We are interested in the gating of mechanically sensitive ion channels, which open in response to force on the channel proteins. We study these channels primarily in vertebrate hair cells—the receptor cells of the inner ear, which are sensitive to sounds or accelerations. Hair cells are epithelial cells and they have a bundle of stereocilia rising from their apical surfaces. Mechanical deflection of a bundle changes the tension in fine "tip links" that stretch between the stereocilia; these filaments are thought to pull directly on the force-gated transduction channels to regulate their opening. A variety of projects in the laboratory explore hair-cell structure and function, as well as some related topics.

Hair bundles of the bullfrog saccule.

Click on the following to read more:

Hair Bundle Mechanics

When the tip of a hair bundle is deflected by a sensory stimulus, the stereocilia pivot as a unit, producing a shearing displacement between adjacent tips. It is not clear how stereocilia can stick together laterally but still shear. We use hair cells from the bullfrog saccule and high-speed video imaging to characterize this sliding adhesion. Movement of individual stereocilia is proportional to height, indicating that stereocilia pivot at their basal insertion points. All stereocilia move by approximately the same angular deflection, and the same motion is observed at 1, 20 and 700-Hz stimulus frequency. Motion is consistent with a geometric model that assumes stereocilia shear without measurable separation; we estimated that stereocilia separation does not increase by more than 5 nm during physiological stimuli (<300nm). The same motion is observed when bundles are moved perpendicular to the tip links, when tip links are cut, or when ankle links are cut, ruling out tip links and ankle links as the basis for sliding adhesion. These observations suggest that the horizontal top connectors, the only links not disrupted, mediate a sliding adhesion. They also indicate that all transduction channels of a hair cell are mechanically in parallel, which simplifies biophysical models of transduction and which may enhance amplification in the inner ear.

Deflection of a bullfrog hair bundle with a mechanical probe. Stereocilia do not separate by more than a few nanometers, even for large stimuli like these.
→ Watch "movieF1.avi"

MovieF1: Low-frequency (1 Hz) motion of a hair bundle. The stimulus probe is attached by suction to the kinociliary bulb. Stimulus amplitude ~400 nm

→ Watch "movieF700.avi"

MovieF700: High-frequency (700 Hz) motion of a hair bundle (same cell and stimulus as in MovieF1)

→ View Geometric Model

Back To Top


Cutting the tip links by treatment of a hair bundle with BAPTA-buffered low Ca2+ solutions immediately abolishes mechanical sensitivity and also abolishes the mechanical correlates of channel opening, indicating that tip links are essential to convey mechanical force to the transduction channel. Yet the connection of the tip link to the channels is still not clear. Earlier Ca2+ imaging from this laboratory reported stimulus-dependent Ca2+ entry in most stereocilia of the bullfrog saccule, including both tallest and shortest of a column, suggesting that transduction channels are at both ends of tip links. More recently, Ricci and Fettiplace saw no Ca2+ entry in the tallest stereocilia of mammalian cochlear hair cells, suggesting instead that transduction channels are only at the lower end of each tip link. We have returned to this problem, using super-resolution optical methods, to resolve the differences between bullfrog and rat hair cells. One possibility is that—in the bullfrog—the tallest stereocilium in a lateral column may send an off-axis tip link to a yet-taller stereocilium of a more-medial column, so that stereocilia morphologically interpreted as being the tallest were functionally not the tallest. We are using STORM and STED microscopy to test this possibility.

Back To Top


Transduction channels open with a mechanical stimulus, but then adapt over a time course of milliseconds. One phase of adaptation was shown to be mediated by a motor complex of myosin-1c molecules relaxing tension on the channels, but a faster phase apparently results from Ca2+ that enters through the channels immediately binding to close them. Where Ca2+ binds, and to what, are less clear. We are characterizing the site of Ca2+ action by photolytically releasing Ca2+ inside stereocilia and measuring the nanometer movements that correspond to channel closing. By understanding the mechanical correlates of fast adaptation we will gain a better sense of how Ca2+ closes channels.

Receptor current (top) and hair bundle motion (bottom) in response to force steps. Fast adaptation in the first few milliseconds is correlated with negative bundle motion.

Back To Top

Cadherin Structure and Mechanics

Tip links are made of two unusual cadherins with long extracellular domains—cadherin 23 and protocadherin 15—whose N-termini join to complete the link. We are interested in the tip link's biophysical properties and how the two cadherins join. However, their molecular structure, elasticity, and deafness-related structural defects are unknown. Together with Wilhelm Weihofen and Rachelle Gaudet (Harvard University) we have solved the x-ray crystallographic structures of the first and second extracellular repeats (EC1+2) of cadherin 23. They show a typical cadherin fold, but reveal an elongated N-terminus that precludes classical cadherin binding interactions and contributes to a novel Ca2+-binding site.

Structure of the first two EC repeats of cadherin 23. Calcium ions hold together adjacent EC repeats, and may similarly hold cadherin 23 to protocadherin 15.

By putting the atomic structure in a computer along with associated water molecules, and by iteratively calculating its movement in response to applied forces, we can understand mechanical properties such as its elasticity and unfolding strength. These steered molecular dynamics simulations suggest that cadherin 23 repeats are stiff, and that either removing Ca2+ or mutating Ca2+-binding residues reduces rigidity and unfolding strength. The deafness mutation D101G, in the linker region between the repeats, reduces the affinity for Ca2+ to allow unfolding at lower forces. The structures define an uncharacterized cadherin family, they suggest a novel binding interaction with protocadherin 15, and they indicate mechanisms underlying inherited deafness.

Back To Top

Protein Turnover in Hair Cells

Hair bundles have an elaborate and highly stereotyped morphology, which is essential for conveying mechanical stimuli to transduction channels. However, hair cells are not replaced during the life of an animal, so they need to replace degraded proteins on a regular basis. Tip links, for instance, can be replaced in 5-10 hours, and actin in the bundles of neonatal hair cells in culture is thought to turn over in two days. Together with Claude Lechene (Brigham & Women's Hospital), we have used a new method, multi-isotope imaging mass spectrometry (MIMS), to quantify protein turnover in defined subcellular compartments. This method can detect atoms of specific isotopic mass and has spatial resolution near that of electron microscopy. We fed precursor amino acids labeled with the stable isotope 15N to frogs and mice, we sacrificed after times of days to months, and we recorded quantitative images at masses corresponding 14N and 15N. The 15N/14N ratios in the inner ear reveal regions of high and low protein turnover. Particularly slow turnover is apparent in the otolithic and tectorial membranes, which convey the stimulus to the hair cells, and in the stereocilia, where mechanotransduction occurs. These results, obtained in adult animals in vivo, indicate that the normal turnover of protein in stereocilia is slower than previously suggested. The most stable structures in cochlea are stiff elements carrying the mechanical stimulus: the tectorial membrane, the pillar cells, the reticular lamina and the stereocilia.

These results are in striking contrast to those reported for actin treadmilling in cultured neonatal hair cells. However MIMS experiments to explore protein turnover in neonatal animals in vivo, and in cultured neonatal hair cells, suggest similarly slow protein turnover. Finally, actin turnover assessed by laser bleaching of GFP-tagged actin in stereocilia is also not consistent with rapid treadmilling.

15N/14N ratios in mouse utricle hair bundles. There is normal turnover in cell bodies of hair cells and supporting cells, but very little in stereocilia, except at the tips.

Back To Top

Gene Expression Profiling

Because there are relatively few hair cells in the inner ear, biochemical characterization of their constituent proteins has been difficult, and there are consequently few good candidates for proteins of the mechanical transduction apparatus. We have sought to determine the transcriptome of hair cells from the mouse utricle using oligonucleotide arrays (GeneChips). In collaboration with Zheng-Yi Chen (Massachusetts Eye & Ear Infirmary) we characterized gene expression in the utricular epithelium at developmental stages from E12 to P20. We have also characterized expression in epithelia lacking hair cells (Atoh1 knockout) and those with hair cells but lacking hair bundles (Brn3c knockout). Finally, we have induced expression of the bHLH transcription factor Atoh1 in a human osteosarcoma cell line to identify genes driven by Atoh1.

We found that expression of HES6, among other genes, is induced by Atoh1. In situ hybridisation showed that the rise and fall of Hes6 expression closely follow that of Atoh1 in cochlear hair cells. Moreover, electrophoretic mobility shift assays and luciferase assays show that human ATOH1 directly activates HES6 transcription through binding on three clustered E boxes of its promoter.

Acetylcholine is a key neurotransmitter of the inner ear efferent system, but only two nicotinic acetylcholine receptor (nAChR) subunits have been described in inner ear hair cells: α9 and α10. From expression profiling, we identified a third nAChR subunit in the inner ear—α1, encoded by Chrna1—and showed that Chrna1 is expressed in all types of inner ear hair cells. Its expression begins at embryonic stage E13.5 in the vestibular system and E17.5 in the organ of Corti. Atoh1 activates CHRNA1 transcription through binding to two E boxes located on the proximal promoter. Moreover, we found that the β1, and γ subunits are also expressed in wild-type mice cochlea, with the γ subunit also under control of Atoh1. The α1 subunit, best known as the α subunit of the muscle nAChR, does not change the electrophysiological properties of the α9/α10 acetylcholine receptor in Xenopus laevis oocytes. Instead, α1 in inner ear hair cells may be part of a novel nicotinic acetylcholine receptor resembling the muscle-type receptor.

These extensive expression databases are now being mined for other proteins that may be associated with hair cell transduction. Refinement of expression profiles by deep sequencing of pure hair-cell mRNA is underway.

GeneChip images for wild-type (top) and α-tectorin knockout (bottom) mice. Boxed are the probes for α-tectorin.

Back To Top

TRP Channels

The force-gated ion channel that mediates mechanotransduction in hair cells has not been identified. One clue comes from its permeation properties: the channel is only weakly selective among cations, passing some as large as 11 Å in diameter, and it is especially permeable to Ca2+. The TRP family of ion channels shares some of these properties and is involved in sensory transduction in a variety of other systems. To determine whether the transduction channel is a member of the TRP family, we used in situ hybridization to determine which, if any, TRP channels are expressed by hair cells. Of the 33 mouse TRP channels, at least five are made by hair cells. We have tested auditory function in mutant mouse lines deficient in four of these.

TRPA1 was an attractive candidate, because it is expressed in hair cells at about the time they become mechanically sensitive. In addition, TRPA1 has an extended N-terminal domain encompassing 17 ankyrin repeats, and we showed with molecular dynamics simulations that such domains show "tertiary-structure elasticity" with a stiffness close to that expected for the hair-cell gating spring. However a TRPA1 knockout mouse that we made has no observable deficit in mechanosensation, and it is unlikely that other TRP channels could compensate for TRPA1.

The PKD2 class of TRP channels are also interesting candidates. We have tested knockouts of PKD2, PKD2L1, and PKD2L2. Individually, none shows significant hearing deficit. Because PKD2 channels are similar to each other, compensation is a possibility and we are making double and triple knockouts.

Back To Top

TRPA1 Structure and Function

TRPA1 is expressed by hair cells, but also by dorsal root ganglion neurons where it has proposed roles in sensing painful cold and irritating chemicals. Knockout mice display behavioral deficits in response to mustard oil and bradykinin, to cold (~0° C), and to punctate mechanical stimuli. Because behavioral assays of sensory function are not very direct, we also used the ex-vivo skin-nerve preparation, together with the laboratory of Cheryl Stucky, to directly determine the contribution of TRPA1 to cold and mechanotransduction at the level of the primary afferent terminal. Cutaneous fibers from TRPA1-deficient mice show no deficits in cold sensitivity, at least not for short exposure, but they display striking deficits in mechanical response properties. C fiber nociceptors from TRPA1-deficient mice exhibit action potential firing rates 50% lower than those in wild-type C fibers across a wide range of force intensities. Aδ fiber mechano-nociceptors also have reduced firing, but only at high intensity forces. Immunostaining of skin for TRPA1 reveals extensive labeling of keratinocytes, and labeling of sensory neurons with a wide range of cell body sizes. Thus, TRPA1 is not restricted to thin-caliber axons and epidermal endings, but is also found in many large caliber axons, Meissner and lanceolate endings. We think that TRPA1 modulates mechanotransduction in two ways: through a cell-autonomous function in sensory nerve terminals, and through a modulatory role in keratinocytes which interact with sensory terminals to modify their mechanical firing properties.

Because TRPA1 is a direct receptor or indirect effector for a wide variety of nociceptive signals, it is a compelling target for development of analgesic pharmaceuticals such as channel blockers. We investigated the pore and selectivity filter of TRPA1, primarily through point mutations of key charged residues. We find that the glutamate at position 920 plays an important role in collecting cations into the mouth of the pore, by changing the effective surface potential by ~16 mV, but acidic residues further out have little effect on permeation. The aspartate at position 915 represents the constriction site of the TRPA1 pore, which is ~8.3 Å in diameter, and its charge is critical in assisting ion permeation.

Sequence and predicted arrangement of acidic residues in the TRPA1 pore.

Back To Top