Biophysical and Pharmacological Characterization of Voltage-Gated Calcium Currents in Turtle Auditory Hair Cells


  • M. E. Schnee,

    1. Neuroscience Center and Kresge Hearing Laboratories, Louisiana State University Health Sciences Center, New Orleans, LA 70112, USA
    Search for more papers by this author
  • A. J. Ricci

    Corresponding author
    1. Neuroscience Center and Kresge Hearing Laboratories, Louisiana State University Health Sciences Center, New Orleans, LA 70112, USA
    • Corresponding author
      A. J. Ricci: Neuroscience Center and Kresge Hearing Laboratories, 2020 Gravier Street Suite D, LSU Health Sciences Center, New Orleans, LA 70112, USA. Email:

    Search for more papers by this author


Hair cell calcium channels regulate membrane excitability and control synaptic transmission. The present investigations focused on determining whether calcium channels vary between hair cells of different characteristic frequencies or if multiple channel types exist within a hair cell, each serving a different function. To this end, turtle auditory hair cells from high- (317 ± 27 Hz) and low-frequency (115 ± 6 Hz) positions were voltage clamped using the whole-cell recording technique, and calcium currents were characterized based on activation, inactivation and pharmacological properties. Pharmacological sensitivity to dihydropyridines (nimodipine, Bay K 8644), benzothiazepines (diltiazem) and acetonitrile derivatives (verapamil, D600) and the insensitivity to non-L-type calcium channel antagonists support the conclusion that only L-type calcium channels were present. Fast activation rise times (< 0.5 ms), hyperpolarized half-activation potentials and a relative insensitivity to nimodipine suggest the channels were of the α1D (CaV1.3) variety. Although no pharmacological differences were found between calcium currents obtained from high- and low-frequency cells, low-frequency cells activated slightly faster and at hyperpolarized potentials, with half-activating voltages of −43 ± 1 mV compared to −35 ± 1 mV. Inactivation was observed in both high- and low-frequency cells. The time course of inactivation required three time constants for a fit. Long depolarizations could result in complete inactivation. The voltage of half-inactivation was −40 ± 2 mV for high-frequency cells and −46 ± 2 mV for low-frequency cells. Calcium channel inactivation did not significantly alter hair cell electrical resonant properties elicited from protocols where the membrane potential was hyperpolarized or depolarized prior to characterizing the resonance. A bell-shaped voltage dependence and modest sensitivities to intracellular calcium chelators and external barium ions suggest that inactivation was calcium dependent.

Calcium channels are fundamental to signal processing in auditory sensory hair cells, regulating both the membrane excitability and neurotransmitter release (Roberts et al. 1990). Electrical resonance, the ability of the hair cell's membrane potential to oscillate at a particular frequency, is the primary tuning mechanism of auditory hair cells in lower vertebrates (Crawford & Fettiplace, 1978; Ashmore, 1983; Lewis & Hudspeth, 1983; Fuchs et al. 1988). Electrical resonance is driven by the interaction between calcium channels and calcium-activated potassium (BK) channels (Art et al. 1986; Hudspeth, 1986; Art & Fettiplace, 1987). Tonotopic variations in the magnitude of both channel types as well as kinetic and calcium sensitivity differences in the BK channels underlie the tonotopic distribution of resonant properties (Art & Fettiplace, 1987; Fuchs et al. 1988; Hudspeth & Lewis, 1988b; Fuchs & Sokolowski, 1990; Art et al. 1995; Wu et al. 1995). Whether similar variations in kinetics or steady-state properties of the hair cell calcium channel occur tonotopically is unknown, and is one of the questions addressed by this work.

Synaptic transmission is driven by calcium entering hair cells through calcium channels. Calcium channels are clustered, presumably at synaptic release sites (Roberts et al. 1990; Issa & Hudspeth, 1994; Tucker & Fettiplace, 1995). The number of calcium channels and the number of release sites, but not the density of channels, increases with characteristic frequency (Sneary, 1988; Wu et al. 1996; Ricci et al. 2000). Whether calcium channels linked to neurotransmitter release are different from those linked to electrical resonance remains to be elucidated.

Calcium channels have been classified biophysically, pharmacologically and molecularly (see Hille, 2001 for review). L-type calcium channels typically activate at depolarized potentials, are sensitive to dihydropyridines and show slow inactivation (Tsien et al. 1988). Calcium channels are multimeric, containing α, β, α2δ and sometimes γ subunits. The α subunits make up the pore-forming region and are mandatory for channel functioning. L-type calcium channels have two main subtypes based on α subunits, the α1C and the α1D. The first identified and characterized L-type channel was the α1C type, which is found largely in skeletal muscle and heart, while the α1D is found in neuronal cells and some epithelial cells. The α1D channels have several unusual properties including a hyperpolarized activation curve, fast (submillisecond) activation rise times and an insensitivity to the L-type dihydropyridine antagonists (Koschak et al. 2001). Recently the α1D channels have been linked to synaptic release proteins and are thought to regulate some forms of synaptic transmission (Yang et al. 1999). In addition, these accessory proteins can modulate channel electrical properties (Yang et al. 1999).

The α1D channel type has been identified in the chick auditory papilla (Kollmar et al. 1997a, b), frog saccule (Rodriguez-Contreras & Yamoah, 2001), trout saccule and mammalian cochlea (Zhang et al. 1999; Platzer et al. 2000; Koschak et al. 2001). Hair cell calcium channels are somewhat different from the α1D channels that are expressed heterologously in vitro, in particular with regard to inactivation. Another purpose of the present work was to compare the properties of hair cell calcium channels to those reported for expressed α1D channels.

Several different types of calcium channel have been identified in hair cells. L-type channels have been identified in a variety of hair cell organs including the frog saccule (Hudspeth & Lewis, 1988a; Roberts et al. 1990), turtle papilla (Art et al. 1986; Art & Fettiplace, 1987), chick papilla (Fuchs et al. 1990; Zidanic & Fuchs, 1995; Spassova et al. 2001), guinea-pig cochlear hair cells (Bobbin et al. 1990; Nakagawa et al. 1991; Oshima et al. 1996; Zhang et al. 1999) and the frog semicircular canal (Prigioni et al. 1992; Martini et al. 2000). N-, R- and T-type channels have also been described in vestibular hair cells (Rennie & Ashmore, 1991; Martini et al. 2000; Rispoli et al. 2000). In particular, N-type channels have been identified recently in frog saccule hair cells, an end-organ traditionally thought to have only L-type channels (Su et al. 1995; Rodriguez-Contreras & Yamoah, 2001). R-type currents have been identified in frog semicircular canal hair cells (Martini et al. 2000; Rispoli et al. 2000). Whether different channel types are responsible for different aspects of signal processing in turtle auditory hair cells is unknown, and is also a focus of the present work.

The data presented here will demonstrate that only L-type channels are present in turtle auditory hair cells. Tonotopic differences in the activation properties of L-type channels are also described. Activation properties and pharmacological sensitivities support the hypothesis that the calcium channels are of the α1D variety. An unexpected novel finding was the identification of calcium-dependent inactivation.


Tissue preparation

Auditory papillae were prepared as described previously (Crawford & Fettiplace, 1985; Ricci & Fettiplace, 1997, 1998). Red-eared sliders (Trachemys scripta elegans), carapace length 8–13 cm (3–5 inches) were decapitated and the inner ear organs removed using procedures approved by the Animal Care Use Committee at LSU Health Sciences Center and by standards established by NIH guidelines. The inner ear organs were placed into external solution containing (mm): 125 NaCl, 0.5 KCl, 2.8 CaCl2, 2.2 MgCl2, 2 each of pyruvate, creatine, lactate and ascorbate, 6 glucose and 10 Hepes. The solution was buffered to pH 7.6 and had a final osmolality of 275 mosmol kg−1. The low-potassium solution was used to maintain the cells in the papilla in a hyperpolarized state, thus lengthening the viability of the tissue. The tissue was pinned to the bottom of a Sylgard-coated dish with minutien pins with the auditory papilla facing upward. The external membrane was removed, exposing the tectorial membrane. Protease type XXIV (Sigma) 0.02–0.04 mg ml−1 was added to the solution and the tissue was incubated for 5–20 min, depending on enzyme potency. The tectorial membrane was removed and the enzyme washed out with multiple rinses of external solution. The papilla was trimmed and placed into a recording chamber with a coverslip at its base. The tissue was held in place with three single strands of dental floss. The recording chamber was perfused at a rate of 0.5–1 ml min−1 with external solution supplemented with 100 nm apamin (Calbiochem) to block the caesium-permeable SK calcium-activated potassium current (for example see Fig. 1B; Tucker & Fettiplace, 1996). A peristaltic pump (Gilson) was used for the bath perfusion. In recordings from low-frequency cells, 100 μm 4-aminopyridine (4-AP) was included to block any delayed-rectifier conductance (Goodman & Art, 1996).

Figure 1.

Recording procedures and calcium current run-up

A, differential interference contrast image of the recording electrode and papilla after clearing that ensured rapid and reproducible solution exchange to cells. B, to estimate drug delivery time, a hair cell was voltage clamped at −84 mV and depolarized to −14 mV for durations that allowed for calcium accumulation to activate the caesium-permeant SK calcium-activated potassium conductance (Tucker & Fettiplace, 1996). Bath application of apamin at 100 nm completely blocked the SK current in about 4 min. C, upon establishment of the whole-cell configuration, calcium currents increased in amplitude, often doubling or tripling in size. D, there was no shift in the plot of peak current against command potential (IV) for different times after rupture of the patch. C and D are from the same cell. E, a plot of the normalized current against time after rupture was best fitted by a single exponential function with a time constant of 4.3 ± 0.3 min (n= 10). After reaching a maximal current, recordings were stable for more than 1 h, showing a less than 10 % reduction in peak current during this time.

Recording procedures

A large blunt pipette was advanced into the papilla from the abneural edge while applying pressure to the back end of the pipette, making a hole from which 1–3 cells could be removed to ensure good access (Fig. 1A). The space around the cell being recorded was carefully cleared before the cell was patched. Whole-cell recordings were obtained as has been described previously (Ricci & Fettiplace, 1997). An EPC8 (Heka) or an Axopatch 1D (Axon Instruments) was used for all recordings. The EPC8 was used for current-clamp measurements. The internal solution contained (mm): 110 CsCl, 3 MgATP, 5 creatine phosphate, 1 BAPTA, 10 Hepes and 2 ascorbate; the pH was 7.2. For current-clamp experiments, potassium was replaced by caesium (Ricci et al. 2000). Where higher BAPTA concentrations were used, CsCl was reduced to maintain a constant osmolality. Caesium was used to block the BK potassium conductance, thus unmasking the calcium current (Art et al. 1993).

Series resistances ranged from 1 to 5 MΩ after up to 70 % compensation. Cell capacitance was 11.6 ± 0.2 pF (n= 75), giving voltage-clamp speeds of ⩽ 60 μs. A junction potential of −4 mV was measured and corrected off-line, as was any residual series resistance. Cells with leak currents greater than 50 pA were excluded as they provided an uncontrolled source of calcium that tended to produce run-down of the currents and a loss of inactivation. Minimal leak subtraction was performed with the amplifier circuit. Cells where series resistance varied by more than 25 % during the recording were also excluded. Cell capacitance tended to increase throughout the recording, presumably due to the interruption of vesicle recycling when the whole-cell configuration was obtained. Mechanoelectric transducer currents were eliminated by spraying internal solution containing BAPTA onto the hair bundles before recording (Assad et al. 1991; Crawford et al. 1991; Marquis & Hudspeth, 1997; and authors' unpublished observations). The location of the cell, recorded as distance from the apical end where the lagena is located, was noted at the end of the recording. All experiments were performed at a temperature of 19–22 °C.

Upon reaching the whole-cell configuration, calcium currents increased in amplitude for the first 10–15 min (Fig. 1CE). No change in the voltage dependence of the current-voltage (IV) plots was observed during the run-up period. Due to run-up, data were not collected until the peak current had stabilized, about 15 min after break-through. Since run-up was an uncontrolled variable in these experiments, high-frequency cells were required to run-up to at least 600 pA maximal current and low-frequency cells were required to reach at least 250 pA. More than 90 % of the cells met these criteria.

Drug application

All drugs were bath applied. The application was either automated using a Gilson peristaltic pump coupled through miniature solenoid switches (Lee Valves), or else drugs were hand pipetted. There was difficulty with precipitation when using the pump for dihydropyridine and barium application, so hand pipetting was used for these and the bath perfusion turned off while pipetting was carried out. Nimodipine dissolved in DMSO comes out of solution within 30 min of preparation, initially giving quite variable results. To circumvent this problem, nimodipine was prepared after the whole-cell recording had been obtained and control currents were recorded so that there was no more than 5 min between solution preparation and application. Data where the level of precipitation was uncontrolled were not included. Barium exchange also gave some technical difficulties. When in contact with normal external solution, a precipitate would form. In addition, exchange appeared to be very slow at the preparation, as if calcium was being released from the tissue. To circumvent these problems, the tissue was washed first in a low-calcium (50 μm) solution before barium was applied. No precipitation was observed and results were quite consistent. At least a 10-fold volume substitution of bath was performed for any drug application. Control experiments showed no effect of either bath application of the normal solution or turning the bath flow off for necessary time periods.

To determine the time it took for a drug to reach the basolateral aspects of the hair cells, experiments were carried out with apamin absent from the external solution. A depolarizing step of sufficient duration to activate the SK calcium-activated potassium current was applied (Fig. 1B). Apamin was then perfused onto the preparation and time until SK block was recorded. Apamin is a large molecular weight peptide (MW 2027) and so should diffuse slowly. On average, 4 min allowed for complete block of the SK current, thereby revealing the calcium current. In general, drugs were applied for at least 10 min before measuring a response and recordings were made over the next several minutes to ensure a steady-state response. Calcium channel blockers ω-agatoxin IVA, calcispetine, ω-conotoxin and MVII, SNX-482 were purchased from Alomone Laboratories. Chemicals were purchased from Fisher or Sigma. Cytochalasin D, colchicine, nimodipine, verapamil, apamin, D600, Bay K 8644 (mixed isomers), diltiazem, ω-conotoxin GVIA and DIDS were purchased from Calbiochem, and cyclodextrin from Sigma. Where necessary, drugs were dissolved in DMSO, final concentrations being < 0.01 %. No effects of DMSO alone were observed.

Data analysis

All data are presented as means ±s.e.m. The number of cells (n) is given with each set of data. Unless otherwise stated, current traces illustrated are averages of four for activation protocols and single traces for inactivation protocols. Data were collected with Signal software (CED) and exported to Origin (Microcal) for analysis. Origin uses the Levenberg-Marquardt algorithm for fitting. Where appropriate, correlation coefficients are given as r2 values. Unless stated otherwise, Student's two-tailed t test was used to assess statistical significance, with P < 0.01 indicating a statistically significant difference.

Typically, tail currents were too fast or were obscured by capacitative artefacts to generate proper activation curves. In addition, the tails could also be contaminated by remaining mechanoelectric transducer currents. In lieu of this, IV curves were plotted as peak current for a given command voltage against the command potential. The curves were fitted with a Boltzmann function of the form:

display math(1)

where, Imax is the maximal current elicited, V1/2 is the voltage of half-activation and δx reflects the steepness of the plots. The same equation was used to fit normalized inactivation plots. where, V1/2 represents the voltage of half-inactivation and δx is the slope of the function, Theoretically, Boltzmann curves are only accurate if changes in driving force are accounted for, hence the use of isopotential tail currents or dividing maximal current by the reversal potential. Neither was done here because tail currents could not be isolated and the reversal potentials could not be clearly determined. Reversal potentials were typically contaminated by remaining transducer current or some unblocked residual SK or BK at these depolarized potentials. The error associated with the method used here will underestimate the peak response for the larger depolarizations, and the slope of the curve will be shallower; however these errors will be consistent between frequency positions. The purpose of these plots is to compare hair cell responses from cells at different frequency positions and also for characterization of the channel types present. Neither of these goals is compromised by the methodology used.

Activation rise times were measured as the time taken for the current to activate from 10 to 90 % of maximum current. Although a less formal estimate, this measurement does not require assumptions about gating schemes or inactivation, yet it does allow for a simple comparison between currents obtained at different papilla positions or under different pharmacological conditions.

Dose-response curves were fitted with a Hill equation:

display math(2)

where k is the half-blocking dose (IC50), nh is the Hill coefficient and Bmax is the maximal block obtained.

Unless otherwise indicated, inactivation time constants were obtained by fitting triple exponential equations to the decay in current during a 1 s depolarization to −14 mV. The equation was of the form:

display math(3)

where, I0 is the current at 1 s, τ1, τ2 and τ3 are the time constants fast to slow and A1, A2 and A3 are the proportionality constants.

The inactivation index is the ratio of peak to steady-state current at a given potential.


Tonotopic variations in calcium current

Hair cells were patch clamped from a high- (0.64 ± 0.01) and a low-frequency (0.34 ± 0.01) position. Using a potassium-based intracellular solution, current-clamp measurements of electrical resonance and resting potentials were obtained (Fig. 2A). High-frequency cells had a resonant frequency of 317 ± 27 Hz and a resting potential of −46 ± 1 mV (n= 11), while low-frequency cells had a resonant frequency of 115 ± 6 Hz and a resting potential of −51 ± 0.5 mV (n= 9; Fig. 2A). Both measurements were statistically significantly different (P < 0.01). Examples of calcium currents obtained using a caesium-based intracellular solution are given in Fig. 2B. Currents activated at potentials negative to −40 mV and peaked at potentials negative to −10 mV. Complex inactivation was observed in most cells. As reported previously, high-frequency cells had larger maximal currents than low-frequency cells (0.75 ± 0.02 nA (n= 75) as compared to 0.36 ± 0.04 nA (n= 10; Art & Fettiplace, 1987; Art et al. 1993; Ricci et al. 2000). Currents activated rapidly, with rise times that decreased exponentially with depolarization (Fig. 2C). Rise times were fast for currents obtained at both papilla locations, (< 0.2 ms for depolarizations to −14 mV), a property found commonly for L-type channels of the α1D variety. The rise times measured from low-frequency cells were slightly faster at hyperpolarized potentials (Fig. 2C). Fits to a single exponential equation demonstrated a difference in the voltage dependence of the rise times, with values of 102 ± 7 and 153 ± 14 mV for high- and low-frequency cells, respectively. The rise times measured were comparable to those reported previously for auditory hair cells (Art & Fettiplace, 1987; Zidanic & Fuchs, 1995). Our plots did not show the bell-shaped response predicted by Hodgkin-Huxley-type kinetics, but this may simply be due to our inability to measure rise times from currents at potentials more negative than −45 mV, where previous work has suggested that the kinetics are faster (Art & Fettiplace, 1987; Zidanic & Fuchs, 1995). Inactivation was observed in hair cells from both locations and will be characterized in the section entitled, ‘Inactivation’. Plots of peak current against command potential (IV) suggest that currents in low-frequency cells activate at a more hyperpolarized level than do high-frequency cells (Fig. 2D). The IV curves were smooth, with no apparent humps which might suggest contributions from different channel types. To better quantify the difference between frequency locations, IV plots were normalized to peak current, replotted and fitted with Boltzmann functions (eqn (1); Fig. 2E). Low-frequency cells activated at more hyperpolarized levels, with half-activating voltages (V1/2) of −43 ± 1 mV compared to −35 ± 1 mV for high-frequency cells. Here too, single Boltzmann functions provided good fits to the data, suggesting that multiple channel types are not involved. Slopes of 4.7 ± 0.3 and 4.2 ± 0.2 mV−1 were measured for high- and low-frequency cells, respectively, and were not significantly different.

Figure 2.

Tonotopic variations in hair cell properties

A, examples of electrical resonance obtained from hair cells at the high (left) and low (right) frequency position in response to 100 pA current step injections using a potassium-based intracellular solution. B, examples of calcium currents obtained from a high- (left) and low-frequency (right) cell with the corresponding stimulus protocol shown above, using caesium-based electrodes. Larger currents were obtained from high-frequency cells. C, a plot of current rise time (10–90 %) against command potential for low- (▵, n= 11) and high-frequency (□, n= 12) cells. Single exponential fits of the form: Y=Y0+Aexp(−x/τ) were obtained. Values of 260 ± 10, 20 ± 5 and 102 ± 7 μs were obtained for Y0, A and τ, respectively (r2= 0.94) for high-frequency cells and 260 ± 1, 53 ± 10 and 153 ± 14 μs (r2= 0.99) for low-frequency cells. * Points that are significantly different. D, IV curves for 10 high-frequency (□) and 10 low-frequency cells (▵). E, normalized plots from the data shown in D, symbols the same, with Boltzmann fits to the data. Half-activation voltages (V1/2) were −35 ± 1 and −43 ± 1 mV and slopes were 4.7 ± 0.3 and 4.2 ± 0.2 mV−1 for high- and low-frequency cells, respectively (r2= 0.99 for each).

Pharmacological characterization of channels

A fundamental question addressed by the present work is: are there multiple types of calcium channels in turtle auditory hair cells? A first step towards answering this question was to ensure that all the measured current was carried by calcium. High concentrations of the divalent cations nickel and cadmium blocked the inward current. Both nickel and cadmium were effective at blocking the current, but at doses that were not selective between channel types (Table 1). Nickel had an IC50 of 0.74 ± 0.2 mm, while cadmium at 1 mm blocked 94 ± 3 % of the current (n= 3). IV plots were smooth and showed little change other than the decrease in amplitude in the presence of blockers, suggesting that additional currents were not unmasked or selectively blocked by these ions (data not shown). The chloride channel blocker DIDS was also used to ensure that all the inward current was carried by calcium. No effect of DIDS was observed.

Table 1. Summary of the pharmacological data of hair cell calcium currents
CompoundBlocker typeConcentration n % blockIC50Hill coefficient
  1. With the exception of calciseptine and taicatoxin, L-type blockers were effective at antagonizing the current. No other class of calcium channel antagonists was effective at blocking the current. NS, non-selective; all other abbreviations have their standard meaning. Where dose–response curves were obtained, IC50 and Hill coefficient values are included as well as the highest dose with its corresponding blocking effecacy.

NiCl2NS3 mM788 ± 2850 ± 180 μM1.5 ± 0.4
CdCl2NS1 mM394 ± 3  
Nimodipine (Vh−80mV)L30 μM568 ± 22.2 ± 0.2 μM2.1 ± 0.4
Nimodipine (Vh−60mVL30 μM589 ± 21.8 ± 0.2 μM1.7 ± 0.2
VerapamilL1 mM394 ± 3192 ± 20 μM1.7 ± 0.4
D600L3 mM398 ± 15367 ± 22 μM1.6 ± 0.2
DiltiazemL1 mM791 ± 3375 ± 20 μM2.5 ± 0.3
CalsiceptineL1 μM6−6 ± 2  
ω-Conotoxin GVIAN1 μM83 ± 4  
ω-AgatoxinP/Q0.3 μM5−11 ± 2  
ω-Conotoxin MVIIAN1 μM30  
TaicatoxinL0.2 μM30-  
SNX-482R0.1 μM30  
DIDS  30  


L-type calcium channels are selectively blocked by the dihydropyridine class of compounds. Nimodipine, a relatively soluble form of dihydropyridine, was tested on the hair cell current (Fig. 3). Nimodipine-induced block of the hair cell-calcium current was weak, requiring concentrations as high as 30 μm to substantially antagonize the current (Fig. 3E and F). Two possible conclusions can be drawn, one that multiple channel types exist (i.e. channels sensitive or insensitive to dihydropyridines) or second, that the channel type present was only partially antagonized by these compounds (i.e. the α1D L-type channel). Before attempting to delineate between these possibilities, a better understanding of the dihydropyridine effect was needed. Dihydropyridine block can be both voltage and use dependent. Initial experiments used a holding potential of −84 mV, where the maximal block was 68 ± 2 % (n= 5). Changing the holding potential to −64 mV gave a maximal block of 90 ± 2 % (n= 5), suggesting that the majority of current could be antagonized by dihydropyridines (Fig. 3A, B and E). IV curves in the presence, absence and difference (i.e. current sensitive to nimodipine block; Fig. 3B) were normalized to their respective maximal current, allowing direct comparisons of their activation properties (Fig. 3C and D). Boltzmann fits (eqn (1)) to these data showed no difference between plots at either holding potential, suggesting that only one channel type was present. In addition, rise times in the absence or presence of nimodipine were not different (data not shown); similar results from chick papilla hair cells argued for the existence of one channel type (Zidanic & Fuchs, 1995).

Figure 3.

Nimodipine blocks the calcium current in a voltage-dependent manner

A, calcium currents elicited by depolarizing the cell to −14 mV from holding potentials of −84 mV or −64 mV in the presence and absence of 5 μm nimodipine for a high-frequency cell. No difference in the control response is seen, but nimodipine block was greater at a −64 mV holding potential. B, IV curves generated from cells using holding potentials of −84 or −64 mV in the absence (−84 mV, □; −64 mV, ○) and presence (−84 mV, ▪; −64 mV, •) of 5 μm nimodipine. C and D, normalized IV curves for data obtained in the absence (□) and presence (▵) of 5 μm nimodipine and the difference (nimodipine-sensitive, •), from a holding potential of −84 mV (C) and −64 mV (D). There was a slight shift in the V1/2 from −40 ± 0.3 to −37 ± 0.3 mV for the difference plot at −84 mV, but no differences in slope (5.2 ± 0.3 mV−1). No difference in IV plots was observed for the holding potential of −64 mV, where the V1/2 was −39 ± 1 mV and the slope 4.2 ± 0.8 mV−1. E, dose-response curves with their corresponding fits to the Hill equation for holding potentials of −84 mV (□) and −64 mV (▵). The number of cells tested is given in parenthesis as (−84 mV, −64 mV, respectively). Values for Bmax, the half-blocking dose and the Hill coefficient were 69 ± 4 %, 2.2 ± 0.2 μm and 2.1 ± 0.4, respectively, for a holding potential of −84 mV and 90 ± 4 %, 1.8 ± 0.2 μm and 1.7 ± 0.2, respectively, for the holding potential of −64 mV (r2 values of 0.989 and 0.99, respectively, were obtained for the two conditions). F, plots normalized to the maximal blocking dose for holding potentials of −84 mV (□) and −64 mV (▵) with the corresponding fit to the Hill equation. Here, values of 1.0, 2.0 ± 0.1 μm and 1.9 ± 0.2 were obtained for Bmax, half-blocking dose and the Hill coefficient, respectively (r2= 0.99). G, the voltage dependence of nimodipine block was investigated further by plotting the ratio of current obtained in the presence of 5 μm nimodipine to control current at different test potentials against the test potential. A single exponential (Y=Y0+Aexp(−xx)) best fitted this plot with values of 0.3 ± 0.01, 0.003 ± 0.001 and 8.8 ± 1 mV for Y0, A and δx, respectively (r2= 0.98). HJ, use dependence of nimodipine block was investigated by depolarizing a cell to −14 mV for 20 ms at a frequency of 1 Hz. Examples of the first and last response from a holding potential of −84 mV (H) or −64 mV (J) are given with the thicker lines representing the first and the thin lines the last. Time zero represents the start of the 1 Hz stimulus protocol that begins 10 min after drug application. Traces at the top are in the absence and the bottom in the presence of 5 μm nimodipine (Nim.). I, plot of the peak current against time where the symbols correspond to their respective condition (H, J). The continuous lines reflect the initial value. From a holding potential of −84 mV a further decrease of 9 % was observed while from a holding potential of −64 mV a further reduction of 32 % was observed for this cell.

Dose-response curves obtained from holding potentials of −84 and −64 mV are plotted in Fig. 3E, normalized to maximal control current. Fits to the Hill equation (eqn (2)) gave IC50 values of 2.2 ± 0.2 and 1.8 ± 0.2 μm and a Hill coefficient of 2.1 ± 0.4 and 1.7 ± 0.2 for holding potentials of −84 and −64 mV, respectively. These values were not statistically different. To further demonstrate the similarity in block at different holding potentials, dose–response curves were normalized to maximal block (Fig. 3F), resulting in overlapping plots where neither the IC50 nor the Hill coefficient were different. Therefore, the major difference between holding potentials appears to be in the Bmax, perhaps suggesting that a particular channel state is required for drug binding and that this state is favoured at the −64 mV holding potential.

α1D channels expressed in oocytes have a u-shaped voltage dependence to block by nimodipine (Xu & Lipscombe, 2001). That is, plotting the percentage of current left in the presence of nimodipine against the command potential revealed a region where the drug was most sensitive (−4 to 0 mV), but at either hyperpolarized or depolarized potentials the block was less. Figure 3G reproduces this plot for hair cell channels, demonstrating that no u-shape was observed, rather a simple exponential relationship was found, perhaps suggesting that differences exist between expressed channels and the native hair cell channel.

Recent evidence from frog saccule hair cells suggests a strong use dependence to nimodipine block similar to that observed for α1C channels (Rodriguez-Contreras & Yamoah, 2001). To investigate use dependence, a stimulus protocol was used that stepped the membrane potential to −14 mV for 20 ms from a holding potential of either −84 or −64 mV at 1 Hz in the absence and presence of 5 μm nimodipine (Fig. 3HJ). The frequency of stimulus was chosen so as not to cause significant inactivation of the current and to match previous experiments (Rodriguez-Contreras & Yamoah, 2001). Comparing the first to the last cycle of depolarization in the presence of 5 μm nimodipine gave a decrease of 4 ± 1 and 10 ± 2 % for holding potentials of −84 and −64 mV, respectively (n= 5). These data indicate that turtle auditory hair cells show much less use dependence of block by nimodipine than saccule hair cells.

The nimodipine data presented are from high-frequency cells. Low-frequency cells were also investigated and showed the same sensitivities to nimodipine as high-frequency cells. A similar dependence on holding potential was found, whereby 5 μm nimodipine blocked 64 ± 3 % of the current at −84 mV, while 88 ± 3 % (n= 5) of the current was blocked at −64 mV, values that are not statistically different from measurements recorded from high-frequency cells.

Bay K 8644 is a dihydropyridine that acts as an agonist for L-type channels by prolonging the open time of the channel. An example of Bay K 8644 action on hair cell calcium currents is shown in Fig. 4A, where the current magnitude was increased, activation kinetics were slowed and the IV relationship shifted in a more hyperpolarized direction. A 46 ± 11 % (n= 4) increase in peak current amplitude was observed. Fitting the normalized IV curves for control and Bay K 8644 with Boltzmann functions demonstrated a shift in V1/2 from −37 ± 0.3 to −51 ± 2 mV and a change in slope from 3.5 ± 0.2 to 4.9 ± 0.2 mV−1 (n= 4), both values being statistically significant (Fig. 4C). Single Boltzmann fits to the IV curves also support the hypothesis that only one class of channel was present in turtle auditory hair cells. Since Bay K 8644 is predicted to shift the activation of the L-type currents, leaving non-L-type currents unaffected, a single Boltzmann function would not be sufficient to fit the resulting complex IV curve. Activation kinetics were slowed by Bay K 8644 (Fig. 4A and D). Plotting rise time against potential demonstrates that activation kinetics at potentials hyperpolarized to −30 mV were slower in the presence of Bay K 8644 (Fig. 4D). Bay K 8644 affects both α1D and α1C L-type calcium channels, but not other classes of calcium channel, so the effects reported suggest the channels present in hair cells were of the L-type.

Figure 4.

Bay K 8644, a dihydropyridine agonist, increases hair cell calcium currents

A, an example of a calcium current obtained in the absence (upper) and presence (lower) of the dihydropyridine agonist Bay K 8644. The stimulus protocol is shown at the top; more negative steps refer to IV plots in the presence of Bay K 8644. B, IV curves demonstrating that the current magnitude increased and voltage sensitivity shifted in a hyperpolarized direction. In each panel, □s are the control and ▵s are in the presence of 1 μm Bay K 8644. C, IV plots normalized to maximum current, demonstrating the shift in the V1/2. Continuous lines represent fits to a Boltzmann function, where the V1/2 values were −37 ± 0.3 and −51 ± 2 mV and the slopes were 4.9 ± 0.2 and 3.5 ± 0.2 mV−1 for control and Bay K 8644-treated cells, respectively (r2= 0.99 for both fits, n= 4). D, activation kinetics were slowed as demonstrated by plotting the 10–90 % rise time of the current activation in the absence and presence of 1 μm Bay K 8644. Coninuous lines represent exponential fits with voltage dependence, δx, of 153 ± 26 and 93 ± 8 mV for control and Bay K 8644-treated cells, respectively (r2 values were 0.99 and 0.96 for control and Bay K 8644-treated (n= 4) cells, respectively).

Non-dihydropyridine L-type blockers

The dihydropyridine data support the hypothesis that only L-type channels, probably of the α1D variety, are present in the turtle auditory papilla. As nimodipine block is incomplete and high doses were needed for the block, the data do not demonstrate conclusively that α1D channels are the only type present. For this reason, other L-type antagonists were used.

Verapamil and D600 are acetonitrile derivatives that were some of the first calcium channel antagonists available (Lee & Tsien, 1983). They are still thought to be relatively selective for L-type channels over other types of calcium channel (Sperelakis, 1987; Striessnig et al. 1998). Dose-response curves with their corresponding fits to Hill equations are given in Fig. 5. Both verapamil and D600 completely antagonized the calcium currents with IC50 values of 192 ± 20 and 375 ± 25 μm and Hill coefficients of 1.7 ± 0.4 and 1.6 ± 0.2, respectively (Fig. 5A and B). Diltiazem, a benzothiazepine derivative also known to be relatively selective for L-type calcium channels blocks the hair cell current (Sperelakis, 1987; Striessnig et al. 1998). Figure 5C shows the dose-response curve and corresponding fit to the Hill equation for diltiazem. An IC50 of 367 ± 22 μm and a Hill coefficient of 2.5 ± 0.3 were obtained. The diltiazem result was somewhat unusual in that low doses consistently potentiated the current amplitude, an effect not reported previously. A fourth compound, calciseptine (1 μm), a peptide toxin of L-type calcium channels, was ineffective at antagonizing the current (de Weille et al. 1991). Calciseptine is thought to act at, or near the dihydropyridine binding site, a site that, based on drug potency, may be different between α1C and α1D subunits (Yasuda et al. 1993). The lack of effect of calcispetine might also be a reflection of the difference between α1C and α1D subunits at the same binding site. Taicatoxin, another L-type blocker was ineffective at antagonizing the hair cell current (Hamilton & Perez, 1987). Taicatoxin also competes for the dihydropyridine-sensitive site and so it is not surprising that this compound was ineffective at blocking the hair cell current (Hamilton & Perez, 1987).

Figure 5.

Traditional L-type channel blockers can completely antagonize the calcium current

Dose–response curves, with corresponding fits to Hill equations, are shown for verapamil (A), D600 (B) and diltiazem (C). Insets are currents elicited from depolarizations to −14 mV from a holding potential of −84 mV in the absence and presence of the three highest drug concentrations (averages of four). The control trace is the largest current in A and B, but in the diltiazem trace it appears that the low dose potentiates the response. The scale bar is 500 pA and 20 ms. The value of n for each dose is given in parenthesis. Half-blocking doses obtained from fitting Hill equations were 192 ± 20, 375 ± 25 and 367 ± 22 μm for verapamil, D600 and diltiazem, respectively. Hill coefficients were 1.7 ± 0.4, 1.6 ± 0.2 and 2.5 ± 0.3 for verapamil, D600 and diltiazem, respectively, with r2 values of 0.997, 0.998 and 0.99, respectively.

Non-L-type antagonists

The pharmacological data suggest that most, if not all the calcium current can be attributed to L-type calcium channels, most probably of the α1D variety (Table 1). However, the high concentrations required for complete channel antagonism, in addition to reports of other calcium channel types present in hair cells, prompted the use of additional pharmacological tools to assess directly the presence of other calcium channel types. ω-Conotoxin GVIA is a potent N-type channel antagonist that does not interact with dihydropyridine or verapamil binding sites on L-type channels. At 1 μm there was no effect of ω-conotoxin GVIA (Table 1 for summary). ω-Agatoxin IVA is a potent blocker of P/Q-type channels that does not interact with L-type channels (Safa et al. 2001). At 0.3 μm, agatoxin was also ineffective at antagonizing the current. Additional blockers used are listed in Table 1, including ω-conotoxin MV11A and SNX-482, a purported R-type antagonist (Bourinet et al. 2001). None was effective at antagonizing the current.

Both the biophysical and pharmacological data suggest that only L-type calcium channels are present in turtle auditory hair cells. Data supporting this hypothesis include: single Boltzmann functions fit IV plots in the absence and presence of antagonists; only L-type antagonists were effective at blocking the current and all current could be blocked by L-type channel blockers. In particular, the α1D type of L-channel is implicated based on: rapid (< 0.5 ms) activation rise times; hyperpolarized IV curves and relative insensitivity to dihydropyridines. All the L-type antagonists used have binding sites on the α subunit (Striessnig et al. 1998) and reveal nothing about accessory subunits. The next section addressing inactivation is analysed with the assumption that one channel type is present.


Time-dependent inactivation was observed in currents recorded from both high- and low-frequency cells (see examples in Figs 1 and 2). Since inactivation has not been characterized in auditory hair cells previous to this work, it was important to ensure that the measured decay in current amplitude was inactivation and not an artefact of contamination by some other current. TEA was included both within the pipette and externally at 25 mm to block any residual potassium currents. No effect on inactivation was observed. 4-AP was also used externally with no effect on inactivation. As will be described, external barium, which should block potassium currents and should not substitute for calcium in activating BK channels, did not remove inactivation. Apamin (100 nm), which was used to block the SK potassium conductance, reduced the current ascribed to the SK without altering the time-dependent component of inactivation (Fig. 1). DIDS, a voltage-dependent chloride-channel blocker was also ineffective in altering inactivation. Bay K 8644 slowed inactivation, suggesting that inactivation was a property of the calcium channels. In addition, inactivation was slowed by chemicals that antagonized the channel, such as verapamil and diltiazem, suggesting that inactivation is a function of calcium entry and is dependent on the calcium channel. Together, these data suggest that inactivation is a function of the calcium channels and was not an artefact of contamination by some other current.

Inactivation could often be labile, cells with leak current (> 30 pA) or currents in cells that were depolarized for extended periods of time could lose inactivation, despite maintaining a reasonably large current amplitude. The most relevant factor in maintaining inactivation was the interpulse interval (IPI). Initially, protocols were run that used brief, 50 ms IPIs. These cells would not show time-dependent inactivation, in spite of having stable, relatively large currents. To better quantify the importance of IPIs, a protocol was developed that used a holding potential of −84 mV and depolarized for 20 ms to −14 mV with varying times between pulses (Fig. 6A). A short IPI yielded currents with no time-dependent inactivation (Fig. 6A, lower panel). That is, there was no decay in current during the depolarizing pulse. The current amplitude was reduced, suggesting that channels had not recovered from the inactivation induced by the prior stimulus. A plot of normalized current against IPI (Fig. 6B) reveals an exponential relationship with a time constant of 174 ± 36 ms (n= 21), indicating that protocols using 20 ms depolarizations required an IPI of 450 ms to maintain a constant-amplitude current. The plot demonstrates that with a short IPI the current remains inactivated and thus does not show any time-dependent inactivation. With short pulses, inactivation does not appear to be very robust; however, as shown in Fig. 6C and D, inactivation could be quite profound if long pulses were used. These data also demonstrate that all the calcium channels present could be inactivated.

Figure 6.

The interpulse interval is a critical determinant of inactivation

A, a protocol that held the cell at −84 mV and delivered a 20 ms depolarization to −14 mV at time 0 and at varying times (P2) following the initial pulse, was used to determine the required interpulse interval needed to preserve inactivation. The upper panel shows the stimulus protocol, the middle traces are currents elicited and the lower trace is an expansion of the first two current responses used to illustrate the loss of a time-dependent component of inactivation (no averaging was used in these protocols). The continuous line is shown to point out the loss of both peak current and time-dependent inactivation. B, a plot of the peak current normalized to the control at time 0 against interpulse duration (P2) illustrates the time to recovery from inactivation and has an exponential relationship with a time constant of 174 ± 36 ms (n= 21). C, the response to a 1 s depolarization to −14 mV with corresponding single (black), double (light grey) and triple (white, dashed) exponential fit. D, that all channels can inactivate is illustrated by depolarizing a cell to −14 mV for 25 s. Both C and D are single-episode responses.

The kinetics of inactivation were quite complex, reductions in current in response to depolarizations of 1 s required three time constants to reflect accurately the time course of current decay. Figure 6C illustrates an example of an inactivating current with single, double and triple exponential fits. Only the fit to three time constants could adequately describe the data. No difference in time constants of inactivation was observed between high- and low-frequency cells. Values of 6 ± 1 and 6 ± 1 ms for τ1, 73 ± 5 ms and 77 ± 14 for τ2 and 903 ± 132 and 1179 ± 122 ms for τ3 were obtained for high- and low-frequency cells, respectively (n= 12 and 6, respectively). The only difference observed between frequency positions was in the proportion of inactivation contributed by the fast component, where low-frequency cells exhibited more fast inactivation than high-frequency cells (55 ± 7 vs. 41 ± 3 %). The total amount of inactivation observed between cells of different frequency locations was comparable at 37 ± 2 and 42 ± 4 % for high- and low-frequency positions, respectively, for depolarizations to −14 mV.

Often cells that were depolarized for 1 s would require tens of minutes to recover. Protocols were developed to allow cells to recover sufficiently to do multiple experiments on a given cell. Inactivation was studied using a prepulse protocol that varied the membrane potential between −114 and +114 mV in 10 mV increments, followed by a test pulse to −14 mV from a holding potential of −84 mV. A 3 s IPI was used. Initial experiments varied the prepulse duration which, simply based on the measured time constants, were required to be of several seconds duration. However, since long prepulses did not allow for cells to adequately recover, shorter duration prepulses were tested. These shorter pulses would favour inactivation driven by the faster time constants. A comparison of data obtained using different prepulse durations is given in Fig. 7AC. An example of the data obtained is given in Fig. 7A for a 20 and a 200 ms prepulse. Plotting the peak current at the test potential of −14 mV against the prepulse potential gave bell-shaped curves (Fig. 7B). This bell shape typically implies calcium-dependent inactivation (Brehm & Eckert, 1978). More inactivation was observed with the 200 ms prepulse. Half-inactivating voltages were obtained by fitting Boltzmann functions to the first portion of the plot in Fig. 7B normalized to the maximal current. No shift in half-inactivating voltage was observed (Fig. 7C), while the fraction of current inactivating increased exponentially as the prepulse duration was increased (Fig. 7C). As can be seen in the example shown in Fig. 8B, 20 ms protocols often resulted in an overshoot in recovery from inactivation, seen as an increase in the peak current; this was less prevalent in response to longer prepulse protocols. A prepulse duration of 20 ms was chosen as the standard protocol because inactivation amplitudes were sufficient to quantify and multiple protocols could be run on a given cell with this short-duration pulse.

Figure 7.

Calcium channel inactivation has a bell-shaped voltage dependence and is similar between hair cells of different papillary locations

A, examples of a hair cell's current response to a stimulus protocol using either a 200 or a 20 ms prepulse to potentials between −114 and 96 mV, incremented by 10 mV, followed by a test pulse to −14 mV for 20 ms and then a return to the holding potential of −84 mV. The stimulus protocol shown above the current records illustrates the shape of the protocol used. The traces shown are for prepulses to −84 mV, −14 mV and +86 mV (bottom to top). Inactivation is measured as the decrease in current amplitude during the test pulse. B, a plot of peak current during the step to −14 mV against the corresponding prepulse potential to demonstrate that the magnitude of current decrease was greater for longer pulse durations, but that the voltage dependence of the process was unaffected. The response to the 20 ms pulses typically showed a rebound effect where the current maximum was greater after depolarizing prepulses than control. C, a plot of percentage current inactivated against prepulse duration has an exponential relationship with a time constant of 35 ± 8 ms (n is given by points in the plot). A similar plot of the V1/2 for inactivation, measured from the Boltzmann fits to the data shown in B (data between −114 and −4 mV fit) showed no difference between prepulse values. D, currents in response to a protocol that varied the prepulse from −114 to 104 mV for 20 ms, followed by a 20 ms pulse to −10 mV from a holding potential of −84 mV are shown for a high- (top) and a low-frequency (bottom) cell. Traces shown are for prepulses to −84 mV, −14 mV and +76 mV. E, normalized plots of test current against prepulse potential were bell-shaped for both frequency positions. F, the V1/2 of inactivation measured from the Boltzmann fits to the data was −40 ± 2 and −46 ± 2 mV, and the slope value was 4 ± 2 mV−1 for both high- and low-frequency cells (r2 for fits were 0.99 and 0.97 for high- and low-frequency cells, respectively). G, a plot of the V1/2 for inactivation against the V1/2 for activation gives a linear relationship with a slope of 1.0 ± 0.1 and a y-intercept of −9 ± 3 mV (r2= 0.88).

Figure 8.

The physiological significance of calcium channel inactivation was investigated in both voltage- and current-clamp experiments

A, hair cell currents measured in response to depolarizing voltage steps between −64 and 0 mV from a holding potential of −84 mV (top) or −44 mV (bottom). B, IV plot of data in A showing reduction in the magnitude of the calcium current when the holding potential (Vh) was varied from −84 to −44 mV. •s represent the return to the control condition of −84 mV. C, current-clamp responses (averages of 20) from a hair cell where current (−25 pA) was first injected to hyperpolarize the cell to −84 mV and then injected (50 pA, depolarizing total of 75 pA) to depolarize the cell to its best resonant voltage (left) compared to the same cell where current was injected (50 pA) to elicit electrical resonance (right). The frequencies of the oscillations were 290 and 312 Hz for left and right responses, respectively. The quality (Q) of the resonance measured from the time course of the decay in the oscillations (Art & Fettiplace, 1987) according to the equation Q=[(πf0τ0)2+ 1/4]1/2, where f0 is the resonant frequency and τ0 is the time constant measured from the exponential decay of the oscillations at current onset, was 5.5 and 5.2, respectively.

Tonotopic differences

Inactivation was present in both high- and low-frequency cells (Fig. 1A and Fig. 7E). Using the 20 ms stimulus protocol, inactivation was greater in low-frequency cells, due most likely to the preponderance of the fast component of inactivation. Mean plots of test current amplitude at −14 mV against prepulse potential are given in Fig. 7E, where it can be seen that low-frequency cells consistently inactivated more than high-frequency cells. In addition, although both cells show a rebound, where currents elicited after positive prepulses were greater than control, low-frequency cells rebounded more. Replotting the data from Fig. 7E and fitting them with Boltzmann functions demonstrates that the low-frequency cells consistently inactivated at more negative potentials than high-frequency cells (Fig. 7F). Half-inactivating potentials were −44 ± 2 and −50 ± 2 mV for high- and low-frequency cells, respectively. No differences between slopes were observed (4 ± 1 mV−1). The difference in inactivation voltage correlated well with the difference in half-activation voltage (V1/2; Fig. 7G). The distribution forms a linear continuum with a slope of 1.0 ± 0.1 and a y-intercept of −9 ± 3 mV, suggesting that a similar mechanism was regulating inactivation at both frequency locations.

Physiological significance

To assess the role of inactivation on membrane excitability, two experiments were performed. First, cells were voltage clamped near the measured resting potential of the cell and calcium currents were elicited by the standard protocol (Fig. 8A). Currents elicited from a holding potential of −44 mV were significantly smaller than those obtained from a holding potential of −84 mV, demonstrating that inactivation can have a profound effect on the availability of channels for activation. The time course of inactivation was slowed to the point where it was not observed during the 20 ms pulse from a holding potential of −44 mV as compared to the obvious kinetic component in current traces elicited from a holding potential of −84 mV. The IV plots in Fig. 8B illustrate the decrease in current amplitude with no shift in V1/2 or voltage dependence. The reduction in current amplitude is in accord with the V1/2 for inactivation obtained from data shown in Fig. 7 and supports the conclusion that inactivation can be a major mechanism for regulating calcium-current amplitude physiologically. The second experiment investigated electrical resonance, the major tuning mechanism found in these hair cells (Crawford & Fettiplace, 1978; Art et al. 1986). In current-clamp mode, using a potassium-based intracellular solution, hyperpolarizing the hair cell to near −84 mV and then eliciting a resonant response near its resting potential was compared to simply eliciting a resonant response from the resting potential of the cell (Fig. 8C). The resonant frequency was consistently higher by 29 ± 8 % (n= 8) when resonance was elicited from the cell's resting position than from −84 mV, the opposite to what would be predicted if additional calcium channels were recruited and was most probably due to the additional charging of the membrane capacitance. The quality of the resonance was assessed from the decay in the onset oscillations (τ0 being the time constant of decay) and obtained from the equation Q= ((πf0τ0)2+ 1/4) (Art & Fettiplace, 1987) where, f0 is the resonant frequency and τ0 is the time constant of decay. Results were more variable, four out of eight cells showed no significant effect, in two cells the quality increased and in two cells the quality decreased. The variability comes, at least in part, from the change in frequency: if τ0 is compared, six out of eight cells had a longer τ0 from a holding potential of −84 mV than at their resting potential. However the data were compared, the effect of doubling the calcium-current amplitude had little effect on resonant properties, implying that hair cells have more calcium channels than are required to elicit electrical resonance.

Underlying mechanisms

The bell shape of the inactivation plots as well as the time course of inactivation and its sensitivity to blockers suggests that inactivation was driven by an increase in intracellular calcium (Brehm & Eckert, 1978). To test this hypothesis further, calcium was replaced with barium. Barium typically does not substitute well for calcium biochemically and thus inactivation should be reduced. In addition, barium is also a test for channel type in that it permeates L-type channels better than calcium, so current amplitudes should increase in the presence of barium. A summary of the data obtained by substituting barium for calcium is given in Fig. 9. Magnesium was removed from the external solution and calcium or barium was incorporated at 5 mm. The maximum current increased from 0.99 ± 0.06 to 1.75 ± 0.07 nA (n= 5) when barium was equimolarly substituted for calcium (Fig. 9A and D). The IV plot shifted in a hyperpolarized direction (Fig. 9D and E) and the activation kinetics slowed in barium (data not shown). The values of V1/2 for activation were −42 ± 1 and −53 ± 1 mV and the slopes were 3.4 ± 0.5 and 2.9 ± 0.2 mV−1 for 5 mm calcium and 5 mm barium, respectively. Similar results have been reported on dissociated turtle hair cells (Art & Fettiplace, 1987). The ion-dependent shift in the IV plot might implicate an interaction between the ion and channel, as suggested by Rodriguez-Contreras & Yamoah (2001). Inactivation remained, but was slowed in the presence of barium (Fig. 9AC). Responses to 1 s depolarizations of the voltage eliciting the maximal current are shown in Fig. 9A for cells bathed in 2.8 mm calcium, 5 mm calcium or 5 mm barium. The first 60 ms of the normalized currents are shown in Fig. 9B to illustrate the effect on inactivation. The fast and intermediate time constants were significantly slowed in barium as compared to calcium. A summary of the time constants is shown as a bar graph in Fig. 9C. No difference was found between time constants in 2.8 and 5 mm calcium. No difference was found in the slowest time constant between any treatment. Slow time constants of 920 ± 192, 961 ± 178 and 1128 ± 118 ms were obtained for 2.8 mm calcium, 5 mm calcium and 5 mm barium, respectively. A summary of the steady-state inactivation properties obtained from the standard prepulse protocol described earlier is given in Fig. 9F and G for 5 mm calcium and 5 mm barium. The voltage of half-inactivation was −47 ± 1 and −61 ± 1 mV with slopes of 4.0 ± 0.5 and 7 ± 1 mV−1 for 5 mm calcium and 5 mm barium, respectively. At first inspection, these results suggest that inactivation is driven by calcium, but barium appears to substitute for calcium.

Figure 9.

Inactivation persists with barium as the charge carrier

A, examples of currents elicited in response to a 1 s depolarization from −84 to −14 mV in the presence of 2.8 and 5 mm calcium or 5 mm barium (single traces). B, expanded view of the initial portion of currents elicited from A to better illustrate the effects on inactivation. C, fast and intermediate time constants of inactivation from fits to data shown in A. D, IV plots for activation using 5 mm calcium (□) and 5 mm barium (▵) demonstrating the increase in maximal current and a leftward shift in the plot. Symbols are the same for each panel. E, IV plots generated from the data shown in D normalized to peak current for each ionic condition and fitted with single Boltzmann functions. The V1/2 for records obtained in 5 mm calcium and 5 mm barium was −42 ± 1 and −53 ± 1 mV, respectively. The slopes for records obtained in 5 mm calcium and 5 mm barium were 3.4 ± 0.5 and 2.6 ± 0.2 mV−1, respectively. All fits had r2 values greater than 0.98. F, current-prepulse voltage plot demonstrating the effects of divalent ion species on inactivation. G, inactivation persists in the presence of barium. For clarity, the Boltzmann fits to the inactivation currents normalized to peak current (data from F) are shown without corresponding data points. The voltages of half-inactivation for 5 mm calcium and 5 mm barium were −47 ± 1 and −62 ± 1 mV, respectively. The slopes for 5 mm calcium and 5 mm barium were 4 ± 1 and 7 ± 1 mV−1, respectively. All fits had r2 values greater than 0.98.

To investigate further the calcium dependence of inactivation, the concentration of the intracellular calcium chelator BAPTA was increased from 1 to 10 or 30 mm. The results are summarized in Fig. 10. The inactivation index decreased from 23 ± 1 % in 1 mm BAPTA to 16 ± 1 % in 30 mm BAPTA. The time constant of the run-up decreased from 7 ± 2 to 5 ± 1 and then 3 ± 1 min in 1, 10 and 30 mm BAPTA, respectively. The similarity in sensitivity of the run-up and inactivation to calcium buffers might indicate a common mechanism. Peak current plotted against time after obtaining whole-cell recording demonstrates the complex effect of increased BAPTA (Fig. 10A: 1 vs. 30 mm BAPTA shown). With both 10 and 30 mm intracellular BAPTA solutions, the calcium current tended to peak after about 12 min and then was reduced by 12 % (10 mm BAPTA, n= 9) and 63 % (30 mm BAPTA, n= 6). The cause of this run-down is unknown. Both the activation plots and the inactivation plots shifted towards more depolarized values (Fig. 10B and C). The values of V1/2 of activation varied from −39 ± 1 to −28 ± 4 to −25 ± 2 mV, while half-inactivation voltages varied from −43 ± 2 to −33 ± 5 to −26 ± 3 mV for 1, 10 and 30 mm BAPTA, respectively (n= 12, 9 and 6, respectively). No difference between slope values was observed. Similar effects on activation have been reported for adrenal chromaffin cells, where a mechanism involving GTP and magnesium has been implicated (Bodding & Penner, 1999). The slopes of these curves were unaffected. The shift in the inactivation curve mirrors that of the IV plot. Confirmation of this is found in the plot of the voltage for activation against that for inactivation as shown in Fig. 10E. The data obtained under varying conditions of external divalent cations and internal buffering all fall along the same linear plot, suggesting a common underlying mechanism. Responses to 1 s depolarizations to −10 mV were also compared (Fig. 10D). The fast time constant was significantly increased by BAPTA from 5.5 ± 0.5 to 22 ± 3 ms in 1 and 30 mm BAPTA, respectively. However, the slowest time constant was reduced from 903 ± 132 to 245 ± 22 ms in 30 mm BAPTA. The proportion of fast:slow inactivation was increased for both the 10 and 30 mm BAPTA when compared to the 1 mm BAPTA. These results support the hypothesis that inactivation is driven by calcium, one possibility being the fast components being close to the channel and the slow component being further removed.

Figure 10.

Calcium buffers alter, but do not prevent inactivation of the calcium currents

A, the time course of run-up of the calcium current is decreased in 30 mm BAPTA (n= 6) as compared to 1 mm BAPTA (n= 12). The time constant of the exponential fit to the data shown in A decreased from 7 ± 2 to 2.7 ± 0.7 min as the concentration of BAPTA increased from 1 mm (□) to 30 mm (▵). In addition, the initial amplitude of the measured peak current was also increased with higher concentrations of BAPTA. Responses in 30 mm reached a peak current and then ran down over time by almost 500 pA. The sum of two exponentials was used to fit this data, with the rising phase having a time constant of 2.7 ± 0.7 min and the falling phase a time constant of 12.8 ± 0.5 min. Activation (B) and inactivation (C) IV plots generated from the protocols described in Fig. 2 and 7D, respectively, show a trend towards depolarized voltages for activation and the persistence of inactivation even at 30 mm BAPTA (n= 12 for 1 mm BAPTA (□) and n= 6 for 30 mm BAPTA (▵)). Boltzmann fits to the normalized IV curves gave V1/2 values of −39 ± 1 and −25 ± 2 mV for 1 and 30 mm BAPTA, respectively, and slopes of 4.1 ± 0.1 and 4.7 ± 0.3 mV−1, respectively. Boltzmann functions to the inactivation curves similarly shifted in a depolarized direction, with half-inactivation voltages of −43 ± 2 and −26 ± 3 mV, respectively. Inactivation protocols were as described earlier using 20 ms prepulses. The magnitude of inactivation was reduced from 23 ± 1 to 16 ± 1 % (1 mmvs. 30 mm BAPTA). D and E, current responses to 1 s depolarizations to −10 mV using an intracellular BAPTA concentration of 1 or 30 mm. Replotting the V1/2 of activation against the V1/2 of inactivation for 1 mm BAPTA (□, low frequency, ▵, high frequency), 10 mm BAPTA (♦), 30 mm BAPTA (▪) and 5 mm external barium (▴) has a linear relationship. The dashed line represents an intercept of 0 and a slope of 1. The continuous line is a linear regression having a slope of 1 ± 0.05 and an intercept of 7.5 ± 0.5 mV (r2= 0.83).

Two additional experiments were performed to attempt to decipher the mechanism of inactivation. The calcium chelator 5,5′-difluoroBAPTA (DFB) has a Kd of about 635 nm, whereas that for BAPTA is 160 nm. By incorporating DFB into the patch pipette it was possible to elevate baseline intracellular calcium levels. At steady state, calcium levels would be expected to approach the Kd of the buffer. An example of this type of experiment is shown in Fig. 11A where recordings at 100, 430 and 660 s after establishment of the whole-cell configuration are illustrated with their corresponding IV plots. The currents run down almost completely in the first 11 min, implicating a calcium-dependent inactivation process.

Figure 11.

Calcium currents run down to near 0 pA when intracellular calcium is elevated by using a low-affinity calcium buffer (1 mm difluoroBAPTA, n= 3)

A, examples of the time course of run-down, with the stimulus protocol given at the top and current responses over time shown below. B, IV plots show the decrease in current amplitude during the recording. No shift in activation was observed. C, electrodes, filled with an internal solution where magnesium was not included, also demonstrated strong run-down behaviour. ATP-dependent processes will fail without the presence of magnesium. D, examples of currents elicited at different times after establishing the whole-cell configuration with corresponding IV plots. Again, no shift in activation curve was observed (n= 3).

A second experiment tested for the possibility that phosphorylation or dephosphorylation was required for inactivation. Removal of magnesium from the intracellular pipette renders ATP inactive, as most processes depending on ATP require MgATP for activity. Under these conditions the calcium current inactivated robustly (Fig. 11C and D), but did not recover from inactivation, suggesting that the ability of the channel to pass current requires phosphorylation and that inactivation is, in part, due to the dephosphorylation of the channel. The increased robustness of inactivation may reflect the unmasking of the inactivation by inhibiting the rephosphorylation of the channel that would normally reduce the total inactivation recorded.


Several important findings have come from the present investigations. First, only L-type calcium channels are functionally present in turtle auditory papilla hair cells. Second, a tonotopic difference exists in the activation properties of the calcium channels, with low-frequency channels activating at hyperpolarized potentials and having faster rise times than high-frequency channels. Third, hair cell calcium channels inactivate. About half of the channels are predicted to be inactivated by a potential value near the cell's resting potential. Inactivation appears to be driven by calcium entry.

Types of channel present

Several pieces of data support the conclusion that the turtle auditory hair cell calcium current is carried only through L-type channels, implying that this channel is responsible for driving both electrical resonance and neurotransmitter release.

At depolarized potentials and high doses, dihydropyridines can block more than 90 % of the whole-cell calcium current (Fig. 3). Bay K 8644 potentiates the current (Fig. 4). Non-dihydropyridine L-type antagonists block more than 90 % of the current (Fig. 5). Selective blockers for other channel types including N, R, T and P/Q types were ineffective in antagonizing any part of the macroscopic current (Table 1). And finally, the current magnitude nearly doubled when barium was equimolarly substituted for calcium; a telltale signature of L-type channels is that barium is more permeable than calcium (Fig. 9). None of the pharmacological agents produced shifts in the IV plots that might be expected if a particular channel type was selectively blocked or unmasked. Single Boltzmann functions fitted all IV plots. In addition, no changes in current rise times were observed that might also suggest the presence of a second channel type (Zidanic & Fuchs, 1995). Together these data support the conclusion that only L-type channels are present. Calciseptine is an exception, since it is an L-type channel antagonist that was ineffective in blocking any part of the current. Calciseptine binds at the α subunit at a region overlapping the dihydropyridine binding site (Yasuda et al. 1993). Since nimodipine potency was limited, perhaps due to a difference in binding, it is possible that calciseptine affinity was also reduced due to this same difference. Nimodipine insensitivity is a hallmark of the α1D subunit of the L-type channel. In this light, calciseptine may be more discriminating in differentiating between L-type channels in that it has no effect on the α1D channel, while dihydropyridines antagonize both types of channels but with different efficacy. Taicatoxin is another example of an L-type antagonist that was ineffective in blocking the hair cell current. Taicatoxin also binds near the dihydropyridine site and so its effectiveness at blocking α1D-type conductances may be limited (Hamilton & Perez, 1987).

Native channel compared with expressed α1D channels

Hair cell calcium channels have similarities to expressed L-type channels of the α1D variety. The IV plots were shifted in a hyperpolarized direction with respect to the α1C type channel and are comparable to expressed α1D channels (Fig. 2; Koschak et al. 2001; Xu & Lipscombe, 2001). Activation kinetics were fast, with rise times of less than 0.5 ms at peak current levels, again similar to the kinetics reported for α1D channels (Koschak et al. 2001; Fig. 2). Nimodipine antagonized the current with an IC50 of 2.0 ± 0.1 μm for the hair cell channel as compared to 2.7 ± 0.3 μm for expressed α1D channels (Xu & Lipscombe, 2001; Fig. 3). The nimodipine block did not exhibit a u-shaped voltage dependence, as has been reported for some expressed α1D channels (Xu & Lipscombe, 2001). In addition, although the hair cell channels exhibited a significant degree of inactivation, it was not nearly as robust as that exhibited by expressed α1D channels, and the kinetics were different. Despite these small differences, the preponderance of the data supports the conclusion that the calcium channels are of the α1D variety of L-type channels.

Comparisons with other hair cell types

The data presented here are largely in agreement with previous work regarding calcium channels in auditory hair cells. That the channels are of the L-type has been suggested in turtle (Art et al. 1986), chick (Fuchs et al. 1990; Kollmar et al. 1997b; Spassova et al. 2001), guinea-pig outer hair cells (Nakagawa et al. 1991; Chen et al. 1995) and guinea-pig and mouse inner hair cells (Zhang et al. 1999; Platzer et al. 2000; Engel et al. 2002). That currents activated rapidly and at negative potentials is also in agreement with data obtained from auditory hair cells (Art & Fettiplace, 1987; Fuchs et al. 1990; Engel et al. 2002) and is consistent with the channel being of the α1D type. Evidence exists for the importance of the α1D subtype of channel in cochlear hair cells (Kollmar et al. 1997b; Platzer et al. 2000; Engel et al. 2002) and the data presented here suggest that this channel type is conserved across different species.

The major difference between the data presented here and previous work is the presence of a significant component of inactivation. Why turtle cells show inactivation and other auditory hair cell types do not is a mystery. It is possible that there are technical differences between experiments. For example, the presence of inactivation was very sensitive to the IPI used, so that if too short a time between pulses was used, time-dependent inactivation may not have been observed and maximal current amplitudes most likely underestimated. It is also possible that enzymatic digestion alters inactivation properties so that the difference observed here comes from recording in an intact papilla as compared to dissociated cells (Armstrong & Roberts, 1998). It is possible that the cells are at different states of metabolic stress, as calcium loading results in a loss of inactivation that is usually irreversible despite the presence of a reasonable amplitude calcium current. It is also possible that the larger current amplitudes measured here make the currents more susceptible to calcium-induced inactivation. Another possibility is that the present data were contaminated by some other outward current and that the calcium current itself was not inactivating. A variety of pharmacological experiments were performed to address this possibility and none was effective in altering inactivation. In addition, inactivation was significantly reduced by blockers that antagonized the calcium current, suggesting that calcium accumulation is important. Together these data suggest that inactivation is not an artefact of contamination by some additional conductance. Other hair cell types have shown inactivation that was attributed to an R-type current, (Martini et al. 2000; Perin et al. 2001). No additional channel types have been identified in turtle auditory hair cells.

Tonotopic differences

Both the current rise times and the steady-state IV plots suggest that low-frequency channels are different from high-frequency channels. Although the steady-state difference between frequency positions was modest, about 7 mV, it was consistent between measurements. Previous work has demonstrated a similar difference in values of V1/2 for the BK channels (Ricci et al. 2000). The present data also demonstrate that low-frequency cells have a hyperpolarized resting potential compared to high-frequency cells. It is probable that the hyperpolarized IV plot of the calcium current is, at least in part, responsible for the hyperpolarized IV curve of the BK channel, which is ultimately responsible for the difference in resting potential. Recent evidence expressing splice variants of the α1D channel reported no difference in activation properties (Xu & Lipscombe, 2001), implying that the difference in hair cell properties may not be a function of the α subunit. However, many splice variants exist for the α1D and not all have been expressed (Safa et al. 2001). Several additional possibilities exist. Differences in the auxiliary subunits, β or α2δ, might alter activation properties. Post-translational differences might also be responsible. In addition, accessory proteins like syntaxin have been purported to cause differences in activation properties (Yang et al. 1999). Future experiments are aimed at determining the underlying mechanism responsible for these differences.

Physiological significance of inactivation

Inactivation was comparable between low- and high-frequency cells despite the fact that high-frequency cells have approximately twice the current magnitude. Given that inactivation appears to be driven by calcium, it might be expected that cells with larger currents would inactivate more. However, previous work has suggested that the density of calcium channels remains constant between frequency positions, but that the surface area and the number of hot-spots increases tonotopically (Ricci et al. 2000). The present results demonstrating that inactivation was comparable between hair cells at different frequency positions would also support this hypothesis.

Although the protocols used for analysis evoked low levels of inactivation, examples demonstrated that long depolarizations can inactivate most of the current and at the hair cell's resting potential about 50 % of the current was inactivated. The conclusion that up to half of the calcium current might be inactivated at the hair cell's resting potential suggests a powerful mechanism for regulating both hair cell excitability and transmitter release. A possible mediator of this regulation is the efferent system, where hyperpolarization induced by efferent stimulation might be necessary to reprime the calcium channels, making more available to open.

What role does inactivation play in shaping the physiological response (i.e. electrical resonance) of the hair cell? Current-clamp results suggest that inactivation does not have a major effect on electrical resonance. Neither the frequency nor the quality of the resonance was affected. Therefore, the temporal component of inactivation does not play a significant role in shaping the hair cell's receptor potential. Current-clamp data support this conclusion. The steady-state inactivation properties suggest that as much as half of the current could be inactivated at physiological potentials and would be predicted to alter both the frequency and quality of resonance. However, no effect on resonance was observed, implying that the hair cell has more calcium channels than needed to drive electrical resonance, and further that addition of calcium channels has little effect on membrane excitability. Previous work has demonstrated that the ratio of calcium channels to potassium channels is 2:1 regardless of tonotopic position (Wu et al. 1995; Ricci et al. 2000). The lack of effect of the calcium channel number on membrane excitability may suggest that at any given time, one calcium channel is sufficient to activate its corresponding potassium channel due to its close proximity (Roberts et al. 1990). What then is the function of the additional calcium channels? The simplest possibilities are that these channels are present to ensure that the BK channel activates whenever a calcium channel does and/or that these channels regulate synaptic strength. Experiments are presently underway to investigate this possibility.

Mechanisms of inactivation

What are the mechanisms responsible for inactivation? Inactivation of L-type calcium channels can be calcium dependent, voltage dependent or phosphorylation dependent. The data presented here cannot distinguish between these three possibilities, but do suggest that, at least, part of the inactivation is regulated by intracellular calcium. Evidence implicating a calcium-dependent process is somewhat equivocal. Inactivation had the bell-shaped voltage dependence that is typical of a calcium-dependent process. Inactivation was driven by current such that the steady-state properties required activation of the current before inactivation occurred. Inactivation, particularly the fast component, was reduced or even eliminated when the calcium current was reduced pharmacologically, implicating the involvement intracellular calcium. Inactivation was persistent but reduced in experiments where barium replaced calcium and also in the face of 30 mm intracellular BAPTA, suggesting that calcium was involved but that the calcium dependence was somehow less sensitive to these manipulations. The insensitivity may be due, in part, to channel clustering. Calcium summating between channels of close proximity will limit the effectiveness of calcium buffers. Also, barium can substitute for calcium, albeit with a Kd that is 100 times that of calcium (Ferreira et al. 1997). Three factors make it possible that the concentration of barium can reach these higher levels. First, the channels are clustered and so local concentrations will be determined by the sum of ion entry through multiple channels. Theoretical arguments have demonstrated that free-calcium concentration can be quite high at the centre of the hot-spots (Roberts, 1994; Wu et al. 1996). Second, barium currents are almost double that of calcium currents at the same divalent concentration, meaning that double the number of ions will be entering the cell. Over the short time courses being investigated, diffusion will have an insignificant effect on the ions available at the channel, so the concentration should be directly proportional to ion entry. And third, barium does not bind BAPTA, so that intracellular buffering will not limit the increase in barium concentration. Together these factors make it likely that barium concentrations will elevate to levels much greater than calcium and thus substitute for calcium in generating inactivation.

In conclusion, the present work has demonstrated that a single class of calcium channels is present in turtle auditory papilla hair cells and that this class of channels must therefore drive both electrical resonance and synaptic transmission. In addition, a tonotopic difference in activation and inactivation properties was established, properties that may underlie the measured difference in resting potentials of the hair cell. Evidence is presented demonstrating that hair cell currents inactivate. Physiologically, inactivation does not alter the cell's electrical resonance properties. However as much as 50 % of the current can be inactivated at the cell's resting potential. Thus, regulation of inactivation may be an important mechanism for altering synaptic strength without altering the frequency or quality of tuning. Inactivation appears to be driven by calcium entry.


This work was supported by an RO1 from NIDCD to A. J. R. and by the Tinnitus Association. Our thanks to Katie Rennie, Chris LeBlanc and Hamilton Farris for careful review of the manuscript.