Application of physiological genomics to the study of hearing disorders



Although the biophysical principles of how the ear operates are reasonably well understood, little is known about the specific genes that confer normal function to the inner ear. Nevertheless, the recent implementation of genomic tools has led to extraordinary progress in the identification of mutated genes that cause non-syndromic and syndromic forms of deafness. Part of this success is directly related to the sequencing of the human and mouse genomes and improved gene annotation methods. This review discusses how physiological genomic tools, such as genomic databases, expressed sequence tag databases and DNA arrays have been applied to find candidate genes for important molecular processes in the inner ear. It also illustrates, using the discovery of genes encoding essential components of cochlear K+ homeostasis as an example, how the combination of physiological genomic tools with physiological and morphological information has led to an in-depth understanding of cochlear ion homeostasis. Finally, it discusses how the use of applied genomic tools, such as gene arrays, will further advance our knowledge of how the inner ear works, develops, ages and regenerates.

Physiological genomic approaches to study inner ear function

The application of multi-disciplinary approaches to physiology is not a new theme. In the study of the inner ear, for example, physiology and biophysics have melded together in the form of exciting and innovative research producing an elegant portrait of how the cochlea and the vestibular system work. A thorough understanding of these processes at a molecular level, however, has followed a rockier path. The inner ear is a morphologically complex organ that is deeply hidden inside the skull and protected by the extremely hard temporal bone that is very accurately named ‘Felsenbein', meaning ‘rock bone’ in German. Careful dissection of the different inner ear organs yields very little cellular material for applied molecular-biological studies. One of the consequences of this paucity of cells, especially of the ear's featured sensory receptors, the hair cells, is that the generation of cDNA libraries is extremely laborious and therefore only a small number of cDNA libraries exist. As a result, hearing researchers do not have a vast number of cDNA libraries at their disposal enabling the use of a wide variety of standard molecular-biological techniques. This limitation has impeded the use of molecular-biological methods to study functionally important inner ear proteins.

Recent genetic approaches, however, have been very successful in the identification of genes that are important for inner ear function and development. This success is predominantly due to the application of genomic tools. For example, advances in linkage analysis and statistical genetics have resulted in identification of more than 80 human loci linked to non-syndromic hearing disorders (G. van Camp & R. J. Smith, Likewise, about 150 murine loci linked to deafness or vestibular dysfunction have been described (Q. Y. Zheng et al., Approximately one-third of the human loci have been linked to a defective gene, mostly via candidate gene approaches. Annotated genomic databases are a major tool for identifying candidate genes for hearing disorders. These databases provide lists of known genes and putative genes that are physically localized within the genomic region implicated in causing the inner ear defect. The investigator who has identified the affected genomic region and is searching for the associated genetic defect has to select appropriate candidate genes from within those candidate lists, based upon the genes' expression patterns or the function of the encoded proteins.

Candidate genes

One criterion for a candidate gene is selective expression of its corresponding mRNA in inner ear cells. For example, a cDNA clone for cochlin – the protein encoded by the COCH gene – was initially found in cDNA that was enriched by subtractive hybridization for clones that are expressed in the inner ear but not in other organs (Robertson et al. 1997; Heller et al. 1998). Subtractive hybridization is a method that allows comparison of two cDNA populations: for example a cochlear cDNA preparation with a mixture of cDNA from other organs, resulting in selective enrichment of cDNAs that are present in one of the source cDNA preparations. The ear-specific expression of cochlin mRNA was confirmed by Northern blot analysis, a method used to detect and to compare the relative amounts of specific mRNA among RNA preparations from different organs.

Mapping of the chromosomal location of COCH placed the gene within the linkage region of the human non-syndromic deafness disorder DFNA9 (DFN stands for deafness, A for autosomal dominant). Sequence analysis of genomic DNA of affected patients revealed that DFNA9 is indeed caused by mutations in COCH (Robertson et al. 1998). Although more work has to be done to determine the in vivo function of cochlin, the fact that missense mutations in the COCH gene lead to deafness and vestibular dysfunction in a dominant manner, provides clues with regard to a potential pathological effect of the mutated cochlin isoform.

Because the human genome has been completely sequenced and assembled, it is now possible to immediately assess whether a cDNA clone whose corresponding mRNA is restricted to cells of the inner ear, is derived from a gene that is located within the linkage region of a hearing disorder. Detailed mapping of the chromosomal location of the candidate gene, as done for COCH, is therefore no longer required and is usually replaced by use of genome database resources, for example at the National Center for Biotechnology Information ( Searching the human genomic database with the candidate cDNA sequence determines whether the corresponding gene is located on the same genomic interval as the hearing disorder.

Expressed sequence tag databases

The finding that mutations in COCH cause deafness emphasizes that gene expression analysis is an important tool of physiological genomics to study inner ear disorders. Candidates for gene expression analysis are usually selected from inner ear cDNA libraries. Although the scarcity of material that can be obtained from inner ear specimens has slowed the construction of cDNA libraries, a number of excellent inner ear-derived cDNA libraries have recently been reported (Crozet et al. 1997; Skvorak et al. 1997; Heller et al. 1998; Cyr et al. 2000; Verpy et al. 2000). Whereas the majority of genes represented by inner ear-derived clones are also expressed in other organs, a small population of these genes is restricted to the inner ear. An efficient method to test expression of individual clones is provided through abundance analysis of expressed sequence tags (ESTs). ESTs are partial sequences of cDNA clones that are the result of high-throughput-sequencing efforts of inner ear and mostly other cDNA libraries. Since ESTs are derived from cDNAs, they represent genes that are expressed in the organ used for construction of the individual cDNA library. Several thousand inner ear ESTs are already publicly available (search Genbank at for EST and COCHLEA or for EST and EAR). Given that EST databases mainly contain ESTs that are not derived from the inner ear, it is relatively easy to ascertain via search algorithms whether an individual inner ear cDNA clone, represented by an EST, is ear-specific or is expressed in other organs. ESTs that can only be found in inner ear cDNA libraries are better candidates for being implicated in non-syndromic deafness than ESTs that are more widespread. As the inner ear harbours a variety of cell types, ESTs that are derived from this organ imply that the corresponding gene is active in the ear but it does not implicate a specific cell type. More detailed analysis of cellular gene expression can be achieved by in situ detection of the corresponding mRNA (for examples see Heller et al. 1998; Robertson et al. 1998; Lanford et al. 1999). This powerful method uses tagged complementary RNA (cRNA) antisense to the targeted mRNA. Antisense cRNA is most commonly labelled by incorporation of radioactive ribonucleotides or by using ribonucleotides that are tagged with the hapten digoxigenin that is suitable for antibody staining. Whereas radioactively labelled antisense cRNA provides higher sensitivity, it lacks the high cellular resolution achieved with digoxigenin-labelled cRNA probes. In situ hybridization has detection limitations, especially with non-radioactive cRNA probes, and cannot be applied to demonstrate low mRNA levels. A more sensitive method to detect mRNA is in situ amplification of reverse transcribed cDNA by polymerase chain reaction (PCR) (Rosenblatt et al. 1997). Although in situ PCR is a powerful method, it requires laborious optimization and multiple experimental controls, and may therefore not be suitable for routine analysis of transcripts found through EST database analysis.

Improved genomic databases

The reader may ask why the complete sequencing of the human and mouse genome is not providing sufficient data for immediate identification of all mammalian deafness genes. The rough sequence data that has so efficiently been assembled by supercomputers lacks the most important information, namely which stretches of nucleotide sequence actually encode genes. Only a minority of mammalian genes are already well characterized and can easily be annotated. New genes are constantly being defined by sophisticated algorithms that analyse genomic sequence (Claverie, 1997; Rouze et al. 1999; Gaasterland et al. 2000; Gaasterland & Oprea, 2001). One of the astonishing predictions that were made after the complete sequencing of the human genome was that the genome harbours only 30 000–40 000 genes (Lander et al. 2001; Venter et al. 2001). However, subsequent comparison of the predicted genes that arose from public and commercial sequencing projects and inclusion of a third cluster of known genes revealed little overlap among these three groups (Hogenesch et al. 2001). The conclusion, therefore, is that there are probably significantly more than 40 000 genes in the mammalian genome and that most are not easy to identify. Despite advances in computational gene prediction, the only proof that a stretch of genomic DNA encodes a protein-coding gene is when a corresponding mRNA can be found.

Fortunately, significant progress has been made in high-throughput partial sequencing of clones derived from cDNA libraries of different organs. The continuous combination of the resulting EST databases with genomic databases provides a new level of annotation, namely the physical genomic location of an expressed gene coupled with preliminary expression data of the corresponding mRNA. Several ESTs can often be assembled to a continuous stretch of cDNA called a contig. Contigs often reveal at least the partial amino-acid sequence of the corresponding protein that usually can be supplemented by using genomic DNA sequence information. Knowledge of a novel protein's sequence is very helpful for preliminary classification and sometimes reveals a possible function. A basic classification could be based on the presence or absence of an amino-terminal signal sequence, putative membrane-spanning domains, or sequence domains that mediate nuclear targeting. Possible function of a novel protein could be revealed when the protein displays obvious homology with members of known protein families. Although several proteins with important functions in the cochlea cannot be classified in detail based on their protein sequence alone, at least some hypothesis can be made based on primary protein structure. The protein harmonin is an example for such a case. It is encoded by the USH1C gene which, when mutated, causes Usher syndrome type 1, a disorder characterized by profound sensorineural deafness, vestibular dysfunction and blindness. Harmonin contains a region with sequence homology to the consensus sequence for PDZ domains, a functional domain that has been implicated in protein interaction (Verpy et al. 2000). Harmonin therefore may functionally interact with other proteins via its PDZ domain. Although this potential feature does not provide evidence for a functional interaction between harmonin and other proteins, it provides a lead to a possible function that can be followed by searching for proteins that are capable of binding to harmonin's PDZ domain.

How do improved databases help in the study of hearing disorders? An obvious strategy is to analyse whether ESTs and annotated genes that are expressed in the cochlea localize within the genomic interval of deafness loci. Identification of physiological abnormalities of the patients or of the affected animals may be very useful in this regard. For example, a mouse mutant that displays reduced endocochlear potential may carry a mutation in a gene affecting cochlear K+ homeostasis (see below for a number of specific examples). Any gene that is related to the already identified components of cochlear K+ homeostasis (Table 1), is expressed in the cochlea, and is located within the linkage interval of the mutated locus would be an excellent candidate gene for causing the mouse strain's deafness. Alternatively, scans of the genome for genes that may be involved in, for example, renal ion homeostasis may identify genes that are relevant for hearing and balance because of the importance of K+ homeostasis for inner function. If the physical location of one of these genes is within the linkage interval of a deafness disorder, it may be worth checking whether the gene is mutated in affected individuals.

Table 1. Genes that alter cochlear K+ homeostasis when mutated
Gene Encoded proteinLocalizationFunction
GJB2 Connexin 26Fibrocytes and non-sensory epithelial cellsGap junction protein
GJB3 Connexin 31FibrocytesGap junction protein
GJB6 Connexin 30FibrocytesGap junction protein
GJA1 Connexin 43FibrocytesGap junction protein
POU3F4 Brn-4(Pou3f4)Ear mesenchymeTranscription factor
KCNE1 (ISK) KCNE1 (ISK)Strial marginal cellsK+ channel β-subunit
KCNQ1 (KvLQT1) KCNQ1(KvLQT1)Stria vascularisK+ channel β-subunit
KCNQ4 KCNQ4Outer hair cellsK+ channel
Slc12a2 NKCC1Strial marginal cellsNa+-K+-Cl co-transporter
BSND BarttinStrial marginal cellsCl channel β-subunit
KCNJ10 KCNJ10(Kir4.1)Strial intermediate and Deiters cellsK+ channel

A helpful accessory tool in this analysis is the before mentioned in situ hybridization technique. In situ hybridization can be applied in a systematic manner to test the cellular cochlear expression of either known genes or of genes that are represented solely by ESTs (Heller et al. 1998). Although this method cannot be used to exclude candidate genes because of its detection limitations, it can provide useful clues about the cellular expression of a gene and perhaps implicate a physiological role. In situ hybridization is unquestionably not a high-throughput method and the parallel analysis of the expression of several thousand different genes necessitates different techniques, such as DNA array technology.

DNA arrays

The current workhorse of the genomics revolution – the DNA array or Gene Chip – comes in three flavours. A reliable technology is the so-called macroarray, which consists of up to several tens of thousands of cDNA clones spotted at a high density on nylon filters. Typically, these filters are hybridized with cell- or organ-specific probes and analysed using storage phosphor autoradiography. Individual arrays can be reused for several hybridization experiments. The more widely used DNA microarrays are single-use glass slides that can harbour up to tens of thousands of polymerase chain reaction-amplified cDNA fragments (Schena et al. 1995). The advantage of DNA microarrays is that two different mRNA-derived probes, prepared from individual cells or whole organs, can be hybridized simultaneously, because the methodology employs two different probe-tags used for detection (Shalon et al. 1996). This feature allows the direct comparison of gene expression profiles (Schena et al. 1996). Finally, Gene Chips are high-density oligonucleotide arrays that are manufactured with photolithographic fabrication techniques employed in the semiconductor industry. Gene Chips combine the features of DNA microarrays with increased specificity and higher density. The increased specificity of Gene Chips is a result of the use of different unique oligonucleotides for each individual gene that is represented on the array. Detection of signals on DNA microarrays and Gene Chips is usually done with scanners that laser-excite the fluorescent probe-tags and convert the emitted light into high-resolution images. All array technologies are commercially distributed.

The main application of gene arrays in hearing research is gene expression profiling. Commercial arrays harbour collections of cDNA or EST clones that can be hybridized with probes derived from mRNA of whole auditory organs or even of individual cell types. Comparison of the results of such hybridization experiments provides information about differential gene expression among the specimens used for probe generation. Potential applications are abundant – although researchers who study the cell types of the inner ear once again face the problems caused by the scarcity of material. It is, for example, much easier to purify millions of immune cells than to obtain a handful of inner ear hair cells to create probes for gene expression profiling using DNA arrays. Such issues slow the widespread use of array technology but these hurdles will soon be minimized or even eliminated. Single-cell probe generation will allow simultaneous interrogation of genes expressed at different developmental stages of cochlear hair cells. Physiology has already provided clues that murine hair cells change their electrophysiological properties during the first three postnatal weeks (Kros et al. 1998; Rusch et al. 1998) and gene expression profiling may elucidate the corresponding changes in gene expression.

Other DNA array experiments may address differences in hair cells along the frequency gradient of the cochlea or differences between inner and outer hair cells of the organ of Corti. Mammalian outer hair cells are very unusual in this regard because they have a uniquely specialized lateral cell membrane that confers voltage-dependent contractility (Brownell et al. 1985; Kachar et al. 1986; Ashmore, 1987). This contractility is an interesting mammalian cochlear specialization that may modulate and supplement the general vertebrate cochlear amplifier that is probably provided by active elements in the hair bundle (Martin & Hudspeth, 1999). The structural components that mediate contractility are associated with the lateral cell membrane of outer hair cells. First morphological indications for this location arose from electron-microscopic studies that revealed highly structured protein complexes associated with the cell membrane and the actin cytoskeleton (Arima et al. 1991; Holley et al. 1992). The high abundance of these structures made it likely that genes encoding specific components of this specialization are highly active in outer hair cells. Inner hair cells do not have such a lateral cell membrane specialization and Zheng and colleagues (2000) made use of this observation by comparing the mRNA population of outer hair cells with mRNA prepared from inner hair cells. This comparison was done by subtracting inner hair cell cDNA from outer hair cell cDNA by homology hybridization and by selective amplification of cDNA molecules that are present only in outer hair cells (Zheng et al. 2000). The resulting cDNA was used to create a library of clones that are specifically expressed in outer hair cells. A jutted population of individual cDNAs from this library encoded the novel protein prestin, a membrane protein that confers voltage-dependent contractility to cell lines that normally do not have this feature (Zheng et al. 2000).

Prestin was found using a combination of cDNA subtraction and gene expression profiling using a relative small number of arrayed clones from an outer hair cell cDNA library. More comprehensive gene expression profiles of a number of inner ear-derived probes have already been created with commercial gene chips and are publicly available (Chen & Corey, 2001; Nevertheless, commercial gene chips do not harbour all possible genes and most inner ear-specific genes are not represented on current releases. It is therefore likely that the inner ear research community will not wait until ear-specific ESTs are explicitly included in commercial gene arrays. Custom-made cochlea cDNA macroarrays have already been used for the identification of genes expressed in the cochlea (Liedtke et al. 2000). The preliminary output of ongoing mammalian cochlea EST projects already displays a number of ESTs that do not occur in other organs (Skvorak et al. 1999). Creating arrays that harbour these human and murine cochlea ESTs will furnish the hearing research community with an exceptional tool for analysis of cochlear gene expression. Future applications of gene expression profiling in hearing research include the study of ear development, regeneration, trauma, ototoxicity and protection from ototoxicity, ageing, and also the analysis of mouse mutants. All parts of the auditory system – from individual hair cells to auditory brainstem nuclei – will provide plenty of opportunity for comprehensive analysis of the expression of genes important for hearing and vestibular function.

Candidate gene selection based on physiological relevance: interplay of genomics and physiology

While specific genetic defects responsible for hearing and vestibular disorders have been identified using gene expression analysis, other strategies have taken advantage of the detailed knowledge of inner ear physiology. The discovery of some of the key players in cochlear potassium homeostasis – a crucial mechanism of inner ear function – is an illustrative example of how knowledge of inner ear physiology and morphology (reviewed in Wangemann & Schacht, 1996) directed the identification of candidate genes for hearing disorders.

The cochlear duct, or scala media, is filled with endolymph, a fluid that is rich in K+ (≈150 mm) and maintained at a positive potential, i.e. the endocochlear potential of ≈85 mV (Fig. 1). Endolymphatic K+ carries the majority of the electrical charge that depolarizes hair cells during mechanoelectrical transduction, which occurs at the apical portion of the cell. Potassium is driven through mechanosensitive ion channels along an electrochemical gradient with the major driving force being the endocochlear potential manifested as a difference of ≈145 mV between the endolymph and a cochlear hair cell's cytosol. Hair cells extrude K+ through basolateral channels into the extracellular space. A possible recycling pathway for K+ through gap junctions of fibrocytes of the spiral ligament back to the stria vascularis has long been defined and supported by physiological and morphological findings (Jahnke, 1975; Wada et al. 1979; Santos-Sacchi & Dallos, 1983; Forge, 1984; Kikuchi et al. 1995; Wangemann et al. 1995).

Figure 1.

Schematic diagram of cochlear cross-section and K+ recycling pathway

The figure depicts a cross-section of the mammalian cochlea (adapted from Kikuchi et al. 1995) and magnified schematic views of the stria vascularis (upper box) and part of the organ of Corti and its support structures (lower box). Upper box: the unusual ionic composition of the endolymph that fills the scala media is created by the stria vascularis, a structure composed of three cell types: marginal, intermediate and basal cells. Marginal cells utilize basolateral Na+-K+-ATPase to pump K+ into the cytoplasm and to create a Na+ gradient across the basolateral membrane. This Na+ gradient drives K+ and Cl into the cell via the Na+-K+-Cl co-transporter NKCC1. Chloride leaves the marginal cell through basolateral channels composed of ClC-Ka/barttin and ClC-Kb/barttin heteromeres. Potassium diffusion into the endolymph happens through apical K+ channels made of KCNQ1 and KCNE1 and driven by the difference between the marginal cell's positive resting potential of ≈95 mV and the endolymphatic potential of ≈85 mV. Potassium in the extracellular space basolateral of marginal cells (also called interstrial space) is replenished from intermediate cells. Gap junctions connect intermediate cells with basal cells and type 1 fibrocytes. The ion channel that releases K+ from intermediate cells is encoded by the KCNJ10 gene (Ando & Takeuchi, 1999; Marcus et al. 2002). Lower box: endolymphatic K+, driven by its high electrochemical gradient, enters hair cells through mechanosensitive ion channels located near the tips of the hair cells' stereocilia. Potassium is rapidly exhausted through the hair cells' basolateral membranes. The KCNQ4 gene encodes a K+ channel that is essential for this extrusion from outer hair cells. Potassium also leaves inner and outer hair cells through other voltage- and Ca2+-regulated channels. Deiters cells take up K+, possibly via inward rectifying channels (Hibino et al. 1997), and convey it through a network of gap junctions to root cells where it is released. Type 2 fibrocytes embrace root cells and take up extracellular K+, probably via interplay of Na+-K+-ATPases and Na+-K+-Cl co-transporters. The circuit is closed because type 2 fibrocytes form gap junctions with type 1 fibrocytes. The mechanisms that mediate K+ uptake into Deiters cells and that release K+ from root cells are not identified.

Due to this proposed role of gap junctions for K+ homeostasis, connexin genes are prime candidates for hearing disorders. Indeed, one of the first loci linked to human non-syndromic deafness was identified as the GJB2 gene, which encodes the gap junction protein connexin 26 (Kelsell et al. 1997). Knowledge of the first deafness gene paired with physiological and morphological considerations (see Kikuchi et al. 1995) qualified other members of the connexin family of gap junction proteins as bona fide candidates for causing hearing disorders when mutated. Not surprisingly, comparison of the chromosomal location of other connexin genes with the genomic intervals linked to non-syndromic deafness disorders led to the identification of deafness-causing mutations in three additional connexin genes (Xia et al. 1998; Grifa et al. 1999; Liu et al. 2001; Table 1).

Additional genes that do not affect gap junctions directly but disturb the fibrocytes interconnected by these junctions also cause inner ear defects. Mice that are deficient for the transcription factor Brn-4, which is also mutated in a human hearing disorder (de Kok et al. 1995), have ultrastructurally abnormal fibrocytes and reduced endocochlear potential (Minowa et al. 1999). Brn-4 is likely to be important for proper development of inner ear mesenchyme, which gives rise to the fibrocytes of the spiral ligament (Phippard et al. 1998, 1999).

Recycled K+, conveyed from the organ of Corti through type II and type I fibrocytes of the spiral ligament, enters the stria vascularis through gap junctions between type I fibrocytes and basal cells (Fig. 1). Basal cells form gap junctions with intermediate cells and K+ is released from intermediate cells into the extracellular space between intermediate cells and marginal cells, where it is actively taken up by marginal cells, which then release K+ into the scala media (Fig. 1, upper box).

As might be expected, perturbation of this portion of K+ homeostasis also results in hearing disorders. For example, pharmacological evidence suggested that the KCNE1 (ISK) gene encoding a K+ channel β-subunit, which is expressed by marginal cells of the stria vascularis, is essential for K+ secretion into the endolymph (Sakagami et al. 1991; Shen et al. 1995; Shen & Marcus, 1998). The importance of the KCNE1 gene for inner ear function was subsequently tested in vivo by generation of KCNE1-deficient mice. Strial marginal cells of KCNE1-knockout mice are highly impaired in K+ secretion causing a failure of cochlear potassium homeostasis (Vetter et al. 1996).

The pathological inner ear defects of KCNE1-deficient mice are strikingly similar to those of patients diagnosed with Jervell and Lange-Nielsen syndrome (Friedmann et al. 1966, 1968). This observation, coupled with the finding that KCNE1 co-assembles with the KCNQ1 (KvLQT1) α-subunit, indicated both human genes as candidates for causing this cardioauditory syndrome (Barhanin et al. 1996; Sanguinetti et al. 1996).

Linkage analysis of patients with Jervell and Lange-Nielsen syndrome revealed that this genetically heterogeneous disorder is indeed caused by mutations in either KCNQ1 or KCNE1 (Neyroud et al. 1997; Schulze-Bahr et al. 1997; Casimiro et al. 2001). It is conceivable that KCNQ1 and KCNE1 form the channel that allows secretion of K+ from marginal cells into the scala media (Fig. 1, upper box). Another member of the KCNQ family of potassium channels – KCNQ4 – was subsequently identified as causing human non-syndromic deafness (Kubisch et al. 1999). KCNQ4 is expressed by outer hair cells of the cochlea and in auditory nuclei of the brainstem and is probably involved in basolateral K+ secretion of outer hair cells (Kharkovets et al. 2000).

Gap junction proteins and K+ channels are two remarkable examples of how genomic research and physiology have interplayed in elucidating inner ear function. After a single member of the connexin and the KCNQ gene family had been linked to deafness, other members of these families became prime candidates for causing human deafness and utilization of genomic databases rapidly led to discovery of additional connexins and KCNQ genes that, when mutated, cause hearing disorders (Table 1). Even more compelling is the amalgamation of morphology, physiology and genomics that led to in-depth understanding of how K+ is actively driven through strial marginal cells. Physiological evidence and expression pattern analysis of genes encoding ion pumps, co-transporters and channels revealed a likely mechanism for K+ secretion in which a basolateral Na+-K+-ATPase in marginal cells creates a Na+ gradient that drives the co-transport of Na+, K+ and Cl into the cell (Fig. 1, upper box). Potassium enters the endolymph through KCNQ1/KCNE1 channels and Cl leaves marginal cells basolaterally. Supporting this scheme, Na+-K+-ATPase and the Na+-K+-Cl co-transporter NKCC1 are both localized in the basolateral membrane of marginal cells (Schulte & Adams, 1989; Crouch et al. 1997). Elegant use of EST databases and partial genomic information led to the identification of mutations in the NKCC1 co-transporter gene Slc12a2 as the cause of deafness in the spontaneous occurring recessive mouse mutation no syndactylism (syns) (Dixon et al. 1999). The phenotype of this mutation is characterized by decreased production of endolymph. Parallel knock-out studies of the same gene resulted in identical findings (Delpire et al. 1999).

Potassium leaves marginal cells through apical KCNQ1/ KCNE1 channels, whereas co-transported Cl is extruded through basolateral channels. The kidney Cl channel ClC-Ka is expressed by marginal cells (Ando & Takeuchi, 2000) and was considered, but then excluded, as a candidate causing a form of Bartter's syndrome, a renal disorder associated with congenital deafness (Landau et al. 1995; Brennan et al. 1998). Ultimately, mutations in a gene encoding the Cl channel β-subunit barttin were identified as the cause for this disorder (Birkenhager et al. 2001; Estevez et al. 2001). Barttin forms heteromeric Cl channels with either α-subunit ClC-Ka or ClC-Kb. The three channel subunits are expressed in marginal cells of the stria vascularis (Estevez et al. 2001). Although the disruption of the murine gene encoding ClK-Ka does not result in deafness (Matsumura et al. 1999), perhaps a double knock-out of the genes encoding both Cl channel α-subunits would result in deafness, if homozygous animals are viable.

Discovery of the genes involved in cochlear K+ recycling is one of the best examples of how physiological genomics has been a central component of inner ear research. The majority of genes that have been linked to deafness were identified because a suitable candidate gene was defined, often based on considering physiological relevance. The major tools for validating candidate genes are the mammalian genomics resources: genomic databases that are available in the public domain ( and commercially (Celera Discovery System at Identification of deafness genes will be even more expedited in the future because these resource services are continuously improving and updating their sequence databases.


I thank Drs J. L. Cyr, H. Hibino and A. J. Hudspeth, and the members of the Eaton-Peabody Laboratory, for insightful comments on the manuscript and for inspiring discussions. S. H. is a Basil O'Connor Scholar.