Rat Auditory Inner Hair Cell Mechanotransduction and Stereociliary Membrane Diffusivity Are Similarly Modulated by Calcium.

The lipid bilayer plays a pivotal role in force transmission to many mechanically-gated channels. We developed the technology to monitor membrane diffusivity in order to test the hypothesis positing that Ca2+ regulates open probability (P o) of cochlear hair cell mechanotransduction (MET) channels via the plasma membrane. The stereociliary membrane was more diffusive (9x) than the basolateral membrane. Elevating intracellular Ca2+ buffering or lowering extracellular Ca2+ reduced stereociliary diffusivity and increased MET P o. In contrast, prolonged depolarization increased stereociliary diffusivity and reduced MET P o. No comparable effects were noted for soma measurements. Although MET channels are located in the shorter stereocilia rows, both rows had similar baseline diffusivity and showed similar responses to Ca2+ manipulations and MET channel blocks, suggesting that diffusivity is independent of MET. Together, these data suggest that the stereociliary membrane is a component of a calcium-modulated viscoelastic-like element regulating hair cell mechanotransduction.

Introduction

Mechanotransduction (MET) is a key step in many sensory processes, including touch, proprioception, pain, osmoregulation, cell adhesion, balance, and hearing. A common feature to each of these is the activation of mechanosensitive ion channels. In principle, force gating of mechanosensitive ion channels can occur through the lipid bilayer (“force-from-lipids” principle) or through tether-forming proteins that connect the channel to the cytoskeletal and/or extracellular matrix (“force-from-filament” model) to translate mechanical stimulus into electrical or biochemical signals (Ursell and Blount, 2008; Martinac and Poole, 2018). These mechanisms are not mutually exclusive, and the lipid bilayer can play a role in force transmission in both scenarios.

The lipid bilayer is indirectly implicated in modulating cochlear hair cell MET. GsMTx4, a lipid-mediated stretch-activated channel modifier (Suchyna et al., 2000; Bowman et al., 2007), shifts the MET activation curve rightward, decreasing the resting open probability (Po) of MET channel, and inhibits the leftward shift induced by reducing extracellular divalent ions and membrane depolarization (Peng et al., 2016). Increasing intracellular Ca2+ buffering also shifts the activation curve leftward, increasing the resting Po independent of the extracellular effects. PIP2, an endogenous lipid, modulates MET channel conductance, open probability, and kinetics (Hirono et al., 2004; Effertz et al., 2017; Cunningham et al., 2020). And finally, computational models support a role for the lipid bilayer in shaping the force exerted on hair cell MET channels (Powers et al., 2012, 2014; Gianoli et al., 2017). Present work directly assesses membrane diffusivity, one component of membrane mechanical properties.

Hair cells detect mechanical stimuli via displacement of their sensory hair bundle. The apically located hair bundle comprises an array of stereocilia, actin-filled microvilli, that increase in height in a staircase-like manner. An extracellular tether, the tip link, connects stereocilia columns such that deflection toward the tallest stereocilia row pulls the tip link (Pickles et al., 1984), exerting force that is transmitted to MET channels located at the top of the shorter stereocilia (Beurg et al., 2009). MET channels are mechanically coupled to the tip link through several membrane-spanning proteins and the lipid bilayer and likely also with proteins spanning from the plasma membrane to the cytoskeleton (Gillespie and Müller, 2009; Peng et al., 2011). The upper tip link is similarly coupled both to the membrane and cytoskeleton through a variety of proteins, likely including myosin molecules. How the lipid bilayer influences mechanical coupling between these molecules and thus modulates force transmission to the MET channel is a focus of the present work.

One method for monitoring membrane properties uses FRAP (fluorescence recovery after photobleaching) to measure lateral diffusivity of fluorescent particles as a measure of membrane properties. Lateral fluorescent particle diffusivity, which we refer to as membrane diffusivity for simplicity, is influenced by factors such as the lipid and protein composition, membrane order (lipid packing efficiency), membrane curvature, membrane tension (stress in the plane of the membrane), and hydrophobic thickness (Butler et al., 2001; Haidekker et al., 2001; Blood et al., 2005; Reddy et al., 2012). FRAP provides an average across a large volume of membrane (optically limited) over a relatively slow time course. FRAP does not provide information regarding fast changes in membrane properties or nanodomains created by individual proteins, but measurements can be influenced by underlying specializations. Changes in these membrane properties might alter diffusivity and indirectly affect membrane protein structure and function (Tillman and Cascio, 2003). Diffusivity is inversely related to membrane viscosity, but the specific relationship is complex and dependent on many of the properties described above.

We explored stereociliary membrane diffusivity using two-photon FRAP of the lipid probe di-3-ANEPPDHQ under conditions known to alter MET channel resting Po in rat inner hair cells (IHCs) and their stereocilia. We found that the stereociliary membrane is about nine times more diffusive than the IHC soma, consistent with previous findings (Boutet de Monvel et al., 2006). Unlike the soma, stereociliary diffusivity is sensitive to Ca2+ and voltage, yet independent of MET channel activity. Our data support a mechanism whereby the stereociliary membrane diffusivity and MET channel Po are co-modulated such that increases in diffusivity are paralleled by decreases in MET channel Po.

Results

One-Dimensional Diffusion in the Stereociliary Membrane

Two-photon FRAP was used to evaluate membrane diffusivity in rat apical IHCs. The tallest (row 1) and middle (row 2) rows of the sensory hair bundle were evaluated using the membrane dye, di-3-ANEPPDHQ, as illustrated in Figures 1A–1D. The laser beam was focused at the top of a stereocilium which allowed for the average fluorescence intensity of a 1.9 μm long region (axial resolution of our optical system) of that stereocilium to be monitored over time. Figure 1C shows a set of stereocilia before and after photobleaching, with the bleached, reference, and background regions highlighted. Following background subtraction and photobleaching correction, a normalized fluorescence intensity vs time plot (FRAP curve) was generated with t = 0 representing the first post-bleach measurement (Figure 1E). To estimate the diffusion coefficient D, the FRAP curve was fitted with a stereocilia-specific one-dimensional diffusion model (Figure 1E). Details of FRAP parameters and diffusion models are presented in Methods and Figures S1–S3. The filled appearance of the stereocilia initially questioned dye localization to the plasma membrane. Further investigations revealed “excitation photoselection effect” (Parasassi et al., 1997; Bagatolli, 2006) that was corrected with a half-wave plate as illustrated in Figure S3. D values were similar with or without the half-wave plate rotation, confirming diffusion was only occurring through the membrane (Figure S3H).

Figure 1

Setup of Two-Photon FRAP of Membrane Dye di-3-ANEPPDHQ in IHC Stereocilia

(A and B) (A) Bright field and (B) two-photon sections of the same IHC bundles stained with di-3-ANEPPDHQ and oriented vertically, focused at the tip of the row 1 (top) and row 2 (bottom) stereocilia. Scale bar represents 5 μm.

(C) Representative FRAP experiment on the row 1 stereocilia with t = 0 indicating the first image taken post-bleaching. The red, blue, and brown circles define the bleached (BL), reference (REF), and background (BK) regions, respectively. Scale bar represents 1 μm.

(D) Depiction of an IHC bundle with red dashed lines indicating the approximate position of the focal plane at the tip of the row 1 and row 2 stereocilia for FRAP experiments. Only one stereocilia was bleached at a time. The gradient filling indicates the degree of bleaching with white representing the highest degree of bleaching.

(E) The experimental FRAP curve is fitted with a one-dimensional diffusion model to estimate the diffusion constant D, yielding a value of 4.6 μm2/s with a fitting error of 1.2%.

(F) The average FRAP curves (mean ± SD) and the corresponding fitted curves (red trace) for the row 1 stereocilia (n = 43, black filled squares), row 2 stereocilia (n = 21, black unfilled squares), and the cell body (n = 8, gray filled squares).

(G) Time constant τ derived from fitting the data, showing significantly faster recovery for the cell body than the stereocilia. Row 1 recovery was significantly faster than that of row 2 (t test, p < 0.001).

(H) Diffusion constants extracted from the model were similar for row 1 and row 2 stereocilia and were significantly larger than those of the soma (t test, p < 0.001).

(I) FRAP data points and fitted curves comparing control stereocilia to stereocilia treated with 50 mM water-soluble cholesterol complex to load cholesterol into the membrane. The estimated D values for the given examples are 5 μm2/s (fitting error = 2.2%) for the control stereocilia and 3.6 μm2/s (fitting error = 1.7%) for the cholesterol loaded stereocilia.

(J) Cholesterol loading resulted in significant reduction in the diffusivity in both row 1 and row 2 stereocilia (t test, p < 0.001).

Boxes in (G), (H), and (J) represent the SD, and the star symbol indicates the mean. Each data point corresponds to a stereocilium (for row1 and row2) or a cell (for soma). ∗∗∗p < 0.001.

Stereociliary Membrane Is Highly Diffusive Compared to the Soma

We compared FRAP measurements from the stereocilia to those of the IHC's soma, near the top of the nucleus, as a control for stereociliary specific behavior. Figure 1F presents average FRAP curves from the soma and both rows of stereocilia illustrating differences in recovery time courses. The soma recovered fastest while the row 1 stereocilia recovered slowest. In agreement with the FRAP curves, the time constants were significantly faster (t test, p < 0.001) for the soma, 0.2 ± 0.05 s (n = 8 cells) compared to the row 2 stereocilia, 1.1 ± 0.2 s (n = 43 cilia), which are faster again (t test, p < 0.001) than the row 1 stereocilia, 4.5 ± 1 s (n = 21 cilia) (Figure 1G). However, the time constants of recovery are greatly influenced by the geometry of the investigated system and the bleach extent. Thus, it is critical to develop a morphologically accurate model to correct for these parameters in order to extract a biologically relevant diffusion constant (see Supplemental Information). The estimated diffusion constants between stereociliary rows were not significantly different from each other, with values of 5.1 ± 1.1 μm2/s for row 1 stereocilia and 5.6 ± 0.8 μm2/s for row 2 stereocilia (Figure 1H). However, the stereociliary diffusion constant was about nine times greater than that of the soma of 0.58 ± 0.13 μm2/s (Figure 1H). These results are in agreement with the previous finding of Boutet De Monvel et al. (2006) who report larger diffusion constants for outer hair cell (OHC) stereocilia compared to soma (Table S1). Despite experimental (i.e., bleach extent, bleach time, dye selection) and morphological differences between our observations and those of Boutet De Monvel et al. (2006), resulting in recovery time constants differing by ∼25 fold, the diffusion constants of the soma are comparable and smaller than those of the stereocilia (Figure S2E). The marked difference between stereocilia and soma might imply a functional relevance to the diffusivity.

FRAP Sensitivity

FRAP sensitivity was evaluated by increasing membrane cholesterol levels, a manipulation previously shown to reduce membrane diffusivity (Nguyen and Brownell, 1998; Organ and Raphael, 2009). Figures 1I and 1J show that cholesterol addition slowed the time course of FRAP curves compared to control, with corresponding reductions in the diffusion constant for both stereociliary rows (t test, p < 0.001), from 5.1 ± 1.1 μm2/s to 4.2 ± 0.8 μm2/s for row 1 stereocilia and 5.6 ± 0.8 μm2/s to 3.8 ± 0.9 μm2/s for row 2 stereocilia. Similarly, the time constant showed significant increases (t test, p < 0.001) with cholesterol addition in both stereociliary rows, with no significant effect (t test, p > 0.2) on the mobile fraction of recovery (Figure S4). Thus, the membrane enrichment with cholesterol decreases membrane fluidity, consistent with previous reports (Owen, 2015; Ayee and Levitan, 2016). This result further demonstrates that FRAP measurement implemented at the level of individual stereocilium could detect the reduction in diffusivity of di-3-ANEPPDHQ in the stereocilia membrane. And finally, the cholesterol-induced change of ∼25% is modest compared to the difference between soma and stereocilia (x9), suggesting additional factors are needed to explain the differences between soma and stereocilia.

Effect of Internal Ca2+ Buffering on Membrane Diffusivity

The lipid membrane is implicated in modulating the resting Po of the hair cell MET channel (Peng et al., 2016; Effertz et al., 2017). Hypothesizing that the viscoelastic properties of the stereocilia membrane impact MET channel properties, we investigated lipid membrane diffusivity as a first step in probing membrane mechanics, under conditions that alter MET channel resting Po. As diffusivity is an indication of viscosity, any correlation between diffusivity and MET channel resting Po can only suggest a common regulatory site for the two processes as viscosity effects can only manifest themselves in a non-steady state condition, i.e., during stimulation.

Elevated internal Ca2+ buffering increases MET channel resting Po (Crawford et al., 1991; Ricci and Fettiplace, 1997). To examine the effect of internal Ca2+ buffering on the stereocilia membrane diffusivity, we compared FRAP results from the stereocilia of cells patched with 0.1 mM BAPTA (1,2-bis(o-aminophenoxy)ethane-N,N,N',N'-tetra-acetic acid) and 10 mM BAPTA in the internal solution. Increasing internal Ca2+ buffering from 0.1 mM BAPTA to 10 mM BAPTA with a holding potential of −84 mV increased resting Po significantly (t test, p < 0.001) from 3.8 ± 1.4% (n = 65 cells) to 14.6 ± 4.6% (n = 47 cells) (Figures 2A and 2B), similar to other reports (Fettiplace, 1992; Ricci and Fettiplace, 1997; Peng et al., 2013).

Figure 2

Internal Ca2+ Buffering Reduces Membrane Diffusivity of the Stereocilia

(A) Voltage-clamped (V) MET current responses of IHCs to fluid jet sinusoidal stimulus (M) with 0.1 mM BAPTA (black) and 10 mM BAPTA (red) internal solutions, with a holding potential of −84 mV.

(B) Summary box plots for the resting open probability (Po) measured for 0.1 mM BAPTA (n = 65 cells) and 10 mM BAPTA (n = 47 cells) internal solutions. Each data point corresponds to a cell.

(C and D) Average FRAP data points (mean ± SD) and fitted curves from the row 1 (C) and row 2 (D) stereocilia comparing 0.1 and 10 mM BAPTA internal solutions. Insets show slower FRAP curves with 10 mM compared to 0.1 mM BAPTA internal solutions.

(E) Summary box plots show that the diffusion constant was significantly lower with 10 compared to 0.1 mM BAPTA for both stereocilia rows (t test, p < 0.001). There was a significant reduction in diffusivity (t test, p < 0.001) of row 2 with 0.1 mM BAPTA compared to unpatched cells. Also, application of 1 mM curare had no effect on the diffusivity of row 1 and row 2 with and without patching (t test, p < 0.001). Each data point corresponds to a stereocilium.

(F) Summary of diffusion constants from the soma shows no significant difference between unpatched cells and patched with 10 mM BAPTA internal solution. Different symbols correspond to different cells.

Boxes in (B), (E), and (F) represent the SD, and the star symbol indicates the mean. 0.1 B, 0.1 mM BAPTA; 10 B, 10 mM BAPTA; NP, not patched; NP CUR, not patched and treated externally with 1 mM curare; 0.1B CUR, patched with 0.1 mM BAPTA internal and treated externally with 1 mM curare. ∗∗∗p < 0.001.

Figures 2C and 2D summarize average FRAP curves measured across the two rows of stereocilia comparing 0.1 mM BAPTA to 10 mM BAPTA with the superimposed model fits for each data set. There is a small but distinct slowing of recovery with higher BAPTA concentrations, similar in scale to that observed with elevating cholesterol. Both stereociliary rows show a similar slowing of recovery with elevated BAPTA. The diffusion constant (Figure 2E) similarly suggests that both stereociliary rows are sensitive to BAPTA concentration in that diffusivity in 10 mM BAPTA (3.9 ± 0.8 μm2/s for row 1 and 3.9 ± 0.5 μm2/s for row 2) and is significantly reduced (t test, p < 0.001) as compared to 0.1 mM BAPTA (4.8 ± 1.1 μm2/s for row 1 and 4.4 ± 0.6 μm2/s for row 2). The time constants also showed a significant increase (t test, p < 0.001) with elevated internal Ca2+ buffering in both stereociliary rows, with no effect on the mobile fraction of recovery (Figure S5).

Role of MET Channel in Modulating Diffusivity

We observed a difference in D values between rows when comparing hair cells prior to obtaining a whole-cell recording to those with 0.1 mM BAPTA internal solution, with row 2 having a statistically significant reduction in D (t test, p < 0.001) while row 1 remained unaffected (Figure 2E). One possible interpretation of this difference is that row 2 has a higher Ca2+ concentration because of the functional MET channels located in them, making row 2 more sensitive to BAPTA. Similarly, the reduced relative change between 0.1 and 10 mM BAPTA for row 2 (19% in row 1 and 11% in row 2) may reflect local buffer saturation in row 2 with 0.1 mM BAPTA because of open MET channels in row 2. To determine whether the difference between the 0.1 mm BAPTA and the non-patched cell was due to changes in MET channel P, we applied curare (1 mM) externally to non-patched hair bundles to block the MET channels (Glowatzki et al., 1997; Farris et al., 2004; Kirkwood et al., 2017). Curare had no effect on membrane diffusivity of stereocilia, arguing MET channels were not driving the difference. We also tested the row 2 membrane sensitivity to Ca2+ entry through MET channels by applying curare after obtaining whole-cell mode (with 0.1 mM BAPTA internally); curare addition again had no effect on the enhanced BAPTA sensitivity of row 2, further arguing that adjustments to MET channel Po were not responsible for changes in diffusivity. The 10 mM BAPTA condition similarly effects both stereociliary rows relative to non-patched cells (24% in row 1 and 30% in row 2), potentially by removing Ca2+ away from the lipid bilayer and altering lipid packing (Ito and Ohnishi, 1974; Melcrova et al., 2016). Thus, we find stereociliary membrane diffusivity to be sensitive to internal Ca2+ buffer but to be independent of MET channel activity.

It is possible that membranes are simply sensitive to Ca2+ buffers, specifically BAPTA, and that stereocilia are not unique in this property. We compared IHC basolateral membrane sensitivity to Ca2+ buffers as a test of this hypothesis and found that unlike stereocilia, the basolateral membrane diffusivity was unaffected by 10 mM BAPTA (Figure 2F). Therefore, stereocilia appear to be uniquely sensitive to Ca2+ buffering. It remains plausible that a common underlying mechanism influences MET channel Po and membrane diffusivity and that MET channel activity does not impact this underlying mechanism.

Effect of Extracellular Ca2+ on Membrane Diffusivity

Lowering extracellular Ca2+ increases MET channel resting Po (Crawford et al., 1991; Peng et al., 2016; Effertz et al., 2017) and its effect on Po is blocked by GsMTx4, a compound that interferes with force transmission through the bilayer to MET channels (Suchyna et al., 2000; Bowman et al., 2007; Peng et al., 2016). We investigated the effect of extracellular Ca2+ on the diffusivity of di-3-ANEPPDHQ with 0.1 mM BAPTA or 10 mM BAPTA as the internal Ca2+ buffer. Lowering extracellular Ca2+ from 2 to 0.02 mM increased MET channel resting Po (Figures 3A and 3B) from 3.1 ± 1.0% to 9.1 ± 6.2% with 0.1 mM BAPTA (paired t test, p < 0.001) and from 12.7 ± 4% to 20.2 ± 10.3% with 10 mM BAPTA (paired t test, p < 0.01). Lowering extracellular Ca2+ to 20 μM increases (paired t test, p < 0.001) the maximum MET current (Figure 3C), from 0.7 ± 0.1 nA to 1.2 ± 0.2 nA with 0.1 mM BAPTA internal solution and from 0.7 ± 0.1 nA to 1.1 ± 0.2 nA with 10 mM BAPTA internal solution, a result due to removal of Ca2+ block from the MET channel (Crawford et al., 1991; Ricci and Fettiplace, 1998). FRAP measurements were obtained after observing the reported increase in resting Po and maximum MET current which ensured reliable hair bundle perfusion and equilibration of the 20 μM Ca2+. As this experiment requires repeated FRAP measurements over time, we first demonstrate that the FRAP values are stable over the duration of the experimental time (Figure S6A). Lowering extracellular Ca2+ results in slower recovery in the stereocilia for both 0.1 and 10 mM BAPTA internal solutions as indicated by average FRAP curves (Figures 3D and 3G). The diffusion constants of the stereociliary membrane were significantly decreased in both rows (Figures 3E and 3H) in the presence of low extracellular Ca2+, from 5.7 ± 1.1 μm2/s to 4.2 ± 1.4 μm2/s (paired t test, p < 0.01) and 4.4 ± 0.8 μm2/s to 3.6 ± 0.8 μm2/s (paired t test, p < 0.05) for the row 1 and row 2 stereocilia, respectively, with 0.1 mM BAPTA internal solution. With 10 mM BAPTA internal solution, the diffusivity reduced from 4.6 ± 0.8 μm2/s to 3.4 ± 0.8 μm2/s (paired t test, p < 0.001) and 4.5 ± 0.8 μm2/s to 3.8 ± 0.7 μm2/s (paired t test, p < 0.01) for row 1 and row 2 stereocilia, respectively. Both internal solutions produced similar shifts in the diffusion constant in both rows of stereocilia (Figures 3F and 3I), suggesting that the effector site was external, similar to a previous report on MET channel Po (Peng et al., 2013).

Figure 3

Lowering External Ca2+ Reduces Stereociliary Membrane Diffusivity

(A) Representative current responses to sinusoidal fluid jet stimulation from IHCs before, during, and after application of 0.02 mM Ca2+ external solution, with 0.1 mM BAPTA internal (top) and 10 mM BAPTA internal (bottom). Note the increase in baseline current and MET current in the presence of low Ca2+ external.

(B and C) (B) Resting Po and (C) peak MET current Imax increases significantly when the external Ca2+ concentration is reduced from 2 to 0.02 mM (paired t test, p < 0.01). Each symbol indicates a cell for a given internal solution.

(D) and (G) Average FRAP data points and fitted curves comparing 2 mM (black traces for 0.1 mM BAPTA internal and red traces for 10 mM BAPTA internal) with 0.02 mM Ca2+ external (gray traces for 0.1 mM BAPTA internal and light red traces for 10 mM BAPTA internal) for row 1 and row 2 stereocilia.

(E) and (H) Summary box plots of diffusion constants with 0.1 mM BAPTA internal solution (E, black and grey symbols) and 10 mM BAPTA internal solution (H, red and light red symbols) show a significant reduction in diffusivity in the presence of 0.02 mM Ca2+ external irrespective of the internal solution and the row of stereocilia (paired t test, p < 0.05). Each symbol in a graph corresponds to a bundle.

(F) and (I) The fractional decrease in D values when the external Ca2+ is reduced from 2 to 0.02 mM was not significantly different between the internal solutions i.e., 0.1 mM BAPTA (F) and 10 mM BAPTA (I) and the rows of stereocilia.

Boxes represent the SDs, and the star symbol indicates the mean. 0.1 B, 0.1 mM BAPTA; 10 B, 10 mM BAPTA. ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001.

Similarly, the time constants showed significant increase (paired t test, p < 0.001) with low extracellular Ca2+ in both rows of stereocilia, with no significant effect on the mobile fraction of recovery (Figures S6B–S6D). The time course of recovery upon returning the extracellular Ca2+ back to 2 mM appears slower than we accounted for in our measurements such that some stereocilia recovered to a lesser degree than the original measurements. The recovery was not statistically different than the control, and there was no buffer concentration or row-specific differences in the recovery measurement. We did not compare the stereociliary membrane to that of the soma in low external Ca2+ due to the inability to reproducibly expose the basolateral hair cell membrane to low Ca2+ solution using the apical perfusion pipette in the semi-intact epithelial preparation.

Our results indicate that extracellular Ca2+ modulation of membrane diffusivity is independent of the internal buffer, as the fractional changes in diffusivity were unaffected (t test, p > 0.05) by the internal buffer. Data support the idea of extracellular binding sites for Ca2+ that could change membrane mechanics potentially by altering the interaction between adjacent lipids (Ito and Ohnishi, 1974; Melcrova et al., 2016). The sensitivity of stereocilia membrane diffusivity to extracellular Ca2+ appears independent of the MET channel activity as similar results were obtained in both rows. However, low extracellular Ca2+ modulates MET channel resting Po, as well as membrane diffusivity, suggesting a common mode of regulation.

Effect of Voltage on Membrane Diffusivity

Voltage modulates the MET channel Po in a time-dependent manner in the mammalian OHCs (Peng et al., 2016). Although resting Po suggests a time independent steady-state measurement reflecting channel open probability when the bundle is not stimulated, with depolarization there is a temporal component where the Po rapidly increases and then slowly reduces, reflecting a non-steady state condition. We term this Po∗, simply to indicate the same measurement (a non-externally stimulated hair bundle) but under non-steady-state conditions. GsMTx4 inhibits the voltage modulation suggesting an involvement of the membrane in regulating Po∗. We investigated the effect of voltage on the stereociliary membrane diffusivity of di-3-ANEPPDHQ during prolonged depolarization to +76 mV. Firstly, we characterized how the Po∗ changes with voltage in mammalian IHCs by stimulating the bundles with large negative and positive sinusoidal displacements delivered with a fluid jet pipette during prolonged depolarization for 22 s (Figure 4A). We also investigated the dependence of voltage modulation of Po∗ on internal Ca2+ buffer using 0.1 mM BAPTA or 10 mM BAPTA internally. The MET channel Po∗ rose to a maximum rapidly upon depolarization, followed by a slow decrease to a steady-state value during prolonged depolarization. The timing and extent of the response were sensitive to the internal Ca2+ buffer (Figures 4B and 4C). The maximum Po∗ with depolarization to +76 mV was 26.1 ± 10.7% with 0.1 mM BAPTA, which was significantly lower (t test, p < 0.001) than that with 10 mM BAPTA which peaked at 45.1 ± 6.5% (Figure 4D). Also, the time to peak Po∗ was significantly longer for cells with 0.1 mM BAPTA internal compared to 10 mM BAPTA internal, measuring 2.7 ± 1.4 s and 0.38 ± 0.4 s for 0.1 mM BAPTA and 10 mM BAPTA, respectively (Figure 4F).

Figure 4

Depolarization Modulates MET Po∗

(A) presents mechanical stimulus (M) and voltage step at the top and a representative current response at the bottom. Depolarization to +76 mV results in an immediate increase in the MET channel Po∗, with Po∗ decreasing and stabilizing with prolonged depolarization in IHCs.

(B) Summary plot showing the change in Po∗ during depolarization for 22 s for 0.1 mM BAPTA (top) and 10 mM BAPTA (bottom) internal solutions. The values measured at −84 mV are represented at time 0. Each symbol indicates a cell for a given internal solution. The peak Po∗ and the steady-state Po∗ (average of Po∗ values after the time point when change in Po∗ is ≤5%) are highlighted with black arrows.

(C) Expanded view of MET current response to a single period of a sinusoidal stimulus (M) from the current plot in (A), highlighted as I, II, and III, for different internal solutions; 0.1 mM BAPTA (black traces) and 10 mM BAPTA (red traces) for −84 mV (top panel), +76 mV at 2 s (middle), and +76 mV at 20 s (bottom).

(D and E) Summary plot of the peak Poand steady state Po in (D) and ΔPo (peak Po∗ - steady state Po∗) in (E) showing significant differences between 0.1 mM BAPTA (black symbols) and 10 mM BAPTA (red symbols) internal solutions (t test, p < 0.001). Each symbol indicates a cell for a given internal solution, with n = 10 cells for 0.1 mM BAPTA and n = 6 cells for 10 mM BAPTA.

(F) Summary plot of the time taken to reach the peak Po∗ for different internal solutions. Each symbol indicates a cell for a given internal solution, with n = 10 cells for 0.1 mM BAPTA and n = 6 cells for 10 mM BAPTA.

Boxes represent SDs and the star symbol indicates the mean. 0.1 B, 0.1 mM BAPTA; 10 B, 10 mM BAPTA. ∗∗∗p < 0.001.

During prolonged depolarization, Po∗ decreased slowly to a steady state of 18.6 ± 9.2% for 0.1 mM BAPTA which was significantly lower (t test, p < 0.001) than that of 10 mM BAPTA internal measured as 25.4 ± 4.5% (Figure 4D). Also, the change in Po∗ from the peak to the steady state was significantly lower for 0.1 mM BAPTA (7.4 ± 4.9%) compared to 10 mM BAPTA (19.6 ± 4.1%). Thus, the data suggest that the time-dependent voltage modulation of Po∗ is dependent on the internal Ca2+ buffering.

We next examined whether voltage affects the stereociliary membrane diffusivity in a time-dependent manner, more specifically during the initial depolarization when we see a rapid increase in the Po∗ and during prolonged depolarization which results in a slow reduction of Po∗ to a steady state. FRAP measurements were done either during the first 12 s of depolarization (early FRAP) to capture changes responsible for the increase in Po∗ or after the first 10 s of depolarization (late FRAP) to capture the reduction of the Po∗ over time, as illustrated in Figures 5A and 5B. Figures 5C and 5D summarize average FRAP curves measured across the two rows of stereocilia comparing early FRAP and late FRAP done at +76 mV to FRAP measured at −84 mV either with 0.1 mM BAPTA or 10 mM BAPTA internal. There was no change in the recovery for early FRAP in any condition while there was a slight increase in the recovery with late FRAP in both stereociliary rows and with both internal buffers. FRAP measurements during the early depolarization resulted in no change in the diffusion constant (Figures 5E and 5G). As FRAP is an average measure across a large volume and an extended time course, it is possible that any localized and fast membrane diffusivity changes due to early depolarization response may not be fully detected in these measurements. The stereociliary membrane diffusion constants were significantly increased in row 2 with late depolarization (paired t test, p < 0.01), from 3.9 ± 0.4 μm2/s to 4.9 ± 1.1 μm2/s with 0.1 mM BAPTA internal (Figure 5F) but not for row 1. Diffusivity reduced significantly for both stereociliary rows with 10 mM BAPTA internal, from 3.8 ± 0.9 μm2/s to 5.1 ± 1.3 μm2/s for row 1 (paired t test, p < 0.01) and 3.7 ± 0.4 μm2/s to 4.9 ± 0.7 μm2/s for row 2 (paired t test, p < 0.05) (Figure 5H).

Figure 5

Prolonged Depolarization Increases Stereociliary Membrane Diffusivity

(A and B) Illustration of the FRAP timing with respect to depolarization showing (A) early FRAP, where FRAP recovery phase starts at time 0 of depolarization, and (B) late FRAP, where FRAP recovery phase starts 10 s after depolarization.

(C) Average early FRAP curves (mean ± SD) measured from row 1 (left side) and row 2 (right side) stereocilia, with 0.1 mM (top) and 10 mM BAPTA (bottom) internal solutions, showing no difference between the measurements at −84 mV (black and red traces) and +76 mV (gray and light red traces).

(D) Average late FRAP curves (mean ± SD) show faster recovery with prolonged depolarization to +76 mV (gray and light red traces) compared to that of −84 mV (black and red traces).

(E) and (G) Summary of cells with 0.1 mM BAPTA (E) and 10 mM BAPTA (G) internal solutions showing that early depolarization from -84 mV (black symbols in E and red symbols in G) to +76 mV (grey symbols in E and and light red symbols in G) has no significant effect on the diffusion constants in both stereocilia rows as well as with 0.1 and 10 mM BAPTA internal solutions. Each colored/patterned symbol corresponds to a bundle/cell.

(F) and (H) Summary plot of D values with 0.1 mM BAPTA (F) and 10 mM BAPTA (H) internal solutions showing that late depolarization significantly increases the diffusion constant of row 2 irrespective of the internal solution and of the row 1 with 10 mM BAPTA (paired t test, p < 0.05). Each colored/patterned symbol corresponds to a bundle/cell.

Boxes represent SDs, and the star symbol indicate the mean. 0.1 B, 0.1 mM BAPTA; 10 B, 10 mM BAPTA. ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001.

To determine whether the differences observed after prolonged depolarization were unique to stereocilia, we also performed FRAP experiments in IHC soma. Using 10 mM BAPTA, we found no difference in diffusivity due to voltage (Figure S7). Thus, we conclude that voltage-dependent diffusivity changes are unique to the stereociliary membrane.

Discussion

Stereocilia membrane properties were investigated under conditions known to alter MET channel resting Po to assess more directly the hypothesis that the lipid bilayer modulates the effect of Ca2+ and voltage on mammalian auditory MET channels. We show that the IHC stereocilia membrane is highly diffusive compared to the soma, with both stereociliary rows having similar membrane diffusivity. We further demonstrate that the diffusivity of the stereocilia lipid membrane is sensitive to Ca2+ and voltage while the basolateral membrane is not. Membrane diffusivity is not modulated by MET; however, the strong correlation between the resting Po and diffusivity suggests that membrane mechanics and MET channel resting Po have a common underlying modulator (Figure 6A).

Figure 6

MET Resting P Is Correlated to Stereociliary Membrane Diffusivity

(A) The mean MET resting P plotted against the mean diffusion constant D measured across experiments shows a strong correlation (r2 = 0.69, p < 0.05), with black symbols representing measurements taken with 0.1 mM BAPTA internal solution/2 mM external Ca2+ at −80 mV holding potential, red symbols representing 10 mM BAPTA internal solution/2 mM external Ca2+ at −80 mV holding potential, gray symbols representing 0.1 mM BAPTA internal solution/20 μM external Ca2+ at −80 mV holding potential, and salmon symbols representing 10 mM BAPTA.

(B) presents the relationship in (A) pictorially, suggesting that the MET Po is sensitive to membrane diffusivity. The stereocilia membrane curvature presented is not scaled as the exact curvature value would depend on the exact position of the MET channel on the stereocilia, which is unknown.

(C) Schematic illustrating the difference in membrane curvature between the stereocilia wrapping around its actin core and the basolateral membrane. The white cross sections on the hair bundle and the cell body indicate the approximate location of the represented membrane curvatures for the stereocilia (in red) and basolateral membrane (in blue). The cartoon denotes the need for specialized conical lipids on the outer and inner leaflets to accommodate for the observed curvature. The cartoon is not a scaled representation of the relation between lipid bilayer thickness and curvature as that would make it impossible to see the individual lipids. The lipids are colored to indicate the range of diffusivity with red representing high diffusivity and blue representing low diffusivity. We hypothesize that the inherent curvature alters the sensitivity of the stereocilia to divalent ions which alters diffusivity and can indirectly alter MET channel Po.

High Membrane Diffusivity of the IHC Stereocilia

IHC stereocilia showed significantly higher diffusion constants (∼9 times) than the soma, consistent with reported elevated diffusion constants for guinea pig OHC stereocilia bundles compared to the soma (Boutet de Monvel et al., 2006). Diffusion constants measured for di-3-ANEPPDHQ in the IHC soma ranged from 0.4 to 0.7 μm2/sec, comparable to the previously reported values for di-8-ANEPPS in IHCs and OHCs ranging 0.2–1 μm2/sec (Oghalai et al., 2000, Boutet De Monvel et al., 2006; Chen and Zhao, 2007; Organ and Raphael, 2009). All measured values are within the range typically reported for living cells, i.e., 0.01–10 μm2/sec with most cells exhibiting values from 0.1 to 1 μm2/sec at 18–22°C (Lee and Jacobson, 1994; Almeida and Vaz, 1995). However, the IHC stereocilia values are significantly higher than soma values at 20–22°C using the same lipid dye; similar high values at room temperature are observed in a few cell types such as rat smooth muscle cells and mouse spleen and lymphocytes (Lee and Jacobson, 1994). Membrane diffusivity differences may arise from different lipid composition, protein content, membrane interaction with underlying cytoskeletal structures, mechanotransduction machinery, active lipid transport, and/or membrane curvature (Hoekstra, 1994; Bigay and Antonny, 2012).

Lipid Composition

Mass spectrometry of chick vestibular hair bundles suggests that the stereocilia lipid composition does not differ significantly from other cell membranes, except for higher level of sphingomyelin and ceramide (Zhao et al., 2012). Therefore, it is less likely that the high stereociliary diffusivity is due to gross differences in lipid composition. However, compartmentalization of specific lipids or even low levels of some lipids can have dramatic effects on membrane mechanics; therefore, we cannot completely exclude lipid composition as a contributor.

Protein Content and Cytoskeleton

The presence of distinct transmembrane proteins, sub-membranous scaffolding proteins, and proteins that link the membrane to the underlying dense actin cytoskeleton would support the stereocilia membrane being less diffusive (Owen, 2015) and so are unlikely responsible for the elevated stereocilia diffusivity. It is possible, though unlikely, that all cell membranes have high levels of proteins embedded and have strong cytoskeletal connections leading to low diffusivity, and it is the lack of proteins and cytoskeletal interactions contributing to the higher stereociliary diffusivity levels.

Mechanotransduction Machinery

MET channels are only present in the shorter stereocilia rows (Beurg et al., 2009). The upper tip-link insertion point has a density with selective molecules like harmonin, whirlin, and cadherin 23, while the lower insertion point is linked via protocadherin 15, CIB2 (Calcium And Integrin Binding Family Member 2), LHFPL5 (Lipoma HMGIC fusion partner-like 5), TMIE (Transmembrane Inner Ear), and TMC (Transmembrane channel-like) molecules (Kazmierczak et al., 2007; Xiong et al., 2012; Kurima et al., 2015; Giese et al., 2017; Pan et al., 2018; Cunningham et al., 2020; Krey et al., 2020). The MET channel at the lower end of the tip links permeates calcium and monovalent ions into this small volume. Yet, our data show that both stereociliary rows have similar baseline membrane diffusivity and similar changes in diffusivity following Ca2+ and voltage manipulations, indicating that the observed effects are independent of MET channel activity. Additionally, pharmacologically blocking MET channels had no effect on diffusivity, suggesting that the MET channel complex does not regulate macroscopic membrane diffusivity. Thus, it is unlikely the mechanotransduction machinery is responsible for the differences between soma and stereocilia.

Active Transport

High stereociliary diffusivity could also arise from an active lipid transport component coupled with passive diffusion. The stereociliary membrane specifically expresses phospholipid-translocating ATPases such as ATP8B1, which are essential in phospholipid transport to maintain membrane asymmetry and curvature (Sebastian et al., 2012; Coleman et al., 2013), and morphological and functional degeneration of the hair bundles due to the deficiency of ATP8B1 and ATP8A2 indicate the importance of this transport mechanism involved in maintaining the lipid composition and the mechanical stability of the stereocilia (Stapelbroek et al., 2009; Coleman et al., 2014). The functional relevance of this active lipid transport is unclear. Whether this transport can account for the large difference in diffusivity is also unclear, and further experiments are needed to address this possibility.

Membrane Curvature

A simple yet plausible mechanism for elevated stereociliary diffusivity stems from the mechanics associated with high stereociliary membrane curvature. Membrane curvature is generated by the membrane monolayer asymmetry due to the presence of specific lipids such as conical lipids and clustering of shaped transmembrane proteins (McMahon and Gallop, 2005; Jarsch et al., 2016) required for establishing the tight wrapping of the actin core (Figure 6C). In its simplest form, lipid asymmetry is required for the membrane to ensheath the actin core of the stereocilia. Importantly, membrane curvature can mechanically control the spatial organization of the lipid bilayer, with regions of high curvature leading to loose lipid packing (i.e., high diffusivity) due to the preferential localization of disordered lipid domains (Parthasarathy et al., 2006; Bigay and Antonny, 2012; Vamparys et al., 2013). Parthasarathy et al. found a critical curvature of <0.8 μm2, where spatial organization of lipids are directly affected by curvature (Parthasarathy et al., 2006). Thus, membranes of IHC stereocilia having curvatures greater than 1 μm2 can be mechanically stressed and thus likely impacting membrane diffusivity by the spatial organization of lipid domains, as seen in other membrane protrusions such as filopodia and microvilli (Zhao et al., 2013; Prévost et al., 2015). The observed similar high stereociliary membrane diffusivity in both rows of stereocilia (with similar diameters) may in part be established by the curvature associated with the dimensions of the stereocilia.

FRAP is a slow measurement that averages across a large membrane volume. Stereocilia membranes are described with bumps potentially representing transmembrane proteins. It is also likely that the membrane is not longitudinally uniform, with MET machinery at the tops of the shorter rows and with the tip membrane undergoing deformation via the tip link pulling (either directly or indirectly). FRAP is not reporting on nanodomain differences in diffusivity; alternate technologies are needed for this level of resolution.

How Might the Stereociliary Membrane Be Affected Differently from the Soma?

Data demonstrate Ca2+ and voltage effects on stereociliary membrane diffusivity but not on the basolateral membrane. Multivalent ions, especially Ca2+, can directly interact with lipids to change the membrane mechanical properties (Ito and Ohnishi, 1974; Shoemaker and Vanderlick, 2003; Melcrova et al., 2016). Ca2+ can change lipid packing or membrane order through conformational changes to lipid headgroups, ordering of acyl chains, affecting the hydration shell of lipid headgroups, and altering the repulsive interactions between the lipids. The baseline diffusivity associated with the bilayer wrapping around the stereocilia may simply reflect the above changes better than the soma such that a 10% change is easily detectable in the stereocilia (∼0.5 μm2/s) but less so in the soma (0.05 μm2/s). Alternatively, the curvature tension of the stereocilia may create an environment more sensitive to divalent modulation (Figure 6C).

The effects of voltage on stereocilia diffusivity are similarly complex as the effects on MET Po. The differential effect on stereocilia as compared to soma may be as described above, a reflection of membrane curvature impacting lipid distributions. The lack of effect on diffusivity during the initial depolarization where Po is increasing is likely due to the slow temporal resolution of FRAP, and better technology is needed to characterize fast responses. The slower increase in diffusivity correlates with a reduction in Po and suggests a reduction in force transference to the channel. The underlying molecular mechanism for this effect and the physiological relevance remain to be elucidated. The simplest explanation may be a rearrangement of lipid molecules due to charged headgroups being less shielded by calcium ions.

Implications for Hair Cell Mechanotransduction

The nine-fold difference in membrane diffusivity between the stereocilia and soma is likely to provide the necessary lipid environment for modulation of mechanotransduction, the key functional output of this organelle. High membrane diffusivity is often associated with elevated membrane tension and reduced hydrophobic thickness (Butler et al., 2001; Haidekker et al., 2001; Blood et al., 2005; Reddy et al., 2012). Studies of other mechanosensitive channels show that decreased membrane fluidity increases their activation threshold, and thinner membrane favors channel opening (Perozo et al., 2002; Nomura et al., 2012). In contrast, stereociliary membrane diffusivity is generally high while the MET resting Po is low, and the Po increases as diffusivity is reduced (Figures 6A and 6B), at least in the range of measured diffusivity. The opposite polarity for hair cell MET channel sensitivity may be simply due to how force is transmitted to the MET channel. While many mechanosensitive channels sense force transmitted directly through the membrane, hair cell channels are thought to be tethered to both an extracellular and intracellular link (Effertz et al., 2015). Thus, the mechanical contribution of the bilayer may be indirect or in parallel to the protein coupling which in its simplest form would reverse sensitivity. More work is needed to clarify this important point.

Our data support a potential mechanism whereby the stereociliary membrane has a significantly higher baseline membrane diffusivity compared to the soma likely due to the high curvature of the stereociliary membrane. The differential Ca2+ effect on the stereociliary membrane may also be a manifestation of curvature-induced stress, where the stereocilia and the soma are at different curvature tension such that Ca2+ differentially affects lipid packing and thus the membrane diffusivity of these structures. MET channels do not contribute to stereociliary diffusivity but may be directly or indirectly modulated by changes in diffusivity as indicated by the resting open probability. We postulate that the stereociliary membrane contributes to a viscoelastic-like component modulating the hair cell MET channel.

Limitations of the Study

FRAP measurements are both spatially and temporally limited given that recovery times are a function of the volume bleached and the ability to image at high rates. The technology does not allow for monitoring rapid changes in membrane properties. Additionally, the volume is typically dictated by the z-resolution of the objective (as set by the numerical aperture of the objective) and so local changes will be filtered or averaged into the larger volume being monitored. Thus, the technology is unlikely to detect changes associated with MET channel gating, for example, because gating is both very local and fast. An example is that FRAP does not detect a change during the rapid change in MET open probability with depolarization. These data cannot be interpreted as there is no membrane effect but simply that FRAP does not detect a change. Related to the spatial and temporal limitations is the fact that the relationship between MET open probability and membrane diffusivity is a correlation; additional technologies are needed to generate causal links between open probability and membrane mechanics. Data presented here provide the impetus to follow-up on developing technologies to investigate causality.

Resource Availability

Lead Contact

Further information and requests for resources should be directed to and will be fulfilled by the lead contact, Prof. Anthony Ricci, aricci@stanford.edu.

Materials Availability

This study did not generate new unique reagents.

Data and Code Availability

The original unprocessed data of live cell imaging and electrophysiological recordings are contained in very large files. These are available upon request from the corresponding author. All codes have been uploaded to Mendeley Data “Mendeley Data:https://doi.org/10.17632/3wrd9xp4gc.1.” and are also available through the authors.

Methods

All methods can be found in the accompanying Transparent Methods supplemental file.