Since the sensory epithelia of goldfish and amphibians have hair cells with different morphologies, we examined the contours and dimensions of crista hair cells to quantitatively determine if these organs have differently shaped and sized hair cells or whether they are homogenous. To examine the crista for multiple hair cell subtypes based on morphology, we created somatic F0 mosaic transgenic zebrafish that express the fluorescent protein Cerulean under the control of the hair cell promoter myosin 6b (Cruz et al., 2015). This allowed us to identify sporadically labeled hair cells. Zebrafish at the larval stage were examined because they are optically clear, permitting scrutinization of hair cells in vivo. Using a confocal microscope, we imaged optical slices of the central thickness of the lateral crista because it is known that the lateral hair cells are developing in larvae (Haddon and Lewis, 1996), and the central region may contain the most mature population. Micrographs revealed two prevalent hair cell types: Tall hair cells with an elongated thin neck region and a nucleus close to the basal area of the cell and short hair cells that are stunted in comparison (Figures 1A,C). To quantitate length differences and determine spatial relationships between these cells, we generated a stable transgenic line, Tg(myo6b:Cerulean). In this section of the crista, tall and short hair cells are interdigitated (Figures 1B,C). Plots of hair cell lengths reveal two discernable populations (Figures 1D,E). The mean somata lengths of the short and tall hair cells are 11.0 ± 0.2 (mean ± SEM) μm and 17.1 ± 0.3 μm, respectively.

Since hair cells between the lateral line and the maculae of fish and between the maculae and the cochleae of mice (Kawashima et al., 2011) depend on different complements of Tmc proteins, we queried if these distinct hair cell subtypes rely on the same Tmc proteins or if their morphologies correlate with particular Tmc dependencies. To address this question, we implemented a genetic strategy using tmc1 mutant zebrafish, tmc1^cwr5, which lack the Tmc1 protein (Supplementary Tables S1, S2; Chen et al., 2020). To visually identify normal mechanosensitive hair cells versus those that may have lost or experienced a reduction in this defining property, we imaged the uptake of cationic fluorescent molecule 4-Di-2-ASP (Figure 2A), which passes through the hair cell’s mechanotransduction channel to act as an indicator of channel function (Gleason et al., 2009; Owens et al., 2009). After direct injection of 4-Di-2-ASP into the otocyst and subsequent confocal imaging, micrographs demonstrated that most hair cells in controls have functional mechanotransduction channels. However, in the tmc1^cwr5 mutant, tall hair cells (length ∼19 μm) had significantly reduced labeling with 4-Di-2-ASP in the central thickness of the lateral crista, indicating that mechanotransduction is greatly diminished in these cells.

Because these outcomes disagree with recent findings by Smith et al. (2020), who stated that Tmc1 and Tmc2b are individually sufficient for MET function in tall hair cells and did not note a difference in uptake of FM dye in their tmc1 mutant, we verified our findings using a second tmc1 mutant strain, tmc1^cwr4 (Supplementary Tables S1, S2; Chen et al., 2020). In the tmc1^cwr4 mutant, tall hair cells in the central thickness of the lateral crista displayed a substantial reduction in uptake after 4-Di-2-ASP presentation (Figure 2A).

To further analyze the relationship between Tmc1 reliance and hair cell morphology, we applied a double-labeling scheme. Specifically, we generated tmc1^cwr4 mutant larvae that express GCaMP6s-CAAX by outcrossing, injected their ears with fluorescent molecule FM4-64, and then imaged the hair cells of their cristae. GCaMP6s-CAAX is a fluorescent protein that labels the plasma membrane, defining the outline of each hair cell through illumination (Lukasz and Kindt, 2018). FM4-64 is a fluorescent molecule that passes through the transduction channel similarly to 4-Di-2-ASP (Meyers et al., 2003), and its emission spectra is distinguishable from that of GCaMP6s-CAAX. This semi-quantitative method detecting FM4-64 demonstrated that the tmc1^cwr4 mutant that expressed GCaMP6s-CAAX contained tall hair cells with reduced mechanotransduction capacity. The flawed mechanotransduction was manifested by a reduction of ∼48% in the normalized fluorescence intensity ratio in the tall hair cell population (Figures 2B’,B”,C).

A potential explanation for the discrepancy between our findings and those of Smith et al. (2020) could be genetic in origin. Smith et al. (2020) used a zebrafish strain with a mutation in exon 3 of the tmc1 gene. In our work, the tmc1^cwr4 and tmc1^cwr5 mutations are on exon 7 and exon 5, respectively. It is possible that exon skipping plays a role (Anderson et al., 2017), and exon 3 of tmc1 is not as critical for the protein’s function in tall hair cells. Alternatively, Smith et al. (2020) may be imaging another region of the crista than the present study where hair cell Tmc dependencies are different. Nevertheless, tall hair cells in the central thickness of the lateral crista depend heavily on Tmc1 (Figure 2G).

Since tall hair cells are heavily dependent on Tmc1 for mechanotransduction, where short hair cells display no such reliance, we considered the possibility that short hair cells use Tmc2 isoforms, who are encoded by tmc2a and tmc2b. After injecting the larval ears of tmc2b^cwr2 tmc2a^cwr3 double mutants (Supplementary Tables S1, S2) with 4-Di-2-ASP and imaging optical sections of the central thicknesses of lateral cristae, we noted that most short hair cells (∼11 μm) do not permit passage of the fluorescent molecule, indicating that short hair cells require Tmc2 isoforms for mechanotransduction (Figures 2A,G). In contrast, in this double mutant, tall hair cell 4-Di-2-ASP labeling revealed no apparent phenotype.

Next, we looked to confirm the presence of unlabeled short hair cells using GCaMP6s-CAAX as a counter label in the double mutant. Specifically, we generated tmc2b^cwr2 tmc2a^cwr3 animals that express GCaMP6s-CAAX, and then imaged uptake of FM4-64 by hair cells of the lateral cristae. Similar to labeling with 4-Di-2-ASP, most short hair cells do not take up the FM4-64, contrasting with the tall hair cells, which permit entry of the fluorescent molecule (Figures 2B”’,D,E). Although most short hair cells lacking Tmc2 isoforms did not take up FM4-64, on occasion a short hair cell did permit entry of this fluorescent molecule. As these cells seemed out of place, we termed them short hair cell erratics (Figure 2B”’, bottom, red asterisk), borrowing from the geology term for rocks that appear incongruous with the local bedrock.

Because tmc2a mRNA is detectable in the crista by in situ hybridization, where tmc2b mRNA is not (Maeda et al., 2014), we aimed to determine if removal of Tmc2a alone could perturb mechanotransduction by short hair cells. Thus, we carried out 4-Di-2-ASP uptake studies on hair cells in tmc2a^cwr3 single mutant zebrafish. Similar to the double mutant, the tmc2a^cwr3 single mutant predominantly lacked 4-Di-2-ASP uptake by short hair cells (Figure 2A). Comparable results were obtained with FM4-64 injections into the ears of the tmc2a^cwr3 single mutant that expresses GCaMP6s-CAAX (Figures 2B””,D,E). These findings confirm that short hair cells mainly depend on Tmc2 isoforms, and Tmc2a plays a significant role in these cells.

We hypothesized that the short hair cell erratics, who permit mechanotransduction despite being devoid of Tmc2 isoforms, are in fact dependent on Tmc1. To test this, we carried out 4-Di-2-ASP uptake studies on the tmc2b^cwr8 tmc1^cwr4 tmc2a^cwr6 triple mutant (Chen et al., 2020; Supplementary Tables S1, S2). None of the lateral crista hair cells permitted 4-Di-2-ASP entry, indicating that short hair cell erratics of the lateral crista are dependent on Tmc1 (Figure 2F). We confirmed these results using FM1-43FX, a fluorescent molecule that enters through the mechanotransduction channel (Gale et al., 2001; Meyers et al., 2003; Chou et al., 2017; Figure 2F). In all, these findings show that hair cell morphology coincides with a bias for specific Tmc use within the ear, but the relationship of a particular Tmc to a morphology need not be absolute.

Next, we considered whether four hair cell anatomical variables – as they relate to Tmc dependencies – play a role in hearing: organ, position within an organ, axis of best sensitivity in which the hair bundle responds, and the hair bundle’s orientation within this axis. More specifically, a hair cell’s dependency on a particular Tmc may be restricted to one hair bundle orientation within an axis of best sensitivity (asymmetric model), or it may be free of this constraint (symmetric model) (Supplementary Figure S1). We explored Tmc dependencies in the saccule, which is considered the major hearing organ in the larval zebrafish ear (Inoue et al., 2013) and offers a sensory epithelium where the hair bundle orientations can be mapped and correlated to function.

To explore these four hair cell anatomical variables and their relationship to Tmc dependencies in the saccule, we took a two-step approach. We mapped the hair bundle orientations of the 7-dpf saccule of the wild-type Tübingen (Tü) strain by labeling with fluorophore-coupled phalloidin and then overlaid hair cell Tmc dependencies onto this organ. The first step was done in quadruplicate and a representative map is displayed in Figure 3. The map of the posterior macula revealed a stereotyped organization of hair bundles similar to those of 5-dpf animals (Inoue et al., 2013), showing that during this developmental time period the hair bundle orientations are not greatly altered. Overall, the organ is light bulb shaped (Figures 3A,D). There is an anterior narrowing where the dorsal hair cells tend to face anteriorly, but the ventral cells face posteriorly – establishing a single anterior–posterior axis of best sensitivity (Figures 3A,D). Hair bundles in the bulbous region have two orientations: the dorsal hair cells face ventrally, and the ventral hair cells face dorsally, yielding a single dorsal-ventral axis of best sensitivity (Figures 3B–D). There seems to be a transition area between the narrow region and the bulbous region that has hair bundles that may be in the process of orienting to the major axes of best sensitivity (Figure 3A,D). Overall, we show that 7-dpf larvae have two major axes of best sensitivity that are perpendicular to each other, and run parallel to those of the lateral line (Lopez-Schier et al., 2004; Chou et al., 2017), suggesting the paramount importance of detecting these directions for larval survival. Interestingly, the adult zebrafish saccule and those of goldfish and carp have just one axis of best sensitivity (Platt, 1993), contrasting with those of otophysan fishes who have two (Popper and Lu, 2000). Thus, the larval zebrafish saccular architecture – with respect to hair bundle orientations – is more similar to that of the latter superorder than it is to adult zebrafish.

Using this map, we explored if hair cell anatomical variables – either position of the hair cell, axis of best sensitivity, or a hair bundle’s orientation within this axis – play a role in Tmc dependencies in the ear. Alternatively, these variables may not be relevant and Tmc dependencies in the maculae may be organized by a different set of criteria or even randomly. To address this question, we used the tmc2b^cwr2 tmc2a^cwr3 double mutant larvae for a number of favorable properties that it offers. One, hair cell mechanotransduction in this fish strain is dependent on just one Tmc protein, Tmc1, eliminating the chances of complex Tmc interactions that could complicate the interpretation of results. Two, the hair cells that still function are in the posterior macula, but functional hair cells have not been mapped with high resolution (Chen et al., 2020; Smith et al., 2020). Three, this mutant can hear. Its capacity to do so is limited, but the presence of Tmc1 is able to mediate a C-start response to vibrational stimuli (Chen et al., 2020).

To evaluate the potential roles of the anatomical variables, we injected FM1-43FX into the ears of mutant zebrafish and labeled the animals with phalloidin to reveal hair bundle orientations and FM1-43FX uptake. High-resolution confocal images revealed that Tmc1-dependent transduction in the larval saccule is limited to just ∼10–18 hair cells. These Tmc1-dependent hair cells were highly constrained, located in the bulged end of the lightbulb-shaped organ in two symmetrical stripes (Figures 4A,B). The Tmc1-dependent stripes are located at opposite poles of the organ, dorsal and ventral. Each stripe is approximately two hair cells in thickness. All Tmc1-dependent hair cells fall within a unitary axis of best sensitivity, dorsal-ventral, and face both directions, preserving symmetry. This pattern of uptake is not observed in the tmc2a^cwr7 single mutant or tmc1^cwr4 tmc2a^cwr7 double mutant, indicating it is a pattern specific to Tmc1 (Figure 4A). These findings reveal a number of rules about the geometrical nature of Tmc1 dependencies. One, position is a major factor in determining Tmc1 dependence because only cells in the peripherally positioned stripes use this protein. Two, Tmc1-dependence falls within a single axis of best sensitivity, dorsal-ventral. Three, not all hair cells depend on Tmc1 in this axis, so it cannot be the only determining factor of Tmc1 dependence. Four, since Tmc1-dependent hair cells have opposite hair bundle orientations, a single hair bundle direction does not govern Tmc1 reliance, thus, satisfying the symmetric model of Tmc1 dependence (Supplementary Figure S1).

Protocols for housing and handling of zebrafish were approved by Case Western Reserve University’s Institutional Animal Care and Use Committee. Wild-type Tübingen (Tü), Tg(myo6b:GCaMP6s-caax) transgenic (Lukasz and Kindt, 2018), tmc1^cwr5 mutant, tmc2a^cwr3 mutant, tmc2b^cwr2 tmc2a^cwr3 double mutant, and tmc2b^cwr8 tmc1^cwr4 tmc2a^cwr6 triple mutant fish (Chen et al., 2020) were used in this study.

CRISPR gene editing was performed as described previously (Chen et al., 2020). Tmc1 (ENSDARG00000056386) was targeted with an sgRNA complementary to exon 7 (5′-TGGGCTG GTCATGGTTCCAG-3′). Tmc2a (ENSDARG00000033104) was targeted with an sgRNA complementary to exon 9 (5′-AGGTC CCAATGCCCACCATG-3′). The mutations in each gene carried by founder fish and F1 fish were identified using sequencing primers described previously (Chen et al., 2020). Tmc1^cwr4 mutant fish with an 8-bp deletion and tmc2a^cwr7 mutant fish with a 1-bp indel were used for assays in this work. The tmc2a^cwr7 single mutant fish was isolated from a crossover recombination event after mating tmc1^cwr4/+ tmc2a^cwr7/+ fish with wild-type animals and subsequent incrossing.

To generate Tg(myo6b:Cerulean) zebrafish, DNA plasmid, pMT/myo6b/cerulean, at 100 ng/μl and Tol2 RNA at 62.5 ng/μl were mixed with phenol red and 0.125 mM KCl solution and then coinjected into Tü zebrafish embryos at the one-cell stage to create somatic transgenics. To generate a stable transgenic line, positive founder fish were outcrossed with Tü zebrafish.

To generate Tg(myo6b:GCaMP6s-CAAX); tmc2b^cwr2 tmc2a^cwr3 zebrafish, we crossed tmc2b^cwr2/+ tmc2a^cwr3 fish with stable transgenic Tg(myo6b:GCaMP6s-CAAX) fish. We then crossed Tg(myo6b:GCaMP6s-CAAX); tmc2b^cwr2/+ tmc2a^cwr3/+ fish with tmc2b^cwr2/+ tmc2a^cwr3 fish to obtain Tg(myo6b:GCaMP6s-CAAX); tmc2b^cwr2 tmc2a^cwr3 zebrafish. Crossing was used to generate Tg(myo6b:GCaMP6s-CAAX); tmc1^cwr4 zebrafish.

Fixation and FM1-43FX imaging were performed as described previously (Chou et al., 2017). For live imaging, the larvae were anesthetized in 0.612 mM ethyl 3-aminobenzoate methanesulfonic acid (Sigma-Aldrich) in fish water and mounted laterally on glass-bottom dishes (MatTek) in 1.0% (wt/vol) low-melting-point agarose (Promega). Z-stack images of the entire lateral crista were collected on a confocal microscope (TCS SP8, Leica) with a 40 × /1.3 NA oil-immersion objective. A laser wavelength of 405 nm was used for Cerulean excitation. A laser wavelength of 488 nm was used for 4-Di-2-ASP excitation. Laser wavelengths of 545 and 488 nm were used for FM4-64 and GCaMP6s excitation, respectively. For the identification of hair bundle orientations, Alexa Fluor 633 phalloidin at 1:50 dilution was used to label actin filaments (Chou et al., 2017; Chen et al., 2020). For Alexa Fluor 633, an excitation wavelength of 633 nm was used.

Confocal images were imported to and analyzed in ImageJ (NIH). A Gaussian blur filter (radius = 1.00) was applied to all images to reduce random noise. For Cerulean, GCaMP6s, and phalloidin channels, the brightness/contrast were adjusted automatically to enhance the morphological appearance of hair cells or hair bundles. For the styryl dye (4-Di-2-ASP, FM4-64, and FM1-43FX) emission channels, raw data are presented. Images of partial Z-stack maximum projections are shown for the lateral crista to minimize hair cell overlap. For mapping, images of whole Z-stack maximum projections are shown for the posterior macula.

Due to the densely overlapping nature of crista hair cells, cell length measurements or enumeration was performed manually by scanning the Z-stack series of Cerulean-labeled or FM4-64/GCaMP6s double-labeled hair cells, respectively. Graphs were generated in GraphPad Prism version 8 (GraphPad Software).

All statistics were performed by GraphPad Prism version 8. Data are reported as mean ± SEM or SD. Comparisons between groups were tested by one-way ANOVA or Student’s t-test. Short hair cells were judged to be positive for mechanotransduction if their fluorescence intensity was 2.33 standard deviations greater than the background intensity (Honegger et al., 2011).

To quantitate FM4-64 uptake in the lateral crista hair cells, a 0.75 μm² square region of interest (ROI) was defined manually within the hair cell, not overlapping with other hair cells, using the GCaMP6s-CAAX images taken simultaneously (LAS X, Leica). Each ROI selected for quantitation was from the optical slice with the maximum fluorescence intensity of that hair cell’s Z-stack. After background subtraction, the average of the FM4-64 intensities of individual hair cells in the lateral crista were normalized to the average of the intensity of the corresponding control group (Figure 2C).