Lack of Bdnf and TrkB signalling in the postnatal cochlea leads to a spatial reshaping of innervation along the tonotopic axis and hearing loss

Signalling of the neurotrophin gene family through their high-affinity receptors, termed Trks, is responsible for neuronal survival and/or target innervation in the peripheral nervous system(Bibel and Barde, 2000). In the inner ear, the neurotrophins brain-derived neurotrophic factor (Bdnf) and neurotrophin 3 (Nt3; now known as Ntf3), and their corresponding receptors TrkB (Ntrk2) and TrkC (Ntrk3), are responsible for the survival of cochlear and vestibular neurones (Rubel and Fritzsch, 2002). In mammals, the dependence of these neurones on Bdnf or Nt3 signalling through their receptors has been demonstrated by the study of mouse mutants in which the corresponding genes have been inactivated by homologous recombination (Ernfors et al., 1995; Schimmang et al.,1995; Bianchi et al.,1996; Fritzsch et al.,1997a). In these studies, it was shown that Bdnf and TrkB mutants lose the majority of vestibular neurones, whereas Nt3- and TrkC-deficient mice revealed a severe reduction of neurones in the cochlea.

Cochlea type I sensory neurones, comprising approximately 90% of the cochlear ganglion, provide afferent innervation to the inner hair cells(IHCs). The remaining part of the ganglion consists of small-sized type II sensory neurones, which innervate outer hair cells (OHCs). Initial studies of knockout animals for Bdnf or TrkB showed a severe reduction of type II sensory neurones, whereas Nt3 and TrkC mouse mutants lost the majority of type I neurones (Ernfors et al.,1995; Schimmang et al.,1995). From these results, and from the study of the afferent innervation patterns using neurofilament antibodies, it was suggested that survival of specific neurone types and their targets was under the control of particular neurotrophins and their receptors.

More detailed studies using DiI as a tracer revealed that innervation in the cochlear sensory epithelium shows a more complex pattern in neurotrophin and Trk receptor mutants (Fritzsch et al.,1997a). Bdnf and TrkB mutants showed the most pronounced defects in the apex, whereas the lack of innervation of TrkC and Nt3 mutants was most severe at the cochlear base (Bianchi et al., 1996; Fritzsch et al.,1997b; Fritzsch et al.,1998). Moreover, these studies demonstrated that both neurotrophins and their receptors are responsible for the innervation of OHCs(Rubel and Fritzsch,2002).

Recent data suggest that the spatial gradients of neuronal loss observed in neurotrophin and Trk receptor mutants is attributable to the spatial-temporal gradients of neurotrophin expression during embryonic development(Fariñas et al., 2001). Based on this study, Bdnf is localised predominantly in the apex of the cochlea, whereas Nt3 is localised in the base. Additionally, Bdnf expression has been shown to correlate with the rearrangement of fibres during postnatal development of the cochlea, and this reorganisation is disturbed in Bdnf knockout mice (Wiechers et al.,1999). However, because Bdnf or TrkB receptor mouse mutants do not survive past the first postnatal weeks(Klein et al., 1993; Ernfors et al., 1994; Jones et al., 1994; Korte et al., 1995), the consequences of a severe loss of TrkB signalling in the mature cochlea could not be analysed.

To study the importance of particular signalling pathways used by Trk receptors for neuronal survival and target innervation, specific point mutations have recently been introduced in the docking site for the Shc adapter protein on the TrkB and TrkC receptor(Minichiello et al., 1998; Postigo et al., 2002). Although in both TrkB^shc/shc- and TrkC^shc/shc-mutant mice the RAS/MAPK and phosphoinositide-3-kinase pathways are similarly affected, neuronal survival in the peripheral nervous system, including the inner ear, is only modestly reduced. However, whereas target innervation in TrkC^shc/shc mice is maintained, in TrkB^shc/shc mice several cranial sensory populations, including vestibular neurones, lose innervation to their peripheral targets (Postigo et al.,2002).

In the present study, we used these animals to create adult mice with a severe lack of TrkB receptor signalling, by crossing them to TrkB^+/- mutants with a targeted deletion within the tyrosine kinase domain of the receptor(Klein et al., 1993). Similar to TrkB^-/- and Bdnf^-/- mutants(Bianchi et al., 1996; Fritzsch et al., 1998), mice carrying the mutant shc allele and the mutated tyrosine kinase domain of the TrkB receptor (TrkB^shc/-) show a region-specific absence of outer hair cell innervation in the apical part of the cochlea during early postnatal development.

Further analysis of Bdnf^-/- and TrkB^shc/- animals revealed that the observed defect is reversed during maturation of the cochlea, thus leading to normal innervation of OHCs in the apex but absence of afferent and retardation of efferent nerve fibres in the basal turn. Although the cellular integrity of OHCs appeared unaffected in the mutant animals in all cochlear turns, hearing thresholds of TrkB^shc/- mice were significantly reduced compared with control mice. The present data are consistent with recent results that suggest that neurotrophins in the mature hearing organ may have a distribution opposite to that found during development(Adamson et al., 2002).

Mice heterozygous for a deletion of the Bdnf gene were kindly provided by Patrick Carroll and Hans Thoenen (MPI Martinsried, Munich,Germany). The offspring of these mice were genotyped by PCR according to Korte et al. (Korte et al., 1995). Heterozygous TrkB^+/--mutant mice were obtained from Jackson Laboratory and their offspring were genotyped by PCR according to Schimmang et al. (Schimmang et al.,1995).

Mice carrying the mutant shc allele of the TrkB receptor(TrkB^shc/shc) have been described in Minichiello et al., 1998. Mutant

TrkB^shc/- mice were created by crossing TrkB^+/- mice with TrkB^shc/shcmutants.

Cochleae were isolated and dissected as previously described(Knipper et al., 2000). Briefly, cochleae were fixed by immersion in 2% paraformaldehyde, 125 mM sucrose in 100 mM phosphate buffered saline (PBS; pH 7.4), for 2 hours,followed by overnight incubation in 25% sucrose, 1 mM protease inhibitor(Pefabloc, Roche) in PBS (pH 7.4). Cochleae of animals older than postnatal day 10 (P10) were decalcified after fixation for 15 minutes to 2 hours in Rapid bone decalcifier (#904687, Eurobio, Fisher-Scientific, 61130 Nidderau,Germany). After overnight incubation, cochleae were embedded in O.C.T. compound (Miles Laboratories, Elkhart, Ind., USA). Tissues were then cryosectioned at 10 μm thickness for in situ hybridisation and immunohistochemistry, mounted on SuperFrost*/plus microscope slides, dried for 1 hour and stored at -20°C before use.

For immunohistochemistry, cochlear sections from different postnatal stages and adult mice or rats were thawed, permeabilised with 0.1% Triton X-100 for 3 minutes at room temperature, preblocked with 1% bovine serum albumin in PBS and incubated overnight at 4°C with primary antibodies. The following primary antibodies were used: anti-NF200 (Sigma N4142); anti-synaptophysin(The Binding Site #PH510) anti-peripherin (Chemicon #AB1530), anti-prestin(Weber et al., 2002) and anti-potassium channel SK2 (Sigma, #P0483). Expression of full-length TrkB in cochlear neurones was confirmed by antibodies directed against aminoacids 794-808 of the TrkB receptor (#794, Santa Cruz Biotechnology). Primary antibodies were visualised with either Cy3-conjugated goat anti-rabbit Ig(0.35 μg/ml; Jackson Immuno Research Laboratories, PA, USA) or with Alexa-Green conjugated goat-anti-mouse antibodies (1:1500, Molecular Probes,Leiden, The Netherlands). Sections were rinsed, mounted with Vectashield(Vector Laboratories, Burlingame, CA, USA) and viewed using an Olympus AX70 microscope equipped with epifluorescence illumination.

Genomic DNA from rat liver was isolated by Easy-DNA kit from Invitrogen,following the protocol provided by the manufacturer. Polymerase chain reaction was used to amplify exon 4 of the Bdnf gene(Timmusk et al., 1993). For the exon 4-specific probe, a sense primer (5′-cca atc gaa gct caa ccg aa-3′) and an antisense primer (5′-tca ggg tcc aca caa agc tc-3′) corresponding respectively to nucleotide position 1732-1751 and 2059-2078 from genomic fragment B were used. During the PCR reaction, genomic DNA was first denatured for 4 minutes at 94°C, followed by 30 cycles consisting of 1 minute at 94°C, 1 minute at 55°C and 1 minute at 72°C. The extension reaction was carried out at 72°C for 10 minutes. The amplified fragment corresponding to the expected length of 347 nucleotides was extracted and sequenced. Clones of full-length rat Nt3 and full-length TrkB were supplied by Regeneron Pharmaceuticals. In situ hybridisation was performed as described(Wiechers et al., 1999). Sections were mounted with Moviol (Sigma) and viewed using an Olympus AX70 microscope.

Auditory evoked brainstem responses (ABR) were recorded in anaesthetised mice. Anaesthesia was achieved by the intraperitoneal injection of 65 mg/kg ketamin hydrochloride (Ketamin 50 Curamed, CuraMED Pharma, Germany), 10.5 mg/kg xylazin hydrochloride (Rompun 290, Bayer Leverkusen, Germany) and 0.33 mg/kg atropine sulphate (Atropinsulfat, Braun, Germany). Recordings were carried out in a sound-proof chamber (IAC, Type 400-A). A Multi IO Card(National Instruments MIO 16 E1) was used for the generation of stimuli and recording of evoked potentials. Tone pips of 3 ms duration (1 ms rise and fall time, cosine-shaped) were presented at a rate of 60/second with alternating phase. Clicks with duration of 100 μs of alternating phase were also used. Sound pressure level was adjusted with a custom-made attenuator and tone pips were amplified with a custom-made amplifier. Stimuli were delivered to the ear in a calibrated (Bruel&Kjaer 2610, 4191) closed system by a Beyer DT911 loudspeaker. In the case of tone pips, the sound pressure was calibrated in situ at all frequencies recorded prior to each measurement taken. As this is not possible for clicks, sound pressure was calibrated off-line once prior to the experiments. In this case, 0 dB attenuation corresponds to 112 dB SPL root means square (rms). To record bio-electrical potentials, subdermal silver wire electrodes were inserted at the vertex (reference), ventro-lateral to the measured ear (active) and at the back of the animal (ground). After amplification (100 dB) and bandpass filtering (0.3-5.0 kHz), electrical signals were averaged (128 repetitions). ABR responses were recorded for frequencies between 2.8 kHz and 45.2 kHz at a resolution of 2 steps per octave. At each frequency ABRs were collected from 20 dB to 100 dB SPL in 5 dB steps. The software averager included an artefact rejection code (all waveforms with a peak to peak voltage exceeding a defined voltage were rejected) to eliminate the ECG and muscle activity. Thresholds were defined as the sound pressure level where a stimulus correlated response was clearly identified in the recorded signal. If no response was obtained we used 100 dB SPL as the threshold value volume for further calculations.

Anaesthesia was performed as described for the ABR measurements. For DPOAE measurements acoustic stimulations were performed by a PC-based system with an A/D converter board using LabWindows software, which also performed fast Fourier transformation (FFT) for data analysis. Pure tone sound stimuli were generated and delivered on two separate channels using dynamic speakers (Beyer DT48). Sound levels were measured with a probe microphone (Bruel&Kjaer 4135) and a measuring amplifier (Bruel&Kjaer 2610). The speakers and microphone were coupled together, with the closed field into the ear canal of the animal, and the sound system calibrated with white noise (50 Hz-20 kHz). DPOAEs were measured as cubic distortion products at the frequency 2f1-f2. For each frequency f1 (7, 12 or 18 kHz), the individual best ratio (BR) f2/f1 was determined using an intermediate sound level (60 dB SPL or higher, if necessary), varying the ratio from 1.06 to 1.44 in steps of 0.02. This ratio was used for the following input/output measurement. The sound level of f1 was kept at a constant 10 dB louder than the f2 level. Sound was initially presented at the lowest level and increased in steps of 5 dB. Recording windows were averaged in order to decrease the background noise to at least 10 dB below the level of emission, but not over more than 1000 repetitions.

As previously reported, we observed defects of the afferent and efferent innervation in all vestibular sensory epithelia in adult TrkB^shc/shc mice(Fig. 1A,B)(Postigo et al., 2002). By contrast, innervation of the cochlear sensory epithelium was found to be normal in these mutants (data not shown). As TrkB^-/- mouse mutants, which carry a defective tyrosine kinase, die shortly after the first or second postnatal week (Klein et al.,1993), an analysis of the lack of TrkB function during further maturation of the inner ear in vivo has not been possible. To monitor the consequences of a severe reduction of TrkB signalling during later development, we crossed TrkB^shc/shc mice with viable heterozygous TrkB^+/- mice, to produce TrkB^shc/- mice. TrkB^shc/- mice were found to have a normal lifespan, thus permitting a further analysis of the inner ear during later postnatal development and adulthood.

Examination of the vestibular system of adult TrkB^shc/- mice revealed a similar reduction of afferent innervation of target sensory epithelia to that observed in TrkB^shc/shc animals. In TrkB^shc/+ control animals bundles of neurofilament-positive afferent fibres project to the vestibular hair cells as shown for the sensory cells of the sacculus, reaching the target hair cells at P0 and forming a calyx-like structure around type I hair cells in adult specimens (Fig. 1A). In both P0 and adult TrkB^shc/shc and TrkB^shc/- mice, the number of neurofilament-positive fibres contacting the target hair cells in the sacculus and utriculus is significantly reduced, and calyx-like structures are not formed in adults (Fig. 1A,B). Similar to TrkB^shc/shc mice, the ampullary cristae show the most severe innervation defects in the TrkB^shc/- mutants (data not shown)(Postigo et al., 2002). Moreover, we observed a lack of nerve fibres in the cochlear sensory epithelium at the level of the OHCs in TrkB^shc/-mice: the medial turns of early postnatal cochleae in TrkB^shc/- mice are shown for comparison with TrkB^shc/+ controls(Fig. 1C). In agreement with earlier findings, a similar phenotype could be observed in Bdnf^-/- and TrkB^-/- mouse mutants(Fig. 1C)(Bianchi et al., 1996; Fritzsch et al., 1998; Wiechers et al., 1999). Thus,the TrkB^shc/- mutant offers a useful model to study the consequences of defective TrkB signalling in the cochlea during further maturation and in adulthood. We therefore decided to perform a detailed analysis of the innervation patterns and the cochlear morphology TrkB^shc/- mice during these timepoints and to compare them with the defects of TrkB^-/- and Bdnf^-/- mutants.

Microscopical examination on serial sections revealed no apparent morphological changes in the organ of Corti of TrkB^shc/- mice compared with control littermates(Fig. 2A). To further confirm the intactness of OHCs, we used antibodies directed against the outer hair cell motor protein prestin (Weber et al.,2002). Similar to TrkB^shc/+ control animals, the characteristic staining pattern of prestin localised in the lateral membrane of the outer hair cell was observed in all cochlear turns of adult TrkB^shc/- mutants(Fig. 2B). Previous studies had reported a specific loss of type II sensory neurones, innervating OHCs in TrkB and Bdnf mutants (Ernfors et al.,1995; Schimmang et al.,1995). To monitor the presence of type II sensory neurones in TrkB^shc/- mice we used antibodies directed against peripherin, a 57 kDa type III intermediate filament that has been defined as a specific marker for these neurones(Hafidi, 1998). On serial sections of P19 and adult control mice distinct small-sized peripherin-positive neurones were detected(Fig. 3A,C). No peripherin-positive neurones could be detected in the cochlear ganglion along the entire length of the hearing organ in adult TrkB^shc/- mutants(Fig. 3B). This type of neurone was also not detected in Bdnf^-/- mutants at P19(Fig. 3D). Therefore, the absence of type II sensory neurones in TrkB^shc/-mutants confirms the dependence of these neurones on TrkB signalling and underlines the validity of TrkB^shc/- mice as a model for a severe loss of TrkB function in the cochlea.

TrkB^-/- and Bdnf^-/- mutants showed a lack of afferent type II fibres to OHCs at birth and after the first postnatal week (Fig. 1)(Bianchi et al., 1996; Fritzsch et al., 1998; Wiechers et al., 1999). Furthermore, a region-specific absence of nerve fibres in the apex could be observed. In agreement with these data, we were unable to detect neurofilament type II fibres in the medial and apical turn of cochleae of P0-P8 of TrkB^-/- and Bdnf^-/- mice(Fig. 4). In comparison, a dense pattern of fibres was observed in these cochlear turns at the base of OHCs in TrkB^shc/+(Fig. 4A) and TrkB^+/- mice (data not shown). Similar to TrkB^-/- and Bdnf^-/- mice,neurofilament-positive projections towards OHCs were not detected in the medial and apical turn in TrkB^shc/- mutants,whereas fibres of this phenotype were found opposite of OHCs in the basal cochlear turn (Fig. 4A). No innervation defects were observed at the level of IHCs in TrkB^shc/- mutants.

To analyse the further development of the innervation pattern in the hearing organ with a severe lack of TrkB signalling, we examined the cochlear sensory epithelium of P19 Bdnf^-/- mutants and adult TrkB^shc/- mice. Surprisingly, at these timepoints, Bdnf^-/- and TrkB^shc/- mice revealed the opposite innervation pattern to the one observed during early postnatal development(Fig. 4B). Apical turns of P19 Bdnf^-/- and adult TrkB^shc/-mice showed a normal pattern of afferent fibres below OHCs, similar to TrkB^shc/+, TrkB^shc/shc and TrkB^+/- mice(Fig. 4B and data not shown). On the contrary, in basal turns, no contacts of fibres with OHCs could be detected in TrkB^shc/- and Bdnf^-/- animals, whereas TrkB^shc/+, TrkB^shc/shc and TrkB^+/- mice showed a normal innervation pattern (Fig. 4B and data not shown). These results demonstrate that the initial innervation pattern observed during early postnatal development is reversed during maturation of the cochlea in mutants deficient for Bdnf or TrkB.

Upon analysis of the efferent innervation pattern using synaptophysin as a marker protein (Knipper et al.,2000), we observed that the loss of afferent type II fibres projecting to OHCs in basal cochlear turns of TrkB^shc/- mice was associated with a severe reduction in the number and size of synaptophysin-positive boutons below these cells (Fig. 5). To further define this defect on the postsynaptic site of the cell membrane, we studied the localisation of the apamin-sensitive, small-conductance Ca²⁺-activated potassium channel SK2, which has been localised in OHCs, in a localisation opposite to the efferents(Fig. 5)(Oliver et al., 2000). In the basal turn of adult TrkB^shc/+ animals,synaptophysin-positive boutons were detected below each hair cell, opposite SK2 channels and restricted to the area of efferent contact. In adult TrkB^shc/- mice, we observed only a few small synaptophysin-positive boutons, with either a very weak or no SK2-positive staining at the outer hair cell contact zone. These findings were also confirmed in the basal turn of control Bdnf^+/+ and Bdnf^-/- mice at P19. These results indicate that the loss of afferents below OHCs in the basal cochlear turns of Bdnf- and TrkB-deficient mice may have consequences on the size of the efferent contact zone on the hair cell, and that the reduced size of the presynaptic efferent contact zone may subsequently result in a reduction or absence of the postsynaptic field, characterised by the expression of SK2 ion channels.

The switch of the dependence on Bdnf observed in afferent fibres along the tonotopic axis during postnatal cochlear development, and the resulting reversal of the innervation pattern, may be explained by the differential expression of Bdnf or its receptor, TrkB, along this axis. Additionally, the expression of Nt3, which is also able to signal through TrkB(Davies et al., 1995; Kaplan and Miller, 1997), may influence the reorganisation of afferent fibres. Although it has been shown that Nt3 is expressed by adult inner hair cells (see also Fig. 6A), neither Bdnf nor TrkB are expressed in the adult sensory epithelium(Ylikoski et al., 1993; Wheeler et al., 1994; Fritzsch et al., 1999). However, mammalian spiral ganglion neurones have been shown to express Nt3, Bdnf and TrkB either postnatally or in adulthood(Mou et al., 1997; Wiechers et al., 1999; Oestreicher et al., 2000; Zha et al., 2001). Transcripts of Nt3, Bdnf and TrkB were also found to be expressed in adult cochlear neurones of rats (Fig. 6A). We next studied the distribution of these transcripts along the tonotopic axis. For Bdnf and TrkB, we observed a gradient of expression increasing from the apical and medial turns towards the midbasal and basal parts of the cochlea(Fig. 6B). By contrast, Nt3 mRNA showed an opposite gradient of expression, with more transcripts present in spiral ganglia neurones of the apical and medial turns compared with the midbasal and basal parts(Fig. 6B). This distribution was first observed around the end of the second postnatal week, and the described expression patterns were also confirmed in mice and gerbils (data not shown). Consequently, the lack of innervation in the more basal part of TrkB and Bdnf mutants during maturation of the cochlea most likely reflects the dependence of afferent fibres on increased levels of Bdnf and TrkB expression.

To monitor the importance of the observed mutant phenotype in TrkB^shc/- mice for function of the cochlea, we performed hearing tests. Frequency-dependent auditory brain stem responses were determined in adult TrkB^shc/- mutant and TrkB^shc/+ control animals. Mutant mice showed significantly elevated thresholds at all frequencies tested(Fig. 7A). Moreover, distortion product otoacoustic emissions above the background level were not detectable in TrkB^shc/- animals(Fig. 7B). These results thus suggest that the observed lack of outer hair cell innervation in TrkB^shc/- mice leads to a qualitative loss of hearing performance.

Until recently, the consequences of a lack of Bdnf or TrkB signalling for the mature inner ear could not be analysed because mouse mutants for these genes die within the first postnatal weeks(Klein et al., 1993; Ernfors et al., 1994; Jones et al., 1994; Korte et al., 1995). The creation of mouse mutants that lack specific parts of the signalling cascade controlled by the TrkB and TrkC receptor results in mice with a normal lifespan, and thus has allowed us to study the phenotypes of these mice(Minichiello et al., 1998; Postigo et al., 2002). Specifically TrkB^shc/shc mice, which lack the binding site of the Shc adapter protein of TrkB, appear to offer a useful model to investigate a reduction of TrkB signalling during further maturation of the inner ear.

In the present study, and a previous study(Postigo et al., 2002), we observed severe afferent and efferent innervation defects in the vestibular sensory epithelia of TrkB^shc/shc mice, which are similar to the ones found in TrkB^-/- mice that lack the entire tyrosine kinase domain (Schimmang et al., 1995; Fritzsch et al.,1998). By contrast, no defects could be observed in cochlear neurones of TrkB^shc/shc mutants. Cochlear neurones may therefore depend on other signalling domains of the TrkB receptor(Minichiello et al., 2002), or on additional growth factors (Fritzsch et al., 1999). By crossing TrkB^shc/shcwith TrkB^+/- mice we have created TrkB^shc/- mutants with a further reduction in TrkB signalling but a normal life span. In the vestibular system of TrkB^shc/- mice, we observed similar innervation defects to the ones observed in TrkB^shc/shcmutants (Postigo et al.,2002). Interestingly however, in contrast to TrkB^shc/shc mutants, TrkB^shc/- mice also show a loss of afferent type II fibres to the OHCs of the adult cochlea, which had so far only been reported in TrkB^-/- mutants during early postnatal development (Schimmang et al.,1995; Fritzsch et al.,1998). TrkB^shc/- mice thus offered an interesting model to study the consequences of severely reduced TrkB signalling in the adult cochlea. The validity of this model was established by the parallel analysis of postnatal Bdnf-mutant mice, which show an expanded postnatal lifespan compared with TrkB^-/- mutants(Klein et al., 1993; Ernfors et al., 1994; Jones et al., 1994; Korte et al., 1995).

Similar to results obtained by neuronal cell counts in TrkB and Bdnf mutants (Schimmang et al.,1995; Ernfors et al.,1995), we observed a loss of type II cochlear sensory neurones using a cell-type specific marker for these neurones(Hafidi, 1998) in TrkB^shc/- mice. This result thus further confirms that the primary defect in Bdnf and TrkB mutants is a loss of type II sensory neurones, and a defect in the innervation of their targets, the OHCs (see below). However, the morphology and structural integrity of OHCs, as demonstrated by staining with antibodies against the OHC motor protein prestin, appears unaffected in TrkB^shc/- mutants. Likewise, only minor or no structural changes of cochlear morphology have been reported in previous analyses of neurotrophin or Trk receptor mutants(Schimmang et al., 1995; Fritzsch et al., 1997b).

As the major defect in TrkB and Bdnf mutants appears restricted to the innervation of OHCs, we have analysed in detail the spatiotemporal pattern of innervation to these cells during postnatal development along the tonotopic axis of the cochlea. As previously reported, we observed a lack of afferent fibres to OHCs in the medial and apical part of the cochlea in Bdnf and TrkB mutants during the first postnatal week(Bianchi et al., 1996; Fritzsch et al., 1998; Wiechers et al., 1999). A similar defect was also observed in TrkB^shc/-mice, once more confirming the validity of these mice as a model for a severe reduction of TrkB signalling in the postnatal cochlea.

The early postnatal phenotype of TrkB and Bdnf mutants is most likely explained by an apical-to-basal gradient of Bdnf expression in the cochlear sensory epithelia during embryonic development(Fariñas et al., 2001). Strikingly, we observed a mirror-image of this phenotype during subsequent postnatal maturation in the adult cochlea of Bdnf^-/- and TrkB^shc/- mutants, respectively. Towards the end of the third postnatal week, Bdnf mutants showed a complete absence of basal afferent innervation, whereas the apical part of the cochlea showed a normal pattern of afferent fibres on OHCs. Likewise, an identical pattern of innervation was observed in TrkB^shc/- adult mutants. How can the reversal of the phenotype observed in Bdnf and TrkB mutants during postnatal development be explained?

At birth, a longitudinal expression of Bdnf with highest expression in the apex (Wheeler et al., 1994; Wiechers et al., 1999) or a homogenous pattern of expression(Fariñas et al., 2001)has been described in the cochlear sensory epithelium. Subsequently, Bdnf has been detected in OHCs of the apical to midbasal turn, but not of the basal turn during the first postnatal week(Wiechers et al., 1999). After this timepoint, and in adulthood, no Bdnf expression has been observed in OHCs(Wheeler et al., 1994; Fritzsch et al., 1999; Wiechers et al., 1999). In summary, these data provide no explanation for the basal loss of afferent innervation of OHCs observed towards the end of the third postnatal week and in adulthood.

However, the basal-to-apical gradient of Bdnf and TrkB expression detected in cochlear neurones in the present report offers an attractive explanation for the dependence of afferent innervation on Bdnf and TrkB signalling in the more basal cochlear turns. Specifically, the co-expression of TrkB and Bdnf suggests an autocrine mechanism that maintains the innervation of the basal turn during further maturation of the cochlea. This process may be initiated by the expression of Bdnf in cochlear ganglion neurones during the first postnatal week (Wiechers et al.,1999). The gradient of Bdnf expression is first observed at the end of the second postnatal week (M.K., unpublished observation). Therefore,the spatial and temporal expression of Bdnf in cochlear neurones during postnatal development and in adulthood fits perfectly with the gradual reversal of innervation of OHCs observed in Bdnf and TrkB mutants during the second and third postnatal weeks. In this context, it is worthwhile mentioning that Bdnf is not expressed in differentiated neurones that express TrkB during embryonic development (Fariñas et al., 2001). Therefore, the postulated autocrine mechanism for the maintenance of basal outer hair cell innervation is restricted to the postnatal and adult phases of life. How can the re-innervation of the medial and apical turns of the cochlea in Bdnf and TrkB mutants be explained?

The most attractive model may be an apical-to-basal gradient of expression of Nt3 in the postnatal inner ear. Indeed, such a gradient with the highest expression in the apical part has been described for the cochlear sensory epithelium at birth (Pirvola et al.,1992; Wheeler et al.,1994; Fritzsch et al.,1997b). However, recently this gradient could not be reconfirmed(Fariñas et al., 2001). In the present analysis we have detected an apical-to-basal gradient of Nt3 expression in cochlear neurones, which develops postnatally in the opposite direction to the Bdnf gradient described above. The increased expression of Nt3 in cochlear neurones in the apical turn thus offers an attractive explanation for the re-innervation of OHCs in this part of the cochlea in Bdnf and TrkB mutants. At present we are analysing whether this expression is accompanied by the expression of its high-affinity receptor TrkC. Similar to the situation in the base, co-expression of the neurotrophin and its high-affinity receptor may initiate an autocrine mechanism. Thus, in the apical turns of the cochleae of TrkB and Bdnf mutants, a presumptive Bdnf-independent, Nt3-dependent process of re-innervation of OHCs may be maintained.

At present, it is unclear which neurones are involved in the process of reinnervation. In theory, one would postulate the involvement of two classes of type II sensory neurones, which naturally may provide embryonal and adult innervation to OHCs. In support of this, 'embryonal' type II neurones have been shown to be severely reduced in the absence of Bdnf, as assessed using cell counts in early postnatal Bdnf and TrkB mutants(Ernfors et al., 1995; Schimmang et al., 1995). In the present study, using peripherin as a marker, we also found no evidence for the presence of type II sensory neurones in adult TrkB^shc/- mutants. However, we cannot rule out the existence of type II neurones that have failed to express peripherin in the Bdnf and TrkB mutants analysed in the present study. Moreover, the presence of type II sensory neurones in Bdnf^-/- mutants has been postulated based on the expression of the neurofilament marker RT97(Bianchi et al., 1996). Therefore, two populations of these neurones may exist, which may be involved in initiating the process of embryonal innervation and postnatal re-innervation of OHCs.

It may be considered that neurones innervating OHCs in the apical region,which provide two times more synaptic contacts to OHCs than basally localised neurones of this type (Rubel and Fritzsch,2002), are possibly Nt3-dependent, type I sensory neurones(Ernfors et al., 1995) sending out collaterals to the OHCs. As such, it has been suggested that initially individual ganglion cells innervate both IHCs and OHCs during development(Echteler, 1992; Pujol et al., 1998). In the mammalian cochlea, an extensive process of afferent and efferent fibre reorganisation takes place during the first and second postnatal week(Wiechers et al., 1999). Therefore, the possibility of an early innervation of OHCs by type I neurones,and a Bdnf-dependent phenotypic switch of type I to type II neurones in more basal and less apical turns during postnatal development, should be further investigated. Finally, as shown previously in Nt3 mutants(Fritzsch et al., 1997b; Coppola et al., 2001),tangential extensions of fibres running underneath the cochlea may also contribute to the innervation patterns observed.

The relevance of the neuronal expression of Bdnf described in the present article is also highlighted by recent results on the electrophysiological behaviour of postnatal cochlear neurones(Adamson et al., 2002). In these experiments it was shown that exposure to Bdnf caused all neurones,regardless of their original cochlear position, to display characteristics of the basal neurones. Conversely, Nt3 caused all neurones to show the properties of apical-cochlear neurones. These data fit perfectly with the opposing basal-to-apical gradients of Bdnf and Nt3 expression observed in the postnatal cochlea in the present study.

Hair cells are the targets of olivocochlear fibres that control efferent inhibitory feedback from the brain. Efferent feedback to OHCs is predominantly provided by the release of acetylcholine (ACh) from olivocochlear nerve fibres, which activates the characteristic high Ca²⁺ conductance heterooligomeric complex of ACh receptors (AChRs)(Elgoyhen et al., 1994; Elgoyhen et al., 2001). The Ca²⁺ flux through AChRs in turn, activates the small-conductance Ca²⁺-activated potassium channel SK2. The fast coupling of both is reflected by their co-localisation (Oliver et al., 2000; Knipper and Zenner, 2003). Although current concepts indicate an autonomous morphological differentiation of hair cells independent from ganglion neurones(Sobkowicz, 1992; van de Water et al.,1992; Silos-Santiago et al.,1997; Ma et al.,2000), which acquire their basic physiological properties in the absence of innervation (He and Dallos,1997; Rüsch et al.,1998; Rubel and Fritzsch,2002), data in the present study indicate that distinct steps of the final functional phenotype of OHCs is influenced by the innervating fibres.

The loss of afferent type II fibres below OHCs in Bdnf and TrkB^shc/- mutants is accompanied by a significant reduction in size of efferent synaptic endings and postsynaptic fields on OHCs, in which SK2 channels are inserted. As we observed no defects in the innervation of IHCs, the alteration of the OHC phenotype may directly be related with the observed hearing loss of 10-30 dB in TrkB^shc/- mice across all frequencies tested. This hearing loss and the lack of DPOAEs suggest a common mechanism that is impaired in the mutant animals. As no DPOAEs were detectable, the defect may reside at the level of the OHC and may be caused by the reduction of efferent input. However, at present it remains unclear whether this defect is caused by structural or physiological changes of the OHC.

Previous analysis has shown that the neurotrophin receptor mutants lacking TrkB and TrkC develop central cochlear nuclei of a reduced size(Schimmang et al., 1997). The reduction of these nuclei may be explained by a direct trophic role of neurotrophins within these nuclei. Alternatively, a lack of trophic support of central auditory neurones by the peripheral auditory neurones may be responsible for the observed phenotype. In this context, it is also tempting to speculate that Bdnf expression, which is involved in several plasticity-driven processes (Berardi et al., 2000; Schinder and Poo,2000), may also influence the tonotopic map present in the central part of the auditory system. Expression of Trk receptors and neurotrophins has been described in central auditory neurones(Hafidi et al., 1996; Hafidi, 1998), and their involvement in activity-dependent trophic interactions with the peripheral neurones has been suggested (Rubel and Fritzsch, 2002). The postnatal gradient of Bdnf expression in the cochlear ganglion may thus set up, and later reinforce or maintain, the tonotopic map in the central cochlear nuclei. Therefore, a central component may also be involved in the hearing loss of TrkB^shc/- mice.

Finally, the observation that Bdnf and TrkB are co-expressed in cochlear neurones reinforces the idea of an autocrine mode to maintain functional activity of these neurones (Hansen et al.,2001). Treatment of cochlear neurones with neurotrophins has been shown to stimulate their survival and alter their functional activity(Hegarty et al., 1997; Adamson et al., 2002). Moreover, depolarisation of spiral ganglion neurones together with neurotrophins promotes survival in an additive manner in vitro(Hegarty et al., 1997). Our present results suggest that the interplay of an autocrine neurotrophic mechanism and neuronal activity may also maintain the survival and correct functioning of auditory neurones in vivo, thus ensuring the intactness of the sensory organ. In this context, it is worthwhile mentioning that the expression of Bdnf described in the present article may also give a functional explanation for the positive activity of neurotrophins in animal models of hearing loss. Application of Bdnf and Nt3 to the inner ear has been shown to ameliorate hearing loss in several of these models(Ernfors et al., 1996; Duan et al., 2002; Shinohara et al., 2002). Our results suggest that the underlying mechanism is the stimulation of cochlear neurones by neurotrophins, which may lead to a functional re-stimulation of neuronal activity. Further experiments will show if re-innervation processes,like the ones observed in the present study, can also be re-initiated in postnatal animals and adults. In summary, our results may provide an important step for understanding the basis of treatment of hearing loss by neurotrophins, and may eventually allow us to improve therapies by targeting specific neurotrophins to subsets of cochlear neurones.

We thank Rüdiger Klein for providing TrkB mutants and for critical reading of the manuscript, Bernd Fritzsch for help in the interpretation of the TrkB^shc/shc mutant, and Michael Sendtner and Bettina Holtmann for providing the Bdnf mutants. This work was supported by the DFGSFB444, DFG KN 316/3-2, fortüne Kni-F1251236, and fortüne Kni-F1251279.