INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION CONCLUSION REFERENCES

Mechanosensitive (MS) ion channels are operationally defined by a particular aspect of their behavior during patch clamp recordings: their P_open changes with changing membrane tension. When MS channels were first reported (Guharay and Sachs, 1984) some questioned whether the cation currents were carried by discrete proteins, but K⁺-selective MS unitary currents (Sigurdson et al., 1987) confirmed that channels can be tension-sensitive. Better yet, K⁺-selective MS channels of two classes, stretch-activated and stretch-inactivated (SA and SI) were found to coexist in membrane patches (Morris and Sigurdson, 1989), suggesting that membrane tension provided channel-specific gating energy. Initially, therefore, the expectation was that special "mechano-gating" motifs would be uncovered in the various channels whose P_open changed substantially with stretch. However, an antithetical view of the diversity of MS channels observed by patch clamp is that inherent mechanosusceptibility is widespread among channels because, for stochastic multistate membrane proteins, the likelihood that all transition rates are tension-independent is low ... with the corollary that ... channels in situ normally rely on cellular mechanoprotection to prevent mechanical interference in gating (cellular mechanoprotection is any long- or short-range feature that minimizes tension changes felt by a channel, e.g., excess bilayer, membrane skeleton, auxiliary subunits).

If this view is valid, then relaxation of mechanoprotection could allow for mechanotransducer currents, and pathological disruption of mechanoprotection (say via membrane skeleton abnormalities) could lead to mechano-leak currents in pathologies (e.g., muscular dystrophy, ischemia, physical trauma).

As a test for the notion that inherent mechanosusceptibility can be unrelated to specialized mechano-gating protein motifs, we turned to an intensively studied channel whose nonmechanical physiology is thoroughly established, namely the voltage-gated K⁺ channel, Shaker. The test system was unabashedly nonphysiological because 1) Shaker's native auxiliary -subunit (Yao and Wu, 1999) was not provided; 2) an N-terminal truncated (inactivation-removed (IR)) version of the channel with a small foreign C-terminal epitope was used; 3) Shaker-IR, an arthropod nerve-muscle channel, was expressed in an amphibian oocyte, an environment unlike its native excitable cells; 4) in most experiments, cytoplasm was missing (i.e., excised patches were used); and 5) the plasma membrane was traumatized because patch clamp plus stretch stimuli unavoidably disrupt membrane skeleton (Small and Morris, 1994) and mechanoprotective membrane undulations (Zhang and Hamill, 2000). Lipid rafts (see Martens et al., 2000), too, may disperse in patches.

Except for stretch, these conditions have been extensively used in studying Shaker gating. Although pipette aspiration of a gigaohm-sealed membrane is a convenient way to stretch a membrane patch, the absolute tension produced is at best only estimated (Sachs and Morris, 1998). Thus, to ensure the robustness of our findings, we used mechanically varied recording conditions including large and small patches, excised and cell-attached configurations, positive and negative pipette pressures. Controlling membrane tension from one stimulus to the next is difficult enough in a given patch, but comparing one patch to another is even more challenging. Accordingly, within-patch comparisons are used where possible, and for between-patch comparisons, data are normalized.

We show that Shaker-IR is an MS channel. The message is not "this is a trivial artifact." Quite the contrary. If inherent mechanosusceptibility such as Shaker's is widespread among nonmechanotransducer channels, then various means for protecting channels from bilayer tension in native membrane (Zhang and Hamill, 2000; discussed in Morris and Homann, 2001) should also be equally widespread and will need more attention.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION CONCLUSION REFERENCES

Stage V-VI Xenopus oocytes defolliculated by collagenase treatment were injected with cRNA for Shaker-IR, that is, with the N-terminal inactivation ball removed (6-46, see Hoshi et al., 1994). The construct, provided by C. Miller, Brandeis University, had an added C-terminal eight-amino acid epitope. The cDNA used to generate RNA was subcloned into a "Melton expression vector," pSP64TM. Immediately before patching, oocytes were briefly shrunk in a hyperosmolar solution (in mM): 200 potassium aspartate, 20 KCl, 1 MgCl₂, 5 EGTA, 10 HEPES pH 7.4) for 3-10 min and manually devitellinized.

Fire-polished pipettes were pulled on a two-stage L/M-3P-A vertical puller (Darmstadt, Germany) using thick-walled borosilicate glass N51A (OD > 1.65 mm, ID 1.15 mm; Garner Glass, Claremont, CA) as described previously (Wan et al., 1999). Changes of pipette pressure were achieved as described previously (Small and Morris, 1994) using two pressure transducers DPM-I (Biotek Instrument, Winooski, VT) arranged in series. Currents were recorded using an Axopatch 200B amplifier and TL1 interface for D/A and A/D conversion (Axon Instruments, Foster City, CA) in conjunction with pClamp 6 software (Axon Instruments). The analog pressure trace was also fed via a gain amplifier into the TL1 interface as described previously (Small and Morris, 1994). A two-electrode voltage clamp (Warner Oocyte Clamp 725C) was used for some tests, as noted. Data were analyzed using pClamp 6, SigmaPlot 5.1 (Jandel Scientific, Corte Madra, CA), and Origin 4.1 (Microcal Software, Northampton, MA).

Patch recordings were made with the following pipette solution (in mM): 140 NaCl, 2 KCl, 6 MgCl₂, 5 HEPES, pH 7.2 according to Hoshi et al. (1994), plus ~100 µM GdCl₃ (see comments below in this section) except where noted. For excised patches the bath ("cytoplasmic") was 140 KCl, 1 CaCl₂, 2 MgCl₂, 11 EGTA, 10 HEPES, pH 7.2, whereas for cell-attached and two-microelectrode voltage clamp recordings, the bath was 115 NaCl, 2.5 KCl, 1.8 CaCl₂, 1 MgCl₂, 10 HEPES, pH 7.2.

Gadolinium solutions are unstable and so were made daily from GdCl₃ · 6H₂O (Aldrich, Milwaukee, WI). For patch clamp work, the salt was weighed and added to 20 ml of pipette solution to give 100 µM GdCl₃. Measuring a small quantity of highly hygroscopic salt probably gave substantial day-to-day variation in pipette solution [Gd³⁺], with further variation developing in the course of seal formation. These variations may contribute to the wide range of apparent V_0.5 we noted for Shaker-IR even in excised patches. For two-microelectrode work, oocyte-to-oocyte variation in bath solution [Gd³⁺] was not an issue because the solution was made in a large quantity and experiments were done over a short period of time using a single batch of solution.

Because our protocols required knowing the foot of G(V), voltage steps were applied to each patch to determine this region. Ideally, the position of G(V) on the voltage axis the should be characteristic from patch to patch, but in practice it was remarkably variable, presumably due to pipette-to-pipette variations in [Gd³⁺]. Oocyte-to-oocyte variations in the surface chemistry of the plasma membrane (and hence variable charge shielding by Gd³⁺) may also have been a factor.

Gigaohm seals (to yield cell-attached patches) were made as previously described (Vandorpe et al., 1994) and excised patches were formed by quickly pulling away from the oocyte once a seal had been achieved. For cell-attached patches, the assumed V_rest was 50 mV; to indicate that membrane voltages for cell-attached patches are approximate, they are preceded by "~" in the Results.

Membrane tension in patches was increased as previously described (Wan et al., 1999) using negative (suction) or positive (blowing) pressure. The pipettes (2-4 M) and pressures (usually <40 mmHg) were well within the usual range for studying MS channels (Sachs and Morris, 1998); because the membrane tension generated by a given pressure varies between patches, direct comparisons and dose-responses were made using within-patch data. As we were principally monitoring for the occurrence (or not) of mechanoresponses, not quantifying sensitivity changes as we did for native SA channels in situ (Small and Morris, 1994), patch sealing/mechanostimulation history was not tracked.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION CONCLUSION REFERENCES

Shaker-IR in the presence of Gd³⁺

Voltage steps that elicited small leak and capacitative currents in patches from uninjected oocytes (Fig 1 a) elicited typical I_Sh currents (Hoshi et al., 1994) in Shaker-IR cRNA-injected oocytes (Fig. 1 b). I_Sh was outward and noninactivating for hundreds of milliseconds and showed voltage-dependent activation and kinetics (S-shaped activation time course) (Fig. 1 b). With long, large depolarizing pulses (2 s or more; not shown) slow "C-type" inactivation (Hoshi et al., 1991) became evident.

View larger version (26K): [in this window] [in a new window]   FIGURE 1   Shaker-IR currents (ISh) in Xenopus oocytes and effects of Gd3+ on ISh. Currents elicited by depolarizing steps (110 to +70 mV in 10 mV increments, Vhold = 100 mV, 1 s between successive steps) from inside-out excised patches of (a) control (uninjected), and (b) Shaker-IR cRNA-injected oocytes (asterisk below a 0 mV current trace). Electronic compensation was used to minimize initial capacitative transients. (c) Conductance-voltage (G(V)) curves from two-microelectrode voltage clamp for a Shaker-IR injected oocyte with control (), 100 µM Gd3+ (), and wash-control (*) solutions perfused into the bath. (d) Normalized version of c. The solid lines are fits to a Boltzmann equation.

Gadolinium (Gd³⁺) used at ~100 µM fully blocked the endogenous MS nonselective cation channels (Yang and Sachs, 1989). Although Gd³⁺ does not abolish TREK-type SA K⁺ channel currents (Small and Morris, 1995), it has diverse actions, so 100 µM Gd³⁺ effects on I_Sh were characterized by two-microelectrode clamp. Steady-state I_Sh (V_m) with or without 100 µM Gd³⁺ data were used to obtain G(V) curves (Fig. 1 c), which were fit to a Boltzmann equation, then normalized (Fig. 1 d) to the fitted G_max. Gd³⁺ significantly decreased G_max to 62 ± 9% (n = 9 oocytes, p < 0.01, paired t-test). G_max recovered completely (e.g., Fig. 1, c and d) upon Gd³⁺ washout. Gd³⁺ also significantly and reversibly shifted half-activation by +22 ± 3 mV (from 19 to +3 mV) and decreased the slope factor by 6 ± 2 mV (from 14 to 20 mV) (n = 9; p < 0.01, paired t-tests) without fundamentally altering the sigmoid voltage-dependence of the channels (e.g., Fig. 1 d).

Comparable trivalent lanthanide effects are reported for mammalian Kv1.1 expressed in Xenopus oocytes (Tytgat and Daenens, 1997) and Gd³⁺ inhibits voltage-dependent Na⁺ and K⁺ currents in Xenopus laevis axons (Elinder and Århem, 1994). For the axonal K⁺ current, 60 and 200 µM Gd³⁺ cause 18 and 26 mV rightward shifts, respectively, comparable to our 22 mV shift with 100 µM Gd³⁺ for Shaker-IR.

Endogenous MS channels

Because of the worry that residual MS current through endogenous SA cation channels might contaminate Shaker-IR records during stretch, we briefly note, for the solutions and recording/stimulating pipettes used in Shaker-IR studies, the characteristics of the endogenous SA channels in the absence of Gd³⁺ and with no Shaker-IR expressed. Pressures of 5 to 30 mmHg activated the channels (the larger the patch, the smaller the pressure required), reversal was near 0 mV, and single-channel conductance was ~30 pS at V_rest (~50 mV). Oocyte patches held at or below V_rest exhibited low-frequency spontaneous unitary inward current events (flickery bursts seldom exceeding 10 ms) whose frequency increased reversibly (Fig. 2 a) with suction. This sustained P_open (open probability) change during high membrane tension represents "classical" MS channel behavior. When both negative (suction) and positive (blowing) pressures were applied to the same patch, each activated the endogenous SA channels (blowing tends to break seals, so suction was routinely used). We did not use "gentle patches" and accordingly did not observe adaptation (Hamill and McBride, 1997).

View larger version (36K): [in this window] [in a new window]   FIGURE 2   Endogenous SA cation channel events and SA events from a Shaker-IR-injected oocyte. (a) Single-channel recording of endogenous MS cation channel events from a control oocyte (cell-attached patch; no Gd3+; Vm = ~50 mV) before, during (including expanded section), and after the indicated pipette suction. (b) Cell-attached patch-clamp recording (Vm = ~20 mV) from Shaker-IR-injected oocyte in the presence of Gd3+, showing steady-state outward unitary current events, then sustained (and reversible) stretch-activation during pipette suction.

For all oocyte batches used for Shaker-IR data, cell-attached or excised patches from control oocytes were tested with 100 µM Gd³⁺ in the pipette, using a range of voltages and suction pressures. Neither spontaneous nor suction-induced ionic current events were observed (n > 20).

Stretch and Shaker-IR

Shaker-IR patches were stretched with 100 µM Gd³⁺ in the pipette and I_Sh ( = iNP_open, where i = unitary current, N = number of channels in the patch) was recorded. Initially we had not predicted that stretch would affect I_Sh. Rather, we had made Shaker-ankyrin fusion constructs to study spectrin-channel-tension interactions (to test the Guharay-Sachs model of MS channels; Gu et al., 1997). Shaker was merely the control. Open and closed conformations are, by definition, equally unstable at the midpoint of the G(V) curve. Our preliminary trials were geared to midregion voltages so that either SI (reduced iNP_open) or SA (increased iNP_open) effects on I_Sh would be evident. The outcome was puzzling. Stretch was not benign, but it had no consistent effect; some patches showed increased steady-state I_Sh, some showed a decrease, and some showed increased low-frequency I_Sh noise with no change in mean current.

Near the foot of G(V), however, with Shaker-IR channels mostly closed yet somewhat destabilizedi.e., exhibiting a detectably non-zero P_opena different story emerged. There, stretch consistently elicited responses like classical SA channels, as in Fig. 2 b. We should mention that, having set gains for the mid-G(V) region, we nearly overlooked such responses. Unitary Shaker-IR currents are poorly resolved because sampling frequency was for macroscopic I_Sh, but outward events are evident. Prestretch P_open was low (<0.001 based on the G(V) relation and unitary conductance), but with stretch, NP_open reversibly increased ~20-fold. Because endogenous SA currents were blocked and could not in any case have carried outward current at 20 mV, the SA events were attributed to the recombinant Shaker-IR channels. Shaker-IR-injected oocytes showed SA outward currents near the G(V) foot in all patches tested under similar conditions (n = 18).

Stretch at the foot of the Shaker-IR G(V)

We therefore reexamined stretch effects at the foot of G(V), using voltage-step-plus-stretch protocols to get around a drawback evident in Fig. 2 b, namely the lack of the Shaker-IR kinetic signature. Generating full current families before, during, and after suction needed at least 2 min per family, during which time rupture was likely. Instead, a truncated approach involving a pair of voltages ("step pair") was used. Patch-to-patch variations in the position of G(V) associated with using Gd ³⁺ (see Methods) were dealt with by locating the foot region of G(V) before applying the step pair protocol. A pressure in the range 5 to 20 mmHg was chosen, aiming for a substantial but nonlytic membrane tension (there is an "art" to this; pipette tip size, sealing behavior, and membrane capacitance are the main gauges). The duration of stretch was <10 s. V_command for a step pair protocol near the foot of G(V) is illustrated in Fig. 3 (top right). The step 1 voltage (duration 150 ms) was chosen to minimally activate I_Sh (usually 20 mV, more rarely 10 mV or 0 mV, depending on G(V) for the patch) and step 2 (also 150 ms) was 10 mV more depolarized. V_hold between step pairs was 100 mV. Step pairs were repeated six times (1 s between pairs) before, during, and after stretch to test for repeatability and/or time-dependent trends.

View larger version (30K): [in this window] [in a new window]   FIGURE 3   Examining SA of Shaker-IR by step pair experiments. The top right "battlement" pattern (6 step pairs labeled 1, 2) shows the Vcommand used to elicit the currents illustrated for excised patches of (a) one control and (b)-(d) three Shaker-IR-injected oocytes. Pipette solutions had 100 µM Gd3+. Vhold was 100 mV and steps 1, 2 were 10 mV apart, as described in the text. Capacitance was compensated before an experiment, then not altered. The protocol was run before, during, and after membrane tension was elevated by pipette suction (solid bar); pipette suction in mmHg is indicated for (a) and (b). In (a) and (b) gaps correspond to ~1 s, during which the data buffer was emptied and the next run initiated (Vhold and pressure were maintained); 100 pA and 60 pA correspond to (a) and (b), respectively. (c) and (d) are higher-resolution recordings (unitary events are evident in (d)) illustrating ISh elicited by step pairs before, during, and after stretch (bars: 15 and 20 mmHg in c and d, respectively); one of the six pairs is illustrated per condition for c, several in d (illustrating large stochastic variation at low NPopen). In c and d most of the current at Vhold is omitted (hence the gaps).

Current traces for this protocol were similar for excised and cell-attached patches. In control patches (n = 15) (Fig. 3 a) the step pairs elicited small leak currents and capacitative transients. As expected, the second capacitative transient (upward deflection) of the step pair, associated with the 10-mV step, was smaller than the first and third (upward and downward deflections, respectively). Neither leak nor capacitative currents, inspected at high resolution, changed with suction. This indicates a fixed membrane area. Control patch and Shaker-IR patch runs were interspersed, with at least one control per batch of oocytes. Occasional tests with suction that would rupture most excised patches (>40 mmHg) showed that even this did not detectably alter capacitative or leak currents. The controls indicated a) no interference from endogenous SA channels and b) suction stimuli did not enlarge the membrane area in the patch.

With Shaker-IR patches (e.g., Fig. 3, b-d), the step pairs elicited outward I_Sh (at step 1, I_leak was often distinguishable from I_Sh). Currents elicited by the step pair corresponded to "family members" below the asterisk in the full family of Fig. 1 b with I_Sh showing recognizable activation kinetics most clearly at step 2. With stretch, step 1 and 2 currents increased reproducibly and reversibly (n = 22 patches) (Fig. 3 b). No time-dependent trends were noted over the six repeats (e.g., Fig. 3, b and d). In most cases (n = 15) the procedure was repeated with the step pair shifted by 5 or 10 mV and again, stretch-augmented I_Sh was evident.

These step pair data were quantified as in Fig. 4. For each patch, mean currents (n = 6, ±SD) before (0 mmHg, i.e., "control"), during ("stretch") and after ("release") suction were determined (steady-state I_Sh or, if current was still rising at 150 ms, maximal I_Sh, or, if unitary currents were obtained, time-averaged I_Sh). Fig. 4 a illustrates a patch whose below-average response allows for display of step 1 and step 2 data on a common y axis. For between-patch comparisons, within-patch data were normalized (Fig. 4 b) as this partially deals with differences in channel density, patch size, and tension. Between-patch comparisons over the whole data set can be summarized thus: for 17 of 17 patches tested using the step pair 20 mV, 10 mV, suction (20 mmHg) increased I_Sh at both voltages (paired t-tests, p < 0.01). The fold-increase of I_Sh (I_test/I _control) was significantly greater for step 1 than step 2 (3.8 ± 0.4 for step 1 and 2.6 ± 0.3 for step 2 (mean ± SE, n = 17, p < 0.01)). Thus, in all patches, stretch was disproportionately more effective near the foot of G(V) than 10 mV more depolarized. Fig. 4 c plots the outcome for all 17 patches, with the stretch-current (not the control) as unity, thereby emphasizing that I_Sh was an "SA channel" over this voltage range.

View larger version (25K): [in this window] [in a new window]   FIGURE 4   Voltage-tension interactions; analysis of step pair data. (a) From a sample patch (different patch, but raw data as illustrated in Fig. 3, b--d), steady-state ISh (mean ± SD, n = 6) elicited by step 1 and by step 2 for pipette pressures 0 and 15 mmHg, before (control, "0"), during (stretch, "15 mmHg"), and after (release, "0") mechanical stimulation. Vcommand for steps 1 and 2 was 20 mV and 10 mV, respectively. Here and elsewhere, steady-state ISh was obtained by cursor measurement (Clampfit from Axon Instruments' p-Clamp). The baseline was reset to the leak current level where this was evident. (b) Mean test ISh (same data as (a)) replotted as ratio of the control ISh for step 1 and step 2. (c) Among-patches comparisons of the relative extent of stretch-activation in 17 excised patches; here, within-patch values have been normalized by setting the mean "ISh stretch" in each patch to unity. For all 17 patches, the step pair was 20 mV, 10 mV, and in most patches stretch was elicited by 10 mmHg suction. Paired t-tests (step 1 [Istretch/Icontrol] versus step 2 [Istretch/Icontrol] in each patch) showed that the fold-increase nearer the foot of G(V) exceeded that for 10 mV more depolarized (p < 0.01).

Higher pressures applied at the end of these experiments consistently elicited larger I_Sh increases, indicative of a dose-dependent effect of stretch. Fig. 5 shows dose-dependence for an excised patch subjected to a broad range of pressures. Current amplitudes (Fig. 5 a) increased with pressure until patch rupture occurred just after 40 mmHg. Fig. 5 b plots fold changes. Patches from three oocytes withstood this treatment, each showing that when the pre-stretch P_open(V) values were low but detectably > 0 (i.e., near the foot of G(V)), stretch increased iNP_open in a dose-dependent manner. In Fig. 5, for step 1, the smallest stimulus (5 mmHg) almost doubled I_Sh and the largest (40 mmHg) elicited a >10-fold increase. As with the 17 patches above, stretch activation was greater on a percent basis at step 1 than step 2 (Fig. 5 b), so over a wide range of tensions, stretch was disproportionately a better I_Sh activator near the foot of G(V) than 10 mV more depolarized. For step pair data, error introduced if I_Sh failed to reach steady-state during a step would underestimate, not overestimate, this disproportion because I_Sh activation speeds up with depolarization. Likewise, error from leak currents would contribute to an underestimate. Stretch's disproportionate effect at lower P_open (V) values is good evidence that it affected Shaker-IR gating. Had stretch increased I_Sh not through tension effects on gating (i.e., on P_open), but by increasing the quantity of Shaker-IR-bearing membrane in the patch (hence increasing N), then I_Sh at the two voltages would have increased proportionately with suction.

View larger version (26K): [in this window] [in a new window]   FIGURE 5   Within-patch dose-response over a wide pressure range. (a) ISh traces for a step pair experiment on an excised patch with successively increasing membrane tension (pipette pressure in mmHg is indicated along the top). The step pair voltages (10 mV, 0 mV) were near the foot of the G(V) for the patch. The step pair was applied once/pressure. (b) The currents in (a) plotted as a dose-response using the same procedure as in Fig. 4 b (except that the step pair was applied only once at each pressure rather than six times). Since step 2 saturated the current amplifier at 40 mmHg, the plot stops at 35 mmHg.

Stretch-activation plus stretch-inactivation of I_Sh

Preliminary stretch tests at high P_open (like Fig. 2 a, but for very depolarized voltages) did not yield increases of outward current. Had stretch enhanced nonspecific leak, had it caused "breakthrough" SA cation channel currents, or had it generated additional current through Shaker channels, outward current would have resulted. Instead, the effects of stretch at large depolarization looked like a decrease (SI) in I_Sh. We therefore examined stretch on I_Sh at both extremes of prestretch P_open (V), using a step protocol so that both voltages extremes were tested during a given mechanical stimulus. For the three patches that withstood a wide range of pressures, effects of stretch near the foot and head of G(V) are shown in Fig. 6. Steps 1 and 2 (60 mV apart) were chosen after locating G(V) for the patch. Fig. 6 a illustrates current traces (patch 1) at two test pressures. To reduce the possibility of patch rupture, step pairs were applied once (not six times). Suction was held constant for 2 s at each level, the step pair was applied, pressure was increased, and so on. For each patch, where the high prestretch P_open(V) plots (step 2 for each patch) show a dose-dependent SI trend (I_Sh decreasing as pressure increases), simultaneously the low prestretch P_open(V) plots show a dose-dependent SA of I_Sh. In other words, in immediately adjacent 150-ms periods (with the stretch stimulus unchanged), the population of Shaker-IR channels in any given patch exhibited dose-dependent SA and SI.

View larger version (21K): [in this window] [in a new window]   FIGURE 6   Within-patch pressure dose-responses for "foot-to-head" step pairs. (a) Raw step pair currents from patch 1 (see (b)) at two different test pressures. (b) Steady-state ISh from three excised patches using 60-mV step pairs at the indicated voltages. ISh magnitudes differed greatly 60 mV apart and so are plotted separately with different y axis gains. The particular step pair for each patch is indicated (, step 1; , step 2).

Possible sources of artifacts

Mechanical stimulation of soft biomaterial is unavoidably messy. Although patch aspiration is the best method for transiently increasing plasma membrane tension, it is a poorly controlled stimulus, making it imperative to probe the robustness of stretch effects. One concern is that after gigaohm seal formation residual pressure of unknown value may persist (Morris and Sigurdson, 1989). If so, application of a small pressure step (say, in the ±1-10 mmHg range) might either increase or decrease the net pressure by opposing the residual value. Accordingly, it was reassuring that the dual outcomes of Fig. 6 (i.e., SA-plus-SI) were observed over a large pressure range. Furthermore, for negative pressures that would have far exceeded any inadvertent positive pressure, exactly the same suction stimuli that augmented I_Sh at one voltage (step 1) diminished it at another (step 2). This essentially eliminates the possibility that stretch effects on I_Sh resulted from transient membrane area increases.

It was also crucial to establish that stretch effects ascribed to Shaker-IR were not in fact "contamination currents" from inadequately blocked endogenous SA cation channels. Two measures to rule this out have been mentioned: noninjected oocytes were examined and SA of Shaker-IR was studied near V_rev of the endogenous SA channels. A third approach is a direct demonstration that even without Gd³⁺ (or with "break-through" of the Gd³⁺ block), endogenous SA currents would not explain effects seen in Shaker-IR patches. Fig. 7 a shows the endogenous SA cation currents (control oocyte, no Gd³⁺) reversing near 0 mV. As is common, a secondary mechanical effect is evident (baseline SA channel activity did not return immediately to the prestretch level). The next traces (Fig. 7, b and c) are for Shaker-IR-injected oocytes over a voltage range. Steady-state voltage-gated I_Sh was established by stepping (not shown) from 100 mV to a depolarized voltage for 10 s, during which time suction was applied twice. In Fig. 7 b, suction increased the steady-state current at the least depolarized voltage (~30 mV), produced an equivocal increase at a voltage 10 mV more depolarized, then reproducibly decreased the current 20 mV beyond that. Unequivocal stretch-induced changes were associated with the voltage "extremes" (foot and head of G(V)) rather than the intermediate voltages (e.g., +20 mV in Fig. 7 c). At intermediate voltages, SA and SI effects on current were presumably sufficiently balanced across the population of channels that they tended to cancel.

View larger version (28K): [in this window] [in a new window]   FIGURE 7   Suction effects on ISh and endogenous SA currents. (a) Endogenous SA Ication elicited by suction of 20 mmHg (bars); cell-attached patch, noninjected oocyte, Vm as indicated, no Gd3+. (b) Cell-attached patch clamp trace, Gd3+ in the pipette, Shaker-IR injected oocyte. Pipette pressure traces are below the current traces; downward deflection indicates negative pressure (suction), Vm as indicated. In the last current trace, slow (C-type) inactivation of ISh is evident. Note each current trace has its own scale. (c) Excised patch clamp trace, Gd3+ in the pipette, Shaker-IR-injected oocyte, suction (15 mmHg) applied as indicated by bars. Again, each trace has its own scale. Pronounced C-type inactivation of ISh is evident in the last two traces. (d) Excised patch, with zero Gd3+ in the pipette, Shaker-IR-injected oocyte (traces as in (b), except that current is shown at 100 mV, then +50 mV, then 100 mV, and the pressure and current traces overlap).

Currents through inadequately blocked endogenous MS channels could not have produced the stretch-induced pattern of changes in Fig. 7, b and c. Voltage-dependent I_Sh was exclusively outward at all test voltages, whereas the endogenous SA channel exhibited inward currents when V_m < V_rev and outward currents when V_m > V_rev (Fig. 7 a). Given their particular V_rev values (~0 mV versus ~80 mV), combining endogenous SA currents with steady-state I_Sh(V) would have produced a strong (mis)impression of SI of I_Sh at hyperpolarized potentials and of SA of I_Sh at depolarized potentials. Precisely the opposite was observed (Fig. 7, b and c).

For ENaC channels reconstituted into bilayers, release from calcium block during application of hydrostatic pressure was misconstrued as MS gating (Ismailov et al., 1997). For Shaker-IR in oocytes patches, SA and SI of I_Sh occurred in the absence of Gd³⁺ and therefore were not release-from-block effects. A trace showing SI of I_Sh in a Shaker-IR patch in the absence of Gd³⁺ (Fig. 7 d) illustrates this point. The endogenous SA channels probably added a component of MS outward current during the two stretch stimuli, but obviously this would not explain a net decrease in outward current during stretch. Although we cannot rule out that stretch/Gd³⁺ interactions at the membrane modified the responses of Shaker-IR to stretch, Gd³⁺ did not cause them.

Tension versus membrane curvature

To determine whether membrane tension was the relevant mechanical stimulus during SA and SI of I_Sh, excised patches at a fixed voltage were subjected to suction, then blowing. Fig. 8 shows that suction and blowing elicited qualitatively the same response. Where suction increased I_Sh so did blowing, and where suction decreased I_Sh so did blowing. Thus, tension, independent of the sign of membrane curvature, was the critical mechanical stimulus. The other optionthat both suction and blowing increased I_Sh by reversibly augmenting membrane area in one voltage range while, in another range, both suction and blowing decreased I_Sh by reversibly diminishing membrane areais wholly implausible.

View larger version (20K): [in this window] [in a new window]   FIGURE 8   Suction and blowing have equivalent effects. ISh in two excised patches (a and b),Vm as indicated, and patch tension generated using both concave patch (suction) and convex patch (blowing) configurations. The pressure traces are shown below each current trace. C-type inactivation is evident in (b).

Stretch-activation I_Sh at the single-channel level

Finally, effects of tension on Shaker-IR were examined at the single-channel level, typically using patches smaller than for macroscopic I_sh. A greater fraction of channels/patch responded to stretch in single-channel studies than in the macroscopic current studies. It was harder forming seals with the smaller patches needed for single-channel recording, so these patches were probably quite traumatized. Because membrane trauma renders various channels more, not less, susceptible to stretch (TREK-type channels, Small and Morris, 1994; Wan et al., 1999; NMDA channels, Paoletti and Ascher, 1994; Na⁺-channel -subunits, Tabarean et al., 1999) it may have enhanced "recruitment" by stretch in the single-channel studies. Applied to patches, Laplace's law dictates that the smaller the patch radius of curvature, the greater the pressure required for a given tension, hence the larger pressures used in this section.

Bearing in mind the G(V)-shifting and G-decreasing effects of Gd³⁺ on I_Sh (Fig. 1 c), single-channel characteristics at resting membrane tension were as expected for Shaker-IR (Hoshi et al., 1994). Because patches had multiple channels, unitary events were best resolved at membrane potentials and/or tensions that yielded small P_open values. Fig. 9 a illustrates single-channel data from an excised patch with a low prestretch P_open. Increasing stretch intensity (pressures 0 to 20 mmHg) increased NP_open, presumably via P_open. The effect was reversible upon release (see bottom trace, 0 mmHg). The pressure-family of current amplitude histograms (Fig. 9 b) indicates that stretch-sensitive current was associated with Shaker-IR, not, say, breakdown noise; equal intervals of ~0.6 pA between baseline and the subsequent amplitude peaks indicate that the same class of channel contributed at low, intermediate, and high current levels. This also confirms that membrane stretch did not increase Shaker-IR iNP_open by changing i, the unitary current amplitude (as might occur if, say, stretch reduced the Shaker-Gd³⁺ interaction that diminishes G_Sh).

View larger version (45K): [in this window] [in a new window]   FIGURE 9   Effect of increasing suction on single ISh and on the Popen of Shaker-IR. (a) Five-second current trace excerpts, excised patch, pipette pressure in mmHg as indicated. Each pressure was applied for 20 s, then switched to the next level. The last trace was after suction was released to 0 mmHg from 20 mmHg. Vm = 30 mV. (b) All-points current amplitude histograms for the same data set as in (a) (and same as open circles, (c)). On the current amplitude axis, "0 pA" (no open channels) is the mean of the Gaussian baseline current (peak truncated in some cases). (c) Dose-response data as in (a) for six excised patches. Each point represents an average NPopen computed from a continuous 20-s recording. Vm values were 30 mV (); 30 mV (); 20 mV (); 10 mV (); 0 mV (); and 5 mV(). Pressure was increased progressively without release between successive pressures, except for one patch (), which was done in reverse order. Inset: a U-shaped dose-response resulted when patch , whose activity was low but non-zero at 0 mmHg, was subsequently subjected to a series of positive and negative pressures (the fit is for display purposes and implies no precise function).

Single-channel dose-responses for this and five other patches (Fig. 9 c) were fit to sigmoids that have been normalized for display purposes. Each dose-response has a steeply rising activity (pressure) region, typical of most SA channels. The total number of stretch-responsive Shaker-IR channels in the patch was unknown because all patches ruptured before showing convincing saturation (patches ruptured after the highest pressure for which data are given). We therefore have not extracted Boltzmann parameters (slope factor and half-maximum P_open) from the curves.

Wide variability in the location of SA curves along the pressure axis, as in Fig. 9 c, is typical for SA channels (Gustin et al., 1988; Sachs and Morris, 1998; Vandorpe et al., 1994) because tension (dependent on the membrane's radius of curvature) is the operative x axis variable, not applied pressure. Probably the three leftmost curves represent larger patches with larger radii of curvature.

For one of the six patches (inset, Fig. 9 c), both positive and negative pressures were applied. For this excised patch, P_open at 0 mmHg was marginally above zero. As mentioned for macroscopic I_Sh, pressure at both signs, "blowing and suction," should elevate membrane tension, with membrane curvature directed out of or into the pipette tip, respectively. The resulting SA dose-response, a classical U-shape, confirms at the single-channel level that increased membrane tension underlies I_Sh responses to changes in pipette pressure.

Stretch-inactivation of I_Sh at the single-channel level

SI of I_Sh, too, was evident at the single-channel level, as illustrated (Fig. 10) for an excised patch where suction, applied twice, reversibly "turned-off" most of the voltage-activated Shaker-IR channels in the patch. Post-suction recovery was noninstantaneous partly because decay of applied pressure on release of suction was slower than pressure onset (see typical pipette pressure traces in Figs. 7 and 8), but beyond this instrumentation effect, asymmetric responses (faster onset than offset) are common for MS channels (e.g., see Fig. 7 a, endogenous channel) including a neuronal SI K⁺ channel (Small and Morris, 1994; Fig. 2 b). This may relate more to patch mechanics than channel properties per se. Rapid, reversible, reproducible switching between multiple- and single-channel recording of I_Sh as seen in Fig. 10 supports the earlier argument (based on macroscopic data alone, as in Figs. 6-8) that suction-induced SI of I_Sh represented a tension-induced decrease in Shaker-IR P_open.

View larger version (26K): [in this window] [in a new window]   FIGURE 10   Reversible stretch-inactivation of ISh at high Popen. ISh from a depolarized (Vm = +20 mV) excised patch. When suction was applied and released twice, ISh changed reversibly from a macroscopic (multichannel) to a single-channel trace. Below, this trace is seen at expanded time resolutions, with locaters provided. The trace is interrupted in the second line. Channel activity turned off via discrete current steps (third line). Suction is indicated (bar) except in the last line, when it was continuous.

Single-channel SI I_Sh dose-response data are illustrated in Fig. 11 a for an excised patch at +20 mV and a series of increasingly negative pressures. By 50 mmHg, stretch completely silenced the voltage-activated events. Next, at 0 mmHg (penultimate trace), the patch was stepped down by 30 mV to 10 mV, i.e., the foot of G(V) for the patch, yielding a low level of channel activity. The pipette pressure was again increased sequentially; a trace at 50 mmHg is illustrated. Here, at the foot of G(V), SA is observed (as in Fig. 9 a). Unequivocally, SI and SA of Shaker-IR can be elicited from the same patch. If it is striking to compare +20 and 10 mV traces for 0 mmHg, it is spectacular to compare them for 50 mmHg; 50 mmHg completely abolished channel activity at +20 mV but had the opposite effect at 10 mV, producing intense channel activity from a near-silent patch. Thus, depending on prestretch P_open, Shaker-IR channels can be either SA channels or SI channels.

View larger version (49K): [in this window] [in a new window]   FIGURE 11   High versus low prestretch Popen (Vm), single-channel data. (a) Five-second current traces from 20-s continuous recordings, collected in the sequence shown. After the series at Vm = +20 mV, the patch was clamped at Vm = 10 mV and taken from 0 to 60 mmHg in 10 mmHg steps (not all shown). The two higher-resolution fragments of 10 mV current trace (bottom) are provided to allow for comparison of current amplitudes during periods of exceptionally high activity at 0 mmHg and of exceptionally frequent transitions at 50 mmHg. (b) SA and SI dose-responses for this patch at Vm = +20 mV () and Vm = 10 mV () using normalized NPopen. The fits are for display. (c) Single-channel dose-responses (protocol as in Fig. 9, with NPopen calculated from 20-s recordings) from five excised patches at depolarized (elevated NPopen) voltages: +60 mV(*); +40 mV (); +60 mV (); +20 mV(, same as in (b)); +60 mV (). NPopen values were normalized for each patch.

As an aside, the 30 mV increase in driving force at +20 mV versus 10 mV probably accounts for most of the increase in unitary current, but relief of fast block of the Shaker-IR channels by Gd³⁺ at more depolarized potentials may also contribute.

Fig. 11 b uses the entire 20-s record traces from the same patch (Fig. 11 a shows 5-s excerpts plus two additional 800-ms traces) to plot normalized NP_open as a function of pressure, showing that SI of single-channel I_Sh was dose-dependent and that SA occurred over the same pressure range. Obtaining SA or SI responses depended on whether channels were mostly quiescent (SA) or mostly active (SI) before stretch, i.e., whether the holding potential was 10 mV (SA) or +20 mV (SI). The overlapping dose responses for the two voltages corroborates the macroscopic data (Fig. 6), indicating again that SI and SA can occur in the same patch.

Although Fig. 11, a and b was the only patch yielding a full single-channel dose response at both low and high prestretch P_open, single-channel SI dose responses for several patches held at potentials between +20 mV and +60 mV are shown in Fig. 11 c. For between-patch comparisons, dose (pressure)-dependent channel activity is plotted as normalized NP_open. Patch rupture was a problem, though near-zero activity levels were obtained before rupture in three of five patches. Single-channel dose responses (plus the macroscopic ones, Fig. 6) at high prestretch P_open showing an inverse relation between Shaker-IR NP_open and applied suction signify that Shaker-IR can be considered an SI channel and an SA channel.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION CONCLUSION REFERENCES

Shaker-IR: an MS channel "ordinaire"

The early view (Guharay and Sachs, 1984) that MS channels are specially designed physiological mechanotransducers has been controversial (Morris, 1992; Sachs and Morris, 1998), and except for prokaryote osmotic safety valvesthe MscL channels (see Sukharev et al., 2001)evidence that particular MS channels are mechanotransducers remains threadbare (e.g., Sachs et al., 2000). MscL, strikingly, is designed to be stretch-insensitive until tension becomes near-lytic (Batiza et al., 1999). In contrast, Shaker-IR routinely responded to stretch at comfortably nonlytic tensions, as do neuronal SIK channels (Morris and Sigurdson, 1989) and SAK (TREK-type) channels (Wan et al., 1999; Patel et al., 1998) and oocyte SA channels. None are mechanotransducers (e.g., Steffensen et al., 1991; Morris and Horn, 1991; Wilkinson et al., 1998). Its tension/voltage interactions aside, Shaker-IR behaved as a "standard" MS channel during patch clamp (macroscopic and unitary patch currents), showing dose-dependent, sustained, reversible responses, with membrane tension the relevant mechanical stimulus, and with stretch able to produce a one-to-two orders of magnitude NP_open. Several other voltage-gated channels respond to membrane stretch. Smooth muscle calcium currents increase reversibly with inflation (Langton, 1993). Ca²⁺-activated large-K⁺ channel can be stretch-activated independent of calcium (Lee et al., 2000). The skeletal muscle Na⁺ channel -subunit, expressed in Xenopus oocytes, does not show reversible effects as seen in Shaker, but its anomalously slow inactivation irreversibly converts to normal inactivation with stretch (Tabarean et al., 1999). Although Shaker-IR's reversible responses to stretch are more akin to those of TREK than the irreversible response of the Na⁺ channel, it would seem absurd to invoke sequence similarities or differences to explain any of this.

The biology of mechanosusceptibility

Shaker-IR was mechanosusceptible under profoundly nonphysiological conditions. Even minus its cytoplasmic S3-S4 linker (Tabarean et al., 2000), Shaker-IR retains mechanosusceptibility. Clearly this is a deeply embedded biophysical trait, but what might it signify in biological, i.e., in evolutionary terms? Here are two opposing evolutionary hypotheses to account for the observation that diverse unrelated channel types, Shaker-IR included, are MS channels: 1) MS channels respond to bilayer tension because evolutionary design provides them with specialized mechano-gating regions and/or a special global structure that renders them susceptible to bilayer tension; 2) MS channels respond to bilayer tension because it has been impossible, undesirable, and/or unnecessary to eradicate protein characteristics whose side effect is mechanosusceptibility. The two likeliest "side effects" are a) at least one of the channel's multiple conformations has a different in-plane area than the others; and b) the channel has domains that are deformable by sublytic bilayer tension. Options a and b are not mutually exclusive.

We prefer hypothesis 2 for Shaker-IR. Increasingly, the idea that mechanosusceptibility is hard to design out of membrane proteins seems more compelling than the idea that mechanosusceptibility is hard to design in to membrane proteins. Ironically, the ample selection of MscL mutants that produce a channel more mechanosusceptible than the wild type (Batiza et al., 1999) make the same point. If 2a and/or 2b apply, it is easy to suggest why natural selection has failed to eradicate mechanosusceptibility from many nonmechanotransducer ion channels. For 2a, restricting a membrane protein to fixed-area conformation changes would greatly restrict its range of molecular motions. For 2b, pervasive internal cross-linking that rendered a channel nondeformable would exact a heavy cost in the guise of low frequency of thermal transitions, and hence long response times to significant stimuli.

Conformation area and/or deformability

Shaker-IR behaved as an MS channel in excised oocyte plasma membrane. Although MscL requires only an artificial bilayer for SA gating (Sukharev et al., 2001) the SA-plus-SI behavior of Shaker-IR may require a more complex environment. Excised oocyte membrane is a potpourri of diffusible and anchored lipids, transmembrane proteins, and membrane skeleton remnants and, possibly, lipid subdomains. Below we consider one model that invokes plasma membrane heterogeneity and one that does not.

Consider a two-state (CO) channel in a Hookean bilayer: if state O occupies more in-plane area, it will be favored by high tension because, under stretch, the channel need do less bilayer-compressing work to expand to O. A "big area O" channel would be an SA channel, whereas a "big area C" channel would be an SI channel (see Sachs and Morris, 1998). Shaker-IR exhibits SA and SI behavior, which a two-state area model cannot accommodate, but addition of one additional state (e.g., CCO or COC) should, in principle, and Shaker-IR has many kinetically distinguishable kinetic states, mostly closed (Hoshi et al., 1994; Bezanilla, 2000). Unless all occupy precisely the same in-plane membrane area, Shaker-IR should exhibit some degree of mechanosusceptibility. Although simultaneous fluorescence/gating charge probes of transitions among Shaker closed states strongly suggest that gating motions have vectors in the bilayer plane, the models depicting the moving parts have not, to date (see Fig. 16 in Bezanilla, 2000) specified net area differences. However, it remains to be established whether Shaker conformations are indeed size-invariant in the plane of the bilayer.

If Shaker-IR deforms uniformly under tension, the deformed channels could have abnormal transition rates, hence SA or SI should result. If instead, bilayer heterogeneity renders a macroscopic membrane tension nonuniform at the level of individual channels, SA-plus-SI behavior might result in the following way: during stretch, two channels could experience different microforces such that one moved to the left (SI) and the other moved to the right (SA) along multistate kinetic schemes like those below. The resulting dual effects would cancel out at some voltages, but not at others. Before elaborating on tension/voltage interactions, however, we note a fundamental difference between -area versus deformation models. A central assumption of the former is that channels have hard discrete-sized states, whereas an assumption of the latter is that channels (or some part of them (Zaccai, 2000)) are soft enough to be reshaped by tensile forces acting in the plane of the bilayer.

Stretch/voltage interactions in Shaker-IR

A simplified kinetic scheme for voltage dependence in Shaker-IR is
Voltage-dependent transitions taking Shaker-IR from Cn to C1 involve rearrangements of the voltage sensor (to a first approximation, the S4s (Bezanilla, 2000)). What if C1 (all S4s repacked) occupies more area than any of the other states? Might that explain our data (i.e., net SA and net SI at slightly depolarized and highly depolarized voltages, respectively)? "Big C1" is probably not the explanation for our observations. If it were, stretch would dramatically increase the duration of within-burst closed events during both SA and SI, and no such effect was seen in single-channel records. Also, we obtained strong SI in cases (e.g., Fig. 11 a) that fell well short of the "O dominates" condition. It is possible but not self-evident that Shaker-IR has a "big intermediate C" (a C less proximal to O) that produced effects such as we observed. Tests for mechanosusceptibility would be worthwhile for bacterial K_CSA (gated by pH rather than voltage), whose transmembrane segments may repack to occupy different areas during gating transitions (Perozo et al., 1998).

Next we sketch a heterogeneous stretch-deformation scheme of Shaker-IR that yields SA plus SI. Individual channels in a heterogeneous bilayer deform differently. Channels not appreciably deformed by stretch would maintain their standard (S) transition rates. Stretch-deformed channels showing a net increase in either forward (F) or backward (B) rates would be designated SA or SI, respectively.
In a patch with 100 Shaker-IR channels, let 80, 10, 10 channels be S, F, B, respectively. If prestretch P_open = 0.01 and if during stretch P_open of channels F and B is 0.1 and 0.001, respectively, then I_Sh (where i = unitary current, say, 1 pA) is



and this constitutes a net stretch-activation. Now consider stretch-inactivation, i.e., an increased "P_closed," where P_closed = 1  P_open. If V_m is changed so that prestretch P_closed = 0.01, comparable channel deformation effects (i.e., 80 unaffected, 10 experiencing 10-fold P_closed increase, 10 experiencing 10-fold P_closed decrease) will yield a net stretch-inactivation of I_Sh. At voltages where pre-stretch P_open = P_closed = 0.5, balanced increases in forward and backward transitions would obscure the microscopic effects of stretch on the steady-state current.

Conti et al. (1984), having analyzed ionic (Na⁺, K⁺) and gating (Na⁺ channel) currents at normal and elevated pressures in squid axon, suggested that a late, rate-limiting activation step not accompanied by easily detected charge movement has a large positive activation volume (volume associated with a reaction's transition state). This could be broadly consistent with a scheme in which voltage-gated channels expand in the plane of the bilayer as they go from C_n to O.

Our underlying assumption has been that what we called stretch-inactivation is really (i.e., "really" in the parlance of voltage-gating) stretch-deactivation. That assumption may, however, be unjustified because in addition to its O and C states, Shaker-IR can also access (albeit on a time scale of >10 s) inactivated (I) states (Bezanilla, 2000). This process is called, in reference to the C-terminus and the pore region, C-type and P-type inactivation. At high P_open Shaker-IR may, as we have assumed, principally re-enter closed (C_n  C₁) states (formally, this would be stretch-deactivation) but our experiments did not rule out the possibility that, instead, at high P_open stretch favors large-area (compared to C_n) C-type and/or P-type inactivated states. In linear schemes, these states would be to the right of and coupled to (Loots and Isacoff, 2000) the "O" state. We note, however, that Meyer and Heinemann (1997) from a thermodynamic analysis of Shaker-IR currents measured over a temperature and pressure range, suggest that the C-type inactivated state occupies a smaller volume than the noninactivated state.

Other MS channels that show both SA and SI gating

MS cation channels in healthy muscle exhibit SA, whereas in dystrophic muscle they exhibit SI (Franco-Obregon and Lansman, 1994). For unknown reasons the dystrophic muscle channels have a significantly higher prestretch P_open than healthy muscle channels (for Shaker, voltage produces such a difference). If the dystrophic situation represents maximal P_open, then stretch could produce SI (or have zero effect), but could not produce SA. Given the different membrane skeletons involved (plus or minus dystrophin), channels may experience microstresses that favor high P_open in healthy membrane and low P_open in dystrophic membrane. In other words, the general factors we invoked for Shaker-IR-prestretch P_open differences coupled with microenvironment-dependent channel mechanics during stress could render the same MS cation channel SA in healthy cells but SI in dystrophic cells.

K_ACh channels are another SA-SI example. One laboratory reported these Kir channels as SA (Pleumsamran and Kim, 1995), and another (using recombinant channels) reported them as SI (Ji et al., 1998). As for muscle MS cation channels and Shaker-IR, this discrepancy might be resolved by reference to the different prestretch P_open prevailing in the two situations. Where they were described as SI channels, the Kir channels were first activated almost maximally (via acetylcholine stimulation of G-proteins), then subjected to stretch. By contrast, the SA version of Kir was at relatively low P_open before stretch was applied.

Mechanosusceptibility and its partner mechanoprotection

An organized membrane skeleton may contribute to mechanoprotection of Shaker in nerve and muscle. Like TREK and NMDA channels (Wan et al., 1999; Paoletti and Ascher, 1994) which are most mechanosusceptible after the cortical cytoskeleton deteriorates, voltage-gated channels in squid axon show evidence of feeling tension when the membrane skeleton is destroyed. Using electron microscopy and electrophysiology plus chaotropic solutions on squid axon, Terakawa and Nakayama (1985) demonstrated that submembranous cytoskeleton dissolves with KCl and KBr, but maintains its integrity with nonchaotropic KF. Inflation of KF-perfused axons has little impact on the action potential or on voltage clamp currents, whereas inflation of KCl- or KBr-perfused axons reversibly depolarizes them and alters their Na⁺ and K⁺ channel currents. Shaker-like channels target preferentially to stiff bilayer microdomains (Martens et al., 2000) which, for the channels they harbor, may be mechanoprotective.

Mechanoinsusceptibility for mechanotransducers?

Shaker-IR showed that it lacks "unshakeable" intrinsic mechanoprotection. For exquisitely sensitive mechanotransducer channels, e.g., hair cell channels (Hudspeth, 1997), what then seems reasonable? Hypermechanosusceptibility? Surely not. Hair cell channels would be better-off inured to bilayer tension fluctuations (unlike Shaker-IR), attentive exclusively to mechanical signals along the tiplink axis. Accordingly, they may be extremely stiff and/or may avoid -area conformations. Expressed heterologously (see Garcia-Anoveros et al., 1998), mechanotransducer channels with these characteristics should "fail" as MS channels.

	CONCLUSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION CONCLUSION REFERENCES

We showed that a molecularly altered voltage-gated channel, Shaker-IR, is an MS channel. Indirectly, this has physiological consequences, as it signifies a need for mechanoprotection. Strip away the overlay of in situ mechanoprotection, and "primitive" mechanosusceptibility emerges. Structure-function studies directed at Shaker-IR's voltage-dependence may inadvertently uncover explanations (e.g., in-plane area changes) for its mechanosusceptibility.