INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Hair cells convert deflections of their hair bundle into electrical signals by gating mechanically sensitive channels (Fettiplace et al. 2001; Hudspeth et al. 2000). In achieving their remarkable sensitivity, hair cells must overcome a fundamental problem; the energy at sound threshold levels is limiting (Hudspeth 1997). Hair cells overcome this barrier by mechanically amplifying low-intensity stimuli, a mechanism termed the active process.

Recent evidence suggest that at least part of the active process may arise from a feedback system regulating hair bundle movement, termed fast adaptation, directly associated with mechano-electric transducer (MET) channels (Benser et al. 1996; Fettiplace et al. 2001; Hudspeth et al. 2000; Ricci et al. 1998, 2000a). Calcium-dependent oscillations in MET currents and hair bundle movements have been measured in hair cells from turtle auditory papilla and frog saccule (Benser et al. 1996; Ricci et al. 1998, 2000a). The frequency of hair bundle oscillations varied tonotopically, suggesting a role in mechanical tuning (Ricci et al. 1998, 2000a).

Adaptation manifests itself as a decrease in MET current during a constant stimulus, presumably by changing the probability of opening of the channel (Crawford et al. 1989; Eatock et al. 1987; Hudspeth and Gillespie 1994). Adaptation is a complex process responsible for maintaining sensitivity while extending the dynamic range of the hair bundle (Crawford et al. 1989; Eatock et al. 1987; Wu et al. 1999). Two components of adaptation have been described in turtle auditory hair cells (Wu et al. 1999). The components are distinct kinetically, pharmacologically, mechanically and in displacement sensitivity (Wu et al. 1999). Fast adaptation is calcium dependent, its rate varies tonotopically, it is thought to be directly associated with MET channels, and it is the predominant form of adaptation observed in turtle auditory hair cells (Ricci and Fettiplace 1998; Ricci et al. 1998, 2000a; Wu et al. 1999). Fast adaptation may be a misnomer in that, coupled with channel gating, it is thought to underlie the mechanical oscillations described above, providing a mechanical tuning and amplification to the hair bundle (Ricci et al. 1998; Wu et al. 1999). A simple gating scheme has been developed that can account for most of the observed results regarding fast adaptation (Crawford et al. 1991; Wu et al. 1999). In this scheme, illustrated in Fig. 1, there are two closed and two open states. The transition between the first closed state and open state is displacement sensitive. The second open state has calcium bound to a site that regulates fast adaptation. The transition from this second open state to the second closed state is dictated by fast adaptation. Implicit in this scheme is the hypothesis that adaptation can modulate transducer channel kinetics. A similar but more complex model has been postulated in order to generate high-frequency oscillations (Choe et al. 1998). Variations in channel kinetics might underlie the tonotopic variation in measured adaptation rates. Experiments are designed to test this hypothesis.

Amiloride and dihydrostreptomycin (DHS) have been used to characterize properties of the MET channels (Jaramillo and Hudspeth 1991; Jorgensen and Ohmori 1988; Kimitsuki and Ohmori 1993; Ohmori 1985; Rusch et al. 1994). Both have been used to localize channels to the tops of stereocilia (Furness et al. 1996; Hackney et al. 1992; Jaramillo and Hudspeth 1991). DHS is an open channel blocker, presumably plugging the channel pore as the channel opens (Ohmori 1985). DHS has also been used to block hair bundle movements (Assad and Corey 1992; Denk et al. 1992; Ricci et al. 2000a), and to estimate gating forces and compliance associated with channel activation (Jaramillo and Hudspeth 1993). Amiloride's action is more complex, although it too requires the channel to be open for its action (Rusch et al. 1994). The present study used these compounds to investigate mechanisms underlying fast adaptation and its tonotopic distribution.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Tissue preparation

All surgical procedures were approved by the Animal Care and Use Committee at LSU Health Sciences Center. The auditory papilla from the red-eared turtle, Trachemys scripta elegans, was dissected from the half head (see Crawford and Fettiplace 1985) treated with 10-40 µg/ml protease type XXIV (Sigma, St. Louis, MO) in a solution containing (in mM) 125 NaCl, 4 KCl, 2.8 CaCl₂, 2.2 MgCl₂, 2 Na pyruvate, 2 ascorbate, 8 glucose, and 10 Na HEPES, pH 7.6, followed by removal of the tectorial membrane to expose hair bundles (Ricci and Fettiplace 1997). The trimmed preparation was mounted in a Plexiglas recording chamber, placed on a Zeiss (Oberkochen, Germany) Axioskop 2 microscope, and viewed, using Nomarski DIC optics, with a ×63 water-immersion objective (numerical aperture 0.9) and a Hamamatsu (Bridgewater, NJ) C2400 CCD camera. The chamber was perfused with a solution containing (in mM) 128 NaCl, 0.5 KCl, 2.8 CaCl₂, 2.2 MgCl₂, 2 Na pyruvate, 2 ascorbate, 8 glucose, and 10 Na HEPES, pH 7.6. An apical perfusion pipette coupled to a peristaltic pump (Gilson, Middleton, WI) through a six port manifold (Warner Instruments) was placed perpendicular to the axis of sensitivity of the hair bundle. Complete exchange of apical solutions took about 1 min (Ricci and Fettiplace 1997). Recordings were made after 3 min of perfusion. The apical perfusion solutions contained (in mM) 130 NaCl, 0.5 KCl, 2.8 CaCl₂, 2 Na pyruvate, 2 ascorbate, and 8 glucose. Amiloride and DHS (Sigma, St. Louis MO) were made as 3-mM stock solutions and serially diluted. Amiloride was heated in the dark to ensure that it was completely dissolved. In experiments with lower calcium concentrations, sodium was equimolar substituted for calcium. Calcium concentrations were directly measured with a calcium-sensitive electrode to ensure no contamination (Microelectrodes). All external solutions had pH of 7.6 and osmolalities ~275 mOsm.

Recording and stimulation procedures

Whole cell recordings were obtained with borosilicate patch electrodes filled with (in mM) 125 CsCl, 3 Na₂ATP, 2 MgCl₂, 5 creatine phosphate, and 10 Cs-HEPES, pH 7.2. The calcium buffer was either 1 or 10 mM bis-(o-aminophenoxy)-N,N,N',N'-tetraacetic acid (BAPTA; Molecular Probes, Eugene, OR), or 0.1 mM EGTA (Fluka, Ronkonkoma, NY). pH was buffered to 7.2 and osmolality fixed at 250 mOsm. CsCl was equimolar substituted with the various intracellular calcium buffers to maintain osmolality. Hair cells were voltage clamped at 80 mV using an Axopatch 200B (Axon Instruments, Foster City, CA). Series resistance, after compensation, ranged between 1 and 5 M giving recording time constants between 10 and 80 µs. The location of each cell along the long axis of the papilla was recorded at the end of each experiment.

Hair bundles were stimulated with a stiff glass fiber, 0.5-1 µm diameter, attached to a piezoelectric bimorph. Stimulus rise times were ~150 µs, measured using a dual photodiode motion detector system (Crawford and Fettiplace 1985; Ricci et al. 2000a). Voltage steps to drive the piezo element were generated with Signal software using a CED D/A converter (CED, Cambridge, UK), filtered at 5 kHz with an 8-pole bessel filter to remove bimorph resonance (Cygnus Technology) and attenuated to appropriate levels with an attenuator (Kay Elemetrics, Lincoln Park, NJ). Data were recorded at 20 kHz.

Data analysis

Maximal currents were measured from transducer activation curves and included the resting current, measured from displacements that closed MET channels, and peak currents measured from large saturating displacements. Adaptation time constants () were measured by fitting an exponential to the current decay in response to a stimulus that elicited less than half the maximal transducer current (Ricci and Fettiplace 1997) (see insets of Fig. 1 for examples). Typically a stimulus between 50 and 100 nm eliciting between 25 and 30% of the maximal current was selected. The resting open probability (P₀) was measured as the fractional current on at the resting hair bundle position. The IC₅₀ was obtained by fitting the normalized current versus dose with the Hill equation: I/(I_max) = ([drug]ⁿ/(EC₅₀ⁿ + [drug]ⁿ), where n is the Hill coefficient. All data are plotted as means ± SE.

View larger version (31K): [in this window] [in a new window]   Fig. 1. High-frequency cells have larger transducer currents and faster adaptation than low frequency cells. A and B: examples of mechano-electric transducer (MET) currents elicited from a low (A) and a high- (B) frequency hair cell. Stimuli are shown above the current traces. In this and subsequent figures, upward stimuli indicate movement toward the kinocilium. The maximum stimulus was 350 nm. In each figure, records are averages of 16. The insets in A and B are the MET current responses to the smallest stimuli evoking an inward current, truncated to illustrate the difference in the kinetics of adaptation. The solid lines represent fits to a single exponential from which the time constant of adaptation () was obtained. C: peak current vs. displacement plots for the cells shown in A and B. Solid lines represent fits to a double Boltzmann functions, I/Imax = {1 + exp(a1x0  a1x)*[1 + exp(b1x0  b1x)]}  1, where x0 = 0.037 µm and 0.06 µm, a1 = 10.9 and 10, and b1 = 26 and 27 µm1 for the low- and high-frequency cell. Resting open probability (P0) was measured as the ratio of current on at zero displacement to the maximal current elicited. D: a simplified schematic of the postulated gating scheme representing fast adaptation's regulation of MET channels (Crawford et al. 1989; Wu et al. 1999). Dashed lines represent possible sites of interaction for dihydrostreptomycin (DHS).

Noise analysis

The transducer current signal from free-standing bundles was AC coupled and amplified up to ×50 to use the full dynamic range of the A/D converter. Current samples of 300 ms duration were filtered at 5 kHz (Butterworth filter, Wavetek, Rockland, NJ) and recorded at 20 kHz. Traces were rescaled and power spectra generated using Origin software (Microcal, Northampton, MA). A significant problem faced was that the rolloff of the spectra quite often was very close to the rolloff of the filter and that set by the recording conditions. It was critical to maintain a very low series resistance; compensated series resistance of less than 3 M were required for high-frequency cells. Only cells where the rolloff due to MET channels was clearly separate from the noise rolloff were included for analysis. The sharp rolloff seen at the end of the spectra was a function of the filter used. It is also important to obtain a good record of background noise in order to account for any voltage-dependent channels that might be flickering as well as to assess directly the filtering properties of the electrode and filter being used for the recordings. Background noise was measured in two manners. First, the bundle was statically biased away from the kinocilium, turning off transducer channels. A problem with this approach was that the bundle tended to adapt back, turning some channels on. The second method used to obtain background spectra was to block MET channels with DHS. Similar results were obtained with both methods. Fifty to 100 traces were obtained. Fourier analysis was performed, and the resulting spectra were averaged. Average background spectra were subtracted from the test conditions. Spectra were fitted with single Lorentzian functions. Single Lorentzians can describe quite complex channel behavior, and in this present study the Lorentzian was being used to estimate the cutoff frequency (f_c) (Colquhoun and Hawkes 1977). Estimating f_c as the frequency at which the power had dropped to half its low-frequency value gave comparable results. In several high-frequency cells, power spectra were limited by voltage-clamp speed; these cells were excluded.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Tonotopic differences in transducer currents

MET currents were recorded from hair cells at a high (d = 0.74 ± 0.01, mean ± SE) and low- (d = 0.27 ± 0.01) frequency region of papilla, measured as relative distance (d) from the lagena (Fig. 1). A tonotopic gradient existed in both the time constant of transducer adaptation high frequency ( = 0.74 ± 0.07 ms, n = 34) and low frequency ( = 2.6 ± 0.3 ms, n = 31) and maximal transduction current high frequency (I_max = 0.87 ± 0.04 nA, n = 34) and low frequency (I_max = 0.34 ± 0.03 nA, n = 31). Activation curves were obtained by plotting the peak current versus displacement. Previous reports have demonstrated that these data are best fit by a double Boltzmann function (Crawford et al. 1989), so this was used to fit activation curves normalized to peak current. The equation used was I/I_max = {[1 + exp(a₁x₀  a₁x)]*[1 + exp(b₁x₀  b₁x)]}¹ where a₁ and b₁ represent displacement sensitivity values and x₀ represents the setpoint. No differences (Student's t-test, P > 0.05) were found between high- and low-frequency cells in the values for a₁ or b₁. Values obtained were, high frequency, 24 ± 5 µm¹ (n = 14) and low frequency, 26 ± 5 µm¹ (n = 19) for a₁. For b₁ values were, high frequency, 26 ± 4 µm¹ and low frequency, 27 ± 4 µm¹. There was a small difference in the setpoint value, x₀ obtained (Student's t-test P < 0.05). Values obtained were 0.07 ± 0.01 µm (n = 14) and 0.10 ± 0.01 µm (n = 19) for low- and high-frequency cells, perhaps reflecting a difference in calcium load.

The schematic of the hypothesized MET channel-gating scheme is shown in Fig. 1D. This schematic, most likely an oversimplification of a complex process, is used here as a tool for designing experiments and a framework from which to interpret data. In this scheme K₁ represents the displacement sensitivity of the process (Crawford et al. 1991; Wu et al. 1999). Since no difference has been observed in either displacement-sensitive component of the activation curves, as measured from the double Boltzmann fits to the data, K₁ is presumed to be constant between MET channels at different papilla locations. This does not mean that the activation kinetics are assumed to be identical. K₂ represents the process underlying fast adaptation. Differences in the rate constants of K₂ might underlie the tonotopic differences in adaptation.

DHS sensitivity

The gating scheme suggests that fast adaptation contributes to the channel kinetics. Cells that adapt at different rates are predicted to have different channel kinetics, and this would make them more or less sensitive to inhibition by an open channel blocker such as DHS (Ohmori 1985). Examples of transducer current responses and the dose-response relationship for DHS from hair cells at two papilla locations are given in Fig. 2. Low-frequency cells, that are predicted to have slower channel kinetics, are more sensitive to DHS block than are the high-frequency cells. The solid lines represent fits to the Hill equation with Hill coefficients of 1 and IC₅₀ values of 14 ± 2 µM and 74 ± 5 µM for low- and high-frequency cells, respectively. There was an n of at least 6 for each point. As an open channel blocker, a time-dependent decay in the current would be expected in response to hair bundle displacements. This was not observed for cells at either papilla location. Nor has this been observed by other investigators (Jaramillo and Hudspeth 1991, 1993; Kroese et al. 1989; Ohmori 1985). The lack of a time-dependent block may be due to DHS blocking very fast, much faster than the stimulus rise-time of the probe so that the channels are being blocked faster than is observable by the recording system. On the other hand the block could be very slow so that further block is not observed during the short duration pulses used to elicit activation curves. And finally, it is possible that DHS is not blocking the open channel as reported (Ohmori 1985) but can block the closed state as well. Experiments described below help to delineate between these possibilities. Regardless of drug mechanism, the lack of time-dependent block is useful because it allows for measurements of adaptation rate uncontaminated by the kinetics of drug binding.

View larger version (28K): [in this window] [in a new window]   Fig. 2. DHS blocks MET currents from low- and high-frequency hair cells with different efficacy. Examples of a low-frequency (A) and high-frequency cell (C) for control (top) and in the presence of 10 or 100 µM, dihydrostreptomycin (DHS), doses selected to be near the IC50 values, as indicated (bottom panels). Stimuli are shown at top (maximum stimuli 300 nm). B: dose-response curves for the low- () and high () frequency cells. Solid lines represent fits to the Hill equation with Hill coefficients of 1 and IC50s of 14 ± 2 µM and 75 ± 5 µM for low- (r = 0.99) and high- (r = 0.99) frequency cells, respectively. Traces are averages of 16, and the n varied from 5 to 12 at each point.

Summation

Previous work suggested that the rate of adaptation varied with the MET current magnitude (Ricci and Fettiplace 1997; Ricci et al. 1998). From this a hypothesis was devised suggesting that the rate of adaptation is modulated by the dynamics of intraciliary calcium, such that a faster or larger change in intraciliary calcium results in faster adaptation. Multiple channels per stereocilia and a tonotopic increase in the number of channels per stereocilia would lead to a gradient in the rate of adaptation. One problem with this previous work has been that it relies on a comparison of adaptation between hair cells. Using DHS allows investigation of changes in the time course of adaptation in individual cells where the magnitude of the current is altered pharmacologically, thus the number of operational stereocilia remains constant. Figure 3 summarizes the results obtained, measuring both the time constant of adaptation () and the resting open probability (P₀) of the channel. As the transducer channels are blocked, adaptation gets slower. In agreement with previous data, these data demonstrate that summation is an important mechanism involved in establishing the tonotopic gradient of adaptation. Summation most likely refers to multiple channels per stereocilia resulting in faster, larger changes in intraciliary calcium levels. More detailed analysis of stereociliary calcium levels are required to better understand this process.

View larger version (19K): [in this window] [in a new window]   Fig. 3. Summation is an important mechanism regulating adaptation. A: example of a high-frequency transducer current elicited from a 50-nm bundle deflection (top trace) illustrating fast adaptation (left) and the slowing of adaptation when MET currents were blocked (right) with DHS. Scale bars represent 5 ms and 0.1 nA. Solid lines represent exponential fits with time constants of 1 and 3.5 ms for control and in the presence of DHS. The fraction of total transducer current blocked was plotted against the adaptation time constant () (B) or the resting open probability (P0), so that more current is blocked as x-axis increases (C). In both plots, filled squares represent low-frequency cells and open circles represent high-frequency cells. The n varied from 5 to 12 for each point. For each plot, linear regressions are shown; the correlation coefficients for each fit were >0.98.

MET calcium permeability

DHS blocks in a calcium-dependent manner (Kroese et al. 1989). It is possible that DHS is competing with calcium in the channel and that the different IC₅₀ values obtained were a function of calcium permeability rather than a function of channel kinetics. It is also possible that summation described above is due to an increased calcium permeability and not an increase in the number of channels per stereocilia. Theoretically, from the schematic in Fig. 1D, differences in calcium permeation could be responsible for the tonotopic differences in adaptation rate and not intrinsic differences in channel kinetics. DHS results alone cannot differentiate between these two possibilities. Calcium both blocks and permeates the transducer channel (Crawford et al. 1991; Lumpkin et al. 1997; Ohmori 1989; Ricci and Fettiplace 1998). To assess calcium blocking properties, external calcium was reduced from 2.8 to 0.25 mM and evaluated by comparing changes in peak current (Ricci and Fettiplace 1998). The ratio of peak current in 0.25 mM calcium to peak current in 2.8 mM calcium were 1.7 ± 0.1 (n = 8) and 1.8 ± 0.3 (n = 7) for high- and low-frequency channels, respectively. Measurements with Tris as the monovalent ion assessed calcium permeability through MET channels (Ricci and Fettiplace 1998). Tris does not permeate the transducer channel (Ricci and Fettiplace 1998). Ratios of peak current in Tris to peak current in Na gave values of 0.22 ± 0.01 (n = 7) for high- and 0.20 ± 0.03 (n = 8) for low-frequency channels. No statistical differences were found between papilla locations for either measurement using two-tailed Student's t-test (P > 0.05). Therefore data that suggest differences in DHS efficacy between frequency positions were not a function of the differences in calcium permeation or block of the MET channel at these locations and were most likely due to differences in channel kinetics.

Calcium dependence of DHS block

DHS block was sensitive to external calcium concentrations. Dose-response curves were obtained for DHS in 0.25 mM calcium to investigate the calcium dependency of the block (Fig. 4). Lowering external calcium shifted the IC₅₀ values for cells at both papilla locations. Values obtained were 6 ± 1 µM and 0.8 ± 0.1 µM for high- and low-frequency cells, respectively. Hill coefficients were 1. Since calcium sensitivity is not a result of a direct interaction between DHS and calcium, the sensitivity is most likely explained by the effect lowering external calcium has on adaptation. Lowering external calcium concentration from 2.8 to 0.25 mM slowed adaptation and increased P₀. For high-frequency cells,  increased from 0.74 ± 0.07 ms to 3.3 ± 0.3 ms (n = 12), while the P₀ varied from 0.04 ± 0.01 to 0.12 ± 0.01. For low-frequency cells,  slowed from 2.6 ± 0.3 ms to 6.9 ± 0.9 ms (n = 11), and P₀ increased from 0.09 ± 0.01 to 0.130 ± 0.001. If fast adaptation is regulating transducer channel kinetics, then slowing adaptation will slow channel kinetics making the channels more vulnerable to DHS block.

View larger version (18K): [in this window] [in a new window]   Fig. 4. Dose-response curves for DHS obtained in 0.25 mM external Ca2+. Lowering external calcium shifted the IC50 of DHS for both low- () and high- () frequency cells (n = 6). The dotted lines through the data points are the fits to Hill equation with Hill coefficients of 1 and IC50s of 0.8 ± 0.2 µM (r = 0.97) and 6 ± 1 µM (r = 0.98) for low- and high-frequency channels, respectively. The dose-response curves from Fig. 2, where 2.8 mM calcium was external, are reproduced for comparison with the solid line representing the high-frequency and the dashed line representing the low-frequency responses.

If DHS efficacy is a function of the MET channel kinetics and adaptation modulates these kinetics, then it should be possible to alter efficacy by altering adaptation. To test this hypothesis, experiments were designed to make adaptation in low-frequency cells fast and adaptation in high-frequency cells slow (Fig. 5). Adaptation was made fast by replacing the intracellular calcium buffer, 1 mM BAPTA with a lower concentration of the slower buffer 0.1 mM EGTA. Under these conditions the time constant of adaptation changed from 2.6 ± 0.3 ms to 0.9 ± 0.1 ms for low-frequency cells. The IC₅₀ for DHS increased to 43 ± 4 µM (n = 6), supporting the argument that adaptation regulated DHS efficacy (Fig. 5). In high-frequency cells adaptation was slowed by increasing the recording solution concentration of BAPTA from 1 to 10 mM. The time constant of adaptation slowed from 0.74 ± 0.07 ms to 2.3 ± 0.5 ms (n = 6), and the IC₅₀ for DHS was reduced to 25 ± 3 µM (n = 6; Fig. 5).

View larger version (30K): [in this window] [in a new window]   Fig. 5. IC50 for DHS is regulated by adaptation. Top panels show stimulus protocol with maximal stimulus equal to 300 nm. A: adaptation was sped up in low-frequency cells by replacing 1 mM BAPTA with 0.1 mM EGTA in the recording pipette (n = 6), and adaptation was slowed in high-frequency cells by using 10 mM BAPTA as opposed to 1 mM BAPTA in the recording pipette (n = 6) (C). Bottom panels show responses in the presence of 30 µM DHS. DHS dose-response curves were obtained and are plotted in C; solid squares represent low-frequency and open circles high-frequency cells. Dotted lines through points are fits to the Hill equation with Hill coefficients of 1. In this case high-frequency cells are more sensitive to DHS, with IC50s of 43 ± 4 µM (r = 0.99) and 26 ± 3 µM (r = 0.99) for low- and high-frequency cells, respectively. Fits from Fig. 2, where 1 mM BAPTA was the intracellular calcium buffer, are shown for comparison with the solid line representing high-frequency and the dashed line low-frequency responses.

A summary plot of DHS IC₅₀ against adaptation time constant , or resting open probability (P₀) obtained under the different buffering and external calcium concentration conditions is given in Fig. 6. The plot illustrates the relationship between adaptation and DHS efficacy. It is unclear from this plot whether the relationship is the same for high- and low-frequency cells. For simplicity a single regression line (r² = 0.98) was plotted; however, a comparison of the lowest efficacy conditions for the two positions, i.e., 1 mM BAPTA in 2.8 mM external calcium for the high-frequency compared to 0.1 mM EGTA in 2.8 mM external calcium for the low-frequency position, is revealing. Although no statistical difference exists in either the P₀ or the time constant of adaptation between these points (Student's 2-tailed t-test, P > 0.05), the IC₅₀ values were statistically different (Student's 2-tailed t-test, P < 0.01). This result might suggest that, although adaptation shifts DHS efficacy at a given papilla location, additional differences between MET channels are needed to fully explain the DHS effects at different frequency positions.

View larger version (16K): [in this window] [in a new window]   Fig. 6. Plots of the IC50s for DHS obtained under the various conditions described against either resting open probability (P0) of the channel (A) or the adaptation time constant () (B). The correlation coefficient for both plots was 0.91.

Amiloride blocking properties

The lack of a use-dependent block and the calcium sensitivity of the DHS block, although consistent with the present hypothesis regarding adaptation, are unusual properties for an open channel blocker. In addition, it is possible, although unlikely, that there are diffusional barriers limiting DHS access differently at the two papilla locations, so an additional control drug is useful. To this end it seemed prudent to use another type of blocker as a control for the DHS experiments. Amiloride and amiloride derivatives are known inhibitors of mechanically gated channels (Jorgensen and Ohmori 1988; Lane et al. 1993; Rusch et al. 1994). Amiloride block is complex; it does not plug the pore of the channel as DHS is purported to do, but it does require the channel to be open (Lane et al. 1993; Rusch et al. 1994). MET current responses from low- and high-frequency hair cells as well as a dose-response relationship for amiloride are given in Fig. 7. No difference in IC₅₀, 24.2 ± 0.5 µM was found. The Hill coefficient was 2, suggesting that two molecules of amiloride were required for block. The presence of a use-dependent component of the block made analyzing effects on adaptation difficult, although it is clear from the examples shown that the responses to small hair bundle deflections were slowed in the presence of amiloride. A time-dependent component of amiloride block has been reported previously, and the kinetics were shown to vary with concentration (Rusch et al. 1994). Figure 8 shows an example of the time-dependent block of MET current by amiloride. The current traces are responses to maximal stimuli where adaptation is not observed, normalized so that the use-dependent effects of amiloride can be observed. Single exponential curves were fit to these traces to measure time constants of amiloride block. A plot of the time constant measured, against amiloride concentration, is given in Fig. 8B. Low-frequency cells are blocked more slowly than high-frequency cells, although with similar efficacy. Although the mechanism responsible for the difference is unknown, it also suggests that there are differences between the MET channels at different frequency locations.

View larger version (28K): [in this window] [in a new window]   Fig. 7. MET currents from a low-frequency (A) and high-frequency cell (C), maximal stimulus 300 nm. Bottom panels were obtained in the presence of 30 µM amiloride. Stimuli (maximum stimuli 300 nm) are shown above current records. Dose response curve (B) for both high- () and low-frequency () cells where the solid line is the regression fit to the Hill equation with a Hill coefficient of 2 and an IC50 of 24.2 ± 0.5 (r = 0.99). The n varied from 3 to 8 for each dose.

View larger version (14K): [in this window] [in a new window]   Fig. 8. An expanded time base of normalized currents in response to the largest mechanical bundle deflection from Fig. 7 in the absence () and presence (- - -) of 30 µM amiloride is shown in A and C. Summary plot of the time constants, measured from exponential fits of the current decay, against amiloride concentration (B).

Thus far, the pharmacological data support the hypothesis that MET channels are kinetically different at different papillary locations. Data suggest that DHS efficacy is modulated by MET channel kinetics and that adaptation contributes to the kinetics. It is possible that variations in the probability of opening of the channel are responsible for the pharmacological results and not the channel kinetics. Increasing the probability of opening could result in channels being in the open state more and so more susceptible to the open channel block of DHS. DHS results alone cannot clearly differentiate between these possibilities. However, experiments that biased the hair bundle did not change DHS efficacy (data not shown), suggesting that channel kinetics and not open probability was responsible for efficacy.

Noise analysis

To more directly determine whether MET channel kinetics vary with papilla location, stationary noise analysis was performed. Noise analysis is based on the fact that a steady-state signal made up of independent unitary events (single channels) exhibits fluctuations about a mean level (Anderson and Stevens 1973; Jackson and Strange 1995). This technique takes advantage of the fact that at the hair bundle's resting position the probability of opening is small. A simple two-state model would predict that a single Lorentzian function would describe the isolated power spectra. The present model system (Fig. 1D) (Wu et al. 1999) predicts at least four states, and other theoretical work has predicted even more (Choe et al. 1998), suggesting that the power spectra would be quite complex. The point of the present experiment is to determine whether adaptation kinetics vary with papilla position. The displacement sensitivity appears to be the same between hair cells at different papilla locations, so for simplicity it is assumed that K₁ (from Fig. 1D) is a constant between positions and is not calcium dependent (Fig. 1). In addition, the calcium permeability has also been demonstrated to be equivalent between positions, so in this model system K₂ is predicted to vary between papilla positions and also with external calcium concentrations. Therefore differences observed in the power spectra obtained under different external calcium conditions or between papilla locations will largely be attributable to K₂. Examples of the background noise as well as the noise in 2.8- and 0.25-mM external calcium concentrations are given in Fig. 9. Details of the recordings and data processing are given in METHODS. The power spectra for the noise were typically two to three orders of magnitude lower than the experimental traces. Power spectra for each condition are given in Fig. 9B. The traces representing MET channel noise in 2.8 and 0.25 mM have had the background spectrum subtracted. Although the spectra can be quite complex, the half power frequency could be obtained for cells at each papilla location. Values obtained were 1,185 ± 148 Hz (n = 6) and 551 ± 145 Hz (n = 5) for high- and low-frequency cells in 2.8 mM external calcium and (Fig. 9B), 768 ± 205 Hz (n = 4) and 289 ± 63 Hz (n = 4) when external calcium was lowered to 0.25 mM. Values were obtained by either fitting a single Lorenzian function to the part of the curve attributable to transduction (solid lines on spectra in Fig. 9B) or by measuring the half power frequency as compared to the low-frequency measurement. In its most simple form, the half power frequency is proportional to the mean open time of the channel. In our more complicated system, the half power frequency is probably a function of the rate limiting channel kinetics, and the differences measured between papilla locations suggest that channel kinetics vary. The half power frequency is calcium sensitive, implicating adaptation kinetics. Slowing adaptation by lowering external calcium shifted the half power frequency to lower values, indicating that fast adaptation can directly alter channel properties. A summary plot of the half power frequency against the macroscopic adaptation time constant is given in Fig. 10, demonstrating a correlation. Together these data support the hypothesis that MET channel kinetics vary tonotopically.

View larger version (37K): [in this window] [in a new window]   Fig. 9. Noise analysis suggests the kinetics of MET channels varied with frequency and were calcium sensitive. A: examples of the background, obtained by blocking with DHS, and channel noise in 2.8 and 0.25 mM calcium for low-frequency (left) and high-frequency (right) hair cells. Power spectra were averaged and background spectra subtracted to give plots shown in B for the data presented in A for the noise acquired in 2.8 and 0.25 mM calcium. Solid lines are fits to a single Lorentzian function. The cutoff frequencies (fc) are indicated with arrows. The higher power traces are from recordings where the external calcium concentration was 0.25 mM.

View larger version (14K): [in this window] [in a new window]   Fig. 10. Plots of the relationship between the macroscopic adaptation time constant and the cutoff frequency measured from noise analysis. High- and low-frequency cells in 2.8 or 0.25 mM external calcium concentration are included. The solid line represents a least-squares regression fit of an exponential demonstrating the correlation between  and Fc.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Summary

The major findings presented in this manuscript include the following: DHS blocks high- and low-frequency MET channels with different efficacy, the efficacy was calcium-dependent and varied with fast adaptation, no use dependence was observed, DHS block slowed adaptation, amiloride's block of MET channels was calcium-independent and had similar efficacies between papilla locations and showed use dependence, the kinetics of amiloride block were different between papilla locations, and finally noise analysis demonstrates calcium-dependent kinetic differences between MET channels of high- and low-frequency hair cells.

Channel kinetics hypothesis

One possible explanation for the above data is that MET channel kinetics vary tonotopically. DHS, an open channel blocker, will be sensitive to the channel state (see Fig. 1D); the longer the channel stays open, the better chance for DHS to block and so the greater efficacy. Low-frequency channels may stay open longer than high-frequency channels, resulting in greater DHS efficacy. Fast adaptation may alter the time the channel stays in the open state, in the steady-state condition by regulating the probability of opening of the channel. On the other hand based on the channel scheme in Fig. 1D, the rate constants associated with K₂ will directly modulate the time the channel spends open, and thus variations in K₂ could alter DHS efficacy. In this way any manipulation that alters fast adaptation would be predicted to change DHS efficacy; thus the calcium dependence of the block. The lack of a use-dependent block by DHS also supports this argument. It suggests that the rate-limiting step in determining DHS block is the channel kinetics. DHS interacts with the channel much faster than the channel opens and closes; thus the differential sensitivity to DHS of high- and low-frequency channels. In contrast amiloride shows a use-dependent block, suggesting that the rate limiting step is drug binding. Since drug binding is considerably slower than channel gating, a difference in channel gating between papilla locations does not alter the effectiveness of the drug. Additional support of this kinetic argument includes the DHS sensitivity to calcium, the noise analysis results indicating a difference in channel kinetics that are calcium sensitive and the amiloride data suggesting amiloride binding kinetics are different between papilla locations. Implicit with this argument is that the calcium-binding site must be tightly coupled to the MET channel.

DHS competes for DHS binding site

An alternative but plausible hypothesis to explain the DHS data presented is that DHS competes with calcium for a binding site on the channel. As depicted in the gating scheme of Fig. 1D, calcium enters the open MET channel and binds triggering channel closure. If DHS competes for the calcium-binding site, then it would be predicted that adaptation would slow in the presence of DHS and DHS efficacy would be a function of intraciliary calcium. Thus conditions where intraciliary calcium concentrations are high would limit DHS efficacy. For example if, as predicted, high-frequency cells adapt fast because of increased calcium load, then DHS efficacy would be less than in low-frequency cells where the calcium load was less. Altering calcium buffer or changing external calcium concentrations would change intraciliary calcium and thus alter DHS efficacy. The covariation of DHS efficacy and fast adaptation would reflect a common regulation by calcium and not necessarily an effect of fast adaptation. Implicit with this argument is that summation of calcium underlies the tonotopic differences in adaptation rate. The other data including amiloride block and noise analysis measurements would be explained by the role of summation. This hypothesis would not require that channels be intrinsically different, only that the calcium load increase tonotopically due to an increase in number of channels per stereocilia. One difficulty with this hypothesis is that fast adaptation is thought to be a calcium-dependent feedback process that serves to maintain calcium concentrations at a constant level. If this feedback system is operating as predicted (Ricci et al. 1998), then at steady-state the calcium concentration at the adaptation binding site is a constant regardless of external calcium concentrations and so DHS efficacy would be predicted to remain a constant as well.

Summation

Do the above arguments require MET channels to be different? The simple answer is no. Summation, a term meant to represent the finding that MET channels interact in such a way that having more channels makes adaptation faster, might account for the apparent pharmacological differences. The simplest mechanism for summation is calcium entering through MET channels sum. Faster changes in calcium concentration or larger calcium changes will speed adaptation (Ricci and Fettiplace 1998; Wu et al. 1999). Since data presented demonstrate calcium permeability properties of MET channels are the same between frequency locations, summation implies an increase in the number of channels per stereocilia. If high-frequency cells have more channels per stereocilia than adaptation would be faster and DHS efficacy would be less even though the channels are physically the same. A similar argument could be used to explain both the noise analysis and amiloride data. A confounding variable in this hypothesis is that the calcium sensitivity of the MET channels might be different. This possibility would also require calcium summation but might explain why high-frequency cells are consistently faster than low-frequency cells even at comparable current magnitudes. It might also explain why the steepness of the plots in Fig. 3 are different between frequency locations. Implicit in this argument is the rate-limiting step in establishing the rate of adaptation is the rate of change of stereocilia calcium. This mechanism is certainly not unprecedented in that it is how the BK channel regulates deactivation kinetics and thereby resonant frequency (Art et al. 1995; Jones et al. 1999; Ricci et al. 2000b). A difference in calcium sensitivity can also account for all the above data and would require the MET channel to be different between papilla locations.

A difficulty with directly estimating the number of MET channels per stereocilia is that the number of stereocilia per hair bundle vary tonotopically, and most likely the number of functioning stereocilia also vary (Hackney et al. 1993). If we assume that the largest MET currents measured at a particular location reflect all the stereocilia operating or at the very least that a comparable proportion of stereocilia are functioning, then an estimate can be obtained. For high-frequency cells, the maximal current recorded at 0.74 position was 1,900 pA, assuming 8 pA (Crawford et al. 1989) per channel gives 238 channels for a hair bundle having about 80 stereocilia resulting in 3 channels per stereocilia (Hackney et al. 1993). Low-frequency cells have a maximal current around 600 pA giving about 75 channels in a hair bundle having about 40 stereocilia or 1.8 channels per stereocilia. Most likely these are underestimates, but they do suggest close to a doubling of the calcium load between these positions. Is this enough of a change to explain the tonotopic distribution of adaptation time constants? Previously, a linear relationship between rate of adaptation and proportion of MET current carried by calcium was described that had a slope of 2.2 ms¹ nA¹ (Ricci and Fettiplace 1998, Fig. 7). Since we know from data presented herein that the fractional current carried by calcium remains a constant between frequency locations, we can use the above relationship to estimate whether the change in calcium load can predict the measured adaptation time constants. The low-frequency time constant of 2.6 ms predicts a calcium current of 175 pA; doubling this to 350 pA predicts a time constant of 1.3 ms, much slower than the measured 0.74 ms of the high-frequency cells. However, the relative difference in maximal current between mean values was 2.6. Using this value to estimate the calcium load difference gives a time constant of 1 ms, still considerably slower than the measured 0.74 ms. This result would argue that additional differences between channels are needed to explain the tonotopic distribution of fast adaptation.

Another approach in attempting to determine whether an increase in the number of channels per stereocilia can adequately explain tonotopic differences in adaptation is to again use the relationship between adaptation rate and the proportion of MET current carried by calcium (Ricci and Fettiplace 1998, Fig. 7). These data were obtained from a papilla region between 0.55 and 0.65, a location between the two positions used in the present study. The low-frequency current of 340 pA predicts an adaptation time constant of 2.3 ms, faster than the measured value of 2.6 ms, and the high-frequency current of 870 pA predicts a time constant of 1 ms slower than the measured high-frequency value of 0.74 ms. A simple explanation for the observed differences is that the MET channels are different between frequency locations. The difference in channel properties may be in calcium sensitivity or in kinetics, much like the BK channel properties in these cells (Ricci et al. 2000b).

Although intrinsic kinetic differences are not mandated by the present data, the above argument suggests that an additional mechanism besides summation is necessary to account for the tonotopic variation in adaptation. Kinetic differences between MET channels can account for the difference in DHS efficacy, amiloride binding kinetics and noise analysis data obtained between frequency positions. If DHS binding to the channel altered the rate constants of channel closing it could also explain the apparent summation response. It should be pointed out that kinetics in this case could equally refer to activation or adaptation kinetics (i.e., K₁ or K₂). That is, for simplicity it has been assumed that K₁ (Fig. 1D) is the same between frequency locations. To date there is no direct evidence to this effect. A difference in activation kinetics could indirectly alter adaptation rates by changing the dynamics of intraciliary calcium, much like slowing the stimulus rise-time slows adaptation rate (Wu et al. 1999). Noise analysis suggests a kinetic difference that may or may not be directly associated with adaptation. Calcium sensitivity implicates adaptation; however, the calcium sensitivity of activation remains to be explored and so cannot be ruled out. Only direct measurements of single-channel properties will distinguish between these possibilities.

DHS as an open channel blocker

It would be expected that as an open channel blocker DHS would show some stimulus dependence or use-dependent block. This was not observed and at first would suggest that perhaps the block was not an open channel as reported (Ohmori 1985). The lack of stimulus-dependent block is most likely a function of DHS binding rapidly to the open channel, more rapid than the stimulus rise-time or recording system allows for detection. It may be the relative difference between DHS binding kinetics and MET channel kinetics that results in the different efficacies at different papilla locations. This argument is supported by data used to localize MET channels to the tops of the stereocilia (Jaramillo and Hudspeth 1991). Slower channel blocking would have dissipated the iontophoretic gradient of DHS masking the positional sensitivity reported. In addition, the rapid change in bundle mechanics used to estimate gating compliance also supports a rapid action of DHS (Jaramillo and Hudspeth 1993). In contrast, amiloride binding is the rate-limiting step for amiloride block because it binds more slowly, as indicated by the time constant of the use-dependent component of the block. The slow binding of amiloride makes it less influenced by MET channel kinetics, and thus there is no difference in efficacy. However, the difference in MET current kinetics in the presence of amiloride implies a difference in amiloride binding kinetics for apical and basal hair cells.

Tonotopic variations in MET channels

Data presented explored a variety of MET channel properties including displacement sensitivity, calcium permeation and block, pharmacology, and channel kinetics. No difference was found in displacement sensitivity, suggesting that the mechanical coupling of the channel is conserved across frequency positions. No difference was found in the calcium permeation and block of the MET channel, suggesting that the pore of the channel may also be conserved across the papilla. The MET channels presented as pharmacologically distinct. However, as described above, the pharmacological differences do not require the channels to be different. No difference in IC₅₀ was found for amiloride. Together these data suggest that the majority of channel properties are conserved but suggest a novel mechanism for regulating the kinetic properties of MET channels through summation of intraciliary calcium. However, although intrinsic differences in MET channel properties are not mandated by the present data, theoretical arguments presented above suggest that summation alone cannot account for the tonotopic differences in adaptation. From this it is hypothesized that intrinsic differences between MET channels are in part responsible for the differences in measured adaptation rates. These differences may be in calcium sensitivity or in channel kinetics.

How fast can MET channels operate?

The fastest adaptation time constant measured is equivalent to the stimulus rise-time of the system, roughly 100 µs. Faster stimulation will be necessary to determine the upper limit of the channel responses. Theoretically, the rate-limiting step in adaptation is the rate of change of stereociliary calcium. Ultimately, the rate of change of calcium will be limited by the activation kinetics of the MET channels. Noise analysis also suggests that MET channels can operate at high frequencies; power spectra often remained flat into the kilohertz range, being limited by the recording system. Considering the physiological operating range of these channels in turtle, it is quite remarkable that they appear to be capable of operating at frequencies an order of magnitude higher than is physiological. Indirect measures of transducer channel activation suggest that the channels can operate into the tens of kilohertz range, so it is quite likely that adaptation and the corresponding mechanical tuning mechanism can work at these rates (Yates and Kirk 1998). It seems quite possible that specializations of mammalian auditory hair cell bundles coupled with summation, i.e., increasing the number of channels per stereocilia, and possibly variations in MET channel kinetic properties, will serve to extend the frequency range of the MET channels even further. Only direct measurements in a mammalian system can answer these important questions.

In summary, data presented included the following. DHS blocked MET channels differently at low- and high-frequency positions. The difference in efficacy was calcium-dependent and may be the result of differences in channel kinetics or perhaps in calcium load of the stereocilia. Amiloride blocked MET channels of different frequency positions with similar efficacy but with different binding kinetics. Noise analysis suggests that MET channel kinetics varied tonotopically. The kinetic differences may be due to intrinsic channel properties or may be due to an increase in number of channels per stereocilia.