Cochlear function in Prestin knockout mice


Corresponding author M. A. Cheatham: 2-240 Frances Searle Building, 2240 Campus Drive, Northwestern University, Evanston, IL 60208-3550, USA. Email:


Gross-potential recordings in mice lacking the Prestin gene indicate that compound action potential (CAP) thresholds are shifted by ∼45 dB at 5 kHz and by ∼60 dB at 33 kHz. However, in order to conclude that outer hair cell (OHC) electromotility is associated with the cochlear amplifier, frequency selectivity must be evaluated and the integrity of the OHC's forward transducer ascertained. The present report demonstrates no frequency selectivity in CAP tuning curves recorded in homozygotes. In addition, CAP input–output functions indicate that responses in knockout mice approach those in controls at high levels where the amplifier has little influence. Although the cochlear microphonic in knockout mice remains ∼12 dB below that in wild-type mice even at the highest levels, this deficit is thought to reflect hair cell losses in mice lacking prestin. A change in OHC forward transduction is not implied because knockout mice display non-linear responses similar to those in controls. For example, homozygotes exhibit a bipolar summating potential (SP) with positive responses at high frequencies; negative responses at low frequencies. Measurement of intermodulation distortion also shows that the cubic difference tone, 2f1f2, is ∼20 dB down from the primaries in both homozygotes and their controls. Because OHCs are the sole generators of the negative SP and because 2f1f2 is also thought to originate in OHC transduction, these data support the idea that forward transduction is not degraded in OHCs lacking prestin. Finally, application of AM1-43, which initially enters hair cells through their transducer channels, produces fluorescence in wild-type and knockout mice indicating transducer channel activity in both inner and outer hair cells.

In the mammalian cochlea, receptor cells of the organ of Corti are fully differentiated into two populations: inner (IHC) and outer hair cells (OHC). While IHCs serve as sensory receptors, sending information to the auditory nerve, OHCs have a motor function. Experimental evidence suggests that OHCs change their length in a voltage-dependent manner (Brownell et al. 1985; Kachar et al. 1986; Santos-Sacchi & Dilger, 1988). This electromotility is thought to provide a means for improving mammalian cochlear sensitivity and frequency selectivity (Ashmore, 1987; Dallos, 1992; Dallos & Fakler, 2002). Discovery of the OHC motor protein prestin, as well as the in vitro demonstration that prestin provides the molecular basis for electromotility (Zheng et al. 2000), both support the idea that prestin is associated with the cochlear amplifier (Davis, 1983). It has also been shown that Prestin gene and protein expression, as well as electromotility, exhibit the same developmental gradient (Zheng et al. 2000; Belyantseva et al. 2000). The importance of prestin for human hearing has also been reported by Liu et al. (2003) who demonstrated that a non-syndromic deafness is linked to a prestin mutation.

In spite of this evidence, an in vivo model is required in order to investigate the association between electromotility and the cochlear amplifier. This is because the OHCs are included in a tightly coupled feedback loop, which comprises the basilar membrane–OHC–tectorial membrane complex. Fortunately, development of the Prestin knockout mouse makes it possible to determine whether loss of OHC motor function affects cochlear sensitivity and/or frequency selectivity. Previous results (Liberman et al. 2002) indicate that OHCs isolated from F2 generation mice lacking prestin do not exhibit electromotility. There is also an ∼50 dB threshold shift for both auditory brainstem responses (ABRs) and otoacoustic emissions (OAEs). However, it is not known if frequency selectivity is also compromised. This information is required before one can conclude that prestin is associated with the cochlear amplifier.

It is also important to determine the functional status of mechanoelectrical transduction, which involves the mechanical gating of transducer channels located at the tips of the stereocilia (Hudspeth & Corey, 1977). In other words, an assessment of transducer function is required in order to learn if the loss of function phenotype reflects changes in forward, i.e. mechanoelectrical transduction, as opposed to the assumed reverse or electromechanical process. If prestin is a key component of the cochlear amplifier, then the functional changes observed in knockout mice should result from loss of OHC motor function, i.e. reverse or electromechanical transduction, and not from a change in forward transduction.


In vivo physiology

Mice were generated from designated breeding pairs to produce litters of homozygous (−/−), heterozygous (+/−) and wild-type (+/+) pups. The physiological data in this report were obtained from a total of 32 mice: 10 +/+, 9 +/− and 13 −/−. Except where noted, the measurements were made using electrical recordings from the round window membrane in F4 generation mice (129/SvEv and C57BL/6 mixed background) between 30 and 58 days of age. Mice were anaesthetized with sodium pentobarbital (∼80 mg kg−1i.p.) and supplemental doses, using dilute anaesthetic, were administered as needed. During surgery and recording, the animal's rectal temperature was maintained at ∼38°C and the headholder was heated to prevent cooling of the cochlea (Shore & Nuttall, 1985; Ohlemiller & Siegel, 1992). Using the traditional ventrolateral approach, the bulla was exposed and a small window was made in the bullar bone using a miniature knife. A recording electrode, fashioned from Teflon-coated silver wire (insulated diameter 140 μm), was placed on the round window membrane. The reference electrode was placed subcutaneously on the opposite side of the head. At the end of all experiments, animals were killed with a lethal dose of anaesthetic (200 mg kg−1i.p.) and then decapitated.

Cochlear sensitivity was determined by measuring thresholds for the compound action potential (CAP) produced in response to 20 ms tone bursts, averaged 32 times and presented with an interstimulus interval of 135 ms. In this procedure, a tracking program (Taylor & Creelman, 1967; Gummer et al. 1987) determined the sound pressure level needed to generate an N1/P1 voltage of 10 μV at any given stimulus frequency. The method of sound calibration used to obtain these values is described elsewhere (Pearce et al. 2000).

In addition to threshold determinations, tone-on-tone masking curves (Dallos & Cheatham, 1976a) were also obtained to generate CAP tuning curves. Although forward masking curves are quantitatively more comparable to single unit and mechanical tuning curves, higher masker levels are required to produce criterion decreases in the N1/P1 voltage. This occurs because masker and probe are non-overlapping in time. Only by obtaining simultaneous tuning curves was it possible to assay frequency selectivity in homozygotes where CAP thresholds are shifted by ∼50 dB. In other words, the maximum sound pressure levels produced by our signal generation system were insufficient to obtain forward masking tuning curves in knockout mice. Hence, our use of simultaneous masking. In this paradigm, a 10 ms probe tone was presented 24 ms after the onset of the masker, which was 44 ms in duration including the 5 ms rise/fall times. The masker was presented in alternating phase to minimize the cochlear microphonic (CM) generated in response to the masker. Prior to collecting the tuning curve, the 12 kHz probe tone was presented alone and adjusted in level until an N1/P1 voltage of 45 μV was recorded. Masker frequency and level were then varied to produce a 3 dB decrease in the probe response, i.e. a decrease to 32 μV. Responses to 20 stimulus presentations were averaged to improve the signal-to-noise ratio. The tuning curve was then constructed by tracing together the masker frequency level/combinations associated with the criterion reductions. The computer generated tone bursts used to elicit the CAP had 1 ms rise/fall times for both CAP threshold and CAP tuning curve determinations.

When collecting CAPs to determine sensitivity and/or frequency selectivity, a custom-made programmable filter was used to bandpass filter the round window responses between 0.2 and 3 kHz. In other cases, input–output functions were collected without bandpass, but with lowpass filtering at 49.9 kHz to prevent aliasing. This wideband configuration allowed the cochlear microphonic and summating potentials (SP) to be recorded, as well as the CAP. Magnitudes were taken from fast Fourier transforms (FFTs) of averaged response waveforms. Waveform segments, obtained during the steady-state portion of the response, were windowed prior to transformation using a Hanning function.

Two-tone inputs were also used to study intermodulation distortion. In these experiments, the stimulus consisted of two tones partially overlapping in time (Cheatham & Dallos, 1997). These and all signals were transduced by a single modified Realistic super tweeter (RS no. 40-1310B). Analysis windows for FFTs were chosen to measure the primaries alone, as well as combination tones generated when both signals were presented together. The level of f1 (10 kHz) was 14 dB higher than that of f2 (12.195 kHz) to produce similar CM responses for each primary. To assure that responses shown in Fig. 7 were not due to distortion in the sound, measurements were made in a knockout mouse after euthanasia. All postmortem responses were less than 0.7 μV, even at the highest input levels.

Figure 7.

Intermodulation distortion observed in the CM
Input-output functions are provided for f1 alone (circles) and for 2f1f2 (triangles) in wild-type (A) and knockout (B) mice. The abscissa represents the level of f2, with f1 being 14 dB higher than f2. C, the level of 2f1f2 measured relative to that of f1 in decibels. In knockouts and their controls, the cubic difference tone is ∼20 dB below f1.

AM1-43 fluorescence

Viability of the hair cell transducer was also evaluated using AM1-43 (Biotium, Hayward CA, USA), the fixable analogue of FM1-43 (Nishikawa & Sasaki, 1996; Gale et al. 2001; Meyer et al. 2001; Meyers et al. 2003; Géléoc & Holt, 2003b; Si et al. 2003). This fluorescent styryl dye is quickly taken up by hair cells through their transducer channels, with subsequent endocytotic entry occurring on a slower time scale. On postnatal day 12 (P12), AM1-43 (3 mg (kg body weight)−1) was injected into a litter of six pups that were produced by mating heterozygotes. At ∼5 weeks of age, the mice were killed with an overdose of sodium pentobarbital and cardiac perfused, first with heparinized saline and then with 4% paraformaldehyde. Cochleae were removed and apical turns dissected and mounted for viewing on a Leica confocal microscope.

Dye uptake into cochlear hair cells was also studied in vitro. Following an overdose of sodium pentobarbital (200 mg kg−1i.p.), cochleae from +/+ and −/− mice at ∼7 weeks of age were harvested. Openings at the apex and base allowed exposure to 0.3 μm AM1-43 for ∼1–2 min. After washing for ∼2 min, the tissue was fixed in 4% paraformaldehyde for ∼30 min. Organ of Corti segments from the apical half of the cochlea were then removed and viewed on the Leica confocal microscope. The medium used in all cases was composed of the following: 5 mm KCl, 10 mm Hepes, 45 mm NaCl, 105 mm NaOH, 24 mm HCO3, 100 mm lactobionic acid. The osmolarity was 317 mosmol l−1 and the pH was 7.3 after bubbling with carbogen (95% O2, 5% CO2).

In order to verify that dye uptake was via the transducer channels and not via endocytosis (Meyer et al. 2001; Griesinger et al. 2002; Si et al. 2003), the transducer channel was blocked using dihydrostreptomycin (DHSM) (Kroese et al. 1989). In these experiments, cochleae were harvested from wild-type mice. The first cochlea dissected was pre-incubated with 1 mm DHSM for ∼5 min. After exposure of one ear to the blocker, both ears were then placed in 3.5 μm AM1-43 for ∼1.5 min. After washing, the cochleae were again fixed in 4% paraformaldehyde for ∼30 min. It should be emphasized that DHSM was added to the AM1-43 and to the wash medium, but only for the ear pre-incubated with the blocker. In contrast to all other results, the DHSM experiments were performed on F5 generation wild-type mice. All procedures were approved by the National Institutes of Health and by Northwestern University's Institutional Review Board.


In vivo physiology

Mean CAP thresholds for all three genotypes are shown in Fig. 1 for F4 generation mice. Although +/+ (circles) and +/− (triangles) mice show similar thresholds, those for −/− (squares) mice are shifted in sensitivity in a frequency-dependent manner, ranging from ∼35 dB at low frequencies to ∼60 dB at high frequencies. These threshold shifts are similar to those from Liberman et al. (2002) in F2 generation homozygous mice using ABRs and OAEs. The mean CAP thresholds in Fig. 1 were obtained from a total of six mice, except at 38 kHz where a threshold was measured in only two animals. In addition, at 45 kHz and above, the sound pressure levels were not sufficient to generate the criterion 10 μV N1/P1 voltage in homozygotes. These data imply that the number of functional hair cells in the basal ∼1.5 mm of the cochlea is probably small in animals older than 46 days of age, which is the mean age of homozygotes represented in Fig. 1. This estimate is based on a physiological frequency-place map (Mueller et al. 2004). This map indicates that the lower frequency limit at the apex of the mouse cochlea occurs at 4.8 kHz. This frequency limit suggests that the smaller threshold shifts measured below 5 kHz probably reflect responses by contributing single units on the tails of their tuning curves where the amplifier has less influence. In other words, frequency-dependent CAP threshold shifts do not imply that the gain of the cochlear amplifier is reduced at 2 vs. 5 kHz. The frequency-dependent gain change, attributable to the cochlear amplifier, is from ∼45 dB at 5 kHz to ∼60 dB at higher frequencies.

Figure 1.

Compound action potential (CAP) thresholds
Means and standard deviations are shown for CAP thresholds in F4 generation mice. Although wild-type and heterozygous mice demonstrate similar sensitivity, that in homozygotes is shifted in a frequency-dependent manner.

CAP input–output functions obtained at 6, 12 and 32 kHz are displayed in Fig. 2. In harmony with the CAP threshold data in Fig. 1, the low-level segments of the functions for knockout mice are shifted when compared to wild-type or heterozygous mice. However, the CAP magnitude differences between wild-type and knockout mice are smaller at high stimulus levels. It is emphasized that the control CAP plots possess two distinct segments: one associated with responses from fibres with best frequencies (BF) near stimulus frequency for the low-level response and the other with fibres whose BFs are above stimulus frequency for the high-level responses (Özdamar & Dallos, 1976). The high-level segments represent neural responses on the tails of the fibres' tuning curves, i.e. responses from cells located basal to the BF place for the stimulus frequency. Consequently, these high-level responses are less influenced by cochlear amplification. In homozygotes, the low-level segment is missing and response magnitudes are reduced for the high-level segment, especially at 12 and 32 kHz. The abnormality of the high-level segment is likely to be a consequence of a reduction in the number of contributing nerve fibres due to basal hair cell loss in −/− mice (Wu et al. 2004). This supposition is consistent with results at 6 kHz where homozygotes produce CAPs at high levels that are nearly commensurate with those in controls, while the high-level responses at 32 kHz are well below those in control mice. In fact, one can surmise from CAP thresholds that there are probably very few remaining IHCs with BFs above ∼38 kHz. This observation limits the number of fibres that can be recruited to provide the high-level segment of the CAP function.

Figure 2.

CAP input–output functions
Means and standard deviations are provided for CAP input–output functions at 6, 12 and 32 kHz in all three genotypes. Responses in homozygotes are reduced when compared to controls, be they wild-type or heterozygous mice.

Simultaneous masking (or tuning) curves are shown in Fig. 3 for the CAP produced in response to a 12 kHz probe tone. These curves indicate that +/+ and +/− mice show similar frequency selectivity. In fact, bandwidths at 10 dB above the tip, the Q10 values, are 5.26 for the average wild-type tuning curve and 5.34 for the average heterozygous curve. These values are only slightly better than those published by Shnerson & Pujol (1981) and very similar to Q10 values obtained from single units (Taberner & Liberman, 2003). The average probe level for both wild-type and heterozygous mice was 53 dB. These functions exhibit the familiar tip and tail segments, reminiscent of tuning curves at both the mechanical and single unit levels (Narayan et al. 1998). Because maskers around the 12 kHz probe are the most effective at reducing the probe response, criterion reductions are achieved at lower sound pressure levels in this region. In contrast to results for +/+ and +/− mice, there is no tuning in mice that lack the Prestin gene.

Figure 3.

CAP tuning curves
Simultaneous masking curves for a 12 kHz probe are provided for wild-type (+/+), heterozygous (+/−) and homozygous (−/−) mice. Means and standard deviations are plotted. The average probe level in both wild-type and heterozygous mice was 53 dB; that in homozygotes, 86 dB. These levels are indicated by the isolated symbols at 12 kHz. No frequency selectivity is seen in knockout mice.

In addition to neural responses, cochlear potentials were also recorded. The CM, which follows the time waveform of the stimulus, provides a gross reflection of the vectorially summed AC receptor currents produced by individual hair cell generators. CM input–output functions at 16 kHz are provided in Fig. 4, again for F4 generation mice. Data obtained at the same frequency by Liberman et al. (2002) for F2 generation mice are also appended and plotted with dotted lines and open symbols. Data shown in the right panel indicate that the average CM in F4 homozygotes is essentially the same as that for F2 generation mice. However, in contrast to the data from knockout mice, there are large differences between the present results and those of Liberman et al. (2002) for both wild-type (left panel) and heterozygous (centre panel) mice. As expected, our data indicate that the CM is decreased substantially

Figure 4.

Cochlear microphonic (CM) at 16 kHz
CM input–output functions at 16 kHz are shown for all three genotypes in F4 generation mice. Wild-type data shown on the left represent the averaged responses from six mice; those for heterozygotes in the centre and homozygotes on the right represent the averaged responses from five mice. Results from Liberman et al. (2002) at the same stimulus frequency are appended and plotted with dashed lines and open symbols. The CM in F4 generation wild-type and heterozygous mice is considerably larger than that for the F2 mice whose data are presented in the Liberman et al. (2002) report.

in knockout mice when compared to controls, be they wild-type or heterozygous mice. We also note that our +/+ and +/− data are essentially identical. The differences between +/+ and +/− CM responses in this study vs. those from Liberman et al. (2002) could be caused in part by variations in mixed strain backgrounds between F2 and F4 generations. In spite of this unlikely possibility, the present findings suggest that evaluation of the hair cell transducer is required because of the large reductions in the CM, which could be due either to loss of prestin or to some alteration in the chain of events associated with forward transduction. Although a reduction in the endocochlear potential (EP) would by necessity decrease receptor currents (Dallos, 1973; Patuzzi et al. 1989b), this is an unlikely possibility because prestin is not expressed in the stria vascularis or lateral wall (Judice et al. 2002). Therefore, the EP was not measured and is not thought to underlie the reduced CM measured in knockout mice.

In order to evaluate the status of the hair cell transducer, CM input–output functions were collected at 6 kHz. This stimulus, which is a relatively low frequency for the mouse, was chosen to minimize the effect of OHC loss at the base of the cochlea in mice lacking prestin (Wu et al. 2004). Figure 5 shows average CM responses, as well as standard deviations, for wild-type controls and knockout mice. Also included are individual functions for the nine mice used to generate the wild-type mean. These results are plotted with dotted lines. The dashed lines show data from the knockout with the largest CM responses. Although shifted in sensitivity, the CM recorded in mice with the largest responses (dashed lines) approaches wild-type magnitudes at high levels.

Figure 5.

CM at 6 kHz
Mean CM input–output functions are plotted for wild-type (n= 9) and homozygous (n= 5) mice at 6 kHz. Standard deviations are provided in only one direction to foster comparisons between knockouts with the largest responses and their wild-type controls. Functions without symbols represent results from the homozygote with the largest CM responses (dashed lines) and from the nine wild-type mice used to obtain the mean (dotted lines).

Cochlear non-linearities were also evaluated by measuring the SP, which is a DC baseline shift produced in response to an AC stimulus. In normal animals, this DC potential is both frequency and level dependent (Dallos et al. 1972), such that the SP recorded at the round window is positive at high frequencies and negative at low frequencies. The bipolar nature of the SP is illustrated in Fig. 6 where averaged response waveforms are provided at 32 (left column) and 6 kHz (right column). Examples from wild-type mice appear at the top and those from knockout mice are at the bottom. Both genotypes indicate a positive SP at 32 kHz and a negative SP at 6 kHz.

Figure 6.

Time waveforms of gross cochlear potentials
Averaged response waveforms are shown on the top row for wild-type mice and the bottom row for knockout mice. Data on the left were obtained for 32 kHz at 99 dB; those on the right for 6 kHz at 100 dB. Positive SPs are recorded at 32 kHz and negative SPs are recorded at 6 kHz in both genotypes.

In addition to the SP, intermodulation distortion was also studied in the CM by recording combination frequencies generated in response to a two-tone input (f1= 10 kHz; f2= 12.195 kHz). The most prominent distortion component, the cubic difference tone, is at 2f1f2 (7.805 kHz). Input–output functions for this component, as well as for f1 at 10 kHz are shown in Fig. 7. These data are plotted as a function of the level of f2, which was 14 dB lower than the level of f1. This adjustment resulted in about equal responses to each primary. Data in Fig. 7A and B are from a wild-type and knockout mice, respectively. Although responses are smaller in homozygotes due to the loss of gain, the relative level of 2f1f2 is similar in mice lacking prestin and their controls. In fact, data in Fig. 7C indicate that both wild-type and knockout mice produce a cubic difference tone which is ∼20 dB below the level of f1. This result is consistent with normal forward transducer function in OHCs (Patuzzi et al. 1989b; Jaramillo et al. 1993).

AM1-43 fluorescence

Hair cell transducer function was also assessed in +/+ and −/− mice using the fixable fluorescent styryl dye AM1-43, which is quickly taken up by hair cells through their transducer channels (Gale et al. 2001; Meyers et al. 2003; Géléoc & Holt, 2003b; Si et al. 2003). Figure 8 indicates that in vivo injection of the dye at P12 results in the labelling of both inner and outer hair cells in wild-type (A) and knockout (B) mice. In addition, both populations of cochlear hair cells were fluorescent when dissected apical segments were exposed to AM1-43 in vitro for ∼2 min as shown in Fig. 8C and D for wild-type and knockout mice, respectively. Finally, Fig. 8E and F show images obtained using DHSM, a known blocker of the hair cell transducer channel (Kroese et al. 1989). In spite of using a 10-fold increase in the concentration of AM1-43, fluorescence is drastically reduced in the presence of DHSM, as shown in Fig. 8F. Therefore, AM1-43 is probably entering the hair cells through transducer channels and not via endocytosis.

Figure 8.

AM1-43 fluorescence
A and B, fluorescence for wild-type and knockout mice, respectively. In these experiments the AM1-43 was injected in vivo at P12 and the mice were killed at P37 for the wild-type mouse and at P32 for the knockout mouse. For the in vitro experiments in C and D, the wild-type control (6 weeks 5 days) was exposed to 0.3 μm AM1-43 for 2 min, while the knockout (7 weeks 2 days) was exposed for 2 min 26 s. In no case did outer hair cells (OHCs) fail to fluoresce. E and F, the effect of pre-incubation in 1 mm dihydrostreptomycin (DHSM) for 5 min 32 s. For these control experiments, the concentration of AM1-43 was 3.5 μm AM1-43. One ear from this wild-type mouse (6 weeks 4 days) was exposed to AM1-43 alone for 1 min 27 s, while the other ear was exposed for 1 min 22 s following pre-incubation with DHSM. The organ of Corti was wet dissected and then mounted in an anti-fading medium and a coverslip was applied; this sometimes results in a folding of the tissue as in E where the inner hair cells are difficult to resolve. AF, calibration bar = 15 μm.


CAP thresholds in knockout mice show a frequency-dependent, sloping loss between 35 and 60 dB. This loss of sensitivity in the CAP recorded at the round window is similar to that obtained by Liberman et al. (2002) in F2 generation mice using ABRs and OAEs. Although frequency selectivity was not measured in the Liberman et al. (2002) study, the loss of tuning in homozygotes is consistent with the idea that prestin is associated with the cochlear amplifier, i.e. those mechanisms that establish sensitivity and frequency selectivity in the peripheral auditory system. Our results in F4 generation mice also show essentially identical CAP responses in +/+ and +/− mice. This result is in conflict with the Liberman et al. (2002) study where a 6 dB threshold deficit was seen in heterozygotes. Taken together, the similarity of CAP thresholds and CAP tuning curves in F4 generation +/+ and +/− mice imply that one copy of the Prestin gene is sufficient for normal cochlear sensitivity and frequency selectivity.

In contrast to CAPs, which are summed neural responses, the CM reflects hair cell receptor currents. Although IHCs probably make some contribution to the CM, evidence suggests that this response is dominated by OHC receptor currents (Dallos & Wang, 1974; Dallos & Cheatham, 1976b; Patuzzi et al. 1989a). In addition, the CM recorded from the round window is a population response with contributing hair cell groups changing with stimulus frequency and level. It is therefore important to exercise caution when interpreting CM results. For example, CM data in normal animals suggests that low-level responses are more dependent on hair cell populations near the BF of the stimulus, whereas high-level responses are strongly dominated by cells near the round-window electrode even at low frequencies (Patuzzi et al. 1989a). This level of dependence is probably minimized in knockout mice where hair cell death begins at ∼3 weeks of age, and by 6 weeks of age IHC and OHC loss is observed over the basal 28% of the cochlea (Wu et al. 2004). To make matters worse, the reported profile of hair cell loss probably underestimates functional hair cell loss. Just because a cell is present does not mean that it is working properly (Steel & Bock, 1983).

Data in Fig. 4 indicate that CM responses in the present study for +/+ and +/− mice are commensurate with those obtained by Legan et al. (2000) who also used the same mixed 129SvEv/C57BL6 background to develop their α-tectorin knockout mice. Both studies show CM responses in wild-type controls that are much larger that those of Liberman et al. (2002). Independent of whether these differences are due to variations between F2 and F4 generations or whether the Liberman et al. (2002) data are abnormal, it is incorrect to assume that the similarity of CM responses in all three genotypes implies normal OHC transducer function (Liberman et al. 2002). In other words, one should not expect identical CM functions from ears that possess normal amplification and from those with impaired amplification. No matter how the amplification-providing feedback loop is opened, the system gain will be reduced and with it the input to the forward transducer. As a result, the transducer current, and thus the CM, will decrease. These reductions in the CM observed in F4 homozygotes could reflect a deficit in either forward or reverse transduction. Therefore, the status of the forward transducer was evaluated.

Results obtained in homozygotes using AM1-43 suggest that both inner and outer hair cells have open transducer channels. Although it is possible that this dye is taken up into hair cells via endocytosis (Nishikawa & Sasaki, 1996; Meyer et al. 2001; Griesinger et al. 2002), the time course is slower than that through functional transducer channels. While the latter occurs in seconds, endocytosis takes at least 2 min dye exposure in isolated sense organs (Si et al. 2003). In our case, however, the entire cochlea was incubated in AM1-43. While isolated organ segments were dissected for examination, this took place only after fixation in order to prevent damage to the transducer channels. Because of the longer diffusion time for the dye to enter scala media in intact preparations, one should probably not use estimates of AM1-43 entry via endocytosis that are derived from studies on organ segments. Furthermore, we have observed OHC fluorescence in knockout mice for cochlear incubation times as short as 68 s. Hence, the different time course, as well as the squelching of fluorescence in the presence of DHSM, suggests that the route of entry by AM1-43 for short exposures is via transducer channels. However, it is conceivable that transducer channels in homozygotes are fully open, allowing dye to enter, but not fully functional. If this were the case, one would not expect signal-dependent modulation of the channel. The presence of a CM argues against this possibility.

The physiological data also support the idea that forward transduction is normal in OHCs lacking prestin. CAP input–output functions at 6 kHz show near-normal magnitudes at high levels where minimal amplification is expected. However, the CM in the knockout mouse with the largest responses is still ∼12 dB down when compared to the wild-type mean. Reduction in the number of contributing hair cells in knockout mice could underlie this observation that homozygotes produce smaller maximal CM responses than do controls at high levels. In the presence of hair cell loss, the high-level CM in knockout mice will be diminished reflecting the reduction in the number of contributing hair cells, their increasing distance from the recording electrode and the complications associated with phase cancellation. For example, contributions from basal-turn OHCs in wild-type mice at 6 kHz are expected to grow linearly with level up to the point where the transducers of contributing hair cells begin to saturate. These contributions, obtained well below the BF of contributing OHCs, would not suffer reductions due to phase cancellation. As a result, the CM in wild-type mice at low frequencies is minimally influenced by the cochlear amplifier because the hair cells dominating the CM are responding linearly on the tails of their tuning curves. In contrast, the CM recorded from homozygotes is probably dominated by OHCs with BFs nearer to the stimulus frequency because of the incidental hair cell death occurring in these mice. As a result, increasing the level of a 6 kHz tone does not recruit as many OHCs from the base of the cochlea in knockout mice. Moreover, the summed responses from hair cells that do contribute are probably subject to phase cancellation. This argument suggests that the reduced CM is probably not due to a modification in forward transduction. This conclusion is supported by the existence of tip links between OHC stereocilia in knockout mice (Wu et al. 2004).

The presence of a bipolar SP is also consistent with normal OHC transducer function. Evidence from intracellular recordings suggests that OHCs respond to inputs below BF with negative, hyperpolarizing dc receptor potentials (Dallos et al. 1982; Russell & Sellick, 1983; Dallos, 1985, 1986; Cody & Russell, 1987), whereas IHCs never produce this polarity. These intracellular responses are reflected at the round window as negative SPs. It is also known that no negative SP is recorded in animals lacking OHCs in the basal half of the guinea pig cochlea due to kanamycin intoxication. Therefore, Dallos & Wang (1974) concluded that OHCs alone generate this SP polarity. This knowledge implies that SP responses, obtained at what are low frequencies in the mouse, should be asymmetrical in the negative direction when OHCs with viable transducers, basal to the BF location of the stimulus, contribute to the responses.

The presence of a cubic difference tone at ∼20 dB below the primaries in both wild-type and knockout mice is also consistent with normal OHC transducer function. Measurement of intermodulation distortion in cochlear mechanics (Robles et al. 1997) or in neural responses (Dallos et al. 1980) suggests that the most prominent distortion product, 2f1f2, is probably generated by OHCs. Because reverse transduction cannot produce this component in mice lacking prestin, present data support the idea that the cubic difference tone is associated with OHC mechanoelectrical transduction (Patuzzi et al. 1989b; Jaramillo et al. 1993), i.e. cochlear distortion originates in hair cell non-linearities (Goldstein, 1967; Dallos et al. 1969). The fact that intermodulation distortion in both knockout and wild-type mice is ∼20 dB down from the primaries also indicates that shortening of OHCs does not alter ciliary mechanics or the residual mechanical response (Santos-Sacchi, 2003).

Although results shown here are consistent with the idea that prestin and OHC motility are required for normal cochlear function, one cannot conclude that prestin alone is the cochlear amplifier. While we show that OHC motility is necessary for amplification, it is yet to be determined if it is sufficient or whether its interaction with other processes is required. This caveat is needed because removal of prestin makes it difficult to determine whether other elements in the feedback path contribute to the sensitivity and frequency selectivity of the peripheral auditory system (Géléoc & Holt, 2003a; Fettiplace & Ricci, 2003). This difficulty is exacerbated by the hair cell loss in the Prestin-null mouse, which complicates comparisons with controls.



Work supported by NIH grants DC00089 to P. Dallos and DC04761 and CA21765 to J. Zuo. We appreciate contributions by R. Edge and J. Zheng to the AM1-43 experiments.