Cyclic Nucleotide-gated Channel α-3 (CNGA3) Interacts with Stereocilia Tip-Link Cadherin 23 + Exon 68 or Alternatively with Myosin VIIa, Two Proteins Required for Hair Cell Mechanotransduction*

Background: Stereocilia tip-link protein cadherin 23 couples mechanical forces to sensory transduction in inner ear receptor hair cells. Results: CNGA3 interacts with cadherin 23 and with cadherin 23 binding partner, myosin VIIa. Conclusion: Cadherin 23 and myosin VIIa form alternate complexes with CNGA3. Significance: CNGA3 interactions with cadherin 23 +68 and myosin VIIa predict a role for CNGA3 in hair-cell mechanotransduction. Previously, we obtained evidence for a photoreceptor/olfactory type of CNGA3 transcript in a purified teleost vestibular hair cell preparation with immunolocalization of CNGA3 protein to stereocilia of teleost vestibular and mammalian cochlear hair cells. The carboxyl terminus of highly Ca2+-permeable CNGA3 expressed in the mammalian organ of Corti and saccular hair cells was found to interact with an intracellular domain of microfibril interface-located protein 1 (EMILIN 1), a member of the elastin superfamily, also immunolocalizd to hair cell stereocilia (Selvakumar, D., Drescher, M. J., Dowdall, J. R., Khan, K. M., Hatfield, J. S., Ramakrishnan, N. A., and Drescher, D. G. (2012) Biochem. J. 443, 463–476). Here, we provide evidence for organ of Corti proteins, of Ca2+-dependent binding of the amino terminus of CNGA3 specifically to the carboxyl terminus of stereocilia tip-link protein CDH23 +68 (cadherin 23 with expressed exon 68) by yeast two-hybrid mating and co-transformation protocols, pulldown assays, and surface plasmon resonance analysis. Myosin VIIa, required for adaptation of hair cell mechanotransduction (MET) channel(s), competed with CDH23 +68, with direct Ca2+-dependent binding to the amino terminus of CNGA3. Based upon the premise that hair cell stereocilia tip-link proteins are closely coupled with MET, these results are consistent with the possibility that CNGA3 participates in hair-cell MET. Together with the demonstration of protein-protein interaction between HCN1 and tip-link protein protocadherin 15 CD3 (Ramakrishnan, N. A., Drescher, M. J., Barretto, R. L., Beisel, K. W., Hatfield, J. S., and Drescher, D. G. (2009) J. Biol. Chem. 284, 3227–3238; Ramakrishnan, N. A., Drescher, M. J., Khan, K. M., Hatfield, J. S., and Drescher, D. G. (2012) J. Biol. Chem. 287, 37628–37646), a protein-protein interaction for CNGA3 and a second tip-link protein, CDH23 +68, further suggests possible association of two different channels with a single stereocilia tip link.

Cyclic nucleotide-gated (CNG) 2 channels mediate sensory transduction in vision and olfaction (1,2) and are highly permeable to Ca 2ϩ , representing a primary mechanism for Ca 2ϩ entry. In previous work, we obtained three full-length sequences for a photoreceptor/olfactory type of CNGA3 transcript expressed by hair cells in a purified teleost vestibular hair cell preparation (3). CNGA3 subunit protein was immunolocalized to sites on the stereocilia and subcuticular plate region for both teleost saccular hair cells and mammalian cochlear cells. Protein-protein interaction protocols indicated for both models that the carboxyl terminus of CNGA3 binds to a cytoplasmic carboxyl terminus of elastin microfibril interface-located protein 1 (EMILIN1), an extracellular matrix protein and member of the elastin superfamily of proteins. EMILIN1 was also localized to hair cell stereocilia along with phosphodiesterase 6C, a phosphodiesterase regulating cGMP gating of CNGA3 in cone photoreceptors (3).
Here we present biochemical evidence that the cytoplasmic amino terminus of ion channel CNGA3 binds the cytoplasmic carboxyl terminus of cadherin 23 (CDH23), a transmembrane protein that forms the upper two-thirds of filaments constituting hair cell stereocilia tip links. The tip links are considered important in transducing the mechanical stimuli of inner ear hair cells without, however, the upper tip-link protein, cadherin 23, being previously identified as participating in protein-pro-10). The transformants were plated on SD/ϪLeu medium, and after 5 days the colonies were collected and used for yeast mating experiments. Generally, 3 g of pGADT7-Rec was mixed with 20 l of the double-stranded cDNA for standard yeast transformation reactions.
Mating and Screening-To screen for interacting proteins, we mixed cells containing bait constructs in strain Y2H gold with the cells containing prey constructs in Y187. These cells were allowed to mate in 2ϫ YPDA medium (Clontech) overnight at 30°C and plated on a selection medium SD/ϪAde/ ϪHis/ϪLeu/ϪTrp/aureobasidin A (QDO/A) (quadruple dropout medium with aureobasidin A). After 5-7 days, the larger colonies that appeared on the plates were analyzed by PCR, amplification products were sequenced, and prey sequences were identified using BLAST (NCBI).
Yeast Two-hybrid Co-transformation Analysis-The rat CNGA3 amino-terminal cytoplasmic region (CNGA3-N) was prepared as a bait plasmid in pGBKT7 vector. The cytoplasmic carboxyl terminus of CDH23 expressed in rat organ of Corti was targeted in PCR using specific primers (upstream primer, 5Ј-GCCGAATTCTTCTGCCGGAACCTGGAGCTG-3Ј; downstream primer, 5Ј-TGCGGATCCTTACAGTTCCGTGATTTC-CAGGG-3Ј; NM_053644, nucleotides 9747-10299). The PCR amplification revealed two fragments corresponding to CDH23 with and without exon 68. These fragments were cloned into the prey vector pGADT7. For the yeast two-hybrid assay, ϳ500 ng of each vector was co-transformed into Y2H gold yeast strain using standard protocol (Clontech Yeast Two-hybrid Manual), and the transformants were plated on nutrient-deficient selection medium (QDO/A) for 3-5 days. Transformants were also plated on a less stringent deficient medium (SD/ϪLeu/ϪTrp) to select transformants. The colonies were picked and streaked on higher stringency medium (QDO/A). Negative controls contained either the bait or the prey along with the prey or bait empty vector, respectively. To further confirm the binding interaction of CNGA3-N with CDH23 ϩ68, we performed the reversal system of bait and prey along with negative controls, CNGA3-N as prey and CDH23 ϩ68 as bait in selection medium (QDO/A).
Western Blot Analysis-Purity of the fusion proteins was ascertained by SDS-PAGE (4 -12% NuPAGE gel followed by overnight blocking with 5% nonfat milk at 4°C) with Coomassie Blue staining. Blots were incubated with anti-Xpress monoclonal antibody (Invitrogen, 1:5,000) for 3 h at room temperature or overnight at 4°C, washed with PBS containing 0.1% Tween 20, and incubated with HRP-conjugated anti-mouse IgG (1:10,000) for 1 h at room temperature, and the proteins were detected using Western Lightning chemiluminescence (Perkin-Elmer Life Sciences).
Pulldown Assay-Pulldown assays for rat organ of Corti proteins were performed as described previously (3) with GST fusion protein (GST-CNGA3-N) and hexahistidine-tagged fusion proteins (CDH23 ϩ68 and CDH23 Ϫ68). GST conjugates were mixed with hexahistidine-tagged CDH23 ϩ68 and Ϫ68 in binding buffer (PBST buffer, pH 7.4, 1ϫ protease inhibitors) for 2 h at room temperature. Glutathione-Sepharose beads (30 l of 50% slurry) were added to the reaction mix for 3 h at room temperature or overnight at 4°C followed by centrifugation for 2 min at 1,000 ϫ g. The beads were washed 5 times in the binding buffer, incubated at 70°C for 5 min in gel loading buffer (30 l), and centrifuged. The supernatant was then electrophoresed in a 4 -12% NuPAGE gel (Invitrogen), electroblotted to a nylon membrane, and immunodetected using anti-Xpress antibody and CDH23-specific goat polyclonal antibody (sc-26338).
Protein Expression and Purification-cDNA was amplified from rat organ of Corti and inserted into pRSET vectors. PCR primers for protein expression in pRSET-B included CNGA3 amino terminus upstream (5Ј-GCCAGATCTATGG-CCAAGGTCAATACCCAA-3Ј) and downstream (5Ј-TGC-GAATTCTTAGTTGCTGGAGGGGTCCAC-3Ј) (AJ272428). For CDH23 ϩ68 and Ϫ68 in pRSET-A, the primers were: upstream, 5Ј-GCCGGATCCTTCTGCCGGAACCTGGAGC-TG-3Ј, and downstream, 5Ј-TGCGAATTCTTACAGTTCCG-TGATTTCCAGGG-3Ј (NM_053644). For the carboxyl terminus of myosin VIIa (amino acids 1608 -2177, composed of SH3, MyTH4, and FERM domains) in pRSET-A, the upstream primer was 5Ј-GCCGGATCCGGCTGGGCCAATGGCATC-AAC-3Ј, and downstream primer was 5Ј-TGCGAATTCTCA-CCTCCCGCTCCTGGAGTTCCT-3Ј (NM_153473). The PCR products were purified and digested with restriction enzymes and cloned into similarly digested pRSET vectors in Escherichia coli DH5␣ cells (Invitrogen). Clones were selected and sequence-verified before protein expression studies. The pRSET expression vectors containing the desired sequences were used to transform E. coli BL21 (DE3) cells (Invitrogen). Selected clones were used for expression of fusion proteins induced by adding 1 mM isopropyl ␤-D-1-thiogalactopyranoside in LB culture medium. The cultures were incubated for 4 h at 37°C, and the cells were harvested, lysed in 10 mM phosphate buffer, 8 M urea (pH 8.0), centrifuged, and analyzed via 4 -12% SDS-PAGE with protein bands visualized by Coomassie Blue staining. Clones that showed robust expression were selected and stored for future use. For purification of fusion proteins, 100 -500-ml cultures of each clone were grown for 4 -6 h with isopropyl ␤-D-1-thiogalactopyranoside, lysed by sonication, and centrifuged at 20,818 ϫ g. The clear supernatant was mixed with 0.5 ml Talon beads (Clontech), and the beads were sedimented and washed. The bound fusion proteins were eluted by the addition of elution buffer (His TALON buffer, Clontech) followed by mixing and centrifugation. The supernatant con-taining the fusion proteins was then exhaustively dialyzed against cold PBST buffer containing 1ϫ protease inhibitors for 24 h at 4°C with constant stirring and 5 or 6 buffer changes. Protein was determined with the Qubit fluorescent system (Invitrogen).
SPR Analysis-The binding interactions of CNGA3-N and CDH23 or myosin VIIa tail expressed in rat organ of Corti were measured using SPR with a Biacore 3000 instrument at 25°C (Biacore, Piscataway, NJ) as previously described (3). Binding of the mobile molecule (analyte) to the immobilized molecule (ligand) produces changes in refractive index which can be detected. Affinity-purified hexahistidine-tagged fusion proteins were immobilized on a CM5 sensor chip by an aminecoupling reaction (cf. Ref. 11). Surfaces were prepared for each immobilized protein (ligand) with different concentrations of purified CDH23 ϩ68 and Ϫ68 fusion proteins as analyte. The reference surface was blocked with ethanolamine and thus contained no ligand. The response was recorded as the ligand response units (RU) minus the reference RU.
SPR Kinetic Measurements of CNGA3-N Binding-The rate and equilibrium binding constants of the interaction of CNGA3-N with CDH23 ϩ68 and myosin VIIa tail were determined in kinetic studies. The CNGA3-N fusion protein was immobilized as a ligand on a CM5 chip. Rat CDH23 ϩ68 or myosin VIIa (analyte) was diluted in a series of concentrations in HBS-N buffer including 68 M Ca 2ϩ . Kinetic values were determined using BIAevaluation 3.0 software (Biacore). A 1:1 Langmuir binding model was generally selected for calculations in the kinetic studies. Affinity-purified CDH23 ϩ68 and myosin VIIa rat fusion proteins were analyzed as analytes, 0 -320 nM and 0 -100 nM, respectively. After each reaction, the chip was regenerated with HBS-N buffer.
Single Hair Cell RT-PCR-cDNA from a pool of 25-30 singly isolated outer hair cells or, separately, a similar number of inner hair cells was prepared according to "Make Your Own Mate & Plate Library System" (Clontech) with Moloney murine leukemia virus reverse transcriptase or, for selected experiments, SMARTScribe reverse transcriptase (12). The specific upstream and downstream primers for CNGA3 amplification were 5Ј-GCCAAGGTCAATACCCAATG-3Ј and 5Ј-CGAATGGAG-ATGATGAAGCG-3Ј, respectively, designed to cross introns with the aid of Accelrys Gene 2.5 software (San Diego, CA) and a comparison with rat genomic sequence (NW_001084882.1). These PCR reactions were carried out using Advantage 2 GCpolymerase mix (Clontech) with denaturation at 95°C for 20 s, annealing at 60°C for 20 s, and extension at 72°C for 30 s, with a total of 65 cycles. Negative controls were derived from the elimination of cDNA and substitution of sterilized water in the PCR reaction. The PCR products were analyzed with 1.5% agarose gel electrophoresis (NuSieve 3:1 agarose, Lonza, Rockland, ME). All amplified PCR products (253 bp) were sequenced.

RESULTS
The Amino Terminus of CNGA3 Binds to the Carboxyl Terminus of Cadherin 23 by Yeast Two-hybrid Analysis-Previously, a role for CNGA3 in vestibular hair cell function was implied from the finding of three forms of CNGA3 transcript expressed in a purified preparation of Ͼ10 6 hair cells (3) (GenBank TM accession numbers HQ542177, HQ542178, HQ542179). CNGA3 protein was immunolocalized to stereocilia, the site of mechanotransduction, in these saccular hair cells. A role for highly Ca 2ϩ -permeable CNGA3 in hair cells, across hair cell type from vestibular to cochlear hair cells, was suggested by immunolocalization of CNGA3 protein to stereocilia of cochlear outer hair cells (3) and cochlear inner hair cells. Evidence of transcript expression of CNGA3 in singly isolated cochlear inner and outer hair cells (Fig. 1B, arrow) further bolsters the conclusion that CNGA3 is expressed in cochlear as well as in vestibular hair cells.
Molecular function for ion channels can be deduced from protein-protein interactions of their intracellular domains. To identify CNGA3 interaction partners, we carried out yeast twohybrid mating screens of a rat organ of Corti prey cDNA library using the intracellular amino terminus of CNGA3 (CNGA3-N) amplified from cDNA derived from rat organ of Corti, representing auditory sensory epithelium, as bait. The bait sequence corresponded to aa 1-146 of rat CNGA3 (GenBank TM accession number AJ272428) extending to the S1 transmembrane domain, with identity to both human and trout saccular hair cell CNGA3 (Fig. 1C, RAT CNGA3-N compared with HUM CNGA3-N and THC CNGA3-N sequence, respectively), indicating conservation of sequence across evolution. Transformation efficiency, based upon the number of colonies that appeared on the SD/ϪLeu plates after 5 days, was 8.4 ϫ 10 5 /3 g of vector.
Overall, 92 prey clones were sequenced for the amino terminus bait (CNGA3-N). We noticed a strong interaction in quadruple drop-out media (i.e. multiple clones) for CNGA3-N and prey, which was identified as rat CDH23, a stereociliary tip-link protein. There are two splice variants for CDH23: Ϯexon 68, with and without an insert of 105 bp encoding 35 amino acids (aa 3210 -3245, Fig. 2, A and D) into the putative internal PDZ binding domain of CDH23 (6). Both splice variants retain the same carboxyl terminus containing a second PDZ binding domain. And both splice variants are expressed in organ of Corti of the adult ( Fig. 2A, arrow ϩ and arrow Ϫ for PCR amplification of CDH23 ϩ68 and CDH23 Ϫ68 transcripts, respectively). Transcript for CDH23 ϩ68 is differentially found in sensory epithelia of the inner ear, specific for cochlear hair cells (4), not expressed in heart, kidney, spleen, brain, and retina, consistent with ϩ68-specific function in the inner ear/hair cells (4,7). Interestingly, it was CDH23 ϩ68 that was identified as the binding partner of CNGA3-N, with 99, 97, and 95% aa identity, respectively, to rat (EDL93044), mouse (NP_001239564), and human (NP_071407.4) CDH23 ϩ68, indicating again highly conserved sequence and theoretically highly conserved molecular function. These results point to hair cell-specific function that is modulated by CDH23 ϩ68 protein-protein interaction with CNGA3, potentially placing CNGA3 in position to be affected by mechanical stimulus via the stereociliary tip-link protein CDH23 ϩ68. A, shown is a diagram illustrating the predicted structure of the CNGA3 protein (GenBank TM EDL99244, rat). The cytoplasmic amino terminus sequence is encircled in a schematic depiction of CNGA3 with S1-S6 transmembrane regions and S4 hatched representing the voltage sensor and CNBD referring to the cyclic nucleotide binding domain. B, shown is PCR amplification of a 242-bp cDNA product corresponding to the amino terminus of rat CNGA3, from the analysis of pooled, singly isolated rat cochlear outer hair cells (OHC) and inner hair cells (IHC). The primers were designed to target the CNGA3 sequence expressed in a subset of rat olfactory sensory neurons (16) (GenBank TM AJ272428) with an identity to human cone photoreceptor CNGA3 (NP_001073347) and trout saccular hair cell CNGA3-2 (HQ542178). The amplification product yielded rat CNGA3 sequence with 99% nucleotide identity for cochlear outer hair cell and 98% nucleotide identity for inner hair cells to rat AJ272428. Lanes S, 1-kb standards; lanes OHC and IHC, CNGA3 242-bp amino terminus products amplified from rat cochlear outer and inner hair cells, respectively, arrow; lane B, negative control in which cDNA was omitted. C, shown is alignment (with OMIGA 2.0) of the amino terminus amino acids of rat organ of Corti CNGA3 (EDL99244) used as bait (RAT CNGA3-N, second line) in yeast two-hybrid protocols aligned with human CNGA3 (HUM CNGA3-N, first line, NP_001073347) and trout saccular hair cell CNGA3-2 (THC CNGA3-N, third line, AER38241). CNGA3 amino termini display partial EF hand Ca 2ϩ binding domains (17), DRDLDQ, aa 23-28, for rat organ of Corti sequence with comparable motifs appearing in the calcium-binding protein Cab45 (Q91ZS3), cadherin 23 (58), and EDH4 (36, NP_647540). In addition, there is a PDZ domain NGLG aa 31-34 (compare with piccolo, EDL99465). The amino terminus of trout saccular hair cell CNGA3-2 variant (third line) indicates identity to both human (first line) and rat CNGA3 (second line) sequences.
Yeast two-hybrid co-transformation analysis confirmed specific interaction between CNGA3 and CDH23 ϩ68 (  There was no band in the pulldown assays corresponding to CDH23 Ϫ68 (Fig. 3, E--G, short arrow plus dot). In these experiments similar amounts of the CDH23 ϩ68 and CDH23 Ϫ68 proteins were used to allow for a quantitative comparison of binding for ϩ68 (Fig. 3E, lanes 4 and 5) versus no binding of Ϫ68 (Fig. 3E, lanes 6 and 7, respectively). Therefore, we conclude that GST-CNGA3-N interacted with CDH23 ϩ68 and not CDH23 Ϫ68, which again suggests that amino acids encoded by exon 68 affect the interaction of CNGA3-N with CDH23. These results confirmed yeast two-hybrid mating and co-transformation findings. Negative controls included rat CDH23 ϩ68 incubated with either GST bacterial lysate (not   Fig. 2D) at 25 kDa, short arrow plus dot; lane 6, pRSET-A bacterial lysate as negative control. C, shown is a Western blot for full-length CDH23 from rat brain lysate detected with an affinity-purified goat polyclonal antibody raised against a peptide in the carboxyl terminus of CDH23 of human origin, crossing to rat sequence (sc-26338, Santa Cruz; 1:500). Lane 1, Magic XP Western standards; lane 2, full-length CDH23 at 365 kDa (arrow). D, shown is a Western blot for CDH23 ϩ68 and CDH23 Ϫ68 in pRSET-A detected with anti-CDH23 antibody (sc-26338, Santa Cruz; 1:500). Lane 1, Magic XP Western standards; lane 2, affinity-purified CDH23 ϩ68 at 30 kDa, short arrow; lane 3, RCDH23 Ϫ68 at 25 kDa, short arrow plus dot. E, shown is a Western blot for GST pulldown assay for GST-CNGA3-N with either CDH23 ϩ68 or CDH23 Ϫ68 pRSET fusion proteins (200 and 50 ng) detected with anti-Xpress monoclonal antibody. Lane 1, CDH23 ϩ68 protein (30 kDa), short arrow; lane 2, CDH23 Ϫ68 protein (25 kDa) used in the pulldown assay, short arrow plus dot; lane 3, Magic XP Western standards; lanes 4 and 5 indicate pulldown for CDH23 ϩ68 (ϳ200 ng and ϳ50 ng, respectively) with GST-CNGA3-N, short arrow; lanes 6 and 7 indicate no pulldown for CDH23 Ϫ68 (ϳ200 and ϳ50 ng, respectively) with GST-CNGA3-N; lane 8, a negative control, indicates no pulldown for CDH23 ϩ68 with GST-CNGA3-N but no beads; lanes 9 and 10 are additional negative controls (lane 9, GST lysate without CNGA3-N plus CDH23 ϩ68 plus glutathione beads; lane 10, CDH23 ϩ68 and glutathione beads), all detected with anti-Xpress monoclonal antibody. F, shown is a Western blot for a GST pulldown assay for GST-CNGA3-N with CDH23 ϩ68 and Ϫ68 pRSET fusion proteins, similar to E except for the CDH23 protein concentrations used (100 and 150 ng, respectively). G, shown is a Western blot for a GST pulldown assay: GST-CNGA3-N with CDH23 ϩ68 and CDH23 Ϫ68 fusion peptides detected with anti-CDH23 antibody (Santa Cruz, 1: 500). Lane 1, CDH23 ϩ68, short arrow; lane 2, CDH23 Ϫ68 used in the pulldown assay, short arrow plus dot; lane 3, Magic XP Western standards; lanes 4 and 5 indicate pulldown for CDH23 ϩ68 (ϳ50 and ϳ200 ng, respectively) with GST-CNGA3-N, short arrow, whereas lanes 6 and 7 indicate no pulldown for CDH23 Ϫ68 (ϳ50 and ϳ200 ng, respectively) with GST-CNGA3-N; lane 8 indicates no pulldown for CDH23 ϩ68 with GST-CNGA3-N (no beads); lanes 9 and 10 are negative controls (lane 9, GST lysate without CNGA3 and CDH23 ϩ68 and glutathione beads; lane 10, CDH23 ϩ68 and glutathione beads). H, shown is a Western blot for myosin VIIa fusion protein.   Table 1 for kinetic constants.
including CNGA3-N) or with Sepharose beads alone (Fig. 3, E  and F, lanes 8 -10) under conditions similar to those for the experimental samples.
The tip-link protein CDH23 is also known to bind myosin VIIa, which itself is required for the adaptation phase of hair cell mechanotransduction (8). This protein-protein interaction between CDH23 and myosin VIIa is also affected by the presence or absence of exon 68 in CDH23. Therefore, we wondered whether the carboxyl terminus of myosin VIIa, which binds CDH23, would disrupt the interaction between CDH23 ϩ68 and CNGA3.
The carboxyl tail of myosin VIIa in pRSET-A, at 70 kDa on Western blots (Fig. 3H, lane 2, arrow), in competition with CDH23 ϩ68 in GST pulldown assays, did prevent binding between CDH23 ϩ68 and CNGA3. CDH23 ϩ68 at 30 kDa was absent in pulldown assays with CNGA3 when myosin VIIa was included (Fig. 3I, lanes 5 and 6) with no band at the position of CDH23 ϩ68 (demarked by the long arrow in Fig. 3I, lane 4, for CDH23 ϩ68 interaction with CNGA3-N). Curiously, a band with a larger molecular mass (ϳ38 kDa) than that of CDH23 ϩ68 (30 kDa) did appear in pulldown assays including both CDH23 ϩ68 and myosin VIIa (Fig. 3I, lanes 5 and 6, short arrow with dot). This band was similar in mass to a protein in the myosin VIIa bacterial lysate (Fig. 3I, lane 2). Furthermore, a protein of very similar mass also appeared in negative controls with only myosin VIIa bacterial lysate and GST-CNGA3 attached to beads (Fig. 3I, lane 7), suggesting the possibility that a protein related to the carboxyl terminus of myosin VIIa may bind to the amino terminus of CNGA3. Confirmation of direct binding of myosin VIIa to the amino terminus of CNGA3 was obtained in multiple pulldown assays with GST-CNGA3-N as illustrated in Fig. 3J, lane 4 (myosin VIIa tail at 70 kDa, arrow). Overall, these results were consistent with the formation of alternate complexes of CNGA3 either with CDH23 ϩ68, yielding a first protein complex, or with myosin VIIa, forming a second protein complex with both interactions for CNGA3 involving proteins with roles in hair cell mechanotransduction.
SPR Analysis of CNGA3-N and Cadherin 23 Interaction-Interaction of the rat CNGA3-N with CDH23 ϩ/Ϫ68 was also examined with SPR, a biochemical protocol capable of yielding kinetic constants for protein-protein interactions, and unequivocally, the dependence of the protein-protein interac-tion on Ca 2ϩ . Affinity-purified his-tag fusion proteins of the CDH23 ϩ68 and Ϫ68 were utilized as analytes (aa 3076 -3353 and 3076 -3317, respectively, with amino acid numbering according to GenBank TM 446096.1), and affinity-purified CNGA3-N protein was immobilized on a CM5 sensor chip as ligand (11). Purified CDH23 ϩ68 and Ϫ68 at 35 and 31 kDa on Western blot (Fig. 4A, lanes 2 and 3, respectively) were diluted to 100 nM in HBS-N SPR buffer (11). The CNGA3 amino terminus distinctly bound to CDH23 ϩ68 in SPR analyses (Fig. 4, B, blue trace and C, magenta trace). There was no binding to CDH23 Ϫ68 (Fig. 4B, 26.5 M Ca 2ϩ , magenta trace) or little binding (Fig. 4C, 68 M Ca 2ϩ , red trace) consistent with results from yeast two-hybrid and pulldown assays. No binding was observed when HBS-N buffer alone was used as a negative control (Fig. 4B, red trace and C, purple trace).
The binding of the purified CNGA3-N fusion protein as ligand to CDH23 ϩ68 as analyte was Ca 2ϩ -dependent (Fig. 4D, 68 M Ca 2ϩ , blue trace; 26.5 M Ca 2ϩ , red trace). No binding was detected for 100 nM CDH23 ϩ68 in HBS-N buffer with 1 mM EGTA (Fig. 4D, green trace). A dramatic increase in binding was observed when the CDH23 ϩ68 was analyzed in HBS-N buffer containing 26.5 M Ca 2ϩ (Fig. 4D, red trace), with a further significant increase in binding at 68 M Ca 2ϩ (Fig. 4D, blue trace). The results were normalized by subtracting the SPR response (RU) for buffer alone and averaged from a single experiment performed in triplicate. All of the experiments were performed multiple times, and samples for different Ca 2ϩ concentrations were analyzed randomly. There was no binding to the reference cell surface, which was blocked by ethanolamine. Thus, the present data indicate that CNGA3-N directly interacts with CDH23 ϩ68 but only in the presence of a finite amount of Ca 2ϩ (26.5 to 68 M Ca 2ϩ ). There was no evidence of interaction of CNGA3-N to CDH23 ϩ68 in the absence of Ca 2ϩ (1 mM EGTA). In a separate experiment we obtained similar results with SPR when ligand and analyte were reversed for further proof of protein-protein interaction (CDH23 ϩ68 as ligand and CNGA3-N as analyte at 100 nM); binding again occurred between CNGA3-N and CDH23 ϩ68 at 26.5 M Ca 2ϩ (Fig. 4E, blue trace), which was further enhanced at 68 M Ca 2ϩ (Fig. 4E, red trace). Response unit plots for binding of CNGA3-N to 0 -320 nM CDH23 ϩ68 as analyte at 68 M Ca 2ϩ (Fig. 4G) indicated a K D ϭ 3.6 ϫ 10 Ϫ7 M (Table 1).
SPR conducted with shorter fragments of CDH23 ϩ68 (amino acids 3134 -3353, Fig. 2D with numbering according to GenBank TM 446096.1) yielded results for Ca 2ϩ -dependent binding of CNGA3-N to CDH23 ϩ68 (Fig. 4F) similar to those obtained for the CDH23 ϩ68 peptide which extended farther in the 5Ј direction from the carboxyl terminus (i.e. Fig. 4E).
SPR Analysis of CNGA3-N and CDH23 ϩ68 Interactions with the Myosin VIIa Carboxyl Tail-Confirming results of GST pulldown assays, the amino terminus of CNGA3 (CNGA3-N, ligand) was found by SPR to bind to the carboxyl terminus of myosin VIIa (100 nM, analyte). This interaction can occur in the absence of Ca 2ϩ (Fig. 5A, blue trace) with increases in levels of binding observed at 26.5 M Ca 2ϩ and further at 68 M Ca 2ϩ (Fig. 5A, red trace and magenta trace, respectively, indicating Ca 2ϩ dependence Fig. 5B). Again, further proof of function was achieved by reversing ligand and analyte, demonstrating Ca 2ϩ -

TABLE 1 Kinetic constants for CNGA3 interaction
Rate constants for association and dissociation were obtained from original surface plasmon resonance plots by the use of BIAevaluation software. Equilibrium binding constants were calculated from the ratio of means of the rate constants. CNGA3-N, carboxyl terminus of CDH23 ϩ68 and myosin VIIa tail his-tag fusion proteins were used in this study. Ca 2ϩ concentration was 68 M for each reaction. Kinetic constants were determined with CNGA3-N as ligand and CDH23 exon ϩ68 as analyte, CNGA3-N as ligand and myosin VIIa as analyte, and CDH23 exon ϩ68 as ligand and myosin VIIa as analyte.

CNGA3 Directly Interacts with Cadherin 23 ؉68 and Myosin VIIa
dependent interaction between CNGA3-N and myosin VIIa (Fig. 5, C and D). The interaction between CDH23 ϩ68 (shorter fragment, amino acids 3134 -3354, illustrated in Fig. 2D) and myosin VIIa, first demonstrated by Bahloul et al. (8), was confirmed and found also to be dependent on Ca 2ϩ (Fig. 6, A and B). Similar to the interaction between CNGA3-N and myosin VIIa, there is finite binding for CDH23 ϩ68 and myosin VIIa in the total absence of Ca 2ϩ (1 mM EGTA) with increases at 26.5 M Ca 2ϩ (Fig. 6A, green trace) and further increases at 68 M Ca 2ϩ (Fig. 6A, red trace). No binding was detected in negative controls with HBS-N and 1 mM EGTA buffer negative controls (Fig. 6A, cyan and magenta traces, respectively). The results were normalized by subtracting the SPR response (RU) for buffer alone and averaged from a single experiment performed in triplicate.
Kinetic analysis for CNGA3-N as the ligand interacting with myosin VIIa as analyte (1.5-100 nM, 68 M Ca 2ϩ , Fig. 5E) yielded a K D ϭ 2.3 ϫ 10 Ϫ8 (Table 1) or, for the reversal of ligand and analyte, a K D ϭ 1.2 ϫ 10 Ϫ8 (calculated from results displayed in Fig. 5F) compared with the interaction of CDH23 ϩ68 (shorter fragment, ligand) and myosin VIIa as analyte (3-100 nM, 68 M Ca 2ϩ , Fig. 6C), with a K D of 4.3 ϫ 10 Ϫ7 . It may be noted that with CNGA3-N as ligand, the K D for binding to myosin VIIa as analyte was a factor of 10ϫ smaller (10 Ϫ8 ), i.e. higher affinity, than for binding of CNGA3-N as ligand to CDH23 ϩ68 as analyte (10 Ϫ7 ) ( Table 1), potentially influencing relative formation of alternate complexes with myosin VIIa A, a SPR sensorgram shows Ca 2ϩ -dependent interaction between CNGA3-N (ligand) and myosin VIIa (100 nM, analyte) with maximum response of 130 RU at 68 M Ca 2ϩ (magenta) compared with 26.5 M Ca 2ϩ (red). A less but finite amount of binding was observed with 1 mM EGTA (blue). Negative controls included HBS-N ϩ 1 mM EGTA (green), 10 g/ml bovine serum albumin (gray), and HBS-N buffer alone (cyan). B, shown is a bar graph representation of results for A and in descending order. C, a SPR sensorgram shows a Ca 2ϩ -dependent interaction between myosin VIIa (ligand) and CNGA3-N (125 nM, analyte) with maximum response of 217 at 68 M Ca 2ϩ (blue) compared with that for 26.5 M Ca 2ϩ (red). Again, there was a smaller but finite response for 1 mM EGTA (magenta). Negative controls included HBS-N ϩ 1 mM EGTA (cyan), HBS-N buffer alone (blue), overlapping in the bottom trace. D, shown is a bar graph representation of results for C in descending order. E, shown is quantitation of CNGA3-N (ligand) and myosin VIIa (analyte) interaction by SPR. Purified myosin VIIa was immobilized on a CM5 sensor chip, and CNGA3-N was diluted in a series of concentrations at 68 M Ca 2ϩ : 100 nM (blue), 50 nM (green), 25 nM (red), 12.5 nM (magenta), 6.25 nM (cyan), 3.0 nM (gray), 1.5 nM (brown), and HBS-N buffer (black). Kinetic constants are summarized in Table 1. F, shown is quantitation of myosin VIIa (ligand) and CNGA3-N (analyte) interaction by SPR. Purified myosin VIIa was immobilized on a CM5 sensor chip, and CNGA3-N was diluted in a series of concentrations at 68 M Ca 2ϩ : 320 nM (blue), 160 nM (red), 80 nM (green), 40 nM (magenta), 20 nM (red), 10 nM (cyan). MARCH

DISCUSSION
CNG channels mediate sensory transduction in olfaction, vision, and gustation, CNGA2 and CNGA4 in primary olfactory sensory neurons, CNGA1 and CNGA3 in rod and cone photoreceptors, respectively, and CNGA3 in chemosensory receptors. In addition, CNGA3 also mediates sensory transduction in a subpopulation of primary olfactory sensory neurons. The identity of the ion channel(s) responsible for sensory transduction in cochlear/vestibular receptor hair cells is unknown. However, transcript and protein for cAMP-gated olfactory forms of CNG, namely CNGA2 and CNGA4, have been detected in mammalian cochlear hair cells (13), and CNGA3 transcript was detected in an organ of Corti cochlear subdissected fraction (14).
Expression of CNGA3 transcript in teleost saccular hair cells has been documented (3). The molecular forms of three fulllength amino acid sequences for CNGA3 expressed in teleost saccular hair cells (3), a model hair cell preparation free of supporting cells, show identity to CNGA3 expressed in human cone photoreceptors (15), trout pineal photoreceptors (AF393839), and human and rat primary olfactory sensory neurons (GenBank TM accession numbers NP_001073347 and EDL99244, respectively) (16), in contrast to the form of CNGA3 used in sensory transduction in rat gustation (NM_053495) (3). This assignment of the hair cell CNGA3 as an olfactory/cone photoreceptor CNGA3 subtype is primarily based upon amino terminus sequence obtained from a trout saccular hair cell preparation with its large numbers of hair cells as the only intact cell type. In the present investigation, it is the amino terminus of CNGA3, expressed in the rat organ of Corti and in rat cochlear outer and inner hair cells and corresponding to the mammalian olfactory/cone photoreceptor CNGA3 with high identity to trout saccular hair cell CNGA3, that was found as bait to bind to tip-link protein CDH23 ϩ68 as prey in yeast two-hybrid mating protocols. Given that CDH23 ϩ68 is specific to cochlear hair cells, the protein-protein interaction between CDH23 ϩ68 and CNGA3 speaks to hair cell-specific function.
CNGA3 in Hair Cell Stereocilia and EMILIN1-Not only was there CNGA3 transcript in teleost saccular hair cells, but CNGA3 protein was immunolocalized to stereocilia and the subcuticular plate region in both teleost vestibular hair cells and mammalian auditory hair cells, predicting a role for this ion channel in stereocilia across hair cell types (3). The carboxyl terminus of CNGA3 was found to bind elastin microfibril interface-located protein (EMILIN1) in both teleost and mammalian hair cell models, pointing to conservation of molecular function across evolution. Although EMILIN1 is considered to be an extracellular matrix protein interacting with ␤1-integrin, we demonstrated that the carboxyl terminus of mammalian EMILIN1, preceded by a predicted transmembrane region, is likely to be intracellular. It was this carboxyl sequence that was specifically responsible for the binding of EMILIN1 to CNGA3, which points to intracellular binding between EMILIN1 and CNGA3 (3).
Transcript for phosphodiesterase 6C, the phosphodiesterase regulating intracellular levels of cGMP gating cone photoreceptor CNGA3, was identified in teleost saccular hair cells (3), consistent with a cone photoreceptor-like cGMP biochemical pathway in saccular hair cells, possibly correlating with the similarity in molecular form of CNGA3 that is expressed by these FIGURE 6. SPR analysis of interactions of CDH23 ؉68 with myosin VIIa. A, shown is CDH23 ϩ68 interaction with myosin VIIa: Ca 2ϩ dependence. CDH23 ϩ68 (aa 3134 -3353, Fig. 2D) was immobilized as ligand and myosin VIIa fusion peptide (300 nM) served as analyte. The sensorgram shows maximum binding of ϳ255 RU at 68 M Ca 2ϩ (red) decreasing to 180 RU at 26.5 M Ca 2ϩ (green). Binding was reduced but finite with 1 mM EGTA (blue). HBS-N buffer and 1 mM EGTA (magenta) and HBS-N alone (cyan) served as negative controls. B, shown is a bar graph for results in A. C, shown is quantitation of CDH23 ϩ68 and myosin VIIa interaction by SPR. Binding (68 M Ca 2ϩ ) was determined for CDH23 ϩ68 (aa 3134 -3353, Fig. 2D) as ligand to myosin VIIa as analyte at 100 nM (green), 50 nM (magenta), 25 nM (red), 12.5 nM (blue), 6.25 nM (cyan), 3.0 nM (gray-green), and HBS-N buffer (gray). See Table 1 for kinetic constants. two sensory cells. EMILIN1 and phosphodiesterase 6C proteins were also immunolocalized to stereocilia of mammalian cochlear hair cells along with CNGA3, potentially reflecting a concentration of proteins required for CNGA3 action.
CNGA3 and Cadherin 23-In addition to interacting via its carboxyl terminus with EMILIN1, CNGA3 has now been found to bind via its amino terminus to CDH23 and in particular to the form of CDH23 containing the alternatively spliced exon 68 (CDH23 ϩ68). The amino terminus of rat organ of Corti CNGA3 as bait in yeast two-hybrid mating protocols, applied to a rat organ of Corti prey cDNA library, interacted with prey protein CDH23 ϩ68. These results were confirmed with yeast two-hybrid co-transformation protocols utilizing appropriate negative controls with empty vectors and the interchange of bait and prey. Pulldown assays and SPR analysis also confirmed the specificity of the interaction and its dependence on intracellular Ca 2ϩ . The amino terminus sequences of CNGA3 in human, trout hair cell, and rat organ of Corti contain calcium binding domains (e.g. DRDLDQ, aa 23-28 for rat organ of Corti sequence), representing partial EF-hand calcium binding motifs (17) present in calcium-binding proteins Cab45, EHD4, and CDH23 (Q91ZS3, NP_747540, EDL93044, respectively). The Ca 2ϩ dependence of the CNGA3 interaction with CDH23 ϩ68 may in part derive from this motif.
CDH23 is a transmembrane protein constituting the upper 2/3 of the hair cell stereociliary tip links, structures linking the stereocilia in the stereociliary array to the transducing mechanical stimulus. The tip links control gating of the mechanotransduction (MET) channels of receptor hair cells in the sensory epithelia of the inner ear (4, 18 -23). The upper part of the filamentous tip link is formed by cadherin 23 homodimers and the lower part by protocadherin 15 homodimers. The two cadherins interact at their amino termini to form the tip link filament (24 -26). CDH23 is thought to be inserted via its cytoplasmic carboxyl terminus at a lateral position on the taller stereocilia, whereas protocadherin 15 inserts, also via its carboxyl cytoplasmic terminus, into the top of the shorter stereocilia (24,27) for adjacent stereocilia, which are arranged in ascending rows in the stereociliary array of each hair cell (see Fig. 7).
CDH23 and protocadherin 15 are members of the cadherin superfamily, and both are protein targets in Usher syndrome Type I (USH1) loci (28,29). Mutations in CDH23 are the specific cause of Usher syndrome deafness USH1D (4) and are also implicated in non-syndromic deafness DFNB12 (30 -32). There are two alternatively spliced CDH23 isoforms differing with respect to the inclusion of exon 68, encoding part of the intracellular CDH23 domain (4,6,7). Messenger RNAs have been detected for both CDH23 ϩ68 and CDH23 Ϫ68 in the inner ears of adult mice (4). CDH23 ϩ68 is considered to be the hair cell-stereocilia-specific splice variant (4) and is not detected in retina (7).
Cadherin 23 Protein-Protein Interactions and Effect of the Presence or Absence of the ϩ68 Exon-It has previously been shown that CDH23 interacts with the PDZ domain-containing proteins harmonin (7,8,33), MAGI-1 (34), PIST (35), and the non-PDZ domain-containing proteins protocadherin 15 (PCDH15) (24), myosin-VIIa (8), myosin-1c (4), and EHD4 (Eps15-homology domain-containing protein 4) (36). It was the EH calcium binding domain of EHD4 that interacted with CDH23. These CDH23 binding partners were identified in yeast two-hybrid protocols and verified in pulldown assays and/or SPR analysis using in vitro expression systems and brain as a source of protein. Immunofluorescence co-localizations of proteins were established in in vitro expression systems in some instances (24,33,35,37). The individual proteins have been immunolocalized to positions on hair cell stereocilia relative to phalloidin-labeled actin, a stereocilia marker protein. Immunofluorescence co-localization of myosin VIIa, cadherin 23, and sans (a myosin VIIa-interacting protein) in hair cell stereocilia has been observed (38).
The interactions of the two PDZ domain-containing proteins harmonin and MAGI-1 with CDH23 are complex and speak to the issue of the ϩ/Ϫ68 exon. Two PDZ domains in harmonin interact with two complementary binding surfaces in the CDH23 cytoplasmic domain, one internal to the cytoplasmic domain sequence and one at the carboxyl terminus. The internal site can be disrupted by sequences encoded by the alterna-  (24 -26). PCDH15, constituting the lower 1/3 of the tip link, is inserted via its cytoplasmic carboxyl terminus at the top of the shorter stereocilium (24). CNGA3 and HCN1, representing stereociliary proteins (3,10,12), interact with CDH23 ϩ68 and PCDH15 CD3, respectively. In one configuration, as a consequence of specific Ca 2ϩ -dependent molecular interactions, HCN1 is localized to the top of the shorter stereocilium and CNGA3 to a lateral position on the next taller stereocilium. MARCH 8, 2013 • VOLUME 288 • NUMBER 10 tively spliced CDH23 ϩ68 with the outcome of 10 times more binding for the Ϫ68 than for the ϩ68 versions of CDH23 (7). The second and weaker binding site is at the carboxyl terminus. MAGI-1, a member of the membrane-associated guanylate kinase proteins, also utilizes the carboxyl-terminal PDZ binding domain of CDH23 ϩ68 for its PDZ4 domain (34) and has been proposed to replace harmonin from CDH23 via its stronger binding. MAGI-1 is known to bind F-actin-binding proteins and has been suggested to play a role in actin cytoskeleton dynamics within polarized epithelial cells (39). The significance of this carboxyl-terminal PDZ binding domain was further pointed out by the action of PIST, a Golgi-associated PDZ protein that also binds this domain, thus retaining CDH23 in the Golgi. PIST is released from this carboxyl terminus site on CDH23 ϩ68 by either MAGI-1 or harmonin (35), presumably reflecting relative binding affinities and possibly the existence of alternate protein complexes.

CNGA3 Directly Interacts with Cadherin 23 ؉68 and Myosin VIIa
Curiously, the CNGA3 amino terminus appears to incorporate not only the possibility for Ca 2ϩ -dependent binding via a Ca 2ϩ -sensitive motif but also includes a PDZ domain NGLG (aa 31-34), corresponding to a polypeptide site on piccolo (EDL99465) recognized as having identity to the GLGF loop, a PDZ-carboxylate binding motif (CDD: 29049, see Ref. 40).
CNGA3, CDH23, and Myosin VIIa-CDH23 ϩ68 also interacts with myosin VIIa, another targeted protein in Usher 1 syndrome, with mutations of myosin VIIa giving rise to USH1B. Myosin VIIa has been implicated as having a role in adaptation of hair cell (MET) (8,41) and is localized to the upper tip-link density (38). Interestingly, a ternary complex can be formed between CDH23, harmonin, and myosin VIIa thought to be involved in "applying tension forces on hair bundle links" (8), affecting mechanotransduction. In this case the binding observed between CDH23 ϩ68 and harmonin was high affinity, involving the harmonin amino terminus-PDZ1 supramodule and not requiring the CDH23 carboxyl terminus PDZ binding domain (8). The affinities between myosin VIIa and either CDH23 or harmonin A were a factor of 10 lower than that for CDH23 and harmonin A.
In our experiments the question that immediately arose was what effect alternative binding of myosin VIIa and CDH23 ϩ68 (cf. Bahloul et al. (8)) would have on the interaction between CNGA3 and CDH23 ϩ68. Pulldown assays with myosin VIIa and CDH23 ϩ68 as competitors in binding to GST-CNGA3 indicated that the myosin VIIa tail (70 kDa) did in fact interfere with binding between CNGA3 and CDH23 ϩ68, leading to the determination that myosin VIIa directly binds CNGA3 by both pulldown and SPR, and this interaction was found to be Ca 2ϩdependent. Thus, binding between the three proteins, CNGA3, CDH23 ϩ68, and myosin VIIa, is modulated by separate interactions; elimination of binding of CNGA3 to CDH 23 ϩ 68 by myosin VIIa in pulldown assays is not simply due to the competing interaction of CDH23 ϩ68 with myosin VIIa, which we have confirmed, but is also the result of a direct interaction between CNGA3 and myosin VIIa. Curiously, in these experiments the K D for CNGA3 and CDH23 ϩ68 was similar in magnitude to that for binding of myosin VIIa and CDH23 ϩ68, whereas the K D for CNGA3 and myosin VIIa was smaller, i.e. affinity higher, by a factor of 10 (Table 1), consistent with the hypothesis that myosin VIIa could substitute for CDH23 ϩ68 in binding to CNGA3 affecting formation of alternate protein complexes with CNGA3. Given that two of these proteins are involved in mechanotransduction, it may be pointed out that distinct alternate protein complexes are presumed to underlie cellular response to mechanical stimuli (discussed in Ref. 12).
Role of CNGA3 in Hair Cells; Ca 2ϩ Permeability-The CNG channels admit Ca 2ϩ in sensory and non-sensory cell types, representing a major source of Ca 2ϩ influx. All CNG channels undergo Ca 2ϩ /calmodulin adaptation; Ca 2ϩ /calmodulin acts at the amino termini of CNGA1 (rod photoreceptors) and CNGA2 (olfactory sensory neurons), effecting adaptation. For heteromeric CNG3 channels comprising ␣ and ␤ subunits, adaptation occurs as a consequence of Ca 2ϩ /calmodulin modulation of CNGB3 involving an interaction between amino and carboxyl termini (42). Both high permeability to Ca 2ϩ and the possibility for adaptation are known properties for the observed hair cell mechanosensory transduction channel. Also, the CNGA3 binding partner EMILIN1, theoretically, as a member of the elastin superfamily, could provide the elastic element, currently unidentified but predicted for hair cell mechanotransduction.
If CNGA3 were to be considered as necessary for hair cell function, there might be an expectation of deafness or vestibular deficits in individuals with CNGA3/CNGB3 mutations. Mutations of CNGA3 (on chromosome 2q11.2) (43) and CNGB3 (on chromosome 8q21.3) (44,45) are known to give rise to the molecular heterogeneous photoreceptor retinal disease of achromatopsia (with degeneration of cone photoreceptors). CNGA3 and myosin VIIa are targeted genes in Leber congenital amaurosis (46). Both of these retinal diseases include congenital nystagmus.
Hair Cell MET Channels-The ion channel composition of the hair cell MET channels is unknown, but there is evidence for two channels per stereocilium (47), with the possibility of one channel at either end of the tip link (48). The two channels would derive from one positioned at the lower end of one tip link and the second associated with the upper end of the previous tip link (see tip link insertion points in Fig. 7). Although the initial rapid onset (Ͻ10 s) and adaptation phases of MET are specific properties that need not invoke the participation of more than one type of channel (49), with CNG channels representing putative candidates (50), the possibility of more than one channel type contributing to what is measured as MET is not obviated. Van Netten et al. (51) have proposed that pairwise coupling of two hair cell transducer channels links auditory sensitivity and dynamic range. In this model two channels with different properties, one at either end of the tip link, could explain the electrophysiological characteristics of MET. The two channels would have different onset and adaptation characteristics, possibly representing different orientations of a single type of channel in shorter and taller stereocilia, respectively. The model did not preclude the possibility of two different channel types at the separate locations. Interestingly, in another study, reassembly of tip links after dissociation with BAPTA suggested a difference in timing, with CDH23 replacements (on the taller stereocilium) occurring at a later time than those for protocadherin 15, a phenomenon associated with alteration in MET adaptation (52), again possibly consistent with differential channel expression at either end of the tip link.
We previously obtained evidence that hyperpolarization-activated HCN1 binds tip-link protein protocadherin 15 CD3 (10,12) and protocadherin 15 has been localized to the top of the shorter stereocilia (24), a putative site for mechanotransduction (53). HCN1 specifically interacts with F-actin-binding filamin A, and filamin A has recently been implicated as a mediator of cellular response to mechanical stimulation (54). Ehrlicher et al. (54) determined that mechanical strain in actin networks directly differentially regulates protein-protein interactions of filamin A. A correlate to this finding would be the existence of alternate protein complexes, a model that we have confirmed for proteins expressed in hair cells (10,12). Filamin A, in immunoprecipitations of full-length proteins, binds HCN1 and tip link-protein protocadherin 15 CD3, but this protein complex does not include HCN2. HCN2 interacts with HCN1 via a conserved amino terminus region utilized in HCN channel formation, and this complex does not include protocadherin 15 CD3. HCN2 also interacts with fascin-2, and this complex includes HCN1 but not protocadherin 15 CD3.
In the study by Ehrlicher et al. (54), mechanical stimulation increased the binding of filamin A to the cytoplasmic tails of ␤-integrins. ␤1-integrin is a binding partner of EMILIN1 (55), that in turn undergoes protein-protein interactions with CNGA3 (3). Therefore, a pathway for HCN1 could extend directly to CNGA3 at one tip-link site, at the top of the shorter stereocilium, the presumed site of Ca 2ϩ entry during mechanosensory transduction. We have obtained evidence that the cGMP pathway enzyme guanylate kinase is also a protein binding partner of protocadherin 15 CD3. 3 1 Guanylate kinase is utilized to recycle GMP produced from cGMP, which is potentially required in conjunction with a cGMP-gated ion channel. The channels formed from HCN isoforms, while having the potential to contribute to the rapid component of MET (via I inst (instantaneous current coded for by HCN isoforms ) ), admit Ca 2ϩ but are not highly Ca 2ϩ -permeable as is MET. Therefore, a protein complex at the top of the shorter stereocilia incorporating highly Ca 2ϩ -permeable CNGA3 might support the property of MET having high Ca 2ϩ -permeability.
However, in addition to the above pathway considerations, the results from the present work are consistent with a second site for CNGA3 interaction, namely, at the upper tip-link CDH23 insert (Fig. 7). CNGA3 is not voltage-dependent; heterologously expressed human CNGA3 channel is activated by cGMP, and the cGMP-elicited current is enhanced by ATP and inhibited by phosphatidylinositol (3,4,5)-triphosphate (PIP 3 ) (56). CNGA3/CNGB3 can undergo adaptation via a Ca 2ϩ -dependent mechanism, also observed for MET. One possible version of a two-channel model would include CNGA3 interacting with CDH23 at the upper tip-link density combined with HCN1 binding to protocadherin 15 CD3 at the lower tip-link density (Fig. 7). A further role for the described protein-protein interactions of CNGA3, CDH23, and myosin VIIa may exist in protein transport as modulated by Ca 2ϩ (57), with all three proteins localized to hair cell subcuticular plate sites.
In conclusion, cGMP-gated CNGA3 binds tip-link protein CDH23 ϩ68 and myosin VIIa, known components of the mechanotransduction apparatus (31). Based upon the premise that hair cell stereocilia tip-link proteins are closely coupled with the mechanotransduction channel(s), we suggest the results of this investigation are consistent with the possibility that CNGA3 participates in hair-cell MET.