Lysophospholipids open the two-pore domain mechano-gated K(+) channels TREK-1 and TRAAK.

The two-pore (2P) domain K(+) channels TREK-1 and TRAAK are opened by membrane stretch as well as arachidonic acid (AA) (Patel, A. J., Honoré, E., Maingret, F., Lesage, F., Fink, M., Duprat, F., and Lazdunski, M. (1998) EMBO J. 17, 4283-4290; Maingret, F., Patel, A. J., Lesage, F., Lazdunski, M., and Honoré, E. (1999) J. Biol. Chem. 274, 26691-26696; Maingret, F., Fosset, M., Lesage, F., Lazdunski, M. , and Honoré, E. (1999) J. Biol. Chem. 274, 1381-1387. We demonstrate that lysophospholipids (LPs) and platelet-activating factor also produce large specific and reversible activations of TREK-1 and TRAAK. LPs activation is a function of the size of the polar head and length of the acyl chain but is independent of the charge of the molecule. Bath application of lysophosphatidylcholine (LPC) immediately opens TREK-1 and TRAAK in the cell-attached patch configuration. In excised patches, LPC activation is lost, whereas AA still produces maximal opening. The carboxyl-terminal region of TREK-1, but not the amino terminus and the extracellular loop M1P1, is critically required for LPC activation. LPC activation is indirect and may possibly involve a cytosolic factor, whereas AA directly interacts with either the channel proteins or the bilayer and mimics stretch. Opening of TREK-1 and TRAAK by fatty acids and LPs may be an important switch in the regulation of synaptic function and may also play a protective role during ischemia and inflammation.

The recently discovered family of mammalian two pore (2P) 1 domain K ϩ channels consists so far of seven members (TWIK-1, TWIK-2, TREK-1, TRAAK, TASK-1, TASK-2, and KCNK6) (4 -12). Although these subunits share the same structural motif with four transmembrane segments, 2P domains, an extended M1P1 external loop (60 -70 residues), and both amino and carboxyl termini intracellularly, they only share 25-50% identity. TWIK-1 and TWIK-2 encode weak inward rectifiers that are inhibited by internal acidosis but stimulated by protein kinase C activation (7,11). The second pore region P2 of the nonexpressing mouse KCNK6 is characterized by a GLG motif identical to that found in TWIK channels and different from the GFG sequence found in other 2P domain K ϩ channels (8). TASK-1 and TASK-2 encode background outward rectifiers that are constitutively active at all voltages and inhibited by mild external acidosis near the physiological pH (5,9,10,12). Finally, TREK-1 and TRAAK are outward rectifier K ϩ channels opened by membrane stretch, cell swelling, and shear stress (1)(2)(3). At atmospheric pressure, basal activity is negligible, and channels are opened by convex curvature of the plasma membrane (3). Mechano-gating does not require the integrity of the cytoskeleton, and the activating force is apparently directly coming from the bilayer (3). Mechanical activation of TREK-1 and TRAAK is mimicked by polyunsaturated fatty acids such as arachidonic acid (AA) and by the anionic amphipath trinitrophenol (1,3,4). Recently, we showed that TREK-1, but not TRAAK, is also opened by inhalational anesthetics including chloroform, ether, halothane, and isoflurane as well as by mild intracellular acidosis (2,13). In mouse tissues, TREK-1 is ubiquitous, with strong expression in brain and heart, whereas mouse TRAAK is restricted to the central nervous system, the spinal cord, and the retina (4,6). In human tissues, TREK-1 is at low level or even absent from heart, whereas TRAAK is also detected in the placenta, testis, prostate, and small intestine (13). In the present report, we demonstrate that extracellular LPs as well as PAF mimic the effect of stretch and are potent openers of the neuronal 2P domain K ϩ channels TREK-1 and TRAAK.
Cell Culture-COS-7 cells were maintained in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum. The 2P domain K ϩ channel cDNAs were subcloned into the pIRES-CD8 vector and transfected using the DEAE-dextran procedure. Cells were visualized 48 h after transfection using the anti-CD8 antibody-coated bead method.
Solution-For whole-cell experiments, bath solution (EXT) contained 150 mM NaCl, 5 mM KCl, 3 mM MgCl 2 , 1 mM CaCl 2 , 10 mM Hepes, pH 7.4 with NaOH, and pipette solution (INT) contained 150 mM KCl, 3 mM MgCl 2 , 5 mM EGTA and 10 mM Hepes, pH 7.2 with KOH. For cell attached and inside-out experiments, the bath solution was INT, and the pipette contained EXT solution (5 mM KCl). All chemicals were obtained from Sigma except ginkolide B, trans-1,3-dioxolane, lyso-PAF, methyl arachidonyl fluorophosphonate, 7,7-dimethyleicosadienoic acid, and RHC-80267 were obtained from Biomol. Stock solutions were kept at Ϫ20°C and renewed weekly. AA and arachidonyl trifluoromethyl ketone were dissolved in ethanol at the concentration of 100 mM. Saturated fatty acids, PAF, lyso-PAF, and PAF receptor antagonists ginkolide B and trans-1,3-dioxolane were dissolved at the concentration of 10 mM in ethanol. LPCs, LPA, LPI, choline, and Gd 3ϩ were dissolved in water at the concentration of 10 mM. LPS and LPE were either dissolved at the concentration of 10 mM in 75% methanol or in a mixture of 70% chloroform, 27% methanol, 3% water. PC was dissolved at a concentration of 30 mM in ethanol. Chlorpromazine was dissolved at the concentration of 10 mM in 10% ethanol. Amiloride was dissolved at the concentration of 1 mM in saline. Phospholipase A 2 antagonists 7,7-* This work was supported by CNRS, the Association Française contre les Myopathies (AFM), and the Conseil Régional PACA. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
‡ To whom correspondence should be addressed: Mutations-Polymerase chain reaction was used to generate aminoterminal deletions in TREK-1 by introducing a methionine just before Val-47. A carboxyl terminus deletion was generated by introducing a stop codon at Gly-308. The external loop M1P1 of mouse TREK-1 (amino acids 68 -135) was substituted with the loop of human TASK-1 (amino acids 30 -85).
Data Analysis-For whole-cell recordings, values of control currents were subtracted from the currents recorded in the presence of the various lipids. Results are presented as means of density current associated with their standard errors (S.E.). Cell capacitance was determined by hyperpolarization pulses of 10 mV from a holding potential of Ϫ80 mV. The number of experiments are indicated on each graph.

RESULTS
The biophysical and pharmacological properties of TREK-1 were investigated in transiently transfected COS cells. Cells expressing TREK-1 were voltage-clamped at 0 mV in the whole-cell patch clamp configuration, and lipids were applied in the bath as illustrated in Fig. 1A. As previously reported, TREK-1 was activated by polyunsaturated fatty acids including AA (10 M) (Fig. 1A) (1). Similarly, 3 M LPC (C14:0) opened TREK-1 and induced a strong outward current (Fig.  1A). Onset and offset kinetics for LPC activation were fast compared with AA (Fig. 1B). In physiological K ϩ conditions, the LPC-induced current displayed an outward-going rectification and reversed at Ϫ81 Ϯ 3 mV (n ϭ 6) (Fig. 1C). In symmetrical K ϩ conditions, the reversal potential was shifted to 0 mV, as expected for a K ϩ -selective channel, and rectification was maintained (Fig. 1D). Activation of TREK-1 was dose-dependent, with a threshold concentration of 30 nM (Fig. 1E). Concentrations higher than 10 M could not be tested since they produce irreversible cell leakage. Current stimulation was sta-ble during continuous application of LPC up to 15 min (not shown). 10 M LPC had no significant effect on mock-transfected or TASK-1-or TASK-2-expressing cells (Fig. 1F).
The cationic cup former membrane modifying agent chlorpromazine completely and reversibly reversed LPC activation of TREK-1 (1 M; n ϭ 5) (Fig. 5A). Interestingly, chlorpromazine had no effect on TRAAK (n ϭ 6) (not shown). Blockers of stretch-sensitive channels such as amiloride (2 mM) and Gd 3ϩ (10 M) (14) similarly reduced LPC stimulation of TREK-1 and TRAAK (Fig. 5, B-C) (n ϭ 6). Basal TREK-1 activity, unlike TRAAK, was inhibited by protein kinase A and protein kinase C stimulation (1,4,6). Similarly, activation of TREK-1 by LPC was rapidly reversed by 500 M 8-(4-chlorophenylthio)-cAMP (n ϭ 5) and 400 nM phorbol 12-myristate 13-acetate addition (n ϭ 3) (Fig. 6). Protein kinase A-mediated inhibition was slowly reversible over a 30-min wash, whereas protein kinase C inhibition was completely irreversible (Fig. 6). Fig. 7A shows that LPC added to the bath rapidly activated TREK-1 channels in the cell-attached patch configuration. In comparison, AA in the same configuration was weakly active and required several minutes of application to open the channels isolated by the patch pipette (Fig. 7A). In the excised inside-out patch configuration, LPC decreased basal activity, whereas AA reversibly and strongly opened TREK-1 (Fig. 7B). Moreover, intracellular LPC application reversed the opening induced by internal AA in the inside-out patch configuration (Fig. 7C). In the outside-out patch configuration, LPC was only very weakly effective, whereas AA fully activated TREK-1 (Fig.  7D). Similar results were obtained with TRAAK (not shown). In the outside-out patch configurations, intracellular ATP (5 mM; n ϭ 3) and GTP (1 mM; n ϭ 7) did not affect LPC activation. In the inside-out patch configuration, the addition of intracellular GTP␥s (100 M; n ϭ 3) or intracellular calcium (3 M free calcium; n ϭ 7) did not open channels. Cells treated for 8 h with the cytoskeleton-disrupting agents colchicine (500 M) and cytochalasine D (3 g/ml) displayed a similar sensitivity to LPC compared with control cells (n ϭ 6).
Site-directed mutagenesis was used to investigate the respective role of amino and carboxyl termini as well as the external M1P1 loop in TREK-1 channel activation by LPC as well as AA (Fig. 8). Deletion of the amino terminus and exchange of external loops between TREK-1 and TASK-1 had no effect on AA and LPC activation (Fig. 8B). However, partial deletion of the TREK-1 carboxyl terminus strongly reduced both AA and LPC activation (Fig. 8B).  1, 3, and 4 and the present report). The relatively low affinity of AA for channel activation as well as the slow kinetics suggest that the accessibility to the "AA site" is limited. Insertion of FAs in the membrane may be required to directly reach the active site of the channel protein or may alter fluidity, tension, curvature, or phospholipid-channel interaction, which in turn may open the mechano-gated ion channels (1,3). We have previously demonstrated that the anionic amphipath trinitrophenol mimics the effect of AA and opens TREK-1 and TRAAK channels (1, 3). These results were interpreted qualitatively on the basis of the bilayer couple hypothesis, assuming that the mechanosensitivity derives entirely from interactions within the bilayer and is independent of the cytoskeleton (1, 3, 23, 28). Anionic amphipaths would preferentially insert in the outer leaflet (because of the natural asymmetric distribution of negative charges in  TRAAK (1, 3).
Both TREK-1 and TRAAK are also opened by neutral amphipaths such as LPC, lyso-PAF, and PAF. The interpretation based on the bilayer couple model may thus be challenged as neutral amphipaths (such as LPC), which may distribute evenly at equilibrium in both leaflets, also mimic the effect of negatively charged compounds and open TREK-1 and TRAAK. An alternative hypothesis is to consider the shape rather than the charge of the molecule (15,16,21). LPC has a large polar head with a diameter of about 5.2 nm and a thin hydrophobic acyl chain tail with a cross-section estimated to be 1.8 -2.0 nm and can be pictured as a cone with its base at the polar interface (29 -32). Hydration of the polar head will also contribute to increase the conic shape of the molecule. Compounds with a small polar head and a large hydrophobic tail such as AA are considered as inverted cones with the base of the cone at its hydrophobic tail (30 -32). The cone shape hypothesis assumes that amphipaths distribute mainly in a single layer that is deformed in opposite directions by cone (LPC)-and invertedcone (AA)-shaped lipids (16, 21, 30 -32). This hypothesis also implies that the potency of cones is proportional to the size of the polar head rather than their charge. LPI and LPC (large head groups) are much more potent openers of TREK-1 and TRAAK than LPE and LPS (small head groups). LPC is neutral, whereas LPI is negatively charged, demonstrating that the effect of LPs is independent of the charge of the molecule. Therefore, the structural requirements for effective activations seem to be the presence of a large head group independently of its charge and a sufficiently hydrophobic domain. The data are apparently in favor of the cone shape hypothesis for LPs, but this hypothesis does not account for activation of TREK-1 and TRAAK by inverted cones such as AA.
In the cell-attached patch configuration, channel opening is induced within seconds when LPC is added to the bath outside of the patch pipette. However, patch excision produces a loss of channel activation by LPC, whereas AA is still able to maximally open channels. The loss of LPC activation after excision suggests the existence of a cytosolic factor, which may be involved in channel regulation. Channel activation by LPs is insensitive to PAF receptor antagonists and is not mimicked by LPA application, ruling out the eventual role of PAF and LPA receptors. External application of large size polar head coneshaped LPs may release a cytosolic factor by either directly activating putative LPs membrane receptor(s) or may indi- rectly affect membrane properties, thus activating a membrane-bound mechano-sensitive effector. Inhibition of AA releasing enzymes phospholipase A 2 and diacylglycerol lipase has no effect on LPC activation. The identity and the mechanism of release of the possible cytosolic factor, which may, similarly to AA, directly interact with the channels or affect membrane curvature, remain to be determined.
Deletional analysis indicates that the carboxyl terminus, but not the amino terminus and the external loop of TREK-1, is critical for LPC and AA activation. The same region was previously found to be important for stretch activation (2). The present data indicate that LPC activation requires the entire carboxyl terminus of TREK-1 and, thus, may involve the mechanosensitivity of the channel. During brain (and heart) ischemia two distinct phospholipase A 2 enzymes are at work. First, cytosolic phospholipase A 2 enzymes release FAs from phospholipids and, thus, generate massive amounts of intracellular LPs (for review, see Refs. 33,34). Second, secretory phospholipase A 2 -type II expression is rapidly induced, as it is also in endotoxic shock (35). The action of secretory phospholipase A 2 enzymes would produce external LPC, which would open TREK-1 and TRAAK channels, whereas activation of the cytosolic phospholipase A 2 would produce internal LPC, which would close the channels and reverse AA activation. These results suggest that secretory (external LPs) and cytosolic phospholipase A 2 (internal LPs) might differentially regulate TREK-1 and TRAAK opening. Cell swelling (opener of TREK-1 and TRAAK) and intracellular acidosis (opener of TREK-1) occurring during ischemia will produce further opening of TREK 1 and TRAAK. Hyperpolarization via TREK-1 and TRAAK activation would tend to reduce Ca 2ϩ influx via voltage-gated Ca 2ϩ channels by setting the membrane potential to values below the voltage threshold for activation. Similarly, hyperpolarization is expected to reduce Ca 2ϩ influx via the N-methyl-D-aspartate receptor-associated Ca 2ϩ -permeable channel by promoting blockade by Mg 2ϩ . Opening of the mechano-gated FAs/LPs/PAF-sensitive TREK-1 and TRAAK K ϩ channels during ischemia may thus have an important neuro-protective effect.