What is the hair cell transduction channel?

Abstract

In contrast to nearly all other sensory systems, the mechanically sensitive ion channel carrying the receptor current into hair cells of the inner ear has not been identified in molecular terms. A number of candidates from at least two different ion channel families have been considered: these include the epithelial sodium channel (ENaC) and acid-sensing ion channel (ASIC) members of the DEG/ENaC superfamily of amiloride-sensitive sodium channels, as well as the TRP channels TRPN1, TRPV4, TRPML3 and TRPA1. For each, initial supportive results were followed by further studies that cast doubts on their involvement. No promising candidates have recently emerged, but the TRP family continues to be attractive in general.

The ion channels carrying the stimulus-activated receptor current in sensory neurons have been identified and the mechanism of their activation elucidated, for a variety of sensory systems and for species ranging from worms and flies to humans. In striking contrast are those mediating hearing and balance in vertebrates – the hair-cell transduction channels. Hypothesized as a distinct class of ion channel and characterized biophysically more than 20 years ago (Corey & Hudspeth, 1983), they have been the object of a continuous quest since then. In the past 10 years, a variety of candidates have appeared, only to disappear, wraith-like, in the clear light of further experiments. At the present time we do not know the molecular identity of the transduction channel, but these previous apparitions are bringing us closer to an answer.

We know very well what the hair cell transduction channel should look like, in physiological terms. It is a non-selective cation channel that is permeable to all the alkali cations and to many divalent cations (Corey & Hudspeth, 1979; Ohmori, 1985). The channel has a particularly high calcium permeability, passing Ca²⁺ at least 10 times better than Na⁺ (Lumpkin et al. 1997; Ricci & Fettiplace, 1998); at the same time, Ca²⁺ in millimolar concentrations acts as a partial blocker, inhibiting current by monovalent ions. Remarkably, the channel is also permeable to small organic cations, even fluorescent styryl dyes like FM1-43 of molecular weight > 450 Da (Corey & Hudspeth, 1979; Gale et al. 2001; Meyers et al. 2003; Farris et al. 2004). There are no highly specific blockers: amiloride and its analogues, as well as aminoglycosides, block at low micromolar concentrations in a voltage-dependent manner. The block is relieved at more negative potentials (Jorgensen & Ohmori, 1988; Kroese et al. 1989; Rusch et al. 1994) suggesting that these charged compounds can be driven through the channel by voltage (Marcotti et al. 2005). A systematic study of blockers indicated that the pore is ∼1.3 nm in diameter at its narrowest, with a larger vestibule extending ∼1.5 nm in from the outside (Farris et al. 2004). Perhaps consistent with the lack of selectivity, the conductance of the channel is large, ranging from 100 to 300 pS or more (Crawford et al. 1991; Geleoc et al. 1997; Ricci et al. 2003). That the conductance can vary by 3-fold or more in different hair cells from a single organ (Ricci et al. 2003) was quite unexpected, but it puts further constraints on prospective candidates.

As there are no high-affinity ligands for the transduction channel (and probably not enough channel protein in the tissue to sequence even if it could be affinity purified), a variety of indirect strategies have been used to search for the transduction channel. These include searches for genes that cause deafness when mutated, searches for channels related to mechanosensory channels in other organs, and searches for channels with similar selectivity and permeability.

Amiloride-sensitive sodium channels

Following initial observations that amiloride reduced the sensitivity of lateral line hair cells, Jorgensen & Ohmori (1988) used single-cell recordings to show that amiloride blocks the channel in a voltage-dependent manner. The half-blocking concentration is nearly 100-fold higher for hair cells than for the amiloride-sensitive sodium channels of transporting epithelia (the ENaC channels); nevertheless this raised interest in ENaCs as candidates for the transduction channels. Hackney et al. (1992) used an antiserum raised against a biochemically purified ENaC for immunogold labelling of stereocilia, and found the highest density of immunoreactivity at the point of closest apposition between adjacent stereocilia. With electron microscopy, this could be distinguished from the location of the tip links (just above the apposition), where channels are thought to be located (Pickles et al. 1984).

Interest in the ENaC family of channels was raised considerably when Rossier and colleagues cloned the α, β and δ subunits of the ENaC channel from rat colon (Canessa et al. 1993, 1994), and recognized that they were part of a gene family that also included two proteins, MEC-4 and MEC-10, which are defective in mechanosensation mutants of the nematode C. elegans (Chalfie & Sulston, 1981; Chalfie & Au, 1989). A related protein, DEG-1, causes degenerative neuronal death when certain residues are mutated, conferring the name ‘degenerins’ or DEGs to this channel family in worms (Chalfie & Wolinsky, 1990) and contributing to the rather cumbersome name DEG/ENaC for the channel superfamily (Corey & García-Añoveros, 1996). Chalfie et al. (1993) revealed more extensive sequence similarity between the two channel groups, and suggested in addition that the hair cell transduction channel might be an ENaC.

Several groups then searched for ENaCs or related channels in mammalian tissues, leading to the discovery of αENaC isoforms in the chick cochlea (Killick & Richardson, 1997), and to the discovery of another branch of the DEG/ENaC family in mammals, now termed the ASICs (Price et al. 1996; Waldmann et al. 1996; García-Añoveros et al. 1997). A number of the ASIC channel subunits are expressed by mechanosensory neurons of the dorsal root and trigeminal ganglia and are transported to the sensory endings in skin (e.g. García-Añoveros et al. 2001), suggesting that they may constitute mechanically activated ion channels.

Despite these intriguing findings, problems soon arose that cast doubt on the involvement of DEG/ENaCs in hair-cell transduction. First, the DEG/ENaC channels have too low a conductance (10–15 pS), too high a Na⁺ selectivity (P_Na/P_K = 5–100), and too low a Ca²⁺ permeability (P_Ca/P_K < 0.4) (Kellenberger & Schild, 2002) to be consistent with the hair cell transduction channel. In addition, knockout mice seem to have normal inner ear function. Rusch & Hummler (1999) found that αENaC knockout mice had normal vestibular behaviour, and normal receptor currents in cochlear outer hair cells. ASIC2 knockouts have normal hearing, as measured by the auditory brainstem-evoked response (ABR) (Peng et al. 2004; Roza et al. 2004), and ASIC1 and ASIC3 knockouts have no reported auditory or vestibular deficit (Xie et al. 2003). At this point, there is no support for involvement of the DEG/ENaCs in hair cell transduction.

NOMPC (TRPN1)

A genetic screen for mechanosensation genes in Drosophila revealed another candidate ion channel family. Kernan et al. (1994) found five ‘nomp’ genes whose mutation caused no mechanoreceptor potential in the bristle organs. One of them, nompC, was found to encode a protein homologous to members of the transient receptor potential (or TRP) family of ion channels (Walker et al. 2000), a group of channels typified by a large conductance and a high Ca²⁺ permeability appropriate for the hair cell channel (Owsianik et al. 2006). nompC is expressed in the sensory bristle complexes in Drosophila and a homolog is in mechanosensory neurons in C. elegans. Three nompC alleles had nonsense mutations; in those the large, transient part of the mechanoreceptor potential is absent. A fourth had a missense mutation that causes the transient to be faster, suggesting that nompC is itself a mechanically gated ion channel or is intimately associated with it (Walker et al. 2000). While there is presently no additional evidence for the function of nompC in bristle mechanotransduction, the involvement of a TRP channel in a ciliated mechanosensory organ not unlike a hair cell focused new attention on the TRP channel family.

Although the nompC (now TRPN1) channel was initially thought to be restricted to invertebrate animals, Sidi et al. (2003) found an ortholog in the zebrafish genome. In situ hybridization showed weak expression in inner ear hair cells by 48 h post-fertilization, when hair cells begin to show mechanosensitivity in zebrafish. To inhibit normal expression of zebrafish TRPN1, Sidi et al. injected fertilized eggs with morpholino oligonucleotides targeting the splice donor site of exon 28, so that missplicing would delete the transmembrane domains. Successful disruption of the mRNA in morphant fish was detected by PCR. Morphants lacked an ‘acoustic’ startle response elicited by tapping the dish, and they swam in circles or sideways suggesting vestibular dysfunction. In fish that lacked the startle response, hair cells of the lateral line neuromast organs did not accumulate the fluorescent dye FM1-43, a marker of functional transduction channels, and they did not generate a microphonic potential when stimulated with sinusoidal fluid flow. In these experiments, however, relatively large amounts of morpholino were injected, and although the fish studied had no visible morphogenic defects there may have been off-target effects that were undetected and that non-specifically affected hair cell function.

TRPN1 has also been detected in two frog species: Xenopus laevis (Shin et al. 2005) and Rana catesbeiana (D. Tamasauskas, K. Y. Kwan and D. P. Corey, unpublished observation). A good antibody to the Xenopus TRPN1 labelled the lateral line hair cells of stage 48 embryos and saccular hair cells of adult Xenopus. However, the label was detected in the kinocilia, but not in the stereocilia of saccular hair cells where the most or all of the transduction channels are located (Shin et al. 2005). Another class of ciliated epidermal cells was strongly labelled in Xenopus, supporting the association of TRPN1 with microtubule-based cilia. Perhaps TRPN1 in kinocilia is somehow involved in the mechanical coupling of the mechanical stimulus to the stereocilia bundle, which could explain a hearing deficit in morphant zebrafish.

Whatever the role of TRPN1 in fish and frogs, it cannot be part of the transduction channel in all vertebrates. The TRPN1 gene is entirely absent (and not simply a pseudogene) in mammals and birds, and is also missing in some other fishes (the pufferfishes Fugu rubripes and Tetraodon nigroviridis).

TRPV4

A genetic screen in C. elegans for genes involved in osmotic sensation revealed a new TRP channel, Osm-9, related to the vertebrate capsaicin receptor, TRPV1. Heller and colleagues (Liedtke et al. 2000), then screened a chicken inner ear library and mammalian brain and kidney libraries for additional homologs of Osm-9 and TRPV1. They discovered (simultaneously with three other groups) a channel now known as TRPV4. TRPV4 is expressed in most cells lining the endolymphatic duct of the mouse ear, including the hair cells and the marginal cells of the stria vascularis (Liedtke et al. 2000; Shen et al. 2006). TRPV4 is mechanically sensitive, in that it is activated by osmotic cell swelling and by fluid flow. It is also activated by heat and by phorbol esters (see O'Neil & Heller, 2005 for review). Like the hair cell transduction channel, it has a high Ca²⁺ permeability and a conductance of about 90 pS (Owsianik et al. 2006). Finally, TRPV4 knockout mice have a hearing deficit (Tabuchi et al. 2005).

While TRPV4 was an intriguing possibility, there was doubt about its role in transduction from the beginning (Liedtke et al. 2000). The strongest expression in the cochlea is not in hair cells but in the stria, which is a transporting epithelium and where an osmoregulatory role makes sense. Its activation by mechanical stimuli is slow, suggestive of a second messenger intermediate and inconsistent with transduction (O'Neil & Heller, 2005). Also the deficit in the mouse knockout only appears at several months of age, and represents no more than about 20 dB of threshold shift (Tabuchi et al. 2005).

TRPML3

A large number of mutant mice exhibit hearing loss, sometimes in combination with other deficits, and some of the affected genes have been identified. For instance, Varitint waddler mutant mice have pigmentation defects, and also show deafness and circling behaviour that is attributable to progressive cytoplasmic pathology of the hair cells and disorganization of hair cell stereocilia in the cochlea (Cable & Steel, 1998; Di Palma et al. 2002). Both alleles of Varitint waddler (Va and Va^J) act in a semi-dominant manner, with hearing loss and stereocilia disorganization in heterozygotes; Va homozygotes are usually lethal but Va^J homozygotes are not. The mutated gene is most probably that encoding another TRP channel, TRPML3 or mucolipin-3 (Di Palma et al. 2002). The more severe Va allele has a single amino acid substitution (A419P) near the cytoplasmic end of the putative fifth transmembrane domain, likely to contribute to the ion conduction pathway; this would explain a dominant phenotype, since one mutant subunit could block conduction in a multimeric channel. The less severe Va^J allele has, in addition, a substitution between the putative third and fourth transmembrane domain (I362T), which apparently compensates in part for the A419P mutation.

There are two other members of the TRPML subfamily in mammals. Both TRPML1 and TRPML2 are associated with the lysosomal compartment of cells and TRPML1 is important for lysosome acidification. Fluorescence resonant energy transfer experiments show that all three TRPML proteins associate with one another, and coexpression of TRPML3 with either of the others localizes it to lysosomes in cultured cells (Venkatachalam et al. 2006). Indeed, in hair cells, antibody labelling of TRPML3 showed it to be primarily in organelles in the cell body, albeit with faint label in stereocilia (Di Palma et al. 2002). Thus it seems that TRPML3 is primarily in cytoplasmic organelles and not the plasma membrane, and that the deafness arises not from mutation of a transduction channel but from indirect effects of abnormal organelle trafficking. While TRPML3 cannot be ruled out as a transduction channel candidate, it seems unlikely.

TRPA1

A growing number of other TRP channels have been implicated in mechanosensation, in both vertebrates and invertebrates (Sukharev & Corey, 2004; Lin & Corey, 2005). In addition, the TRP family is attractive for its generally high conductance, low selectivity and high Ca²⁺ permeability (Owsianik et al. 2006). These led Corey et al. (2004) to screen all 33 mouse TRP channels, using in situ hybridization. Several TRPs were found to be expressed in hair cells (including TRPML3), but a particularly good candidate was TRPA1. First, TRPA1 is first expressed in the sensory epithelium of the mouse utricle (based on quantitative RT-PCR) at embryonic day 17, the same time during development that these hair cells become mechanically sensitive (Geleoc & Holt, 2003). An affinity-purified antibody to a C-terminal fragment of TRPA1 labelled the tips of stereocilia in bullfrog and mouse vestibular hair cells. Label was weaker but present in cochlear hair cells. Label in bullfrog hair cells was nearly eliminated by brief treatment with BAPTA or La³⁺, which cuts the tip links and is thought to promote recycling of transduction components from the stereocilia (Siemens et al. 2004). In the zebrafish genome, two TRPA1 orthologs were discovered. Injection of eggs with morpholino oligonucleotides targeting TRPA1a but not TRPA1b caused diminished FM1-43 loading of both lateral line and inner ear hair cells (at 50–60 h post-fertilization), and caused smaller microphonic potentials in the inner ear in response to vibratory stimulation of the body. The mouse genome harbours a single TRPA gene, whose expression was inhibited with siRNAs. Hair cells in utricules cultured from embryonic day (E) 15 mice – chosen because the mRNA has not appeared at that point – were infected at E16 with adenoviruses encoding one of two siRNAs to TRPA1 and were tested physiologically at E17–E18. Infected cells (identified by the GFP also encoded in the virus) had little or no transduction current and only slight loading with FM1-43. A control adenovirus encoding only GFP did not reduce the transduction current (Corey et al. 2004). Together, this evidence strongly supported TRPA1 as a candidate for the channel.

TRPA1 was additionally attractive because it is unique among mammalian TRPs in having a large number (17) of ankyrin repeats in its extended N-terminus. The alpha helices of polyankyrin domains pack into a curved structure, which was shown by both steered molecular dynamics and atomic force microscopy to have an elasticity nearly the same as that measured for the hair cell ‘gating spring’ (Sotomayor et al. 2005; Lee et al. 2006). Moreover, the only other TRP channel with a large number of ankyrin repeats is TRPN1, also thought to be a mechanosensitive channel. The intriguing possibility arose that TRPA1 has both a mechanically sensitive channel domain and an elastic domain that pulls the channel open (Howard & Bechstedt, 2004; Corey & Sotomayor, 2004).

Trouble with the hypothesis appeared almost immediately. Hair cell transduction is not affected by allyl isothiocyanate (M. A. Vollrath and D. P. Corey, unpublished observations), a known activator of TRPA1 (Jordt et al. 2004). The transduction channel has a sensitivity to blockers mostly like that of TRPA1, but the transduction channel is 10–20 times more sensitive to amiloride than is TRPA1 (Nagata et al. 2005), and is 100 times less sensitive to Gd³⁺ than is TRPA1. TRPA1 is thought to be activated by intracellular Ca²⁺ (Jordt et al. 2004), but Ca²⁺ inhibits the transduction channel (Howard & Hudspeth, 1988; Cheung & Corey, 2006). However, all of these differences could be attributed to the possibility that TRPA1 in vivo assembles in a multimeric channel with other subunits, which alter its pharmacology.

To further test its function, both Bautista et al. 2006 and Kwan et al. (2006) deleted critical exons from the mouse TRPA1. Although the mice showed the expected deficits in cutaneous pain sensation, consistent with the known expression of TRPA1 in small diameter neurons of the dorsal root and trigeminal ganglia, they were not deaf. A detailed evaluation showed normal acoustic startle reflex and vestibular behaviour, normal ABR, and normal transduction and adaptation in the receptor currents of single utricular hair cells (Kwan et al. 2006). These results are difficult to reconcile with the idea of TRPA1 as the transduction channel.

What went wrong? There are two classes of explanation. It might be, on the one hand, that the absence of TRPA1 throughout hair cell development causes the up-regulation of another (TRP?) channel to compensate. Compensation is a common story in knockout mice; however, if the polyankyrin domain is important for function, there is no other mammalian TRP that could compensate. It might be that TRPA1 is a subunit of one group of transduction channels, for instance at the lower end of each tip link, while other proteins constitute channels at the upper end. Deletion of half might not produce deafness. However, the total transduction current was not smaller in the knockout mice (Kwan et al. 2006) and in any case there is no kinetic or pharmacologic evidence for two distinct mechanically activated conductances in hair cells. On the other hand, it might be that each experiment supporting TRPA1 was somehow flawed: Even affinity-purified antibodies can bind to other proteins, perhaps even proteins of the transduction complex. Morpholinos can disrupt development non-specifically, and the inhibition of hair cell function with TRPA1 morpholinos was 50–70% at best. Even though the adenovirus was not toxic to mouse hair cells, siRNAs can have off-target effects. A variety of further tests, no doubt underway in several laboratories, should distinguish these possibilities.

Conclusions

What, then, is the hair cell transduction channel? TRPA1 is still an attractive candidate, but only if another TRP channel could compensate for it in knockout mice. TRP channels in general are enticing, for their high single-channel conductance, their non-selective pore with high Ca²⁺ permeability, and for their association with sensory transduction and mechanosensation in a variety of other tissues (Clapham, 2003; Lin & Corey, 2005; Owsianik et al. 2006). Since most TRP channels were identified by sequence similarity rather than by functional or positional cloning, functions are not well understood for the majority of the 30-odd TRPs in mammalian genomes. Most remain potential candidates. Some other channel classes have high conductance and/or non-selective pores; these include acetylcholine receptors, glutamate receptors, cyclic-nucleotide-gated channels, the P2X ATP-gated channels, and connexin hemichannels. Yet without more systematic search strategies, there may be more phantoms to chase before we have the real channel in hand.