TRAAK Is a Mammalian Neuronal Mechano-gated K+Channel*

The novel structural class of mammalian channels with four transmembrane segments and two pore regions comprise background K+ channels (TWIK-1, TREK-1, TRAAK, TASK, and TASK-2) with unique physiological functions (1-6). Unlike its counterparts, TRAAK is only expressed in neuronal tissues, including brain, spinal cord, and retina (1). This report shows that TRAAK, which was known to be activated by arachidonic acid (3), is also opened by membrane stretch. Mechanical activation of TRAAK is induced by a convex curvature of the plasma membrane and can be mimicked by the amphipathic membrane crenator trinitrophenol. Cytoskeletal elements are negative tonic regulators of TRAAK. Membrane depolarization and membrane crenation synergize with stretch-induced channel opening. Finally, TRAAK is reversibly blocked by micromolar concentrations of gadolinium, a well known blocker of stretch-activated channels. Mechanical activation of TRAAK in the central nervous system may play an important role during growth cone motility and neurite elongation.

A key feature of all classes of K ϩ channel pore-forming subunits is a conserved signature sequence constituting the pore segment (P) 1 (7). Mammalian K ϩ channels can be divided into three major structural classes encoding channels with six transmembrane segments (TMS), 4TMS and 2TMS (7). The 6TMS class comprises voltage-gated as well as Ca 2ϩ -gated K ϩ channels. The inward rectifier IRK, the G protein-coupled GIRK and the ATP-gated K ATP K ϩ channel subunits are members of the 2TMS structural class. The most recent structural class of 4TMS mammalian K ϩ channel subunits consists so far of five members (TWIK-1, TREK-1, TRAAK, TASK, and TASK-2) (1)(2)(3)(4)6). Besides the presence of 4TMS, the other major structural characteristic is the presence of two P regions as well as an extended M1P1 extracellular loop (8). Although these subunits display the same structural motif (4TMS/2P), they only share 25-40% sequence identity. These unusual K ϩ channel structures are associated with unique physiological properties. TWIK-1 is an ubiquitous subunit, which directs the expression of a time-and voltage-independent K ϩ -selective weak inward rectifier in Xenopus oocytes (4,8). TREK-1 is highly expressed in brain, heart, lung, kidney, and several other tissues and encodes a mammalian serotonin-sensitivelike K ϩ channel (3,5). TASK and TASK-2 encode background K ϩ channels that are blocked by external acidification near the physiological pH (2,6). TASK is highly expressed in heart and brain, while TASK-2 is mainly expressed in kidney (2,6).
TRAAK is the only 4TMS subunit to be specifically expressed in the nervous system (1). TRAAK is opened by polyunsaturated fatty acids, including arachidonic acid (AA) (1). The present report demonstrates that TRAAK is also opened by membrane stretch and thus belongs to the class of mechano-gated ion channels (9 -16).

MATERIALS AND METHODS
Cell Culture-COS-7 cells were maintained in Dulbecco's modified Eagle's medium (Life Technologies, Inc.) supplemented with 10% fetal bovine serum. The cDNA for TRAAK was subcloned into the pIRES-Neo expression vector (CLONTECH), and the resulting construct was transfected in COS cells using the phosphate calcium method. After 48 h in normal medium, the transfected cells were dissociated and subcultured by 1:20 dilution in medium containing 500 g/ml Geneticin (Life Technologies, Inc.). Antibiotic-resistant clones were selected at random after 2 weeks. A TRAAK-expressing clone was then identified by using an 86 Rb ϩ efflux assay as well as electrophysiology and expanded to maintain a stock culture.
Western Blots and Immunocytology-Anti-TRAAK antibodies were raised against a glutathione S-transferase fusion protein containing the carboxyl-terminal 102 amino acids of TRAAK (residues Pro 296 to Val 398 ). The antibodies were purified by using a glutathione S-transferase fusion protein containing the same domain of TRAAK. The preparation of fusion proteins, rabbit immunization, and antibody purification were performed as described previously (8).
Proteins from COS cells and from synaptic plasma membranes of mouse spinal cord were prepared and analyzed in the presence of reducing agents (8). For immunocytology, TRAAK immunodetection was performed as described previously (8), except that cells were permeabilized by adding 0.1% Triton X-100 in the blocking solution (phosphate-buffered saline supplemented with 2% BSA and 5% goat serum).
Efflux of 86 Rb ϩ -COS-7-TRAAK cells were plated at a density of 40,000 cells/well (Falcon, 24 wells) and used 2-3 days later. Cells were preloaded for 3 h with 0.5-1 Ci/ml 86 Rb ϩ in 0.5 ml of Dulbecco's modified Eagle's medium, 10% fetal bovine serum, and 500 g/ml Geneticin at 37°C. Release experiments were then carried out in a standard (EXT) saline solution containing (in mM: 150 NaCl, 5 KCl, 3 MgCl 2 , 1 CaCl 2 , 10 Hepes, pH 7.4, with NaOH) during consecutive intervals of 5 min at room temperature. For each point six independent experiments were performed at the same time. An hour washing with EXT solution was then performed before collecting 1-ml fractions. 86 Rb ϩ at 2-7 mCi/mg was from NEN Life Science Products. Fractional rates of release were calculated as 86 Rb ϩ released during each 5-min interval and expressed as the percentage of 86 Rb ϩ content in the cells at the beginning of the respective intervals. 86 Rb ϩ was counted on a Packard Tri-Carb with a Cerenkov program.
Patch Clamp Experimental Protocols, Recordings and Data Analysis-The electrophysiology procedure has been described elsewhere previously (5). INT. Mechanical stimulation was applied through an open-loop pressure generating system and monitored at the level of the patch pipette throughout the experiment by a calibrated pressure sensor. This system provides a stable pressure pulse (5). Colchicine was dissolved daily in the saline solution at the concentration of 500 M. Cytochalasin D was dissolved at the concentration of 1 mg/ml in Me 2 SO and kept at Ϫ20°C. AA was dissolved in ethanol at the concentration of 100 mM, flushed with argon, and kept at Ϫ20°C for a week. TNP was mixed with the saline solutions and pH adjusted. 4-Bromophenacyl bromide was dissolved daily at 100 mM in Me 2 SO and AACOCF 3 at 100 mM in ethanol. All chemicals were obtained from Sigma.

RESULTS
TRAAK was transfected in COS cells, and a stable TRAAK cell line was characterized by immunoblot using affinity-purified antibodies directed against the carboxyl terminus of TRAAK ( Fig. 1, A-C). A major band is detected with a relative molecular mass (M r ) of 54,000 -64,000 and two minor bands of M r 49,000 and 52,000 (Fig. 1A, lane 2). No signal is detected in nontransfected COS cells (Fig. 1A, lane 1). The difference between the observed M r values and the calculated molecular mass of TRAAK (43,000 Da) is probably due to glycosylation. This hypothesis is supported by the presence of two consensus sites for N-linked glycosylation (residues 81 and 84) in a region of the channel that is expected to be extracellular (1). This probable glycosylation of TRAAK is also observed in the spinal cord (Fig. 1A, lane 3), a tissue that is known to express a high level of TRAAK transcript (1). Additional bands of M r 49,000 and 52,000, which are detected in transfected COS cells, could correspond to incompletely glycosylated or partially degraded forms of the protein. The heterologous expression of TRAAK was confirmed by immunocytology ( Fig. 1, B and C). A strong signal is observed at the plasma membrane of the cells expressing TRAAK (Fig. 1C), while it is absent in control WT COS cells (Fig. 1B).
The expression of exogenous K ϩ channel activity was initially detected in the TRAAK cell line using 86 Rb ϩ efflux ( Fig.  1, D-F). Control COS cells (WT) do not show any significant increase in 86 Rb ϩ efflux in the presence of AA (Fig. 1D). However, cells expressing TRAAK display a reversible 3-4-fold increase in 86 Rb ϩ efflux induced by AA (Fig. 1, D and E). AA activation is reversed by the addition of micromolar concentrations of Gd 3ϩ (IC 50 : 5 Ϯ 1 M) (Fig. 1E).
The TRAAK cell line was then used for electrophysiological investigations. In the whole cell configuration, basal channel activity is low (TRAAK: 12.2 Ϯ 1.3 pA/picofarad, n ϭ 12; COS WT: 4.2 Ϯ 0.5 pA/picofarad, n ϭ 18; at 100 mV) ( Fig. 2A). An outward current is slowly and reversibly induced at a holding potential of 0 mV by AA superfusion (4.8 Ϯ 0.4-fold, n ϭ 12) ( Fig. 2A, bottom inset). This current is absent in mock-transfected COS cells (n ϭ 18) (5). The I-V curve of the current induced by AA displays a strong outward-going rectification ( Fig. 2A). The reversal potential of the current activated by AA reverses at Ϫ83.7 Ϯ 0.7 mV (n ϭ 12), which is the predicted value for E K ϩ. Opening of TRAAK is also reversibly induced by the amphipathic membrane crenator TNP (6.7 Ϯ 1.1-fold, n ϭ 6 with 400 M TNP) (15,17). The kinetics for current activation is faster in the presence of TNP compared with AA ( Fig. 2A,  insets). I-V curves in the presence of TNP and AA are similar. Finally, as observed with 86 Rb ϩ efflux experiments, TRAAK activation by AA is also blocked by Gd 3ϩ (Ϫ64.2 Ϯ 4.7%, n ϭ 8 with 10 M Gd 3ϩ ) (Fig. 2B).
In the cell-attached patch configuration, no channel activity is detected under resting conditions (NPo: 0.05 Ϯ 0.03, n ϭ 6). However, we observed that TRAAK opens when a negative pressure is applied to the patch pipette (Fig. 3A). No channel activity is observed in mock-transfected COS cells (5). In the experiment illustrated in Fig. 3, A and B, channel activity is absent at atmospheric pressure but three TRAAK 38-picosiemens channels readily and reversibly open during the application of a Ϫ66 mm Hg pressure. As reported previously for AA activation, TRAAK opening was characterized by flickering kinetics (1).
Both the number of active channels and the sensitivity to mechanical stretch are strongly enhanced after a treatment for 20 min with colchicine (500 M) (Fig. 3, C-E). The threshold for mechanical activation is lowered from Ϫ70 to Ϫ20 mm Hg, and maximal channel activity is enhanced at all indicated pressures in the presence of colchicine (Fig. 3, C and D). Similarly, addition of cytochalasin D (5 g/ml) enhances channel activity with, however, a weaker effect (Fig. 3E). Finally, excision of the patch in the inside-out configuration, which is known to disrupt cytoskeletal elements, produces an almost 10-fold increase in channel activation induced by a Ϫ70 mm Hg pressure (Fig. 3E). Moreover, the threshold for mechanical activation is also significantly lowered after excision of the patch (Fig. 4A). In the inside-out patch configuration, channel activity is basically absent at atmospheric pressure, and only negative, but not positive, pressure induces channel opening (Fig. 4B). The I-V curve performed with a voltage ramp protocol in physiological K ϩ conditions is outwardly rectifying and reverses at Ϫ80 mV, the predicted value for the equilibrium potential for K ϩ (Fig. 4B).
The absence of channel activation by positive pressure in the inside-out patch configuration suggests that channel opening is mediated by a specific membrane deformation. Indeed, in the outside-out patch configuration, TRAAK opening is only in-duced by positive pressure (Fig. 5A). The activation of TRAAK is purely pressure-dependent, and moreover, as observed in both the cell-attached and inside-out configurations, channel activity is absent at atmospheric pressure. Opening of TRAAK by positive pressure in outside-out patches is reversibly blocked by the addition of 30 M Gd 3ϩ (n ϭ 3) (Fig. 5B).
We then investigated the effect of voltage on TRAAK activa- tion in the inside-out patch configuration. Fig. 6 shows that opening of TRAAK by membrane stretch is enhanced by membrane depolarization. The threshold for mechanical activation by negative pressure is lowered at depolarized voltage, and moreover maximal NPo is significantly enhanced (Fig. 6C). The basal NPo measured at atmospheric pressure is also signifi-cantly increased at depolarized potential (Fig. 6, A-C). The steep modulation of TRAAK activation occurs at voltages between Ϫ25 and 50 mV.
We investigated the possible interaction between the membrane crenator TNP and membrane stretch. Fig. 7, A and B, show that in the inside-out configuration, internal application of TNP has no effect at atmospheric pressure. However, TNP produces a significant reversible stimulation of the opening of TRAAK by negative pressure (Fig. 7, A and B). Interestingly, in the outside-out configuration, external TNP produces a significant activation of TRAAK at both atmospheric and positive pressures (Fig. 7, C and D).
Finally, to assay for the possible involvement of AA in the mechano-activation of TRAAK, we investigated the effect of 30 M lipid-free BSA, which is known to bind fatty acids and remove them from membranes (18). Application of BSA on both sides of the membrane did not alter the activation of TRAAK by a Ϫ50 mm Hg pressure in the inside-out patch configuration (n ϭ 6; data not shown). Moreover, the addition of the phospholipase A 2 blockers 4-bromophenacyl bromide (30 M) and arachidonyltrifluoromethyl ketone (AACOCF 3 ) (30 M) to the cytosolic face of inside-out patches for 5 min did not alter activation of TRAAK by a Ϫ50 mm Hg pressure (n ϭ 5; data not shown). DISCUSSION TRAAK is a member of the novel K ϩ channel family with two P domains and four transmembrane segments (1). Despite a similar membrane topology, the 2P/4TMS channels display little sequence identity (about 25%). When compared with TWIK-1 and TREK-1, TRAAK has a shorter amino-terminal region but an extended C terminus (3,4). TRAAK has a large extracellular loop between M1 and P1, with a cysteine residue at position Cys 52 analogous to the cysteine residue Cys 69 involved in the disulfide-bridged homodimerization of TWIK-1 (8). TRAAK is only expressed in brain, retina, and spinal cord (1). The most intense levels of expression are present in the olfactory system, cerebral cortex, hippocampal formation, habenula, basal ganglia, and cerebellum. This laboratory demonstrated that TRAAK is opened by AA (1). The activation of TRAAK by AA is reversible, concentration-dependent, and direct (i.e. not via protein kinase C). Other polyunsaturated fatty acids also open TRAAK, while saturated fatty acids are without effect (1). TRAAK current is instantaneous and outwardly rectifying in a physiological K ϩ gradient. Finally, TRAAK is only inhibited by high concentrations of Ba 2ϩ and is insensitive to the other classical K ϩ channel blockers, including tetraethylammonium, 4-aminopyridine, and Cs ϩ (1).
The anionic amphipath TNP has been shown previously to expand the exterior half of the lipid bilayer and thereby induce human erythrocytes to crenate (17). Moreover, TNP was shown to activate bacterial mechano-gated cationic channels, and this effect was attributed to membrane crenation (15). In the present study, we demonstrate that TNP is a strong activator of TRAAK. TRAAK opening by TNP is observed in the whole cell as well as in the excised outside-out patch configurations, suggesting a rather direct mechanism (i.e. without second messenger). Interestingly, no significant effect of TNP (at the concentration used) is observed under resting conditions in the insideout patch configuration. This observation suggests that TNP may act preferentially via the external side of the membrane. It has been hypothesized previously that because of the negative charges (mainly phosphatidylserine) of the inner leaflet, anionic amphipaths may preferentially insert in the external leaflet of the plasma membrane (15,17). The differential insertion of these amphipaths in the external lipid monolayer induces a curvature of the membrane which generates transversal forces that may alter channel activity. This model implies that TRAAK may be sensitive to mechanical forces transmitted via the membrane.
In every configuration tested, basal channel activity measured at atmospheric pressure was basically absent. In the inside-out patch configuration, TRAAK is opened by negative pressure, while it is opened by positive pressure in the outsideout patch configuration. These results suggest that TRAAK is selectively activated by a convex curvature of the membrane. This result fits well with the bilayer couple model of the membrane described above (15,17). Channel activity was graded with applied pressure and was maintained throughout stimulation, demonstrating the absence of an adaptation process. TRAAK can be opened by both arachidonic acid as well as membrane stretch. A mechanically sensitive phospholipase A 2 (19) could therefore couple membrane stretch to TRAAK opening as suggested for stretch-activated K ϩ channels in gastric smooth muscle (18). Although the addition of lipid-free BSA as well as the phospholipase A 2 (PLA 2 ) inhibitors do not alter the activation of TRAAK by mechanical stimulation, we cannot entirely rule out such a mechanism.
Disruption of the cytoskeleton by either chemical or mechanical means (colchicine, cytochalasin D, or membrane excision) potentiates TRAAK opening by membrane stretch. These results suggest that mechano-gating does not require the integrity of the cytoskeleton. Moreover, these results imply that the activating force is coming directly from the bilayer membrane. Finally, the up-regulation of channel activity suggests that TRAAK is tonically repressed by the cytoskeleton. Dynamic regulation of cytoskeletal components through, for instance, phosphorylation/dephosphorylation is thus expected to have an important effect on the indirect regulation of TRAAK channel activity.
Opening of TRAAK is clearly facilitated at depolarized potentials (between Ϫ25 and 50 mV). Depolarization lowers the threshold for mechanical activation and increases the maximal channel activity (NPo). The 2P/4TMS channels lack the positively charged S4 voltage sensor found in Shaker-type K ϩ channels (7). Thus the region responsible for the voltage-dependent modulation of TRAAK remains to be determined. An important physiological implication of these findings is that activation of TRAAK (induced mechanically or chemically) will be mostly efficient at positive potentials (i.e. during an action potential). The inside-and outside-out experiments indicate that TNP synergize with pressure activation. This result suggests that physiologically, mechanical activation of TRAAK may be most effective when endogenous amphipathic crenators are produced.
In conclusion, this report demonstrates that the mammalian neuronal 2P/4TMS K ϩ channel TRAAK belongs to the mechano-gated ion channel family (9 -16, 18, 20 -23). Stretchactivated K ϩ channels have been described in several neuronal cell types, including snail and Aplysia neurons as well as in rat hippocampal neurons (20 -24). The presence of stretch-activated K ϩ channels in the growth cones of Lymnaea stagnalis neurons suggested that these channels may be involved in the modulation of axonal pathfinding and guidance (22). Indeed, growth cone motility is a mechanical process where neurite membranes experience important tension changes. Activation of TRAAK during cell locomotion may participate via membrane hyperpolarization to the fine regulation of intracellular calcium, actin-myosin contractions and thus neurite elongation. Finally, the expression of TRAAK in dorsal root ganglia 2 may suggest its possible role in more specialized sensory functions such as touch detection and/or pain sensation. We previously identified a charged region in the COOH terminus of TREK-1 (Arg 297 -Glu 306 : RVISKKTKEE), which is critically involved in channel activation by AA as well as mechanical stimulation. This region is conserved in TRAAK (Arg 259 -Glu 268 : RAVSRRTRAE) and may also fulfill an important function for TRAAK activation.