Mechano- or Acid Stimulation, Two Interactive Modes of Activation of the TREK-1 Potassium Channel*

TREK-1 is a member of the novel structural class of K+ channels with four transmembrane segments and two pore domains in tandem (1, 2). TREK-1 is opened by membrane stretch and arachidonic acid. It is also an important target for volatile anesthetics (2, 3). Here we show that internal acidification opens TREK-1. Indeed, lowering pH i shifts the pressure-activation relationship toward positive values and leads to channel opening at atmospheric pressure. The pH i -sensitive region in the carboxyl terminus of TREK-1 is the same that is critically involved in mechano-gating as well as arachidonic acid activation. A convergence, which is dependent on the carboxyl terminus, occurs between mechanical, fatty acids and acidic stimuli. Intracellular acidosis, which occurs during brain and heart ischemia, will induce TREK-1 opening with subsequent K+ efflux and hyperpolarization.

The near completion of the sequencing of the nematode Caenorhabditis elegans genome recently identified more than 80 K ϩ channel genes divided into three major structural classes: (i) the inward rectifiers with two TMS 1 and a single P domain; (ii) the Shaker types with six TMS and a single P domain comprising the voltage-gated Kvs, the calcium-activated Slo, the calcium-regulated SK, the Eag/Erg, and the KQT channels; and (iii) the two P types with 4TMS being the largest structural class (about 50 genes) (4 -6). Despite an overall similar 4TMS/2P structure, the sequence identity between these channels is very low (less than 30%) (5,6).
The mammalian family of 4TMS/2P K ϩ channels comprises TWIK-1, TWIK-2, TASK-1, TASK-2, TREK-1, and TRAAK (1,(7)(8)(9)(10)(11). TWIK-1 and TWIK-2 are widely distributed and encode K ϩ -selective channels with a characteristic weak inward rectification (8,11). TASK-1 is found principally in the pancreas, placenta, lung, brain, and heart (9,12,13). TASK-1 lacks intrinsic voltage sensitivity and is thus a pure background K ϩ -selective channel. Moreover, TASK-1 is extremely sensitive to variations of extracellular pH in a narrow physiological range, with 90% of the maximum current recorded at pH 7.7 and only 10% at pH 6.7 (9). TASK-2, another background K ϩ channel recently isolated from human kidney, shares the ex-ternal pH sensitivity of TASK-1 (10). Unlike the other 4TMS/2P channels, TASK-2 is almost absent from the brain and is mainly expressed in the kidney. Murine TREK-1 is widely distributed with a strong expression in the brain and in the heart (1). It is activated by membrane stretch, by AA as well as inhalational anesthetics, while it is inhibited by a cAMP-dependent phosphorylation (2,3). Interestingly, TREK-1 shares the properties of the Aplysia S-type K ϩ channel, which is involved in presynaptic facilitation underlying a simple form of learning (14,15). TRAAK, another mouse mechano-gated AA-sensitive 4TMS/2P K ϩ channel, is only expressed in neuronal tissues including brain, spinal cord, and retina and lacks sensitivity to cAMP (7,16).
In the present report, we demonstrate that the AA-sensitive mechano-gated K ϩ channel TREK-1 is opened by intracellular acidosis. Mutagenesis experiments identify the carboxyl-terminal region of TREK-1 as critical for the integration of both mechanical and acidic stimuli.
For 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 (16). Pressure-effect relationships were fitted with Boltzmann equations. AA was dissolved in ethanol at a concentration of 100 mM, flushed with argon, and kept at Ϫ20°C for 1 week. All chemicals were obtained from Sigma.

RESULTS
TREK-1 cDNA was transiently transfected in COS cells, and channel activity was recorded using the whole cell patch clamp configuration. In a cell voltage-clamped at 0 mV, AA superfusion induces a strong outward current (Fig. 1A). In the same cell, repetitive application of 90 mM HCO 3 Ϫ , which produces intracellular acidification (31,32), mimics AA stimulation (5.5 Ϯ 0.5-fold increase, n ϭ 29 at 0 mV). Both AA and HCO 3 Ϫ are ineffective on control mock-transfected cells (0.5 Ϯ 0.4, n ϭ 7) (Fig. 1B). Moreover, substitution of 90 mM NaCl by an equivalent concentration of sodium gluconate does not mimic the activation of TREK-1 by NaHCO 3 (n ϭ 6). The I-V curve of the current induced by HCO 3 Ϫ shows a prominent outward going rectification in physiological K ϩ conditions and reverses at the predicted E K ϩ value of Ϫ80 mV (Fig. 1C). When external Na ϩ is substituted with K ϩ , the reversal potential shifts to 0 mV, and the I-V curve remains outwardly rectifying (Fig. 1D). AA similarly activates TRAAK, while HCO 3 Ϫ is ineffective (0.9 Ϯ 0.1, n ϭ 17) (Fig. 1E).
The activation of TREK-1 by the addition of HCO 3 Ϫ is also observed at the single channel level in the cell-attached patch configuration (n ϭ 13) ( Fig. 2A). In this configuration, the pH of the external solution bathing channels under recording is clamped by the pipette medium (pH 7.2), and channel modulation is expected to be due to intracellular effects. In this experiment, the activity of the mechano-gated channel TREK-1 is recorded both at atmospheric pressure and during the application of a Ϫ66 mm Hg pressure stimulation ( Fig. 2A, inset). At atmospheric pressure, channel activity (NPo) is very low (0.80 Ϯ 0.22, n ϭ 27). Both the resting and the pressureinduced activities are reversibly stimulated by the HCO 3 Ϫ addition ( Fig. 2A). Superfusion of a CO 2 -rich solution, which also produces a strong intracellular acidification (32), leads to TREK-1 opening in the cell-attached patch configuration (n ϭ 15) (Fig. 2B). Again, both basal and stretch-induced activities are strongly stimulated. The current induced by CO 2 is outwardly rectifying and reverses at Ϫ80 mV ( Fig. 2B, inset). Another classical approach to alter pH i -regulated mechanisms is the NH 4 Cl prepulse technique, which relies on the greater membrane permeability for NH 3 than for NH 4 ϩ ions (32). The addition of NH 4 Cl (producing intracellular alkalinization) does not affect TREK-1 channel activity, while washout of NH 4 Cl (producing intracellular acidosis) strongly stimulates TREK-1 channel activity (n ϭ 6) (Fig. 2C). The current activated by NH 4 Cl withdrawal similarly displays a strong outward rectification (Fig. 2C).
The effects of intracellular acidification were also studied on excised inside-out patches expressing TREK-1 (Fig. 3). Channel activity was recorded at both atmospheric pressure and during membrane stretch. Gradual intracellular acidification from 7.2 to 5.0 induces channel opening at atmospheric pressure (Fig. 3, A and B). NPo is strongly increased, while the single channel conductance is gradually decreased by internal acidosis (Fig. 3B). Half-maximal activation is induced at pH i 6.0, and a drop of 0.7 pH i unit already produces a significant increase in channel activity (Fig. 3B). A Ϫ42 mm Hg stretch induces a robust channel opening at intracellular pH between 7.2 and 6.0, although it fails to further open TREK-1 at pH i 5.0 (Fig. 3A). The pressure-activity relationship of TREK-1 is presented in Fig. 3C. At physiological intracellular pH 7.2, the pressure-activity relationship is described by a Boltzmann function with a pressure required for half-maximal activity (P 0.5 ) of Ϫ36.8 Ϯ 2.1 mm Hg (n ϭ 13). Progressive lowering in pH i gradually shifts the P 0.5 toward positive values, leading to constitutive channel activity under atmospheric pressure at pH i 5.0 (Fig. 3, C and D). Half-maximal effect is observed at pH 5.9 (Fig. 3D).
The modulation of channel activity by intracellular acidification was studied on deleted TREK-1 mutants in the inside- Ϫinduced current. The holding potential is Ϫ80 mV, and voltage ramps of 800 ms in duration are applied from Ϫ130 to 100 mV every 10 s. Both control and HCO 3 Ϫrich solutions contain 5 mM K ϩ . D, same cell as C in the presence of 155 mM K ϩ . E, whole cell recording of a COS cell expressing TRAAK (same conditions as A).
out patch configuration. We used a protocol including two acidic steps to pH i 6.0 and 5.0 at both atmospheric pressure and during a Ϫ66 mm Hg stretch (Fig. 4, B-D). Lowering internal pH from 7.2 to 6.0 reversibly induces TREK-1 wild type opening at atmospheric pressure and slightly potentiates channel activity during stretch (Fig. 4B). Lowering internal pH to 5.0 further opens TREK-1 at atmospheric pressure with no additional effect at Ϫ66 mm Hg. Deletion of the N-terminal region of TREK-1 does not alter pH i sensitivity or mechanogating (n ϭ 5; data not shown). However, serial deletion of the C terminus progressively impairs TREK-1 activation by pH i (Fig. 4E). Like TREK-1 wild type, at atmospheric pressure, ⌬46 is opened at pH 6.0 and 5.0 (Fig. 4E, second panel), while the next deletion inward, ⌬89, lacks activation at pH 6.0 and is only mildly opened at pH 5.0 (Fig. 4, C and E, third panel). Activation of ⌬89 by a membrane stretch of Ϫ66 mm Hg is strongly potentiated at both pH 6.0 and pH 5.0 (Fig. 4, C and E,  third panel). ⌬100 is only weakly activated at atmospheric pressure by pH 5.0, and channel activity induced by a Ϫ66 mm Hg stretch is only enhanced at pH 5.0 (Fig. 4E, fourth panel). ⌬103 is not opened at atmospheric pressure by internal acidification at either pH 6.0 or pH 5.0, and stimulation of the stretch-induced activity is only observed at pH 5.0 (Fig. 4, D  and E, fifth panel). No channel activity is detected with ⌬113 at both pressures and at all pH conditions tested (n ϭ 7).
Progressive deletion of the cytosolic region of TREK-1 shifts the pressure-activity curve toward more negative values, leading to less sensitive mutant channels (Fig. 5, A-C). For instance, the relationship for ⌬103 is about 60 mm Hg more negative compared with that of TREK-1 at pH 7.2 (Figs. 3C and 5C). Lowering pH i to 6.0 and then 5.0 gradually shifts the relationship toward more positive values and strongly stimulates channel activity elicited by pressure (Figs. 4D and 5C).

DISCUSSION
Mammalian mechano-gated K ϩ channels have been previously described in atrial and ventricular cardiac myocytes, in neurons from mesencephalic and hypothalamic areas of the brain as well as in kidney (17)(18)(19)(20)29). Negative pressure applied to cell-attached patches activates K ϩ channels (17)(18)(19)(20). The pressure to induce half-maximal activation is between Ϫ12 and Ϫ18 mm Hg at ϩ40 mV. I-V curves are outwardly rectifying, and single channel conductances are 94 and 143 picosiemens at ϩ60 mV in symmetrical K ϩ for cardiac and brain cells, respectively. Openings induced by stretch are typically bursty and flickery. The probability of these channels to open at a fixed pressure is voltage-dependent with a higher opening at depolarized potentials. Both cardiac and neuronal channels are similarly opened by AA and other lipophilic compounds in the micromolar range (17)(18)(19)(20). AA activation is found in cells treated with cyclo-oxygenase and lipoxygenase inhibitors, indicating that AA itself can directly activate these channels. Unsaturated fatty acids (linoleic, linolenic, and docosahexaenoic acids) but not saturated fatty acids also activate these channels. A very important property of these native mechanogated arachidonic-sensitive K ϩ channels is that they are also stimulated by lowering cytoplasmic pH over the range 7.2-5.6 (17,20), and the channels are more sensitive to pressure at acidic intracellular pH.
Two recently cloned 4TMS/2P channels, TREK-1 and TRAAK (1,2,7,16), have many of the properties of endogenous mechano-gated K ϩ channels. Both channels are activated by shear stress, cell swelling, and membrane stretch (2,16). Moreover, TREK-1 and TRAAK are opened by AA, linoleic, linolenic, and docosahexaenoic acids but are resistant to saturated fatty acids (2,7). The single channel conductance of about 100 pS at ϩ50 mV in symmetrical K ϩ , the outward going rectification, the bursty and flickery openings, and the voltage dependence are identical to described endogenous channels (2, 16 -20).
The mechano-gated K ϩ channel TREK-1 is opened, like native cardiac and neuronal mechano-sensitive K ϩ channels (17- Ϫ (substituting KCl) in the bath is indicated by a horizontal bar. The inset shows currents induced by membrane stretch (Ϫ66 mm Hg) in control and during HCO 3 Ϫ superfusion. Zero current is indicated by a horizontal dashed line. B, effects of a CO 2 -rich solution on a cell-attached patch from a cell expressing TREK-1 (same conditions as A). The inset shows I-V curves performed with voltage ramps of 800 ms in duration from Ϫ130 to 100 mV in control and during CO 2 addition (different cell from B). C, effects of addition and washout of 20 mM NH 4 Cl (substituting KCl) on a cell-attached patch from a cell expressing TREK-1 (same conditions as A). During wash-out of NH 4 Cl, voltage is stepped to Ϫ50 and ϩ50 mV (as indicated by arrows). 20), by intracellular acidosis. The mechanism by which pH i variations activate TREK-1 involves a shift of the pressureactivity curve toward positive pressures, leading to constitutive channel opening at atmospheric pressure under mild internal acidosis. The acidic stimulation of TREK-1 was seen in whole cell (mediated by HCO 3 Ϫ ), cell-attached (mediated by HCO 3 Ϫ , CO 2 , and NH 4 Cl prepulse), and excised inside-out patch configuration (internal acidification), suggesting that protons directly affect channel activity. To our knowledge, TREK-1 is the only cloned K ϩ channel reported so far to be directly opened by intracellular acidosis.
TRAAK is significantly (p Ͻ 0.05) less sensitive to stretch (P 0.5 ϭ Ϫ46 Ϯ 2 mm Hg) (16) compared with TREK-1 (P 0.5 ϭ Ϫ36 Ϯ 2 mm Hg). It is also much less sensitive to pH i variations. This is a major difference between the two channels, which can now be used to ascribe a native AA-and mechanosensitive K ϩ channel to TREK-1 rather than to TRAAK. Another major difference between TREK-1 and TRAAK is the lack of inhibition by cAMP (7). Phosphorylation of Ser-333 in the carboxyl terminus of TREK-1 is responsible for the cAMPinduced down-modulation (2). The possibility that intracellular acidosis mediates TREK-1 opening via an interaction with the protein kinase A phosphorylation site, which is not present in the TRAAK structure, is ruled out, since ⌬89, which lacks Ser-333 and is not sensitive to cAMP, remains stimulated by low pH i .
Deletion and chimeric analysis indicate that the pH i -sensitive region of TREK-1 is located in the carboxyl-terminal region between Val-298 and Thr-368 (⌬113-⌬46 region). Further deletion in this carboxyl-terminal region impairs activation by stretch, AA, and pH i , demonstrating its critical importance for channel function (2). The sensitivity to pH i is conferred to TRAAK when the proximal carboxyl terminus of TREK-1 is exchanged with TRAAK, demonstrating that the region between ⌬46 and the fourth TMS is necessary and sufficient to provide pH sensitivity. Progressive deletions of the carboxyl terminus of TREK-1 show that pH i sensitivity as well as mechano-gating is gradually altered. These results indicate that the whole segment (Val-298 to Thr-368) is probably involved in acidic and stretch modulation and that several amino acids may indeed be involved.
TREK-1 is modulated by a variety of mechanical stimuli such as stretch, swelling, and shear stress and by a variety of chemical stimuli including AA, ligands producing cAMP-dependent phosphorylation, and acidic stimuli (2, 3). TREK-1 is therefore an example of molecular integrator. Furthermore, mechanical activation of TREK-1 is enhanced by intracellular acidosis, demonstrating that a response to one type of stimulus alters the sensitivity to others.
By integrating multiple stimuli, TREK-1 probably fulfills an essential physiological function in the nervous and cardiovascular systems. Under physiological conditions, effectors of TREK-1 activity are probably stretch, AA, and neurotransmitters or hormones that increase intracellular cAMP (2). It is unlikely that activation of TREK-1 by acidification of the intracellular medium is an important physiological stimulus, although one cannot eliminate transient variations below pH i 7.0 (33).
Protection against an exaggerated cellular Ca 2ϩ invasion is one of the important roles of several types of K ϩ channels (34,35). This has been particularly well established for the large conductance Ca 2ϩ -activated K ϩ channels (K Ca 2ϩ channels of the BK type) as well as for ATP-sensitive K ϩ channels (K ATP channels) (36 -38). When Ca 2ϩ invades a cell, one of the ways to resist further Ca 2ϩ invasion (which would be deleterious), from voltage-sensitive Ca 2ϩ channels and/or through NMDA receptors, is hyperpolarization. Hyperpolarization puts the cell membrane potential far from the threshold of voltage-sensitive Ca 2ϩ channel activation and favors the NMDA receptor-associated Ca 2ϩ -permeable channel blockade by Mg 2ϩ . It has been particularly well demonstrated that, when hypoxia, anoxia, or/and hypoglycemia occur, the associated decrease of intracel-lular ATP and the related increase of intracellular ADP result in the activation of K ATP channels (34, 35, 39 -41). This activation produced by inhibition of the energetic metabolism is protective in tissues where the channel is expressed such as in the heart or brain (35). In conditions of intracellular acidification, which would occur in many physiopathological situations such as brain or heart ischemia, opening of TREK-1 (K Hi channel mode) may fulfill a similar protective role. TREK-1 channels in the K Hi channel mode could work in concert with K Ca 2ϩ (BK type) and K ATP channels. Cellular swelling, which also accompanies ischemia (42), would increase the effect of intracellular acidification on the activity of the TREK-1 channel (2). Ischemia in heart and brain is associated with a major K ϩ efflux (34,35,43). This efflux is supposed to be through K ϩ channels that open under these conditions (34,35). However, specific blockers of K Ca 2ϩ channels or of K ATP channels do not eliminate this efflux (43,44). TREK-1 channels may thus constitute an important pathway for this K ϩ efflux. FIG. 5. The carboxyl terminus of TREK-1 is critical for mechano-gating. A, mean pressure at 0 mV required for half-maximal activation (P 0.5 value) of various TREK-1 mutants. P 0.5 is determined by fitting Boltzmann equations to the pressure-effect relationships as illustrated in B and C. B and C, pressure-effect relationships for mutants ⌬89 and ⌬103. Pressure-effect relationship for TREK-1 wild type (WT) at pH 7.2 is indicated with a dashed line. P 0.5 values are indicated by vertical arrows.
FIG. 6. Transfer of proton sensitivity to TRAAK by fusing the proximal carboxyl terminus of TREK-1. A, intracellular acidification (as indicated by steps) does not alter TRAAK activity in an insideout patch. The holding potential is 0 mV, and Ϫ66 mm Hg pressure applications are indicated by vertical arrows. Zero current is indicated by a horizontal dashed line. B, diagram illustrating the chimeric construct (see "Experimental Procedures" for further details). C, inside-out patch showing that internal acidosis stimulates TRAAK/TREK-1 chimera at both atmospheric pressure and during a stretch at Ϫ66 mm Hg (same conditions as A). D, histogram illustrating the effects of various pH i (as indicated) on TRAAK NPo recorded at atmospheric pressure and during a Ϫ66 mm Hg stretch. The number of experiments is indicated, and conditions are identical to A. E, same as D for chimera TRAAK/TREK-1.