Rab3-interacting molecules 2α and 2β promote the abundance of voltage-gated CaV1.3 Ca2+ channels at hair cell active zones

Significance Sound encoding relies on Ca2+-regulated transmitter release from inner hair cells (IHCs). Here we demonstrate a role of Ras-related in brain 3 (Rab3)–interacting molecule 2 (RIM2) in Ca2+ channel-clustering and vesicle-tethering at the active zones of IHCs. Active zones of RIM2α-deficient IHCs cluster fewer synaptic voltage-gated CaV1.3 Ca2+ channels, resulting in reduced synaptic Ca2+ influx. Exocytosis was diminished in RIM2α-deficient IHCs, likely contributing to the mild hearing impairment of RIM2α knockout mice. Hair cell-specific disruption of all RIM2 isoforms caused a stronger decrease of Ca2+ current and exocytosis in IHCs and impaired the encoding of sound onset in spiral ganglion neurons. We conclude that RIM2α and RIM2β promote synaptic clustering of Ca2+ channels at IHC active zones and are required for normal hearing. Ca2+ influx triggers the fusion of synaptic vesicles at the presynaptic active zone (AZ). Here we demonstrate a role of Ras-related in brain 3 (Rab3)–interacting molecules 2α and β (RIM2α and RIM2β) in clustering voltage-gated CaV1.3 Ca2+ channels at the AZs of sensory inner hair cells (IHCs). We show that IHCs of hearing mice express mainly RIM2α, but also RIM2β and RIM3γ, which all localize to the AZs, as shown by immunofluorescence microscopy. Immunohistochemistry, patch-clamp, fluctuation analysis, and confocal Ca2+ imaging demonstrate that AZs of RIM2α-deficient IHCs cluster fewer synaptic CaV1.3 Ca2+ channels, resulting in reduced synaptic Ca2+ influx. Using superresolution microscopy, we found that Ca2+ channels remained clustered in stripes underneath anchored ribbons. Electron tomography of high-pressure frozen synapses revealed a reduced fraction of membrane-tethered vesicles, whereas the total number of membrane-proximal vesicles was unaltered. Membrane capacitance measurements revealed a reduction of exocytosis largely in proportion with the Ca2+ current, whereas the apparent Ca2+ dependence of exocytosis was unchanged. Hair cell-specific deletion of all RIM2 isoforms caused a stronger reduction of Ca2+ influx and exocytosis and significantly impaired the encoding of sound onset in the postsynaptic spiral ganglion neurons. Auditory brainstem responses indicated a mild hearing impairment on hair cell-specific deletion of all RIM2 isoforms or global inactivation of RIM2α. We conclude that RIM2α and RIM2β promote a large complement of synaptic Ca2+ channels at IHC AZs and are required for normal hearing.

Ca 2+ influx triggers the fusion of synaptic vesicles at the presynaptic active zone (AZ). Here we demonstrate a role of Ras-related in brain 3 (Rab3)-interacting molecules 2α and β (RIM2α and RIM2β) in clustering voltage-gated Ca V 1.3 Ca 2+ channels at the AZs of sensory inner hair cells (IHCs). We show that IHCs of hearing mice express mainly RIM2α, but also RIM2β and RIM3γ, which all localize to the AZs, as shown by immunofluorescence microscopy. Immunohistochemistry, patch-clamp, fluctuation analysis, and confocal Ca 2+ imaging demonstrate that AZs of RIM2α-deficient IHCs cluster fewer synaptic Ca V 1.3 Ca 2+ channels, resulting in reduced synaptic Ca 2+ influx. Using superresolution microscopy, we found that Ca 2+ channels remained clustered in stripes underneath anchored ribbons. Electron tomography of high-pressure frozen synapses revealed a reduced fraction of membrane-tethered vesicles, whereas the total number of membrane-proximal vesicles was unaltered. Membrane capacitance measurements revealed a reduction of exocytosis largely in proportion with the Ca 2+ current, whereas the apparent Ca 2+ dependence of exocytosis was unchanged. Hair cell-specific deletion of all RIM2 isoforms caused a stronger reduction of Ca 2+ influx and exocytosis and significantly impaired the encoding of sound onset in the postsynaptic spiral ganglion neurons. Auditory brainstem responses indicated a mild hearing impairment on hair cell-specific deletion of all RIM2 isoforms or global inactivation of RIM2α. We conclude that RIM2α and RIM2β promote a large complement of synaptic Ca 2+ channels at IHC AZs and are required for normal hearing.
Rab3-interacting molecule | active zone | ribbon synapse | Ca 2+ channel | vesicle T ens of Ca V 1.3 Ca 2+ channels are thought to cluster within the active zone (AZ) membrane underneath the presynaptic density of inner hair cells (IHCs) (1)(2)(3)(4). They make up the key signaling element, coupling the sound-driven receptor potential to vesicular glutamate release (5)(6)(7). The mechanisms governing the number of Ca 2+ channels at the AZ as well as their spatial organization relative to membrane-tethered vesicles are not well understood. Disrupting the presynaptic scaffold protein Bassoon diminishes the numbers of Ca 2+ channels and membrane-tethered vesicles at the AZ (2,8). However, the loss of Bassoon is accompanied by the loss of the entire synaptic ribbon, which makes it challenging to distinguish the direct effects of gene disruption from secondary effects (9).
Among the constituents of the cytomatrix of the AZ, RIM1 and RIM2 proteins are prime candidates for the regulation of Ca 2+ channel clustering and function (10,11). The family of RIM proteins has seven identified members (RIM1α, RIM1β, RIM2α, RIM2β, RIM2γ, RIM3γ, and RIM4γ) encoded by four genes (RIM1-RIM4). All isoforms contain a C-terminal C 2 domain but differ in the presence of additional domains. RIM1 and RIM2 interact with Ca 2+ channels, most other proteins of the cytomatrix of the AZ, and synaptic vesicle proteins. They interact directly with the auxiliary β (Ca V β) subunits (12,13) and poreforming Ca V α subunits (14,15). In addition, RIMs are indirectly linked to Ca 2+ channels via RIM-binding protein (14,16,17). A regulation of biophysical channel properties has been demonstrated in heterologous expression systems for RIM1 (12) and RIM2 (13).

Significance
Sound encoding relies on Ca 2+ -regulated transmitter release from inner hair cells (IHCs). Here we demonstrate a role of Rasrelated in brain 3 (Rab3)-interacting molecule 2 (RIM2) in Ca 2+ channel-clustering and vesicle-tethering at the active zones of IHCs. Active zones of RIM2α-deficient IHCs cluster fewer synaptic voltage-gated Ca V 1.3 Ca 2+ channels, resulting in reduced synaptic Ca 2+ influx. Exocytosis was diminished in RIM2α-deficient IHCs, likely contributing to the mild hearing impairment of RIM2α knockout mice. Hair cell-specific disruption of all RIM2 isoforms caused a stronger decrease of Ca 2+ current and exocytosis in IHCs and impaired the encoding of sound onset in spiral ganglion neurons. We conclude that RIM2α and RIM2β promote synaptic clustering of Ca 2+ channels at IHC active zones and are required for normal hearing.
A role of RIM1 and RIM2 in clustering Ca 2+ channels at the AZ was demonstrated by analysis of RIM1/2-deficient presynaptic terminals of cultured hippocampal neurons (14), auditory neurons in slices (18), and Drosophila neuromuscular junction (19). Because α-RIMs also bind the vesicle-associated protein Ras-related in brain 3 (Rab3) via the N-terminal zinc finger domain (20), they are also good candidates for molecular coupling of Ca 2+ channels and vesicles (18,21,22). Finally, a role of RIMs in priming of vesicles for fusion is the subject of intense research (18,(21)(22)(23)(24)(25)(26)(27). RIMs likely contribute to priming via disinhibiting Munc13 (26) and regulating vesicle tethering (27). Here, we studied the expression and function of RIM in IHCs. We combined molecular, morphologic, and physiologic approaches for the analysis of RIM2α knockout mice [RIM2α SKO (28); see Methods] and of hair cell-specific RIM1/2 knockout mice (RIM1/2 cDKO). We demonstrate that RIM2α and RIM2β are present at IHC AZs of hearing mice, positively regulate the number of synaptic Ca V 1.3 Ca 2+ channels, and are required for normal hearing.

Results
RIM2α, RIM2β, and RIM3γ Are Present at IHC AZs of Hearing Mice. The mature IHC ribbon synapse shows an unconventional molecular composition (reviewed in refs. [29][30][31]. Recently, the expression of RIM2 and its colocalization with synaptic ribbons was demonstrated for immature mouse IHCs. However, IHCs of hearing mice were reported to lack RIM2 protein as well as RIM1-RIM4 mRNA (13). Here we revisited the expression of RIMs in IHCs of hearing mice, using RT-PCR and immunohistochemistry, using knockout tissue as control. We used the previously published conditional RIM1/2 DKO (14) and crossed it to a new transgenic mouse line expressing Cre-recombinase under the promoter of Vglut3 [Vglut3-Cre, generated as described in SI Appendix, Fig.  S1A, active in IHCs (31,32); a detailed characterization of the Vglut3-Cre line will be published as a separate study] for hair cell-specific deletion of all RIM1 and RIM2 isoforms (nicknamed RIM1/2 cDKO). Using a floxed-GFP reporter (33), Cre recombination was observed in >99% of IHCs and in some outer hair cells, but not in neurons or supporting cells within the cochlea (SI Appendix, Fig. S1B). Different from a previous report (13), we detected mRNA for RIM1-RIM4 in organs of Corti of wild-type mice after the onset of hearing [SI Appendix, Fig. S2A and SI Appendix, Experimental Procedures, postnatal day (p) 14-16, not differentiating the individual RIM1/2 isoforms]. Nested RT-PCR revealed RIM2 and RIM3γ mRNA in single IHCs (p14-p16; SI Appendix, Fig. S2B). Nested RT-PCR for specific RIM2-isoforms detected RIM2α mRNA in most IHCs (SI Appendix, Fig. S2C). mRNAs for RIM2β and RIM2γ were more rarely detectable in single IHCs but were always found at the organ of Corti level (SI Appendix, Fig. S2C).
Next, we studied the presence and localization of RIM proteins by labeling IHCs of wild-type, RIM2α SKO, and RIM1/2 cDKO mice with antibodies to RIM1, RIM2, and RIM3γ. In wild-type IHCs, we found spot-like RIM2 (Fig. 1A) and RIM3γ (SI Appendix, Fig. S3A) immunofluorescence that colocalized with C-terminal binding protein 2 (CtBP2)/RIBEYE-marked ribbon-occupied AZ of IHCs, whereas no obvious RIM1 immunofluorescence was detectable (SI Appendix, Fig. S3B). In addition, we found RIM2 immunofluorescence separated from ribbons ( Fig. 1 A and C-F). This most likely reflects RIM2 expression at synapsin-expressing conventional efferent synapses (blue in Fig. 1A, synapsin is absent from IHCs) (34) of lateral olivocochlear neurons with spiral ganglion neurons (SGNs; connectivity is illustrated in Fig. 1B) underneath the IHCs (Fig. 1D). Using two-color stimulated emission depletion (STED) microscopy, we observed a stripe-like shape of the RIM2 immunofluorescence, which, according to synapse orientation, appeared to mark the base of the synaptic ribbon (Fig. 1C), indicating that RIM2 is part of the Bassoon-positive presynaptic density of the AZ, which occupies the space between the ribbon and membrane (4). RIM2 immunofluorescence, Fig. 1. RIM2α, RIM2β, and RIM3γ are present at IHC AZs of hearing mice. (A) Projections of confocal sections of a wild-type organ of Corti immunolabeled with antibodies to RIM2 (green), CtBP2/RIBEYE (red, antibody labeling the presynaptic ribbon), and synapsin1/2 (blue, labeling conventional efferent presynaptic terminals). Each ribbon synapse also showed RIM2-immunofluorescence, indicating the presence of RIM2 at the ribbon-type AZ of IHCs. RIM2 spots were also present away from ribbon often colocalizing with the synapsin-stain of efferent synapses onto the postsynaptic boutons of type I SGNs formed by lateral olivocochlear neurons, illustrated in the simplifying schematic in B. (Scale bar: 10 μm.) (C) 2-color STED image of a wild-type IHC AZ immunolabeled for RIM2 (green) and CtBP2/RIBEYE (magenta), showing a stripe-like morphology of the synaptic RIM2 cluster that is juxtaposed to the ribbon, as expected for a protein of the presynaptic density. The image is a maximum-intensity projection after smoothing with a Gaussian (σ = 1 pixel) in ImageJ. (Scale bar: 1 μm.) (D) XY-projection of confocal sections of a RIM2α con organ of Corti immunolabeled with a RIM2-specific antibody (green, Middle) and a CtBP2/RIBEYE antibody labeling the presynaptic ribbon (magenta, merged with green, Left). (Right) ZY projections of the same confocal sections with two IHCs (basolateral pole outlined by dashed line) illustrating RIM2 immunofluorescence underneath the IHCs. [Scale bars: 10 μm (Middle), 2 μm (Right).] (E) XY projection of confocal sections of a RIM2α SKO organ of Corti immunolabeled with a RIM2-specific antibody (green, Middle) and a CtBP2/RIBEYE antibody labeling the presynaptic ribbon (magenta, Right). Each ribbon also showed RIM2-immunofluorescence, indicating the presence of RIM2β at the ribbon-type AZ of RIM2α SKO IHCs. (Scale bar: 10 μm.) (F) XY projection of confocal sections of a RIM1/2 cDKO organ of Corti immunolabeled with a RIM2-specific antibody (green, Middle) and a CtBP2/RIBEYE antibody labeling the presynaptic ribbon (magenta, Right). RIM2 spots were always separated from ribbons, consistent with RIM2 expression at efferent synapses in the hair cell-specific cDKO. (Scale bar: 10 μm.) although weaker, remained at the ribbon-occupied AZs of RIM2α SKO IHCs (Fig. 1E) but was absent from AZs of IHCs of RIM1/2 cDKO mice lacking all RIM2 isoforms (Fig. 1F). This indicates the additional presence of RIM2β at IHC AZs, as the RIM2 antibody does not recognize RIM2γ.
Reduced Ca V 1.3 Ca 2+ Channel Immunofluorescence at AZs of RIM2α-Deficient IHCs. First, we studied the role of RIM2α, the likely most abundant RIM2 isoform in IHCs, in regulating presynaptic Ca V 1.3 Ca 2+ influx. We approached the overall organization of RIM2α SKO IHC synapses using immunohistochemistry for Ca V 1.3 and CtBP2/RIBEYE, marking the presynaptic Ca 2+ channel cluster and ribbons as well as GluA2/3, indicating the postsynaptic AMPA receptor cluster. Ribbon-occupied synapses, identified as CtBP2/RIBEYE immunofluorescent spots juxtaposed to GluA2/3 immunofluorescence ( Fig. 2A), were found in normal number in RIM2α SKO IHCs (Fig. 2B). Ca V 1.3 Ca 2+ channels remained clustered at each ribbon-type IHC AZ (Fig. 2 C and D), but the Ca V 1.3 immunofluorescence intensity was reduced in RIM2α SKO IHCs. Using semiquantitative immunohistochemistry with identical and parallel tissue processing, imaging, and analysis of immunofluorescence, we found a reduction by 48% ( Fig. 2 D and E; P < 0.001, 417 RIM2α SKO AZs in 56 IHCs in three organs of Corti from two mice; 390 RIM2α control (con) AZs in 50 IHCs in four organs of Corti from three mice). A similar result was obtained in two further sets of experiments. Using two-color STED microscopy, we found stripe-like Ca 2+ channel clusters at ribbon-occupied RIM2α SKO AZs, which, on inspection, in single XY sections, seemed to be qualitatively comparable in size and shape to those of RIM2α con IHCs (Fig. 2C). We did not detect significant differences in immunofluorescence intensity for RIBEYE ( Fig.  2D; P = 0.27, same AZs analyzed as for Ca V 1.3 earlier) and GluA2/3 (P = 0.26; n = 650 synapses in 72 IHCs in four organs of Corti from RIM2α SKO mice; n = 673 synapses in 78 IHCs and four organs of Corti for RIM2α con).

Disruption of RIM2α and RIM2β Reduces Synaptic Ca 2+ Influx in IHCs.
To study the effect of RIM2 disruption on presynaptic Ca 2+ influx, we performed perforated patch-clamp recordings of Ca 2+ current (I Ca ) and confocal imaging of presynaptic Ca 2+ signals in IHCs of RIM2α SKO. We found a mild reduction of the wholecell I Ca amplitude on deletion of RIM2α (16.8%; Fig. 3 A and B). The deletion of all RIM2 isoforms in IHCs of RIM1/2 cDKO mice decreased I Ca by 51.4% ( Fig. 3A and SI Appendix, Fig. S5C; P = 0.01 relative to Cre-negative littermate controls: RIM1/2 con). I Ca amplitudes were not significantly different between IHCs of RIM1/2 cDKO (lacking all RIM1 and RIM2 isoforms, eight IHCs in five organs of Corti) and RIM2 cSKO [lacking all RIM2 isoforms (see Methods) seven IHCs in four organs of Corti; P = 0.5; SI Appendix, Fig. S5C], supporting the notion that IHC AZs operate without RIM1. Next, we performed a detailed analysis of the biophysical properties of the Ca 2+ channels in RIM2α SKO IHCs. Voltage dependence (V 0.5 , −14.8 ± 2.7 mV for nine RIM2α SKO IHCs vs. −12.4 ± 3.5 mV for eight RIM2α con IHCs; P = 0.58) and kinetics of activation of I Ca as well as I Ca inactivation were unaltered (SI Appendix, Fig. S4 A and B). The reduced synaptic Ca V 1.3 immunofluorescence in RIM2α SKO IHCs suggested a diminished number of synaptic Ca V 1.3 Ca 2+ channels. To test this hypothesis and probe for potential changes in single-channel current or open probability, we performed a nonstationary fluctuation analysis as described ( Fig. 3B) (35). The estimated number of Ca 2+ channels in RIM2α SKO IHCs (eight IHCs) amounted to 1,102 ± 65, which was 17.8% less than in RIM2α con IHCs (1,341 ± 73; seven IHCs), and hence, the reduced number of Ca 2+ channels accounts for the reduction of the whole-cell Ca 2+ current. Neither the single channel current nor the open probability changed significantly (SI Appendix, Table S3). Because the whole-cell recordings sum I Ca of all AZs and extrasynaptic membranes, we sought to further analyze Ca 2+ influx at the single AZ, using confocal Ca 2+ imaging at fluorescently tagged AZs (2,36,37). IHCs were depolarized for 20 ms to −7 mV to fully activate Ca 2+ channels, and Fluo-5N fluorescence was studied using line scans across the center of "Ca 2+ microdomains" (Fig. 3 C and D), identified as hotspots of Fluo-5N fluorescence change during a preceding 254-ms depolarization. Ca 2+ microdomains of RIM2α SKO (40 AZs in 13 IHCs in seven organs of Corti) and RIM2α con (35 AZs in 15 IHCs in seven organs of Corti) showed comparable spatial extents ( Fig. 3C and SI Appendix, Fig. S5B) and kinetics (Fig. 3D). However, we found a 36% reduction of the maximal fluorescence change ( Fig.  3D and SI Appendix, Fig. S5A; P < 0.002), which is less than the decrease of synaptic Ca V 1.3 immunofluorescence (48%) but exceeds the loss of the whole-cell Ca 2+ current (18%). The fluorescence change varied greatly among the AZs in IHCs of RIM2α SKO mice, as previously described for wild-type IHCs (36). The coefficients of variation were comparable in both genotypes (0.65 for RIM2α con and 0.66 for RIM2α SKO IHCs), again contrasting the synaptic phenotype of Bassoon-deficient IHCs that showed reduced Ca 2+ signaling heterogeneity. In summary, RIM2α and RIM2β promote a large complement of synaptic Ca V 1.3 Ca 2+ channels. This function of RIM2 seems to involve the immobilization of the channels at the AZ, at least in addition to promoting channel trafficking to the plasma membrane, as the reduction of synaptic Ca 2+ channels exceeded that of the whole-cell Ca 2+ current.
Disruption of RIM2α Reduces the Fraction of Membrane-Tethered Synaptic Vesicles. Next, we studied the ultrastructure of the AZ, using electron microscopy and tomography. To obtain a close to native structural preservation, we used high-pressure freezing in combination with freeze-substitution. Tomographic reconstructions revealed that the AZ ultrastructure was generally maintained in RIM2α SKO IHCs showing normally anchored, sized [average ribbon height: 258.4 ± 24.6 nm for RIM2α con (n = 10 tomograms) vs. 270.2 ± 18.0 nm for RIM2α SKO (n = 10 tomograms); Fig. 4 A and G] and shaped synaptic ribbons with an unaltered complement of ribbon-associated synaptic vesicles (29.0 ± 3.9 for RIM2α SKO vs. 28.2 ± 3.5 for RIM2α con per tomogram, typically containing approximately half of the ribbon). The total number of membrane-proximal vesicles within 50 nm from the AZ membrane ( Fig. 4C) was unaltered in RIM2α SKO compared with controls (8.0 ± 1.0 for RIM2α SKO vs. 6.9 ± 0.7 for RIM2α con; Fig. 4 D and F). Performing electron tomography allowed the identification of tethers to the vesicles as well as the determination of the tether length in 3D (SI Appendix, Fig. S6 A and B). Interestingly, the fraction of membranetethered vesicles (number of tethered vesicles/total number of membrane-proximal vesicles) was significantly reduced at RIM2α SKO AZs (0.24 ± 0.03 for RIM2α SKO vs. 0.41 ± 0.02 for RIM2α con; P < 0.002; Fig. 4 D and E), suggesting RIM2α plays a role in tethering vesicles to the AZ membrane of IHCs. The tether length, however, was unchanged (average tether length, 22.7 ± 1.6 nm for RIM2α con vs. 22.9 ± 1.5 nm for RIM2α SKO; SI Appendix, Fig. S6C).
Disruption of RIM2α and RIM2β Reduces IHC Exocytosis. How does the disruption of RIM2 affect hair cell exocytosis? The reduction of Ca 2+ channels and the impaired vesicle tethering are likely to diminish exocytosis. Moreover, as RIM2α is a candidate for linking vesicles and Ca 2+ channels, a potential disruption of Ca 2+ influx-exocytosis coupling might result. We tested these hypotheses by measuring exocytic membrane capacitance changes (ΔC m ) in RIM2α SKO and RIM1/2 cDKO IHCs in response to depolarizations of various durations to the maximum Ca 2+ current potential in perforated-patch recordings (Fig. 5 A and B). On the basis of previous work (35,(38)(39)(40), we interpret the fast and saturating component (0-20 ms depolarization duration) as exocytosis of a standing readily releasable pool (RRP) (41), and the slower component (>20 ms depolarization duration) as sustained exocytosis, primarily reflecting vesicle supply to the RRP and subsequent fusion (Fig. 5B). The exocytic ΔC m , in response to 20-ms depolarization, was reduced by 36% in RIM2α SKO IHCs (14 IHCs, P = 0.015 relative to RIM2α con) and by 52% in RIM1/2 cDKO (15 IHCs, P = 0.003 relative to RIM1/2 con and P < 0.001 relative to RIM2α SKO). The rate of sustained exocytosis was diminished, as well [from 0.76 ± 0.14 fF/ms in RIM2α con IHCs to 0.38 ± 0.07 fF/ms in RIM2α SKO IHCs (P = 0.02 relative to RIM2α con) and 0.09 ± 0.02 fF/ms in RIM1/2 cDKO IHCs [P < 0.001 relative to RIM1/2 con (nine IHCs, 0.5 ± 0 0.08 fF/ms; P < 0.001 relative to RIM2α SKO)]. Exocytosis was not significantly different between IHCs of RIM1/2 cDKO and RIM2 cSKO (P = 0.5; SI Appendix, Fig. S5D), supporting the notion that hair cell AZ operates without RIM1. Relating the exocytic ΔC m to the RIM2α SKO RIM2α con corresponding integrated Ca 2+ current (Q Ca ; Fig. 5C) reduced the discrepancy of the exocytic responses between the RIM2α SKO and control IHCs, indicating that the reduction of presynaptic Ca 2+ influx explains impaired exocytosis to a large extent.
As an interaction partner of Rab3, which is present in hair cells (42), and of the Ca V 1.3 channel (13), RIM2α is a candidate molecular linker between Ca 2+ channel and vesicular release site at the IHC AZ. Hence, we studied Ca 2+ influx-exocytosis coupling by estimating the apparent Ca 2+ cooperativity m of exocytosis upon manipulation of the single Ca 2+ channel current via changes of [Ca 2+ ] e (Fig. 5D, Upper) and reduction of open Ca 2+ channels by slow perfusion of the dihydropyridine channel antagonist isradipine (10 μM; Fig. 5D, Lower). We obtained m by fitting the relationship of ΔC m,20 ms and Q Ca for each cell with a power function: ΔC m = A(Q Ca ) m (43), and for the case of the single Ca 2+ channel current change, we restricted fitting to the Q Ca range that did not show obvious saturation of ΔC m,20 ms (4) (Fig. 5D). The estimates of m were statistically indistinguishable between RIM2α SKO and RIM2α con IHCs for both manipulations [isradipine, 1.58 ± 0.12 vs. 1.51 ± 0.11 (P = 0.62); Δ[Ca 2+ ] e , 2.63 ± 0.09 vs. 2.59 ± 0.14 (P = 0.8); n = 6 IHCs for both genotypes]. We conclude that the RIM2α disruption impairs RRP size and replenishment, but leaves Ca 2+ influx-exocytosis coupling largely intact.
RIM2 Disruption Impairs Sound Onset Encoding. What are the consequences of RIM2 disruption for hearing? The thresholds of auditory brainstem responses (ABRs) were mildly but significantly elevated in the RIM mutants (Fig. 6 B and D; by 5 dB SPL in RIM1/2 cDKO mice and by 16 dB in RIM2α SKO mice compared with their respective littermate controls; both P < 0.0001).
As the presence of otoacoustic emissions with normal amplitudes (SI Appendix, Fig. S7 A-D) indicates normal mechanoelectrical transduction and cochlear amplification, this suggests a synaptopathic hearing impairment (44).      The significant delay of ABR waves 2-5 ( Fig. 6C and SI Appendix, Fig. S7E; P < 0.0001 for each) and the reduction in ABR wave 3 amplitude in RIM2α SKO mice ( Fig. 6D; P < 0.0001) indicate an additional impairment of synaptic transmission deficit at central auditory synapses. Consistent with the above notion that the IHC synapse seems to operate without RIM1, RIM1α SKO mice showed normal ABR thresholds and waveforms (SI Appendix, Fig. S7 G and H), also indicating that the sole deletion of RIM1α does not impair transmission at central auditory synapses. Next, we performed extracellular recordings from individual SGNs, each of which is driven by a single AZ (45), using glass microelectrodes targeted toward the entry point of the auditory nerve into the anteroventral cochlear nucleus (46). SGNs were distinguished from cochlear nucleus neurons on the basis of their response patterns and the electrode position (Methods) (46). Sound thresholds (SI Appendix, Fig. S8A) and frequency tuning of SGN were comparable between RIM mutant mice and their respective littermate controls. Responses of SGNs to suprathreshold tone bursts followed an overall normal pat-tern, with spike rate peaking at sound onset and then gradually declining [short-term adaptation, likely reflecting partial RRP depletion (2,46)] to an adapted rate that was maintained throughout stimulation (Fig. 6 E and F). SGNs of RIM1/2 cDKO mice showed a significantly reduced peak rate ( Fig. 6E; P = 0.02, paired t test) and a delayed first spike latency (P = 0.02, paired t test), whereas the jitter of the first spike was unaltered (P = 0.45 for comparison of the variance of first spike latency). Steadystate spike rates (Fig. 6E) and spontaneous spike rates (SI Appendix, Fig. S8B) showed a nonsignificant trend toward smaller values. Both spontaneous (SI Appendix, Fig. S8B) and evoked rates (Fig. 6F) also tended to be smaller in putative SGNs of RIM2α SKO mice, although differences did not reach significance. The increase of average spike rate with sound intensity was significantly less steep, and the dynamic range tended to be broader in RIM2α SKO SGNs compared with SGNs of littermate controls (SI Appendix, Fig. S8 C-E). The time course of recovery from adaptation (likely reflecting RRP recovery from depletion; SI Appendix, Fig. S8F) was normal in RIM2α SKO SGNs, as was the temporal precision of coding, assessed as the jitter of the first spike after sound onset (SI Appendix, Fig. S8G) and as the synchronicity of firing to the sound envelope of amplitude-modulated tones (SI Appendix, Fig. S8H). Similar to SGNs, neurons of the cochlear nucleus in RIM2α SKO mice showed only subtle alterations in their spontaneous and soundevoked spiking activity, such as a small but significant reduction of peak rates in putative bushy cells and in multipolar cells (SI Appendix, Fig. S8 I and J). Taken together, RIM2 disruption caused a mild sound-encoding phenotype that reached significance only in RIM1/2 cDKO SGNs, compatible with the stronger impairment of the presynaptic function of their IHCs compared with RIM2α SKO IHCs. RIM2α disruption at the central auditory synapses of RIM2α SKO mice was reflected by reduced peak firing rates in bushy cells and alteration of the central ABR peaks.

Discussion
Here we identified RIM2α and RIM2β as the long RIM isoforms at the afferent IHC synapse of hearing mice. RIM2α and RIM2β positively regulate the number of synaptic Ca V 1.3 Ca 2+ channels in IHCs, likely by clustering them at the AZ. In addition, RIM2α seems to be involved in tethering synaptic vesicles to the AZ membrane. Disruption of RIM2α and RIM2β diminishes IHC exocytosis primarily via reducing synaptic Ca 2+ influx and mildly impairs synaptic sound encoding.
AZs of Cochlear IHCs of Hearing Mice Use Multiple RIM Isoforms. The molecular composition of the IHC ribbon synapse is highly specialized to suit the needs of fast and indefatigable transmitter release and deviates from "conventional" synapses of the CNS in several ways (28)(29)(30). Here we show that IHCs of hearing mice express RIM2α, RIM2β, and RIM3γ, which corrects a previous notion of a developmental loss of RIM expression in IHCs (13). However, unlike many other synapses studied so far (14,18,(47)(48)(49), IHC ribbon synapses seem to lack RIM1. Obviously, negative single-cell PCR and immunohistochemistry do not rule out weak RIM1 expression. However, the normal ABR in RIM1α SKO mice and the indistinguishable presynaptic dysfunction of IHCs from RIM2 cSKO and RIM1/2 cDKO mice support the notion that IHC synapses operate without RIM1. Therefore, RIM2α seems to serve as the only α-RIM capable of interacting with the vesicle-associated protein Rab3, which is also expressed in hair cells (42). Beyond the Rab3 interaction, the long isoforms RIM2α and RIM2β have overlapping interaction partners such as Munc13, ELKS, and Ca 2+ channels. The further reduction of synaptic Ca 2+ influx by additional deletion of RIM2β likely reflects additive and potentially nonoverlapping roles of RIM2α and RIM2β in clustering synaptic Ca 2+ channel, although we cannot definitively exclude a compensatory up-regulation of   6. Mild impairment of sound encoding in RIM1/2 cDKO and RIM2α SKO mice. (A) ABR waveforms of RIM1/2 cDKO mice (gray, n = 8) and RIM1/2 con (black, n = 12) in response to 80 dB (pe) click stimulation were generally preserved, but in RIM1/2 cDKO, the latency of peak I was prolonged (SI Appendix, Fig. S7F). (B) Average ABR thresholds of RIM1/2 cDKO mice (gray, n = 8) were significantly elevated compared with RIM1/2 con (black, n = 11).
(C) ABR in RIM2α SKO mice (gray, n = 15) were preserved but had prolonged latencies of peaks II-V compared with RIM2α con (black, n = 15) littermate controls (SI Appendix, Fig. S7E). (D) Average ABR thresholds of RIM2α SKO mice (gray, n = 11) were significantly elevated compared with littermate controls (black, n = 14). (E) Mean peristimulus time histogram of RIM1/2 cDKO (gray, n = 10) and RIM1/2 con (black, n = 22) SGNs to tone burst stimulation at the characteristic frequency of each fiber, 30 dB above threshold. Shaded areas indicate the SEM. (Inset) Quantification of peak rates at sound onset (maximum, 0.5 ms bin) and steady state rates (averaged between 37 and 47 ms after sound onset) of individual fibers. Peak rates were significantly reduced in RIM1/2 cDKO (P = 0.02), and sustained rates tended to be reduced (n.s.); the overall response pattern was preserved. (F) Mean peristimulus time histogram of RIM2α SKO (gray, n = 32) and RIM2α con (black, n = 42) SGNs, illustrated as in E. There was a nonsignificant trend to lower spike rates in the RIM2α SKO SGNs.
RIM2β in RIM2α SKO IHCs. However, neither a compensatory expression of RIM1 nor an up-regulation of RIM2β can fully substitute for the lost RIM2α in the regulation of synaptic Ca 2+ channels in IHCs. These findings contrast recent observations at the calyx of Held synapse of the central auditory pathway, in which RIM1 and RIM2 isoforms can largely replace each other in regulating the abundance of presynaptic Ca 2+ channels (50). The localization of RIM3γ at the IHC AZ was unexpected, given that γ-RIMs assume postsynaptic roles in CNS neurons (51,52). Future studies on RIM3γ knockout mice will be required to confirm the specificity of RIM3γ immunolabeling at IHC AZs and to elucidate its relevance for IHC synaptic transmission and hearing.
Positive Regulation of Presynaptic Ca 2+ Signaling of IHCs by RIM2α and RIM2β. Here we identified the IHC ribbon synapse as a system that requires RIM2α and RIM2β for normal function. Importantly, the functional synaptic deficit arose despite a normal presence of synaptic ribbons, which is different from the Bassoon disruption that also reduced synaptic Ca 2+ channels but, in addition, impaired the anchorage of synaptic ribbons (2,9). Immunohistochemistry and patch-clamp recordings of Ca 2+ influx and confocal Ca 2+ imaging consistently demonstrated a reduction of synaptic Ca 2+ channels on disruption of RIM2α. Similar to previous findings on Bassoon-deficient IHCs (2), the reduction of synaptic Ca 2+ channels was greater than that of the total number of channels, which likely indicates that RIM2 primarily recruits and/or retains Ca V 1.3 channels at the AZ. In addition, RIM2 may have a role in Ca V 1.3 trafficking to the plasma membrane, but the extent to which this mechanism contributes is less clear, given that palmitoylated Ca V β2a, which is likely expressed in IHCs (35), might take this role. Unlike in Bassoon-deficient IHCs that primarily lack the population of AZs with a large Ca 2+ channel complement (2,9), the cumulative distribution function of Ca V 1.3 immunofluorescence intensities of RIM2α SKO AZs is overall shifted toward smaller values. The shape of the Ca 2+ channel clusters at RIM2α SKO IHC AZs remained stripe-like, indicating that channel localization at the AZ is maintained on the tens of nanometer scale, whereas clusters are disintegrated into two to three round spots on Bassoon-disruption. Moreover, we found a generally well-preserved synaptic ultrastructure in IHCs of RIM2α SKO mice. Most notably, unlike the ribbonless synapses in Bassoondeficient IHCs (8), ribbons were properly anchored and displayed a normal morphology in RIM2α-deficient IHCs. Hence, although both RIM (14,16) and Bassoon (53) may use binding to RIM-binding protein to interact with Ca 2+ channels, the effects of RIM2α and Bassoon disruption on Ca V 1.3 Ca 2+ channel clustering at the IHC AZ differ. This may reflect a stronger perturbation of the AZ in Bassoon-deficient IHCs, which is also evident from the additional loss of the synaptic ribbon. Most likely, RIM2 also indirectly interacts with the Ca V 1.3 Ca 2+ channel complex via binding to the Ca V β subunits (13); for instance, to Ca V β2, the predominant β-subunit in IHCs (35). Future studies of RIM interactions with the Ca V 1.3 channel complex will be required to explore additional binding mechanisms and evaluate the relative contribution of the interactions for mediating the positive regulation of the number of synaptic Ca 2+ channels by RIM2α and RIM2β.
Impaired Exocytosis and Hearing Upon RIM2 Disruption. Given the concurrent reduction of Ca 2+ influx and exocytosis observed in the present study for both RIM2α-and RIM2α/β-deficient IHCs, we favor the interpretation that RIM2 disruption reduces exocytosis primarily by diminishing synaptic Ca 2+ channels. The reduced presynaptic exocytosis underlies the impaired sound encoding that was more pronounced on deletion of both RIM2α and RIM2β, which also caused a stronger reduction of presynaptic Ca 2+ influx and exocytosis. This view is in line with previous studies of IHCs with fewer channels resulting from the deletion of Ca V β2 (35) and Ca V 1.3α (6), where we also observed a linear reduction of exocytosis with the Ca 2+ current (by 70% and 90% in Ca V β2and Ca V 1.3-deficient IHCs, respectively). The apparent Ca 2+ cooperativity of exocytosis (Fig. 5C) and Ca 2+ efficiency (Fig. 5D) suggests that coupling of Ca 2+ influx to exocytosis (1,4,54) was largely unaltered in RIM2α-deficient IHCs, which contrasts findings at the in RIM1/2-deficient IHCs calyx of Held (18), which might reflect compensation by other molecular links between channels and release sites at the IHC AZ. Alternatively, a subtle alteration of Ca 2+ channel position relative to the vesicular Ca 2+ sensor on the scale of few nanometers might have gone undetected in our measurements of exocytic membrane capacitance changes. Our morphological analysis did not reveal major structural deficits of RIM2α-deficient synapses: synapse number, ribbon anchorage and the shape of the Ca 2+ channel cluster appeared intact, and the number of ribbon-associated and membrane-proximal vesicles was not changed. This contrasts the synaptic phenotype elicited by disruption of two other presynaptic scaffold proteins, CAST and Bassoon, for which a reduction in ribbon size in rod photoreceptors (55) and a loss of the ribbon, respectively, from AZs of photoreceptors (56) and hair cells (8) was reported that led to a reduction of membrane-proximal synaptic vesicles at the presynaptic AZ in each case. Our observation of a reduced fraction of membrane-tethered vesicles adds further support to a role of RIMs in regulating vesicle tethering to the membrane of the AZ (27), although RIM2α seems not to be essential for this process.
The contribution of the impaired tethering to reduced exocytosis and the consequences of deleting all RIM isoforms on synaptic ultrastructure remains to be probed in future studies. Deletion of all long RIM isoforms reduced the number of docked synaptic vesicles at AZs of both cultured neonatal hippocampal neurons (14) and the calyx of Held (18). Interestingly, neither the isolated deletion of RIM1 nor RIM2 isoforms affected synaptic transmission at the calyx of Held in vitro (50), and ABR recordings suggest normal auditory signaling in RIM1α knockout mice (SI Appendix, Fig. S7). In contrast, delayed latencies of central auditory ABR peals and extracellular recordings from single cochlear nucleus neurons suggested a mild impairment of transmission at central auditory synapses in the absence of RIM2α. Thus, there seems to be a requirement for RIM2 at central auditory synapses for normal hearing. Future in vivo recordings from the central auditory pathway on cell-specific RIM deletion mutants will help reconciling these data sets.
Generation of Vglut3-Cre mice is described in SI Appendix, Fig. S1. Most experiments compared RIM1/2 cDKO and RIM2 cSKO with their Cre-negative littermate controls (RIM1/2 con and RIM2 con), or RIM2α SKO with their wildtype littermates (RIM2α con), respectively. C57BL/6 mice were used for some experiments related to Fig. 1  semiquantitative analysis of immunofluorescence, samples of mutant and control were processed identically and in parallel throughout immunohistochemistry, confocal imaging, and analysis. Intensity per synapse was estimated in the optical section with peak intensity. Immunofluorescence intensity was measured as the integral of the pixel values under the 2D Gaussian after subtraction of the background [estimated by fitting a linear function to the region outside of the 2D Gaussian, and subsequently subtracting this function from the entire image (2)]. The average voxel intensity of the stack was subtracted as background. As the absolute immunofluorescence intensity varied between the sets of three (Ca V 1.3 and RIBEYE) and four (GluA2/3) experiments, the analysis was presented for one experiment each, but the findings in mutant relative to control were similar in all experiments. Two-color STED images were acquired on a custom-built setup, as described previously (57). A pair of pulsed laserdiodes at 595 and 640 nm wavelength were used to excite the dyes Atto590 (Atto-Tec) and KK114 (58). The fluorescence was detected by avalanche photodiodes (SPCM-AQRH13, Excelitas, and PR-050-CTB; Micro Photon Devices) with detection windows ranging from 600 to 640 nm and from 660 to 720 nm, respectively. The resolution enhancement is achieved with a pulsed STED laser of 775 nm wavelength. As both color channels are recorded quasi-simultaneously, and the same STED laser beam is used for both dyes, the color channels are inherently aligned, and no postprocessing is required.
High-Pressure Freezing. For high-pressure freezing, organs of Corti (p14-p15) were explanted in Hepes-Hanks' solution and placed in aluminum specimen carriers of 200 μm (type A, Leica) depth filled Hepes-Hanks' solution. The aluminum lid (type B, Leica) has been dipped in hexadecen (Sigma-Aldrich) before placing onto the sample. Samples were frozen immediately using the HPM100 (Leica) and rapidly transferred into liquid nitrogen for storage. Freeze substitution was performed in an EM AFS2 (Leica). Samples were incubated for 4 d in 0.1% tannic acid in acetone at −90°C and subsequently washed three times for 1 h in acetone. Before increasing the temperature from −90°C to −20°C (5°C/h), 2% (wt/vol) osmium tetroxid in acetone was added and incubated for 16 h at −20°C. Finally, the temperature was raised to 4°C (10°C/h) and samples were washed in acetone to remove the osmium tetroxid and warmed to room temperature (RT). Samples were infiltrated and embedded in Epon resin. After the embedding, 250-nm sections for electron tomography were obtained on an Ultracut E ultramicrotome (Leica) with a diamond knife (Diatome). Sections were applied to Formvar-coated copper mesh grids and poststained with 4% (wt/vol) uranyl acetate and Reynold's lead citrate.
Electron Tomography. To both sides of the stained grids, 10-nm gold particles (British Bio Cell) were applied. Single-tilt series were acquired using a JEOL JEM 2100 transmission electron microscope at 200 kV from −60 to +60 with 1°i ncrement, using the Serial-EM software. The tomograms were generated using the IMOD package etomo, and models were rendered using 3dmod (bio3d.colorado.edu/imod/). Quantitative image analysis was performed blinded, using the ImageJ software package as follows: The height of the ribbon was determined as the maximum extent from the presynaptic density in virtual sections. For ribbonassociated synaptic vesicles, the first row of vesicles with a maximal distance of 80 nm from the ribbon to the vesicle membrane was counted per tomogram. Membrane-proximal synaptic vesicles were defined as vesicles with a membrane-to-membrane distance to the AZ of maximal 50 nm. For tether length analysis, the coordinates of the tether starting (X 1 , Y 1 , Z 1 ) and ending (X 2 , Y 2 , Z 2 ) points were defined in virtual sections, using the ImageJ software package, and length was calculated according to Tether Length = √(( Fig. S6).
Patch-Clamp and Confocal Ca 2+ Imaging in IHCs. p14-p16 IHCs from apical coils of freshly dissected organs of Corti were patch-clamped as described (41 2+ ] was varied between nominally Ca 2+ -free to 5 mM for experiments manipulating the single Ca 2+ channel current and was 10 mM for fluctuation analysis. In all cases, NaCl concentration was adjusted for osmolarity of solutions with different CaCl 2 concentrations. Currents were low-pass filtered at 2.9 kHz (8.5 kHz for fluctuation analysis) and sampled at 50 kHz (100 kHz for fluctuation analysis). Cells with holding current greater than −50 pA were discarded. An EPC-9 amplifier and "Pulse" or an EPC-10 amplifier and "Patchmaster" software (HEKA Elektronik) were used for measurements. Whole-cell capacitance measurements were performed in perforated-patch configuration, as previously described (41), except that ΔC m was estimated as the difference of the mean C m over 100 ms after the end of the depolarization (the initial 80 ms were skipped). All voltages were corrected for liquid-junction potentials and voltage drops across series resistance. Ca 2+ currents were further isolated from background current, using a P/n protocol. Confocal Ca 2+ imaging was performed essentially as described (37). In brief, presynaptic Ca 2+ -signal of IHCs were observed as changes of Ca 2+ indicator fluorescence in line scans at the ribbon location, using (20 ms) step depolarizations to −7 mV.

Recordings of Auditory Brainstem Response and Distortion Product Otoacoustic
Emissions. Animals aged 5-9 wk were anesthetized intraperitoneally with a combination of ketamine (125 mg/kg) and xylazine (2.5 mg/kg). The core temperature was maintained constant at 37°C, using a rectal temperaturecontrolled heating blanket (Hugo Sachs Elektronik-Harvard Apparatus GmbH). For stimulus generation, presentation, and data acquisition, we used the TDT System III (Tucker-Davis-Technologies) run by BioSig32 software (Tucker-Davis-Technologies). Sound pressure levels are provided in dB SPL RMS (tonal stimuli) or dB SPL peak equivalent (PE, clicks) and were calibrated using a 1/4" microphone (D 4039; Brüel and Kjaer GmbH). Tone bursts (12 kHz, 10 ms plateau, 1 ms cos 2 rise/fall, stimulus rate 40 Hz) or clicks of 0.03 ms (stimulus rate 20 and 100 Hz) were presented in the free field ipsilaterally using a JBL 2402 speaker (JBL GmbH & Co.). The difference potential between vertex and mastoid subdermal needles was amplified (50,000 times), filtered (low pass, 4 kHz; high pass, 400 Hz) and sampled at a rate of 50 kHz for 20 ms, 2 × 2,000 times to obtain two mean auditory brainstem response traces for each sound intensity. Hearing threshold was determined with 10 dB precision as the lowest stimulus intensity that evoked a reproducible response waveform in both traces by visual inspection. For distortion product otoacoustic emissions, the ED1/EC1 speaker system (Tucker-Davis-Technologies) was used to generate two primary tones (frequency ratio f2/f1, 1.2; intensity f2 = intensity f1 + 10 dB). Primary tones were coupled into the ear canal by a custom-made probe containing an MKE-2 microphone (Sennheiser) and adjusted to the desired sound intensities at the position of the ear drum, as mimicked in a mouse ear coupler. The microphone signal was amplified and digitalized (Terratec DMX 6 Fire) and analyzed by fast Fourier transformation (Matlab).
Extracellular Recordings from Single. SGNs were performed as previously described (9,46). In brief, mice aged 4-12 wk were anesthetized by i.p. injection of urethane (1.32 mg·kg −1 ), xylazine (5 mg·kg −1 ), and buprenorphine (0.1 mg·kg −1 ). An occipital approach with partial removal of the cerebellum was taken to advance a glass microelectrode during noise burst stimulation, aiming to identify SGNs near the auditory nerve's entry zone into the cochlear nucleus and avoiding the posterior region of the anteroventral cochlear nucleus, where spherical bushy cells are mainly located. Putative SGNs were distinguished from primary cochlear nucleus units based on the electrode position (electrode aiming at internal auditory canal, >1 mm below the surface of cochlear nucleus) and their primary-like response characteristics to 200 suprathreshold 50-ms tone bursts presented at the characteristic frequency of the SGN, 30 dB above threshold. The irregular firing pattern observed in SGNs was confirmed by a coefficient of variation of interspike intervals of adapted responses greater than 0.5. We first determined the spontaneous rate in a 10-s interval, and if it was below 10 or 2 Hz, an additional 10-or 20-s recording interval was added, respectively. Next, we determined the tuning curve by varying the frequency and intensity of 15-ms tone bursts by an automatically adapting procedure (9). For tuning curves, spike detection was performed online based on a manually set trigger, whereas all other data were analyzed offline using custom-written Matlab routines using waveform-based spike detection. For rate-intensity function, we varied the intensity of 50-ms tone bursts at characteristic frequency (CF) in 5-dB steps between 20 dB below and 50 dB above threshold with 25 repeats for each intensity; the dynamic range was calculated as in ref. 46.
Data Analysis and Statistical Tests. Data are presented as mean ± SEM unless otherwise specified. Normality was assessed with the Jarque-Bera test. F test was used to assess equality of variance in normally distributed data sets. The unpaired, two-tailed Wilcoxon rank sum test (also known as Mann-Whitney U test) was used to compare data of nonnormal distribution or when variances of experimental groups were unequal. In the case of normally distributed equal-variance data, Student's unpaired two-tailed t test was used