Electromechanical coupling in the membranes of Shaker-transfected HEK cells

Abstract

Membranes flex with changes in transmembrane potential as a result of changes in interfacial tension, the Lippman effect. We studied the membrane electromotility of Shaker K⁺-transfected HEK-293 cells in real time by using combined patch-clamp atomic force microscopy. In the voltage range where the channels were closed, Shaker expression had little effect on electromotility relative to wild-type cells. Depolarization between −120 and −40 mV resulted in a linear upward cantilever deflection equivalent to an increase in membrane tension. However, when depolarized sufficiently for channel opening, the electromotility saturated and only recovered over 10 s of milliseconds. This remarkable loss of motility was associated with channel opening, not ionic flux or movement of the voltage sensors. The IL mutant of Shaker, in which the voltage dependence of channel opening but not sensor movement is shifted to more positive potentials, caused the loss of electromotility saturation also to shift to more positive potentials. The temporary loss of electromotility associated with channel opening is probably caused by local buckling of the bilayer as the inner half of the channel expands as expected from X-ray structural data.

We wanted to measure the normal component of voltage sensor movement (1) by attaching S4 to an atomic force microscopy (AFM) cantilever. As opposed to measurements using fluorescence resonance energy transfer (FRET) (2) or cross-linkers (3) that respond to the relative position of subunits, AFM measures absolute displacements with respect to the laboratory reference frame (4). To measure S4 movement with AFM we need to distinguish the background motion of the membrane from the motion of the channel relative to the membrane. As a first step, we measured electromotility (EM) of the membrane in both wild-type cells and cells transfected with Shaker and its IL mutant (5).

Membrane electromotility (MEM) is primarily a result of repulsive forces generated by excess charge at the interface, a process known as the Lippman effect that forms the basis for the hanging-drop Hg²⁺ electrode (6). The Lippman equation relates interface tension to potential, ∂γ/∂V = −σ, where γ is the tension, V the interfacial potential, and σ the surface charge. Because membranes are composed of two polarizable interfaces ≈3 nm apart, the two interfaces attract each other, leaving only second-order Lippman effects (7). When a cell is indented with an AFM tip, the cytoskeleton and the bilayer tension push back (8–10). Changes in voltage produce changes in tension resulting in movement of the probe tip (11–13). In a neutral membrane, changes in voltage produce minimal movement, but the presence of asymmetric fixed charges perturbs the system, producing a parabolic voltage dependence centered at the mean surface potential. In the accessible voltage range, the MEM of normal membranes is linear, and the more a cell is indented by the probe, the larger the area of membrane in contact with the tip and the greater is force per mV (12). If ion channels changed their lateral dimensions significantly with voltage, we would expect to see a change in tension; i.e., if channels became larger, the membrane tension would fall, and the AFM tip would sink deeper into the cell. If a voltage-dependent channel were located below the tip and it changed its normal dimension, as proposed for hydrophobic mismatch in mechanosensitive channels (14), this too would appear as a voltage-dependent displacement and would fit with some models of voltage sensor movement (1).

We expected that Shaker transfection would cause large changes in MEM from either the voltage sensor pushing the probe outward with depolarization (3), or a large change in surface potential produced by the gating currents changing the Lippman tension (1). Contrary to expectation, Shaker transfection produced an abrupt loss of MEM at the potentials associated with channel opening, but no change of MEM with sensor activation. The change in tension was not associated with ion flux. Based on structural data (15–18), Shaker opening appears to be accompanied by a large increase in lateral area of the intracellular half. The voltage sensor movement normal to the membrane is in some dispute (19, 20) but would appear to be <6–20 Å and mostly interior to the bilayer.

Results

Wild-type HEK (wtHEK) cells were voltage clamped in whole-cell mode with the AFM in force-clamp (FC) (Fig. 1A). Depolarizing voltage steps induced outward membrane movement, i.e., upward motion of the cantilever (Fig. 1B). In agreement with earlier data (12), displacement from −120 mV to +60 mV MEM was linear at all set points of force from 100 to 500 pN (n = 8) (Fig. 2A). The amplitude of the displacements, ≈1 nm/100 mV, was comparable with that of previous studies (12, 13). At 500 pN, a 100-mV depolarization caused a peak outward displacement of 6.75 Å, corresponding to 13.5 pN for a 0.02 N/m cantilever, similar to the published value of ≈10 pN/100 mV (12).

Fig. 1.

Whole-cell voltage clamp/AFM (VC-AFM) experimental setup (A) and typical experimental results (B and C). (A) Experimental setup. Cultured HEK cells on coverslips are placed atop an inverted microscope, and a patch pipette and AFM cantilever are positioned nearby. (Upper) Optical image (10×) of a typical cell with patch pipette and an AFM cantilever shown here for a sense of size. (Lower) whole-cell voltage-clamp (VC) is established first, and membrane voltage (V_m) is controlled by using a dedicated amplifier (VC Amp). Then, an AFM cantilever is placed atop the cell surface in force-clamp mode at 20–500 pN. Cantilever positioning is controlled by another amplifier (AFM Amp), which also acquires cantilever movement information by using an optical lever (red laser), the deflection of which off the back of the cantilever is sensed by a 4-quadrant position-sensitive photodetector (PSPD). VC-AFM is controlled by a personal computer (PC). (B) Typical data for wtHEK cells. (Top) Voltage step protocol with a holding potential of V_m = −80 mV. V_m is prepulsed to −120 mV and then randomly (to avoid introducing trends) with 10-mV increments through the voltage range of physiologic interest to + 80 mV. (Middle) Voltage-induced whole-cell currents. For wtHEK, maximum depolarization produces a few hundred pA. (Bottom) voltage-induced membrane displacement (EM). A voltage step produces a jump in membrane tension reported by bending of the cantilever parked on the cell membrane. For wtHEKs, depolarization results in an increase in membrane tension and thus an upward cantilever deflection (positive). Hyperpolarization decreases membrane tension, resulting in cantilever sinking into the cell (negative). Displacements are calculated from a baseline at −120 mV, averaging between 25 and 30 ms of the pulse. After the early voltage-induced displacement (average 33–35 ms), the cantilever drifts toward baseline because of cortical relaxation, partially flattening the late EM. Downward drift is observed even after membrane voltage is stepped back to the −80 mV holding potential (off, average 65–70 ms). (C) A typical ShHEK experiment. (Top) Voltage ladder protocol. (Middle) Upon channel activation (V_1/2 = −41 mV), the voltage-induced membrane current is several nA, rising linearly with voltage. (Bottom) EM ShHEK is similar in behavior to wtHEK until the channel open (green and red traces). Note, in the early times of the voltage step, no increase in movement is seen with additional depolarization. Downward drift is seen late in the pulse with nonactivating depolarizations as with the wtHEK, although the late response to the activating pulses is opposite to wtHEK (green and red). Finally, unlike the continued downward drift of wtHEK when stepped off, activating voltages resulted in a continued increase in tension.

Fig. 2.

MEM curves for wtHEK (A, C, and E) and shHEK (B, D, and F) at a range of set point forces (50–500 pN). (Insets) Normalized (Δz/∣z∣_max) VD curves (n = 5). (A) Early wtHEK MEM is linear through the V_m range between −120 and +60 mV. MEM is also positively correlated with FC. (B) Late wtHEK MEM shows flattening but force and voltage dependence are maintained. (C) Off wtHEK MEM shows continued downward drift that is more pronounced for larger depolarizations or larger step displacements. (D) Early ShHEK MEM saturates. MEM is linear up to Shaker activation at −40 mV. This is followed by saturation over a wide range of V_m and for all FCs. (E) Late ShHEK MEM straightens, and displacement magnitude is similar to wtHEK. (F) Off ShHEK is notable for its nonlinearity. For nonactivating steps that preceded this time point, the downward drift caused by earlier stimuli is similar to wtHEK (above), but after the activating steps, the membrane maintains some tension above baseline.

After the transfection with Shaker (ShHEK), MEM was similar to wild type, linear over the voltage range of −120 to −40 mV where Shaker remains closed (Fig. 1C), but the amplitude was somewhat larger for ShHEK than for wtHEK at all holding forces (MEM_ShHEK/MEM_wtHEK = 1.48 ± 0.21) (Fig. 2D). This difference would suggest channel expression produced a larger surface charge. This effect is not specific to Shaker because we observed similar behavior for acetylcholine receptor expression.

Remarkably, in the voltage range where Shaker is open (−40 mV to +60 mV), MEM saturated (Fig. 2D) at the earliest time points, and further depolarization produced no change in probe position. We saw saturation in 83% (n = 23) of the experiments, and the only the experiments without obvious saturation were those performed at the lowest-force set points where mechanical noise often dominated the recording (Fig. 2D).

Kinetic Responses.

The basic AFM response to a jump in potential (early) is an instantaneous jump in probe position (Fig. 2A) followed by a relaxation to a potential-dependent steady-state position (late) (Fig. 2B). At the end of the voltage pulse (off), there was a slow relaxation back to rest (Fig. 2C). The early MEM of wild-type cells and Shaker-transfected cells was always linear in voltage negative to −40 mV. The saturation occurred at the early time points of the activating voltage steps where the response should have been maximal (Fig. 2D). MEM did recover late in the pulse (20 ms after activation) in 56% of the experiments (n = 18), and this was independent of the FC set point (Fig. 2E). At maximum depolarization, displacement late in the pulse was similar to wtHEK and ShHEK with an average difference only 2 ± 1 Å.

ShHEK MEM continued to be nonlinear with respect to voltage even after the membrane potential was stepped back to baseline (off). Displacement at constant voltage showed a downward drift similar to that seen for wtHEK or for Shaker with voltage steps that were nonactivating. This relaxation of displacement was probably a result of cytoskeletal rearrangements because these were visible after repolarization as upward offsets in the baseline (Fig. 1C Bottom and Fig. 2F). We expected that Shaker would produce only local changes in membrane properties so that the background motion would be superimposed on any channel-induced motion. Thus, saturation was unexpected.

The return of MEM at long times suggested that the saturation was probably a kinetic effect, but all of the simple explanations appeared to require a high channel density.

Channel Density.

We estimated the number of active channels by using N_ch = I [g(V_m − V_K)]⁻¹, where I is the current, V_m is the membrane potential, V_K is the reversal potential for K⁺ (V_K = −84 mV), and g is the unitary channel conductance (pS) (21). For Shaker, g ≈10 pS in our conditions (22). The maximal currents were ≈10 nA/100 mV, suggesting ≈10³ active channels per cell. We chose small rounded cells for our experiments and inflated them to increase stiffness, and that made them spherical (10). Typical membrane capacitance was C_m ≈10 pF. Assuming a specific capacitance of 10 fF/μm² (21), the membrane area was ≈1,000 μm², and thus the mean channel density was ≈1/μm².

Assuming that the channels were evenly distributed, they would be separated by distances much larger than the Debye length and hence could not readily alter the mean surface charge of the membrane. According to the Lippman equation (7, 12), saturation of MEM implies a complete loss of surface charge. The surface charge of HEK cells is approximately −20 mV (12), and for the gating current associated with opening (1–2 e⁻) to neutralize that amount of charge would have required a much higher expression density. Furthermore, a sudden change in charge density would cause a jump in MEM rather than saturation because a change in charge density at constant potential will produce motion.

Search for the Source of Saturation.

Flux coupling.

The saturation of MEM occurred when the channel opened, suggesting that an opposing solute or water flux might cancel the background MEM. However, from approximately −20 mV to +60 mV where MEM was saturated, the current increased linearly with voltage so that fluxes through the channel were an unlikely source. We further tested the flux hypothesis by using ion substitution to reverse the K⁺ gradient.

We replaced intracellular K⁺ with N-methyl-d-glucamine (NMDG) (21, 23), leaving 1 mM K⁺ to trace the conductance (24). There was a reduction in the K⁺ current to ≪1 nA, but the gating kinetics remained the same, and MEM saturated as it did in normal saline (Fig. 3A). As a final test of the flux hypothesis, we used symmetrical K⁺ so that the reversal potential was 0 mV. Channel opening resulted in an initial inward K⁺ flux, and further depolarization, an outward K⁺ flux. Again, there was no effect on MEM (Fig. 3B). Thus, K⁺ flux (and the coupled water) or ionic currents are not the source of the saturation.

Fig. 3.

Searching for the source of ShHEK MEM nonlinearity. Voltage-dependent fractional lever displacement (Δz/∣z∣_max) is overlaid on ShHEK displacements from Fig. 2D (gray). (A) Intracellular NMDG (n = 3). (B) Symmetric K⁺ (n = 5). (C) ShIL (n = 4).

Shaker IL.

If saturation of MEM was related to channel opening, then the IL mutant, in which gating is shifted to more depolarized levels with respect to S4 movement (25, 26), should also shift the saturation voltage. The voltage-sensing apparatus of the ILT mutant, which is a close relative of the IL mutant but poorly expressing, is intact and behaves similarly to Shaker. MEM saturation for the IL mutant shifted to more positive potentials where the channel opened (Fig. 3C). This further suggests strongly that the displacements are caused by channel opening and not S4 motion.

Discussion

Our results showing nonlinear electromechanical behavior of Shaker-transfected HEK cell membranes are consistent with the earlier study by Mosbacher et al. (11) and agreed that at negative potentials, the ShHEK MEM amplitude was 25–125% larger than wild-type cells. In the present work with higher time resolution, we observed an early saturating behavior related to channel opening (not voltage sensing). This saturation disappeared by the end of the voltage step (20 ms) as background MEM was reestablished. Any model to explain the data must address the presence of saturation, a channel density of 1/μm², the absence of an effect of ion flux, the IL mutant translation of the voltage dependence of saturation, independence of the response from set point force, and the slow relaxation of displacements. There are two types of models that come to mind: changes in surface charge and changes in channel dimensions.

Surface Charge Models.

The Lippman effect changes the mean tension of the interface and a thus the force on an indenting probe. When the interfaces are asymmetric, the membrane will bend at zero applied force, and this is known as flexoelectricity (27). Because Shaker has highly charged mobile voltage sensors (1) and fixed charges on the extracellular surface (28, 29), rearrangement of any or all of those charges during channel opening could influence the charge distribution. Most charge movement in Shaker occurs during voltage sensing when ≈13.6 e per channel move across the electric field (30). Our data show that the voltage sensor motion is not the origin of MEM saturation because the ShIL mutant shifted the voltage for saturation, and the saturation occurred at potentials more positive than most of the S4 movement.

There is a small charge displacement during channel opening, equivalent to ≈2 e⁻ per channel (25), or <1 fC per cell. At ≈1 channel per μm², this charge is not sufficient to explain the striking saturation of MEM. The Lippman equation would require ≈10× increase in mean negative charge density on the extracellular face or a similarly positive increase on the intracellular face to explain the saturation. Thus, we abandoned the surface change models.

Geometric Models.

If Shaker changed dimensions when it opened, this could translate into movement of the probe, either via a direct movement normal to the membrane or by a change of in-plane area leading to a change in tension. For Shaker, opening has been proposed to involve a large splaying of the inner pore helices at a conserved glycine (16, 31). This would be associated with an increase in mean channel radius of 1–2 nm.

Modeling the Change of In-Plane Area.

If we approximate the cell as a sphere, Laplace's law predicts that a change in cell area, ΔA, at constant hydrostatic pressure would result in a change of tension ΔT

and this will produce a change in force on the probe. To a first approximation we can assume the tip to be conical. Used cantilever tips were randomly scanned with scanning electron microscopy (SEM) and were on the order of 50-nm diameter. Then, the force on the indenting tip is ΔF = 2πr_zΔT = kΔz, where r_z is the radius of the tip where the membrane separates from the indenting tip (not the radius of curvature of the tip), k is the cantilever stiffness, and Δz is the change in tip position (12). The area of the membrane in contact with the cantilever (A) is

Solving for z after substituting r = z/tanθ, we have

Differentiating this solution for the change in cantilever height (Δz) given a change in area (ΔA) associated with the opening of a single channel,

Assuming that the mean radius of Shaker in the closed state is 5 nm, and the mean radius increases by ≈1 nm, and the cantilever–membrane contact area is ≈1 μm², dz would only be ≈4 pm, below our level of resolution. There are a number of ways that the assumption of the calculation can be in error.

There is a possibility the channels were not uniformly distributed but clustered beneath the tip (32–34) so the local area charge was large, or the membrane tension was nonuniform. To explain the data, the relative change of in-plane area would only have to be ΔA/A = 0.15%, requiring 150 channels under the AFM tip. However, we have not been able to measure the local channel density, and it is unlikely that in hundreds of experiments we would always hit a clump of channels. It is possible that the functional channels, i.e., those capable of conducting current, were a small fraction of all of the channels expressed, but the “silent majority” were capable of changes in shape or surface charge density and were visible to the AFM.

The assumption of uniform tension might be wrong because the cell cortex is inhomogeneous and viscoelastic (35). Sudden expansion of a channel during opening could cause the surrounding bilayer to buckle (36), and this excess bilayer would allow the cantilever to settle toward the cell interior, as observed (Fig. 4). The buckled folds would appear unresponsive to voltage because the mean membrane position would not change. As the stresses relaxed in time, the bilayer would return to sharing a fraction of the mean cortical stress (37), and MEM would reappear. This is the behavior we observed. MEM saturated over a wide voltage range just after channel opening (Fig. 2D) as expected from the high compliance of a buckled bilayer. With time, MEM reappeared as the membrane supposedly flattened (Fig. 2E).

Fig. 4.

Diagram of how channel opening could reduce EM. (Upper) Approach of the AFM tip to a resting channel in the membrane. (Lower) When the channel opens as a cone, it pushes out the inner monolayer, causing the membrane to buckle. Applied potentials then have little effect on moving the tip because they will only change the local curvature.

Modeling Movement Normal to the Membrane.

Channel opening may be associated with a movement normal to the membrane. In the extreme case where the channel sits directly under the AFM tip, the channel has a low compliance, and a change in height drives the cantilever. The predicted change in cantilever position is a function of the compliance of the cantilever and the cell according to

where k_c is the cantilever stiffness, k_s the sample (cell) stiffness, and ΔF the force generated by the channel. From the mean force–distance curves, we estimated k_s ≈0.6 mN/m (10) and from the Veeco specifications k_c = 2mN/m. From the difference in linear fits of pre- and post-MEM over the channel activation range (≈50 pN FC), we estimated the force required to cancel the background MEM as ΔF ≈0.7 pN. This would correspond to a change of system height change (ΔZ) of 1.4 nm and a channel with a linear change in height with voltage to cancel the background MEM.

When closed, KcsA is ≈3.5 nm in height, whereas the homologous MthK purportedly in the open state is ≈2.6 nm in height (16). If the channels were not directly beneath the tip (32–34), but pressed against the side of the cantilever, Δz = Δh_c cosθ, where Δh_c is the change in height change of the channel and θ = 54° is the taper angle of the tip. Taken literally, this model predicts that the observed sinking of the cantilever upon channel opening is a result of the channel getting thinner or translating toward the cell interior upon opening. If the channel density was 1/μm², the observed forces would be proportionately less and scaled by the contribution of the various constitutive properties. The difficulty with this model as with other geometric models is that they tend to predict an inflection on the background MEM rather than saturation.

The lack of movement of the probe in the voltage range in which S4 motions takes place would appear to place an upper limit on S4 motion normal to the membrane. The rms noise level of the AFM data was 1 Å at the bandwidth of 1 kHz, and we saw no measurable response in the gating voltage range so that if channel was pushing against the tip we would predict the motion to be ≤1 Å. Further increase in dynamic z-resolution is possible with the AFM, but it would require use of high-gain soft AFM probes (38).

Channel–Cytoskeleton–Lipid Bilayer Interaction.

Electrostatic models of ideal membrane mechanics (12) do not predict relaxation times in the range of 10–1,000 ms. The time dependence presumably arises from viscoelastic relaxation of the cell cortex in response to the mechanical stress induced by changes in channel shape upon opening. Wild-type MEM showed a slow drift back to baseline after the voltage step and undershoot upon release of the voltage, a response characteristic of linear viscoelastic elements (Fig. 2C). With Shaker, however, depolarizations large enough to open the channel led to a rapid relaxation, with no undershoot. Upon release of the stimulus, there was a return to an offset baseline representing slower relaxing components (Fig. 2F). This plastic behavior is reminiscent of domain unfolding in large polymers such as titin (39). If channels are coupled to the cytoskeleton (40), then opening may provide sufficient stress to cause plastic unfolding. The MacKinnon group (41) recently reported similar prolonged effects of pressure on K channel function and presumed these to be caused by the channel–cytoskeleton interactions.

Channel opening, although inherently rapid, is a step conformation that will drive long-lived mechanical perturbations of the cell cortex. For Shaker, the saturation of MEM that occurred early in the voltage pulse disappeared later in the pulse as Δz decayed exponentially to a plateau restoring the background MEM. In the plateau region, MEM was similar to the background motion of wild-type cells. However, for voltage jumps large enough to open the channels, the displacements decayed with two time constants. There was a fast relaxation with a time constant of several ms and a much slower one visible as an offset in baseline at the end of the record (Fig. 2E). The wild-type cells did not show the long relaxation time constant, suggesting that the channels introduced new stresses in the cortex, possibly via direct linking to the cytoskeleton (Fig. 5) (42–47).

Fig. 5.

Diagram of how channel opening results in long-term changes in MEM. (A) MEM is at background sensitivity until the channel opens, reducing tension (B). As the underlying cytoskeleton relaxes, MEM returns to background (C). However, long-term nonlinearity in VD persists even after the membrane voltage is stepped back to rest. (D) MEM returns to baseline as the cytoskeleton reforms (E).

The coupling of voltage-dependent channels to cell mechanics may have physiological consequences for remodeling the cytoskeleton by channel activation. Channel expression alone is known to restructure the cytoskeleton (48), although in that case, the effects of channel activation were minimal. The high-speed modulation of cell mechanics by voltage serves as an active amplifier in the cochlear hair cells (49, 50), but its functional role in neurons and other cells has not yet been tested. MEM serves as a useful tool to examine high-speed conformational changes of membrane proteins because the AFM and the voltage clamp respond on similar time scales, and there is little inherent cross-talk.

Methods

Cell and Molecular Biology.

Noninactivating ShakerH4 (ShIR) with A359C mutation was from Richard Horn (Thomas Jefferson University, Philadelphia, PA) (51). The IL mutant of ShakerH4 with additional V369I and I372L mutations was from Gary Yellen (Harvard University, Cambridge, MA). Acetylcholine receptor α, β, δ, ε subunits used in 2:1:1:1 ratio, respectively, were from Anthony Auerbach (University at Buffalo, State University of New York, Buffalo, NY).

All experiments were performed on tsA201 cells (HEK-293; American Type Culture Collection). Cells were grown in Dulbecco's modified Eagle's medium, supplemented with 10% FBS and 1% penicillin–streptomycin, at 37 °C in air/5% CO₂ incubator. Cells were transfected (12–24 h) with cDNA for the channel and GFP by using FuGENE 6 (Roche) according to the manufacturer's protocol. Immediately before the experiments, cells were taken out of the growth media and placed into the recording bath solution.

Physiologic bath solution was 137 mM NaCl, 5.4 mM KCl, 0.5 mM MgCl₂, 1.8 mM CaCl₂, 10 mM Hepes, 5 mM d-glucose. NMDG bath solution contained 142.4 mM NMDG instead of NaCl and KCl. Physiologic intracellular (pipette) solution contained 145 mM KCl, 5 mM NaCl, 0.5 mM MgCl₂, 10 mM EGTA, 10 mM Hepes, 5 d-glucose (pH 7.4) with KOH and 300 mOsm. NMDG pipette solution contained 149 mM NMDG and 1 mM KCl instead of 145 KCl and 5 NaCl. Symmetric K⁺ solutions were an extracellular solution with NaCl replaced with KCl (142.4 mM) All solutions were adjusted to pH 7.4 and 300 mOsm with mannitol.

Patch-Clamp/AFM.

Transfected cells were identified by GFP fluorescence, and small, rounded cells were selected. In a typical experiment, a cell was whole-cell voltage clamped according to established guidelines (52). To improve the signal-to-noise of the AFM recording and increase the speed of response, the cells were stiffened by inflation with hydrostatic pressure of 20 mmHg through the patch pipette (10). Once whole-cell configuration was established, the AFM cantilever was positioned ≈50 μm above the cell surface. The cantilever was then stepped down in 2-μm steps, and a force–distance routine was performed at the end of each step to detect contact with the cell. Upon reaching the membrane, the cantilever was engaged at the desired force (0.02–0.5 nN) and maintained at that level by using a slow proportional/integral/derivative feedback loop driven mostly by integral gain with τ_I = 500 ms. Cells were stimulated with the voltage protocols and ionic currents and corresponding cantilever deflection recorded by a data acquisition board (AT-MIO-16E2; National Instruments) controlled by custom software in Labview (53). The displacement of the cantilever was recorded as a voltage output from the bottom to top (B–T) of the photodetector and was typically ≈3 mV/nm. Displacement data were filtered (typical bandwidth, 0.1–500 Hz) and amplified (100×) by using an active 8-pole Bessel filter (Krohn–Hite model 3341). Data were averaged 5–50 (mean 12) traces to reduce random noise.

Data Selection and Analysis.

Data from seals with R_seal >1 GΩ and series resistance <5 MΩ were selected for analysis. For each dataset, the cantilever displacement was calculated at multiple time points during the pulse. Displacements were smoothed by a windowed average of ≈1 ms.