An Extracellular Ion Pathway Plays a Central Role in the Cooperative Gating of a K2P K+ Channel by Extracellular pH*

Background: TASK-3 is gated cooperatively by extracellular pH. Results: Mutual electrostatic interaction between K+ ions and two pH-sensing histidines occurs in a recently discovered extracellular ion pathway. Conclusion: Channel opening requires neutralization of both sensing histidines, with neutralization of the second sensor becoming favored by an electrostatic effect K+ ions. Significance: The work suggests a central role for the extracellular ion pathway in the gating of K2P K+ channels. Proton-gated TASK-3 K+ channel belongs to the K2P family of proteins that underlie the K+ leak setting the membrane potential in all cells. TASK-3 is under cooperative gating control by extracellular [H+]. Use of recently solved K2P structures allows us to explore the molecular mechanism of TASK-3 cooperative pH gating. Tunnel-like side portals define an extracellular ion pathway to the selectivity filter. We use a combination of molecular modeling and functional assays to show that pH-sensing histidine residues and K+ ions mutually interact electrostatically in the confines of the extracellular ion pathway. K+ ions modulate the pKa of sensing histidine side chains whose charge states in turn determine the open/closed transition of the channel pore. Cooperativity, and therefore steep dependence of TASK-3 K+ channel activity on extracellular pH, is dependent on an effect of the permeant ion on the channel pHo sensors.

Proton-gated TASK-3 K ؉ channel belongs to the K 2P family of proteins that underlie the K ؉ leak setting the membrane potential in all cells. TASK-3 is under cooperative gating control by extracellular [H ؉ ]. Use of recently solved K 2P structures allows us to explore the molecular mechanism of TASK-3 cooperative pH gating. Tunnel-like side portals define an extracellular ion pathway to the selectivity filter. We use a combination of molecular modeling and functional assays to show that pH-sensing histidine residues and K ؉ ions mutually interact electrostatically in the confines of the extracellular ion pathway. K ؉ ions modulate the pK a of sensing histidine side chains whose charge states in turn determine the open/closed transition of the channel pore. Cooperativity, and therefore steep dependence of TASK-3 K ؉ channel activity on extracellular pH, is dependent on an effect of the permeant ion on the channel pH o sensors.
K ϩ leak or background conductances are responsible for setting the resting membrane potential (1) and play crucial roles in the regulation of neuronal firing, muscle contraction, hormone, neurotransmitter and enzyme secretion, and transepithelial transport (2). K ϩ channels of a family known as K 2P have been found to be the molecular counterparts of these widespread background K ϩ conductances. Best known families of K ϩ channels are membrane proteins with six (K v ) or two (K ir ) transmembrane ␣-helices (TMs) 5 and one highly conserved P-domain that forms the K ϩ selectivity filter in a tetrameric arrangement around the conduction pathway (3). In contrast, K 2P K ϩ channels have four TMs and two P-domains (4,5) and form dimers with their P-domains arranged in pseudo-fourfold symmetry (6). Although K v and K ir channels are active at depolarized or hyperpolarized potentials, respectively, and play important roles during large membrane potential excursions typical of excitable cells, ubiquitously expressed K 2P channels are essentially voltage-independent and behave as open rectifiers as expected for K ϩ leakage pathways (2).
Although appearing as passive leaks, K 2P channels are finely tuned by a variety of stimuli. Important among these are intraand extracellular [H ϩ ] that gate the activity of 10 of the 15 known K 2P mammalian proteins (7). Examples are TASK subfamily K 2P channels TASK-1 (K 2P 3.1) and TASK-3 (K 2P 9.1), which are exquisitely sensitive to extracellular proton concentration (8 -11). TASK-3 is fundamental to various cellular functions including the modulation of sleep (12), aldosterone/renin secretion (13,14), and O 2 sensing (15), and its dysfunction has been linked to the Birk Barel mental retardation dysmorphism syndrome (16). TASK-3 is inactive (closed) at acid extracellular pH (pH o ) but activates (opens) as pH o is increased. The effect of pH o on gating is cooperative occurring with a Hill number n H of 2 (10). Although it is known that the pH o sensor is a histidine residue (His-98) residing near the extracellular mouth of the pore (9, 10), the cooperativity mechanism of the gating process remains unexplored.
Recent insight into the function of K 2P channels has come from the x-ray crystallographic resolution of TWIK-related arachidonic acid-stimulated K ϩ channel (TRAAK) and TWIK-1 structures (17,18). A distinguishing feature of these structures is an extracellular cap, corresponding to the TM1-P1 extracellular loop, which obstructs direct movement of ions into the pore normal to the plane of the membrane and defines instead two tunnel-like side portals, the extracellular ion pathway (EIP), which gives bilateral access to K ϩ binding sites in the selectivity filter (SF). We now explore the consequences of such an arrangement on the gating of TASK-3 by extracellular protons, in particular on the cooperativity mechanism. Our data strongly suggest that in these dimeric channels, both His-98 pH o sensors, which lie close to the wall of the EIP side portals, need to be neutralized for TASK-3 to become active and that neutralization of the sensor in one subunit enhances the propensity of that in the second subunit to become neutral. The permeant ion plays a central role in this change of acidity of the histidine sensor, an effect that is enhanced by the electrostatic effect of a glutamate residue (Glu-70) lining the wall of EIPs communicating the extracellular space and the outer mouth of the SF.

EXPERIMENTAL PROCEDURES
Electrophysiological Assays-Mus musculus TASK-1 (K 2P 3.1, GenBank TM accession number NP_034738) plasmid was obtained from Dr. Donghee Kim (Rosalind Franklin University, Chicago, IL). Cavia porcellus TASK-3 (K 2P 5.1, Gen-Bank accession number AF212827) plasmid was provided by Dr. Jürgen Daut (Marburg University, Germany). DNAs were subcloned into the pCR3.1 vector. Transient transfections were done in HEK-293 cells using CD8 cotransfection to identify effectively transfected cells (19). The CD8 antigen was revealed with microspheres (Dynabeads) coated with an anti-CD8 antibody. Site-directed mutagenesis and chimera constructions were done by PCR (19,20). Tandem dimer TASK-3 constructs were made after insertion of XbaI restriction sites to facilitate concatenation (21). The validity of the concatenated channel approach rests on the assumption that the linked subunits are both correctly inserted in the membrane and form the dimer. This sometimes requires shortening the linker between subunits, as reported before for the TASK-2 channel (21). There was no need for amino acid deletion for proper function of the concatenated TASK-3 constructs. Indeed the identical behavior of WT-H98N and H98N-WT TASK-3 constructs strongly suggests the correct assembling of concatenated channels.
Standard whole-cell patch clamp recordings were performed as described elsewhere (22,23). Currents were measured at several potentials as indicated. All chemicals were from Sigma.
The effect of pH on currents was evaluated by plotting current (I) measured at the indicated membrane potential, against extracellular [H ϩ ]. For graphical representation, average Ϯ S.E. values of I/I max values are used. The data were obtained from individual experiments. When appropriate, fit of a Hill equa-tion to the data was done for each individual experiment. The parameters are defined in Equation 1.
A similar analysis was done for inhibitor effect. Fits were done using the Marquardt-Levenberg algorithm as implemented in the SigmaPlot software.
Molecular Model for TASK-3-The TASK-3 homology model was built using Modeller (24). The template used was the TWIK-1 structure (Protein Data Bank (PDB): 3UKM) (18). Modeller created 10 structures, and the homology model with the lower value of root mean squared deviation of the backbone atoms regarding the template (root mean squared deviation ϭ 0.185 Å) was selected. The multiple alignment used to build TASK-3 model is based on that published by Brohawn et al. (17). The homology model of TASK-3 was validated using PROCHECK (25), and the tautomeric state of the neutral His-98 residue was assigned according to Stansfeld et al. (26). The homology model was embedded into a pre-equilibrated phosphatidyl oleoyl phosphatidylcholine bilayer in a periodic boundary condition box (10.2 ϫ 10 ϫ 10.5 nm) with pre-equilibrated TIP3P water molecules (27). Three K ϩ ions were associated to the models in positions S0, S2, and S4 of the selectivity filter and two water molecules at sites S1 and S3, respectively (26). Fifty-eight potassium ions were added to the aqueous phase to simulate a concentration of 150 mM K ϩ . The same number of chloride ions was added for electroneutrality. The initial configuration of the system was first optimized using energy minimization followed by a molecular dynamics (MD) simulation at 300 K for 1 ns. The protonation state of His-98 was assigned to the optimized homology model of TASK-3 using the CHARMM27 (28) force field. Models bearing both neutral (His 0 -His 0 ) and one (His ϩ -His 0 ) or both (His ϩ -His ϩ ) His-98 residues protonated were improved using energy minimization followed by two equilibrated MD simulations at 300 K for 8 ns (supplemental Fig. S1). The MD simulations were done in an isobaric isothermic ensemble using harmonic restraints of 0.5 kcal/mol Å 2 applied to the backbone atoms, but not to the selectivity filter. All MD simulations were performed using the NAMD program (29). The electrostatic interactions were computed with no truncations using the particle mesh Ewald algorithm (30) under periodic boundary conditions. The mutant TASK-3-E70K model was built in silico using the plug-in mutate of the VMD program on the basis of the optimized homology model of the TASK-3 wild type bearing both neutral His-98 residues. The TASK-3-E70K mutant was inserted into the membrane, minimized, and equilibrated following the same protocol as applied for wild-type TASK-3.
Electrostatic Potential Calculations-The electrostatic potential, (r), was calculated using the Poisson-Boltzmann equation implemented in the PBEQ module of the program CHARMm following a previously published protocol (31) (see also supplemental Fig. S2).
TASK-3 to gain a better understanding of the gating process. TWIK-1 shares with TASK channels the histidine sensor in P-domain 1. Amino acid sequences for TWIK-1 and TASK-3, on the other hand, are 23% identical and 62% similar in the extracellular cap/P1 region. Secondary structure prediction gives the ␣-helices likely to constitute the conserved extracellular cap seen in K 2P channel structures. Models bearing both neutral (His 0 -His 0 ) and one (His ϩ -His 0 ) or both (His ϩ -His ϩ ) His-98 residues protonated were subjected to MD. Fig. 1, A-C, show representations of the models at the end of 8-ns MD runs, with EIP depicted as a colored cavity. In all cases, the neutral His-98 side chains remain behind the SF at the bottom surface of the side portals near the pore. Protonation is accompanied by a movement of the charged His-98 side chains toward the EIP, which narrows the cavities as indicated by the change in color of EIP walls (Fig. 1, B and C). Both in the singly (His ϩ -His 0 ) and in the doubly (His ϩ -His ϩ ) protonated His-98 situations, but not in the His 0 -His 0 model, the outermost K ϩ ion bound at the selectivity filter becomes destabilized, implying a nonconductive channel. This is shown in Fig. 1D, which plots the position of the K ϩ ion sited at S0 at the beginning of separate 8-ns MD runs as a function of time. A previous MD study using a TASK-1 homology model proposed that a water molecule located behind the selectivity filter is critical for the stability of the SF, which is disrupted upon His-98 protonation (26). This water molecule is not present in the resolved K 2P structures and was not essential for the conformational changes described here. The MD data shown above suggest a requirement for side chain neutralization of both His-98 sensors of TASK-3 for channel activity, with protonation of single sensor leading to channel closure.

Role of Individual Histidine pH o Sensors of TASK-3-To test whether neutralization of both pH o sensors is indispensable to
lead to the open state of TASK-3, we have performed functional assays using concatenated channels (21). Tandem channels containing either a normal set of pH o sensors, mixed structures containing one able and one neutralized (H98N mutation) pH o sensor, or two neutralized pH o sensors were assayed for pH o sensitivity by patch clamp after expression in HEK-293 cells. Gating by pH o of wild-type TASK-3 channels was cooperative with pK1 ⁄ 2 6.0 Ϯ 0.05 and n H 2.1 Ϯ 0.09 (means Ϯ S.E., n ϭ 9). Extracellular H ϩ gating of the concatenated TASK-3 wild-type channels (WT-WT) did not differ significantly from that of the nonconcatenated equivalent channels ( Fig. 2A). Doubly mutated TASK-3-H98N proteins (H98N-H98N) were pH o -insensitive, but the mixed WT-H98N and H98N-WT constructs followed pH o in a noncooperative manner with pK1 ⁄ 2 values of 4.9 and 5.1 ( Fig. 2A and supplemental Table S1). Similar results were obtained at physiological 5 mM extracellular K ϩ (Fig. 2B  and supplemental Table S1). These results confirm the prediction of a requirement for both sensing histidines to be neutralized for TASK-3 to open. They also suggest that neutralization of a first His-98 sensor affects the pK a of the second, making it more susceptible to H ϩ loss.
Minimal Kinetic Model for Cooperative pH o Gating of TASK-3-The results with concatenated TASK-3 channels shown in Fig. 1 suggest that there is a requirement for both sensing histidines (His-98) to be neutralized for TASK-3 to open. The simplest model representing this situation is where CH 2 2 ϩ , CH ϩ , and C are doubly protonated, singly protonated, and unprotonated closed states and O is an open state. The dissociation constant for the first transition is 0.5K d1 , as two charged sensors, in monomers 1 and 2, could be deprotonated. The second transition is governed by 2K d2 , as the unprotonated closed state C can be generated from a complex containing a neutralized sensor at monomer 1 or 2. The relative open probability of the system will be given by and Fit of Equation 2 to the data for the WT-WT construct shown in Fig. 1A Fig. S3a) along the ion conduction pathway show a strong peak of positive potential imparted upon the SF by the doubly protonated His-98 side chains, the His ϩ -His ϩ configuration. This is much decreased in the His ϩ -His 0 situation. On this basis, we would propose that K ϩ ion occupation of the outermost binding sites of SF would be higher in His ϩ -His 0 than in His ϩ -His ϩ . This provides a rationale for cooperativity due to an increased facility of His-98 deprotonation in His ϩ -His 0 when compared with that in His ϩ -His ϩ because of an electrostatic influence of K ϩ ions.
The side portals predicted by modeling TASK-3 are, like those of TWIK-1 (18), funnel-shaped, becoming narrower nearer to the SF opening (Fig. 1A). Amino acids lining the walls of this EIP, including Gln-68, Glu-70, Pro-71, Gly-75, and Gln-209 as well as His-98, should confer a predominantly negative electrostatic potential to the cavity. As the cavity size (supplemental Fig. S4) is of the same order as a scantily hydrated monovalent ion, the electrostatic potential is expected to make a large contribution to local cation concentrations within the EIP. Glutamic acid 70 has been shown to lie near proton-sensing His-98 in TASK-3 and to modulate pH o sensitivity (32). Given the possible role of Glu-70 in determining EIP electrostatic potential, we wondered whether it might be important in the cooperativity between pH o sensors in TASK-3. The sensitivity of TASK-3 to pH o , but not the cooperative nature of gating, is affected by extracellular [K ϩ ]. Fig. 3A compares the pH o gating of TASK-3 at [K ϩ ] 0.5, 5, and 140 mM, together with fits of the Hill equation with n H not differing from 2 and respective pK1 ⁄ 2 values of 7.13, 6.67, and 5.97. The effect of [K ϩ ] agrees with the idea that increased occupancy by K ϩ favors the open state by affecting His-98 pK a . Mutation E70K acid-shifted the pH o dependence curves (Fig. 3B), as expected if Glu-70 also contributed to a local rise in [H ϩ ]. Remarkably, cooperativity was now absent at 0.5 mM [K ϩ ] and markedly diminished at 5 mM, but the fit to 140 mM [K ϩ ] data gave an n H 2.0 Ϯ 0.2 (see also supplemental Table S2). Graphs in Fig. 3, C and E, show a summary of fitted parameters of TASK-3 experiments at various voltages. There was no marked change in cooperativity with voltage, but there was a tendency for lower pK1 ⁄ 2 with depolarization, especially at the lower K ϩ concentrations. Cooperativity was low, as judged by n H values, after E70K mutation, and there was an increased effect of depolarization on both n H and pK1 ⁄ 2 (Fig. 3, D and F). This we attribute to an increased [K ϩ ] at the EIP by depolarization-promoted efflux of K ϩ from the high intracellular concentration (23).
Our results suggest an effect of residue Glu-70 on [H ϩ ] and [K ϩ ] in the vicinity of His-98 pH sensor so that charge reversal in TASK-3-E70K would lead to an apparent acid shift in pK1 ⁄ 2  Table 1. and a decrease in cooperativity. The loss of cooperativity in TASK-3-E70K therefore would be due to the increase in positive electrostatic potential (see supplemental Fig. S3b) because of charge reversal leading to a decrease in local K ϩ concentration. The recovery of TASK-3-E70K cooperativity at higher K ϩ concentration corroborates this interpretation.
Interestingly, TASK-1 K ϩ channel, which is another closely related member of the TASK subgroup of K 2P channels, has a lysine at the site equivalent to TASK-3 Glu-70, and its pH o gating is noncooperative. As shown before (11), and confirmed here (supplemental Fig. S5 and supplemental Table S3), pH o gating of TASK-1 becomes cooperative at high [K ϩ ] o . TASK-1 has a lysine residue at the place of TASK-3 Glu-70, but mutating this to glutamic acid, as in TASK-1-K70E, although shifting the pH o dependence curves did not confer cooperativity at the lower [K ϩ ] o . It is interesting to note that side chains on the wall of the putative EIP of TASK-1 make it hostile to cations, with a further positively charged residue, Arg-68 in place of Gln-68 of TASK-3.
Pharmacological Blockade of TASK-3 Is Also Dependent upon EIP Electrostatics-Independent evidence that Glu-70 might be an important determinant of electrostatic potential at the external ion pathway of TASK-3 comes from inhibition experiments by positively charged blockers Zn 2ϩ and ruthenium red (RR). Blockade of TASK-3 by Zn 2ϩ and RR are K ϩ -dependent and show virtual voltage independence, as expected from an interaction with superficial sites at the selectivity filter. Both Glu-70 and His-98 are proposed to form the blocker binding site. It is also possible that Glu-70 side chain additionally favors the interaction of the blockers by increasing their local concentration electrostatically. Fig. 4A shows that both TASK-3 and its mutant TASK-3-E70K are blocked by Zn 2ϩ and that the potency of the effect is decreased by increasing K ϩ concentration. Interestingly, as for the H ϩ concentration effect on gating, the effect of Zn 2ϩ was also cooperative, suggesting more than one blocker site ( Fig. 4A and supplemental Table S4). The shift in the K D (K D of TASK-3-E70K versus K D of TASK-3) for inhibition is compatible with a change in electrostatic potential (⌬) (31) of 46 mV at the site of action of the blocker. We have used the TWIK-1-based model of TASK-3 (His-98 0 -His-98 0 configuration) to explore the interaction of Zn 2ϩ with the channel. Fig. 4B shows a Zn 2ϩ ion positioned at a site in the EIP where the switch from Glu-70 to Lys is predicted to give a ⌬ of ϳ46 mV. After an 8-ns MD run (Fig. 4C), the Zn 2ϩ appears coordinated by His-98, which hat has flipped upwards, Glu-70, and Gln-209 (not shown). This conformational change is accompanied by disruption of the selectivity filter and a switch of the position of the second His-98 from behind the SF toward the EIP as though forming a second Zn 2ϩ site and thus providing a rationale for cooperative inhibition. Fig. 4D shows the result of an MD run with two Zn 2ϩ ions originally positioned as for Fig. 4B but at opposite ends of the EIP.
In contrast with the Zn 2ϩ effect, RR inhibits TASK-3 in a noncooperative manner with a K D of 10 M (supplemental Fig.  S6a). This is explained by the shape of the inhibitor that could plausibly be coordinated by both His-98-Glu-70 pairs in each monomer and thus span across the SF mouth (supplemental Fig. S6b). Use of the TASK-3 minus TASK-3-E70K ⌬ value calculated above predicts an RR (valence 6) ratio K D , TASK-3-E70K/K D , of 6 ϫ 10 4 . This prediction matches well the lack of effect of RR up to 1 mM on TASK-3-E70K channels.

DISCUSSION
TASK-3 shares with congener K 2P channels TASK-1 and TWIK-1 a sensitivity to extracellular pH that is transduced through protonation and neutralization of histidine 98 located at P1. Although TWIK-1 pH o gating is complex, involving a profound change in selectivity only recently begun to be understood (33), more is known about TASK-1 gating. Gating of K 2P channels by extracellular H ϩ is known to occur at the selectivity filter by a process that has been thought to be, at least superficially, akin to C-type inactivation commonly encountered in other K ϩ channels (reviewed in Ref. 34). K ϩ dependence of TASK-1 pH o gating is in accord with this concept (11,35). Yuill et al. (36) modeled TASK-1 based on the structure of the KcsA channel and first suggested that His-98 sensor was located behind the selectivity filter where its protonation might alter the shape of the SF producing a nonconducting state. The same authors show that site-directed mutagenesis affecting ion selectivity produced concomitant changes in pH o dependence, consistent with SF gating. Later MD studies provide the important observation that protonation of His-98 leads to a flipping upwards of the sensing His-98, which, in addition to producing a deformation of the SF moving away K ϩ -coordinating backbone carbonyl groups from its lumen, would create "an electropositive barrier to K ϩ ions at the outer mouth of the channel" (26). This proposal is interesting as it links pH o gating with its known K ϩ dependence. It necessitates, however, that the protonated His-98 side chain would somehow avoid the bulk solution bathing the extracellular aspect of the channel to exert this proposed electrostatic effect. Differences in pH o sensitivity between TASK-1 and TASK-3 have been attributed to the effect of residues in the large TM1-P1 extracellular loop on the sensing machinery. Indeed Clarke et al. (32) employed TM1-P1 loop exchange TASK-1/TASK-3 chimeras to show its importance in pH o dependence and Zn 2ϩ blockade, concluding that the loop must lie close to the channel pore. A point not addressed in the literature cited above is that of the higher sensitivity of TASK-3 to external H ϩ that is expressed through the cooperative nature of the gating process.
The recent discovery of the structure of TWIK-1 and TWIKrelated arachidonic acid-stimulated K ϩ channel (17,18), two K 2P K ϩ channels, has revealed the presence of an extracellular helical cap that is likely to be a conserved feature among members of this family of transport proteins. The extracellular cap, made of the TM1-P1 linkers, impedes extracellular access to the SF on the main axis of conduction of the channel. Instead it defines what has been called side portals or the EIP that only allows bilateral access to the mouth of the selectivity filter through rather narrow tunnel-like structures. We show that the presence of these side entrances crucially determines the main gating process of TASK-3, namely the opening and closing of the pore in response to changes in extracellular pH. Our data  supplemental Table S4. The model for TASK-3 has been used to explore the binding of Zn 2ϩ to the extracellular entrance pathway of the channel. Results are means Ϯ S.E. In B, a Zn 2ϩ ion was positioned at a point within the EIP such that replacement of Glu-70 by lysine, as in TASK-3-E70K, would give a ⌬ in electrostatic potential of ϳ46 mV (see "Results"). The sensing histidines are in the neutral configuration. In C, the position adopted by the Zn 2ϩ ion at the end of an 8-ns MD run is shown. Notice the bending upwards from behind SF into the EIP of the His-98 side chains in both monomers as well as the twisting of the selectivity filter hydroxyls. D shows the positions of the Zn 2ϩ ions at the end of an 8-ns MD run after positioning one Zn 2ϩ at each side of the EIP as in B. FEBRUARY 22, 2013 • VOLUME 288 • NUMBER 8 indicate that the concentration of K ϩ in the cavity must be a major determinant of pH sensitivity and the cooperativity exhibited by TASK-3 in its gating process. The side chains of pH-sensing histidine residues, one in each monomer of these homodimeric channels, line the side portals and constrain their diameter, bursting into the cavity when protonated. In channel structures lacking the extracellular cap characteristic of K 2P channels, we would expect the protonated His-98 imidazole charge to be screened out by the external solution (31). Instead its electrostatic effect should be heightened in the constrained space of the EIP, particularly after its narrowing by flipping upwards of the His-98 side chain. The negatively charged Glu-70 side chains, already known to be located close to the His-98 sensors (32), also form part of the side portal wall and influence the electrostatic potential in EIP, presumably becoming a major determinant of the EIP K ϩ concentration and affecting the occupancy of the outermost SF K ϩ binding sites.

K 2P Extracellular Ion Pathway and pH o Gating
Independent evidence for the importance of electrostatics of the EIP comes from experiments with TASK-3 blockers Zn 2ϩ and RR. The inhibition by these, and other divalent cations, has been shown to depend on the presence of the Glu-70 residue (32,(37)(38)(39) and a neutral His-98 (32,38). Here we show that inhibition of TASK-3 by Zn 2ϩ , just as pH o gating, is cooperative and strongly impeded by increasing extracellular K ϩ concentration. MD experiments suggest that two Zn 2ϩ ions might plausibly bind at sites defined by neutral His-98 side chains flipped upwards into the EIP at both sides of the SF entrance. The presence of Glu-70, however, does not appear to be necessary as Zn 2ϩ still inhibited TASK-3-E70K channels, albeit with a higher K D . The results suggest that the main effect of Glu-70 on Zn 2ϩ inhibition is in electrostatically affecting its concentration in the EIP. A similar role for Glu-70 was surmised from the effect of divalent cations on TASK-3 single channel conductance (37). Our experiments comparing Zn 2ϩ inhibition of TASK-3 and TASK-3-E70K allowed us an estimation of the electrostatic potential due to the charge replacement at a site of action of the divalent cation. Use of this value predicted a lack of effect of RR 6ϩ on TASK-3-E70K that has been previously reported (39) and is corroborated here. In contrast to Zn 2ϩ inhibition, RR blockade was not cooperative, in line with the binding of a single inhibitory molecule. Using the x-ray structure of RR and our model of TASK-3, we show the plausibility of such an interaction with RR spanning between the two Glu-70/ His-98 sites. Czirják and Enyedi (39), although lacking structural information, had already suspected such an arrangement when proposing that RR acted by tethering the two subunits of TASK-3 through their Glu-70 side chains.
Our results suggest a molecular explanation for the cooperativity and [K ϩ ] dependence in the gating of TASK-3. We postulate that the vicinity of His-98 charged side chains to the outer part of the SF leads to a decrease in K ϩ occupancy because of an electrostatic effect of the charged imidazole rings. Neutralization of one of the His-98 sensors results in a decrease in electropositive potential accompanied by an increased occupancy of the outermost selectivity filter K ϩ binding sites. Increased K ϩ occupancy should in turn affect the readiness with which the imidazole of the remaining charged His-98 loses its H ϩ , thus accounting for the observed increase in acidity of pK a observed experimentally for the neutralization of a second pH o sensor in TASK-3. Our interpretation provides a good explanation for the loss of cooperativity in TASK-3-E70K and its recovery at higher K ϩ concentration.
The presence of side portals with predominantly negative walls in TASK-3 is crucial in providing a restricted space for the gating process we postulate. The effect of local environment in the modulation of the pK a of key residues in both enzymes and ion channels is known to play a role in their catalysis/gating mechanisms (22,40). The mechanism for cooperativity of channel gating proposed here, a perturbation in environment caused by the transported ion itself leading to a shift in pK a of a gating residue, is, to our knowledge, novel, as is the proposed role of the newly discovered extracellular cap structure of K 2P channels in the process. TASK-3 is fundamental to various cellular functions and has been implicated in disease. The structural and functional insights gained here might help in the design of molecules to interact specifically with the gating machinery of the channel and modulate its function.