FCHSD2 is required for stereocilia maintenance in mouse cochlear hair cells

Hair cells are the auditory and vestibular receptor cells in the inner ear. They are responsible for converting mechanical signals into electrical signals – a process that is known as mechano-electrical transduction (MET) (Fettiplace, 2017; Gillespie and Müller, 2009; Maoileidigh and Ricci, 2019). Hair cells are characterized by their hair bundles, which consist of dozens to hundreds of F-actin-based stereocilia and one microtubule-based kinocilium on the apical surface of each hair cell. Stereocilia are organized into several rows of increasing heights, forming a U-shaped or V-shaped staircase-like pattern. There are typically three rows of stereocilia in each mouse cochlear hair cell, usually referred to as the first (tall), second (middle) and third (short) rows. The kinocilium localizes at the vertex of each hair bundle, juxtaposed to the first-row stereocilia. The kinocilium plays important roles in hair bundle development but is not necessary for MET, and it is lost when cochlear hair cells mature (Hudspeth and Jacobs, 1979; Jones et al., 2008). In contrast, stereocilia play a pivotal role in MET (Hudspeth and Jacobs, 1979). Deflection of stereocilia towards the taller edge increases the open probability of MET channels at the tips of stereocilia in the shorter rows, which then leads to the MET currents (Beurg et al., 2009; Hudspeth and Jacobs, 1979).

The actin filaments in each stereocilium are interconnected by various actin crosslinkers – such as PLS1, FSCN2, ESPN and TRIOBP, among others (Kitajiri et al., 2010; Krey et al., 2016; Shin et al., 2010; Taylor et al., 2015; Zheng et al., 2000) – and are tightly packed into a paracrystalline array with barbed (plus) ends pointing towards the stereociliary tips (Flock and Cheung, 1977; Tilney et al., 1980). Stereocilia develop into the final mature structure in a multistep, precisely determined order by increasing the length and number of actin filaments (Krey et al., 2020; Tilney et al., 1992). In mature stereocilia, the F-actin core is quite stable, and actin monomer exchange only happens at the stereociliary tips (Drummond et al., 2015; Narayanan et al., 2015; Zhang et al., 2012). Consequently, stereocilia development and maintenance rely on the precise regulation of F-actin core assembly. The importance of stereociliary F-actin assembly is highlighted by the large number of deafness genes that encode actin or actin-regulatory proteins (Barr-Gillespie, 2015; Miyoshi et al., 2021).

The detailed mechanisms of stereocilia development and maintenance are being unraveled, and several proteins have been identified to play essential roles in these processes (Barr-Gillespie, 2015; Krey and Barr-Gillespie, 2019; Nayak et al., 2007; Vélez-Ortega and Frolenkov, 2019). One of these proteins is cell division cycle 42 (CDC42), a key regulator of the actin cytoskeleton. CDC42, Rac1 and RhoA belong to the Rho family of small GTPases and are all expressed in cochlear hair cells (Kalinec et al., 2000; Liu et al., 2014; Scheffer et al., 2015). CDC42 has been shown to stimulate ARP2/3-dependent actin polymerization in Xenopus egg extracts mainly through binding to actin nucleator N-WASP (also known as WASL; Rohatgi et al., 1999). Previously, we and others have shown that CDC42 is localized in cochlear hair cell stereocilia, and that Cdc42 gene knockout affects stereocilia development and maintenance in mice (Du et al., 2021; Kirjavainen et al., 2015; Ueyama et al., 2014).

FCH domain and double SH3 domains-containing protein 2 (FCHSD2) is a member of the FCH and Bin-Amphiphysin-Rvs (F-BAR) protein family, which consists of cytosolic proteins that regulate actin polymerization, membrane curvature and cell motility (Carman and Dominguez, 2018; Salzer et al., 2017). Previously, we showed that FCHSD2 is localized along the hair cell stereocilia and is able to stimulate WASP–ARP2/3-mediated actin polymerization in vitro, suggesting that FCHSD2 might play a role in stereocilia development and/or maintenance (Cao et al., 2013). Recently, we reported that FCHSD2 cooperates with CDC42 and N-WASP to regulate cell protrusion formation in COS-7 cells (Zhai et al., 2022). In the present work, we establish two Fchsd2 knockout mouse lines and show that loss of FCHSD2 leads to deficits in stereocilia maintenance as well as hearing loss. Furthermore, Fchsd2/Cdc42 double knockout mice show more severe stereocilia deficits and hearing loss, suggesting that FCHSD2 and CDC42 might cooperatively regulate stereocilia maintenance in mice.

The mouse Fchsd2 gene has 21 exons – with the translational start codon localized in exon 1 and the translational stop codon localized in exon 21 – and encodes a protein of 749 amino acids. Exon 11 is subjected to alternative splicing, which gives rise to two FCHSD2 isoforms with or without an insertion of 24 amino acids (NP_950177.2 and NP_001139482.1). We introduced two loxP sequences on each side of exon 2 through homologous recombination and obtained Fchsd2^loxP/+ mice (Fig. 1A). By crossing Fchsd2^loxP/+ mice with EIIa-Cre^+/− transgenic mice, we obtained EIIa-Cre^+/−; Fchsd2^loxP/loxP mice, in which exon 2 of the Fchsd2 gene is ubiquitously deleted in all somatic and germline cells (Fig. 1A). Thereafter, all of the homozygous offspring of these mice have exon 2 of the Fchsd2 gene deleted and are therefore referred to as Fchsd2^−/− in the following text. Exon 2 consists of 98 bp, and its deletion causes a premature translational stop in the Fchsd2 transcript. Western blotting using a custom antibody raised against amino acids 1–631 of mouse FCHSD2 was employed to examine the expression of FCHSD2 in Fchsd2^−/− mice. The predicted molecular mass of mouse FCHSD2 is 84.3 kDa. A band with a molecular mass of ∼90 kDa was detected in all the tissues examined of wild-type and Fchsd2^+/− mice, including the inner ear, but this band was not detected in Fchsd2^−/− mice (Fig. 1B). Several additional weak bands were also detected, which might result from unspecific binding of the antibody (Fig. 1B). Furthermore, whole-mount immunostaining of cochlear sensory epithelia showed that FCHSD2 immunoreactivity is localized on the perimeter of the stereociliary F-actin core in Fchsd2^+/− mice, but not in Fchsd2^−/− mice (Fig. 1C–C″). Taken together, the western blot and immunostaining results confirm that Fchsd2^−/− mice are devoid of FCHSD2 protein in the inner ear.

The Fchsd2^−/− mice were found to be morphologically and behaviorally indistinguishable from Fchsd2^+/− or wild-type mice. Furthermore, interbreeding of Fchsd2^+/− mice gave rise to offspring in the expected Mendelian ratio (25% wild type, 47.7% Fchsd2^+/− and 27.3% Fchsd2^−/−; n=44; P=0.967; χ² test) with normal viability, and both male and female Fchsd2^−/− mice were found to be fertile. These results suggest that FCHSD2 is dispensable for general development and reproduction in mice. Auditory brainstem response (ABR) measurements were then performed to evaluate the auditory function of Fchsd2^−/− mice. When examined at age of 1 month or 2 months, the ABR thresholds of Fchsd2^−/− mice to click stimuli were not significantly different from those of Fchsd2^+/− mice (Fig. 1D). However, a 20 decibels (dB) ABR threshold elevation was observed in Fchsd2^−/− mice at an age of 3 months (Fig. 1D). The hearing threshold elevation of Fchsd2^−/− mice reached 40 dB at an age of 9 months, and Fchsd2^−/− mice were profoundly deaf, with a hearing threshold above 90 dB sound pressure level (SPL), by 12 months of age (Fig. 1D). Moreover, our data show that ABR thresholds of Fchsd2^−/− mice to pure-tone stimuli are elevated in all frequencies when examined at an age of 5 months (Fig. 1F). Taken together, the ABR measurements suggest that Fchsd2^−/− mice suffer progressive hearing loss. Our present data also reveal that the hearing thresholds are not significantly different between Fchsd2^+/− mice and Fchsd2^+/+ mice (Fig. 1E), therefore Fchsd2^+/− mice were used as controls in the following work.

We also established another Fchsd2 knockout mouse line using the clustered regularly interspaced short palindromic repeat (CRISPR)-associated Cas9 nuclease (CRISPR/Cas9) genome editing technique. A deletion of 37 bp was introduced into exon 2, which causes a frameshift and premature translational stop in the Fchsd2 gene (Fig. S1A). The resulting mouse line is referred to as Fchsd2^+/Δ37 hereafter. Western blotting and whole-mount immunostaining confirmed that FCHSD2 protein is absent in Fchsd2^Δ37/Δ37 mice (Fig. S1B and data not shown). Consistent with the results obtained from Fchsd2^−/− mice, Fchsd2^Δ37/Δ37 mice were found to be developmentally normal but manifest progressive hearing loss (Fig. S1C,D).

Scanning electron microscopy (SEM) was performed to examine the morphology of hair cell stereocilia in the Fchsd2 knockout mice. Stereocilia of both inner hair cells (IHCs) and outer hair cells (OHCs) in 1-month-old Fchsd2^−/− mice were found to have intact stereocilia similar to those of Fchsd2^+/− control mice (Fig. S2). At an age of 2 months, however, complete stereocilia loss could be observed in some Fchsd2^−/− OHCs at the basal turn (Fig. 2A,C). Complete OHC stereocilia loss became more severe in 3-month-old Fchsd2^−/− cochlea and extended from the basal to the middle turn (Fig. 2B,C). Meanwhile, IHC stereocilia were found to largely remain intact in Fchsd2^−/− cochlea (Fig. 2A,B,D).

Hair cell survival was then examined by performing whole-mount immunostaining using an antibody against the hair cell marker myosin VIIA (MYO7A) (Hasson et al., 1997). There was no obvious hair cell loss in Fchsd2^−/− mice at 1 month, but significant OHC loss was observed by an age of 3 months at the basal and middle turns in Fchsd2^−/− mice (Fig. 2E,F). The percentage of OHC loss was ∼36% at the basal turn of 3-month-old Fchsd2^−/− mice, compared to 4% in age-matched control Fchsd2^+/− mice. By an age of 5 months, OHC loss became more severe in Fchsd2^−/− mice, whereas IHCs were still largely unaffected at this age (Fig. 2E,F).

We then used SEM imaging to examine the stereocilia in more detail. By an age of 2 months, degeneration of third-row stereocilia could be observed in some OHCs at middle and basal turns (Fig. 3A,C). The third-row stereocilia degeneration became more severe and extended to all the three turns by 3 months of age (Fig. 3B,C). Meanwhile, IHC stereocilia remained intact in both 2-month-old and 3-month-old Fchsd2^−/− mice (Fig. 3A,B). Similar phenotypes were observed in Fchsd2^Δ37/Δ37 mice (Fig. S1E–H). We also measured the length of first-row stereocilia in IHCs and OHCs, which did not reveal any significant difference between 3-month-old Fchsd2^−/− and Fchsd2^+/− mice (Fig. 3D). Taken together, our data suggest that knockout of Fchsd2 affects stereocilia maintenance in OHCs.

It has been shown that stereocilia deficits are associated with noise-induced hearing loss (Avinash et al., 1993; Engström et al., 1983; Liberman, 1987; Liberman and Dodds, 1987; Tilney et al., 1982). We examined whether Fchsd2 knockout mice are more vulnerable to noise insults. Fchsd2^−/− and Fchsd2^+/− mice at an age of 1 month were exposed to a broadband noise of 2–20 kHz at 106 dB SPL for 2 h, and the ABR thresholds were measured at different time points before and after the noise exposure (Fig. 4A). Similar protocols have been shown to induce a slight or moderate permanent threshold shift (PTS) at 2 weeks after noise exposure in mice (Chen et al., 2012; Gong et al., 2012). During the experiment, some of the mice were euthanized, and the stereocilia and ribbon synapses were examined using SEM and immunostaining at certain time points (Fig. 4A). ABR measurements show that the noise exposure induced a slight PTS in control Fchsd2^+/− mice (Fig. 4B). However, the same noise exposure induced significantly increased PTS in Fchsd2^−/− mice, suggesting that Fchsd2^−/− mice are more vulnerable to noise insults (Fig. 4B).

SEM analysis revealed that at 2 h after noise exposure, the stereocilia morphology was largely normal in Fchsd2^−/− mice (Fig. 4C). However, significant OHC stereocilia loss at middle and basal turns was observed in Fchsd2^−/− mice at 4 days after noise, which became more severe at 14 days after noise (Fig. 4C,D). Meanwhile, IHC stereocilia in Fchsd2^−/− mice largely remained intact at all the time points examined (Fig. 4C). As a control, stereocilia in both OHCs and IHCs were largely unaffected by the noise in Fchsd2^+/− mice (Fig. 4C,D). Consistently, phalloidin and DAPI staining showed that loss of OHCs, but not of IHCs, occurred at 4 days after noise exposure in Fchsd2^−/− mice, and this loss was exaggerated at 14 days after exposure to noise (Fig. 4E; Fig. S3).

High-magnification SEM imaging was then used to examine the stereocilia in more detail. Increased third-row stereocilia degeneration was detected in OHCs of Fchsd2^−/− mice at 2 h after noise exposure, especially at the basal turn, and this degeneration became more severe at 4 and 14 days after noise exposure (Fig. 5A,D). Meanwhile, first-row stereocilia degeneration was also observed in Fchsd2^−/− OHCs following noise exposure (Fig. 5A,C). IHCs of Fchsd2^−/− mice were found to be largely unaffected (Fig. 5B). Taken together, our data suggest that exposure to noise causes robust OHC stereocilia degeneration, which may eventually lead to increased OHC loss and hearing threshold elevation in Fchsd2 knockout mice.

Loss of IHC ribbon synapses has been suggested to be associated with noise-induced hearing loss (Kobel et al., 2017; Kujawa and Liberman, 2009; Liberman, 2017). We examined the ribbon synapses by performing whole-mount immunostaining using antibodies to detect CtBP2 and GluR2 (also known as GRIA2). The results showed that in control mice, the numbers of double-positive ribbon synapses decreased at 2 h after noise but recovered to the normal level by 14 days after noise exposure (Fig. S4). Interestingly, similar recovery of IHC ribbon synapses was observed in Fchsd2^−/− mice (Fig. S4), suggesting that ribbon synapse deficits do not contribute to the severe PTS observed in Fchsd2^−/− mice.

Nervous wreck (Nwk), the Drosophila homolog of FCHSD2, shares 41% amino acid identity for the F-BAR domain and 54% amino acid identity for the SH3 domains with FCHSD2 (Cao et al., 2013). It has been shown that Nwk cooperates with Cdc42 to regulate actin polymerization and synaptic growth (Rodal et al., 2008). Nwk and Cdc42 colocalize in periactive zones of the neuromuscular junction, and cdc42; nwk double mutants show an additive effect on synaptic growth (Rodal et al., 2008). Moreover, FCHSD2 cooperates with CDC42 to regulate cell protrusion formation in COS-7 cells (Zhai et al., 2022). CDC42 has been shown to localize in the stereocilia, and Cdc42 knockout in mouse leads to stereocilia deficits such as third-row stereocilia degeneration, extra fourth-row stereocilia, as well as dysregulated planer cell polarity (Du et al., 2021; Kirjavainen et al., 2015; Ueyama et al., 2014). Therefore, we next examined whether FCHSD2 and CDC42 could cooperate in mammalian hair cells by evaluating the auditory function of the corresponding knockout mice. By an age of 1 month, when Fchsd2^−/− mice still show normal ABR threshold, Atoh1-Cre^+/−; Cdc42^loxP/loxP mice exhibited a 20 dB threshold elevation to click stimuli (Fig. 6A). Interestingly, loss of one copy of the Cdc42 gene in a Fchsd2^−/− background or loss of one copy of the Fchsd2 gene in an Atoh1-Cre^+/−; Cdc42^loxP/loxP background resulted in threshold elevation of 30 and 40 dB, respectively (Fig. 6A). Furthermore, Atoh1-Cre^+/−; Cdc42^loxP/loxP; Fchsd2^−/− double knockout mice showed the highest threshold elevation, which was nearly 60 dB (Fig. 6A). The hearing thresholds of knockout mice were further elevated at an age of 2 months (Fig. S5). These results suggest that FCHSD2 and CDC42 cooperate to regulate auditory function in mice.

SEM was then employed to examine the cooperative role of FCHSD2 and CDC42 in stereocilia maintenance. At an age of 1 month, Cdc42^loxP/loxP and Cdc42^loxP/loxP; Fchsd2^−/− mice were found to have intact stereocilia, whereas Atoh1-Cre^+/−; Cdc42^loxP/loxP mice exhibited significant OHC stereocilia loss (Fig. 6B,C). Loss of one copy of the Cdc42 gene in a Fchsd2^−/− background or vice versa resulted in more robust OHC stereocilia loss, whereas Atoh1-Cre^+/−; Cdc42^loxP/loxP; Fchsd2^−/− double knockout mice showed the most severe OHC stereocilia loss (Fig. 6B,C). Our data reveal that Cdc42 knockout also affects IHC stereocilia maintenance, and we found that this effect is exaggerated by Fchsd2 knockout (Fig. 6B,D). Consistently, more robust hair cell loss was observed in Atoh1-Cre^+/−; Cdc42^loxP/loxP; Fchsd2^−/− double knockout mice than in each single knockout mouse line (Fig. 7A–C).

We then examined the detailed stereocilia morphology of double knockout mice at an age of 1 month using higher magnification SEM. In Cdc42^loxP/loxP; Fchsd2^−/− mice, OHC stereocilia were found to be nicely organized into three rows of increasing heights, similarly to those of Fchsd2^−/− mice (Fig. S2B) and Cdc42^loxP/loxP mice (Fig. 8A). However, sporadic degeneration of individual third-row stereocilia was observed in Atoh1-Cre^+/−; Cdc42^loxP/loxP OHCs (Fig. 8A,C). OHC stereocilia degeneration was exaggerated in Atoh1-Cre^+/−; Cdc42^loxP/loxP; Fchsd2^+/− OHCs as well as in Atoh1-Cre^+/−; Cdc42^loxP/+; Fchsd2^−/− OHCs, and the most robust stereocilia degeneration was observed in Atoh1-Cre^+/−; Cdc42^loxP/loxP; Fchsd2^−/− OHCs (Fig. 8A,C). Stereocilia fusion could be observed in some Atoh1-Cre^+/−; Cdc42^loxP/loxP IHCs, and this fusion was exacerbated in Atoh1-Cre^+/−; Cdc42^loxP/loxP; Fchsd2^−/− IHCs (Fig. 8B). Taken together, our data suggest that FCHSD2 and CDC42 cooperate to regulate stereocilia maintenance and auditory function in mice.

Stereocilia development and maintenance rely on the precise regulation of F-actin core assembly. With genetic and proteomic advances, many proteins have been identified to play essential roles in this process (Barr-Gillespie, 2015; Krey and Barr-Gillespie, 2019; Nayak et al., 2007; Vélez-Ortega and Frolenkov, 2019). Previously, we have shown that FCHSD2 is localized along the stereocilia of auditory hair cells and is able to promote cell protrusion formation in cultured cells, suggesting that it might play a role in stereocilia development and/or maintenance (Cao et al., 2013; Zhai et al., 2022). In the present work, we established Fchsd2 knockout mice and showed that FCHSD2 is indispensable for stereocilia maintenance. As a result, Fchsd2 knockout mice suffer from progressive hearing loss.

FCHSD2 is a member of the BAR protein superfamily, which is characterized by the presence of the BAR domain that promotes membrane curvature in antiparallel homodimers (Peter et al., 2004; Simunovic et al., 2019). BAR proteins are categorized into three major subfamilies, namely N-BAR, F-BAR and I-BAR proteins, according to the sequence homology and structure similarity of the BAR domains (Carman and Dominguez, 2018). The classical crescent N-BAR domain and the more extended and less curved F-BAR domain usually promote membrane invaginations, whereas the inverse curvature I-BAR domain usually generates membrane protrusions (Carman and Dominguez, 2018). FCHSD2 has long been considered to be an F-BAR protein. Evidence from others and our group show that overexpression of the FCHSD2 F-BAR domain promotes formation of plasma membrane protrusions in cultured cells (Almeida-Souza et al., 2018; Becalska et al., 2013; Kelley et al., 2015; Zhai et al., 2022). Our present work shows that FCHSD2 is indispensable for the maintenance of stereocilia, a special type of actin-based, microvilli-like cell protrusions in inner-ear hair cells, further supporting that FCHSD2 behaves as an atypical F-BAR protein. A similar scenario has been reported for SRGAP1, SRGAP2 and SRGAP3, whose F-BAR domains induce cell protrusions in cultured cells and are therefore referred to as inverse F-BAR (IF-BAR) domains (Carlson and Soderling, 2009; Carlson et al., 2011; Coutinho-Budd et al., 2012; Guerrier et al., 2009). Our results suggest that FCHSD2 could be considered as another member of the IF-BAR protein subfamily.

In Fchsd2 knockout mice, stereocilia were intact in both OHCs and IHCs when examined at an age of 1 month. Stereocilia degeneration was first observed in OHCs of 2-month-old Fchsd2 knockout mice, when hearing impairment had not yet manifested. Noticeably, we found that stereocilia degeneration in Fchsd2 knockout OHCs is mostly limited to the third-row stereocilia, leaving the first- and second-row stereocilia largely unaffected. The degeneration of the mechanosensitive third-row stereocilia might affect the MET function of OHCs. Complete OHC stereocilia loss as well as OHC loss became apparent by an age of 3 months, which is consistent with the progressive hearing loss observed in Fchsd2 knockout mice. Mice lacking EPS8L2, MYO15A-L (the long isoform of MYO15A), BAIAP2L2 or PJVK have been reported to show similar third-row stereocilia degeneration in both OHCs and IHCs (Carlton et al., 2021; Fang et al., 2015; Furness et al., 2013; Kazmierczak et al., 2017; Yan et al., 2022). Different from the ubiquitous expression pattern of FCHSD2 along the stereocilia, EPS8L2, MYO15A-L and BAIAP2L2 are localized at the tips of shorter-row stereocilia, whereas PJVK is suggested to mainly localize at the stereociliary rootlet (Carlton et al., 2021; Fang et al., 2015; Furness et al., 2013; Kazmierczak et al., 2017; Yan et al., 2022). It awaits further investigation whether FCHSD2 could functionally interact with these proteins to regulate stereocilia maintenance.

Stereocilia damage has been known as a hallmark of permanent noise-induced hearing loss for decades (Liberman, 1987). Noise exposure can induce stereocilia deficits ranging from stereocilia degeneration and fusion to rootlet breakage (Avinash et al., 1993; Engström et al., 1983; Tilney et al., 1982). Generally speaking, mild stereocilia changes are associated with temporary threshold shift-inducing noise exposure, whereas severe stereocilia damage is associated with PTS-inducing noise exposure (Liberman, 1987; Liberman and Dodds, 1987). Consistent with the potential role of FCHSD2 in stereocilia maintenance, we found that Fchsd2 knockout mice show increased vulnerability to acoustic stimuli. Upon the same acoustic stimuli, Fchsd2 knockout mice developed more severe PTS compared with that of control mice. Further examination revealed that noise exposure induced temporary IHC ribbon synapse loss in both Fchsd2 knockout and control mice, whereas noise exposure caused OHC stereocilia degeneration as well as permanent OHC loss in Fchsd2 knockout mice only. These results suggest that Fchsd2 knockout mice show increased OHC stereocilia degeneration and more profound PTS upon acoustic stimulation, further supporting an important role of FCHSD2 in stereocilia maintenance. Interestingly, besides the third-row stereocilia, the tallest first-row stereocilia in Fchsd2^−/− OHCs also degenerated after acoustic stimulation. First-row stereocilia degeneration has been reported in OHCs deficient for the TRIOBP-5 isoform of TRIOBP, which localizes at the stereocilia rootlet (Katsuno et al., 2019). Whether FCHSD2 plays a role in stereocilia rootlet awaits further investigation.

As a BAR protein, FCHSD2 contains a N-terminal BAR domain, two SH3 domains and a C-terminal PDZ domain-binding interface (PBI). The BAR domain of FCHSD2 mediates protein dimerization as well as plasma membrane binding, while SH3 domains interact with other proteins such as ITSN1, WASP and N-WASP (Almeida-Souza et al., 2018; Cao et al., 2013). The PBI of FCHSD2 interacts with PDZ domain-containing proteins such as MAGI-1 and CASK (Ohno et al., 2003). Recently, we have shown that FCHSD2 and CDC42 simultaneously bind to N-WASP and cooperatively regulate actin-based cell protrusion formation in COS-7 cells (Zhai et al., 2022). Here, our results suggest that FCHSD2 and CDC42 may also cooperatively regulate stereocilia maintenance in mouse cochlear hair cells. CDC42 has been shown to localize along the hair cell stereocilia, and Cdc42 knockout affects stereocilia development and maintenance in mice (Du et al., 2021; Kirjavainen et al., 2015; Ueyama et al., 2014). Our present data show that simultaneous knockout of Fchsd2 and Cdc42 causes more severe stereocilia degeneration as well as more profound hearing loss in mice, suggesting that they may work together to regulate stereocilia maintenance.

FCHSD2 and CDC42 act on the same downstream target proteins – WASP or N-WASP – to regulate ARP2/3-dependent F-actin assembly and cell protrusion formation (Alekhina et al., 2017; Zhai et al., 2022). Proteomic screening studies have revealed the existence of several ARP2/3 subunits in the stereocilia, although their precise localization remains unknown (Krey et al., 2015; Shin et al., 2013). WASP and N-WASP have not been detected in the stereocilia, but low levels of mRNA expression in the hair cells have been reported in the gEAR and SHIELD database (Orvis et al., 2021; Shen et al., 2015). It is noteworthy that the stereocilia are based on unbranched, parallel F-actin, whereas WASP and N-WASP usually activate ARP2/3 and mediate branched F-actin polymerization (Alekhina et al., 2017). Therefore, whether FCHSD2 and CDC42 regulate stereocilia maintenance through WASP or N-WASP and ARP2/3 awaits further investigation. Alternatively, our data reveal that FCHSD2 immunoreactivity is localized at the perimeter of the stereociliary F-actin core and inside the plasma membrane. Therefore, FCHSD2 might regulate stereocilia stability by bridging between the stereociliary F-actin core and the plasma membrane through its C-terminal domains and its N-terminal BAR domain, respectively.

Our data show that stereocilia morphology in Fchsd2 knockout mice is largely normal until 2 months of age and that only OHCs are affected by Fchsd2 knockout, suggesting that other proteins might compensate for the loss of FCHSD2 in Fchsd2 knockout mice to some extent. A good candidate is FCHSD1, the homolog of FCHSD2, which is also expressed in the auditory hair cells (Cao et al., 2013). Although FCHSD1 has lower binding affinity toward WASP and N-WASP when compared with that of FCHSD2, it can efficiently stimulate WASP–ARP2/3-mediated F-actin polymerization through binding to sorting nexin 9 (SNX9) (Cao et al., 2013). Fchsd1/Fchsd2 double knockout mice would certainly help us to learn more about the mechanism of how FCHSD1 and FCHSD2 regulate stereocilia development and/or maintenance.

Animal experiments were approved by the Animal Ethics Committee of Shandong University School of Life Sciences (permit number SYDWLL-2017-05) and were performed accordingly. Fchsd2^loxP/+ mice in C57BL/6 background were generated by Shanghai Biomodel Organism Science & Technology Development Co., Ltd. Briefly, a targeting vector was constructed to insert two loxP sites flanking exon 2 of the mouse Fchsd2 gene. Mouse Pgk1 promoter-driven neomycin resistance gene and herpes simplex virus thymidine kinase gene were included for positive and negative selection, respectively. The targeting vector was electroporated into embryonic stem cells (ESCs), and G418 and ganciclovir double-resistant colonies were screened by performing genotyping PCR (Fchsd2-F1, 5′-CAGTCTCATGGACTTCCTTTTTGGC-3′; Neo-R1, 5′-GGCCTACCCGCTTCCATTGCTC-3′; Neo-F2, 5′-TCGCATTGTCTGAGTAGGTGTC-3′; Fchsd2-R2, 5′-AAGGGAAAGAGGGAGGAAAGGAACCA-3′). Homologous recombinant ESCs were microinjected into C57BL/6 female mice to obtain heterozygous Fchsd2^loxP/+ mice, which were then crossed with EIIa-Cre transgenic mice (Lakso et al., 1996) in a mixed background to eventually obtain Fchsd2^−/− mice.

Fchsd2^+/Δ37 mice in C57BL/6 background were generated using the CRISPR/Cas9 genome editing technique by Shanghai Biomodel Organism Science & Technology Development Co., Ltd. Briefly, three small guide RNAs (sgRNAs) that target different sites were in vitro transcribed (site-1, 5′-CAAGAGCTGAGAAACATTCAAGG-3′; site-2, 5′-AAAACTTCAAGCCAAACATCAGG-3′; site-3, 5′-GTCACATTCTGCCTGATGTTTGG-3′) in exon 2 of the mouse Fchsd2 gene. The sgRNAs were mixed with Cas9 mRNA and injected into the cytoplasm of C57BL/6J zygotes. After culturing in vitro for 24 h, the injected zygotes were transferred into the oviduct of a pseudopregnant ICR female mouse at 0.5 days post coitus (dpc) to give rise to F0 mice. To determine the sequences of mutant alleles, F0 mice were examined by performing genotyping PCR (Fchsd2-F, 5′-GGTCTTTGGGGATAGGAAAAC-3′; Fchsd2-R, 5′-CAGATACAGAAGTTCTCCAGC-3′) and Sanger sequencing. F0 mice were crossed with C57BL/6J mice to obtain F1 Fchsd2^+/Δ37 mice. Fchsd2^+/Δ37 mice were then crossed with wild-type mice in a mixed background, and the offspring were used for the functional studies.

Atoh1-Cre^+/−; Cdc42^loxP/loxP mice were obtained by crossing Cdc42^loxP/+ mice (van Hengel et al., 2008; Wu et al., 2006) in C57BL/6 background with Atoh1-Cre^+/− mice in a mixed background that express Cre recombinase under the control of Atoh1 promoter (Yang et al., 2010).

Rabbit anti-FCHSD2 polyclonal antibody was developed by Abmart (Shanghai, China). Briefly, mouse FCHSD2 (1–631 amino acids) was expressed as GST fusion protein in Escherichia coli and purified using Glutathione Sepharose 4B (GE Healthcare), as described previously (Cao et al., 2013). The fusion protein was then used to immunize two rabbits, and IgG was purified using Protein A agarose affinity chromatography. The dilutions used in immunostaining and western blotting were 1:200 and 1:1000, respectively. Prestained protein molecular mass marker was purchased from Thermo Fisher Scientific (cat. no. 26612). Mouse anti-GAPDH polyclonal antibody was purchased from Millipore (cat. no. MAB374; 1:5000 dilution for western blotting). Rabbit anti-MYO7A polyclonal antibody was purchased from Proteus Biosciences (cat. no. 25-6790; 1:400 dilution for immunostaining). Mouse anti-CtBP2 antibody (IgG1) was purchased from BD (cat. no. 612044; 1:400 dilution for immunostaining). Mouse anti-GluR2 antibody (IgG2a) was purchased from Millipore (cat. no. MAB397; 1:1000 dilution for immunostaining). Alexa Fluor 488-conjugated donkey anti-rabbit IgG secondary antibody was purchased from Thermo Fisher Scientific (cat. no. A21206; 1:400 dilution for immunostaining). Alexa Fluor 568-conjugated goat anti-mouse IgG1 was purchased from Invitrogen (cat. no. A21124; 1:400 dilution for immunostaining). Alexa Fluor 488-conjugated goat anti-mouse IgG2a was purchased from Invitrogen (cat. no. A21131; 1:400 dilution for immunostaining). HRP-conjugated goat anti-rabbit IgG secondary antibody was purchased from Bio-Rad (cat. no. 170-6515; 1:3000 dilution for western blotting). HRP-conjugated goat anti-mouse IgG secondary antibody was purchased from Bio-Rad (cat. no. 170-6516; 1:5000 dilution for western blotting). TRITC-conjugated phalloidin was purchased from Sigma-Aldrich (cat. no. P1951; 1:1000 dilution for staining). SF488-conjugated phalloidin was purchased from Solarbio (cat. no. CA1640; 1:1000 dilution for staining). DAPI was purchased from Gene View Scientific Inc. (cat. no. GD3408; 1:3000 dilution for staining).

For western blotting, dissected tissues or cells were homogenized and lysed in ice-cold RIPA buffer (Beyotime, P0013B) with 1 mM PMSF and 1× protease inhibitor cocktail (Roche). After centrifugation, the supernatant was collected and separated by polyacrylamide gel electrophoresis (PAGE), then transferred to PVDF membrane. After blocking in TBS (20 mM Tris-HCl, 150 mM NaCl) containing 5% non-fat dry milk and 0.05% Tween-20, the membrane was incubated with primary antibody at 4°C overnight, followed by incubation with HRP-conjugated secondary antibody at room temperature for 1 h. The signals were detected with the ECL system (Thermo Fisher Scientific) (Fig. S6).

For whole-mount immunostaining, organ of Corti explants were dissected and fixed with 4% paraformaldehyde (PFA) in PBS for 30 min, followed by permeabilization with 0.5% Triton X-100 for 20 min and blocking with 5% donkey serum (Solarbio, cat. no. SL050) for 1 h at room temperature. For FCHSD2 staining, the samples were additionally incubated in 10% methanol for 15 min at −20°C before permeabilization. Samples were then incubated with primary antibody in PBS overnight at 4°C, and with secondary antibody in PBS for 1 h at room temperature. For stereocilia visualization, samples were incubated with phalloidin in PBS for 30 min at room temperature. For nuclei visualization, samples were incubated with DAPI in PBS for 1 h at room temperature. For CtBP2 and GluR2 staining, samples were incubated with primary and secondary antibodies at 37°C overnight or for 1 h, respectively. Afterwards, samples were mounted in PBS and glycerol (1:1 mixture) and imaged with a confocal microscope (LSM 700 or 880, Zeiss, Germany) with a 1.25 NA/Apo 63× M27 or 1.4 NA/Apo 100× M27 objective. OHCs and IHCs were counted along successive 0.2 mm intervals at different cochlear turns by a second person blind to the experimental conditions. A hair cell was considered as missing if MYO7A immunoreactivity or the phalloidin and DAPI signals were absent.

ABR measurement was performed as described previously (Du et al., 2020). Stimulus generation, presentation, ABR acquisition and data management were achieved using a RZ6 workstation and BioSig software (Tucker Davis Technologies, Inc.). Mice were anesthetized by intraperitoneal injection of pentobarbital (8.4 mg/100 g body weight). The electrodes were then put subcutaneously at the vertex, the pinna and near the tail. Acoustic stimuli (clicks or pure-tone bursts) of sound level decreasing from 90 dB SPL in 10 dB SPL steps were delivered to the mouse ear through an earphone. Hearing threshold was indicated by the lowest sound level at which all ABR waves were detectable.

Mouse temporal bones were dissected and fixed with 2.5% glutaraldehyde in 0.1 M phosphate buffer at 4°C overnight. The cochleae were dissected out of the temporal bone and post-fixed with 1% osmium tetroxide in 0.1 M phosphate buffer at 4°C for 2 h. The samples were dehydrated in ethanol, then underwent critical point drying using a Leica EM CPD300 (Leica, Germany), followed by mounting and sputter coating with platinum (15 nm) using a Cressington 108 sputter coater (Cressington, UK). The samples were imaged with a Quanta250 field-emission scanning electron microscope (FEI, The Netherlands). Hair bundle numbers were counted along successive 0.2 mm intervals at different cochlear turns by a second person blind to the experimental conditions. To quantify the loss of first- or third-row stereocilia, the stereocilia numbers in each row from individual OHCs were counted by a second person blind to the experimental conditions. The percentage of first- or third-row stereocilia loss was calculated by subtracting the number of first- or third-row stereocilia from the number of second-row stereocilia, then dividing the resulting value by the number of second-row stereocilia before multiplying by 100. Stereocilia length measurement was performed as previously described (Li et al., 2020). Briefly, SEM images were taken in two imaging planes of different angles. The relative stereocilia projection length and angles in the reconstructed three-dimensional coordinates were measured using Photoshop (Adobe). The stereocilia length was then calculated using equations described previously (Li et al., 2020).

Mice were located in a sound-exposure chamber with loudspeakers driven by a power amplifier (AWA5870B; Aihua Instruments Co., Hangzhou, China). Audio sound files were created and equalized using audio editing software (AWA6290M; Aihua Instruments Co.). Sound levels were calibrated using a sound level meter (AWA6221A; Aihua Instruments Co.) at multiple locations within the chamber. Mice of different genotypes were exposed to 2–20 kHz noise at 106 dB SPL for 2 h, and then put back in their home cage. ABR thresholds were measured before exposure and at different post-exposure time points. At certain time points, some of the mice were euthanized, and the stereocilia and ribbon synapses were examined by performing SEM and immunostaining.

The numbers of independent animals are indicated in the figure legends. Data are shown as mean±s.e.m. Appropriate tests used to determine statistical significance are indicated in the figure legends.

We thank Sen Wang, Xiaomin Zhao and Haiyan Yu from the core facilities for Life and Environmental Sciences, Shandong University, for technical support in SEM and confocal microscopy.