Formation of Functional Heterodimers by TREK-1 and TREK-2 Two-pore Domain Potassium Channel Subunits*

Two-pore domain (K2P) potassium channels are the major molecular correlates of the background (leak) K+ current in a wide variety of cell types. They generally play a key role in setting the resting membrane potential and regulate the response of excitable cells to various stimuli. K2P channels usually function as homodimers, and only a few examples of heteromerization have been previously reported. Expression of the TREK (TWIK-related K+ channel) subfamily members of K2P channels often overlaps in neurons and in other excitable cells. Here, we demonstrate that heterologous coexpression of TREK-1 and TREK-2 subunits results in the formation of functional heterodimers. Taking advantage of a tandem construct (in which the two different subunits were linked together to enforce heterodimerization), we characterized the biophysical and pharmacological properties of the TREK-1/TREK-2 current. The heteromer was inhibited by extracellular acidification and by spadin similarly to TREK-1, and its ruthenium red sensitivity was intermediate between TREK-1 and TREK-2 homodimers. The heterodimer has also been distinguished from the homodimers by its unique single channel conductance. Assembly of the two different subunits was confirmed by coimmunoprecipitation of epitope-tagged TREK-1 and TREK-2 subunits, coexpressed in Xenopus oocytes. Formation of TREK-1/TREK-2 channels was also demonstrated in native dorsal root ganglion neurons indicating that heterodimerization may provide greater diversity of leak K+ conductances also in native tissues.

Two-pore domain (K 2P ) potassium channels are the major molecular correlates of the background (leak) K ؉ current in a wide variety of cell types. They generally play a key role in setting the resting membrane potential and regulate the response of excitable cells to various stimuli. K 2P channels usually function as homodimers, and only a few examples of heteromerization have been previously reported. Expression of the TREK (TWIKrelated K ؉ channel) subfamily members of K 2P channels often overlaps in neurons and in other excitable cells. Here, we demonstrate that heterologous coexpression of TREK-1 and TREK-2 subunits results in the formation of functional heterodimers. Taking advantage of a tandem construct (in which the two different subunits were linked together to enforce heterodimerization), we characterized the biophysical and pharmacological properties of the TREK-1/TREK-2 current. The heteromer was inhibited by extracellular acidification and by spadin similarly to TREK-1, and its ruthenium red sensitivity was intermediate between TREK-1 and TREK-2 homodimers. The heterodimer has also been distinguished from the homodimers by its unique single channel conductance. Assembly of the two different subunits was confirmed by coimmunoprecipitation of epitope-tagged TREK-1 and TREK-2 subunits, coexpressed in Xenopus oocytes. Formation of TREK-1/TREK-2 channels was also demonstrated in native dorsal root ganglion neurons indicating that heterodimerization may provide greater diversity of leak K ؉ conductances also in native tissues.
Heteromerization of subunits is a widespread mechanism for increasing diversity in voltage-gated (K v ) and inwardly rectifying (K ir ) K ϩ channel subfamilies. The resulting heteromers are typically characterized by different biophysical and regulatory properties from the homomers of the parent subunits. Certain K v or K ir (e.g. K v 8.1 or K ir 5.1) subunits cannot form functional homomers on their own; however, via heteromerization they can also gain important physiological function (1); deletion or mutations of their genes have been associated with genetic diseases and pathophysiological conditions (2)(3)(4).
Two-pore domain (K 2P ) K ϩ channels are the molecular correlates of background (leak) K ϩ currents. The mammalian K 2P family consists of 15 different members, which are grouped into six subfamilies as follows: TWIK (tandem of pore domains in a weak inward-rectifying K ϩ channel); TREK (TWIK-related K ϩ channel); TASK (TWIK-related acid-sensitive K ϩ channel); TALK (TWIK-related alkaline pH-activated K ϩ channel); THIK (tandem pore domain halothane-inhibited K ϩ channel); and TRESK (TWIK-related spinal cord K ϩ channel) (5). In contrast to K v and K ir tetramers containing one pore domain per subunit, K 2P subunits possess two pore domains and accordingly dimerize to form functional channels. Only a few examples of heterodimerization have been described in the K 2P channel family so far. The first example, heteromerization of TASK-1 with TASK-3, was characterized in our laboratory (6). The preferentially formed TASK-1/TASK-3 heterodimer has since been detected in several native cells (7)(8)(9)(10)(11)(12). The heterodimer plays a role in fine-tuning the resting membrane potential and the sensitivity of the cells to physiological stimuli (extracellular pH changes and agonists of G q -protein-coupled receptors) in glomus cells, cerebellar granule neurons, and motoneurons. Subsequently, further heterodimers of K 2P channel subunits have been recognized. The heterodimerization of THIK-1 and THIK-2 has been reported, and apparently THIK-1 was able to counteract the endoplasmic retention of THIK-2 (13). Another K 2P channel interaction (in which the partner subunits belong to different subfamilies), the heteromerization of the silent TWIK-1 subunit with TASK-1 (or TASK-3), has recently been revealed. Properties of the heteromer were analyzed in detail in cerebellar granule cells and also in a heterologous expression system; TWIK-1 was found to confer SUMOylation sensitivity to the heteromer (14). In this study TWIK-1 and TREK-1 coexpression served as a negative control, as no sign of interaction was found between these two subunits. In contrast, TWIK-1/TREK-1 heteromerization was reported by another group in hippocampal astrocytes (15). Quite unexpectedly, activation of a G i -protein-coupled receptor was claimed to convert the heterodimer into an anion conductive channel (15); the authors have suggested that the binding of the ␤␥ subunits to the heterodimer induces channel pore dilation resulting in glutamate permeability. This intriguing hypothesis, however, remains controversial and has yet to be proven experimentally.
The closely related TREK-1 and TREK-2 channels (encoded by the KCNK2 and KCNK10 (potassium channel subfamily K member 2 and 10) genes) are thermo-and mechanosensitive, and they are activated by volatile anesthetics, intracellular acidosis, arachidonic acid, and also similarly modulated by a wide variety of signaling cascades initiated by activation of G-protein-coupled receptors. In contrast, TREK-1 and TREK-2 chan-nels are regulated differently by extracellular (EC) 2 acidification; TREK-1 is inhibited, and TREK-2 is activated (16,17). TREK-1 is inhibited by the sortilin-derived neuropeptide spadin, and TREK-2 is insensitive to spadin (18,19). Recently, we described the differential pharmacological effect of the polycationic dye ruthenium red (RR) on TREK channels; TREK-2 is potently inhibited by the dye, whereas TREK-1 is insensitive to RR (20). The overlapping expression of the two TREK subunits raises the possibility of heteromerization in native tissues, which may result in a channel with new attributes (e.g. different single channel conductance, pH sensitivity, and distinct regulatory profile). In this paper, by applying pharmacological tools combined with single channel measurements, we demonstrate that TREK-1 and TREK-2 subunits form functional heterodimers when coexpressed in Xenopus laevis oocytes, HEK293T cells, and isolated dorsal root ganglion neurons. The properties of the heterodimer proved to be different from those of the homodimers, thus further expanding the diversity of leak K ϩ currents.

Experimental Procedures
Materials and Methods-Chemicals of analytical grade were purchased from Sigma, Fluka (Milwaukee, WI), or Merck. Enzymes and kits for molecular biology applications were purchased from Ambion (Austin, TX), Thermo Scientific (Waltham, MA), New England Biolabs (Beverly, MA), and Stratagene ( La Jolla, CA). The sortilin-derived peptide spadin was synthesized by the MTA-ELTE Research Group of Peptide Chemistry (Budapest, Hungary), dissolved in distilled water, and stored as 1 mM stock solution at Ϫ20°C.
Molecular Biology-The plasmid coding mouse TREK-1 (mTREK-1) channel was kindly provided by Professor M. Lazdunski and Dr. F. Lesage. The cloning of mouse TREK-2 channel from mouse brain has been described previously (21). For expression in mammalian cell lines, the coding sequences of TREK-1 and TREK-2 were subcloned into the pcDNA3.1 expression vector. For coexpression of TREK-1 and TREK-2 channel subunits in the same cell, the two channel subunits were linked by a viral peptide, T2A, which displays a self-cleaving property during translation (22) and was shown to ensure equimolar separate expression of the two proteins (23).
For assembly of the TREK-2/TREK-1 concatenated (tandem) construct, the coding sequence of TREK-2 (without the stop codon) and that of TREK-1 (without the start codon) were amplified from mTREK-2-pEXO and mTREK-1-pEXO plasmids by PCR, respectively. The TAA stop codon of the TREK-2 subunit and the ATG start codon of the TREK-1 subunit were replaced with unique MunI restriction enzyme sites. This introduced a two-amino acid (glutamate and leucine) linker between the two concatenated subunits after final ligation into a pXEN vector between Eco81I and XbaI restriction sites.
Cloning of TREK Mutants-The "long" version of mTREK-1 (mTREK-1-L), from which translation can only initiate from the first ATG codon of the mTREK-1 coding sequence, was created by mutating the alternative translation initiation site (Met-58) of mTREK-1 to isoleucine (24) with the QuikChange site-directed mutagenesis kit (Stratagene, La Jolla, CA) according to the manufacturer's instructions. The long version of TREK-2 (M55L and M67L), used in coimmunoprecipitation experiments, was constructed with a similar approach. The "short" version of mTREK-2 (mTREK-2-S), where translation can only initiate from the last downstream start codon of the wild type TREK-2 sequence, was created by amplifying the coding sequence of mTREK-2-pEXO plasmid without the first 66 amino acids of the N terminus. The PCR product was inserted into pXEN vector between Eco81I and XbaI restriction sites.
During the assembly of TREK-2-S/TREK-1-L tandem construct, the same strategy was followed as detailed above at the TREK-2/TREK-1 construct; however, the templates of the PCRs were replaced by the mutant subunits.
TREK-1-L and TREK-2-L subunits were labeled with epitope tags (V5 and FLAG respectively) using PCR. The fidelity of all the constructs generated by PCR was verified by automated sequencing.
cRNA Synthesis-The plasmids coding wild type, tagged, or mutant channel constructs were linearized with NotI, XhoI, or XbaI restriction enzymes. These templates were used for in vitro RNA synthesis. RNA was synthesized using the Ambion mMESSAGE mMACHINE TM T7 in vitro transcription kit (Ambion, Austin, TX). The structural integrity of the RNA was checked on denaturing agarose gels, and RNA was quantified spectrophotometrically.
Animals, Tissue Preparation, Xenopus Oocyte Microinjection, and Isolation of Dorsal Root Ganglion Neurons-X. laevis oocytes were prepared as described previously (6). Oocytes were injected 1 day after defolliculation with 2-6 ng of RNA per oocyte for two-electrode voltage clamp and membrane preparation experiments. For single channel recordings, 0.025-0.5 ng of RNA per oocyte was used. Fifty nanoliters of the appropriate RNA solution was delivered with Nanoliter Injector (World Precision Instruments, Sarasota, FL). Dorsal root ganglion (DRG) neurons were isolated from 3-month-old FVB/Ant mice obtained from the Institute of Experimental Medicine of the Hungarian Academy of Sciences (Budapest, Hungary). Their preparation and culturing were described previously (20). All the animal experiments were approved by the Animal Care and Ethics Committee of Semmelweis University (approval ID: XIV-I-001/2154-4/2012).
Electrophysiology-Two-electrode voltage clamp experiments were performed 2-3 days after the microinjection of cRNA into Xenopus oocytes, as described previously (25). The holding potential was Ϫ75 mV. Background K ϩ currents were measured at the end of 300-ms-long voltage steps to Ϫ100 mV applied every 4 s. Low [K ϩ ] solution contained (in mM): NaCl 95.4, KCl 2, CaCl 2 1.8, HEPES 5. High [K ϩ ] solution contained 40 mM K ϩ (38 mM Na ϩ of the low [K ϩ ] solution was replaced with K ϩ ). The pH of the solutions was adjusted to 8.5 or 6.5 with NaOH on the day of the experiment. RR was diluted in high [K ϩ ] solution before the experiment from a 10 mM stock solution in 0.1 M ammonium acetate. Experiments were carried out at room temperature (21°C). Solutions were applied using a gravity-driven perfusion system. Data were analyzed by pCLAMP 10 software (Molecular Devices, Sunnyvale, CA).
For patch clamp experiments, pipettes were pulled from thick-walled borosilicate glass (Standard Glass Capillaries, 4 inches, 1.2/0.68 outer/inner diameter, filament/fire-polished (item number 1B120F-4) from World Precision Instruments, Sarasota, FL) by a P-87 puller (Sutter Instrument Co., Novato, CA) and fire-polished. Pipettes were filled with pipette solution and connected to the head stage of an Axopatch-1D patch clamp amplifier (Axon Instruments, Inc., Foster City, CA). Experiments were carried out at room temperature (21°C). Solutions were applied using a gravity-driven perfusion system. Data were digitally sampled by Digidata 1200 (Axon Instruments, Inc.). Data were analyzed by pCLAMP 10 software.
Single channel currents were recorded from X. laevis oocyte membrane patches 2 days after injection or from DRG neurons 1-2 days after isolation. Oocytes were devitellinized manually in modified Barth's saline after a brief exposure to hyperosmotic conditions (500 mosM). Cutoff frequency of the eightpole Bessel filter was adjusted to 2 kHz, and data were acquired at 20 kHz. Data were analyzed by pCLAMP 10 software. Recordings were low pass filtered at 200 Hz with pCLAMP10 software before analysis.
On inside-out membrane patches single channel currents were recorded at a membrane potential of ϩ60 mV. The pipette solution contained (in mM): 136 NaCl, 4 KCl, 5 EGTA, 1 MgCl 2 , and 10 HEPES (pH 7.4 adjusted with NaOH). Pipette resistances were 3-7 megaohms when filled with pipette solution. Bath solution contained (in mM) the following: 140 KCl, 5 EGTA, 1 MgCl 2 , and 10 HEPES (pH 7.1 adjusted with KOH). The charge carrier of the currents recorded was verified as K ϩ in all patches by transiently substituting the bath solution with a K ϩ -free bath solution (140 NaCl, 5 EGTA, 1 MgCl 2 and 10 HEPES (pH 7.1), adjusted with NaOH).
During the measurements on outside-out membrane patches, single channel currents were measured at a membrane potential of Ϫ60 mV. The pipette solution contained (in mM) the following: 140 KCl, 5 EGTA, 1 MgCl 2 and 10 HEPES (pH 7.1 adjusted with KOH). Pipette resistances were 3-6 megaohms when filled with pipette solution. Bath solution contained (in mM) the following: 140 KCl, 5 EGTA, 1 MgCl 2 , and 10 HEPES (pH adjusted to 7.4 or 6.4 with KOH). RR or spadin were dissolved in the bath solution on the day of the measurement. Verification of the charge carrier as K ϩ was done by determining the reversal potential of the current.
Data were analyzed to obtain amplitude histograms and channel activity (NP o , where N is the number of channels in the patch, and P o is the probability of a channel being open). Single channel current amplitudes were determined from the amplitude histograms. NP o was determined from 30 to 60 s of current recording.
Statistics-Results are expressed as means Ϯ S.E. Statistical significance was determined using Student's t test for independent samples or analysis of variance followed by Tukey's post hoc test. The difference was considered to be significant at p Ͻ 0.05. All statistical calculations were done using Origin8 (OriginLab Corp., Northampton, MA).
Pulldown Experiments-Membrane fraction was prepared from X. laevis oocytes coinjected with wild type or tagged TREK-1 and TREK-2 cRNA after 2 days of expression. All preparation steps were performed at 4°C. Thirty oocytes in each group were homogenized in a Potter homogenizer in buffer H contained (in mM) the following: 90 NaCl, 1 MgCl 2 , 10 HEPES (pH 7.9), and a protease inhibitor mixture. The lysate was centrifuged at 1,000 ϫ g for 10 min. The supernatant was saved and the centrifugation step repeated. The "low speed" supernatant was centrifuged at 16,000 ϫ g for 20 min, and the pellet (containing the membrane fraction) was washed with buffer H and resuspended in solubilizing (S) solution containing (in mM) the following: 250 NaCl, 2 EDTA, 50 HEPES (pH 7.4), 10% glycero, 1% Triton X-100, and a protease inhibitor mixture.
Solubilization for 3 h was followed by centrifugation (16,000ϫ g, 20 min), and the supernatant (solubilized membrane fraction) was added to an equilibrated 30 l of immunoaffinity agarose gel containing goat polyclonal Ab to V5 tag (1229, Abcam, Cambridge, UK). Reactions were incubated overnight by rotating the slurry. The resin was pelleted at 5,000 ϫ g for 1 min and washed three times with solution S. Proteins were eluted with SDS sample buffer, and aliquots (equivalent with the membrane fraction of 1 oocyte) were heated to 90°C for 15 min and were subjected to SDS-PAGE and Western blot. Interaction between TREK-1-V5 and TREK-2-FLAG subunits was detected using an anti-FLAG M2 mouse monoclonal antibody (F3165, Sigma) 1:10,000 and a horseradish peroxidase-conjugated anti-mouse IgG from goat 1:10,000 (R05071), Advansta, Menlo Park, CA).

Construction of a Functional TREK-2/TREK-1 Tandem
Channel-We have previously shown that functional TREK-2/ TREK-2 tandem channels can be engineered by concatenating two identical coding regions (20). Electrophysiological and pharmacological properties of these tandem constructs were indistinguishable from the native post-translationally dimerized TREK-2 channels. In the present experiments, the C terminus of TREK-2 was joined to the N terminus of TREK-1 via a two amino acid linker. Expression of this construct in X. laevis oocytes resulted in a potassium conductance as measured by the two-electrode voltage clamp technique. Assembly of the covalently linked subunits into a heterodimer was indicated by the absence of leak K ϩ current expression by a mutant tandem TREK-2/TREK-1-mut construct in which the signature sequence in the first pore domain of the TREK-1 subunit was changed from GFG to GFE (the K ϩ current of the oocytes expressing the mutant tandem was 6% compared with those expressing the wild type TREK-2/TREK-1 tandem (n ϭ 5); this current was in the range of non-injected oocytes). Accordingly, the tandem channel was considered as a plausible model for heteromerization, and it has been used to predict the pharmacological properties of the heterodimer composed of native TREK-1 and TREK-2 subunits.
TREK-2/TREK-1 Tandem Channel Has Intermediate RR Sensitivity and Is Inhibited by Extracellular Acidification-Pharmacological effects of extracellular acidification and RR on different TREK currents were determined in Xenopus oocytes at Ϫ100 mV in high (40 mM) [K ϩ ] solution. In good accordance with the previous results (16,17), TREK-1 was inhibited by changing the pH from 8.5 to 6.5 (49.7 Ϯ 4.3% inhibition, n ϭ 10, Fig. 1, A and D), whereas TREK-2 was substantially activated by the same treatment (the current increased to 243.3 Ϯ 17.5% of the value measured at pH 8.5, n ϭ 6, Fig. 1, B and D). The TREK-2/TREK-1 tandem construct was inhibited by EC acidification (by 32.7 Ϯ 1.4%, n ϭ 8, Fig. 1, C and D). This degree of inhibition of the tandem channel is not significantly different from that of the TREK-1 homodimer (p ϭ 0.34), suggesting that the inhibitory mechanism responsible for the pH sensitivity of TREK-1 is dominant over the activating effect on TREK-2 subunit.
As we have recently reported (20), RR polycation unambiguously differentiates between the K ϩ currents of TREK-1 and TREK-2 homodimeric channels on the basis of a single amino acid residue difference in the closely related subunits (negatively charged aspartate in TREK-2 and neutral isoleucine in TREK-1). To test whether EC pH influences the inhibition elicited by RR, the oocytes expressing TREK-2 were challenged with the polycationic dye (3 M) in two different solutions adjusted to pH 6.5 or 8.5. TREK-2 current was reduced by 76 Ϯ 5% at pH 6.5 and 81 Ϯ 3% at pH 8.5, respectively (p ϭ 0.39), indicating that EC pH does not affect the inhibition evoked by RR (data not shown). Therefore, in further experiments, the effects of acidification and RR were measured in the same oocyte by reducing pH (from 8.5 to 6.5) and subsequently applying RR (30 M) in the acidified solution. TREK-2 current FIGURE 1. Different inhibitory profiles of TREK-1, TREK-2, and the TREK-2/TREK-1 tandem channel. TREK-1, TREK-2, and TREK-2/TREK-1 tandem channels were expressed in Xenopus oocytes. Oocytes were measured with the two electrode voltage clamp method 2-3 days after injection. Currents were measured at Ϫ100 mV. Holding potential was Ϫ75 mV. [K ϩ ] was switched from 2 to 40 mM and then back to 2 mM as shown above the graphs. Acidification of the bath solution and application of 30 M RR are indicated by the horizontal bars above the recordings. The nonspecific leak current measured in 2 mM K ϩ was subtracted from the currents. The currents were normalized to the value measured in 40 mM K ϩ . The average of multiple recordings was plotted; error bars indicate S.E. A, TREK-1 is inhibited by extracellular acidification and is RR-insensitive (n ϭ 10 oocytes). B, TREK-2 is activated by extracellular acidification and is potently inhibited by RR (n ϭ 6 oocytes). C, TREK-2/TREK-1 tandem channel is inhibited by extracellular acidification and has an intermediate RR sensitivity (n ϭ 8 oocytes). D, sensitivity of TREK channels to extracellular acidification. TREK-1 (white circles) and the tandem channel (black diamonds) are inhibited (there was no significant difference between the degree of inhibition, p ϭ 0.34), although TREK-2 (gray triangles) is activated by lowering the extracellular pH. Averages of the groups are plotted as columns. E, sensitivity of TREK channels to RR. TREK-1 (white circles) is insensitive, although TREK-2 (gray triangles) is greatly inhibited by RR. The tandem channel (black diamonds) has an intermediate sensitivity. The degree of inhibition is significantly different between all three groups (p Ͻ 10 Ϫ6 ). Averages of the groups are plotted as columns.
Thus, we have identified two pharmacological properties of the TREK-2/TREK-1 tandem channel, which were inherited in different manners from the parent subunits. TREK-2/TREK-1 tandem was nearly identical to TREK-1 with respect to acidification, but the response of the tandem channel to RR was intermediate between TREK-1 and TREK-2 homodimers. To present this graphically, the effects of decreased EC pH and RR application on TREK currents were plotted against each other (one point corresponds to one oocyte, Fig. 2A). The data point averages of TREK-1 and TREK-2 homodimers were connected with a straight line. The data points of oocytes expressing the TREK-2/TREK-1 tandem channel clearly deviated from the connecting line ( Fig. 2A). This illustrates that the covalent linkage of the different subunits resulted in an offspring that inherited the pharmacological properties of the parent channels asymmetrically. This asymmetrical inhibitory profile was exploited in the further experiments to investigate the heterodimers formed by separate TREK-1 and TREK-2 subunits.
Coexpression of TREK-1 and TREK-2 Subunits Results in TREK-1/TREK-2 Heterodimer Formation-Covalent linkage of TREK-1 and TREK-2 subunits forces the formation of functional "heterodimeric" channels, which are characterized by pharmacological properties different from the homodimers. To test whether separate TREK-1 and TREK-2 subunits can also coassemble spontaneously, oocytes were injected with different ratios of TREK-1 and TREK-2 cRNA. The pH and RR sensitivities of the resulting TREK currents were determined and plotted in the way established above (Fig. 2B). Theoretically, if the two coexpressed subunits did not form heterodimers then two independent populations of homodimers would be produced in each cell. In this case, the data points corresponding to the oocytes would fall on the line connecting the averages of TREK-1 and TREK-2 homodimers, and the exact position of the data point on the line would reflect the ratio of the two coexpressed homodimeric channel types. However, data points from the coexpression experiment clearly deviated from the connecting line, and approximated the position of tandem channel data points, depending on the ratio of TREK-1 to TREK-2 expression. This suggests that functional heterodimers were spontaneously constituted by individual TREK-1 and TREK-2 subunits.
Verification of Heterodimer Formation by Measuring Single Channel Conductance-Another approach to distinguish TREK-1, TREK-2, and the heteromeric channel is to measure their characteristic single channel conductance. However, this approach is complicated by the fact that multiple conductance levels arise for both TREK-1 and TREK-2 channels due to alternative translation initiation (24,26). To overcome this heterogeneity, mutant channels were constructed. On the basis of literature data, the isoforms of native TREK-1 and TREK-2 with maximum difference between their single channel conductance were produced. N-terminally deleted short TREK-2 (TREK-2-S) was translated only from the downstream native initiation site, but in long M58I mutant TREK-1 (TREK-1-L) the additional site of translation initiation was disabled by mutation. In this way, each channel was translated starting only from one initiation site, and multiple conductance levels have been eliminated. A version of the tandem channel (TREK-2-S/ TREK-1-L) was also constructed from the modified subunits. The pH and RR sensitivities of the mutants were checked and found to be similar to those of the wild type channels. When TREK-1-L and TREK-2-S were coexpressed, the sensitivity of the macroscopic current to acidification and RR showed a similar pattern as that obtained with wild type TREK-1 and TREK-2 (data not shown).
Single channel currents were recorded from inside-out membrane patches of oocytes expressing TREK-1-L, TREK-2-S, or TREK-2-S/TREK-1-L tandem constructs. Unitary currents and single channel conductances were calculated from FIGURE 2. TREK-1 and TREK-2 subunits form functional heterodimers in Xenopus oocytes. A, Xenopus oocytes were injected with TREK-1 (circles), TREK-2 (triangles), or TREK-2/TREK-1 tandem (diamonds) cRNA. TREK currents were measured as described in Fig. 1. The sensitivity of the currents to extracellular acidification and RR application was plotted against each other. Averages of the three groups (empty diamonds) are shown with error bars representing S.E. The straight line connects the averages of the groups expressing TREK-1 or TREK-2 homodimers, respectively. B, Xenopus oocytes were injected with different ratios of TREK-1 and TREK-2 cRNA (squares). The pH and RR sensitivity of the expressed TREK currents were measured as in Fig. 1 and plotted as in A. The averages (with error bars) of the three groups and the straight line from A are also displayed.
30-to 60-s periods of the recordings. Under our experimental conditions, the single channel conductance of TREK-1-L was 39.2 Ϯ 2.7 pS (n ϭ 6 patches, Fig. 3, A and C). TREK-2-S had a large conductance of 148.8 Ϯ 4.2 pS (n ϭ 7 patches, Fig. 3, A and   C). The conductance of TREK-2-S/TREK-1 tandem channel was 83.5 Ϯ 5.0 pS, an intermediate value (n ϭ 8 patches, Fig. 3, A and C). The difference between the conductance of the homodimers and the tandem channel was statistically significant (p Ͻ 0.01). Thus, the TREK-1-L/TREK-2-S heterodimer can be distinguished from both TREK-1-L and TREK-2-S based on its single channel conductance (Fig. 3D).
Patches from Oocytes Coexpressing TREK-1-L and TREK-2-S Contain Heterodimeric Channels-Xenopus oocytes were injected with different ratios of TREK-1-L and TREK-2-S cRNA. Unitary currents and single channel conductances were calculated from patches containing one or two channels and plotted against the different RNA ratios (Fig. 4). The recorded channels clearly formed three distinct groups. Based on the single channel conductances in Fig. 3, the three groups correspond to TREK-1-L homodimers (circles), TREK-2-S homodimers (triangles), and TREK-1-L/TREK-2-S heterodimers (diamonds). Increasing the amount of TREK-2-S cRNA, while keeping the amount of total cRNA constant (0.025 ng of cRNA/oocyte), leads to an increased amount of heterodimer and TREK-2-S homodimer formation (see Fig. 4, compare the column marked 1:1 to the columns marked 1:4 and 1:12). Thus, the formation of TREK-1/TREK-2 heterodimers was confirmed at the single channel level.

RR Inhibits Single TREK-2-S and TREK-2-S/TREK-1-L
Channels-To relate characteristic channel conductances to the specific pharmacological profiles, TREK-1-L, TREK-2-S, and TREK-2-S/TREK-1-L tandem single channel currents were measured in the outside-out configuration. Channel activity (NP o ) was determined before, during, and after application of 30 M RR to the bath solution. The dye had a small inhibitory effect on TREK-1-L channel activity (Ϫ19 Ϯ 3%, n ϭ 4 patches) but efficiently inhibited TREK-2-S (Ϫ90 Ϯ 6%, n ϭ 7 patches, Fig. 5, A and B). In the case of TREK-2-S, a decrease was also observed in the single channel unitary current, but it was not statistically significant (p ϭ 0.066). The inhibition of the TREK-2-S/TREK-1-L tandem channel activity by RR (Ϫ55 Ϯ 11%, n ϭ 4 patches) was greater than the inhibition of TREK-1-L but less than the effect of the dye on TREK-2-S (Fig. 5C). The effect of RR was reversible. The different degree of inhibition between the homodimers and the tandem channel was statistically significant (p Ͻ 0.025). Thus, the effect of RR on TREK single channel currents proved to be consistent with that observed at the whole cell level.

Extracellular Acidification Inhibits Single TREK-1-L and TREK-2-S/TREK-1-L Channels and Activates TREK-2-S-
TREK-1-L, TREK-2-S, and TREK-2-S/TREK-1-L tandem channels were expressed in Xenopus oocytes. Single channel currents were then recorded from outside-out membrane patches. Channel activity was determined before, during, and after acidification (from pH 7.4 to 6.4) of the bath solution. The effect of the acidic extracellular pH on single TREK channels was similar to its effect on TREK channels observed at the whole cell level as seen on Fig. 1. Extracellular acidification had a quantitatively similar inhibitory effect on TREK-1-L (Ϫ50 Ϯ 8%, n ϭ 5 patches) and TREK-2-S/TREK-1-L tandem activity (Ϫ42 Ϯ 20%, n ϭ 4 patches, Fig. 6, A and C). TREK-2-S channel activity (Fig. 6B) was augmented by acidification (192 Ϯ 19%, n ϭ 5 patches). The effect of the pH change was reversible. Modification of the extracellular pH is a potential tool to separate homodimeric TREK-2 channels, where NP o is increased by EC acidification, from TREK-1 homodimers and heterodimers (where NP o is decreased by low pH).

Spadin Inhibits Single TREK-1 and TREK-2-S/TREK-1-L Channels-
The effect of spadin on TREK channels was measured on membrane patches expressing TREK-1-L, TREK-2-S, and TREK-2-S/TREK-1-L tandem channels in the outside-out configuration. Channel activity was determined before, during, and after the addition of spadin (1 M) to the EC solution. Spadin is a potent inhibitor of TREK-1 channel activity (NP o was inhibited by 60 Ϯ 11.2%, n ϭ 4 patches, Fig. 7A), although The difference between the conductance of all three groups was found to be statistically significant (p Ͻ 0.01). * represents significant differences. Currents were recorded from inside-out membrane patches as described in Fig. 3. Unitary currents and single channel conductances were calculated from patches containing one or two channels and plotted in the function of injected RNA ratios. For comparison, the averages (Ϯ S.D.) of TREK-1, TREK-2, and TREK-2-S/TREK-1-L tandem channel unitary currents were also plotted.
TREK-2 activity was not influenced by the application of the peptide (101 Ϯ 14.4% remaining current, n ϭ 4 patches, Fig.  7B). Interestingly, TREK-2-S/TREK-1-L tandem channel was also inhibited by the peptide (Fig. 7C, 85 Ϯ 12%, n ϭ 5 patches). The difference in the degree of inhibition between TREK-2-S homodimer and TREK-2-S/TREK-1-L was statistically significant (p Ͻ 0.005), although the inhibition of TREK-1-L and the tandem channel by spadin was not different (p ϭ 0.38, Fig.  7D). The inhibitory effect of the peptide was reversible on both TREK-1-L and TREK-2-S/TREK-1-L. These data indicate that the association of the TREK-1 subunit to the otherwise spadin-resistant TREK-2 confers susceptibility to the peptide, and the process of heteromerization may increase the sensitivity of the background K ϩ current to spadin in the cells that coexpress TREK-1 and TREK-2.
TREK-1 and TREK-2 Coimmunoprecipitate in Xenopus Oocytes-Wild type and tagged TREK channels were coexpressed in Xenopus oocytes in different combinations. Solubilized membrane preparation from oocytes expressing TREK-1-V5 ϩ TREK-2-FLAG and TREK-1 ϩ TREK-2-FLAG showed a similar signal on Western blot when probed with anti-FLAG antibody (Fig. 8A). The specific signal at 72 kDa (Fig. 8A, marked by an arrow) was missing in the sample from TREK-1-V5 ϩ TREK-2-injected oocytes (although nonspecific bands were apparent at higher and lower molecular weight regions). After immunoprecipitation of the solubilized samples by the V5 immunoaffinity resin, only the preparation from the TREK-1-V5 ϩ TREK-2-FLAG-expressing oocytes showed FLAG immunoreactivity indicating the assembly of the two tagged subunits (Fig. 8B). The blot shown is representative of three independent experiments.
TREK-1 and TREK-2 Form Functional Heterodimers in HEK293T Cells-In addition to Xenopus oocytes, heterodimerization of TREK-1 and TREK-2 subunits has also been demonstrated in a mammalian cell line, in HEK293T cells. For the verification of heterodimer formation, a similar pharmacological approach was used as in the case of the oocytes (illustrated in Fig. 2).
HEK293T cells expressing TREK-1, TREK-2, or the TREK-1-T2A-TREK construct (which leads to equimolar formation of TREK-1 and TREK-2 subunits) were used for the experiments. In contrast to the experiments in oocytes, the pH was only decreased from the physiological 7.4 by one pH unit to 6.4. After EC acidification, 30 M RR was added to the medium. The effects of these manipulations on the TREK current were plotted against each other (Fig. 9). The data point averages of the TREK-1 and TREK-2 homodimers were connected as in Fig. 2.
In good accordance with the literature data and our previous experiments, TREK-1 was inhibited by EC acidification (37.4 Ϯ 5.9% inhibition, n ϭ 5), although TREK-2 was activated by the decrease in EC pH (current increased by 44.3 Ϯ 8.1%, n ϭ 4). TREK-1 was RR-insensitive (1.1 Ϯ 7.1% activation, n ϭ 5), and Channel activity (NP o ) was determined from 30 to 60 s of recording before, during, and after application of RR. D, effect of 30 M RR on channel activity of TREK-1-L (white circles, n ϭ 4 patches), TREK-2-S (gray triangles, n ϭ 7 patches), and TREK-2-S/TREK-1-L (black diamonds, n ϭ 4 patches) channels is displayed as a scatterplot. The averages of the three groups are plotted as columns. The difference between the effect of RR on the channel activity of all three groups was found to be statistically significant (p Ͻ 0.05). * represents significant differences.
TREK-2 was potently inhibited by the dye (75.9 Ϯ 2.9% inhibition, n ϭ 4). The pH and RR sensitivity of the TREK current resulting from expression of the TREK-1-T2A-TREK-2 construct deviated from the line connecting the homodimers (in a matter similar to the oocytes in Fig. 2B). These data indicate that TREK-1 and TREK-2 subunits also form functional heterodimers when expressed in mammalian cells.
TREK-1 and TREK-2 Subunits Assemble into Heterodimeric Channels in DRG Neurons-Expression of both TREK-1 and TREK-2 subunits in the neurons of the dorsal root ganglia has been previously described by different groups (27)(28)(29). Based on these data, we decided to investigate whether TREK-1/ TREK-2 heterodimers are present in these neurons. In native tissues, alternative translation initiation cannot be ruled out, leading to different variants of TREK-1 and TREK-2 subunits. Thus, distinct unitary conductance levels may correspond to the heterodimers of different subunit composition, making the identification of heterodimer formation difficult on the basis of conductance. Therefore, we applied a pharmacological approach in single channel measurements.
Based on our results on outside-out membrane patches from Xenopus oocytes, sensitivity to spadin can be used to recognize a potassium channel as a TREK-1 homodimer or a TREK-1/ TREK-2 heterodimer. After the recovery of the current from the inhibition by spadin, RR can be used for the identification of the channel. If the channel is RR-sensitive, then the channel is a TREK-1/TREK-2 heterodimer, if not, then it is a TREK-1 homodimer.
Outside-out membrane patches were excised from DRG neurons. Potassium channel activity was detected in 16 out of 25 membrane patches. Only patches containing 1 or 2 channels were analyzed. The effects of spadin (1 M) and RR (30 M) on channel activity were measured and plotted against each other (see Fig 10A). In 16 membrane patches, 21 distinct potassium channels were identified. Four channels were found to be resistant to both RR and spadin. Eight channels were spadin-resistant but potently inhibited by RR (for representative recording see Fig. 10B). Nine channels were sensitive to both spadin and RR (Fig. 10C). This inhibitory profile indicates that these potassium channels found in DRG neurons are formed by the heterodimeric assembly of TREK-1 and TREK-2 subunits. Heterodimeric assembly was thus confirmed in a native tissue.

Discussion
TREK channels are known to participate in diverse physiological processes, such as nociception and thermoreception (30), relaxation of visceral and vascular smooth muscle cells (31), and regulation of hormone secretion (32). They were also reported to have potential therapeutic importance in different disease states, e.g. clinical depression (33) and prevention of post-ischemic brain damage (34). In this report, using a pharmacological approach combined with measurements at the single channel level, we demonstrate heteromerization of TREK-1 and TREK-2 subunits in different expression systems (Xenopus oocytes, HEK293T cells) and also in native cells (in DRG neurons).
We have previously obtained evidence that functional channels are formed when two identical (TREK-2/TREK-2) subunits are concatenated. The current of this tandem channel was indistinguishable from the naturally assembled TREK-2 homodimers. In this study we have experimentally confirmed, by mutating a pore domain in only one of the subunits, that the  Xenopus oocytes were coinjected with cRNA of TREK-1 and TREK-2 channel constructs, in different combinations of wild type or epitope-tagged forms, as indicated. A, total membrane lysates were prepared, solubilized, and subjected to SDS-PAGE. Size-fractionated proteins were transferred to nitrocellulose membrane; the blots were probed with monoclonal anti-FLAG antibody, and visualized by HRPconjugated anti-mouse antibody. The arrow marks the specific TREK-2-FLAG immunoreactivity. B, coimmunoprecipitation was performed on solubilized membrane preparations using beads to which polyclonal anti-V5 antibody was bound. The affinity-purified proteins were analyzed as described at A. The blot shown is representative of three independent experiments. assembly of the joined subunits is highly preferred over the intermolecular random dimerization. To characterize the functionality and properties of the TREK-1/TREK-2 channel, we created a model, i.e. a concatenated construct in which the two different subunits were covalently linked together and thereby forced to heterodimerize. The tandem TREK-1/TREK-2 turned out to be functional; its acid sensitivity was similar to that of the TREK-1 homodimer, and it was moderately inhibited by RR. The reduced RR sensitivity is in good accordance with our previous report, in which mutation of the RR interacting amino acid (Asp-135) in one of the TREK-2 subunits converted the TREK-2/TREK-2 tandem to be moderately RR-sensitive (20). Based on this asymmetrically inherited pharmacological profile, the tandem channel could be distinguished from the mixture of homodimers. Coexpression of TREK-1 and TREK-2 subunits leads to a current that has a different pH and RR sensitivity than expected from a linear combination of TREK-1 and TREK-2 currents. This can only be explained by spontaneous formation of functional TREK-1/TREK-2 heterodimers from the individual subunits. Interaction of the two subunits was confirmed by coimmunoprecipitation of tagged TREK-1 and TREK-2 subunits in Xenopus oocytes. Heterodimerization also occurred in mammalian cells (HEK293T), ruling out the possibility that TREK-1/TREK-2 only heterodimerize in the amphibian system.
TREK-1 and TREK-2 channels have different single channel conductances. Differentiating the two channels on the single channel level is complicated by the fact that both channels have multiple isoforms due to alternative translation initiation, which possess different conductance levels (24,26). To advance our study of the heteromerization of TREK-1 and TREK-2 subunits, we eliminated the multiple conductance levels arising from alternative translation variants by creating mutant versions of the channels (TREK-1-L, TREK-2-S, and TREK-2-S/ TREK-1-L tandem) with a single translation initiation site. As expected, the mutant channels had a pH and RR sensitivity similar to their wild type counterparts and also formed functional heterodimers.
When expressed in Xenopus oocytes, each mutant channel (TREK-1-L and TREK-2-S) had one distinct single channel conductance level in excised inside-out membrane patches. The tandem construct had an intermediate conductance level compared with the two homodimers; thus, the three mutants could be distinguished on the basis of their single channel conductance. In oocytes coexpressing TREK-1-L and TREK-2-S subunits, we recorded single potassium channels with conductances corresponding to the TREK-1-L homodimer, the TREK-2-S homodimer, and channels with an intermediate conductance similar to the tandem channel. The existence of this intermediate conductance channel can be explained by the formation of functional TREK-1-L/TREK-2-S heterodimers, thus demonstrating the heteromerization of TREK-1 and TREK-2 subunits also on the single channel level.
The effect of RR, EC acidification, and the sortilin-derived neuropeptide spadin on single TREK channels was examined in outside-out membrane patches excised from Xenopus oocytes. EC acidification influenced single channel activity similarly to its effect on the macroscopic TREK currents. Acidification inhibited the activity of TREK-1 and tandem channels and augmented TREK-2 channel activity by decreasing or increasing the open probability of the channels, respectively. The degree of inhibition of channel activity by RR was also similar to the effect of RR on macroscopic TREK currents. TREK-1 was RR-insensitive, and TREK-2 was potently inhibited by the dye. The tandem channel was moderately inhibited by RR. The decrease of channel activity by RR was due to a decrease in the open probability of TREK-2 and tandem channels. Application of RR also leads to a moderate, statistically not significant decrease of the unitary TREK-2 current.
In good accordance with the previous reports, TREK-1 was potently inhibited by spadin, although TREK-2 was not affected. The peptide inhibited the tandem channel, the degree of the effect was similar to that observed in the case of the TREK-1 homodimer. Inhibition of TREK-1 and tandem channel activity was the consequence of the decreased open probability. Spadin has been described as a selective TREK-1 inhibitor (TREK-2, TRAAK, TRESK, and TASK-1 have all been described to be insensitive to spadin). We have demonstrated that the TREK-1/TREK-2 heterodimer is also a target of the peptide. The spadin-interacting site has not been identified, but our results indicate that one TREK-1 subunit is sufficient to confer sensitivity to the peptide. Accordingly, spadin is a promising candidate for testing or confirming experimental results in which TREK-1 is suspected as a heteromeric partner (15).
The TASK-1/TASK-3 heterodimer has been identified in native cells based on its different inhibitory profile from the homomeric TASK-1 and TASK-3 channels (7-9). Based on our results, the TREK-1/TREK-2 heterodimer can be differentiated from other K 2P channels based on its unique inhibitory profile. Application of spadin can be used to identify the current of TREK-1 or TREK-1/TREK-2 heterodimer. The spadin-sensitive current can be further differentiated by RR; the insensitive component reflects the TREK-1 homodimer, and the spadinsensitive current with intermediate RR sensitivity can be identified as TREK-1/TREK-2.
TREK-1 and TREK-2 subunits are widely expressed in both the central nervous system and in peripheral tissues. Overlapping expression of the two subunits has been reported in the striatum and in the hippocampus (35). Native channels with properties corresponding to TREK-1 and TREK-2 and a TREKlike channel (with properties different from TREK-1, TREK-2, and TRAAK) have been described in magnocellular neurosecretory cells of the supraoptic nucleus (36). The expression of both subunits has also been detected in the peripheral (27)(28)(29) superior cervical ganglia (37). Coexpression of TREK-1 and TREK-2 subunits has also been described in non-neuronal cells, for example in pulmonary vascular smooth muscle cells (31).
Heterodimerization of TREK-1 and TREK-2 subunits has also been examined in neurons of dorsal root ganglia. We excised outside-out membrane patches from cultured DRG neurons and examined the effect of spadin and RR on the channel activity. We could group the channels into three distinct groups as follows: spadin-and RR-resistant channels, spadinresistant and RR-sensitive channels, and a spadin-and RR-sensitive group. The spadin-and RR-insensitive currents may correspond to TRESK or TASK-1 or other K ϩ channels. The spadin-resistant and RR-sensitive ones may correspond to TREK-2, TRAAK, or TASK-3. Exact identification of these channels, however, was beyond the scope of this study. The channels of the third group were inhibited by both spadin and RR. Sensitivity to the specific peptide inhibitor spadin suggests that these channels contained the TREK-1 subunit. Sensitivity of the same channel to RR clearly indicated the presence of another RR-sensitive subunit in the channel complex, i.e. the formation of the heterodimer.
Changes in EC pH regulate the membrane potential and excitability of pulmonary vascular smooth muscle cells and somatosensory neurons. These cells express both TREK-1 and TREK-2 subunits. The TREK-1/TREK-2 heterodimer and the TREK-1 homodimer have a similar sensitivity to EC acidi-fication, but the heterodimeric channel has a larger conductance than the TREK-1 homodimer. Assembly of identical number of heterodimers instead of TREK-1 homodimers may lead to a larger background potassium current. A change in the ratio of the heterodimers compared with TREK-1 homodimers can regulate the amplitude of the background potassium conductance, thus controlling the excitability of the cell without changing the overall sensitivity of the current to EC pH. The TREK-1/TREK-2 heterodimer can thus contribute to the control of DRG neuron excitability.
Spadin is a by-product of the maturation of the sortilin receptor. In mouse models of depression, spadin had an effect similar to fluoxetine, an antidepressant used in the treatment of clinical depression. The antidepressant effects of spadin were apparent after 5-6 days of application, although the effect of conventional antidepressants only manifest after 2-3 weeks of treatment. Therefore, spadin has been considered as an endogenous antidepressant (18). According to the present results, the peptide might well be an endogenous regulator of TREK-1/TREK-2 heterodimer activity. In neurons, inhibition of the TREK-1/ TREK-2 heterodimer by spadin can lead to membrane depolarization and increased transmitter release. We have demonstrated that TREK-1 and TREK-2 heterodimerize in neurons of the peripheral nervous system; however, heterodimer formation may also happen in the central nervous system. In this case inhibition of the heterodimeric channel may contribute to the antidepressant effect of spadin.
Author Contributions-M. L. designed and performed the experiments and analyzed data and wrote the paper. G. C. and P. E. designed the experiments, analyzed the data, and wrote the paper.