Organ of Corti vibration within the intact gerbil cochlea measured by volumetric optical coherence tomography and vibrometry

Abstract

There is indirect evidence that the mammalian cochlea in the low-frequency apical and the more commonly studied high-frequency basal regions function in fundamentally different ways. Here, we directly tested this hypothesis by measuring sound-induced vibrations of the organ of Corti (OoC) at three turns of the gerbil cochlea using volumetric optical coherence tomography vibrometry (VOCTV), an approach that permits noninvasive imaging through the bone. In the apical turn, there was little frequency selectivity, and the displacement-vs.-frequency curves had low-pass filter characteristics with a corner frequency of ~0.5–0.9 kHz. The vibratory magnitudes increased compressively with increasing stimulus intensity at all frequencies. In the middle turn, responses were similar except for a slight peak in the response at ~2.5 kHz. The gain was ~50 dB at the peak and 30–40 dB at lower frequencies. In the basal turn, responses were sharply tuned and compressively nonlinear, consistent with observations in the literature. These data demonstrated that there is a transition of the mechanical response of the OoC along the length of the cochlea such that frequency tuning is sharper in the base than in the apex. Because the responses are fundamentally different, it is not appropriate to simply frequency shift vibratory data measured at one cochlear location to predict the cochlear responses at other locations. Furthermore, this means that the number of hair cells stimulated by sound is larger for low-frequency stimuli and smaller for high-frequency stimuli for the same intensity level. Thus the mechanisms of central processing of sounds must vary with frequency.

NEW & NOTEWORTHY A volumetric optical coherence tomography and vibrometry system was used to probe cochlear mechanics within the intact gerbil cochlea. We found a gradual transition of the mechanical response of the organ of Corti along the length of the cochlea such that tuning at the base is dramatically sharper than that at the apex. These data help to explain discrepancies in the literature regarding how the cochlea processes low-frequency sounds.

INTRODUCTION

The study of cochlear mechanics has been a central focus of auditory research for decades because cochlear properties underlie the neural encoding of speech and music. The sound energy propagates in the cochlea via a traveling wave (von Békésy 1960), which is mostly documented in the vibration of the basilar membrane (BM) at the basal high-frequency region (Robles and Ruggero 2001). The cochlea responds to sound in a tonotopic manner in that there is a frequency-to-place mapping along the length of the cochlea, whereby low-frequency sounds create maximal vibration of the BM in the apex, and high-frequency sounds create maximal vibration of the BM in the base. Auditory nerve fibers (AN) that innervate different cochlear regions carry information from different sound frequencies. The mechanical properties of the organ of Corti (OoC) cells and tissues that lie on top of the BM are important in determining the stiffness and mass to support the traveling waves. In mammals, force generation by outer hair cell (OHC) electromotility and possibly active stereociliary bundle mechanics selectively add energy to the traveling wave in the living cochlea (Brownell et al. 1985; Dong and Olson 2013; Nin et al. 2012; Oghalai 2004). This is called cochlear amplification, evidenced in that the vibratory amplitude of the BM increases in a compressive nonlinear manner that is largest for lower sound pressure levels (Davis 1983; Rhode 1971). This evolutionary adaptation extends the frequency range of hearing beyond the ~5-kHz limit that is typical of most nonmammals by overcoming the problem of viscous damping at high frequencies (Gold and Hearing 1948; Liao et al. 2007; Manley 2017).

Cochlear amplification has been carefully measured in the cochlear base of the gerbil, guinea pig, chinchilla, and in the apex of the mouse, all of which are tuned to frequencies above 9 kHz. Associated with the gain of amplification, these data have found an increase in the sharpness of tuning such that frequency selectivity is improved with maximal cochlear amplification at the lower input levels. Furthermore, the frequency responses of BM vibration and AN firing rate closely approximate each other near the threshold of hearing (Narayan et al. 1998). Thus the concept that AN tuning is defined by BM mechanics has been ingrained as a fundamental property of mammalian hearing (Davis 1958). However, this notion has been challenged for AN suprathreshold stimulations that are efferent inhibited (Stankovic and Guinan 1999, 2000) and for lower-frequency cochlear regions (Guinan et al. 2012).

There is significant energy content in human speech and music for frequencies below 5 kHz. However, measurements from regions of the mammalian cochlea that are tuned to these lower frequencies have been sparse. The studies indicate that cochlear mechanics in the low-frequency apical region is different than in the high-frequency basal region. Responses in the low-frequency region exhibit a smaller degree of nonlinearity and broader tuning (Cooper and Dong 2003; Cooper and Rhode 1997, 1995; Dong and Cooper 2006; Zinn et al. 2000). Because of technological limitations at the time these experiments were performed, the otic capsule bone had to be opened, and Reissner’s membrane needed to be incised to drop reflective beads down onto the cochlear partition to measure sound-induced vibratory responses. This disruption of the fluid chambers made these experiments challenging and prone to artifact. Recently, optical coherence tomography (OCT) was developed to measure vibration in the mouse cochlear apex noninvasively by imaging through its relatively thinner bone (Gao et al. 2013, 2014; Lee et al. 2016). However, the mouse cochlear apex is not tuned to frequencies below 5 kHz and thus does not adequately represent low-frequency hearing. Recent studies have used OCT to study the extreme low-frequency region (<2 kHz) of the guinea pig cochlear apex noninvasively, and this demonstrated broad tuning with substantial cochlear amplification above and below the peak of the tuning curve (Recio-Spinoso and Oghalai 2017). Here, we studied all three turns of the gerbil cochlea. This is particularly important because the auditory range of the gerbil at low frequencies approximates that of human hearing (Ryan 1976). We found that, although mechanical amplification and compression were strong along the length of the cochlea, the sharpness of frequency tuning varied dramatically.

MATERIALS AND METHODS

Gerbil preparation.

Experiments were performed initially at Stanford University and later at the University of Southern California, when J. Oghalai changed institutions. The experiments were approved by the Institutional Animal Care and Use Committees at both institutions. Adult gerbils (Charles River) of either sex weighing 40–60 g were deeply anesthetized using ketamine (100 mg/kg) and xylazine (10 mg/kg). The head was fixed in a head holder. The conchal bowl of the left pinna was removed to create a clear view of the tympanic membrane. The bulla was then widely opened to expose the cochlea. We removed the posterior two-thirds of the jaw and infratemporal fossa musculature to widely open the bulla and reach the cochlear apex. Nearly half (12/26) of the animals contributed to the presented data, whereas the other animals were used to figure out the surgical approaches and contributed to postmortem responses.

An earbud was then coupled to the ear canal and calibrated to sound stimuli, which were generated by custom software (Lee et al. 2016). In three gerbils, we tested cochlear function before and after surgically opening the bulla by measuring click-evoked auditory brainstem response (ABR) wave I using subdermal electrodes as previously described (Xia et al. 2007, 2013, 2016). ABR measurements were not performed for later experiments because there was little difference (Fig. 1A) caused by the surgical approaches to the cochlea. In addition, the cochlear condition was also evaluated at different time points by repeating measurements from the same location. The cochlear condition was considered to be healthy when the input-output function of responses changed little. This was typical for the first 2–3 h in every experiment.

Fig. 1.

The experimental preparation and representative data from 3 turns of the gerbil cochlea. A: auditory brainstem response (ABR) wave I peak-to-peak magnitudes in response to click stimuli immediately after induction of anesthesia (before surgery) and after surgically opening the middle ear, the bulla, to visualize the cochlea (after surgery). There were no deviations from normal cochlear function postsurgery. Error bars are the SE. B: in vivo volumetric optical coherence tomography vibrometry (VOCTV) image of the 3 turns of the gerbil cochlea. The measurement locations within the organ of Corti (OoC) of turn 3, turn 2, and turn 1 (X) and the optical path of the VOCTV system (dashed lines) are marked. C: plastic-embedded cross section of the gerbil cochlea. D–F: magnitude of vibratory displacement in turn 3, turn 2, and turn 1 of live gerbils. Darker lines indicate less intense stimuli. G–I: magnitude of the vibratory displacement measured from the same spot in each gerbil cochlear turn postmortem. J–L: sensitivity in the live (solid lines) and dead (dashed lines) gerbil calculated by dividing the magnitude of OoC vibration by that of the middle ear (ME) ossicle, the umbo. M–O: phase in the live (solid lines) and dead (dashed lines) gerbil calculated by subtracting the phase of the umbo responses from that of the OoC. Data of the first, second, and third turns were from gerbils G12, 13, and 21, respectively. BM, basilar membrane; RM, Reissner’s membrane; TM, tectorial membrane; SPL, sound pressure level.

For the vibrometry measurements, pure tone sound stimuli of 100 ms were swept in frequency from 0.1 to 25.0 kHz in 0.1-kHz steps, and the intensity ranged from 10 to 80 dB sound pressure level (SPL). However, not every frequency/intensity combination was presented because, attributable to speaker limitations, some had harmonic distortion peaks that were not 40 dB below the stimulus primary frequency (i.e., <1%). These tones were not used in the experiments. After we performed vibrometry in the live condition, the gerbil was then euthanized, and the vibratory measurements were repeated in the postmortem condition. In addition, vibrometry recordings were made from the middle ear ossicles (i.e., the umbo) to quantify the magnitude and phase of the motion entering the cochlea.

Volumetric optical coherence tomography and vibrometry.

Vibrations of the OoC were measured using volumetric optical coherence tomography vibrometry (VOCTV). Our VOCTV system is custom built and has been described previously (Lee et al. 2015). Briefly, it is composed of a broadband swept source with a center wavelength of 1,300 nm and 200-kHz sweep rate (MEMS-VCSEL; Thorlabs, Newton, NJ), a dual-balanced photodetector (WL-BPD600MA; Wieserlabs, Munich, Germany), and a digitizer (NI-5761; National Instruments, Austin, TX). An adaptor, equipped with galvo motors, attached to the bottom of a dissecting microscope (Stemi-2000; Zeiss, Jena, Germany), scanned the beam in both the x and y directions. For all experiments, the power on the sample was 16 mW. Each individual pixel imaged by our system has a lateral resolution of 9.8 μm and axial resolution of 11.4 μm in water (Lee et al. 2015).

A cross-sectional image was first collected (Gao et al. 2011), and then sound-induced vibrations were measured from a selected point. During the sound stimulus, we recorded the vibratory response from a point of interest that had strong signal intensity on the enlarged image because these provided the lowest noise floor for the vibratory measurements and minimized the risk of phase corruption from other nearby vibrating points (Lin et al. 2017). No averaging of adjacent spatial locations was performed. We analyzed the data by performing a Fast Fourier transform (FFT) of the vibratory response and determined the peak vibratory magnitude at the stimulus frequency. We next determined the mean and SD of the noise floor by averaging the vibratory magnitude over the FFT bins contained within a 100-Hz band adjacent to the stimulus frequency. If the peak vibratory magnitude at the stimulus frequency was less than the noise floor threshold defined by the mean + 3 SD of the vibratory noise floor, we discarded that measurement.

Histology.

Gerbil histological sections were made and imaged using previously described techniques (Cho et al. 2013; Xia et al. 2010). After fixation, osmium staining, decalcification, and embedding, 10-μm-thick sections were cut. Imaging was performed using an upright light microscope with a camera attachment.

Data adjustment and statistical analysis.

To average data collected from different gerbils, vibratory magnitudes and phases were averaged separately, and the mean and SE for all data points were calculated. Phase unwrapping was performed so that all phases that were averaged together at each frequency were within π radians of the mean. Data from an animal were not included in statistical analyses if the vibratory response at that frequency/intensity combination was below the noise floor threshold. Similarly, for all figures, we did not plot data at those frequency/intensity combinations in which the vibratory responses from 3 or more of the 12 gerbils studied were below the noise floor threshold. R studio was used to perform all statistical analyses (R Core Team 2017). Nonpaired two-tailed t-tests were used to compare cochlear gain and Q_6dB between different turns. Multiple linear regression was used to compare ABR peak-to-peak values before and after the surgical opening. For all statistics, a P value <0.05 was considered significant. All values given in the text are the means ± SE.

Frequency to spatial mapping.

To estimate the spatial excitation pattern along the cochlea, we used the gerbil cochlear map (Müller 1996) to define the frequency-to-place mapping of the cochlea. We interpolated the data we had from three different tonotopic locations (base, middle, and apex) to estimate the vibratory response over the intervening cochlear regions. We then could simulate the cochlear traveling wave in response to pure tones of other frequencies and intensities, as we have done previously (Xia et al. 2016). This was done using custom software written in Python.

RESULTS

ABR thresholds remain stable in this animal preparation.

After anesthesia, we measured ABR responses to click stimuli in three gerbils. The ABR responses, measured as the peak-to-peak value of wave I, demonstrated growth with stimulus intensity reflective of a normal pattern of auditory sensitivity (Fig. 1A). We then surgically opened the bulla and repeated the measurements. The ABR responses did not change after the surgical opening (P = 0.281, multiple linear regression including intensity and animal as additional independent variables). These data suggest that the experimental exposure did not affect cochlear function.

Anatomic imaging of the gerbil cochlea in vivo.

We used the VOCTV system to scan the beam and obtain a cross-sectional image (Fig. 1B). The most obvious cochlear structures were the three cochlear chambers (scala vestibuli, media, and tympani), which were separated by Reissner’s membrane and the BM. The imaging resolution of our system did not permit differentiation between subcomponents of the cochlear partition, such as the BM, tectorial membrane, hair cells that were composed of inner hair cells (IHCs) and OHCs, or supporting cells as indicated in the histological cross-section image of a gerbil cochlea (Fig. 1C).

Although the optical path into the cochlea was similar for all turns (dashed lines in Fig. 1B), because the angle of the BM was slightly different for each turn, the relative components of radial and transverse motion we measured in each turn were definitely somewhat different. However, these angles were all ± 20° to the radial direction, which means that >94% of the radial motion was detected. For each turn, we then selected a point providing the strongest signal within the OoC, close to the apical end of OHCs as indicated by crosses in the OCT image to measure sound-induced vibrations (white and color X marks, Fig. 1, B and C).

Although it may be reasonable to use these data to interpret gerbil cochlear mechanics according to present theories and models, which are based on the measurements of transverse vibration, one must remember that we measured radial, not transverse, vibratory measurements. At this point, we do not know what the implications of this different measurement angle are in gerbils, but our studies in the mouse cochlea indicate that the transverse and radial vibratory patterns are quite similar (Lee et al. 2016). In addition, these measurements are closer to the reticular lamina than to the BM. Previous studies have demonstrated that these two structures have subtle differences in their vibratory patterns (Cooper et al. 2018; He et al. 2018; Lee et al. 2016; Ren et al. 2016).

Representative vibrometry measurements at different frequency regions along the cochlea.

Sound-induced motions of the cochlear partition in response to pure tone stimuli from the first, second, and third turns (base, middle, and apex) in the live gerbil were measured and repeated after euthanasia in the dead gerbil. To demonstrate characteristics of vibrometry displacements in the living and postmortem conditions, representative data from the same recording spot are shown (Fig. 1, D–O) with recordings from turn 3, turn 2, and turn 1, respectively. In turn 3, the vibratory response exhibited low-pass responses with a corner frequency around 800 Hz (Fig. 1D). There was also strong compressive nonlinear growth across the entire frequency spectrum measured, with the vibratory response to 70-dB SPL stimuli being only ~10 times more than that measured at 20-dB SPL stimuli, despite an increase in the sound pressure level of 50 dB, or ~300 times. Postmortem, the vibratory responses kept the same low-pass filter shape independent of input level, indicating a linear response (Fig. 1, G and J). Thus the nonlinearity was shown to be physiologically vulnerable.

The difference in the vibratory magnitudes between live and dead conditions is best seen in sensitivity plots, whereby OoC displacement was normalized to umbo displacement (Fig. 1J). In the postmortem condition, the sensitivity curves overlap, indicating that the OoC vibrates linearly in proportion to the stimulus intensity. In the live condition, the sensitivity was greater for 20-dB SPL stimuli compared with 70-dB SPL stimuli. This nonlinear cochlear gain, the difference between the sensitivities of the lowest and highest stimulus intensities, was found over the entire frequency spectrum measured. Finally, the phase demonstrated progressive lags with increasing frequency, which is consistent with traveling wave propagation (Fig. 1M).

In turn 2, responses to 10-dB SPL stimulation were only above the noise floor at frequencies between 1.8 to 2.8 kHz. A band-pass filter shape superimposed on the low-pass-like results were found with a peak at ~2.5 kHz (Fig. 1, E, H, K, and N). Furthermore, the corner frequency was higher (~2.8 kHz), consistent with the fact that this location was closer to the basal end of the cochlea. The nonlinear cochlear gain was greatest, i.e., ~50 dB between 10 and 80 dB SPL, at frequencies around the peak or the best frequency (BF) and was smaller at lower frequencies, i.e., ~35 dB. The data suggested that there was frequency selectivity at the BF of ~2.5 kHz. Cochlear gain disappeared postmortem. Up to approximately four cycles of phase accumulation was found, which is consistent with similar phase accumulation in the high-frequency cochlear base (Ren and Nuttall 2001; Robles and Ruggero 2001).

Sound-induced motions of the cochlear partition in response to pure tone stimuli of the high-frequency turn 1 were also measurable (Fig. 1, F, I, L, and O). Because of the thicker cochlear bony wall and smaller magnitude, the clear responses were presented at intensities above 30 or 40 dB SPL. Similar to well-established gerbil BM responses measured transversely using laser interferometer (Ren et al. 2011), the OoC responses peaked at ~12 kHz, the BF, at 40 dB SPL. Vibratory magnitudes then increased compressively at the BF with increasing sound pressure levels. The sensitivity was greater for 40-dB SPL stimuli compared with 80-dB SPL stimuli. This nonlinear cochlear gain between 40 and 80 dB SPL was ~23 dB at the BF. However, nonlinear gain was also found at frequencies down to 5 kHz with higher intensities, i.e., 70–80 dB SPL, which was consistent with observations at the reticular lamina in the mouse and gerbil (Cooper et al. 2018; He et al. 2018; Lee et al. 2016; Ren et al. 2016). Finally, the phase demonstrated progressive lags with increasing frequency, which is consistent with traveling wave propagation.

Averaged vibratory measurements.

Altogether, twelve gerbils contributed to the main findings. We found similar vibratory responses from all of them measured at multiple points within the OoC throughout the three turns. Averaging the data between animals is a common technique used to reduce noise and highlight the vibratory characteristics important to cochlear function. However, because it was not always possible to measure at identical locations, we could not average all the data together. Therefore, we created averaged data sets from live gerbils that included only those responses that were collected at the three measurement spots close to that shown in Fig. 1B. Thus the averaged data came from five gerbils in turn 1, eight gerbils in turn 2, and four gerbils in turn 3.

The frequency responses of the averaged displacement magnitudes demonstrated several key differences (Fig. 2, A–C). In turn 1, averaged data were performed at the normalized frequency relative to the BF defined by peak frequency at the low sound intensity of 40 dB SPL. Consistent with well-established BM responses measured from the transverse direction of the BM (Ren et al. 2011), there was the classical sharply tuned response to low-intensity stimuli that peaked at the BF, had a normalized frequency equal to 1, and became broader as the stimulus intensity increased. However, in turn 3, the response was nearly completely low pass in nature, and tuning did not vary with the stimulus intensity. In turn 2, the vibratory response appeared to be in between turn 1 and turn 3 characteristics. The response was low pass in nature, with a superimposed area of sharpened tuning to low-intensity stimuli around 2.5 kHz. In addition, the maximum displacement near the corner frequency varied with the turn and roughly increased by an order of magnitude with each turn.

Fig. 2.

Comparison of vibratory data between turns 1, 2, and 3. Data from turns 3, 2, and 1 are averaged responses from 4, 8, and 5 gerbils, respectively. All error bars are the SE. A–C: displacement magnitudes. D–F: sensitivity calculated by normalizing the displacement magnitude by the sound intensity. G–I: input-output (IO) curves for the frequency of maximal gain (F_maxG), a half octave below this (0.71*F_maxG), and a full octave below this (0.5*F_maxG). Average for turn 1 data was performed by normalizing frequencies relative to the peak frequencies of each recording defined as the maximum displacement at low stimulus intensities.

We normalized the displacement by the stimulus intensity to calculate the sensitivity (Fig. 2, D–F). This highlighted the differences in cochlear amplification for low-intensity stimuli in the different turns. In turn 1, although the sensitivity presented here was from 40 to 80 dB SPL, the sensitivity was sharply tuned, whereas in turn 3 it was broadband. In turn 2, sensitivity was broadband but with a peak at ~2.5 kHz for low-level stimuli (10–20 dB SPL).

To quantify these differences, we calculated the gain of cochlear amplification by dividing the sensitivity to the lowest stimulus intensity (10–40 dB SPL) by the sensitivity to the highest stimulus intensity (70–80 dB SPL). In high-frequency cochlear regions, it is common to measure gain at the frequency of maximal vibration to low-stimulus intensity, the BF, because the gain is maximal at that frequency. However, given the differences in vibratory characteristics, this is not necessarily appropriate for low-frequency cochlear regions because there was not a well-defined BF. Because the turn 1 gain was only available between 40–80 dB SPL, we made these calculations between 40 and 80 dB SPL at the frequency with a local maximum gain (F_maxG) and one octave below this frequency (0.5*F_maxG) based on available measurements (Fig. 3, A and B).

Fig. 3.

Comparison of gain, tuning, and traveling wave morphology between turns 1, 2, and 3. A: peak gain. All error bars are the SE. B: gain 1 octave below the frequency of the peak gain. C: difference in gain between A and B. D: Q_6dB for the displacement threshold vibratory tuning curves. E: simulated traveling wave profiles created by interpolating the average displacement and phase data for each turn along the length of the cochlea. The predicted number of inner hair cells (IHCs) that would have bundle deflections large enough to evoke a receptor potential is given for each turn. SPL, sound pressure level. *Statistically different, P < 0.05.

This demonstrated that the local maximum gain between 40 and 80 dB SPL was higher for turn 2 than for turn 3 (P = 0.0004, nonpaired 2-tailed t-test) and for turn 1 (nonpaired 2-tailed t-test, P = 0.001). The maximum gain for turn 3 was not statistically different from turn 1 (nonpaired two-tailed t-test, P = 0.86). As described above, at 0.5*F_maxG, there was cochlear gain in turns 2, 3, and also in turn 1 at high sound intensities (nonpaired 2-tailed t-test, P = 0.05 and P = 0.06 for turn 2 and 3 vs. turn 1, respectively). Thus the difference between the gains at F_maxG and 0.5*F_maxG (ΔGain) dropped moving from turn 1 to turn 2 to turn 3 (Fig. 3C). Therefore, the turn affected the difference in gain between an octave by ~3.8 ± 6.5 dB per turn (linear regression, R² = 0.38, P = 0.75). Together, these data demonstrated that the frequency tuning of cochlear amplification becomes progressively sharper moving from the apex to the base although the gain does not change correspondingly.

To further demonstrate differences in the pattern of cochlear amplification between the three turns, we plotted the vibration magnitude vs. stimulus intensity (also called input-output curves, Fig. 2, G–I). This was done at F_maxG, a half-octave below F_maxG (0.71*F_maxG), and one octave below F_maxG (0.5*F_maxG). For comparison, a linear growth response at 1 dB/dB was also plotted (dashed line). In turn 3, there was strong compression at all three frequencies, and it was present even at the lowest stimulus intensities. The responses increased nonlinearly with a slope of 0.2–0.4 dB/dB consistently over the entire range of stimulus intensities we tested. In turn 2, compression was strongest at F_maxG, weaker at 0.75*F_maxG, and weaker still at 0.5*F_maxG. Additionally, the responses increased at ~0.8 dB/dB (nearly linear) from 10 to 20 dB SPL at all three frequencies. At F_maxG, the responses nearly saturated with a slope of 0.1–0.3 dB/dB up to 80 dB SPL, whereas, at the other two frequencies, the degree of compression was less and the responses became almost linear above 60 dB SPL. In turn 1, the responses increased in a compressive nonlinear manner at ~0.4 dB/dB only at F_maxG. At the other two frequencies, the responses increased almost linearly up to 60 dB SPL, then compressively at higher intensities. The compression at higher intensities was similar at all three frequencies. Together, these data confirm that the frequency dependence of amplification differs between the three turns.

We then quantified the Q_6dB of the displacement threshold curves for the three turns at 40 dB (Fig. 3D). This was done by dividing the frequency of the local maximum by the bandwidth 6 dB below this local maximum (i.e., Fig. 2, D–F). In data sets where the low-frequency side of the tuning curve did not decrease 6 dB (i.e., Fig. 2, D–F), we used 0 Hz as the low-frequency side of the bandwidth calculation. This method of quantifying tuning curve sharpness demonstrated the dramatic loss of frequency tuning in lower-frequency regions of the cochlea, with each turn changing the Q_6dB of the mechanical response by ~1.00 ± 0.49 (linear regression, R² = 0.8, P = 0.29). Taken together with the previous findings, these data indicate that the mechanical tuning of the OoC changes along the length of the cochlea and that mechanical tuning of the OoC does not solely define AN tuning.

Estimating the spatial pattern of IHC stimulation.

Based on our new measurements in turn 1, 2, and 3, we mapped the data to predict the spatial response of the traveling wave. We used the tonotopic map of the gerbil cochlea based on single-unit AN recordings (Müller 1996) to define the frequency layout of the cochlea. We then interpolated the vibratory data from Fig. 2 of the three turns to create predicted mechanical responses along the entire length. Thus we did not assume scaling symmetry. We then calculated the traveling wave pattern to three 40-dB SPL tones using frequencies tuned to each of the three turns and evenly spaced logarithmically (Fig. 3). On the basis of the shape of the traveling wave envelope, also known as the spatial excitation pattern, the model predicted that lower-frequency sounds produce OoC vibrations that are greater in magnitude and span a greater longitudinal distance.

IHC stereocilia are arranged in rows, running almost parallel to the spiraling longitudinal axis of the cochlea (Slepecky 1996), which make them sensitive to radial deflections. To quantify this finding, we attempted to estimate the number of IHCs that would theoretically be stimulated. Because of variations in the OoC anatomy, hair cell size, and stereociliary bundle length along the length of the cochlea, there is no agreed way that best defines what measurement to use as the threshold to define a region of activation. Therefore, for simplicity, we decided to use an OoC displacement of 1 nm to define this threshold, and we assumed that 10 μm is the average longitudinal distance between each column of hair cells in the gerbil cochlea (Karavitaki and Mountain 2007a). On the basis of these assumptions, we estimate that the number of IHCs predicted to be stimulated by OoC vibration increases as the sound frequency decreases, from 105 in turn 1 to 344 in turn 2 to 572 in turn 3. As a guideline, gerbils have ~1,110 IHCs. Thus we predict that about half of the hair cells in the cochlea are mechanically stimulated in response to a low-frequency tone presented at 40 dB SPL.

DISCUSSION

Much needs to be understood about low-frequency cochlear mechanics below ~5 kHz, in which there are important speech and music contents. In this study, we report the first in vivo vibratory measurements within OoC of all three turns from the intact gerbil cochlea. One key finding of this study was that the mechanical compressive nonlinear cochlear gain was as large in low-frequency turn 2 as previously reported in high-frequency turn 1 measured along the transverse direction at the level of the BM, ~60 dB between active and passive preparations (Robles and Ruggero 2001), but lower in the very-low-frequency turn 3. The compressive cochlear gain of turn 3 was not tuned with a BF as described for high-frequency basal BM turn 1 responses. Instead, the compressive gain was distributed over a much larger frequency range. In turn 1, the compressive cochlear gain of OoC presented in a broader region more than a half octave below the BF at moderate stimuli intensities and more than one octave at high sound pressure level, i.e., 70–80 dB SPL. This finding was consistent with the new data found in the mouse and gerbil reticular lamina and the BM vibrations (Cooper et al. 2018; He et al. 2018; Lee et al. 2016; Ren et al. 2016). This was likely because our measurements came from within the OoC near the top of the OHCs, the reticular lamina. Another key finding was that our data demonstrated within the OoC mechanical tuning became broader moving apically from the base, i.e., bandpass at turn 1, bandpass superposed on low pass at turn 2, and ultimately turning into a low-pass filter at turn 1. Thus compressive cochlear gain and sharp frequency tuning are not inextricably linked in the mammalian cochlea. Third, phase responses confirmed wave propagation from the cochlear base to apex via traveling waves.

A strength of this study was that, using VOCTV, we could record vibratory responses within the OoC noninvasively, thus increasing the reliability of the recordings over previous approaches that required opening the cochlea (Cooper and Rhode 1997, 1995; Dong and Cooper 2006; Khanna and Hao 1999; Zinn et al. 2000). This has the benefit in that it removes the possibility of changing cochlear hydrodynamics because opening the cochlea creates a high-pass filter effect (Dong and Cooper 2006; Recio-Spinoso and Oghalai 2017). Another strength was that our approach permitted vibratory measurements to low-intensity stimuli, especially at turn 2 and 3, where cochlear amplification typically has the most impact. A recent publication described experiments performed in the apex of unopened guinea pig cochleae using Fourier domain optical coherence tomography (Warren et al. 2016). However, the approach appears to be limited to intense sound stimuli, so the level dependence of amplification and frequency tuning could not be studied.

There are a few published measurements of mechanical responses at low-frequency apical region using the classical laser interferometer, which requires opening of the apex of the cochlea and mixing the perilymph and endolymph while tearing the Reissner’s membrane for delivering shinning reflective beads on the recoding structure, such as the transparent tectorial membrane (Cooper 2003; Cooper and Dong 2003; Cooper and Rhode 1995; Robles and Ruggero 2001). These manipulations lead to acoustic trauma and damage of the hearing. From these studies, the degree of nonlinearity at the low-frequency apical region with corner frequency ~1 kHz has been reported to be less than those well-established BM responses at high-frequency basal regions using laser interferometry (Robles and Ruggero 2001). Parallel to the present measurements, we recently reported measurements from the cochlear apex of the guinea pig using VOCTV technique (Recio-Spinoso and Oghalai 2017). Similar low-pass responses with corner frequencies of 0.2 kHz in turn 4 and 0.4–0.6 kHz in turn 3 were reported. Importantly, nonlinear gain and compression were reported both below and above the low-pass corner frequencies. Our turn 3 responses were consistent with this conclusion. The results confirmed the conclusion that AN low-frequency suppression was cochlear originated (Temchin et al. 1997). The differences in the degree of nonlinearity between further basal, i.e., turn 2, and the very apical regions, turn 3, are still under investigation.

The acousto-mechanical tunings measured at low-frequency apical regions using classical laser interferometer are commonly reported to have two peaks, separated by a distinct mid-band sensitivity notch or bandpass filter shape (Cooper and Rhode 1997, 1995; Gummer et al. 1996; Khanna et al. 1989), which was found to be due to artifacts in the mechanical responses caused by opening of the otic capsule bone. In a detailed study of acousto-mechanical effects of invading the cochlea, others (Dong and Cooper 2006) demonstrated that the degree of sealing of the opening varied the relative size of the slow-traveling and fast-compression-wave components. The interaction between the two components led to different tuning shapes. Under the well-sealed condition (i.e., near intact), the mechanical tuning at the low-frequency apical region was more like low-pass filter shape, which differs from the filtering seen in individual AN fibers with similar characteristic frequencies (Huang and Olson 2011; van der Heijden and Joris 2003). The low-pass filter shape of tuning was further confirmed by a recent study using VOCTV technique (Recio-Spinoso and Oghalai 2017) in guinea pig as well as in our turn 3 responses from an intact gerbil cochlea.

A downside of this study is that we measured from a location within the OoC because the signal-to-noise ratio was best, and we had the lowest noise floor. In contrast, it has been much more common to record vibrations from the BM at the cochlear base or from the tectorial membrane and/or Henson cells at the cochlear apex. In addition, our measurement angle was more radial than transverse. Figure 4 illustrates similarities and differences between radial motion within the OoC and well-established transverse motion of the BM responses measured at turn 1 (Ren et al. 2016). In general, the mechanical characteristics within the OoC have been shown to be similar to those measured from the transverse direction at the BM in that responses increased with stimuli intensities in a compressive manner, the degree of cochlear gain was similar, and phases accumulated with increasing frequencies, confirming the propagation of traveling waves. However, there are some differences between responses of the OoC and the BM, i.e., responses of the OoC showed greater amplitude, broader tuning, and the existence of nonlinearity at high SPL (Cooper et al. 2018; He et al. 2018; Ren et al. 2016).

Fig. 4.

Comparison of turn 1 vibratory data between the organ of Corti (OoC) measured by the optical coherence tomography (OCT) system and the basilar membrane (BM) using traditional laser interferometer. The magnitude (A) and phase (B) of OoC vibratory displacement (Expt. #20) are shown. The magnitude (C) and phase (D) of BM responses are measured by a laser interferometer (Ren et al. 2011). Sound pressure levels (SPL) were 30 to 80 dB SPL in 10-dB steps. The phase responses were normalized to middle ear (ME) responses.

A classical cochlear mechanics model (Davis 1965) suggests that hair cells within the OoC move together with the traveling wave. Mechanoelectric transduction of hair cells depends on the shear motion between the reticular lamina and tectorial membrane that causes the stereocilia to pivot. When stereocilia pivot toward the tallest stereocilia, transducer channels open, and the hair cell depolarizes. In OHCs, which are unique to mammals, voltage changes elicit mechanical forces via electromechanical transduction. These forces act within the traveling wave to boost and sharpen the mildly frequency-resolved pattern of vibration of the passive cochlea, which is known as the cochlear amplification process. Within this framework, all cellular structures are assumed to be solid except that only the OHCs have motility. A recent study made simultaneous measurements of OHC extracellular voltage and BM displacement within turn 1 of the gerbil cochlea and demonstrated that, around the BF, OHC force and BM velocity in the transverse direction were in phase (Dong and Olson 2013).

Understanding the cochlear amplifier relies on observation of the motions of the cochlear partition both in the transverse and radial directions. More direct measurements in mouse and gerbil at both the reticular lamina and BM demonstrate that the top and bottom of the OoC do not move in unison (He et al. 2018; Ren et al. 2016) along the transverse direction with further evidence that Deiters’s cells change length cyclically during sound stimulation by about two times that of the OHCs (Cooper et al. 2018). In addition, substantial radial motions in the BM, reticular lamina, and the tectorial membrane were evidenced in intact mouse cochleae recently using VOCTV (Lee et al. 2016) and at the high-frequency basal region in gerbil using OCT (Cooper et al. 2018). These pieces of evidence seem to suggest that motions within the cochlear partition are more complex than that described in the classical Davis model and consistent with the suggestion that there are multiple motions within the cochlea (Karavitaki and Mountain 2007a, 2007b). In a recent theoretical study, complex motions among cellular structures of the OoC are achieved by a finite-element model (Motallebzadeh et al. 2018). With realistic material properties and cellular arrangement within the OoC of a mouse cochlea, including OHCs, Deiters’s cells, and Y-shaped phalangeal process, the model predicts that the vibrations at the reticular lamina have a higher degree of compressive nonlinearity and broader tuning compared with those of the BM, consistent with experimental observations. The finite-element model provides insights on how OHCs work together to achieve high hearing sensitivity and frequency selectivity and appears to be the direction to provide further explanation on cochlear micromechanics that is still not well understood.

The cellular structure of the OoC displays a longitudinal gradient in which there is an increase in size from the base to the apex (Slepecky 1996), i.e., the cells in the apex are larger than those in the base, the stereocilia on the hair cells are longer and less stiff, the BM is wider, and the tectorial membrane has a greater mass. Because the IHC stereocilia are freestanding, not like the top of the OHC stereocilia firmly attached the tectorial membrane, the mechanical drive to the IHC stereocilia is proposed to be the fluid flow radially past the IHC stereocilia, not by direct mechanical shear as in the OHCs (Freeman and Weiss 1990; Guinan 2012; Sellick and Russell 1980). Radial motion of the BM, reticular lamina, and the tectorial membrane appeared to be controlled by passive mechanics of the cochlear partition, which may play a role in deflecting hair bundles and exciting hair cells. Therefore, our radial measurements, which demonstrated different tuning curves along the cochlea, suggest that there are fundamental differences between how or how many IHC stereociliary bundles are stimulated for different frequency sounds.

Although significant differences between the AN fiber threshold tuning curves and mechanical tuning responses have long been appreciated below the characteristic frequency at high-frequency basal regions (Narayan et al. 1998; Yoon et al. 2011), our data showed that these differences are significantly greater in the low-frequency apical regions of the cochlea at the level of OoC. Ultimately, the alterations in the shape of the mechanical tuning curve enlarge the bandwidth so much that this frequency tuning is obviously broader than that of the AN fibers that innervate the IHCs in that region of the cochlea (Guinan 2017). Thus our data dispel the long-held notion that the electrochemical responses of AN fibers mirror the vibratory responses of the OoC. Our data suggest that OoC vibration alone does not establish AN tuning in low-frequency regions of the mammalian cochlea. Whether stereociliary bundle deflection is sharply tuned although OoC vibration is not remains to be determined. On the other hand, in nonmammals, BM mechanical responses are either not tuned or broadly tuned, whereas their AN responses are sharply tuned (O’Neill and Bearden 1995; Peake and Ling 1980; Xia et al. 2016). This is because electrical resonance intrinsic to the hair cell produces sharp tuning (Crawford and Fettiplace 1981; Fuchs et al. 1988; Lewis and Hudspeth 1983). Although gerbil hair cells have not been carefully tested for electrical resonance, guinea pig OHCs do not have it (Ashmore and Meech 1986). However, in vivo intracellular recordings from both OHCs and IHCs indicate that their receptor potentials are as sharply tuned as single-unit AN responses in the guinea pig (Dallos 1986; Dallos et al. 1982; Russell and Sellick 1978). At low-frequency apical regions, the quick drop off of the mechanical response above the corner frequency seems to match that of the neural response; the low-frequency side does not. Both the gerbil and guinea pig apex data suggest that a high-pass filter that is not detectable with our vibratory measurements must be altering the sound-encoding process.

There may be three possible mechanisms that could underlie this additional filtering. The first possibility is that the IHC stereociliary bundle may not be deflected with the same frequency response as the point within the OoC we measured. Other than a direct measurement of IHC stereociliary bundle deflection, it is unclear what recording position provides the best surrogate measurement. Still, the point we measured near the apical surface of the OHCs is certainly closer to the IHC stereociliary bundle than the more commonly measured BM. Nevertheless, the fluid mechanics of bundle stimulation within the subtectorial space are unknown, and it is possible that fluid velocity within the subtectorial space (rather than fluid displacement) is what stimulates the IHC bundle, creating what would appear to be a high-pass filtering process on displacement tuning curves (Robles and Ruggero 2001). The second possible mechanism is via fast adaptation of the stereociliary bundle. Adaptation acts as high-pass filter transduction currents in ex vivo studies (Kennedy et al. 2003; Ricci et al. 2005). In vivo, this should make the hair cell receptor potential and AN responses appear to be sensitive to velocity rather than displacement, consistent with published results (Russell and Sellick 1983; Zwislocki and Sokolich 1973). The third possible mechanism is that tonotopic gradients in voltage-dependent ion channels within IHCs provide the filtering, possibly by altering the kinetics of synaptic release (Johnson et al. 2011; Johnson and Marcotti 2008).

The data confirmed the findings from other species that the low-frequency cochlear apex has different mechanical responses than in the high-frequency cochlear base. The key difference is that tuning in the apex is much broader than tuning in the base (Shera et al. 2010). Furthermore, we extended these findings to show that cochlear amplification and compression in the apex is as large (or nearly as large) as in the base but that it operates over a much broader frequency range above and below the low-pass corner frequency. Recent data suggest AN correlates to the enhanced reticular laminar mechanical measurements from the apex (Nam and Guinan 2018). They have found regions at the side lobe frequency regions of low-frequency tuning curves that are efferent inhibited (Stankovic and Guinan 1999, 2000) and have strong effects attributable to low-frequency bias tones. From this, they hypothesize that cochlear regions on either side of the characteristic frequency region have a high degree of mechanical amplification. The side lobe may correspond to the above corner frequency mechanical response seen in the gerbil (present work) and in the guinea pig (Recio-Spinoso and Oghalai 2017).

These observations are consistent with the notion that the mammalian cochlea is mechanically different along its length. It has been described as being split into two regions, the high-frequency basal and the low-frequency apical regions, with apical-basal transition frequency of 1 kHz (human) or 4 kHz (chinchilla) corresponding to a cochlear map fraction of 0.4 to 0.7, respectively (Shera et al. 2010). In the basal region, normalized tuning is roughly shifted similarly and is thus constant across cochlear location, known as cochlear local scaling symmetry, according to this theory. Conversely, in the apical region, normalized tuning changes rapidly as a function of distance from the cochlea and thus is not shifted similarly. Although we only compared data from three different cochlear locations and cannot definitively prove or disprove this concept, our data were instead consistent with a smoother gradient in tuning along the length of the cochlea, progressively morphing from a bandpass filter response with sharply tuned gain in the basal high-frequency region into a low-pass filter response with broadband compressive gain in the apical low-frequency region of the cochlea (Figs. 1 and 2). Furthermore, the data suggest that the number of hair cells stimulated by a tone stimulus increases as the frequency is lowered. Thus the encoding of sound within the AN must be based, not only on the mechanical response of the OoC and the gain of cochlear amplification provided by OHCs, but also on an additional biophysical filtering mechanism. This filter is superimposed on the vibratory response to sharpen tuning by high-pass filtering the signal before it reaches the AN.

GRANTS

The research is supported by NIH-NIDCD DC011506 (W. Dong), DC07910 (S. Puria), DC014450, DC013774, and DC010363 (S. Oghalai).

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

W.D., A.X., and J.S.O. performed experiments; W.D., A.X., P.D.R., B.E.A., and J.S.O. analyzed data; W.D., A.X., S.P., and J.S.O. interpreted results of experiments; W.D., A.X., and J.S.O. prepared figures; W.D., A.X., S.P., and J.S.O. drafted manuscript; W.D., A.X., S.P., and J.S.O. edited and revised manuscript; W.D., A.X., and J.S.O. approved final version of manuscript.