Unconventional Myosin VIIA Is a Novel A-kinase-anchoring Protein*

To gain an insight into the cellular function of the unconventional myosin VIIA, we sought proteins interacting with its tail region, using the yeast two-hybrid system. Here we report on one of the five candidate interactors we identified, namely the type Iα regulatory subunit (RIα) of protein kinase A. The interaction of RIα with myosin VIIA tail was demonstrated by coimmunoprecipitation from transfected HEK293 cells. Analysis of deleted constructs in the yeast two-hybrid system showed that the interaction of myosin VIIA with RIα involves the dimerization domain of RIα. In vitrobinding assays identified the C-terminal “4.1, ezrin, radixin, moesin” (FERM)-like domain of myosin VIIA as the interacting domain. In humans and mice, mutations in the myosin VIIAgene underlie hereditary hearing loss, which may or may not be associated with visual deficiency. Immunohistofluorescence revealed that myosin VIIA and RIα are coexpressed in the outer hair cells of the cochlea and rod photoreceptor cells of the retina. Our results strongly suggest that myosin VIIA is a novel protein kinase A-anchoring protein that targets protein kinase A to definite subcellular sites of these sensory cells.

Myosin VIIA is an unconventional myosin, which plays a key role in both the eye and inner ear functions. Mutations in the MYO7A gene are responsible for Usher syndrome type 1B (USH1B), 1 characterized by congenital sensorineural deafness, vestibular dysfunction, and retinitis pigmentosa (1). Mutations in this gene also underlie autosomal recessive and dominant forms of isolated deafness (2)(3)(4). In the retina, myosin VIIA is expressed in the photoreceptor cells and the pigment epithelium (5)(6)(7), where it is believed to be involved in the transport of opsins (8) and the distribution of melanosomes (9), respectively. In the inner ear, the expression of myosin VIIA is restricted to the sensory hair cells in both the cochlea (auditory apparatus) and the vestibule (balance organ) (5,6,10,11). Myosin VIIA-defective mice (shaker-1 mutants) are deaf (12) and harbor a disorganization of the hair cell stereocilia, i.e. the mechanoreceptive structure (13). In addition, the uptake of aminoglycosides by the hair cells is impaired (14). Interestingly, a crucial role of myosin VIIA in phagocytosis has recently been demonstrated in the Dictyostelium amoeba (15), which suggests that the role of this protein in endocytosis has been conserved throughout evolution.
In addition to the eye and inner ear, myosin VIIA is expressed in a variety of organs or tissues including the olfactory epithelium, brain, choroid plexus, intestine, liver, kidney, adrenal gland, and testis (16,17). However, the phenotypes associated with myosin VIIA mutations in humans and mice have so far revealed deleterious effects only in the inner ear and the eye.
The primary structure of myosin VIIA (Fig. 1A) is typical of unconventional myosins (18). Myosin VIIA is composed of a N-terminal motor head domain (aa 1-729) containing the ATPand actin-binding sites, a neck region (aa 730 -855) composed of five isoleucine/glutamine (IQ) motifs, which are thought to bind to calmodulin, and a long tail (aa 856 -2215) (10,19). The tail segments of unconventional myosins contain domains that are believed to interact with other proteins, such as cargo molecules and components of the signal transduction pathways. The tail of myosin VIIA (Fig. 1A) begins with a short coiled-coil domain (79 aa), which has been shown, using the yeast two-hybrid system, to be implicated in the formation of myosin VIIA homodimers (2). This dimerization domain is followed by two large repeats (I and II) of about 460 aa each, with a poorly conserved Src homology 3 domain in between (19,20). Each repeat is composed of a "myosin tail homology-4" (MyTH4) domain and a "4.1, ezrin, radixin, moesin" (FERM)like domain (21), which characterizes the protein 4.1 superfamily. This family includes protein 4.1, talin, filopodin, and merlin/schwannomin and the ERM proteins, ezrin, radixin, and moesin. The FERM domains of these proteins have been shown to bind, either directly or via an adaptor protein, to various integral membrane proteins (for a review, see Ref. 22). Accordingly, myosin VIIA could be one of the molecules that cross-link the cortical actin filaments to the plasma membrane (22)(23)(24).
The identification of proteins that interact with myosin VIIA tail is expected to provide helpful clues for understanding its role at the cell level. We used the yeast two-hybrid system to seek such interacting proteins. Five specific cDNAs were thereby isolated. Here we report on one of these that encodes the type I␣ regulatory subunit (RI␣) of cAMP-dependent protein kinase (PKA).
Human Retina cDNA Fusion Library-Human retina poly(A) ϩ RNA was purchased from CLONTECH (Palo Alto, CA). The cDNA library was constructed by generating random and oligo(dT)-primed cDNAs using the SuperScript cDNA synthesis kit (Life Technologies, Inc.). cDNAs were cloned downstream GAL4AD in pGAD-GE. About 7 ϫ 10 5 independent clones were obtained.
Yeast Two-hybrid Screening-A yeast two-hybrid screening was performed according to Vojtek et al. (26). Briefly, the yeast cells expressing the LexABD-bait fusion protein were transformed with the human retinal cDNA expression library (see above). The positive clones were rescued, retransformed into fresh L40 yeast cells, and confirmed by growth on plates lacking histidine and ␤-galactosidase assay. The positive clones were further analyzed by cotransformation with irrelevant baits, used as negative controls (see "Results and Discussion").
DNA Sequencing and Sequence Analysis-DNA sequencing was performed using the Sequenase kit (U.S. Biochemical Corp.) on an Applied Biosystems ABI 377 or ABI 373 automatic DNA sequencer. Sequences were analyzed using the Wisconsin Package (Genetics Computer Group, Inc., Madison, WI).
Data base searches were performed with the BLAST 2.0 algorithm in the nonredundant nucleotide sequence data base (GenBank TM /EMBL/ DDBJ/PDB) and nonredundant peptide sequence data base (Gen-Bank TM CDS translations/PDB/SwissProt/PIR/PRF) maintained at NCBI (available on the World Wide Web).
Cell lysates containing biotinylated myosin VIIA fusion peptides or an unrelated control protein (chloramphenicol acetyltransferase) were incubated with tetrameric avidin resin (TetraLink resin; Promega) on a rotating wheel at 4°C overnight. After incubation, the coated resin was washed three times with buffer B supplemented with 0.1% Nonidet P-40 (Sigma). For in vitro binding assay, 100 l of a bacterial cell lysate containing His 6 -RI␣-(1-286) in buffer B with 0.05% Nonidet P-40 were incubated with 3-6 l of the coated resins for 2 h at 4°C on a rotating wheel. The quantity of the protein bound to resin was visualized on a Western blot in order to adjust the amounts of coated resins in the binding experiments. After incubation, the resin was washed five times with 1 ml of buffer HS (50 mM Tris-HCl, pH 7.5, 250 mM NaCl, 5 mM MgCl 2 , 1 mM EDTA, 10% glycerol, 1% Nonidet P-40). Bound proteins were eluted from the resin by boiling in SDS gel-loading buffer, separated by electrophoresis on SDS-polyacrylamide gel (12%) (27), electrotransferred to nitrocellulose sheets, and then submitted to immunoblot analysis. The blot was blocked in 5% powdered nonfat milk. To reveal His 6 -RI␣-(1-286), the blot was incubated with monoclonal antibodies against the polyhistidine tag at 1:2000 dilution (anti-RGS-His, Qiagen). Horseradish peroxidase-conjugated goat anti-mouse antibodies (Bio-Rad; 1:5000 dilution) and the ECL chemiluminescence system (Amersham Pharmacia Biotech) were used for detection. To visualize the amounts of biotinylated peptide/proteins used in each binding assay, the same blot was stripped of bound antibodies (by incubation in a buffer containing 62 mM Tris-HCl, pH 6.8, 2% SDS, and 100 mM ␤-mercaptoethanol at 50°C for 30 min), washed in PBS, blocked in dry milk and incubated with horseradish peroxidase-conjugated streptavidin at 1:2000 dilution (Amersham Pharmacia Biotech). The protein bands were revealed with the ECL chemiluminescence system.
Immunoprecipitation-HEK293 cells were cultured in high glucose Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum, 2 mM glutamine, 150 units/ml gentamicin. The cells were grown on 10-cm dishes until 50% confluency. They were transfected with a cDNA fragment encoding the entire myosin VIIA tail (aa 848 -2215) cloned in pEGFP-C2 (CLONTECH) (10 g of total DNA/10-cm dish), using LipofectAMINE (Life Technologies). After 36 h, cells were rinsed, pelleted, and resuspended in 50 mM Tris-HCl, pH 7.4, 150 mM NaCl (Tris-buffered saline). Cell extracts were prepared using 1% deoxycholate as described previously (28). Polyclonal antibodies to either RI␣ or myosin VIIA were preincubated with protein A-agarose (Amersham Pharmacia Biotech) for 1.5 h at room temperature. The cell extract was then added, and the mixture incubated was for an additional 2 h at 4°C. The resin was pelleted by centrifugation, washed three times with 50 mM Tris-HCl, pH 7.4, 0.1% Triton X-100, resuspended in SDS gel-loading buffer (62 mM Tris-HCl pH 7.4, 2% SDS, 10%  (20) and the BG2137E construct carries the Gly 2137 3 Glu missense mutation found in an Usher 1B-affected individual (25). Diploids were tested for growth on a selective medium (Leu Ϫ , Trp Ϫ , His Ϫ , with 3 mM 3-AT) and for ␤-galactosidase (lacZ) activity by a filter assay. Only the original bait construct interacts with the D10 peptide in yeast. None of the other constructs was able to activate the reporter genes when coexpressed with D10.
Immunohistofluorescence-Adult guinea pig inner ears were fixed for 3 h in 4% paraformaldehyde in PBS, pH 7.4, and then decalcified for 4 days in 10% EDTA-PBS at 4°C. Human retinas were fixed in 4% paraformaldehyde-PBS overnight at 4°C. After three washes with PBS, the samples were immersed in 20% sucrose-PBS for 12 h at 4°C and then frozen in Tissue-Tek O.C.T. embedding medium (Miles). Cryostat sections (10 -14 m) were processed as described previously (6). The polyclonal antibodies against the RI␣ (sc-906) and RI␤ (sc-907) regulatory subunits of PKA were purchased from Santa Cruz Biotechnology, Inc. (Santa Cruz, CA); these antibodies are partially cross-reacting. The inner ear hair cells were identified using the affinity-purified polyclonal anti-myosin VIIA antibody (6). The specificity of the immunolabeling was verified by omission of the first antibody and use of the preimmune serum.

Yeast Two-hybrid Screening for Interacting Partners of the Myosin VIIA Tail Identifies the RI␣ Regulatory Subunit of Protein Kinase A-
The presence of a FERM domain at the C-terminal end of myosin VIIA is expected to target this molecule to cell membrane proteins and possibly serves as a "cargo binding domain" (21). This domain is preceded by a MyTH4 domain of unknown function. The same tandem arrangement of the two domains is present in other motor proteins, namely myosin X, myosin XV, and a plant kinesin (18,19,29,30), suggesting that the association of these domains has a functional significance. Therefore, two peptides covering the Cterminal MyTH4 and FERM domains (repeat II), which differed by their N termini, were tested for their ability to be used as baits in a two-hybrid screen. They were expressed as a fusion protein with LexABD in the pNLX3 vector. The longer peptide (aa 1698 -2215) caused a strong activation of the two reporter genes on its own and could not be used. The shorter one (aa 1752-2215) gave a weak activation of the histidine synthetase reporter gene; this was overcome by growing the cells in the presence of 3 mM 3-AT, which restores histidine auxotrophy. This C-terminal 464-aa peptide, comprising the MyTH4 domain (except for the first four aa) and the entire FERM domain, was used as bait (see Fig. 3). Since loss of myosin VIIA in humans results in retinal degeneration (see Ref. 31), we constructed a two-hybrid cDNA library from human retina (see "Experimental Procedures"). Approximately 6 ϫ 10 6 yeast transformants were screened, representing about 10 times the complexity of the cDNA fusion library. More than 1000 colonies were found to grow on histidine-free plates with 3 mM 3-AT; all were also ␤-galactosidase-positive. Two hundred fifty well-growing colonies were retained for further analysis.
To identify redundant clones, library inserts of the 250 colonies were polymerase chain reaction-amplified with primers specific to the library vector, blotted onto Hybond-N ϩ membrane, and hybridized with randomly chosen library inserts. This reduced the 250 colonies to 141 clones.
To test the specificity of their interaction with the bait, the 141 plasmids were isolated, retransformed into yeast strain AMR70, and mated with the L40 strain containing either the LexABD-bait or the LexABD-lamin C, which served as a negative control. For 27 (19%) of these plasmids, upon coexpression with the LexABD-bait, the His ϩ LacZ ϩ phenotype could not be confirmed. Sixty-nine (49%) plasmids showed also an interaction with LexABD-lamin C and were discarded. The remaining 45 plasmids were analyzed further. They were used to transform the Y187 strain, which was mated with the HF7 strain expressing the GAL4BD in fusion with the bait (GAL4BD-bait in pAS2). This switch from LexABD-to GAL4BD-bait fusions eliminates false positives that may arise as a result of binding to the junction region between LexABD and the bait. Finally, clones that would be able to interact with various FERM domains were eliminated by mating these 45 transformed Y187 yeasts with the HF7 strain carrying either the entire or the N-terminal part (containing the FERM domain) of merlin/schwannomin in fusion with GAL4BD (pGBT10-DH3 and pGBT10-DH1, respectively).
We thereby identified five independent prey clones that 1) grew well on histidine-free plates with 3 mM 3-AT and displayed strong ␤-galactosidase activity, when coexpressed with the bait, and 2) showed no interaction with the reporter genes in coexpression with lamin C or merlin/schwannomin. This strongly suggested that the proteins encoded by these clones are binding partners of myosin VIIA. Sequencing of the cDNA inserts of the five clones showed in all of them an open reading frame in direct fusion with the GAL4AD. One of them, clone D10, contained a fragment of the human cDNA coding for the regulatory subunit RI␣ of PKA (GenBank TM accession number M33336; Ref. 32).
The cDNA fragment that is present in clone D10 starts by 90 nt of the 5Ј-UTR, followed by the first 857 nt of the RI␣ coding sequence. The 90 nt of the 5Ј-UTR form an open reading frame (30 aa) in frame with the RI␣ coding sequence. The structure of the regulatory subunits (R) of PKA comprises a N-terminal region involved in homodimerization, a hinge that binds to the active site of the catalytic subunit (C) in the absence of cAMP, and two contiguous cAMP-binding domains, A (aa 145-262) and B (aa 263-381). D10 encodes the first 286 aa of RI␣, which is all except for the C-terminal cAMP-binding domain (see Fig. 3).
Immunoprecipitation experiments performed on lysates from transfected HEK293 cells expressing the entire myosin VIIA tail (see "Experimental Procedures") confirmed the interaction of the endogenous PKA RI␣ subunit to myosin VIIA, and also showed the coimmunoprecipitation of the ␤ isoform of RI (RI␤) (Fig. 2). The existence of RI␣-RI␤ heterodimers (33) in these cells may account for the RI␤ coimmunoprecipitation, although a direct binding of the RI␤ subunit to myosin VIIA cannot be excluded.
The Dimerization Domain of RI␣ Is Involved in the Binding to Myosin VIIA-The PKA holoenzyme is composed of two  (lanes 3 and 4). In lane 5, untransfected cells were used as a negative control for immunoprecipitation with the anti-myosin VIIA. regulatory (R) and two catalytic (C) subunits. Two classes of R subunits exist, RI and RII, which define the type I and type II PKA, respectively (34). The activation of PKA is triggered by the intracellular second messenger cAMP, which binds to the R subunits, resulting in the dissociation of the holoenzyme and the release of free active C subunits. Active PKA controls the function of various structural proteins and key enzymes by phosphorylating specific serine and threonine residues in these proteins. PKA molecules are targeted to specific subcellular locations, such as the cytoskeleton, plasma membrane, nucleus, Golgi apparatus, endoplasmic reticulum, and other organelles (for a review, see Refs. 34 and 35), via interactions of their R subunits with protein kinase A-anchoring proteins (AKAPs). The current model proposes that the anchoring of PKA to subcellular structures is a mechanism that ensures a high concentration of PKA near its substrate proteins; this would facilitate their rapid and preferential phosphorylation upon an increase of the intracellular cAMP concentration. In addition, this may allow for activation of localized pools of PKA. The dimerization domain of the PKA RII subunit and, more recently, of the RI subunit, has been implicated in the interaction with AKAPs (36 -40). To determine if the binding of RI␣ to myosin VIIA involves the dimerization domain (aa 14 -63), several D10-deleted constructs were tested for their ability to bind to the bait in the two-hybrid system (Fig. 3). A construct encoding the first 249 aa of RI␣ (pGAD-GE RI␣ 1-249) conferred a His ϩ LacZ ϩ phenotype to the yeast strain L40 cotransformed with the LexABD-bait, thus excluding the involvement of the extra 30 aa encoded by the 5Ј-UTR in the interaction of D10 with the bait. An in-frame internal deletion in D10 (D10⌬SacI), which encompasses the last 22 aa encoded by the 5Ј-UTR and the first 18 aa of RI␣, did not abolish the binding of D10 to the bait either (Fig. 3). In contrast, a shorter construct (RI␣ 77-249) was unable to interact with myosin VIIA, demonstrating that the N-terminal 76 aa residues of RI␣ are necessary for the interaction with the bait (Fig. 3). Finally, another deleted construct (RI␣-(47-249)) lacking the first 46 aa of RI␣ led to a LacZ ϩ phenotype, but the cotransformed yeast grew poorly on histidine-free medium with 3 mM 3-AT, indicating that this deletion affects the binding activity of D10 to the bait. Therefore, in the yeast two-hybrid system, we were able to map the region of RI␣ interacting with myosin VIIA to an N-terminal segment (aa 19 -76) encompassing the dimerization domain. This result strongly suggests that myosin VIIA is an AKAP.

RI␣ Binds to the C-terminal FERM Domain of Myosin
VIIA-In a first attempt to localize the binding site of peptide D10 within the bait, using the yeast two-hybrid system, several truncated or mutated versions of the bait were generated (Fig. 1B), namely two C-terminal truncated baits (BdC1 (aa 1752-2006) and BNdC3 (aa 1698 -1859)), two N-terminal truncated baits (BdN1 (aa 1896 -2215) and BdN2 (aa 2007-2215)), and the BF1800I and BG2137E mutated baits (see "Experimental Procedures"). BF1800I carries, in the MyTH4 domain, the Phe 1800 3 Ile substitution, which is observed in the 26SB allele of the shaker-1 mouse mutant (20). BG2137E carries, in the FERM domain, the Gly 2137 3 Glu substitution, which has been found in an USH1B-affected individual (25). The expression of all bait constructs in the yeast was analyzed on a Western blot with an anti-LexA antibody. Whereas all fusion proteins were expressed at similar levels, a high proportion of smaller bands was observed for both truncated and mutated baits as compared with the normal bait, indicating an elevated proteolytic degradation in the former (data not shown). Consistently, neither the truncated nor the mutated baits were able to activate the reporter genes when coexpressed with the D10 peptide, in the yeast two-hybrid system (Fig. 1B). The possibility that the inability of D10 to interact with any of the truncated or mutated baits was due to a reduced stability of these peptides is further supported by the fact that the two mutated baits were not able to activate the reporter genes when coexpressed with any of the original 141 library plasmids (including those that were positive with the negative controls lamin C and merlin/schwannomin used in the screening). Therefore, it appears that mutations or truncations of the myosin VIIA tail make the protein susceptible to proteolytic degradation in yeast cells. Interestingly, a marked decrease in the amount of myosin VIIA has been found in tissues of the 26SB shaker-1 mouse (41), which carries the mutation that was introduced in our BF1800I construct. FIG. 3. The dimerization domain of RI␣ is involved in the interaction with myosin VIIA. Schematic structure of the RI␣ protein is shown at the top. RI␣ consists of a N-terminal region involved in the homodimerization of the PKA regulatory subunit, a hinge that binds to the active site of the catalytic subunit in the absence of cAMP (inhibitor site), and two contiguous cAMP-binding domains A and B. The black bar below depicts the region of RI␣ protein contained in the D10 clone isolated in the two-hybrid screening. The ability of various GAL4AD-RI␣ truncated peptides (represented by gray bars) to activate the reporter genes when coexpressed with the LexABD-bait fragment of myosin VIIA in the yeast two-hybrid system is indicated on the right. Bait and prey plasmids were cotransfected into the L40 yeast strain and tested both for growth on a selective medium (Leu Ϫ , Trp Ϫ , His Ϫ , with 3 mM 3-AT) and for ␤-galactosidase (lacZ) activity by a filter assay. Residues 19 -76 of RI␣ are necessary for the interaction of D10 with the myosin VIIA bait. To circumvent the difficulty in determining the binding site within myosin VIIA using the yeast two-hybrid system, in vitro binding experiments were performed. The bait, its truncated or mutated versions, and an unrelated control protein (chloramphenicol acetyltransferase) were expressed with an N-terminal biotin tag in bacteria (see "Experimental Procedures"). These peptides were immobilized on an avidin-coupled resin and tested for their ability to bind the His 6 -RI␣-(1-286). The biotinylated baits bound to the resin and the associated proteins were eluted and analyzed on Western blots (Fig. 4). Peptide His 6 -RI␣-(1-286) was found to bind to the resin coated with the bait but not to the resin alone or coated with the control protein (Fig. 4). The resin coated with the C-terminally truncated bait peptides BdC1 (aa 1752-2006) or BdC2 (aa 1752-1931) failed to interact with His 6 -RI␣- . In contrast, peptide BdN2, comprising the C-terminal 209 aa of the bait (aa 2007-2215), retained His 6 -RI␣-(1-286). Interestingly, the bait carrying the G2137E mutation was also observed to bind to His 6 -RI␣-(1-286) (Fig. 4). These results establish that 1) RI␣ binds to myosin VIIA within the last 209 aa of the C-terminal FERM domain and 2) the G2137E mutation, which has been found in an USH1B-affected patient, does not affect this interaction.
Sequence analysis detected no homology between myosin VIIA FERM domain and any AKAP domain. More than 20 AKAPs have been identified (42). Although the primary sequence of their PKA binding sites varies substantially, their predicted secondary structure indicates that they form an am-phipathic ␣-helix (36,43). Their hydrophobic surfaces have been shown to play a key role in the interactions of AKAPs with RI or RII PKA regulatory subunits (36,44). Interestingly, helical wheel analysis (PEPWheel program) revealed that within the myosin VIIA FERM domain, several sequences (containing 14 -18 aa) are predicted to fold into an amphipathic ␣-helix, i.e. between aa positions 2007 and 2024, 2053 and 2066, and 2193 and 2207.
It has been proposed that AKAPs also act as scaffolds for assembling multiprotein complexes. In particular, AKAPs may assemble signaling complexes through association with multiple enzymes and potential substrates (35,42). Since the tail of myosin VIIA contains at least three domains probably involved in protein-protein interactions, namely two FERM domains and a putative Src homology 3 domain, we propose that myosin VIIA also acts as a signaling organizer. Interestingly, whereas myosin VIIA does not contain any clear consensus motif for PKA phosphorylation (see Ref. 45), partial sequences of the four other putative ligands of the myosin VIIA tail that have been isolated so far have revealed such a motif in at least one of them. 2 Myosin VIIA and RI␣ Are Coexpressed by Some but Not All Sensory Cells of the Retina and Inner Ear-The expression of RI␣ was studied by immunohistofluorescence on tissue sections  of the eye and the inner ear, which are the two target organs of the MYO7A defect in USH1B-affected individuals. Several studies have shown that myosin VIIA in the retina is restricted to the photoreceptor cells and the pigment epithelium (5)(6)(7)10). In the human adult retina (Fig. 5, A-C), RI␣ was found to be expressed in the photoreceptor cells but not in the pigment epithelium. However, whereas all rod photoreceptors were labeled with anti-RI␣ antibodies, at least some identified cones showed no RI␣ staining (see Fig. 5B). The RI␣ and myosin VIIA labelings of the rods were both concentrated in the inner segment (Fig. 5, B and C). No RI␣ or myosin VIIA labeling was observed in the outer segment. RI␣ was detected in several other retinal cell types, which do not express myosin VIIA, such as the horizontal, amacrine, and ganglionic cells (Fig. 5, A and  B). A similar expression pattern was obtained with an antibody directed against RI␤ (data not shown). Therefore, in the human retina, myosin VIIA and PKA RI subunits are colocalized in the inner segment of rod photoreceptor cells. However, whereas the role of PKA in the outer segment has been studied for many years (46 -48), its role in the inner segment remains unknown.
In the guinea pig organ of Corti (the auditory transduction apparatus), a strong RI␣ immunoreactivity was detected in the outer hair cells (OHCs), whereas no labeling was observed in the inner hair cells (IHCs) and supporting cells. RI␣ was also detected in the epithelial cells of the inner and outer sulcus and in the fibrocytes of the spiral limbus (Fig. 5, D and E). A similar expression pattern was obtained with an antibody directed against RI␤ (data not shown). As previously shown (5,6,11), the myosin VIIA staining in the cochlea was restricted to the IHCs and OHCs (Fig. 5F). Thus, in the cochlea, myosin VIIA and RI subunits of PKA are coexpressed only by the OHCs. Until recently, limited attention had been paid to a possible regulation of the activity of the sensory hair cells by cAMP. Indeed, contrary to photo-or chemotransduction, auditory mechanotransduction does not involve a second messenger, the transduction channels being directly opened by the deflection of the hair cell stereocilia, which is produced by sound pressure waves (49). However, cAMP has been shown to regulate the activity of the auditory transduction channel in the turtle (50). In addition, PKA regulates at least three different voltage-dependent potassium channels and a Ca 2ϩ -activated nonselective cation channel in the basolateral membrane of OHCs (51-53). Hence, one or several AKAPs interacting with the plasma membrane or the cortical cytoskeleton are likely to be involved in the regulation of these ion channels. Myosin VIIA is a possible candidate, since it is present in the basolateral region of the hair cells and is expected to interact with both the cytoskeleton, by its actin-binding domain, and the membrane, by its C-terminal FERM domain (22). Characterization of the other ligands of the myosin VIIA tail should contribute to assessing the proposed scaffolding role of myosin VIIA. It could also contribute to a better understanding of the modulatory role of PKA in the visual and auditory sensory cells.