Authors' present addresses Z.-L. Mo: Pharmacia Co., Department of Pharmacology Room 319, 7250-209-319, 301 Henrietta Street, Kalamazoo, MI, USA.
C. L. Adamson: W. M. Keck Center for Collaborative Neuroscience, Rutgers University, Piscataway, NJ 08854, USA.
We have previously identified two broad electrophysiological classes of spiral ganglion neuron that differ in their rate of accommodation (Mo & Davis, 1997a). In order to understand the underlying ionic basis of these characteristic firing patterns, we used α-dendrotoxin (α-DTX) to eliminate the contribution of a class of voltage-gated K+ channels and assessed its effects on a variety of electrophysiological properties by using the whole-cell configuration of the patch-clamp technique. Exposure to α-DTX caused neurons that initially displayed rapid accommodation to fire continuously during 240 ms depolarizing test pulses within a restricted voltage range. We found a non-monotonic relationship between number of action potentials fired and membrane potential in the presence of α-DTX that peaked at voltages between –40 to –10 mV and declined at more depolarized and hyperpolarized test potentials. The α-DTX-sensitive current had two components that activated in different voltage ranges. Analysis of recordings made from acutely isolated neurons gave estimated half-maximal activation voltages of –63 and 12 mV for the two components. Because α-DTX blocks the Kv1.1, Kv1.2 and Kv1.6 subunits, we examined the action of the Kv1.1-selective blocker dendrotoxin K (DTX-K). We found that this antagonist reproduced the effects of α-DTX on neuronal firing, and that the DTX-K-sensitive current also had two separate components. These data suggest that the transformation from a rapidly adapting to a slowly adapting firing pattern was mediated by the low voltage-activated component of DTX-sensitive current with a potential contribution from the high voltage-activated component at more depolarized potentials. In addition, the effects of DTX-K indicate that Kv1.1 subunits are important constituents of the underlying voltage-gated potassium channels.
Systematic analysis of cells that comprise the auditory system has revealed a multitude of mechanical and electrical specializations that are essential for reliable and precise neural encoding of sound stimuli (For review, see Trussell, 1997; Oertel, 1999). In the cochlea, sounds are decomposed into their component frequencies while preserving temporal features of the stimulus. This is accomplished by the combined action of the basilar membrane, hair cell mechanoreceptors, and spiral ganglion neurons, which receive synaptic input from the hair cells and project to the cochlear nucleus in the brainstem.
Recordings from spiral ganglion neurons in vivo have been used to examine the composite behaviour of the cochlea during presentation of simple and complex sounds. Features of auditory encoding elucidated by these studies include frequency tuning, frequency-dependent phase locking, varied levels of spontaneous activity, adaptation in response to maintained sound stimuli and a host of other linear and non-linear responses to sound (For review, see Ruggero, 1992). Studies aimed at understanding the basic mechanisms of these processes have focused on properties of the basilar membrane, hair cell transduction and hair cell transmitter release, with the assumption that spiral ganglion neurons contribute little and serve primarily as passive conduits of auditory information into the central nervous system.
To begin to address this issue, we and others have developed preparations for studying spiral ganglion neurons in which the intrinsic features of the cells can be examined independently of their connections with hair cells (Santos-Sacchi, 1993; Davis, 1996; Lin, 1997; Mo & Davis, 1997a,b; Lin & Chen, 2000). For our studies, we have utilized postnatal neurons, just prior to myelination, which allows us to make whole-cell current-clamp and voltage-clamp recordings from these cells before the soma becomes inaccessible and enzymatic treatments become necessary to prepare cells for recording. This work revealed that spiral ganglion neurons display heterogeneous and complex responses to electrical stimulation, and can be divided into two distinct classes based upon their rate of accommodation (Mo & Davis, 1997a; Mo & Davis, 1997b).
By using cell-attached current-clamp recording we showed that primary auditory neurons, like their central counterparts, show changes in their firing patterns upon exposure to α-DTX. Phasic responses to depolarization became sustained, particularly at relatively low depolarizations. Voltage-clamp recording revealed both low and high voltage activated components of α-DTX-sensitive current, implicating Kv1.1 in termination of action potential firing. We tested this explicitly with the Kv1.1-selective blocker dendrotoxin-K (DTX-K) and found that this toxin reproduced the action of α-DTX. This conclusion is consistent with immunohistochemical staining, and further supports the idea that spiral ganglion neurons, as with the other elements of the auditory pathway, contain specific populations of ion channels that tailor their firing for specific functions.
Experiments were performed on mouse (CBA/CaJ) spiral ganglion neurons. The procedures involving animals in this study have been approved by The Rutgers University Institutional Review Board for the Use and Care of Animals (IRB-UCA), protocol #90-073. Neonatal animals, postnatal day 3 to 8 (P3-P8) were decapitated and both inner ears were removed from the base of the cranium for further dissociation. Neuronal cultures were plated in 35 mm culture dishes coated with poly-l-lysine and maintained for up to 21/2 weeks in growth medium: DMEM (Sigma, D-6171), supplemented with 10 % fetal bovine serum (Sigma, F-2442), 4 mm l-glutamine (Sigma, G-6392) and 0.1 % penicillin-streptomycin (Sigma, P-0781).
The whole-cell configuration of the patch-clamp technique was used to obtain current and voltage-clamp recordings from spiral ganglion neurons in vitro (Hamill et al. 1981). Whole-cell, rather than perforated patch-clamp recording, was utilized in order to obtain optimal electrical access to the cell soma. Electrodes were pulled on a two-stage vertical puller (Narishige, PP-83) and the shafts were coated with sylgard (Dow Corning) to reduce pipette capacitance. Just prior to use, electrode tips were fire-polished (Narishige MF-83 microforge). Electrode resistances ranged from 2 to 5 MΩ in standard pipette and bathing solutions (see below). Pipette offset current was zeroed immediately prior to contacting the cell membrane and electrode capacitance was cancelled after seal formation. After establishing the whole-cell recording mode, series resistance was compensated in accord with the amplifier manufacturer recommended procedure with lag set to 10 μs and prediction set to 80–90 %. On-line leak subtraction was not used. For the thirteen recordings from acutely isolated neurons that satisfied our selection criteria (see below), the whole-cell capacitance transient in the absence of whole-cell capacitance compensation showed no evidence of multiple decay components that could be indicative of incomplete separation of the soma from its processes. This transient was produced by a 10 mV hyperpolarizing (n= 3) or depolarizing (n= 10) voltage pulse from a holding potential of −80 mV. The voltage change inside the cell would be expected to be approximately 10 mV although the actual voltage command generated by the amplifier series resistance correction circuitry would be larger (proportional to the series resistance setting) and a feedback function of the current flow. The whole-cell capacitance transient was also used for measurement of the compensated series resistance (Scroggs & Fox, 1992) which ranged from 0.4 to 0.7 MΩ. For three of the recordings the compensated series resistance was also measured from the calibrated series resistance and correction dials on the Axopatch 200 amplifier (Rs= 5–9 MΩ, 85–95 % compensation).
A standard set of solutions was used to approximate physiological conditions. The basic internal solution was (in mm): 112 KCl, 2 MgCl2, 0.1 CaCl2, 11 EGTA, 10 Hepes, pH 7.45 (with KOH). On the day of the recording 2 mm NaATP was added to the stock solution and the pH was readjusted to 7.4 with KOH. Neurons were exposed to the following bath solution (in mm): 137 NaCl, 5 KCl, 1.7 CaCl2, 1 MgCl2, 17 glucose, 50 sucrose, 10 Hepes, pH 7.45 (with NaOH). For selected experiments, rapid solution changes were achieved with a micro-perfusion system (Ogata & Tatebayashi, 1991). α-DTX and DTX-K were obtained from Alomone Labs (Jerusalem, Israel) and used at a concentration of 100 nm to produce maximal block of Kv1.1 (Owen et al. 1997). In early experiments, this point was confirmed by applying a series of toxin concentrations from 50–200 nm. The toxins were prepared as stock solutions in distilled water and were dissolved into bath solution for application to neurons.
Recordings were made at room temperature (19 to 22 °C) using an Axopatch 200 (Axon Instruments) patch-clamp amplifier. Data were digitized with an Indec IDA 15125 interface in an IBM-compatible personal computer. Unless otherwise indicated, each segment of data was digitized at 5 kHz and filtered at 1 kHz. The programs for data acquisition and analysis were written in Borland C++ and Microsoft Visual Basic (generously contributed by Dr Mark R. Plummer, Rutgers University). To ensure that cells studied had not been previously exposed to toxin, we only recorded from one neuron per culture dish.
Because our cultures contained glia as well as neurons, we only evaluated cells that possessed TTX-sensitive, rapidly activating Na+ currents in voltage-clamp and action potentials in current-clamp. Current-clamp recordings were considered acceptable when they met the following criteria: low noise levels, stable membrane potential in response to maintained constant current injection, discernible membrane time constant upon step current injection and overshooting action potentials (magnitudes of at least 70 mV). If any of these parameters changed during an experiment, the remaining data were not analysed. To distinguish between action potentials and other active membrane processes we defined action potentials as depolarizing events that showed a distinct inflection point as distinguished by eye (Mo & Davis, 1997a). Whether depolarizations without a discernable inflection point were due to distantly generated action potentials or to local active membrane currents was not determined in this study.
For current-clamp recordings, we quantified action potential number and input resistance as a function of the steady-state voltage achieved during a 240 ms square pulse of injected current. Because of the known distortions produced by patch-clamp amplifiers on action potential waveforms during current-clamp recordings (Magistretti et al. 1996, 1998), detailed measurements of absolute action potential characteristics such as amplitude and rate of rise were not undertaken in this study. We did, however, do a comparative analysis in which we quantified interspike interval and action potential amplitude before and after DTX application. For interspike interval in particular, we concentrated on the lower frequencies that are less affected by current-clamp artifacts (Magistretti et al. 1998).
All measurements are presented as means ±s.e.m. and Student's t test (two-tailed) was used to determine statistical significance. Maintained current injection was typically used to keep the membrane potential of a recorded neuron at −60 mV. Superimposed on this were 240 ms square pulse injections delivered at 3 s intervals of either depolarizing or hyperpolarizing current. The steady-state voltage level was defined as the absolute value of the membrane potential achieved close to the termination of the constant current injection. For cells that exclusively displayed rapid accommodation, this value was measured with a cursor placed by eye. For cells in which there was repetitive firing throughout the current pulse, the steady-state voltage was estimated for all traces (even those that showed accommodation) by averaging the voltage recorded during the entire 240 ms stimulus, as carried out in a previous investigation (Mo & Davis, 1997a). This approach yielded values that mapped onto current-to-voltage relationships that were constructed mainly of directly measured values without introducing any discontinuities in the function, suggesting that the method was valid.
To study the characteristics of the DTX-sensitive current, cells were voltage-clamped using the same solutions as for current-clamp. Under our recording conditions, we found the effects of α-DTX to be irreversible, so we could not check for the recovery of the currents after washing. Therefore, we took several precautions to eliminate sources of variability during a recording that could alter our observations of the α-DTX-sensitive current. We repeatedly measured the current-to-voltage relationship for the TTX-sensitive Na+ current throughout an experiment, and only used recordings in which these parameters remained stable. This precaution established that decreases in current magnitude could be attributed to DTX application. The Na+ current-to-voltage relationship was also utilized as an indication of the effectiveness of our voltage-clamp. For quantitative analysis, we studied acutely isolated neurons devoid of substantial neurites in order to evaluate DTX-sensitive currents under optimal voltage control. Cells were isolated with the same procedures described above, but were used immediately rather than being plated.
Tissues were fixed in 100 % methanol at −20 °C for 6 min; cultures were then incubated for 1 h in a 5 % solution of serum of the species in which the secondary antibody was raised. The primary antibody was applied and left for 1 h at room temperature or overnight at 4 °C. A fluorescent-conjugated secondary antiserum was subsequently applied for 1 h. Between each step, the preceding solution was removed by washing three times in 0.01 m phosphate buffered saline (pH 7.4) except after the application of blocking solution. For control cultures, tissue was treated identically except that the primary antibody was not included.
Neurofilament 200 (NF200) antisera were used to distinguish neurons from background satellite cells that survived in vitro. NF200 monoclonal antibody (Sigma, N-0142) stained both the cell soma and processes, whereas NF200 polyclonal antibody (Sigma, N-4142) primarily stained the processes, labelling only an occasional cell soma. Because neurofilament expression can be developmentally regulated (Schwartz et al. 1994; Liu et al. 1994), a previous study confirmed that all cells that contained neuron specific enolase epitopes also labelled with NF200 (Mou et al. 1997).
Kv1.1 polyclonal and monoclonal antibodies were obtained from two different sources; however, the staining patterns that resulted from the two different antibodies to the same ion channel protein were indistinguishable. Kv1.1 polyclonal antibody (Alamone, APC-009) was made against amino acid residues 416–495 of the carboxy terminus of mouse full-length Kv1.1 protein. Kv1.1 monoclonal antibody (Upstate Biotechnology, 05-407, Waltham, MA, USA) was made from a synthetic peptide corresponding to amino acid residues 458–476 of rat brain Kv1.1.
Results from this study were based on a total of 38 current-clamp recordings, 37 voltage-clamp recordings and nine separate immunocytochemical experiments with monoclonal and polyclonal antibodies to Kv1.1. As observed previously (Mo & Davis, 1997a), current-clamp recordings revealed that the majority of neurons (32/38; 84 %) fired fewer than six action potentials in response to the 240 ms duration constant current injections and were therefore placed in the rapidly adapting category. The remaining six recordings (16 %) were classified as slowly adapting because they fired greater than six action potentials (17.3 ± 2.1) in response to step depolarizations of the same duration.
The mamba snake toxin, α-dendrotoxin (α-DTX), is a specific blocker of potassium channels containing the Kv1.1, Kv1.2 and Kv1.6 alpha subunits (For review, see Harvey, 2001). The concentration utilized in this study, 100 nm, was chosen to produce maximal inhibition. Application of α-DTX to the external recording solution decreased the prevalent accommodation displayed by spiral ganglion neurons at low voltage levels, converting rapidly adapting neurons into ones that fired multiple action potentials. This transformation is shown for one neuron in Fig. 1A. Voltage responses produced by four different depolarizing current injections are shown on the left side of the figure (labelled as ‘control’). For the lowest amount of current injection (50 pA, which changed the membrane potential to −42 mV), the neuron bathed in control solution fired only a single action potential. Further depolarization evoked a maximum of two action potentials.
Exposure to α-DTX dramatically altered the firing pattern described above. Current-clamp recordings at comparable constant current injections, which depolarized the cell to similar voltage levels, showed that the action potential number increased substantially. In the example shown (Fig. 1A, traces on right), the maximum number of action potentials was evoked by moderate current injections, with action potential number decreasing at more depolarized membrane potentials.
The high degree of specificity of α-DTX action (Grissmer et al. 1994; Owen et al. 1997) allowed us to assess the involvement of channels containing a subset of Drosophila Shaker-type subunits to spiral ganglion neuron firing patterns. Four different parameters were evaluated: APmax (the maximum number of action potentials that a cell was capable of firing), spike amplitude decrement, interspike interval and input resistance. As described in detail below, three of these parameters, APmax, interspike interval and input resistance, were significantly altered by DTX, spike amplitude decrement was not.
Of the 32 current-clamp recordings made from rapidly adapting neurons, 31 of the them showed a 2- to 14-fold increase in APmax upon exposure to α-DTX (Fig. 1). The plot of APmax against voltage had an inverted V-shape (Fig. 1B); the voltages at which peak firing occurred were not the same from cell to cell, but ranged from −13.5 to −37 mV. As a population, rapidly adapting spiral ganglion neurons showed a significant increase in APmax upon exposure to α-DTX (P < 0.01), and the maximum number of action potentials fired became indistinguishable from that of slowly adapting (SA) neurons that were not exposed to α-DTX (Fig. 1C).
We analysed interspike interval for rapidly adapting neurons, but because these cells fired very few action potentials, we restricted our measurements to the time period between the first and second spikes. Using this approach we noted that this first interspike interval was consistently altered when rapidly adapting neurons were exposed to α-DTX (Fig. 2). An example of interspike interval measurements made from a single experiment is shown in Fig. 2A. For our initial analysis, interspike interval was plotted against the amount of injected current. As expected, the interspike interval decreased systematically with current magnitude. After α-DTX exposure, the function was shifted to the left. It is noteworthy that the difference between control and α-DTX is pronounced at the low firing frequencies less affected by current-clamp artifacts (Magistretti et al. 1998).
In order to place these results in a more physiological context, we also plotted interspike interval against the steady-state voltage achieved in response to depolarizing current injection (Fig. 2C). As expected, the shape of the function obtained was similar to that seen when plotted against current. For large depolarizing currents, interspike interval was not affected by exposure to α-DTX (7.6 ± 0.56 vs. 7.8 ± 0.5 ms for control vs.α-DTX at −15 mV, P > 0.05), but there was, however, a significant reduction in the voltage needed to elicit two action potentials (-38.1 ± 1.9 vs.−49.4 ± 1.5 mV, P < 0.01). In order to compare the overall dependence of interspike interval on membrane voltage quantitatively, independent of interspike intervals at particular voltages, we fitted the interspike interval distributions with a single exponential and determined the current (Fig. 2A) or voltage (Fig. 2C) at which the maximum change in slope occurred (arrows) both before (shaded diamonds) and after (filled triangles) α-DTX exposure. The point of maximal change in slope is an approximate indicator of the upper extent of the voltage range in which interspike interval is most sensitive to changes in membrane potential. In the example from a single recording shown in Fig. 2C, this value was shifted by approximately 10 mV in the hyperpolarizing direction after application of α-DTX. For the population, the mean voltage of maximal change in slope was significantly reduced by application of α-DTX, changing from −32.9 ± 1.9 mV for control recordings to −40.2 ± 1.5 mV (P < 0.01, Fig. 2D).
We also analysed the effects of α-DTX on APmax and interspike interval in slowly adapting neurons. On average, α-DTX had no significant effect on APmax in this population of cells (17.3 ± 0.9 vs. 19 ± 2.2, P > 0.05, n= 6). Inspection of individual recordings, however, showed that α-DTX did increase APmax in two out of the six recordings of slowly adapting neurons. An example of one of these cells is shown in Fig. 3. The five vertically aligned traces labelled ‘control’ (Fig. 3A) demonstrate that before α-DTX exposure the neuron was capable of firing multiple action potentials at moderate membrane potentials with accommodation evident at more depolarized levels (top traces). This was the typical pattern observed from slowly adapting spiral ganglion neurons, which also resulted in an inverted V-shaped relationship between action potential number and steady-state membrane potential (Fig. 3B, shaded diamonds; also see Mo & Davis, 1997a). The five vertically aligned traces labelled ‘α-DTX’ in Fig. 3A were taken from the same control neuron after exposure to 100 nmα-DTX. The absolute number of action potentials increased at voltages ranging from −35 to −45 mV; however, as under control conditions, accommodation was present at more depolarized levels (Fig. 3B, filled triangles).
For the slowly adapting neurons that responded to α-DTX, average interspike interval was also decreased by the toxin. Comparison of traces (second from the bottom; Fig. 3A) shows that even though the cell fired throughout the duration of the step current injection under control conditions (yielding 13 action potentials) an even greater number of action potentials were fired following α-DTX application (17 action potentials). In this case, the average interspike interval decreased by 5 ms (from 19.1 to 14.1 ms). Systematic analysis showed, however, that a change in interspike interval did not correlate completely with the increase in APmax (Fig. 3C). For membrane potentials < −40 mV (Fig. 3C, arrow), it is evident that interspike interval decreased, and may account for, the increase in APmax. For membrane potentials > −40 mV, however, APmax was still increased by α-DTX but interspike interval was not. The other slowly adapting neurons that showed increased APmax following α-DTX application displayed similar responses (data not shown).
Another feature of the spiral ganglion neuron firing patterns is the decline in action potential amplitude with each subsequent spike (Lin, 1997; Mo & Davis, 1997a; Lin & Chen, 2000), which we also observed in this study. This was evident in the rapidly adapting neurons (Fig. 1A) as well as the slowly adapting ones (Fig. 3A). The activation of homomeric Kv1.1 channels by low voltages suggests that they would be unlikely to contribute to this phenomenon. Our prediction, therefore, is that action potential amplitude should be unaffected by DTX application as has been seen by other investigators (Wang et al. 1998). Nonetheless, the fact that Kv1.1 can form heteromultimers with the other Kv1 subunits makes such predictions uncertain, and so for completeness we measured the difference in amplitude between the first and second spikes for each voltage level that was tested and determined whether this parameter was altered by α-DTX. The change in action potential amplitude as a function of membrane potential was linear and did not differ substantially between the control and α-DTX conditions (not shown). The slopes of the linear fits compared between experiments and to evaluations of slowly adapting neurons showed no significant differences (P > 0.05). The average slope in the control group which consisted of rapidly adapting neurons before α-DTX application was −0.99 ± 0.06. After α-DTX exposure the average slope remained essentially unchanged at −0.94 ± 0.08. These values also did not differ significantly from measurements of the amplitude difference between the initial two spikes fired by slowly adapting neurons (-0.87 ± 0.1).
In order to determine the voltage range in which α-DTX was having its effect, we evaluated current-to-voltage relationships in rapidly adapting neurons before and after α-DTX application. As shown in Fig. 4, measurements of input resistance after the membrane potential had achieved steady-state were altered after exposure to α-DTX at low membrane potentials, but not at the higher voltage levels. Figure 4A shows an example of the pattern that we consistently observed. Input resistance (Ri) calculated at high voltages (-25 to −5 mV) was not substantially different between control and α-DTX conditions (23.1 and 37.8 MΩ, respectively). However, membrane resistance changed considerably at low voltage levels (42.5 and 199 MΩ, respectively at −50 to −40 mV). Averaged data shows that the differences observed at low voltage levels was significant (P < 0.01) whereas the change at higher voltage levels was not. From a membrane potential of −60 mV, the average input resistance of rapidly adapting cells at low voltage levels was 63.4 ± 6.4 MΩ, which increased to 146 ± 12.4 MΩ after α-DTX exposure (Fig. 4B). In contrast, the input resistance changed little at higher voltage levels (Fig. 4C; control conditions Ri= 31.3 ± 2.8 MΩ; after α-DTX exposure Ri= 36.4 ± 2.4 MΩ).
α-DTX blocks ion channel subunits; it is not surprising, therefore, that application of this toxin to neurons that possess these channels would cause increased membrane resistance. It is interesting, however, that the change in resistance was far more evident at lower voltage levels, suggesting that a predominant action of α-DTX is mediated via low voltage activated K+ channels such as Kv1.1.
The ionic current responsible for mediating the effect of α-DTX on adaptation and firing rate in spiral ganglion neurons was observed directly with whole-cell voltage-clamp recordings. In order to compare these experiments with the results from the current-clamp recordings, the voltage-clamp data was obtained from cells in culture maintained under similar conditions. A recording representative of the experiments of this type is shown in Fig. 5. To facilitate analysis of K+ currents, the transient Na+ current was blocked with 1 μM TTX and Ih (inward rectifying cationic current) was blocked with 5 mm CsCl. The large outward current that remained after these treatments was reduced by addition of α-DTX (Fig. 5A). To examine the timecourse of the α-DTX blockage, cells were held at −100 mV and stepped to −40 mV, a potential that would reveal mainly low voltage activated currents. With this voltage protocol, 86 % of the outward current was blocked 140 s after α-DTX application (Fig. 5B). Although a substantial amount of current remained at more depolarized step potentials, outward currents activated at voltages between −70 and −30 mV were all but eliminated by α-DTX (Fig. 5C). The current-to-voltage relationship measured from the subtracted currents (Fig. 5D) and the calculated conductance (Fig. 5E) suggested that the α-DTX current consisted of multiple components. There was an initial increase in conductance at approximately −70 to −30 mV and a second increase at more depolarized levels (+10 to +30 mV).
Finding two components of α-DTX-sensitive current was somewhat surprising, but might have been an artifact of the compromised voltage control seen in recordings from cultured neurons. To address this issue, our quantitative analysis of α-DTX-sensitive current was made from recordings of acutely prepared neurons, isolated only hours before recording. The fragility of the cells made this method substantially more time consuming, but, as described below, the voltage control attained in successful recordings was excellent. Na+ currents were not blocked in these experiments, but were instead used to gauge the quality of voltage control achieved in each of these recordings. The peak of the inward transient current that was blocked previously with TTX showed a smooth and graded increase in amplitude at low test potentials and became progressively smaller as the depolarizations increased (not shown). Discontinuities associated with a lack of voltage control were not observed in these recordings. In addition, the amplitude and voltage dependence of the Na+ current were unaffected by the application of α-DTX (not shown). This observation strongly supports the conclusion that the changes in the outward currents could be attributed to toxin exposure.
Outward currents were reduced in acutely isolated neurons exposed to α-DTX. This can be seen by comparing control recordings (Fig. 6A) to post exposure recordings (Fig. 6B). The resultant subtraction currents (inset, Fig. 6C) were used to construct current-to-voltage relationships. Averages made from multiple recordings again reveal and confirm the multiple components of α-DTX-sensitive current (Fig. 6C). Although they differed in their relative proportion from cell to cell, all recordings showed evidence of separate low voltage and high voltage components. Using an estimated reversal potential of −80 mV, we calculated the conductance of the α-DTX-sensitive current which was normalized to the maximal conductance of the low voltage component (Fig. 6D). We then fitted the data with either a single (low voltage component only) or double (the sum of both components) Boltzmann relationships. Based on these fits, the average half-maximal activation voltage of the low voltage component was −62.3 mV when fitted alone, or was −62.9 mV when fitted together with the high voltage component. The half-maximal activation voltage of the high voltage component was estimated to be 11.9 mV, but this value must be interpreted with caution because the acutely isolated neurons did not tolerate sufficiently large depolarizations to determine the maximal conductance of this component.
α-DTX blocks three potassium channel subunits, Kv1.1, Kv1.2 and Kv1.6, which can combine either singly or in combination to form homomeric or heteromeric ion channels. It is possible, therefore, that the functional consequences of α-DTX block are mediated by any combination of these subunits or even other subunits, such as Kv1.4, known to associate with the α-DTX-sensitive subunits. To begin to address this issue, we focused on the role of the Kv1.1 subunit by doing immunohistochemcial labelling and by using the selective Kv1.1 blocker dendrotoxin-K (DTX-K).
In correspondence with the electrophysiological recordings, the majority of spiral ganglion somata and processes that were identified with neurofilament-200 (NF200), which we have previously shown to label the entire population of neurons in vitro (Mou et al. 1997), also co-labelled with Kv1.1 antibodies (Fig. 7A). This double labelling revealed, however, that not all NF200 positive cells were also Kv1.1 positive (orange neuronal profiles in double exposures; the arrows in Fig. 7Aa, b and c show an example of one such neuron). Therefore, our electrophysiological and immunocytochemical results show that most, although not all, postnatal spiral ganglion neurons display rapid accommodation and stain with Kv1.1 antibodies in tissue culture.
Application of DTX-K to cultured neurons affected firing pattern in the same way as α-DTX. Rapidly adapting neurons that fired phasic bursts of action potentials under control conditions showed sustained firing to maintained depolarization in the presence of DTX-K (Fig. 7A), and APmax was significantly increased from 2.9 ± 0.3 to 24.8 ± 2 (P < 0.01, n= 5, Fig. 7C).
As also observed with α-DTX, we studied the ionic current blocked by DTX-K in acutely isolated neurons. The current-to-voltage relationship again was composed of two distinct components that activated in different voltage ranges. Analysis of conductance gave estimates for the half-maximal activation voltage of the low voltage component of −59 mV when fitted alone, or −59.5 mV when fitted together with the high voltage component (Fig. 7D). The half-maximal activation voltage of the high voltage component was estimated to be 0.1 mV, but as above, this value must be interpreted carefully.
DTX-sensitive ion channels are widely distributed in both neural and non-neural cell types, and found in both central and peripheral neurons. By using a combination of electrophysiological and immunohistochemical approaches, we have shown that presumptive Kv1.1-containing voltage-gated potassium channels shape the firing patterns of postnatal spiral ganglion neurons by influencing the number and rate of action potentials generated in response to a maintained depolarization. The findings suggest that these channels may be important regulators of neuronal firing during development, and also indicate that intrinsic properties of spiral ganglion neurons may contribute to the adaptation observed in the adult peripheral auditory system.
Application of both α-DTX and DTX-K increased the number of action potentials fired by spiral ganglion neurons, suggesting that normal function of the channels is to limit firing in response to sustained input. When considered in the context of auditory stimulation, this could contribute to the known adaptation of auditory neuron responses to maintained sound. This point is considered in more detail below. In addition, we noted that the α-DTX shifted the dependence of interspike interval on membrane potential to more hyperpolarized values. If auditory information is encoded to some extent by this parameter, then the DTX-sensitive channels may also regulate the voltage range in which this aspect of cell firing is most sensitive to small changes in input.
We also observed that the α-DTX-sensitive current was comprised of two components with different half-maximal activation voltages. Use of acutely isolated neurons without extensive processes made us confident that this finding was not a result of poor voltage control. The fact that both components were also seen in the DTX-K-sensitive current suggests that the different components cannot be simply attributed to α-DTX block of multiple kinds of potassium channels subunits. Thus we predict that the two components result from different combinations of Kv1.1 with other members of the Kv1 family, regardless of whether the other subunits are blocked by dendrotoxins or not. It has been well established that both α-DTX and DTX-K can block with high affinity both Kv1.1 homomeric channels as well as Kv1.1-containing heteromeric channels (Wang et al. 1999; Hatton et al. 2001), so our findings are consistent with this possibility. Specific tests of this hypothesis, however, must await development of more specific channel blockers.
At present, we tentatively conclude that it is the low voltage activated component of DTX-sensitive current that is responsible for the effects observed on action potential firing. Both α-DTX and DTX-K produced maximal changes in APmax in response to modest depolarizing voltages, and the change in input resistance seen in the low voltage range was much more pronounced than that seen in the high voltage range. The effects observed were also much more prominent in the rapidly adapting category of spiral ganglion neuron than in the slowly adapting category.
The functional significance of the high voltage activated component of DTX-sensitive current is not yet clear. Our voltage-clamp recordings demonstrate its presence, but measurements of input resistance did not reveal a large effect of α-DTX in the high voltage range. It must be stated, however, that although we did not observe a statistically significant difference in the population of neurons studied (Fig. 4C), a change in Ri was evident in individual recordings, although its overall contribution may be minor when considering the numerous other potassium channel conductances active in this higher voltage range.
Our data were largely obtained by making recordings from the cell bodies of cultured spiral ganglion neurons obtained from P3-P8 animals and placed in tissue culture. The reason for this approach was technical. Cells studied at this time are not yet myelinated, although they are excitable and survive the procedure well. We can be confident, therefore, that electrophysiological response properties (or the absence thereof) are not influenced by poor viability. We also controlled for changes in cellular properties during time in tissue culture by recording from acutely isolated neurons, and found no obvious differences between cells obtained with the different procedures. It has been demonstrated, however, that Type I spiral ganglion neurons (including their cell bodies) eventually become myelinated and by analogy with other peripheral nerves (Vabnick et al. 1999), the Kv1.1 channel protein observed in the adult neurons (Tempel et al. 1996) may become redistributed. In addition, mature auditory function is not seen in the mouse until 2 months of age, although the fundamental response properties of cells in the murine cochlear nucleus are established as early as P7 (Wu & Oertel, 1987). Implications of these issues with regard to the functional role of Kv1.1 in spiral ganglion neurons are considered below.
Developmental changes in the distribution of Kv1.1
During development of the sciatic nerve, the function of Kv1.1 and Kv1.2 potassium channels has been suggested to progress through three distinct stages (Vabnick et al. 1999). At early postnatal times, the channels contribute to action potential generation and speeding of repolarization. At intermediate times, the channels prevent bursting behaviour in response to a single stimulus and thus guard against excessive repetitive firing. At later times, as myelination nears completion, the Kv1.1/1.2 channel types are sequestered to juxtaparanodal regions, where their function is controversial. Although some investigators postulate that they become electrically silent as animals mature (Vabnick et al. 1999), others find that elimination of specific internodal K+ channel subunits can have an impact on conduction under particular conditions (such as in the presence of TEA or during cooling). Data from these and related studies have lead to the suggestion that internodal K+ channels prevent excessive depolarization in 3-to 5-month-old rats (David et al. 1993, 1995), alter sciatic nerve conduction in 2-to 4-month-old mice (Smart et al. 1998) and prevent hyperexcitability in both young and older (> 3 months) mice at the transition zone between myelinated and unmyelinated membrane where axons form presynaptic terminals on postsynaptic cells and are subject to an impedance mismatch (Zhou et al. 1998, 1999).
There are useful comparisons to be made between those studies and our results with postnatal spiral ganglion neurons. First, as with sciatic nerve, we find that Kv1.1 and Kv1.2 channels are present during early postnatal development and they act to prevent bursting in response to maintained depolarization. Second, immunohistochemical labelling has shown that these potassium channel types are present in adult spiral ganglion neurons (Tempel et al. 1996), and thus may act at the transition zone between unmyelinated and myelinated membrane where action potential initiation occurs in spiral ganglion neurons in vivo. Third, there is another area of impedance mismatch in spiral ganglion neurons that may be unique to this cell type. In these neurons, the cell body is myelinated, and there are dramatic changes in longitudinal resistance as the action potential propagates from the ‘dendritic’ portion of the axon, through the cell body, and into the central axon. This impedance mismatch has been suggested to be a potential site of conduction failure (Robertson, 1976; Yates et al. 1985), and spiral ganglion neurons show specializations that may counter this. One of the most obvious is the difference in somatic myelination as compared to axonal myelination (Romand et al. 1980; Romand & Romand, 1986; Romand & Romand, 1987, 1990; Goycoolea et al. 1990). Instead of the numerous wrappings of compact myelin, somatic myelin is relatively loose and heterogeneous. These differences have lead to the suggestion that the soma may actually be excitable (Robertson, 1976). It is evident, therefore, that this region of the cell cannot simply be considered as an internode comparable to that of a true axon. Indeed, the somata of spiral ganglion neurons may be more like the initial spike initiation area than axon. In any case, recordings from the soma are likely to reveal interesting specializations that impact normal action potential conduction in these cells.
Composition of DTX-sensitive current
The V1/2 (half-maximal voltage) of the low and high voltage activated components of the α-DTX-sensitive currents in spiral ganglion neurons were −62.9 and 11.9 mV, respectively. The low voltage activated component of the α-DTX-sensitive current showed little inactivation and is similar to what has been observed by others in auditory neurons. For example, the V1/2 was −58 mV for the avian nucleus magnocellularis (Rathouz & Trussell, 1998), −50 mV for neurons in the rat medial nucleus of the trapezoid body (Brew & Forsythe, 1995) and −45 mV in octopus cells in the cochlear nucleus (Bal & Oertel, 2001). Of the known α subunits blocked by DTX, Kv1.1 activates at the lowest voltage levels, and is, therefore, the best candidate for the low voltage activated current. Nevertheless, the V1/2 in these auditory neurons is more negative than values of −32 to −37 mV measured from Kv1.1 expressed subunits in heterologous expression systems (Grissmer et al. 1994; Hopkins et al. 1994; Tempel & Hopkins, 1995). This could indicate that DTX-sensitive ion channels are composed of as yet unidentified α subunits other than, or in addition to, Kv1.1 that have lower voltages of activation. Alternatively, the difference in the voltage of activation could indicate that other modulatory elements, such as β subunits, or glycosylation and/or phosphorylation alter the voltage dependence of the Kv1.1 α subunits in these auditory neurons. The lack of an inactivating component suggests that Kv1.4 is not a prominent component of the channel.
The identity of the high voltage activated component of DTX-sensitive current is less certain than that of the low voltage component. One possibility is that the channel is composed of other subunits such as Kv1.2 in addition to Kv1.1. The half-maximal activation voltages found for Kv1.2 subunit homomultimers in heterologous expression systems are more depolarized than Kv1.1, but values from different reports span a wide voltage range from −27 to +27 mV (Grissmer et al. 1994; Hopkins et al. 1994). Moreover, native DTX-sensitive currents that could be predicted to be dominated by the Kv1.2 subunit based on high Kv1.2 expression have been shown to have half-maximal activation voltages of −42 mV (Southan & Robertson, 2000) and −22 mV (Bekkers & Delaney, 2001). Thus it is difficult to isolate the contribution of a particular subunit when the complete composition of the underlying channels is not known.
Role of Kv1.1 in adult spiral ganglion neurons
An intriguing possibility for the function of Kv1.1 in adult spiral ganglion neurons is to contribute to the adaptation of auditory nerve responses to sound. When presented with maintained tones, spiral ganglion neurons produce a characteristic response in which neural activity rises quickly to a peak and is followed by rapid adaptation (Kiang et al. 1965). Detailed analysis of this adaptation has revealed a variety of complexities, including separation into at least three components, named rapid, short-term, and long-term, with time constants of > 10 ms, 50–100 ms, and tens of seconds, respectively (For review, see Ruggero, 1992).
At present, adaptation in the peripheral auditory system is generally attributed to depletion of transmitter in sensory hair cells. Initial evidence for this was provided by studies of transmitter release from hair cells in goldfish (Furukawa & Matsuura, 1978). In that study, intracellular recordings from eight nerve fibres during sound stimulation showed a gradually decreasing postsynaptic response, without any changes in the characteristics of the microphonic potential. More recent evidence has been obtained by direct measurement of exocytosis in murine hair cells (Moser & Beutner, 2000). The authors found that membrane capacitance increased exponentially upon hair cell depolarization, but rapidly (within 20 ms) slowed to a constant rate that could be maintained for depolarizations up to 1 s in duration. The rapid phase of transmitter release, attributed to exocytosis of a readily releasable pool of vesicles, could produce the initial response to sound, and its decline could account for aspects of rapid adaptation (Moser & Beutner, 2000). It is important to emphasize, however, that a postsynaptic contribution to adaptation has not been ruled out, and may actually be required to account fully for properties such as the non-monotonic decrease in discharge probability in response to a maintained stimulus (Lutkenhoner & Smith, 1986). As discussed above, it is possible that the function of Kv1.1 persists into adulthood and serves the same function in spiral ganglion neurons that is also seen in central neurons (discussed in the next section), which is to provide a mechanism by which the responses of spiral ganglion neurons are terminated quickly in response to sustained stimuli or are appropriately discrete in response to brief stimuli. This may be especially important during the more sustained phase of hair cell exocytosis that could occur in response to maintained high frequency tones which would produce a tonic depolarization of inner hair cells.
With regard to the entire population of cells studied, adaptation has not been found to be uniform, but can vary depending upon a variety of factors, and can even be absent altogether (Furukawa & Ishii, 1967; Furukawa & Matsuura, 1978; Westerman & Smith, 1984; Rhode & Smith, 1985; Westerman & Smith, 1985; Yates et al. 1985; Lin, 1997; Mo & Davis, 1997a). In a previous study, we showed that APmax falls into a biphasic function with a skewed distribution of neurons that display a variety of APmax levels in response to the same stimulus (Mo & Davis, 1997a). It is possible that varying density of DTX-sensitive currents, among other factors, could account for this phenomenon. There is a low percentage of neurons within each frequency region, in both paraffin-embedded sections and tissue culture, that have substantially less antibody staining and presumably fewer K+ channel subunits (Adamson et al. 1999). This could produce the graded differences in action potential number that we see within the spiral ganglion.
Factors beyond the DTX-sensitive currents may also contribute to adaptation in these neurons. The non-monotonic APmax/voltage function shows a prominent decline at greater depolarizations which correlates with in vivo recordings in which the adaptation rate increases with stimulus intensity (Rhode & Smith, 1985). In addition to Na+ channel inactivation, this may also be due to other K+ channel types that activate at more depolarized levels. We know that spiral ganglion neurons also possess a wide variety of K+ channel types, such as calcium-activated K+ channels, (Adamson et al. 1999; Bowne-English & Davis, 1999) that could contribute to the adaptation we observe at higher voltage levels.
Comparison with central auditory neurons
The spiral ganglion is the first neural element in the complex auditory neural pathway. Axons of cells that comprise the ganglion split into ascending and descending branches that terminate in the anterior ventral, posterior ventral and dorsal cochlear nucleus (Ryugo, 1992), which marks the beginning of separate parallel neural pathways to higher brain centres, presumably mediating different functions (For review, see Oertel, 1999). Because timing is an essential element of auditory information, it is not surprising that many of the neurons in this system have specific features, including distinctive K+ channel distributions (Peusner et al. 1998; Grigg et al. 2000) that preserve temporal fidelity. For example, neurons thought to be responsible for forming a spatial map of low frequency sounds display the distinctive low voltage activated currents also observed in the current study (Manis & Marx, 1991; Reyes et al. 1994; Smith, 1995; Rathouz & Trussell, 1998).
In addition to the presence of a low threshold activated K+ current, described in the present report, Ih has also been observed in spiral ganglion neurons (Chen, 1997; Mo & Davis, 1997b). Both of these currents are active close to or at the resting potential of the cell and have the ability to speed up the response of the cell. The Ih in spiral ganglion neurons has the added complexity of showing a particularly wide range of half-maximal voltages that are subject to modulation by cAMP (Mo & Davis, 1997b). This, in combination with the low voltage activated K+ currents, may, among other things, fine-tune a cell's temporal processing capabilities. We have found, therefore, that the biophysical properties necessary for carrying out the complex tasks of sound localization and temporal acuity are found in primary as well as in higher order neurons. The goal for the future is to determine with certainty whether the properties observed herein are developmental in nature, perhaps contributing to activity-dependent development of neural pathways (Gamkrelidze et al. 1998; Bekkers & Delaney, 2001; Beutner & Moser, 2001) or whether they persist into adulthood and contribute to signalling in the mature auditory system.
We thank Dr Mark R. Plummer for helpful discussions and for critically reading an earlier version of the manuscript and Dr Kewa Mou for expert technical support. This work was supported by the NIH (NIDCD01856).