Abstract
The mammalian cochlea receives efferent feedback from the brain. Many functions for this feedback have been hypothesized, including on short timescales, such as mediating attentional states, and long timescales, such as buffering acoustic trauma. Testing these hypotheses has been impeded by an inability to make direct measurements of efferent effects in awake animals. Here, we assessed the role of the medial olivocochlear (MOC) efferent nerve fibers on cochlear amplification by measuring organ of Corti vibratory responses to sound in both sexes of awake and anesthetized mice. We studied long-term effects by genetically ablating the efferents and/or afferents. Cochlear amplification increased with deafferentation using VGLUT3−/− mice, but only when the efferents were intact, associated with increased activity within OHCs and supporting cells. Removing both the afferents and the efferents using VGLUT3−/− Alpha9−/− mice did not cause this effect. To test for short-term effects, we recorded sound-evoked vibrations while using pupillometry to measure neuromodulatory brain state. We found no state dependence of cochlear amplification or of the auditory brainstem response. However, state dependence was apparent in the downstream inferior colliculus. Thus, MOC efferents upregulate cochlear amplification chronically with hearing loss, but not acutely with brain state fluctuations. This pathway may partially compensate for hearing loss while mediating associated symptoms, such as tinnitus and hyperacusis.
Significance Statement
The functional role of efferent innervation of the mammalian cochlea has remained in question. Here we show that the medial olivocochlear efferent system chronically potentiates cochlear sensitivity in response to removing the afferent signal but does not affect sensitivity in response to fluctuations in pupil-indexed brain state. While partially compensating for hearing loss, the efferent-mediated chronic potentiation may also contribute to associated symptoms of hearing loss, such as tinnitus and hyperacusis.
Introduction
The cochlea, like the retina and other sensory organs, is traditionally viewed as a feedforward system, stably providing sensory information to the brain. However, the cochlea receives neural feedback in the form of the olivocochlear efferent systems. Inputs from many sources, including widespread auditory and neuromodulatory structures, impinge on the efferent neurons in the olivary complex (Romero and Trussell, 2022). The efferents then carry these signals back into the cochlea via the olivocochlear bundle (Fig. 1). Lateral olivocochlear (LOC) efferent fibers synapse on the dendrites of the auditory nerve fibers under the inner hair cells and modulate hearing at the level of the auditory nerve (Lopez-Poveda, 2018). In contrast, the medial olivocochlear (MOC) efferent fibers synapse onto outer hair cells (OHCs) and release acetylcholine (ACh). Alpha9/10 ACh receptors detect this signal and activate calcium-dependent potassium channels, reducing OHC electromotility (Elgoyhen et al., 1994; Fuchs and Lauer, 2019). Supporting cells around the OHCs also have MOC efferent synapses (Zhao et al., 2022). Direct stimulation of MOC efferents has been shown to reduce basilar membrane vibration (Cooper and Guinan, 2006) and lower the voltage response of inner hair cells (Brown and Nuttall, 1984).
Inputs and outputs of the descending medial olivocochlear (MOC) efferent system. Medial olivocochlear (MOC) neurons located in the ventral nucleus of the trapezoid body (VNTB), one of the periolivary nuclei within the superior olivary complex (SOC; green), project to the cochlea, where they synapse on hair cells (bottom right). The MOC neurons receive descending input from the reticular activating system (RAS; dark blue), the inferior colliculus (purple), the auditory cortex, and other areas (top). These descending inputs could convey neuromodulatory brain state and acoustic context, respectively, to the cochlea. The descending RAS inputs also tightly regulate the size of the pupil. Therefore, pupil diameter in constant luminance conditions can be used as a readout of moment-to-moment changes in brain state. Central input onto the efferent MOC cells may affect cochlear function, since they project to the outer hair cells (OHCs) and modulate their ability to amplify the traveling wave, in a process known as cochlear amplification (top right). Sensitivity is the amount the organ of Corti vibrates in response to a sound stimulus. The characteristic frequency (CF) is the sound frequency at which vibration peaks when presenting low stimulus levels.
Functions of the MOC feedback pathway in hearing and sound-driven behavior are widely debated (Guinan, 2018; Lauer et al., 2022). On short time scales, ascending auditory input drives contralateral and ipsilateral MOC reflex arcs, slightly reducing otoacoustic emissions and compound auditory nerve responses shortly after the onset of an acoustic stimulus (Puria et al., 1996). These ascending inputs, combined with descending auditory and neuromodulatory inputs to the MOC efferents, are thought to enhance auditory signals, particularly in noisy backgrounds, by suppressing the response to unattended or ignored auditory or visual stimuli (Guinan, 2006; Delano et al., 2007; Abdala et al., 2009; Mishra and Lutman, 2014; Bidelman and Bhagat, 2015; Smith and Keil, 2015; Maruthy et al., 2017). Results on these top-down effects have been variable and often conflicting (Beim et al., 2019). This may, in part, be due to methodological challenges in directly measuring MOC effects and separating them from LOC or middle ear reflex effects in awake animals, including humans (Liberman and Guinan, 1998). One overarching hypothesis is that the top-down influences on the MOCs impart broad state dependence on the cochlea, analogous to the role of brain state in controlling pupil size for vision (McGinley, 2020).
Long time scale effects suggest that MOC efferents help “toughen” OHCs by sound conditioning so that they are less subject to loud noise trauma after chronic exposure to lower levels of noise (Kujawa and Liberman, 1997, 1999). Furthermore, acoustic injury to the cochlea is mitigated by MOC efferents via alpha9 receptors on OHCs (Maison et al., 2002; Slika and Fuchs, 2024). Another long-term role for MOC efferents may be adjusting cochlear gain and dynamic range after hearing loss. Besides having difficulty hearing quiet sounds, patients with hearing loss have poor speech understanding in noisy environments and tinnitus. They also experience hyperacusis, where certain sounds feel painfully loud to them but are not bothersome to others (Langguth et al., 2013, 2024). The common explanation for tinnitus and hyperacusis is that because the afferent auditory signal from the cochlea provides less information than the brain expects, central gain within the central nervous system, primarily the cortex, is increased as a compensation mechanism (Chambers et al., 2016; Salvi et al., 2017, 2021; McGill et al., 2022).
Thus, two functional categories for the MOC efferents have been proposed: dynamic changes in cochlear function to moment-to-moment changes in attention and brain states and chronic cochlear adaptations to maintain dynamic range while protecting against loud environmental conditions. Here we adapted optical coherence tomography (OCT) to measure organ of Corti vibration in awake mice and assess these possibilities. We hypothesized that deafferentation would increase amplification due to MOC efferent effects. Our data demonstrate that the MOCs produce chronic, but not short-term state-related, modulation of cochlear amplification.
Materials and Methods
Animal preparation
For OCT experiments, we used wild-type (WT) CBA/CaJ mice (JAX stock # 000654), Alpha9−/− (JAX stock #005696), and VGLUT3−/− mice (JAX stock # 016931) that were bred on a mixed background of C57BL/6J and 129S1. The number of males and females for each experiment were within 25% of each other. Animals were housed under a 12 h light/dark cycle with ad libitum access to food and water. When mice required anesthesia, intraperitoneal injections of ketamine/xylazine (100 and 10 mg/kg, respectively) were used, and body temperature was maintained at 37 ± 0.5°C with a heating pad. Supplemental doses of anesthesia were provided as needed to maintain areflexia throughout the procedure. All surgical and animal handling procedures were carried out in accordance with the ethical guidelines of the National Institutes of Health and were approved by an Institutional Animal Care and Use Committee. Depending upon where the experiments were performed, approval was either at the University of Southern California (OCT experiments and immunolabeling) or at Baylor College of Medicine [auditory brainstem response (ABR) and inferior colliculus (IC) recordings].
OCT vibrometry of anesthetized mice
We used our previously published experimental preparation for these studies (Applegate et al., 2011; Jawadi et al., 2015; Lee et al., 2015; Dewey et al., 2018, 2019, 2021; Badash et al., 2021a; Nankali et al., 2022; Quiñones et al., 2022). The custom-built system consisted of a swept laser (Insight Photonic Solutions, SLE-101) with center wavelength of 1,310 nm, bandwidth of 90 nm, and sweep rate of 100 kHz. The optical signal was transduced by a dual-balanced photodetector (Insight Photonic Solutions, BPD-1) and then digitized at 400 ms/s (AlazarTech, ATS9373). The axial resolution was 11.8 μm and the lateral resolution was 9.8 μm.
All mice for these experiments ranged in age from postnatal days (P) 37 to 54. We studied 35 mice (9 WT, 9 Alpha9−/−, 8 VGLUT3−/−, 9 Alpha9−/−VGLUT3−/−). After the induction of anesthesia, the skull was fixed with dental cement to a custom head-holder, and a ventrolateral surgical approach was used to access the left bulla. The bulla bone was then carefully chipped away to allow visualization of the otic capsule and ossicular chain. A small earphone (ER2SE; Etymotic Research) was positioned in front of the opening of the ear canal so it was ∼3 mm from the eardrum. Our 1D microscopic VOCTV system was then used to image and measure vibrations from within the cochlea peering through the otic capsule bone noninvasively. We studied the apical turn, approximately a half-turn down from the helicotrema. Vibratory measurement points were chosen to be local intensity maxima near the basilar membrane. Calibrated pure tone stimuli were used. After the measurements were completed, mice were killed by anesthetic overdose, and the cochleae were harvested for immunohistochemistry.
Dynamic optical coherence microscopy of subcellular activity
We also used a custom-built OCT experimental rig to image the organ of Corti at high resolution, noninvasively. We studied 10 mice (5 WT, 5 VGLUT3−/−). An upright microscope (Eclipse LV100, Nikon) was adapted to relay frequency-domain OCT signals to and from the tissue. A 165 nm broadband light source centered at 845 nm (T-850-HP, Superlum) and a spectrometer (Cobra-S 800, Wasatch Photonics) powered the OCT unit. We used a 10×, 0.3 NA objective (Plan Fluor MRH07120, Nikon) to image through the round window membrane (RWM) of anesthetized mice and collect image stacks of the organ of Corti. Because the objective has a high numerical aperture, the depth over which the OCT system can image is optically limited; thus this technique is named optical coherence microscopy (OCM). The axial resolution was 1.9 μm and the lateral resolution was 1.7 μm.
After finding the organ of Corti under the microscope, we collected a 3D volume stack for imaging purposes. Then, for the dynamic OCM measurements, we collected 150 B-scan (x–z) images at 130 Hz and performed FFT analysis on each voxel within the magnitude images. The frequency response of changes in pixel intensity was broken down into three bands (low, 0–5 Hz; middle, 5–15 Hz; and high, 15–64 Hz) and the mean of the power spectrum within each band calculated. This quantifies the frequency response and the magnitude of changes in reflectivity within cells, thus giving a measure of their subcellular activity.
Awake OCT mouse procedures
Mice used for awake OCT experiments went through several sequential steps. They ranged from 5 to 14 weeks of age. First, they had a head post glued to their skull. Buprenorphine (0.5 mg/kg) was administered subcutaneously ∼5 min before this procedure for pain control. A portion of the scalp was resected, the periosteum was elevated off the skull, and a lightweight (∼0.3 g) head post was affixed to the skull with dental adhesive (RelyX Unicem, 3M ESPE). The left pinna was resected to permit a direct view of the tympanic membrane through the bony ear canal. Animals were then awakened, housed individually, and allowed to recover for 3–5 d.
After recovery from placement of the head post, the mice were habituated to head fixation while standing on top of a cylindrical treadmill. The treadmill spun freely, and mice were permitted to spontaneously walk or stay still on top it. The treadmill was fixed to an air table within a double-walled sound booth. We progressively increased their time on the treadmill to 60 min over 3–5 d using walnuts as a treat to create a positive experience. Once the mouse was comfortable on the treadmill, experiments were run for several days over the following 1–2 weeks. Thus, each mouse had between two and five experimental recordings to pick from. During experiments, we recorded vibrations from the organ of Corti to assess cochlear physiology, pupil diameter as a measure of brain state, and body movements for artifact rejection. We selected the recording that by eye demonstrated many changes in pupil diameter (indicating change in brain state) with many high-quality vibratory tuning curves as assessed by a low noise floor and with minimal rejection artifacts. Thus, only one recording from each mouse was included in the averaged data.
Vibrometry of the organ of Corti in awake mice
Sound-evoked vibrations were recorded from the apical turn of the cochlea using a miniature version of our 1D OCT device (Lui et al., 2021). This device was small enough to be mounted on a flexible arm mounted to the air table that could be locked into place once a good imaging position was identified. The axial resolution was 11.8 μm and the lateral resolution was 26 μm.
The OCT device was positioned ∼10 mm from the entrance of the ear canal and once we got to a good imaging position, locked in place. We selected the imaging position by seeing the tympanic membrane on the built-in camera (x–y image) and the cochlea and organ of Corti on the OCT B-scan (x–z cross-sectional image). Then, vibrations were measured from single voxels representing points within the OoC at a location approximately one half-turn from the apex. Vibratory measurement points were chosen to be local intensity maxima near the basilar membrane.
Acoustic stimuli were presented free field through a horn tweeter (Fostex FT17H) positioned 6 cm from the left ear. Before recording, stimuli were calibrated using a 0.25 inch condenser microphone (B&K, 4939). Auditory stimuli were either single tones of varying frequency or multitone Zwuis complexes (van der Heijden and Joris, 2003). The single-tone stimuli ranged from 5 to 13 kHz in 1 kHz steps and were presented at 50 dB SPL for 100 ms to 14 mice (6 WT and 8 Alpha9−/−). The multitone stimuli had 18 spectral components, spanning from 4.5 to 13.8 kHz with an average spacing of 541 Hz. All components had equal amplitudes and were presented from 10 to 80 dB SPL in 10 dB steps for 100 ms to 12 different mice (6 WT and 6 Alpha9−/−).
After collecting data in the awake state, we anesthetized the mice and kept them in the same position. Thus, we could acquire OoC vibratory responses from the same cochlear location in both awake and anesthetized conditions in the same animals down the ear canal. The only adjustment was that a heating pad was positioned under the mouse's body during these anesthetized recordings. We used the multitone stimuli to study 12 mice (6 WT and 6 Alpha9−/−). These were the same 12 mice used for the awake measurements with multitone stimuli described above.
Artifact rejection during vibrometry in awake mice
When we initiated these experiments, we first tested the rigidity of the head fixation system. We imaged first the head post and then the cochlea in mice that were awake and moving on the treadmill. Even while running, the post and the area of the skull glued to the post did not move. In contrast, the cochlea was found to move up to 50 µm, presumably because of flex in the suture lines within the different portions of the skull. When the mouse stopped moving, the cochlea always returned to its previous position. These movements created artifacts in our vibrometry recordings. Thus, we used a camera to assess for movements of the mouse and did not analyze data that was collected when the mouse was moving. With this approach, we could remove artifact-ridden data from our analysis in an objective manner.
Pupillometry for OCT experiments
During the awake mouse experiments, brain state was tracked using pupillometry, which has previously been demonstrated to track neuromodulatory brain state in the auditory neocortex and thalamus and hippocampus (McCormick et al., 2015; McGinley et al., 2015a,b; Reimer et al., 2016). The right eye was illuminated with infrared light (850 nm, LED Engin, LZ1-00R602), and images were acquired at 10 Hz using a camera (Allied Vision, Alvium 1800 U-501m) fitted with a zoom lens (Newport, M-5X) and bandpass optical filter (MidOpt, BN810). Additionally, an ambient UV light (365 nm, Seoul Viosys, TCN6MA1A) was used to provide low-intensity illumination. First, we turned up this light such that the pupil of the animal was approximately mid-range in diameter. We further adjusted it until we were confident that it increased with walking and decreased when still to verify that the measurement provided adequate sensitivity to detect brain state.
Using custom software, pupil diameters were determined in real time. The images of the pupil were thresholded by signal intensity and fitted with an ellipse. The diameter of a circle with an area equivalent to the ellipse was calculated for each image frame and taken to be the pupil diameter. Pupil diameters were normalized by dividing by their maximal value in each session. Timestamps of individual video frames were used to associate the pupil measurements with the vibrometry and neural recordings.
Assessment of body movements for OCT experiments
Body movements were monitored with another camera (Jiusion 40×) situated in front of the mouse at ∼15 cm. The sum of absolute differences method (Richardson, 2003) was used to detect animal movements in the video recordings. As nothing in the frame except the mouse could move, the total motion energy of the pixels of the video was calculated as an index of perceptible movements of the mouse. This was computed as follows:
Brainstem recordings in awake mice
A total of 19 WT C57BL/6J mice were used (Jackson Laboratory). Animals at 7–8 weeks postnatal were anesthetized with isoflurane (1.5–2% in oxygen) and were first implanted with a head post for securing them on the head-fixed experimental setup.
For IC experiments, immediately after implanting the head post, a 3 mm craniotomy was made over IC, centered at the coordinates: AP = −5.1 mm, ML = 1 mm. The craniotomy was covered with a glued stack of two coverslips (3 and 5 mm). A small hole (∼0.5 mm) was made through the coverslips beforehand to provide access to the Neuropixels probe. The coverslip stack was sealed around the edges of the craniotomy using Metabond. The hole in the coverslip was covered with silicon polymer (Kwik-Vast). After 3–4 d of postsurgery recovery, animals were habituated to the cylindrical treadmill for a single 1.5 h session. Recordings were conducted starting the day after habituation.
Pure tones of duration 40 ms (with 1 ms rising and falling cosine gate) and carrier frequency ranging from 3 Hz to 96 kHz in steps of 0.2 octaves, at two sound levels (50 and 60 dB SPL), were presented in random order to the animals. The interstimulus interval was a silent period lasting 140 ms. After each unique stimulus was presented 20 times, there was a 30 s silent period before the subsequent randomized collection of tones. Each unique stimulus was presented 500 times in each session. Neuropixels probes 1.0 were used to acquire LFP responses in IC, sampled at 2,500 Hz. At the headstage, LFP was filtered with a first-order passive low-pass switched-capacitor filter with a cutoff frequency of 500 Hz and a high-pass filter at 0.5 Hz. The gain was set to 250X. A total of 22 sessions were conducted across 13 mice.
To confirm the electrode location in IC, Neuropixels probes were coated with DiI before insertion. The brain was harvested after transcardial perfusion with 4% paraformaldehyde. Sagittal sections with thickness of 100 μm were made using a vibratome and stained with DAPI before visualization.
For ABR experiments, four stainless steel screw electrodes with presoldered silver wire and gold pin (Plastics One) were implanted at the following locations:
Active electrode: anterior-posterior (AP) = −6.25 mm, medial-lateral (ML) = 3 mm, ipsilateral to the speaker
Reference: AP = −2.0 mm, ML = 0 mm
Ground: AP = 4 mm, ML = 0 mm
Additional ground: AP = −6.25 mm, ML = 3 mm, contralateral to the speaker
Mice were habituated as described above for IC experiments. Pure tones of duration 5 ms with 1 ms rise–fall time, carrier frequency of 8,16 or 32 kHz, and five sound levels (40–80 dB SPL, in steps of 10 dB) were presented in random order to the animals. The interstimulus interval was a silent period lasting 20 ms. After each unique stimulus was presented 500 times, there was a 30 s silent period. Overall, in a session, each unique stimulus was presented 15,000 times. ABRs were bandpass filtered between 1 and 10,000 Hz and amplified 100× by a bioamplifier (A-M Systems Model 1,800) before sampling by the NI A/D board at 30 KHz. Stimuli were presented in opposite polarity successively to minimize the contribution of microphonic potentials to the ABR waveforms. A total of 32 sessions were acquired from six mice.
Pupillometry in IC and ABR experiments was carried out as described above for OCT experiments but with slight differences in methodology. Images of right eye was acquired using Basler GigE camera and a fixed focal length lens (55 mm EFL, f/2.8, for 2/3″; Computar) placed at ∼8 inches from the animal. An infrared light source (850 nm, DigiKey) was used to illuminate the eye. A near-ultraviolet LED (405 nm) was placed above the animal head to provide ambient illumination. The intensity of the near-uv LED was adjusted for each session to capture the full range of pupil fluctuation, by increasing the light level until the pupil slightly constricted while the animal was walking on the treadmill. Frames were acquired at 15 Hz using custom LabVIEW code. Eye videos were labeled by DeepLabCut to mark pupil boundary using 8 points, and the pupil size was computed as the area of the ellipse fitted to these points.
Walk velocity was acquired by measuring the wheel motion using a rotary encoder (Accu, SL# 2204490). Data during which the animal was walking on the wheel were excluded.
All analyses were performed using MATLAB. LFPs were low-pass filtered at 300 Hz. Responses to only 60 dB were used in the subsequent LFP analysis. To remove responses from the putative dorsal cortex of IC, LFPs on electrodes lying within 300 μm from the IC surface were excluded. Electrode sites with significant responses around trough (N1; 2 ms window centered at N1) to at least one of the 26 frequencies presented (Wilcoxon rank sum test, alpha = 0.01, compared with 10 ms preceding baseline period) were included for further analysis. Peaks and troughs were identified using findpeaks in MATLAB. For each responding electrode site, the best frequency (BF) was calculated as the frequency with maximum (minimum for negative-going responses) response at N1. Responses to only the BF were used for state dependence analysis. Median pupil size in a 500 ms window preceding each BF stimulus (500 trials) was taken as the measure of neuromodulatory brain state for that trial (McGinley et al., 2015a,b). For each session, the distribution of pupil sizes for all stimuli was divided into 10 bins making sure that each bin contained at least 30 trials of each unique stimulus and that the remaining bins were approximately equal in pupil size.
ABRs were bandpass filtered between 300 and 3,000 Hz. ABRs in response to only 16 kHz were used in the analysis. The sound level of ABRs used in the analysis were chosen separately for each session (ranging from 60 to 80 dB SPL) by selecting the lowest sound level for which all five ABR waves had significant peaks. Given the higher number of stimulus trials in ABR, the distribution of pupil sizes preceding each stimulus was divided into deciles using prctile function in MATLAB. ABR peaks were identified using findpeaks in MATLAB.
Pupil-binned mean peak responses at each electrode site/session (LFP or ABR) were bootstrapped 1,000 times to fit a quadratic model to compute R2. Briefly, mean response in each bin pupil was computed at each electrode site for LFP and each session for ABR. This population data was bootstrapped, and a quadratic model was fit on the mean responses as a function of pupil size using polyfit function in MATLAB. The performance of the model was cross-validated using the leave-one-out method and R2 was computed as follows:
Prestin immunolabeling to measure OHC prestin-containing membrane surface area
Our immunohistological methods were previously published (Badash et al., 2021b). We studied 26 mice (6 WT, 7 Alpha9−/−, 7 VGLUT3−/−, 6 Alpha9−/−VGLUT3−/−). Briefly, excised cochlea were fixed in 4% paraformaldehyde at room temperature for at least 30 min. The cochleae were then decalcified, and the sensory epithelium was dissected and immunolabeled for prestin (1:200, Santa Cruz Biotechnology), myosin VIIa (1:400, Proteus BioSciences), and neurofilament H (1:1,000, MilliporeSigma) and stained for actin using phalloidin 405 (1:200, Thermo Fisher Scientific). Secondary antibodies were Alexa Fluor 488 (1:500, Thermo Fisher Scientific), Alexa Fluor 555 (1:500, Thermo Fisher Scientific), and Alexa Fluor 647 (1:500, Thermo Fisher Scientific). Tissues were mounted onto slides using Fluoromount-G (Thermo Fisher Scientific), and confocal image stacks were acquired with a Leica SP8 microscope with a 63× oil-immersion objective (1.4 NA; Leica Microsystems).
3D reconstructions of image Z-stacks were segmented for prestin using the machine learning algorithm in Imaris (v 10.1.1, PerkinElmer). A mask was generated from which the prestin-containing membrane surface area for each OHC could be calculated. For the middle portion of each cochlea (CF 11.5–26 kHz), the average prestin intensity was quantified as the average prestin intensity of the ∼45–70 cells contained within the field of view.
Experimental design and statistical analyses
Data were collected using customized software written in Python, plotted using Matlab 2023a (MathWorks) and analyzed statistically using RStudio (Build 561, Posit software) running R (4.3.1, The R Foundation). Reported values are the mean ± standard error (SE) unless otherwise noted. For single comparisons, two-tailed unpaired or paired Student's t tests were used to assess statistical significance at the p = 0.05 level. For comparisons of multiple groups, ANOVA was performed. If p < 0.05, this was followed by post hoc t tests with Bonferroni’s correction for each pair combination.
For comparisons of vibratory responses, regression was performed using linear mixed effect models that included random effects using our published technique (Oghalai, 2023). We fit the vibratory data with a third-degree polynomial function and included stimulus frequency and stimulus level as cross-terms. We ran the fit from the LMER function into an ANOVA function, which provides one p value that describes the relationship between pupil size and the curves (i.e., magnitude or phase) crossed with frequency and level. When differences were found, t tests at different frequency/level combinations were performed as needed. Comparisons between other curves, such as gain, Q10dB, etc. were performed in a similar fashion.
Significance is indicated in the figures by asterisks (*p < 0.05, **p < 0.005, ***p < 0.0005). Also, we independently compared males and females within each cohort for every statistical test to assess for sex differences. There were no statistically significant sex differences in the effects. Thus, males and females were combined for the presented results.
Data supporting the findings of this study are available on our GitHub site (Oghalai, 2024).
Results
The MOC efferent pathway increases cochlear amplification with hearing loss
As a first assessment of how cochlear amplification was affected by the MOCs, we compared CBA/CaJ wild-type (WT) mice to VGLUT3−/− mice, a well-established model of profound hearing loss due to defective neurotransmission across the inner hair cell→auditory nerve synapse (Ruel et al., 2008; Seal et al., 2008; Akil et al., 2012). We hypothesized that if chronic noise exposure makes mice less susceptible to noise exposure via the MOC efferent pathway (Kujawa and Liberman, 1997, 1999), presumably by downregulating the OHC electromotile response, chronic hearing loss should have the opposite effect. Instead of measuring threshold shifts after noise exposure—which are affected by synapse and other changes, including through LOCs—we directly measured cochlear amplification by recording basilar membrane vibratory tuning curves from a position that is half a turn down from the apex in anesthetized mice (Lee et al., 2015).
Representative recordings demonstrate increased vibratory magnitudes to low-intensity stimuli in VGLUT3−/− mice (Fig. 2). This is most easily seen as higher sensitivity to sounds presented at 10 dB SPL (e.g., compare sensitivity curve peak against the red dotted lines in B and F). Importantly, both the vibratory magnitudes to high-intensity stimuli and the phase responses were similar between the two mouse strains. This is consistent with a difference in cochlear amplification but no difference in passive cochlear mechanics. Time domain recordings in response to a 30 dB SPL click stimulus revealed larger vibratory responses and more ringing behavior in VGLUT3−/− mice compared with wild-type mice (compare Fig. 2D vs Fig. 2H). Again, this pattern is consistent with an increase in cochlear amplification in VGLUT3−/− mice.
Deafferented VGLUT3−/− mice have more cochlear amplification than WT mice. Representative data from one WT mouse (top row) and one VGLUT3−/− mouse (bottom row). A, E, Raw vibratory responses to sounds of different frequencies (4–15 kHz) and intensities (10–90 dB SPL). The characteristic frequencies were ∼9–10 kHz. B, F, Sensitivity curves were created by referencing the vibratory magnitude to the stimulus intensity. For lower stimulus intensities (10–30 dB SPL), VGLUT3−/− mice demonstrated more cochlear amplification, as noted by the responses above the dotted red line. C, G, Phase responses were similar between the genotypes. The phase above 10 kHz is noisy because of the low vibratory magnitude and, thus, not significant. D, H, Responses to a 30 dB click revealed larger amplitude vibrations and more ringing (arrow) in VGLUT3−/− mice, consistent with more cochlear amplification.
To determine if the MOC efferents were the causative factor in the increased cochlear amplification in these hearing-impaired mice, we crossed VGLUT3−/− mice with Alpha9−/− mice. Alpha9−/− mice lack this acetylcholine receptor subunit on their OHCs so they do not respond to MOC efferents (Elgoyhen et al., 1994; Brown and Vetter, 2009). Littermates were then studied so we could compare four different genotypes: wild-type, Alpha9−/−, VGLUT3−/−, and Alpha9−/−VGLUT3−/− mice. Only +/+ and −/− genotypes were used; we did not study any mice that were heterozygous for VGLUT3 or Alpha9. All four genotypes demonstrated similar vibratory responses to high-intensity stimuli, but the VGLUT3−/− mice demonstrated larger vibratory responses to low-intensity stimuli (Fig. 3). This was particularly noticeable in the sensitivity curves, where only VGLUT3−/− mice had peak sensitivities above 75 dB re:1 nm/Pa. Statistical analyses revealed no differences in vibratory magnitudes between WT, Alpha9−/−, and Alpha9−/−VGLUT3−/− mice. However, VGLUT3−/− mice had larger vibratory responses compared with Alpha9−/−VGLUT3−/− mice.
Increased cochlear amplification in VGLUT3−/− mice requires a functional MOC efferent pathway. Averaged vibratory responses for (A) WT mice (n = 9), (B) Alpha9−/− mice (n = 9), (C) VGLUT3−/− mice (n = 8), and (D) VGLUT3−/−Alpha9−/− mice (n = 9). Increased vibratory responses for lower sound levels are noted in the magnitude responses (first column) and sensitivity curves (second column) of VGLUT3−/− compared with VGLUT3−/−Alpha9−/− mice (compare curves above the red dotted line in the sensitivity plots; linear mixed effect model comparison p < 0.001, F = 9.11). However, their phase responses (third column) were similar (p = 0.064, F = 2.42). There were no differences in vibratory response magnitudes between the other three genotypes (p = 0.372, F = 1.08).
To aid the comparison of the four genotypes, we quantified several key characteristics. The gain was calculated for frequencies near the CF by subtracting the sensitivity at 80 dB from that at 20 dB SPL. This demonstrated that VGLUT3−/− mice had higher gain at 9.0 and 9.5 kHz (Fig. 4A). There were no differences in the frequency of maximal vibration (best frequency, or BF) or the sharpness of tuning (Q10dB) between the four genotypes (Fig. 4B,C).
Quantification of the increased cochlear amplification in VGLUT3−/− mice. A, Comparison between WT (n = 9), Alpha9−/− (n = 9), VGLUT3−/− (n = 8), and Alpha9−/−VGLUT3−/− (n = 9) mice. Gain between 20–80 dB SPL was higher in VGLUT3−/− mice at 9.0 and 9.5 kHz (linear mixed effect model comparisons p = 0.020, F = 2.278; follow-up t tests for 9.0 kHz: p = 0.038, t = 2.283 and 9.5 kHz: p = 0.033, t = 2.418), but similar among the other three genotypes (p > 0.05 for all other comparisons). B, There were no differences in the best frequency (BF) between the genotypes for any intensity level (linear mixed effect model comparisons p = 0.822, F = 0.570). C, There were no differences in the sharpness of frequency tuning (Q10dB) between the genotypes for any intensity level (linear mixed effect model comparisons p = 0.855, F = 0.526). D, The maximum gain was largest in VGLUT3−/− mice (ANOVA p = 0.011, F = 4.363; follow-up t test vs Alpha9−/−VGLUT3−/− mice p = 0.026, t = 2.476). There were no differences in the maximum gain between the other three genotypes (ANOVA p = 0.497, F = 0.721). E, The sensitivity at the characteristic frequency (CF) was largest in in VGLUT3−/− mice (ANOVA p = 0.001, F = 6.668; follow-up t test vs Alpha9−/−VGLUT3−/− mice p = 0.004, t = 3.423). There were no differences in the sensitivity at CF between the other three genotypes (ANOVA p = 0.736, F = 0.31). F, The sensitivity at 5 kHz, which was roughly half the CF, was similar among the genotypes (ANOVA p = 0.605, F = 0.623). For D&E, ANOVA was performed first, followed by post hoc t tests with Bonferroni’s correction for each pair combination. For F, only ANOVA was performed because it demonstrated no significance.
We then calculated the largest gain at each frequency by subtracting the sensitivity at 90 dB SPL from the lowest stimulus level that produced a measurable vibratory response at least 3 SD above the noise floor for each mouse. The highest was termed the maximal gain. Both the maximal gain and the sensitivity at the characteristic frequency (CF, the BF for the lowest stimulus intensity) were higher in VGLUT3−/− mice than in Alpha9−/−VGLUT3−/− mice (Fig. 4D,E). There were no differences in the maximum gain or sensitivity at the CF between wild-type, Alpha9−/−, and Alpha9−/−VGLUT3−/− mice. There was no difference in the sensitivity at 5 kHz, which is about half the CF, between any of the genotypes (Fig. 4F). Together, these data demonstrate that cochlear amplification is larger in VGLUT3−/− mice and that this effect is not found when the Alpha9 receptor is knocked out. Thus, the increase in cochlear amplification in VGLUT3−/− mice requires the MOC efferent pathway.
MOC efferents mediate increased OHC and supporting cell activity in VGLUT3−/− mice
Prestin levels have been demonstrated to increase in some states of hearing loss (Yu et al., 2008; Xia et al., 2013; Song et al., 2015), but the mechanism for this effect is unknown. We sought to determine if MOC efferent pathways are involved. Therefore, we performed immunofluorescence labeling and quantified prestin levels in the four homozygous littermate cohorts that resulted from the VGLUT3/Alpha9 double-crossing. Across all genotypes, there were no differences in the amount of prestin per OHC (Fig. 5). This argues that the effect of the MOC efferents on cochlear amplification is not mediated by changes in prestin levels within OHCs.
Increased cochlear amplification in VGLUT3−/− mice are not due to increased OHC prestin levels. Top, Representative immunofluorescence images from the mid-portion of cochlea from the four genotypes. We labeled nerve fibers (α-NF, gradient glow), actin (phalloidin, cyan), hair cells (α-Myo7a, green), and prestin (α-prestin, red). Images of prestin labeling alone are shown to the right. Scale bars, 10 µm. Bottom, We quantified whole cell prestin immunofluorescence from each individual OHC and then averaged these data for each mouse. We studied WT (n = 6), Alpha9−/− (n = 7), VGLUT3−/− (n = 7), and Alpha9−/−VGLUT3−/− (n = 6) mice. There were no significant differences in OHC prestin levels between the four genotypes (ANOVA, p = 0.404, F = 1.018).
MOC efferents may alter cellular physiology in many other ways that could modulate the production of cochlear amplification. Within the OHC, some examples of potential mechanisms include cell stiffness, turgor pressure, membrane fluidity, chloride levels, calcium levels, and resting potential (Santos-Sacchi, 1991; Chertoff and Brownell, 1994; Kakehata and Santos-Sacchi, 1995; Sziklai and Dallos, 1997; Santos-Sacchi et al., 1998, 2001, 2006; He and Dallos, 1999; Oghalai et al., 2000; Sugawara et al., 2000; Lue et al., 2001; Szonyi et al., 2001; He et al., 2003; Oghalai, 2004; Santos-Sacchi and Wu, 2004; Rajagopalan et al., 2007; Zheng et al., 2007). Outside of the OHC, there are MOC efferent terminals that innervate supporting cells and alter gap junction physiology with long-term effects on cochlear amplification (Zhao et al., 2022). Furthermore, MOC effects on OHCs may impact nearby supporting cells, because the two types of cells are both embedded within the organ of Corti and have structural connections. To assess for chronic changes in OHC and supporting cell physiology and/or biomechanical properties related to MOC efferents, we measured changes in reflectivity with subcellular resolution using dynamic OCM (Münter et al., 2020). OCM is essentially OCT but with a high-power objective that permits subcellular resolution.
First, we imaged the organ of Corti in live mice through the RWM using our custom-built OCM system. We then killed the animal and waited 10 min. This was done to remove any dynamic effects by the MOC efferent fibers and only assess for chronic changes. We rapidly collected cross-sectional images and then analyzed the variance in reflectivity at each voxel by bandpass filtering them into low-, middle-, and high-frequency bands (Fig. 6). Simply speaking, this provides an assessment of how fast particles are moving within the cells, not a vibrational measurement. VGLUT3−/− mice demonstrated similar low-frequency activity to control mice, but more mid- and high-frequency activity. While individual cells could not be clearly distinguished, the extent of the spread was large enough to tell that the differences were found both in the OHCs and in the nearby supporting cells, including Deiters’ cells and Hensen cells. To quantify this, the activity within the OHC region and within the Deiters’ cell region was normalized to the activity within the RWM for the frequency band for five mice in each cohort. This demonstrated that the OHC activity was higher within the mid- and high-frequency bands, but similar in the low-frequency band. In the Deiters’ cell region, only the mid-frequency band demonstrated more activity in VGLUT3−/− mice. These data argue that the reason the VGLUT3−/− mice have more cochlear amplification is because the OHCs and supporting cells around the OHCs have long-lasting alterations in physiological and/or biophysical properties.
Dynamic OCM reveals increased middle- and high-frequency activity within the OHCs and nearby supporting cells of VGLUT3−/− mice. Cross-sectional (x–z) images through the round window membrane (RWM) were taken 10 min after killing. The variation in pixel intensity was analyzed to assess the movements of subcellular particles in the tissues. Pixel intensity over time was bandpass filtered into three bins: low (0–5 Hz), mid (5–15 Hz), and high (15–64 Hz). Data from representative WT and VGLUT3−/− mice are shown. The outer hair cell (OHC) and inner hair cell (IHC) regions are noted. VGLUT3−/− mice (n = 5) had relatively more activity in the OHC region and the Deiters’ cell (DC) region in the middle- and high-frequency bands compared with WT mice (n = 5; cyan arrow and orange arrow, respectively). Nonpaired t tests OHClow: p = 0.266, t = 1.19; DClow: p = 0.543, t = 0.635; OHCmid: p = 0.006, t = 3.67; DCmid: p = 0.041, t = 2.43; OHChigh: p = 0.015, t = 3.07; DChigh p = 0.091, t = 1.92.
MOC efferents do not modulate cochlear physiology with changes in brain state
It is well recognized that MOC efferents are activated by sound stimuli via a brainstem reflex arc that feeds back onto OHCs and decreases cochlear gain (Kujawa et al., 1992; Cooper and Guinan, 2003, 2006; Guinan and Cooper, 2008). This is a rapid response and primarily due to ACh-mediated activation of K+ channels that transiently lowers the OHC resting potential (Bobbin and Konishi, 1971; Kakehata et al., 1993; Sziklai and Dallos, 1993; Sziklai et al., 1996; Dallos et al., 1997; Kalinec et al., 2000). The magnitude of the drop in cochlear amplification due to this reflex arc is relatively small, possibly because the anesthesia used during the animal experiments inhibits a potentiating effect of descending inputs to the MOC efferents (Puria et al., 1996; Liberman and Guinan, 1998; Boyev et al., 2002).
To test whether brain state modulates cochlear amplification, we recorded sound-evoked vibrations from the organ of Corti within the apical turn of the cochlea in awake wild-type (WT) and Alpha9−/− mice through the ear canal (Fig. 7). Pupillometry, a proxy for brain state, was tracked simultaneously. During the course of these 2–3 h experiments, we correlated spontaneous changes in pupil diameter with cochlear vibratory responses. By using a session duration of ∼2–3 h, we ensured sufficient time to capture the full range of pupil-indexed brain states in mice. This is supported by the fact that the pupil was sometimes very small and also became very large when the animal walked (largest when running), thus spanning a wide range of states, similar to prior studies (McGinley et al., 2015a, b). The average pupil size was measured in a 500 ms window preceding the sound onset, corresponding to the spontaneous period. This approach was intentional to avoid contamination by sound-evoked pupillary transients (Beatty, 1982; de Gee et al., 2020), allowing us to accurately capture the ongoing brain state at each moment. In separate experiments, we recorded ABRs or local field potentials (LFPs) in the IC to relate possible state-dependent changes in the cochlea to changes in the brainstem at progressive processing stages encompassing the location of efferent neurons.
Experimental setup for recording organ of Corti (OoC) vibrometry and pupillometry in awake mice. A, Head posting and resection of the left pinna was performed. After 1 week, the mouse was habituated to being comfortable on the free-spin wheel and to hearing the sound stimuli. This took ∼3 d, and then we began performing basilar membrane vibrometry through the ear canal. B, The infrared light-emitting diode (IR-LED) and IR camera were used to image the pupil; the ultraviolet LED (UV-LED) was titrated at the beginning of the experiment so that the pupil diameter was in the middle of its range. The visible light (RGB) camera was used to monitor for movements of the mouse that might produce artifacts. C, Image down the ear canal of the mouse. The tympanic membrane and ossicular mass are visible. D, OCT cross-sectional image through the tympanic membrane (Tym) and otic capsule bone reveals the organ of Corti (OoC) within the apical turn of the mouse cochlea. E, Diagram of the cross section of the cochlea. SV, scala vestibuli; SM, scala media; ST, scala tympani; TM, the tectorial membrane; RM, Reissner's membrane. F, Representative data from one mouse. Pupil diameter is shown versus time (red tracing). OoC vibrometry response curves were collected at each black dot. Three sets of vibrometry response curves are shown, each with a different size pupil (small, medium, and large).
Pupil-indexed brain state does not affect organ of Corti vibration
To assess for effects of brain state on cochlear amplification, we measured pupil diameter continuously throughout the experiments in six WT mice (Fig. 8A). Pupil diameter varied spontaneously from small to large during these experiments, indicating spontaneous changes in brain state. We also measured vibratory magnitudes to 50 dB SPL pure tones ranging from 5 to 13 kHz (Fig. 8B). For these initial experiments, phase was not analyzed. We plotted the peak vibratory magnitude versus pupil diameter in a representative example from one mouse (Fig. 8C). There was no correlation. We split these data up into three different pupil sizes (small, medium, and large) for all mice and then plotted the averaged magnitude response (Fig. 8D). There was no significant difference between these three curves.
Brain state, as measured by pupil diameter, does not affect organ of Corti vibration. A, Spontaneous variations in pupil diameter measured in one representative mouse. Over this 100 s cropped portion of the recording, the pupil dilated and then constricted. B, The peak magnitude of the vibratory response measured in the same mouse during the same time calculated from repeated measurements to 50 dB SPL stimuli ranging in frequency from 5 to 13 kHz. There were no obvious changes in vibratory magnitude that correlated with the pupil diameter. C, Scatterplot of data recorded from a full experiment from the same mouse demonstrates no obvious correlation between pupil diameter and the peak magnitude of the vibratory response (linear fit R2 = 0.0045; p = 0.496). D, E, The data for each mouse were binned into small, medium, and large pupil sizes and the vibratory responses averaged. Then, mice within each cohort were averaged (WT: n = 6, Alpha9−/−: n = 8). There were no correlations between vibratory responses and pupil diameter in either genotype (linear mixed effect model comparisons WT: p = 1.00, F = 0.022; Alpha9−/− p = 0.976, F = 0.203).
We performed the same experiment in eight Alpha9−/− mice, in which the OHCs do not receive synaptic input from the MOC efferent fibers (Vetter et al., 1999, 2007; Elgoyhen, 2023). These functioned as controls for these experiments, as their MOC efferent pathway is nonfunctional. Like WT mice, there was no obvious relationship between pupil diameter and vibratory magnitudes in Alpha9−/− mice (Fig. 8E).
Follow-up experiments confirmed that pupil-indexed state does not affect organ of Corti vibration
These initial results surprised us because we considered it likely that since brain state modulates the peripheral visual system by changing pupil diameter (McGinley, 2020), and the efferent neurons receive extensive innervation by neuromodulatory systems (Romero and Trussell, 2022), a similar effect would be found within the peripheral auditory system. Thus, a different experimenter performed a second, more comprehensive series of experiments. This time, we varied the stimulus intensity from 10 to 80 dB SPL. We also increased the frequency resolution of the sound stimuli. To maintain the ability to measure each set of vibratory tuning curves quickly and reduce the impact of animal movements, we used a multitone simultaneous stimulus approach. One downside of this approach is that the presentation of multiple-frequency stimuli mildly decreases cochlear amplification compared with the sequential presentation of pure-tone stimuli (Versteegh and Van Der Heijden, 2012; Fallah et al., 2019).
Each mouse had repeated vibratory measurements performed during the experiment. Pupil diameter was measured simultaneously, and we divided the vibratory data into three categories as described previously (small, medium, and large pupil diameters). We then averaged these three categories of vibratory responses for each mouse. A representative set of vibratory magnitude data measured in one WT mouse is shown (Fig. 9A, left panel). There were no obvious differences between the magnitude responses with different pupil diameters in this example.
Vibratory responses did not vary with brain state. While measuring each set of vibratory responses, the pupil diameter was also measured and categorized as being small, medium, or large. All vibratory responses within each category were then averaged together to create three sets of responses for each mouse. A, B, Representative data from one WT and one Alpha9−/− mouse (left). The three curves from each mouse in each cohort (WT: n = 6, Alpha9−/−: n = 6) were then averaged to get magnitude (center) and phase (right) responses. There were no differences in vibratory magnitude with pupil size (linear mixed effect model comparisons WT: p = 0.505, F = 0.885; Alpha9−/− p = 0.842, F = 0.455). Similarly, phase did not correlate with pupil size (linear mixed effect model comparisons WT: p = 0.107, F = 1.743; Alpha9−/− p = 0.649, F = 0.549). C, D, The gain (left), best frequency (BF, center), and sharpness of frequency tuning (Q10dB, right) were analyzed. There were no correlations between these measures of cochlear amplification and pupil diameter. Linear mixed effect model comparisons were performed for gain (WT: p = 0.882, F = 0.394; Alpha9−/− p = 0.928, F = 0.316), BF (WT: p = 0.918, F = 0.333; Alpha9−/− p = 0.992, F = 0.135), and Q10dB (WT: p = 0.819, F = 0.484; Alpha9−/− p = 0.979, F = 0.189).
The vibratory magnitude and phase data for the 6 WT mice we studied were then averaged (Fig. 9A, center and right panels). Regression analyses revealed no differences associated with pupil diameter. We performed the same experiments with 6 Alpha9−/− mice and found similar results (Fig. 9B). No differences were associated with pupil diameter.
To assess for changes in cochlear amplification more specifically, we calculated the gain between 20 and 80 dB SPL stimuli at frequencies where the vibratory response was largest (Fig. 9C, left panel). There were no differences in gain linked to pupil diameter. We also looked for shifts in the tonotopic map by determining the BF for each stimulus level (Fig. 9C, center panel). However, statistical analyses revealed no differences linked to pupil diameter. Finally, we assessed for changes in the sharpness of tuning by calculating the Q10dB for each stimulus intensity (Fig. 9C, right panel). Again, there were no differences associated with pupil size. In Alpha9−/− mice, these analyses revealed similar results (Fig. 9D).
We then directly compared WT and Alpha9−/− mice by binning pupil diameter from 0 to 100% (minimum to maximum diameter for each mouse) and comparing several parameters. We assessed only for differences linked to genotype or the combination of genotype and pupil diameter, with the assumption that correlations related to pupil diameter, but not genotype, do not involve the MOC efferent pathway. The gain from 20 to 80 dB SPL at the characteristic frequency (CF, the BF at the lowest stimulus intensity) was not statistically higher in Alpha9−/− mice than that in wild-type mice, and there was no correlation of pupil diameter with gain (Fig. 10A). Interestingly, even though there seemed to be a slight difference in tuning between WT and Alpha9−/− mice, the characteristic frequency (CF, or the BF to 20 dB SPL stimuli) was not statistically different between these cohorts and, more importantly, there were no differences linked to pupil diameter (Fig. 10B). Tuning curve sharpness at 20 dB SPL (Q10dB) was not different between the genotypes nor did it vary with pupil diameter (Fig. 10C).
There were no correlations between cochlear amplification and pupil diameter in either WT (n = 6) or Alpha9−/− (n = 6) mice. Data were binned by pupil diameter into 10 bins (i.e., by decile), and linear mixed effect model comparisons were performed to assess for effects of genotype and pupil diameter on each measurement. A, Gain (genotype: p = 0.420, F = 0.705; pupil: p = 0.570, F = 0.684), (B) CF (genotype: p = 0.101, F = 3.28; pupil: p = 0.595, F = 0.642), (C) Q10dB (genotype: p = 0.252, F = 0.1.49; pupil: p = 0.353, F = 1.32), (D) the vibratory magnitude at the BF (genotype: p = 0.856, F = 0.035; pupil: p = 0.124, F = 2.10), (E) the vibratory magnitude at half the BF (genotype: p = 0.556, F = 0.369; pupil: p = 0.762, F = 0.389), and (F) the phase at the CF (genotype: p = 0.592, F = 4.51; pupil: p = 0.172, F = 1.79) all demonstrated no correlations.
Next, we analyzed several raw values from the vibratory data curves that we thought might be even more sensitive to genotype and/or pupil diameter differences. The vibratory magnitude at the peak of the tuning curve with low-intensity stimuli would be most likely to change with the modulation of cochlear amplification. However, the vibratory magnitude at the BF to 20 dB SPL stimuli did not vary with genotype or pupil diameter (Fig. 10D). Another possibility was that cochlear amplification below the BF might be modulated by the MOC efferents. However, the vibratory magnitude at half the BF in response to 50 dB SPL stimuli also did not vary with genotype or pupil diameter (Fig. 10E). Finally, if the speed of the traveling wave was being modulated, the phase at the BF to low-intensity stimuli would change. However, it did not vary significantly with genotype or pupil diameter (Fig. 10F). Altogether, these data from two sequential series of experiments in awake mice demonstrate that dynamic changes in brain state, as measured by pupillometry, do not modulate cochlear amplification.
General anesthesia does not affect cochlear amplification
In order to test whether larger variations in brain state affect cochlear amplification, we next determined the effect of general anesthesia, where cortical input to the MOC efferent pathway is removed, on cochlear amplification. We first measured organ of Corti vibratory responses in awake mice and then remeasured them after the induction of general anesthesia. We hypothesized that the activity of the MOC efferents reflecting brain state would be blocked by anesthesia and thus detected with our vibratory measurements. However, we found that WT mice demonstrated no change in vibratory magnitude or phase between the awake and anesthetized state (Fig. 11A). Similarly, Alpha9−/− mice also had no changes in these measurements between the awake and anesthetized state (Fig. 11B). Further analysis to quantify the gain, BF, and Q10dB also revealed no differences between awake and anesthetized wild-type mice (Fig. 11C). In Alpha9−/− mice, there were no differences in the BF between the awake and anesthetized states (Fig. 11C, center panel). However, there were statistically significant differences in the gain and Q10dB between these states in Alpha9−/− mice (Fig. 11C, left and right panels). To better understand these differences, we then performed paired t tests with Bonferroni’s correction at each frequency. Significant differences were found for three frequencies in the gain curves (8.3, 9.9, and 12.7 kHz) and for two levels in the Q10dB curve (20 and 30 dB SPL). Our interpretation of these areas of significance is that, even with the Bonferroni’s correction, they most likely represent random variation due to the large number of statistical comparisons we have done to analyze these data. Furthermore, the fact that they were only found in Alpha9−/− mice indicates that, even if these findings have a true physiologic basis, the underlying cause is not related to the MOC efferent pathway. Thus, none of our experiments supported the hypothesis that brain state dynamically modulates cochlear function.
Anesthetized brain state does not alter BM vibration. A, B, Vibratory responses from awake (solid lines) and anesthetized (dotted lines) WT (n = 6) and Alpha9−/− (n = 6) mice were similar for one representative mouse (left), averaged magnitude responses (center), and averaged phase responses (right). There were no differences in vibratory magnitude between the awake and anesthetized conditions in WT or Alpha9−/− mice (linear mixed effect model comparisons WT: p = 0.363, F = 1.06; Alpha9−/− p = 0.207, F = 1.52). Similarly, the phase did not change (linear mixed effect model comparisons WT: p = 0.649, F = 0.549; Alpha9−/− p = 0.615, F = 0.600). C, D, The gain (left), BF (center), and Q10dB (right) were calculated. Linear mixed effect model comparisons were performed to assess for effects of anesthesia on each measurement. Gain (WT: p = 0.669, F = 0.521; Alpha9−/−: p < 0.001, F = 7.85), BF (WT: p = 0.469, F = 0.855; Alpha9−/−: p = 0.102, F = 2.15), Q10dB (WT: p = 0.485, F = 0.825; Alpha9−/−: p = 0.039, F = 2.93). There were no significant differences found in WT mice. However, there were a few occasional points of statistical significance in Alpha9−/− mice that we followed up with t test analyses (e.g., Gain 8.3 kHz p = 0.036; Gain 9.9 kHz p = 0.022; Gain 12.7 kHz p = 0.014; Q10dB 20 dB SPL p = 0.015; Q10dB 30 dB SPL p = 0.035). However, these appear to have little physiological relevance, and our interpretation is that they represent statistical outliers.
State dependence of overall sound response strength emerges in the inferior colliculus
Brain states dynamically modulate auditory responses in the cortex and thalamus (McCormick et al., 2015; McGinley et al., 2015a; Reimer et al., 2016). However, since we did not find any state-dependent changes within the cochlea, we wanted to determine whether these brain state effects are confined to the cortical and thalamic structures, or if subthalamic structures, including where the MOC neurons are located, are also influenced by brain state. Using head-posted awake mice on a cylindrical treadmill, we first measured pure tone-evoked LFPs from the central nucleus of the IC in the midbrain using Neuropixels probe 1.0 (Fig. 12A, left). In addition to LFPs, we also tracked pupil diameter and walking velocity (Fig. 12B). The LFPs exhibited two distinct peaks: an early positive peak (P1) followed by a large negative trough (N1; Fig. 12C). We found that the mean amplitude of N1 was significantly modulated by pupil-indexed brain state (Fig. 12E,F), showing higher responses at mid-arousal levels, whereas P1 was only weakly modulated (Fig. 12E,F; note the N1 amplitude sign is reversed). The error bars in Figure 12E represent the 68% BCA bootstrap confidence interval of the mean (N = 2,687 electrode sites from 22 sessions in 13 animals).
Brain state correlates with responses in the inferior colliculus (IC), but not auditory brainstem responses (ABR). A, Left, Schematic of Neuropixels probe and image of the probe tip used for LFP data acquisition. Right, Sagittal view of the brain section showing probe track in IC. B, Snippet from an example session showing LFP aligned to pupil size. C, Population average tone-evoked LFP in the IC grouped by pre-stim pupil size showing state-dependent modulations in the negative peak of the LFP (N1). The error bars represent 68% BCA bootstrap confidence interval of the mean (N = 2,687 electrode sites from 22 sessions in 13 animals). D, Population average ABR grouped by pre-stim pupil size show lack of state dependence (n = 32 sessions in 6 animals). The error bars represent 68% BCA bootstrap confidence interval of the mean. E, Peak amplitude responses for LFP and ABR waves as a function of pupil size. N1 in the LFP responses from the IC shows strong state-dependent modulations. The ABR waves show no obvious state dependence. The error bars represent 68% BCA bootstrap confidence interval of the mean. The horizontal error bars (gray) for pupil size are across sessions whereas vertical error bars are across electrode sites. F, Using a quadratic model, we show that pupil-indexed brain state better predicts tone-evoked IC response but not for ABR waves. Explained variance (R2) is much larger for N1 than those observed for ABR waves. Data is bootstrapped 1,000 times with replacement to get R2 replicates. R2 is cross-validated using leave-one-out method. The error bars represent 68% BCA bootstrap confidence interval of the mean. ABR-I: p = 0.61, ABR-II: p = 0.53, ABR-III: p = 0.55, ABR-IV: p = 0.86, ABR-V: p = 0.87, log(V–I ratio): p = 0.99, LFP-P1: p = 0.04, LFP-N1: p < 0.0005.
Having demonstrated that state dependence is apparent at the level of IC responses, we wondered if it was also apparent in earlier brainstem structures. Thus, we measured ABRs to pure tones using scalp electrodes while tracking pupil diameter in awake head-fixed mice (Fig. 12D; N = 32 sessions in 6 animals). We analyzed the amplitudes of the first five ABR waves as a function of pupil size and found no state dependence in any of the peaks (Fig. 12E). The ABR wave V–I ratio, commonly used to quantify collicular central gain, also showed no state dependence (Fig. 13).
Local averaging of adjacent trials improves R2 for IC responses but not for ABR waves. A, The R2 for IC responses increased as the number of trials averaged increased. No such increase was observed in ABR waves. The error bars represent 68% BCA bootstrap confidence interval of the mean. B, Ratio of ABR wave V–I does not show state dependence.
To quantify state dependence in both LFPs and ABRs, we fit a quadratic model to the peak responses as a function of pupil-indexed brain state and computed the explained variance (R2). This analysis showed that incorporating brain state improved prediction of the tone-evoked N1 response in the IC, but not the ABR waves or IC P1 (Fig. 12F). The small effect size of state dependence, as reflected in low R2 values, could be due to high single-trial variability resulting from the low SNR of ABR and LFP measurements. To address this possibility, we averaged trials that were adjacent in pupil size, before computing R2. The R2 for IC N1 responses increased with the number of trials averaged and a minimal increase was seen in P1, whereas no such increase was observed in ABR waves. The error bars represent the 68% BCA bootstrap confidence interval of the mean (Fig. 13). These results demonstrate that, even when locally denoising ABR and LFP estimates, pupil-indexed brain state does not impact ABR response magnitudes. Consistent with our vibrometry recordings, which did not show a change in brain state, the most reliable peak in the ABR, wave 1, reflecting the early auditory nerve response, did not show changes. Furthermore, waves 4–5, which encompass the superior olive (Melcher et al., 1996), where MOC cell bodies are located, did not show state dependence. Thus, these data demonstrate that pupil-indexed brain state modulate the overall response gain in the auditory pathway at higher levels than the cochlea and its MOC inputs and validates our experimental protocol for monitoring brain state.
Discussion
Our studies reveal that MOC efferent feedback causes cochlear gain to increase in chronic states of hearing loss. This occurs through chronic changes in biophysiological properties of the OHCs and supporting cells within the organ of Corti. Furthermore, we found that, unlike the peripheral visual system, which is modulated dynamically by the effects of spontaneous changes in brain state on pupil diameter, the cochlear amplifier is not. Together, these data suggest that the effect of brain state on processing within the mammalian cochlea is likely to be more chronic in nature and related to long-standing changes in acoustic exposure. Age, noise, and ototoxic exposure cause sensorineural hearing loss through the loss of cochlear hair cells and auditory neurons. Significant hearing loss affects 15% of American adults, and almost everyone becomes affected with age (NIH-NIDCD, 2023). Given that LOC efferents are thought to balance the afferent input between the two ears (Darrow et al., 2006), it is intriguing to consider that MOC efferents may provide a way for the brain to increase cochlear amplification and partially compensate for progressive sensorineural hearing loss.
The source of increased cochlear amplification in VGLUT3−/− mice might be due to increased cortical stimulation of the MOC efferents in response to the state of hearing loss. This might occur through the loss of afferent auditory input directly on the MOC brainstem neurons at the level of the reflex arc. Alternatively, it is important to consider that VGLUT3 is present in neuromodulatory brain structures (Schäfer et al., 2002; Amilhon et al., 2010; Balázsfi et al., 2018). VGLUT3−/− mice have been reported to have increased stress and anxiety (Balázsfi et al., 2018). Thus, it is possible that these effects also modulate MOC efferent activity. Given the association between stress, anxiety, and hearing loss found in large public health studies (Chung et al., 2015; Grewal and Golub, 2023; Zhang et al., 2024), this potential mechanism is also intriguing.
Using OCT, we were able to record basilar membrane vibrations within the mammalian cochlea of awake animals for the first time. OCT provides a direct measure of cochlear amplification, whereas most of the previous work on MOC efferent physiology uses indirect measures of cochlear function such as ABRs, otoacoustic emissions, and threshold shifts after noise exposure. Most importantly, this technique permitted us to study awake mice, a necessity in understanding the connection between the brain and the cochlea (Chambers et al., 2012).
It could be argued that there are prominent effects of brain state on the MOC reflex that occurs in response to contralateral or ipsilateral noise, which our approach neglected. We considered this when designing the study and decided against it. We reasoned that while there may be effects of brain state on this reflex arc, the strength of the MOC reflex is relatively small (<10 dB) in mice and has high variability (Chambers et al., 2012; Froud et al., 2015), and so the impact of neuromodulatory brain state would be expected to be even smaller and difficult to detect. It should be noted that some of the variability between studies may reflect the species too, with humans and nonhuman primates perhaps having larger responses (Tavartkiladze et al., 1996; Mehta et al., 2021). However, it is difficult to separate out MOC effects from other feedback mechanisms (e.g., LOC effects and brainstem, subcortical, and cortical circuitry) in less-controlled species. At any rate, we concluded that the significance of such an effect on the behavior of mice would be challenging to unravel. Nevertheless, we are excited about the potential for brain state modulation of the MOC reflex to have relevance to human communication. Intriguingly, recent data suggests that there may also be auditory processing abnormalities related to loss of MOC efferent activity (Mondul et al., 2024). Additional data indicates auditory contrast enhancement mostly occurs at the cortical and not subcortical level (Mehta et al., 2021). Finally, the experimental approach we developed to measure organ of Corti vibrations in awake mice using OCT is the first step to advancing the technology to measure the cochlear traveling wave in humans. This will be a critical step to understanding how efferent nerves modulate cochlear physiology when attending to speech in noise.
A weakness of this study is that we only assessed for effects of spontaneous changes in brain state. While this was adequate to detect changes within the IC, it may not have been strong enough to evoke changes within the cochlea. Future work is needed to understand if larger, evoked changes in brain state, such as hearing, seeing, or feeling something that piques the interest of the animal, create MOC efferent effects that are detectable within the cochlea. Our robust experimental approach will permit these more advanced studies of dynamic and chronic changes in cochlear physiology in future experiments; sequential measurements over long periods of time can even be performed in the same animal. Another weakness is that although counts of MOC efferent to OHC synapses are roughly similar in the apex, middle, and base of the mice at the ages we studied, there is, on average, 0.5 more synapses in the middle than in the apex or base (Park et al., 2017; Dörje et al., 2024). Thus, it is possible that the dynamic OCM data we measured in the base may not fully reflect what is happening in the mid-apical turn where we performed vibrometry.
This study also reports the first use of dynamic OCM to assess cellular physiology within the organ of Corti. Together with the analysis of changes in reflectivity, an assessment of dynamic activity within cells can be obtained. The changes are due to the movement of intracellular molecules and are reflective of intracellular transport, synthesis, and metabolism (Münter et al., 2020). The changes we found related to increased cochlear amplification were not only within OHCs, but also in supporting cells, suggesting that either MOC efferents directly alter all these cells or that increased activity of OHCs led to downstream effects on the cells around them. This is certainly feasible since K+ recycling occurs through these cells, and they are all connected via gap junctions. The fact that we saw these changes after MOC efferent activity had been removed (by killing) but during the time window when these cells have normal physiology as assessed by patch-clamp studies (Sato and Santos-Sacchi, 1994; Oghalai et al., 1998; Zhao and Santos-Sacchi, 1998, 1999, 2000; Rajagopalan et al., 2007) indicates that MOC efferents spur long-term changes in cell activity. This result is consistent with data on sound conditioning, whereby chronic exposure to low levels of noise reduces susceptibility to loud noise exposure and is linked to changes in OHC function (Kujawa and Liberman, 1997; Yoshida and Liberman, 2000).
Finally, it is quite common for patients to have normal auditory thresholds yet have symptoms of tinnitus and hyperacusis. While “hidden hearing loss” due to cochlear synaptopathy is a well-recognized cause of this (Sergeyenko et al., 2013; Fernandez et al., 2015; Liberman et al., 2015; Liberman and Kujawa, 2017), it is possible that increased cochlear gain is also a factor. Although we did not detect dynamic effects of brain state on cochlear amplification within the time frame of our measurements (seconds to minutes), long-lasting changes in brain state may modulate cochlear amplification. Stress, anxiety, and depression may thus all play a role in modulating how the cochlea processes sound, such as has been shown in mice with restraint stress (Kujawa and Liberman, 1997; Wang and Liberman, 2002). Thus, we consider it possible that MOC efferent-mediated control of cochlear gain may also underlie some of the associated symptoms of hearing loss that are well established clinically to be modulated by brain state, such as tinnitus and hyperacusis (Langguth et al., 2013, 2024).
Data Availability
Data supporting the findings of this study will be shared upon request. We also have uploaded key datasets to our GitHub site.
Footnotes
This work was supported by the National Institutes of Health/National Institute on Deafness and Other Communication Disorders (NIH/NIDCD) R01 DC014450 (J.S.O and M.J.M), R01 DC013774, R01 DC017741, R25 DC019700 (J.S.O), NIH/National Institute of Biomedical Imaging and Bioengineering R01 EB027113 (B.E.A), NIH/NIDCD R21 DC019209 (J.B.D), R01 DC017797 (M.J.M), and the Keck School of Medicine Dean's Research Scholar Program. We thank the Optical Imaging Facility at the Broad CIRM Center of USC and the Translational Imaging Center at the USC Viterbi School of Engineering. We also thank Naomi Quiñones for illustrating the Figure 1 schematic.
J.S.O. and B.E.A. are founders of AO technologies, with the goal of translating inner ear imaging technologies for clinical purposes. The other authors declare no competing financial interests.
- Correspondence should be addressed to John S. Oghalai at oghalai{at}usc.edu.
