Biased auditory nerve central synaptopathy is associated with Age-related Hearing Loss

Abstract

Sound information is transmitted from the cochlea to the brain by different subtypes of spiral ganglion neurons (SGN), which show varying degrees of vulnerability under pathological conditions. Selective cochlear synaptopathy, the preferential damage of certain subtypes of SGN peripheral synapses, has been recognized as one of the main mechanisms of hearing loss. The organization and function of the auditory nerve (AN) central synapses from different subtypes of SGNs remain unclear, including how different AN synapses reassemble onto individual neurons in the cochlear nucleus (CN), and how they differentially change during hearing loss. Combining immunohistochemistry with electrophysiology, we investigated the convergence pattern and subtype-specific synaptopathy of AN synapses at the endbulb of Held, as well as the response properties of their postsynaptic bushy neurons in CBA/CaJ mice of either sex under normal hearing and age-related hearing loss (ARHL). We found that calretinin-expressing (type I_a) and non-calretinin-expressing (type I_b/I_c) endbulbs converged along a continuum of different ratios onto individual bushy neurons with varying physiological properties. Endbulbs degenerated during aging in parallel with ARHL. Furthermore, the degeneration was more severe in non-calretinin-expressing synapses, which correlated with a gradual decrease in bushy neuron subpopulation predominantly innervated by these inputs. These synaptic and cellular changes were profound in middle-aged mice when their hearing thresholds were still relatively normal and prior to severe ARHL. Our findings suggest that biased AN central synaptopathy and the correlated shift in CN neuronal composition play significant roles in weakened auditory input and altered central auditory processing during ARHL.

INTRODUCTION

Spiral ganglion neurons (SGN) of the peripheral auditory system convey sound information from sensory hair cells to the cochlear nucleus (CN) (Nayagam et al., 2011), which is the first station of the central auditory system. Most SGNs (90–95%) are type I (Spoendlin, 1969; Kiang et al., 1982) and can be further divided into three subtypes based on their spontaneous firing rate and threshold, fiber caliber and preferential terminal distribution around the hair cell basal circumference (Liberman, 1978, 1980, 1982b, a; Liberman & Oliver, 1984). These SGN subtypes differentially encode sound intensity, and collectively ensure a comprehensive representation of the acoustic environment. Transfer of sound information through the bipolar SGNs begins at their dendritic terminal of cochlear synapses on hair cells, and ends at AN central synapses in the CN (Nayagam et al., 2011; Yu & Goodrich, 2014). Pathology of the cochlear synapses (known as “cochlear synaptopathy”) due to noise insult and aging preferentially affects the low spontaneous rate/high threshold subtype of type I SGNs (Furman et al., 2013; Liberman et al., 2015), and is recognized as a major mechanism that underlies hearing loss (Kujawa & Liberman, 2015; Liberman et al., 2015; Liberman, 2017; Parthasarathy & Kujawa, 2018). However, it remains unknown how different subtypes of AN central synapses converge onto individual CN neurons (Spirou et al., 2005; Nayagam et al., 2011; Yu & Goodrich, 2014), how different innervation patterns correlate with postsynaptic neuronal response properties, and most importantly, how subtype-specific changes of AN synapses affect the neural processing of target CN neurons under pathological conditions. Research to answer these questions has been hindered by the lack of known markers that can specifically label different subtypes of type I SGNs. Recent studies using single-cell RNA sequencing technique reported various molecular markers that can selectively classify type I SGNs into three distinct subtypes (type I_a, I_b, and I_c) (Petitpre et al., 2018; Shrestha et al., 2018; Sun et al., 2018). In particular, calretinin was shown to be selectively expressed in type I_a SGNs, whose peripheral synapses contact the pillar side of inner hair cells (Sharma et al., 2018) and presumably give rise to auditory nerve (AN) fibers with high spontaneous rate/low threshold (Liberman, 1982b).

The most prominent AN synapse is the endbulb of Held (Ryugo & Fekete, 1982; Rouiller et al., 1986), which reliably transmits information about the temporal fine structure of sound to its postsynaptic bushy neuron in ventral CN (Joris et al., 1994a; Joris et al., 1994b; Manis et al., 2011). Such temporal information is crucial for pitch perception and speech recognition (Shannon et al., 1995; Moore, 2008), which are often compromised in hearing impaired patients (Lorenzi et al., 2006; Grose & Mamo, 2010; Anderson et al., 2012). After hearing loss, endbulb of Held synapses show modified synaptic morphology with reduced size and simplified complexity (Ryugo et al., 1997; Ryugo et al., 1998; Wright et al., 2014), as well as compromised synaptic transmission (Oleskevich & Walmsley, 2002; Wang & Manis, 2005; Wright et al., 2014; Xie & Manis, 2017b; Zhuang et al., 2017) and reduced auditory input to target neurons (Xie, 2016). Despite decades of research on the endbulbs of Held, however, it remains unknown how different subtypes of endbulbs converge onto individual bushy neurons, as well as whether and how such synaptic innervation changes during hearing loss. It is also unclear how subtype-specific changes of the endbulbs impact the response properties of postsynaptic bushy neurons.

To address these issues, we combined immunohistochemistry with electrophysiology using acute brain slices from CBA/CaJ mice at three different age groups with normal hearing, moderate ARHL that mimics hidden hearing loss, and prominent ARHL (Sergeyenko et al., 2013). Endbulb of Held synapses from type I_a SGNs (high spontaneous rate/low threshold) were differentiated from other subtypes (type I_b/I_c SGNs with medium/low spontaneous rate and medium/high threshold) based on immunostaining against calretinin (Liberman, 1982b; Petitpre et al., 2018; Sharma et al., 2018; Shrestha et al., 2018; Sun et al., 2018). We report the innervation patterns of type I_a and non-type I_a endbulbs on individual bushy neurons and their intrinsic and AN-evoked response properties, as well as subtype-specific AN central synaptopathy during ARHL.

METHODS

Ethical Approval

All experiments were conducted under the guidelines of the protocols approved by the Institutional Animal Care and Use Committee of The Ohio State University (IACUC protocol number: 2018A00000055), which maintains an Animal Welfare Assurance (#D16–00168 / A3261–01) in compliance with the United States Public Health Service Policy on Humane Care and Use of Laboratory Animals. We understand the ethical principles under which the journal operates, and our work complies with the animal ethics checklist.

Origin and Source of the Animals

The study used CBA/CaJ mice of either sex, which were purchased from The Jackson Laboratory, bred and maintained at the animal facility at The Ohio State University. Three age groups of mice were used, including 32 young (1.5 – 4.5 months), 22 middle-aged (17 – 19 months), and 55 old (28 – 32 months) mice.

Access to Food and Water

All mice were maintained in 12-hour light/dark cycle, and had ad libitum access to food and water.

Anaesthetic Protocols and Euthanasia

Mice were deeply anaesthetized with IP injection of ketamine (100 mg/kg) and xylazine (10 mg/kg), which was verified by the absence of toe-pinch reflex. Euthanasia was performed under anesthesia by decapitation.

Auditory Brainstem Response (ABR)

Hearing status of the mice were assessed by measuring ABR to clicks as previously described (Wang et al., 2019). Briefly, mice were anaesthetized and placed inside a sound-attenuating chamber. Body temperature was maintained at ~ 36 °C using a feedback-controlled heating pad. ABR to clicks were acquired using a RZ6-A-P1 system with BioSigRZ software (Tucker-Davis Technologies). Clicks (0.1 ms, monophasic with alternating phase; 21 times/s) were delivered through a free field MF1 magnetic speaker located 10 cm away from the pinna. Recording needle electrodes were placed at the ipsilateral pinna and vertex, with the ground electrode at the rump. ABR at each sound level was repeated 512 times and averaged (Fig. 1).

Figure 1. CBA/CaJ mice show age-related hearing loss (ARHL).

(A) Example ABR waveforms from mice at different ages to clicks at different intensities. Red dots mark the positive and negative peaks of ABR wave I at 80 dB SPL. Thick traces: ABR threshold trace. Scale: 2 ms, 5 μV.

(B) ABR click thresholds in 32 young, 22 middle-aged, and 43 old mice. Dunn’s multiple comparisons test: **p < 0.001; ****p < 0.0001. ABR threshold was beyond 85 dB SPL in 12 old mice and not determined.

(C) Growth curves of ABR wave I amplitude in all mice. Thin lines: individual mice; thick lines: average of each age group, with error bars represent SEM.

Brain Slice Preparation

Under ketamine/xylazine anesthesia, mice were decapitated and the skulls were opened to retrieve the brainstem. Parasagittal slices containing the CN were cut at a thickness of 225 – 240 μm using a Vibratome 1000 (Technical Products, Inc.) or a VT1200S Microtome (Leica Biosystems). Tissue dissection and slicing were performed in artificial cerebral spinal fluid (ACSF) at 34 °C. ACSF contained (in mM): 122 NaCl, 3 KCl, 1.25 NaH₂PO₄, 25 NaHCO₃, 20 glucose, 3 myo-inositol, 2 sodium pyruvate, 0.4 ascorbic acid, 1.8 CaCl₂ and 1.5 MgSO₄, and was gassed with 95% O₂ and 5% CO₂. Slices were then incubated in the same ACSF at 34 °C for 30–45 minutes before recordings began.

Electrophysiological Recording

After incubation, the brain slice was moved to an ACSF-bathed recording chamber under an Axio Examiner microscope (Carl Zeiss). Whole-cell recordings under current clamp mode were performed from bushy neurons of the anteroventral CN (AVCN). Recording pipettes were made using a P-2000 micropipette puller (Sutter Instrument) and filled with electrode solution that contained (in mM): 126 potassium gluconate, 6 KCl, 2 NaCl, 10 HEPES, 0.2 EGTA, 4 MgATP, 0.3 GTP, 10 Tris-phosphocreatine, and pH adjusted to 7.20. Alexa Fluor 594 was added to the electrode solution at the final concentration of 0.01% by weight to dye-fill the neuron for online visualization of cell morphology as well as cell labeling upon completion. Data acquisition used hardware and software from Molecular devices, including Multiclamp 700B amplifier, Digidata 1550B acquisition system and pClamp 11 software. All recordings were made at 34 °C. Recordings were made from neurons in the middle and dorsal area of the AVCN. Only bushy neurons were included in the study, which were identified based on electrophysiological response properties and morphological features as previously described (Cant & Morest, 1979; Webster & Trune, 1982; Wu & Oertel, 1984; Manis et al., 2019). Responses to current step injections were obtained to assess intrinsic membrane properties. AN stimulation (Fig. 2A) was delivered through a 75 μm diameter concentric stimulating electrode (Frederick Haer Company) placed at the nerve root. The stimulus pulse had a duration of 0.1 ms, with intensity set at ~30% above the threshold intensity level that first triggered spikes in the target neuron. For neurons that failed to fire any spikes, stimulus intensity was tested up to the level that would cause tissue damage at the stimulation site. Responses to trains of AN stimulation at 100 Hz for 9 s and at 400 Hz for 10 s were recorded. Liquid junction potential was - 12 mV and was corrected. After completion of data acquisition, the recording electrode was slowly withdrawn until it broke off from the target neuron, which usually resealed itself with preserved neuronal morphology and remained filled with Alexa Fluor 594 dye. The brain slice was immediately fixed in 4% paraformaldehyde in PBS for 15 minutes, followed by post-hoc immunohistochemistry.

Figure 2. Different subtypes of AN synapses converge onto individual CN neurons.

(A) Diagram of the CN and experimental setup. It depicts two endbulb of Held synapses (E1 and E2) from different AN fibers (AN1 and AN2) that innervate the target bushy neuron, which is recorded and filled with fluorescent dye (magenta). AN is activated by electric stimulation. CN: cochlear nucleus; AVCN: anteroventral CN; PVCN: posteroventral CN; DCN: dorsal CN.

(B) Example responses of a bushy neuron to current step injections (top) and trains of AN stimulation at 100 Hz (bottom). Ticks mark the stimulus onset. Scale: 10 mV and 20 ms.

(C) Single frame confocal images of the bushy neuron in (B) filled with Alexa Fluor 594 dye (magenta), with immunostained VGluT1 (green) and calretinin (CR; red). The bushy neuron receives a large type I_a endbulb of Held synapse (asterisk), which contains VGluT1-labeled puncta (arrow). Double-labeled puncta are shown in yellow in the merged panel (arrow). VGluT1-labeled puncta that do not overlap with CR staining are shown in green in the merged panel (arrowhead), which are from AN synapses that do not express calretinin (non-type I_a synapse). Scale: 10 μm.

(D) 3D reconstruction of the bushy neuron and AN synapses in (C). Red: CR-stained type I_a endbulb of Held synapse. Scale: 10 μm, also applies to E-H. See also Supplemental Video 1.

(E) same as in (D) except showing only the VGluT1-labeled puncta. Yellow (arrow): VGluT1-labeled puncta inside the calretinin-expressing type I_a synapses; green (arrowhead): puncta from non-calretinin-expressing (non-type I_a) synapses.

(F) View of all VGluT1-labeled puncta from (E) without revealing the postsynaptic bushy neuron. Note that this bushy neuron received mostly type I_a synaptic inputs from the AN in terms of VGluT1-labeled puncta volume.

(G) Morphology of the single type I_a endbulb of Held synapse from (D).

(H) Type I_a VGluT1-labeled puncta inside the single endbulb in (G).

Immunohistochemistry

Fixed brain slices with dye-filled neurons were processed for immunostaining as previously described (Karadottir & Attwell, 2006; Lin & Xie, 2019). After rinsing in PBS (3 times, 15 minutes each), slices were pre-incubated in blocking solution (10% horse serum, 0.5% Triton X-100, and 0.05% NaN₃ in PBS) for 6 hours at room temperature, followed by overnight incubation with primary antibodies against VGluT1 (guinea pig anti VGluT1; Cat# 135304, Synaptic Systems, 1:500) and calretinin (rabbit anti calretinin; Cat# 214102, Synaptic Systems, 1:500) at 4 °C. Slices were then rinsed again in PBS (4 times, 20 minutes each), incubated with secondary antibodies (goat anti guinea pig IgG, Alexa 488 conjugated, Cat # A11073, Thermo Fisher Scientific, 1:500; and goat anti rabbit IgG, Alexa 647 conjugated, Cat# A21245, Thermo Fisher Scientific, 1:500) at room temperature for 4 hours, re-rinsed, and mounted on slide with DAPI-Fluoromount-G mounting media (Southern Biotech). Filled neurons and labeled AN synapses were imaged using an Olympus FV3000 confocal microscope. Images were sampled using a 60x oil immersion objective, 3.0 times software zoom, a z-step size of 0.3 μm, and at a resolution of either 800 × 800 or 1024 × 1024 pixels.

Image processing

All image processing and measurement were performed age-blinded and done using Imaris software (version 9.5.0; Oxford Instruments). Three-dimensional reconstruction of the dye-filled target bushy neuron (magenta), VGluT1-labeled puncta (green) and type I_a endbulb of Held terminals (red) in close proximity to the soma of the target bushy neuron were made using semi-automated Imaris tools, including masking uninterested area and rendering surfaces in three separate channels (see Supplemental Video 1 for detailed step-by-step process). VGluT1-labeled puncta located inside type I_a endbulbs were determined as puncta from type I_a synapses and shown in yellow (overlap of red and green), while those outside of type I_a endbulbs were from non-type I_a synapses and remained in green (Fig. 2D–F, Supplemental Video 1). Volumes of VGluT1-labeled puncta and individually identified endbulb of Held synapses were measured.

Electrophysiology Analysis

Electrophysiological data were analyzed in Igor Pro (WaveMetrics) using custom-written functions. Responses to current step injections were used to quantify the intrinsic membrane properties as previously described (Xie & Manis, 2017a). Input resistance was calculated as the slope of the current-voltage relationship curve from current step injections of −25, −50, −75, and −100 pA and the resulted peak hyperpolarization amplitudes. Depolarizing current step injection was gradually increased by 25 pA steps until the target neuron fired the first spike, i.e. the threshold spike. This current level was determined as the threshold current of the neuron. Threshold spike amplitude was calculated as the voltage difference between the resting membrane potential and the peak of the threshold spike. Spike adaptation index of the responses to 100 and 400 Hz trains were calculated as the ratio between the firing rate during the last 5 seconds of the train and the firing rate during the first second of the train. Vector strength (Goldberg & Brown, 1969) of the stimulus train-evoked spikes was calculated using the formula: $\text{VS} =\sqrt{{\left(\Sigma sin\theta i\right)}^2+{\left(\Sigma cos\theta i\right)}^2}/n$, where θ_i is the phase angle of the spike i relative to the stimulation cycle, and n is the total number of spikes throughout the stimulus trains (Xie & Manis, 2013). Rayleigh statistics was calculated as 2n(VS)². Neurons with 2n(VS)² larger than 13.8 were considered to fire phase-locked spikes (p < 0.001) (Buunen & Rhode, 1978; Lu & Wang, 2000). Neurons with n less than 30, or fired less than 5 spikes per stimulus train were excluded from VS analysis. Spike jitter was calculated as the standard deviation of the spike latency relative to the stimulus onset.

Experimental design and statistical analysis

The study was cross-sectional using three different age groups of mice. The first set of study was conducted in young mice to characterize features of AN central synapses under normal hearing. Changes of AN synapses and postsynaptic neurons during hearing loss were further investigated in middle-aged and old mice groups with moderate and prominent ARHL, respectively. Statistical analysis used GraphPad Prism (version 6.0h). Kolmogorov-Smirnov test was first performed to examine whether population data were normally distributed. For data with normal distribution, unpaired t-test or one-way ANOVA followed by Tukey’s multiple comparisons test was used as stated. For data that were not normally distributed, Mann Whitney test or Kruskal-Wallis test followed by Dunn’s multiple comparisons test was used. Two-way ANOVA was performed to analyze the population data variances with calculated main effects of each of the factors and their interactions. All data were presented as mean ± SD except in Fig. 1C, which used mean ± SEM for clarity purpose of the panel.

RESULTS

CBA/CaJ mice and ARHL

Experiments in this study used CBA/CaJ mice, which developed progressive ARHL as shown by elevated hearing threshold and reduced wave I amplitude in ABRs to clicks (Fig. 1A). On average, the click threshold was 28 ± 4 dB SPL in young mice (n = 32), 37 ± 4 dB SPL in middle-aged mice (n = 22), and 70 ± 10 dB SPL in old mice (n = 43) (Fig. 1B; Kruskal-Wallis test: p < 0.0001; Dunn’s multiple comparisons test: young vs. middle-aged, p < 0.01; young vs. old, p < 0.0001; middle-aged vs. old, p < 0.0001). With increasing sound intensity, the growth curve of the wave I amplitude became considerably flatter in middle-aged and old mice with significantly decreased peak amplitude (Fig. 1C; Two-way ANOVA: age effect, F_{(2, 1477)} = 3309, p < 0.0001; intensity effect, F_{(13, 1477)} = 318, p < 0.0001; interaction, F_{(26, 1477)} = 121, p < 0.0001). In particular, middle-aged mice showed relatively normal ABR threshold but greatly reduced ABR wave I amplitude, which resembles hidden hearing loss. The observed hearing profile is consistent with previous reports about CBA/CaJ mice (Sergeyenko et al., 2013).

Different subtypes of AN central synapses converge onto individual CN neurons

Bushy neurons are the principal neurons of the AVCN that receives AN inputs via large endbulb of Held synapses on the soma (Manis et al., 2011). We combined electrophysiology with post-hoc immunohistochemistry to identify the subtype-specificity of the AN synaptic inputs onto individual bushy neurons in parasagittal CN slices. Whole-cell current clamp recording was performed from target bushy neurons, which were filled with Alexa Fluor 594 included in the electrode solution (Fig. 2A–B). Slices were post-hoc immunostained using antibodies against VGluT1, which labels glutamate vesicles in all AN synaptic terminals, and calretinin, which labels only the AN fibers and synapses from type I_a SGNs (Materials and Methods). As shown in Fig. 2C, the bushy neuron received inputs from a large type I_a endbulb of Held synapse (asterisk), which contained glutamate vesicles revealed by VGluT1-staining (arrow). In addition, there were VGluT1-labeled puncta (arrowhead) that were not enclosed in calretinin-labeled synapses, indicating that they were from non-type I_a AN synapses (presumably type I_b or I_c). Z-stack images were acquired to reconstruct the 3-dimentional structure of the type I_a endbulbs as well as VGluT1-labeled puncta surrounding the soma of the labeled bushy neuron (Fig. 2D–H; Supplemental Video 1). Using this approach, we show that type I_a and non-type I_a AN synapses converge onto individual CN bushy neurons (Fig. 2F).

The proportion of type I_a synaptic inputs correlates with physiological properties of postsynaptic neurons.

We studied the innervation patterns of type I_a and non-type I_a AN synapses onto 49 bushy neurons in normal hearing young mice as well as their electrophysiological response properties. Across the population, bushy neurons received different convergence of inputs from type I_a and non-type I_a synapses as shown in Fig. 3A, D, G, and J by different colors of VGluT1-labeled puncta. We measured the puncta volume and quantified the proportion of type I_a inputs (type I_a volume %) onto each bushy neuron, which varied on a continuum between those that received type I_a-dominant (I_a-D; with I_a volume > 50%; Fig. 3A, D) inputs and those with non-type I_a-dominant (Non-I_a-D; with I_a volume < 50%; Fig. 3G, J) inputs (Fig. 3M). The results suggest that auditory information from different subtypes of SGNs is systematically reassembled and processed in target CN neurons with gradually changing synaptic convergence. Interestingly, there was a significant correlation between the total volume of VGluT1-labeled puncta and the proportion of type I_a puncta volume (Fig. 3M), which indicates that neurons with higher proportion of type I_a inputs have larger volume of total VGluT1-labeled puncta, and thus stronger overall inputs from AN synapses.

Figure 3. The proportion of type I_a synaptic inputs correlates with physiological properties of postsynaptic neurons in young mice.

(A-C) An example bushy neuron that received mostly type I_a synaptic inputs (I_a-dominant, or I_a-D) from the AN. (A) Reconstructed VGluT1-labeled puncta surrounding the soma of the target bushy cell (not shown) from type I_a (yellow) and non-type I_a (green) synapses. I_a%: volume proportion of the VGluT1-labeled puncta from type I_a synapses over total. Scale: 5 μm. (B) Responses of the bushy cell to current step injections. Red traces: threshold current injection (bottom) and response (top). Scale: 10 mV and 20 ms. (C) The neuron fired sustained spikes to a train of auditory nerve stimulation (black bar) at 100 Hz; scale: 10 mV and 1 s. Arrowhead: last spikes of the train expanded in the inset on the right; scale: 10 mV and 10 ms. PSTH: post-stimulus time histogram. Stim Num: stimulus number. Adap: spike adaptation index. PH: period histogram. VS: vector strength.

(D-F) Another example bushy neuron with I_a-D but lower proportion of type I_a inputs.

(G-I, J-L) Two example bushy neurons that received mostly non-type I_a synaptic inputs (Non-type I_a-dominant, or Non-I_a-D), and fired only transient or onset spikes to AN stimulus trains.

(M) Bushy neurons with different proportion of type I_a inputs (x-axis) show correlated distribution in the total volume of VGluT1-labeled puncta (black), volume of type I_a only puncta (yellow), and volume of non-type I_a only puncta (green). Linear regression lines: black, r² = 0.09, p = 0.039; yellow, r² = 0.57, p < 0.0001; green, r² = 0.38, p < 0.0001.

(N) The proportion of type I_a inputs (x-axis) in bushy neurons negatively correlates with their threshold spike amplitude, as shown by example responses in B, E, H and K. Linear regression line: r² = 0.40, p < 0.0001.

(O) Bushy neurons with different proportion of type I_a inputs (x-axis) show different firing patterns to 100 Hz stimulus trains, as quantified by spike adaptation index. Linear regression line: r² = 0.13, p = 0.015.

(P-V) Comparisons between bushy neurons with I_a-D and Non-I_a-D inputs in resting potential (P), input resistance (Q), threshold spike amplitude (R), threshold current injection level that triggered the first spike (S), firing rate throughout the 100 Hz train (T), spike adaptation index (U), and vector strength of the spikes (V). Unpaired t-test or Mann-Whitney test: NS, p > 0.05; *p < 0.05; **p < 0.01; ***p < 0.001.

To assess the intrinsic membrane properties of the target bushy neurons, we acquired electrophysiological responses to current step injections (Figure 2A–B; see Materials and Methods). Example responses from four bushy neurons in Fig. 3A, D, G, J are shown in Fig. 3B, E, H, K, respectively. Neurons with I_a-D inputs (Fig. 3B, E) had smaller threshold spike amplitude than those with Non-I_a-D inputs (Fig. 3H, K), which is shown by the negative correlation between threshold spike amplitude and the proportion of type I_a inputs among all bushy neurons (Fig. 3N). Compared to those with Non-I_a-D inputs, bushy neurons with I_a-D inputs showed significantly more depolarized resting membrane potential (Fig. 3P), smaller spike amplitude (Fig. 3R), required smaller current injection to trigger threshold spikes (Fig. 3S), but had similar input resistance (Fig. 3Q) (Table 1). The data suggests that bushy neurons with I_a-D inputs are more excitable than those with Non-I_a-D inputs.

Table 1.

Electrophysiological properties of bushy neurons.

		Ia-D			Non-Ia-D			p value (unpaired t-test or Mann Whitney test)
		n	mean	SD	n	mean	SD
Young	resting potential (mV)	25	−65.2	2.4	24	−67.2	2.6	0.0075**
	input resistance (MΩ)	25	60	23	24	67	38	0.851
	threshold spike amp (mV)	25	27.8	5.4	24	38.3	12.4	0.0007***
	threshold current (pA)	25	203	84	24	326	193	0.025*
	spikes per stim at 100 Hz	25	0.50	0.31	24	0.23	0.33	0.0014**
	adaptation index at 100 Hz	25	0.47	0.31	24	0.21	0.34	0.0004***
	vector strength at 100 Hz	25	0.993	0.004	24	0.985	0.015	0.030*
	spike jitter at 100 Hz (ms)	25	0.180	0.060	24	0.257	0.126	0.030*
	spikes per stim at 400 Hz	23	0.068	0.080	21	0.016	0.042	< 0.0001****
	adaptation index at 400 Hz	23	0.17	0.22	21	0.02	0.05	< 0.0001****
	vector strength at 400 Hz	22	0.74	0.15	9	0.73	0.15	0.823
	spike jitter at 400 Hz (ms)	22	0.344	0.181	9	0.347	0.164	0.781
Middle-aged	resting potential (mV)	23	−65.5	1.6	8	−66.0	1.9	0.449
	input resistance (MΩ)	23	58	17	8	58	13	0.982
	threshold spike amp (mV)	23	28.2	6.5	8	35.5	12.9	0.049*
	threshold current (pA)	23	205	84	8	338	136	0.0005***
	spikes per stim at 100 Hz	23	0.50	0.34	8	0.26	0.36	0.080
	adaptation index at 100 Hz	22	0.48	0.34	5	0.32	0.35	0.306
	vector strength at 100 Hz	20	0.995	0.003	5	0.982	0.020	0.0074**
	spike jitter at 100 Hz (ms)	20	0.166	0.069	5	0.279	0.142	0.015*
	spikes per stim at 400 Hz	22	0.05	0.10	8	0.01	0.02	0.030*
	adaptation index at 400 Hz	20	0.035	0.049	5	0.002	0.005	0.025*
	vector strength at 400 Hz	19	0.806	0.082	4	0.679	0.157	0.138
	spike jitter at 400 Hz (ms)	19	0.267	0.079	4	0.381	0.155	0.188
Old	resting potential (mV)	29	−65.3	2.5	6	−66.2	2.6	0.445
	input resistance (MΩ)	29	72	33	6	72	36	0.982
	threshold spike amp (mV)	29	28.5	4.8	6	36.0	8.1	0.0047**
	threshold current (pA)	29	185	94	6	254	145	0.273
	spikes per stim at 100 Hz	29	0.38	0.32	6	0.06	0.08	0.019*
	adaptation index at 100 Hz	24	0.37	0.30	4	0.02	0.02	0.0007***
	vector strength at 100 Hz	23	0.986	0.010	3	0.959	0.039	0.182
	spike jitter at 100 Hz (ms)	23	0.257	0.118	3	0.420	0.241	0.211
	spikes per stim at 400 Hz	29	0.026	0.037	5	0.001	0.001	0.018*
	adaptation index at 400 Hz	24	0.098	0.144	3	0.018	0.031	0.230
	vector strength at 400 Hz	21	0.689	0.179	3	0.808	0.158	0.271
	spike jitter at 400 Hz (ms)	21	0.361	0.154	3	0.247	0.131	0.172

We next investigated the AN evoked firing properties of target bushy neurons using responses to trains of AN stimulation (Fig. 2A–B). AN stimulation at 100 Hz evoked sustained spikes with moderate adaptation in neurons with I_a-D inputs (Fig. 3C, F), but only elicited transient or onset spikes in those with Non-I_a-D inputs (Fig. 3I, L). Spike adaptation index was calculated (see Materials and Methods) to quantify AN evoked firing patterns, in which higher adaptation index value corresponds to the firing of more sustained spikes and lower adaptation index value corresponds to the firing of transient/onset spikes. As shown in Fig. 3O, there was a significant correlation between spike adaptation index and the proportion of type I_a inputs across all bushy neurons. In summary, neurons with I_a-D inputs showed significantly higher AN-evoked firing rate (Fig. 3T) and spike adaptation index (Fig. 3U) than neurons with Non-I_a-D inputs (Table 1). They also fired spikes with higher temporal precision, as revealed by significantly larger vector strength (Fig. 3V) and smaller spike jitter (Table 1).

AN can fire at sustained rates above 400 Hz under physiological conditions (Taberner & Liberman, 2005; Wen et al., 2009). We further tested the firing properties of bushy neurons to high rate AN stimulation at 400 Hz (Fig. 4A–D). Similar to 100 Hz trains, 400 Hz AN stimulation evoked significantly higher firing rate (Fig. 4E) and spike adaptation index (Fig. 4F) in bushy neurons with I_a-D inputs than those with Non-I_a-D inputs (Table 1). No significant difference was found in the vector strength of evoked spikes (Fig. 4G) or spike jitter (Table 1). Most neurons (12 out of 21) with Non-I_a-D inputs fired insufficient number of spikes to 400 Hz stimulus trains and no vector strength or spike jitter were calculated from those neurons.

Figure 4. Bushy neurons with different convergence of AN synaptic inputs show different firing patterns to trains of AN stimulation at 400 Hz in young mice.

(A) The same example bushy neuron in Fig. 3A fired sustained spikes to trains of AN stimulation at 400 Hz. Scale: 10 mV and 1 s.

(B-D) Responses of bushy neurons in Fig. 3D, G, and J to AN stimulation at 400 Hz. Vector strength of the cell in D was not calculated due to insufficient number of spikes.

(E-G) Comparisons between bushy neurons with I_a-D inputs and Non-I_a-D inputs to 400 Hz AN stimulus trains in firing rate (E), spike adaptation index (F), and vector strength (G). Unpaired t-test or Mann Whitney test: NS, p > 0.05; ****p < 0.0001.

These results indicate that under normal hearing condition, AN synapses from different subtypes of SGNs innervate CN bushy neurons with systematically varying ratio of convergence, which correlates with distinct intrinsic property and AN evoked firing property of target neurons. The findings suggest a new dimension of neuronal organization within CN bushy cell population that dedicates to the reassembling and processing of auditory intensity information from different subtypes of SGNs.

AN synaptopathy and altered composition of bushy neuron population during ARHL

We next examined the innervation patterns of different AN synapses and the response properties of postsynaptic bushy neurons in middle-aged and old mice with different levels of ARHL (Fig. 1). Recordings were obtained from 31 bushy neurons (Figs. 5, 6) in middle-aged and 35 bushy neurons (Figs. 7, 8) in old mice.

Figure 5. Convergence of different AN synapses and the response properties of postsynaptic bushy neurons in middle-aged mice.

(A-C) Example bushy neuron with I_a-D inputs (I_a: 92%). (A) Reconstructed VGluT1-labeled puncta from type I_a (yellow) and non-type I_a (green) synapses. Scale: 5 μm. (B) Responses of the bushy cell to current step injections. Scale: 10 mV and 20 ms. (C) The neuron fired sustained spikes to a train of auditory nerve stimulation at 100 Hz; scale: 10 mV and 1 s. Arrowhead: last spikes of the train expanded in the inset on the right; scale: 10 mV and 10 ms.

(D-F) Example bushy neuron with I_a-D but lower proportion of type I_a inputs (I_a: 69%).

(G-I, J-L) Two example bushy neurons that received Non-I_a-D inputs, and fired only transient spikes or failed to fire any spike to AN stimulus trains.

(M) Bushy neurons with different proportion of type I_a inputs show correlated distribution in the total volume of VGluT1-labeled puncta (black), volume of type I_a only puncta (yellow), and volume of non-type I_a only puncta (green). Linear regression lines: black, r² = 0.16, p = 0.025; yellow, r² = 0.62, p < 0.0001; green, r² = 0.40, p = 0.0001.

(N) The proportion of type I_a inputs in bushy neurons negatively correlates with their threshold spike amplitude. Linear regression line: r² = 0.15, p = 0.043.

(O) Spike adaptation index from 100 Hz trains in bushy neurons with different proportion of type I_a inputs. Linear regression line: r² = 0.08, p = 0.160.

(P-V) Comparisons between bushy neurons from middle-aged mice with I_a-D and Non-I_a-D inputs in resting potential (P), input resistance (Q), threshold spike amplitude (R), threshold current level (S), firing rate throughout the 100 Hz train (T), spike adaptation index (U), and vector strength of the spikes (V). Unpaired t-test or Mann-Whitney test: NS, p > 0.05; *p < 0.05; **p < 0.01; ***p < 0.001.

Figure 6. Bushy neurons in middle-aged mice show different firing properties to trains of AN stimulation at 400 Hz.

(A-D) Responses of four example bushy neurons to trains of AN stimulation at 400 Hz. Scale: 10 mV and 1 s. No spike was evoked in the cell in D.

(E-G) Comparisons between bushy neurons with I_a-D and Non-I_a-D inputs to 400 Hz AN stimulus trains in firing rate (E), spike adaptation index (F), and vector strength (G). Mann Whitney test: NS, p > 0.05; *p < 0.05.

Figure 7. Convergence of different AN synapses and the response properties of postsynaptic bushy neurons in old mice.

(A-C) Example bushy neuron with I_a-D inputs (I_a: 91%). (A) Reconstructed VGluT1-labeled puncta from type I_a (yellow) and non-type I_a (green) synapses. Scale: 5 μm. (B) Responses of the bushy cell to current step injections. Scale: 10 mV and 20 ms. (C) The neuron fired sustained spikes to a train of auditory nerve stimulation at 100 Hz; scale: 10 mV and 1 s. Arrowhead: last spikes of the train expanded in the inset on the right; scale: 10 mV and 10 ms.

(D-F) Example bushy neuron with I_a-D but lower proportion of type I_a inputs (I_a: 66%).

(G-I, J-L) Two example bushy neurons that received Non-I_a-D inputs (I_a: 10% and 39%).

(M) Bushy neurons with different proportion of type I_a inputs show correlated distribution in the total volume of VGluT1-labeled puncta (black), volume of type I_a only puncta (yellow), and volume of non-type I_a only puncta (green). Linear regression lines: black, r² = 0.13, p = 0.033; yellow, r² = 0.40, p < 0.0001; green, r² = 0.49, p < 0.0001.

(N) Proportion of type I_a inputs (x-axis) in bushy neurons negatively correlates with their threshold spike amplitude. Linear regression line: r² = 0.30, p = 0.001.

(O) Spike adaptation index from 100 Hz trains in bushy neurons with different proportion of type I_a inputs (x-axis). Linear regression line: r² = 0.20, p = 0.019.

(P-V) Comparisons between bushy neurons from old mice with I_a-D and Non-I_a-D inputs in resting potential (P), input resistance (Q), threshold spike amplitude (R), threshold current level (S), firing rate throughout the 100 Hz trains (T), spike adaptation index (U), and vector strength of the spikes (V). Unpaired t-test or Mann-Whitney test: NS, p > 0.05; *p < 0.05; ***p < 0.001.

Figure 8. Bushy neurons in old mice show different firing properties to trains of AN stimulation at 400 Hz.

(A-D) Responses of four example bushy neurons to trains of AN stimulation at 400 Hz. Scale: 10 mV and 1 s. No spike was evoked in the cell in D.

(E-G) Comparisons between bushy neurons with I_a-D and Non-I_a-D inputs to 400 Hz AN stimulus trains in firing rate (E), spike adaptation index (F), and vector strength (G). Mann Whitney test: NS, p > 0.05; *p < 0.05.

Similar to those in young mice, we observed both bushy neurons with I_a-D inputs and those with Non-I_a-D inputs in middle-aged (Fig. 5A–M) and old (Fig. 7A–M) mice, except that the prevalence of neurons with Non-I_a-D inputs decreased with age (also see Fig. 9D and the discussion below). In general, the proportion of type I_a inputs remains correlated with the intrinsic and AN evoked response properties of the postsynaptic bushy neurons in both age groups with ARHL. Specifically, in middle-aged mice, higher proportion of type I_a inputs was correlated with smaller threshold spike amplitude (Fig. 5N), with target neurons tended to fire more sustained spikes (Fig. 5O). In summary, compared to those with Non-I_a-D inputs, bushy neurons with I_a-D inputs showed similar resting membrane potential (Fig. 5P) and input resistance (Fig. 5Q), but had significantly smaller threshold spike amplitude (Fig. 5R), were more excitable and required significantly smaller current injection to trigger threshold spikes (Fig. 5S) (Table 1).

Figure 9. AN central synaptopathy is more severe in non-type I_a synapses and associated with altered composition of bushy neuron population during ARHL.

(A-C) Comparisons of the total volume (A), type I_a only volume

(B) and non-type I_a only volume (C) of VGluT1-labeled puncta onto bushy neurons from three age groups of mice. Tukey’s or Dunn’s multiple comparison tests: *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001.

(D) Prevalence of bushy neurons that receive I_a-D and Non-I_a-D synaptic inputs among three age groups. Numbers mark the cell count of each type over total.

(E-J) Comparisons of bushy neurons among three age groups in resting potential (E), threshold spike amplitude (F), threshold current injection that triggered the first spike (G), firing rate to 100 Hz stimulation (H), spike adaptation index (I), and vector strength of the spikes throughout the 100 Hz trains (J). Two-way ANOVA revealed significant cell type effect (I_a-D vs. Non-I_a-D) in all panels (E-J): **p < 0.01; ***p < 0.001; ****p < 0.0001. Age effect was only significant in vector strength (J): p < 0.001.

In response to AN stimulus trains at 100 Hz, middle-aged neurons with I_a-D inputs showed a tendency to fire more spikes (Fig. 5T) with higher spike adaptation index (Fig. 5U; excluding 3 Non-I_a-D neurons that failed to fire any spike), as well as significantly better temporal precision as revealed by larger vector strength (Fig. 5V) and smaller spike jitter (Table 1). In response to 400 Hz stimulus trains (Fig. 6A–D), bushy neurons from middle-aged mice showed similar differences between neurons with I_a-D inputs and those with Non-I_a-D inputs, in which the former fired at significantly higher rate (Fig. 6E) with higher spike adaptation index (Fig. 6F) (Table 1). No significant difference was found in their spike vector strength (Fig. 6G) or spike jitter (Table 1).

In old mice, higher proportion of type I_a inputs was also correlated with smaller threshold spike amplitude (Fig. 7N), with target neurons fired more sustained spikes (Fig. 7O). Compared to bushy neurons with Non-I_a-D inputs, neurons with I_a-D inputs showed similar resting membrane potential (Fig. 7P) and input resistance (Fig. 7Q), but had significantly smaller threshold spike amplitude (Fig. 7R) (Table 1). They had similar excitability with no significant difference in their threshold current level (Fig. 7S; Table 1). In response to AN stimulus trains at 100 Hz, neurons with I_a-D inputs fired significantly more spikes (Fig. 7T) with higher spike adaptation index (Fig. 7U) (Table 1). No significant difference was found in spike vector strength (Fig. 7V) or spike jitter (Table 1). In response to 400 Hz stimulus trains (Fig. 8A–D), bushy neurons with I_a-D inputs fired at significantly higher rate (Fig. 8E), but with similar spike adaptation index (Fig. 8F), vector strength (Fig. 8G) and spike jitter (Table 1).

To evaluate the morphological changes of AN synapses during ARHL, we compared the total volume as well as the volumes of subtype-specific VGluT1-labeled puncta in bushy neurons from mice of all three age groups. As shown in Fig. 9A, the total volume of VGluT1-labeled puncta in bushy neurons significantly decreased with age (young: 100 ± 50 μm³, n = 49; middle-aged: 77 ± 31 μm³, n = 31; old: 46 ± 25 μm³, n = 35; one-way ANOVA: F_(2,112) = 19.26, p < 0.0001), suggesting an overall degeneration of AN synapses during ARHL. Further investigation of different AN subtypes showed that synaptic degeneration was significant in both type I_a synapses (Fig. 9B; type I_a VGluT1-labeled puncta, young: 61 ± 46 μm³, n = 49; middle-aged: 52 ± 29 μm³, n = 31; old: 36 ± 24 μm³, n = 35; one-way ANOVA: F_(2,112) = 5.00, p = 0.0083) and non-type I_a synapses (Fig. 9C; Non-type I_a VGluT1-labeled puncta, young: 39 ± 31 μm³, n = 49; middle-aged: 26 ± 17 μm³, n = 31; old: 11 ± 9 μm³, n = 35; Kruskal-Wallis test: p < 0.0001). Remarkably, the degeneration was more profound in non-type I_a synapses (Fig. 9B–C; two-way ANOVA: age effect, F_(2,224) = 16.2, p < 0.0001; synaptic subtype effect, F_(1,224) = 36.2, p < 0.0001; interaction, F_(2,224) = 0.11, p = 0.900). On average, the volume of VGluT1-labeled puncta from type I_a synapses decreased by 15% in middle-aged mice and 41% in old mice compared to young mice (Fig. 9B). In contrast, the volume of non-type I_a puncta decreased by 34% and 73% in middle-aged and old mice (Fig. 9C), respectively. These results indicate that there is a progressive AN central synaptopathy during aging, and this synaptopathy is biased toward being more severe in non-type I_a synapses.

Consistent with the more profound degeneration of non-type I_a synapses, the prevalence of postsynaptic bushy neurons with Non-I_a-D inputs progressively decreased with age (Fig. 9D; 49% in young, 26% in middle-aged, and 17% in old mice; also see Fig. 3M, 5M and 7M for the distribution of all individual neurons). Across different age groups, the differences between bushy neurons with I_a-D inputs and those with non-I_a-D inputs in both intrinsic membrane properties (Fig. 9E–G) and AN-evoked firing responses (Fig. 9H–J) were retained. Specifically, bushy neurons with I_a-D inputs showed significantly depolarized resting membrane potential (Fig. 9E; two-way ANOVA: cell type effect, F_(1,109) = 4.98, p = 0.028; age effect, F_(2,109) = 0.42, p = 0.656; interaction, F_(2,109) = 1.05, p = 0.352), smaller threshold spike amplitude (Fig. 9F; two-way ANOVA: cell type effect, F_(1,109) = 20.88, p < 0.0001; age effect, F_(2,109) = 0.17, p = 0.841; interaction, F_(2,109) = 0.39, p = 0.678), were more excitable with smaller threshold current (Fig. 9G; two-way ANOVA: cell type effect, F_(1,109) = 15.18, p = 0.0002; age effect, F_(2,109) = 1.16, p = 0.319; interaction, F_(2,109) = 0.44, p = 0.648), fired more spikes to AN stimulation at 100 Hz (Fig. 9H; two-way ANOVA: cell type effect, F_(1,109) = 15.55, p = 0.0001; age effect, F_(2,109) = 1.67, p = 0.193; interaction, F_(2,109) = 0.09, p = 0.912) with higher spike adaptation index (Fig. 9I; two-way ANOVA: cell type effect, F_(1,94) = 9.74, p = 0.0024; age effect, F_(2,94) = 1.61, p = 0.205; interaction, F_(2,94) = 0.35, p = 0.702) and higher vector strength (Fig. 9J; two-way ANOVA: cell type effect, F_(1,88) = 24.08, p < 0.0001; age effect, F_(2,88) = 8.60, p = 0.0004; interaction, F_(2,88) = 2.70, p = 0.073). However, there were increasingly more neurons during aging that failed to fire any spike to AN stimulation (Fig. 5L, 7L) (3/49 in young, 5/31 in middle-aged, 7/35 in old mice). The results suggest an age-related shift in the composition of bushy neuron population during ARHL in the direction of having relatively more neurons with I_a-D inputs.

Endbulbs show decreased synaptic strength with age during ARHL

To evaluate the synaptic strength of the endbulb of Held onto bushy neurons during ARHL, we analyzed the evoked EPSPs that failed to trigger any spikes during the response to 100 Hz stimulus trains (Fig. 10). EPSPs that did trigger spikes were often merged with the spike waveform (Fig. 10A) and therefore excluded from this analysis. As shown in Fig. 10B, the average EPSP amplitude showed no difference overall between neurons with I_a-D and Non-I_a-D inputs, but was significantly decreased with age (two-way ANOVA: cell type effect, F_(1,94) = 0.149, p = 0.701; age effect, F_{(2, 94)} = 5.17, p = 0.0074; interaction, F_{(2, 94)} = 0.163, p = 0.850). We used two additional EPSP measurements to estimate the threshold of the synaptic inputs in triggering postsynaptic spikes. The first was EPSP amplitude at 90^th percentile, which was significantly decreased with age (Fig. 10C; two-way ANOVA: cell type effect, F_(1,94) = 0.117, p = 0.733; age effect, F_{(2, 94)} = 3.82, p = 0.025; interaction, F_{(2, 94)} = 0.163, p = 0.850). As bushy neurons often exhibit a rate threshold of the membrane potential depolarization (McGinley & Oertel, 2006), we secondly calculated the maximum rising slope of each EPSP, which was determined as the peak of the first derivative during the EPSP rising phase (Fig. 10D) (Xie & Manis, 2013). The rate threshold was estimated to be the 90^th percentile of the EPSP maximum slope in each bushy neuron, which was also significantly decreased with age (Fig. 10E; two-way ANOVA: cell type effect, F_(1,94) = 2.73, p = 0.102; age effect, F_{(2, 94)} = 5.11, p = 0.0078; interaction, F_{(2, 94)} = 0.603, p = 0.549). The results suggest that the synaptic strength at the endbulb of Held was weakened during aging, which is consistent with the overall degeneration of AN synapses (Fig. 9A–C). However, preferential reduction in synaptic strength of the Non-I_a-D inputs was not detected (Fig. 10B), possibly due to the small number of neurons with Non-I_a-D inputs in aged groups. The decreased threshold to AN synaptic input with age (Fig. 10C, E) may reflect homeostatic changes of CN neurons that compensate the reduced sensory inputs from the AN during ARHL (Xie, 2016).

Figure 10. Evoked EPSPs decrease in amplitude with age during ARHL.

(A) Example response of a bushy neuron to show AN stimulation evoked spikes and EPSPs. Dashed rectangle: an example EPSP that failed to trigger any spike. Scale: 10 mV and 10 ms.

(B-C) Comparisons of bushy neurons among three age groups in average amplitude of EPSPs that failed to trigger spikes (B), and EPSP amplitude at 90^th percentile (C).

(D) Maximum rising slope of the example EPSP from (A), calculated as the peak of the first derivative (dv/dt) during the rising phase.

(E) Comparison of EPSP maximum slope at 90^th percentile.

Two-way ANOVA revealed significant age effect (I_a-D vs. Non-I_a-D) in B, C and E: *p < 0.05; **p < 0.01.

Synaptopathy in individual type I_a endbulb of Held synapses during ARHL

As calretinin-staining labels the entire type I_a synaptic terminal, we reconstructed the 3-dimensional profile of individual type I_a endbulb of Held synapses onto bushy neurons from three age groups of mice (Fig. 11A–C). Bushy neurons with identifiable type I_a endbulb terminals received between 1 – 3 type I_a endbulbs, with an average of 1.5 ± 0.6 (n = 64) in young, 1.5 ± 0.6 (n = 20) in middle-aged, and 1.7 ± 0.7 (n = 32) in old mice (Kruskal-Wallis test: p = 0.472). Individual type I_a endbulbs degenerated during aging with significantly decreased volume of the synaptic terminal (Fig. 11D; young: 218 ± 101 μm³, n = 98; middle-aged: 126 ± 63 μm³, n = 29; old: 104 ± 64 μm³, n = 53; Kruskal-Wallis test: p < 0.0001) as well as the volume of VGluT1-labeled puncta within each endbulb (Fig. 11E; young: 70 ± 37 μm³, n = 98; middle-aged: 35 ± 18 μm³, n = 29; old: 29 ± 15 μm³, n = 53; Kruskal-Wallis test: p < 0.0001). However, there was no change in the volume ratio between the two (Fig. 11F; young: 0.32 ± 0.11, n = 98; middle-aged: 0.29 ± 0.10, n = 29; old: 0.30 ± 0.11, n = 53; Kruskal-Wallis test: p = 0.416), suggesting that the relative quantity of synaptic vesicles within given synaptic terminal space remains unchanged during aging. Thus, AN synaptopathy during ARHL occurs with reduced synaptic terminal volume as well as a balanced decrease in the functional components of the synaptic machinery, at least in type I_a endbulb of Held synapses.

Figure 11. Synaptopathy of individual type I_a endbulb of Held synapses during ARHL.

(A) Representative morphology of individual type I_a endbulb of Held synapses in young mice. Panels from left to right: filled neurons (magenta) with reconstructed individual type I_a endbulbs, type I_a endbulbs alone, type I_a endbulbs (semi-transparent) with enclosed VGluT1-labled puncta (yellow), and VGluT1-labeled puncta alone. Top panels: example bushy neuron with only one type I_a endbulb of Held synapse. Bottom panels: example neuron with two type I_a endbulb of Held synapses, which are shown in red and blue respectively.

(B) Representative morphology of individual type I_a endbulb of Held synapses in middle-aged mice.

(C) Representative morphology of individual type I_a endbulb of Held synapses in old mice. Scales in A-C: 5 μm.

(D-F) Comparisons of individual type I_a endbulb volume (D), enclosed VGluT1-puncta volume (E), and VGluT1/endbulb volume ratio (F) among three age groups. Dunn’s multiple comparison test: NS, p > 0.05; ****p < 0.0001.

DISCUSSION

Convergence of AN synaptic inputs onto individual CN neurons and intensity processing

Information about sound intensity from each inner hair cell is conveyed by distributed firing rates amongst three subtypes of type I SGNs that operate over different ranges (Liberman, 1978). How this information is reassembled in CN neurons for central processing has remained unclear despite decades of research. Classical studies relied on tracing individual HRP filled AN fibers with known spontaneous rates (Fekete et al., 1984; Sento & Ryugo, 1989; Liberman, 1991; Ryugo & Sento, 1991; Liberman, 1993; Ryugo et al., 1996; Tsuji & Liberman, 1997; Ryugo, 2008). Because most filled AN fibers ended on different cells, it was not clear how different subtypes of AN synapses converge onto individual CN neurons. The physiological properties of these postsynaptic neurons were also unknown. Recent identification of molecular markers for different subtypes of type I SGNs (Petitpre et al., 2018; Shrestha et al., 2018; Sun et al., 2018) made this study possible. In particular, we used calretinin as the marker to differentiate AN central synapses of type I_a SGNs from those of other non-type I_a SGN subtypes. It remains a challenge to identify type I_b and I_c AN central synapses, which were collectively analyzed as non-calretinin-expressing (non-type I_a) synapses in this study.

Sharma et al (2018) showed that calretinin-expressing SGNs synapse onto inner hair cells on the pillar side and non-calretinin-expressing SGNs synapse on the modiolar side, which respectively correspond to the SGN subtype with high spontaneous rate/low threshold and those with medium/low spontaneous rate and medium/high threshold (Liberman, 1978, 1982b). Therefore, our findings about the innervation pattern of calretinin-expressing and non-calretinin-expressing endbulbs onto CN bushy neurons suggest that sound intensity information from different subtypes of SGNs is systematically transformed and processed in the central auditory system in a new dimension that extends even within a traditionally classified cell type. In this case, CN bushy neurons have been long recognized as a specialized cell population in the CN that dedicates to the processing of information about the temporal fine structure of sound (Joris et al., 1994a; Joris et al., 1994b; Manis et al., 2011). Our findings showed that within this bushy neuron population, cells differentiate physiologically based on the subtype composition of their AN synaptic inputs, i.e. the relative amount of information they receive about different sound intensities. Specifically, neurons with I_a-D inputs preferentially process information about low intensity sound, are more excitable and can fire sustained spikes to prolonged AN inputs (optimal for encoding persistent sound); while neurons with Non-I_a-D inputs are devoted to process information about medium and high intensity sound, are less excitable and only fire transiently to prolonged AN inputs (optimal for detecting loud auditory transients). It is worth noting that we did not classify our recorded bushy cells into the traditional spherical and globular bushy cell types, as it is difficult to unambiguously identify these neurons in mice (Lauer et al., 2013). It is unclear if spherical and globular bushy neurons indeed receive different convergence of inputs from type I_a and non-type I_a AN synapses.

Cochlear synaptopathy and AN central synaptopathy

Cochlear synaptopathy has been a research focus in recent years (Kujawa & Liberman, 2015; Liberman, 2017), which characterizes the anatomical loss of SGN peripheral synapses under pathological conditions. Due to the small size, it is hard to investigate whether the surviving cochlear synapses retain normal physiological function or are functionally compromised after damage. In this study, we characterized AN central synaptopathy from a different perspective. While we could not investigate the lost AN central synapses, we were able to elucidate the pathological changes of the surviving AN central synapses both physiologically using whole-cell recording and anatomically using post-hoc immunohistochemistry on the recorded neurons. As most cochlear synapses and SGNs (> 70%) survive during normal aging (Sergeyenko et al., 2013), it is of significant importance to understand the physiological changes of the existing synapses, whose function is critical in supporting the remaining auditory system and therefore the residual hearing of the subject. Indeed, our findings on AN central synaptopathy showed that the surviving synapses degenerate both anatomically and physiologically during aging, and may play significant roles in the development of ARHL.

AN synaptopathy and altered CN neuronal composition during ARHL

The observed AN central synaptopathy could be the consequence of aging at the synapse, as aging universally impacts every structure. In this case, however, it is hard to justify the preferential degeneration of non-type I_a synapses in ARHL groups. Therefore, AN central synaptopathy are likely secondary changes after primary damages of SGNs at the peripheral synapses, which lead to diminished physiological activity as well as loss of SGNs following long-term delay (Furman et al., 2013; Sergeyenko et al., 2013; Kujawa & Liberman, 2015; Liberman, 2017). The decrease in VGluT1-labeled puncta per bushy neuron (Fig. 9A–C) probably reflects both the degeneration of central synapses from the surviving SGNs (Fig. 10) as well as the loss of entire central synapses from lost SGNs. Given that low spontaneous rate/high threshold (type I_c) SGNs are preferentially damaged under pathological conditions (Furman et al., 2013; Shrestha et al., 2018), it is not surprising that central synaptopathy is more severe in non-type I_a AN synapses during ARHL. Such biased AN synaptopathy consequently results in relatively more abundant type I_a synapses among the surviving population.

ARHL is also associated with changes throughout the central auditory system (Frisina & Walton, 2006; Caspary et al., 2008; Profant et al., 2013; Groschel et al., 2014; Ouda et al., 2015), particularly the compromised ability in processing sound information during auditory tasks (Lorenzi et al., 2006; Grose & Mamo, 2010; Anderson et al., 2012; Atcherson et al., 2015). The link between cochlear pathology and central auditory processing deficits during ARHL remains unclear. For the first time, we show in this study that AN central synaptopathy, which presumably links to cochlear pathology, is associated with significant changes in the CN that include shifted neuronal composition of bushy neurons. It is conceivable that such changes alter the physiological responses across the CN neuronal population and the overall representation of auditory information. Since CN neurons initiate parallel pathways in the central auditory system (Cant & Benson, 2003), the shifted composition of CN neuronal population could be the trigger of many central auditory processing disorders including hyperacusis (loudness intolerance) and tinnitus (Vogler et al., 2011; Wang et al., 2011; Niu et al., 2013; Caspary & Llano, 2017; Luo et al., 2017; Radziwon et al., 2019; Sheppard et al., 2019; Shore & Wu, 2019). Further investigation of the synaptic and cellular changes of the CN is essential in understanding the relationship between cochlear pathology and central auditory processing disorders under different hearing impairment conditions.

Remarkably, CBA/CaJ mice in the middle-aged group had relatively normal ABR threshold but greatly reduced wave I amplitude (Fig. 1), a scenario that mimics hidden hearing loss (Sergeyenko et al., 2013; Kujawa & Liberman, 2015). Both biased AN synaptopathy (Fig. 9A–C) and shifted composition of neuronal population (Fig. 9D) were profound in this age group. The findings suggest that in addition to selective cochlear synaptopathy (Kujawa & Liberman, 2015), which is considered the primary cause of hidden hearing loss, significant changes also occur in central structure along the auditory pathway before hearing loss becomes substantial. In this case, decreased prevalence of bushy neurons with Non-I_a-D inputs (Fig. 9D) in middle-aged mice suggests that temporal processing of middle/high intensity sound from non-type I_a SGNs is compromised in CN neural circuitry and probably also in connected higher nuclei of the auditory system.

Response properties of CN bushy neurons during ARHL

CN bushy neurons were previously shown to vary in intrinsic membrane properties due to different composition of voltage gated ion channels (Manis & Marx, 1991; Rothman & Manis, 2003; Cao et al., 2007; Manis et al., 2019). We found in this study that the intrinsic membrane property as well as the AN evoked firing property of bushy neurons correlate with the proportion of type I_a synapses they receive, which suggests that bushy neurons are each optimized with a unique combination of anatomical and physiological features to process the distinctively converged auditory information from different SGNs. In young mice with normal hearing, bushy neurons showed a continuous ratio of distribution in the proportion of type I_a synaptic inputs they receive (Fig. 3M) as well as their physiological properties (Fig. 3N–O). The distribution became bimodal during ARHL, especially in the old mice group (Fig. 7M–O), presumably due to the preferential synaptopathy of non-type I_a synapses. We hence classified bushy neurons into two subgroups, including neurons with I_a-D inputs and those with Non-I_a-D inputs. Despite the changed neuronal prevalence (Fig. 9D), the physiological properties of two bushy neuron subpopulations remain distinct in all age groups during ARHL (Fig. 9E–J), suggesting that the function of CN bushy neurons are tightly regulated by the types of AN inputs they receive. It remains unclear whether the altered bushy neuron prevalence during ARHL is due to the loss of bushy neurons with Non-I_a-D inputs after biased AN synaptopathy, or the conversion of neurons with Non-I_a-D inputs to those with I_a-D inputs after adaptive changes like altered expression of voltage-gated ion channels.

Clinical implications

Cochlear implantation has been widely used in the clinic to treat profound hearing loss. Its performance relies on the function of SGNs, especially the SGN central structure. Age-related AN central synaptopathy is no doubt one of the key factors that limit the performance of cochlear implants. Indeed, duration of deafness is the most important predictor of the post implant outcomes (Rubinstein et al., 1999). The stimulation parameters of modern cochlear implants presumably activate all SGNs indiscriminately. This study suggests that, if possible, differential stimulation of different subtypes of SGNs in cochlear implants with enhanced activation of non-type I_a SGNs may compensate the biased AN central synaptopathy and altered CN neural circuitry, which would enhance compromised auditory processing, improve cochlear implant performance and overall hearing.

Supplementary Material

Supplementary - Statistical Summary Table

Supplemental Video 1. Three-dimensional reconstruction of the target bushy neuron and connected AN synapses.

Target bushy neuron was filled with Alexa Fluor 594 in magenta. Staining of calretinin is shown in red, which marks both type I_a AN fiber and synapses. Staining of VGluT1 is shown in either green, which marks the puncta from non-type I_a synapses, or yellow, which marks the puncta from type I_a synapses (overlap of red and green). See Materials and Methods for details. Background music: The Story Unfolds - Jingle Punks (https://youtu.be/_8iypSHvdx0).

KEY POINTS SUMMARY.

• Sound information is transmitted by different subtypes of SGNs from the ear to the brain.

• Selective damage of SGN peripheral synapses (cochlear synaptopathy) is widely recognized as one of the primary mechanisms of hearing loss, while the mechanisms at the SGN central synapses remain unclear.

• We report that different subtypes of SGN central synapses converge at different ratios onto individual target CN neurons with distinct physiological properties, and show biased morphological and physiological changes during ARHL.

• The results reveal a new dimension in CN neural circuitry that systematically reassembles and processes auditory information from different SGN subtypes, which is altered during aging and likely contributes to the development of ARHL.

• In addition to known cochlear synaptopathy, this study shows that SGN central synapses are also pathologically changed during aging, which collectively help us better understand the structure and function of SGNs during ARHL.