retention signals on the KCNK6 cytoplasmic parts, chimeras have been prepared between KCNK6 and intracellular parts of the two-P domain TASK1 or TREK1 channels, which by themselves reach the plasma membrane (5,9). The KCNK6/TASK-C and KCNK6/ TREK-loop chimeras presented in Fig. 4 were produced and expressed. Again, the same characteristic ER lacy pattern was observed by immunocytolocalization of these chimeras, and no current was recorded. Taken together, these results show that the KCNK6 subunit remains inside of the cell but that it is probably not specifically expressed in the ER, since this subunit does not contain specific retention signals. The results suggest that KCNK6 necessitates an association with a partner to reach the plasma membrane. This possible partner seems to be absent in the expression systems used.
Cloning of KCNK7, a Human Subunit Closely Related to KCNK6 -Since KCNK6 was cloned from mice and did not produce a functional channel, we then decided to clone a homologue of KCNK6 in humans in order to verify whether changing animal species would lead to functional expression. A Blast search was performed on human EST data bases with the KCNK6 peptide sequence. Two sequences (EST accession numbers AA777982 and N39619) were identified corresponding to the P1-M2-M3-P2 region of the two-P domain channel subunits and were very close to KCNK6. The P2 domain showed a very unconventional sequence GLE instead of GLG found in TWIK-1 and KCNK6 pore domains. As for KCNK6, 5Ј-and 3Ј-RACE reactions allowed the identification of a start codon preceded by several stop codons. Interestingly, two 3Ј-RACE products were found containing the poly(A) tail. Sequencing revealed that both fragments are produced by splicing of a unique mRNA. Further analyses, detailed below, revealed that exon skipping gives rise to five spliced mRNAs. These five mRNAs code for three protein forms, a subunit called KCNK7-A and two shorter variants called KCNK7-B and KCNK7-C with truncated cytoplasmic carboxyl termini. PCR was used to clone the complete coding sequences of KCNK7-A, KCNK7-B, and KCNK7-C from human brain cDNA. The independent cloning of these three forms confirmed sequences, and especially the peculiar sequence GLE found in the P2 domain. The KCNK7-A, KCNK7-B, and KCNK7-C cDNA have open reading frames of 921, 756, and 771 base pairs and code for proteins of 307, 252, and 257 amino acids (Fig. 5) with a calculated molecular mass of 32, 26, and 27 kDa, respectively. KCNK7-A has the hallmarks of a functional subunit of two-P domain potassium channels. KCNK7-A shows 80 and 42% amino acid identity, and 94 and 64% homology with the core region of KCNK6 and TWIK-1, respectively (Fig. 1C). As expected, KCNK7-A is very close to KCNK6 (Fig. 5). Several sites are conserved between both subunits such as the consensus sites for N-linked glycosylation (residue 82) and the two sites for casein kinase II (residues 255 and 276). The short C-terminal tail of KCNK7-A also presents two potential Src homology 3 binding motifs. However, the EF-hand domain found in KCNK6 is not conserved. A site for Ca 2ϩ calmodulin kinase (residue 8) at the cytoplasmic N-terminal tail could confer to KCNK7 an indirect regulation by Ca 2ϩ . The spliced forms, KCNK7-B and KCNK7-C, have shorter cytoplasmic carboxyl termini (Fig. 5). The M4 domain of KCNK7-B and KCNK7-C presents an abnormally short ␣-helical structure (about 16 residues as deduced from the nnPredict program and hydropathy analyses (data not shown)) in comparison with the equivalent domains of the other K ϩ channels.
As previously indicated, the analysis of the variable 3Ј region by reverse transcription-PCR led us to the identification of five KCNK7 mRNA spliced forms (Fig. 6B). These different variants contain three alternative stop codons called stop-A, -B, and -C, as depicted in Fig. 6A. Sequencing of the different splice variants allowed us to identify the donor and acceptor splice sites (Fig. 6D). The longest protein form (KCNK7-A) is produced as the result of exclusion of an exon (exon skipping) containing an in-frame stop codon. The quantification of the different forms was performed by Southern blot and revealed that mRNAs coding for KCNK7-A, KCNK7-B, and KCNK7-C subunits are in proportions of 46, 39, and 8%, respectively (Fig.  6C). We cannot distinguish between KCNK7-B and KCNK7- C   FIG. 3. A and B, microsomal proteins of TWIK-1-, N-terminally tagged KCNK6-, or KCNK6-expressed COS-7 cells were analyzed in nonreducing (␤ME: Ϫ) or reducing (␤-ME: ϩ) conditions. Immunodetection was performed either with the commercial mouse anti-Tag antibody M2 (A) or with the rabbit anti-KCNK6 antiserum (B). C-E, immunofluorescence localization of the KCNK6 subunit on permeabilized COS-7 cells. Immunocytolocalization was realized on cells expressing native KCNK6 subunit (C) with anti-KCNK6 polyclonal antibodies. For comparison, D shows the plasma membrane expression of the two-P domain K ϩ channel TRAAK. Negative controls for anti-KCNK6 polyclonal antibodies (E) were performed on cells expressing TWIK-1 subunit. Anti-KCNK6 antibodies were visualized using a goat anti-rabbit IgG-tetramethylrhodamine isothiocyanate-conjugated secondary antibody (red detection), and anti-TRAAK antibodies were detected using a goat anti-rabbit IgGfluorescein isothiocyanate-conjugated secondary antibody (green detection).
for the last 8% (Fig. 6, B and C). KCNK7 is expressed in the brain.
As for KCNK6, neither human KCNK7 nor its spliced forms are able to generate channel activity in transfected COS-7 cells. Attempts to record currents in cells transfected with all possible combinations of two of the KCNK6 and KCNK7-A, -B, and -C subunits also failed to produce any ion current.
Gene assignment by radiation hybrid panel mapping localized the human KCNK7 channel gene to chromosome 11q13 at 6.4 cRays from the framework marker WI-1409 (logarithm of odds score Ͼ21) (Fig. 7). DISCUSSION This paper reports the cloning of KCNK6 (mouse) and KCNK7 (human) subunits, which are structurally related to the growing family of K ϩ channels with four TMDs and two P domains (5-10). They are more particularly related to the TWIK-1 channel (40 and 42% amino acid identity, respectively). The previously cloned channels of this family all produce instantaneous and noninactivating currents, except TASK-2, which displays a relatively slow kinetic of activation. They are open at resting potential and are able to drive the membrane potential near the K ϩ equilibrium potential. There are three distinct functional subfamilies of two-P domain channels. TASK-1 and TASK-2 are background K ϩ channels sensitive to external pH variations near physiological pH (9,10). Both TREK-1 (13) and TRAAK (6,14) produce arachidonic acidactivated and mechanosensitive K ϩ currents, and TREK-1 but not TRAAK is inhibited by both protein kinase A and C (5). TWIK-1 is a weakly inward rectifying K ϩ channel that is stimulated by protein kinase C and inhibited by internal acidification (8). These K ϩ channels probably contribute in a major way to the regulation of the resting membrane potential and are probably endowed with a diversity of physiological roles.
Voltage-sensitive Ca 2ϩ -dependent K ϩ channels have been identified in many cell types and have now been cloned (40). Background Ca 2ϩ -activated K ϩ channels have also been re- corded, but they have not yet been cloned in mammals. They are particularly well expressed in sensory neurons of the nodose ganglion, where they contribute to the lack of spontaneous activity observed in these cells (41). Background Ca 2ϩ -activated K ϩ channels have been cloned in plant cells and correspond to a two-P domain K ϩ channel, which is activated by cytosolic Ca 2ϩ via two EF-hand domains (42). A reasonable assumption would then be that the mouse channel subunit KCNK6, which comprises an EF-hand domain, would also be a background channel regulated by internal Ca 2ϩ . Sequences of the human subunits KCNK7-A, -B, and -C do not contain an EF-hand domain, but an indirect Ca 2ϩ regulation would also be possible in that case through the Ca 2ϩ -calmodulin kinase site.
Electrophysiological experiments have revealed that neither KCNK6 nor KCNK7 can generate channel activity by themselves in Xenopus oocytes or transfected COS-7 cells. Reasons for this lack of expression have been particularly analyzed for KCNK6. Biochemical experiments showed KCNK6 protein expression and, as for TWIK-1 (11), subunit dimerization in transfected cells, indicating that the absence of channel activity was not due to a lack of expression or dimerization, or to a degradation of the protein. Immunocytolocalization on transfected COS-7 cells has shown that the absence of current is due to the fact that KCNK6 is unable to reach the plasma membrane. A first interpretation of this result would be that KCNK6 is an intracellular channel. If this were the case, then the putative K ϩ channel subunit would be expected to carry one of the signals for intracellular retention on its cytoplasmic regions. One such possible signal is indeed observed in a KCNK6 region that overlaps with the EF-hand region. However, suppression of this hypothetical retention signal by deletion did not confer access to the plasma membrane. Chimeras formed with KCNK6 and cytoplasmic elements of TREK1 or TASK1 channels (which have an easy access to the cell surface) also failed to reach the plasma membrane and to form electrophysiologically recordable channels. Another possibility, which is at present the most probable one, would be that KCNK6 needs a partner to reach the plasma membrane. However, this partner remains to be identified, and is not present in the heterologous expression systems used. An association between partner channel subunits essential for the expression of channel activity has been observed with G-protein-gated inward rectifiers (43,44), with some of the voltage-sensitive K ϩ channels (17,19,(45)(46)(47), or with epithelial Na ϩ channels (48 -50). The required association can also take place between channel subunits and other types of subunits, themselves acting as chaperones (51,52) or carrying other types of functions such as for the ATP-sensitive K ϩ channels (53, 54) or for the channel generating the I-Ks cardiac K ϩ current (55)(56)(57).
KCNK6 is mainly expressed in tissues such as eyes, lung, and stomach. It is also significantly expressed in the embryo. The highest level of expression is found in the eyes, where in situ hybridization and immunohistochemistry showed that KCNK6 is only expressed in ganglion cells and in some neurons of the inner nuclear layer. This very restricted localization probably explains the absence of detection by Northern blot. In the mammalian retina, the first spontaneous Ca 2ϩ waves are observed at postnatal day 2 and are thought to result from Ca 2ϩ influx associated with burst of action potentials seen in ganglion cells at this developmental time (58 -60). The early appearance of the KCNK6 channel in development, the fact that it has a Ca 2ϩ -binding domain probably conferring Ca 2ϩ sensor properties, and its selective expression in ganglion cells suggest that this channel could play a role in the modulation of the electrical signal in the retina.
The human channel subunit KCNK7 is very close in structure from KCNK6 (80% of identity on the core channel), but it does not contain an EF-hand domain. KCNK7 RNA is alternatively spliced by an exon skipping mechanism, which generates five different mRNAs in the human brain. These five mRNAs present three alternative stop codons (named A, B, and C) and code for three subunits named KCNK7-A, -B, and -C. KCNK7-B and -C subunits are truncated in the C-terminal sequence. The mRNA coding for these two short forms represent about 50% of the total KCNK7 mRNA. Interestingly, KCNK7-B and -C have a very short M4 domain with a potential anchor at the C-terminal extremity (myristoylation site and amidation site, in KCNK7-B and KCNK7-C, respectively). Short TMDs are considered as important factors for the ER retention (27,28,61). Plasma membrane proteins generally have longer TMDs than Golgi membrane proteins (62). Therefore, the short M4 domain of KCNK7-B and -C could confer an intracellular localization.
The structures of the second P region (P2) of the different mammalian two-P domain channels are compared in Fig. 8. An important element of the signature of K ϩ channel function has long been recognized as being the P domain GYG sequence (63). This sequence is found in most voltage-sensitive, Ca 2ϩ -sensitive, and inward rectifier K ϩ channels (3,64,65). In the two-P domain channels, the canonical GYG structure is replaced by a GFG (TREK1, TASK1, TASK2, and TRAAK) or by a GLG (TWIK1 and KCNK6). A tryptophan residue (Trp 213 in TWIK-1) is conserved in TWIK1, TREK1, TRAAK1, TASK1, and TASK2, but it is replaced by a cysteine residue in KCNK6 and KCNK7. In addition, two adjacent residues, a tyrosine and a valine (Tyr 234 -Val 235 in TWIK-1), which are also present in all other members of the family, are replaced by a leucineleucine sequence (Leu 221 -Leu 222 in KCNK6) in KCNK6 and KCNK7. KCNK7 is unique, since a glutamic acid residue (Glu 219 in KCNK7) is found instead of the strictly conserved glycine residue that is present in the filter part of all of the other members of the K ϩ channel family. Moreover, a negatively charged residue (Asp or Glu residues in the different members of this family) is always found at a 13-amino acid distance of the G(F/L)G sequence filter but is replaced by a glycine (Gly 204 ) in KCNK7. Interestingly, the comparison with the recently published tertiary structure of the selectivity filter of the Kcsa K ϩ channel (3) reveals that the side chains of the amino acids found in the ion filter region, Glu 219 and Leu 221 -Leu 222 in KCNK7, are close in space to the side chains of residues observed in the pore helix region, Gly 204 and Cys 200 , respectively (see model in Fig. 8), suggesting an adaptation of the pore structure.
There is a growing number of pathological situations associ-ated with mutations in diverse ion channel genes (66 -75), and it would then be not very surprising if diseases were also associated with the two-P domain class of channel subunits. The human KCNK7 gene has been mapped to the 11q13 region of the human genome near the markers WI-1409. Four different diseases are associated with this genome region: the Best's vitelliform dystrophy, which is an autosomal dominant juvenile onset macular degeneration and associated with an electrophysiological abnormality and a degeneration of the retinal pigment epithelium (76); the Bardet-Biedl syndrome, which combines mental retardation, obesity, hypogenitalism, and progressive retinal pigmentary dystrophy (77); the insulindependent diabetes mellitus named IDDM4 (78); and the cerebellar ataxia named CLA1 (79).
FIG. 8. Top, alignment of the second pore region of mammalian two-P domain channels. The arrows indicate the singular amino acids found in the KCNK6 and/or KCNK7 pore domain. Junctions between arrows indicate the regions that are assumed to be close to each other in the tertiary structure (3). Bottom, pore model for the two-P domain channels, KCNK6 and KCNK7, based on the tertiary structure of the selectivity filter of the Kcsa K ϩ channel (3).