Zebrafish (Danio rerio) were maintained and bred using standard procedures and animal research complied with guidelines of Laboratory Animal Care and Use stipulated by Stanford University. The wild-type zebrafish line used for this study was Topfin Long. All embryos and larvae were 1–15 dpf, before sex determination is possible.

The Hybridization Chain Reaction (HCR) assay was used to detect various mRNAs in whole-mount samples (HCRv3; Molecular Instruments). DNA oligonucleotide probes targeting the genes in this study were designed by Molecular Instruments and used per the manufacturer’s instructions. Each probe was assigned to a separate fluorophore channel to enable multiplex detection in each specimen.

Embryos and larvae were raised in 1× E3 with 0.003% PTU starting at 20–24 hpf to prevent pigmentation, and E3 + PTU was replaced daily. Embryos and larvae were collected, anesthetized, and fixed overnight in 4% p-formaldehyde (PFA) in 1× PBS at 4°C (n ≥ 8 for each stage). After fixation, specimens were rinsed in PBS, dehydrated and permeabilized with 100% methanol, and stored in 100% methanol at −20°C until use.

RNA-FISH followed the official protocol for HCR in whole-mount zebrafish, from the manufacturer. Briefly, specimens were rehydrated from methanol into PBST by stepwise, 5-min incubations in PBST with 75, 50, and 25% methanol, followed by five incubations in PBST. Rehydrated samples were permeabilized with 30 μg/ml proteinase K in PBST at room temperature for 15–20 min. Specimens were then rinsed with PBST, post-fixed in 4% PFA in 1× PBS for 20 min, and rinsed 5× for 5 min in PBST.

Specimens incubated in pre-hybridization buffer for 30 min at 37°C prior to addition of the probes. A pooled solution of the probes was prepared in probe hybridization buffer as described in the manufacturer’s protocol. Specimens were incubated in probe hybridization buffer with probes overnight at 37°C. The next day, specimens were washed 4× for 15 min in probe wash buffer and then 2× for 5 min in 5× SSCT at room temperature.

Amplification hairpin oligonucleotides were prepared and applied as described in the manufacturer’s protocol (HCRv3; Molecular Instruments). Individual hairpin h1/h2 solutions were denatured at 95°C for 90 s and cooled to room temperature prior to pooling in amplification buffer. Specimens incubated overnight in the dark at room temperature in the hairpin solution. Hairpin solution was removed by multiple wash steps in 5× SSCT and stored at 4°C in the dark prior to imaging.

All imaging was performed with LSM700 laser-scanning confocal microscopes (Carl Zeiss). Three-dimensional views of neuroepithelia were generated using Imaris software; two-dimensional views were generated using ZEN (Carl Zeiss) and Image J software. For all stages, more than 8 embryos or larvae were examined in at least two independent experiments.

The single-cell RNA sequencing dataset that was utilized for this portion of the analysis comes from the Danio-cell database (Sur et al., 2023). The data can be download via GEO with accession number: GSE223922. This dataset provided the raw data count matrix which was then processed in R via the Seurat Package (v4; Stuart et al., 2019). Based on existing clustering labels, the inner ear portion of the atlas was isolated and then converted to an h5ad format for further analysis in Python via the scanpy package (Wolf et al., 2018). Standard downstream pre-processing steps were taken to log normalize the data (sc.pp.normalize_total, sc.pp.log1p). UMAPs were generated through creation of PCA dimensional reduction to 40 components. Additional analysis tools such as differential gene expression were employed via “t-test” through scanpy’s function: rank_genes_groups while the rest of the data manipulation was done via other scanpy package functions. See full notebook code here: https://colab.research.google.com/drive/1lP2npK34-T5qWeb1yztSLncf4tZ9Pyr5?usp=sharing.

Nascent hair cells in the developing otic capsule first appear approximately at 18 hpf (Whitfield, 2020). By 24 hpf, four to six hair cells are present in two macular organs, which include a tiny otolith (Figure 1A). The anterior macula will develop into the utricle and the posterior macula will become the saccule. We examined expression of tmc1, tmc2a, and tmc2b transcripts in 1 dpf specimens using HCRv3 RNA-FISH (Figures 1B–D). At this early stage, expression of all three genes overlaps in all hair cells (Figures 1E,F). To further corroborate the initial phase of expression of all three tmc transcripts, we examined single optical sections of hair cells in both maculae at 1 dpf (Figures 1G–L) In single optical sections, triple labeling was detected in all hair cells (n = 17 hair cells). One day later at 48 hpf when the number of hair cells has more than doubled in each macula (Figures 2A–C), a mixed pattern exists with most hair cells still expressing two or three of the tmc paralogues along with a small subset of hair cells expressing just a single gene, tmc2a (Figures 2D,E). The emergence of hair cells expressing tmc2a alone is particularly noticeable in the central region of the saccule (Figure 2E). These results suggest that early-stage hair cells express all three tmc1/2 genes and gradually transition to assorted patterns of expression.

We then examined expression at 5 dpf, which is a free-swimming larval stage of development where postural control and acoustic startle reflexes are robust (Figure 3). At this stage, cristae in the developing semicircular canals are present, although behavioral responses mediated by this cell type are not apparent until 25 dpf (Beck et al., 2004). Figures 3A,B shows a lateral view of the inner ear with a schematic of the five sensory end organs. The inferred striolar region of the utricle at 5 dpf was added based on previous publications (Hammond and Whitfield, 2006; Inoue et al., 2013; Liu et al., 2022). At 5dpf, the predominance of hair cells expressing tmc2a alone is more evident (Figures 3C–F). In each end organ, tmc2a is expressed throughout the neuroepithelium (Figures 3D–F). In contrast, tmc1 expression is restricted to the medial extrastriolar region of the larval utricle and peripheral cells in the larval saccule and cristae (Figures 3D–F). This pattern is consistent with previous findings showing that vital dye labeling of hair cell somas remains in the posterior lobe of the posterior macula in tmc2a/2b mutants (Smith et al., 2020; Zhu et al., 2021). The tmc2b paralogue varies according to end organ. In the larval utricle, tmc2b is expressed in tmc1+ cells within the extrastriola and in addition, tmc2b transcripts are observed in a subset of tmc2a+ cells in the striola (Figure 3D). In the larval saccule and crista, tmc2b is predominantly expressed in peripheral hair cells, either outlining the entire saccule or at both ends of the cristae (Figures 3E,F). In addition, tmc2b was seen in a few hair cells with more central locations within the macular organs.

Although the neuroepithelia of larval macular end organs appear to have single layers of hair cells, regions of the cristae exhibit two layers of hair cells that are reminiscent of the pattern in other species (Lysakowski and Goldberg, 2008). As previously reported (Smith et al., 2020; Zhu et al., 2021), round, tear drop shaped hair cells are present in an upper layer and gourd shaped hair cells with nuclei positioned in a lower layer (Figure 4A). The upper layer hair cells show a dependence on tmc2a function in terms of vital dye labeling, whereas the lower layer is tmc1/2b dependent (Smith et al., 2020; Zhu et al., 2021). To determine whether differential expression extends to these two layers, we examined optical cross sections of the lateral crista (Figures 4B–D; n = 4 cristae). Our data indicate that the top layer largely expresses tmc2a, whereas the bottom layer of hair cells expresses all three tmc paralogues (Figure 4E).

In the mouse cochlea, hair cells transition from expressing Tmc2 to Tmc1 at more mature stages (Kawashima et al., 2011). To determine whether the patterns observed at early larval stages were transient or if they persisted to later stages of development, we examined expression of tmc1/2 transcripts at 15 dpf (Figure 5A). We labeled whole mounts using the same HCRv3 method as was done for earlier stages (Figure 5B). We found that the general pattern of tmc2a transcripts being expressed predominantly in the lateral region of the utricle and central regions of each end organ was nearly identical to the patterns seen at 5 dpf (Figures 5C–E). In addition, the restriction of tmc1 to peripheral hair cells was similar. The pattern of tmc2b was also largely unchanged. One notable exception was the addition of what appeared to be another row of tmc2b+ positive cells formed near the lateral edge of the utricle (Figure 5C).

To determine whether other genes encoding components of the mechanotransduction apparatus were universally or differentially expressed like the tmc1/2 genes, we analyzed the transcriptional profile of key genes in specific end organs. Utilizing the daniocell single cell RNA sequencing dataset (Sur et al., 2023), we mapped the distribution of tmc genes within the inner-ear hair cells. We identified 528 hair cells from 40 hpf to 120 hpf and further classified cells based on known transcriptomic markers. We identified 136 hair cells as lateral line hair cells while for the inner ear, we identified 165 striolar hair cells and 78 extrastriolar hair cells (Figure 6A). We used previously identified markers for these distributions: cabp2b and pvalb9 for striolar hair cells and zgc:153395 (skor2) and cabp1b for extrastriolar hair cells (Shi et al., 2022). Based on these classifications, the single cell data correlate well with the mRNA in situ analysis, showing that tmc transcripts describe unique subpopulation of cells in the stages of development. We observed that tmc1 expression is predominantly in the extrastriolar hair cells, while there is expression of tmc2a in both striolar and extrastriolar populations.

With the single cell data, we further explored the expression of other known mechanotransduction components in these cells (Figure 6C). We compared the expression of these corresponding transcripts in the striolar and extrastriolar populations (Figures 6B,C). The subpopulations of the striolar hair cells could be further separated into tmc2a+ and tmc2a/2b+ subtypes (Figure 6D) and the extrastriolar hair cells could be grouped into a predominantly tmc2a+ population versus a tmc1+ subpopulation (Figure 6E).

To confirm the above findings, we examined the expression of candidate genes that were differentially expressed among the various subtypes of hair cells using RNA-FISH. We used 5 dpf larvae for our experiments as the tmc1/2 pattern did not vary considerably after this stage. Candidates with the most striking differences in the expression dot plots in Figures 6D,E include cib2 and cib3, which encode calcium binding proteins that are known to interact with the TMCs in other species (Giese et al., 2023; Wang et al., 2023), loxhd1a and loxhd1b, which encode multi-PLAT domain proteins whose orthologue in mice is implicated in mechanotransduction in cochlear hair cells (Trouillet et al., 2021), and ush1c, encoding a component of the upper tip link (Grillet et al., 2009).

In all end organs we observed that the cib2 and cib3 genes had patterns that mirrored those of tmc2a and tmc1, respectively (Figure 7). cib2 was broadly expressed like tmc2a whereas cib3 was confined to tmc1+ cells (Figures 7A–C). These results suggest a preference for specific combinations of Tmc1/2 subunits and their closely associated Cib2/3 binding partners among different subtypes of hair cells.

The loxhd1 paralogues 1a and 1b exhibited the most complex patterns (Figure 8). Like tmc2a and cib2, loxhd1a is broadly expressed in all hair cells with higher levels predominantly seen in the striolar regions (Figures 8A–C, left panels). In the utricle, loxhd1b is detectable at low levels in the striolar zone (Figure 8A). In contrast, loxhd1b is highly expressed in peripheral hair cells of the saccule’s anterior lobe, and in a small subset of hair cells of the posterior lobe (Figure 8B). We observed that hair cells that are positive for tmc1 in the posterior lobe show much lower levels of lohxd1a and that loxhd1b is not detectable. In cristae, we observed expression of loxhd1b primarily in peripheral hair cells (Figure 8C).

Finally, we examined the expression pattern of ush1c (Figure 9). Unlike the other mechanotransduction components, USH1C has been shown to localize to the upper tip link region in mouse hair cells (Grillet et al., 2009). Like cib3 and tmc1, ush1c is expressed at high levels in extrastriolar or peripheral hair cells (Figures 9A–C); however, ush1c was also detected in central regions of each end organ, indicating that it is a universal component. Curiously, in the saccule ush1c signal was highest in a patch of ventral cells in the posterior lobe and in most of our specimens we noted a single central hair cell in the posterior lobe that expressed high levels of ush1c (Figure 9B).

In general, the subtype grouping identified via bioinformatic analysis was validated with RNA-FISH. One exception was the profiling of tmc2b, where scRNAseq data suggested low copy numbers of transcripts for this gene, whereas RNA-FISH suggested that expression was robust in specific cell types. These differences could be due to a high affinity of the tmc2b RNA probe sets used here or the relatively small number of hair cells captured in the RNAseq datasets that may be resolved with additional sequencing. Nevertheless, the RNA-FISH results validated the grouping of tmc1/2 subtypes identified by bioinformatic analysis and demonstrated further differential expression of genes encoding essential components for mechanotransduction in hair cells.