Asymmetric mechanosensitivity in a eukaryotic ion channel

Significance

One important way in which living organisms are able to detect and respond to their environment is via the conversion of mechanical forces into electrical signals. However, the molecular mechanisms that enable mammalian “mechanosensitive” ion channels to detect a wide profile of forces within the membrane remain unclear. By studying the functional activity of individual TREK-2 K2P channels inserted in different directions into a lipid bilayer, we are now able to describe how the asymmetric structure of this channel enables it to sense such a broad profile of forces. These results help us understand how eukaryotic ion channels respond to a rich variety of sensory stimuli.

Abstract

Living organisms perceive and respond to a diverse range of mechanical stimuli. A variety of mechanosensitive ion channels have evolved to facilitate these responses, but the molecular mechanisms underlying their exquisite sensitivity to different forces within the membrane remains unclear. TREK-2 is a mammalian two-pore domain (K2P) K⁺ channel important for mechanosensation, and recent studies have shown how increased membrane tension favors a more expanded conformation of the channel within the membrane. These channels respond to a complex range of mechanical stimuli, however, and it is uncertain how differences in tension between the inner and outer leaflets of the membrane contribute to this process. To examine this, we have combined computational approaches with functional studies of oppositely oriented single channels within the same lipid bilayer. Our results reveal how the asymmetric structure of TREK-2 allows it to distinguish a broad profile of forces within the membrane, and illustrate the mechanisms that eukaryotic mechanosensitive ion channels may use to detect and fine-tune their responses to different mechanical stimuli.

Many forms of sensory perception in higher organisms depend on the rapid conversion of mechanical and physical stimuli into the electrochemical language of nerve and muscle cells (1). These transduction mechanisms often involve “mechanosensitive” (MS) ion channels that are able to rapidly convert physicomechanical stimuli into the electrical signals required for complex physiological responses (e.g., touch, pain, hearing, proprioception), as well those mechanical signals that play key roles in development and cell–cell communication (2). The enormous complexity and diversity of these sensory responses suggests that a variety of molecular mechanisms are likely responsible; for example, while some channels are regulated via physical coupling to proteins of the cytoskeleton and/or cell matrix, others respond directly to changes in lipid membrane tension (3). The precise structural and physical mechanisms underlying these processes remain poorly understood, however (4–6).

Our current understanding of how membrane tension directly regulates ion channel gating is largely derived from studies of prokaryotic MS ion channels involved in the control of bacterial turgor pressure (7, 8). The open state of these MS channels has a larger cross-sectional area than the closed state within the membrane, and stretch-induced changes in the forces within the bilayer stabilize the expanded open-state conformation. This is commonly referred to as the force-from-lipid principle of mechanosensitivity (9). Similar physical principles are thought to regulate many eukaryotic MS channels (5, 10).

Typically, the functional activity of MS ion channels is examined in pipette-based patch-clamp recordings in which negative or positive pressures are used to alter bilayer tension. These pressure changes stretch the membrane to alter the lateral pressure profile, the complex profile of forces that vary with location across the bilayer (11–13). MS channels embedded within the bilayer have been shown to directly respond to these changes (14, 15). The application of either positive or negative pressure increases the overall tension within both leaflets of the bilayer, and there is now a large body of evidence demonstrating how such changes in lateral tension underlie the activation of MS channels (2–5). However, within their cellular environment eukaryotic MS channels are able to respond to a diverse and complex range of mechanical stimuli, such as changes in osmotic and mechanical pressure, strain and shear forces, acceleration, vibration, and sound, which may modify membrane mechanics in myriad ways. Thus, for intrinsically MS channels gated solely by membrane tension, it seems unlikely that symmetrical changes in lateral membrane tension alone (i.e., equal changes in both leaflets) can fully explain the complexity of the response and the rich variety of physiological processes that MS channels are able to control. Interactions with the cytoskeleton and other proteins will undoubtedly influence the response (1), but a fundamental understanding of their intrinsic response to changes in bilayer tension is still required. Previous computational and theoretical studies of membrane tension have emphasized how stretch-induced changes in the lateral pressure profile might not be symmetrical between the two leaflets (i.e., inner and outer) of the lipid bilayer and explored how this may influence gating (16–20), but these studies were restricted mostly to prokaryotic MS channels. More tractable and better-defined experimental systems for the study of eukaryotic MS channels are needed.

TREK-2 is a eukaryotic MS ion channel that belongs to the two-pore domain (K2P) family of K⁺-selective channels and is important for mechanical and thermal nociception in many sensory neurons (21, 22). Other members of this subgroup of MS K2P channels include TREK-1 and TRAAK. All three channels have now been shown to respond directly to changes in membrane tension (23, 24), and their mechanosensitivity is thought to be important for many diverse physiological processes, including electromechanical feedback in the heart, cell volume regulation and perception of shear flow stress in epithelial tissues, pressure-induced vasorelaxation in endothelia, and even cell migration (25–28). Crystal structures are also now available for these three channels in multiple conformations (29–32); thus, they provide an excellent opportunity to investigate the molecular mechanisms underlying mechanosensitivity in eukaryotic ion channels.

We have recently shown how increased (symmetrical) lateral tension within the lipid bilayer induces a structural change in TREK-2 from the “down” state of the channel to the more expanded “up” state conformation (31, 33, 34). However, unlike the nonselective prokaryotic MS channels, in which stretch-induced expansion dilates the central pore, K2P channels remain highly K⁺-selective (35). Therefore, it is perhaps not surprising that stretch-induced expansion in TREK-2 occurs primarily via movement in the lower half of the protein, i.e., away from the filter and within the cytoplasmic inner leaflet alone (Fig. 1). Thermodynamic descriptions of mechanogating refer to asymmetric structural changes like this as “shape sensitivity.” Along with changes in overall cross-sectional area, this represents one of the main types of conformational change thought to underlie the mechanosensitivity of integral membrane proteins (13, 36, 37).

Fig. 1.

Asymmetric structural changes in response to membrane stretch. (A) The TREK-2 (down-state) K2P channel structure embedded within a lipid bilayer. Only the TM helices (M1–M4) are colored. (B) Overlay of surface representations of the down (gray) and up (red) conformations of the channel to illustrate the difference in cross-sectional area of the two conformations. Note that most of the structural changes are seen in the lower half of the protein. (C) As reported previously (34), these asymmetric changes are seen more clearly in this comparison of the stretch-induced change in cross-sectional area. (D) The cross-sectional area that TREK-2 assumes within the bilayer also can be considered a funnel shape. On membrane stretch (red), the lower half expands, and the overall structure becomes more symmetrical. The upper half of the protein remains relatively constant to maintain the structural integrity of the selectivity filter.

One prediction of this shape sensitivity for TREK-2 is that it also should allow the channel to detect an increase in tension within an individual leaflet of the bilayer. Differences in tension between the two leaflets not only result in an asymmetry of the lateral pressure profile, but also generate additional forces across the membrane known as membrane torque (36, 38). Typically, such asymmetric changes are elicited either by alterations in membrane curvature or via insertion of noncylindrical lipids into individual leaflets (36, 39). There is extensive evidence of the effects of noncylindrical lipids on TREK channel function (40, 41), as well as of the asymmetric effects of the MS toxin, GsMTx4 (42). However, conflicting reports on the response of TREK channels to negative vs. positive pressure (23, 24, 43) means that the contribution of membrane curvature and shape-sensitivity to K2P channel mechanosensitivity remains unclear.

In this study, we used a combination of functional and computational approaches to investigate how the different structural states of TREK-2 define its response to membrane tension. Importantly, our results provide clear evidence of an asymmetric response to changes in bilayer tension that allows TREK-2 to detect a broad profile of forces within the membrane. Overall, this highlights the complex role of the bilayer environment in the gating of a eukaryotic MS ion channel.

Results

Structural Changes in Response to Asymmetric Membrane Tension.

Increasing lateral membrane tension within the bilayer is associated with increases in the area per lipid and free volume within the membrane (14, 44). In a previous study, we used a well-established molecular dynamics simulation approach to simulate these effects of membrane stretch (34). This method of symmetrically stretching the membrane increased the area per lipid in both the inner and outer leaflets of the bilayer at the same time and elicited a structural transition in TREK-2 from the down state to the up state (Fig. 1B); however, the resultant change in cross-sectional area of the channel was confined almost exclusively to the inner leaflet and thus was highly asymmetric (Fig. 1 C and D). Thus, we examined whether an increase in tension confined to the inner leaflet alone is sufficient to drive this structural transition. Previous studies of prokaryotic MS channel activation have used the addition of molecules to a single leaflet to generate differences in monolayer stresses (45); however, in the approach that we used here, the area of the simulation box in the plane of the bilayer was kept constant, and lipids were selectively removed from a single leaflet (either inner or outer). This generated an increase in area per lipid in a single leaflet equivalent to the increase in tension used previously in our (symmetric) stretch protocols (34) (Fig. 2A and Materials and Methods).

Fig. 2.

Structural changes in TREK-2 induced by pressure profile asymmetry. (A) The NPAT simulation method used to produce asymmetric changes in tension. The dimensions of the simulation box are kept constant (box shown in black) and lipids are removed at random from either the inner leaflet or the outer leaflet to increase the area per lipid in that leaflet. (Left) Structure of the down state before lipid removal. (Right) Structure after tension is increased only in the inner leaflet. Orange spheres represent the lipid headgroups, and shading relates to lipid density. Note that this protocol does not distort the bilayer. (B) Pressure profile asymmetry induced by removal of lipids from the inner leaflet (blue). A symmetrical unstretched bilayer is shown in black and reveals major changes in tension in the inner leaflet and pressure profile asymmetry across the bilayer. (C, Left) An rmsd comparison of the simulated TREK-2 down-state structure vs. the up-state crystal structure. When tension is selectively increased in the lower leaflet by removal of lipids, the down state quickly moves toward the up state (blue); however, when tension is increased only in the outer leaflet (black) by lipid removal, there is no movement toward the up state. Also no movement is seen if the number of lipids is increased in the outer leaflet alone suggesting that a change in torque alone is not sufficient. (C, Right) Overlay of the simulated structures after 200 ns with the up-state structure induced by symmetrical stretch shown in red (34).

This asymmetric stretch protocol did not cause any major alteration in the shape of the bilayer (Fig. 2A), but it did produce an asymmetric change in the lateral pressure profile across the bilayer (Fig. 2B). In particular, a large decrease in the positive pressure was produced by the lipid tails within the core of the bilayer when their number was reduced, with a concomitant increase in lateral tension at the interfacial region near the headgroups. A small compensatory increase in positive pressures was also seen in the opposite leaflet, but overall the changes are restricted to the leaflet in which lipids are removed. To monitor any conformational changes in TREK-2, we compared the rmsd of the Cα atoms for the TM helices (M1–M4) against the up-state crystal structure [Protein Data Bank (PDB) ID code 4BW5]. This measurement provides an estimate of how closely the structure resembles the up-state conformation. Interestingly, increasing tension within the inner leaflet alone induced a conformational change toward the up-state over the course of a 200-ns simulation to produce a structure similar to that induced by symmetrical tension (Fig. 2 B and C). In contrast, no movement toward the up state was seen when tension was altered in the outer leaflet alone by either increasing or decreasing the number of lipids in the outer leaflet (Fig. 2C).

This structurally asymmetric response suggests that TREK-2 also should exhibit a functionally asymmetric response to the stretch induced by positive vs. negative pressure, and although they have been questioned (46), the functionally asymmetric effects of noncylindrical lysolipids are consistent with this hypothesis. Nevertheless, direct functional studies of membrane curvature and leaflet asymmetry are experimentally challenging (6, 47), particularly with eukaryotic MS channels. Traditionally, pipette-based electrophysiological recordings are used in conjunction with the application of either positive or negative pressure to stretch the membrane; however, numerous factors arising from the Ω shape of the membrane within the pipette and preferential adhesion of the outer leaflet to the glass complicate the interpretation of such experiments (6, 18, 48), and might have contributed to the differences in the reported effects of negative vs. positive pressure on K2P channels (23, 24, 43). Therefore, we chose a more controlled approach and examined the mechanosensitivity of purified TREK-2 proteins reconstituted in a lipid bilayer system with a defined symmetrical composition and less-complex geometry.

TREK-2 Single-Channel Activity in a Planar Lipid Bilayer.

In previous studies, we showed that a purified truncated version of TREK-2 retains its core functional properties, including its intrinsic mechanosensitivity, when reconstituted into a planar lipid bilayer (31, 34). For these experiments, we reconstituted TREK-2 into giant unilamellar vesicles (GUVs) and used a planar patch clamp system, as described in Materials and Methods) (49). This GUV approach has the advantage of a less-complex geometry than pipette-based recordings and fewer of the complications arising from the Gibbs plateau border in traditional BLMs.

However, one problem is that TREK-2 orients randomly when reconstituted into GUVs, which precludes any macroscopic analysis of membrane asymmetry. Thus, we examined whether it was possible to measure responses to positive and negative pressure at the single channel level. Fig. 3A shows a typical recording of these purified TREK-2 channels. Under these conditions (symmetrical 200 mM K⁺, −80 mV), two main conductances were observed; one large conductance of ∼305 pS and another of ∼155 pS. These arose from oppositely oriented channels that have an asymmetric unitary conductance, however. This rectification is observed more clearly in recordings of single channels that reveal only a single channel type with either the larger unitary conductance at negative potentials or vice versa for a channel in the opposite orientation (Fig. 3B). A difference in channel activity is also seen at the equivalent of positive potentials, with the smaller conductance exhibiting increased open probability and a less flickery behavior (Fig. S1 A–D).

Fig. 3.

Recording of single TREK-2 channels in a planar lipid bilayer. (A) Single-channel recordings of purified TREK-2 channels in an unstretched planar patch-clamp lipid bilayer. Cartoon showing how the channel randomly incorporates in either the inside-out (green) or outside-out orientation (red). (Inset) A recording from a bilayer with two oppositely oriented channels. (B) Two types of single channel activity are observed, denoted TREK-2^in/out or TREK-2^out/out. These opposite orientations can be distinguished based on their rectification. (Left) Single TREK-2^in/out channel with lower (∼155 pS) conductance and long openings at +80 mV, but higher (∼305 pS) conductance and flickery openings at −80 mV; also see Fig. S1A. (Right) Single TREK-2^out/out channels exhibit the opposite characteristics. (C) Confirmation of orientation with specific blockers. TPA blocks TREK-2 from the inside at a site within the inner cavity. A bilayer with at least four TREK-2^in/out channels is blocked when TPA is applied from the equivalent of the inside. (D) In contrast, ruthenium red blocks at a site within the outer mouth of the pore, just below the CAP domain. This bilayer contains at least two TREK-2^out/out channels, which are blocked by the addition of ruthenium red from the equivalent of the outside. Control experiments in the reverse orientation for each blocker are shown in Fig. S1B.

Fig. S1.

(Related to Fig. 3.) TREK-2 single-channel characteristics and orientation. (A) I/V plot showing that the single-channel conductance of TREK-2^in/out channels exhibit voltage-dependent rectification. (B) Voltage dependence of the open probability for the same channels. (C) Plot highlighting the rectification seen in the conductance of single inside-out channels recorded at +80 mV and −80 mV, and the inverse relationship for single outside-out channels. (D) Inside-out channels exhibit a higher open probability at positive potentials compared with negative potentials. The inverse relationship is seen for single outside-out channels. All data represent mean ± SEM (n = 4). (E) TREK-2^in/out channels are unaffected by ruthenium red (10 µM) applied from the equivalent of the inside. (F) TREK-2^out/out channels are not inhibited by TPA (200 µM) applied from the equivalent of the outside, thus confirming orientation of these channels.

The single-channel conductance of TREK-2 reportedly varies depending on the length of the N terminus, with shorter splice variants and truncations stabilizing a larger, ∼250 pS conductance with similar rectification (50, 51). Therefore, given the truncated N terminus of the purified protein used in these studies, the single-channel conductances that we found here (in 200 mM KCl) are entirely consistent with those expected for TREK-2. Nonetheless, to confirm channel identity and orientation, we made use of blockers known to inhibit TREK-2 from either the outside or the inside of the membrane.

Tetrapentylammonium (TPA) blocks TREK-2 channels from the intracellular side by binding within the inner pore (52). In contrast, ruthenium red selectively blocks TREK-2 from the outside by binding just below the external CAP domain (53). To enable comparisons of orientation and voltage with pipette-based recordings, we refer to channels oriented with their extracellular cap domain downward toward the glass chip as “inside-out” (TREK-2^In/Out), with the opposite orientation termed “outside-out” (TREK-2^Out/Out). Using this notation, we found that TPA inhibited channel activity only when applied from the equivalent of the inside, and that ruthenium red inhibited this activity only from the outside (Fig. 3 C and D and Fig. S1 E and F). This finding confirms that these recordings represent TREK-2 inserted in opposite orientations, and that these different orientations can also be distinguished based on their different unitary conductances.

Asymmetric Mechanosensitivity of TREK-2.

Previous models have suggested that a functionally asymmetric response to positive vs. negative pressures results from differences in stretch-induced membrane curvature and the resulting pressure profile asymmetry. This appears to be supported by the asymmetric action of lysolipids and the MS toxin GsMTx4 on channel function (40, 42). However, the overall lateral tension within the membrane will increase in both leaflets when it is stretched by either negative or positive pressure. Therefore, if increased tension per se is sufficient to activate TREK-2, then the channel should be activated by both positive and negative pressure changes. However, if the asymmetric structure of the channel allows it to differentiate between changes in tension in the two leaflets, then this will be orientation-dependent and also may produce a difference in response to the different types of curvature induced by positive vs. negative pressures.

To examine this question, we applied either positive or negative pressure to bilayers containing a single TREK-2 channel. Fig. 4A shows that an inside-out channel is activated by an increase in negative pressure (−80 mbar), but when the same channel was exposed to an equivalent jump in positive pressure (+80 mbar), no equivalent activation was observed, but rather a small decrease in activity was seen. This functionally asymmetric response to different pressure changes between −120 and +120 mbar is shown in Fig. 4B. Importantly, the opposite behavior was also seen in a single outside-out channel.

Fig. 4.

Asymmetric response to positive vs. negative pressure. (A) Bilayers with single channels in either the inside-out (Left) or outside-out orientation (Right) were stimulated by pressure jumps from 0 to ±80 mbar. The pump used to apply pressure is positioned underneath the bilayer and the direction of the pressure change on the bilayer is shown by the arrows. (B) The mean log₁₀ fold change in Po for each channel orientation is shown in response to various pressure changes from −120 to +120 mbar. The values above each error bar indicate the number of independent replicates, and error bars represent SEM. Asterisks represent the significance of difference between the two orientations at opposite pressures: ***P ≤ 0.001, **P ≤ 0.01, *P ≤ 0.05 (two-way unpaired t test). Note that while negative pressure activates inside-out channels, the outside-out channels are preferentially activated by positive pressure (i.e., the equivalent of negative pressure for inside-out channels).

Inverse Asymmetric Regulation of Oppositely Oriented Channels.

These results suggest that, under these particular recording conditions, the asymmetric changes in shape and cross-sectional area of TREK-2 are reflected in a functionally asymmetric response to changes in positive vs. negative pressure. However, although these recordings are from purified proteins in GUVs using a planar patch clamp configuration, the precise geometry and mechanics of these bilayers is unknown and they may already possess some form of inherent asymmetry that might account for this behavior.

To examine this further, we next studied the simultaneous behavior of two oppositely oriented channels within the same bilayer. In this important control experiment, if these two channels exhibit equally asymmetric (but inverse) responses to positive vs. negative pressure changes, then this effect is unlikely to result from any intrinsic artifacts in the recording system, e.g., curvature when the bilayer forms a Giga-Ohm seal on the glass chip, or preferential adhesion of a single leaflet to the glass surface. If any such intrinsic asymmetry exists within the recording system, then it is unlikely that a channel in the opposite orientation also would exhibit the inverse behavior, particularly if this behavior is seen simultaneously in two or more such oppositely oriented channels.

Fortunately, the asymmetric single-channel conductance of TREK-2 makes it possible to distinguish between channels in opposite orientations. However, when more than one channel exists, openings of identical conductance can occur concomitantly (Fig. 2A). We therefore created an in-house analysis protocol to facilitate sorting of channel activity into contributions from either ‘inside-out’ or ‘outside-out’ channels (Materials and Methods).

Fig. 5A and Fig. S2 show how this approach can be used to analyze the pressure response in recordings containing up to two channels of each orientation. In response to a step of −80 mbar, the inside-out channels become activated, while the oppositely-oriented outside-out channels are not activated and exhibit a much smaller fold decrease. In contrast, when the same channels are then exposed to +80 mbar, the exact opposite is seen, with the outside-out channels becoming activated preferentially over inside-out channels. The relative responses of these oppositely oriented channels to steps between −120 and +120 mbar, shown in Fig. 5B, clearly demonstrate a functionally asymmetric response with preferential activation by the equivalent of negative pressure.

Fig. 5.

Inverse regulation of oppositely orientated channels. Simultaneous recording of bilayers with at least one channel of each orientation, and up to a maximum of two of each. (A) Example of change in channel currents in response to either positive or negative pressure (recording at −80 mV). The color chart is the result of our in-house analysis used to classify channels according to their orientation. (B) The mean log₁₀ fold change in Po for these oppositely oriented channels is shown in response to various pressure changes from −120 to +120 mbar. The values above each error bar indicate the number of independent replicates, and error bars represent SEM. Asterisks represent the significance of difference between the two orientations: ***P ≤ 0.001, **P ≤ 0.01, *P ≤ 0.05 (two-way unpaired t test). Note that while negative pressure activates only inside-out channels (green), only positive pressures activate outside-out channels (red), i.e., the equivalent of negative pressure for inside-out channels.

Fig. S2.

(Related to Fig. 5.) Custom-made MATLAB analysis for isolation of activity based on single channel properties. This example illustrates analysis of five data segments representing various combinations of channel openings from a bilayer with two TREK-2^in/out channels and two TREK-2^out/out channels. (A) Red dots on top of the trace indicate slope evaluation points (see B and C). Data segments (black) are overlaid on colored lines that represent amplitude thresholds used to assign conductance (see D for color key). Each cluster is evaluated based on various parameters (e.g., slope, closing frequency, immediate neighbor). (B) Two simultaneously opened TREK-2^out/out (Left) have similar conductance as a single TREK-2^in/out (Right), but different opening kinetics. (C) The slope evaluation points distinguish TREK-2^in/out from TREK-2^out/out, and are given for each data point as the amplitude difference between the preceding point and before the subsequent point. Examples of data slope points are shown in red dots above the traces; the cutoff threshold is set at 8 pA. (D) To distinguish 2× TREK-2^out/out + 1 TREK-2^in/out from 2 TREK-2^in/out, both the slope value and the immediate neighbor are considered, because a single channel opening or closing is more likely than two channels opening and/or closing simultaneously.

Stretch Activation Preferentially Destabilizes a Closed State of the Channel.

Although both the up and down conformations of TREK-2 can be conductive, our previous studies have shown that stretch activation preferentially stabilizes the up conformation, and that this has a positive allosteric effect on the filter gating mechanism (31, 33). The precise relationship between the different conformational states of any ion channel and their behavior at the single-channel level is complex and not fully understood, however. Previous single-channel recordings of wild-type TREK-2 have been difficult to analyze because initiation of alternative translation produces multiple isoforms with different conductance levels (51), but the purified TREK-2 protein studied here represents a single species with a well-behaved large conductance. Therefore, we examined the response of these single channels to stretch activation in more detail.

The typical behavior of a single inside-out channel at −80 mV is shown in Fig. 6A. Calculation of the open and closed dwell time distributions indicates that they can be well fit with one main open time and three distinct closed times (Fig. 6B). The application of membrane stretch increases the overall channel open probability, but does not affect either the open dwell time or the two shorter closed times (Fig. 6C). Instead, the increase in burst duration and open probability appears to arise through a selective reduction in the longest of the three closed times (Fig. 6D), thus suggesting a possible link between these structural and functional states.

Fig. 6.

Selective effect of pressure activation on the long closed state of TREK-2. (A) Representative segment of a single TREK-2^in/out channel recording with a burst expanded as indicated by the arrows. (B) Maximum likelihood fits to open, closed, and burst dwell time distributions from an unstretched bilayer at 0 mbar. (C and D) The same, but recorded after a pressure jump to −80 mbar. Note the selective reduction in the long closed time (black arrow).

Discussion

In this study, we used a multidisciplinary approach to demonstrate that the MS TREK-2 K⁺ channel can detect a pressure profile asymmetry within a lipid bilayer membrane. Furthermore, we have demonstrated that this asymmetric mechanosensitivity results from uneven changes in the shape and cross-sectional area of the protein. This confers TREK-2 with the ability to sense a complex profile of forces across the membrane, and is likely to be critical for the diversity of physicomechanical coupling processes that MS K2P channels are required to regulate.

Living organisms are able to perceive and respond to a diverse range of mechanical stimuli using a variety of MS ion channels. Structures of the prokaryotic MscS/MscL channels have provided valuable insight into the underlying molecular mechanisms, but structures of some of the key eukaryotic MS channels, such the TREK/TRAAK family of K2P channels and the piezo channels, have been obtained only recently. The propeller- like curved arms and overall size of the piezo channels immediately suggest possible mechanisms for sensing changes in membrane curvature, but current structures are of insufficient resolution for detailed studies (2, 54). In contrast, a large number of high-resolution structures for the TREK/TRAAK family have now been identified, providing an excellent model system for how these MS channels move between their different conformations in response to membrane stretch.

For MS channels to exhibit the diversity of responses required to control so many different physiological processes, they must be able to respond to a more complex profile of forces beyond symmetrical changes in lateral tension. An example of this is the ability to differentiate between changes in tension in individual leaflets of the bilayer, that is, to detect a pressure profile asymmetry. Several thermodynamic descriptions of mechanosensitivity have proposed mechanisms by which this might occur, including “shape-sensitivity,” whereby the change in cross-sectional area of the protein differs between the inner and outer leaflets.

Several ideas for the asymmetric mechanosensitivity in K2P channels were proposed before their structures were solved (41), but up to now direct evidence for this has been either lacking or contradictory, possibly owing to differences in the recording systems used. A cautionary tale highlights many of the experimental challenges involved (6), and it is unlikely that our recording system is free of all geometric and/or mechanical artifacts. Nonetheless, the ability to study the simultaneous inverse responses of oppositely oriented channels provides a powerful argument against such artifacts as the origin of the functional asymmetry observed.

Three basic types of structural deformation that underlie the thermodynamics of MS channel gating have been proposed (36): (i) a symmetrical change in cross-sectional area within the plane of the membrane that responds to lateral tension; (ii) a change in shape (without an overall change in area) that responds to differences in tension or “torque” between the leaflets; and (iii) a change in length that responds to hydrophobic mismatch. It is highly unlikely that these different mechanisms are mutually exclusive, but our previous study suggests that hydrophobic mismatch by itself is not sufficient to drive this structural transition in TREK-2 (34). In addition, these new results implicate a role for a combination of the other two mechanisms, a change in cross-sectional area and a change in shape, during TREK-2 mechanogating.

The structural and functional response of these MS TREK channels to a symmetrical change in the pressure profile is consistent with previous predictions for the required change in cross-sectional area (37). We now also provide evidence of the ability of these channels to detect pressure profile asymmetry. The asymmetric structural changes in TREK-2 that underlie this response result not only from the need to maintain the integrity of the selectivity filter (35), but also from the requirement to detect a wide variety of mechanical stimuli.

Symmetric changes in the lateral pressure profile will elicit conformational changes in TREK-2 independent of channel orientation. However, when this asymmetric channel structure inserts into the membrane, it has been shown to affect the local environment (i.e., pressure profile) differently in the outer leaflets compared with the inner leaflets (34). Thus, although changes in global curvature per se are thought to be of negligible energetic consequences for MS channel activation, the direction of the curvature produced by positive vs. negative pressure will be relevant, because it has different effects on the two leaflets of the bilayer (19, 20, 55). Furthermore, many of the biologically relevant mechanical forces that act on the membrane are not likely to act equally on both leaflets and also will produce asymmetric effects on the pressure profile, such as those shown in Fig. 7. Similar effects also may result from different lipid compositions in the two leaflets, and thus the inherent structural asymmetry of TREK-2 is well designed to detect any such pressure profile asymmetry resulting from this.

Fig. 7.

Response of TREK-2 to an asymmetric pressure profile. This cartoon illustrates a model of how TREK-2 detects a pressure profile asymmetry. To maintain the structural integrity of the selectivity filter, stretch-induced expansion occurs primarily within the lower leaflet. This shape sensitivity means that the channel can detect an asymmetry increase in tension such as that induced by membrane curvature or other asymmetric forces, and when analyzed in a bilayer, these responses will be orientation dependent. Overall, such asymmetric functional responses will allow these MS channels to focus the effects of different mechanical forces and to fine-tune cellular electrical responses accordingly.

Sensitivity to asymmetry within the pressure profile will endow these channels with the ability to focus the effects of different mechanical forces that act on the membrane and to fine-tune cellular electrical responses accordingly. This functional diversity also may be further modified by the intracellular C-terminal domain that extends from M4; however, this domain is missing in the current crystal structures, and no structural information is available, so speculation as to precisely how this domain might modulate the response of these channels to tension and/or curvature may be premature.

Our results also suggest other possible advantages of this planar patch-clamp bilayer recording system. In particular, the ability to obtain high-quality single-channel recordings of TREK-2 channel activity may provide further insight into the relationship between channel structure and single-channel activity during mechanogating. Thus, this recording system provides an excellent tool for use in future studies in this area.

In summary, the results presented here provide compelling evidence for a versatile mechanism for sensing membrane tension by a eukaryotic MS ion channel that allows the functional integration of a broad profile of membrane forces. These effects not only are consistent with its structurally asymmetric structural architecture, but also are more representative of the complex sensory mechanisms that underlie its physiological role. The structures of other eukaryotic MS channels are also now becoming available, and it will be interesting to examine how the mechanisms that we describe here are related to the function of other MS ion channels.

Materials and Methods

Molecular Dynamics.

Increased bilayer tension results in an increase in the lateral area per lipid (APL) (34, 44). Thus, asymmetric tension was simulated in each leaflet using an NPAT ensemble and creating mismatch in the lateral APL in either the outer or inner leaflet; that is, the simulation box size was fixed and the number of lipids in each leaflet was varied. In our previous study (34), the maximum P-50 stretch protocol increased the APL by ∼35%; thus, here we used a lipid mismatch of 35% in the NPAT ensemble simulations to mimic these increases. To mimic a similar change in torque (without an increase in tension in the inner leaflet), lipids were added at random into the outer leaflet by gradually increasing their number and energy minimization to avoid an infinite force on the atom(s). The maximum mismatch in this case was limited to +16% to prevent bilayer distortion and instability. All-atom simulations of TREK-2 in these asymmetric bilayers were then run as before (34). In brief, a 200-ns equilibrium NPT ensemble simulation of the TREK-2 down state (PDB ID code 4XDJ) was performed to obtain the equilibrium APL. The last frame was then used as the initial structure for 200-ns NPAT ensemble simulations after randomly removing (or adding) lipids in the relevant leaflet. Simulations were repeated with different randomly removed lipids and random velocities. All simulations were performed with GROMACS 4.6 (56), Charmm36 force field (57) and a 2-fs integration time step. Each simulation contained approximately 200 lipids, 25,000 TIP3P water molecules, and 150 mM KCl. Semi-isotropic pressure was controlled using a Parrinello–Rahman barostat (58) with a compressibility of 0 in the x-y plane and 4.5 × 10⁻⁵ bar⁻¹ in the z direction. Lateral pressure profiles were calculated as described previously (34).

Preparation of TREK-2 and GUVs.

The human TREK-2 (KCNK10) crystal construct protein (Gly⁶⁷ to Glu³⁴⁰) was expressed and purified as described previously (26). In brief, 1,2-diphytanoyl-sn-glycero-3-phosphocholine was dissolved in chloroform to a concentration of 10 mM and stored at −20 °C. The GUVs where then made by electroformation in 1 M sorbitol using Vesicle Prep Pro (Nanion Technologies). Purified TREK-2 was mixed with GUVs to a final concentration of ∼1–5 µg/mL and then incubated overnight at 4 °C with Bio-Beads (Bio-Rad) before use.

Bilayer Recordings.

All electrophysiological recordings were performed with a Nanion Port-a-Patch planar patch-clamp system connected to an Axopatch 200B amplifier via a Digidata 1440A digitizer (Molecular Devices). The microstructured apertures in the glass chips of this planar patch-clamp system are used to form Giga-Ohm seals on TREK2-containing GUVs that burst on contact with the glass surface (49). All data were filtered at 10 kHz and recorded at a sampling rate of 250 kHz. Experiments were carried out in symmetrical 200 mM KCl, 10 mM Hepes pH 6.0 solutions. Channel open probability (Po) was evaluated over 5-s periods before and after each stated pressure jump. Data from single-channel patches were idealized in MATLAB with a half-amplitude threshold crossing algorithm, and a 43-µs resolution was imposed before distributions were calculated using the maximum likelihood method. The critical time for classification of bursts was found by numerically solving 1 − e^−tc/tm = e^−tc/tl for tc, where tm and tl are the intermediate and long closed dwell times, respectively (59). Plots and statistical analysis were performed with GraphPad Prism 5 software. Procedures for the analysis of channel activity based on orientation are described below.

Single-Channel Sorting Protocol.

The voltage dependence of the single-channel properties of TREK2 allows dissection of channel orientation. To achieve this, a custom MATLAB analysis protocol was created that facilitates sorting of channel activity into contributions from TREK-2^In/Out and TREK-2^Out/Out channels (Fig. S2). Multiple channel activity is sorted into clusters (based on signal amplitude) that represent possible channel combinations, and then joined together, excluding brief deflections below the threshold limits. After removal of clusters shorter than 40 µs, overlapping prospective channel combinations are evaluated and classified based on a set of parameters that include closure frequency, slope evaluation (sum of slope evaluation points/cluster length) (Fig. S2C), ideal TREK-2^In/Out (frequency of fast 28–32 pA amplitude jumps/cluster length), and immediate neighbor signals. The rationale behind the immediate neighboring signals is that a fast transition between different cluster categories should be as simple as possible; that is, a single channel opening or closing is more likely than two channels opening and/or closing simultaneously.

A key feature of the script is the ability to distinguish one TREK-2^In/Out from two simultaneously opened TREK-2^Out/Out channels that have similar combined amplitude. To validate this approach, we evaluated 30-s recordings of patches with either two TREK-2^In/Out or two TREK-2^Out/Out channels. In a 30-s recording of a bilayer with two TREK-2^In/Out channels, the script correctly identified 94.2% of the activity as being from TREK-2^In/Out, while 1.8% was incorrectly categorized as TREK-2^Out/Out. When a bilayer with two TREK-2^Out/Out channels was evaluated for 30 s, the script correctly identified 96.9% of the signals as TREK-2^Out/Out, misinterpreted 0.1% as TREK-2^In/Out and missed 3% (Fig. S2). The MATLAB script for channel sorting is available at: https://github.com/biologichael/Channel-sorting.