Abstract
The detection of sound by the mammalian hearing organ involves a complex mechanical interplay among different cell types. The inner hair cells, which are the primary sensory receptors, are stimulated by the structural vibrations of the entire organ of Corti. The outer hair cells are thought to modulate these sound-evoked vibrations to enhance hearing sensitivity and frequency resolution, but it remains unclear whether other structures also contribute to frequency tuning. In the current study, sound-evoked vibrations were measured at the stereociliary side of inner and outer hair cells and their surrounding supporting cells, using optical coherence tomography interferometry in living anesthetized guinea pigs. Our measurements demonstrate the presence of multiple vibration modes as well as significant differences in frequency tuning and response phase among different cell types. In particular, the frequency tuning at the inner hair cells differs from other cell types, causing the locus of maximum inner hair cell activation to be shifted toward the apex of the cochlea compared with the outer hair cells. These observations show that additional processing and filtering of acoustic signals occur within the organ of Corti before inner hair cell excitation, representing a departure from established theories.
Introduction
A major focus of hearing research is to elucidate how sound stimuli are transduced into neural responses. Auditory nerve excitation is triggered by the depolarization of the inner hair cells (IHCs), which is induced by sound-evoked deflection of their stereocilia (ter Kuile, 1900; Flock, 1965; Rhode and Geisler, 1967; Hudspeth and Corey, 1977; Dallos, 2003). Although the sound-evoked motion is generated primarily by the vibrations of the basilar membrane, because of its fluid coupling via the traveling wave, active force production by the outer hair cells (OHCs) enhances these vibrations by a factor close to 1000 (Brownell et al., 1985; Ashmore, 1987; Chen et al., 2011; Zheng et al., 2011). It is, however, not clear how the vibrations of the basilar membrane are transmitted to the IHCs or how IHCs are influenced by the active forces produced by the OHCs (Ashmore et al., 2010).
With the exception of a recent report (Chen et al., 2011), high-frequency sound-evoked vibrations have been measured in vivo only at the basilar membrane (Rhode, 1971; Khanna and Leonard, 1982; Robles et al., 1986; Nuttall et al., 1991; Nilsen and Russell, 1999; Cooper, 2000; Rhode and Recio, 2000; Ren, 2002). Vibrations of the stereociliary side of the organ of Corti have mostly been measured in vitro in the low-frequency areas of the cochlea (Ulfendahl et al., 1996; Hemmert et al., 2000; Zinn et al., 2000; Chan and Hudspeth, 2005; Fridberger et al., 2006; Nowotny and Gummer, 2006; Karavitaki and Mountain, 2007; Tomo et al., 2007; Hakizimana et al., 2012; but see Cooper and Rhode, 1995). The paucity of in vivo vibration measurements from structures other than the basilar membrane leads to a significant gap in our understanding of inner ear function. Mathematical models as well as auditory nerve recordings suggest that additional processing of the acoustic signal could occur within the organ of Corti, and early, now largely disregarded work even suggested the presence of a “second filter” that would sharpen the frequency resolution of the hearing organ (Evans and Wilson, 1975). Direct experimental evidence has, however, been lacking.
In the current study, we investigated sound-evoked vibrations across the width of the hearing organ. We show that different cell types have different vibration patterns and different dependence on the sound stimulus level. The results suggest that multiple vibration modes exist at the stereociliary side of IHCs and OHCs, and show that the frequency tuning of the IHCs is different from other cell types nearby. Additional filtering of acoustic signals therefore occurs en route to the IHCs, thus reinstating the second filter, although transformed in shape.
Materials and Methods
Surgery and experimental procedures.
Normal-hearing albino guinea pigs (250–350 g) of either sex were prepared for recording of cochlear vibrations using procedures approved by the Institutional Animal Care and Use Committee of Oregon Health & Science University. These procedures were described previously (Nuttall et al., 1991; Fridberger et al., 2004; Zheng et al., 2011). In brief, after inducing anesthesia with a combination of ketamine (40 mg/kg) and xylazine (10 mg/kg), the animal was placed in a customized heated head-holder and the cochlea surgically exposed through a ventrolateral approach. The condition of the inner ear was monitored throughout surgery by a pair of low-level continuous tones at 18 and 18.9 kHz. These tones generated a 900 Hz electrical distortion product (Nuttall and Dolan, 1994, 1996), which was measured using a silver wire electrode positioned in the round window niche. Whenever the distortion product amplitude decreased, surgery was halted to allow recovery. Audiograms were also acquired at regular intervals by measuring the compound action potentials of the auditory nerve (10 μV response criterion), using the round window electrode. All 12 animals included in the analysis had <10 dB loss of compound action potential threshold at the time when vibration measurements began.
After the exposure and opening of the bony bulla, the middle ear muscles were cut and a window created in the basal turn of the cochlea by gradually thinning the bone. To identify the structure where vibrations are measured, it is necessary to have a good image of the cochlear partition. This requires a 400- to 500-μm-diameter opening located close to the round window. This relatively large opening allowed more focused light to fall on the hearing organ and permitted observation across its entire width.
During the experiment, a cross-section image was first acquired (Fig. 1A,B), guiding the selection of the measurement site. Two precision galvanometric mirrors then directed the light from the optical coherence tomography (OCT) system to the selected site, and vibration measurements were conducted with this homodyne interferometer system.
OCT.
Sound-evoked vibrations were measured at the luminal surface (the reticular lamina [RL]) of the hearing organ, near the base of OHC and IHC stereocilia and their surrounding supporting cells, using OCT (Choudhury et al., 2006; Chen et al., 2011). In brief, the light from a wide bandwidth superluminescent diode (1310 ± 47 nm) was focused on the inner ear through a custom microscope, and the back-reflected light made to interfere with a reference beam on a sensitive detector. Because of the short coherence length of the light source, the length of the sample and reference arms of the interferometer need be within 7 μm of each other for interference to occur. This distance is also the axial resolution of the system, which therefore allows measurements from structures deep inside the hearing organ without influence from other structures in the beam path. The numerical aperture of the microscope lens is not a limiting factor because axial resolution arises from the short coherence length of the light source. The infrared light has deep penetration capability allowing vibration inside the organ of Corti to be measured.
Data analysis and statistics.
Image processing was performed using MATLAB (MathWorks; release 2012b). The areas of the outer and inner tunnel of Corti were computed using an algorithm that evaluated the 2 × 2 neighborhood of each pixel. Lines corresponding to the RL, OHC long axis, and the basilar membrane were drawn on each individual image using standard MATLAB tools. Figure 1C shows the average angles used. The data were brought into a standard coordinate system using affine transformations and Student's t test used for evaluating differences in inclination between the basilar membrane and other structures.
The experimental protocol involved vibration measurements at five positions (the IHC apex, pillar cells [PCs], OHC row 1 and 3, and the Hensen cells [HeCs]) and four different sound pressure levels, leading to a requirement for 20 measurements to be performed in each animal. Repeated measurements led to correlations within the dataset that need to be considered during statistical analysis. Experimental difficulties made it impossible to acquire complete datasets from every preparation, and the optical reflectivity varied between cell types (Fig. 1D), leading to higher variance in measurements obtained from IHCs, PCs, and HeCs. Linear mixed models (Laird and Ware, 1982; see also Galecki and Burzykowski, 2013) are designed to account for such factors and were therefore used to determine whether significant differences in response magnitude were present among cell types. The dependent variable was the log-transformed displacement at the best frequency. The cell type and sound pressure level were treated as fixed categorical predictors, and the random effect term included subject-dependent intercepts and slopes. The model assumed different variances across cell types.
To determine whether there were significant differences in best frequency among cell types, we used the model described above, but the random effect only included a subject-dependent intercept because inclusion of a subject-dependent slope did not improve the fit. All calculations were performed using the lme package in R (version 2.15.3; R Core Team, 2013).
Results
Imaging the organ of Corti in vivo
All experiments began by imaging the organ of Corti, to identify structures where vibration measurements were to be performed. Images of the undisturbed, normally functioning hearing organ may be useful for mathematical modeling; a summary of key features is therefore presented here. The resolution does not permit subcellular structures to be examined, but key elements are clearly visible (Fig. 1A). To quantify morphological features across preparations, lines were drawn parallel to the basilar membrane (Fig. 1B, red line), RL (green line), and the long axis of the cylindrical OHCs (blue line). The outline of the outer and inner tunnel of Corti was marked. The averaged parameters (n = 11) were used to draw the relations shown in Figure 1C. The RL, which connects IHCs and OHCs, was inclined 16 ± 1° (mean ± SEM) with respect to the basilar membrane (p < 0.001, t test). The corresponding angle for the OHCs was 46 ± 4° (p < 0.001, t test). The average area of the outer tunnel of Corti (1100 ± 170 μm2) was not significantly different from the inner tunnel (1700 ± 300 μm2; p = 0.11), which is surprising because most previous images of the organ of Corti show a quite small outer tunnel of Corti. This suggests that fixation and sample processing can significantly alter hearing organ structure, particularly with regard to the fluid-filled spaces.
The optical reflectivity varied across cell types (Fig. 1D). Using the basilar membrane as the reference, HeCs and IHCs returned less of the incident infrared light to the detectors, whereas the first row of OHCs had high reflectivity. The signal-to-noise ratio is therefore better when measuring OHC vibrations, a fact that was considered in the statistical analysis by allowing variances to differ according to cell type.
Figure 1E shows data from a preparation where basilar membrane, OHC, and IHC vibrations were measured at sound pressures ranging from 34 to 94 dB SPL. The basilar membrane showed maximal responses at 21 kHz and compressive nonlinearity. This nonlinearity reduced the gain from 370 to 135, 24, and 4 nm/Pa as the sound pressure level increased from 34 to 94 dB. In agreement with Chen et al. (2011), larger displacements were found at the first row of OHCs (Fig. 1E, middle), and the best frequency was higher (21.5 kHz). IHCs showed smaller vibrations than OHCs, and peak vibration was found at an even higher frequency (∼22 kHz; Fig. 1E, right).
The phase of sound-evoked motions is shown in Figure 1F. In all cases, there was a point where the phase was invariant as the stimulus level increased. This phase-invariant point coincided with the point of maximal vibration at both the basilar membrane and at the OHCs. The IHCs showed more pronounced phase variations at low frequencies, and the phase curves crossed at 22.75 kHz, slightly higher than the location of maximum vibration.
Because basilar membrane vibrations were extensively characterized in the past, and our basilar membrane results do not differ from previously published data from sensitive ears, we focused data collection on structures along the RL.
Frequency-response varies across the width of the hearing organ
The single-animal data shown in Figure 1E, F suggest that sound stimuli are locally filtered within the hearing organ. To determine whether this truly is the case, we acquired data from 11 additional animals. We made attempts to always acquire data from the same position along the cochlear spiral, but the best frequency of the recording location nonetheless varied slightly across preparations. To create averaged tuning curves for each sound stimulus level and cellular structure, we therefore normalized the frequency axis before averaging the data, resulting in the tuning curves shown in Figure 2 (mean ± SEM, n = 12).
The largest movements were found in the OHC region, where the mean displacement at the best frequency was 18 ± 2 nm at 94 dB SPL. Other cell types had smaller motions (PCs, 10 ± 1 nm; IHCs, 7 ± 1 nm; HeCs, 4 ± 1 nm). Using the IHC displacement as the reference, these differences were statistically significant (p = 0.03 for the PCs; p < 0.001 for the first and third rows of OHCs, p = 0.02 for the HeCs; linear mixed model). As expected, alterations in the stimulus level changed the displacement magnitudes, but OHCs continued to show the largest displacements at all sound levels. At 54 and 74 dB SPL, displacements of PCs are larger than IHCs, but a transition appears to occur near 54 dB SPL, where PCs have slightly smaller average displacement than the IHCs. It is challenging to measure responses to soft sounds, and displacements were found to be close to the noise floor at 34 dB SPL in many cases. We therefore restricted all subsequent analysis to data acquired at 54, 74, and 94 dB SPL.
Alterations in stimulus level changed the frequency to which the hearing organ responded maximally. At 94 dB SPL, tuning curves from OHCs and IHCs had similar shape (Fig. 3A), without apparent differences in their best frequencies. When decreasing the stimulus level, IHCs began to respond better at higher frequencies than the OHCs (Fig. 3B,C).
These differences were apparent also when examining averaged data (Fig. 3D–F). At 94 dB SPL, all structures shared a best frequency near 16.5 kHz (range 16.4–16.7 kHz; n = 9 for IHCs, 10 for OHC row 1, 11 for OHC row 3, and 8 for HeCs), but at lower stimulus levels, the best frequency increased. The IHC best frequency was 16.4 kHz at 94 dB SPL, 20.3 kHz at 74 dB SPL, and 21.3 kHz at 54 dB SPL. Similar trends were observed for the other cellular structures, and the effect was significant across cell types (p < 0.001). The slope of the linear mixed model was 150 Hz/dB, indicating that a 20 dB decrease in the stimulus level increased the best frequency by 3 kHz on average. This shift is expected to reach a maximum of half an octave, ∼6.8 kHz in this case, for the lowest stimulus level (Cody and Johnstone, 1981; Johnstone et al., 1986; Nuttall and Dolan, 1996; Ruggero et al., 1997); the best frequency at the top of the OHCs was previously shown to be higher than at the basilar membrane (Chen et al., 2011)). Thus, in all cell types, the best frequency shows a significant variation dependent on the stimulus level.
For stimulus levels <94 dB SPL, the best frequency also varied across cell types and an orderly progression was evident, where the IHCs had the highest best frequency (21.3 ± 0.5 kHz; Fig. 3F), followed by OHCs (20.2 ± 0.7 kHz for row 1 and 20.1 ± 0.3 kHz for row 3) and HeCs (18.8 ± 0.8 kHz). These differences were statistically significant (p = 0.03 for the comparison between IHCs and the first row of OHCs; p = 0.005 for the third row of OHCs; and p = 0.001 for HeCs). The best frequency of PCs (21.1 ± 0.7 kHz), however, was not significantly different from the IHCs (p = 0.99).
These radial frequency gradients did not arise because of a longitudinal shift in the measurement location. We ensured the scan direction was aligned correctly along the radial direction by using a cover glass as a reference in the OCT images. An excessive longitudinal misalignment of ∼200 μm would be necessary to explain the >1 kHz radial frequency gradients observed in Figure 3. From our OCT images, the RL width, defined as the distance from IHC to HeCs, is ∼70 ± 10 μm. This value is close to 60 μm derived from Yarin et al. (2014). According to Yarin et al. (2014), the average RL width is 23 μm from the inner edge of IHC cuticular plate to the outer edge of the cuticular plate of third row of OHC for the basal location where our measurements are made. The shrinkage resulting from fixation is ∼50%, which leads to an RL width of 46 μm. The distance from the IHC and the HeCs, which is ∼30% longer than the RL width measured by Yarin et al. (2014), is 60 μm. In the presence of 200 μm longitudinal misalignment, the RL width in our OCT images would have to be ∼210 μm, which is not the case.
Differences in frequency tuning imply that different structures will vibrate with different phase because of the large phase changes inherent to traveling waves in the region near the best frequency. Because the vibrations of the organ of Corti are thought to be driven largely by the OHCs at low stimulus levels, we analyzed phases by computing phase curves relative to the third row of OHCs. At the highest sound level (Fig. 4, top row of plots), there were no consistent phase differences among the different structures. This implies that all structures vibrated as a single unit and that only a single mode of vibration was present. At lower sound levels, a phase lead became apparent for the IHCs (Fig. 4, left column of plots). A poor signal-to-noise ratio at 54 dB SPL prevented recordings of phase difference curves at frequencies higher than the peak of the tuning curve, but it nonetheless appears that the IHC phase lead extends to lower stimulus frequencies at this level. Recordings from PCs were quite difficult, and reliable phase data could only be acquired at 94 dB SPL, where the phase did not differ from that measured at the OHCs (data not shown).
Among the OHCs, there were no consistent differences in response phase between rows 1 and 3 at any level (Fig. 4, center column of plots), implying that all three rows of OHCs vibrate as a single body. This was not the case for the HeCs. Although data from these cells had a relatively high noise level, the curves indicate a phase lead reaching 1 radian at the best frequency at 54 dB SPL (Fig. 4, right column).
In summary, when a moderate stimulus level is used, the organ of Corti has three different functional regions: a medial region consisting of IHCs and PCs, where the best frequency is high and where relatively small vibrations phase-lead the central zone, which is composed of the OHCs. The central zone has higher best frequency than the lateral zone (composed of HeCs), but the lateral zone has a phase lead. At high levels, the pattern is different and all structures vibrate at similar phase.
Spatial vibration pattern along the width of the hearing organ
Instantaneous displacement profiles are plotted for different sound pressure levels in Figure 5. At 94 dB SPL, all structures vibrate in phase, characteristic of a structure vibrating in its fundamental mode (Fig. 5A). As the sound pressure decreases, the pattern of vibration becomes increasingly complicated; and at 74 dB SPL, a phase shift starts becoming apparent at the IHCs. By following the red trace in Figure 5B, it is seen that maximum OHC displacement no longer coincides with maximum IHC motion. This difference is clear also at 54 dB SPL (Fig. 5C), where a substantial phase difference between OHCs and IHCs is apparent, along with complicated displacement profiles at both HeCs and PCs. The average amplitudes for the different cellular structures are plotted in Figure 5D–F. Note the small displacements of IHCs and HeCs at all sound pressure levels. As a first approximation, we would consider this radial pattern of vibration to be a consequence of superposition or weighted sum of the first (symmetric) and the second (antisymmetric) vibration modes of a simply supported structure. The presence of the second mode becomes more prominent as the stimulus level decreases, which is consistent with the excitation of this mode by an off-centric force emanating from the OHCs.
Variation in compressive nonlinearity along the width of the RL
The data shown above demonstrate that IHCs and OHCs vibrate differently when the sound level decreases. To examine this more closely, Figure 6 shows their displacement amplitudes at the best frequency as a function of stimulus level. IHCs have lower displacement magnitudes and a more pronounced compressive nonlinearity than the third row of OHCs. The more pronounced nonlinearity of the IHCs will mean that they map a large range of stimulus intensities onto a much smaller dynamic range of outputs, which is consistent with the known behavior of the auditory nerve.
Physiological basis for the radial variation in compressive nonlinearity
A plausible physiological basis for the observed variation in nonlinearity along the width of the RL is elucidated in this section. A rigorous mechanical model would require 3D representation of the RL, including its coupling to other structures in the organ of Corti. However, at this time, many of the distributed mechanical properties of the organ of Corti structures are unknown.
Therefore, for the purpose of elucidating the radial variation in compressive nonlinearity, we take a simplified approach and consider the RL to be analogous to a beam subjected to transverse loading (perpendicular to its long axis). Such a loading would involve a combination of a linear, passive, and uniform load because of the pressure from the acoustic stimulus and a nonlinear point load representing the OHC somatic active force. The beam is assumed to have simply supported boundaries formed by the IHCs and HeCs, where small but not zero vibrations are measured. However, the qualitative observations from this modeling analogy do not depend on the boundary condition. The beam analogy would be a closer representation of the dynamics of the RL under the following simplifying assumptions: (1) The RL has negligible longitudinal coupling. Under this condition, the RL could be considered as an array of beams. (2) RL can be considered as a “free body,” on which the passive pressure and the OHC somatic active force are applied as “external” forces.
The transverse displacement w(x) of a beam simply supported on both ends and subjected to transverse loading can be represented as a sum of multiple modes as follows: where L is the length of the beam (corresponds to RL width); the sine series represents the mutually orthogonal Eigen functions of a transversely loaded simply supported beam; coefficients Cn represent the normalized coordinates; and N is the number of modes needed for asymptotic convergence of the series. To represent RL modes excited near the characteristic frequency, the first two modes of the beam are sufficient (i.e., N = 2). As discussed in the previous section, the RL width in this study is longer than previous studies, such as Nowotny and Gummer (2006), which describe it as extending from the IHC cuticular plate to the third OHC cuticular plate. Those studies considered the RL as a rigid bar that translates vertically and rotates about the pillar cell. However, our measurements show that RL vibration pattern is indicative of a structure more flexible than the basilar membrane and resembles a combination of two sinusoidal modes as described in Equation 1. Nevertheless, over the mid-portion of its length, these two modes are consistent respectively with “vertical translation” and “rotation about the center.”
The net forcing per unit length on the beam f(x) = fp(x) + fa(x) where fp(x) = fp is a passive uniform load, and fa(x) = Fa δ(x − xa) is the active force applied as a point-force at x = xa corresponding to the position of the second row of OHC. Many of the parameters necessary to numerically determine the forces fp(x) and fa(x) are not known. For xa = L/4, the active force primarily excites the second mode, whereas the passive (uniform) pressure arising from the acoustic stimulus primarily excites the first mode.
Figure 7 shows the response of the beam subjected to transverse loading, an analogy to the dynamics of the RL. Figure 7A shows the transverse displacement of the beam at two different radial positions x/L = 0.25 (corresponding to OHC) and x/L = 0.6 (corresponding to IHC) showing different levels of nonlinear compression resulting from the same active force applied at xa/L = 0.25. In Figure 7B, the transverse displacement ratio of IHC relative to OHC from Figure 7A is compared with experimental data from Figure 6. Both model and experiment qualitatively agree to show that the IHC has more nonlinear compression than OHC.
This phenomenon could be explained as follows. With a single mode of vibration excited by a compressive nonlinear force, the ratio between the displacements at any two points along the beam is constant versus stimulus level. However, when there are multiple modes of vibration where one mode is preferentially excited by the compressive nonlinear force, then the ratio between the displacements at any two points along the beam is itself a function of stimulus level. This analysis shows that the presence of multiple modes of vibration could lead to a radial variation in compressive nonlinearity resulting from a single compressive nonlinear point force (analogous to active force in the organ of Corti). The change in the net radial vibration pattern with change in stimulus level, as seen in the measured data in Figure 5, is another manifestation of the same phenomenon.
Discussion
A fundamental tenet of auditory neuroscience is that the frequency tuning of the cochlea is expressed in the vibrations of the basilar membrane, which are controlled to a large extent by the OHCs. In this view, the OHCs also determine the input to the IHCs, which perform little further signal processing before releasing transmitter substance onto auditory nerve dendrites. In contrast, we demonstrate significant differences in frequency tuning, amplitude, and response phase among IHCs and OHCs as well as their surrounding supporting cells.
Our previous study demonstrated that the basilar membrane and RL, coupled through the OHCs, vibrated at different phase and were tuned to different frequencies (Chen et al., 2011). We also showed (Zha et al., 2012) sound level-dependent OHC length changes in vivo, which suggested power transfer to the organ of Corti by OHC somatic electromotility. Our current study, in addition to confirming these data, demonstrates three new findings. First, there are significant differences in the spatial vibration pattern across the width of the hearing organ. Whereas the basilar membrane has a unimodal vibration pattern, the RL has a sound level-dependent multimodal vibration with different response phase across the width. Second, the compressive nonlinearity varies along the width, with higher compression at the IHCs than at the OHCs. Third, as the stimulus level decreases, the frequency of peak vibration decreases along the width. Higher best frequency is seen at IHC and lower best frequency at the HeCs. The implications and plausible physiological basis of these findings are discussed next.
In most simple models of hearing organ function, a unimodal vibration pattern is assumed. This unimodal pattern means that all structures move together, with the same phase, which clearly is the case at high stimulus levels (Figs. 3, 4, and 5, top row). At low stimulus levels, vibration is driven by OHCs, resulting in a more complex vibration pattern where the response phase varies across the width of the hearing organ. The phase difference between IHCs and OHCs has important implications. Because the IHCs phase-lead the OHCs, which phase-lead the basilar membrane (Chen et al., 2011), the combined lead at the best frequency would be close to half a cycle, implying that IHCs are maximally excited when the basilar membrane is moving toward scala tympani. This response polarity is consistent with response polarities inferred from auditory nerve recordings (Ruggero et al., 2000), but not with simple models of IHC function.
The observed differences in compressive nonlinearity along the RL width, especially, larger compression at the IHCs, might explain how the IHC maps a large range of stimulus intensities onto a much smaller dynamic range of outputs. The mathematical analysis shows that such a radial variation in nonlinearity could arise from a nonlinear point-force excitation on a mechanical structure with multiple modes of vibration.
These data have important physiological consequences for hearing. First, they provide experimental confirmation that the mechanical vibratory input to the IHC does not solely arise out of the traveling wave that excites the basilar membrane, consistent with the hypothesis by Mountain and Cody (1999). Second, this local signal processing adds complexity to cochlear responses by shifting the locus of maximum IHC activity resulting from any particular sound. The magnitude of this effect may be obtained, as a first approximation, from Greenwood's frequency–position map (Greenwood, 1990), which suggests that the present 1 kHz frequency difference would shift the maximum IHC response ∼200 μm toward the apex of the cochlea compared with the OHCs. Such a shift may be inferred from comparing auditory nerve data (Liberman, 1982) with basilar membrane recordings (Khanna and Leonard, 1982), but subsequent work (Greenwood, 1990; Narayan et al., 1998) dismissed the shift as an effect of cochlear damage. The present direct measurements reveal that the shift is real and that the frequency tuning of the IHC is not fully expressed at the OHC or the basilar membrane, as previously thought (Narayan et al., 1998).
Finally, the differences in frequency tuning of the different structures in vivo necessitate reconsidering the existence of a transformed “second filter” in the cochlea. It has been suggested that cochlear filtering is a two-stage process: the first filtering by the basilar membrane is followed by a physiologically vulnerable second filter (Zwislocki and Kletsky, 1979). Investigations on auditory nerve fibers and the basilar membrane in the same cochleae support the second filter (Evans and Wilson, 1975; see also Lamb and Chadwick, 2014). However, the validity of these comparisons was questioned (Narayan et al., 1998; Ruggero et al., 2000) because the measurements of Evans and Wilson (1975) were performed on ears with substantial hearing loss. Our measurements clearly show that there is a difference in frequency tuning between the OHC and IHC, as well as between the OHCs and the basilar membrane (Chen et al., 2011). This suggests that a second filter does indeed exist in the mammalian cochlea, although in a form different from envisioned earlier.
Footnotes
This work was supported by NIH Grants NIDCD R01 DC 000141, NIDCD R01 DC 010399 (A.L.N.), and NIDCD R01 DC 005983 (ALN-Core PI); the Swedish Research Council Grant K2011-63X-14061-11-3, the Research Council for Health, Working Life Welfare (2006-1526), the Torsten Söderberg Foundation, the Tysta Skolan Foundation, and the Hörselskadades Riksförbund (A.F.); National Natural Science Foundation of China No. 81271077, The National Basic Research Program of China 2014CB541706, and The Fourth Military Medical University Started fund for students returned (D.Z.); and NIH Grant NIDCD R01 DC 010201 (R.W.).
The authors declare no competing financial interests.
- Correspondence should be addressed to either of the following: Dr. Alfred L. Nuttall, Oregon Hearing Research Center, Oregon Health & Science University, 3181 SW Sam Jackson Park Road, Portland, OR 97239, nuttall{at}ohsu.edu; or Dr. Dingjun Zha. Xijing Hospital, Fourth Military Medical University, Department of Otolaryngology/Head & Neck Surgery, Xi'an 710032, People's Republic of China, zhadjun{at}fmmu.edu.cn