Abstract
Neural response properties that typify primary sensory afferents are critical to fully appreciate because they establish and, ultimately represent, the fundamental coding design used for higher-level processing. Studies illuminating the center-surround receptive fields of retinal ganglion cells, for example, were ground-breaking because they determined the foundation of visual form detection. For the auditory system, a basic organizing principle of the spiral ganglion afferents is their extensive electrophysiological heterogeneity establishing diverse intrinsic firing properties in neurons throughout the spiral ganglion. Moreover, these neurons display an impressively large array of neurotransmitter receptor types that are responsive to efferent feedback. Thus, electrophysiological diversity and its neuromodulation are a fundamental encoding mechanism contributed by the primary afferents in the auditory system. To place these features into context, we evaluated the effects of hyperpolarization and cAMP on threshold level as indicators of overall afferent responsiveness in CBA/CaJ mice of either sex. Hyperpolarization modified threshold gradients such that distinct voltage protocols could shift the relationship between sensitivity and stimulus input to reshape resolution. This resulted in an “accordion effect” that appeared to stretch, compress, or maintain responsivity across the gradient of afferent thresholds. cAMP targeted threshold and kinetic shifts to rapidly adapting neurons, thus revealing multiple cochleotopic properties that could potentially be independently regulated. These examples of dynamic heterogeneity in primary auditory afferents not only have the capacity to shift the range, sensitivity, and resolution, but to do so in a coordinated manner that appears to orchestrate changes with a seemingly unlimited repertoire.
SIGNIFICANCE STATEMENT How do we discriminate the more nuanced qualities of the sound around us? Beyond the basics of pitch and loudness, aspects, such as pattern, distance, velocity, and location, are all attributes that must be used to encode acoustic sensations effectively. While higher-level processing is required for perception, it would not be unexpected if the primary auditory afferents optimized receptor input to expedite neural encoding. The findings reported herein are consistent with this design. Neuromodulation compressed, expanded, shifted, or realigned intrinsic electrophysiological heterogeneity to alter neuronal responses selectively and dynamically. This suggests that diverse spiral ganglion phenotypes provide a rich substrate to support an almost limitless array of coding strategies within the first neural element of the auditory pathway.
Introduction
Neurons in different areas of the brain can be distinguished from one another by their unique electrophysiological characteristics (Bean, 2007). These distinct firing patterns are not an epiphenomenon but instead represent different information processing systems for individual brain nuclei (Manis and Campagnola, 2018). Electrophysiological diversity, however, is not only found between neuron categories, it is also observed within them. While it is becoming more evident that similar cell types also differ electrophysiologically from one another, a uniform class of primary auditory afferents exemplifies this behavior (Mo and Davis, 1997a,b). Type I spiral ganglion neurons (SGNs), with their simple bipolar or pseudomonopolar morphology and one-to-one peripheral synaptic connections (Spoendlin, 1973), are a model of intrinsic electrophysiological heterogeneity. When evaluated in isolation, these neurons can be grouped into three separate categories based upon in vitro electrophysiology (Crozier and Davis, 2014) and transcriptomic profiling of developing and adult ganglia (Petitpre et al., 2018; Shrestha et al., 2018; Sun et al., 2018), similar to those identified in vivo (Liberman, 1978). Yet their complexity extends beyond the three subclasses because neurons within each group are not all electrophysiologically identical (Davis and Crozier, 2016).
The existence of such a large variety of spiral ganglion phenotypes suggests that neurons encode as a population. While this organization could represent an extensive and complete afferent coding mechanism, invariant neuron properties would effectively lock the intrinsic electrophysiological characteristics of these neurons into a static profile. This scenario, consistent with enhancing the system's ability to encode a wide range of input, would necessarily limit resolution, since there is a restricted number of afferent neurons that can be used for this purpose. Yet the auditory system has an extensive coding capacity relative to both dynamic range and resolution. The intensity submodality, for example, mediates the processing of sounds as soft as a whisper to those as loud as the roar of a jet engine, while still possessing the fine resolution to discriminate the smallest of intensity differences that underlie the capacity to localize high-frequency sounds in space with precision (Allen and Ison, 2010).
SGNs with their complex intrinsic properties reside at the intersection of sensory receptor glutamatergic input and efferent regulatory output. A wealth of neurotransmitters and neuromodulators identified within the lateral olivocochlear efferents, including acetylcholine, dopamine, GABA, and neuropeptides, innervate the postsynaptic membrane of Type I SGNs that possess corresponding neurotransmitter receptor subtypes (Sewell and Starr, 1991; Felix and Ehrenberger, 1992; Eybalin, 1993; Sewell, 2011; Reijntjes and Pyott, 2016). This complexity extends beyond the Type I neurons to include presynaptic innervation of inner hair cells and cholinergic medial olivocochlear efferent fibers (Maison et al., 2012). Not surprisingly, studies document diverse outcomes of lateral olivocochlear efferent modulation from upregulating or downregulating the firing behavior of auditory nerve fibers to balancing binaural excitability levels, upregulating TH, and providing neural protection (Niu and Canlon, 2002; Darrow et al., 2006; Maison et al., 2010, 2012; Le Prell et al., 2014a,b; Wu et al., 2020). Because the complexity of the system does not permit a clear view of the efferent impact on each individual element, we used an in vitro model to examine the effect of specific neuromodulatory elements directly on isolated Type I SGNs.
Herein we found that hyperpolarization and increased cAMP levels have defined and differential outcomes, which selectively target unitary, rapid, and slow adaptation (UA, RA, and SA, respectively) categories (Crozier and Davis, 2014). Most importantly, we found evidence that the broad heterogeneity in multiple electrophysiological parameters can be differentially delimited. These results demonstrate that the complex plasticity necessary to widen or narrow dynamic range, increase or decrease response resolution, shift individual electrophysiological parameters along the input current axis, and more, can all be accomplished by dynamically modifying the signature electrophysiological heterogeneity of SGNs.
Materials and Methods
Tissue culture
Cochleae were removed from postnatal age (P) 14-15 CBA/CaJ mice (The Jackson Laboratory) of either sex as previously described (Crozier and Davis, 2014). Briefly, basal, middle, and apical fifths of the spiral ganglion were isolated for tissue culture and plated as explants without exposure to enzymes. Explant tissues were positioned onto the center of poly-l-lysine-coated dishes and maintained in growth medium (DMEM, supplemented with 10% FBS, 4 mm L-glutamine, and 0.1% penicillin-streptomycin) for 13-27 DIV at 37°C in a humidified incubator with 95%/5% O2/CO2. Time in culture had little effect on the electrophysiological response properties of SGNs. Linear fits to key parameters (maximum number of action potentials fired to suprathreshold stimulation [APmax], APD threshold, resting membrane potential [RMP], action potential latency and duration) plotted as a function of DIV yielded low R2 values (R2 = 0.004, 0.0244, 0.0321, 0.0777, 0.0304, respectively), indicating that there was no systematic change over time. Procedures performed using CBA/CaJ mice were approved by the Rutgers University Institutional Review Board for the Use and Care of Animals, protocol 90-073.
Electrophysiology
Filamented borosilicate glass capillary tubes (Sutter Instruments, catalog #BF150-110-10) were pulled on a 2-stage vertical puller (Narishige, PP-83). Electrodes were coated with silicone-elastomer (Sylgard, Corning) and fire-polished (Narishige, MF-83) just before use. Electrode resistances ranged from 2 to 6 mΩ in the bath solution used for this study. Pipette offset current was zeroed just before contacting the cell membrane. Current-clamp measurements were made using the Ifast circuitry of the Axopatch 200A amplifier to reduce error currents (Magistretti et al., 1996, 1998). The internal solution was as follows (in mm): 112 KCl, 2 MgCl2, 0.1 CaCl2, 11 EGTA, 10 HEPES-KOH, 2 Na2-ATP, 0.3 Na3-GTP, and 20 phosphocreatine, pH 7.5. The bath solution (in mm) was 137 NaCl, 5 KCl, 1.7 CaCl2, 1 MgCl2, 17 glucose, 13 sucrose, and 10 HEPES, pH 7.5 (∼330 mOsm). A liquid junction potential of −5 mV was not corrected. Recordings were made from neuronal somata at room temperature (19°C-22°C). Although differential temperature-dependent changes were found for current magnitude and kinetics (Cao and Oertel, 2005), and could potentially alter cAMP actions, parameters that include voltage dependence, threshold, and adaptation were found to be temperature-insensitive under standard recording conditions (Cao and Oertel, 2005; Crumling and Saunders, 2005).
Because the cultures contain neuronal and non-neuronal cells, neurons were recognized by the presence of a large, transient inward sodium current in whole-cell voltage-clamp mode with a depolarizing voltage step from −60 to −30 mV that was blocked by 0.2 μm TTX in a previous study (Liu et al., 2014a). For current-clamp recordings, current steps of 240 ms duration were elicited every 5 s in 1-10 pA increments. Parameters assessed include action potential latency and duration, voltage threshold, and adaptation, also referred to as APmax, which is measured as the maximum number of action potentials fired in response to suprathreshold current injections. Voltage threshold was measured as the peak amplitude of the just-subthreshold response using 1 pA current injection increments. A holding potential of −60 mV was chosen to assess responses, unless otherwise noted. Data were digitized at 10 kHz with a CED Power 1401 interface using a personal computer and filtered at 5 kHz; the programs for data acquisition and analysis (written by M.R.P.). Acceptable current-clamp recordings met the following criteria: stable membrane potentials, discernible membrane time constant on step current injection, and overshooting initial action potentials (magnitudes of ≥70 mV from baseline); any changes to these parameters during the course of an experiment rendered subsequent activity unacceptable for analysis. The RMP for each neuron was measured as the voltage level recorded at the outset of the experiment in current-clamp mode before any current had been injected. A previous study showed that this protocol generated robust data that, while slightly depolarized, reproduced noninvasive, single-channel assessments (Liu et al., 2014a).
8-Bromo-cAMP, sodium salt (8-Br-cAMP, 100 μm) (Tocris Bioscience) was prepared as 1000× stock solution using sterile water. Cultures were pretreated with either 8-Br-cAMP or vehicle control (2 µl of sterile distilled water) added to the media 1 h before recording. The same concentration of each reagent was added to the external recording solution and the internal solution at the same concentration as the pretreatment condition. All recordings were obtained from multiple platings and animals.
Stimulus parameters
Square-pulse stimuli used to assess threshold responsiveness to direct depolarization (APDϴ) and rebound hyperpolarization (APRϴ) were 240 ms in duration. Biphasic stimuli to evoke rebound excitation (APREϴ) were 160 ms in duration in the hyperpolarizing direction, immediately followed by 80 ms in the depolarizing direction. Oscillating stimuli (41.7 or 83.3 Hz) delivered from a holding potential of −60 mV were used to determine lowest threshold response frequencies, from which we evaluated sinusoidal APREϴ.
Statistical analyses
Comparisons between the three adaptation groups (UA, RA, and SA) and three tonotopic regions (base, middle, and apex) were made with Levene's test to assess the equality of variances followed by a one-way ANOVA with either a Tukey–Kramer or a Games–Howell post hoc test for p > 0.05 or p < 0.05, respectively. To compare between the two experimental conditions (vehicle control and cAMP), the F test was used to assess the equality of the variances. If F test p values were <0.05, then a two-tailed unequal variances t test was used, which is the preferred statistical analysis whether a heterogeneous dataset was normally or non-normally distributed (Zimmerman, 1998). If F test p values were >0.05, then a two-tailed equal variances t test was used. Significant differences at p values of <0.05 (*) and <0.01 (**) were noted, unless otherwise indicated. The data are displayed as mean ± SEM as indicated.
Results
Characterizations of electrophysiological heterogeneity necessitate the evaluation of many neurons isolated from the spiral ganglion to assess the full spectrum of their endogenous membrane properties. For that reason, we chose to examine murine SGNs at P14-P15, immediately after adult hearing is attained (Mikaelian and Ruben, 1965), yet at an age still amenable for isolating and patch clamping many healthy neurons for extensive analysis. This timing follows an active phase of spiral ganglion development between P0 and the onset of hearing (Coate et al., 2019), which was observed at P10 in mice (Ehret, 1976). Neuronal development proceeds from base to apex in the ganglion (Rubel and Fritzsch, 2002), a property that is also reflected in the maturation of intrinsic membrane properties. Basal neurons mature faster as evidenced by changes in APmax, threshold, and kinetic properties that plateau by P14 (Crozier and Davis, 2014). Apical neurons show comparable changes; however, they do so with a delayed time course, indicating that they likely reach full maturity at or slightly after P14. Thus, the basic Type I intrinsic membrane properties analyzed herein mature within an early developmentally active time window compared with subsequent refinements that continue into the fourth postnatal week, such as synaptic interactions, spontaneous rate, myelination, and some gene expression profiles (Kim and Rutherford, 2016; Liberman and Liberman, 2016; Wu et al., 2016; Frank and Goodrich, 2018; Sun et al., 2018). In sum, this study includes recordings from a total of 150 neurons: 90 in vehicle control and 60 in 8-Br-cAMP-supplemented conditions.
Three adaptation categories: baseline
We first characterized the properties of the neuronal population maintained under baseline conditions and found that, consistent with previous electrophysiological findings (Crozier and Davis, 2014), three classes of Type I SGNs with distinct adaptation patterns were dependably observed in response to graded intensity, 240 ms constant current injections. Examples of UA, RA, and SA neurons are shown in Figure 1A1, 1A2, and 1A3, respectively.
Action potential amplitude decay rate differed significantly between the three distinct adaptation classes of SGN. A1-A3, stacked sweeps from UA, RA, and SA neurons. Traces from each experiment were aligned vertically from low to high current amplitude (step current levels in pA: A1, 270, 480, 700, 900; A2, 185, 305, 365, 395; A3, 246, 286, 546, 956). The constant current injection time course is shown below each stack of sweeps. Holding potential = −60 mV; calibration bar lower right (50 ms, 40 mV) applies to all voltage traces. B1, Action potential (AP) amplitude decay rate (m) was significantly different between RA and SA neurons when assessed for the second to last action potential fits (m2), but not for the first and second action potential fits (m1). B2, Overlapping sweeps and linear fits from 42 RA neurons. B3, Overlapping sweeps and linear fits from 14 SA neurons. C1-C3, Input-output functions for UA, RA, and SA neurons, respectively.
It should be emphasized that, because neurons were isolated from their peripheral and central targets in vitro, the firing features shown in Figure 1 and throughout the current study are purely intrinsic responses to constant current injections. This indicates that, although acoustic stimuli and the resultant sensory receptor synaptic input are phasic in nature, accommodation is an integral part of the response pattens of SGNs. Comparisons between the three classes showed that a feature they all have in common is their initial fast, overshooting action potential (Fig. 1A). It is the rate of decay of subsequent action potentials, however, that distinguishes between the UA, RA, and SA categories (Fig. 1A1, 1A2, and 1A3, respectively). The first to second action potential decay (black line) was assessed separately from fits of the second through last action potential decay (red line). Example fits for individual RA and SA neurons (Fig. 1A2 and 1A3, respectively) and all RA and SA examples (Fig. 1B2 and 1B3, respectively) were made at the current injection level that produced the maximum number of action potentials (APmax). A statistical comparison for each category showed that action potential decay rate for the first to second action potential (m1) was not significantly different between SA and RA classes (Fig. 1B1), indicating that a similar mechanism controls this initial firing period. The second through last action potential decay rates (m2), however, differed significantly between adaptation categories (Fig. 1B1), signifying differential regulatory mechanisms or proportions of contributing ion channels. That individual fits in Figure 1B2 and 1B3 were not identical from neuron to neuron is an example of the diversity observed within each adaptation category.
The overall impact of this electrophysiological behavior ultimately has an important influence on the input-output functions that typify each of the three categories. UA neurons are essentially homogeneous in this regard because they fire only a single action potential in response to prolonged depolarizing current injection (Fig. 1C1). In contrast, SA neurons can fire a wide range of action potentials at APmax, resulting in the greatest input-output heterogeneity of the three categories (Fig. 1C3). The RA neurons, which fire >1 and <7 action potentials herein, similar to previous studies (Mo and Davis, 1997a; Crozier and Davis, 2014), display intermediate input-output heterogeneity because of the limitation imposed by enhanced action potential amplitude decay (Fig. 1C2, middle). A comparison of each SGN adaptation category (Fig. 1C1–C3) shows that these groups are not uniform as can be readily observed from the input-output functions of RA and SA neurons (Fig. 1C2,C3), indicating that the underlying firing rates also differ from neuron to neuron.
Intrinsic electrophysiological heterogeneity within the three adaptation categories: baseline
In order to characterize specific aspects of baseline heterogeneous firing features within the Type I adaptation categories before assessments with neuromodulators, we quantitated RA and SA instantaneous (i) and average (a) interspike intervals (ISI) at APmax (Fig. 2A, double-headed arrows) for the control dataset. While the SA neurons make up the smallest percentage of the total population (14 of 85 = 16.5%), their kinetic heterogeneity is striking because of their activity profile (Fig. 2A, light gray traces): APmax ranged from 13 to 42; instantaneous ISI values (ISI(i)) ranged from 4.6 to 16.2 ms (Fig. 2B, dark gray diamonds); ISI(a) ranged from 5.84 to 19.6 ms (Fig. 2B, light gray diamonds), the coefficient of variance (CV) for each was 42% and 38%, respectively. Not as evident in RA neurons (42 of 85 = 49.4%), because of their smaller range of APmax values (2-6), was the wide range of firing rates: ISI(i) ranged from 5.3 to 20.4 ms (Fig. 2B, dark gray triangles); average ISI values (ISI(a) ranged from 6.5 to 24.2 ms (Fig. 2B, light gray triangles), the CV for both was 31%. The ISIs of RA neurons were overlapping with and not significantly different from SA neurons (ISI(i) RA mean = 9.6 ± 0.46 ms; ISI(i) SA mean = 8.2 ± 0.92 ms; ISI(a) RA mean = 10.5 ± 0.51 ms; ISI(a) SA mean = 11.1 ± 1.13 ms), consistent with the distribution of action potentials in RA neurons close to stimulus onset (Fig. 2A, dark gray traces). These firing patterns emphasize our initial observations that action potential decay rate, rather than firing rate, is a determining factor that distinguishes between the RA and SA categories under control conditions. One might imagine that, by merely changing the time course of the voltage-dependent ion currents responsible for decreasing the action potential amplitude, some of the RA and SA neurons may switch between categories. A previous study has shown that holding potential changes from −80 to −60 mV, permit a subgroup of P10 neurons to switch from UA to RA and others to switch from RA to SA (Crozier and Davis, 2014). Moreover, analysis of P10-P14 SGNs held at −60 mV showed proportions of UA, RA, and SA firing patterns (38.2%, 44.7%, and 17.1%, respectively), similar to that found herein, yet predictably different from those measured from cells held at −80 mV where >75% of the cells were UA. Consistent with the idea that kinetics are regulated separately, the ISI for each neuron remained unchanged, despite altering the APmax by depolarizing the holding potential relative to a stable step depolarization. This indicates that, for each SGN, firing rate remains stable at a particular step depolarization regardless of holding potential, whereas the number of action potentials fired varies. Primary sensory afferents displaying distinct firing pattern categories typifies the electrophysiological substrate for parallel processing, such as that observed in the somatosensory system (Abraira and Ginty, 2013). Auditory afferents not only possess this feature but, depending on holding potential, can shift from one class to another, similar to that observed in neurons at higher processing levels (McCormick and Pape, 1990a,b).
Heterogeneous firing rates at APmax overlap for SA and RA neurons. A, SA and RA neurons with similar ISIs were overlaid in 5 different pairs and arranged vertically by firing rate. Double-headed arrows indicate the time intervals used to calculate the ISI(a) ([t last AP − t first AP]/[number of APs −1]); and the ISI(i) (t second AP – t first AP). Bottom right, Calibration (50 ms, 40 mV) applies to all voltage traces. B, ISI(i) and ISI(a) plotted as functions of APmax for 42 RA and 14 SA neurons; 29 UA neurons fired a single spike only. SA ISI(a) and ISI(i) functions were fitted with exponentials having R2 values of 0.95 and 0.86, respectively.
The intrinsic properties that regulate SGN sensitivity are also distinctly heterogeneous between and within the three adaptation categories. The parameters that determine the overall excitability of a neuron are twofold: the threshold level at which the action potential fires 50% of the time and the RMP that sets the voltage level at or more hyperpolarized than the neuron's threshold. A previous study showed that these two parameters were independently regulated and accounted for the broad intrinsic excitability noted in vitro (Liu et al., 2014a), which is consistent with the wide range of in vivo responses (Muller et al., 2005; Taberner and Liberman, 2005). For the P14 dataset shown herein, the overall endogenous threshold voltage levels to step depolarizations (APDϴ) range from those displaying high and intermediate threshold levels (Fig. 3A1,A2) to those with thresholds close to the holding potential of −60 mV (Fig. 3A3). Not surprisingly, we observed that some neurons having low threshold levels and high RMP levels fired spontaneously in vitro (Fig. 3B). Moreover, these neurons display obvious membrane oscillations at holding potential (Fig. 3B, arrows), likely resulting from the interplay between the low voltage-activated and Ca2+-activated currents that are prevalent in these neurons (Adamson et al., 2002b; Chen et al., 2011; Davis and Crozier, 2015).
Depolarization evoked action potential threshold levels (APDϴ) and RMP are heterogeneous yet show significant differences between adaptation categories. A1-3, Superimposed increasing step depolarization traces from UA, RA, and SA neurons, respectively, were used to measure APDϴ. The peak voltage of the highest subthreshold trace (gray) is the most accurate current-clamp measurement of threshold voltage (dashed line). Bottom right, Calibration (5 ms, 20 mV) applies to all traces. B, Ten sequential sweeps recorded from the spontaneously active SA neuron in A3 held at −60 mV. Arrows indicate examples of membrane oscillations. Bottom left, Calibration (40 ms, 30 mV) applies to all sweeps. C, APDϴ voltage and RMP plotted as a function of APmax for each adaptation category. D1, D2, Significant differences are found between UA, RA, and SA neurons for APDϴ and RMP, respectively.
Threshold and RMP levels are heterogeneous within each group (Fig. 3C), yet the overlapping ranges do not preclude significant differences between the categories. Threshold levels for UA, RA, and SA neurons were −42.24 ± 0.88 mV, n = 29; −48.47 ± 0.54 mV, n = 42; and −51.76 ± 0.63 mV, n = 14, respectively (Fig. 3D1); RMPs for UA, RA, and SA neurons were −66.96 ± 0.37 mV, n = 29; −66.04 ± 0.32 mV, n = 42; and −63.78 ± 0.80 mV, n = 14, respectively (Fig. 3D2). Thus, the limited firing of the UA neurons is consistent with the highest voltage threshold levels and lowest RMPs, whereas SA neurons with their slowly adapting firing profile are most excitable having the lowest thresholds and highest RMPs. The differences are notable considering that in combination these parameters ultimately determine the responsivity of each afferent category.
Neuromodulation of SGN heterogeneity: hyperpolarization
Depolarizing current injections that evoke action potentials (APD) are not the only stimulus type that activate the spiral ganglion. As described for many kinds of neurons (Getting, 1989; Grenier et al., 1998; Felix et al., 2011), SGNs display postinhibitory rebound action potentials (APR) activated at the offset of hyperpolarizing stimuli. The consistency of this feature in SGNs at −60 mV holding potential allows assessments of APR to determine its voltage threshold and the current required to activate it compared with APD for each adaptation category. Similar to previous analysis, small-amplitude gradations of constant current step potentials were used to characterize action potential thresholds, yet in this case, action potentials were also evoked by hyperpolarizing stimuli (Fig. 4A1–A3, asterisks).
SGNs responded at the offset of hyperpolarizations at levels linearly related to their depolarization-evoked onset responses. A1-A3, Superimposed step depolarization and hyperpolarization traces used to assess APD and APR thresholds from UA, RA, and SA neurons, respectively. Gray represents APmax and subthreshold sweeps. Calibration bars and time course of current injection shown in the top and bottom right in A3, respectively, apply to all panels. Input-output functions for each recording are below each set of superimposed sweeps. Double-headed red arrows indicate the difference in current required to generate an APD and APR threshold response. Blue asterisks indicate the APRϴ trace and its location within the plot below. Red symbols represent the recordings for reference in B-D. B, APD and APR voltage thresholds were both graded by adaptation. Values for individual neurons in each adaptation group (n = 23 UA, 38 RA, 13 SA) were connected by a solid black line. Lower voltage thresholds for APR were consistently observed compared with APD; this difference was greater for the highest threshold UA neurons compared with the lowest threshold SA neurons. Inset, The difference between APD and APR threshold voltages was significantly larger for UA neurons compared with RA and SA neurons (p < 0.01). C, APR plotted as a function of APD threshold voltage. A single linear fit (black line, R2 = 0.86) deviated from unity (gray dashed line) toward lower voltage threshold levels for APR compared with APD. The RA/SA category indicates measurements made from neurons that fired >1 action potential yet were not evaluated for APmax. D, APR plotted as a function of APD threshold current. Two linear functions were fitted to the data. The low threshold fit (m = 0.94; R2 = 0.89) had a slope close to unity (gray dashed line); the high threshold fit (m = 1.57; R2 = 0.69) deviated toward higher current levels for APR. Values derived from A1-A3, are shown as red symbols in B–D. Legend in C applies to C, D.
This analysis showed that the least sensitive, high APDϴ UA neurons (Fig. 4A1) required greater currents to evoke an APR than the most sensitive, low APDϴ SA neurons (Fig. 4A3), with the RA neurons in between (Fig. 4A2). Action potential number plotted against current injection amplitude illustrated that the gap between the onset and offset responses was quite wide in the UA example (Fig. 4A1, red double arrow). Indicative of the UA category, the high threshold example neuron was relatively silent, firing action potentials only to stimuli of the highest input current magnitudes (306 and −345 pA, respectively) reaching the highest thresholds (−36.9 and −44.96 mV, respectively). In contrast to this, the most sensitive SA neurons were essentially active at most current injection levels, with only a small window of current in which they were silent or less active (Fig. 4A3, red arrowheads). Moreover, their APR and APD thresholds were relatively similar as were the low currents required to stimulate them, exemplified by the SA neuron shown in Figure 4A3 (−54.0 vs −52.0 mV and 49 vs −39 pA, respectively). This highly sensitive response pattern indicates that depolarizing input likely shapes ongoing activity, rather than initiates it. In contrast, hyperpolarization eliminates the firing of spontaneously active neurons yet subsequently activates a dependable action potential response at its termination, effectively choregraphing an abrupt and unambiguous offset response. SGNs with intermediate APD and APR threshold levels and response properties were also observed, filling in the range of voltage sensitivity (Fig. 4A2).
Whether neurons were of the UA, RA, or SA category, the APR was consistently found at lower voltages compared with the APD, indicating that responses at the offset of hyperpolarizations had greater sensitivity than their depolarizing counterparts (Fig. 4B). The difference between APD and APR threshold voltages was larger for UA neurons compared with RA and SA neurons (Fig. 4B, inset). The progressive increase for neurons with higher threshold levels was also observed by larger deviations from unity as threshold increased (Fig. 4C, dashed line). There was, however, an inverse relationship for current. While APD and APR current levels closely followed the line of unity for low threshold neurons, higher threshold neurons required more current to respond at the offset of a hyperpolarization (APR) than to the onset of a depolarization (Fig. 4D). Enhanced heterogeneity was also observed at these higher current levels, reflected in the higher CV for APR (67%) compared with the CV at lower levels (42%). The CV for APD was similar for high and low current levels (44% and 38%, respectively). Overall, despite prominent heterogeneity, the decidedly linear relationship between APD and APR is striking whether at high or low current levels (R2 = 0.658 and R2 = 0.886, respectively).
The greater APR voltage sensitivity compared with that of APD opens the question of whether rebound excitation (RE), a process by which hyperpolarizing prepulses reduce the action potential threshold to a depolarizing stimulus (APRE), is also an intrinsic feature of these neurons. To evaluate this possibility, we administered biphasic constant current square pulses and assessed the resulting thresholds. As expected for this mechanism (Mitra and Miller, 2007; Goaillard et al., 2010), we noted that action potential threshold was indeed reduced when depolarizing square pulses were preceded by hyperpolarizations (Fig. 5A1, black and blue square pulses and associated traces, respectively). The average threshold current for neurons stimulated with a prepulse was significantly lower than those stimulated without a prepulse: 39.2 ± 4.5 pA, n = 38 versus 111.9 ± 9.8 pA, n = 65, respectively (p < 0.01; Fig. 5B). Similarly, an abbreviated hyperpolarizing sinusoidal stimulus at 41.7 Hz was also effective in lowering threshold. Compared with a sinusoid initiated with a depolarizing phase, those initiated with a hyperpolarization enhanced action potential sensitivity (Fig. 5A2), indicating that APRE does not require a long duration hyperpolarization to affect a significant current difference (p < 0.01; Fig. 5B). The relationship between APD and APRE current thresholds initiated with both a square pulse and sinusoid was linear (Fig. 5C), suggesting that the systematic sensitivity shifts within the population of neurons observed for APD and APR are also maintained for APRE. The slopes of the APD versus APRE sinusoidal functions compared with APD versus APRE double square pulses did differ, however (Fig. 5C, gray and black lines, respectively). Although threshold voltage levels were not significantly different (Fig. 5D,E), the current level required to activate the slower onset time course sinusoidal stimulus APRE was higher than for the rapid onset double square pulse APRE. This is consistent with the idea that input current kinetic parameters directly affect neural responsiveness as assessed by threshold current level. Voltage threshold levels for double square pulse hyperpolarizing biphasic stimuli were significantly lowered from the baseline condition (Fig. 5D; p < 0.01). Both square pulse and sinusoidal rebound excitation stimuli demonstrated overlapping linear relationships when plotted as a function of APD (Fig. 5E).
Hyperpolarizing-leading biphasic square pulses and sinusoids lowered action potential threshold by rebound excitation (APRE). A1, A2, Example recordings (top traces) from an individual neuron in response to double square pulses and sinusoids, respectively; with (blue) and without (black) a leading hyperpolarization. Red dotted lines indicate APD threshold. Red dashed lines indicate APRE threshold. B, Average APRE threshold current levels, for both square pulses and sinusoids, were significantly lower than APD values; APRE square pulses, however, showed the greatest change. C, APD current threshold levels plotted as a function of APRE square pulses and sinusoids showed nonoverlapping linear relationships (m = 0.4, R2 = 0.87; m = 0.6, R2 = 0.96, respectively). D, Average APRE threshold voltage levels, for square pulses and sinusoids, were significantly lower than APD values; no significant difference was found, however, between APRE square pulses and sinusoids. E, APD voltage thresholds plotted as a function of APRE square pulses and sinusoids showed essentially overlapping linear relationships (m = 0.79, R2 = 0.79; m = 0.86, R2 = 0.79, respectively). Both linear fits vary from unity (black dashed line) at depolarized threshold levels.
Threshold voltage plotted as a function of input current for the full population of neurons shows that the relationships between voltage and current are uniquely different between APD, APR, and double square pulse APRE. Compared with APD (Fig. 6A, black circles and fitted line), both voltage threshold and input current for double square pulse APRE are compressed, resulting in a steeper sloping function (Fig. 6A, gray circles and fitted line). In contrast, responses at the offset of hyperpolarizations, plotted against the absolute values of input current for ease of comparison, demonstrate expanded current levels along with lowered threshold voltages (Fig. 6A, white circles, dashed fitted line), resulting in a shallow sloping function for APR. The lowered thresholds displayed by APRE and APR result in functions having different slopes because of the starkly different input current range. Significantly, the graded order of this parameter relative to current level is retained despite different underlying mechanisms controlling the APRE and APR functions (Fig. 6A). Thus, voltage modulation of heterogeneity acted across adaptation categories by preserving graded responses among all neurons while simultaneously altering resolution as defined by the slope of the voltage-to-current relationship.
Hyperpolarization voltage protocol impacts the relationship between input current and threshold voltage. A, Threshold voltage plotted as a function of input current for APD, APRE (double square pulse), and absolute current values for APR show different resolution, based on the slopes from linear fits: black (m = 0.595, R2 = 0.74, n = 90), gray (m = 0.1198, R2 = 0.795, n = 38), and dashed (m = 0.0222, R2 = 0.599, n = 77) lines, respectively. B, Threshold sweeps from a single neuron when holding potential is hyperpolarized from −60 to −80 mV in 5 mV increments. Inset, dV/dt (mV/ms) plotted as a function of voltage for each high-resolution action potential stimulated with 24 ms square pulses. C1-C3, Overlapping sweeps from low, middle, and high threshold neurons, respectively. Bottom right, Calibration bars apply to all sweeps. Symbols represent the time where measurements were made. D, Voltage to current relationships at −60 mV holding potential for the three recordings shown in C1-C3. Fits were made to the linear regions (thick lines) of the function that extended from just below threshold, through holding potential to ∼−70 mV. The resulting conductance values were 57.7, 72.7, and 95.7 mΩ for the high, middle, and low threshold neurons, respectively. E, Measurements from the three example neurons of APD at −60 to −80 mV holding potentials (colored diamond symbols) are superimposed on average linear fits of APD, APR, and APRE population data replotted from A (gray lines). Colored lines fitted to a larger dataset ranging from 7 to 9 measurements from 9 different neurons had relatively stable slopes (red: −60 mV, m = 0.051, R2 = 0.77, n = 9; amber: −65 mV, m = 0.058, R2 = 0.77, n = 9; light green: −70 mV, m = 0.062, R2 = 0.77, n = 9; light blue: −75 mV, m = 0.062, R2 = 0.78, n = 8; purple: −80 mV, m = 0.061, R2 = 0.82, n = 7). To show the consistency in measurements for each category of response, APRE (gray triangles) and APR (gray squares) measurements shown for the three example neurons aligned closely with the population data replotted from A (top and bottom light gray lines, respectively). APD at −60 mV holding potential for the three neurons (red diamonds), similarly aligned with the population data replotted from A (middle light gray line).
To test whether APD thresholds were affected by prolonged hyperpolarizations, in a subset of experiments, we compared threshold responses at different holding potentials from −60 to −80 mV in 5 mV increments (Fig. 6B). Although a smaller population of neurons composed this dataset (n = 9) our measurements included neurons at low, middle, and high threshold levels (Fig. 6C1, C2, and C3, respectively), in which input resistance (Rin) measurements at −60 mV could be obtained (Fig. 6D). Three experimental examples (Fig. 6C,D) superimposed on group data fits of APD, APR, and APRE measurements (Fig. 6A, light gray lines) systematically exhibited increased threshold sensitivity as the holding potential was hyperpolarized (Fig. 6E; red, amber, green, blue, purple symbols represent three examples shown in Fig. 6C1-C3,D; red, amber, green, blue, purple lines fitted to the total of 9 recordings). APR and APRE values measured from the three example neurons fell on or close to the fitted lines from the population (Fig. 6E, dark gray squares and triangles, respectively). Greater input current level changes were observed at each hyperpolarized holding potential for the neuron displaying the highest APD threshold compared with neurons with low and intermediate threshold levels (Fig. 6E; compare slopes of three thin black dashed lines fitted to holding potential data). Again, we found that threshold sensitivity shifted uniformly as a population, while maintaining a significant level of heterogeneity. In contrast to APR and APRE, however, the relationship between current and voltage remained relatively constant for APD compared at each holding potential (Fig. 6E; red, amber, green, blue, purple fitted lines to n = 9 experiments; R2 ranged from 0.77-0.82 for −60, −65, −70, −75, −80 mV holding potentials). The relatively large shift in current to evoke APD at lower holding potentials in high threshold neurons compared with lower threshold neurons is correlated with graded input resistances. This simple mechanism results in greater sensitivity with slightly expanded current ranges without altering resolution.
Thus, the relationship between voltage and current at threshold paints a complex picture of neuronal voltage sensitivity and input current responsivity for each of the distinct hyperpolarizing voltage protocols. Three different scenarios were observed based on the relationship between voltage and current. The first, represented by APRE, exhibited the potential for increased coding resolution based on the steeper slope that resulted from the combination of increased threshold sensitivity and lower input current levels. The second example, APR, demonstrated decreased resolution because of the shallow slope of the function that resulted from the larger current levels required to reach lower voltage threshold levels. The third example, APD at −60 to −80 mV holding potentials, showed that the sensitivity of the neuronal population could be increased without significant alteration in resolution (relatively uniform slopes observed at each holding potential), because of the graded threshold changes relative to current input at low versus high thresholds.
Neuromodulation of SGN heterogeneity: cAMP
The differential regulation of spiral ganglion intrinsic electrophysiological heterogeneity by distinct hyperpolarizing voltage protocols showed a complex relationship between sensitivity, responsiveness, and resolution at threshold. But are other heterogeneous attributes, such as the separately regulated firing kinetics of the SGNs, also dynamically controlled by neuromodulators? In order to answer this question, we tested cAMP because it is a second messenger that is activated by a subset of dopaminergic receptors (D1, D5) that have been localized to the Type I postsynaptic membrane (Maison et al., 2012). In addition to having a long-term impact on spiral ganglion survival (Hegarty et al., 1997; Hansen et al., 2001), cAMP can exert relatively short-term electrophysiological changes via its direct and indirect effects on numerous ion channel types (DiFrancesco and Tortora, 1991; Dai et al., 2009). These actions can potentially compress neuronal responsivity and kinetics without necessarily changing threshold voltage levels.
To characterize the effects of cAMP, we first assessed neuronal Ih sag magnitude and RMP levels to determine that the second messenger produced the expected outcome of altering Ih channel activation resulting in depolarized RMPs. The sag magnitude in our recordings was quantified as the voltage difference between plateau responses compared with the nadir at −185 mV (Fig. 7A, gray and black double-headed arrows for cAMP and vehicle control, respectively). When assessed with current clamp, we noted a reduced Ih sag magnitude in the cAMP condition compared with vehicle control for each adaptation class (Fig. 7B). This result is consistent with the idea that in the cAMP condition more open Ih channels at relatively depolarized levels would leave fewer channels to open during the hyperpolarization, resulting in lower sag magnitudes compared with controls. The expected relative change in kinetics between the vehicle control and cAMP examples (Fig. 7A, black and gray arrowheads, respectively) was representative of the overall population of cells. For example, the onset time course of hyperpolarizing traces that reached a −185 mV nadir was significantly different between vehicle control and cAMP conditions at p < 0.01. Thus, at the same hyperpolarizing voltage range, time-dependent voltage changes were observed with faster kinetics for the cAMP experimental condition.
SGNs supplemented with 8-Br-cAMP displayed selective responses by electrophysiological parameter and adaptation class. A, Traces from two UA neurons used to assess Ih sag magnitude (hyperpolarization from −60 to −185 mV) for vehicle control (black) and 8-Br-cAMP (gray). Inset, RMP was defined herein as the voltage level recorded at the outset of the experiment in current-clamp mode before any current had been injected. Black traces represent control condition (asterisks). Gray traces represent 8-Br-cAMP supplemented condition. Arrowheads indicate the membrane kinetic differences between control (black) and cAMP (gray) conditions. B, Ih sag magnitude changed significantly between vehicle control and 8-Br-cAMP conditions for each adaptation class. C, RMP level was significantly elevated in the 8-Br-cAMP condition compared with vehicle control for each adaptation class. D, E, Graded average APD (D) and APR (E) threshold current level differences for adaptation classes was unchanged between experimental conditions. The distribution of current values, assessed with f-distribution statistics, was significantly compressed in the 8-Br-cAMP condition compared with vehicle control for RA neurons (gray line and #). F, Graded average APRE threshold current level differences for adaptation classes was unchanged between experimental conditions for UA and SA neurons. A significant difference was noted, however, for RA neurons (p < 0.05). G-I, Average voltage thresholds of APD (G), APR (H), and APRE (I) retained levels and graded differences by adaptation type between treatment groups. J, K, APD voltage threshold plotted as a function of RMP for control and cAMP conditions, respectively. Two different voltage magnitudes at 15 and 25 mV are shown as reference. L, Average input resistance (Rin) showed similar patterns of significance to APD, APR, and APRE threshold current levels shown in D-F.
The predictable changes in RMP were observed between vehicle control and cAMP conditions (Fig. 7A, asterisk) for all adaptation classes (Fig. 7C). Larger average shifts in RMP were noted for UA neurons from their relatively hyperpolarized initial RMPs. In contrast to the significant differences in Ih sag magnitude and RMP, the systematic differences in threshold voltage (Fig. 7D–F) and input current levels (Fig. 7G–I) for UA, RA, and SA neurons observed in vehicle controls remained essentially unchanged in the cAMP condition compared at the same −60 mV holding potential. Only RA neuronal current levels for APRE thresholds were significantly different between experimental conditions (Fig. 7F; p < 0.05). However, RA neurons also showed a significant difference between vehicle control and cAMP conditions in heterogeneity assessed by the F test for current levels measured at APD and APR, (Fig. 7D,E, gray bars and # symbol representing p < 0.05). These findings indicate that current levels are compressed or shifted in the cAMP condition compared with vehicle control, but only for RA neurons and without affecting voltage threshold. Thus, the diversity in neuron excitability is retained even when RMPs were shifted to more depolarized levels, as can be observed for APD in control (Fig. 7J) and cAMP (Fig. 7K) conditions. Stable voltage threshold levels indicate that KV1.1 and KV1.2 ion channels, which regulate threshold voltage level, are unaffected by cAMP. This finding is consistent with maintained APmax in the cAMP condition (Fig. 8A) and indicates that cAMP changes responsivity to current without altering threshold sensitivity. Expectedly, input resistance within the voltage range evaluated was only significantly different for RA neurons compared between control and cAMP conditions, substantiating their selective modulation by cAMP (Fig. 7L). This finding reveals a further distinction between the RA and SA neurons beyond APmax, threshold, and action potential decay rate, consistent with the differential neuromodulation that might be expected between parallel coding pathways.
Kinetic differences affected by 8-Br-cAMP were preferentially targeted to RA neurons. A, Action potential firing numbers within each APmax category did not change significantly from vehicle controls for RA and SA classes supplemented 8-Br-cAMP. Inset, Action potential latency was measured (in ms) between the onset of the step depolarization and the peak of the action potential. Action potential duration (in ms) was measured between data points located halfway between the peak and the nadir of the spike. B, Average values and graded latency trends between adaptation categories in the vehicle control condition were retained in the 8-Br-cAMP experimental group. The significant difference between control UA and SA neurons was not observed in the cAMP condition. C, Action potential duration was uniformly heterogeneous for each adaptation group in vehicle control condition. The average value (p = 0.037) and f distribution (p = 0.040) of this parameter changed significantly, but only for RA neurons in the 8-Br-cAMP experimental condition. D1, D2, Action potential duration plotted as a function of input current for vehicle control and cAMP conditions, respectively. Colored symbols represent measurements from traces shown in the inset. Inset, Traces from example recordings were normalized for action potential amplitude and aligned to the initial depolarizing upstroke of the spike to enable direct comparison of duration differences. D2, Double arrowhead indicates active onset kinetics noted at higher thresholds. E, ISI(i) reduced significantly for RA neurons in the 8-Br-cAMP condition compared with vehicle controls. F, ISI(i) plotted as a function of current for 42 RA recordings obtained in the vehicle control condition showed a wide range of responses. G, ISI(i) plotted as a function of current for 28 RA neurons in the 8-Br-cAMP condition displayed more compressed responses. H, ISI(i) did not differ significantly for SA neurons between the vehicle control and 8-Br-cAMP conditions. I, ISI(i) plotted as a function of current from 14 SA neuron recordings obtained in the vehicle control condition. Inset, Percentage of neurons in each adaptation category for the vehicle control condition. J, ISI(i) plotted as a function of current from 12 SA neuron recordings in the 8-Br-cAMP condition. Inset, Percentage of neurons in each adaptation category for the 8-Br-cAMP condition.
While the average APmax for each class of neuron was stable (Fig. 8A), other select changes in the kinetics were found between control and cAMP conditions. For this analysis, we compared measurements of action potential latency and duration at threshold (Fig. 8A, inset). We found that the systematic trend in latency between adaptation groups in vehicle control was retained in the cAMP condition, yet the significant difference between control UA and SA neurons was not (Fig. 8B). Action potential duration levels, on the other hand, that did not differ between adaptation groups in vehicle control, were significantly different between vehicle control and cAMP conditions, but again, only for RA neurons (p < 0.05; Fig. 8C). We noted that APR durations evoked from −185 mV for RA neurons also differed between experimental conditions (p < 0.05; data not shown). To examine this result in greater detail, we plotted action potential duration as a function of input current for the vehicle control (Fig. 8D1) and cAMP (Fig. 8D2) experimental groups. As expected, the longest action potential durations were found at the lowest input currents for the vehicle control group; however, the shortest durations were also found at similar current levels, resulting in the largest range at the lowest current levels (Fig. 8D1, colored symbols and traces). In contrast to this, the largest range of action potential durations at threshold was noted at the highest current inputs for the cAMP condition (Fig. 8D2, colored symbols and traces). Thus, action potential durations at threshold were reduced for RA neurons supplemented with cAMP while shifting their largest range of responses to higher current levels. The faster responses that are shifted to higher current levels are consistent with the significantly lower Rin of RA neurons, while the prominent onset currents (Fig. 8D2, double arrowhead) suggest a role of LVA Ca2+ channels in this process (Chen et al., 2011). These results indicate that, by selectively modifying Rin, the responses of a single adaptation group can be shifted both in current responsiveness and membrane kinetics. The broad application of cAMP to all adaptation classes having restricted effects on RA neurons further distinguished our electrophysiological categorizations of neuron classes and indicates that selective phenotypic differences between adaptation groups could convey specific effects without requiring targeted neuromodulation. As a neural encoding strategy, this apparent realignment of kinetic parameters with signal input for a subclass of neurons would focus the widest dynamic range of a particular parameter on a limited current range, ultimately amplifying resolution for a targeted input. Taken at face value, the observed shifts in kinetic heterogeneity from low to high input levels enables afferent processing power to focus on sensory signals of restricted intensity.
We followed up on the significantly different action potential durations between experimental groups for RA neurons by evaluating ISI(i) at APmax. We hypothesized that the time between action potentials could shorten should action potential duration measurements at threshold contribute to differential firing properties at suprathreshold levels. Not surprisingly, the group data showed that ISI(i) values were significantly abbreviated in the cAMP condition for RA neurons compared with vehicle controls (Fig. 8E). ISI(a) also differed between conditions, but only for RA neurons (p < 0.01, data not shown). ISI(i) plotted as a function of input current for RA neurons showed that the dataset was indeed compressed in the kinetic dimension for the cAMP condition (Fig. 8G; n = 28) compared with vehicle control (Fig. 8F; n = 42). SA neurons, by contrast, did not show significant differences between experimental conditions. While the APmax values were somewhat compressed for the slowest firing SA neurons between conditions (Fig. 8A), this observation was not significantly different when assessed with an F test or Student's t test. Nor did we find action potential durations at threshold (Fig. 8C), ISI(i) (Fig. 8H), or ISI(a) (p > 0.05; data not shown) measurements at APmax significantly different for SA neurons. Only minor changes were observed between vehicle control and cAMP conditions for ISI(i) versus input current functions (Fig. 8I, n = 14; and Fig. 8J, n = 12, respectively). Thus, RA neurons were selectively altered in the cAMP experimental condition, displaying significantly faster threshold and suprathreshold kinetics that ultimately increased firing rate.
The same dataset, reordered according to tonotopic location, did not show just a single kinetic organization, but instead revealed evidence of dual mechanisms underlying SGN input responsivity. Tonotopic comparisons were made between neurons derived from the base, middle, and apex of the cochlea. The dataset had evenly distributed averaged APmax values that were unchanged with cAMP (Fig. 9A), representing all adaptation categories across the cochlea in control (base: 17% SA, 50% RA, 33% UA, n = 24; middle: 14% SA, 43% RA, 43% UA, n = 35; and apex: 19% SA, 58% RA, 23% UA, n = 26) and cAMP conditions (base: 36% SA, 43% RA, 21% UA, n = 14; middle: 15% SA, 45% RA, 40% UA, n = 20; and apex: 17% SA, 54% RA, 29% UA, n = 24). Further, there were no significant differences in APD latency and duration between regions and conditions (Fig. 9B,C). Base, middle, and apex ISI(i) control (9.44 ± 0.89, 8.68 ± 0.63, and 9.44 ± 0.61 ms, respectively), ISI(i) cAMP (8.07 ± 0.55, 7.18 ± 0.47, and 7.88 ± 0.37 ms, respectively), ISI(a) control (10.70 ± 0.93, 10.61 ± 0.78, and 10.63 ± 0.64 ms, respectively), and ISI(a) cAMP (10.10 ± 0.67, 8.42 ± 0.67, and 9.32 ± 0.39 ms, respectively) were not significantly different regionally. When compared across conditions only, the apical ISI(i) differed between the control and cAMP conditions (p < 0.05); base and middle ISI(i) and all ISI(a) statistical comparisons between conditions were not significantly different (data not shown). In contrast, significant differences were found for both voltage and current assessments of Ih, showing a robust tonotopic gradation with the basal neurons requiring greater injected currents to reach −185 mV than middle and apical neurons (Fig. 9D) and having a significantly smaller sag magnitude for the base compared with the apex (Fig. 9E). This suggests that Ih current activity increased along the P14 cochlear contour from apex to base. While retaining heterogeneity, the significant differences in sag magnitude currents and voltages between the cAMP and control conditions eliminated the average differences between regions in the cAMP condition (Fig. 9D,E). As expected from adaptation comparisons, RMP levels were significantly depolarized in cAMP compared with the control condition (Fig. 9F). By comparison, the gradations and significant differences in Ih current parameters observed tonotopically were not observed across adaptation categories (Ih sag current for control SA = −752.07 ± 68.27 pA, RA = −740.76 ± 39.72 pA, UA = −771.08 ± 39.74 pA; cAMP SA = −978.82 ± 114.07 pA, RA = −1192.17 ± 57.41 pA, UA = −1204.67 ± 103.54 pA; p > 0.05). This indicates that averaged Ih activity is graded tonotopically, but not between adaptation groups.
SGNs supplemented with 8-Br-cAMP displayed selective responses by electrophysiological parameter and tonotopic location. A-C, APmax, action potential latency and duration, respectively, did not differ between region or condition. D, The current required to reach a constant negative potential from which comparisons between sag magnitude were made (−185 mV) was significantly different between regions in control, but not in the cAMP condition. E, Ih sag magnitude differed significantly between base and apex neurons in vehicle control and between vehicle control and 8-Br-cAMP conditions for each tonotopic region. F, RMP level was significantly elevated in the 8-Br-cAMP condition compared with vehicle control for each tonotopic region. G, H, APD and APR threshold current levels, respectively, differed only between base and apex neurons in the vehicle control condition. I, APRE threshold current was not significantly different between regions or conditions. J-L, APD, APR, and APRE threshold voltage levels, respectively, did not differ between regions or conditions. M, N, Linear fits to apex, middle, and base APR threshold voltages plotted as a function of input current showed slopes with a greater divergence between tonotopic regions in control than in the 8-Br-cAMP condition, respectively. The slopes (mV/pA) and R2 values are as follows: control: apex = 0.0413, 0.886, middle = 0.0319, 0.713, base = 0.0164, 0.561; cAMP: apex = 0.0328, 0.835, middle = 0.0243, 0.414, base = 0.0259, 0.767). O, Significant differences in Rin between tonotopic categories in the control condition are no longer present in the cAMP condition.
Also shown for adaptation categories (Fig. 7D,E), APD, and APR threshold current differed significantly according to cochlear location (Fig. 9G,H), suggesting the presence of dual neural responsivity gradients within the ganglion. This did not apply to APRE threshold current. The significant differences observed for APRE between adaptation categories (Fig. 7F) were not observed regionally (Fig. 9I), indicating that channel types in addition to or other than Ih correlate with this parameter. Overall, greater current levels were required to reach APD and APR threshold for basal neurons compared with apical ones (Fig. 9G,H). This was consistent with significant differences in regional Rin levels between control base and apex neurons (Fig. 9O; 83.18 ± 4.85 and 134.996 ± 11.58 mΩ, respectively; p < 0.01) but not between the base and apex neurons in the cAMP condition (90.55 ± 8.77 and 100.74 ± 10.47 mΩ, respectively), similar to findings across adaptation categories (Figs. 6D and 7L). What is dissimilar from the adaptation categorization, however, was that the gradations in APD and APR current threshold levels were not observed in cAMP (Fig. 9G,H), nor were there significant differences in voltage threshold at the −60 mV holding potential used herein (Fig. 9J–L). In combination, these observations are consistent with the idea that Ih currents play a role in the tonotopic regulation of APD and APR current thresholds. Consistent with this, cAMP has been shown to increase Ih activity in SGNs by shifting the voltage dependence of activation to depolarized levels (Mo and Davis, 1997b); moreover, the same pattern of RMP shifts, without altered voltage threshold, have been previously observed with an Ih pharmacological antagonist (Liu et al., 2014a). By contrast, LVA ion channel α-subunits not directly modulated by cAMP with the capacity to directly alter threshold and APmax simultaneously, such as KV1.1 and KV1.2, have the potential to regulate responsivity and sensitivity gradients associated with accommodation that potentially account for increased APRE resolution.
Despite the significant differences in averaged APD and APR threshold current levels, heterogeneity is, nevertheless, retained. Assessments in both control and cAMP conditions would be expected to show that this response range is shifted relative to voltage. Most clearly observed for the hyperpolarization-activated APR responses, the tonotopic substructure observed within the population (Fig. 9M) becomes essentially overlapping in the cAMP condition (Fig. 9N). Expectedly, Rin is no longer significantly different between adaptation categories in the cAMP condition (Fig. 9O). These results indicate that, separate from AP duration regulation by cAMP, which is associated with a single adaptation class, cAMP shifts threshold current for the full neuron population according to cochlear location. These different responses, in addition to the dual adaptation- and location-related responsivity gradations, indicate that the voltage-gated ion channels that regulate duration, RMP, adaptation, and voltage threshold have unique distributions within the ganglion to coordinate their distinct electrophysiological response properties.
To examine the significant differences observed tonotopically in greater detail, APD and APR current thresholds and Rin were separately evaluated in each location for UA, RA, and SA neurons in control and cAMP conditions (Fig. 10). This analysis sheds light on the dual adaptation and regional gradients postulated for these parameters. The first gradient between adaptation categories was significantly different for each electrophysiological parameter in the middle and apical regions for control conditions (Fig. 10A–C). By contrast, the second gradient extended across the tonotopic regions for all three parameters measured in control conditions, but only for the RA neurons (Fig. 10A–C). As noted previously, the tonotopic gradient displayed by RA neurons was eliminated in the cAMP condition, as was the significance in the middle for adaptation classes (Fig. 10D–F). Although the low percentage of SA recordings were limited in number when divided into multiple categories, a significant tonotopic difference for Rin emerged in this adaptation category for the cAMP condition (Fig. 10F), further distinguishing RA from SA neurons. What is common between each of the parameters, however, was that significant differences between adaptation classes were restricted to the apex and that the RA tonotopic gradient was abolished in the cAMP condition (Fig. 10D–F). Overall, these results show that the orthogonal gradients are separately established and differentially modulated, allowing different patterns of activity to emerge within the ganglion because of neuromodulation.
Dual gradients of APD and APR threshold current levels and Rin observed between adaptation classes and cochlear locations were altered by cAMP. A-C, For the control condition, significant tonotopic differences were restricted to RA neuron measurements, while significant differences between adaptation categories were consistently found in the middle and apex. D-F, In the cAMP condition, tonotopic differences displayed by RA neurons were eliminated and adaptation-specific differences were limited to the apex. Interestingly, a new tonotopic difference emerged for SA neurons, but only for Rin (F).
Summary of results
The dynamic compression, expansion, shifts, and realignment of intrinsic electrophysiological properties in response to hyperpolarization and increased cAMP reveal neuromodulation of electrophysiological heterogeneity as a hallmark of neural coding in the spiral ganglion. This mechanism depends upon both the types and distributions of endogenous voltage-gated ion channels within primary auditory afferents and their independent regulation by multiple neuromodulatory elements. While it was not possible in a single study to examine the full range of neuromodulators or to target every possible ion channel type within these neurons, our focused study, nevertheless, revealed something quite remarkable about the spiral ganglion. This small population of neurons has the capability of orchestrating countless responses considering how the vast array of neuromodulators present in the system can target myriad ion channel types differentially distributed within and among three different neuron classes.
Discussion
The contribution of complex electrophysiological regulation to peripheral sensory coding is not a unique observation in the auditory system. In reptiles, for example, frequency-specific electrical “ringing” in cell membranes resulting from the interplay between calcium channels and calcium-activated K+ channels has been observed in turtle cochlea hair cells (Art and Fettiplace, 1987; Art et al., 1995). While tonotopic variations remain (Fettiplace, 2017; Coate et al., 2019), electrical tuning is either missing or minor in inner hair cells of the mammalian cochlea (Marcotti et al., 2003; Johnson, 2015). Thus, it appears that these types of intrinsic electrophysiological specializations in mammals are first observed at the next stage of auditory processing: the primary afferents of the spiral ganglion.
These neurons are replete with endogenous membrane specializations that have been observed using multiple approaches (Adamson et al., 2002a,b; Langer et al., 2003; Chen et al., 2011; Flores-Otero and Davis, 2011; Crozier and Davis, 2014; Liu et al., 2014a,b; Petitpre et al., 2018; Shrestha et al., 2018; Sun et al., 2018). Thus, the many voltage-gated ion channel types that underlie complex firing behaviors observed between and within three adaptation categories are perfectly suited to serve as targets for the prolific array of postsynaptic efferents (Eybalin, 1993; Sewell, 2011; Reijntjes and Pyott, 2016). Moreover, there are multiple receptor types in the postsynaptic membrane that support the idea that fine control can potentially be exerted on spiral ganglion responses. Hyperpolarization, for example, can be directly generated with different time courses by ionotropic and metabotropic GABAA and GABAB receptors, respectively, and negative regulators of activity are effectuated by muscarinic, dopaminergic, and metabotropic glutamate receptors associated with the Gi α subunit (Huang and Thathiah, 2015; Reijntjes and Pyott, 2016). On the other hand, two classes of metabotropic dopaminergic receptors, D1 and D5, associated with Gs α subunits, are also present (Maison et al., 2012), which, when activated, elevate cAMP levels (Neve et al., 2004).
We first chose to examine one of the most basic functions of primary sensory afferents: their capacity to respond accurately and predictably to sensory stimuli of graded intensity. Rather than using action potential number, which differs in each adaptation category, we chose instead to evaluate neuronal threshold responses for each experimental condition. This approach permits assessment of the entire neuron population to determine how heterogeneity shapes the coding capability of the system. Results show that, by simply changing hyperpolarization protocol, the overall responsiveness, sensitivity, and the resulting resolution of the neuron population can be systematically modified. Compared with our baseline condition (Fig. 11A, APD, thick black line), the slope of this function can be increased (Fig. 11A, APRE, thick gray line), decreased (Fig. 11A, APR, thick blue-gray dashed line), or maintained (Fig. 11B, APD thick colored lines) depending on how individual neurons within the dataset contribute to the voltage-to-current relationship.
Diverse hyperpolarization protocols are powerful modulators of stimulus detection and resolution by shifting the relationship between current responsiveness and output sensitivity. A, Fitted functions from data in Figure 6A (APRE, APD, and APR, light gray, and black solid lines, and blue-gray dashed line, respectively) exemplify how voltage to current relationships can be shifted along the x and y axes (thin gray, black and blue-gray arrows) within a uniform population of neurons. The rebound excitation voltage protocol lowers APRE threshold voltage compared with APD, while requiring lower depolarizing current input to do so. The compression of both voltage and current heterogeneity of these parameters results in enhanced resolution (steeper slope). Thus, by limiting the voltage and current dynamic ranges for the same group of cells, neurons within a uniform population are endowed with enhanced responsivity, sensitivity, and resolution. APR in response to hyperpolarizations, on the other hand, show the lowest resolution. The compression of voltage sensitivity (y axis, dotted arrows) and expansion of current input responsiveness (x axis, dotted arrows) are responsible for changing the slope of their relationship. B, Fitted functions from Figure 6E progressively shift in sensitivity (y axis, −60 and −80 mV holding potential for red and navy blue, respectively) and responsiveness (y axis, colored arrows), while maintaining resolution (slope of the functions). Thus, amplifying dissimilarities of a specific parameter (i.e., APRϴ) permits greater accuracy in comparative discriminations, such as those required for sound localization, whereas enhanced resolution is classically used by the nervous system to improve the detection of stimulus form (i.e., APRE).
The complement of endogenous ion channels found among the neuron population provides a high degree of electrophysiological flexibility that can be modified to exhibit distinct encoding specializations. For example, lower currents required to reach APRE voltage threshold are consistent with the closure of inhibitory ion channels, such as KV1.1 and KV1.2, as they are active in the voltage range examined and found in greater densities in high threshold neurons. The shallow slope of the APR function, on the other hand, indicates the opening of excitatory ion channels, such as activation of Ih channels and noninactivation of Na+ and LVA Ca2+ channels (Davis and Crozier, 2015; Reijntjes and Pyott, 2016). This contrasts with the maintained slope observed with holding potential changes associated with input resistance shifts that are likely due to Ih channels in combination with the plethora of leak channels found within these neurons, some of which are weakly voltage dependent (Chen and Davis, 2006).
Yet, do these resolution changes serve solely to shape the responsiveness of the neuronal population, or can they also improve the detection of stimulus input? Since acoustic level and timing detection are important components of sound localization that can potentially be enhanced in initial neural processing pathways (Joris et al., 2006; Grothe et al., 2010; Pecka et al., 2010; Ashida and Carr, 2011), it is possible that both outcomes are critical aspects of sound processing by the spiral ganglion.
What is most striking about these observations is the implication that the full population of neurons is capable of dynamically contributing to coding broad levels of stimulus input. For example, high threshold neuron responses are likely not reserved only for loud sounds. Instead, their firing features can be shifted under appropriate conditions allowing them to respond to much lower input levels. This mechanism permits the full neuronal population to potentially be available to encode acoustic signals when shaped by neuromodulation, which in the case of APRE, enhances resolution as well as sensitivity and responsivity (Fig. 11A). This scenario is quite different from how coding occurs in the visual and somatosensory systems that are “hard-wired” to trade off high resolution for enhanced sensitivity in their receptive field circuitry design.
The alternate strategy, to dynamically regulate intrinsic electrophysiological heterogeneity, observed herein, requires that a larger number of primary afferents innervate cochlear regions with the greatest processing requirements. This condition is achieved by the increased number of neurons innervating receptors within broad mid-frequency regions (Meyer et al., 2009) that display the largest range of threshold sensitivities both in vivo (Muller et al., 2005; Taberner and Liberman, 2005) and in vitro (Liu and Davis, 2007). Yet rather than solely providing high resolution, which is usually paired with low sensitivity, dynamic heterogeneity by efferent modulation can enhance coding flexibility instead. This suggests that the auditory equivalent of the visual fovea is a dynamic construct that is created, rather than hard-wired. Moreover, this design essentially solves the trade-off between high resolution/low sensitivity versus low resolution/high sensitivity. As can be seen in Figure 11A, the high resolution due to the compressed responses observed for APRE is also highly sensitive. Instead, the trade-off appears to be dynamic range versus resolution for primary auditory afferents.
Modulation of the holding potential by prolonged hyperpolarization has a distinctly different effect on SGNs. In this case, neuronal voltage threshold is shifted to more sensitive ranges, yet the resolution of the system remains essentially constant (Fig. 11B). APD threshold levels at −60 mV holding potential are more elevated, or less sensitive than those recorded from the same neurons at more hyperpolarized levels. This pattern of responsivity is similar to individual retinal ganglion cell activity in the presence of differing levels of luminance (Sakmann and Creutzfeldt, 1969). If used for the same purpose in the auditory system, then one would predict an inverse relationship between holding potential and background noise levels.
The ability of spiral ganglion afferents to adapt their responsiveness and sensitivity to various types of hyperpolarizing modulation shows that they are capable of shaping stimulus representation, rather than merely relaying it. As revealed by cAMP neuromodulation, the underlying mechanism for this is dependent on a complex and dynamic electrophysiological organization. For example, input responsiveness was structured into two orthogonal maps: one represented by the RA category, which was graded tonotopically (Fig. 12A, dashed double arrow); and the other graded regionally between the three adaptation classes (Fig. 12A, dotted double arrows). The selective regulation of the RA tonotopic gradient by cAMP (Fig. 12B) indicates that the dual maps are differentially modulated and are therefore independently biophysically regulated. As other dual gradients have been observed morphologically and immunocytochemically in spiral ganglion presynaptic phenotypes (Flores-Otero and Davis, 2011; Nayagam et al., 2011), the orthogonal electrophysiological maps noted herein are likely exemplars of a more complex organization within the ganglion encompassing additional signaling attributes. Together, voltage and cAMP neuromodulation can reconfigure the electrophysiological heterogeneity of SGNs, dynamically switching between adaptation-, tonotopic-, and ganglion-specific signaling strategies. The adaptation-specific organization classically underlies parallel processing as exemplified by somatosensory mechanoreceptors (Abraira and Ginty, 2013). In contrast, the suborganization of electrophysiological heterogeneity within an adaptation class, such as the RA-specific tonotopic gradient and its specific modulation (Fig. 12), indicates the presence of flexibly organized encoding strategies within each adaptation category. And finally, the systematic shifts in threshold resolution across all neurons indicate that the primary afferents are uniquely capable of simultaneously conveying coordinated activity across the entire cochlear contour.
Patterns of input responsivity form multiple gradients in the spiral ganglion that were dynamically and selectively modulated. A, In the control condition, average APD input responsivity was graded as a function of adaptation category in the middle and apex regions (dotted double arrows). These electrophysiological features were also tonotopically graded, yet interestingly, this observation was limited to the RA neurons (dashed double arrow). The electrophysiological behavior summarized for APD was also observed for APR current thresholds, as was the expected reciprocal activity for Rin. B, Selective modulation by cAMP eliminated the RA tonotopic gradient (dashed double arrow) and restricted the significance of the local gradients between adaptation groups to the apex.
In conclusion, our findings show that dynamic heterogeneity within the spiral ganglion is the key to understanding the initial stages of auditory coding. The myriad voltage-gated ion channels and neurotransmitter receptors within the primary auditory afferents indicate that the inherent plasticity of the system is considerable. This mechanism results in compression and expansion of diverse firing patterns to shift kinetics, resolution, and dynamic range within and among adaptation-, tonotopic-, and ganglion-specific regions. Based on what we already know for the modulation of other sensory afferents, these processes likely serve to improve resolution, to scale neuron responses, to refine receptive fields, and to enhance aspects of signal encoding for subsequent higher-level processing.
Footnotes
The work was supported by National Institutes of Health, National Institute on Deafness and Other Communication Disorders RO1 DC01856 and Action on Hearing Loss, United Kingdom. We thank Hui Zhong (Susan) Xue for expert technical support.
The authors declare no competing financial interests.
- Correspondence should be addressed to Robin L. Davis at rldavis{at}dls.rutgers.edu
