Type II spiral ganglion afferent neurons drive medial olivocochlear reflex suppression of the cochlear amplifier

The dynamic adjustment of hearing sensitivity and frequency selectivity is mediated by the medial olivocochlear efferent reflex, which suppresses the gain of the ‘cochlear amplifier' in each ear. Such efferent feedback is important for promoting discrimination of sounds in background noise, sound localization and protecting the cochleae from acoustic overstimulation. However, the sensory driver for the olivocochlear reflex is unknown. Here, we resolve this longstanding question using a mouse model null for the gene encoding the type III intermediate filament peripherin (Prph). Prph(−/−) mice lacked type II spiral ganglion neuron innervation of the outer hair cells, whereas innervation of the inner hair cells by type I spiral ganglion neurons was normal. Compared with Prph(+/+) controls, both contralateral and ipsilateral olivocochlear efferent-mediated suppression of the cochlear amplifier were absent in Prph(−/−) mice, demonstrating that outer hair cells and their type II afferents constitute the sensory drive for the olivocochlear efferent reflex.

H earing performance is dependent upon the active enhancement of sound-induced vibration of the cochlear organ of Corti through prestin-mediated outer hair cell (OHC) electromotility 1 . The action of this 'cochlear amplifier' 2,3 is responsible for the sensitivity and exquisite frequency resolution of mammalian hearing and can be assessed by measuring distortion product otoacoustic emissions (DPOAEs) 4,5 in the ear canal. The cochlear amplifier selectively amplifies cochlear vibrations in a focused region of the cochlea, thereby enhancing sound transduction at inner hair cells (IHCs), each of which receives exclusive sensory innervation from several type I spiral ganglion neurons (SGNs), whose central projection to the brainstem cochlear nuclei drives hearing perception. The relative levels of sound transduced by IHCs between the two ears is principally regulated by a sensorimotor reflex pathway, which responds to elevation of sound in one ear by rapidly reducing the gain of the cochlear amplifier in the opposite cochlea [5][6][7] . This 'contralateral suppression' is produced by medial olivocochlear (MOC) efferent neurons located in the superior olivary complex of the brainstem, whose myelinated axons approach the floor of the fourth ventricle from both sides before migrating laterally to join the ipsilateral auditory nerve and terminate on the OHCs 8 . These cholinergic MOC synapses hyperpolarize the OHCs, inhibiting OHC electromotility and thus reducing the gain of the cochlear amplifier [9][10][11][12] . In addition to contralateral suppression, there is also an ipsilateral pathway through which MOC neurons suppress the cochlear amplifier 7,13 . Efferent control of the cochlear amplifier has been shown to be necessary for speech discrimination in noise 14,15 , sound localization 16 and protection from noise-induced hearing loss 17 . However, the sensory input that drives this neural control process has not been identified 8 .
The cochlear amplifier gain control reflex must either originate with primary auditory encoding of the sound perception channel by type I SGN, which make up 95% of the spiral ganglion, and/or with the residual 5% of the SGNs-designated type II, which exhibit a dispersed sensory innervation of OHCs [18][19][20][21][22][23][24][25][26] . The function of the type II SGN remains enigmatic 8 . Here, we assess the role of type I versus type II SGNs in the MOC reflex by utilizing a mouse model, null for gene encoding the type III intermediate filament peripherin (Prph ( À / À ) ), that is found to lack type II SGN innervation of OHCs. Absence of suppression of the cochlear amplifier-derived DPOAEs in response to contralateral or ipsilateral sound in these Prph ( À / À ) mice supports the hypothesis that the putative OHC-type II SGN sensory transmission drives the MOC efferent regulation of cochlear amplifier gain. Thus, MOC reflex control of hearing sensitivity utilizes a closed-loop negative-feedback pathway with the OHCs operating as both the sensor and effector.

Results
Prph ( À / À ) mice lack type II SGN innervation. In the mouse cochlea, peripherin (PRPH) is exclusive to the type II SGNs during the establishment and maturation of the OHC innervation [27][28][29][30] . We examined cochlear innervation patterns in an established Prph ( À / À ) mouse line 31 and found complete absence of OHC innervation by unmyelinated type II SGN neurites, consistent with key roles for PRPH in the regulation of peripheral sensory fibre extension and maintenance 28,[32][33][34] . The Prph ( À / À ) mouse line was initially characterized as lacking a substantial proportion of unmyelinated sensory fibres from the L5 dorsal root ganglion 31 , which prompted our assessment of this model to investigate the physiological significance of the unmyelinated cochlear type II SGN sensory fibres.
The type II SGN innervation of the OHCs was identified using PRPH immunofluorescence in cochlear cryosections and surface mounts of the Prph ( þ / þ ) organ of Corti from postnatal day 4 ( Supplementary Fig. 1). The PRPH-positive type II SGN neurites extend from the somata in Rosenthal's canal, past the type I SGN (radial fibre) innervation of the IHCs at the inner spiral plexus, crossing the floor of the tunnel of Corti. The type II SGN fibres then turn basally at the Deiters' cell level to form the outer spiral bundle (OSB) that innervates multiple OHCs 19,22 . This was most evident in the neonatal cochlea ( Supplementary Fig. 1a,c-f), whereas PRPH immunolabelling in the adult cochlear type II SGN fibres diminished beyond the inner spiral plexus ( Supplementary Fig. 1b,g,h), compared with resolution of the OSB with neurofilament 200 kDa (NF200) immunolabelling (Fig. 1a,c). The absence of PRPH in the Prph ( À / À ) spiral ganglion was confirmed ( Supplementary Fig. 1b    The OSB was absent in the Prph ( À / À ) organ of Corti (Fig. 1b,d), a few residual type II SGN fibres were occasionally observed extending toward the OHCs, although none are visible in the fields shown here. In contrast, the MOC efferent innervation of the OHCs was preserved (Fig. 1b,d and Supplementary Movie 2). The loss of the OSB in the Prph ( À / À ) cochleae was confirmed in eight adult mice, alongside well-resolved OSBs imaged in cochleae from five Prph ( þ / þ ) and one Prph ( þ / À ) animal, using NF200 immunofluorescence. Thus knockout of Prph expression selectively disrupted the type II SGN afferent innervation of the OHCs.

Figure 1 | Loss of afferent innervation of outer hair cells (OHC) in
Serial block face imaging in the (field emission) scanning electron microscope (SBF-SEM) resolved the small type II SGN afferent synaptic boutons adjacent to the large MOC efferent boutons at the base of the Prph ( þ / þ ) OHCs in the mid-apical turn region of the cochlea ( Supplementary Fig. 2a,c,d,g). No type II SGN synaptic boutons were identified in Prph ( À / À ) OHCs ( Supplementary Fig. 2b,e,f,g). The Prph ( þ / þ ) OHCs typically had a single type II SGN bouton with a mean volume of 1.20±0.10 mm 3 (average of data from three blocks from different mice), as well as three MOC efferent boutons (mean volume ¼ 14.22±0.83 mm 3 ), consistent with data from conventional transmission electron microscopy (TEM) analysis, including discrimination of the efferent synapses by their size, abundant mitochondria and typical presence of electron-dense postsynaptic cisternae 35 . In the Prph ( À / À ) , the efferent bouton number and distribution was comparable to the Prph ( þ / þ ) OHCs, with some hypertrophy (mean volume 19.05 ± 0.95 mm 3 (n ¼ 3); P ¼ 0.0186 (SA1)) (Supplementary Fig. 2g; Supplementary Table 1).
We validated our contralateral suppression model in two control studies: acoustic cross-talk between cochleae was excluded by physically disrupting the contralateral tympanic membrane and ossicles (contralateral suppression 5.5 ± 0.9 dB before versus 0.2±0.2 dB after; P ¼ 0.004 (SA6), n ¼ 3; 30 s 82 dB SPL 10-17 kHz noise; primaries at 65 dB SPL B28 kHz). To confirm that the MOC efferent pathway was involved (and eliminate a role for the middle ear muscle reflex, which drives the ossicular chain tensor tympani and stapedius muscles, increasing middle ear impedance and reducing sound transmission during loud sound and vocalization 37 ), we made a rostral-caudal incision at the floor of the fourth ventricle ( Supplementary Fig. 5, contralateral suppression 3.9 ± 0.6 dB before versus 1.2 ± 0.3 dB post section, P ¼ 0.043 (SA6), n ¼ 3; ref. 7).
Prph ( À / À ) mice lack ipsilateral suppression. Ipsilateral suppression is stronger than contralateral suppression, but more challenging to quantify because the sound used to elicit the MOC reflex is delivered over both the primary tones and the DPOAEs 13 . We established a mouse model for ipsilateral suppression of the cochlear amplifier by measuring the recovery from a brief ipsilateral noise during on-going quadratic DPOAE measurement.
MOC-recruiting ipsilateral sound stimulation (82 dB SPL, 10-25 kHz, 5 s) was presented over the DPOAE primary tones (60 dB SPL, around 20 kHz). Immediately after the suppressor noise ended, DPOAE measurements were restored and a pronounced reduction in quadratic DPOAE was evident in Prph ( þ / þ ) and Prph ( þ / À ) mice. This ipsilateral suppression adapted within B10 s, to a sustained offset from the pre-noise intensity (immediate DPOAE intensity reduction 9.7 ± 2 dB; average DPOAE reduction at 21 to 45 s post noise 4±1.4 dB; P ¼ 0.004 (SA3); n ¼ 9; Fig. 4a). These results contrasted with a smaller immediate DPOAE reduction post noise (5.0±1.9 dB) in the Prph ( À / À ) mice, that very rapidly (B10 s) returned to   . This showed that the immediate post noise reduction in DPOAE was reduced to that seen in the Prph ( À / À ) mice, and the measurements from B9 s post noise similarly returned to baseline. Thus ipsilateral suppression without the MOC contribution in Prph ( þ / þ ) mice was indistinguishable from that observed in Prph ( À / À ) mice with an intact MOC fibre tract.

Discussion
Our findings determine that the OHC-type II SGN sensory pathway drives the MOC efferent reflex-mediated control of the cochlear amplifier. Thus, not only does sound transduction by OHCs produce the electromotility that is the substrate of the cochlear amplifier, but integration of OHC transduction at the type II SGN afferent synapses provides the sensory coding of cochlear amplifier gain that drives the negative feedback control via the MOC efferent innervation to both cochleae (evident as contralateral (Fig. 5a) and ipsilateral (Fig. 5b) suppression).
No functional contribution to hearing has previously been established for type II SGN 8,25 ; although central projections to the cochlear nuclei alongside the type I SGN have been documented 25,39 . Type II SGN excitability has been demonstrated in vivo by cochlear electrical stimulation 40 , and their membrane properties demonstrated in vitro [22][23][24] . Significantly, type II SGN make multiple en passant synapses across multiple rows of OHCs approximately a quarter of an octave basal to the corresponding innervation of IHC by the type I SGN radial fibres 8,25 . This is consistent with the coding of the active region of the cochlear amplifier relative to the tonotopic coding by the corresponding IHC 8 . The match of the type II SGN OHC innervation to the MOC efferent regulation of the OHCcontrolled cochlear biomechanics was originally highlighted by Kim 20 . The concept has also been supported by the recent electrophysiological evidence from isolated rat organ of Corti, showing integration of synaptic input from multiple OHCs by type II SGN 26 .
The role of type II SGN fibres in conveying the operational state of the cochlear amplifier to the MOC neurons is also compatible with the known central projections of these fibres to the cochlear nuclei. The type II SGN input overlaps extensively with the type I SGN projections in the dorsal and ventral nuclei, as well as having differential input to the granule cell region (reviewed by Nayagam et al. 25 ). Transneuronal retrograde labelling via pseudorabies virus injections into the cochlea indicate that the cochlear sensory drive to the MOC efferent neurons within the superior olivary complex is most likely relayed through planar multipolar cells of the posterior and anterior ventral cochlear nuclei 41 , which overlaps with the type II SGN projection.
The quadratic DPOAE contralateral suppression adaptation profile, which takes tens of seconds, is consistent with slow adaptation of MOC efferent-mediated inhibition of the cochlear amplifier downstream of the OHC a9/a10 nAChRs [42][43][44] . Our Prph ( À / À ) model has also enabled the isolation of the sensory drive for MOC efferent suppression arising from ipsilateral sound. These data show that the ipsilateral suppression is substantially stronger than the contralateral suppression for the same input intensity, and adapts more slowly than the contralateral suppression. This is evident from the sustained B5 dB suppression after 30 s of ipsilateral noise (82 dB SPL; Fig. 4b,b inset) compared with o1 dB residual suppression after 30 s contralateral noise (82 dB SPL; Fig. 3a). Consequently, the rapid adaptation of contralateral suppression likely reflects ipsilateral adaptation of the sensory coding from the cochlear amplifier via the OHC-type II SGN in the (contralateral) cochlea receiving the noise probe, rather than adaptation of the downstream MOC efferent output to the opposite cochlea. Hence, the sensory drive for contralateral suppression attenuates rapidly. The dominance of ipsilateral suppression is correlated with the bias in neural connectivity from the ipsilateral cochlear nuclei to the MOC neurons on the opposite side of the brainstem 41 . The discrimination between ipsilateral and contralateral suppression properties achieved here by isolating the type II fibre sensory driver has considerable significance for understanding the weighting and differential adaptation of the gain control of the cochlear amplifiers in each ear, and is likely to have bearing on the design of cochlear implant systems for binaural implantation.
Our study is also informative with regard to the role of the intermediate filament PRPH, in sustaining neural connectivity. Several studies have demonstrated its action in other neural populations, including the finding that a subpopulation of unmyelinated dorsal root ganglion sensory neurons was absent in the Prph ( À / À ) mouse 31 .
Atrophy of peripheral neurites in the type II SGN OSB fibres was nearly complete in the Prph ( À / À ) mice. This is likely due to loss of PRPH-dependent neurite extension and stabilization cues 33 in the critical postnatal period for establishment of the cochlear afferent innervation 25,27,29 . In this regard, the mouse model provided a fortuitous advantage, as Prph expression by SGN is constrained to the type II fibres from around E18 in the mouse, which precedes the afferent innervation of the hair cells 27,29,30 . In the rat, Prph expression is sustained in the type I SGN over the early critical afferent innervation period 34,45 . It has not been established whether mouse type I SGN fibres express Prph at earlier embryonic periods, but the normality of the ABR function in the Prph ( À / À ) mice indicates that, even if this is the case, it has not affected the establishment of the central auditory circuitry, making it unlikely that an indirect action on type I fibres is a factor in the loss of the MOC reflex suppression of outer hair cell electromotility evident in these mice. Peripherin expression in the adult mouse brainstem is highly constrained and, outside of the central projections of the type II SGN, is not correlated with the MOC reflex pathway 46 .
The elimination of the type II afferent boutons may also have effects on the local neural circuitry within the organ of Corti, as some interconnectivity between type II afferent fibres, MOC efferent varicosities and inner hair cells has been identified 47 . The hypertrophy of the MOC efferent boutons at the base of the OHCs in the Prph ( À / À ) mice is consistent with the reduction in sensory drive 48 to the central MOC efferent reflex circuit, evident as a loss of dynamic suppression of the DPOAEs when sound is presented to the ears.
Our data are consistent with the concept that the binaural regulation of hearing sensitivity via the MOC efferent fibremediated suppression of cochlear amplifier gain is largely an autonomous reflex that utilizes a putative non-perceptive auditory coding channel provided by the OHCs and type II SGNs. This complements the lateral superior olivary complex efferent fibres, which directly regulate the type I SGN output via presynaptic termination beneath the IHCs 12  encoding of OHC transduction that underpins the cochlear amplifier state, provides the sensory drive to adjust gain between the two ears. This is evidently critical for balanced hearing, which enables optimal sound localization 16 and speech discrimination in noise 14,15,50 . The MOC efferent pathway has also been implicated in protection from hearing loss from sustained noise 17 . Given that type II SGN loss is prevalent with aging 51 , degradation of the OHC/type II SGN drive to MOC efferentbased suppression of the cochlear amplifier may contribute to the decline in hearing performance in noise with aging, and to agerelated hearing loss.
Hearing function tests. Mice were anaesthetized using a ketamine cocktail containing ketamine (40 mg kg À 1 ), xylazine (8 mg kg À 1 ) and acepromazine (0.5 mg kg À 1 ) via intraperitoneal injections. Hearing testing was carried out using an auditory-evoked potential and DPOAE workstation (TDT system 3 with RX6 and RX6-2 signal processors, Tucker Davis Technologies, Ft Lauderdale, FL, USA) with BioSig32 software. Sound levels were calibrated using a one-quarter-inch Free Field Measure Calibration Microphone (model 7016; ACO). Hearing testing was carried out in a sound-attenuating chamber (Sonora Technology, Japan). The level of anaesthesia (indicated by heart rate and breathing rate) was monitored using a mouse oximetry system (MouseOx, STARR Life Sciences) and depth of anaesthesia was further assessed by lack of tail-pinch response. The core body temperature of the mice was clamped at 37°C by feedback control of a heating pad (Right Temp, Able Scientific). DPOAE: The right external auditory meatus of each mouse was coupled to a small microphone used to detect changes in sound pressure which represent DPOAEs (Etymotic ERB10 þ , Etymotic Research). Two EC1 electrostatic speakers, controlled by the Tucker Davis Technologies (TDT) system (BioSigRP software), generated equal intensity primary tones (f 1 and f 2 ; f 2 /f 1 ratio: 1.25). Recordings were taken at 8, 12, 16 and 24 kHz presented from 0 to 70 dB SPL (increasing in 5 dB steps). Fifty measurements at 6.7 per second were averaged for each stimulus intensity and frequency, which were analysed by Fast Fourier transformation. Thresholds at the different frequencies were determined based on when the cubic (2f 1 -f 2 ) distortion product reached 5 dB above the noise floor, as the tone intensity increased.
ABRs were recorded using three platinum subdermal electrodes inserted subcutaneously at the vertex ( þ ), over the mastoid process ( À ) and near the base of the tail (ground). An EC1 speaker provided either click, or puretone pip (8, 16 and 24 kHz) stimuli at 10 per second, in descending 5 dB steps (starting at 70 to 80 dB SPL). Each trace studied is an average of up to 512 raw recordings at each intensity and frequency. Thresholds at the different frequencies were determined as the lowest intensity at which ABR waves could be observed.
MOC efferent-mediated suppression of DPOAE: Either the contralateral or ipsilateral ear was exposed to broadband suppressor sound stimulation while DPOAEs were elicited using primary tones in the ipsilateral ear. 10-50 measurements were averaged (6.7/s) for each recording. Baseline DPOAE measurements were taken before and recovery was monitored following the suppressor stimulus. The amplitude of the quadratic (f 2 -f 1 ) distortion product relative to the noise floor was then analysed before, during and after the noise stimulation. Broadband noise was generated using a National Instruments A/D, with custom software to create the noise. The output was amplified using an SA1 amplifier (TDT) and delivered closed-field via either an MF1 or FF1 speaker (TDT) using a coupler.
Surgical disruption of MOC efferent activity. These were all carried out in adult CBA/129mice under ketamine (40 mg kg À 1 ), xylazine (8 mg kg À 1 ) and acepromazine (0.5 mg kg À 1 ) anaesthesia. The contralateral tympanic membrane and ossicles were surgically disrupted. Sectioning of the MOC neurons: The contralateral suppression was robust and repeatable after an hour under the ketamine/ xylazine/acepromazine anaesthesia. Subsequently the MOC efferent fibres beneath the floor of the fourth ventricle were sectioned with a scalpel blade incision just ipsilateral to the midline, achieved via a dorsal fenestration between C1 and the foramen, with supplementary isoflurane anaesthesia.
Immunohistochemistry. Immunohistochemical labelling was carried out on fresh tissue from unexposed animals. Mouse cochleae were fixed by scali perfusion of 4% paraformaldehyde, decalcified for 14 days in 8% EDTA and then dissected for surface mount preparation or cryoprotected using sucrose and OCT before being cryosectioned at 50 mm. Nonspecific binding was blocked with 10-15% normal goat or donkey serum, 1% Triton X-100 in phosphate-buffered saline (PBS) for 1 h at room temperature (RT) and sections were then transferred to the primary antibody solution; Neurofilament 200 kDa (Sigma, rabbit, 1:5,000); peripherin (EB12405, Everest Biotech, goat, 1:10,000); CtBP2 (BD Bioscience, mouse, 1:500); vesicular acetylcholine transporter (VAChT; Phoenix, goat, 1:200) in 5-15% normal goat or donkey serum, 0.1% Triton X-100 in PBS) overnight at RT. The sections were washed in PBS and appropriate secondary antibody was applied overnight RT (Molecular Probes, 1:500 anti-rabbit IgG AlexaFluor 594, anti-goat IgG AlexaFluor 488, anti-mouse IgG AlexaFluor 488). Following a further PBS wash, DAPI was applied for 5 min at 1:5,000. Two final washes in PBS were carried out before the tissue was mounted using Vectashield (Vector Laboratories) and imaged on a Zeiss confocal microscope (Zeiss 710 NLO). Data were obtained from 1 to 3 organ of Corti z stacks (30 mm) analysed per animal.
Conventional TEM was used to identify tissue preparation quality, and the outer hair cell region, with samples imaged at x1-40 k using a digital TEM camera at 4 MP. Block faces were then trimmed to B50-80 mm 2 to facilitate optimal imaging for subsequent SBF-SEM (3View, Gatan). Serial 50 nm sections were cut and block faces imaged through B50 mm of tissue per block. Blocks from three mice per genotype were scanned. Serial section EM, along with the increased electron density of the afferent boutons and evident synaptic vesicles and an electron-dense postsynaptic cistern in the efferent boutons provided unequivocal discrimination of the type II afferent boutons from the MOC efferent boutons.
Outer hair cells and their associated afferent and efferent synaptic boutons were manually identified from their characteristic morphology, and segmented using TrakEM2 (ref. 52), a plugin developed for ImageJ (NIH) and packaged in the FIJI distribution 53 . Segmentation of multiple slices along the axial plane yielded threedimensional reconstructions on which volume measurements were made.
Statistical analysis. Data are expressed as population mean±s.e.m. Statistical analysis was carried out in SigmaPlot (V. 11, Systat Software, Germany). Statistical tests cited in the text are indicated in Table 1. Holm-Sidak post hoc analysis was utilized for multiple pairwise comparisons within analysis of variance, providing a correction for repeated testing to maintain the type I error rate (false positive). Data were tested for normal data distribution and a rank transformation was applied where required. Power calculations based on pilot experiments informed experimental design.