The neuropeptide GsMTx4 inhibits a mechanosensitive BK channel through the voltage-dependent modification specific to mechano-gating

The cardiac mechanosensitive BK (Slo1) channels are gated by Ca2+, voltage, and membrane stretch. The neuropeptide GsMTx4 is a selective inhibitor of mechanosensitive (MS) channels. It has been reported to suppress stretch-induced cardiac fibrillation in the heart, but the mechanism underlying the specificity and even the targeting channel(s) in the heart remain elusive. Here, we report that GsMTx4 inhibits a stretch-activated BK channel (SAKcaC) in the heart through a modulation specific to mechano-gating. We show that membrane stretching increases while GsMTx4 decreases the open probability (Po) of SAKcaC. These effects were mostly abolished by the deletion of the STREX axis-regulated (STREX) exon located between RCK1 and RCK2 domains in BK channels. Single-channel kinetics analysis revealed that membrane stretch activates SAKcaC by prolonging the open-time duration (τO) and shortening the closed-time constant (τC). In contrast, GsMTx4 reversed the effects of membrane stretch, suggesting that GsMTx4 inhibits SAKcaC activity by interfering with mechano-gating of the channel. Moreover, GsMTx4 exerted stronger efficacy on SAKcaC under membrane-hyperpolarized/resting conditions. Molecular dynamics simulation study revealed that GsMTx4 appeared to have the ability to penetrate deeply within the bilayer, thus generating strong membrane deformation under the hyperpolarizing/resting conditions. Immunostaining results indicate that BK variants containing STREX are also expressed in mouse ventricular cardiomyocytes. Our results provide common mechanisms of peptide actions on MS channels and may give clues to therapeutic suppression of cardiac arrhythmias caused by excitatory currents through MS channels under hyper-mechanical stress in the heart.

Mechanosensitive (MS) 4 channels are membrane proteins that play important roles in multiple sensory processes, including hearing, touching, proprioception, pain, and numerous cellular functions, including gene expression, vesicular transport, and fluid homeostasis (1)(2)(3)(4)(5). In the heart, mechanical stress could increase excitability and initiate arrhythmias and failure. Thus, specific inhibition of cardiac MS channels could provide a novel therapy for cardiac arrhythmogenesis (6,7). Recently, the neuro-active peptide GsMTx4, extracted from the venom of the tarantula Grammostola spatulata (8), was found to specifically inhibit MS channels (9 -12). It has become an important pharmacological tool to investigate the functional roles of the MS channel in excitatory systems for normal physiology and pathogenesis (3,5,(13)(14)(15)(16). Nevertheless, the mechanism of inhibition is unclear.
GsMTx4 is a small amphipathic molecule with a conserved inhibitory cysteine-knot (ICK) backbone similar to other channel-sensitive peptides from the spider venom (17,18). GsMTx4 is known to influence MS channel gating (the processes of opening and closing) by partitioning into the lipid membrane as do other ICK peptides, such as the voltage-sensor toxin (VsTx) (8, 19 -22). However, the inhibitory action of GsMTx4 is not stereospecific (19), a property that differs from other ICK peptides (23), implying a different mechanism from others.
BK channels are expressed in nearly all excitable cells, and their activation depends on both voltage and intracellular Ca 2ϩ . All types of BK channels are tetramers. Each ␣ subunit contains three functional components: a voltage-sensor domain (VSD, S1-S4); a pore-forming domain (PD: S5-S6); and a large cyto-solic tail domain that comprises the two tandem regulators of K ϩ conductance (RCK) domains (RCK1 and RCK2), referred to as gating-rings (24 -27). Some BK (Slo1) channels are mechanosensitive (28 -30), including the stretch-activated BK (SAKca) channel that we previously identified in chick heart (31)(32)(33). Like other BK channels (24 -26), SAKcaC can be activated by both membrane depolarization and intracellular Ca 2ϩ . However, SAKcaC demonstrates stronger Ca 2ϩ -and mechano-sensitivities than other BK channels (31)(32)(33). Molecular cloning and subsequent mutagenesis studies have shown that the pore-forming ␣-subunit of SAKcaC contains an additional exon located between the RCK1 and RCK2 domains in the conserved BK C terminus (referred to as the stress-axis regulated exon or STREX) (Fig. 1A), which is critical for both the high Ca 2ϩ and mechano-sensitivities of this channel (31)(32)(33).
The atomic structure of SAKcaC is not known. However, the structure of the related BK channel (34,35) shows that Ca 2ϩ binding at the two binding sites per subunit stabilizes an expanded conformation of the gating ring, such that the conformational changes are propagated from the gating ring to the channel gates through the S6-RCK1 linkers. This is proposed to open the pore, a mechanism consistent with the spring-based BK gating-ring mode previously proposed by Niu et al. (24). For SAKcaC, membrane tension created by membrane stretching is supposed to be transferred to expand the gating ring (RCK1-STREX-RCK2) and subsequently impact the channel gates by pulling them open in a similar way as the gating-ring mechanism used by Ca 2ϩ in regular BK channels (33). In this study, we investigated the stretch-dependent gating mechanism of SAKcaC and its modulation by GsMTx4 using patch-clamp recordings of the WT channel and its mechano-reducing, STREX-deleted (STREX-del) mutation (31)(32)(33). We observed that GsMTx4 failed to inhibit the mechano-reducing mutation as well as the nonmechanosensitive mSlo1 channel. These results are consistent with a necessary role for the channel region encoded by the STREX-exon and lend support to the linkergating ring hypothesis in which membrane force is transferred to the gates, possibly by an unknown membrane protein (32,33). We also found that GsMTx4 exerts distinct effects on WT SAKcaC under the condition of membrane depolarization when compared with membrane hyperpolarized/resting states. With the combination of molecular dynamics (MD) simulation technology, we proposed two inhibitory modes for GsMTx4, a shallow-and a deep-inhibitory mode, which differentially impact mechano-gating of the SAKcaC. This paper presents our experimental observations and MD simulation outcomes, and we then present two proposed inhibitory modes as a working hypothesis for the mechano-gating of SAKcaC modulated by both membrane stretch and GsMTx4.

GsMTx4 inhibits stretch-activated BK channel from the extracellular side of cell membrane
Compared with other BK channels, SAKcaC is unique in that its gene contains an extra exon (STREX) encoding a 59-amino acid stretch-sensitive domain of the channel, located between the RCK1 and RCK2 domains (Fig. 1A). Because the cell membrane generates forces automatically upon the formulation of inside-out patch-clamp configurations (36), we first tested the GsMTx4 effect on the SAKcaC under the condition without additional pressure applied on the cell membrane.
Similar to our observations with other BK channels, in inside-out patch-clamp recordings from cultured ventricular myocytes isolated from the chick heart, we observed ongoing spontaneous activity of SAKcaC with a high concentration of Ca 2ϩ in the intracellular side (26), and the channel activity did not inactivate or rundown during the recordings. Furthermore, upon the background force automatically formed, we added Ϫ40 mm Hg membrane force and observed that the open probability (P o ) of SAKcaC was significantly increased; removal of the force reversed this increase in P o (Fig. 1, B-D), confirming the mechanosensitivity of this channel. The single channel conductance of SAKcaC was ϳ270 pS (Fig. S1), consistent with the "big" conductance character of BK channels as characterized previously (31)(32)(33).
We subsequently tested the effect of GsMTx4 on the ventricular myocyte SAKcaC by the back-fill method (See "Experimental procedures"), in which GsMTx4 gradually disperses to the outer surface of the ion channel protein (37,38). Fig. 1E shows typical single channel currents at different time points following the onset of 100 nM GsMTx4 backfilled in the pipette with 1 mM Ca 2ϩ in the bath (facing the intracellular side of the cell membrane). The data indicated that 100 nM GsMTx4 decreased P o of SAKcaC in a time-dependent way (Fig. 1, E and F; time courses of inhibition are summarized in Fig. 1G). The observed gradual decrease in P o for SAKcaC reflects the diffusion of GsMTx4 to the surface of the patch membrane (37,38). A higher concentration of GsMTx4 (500 nM) completely inhibited SAKcaC at the hyperpolarized voltage (Ϫ80 mV). However, the magnitude of channel inhibition by the peptide toxin was reduced by membrane depolarization, such as at ϩ50 mV (Fig. 1G), 500 nM GsMTx4 produced incomplete inhibition, whereas a 5-fold lower concentration produced nearly complete inhibition at Ϫ80 mV ( Fig. 1G; also see Fig.  9). This indicates that the peptide action is altered by voltages across the cell membrane (also see Fig. 9). Notably, the ion selectivity (K ϩ /Na ϩ ) and single channel conductance were not altered by GsMTx4 (Fig. S1, A and B), consistent with the idea that GsMTx4 acts as a gating modifier on MS channels.
We also compared GsMTx4 effects on SAKcaC under two different [Ca 2ϩ ] i concentrations. With 0.8 M Ca 2ϩ in the intracellular side of the cell membrane, 50 nM GsMTx4 inhibited SAKcaC by ϳ69.8% of the P o (P o was decreased from 61.9 Ϯ 10.1 to 18.7 Ϯ 6.8%). For comparison, the same concentration of GsMTx4 inhibited SAKcaC P o up to ϳ51.1% with 1 mM Ca 2ϩ in the intracellular side of the cell membrane (P o was decreased from 85.1.9 Ϯ 6.8 to 42.7 Ϯ 8.5%, Fig. S2). These results suggested that the effects of GsMTx4 on SAKcaC were not significantly impacted by the intracellular Ca 2ϩ concentration.

Deletion of the STREX-exon mostly abolishes the mechanosensitivity and GsMTx4 inhibition on SAKcaC
To investigate whether GsMTx4 acts by influencing channel mechano-gating, rather than interacting with the VSD as the voltage-sensor toxins do (e.g. VsTx1 and VsTx3) (23,39), we tested the impact of the toxin on STREX-del mutant channel (Fig. 2). Fig. 2A shows the linear map of STREX-del mutation, of which the STREX-exon located between RCK1 and RCK2 in the C terminus of SAKcaC is removed. However, deletion of this domain did not change the unitary single channel conductance nor the ion (K ϩ /Na ϩ ) selectivity (Fig. S3) of the channel. Up to Ϫ40 mm Hg of negative pressure applied on the cell membrane had no significant effect on channel activation at the voltages we tested (Fig. 2, B and C), confirming that the region encoded by STREX-exon acts as a mechanosensitive domain for SAKcaC. More interestingly, in the absence of the mechanosensitive domain, application of even a saturated concentration of GsMTx4 (7.5 M) from the extracellular side of the channel did not inhibit channel activity at any of the tested membrane potentials (Fig. 2, E-J). This result suggests that the "STREX" domain in the C terminus is critical for the inhibitory action of the peptide toxin on SAKcaC.

Voltage-dependent mechano-specific modification of GsMTx4 on SAKcaC
are possible because the voltage-sensor toxin was reported to have high channel specificity (36).

GsMTx4 does not inhibit stretch-activated BK channel from intracellular side of the cell membrane
To further explore the mechanism of GsMTx4 action, we tested whether GsMTx4 inhibits SAKcaC by directly interacting with the STREX domain of SAKcaC, which is located on the intracellular side of the membrane. GsMTx4 is known to inhibit MS channels by partitioning into the lipid bilayer (19,21,40) as do other ICK peptides (e.g. HaTxI and VsTx). Because there are several hydrophobic residues in the STREX-exon, it would seem possible for this domain to be partially attached on or bound to the inner leaflet of the membrane bilayer. If so, GsMTx4 might inhibit SAKcaC more potently from the intracellular side of the channel protein because GsMTx4 would have better access to the STREX domain when applied directly from the intracellular where V1 ⁄2 represents the voltage required for half of the maximum channel opening, and K represents the slope factor. Data show that even a saturation concentration of GsMTx4 had no effect on STREX-del mutation channel in the range of voltages examined. STREX-del mutation currents were recorded from the CHO-expressing system.

Voltage-dependent mechano-specific modification of GsMTx4 on SAKcaC
side of the cell membrane. Therefore, we tested the GsMTx4 effect on SAKcaC by direct perfusion from the bath, which accessed the intracellular side of the ion channel in the excised inside-out patch configuration. Fig. 4 shows that 100 nM GsMTx4 in the pipette (the extracellular side of the cell membrane) strongly inhibited SAKcaC P o at Ϫ80 mV, but even a 5-fold higher concentration of 500 nM GsMTx4 applied from the intracellular side did not show a significant effect on SAKcaC under the same conditions.
Perhaps surprisingly, there was also no significant effect of GsMTx4 on SAKcaC with higher concentrations applied to the extracellular side of the cell membrane using the cell-attached configuration (data not shown). This may be because the resting tension near the channel protein in the on-cell configuration is minimal (36), consistent with the specific effect of GsMTx4 on MS channels when mechanical stretch (and channel activation by this mechanism) is present.

Extracellular GsMTx4 inhibits SAKcaC gating in a dose-dependent manner
To investigate peptide-induced changes in the SAKcaC gating properties, we analyzed the kinetics of opening and closing in single channel recordings (41)(42)(43). With 1 mM Ca 2ϩ in a bath that was used to limit BK states (43), SAKcaC showed a high level of activity and high frequencies of opening and closing transitions, known as "flickering" (Fig. 5A, upper panel). Following the application of 50 or 100 nM GsMTx4 extracellularly, the channel opened less frequently and showed dose-dependent decreases in channel flickerings (  The inset at top left shows the voltage protocol used for current recordings. For clarity, the currents shown are at the voltages with a 20-mV increase from Ϫ140 to ϩ160 mV. For easy comparison, the red-colored traces on the left set highlight the currents activated at ϩ120 mV, and cyan-colored traces at right show those in the presence of 100 nM VsTx3 applied from the extracellular side of cell membrane. D and E, normalized G-V curves for mSlo1 (D) and mSlo1-STREX (E) channels at the concentrations of VsTx3 as indicated. The solid lines are fits to the Boltzmann equation (see "Experimental procedures"). The V1 ⁄2 obtained for mSlo1 is 54.1 Ϯ 7.3 mV for control, 61.1 Ϯ 9.2 mV for 25 nM, 66.5 Ϯ 7.0 mV for 100 nM, and 64.9 Ϯ 6.6 mV for 300 nM VsTx3 applied from the extracellular side. The V1 ⁄2 obtained for mSlo1-STREX chimera are 19.9 Ϯ 3.5 mV for control, 33.5 Ϯ 2.1 mV for 25 nM, 35.9 Ϯ 7.2 mV for 100 nM, and 37.3 Ϯ 6.8 mV for 300 nM VsTx3. F, summarized blocked currents (blocked I) for the effects of 100 nM VsTx3 on mSlo1 and mSlo1-STREX as indicated. The blocking effects were compared at the voltages near V1 ⁄2 (at ϩ50 mV for mSlo1 and at ϩ20 mV for chimera) or at 120 mV, where both channels were fully opened. Blocked I ϭ (1 ϪI VsTx3 /I Control ) ϫ 100%. G, comparisons of the effects of 100 nM VsTx3 on the voltage-dependent activation between mSlo1 and mSlo1-STREX. ⌬V1 ⁄2 ϭ V1 ⁄2(VsTx3) Ϫ V1 ⁄2(Control) . n.s., not significant. n ϭ 4 -7. mSlo1 and mSlo1-STREX currents were recorded from Xenopus oocytes with outside-out patch configuration. VsTx3 was applied from bath (the extracellular side of the cell membrane).

Voltage-dependent mechano-specific modification of GsMTx4 on SAKcaC
nents. GsMTx4 significantly abbreviated the open-time duration ( O ) (Figs. 5B, left, and 4C) and prolonged both the fast ( C1 ) and the slow ( C2 ) closed-time components (Figs. 5B, right, and 4D) in a dose-dependent manner, accompanied by a significant increase in the relative weight of A C1 , which reflects the percentage of the closed events for C1 (Fig. 5E). However, channels exhibited a decreased P o because C1 was significantly prolonged. In summary, the net effect of GsMTx4 was to abbreviate open-time duration and prolong closed-time constants, resulting in shorter and less frequent opening.

GsMTx4 inhibits SAKcaC P o increased by membrane tension
We next considered how GsMTx4 action specifically targets SAKcaC mechano-gating. Because it is not possible to estimate the resting tension automatically formed from the excised inside-out patch configuration, we studied the GsMTx4 effect on channel activation induced by additional membrane stretch (caused by suction of the membrane patch). As shown in Fig. 6, A-C, Ϫ40 mm Hg applied on the cell membrane increased SAKcaC P o dramatically, consistent with our previous data on the channel's mechanosensitivity (31)(32)(33). Interestingly, application of a low concentration of GsMTx4 (50 nM) from the extracellular side (with continued membrane stretch of Ϫ40 mm Hg) fully reversed the stretch-induced channel activation and backed to the resting level. As also shown in Fig. 2, the STREX-del channel failed to respond to the additional membrane stretch and failed to show inhibition by the same concentration of GsMTx4 (Fig. 6, A, lower panel, and C). Moreover, Ϫ40 mm Hg of membrane stretch caused a marked leftward shift of the P o -V curve in the hyperpolarized direction by about 40 mV, whereas subsequent application of 50 nM GsMTx4 from the extracellular side of the cell membrane (in the presence of continuous membrane stretching of Ϫ40 mm Hg) shifted the curve in the opposite direction by ϳ40 mV, as if GsMTx4 antagonized the gating induced by membrane stretch (Fig. 6D). Finally, GsMTx4 (50 nM) applied in the absence of additional suction induced a rightward shift of the channel's P o -V curve, indicating that even the resting levels of membrane stretchinduced channel activity are antagonized by the toxin. ) during GsMTx4 diffusion to the patch membrane upon the onset of backfilling in the pipette (olive squares) or perfusion from bath (green squares). The time points were measured from the onset of back-filling from the extracellular side of the cell membrane (olive squares) or application from bath (green squares). GsMTx4 concentrations used were 100 nM (olive squares) for backfilling or 500 nM (green circles) applied from the bath (n ϭ 6 -8). GsMTx4 concentrations used were 100 nM in A and B and 500 nM in C and D, respectively. MPs were held at Ϫ80 mV in all experiments. SAKcaCs were recorded from the CHO-expressing system.

Voltage-dependent mechano-specific modification of GsMTx4 on SAKcaC
Next, we examined whether GsMTx4 itself could affect SAKcaC activity through the modulation of membrane tension. Fig. 7A shows SAKcaC activation induced by negative pressure (Ϫ40 mm Hg) applied to the cell membrane. Although P o was decreased to 6.9% with 50 nM GsMTx4 applied from the extracellular side, negative pressure (Ϫ40 mm Hg) still increased P o to 22.9 for this particular patch recording (Fig. 7B). However, there was no significant difference among the inhibited channel activations by GsMTx4 under the different membrane tensions (Fig. 7, C and D). We thus concluded that GsMTx4 itself does not control channel activation as does membrane stretch.

GsMTx-4 inhibits SAKcaC through the specific gating modulated by membrane stretch
To explore the biophysical mechanism of how GsMTx4 modulates the mechano-gating, we performed detailed analysis of single channel kinetics, which directly reflects the gating characteristics of the channel. Fig. 8 shows SAKcaC channel kinetics induced by membrane stretch alone (Ϫ40 mm Hg) or membrane stretch in the presence of GsMTx4 (50 nM). The results show that membrane stretch alone (Ϫ40 mm Hg) significantly prolongs SAKcaC open-time duration ( O ) at all membrane potentials to a similar fold increase, and this was subsequently reduced approximately to the control levels following addition of an extracellular 50 nM GsMTx4 (with continued stretch in the pipette) (Fig. 8, A-C, left, and D). Furthermore, membrane stretch (Ϫ40 mm Hg) significantly shortened the closed-time constants ( C1 and C2 ) (Fig. 8E) but only at hyperpolarization or resting potentials (e.g. at Ϫ80 or Ϫ50 mV, Fig. 8, E and F); there was no measurable effect of stretch on the closed-time constants at more depolarized voltages (e.g. at Ϫ20 mV). Again, the channel gating induced by membrane stretch was reversed by GsMTx4 at hyperpolarized/resting potential (Figs. 8, A-C, right, 6, E and F). Finally, membrane tension did not have a significant effect on the relative area of A C2 (e.g. at Ϫ20 mV), which reflects the percentage of the closed events for C2 , but tension dramatically decreased A C2 at the membrane potentials close to resting or hyperpolarization levels (for example, A C2 was decreased more than 2.0-fold by Ϫ40 . Data for GsMTx4 effects were obtained 25 min later following the onset of back-filling when drugs were completely diffused to the cell membrane. MPs were held at Ϫ50 mV. Data points at each condition represent at least three determinations (n Ͼ3); *, p Ͻ 0.01; **, p Ͻ 0.001; n.s, not significant. SAKcaCs were recorded from isolated chick ventricular myocytes.

Voltage-dependent mechano-specific modification of GsMTx4 on SAKcaC
mm Hg at Ϫ80 mV from 44.5 Ϯ 7.1 to 20.3 Ϯ 3.3%). Again, these effects were almost reversed by further application of 50 nM GsMTx4. Together, these results from single channel analysis suggest that GsMTx4 targets the mechano-gating of SAKcaC.
In summary, the net effects of membrane stretch on SAKcaC at resting/hyperpolarized voltages (such as Ϫ50 and Ϫ80 mV) resulted in a prolongation in channel opening ( O ) and a decrease in the closed-time constants with a concomitant decrease in the slow-closed events (indicated by A C2 in Fig. 8G), all of which led to an increased P o . However, 50 nM GsMTx4 (under the continuous suction of Ϫ40 mm Hg in the pipette) reversed all these effects induced by membrane tension, an effect that could be explained by GsMTx4 inhibition of SAKcaC activation by mechanic stretching. Nevertheless, at more depolarized voltages (e.g. at Ϫ20 mV), both membrane stretch (Ϫ40 mm Hg) and GsMTx4 (50 nM) did not significantly affect the factors for the closed-time components (e.g. C1 , C2 , and A C2 ), resulting in a less inhibitory effect of GsMTx4 on SAKcaC (see "Discussion"). These results also suggested that SAKcaC channel gating is modulated by both membranes stretch and the peptide toxin by a mechanism that depends on membranes potential (see "Discussion").
Taken together, these results demonstrated that membrane stretch activated SAKcaC as a positive-gating modifier, whereas the spider peptide GsMTx4 antagonized the effect of membrane stretch and thus acted as a negativegating modifier. We concluded that GsMTx4 inhibited SAKcaC gating through the specific modification of the channel mechano-gate.

GsMTx4 shows more efficiency on SAKcaC under hyperpolarized/resting conditions
In addition to being sensitive to membrane stretch, SAKcaC is also regulated by membrane voltages as the normal BK (Slo1) channels (25,26,31). We compared the effects of GsMTx4 at a voltage range from hyperpolarization (e.g. at Ϫ80 mV) to depolarization potential (e.g. at ϩ30 mV), where the channel reaches maximal levels of opening with 1 mM Ca 2ϩ in the intracellular side (see Fig. 6D). Although 100 nM GsMTx4 inhibited SAKcaC open probability approximately ϳ74.5% at Ϫ80 mV, the same concentration of GsMTx4 decreased P o by only ϳ20% at ϩ30 mV (Fig. 9, A-D). To exclude the possibility that GsMTx4 exerts its effect by interaction with the VSD as is the case for the voltage-sensor toxins (44, 45), we further tested for an effect of GsMTx4 on mSlo1, the regular BK channel that contains VSDs

Voltage-dependent mechano-specific modification of GsMTx4 on SAKcaC
just like SAKcaC, but lacking the STREX-exon in the C terminus. Under similar conditions, concentrations as high as 10 M GsMTx4 did not show a significant effect on mSlo1 (Fig. 9, E and F), confirming again that GsMTx4 effect was not achieved by interacting with the VSDs.

Lipid-peptide interaction is influenced by membrane voltages
With a belt of positively charged residues (ϩ5e) around the periphery and a hydrophobic protrusion (18,19), GsMTx4 is proposed to interact with carbonyl oxygen atoms of both leaflets of the lipid bilayer membrane, similar to the action of HaTx1 and VsTx. Using MD simulations under free production runs, Nishizawa and Nishizawa (40) have suggested two peptide-lipid interaction modes: a shallow binding mode in which the positively charged residues of GsMTx4 interact with the outer leaflet lipid, and the deep-binding mode in which the electrostatic interactions bring the positively charged groups of GsMTx4 into the vicinity of the inner leaflet of the cell membrane. To examine the impact of membrane voltage on the inhibitory effect of the toxin, we performed MD simulations to mimic toxin's actions at different voltages. Fig. 10 shows the representative trajectories and the corresponding snapshots under depolarized (A-C) and hyperpolarized (D-F) states. GsMTx4 com (the center of GsMTx4) was initially put at the location of ϳ0.5 nm above the center of the DPPC bilayer (Fig.   10, inset). Following a voltage-restricted production run (see "Experimental procedures"), GsMTx4 was driven outward (under depolarized state) or inward (under hyperpolarized/ resting states), respectively, by the electrostatic force. At 60 -70 ns of the system runs, the GsMTx4 com positions corresponded to two different modes of interactions, namely the shallow interaction mode (under depolarization condition, Fig.  10, A-C) and the deep interaction mode (under the hyperpolarized/resting states, Fig. 10, D-F), respectively. Under the depolarized state (shallow interaction mode), GsMTx4 com moved upward (Fig. 10, A-C), and the positively charged residues (e.g. Lys-8, Lys-15, Lys-18, Lys-20, Lys-22, Lys-25, and Lys-28) become stabilized near the headgroups of the outer leaflet DPPC by interaction with the carbonyl oxygen atoms (Fig. S4). This mode is consistent with the shallow binding mode suggested by Nishizawa and Nishizawa (40). By contrast, under the hyperpolarized condition (deep-binding mode), GsMTx4 com moves downward (Fig. 10, D-F), and the positively charged residues become split into two functional regions, one interacting with the outer leaflet (Lys-18, Lys-20, and Lys-22) and the other interacting with the inner leaflet (e.g. Lys-15, Lys-8, Lys-28, and Lys-25). This later mode is consistent with the deepbinding mode (40). In the deep-binding mode, GsMTx4-induced membrane deformation was uneven around the peptide; it inserted deeply and interacted with the inner monolayer

BK variants containing STREX insert are expressed in mouse cardiac myocyte membrane
To investigate whether SAKcaC is also expressed in mouse ventricular myocytes, we used an anti-STREX Ab, which is specific against STREX-exon in the BK C terminus, to detect STREX-containing BK variants. Notably, immunostaining of STREX indicates a relatively high STREX signal in the heart, in both atrium and ventricular myocytes (Fig. 11, A-F). The overlay of the STREX signals with the plasma membrane marker (WGA), as indicated by white arrows in Fig. 11, I and J, further confirmed the co-localization of STREX with WGA on the plasma membrane of mouse ventricular myocytes (Fig. 11L).
Recently, different BK variants, e.g. BK-STREX, BK-DEC, and BK-STREX-DEC, were identified in ventricular cardiomyocytes (46 -49). Among them, the BK variant containing both STREX and DEC (BK-STREX-DEC) was found to be mechanosensitive. To test the possible contribution of DECcontaining BK variants to the native SAKcaC (BK-STREX) currents we obtained, we used a poly-cloned antibody against DEC at the end of the BK ␣ subunit, generated by the same antigen peptide used by Singh et al. (46). Although BK variants containing STREX inserts (e.g. BK-STREX and/or BK-STREX-DEC) are clearly observed in ventricular myocyte membranes (Fig. 11), as shown in Fig. S5, we did not observe signals of overlaid DEC with WGA in ventricular myocytes (Fig. S5), suggesting that BK variants containing DEC, e.g. BK-DEC and BK-STREX-DEC, are not located on the surface of ventricle myocytes (Fig. S5). Thus, we con- The histograms in C were obtained from the single channel traces 25 min later following the back-filling when GsMTx4 was completely diffused to the patch membrane. Data points in D-G at each membrane potential represent at least three determinations. *, p Ͻ 0.05; **, p Ͻ 0.01. SAKcaCs were recorded from isolated chick ventricular myocytes.

Voltage-dependent mechano-specific modification of GsMTx4 on SAKcaC
cluded BK-STREX-DEC did not contribute to native cell currents that we recorded from the ventricular myocytes.

Discussion
GsMTx4 inhibits mechanosensitive channels (MSCs) by interacting with the lipid bilayer rather than physically occupying the channel pore (19). MD simulation with free production runs suggested two binding modes of GsMTx4 interaction with the lipid bilayer, which may influence the magnitude of GsMTx4 effects on the MSC gating (22,40). In this study, we presented data to address the specific inhibitory mechanism that GsMTx4 acts directly on the mechano-gating of a stretchactivated BK channel (SAKcaC). Validation of the expression of these channels in mouse ventricular myocyte membranes (Fig.  S5) provides justification for understanding. We also explored the peptide-lipid interaction modes, under depolarized versus hyperpolarized/resting states, to interpret the observations that GsMTx4 inhibited SAKcaC stronger at hyperpolarized/resting states compared with that at the depolarized condition. Each mode appears in a voltage-dependent way parallel to the two inhibitory modes.

Mechano-sensing domain, STREX in SAKcaC, is the target for GsMTx4 inhibition
The mechanosensitive channel inhibitor, GsMTx4, belongs to the same peptide family as voltage sensor toxins, and they have structural similarity with hydrophobic patches in the surface and a belt of positively charged residues around the periphery (18). It is known that the voltage-sensor toxins inhibit voltagedependent ion channels by partitioning into the bilayer to interact with the voltage sensor paddle (VSP) buried in the lipids (39,44,45). SAKcaC possesses a similar VSD to those of regular BK (Slo1) and voltage-dependent K ϩ channels. Thus, the question first to consider was whether GsMTx4 inhibits SAKcaC by directly targeting its VSD domains.
Generally, the mechanism of BK channels regulated by membrane potentials is explained by having the four voltage sensors move in (or through) the fluid membrane interior, as for volt-

Voltage-dependent mechano-specific modification of GsMTx4 on SAKcaC
age-gated Kv channels (24,27,35). Judging from the previous reports that GsMTx4 inhibits MS channels by partitioning into the cell membrane to modulate channel gating (19,40), a reasonable hypothesis is that GsMTx4 inhibits SAKcaC by acting at the VSDs, as in the case of VsTx. This hypothesis is also attractive because the positively charged GsMTx4 (ϩ5e) could repel electrostatically the positively charged VSDs upon its portioning into the cell membrane, thus facilitating the VSDs moving from the exterior (activated state) to the interior (inactivated state), an action similar to that achieved by membrane hyperpolarization. However, the observation that even at saturating concentrations GsMTx4 did not inhibit mSlo1 (Fig. 9, E and F) nor the STREX-del mutant (Figs. 2 and 6) clearly excluded the possibility that GsMTx4 achieves its effect through interaction with the channel's VSDs, because both mSlo1 and STREX-del mutant possess VSDs similar to those of regular Kv channels. Rather, our data support an alternative hypothesis that the STREX-exon located between RCK1 and RCK2 domains in the SAKcaC C terminus is the target (direct or indirect) for the inhibition by GsMTx4. Key data in support of this explanation is that deletion of STREX, one of the mechano-sensing domains (or sites) in SAKcaC, completely abolishes the effect of the peptide at all voltage ranges tested (from Ϫ100 to ϩ70 mV). Nevertheless, we did not observe a significant inhibitory effect of GsMTx4 on mSlo1-STREX, a chimeric construct containing the STREX insert between the RCK1 and RCK2 domains in the mSlo1 channel, suggesting that the different structures between mSlo1 and SAKcaC may also account for GsMTx4 effects.
Recently, stretch activation of the BK-STREX-DEC variant was also noticed in cardiac myocytes (46,47). We thus considered whether this channel variant could have contributed to the results obtained from native myocytes. The SAK-caC clone used in our experiments does not contain the DEC insert in the end of C terminus, and the characteristics of the channel were consistent with our results from experiments done in native myocytes (31,33). We also failed to detect a signal against the DEC region at the surface of myocytes isolated from newborn mice (Fig. S5). This is consistent with the view that STREX and DEC are located in different places.  (as in B). The inset at top illustrates the method of the distance measurement between GsMTx4 com and inner monolayer measured in A and D. Ocher spheres represent phosphorus atoms, and the red represents carbonyl oxygen atoms; GsMTx4 is represented in surface structure. Basic residues (Arg and Lys) are blue, and acidic residues (Asp and Glu) are red. Trp residues are orange; Phe residues are yellow, and other hydrophobic residues (Ala, Cys, Ile, Leu, Met, Pro, and Val) are green. For clarity, water molecules and chloride ions are not shown.

Voltage-dependent mechano-specific modification of GsMTx4 on SAKcaC
DEC-containing splice variants are reported to be specifically localized to the mitochondria (46). Thus, there is no evidence that BK variants containing the DEC exon played a role in our experiments.

GsMTx4 inhibits SAKcaC through the specific modulation of the mechano-gating
GsMTx4 has been widely used to detect MS channel functions in many excitable systems because of its specific effects on MS channels. Nevertheless, the mechanism of how this peptide targets the mechano-gates of MS channels is unknown. In this study we show that the P o of SAKcaC was significantly increased by membrane stretch (Figs. 1B, and 6,  A and B), which caused a leftward shift of the P o -V curves (Fig. 6C), suggesting that the channel gating was positively modulated by membrane stretch. This "shift" was subsequently reversed by further application of GsMTx4 (Figs. 6 and 8), indicating that GsMTx4 acts as a negative gating modifier to antagonize the opening of channel mechanogates activated by membrane tension.
GsMTx4 is reported to induce membrane (bilayer) stretch by membrane deformation like an immersible wedge; hence, the deformation of lipid membrane is different from that generated by ordinary membrane stretch, e.g. negative pressure. Thus, GsMTx4 could not induce changes in channel kinetics to control channel activation as membrane stretch does. It only acts to antagonize the effect of membrane stretch. In addition, the extent (%) of the inhibition by GsMTx4 was also not altered by changes in membrane tensions (Fig. 7), suggesting again that GsMTx4 itself could not induce membrane stretch to modulate channel gating. To clarify our understanding of the gating relationships between membrane stretch and GsMTx4, we performed detailed kinetic analysis of the single channel gating. The observations that the prolonged channel open-time ( O ) and shortened closed-time constants ( C1 and C2 ) by membrane stretch are consequently reversed by further application of GsMTx4 (Fig. 8) confirm that GsMTx4 inhibits SAKcaC through the specific modification of the channel mechanotransduction gating. Furthermore, this is consistent with the elimination of GsMTx4 effects at all voltages on the STREXdel mutant channel. Based on the above observations, we thus hypothesize that stretching of the cell membrane generates a force on channel gate through the expansion of the gating-ring complex (RCK1-STREX-RCK2 in SAKcaC), which opens the channel gates (Fig. 12, A and B), a mechanism analogous to what Ca 2ϩ does on BK channel mode (24,27,35). However, the peptide GsMTx4 likewise generates compression forces by specifically targeting the mechanical tension on the channel gates, through the same pathway (gating-ring complex), to push the SAKcaC gate to close (additional details below). There is still the possibility that the toxin binds directly to the STREX domain and thus disrupts the communication between sensing the stretch stimulus and opening the channel. Nevertheless, this seems unlikely, because GsMTx4 shows no inhibitory action when applied to the intracellular side of the cell membrane (Fig. 4), which gives the toxin access directly to the STREX domain.

Two inhibition modes gated by membrane stretch and GsMTx4 inhibition
We have presented data to show that GsMTx4 inhibits SAKcaC by specifically targeting the mechano-gate without interaction with VSDs. Our hypothesis for the different inhibitory effects of GsMTx4 on SAKcaC between membrane hyperpolarization and depolarization (Fig. 9) is that this may arise from the different lipid-peptide interaction modes.
Recently, with MD simulation technology, Nishizawa and Nishizawa (40) showed two binding modes of GsMTx4 with the lipid bilayer followed by a free 30-ns production run: the shallow binding mode where all the positively charged residues of GsMTx4 interact with the outer leaflet lipid, and the deepbinding mode where the electrostatic interaction brings the positively charged groups into the inner leaflet lipids, accompanied by membrane deformation (40). To explore the possible different peptide-lipid interaction modes, we also performed MD simulation to mimic the action of GsMTx4 inside the lipid bilayer under hyperpolarizing/resting versus depolarizing conditions. Under the depolarized state, GsMTx4 that partitions into the cell membrane was driven back along the electric field toward the outer leaflet, resulting in binding shallowly with the outer monolayer. This induces weaker membrane defor-

Voltage-dependent mechano-specific modification of GsMTx4 on SAKcaC
mation, which we call the shallow lipid-peptide interaction mode (Figs. 10, A-C, and 12). In contrast, under membrane hyperpolarization/resting states, the positively charged GsMTx4 moves down the electric field and binds deeply with the inner monolayer of lipid bilayer (Fig. 10). This resulted in deep interaction of GsMTx4 with the inner monolayer, where the mechano-sensing domain (STREX-exon) of SAK-caC is closed (or attached); we call this the deep interaction mode (Fig. 10, D-F). Based on our observations and BKgating mechanisms established previously (24,27,33,35), we thus predict two inhibition modes, referred as "deep inhibition" (Fig. 12C) and "shallow inhibition" (Fig. 12E), to further interpret our observations.
In terms of the two modes in Fig. 12, membrane tension first turns the passive spring of the gating ring (RCK1-STREX-RCK2 in SAKcaC) into a force-generating machine with a characteristic force that uses the free energy to expand the gating ring (four RCK1-STREX-RCK2 in one ␣ subunit) and thus opens the channel gates, as predicted previously in BK channel (24,35). Under membrane hyperpolarizing/resting conditions (Fig. 12B), where negative charges accumulated in the inner monolayer (the intracellular side of cell membrane) and the positive charges outside the outer monolayer (the extracellular side of cell membrane), the positively charged (ϩ5e) peptide is driven down deeply along the electrochemical gradients to bind with both leaflets of the lipid bilayer based on our observations from MD simulation outcome (Fig. 10, D-F). This generates a large closing force and strongly pushes the unknown membrane protein (MP) back (as we predicted previously (33)), thus reversing the effects induced by membrane stretching (Fig. 12, B and C). In this mode, GsMTx4 is further stabilized deeply inside the lipid bilayer by interaction simultaneously with the carbonyl oxygen atoms in both leaflets of the lipid bilayer as observed in MD simulation. It is placed deeply and binds tightly with the MP/inner monolayer to reverse the membrane force generated by membrane stretch that expands the four-gating rings (in one SAKcaC ␣ subunit) to open the channel gate.
Under the depolarization condition, the outwardly moved VSD further opens the S6 gate of SAKcaC (Fig. 12, B-D) as observed in BK channel. Our hypothesis is that the less inhibition of GsMTx4 on SAKcaC is reached by the release of lipid-peptide interaction, from the deep interaction mode (Fig. 12C) to the shallow interaction one (Fig. 12E). As we observed in the MD simulation outcome, GsMTx4 is driven back along the electrochemical gradients, thus the closing forces generated by GsMTx4 on the inner monolayer of cell membrane is certainly released, resulting in less inhibitory effect of GsMTx4 on SAKcaC unless a higher concentration of GsMTx4 is used to neutralize the negative charges accumulated outside of the cell membrane. Alternatively, membrane depolarization prevents the positively charged extracellular peptide from partitioning into and moving down against the electrochemical gradients, thus generating less closing force on the inner leaflet of the cell membrane (Fig.  12E). Six lysines in GsMTx4 have been proposed to be important in affecting lipid-peptide binding (21,22,40). In our MD simulation outcome, lysines (e.g. Lys-8, Lys-15, Lys-25, and Lys-28) are interacting with carbonyl oxygen atoms to cause membrane deformation (Fig. S4); thus, it is not difficult for one to imagine that mutation of these positively  (24,33). A and B, membrane force (F m ) first pulls SAKcaC gate opening through STREX and MP (33). C, deep inhibition mode: under hyperpolarized/resting conditions, GsMTx4 was driven down (inward) along the electrochemical gradients upon partitioning into the lipid bilayer. It was placed at a deep position to interact with both inner and outer monolayers and induce strong membrane deformation as observed in the MD simulation. Thus, GsMTx4 has the ability to push STREX back strongly through MP, to close the channel gate firmly. D, membrane depolarization drives VSP upward, which generates pulling force to further open channel gate (from B). E, shallow inhibition mode: under the depolarization condition, GsMTx4 was driven back (move outward) by membrane potential. Alternatively, the outward electrostatic forces may also prevent the absorption of GsMTx4 from the extracellular side into the lipid bilayer. In either case, GsMTx4 interacts with the outer monolayer only, and induces weak membrane deformation as observed in the MD simulation. The SAKcaC gating mode is drawn based on BK gating mode (24,27,33,35), in which VSP and STREX-exon are both connected to the channel gate.

Voltage-dependent mechano-specific modification of GsMTx4 on SAKcaC
charged lysines in GsMTx4 could reduce the interaction force between lipid-peptide and thus decrease membrane deformation and release the closing force generated by partitioning of peptide. Therefore, the final results of mutating the lysines would result in decreased efficiency of peptide inhibition on SAKcaC.
In these modes, because SAKcaC structure is not known, it is not certain how the mechano-sensing domain STREXexon in the intracellular side of the SAKcaC C terminus is communicating with the lipid bilayer. Although the hydrophobic residues in STREX-exon (Fig. S3A) provide the possible interaction of STREX with the inner monolayer, an unknown MP is supposed to be in the modes to serve as the communicator to link the peptide and STREX-exon, as predicted previously (32,33). Although we observed a weak response of the STREX-del mutation to membrane stretch at higher voltages, this does not affect our conclusion for the specific effect of GsMTx4.

Independent and synergistic gating by voltage-sensor domain and membrane stretch/GsMTx4
In the modes of SAKcaC we presented in Fig. 12, the movement of VSDs and the expansion/compression of cytosolic gating-ring complex (four in one ␣ subunit) are relatively independent of one another. However, as the gating ring and the VSDs in SAKcaC are finally converged on the same gate (S6), the gating by each of the voltage sensors in SAKcaC and membrane stretch/GsMTx4 would be synergistic as observed in regular BK channels, for which there are independent and synergistic effects on the gating ring due to the modulation of the VSD moving and Ca 2ϩ binding (24,27,33,35).

Cell culture, mutagenesis, and heterologous expression
Native SAKcaC was recorded from chick ventricular myocytes, which were dissected from 10-to 12-day-old White Leghorn embryos as described previously (31,33). The chick SAKcaC gene was cloned from a cDNA library made from chick embryonic hearts (31)(32)(33). STREX-del mutation was verified by sequencing as published previously (32,33).
Cloned SAKcaC or STREX-del cDNAs were transiently transfected in Chinese hamster ovary (CHO-K1) cells using the Lipofectamine TM 2000 transfection reagent (Thermo Fisher Scientific) according to the manufacturer's protocol. pEGFP (Clontech) was co-expressed with the channel cDNA at a ratio of 5:1 (weight/weight) to identify successfully transfected cells as used previously (31)(32)(33).
Mouse Slo1 (mSlo1) cDNA clone was a gift from Dr. Christopher Lingle (Washington University in Saint Louis). cRNA was transcribed in vitro with MessageMachine kit SP6 (Ambion) and injected in Xenopus oocytes as described previously (50,51). No endogenous BK channel activation was detected in untransfected CHO-K1 or uninjected Xenopus oocytes.
Macroscopic currents for BK (mSlo1) and mSlo1-STREX chimera were recorded from Xenopus laevis oocytes under standard excised inside-out (Fig. 9, C and D) or outside-out (Fig. 3) configurations (25,52) with an A-M 2400 patch-clamp amplifier (A-M Systems, Inc.). pClamp (Molecular Devices) was used to drive stimulus protocols and digitize currents (26). Whether for outside-out or inside-out patches, the standard extracellular solution was as follows (in mM): 140 KMES (methanesulfonate), 20 KOH, 10 HEPES, 2 MgCl 2 , pH 7.0. The composition of internal solution contained the following (in mM): 140 KMES, 20 KOH, 10 HEPES, 5 EGTA, pH 7.0. The [Ca 2ϩ ] i solutions were buffered as described previously (25,26). To obtain the conductance-voltage (G-V) curves, currents were elicited by voltage pulses from Ϫ140 to Ϫ160 mV (20 ms) with 10-mV increments, whereas the voltages before and after the pulses were held at Ϫ140 mV to close the channels.
The extracellular GsMTx4 effect (facing the external side of ion channels) was tested by application in the pipette solution using the standard back-fill method established previously (37,38). In brief, the tips of the electrodes were first filled with the normal pipette solution and then backfilled with the same solution containing a concentration of GsMTx4. The intracellular GsMTx4 effect (facing the internal side of channels) was examined by directly perfusion of GsMTx4 in the bath solution. No significant difference was observed for SAKcaC recorded from chick ventricular myocytes and CHO cells. The intracellular GsMTx4 effect (facing the internal side of channels) was examined by direct perfusion of GsMTx4 in the bath solution. Membrane stretch was reached by negative pressure applied to the patch pipette using a pneumatic transducer tester (DPM-IB, BioTek Instruments Inc., Winooski, VT) (31,33). All experiments were performed at room temperature (22-25°C).

Single channel analysis
Single channel conductance was determined by the slope of current-voltage (I-V) curves from Ϫ80 to ϩ80 mV, where I-V data could be fitted to a linear line. Amplitude histograms were measured in inside-out patches with 1-4 channels. Histograms were fitted with a Gaussian function using the software pCLAMPFIT. The P o (%) was determined as described previously (51,53). The P o was calculated using the area under each peak (aj) at each current level (j) in the histogram along with the number of channels (N) as shown in Equation 1, The channel number (N) in each patch was defined as the largest open channel numbers that can be observed at ϩ80 mV under 1 mM [Ca 2ϩ ] i . To assess the effect of GsMTx4 on the kinetic properties of SAKcaC gating, patches with only one single channel were used. The opening and closing transitions of channels were detected by the half-amplitude threshold method after Gaussian filtering at 1 kHz. Open or closed channel events with durations of Ͻ0.6 ϫ 10 Ϫ3 ms were eliminated from the data set, of which this amount of "flickering" increases with depolarization but was not changed either by membrane stretch or by peptide inhibition. These events were log binned into open-or closed-time histograms for better understanding the mechanism of gating states (10 bins/decade (41,42)). The open and closed (or inhibited) times were fitted to exponential functions using the Marquardt-Levenberg algorithm (Igor Pro, WaveMetrics). To minimize the number of single channel states, the saturating [Ca 2ϩ ] i concentration of 1 mM was used (43).

Immunohistochemistry
Newborn Kunming mice were used to detect the expression of SAKcaC (BK-STREX variant) in the hearts. All animal protocols were approved by the Animal Experiment Administration Committee in Xuzhou Medical University.
Heart slides were prepared and stained as described previously (54). In brief, mouse hearts were fixed in 4% paraformaldehyde, equilibrated in 10% sucrose, and embedded in OCT. Then cryosections of 6 -10 m thick were prepared and incubated overnight with individual antibodies at 4°C, followed by 1-h secondary antibody incubations at 37°C. The slides were mounted with or without the unclear counterstain (DAPI, Beyotime).

Simulation of GsMTx4 interaction with lipid bilayer
The peptide-lipid bilayer system was built by CHARMM-GUI (55,56). The coordinates for GsMTx4 (PDB code 1TYK) were obtained from RCSB Protein Data Bank (18). The simulation programs, conditions, and water models were basically the same as those described previously by Nishizawa et al. (22,40). Specifically, the system was composed of 64/64 DPPC bilayers (64 molecules in the upper/lower leaflets), one peptide, and 3895 H 2 O along with Cl ions that were used to neutralize the positive charges of the peptides. The XYZ axes of the simulation box (Å) was 65 ϫ 65 ϫ 110, and the z-axis was defined as the bilayer normal. GsMTx4 com was embedded into DPPC in the bilayer center (the plane that is parallel to the membrane and contains the com of the membrane). MD simulation was performed with an 80-ns production run followed by an energy minimization and an equilibration run. The temperature was set at 323.15 K with Nose-Hoover coupling. The pressure was controlled by the Parrinello-Rahman at 1 atm with the independent (semiisotropic) coupling in the xy and z directions. For the simulations containing a bilayer, the bond lengths of z-axis were constrained with LINCS. For the simulations of GsMTx4 interaction with lipid bilayer under the hyperpolarized/depolarized conditions, the systems were settled in the electric fields along the z axis as used by Su and Guo (57). A time step of 2 fs was used. Trajectories were saved every 5 ps for analysis. The simulation was carried out at the National Supercomputer Center, LvLiang, China. All molecular images were made with Visual Molecular Dynamics.

Data analysis
Data acquisition and analysis were carried out using pClamp9 (Molecular Devices), PULSEϩPULSEFIT TAC 4.0 (HEKA Elektronik), ANA.EXE and Origin 7.5 software. Data in all figures are presented as means Ϯ S.E. Statistical significance was evaluated by a two-sample t test and p Յ 0.05 was considered as significant.

Antibodies and chemicals
A SREX-specific antibody (anti-STREX) was raised in rabbits targeting STREX-exon in BK variants containing the STREX insert. Primary antibody for BK was raised in guinea pigs against the C-terminal end of the BK channel encoded by the Kcnma1 gene (catalog no. AGP-014, Alomone Labs). The surface membrane was labeled by WGA (Thermo Fisher Scientific). The peptides GsMTx4 and VsTx3 were purchased from Alomone Labs. The aqueous stock solutions were prepared at 5 or 10 mM and kept at Ϫ20°C. An appropriate amount of the stock aliquot was diluted to the concentration used on the day of the experiment. The working solutions had concentrations of 50, 100, and 500 nM and 7.5 and 10 M for GsMTx4 and 20, 100, and 300 nM for VsTx3. Other chemicals were purchased from Sigma.