The Presynaptic Ribbon Maintains Vesicle Populations at the Hair Cell Afferent Fiber

1 The presynaptic ribbon maintains vesicle populations at the hair cell afferent fiber

2 synapse

3 Lars Becker1, Michael E. Schnee1, Mamiko Niwa1*, Willy Sun2, Stephan Maxeiner3**,

4 Sara Talaei1, Bechara Kachar2, Mark A. Rutherford4, Anthony J. Ricci1,3

5 1 Department of Otolaryngology Stanford University, United States, 2 National Institute

6 of Deafness and Communicative Disorders, United States, 3 Molecular and Cellular

7 Physiology, Stanford University, United States, 4 Department of Otolaryngology,

8 Washington University in St. Louis, United States,

9 *Present address: Department of Otolaryngology Johns Hopkins University, United

10 States

11 ** Present address: Department of Neuroanatomy, Institute for Anatomy and Cell

12 Biology, Medical School Saarland University, Germany

20 Abstract: The ribbon is the structural hallmark of cochlear inner hair cell (IHC) afferent

21 synapses, yet its role in information transfer to spiral ganglion neurons (SGNs) remains

22 unclear. We investigated the ribbon’s contribution to IHC synapse formation and

23 function using KO mice lacking RIBEYE. Despite loss of the entire ribbon structure,

24 synapses retained their spatiotemporal development and KO mice had a mild hearing

25 deficit. IHCs of KO had fewer synaptic vesicles and reduced exocytosis in response to

26 brief depolarization; high stimulus level rescued exocytosis in KO. SGNs exhibited a

27 lack of sustained excitatory postsynaptic currents (EPSCs). We observed larger

28 postsynaptic glutamate receptor plaques, potentially in compensation for the reduced

29 EPSC rate in KO. Surprisingly, large amplitude EPSCs were maintained in KO, while a

30 small population of low amplitude slower EPSCs was increased in number. The ribbon

31 facilitates signal transduction at physiological stimulus levels by retaining a larger

32 residency pool of synaptic vesicles.

35 Key words: auditory hair cell, synaptic transmission, RIBEYE, glutamate receptor,

36 multivesicular release

41 Introduction

42 The synaptic ribbon is an electron dense structure associated with presynaptic active

43 zones at sensory synapses in the retina, lateral line and inner ear organs. RIBEYE, the

44 major protein associated with the ribbon (Zanazzi and Matthews 2009), consists of a

45 unique amino-terminal A domain that forms the ribbon scaffold and a carboxy-terminal B

46 domain with lysophosphatidic acid acyltransferase activity which is almost identical to

47 CtBP2, a transcriptional co-repressor (Schmitz, Konigstorfer et al. 2000). The ribbon is

48 anchored to the presynaptic membrane by bassoon protein in mammalian IHCs

49 (Khimich, Nouvian et al. 2005) as well as retinal photoreceptors and bipolar cells (dick

50 2001). Although it is comprised predominantly of RIBEYE (Zenisek, Horst et al. 2004),

51 the ribbon complex may associate with over 30 different synaptic proteins (Uthaiah and

52 Hudspeth 2010, Kantardzhieva, Peppi et al. 2012).

53 IHCs respond to sound with graded receptor potentials that drive rapid and precise

54 synaptic release, encoding intensity and timing information (Fuchs 2005, Matthews and

55 Fuchs 2010). The ribbon is thought to sustain continuous encoding by tethering a pool

56 of vesicles that replenishes the local pool of docked and primed vesicles upon

57 depletion. Other functions ascribed to the ribbon include: regulating vesicle trafficking

58 from the cytoplasm to the active zone, serving as a vesicle trap to bind freely diffusing

59 vesicles, a location for vesicle formation, a physical barrier for Ca2+ ensuring a high

60 local Ca2+ signal, and as a primary structural component of the synapse anchoring Ca2+

61 channels close to synaptic vesicle release sites (Parsons 1994, Frank, Rutherford et al.

62 2010, Graydon, Cho et al. 2011, Kantardzhieva, Peppi et al. 2012).

63 Postsynaptic EPSCs at ribbon synapses range in amplitude by up to 20X. The smallest

64 and largest EPSCs have similar kinetics, which together suggested the existence of a

65 mechanism to synchronize the release of multiple quanta (multivesicular release, MVR;

66 (Glowatzki and Fuchs 2002, Li, Keen et al. 2009, Schnee, Castellano-Munoz et al.

67 2013, Rudolph, Tsai et al. 2015). By holding vesicles and voltage-gated Ca2+ channels

68 at the active zone the ribbon may support MVR, occurring through various mechanisms

69 including: i) coordinated fusion of multiple single vesicles, ii) release of large pre-fused

70 vesicles, iii) rapid sequential fusion in which the 1st fusion event triggers fusion of

71 additional vesicles, or some combination of these mechanisms (Matthews and Fuchs

72 2010).

73 Previous attempts to study ribbon function used genetic disruption of the anchoring

74 protein Bassoon or acute destruction of the ribbon by fluorophore assisted light

75 inactivation (FALI). These manipulations resulted in smaller vesicle pool size, calcium

76 channel mislocalization, increased auditory thresholds, asynchrony in neural responses

77 and reductions in both tonic and phasic responses (Snellman, Mehta et al. 2011, Jing,

78 Rutherford et al. 2013). However, neither of those manipulations were selective for the

79 ribbon and so might be manifestations of accessory protein denaturing.

80 Recently, manipulation of the gene encoding CtBP2/RIBEYE allowed for deletion of

81 RIBEYE by removing the A-domain of ctbp2 to eliminate ribbons in mice, enabling a

82 more precise examination of ribbon function. This mouse will be referred to as KO and

83 the littermate controls as WT throughout the manuscript. Elimination of ribbons in the

84 retina reduced phasic and tonic transmission and rendered release sites sensitive to the

85 slow calcium buffer EGTA (Maxeiner, Luo et al. 2016). Here we present analysis of

86 cochlear synaptic anatomy, hair cell and afferent fiber physiology and hearing function

87 in the RIBEYE knockout mouse. Results are distinct from those in retina and from a

88 RIBEYE deletion in zebrafish (Lv, Stewart et al. 2016) , suggesting a more subtle

89 function for the ribbon than previously assumed.

90 Results:

91 IHC-Spiral Ganglia Neuron (SGN) synapses are formed and maintained despite

92 absence of the ribbon: The inner ear structure of the KO was assessed first for the

93 presence of synaptic ribbons and synapses (Maxeiner, Luo et al. 2016). Validating loss

94 of hair cell ribbons was accomplished using immunohistochemistry for the RIBEYE B-

95 domain (CtBP2). Anti-CtBP2 labelled the ribbons in wildtype (WT) but not in the KO

96 animals (postnatal day 21, P21)(Fig. 1A). Nuclear CtBP2 was still observed in both WT

97 and KO (not shown) because the KO is specific to the RIBEYE A-domain. The nuclear

98 staining served as a usful control for antibody function.

99 Using antibodies to Homer for postsynaptic labeling and Bassoon for presynaptic

100 labeling (Fig. 1B,C) we quantified the number of puncta and whether puncta were

101 juxtaposed as synaptic pairs (Fig. 1I,J). Bassoon and Homer puncta were similar in

102 number between WT and KO (Bassoon: 40 ± 15 in WT, vs. 43 ± 12 in KO, p = 0.76;

103 Homer: 17 ± 3 for both WT and KO; WT: n = 20 mice, KO: n = 19 mice, p = 0.64). The

104 number of paired Homer and Bassoon puncta (Fig. 1D) were similarly comparable:

105 14 ± 2 (n = 20) for WT and 13 ± 2 (n = 19, p = 0.18) for KO IHCs, suggesting that

106 synapses were formed and maintained in comparable numbers regardless of the

107 presence of ribbons. The higher number of Bassoon puncta likely reflects its localization

108 at efferent synapses. Figure 1- Supplemental Fig. 1 presents a lower magnification

109 image that illustrates the multiple roles of Bassoon at the auditory periphery. Bassoon is

110 a major component of the presynaptic elements at the auditory periphery including the

111 efferent fibers and the hair cell afferent synapses. Thus it is not surprising that paired

112 Bassoon puncta are much less than total puncta because Bassoon serves multiple

113 synapses. Similarly, when interpreting Bassoon KO data it is not surprising that its

114 phenotype may be more dramatic than simply the RIBEYE effect alone due to number

115 of synapses involved and schaffolding function of Bassoon that leads to structural

116 changes at the synapse.

117 RIBEYE KO in zebrafish revealed a non osmiophilic structure resembling a ribbon that

118 maintained organization of synaptic vesicles in the absence of the RIBEYE protein in

119 hair cells of the lateral line neuromasts (Lv, Stewart et al. 2016). We examined whether

120 such a structure existed in the mammalian cochlea in the IHCs of the KO using

121 transmission electron microscopy (TEM) and electron-tomography. Representative

122 examples of conventional TEM (Fig. 1E, 70 nm sections) and tomographic

123 reconstructions (Fig. 1F,G at 10 nm tomographic sections) are presented. The KO

124 presynaptic synapses consistently showed a small osmiophilic density in contact with

125 the pre-synaptic membrane at the center of the synapse that resembled the arciform

126 density seen at ribbon synapses in the retina. This density may be the Bassoon

127 enriched linker between the membrane and the ribbon. Interestingly, this density is not

128 observed in the Bassoon KO mouse (Frank et al., 2010). Tomographic reconstructions,

129 (Fig. 1G, right) clearly show the presynaptic density, which is smaller than the ribbon

130 and is present also at ribbon synapses in WT IHCs (Fig. 1G, left). We found no spatial

131 organization of vesicles near active zones in KO IHCs, suggestive of the ‘ghost’ ribbons

132 observed in the zebrafish lateral-line hair cells (Lv, Stewart et al. 2016). Thus, in the KO

133 mouse, IHCs lack synaptic ribbons and have no obvious compensatory structures. We

134 surmise that the differences observed in the zebrafish model are a function of the KO

135 being incomplete or there being a compensation mechanism in zebrafish, not found in

136 mammal.

137 KO IHCs may upregulate a similar protein to replace RIBEYE loss. CtBP1 is a likely

138 candidate because it is similar to the RIBEYE-B domain, is expressed in hair cells and

139 localizes to the synaptic region (Kantardzhieva, Liberman et al. 2013). Antibody labeling

140 (Fig. 1H,K,L) demonstrates the presence of CtBP1 in both WT and KO animals which

141 overlapped with Homer in WT. A reduction in puncta number was identified in the KO

142 animal due to loss of localization at the afferent synapses (Fig. 1K,L). Thus, the CtBP1

143 localization at the synapse requires the presence of RIBEYE, and CtBP1 does not

144 substitute for RIBEYE in the KO animal.

145 Synaptic vesicles are normal in size but fewer in number: Vesicle diameters

146 (distance between the outer edges of the bilayer in each vesicle) were measured in the

147 reconstructed tomograms. We measured in both X and Y directions with no differences

148 observed. In the WT the average vesicle diameter was 33.6 ± 3 nm (n = 21) and in the

149 KO it was 33.8 ± 3 nm (n = 21).

150 Synaptic vesicles were conspicuously fewer in number the KO. Vesicles were counted

151 in volumes around the center of the reconstructed synapses (Fig. 1G), where one

152 boundary was the presynaptic membrane. The volumes had XY dimensions of 453 ± 6

153 nm for WT (n = 3) and 569 ± 117 nm for KO (n = 11), and Z dimensions of 88 ± 3 and

154 95 ± 10 nm for WT and KO, respectively. Total vesicle numbers were 20 ± 2 for WT and

155 7 ± 3 for KO, significantly greater for WT compared with KO synapses (p < 0.001, Welch

156 correction for unequal variance). Our TEM data is qualitatively similar to that reported in

157 the companion paper and supports the conclusion that ribbons are not present in the

158 KO.

159 Auditory nerve activity is reduced in absence of synaptic ribbons: Previous

160 investigation of ribbonless synapses in the KO mouse focused on the retina and did not

161 assess function at the systems level (Maxeiner et al., 2016). To determine how hearing

162 was affected, we measured ABRs and distortion product otoacoustic emissions

163 (DPOAEs). DPOAEs are a direct measure of outer hair cell function and cochlear

164 amplification. DPOAEs are not expected to be affected by absence of RIBEYE and

165 serve as a control for developmental acquisition of mature cochlear mechanics. DPOAE

166 threshold measurements were not different across frequencies from 4 - 46 kHz (Fig.

167 2A). Data are presented in decibels (dB) as referenced to sound pressure level (SPL)

168 with lower values indicating enhanced sensitivity in the mid-cochlea. Normal DPOAEs

169 indicate that cochlear amplification was not effected mechanically by the loss of outer

170 hair cell ribbons, suggesting that sound-driven activation of IHCs was intact, and that

171 any differences in auditory nerve activity should result from changes at the IHC afferent

172 synapses.

173 Wave 1 of the ABR is a direct measurement of the synchronized firing of SGNs at

174 sound onset, triggered by synaptic output from cochlear IHCs. The subsequent waves

175 of the ABR reflect the ensuing activation of brainstem nuclei in the ascending auditory

176 pathway (Zheng, Henderson et al. 1999). ABR thresholds were elevated by

177 approximately 10 dB at all probe frequencies in KO mice (Fig. 2B), except for 48 kHz

178 where hearing threholds are already relatively high. The variance in our control data

179 ranged from 8-12dB depending on frequency; this dictated the sample size needed to

180 detect a 5% change as statistically different. ABR threshold was identified as the

181 stimulus intensity required to evoke a voltage response 5x the RMS noise floor for the

182 measurement. The companion paper found a similar elevation in threshold though it did

183 not reach statistical significance. The difference is likely due to sampling size where we

184 powered our sample sized based on variance found in control measurements. The ABR

185 waveform (Fig. 2C: 23 kHz, 60 dB) shows a large reduction in wave 1 amplitude with no

186 effect on the subsequent peaks. This was surprising and suggests considerable

187 plasticity in central processing of data translated from cochlea periphery. KO mice had

188 smaller wave 1 peak amplitudes, for all sound levels that evoked a response (Fig. 2D).

189 When wave 1 growth functions were plotted relative to the maximal response amplitude

190 within genotype, the amplitude versus level function for KO scaled up to overlap with

191 WT (Fig. 2E). Considering wave 1 amplitude as a fraction of the maximal amplitude, the

192 overlap of the normalized wave 1 growth curves suggested that loss of maximum output

193 could produce an elevation of threshold in the ABR measurement that does not

194 necessarily reflect a sound detection limit or a change in sensitivity of the auditory

195 system. Indeed, normal ABR amplitudes beyond wave 1 indicate compensation in the

196 cochlear nucleus. Future work should determine if KO mice have increased behavioral

197 thresholds.

198 In mice lacking Bassoon function, where IHCs lacked most of their membrane-anchored

199 ribbons, there was an increase in wave 1 latency (Khimich, Nouvian et al. 2005) and in

200 first-spike latency in recordings of SGNs in vivo (Buran, Strenzke et al. 2010). ABR

201 wave 1 latencies were not different between WT and KO at any stimulus intensity (Fig.

202 2F), suggesting a milder phenotype in the KO as compared to Bassoon.

203 As the ABR measurements are averaged responses of 260 stimulus repetitions, a

204 possible explanation for the reduced ABR wave 1 amplitude in KO is that the response

205 reduces over time. If the ribbon is providing a means of maintaining vesicles near the

206 synapse, the concomitant loss of vesicles at ribbonless synapses might reduce the

207 ongoing response to repetitive stimulation. Comparison of ABR amplitudes where the

208 number of repeats was varied from 260 to 1000 showed no difference in amplitudes, nor

209 did repeated measurements meant to evoke depletion.Thus, rundown alone is unlikely

210 to account for the decrease in ABR wave 1.

211 Characterization of molecular synaptic components indicates an increase in

212 postsynaptic glutamate receptor distribution: To begin identifying cellular

213 mechanisms that might underlie the measured reduction in ABR wave 1 amplitude we

214 investigated the distribution of presynaptic Ca2+ channels and postsynaptic glutamate

215 receptors as indicators of functional synapses. Figure 3 presents immunohistochemical

216 localizations of presynaptic ribbons with anti-CtBP2, presynaptic voltage-gated Ca2+

217 channels with anti-CaV1.3, and postsynaptic AMPA-type glutamate receptors with anti-

218 GluA3 in P35 mice organ of corti. There was no difference in the number of puncta

219 between WT and KO for CaV1.3 (20 ± 6, n = 9 vs. 18 ± 5, n = 13, p = 0.51) or GluA3 (15

220 ± 3, n = 9 vs. 15 ± 2, n = 13, p = 0.43) (Fig. 4A,B). There was also no difference in the

221 number of paired synaptic puncta between genotypes, with values of 14 ± 2 (n = 9) for

222 WT and 13 ± 2 (n = 13) for KO (p = 0.13). Thus, the presence of the ribbon did not

223 influence the number of apparently functional synapses per IHC.

224 Careful inspection of GluA3 immunoreactivity showed that the glutamate receptor

225 puncta were larger on average in the KO (Fig. 4D). Puncta volumes were measured for

226 presynaptic markers CaV1.3 and Bassoon as well as postsynaptic markers Homer and

227 GluA3 (Fig. 4C). Like GluA3, Homer puncta were significantly larger (KS-test,

228 p ≤ 0.0001) on average in the KO (Fig. 4D). The volumes of CaV1.3 and Bassoon

229 distributions were very similar among genotypes however the volume of Bassoon was

230 significantly higher in KO. Most WT and KO synapses exhibited the typical elongated

231 stripe-like appearance of CaV1.3 puncta. However, we noted a tendency for KO

232 synapses to have CaV1.3 clusters that were fragmented into smaller puncta (Fig. 3C,

233 arrow; see also Companion Paper). These anatomical data support the hypothesis that

234 the ribbon is not a major regulator of synapse formation or stability; however, it does

235 perhaps suggest that postsynaptic receptors are sensitive to activity and the increased

236 plaque size might reflect a compensation for reduced activity.

237 Basolateral IHC properties: We compared the properties of the IHC voltage-gated

238 Ca2+ current (Fig. 5), which drives exocytosis. We first blocked outward K+ currents

239 using Cs+ and tetraethylammonium (TEA) in the internal solution to isolate Ca2+ currents

240 in WT and KO IHCs (Fig. 5A,B). Current-voltage plots showed no difference in voltage

241 of half-activation (V1/2) or maximum current amplitude (Fig.5C-E). Half activation

242 voltages were -34 ± 0.3 mV (n = 18, WT) and -33.4 ± 0.3 mV (n = 26, KO) for animals

243 P10-P13. The companion paper found small changes in voltage dependent properties

244 of the calcium channels at older ages and in higher external calcium; differences that

245 might suggest compensatory effects of ribbon loss or later specializations being ribbon

246 dependent. Examples of current elicited with K+ as the major internal cation and without

247 TEA are presented in Fig. 5F,G. Recordings were made at P20-24. No differences were

248 observed in maximal current or in V1/2 (Fig. 5H). Zero-current potentials monitored in

249 current-clamp mode (Fig. 5I) were not different, with means of -63 ± 10 mV (n = 5, WT)

250 and -65 ± 8 mV (n = 8, KO).

251 Release is slower and smaller for naturalistic stimuli in absence of the ribbon: We

252 measured presynaptic exocytosis with the two-sine wave method for tracking vesicular

253 fusion by monitoring real time changes in cellular membrane capacitance (Cm) from

254 IHCs (Santos-Sacchi 2004, Schnee, Castellano-Munoz et al. 2011). The two-sine

255 method allows continuous monitoring of Cm during the depolarizing stimulus as

256 compared to a before and after monitoring of capacitance changes, which is particularly

257 helpful during smaller, longer depolarizations (Schnee, Santos-Sacchi et al. 2011).

258 Typically, small depolarizations lead to relatively slow, linear increases in Cm during

259 depolarization. In contrast, larger depolarizations induce an abrupt departure from

260 linearity termed the ‘superlinear’ response (Schnee et al. 2011). Both types of

261 responses were present in recordings from IHCs of WT and KO mice alike (Fig. 6A,B).

262 From Cm measurements, we estimated the sizes of synaptic vesicle pools, their rates of

263 release and the Ca2+-dependence of release. In response to strong depolarizations, we

264 found no difference in maximal release between genotypes for either the linear or

265 superlinear release components (Fig. 6C). The linear component was 52 ± 24 fF (n =

266 15, WT) compared to 42 ± 22 fF (n = 15, KO); and the superlinear component of release

267 was 223 ± 79 fF (n = 17, WT) compared to 253 ± 116 fF (n = 17, KO). Likewise, no

268 differences in release rates (Fig. 6D) were observed for the linear component: 101 ± 72

269 fF/s (n = 15, WT) compared to 75 ± 50 fF/s (n = 14, KO) or for the superlinear

270 component: 256 ± 164 fF/s (n = 17, WT) and 253 ± 117 fF/s (n = 18, KO). In addition,

271 the Ca2+-dependence of release for the linear component was not different (0.9 ± 0.03

272 fF/pC (n = 37, WT) as compared to 0.93 ± 0.03 fF/pC (n = 34, KO). The voltage at which

273 the superlinear component of release first appears was also not different between mice

274 (Fig. 6E). Fitting the data with a simple Gaussian function gave values of -40 ± 9 mV for

275 WT and -41 ± 3 mV for KO for the center values. However the width of the functions

276 were different, with values of 7.6 ± 2 for WT compared to 14.3 ± 3 for KO (values p<

277 0.002 level). Thus, although the variance in release properties increased for the KO

278 animal, there were no major differences in the ability of the synapses to respond to

279 strong stimuli.

280 In contrast to the comparable maximal responses, smaller depolarizations revealed a

281 reduction in release at early time points in the KO. The difference observed was

282 masked at longer time points or by stronger depolarization (Fig. 6F-J). Comparing

283 capacitance changes at short intermediate and longer time points in response to smaller

284 depolarizations showed that the KO consistently had reduced release as compared to

285 the WT data. Fitting the data in Fig. 6G,H with a simple exponential demonstrated that

286 although both genotyes reached the same values of release with strong depolarizations

287 the relationships varied in two ways. For the shortest time point release started at -40

288 mV as compared to -45 mV with all other variables remaining constant. Given that the

289 numbers for release are quite small and variable, conclusions should be conservative,

290 but support the argument that fewer vesicles are available for release at this early time

291 point and recruitment is minimal. For the longer time point Fig. 6H, the voltage-

292 dependence (Cm/Vm) varied with the KO, 7.3 ± 3.6 for WT as compared to 12.7 ± 5.8

293 for KO. F-tests show these fits to be different at the p < 0.002 level. Longer time points

294 as depicted in Fig. 6I,J show no difference between groups. In addition, for smaller

295 depolarizations the delay to release onset was greater in the KO IHCs, consistent with a

296 reduction in vesicles immediately available for release (Fig 6K). For example,

297 depolarizations to -44 mV resulted in significantly shorter latency in WT (102 ± 98 ms, n

298 = 16) compared to KO (204 ± 166 ms, n = 17, p = 0.039). Stronger depolarizations (> -

299 39 mV) resulted in no difference in release latency. The time constants of endocytosis

300 were not significantly different (6.0 ± 4s, n = 13, WT and 7 ± 5 s, n = 13, KO). Together

301 these data suggest that fewer vesicles were available for release at ribbonless

302 synapses, but that with stronger depolarizations or longer time courses of depolarization

303 the recruitment of additional vesicles masked this immediate difference in release

304 latency and amplitude.

305 The ribbon may be a barrier to Ca2+ diffusion, creating a restricted space for sharp Ca2+

306 nanodomains beneath the ribbon (Roberts 1994, Graydon, Cho et al. 2011).

307 Intracellular Ca2+ buffering can alter release properties depending on the relative

308 distribution of Ca2+ channels with release sites. In retina the KO increases buffering

309 efficacy (Maxeiner et al. 2016), suggesting looser coupling between Ca2+ channels and

310 release sites. To investigate the effects of ribbon absence on stimulus-secretion

311 coupling in p11 to p13 IHCs, we compared between genotypes the effects on release

312 properties of changing concentration of the slow Ca2+ buffer EGTA (Fig. 7). Examples of

313 release obtained with either 0.1 or 5 mM EGTA are presented in Fig. 7A (control

314 experiments all use 1 mM EGTA). Although increasing buffer concentration reduced

315 release and lowering buffer increased release, the relative changes were comparable

316 between WT and KO mice for release measured at 1 sec (Fig. 7B). Data previously

317 demonstrated that increasing Ca2+ buffering increased the time to superlinear release

318 (Schnee, Santos-Sacchi et al. 2011). Fig. 7C demonstrates that this effect is retained in

319 the KO and that there are no major differences between WT and KO modulation by

320 Ca2+ buffering. These results contrast with findings from retinal bipolar cells and in the

321 zebrafish KO (Maxeiner, Luo et al. 2016); see Discussion). Similarly, release rates

322 measured for steps to -45 mV were not significantly different for the initial release

323 components (Fig. 7D). We further investigated the first component of release, typically

324 thought to represent vesicles in early release pools, not requiring trafficking. Comparing

325 release at two voltages again showed that the buffer greatly altered release properties

326 but not in a differential way between genotypes. However, it should be noted that the

327 difference in early release observed in Fig. 6 with 1 mM EGTA was lost with 0.1 mM

328 EGTA. Thus, it is possible that a sensitivity exists that is below a clear detection limit for

329 our capacitance measurements.

330 EPSCs are more heterogeneous in size and kinetics in the absence of the ribbon:

331 Cm measurements are a powerful tool for investigating whole-cell exocytosis but they

332 lack sensitivity (1 fF reflects fusion of ~20 vesicles), they reflect information from all

333 synapses and they can be quite variable. To obtain a more refined view of activity from

334 individual synapses we recorded postsynaptically, from the afferent fiber bouton. Hair

335 cells bathed in artificial perilymph lack the standing inward current through

336 mechanoelectric transduction channels in hair bundles, causing the resting membrane

337 potential to be relatively hyperpolarized (Farris, Wells et al. 2006, Johnson, Kennedy et

338 al. 2012). To depolarize hair cells and increase the rate of vesicle fusion, we applied a

339 high potassium (40 mM K+) solution to depolarize the hair cell. K+ application to the hair

340 cell in current-clamp mode depolarized hair cells by an average of 32 ± 5 mV (n = 8;

341 Fig. 8A). There was no difference in depolarization between genotype thus allowing

342 conclusions regarding postsynaptic responses to be ascribed to presynaptic effects.

343 Afferent boutons of type I auditory nerve fibers were recorded ex vivo at ages of P17-

344 P21. Due to the approach vector of the recording pipette, recordings were made from

345 boutons located on the side of the IHC facing the ganglion (modiolar face) rather than

346 the pillar cell side. The start of the high K+ response was defined when the mean EPSC

347 rate exceeded the rate in normal external solution by 2 standard deviations. Examples

348 of postsynaptic responses to K+ applications are provided in Figure 8B-D. Two types of

349 responses were observed, sustained responses (Fig. 8B,C) and transient responses

350 (Fig. 8D). WT responses were mainly of the sustained type (7 of 8) while the KO

351 responses had a smaller proportion of sustained responses (7 of 11). EPSC rates are

352 plotted for transient (Fig. 8E) and sustained responses (Fig. 8F) for both WT and KO.

353 EPSC rates in 40 mM K+ perilymph were quantified over one second windows at 100

354 ms intervals and the means are displayed as thicker lines (Fig. 8F). The peak release

355 rates varied considerably across fiber populations (0-120 s-1) and there was a tendency

356 for the KO neurons to reach a lower peak EPSC rate on average: 57 ± 30 s-1, (n = 11,

357 WT) compared with 44 ± 31 s-1, (n = 18, KO) (Fig. 8G).

358 The time to peak EPSC rate was less variable than absolute rates (Fig. 8H). The KO

359 had one outlier that steadily increased throughout the K+ application, removal of this

360 point from the statistical analysis resulted in the KO reaching peak EPSC rate sooner (9

361 ± 10 sec, n = 10) than WT (21 ± 12 sec, n = 9; p = 0.037). Observing the sustained

362 release group shows that both WT and KO reduced EPSC rate over time but that the

363 KO reduced more completely than WT (Fig. 8G). 8/11 KO reduced to less than 5% of

364 peak as compared to 3/8 WT. Integrating the rates over time (Fig. 8I) confirmed an

365 increase in total EPSCs in WT (1.8 ± 1.1 x 103) compared to KO (0.81 ± 0.66 x 103, p =

366 0.021).

367 EPSC amplitude distributions were generated for each recorded fiber. Frequency

368 histograms of EPSC amplitudes could be fit with either a double Gaussian function or a

369 Gaussian and gamma function where the smallest peak was always fit by a Gaussian

370 function (e.g. Fig. 8J,K). The area under the smaller peak was calculated and plotted as

371 a percentage of the total in Fig. 8L, where the KO showed a statistically larger (p =

372 0.034) percentage of small events with 5.5 ± 5.6% (n = 12) as compared to 1 ± 0.7% (n

373 = 7). The increase in number of small events might suggest less coordination in release

374 because there are fewer vesicles filling release sites. It might also suggest that smaller

375 responses are visible because of an increased in receptor plaque size as well. Only

376 fibers with more than 100 EPSCs were included for this analysis. No difference was

377 found in the proportion of simple events for each unit, where a simple event is described

378 as having a single rise and decay phase (Fig. 8M). We also compared the median

379 amplitudes (Fig. 8N). Although there was a tendency for EPSCs in the KO to be greater

380 in amplitude 212 ± 46 pA (n = 13) compared to 173 ± 36 pA (n = 7), the values were not

381 statistically different. A difference here would support the argument of a postsynaptic

382 compensatory increase in glutamate receptor plaque size. The coefficient of variation

383 (CV) for the amplitude distributions were different on average (p = 0.017; Fig. 8O). The

384 control values are similar to those obtained by (Li, Keen et al. 2009, Grant, Yi et al.

385 2010). Recordings with a larger proportion of smaller events had larger CV. When small

386 events were removed from the histogram, the CVs were not statistically different

387 between KO and WT. The postsynaptic responses in KO also had a greater proportion

388 of EPSCs > 350 pA (Fig. 8J,K, 9D): 1.3 ± 1.3% in WT compared to 8.0 ± 10% for KO

389 EPSCs (t-test for unequal variances, p = 0.04). The change in amplitude was not

390 accompanied by a change in kinetics. The increased amplitude might reflect the

391 increase in postsynaptic glutamate receptor plaque (Fig. 3,4).

392 Examples of EPSCs are presented in Fig. 9A to show the variation in amplitudes and to

393 identify a smaller population of EPSCs with slower kinetics (see inset). As the ribbon is

394 proposed to be involved with multivesicular release (MVR), and the kinetic similarities

395 between large and small EPSCs are a signature of MVR, we investigated the decay of

396 EPSCs in auditory nerve fiber recordings from WT and KO animals. Fitting the EPSC

397 decay with single exponentials provided time constants (Fig. 9B). Calculating the CVs of

398 EPSC decay time constants (Fig. 9C) revealed more variability in KO but no significant

399 difference compared to WT. The skewness (Fig. 9F) of the decay time constant

400 distributions was significantly larger for the KO (0.46 ± 0.2, n = 13) compared to WT

401 (0.29 ± 0.07, n=7; p = 0.04); the KO mean EPSC decay time was greater than the

402 median decay time. As a population 0.44% of WT EPSCs were < 100 pA in amplitude

403 with decay time constants > 1 ms while 1.9% of KO events fell in that category. Plotting

404 the decay time constant against amplitude for each EPSC in all fibers (Fig. 9D) (WT: n =

405 7 recordings, 15,564 EPSCs; KO: n = 13 recordings, 17,210 EPSCs) identifies a

406 population of small slow EPSCs in the KO that are present in greater proportion than in

407 WT (upper left quadrant). The mean EPSC decay time constants plotted as a function of

408 EPSC amplitude for each recording (bin of 10 pA) illustrates that all the KO fibers had a

409 proportion of EPSCs with slow time courses while the WT have only 2 fibers with these

410 slower events (Fig. 9E). An example of the smaller, slower EPSC is provided in the

411 inset of Fig. 9A. The distributions of EPSC durations for all EPSCs < 100 pA (Fig. 9G,H)

412 demonstrates the shift toward slower longer duration EPSCs in KO fibers (p > 0.001,

413 using the Kolmogorov-Smirnov test). The cumulative probability functions of EPSC

414 duration for all fibers by genotype (thick red and blue lines) illustrates the larger

415 proportion of slow EPSCs in the KO fibers (Fig. 9H). These data may also support the

416 argument for a larger postsynaptic glutamate receptor plaque, where diffusion of

417 glutamate, coupled with the lower glutamate concentration might serve to slow the

418 decay kinetics.

419 Discussion:

420 The KO mouse provides an unprecedented opportunity to investigate the functional

421 relevance of synaptic ribbons in driving information transfer at the primary auditory

422 synapse. Synapse formation and stability appeared normal based on synapse counts

423 and protein localization, despite complete absence of the ribbon (Figs. 1, 3-4). No

424 evidence was found for partial or defective ribbons structures, nor did we detect

425 compensation by other proteins in the absence of RIBEYE (Fig. 1). Also, cochlear

426 mechanics appeared normal as assessed by DPOAE (Fig. 2). Thus, this mouse

427 provides an excellent model to investigate the ribbon’s role in synaptic transmission.

428 Whole animal function: ABR responses reported a 10 dB threshold shift and about a

429 50% reduction in ABR amplitude with no significant effect on the shape of the

430 waveforms or on the delay (Fig. 2). These data agree qualitatively with the companion

431 paper data, the lack of significance likely being a function of the sample size. The minor

432 auditory deficit may not be expected to produce a change in behavioral sound detection

433 threshold as the ensuing ABR peaks, representing brainstem nuclei, were unaffected.

434 Alternatively, central compensation might occur that mask peripheral deficits. These

435 data suggest that the ribbon is not essential for setting hearing threshold, and that its

436 role in auditory function is more nuanced than for example otoferlin which when absent

437 or altered leads to deafness with threshold elevations of more than 100 dB (Roux,

438 Safieddine et al. 2006, Pangrsic, Lasarow et al. 2010). Further work is needed to better

439 understand ribbon function at the systems level. Additional testing to better investigate

440 auditory processing such as sound localization experiments may provide additional

441 insight into how the ribbon might serve systems level function.

442 Synapse formation: Synapse formation is surprisingly normal with appropriate number

443 of synapses and synaptic markers present (Figs. 1,4). Little change in Ca2+ channel

444 distribution or biophysics were observed in the KO (Fig. 5) relative to Bassoon

445 disruption, suggesting that Ca2+ channel distribution or regulation does not depend

446 critically on the ribbon (Frank, Rutherford et al. 2010). Our companion paper identifies

447 minor changes in Ca2+ channel distribution and voltage sensitivity at later ages; whether

448 these are a function of ribbon absence or activity dependent modulation remains to be

449 explored. The lack of effect on calcium channel properties and distributions at early

450 ages supports the hypothesis of a compensatory mechanism occurring later during

451 synapse maturation. Given the reports of late development changes in ribbon size and

452 glutamate receptor plaque sizes, activity driven shaping of the synapses is a reasonable

453 expectation (Fernandez, Jeffers et al. 2015). Postsynaptically, both Homer and the

454 glutamate receptors had larger volumes in the KO animal than in WT (Fig. 4). This

455 change also may be an indication of activity dependent synaptic shaping where the

456 reduced activity in the KO animal results in a postsynaptic up regulation of receptors.

457 The increase in the population of larger amplitude EPSC’s in KO animals further

458 supports this conclusion (Figs 8I,J, 9D).

459 Comparison to other systems: In zebrafish a frameshift mutation was used to

460 eliminate ribeye from ribbon synapses. This work revealed ghost like ribbons where

461 vesicles were organized about invisible ribbon-like structures (Lv C et al.,2016). Neither

462 our investigations, nor the companion paper nor the original work investigating the

463 retina (Maxeiner et al., 2015) found comparable ghosts in this mouse. We would

464 conclude that either the frame shift approach did not account for all ribeye or there is a

465 strong compensatory mechanism in zebrafish that is not found or is much weaker in

466 mammal. We investigated CTBP1 as a potential compensatory molecule but as Fig. 1

467 shows there is no labeling near synapses in the KO mouse. The presence of ribbons in

468 photoreceptors of zebrafish support the idea that there are fundamental differences

469 between zebrafish and mammalian systems. Additionally, work in zebrafish suggests a

470 tight reciprocal relationship between calcium channel numbers, calcium current

471 amplitude and ribbon size (Lv, Stewart et al. 2016). As no changes in calcium channel

472 distribution or numbers were observed in the KO animal despite the loss of the ribbon,

473 this relationship does not appear to exist in mammal or is greatly reduced, again

474 supporting fundamental differences between model organisms.

475 Mechanism of ABR amplitude reduction: The decrease in ABR amplitude is likely a

476 function of the reduced frequency of EPSCs leading to fewer action potentials

477 generated (Fig. 8F). In addition, the increase in proportion of EPSCs with lower

478 amplitudes and longer durations (Fig. 9H) is expected to increase the likelihood of spike

479 generation failure (Rutherford, Chapochnikov et al. 2012). As larger EPSCs are better

480 phase-locked than smaller EPSCs (Li, Cho et al. 2014) and present works finds a 5-

481 10% increase in smaller EPSCs one might predict a lower vector strength and reduced

482 phase locking. Our companion paper in fact does observe these differences in single

483 unit firing. Similarly the increased population of smaller EPSCs is reminiscent of

484 observations in young animals when compared to older animals and so might reflect a

485 developmental delay in the KO animal. Fewer fibers fire together at sound onset,

486 reducing auditory nerve synchrony and wave 1 amplitude (Fig 2C,D). Measurements of

487 exocytosis from IHCs in response to small depolarizations, clearly showed a reduction

488 in release and an increase in time to release in the KO (Fig. 6E,F). This suggests a

489 decrease in vesicles available at the synapse for immediate release (supported by TEM

490 counts). The reduction in available vesicles for release as well as the population of

491 reduced amplitude EPSCs might also predict a spike timing deficit as described in the

492 companion paper. However, the more rapid response in WT was masked by stronger

493 stimulation (Fig. 6F), indicating that vesicle replenishment remained robust despite the

494 absence of the ribbon-associated synaptic vesicle pool. Given the robust nature of

495 vesicle trafficking, the primary role for the ribbon maybe in ensuring a population of

496 vesicles is present for the timing of onset responses, whereas vesicle trafficking can

497 serve to replenish vesicle populations during sustained release. A similar result was

498 reported by (Maxeiner, Luo et al., 2016) in retina.

499 Postsynaptic recordings suggest that the KO reaches peak release sooner than WT and

500 to a similar peak rate, implying that vesicle numbers are comparable for immediate

501 release. In contrast, TEM suggests fewer vesicles available at the synapse and the

502 small increase in population of small EPSCs also suggests a reduction in filled release

503 sites. Thus, it is possible that the ribbon acts as a barrier for vesicle interaction with the

504 release site while simultaneously serving to increase the total vesicles number

505 available. Interestingly, this idea was first suggested by (Furukawa and Matsuura 1978)

506 and later supported by (Schnee, Castellano-Munoz et al. 2013) as a possible

507 mechanism for explaining neural adaptation. However, the companion paper suggests

508 modifications at the synapse that might also be in accord with the postsynaptic

509 measurements here as well, i.e. more vesicles at the membrane. If there is an increase

510 in presynaptic density area and more vesicles per active zone, it might be predicted that

511 faster immediate release occurs but sustainability might be reduced, EPSC data also

512 supports this idea. To resolve these possibililities we need a better understanding of the

513 molecular underpinnings of the active zone and number of available release sites.

514 Synaptic vesicle replenishment for sustained release: Our data would argue against

515 a role for the ribbon in vesicle formation and replenishment. Total exocytosis, from both

516 the linear and superlinear components, was not reduced in the KO, implying a similar

517 total number of vesicles that could be recruited for release (Figs. 6A,B and 7A). In

518 contrast to zebrafish lateral line (Lv, Stewart et al. 2016), we measured no change in

519 vesicle size between KO and WT, indirectly supporting our view that the ribbon is not

520 regulating or involved in vesicle formation, a result corroborated by the companion

521 paper. Thus, no change in vesicle size, or maximal release supports the conclusion that

522 the ribbon is not involved in vesicle formation.

523 Ca2+ influx and coupling to exocytosis: The ribbon may act as a barrier to Ca2+

524 diffusion, creating a higher level of local [Ca2+] to synchronize vesicle fusion of multiple

525 vesicles (Graydon, Cho et al. 2011). In this case, the ribbon’s presence would reduce

526 the sensitivity of exocytosis to exogenous Ca2+ buffers (by elevating concentrations of

527 Ca2+). Likewise in retina from KO mice, EGTA reduced spontaneous release in the KO

528 but not in WT (Maxeiner, Luo et al., 2016). In contrast, our data shows that buffering

529 efficacy was not altered in the KO (Fig. 7), suggesting the ribbon does not play a major

530 role in regulation of intracellular Ca2+ concentration at IHC active zones. The similar

531 response between WT and KO might indicate a small effect though at a level below a

532 clear detection limit for capacitance measurements. It is possible that the geometric

533 differences and synaptic architecture between retina and cochlea alter the sensitivity to

534 Ca2+ buffers or that the effect noted in retina is indirect, for example EGTA-AM might

535 alter the presynaptic cell’s resting potential. Thus, our results are not fundamentally

536 different from (Maxeiner, Luo et al., 2016), both supporting the idea that the ribbon

537 serves to consolidate vesicles near to release sites as suggested by (von Gersdorff,

538 Vardi et al. 1996).

539 Ribbon and Multiquantal release: The underlying mechanism for multiquantal release

540 remains an open question but might be driven by simultaneous multiquantal or

541 multivesicular release (Glowatzki and Fuchs 2002). The majority of EPSCs measured in

542 the KO remain large and uniformly fast in onset and decay times and the EPSC

543 amplitude is not reduced suggesting that the ribbon does not play a pivotal role in

544 multivesicular release. Thus, mechanisms that involve sequential fusion of vesicles

545 along the ribbon and then release or rapid sequential vesicle release that piggyback

546 onto a release site from the ribbon are unlikely to be valid as these should be

547 dramatically altered by the loss of the ribbon. One remaining possibility is the

548 synchronous release of vesicles from multiple release sites. Although the ribbon is not

549 required for this mechanism, it may indirectly modulate this mechanism by influencing

550 the number of docked vesicles available for release. The increase in number of smaller

551 slower EPSCs likely arising from fewer docked vesicles available for synchronous

552 release may support a ribbon role; however, the small relative contribution of this

553 vesicle pool indicates a minor role. The phenomenon of large EPSC amplitude

554 variability may still be explained by a univesicular mechanism including a dynamic

555 fusion pore, which presumably would not require a ribbon (Chapochnikov, Takago et al.

556 2014). Where identified, the slow footlike response described by (Chapochnikov,

557 Takago et al. 2014) was not different between genotypes.

558 Comparison to Bassoon KO: In mice, the effects of KO are subtle compared with

559 disruption of Bassoon protein. Loss-of-function mutation in Bassoon led to a reduction in

560 the number of IHC ribbons associated with presynaptic elements, ABR threshold

561 elevation of 30 dB, longer ABR wave 1 latency and first-spike latency, and fewer Ca2+

562 channels at active zones (Khimich, Nouvian et al. 2005, Buran, Strenzke et al. 2010,

563 Frank, Rutherford et al. 2010, Jing, Rutherford et al. 2013). Thus, the effects of loss of

564 Bassoon protein (and membrane-anchored ribbons) were greater than loss of RIBEYE

565 protein, consistent with a more critical role for Bassoon in scaffolding synaptic proteins.

566 Also given Bassoon’s presence at efferent synaptic endings in the organ of Corti, it is

567 not surprising that the functional loss of Bassoon has a greater phenotype than that of

568 RIBEYE (Supplemental Fig. 1).

569 Summary: A central tenet for ribbon function is maintaining a large pool of vesicles

570 near the synapse. The ribbon may serve as a conveyor belt to both maintain and to

571 traffic vesicles to release sites (Parsons 1994, Nouvian, Beutner et al. 2006). Our data

572 suggests the ribbon increases the time a vesicle remains in the synaptic region. Robust

573 vesicle trafficking and replenishment either via recycling or transporting new vesicles to

574 the synaptic zone provides vesicles to the synaptic zone, but the time vesicles remain

575 near to release sites may be shortened in the absence of the ribbon serving to maintain

576 them close to the synapses. A model for this basic idea was proposed (Graydon, Zhang

577 et al. 2014) that reproduced data from bipolar cell synapses. The addition of a more

578 robust trafficking mechanism may allow this model to also reproduce hair cell synapses.

579 The increase in variance, whether measured as CV, skewness or simply standard

580 deviation in measured parameters such as time to release, release rate and EPSC

581 amplitude was greater for KO. The increased variance may simply reflect a nonuniform

582 population of vesicles near the synapse in the absence of the ribbon. Thus, the ribbon

583 may serve as a biological buffer, maintaining a uniform population of vesicles near

584 release sites.

585 Materials and Methods

586 Animals

587 Mice of both sexes from a mixed genetic background (C57bl6, 129/Ola) were bred from

588 heterozygous breeding pairs and genotyped by PCR for the knockout of the A-domain

589 in Ctbp2 (Maxeiner et al., 2016) while wild-type littermates served as controls. The

590 Administrative Panel on Laboratory Animal Care at Stanford University (APLAC

591 #14345) and Animal Care and use Committee and Washington University in St. Louis

592 approved all animal procedures. Experiments were performed blinded to genotype.

593 Auditory measurements: We measured auditory brainstem responses (ABRs),

594 distortion products of optoacoustic emissions (DPOAEs) in mice of postnatal day 21 as

595 previously described (Oghalai, 2004). Mice were anesthetized by administering

596 100 mg/kg ketamine and 10 mg/kg xylazine intraperitoneally and their body temperature

597 held constantly at 37°C (FHC, DC-temperature controller and heating pad) until they

598 fully recovered.

599 Sound stimuli were generated digitally by a self-written software (MATLAB 2009b,

600 MathWorks), controlling a custom-built acoustic system using a digital-to-analog

601 converter (National Instruments, NI BNC-2090A), a sound amplifier (TDT, SA1), and

602 two high frequency speakers (TDT, MF1). The speakers were connected to an ear bar

603 and calibrated in the ear canal prior to each experiment using a probe-tip microphone

604 (microphone type 4182, NEXUS conditioning amplifier, Brüel and Kjær).

605 Auditory Brainstem Responses (ABRs) were recorded by placing three needle

606 electrodes subcutaneously at the vertex, below the left ear, and a ground electrode

607 close to the tail. The signals were amplified 10,000 times using a biological amplifier

608 (Warner Instruments, DP-311) digitized at 10 kHz, and digitally band-pass filtered from

609 300 to 3,000 Hz. The stimulus for eliciting the ABR was a 5 ms sine wave tone pip with

610 cos2 envelope rise and fall times of 0.5 ms. The repetition time was 50 ms, and 260

611 trials were averaged.

612 At each frequency, the peak to peak voltages of ABR signals at stimulus intensities

613 ranging from 10 to 80 dB sound pressure levels (SPL) were measured in 10 dB steps

614 and fitted and interpolated to find threshold five standard deviations above the noise

615 floor. For statistical purposes we defined threshold as 80 dB SPL, if no ABR was

616 detected.

617 Distortion Products of Otoacoustic Emission (DPOAE): For stimulation of DPOAEs

618 two sine wave tones of equal intensity (l1=l2) and 1-second duration were presented to

619 the ear. The tones ranged from 20 to 80 dB SPL attenuated in 10 dB increments and

620 the two frequencies (f2 = 1.2f1) ranged from f2 with 4 to 46 kHz. The acoustic signal was

621 detected by the microphone in the ear bar was sampled at 200 kHz and the magnitude

622 of the cubic distortion product (2f1-f2) determined by FFT. Averaging 20 adjacent

623 frequency bins surrounding the distortion product frequency determined the noise floor.

624 DPOAE thresholds were reached when the signal was greater than three standard

625 deviations above the noise floor.

626 Tissue Preparation, Staining and Confocal Microscopy: For staining of pre- and

627 postsynaptic markers we used antibodies against Homer (Synaptic Systems, 160003,

628 1:800 dilution), Bassoon (AbCam, ab82958, 1:500 dilution), CtBP1 (BD-Biosciences,

629 612042, 1:600 dilution), CtBP2/RIBEYE (Santa Cruz Biotechnology, sc-5966, 1:300

630 dilution; BD Transduction Lab, 612044, 1:400), CaV1.3 (Alomone Labs, ACC-005, 1:75),

631 GluA3 (Santa Cruz Biotechnology, sc-7612, 1:200 dilution). Cochlea of P22 or P35 mice

632 were isolated in HBSS solution or artificial perilymph, a small hole was introduced at the

633 apex, cochlea were gently perfused through round and oval windows with freshly

634 prepared 4 % (w/v) PFA (Electron Microscopy Sciences) in PBS until the fixative

635 became visible at the apex. Organs were incubated in the same fixative solution for 30

636 min at room temperature. Whole mount organs of Corti were micro dissected, the

637 tectorial membrane removed, the tissue permeabilized for 30 min in 0.5 % (v/v)

638 Triton X-100/PBS, blocked for 2 h in 4 % (w/v) BSA/PBS at room temperature. Staining

639 with all primary antibodies was performed overnight at 4°C in the same blocking buffer.

640 For staining with antibodies against GluA2 and CaV1.3 the protocol was slightly varied:

641 fixation was 20 min at 4°C, followed by decalcification in 10 % EDTA (Sigma) for 2 h,

642 permeabilization and blocking was carried out in a blocking buffer containing 16%

643 donkey serum and 0.3 % Triton X-100. After washing all species appropriate secondary

644 antibodies (Alexa Fluor 488, 546 or 555, 647: 1:600 dilution in 4 % (w/v) BSA/PBS)

645 were applied for 1 h at room temperature. After washing, the organs of Corti were

646 mounted hair bundles facing upward in ProLong Gold© Antifade (Life Technologies).

647 Samples were imaged using confocal microscopy (LSM700 or LSM880, Zeiss) on 1.4

648 NA oil immersion objectives for confocal imaging, and 1.2 NA water immersion objective

649 for Airyscan imaging. The optimal voxel size was adjusted to the green channel

650 (confocal x/y/z: 100/100/350 nm; Airyscan x/y/z: 33/33/222 nm). RIBEYE WT and KO

651 samples were prepared in parallel and images were collected with identical settings.

652 All images were taken in the apical region, between 200-1000 µm from the apex, of the

653 organ of Corti.

654 Image analysis

655 For three-dimensional analysis of synaptic punctum volume, intensity and

656 colocalization, we used the Imaris software package (Version 8.4, Bitplane). Puncta

657 were initially detected using the spot algorithm allowing for different spot sizes and

658 elongated point spread function in z-direction (initial seed size x-y/z: CtBP1, Ctbp2,

659 Homer, 0.44/1.13 µm; Bassoon 0.37/1.13 µm; GluA3 0.7/1.4 µm; CaV1.3 0.56/1.4 µm),

660 followed by co-localizing the spots with a threshold lower than 1. All counts were

661 normalized to the number of Inner hair cells in a region of interest.

662 To estimate the size of synaptic puncta we used the surface algorithm in Imaris. For all

663 stainings, the surface detail was set to 0.15 µm, background subtraction to 0.562 µm,

664 and touching object size was 0.750 µm (Bassoon puncta were split at 0.36 µm).

665 TEM and Tomography

666 Temporal bones of P21 mice (WT: n = 5, KO: n = 5) were isolated in HBSS, a small

667 hole was introduced at the apex of the cochlea, gently perfused through round and oval

668 windows with freshly prepared 4 % (w/v) PFA, 2 % (w/v) Glutaraldehyde (Electron

669 Microscopy Sciences) in a buffer containing (in mM) 135 NaCl, 2.4 KCl, 2.4 CaCl2, 4.8

670 MgCl2, 30 HEPES, pH 7.4 until the fixative became visible at the apex. Temporal bones

671 were soaked for 4 hours in the same buffer and shipped in PBS. Cochleae were

672 microdissected out from fixed adult mouse temporal bones and slowly equilibrated with

673 30% glycerol as cryoprotection. Tissues were then plunge frozen in liquid ethane

674 at -180C using a Leica Biosystems grid plunger. Frozen samples were freeze

675 substituted using Leica Biosystems AFS in 1.5% uranyl acetate in absolute methanol

676 at -90C for 2 days and infiltrated with HM20 Lowicryl resin (Electron Microscopy

677 Sciences) over 2 days at -45C. Polymerization of the resin was carried out under UV

678 light for 3 days at temperatures between -45C and 0C. Ultrathin sections at 100 nm

679 were cut using a Leica ultramicrotome and collected onto hexagonal 300-mesh Cu grids

680 (Electron Microscopy Sciences).

681 Samples were imaged using a 200kV JEOL 2100 with a Gatan Orius 832 CCD camera.

682 Single images were acquired using DigitalMicrograph (Gatan). Tilt series were acquired

683 using SerialEM (Mastronarde 2005) at 1 tilt increment from -60 to +60 and

684 tomographic reconstruction of tilt series and segmentation of tomograms was done with

685 IMOD software suite (Kremer, Mastronarde et al. 1996). Fiji was used to generate

686 projection images from tomograms and for cropping, rotating and adjusting the

687 brightness and contrast of images.

688 Hair cell recordings:

689 IHCs from the apical coil of the organ of Corti were dissected from p10-13 Ctbp2

690 littermates of either sex, placed on a plastic coverslip and secured with two strands of

691 dental floss. Cells were visualized with a water immersion lens (100x) attached to a

692 BX51 Olympus upright microscope. Patch-clamp recordings were obtained at room

693 temperature (21-23 C) in an external solution containing in (mM) NaCl 140, KCl, 5.4,

694 CaCl2, 2.8, MgCl2 1, 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid (HEPES) 10, D-

695 glucose 6, creatine, ascorbate, Na pyruvate 2, pH 7.4 adjusted with NaOH. 100 nM

696 apamin was included to block the sK K+ channel. Cesium at 3 mM was used externally

697 to reduce inward leak current. Pipette solution contained (mM) CsCl 90, TEA,

698 tetraethylammonium chloride 13, 3 NaATP, MgCl2 4, creatine phosphate 5, ethylene

699 glycol-bis(β-aminoethyl ether)-N,N,N',N'-tetraacetic acid (EGTA), HEPES 10, pH

700 adjusted with CsOH to 7.2.

701 Two-Sine Method

702 Hair cells were voltage-clamped with an Axon Multiclamp 700B (Axon Instruments-

703 Molecular Devices, Sunnyvale, CA.) and data recorded at 200-300 kHz with a National

704 Instrument USB-6356 D/A acquisition board and JClamp software (SciSoft). Vesicle

705 fusion was determined by monitoring membrane capacitance as a correlate of surface

706 area change. Capacitance was measured with a dual sinusoidal, FFT-based method

707 (Santos-Sacchi 2004). In this RC analysis method, two voltage frequencies, f1 and f2

708 (twice the f1 frequency) are summed and the real and imaginary components of the

709 current response are used to determine the magnitudes of the three model components.

710 The time resolution of the Cm measurement is the period of f1, which we varied from

711 0.16 to 0.32 ms, corresponding to 6250 to 3125 Hz. Stimulation protocols were

712 performed after 10 minutes in the whole-cell mode to allow Ca2+ current run up and full

713 exchange of the internal solution into the cell (Schnee, Castellano-Munoz et al. 2011). A

714 dual sinusoid of 40 mV amplitude was delivered superimposed to the desired voltage

715 step. The mean IHC Cm was 10 ± 1 pF N = 25 and uncompensated series resistance of

716 10 ± 3 M. All voltages were corrected for a 4 mV junction potential.

717 Postsynaptic Recordings

718 Whole-cell voltage-clamp recordings were made at terminal endings of SGNs from P17-

719 21 animals (Grant, Yi et al. 2010). Cochlea were acutely dissected from the temporal

720 bone in chilled (~ 4°C) standard extracellular solution containing 1 mM curare. The

721 apical turn was excised after removing stria vascularis and tectorial membrane, and

722 placed under two dental floss fibers on a recording chamber filled with the standard

723 extracellular solution at room temperature. Recording pipettes were made from 1.5 mm,

724 thick-wall borosilicate glass (WPI). Pipettes were pulled using a multistep horizontal

725 puller (Sutter) and fire polished while applying a pressure to the back end of pipettes to

726 an impedance of 6-8 MΩ (Johnson, Forge et al. 2008). When whole-cell recording was

727 achieved from a bouton, 5 µM TTX was perfused into the chamber, and EPSCs were

728 recorded in the standard solution (5.8 mM K+) for at least 3 min, then in 40 mM KCl for

729 1-5 min (depending on the duration of synaptic activity on each nerve). Standard

730 external solution contained (in mM): 5.8 KCl, 140 NaCl, 0.9 MgCl2, 1.3 CaCl2, 0.7

731 NaH2PO4, 5.6 D-glucose, 10 HEPES, 2 Ascorbate, 2 Na-Pyruvate, 2 Creatine

732 Monohydrate, Osmolarity was 305 mosmls, pH = 7.4 (NaOH). Intracellular solutions

733 contained (in mM): 135 KCl, 3.5 MgCl2, 0.1 CaCl2, 5.0 EGTA, 5.0 HEPES, 2.5 Na2-ATP,

734 pH = 7.2 (KOH), osm ~290 mosmls.

735 Data Analysis:

736 Data were collected blinded to genotype for the experimentalist. Genotype was revealed

737 during analysis phase to ensure appropriate n values were obtained between groups.

738 Although gender was noted, no gender differences were observed between groups and

739 so data is presented combined. Data analysis was largely performed using unpaired

740 two-tailed t-tests unless otherwise stated. When non normal distributions were

741 observed, as in the EPSC histograms we selected different comparators as described

742 with the data. Similarly when large n values would bias comparisons, such as with

743 analyzing immunohistochemistry data, we moved to alternative more appropriate tests

744 as described in the text. Data is presented as mean ± standard deviation unless

745 otherwise stated. Sample size is presented in legends or in text. For physiology

746 experiments sample size reflects cell numbers which also indicates animal numbers as

747 we did not record from more than one cell in a preparation. For immunohistochemistry,

748 data is presented for animal numbers as well as image, cell and puncta numbers.

749 The voltage against capacitance plots were arbitrarily fit with an exponential equation of

(-x(x-x )/t) 750 the form: Y = Y0 + A1 exp 0 where: Y0 is the baseline change in capacitance and is

751 near 0, A1 is the asymptote for the maximal change in capacitance, x0 is the voltage

752 where changes in capacitance measurements were first noted and t reflects the voltage-

753 dependence of the release (Cm/Vm). The equation for the Gaussian fit used in Fig.

(-2*((x-x )/w)^2) 754 6E is: Y = Y0 + (A/(w*√(π/2)))*exp c where Y0 is baseline, typically 0, A is the

755 area under the curve, w is the width and xc is the peak of the Gaussian.

756 Acknowledgments

757 This work was supported by National Institutes of Health, National Institute on Deafness

758 and Other Communication Disorders (NIDCD), Grants R01DC014712 to M.A.R.; M.A.R.

759 was also supported by an International Project Grant from Action on Hearing Loss.

760 NIDCD; F32 4F32 DC013721 grant from NIDCD to MN, Intramural Research Program

761 (Z01-DC000002) and NIDCD Advanced Imaging Core ZIC DC000081) to W.S. and

762 B.K.: RO1DC009913 to AJR, core grant P30 44992. We thank Autefeh Sajjadi for her

763 help with genotyping animals and basic husbandry. We thank Bill Roberts for comments

764 on the manuscript.

765 The authors declare no competing financial interests

766 References

767 Buran, B. N., N. Strenzke, A. Neef, E. D. Gundelfinger, T. Moser and M. C. Liberman

768 (2010). "Onset coding is degraded in auditory nerve fibers from mutant mice lacking

769 synaptic ribbons." J Neurosci 30(22): 7587-7597.

771 Chapochnikov, N. M., H. Takago, C. H. Huang, T. Pangrsic, D. Khimich, J. Neef, E.

772 Auge, F. Gottfert, S. W. Hell, C. Wichmann, F. Wolf and T. Moser (2014). "Uniquantal

773 release through a dynamic fusion pore is a candidate mechanism of hair cell

774 exocytosis." Neuron 83(6): 1389-1403.

776 dick, o. (2001). "localization of the presynaptic cytomatrix protein piccolo at ribbon and

777 conventional synapses in the rat retina: comparison with basson." J Comp Neurol 439:

778 10.

780 Farris, H. E., G. B. Wells and A. J. Ricci (2006). "Steady-state adaptation of

781 mechanotransduction modulates the resting potential of auditory hair cells, providing an

782 assay for endolymph [Ca2+]." J Neurosci 26(48): 12526-12536.

783 Fernandez, K. A., P. W. Jeffers, K. Lall, M. C. Liberman and S. G. Kujawa (2015).

784 "Aging after noise exposure: acceleration of cochlear synaptopathy in "recovered" ears."

785 J Neurosci 35(19): 7509-7520.

787 Frank, T., M. A. Rutherford, N. Strenzke, A. Neef, T. Pangrsic, D. Khimich, A. Fejtova,

788 E. D. Gundelfinger, M. C. Liberman, B. Harke, K. E. Bryan, A. Lee, A. Egner, D. Riedel

789 and T. Moser (2010). "Bassoon and the synaptic ribbon organize Ca(2)+ channels and

790 vesicles to add release sites and promote refilling." Neuron 68(4): 724-738.

792 Fuchs, P. A. (2005). "Time and intensity coding at the hair cell's ribbon synapse." J

793 Physiol 566(Pt 1): 7-12.

795 Furukawa, T. and S. Matsuura (1978). "Adaptive rundown of excitatory post-synaptic

796 potentials at synapses between hair cells and eight nerve fibres in the goldfish." J

797 Physiol 276: 193-209.

799 Glowatzki, E. and P. A. Fuchs (2002). "Transmitter release at the hair cell ribbon

800 synapse." Nat Neurosci 5(2): 147-154.

802 Grant, L., E. Yi and E. Glowatzki (2010). "Two modes of release shape the postsynaptic

803 response at the inner hair cell ribbon synapse." J Neurosci 30(12): 4210-4220.

804 Graydon, C. W., S. Cho, G. L. Li, B. Kachar and H. von Gersdorff (2011). "Sharp

805 Ca(2)(+) nanodomains beneath the ribbon promote highly synchronous multivesicular

806 release at hair cell synapses." J Neurosci 31(46): 16637-16650.

808 Graydon, C. W., J. Zhang, N. W. Oesch, A. A. Sousa, R. D. Leapman and J. S.

809 Diamond (2014). "Passive diffusion as a mechanism underlying ribbon synapse vesicle

810 release and resupply." J Neurosci 34(27): 8948-8962.

812 Jing, Z., M. A. Rutherford, H. Takago, T. Frank, A. Fejtova, D. Khimich, T. Moser and N.

813 Strenzke (2013). "Disruption of the presynaptic cytomatrix protein bassoon degrades

814 ribbon anchorage, multiquantal release, and sound encoding at the hair cell afferent

815 synapse." J Neurosci 33(10): 4456-4467.

817 Johnson, S. L., A. Forge, M. Knipper, S. Munkner and W. Marcotti (2008). "Tonotopic

818 variation in the calcium dependence of neurotransmitter release and vesicle pool

819 replenishment at mammalian auditory ribbon synapses." J Neurosci 28(30): 7670-7678.

821 Johnson, S. L., H. J. Kennedy, M. C. Holley, R. Fettiplace and W. Marcotti (2012). "The

822 resting transducer current drives spontaneous activity in prehearing mammalian

823 cochlear inner hair cells." J Neurosci 32(31): 10479-10483.

825 Kantardzhieva, A., M. C. Liberman and W. F. Sewell (2013). "Quantitative analysis of

826 ribbons, vesicles, and cisterns at the cat inner hair cell synapse: correlations with

827 spontaneous rate." J Comp Neurol 521(14): 3260-3271.

829 Kantardzhieva, A., M. Peppi, W. S. Lane and W. F. Sewell (2012). "Protein composition

830 of immunoprecipitated synaptic ribbons." J Proteome Res 11(2): 1163-1174.

832 Khimich, D., R. Nouvian, R. Pujol, S. tom Dieck, A. Egner, E. D. Gundelfinger and T.

833 Moser (2005). "Hair cell synaptic ribbons are essential for synchronous auditory

834 signalling." Nature 434(7035): 889-894.

836 Kremer, J. R., D. N. Mastronarde and J. R. McIntosh (1996). "Computer visualization of

837 three-dimensional image data using IMOD." J Struct Biol 116(1): 71-76.

839 Li, G. L., S. Cho and H. von Gersdorff (2014). "Phase-locking precision is enhanced by

840 multiquantal release at an auditory hair cell ribbon synapse." Neuron 83(6): 1404-1417.

842 Li, G. L., E. Keen, D. Andor-Ardo, A. J. Hudspeth and H. von Gersdorff (2009). "The

843 unitary event underlying multiquantal EPSCs at a hair cell's ribbon synapse." J Neurosci

844 29(23): 7558-7568.

846 Lv, C., W. J. Stewart, O. Akanyeti, C. Frederick, J. Zhu, J. Santos-Sacchi, L. Sheets, J.

847 C. Liao and D. Zenisek (2016). "Synaptic Ribbons Require Ribeye for Electron Density,

848 Proper Synaptic Localization, and Recruitment of Calcium Channels." Cell Rep 15(12):

849 2784-2795.

851 Mastronarde, D. N. (2005). "Automated electron microscope tomography using robust

852 prediction of specimen movements." J Struct Biol 152(1): 36-51.

854 Matthews, G. and P. Fuchs (2010). "The diverse roles of ribbon synapses in sensory

855 neurotransmission." Nat Rev Neurosci 11(12): 812-822.

857 Maxeiner, S., F. Luo, A. Tan, F. Schmitz and T. C. Sudhof (2016). "How to make a

858 synaptic ribbon: RIBEYE deletion abolishes ribbons in retinal synapses and disrupts

859 neurotransmitter release." Embo j 35(10): 1098-1114.

861 Nouvian, R., D. Beutner, T. D. Parsons and T. Moser (2006). "Structure and function of

862 the hair cell ribbon synapse." J Membr Biol 209(2-3): 153-165.

864 Pangrsic, T., L. Lasarow, K. Reuter, H. Takago, M. Schwander, D. Riedel, T. Frank, L.

865 M. Tarantino, J. S. Bailey, N. Strenzke, N. Brose, U. Muller, E. Reisinger and T. Moser

866 (2010). "Hearing requires otoferlin-dependent efficient replenishment of synaptic

867 vesicles in hair cells." Nat Neurosci 13(7): 869-876.

869 Parsons, T. D. (1994). "Calcium-triggered exocytosis and endocytosis in an isolated

870 presynaptic cell: capacitance measurements in sacucular hair cells." Neuron 13: 8.

871 Roberts, W. M. (1994). "Localization of calcium signals by a mobile calcium buffer in

872 frog saccular hair cells." J Neurosci 14(5 Pt 2): 3246-3262.

874 Roux, I., S. Safieddine, R. Nouvian, M. Grati, M. C. Simmler, A. Bahloul, I. Perfettini, M.

875 Le Gall, P. Rostaing, G. Hamard, A. Triller, P. Avan, T. Moser and C. Petit (2006).

876 "Otoferlin, defective in a human deafness form, is essential for exocytosis at the

877 auditory ribbon synapse." Cell 127(2): 277-289.

879 Rudolph, S., M. C. Tsai, H. von Gersdorff and J. I. Wadiche (2015). "The ubiquitous

880 nature of multivesicular release." Trends Neurosci 38(7): 428-438.

882 Rutherford, M. A., N. M. Chapochnikov and T. Moser (2012). "Spike encoding of

883 neurotransmitter release timing by spiral ganglion neurons of the cochlea." J Neurosci

884 32(14): 4773-4789.

886 Santos-Sacchi, J. (2004). "Determination of cell capacitance using the exact empirical

887 solution of partial differential Y/partial differential Cm and its phase angle." Biophys J

888 87(1): 714-727.

890 Schmitz, F., A. Konigstorfer and T. C. Sudhof (2000). "RIBEYE, a component of

891 synaptic ribbons: a protein's journey through evolution provides insight into synaptic

892 ribbon function." Neuron 28(3): 857-872.

894 Schnee, M. E., M. Castellano-Munoz, J. H. Kong, J. Santos-Sacchi and A. J. Ricci

895 (2011). "Tracking vesicle fusion from hair cell ribbon synapses using a high frequency,

896 dual sine wave stimulus paradigm." Commun Integr Biol 4(6): 785-787.

898 Schnee, M. E., M. Castellano-Munoz and A. J. Ricci (2013). "Response properties from

899 turtle auditory hair cell afferent fibers suggest spike generation is driven by

900 synchronized release both between and within synapses." J Neurophysiol 110(1): 204-

901 220.

903 Schnee, M. E., J. Santos-Sacchi, M. Castellano-Munoz, J. H. Kong and A. J. Ricci

904 (2011). "Calcium-dependent synaptic vesicle trafficking underlies indefatigable release

905 at the hair cell afferent fiber synapse." Neuron 70(2): 326-338.

907 Snellman, J., B. Mehta, N. Babai, T. M. Bartoletti, W. Akmentin, A. Francis, G.

908 Matthews, W. Thoreson and D. Zenisek (2011). "Acute destruction of the synaptic

909 ribbon reveals a role for the ribbon in vesicle priming." Nat Neurosci 14(9): 1135-1141.

911 Uthaiah, R. C. and A. J. Hudspeth (2010). "Molecular anatomy of the hair cell's ribbon

912 synapse." J Neurosci 30(37): 12387-12399.

914 von Gersdorff, H., E. Vardi, G. Matthews and P. Sterling (1996). "Evidence that vesicles

915 on the synaptic ribbon of retinal bipolar neurons can be rapidly released." Neuron 16(6):

916 1221-1227.

918 Zanazzi, G. and G. Matthews (2009). "The molecular architecture of ribbon presynaptic

919 terminals." Mol Neurobiol 39(2): 130-148.

921 Zenisek, D., N. K. Horst, C. Merrifield, P. Sterling and G. Matthews (2004). "Visualizing

922 synaptic ribbons in the living cell." J Neurosci 24(44): 9752-9759.

924 Zheng, X. Y., D. Henderson, S. L. McFadden, D. L. Ding and R. J. Salvi (1999).

925 "Auditory nerve fiber responses following chronic cochlear de-efferentation." J Comp

926 Neurol 406(1): 72-86.

928 Figure Legends

929 Figure 1. Number of synapses stays unaffected in ribbon less organs of corti.

930 A-D Max intensity projections of confocal z-sections of inner hair cells (IHCs) in apical

931 cochlea turns of P22 KO and wildtype littermates, immunolabeled for CtBP2/RIBEYE

932 (A), Homer (B), Bassoon (C) and Overlay (D). Virtually no punctated extra nuclear

933 CtBP2 staining can be seen in KO mice that is colocalized with other synaptic proteins,

934 while Bassoon and Homer appear generally unaffected. (E) 100 nm thick TEM sections

935 of pre- and postsynaptic densities in WT and KO mice at P28 confirm the absence of

936 electron dense ribbon structures in KO mice. (For E-G) Black arrows point to ribbon

937 structure, white arrows point to synaptic vesicles and white arrowheads point to the

938 remaining density resembling the arciform density. (F) Single slice from a 10 nm thick

939 tomographic section. (G) Reconstruction of complete synapses from data obtained used

940 to create E,F. Presynaptic density in cyan and postsynaptic density in green.Tethered

941 synaptic vesicles modeled in yellow, electron dense structure attributed to synaptic

942 ribbons in magenta. (H) Confocal max intensity projection of IHCs immunolabeled for

943 CtBP1 (green) and Homer (magenta). (I) Quantification of RIBEYE, Homer and

944 Bassoon puncta in the organ of corti normalized to the number of IHCs in analyzed

945 regions of interest, showing complete loss of synaptic RIBEYE staining in KO. The

946 amount of Homer and Bassoon staining stays unaffected (p > 0.05). WT in blue, n = 20

947 mice, quantified puncta: RIBEYE = 3796, Bassoon = 10215, Homer = 4340; KO in red,

948 n = 19 mice, n puncta: RIBEYE = 33, Bassoon = 10825, Homer = 4201. (J) Paired

949 synaptic puncta per IHC is lost in absence of RIBEYE staining (p ≤ 0.001) between WT

950 and KO. The number of synapses defined by colocated postsynaptic Homer and

951 presynaptic Bassoon puncta (p = 0.18) is unaffected in KO. (K) The number of extra

952 nuclear CtBP1 puncta is greatly reduced (p = 0.003) in KO mice. (WT in blue, n = 9

953 mice, n puncta: CtBP1 = 19929, Homer 4940; KO in red, n = 7 mice, n puncta: CtBP1 =

954 11960, Homer = 5033). (L) Remaining CtBP1 puncta show significantly reduced

955 (p ≤ 0.001) pairing with postsynaptic Homer staining. Scale Bars: A-D, H: 5 µm; E, F:

956 100 nm. Boxes represent standard deviations of the mean. Significance levels of two

957 tailed unpaired t-tests: n.s., not significant; * p ≤ 0.05; ** p ≤ 0.01; *** p ≤ 0.001.

959 Figure 1 Supplemental Figure 1: Immunohisotchemistry from WT (A) and KO (B) for

960 Bassoon (green) CTBP2 (red) and myosin VIIa (Blue). Low power images show

961 Bassoon is present in efferents of outer hair cells as well as presynaptically in IHCs.

962 Bassoon localization is much broader than RIBEYE’s and is minimally effected in the

963 mutant animal.

965 Figure 2 Hearing in RIBEYE knockout mice. A Distortion products of otoacoustic

966 emissions (DPOAE) show no significant differences in the tested frequency range (4-

967 46 kHz in half octave increments) between KO (red, n = 30) and WT littermates (blue,

968 n = 22), suggesting normal outer hair cell function. (B) Auditory brainstem response

969 (ABR) measurements in the tested frequency range (4-46 kHz in half octave

970 increments) reveal a small (≈ 10 dB SPL), but significant threshold shift in KO (red, n =

971 28) over their WT littermates (blue, n = 22) at P21-P22. (C) Mean ABR waveforms to

972 5 ms tone pips at 23 kHz at 60 dB SPL from KO (red, n = 30) and wild type littermates

973 (blue, n = 24) reveal a decrease in amplitude of the first ABR wave in KO. (D)

974 Quantification of peak to peak amplitude (wave I: p1-n1) differences between KO (red)

975 and WT littermates (blue) at 23 kHz were significant over most sound pressure levels

976 (20 dB SPL: p > 0.05; 30 & 60 dB SPL: p ≤ 0.01; 40-60 dB SPL: p ≤ 0.001) (E)

977 Amplitudes of first peak amplitudes normalized to their respective values at 60 dB SPL.

978 Amplitudes scale with their maximum value in WT and KO. (F) First peak latency (p1) is

979 not significantly altered in KO mice. For wave I amplitude and latency (D-F) analysis the

980 n-number varied between sound pressure levels by a clearly detectable first wave and

981 speaker calibration, n of WT/KO: 20 dB, 11/7; 30 dB, 21/11; 40 dB, 23/22; 50 dB, 23/26;

982 60 dB, 23/28; 70 dB, 5/6. Boxes represent standard deviations of the mean; whiskers

983 and shaded line in (C) represent standard errors of the mean. Significance levels of two

984 tailed unpaired t-tests: n.s., not significant; * p ≤ 0.05; ** p ≤ 0.01; *** p ≤ 0.001.

986 Figure 3. Ribbonless synapses contain clusters of presynaptic voltage-gated Ca2+

987 channels juxtaposed to postsynaptic glutamate receptors. A-D Max intensity

988 projections of airyscan z-sections of IHCs in apical cochlea turns of P35 KO and WT

989 littermates, immunolabeled for CtBP2/RIBEYE (A), GluA3 (B), CaV1.3 (C) and overlay

990 between calcium channels and glutamate receptors (D). Inner hair cell afferent

991 synapses from WT have a synaptic ribbon: a cytoplasmic presynaptic density comprised

992 mainly of RIBEYE protein, recognized by anti-CtBP2 (green). Co-labelled are the

2+ 993 closely associated presynaptic voltage-gated Ca channels (CaV1.3, gray) and

994 postsynaptic AMPA receptors (GluA3, magenta). Inner hair cell afferent synapses from

995 KO lacked synaptic ribbons. The AMPA receptors in KO mice exhibited a ring-like

996 morphology, and the CaV1.3 channels clustered in the form of round spots or elongated

997 stripes, both similar to WT synapses. Scale bar: 2 µm

998 Figure 4. Changed size distribution but unaffected synapse number and

999 organization in ribbonless synapses (A) Quantification of RIBEYE, GluA3 and

1000 CaV1.3 puncta in the organ of corti normalized to the number of IHCs in analyzed

1001 regions of interest. The number of GluA3 receptor and CaV1.3 puncta is unaffected (p >

1002 0.05). WT in blue, n = 9 images from 3 mice, 95 cells quantified puncta: RIBEYE =

1003 1456, GluA3 = 1459, CaV1.3 = 1897 KO in red, n = 13 images from 3 mice, 248 cells, n

1004 puncta: RIBEYE = 28, GluA3 = 2149, CaV1.3 = 2681. (B) Colocalization of synaptic

1005 puncta per IHC. In absence of RIBEYE, the number of synapses, defined by

1006 colocalization of postsynaptic GluA3 and presynaptic CaV1.3, is unaffected (p = 0.13) in

1007 KO. (C) Volume of presynaptic CaV1.3, Bassoon and postsynaptic Homer and GluA3 in

1008 WT (blue, n synapses: CaV1.3 = 1370; Bassoon = 1917; Homer = 5219; GluA3 =

1009 1611) and KO (red, n synapses: CaV1.3 = 2132; Bassoon = 1817; Homer = 4990;

1010 GluA3 = 2334) mice. (D) Normalized cumulative distributions of volumes show

1011 increased sizes of postsynaptic GluA3 and Homer immunoreactivity. (C) Boxes

1012 represent standard deviations of the mean. Significance levels of two tailed unpaired t-

1013 tests (A/B) or two sample Kolmogorov-Smirnov test (C): D represents the maximum

1014 distance of the Kolmogorov-Smirnov distribution functions. n.s., not significant; * p ≤

1015 0.05; ** p ≤ 0.01; *** p ≤ 0.001; **** p ≤ 0.0001.

1016 Figure 5: No differences were observed in calcium or potassium currents

1017 between KO and WT animals. Ca2+ currents were obtained from mice ranging in age

1018 from P10-P12 for WT (A, blue) and P10-13 KO (B, red), with time course of stimulus

1019 shown above, are comparable in all measured properties. Voltage-current plots,

1020 presented in (C) as mean ± SD show no difference (n provided in figure). Summaries of

1021 peak current (D) and voltage of half activation (V1/2, E) are not statistically different (n=

1022 18 and 26 for WT and KO, p = 0.13 for peak current and 0.326 for V1/2). Potassium

1023 currents, obtained from responses to 50 ms voltage pulses ranging from -124 mV to 16

1024 mV in 10 mV increments, from a holding potential of -84 mV are presented in (F) for WT

1025 and (G) for KO mice at P24. A summary of the half activation voltage as estimated from

1026 a Boltzmann fit to the tail current voltage-current (H) found no statistical difference (n =

1027 6 and 8 for WT and KO, respectively with p = 0.084). Current-clamp measurements

1028 found no difference in resting potential (I) (n = 5 and 8 for WT and KO, respectively, p =

1029 0.69).

1031 Figure 6. Capacitance responses recorded from RIBEYE knockout mice show

1032 little difference between phenotypes. Example records of Cm responses (top),

1033 showing two release components and Ca2+ currents (below) recorded from a WT (blue)

1034 (A) and KO (red) (B) in response to 1s depolarizations from -84 mV to -49 and -44 mV.

1035 Dashed line indicates pre-stimulus baseline Cm. Black solid line indicates the 1st linear

1036 component of release and cyan line the 2nd, superlinear component. (C) Summary plot

1037 of 1st and 2nd component magnitudes. No statistical difference was found between

1038 groups. For the linear component n = 15 for WT and 15 for KO, p = 0.23. For 2nd,

1039 superlinear component n = 17 for both groups and p = 0.38. Filled stars indicate mean

1040 for this and all subsequent figures. (D) Summary plot of 1st and 2nd component rates

1041 obtained by linear fits to data points, as indicated by lines in (A,B). No statistical

1042 difference was found between groups where, n = 16 for WT amd 13 for KO, with p =

1043 0.266. For the second component rate, n = 17 for both, with p = 0.085. (D). (E) presents

1044 a population histogram for the voltage at which the superlinear response is first

1045 observed. Fits to Gaussian distributions show no change in the mean value but an

1046 increase in the width of the plot for the KO. (F) Example of delay in onset of capacitance

1047 change with small depolarizations for a KO cell. Plots of Cm verses voltage responses

1048 measured at 0.1 (G), 0.5 seconds (H) 1 sec (I) or 3 sec (J) show a significant difference

1049 for release at 0.1s and -44 mV (p=0.03), error bars represent SEM. Fitting the

1050 relationships with simple exponential functions reveal differences in the shapes of the

1051 curves but not in maximum release for the 0.1 and 0.5 plots but not the 1 and 3 sec time

1052 points. N provided in figure for each group. (K) Shows time to release is significantly

1053 longer for KO compared to WT for depolarizations eliciting less than 50% of the

1054 maximal calcium current; n = 32 for WT and 26 for KO, p = 0.0012.

1055 Figure 7. Changing internal Ca2+ buffering does not distinguish phenotypes. (A)

1056 Example records of Cm changes in response to 1s depolarizations from -49 mV in 0.1

1057 EGTA and -39 mV in 1 and 5 mM EGTA for WT (blue) and KO (red). (B) Plot of the

1058 change in Cm in response to 1s voltage steps of -49, -44 and -29 mV for 0.1 EGTA and

1059 -49,-44,-39 and -34 mV for 5 EGTA, n provided in figure. (C) The time to the 2nd

1060 component increases significantly with elevated Ca2+ buffer (p value indicated in figure)

1061 but is similar between phenotypes; for 0.1 EGTA n = 7 for WT and 6 KO , p = 0.93,

1062 steps to -29 mV for WT and KO, for 1 EGTA n = 15 for WT and13 for KO , with p = -

1063 0.13, steps to -36 + 8 mV WT and -37 + 9 mV KO and for 5 EGTA n = 3 for WT and 4

1064 for KO, with p = 0.67, steps to -39 + 5 mV WT and -35 + 2 mV KO. (D) The rate of

1065 release to the -49 mV step declines equally between phenotypes with increasing

1066 [EGTA] for 0.1 EGTA, n = 7 for WT and 6 for KO, p = 0.323, for 1 EGTA n = 20 for WT

1067 and 22 for KO, p = 0.99 and for 5 EGTA n = 7 for WT and 6 for KO, p = 0.235. (E,F)

1068 Provide box plots of the first component of release for depolarizations to -49 mV (E) or -

1069 44 mV (F). Effects of buffers were similar between the two groups.

1070 Figure 8: Subtle differences exist between WT and KO mice postsynaptic

1071 properties. Postsynaptic bouton recordings were obtained from WT or KO in control

1072 and during application of 40 mM K+ to the bath. Hair cell recordings demonstrate similar

1073 depolarizations in WT and KO, suggesting that any differences in response properties

1074 may be synapticaly mediated. (A). Cells rapidly recovered following washout of K+. The

1075 solid line below the trace indicates the duration of the application.

1076 Postsynaptic recordings from afferent boutons in response to 40 mM K+ external

1077 solution application are presented in (B) for WT and (C) for KO with a sustained

1078 response and (D) for KO with a transient response. Recordings were made in the

1079 presence of 5 µM tetrodotoxin (TTX). Frequency responses were generated using a 1

1080 second window sliding at 100 ms intervals (E,F). Populations were defined as transient

1081 or sustained by whether firing persisted after 20 seconds. Number of each response

1082 type is presented as fraction in each panel. Maximum frequency responses are

1083 presented in box plot format with mean value shown as start (G); n = 10 for WT and 18

1084 for KO, p = 0.313. The time to peak is presented in (H) where onset of K+ application

1085 response was defined as the time point when the frequency exceeded the spontaneous

1086 frequency +3x SD for two overlapping time windows (200 ms). The time to peak was

1087 statistically different when the one outlier was removed (time point at 181 sec); n = 8 for

1088 WT and 10 for KO. Time to peak only includes fibers whose peak frequency reached 30

1089 Hz because those lower had too few points to make an accurate measure. (I)

1090 Integrating the frequency response over 50 seconds shows that WT release in a more

1091 sustainable manner than the KO, n = 8 for WT and 10 for KO, p = 0.021. Representative

1092 probability density functions for EPSC amplitudes are presented for WT number of

1093 events = 6583 WT and 1551 KO in (J) and KO in (K). Solid lines represent fits to either

1094 double Gaussian functions or Gaussian and gamma functions (smaller peak always fit

1095 with Gaussian function). The curve fitting the small events were integrated to measure

1096 the % of events under the small peak (shown next to each example). Only nerves that

1097 had at least 100 events are included to ensure robustness of the fitting. (L) presents a

1098 box plot of the WT and KO percent of area contributed by the small peak. Comparisons

1099 using a t-test for unequal variance (Welch test) suggest a significant, p = 0.032, n = 7

1100 for WT and 13 for KO. (M) presents a box plot comparing the percentages of simple

1101 events, defined as single rise and decay times plots are not significantly different, n = 7

1102 for WT and 13 for KO, p = 0.3. (N) provides box plots of the median amplitude, where

1103 no significant difference was observed, n = 7 for WT and 13 for KO for each, p = 0.105,

1104 while (O) are box plots of the coefficient of variation for the distributions in (N). Plots

1105 were different, p = 0.017, when equal variance was not assumed.

1106 Figure 9: KO synapses have a larger population of small slow EPSCs. 7 WT

1107 including 15,564 events and 13 KO fibers including 17210 events were analyzed. (A)

1108 shows examples of EPSCs that range in amplitude and kinetics. The inset highlights

1109 one of the small slower EPSCs identified in KO fibers. Median decay times (B) and the

1110 Coefficient of Variation for the distribution in decay times (C) are presented as box plots.

1111 Neither population show statistical significance. (D) plots the skewness of the

1112 distributions of decay times and demonstrates that the KO decay time is significantly

1113 greater (away from 0), suggesting that the mean is larger than median, largely because

1114 they have a long tail of large decay time. (E), plots the EPSC decay time constant for

1115 every fiber (each fiber a different color) for both WT and KO. (F), summarizes the data

1116 from C as average decay time constant with EPSCs binned over 10 pA. (G), pooling

1117 fibers for WT and KO recordings allowed for the generation of a probability density

1118 function for event duration. These distributions are significantly different by KS test

1119 (p<10 -18). KS test (p<10-18). (H) is the cumulative probability function for the duration of

1120 EPSCs. Dashed lines are the individual fiber distributions. Solid red and Blue lines

1121 represent the population aggregate (summed responses) for KO and WT respectively.

1122 The cyan and magenta lines are the averages for WT and KO respectively.