How to Bury the Dead: Elimination of Apoptotic Hair Cells from the Hearing Organ of the Mouse

Abstract

Hair cell death is a major cause of hearing impairment. Preservation of surface barrier upon hair cell loss is critical to prevent leakage of potassium-rich endolymph into the organ of Corti and to prevent expansion of cellular damage. Understanding of wound healing in this cytoarchitecturally complex organ requires ultrastructural 3D visualization. Powered by the serial block-face scanning electron microscopy, we penetrate into the cell biological mechanisms in the acute response of outer hair cells and glial-like Deiters’ cells to ototoxic trauma in vivo. We show that Deiters’ cells function as phagocytes. Upon trauma, their phalangeal processes swell and the resulting close cellular contacts allow engulfment of apoptotic cell debris. Apical domains of dying hair cells are eliminated from the inner ear sensory epithelia, an event thought to depend on supporting cells’ actomyosin contractile activity. We show that in the case of apoptotic outer hair cells of the organ of Corti, elimination of their apices is preceded by strong cell body shrinkage, emphasizing the role of the dying cell itself in the cleavage. Our data reveal that the resealing of epithelial surface by junctional extensions of Deiters’ cells is dynamically reinforced by newly polymerized F-actin belts. By analyzing Cdc42-inactivated Deiters’ cells with defects in actin dynamics and surface closure, we show that compromised barrier integrity shifts hair cell death from apoptosis to necrosis and leads to expanded hair cell and nerve fiber damage. Our results have implications concerning therapeutic protective and regenerative interventions, because both interventions should maintain barrier integrity.

Electronic supplementary material

The online version of this article (doi:10.1007/s10162-014-0480-x) contains supplementary material, which is available to authorized users.

INTRODUCTION

Upon trauma and during normal cellular turnover, epithelial integrity is maintained by the extrusion of whole dying cells. In some epithelia, only the cell’s apical portion is extruded and the remaining cell body degenerates apoptotically within the epithelium. The resulting apoptotic debris is cleared by migrating professional or tissue-resident nonprofessional phagocytes (Parnaik et al. 2000; Monks et al. 2005). Cellular extrusion and the maintenance of the protective surface barrier have been shown to depend on the actomyosin contractile mechanisms of adjacent cells and, in some cases, on the force originating from the apoptotic cell itself (Rosenblatt et al. 2001; Sonnemann and Bement 2011; Wang et al. 2011; Kuipers et al. 2014).

The mammalian auditory sensory epithelium, the organ of Corti of the cochlea, comprises the mechanosensory hair cells responsible for detection, transmission, and amplification of sound energy. In addition, this organ contains different types of supporting cells. These cells possess rigid actin and microtubule cytoskeletons that provide structural support essential for normal hearing function (Raphael and Altschuler 2003). Supporting cells are also key cellular players in hair cell regeneration and in epithelial repair. As opposed to the nonmammalian inner ear sensory epithelia, supporting cell-based hair cell regeneration does not occur in the organ of Corti, perhaps at the expense of its morphological complexity (Burns and Corwin 2013). Noise, otototoxic drugs and the effects of aging are the most common triggers of hair cell death. Of the different types of supporting cells, Deiters’ cells (DCs) have a pivotal role in repair after hair cell loss. Prompt sealing of the epithelial surface by DCs prevents expansion of damage by limiting the entry of the potassium-rich endolymph into the organ (McDowell et al. 1989; Monzack and Cunningham 2013; Oesterle 2013).

Compared with simpler epithelia, unique structural features limit wound healing in the organ of Corti. Surface closure is complicated by the actin-rich cuticular plates of hair cells and by the F-actin belts of DCs, both of which potentially limit apical contractile forces. Moreover, cell debris clearance within the epithelium is hampered by the large fluid-filled lumens and by a complex cellular organization. Understanding how hair cells die and how they interact with DCs might facilitate the development of protective interventions and broaden the understanding of the reasons underlying the absence of hair cell regeneration in mammals.

Ultrastructural analysis of wound healing in the organ of Corti by transmission electron microscopy (TEM) is limited to two-dimensional projection images of thin sections, neglecting information from the third dimension. A three-dimensional (3D) view can be achieved by laborious manual work with serial section TEM, as has been done to reveal the synaptic organization below outer hair cells (OHCs) (Fuchs et al. 2014) or by modern 3D imaging methods. The serial block-face scanning electron microscopy (SBEM) provides a streamlined 3D data acquisition process, producing TEM-like images with perfect image stack alignment in an automated fashion (Denk and Horstmann 2004). We have here used SBEM for ultrastructural 3D modeling of acute wound healing in the organ of Corti after ototoxic trauma.

MATERIALS AND METHODS

Animals

Embryonic day 18 (E18), postnatal day 10 (P10), and P22 NMRI mice were used to study the development of the cytoarchitecture of the organ of Corti. For timed pregnancies, the day of observation of the vaginal plug was taken as E0. The day of birth was assigned as P0. For ototoxic lesioning, NMRI mice and the Cdc42^loxP/loxP; Fgfr3-iCre-ER^T2 mice homozygous for the floxed Cdc42 allele were used. In addition, from the same crossings, mice carrying one wild-type Cdc42 allele were used as control animals. The generation and genotyping of Cdc42^loxP/loxP; Fgfr3-iCre-ER^T2 and Cdc42^loxP/wt; Fgfr3-iCre-ER^T2 mice has been previously described (Wu et al. 2006; Young et al. 2010; Anttonen et al. 2012). This mouse line was maintained in a mixed background. Both females and males were used in the analyses. All animal work has been conducted according to relevant national and international guidelines. Approval for animal experiments has been obtained from the National Animal Experiment Board.

Tamoxifen-Induced Recombination in Auditory Supporting Cells

Tamoxifen (Sigma, T5648) was prepared and intraperitoneally injected (50 μg/g body weight) once daily between P2 and P4 as previously described (Anttonen et al. 2012). That study also described the characteristics of Fgfr3-iCre-ER^T2-mediated recombination. Briefly, the regimen used led to efficient iCre-mediated recombination in the majority (>95 %) of DCs and pillar cells, in addition to recombination in some Hensen’s cells of the organ of Corti. Inner hair cells and OHCs were unrecombined, except for a few OHCs in the most apical and basal parts of the cochlea duct.

Ototoxic Lesion

OHC loss was induced at P22 by a single subcutaneous injection of 1 mg/g kanamycin (Sigma) in phosphate-buffered saline (PBS), pH 7.4, followed 45 min later by intraperitoneal injection of 0.4 mg/g furosemide (Fresenius Kabi) as previously described (Oesterle et al. 2008; Taylor et al. 2008; Anttonen et al. 2012). Mice were killed 36 and 48 h after the ototoxic challenge.

Whole Mount Specimens for Confocal Microscopy

Cochleas were fixed by perilymphatic perfusion with 4 % paraformaldehyde (PFA) in PBS, followed by immersion in the fixative for 8 h. The organ of Corti was dissected from the cochlear tissue, and tectorial membrane was removed. Full-length organ of Corti was divided into upper, medial, and basal parts. For immunofluorescence, whole mounts were blocked for 30 min with 10 % donkey serum in PBS containing 0.25 % Triton-X-100 (PBS-T), followed by incubation overnight at +4 °C with the mouse monoclonal synapsin-1 antibody (1:200; clone 223, Affiniti Research/Enzo Life Sciences). Secondary antibody made in donkey and conjugated to Alexa 594 was used. Following antibody incubations, F-actin filaments were visualized using Oregon Green 488 phalloidin (1:400 in PBS-T) for 30 min at RT. ProLong Gold anti-fade reagent with or without DAPI was used for mounting (all from Molecular Probes/Invitrogen). Confocal images were acquired using a Leica TCS SP5 laser scanning microscope with Plan Apochromat 63×/1.3 NA glycerol objective. The acquisition software was Leica LAS AF. Z-projections were processed with ImageJ (NIH). Both 36 and 48 h after administration of ototoxins, three nontraumatized and three traumatized cochleas (P22) were fixed and prepared for whole mount confocal analysis.

Histological Sections

Three nontraumatized and three traumatized cochleas (P22) were perilymphatically fixed with 4 % PFA and decalcified in 0.5 M EDTA, pH 8.0, for 12 h at +4 °C. The specimens were embedded into paraffin and cut to 5-μm-thick sections. Epitopes were unmasked by microwave heating (800 W) in 10 mM citrate buffer, pH 6.0, for 10 min. Sections were blocked for 30 min with 10 % goat serum in PBS-T, followed by overnight incubation at +4 °C with the rabbit polyclonal NKCC1 antibody (Kurihara et al. 1999; 1:5,000 in PBS-T). Detection was done with the Vectastain Elite ABC kit and diaminobenzidine substrate (DAB Detection kit, both from Vector Laboratories). Sections were counterstained with 3 % methyl green and mounted in Permount (Fisher Scientific). Sections were analyzed with a BX61 microscope (Olympus). Images were acquired through the DP70 CCD color camera and cell ‘F software (Olympus) and processed using Adobe Photoshop CS4 (Adobe Systems).

Sample Preparation for SBEM

Cochleas (E18, P10, P22, and P22 plus 36/48 h post-trauma) were rapidly dissected out from the heads of decapitated mice and fixed by perilymphatic perfusion with 2.5 % glutaraldehyde in 0.1 M phosphate buffer, pH 7.4. Specimens were immersed in the fixative overnight at +4 °C. For microdissection of the organ of Corti, the bony (P10 and P22) cochlear capsule was removed piece by piece starting from the apex and working toward the base of the cochlear duct with self-modified fine tweezers. The modiolus and the cochlear nerve were removed with a needle. The Reissner’s membrane was cut open and the exposed stria vascularis was cut lengthwise, so that a small ledge remained at the border of the organ of Corti. This ledge was later used to determine the orientation of the specimen during mounting. The tectorial membrane was removed with fine tweezers. The coiled organ of Corti was divided into three pieces with a needle to prevent sample folding and tearing during the following steps. Regions from the cochlea representing acute (36 h post-trauma, upper medial coil), late (36 h post-trauma, lower medial coil), and recovery (48 h post-trauma, medial coil) lesion sites (see Results) were selected under DIC optics (Olympus BX61 microscope). Also in the case of E18, P10, and P22 nontraumatized cochleas, the medial coil was used for analysis.

Selected pieces were processed in 2 ml Eppendorf tubes for embedding as previously described (Wilke et al. 2013). In brief, after fixation, specimens were washed extensively with Na-cacodylate buffer, pH 7.4, supplemented with 2 mM CaCl₂, and subjected to double-osmication comprising incubations with 2 % OsO₄–1.5 % K₄[Fe (CN)₆] in Na-cacodylate/CaCl₂ buffer for 1 h on ice, 1 % thiocarbohydrazide for 20 min at RT and 2 % OsO₄ for 30 min at RT. To increase the contrast, specimens were then incubated in 1 % uranyl acetate overnight at +4 °C and in Walton’s lead aspartate solution for 30 min at +60 °C. In between each step, specimens were washed 5 × 3 min with distilled water. Next, specimens were dehydrated using increasing gradient of ethanol starting from 20 %, and ice cold acetone as transitional solvent, and finally embedded in Durcupan ACM resin (Fluka, Sigma-Aldrich). Durcupan infiltration was done as follows: First, 2 h in 30 %, 2 h in 50 %, and overnight in 75 % mixture of Durcupan w/o accelerator at RT, followed by two times 2 h in fresh mixture w/o accelerator at +50 °C. Finally, specimens were transferred to full Durcupan mixture containing accelerator and incubated first at +50 °C for 2 h, then mounted in a desired orientation between two objective slides separated by 1 mm thick silicon spacers, and polymerized in a +60 °C oven for 48 h.

SBEM Data Acquisition

Excess resin around the specimen was trimmed away, and the tip of the pyramid was then cut from the block with a razor blade and mounted on a 3View pin using conductive cyanoacrylate glue (CircuitWorks, Chemtronics). The sides of the pyramid were covered with silver paint (Agar Scientific Ltd.) and the pin was coated with a 5-nm-thick layer of platinum in a sputter coater (Quorum Q150TS; Quorum Technologies) to further enhance conductivity.

Specimens were imaged on an FEI Quanta 250 FEG SEM equipped with a Gatan 3View system (Gatan, Inc.). The imaging was conducted at low vacuum to prevent charging; the chamber pressure was kept at 0.2–0.3 Torr (26–40 Pa). The accelerating voltage was set to 2.5 kV and spot size to 3.0 to get the best signal to noise ratio. Decrease of accelerating voltages below 2.5 kV was also tested, but it resulted in increase of noise and was therefore not used. On the other hand, increase of accelerating voltages above 2.5 kV is not recommended due to increase of the excitation depth within the sample above 40 nm, which was the typical z-step used during cutting. The final lens aperture was 30 μm, the pixel dwell time was 15–20 μs. The cutting z-step size was selected to 40 nm, which we found suitable for modeling of organelles. The diamond knife was oscillating during cutting, and cutting speed was 0.1 mm/s. The specimens were imaged with the final pixel size around 20 nm (twice less than the z-step). Before and after each volume was collected, a low magnification (∼200–300×) image of the block face was collected to map the collected dataset within the whole organ of Corti. The images were stored in Digital Micrograph (.DM4) format that, in addition to images, contains all corresponding meta-tags with parameters of the run, such as pixel size, dwell time, and chamber pressure conditions.

Image Processing for SBEM

Once a volume was collected, the images were batch converted to 16-bit TIFs using a custom script written under Gatan Digital Micrograph software. The purpose of the script was to convert DM4 to TIF with simultaneous extraction of all meta-tags and saving them into a separate header file, so that the parameters may be easily accessed later. The 16-bit TIFs were further processed in Microscopy Image Browser (MIB), using a MatLab-based, self-developed, freeware, open source software for image processing, segmentation, and analysis of multidimensional datasets. During the initial steps of image processing, the individual TIF-files were assembled in 3D stacks, which sometimes included stitching and alignment of several individual data fragments. Furthermore, the histograms for the slices throughout the volume stack were normalized to correct the drift in image intensity during acquisition. In most cases, a masked normalization was used instead of a global one. During mask-based normalization, the coefficients used to normalize the dataset are calculated from the masked areas only and are then applied globally for the whole dataset. This approach eliminates normalization artifacts that appear on slices that have a lot of background. After normalization, the contrast was adjusted to make the structures of interest as good as possible, and the stacks were converted to 8-bit format. The voxel sizes and coordinates of the 3View stage were used to define bounding box information for each dataset. The bounding box information is an important step in the routine, because it allows placing each dataset to exactly right part of the global 3D space. This minimizes the need for further alignment of individual datasets between each other. The processed data was stored in Amira Mesh format, while the original DM4 files were compressed and backed up to the RAID storage.

3D Reconstruction and Analysis of SBEM Data

The images were first processed and segmented using Amira (Visage Imaging, Inc.) and Microscopy Image Browser and further rendered in Amira. Shortly, segmentation was carried out by selecting object contours from different levels by hand or by threshold selection, followed by interpolation. Especially, cellular structures exposed to the lumenal spaces were easily segmented by local threshold selection against lumenal background. Mistakes produced by automated segmentation tools were corrected by hand. 3D surfaces were rendered from the segmented objects and are presented in shaded or partly transparent mode. Used colors were selected so that they are easily identifiable from overlays with single TEM-like images. The general analysis and comparison between datasets were done using ObliqueSlice module of Amira. Despite all the efforts for uniform orientation of specimens, there was always a difference. The ObliqueSlice module allows the display of arbitrarily oriented slices through a 3D volume, which in general allows comparision of datasets with different orientation. All images are presented in orthographic mode, allowing quantitative length measurements.

Two SBEM specimens from the medial coil of the cochlea were analyzed at E18, P10 and P22 each. Two specimens from both the upper medial and lower medial coil at 36 h post-trauma (acute and late lesion sites) and two specimens from the medial coil at 48 h post-trauma (recovery lesion site) were studied. In addition, two specimens from the medial coil of cochleas of Cdc42^loxP/loxP; Fgfr3-iCre-ER^T2 mice at 36 h post-trauma as well as one specimen from the same cochlear region of Cdc42^loxP/wt; Fgfr3-iCre-ER^T2 mice were studied by SBEM.

RESULTS

Development of the Cytoarchitecture of the Organ of Corti Revealed by SBEM

Cytoarchitectural complexity of the mature organ of Corti separates it from other types of mammalian and nonmammalian inner ear sensory epithelia. In rodents, structural maturation takes place early postnatally, just before the onset of hearing (Souter et al. 1997). We hypothesized that the unique structural features of the mature organ of Corti bring about specializations in wound healing. To study this, we first compared structural relationship between OHCs and DCs in the late-embryonic (E18), juvenile (P10) and mature (P22) mouse organ of Corti using SBEM and 3D modeling (Figs. 1A–E and 2A–L).

FIG. 1.

Cytoarchitectural development of the organ of Corti. A–D SBEM imaged volumes with 3D-modeled OHCs and DCs of the organ of Corti at E18 (A) and P10 (C). Bending of DC phalangeal processes (open arrows) early postnatally creates distinct apical (arrowheads) and basal domains (arrows), which contact different OHCs. The basal domain of DCs differentiates into a cup-like structure. Surface views (B, D) of the imaged volumes reveal the spatial relationship between OHCs and DCs. The surface of a DC that basally contacts an OHC and the surface of this OHC are marked by identically colored spheres. E Schematic cross-sections through the organ of Corti at birth and adulthood. Schematic representations of developmental changes in DC morphology and the contact areas between these cells and OHCs. Abbreviations: T tunnel of Corti, N space of Nuel, IHC inner hair cell, IP inner pillar cell, He Hensen’s cell. Scale bars, 10 μm.

Segmented data revealed the well-characterized mosaic-like organization of the three rows of OHCs and supporting cells at the epithelial surface, the reticular lamina. This organization was evident already at late embryogenesis (Fig. 1B, D). 3D models showed that late-embryonic OHCs and DCs have a perpendicular orientation with respect to the epithelial surface (Fig. 1A; Video 1), as do cells of simpler epithelia (Guillot and Lecuit 2013). This organization changes dramatically during early postnatal development when the phalangeal processes of DCs elongate and bend lengthwise, as shown at P10 (Fig. 1C). In addition, a cup-like subdomain that encapsulates the basal part of an OHC differentiates from the DC soma (Fig. 1C, Video 2). The DC cup docks the OHC of the nearest row (e.g., 1. row OHC-1. row DC), but due to phalangeal bending, these cells are not apically rejoined. Rather, the DC reaches the reticular lamina farther away and makes apical contacts with different OHCs (Raphael and Altschuler 1991; Fig. 1C–E). In other types of inner ear sensory epithelia, a straight supporting cell can contribute to the elimination of an adjacent OHC throughout its full length. Thus, unique to the mature organ of Corti, the apical surface closure and the basal clearance of a given dying OHC are spatially regulated by different DCs.

3D models of developing and mature OHCs together with sections from different height of the respective OHCs (Fig. 2A–L) show the prominent elements of structural differentiation: (1) increase in cell length, (2) differentiation of the apical mechanotransduction domain that comprises the stereociliary bundle and cuticular plate, (3) changes in the shape and localization of mitochondria, and (4) changes in the circumferential shape of OHCs. Important for the present study is the opening of fluid-filled spaces within the organ of Corti, leading to the separation of lateral plasma membranes of OHCs and DCs (Fig. 2J, K), a fact that adds further complexity to wound healing by limiting cellular contacts.

FIG. 2.

Morphological maturation of outer hair cells. A, B Side, top–down and down–top views of 3D modeled E18 (A) and P22 (B) OHCs from SBEM datasets. In the apical domain, note the increase in the length of stereocilia, formation of the cuticular plate, and disappearance of the primary cilium, the kinocilium, by adulthood. Note also the relocalization and shape changes of mitochondria along maturation. In the basal domain, the amount of mitochondria beneath the nucleus decreases by adulthood. C–G, H–L Single images of the respective hair cells at the plane parallel to the apical surface. Section levels are indicated by arrows in (A, B). Note the stereociliary bundle maturation (C, H), cuticular plate formation (D, I), change of the OHC circumferencial shape from pentagonal shield like to round or rounded triangle (E, F, J, K), central-to-lateral relocation of mitochondria and tubular-to-round change in mitochondrial shape (D, E, F, J, K), and appearance of open spaces between OHC lateral plasma membranes (E, F, J, K), except at the level of the nucleus (G, L). Scale bar, 5 μm.

OHC Apoptosis and DCs Swelling Within the Sensory Epithelium

To study acute wound healing events in the mature organ of Corti, we exposed mice at P22 to severe ototoxic trauma, applying a protocol based on single injections of an aminoglycoside antibiotic and a loop diuretic drug. Synergism between these compounds triggers rapid, selective, and reproducible OHC loss that proceeds in a base-to-apex gradient along the cochlear duct (Oesterle et al. 2008; Taylor et al. 2008; Anttonen et al. 2012); 36 h after injection, the wave of OHC loss had proceeded to the middle coil of the cochlea. We term this region where approximately half of OHCs are lost as the acute lesion site, and the immediate lower part of this coil showing total OHC loss as the late lesion site. 48 h after ototoxic challenge, the middle coil of cochlea showed complete OHC loss and is termed as the recovery lesion site (Fig. 3A). To exclude the possible direct effects of loop diuretics on supporting cells, we confirmed that the Na⁺–K⁺–2Cl⁻ ion cotransporter (NKCC1), a primary target of loop diuretics (MacVicar et al. 2002), is not expressed in the nontraumatized (Crouch et al. 1997; Sakaguchi et al. 1998) or traumatized organ of Corti (Fig. 3B, B’).

FIG. 3.

Outer hair cell apoptosis following ototoxic trauma. A Schematic surface views of nontraumatized and traumatized organ of Corti, and the nomenclature used for the lesion sites. B, B’ Hematoxylin-stained paraffin section through a traumatized cochlea (B). Adjacent section shows that NKCC1 is not expressed in the organ of Corti but is found in the laterally located supporting cells, the Claudius cells, and in the stria vascularis (B’). C–G, H–L Single images at different heights of OHCs undergoing early and advanced apoptotic degeneration at the acute lesion site. Corresponding heights are marked by arrows in (M). Gradual shrinkage of the OHC body, mitochondrial delocalization from beneath the plasma membrane (arrowhead in F), and their degradation (K) are seen. Also, DC swelling and the resulting filling of the interstitial spaces is evident. M Side, top–down and down–top views of a 3D-modeled OHC (P22 + 36 h post-trauma) undergoing advanced apoptotic degeneration show prominent cell body shrinkage, but a largely intact apical domain (compared with Fig. 2B). Minor fusion of stereocilia can be seen (open arrows in H, M). Abbreviations: SV stria vascularis, OC organ of Corti, Cl Claudius cells, T tunnel of Corti. Scale bar, 5 μm.

By comparing the acute, late and recovery lesion sites, we aimed to ultrastructurally characterize the sequence of OHC death and wound healing performed by DCs (Fig. 3C–M). At the acute lesion site, remaining OHCs (n = 22 OHCs in SBEM datasets) showed variable degree of degeneration. Stereociliary bundles of both early and late apoptotic OHCs appeared surprisingly intact, apart from minor fusion and absence of cilia (Fig. 3C, H, M). Also the cuticular plate was intact (Fig. 3D, I, M), except for the appearance of vacuoles underneath it, at the pericuticular level (Fig. 3E, J), in agreement with prior studies (Ylikoski 1974; Raphael and Altschuler 1991). However, apoptotic OHC bodies showed progressive shrinkage coupled with the loss of the circumferential cell shape (Fig. 3E, F, J, K, M). Shrinkage of the cell body of OHCs with an intact apical domain was also evidenced in phalloidin-labeled whole mount specimens analyzed by confocal microscopy (data not shown). In early apoptotic OHCs (7/22), mitochondria had delocalized from beneath the lateral cell membrane and were swollen (Fig. 3E, F). The strongly shrunken OHCs of the late apoptosis stage (15/22) contained tightly packed and degraded mitochondria as well as other cell organelles and vacuolar structures (Fig. 3J, K). Nuclei of apoptotic OHCs showed condensed chromatin (Fig. 3G, L), an additional feature that separated these cells from age-matched undamaged hair cells (Fig. 2B, H–L).

We next focused on the interaction between apoptotic OHCs and adjacent DCs (Fig. 4A–H). Fluid-filled spaces are clearly visible around the lateral cell membranes of OHCs in the undamaged sensory epithelium (Figs. 2J, K and 4A, F). At the acute lesion site, phalangeal processes of DCs became distinctly swollen towards shrunken OHCs and filled the interstitial spaces (Fig. 4B, C, G). The swelling appeared concomitantly with early signs of OHC degeneration (Fig. 3E, F), but it did not produce constricting force towards OHC bodies. For example, the most strongly shrunken area of OHCs at the pericuticular level remained free of DC contacts (21/22, Fig. 4B, C). Thus, no DC protrusions towards degenerating OHCs could be seen that would indicate active OHC excision by external contractile forces. Together, our data show that, upon the ototoxic trauma, degeneration of OHC bodies has characteristics of apoptosis. Notably, degeneration of the OHC body and swelling of adjacent DC phalanges preceded major structural changes at the apical surface of the organ of Corti.

FIG. 4.

Elimination of outer hair cells within the organ of Corti. A The nontraumatized epithelium shows distinct interstitial spaces between OHC’s lateral plasma membranes. B, C Two images from the same orientation showing swelling of the DC phalanges at the acute lesion site and filling of interstitial spaces. No indications of protrusive activity by DCs can be seen at the pericuticular level (thin arrows in A–C). Note that in (B), the OHC on the right side appears to be in two portions, but in fact its intact middle portion bends out from the image plane, showing the importance of 3D analysis for reliable documentation. Apoptotic body-like debris and a degrading nucleus (arrows) are seen inside the swollen DC phalanges. D A higher magnification view of these engulfed apoptotic cell particles (arrows). In addition, apoptotic bodies waiting to be internalized are seen between swollen DC phalanges (arrowhead). E At the late lesion site, lumenal spaces re-open along with partial recovery of the volume of DC phalanges. F–H Lumenal re-opening at the late lesion site is also evident in sections viewed at the plane parallel to the epithelial surface. The OHC rows are numbered. Necrotic-like cell debris (asterisks) persists in lumens. Note the OHC (nucleus marked by open arrow) of the first row that is hanging free inside the lumen. The inset shows this cell in another orientation (G). Scale bar, 5 μm.

Swollen Phalangeal Processes of DCs Are Responsible for Apoptotic OHCOHC Clearance

At the early lesion site, electron-dense particles of variable size were captured by swollen phalangeal processes of DCs (n = 18) at the site of lost OHCs. Some of these particles contained degrading organelles and resembled apoptotic bodies. Even a whole OHC nucleus was found to be engulfed by a DC phalangeal process (Fig. 4C, D; Video 3). This engulfment lacked features characteristic for professional phagocytes. Rather, the sessile nature of DCs, the lack of lamellipodial extensions and membrane ruffles around apoptotic OHCs and the small size of phagocytosed material are features described for nonprofessional phagocytes (Parnaik et al. 2000). Apoptotic body-like structures were not anymore found at the late lesion site, suggesting for prompt apoptotic OHC clearance (Fig. 4E, H). Interestingly, phalangeal swelling decreased concomitantly with successful clearance of engulfed material (Fig. 4E, H). Migrating phagocytes were not found in any organ of Corti specimens studied.

Phagocytosed Apoptotic Debris Stands Out from Necrotic Debris

In addition to the apoptotic material processed by phagocytosis, large quantities of floating cellular debris were seen within the large lumens, the tunnel of Corti and the space of Nuel, both of which remain open at the acute lesion site. This debris was smaller sized and less electron dense than apoptotic body-like material. It comprised ruptured membranes and cell organelles, but intact nuclei were not seen (Fig. 4G). Similar debris has been earlier reported following ototoxic damage and has been suggested to represent necrosis or, more likely, secondary necrosis of OHCs (Taylor et al. 2008). Secondary necrosis is an autolytic cell degeneration process following apoptosis that occurs when the apoptotic cell is not cleared by phagocytes (Silva 2010). At the acute lesion site, 3D dataset allowed us to track the origin of this cellular debris mainly to the first OHC row. At this location, degenerating OHCs displayed an intact apical portion and a shrunken cell body, sometimes even hanging free in the fluid-filled lumen (Fig. 4G). Because of the location close to the tunnel of Corti and the lack of surrounding DCs, the fate of these apoptotic OHCs is likely secondary necrosis. In contrast to apoptotic body-like material, floating debris was still seen at the late lesion site (Fig. 4H).

Surprisingly, when analyzing the late lesion site, one OHC lacking the apical domain was found to have survived the lesion. This bundleless OHC was not shrunken and had unswollen mitochondria, but its nucleus showed signs of early degeneration (Fig. 5A–C; Video 4). These findings show that OHCs can survive for a short time period as bundleless cells in vivo, albeit this appears to be a rare event. At the same lesion site, necrotic cellular debris was found within two DC cups, suggesting that this debris originates from bundleless OHCs fated to secondary necrosis (Fig. 5A, D). Collectively, these data provide direct evidence of engulfment of apoptotic OHC fragments by DCs and show the function of DCs as tissue-resident phagocytes. The data also suggest for the contribution of secondary necrosis in cases where apoptotic OHCs are not surrounded by swollen DC phalanges.

FIG. 5.

Rare outer hair cells persist as bundleless cells after lesion. A Overview of the late lesion site where a bundleless OHC (blue) was found. Note also necrotic-like debris in a Deiters’ cell cup (asterisk) in the proximity to the bundleless OHC. B A single section shows that this cell is innervated by nerve fibers (arrow), demonstrating that it is not a transdifferentiated supporting cell. C A 3D reconstruction of bundleless OHC. Note the collapsed appearance of the cell body and the mitochondria with no overt morphological changes. D A single image section showing necrotic-like cell debris in the Deiters’ cell cup. This debris might originate from the bundleless OHC degenerating through secondary necrosis. Scale bar, 5 μm.

Disappearance of the DCs Cup and Nerve Terminals Upon OHC Apoptosis

We next focused on the role of the cup of DCs in wound healing, a poorly characterized cellular domain. It does not exist in supporting cells of other types of inner ear sensory epithelia. DC cups could be thought as major processing sites of apoptotic OHC debris due to their direct contact with OHCs. However, following ototoxic challenge, DC cups did not swell and engulf OHC debris, in contrast to phalangeal processes of the same cells.

In addition that the DC cup docks the OHC base, nerve terminals synapsing on the OHC were found to be clustered inside the cup, suggesting that the cup physically supports nerve endings as well (Fig. 6A, B; Video 5). After OHC loss, nerve terminals persisted for a short time in the cup, but they disappeared along with the cup closure (Fig. 6C–F). Notably, a part of neuronal contacts on OHCs were lost already during early stages of apoptosis of these cells. Quantification from 3D reconstructions showed on the average 5.9 ± 0.5 terminals per OHC in undamaged specimens (mean ± SEM, n = 10 OHCs). This value was 2.0 ± 0.3 per degenerating OHC at the acute lesion site (mean ± SEM, n = 12 OHCs). At the recovery lesion site, the majority of nerve terminals were lost from the area of former DC cups. This decrease in the number of nerve endings was confirmed by synapsin-1 immunolabeling in whole mount specimens under confocal microscopy (Fig. 7A–C).

FIG. 6.

Closure of the Deiters’ cell cup and loss of synaptic terminals following ototoxic lesion. 3D models and corresponding single sections of the DC cup and associated synaptic terminals in an undamaged specimen (A, B) and at the acute (C, D) and recovery (E, F) lesion sites. While the DC cup normally fully enfolds the OHC base and clusters nerve terminals, the cup closes at the acute lesion site and loses synaptic contacts. This closure starts after detachment of the dying OHC. Note the dissociation of the cell membrane from the degrading actin plaque (asterisk in D). Note also progressive straightening of the phalangeal process (compare A and E). When the cup closure is complete at the recovery site, nerve terminals are lost and the cup’s actin plaque has disappeared (F). Open arrows mark the electron-dense F-actin plaque. Scale bar, 5 μm.

FIG. 7.

Cytoskeletal changes in Deiters’ cells and alterations in innervation density to outer hair cells upon ototoxic trauma, as revealed by confocal microscopy. Confocal Z-stack projections of DC cups from a nontraumatized specimen (A, A”) and from acute (B, B”) and late (C, C”) lesion sites show merged and single z-projections of phalloidin, synapsin-1 and DAPI labeling in the basal domain. Merged images (A–C) show the loss of synapsin-1-immunostained nerve terminals, evident already at the acute lesion site. Note the prominent actin plaque (arrowheads) of DC cups in the control specimen and the gradual degradation of the plaque upon trauma (A’–C’). A part of DAPI-labeled OHC nuclei are shrunken and fragmented (open arrows), indicating a late stage of apoptosis (A”–C”). D–F Confocal Z-stack projections of phalloidin-labeled reticular lamina show the apical domain. The dynamic remodeling and reinforcement of new surface extensions by F-actin at the site of lost OHC can be seen by comparing the intensity of phalloidin labeling at acute and late lesion sites. Insets show these events on single surface extension sites. Note also the progressive flattening of the sensory epithelium and collapse of supporting cells (arrows in B’, C’). Abbreviation: He Hensen’s cell, OHC outer hair cell, DC Deiters’ cell. Scale bar, 10 μm.

In undamaged cochleas, an electron-dense F-actin plaque is present in the cup cytoplasm, a structure that likely promotes the cup’s function as a docking site for an OHC (Figs. 1C and 6A, B). Interestingly, trauma-induced closure of the DC cup was not accompanied by remodeling of its F-actin plaque, as evidenced by SBEM dataset (Fig. 6D, F) and by confocal imaging of phalloidin-labeled whole mount specimens (Fig. 7A’–C’). This is a major difference to the F-actin network in the apical domain of the same cells that undergo dynamic changes during surface closure, as shown by phalloidin labeling (Fig. 7D–F) and SBEM images (Fig. 8A–H).

FIG. 8.

Epithelial surface closure is reinforced by F-actin belts. A, B Surface rendering from the acute lesion site shows variation in the geometry of DC junctions (arrows) that close the sites of lost OHCs (A). This variation is minimal at the late lesion site, showing that dynamic junctional remodeling takes place after the initial surface closure (B). C–E’ Single block-face images and 3D models of DCs of the second row from an undamaged specimen (C, C’) and from the acute (D, D’) and late (E, E’) lesion sites. In the undamaged specimen, note that apical junctions bend at some sites and are not perpendicular to the epithelial surface. The frames of apical extensions formed upon hair cell loss are depicted (C, C’). At the acute lesion site, junctional extensions reach the sites of lost OHCs but F-actin recruitment to these extensions is minimal. The original F-actin belt is distinct (D, D’). At the late lesion site, actin recruitment to extensions has taken place and the original belts have partly disassembled (E, E’). Note that the tricellular junctions are devoid of strong actin recruitment (thin arrows in E’). Scale bar, 5 μm.

Apical Surface Closure Is Reinforced by New F-actin Belts

In the organ of Corti, apical extensions and new F-actin belts of DCs reseal the places of lost OHCs through a process referred as scar formation (Forge 1985; Raphael and Altschuler 1991; Leonova and Raphael 1997). The current SBEM analysis revealed that DC apical extensions undergo dynamic reorganization at the acute lesion site, leading to more uniform shape of scars (Fig. 8A, B). At this lesion site, F-actin accumulation to the extensions was minimal and the original, rigid actin belts were still clearly visible (n = 16 scars in SBEM datasets). At the late (n = 31 scars) and recovery (n = 20 scars) lesion sites, remodeled DC extensions had acquired new F-actin belts and original belts were partly disassembled (Figs. 7D–F and 8C–E’). Thus, these results showed that the thick F-actin belts cannot close the surface as a contractile apparatus. Instead, the apical extensions of DCs are reinforced by new F-actin belts following surface closure.

Impaired Apical Surface Closure Triggers Necrotic Hair Cell Death

By analyzing the Cdc42^loxP/loxP; Fgfr3-iCre-ER^T2 mutant mice, we have previously shown that inactivation of the Rho GTPase Cdc42 impairs structural maturation of the actin-rich apices of auditory supporting cells (Anttonen et al. 2012). This is consistent with the fact that Cdc42 is a major regulator of actin dynamics (Ridley 2006). We also showed that this apparent locking perturbs actin remodeling-based surface closure following ototoxic trauma (Anttonen et al. 2012). Herein, by studying traumatized cochleas of these Cdc42 mutant mice at ultrastructural level, we sought to find out the impact of proper surface closure on the mode of OHC death.

At the acute lesion site, SBEM analysis revealed a highly distorted reticular lamina (Fig. 9A–C). Multiple sites of surface breaches, partly filled by cellular debris, were found at places of lost OHCs. Former apical junctions could still be seen at these sites, consistent with prior analysis by confocal microscopy (Anttonen et al. 2012; Fig. 9A, A’). Hensen’s cells, located laterally next to the third row of DCs, were found to close some surface breaches. The sealing of the surface was achieved by Hensen’s cells extensions traveling within the epithelium and finally reaching the surface area. In some locations, DC extensions at the epithelial surface could be seen, but these extensions were highly irregular and were only minimally reinforced by F-actin belts (Fig. 9A, A’). Collectively, these findings show that the initial surface closure, subsequent junctional remodeling and reinforcement by F-actin are severely impaired in Cdc42 mutant mice, leaving direct leakage sites for the endolymph into the sensory epithelium.

FIG. 9.

Compromised barrier integrity causes necrotic outer hair cell death. Analysis of the acute lesion site of the organ of Corti of Cdc42 ^loxP/loxP; Fgfr3-iCre-ER ^T2 mutant mice (P22 + 36 h) exposed to ototoxic trauma. A, A’ Reticular lamina of a mutant specimen displays two direct leakage sites (arrowheads), shown in different planes, at places where DCs have not established surface extensions. Note the maintenance of junctions that contacted the former OHC (arrows). B A cross-section through the sensory epithelium of a mutant animal shows one apoptotic (on the right side) and two necrotic OHCs (asterisks). Note the disintegrated stereociliary bundle of the necrotic hair cell, as opposed to the apoptotic cell. B’ Side and top–down views of a 3D-modeled necrotic OHC show prominent swelling of the cell body. Stereocilia are internalized by the bulging apical membrane (compared with nontraumatized OHC in Fig. 2B and to apoptotic OHC in Fig. 3M). C Necrotic OHCs burst and release their cellular content, causing formation of large cavities (asterisks) within the epithelium. The reticular lamina becomes highly distorted upon cell collapse and noneliminated cuticular plates are seen. D, E In the wildtype, traumatized organ of Corti, the acute lesion site still comprises nerve terminals (open arrows), despite closure of DC cups. In contrast, the mutant sensory epithelium lacks terminals and only longitudinal nerve fibers are left. The ototoxic protocol used does not cause major damage to inner hair cells (B, C). Abbreviations: He Hensen’s cell, cp cuticular plate, IHC inner hair cell, Mut mutant specimen. Scale bar, 5 μm.

In the traumatized Cdc42 mutant mice, we next analyzed the status of OHCs and innervating neurons, both of which are known to be vulnerable to endolymph leakage. Notably, although a few apoptotic OHCs were found (n = 2), the majority of remaining OHCs were distinctly swollen (n = 4), typical to necrotic cells. These necrotic OHCs showed degraded stereocilia, often internalized by the bulged apical cell membrane (Fig. 9B, B’). OHC swelling culminated in cell membrane rupture and release of cellular contents, including cell organelles and a swollen nucleus, all characteristics of necrotic death (n = 3 just ruptured OHCs). Necrotic OHC death was associated with the formation of wide cavities filled with cellular debris within the organ of Corti. This debris showed up as large, irregular, nonvesicular clumps, unlike the apoptotic body-like particles found in wildtype specimens. Interestingly, at places previously occupied by necrotic OHCs, some cuticular plates were still attached to the reticular lamina, without signs of surface closure activity (Fig. 9C). Only longitudinal nerve fibers, but no nerve endings could be seen at the acute lesion site (Fig. 9D, E). Collectively, these results give direct evidence that the compromised surface barrier causes severe damage to distal parts of nerve fibers and shifts the mode of OHC death from apoptosis to necrosis.

DISCUSSION

Due to the high prevalence of hair cell and hearing loss in humans, there is a great need for the development of a clinically relevant protective strategy. Pharmacological targeting of intracellular signaling cascades has provided attenuation of trauma-induced hair cell loss in experimental animal models, but the establishment of an effective therapy in humans is still to be realized. Understanding of hair cell death and wound healing mechanisms is required for therapy development. Obtaining this knowledge has been hampered by technical limitations to ultrastructurally analyze the complex 3D organization of the organ of Corti. The current study shows the potential of SBEM in this analysis. We show that degeneration of OHC bodies is initiated rapidly after ototoxic insult, well before any major signs of damage in the mechanotransduction domain at the cell surface. Thus, ultrastructural analysis within the organ of Corti is required to capture the early events of hair cell degradation.

The present results demonstrate classical features of apoptosis in OHCs challenged with ototoxic drugs, confirming prior TEM data (Taylor et al. 2008). The prominent shrinkage of OHCs was linked with the intrinsic apoptotic process, rather than with concomitant swelling of phalangeal processes of neighboring DCs. Phalangeal swelling was found to promote the function of DCs as phagocytes, as engulfed apoptotic body-like material was concentrated to these swollen processes. Phagocytosis by DCs has been previously suggested based on the localization of prestin-immunostained cellular debris to the trauma area. However, by light microscopy it has been challenging to characterize apoptotic debris and to localize it to DC cytoplasm (Abrashkin et al. 2006; Taylor et al. 2008). SBEM and 3D rendering allowed clear identification of hair cell particles in the DC phalanges. Phagocytosis by DCs was directed towards late apoptotic particles of OHCs, as opposed to professional phagocytes that can engulf even whole dying cells (Hochreiter-Hufford and Ravichandran 2013). DC cups lacked detectable phagocytotic activity, despite their direct contact with OHCs. The stout F-actin plaque in the cup cytoplasm may physically limit engulfment, while the swollen phalangeal process of the same cell is more accessible to apoptotic debris. Our data suggest that apoptotic OHCs of the first row are prone to secondary necrosis, as they lack sufficient contact with swollen phalanges. These findings highlight the importance of cellular environment in directing apoptotic hair cell clearance and suggest that cells lacking contact with tissue-resident phagocytes are prone to die by secondary necrosis, without prompt debris clearance.

DC phalangeal swelling was a rapid and transient response to ototoxic trauma. Swelling has been previously reported following ototoxic and noise traumas (Fredelius 1988; Flock et al. 1999; Ahmad et al. 2003; Abrashkin et al. 2006; Taylor et al. 2012; Forge et al. 2013). A possible trigger for this event is altered ion transport coupled with intercellular signaling between dying hair cells and supporting cells (Piazza et al. 2007; Forge et al. 2013). Thus, phalangeal swelling of DCs can serve as a protective ion buffering mechanism in the lesioned organ of Corti. Pillar cells lacked prominent swelling, suggesting that DC volume changes depend on ion transport mechanisms specific to this cell type.

In nonmammalian species, traumas often trigger extrusion of entire hair cells from the inner ear sensory epithelia. In some cases, only the apical portions of hair cells are extruded (Cotanche and Dopyera 1990; Gale et al. 2002; Hirose et al. 2004; Hordichok and Steyger 2007; Bird et al. 2010). Decapitation is a predominant mechanism of hair cell elimination in the adult organ of Corti (Forge 1985). Why whole cell extrusion is rarely if never seen in this epithelium? It might be antagonized by physical restrictions created by the apical, rigid actin cytoskeleton and the tallness of hair cells. Additionally, interstitial fluid-filled spaces within the organ of Corti potentially limit forces required for cell extrusion.

Live cell imaging in the chick vestibular organ has suggested that the apices of hair cells are excised at the pericuticular level by contractile actin cables emanating from adjacent supporting cells (Bird et al. 2010). Also, in the lesioned bullfrog vestibular organ in vitro, TEM analysis has described finger-like supporting cell protrusions beneath the cuticular plate (Gale et al. 2002). We did not capture the putative active excision in our SBEM specimens. One reason might be that this event is very rapid, as shown by live cell imaging (Bird et al. 2010). We found that OHC’s pericuticular region was prominently shrunken as a part of the cell’s apoptotic process, without concomitant signs of excision activity by adjacent DCs. Therefore, OHC itself might be responsible for its own cleavage through apoptotic shrinkage or, less likely, through intrinsic contractile forces. This possibility implies that the elimination of OHC head and the expansion of apices of DCs to reseal the epithelial surface are independent processes. In agreement, from recent live cell imaging in the mouse vestibular organ (Burns and Corwin 2014), it can be concluded that the cuticular plates of ototoxically challenged hair cells detach before initiation of surface closure (Video 2 in Burns and Corwin 2014). Also recent studies in vitro have suggested that dying epithelial cells contribute to their own elimination by intrinsic apical contraction (Wang et al. 2011; Kuipers et al. 2014). However, the finding in one SBEM specimen of a bundleless, largely nonapoptotic hair cell is surprising. Still in this case it remains to be shown if mildly damaged hair cells can independently cleave their apical domains or if this elimination is a consequence of the activity of adjacent supporting cells. Together, it is possible that both cell autonomous and nonautonomous mechanisms regulate hair cell elimination in the traumatized organ of Corti.

Following trauma-induced OHC loss, epithelial surface is closed by apical extensions and new F-actin belts of DCs, as shown in prior TEM studies (Forge 1985; Raphael and Altschuler 1991; Leonova and Raphael 1997). Our SBEM data bring novel insights into F-actin dynamics in this process. The results show that the junctional geometry of scars is rapidly remodeled to a uniform pattern and that the junctions are reinforced by new F-actin belts. We found that the original F-actin belts were slowly disassembled, concomitantly with gradual strengthening of new belts. These results on DCs of the organ of Corti are consistent with the recent results on the mouse vestibular organ, which demonstrated that thickened F-actin belts in supporting cells become less mobile with age, causing retardation of surface closure (Burns and Corwin 2014).

The rigid F-actin belts are thought to limit the capacity of mature mammalian vestibular supporting cells for hair cell replacement (Meyers and Corwin 2007; Burns and Corwin 2013, 2014; Burns et al. 2013). While cytoskeletal modulation of supporting cells might elicit regenerative capacity, structural dedifferentiation is likely to compromise surface barrier and, thus, cellular survival. This is most evident in the organ of Corti where a single DC has a dual function in providing support to the basal domain of an OHC and to the apical domain of other OHCs (Fig. 1B). Thus, it is unlikely that DCs can produce replacement hair cells without abandoning at least a part of these critical supportive functions.

By analyzing the Cdc42^loxP/loxP; Fgfr3-iCre-ER^T2 mice where Cdc42 is inactivated specifically in auditory supporting cells, we provide direct evidence that impaired barrier function triggers necrotic OHC death and damage of nerve endings following ototoxic insult. F-actin belts of Cdc42-depleted supporting cells were improperly developed, causing severe defects in the formation of new belts at places of lost hair cells (Anttonen et al. 2012; current study). SBEM allowed reliable documentation of unclosed sites at the reticular lamina. Although apoptotic OHCs were found in mutant specimens, their numbers were outnumbered by necrotic OHCs, causing massive structural damage to the organ of Corti. In contrast to wildtype specimens where apoptotic debris was cleared by phagocytotic DCs, necrotic cell debris accumulated in the mutant auditory organs, strengthening the view that apoptotic mode of OHC death promotes efficient debris clearance. Thus, apoptosis is an altruistic and controlled way for elimination of a damaged hair cell from the hearing organ without compromising viability of residual hair cells.

Traumatized cochleas of Cdc42 mutant mice resemble in many respects cochleas exposed to intense noise where mechanical force is thought to damage the integrity of reticular lamina, leading to ion leakage and hair cell necrosis (Saunders et al. 1985; Bohne et al. 2007). However, in the case of noise trauma, it has been difficult to directly link barrier loss to necrotic hair cell death, because noise can also mechanically damage hair cells. The current data with an ototoxic lesion and an optimal mutant mouse model directly show that compromised barrier is sufficient to drive hair cells from apoptosis to necrosis and to induce rapid and massive damage to nerve endings.

To conclude, the current study brings novel findings how structural specializations of mature epithelial cells underlie unique limitations to wound healing in vivo. How intercellular signaling, ion homeostasis, hair cell apoptosis and apical surface closure are linked together in the organ of Corti is a complex but important question regarding the development of therapeutic approaches to prevent hair cell loss. The present findings emphasize that future therapeutic protective and regenerative interventions should not compromise barrier integrity.