Adaptation Independent Modulation of Auditory Hair Cell Mechanotransduction Channel Open Probability Implicates a Role for the Lipid Bilayer

Abstract

The auditory system is able to detect movement down to atomic dimensions. This sensitivity comes in part from mechanisms associated with gating of hair cell mechanoelectric transduction (MET) channels. MET channels, located at the tops of stereocilia, are poised to detect tension induced by hair bundle deflection. Hair bundle deflection generates a force by pulling on tip-link proteins connecting adjacent stereocilia. The resting open probability (P_open) of MET channels determines the linearity and sensitivity to mechanical stimulation. Classically, P_open is regulated by a calcium-sensitive adaptation mechanism in which lowering extracellular calcium or depolarization increases P_open. Recent data demonstrated that the fast component of adaptation is independent of both calcium and voltage, thus requiring an alternative explanation for the sensitivity of P_open to calcium and voltage. Using rat auditory hair cells, we characterize a mechanism, separate from fast adaptation, whereby divalent ions interacting with the local lipid environment modulate resting P_open. The specificity of this effect for different divalent ions suggests binding sites that are not an EF-hand or calmodulin model. GsMTx4, a lipid-mediated modifier of cationic stretch-activated channels, eliminated the voltage and divalent sensitivity with minimal effects on adaptation. We hypothesize that the dual mechanisms (lipid modulation and adaptation) extend the dynamic range of the system while maintaining adaptation kinetics at their maximal rates.

SIGNIFICANCE STATEMENT Classically, changes in extracellular calcium and voltage affect open probability (P_open) through mechanoelectric transduction adaptation, and this mechanism is the only means of controlling the set point of the channel. Here, we further characterize the effects of extracellular calcium and voltage on the channel and for the first time determine that these manipulations occur through a mechanism that is independent of fast adaptation and involves the lipid bilayer. These data additionally demonstrate that effects on P_open are not enough to characterize adaptation and thus may clarify some of the discrepancies within the literature as to mechanisms underlying adaptation.

Introduction

Sound or head motion induce mechanical stimulation of stereociliary hair bundles protruding from the apical surface of inner ear sensory hair cells (Hudspeth and Corey, 1977). Hair bundle deflection applies force to tip-link proteins that connect the stereocilia of adjacent rows (Pickles et al., 1984). This force is coupled to mechanoelectric transduction (MET) channels located near the tops of the shorter stereocilia and regulates their open probability (P_open) (Beurg et al., 2009).

The auditory system is able to detect movement at atomic dimensions (Bialek, 1987). To maximize the sensitivity of the system, most auditory hair cells have channel open probabilities set at the steepest part of their activation curve [current–displacement (I–X) curve], so ∼50% of the channels are open at rest (Ricci et al., 1998; Farris et al., 2006; Johnson et al., 2011). An adaptation process regulates the set point of the activation curve to maintain sensitivity and extend the dynamic range of the hair bundle (Eatock et al., 1987; Gillespie and Cyr, 2004; Fettiplace and Kim, 2014). Tonotopic variations in adaptation kinetics contribute to the mechanical filtering of incoming signals (Ricci and Fettiplace, 1997; Kennedy et al., 2003; Ricci et al., 2005). In lower vertebrates (e.g., turtles and frogs), calcium (Ca²⁺) entry through the MET channel drives two components of adaptation: (1) Ca²⁺ binding to calmodulin modulates a slow motor process; and (2) Ca²⁺ acts at an unknown intracellular site to regulate a fast adaptation process (Crawford et al., 1989, 1991; Assad and Corey, 1992; Walker and Hudspeth, 1996; Ricci et al., 1998; Wu et al., 1999). The current view is that adaptation equilibrates the local force and intracellular Ca²⁺ to determine the channel set point. Recent work in the mammalian cochlea suggests that the fast component of adaptation, defined as adaptation occurring in <5 ms, is independent of Ca²⁺ and that there is very little slow adaptation (Peng et al., 2013).

The resting P_open of the MET channel is an indicator of the set point of the steady-state activation curve. The set point is sensitive to extracellular Ca²⁺ and membrane potential (Corey and Hudspeth, 1983; Eatock et al., 1987; Assad et al., 1989; Crawford et al., 1989, 1991; Ricci et al., 1998; Farris et al., 2006; Johnson et al., 2011; Corns et al., 2014). Classically, adaptation is the sole regulator of P_open, so that changes in P_open are often used as indicators of the state of adaptation (Corns et al., 2014, 2016; Fettiplace and Kim, 2014). Similar to lower vertebrates, P_open in mammalian auditory hair cells is also modulated by extracellular Ca²⁺ and voltage (Beurg et al., 2010; Johnson et al., 2011; Zampini et al., 2011; Peng et al., 2013; Corns et al., 2014). However, as fast adaptation appears independent of Ca²⁺, what mechanism regulates P_open?

Here, we show that the effects of voltage and extracellular Ca²⁺ modulate a mechanism that acts via the lipid bilayer. The set point of these MET channels is modulated by extracellular divalent ions with a selectivity based on their ionic radius and by membrane potential. GsMTx4, which is an amphiphilic modifier of stretch-activated channels, can inhibit the effect of divalents and voltage while fast adaptation is intact, therefore, by itself, P_open is not a unique readout of adaptation. This new mechanism we discovered may enhance the dynamic range of the cell while preventing saturation of adaptation.

Materials and Methods

Preparation and recordings.

Animals were killed by decapitation using methods approved by the Stanford University Administrative Panel on Laboratory Animal Care. Organs of Corti were dissected from postnatal day 6 (P6) to P10 Sprague Dawley rats (large majority of experiments used P7–P8) of either sex and placed in recording chambers as described previously (Beurg et al., 2009). Tissue was viewed using a 60 or 100× (1.0 numerical aperture; Olympus) water-immersion objective with a digital Rolera XR (Qimaging) or Phantom Miro 320s (Vision Research) camera on a BX51 microscope (Olympus). Tissue was dissected and perfused with extracellular solution containing the following (in mm): 140 NaCl, 2 KCl, 2 CaCl₂, 2 MgCl₂, 10 HEPES, 2 creatine monohydrate, 2 Na-pyruvate, 2 ascorbic acid, and 6 dextrose, pH 7.4 (300–310 mOsm). In addition, an apical perfusion, with pipette tip sizes of 150–300 μm, provided local perfusion to the hair bundles (Fig. 1). In all preparations, the tectorial membrane is peeled off the tissue.

Figure 1.

Schematic of the experimental setup used. Organs of Corti were acutely dissected, placed in the recording chamber, and held in place by two strands of dental floss (black arrowheads). Shown are the positions of the patch pipette and mechanical stimulator. The apical perfusion size ranged from 150 to 300 μm wide and approached the preparation from two possible locations. Scale bar, 150 μm.

Electrophysiological recordings.

Whole-cell patch clamp is achieved on first- or second-row outer hair cells (OHCs) from middle to apical cochlear turns using an Axon 200B amplifier (Molecular Devices) with thick-walled borosilicate patch pipettes (2–6 MΩ) filled with an intracellular solution containing the following (in mm): 125 CsCl, 3.5 MgCl₂, 5 ATP, 5 creatine phosphate, 10 HEPES, 1 cesium BAPTA, and 3 ascorbic acid, pH 7.2 (280–290 mOsm). For the 4 mm Ba²⁺ with 1 mm BAPTA internal solution, 4 mm BaCl₂ was added. For the EGTA internal solution, 1 mm EGTA replaced cesium BAPTA and ascorbic acid increased to 4 mm. The 1.4 Ca²⁺ internal solution contained the following (in mm): 121 CsCl, 3.5 MgCl₂, 3.5 CaCl₂, 3.5 ATP, 5 creatine phosphate, 10 HEPES, and 2 ascorbic acid, pH 7.2 (280–290 mOsm). Experiments were performed at 18–22°C. Whole-cell currents were filtered at 10 or 100 kHz and sampled at 0.05–1 MHz using USB-6356 (National Instruments) controlled by jClamp (SciSoft). Voltages were corrected offline for liquid junction potentials. All experiments used −84 mV holding potential unless otherwise noted. For inclusion, initial MET currents >600 pA in 2 mm extracellular Ca²⁺ and Mg²⁺ were required.

Hair bundle stimulation.

Borosilicate pipettes were fire polished to shapes that matched the hair bundle structures for OHCs. Probes were driven by a piezoelectric stack [AE0505D08F (Thorlabs) or PSt 150/7x7/7 (APC International)] whose input was filtered using an eight-pole Bessel filter (L8L 90PF; Frequency Devices) at 10–30 kHz and variably attenuated (PA5; Tucker Davis) before being sent to a high-voltage/high-current amplifier to drive the piezoelectric stack. Experiments with the Pep Intl GsMTx4 used a fluid jet stimulation. Position of the hair bundle was determined using high-speed imaging of the stimulation at 10,000 frames/s using the Phantom Miro 320s and extracted using a Gaussian fit to a bandpass-filtered hair bundle image intensity profile.

Extracellular solution perfusion.

N-(Hydroxyethyl)-ethylenediaminetriacetic acid (HEDTA)-buffered solution contained the following (in mm): 150 NaCl, 2 KCl, 3.3 CaCl₂, 4 HEDTA, and 10 HEPES, pH 7.4 (310 mOsm). Free Ca²⁺ concentrations were measured at 20 μm using a MI-600 Ca²⁺ electrode (Microelectrodes) calibrated using Ca²⁺ buffer standards (CALBUF-2; WPI). The extracellular perfusion solutions at 2 mm, 1 mm, 300 μm, and 100 μm Ca²⁺ contained the following: 157 mm NaCl, 2 mm KCl, 10 mm HEPES, pH 7.4, 305–320 mOsm, and 2, 1, 0.3, and 0.08 mm CaCl₂, respectively. These solutions had measured Ca²⁺ concentrations of 2.34, 1.14, 0.41, and 0.15 mm, respectively. The Mg²⁺, Ca²⁺, Sr²⁺, Ba²⁺, Mn²⁺, and Cd²⁺ solutions were composed of the following (in mm): 154 NaCl, 2 KCl, 10 HEPES, and 2 MgCl₂, CaCl₂, SrCl₂, BaCl₂, MnCl₂, or CdCl₂, respectively, pH 7.4 (305–320 mOsm). The Mg²⁺, Sr²⁺, Ba²⁺, Mn²⁺, and Cd²⁺ solutions had measured free Ca²⁺ concentrations of 20, 45, 19, 18, and 43 μm, respectively.

An apical perfusion pipette was used for solution application via a six-inlet manifold (Warner Instruments) using a custom gravity flow perfusion system. Pipette tips were broken to 150–300 μm in diameter to allow perfusion of a wide area around the recorded hair cell, thus ensuring that the concentration of Ca²⁺ at the hair bundle was that in the perfusion solution. The perfusion rate was set by looking at the flow into a bath of water to give a boundary of the flow that did not diverge greatly from the pipette tip size.

The l and d isomers of GsMTx4 (Protonin) was synthesized as described previously (Suchyna et al., 2004). GsMTx4 at 0.5 mg was dissolved into 610 μl of extracellular solution for a 200 μm stock solution and was frozen. The solution was diluted to 2 or 3 μm on the day of the experiment. Data for these two concentrations are pooled. For the Peptides International GsMTx4, 0.11 mg was dissolved directly into 13.5 ml of extracellular solution.

Data analysis.

I–X plots, generated by subtracting leak current defined as the smallest remaining current during the negative steps and normalizing to the peak current, were fit with a double Boltzmann equation:

where Z₁ and Z₂ are the slope factors, and x₀ is the set point.

For mechanical stimulus steps, adaptation time constant fits were obtained at ∼50% peak current using a double-exponential decay:

where τ₁ and τ₂ are the decay constants, and A₁ and A₂ are the respective amplitudes.

The dose–response curve was fit using the following equation:

where A₁ and A₂ are the asymptotes, log(x₀) is the reported K_d, and p is the Hill slope. To generate the activation curve for the fluid jet stimulation, the bundle position at the time of the peak current was used. Activation curves were fit with a double Boltzmann according to Equation 1.

Data were analyzed using jClamp, MATLAB (MathWorks), and Excel (Microsoft). Graphs were created using MATLAB, Origin 8.6, and Adobe Illustrator. The resting mechanosensitive current/maximum mechanosensitive current is P_open, in which we assume we can achieve a P_open of 100%. Mechanosensitive current is the difference between current values elicited from the maximal negative and maximal positive stimulation. Occasionally, the most negative current was identified during the return to rest after the positive stimulation.

Statistical analysis used Student's two-tailed t tests from Excel (Microsoft), except when noted. All p values presented used paired t tests with comparisons within a cell and unpaired unequal variance tests across cell conditions. Significance (p values) is as follows: *p < 0.05, **p < 0.01, ***p < 0.001, and ****p < 0.0001. Data are presented as mean ± SD unless otherwise noted.

Results

Divalent specificity of extracellular Ca²⁺ effect on resting P_open

Lowering extracellular Ca²⁺ increases MET channel resting P_open (Fig. 2a, 3), identifying binding sites for Ca²⁺ that, when bound, lower the resting P_open. Lowering extracellular Ca²⁺ to 20 μm increases the resting P_open to near 50% regardless of internal Ca²⁺ concentration (Fig. 2). Internal solutions containing BAPTA, EGTA, 4 mm barium, or 1.4 mm Ca produced shifts of a similar degree, suggesting that the effector site was extracellular. The magnitude of the shift was variable, with 26 of 53 cells reaching >40% P_open (Fig. 2b). The increase to near 50% P_open is consistent with other reports (Beurg et al., 2010; Johnson et al., 2011). To maximize stability, we tried a variety of perfusion methods. We used the Picospritzer III (Parker Hannifin) for local perfusion with pipette diameters of 3–15 μm placed 5–60 μm from the cell, as well as an apical perfusion with pipette diameters of 150–300 μm located 100–400 μm from the cell. Every method produced scattered results, suggesting a biological origin. We chose to use the large apical perfusion method because this had the largest percentage of shifted cells. The results with 1 mm EGTA differ from a previous report (Johnson et al., 2011) but are in line with a recent report from the same group (Corns et al., 2014), attesting to the inherent variability within the measurements. We generated a dose–response curve of resting P_open versus extracellular Ca²⁺ (Fig. 3). Fitting these data yielded an IC₅₀ of 0.23 ± 0.001 mm (value ± SE) and a Hill slope of 2.8 ± 0.01 (value ± SE) with p < 0.002 (F test). This dose–response curve is different from the blocking action of divalent ions on the MET channel (IC₅₀ of 1.1 ± 0.03 mm, Hill slope of 5.5 ± 6.5, value ± SE; p < 0.007, F test; this IC₅₀ is similar to that reported by Kim et al., 2013), arguing that we are probing a different site.

Figure 2.

Low extracellular Ca²⁺ modulation of P_open is attributable to extracellular binding sites. a, When decreasing extracellular Ca²⁺ to 20 μm, resting P_open increases. A family of step responses for control traces (black; 2 mm Ca²⁺ and 2 mm Mg²⁺ divalent ions) are presented in the left column, and a single protocol with a large negative followed by a large positive stimulation for 20 μm Ca²⁺ (red) are presented to illustrate the resting P_open of the channels. In a normal 1 mm BAPTA internal solution (top traces), resting P_open shifts to ∼50% when decreasing extracellular Ca²⁺. Adding 4 mm Ba²⁺ to the internal solution (bottom traces) had no effect on the ability to shift the resting P_open, suggesting that the site of divalent action is extracellular. M indicates the driving voltage to the piezoelectric stack. The maximum positive stimulation is shown in magenta. b, Summary of resting P_open in 2 mm Ca²⁺ and Mg²⁺ (Hi) and 20 μm extracellular Ca²⁺ (Lo) in different internal solutions. 1B, 1 mm BAPTA (n = 16); 1E, 1 mm EGTA (n = 28); Ba, 1 mm BAPTA with 4 mm Ba²⁺ (n = 4); Ca, 1.4 mm free Ca²⁺ (n = 5, data reanalyzed from Peng et al., 2013). Each point is a different cell with mean ± SD shown.

Figure 3.

Dose–response curves differ for the resting P_open effect and the peak current change. a, A family of curves was generated in the Ca²⁺ concentrations stated above the plots. For the lower concentrations, a single stimulus was used with a large negative stimulation proceeded by a large positive stimulation to estimate the resting P_open. A full family of curves was not generated in these cases because the hair bundle became more susceptible to mechanical damage during stimulation at lower Ca²⁺ concentrations. M indicates the mechanical stimulus used. The maximum positive stimulation is shown in magenta when family of curves was obtained. b, Resting P_open was calculated from curves generated in a, and dose–response curves for P_open were generated using all cells, but some cells died before a complete series was obtained. From lowest to highest Ca²⁺ concentration, n = 12, 15, 17, 17, and 17. IC₅₀, 0.23 ± 0.001 mm; Hill slope, 2.8 ± 0.01 (value ± SE). Additionally, the peak current change was also measured and normalized to the peak current in 2.3 mm Ca²⁺. For the lowest Ca²⁺ concentration, the current was often smaller likely attributable to damage, presumably by breakage of tip-links, so the red fit ignores this data point [IC₅₀, 1.1 ± 0.3 mm; Hill slope, 5.5 ± 6.5 (value ± SE)]. Each data point shows mean ± SD.

To determine whether the extracellular binding sites are specific to Ca²⁺, we tested other alkaline metal ions extracellularly (Fig. 4). We found that, in the presence of 2 mm Ca²⁺, Sr²⁺, or Ba²⁺, the resting P_open remained low, but in the presence of 2 mm Mg²⁺, the resting P_open increased significantly compared with Ca²⁺, Sr²⁺, or Ba²⁺ (unpaired t test, p < 0.001; Fig. 4). These data suggest that the extracellular binding sites are not highly specific to Ca²⁺ and that Mg²⁺ has lower affinity and/or efficacy for these sites. This divalent profile suggests that we are examining global effects, such as surface charge (Träuble and Eibl, 1974).

Figure 4.

Extracellular divalent ions modulate the MET resting P_open. a, A family of step responses recorded in each solution containing 2 mm of the indicated divalent ion (radius of the ions is also indicated; Persson, 2010) shows that only specific divalent ions can prevent a shift in the resting P_open. Low divalents indicate a solution with no added divalent ions and is buffered with HEDTA (see Materials and Methods, HEDTA-buffered solution). The maximum positive stimulation is shown in magenta. b, Summary of the resting P_open plotted against the divalent ion radius (left axis; black symbols; from left to right, Mg²⁺, Mn²⁺, Cd²⁺, Ca, ²⁺ Sr²⁺, and Ba²⁺ with n = 11, 8, 7, 9, 6, and 7, respectively; shown as mean ± SD). A sigmoidal fit (black line) shows a monotonic function of P_open versus ionic radius. The peak mechanically sensitive current normalized against the control current (in 2 mm Ca²⁺ and 2 mm Mg²⁺) shows a non-monotonic change, indicating that the binding sites responsible for the resting P_open are unique from the ones controlling peak current (right axis; blue symbols). Only significance values to neighboring points that indicate the non-monotonicity are shown. Each data point shows mean ± SD. **p < 0.01, ***p < 0.001.

The divalent profile is not typical of the common EF-hand motif for Ca²⁺ binding sites (Chao et al., 1984; Vyas et al., 1989) and suggests that our effects are dependent on ion size. We tested this hypothesis using two divalent transition metal ions (Cd²⁺ and Mn²⁺) whose naked ionic radii (Persson, 2010) lie between Ca²⁺ and Mg²⁺. Consistent with the hypothesis, both of these ions produced intermediate shifts in P_open (Fig. 4). Divalent ions whose radii were larger than Ca²⁺ produced a low resting P_open (<10%), whereas smaller divalent ions had lower affinities and/or efficacy (Fig. 4b, black). The variation of the response with ionic radius shows that we cannot apply classic Debye layer analysis in which the ions are treated as point charges (Israelachvili, 2011).

Ca²⁺ is a permeable blocker of the channel; thus, the conductance of the channel reduces with increasing Ca²⁺ concentration (Fig. 3; Crawford et al., 1991; Ricci and Fettiplace, 1998; Pan et al., 2012). To test whether the permeation site is different from the site affecting P_open, we analyzed divalent ion effects on peak current (Fig. 4b, blue). Normalizing the current amplitude by the control current in 2 mm Ca²⁺ and Mg²⁺, permeation had the opposite selectivity profile compared with the P_open; Ba²⁺ and Sr²⁺ had less affinity as permeable blockers than Ca²⁺ or Mg²⁺. The peak current was not monotonic with naked ionic radius, in contrast to P_open. The inhibition of peak current also had a different Ca²⁺ affinity compared with the P_open (Figs. 3, 4b). These differences suggest that the P_open and conductance effector sites are distinct.

We previously used a high internal Ca²⁺ solution to argue for an extracellular site of divalent ion action (Peng et al., 2013). A confounding factor here could be that many processes use Ca²⁺ signaling so that the local concentration near open channels might be different from the concentration in the pipette as a result of local buffering and Ca²⁺ pumps. To decrease these potential Ca²⁺ regulation artifacts, we added 4 mm Ba²⁺ to the 1 mm BAPTA internal solution (Fig. 2a). Ba²⁺ keeps the resting P_open low when applied extracellularly (Fig. 4), does not bind well to EF-hand motifs (Chao et al., 1984), and exhibits slower extrusion by plasma membrane Ca²⁺ ATPase pumps (Thomas, 2009). Even with high intracellular Ba²⁺, resting P_open increased in 20 μm extracellular Ca²⁺ (Fig. 2), consistent with an extracellular site of divalent ion action.

Voltage control of resting P_open

Voltage is known to modulate resting P_open (Assad et al., 1989; Crawford et al., 1989; Zampini et al., 2011; Peng et al., 2013; Corns et al., 2014). To characterize how the P_open changed with voltage, we delivered large negative and positive displacements at different time points during a prolonged depolarization (Fig. 5a,b). The maximum P_open with depolarization to 76 mV was 58 ± 9% (n = 26; Fig. 5d), consistent with other reports that use fluid jet stimuli (Zampini et al., 2011; Corns et al., 2014). With depolarization, P_open rose to a maximum with a single-exponential time constant of 126 ± 89 ms (n = 25; Fig. 5c,d). For short depolarizations as shown in Figure 6a, the recovery of P_open at −84 mV had a time constant of 9.2 ± 5.3 ms (n = 25). As reported previously (Peng et al., 2013), P_open also decreased slowly during a prolonged depolarization on a timescale >1 s (Fig. 5a,c) because of unknown mechanisms that can bias the behavior of subsequent depolarizations. Here, we only investigated the increase in P_open, which was much slower than the 5.4 ± 2.3 ms time constant reported in lower vertebrates (Ricci et al., 2000) and slower than any identified mammalian adaptation process (Stauffer and Holt, 2007; Waguespack et al., 2007).

Figure 5.

Voltage modulates the MET resting P_open. a, In response to a depolarization to 76 mV, the resting P_open of an OHC channel increases to ∼50%. Resting P_open is determined at each point using a fast negative and positive stimulus (M) to get the peak-to-peak mechanically sensitive current. Cyan × symbols indicate the negative displacement current, orange × symbols the resting current, and magenta × symbols the positive displacement current. The current work focuses on the changes within the first second of depolarization (black arrow). Previous work described the longer timescale changes (gray arrow). b, Magnified current response of single fast negative and positive stimuli from the protocol in a for an OHC shows the changes in resting P_open before the depolarization (black), immediately after depolarization (red), and ∼300 ms after depolarization when the resting P_open has peaked (blue). c, Plot of the change in P_open during the depolarization. Traces shown in b are indicated here as filled color symbols. The negative potential data point has an arbitrary negative time value. d, Summary of the maximum P_open reached at 76 mV for different internal solutions (left) and the time constant of a single-exponential fit to the rise of P_open. Each point is a different cell with mean ± SD shown. 1B, 1 mm BAPTA (n = 25); Ca, 1.4 mm free Ca²⁺ (n = 9); Ba, 1 mm BAPTA with 4 mm Ba²⁺ (n = 9). *p < 0.05, ****p < 0.0001.

Figure 6.

Voltage dependence of P_open shift. a, Depolarization to different potentials in an OHC shows the change in resting P_open is dependent on potential. Plots are similar to those shown in Figure 5a but for depolarization to 36, 76, and 116 mV. b, Plot of the resting P_open against depolarized time from the cell in a shows the difference in kinetics and maximum P_open for each potential. Colors correspond to a. Solid lines are single-exponential fits to the data with τ = 201 ms (black), 120 ms (red), and 80 ms (blue). c, Plot of the maximum P_open reached for each potential. Lines connect data points from the same cell. For 36, 76, and 116 mV, n = 11, 11, and 10, respectively. Depolarization to 136 mV was performed in a separate set of cells (n = 11). d, Plot of the single-exponential rise time of P_open plotted for the different voltages, indicating a significant difference in rise time for 76, 116, and 136 mV, which all had the same max P_open. Red symbols indicate the cell shown in a and b. Black bars with error bars represent the mean ± SD. M indicates the mechanical stimulus and V the voltage stimulus. *p < 0.05, **p < 0.01, ***p < 0.001.

We investigated the Ca²⁺ dependence of voltage modulation using 1.4 mm internal Ca²⁺. The maximum P_open was slightly lower, but the time constant for rising to the maximum was slowed approximately threefold (Fig. 5d). To determine the specificity of divalent dependence, we repeated the experiment with a high Ba²⁺ internal solution and observed a similar shift in the resting P_open as we did with high internal Ca²⁺ but with no effect on the rising time constant (Fig. 5d). These data are consistent with the internal Ca²⁺ modulating the depolarization effect.

We measured the voltage dependence of P_open changes by depolarizing OHCs to 36, 76, and 116 mV (Fig. 6). At 36 mV, P_open did not increase as much as 76 or 116 mV (Fig. 6b,c). The maximum resting P_open at 76, 116, and 136 mV was similar (p = 0.48, single-factor ANOVA; Fig. 6c); however, the rising time constants were significantly different (Fig. 6d). The increase in P_open likely does not involve Ca²⁺ influx, because a change in driving force did not change the maximum P_open. However, the change in kinetics may support a modulatory role for intraciliary Ca²⁺.

GsMTx4 modulates P_open

The peptide GsMTx4, originally isolated from a tarantula venom, is a gating modifier of mechanically sensitive ion channels through modulation of lipid tension near the channel (Suchyna et al., 2000, 2004; Bowman et al., 2007; Nishizawa and Nishizawa, 2007; Posokhov et al., 2007a; Kamaraju et al., 2010). We tested the effects of GsMTx4 on MET channels using the d enantiomer of GsMTx4 (d-GsMTx4), a mirror image of the ∼3200 Da natural peptide. At 2–3 μm, d-GsMTx4 shifted the activation curve toward larger displacements in the presence of 2 mm Ca²⁺ and Mg²⁺ (Δx₀ = 119 ± 38 nm, n = 17), and this effect was reversible (from control |Δx₀| = 14 ± 11 nm, n = 15; Figs. 7a,b, 8b). The shift from d-GsMTx4 occurred with a minimal decrease in the peak current (Fig. 8a), suggesting that GsMTx4 acts as a gating modifier, similar to its effect on stretch-activated channels in rat astrocytes and Piezo1 channels (Suchyna et al., 2004; Bae et al., 2011, 2013). The resting current, which is independent of probe placement, also decreased reversibly with GsMTx4 (Fig. 7a arrowheads, c). The 10–90% width of the I–X curve increased slightly (11 ± 20%, n = 17; Fig. 8c). Overall, these data demonstrate that GsMTx4 is a gating modifier of the MET channel.

Figure 7.

GsMTx4 acts as a gating modifier of MET channels. a, Families of curves before (black), during application of 3 μm GsMTx4 (red), and after washout (blue) shows the effect of d-GsMTx4 on MET. Note the decrease in resting current (arrowhead) to near zero in the presence of d-GsMTx4, indicating a shift in the activation curve. Traces were post-filtered at 15 kHz. Mechanical stimulus used here is similar to those presented in Figure 2a, left traces. b, Activation curves for each condition in a (control, black square; GsMTx4, red circle; wash, blue triangle) shows that d-GsMTx4 shifts the activation curve to the left, characteristic of a gating modifier. Data were fit with double Boltzmann curves. c, Resting P_open decreases 52 ± 12% (n = 17) in GsMTx4 and washes out, in line with a shift in the activation curve toward larger displacements. Lines connect data points for a given cell. Red points are from the cell presented in a and b. Black bars with error bars represent the mean ± SD. ***p < 0.001.

Figure 8.

Parameters that characterize GsMTx4 effects. a, Peak current decreases 15 ± 9% (n = 17) in the presence of d-GsMTx4 (left) and washed out. Similar results are observed in the presence of l-GsMTx4 (middle). With commercially obtained GsMTx4 from Peptides International, a larger block was observed (right). b, All compounds observed a shift in the x₀ position in the presence of GsTMx4. c, Changes in the 10–90% width of the activation curve is plotted for the various compounds. The Peptides International data were obtained using the fluid jet and therefore has lower observed widths and x₀ (attributable to the smaller width). This discrepancy in stimulation sizes has been seen by others as well (Kros et al., 2002; Beurg et al., 2014). Lines connect data points for a given cell and black error bars represent mean ± SD. *p < 0.05, **p < 0.01, ***p < 0.001.

We tested whether the effect of GsMTx4 was stereospecific as an indication of its site of action being the lipid bilayer. Previous work argued for GsMTx4 affecting the lipid bilayer on stretch-activated channels because of bilayer partitioning (Posokhov et al., 2007a,b) and similar efficacy of the l and d enantiomers (Suchyna et al., 2004). This argues against a stereospecific, chiral, binding site (e.g., lock and key mechanism). In hair cells, l-GsMTx4 had an efficacy (Δx₀ = 162 ± 91 nm, n = 6; Fig. 8b) similar to d-GsMTx4, consistent with a lipid-based mode of action.

Previous work in hair cells reported two actions of GsMTx4: open channel blocking and gating modification (Beurg et al., 2014). We obtained GsMTx4 through collaboration with the inventors in which the peptide was assayed by mass spectrometry, circular dichroism spectroscopy, calorimetry, and HPLC (Oswald et al., 2002; Ostrow et al., 2003) and functionally on Piezo1 channels (Bae et al., 2011). To compare results from different suppliers, we also obtained a commercially available form of l-GsMTx4 (PCB-4393-s; Peptides International), which elicited a stronger open channel block closer to that reported previously (Beurg et al., 2014) than we observed with compounds obtained from our collaborator (Frederick Sachs, SUNY Buffalo, Buffalo, NY; Fig. 8a). There are apparently differences in structure associated with synthesis, purification, or storage of the peptide from different suppliers. However, we observed gating modifier activity as a shift in the set point of the activation curve with all compounds, as well as a slight increase in activation curve width (Fig. 8b,c).

GsMTx4 differentially modulates P_open and adaptation

The data in Figure 7a suggest that GsMTx4 affects resting P_open without a significant change in the percentage of adaptation. To determine whether P_open modulation and fast adaptation are indeed two separate mechanisms, we tested whether GsMTx4 affects P_open modulation differentially from fast adaptation. Because GsMTx4 shifts the activation curves to larger displacements, changes in resting P_open could appear to minimize the underlying activation curve shifts caused by lowering extracellular Ca²⁺. To reduce this potential artifact, we generated full activation curves in each condition to monitor shifts in the activation curves. We found that the negative activation curve shift in 20 μm Ca²⁺ of 114 ± 28 nm (n = 4; Fig. 9a,b, black to green traces) decreased by 93 ± 13% to 9 ± 16 nm (n = 4) in the presence of GsMTx4 (Fig. 9). Thus, GsMTx4 inhibits the modulation of P_open by extracellular Ca²⁺.

Figure 9.

GsMTx4 blocks the extracellular Ca²⁺ effect. a, Fast families of curves (no averaging) taken in the presence of control solution with 2 mm Ca²⁺ and 2 mm Mg²⁺ divalent ions (top row) and 20 μm Ca²⁺ (bottom row) extracellular solutions illustrate the effect of 2 μm d-GsMTx4 in 2 mm Ca²⁺ and 2 mm Mg²⁺ (right top) and 20 μm Ca²⁺ extracellular solution (right bottom). b, Activation curves for each condition (colors as indicated by traces in a; squares are 2 mm Ca²⁺ and 2 mm Mg²⁺, and circles are 20 μm Ca²⁺) show that the 20 μm Ca²⁺-induced negative activation curve shift is abolished in the presence of d-GsMTx4. c, Summary of cells shows that d-GsMTx4 significantly decreases the shift in x₀ of the activation curve from 2 mm Ca²⁺ and 2 mm Mg²⁺ to 20 μm Ca²⁺ extracellular solutions (93 ± 13%, n = 4). Δx₀ = x₀^{(20μm Ca)} − x₀^{(2 mm Ca)}. Red points are from the cell presented in a and b. Lines connect data points for a given cell and black error bars represent mean ± SD. **p < 0.01.

We next tested whether GsMTx4 affects the voltage regulation of P_open using full activation curves before depolarization and ∼500 ms after depolarization to 76 mV. Again, we found that the negative activation curve shift attributable to depolarization (Fig. 10a,b, black to gray traces) was decreased by GsMTx4 (Fig. 10a,b, red to light red traces). The inhibition of the voltage effect was reversible (Fig. 10b, blue to light blue traces). In four cells, we depolarized in 20 μm extracellular Ca²⁺ with 2 μm GsMTx4 and found that depolarization also had little effect. In all cells tested, the depolarization-induced negative shift was inhibited 71 ± 17% (n = 8) in the presence of GsMTx4 (Fig. 10c). These results are in line with early studies by Marcotti et al. (2002), who showed that 1 μm GsMTx4 caused reduced P_open at depolarized potentials (Palmer and Rosen, 2002). Thus, GsMTx4 can inhibit both extracellular Ca²⁺ and voltage modulation of P_open.

Figure 10.

GsMTx4 inhibits voltage modulation. a, Fast families of curves (no averaging) taken before depolarization (top left) and ∼500 ms after depolarization to 76 mV (bottom left) illustrate the effect of 2 μm d-GsMTx4 (right top) on the activation curve shift attributable to depolarization (right bottom). b, Activation curves show that the x₀ shift attributable to depolarization in control solutions (black trace, −84 mV; gray trace, 76 mV) is blocked with d-GsMTx4 (red trace, −84 mV; light red trace, 76 mV), and this effect can be washed out (blue trace, −84 mV; light blue trace, 76 mV). c, Summary of cells showing that d-GsMTx4 significantly reduced the change in x₀ attributable to depolarization (71 ± 17%, n = 8) and was washed out (4 ± 13% decrease from control, n = 6). Δx₀ = x₀^{(76 mV)} − x₀^{(−84 mV)}. Red points are from the cell presented in a and b. Lines connect data points for a given cell and black error bars represent mean ± SD. **p < 0.01, ***p < 0.001.

Finally, we examined GsMTx4 effects on fast adaptation to assess the independence of adaptation from P_open modulation. The rates and percentage of adaptation in the presence of GsMTx4 were not significantly different from controls (Fig. 11d). We also investigated the effect of GsMTx4 on adaptation using a two-pulse protocol (Fig. 11a). Because lowering extracellular Ca²⁺ or depolarization caused a leftward activation curve shift of 100–150 nm, we mimicked this shift for adaptation by using a negative adaptation step of ∼150 nm. The negative adaptation step served a second purpose of decreasing the likelihood of hair bundle damage by overstimulation because, with GsMTx4, the activation curves are already right shifted. In the two-pulse protocol, control activation curves shifted −126 ± 11 nm (n = 8; Fig. 11b,c). With GsMTx4, the shift decreased 27 ± 9% to −93 ± 16 nm (n = 8; Fig. 11b,c). This effect could be partially washed out in 2–7 min (118 ± 5 nm, n = 8; Fig. 11c). Thus, using comparable activation curve perturbations, the effect of GsMTx4 on adaptation is small and is very different from the near abolition of the effects of extracellular Ca²⁺ and voltage. Our data indicate that P_open modulation by divalents and voltage is a process separate from fast adaptation. Extracellular divalent ions, voltage, and GsMTx4 affect the newly identified P_open modulation mechanism in a manner consistent with a lipid-based mechanism.

Figure 11.

GsMTx4 effect on adaptation is minimal. a, Two-pulse experiments with a negative displacement adaptation step investigate d-GsMTx4 effects on adaptation. Activation curves after the adaptation step are colored lighter. M indicates the mechanical stimulus waveform. b, Activation curves for before (darker traces) and after (lighter traces) the adaptation step show that the adaptation shift is slightly decreased in the presence of 3 μm d-GsMTx4 (red and light red circles) compared with control (black and gray squares). c, Summary of cells shows the adaptation shift is decreased by 27 ± 9% (n = 7). Δx₀ = x₀^(after) − x₀^(before). Red points are from the cell presented in a and b. *p < 0.05, **p < 0.01, ***p < 0.001. d, No significant differences were found in adaptation properties in the presence of d-GsMTx4 from the analysis of step responses like that seen in Figure 7a. Percentage adaptation was calculated as (1 − I_{steady state}/I_peak) × 100. Double-exponential fits of the ∼50% peak current traces do not reveal significant differences in the time constants of adaptation (τ₁ is the faster time constant and τ₂ is the slower time constant) or the percentage contribution of τ₂ [A₂/(A₁ + A₂)]. Lines connect data points for a given cell and black error bars represent mean ± SD.

Discussion

We investigated mechanisms controlling the resting P_open of the MET channel in mammalian auditory hair cells, extending our previous work that identified an extracellular binding site. Here, we demonstrate that these sites are not Ca²⁺ specific. We also show that the depolarization-induced increase in P_open has a voltage dependence in which the amount of depolarization affects the rate of P_open rise. Using GsMTx4, we inhibited P_open modulation with minor effects on the fast component of adaptation, confirming the existence of a second mechanism of P_open modulation. Finally, our results with GsMTx4 are consistent with a mechanism in which P_open modulation occurs through local lipid stress surrounding the MET channel.

Separation of P_open modulation and adaptation

Classically, adaptation is what sets the operating point of the MET channel (Fettiplace and Kim, 2014). There are multiple modes of adaptation with reported time constants on the order of hundreds of microseconds, single milliseconds, and tens of milliseconds in mammalian auditory hair cells, with the slowest time constant purported to be slow myosin motor driven adaptation (Holt et al., 2002; Kennedy et al., 2003; Stauffer and Holt, 2007; Waguespack et al., 2007; Peng et al., 2013; Corns et al., 2014). Historically, P_open changes associated with membrane potential or extracellular Ca²⁺ were considered a manifestation of adaptation and were used to argue that adaptation is Ca²⁺ dependent (Crawford et al., 1991; Assad and Corey, 1992; Corns et al., 2014; Fettiplace and Kim, 2014). Our new data demonstrate that, in mammalian cochlear hair cells, P_open modulation can occur independently of fast adaptation. This is supported by the inhibition of P_open modulation, whereas fast adaptation remains intact in the presence of GsMTx4. Our experiments used 5 ms pulses, which rules out adaptation processes in the hundreds of microseconds to single millisecond range. This process is also likely not associated with the slowest time constant of adaptation, because the rising P_open time constant attributable to depolarization is an order of magnitude slower (Fig. 5). Previously, data from the bullfrog saccule showed a divalent ion sensitivity similar to calmodulin; thus, a calmodulin-mediated regulation of myosin motors was proposed (Corey and Hudspeth, 1983; Chao et al., 1984; Hacohen et al., 1989; Gillespie and Cyr, 2004). However, this hypothesis is not supported in mammalian hair cells because the divalent specificity does not match that of calmodulin (Fig. 4) and the divalent effect is extracellular (Fig. 2), further suggesting that P_open control is not associated with the slowest time constant of adaptation (motor adaptation). With this new mechanism of P_open modulation separate from adaptation, previous data need reexamination because the underlying assumption that only adaptation modulates P_open is no longer valid.

Implications for adaptation

Recently, there is much debate about the mechanisms of adaptation. In mammalian auditory hair cells, our group has shown Ca²⁺-independent adaptation (Peng et al., 2013), whereas others have reported Ca²⁺-dependent adaptation (Corns et al., 2014, 2016; Beurg et al., 2015). The major difference in the two approaches is in the mode of stimulation (stiff probe vs fluid jet), each having merits and caveats (Nam et al., 2015). The 500 μs rise time of the fluid jet precludes accurately assessing the fastest adaptation, which is argued to be Ca²⁺ independent (Peng et al., 2013). Present work suggests that fast adaptation and classical motor adaptation is separate from P_open modulation by extracellular Ca²⁺ and voltage. Therefore, set point changes using these manipulations to support Ca²⁺-dependent adaptation requires reinterpretation. However, still unexplained are the differences with stimulus modality seen in two-pulse experiments and time-dependent current decay, which require additional exploration.

Voltage modulation of mechanotransduction

The data suggest an effect of intracellular Ca²⁺ that is specific to Ca²⁺ as seen by the kinetic changes with various internal solutions (Fig. 5d). The change in kinetics attributable to depolarization (Fig. 6d) suggests a mechanism that lies within the electric field of the membrane. However, we cannot rule out the possibility of a Ca²⁺ dependence on intracellular Ca²⁺ concentration, because stereocilia Ca²⁺ concentrations may differ from cytoplasmic Ca²⁺ concentrations (Delling et al., 2013; Beurg et al., 2015). Based on our results with high internal Ca²⁺, Ca²⁺ modulates the change in P_open attributable to voltage.

Unifying mechanism of resting P_open control

How do our manipulations (divalent ions, voltage, and GsMTx4) affect the channel? The manipulations could affect the lipid–channel interface to change resting forces applied to the channel, the energy for different conformational states, or the energy barrier between channel states (Andersen et al., 1999). GsMTx4 affects the energy landscape by embedding in the membrane, altering local stress, and increasing the energy required to reach the open state (Andersen et al., 1999; Sachs, 2010; Nishizawa et al., 2015). The shift in the activation curve set point and the increase in the activation curve width support this model of GsMTx4 action (Fig. 8). Thus, GsMTx4 alters the force translation to the MET channel. The fact that GsMTx4 can abolish the effects of voltage and divalent ions (Figs. 9, 10) suggests that GsMTx4, when inserted into the lipid, dominates the interaction with the channel. Voltage and divalent ions can also create mechanical changes in the lipid (Akinlaja and Sachs, 1998; Beyder and Sachs, 2009). These effects of voltage and divalent ions may be acting through the electric field in the membrane (Fig. 12a; Zhang et al., 2001). For negative surface potentials, decreasing extracellular divalent ions changes the membrane field similar to depolarization. Mobile counter-ions along the membrane surface repel each other laterally, creating membrane tension (the Lippmann effect; Petrov and Sachs, 2002; Sachs et al., 2009) or the converse flexoelectric effect (Todorov et al., 1994; Petrov and Sachs, 2002; Beyder and Sachs, 2009; Harland et al., 2010).

Figure 12.

Schematic drawing of the mechanisms involved in MET channel modulation. a, Top left, Ca²⁺ is normally adsorbed into the outer leaflet of the membrane, causing decreased force sensed by the channel, leading to a low P_open. Top right, During depolarization, Ca²⁺ ions are repelled from the membrane, causing increased force sensed by the channel leading to a higher P_open. Bottom left, Decreasing the extracellular divalent ions from the extracellular side decreases the adsorbed Ca²⁺ ions, leading to increased force sensed by the channel (higher P_open). Bottom right, GsMTx4 inserts into the outer leaflet, decreases force sensed by the channel (lower P_open), and shunts any other lipid effects. b, Illustration of two independent mechanisms that control the midpoint of the activation curve. Adaptation controls the position (shaded gray regions) around the set point (black line), whereas extracellular divalent ions and voltage shift the set point (red lines). c, Schematic drawing of the three mechanisms that control channel P_open. Red arrow represents myosin motors that provide constant tension. Blue arrows represent adaptation, which has an unknown mechanism. Orange arrows represent the lipid modulation of the channel resting P_open.

Multivalent ions can change membrane mechanical properties by binding between adjacent lipids (Ito and Onishi, 1974; Onishi and Ito, 1974; Träuble and Eibl, 1974; Jacobson and Papahadjopoulos, 1975; Lösche and Möhwald, 1989; Ermakov et al., 2010). This can be a large effect producing lipid phase separations that may exhibit the divalent ion specificity as we observed (Fig. 4; Ito and Onishi, 1974). Forces at the lipid–channel interface affect mechanosensitive channels (Perozo et al., 2002; Sukharev and Sachs, 2012). The line tension between the channel and the lipids (Markin and Sachs, 2004; Wiggins and Phillips, 2005) is a component of the work the channel must perform to open (Anishkin and Kung, 2013). Soft lipids would provide a smaller energy barrier to overcome so that a decrease in extracellular Ca²⁺ that reduces the lipid stiffness makes it easier for channels to open.

Alternative possibilities

Our results are consistent with a common pathway involving the lipid (Figs. 9, 10), but there are other potential mechanisms. Changes in extracellular Ca²⁺ will change the local electrostatics and perhaps affect proteins that are located near the channel, such as Tmhs, Tmie, Tmc1, Tmc2, or the tip-link (Ahmed et al., 2006; Kazmierczak et al., 2007; Kawashima et al., 2011; Xiong et al., 2012; Beurg et al., 2014; Marcotti et al., 2014; Zhao et al., 2014). However, the voltage sensitivity suggests that tip-links are not involved because there are no voltage drops on the extracellular links (Ahmed et al., 2006; Kazmierczak et al., 2007). Another potential mechanism could be an effect on a slower adaptation process that we have not directly tested.

Implications for hair cell MET

In the mammalian auditory system, three separate mechanisms likely modulate the MET channel (Fig. 12c). The first is a motor, likely Myosin VIIa or Ic, that climbs along the sides of stereocilia to generate tension in the tip-link (Holt et al., 2002; Kros et al., 2002; Stauffer et al., 2005; Grati and Kachar, 2011). This motor does not represent fast adaptation as classically defined with a stiff probe and could be the Ca²⁺-dependent mechanism observed by others (Corns et al., 2014), in which case it would be a form of slow adaptation that may be inhibited by the mass of the stiff probe. Fast adaptation allows the system to increase its dynamic range by reducing low-frequency gain (Fig. 12b). This mechanism acts to decrease the net force applied to the channel, but it is not solely responsible for setting the resting state of the channel as thought previously. The lipid modulation described here is a new mechanism that clearly plays a significant role in setting the resting P_open (Fig. 12). This system reduces the chance of saturating the fast adaptation mechanism and thus allows it to maintain its role as a high-pass filter. In the cochlea, this mechanism would probably not dynamically regulate the resting P_open but would rather set the resting P_open in the steady state to position the MET channel in the range of greatest sensitivity. Our data support a new model for regulation of the MET activation curve in which a molecular motor provides a constant tip-link tension, adaptation provides fast dynamic adjustment of force applied to the channel (AC response), and Ca²⁺ ion concentration coupled with membrane potential modulates the resting P_open via a lipid-based mechanism (DC response).