THIK-1 and THIK-2, a Novel Subfamily of Tandem Pore Domain K+ Channels* 210

Two cDNAs encoding novel K+channels, THIK-1 and THIK-2 (tandem pore domainhalothane inhibited K +channel), were isolated from rat brain. The proteins of 405 and 430 amino acids were 58% identical to each other. Homology analysis showed that the novel channels form a separate subfamily among tandem pore domain K+ channels. The genes of the human orthologs were identified as human genomic data base entries. They possess one intron each and were assigned to chromosomal region 14q24.1–14q24.3 (human (h) THIK-1) and 2p22–2p21 (hTHIK-2). In rat (r), THIK-1 (rTHIK-1) is expressed ubiquitously; rTHIK-2 expression was found in several tissues including brain and kidney. In situ hybridization of brain slices showed that rTHIK-2 is strongly expressed in most brain regions, whereas rTHIK-1 expression is more restricted. Heterologous expression of rTHIK-1 in Xenopus oocytes revealed a K+channel displaying weak inward rectification in symmetrical K+ solution. The current was enhanced by arachidonic acid and inhibited by halothane. rTHIK-2 did not functionally express. Confocal microscopy of oocytes injected with green fluorescent protein-tagged rTHIK-1 or rTHIK-2 showed that both channel subunits are targeted to the outer membrane. However, coinjection of rTHIK-2 did not affect the currents induced by rTHIK-1, indicating that the two channel subunits do not form heteromers.

Two cDNAs encoding novel K ؉ channels, THIK-1 and THIK-2 (tandem pore domain halothane inhibited K ؉ channel), were isolated from rat brain. The proteins of 405 and 430 amino acids were 58% identical to each other. Homology analysis showed that the novel channels form a separate subfamily among tandem pore domain K ؉ channels. The genes of the human orthologs were identified as human genomic data base entries. They possess one intron each and were assigned to chromosomal region 14q24.1-14q24.3 (human (h) THIK-1) and 2p22-2p21 (hTHIK-2). In rat (r), THIK-1 (rTHIK-1) is expressed ubiquitously; rTHIK-2 expression was found in several tissues including brain and kidney. In situ hybridization of brain slices showed that rTHIK-2 is strongly expressed in most brain regions, whereas rTHIK-1 expression is more restricted. Heterologous expression of rTHIK-1 in Xenopus oocytes revealed a K ؉ channel displaying weak inward rectification in symmetrical K ؉ solution. The current was enhanced by arachidonic acid and inhibited by halothane. rTHIK-2 did not functionally express. Confocal microscopy of oocytes injected with green fluorescent protein-tagged rTHIK-1 or rTHIK-2 showed that both channel subunits are targeted to the outer membrane. However, coinjection of rTHIK-2 did not affect the currents induced by rTHIK-1, indicating that the two channel subunits do not form heteromers.
Relatively little is known about the function of 2P K ϩ channels in vivo. Recently, endogenous currents with properties similar to those of TASK channels have been found in the heart (17), in arterial chemoreceptor cells (18), in zona glomerulosa cells of the adrenal cortex (19), in cerebellar granular cells (20), and in motoneurons (21). TASK-1 currents have been shown to be coupled to the activation of thyrotropin-releasing hormone receptor 1 (TRH-R1; Ref. 21) and angiotensin II receptors (AT1a; Ref. 19); thus, TASK-1 channels are regulated also by mechanisms other than extracellular pH. Furthermore, another member of this subfamily, TASK-3, is activated by depolarization and modulated by extracellular divalent cations (5).
With TREK-1, TREK-2, and TRAAK, three mechanosensitive channels have been cloned (10,22). Other factors that have been reported to modulate the activity of these mechanosensitive channels include heat (TREK-1; Ref. 23 25). The third subfamily of 2P K ϩ channels so far comprises TWIK-1 and TWIK-2, both of which are expressed in multiple tissues (12,13,15). TWIK-1 shows weak inward rectification with symmetrical K ϩ concentrations (12). 2 The current carried by TWIK-2 is heat-sensitive and shows rapid time-dependent inactivation at 37°C (15). Both channels may contribute to setting the resting membrane potential, but their specific function in various tissues is not yet clear.
In this paper we describe the cloning of the first two members of a novel subfamily of 2P K ϩ channels. One of these channels, rTHIK-1, was found to be expressed in all tissues tested. Heterologous expression of rTHIK-1 in Xenopus oocytes induced a current that could be activated by arachidonic acid and inhibited by the volatile anesthetic halothane. The second novel 2P K ϩ channel, THIK-2, is closely related to THIK-1 (58% identity at the amino acid level) but could not be functionally expressed. THIK-2 was strongly expressed in several tissues including stomach, liver, and kidney and was particularly abundant in the brain.

EXPERIMENTAL PROCEDURES
Data Base Analysis-BLAST searches of the expressed sequence tag data base (dbEST), the genomic survey sequence data base (dbGSS), and the human genomic data base (htgs) identified several human and rat EST clones, one human GSS clone, and three human genomic data base entries of novel 2P K ϩ channel subunits. One rat EST clone (GenBank TM accession number AI070460) and one human GSS clone (GenBank TM accession number AQ898820) were purchased from Research Genetics (Huntsville, AL) and completely sequenced.
Isolation of Rat THIK-1 and THIK-2 cDNAs-About 1 ϫ 10 6 clones of a -ZAPII rat brain or a -ZAPII rat heart cDNA library (Stratagene) plated with XL1-Blue MRFЈ cells were screened with Digoxigeninlabeled THIK-2 and THIK-1 fragments derived from the EST or GSS clones, respectively, and positive clones were detected using CSPD (Roche Molecular Biochemicals) as chemoluminescence substrate. After purification of isolated plaques by two further screenings, pBSKϩ plasmids containing the cDNAs were excised from -clones and sequenced.
Sequence Analysis-For sequence analysis, the GCG program package implemented at the Heidelberg Unix Sequence Analysis Resources (HUSAR) was used. The program PROSITE was used for identification of sequence consensus sites, the program GAP for pairwise sequence alignments, the program CLUSTAL W for creation of multiple alignments, and the program CLUSTREE for computation of the phylogenetic tree. Standard default parameters were used for the calculations.
Electrophysiological Analysis-cDNAs encoding rTHIK-2, rTHIK-1, and chimeras between the two were cloned into the expression vector pSGEM (a gift of Dr. M. Hollmann) for expression in Xenopus laevis oocytes. Capped run-off poly(A ϩ ) cRNA transcripts were synthesized and injected individually or in combination into defolliculated oocytes at constant amounts (ϳ3 ng each). Oocytes were incubated at 19°C in ND96 solution (96 mM NaCl, 2 mM KCl, 1 mM MgCl 2 , 1 mM CaCl 2 , 5 mM HEPES, pH 7.4 -7.5), supplemented with 100 g/ml gentamicin and 2.5 mM sodium pyruvate, and assayed 48 h after injection. Two-electrode voltage-clamp measurements were performed with a Turbo Tec-10 C amplifier (npi, Tamm, Germany), which has a settling time of 100 s for a 100-mV voltage pulse. Data were recorded via an EPC9 (Heka Electronics, Lambrecht, Germany) interface at sampling rates up to 20 kHz using Pulse/Pulsefit software (Heka). The Oocytes were placed in a small-volume perfusion chamber and bathed with ND96 or "high K ϩ " solution (96 mM KCl, 2 mM NaCl, 1 mM MgCl 2 , 1 mM CaCl 2 , 5 mM HEPES, pH 7.4 -7.5).
In Situ Hybridization-Wistar rats were decapitated under ether anesthesia, and their brains were removed and frozen on powdered dry ice. Tissue was stored at Ϫ20°C until cutting. Sixteen-m sections were cut on a cryostat, thaw-mounted onto silane-coated slides, and airdried. After fixation for 10 min in 4% paraformaldehyde dissolved in PBS, slides were washed in PBS, dehydrated, and stored in ethanol until hybridization.

RESULTS
Cloning of Two Novel 2P K ϩ Channel cDNAs-Two cDNA populations coding for novel 2P K ϩ channels were isolated from two rat cDNA libraries. The first channel was isolated from a rat brain and a rat heart library and was named THIK-1 for tandem pore domain halothane inhibited K ϩ channel. The second, closely related channel was found only in the rat brain library and was named THIK-2 despite the fact that it was not functional when expressed in Xenopus oocytes. Sequence analysis of the isolated cDNA clones revealed complete open reading frames ( Fig. 1) for rTHIK-1 (2032 bp) and rTHIK-2 (1900 bp). The nucleotide sequences predict proteins of 405 (rTHIK-1) and 430 (rTHIK-2) amino acids, respectively, which show the typical features of 2P K ϩ channels: four transmembrane regions, two pore-forming regions, a large extracellular linker between M1 and P1, a short N terminus and a larger C terminus. As indicated in Fig. 1, A and B, several putative intracellular phosphorylation sites for protein kinase A, protein kinase C, and casein kinase were found in both sequences. In the M1-P1 linker region, rTHIK-1 harbors two putative glycosylation sites (N-{P}-[S/T]-{P}), whereas rTHIK-1 possesses only one. The rTHIK-2 protein is significantly longer at the N terminus, including three putative protein kinase C phosphorylation sites and several repetitive amino acid elements.
Sequence comparison identified the human orthologs of THIK-2 in data base entry AC009600 and of THIK-1 in two overlapping data base entries (AL355074, AL137128). In accordance with the human nomenclature committee (HUGO), the two novel channel genes were named KCNK12 (THIK-2) and KCNK13 (THIK-1), respectively. The human channel subunits are 98% (hTHIK-2) and 84% (hTHIK-1), identical to their rat orthologs. hTHIK-1 and hTHIK-2 are 61.8% identical and 67.5% similar to each other, whereas only about 25-35% identity was found to other 2P K ϩ channels (Fig. 2B). Interestingly, the intracellular M2-M3 linker was 16 -20 amino acids longer than that of other cloned 2P K ϩ channels. The cytosolic Cterminal domains of the THIK channels are 57% identical to each other (see supplemental material), whereas no significant homology was found to other C-terminal domains of 2P K ϩ channel. The pore regions of both THIK channels are nearly identical (no mismatch in a stretch of 13 amino acids in the first pore region and only one mismatch in the second pore), whereas there is considerable difference to other 2P K ϩ channels (Fig. 2C). From the homology scores, the unique structural features of THIK channels and the phylogenetic clustering ( Fig. 2A), it is obvious that THIK 1 (KCNK13) and THIK-2 (KCNK12) form a novel subfamily among the 2P K ϩ channels. Supplemental material (multiple sequence alignment excluding the unrelated C-terminal domains and pairwise alignment of THIK-1 and THIK-2 C-terminal domains) is available in the on-line version.
Analysis of the entries of the human genome databases showed that both genes described here have a similar structure, with a single large intron of 48.1 kilobases (KCNK12) and 121.5 kilobases (KCNK13), respectively, splitting the coding region at the first pore GYG motif. This intron is conserved in all mammalian 2P K ϩ channel genes (KCNK1-10) cloned so far. Chromosomal localization assigned KCNK12 to the human chromosomal region 2p22-2p21 between markers WI-10633 and SGC34238. KCNK13 was assigned to chromosomal region 14q24.1-14q24.3 between markers IB3608 and SGC30527, which is very close to the human TREK-2 (KCNK10) gene. The results of the chromosomal assignments are summarized in Table I. The tissue distribution of the rTHIK-1 and rTHIK-2 was obtained by reverse transcription-PCR analysis. Intronspanning primers were used to avoid false positive results due to genomic contamination. rTHIK-1-specific products were amplified from all tissues tested. In contrast, rTHIK-2 expression was found in brain, lung, kidney, liver, stomach, and spleen but not in skeletal muscle, heart, and testis (Fig. 3). In addition, putative glycosylation sites (h), protein kinase A phosphorylation sites (a),protein kinase C phosphorylation sites (1) and casein kinase phosphorylation sites (⌬) are indicated below the amino acid sequence. Consensus sites were identified using the program PROSITE.
In Situ Hybridization of Rat Brain-The mRNA distribution of rTHIK-1 and rTHIK-2 in the adult rat brain as detected by in situ hybridization is highly differential, with little overlap (summarized in Table II). Two different antisense probes were used for either subunit. The resulting labeling profiles were identical for each subunit, indicating that the probes were specific. rTHIK-2 mRNA was found to be widely expressed in most brain regions, with highest levels in the cerebellar granule cell layer, the mitral and granule cell layers of the olfactory bulb (Fig. 4E), and in the anterodorsal and anteroventral nuclei of the thalamus (Fig. 4F). Strong signals were also found in the thalamic ventral posterior nuclei (Fig. 4G), cortex, hippocampus, the pontine nucleus, the red nucleus (Fig. 4E), the oculomotor nucleus, and some nuclei of the amygdala. In the brainstem, elevated levels were found in all trigeminal sensory nuclei, in the tegmental and reticular nuclei, and in the ventral cochlear nuclei. rTHIK-2 mRNA was also found in some non-neuronal cells such as the ependymal lining of the ventricles. rTHIK-2 was found to be absent from the substantia nigra, gracile nuclei, inferior olive, and from most parts of the caudate putamen and septal nuclei as well as all white matter pathways.
In contrast, rTHIK-1 mRNA expression was found to be rather weak and restricted to only a few brain regions and nuclei (Fig. 4, A-D). Substantial expression levels were detected only in the granule cell layer of the olfactory bulb, in the olfactory tubercle, the lateral septum (Fig. 4A), and in distinct hypothalamic and thalamic nuclei (ventromedial hypothalamic nucleus, lateral mamillary nucleus, reticular nucleus, reunions nuclei (Fig. 4, B and C)). Other rTHIK-1-positive structures were only partially labeled. The cortex, for example, was positive only in layer II, the striatum was labeled only in the caudal part, and the dentate gyrus granule cell layer of the dentate gyrus exhibited particularly high mRNA levels in the caudalventral part (Fig. 4D). Similarly, within a nuclear group, only specific subnuclei were positive, e.g. the dorsal subnucleus of the lateral septum (Fig. 4A), the magnocellular part of the red nucleus (Fig. 4D), a subnucleus of the lateral habenula (Fig. 4,  A and C), one lateral parabrachial subnucleus, and the ventral nucleus of the cochlear nuclei.
Heterologous Expression in Xenopus Oocytes-At day 2 after injection of rTHIK-1 cRNA into Xenopus oocytes, large currents were recorded using the two-electrode voltage clamp (Fig. 5A), which were not seen under control conditions. The steady-state amplitude of the currents measured at ϩ60 mV was 21.85 Ϯ 4.45 A (n ϭ 7) in the presence of 2 mM external K ϩ . Current activation in response to depolarizing voltage steps was not FIG. 2. Sequence comparison of eleven 2P K ؉ channels. A, phylogenetic tree. The GenBank TM accession numbers for the different channels are: hTHIK-1, AF287303; hTHIK-2, AF287302; hTASK-1, AF006823; hTASK-3, AF212829; hTREK-1, AF129399; rTREK-2, AF196965; hTRAAK, AF247042; hTWIK-1, U33632; hTWIK-2, AF117708; KCNK7, AF110522; and mTASK-2, AF084830. The CLUSTAL/CLUSTREE algorithm (Heidelberg UNIX Sequence Analysis Resources) was used for computation. B, pairwise sequence identity scores of THIK channels. The program GAP was used for calculation. C, multiple sequence alignment of the two pore regions of the 2P K ϩ channels. Conserved residues are shown in bold, and the only residue different in THIK-1 and THIK-2 pore domains is labeled with an asterisk. Additional information on sequence comparison (multiple alignment excluding C-terminal domains of the above mentioned channels, pairwise alignment of THIK-1 and THIK-2 C-terminal domains) is available on the on-line version. instantaneous but, when fitted to a single exponential, showed an activation time constant of 0.97 Ϯ 0.25 ms at ϩ60 mV. During a 500-ms voltage pulse, currents did not inactivate. The steady-state current-voltage relationship of rTHIK-1 showed only weak voltage dependence. With low external K ϩ , moderate outward rectification was observed, whereas with nearly symmetrical K ϩ , weak inward rectification was found (Fig. 5B). When external K ϩ was altered to examine the ion selectivity of rTHIK-1, the measured reversal potentials were very close to the K ϩ equilibrium potential (calculated on the assumption of an intracellular K ϩ concentration of 100 mM). The linear regression fit to the data gave a slope of Ϫ53 mV/decade (Fig. 5C), as would be expected for highly selective K ϩ channels. THIK-1 could be partially blocked by Ba 2ϩ ions; application of 1 mM Ba 2ϩ inhibited the instantaneous outward current measured at ϩ60 mV (in the presence of 2 mM external K ϩ ) by about 60% (n ϭ 4, not shown).
In contrast, the currents measured in Xenopus oocytes after injection of rTHIK-2 (Fig. 5D) were not significantly different from those of noninjected or water-injected oocytes (n ϭ 10). To test whether both channel proteins were translated and targeted to the surface membrane, rTHIK-1 and rTHIK-2 were tagged with EGFP at the N terminus. 48 h after injection of EGFP-rTHIK-2, confocal microscopy showed strong membrane fluorescence that was very similar to that found after injection of EGFP-rTHIK-1 (Fig. 6A). Thus rTHIK-2 subunits appear to be targeted to the outer membrane. The reason for the lack of functional expression of rTHIK-2 was further studied by constructing chimeric subunits in which the first part of the rTHIK-2-coding region (M1-P1-M2) was fused to the second half of rTHIK-1 (M3-P2-M4) and vice versa. Injection of the chimeric cRNAs did not induce any current in Xenopus oocytes.
Some voltage-activated (Kv) or inwardly rectifying (Kir) K ϩ channel subunits that do not functionally express as homomers can modify the activity of other subunits by heteromerization. Therefore we investigated the possibility that rTHIK-1 and rTHIK-2 subunits might coassemble to form heterodimeric channels. When rTHIK-1 and rTHIK-2 cRNAs were injected at equimolar amounts (n ϭ 5), all macroscopic current properties were virtually indistinguishable from rTHIK-1 currents, and current amplitudes were not significantly different (Fig. 6, B and C). Furthermore, the currents induced by injection of rTHIK-1 were unaffected by injection of larger amounts of rTHIK-2 cRNAs (ratio 1:5). The apparent absence of heteromerization of rTHIK-2 and rTHIK-1 subunits suggests that rTHIK-2 is not a regulatory subunit of the rTHIK-1 conductance.
Regulation of rTHIK-1 Channels-To test for functional similarities to other members of the two pore domain K ϩ channel family, we studied the modulation of rTHIK-1 channel activity by various experimental interventions such as extracellular acidification, application of the polyunsaturated fatty acid arachidonic acid, or application of the volatile anesthetic halothane (Fig. 7A). The outward current carried by rTHIK-1 was FIG. 3. Tissue distribution of rTHIK-1 and rTHIK-2. cDNA from the tissues indicated was analyzed for the presence of a 326-bp fragment of rTHIK-1 or a 250-bp fragment of rTHIK-2, indicated as a white arrow. Note that both primer pairs were intron-spanning to exclude possible genomic contamination. A pBR322/HaeIII marker was used in the left-most lane.

TABLE II
Distribution of rTHIK-1 and rTHIK-2 mRNA in the adult rat brain In situ hybridization signals obtained for 33 P-labeled oligonucleotide probes on adult rat brain sections were rated according to the relative grain density: ϩϩϩϩ, very abundant; ϩϩϩ, abundant; ϩϩ, moderate; ϩ, low; Ϯ, just above background; 0, no expression. Note that only selected brain regions with elevated expression levels or markedly differential expression patterns are included in the table. AHiPM, amygdalohippocampal area, posteromedial part; PMCo; posteromedial cortical amygdaloid nucleus; Mo5, -7, -12, motor nuclei of cranial nerves 5, 7, and 12. only weakly inhibited by extracellular acidification to pH 6, which is in marked contrast to the pronounced pH sensitivity of TASK-1 (KCNK3; half-maximal block at pH 7.38), as illustrated in Fig. 7B. Lowering the pH to 4.5 inhibited the outward current carried by rTHIK-1 by 34 Ϯ 8% (n ϭ 5). Similar to TASK-3, proton block of rTHIK-1 occurred with a fast time course (Fig. 7A) and was independent of the membrane potential (Fig. 7B). The sensitivity of rTHIK-1 to intracellular pH was tested by the "rebound acidification" technique. 20 mM NH 4 Cl was applied for 5 min and then removed, which should decrease the intracellular pH by about 1 unit. This intervention produced no measurable change in the current carried by rTHIK-1 (n ϭ 3), indicating that rTHIK-1 is not modulated by intracellular pH.
Since TREK-1 has been reported to be heat-sensitive (23), we also studied the temperature dependence of rTHIK-1. Raising the temperature from 22 to 37°C increased the current amplitude by a factor of 1.6 (n ϭ 3), in agreement with van't Hoff's rule, whereas TREK-1 is augmented by a factor of 10 under the same conditions (23). These findings suggest that, unlike TREK-1, rTHIK-1 is not a heat-sensitive channel. Lysophosphatidylcholine (3 M) has also been shown to activate TREK-1 and TRAAK (24). Application of 10 M lysophosphatidylcholine to Xenopus oocytes expressing rTHIK-1 induced only a very minor (up to 20%) increase in current amplitude, which may be attributable to a nonspecific effect of the lysophospholipid.
Application of arachidonic acid to the bath solution induced a rapid increase in the outward current carried by rTHIK-1 channels (Fig. 7A). This effect could be washed out within 5 min. In the presence of 5 M arachidonic acid, the current was increased by 85 Ϯ 24% (n ϭ 5) at ϩ30 mV. As can be seen from Fig. 7C, the current activated by arachidonic acid was outwardly rectifying. It reversed at the calculated potassium equilibrium potential when the external K ϩ concentration was altered. The concentration dependence of the effect of arachidonic acid on rTHIK-1 could be described by a K d of 0.98 M and a Hill coefficient of 1.97 (Fig. 7C). The effects of arachidonic acid reported here are similar to the effects found in TREK-1 (KCNK2) and TRAAK (KCNK4) (7,9,24).
Two of the known 2P K ϩ channels, TREK-1 and TASK-1, are activated by the volatile anesthetic halothane (27). Surprisingly, our whole-cell recordings in Xenopus oocytes showed that the current carried by rTHIK-1 was rapidly and reversibly inhibited by halothane (Fig. 7, A and D). When the membrane potential was held at ϩ30 mV, application of 5 mM halothane reduced rTHIK-1 currents by 56 Ϯ 5% (n ϭ 5). The fit of the concentration-effect curve gave a K d of 2.83 mM and a Hill coefficient of 1.06. Under the same experimental conditions, application of chloroform (1 mM) had no effect (n ϭ 3; data not shown). DISCUSSION THIK-1 and THIK-2 are the first two members of a novel 2P K ϩ channel subfamily. Although they show the typical features of other 2P K ϩ channels, such as four transmembrane regions, two pore-forming regions, and a large extracellular M1-P1 linker region, the two novel channels are only about 25-35% identical to the other known 2P K ϩ channels but 61.8% identical to each other. One notable difference is the larger cytosolic M2-M3 linker region, containing three (rTHIK-2) or one (rTHIK-1) putative phosphorylation site(s). In addition, the THIK-2 protein has an unusual N terminus containing several repeats of proline, arginine, and cysteine residues. The Cterminal domain of THIK-1 and THIK-2 shows no significant homology to other mammalian 2P K ϩ channels.
The reversal potential of the current induced by injection of rTHIK-1 cRNA in Xenopus oocytes followed the calculated K ϩ equilibrium potential when external K ϩ was changed. This suggests that the THIK channels are mainly permeable to K ϩ ions. The whole-cell current induced by heterologous expres- FIG. 5. Heterologous expression of rTHIK-1 and rTHIK-2 in Xenopus oocytes. Whole-cell current recordings from oocytes injected with cRNA of rTHIK-1 (A-C) and rTHIK-2 (D), respectively. A, voltage steps of 500-ms duration from a holding potential of Ϫ60 mV to potentials between Ϫ140 to ϩ60 mV. B, voltage ramps (105 mV s Ϫ1 ) from Ϫ150 to ϩ60 mV at external potassium concentrations of 2, 5, 10, 25, and 96 mM K ϩ , respectively (the holding potential preceding the ramp pulses was Ϫ60 mV). C, semi-logarithmic plot of the measured zero-current (reversal) potentials versus [K ϩ ] e . A linear regression line with a slope of 53 mV per decade was fitted to the data points. D, voltage ramps in a Xenopus oocyte injected with rTHIK-2 cRNA. The ramp protocol was identical to that shown for rTHIK-1 in panel B.

FIG. 6. Targeting of rTHIK-1 and rTHIK-2 and coexpression in
Xenopus oocytes. A, fluorescence signal of EGFP-labeled rTHIK-1 and rTHIK-2 in the oocyte membrane 48 h after cRNA injection. B, wholecell currents recorded from Xenopus oocytes after expression of rTHIK-1 alone and after coexpression of rTHIK-1 together with rTHIK-2 at RNA ratios of 1:1 and 1:5. Voltage ramps from Ϫ150 mV to ϩ60 mV were applied; the preceding holding potential was Ϫ60 mV. C, bar graph of the currents measured at ϩ60 mV. No significant difference was found between oocytes injected with rTHIK-1 alone and in combination with rTHIK-2.
sion of rTHIK-1 displayed outward rectification at physiological external K ϩ and weak inward rectification with approximately symmetrical K ϩ concentrations. It could be activated by arachidonic acid (K d , 0.98 M; Hill coefficient, 1.97) and inhibited by halothane (K d , of 2.8 mM; Hill coefficient, 1.06). Chloroform had no effect on rTHIK-1. Another 2P K ϩ channel, TWIK-2, which is almost absent in the brain, has recently been found to be inhibited by both halothane and chloroform (15). In contrast, both TREK-1 and TASK-1 are activated by halothane and isoflurane, and TREK-1 is additionally activated by chloroform and diethyl ether (27). TREK-1, TASK-1, and THIK-1 are all expressed in specific regions of the brain. The findings reported here suggest that the effects of volatile anesthetics on the brain may be more complex than hitherto assumed. We are aware of the fact that the IC 50 for the effects of halothane on THIK-1 is higher than the EC 50 for the anesthetic effects in vivo (Ϸ250 M; Ref. 28). Nevertheless, we decided to name the new channels tandem pore-domain halothane inhibited K ϩ channels because this discriminates them from some of the other 2P K ϩ channels.
Injection of rTHIK-2 cRNA in Xenopus oocytes did not produce any measurable currents. The lack of functional expression of rTHIK-2 was apparently not due to inadequate targeting, because confocal microscopy showed EGFP-tagged rTHIK-2 channels in the outer cell membrane (Fig. 6). To localize possible structural constraints in rTHIK-2 that prevent expression of functional channels, we constructed rTHIK-1/ rTHIK-2 and rTHIK-2/rTHIK-1 chimeras. However, since neither of the two chimeras was functional, the reason for the nonfunctional state of rTHIK-2 in Xenopus oocytes remains unclear. Another possibility is that rTHIK-2 might be a regulator of rTHIK-1 conductance by coassembling with rTHIK-1. This is unlikely, because whole cell currents produced by injection of rTHIK-1 cRNA were unaffected by coinjection even of 5-fold larger amounts rTHIK-2 cRNA. In situ hybridization in the brain showed little overlap between rTHIK-1 and rTHIK-2, which also argues against heteromerization. In conclusion, the high expression of rTHIK-2 in cerebral cortex, hippocampus, and olfactory bulb and the specific expression in several nuclei (Fig. 6) support the idea that rTHIK-2 is functionally important in neurons, but the available experimental evidence suggests that rTHIK-2 requires an accessory subunit or specific intracellular ligands to form a conducting pore. The strong expression in lung, kidney, and stomach suggests that rTHIK-2 may also play a role in epithelial cells.
Both human THIK genes described here have only one very large intron, 121.5 kilobases in the THIK-1 gene (KCNK13) and 48.1 kilobases in the THIK-2 gene (KCNK12), that splits the coding region of the first pore (GYG motif). In this respect, the THIK genes are similar to the TASK-3 gene (KCNK9). The genes of the TWIK and the TREK-/TRAAK families possess several introns in the coding region (10,29,30), but an intron splitting the GYG motif of the first pore region is conserved in all mammalian 2P K ϩ channel genes cloned so far and, in addition, in most of the 2P K ϩ channels of Caenorhabditis elegans (31) and Drosophila melanogaster. Another interesting feature is the colocalization of 2P K ϩ channel genes at the same chromosomal region. We have shown that THIK-1 (KCNK13) maps to the same region on the long arm of chromosome 14 as TREK-2 (KCNK10). The other 2P K ϩ channel gene pairs known so far are TRAAK (KCNK5) and KCNK7 on chromosome 11q13 (29) and TWIK-1 (KCNK1) and TREK1 (KCNK2) on chromosome 11q41-43 (32). These data are consistent with a common ancestral gene for all 2P K ϩ channels that was subject to several duplication events.
As illustrated in Fig. 2, mammalian 2P K ϩ channels can be subdivided in five subfamilies: 1) TWIK/KCNK7, 2) TREK/ TRAAK, 3) TASK-1/TASK-3, 4) THIK, and 5) TASK-2. These channels display a wide variety of electrophysiological and regulatory characteristics that are usually not confined to one of the subfamilies. The steady-state current voltage relation measured in the whole-cell configuration at symmetrical K ϩ concentrations was found to be inwardly rectifying (TWIK-1, TWIK-2, and THIK-1), outwardly rectifying (TREK-1, TASK-3), or linear (TREK-1, TASK-1, TRAAK). Some of the 2P K ϩ channels show a pronounced sensitivity to intracellular (TREK-1) or extracellular pH (TASK-1, TASK-2, TASK-3). The pharmacological profile of the 2P K ϩ channels is also very diverse and not related to subfamilies. Some 2P K ϩ channels are activated by volatile anesthetics (TREK-1, TASK-1); other channels are inhibited (TWIK-2, THIK-1). Some channels are activated by fatty acids such as arachidonic acid (TREK-1, TRAAK, TWIK-2, THIK-1) or by phospholipids such as lysophosphatidylcholine (TREK and TRAAK). In addition, some of the 2P K ϩ channels are mechanosensitive (TREK-1, TREK-2, and TRAAK) or heat-sensitive (TREK-1). The most remarkable common characteristic of all 2P K ϩ channels known so far is that their regulation by physical and chemical stimuli is very complex. The difficulty in identifying their function is probably related to this complex regulation, which needs to be studied in the native cells in which the channels are expressed.