Direct Interaction of Otoferlin with Syntaxin 1A, SNAP-25, and the L-type Voltage-gated Calcium Channel CaV1.3*

The molecular mechanisms underlying synaptic exocytosis in the hair cell, the auditory and vestibular receptor cell, are not well understood. Otoferlin, a C2 domain-containing Ca2+-binding protein, has been implicated as having a role in vesicular release. Mutations in the OTOF gene cause nonsyndromic deafness in humans, and OTOF knock-out mice are deaf. In the present study, we generated otoferlin fusion proteins containing two of the same amino acid substitutions detected in DFNB9 patients (P1825A in C2F and L1011P in C2D). The native otoferlin C2F domain bound syntaxin 1A and SNAP-25 in a Ca2+-dependent manner (with optimal 61 μm free Ca2+ required for binding). These interactions were greatly diminished for C2F with the P1825A mutation, possibly because of a reduction in tertiary structural change, induced by Ca2+, for the mutated C2F compared with the native C2F. The otoferlin C2D domain also bound syntaxin 1A, but with weaker affinity (Kd = 1.7 × 10–5 m) than for the C2F interaction (Kd = 2.6 × 10–9 m). In contrast, it was the otoferlin C2D domain that bound the Cav1.3 II-III loop, in a Ca2+-dependent manner. The L1011P mutation in C2D rendered this binding insensitive to Ca2+ and considerably diminished. Overall, we demonstrated that otoferlin interacts with two main target-SNARE proteins of the hair-cell synaptic complex, syntaxin 1A and SNAP-25, as well as the calcium channel, with the otoferlin C2F and C2D domains of central importance for binding. Because mutations in the otoferlin C2 domains that cause deafness in humans impair the ability of otoferlin to bind syntaxin, SNAP-25, and the Cav1.3 calcium channel, it is these interactions that may mediate regulation by otoferlin of hair cell synaptic exocytosis critical to inner ear hair cell function.

Calcium is a key regulator of synaptic vesicle fusion (reviewed in Ref. 1). In mechanosensory hair cells, calcium microdomains (2) and possibly nanodomains (3) are formed when voltage-gated calcium channels open upon depolarization. Calcium at these sites is thought to activate protein interactions, leading to vesicle fusion. Some of the key players in this process are the target-SNARE 2 proteins, syntaxin 1A and SNAP-25, and the vesicle-SNARE, synaptobrevin (4). Vesicle-SNARE synaptotagmin 1 plays a crucial role as a calcium sensor at the neuronal synapse, modulating calcium channels and vesicle release by a Ca 2ϩ -dependent interaction with other SNARE proteins in the presence of lipid molecules (4 -6). However, in vertebrate mechanosensory hair cells, synaptotagmin 1 is not detected (7). Instead, fast neurotransmitter release in auditory and vestibular hair cells, facilitated largely by an L-type voltagegated calcium channel, Ca v 1.3 (8,9), is thought to be modulated by a newly discovered protein, otoferlin, acting as the Ca 2ϩ sensor and vesicle-binding protein. When mutated, otoferlin causes DFNB9 nonsyndromic deafness (10). Gene sequences of different deaf families show that the OTOF gene can undergo mutation at multiple locations (11)(12)(13). Recently, it has been demonstrated that otoferlin is necessary for synaptic exocytosis from hair cells (14). Further, an engineered mutation in the C2B domain of otoferlin has been shown to cause deafness in mice (15). However, the precise function of otoferlin as a synaptic protein is not well understood.
Specific mutations in the otoferlin C2F (P1825A) or C2D (L1011P) domains in humans have been documented to cause DFNB9 deafness (11,12). Previous studies suggested that a region of otoferlin containing all three C2 domains, D, E, and F, binds directly to the t-SNARE molecules syntaxin 1A and SNAP-25 in response to an increase in Ca 2ϩ concentration (14). However, it is not understood how a single amino acid substitution in one domain of otoferlin, such as C2F (11) or C2D (12), might independently lead to deafness. Here, we examine the role of otoferlin as a Ca 2ϩ sensor as well as a facilitator of vesicle fusion, as indicated by protein-protein interactions and their [Ca 2ϩ ] dependence.
Expression of Hexahistidine Fusion Proteins-pRSET vectors containing the desired sequences were used to transform E. coli BL21 (DE3) cells. Clones were selected, and expression of fusion proteins was induced by adding 1 mM isopropyl ␤-D-1thiogalactopyranoside (IPTG) in LB culture medium. The cultures were incubated for 5 h at 37°C. The cells were harvested, lysed in 10 mM phosphate buffer, 8 M urea (pH 7.4), 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 protein expression induced by IPTG. The cells were harvested, lysed, and centrifuged, and the clear supernatant passed through a nickel-nitrilotriacetic acid affinity spin column (Qiagen). The columns were washed with wash buffer (10 mM phosphate, 8 M urea, pH 6.3) three times to remove nonspecific components, and the affinity-bound proteins were eluted with elution buffer (10 mM phosphate, 8 M urea, pH 4.5). For thorough removal of urea, the samples were diluted in an equal volume of PBST buffer and dialyzed extensively against PBST containing 1ϫ protease inhibitors, with five or six buffer changes for 24 h at 4°C and constant stirring. At the end of the dialysis, the proteins were collected from the dialysis tubes, and protein concentration was determined with the Qubit fluorescent system (Invitrogen). Purity of the proteins was ascertained by SDS-PAGE with Coomassie Blue staining and Western immunoblotting using anti-His monoclonal antibody.
In an alternate procedure not involving urea, cells induced with IPTG were lysed by sonication, followed by centrifugation. The clear supernatant was mixed with 0.5 ml of Talon beads (Clontech). The beads were sedimented and washed, and the bound fusion proteins were eluted by the addition of elution buffer, followed by mixing and centrifugation. The supernatant containing the fusion proteins was then exhaustively dialyzed against cold PBST buffer containing 1ϫ protease inhibitors.
Cloning of Syntaxin 1A and Ca v 1.3 II-III Loop in pGEX-6P-1 for Expression of GST Fusion Proteins-GST fusion proteins were prepared for pull-down assay. PCR primers for syntaxin 1A and Ca v 1.3 II-III loop containing the desired restriction sites were used in PCRs to clone syntaxin and the calcium channel domain in pGEX-6P-1 vector.
Expression of GST Fusion Proteins-pGEX-6P-1 vectors containing syntaxin 1A or the Ca v 1.3 II-III loop were utilized to transform E. coli BL21 (DE3) cells. The clones were selected for expression of the fusion protein by IPTG induction. Bacterial cell lysates were either used directly in pull-down assays, or the fusion products were affinity-purified using Glutathione-Sepharose beads (GE Healthcare).
Yeast Two-hybrid Assays-Appropriate prey and bait vectors were used to co-transform yeast strain AH109 and were incubated in a low stringency medium, SD/ϪTrp/ϪLeu, to select the double transformants. The colonies that grew were streaked onto a higher stringency medium, SD/ϪTrp/ϪLeu/ϪHis/ ϪAde, containing 3 mM 3-amino-1,2,4-triazole (3-AT) and grown for at least 5 days. Negative controls comprising empty bait or prey vectors were paired with vectors for their counterpart binding partners.
Pull-down Assay-Pull-down assays were performed using GST fusion proteins (GST-syntaxin 1A and GST-Ca v 1.3 II-III loop) and hexahistidine-tagged fusion proteins (otoferlin C2 domains). GST conjugates were mixed with hexahistidinetagged C2 domains in binding buffer (PBST buffer, pH 7.4, 1ϫ protease inhibitor mixture) for 2 h at room temperature. Glutathione-Sepharose beads (30 l of a 50% slurry) were added to the reaction mix for 1 h at room temperature, followed by centrifugation for 2 min at 1,000 ϫ g. The beads were washed five 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.
Surface Plasmon Resonance Analysis-SPR was performed with a Biacore 3000 instrument. Affinity-purified fusion proteins were immobilized on a CM5 chip (research grade; Biacore, Piscataway, NJ) by an amine-coupling reaction (16). Surface tests (17) were performed on each immobilized protein (ligand), with different concentrations of purified otoferlin C2 domain fusion protein (analyte). The reference surface was blocked with ethanolamine and thus contained no ligand. The response was recorded as the ligand RU minus the reference RU. Once appropriate binding and regeneration conditions were established, binding studies were performed.
SPR Kinetic Measurements of Otoferlin Binding-To determine the rate and equilibrium binding constants of the interaction of otoferlin C2 domains with syntaxin 1A, SNAP-25, and the Ca v 1.3 II-III loop, the fusion proteins were immobilized as ligands on a CM5 chip. Otoferlin C2D and C2F domains (analytes) were diluted to a series of concentrations in HBS-P buffer, with added 61 M Ca 2ϩ for optimal binding. Kinetic values were determined using BIAevaluation software (Biacore), and the data were fitted with the model showing closest match (17). A 1:1 Langmuir binding model was generally selected. However, when there was a mass transfer limitation (18) in the binding, 1:1 binding with mass transfer was chosen. The "residuals" (19) were checked, and the values with the lowest 2 values were selected. Residuals reflect the deviation of the actual binding from that of the model chosen and were in the range of Ϯ2 RU. Generally, we employed the global fitting method (20), in which all the sensorgrams representing the different analyte concentrations were fitted simultaneously with the wide window of association and dissociation phases. Individual concentration curves were also evaluated to confirm the fitting data.
Optical Analyses-For fluorometric analysis, 50 nM each of C2F and C2Fm in PBS (pH 7.4) were analyzed in a FluroMax-2 fluorometer (Jobin Yvon Inc., Edison, NJ) with excitation at 282 nm, and the emission was recorded from 295 to 450 nm.
CD spectra were recorded with a Chirascan CD spectrometer (Applied Photophysics Ltd., Leatherhead, Surrey, UK) at 200 -260 nm, using a 1-cm path length cell. The samples were diluted in 1ϫ PBS to yield a final concentration of 20 M, and the PBS blank was subtracted from the sample.
Determination of Ca 2ϩ Concentrations-Ca 2ϩ concentrations of all solutions were directly checked by mass spectral analysis, because of the possibility of background Ca 2ϩ in reagents. Thus the Ca 2ϩ values reported in the present experiments are actual, assayed values.

RESULTS
C2 Domains of Otoferlin-Many isoforms of otoferlin, resulting from single amino acid substitutions and truncations caused by differential splicing (12), are seen in DFNB9 patients ( Fig. 1). A short otoferlin isoform in humans contains only C2D, C2E, C2F, and the transmembrane domain. It is not yet clear whether the short form or the long form of the otoferlin transcript is expressed in the human cochlea, although there are indications that the long forms are functionally important (13).
Otoferlin C2F Domain Directly Binds to Syntaxin 1A-In pilot studies (not illustrated), mouse otoferlin domains C2A, C2B, C2C, C2D, C2E, and C2F, cloned as bait in pGBKT7 vector, were tested in yeast strain AH 109 versus syntaxin 1A fusion protein, which was cloned as prey in pGDAT7 vector. In SD/ϪTrp/ϪHis/ϪAde/ϪLeu medium containing 2 mM 3-AT, colonies that appeared represented syntaxin 1A interaction with C2A, C2B, and C2F. Few colonies were present when C2D and C2E were used as bait. When C2A, C2B, and C2F colonies were screened in a higher stringency medium containing 3 mM 3-AT, colonies appeared only for C2F ( Fig. 2A) and C2A (not illustrated). There were no colonies on control plates where either bait or prey was substituted with empty vectors.
Otoferlin Domains Interact with Syntaxin 1A in Pull-down Assays-Pilot pull-down assays using GST-syntaxin 1A fusion protein and purified otoferlin-C2 domain His tag fusion proteins showed that syntaxin 1A binds C2A, C2B, and C2E (not illustrated) as well as C2F (Fig. 2B, lane 3). However, it is the C2F interaction that is strictly Ca 2ϩ -dependent (see below), unlike the interaction for C2A, C2B, and C2E where Ca 2ϩ is not absolutely required.
SPR Analysis of Otoferlin Domain-Syntaxin 1A Interactions-Interaction of the six otoferlin C2 domains with syntaxin 1A was also examined with SPR, utilizing affinity-purified fusion proteins of the C2 domains as analytes and affinity-purified syntaxin 1A fusion protein immobilized on a CM5 sensor chip as ligand. The purified C2 domains were diluted to 200 nM concentration. The C2F domain distinctly bound syntaxin 1A  (residues 1964 -1986). The large arrowheads indicate the positions of two mutations that were the focus of the present study, in domains C2D (1011 human; 1010 mouse) and C2F (1825). The four thick arrows relate to other single amino acid substitutions known to cause deafness (left to right): P490Q, I515T, R1939Q, and P1987R (12,13). The thin arrows correspond to points of truncation: Arg 237 , Glu 551 , Arg 708 , Gln 829 , Trp 1425 , and Tyr 1497 (12). Mouse otoferlin sequence (NM_031875) is 98% identical to human sequence and exactly the same length (1997 amino acids).
in SPR analysis (Fig. 2C). There was no binding when bovine serum albumin was used as a negative control (Fig. 2C, Blank).
Ca 2ϩ Sensitivity of C2F Interaction with Syntaxin 1A-We examined the binding of the purified C2F fusion protein to syntaxin 1A at various Ca 2ϩ concentrations. No binding was observed when 200 nM C2F was analyzed in HBS-P buffer containing 3 mM EGTA (Fig. 2D, bar 1). However, a dramatic increase was observed when the C2F binding was analyzed in buffer containing 30 M free Ca 2ϩ (Fig. 2D, bar 2). With each increment in free Ca 2ϩ , there was a steady increase in binding that peaked at 61 M Ca 2ϩ (bar 7). However, there was a drop of more than 50% in the binding when the Ca 2ϩ concentration was increased to 95 M (bar 8), and there was no substantial decrease in binding when the level of Ca 2ϩ was further increased to 199 and 371 M (bars 9 and 10). 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 the C2F domain of otoferlin directly interacts with syntaxin 1A; this interaction requires free Ca 2ϩ optimally in the range of 30 -61 M, and there is no interaction if Ca 2ϩ is absent. The C2F domain is 98 -100% conserved among the mammalian species shown in Table 1 and would be expected to interact in a similar manner in other species, including human.
To test whether C2A binding was sensitive to Ca 2ϩ , the C2A domain was diluted to 200 nM in buffer containing either 44 M free Ca 2ϩ or 3 mM EDTA and analyzed by SPR (Fig. 2E). There was little difference between the traces. Fig. 2F shows quantitatively that there was no significant difference in binding when free Ca 2ϩ was absent (with EGTA, bar 1) versus present (bar 2). Thus it appears that the C2A interaction is insensitive to Ca 2ϩ , in contrast to the calcium sensitivity of the C2F interaction.
C2B and C2E domains, not requiring Ca 2ϩ for binding to syntaxin 1A in pull-down assays (above), exhibited relatively low SPR binding values, 35 and 14 RU, respectively (not illustrated).
A P1825A Substitution in the C2F Domain Abolishes C2F Ca 2ϩ Binding Affinity-To explore how a single amino acid substitution in the C2F domain might correlate with nonsyndromic deafness in humans (11), we produced the same amino acid substitution (P1825A) in mouse that is known to cause deafness in humans, by oligonucleotide-directed PCR-based alteration in C2F. The purified mutant C2Fm fusion protein showed no interaction with syntaxin 1A in SPR experiments (Fig. 3, A and B), in contrast to the prominent binding of native C2F to syntaxin 1A. Pull-down assays using GST-syntaxin 1A and C2F native and C2Fm mutated forms also demonstrated that the C2Fm does not interact with syntaxin 1A (Fig. 3C). The position of the P1825A mutation in C2F lies between strands 6 and 7 (␤ sheet) of the secondary structure (Fig. 3D), with proline demarking a probable bend in the polypeptide chain for native C2F.

TABLE 1 Percent identity of C2 domains across vertebrates
Amino acid sequences from vertebrate classes were analyzed to identify the C2 domains using the Simple Modular Architecture Research Tool (SMART; EMBL, Heidelberg, Germany). The predicted C2 domains were aligned against the corresponding predicted mouse C2 domains. The amino acid identity for each was determined by the BLAST program (NCBI, Bethesda, MD). The percent identity of each C2 domain is referenced to the corresponding mouse sequence. GenBank TM accession numbers (protein) are listed in parentheses.

JOURNAL OF BIOLOGICAL CHEMISTRY 1367
Calcium Binding of the C2F Domain-The intrinsic fluorescence of the C2F domain (contributed by Trp and other aromatic amino acids; Fig. 3D) was markedly elevated with Ca 2ϩ for native but not the mutated form. Panels E (mutated form) and F (native form) of Fig. 3 show emission spectra in the presence of 61 M free Ca 2ϩ (solid traces) and emission in the absence of Ca 2ϩ (1 mM EGTA) by the dashed traces. Fig. 3E shows that the C2Fm solution does not exhibit any increase in fluorescence in the presence of 61 M free Ca 2ϩ compared with the absence of free Ca 2ϩ . However, for the native C2F domain, emission is clearly increased by the addition of Ca 2ϩ to a 61 M free Ca 2ϩ concentration (Fig. 3F, solid trace). This effect was reversed by 2 mM EGTA (dashed trace).
To test whether the native state of C2 fusion protein was maintained after purification with urea-containing buffers followed by dialysis (see "Experimental Procedures"), we analyzed with circular dichroism the C2F domain fusion protein prepared under denaturing versus nondenaturing conditions (Fig.  3G). The CD spectra proved similar, suggesting that fusion proteins such as C2F extracted with urea, after exhaustive dialysis against nondenaturing buffers, were properly refolded in the present studies. The spectra show a broad minimum at 206 -207 nm as well as a tendency for a minimum in the 221-222-nm region. The spectral trace was flat above 240 nm and increased below 205 nm (Fig. 3G).
Ca 2ϩ Dependence of Otoferlin C2F Binding to SNAP-25-SPR binding assays involving affinity-purified SNAP-25 as ligand and affinity-purified native and mutated C2F domains as analytes showed that the native C2F binds SNAP-25 (Fig. 4A,  upper trace), whereas the mutated, C2Fm does not (lower trace). Fig. 4B illustrates that the SNAP-25-C2F binding is sensitive to Ca 2ϩ . There is minimal binding when there is no free  (21) program. Otoferlin C2F exhibits type II C2 topology (30), with seven strands of predicted ␤ structure (strands 1-7, arrows and lines above sequence), one strand of predicted helical structure (strand 8), and intervening loops of extended (random coil) structure (21). Tryptophan residues that might increase intrinsic fluorescence of C2F when folded correctly are underlined. Letters at Pro-Ala mutation site (thought to reduce Ca 2ϩ affinity of otoferlin C2F) are in bold. Four carboxylic amino acid residues, positioned to possibly coordinate Ca 2ϩ binding (22), are shaded, and other nearby carboxylic residues are marked with dots. Predictions for turns in the secondary structure are marked by ϳ and bends byˆ. The lowest line shows consensus amino acids with identity of 50% or more, as derived from sequence alignment of 65 different C2 domains (22), with hyphens indicating nonconsensus residues. E, normalized fluorescence spectra for affinity-purified C2Fm domain. The Pro-Ala mutated C2Fm domain protein (20 M concentration) is insensitive to Ca 2ϩ , with no difference in the fluorescence spectrum between conditions of 61 M Ca 2ϩ (solid trace) and EGTA (dashed trace). F, normalized fluorescence spectra for affinity-purified native C2F domain. When 61 M Ca 2ϩ is added to the native C2F protein solution (20 M), the fluorescence is considerably elevated (solid trace), compared with lowered fluorescence for the same solution with added 2 mM EGTA (dashed trace). G, circular dichroism spectral analysis of the C2F domain fusion proteins purified with urea buffer followed by dialysis against PBST buffer (thick trace; see text) versus CD spectrum for C2F purified under nondenaturing conditions, in PBST (dashed trace). Ca 2ϩ (bar 1). However, when the free Ca 2ϩ increases to 30 M, there is an increase in binding (bar 2), and a further, almost 3-fold increase in binding when the free Ca 2ϩ increases to 33 M (bar 3). An additional increase in binding occurs, until the free Ca 2ϩ concentration reaches 61 M (bar 7). Then, when the free Ca 2ϩ is increased further (to 95 M), the binding drops by ϳ50% and continues to decrease up to a free Ca 2ϩ concentration of 371 M (bar 10).
Otoferlin C2D Binds to the Voltage-gated Calcium Channel's Ca v 1.3 II-III Loop-In pilot studies, otoferlin C2 domains A-F were constructed in pGBKT7 vector and with the Ca v 1.3 II-III loop prey construct were used to co-transform yeast strain AH109. In high stringency selection medium containing 3 mM 3-AT, the C2D bait, paired with Ca v 1.3 II-III loop prey, exhibited growth, whereas the C2D bait with the empty prey vector did not (Fig. 5A). In general, the number of colonies that developed on the selection plate indicated the strength of the interaction. Although we also detected interaction of Ca v 1.3 with the C2A, C2B, and C2F domains, the latter interactions were clearly weaker (not shown).
Otoferlin C2D-Ca v 1.3 II-III loop interaction was also demonstrated in pull-down assays, where native C2D bound the GST-Ca v 1.3 II-III loop, yielding a prominent band on Western blot (Fig. 5B, lane 4). In contrast, the mutated C2Dm did not bind the GST-Ca v 1.3 II-III loop (Fig. 5B, lane 3).
The affinity-purified fusion proteins of the otoferlin C2 domains were further analyzed for interaction with the purified Ca v 1.3 II-III loop using SPR. C2 domains (250 nM) in HBS-EP binding buffer were employed as analytes. C2D showed the maximum response (Fig. 5C, upper trace), whereas other C2 domains (not illustrated) yielded much lower interaction responses; C2F produced 60, C2B 35, C2A 30, and C2C 10 RU, whereas C2E did not bind. There was no response when 1 mM bovine serum albumin was used as a negative control. Using a mouse mutant C2Dm fusion protein with an amino acid substitution (L1011P; position 1010 in mouse) identical to that in DFNB9 human deafness (12), we analyzed binding to the Ca v 1.3 II-III loop. With SPR, we detected that the C2Dm binding was reduced 2-3-fold (Fig. 5C, lower trace), compared with the native C2D.
Otoferlin C2D-Ca v 1.3 II-III Loop Interaction Is Calciumdependent-With SPR analysis, we detected that the otoferlin C2D-Ca v 1.3 interaction occurs in the presence of free Ca 2ϩ . An increase in free Ca 2ϩ enhances this interaction, whereas Ca 2ϩ chelation (with EGTA) diminishes it. We immobilized the affinitypurified Ca v 1.3 II-III loop on a CM5 sensor chip and injected C2D analyte (250 nM in HBS-P buffer) containing a series of free Ca 2ϩ concentrations (0 -371 M; Fig. 5D, dotted bars). There was no substantial binding with zero free Ca 2ϩ (Fig. 5D, dotted bar 1). However, a significant increase in the binding occurred when the free Ca 2ϩ was increased to 61 M (dotted bar 3), indicating that Ca 2ϩ regulates the interaction under in vitro conditions. A decrease in binding occurred at the higher Ca 2ϩ concentrations (dotted bars 4 -6). In contrast, we observed that for mutant C2Dm binding to the Ca v 1.3 II-III loop, even though free Ca 2ϩ is required, there was no substantial increase in binding with an increase in free Ca 2ϩ (Fig. 5D, clear bars). To summarize, the Ca v 1.3 channel interacts with otoferlin via the C2D domain (not necessarily exclusively; see "Discussion"), with native C2D requiring an optimum of 61 M free Ca 2ϩ .
Quantitative Syntaxin 1A-C2F Binding Kinetics-Using syntaxin 1A as ligand (Fig. 6A), we analyzed the binding of otoferlin domains C2F (and C2D; not illustrated) at 0, 25, 50, 75, and 100 nM concentrations, with SPR. Fig. 6A shows the set of data for varying concentrations of C2F samples containing 61 M Ca 2ϩ . There was no mass transfer limitation (17) in the interaction, as determined by analyzing 40 nM C2F at several flow rates (5, 15, and 75 l/min). Using BIAevaluation software (GE Healthcare), the curves for the above concentrations of C2F (and C2D) were globally fit to a 1:1 Langmuir binding model to determine the rate constants and equilibrium constants. For global fitting we achieved a 2 value of 2.1, whereas for individual curves we obtained values less than 1.0. (low 2 values indicate closeness of the actual values to the model chosen.) The equilibrium dissociation constants for syntaxin 1A-C2F versus syntaxin 1A-C2D binding ( Table 2; data from individual curves) suggested weaker interaction for the latter binding couple. A similar experiment with 1 mM EGTA in the binding buffer and the same set of C2F concentrations as above resulted in no binding of C2F to syntaxin 1A.
SNAP-25-C2F Binding Kinetics-Using SNAP-25 as ligand in SPR analyses, we also examined the binding for 0, 40, 80, 160, and 320 nM concentrations of otoferlin C2F containing 61 M Ca 2ϩ (Fig. 6B). Again, there was no significant mass transfer limitation. Kinetic constants were determined as for Fig. 6A. With global fitting, we achieved a 2 value of 6.8, whereas for individual curves we obtained values less than 3.0. The equilibrium constant for dissociation of the SNAP-25-C2F interaction was determined to be 1.7 ϫ 10 Ϫ7 M ( Table 2).
Ca v 1.3-C2D Binding Kinetics-The II-III loop of the L-type voltage-gated calcium channel Ca v 1.3 was used as the ligand in further SPR experiments to determine the binding kinetics for the otoferlin C2D domain (Fig. 6C). Kinetic constants were evaluated as above. A value of 1.4 ϫ 10 Ϫ7 M for the dissociation equilibrium constant for Ca v 1.3-C2D domain interaction was obtained ( Table 2).

DISCUSSION
Otoferlin as a SNARE-binding Ca 2ϩ Sensor-C2 domains form calcium-binding modules in diverse groups of proteins, such as those represented by synaptotagmin 1 and protein kinase C (reviewed in Ref. 22). Otoferlin, expressed in brain as well as in mechanosensory hair cells (23), contains six C2 domains and is hypothesized to be the main Ca 2ϩ sensor of the hair cell synapse. Mouse knock-outs for the OTOF gene are deaf and show no exocytosis from the inner hair cells (14). In human, a number of OTOF gene mutations cause either truncation of the protein (10,13) or single amino acid substitutions in C2F (11), C2D (12), or C2C (24) (Fig. 1). In the present study, we chose two functionally important amino acid substitutions, one in C2F and one in C2D, to investigate how these domains may individually interact with the proteins involved in synaptic-vesicle fusion in the hair cell.
Ca 2ϩ Dependence of Otoferlin C2 Domain Interactions-Our results demonstrated that otoferlin interacts with the t-SNARE synaptic proteins syntaxin 1A and SNAP-25 and that this interaction requires free Ca 2ϩ , as is the case for the Ca 2ϩ sensor synaptotagmin 1, which mediates neuronal exocytosis (25). However, otoferlin may also interact with syntaxin 1A in a Ca 2ϩ -independent manner for certain C2 domains. For example, our binding results involving the C2A domain and syntaxin 1A show that Ca 2ϩ is not necessary for the C2A interaction, suggesting that C2A is unlikely to contribute directly to the Ca 2ϩ -sensitive exocytosis at the mammalian hair cell synapse. Moreover, our work involving the C2F domain interaction with syntaxin 1A or SNAP-25 shows that optimal binding is attained when the free Ca 2ϩ concentration reaches 61 M but that binding decreases when the Ca 2ϩ concentration is increased further. The latter finding differs from the results of an earlier report (14) where the strength of binding of syntaxin 1A to an otoferlin polypeptide containing the complete stretch of C2D,  Table 2). A, syntaxin 1A interaction with the otoferlin C2F domain. Purified syntaxin 1A was immobilized on a CM5 sensor chip, and C2F was diluted in a series of concentrations (0, 25, 50, 75, and 100 nM, bottom to top). The curves were selected and fit using a global fitting method and a 1:1 Langmuir binding model (not illustrated) with drifting base line (17). Deviation of fit from model, along the curves, is indicated by the residual (RES) plots for association (0 -300 s) and dissociation (300 -650 s). B, SNAP-25 interaction with the otoferlin C2F domain. The curves represent C2F concentrations of 0, 40, 80, 160, and 320 nM, evaluated as in A. C, Ca v 1.3 II-III loop interaction with the otoferlin C2D domain. The purified Ca v 1.3 II-III loop fusion protein was immobilized as before, and C2D analyte was tested for interaction at 0, 75, 150, 300, and 600 nM concentrations. C2E, and C2F domains regularly increased up to 1 mM Ca 2ϩ , with a calculated half-maximal effective concentration of 80 M. This apparent discrepancy might be explained by the fact that we examined solely the otoferlin C2F domain, whereas the previous study (14) employed a polypeptide containing the three contiguous otoferlin domains. Interestingly, Ca 2ϩ -triggered exocytosis in permeabilized PC12 cells expressing synaptotagmin 1 shows a bell-shaped curve, with peak exocytosis at 10 M Ca 2ϩ (26), thus manifesting a response pattern similar to ours. It should also be noted that different synaptotagmin isoforms show different affinity for Ca 2ϩ , as determined from their phospholipid binding as well as their Ca 2ϩ -triggered exocytosis (26). In hair cells, Ca 2ϩ -dependent exocytosis saturates beyond 50 M Ca 2ϩ and shows no detectable exocytosis below 7 M Ca 2ϩ (27), which would be in accord with our Ca 2ϩ dependence results for binding between C2F and syntaxin 1A and SNAP-25.
The mechanism for the decreased binding we observe at the higher Ca 2ϩ concentrations is currently unknown but may be due to electrostatic properties of divalent cations such as Ca 2ϩ (28) that can change the molecular electrostatic environment, leading to the reduced binding. C2F as the Main Otoferlin C2 Domain Interacting with Vesicular Proteins-We found that the P1825A mutation in the C2F domain almost completely abolished C2F interaction with the SNARE proteins examined in the current study. Prediction of secondary structure for C2F shows that the proline residue undergoing mutation is situated between ␤ strands 6 and 7 ( Fig.  3D; strands numbered according to C2 type II topology) (22,29,30), a location that is near the presumptive Ca 2ϩ -binding region in the molecule.
Although the actual residues responsible for binding Ca 2ϩ in many C2 domains, such as C2F, cannot be known with certainty, because of the paucity of crystallographic data, the four acidic residues highlighted in the otoferlin C2F sequence (Fig.  3D) represent a reasonable assignment of the coordination sites, reflecting the spacing of amino acids suggested by a consensus line-up of 65 different C2 sequences (22). One-to-two Ca 2ϩ ions are thought to bind at the edge of a ␤-sandwich, in a concave depression formed by the two extended loops between ␤ strands 1 and 2 and ␤ strands 5 and 6 (30). Fig. 3 (E and F) illustrate the behavior of tryptophan fluorescence spectra for purified C2Fm and native C2F fusion proteins in the presence and absence of Ca 2ϩ . Significantly, the mutated C2Fm domain (Fig. 3E) is insensitive to Ca 2ϩ , compared with the native C2F (Fig. 3F). Calcium-binding aspartate and glutamate residues (Fig. 3D, shaded) in the C2 domains are separated from one another in the C2 primary structure. Calcium enhancement of tryptophan fluorescence is possible only if the C2 domain folds correctly, positioning the tryptophan (underlined) and carboxylic residues located in the different loops. Proper clustering of tryptophan-containing loops increases the tryptophan fluorescence. Our results for the mutated C2Fm show that even though there is tryptophan fluorescence, the addition of calcium does not increase the fluorescence intensity (Fig. 3E), suggesting that the C2Fm does not fold properly to bind Ca 2ϩ and thus change the tertiary structure appropriately. In addition to the C2F domain, the C2D domain showed Ca 2ϩ -dependent enhancement of fluorescence in the presence of Ca 2ϩ (not illustrated).
The fact that Ca 2ϩ -dependent enhancement of the tryptophan-based fluorescence clearly occurred for the purified fusion proteins in the current investigation, in addition to the occurrence of overlapping CD spectra corresponding to the different isolation methods (Fig. 3G), marshals evidence that the proteins were able to refold to native configuration during our isolation procedures. The CD spectra of Fig. 3G are, in fact, consistent with the presence of C2F secondary structure, as predicted in Fig. 3D. The Fig. 3G plot can be interpreted as a sum of spectra for ␤ sheet and ␣ helix (31). There is clearly no shift of minimum to values below 200 nm that would indicate denaturation (32).
Although not examined, it is possible that an additional degree of molecular organization would be introduced by the interaction of otoferlin C2 domains with the synaptic complex proteins. The phenomenon of induced secondary structural order (33,34) may be relevant to the association of C2 domains with their binding partners, suggesting future studies.
It can be inferred that the C2F domain is the main calciumsensing module within otoferlin that directly interacts with SNARE proteins. Even though there might be cooperativity among the C2 domains, each C2 domain may have a specific and independent role in the exocytotic process. It should also be noted that the in vitro studies reported here do not consider the complexity of the lipid environment, which has been shown to support Ca 2ϩ dependence of exocytosis in the case of synaptotagmin 1 (6).
C2 Domain Binding Affinity-Another important factor for exocytosis could be the differential C2 domain binding affinity for the SNARE proteins. Each C2 domain of otoferlin may have a different affinity, in terms of rate as well as equilibrium constant, dictating which C2 domain or domains might bind more avidly and rapidly to initiate Ca 2ϩ -triggered vesicle docking and exocytosis. For example, in our study, we compared the C2D and C2F affinity for syntaxin 1A and found that the C2F binds syntaxin more strongly than does C2D (Table 2).
C2D Domain Interaction with the L-type Calcium Channel Ca v 1.3-It is known that synaptotagmin 1, the presumed calcium-sensing counterpart of otoferlin in neurons, directly interacts with and modulates voltage-gated calcium channels (35). In the present work, we found that otoferlin through its C2D domain binds to the II-III cytoplasmic loop of Ca v 1.3, which is the major L-type calcium channel in mammalian hair cells (9). We also found that this interaction is regulated by Ca 2ϩ and that a maximum of 61 M Ca 2ϩ is required for optimal binding under in vitro conditions. Moreover, a mutation (L1011P) that causes deafness in humans (12) diminishes the C2D binding to Ca v 1.3 almost 3-fold, and the C2D mutant is insensitive to Ca 2ϩ .
Although our evidence suggests that, of the otoferlin C2 domains, C2D binds the Cav1.3 II-III loop with the highest affinity, other C2 domains may also be involved. It is beyond the scope of the present investigation to measure all interactions (most of them significantly weaker) of the otoferlin domains, Ca 2ϩ -dependent and -independent, with the calcium channel and SNARE proteins. Several published studies describe Ca 2ϩ -dependent and independent binding of synaptotagmin to syntaxin. Kee and Scheller (36) showed that synaptotagmin binding to syntaxin 1A increases as the calcium concentration increases. The latter paper also describes changes in the calciumdependent binding when portions of the synaptotagmin protein are deleted. One such syntaxin mutant, Syt1-5, showed binding in the absence of calcium, as well as a decrease in binding in response to an increase in calcium concentration.
In summary, our study shows that individual otoferlin C2 domains may act independently of each other in the exocytotic process. Otoferlin, through C2 domains, binds two key SNARE synaptic proteins of hair cells, syntaxin 1A and SNAP-25. Otoferlin directly interacts with syntaxin 1A via the C2F domain. A mutation in C2F diminishes its Ca 2ϩ binding ability as well as its interaction with syntaxin 1A. SNAP-25, another t-SNARE protein, also interacts with the C2F domain. We further find that the C2D domain of otoferlin interacts with the calcium channel Ca v 1.3 in a calcium-dependent manner and that a mutation in C2D leads to a substantially decreased binding to Ca v 1.3, independent of calcium. Thus otoferlin, through its C2 domains, may be an important contributor to Ca 2ϩ -dependent synaptic transmission in hair cells.