Live-cell single-molecule fluorescence microscopy for protruding organelles reveals regulatory mechanisms of MYO7A-driven cargo transport in stereocilia of inner ear hair cells

Stereocilia are unidirectional F-actin-based cylindrical protrusions on the apical surface of inner ear hair cells and function as biological mechanosensors of sound and acceleration. Development of functional stereocilia requires motor activities of unconventional myosins to transport proteins necessary for elongating the F-actin cores and to assemble the mechanoelectrical transduction (MET) channel complex. However, how each myosin localizes in stereocilia using the energy from ATP hydrolysis is only partially understood. In this study, we develop a methodology for live-cell single-molecule fluorescence microscopy of organelles protruding from the apical surface using a dual-view light-sheet microscope, diSPIM. We demonstrate that MYO7A, a component of the MET machinery, traffics as a dimer in stereocilia. Movements of MYO7A are restricted when scaffolded by the plasma membrane and F-actin as mediated by MYO7A’s interacting partners. Here, we discuss the technical details of our methodology and its future applications including analyses of cargo transportation in various organelles.


Introduction
Hearing loss affects approximately 20% of the population world-wide 1 and is classified into either conductive or sensorineural hearing loss or a mixture of these two conditions 2 .Sensorineural hearing loss is usually caused by lesions in the inner ear originating from factors including pathogenic variants of genes, noise exposure, ototoxic drugs and aging.The exact etiology for hearing loss often remains unidentified in actual clinical practice 2 .In the inner ear, stereocilia of cochlear hair cells convert mechanical sound vibrations into electrochemical activities of neurons.Signals from hair cells are transmitted to the central nervous system through the afferent fibers of spiral ganglion neurons 3 .A similar mechanism is utilized by vestibular hair cells to detect acceleration including gravity 4 .Stereocilia are cylindrical F-actin protrusions formed on the apical surface of each hair cell.Degeneration of stereocilia often accompanies sensorineural hearing loss including those caused by aging and genetic pathogenic variants 5 .Despite recent advances in gene therapy such as for OTOF [6][7][8] , there are presently no clinically useful treatments to restore degenerated stereocilia or to regenerate lost hair cells in the cochlea 9 .Understanding the molecular events in stereocilia of live hair cells can be a basis to elucidate the pathophysiology of sensorineural hearing loss and to eventually formulate therapeutic strategies to restore normal hearing.
Single-molecule fluorescence microscopy is a powerful technique to localize molecules coupled with a fluorophore, such as GFP and organic fluorescent dyes 10 .One major application is real-time functional analyses of proteins and other molecules in live cells 11 and on a coverslip between immobilized proteins and those in solution 12 .These techniques detect fluorescent molecules whose displacement, usually by diffusion, is smaller than the localization precision of an optical system 11 .Interactions between two fluorescently-labeled molecules can be detected more directly by Förster resonance energy transfer (FRET) 13 , optical tweezers 14 and fluorescence correlation spectroscopy 15 .In addition, super-resolution images can be produced by serially localizing sparse subsets of fluorophores conjugated to imaging targets, for example, by STORM 16,17 , PALM 18 and PAINT 19,20 methodologies.These methods have been extended to three-dimensional tissue specimens by introducing a thin illumination plane and, if possible, minimizing out-of-focus interference using confocal microscopy 11,21 .However, it has been challenging to apply single-molecule microscopy to organelles protruding from the apical plane, such as stereocilia, because most microscopes' focal plane coincides with the plane of the imaging substrate.Volume scanning along the vertical axis is useful for these samples but has disadvantages in resolution, real-time acquisition and phototoxicity 22,23 .One solution is to position a coverslip near the organelle of interest by placing samples upside-down on an inverted microscope 24 or using a thin chamber 25 although these techniques risk damaging the organelles and are not suitable for drug treatments.We solved this problem by employing a dual-view inverted selective plane illumination microscope (diSPIM) 26 and established a workflow of single-molecule microscopy with a more flexible spatial arrangement.We first imaged the dynamics of MYO7A, one of the unconventional myosins essential for normal hearing and vision 27,28 .
Development of functional stereocilia and their maintenance for life-long hearing require regulated motor activity of the wild-type unconventional myosins MYO3A, MYO6, MYO7A and MYO15A 29,30 .Pathogenic variants of these myosin genes are associated with human hereditary nonsyndromic hearing loss 27 .Variants of MYO7A are also associated with Usher syndrome type 1 characterized by congenital hearing loss, vestibular dysfunction and progressive retinal degeneration leading to blindness in the second to third decade of life 28 .During the development of hair cells, microvilli-like F-actin protrusions are formed on the apical surface of immature hair cells, of which a majority are "pruned" and only a small subset of microvilli grow thicker and taller and develop cross-linked paracrystal-like F-actin bundles in them.These maturing stereocilia are organized in rows of increasing height.All rows of stereocilia except the tallest are equipped with mechanoelectrical transduction (MET) channels at the distal ends.
Gating of MET channels is mediated by tip-links physically connected to a spot on the side of adjacent longer stereocilia 31 , which is referred to as the upper tip-link density (UTLD) because of the high electron scattering in transmission electron micrographs 3 (see Fig. 6a).Opening of MET channels allows K + and Ca 2+ from the endolymph to enter stereocilia and depolarizes the plasma membrane, which leads to release of glutamate from the basal surface of the hair cell 32- 34 .Unconventional myosins are involved in these processes by transporting and anchoring specific proteins and phospholipids as "cargo" using their unique tail domains as binding sites 35 .
A number of proteins have been identified as "cargo" of unconventional myosins in stereocilia.For example, MYO15A transport factors to elongate F-actin cores including WHRN (whirlin) and EPS8 (epidermal growth factor receptor kinase substrate 8) 36,37 .The motor domain of MYO15A itself is also reported to nucleate actin monomers 38 .MYO3A can interact with ESPN (espin) isoform 1 and ESPNL (espin-like), both of which are crucial for elongating F-actin protrusions 39,40 , although localization of these proteins in stereocilia may not completely depend on the motor activities of class-III myosins 41 .MYO7A localizes in the UTLD along with two scaffolding proteins, SANS and Harmonin, and tethers the tip-link consisting of CDH23 and PCDH15 to the F-actin core [42][43][44] .It is obvious that the motor activities of these myosins are necessary for localizing their cargo because mice with missense mutations disabling the motor domain of MYO7A and MYO15 have profound hearing loss 45,46 .However, less is understood about how these unconventional myosins traffic and transport cargo in stereocilia using energy from ATP hydrolysis.Mutant mouse models may not always be suitable to approach this question because stereocilia are often severely deformed in mice with impaired activity of these myosins, which likely disrupts cargo transport 45 .
In this study, we develop the methodology for live-cell single-molecule microscopy applicable to organelles protruding from the apical surface of tissue using a dual-view lightsheet microscope, diSPIM 26 , and bright fluorescent dyes (Fig. 1).Here, we experimentally address the unresolved question of how MYO7A molecules can transport components of the tiplink complex (Figs.2-6).MYO10, an unconventional myosin crucial for filopodia formation, was used as a positive control.We found that MYO7A shows directional, processive movements toward stereocilia tips when the motor domain is exposed by disabling its tail-mediated motor inhibition 47,48 .We further compared movements of MYO7A motor domains in stereocilia among (1) homodimers, (2) monomers anchored to the plasma membrane and (3) monomers tethered to F-actin at the C-terminus, all of which are possible through interacting partners of MYO7A.Among these conditions, only homodimers showed processive movements suggesting that MYO7A moves as a dimer in a stereocilium.The knowledge of how myosin-driven cargo transport occurs in stereocilia will be applicable to other F-actin protrusions, microvilli and filopodia, where diffusion of proteins are limited in a tightly-packed F-actin core and in an envelope of the plasma membrane 49 .Here, we introduce the technical details and methodology of our single-molecule microscopy of live hair cell stereocilia and its general application to threedimensional specimens including analyses of cargo transportation mechanisms in various organelles.

Single-molecule microscopy in stereocilia of live hair cells
The workflow for our single-molecule microscopy approach was developed using explant cultures of mouse utricles and saccules, hereafter referred to as "vestibular sensory epithelia", harvested from postnatal day (P) 2 to 5 (Fig. 1a).Stereocilia of utricles and saccules are suitable for single-molecule microscopy because they are straight and can be as long as 10 µm 50 .We employed a dual-view inverted selective plane illumination microscope (diSPIM) 26 in order to image stereocilia protruding upward from the apical surface of hair cells as we previously reported for stereocilia of fixed hair cells 12 .Using a Helios® gene-gun 51 , explant cultures of vestibular sensory epithelia were co-transfected with expression plasmids encoding a HaloTag-fused protein of interest and an EGFP (or EGFP-fused protein) to function as a transfection marker.Transfected vestibular sensory epithelia were maintained in DMEM/F12 culture media (37°C, 5% CO 2 ) and allowed to express these proteins for 16-24 hours.HaloTagfused protein was fluorescently labeled at a low density with JFX554-conjugated HaloTag ligands 52 and imaged with single-molecule microscopy using a diSPIM 26 illuminating a 561-nm laser.To image single protein molecules, we took advantage of our previously reported methodology of multiplexed super-resolution microscopy, in which we also detected single molecules of fluorescently labeled imaging probes with a diSPIM 12 .
The concentration of JFX554-conjugated HaloTag ligands was optimized using vestibular hair cells expressing HaloTag-fused human β-actin (HaloTag-actin) (Fig. 1b).With ligands applied at 0.3 nM or higher, the entire cell was labeled although more densely at stereocilia tips and in the cuticular plate (Fig. 1b, arrows), similarly to hair cells expressing EGFP-actin 24 .Fluorescent puncta of single HaloTag-actin molecules appeared in the cell body at concentrations of 0.1 nM (Fig. 1b, arrowheads) and in stereocilia at ligand concentrations of 0.03 nM or lower (Fig. 1b, circles, image of 0.01 nM shown).The optimal concentration was slightly different between cells depending on the amount of expressed HaloTag-actin.With dyes at 3 nM or above, unreacted fluorescent dyes were not completely washed away and remained in the tissue (image not shown).Thus, we considered that fluorescent ligands should be applied below 3 nM and be optimized depending on the expression level of HaloTag-fused protein.We also confirmed that single protein molecules are visualized by calculating the summed fluorescence intensity for each fluorescent punctum (Fig. 1c) using the cell in Fig. 1b, 0.01nM.
Fluorescent puncta were able to be classified into two populations in this cell (indicated by magenta and cyan circles in Fig. 1b, 0.01 nM).The average intensity of the Pop2 population was twice as large as that of the Pop1 population in Fig. 1c, indicating that fluorescent puncta in Pop1 originated from one fluorophore and that the average intensity of Pop1 corresponds to the quantal intensity.Concordantly, the average line scan of puncta in Pop1 was consistent with the point spread function of the objective lens (Fig. 1d).Time-lapse images of HaloTag-actin and non-fused HaloTag were acquired to evaluate how proteins are visualized when stably bound to the F-actin core and diffusing in stereocilia, respectively (Fig. 1, e and f).Single-plane time-lapse images were acquired every 1 second (s) to compare with subsequent MYO7A imaging (Figs.2-6).In the kymogram and movie, most HaloTag-actin molecules showed trajectories parallel to the time axis (X-axis) and disappeared suddenly due to photobleaching or transition to the dark state (Fig. 1e, arrows and Movie S1).
This behavior is consistent with single fluorophores staying in the same position.Consistent with diffusion, almost all non-fused HaloTag molecules showed trajectories no longer than one frame without staying in the same position (Fig. 1f, arrows and Movie S2) except for a few molecules stuck in stereocilia (Fig. 1f, open arrows).From these data, we concluded that fluorescent puncta in the time-lapse images reflect the dynamics of single protein molecules.

Visualization of directional movement of MYO7A dimers in stereocilia
After establishing the workflow for single-molecule microscopy, we developed imaging conditions suitable for detecting directional movements of HaloTag-fused MYO7A molecules in stereocilia (Fig. 2).We employed the heavy meromyosin-like fragment of mouse MYO7A (MYO7A-HMM) for this purpose (Fig. 2a).MYO7A-HMM is designed to be similar in domain composition to the heavy meromyosin (HMM) 53 , a protein fragment obtained by myosin II trypsinization and consisting of the motor and neck domains necessary for a power stroke on Factin 54,55 .MYO7A-HMM dimers show directional movements in filopodia and microvilli 53,56 and are expected to be a useful benchmark in stereocilia.In addition, MYO7A-HMM can be conditionally dimerized in live cells by fusing the p.F36V substitution mutation of FK506 binding protein 12 (FKBP) to the C-terminus and by adding a FK506-derived bivalent ligand, AP20187, to the culture medium 57 .This chemically inducible dimerization can "turn on" trafficking of MYO7A-HMM when the cells are ready for imaging.Thus, we constructed an expression vector for HaloTag-MYO7A-HMM-FKBP, which has a HaloTag for fluorescent labeling at the Nterminus and the FKBP for conditional dimerization at the C-terminus (Fig. 2b).HaloTag-MYO7A-HMM-FKBP expressed in vestibular hair cells formed large protein blobs at stereocilia tips only when AP20187 was added to the culture medium indicating that MYO7A-HMM dimers move toward the barbed ends of unidirectional F-actin bundles in stereocilia cores (Fig. 2c).
Single HaloTag-MYO7A-HMM-FKBP molecules were successfully detected using the imaging condition established with HaloTag-actin but at a slightly higher concentration of JFX554-ligands, 0.3-0.6 nM (Fig. 2d, Movie S3).MYO7A required a higher concentration of JFX554-ligand than β-actin likely reflecting the different expression levels of these two proteins.Time-lapse images after the AP20187 treatment visualized HaloTag-MYO7A-HMM-FKBP molecules moving directionally toward stereocilia tips (Fig. 2d, magenta circles).Kymograms showed continuous trajectories consistent with processive movements of MYO7A-HMM dimers (Fig. 2d, arrows).The velocity of movement was different between dimers (Fig. S1; representative kymograms) although there are no distinct populations of "slow" and "rapid" movements (Fig. 2e).The average velocity of movements was 101 ± 53 nm/s (n = 42; mean ± standard deviation), which is 10-fold faster than the movements of human recombinant MYO7A-HMM dimers on permeabilized filopodia (9.5 ± 0.4 nm/s) 58 .As we discuss later, this difference can be partially attributed to the temperature (37°C in our study vs. 25°C in the previous study) considering that similar difference was observed for single-molecule microscopy of MYO10 in live-cell filopodia (578 ± 174 nm/s at 25°C vs. 840 ± 210 nm/s at 37°C) 59 .HaloTag-MYO7A-HMM-FKBP did not show directional movements without AP20187 (Fig. 2f, Movie S4).

Constitutively active MYO7A mutants move directionally in stereocilia
Using the imaging condition established with MYO7A-HMM dimers, we tested the hypothesis that MYO7A traffics as dimers (or oligomers) in stereocilia (Fig. 3).However, HaloTag-fused fulllength MYO7A did not show directional movements at a detectable frequency in stereocilia of vestibular hair cells (image not shown).We considered the possibility that full-length MYO7A takes a backfolded autoinhibitory conformation between the tail and motor domains 47,48 .These studies also indicate that this autoinhibitory interaction is mediated by the "RGSK" motif in the second MyTH4-FERM domain (M/F2) and can be disabled by substituting arginine and lysine in this motif with alanine residues.Thus, we evaluated movements of two HaloTag-fused MYO7A mutants, HaloTag-MYO7A-RK/AA and HaloTag-MYO7A-ΔSH3-ΔM/F2, which disable the tailmediated autoinhibition by missense mutations of RK to AA residues and by a deletion of SH3 and M/F2 domains, respectively (Fig. 3a).While wild-type MYO7A diffusely distributed in stereocilia (Fig. 3b, arrowhead in Wild-type), these MYO7A mutants accumulated at stereocilia tips in a few transfected hair cells (Fig. 3b, arrows in RK/AA and ΔSH3-ΔM/F2).We speculate that accumulation at stereocilia tips may occur only when sufficient amounts of HaloTag-fused MYO7A and its interaction partners are present in the cell.
Under single-molecule imaging conditions, a few HaloTag-MYO7A-RK/AA molecules moved toward stereocilia tips (Fig. 3c, Movie S5).Movements of MYO7A-RK/AA were directional and processive resembling the movements of MYO7A-HMM dimers.This result suggests that MYO7A can dimerize spontaneously on the F-actin cores of stereocilia when the motor domain is exposed, for example, by cargo bound to the tail 47,60,61 .Similar processive movements were observed in cells expressing HaloTag-MYO7A-ΔSH3-ΔM/F2 (Fig. 3d, Movie S6) indicating that MYO7A can dimerize using motifs in the neck or in the first MyTH4-FERM domain (M/F1).

Distinct behavior of MYO7A and MYO10 anchored to the plasma membrane
Our data indicate that MYO7A can traffic in stereocilia as a dimer or perhaps an oligomer.
Another possibility is that MYO7A is anchored to the plasma membrane of stereocilia and showed directional movements on the adjacent F-actin core (Fig. 4).Previous studies show that MYO7A can interact with CDH23 via two scaffolding proteins in the tip-link complex, SANS and Harmonin [62][63][64] , and directly with PCDH15 65 (see Fig. 6b).In addition, anchoring to the plasma membrane can induce directional movements of MYO10 in filopodia 66 .Thus, we tested if MYO7A-HMM can show directional movements when tethered to the plasma membrane.
Anchoring of MYO7A-HMM to the plasma membrane was achieved using a small transmembrane motif from the human Interleukin 2 receptor alpha chain (IL2Rα) 67 (Fig. 4a).
IL2Rα and MYO7A-HMM were conditionally heterodimerized by fusing FKBP and FKBP-Rapamycin binding protein (FRB) to the C-terminus of each protein and supplementing the culture medium with a Rapamycin analog, AP21987 68 .Bovine MYO10 lacking the entire tail and the coiled-coil domain for anti-parallel dimerization (MYO10-MD) in the neck 66 was used as a positive control.Fluorescence confocal microscopy shows that HaloTag-MYO7A-HMM-FRB is successfully anchored to the plasma membrane after AP21987 treatment (Fig. 4b).The positive control, HaloTag-MYO10-MD-FRB, accumulated weakly at stereocilia tips in a few cells without AP21987 in the medium (Fig. 4c, arrowhead) and more densely at stereocilia tips after the addition of AP21987 (Fig. 4c, arrows) as previously reported for filopodia 66 .Excess IL2Rα-EGFP-FKBP sometimes accumulated in vesicles but did not cause apparent damage to stereocilia (Fig. 4b, open arrowheads).
Single-molecule microscopy showed that MYO7A-HMM can move in a stereocilium using the plasma membrane as a scaffold although the movements were restricted and different from those of dimers (Fig. 2d and Fig. 4d).In contrast, MYO10-MD moved efficiently toward stereocilia tips when anchored to the plasma membrane (Fig. 4e).In cells expressing HaloTag-MYO7A-HMM-FRB and IL2Rα-EGFP-FKBP, only a small number of MYO7A-HMM molecules moved directionally toward stereocilia tips after AP21987 treatment (Fig. 4d, Movie S7).
Movements of membrane-anchored MYO7A-HMM were intermittent in a kymogram as if MYO7A-HMM moves toward stereocilia tips only when they dissociate from F-actin (Fig. 4d, arrows and open arrows).As indicated by the weak accumulation in fluorescent histochemistry (Fig. 4c, arrowhead), a small number of MYO10-MD molecules moved directionally even in untreated cells (Fig. 4e, arrows and open arrows in the upper kymogram, Movie S8).After AP21987 treatment, MYO10-MD anchored to the plasma membrane showed processive movements toward stereocilia tips (Fig. 4e, arrows and open arrows in the lower kymogram, Movie S9; representative kymograms also in Fig. S2b).The average velocity of MYO10-MD was 2.01 ± 0.37 μm/s (n = 12) before treatment and 0.72 ± 0.34 µm/s (n = 23) after treatment (Fig. S2a) indicating that movements of MYO10-MD were affected by membrane anchoring.
Movements of membrane-anchored MYO7A-HMM were different from MYO7A-HMM dimers or constitutively active MYO7A mutants.Directional movements of MYO10-MD indicate that the restricted movements of MYO7A-HMM are derived from the kinetic differences between the motor domains of MYO7A and MYO10.

Step-wise movements of MYO7A and MYO10 when tethered to F-actin
We tested an additional scenario where F-actin functions as a scaffold for MYO7A to move in a stereocilium (Fig. 5).In the UTLD of stereocilia, a class of Harmonin isoforms that contain the Proline, Serine and Threonine-rich (PST) domain (collectively referred to as Harmonin b) connect the tip-link complex to the F-actin core including MYO7A 62 .Thus, we conditionally tethered HaloTag-MYO7A-HMM-FKBP or HaloTag-MYO10-MD-FKBP to F-actin under AP21987 treatment using the PST domain of Harmonin b (residues 296-728 of NM_01163733) fused with FRB and EGFP at the N-and C-termini (FRB-PST-EGFP) (Fig. 5a).Before AP21987 treatment, both HaloTag-MYO7A-HMM-FKBP and HaloTag-MYO10-MD-FKBP diffusely distributed in stereocilia (Fig. 5, b and c, arrowheads).FRB-PST-EGFP also distributed diffusely in stereocilia and weakly accumulated at stereocilia tips (Fig. 5, b and c, open arrowheads).
Single-molecule microscopy demonstrated that MYO7A-HMM can move in a stereocilium using F-actin as a scaffold, but the trajectories in kymograms are different from those of dimers (Fig. 2d and Fig. 5d).After AP21987 treatment, only a small number of MYO7A-HMM molecules moved directionally toward stereocilia tips (Fig. 5d, Movies S10).Movements of MYO7A-HMM molecules tethered to F-actin were step-wise as observed for those anchored to the plasma membrane.MYO10-MD molecules also showed step-wise movements after the AP21987 treatment (Fig. 5e, Movie S11).These results suggest that myosin molecules tethered to F-actin move only when the tail is released from F-actin.These observations are not consistent with an "inchworm-like" movement proposed by others 40,69 because the step-sizes were 100-200 nm, which is much larger than the size of heavy meromyosin (~15 nm) 70 .The restricted movements of myosins tethered to F-actin may be advantageous to anchor the components of the tip-link complex, including MYO7A, in the UTLD after being transported on the F-actin core by MYO7A dimers.

Discussion
The methodology proposed here expands the targets of single-molecule microscopy in live cells to three-dimensional organelles beyond stereocilia to include cilia, microvilli, filopodia and their associated diseases [71][72][73][74][75] .In this study, we successfully visualized and localized single HaloTagfused protein molecules in live hair cells and used our method to elucidate how a single unconventional myosin can traffic in a stereocilium using their motor activities.Stereocilia are intricate mechanosensors consisting of more than 500 different proteins including actin monomers, F-actin bundling proteins, a multitude of MET channel and tip-link components and unconventional myosins 76 .Among these molecules, unconventional myosins are crucial molecular players during the development and maintenance of functional stereocilia because they transport and anchor specific stereocilia components using their highly diverse tail domains.Some unconventional myosins also interact with phosphatidylinositol 4,5-bisphosphate (PIP2) in the plasma membrane and are involved in adaptation of the MET channels (MYO1C) 77 and maintenance of stereocilia architecture (MYO6) 78 .While decades of studies have identified various "cargo" of unconventional myosins in stereocilia, less is understood about how each myosin traffics in a stereocilium including whether or not each myosin is dimerized (or oligomerized).Elucidating the mechanisms underlying trafficking of unconventional myosins will be a basis for approaching questions more closely related to clinical practice, such as (1) why variants of some myosins (MYO6 and MYO7A) are associated with both autosomal dominant or recessive nonsyndromic hearing loss [79][80][81][82] including the possibility of dominant-negative effects and (2) how myosin function could be restored therapeutically, especially for variants causing autosomal dominant progressive loss of hearing.Our single-molecule microscopy is a methodology of choice to explore these open questions through in vivo real-time functional analyses.
We speculate that the phenotype of constitutively active MYO7A mutants can be a good starting point to elucidate the formation of tip-links and the UTLD in developing stereocilia.
MYO7A localizes in the UTLD region and tethers tip-links consisting of PCDH15 dimers and 12 CDH23 dimers to the F-actin core on the CDH23 side 43,83,84 (Fig. 6a).Two scaffolding proteins, Harmonin and SANS, bridge the interaction between MYO7A and CDH23 [62][63][64] and help to form a network of interactions (Fig. 6b).Harmonin is expressed in several isoforms in hair cells, which are collectively referred to as Harmonin b harboring the PST domain to interact with F-actin and as Harmonin a and c lacking the PST domain 42,85 .Processive movements of MYO7A-RK/AA and MYO7A-ΔSH3-ΔM/F2 (Fig. 3) suggest that MYO7A can dimerize (or perhaps oligomerize) when the tail-mediated autoinhibition 47,48 is disabled.Cargo binding is one major mechanism to unleash MYO7A from the autoinhibitory state 53 although further analyses are required to elucidate how MYO7A can be dimerized in stereocilia.SANS may be involved in this activation through the interaction with the first MyTH4-FERM domain (M/F1) because MYO7A-HMM-FKBP did not show movements without AP20187 at a detectable frequency (Fig. 2f).A previous study indicates that Harmonin is transported by MYO7A and SANS since localization of Harmonin b at stereocilia tips is lost in mice with defective MYO7A (Myo7a 4626SB/4626SB ) or SANS (Ush1g js/js ) 45 .
In addition, MYO7A-HMM did not move efficiently when anchored to the plasma membrane or tethered to F-actin (Figs. 4 and 5), which is consistent with previous studies showing that intact tip-links are not required to localize MYO7A and Harmonin b at stereocilia tips 45,86 and that the PST domain of Harmonin b is not necessary to form tip-links 42 .However, several studies suggest that there are alternative pathways to localize components of tip-links and the UTLD.
Localization of CDH23 and PCDH15 in stereocilia tips, as well as that of MYO7A, are not lost in mice lacking Harmonin isoforms (Ush1c −/− ) 45 .SANS can still localize in stereocilia of Myo7a 4626SB/4626SB and in Ush1c −/− mice although its localization is not limited to the UTLD 87 and different from that in wild-type mice.
Our data support several scenarios that might localize tip-link components: (1) scaffolds of MYO7A, SANS and Harmonin b settle on the F-actin core and then recruit tip-link components (Fig. 6c, "scaffold first" scenario), (2) tip-links are formed at the future UTLD region and then are anchored to the F-actin core by Harmonin b (Fig. 6d, "tip-link first" scenario), and (3) fully formed tip-links are transported toward stereocilia tips and are then anchored to the Factin core by Harmonin b (Fig. 6e, "walking links" scenario).In these scenarios, SANS, PCDH15 and CDH23 may dimerize MYO7A since these proteins can homodimerize with each other 64,83,84 .
Harmonin b might also be another mediator in dimerization or oligomerization through its oligomerization activity 63 .Among these three scenarios, the "scaffold first" scenario is less likely because it cannot explain how MYO7A (and SANS) can settle down at the future UTLD region without a landmark.The "tip-link first" scenario can be supported by remodeling experiments of tip-links because tip-links are newly formed after being disassembled by BAPTA-mediated extracellular calcium chelation 88 .This study also reports that temporary links consisting of only PCDH15 are formed after BAPTA treatment (Fig. 6d, double asterisks) subsequently replacing PCDH15 on the taller stereocilia side with CDH23 to form mature tip-links 88 .In contrast, the "walking links" scenario is supported by a previous cryoelectron microscopy study using anti-PCDH15 antibodies, which detected multiple "lateral links" at the shaft region of stereocilia 89 .
These links consist of PCDH15 (~50 nm long) on one side and a longer partner on the other side (~120 nm long), putatively CDH23 89 .However, the "tip-link first" and "walking links" scenarios may coincide because many uncoupled PCDH15 molecules are reported to localize near stereocilia tips in the cryoelectron microscopy study 89 .These uncoupled PCDH15 molecules may be actively transported, for example, by the interaction with the SH3 domain of MYO7A 65 (Fig. 6d, asterisk) and also provokes the idea that active transport of fully-formed tiplinks may occur on the PCDH15 side (Fig. 6e, triple asterisks).In addition, our study shows that movements of MYO7A is restricted when its tail is anchored to the F-actin core by the PST domain of Harmonin b (Fig. 5).Harmonin a and c lacking the PST domain should be advantageous for effective active transport on the CDH23 side (Fig. 6, d and e).Further analyses, especially on trafficking of PCDH15 and CDH23, are necessary to conclude which circumstance most likely explains how tip links arise in vivo.
In addition to MYO7A, our single molecule imaging studies can be applied to other unconventional myosins functioning in F-actin protrusions to answer unresolved questions.For example, it is still uncertain how class-III myosins move in F-actin protrusions including stereocilia.Currently, there is no evidence that MYO3A has a dimerization sequence, such as a coiled-coil domain 90 .Instead, MYO3A can move directionally in F-actin protrusions using the THDI and THDII domains in its tail, which interact with F-actin binding proteins, ESPN isoform 1 or ESPNL, and F-actin, respectively 39,40,91 .Our method may elucidate how MYO3A can move toward the barbed ends when their tail is tethered to F-actin and how directional movements of MYO3A is regulated by the autoinhibitory kinase domain 92 .Stepwise movements of MYO7A-HMM and MYO10-MD tethered to F-actin suggest that MYO3A may traffic in stereocilia similarly, not by "inchworm-like" movements as previously proposed 40,69 .In addition, MYO10-MD can move in stereocilia using various scaffolds and even as a monomer (Figs.4e and 5e).
This observation gives us a clue as to why MYO7A and MYO7B are utilized in stereocilia and microvilli, both of which are more stable and long-lived than filopodia 93 .It is known that recruitment of MYO10 to the plasma membrane can induce formation of filopodia-like F-actin protrusions 66 .Movements of MYO7A are restricted when anchored to plasma membranes, which might prevent MYO7A from forming unwanted F-actin protrusions that would destroy the stereocilia architecture.MYO10-MD tethered to the plasma membrane or F-actin accumulates at stereocilia tips and can alter the architecture of stereocilia (Figs.4c and 5c).
Our methodology enables functional analyses of protein molecules in the context of live cells.Protein kinetics in stereocilia are complex due to the multitude of components and to the limited diffusion from the tightly-packed F-actin and the enveloping plasma membrane 49 .
Moreover, the behavior of proteins in vivo can be completely different from that in vitro as we demonstrated for CapZ, an F-actin barbed end capper, in lamellipodia 94 .In this study, we noticed that MYO7A-HMM dimers move 10-fold faster than in a previous study 58 .Although this difference can be partially attributed to the difference in temperature (37°C in this study vs. 25°C in the previous study 58 ), a previous MYO10 dimer study shows only less than 2-fold increase in velocity between 25°C and 37°C 59 .In filopodia, it was shown that unidirectionally bundled Factin is advantageous for MYO7A-HMM dimers to move rapidly without changing the "actin tracks", but the increase in velocity is less than 2-fold 58 .One possibility is that some dimers "jump" toward the barbed ends between two processive movements or binding events using restricted diffusion in stereocilia 49 .Further analyses are required to test this hypothesis.In addition, analyses of molecular turnover in a semi-equilibrium state could also benefit from our technique.Actin and regulatory proteins are replenished in the F-actin core while the shape, width and length of stereocilia remain largely unchanged on the apical surface of hair cells 12,24,95 .Our methodology may answer another crucial question as to how the architecture of stereocilia remains stable for the lifespan of a normal hearing person 96 .
In this study, we used a symmetrical diSPIM microscope equipped with two 40×, 0.8 NA water-immersion objective lenses 26 used for single dye detection in our previous study 12 .Our method could be improved by imaging with a higher numerical aperture (NA) objective to increase localization precision and light collection.Improved NA could enable richer analyses of myosin trafficking in stereocilia including investigation of their step-sizes 97 and processivity and also may allow single-molecule microscopy in stereocilia of cochlear hair cells.
Step-size analysis could be facilitated in future studies by using an asymmetric diSPIM design, featuring higher NA lenses and a suitably long working distance 98 .Another future improvement would be to avoid the necessity for an activation mechanism.This study employed chemically-induced dimerization techniques to "turn on" myosin trafficking after cells were ready for imaging.
Methods for direct delivery of fluorescently-labeled protein molecules, such as microinjection, will be useful for this purpose and also for visualizing replenishment and dissociation of protein molecules in stereocilia avoiding the complexities associated with blinking fluorescent dyes 99,100 .
Single-molecule microscopy enables real-time observation of molecules at work, such as unconventional myosins trafficking in stereocilia, as we demonstrate in this study.Recent advances in light-sheet microscopy allowed us to apply this powerful technique to stereocilia, which had been challenging to image due to their architecture protruding upward from the apical cell surface.Although mechanosensory stereocilia are unique to the inner ear, our data should be applicable to analyze active cargo transport in other F-actin protrusions, microvilli and filopodia, where diffusion of proteins are restricted 49 .In addition, the flexible spatial arrangement in our methodology will be useful for other three-dimensional organs and organelles including primary cilia, kinocilia, migrating cells and perhaps neuronal cell layers.We speculate that our methodology will contribute to expanding the target of single-molecule fluorescence microscopy and help formulating therapeutic strategies not only for sensorineural hearing loss but also for other diseases, such as ciliopathies 71,72 , neurodegenerative diseases 101 , inflammatory bowel diseases 73,74 and cancer 75 .

Plasmids and cDNAs
To express HaloTag-fused proteins, we modified the pEGFP-C1 (Clonetech) vector by replacing the EGFP sequence with a HaloTag sequence that was PCR amplified from the pHTC HaloTag® CMV-neo (Promega).Two silent mutations were introduced to the wild-type HaloTag sequence to disable the XhoI and SalI endonuclease restriction sites.The resulting vector pHaloTagXS-C1 was used to express HaloTag-fused proteins.A plasmid encoding HaloTagfused human β-actin was constructed using this vector and the actin sequence deriving from pEGFP-actin (Clonetech).Plasmids encoding HaloTag-fused MYO7A fragments were constructed using the pHaloTagXS-C1 vector and inserts amplified from a plasmid encoding EGFP-fused mouse Myo7a isoform 1 (NM_001256083.1) 36, a gift from Erich Boger, NIDCD.
Plasmids encoding HaloTag-fused MYO10 fragments were constructed similarly using Thus, the C-terminal FKBP and FRB attached to HaloTag-fused myosin fragments have an additional HA tag in this study.A plasmid to express IL2Rα-EGFP-FKBP was constructed from pEGFP-C1 inserting a DNA fragment encoding the Tac antigen (human IL2 receptor alpha subunit; NM_000417.3)amplified from TAC-GFP 103 (Addgene # 162494) and the FKBP fragment.A mouse USH1C fragment (residues 296-728 of NM_01163733) was expressed with FRB at the N-terminus and EGFP at the C-terminus using a plasmid constructed from pEGFP-N3 (Clonetech) inserting DNA fragment amplified from pEGFP-FRB 102 and a USH1C fragment amplified from the cDNA of USH1C isoform b4, a gift from Nicolas Grillet at Stanford University.

Animals
All animal experiments were performed in accordance with the National Institutes of Health Guidelines for the Care and Use of Laboratory Animals and approved by the Animal Care and Use Committees at the NIH (No. 1263 to TBF).Mouse neonates were obtained from timed pregnant C57BL6/J females purchased from the Jackson Laboratory or from our in-house C57BL6/J colony.Vestibular sensory epithelia were harvested from neonates at postnatal day (P) 2-5 after being euthanized by decapitation.

Explant culture and transfection of vestibular sensory epithelia
Inner ear sensory epithelia were cultured and transfected using a Helios® Gene Gun System (Bio-Rad) as previously described with slight modification 51 .Briefly, utricles and saccules of P2-5 mice were isolated in Leibovitz's L-15 medium (Thermo Fisher) after removing the otoliths using a 30G needle.Isolated sensory epithelia were placed on a glass-bottom dish (MatTek Corporation) coated with rat-tail collagen I (A1048301, Thermo Fisher) matrix and cultured in Dulbecco's Modified Eagle Medium/Nutrient Mixture F-12 (DMEM/F12, Thermo Fisher) supplemented with 7% FCS (Atlanta Biologicals) and 20 µg/mL ampicillin (Sigma) at 37°C in 5% CO 2 .After culturing for 6-20 h, vestibular hair cells were transfected using a gene-gun and 1.0 µm Gold Microcarriers (1652263, Bio-Rad) propelled by helium pulses at 110-115 psi.Gold microcarriers were coated with a 1:3 or 1:1 mixture of two plasmids encoding a HaloTag-fused protein and EGFP (or an EGFP-fused protein), respectively.

Single molecule microscopy of live hair cell stereocilia
Vestibular sensory epithelia co-expressing a HaloTag-fused protein and EGFP (or EGFP-fused protein) were labeled using 0.01-1 nM JFX554-conjugated HaloTag ligands 52 (gift from Luke Lavis at Janelia Research Campus) diluted in culture medium (DMEM/F12 supplemented with 7% FCS and 20 µg/mL Ampicillin) for 30 min at 37°C in 5% CO 2 .Unreacted HaloTag ligands were removed by washing the tissue in the culture medium for 3-5 s three times and by incubating samples at 37°C in 5% CO 2 up to 6 h until live imaging was performed.Samples on a collagen matrix were detached from glass-bottom dishes with the underneath coverslip, mounted in a 10-cm plastic culture dish on a 1-2 mm droplet of vacuum grease (Fisher Scientific) and then incubated in Leibovitz's L-15 Medium without phenol red (21083027, Thermo Fisher) warmed at 37°C.For conditional dimerization, the medium was gently removed using plastic transfer pipettes and replaced with a new L-15 medium containing 200 nM AP20187 (Sigma) or 500 nM AP21987 (Takara Bio).
Images were acquired using a custom-made symmetrical dual-view inverted selective plane illumination microscope (diSPIM) 26  The entire architecture of stereocilia was visualized using fluorescence from co-expressed EGFP or EGFP-fused proteins by a volume scan of 0.5-µm thickness illuminating the 488-nm laser at approximately 0.01 kW/cm 2 for 100 ms per slice.HaloTag-fused protein molecules labeled with JFX554 were visualized illuminating the 561-nm laser at approximately 0.2 kW/cm 2 for 100 ms per plane.HaloTag-actin in Fig. 1, a and b, was imaged by a single volume scan of 0.5-µm thickness.Other images were acquired every 0.1 to 1 s by single-plane time-lapse imaging.Sample drift was corrected using our custom-made Python scripts available at GitHub (http://github.com/takushim/momomagick)implementing phase-only correlation with subpixel matching 107 and least-square image matching 108 .Fluorescent puncta were manually tracked using our custom-made Python script with a graphical user interface (http://github.com/takushim/momotrack).The point spread function of the 40× objective lens was obtained from a previous study 109 .Kymograms were generated using the Fiji platform 110 .

Fluorescence histochemistry
Explant cultures of vestibular sensory epithelia were fixed in 4% paraformaldehyde (PFA; Electron Microscopy Sciences) in PBS for 30 min at room temperature (RT) and washed in PBS.For conditional dimerization, samples were treated with 100 nM AP20187 (Sigma) or 500 nM AP21987 (Takara Bio) for 2 h.The concentration of AP20187 was lowered because 200 nM AP21087 caused strong accumulation of MYO7A-HMM dimers at stereocilia tips and often damaged the architecture of stereocilia.Samples were permeabilized and blocked in PBS containing 1% bovine serum albumin (BSA) and 0.2% Triton-X100 for 20-30 min at RT.
HaloTag-fused proteins were visualized by reacting with 200 nM JFX554-HaloTag-ligand 52 (gift from Luke Lavis) in PBS with 0.2% Triton-X 100 for 1 h at RT. F-actin was visualized by supplementing the HaloTag ligand solution with 10-30 nM of Alexa Fluor™ Plus 405 Phalloidin (Thermo Fisher).Confocal images were acquired using a Zeiss LSM880 with Airyscan processing (Zeiss).

Acknowledgements
We thank Drs. Dennis Winkler and Mhamed Grati for valuable comments; Elizabeth Bernhard, Sherly Michel and Alexander Callahan for managing mouse colonies; Jiji Chen and Min Guo for comments on single-molecule microscopy and image processing; and Erina He for her beautiful diagrams.Images using the diSPIM were acquired in the Advanced Imaging and Microscopy Resource of the NIBIB.This article is subject to Howard Hughes Medical Institute (HHMI)'s Open Access to Publications policy.HHMI laboratory heads have previously granted a nonexclusive CC BY 4.0 license to the public and a sub-licensable license to HHMI in their research articles.Pursuant to those licenses, the author-accepted manuscript of this article can be made freely available under a CC BY 4.0 license immediately upon publication.

Declaration of interests
The authors declare no conflict of interest associated with this manuscript.The SAH domain of MYO7A has weak dimerization activity 53,113 .SANS, PCDH15 and CDH23 can dimerize with each other 64,83,84 and may keep multiple MYO7A molecules in proximity.
Oligomerization has been reported for Harmonin 63 .PCDH15 and CDH23 each have one transmembrane motif (TM) 83,84 .The SH3 domain of MYO7A can interact with the intracellular portion (IC) of PCDH15 (CD2 isoform shown) 65 5).PCDH15 may be transported in both shorter and longer stereocilia and form temporary links as indicated by BAPTA-mediated remodeling experiments 88 .The "walking links" scenario (e) supported by a previous cryo-electron microscopy study assumes that tip-links are formed on the shaft of stereocilia and then transported toward stereocilia tips 89 .The "walking links" and "tip-link first" scenarios may coincide because many uncoupled PCDH15 molecules were reported at stereocilia tips in the same study 89 .
, YI, SMA and TBF were supported (in part) by NIDCD intramural research funds DC000039 (to TBF).HDV and HS were supported by NIBIB intramural research funds (to HS).HS was also supported by HHMI.TM was supported by JSPS Overseas Research Fellowships.TM and MS were supported by start-up funds from Southern Illinois University School of Medicine (to TM) and the R00 Pathway-to-Independence Award (1R00DC019949 to TM).

Fig. 1 :
Fig. 1: Development of single-molecule microscopy in live hair cells a, Illustration showing our workflow for single-molecule microscopy.b, Optimization of labeling density using vestibular hair cells expressing HaloTag-actin.JFX554-conjugated HaloTagligands applied at different concentrations.At 0.3 nM or above, JFX554-ligands distribute throughout the cell and accumulated at stereocilia tips (arrows) and in the cuticular plate.Fluorescent puncta begin to appear at 0.1 nM in the cell body (arrowheads) and becomes distinguishable in stereocilia around 0.01 nM (indicated by colored circles for quantification in c).Maximum projections of volume scans are shown.Exposure, 100 ms per plane at 0.2 kW/cm 2 .Bars, 5 µm.c, Intensities of fluorescent puncta in b, 0.01 nM JFX554-ligand.The sum intensity of each punctum is calculated by adding all pixel values encompassing the punctum and then subtracting the background intensity.There are at least two populations of fluorescent puncta, Pop1 and Pop2, indicated by magenta and cyan circles in b, respectively.The average intensity of Pop2 is 76 ± 10 (n = 12; mean ± standard deviation) and approximately twice as large as that of Pop1 (34 ± 10; n = 16) indicating puncta in Pop1 and Pop2 are emitted from one and two fluorophores, respectively.Puncta of high intensity in Pop2 (asterisks) may be emitted from more than two fluorophores.Pixel values are calculated using an average projection of the volume scan.d, Comparison between the average line-scan of fluorescent puncta (magenta circles in b, 0.01 nM) and the point spread function of the objective lens calculated using PSF Generator (https://bigwww.epfl.ch/algorithms/psfgenerator/).Similarity between both intensity curves suggests that these puncta are emitted from a point source.e and f, Representative kymograms of controls, HaloTag-actin (e) and non-fused HaloTag (f) labeled with 0.01 nM and 0.1 nM JFX554-ligands, respectively.Single-plane images are acquired every 1 s for comparison with MYO7A movements.Kymograms are generated from the line scans between arrowheads.Most HaloTag-actin molecules stay in the same place and disappear suddenly due to photobleaching or transition to the dark state (e, arrows), which also suggests that these puncta are emitted from single fluorophores.Most non-fused HaloTag molecules disappear after one frame (f, arrows) due to diffusion except for a few molecules likely stuck in stereocilia (f, open arrows).Imaging conditions are similar to b. Bars, 20 s and 2 µm.

Fig. 6 :
Fig. 6: Possible scenarios for MYO7A-driven cargo transport in stereocilia a, Architecture of stereocilia and the MET and tip-link complexes.The MET channels are localized at distal ends of stereocilia and physically gated by extracellular tip-links connected to the sides of adjacent stereocilia in the longer row at the region referred to as the upper tip-link . c-e, Possible scenarios of MYO7A-driven localization and formation of tip-links and the UTLD.The "scaffold first" scenario (c) assumes that MYO7A (and SANS) traffics toward stereocilia tips, settles at the future UTLD region via the PST domain of Harmonin b and then recruits tip-link components.However, this scenario cannot explain how MYO7A (and SANS) can settle down at the future UTLD region without landmarks.The "tip-link first" scenario (d) assumes that tip-link components are transported toward stereocilia tips and then form tip-links at the UTLD region.Different isoforms of Harmonin might be used during the transport (Harmonin a or c lacking the PST domain) and after the transport (Harmonin b with the PST domain) because MYO7A anchored to the F-actin by the PST domain of Harmonin b cannot traffic efficiently (see Fig.
was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.It is made