Effects of GsMTx4 on Bacterial Mechanosensitive Channels in Inside-Out Patches from Giant Spheroplasts

Abstract

GsMTx4 is a 34-residue peptide isolated from the tarantula Grammostola spatulata folded into an inhibitory cysteine knot and it selectively affects gating of some mechanosensitive channels. Here we report the effects of cytoplasmic GsMTx4 on the two bacterial channels, MscS and MscL, in giant Escherichia coli spheroplasts. In excised inside-out patches, GsMTx4 sensitized both channels to tension by increasing the opening rate and decreasing the closing rate. With ascending and descending pressure ramps, GsMTx4 increased the gating hysteresis for MscS, a consequence of slower gating kinetics. Quantitative kinetic analysis of the primary C↔O transition showed that the hysteresis is a result of the decreased closing rate. The gating barrier location relative to the open state energy well was unaffected by GsMTx4. A reconstructed energy profile suggests that the peptide prestresses the resting state of MscS, lowering the net barrier to opening and stabilizes the open conformation by ∼8 kT. In excised patches, both MscL and MscS exhibit reversible adaptation, a process separable from inactivation for MscS. GsMTx4 decreased the rate of reversible adaptation for both channels and the MscS recovery rate from the inactivation. These measurements support a mechanism where GsMTx4 binds to the lipid interface of the channel, increasing the local stress that is sensed by the channels and stabilizing the expanded conformations.

Introduction

Specific inhibitors are a vital pharmacological tool to study ion channels because they perturb the transitions between states allowing a detailed examination of mechanisms. While mechanosensitive channels are inhibited by lanthanides and cationic antibiotics, these agents are nonspecific (1,2). The development of a specific reagent for MS channels (3–9), GsMTx4, a peptide from tarantula venom, has begun to shed light on the gating mechanisms for mechanosensitive channels (10) and channel-forming peptides (11). The peptide has nonchiral interactions with channels that appear to alter their lipid boundaries during gating.

GsMTx4 is a 34-amino-acid positive pentavalent peptide that belongs to the Inhibitory Cysteine Knot family (10,12). The structure of the peptide has been solved by NMR and is amphipathic (13). The affinity for lipid membranes demonstrated by measuring binding to large unilamellar vesicles showed that GsMTx4 had a significant affinity for zwitterionic POPC (−6.1 kcal/mol) and anionic 25POPC:75POPG (−8.3 kcal/mol) lipids (14,15). Based on brominated lipid quenching of tryptophan fluorescence measured on analogous Hanatoxin (HaTx) (16) and SGTx (17), GsMTx4 likely penetrates the bilayer to within 8–9 Ǻ of the midplane (15,18).

GsMTx4 has structural similarity with Hanatoxin (HaTx) (19) VSTx1 (20), and SGTx (17) with an exposed hydrophobic (largely aromatic) surface surrounded by a ring of positive and negative charges. However, extracellular GsMTx4 has no effect on voltage-gated channels (21). While all three peptides partition into lipids, both VSTx and HaTx interact with the S3-S4 voltage sensor paddles buried in the lipids (16,22,23). VSTx1 acts on the paddle chimera Kv channel in a manner that depends on the lipid composition and mechanical state of the membrane. This led to the proposition that modifiers of voltage sensitivity are often modifiers of lipid-channel interactions (24–26).

Experiments using D and L enantiomers of GsMTx4 indicated no chiral interactions with cationic stretch-activated channels from astrocytes or gramicidin channels, thereby excluding the lock-and-key recognition mechanism and implying long-range interactions (11). As predicted by the Poisson-Boltzmann equation, GsMTx4, with a valence of +5, slightly reduced cation conductance for currents originating from the same side as the peptide. This suggests that the peptide is within a Debye length of the pore (11). However, the primary inhibitory effect of GsMTx4 on endogenous MS channels is as a gating modifier, shifting their activation curves to higher tension and thus lowering occupancy of the open state (11). For gramicidin, GsMTx4 shifts channel activity toward the open state as though locally thinning the membrane.

Jung et al. (27) recently demonstrated that GsMTx4 has antimicrobial effects when applied to the periplasmic side of Escherichia coli. The effect may be due to GsMTx4 affecting MS channel activity that alters the physiology (28). Similarly, when GsMTx4 was expressed in bacteria as a fusion protein, it inhibited normal growth unless transcription was tightly controlled (P. A. Gottlieb and F. Sachs, unpublished results). The recent publication by Hurst et al. (29) demonstrated that GsMTx4 (>12 μM) applied to the extracellular side of E. coli increases the tension sensitivity of MS channels, potentially making the bacteria leaky.

The mechanosensitive channels MscS and MscL are the primary tension-driven osmolyte-releasing valves that limit turgor pressure in bacteria. If they become hypersensitive to membrane tension (say in the presence of GsMTx4), they would disrupt vital ion and metabolite gradients and exert antimicrobial effects. Apart from their physiological role as bacterial osmoregulators, these channels are convenient model systems for studies of other tension-driven conformational transitions in membrane proteins (30,31). The crystal structures of these channels (32) are known, and derived models (33–36) have allowed identification of specific functional groups that can illuminate GsMTx4's mode of action.

In this work, we first examined the phenomenological effects of GsMTx4 on two different mechanosensitive channels, MscS and MscL, in E. coli spheroplasts. GsMTx4 applied to the cytoplasmic face lowered the tension required to open the channels and, based on kinetic analysis of MscS, it slowed channel closing rates. GsMTx4 also delayed recovery of MscS from the inactivated state. The results can be explained if the peptide stabilizes the most expanded conformations that have a larger circumference and more boundary energy (line tension).

Materials and Methods

GsMTx4 was chemically synthesized and purified as previously described (12). The majority of recordings were done in E. coli MJF465 (mscS⁻, mscL⁻, mscK⁻) triple knock-out strain (37) expressing wild-type (WT) MscS inserted in pB10b vector (38). Additional experiments were done in PB113 strain carrying a native copy of mscL in its chromosome (kindly provided by P. Blount, University of Texas-Southwestern, Dallas), a derivative of MJF429 (37). In some experiments, MscL was recorded from MJF465 triple knock-out cells expressing WT MscL from the p5-2-2 vector (39). Spheroplasts were prepared as described previously (40).

Recordings were performed on inside-out patches with +20 mV in the pipette in the standard spheroplast recording solution (200 mM KCl, 5 mM CaCl₂, 45 mM MgCl₂, 5 mM HEPES-KOH, pH 7.4). The bath solution was supplemented with 0.4 M sucrose to osmotically stabilize the spheroplasts. Sucrose did not affect the channel behavior. Bath perfusion delivered the peptide to the cytoplasmic side of the membrane. Pressure stimuli were applied to the pipette from a high-speed pressure-clamp apparatus (ALA Instruments, Westbury, NY). Most analysis of current relaxation times was performed using the Clampfit (Axon Instruments, Foster City, CA) fitting routine. Kinetic fitting of ramp responses was done using the MAC routine of QuB (www.qub.buffalo.edu) that allows for time-varying stimuli (41,42). The data are based on 30 stable patches collected from nine independent spheroplast preparations and two separate batches of synthetic GsMTx4.

Results

For simplicity in the analysis and discussion of the data, we will assume that the primary stimulus driving channel opening is tension generated by the applied pressure according Laplace's law (43–46). In patches, however, the tension and its distribution may change with time as the membrane flows (46) and therefore the responses of channels to sustained pressure steps may be transient. This change of stimulus with time is called adaptation and is manifested as a gradual 10–20% shift of activation curves to higher tension with a relaxation time of 0.1–0.3 s. In excised patches specifically, the tension distribution in x, y, and z may change with time due to possible lipid redistribution between the leaflets through the rim (47). Adaptation is likely linked to stress relaxation in the inner membrane leaflet that does not interact with the pipette (43,47,48).

In pressure-clamp experiments, the rate of pressure onset can be <5 ms, permitting the study of channel activation before significant adaptation occurs. With slow time-varying stimuli, one can estimate the equilibrium properties of the adapted channels. Previous analyses of MscL suggested that, apart from short-lived substates, gating can be well approximated with a two-state C↔O model (44,49), where the equilibrium energy is referenced to either nonadapted or adapted stimuli (Fig. 1 A). In contrast, MscS exhibits a more complex behavior.

Figure 1.

Kinetic schemes for main transitions in MscL and MscS. MscL gating can be generally summarized in a two-state approximation (44). For MscS, reversible opening at subsaturating tension leads to adaptive closure which is then followed by complete inactivation. Normally, open channels do not inactivate. The inactivated state (I) is nonconductive and tension-insensitive. Return from the inactivated back to the closed (C) state is most effective in the absence of tension. The states are aligned according to the putative in-plane area of each of the conformations in the lipid bilayer.

A step of nonsaturating tension to MscS results in opening (C→O) usually followed by adaptive closure (O→C), previously termed desensitization (50). Under sustained tension, closed channels can enter a tension-insensitive inactivated state (C→I). The rates of both opening and inactivation increase with tension, but the tension dependency of the opening rate is steeper, allowing channels to open first. The channels inactivate most effectively under moderate (nonsaturating) tension after adaptive closure (40,51). Release of applied tension results in a fast (1–2 s) recovery from inactivation (I→C, see Fig. 1 B). We have measured the effect of GsMTx4 on gating midpoints for activation and closure for both channel types as well as on the rates of adaptation and recovery from inactivation for MscS. A typical experiment started with application of a 1-s pressure ramp to determine both the saturating and midpoint pressures that depend on the patch geometry. Pressures in the following step protocols were chosen according to the midpoint (p_0.5) and presented in the normalized form.

Fig. 2 A depicts pure MscS currents from an excised patch of an MJF465 spheroplast stimulated with 1-s linear ramps in the presence and absence of GsMTx4. We increased the concentration of GsMTx4 in the bath sequentially in 5 μM steps from 0 to 20 μM, and allowed 5 min to equilibrate at each concentration. The addition of 5 μM peptide shifted the gating midpoint from −95 mmHg to −82 mmHg (∼14%), and the shift saturated at −78 mm Hg (or 18%) with 15–20 μM. From the dose-response curve (Fig. 2 B), we estimated an effective K_d of ∼3.1 μM with a noncooperative binding model, but fitting the data with a cooperative equation produced a Hill coefficient of 2.4. At 20 μM the patches became less stable and rarely survived repeated stimulation. Fig. 2 C presents the statistics from nine independent patches of midpoint shifts toward lower tension as a function of peptide concentration. When applied from the cytoplasmic side, the peptide does not change the single channel current/voltage relationship (see Fig. S1 in the Supporting Material), suggesting that bound GsMTx4 may be more than a Debye length from the pore (52).

Figure 2.

Sensitization and hysteresis of MscS and MscL observed with pressure-ramps with GsMTx4 applied to the cytoplasmic side of an excised patch. (A) Open probability of MscS in response to 1-s pressure ramps to −120 mmHg for different concentrations of GsMTx4. (B) Concentration dependency of the midpoint shift fitted with a noncooperative Langmuir equation (dashed line, K_d = 3.1 μM) and with a cooperative Hill equation (solid line, K_d = 24 μM, n = 2.4). (C) Leftward shifts of activation midpoints as a function of GsMTx4 concentration measured in nine patches. (Bars) Standard deviations. (D) The hysteresis of MscS and MscL gating in response to triangular ramps for controls and 20 μM GsMTx4. (Arrows) Ascending and descending branches of the pressure ramp. (E) Midpoint pressures reflecting the magnitude of hysteresis for MscS and MscL and near-saturation of the effect in the range 0–20 μM GsMTx4.

To compare the effects of GsMTx4 on MscS and MscL under identical conditions in the same patch, we used sawtooth ramps (1 s up and 1 s down) with PB113 cells that express both channels. Fig. 2 D shows the biphasic response expected from two channel populations with different midpoints. The initial activation and the plateau at 80% of saturation reflected activation of MscS, and this was followed by activation and saturation of MscL. This patch contained ∼400 MscS and ∼35 MscL channels. The slope of the pressure ramp (ascending or descending denoted by arrows) affected the currents, and this hysteresis (Fig. 2 D) is a result of the channel kinetics being too slow to reach equilibrium during the ramp. Slower ramps (∼30 s) had negligible hysteresis (data not shown). Hysteresis was more pronounced for MscS that has a slower closing rate.

With increasing concentration of GsMTx4, less tension was required to open the channels and the hysteresis increased for MscS and decreased for MscL. The activation curves of both MscL and MscS were shifted equally to lower tension as though both channels sensed an equivalent change in local stress. The GsMTx4 concentration dependence for the midpoints of ascending and descending ramps is plotted in Fig. 2 E.

We used the MJF465 strain to express MscL alone and measured its response to 10 μM GsMTx4 (Fig. 3). GsMTx4 shifted the activation midpoint to lower pressures by 18 ± 4% (n = 4). In the following double-pulse protocol, the first pulse (saturating pressure) activated the entire MscL population and the kinetics of closure was monitored during a smaller second step. The rate of closure (k) decreased with increasing tension and became even slower with addition of peptide (Fig. 3, B and C). Fitted with single exponentials, these traces reveal that log (k) (where the time constant τ = 1/k) is almost linear with tension (presented in units of normalized pressure) and the 15% left shift caused by the peptide (Fig. 3 D) is comparable to the shift of the activation curve midpoint. This suggests that the change in the C↔O equilibrium caused by the peptide is primarily due to stabilization of the open state. The 18% left shift of the activation midpoint for MscL in the presence of 10 μm GsMTx4 translates into ∼9 kT decrease of energy of O relative to C, assuming ΔE_C→O = 50 kT for MscL (49). GsMTx4 caused the closing rate to decrease approximately fivefold near the activation midpoint (normalized pressure = 1), corresponding to an increase of the barrier height (ΔE_O→B) or lowering of the open state energy relative to the barrier by 1.6 kT. Previous kinetic and substate analysis suggested that the transition barrier in MscL is positioned at ∼2/3 of the way toward the full open state using an in-plane area scale (44,53) as a reaction coordinate. If the distortion of the energy landscape by GsMTx4 was linear with the expansion, as expected for a first-order approximation (46), we would expect ΔE_O→B to increase by ∼3 kT when ΔE_C→O decreases by 9 kT. The change of ΔE_O→B by only 1.6 kT suggests that the peptide affects the barrier height less than the open state energy.

Figure 3.

Effect of GsMTx4 on MscL expressed alone in MJF465 strain. (A) The shift of activation curve with 10 μM GsMTx4 under stimulation with the same linear ramp. The p_0.5 for this particular patch with and without peptide was −163 mmHg and −203 mmHg, respectively; the pressure scale is normalized to the midpoint pressure in control. The time course of MscL closing in control (B) and in the presence of 10 μM GsMTx4 (C) at different pressures. The double-pulse protocol (bottom) shows a 100-ms saturating-pressure pulse followed by varying subthreshold pulses. The current relaxation kinetics was fit with single exponentials. (D) Closing rates versus suction normalized to p_0.5.

As seen from Fig. 2 D, GsMTx4 exerts a stronger effect on MscS than on MscL. We measured the dose-response curves of MscS with a series of pressure steps from subthreshold to saturation. The pulse length (30 s) was sufficient to observe adaptation. Fig. 4 A shows the response of the control and 5 μM GsMTx4 from MJF465 cells expressing only MscS. GsMTx4 clearly sensitized the response to pressure steps and slowed the closing kinetics. Fig. 4 B shows the dose-response curves measured with 1-s ramps and a series of steps (peak values) on a single patch before and after peptide perfusion. The curves are plotted as a function of pressure normalized to the midpoint of the control ramp. When normalized by the maximal current at saturating pressure, the values represent open probability. The curves measured with ramps are right-shifted toward higher pressures relative to those measured with discrete steps as a result of stimulus adaptation. The presence of peptide shifts both dose-response curves to lower pressures—8% for steps versus 6% for ramps. Monoexponential fits of the decaying currents (Fig. 4 A) yield rates for adaptation (Fig. 4 C) that reflect the O→C transition (Fig. 1) under a gradually changing stimulus. Because the semilog plots showed almost linear dependencies, the downshift of log(k) caused by the peptide can also be interpreted as a left-shift along the pressure scale, comparable (11%) to the left-shift of the activation curves.

Figure 4.

Sensitization of MscS and reduction of the apparent adaptation rate in the presence of GsMTx4. (A) Activation of MscS with 30-s pressure steps of varied amplitude for 0 μM GsMTx4 (top) and 5 μM GsMTx4 (bottom). (B) Maximal current elicited by the pressure steps versus suction (symbols) and the ramp responses (continuous curves) taken on the same patch before and after application GsMTx4. (C) Rate constants of MscS desensitization versus suction pressure normalized to p_0.5.

As mentioned before, adaptation is a time-dependent shift of activation curve that occurs more rapidly at higher tension (47,50). It can be attributed to a relaxation of mechanical stress in the inner leaflet of the patch membrane (47,48). To test whether GsMTx4 could affect the time-dependent shift of the activation curve, we used a protocol consisting of two 1-s saturating ramps separated by a 45-s step of subthreshold pressure. The first ramp activated the channels, the constant subthreshold test-pressure facilitated closure and desensitization, and the second ramp assayed the remaining channels. The results are presented in Fig. S2. The peptide shifted the activation curves in response to both ramps toward lower pressure, while not changing the magnitude of the time-dependent shift of the activation midpoint toward higher pressure in response to the second ramp. Assuming that adaptation is caused by slippage/relaxation of the inner leaflet (47), the peptide does not change the overall membrane mechanics.

To further characterize the peptide-induced changes in the main opening and closing transitions (C↔O), we subjected the MscS population to two different protocols.

First was a two-step protocol: a 100-ms saturating pressure-pulse followed by a test pulse to lower pressure. All channels opened in response to the initial pulse and the test pulse revealed the kinetics of closing/adaptation at the reduced pressure (Fig. 5 A). The decay was fit with the sum of two exponentials, with the faster component taken as the closing rate. As shown in Fig. 5 C, log(k) for the closing process is a linear function of tension (proportional to pipette pressure), consistent with a single rate-limiting barrier linearly dependent on tension. The closing rates extrapolated to zero pressure were 2700 s⁻¹ and 200 s⁻¹, without and with 10 μm GsMTx4, respectively. The activation midpoint in this patch produced by 10 μM GsMTx4 decreased by 13% (p_0.5 = −123 mm Hg before and −108 mm Hg after addition of peptide), whereas the shift of log(k) versus pressure for closing was ∼26%. The almost identical slopes of log(k) versus pressure suggest that the peptide does not change the location (the in-plane area) of the energy barrier relative to the open-state well.

Figure 5.

MscS population responses to double-pulse and trapezoidal ramp stimuli are consistently affected by 10 μM GsMTx4 in the same patch. (A) Double-pulse recordings show the kinetics of MscS closing in the presence of peptide (bottom) compared to the control (top). (B) MscS population responses to ascending and descending ramps show increased hysteresis in the presence of the peptide. The blue (short-dashed) and red (long-dashed) lines represent the results of kinetic fitting with QuB (42) to the two-state model (see panel D). (C) Exponential fits of the current relaxation traces in panel A show linear dependencies of the closing times on pressure. Extrapolation of the linear fits to zero pressure yield intrinsic closing rates for the control population and in the presence of peptide. (D) The two-state model with the forward and backward rates exponentially dependent on tension (γ). For pressure-to-tension conversion, the pressure midpoint of the control trace recorded with the ascending ramp (p_0.5 = −123 mm Hg) was assumed to generate tension of 5.5 mN/m in the patch (67). (E) The reconstructed energy profiles for the C↔O transitions from the fitting parameters presented in Table 1 (line coding as in panel B). The horizontal axis represents the in-plane expansion of MscS complex taken as reaction coordinate.

A second pressure protocol utilized a trapezoidal stimulus of a 1-s ramp to saturation, a 1-s plateau at saturation, and a 1-s return to zero pressure. The response in control spheroplasts (panel B, blue trace) showed visible hysteresis; the midpoint ratio between the descending and ascending ramps was 0.72 (p_0.5 = 88 mm Hg vs. 123 mm Hg). Upon addition of 10 μM GsMTx4, the response to the ascending ramp was left-shifted (p_0.5 = 108 mm Hg), but the descending branch was right-shifted (p_0.5 = 43 mmHg, red trace) with a midpoint ratio of 0.40.

The real-time currents (Fig. 5 B) were corrected for series-resistance errors and fit to a two-state model using QuB with exponential dependencies on tension for the forward and backward rates (Fig. 5 D). To constrain the models, we fixed the tension-free value (preexponential term coefficient of the closing rate, k_o) to be the same as that measured in the previous experiment. The fits converged quickly (Fig. 5 B). The two-state model accurately reproduced the slopes of the ascending and descending branches, midpoints, and the increased hysteresis in the presence of peptide. The model slightly overestimated the current at the foot of the ascending branch and near the end of the descending branch. The parameters of the model (Table 1) allowed us to back-calculate the closing rates as a function of tension (Fig. 5 C, open symbols). They are in good agreement with experimental data for GsMTx4-modified channels, but are slightly overestimated for the control population at higher pressures.

Table 1.

MscS kinetic parameters obtained from the two-state fits (Fig. 5B)

	GsMTx4
	0 μM	10 μM
Ascending ramp parameters	p1/2 = −123 mm Hg	p1/2 = −108 mm Hg
	γ1/2 = 5.50 mN/m	γ1/2 = 4.83 mN/m
	ko = 7.01e-6 s−1	ko = 2.4e-3 s−1
	k1 = 2.54	k1 = 1.54
Descending ramp parameters	p′1/2 = −88 mm Hg	p′1/2 = −43 mm Hg
	γ′1/2 = 3.93 mN/m	γ′1/2 = 1.92 mN/m
	k′o = 2700 s−1 (constrained)	k′o = 200 s−1 (constrained)
	k′1 = −1.37	k′1 = −1.49
Thermodynamic parameters for the C↔O transition	ΔA = 16.2 nm2	ΔA = 12.6 nm2
	ΔE = 19.7 kT	ΔE = 11.3 kT

In control, the midpoint pressure (p_0.5) measured with the ascending ramp was assumed to correspond to tension of 5.5 mN/m, which reflects average tension in both monolayers in adapted patches. The energy (ΔE) and in-plane expansion (ΔA) for the MscS main transition were deduced from the kinetic parameters using the relationships: ln(k_o/k′_o) = −ΔE (kT); (k₁-k′₁) · 4.14 pN·nm = ΔA (nm²).

From the parameters of the model, we constructed an energy profile with a reaction coordinate taken to be the in-plane area of the channel (inferred from the tension sensitivity, Fig. 5 E). The coefficient of the tension dependence of the rates estimates the area difference from the wells to the top of the barrier, and hence the difference is the equilibrium change in area ΔA for the transition. The ratio of preexponential factors (k₀) gives the equilibrium constant and the free energy difference (ΔE) between the end states at zero pressure, i.e., the resting tension in the patch (54). For WT MscS, ΔE =19.7 kT and ΔA ∼16 nm². Based on the tension sensitivity coefficient, k₁, for the forward and backward reactions, the peak of the barrier is located 0.7 ΔA from the closed well. GsMTx4 decreased ΔE and ΔA by 8.4 kT and 3.6 nm², respectively. The decrease in ΔA can be attributed to either a slightly preexpanded resting conformation in the presence of peptide, or to the lack of membrane prestress in the model. The 300-fold increase of k₀ for the forward rate induced by GsMTx4 implies a decrease of the energy between the closed channel and the barrier peak by 5.8 kT. The reconstructed profiles suggest that the peptide exerts stronger effects on more-expanded states.

It has been shown previously (40) that the rate of MscS inactivation increases with tension, implying that the inactivated state has an in-plane area larger than the desensitized (closed) state. We attempted to record the rates of desensitization and inactivation with a protocol similar to one used previously (55), but the extremely slow rate of closure in the presence of peptide precluded the experiment. Instead, we measured the rate of recovery from inactivation with a train of short saturating test pulses (Fig. 6). The time-dependence of peak current recovery fit with an exponential provided the characteristic time. The control population recovered with a τ = 1.3–1.6 s, and 5 μM peptide slowed the recovery approximately fourfold. The stabilization of the inactivated state by GsMTx4 is consistent with the notion that this conformation of MscS should also be larger in diameter and thus expose a larger amount of the protein-lipid boundary.

Figure 6.

Kinetics of MscS recovery from inactivation after a prolonged subsaturating step revealed by a train of test pulses. The traces of MscS inactivation and recovery are taken before (A) and after application of 5 μM GsMTx4 (B). (C) Time course of recovery fitted with an exponential indicates that GsMTx4 slows the process of recovery ∼4.5-fold.

Discussion

We examined the effect of GsMTx4 on two bacterial mechanosensitive channels, MscS and MscL, and have shown that they respond differently from the gating modifier inhibition of stretch-activated channels observed in astrocytes (10,56) and other mammalian cells (21,57,58). Instead of inhibiting channel activity, the peptide stabilizes both the open and inactivated states, similar to the previously reported increase of the opening rate and decrease of the closing rate for gramicidin (11). Consistent with the previous results by Hurst et al. (29) who showed that extracellular GsMTx4 sensitizes MscS and MscK to tension, we have showed that cytoplasmic peptide also resulted in channel sensitization for MscS and MscL. Applying the peptide to the intracellular side of the patch, however, we never observed the biphasic behavior (29) at higher peptide concentrations. Because we expect that the local bending stress produced by extracellular and intracellular GsMTx4 will be of opposite polarity, this combined data suggest that torque on the channel (at moderate peptide concentrations) is not the relevant stimulus, while local thinning of the membrane around the channel (11,59) may be a key factor for the mechanism of GsMTx4.

We expect that in a single patch MscS and MscL are subjected to the same mean tension and both channels were affected by GsMTx4. If the membrane is heterogeneous so that stress is shared among subdomains (60), GsMTx4 could possibly alter the tension sharing and thus change the channel kinetics. Partitioning of GsMTx4 into the membrane and/or protein-lipid boundary is likely to be dependent on tension as expected for all amphipaths (46), but explicit measures of this effect is beyond the scope of this study. GsMTx4 should alter the curvature of lipids local to the channel and prestress in a direction favoring the open state. The reaction profiles presented in Fig. 5 E show that GsMTx4 reduces not only the energy of the open state but the barrier as well so as to increase the rate of opening at a given tension. The reduction of ΔA in the presence of peptide suggests that the resting conformation should be prestressed. The dose-response data do not show sharp discontinuities that might indicate domain formation. The effective Hill coefficient for binding is near 2.4 (Fig. 2), suggesting a moderate positive cooperativity with more than one GsMTx4 bound to a given channel.

As illustrated in Fig. 6, recovery from inactivation also slows in the presence of peptide. A parallel study of WT MscS inactivation (K. Kamaraju, V. Belyy, A. Anishkin, and S. Sukharev, unpublished) indicates that inactivation is accelerated and recovery slowed by increased tension, satisfying the assumption that the inactivated conformation has a larger in-plane area than the closed state. Because both closing and recovery from the inactivated state(s) are slowed by the peptide, expanded states may have higher affinity for the peptide. The observed phenomenology is consistent with observations in other channels where some cysteine-knot toxins exert their effects at the protein-lipid boundary (16,23,61). We should expect that, as a globular amphipath, GsMTx4 will affect the local curvature of the lipid bilayer. The stabilizing effects of 10 μM GsMTx4 on the open states of MscL or MscS are moderate, constituting 8–9 kT per complex or 1.2–1.8 kT per subunit, roughly corresponding to one hydrogen bond or salt bridge present with a probability of 0.2.

The benefit of observing such effects on structurally defined channels is that we can construct models of the resting and open conformations for both MscL (33,62) and MscS (35,36,63,64) and eventually make x-ray structures of the complex (32,32). The models have a set of residues at the cytoplasmic rim of the transmembrane barrel that may be a potential site for interactions with the peptide. In MscS, polar residues N50, N53, S70, and Y75 change their exposure to the lipid during opening and may potentially provide extra hydrogen bonding with the peptide that likely penetrates almost halfway through the membrane (16,17). Positively-charged R46, R54, K60, and R74 are not good sites for interaction because the peptide itself is positively charged (+5) and there is competition with polyvalent inorganic ions (Ca²⁺ or Mg²⁺). Another possible site is the negatively charged D67 in the TM1-TM2 loop that moves closer to the midplane of the membrane during opening (35), and may potentially go deeper into the membrane in the inactivated state whose structure is not yet defined.

With regard to MscL, a charged cluster RKKEEP at the C-terminal end of the lipid-facing TM2 helix is predicted to move deeper into the membrane in the open state (65). While residues R104, K105, and K106 would be interacting with the phosphate groups of phospholipids, the negative charges of E107 and E108 can provide a binding region for GsMTx4. It remains to be determined whether thinning of the lipid bilayer near the flattened MscL barrel in the open state (59,66) makes interactions with the peptide more favorable. We propose that GsMTx4 partitioned into the bilayer distorts the local lipids to favor the open state. The experiments with mirror-image D-peptides show that interactions between GsMTx4 and the channels, whether endogenous or gramicidins, are not stereospecific (11), and it seems unlikely that such highly different types of channel gating would involve specific salt bridges or hydrogen bonds. Clearly, experiments on structurally defined bacterial channels with the D-enantiomers could help elucidate the proximity and the specificity of interactions.

While more attention is required to understand the similarities, differences, and possible cross-reactivity among different Inhibitory Cysteine Knot toxins, GsMTx4 is becoming a promising tool for the studies of channel mechanisms that effect the protein-lipid boundary. Not only does GsMTx4 target MS channels ranging from cationic MS channels in cardiomyocytes and astrocytes to gramicidin (11), it is remarkably specific, having no toxic effects in cells (5,6,9) and organs (7) and whole animals in acute (3,4) or chronic treatment (F. Sachs, P. A. Gottlieb, and K,. Nagaraju, unpublished). The data presented here on the best-studied mechanosensitive ion channels support the hypothesis that the protein-lipid boundary is the primary site of GsMTx4 action and provide the first estimates of the energetic scale of GsMTx4 interactions.