Cytoarchitecture and innervation of the mouse cochlear amplifier revealed by large‐scale volume electron microscopy

Abstract In mammalian cochlea, sound‐induced vibration is amplified by a three‐row lattice of Y‐shaped microstructures consisting of electromotile outer hair cell and supporting Deiters cell. This highly organized structure is thought to be essential for hearing of low‐level sounds. Prior studies reported differences in geometry and synaptic innervation of the outer hair cells between rows, but how these fine features are achieved at subcellular level still remains unclear. Using serial block‐face electron microscopy, we acquired few‐hundred‐micron‐sized cytoarchitecture of mouse organ of Corti at nanometer resolution. Structural quantifications were performed on the Y‐shapes as well as afferent and efferent projections to outer hair cells (OHCs). Several new features, which support the previously observed inter‐row heterogeneity, are described. Our result provides structural bases for the gradient of mechanical properties and diverse centrifugal regulation of OHC rows.


| INTRODUCTION
In the mammalian cochlea, sound pressure changes are converted into traveling waves on the basilar membrane (BM). The resultant vibration is then transduced by the organ of Corti (OC), which is arranged along the coiled BM, into electrical signals in auditory nerve fibers (ANFs).

Two types of hair cells reside in the OC, with inner hair cells (IHCs)
being the true sensory cell type for detecting the BM motion and coding the acoustic information. In contrast, outer hair cells (OHCs), which are located on the lateral side of the OC, are responsible for modulating the BM motion via their electromotility and in turn the IHC sensation. This remarkable structure plays an essential role in the cochlear amplification for the perception of low-level sounds (Dallos et al., 1996;Meyer & Moser, 2010;Raphael & Altschuler, 2003).
There are three rows of OHCs docked on a honeycomb-like network of supporting Deiters cells (DCs) along the BM. At their epithelial surface, the reticular lamina (RL), OHCs and the phalangeal processes (PhPs) from the DCs are adjoined and precisely arranged in a mosaic-like organization.
As the PhPs are tilted in the opposite direction of the OHCs, it yields between the RL and BM a lattice of characteristic Y-shaped structures arranged in the direction of the basilar membrane's traveling wave. The resultant structural coupling between the electromotile OHCs and Haoyu Wang, Shengxiong Wang, and Yan Lu contributed equally to this work. passive PhPs, as proposed in prior studies (Motallebzadeh et al., 2018;Wen & Boahen, 2003;Yoon et al., 2011), has an important effect on the mechanical behavior of the cochlea. However, early structural investigations using electron microscopy (EM) were limited by either specimen volume or structure accessibility. By far, the most knowledge about this Y-shaped building block was gained from the third (outermost) OHC-DC row in fixed OC of mole rat (Raphael et al., 1991), hence lacking anatomical information of the first two rows (inner rows). More recent quantification in the mouse cochlea using in situ two-photon imaging revealed significant differences in structural details of the Y-shapes between rows, including longitudinal angles and PhP lengths (Soons et al., 2015). Assuming negligible changes in tissue morphology caused by opening the cochlea, this result indicates putative heterogeneous mechanical properties of the Y-shapes in different rows.
Moreover, the OHC motility is tightly regulated by centrifugal innervations to enable signal discrimination in a noisy background and binaural signal localization (see reviews, Fuchs & Lauer, 2019;Wersinger & Fuchs, 2011;Zhang & Coate, 2017). At the OHC basal pole, DC forms a cup-like structure housing both afferent and efferent projections of the central auditory nervous system. In mammals, thin unbranched afferent fibers of type-2 spiral ganglion neurons (SGNs) turn 90 toward basal cochlear region after crossing the floor of the tunnel of Corti (TC) and travel a characteristic 100-200 μm distance in parallel to the OHC rows before making synaptic contact (Berglund & Ryugo, 1987;Dannhof & Bruns, 1993;Spoendlin, 1972). Currently, much less is known about the exact function of the type-2 SGN and the linkage to its characteristic anatomical features. Several concepts have been proposed including auditory nociception (Flores et al., 2015), cochlear damage report (Liu et al., 2015) as well as efferent control (Froud et al., 2015;Maison et al., 2016). In addition, OHCs are also the primary targets of medial olivocochlear (MOC) efferent neurons (Warr et al., 1997). The MOC fibers cross at the middle level of the TC and synapse onto multiple OHCs with an unusual inhibitory action of acetylcholine (Blanchet et al., 1996;Dallos et al., 1997;Evans et al., 2000;Wersinger & Fuchs, 2011). In rat, two subtypes of MOC fibers were identified based on their abundance of the tunnel-crossing fibers and boutons (Warr & Boche, 2003). However, it remains unclear the functional implication behind such diversity of axonal ramifications.
In order to elucidate the structural basis of the cochlear amplifier as well as its circuit wiring principle, we performed comprehensive ultrastructural analysis in large-scale EM volume of mouse mid-cochlea region using serial block-face EM (SBEM, Denk & Horstmann, 2004

| Sample preparation
SBEM sample preparation was performed following the previously published procedure (Hua et al., 2021). In brief, fresh temporal bones were harvested from adult CBA mice. The cochlea was fixed by perfusion with ice-cold fixative mixture (2% paraformaldehyde and 2.5% glutaraldehyde buffered in 0.08 M cacodylate, pH 7.4) through round window followed by a 5-hour postfixation at 4 C. Decalcification was done by 4-hour immersion in the same mixture with addition of 5% EDTA. The decalcified cochleae were washed twice then en bloc EM stained by sequentially incubating in 0.15 M cacodylate buffer (pH 7.4) containing 2% OsO 4 , 2.5% ferrocyanide, and again 2% OsO 4 at room temperature for 2, 2, and 1.5 hours, respectively. After two 30-min washes with nanopore-filtered water, the cochleae were incubated at room temperature in filtered thiocarbonhydrazide (saturated aqueous solution) for 1 hour, unbuffered OsO 4 aqueous solution for 2 hours and lead aspartate solution (0.03 M, pH 5.0 adjusted by KOH) at 50 C for 2 hours with intermediate washing steps. Dehydration and embedding were done through a graded acetone series (50%, 75%, 90%, 30 min each, all cooled at 4 C) into pure acetone (three times 100%, 30 min at room temperature) followed by sequential infiltration with 1:1 and 1:2 mixtures of acetone and Spurr's resin monomer (4.1 g ERL 4221, 0.95 g DER 736, 5.9 g NSA, and 1% Dimethylaminoethanol) and the pure resin at room temperature for 6 hours each. The infiltrated cochleae were then placed in embedding molds and incubated in a prewarmed oven (70 C) for 72 hours.
For the data acquisition, serial images were registered in single-tile mode (20,000 × 15,000 pixels) at 11 nm pixel size and a nominal cutting thickness of 40 nm; incident beam energy 2 keV; dwell time 1 μs. Focal charge compensation (Deerinck et al., 2018) was set to 100% with a vacuum chamber pressure of 2.8 × 10 −3 mbar. In total, 5193 consecutive slices were collected and aligned using self-written MATLAB script based on crosscorrelation maximum, yielding an EM volume of 325.0 × 204.4 × 207.7 μm 3 .

| Cytoarchitecture reconstruction and data processing
Structure inspection and annotation were done in a browser-based annotation tool, webKNOSSOS (Boergens et al., 2017). Each neurite was at least two-fold traced by multiple observers. The results of the first annotator were proofread by another one for missing branches and for few confusing cases a third annotator was involved in voting.
For quantification of geometrical features such as 3D angle and length, landmarks were created: OHCs were skeletonized along the RL-BM axis; DCs and PhPs were annotated by cytoskeleton tracing; the plane of BM was determined by the roots of DCs; the intersection angle (α) between two branches of Y-shapes were measured from vectorized OHC and PhP (svd, MATLAB build-in function). Longitudinal angles of OHCs, PhPs, DC main trunks, and OPCs were measured in reference to the linear-fitted DC rows. Efferent and afferent fibers were identified based on their characteristic morphology such as fiber trajectory and the presence of presynaptic bouton filled with vesicle cloud.

| Statistics
All data analysis including statistical tests were performed using self- (c), as well as paired t-test (ttest) for Figure 2(e). Data were reported as mean ± SD and the significance level of statistical tests was denoted as n.s. for p-value > .05, * for p < .05, ** for p < .01, and *** for p < .001.

| Cytoarchitecture of the Y-shaped OHC-DC complex
The large EM volume allowed quantitative structural analysis of the 88 intact OHC-DC complexes in 3D (Figure 1(a)) with 29 in the first row (innermost), 30 in the second row (middle), and 29 in the third row (outermost). We skeletonized the OHCs, OPCs and the characteristic microtubule bundles in the DCs as shown in Figure 1 Slightly increased OHC length was found across three rows with the longest OHC in the third row ( Figure 1(c), OHC1: 22.80 ± 0.80 μm; OHC2: 23.45 ± 1.05 μm; OHC3: 24.75 ± 1.06 μm), consistent with previous OHC length measurement (Dannhof et al., 1991) in the middle turn region (50%-60% of the cochlear length, frequency range of 12-16 kHz). For the PhPs, similar length was observed from the first two rows, while DCs in the third row have significantly shorter PhPs (Figure 1( 66.58 ± 3.75 ; PhP3: 75.01 ± 3.25 , Figure 1(h)), while the tilt of OHCs is similar between rows (OHC1-3: 107.28 ± 3.44 , Figure 1(h)).
Given that the relative abundance of afferent fibers per DC decreases from DC1 to DC3 (Figure 2(g), DC1: 20.4 ± 1.3, n = 5; DC2: 13.6 ± 1.7, n = 5; DC3: 7.6 ± 1.5, n = 5), our result suggests an outspiral morphology of type-2 SGNs across DC rows. We next focused on the OHC innervation of SGNs and found that on average the first row OHCs contact with 20% more afferent terminals than those in the second and third rows (Figure 2

| Confirmation of structural findings in second animal
In order to validate these structural findings, we analyzed randomly sampled structures in a SBEM dataset acquired from another CBA animal (p60, middle turn). The volume is 214.8 × 252.6 × 147.6 μm 3 sized at slight lower resolution (12 × 12 × 50 nm 3 ). In total 45 Yshapes (Figure 4(a)), 10 type-2 SGNs (Figure 4(c)) as well as 16 MOCfibers (Figure 4(e)) were annotated. Note that a small region with four rows of OHCs was excluded in the analyses. First, we observed comparable structural changes in Y-shapes across OHC rows (Table 1)

| DISCUSSION
Serial section EM approaches have powered 3D ultrastructure investigations in both inner and outer hair cells (Anttonen et al., 2014;Bullen et al., 2015;Dannhof & Bruns, 1993;Fuchs et al., 2014;Glueckert et al., 2005;Hashimoto et al., 1990;Hua et al., 2021;Liberman et al., 1990;Thiers et al., 2008). In this study, we presented large-scale cytoarchitecture of the mouse cochlear amplifier revealed by SBEM technique (Denk & Horstmann, 2004). To the best of our knowledge, the ultrastructural quantifications in the OSB region were for the first time achieved for a BM of more than 200 μm in length.
Extended structural insights were gained in well-organized DC framework, type-2 SGN morphology as well as row-specific tuning of afferents and efferents on OHCs, allowing model refinement of the mouse OC in 3D. In comparison to dissecting and flattening cochlear tissues, whole cochlea en bloc preparation has the advantages in preserve 3D structural information and minimize mechanical damage. However, it is known that slow chemical fixation can introduce artifacts mainly F I G U R E 4 Comparison with the second CBA animal. (a) Top view (top) and side view (below) of skeletonized outer hair cells (OHCs) and Deiters cells (DCs). One example column of OHCs (red) and DCs (with PhP, green) were highlighted in colors. Note different intersection angles between OHC and PhP in different rows. (b) The intersection angles between the OHC and PhP in the second CBA animal. The first row: 52.36 ± 3.14 (n = 15); the second row: 44.73 ± 2.45 (n = 15); the third row: 35.65 ± 3.07 (n = 15). Two-sample t-test: ***p < .001 (Row1 vs. Row2); ***p < .001 (Row2 vs. Row3). One-way ANOVA: ***p < .001. (c) Top view of the traced type-2 SGNs in the second CBA (gray). Three representative fibers (Row1: Dark blue; Row2: Blue; Row3: Light blue) were illustrated with row-specific innervation and ribbon-associated terminals (red).  Note: Data are presented as mean ± SD. The significance level of statistical tests was denoted as *** for p < .001 tissue shrinking. In our hands, best ultrastructure preservation was achieved by cochlear perfusion with ice-cold fixative, which was further supported by largely comparable measurements to those obtained in vivo (Soons et al., 2015). Moreover, improved spatial resolution of EM compared to light microscopy allows visualization of structural features in cellular organization, fiber morphology as well as synaptic contact, enabling comparative structural investigations at neural circuit level in disease animal models of hearing loss, for example, cochlear synaptopathy (Kujawa & Liberman, 2009;Stamataki et al., 2006), cell death (Anttonen et al., 2014;Stamataki et al., 2006) and aging (Lauer et al., 2012;Sergeyenko et al., 2013).
In line with prior study (Soons et al., 2015), inter-row structural differences in Y-shaped OHC-DC complexes were observed, including slightly longer OHC but significantly shorter PhP in the third row ( Figure 1(c,d)). This is interesting, because it was found recently that DCs in the third row are transcriptionally different than those in the first two rows (Kolla et al., 2020). Unexpectedly, we found the RL span between OHC and PhP of the same DC is exact four, three and two OHC columns for the first, second, and third rows, respectively ( Figure 1(i)), introducing distinct angles of the Y-shapes in different rows, at least at the mid-cochlear region. Our data strengthen the observation from previous study using two-photon light microscopy (Soons et al., 2015) and argue for potential biological relevance behind the inter-row heterogeneity, as such well-organized cytoarchitecture is unlikely caused by tissue damage or deformation during opening the cochlea. Although coordinated movement of three rows of OHCs is believed essential for cochlear amplification, minimal contribution of the third row to the cochlear sensitivity was argued (Chen et al., 2008). Indeed, the observation of steeper upper branches of the third row Y-shapes implies increased structural rigidity. In contrast, wide angle in the first two rows may allow more dramatic structure deformation upon increased shear force as calculated from the modeling (Liu et al., 2011). This feature may provide the reason why the first row OHCs are most vulnerable to permanent acoustic injury (Liberman & Dodds, 1984;Liberman & Kiang, 1978;Robertson, 1982;Yoshida & Liberman, 1999). Moreover, the OPCs, which are tilted in parallel to OHCs, form a crisscross pattern with DCs ( Figure 1(e); Dallos et al., 1996) and thereby can provide extra support for the first row.
Owing to unprecedent large EM volume, comprehensive fiber anatomy of the mouse cochlear amplifier was accessible. Several interesting features in fiber morphology were found in the OSB.
First, turning of the type-2 SGN fibers towards the cochlear base occurs around the OPCs (Figure 2(c)) instead of, as generally thought, after entering the OSB region. This may suggest that OPC is the potential on-site where type-2 fiber receive its guidance cue for turning. Second, the observed biphasic trajectory of the fibers along DCs indicates two distinct functional compartments, namely the climbing and the synapsing region (Figure 2(e)). The anchor-like structures on DCs (Figure 2(d)) not only secure surface attachment of the type-2 fibers but also seem to position them in a wellorganized manner, distributing fibers with partially overlapped synapsing regions underneath their presynaptic partners, the OHCs. Considering almost row-specific innervation of the type-2 fibers (Figure 2(i)), inter-row difference in the fraction of overlapped fibers is most likely the structural basis for heterogeneous afferents on OHCs (Figure 2(h)). If that is the case, an inter-DC coordination is expected for regulated fiber positioning, presumably involving DC gap-junction and OHC-activity-dependent cue during development.
Similar to afferents, more efferent terminals were found on OHCs of the first two rows (Figure 3(c)). This result, consistent with prior studies in guinea pig (Altschuler et al., 1984;Hashimoto & Kimura, 1988) and cat (Liberman et al., 1990;Thiers et al., 2008), implies a conserved radial gradient of centrifugal modulation of OHC motility across mammalian cochleae, which decreases from the first to the third OHC row. Further analysis based on classified MOC fibers revealed that this innervation bias was created by a minor population with multiple tunnel-crossing fibers and strong row-specific innervation ( Figure 3(d)). As highly branched MOC fibers were found originate exclusively from ipsilateral brainstem in rat (Warr & Boche, 2003), our data suggest contra-and ipsilateral MOC fibers differ in OHC innervation pattern and thereby may modulate BM motion in different ways.
However, it remains to elucidate the exact role of this feature for binaural loudness balancing (Froud et al., 2015) in complementing direct gain control of type-1 SGNs by the lateral olivocochlear efferent fibers (Darrow et al., 2006;Maison et al., 2013).
In conclusion, large-scale 3D EM reconstruction of mouse cochlear amplifier is presented in this study. It provides structural insights for the inter-row heterogeneity in morphology of OHC-DC complex as well as gradient innervation of the type-2 SGNs and the MOC fibers, although possible implications of these structural observations remain to be determined by future studies.