We and others previously demonstrated that CIB2 is essential for MET in auditory hair cells (Giese et al., 2017; Michel et al., 2017; Wang et al., 2017). In contrast to auditory hair cells, we found that the vestibular hair cells in Cib2^KO/KO mice apparently have MET. We assessed MET via uptake of FM 1–43 (Figure 1A), a styryl dye that mostly permeates into hair cells through functional MET channels (Meyers et al., 2003), indicating that there may be another CIB protein playing a functionally redundant role. Bulk RNA sequencing revealed that all four members of the Cib family are expressed in mouse inner ear tissues (Figure 1B). However, when assessed by single-cell RNA sequencing analysis, only Cib2 and Cib3 are expressed in immature hair cells in the mouse utricle (Figure 1C), a vestibular sensory organ required to detect linear acceleration (McInturff et al., 2018). Furthermore, we find that the expression of Cib2 and Cib3 is maintained in mature type I and type II utricular hair cells (Figure 1C). We also examined the expression of Tmc1 and Tmc2 in utricular hair cells and found that in immature hair cells, Tmc1 is highly expressed, while Tmc2 expression is low (Figure 1C). Tmc2 expression increases during development but remains below Tmc1 levels in both type I and type II hair cells upon maturation (Figure 1C). CIB2 and CIB3 have been reported to interact with TMC1/2, and CIB2 is essential for auditory hair cell MET (Giese et al., 2017; Liang et al., 2021). The co-expression of Cib2 and Cib3 in developing and mature vestibular hair bundles, and the ability of CIB2/3 to interact with TMC1 and TMC2, all make them attractive candidates to regulate MET function in vestibular hair cells.

To evaluate whether CIB3 regulates MET function both in auditory and vestibular hair cells, we generated Cib3 knockout (Cib3^KO) mice using CRISPR/Cas9 technology (Figure 1D). An indel in exon 4 of Cib3 was introduced leading to a nonsense mutation (p.Asp54*), thus causing a loss of allele function (Figure 1E–F). We first characterized hearing function in Cib3^KO/KO and control littermate mice at P16 by measuring auditory-evoked brainstem responses (ABRs). Normal ABR waveforms and thresholds were observed in Cib3^KO/KO indicating normal hearing. Next, we investigated if there was a functional redundancy between CIB2 and CIB3 in auditory hair cells. For this analysis, the Cib3^KO allele was introduced into the Cib2^KO background. Similar to what has been previously reported in Cib2^KO/KO mice (Giese et al., 2017; Michel et al., 2017; Wang et al., 2017), the Cib2:Cib3 double knockout (Cib2^KO/KO;Cib3^KO/KO) mice were also profoundly deaf. The Cib2^KO/KO;Cib3^KO/KO mice did not respond to click or strong tone burst stimuli at 100  dB sound pressure level (SPL) (Figure 2A, B). As reported previously for Cib2^KO/KO mice, we observed that auditory hair cells in Cib2^KO/KO;Cib3^KO/KO mice did not accumulate FM 1–43 dye (Figure 2C). Also, similar to Cib2^KO/KO mice (Giese et al., 2017; Michel et al., 2017; Wang et al., 2017), we observed progressive degeneration of auditory hair cells in Cib2^KO/KO;Cib3^KO/KO mice (Figure 2D). Taken together with previous studies, our results confirm that CIB2 is indispensable for hearing and MET function in the auditory hair cells.

After we assessed auditory function, we turned our analyses toward vestibular function. Similar to Cib2^KO/KO mice (Giese et al., 2017), Cib3^KO/KO mice did not display any obvious indications of vestibular dysfunction, such as circling, hyperactivity, or head bobbing, performed well when exploratory behavior was tested, and had normal vestibulo-ocular reflexes (Figure 3). In contrast, visual observation of the Cib2^KO/KO;Cib3^KO/KO mice indicated altered exploratory behavior in an open-field test, suggesting an impairment of the vestibular system (Figure 3A–B). To assess the implications of the Cib2^KO/KO;Cib3^KO/KO mice on vestibular function, we next performed a series of tests, including: the vestibulo-ocular reflex (VOR), performance in a rotating rod, and head movement analysis (Figure 3C–H). The VOR is mediated by a direct three-neuron pathway linking the vestibular nerve to the extraocular muscles and produces compensatory eye movements in the direction opposite of the head movement to provide a stable gaze (reviewed in Cullen, 2019). The dynamic response properties (i.e. gain and phase) of the VOR can be precisely quantified in response to rotations applied at different frequencies in darkness (VORd). In response to VORd testing, Cib2^+/+;Cib3^KO/KO mice generated robust compensatory eye movements (Figure 3C, black trace). In contrast, Cib2^KO/KO;Cib3^KO/KO mice generated negligible eye movements (Figure 3C, red trace). Consistent with this observation, quantification of the VORd across frequencies (Figure 3D) revealed gains near zero in the Cib2^KO/KO;Cib3^KO/KO mice that were significantly reduced (p<0.05) compared to those measured in Cib2^+/+;Cib3^KO/KO mice. Interestingly, the VORd gains of Cib2^KO/+;Cib3^KO/KO mice also showed attenuation with a significant decrease at 2 Hz (p<0.01) compared to Cib2^+/+;Cib3^KO/KO mice. On the other hand, the response phase of VORd eye movements was comparable for Cib2^KO/+;Cib3^KO/KO and Cib2^+/+;Cib3^KO/KO mice at all frequencies (p>0.05) (Figure 3D).

As a control for possible changes in the visual system and/or ocular motor pathway, we also quantified the optokinetic reflex (OKR), which generates eye movements in response to a moving visual scene and in light. The OKR normally works together with the VOR to stabilize the visual field on the retina. Both Cib2^+/+;Cib3^KO/KO and Cib2^KO/+;Cib3^KO/KO showed robust OKN responses which similarly decreased as a function of frequency (Figure 3E; p>0.05). Likewise, OKR gain values in Cib2^KO/KO;Cib3^KO/KO were comparable with the single exception of a significant gain decrease at 0.2 Hz (p<0.05). Analysis of OKN similarly revealed a comparable response phase for all three groups of mice at all frequencies (p>0.05) with the exception that the Cib2^KO/KO;Cib3^KO/KO mice demonstrated less lag at 0.8 and 2 Hz (Figure 3E–F). Overall, the OKR responses of mutant and control mice were generally comparable.

In addition to generating compensatory VOR eye movements to stabilize gaze, the vestibular system also plays a key role in the maintenance of posture and control of balance. Thus, we also tested all three groups of mice on several vestibular tasks that assessed balance and postural defects. First, quantification of open-field testing experiments confirmed that mice exhibited circling – a behavior that is commonly associated with loss of vestibular function in mice (Figure 3B; p<0.002). We also completed rotarod testing in which performance was quantified by measuring the amount of time the mouse remained on the rotating rod as its acceleration increased. While Cib2^+/+;Cib3^KO/KO and Cib2^KO/+;Cib3^KO/KO mice exhibited similar performance, Cib2^KO/KO;Cib3^KO/KO mice exhibited significantly poorer performance across all time points from day 1 through day 4 (Figure 3G, p<0.002). Finally, it was also possible to distinguish Cib2^KO/KO;Cib3^KO/KO mice from the other two groups in their home cages since they commonly displayed head bobbing, a trait also generally associated with vestibular dysfunction in mice. To quantify head movement in our mice, we attached a 6D MEMS module consisting of three gyroscopes and three linear accelerometers to the mouse’s head implant (see Methods). Analysis of power spectra of head motion revealed significantly higher power for Cib2^KO/KO;Cib3^KO/KO mice in all six dimensions of motion, as compared to Cib2^KO/+;Cib3^KO/KO and Cib2^+/+;Cib3^KO/KO mice (Figure 3H, compare red vs. blue and black). Furthermore, there were no significant differences in the head movements power spectra of Cib2^KO/+;Cib3^KO/KO and Cib2^+/+;Cib3^KO/KO mice (Figure 3H, blue vs. black). Taken together our findings indicate that Cib2^KO/KO;Cib3^KO/KO mice show marked deficits in vestibulomotor function when compared to both Cib2^KO/+;Cib3^KO/KO and Cib2^+/+;Cib3^KO/KO mice.

To understand the root cause of vestibulomotor function deficits described above for Cib2^KO/KO;Cib3^KO/KO mice, we first analyzed vestibular hair bundle morphology using confocal microscopy. High-resolution confocal imaging did not reveal any obvious vestibular hair cell loss in Cib2^KO/KO;Cib3^KO/KO mice (Figure 4A). Next, we visualized MET activity in vestibular hair cells of saccular and utricular end organs of Cib2^KO/KO;Cib3^KO/KO mice using the styryl dye FM 1–43. These vestibular end organs are divided into two main physically and functionally distinct domains, a central striolar domain and a peripheral extrastriolar domain. At P5, FM 1–43 labeling of Cib2^KO/KO; Cib3^KO/KO tissue failed to label vestibular hair cells in appreciable numbers (Figure 4B). Similarly, fixable FM 1–43 FX intraperitoneal injections at P60 also showed no dye uptake in Cib2^KO/KO; Cib3^KO/KO mice (Figure 4C). Cib2^KO/KO and Cib3^KO/KO single mutant mice had dye uptake both in striolar and extrastriolar hair cells, although a subtle reduction in FM 1–43 FX was observed in extrastriolar hair cells of both single mutants at P60 (Figure 4C). Taken together, these results support compulsory but functionally redundant roles for CIB2 and CIB3 in the vestibular hair cell MET complex.

Numerous studies have shown that the same proteins are required for hair cell MET in zebrafish and mammals (Chen et al., 2020; Erickson et al., 2017; Ernest et al., 2000; Gleason et al., 2009; Gopal et al., 2015; Maeda et al., 2017; Maeda et al., 2014; Phillips et al., 2011; Seiler et al., 2004; Söllner et al., 2004). Unlike mammals, the zebrafish’s inner ear does not have a cochlea. Instead, zebrafish rely on their saccule for hearing and utricle for balance. Zebrafish also utilize hair cells in a specialized sensory system, the lateral line, to detect local water movement. The hair cells within both the zebrafish inner ear and lateral line share many morphological similarities to mammalian vestibular hair cells (Nicolson, 2017). Zebrafish scRNAseq data show that both cib2 and cib3 are expressed in the inner ear while primarily cib2 is expressed in lateral-line hair cells (Baek et al., 2022; Shi et al., 2023). Given the zebrafish reliance on the saccule for hearing and the similarities between zebrafish hair cells to mammalian vestibular hair cells, we tested whether Cib2 and/or Cib3 were required for hair cell MET in the zebrafish inner ear or lateral line.

To examine the impact of Cib2 and Cib3 loss in zebrafish, we created mutants using CRISPR-Cas9-based methods to target each homologue (Varshney et al., 2016). Both mutants lead to a stop codon in the last EF-hand domain (Figure 5—figure supplement 1). We first examined auditory function by testing the acoustic startle reflex in cib2 and cib3 mutants. For our experiments, we used a Zantiks behavioral system to quantify the probability of generating a startle in response to a vibrational tap stimulus. This response is thought to rely primarily on hair cells in the zebrafish saccule and lateral-line system. We found that, compared to controls, the probability of a startle response was significantly reduced in cib2 mutants (Figure 5A, control: 0.71 ± 0.04; cib2: 0.28 ± 0.06, p=0.005). In contrast, cib3 mutants startle with a similar frequency compared to controls (Figure 5A, p=0.36). Interestingly, contrary to mammals, where CIB2 is crucial for hearing, acoustic startle responses were not completely eliminated in cib2 mutants. One possibility for this partial loss of acoustic startle is that Cib3 may be able to compensate for Cib2 in the zebrafish. Therefore, we examined acoustic startle responses in cib2;cib3 mutants. In cib2;cib3 mutant animals we observed a complete absence of startle responses (Figure 5A). Furthermore, we observed that cib2;cib3 mutants but not cib2 or cib3 mutants exhibited spontaneous circling behavior, an additional indicator of inner ear dysfunction in zebrafish (Nicolson et al., 1998). Together our behavioral results indicate that Cib2 and Cib3 are both required to ensure normal acoustic startle responses in zebrafish. Furthermore, just loss of Cib2, but not Cib3 can impair auditory function in zebrafish.

Our results demonstrate that both Cib2 and Cib3 are required for proper auditory function in zebrafish. This raises the question of whether, similar to mice, these proteins are critical for MET function in zebrafish hair cells. Therefore, we next turned our analysis to hair cells in neuromast organs of the lateral line where methods to assess MET are well-documented and straightforward (Lukasz and Kindt, 2018; Seiler et al., 2004).

First, to assess whether there were gross alterations to MET in lateral-line hair cells we examined uptake of the FM 1–43. We found that, compared to controls, the percent of hair cells per neuromast labeling with FM 1–43 was significantly reduced in cib2 mutants (Figure 5B, D and E, control: 97.6 ± 1.2; cib2: 37.5 ± 6.9, p<0.0001). In contrast, a similar percent of hair cells per neuromast labeled with FM 1–43 in cib3 mutants, compared to controls (Figure 5B, D and F, cib3: 95.7 ± 1.9, p=1.0). We also quantified the average intensity of the FM 1–43 in hair cells (Figure 5C). We found that, compared to controls, the average FM 1–43 label was reduced in cib2 but not cib3 mutants (Figure 5C). In cib2 mutants we did observe residual hair cells that were labeled with FM 1–43. We hypothesized that in lateral-line hair cells, similar to its role in zebrafish auditory function, Cib3 may be able to substitute for Cib2 in MET. To assess whether Cib3 can substitute for Cib2, we examined FM 1–43 labeling in cib2;cib3 mutants. In cib2;cib3 mutant animals we observed a complete absence of FM 1–43 labeling in lateral-line hair cells (Figure 5B, D and G). Thus, taken together the results of our FM 1–43 labeling analysis are consistent with a requirement for both Cib2 and Cib3 to ensure normal MET in all lateral-line hair cells.

To obtain a more quantitative assessment of MET we used Ca²⁺ imaging to perform evoked measurements of hair cell function. For this assessment, we used a transgenic line that expresses a membrane-localized GCaMP6s (memGCaMP6s) specifically in hair cells (Figure 5H–I’). This transgenic line has previously been used to measure Ca²⁺ influx into apical hair bundles while stimulating lateral-line hair cells with a fluid-jet (Figure 5H–I’; Lukasz and Kindt, 2018). Using this approach, we were able to stimulate and monitor stimulus-evoked mechanosensitive Ca²⁺ responses in hair bundles within neuromast organs of each genotype (Representative responses: Figure 5I–L). We quantified the number of hair bundles per neuromast with mechanosensitive Ca²⁺ responses, and found that compared to controls, significantly fewer cells were mechanosensitive in cib2 and cib2;cib3 mutants (Figure 5—figure supplement 2A, control: 92.2 ± 2.5; cib2: 49.9 ± 5.8, cib2;cib3: 19.0 ± 6.6, p<0.0001). In contrast, in cib3 mutants, the percent of mechanosensitive cells per neuromast was similar compared to controls (Figure 5—figure supplement 2A, cib3: 94.5 ± 1.8, p=1.0). We also quantified the magnitude of the mechanosensitive Ca²⁺ signal in the hair bundles, focusing on only the responsive hair cells for each genotype. This quantification revealed that, while cib3 mechanosensitive responses were not different compared to controls (Figure 5—figure supplement 2B–C, GCaMP6s % ∆F/F control: 70.6±5.9; cib3: 71.6±5.2, p>1.0), the magnitude of responses in cib2 and cib2;cib3 hair bundles were significantly reduced (Figure 5—figure supplement 2B–C, GCaMP6s % ∆F/F control: 70.6 ± 5.9; cib2: 33.0 ± 5.8, cib2;cib3: 14.7 ± 1.9, p<0.0001). Overall, our in vivo FM 1–43 and Ca²⁺-imaging analyses revealed that loss of Cib2 alone, but not Cib3 can impair mechanosensory function in the majority of zebrafish lateral-line hair cells. Furthermore, Cib2 and Cib3 are both required to ensure normal MET function in all lateral-line hair cells.

Previous work has shown that within the zebrafish inner ear and lateral line there are distinct hair cell subtypes that rely on different complements of the mechanosensitive ion channel Tmc: Tmc1, 2a, and 2b. In the primary posterior lateral line (pLL) hair cells are oriented to respond to either anterior or posterior fluid flow. The subset of hair cells that detects posterior flow relies on Tmc2b while the subset that detects anterior flow relies on both Tmc2b and Tmc2a (Figure 6A; Chou et al., 2017). Furthermore, work with the crista of the zebrafish inner ear has shown that there are two morphologically distinct types of hair cells: short, teardrop-shaped cells with taller hair bundles and tall, gourd-shaped cells with shorter hair bundles (Smith et al., 2020; Zhu et al., 2020). Short cells rely on Tmc2a, while tall cells rely on either Tmc1 or Tmc2b for mechanosensitive function (Figure 6E; Smith et al., 2020). Based on this work, we hypothesized that Cib2 or Cib3 could have specific roles in zebrafish hair cell sensory organs linked to hair cell orientation in the lateral line or morphology type in the crista of the inner ear.

Our FM 1–43 labeling in neuromasts revealed no defect in cib3 mutants, suggesting that Cib2 can compensate for Cib3 in all pLL hair cells. In contrast to cib3 mutants, only a subset of hair cells per neuromast were labeled with FM 1–43 in cib2 mutants (Figure 5E). Previous work has shown that in the pLL of tmc2b mutants, a subset of hair cells that respond to anterior flow label with FM 1–43 (Chou et al., 2017). This residual FM 1–43 label is dependent on Tmc2a. Therefore, we examined whether the remaining, FM 1–43 positive hair cells in cib2 mutants have a specific hair cell orientation. For this analysis, we used a fixable form of FM 1–43, FM 1–43 FX to label hair cells. After FM 1–43 FX labeling, we fixed the larvae and labeled the apical hair bundles with phalloidin to reveal hair cell orientation. This analysis revealed that in sibling controls, a similar percentage of FM 1–43 FX positive hair cells are oriented to respond to anterior and posterior flow (Figure 6B, B’ and D). In contrast, when we examined cib2 mutants, we found that significantly more of the FM 1–43 positive hair cells were oriented to respond to posterior flow (Figure 6C–D). This indicates that in cib2 mutants, the residual Cib3 present in pLL hair cells preferentially functions in hair cells that express primarily Tmc2b and are oriented to respond to posterior flow (Figure 6A). On the other hand, Cib2 can function with hair cells that express Tmc2b (respond to posterior flow) as well as hair cells that express Tmc2b and Tmc2a (respond to anterior flow).

Next, we examined whether different hair-cell subtypes (short versus tall morphology) in the zebrafish inner ear preferentially rely on Cib2 or Cib3 for mechanosensitive function. To examine MET function, we injected FM 4–64 into the zebrafish ear in a transgenic background that labels the hair cell membrane (memGCaMP6s). FM 4–64 is a redshifted version of FM 1–43 and provides a similar readout of MET (Erickson et al., 2017; Smith et al., 2020). After injection, we imaged FM 4–64 label in the medial cristae, an inner ear sensory organ that allows for excellent delineation of short and tall cell types (Figure 6F–F’). We found that similar to our FM 1–43 results in the pLL, no hair cells in the medial cristae were FM 4–64 positive in cib2;cib3 double mutants (Figure 6I–I’). In addition, similar to the pLL we observed no defects in FM 4–64 labeling in the crista of cib3 mutants compared to controls (Figure 6G–G’). Interestingly we found that short hair cells in the crista of cib2 mutants failed to label with FM 4–64 (Figure 6H–H’). Previous work has shown that short hair cells rely on Tmc2a for mechanosensitive function (Smith et al., 2020). Overall, our functional assessment of MET based on hair cell subtypes within the zebrafish pLL and inner ear suggests that Cib2 is required for mechanosensitive function in hair cells that express Tmc2a.

To gain molecular insight into the role played by CIB proteins in hair-cell MET, we generated AlphaFold 2 (AF2) (Jumper et al., 2021; Tunyasuvunakool et al., 2021) models of the human TMC1 dimer, alone and in complex with human CIB2/3 (Figure 7A–B; Figure 7—figure supplement 1; Figure 7—video 1). The AF2 model of TMC1 alone resembles previous TMEM16/OSCA-based models of TMC1 and is also consistent with a recent cryo-EM structure of the worm TMC-1 protein (Ballesteros et al., 2018; Jeong et al., 2022; Pan et al., 2018; Walujkar et al., 2021). In these models, TMC1 has ten transmembrane domains (α1 to α10; Figure 7—figure supplement 1; Figure 7—source data 1) and a putative pore is formed by transmembrane helices α4, α5, α6, and α7 at the periphery of each monomer (Pan et al., 2018; Walujkar et al., 2021), away from the dimeric interface involving α10. Our new models feature an additional amphipathic helix, which we denote α0, extending almost parallel to the expected plane of the membrane bilayer without crossing towards the extracellular side (as observed for a mostly hydrophobic α0 in OSCA channels and labeled as H3 in the worm TMC-1 structure) (Jeong et al., 2022; Jojoa-Cruz et al., 2018; Liu et al., 2018; Zhang et al., 2018b). Extracellular and intracellular domains linking transmembrane helices, which were either omitted or poorly predicted in previous work (Ballesteros et al., 2018; Pan et al., 2018; Walujkar et al., 2021), have well-defined secondary structure and a high-confidence score (pLDDT >70 for most regions; Figure 7—figure supplement 1A). The intracellular domain linking helices α2 and α3, denoted here as IL1, adopts a helix-loop-helix with the two helices running parallel to each other and differing in length (Figure 7—figure supplements 1–4; Figure 7—source data 1). This is the same fold observed in its crystal structure in complex with CIB3 (Liang et al., 2021), which validated the modeling approach.

Interestingly, AF2 models of TMC1 in complex with CIB2 suggest that the N-terminus (NT) of TMC1 also interacts with CIB2 (Figure 7A–B). In our model, CIB2 is clamped between TMC1-NT and TMC1-IL1 while the first helix of TMC1-IL1 runs parallel to the membrane plane interacting with the base of α1 and the intracellular linkers between α4 and α5, α6 and α7, and α8 and α9. Notably, the predicted interaction between TMC1-NT and CIB2 is consistent with previous biochemical data (Giese et al., 2017), while the overall model resolves seemingly contradictory results indicating that either the NT or IL1 of TMC1 was responsible for interactions with CIB proteins (Giese et al., 2017; Liang et al., 2021). Additional AF2 models of TMC1 with CIB3 and of TMC2 with CIB2/3 (Figure 7—figure supplement 1) predicted the same architecture for the TMC1/2 + CIB2/3 complexes where CIB proteins are clamped between the IL1 and NT domains of TMCs. Predictions for fish complexes also show the same general architecture (Figure 7—figure supplement 1A).

Nuclear magnetic resonance (NMR) experiments were used to validate the clamp model for Homo sapiens (hs) TMC + CIB interactions. We produced uniformly ¹⁵N labeled hs CIB2 (22.7 kDa) and CIB3 (22.9 kDa) proteins ([U-¹⁵N]-hs CIB2/3), alone or co-refolded with unlabeled hs TMC1-IL1 (6.3 kDa). In parallel, we produced an unlabeled soluble TMC1-NT fragment (6.9 kDa) and used these CIB and TMC1 samples to test protein-protein interactions by monitoring their effects on 2D ¹H-¹⁵N-correlation spectra of the labeled CIB2 and CIB3 proteins. All protein fragments were purified using size exclusion chromatography (SEC) (Figure 7—figure supplement 5), revealing a single monodisperse peak for [U-¹⁵N]-hs CIB2 (69.2 mL elution volume in S75 16/600 column) and a slightly displaced peak for [U-¹⁵N]-hs CIB2 + hs TMC1-IL1 (68.5 mL). In contrast, SEC experiments revealed two peaks for [U-¹⁵N]-hs CIB3 alone (potential dimer at 62.8 mL and potential monomer at 73.1 mL) and a single peak for [U-¹⁵N]-hs CIB3 + hs TMC1-IL1 (71.3 mL). ¹H-¹⁵N-correlation spectra of [U-¹⁵N]-hs CIB2 alone were consistent with prior data indicating that many, but not all potential peaks are resolved (Figure 7C, Figure 7—figure supplement 6A; Huang et al., 2012). Spectra for the [U-¹⁵N]-hs CIB2 + hs TMC1-IL1 complex exhibit widespread perturbations, consistent with extensive remodeling of the CIB2 structure. Moreover, the addition of hs TMC1-NT induced additional chemical shift perturbations in many, but not all peaks, demonstrating that the TMC1-IL1 and TMC1-NT ligands bind simultaneously to [U-¹⁵N]-hs CIB2 (Figure 7C, Figure 7—figure supplement 6B). We observed similar results when comparing spectra for [U-¹⁵N]-hs CIB3 alone, co-refolded with hs TMC1-IL1, and upon addition of hs TMC1-NT (Figure 7—figure supplement 6C–D). A control experiment with the co-refolded mix of hs CIB3 + hs TMC1-IL1 and [U-¹⁵N]-hs CIB2 (Figure 7—figure supplement 6E) revealed no change in chemical shifts indicating that direct heteromeric interactions between CIB2 and CIB3 are unlikely. Thus, taken together, the NMR data strongly support a model in which human CIB2/3 proteins are clamped between TMC1-IL1 and TMC1-NT as suggested by AF2 and consistent with prior biochemical data revealing interactions with either TMC1-NT or TMC1-IL1 (Giese et al., 2017; Liang et al., 2021).

Our dimeric AF2 models of TMC1/2 alone and in complex with CIB2/3 are consistent with previous experimental data that validated the location of the putative pore in similar TMEM16/OSCA-based homology models of TMC1 (Pan et al., 2018). Our models are also consistent with our NMR data supporting the clamp model for the TMC + CIB complex (Figure 7C). To improve the AF2 predictions and to gain mechanistic insights into TMC and CIB function we turned to all-atom equilibrium molecular dynamics (MD) simulations (Karplus and Petsko, 1990). We built ten different models that included dimeric TMC1 or TMC2 alone, in complex with CIB2 or CIB3 with and without bound Ca²⁺. These models were embedded in either pure POPC or stereocilia-like mixed composition bilayers and solvated (150 mM KCl) to generate a total of twenty systems that were equilibrated for 100 ns each (Figure 8A, Figure 8—source data 1).

Equilibrium simulations of TMC1 and TMC2 alone revealed stable conformations for the transmembrane domains of these proteins, but a highly mobile NT domain (Figure 8B, Figure 8—figure supplement 1). In simulations where CIB2 or CIB3 were present, with or without bound Ca²⁺, we observed more stable conformations of N-termini along with stable salt bridges (Figure 8C). Buried surface area (BSA) for the TMC + CIB complexes (Figure 8—figure supplement 2A, C and E, Figure 8—source data 2) revealed a possible trend in which TMC1 and TMC2 have more contacts with CIB3 than with CIB2. While additional sampling and longer simulations might be needed for a robust prediction, these results suggest variable affinities for TMC + CIB protein complexes that might be modulated by Ca²⁺ (Figure 8—figure supplement 2F).

Additional analysis of the putative pore in our simulations of TMC1 and TMC2 revealed hydration as observed in previous simulations of TMEM16/OSCA-based TMC1 models, as well as movement of lipids that lined up the predicted permeation pathway for cations suggesting a membrane-based gating mechanism (Pan et al., 2018; Sukharev and Corey, 2004; Walujkar et al., 2021). Longer, Anton 2, 960-ns long equilibrium trajectories of TMC1 with CIB2 and bound Ca²⁺ in a mixed bilayer and TMC1 with CIB2 and (un)bound Ca²⁺ in a POPC bilayer revealed how parts of the amphipathic α0 helices spontaneously partitioned at the membrane interface with their hydrophobic sides pointing toward the lipid bilayer core (Figure 8D). The N-terminal cytoplasmic domain of LHFPL5 has been shown to interact with this TMC1 amphipathic helix, which might be essential for channel function in mice (Qiu et al., 2023). The BSA for the dimeric TMC1 + CIB2 complex in these longer simulations (Figure 8—figure supplement 2E) decreased as contacts between TMC1-α0 and CIB2 were lost for one of the monomers in the mixed bilayer system. In the monomer in which contacts were maintained, a single cholesterol was situated in the middle of the putative pore. Similarly, TMC1-α0 and CIB2 contacts were lost for both monomers in the POPC bilayer systems. Regardless, in all Anton 2 equilibrium trajectories contacts between CIB2 and the TMC1-IL1 and -NT remained, thus confirming the stability of the complex over a microsecond timescale.

To further explore the properties of our models of TMC + CIB complexes we took equilibrated TMC1 systems and applied a –0.5 V transmembrane voltage while monitoring ion permeation through the putative pore during 100-ns long trajectories (Figure 8—figure supplements 3–4, Figure 8—source data 2, Figure 8—video 1). In systems with pure POPC bilayers, we observed permeation of K⁺ in either one or both pores of the TMC1 dimer, with or without CIB2 or CIB3 and with or without bound Ca²⁺, despite the presence of Cl^- (150 mM KCl). Similar results were obtained for four TMC1 systems simulated at –0.25 V. Variability of the conductance state was attributed to partial blockage by lipids and poor sampling, but in all cases in which a clear conductive state was monitored, we observed the almost exclusive passage of K⁺ with conductance values that reached up to ~138 pS for a single pore. While simulations tend to overestimate bulk KCl conductivity and in some cases, predictions need to be scaled down (by a factor of ~0.6) (Walujkar et al., 2021), the conductance values monitored here are consistent with previous simulations and with the expected conductance of the inner-ear transduction channel (Beurg et al., 2021; Walujkar et al., 2021). In contrast, we observed smaller conductance values for TMC1 systems with mixed lipid bilayers (at most ~22 pS for a single pore), regardless of the presence of CIB2 or CIB3, with or without bound Ca²⁺ (Figure 8—figure supplement 4B). Additional Anton 2 simulations for one of these TMC1 + CIB2 + Ca²⁺ systems with a mixed membrane, equilibrated for 240 ns and then subjected to –0.43 V for 480 ns, revealed a low conductance value of ~6.2 pS. The reduced conductance values observed for mixed-membrane systems were attributed in part to pore blockage by cholesterol and lipids in the initial setup. Interestingly, Anton 2 simulations of TMC1, TMC1 + CIB2, and TMC1 + CIB2 + Ca²⁺ in pure POPC membranes subjected to –0.5 V, –0.25 V, and –0.125 V for 480 ns, 480 ns, and 960 ns, respectively, revealed a trend in which TMC1 exhibits higher conductance with CIB2 bound (Figure 8E, Figure 8—source data 2). This may begin to explain why CIB2 is essential for normal TMC1 function.

Simulations of TMC2 with an applied transmembrane voltage of –0.5 V revealed a more closed state when compared to our TMC1 model, regardless of membrane composition and presence or absence of CIB2, CIB3, and bound Ca²⁺ (Figure 8—figure supplement 4). Conduction of K⁺ was observed in most cases, but the maximum conductance did not reach above ~64 pS, suggesting that the conformation simulated is less open than the ones explored in simulations of TMC1. We also speculate that this is due to TMC2 having an intrinsic lower single-channel conductance than TMC1, as has been suggested by some experiments (Kim et al., 2013), but not others (Pan et al., 2013). It is also possible that our TMC2 model is not in a fully open conformation, which can only be reached upon mechanical stimulation. Regardless, our simulations unequivocally indicate that TMC1 and TMC2 are cation channels and that these proteins can form stable complexes with CIB2 and CIB3.

Mouse heart tissues were homogenized in CHAPS buffer containing 1 X protease inhibitors, 1  mM Na orthovanadate, 10  mM Na glycerophosphate, and 10  mM NaF. Equivalent protein concentrations were fractionated on a 4–20% Bis-Tris gel (Invitrogen) and transferred to polyvinylidene fluoride (PVDF) membrane (Millipore). Membranes were blocked and then probed with the rabbit anti-CIB2 (1:500), CIB3 (1:500), and GAPDH (1:500) overnight at 4 °C, followed by three washes of 30 min each in 1 X TBST. After washing, blots were incubated with horseradish peroxidase-conjugated anti-rabbit antibody (1:1000, Sigma NA934V, Lot # 17640116) for 2 hr at room temperature, followed by detection using the ECL Prime Western Blotting System (Thermo Fisher Scientific 32,106).

RNAs from brain were extracted from the Cib2;Cib3 mutant and control mice using the Ribopure kit (Ambion). SMARTScribe Reverse Transcriptase kit (Clontech) was used to generate cDNAs, and SYBRgreen technology (Qiagen) was used to perform the qPCR. The following primers were used to amplify Cib2 and Cib3: Cib2_exon4_F: CTCTGTGCTCTGCGAATCAG, Cib2_exon5_R: GGCCAGCGTCATCTCTAAGT, Cib3_exon3_F: TGGTGCCTCTTGACTACACG, Cib3_exon5_R: CAGCACCTTCTCACAGACCA. hprt amplification was used to normalize the samples: hprt_exon8_F: TGTTGTTGGATATGCCCTTG, hprt_exon9_R: GGCTTTGTATTTGGCTTTTCC.

Surgical techniques and experimental setup have been previously described (Beraneck and Cullen, 2007). Eye movement data were collected using an infrared video system (ETL-200, ISCAN system). The rotational velocity of the turntable (head velocity) was measured using a MEMS sensor (MPU-9250, SparkFun Electronics). Eye movements during OKR were evoked by sinusoidal rotations of a visual surround (vertical black and white stripes, visual angle width of 5°) placed around the turntable at frequencies 0.2, 0.4, 0.8, 1, and 2 Hz with peak velocities of ±16°/s. To record VOR responses, the turntable was rotated at sinusoidal frequencies 0.2, 0.4, 0.8, 1, and 2 Hz with peak velocities of ±16°/s in both light and dark. In the light condition, the visual surround remained stationary, whereas in the dark condition, both the visual surround and turntable rotated in phase. Head and eye movement signals were low-passed filtered at 125 Hz and sampled at 1 kHz. Eye position data were differentiated to obtain velocity traces. Cycles of data with quick phases were excluded from the analysis. Least-square optimization determined the VOR and OKN gains and phases (Beraneck et al., 2008) plotted as mean ± standard deviation (SD) against all frequencies for all mice. Cib2^+/+; Cib3^KO/KO (n=4), Cib2^KO/+;Cib3^KO/KO (n=5), and Cib2^KO/KO;Cib3^KO/KO (n=4) mice were tested for all frequencies. Two-way ANOVAs and Bonferroni post hoc tests were used to test statistical significance.

Cib2^+/+;Cib3^KO/KO (n=6), Cib2^KO/+;Cib3^KO/KO (n=6), and Cib2^KO/KO;Cib3^KO/KO (n=6) mice were used. Motorized rotating rod (Rotarod, IITC Life Science) was programmed to accelerate from 4 rpm to 40 rpm with a ramp of 300 s for each test trial. The time taken for a mouse to fall off the rotarod was recorded. Three trials were performed each day for five consecutive days. First day (day 0) was considered the training day.

Head motion was recorded in mice via a miniature motion sensor (MPU-9250, TDK Corporation) attached to the top of the skull. The motion sensor recorded 3-axis linear acceleration (i.e. fore-aft, lateral, and vertical) and 3-axis angular velocity (roll, pitch, and yaw) at a 100 Hz sampling rate. The power spectra of the motion signals in all six dimensions were computed using Welch’s averaged periodogram with nfft=512 and a Bartlett window (pwelch function, MATLAB, MathWorks).

Mice were placed in a cylindrical-shape container (27 cm in diameter, 30 cm in height) and their behavior was recorded using a cell-phone camera. The number of rotations made during the 120 second recording were counted offline by watching the recorded videos. A rotation is defined as a completed 360 degree rotation of the mouse’s heading.

The cochlear and vestibular sensory epithelia were isolated, fine dissected, and permeabilized in 0.25% Triton X-100 for 1 hr and blocked with 10% normal goat serum in PBS for 1  hr. The tissue samples were probed with primary antibody overnight and after three washes were incubated with the secondary antibody for 45  min at room temperature. Rhodamine phalloidin or Alexa fluor phalloidin 488 were used at a 1:250 dilution for F-actin labeling. Nuclei were stained with DAPI (Molecular Probes). Images were acquired using either LSM 700 laser scanning confocal microscope (Zeiss, Germany) with a 63×1.4 NA or 100×1.4 NA oil immersion objectives or Leica SP8 laser scanning confocal microscope with a 100×1.44 NA objective lens. Stacks of confocal images were acquired with a Z step of 0.05–0.5  µm and processed using ImageJ software (National Institutes of Health). Experiments were repeated at least 3 times, using at least three different animals.

In the Ahmed lab, cochlear explants from Cib2 and Cib3 mutant mice were dissected at postnatal day 5 (P5) and cultured in a glass-bottom petri dish (MatTek, Ashland, MA). They were maintained in Dulbecco’s modified Eagle medium (DMEM) supplemented with 10% FBS (Life Technologies) for 2 d at 37 °C and 5% CO₂. Explants were incubated for 10  s with 3  µM FM 1–43, washed three times with Hank’s balanced salt solution, and imaged live using a Zeiss LSM 700 scanning confocal microscope.

In the Holt Lab, P60 mice were injected intraperitoneally with 5 mg/g mouse weight of FM 1–43 FX fixable form as previously described (Meyers et al., 2003). Tissue was collected after ~24 hr and fixed for 1 hr at 25 °C in 4%PFA/1 X PBS protected from light. Tissue was then transferred into 120 mM EDTA pH 7.4 for 48–72 hr protected from light. Then tissue was transferred to 1XPBS, dissected, incubated in 1:1000 dilution of Phalloidin (Invitrogen), washed three times in 1XPBS and mounted under glass coverslips for microscopy. Imaging was completed using a 20x objective on a Zeiss 800 confocal microscope with laser intensity, detector gain, and post processing set equally for each sample.

Hearing thresholds of Cib2 and Cib3 mutant mice (n = 5–10 for each genotype) were evaluated by recording ABRs. All ABR recordings, including broadband clicks and tone-burst stimuli at three frequencies (8, 16, and 32  kHz), were performed using an auditory-evoked potential RZ6-based auditory workstation (Tucker-Davis Technologies) with high frequency transducer RA4PA Medusa PreAmps. Maximum sound intensity tested was 100  dB SPL. TDT system III hardware and BioSigRZ software (Tucker Davis Technology) were used for stimulus presentation and response averaging.

Zebrafish (Danio rerio) were grown at 28 °C using standard methods. Zebrafish were maintained in a Tubingen or TAB background. Larvae were raised in E3 embryo medium (5  mM NaCl, 0.17  mM KCl, 0.33  mM CaCl₂, and 0.33  mM MgSO₄, buffered in HEPES, pH 7.2). Zebrafish work at the National Institute of Health (NIH) was approved by the Animal Use Committee at the NIH under animal study protocol #1362–13. All experiments were performed on larvae at 5–6 d post fertilization (dpf). At this age, sex determination is not possible. Neuromasts (L1-L5) in the primary posterior lateral line were used for our analyses. For Ca²⁺ imaging the following transgenic line was used: Tg(myo6b:memGCaMP6s)^idc1 (Jiang et al., 2017).

Cib2^idc21 and cib3^idc22 mutants were generated in-house using CRISPR-Cas9 technology as previously described (Varshney et al., 2016). Exon 3 and exon 6 were targeted for cib2 and cib3, respectively. Guides for cib2 and cib3 are as follows: 5’-GGAGGAGCTTACACCAG(AGG)–3’ and 5’-GGAAATCCTCCAGAGAAAGG(CGG)–3’. Founder fish were identified using fragment analysis of fluorescent PCR products (Varshney et al., 2016). Founder fish containing a 10 bp insertion in cib2 5’-GCTTACACCA(CCTTTGGTAG)GAGGAGGTTA-3’ resulting in a stop codon in the third EF-hand domain, at amino acid 148, was selected for analysis (Figure 5—figure supplement 1). In addition, a founder fish containing a 4 bp deletion in cib3 5’-TGGCCGCCT(TTCT)CTGGAGG-3’ resulting in a stop codon in the third EF-hand domain, at amino acid 159, was selected for analysis (Figure 5—figure supplement 1). After behavior or imaging analyses, genotypes were confirmed using standard PCR and sequencing. The following primers were used for genotyping: cib2_FWD 5’- ACTGAAGCAACTCCTTTTGTGG-3’ and cib2_REV 5’-GAACCATGCATTTTTACAGC-3’ and cib3_FWD 5’-ATAGTCTCTCCATCTATAGACT-3’ and cib3_REV 5’-TTTTAGGGTGAAATAAAACAGA-3’. Crosses of cib2^+/-;cib3^-/- X cib2^+/-;cib3^+/- or cib2^-/-;cib3^+/- X cib2^+/-;cib3^+/- +/- were used to propagate larvae for our analyses. All controls or siblings are double heterozygous animals. Larvae were chosen at random, except for cib2;cib3 homozygous mutants which, similar to other zebrafish mutants with hearing and balance defects, had no swim bladder and no acoustic startle response.

To label lateral-line hair cells with FM 1–43, larvae were immersed in 1 µM FM 1–43 (Thermo Fisher, T3163) in E3 for 30 s, and then washed 3 times in E3. After washing, larvae were mounted in 1% low melt agarose containing an anesthetic, 0.02% tricaine methanesulfonate (Syndel, ANADA 200–226). FM 1–43 labeled hair cells were then imaged on an A1 Nikon upright laser-scanning confocal microscope with a 60×1.0 NA water objective. FIJI was used to process image Z-stacks. To estimate the number of hair cells per neuromast, laser scanning DIC images were used to count the number of hair bundles per neuromast. To quantify FM 1–43 label, a 5 µm circular region of interest (ROI) was placed on the soma of each hair cell with a hair bundle in the DIC image. These ROIs were used to quantify the mean FM 1–43 intensity per hair cell. These intensity values were averaged to calculate the average FM 1–43 intensity per neuromast.

FM 4–64 (Thermo Fisher, T13320) was used to label hair cells in the crista within the zebrafish inner ear. For this labeling, Tg(myo6b:memGCaMP6s)^idc1 larvae were mounted on their side on Sylgard chamber in E3 embryo media containing 0.02% MESAB. Approximately 2–3 nL of 10 µM FM 4–64 in 0.1 M KCl were injected into the otic capsule as previously described (Smith et al., 2020). FM 4–64 labeled hair cells of the medial crista were then imaged on an A1 Nikon upright laser-scanning confocal microscope with a 25×1.10 NA or 60×1.0 NA water objective. FIJI was used to process image Z-stacks. To assess tall versus short hair cells in each crista memGCaMP6s label was used.

To correlate hair bundle orientation in lateral line hair cells with intact mechanotransduction, larvae were immersed in 3 µM FM 1–43 FX (Thermo Fisher, F25255) in E3 for 30 s, and then washed three times in E3. After washing, larvae were fixed with 4% paraformaldehyde in PBS at 4 °C overnight. After fixation, larvae were washed 4×5 min in 0.1% Tween in PBS (PBST). Larvae were then placed in blocking solution (2% goat serum, 1% bovine serum albumin, 2% fish skin gelatin in PBS +0.3% triton) for 2 hr at room temperature. After block Alexa Fluor 633 Phalloidin (Thermo Fisher, A22284) was added at 1:100 in PBST. Larvae were incubated for 4 hr at room temperature and then washed 4×5 min in PBST. Larvae were mounted on glass slides with Prolong Gold (Thermo Fisher Scientific) using No. 1.5 coverslips. To image hair bundle orientation and FM 1–43 FX label in lateral-line hair cells, a Zeiss LSM780 confocal microscope with Airyscan was used. Hair cells were imaged using a 63×1.3 NA oil objective and processed with an Airyscan processing factor of 7.0. For analysis, Z-stacks were max-projected in FIJI. Images were background subtracted using a rolling ball correction. Hair cells with a FM 1–43 FX intensity more than twice the background intensity were scored as FM 1–43 FX positive.

For imaging of mechanosensitive Ca²⁺ signals in neuromast hair bundles, Tg(myo6b:memGCaMP6s) animals were mounted on their side on Sylgard chamber in E3 embryo media containing 0.2% MESAB (Lukasz and Kindt, 2018). Larvae were then pinned and a solution containing α-bungarotoxin (125 µM, Tocris) was injected into the heart cavity for paralysis during imaging. After paralysis, larvae were immersed in extracellular imaging solution (in mM: 140 NaCl, 2 KCl, 2 CaCl₂, 1 MgCl₂, and 10 HEPES, pH 7.3, OSM 310+/-10). To image Ca²⁺ signals in the hair bundles we used a Bruker Swept-Field confocal system equipped with a Rolera EM-C2 CCD camera (QImaging, Surrey, Canada) and a Nikon CFI Fluor 60×1.0 NA water immersion objective as previously described (Zhang et al., 2018a). The system includes a band-pass 488/561 nm filter set (59904-ET, Chroma) and is controlled using Prairie View software (Bruker Corporation). A piezoelectric motor (PICMA P-882.11–888.11 series, PI Instruments) attached to the objective and used to acquire 5-plane Z-stacks every 0.5 µm using a 35 µm slit at a 50 Hz frame rate for a 10 Hz volume rate. A fluid-jet composed of a glass capillary attached to a pressure clamp system (HSPC-2-SB, ALA) was used to deliver a 500 ms anterior and posterior step stimulation to deflect hair bundles along their axis of sensitivity.

To quantify GCaMP6s fluorescent intensity changes during stimulation, we registered the Z-stacks and then average projected the Z-stacks into a single plane. We then loaded the projected images into FIJI and used the Time Series Analyzer V3 plugin to create circular ROIs with a ~1.7 μm diameter. ROIs were placed on the center of each individual bundle. We then measured and plotted change in the mean intensity (∆F/F₀) within the region during the recording period. The mean intensity within each ROI was computed for each hair bundle. A hair bundle with a signal magnitude (peak value of intensity change upon stimulation) above 10% ∆F/F₀ was considered mechanosensitive. To create the AVG hair bundle GCaMP6s (∆F/F₀) per neuromast plot in Figure 5—figure supplement 2C, only traces from mechanosensitive cells were averaged.

A Zantiks MWP behavioral system was used to examine startle responses. The Zantiks system tracked and monitored behavioral responses via a built-in infrared camera at 30 frames per second. A 12-well plate was used to house larvae during behavioral analysis. Each well was filled with E3 and 1 larva. All fish were acclimated in the plate within the Zantiks chamber in dark for 15 min before each test. To induce startle, an integrated stepper motor was used to drive a vibration-induced startle response. A vibrational stimulus that triggered a maximal % of animals startling in controls without any tracking artifacts (due to the vibration) was used for our analyses. Each larva was presented with the vibrational stimulus 5 times with 100 s between trials. During the tracking and stimulation, a Cisco router connected to the Zantiks system was used to relay x, y coordinates of each larva every frame. To qualify as a startle response, a distance above four pixels or ~1.9 mm was required within two frames after stimulus onset. Animals were excluded from our analysis if no tracking data was recorded for the animal.

All data shown are mean ± standard error of the mean (SEM). All experiments were performed on at least four animals and eight neuromasts and four cristae. Behavior analysis and Ca²⁺ imaging was repeated at least twice on two independent days. For in vivo work (calcium imaging and FM 1–43 labeling) fish were chosen at random and experimenter was blinded when possible. All replicates represented in graphs are biological and no samples were excluded. A power analysis was used to ensure that a sufficient sample size was used for our in vivo work. All statistical analysis was performed using GraphPad Prism 8.0. A D’Agostino and Pearson test was used to test data for normality. A one-way ANOVA with a Dunnett’s correction for multiple comparisons was used to compare differences between siblings and controls regarding the % of mechanosensitive hair cells, number of hair cells, % of FM 1–43 positive hair cells, average FM 1–43 label. For proportion of animals startling and comparisons for FM 1–43 FX label, a Kruskal-Wallis test with a Dunnett’s correction for multiple comparisons was used.

DNA sequences encoding for hs CIB2, CIB3, TMC1-IL1, and TMC1-NT were subcloned into NdeI and XhoI sites of the pET21a vector. All DNA constructs were sequence verified. Protein fragments were expressed in Escherichia coli BL21 Rosetta (DE3) (Novagen) cells, which were cultured in LB or TB media, induced at OD₆₀₀~0.4–0.6 with 1 mM IPTG and grown at 20 °C or 30 °C for ~16 hr. Protein fragments for NMR were produced using the same cells cultured in M9 minimal media supplemented with MEM vitamin solution (Gibco) and 1 g/L ¹⁵NH₄Cl (Cambridge Isotopes) to produce uniformly labeled [U-¹⁵N] protein. Cells were lysed by sonication in denaturing buffer (20 mM Tris HCl, pH 7.5, 6 M guanidine hydrochloride, 10 mM CaCl₂, and 20 mM imidazole) or non-denaturing buffer (20 mM Tris HCl, pH 7.5, and 20 mM imidazole for hs TMC1-NT). The cleared lysates were loaded onto Ni-Sepharose (GE Healthcare) and eluted with denaturing buffer or non-denaturing buffer supplemented with 500 mM imidazole. Denatured proteins were refolded using MWCO 2000 membranes (Spectra/Por 7) in dialysis buffer (400 mM Arg-HCl, 20 mM Tris-HCl, pH 7.5, 150 mM KCl, 50 mM NaCl, 2 mM CaCl₂, 2 mM DTT) for ~16 hr. Before starting refolding reactions (when applicable), elution solutions were diluted to ~0.5 mg/mL using the denaturing buffer, and then reduced by using 2 mM DTT in the diluted sample. Refolded and soluble proteins were concentrated using Vivaspin 20 or Amicon 15 centrifugal concentrators (10 kDa, 5 kDa, or 3 kDa molecular weight cutoff) and further purified on Superdex S75 columns (GE Healthcare) in SEC buffer (50 mM HEPES, pH 7.5, 126 mM KCl, 10 mM NaCl, 3 mM CaCl₂, 5 mM DTT). Proteins were concentrated to ~0.5 mL for NMR experiments.

Purified NMR samples were at concentrations of ~1.4 mM, ~0.2 mM, ~0.8 mM, and ~0.2 mM for CIB2, CIB3, CIB2 + TMC1-IL1, and CIB3 + TMC1-IL1, respectively, in SEC buffer (50 mM HEPES, pH 7.5, 126 mM KCl, 10 mM NaCl, 3 mM CaCl₂, 5 mM DTT), supplemented with 10% (v/v) D₂O and 0.2 mM 2,2-dimethyl-2-silapentane-5-sulphonate (DSS) for field-frequency locking and chemical shift referencing. Two-dimensional ¹H-¹⁵N TROSY-HSQC spectra (Nietlispach, 2005) were recorded for each sample at 25 °C on a Bruker Avance III HD 800 MHz spectrometer equipped with a 5 mm TXI cryoprobe with z-gradients. Data were collected using 20–24 scans, 2048×280 datapoints, and spectral widths of 17482.518×4054.298 Hz in the ¹H and ¹⁵N dimensions, respectively. Data were processed and visualized using NMRFx (Norris et al., 2016).

Models for monomeric hs TMC1 (residues 84–760; NCBI ID: NP_619636.2) and hs TMC2 (148–863; NCBI ID: NP_542789.2) in complex with full-length hs CIB2 (NCBI ID: NP_006374.1) and hs CIB3 (NCBI ID: NP_473454.1), as well as truncated dimeric models of hs TMC1 (residues 260–760; NCBI ID: NP_619636.2) and hs TMC2 (residues 324–863; NCBI ID: NP_542789.2) were generated using AF2 (Evans et al., 2021; Jumper et al., 2021). The TMC2 C-terminal end (residues 818–863) was truncated to match the C-terminal end of TMC1. Using VMD (Humphrey et al., 1996), a model of dimeric hs TMC1 + hs CIB3-Ca²⁺ was constructed by aligning the monomeric hs TMC1 + hs CIB3 complex to the transmembrane region of the truncated dimeric hs TMC1 model. Coordinates for Ca²⁺ ions were obtained by aligning the crystal structure of Mg²⁺-bound hs CIB3 in complex with a hs TMC1-IL1 peptide (PDB code: 6WUD) (Liang et al., 2021) to the AF2 model of monomeric hs TMC1 + hs CIB3, then manually changing Mg²⁺ for Ca²⁺. This model was minimized for 1000 steps in vacuum using NAMD (Phillips et al., 2005) and the CHARMM36 force field (Best et al., 2012; Huang and MacKerell, 2013; Vanommeslaeghe and MacKerell, 2015) and served as a reference model for aligning other combinations of hs TMC1 or hs TMC2 with hs CIB2 or hs CIB3. All models were either embedded into pure POPC or mixed membrane lipid bilayers using CHARMM-GUI (Jo et al., 2009; Jo et al., 2008; Jo et al., 2007; Lee et al., 2019) as summarized in Figure 8—source data 1. The mixed membrane mimicked stereocilia membrane composition (Zhao et al., 2012). The protein-lipid system was then solvated using TIP3P explicit water and neutralized with 0.150 M KCl, representative of endolymph ionic concentration (Bosher and Warren, 1978).

All-atom MD simulations were performed and analyzed using NAMD 2.13 (Phillips et al., 2005) and the CHARMM36 force field with CMAP corrections (Best et al., 2012; Huang and MacKerell, 2013; Vanommeslaeghe and MacKerell, 2015). For all systems, equilibrium simulations consisted of 2000 steps of minimization, 0.5 ns of dynamics with everything excluding lipid tails fixed/constrained, 0.5 ns of dynamics with harmonic constraints applied to the protein (1 kcal mol^–1 Å^–2), 1 ns of free dynamics in the NpT ensemble (γ = 1 ps^–1), and up to 100 ns of free dynamics in the NpT ensemble (γ = 0.1 ps^–1) prior to non-equilibrium production runs. A piston period of 200 fs and a damping timescale of 50 fs was used for pressure control and atomic coordinates were saved every 5 ps. All simulations were performed at 310 K and 1 atmosphere of pressure by using the Langevin thermostat and Nosé-Hoover Langevin pressure control. A 12 Å cutoff distance was employed with a force-based switching function starting at 10 Å. Periodic boundary conditions and the PME method were used to calculate long-range electrostatic interactions with a grid density greater than 1 Å^–3. An integration timestep of 2 fs with electrostatic interactions computed every other timestep was utilized for every simulation in tandem with the SHAKE algorithm for the constraint of hydrogen atoms. A constant electric field was applied to all atoms in non-equilibrium simulations with transmembrane potential (Gumbart et al., 2012).

We used Anton 2, a massively parallel special purpose machine for MD simulations (Shaw et al., 2014), to run simulations of hs TMC1 + hs CIB2-Ca²⁺ in POPC (simulations S1d-g; Figure 8—source data 1), hs TMC1 + hs CIB2 in POPC (simulations S2d-2g; Figure 8—source data 1), hs TMC1 in POPC (simulations S3d-g; Figure 8—source data 1), and hs TMC1 + hs CIB2-Ca²⁺ in a mixed membrane (simulations S4c-d; Figure 8—source data 1). Coordinates of the pre-equilibrated systems were converted to an Anton 2 compatible dms file format using the NAMDtoDMS.py script provided by the Pittsburgh Supercomputing Center. Simulations used CHARMM36 forcefields and were performed in the NpT ensemble (310 K, 1 atm) using the multigrator integration framework (Lippert et al., 2013) with a 2.5 fs integration timestep and saved every 120 ps. The MTK barostat and Nosé-Hoover thermostat was updated every 480 and 24 steps, respectively. Semi-isotropic pressure coupling was used. Long range electrostatic interactions were calculated using u-series electrostatics with a 64×64×64 grid. Van der Waals interactions were calculated with a cutoff of 12 Å. Transmembrane voltages were generated by applying a constant electric field to all atoms (Gumbart et al., 2012).

BSA was computed in VMD by measuring SASA for both TMC and CIB monomers, then subtracting SASA for the complex. BSA for interacting molecules A and B were computed as BSAAB=1/2(SASAA+SASAB−SASAAB). RMSD values were computed in VMD using Cα atoms and initial models as references, with alignments carried out using the corresponding subdomains when indicated. VMD was used to analyze trajectories, render molecular images, and create videos. Xmgrace and Gnuplot were used to generate plots. Sequences for CIB and TMC isoforms from various species were obtained from the NCBI protein database. Sequences were aligned utilizing the MUSCLE algorithm within Geneious (Kearse et al., 2012). Aligned sequences were loaded into JalView (Waterhouse et al., 2009) and colored based on a sequence identity threshold of 45%. Ionic currents were computed by counting the number of crossing, with conductance values estimated as in Walujkar et al., 2021.