Usher syndrome type IIa (USH2A) combines moderate to severe congenital hearing impairment and retinitis pigmentosa. It is the most common genetic form of USH. USH2A encodes usherin, which was previously defined as a basement membrane protein. A much larger USH2A transcript predicted to encode a transmembrane (TM) isoform was recently reported. Here, we address the role of TM usherin in the inner ear. Analysis of the usherin alternative transcripts in the murine inner ear revealed the existence of several predicted TM usherin isoforms with modular ectodomains of different lengths. In addition, we identified in the usherin cytoplasmic region a predicted 24 amino acid peptide, derived from a newly defined exon that is predominantly expressed in the inner ear but not in the retina. In mouse and rat inner ears, we show that TM usherin is present at the base of the differentiating stereocilia, which make up the mechanosensitive hair bundles receptive to sound. The usherin immunolabeling is transient in the hair bundles of cochlear hair cells (HCs), but persists in mature hair bundles of vestibular HCs. Several lines of evidence support the involvement of TM usherin in the composition of the ankle links, a subset of filamentous lateral links connecting stereocilia at the base. By co-immunoprecipitation and in vitro binding assays, we establish that the usherin cytodomain can bind to whirlin and harmonin, two PDZ domain-containing proteins that are defective in genetic forms of isolated deafness and USH type I, respectively. These PDZ proteins are suitable to provide the anchoring of interstereocilia lateral links to the F-actin core of stereocilia. Our results strongly suggest that congenital deafness in USH type I and type II shares similar pathogenic mechanisms, i.e. the disruption of hair bundle links-mediated adhesion forces that are essential for the proper organization of growing hair bundles.
Usher syndrome (USH) is the most frequent cause of hereditary deaf–blindness in humans. Three clinical subtypes are distinguished on the basis of differences in the severity of hearing impairment and the presence of vestibular dysfunction, whereas progressive visual loss due to retinitis pigmentosa with variable age of onset occurs in all three USH types. USH type II (USH2) is characterized by moderate congenital hearing loss and normal vestibular function ( 1 ). Three different USH2 loci, USH2A–C, have been defined by linkage analysis of affected families ( 2 – 4 ). Mutations in the gene encoding usherin underlie USH2A, the most common genetic form of USH ( 5 , 6 ). Four main types of domains compose the originally identified USH2A protein, hereafter referred to as extracellular (EC) usherin: (i) an N-terminal domain with homology to the laminin globular domain present in thrombospondin (TSPN-LG) ( 7 ), (ii) a common N-terminal globular feature of laminins and laminin-related proteins called the LN module ( 8 ), (iii) a series of 10 rod-like laminin EGF-like (LE) modules ( 9 ) and (iv) four fibronectin type III (FnIII) repeats (Fig. 1 A). Another USH2A transcript, predicted to encode a much larger usherin isoform (Fig. 1 A), was subsequently reported, and mutations in exons, which are not included in the short transcript, were shown to underlie USH2A ( 6 ). The long putative protein, hereafter referred to as transmembrane (TM) usherin, harbours, in addition to the previously described functional domains, two more laminin globular-like (LG) domains, 28 additional FnIII repeats, as well as a TM region followed by an intracellular domain with a class I consensus PDZ (Postsynaptic density 95, Discs large, Zonula occludens-1) domain-binding motif, THL, at its C-terminal end ( 6 ). Semi-quantitative differential RT-PCR analysis showed that, although the short transcript is present in various fetal and adult tissues, with a notably strong expression in the neural retina, the long transcript(s) is/are found in fewer tissues but also are abundant in the neural retina ( 6 ). By immunohistochemistry analysis, using an antibody generated against the usherin LN domain, the protein was localized in various basement membranes, including the retina and the cochlea ( 10 , 11 ). An LE usherin–collagen IV interaction was demonstrated and suggested to stabilize the integration of EC usherin in the basement membrane ( 12 ).
Sensory cells of the inner ear, HCs, convert mechanical stimuli, i.e. sound and acceleration caused by head movements, into an electrical signal, manifested by variations of the cell membrane potential ( 13 ). The HC apical surface is crowned by the hair bundle, a mechanoreceptive structure comprised of 30–300 rigid microvilli called stereocilia, filled with a core of unidirectionally oriented actin filaments and one microtubule-based cilium, the kinocilium, which is transient in mammalian cochlear HCs ( 14 ) (Fig. 2 ). Mature hair bundles consist of uniformly oriented rows of stereocilia organized in a regular staircase pattern ( 15 ). Each stereocilium tip is connected to the side of the adjacent taller stereocilium by a single oblique link called the tip link ( 16 ) (Fig. 2 ), which plays a key role in the mechanoelectrical transduction (MET) process. Changes in the tip link tension, caused by mechanically induced deflection of the hair bundle, are thought to modulate the opening probability of associated MET channels, which results in HC membrane potential variations ( 13 ). In addition, at least two different types of fibrous lateral links, namely horizontal top connectors and shaft connectors, interconnect stereocilia to one another within and across rows (Fig. 2 ). In the differentiating hair bundle, numerous transient lateral links interconnect growing stereocilia together (stereociliary links) and some tall stereocilia to the adjacent kinocilium (kinociliary links), from early developmental stages onward ( 17 ). Several lines of evidence have implicated two cadherins with long ectodomains, cadherin 23 and protocadherin 15, underlying two genetic forms of USH1, as components of these links that seem to be necessary for the proper formation and cohesion of the growing hair bundle ( 18 – 21 ). Cadherin 23 has also been proposed to be a component of the tip links ( 22 , 23 ).
Previous studies on usherin ( 10 – 12 ) have focussed on the short isoform. In this study, we address the role of TM usherin in the inner ear. We show that the protein is likely to be involved in the formation of a subset of interstereocilia links called the ankle links, and shed lights on the mechanisms leading to hearing impairment in USH2A syndrome.
RESULTS AND DISCUSSION
Characterization of Ush2a long transcripts in the mouse inner ear
On the basis of the sequence of GenBank ESTs and gene predictions, 11 pairs of primers (Supplementary Material, Table S1) were designed in order to amplify overlapping fragments of the long Ush2a transcript. cDNA fragments that derived from 73 exons were amplified by RT–PCR from P6 mouse inner ear mRNA and sequenced. The longest transcript assembled from these overlapping fragments is over 15.8 kb long. It is predicted to encode a 5213 amino acid protein, which shares 70% identity and 81% similarity with the human TM usherin. Five different regions of alternative splicing were detected along the long Ush2a transcript (Fig. 1 A). Only one of these, which encodes the N-terminal part of EC and TM usherins, is fully included in the two Ush2a transcripts. The corresponding transcript(s) is/are predicted to encode isoform(s) that start with an ATG initiation codon within the context of a Kozak consensus sequence at amino acid position 421 of the complete transcripts, i.e. within the LN module of usherin isoforms; thus, the first defined modules along its sequence are the 10 LE repeats. The second alternatively spliced region is predicted to encode one FnIII repeat and the first LG domain, which is present only in TM usherin. The three other alternatively spliced regions are predicted to encode three to six FnIII repeats (Fig. 1 A). RT–PCR analysis of various mouse tissues suggests that the expression of long Ush2a transcripts is more abundant in the adult retina than in the adult inner ear or testis. TM usherin could not be amplified from lung mRNA (data not shown). Another region of alternative splicing is exon 71, which is predicted to encode a 24 amino acid peptide of the usherin cytoplasmic fragment. This exon was not included in the previously defined long USH2A transcripts ( 6 ). Sequence analysis revealed the presence of this exon in human, rat, cow, dog and zebrafish genomes (Fig. 1 B). RT–PCR analysis of mouse inner ear, retina and testis tissues using primers designed to amplify mRNA sequences encompassing the full cytoplasmic region of TM usherin (Fig. 1 C), followed by cloning and sequencing of the RT–PCR products, confirmed that exon 71 is highly expressed in the inner ear. Among the 50 Ush2a cDNA clones derived from P6 inner ear, 49 included the exon 71-transcribed sequence. In contrast, this sequence was not included in any of the 16 Ush2a cDNA clones amplified from the retina and was found in only one out of seven cDNA clones from the testis (Fig. 1 C). To investigate whether cytosolic isoforms of usherin may exist, rapid amplification of cDNA ends (RACE) analysis was carried out on P6 mouse inner ear mRNA (see Materials and Methods). Sequence analysis showed that all the 37 obtained independent cDNA clones contained regions encoding the usherin TM domain. Therefore, there is no evidence of the existence of cytosolic usherin isoforms in the P6 inner ear. Consistent with RT–PCR analysis results, all these RACE clones contained exon 71.
Thus, Ush2a gene expression in the murine inner ear predicts several TM usherin isoforms with ectodomains of different sizes (Fig. 1 A). In addition, a newly identified alternatively spliced exon which encodes a 24 amino acid peptide of the usherin cytoplasmic domain is conserved throughout vertebrate evolution and was found to be expressed in almost all the inner ear long transcripts analyzed, but not in the retinal ones. The modular ectodomain of usherin is composed of FnIII repeats and various laminin-like domains (TSPN-LG, LN, LE, LG). After an N-terminal thrombospondin-type laminin G domain, the usherin ectodomain continues with an LN and 10 LE domains; both module types are detected in all laminin chains and in netrins. Then, following the LN and LE modules are two series of FnIII repeats separated by two LG domains, which are also present at the C-terminus of α-laminin chains, agrin and perlecan. The structure and function of LE and LG domains have been extensively studied. LE domains are 60 amino acid modules with high mechanical stability. Indeed, they contain eight conserved cysteine residues involved in four disulfide bonds that prevent domain extension under mechanical strength ( 9 , 24 ). LG modules are about 200 amino acids in length and are formed by two anti-parallel seven-stranded β-sheets ( 25 ). They possess a single disulfide bond at their C-terminal end. Three-dimensional structures of LG domains outline a common multiligand-binding site that, in some LG domains, contains a Ca 2+ -binding site involved in Ca 2+ -dependent interactions ( 26 ). It is unclear whether any of the usherin LG domains contains a Ca 2+ -binding site because residues equivalent to Asp2982 in the LG5 domain of laminin α2 are present in the two LG domains of usherin, but the equivalent residue to laminin α2 LG5 Asp3055 is absent from usherin LG1 and uncertain in LG2. A striking feature of TM usherin is the predicted high number of FnIII repeats, between 22 and 33 depending on the composition of alternatively spliced regions, none of which contains the RGD consensus sequence involved in the binding of EC matrix proteins to integrins ( 27 ). In this study, these repeats are split into two clusters and, as a whole, would represent up to two-thirds of the ectodomain size ( 6 ). FnIII modules are approximately 90 amino acid long domains with no disulfide bond, which form seven β-pleated sheets ( 28 ). They are present in ∼2% of mammalian proteins, mainly EC matrix components, but also TM proteins and intracellular proteins ( 29 ). FnIII repeats are present in proteins that experience mechanical stress in vivo , such as the EC matrix proteins, fibronectin ( 30 , 31 ) and tenascin ( 32 , 33 ). These domains are considered as reversible ‘shock absorbers’ ( 34 ). They can unfold under mechanical stress, fold again when tension is released and also exhibit partially folded conformations. A mechanical hierarchy was defined among the 15 FnIII repeats of fibronectin, which can be ranked according to their relative mechanical stability. However, it has also been shown that the FnIII hierarchy of mechanical unfolding can be changed by environmental conditions (such as pH) or by forming complexes with other molecules (e.g. heparin binding) ( 35 , 36 ). Alternative usherin isoforms containing variable numbers of FnIII repeats would therefore allow for different compliances of the ectodomains, which would increase with the number of FnIII modules.
TM usherin localization in the mouse inner ear
We first checked whether long Ush2a transcripts are expressed in the inner ear sensory cells, using single cell RT–PCR experiments performed on HCs from P6 mouse cochleas. Using primers specific for the cytoplasmic region of long Ush2a transcripts, the expected Ush2a product were successfully amplified (data not shown).
In the cochlea, HC differentiation proceeds from the base to the apex. Stereocilia sprout from the apical surface of cochlear HCs at E15 in the mouse and by P4–P6, the hair bundles have reached their final length ( 37 , 38 ). In the vestibular organs, hair bundles start to grow 2 days earlier than in the cochlea. We studied the distribution of usherin during the period of hair bundle differentiation, in the mouse inner ear, by immunostaining. Taking into account the expression of several TM isoforms with various ectodomains and the absence of cytosolic forms in the inner ear, two antibodies, U2aCyt1 and U2aCyt2, were produced against the entire (164 amino acids) usherin cytoplasmic region (see Materials and Methods) in order to detect TM isoforms. The two antibodies recognized the myc-tagged usherin cytodomain produced by transfected HeLa or COS7 cells (data not shown). In the mouse inner ear, similar results were obtained with either antibody. Usherin was detected in the differentiating HCs (Fig. 3 ). Detailed confocal microscopy analysis revealed intense usherin immunoreactivity in the growing stereocilia. At E18 (the earliest stage analyzed), usherin is detected all along the stereocilia of outer HCs (OHCs), whereas no labeling is found in the stereocilia of inner HCs (IHCs) (Fig. 3 A). At E20, the OHC labeling becomes restricted to the base of the stereocilia and also appears at the base of IHC stereocilia (data not shown). Intense usherin labeling at the base of growing stereocilia persists between P0 and P10, when it starts to fade out, first from the IHCs and later from the OHCs (Fig. 3 B and C). Similar results were obtained in rat cochlear HCs (data not shown). Usherin was also detected in the soma of HCs and supporting cells. This labeling became more significant by P15, at the time when usherin is no longer detected in the HCs' stereocilia (Fig. 3 D). The usherin staining was more intense in the apical region of the HCs, corresponding to the junctions with adjacent supporting cells (Fig. 3 D). In the vestibular HCs, usherin was detected at the base of the stereocilia at all stages analyzed, persisting also at P15 (Fig. 3 E). Therefore, TM usherin labeling at the base of hair bundles' stereocilia is transient in the auditory HCs, whereas it persists in mature vestibular HCs.
Ankle links are a subset of lateral links that connect two adjacent stereocilia near their basal ends. Because the gap between adjacent stereocilia is larger at their proximal ends than at their distal ends, ankle links are the longest among the various lateral links and their length exceeds 150 nm. The developmental dynamics of these links has been recently described in the mouse cochlea by electron microscopy studies ( 17 ) (Fig. 3 F). Immature-looking ankle links first appear at the base of the stereocilia around P0. By P2, they become more distinct and are detected in the hair bundles of OHCs and IHCs from both the basal and apical coils of the cochlea. These persist until approximately P9, when they start to disappear first from IHC ( 17 ) (Fig. 3 F). In contrast, ankle links persist in the mature vestibular HCs. Therefore, the spatio-temporal profile of TM usherin is quite similar to that of the ankle links. Moreover, on the basis of the three-dimensional structures of FnIII ( 28 ), LE ( 39 ) and LG ( 40 ) modules, which extend 4, 2.5 and 3.5 nm, respectively, and assuming a 4 nm length for the LN domain as well as a linear arrangement of all domains, the lengths of the TM usherin ectodomains could be between 125 nm (22 FnIII modules) and 170 nm (33 FnIII modules). Dimers that could be formed by two usherin molecules would thus be long enough to connect adjacent differentiating stereocilia. We conclude that TM usherin is likely to be involved in the composition of these links that connect the differentiating stereocilia at the base. We had previously proposed that vezatin, a putative TM protein of adherens junctions which is also present at the base of the growing stereocilia ( 41 ), was tightly associated to the ankle links. However, vezatin cannot by itself account for these interstereocilia links because further extensive transcript analysis did not yield any indication of an ectodomain exceeding 200 amino acids (unpublished data) ( 42 ). Ultrastructural studies have described the ankle links as single-stranded links with a central density zone ( 17 ), but the type of the molecular interactions that make these links is still conjectural. One possibility is trans homophilic or heterophilic interaction between TM proteins. Notably, very large G protein-coupled receptor-1 (VLGR1), the putative TM protein defective in USH2C ( 43 ), also has a long ectodomain, possibly extending 180 nm in length. Hair bundle anomalies that have just been reported in the USH2C mouse model ( 44 ) strongly suggest that the protein is present in the differentiating stereocilia. A heterophilic interaction between usherin and VLGR1 ectodomains cannot be excluded. In addition, considering that TM usherin and VLGR1 both contain LG domains, which from an evolutionary perspective are the most closely related modules, the two proteins may interact with common binding partners. Finally, an EC protein could bridge two TM usherin molecules from adjacent stereocilia. The EC matrix protein nidogen is a possible candidate because it binds to LE domains with high affinity and also to FnIII domains ( 45 ).
The spatio-temporal distribution of a chick ankle link antigen (ALA) has been studied by immunolabeling in the chick ( 46 ). ALA is evenly distributed over most of the hair bundle surface at early differentiation stages and progressively becomes concentrated in a narrow zone around the base of each bundle. Such an expression profile is similar to that of TM usherin in the mouse OHCs. Irrespective of whether the ALA antigen is derived from the usherin ortholog in the chick, which is still unknown, the diffuse TM usherin immunoreactivity in the hair bundles of OHC at early differentiation stages raises the possibility that the protein is involved also in the formation of early transient lateral links.
Usherin cytodomain directly interacts with whirlin and harmonin b
The class I PDZ-binding consensus motif that is present at the C-terminal end of the usherin cytodomain suggests that usherin might interact with PDZ domain-containing proteins. Such proteins are known as organizers of molecular complexes, and some of them are involved in anchoring TM proteins to the underlying cytoskeleton ( 47 ). Two PDZ domain-containing proteins, whirlin and harmonin, have been identified in the differentiating hair bundle. Mutations in the genes encoding these proteins lead to severe hearing impairment in humans and mice ( 48 – 51 ).
Whirlin is present at the tip of the growing stereocilia ( 52 – 54 ), where the protein is required for stereocilia elongation ( 55 ). Transient whirlin labeling has also been detected at the base of growing stereocilia, in a time window similar to that of the ankle links and TM usherin ( 52 ). Moreover, the anchoring of ankle links to the stereocilia actin core may involve the whirlin–myosin VIIa direct interaction ( 52 ). We therefore tested the possibility of a direct interaction between usherin and whirlin. Alternative transcription start sites result in the expression of two groups of whirlin isoforms. Long whirlin forms contain a proline-rich (PR) domain and three PDZ domains, whereas the short C-terminal forms contain only the PR and the third PDZ domain (Fig. 4 A). In co-transfected HeLa cells producing a myc-tagged usherin cytodomain (164 C-terminal amino acids) and the long whirlin isoform, the two proteins entirely co-localized throughout the cytoplasm (data not shown). To determine whether the usherin cytodomain could recruit whirlin to the cell membrane, we analyzed whirlin distribution in the presence of a human E-cadherin (hEcad)–usherin chimeric protein, composed of the EC and TM domains of the hEcad directly fused to the cytodomain of usherin (Fig. 4 A). In transfected HeLa cells producing long whirlin isoforms alone, the whirlin labeling was diffuse in the cytoplasm and was absent from the cell membrane including cell–cell contacts (Fig. 4 B). In co-transfected HeLa cells producing both hEcad–usherin and long whirlin, hEcad–usherin was detected at the cell membrane as expected, with a more intense staining at regions of cell–cell contacts, where it recruited the long whirlin isoform (Fig. 4 C), thus suggesting the existence of a molecular interaction between whirlin and the usherin cytodomain. No such recruitment was observed in co-transfected HeLa cells producing hEcad–usherin and a short whirlin isoform (Fig. 4 D). The usherin–whirlin interaction was shown by a co-immunoprecipitation assay. HEK293 cells were co-transfected with plasmids encoding the whirlin long isoform and a myc-tagged usherin cytodomain. Incubation of the cell extracts with an anti-whirlin antibody yielded co-immunoprecipitation of the two proteins (Fig. 4 E). Usherin–whirlin direct interaction was confirmed by in vitro binding assays. The in vitro translated usherin cytodomain bound to the immobilized glutathione S -transferase (GST)-tagged long whirlin isoform. In the reverse experiment, the in vitro translated long whirlin isoform also bound to a GST–fusion protein including the usherin cytodomain (GST–cytoUsherin). In contrast, binding was not observed when the short whirlin isoform was incubated with GST–cytoUsherin or when the long whirlin isoform was incubated with a GST-tagged truncated usherin cytodomain (GST–cytoUsherinΔ5C′ter) that lacks the five C-terminal amino acids including the PDZ domain-binding motif (Fig. 4 F). Together, these results establish that usherin, through its C-terminal PDZ domain-binding motif, binds to the first and second PDZ domains of whirlin.
The other known hair bundle PDZ protein is the USH1C gene product, harmonin. Interestingly, the three PDZ domains of harmonin b (the longest of three isoform classes) share the highest degree of sequence similarity with those of whirlin. Of the three alternative groups of harmonin transcripts (Fig. 5 A), only class b isoforms are predominantly expressed in the inner ear ( 49 ), where they are present in the differentiating hair bundle ( 18 ). Unlike whirlin, harmonin b has so far not been detected in the basal part of the hair bundles beyond the earliest stage of hair bundle growth. However, immunolabeling analysis on P15 mouse cochlea as performed in this study revealed the presence of harmonin b also at the apical HC junctions with surrounding supporting cells (Fig. 5 B), in the same location as TM usherin (Fig. 3 D). This prompted us to test the possibility that the two proteins might interact. In co-transfected HeLa cells producing the usherin cytodomain and harmonin b, the usherin cytodomain was entirely co-localized with harmonin b and actin. Indeed, we had previously shown that harmonin b directly interacts with F-actin ( 18 ). Moreover, the presence of harmonin b profoundly modified the localization of the usherin cytodomain in transfected HeLa cells (Fig. 5 C–E). Similar co-transfection experiments with harmonin a also showed full co-localization between the usherin cytodomain and harmonin a (data not shown). By co-immunoprecipitation assays, we confirmed that usherin interacts with harmonin. HEK293 cells were co-transfected with plasmids encoding green fluorescent protein (GFP)-tagged harmonin a and myc-tagged usherin cytodomain. Incubation of the cell extracts with an anti-GFP antibody yielded co-immunoprecipitation of the two proteins (Fig. 5 F). Usherin–harmonin direct interaction was confirmed by in vitro binding assays. The in vitro translated usherin cytodomain bound to immobilized GST-full length harmonin a, and in vitro translated harmonin a bound to immobilized GST–cytoUsherin (Fig. 5 G). Binding was not detected between an in vitro translated PDZ domain-containing fragment of Apxl (NP_766029), another PDZ protein also present in HCs, and the GST-tagged cytoUsherin (data not shown). Detailed dissection of the harmonin–usherin interaction showed that harmonin binds to usherin through its first PDZ domain (PDZ1), whereas binding was not detected with any of the separated PDZ2, PDZ3 or CC2-PST domains. In addition, binding was not detected when 35 S-labeled harmonin a was incubated with the GST–cytoUsherinΔ5C′ter construct (Fig. 5 G). We conclude that, through its C-terminal PDZ domain-binding motif, usherin binds to the first PDZ domain of harmonin.
The proteins encoded by the five known USH1 genes are all present in the inner ear sensory cells, and multiple molecular interactions have been shown between these proteins ( 56 ) (Fig. 6 ). In particular, harmonin (USH1C) directly interacts with the four other USH1 proteins, namely myosin VIIa (USH1B), cadherin 23 (USH1D), protocadherin 15 (USH1F) and Sans (USH1G). Moreover, harmonin's first PDZ domain appears to play a key role in this interaction network ( 56 ). As harmonin was also found to bind to usherin, we tested whether usherin can directly interact with other USH1 proteins. Expression of each of these proteins together with the usherin cytodomain in HeLa cells did not yield any co-localization within co-transfected cells. In addition, in vitro binding assays excluded potential interaction between the usherin cytodomain and any of these USH1 proteins (data not shown). Figure 6 A summarizes all the known hair bundle protein interactions.
It is worth noting that there is an overlap in the distributions of harmonin b and TM usherin along the stereocilia of OHCs during early developmental stages, up to P0 in the mouse cochlea. Assuming that usherin is a component of early interstereocilia lateral links too, in this study, harmonin b could anchor these links to the stereocilia actin core via a TM usherin–harmonin b–F actin sequence of direct interactions ( 18 ). Thereafter, harmonin b becomes concentrated at the apex of the stereocilia, where it is likely to play a similar role on the co-localized cadherin 23-links ( 18 , 20 , 21 ).
Pathophysiology of USH
Disorganized hair bundles characterize all USH1 mouse models (reviewed in 57), as well as the USH2C mouse model lacking functional Vlgr1 ( 44 ). In addition, cadherin 23 (USH1D) has been shown to form transient interstereocilia links ( 20 , 21 ), and protocadherin 15 (USH1F), which is present along the stereocilia ( 19 ), is also qualified to form interstereocilia links. As mentioned earlier, the hair bundle anomalies observed in the USH2C mouse ( 44 ) suggest that this integral membrane protein also may form interstereocilia links. By immunolabeling, Vlgr1 showed a spatio-temporal pattern of expression in the hair bundle similar to that of TM usherin. Vlgr1 was observed at the base of stereocilia in cochlear and vestibular HCs in P2–P5 rats, and at adult stages, it persisted at the base of stereocilia in vestibular HCs while disappearing from the cochlear hair bundles ( 58 ) (unpublished data). It is worthy to note that a possible direct interaction of Vlgr1, which possesses a C-terminal PDZ-binding motif, with harmonin and whirlin has also been suggested ( 58 , 59 ). Although an USH2A animal model is not available yet, our results thus raise the interesting possibility that similar pathogenic mechanisms, i.e. the disruption of hair bundle links-mediated adhesion forces, are involved in congenital deafness of USH1 and USH2 (Fig. 6 ). Deafness, however, is more severe in USH1 than in USH2. Several tentative explanations may account for the difference. First, there could be some degree of functional redundancy between USH2 proteins in the developing cochlea. Secondly, ankle links may be less critical than cadherin 23- and protocadherin 15-based links for the cohesion of the growing hair bundle. Thirdly, the developmental role of TM usherin and Vlgr1 could be more important in the OHCs, which act as cochlear amplifiers, than in the IHCs, the genuine sensory cells. Finally, the USH1 proteins could play additional roles in the HCs; for instance, in the mechanotransduction machinery ( 22 , 23 ).
MATERIALS AND METHODS
Amplification and sequencing of the long Ush2a transcripts
The mRNA was isolated as described by Chirgwin et al . ( 60 ) and treated with DNase I (Invitrogen). Synthesis of cDNA was performed using random hexanucleotides as primers, as described by Luijendijk et al . ( 61 ). After the second PCR, products were separated by electrophoresis on an agarose gel. Specific bands were purified from the gel using the QIAquick Gel Extraction kit (Qiagen) and sequenced by the ABI PRISM BigDye Terminator Cycle Sequencing version 2.0 kit (Applied Biosystems) and the ABI PRISM 3730 DNA analyzer (Applied Biosystems).
RACE and RT–PCR
RACE was performed with the BD-Smart RACE cDNA Amplification kit (BD-Clontech) on P2–P6 vestibular polyA + RNA using as reverse primer 5′-CTTCCGTAACAACCTTCTTGTCTGCCATGTC-3′. RT–PCR was performed on P15 cochlea, retina and testis total RNA using as primers A: 5′-GTATCAGAGAGCGACCTCCCTTGG-3′ and B: 5′-TCAGAGGTGGGTGTCGGTAAAGG-3′ derived from exons 70 and 73, respectively.
Single cell RT–PCR
Single cell RT–PCR experiments were carried out as described ( 52 ). PCR for the detection of usherin expression was initially carried out with usherin sense (5′-CCACTG AGCGTCTACCCAC-3′) and antisense (5′-TTACCGACACCCACCTCTGA-3′) primers derived from exons 70 and 73, respectively. A second, nested PCR was carried out with usherin sense (5′-GACACCTATGAGTATTCGGAG-3′) and antisense (5′-CATAATAGTTTCCCACAGTGAG-3′) primers derived from exons 72 and 73, respectively.
Antibodies U2aCyt1 and U2aCyt2 were produced against the entire cytoplasmic region of inner ear usherin isoforms (164 amino acids including the 24 amino acids fragment encoded by exon 71), which was expressed using the FLAG-ATS system (Sigma, St Louis, MO, USA). Both antibodies were raised in rabbits and the reactive immunoglobulin was affinity-purified using the immunogen. The specificity of the affinity-purified antibodies was confirmed by immunofluorescence analysis on transfected COS7 and HeLa cells producing the myc-tagged usherin cytodomain. The data presented in this manuscript derive from the use of both antibodies, which gave similar results in all experiments. Antibodies against whirlin and harmonin b have been previously characterized ( 18 , 52 ).
Cochlear dissections and staining
Mouse inner ears were fixed and treated for immunofluorescence as described ( 62 , 63 ). For whole-mount preparations of the organ of Corti, inner ears were fixed and decalcified, then half turns of the cochlea were carefully dissected to separate the organ of Corti and immediate surrounding tissues. Whole organs of Corti were then used for indirect immunofluorescence ( 62 ). Stained whole-mounted preparations were analyzed on a laser scanning confocal microscope (LSM-510META, Zeiss).
The cytoplasmic region of the mouse usherin (cytoUsherin) was defined according to the previously published human USH2A transcript b (NM_206933). A 495 bp fragment was RT–PCR amplified from mouse inner ear mRNA and cloned into a pCMVtag3B vector (Myc tag, Stratagene) for expression in HeLa and COS7 cells and into a pGex-4T1 vector (GST tag, Amersham) for protein production. Another cytoplasmic Ush2a cDNA fragment, in which the last 15 bp (encoding the C-ter PDZ binding motif) was deleted, was also amplified and cloned into pGex-4T1. Full and partial harmonin a, harmonin b, long and short whirlin were produced as described ( 18 , 52 , 64 )
Cell lines and immunofluorescence analysis
HeLa and COS7 cell lines were cultivated in Dulbecco's modified Eagle medium supplemented with 10% fetal calf serum. Cells were collected 2 days after transfection by Effectene Transfection Reagent (Qiagen) and processed for immunocytofluorescence as described ( 41 ). Briefly, after paraformaldehyde fixation, cells were incubated for 15 min with 50 m m NH 4 Cl in phosphate-buffered saline (PBS) and then washed in 0.01% saponin in PBS. The cells were incubated for 1 h in 10% goat serum in PBS and then with the anti-myc and/or anti-protein antibodies for 1 h at room temperature, followed by the secondary antibody (1 h at room temperature). The mouse monoclonal anti-myc antibody used in our experiments is derived from clone 9E10 (Santa Cruz). Rhodamine-phalloidin (Sigma) staining was used to visualize actin filaments.
To verify the interactions between either the long whirlin isoform or the GFP–harmonin a and the intracellular region of usherin, co-transfected HEK293 cells were lyzed and immunoprecipitated with anti-whirlin (CIP98; a gift from Y. Yamasaki, RIKIN, Japan) or anti-GFP antibodies previously coupled to protein-A Sepharose. Immunoprecipitated proteins were analyzed for the presence of cytoUsherin by western blotting, using cMyc antibody (Santa Cruz, 1:500 dilution). HEK293 cell lysates were prepared by using lysis buffer (PBS pH 7.4, 0.5% Triton X-100, 0.1% DOC and a protease inhibitor cocktail), and the lysate was clarified by centrifugation (45 min, 13 000 g ). Aliquots of the extracts were immunoprecipitated for 6 h at 4°C. Lysates from transfected HEK293 producing either cytoUsherin alone or cytoUsherin and GFP were used as controls.
The in vitro binding assays were carried out using GST-tagged fusion proteins as follows: radiolabeled proteins were translated in vitro with the T7-coupled transcription–translation system (Promega) according to manufacturer's instructions. To test usherin interactions with whirlin, harmonin and other USH1 proteins, a bacterial lysate containing GST–constructs of either usherin, harmonin a and whirlin, or GST alone, was incubated with pre-equilibrated glutathione–Sepharose beads (Pharmacia) for 90 min at 4°C on a rotating wheel. The beads were washed three times with binding buffer (PBS with 5% glycerol, 5 m m MgCl 2 and 0.1% Triton X-100) supplemented with an EDTA-free protease inhibitor cocktail (Roche) and then incubated with 35 S-labeled usherin/whirlin/harmonin/myosin VIIa tail/cadherin 23/protocadherin 15/ sans for 3 h at 4°C on a rotating wheel. The beads were then washed four times with binding buffer supplemented with 150 m m NaCl, and bound proteins were resuspended in 20 µl 2×SDS sample buffer and then analyzed on a 4–12% SDS–polyacrylamide gel.
Supplementary Material is available at HMG Online.
We thank J.-P. Hardelin and S. Cure for critical reading of the manuscript and J. Levilliers for her inexhaustible and valuable help. This work was supported by grants from the R. and G. Strittmatter Foundation, the A. and M. Suchert Forschung contra Blindheit-Initiative Usher Syndrome and the European Commission FP6 Integrated Project EUROHEAR, LSHG-CT-20054-512063. A.A.'s postdoctoral fellowship was granted by the Pasteur-Weizmann Foundation.
Conflict of Interest statement. None declared.