Age-Related Deficits in Binaural Hearing: Contribution of Peripheral and Central Effects

Abstract

Pure-tone audiograms often poorly predict elderly humans’ ability to communicate in everyday complex acoustic scenes. Binaural processing is crucial for discriminating sound sources in such complex acoustic scenes. The compromised perception of communication signals presented above hearing threshold has been linked to both peripheral and central age-related changes in the auditory system. Investigating young and old Mongolian gerbils of both sexes, an established model for human hearing, we demonstrate age-related supra-threshold deficits in binaural hearing using behavioral, electrophysiological, anatomical, and imaging methods. Binaural processing ability was measured as the binaural masking level difference (BMLD), an established measure in human psychophysics. We tested gerbils behaviorally with “virtual headphones,” recorded single-unit responses in the auditory midbrain and evaluated gross midbrain and cortical responses using positron emission tomography (PET) imaging. Furthermore, we obtained additional measures of auditory function based on auditory brainstem responses, auditory-nerve synapse counts, and evidence for central inhibitory processing revealed by PET. BMLD deteriorates already in middle-aged animals having normal audiometric thresholds and is even worse in old animals with hearing loss. The magnitude of auditory brainstem response measures related to auditory-nerve function and binaural processing in the auditory brainstem also deteriorate. Furthermore, central GABAergic inhibition is affected by age. Because the number of synapses in the apical turn of the inner ear was not reduced in middle-aged animals, we conclude that peripheral synaptopathy contributes little to binaural processing deficits. Exploratory analyses suggest increased hearing thresholds, altered binaural processing in the brainstem and changed central GABAergic inhibition as potential contributors.

Significance Statement

Older people often have difficulty communicating in everyday situations involving multiple speakers and other sound sources. Binaural processing facilitates hearing in such complex listening situations. Peripheral damage, such as synapse loss in the inner ear, is thought to be an important factor contributing to these difficulties. However, the role of central processing deficits has received less attention. Here, we investigate age-related binaural processing in Mongolian gerbils using brainstem responses, midbrain single-unit responses, positron emission tomography, and reward-based animal psychophysics. Behavioral deficits in binaural processing precede increases in audiometric threshold, as do changes in central inhibition and physiological binaural processing. Synapse loss in cochlear regions tuned to the stimulus frequency, however, does not change and unlikely contributes to age-related binaural deficits.

Introduction

Even before elderly humans show increased auditory thresholds in a pure-tone audiogram, they often experience compromised perception of signals well above their hearing threshold in quiet. Examples for such supra-threshold hearing deficits are speech reception in noise, sound discrimination based on temporal fine structure, or sound localization (Füllgrabe et al., 2015; Eddins and Eddins, 2018; Füllgrabe and Moore, 2018; Gallun, 2021). The observation of an age-related or noise-induced loss of inner hair cell synapses in animal models (Kujawa and Liberman, 2009; Gleich et al., 2016; Liberman and Kujawa, 2017) has led to the hypothesis that this peripheral synaptopathy may be a major cause of the dysfunctional supra-threshold perception in humans while it leaves auditory thresholds in quiet unaffected (termed “hidden hearing loss”; Schaette and McAlpine, 2011). Because synaptopathy can be demonstrated directly only postmortem, attempts to demonstrate the relation between synaptopathy and perception in humans have often been inconclusive (Plack et al., 2016; Bramhall et al., 2019). Furthermore, changes in central processing mechanisms likely accompany changes in the auditory periphery (e.g., Ibrahim and Llano, 2019; Dobri and Ross, 2021; Resnik and Polley, 2021). Thus, identifying the etiology of supra-threshold auditory perceptual deficits requires a study in an animal model that allows combining anatomical, electrophysiological and psychoacoustical investigations. Here, we explore binaural hearing in Mongolian gerbils as a proxy for supra-threshold hearing and its deterioration with age. It has been demonstrated that old gerbils localize sounds less accurately than young gerbils (Maier et al., 2008).

Binaural hearing relies on the processing of differences in acoustic information reaching the two ears. It is essential for sound localization in the horizontal plane and relevant for understanding speech under adverse conditions (Licklider, 1948; Moore, 2014; Joris and van der Heijden, 2019). Binaural hearing performance can be judged by examining a listener's ability to detect a tone in noise under different target tone/masking noise conditions that are related to the processing of binaural cues. A widely used paradigm in the study of human perception is the binaural masking level difference (BMLD) paradigm investigating tone detection in noise (Hirsh, 1948; Jeffress et al., 1962; Pichora-Fuller and Schneider, 1991). The tone detection in a diotic noise is facilitated if the tone exhibits dichotic differences compared with the detection of a diotic tone (Fig. 1a), and the BMLD is the difference in tone-in-noise detection thresholds between the two conditions. In one commonly used BMLD condition providing the largest effect, the noise has no phase difference between the ears and the tone is presented 180° out of phase (N0Sπ). In an alternative condition the tone is only presented monaurally (N0Sm). In both conditions, the reference has both diotic tones and noise (N0S0). The BMLD indicates the advantage gained from the processing of binaural cues provided by the temporal stimulus properties supporting the understanding of speech in complex acoustic scenes (for a review see Moore, 2014). BMLD in human subjects decreases with age and also with loss of audiometric sensitivity (Pichora-Fuller and Schneider, 1991; Santurette and Dau, 2012; Anderson et al., 2018) as does binaural processing in general (Ross et al., 2007; Moore, 2014; Füllgrabe and Moore, 2018; Vercammen et al., 2018), and the BMLD is thus very well suited to investigate age-related deficits in supra-threshold hearing.

Figure 1.

Gerbils’ tone detection thresholds and binaural masking level differences (BMLDs) depend on age and target/masker condition. a, Schematic of the tone-in-noise conditions presented in the experiments. In all conditions, the noise (N0) is presented identically to both the left (L) and the right (R) ear. In the N0S0 condition (top), the tone is presented identically to both ears (S0). In the N0Sm condition (middle), the tone is presented monaurally (Sm), that is, to only the left (depicted here) or the right ear. In the N0Sπ condition (bottom), the tone is presented with a 180° phase shift between the two ears (Sπ). b, A gerbil during trial initiation (top) and a schematic illustrating the acoustic stimulation using free-field loudspeakers and cross-talk cancellation (bottom). The indirect signal paths (dotted lines) are eliminated by adequate digital filtering, that is by mathematically inverting the system of direct and cross paths between the speakers and the ears. Only the direct signal paths (solid lines) reach the ears and allow for independent manipulation of the acoustic signal channels, including phase-shifting that was used in the experiments. c, Psychometric functions of young (orange, dash-dotted, n = 8), middle-aged (blue, dashed, n = 9) and old (pink, solid, n = 3) gerbils in the N0S0 condition (top), the N0Sm condition (middle) and the N0Sπ condition (bottom). d, Most tone detection thresholds (in dB SPL) measured in the N0Sπ condition (ordinate) were smaller than tone detection thresholds measured in the N0S0 condition (abscissa) in all age groups (young: triangles, middle-aged: circles, old: squares) and consequently fell below the unity line (top). Tone detection thresholds measured in the N0Sm condition (ordinate) fell clearly below the unity line only for young animals, and around or above unity line for middle-aged and old animals, respectively (bottom). e, Tone detection thresholds across conditions (N0S0, N0Sm, N0Sπ) were lowest for young animals (orange, filled black), followed by tone detection thresholds of middle-aged animals (blue, filled gray) and highest for old animals (pink, open). Box plots indicate the 25th and 75th percentiles (box) and the median (horizontal line) and the most extreme data points that are not outliers (whiskers). f, BMLDs resulting from the difference in tone detection thresholds in the N0S0 and the N0Sπ conditions (top) and BMLDs resulting from the difference in tone detection thresholds in the N0S0 and the N0Sm conditions (bottom) decreased significantly with the animals’ age (Pearson's correlation, N0Sπ-derived: r = −0.473, p = 0.035, n = 20; N0Sm-derived: r = −0.629, p = 0.003, n = 20).

In the present study, we investigate the BMLD in Mongolian gerbils using psychophysical, electrophysiological and imaging methods. Our study tests whether gerbils’ BMLDs decrease with age and if this is related to other measures of auditory function that change with age. Peripheral auditory function is evaluated using behavioral audiometric thresholds, histological counts of intact synapses in the inner ear, and the auditory brainstem response (ABR) measures indicating auditory-nerve function. Central auditory function is tested using ABR measures of binaural processing in the auditory brainstem, and central GABAergic inhibition is measured by [¹⁸F]flumazenil (FMZ) positron emission tomography (PET). Our study evaluates the consequences of age-related peripheral and central change (that is, processing beyond the auditory nerve) on binaural processing and perception providing an explanation for the binaural hearing deficits in humans with advancing age.

Materials and Methods

Animal subjects

Forty-four adult, agouti-colored Mongolian gerbils (Meriones unguiculatus, 20 females) were used in the study. All animals were bred in the animal facilities of the Carl von Ossietzky University of Oldenburg, Germany, and originated from animals purchased at Charles River laboratories. The care and treatment of the animals were in accordance with the procedures of animal experimentation approved by the Government of Lower Saxony, Germany (reference #33.9-42502-04-15/1990, Niedersächsisches Landesamt für Verbraucherschutz und Lebensmittelsicherheit). The experiments took place at the Carl von Ossietzky University of Oldenburg [behavioral testing, single-unit recording, auditory brainstem recordings (ABRs), immunohistochemistry] and at the Hanover Medical School (PET experiments). Four animals (age: >36 months) had to be excluded due to insufficient ABR responses. The other animals took part in 3–6 out of the six experiments described below. At the beginning of the experiments, they had an age of 4–38 months. Due to the organization and duration of measurements, a given animal's age differed between experiments. The age group an animal was assigned to could therefore differ between experiments. Age groups were defined as follows: young animals, <18 months old; middle-aged animals, 18 to <36 months old; old animals, ≥36 months. Animals of both sexes were randomly chosen from the stock at the animal facility and assigned to an age group for a single experiment depending on their actual age at the time of the experiments. The experimenters were not blinded to the animals’ age except for the experiment on immunohistochemistry and synapse counts.

Gerbils were housed singly (if no littermate was available) or in groups of up to three animals in macrolon type IV cages, containing paper towels for nesting, wooden chipping as bedding and wooden sticks for gnawing. Animals had free access to autoclaved tap water and food pellets (Altromin 1324 TPF, Altromin Spezialfutter GmbH & Co. KG). All animals remained in constant laboratory conditions with a 14/10 h day/night rhythm, a temperature between 21°C and 23°C, and a humidity of 45–55%.

Stimuli and test conditions

The same stimuli were used in the behavioral experiment, the PET experiment using ¹⁸F-fluorodeoxyglucose (FDG) and the single-unit recordings in the inferior colliculus (IC). They will be briefly described in the following. Stimulus details specific to the single experiments will be indicated in the appropriate sections. Stimuli were pure tones with a frequency of 700 Hz presented in a continuous broadband noise masker encompassing frequencies from 250 Hz to 10 kHz and presented identically to both ears (diotic, N0). Tones were presented binaurally, either identically to both ears (the diotic condition, S0) or with a 180° phase-shift between the two ears (the dichotic condition, Sπ) (Fig. 1a). In the behavioral experiment and the single-unit recordings, tone levels could assume values between 30 and 70 dB sound pressure level (SPL), and tone duration was 125 ms including 25 ms on/off cosine ramps. For the PET experiment, tones had an SPL of 65 dB and a duration of 400 ms followed by 600 ms of silence. The overall level of the noise was 65 dB SPL. Deviating, during the single-unit recordings in the IC, the overall level of the noise could assume values between 30 and 70 dB SPL. The combination of either type of tone and noise resulted in two test conditions: N0S0 and N0Sπ. A third condition, the N0Sm condition, was tested only in the behavioral experiment. The tone in this condition was presented monaurally (Sm), either to the left or the right ear (Fig. 1a). All other parameters were identical to the N0S0/N0Sπ conditions.

BMLDs were derived as the difference between the tone-in-noise detection thresholds in the N0S0 and the N0Sm conditions and the difference between the tone-in-noise detection thresholds in the N0S0 and the N0Sπ conditions. Positive BMLDs indicate a binaural advantage, that is, an enhanced tone detection in noise due to binaural processing. In humans, the largest binaural advantage (i.e., large positive BMLDs) is found for the difference in thresholds when detecting a diotic tone in diotic noise (N0S0, Fig. 1a, top) and when detecting a dichotic tone reversed in phase at one ear relative to the other in diotic noise (N0Sπ, Fig. 1a, bottom; Hirsh, 1948; Jeffress et al., 1962; Pichora-Fuller and Schneider, 1991). Large BMLDs in humans can also be found for the difference between tone detection in noise in the N0S0 condition and the instance when the tone is presented to only one ear in diotic noise (N0Sm, Fig. 1a, middle; Hirsh, 1948).

PET recordings

PET imaging using ¹⁸F-fluorodeoxyglucose (FDG)

Scans with FDG were performed in 15 young gerbils (7 females, age: 4–6 months), 10 middle-aged gerbils (6 females, age: 24 months) and 5 old gerbils (1 female, age: 35–38 months). During stimulation, animals were anesthetized via intra-peritoneal injection (i.p.) using a mixture of fentanyl (0.03 mg/kg bodyweight, Fentadon, Albrecht GmbH), medetomedine (0.15 mg/kg bodyweight, Domitor, Orion Pharma GmbH) and midazolam (7.5 mg/kg bodyweight, Dormicum, Roche). Animals were injected with 19.08 ± 1.73 MBq FDG in a volume of 0.3 ml physiologic isotonic saline (average and standard deviation over all test conditions and age groups). The tracer was applied via a catheter of one meter length, placed i.p. The acoustic stimulation was performed in a sound-attenuated box and followed the procedure described in the paragraph on acoustic stimulation during the single-unit recording in the IC (see below). Four conditions were tested: N0S0, N0Sπ, N0 noise only, and silence. One minute after starting the acoustic stimulation, FDG was injected, followed by 0.2 ml of heparinized saline (0.1%). Acoustic stimulation then lasted for 40 min, covering the whole uptake phase of FDG. Then the PET scan followed employing a high-resolution Siemens Inveon PET-computer tomography (CT) hybrid system (Siemens Medical Solutions). During the PET scan, the anesthetic state of animals was maintained using 1.0–2.5% isoflurane (Forene, Abbvie Deutschland GmbH & Co. KG), vaporized with an Ohmeda Isotec 4 Vaporizer (UniVet Porta, Groppler Medizintechnik). For image acquisition, a 30 min list mode scan was followed by a low dose CT scan. Emission tomograms with a voxel size of 0.8 × 0.8 × 0.79 mm³ and a matrix size of 128 × 128 × 159 were reconstructed using OSEM 3D algorithm, including attenuation correction. Attenuation correction was performed. Spatial coregistration was performed in PMOD3.7 software as described previously (Kessler et al., 2018). For brain regions of interest, standardized uptake value ratios (SUVrs; SUVs divided by the mean SUV in the pons) for spatially normalized 3D data sets of each animal were calculated. Further statistics were performed on voxel level using statistical parametric analysis (SPM8, Institute of Neurology, University College of London) implemented in MATLAB (MathWorks, MA, USA). Here, using SUVr-based images, significant differences between N0S0 and N0Sπ conditions on voxel-based level were calculated, resulting in an SPM t-map (Fig. 2a). The statistical significance level for all conditions was set at p < 0.01 (uncorrected).

Figure 2.

Brain activity as measured by ¹⁸F-fluorodeoxyglucose (FDG) positron emission tomography (PET) shows the same dependence on target/masker condition as the behavioral thresholds. a, The statistical parametric map (SPM) reflects results of the statistical voxel-based comparison (paired t-test) of FDG uptake (normalized to the pons) as an indication for auditory brain activation that is significantly different (p < 0.01, uncorrected) after stimulation with N0S0 compared to N0Sπ for each voxel (anatomical resolution element). For each dimension, an image of the three-dimensional SPM is displayed (top, z = 25; middle, y = 15; bottom, x = 44). b,e, Symbols indicating FDG uptake in (b) the inferior colliculus (IC) and (e) the auditory cortex (AC) of young (triangles, n = 15), middle-aged (circles, n = 10) and old gerbils (squares, n = 5) tend to fall below the line of unity indicating higher focal brain activity after N0S0 stimulation (abscissa) than after N0Sπ stimulation (ordinate). c,f, Brain activation measured by FDG-PET (standardized uptake value ratio, SUVr) in the IC (c) and the AC (f) of young (orange, filled black), middle-aged (blue, filled gray) and old gerbils (pink, open) was higher after N0S0 stimulation than after N0Sπ stimulation. Box plots indicate the 25th and 75th percentiles (box) and the median (horizontal line) and the most extreme datapoints that are not outliers (whiskers). d,g, BMLDs measured behaviorally (N0Sπ-derived) tend to increase with increasing BMLDs derived from the difference in FDG-PET values after N0S0 stimulation and N0Sπ stimulation in (d) the IC and (g) the AC (IC: r = 0.443, p = 0.065, n = 20, AC: r = 0.415, p = 0.087, n = 20, Pearson's correlation coefficient).

PET imaging using [¹⁸F]flumazenil (FMZ)

Scans with FMZ were performed in 9 young gerbils (3 females, age: 4–6 months), 10 middle-aged gerbils (6 females, age: 26 months) and 11 old gerbils (3 females, age: 35–39 months). Radiolabeling FMZ, PET acquisition, and data analysis were performed as described in Kessler et al. (2020). In short, 13.96 ± 2.04 MBq FMZ were injected transcutaneously in the vena femoralis of each animal. Simultaneously to tracer injection, a 60 min PET list-mode acquisition was started employing a high-resolution Siemens Inveon PET-CT hybrid system (Siemens Medical Solutions). During the scan, animals were anesthetized using 2.5% isoflurane (Forene, Abbvie Deutschland GmbH & Co. KG), vaporized with an Ohmeda Isotec 4 Vaporizer (UniVet Porta, Groppler Medizintechnik). Data were reconstructed iteratively into 32 consecutive frames (5 × 2 s, 4 × 5 s, 3 × 10 s, 8 × 30 s, 5 × 60 s, 4 × 5 min, 3 × 10 min) including a germanium-68 source-based attenuation-correction. Analyses were performed using PMOD3.7 software. Values of the nondisplaceable binding potential (BPnd) were calculated using the simplified reference tissue model (SRTM) with pons as reference region. For spatial coregistration, a PET template and a previously created voxel of interest atlas (Kessler et al., 2018) for auditory cortex (AC), IC, and medial geniculate body (MGB) and pons as a reference region were used.

Behavioral testing

Twenty gerbils were trained in a Go/NoGo tone detection task with operant conditioning with food rewards. Eighteen of these animals had been measured with PET before. Initial training lasted about 0.5–3 months, subsequent data acquisition about 1–2 months per animal. At the time of final data collection, the eight young gerbils (three females) were 11–17 months of age, the nine middle-aged gerbils (seven females) 20–33 months of age, and the three old gerbils (no females) 39–45 months of age. Animals were usually housed in groups of 2–3 animals in cages outfitted with litter and nesting material. They had unrestricted access to water but were food-restricted during testing. Their body weight was monitored daily during that time (65–95 g depending on an animal's sex and age). Food rewards in the experiments were 10 mg custom-made food pellets. Food was supplemented daily after the experiments, if necessary, to keep the animal's body weight and motivation approximately constant.

Experimental setup

Behavioral experiments took place in an infrared-illuminated sound-attenuating booth (IAC 403-A, Industrial Acoustic Company) lined with sound-absorbing acoustic foam (PLANO 50/0 covered with PYRAMIDE 100/100 WILLTEC; Seyboth & Co.). The reverberation time T30 of a broadband white noise measured with a microphone positioned in the typical position of the animal's head was <20 ms at frequencies above 600 Hz and 33 ms at 500 Hz. Gerbils ran in darkness on a mesh platform (5 mm openings, 11 × 31 cm) installed about 100 cm above the ground. An infrared camera (Conrad Electronics) allowed for monitoring of the animals from the outside. In the rear half of the platform, a pedestal and an adjacent semicircle-shaped headrest were installed, both equipped with light barriers (Fig. 1b). They allowed for the initiation of trials and the registration of animals’ responses, that is, the animals’ mounting and dismounting of the pedestal. At the front of the platform, a feeding bowl was positioned and connected, via a plastic tube, to a custom-built food dispenser, mounted about 50 cm above the platform. The setup was controlled by a Linux computer. For acoustic stimulation, two loudspeakers (Canton Plus XS) were mounted at 20° to the left and the right of the midline at a distance of 74 cm from the center of the semicircle-shaped headrest.

Acoustic stimulation

Stimuli were presented using cross-talk cancellation to eliminate undesired signal paths between the loudspeakers and the respective contralateral ears (for example, Damaske, 1971; Hartmann et al., 2016; Tolnai et al., 2017; Fig. 1b). That way, acoustic stimulation, though in the free-field, approximated acoustic stimulation during the use of headphones allowing the manipulation of the phase relations of the tones reaching the left and the right ear. Details of the stimulus generation can be found elsewhere (Tolnai et al., 2017). Stimuli were produced by a sound card (Hammerfall DSP Multiface II, RME; sampling frequency 48 kHz), passed through a digital signal processor for filtering (Tucker-Davis Technologies RP2.1), and amplified (Rotel, High Current 8 Channel Power Amplifier RMB-1048).

Procedure

The continuous noise masker was presented throughout the experiment. Gerbils initiated trials by mounting the central platform and positioning their nose in the semicircle with light barrier (Fig. 1b). Upon remaining in that position for 1–5 s (randomized waiting time interval), a tone was presented with a probability of 0.7 (Go stimulus, test trial). If the animal detected the tone and left the semicircle and the platform within 0.75 s after tone onset, a hit (H), that is a correct response, was registered and the animal was rewarded with food two pellets (totaling 0.02 g). If the animal missed the Go stimulus, that is, it remained in the semicircle, a miss was registered, and the next trial initiated automatically. With a probability of 0.3, no tone was presented during a trial (No Go stimulus, catch trial). If in such a case the animal remained in the semicircle for 0.75 s until after the end of the randomized waiting time interval, the trial was registered as a correct rejection. If the animal left the semicircle within that time, the trial was registered as false alarm (FA). Either S0, Sm_Left, Sm_Right, or Sπ tones were tested in a single session. Seven tone levels at 10 repetitions each were presented in pseudo-randomized order. Additionally, ten so-called warm-up trials at the beginning of the session, during which either tones at the highest level (n = 7) or no tone (n = 3) were presented totaling 110 trials per session (10 warm-up trials, 70 test trials, 30 catch trials). In some instances, sessions with only five repetitions per tone level were collected, that is, they contained 60 trials in total (10 warm-up trials, 35 test trials, 15 catch trials). Tone levels of targets were individually adjusted to test both sub- and supra-threshold levels. A test session was deemed valid if the animals detected tones of the two highest levels on average with a probability of ≥0.8 and if false alarms occurred with a probability of <0.2. At least two valid test sessions per condition were collected and merged, that is the analysis was based on 20 repetitions per tone level and 60 catch trials. A (scaled) cumulative normal distribution function with four parameters (slope, inflection point, offsets from 0 and 1) was fitted to the hit rates as a function of tone levels using a least-squares procedure for calculating the psychometric function. Tone detection threshold levels were derived from the fitted function at a fixed criterion hit rate. The criterion hit rate was set to correspond to a sensitivity value d’ of 1.8, derived from signal detection theory (Green and Swets, 2000) by d’ = Φ⁻¹(H)−Φ⁻¹(FA), where H is the hit rate, FA is the false alarm rate and Φ⁻¹ is the inverse cumulative standard normal distribution function.

Tone detection thresholds in silence

Data on the animals’ audiometric sensitivity were collected for all behavioral animals after the end of data collection for the tone-in-noise stimuli. The same procedure was used as described above. Tones without a masking noise were presented diotically from the free-field loudspeakers that were equalized to have a flat spectrum between 250 Hz and 11 kHz. Tones had a frequency of 700 Hz. Tone levels could assume values between −3 and 54 dB SPL. Tone duration was 125 ms including 25 ms on/off cosine ramps. Analogously to tone-in-noise detection threshold levels, hearing threshold levels were derived from fitted functions at a hit rate corresponding to a sensitivity value d’ of 1.8.

Single-unit recordings

Twelve adult, agouti-colored Mongolian gerbils were used in the electrophysiological experiments. Four young animals were used (≤18 months, two females) and eight middle-aged animals (25–28 months, four females); old animals were not used in this experiment. Ten out of 12 animals were used previously in the behavioral experiments, 10 out of the 12 animals had been measured previously with PET.

Anesthesia

A mixture of xylazine-hydrochloride and ketamine-hydrochloride (Rompun 2%, Bayer, xylazine 6 mg/kg body weight and Ketamin 10%, CP-Pharma, ketamine 135 mg/kg body weight i.p. initial dose) was used to initiate anesthesia. During the experiment, xylazine-hydrochloride and ketamine-hydrochloride in saline (0.33 and 7.5 mg/ml, respectively) were administered subcutaneously with a syringe pump (AL-1000, WPI), run at 300–450 µl/hr. The use of a homeothermic heating blanket (Harvard Apparatus) ensured a constant body temperature of 36°C–38°C throughout the experiment.

Recordings

Recordings were made in a sound-attenuated booth (IAC 401-A, Industrial Acoustics Company). Animals were held in the recording setup using a metal bolt, which was fixed to the bone of the skull on the bregma suture. The left IC was entered dorsally with tungsten electrodes (Frederick Haer & Co., impedances: 3.0–3.9 MΩ) driven by a piezo microdriver (EXFO 8200 Inchworm, Burleigh) through a craniotomy (diameter 2,000 µm) positioned 500 µm caudal and 2,000 µm left of the lambda suture, established coordinates in our laboratory for recording from the IC (e.g., Eipert et al., 2018). The reference electrode was a silver wire (EP1, WPI) positioned within skin tissue close to the craniotomy. The voltage signal was amplified (1.5–3.2 × 10⁴), filtered (150 Hz to 9 kHz) and digitized at 40 kHz sampling rate (Plexon MAP Data Acquisition system). Action potentials (spikes) were triggered using the box method (Plexon Rasputin software), and spike times were extracted.

Acoustic stimulation

Stimuli were generated using MATLAB (R2013b, The MathWorks), produced at a sampling rate of 48 kHz by a sound card (Hammerfall DSP Multiface II, RME) and presented through a closed-field delivery system via earphones (IE800, Sennheiser). FIR filters (300th order) derived from the impulse responses of the earphones, which were measured during the presentation of a sine sweep (0.02–22 kHz, logarithmic scaling at 1 octave/s) with microphones (ER-7C, Etymotic) placed at the entrance of the ear canal, corrected the earphones’ output during the experiment. They ensured flat output levels within ±3 dB for frequencies between 125 Hz and 20 kHz.

Auditory brainstem recordings

ABRs were measured in three young gerbils (1 female, age: 15–16 months), 13 middle-aged gerbils (8 females, age: 27–34 months) and one old gerbil (male, age: 39 months). ABRs were measured after behavioral testing, either in a dedicated measurement before sacrificing the animal for subsequent anatomical studies or as part of the electrophysiological recordings in the IC. In the latter case, the needle electrode for recording was placed in the skin just rostral of the bregma suture, in the other case, in the skin on the midline of the interaural axis. A needle electrode placed under the skin in the animals’ neck served as reference. ABRs were amplified (×10⁴) and band pass filtered (300 Hz to 3 kHz; ISO 80, WPI) and digitized at 48 kHz sampling rate using an A/D converter (Hammerfall DSP Multiface II, RME). Animals were anesthetized by an intraperitoneal injection of a mixture of xylazine-hydrochloride and ketamine-hydrochloride (Rompun 2%, Bayer, xylazine 6 mg/kg bodyweight and Ketamin 10%, CP-Pharma, ketamine 135 mg/kg bodyweight). Acoustic stimulation followed the procedure described in the paragraph on acoustic stimulation during the single-unit recording in the IC. Acoustic stimuli were clicks presented monaurally to the left or the right ear or binaurally. Binaurally presented clicks had an interaural time difference of 0 ms and an interaural level difference of 0 dB. Clicks were presented at a number of levels usually from 20 to 90 dB SPL (typically in 10 dB steps). Up to 500 responses were averaged. For more details, also on artefact rejection, see elsewhere (Beutelmann et al., 2015; Laumen et al., 2016b).

The slopes of the ABR wave I growth functions with respect to stimulus level were estimated from the responses to binaural clicks. The binaural interaction component (BIC) was derived by subtracting the sum of responses to monaural left and monaural right stimulation from the responses to binaural stimulation with 90 dB SPL clicks. The BIC was further characterized by extracting the amplitude of the largest negative deflection, the so-called DN1, in an analysis window starting 3.5 ms after stimulus onset and ending 6 ms after stimulus onset (Tolnai and Klump, 2020). ABR thresholds for clicks being presented monaurally to the left and right ear were estimated according to the procedures described by Suthakar and Liberman (2019).

Immunohistochemistry and synapse counts

The histology of the cochleae and the synapse count were done for sixteen animals. They were between 15 and 38.5 months old at the time of perfusion (young: n = 4, middle-aged: n = 11, old: n = 1). The experimenter was blinded to the age of the animals. Details on the transcardial perfusion with phosphate buffer and paraformaldehyde can be found elsewhere (Tolnai and Klump, 2020). Established protocols were used for histology and synapse count (Zhang et al., 2018; Steenken et al., 2021). In short, after fixation the cochleae were decalcified in 0.5 M EDTA, nonspecific binding sites were blocked using 3% BSA and the tissue was permeabilized in 1% Triton X-100. The following primary antibodies were applied: anti-MyosinVIIa to label hair cells (IgG polyclonal rabbit; Proteus Biosciences; catalog #25e6790; RRID: AB_10015251), anti-CtBP2 (C-terminal binding protein) to label presynaptic ribbons (IgG1 monoclonal mouse; BD Biosciences; catalog #612044; RRID: AB_399431) and anti-GluA2 to label postsynaptic receptor patches (IgG2a monoclonal mouse; Millipore; catalog #MAB397; RRID: AB_2113875). Secondary antibodies matching the hosts of the primary antibodies were used: goat anti-mouse (IgG1)-AF 488 (Molecular Probes Inc.; catalog #A21121; RRID: B_141514), goat anti-mouse (IgG2a)-AF568 (Invitrogen; catalog #A-21134; RRID: AB_10393343), and donkey anti-rabbit-AF647 (polyclonal secondary antibody; Life Technologies-Molecular Probes; catalog #A-31573; RRID: AB_162544). Cochleae were treated with an autofluorescence quencher (TrueBlack Lipofuscin Autofluorescence Quencher, 20× in DMF, Biotum) and subsequently dissected and mounted on microscope slides using Vectashield Mounting Medium (Vector Laboratories, H-1000). For an example of the staining see Figure 1 in the publication by Steenken et al. (2021). Synapses were analyzed at 3.8 mm from the apex of the cochlea corresponding to 2 kHz (Müller, 1996). We chose this site because it was available from the largest number of animals tested in the behavioral part of the study. Since it was demonstrated by Steenken et al. (2021) that the synapse counts for cochlear locations representing 0.5 and 1 kHz are correlated with those at the 2 kHz location (personal communication, 0.5 and 2 kHz: r = 0.461, p = 0.0047; 1 and 2 kHz: r = 0.4899, p = 0.00099), measuring synapse numbers at 2 kHz provides a measure of synaptopathy also at 700 Hz due to the correlation. Confocal stacks were obtained (Leica TCS SP8 system, Leica Microsystem CMS GmbH) and deconvolved (Huygens Essentials, Version 15.10, SVI). Following Steenken et al. (2021) procedure, the functional synapses of five hair cells, where both the pre- and postsynaptic labels were present, were manually counted using ImageJ (RRID:SCR_003070, http://fiji.sc/Fiji). Both cochleae of an animal were examined and the number of synapses averaged across ears.

Experimental design and statistical analyses

In the present study, we asked how the BMLD changes with age in the gerbil and if other factors than age influence the relation. Due to the time course of the experiments, a single animal's age differed between the experiments. In the description of the methods, we report the age at which the measurements were obtained. The individually measured values for the BPnd (that is GABA_A R binding), the DN1 amplitude, the slope of the ABR wave I growth function and the number of functional synapses were adjusted to each animal's age at the time when the behavioral experiments were conducted using the linear regression between the values measured in all individuals and those individuals’ ages. Those age-adjusted values were then used for further analysis.

Repeated measures analyses of variance (rm ANOVA) were used to detect statistical differences in tone detection thresholds [factors: target tone/masking noise condition (N0S0, N0Sm, N0Sπ) and age group (young, middle-aged, old)], differences in PET activation following tone-in-noise stimulation [factors: brain region (IC, AC), target tone/masking noise condition (N0S0, N0Sπ), and age group (young, middle-aged, old)], and differences in GABA_A R density [factors: brain region (IC, MGB, AC), age group (young, middle-aged, old)]. If Mauchly's test indicated that the assumption of sphericity had been violated a Greenhouse–Geisser correction was applied to the degrees of freedom. A one-way ANOVA was used to detect differences in hearing thresholds of the three age groups (young, middle-aged, old). In post hoc analyses, t-tests were performed using Bonferroni-adjusted α levels to correct for multiple comparisons (IBM SPSS Statistics, Version 27.0; RRID:SCR_016479). Exact p-values are given. An α level of 0.05 was considered indicating significant differences. If a one-tailed p-value would be significant (p ≤ 0.05) we refer to a “trend.”

Correlation analyses were done using MATLAB (R2019b; The MathWorks; RRID:SCR_001622) determining two-tailed Pearson's correlation coefficients and p-values. Correlation coefficients and p-values in the partial correlations were determined as described in Kim (2015). If data were not normally distributed as indicated by a significant one-sample KS test as in the case of the hearing thresholds and the slopes of the ABR wave I intensity growth function, we used rank-transformed data in the correlation analysis. The correlation analyses were regarded as exploratory. In particular, no adjustment for multiple testing was done, and p-values are displayed for descriptive reasons to indicate potentially meaningful effects. The p-values are considered as noticeable in case of p ≤ 0.05. We also report R² values. The results of the statistical analyses can be found in the Results section and in the figures and figure legends.

Results

Gerbils show an age-dependent BMLD

We investigated the effect of age on binaural processing, a supra-threshold hearing task, in a group of young, normal-hearing (n = 8), middle-aged, normal-hearing (n = 9), and old, hearing-impaired (n = 3) Mongolian gerbils. The animals’ hearing status was determined by ABR measurements and behaviorally by a tone-detection-in-silence task. The animals were trained to detect 700 Hz tones in broadband noise in a Go–NoGo paradigm receiving food rewards for the correct detection of tones. Target tones and masking noise were presented using virtual headphones, a technique that enabled the presentation of independent acoustic signals to the left and right ear like in headphone listening (Fig. 1b). We tested three target tone/masking noise conditions in the gerbils: N0S0, N0Sπ, and N0Sm (Fig. 1a). In all three conditions and in all three age groups, the probability of detecting a target tone in masking noise increased with increasing sound pressure level of the tone (Fig. 1c). In the young normal-hearing animals (orange, dash-dotted), the steep portions of the psychometric functions tended to occur at higher sound pressure levels in the N0S0 condition (top) than in the N0Sm condition (middle) or the N0Sπ condition (bottom). Both N0Sm and N0Sπ conditions led to tone detections at smaller sound pressure levels and thus a shift of the psychometric functions to the left in young animals. A similar but less pronounced shift occurred in the middle-aged, normal-hearing group (blue, dashed) but not in the old, hearing-impaired animals (pink, solid). Taking false alarm rates into account, that is, the probability of gerbils to respond even though no tone was presented, tone detection thresholds were determined applying signal-detection theory as the sound pressure level at which a detectability index d’ of 1.8 was reached. The comparison of tone detection thresholds on an individual basis emphasized the differences between age groups (Fig. 1d). Young animals’ tone detection thresholds in the N0Sπ and N0Sm conditions were lower than tone detection thresholds in the N0S0 condition as indicated by the triangles falling below the unity line. This was also the case for tone detection thresholds in the N0Sπ condition for the middle-aged animals but neither for thresholds in the N0Sm condition nor for thresholds in either dichotic condition in the old animals. Boxplots showing tone detection thresholds for the three age groups and target tone/masking noise conditions illustrate these differences (Fig. 1e). Lowest tone detection thresholds across all conditions were found for young animals (orange, filled black), highest ones for old animals (pink, open) and intermediate tone detection thresholds for middle-aged animals (blue, filled gray). Tone detection thresholds were significantly influenced by target tone/masking noise condition and age [rmANOVA, tone-in-noise condition (N0S0, N0Sm, N0Sπ): F_(2,34) = 5.045, p = 0.012, η² = 0.229; age (young, middle-aged, old): F_(2,17) = 10.533, p = 0.001, η² = 0.544]. Qualitatively, the same results persisted when we excluded the data of the three old animals, which showed elevated hearing thresholds (see below), in the statistical analysis. Tone detection thresholds were significantly influenced by the tone-in-noise condition (N0S0, N0Sm, N0Sπ) (F_(2,30) = 10.460, p < 0.001, η² = 0.411) and, importantly, by age group (young, middle-aged; F_(1,15) = 4.522, p = 0.050, η² = 0.232) indicating that binaural processing differed already between young and middle-aged animals. There was a significant interaction between the effects tone-in-noise condition and age group (F_(2,30)= 3.551, p = 0.041, η² = 0.191) reflecting the observation that the difference in the three conditions was larger in the young animals than in the middle-aged animals. Gerbils’ BMLDs resulting from the differences in tone detection thresholds between the N0S0 condition and the N0Sm or the N0Sπ condition decreased significantly with age (BMLD_N0Sπ: r = −0.472, p = 0.036, BMLD_N0Sm: r = −0.623, p = 0.003; Fig. 1f) indicating that age affects processes that are essential for the BMLD to occur.

FDG-PET measurements demonstrate neural correlate of BMLD

In an attempt to establish an objective measure of binaural processing, neural correlates of the BMLD have been investigated with imaging studies in human subjects (e.g., Wack et al., 2012, 2014; Gascoyne, 2015). In the present study, the activation of the auditory pathway following the exposure to tone-in-noise stimuli was investigated using FDG-PET in anesthetized gerbils (Fig. 2). Activation in the IC and the AC was measured in the form of standardized uptake values scaled to the pons region (Fig. 2a). Comparing PET activation in the IC and AC during N0S0 and N0Sπ stimulation on an individual basis demonstrated the stronger activation with N0S0 than N0Sπ stimulation indicated by the symbols falling below the line of unity (Fig. 2b,e). The lower PET activation found for the N0Sπ condition, an apparent paradox, can be explained by the single-unit responses in the auditory midbrain (see below). Note that the only difference between the N0S0 and the N0Sπ stimulation was the phase of the target tone. Any difference in PET activation with N0S0 and N0Sπ stimulation can therefore only be due to differences in binaural processing of these two stimuli. A rmANOVA showed that PET activation differed significantly between brain regions (IC, AC) [F_(1,27) = 643.610, p < 0.001, η² = 0.960] and between the tone-in-noise conditions (N0S0, N0Sπ) [F_(1,27) = 12.874, p = 0.001, η² = 0.323]. The interaction of brain region and tone-in-noise condition did not significantly influence PET activation indicating that while PET activation was stronger in the IC than in the AC, in both brain regions activation was stronger with N0S0 stimulation than N0Sπ stimulation (Fig. 2c,f). Age group (young, middle-aged, old) also significantly influenced PET activation [F_(2,27) = 8.696, p = 0.001, η² = 0.392]. PET activation measured in middle-aged animals was significantly stronger than in old animals (post hoc test, p = 0.001) and tended to be stronger than in young animals (p = 0.066). PET activation measured in young animals tended to be stronger than in old animals (p = 0.063). When excluding the data of the old animals (n = 5), which showed hearing loss as identified by behavioral hearing thresholds and/or based on ABR, PET activation with tone-in-noise stimulation was still significantly influenced by brain area (IC, AC; F_(1,23) = 15.379, p = 0.001, η² = 0.401), by tone-in-noise condition (N0S0, N0Sπ; F_(1,23) = 696.019, p < 0.001, η² = 0.968), and, importantly, by age group (young, middle-aged; F_(1,23) = 7.459, p = 0.012, η² = 0.245) providing further evidence that binaural processing is already altered in middle-aged animals. A PET-based BMLD (BMLD_PET) was calculated as the difference in PET activation with N0S0 stimulation and N0Sπ stimulation. There was no significant correlation between age and BMLD_PET in the IC (r = −0.021, p = 0.930) or AC (r = 0.171, p = 0.460). Both the BMLD_PET in the IC and in the AC tended to correlate with the BMLD_N0Sπ that was measured behaviorally (BMLD_PET IC: r = 0.443, p = 0.065; BMLD_PET AC: r = 0.415, p = 0.087; Fig. 2d,g). Both the lack of a significant correlation between age and BMLD_PET and the lack of a significant correlation between the behaviorally measured BMLD_N0Sπ and the BMLD_PET are likely due to the large variation of PET activation (measured as SUVr) within the different target tone/masking noise conditions. BMLD_PET is calculated from the differences of activation measures observed with N0S0 stimulation and N0Sπ stimulation. This difference is smaller than the activation values per se whereas the large variation within the age groups remains, exceeding the value of BMLD_PET (Fig. 2c,f). Additionally, the time delay between PET measurements, obtained at the beginning of the experimental series, and the behavioral data collection that were conducted in the same subjects up to several months after the PET measurements might have reduced the correlation.

Auditory midbrain units’ responses explain FDG-PET activation pattern

How can the finding that FDG-PET activation in the gerbil was clearly lower for N0Sπ stimulation than for N0S0 stimulation be related to results from earlier investigations in IC and AC units in the guinea pig? They showed different responses with increasing signal level during N0S0 stimulation compared with N0Sπ stimulation (Jiang et al., 1997; Gilbert et al., 2015). To consider potential species differences, we recorded responses of units in the IC of gerbils to the same tones in noise that were used in the behavioral experiments. The tones were presented with a signal-to-noise ratio of 0 dB either in the N0Sπ condition or in the N0S0 condition. Twenty-one units were recorded from the IC of young animals and 50 units from the IC of middle-aged animals (Fig. 3 top). The units had characteristic frequencies between 0.5 and 8 kHz (median: 2.8 kHz, quartile range: 1.4, 4.0 kHz). Characteristic frequencies were not significantly different between units from young and middle-aged animals (Mann–Whitney U test, p = 0.456). The units’ bandwidth was sufficiently large such that a response by the 700 Hz tone could be elicited. The units’ response rates changed depending on the target tone/masking noise condition. The difference in firing rates between the N0S0 condition and the N0Sπ condition between units recorded in young animals and units recorded in middle-aged animals was not significantly different (Mann–Whitney U test: p = 0.412), mirroring the lack of a significant correlation between the BMLD_PET and age. All units were therefore merged for the analysis. Firing rates were significantly smaller in the N0Sπ condition than in the N0S0 condition (Wilcoxon signed rank test, p = 0.002; Fig. 3, top) and the percentage change was significantly different from 0 (One-sample Wilcoxon signed-rank test, p = 0.020; Fig. 3, bottom). In the gerbil, these different relations in the response rate between the N0S0 and the N0Sπ conditions can therefore explain the decrease of PET activation with N0S0 and N0Sπ stimulation. This corresponds to the response patterns observed in in the guinea pig IC and AC where the population of neurons showing on average a substantial BMLD was characterized by a reduction of the rate response with increasing signal level in the N0Sπ condition but an increase of the rate response with increasing signal level in the N0S0 condition which would be consistent with the present PET results (Jiang et al., 1997).

Figure 3.

Auditory midbrain units’ firing rates depend on the target/masker condition. Firing rates of units in the IC (filled triangles: young, n = 21; open triangles: middle-aged, n = 50) measured in response to a phase-shifted tone presented in noise (N0Sπ) were smaller than firing rates of the same units measured in response to a tone-in-noise presented identically to both ears (N0S0) indicated by the symbols falling below the unity line (top, p = 0.002, Wilcoxon signed rank test) and by the skewed distribution of the percentage of firing rate differences between responses in the N0S0 and the N0Sπ conditions (bottom; p = 0.020, One-sample Wilcoxon signed rank test).

Relation between audiometric sensitivity and synapse numbers and the age dependence of BMLDs

In humans, binaural detection was shown to be affected by even small increases in hearing thresholds (e.g., Bernstein and Trahiotis, 2016; Pichora-Fuller and Schneider, 1991; Santurette and Dau, 2012) suggesting a potential contribution of the auditory periphery to dysfunctional supra-threshold hearing. Here we assessed gerbils’ audiometric sensitivity behaviorally and analyzed the number of intact synapses between inner hair cells and auditory nerve fibers after the end of the experiments, an approach not feasible in human subjects.

Similarly to the measurement of tone-in-noise detection thresholds, gerbils’ hearing sensitivity to 700 Hz tones was measured as animals’ tone detection thresholds in silence in a Go–NoGo paradigm. Behavioral hearing thresholds deteriorated significantly with age (r = 0.767, p < 0.001, Fig. 4a). A one-way ANOVA showed that age group (young, middle-aged, old) significantly influenced hearing thresholds (F_(2,19) = 57.421, p < 0.001). Post hoc tests demonstrated that hearing thresholds in the group of old animals were significantly higher than hearing thresholds in the group of young (p < 0.001) and middle-aged animals (p < 0.001). Notably, hearing thresholds of young and middle-aged animals did not differ significantly (p = 0.672) indicating the unimpaired ability of middle-aged animals to detect 700 Hz tones in silence. In partial correlation analyses (Kim, 2015), we controlled for the animals’ hearing thresholds on the relation between age and BMLD_N0Sm and age and BMLD_N0Sπ evident in the zero-order correlations (Table 1). Three outcomes for this exploratory approach (see Methods) are possible: (1) If the partial correlation is closer to 0 than the zero-order correlation, then the control variable (e.g., hearing threshold) might partly explain the correlation between age and BMLDs. (2) If the partial correlation is closer to +1 or −1 than the zero-order correlation, then the control variable obscured the correlation between age and BMLDs by adding unexplained variability to the correlation. (3) If the partial correlation approximates the zero-order correlation, then the control variable unlikely explains the correlation of age and BMLDs. BMLD_N0Sm did not noticeably correlate with age when controlling for animals’ audiometric sensitivity while it did in the zero-order correlation (Table 1). In contrast, the correlation between BMLD_N0Sπ and age hardly changed between controlling for audiometric sensitivity or not (Table 1). Thus, the increased hearing thresholds with age might explain the age dependence of BMLD_N0Sm but not the age dependence of BMLD_N0Sπ. Other age-related changes therefore likely contribute to deficits in binaural processing with older age.

Figure 4.

Behavioral hearing thresholds and peripheral and central measures of auditory function and their relation to age (first column), BMLD_N0Sm (second column) and BMLD_N0Sπ (third column). a–c, depict behavioral hearing thresholds (young: n = 8, middle-aged: n = 9, old: n = 3), (d–f) the number of functional synapses between inner hair cells and auditory nerve fibers (young: n = 6, middle-aged: n = 9, old: n = 1), (g–i) the slopes of wave I of the auditory brainstem response (ABR) growth function (young: n = 8, middle-aged: n = 8, old: n = 1), (j–l) the amplitude of the first negative deflection (DN1) of the binaural interaction component (young: n = 5, middle-aged: n = 7, old: n = 1) and (m–o) the GABA_A receptor (GABA_AR) binding potential measured by [¹⁸F]flumazenil (FMZ) PET (GABA_PET, BPnd; young: n = 5, middle-aged: n = 7, old: n = 3). Age groups are marked (orange triangles: young, blue circles: middle-aged, pink squares: old). Values in (d–o) are adjusted for the time between actual measurement and behavioral data collection using the linear regression between the measured values and the animals’ age at the time of the actual measurement (see Methods). The coefficients of determination (R²) and the number of animals included (n) are given. In (g–l), data are marked for the type of experiment they stem from (filled symbols: ABR measurements during electrophysiological recordings in the IC, open symbols: dedicated ABR measurements) In (m–o), R² is shown for GABA_PET in the IC (open symbols). R² for GABA_PET in the AC (colored symbols) and the MGB (black symbols) were very similar.

Table 1.

Pearson's correlation coefficients, p-values of correlations and coefficients of determination between BMLD_N0Sπ or BMLD_N0Sm with age for the zero-order correlation and when controlling for the following measures in partial correlations: hearing threshold, the number of intact synapses in the inner ear (synapse count), the slope of the ABR wave I growth function, the first negative deflection of the binaural interaction component of the ABR (DN1 of the BIC), and GABA_A receptor binding potential

	Correlation with age		Number of animals
	BMLDN0Sπ	BMLDN0Sm
Zero-order correlation	r = −0.472, p = 0.036, R2 = 0.223	r = −0.623, p = 0.003, R2 = 0.388	20
Measure controlled for in a partial correlation
Hearing threshold	r = −0.488, p = 0.034, R2 = 0.238	r = −0.375, p = 0.114, R2 = 0.141	20
Synapse count	r = −0.468, p = 0.050, R2 = 0.219	r = −0.619, p = 0.006, R2 = 0.383	16
Slope wave I growth function	r = −0.591, p = 0.010, R2 = 0.349	r = −0.434, p = 0.072, R2 = 0.188	17
DN1 of the BIC	r = −0.627, p = 0.007, R2 = 0.393	r = −0.446, p = 0.072, R2 = 0.199	13
GABAA R binding potential	r = −0.386, p = 0.126, R2 = 0.149	r = −0.451, p = 0.070, R2 = 0.203	15

Note that there was a strong correlation between the slope of the ABR wave I growth function and the amplitude of the DN1 component of the BIC (r = −0.821, p < 0.001).

While it was demonstrated that in the gerbil the number of intact synapses decreases with age (Gleich et al., 2016; Steenken et al., 2021), in our sample of animals, the number of intact synapses in the cochlear region most relevant to the behavioral stimuli did not decrease significantly with age (r = −0.334, p = 0.206, Fig. 4d). This may be due to the smaller age range of the animals tested here [11–39 months of (adjusted) age] than in the earlier studies, which also tested younger animals (Steenken et al., 2021: 3–46 months of age; Gleich et al., 2016: 4 to >35 months of age). But note that also in Steenken et al. (2021) the decline in synapse numbers from young to middle-aged gerbils was relatively small for frequencies in the range of the BMLD stimuli compared to frequencies of 4 kHz and above. The group of young behavioral animals [median (quartile range): 23.1 (21.6, 24.1), n = 6] had a similar number of intact synapses as the group of middle-aged gerbils [median (quartile range): 21.8 (21.6, 23.0), n = 9] (Mann–Whitney U test, p = 0.388). This indicated a similar state of the auditory periphery in young and middle-aged gerbils and mirrored the lack of differences in audiometric sensitivity between young and middle-aged gerbils. The only old animal analysed had 21 intact synapses per inner hair cell. When controlling for the influence of the number of functional synapses on the correlation of BMLD_N0Sm and BMLD_N0Sπ with age, in both cases the correlation coefficients remained noticeable (Table 1). Thus, the number of functional synapses in the low-frequency cochlear region of the behavioral animals is not likely to contribute to the age-dependent deficits in binaural processing.

Relation between ABR wave I and the age dependence of BMLDs

Loss of neural synchrony has been discussed as contributing factor to age-related deficits in binaural processing (Ross et al., 2007; Eddins and Eddins, 2018; Vercammen et al., 2018). Wave I of the ABR could serve as an indicator of neural synchrony (as discussed in Hamann et al., 2002). Yet, there is no evidence that there is deterioration of temporal coding properties of auditory nerve fibres with age (Scheidt et al., 2010; Heeringa et al., 2020). Wave I and its growth with increasing level can, however, be an indicator of an age-related change in the endocochlear potential affecting neural activity (Schmiedt, 2010). In the low-frequency range, the age-related reduction of the endocochlear potential was shown to decrease the slope of the ABR wave I growth function and to affect hearing thresholds less than at high frequencies (Schmiedt, 2010; Vaden et al., 2022). Thus, particularly in the middle-aged normal hearing gerbils of the present study, metabolic presbycusis is a candidate source for dysfunctional supra-threshold hearing. Moreover, the growth of the wave I response with level has been used as an indicator of synaptopathy (Liberman and Kujawa, 2017). In the present case, having ruled out synaptopathy in the low-frequency range in the cochlea of the gerbils, changes in the slope of the ABR wave I growth function might still stem from synaptopathy at higher-frequency regions of the cochlea (Gleich et al., 2016; Steenken et al., 2021) as the click stimuli used to measure the ABR wave I are dominated by high-frequency energy. Although the wave I growth functions may have multiple causalities, they still reflect the functional status of the cochlea. We therefore measured gerbils’ ABR wave I amplitude and determined the slopes of wave I growth functions with respect to stimulus level. The slopes of the wave I growth functions decreased significantly with age (r = −0.855, p < 0.001, Fig. 4g) suggesting that the growth of the wave I amplitudes with level decreased with age. When controlling for the influence of the slopes of the wave I growth function on the relation between BMLDs and age in partial correlation analyses, the correlation coefficient between BMLD_N0Sm and age decreased and its p-value increased (Table 1). The age dependence of BMLD_N0Sπ remained noticeable with a higher correlation coefficient (Table 1). Thus, while the slope of the wave I growth function contributed to the change of BMLD_N0Sm with age, it may be a confounding factor to the relation between BMLD_N0Sπ and age. The effect on BMLD_N0Sm might reflect the influence of metabolic presbycusis. The effect on BMLD_N0Sπ might indicate the influence of potential high-frequency synaptopathy, which is unrelated to the low-frequency behavioral sensitivity.

We also determined the monaural click-evoked ABR thresholds in both ears of the gerbil. Median difference in thresholds for the two ears was 5 dB (maximum 17 dB). This small difference renders it unlikely that it may have affected the BMLD (see Jerger et al., 1984). There was no correlation between the ABR threshold difference and the BMLD_N0Sπ (r = 0.110, p = 0.664).

BMLDs in relation to a neural marker for binaural processing in the auditory brainstem

The age-dependent decrease of BMLDs in gerbils might be linked to age-dependent deficits of binaural processing in the auditory brainstem. A neural marker for binaural processing in the auditory brainstem is the binaural interaction component (BIC; reviewed in Laumen et al., 2016a). In the gerbil it was shown to be age-dependent, particularly its first negative deflection, the so-called DN1 component. We determined the DN1 of the BIC from ABR measurements in the behavioral animals. The absolute amplitude of the DN1 decreased significantly with age (r = 0.880, p < 0.001, Fig. 4j). When controlling for the DN1 amplitude in the partial correlation analyses (Table 1) the correlation coefficient between BMLD_N0Sm and age decreased and its p-value increased suggesting that binaural processing in the auditory brainstem (Benichoux et al., 2018; Tolnai and Klump, 2020) affects the age dependence of the BMLD_N0Sm. The correlation coefficient between BMLD_N0Sπ and age increased when controlling for the DN1 amplitude suggesting that changes in the DN1 amplitude added unexplained variability to the age dependence of the BMLD_N0Sπ; it more clearly reveals the effects of changed central processing with age on the BMLD_N0Sπ once the effects contributed by processing in the auditory brainstem have been removed.

Relation between GABA_A receptor binding and age dependent BMLDs

Changes in central neural inhibition have been discussed as contributing to age-related auditory processing deficits and deteriorated performance with age in listening tasks (Ibrahim and Llano, 2019; Dobri and Ross, 2021; Resnik and Polley, 2021). Recently, GABA_A receptor (R) binding in the gerbil was measured with PET imaging using FMZ, a selective GABA_AR antagonist at the benzodiazepine binding site, and found to be reduced in older gerbils (Kessler et al., 2020). Notably, the reduction occurred before a peripheral hearing loss was apparent. A change in inhibition indicated by GABA_AR binding might thus explain the reduced BMLDs observed in the normal-hearing middle-aged gerbils tested in the present study. A change in inhibition could also explain the PET activation being higher in middle-aged animals than in young animals. We therefore measured FMZ binding to the GABA_AR in the AC, the auditory thalamus (MGB), and in the IC in the behaviorally tested animals, calculated as BPnd. We found that GABA_AR binding was significantly influenced by brain region [AC, MGB, IC] [F_{(1.066,12.797)} = 37.468, p < 0.001, η² = 0.757] and by age group (young, middle-aged, old) [F_(2,12) = 12.162, p = 0.001, η² = 0.670] (Fig. 4m). Post hoc tests showed a significant difference in GABA_AR binding between young and middle-aged gerbils (p = 0.006) and between young and old gerbils (p = 0.004) but not between middle-aged and old animals (p = 1.000). Thus, middle-aged gerbils’ GABA_AR binding was at the level of the old gerbils’ even though their hearing thresholds matched those of young gerbils’ (see above). When controlling for the GABA_AR binding in partial correlation analyses, both the correlation between BMLD_N0Sm and age and the correlation between BMLD_N0Sπ and age decreased and were not noticeable (Table 1). This might suggest that GABAergic inhibition, as measured by GABA_AR binding, may be a decisive factor contributing to deficits in binaural processing occurring with age.

Comparing BMLD_N0Sπ and BMLD_N0Sm

The partial correlation analyses showed different patterns of the effect of the control variables on the age dependence of BMLD_N0Sπ and BMLD_N0Sm. This might reflect different processing mechanisms of BMLD_N0Sπ and BMLD_N0Sm stimuli bearing in mind, however, that, due to the small sample sizes, all partial correlations fall within the 95% confidence intervals for the zero-order correlations of age with BMLD_N0Sm (−0.249 to −0.835) and BMLD_N0Sπ (−0.037 to −0.757), indicative of the exploratory nature of the analysis. The age dependence of the BMLD_N0Sm was reduced when controlling for the animals’ audiometric sensitivity while the age dependence of the BMLD_N0Sπ remained unaffected (Table 1). When controlling for the slope of the ABR wave I growth function and the DN1 of the BIC, the age dependence of the BMLD_N0Sm decreased while the correlation between age and BMLD_N0Sπ increased (Table 1). Common to both BMLDs, when controlling for the GABA_A R density, the correlations between age and BMLD decreased and when controlling for the synapse count in the low-frequency region of the cochlea, the correlations between age and BMLD remained essentially the same (Table 1). The bivariate correlations between BMLD_N0Sπ and BMLD_N0Sm and the control variables can be explored (Fig. 4) as another indicator for potential factors contributing to the age dependence of the BMLDs, keeping in mind that the bivariate correlations ignore the influence that the respective other factors exerted on the BMLDs. The values for r² in each subpanel of Figure 4 represent the amount of variance that could be explained by the respective factor, with an r² above 0.1 indicating a moderate correlation (Bosco et al., 2015). With the exception of the number of synapses per inner hair cell that did not explain much variance, all control variables appear to have at least a moderate effect on BMLD_N0Sm (Fig. 4, second column). In general, the amount of variance that could be explained by the effects of the control variables on BMLD_N0Sπ was smaller (Fig. 4, third column). The largest although weak effects on BMLD_N0Sπ were exerted by hearing threshold and GABA_AR density.

Taken together, this suggests that in the case of BMLD_N0Sm, the analysis of neural excitation patterns might affect tone detection, which is likely to be hampered when auditory-nerve activity is reduced. The effects of the slope of the ABR wave I growth function could indicate the importance of envelope encoding in the BMLD_N0Sm task being compromised by reduced levels of excitation. The influence of the DN1 amplitude on the relation between age and BMLD_N0Sm suggests that the processing of envelope information in the lateral superior olive may be relevant for the BMLD_N0Sm (Benichoux et al., 2018; Tolnai and Klump, 2020). The processing of BMLD_N0Sπ stimuli more likely relies on the analysis of temporal patterns present in the fine structure of the stimuli. Thus, changes in audiometric thresholds, wave I slope and DN1 amplitude, which are related to the level of excitation, are less influential for BMLD_N0Sπ. In either BMLD, the weakest effect was demonstrated for the number of functional synapses in the low-frequency region of the inner ear strengthening the idea that synaptopathy is unlikely to contribute to the age dependence of the BMLD that we observed in the gerbil but central processing reflected by the measure of GABA_AR density may do.

Discussion

Here, we demonstrated in the Mongolian gerbil that binaural processing deficits, i.e., reduced BMLDs, occur with increasing age. Those deficits occurred already in middle-aged animals with normal audiometric sensitivity and were more pronounced in old animals with impaired audiometric sensitivity. Measures of peripheral and central auditory function, obtained in the gerbils, also changed with age. Partial correlation analyses suggested that of those measures GABAergic inhibition may have an effect on the age dependence of both BMLDs. Audiometric sensitivity, the slope of the ABR wave I growth function, and binaural processing in the auditory brainstem may affect the age-dependence of the BMLD_N0Sm. In either BMLD, the partial correlation analysis suggested that the number of functional synapses in the low-frequency region of the inner ear unlikely contributes to reduced BMLDs. The combination of a range of methods, however, constrained the sample size limiting the statistical power. The partial correlation analyses must therefore be regarded as exploratory. Since the gerbils of the present study showed little decrease in synapse number in the cochlear frequency range relevant for the behavioral stimuli with age, synaptopathy is not likely to explain the decline in BMLD. Rather, central factors such as an age-related decline in inhibition as reflected by the BIC or by the change in the GABA_A receptor binding potential may contribute to the deterioration of binaural processing with increasing age. Furthermore, the finding that hearing threshold and the slope of the ABR wave I growth function are affected by age and their possible relation to the age-dependence of the BMLD hints at the importance of peripheral neuronal activity for binaural unmasking in BMLD.

Deterioration of binaural processing precedes impaired audiometric sensitivity

We obtained BMLDs as a measure of binaural processing, a supra-threshold hearing task, and found decreasing BMLD_N0Sm and BMLD_N0Sπ with advancing age of the animals, indicating age-dependent deteriorated binaural processing. Notably, the deterioration occurred in middle-aged animals before significant changes in behavioral hearing thresholds or in the number of functional synapses in the inner ear were apparent. Studies in humans that either statistically controlled for subjects’ audiometric sensitivity or included solely normal-hearing subjects also showed a decreased sensitivity to binaural cues or reduced magnitudes of neural markers of binaural processing with older age (Anderson et al., 2018; Eddins and Eddins, 2018; Füllgrabe and Moore, 2018; Ross et al., 2007; Vercammen et al., 2018; Gallun, 2021). Though those findings suggested additional factors beyond audiometric sensitivity as decisive factors for binaural deficits with age, those studies in human subjects could not link the condition of the cochlea as directly to the perceptual deficit to binaural perception as we could in the gerbil. Our findings emphasize that normal hearing sensitivity for tones in quiet is not sufficient for the sensitive processing of binaural information and can thus not explain supra-threshold hearing deficits. In such a condition resembling hidden hearing loss, more central processes are likely involved as also suggested recently by Henry (2022) reviewing studies in animals regarding the effects of synaptopathy on perception. However, the reduction of the slope of the ABR wave I growth function with increasing age can be observed without a change in synapse numbers in the behaviorally relevant frequency range. Though the reduced slope might be linked to a synapse loss in the high-frequency range of the cochlea (Gleich et al., 2016; Steenken et al., 2021), it can also be due to an age-related reduction in the endocochlear potential reducing sound-evoked firing in the auditory nerve (Schmiedt, 2010). Such a reduced activity could affect further central processing in a similar way as does synaptopathy.

Age impedes the use of binaural temporal fine structure cues

With increasing age the auditory system loses sensitivity for binaural phase and/or temporal envelope information (e.g., Ross et al., 2007; Walton, 2010; Mao et al., 2015; Füllgrabe et al., 2018; Füllgrabe and Moore, 2018). The cues used by young, normal-hearing and old, normal-hearing listeners were directly identified when listeners had to detect diotic and dichotic 500 Hz tones in noise (Mao et al., 2015). Temporal fine structure (TFS) cues contributed to tone detection in noise only in young, normal-hearing listeners. Further studies in humans stressed the effects of age and hearing threshold on the relevance of TFS cues in binaural hearing. When controlling for hearing threshold, an age-related decline in TFS perception persisted indicating that central processing deficits (e.g., due to a loss of inhibition) may be important (Füllgrabe et al., 2018; Füllgrabe and Moore, 2018). An age-related decline of the behavioral BMLDs in human subjects occurred only for 500 Hz tones but not for 4 kHz tones, a frequency at which TFS cues are unavailable, indicating that young but not old listeners were able to use binaural information related to TFS cues (Eddins and Eddins, 2018).

At the level of the single auditory nerve fibers, however, deficits with regard to the encoding of the TFS and the envelope of sounds were not apparent in quiet-aged gerbils apart from differences related to increased hearing thresholds with old age (Heeringa et al., 2020). Based on the altered processing of temporal information with age by auditory neurons at higher stations (Walton, 2010; Caspary and Llano, 2019), however, we hypothesize that temporal cues become less usable with increasing age due to altered central processes, most importantly due to an altered interaction of excitation and inhibition (reviewed in Caspary and Llano (2019)).

Changes in central inhibition may explain deficits in binaural processing with age

Though inhibition and its influence on auditory perception have been the focus of recent research (Lalwani et al., 2019; Dobri and Ross, 2021), little is known about the link between inhibition and perception with respect to binaural processing. The decreasing BIC with increasing age suggests that already at the level of the superior olivary complex reduced inhibition mediated mostly by glycine may affect binaural processing (Laumen et al., 2016a). GABAergic inhibition plays an important role in shaping auditory responses along the entire auditory pathway and contributes to temporal processing (LeBeau et al., 2001; Dehmel et al., 2010; Caspary and Llano, 2019). With increasing age, GABA levels in the auditory system decrease (Ouda et al., 2015; Caspary and Llano, 2019; Pal et al., 2019). When the concentration of GABA in the brain was increased pharmacologically, performance in a gap-detection task improved in old gerbils (Gleich et al., 2003) supporting the idea that GABA is essential for good temporal processing. Our results in the gerbil suggest that changes in central GABAergic inhibition with advancing age may also be responsible for compromised binaural processing as revealed by the decreased behavioral BMLD. GABA_AR binding was already strongly reduced in middle-aged, normal-hearing animals that showed a smaller BMLD compared to young normal hearing animals with similar peripheral hearing. This observation is in line with studies showing that the age-related change of GABAergic inhibition was independent of the integrity of the auditory periphery (Burianova et al., 2009; Kessler et al., 2020; Rogalla and Hildebrandt, 2020) and can also be found in nonauditory brain regions (Kessler et al., 2020; Chamberlain et al., 2021). It is thus conceivable that general central processing deficits due to age-related changes in GABAergic inhibition affect auditory binaural perception as revealed by the decreased performance in a binaural supra-threshold detection task such as the BMLD.

Inference from experimental evidence obtained in gerbils regarding human BMLD

Previous studies on age-related change in binaural processing in humans suffer from the limitation that physiological and anatomical conditions can only indirectly be assessed. In contrast, in the present study on gerbils we were able to narrow down the factors contributing to age-related deficits. The BMLD we collected in gerbils of different ages shared many characteristics with BMLD obtained from human subjects. Gerbil BMLD decreased with increasing age and hearing loss, as was observed in humans (e.g., Pichora-Fuller and Schneider, 1991; Santurette and Dau, 2012; Eddins and Eddins, 2018). It showed a large variability in hearing-impaired gerbils as it did in humans (Santurette and Dau, 2012). And finally, when controlling for hearing status, BMLD_N0Sπ persistently decreased with age in the gerbil, as also was observed in human subjects (e.g., Anderson et al., 2018; Eddins and Eddins, 2018). Like other temporal processing abilities in humans (e.g., monaural and binaural TFS sensitivity; Ross et al., 2007; Grose and Mamo, 2010; Füllgrabe, 2013), the BMLD was affected from midlife onwards in gerbils despite normal audiometric thresholds. Similarly to the present data in which the BMLD_PET tended to be related to the behaviorally measured BMLD, human studies found significant correlations between physiologically measured BMLD and behaviorally measured BMLD (Sasaki et al., 2005; Gascoyne, 2015; Clinard et al., 2017; Eddins and Eddins, 2018). Furthermore, in humans the magnitude of central auditory evoked potentials depended on the tone-in-noise condition (Ishida and Stapells, 2009; Anderson et al., 2018) and so did the brain activation determined by functional magnetic resonance imaging (Wack et al., 2012, 2014) as did the FDG-PET activation measured in the present study in the gerbil. Thus, the observations made in the present study in gerbils transfer to the observations in humans. Based on experimental evidence of the present study usually not available in human subjects, we infer that altered binaural processing in the brainstem reflected by the BIC, changed central GABAergic inhibition and/or metabolic presbycusis appear to outweigh synaptopathy as pivotal factor for deteriorated binaural processing with age.