Synaptic mitochondria are critical for hair-cell synapse formation and function

Sensory hair cells in the ear utilize specialized ribbon synapses. These synapses are defined by electron-dense presynaptic structures called ribbons, composed primarily of the structural protein Ribeye. Previous work has shown that voltage-gated influx of Ca2+ through CaV1.3 channels is critical for hair-cell synapse function and can impede ribbon formation. We show that in mature zebrafish hair cells, evoked presynaptic-Ca2+ influx through CaV1.3 channels initiates mitochondrial-Ca2+ (mito-Ca2+) uptake adjacent to ribbons. Block of mito-Ca2+ uptake in mature cells depresses presynaptic Ca2+ influx and impacts synapse integrity. In developing zebrafish hair cells, mito-Ca2+ uptake coincides with spontaneous rises in presynaptic Ca2+ influx. Spontaneous mito-Ca2+ loading lowers cellular NAD+/NADH redox and downregulates ribbon formation. Direct application of NAD+ or NADH increases or decreases ribbon formation respectively, possibly acting through the NAD(H)-binding domain on Ribeye. Our results present a mechanism where presynaptic- and mito-Ca2+ couple to confer proper presynaptic function and formation.


Introduction 38
Neurotransmission is an energy demanding process that relies heavily on mitochondria. 39 In neurons, mitochondria dysfunction has been implicated in synaptopathies that impact 40 neurodevelopment, learning and memory, and can contribute to neurodegeneration (Flippo 41  presynaptic ribbons that triggers vesicle fusion. Tight spatial regulation of presynaptic Ca 2+ is 54 important for ribbon-synapse function and requires efficient Ca 2+ clearance through a 55 combination of Ca 2+ pumps, Ca 2+ buffers and intracellular Ca 2+ stores (Carafoli, 2011;Mulkey 56 and Malenka, 1992; Tucker and Fettiplace, 1995;Yamoah et al., 1998;Zenisek and Matthews, 57 2000). While ER Ca 2+ stores have been implicated in hair-cell neurotransmission, whether 58 mitochondrial-Ca 2+ (mito-Ca 2+ ) stores play a role in this process remains unclear (Castellano-59 Muñoz and Ricci, 2014;Kennedy, 2002;Lioudyno et al., 2004;Tucker and Fettiplace, 1995). 60 In addition to a role in hair-cell neurotransmission, presynaptic Ca 2+ and CaV1.3 channels 61 also play an important role during inner-ear development. In mammals, prior to hearing onset, 62 auditory hair cells fire spontaneous Ca 2+ action potentials (Eckrich et al., 2018;Marcotti et al., 63 2003; Tritsch et al., 2007Tritsch et al., , 2010. In mammalian hair cells, these Ca 2+ action potentials are 64 CaV1.3-dependent and are thought to be important for synapse and circuit formation. In support of this idea, in vivo work in zebrafish hair cells found that increasing or decreasing 66 voltage-gated Ca 2+ influx through CaV1.3 channels during development led to the formation of 67 smaller or larger ribbons respectively (Sheets et al., 2012). Furthermore, in mouse knockouts of 68 CaV1.3, auditory outer hair cells have reduced afferent innervation and synapse number 69 (Ceriani et al., 2019). Mechanistically, how CaV1.3-channel activity regulates ribbon size and 70 innervation, and whether hair-cell Ca 2+ stores play a role in this process is not known. 71 Cumulative work has shown that ribbon size varies between species and sensory 72 epithelia (reviewed in Moser et al., 2006); these variations are thought to reflect important 73 encoding requirements of a given sensory cell (Matthews and Fuchs, 2010). In auditory hair 74 cells, excitotoxic noise damage can also alter ribbon size, and lead to hearing deficits (Jensen et  a splice variant that is unique to vertebrate evolution (Schmitz et al., 2000a). Ribeye contains a 85 unique A-domain, and a B-domain that is nearly identical to full-length CtBP2. The B-domain 86 contains a nicotinamide adenine dinucleotide (NAD + , NADH or NAD(H)) binding site (Schmitz et 87 al., 2000;Magupalli et al., 2008). NAD(H) redox is linked to mitochondrial metabolism 88 (Srivastava, 2016). Because CtBP is able to bind and detect NAD + and NADH levels, it is thought 89 to function as a metabolic biosensor (Stankiewicz et al., 2014 To study the impact of mito-Ca 2+ and NAD(H) redox on ribbon synapses, we examined 104 hair cells in the lateral-line system of larval zebrafish. This system is advantageous for our 105 studies because it contains hair cells with easy access for in vivo pharmacology, mechanical 106 stimulation and imaging cellular morphology and function. Within the lateral-line, hair cells are 107 arranged in clusters called neuromasts. The hair cells and ribbon synapses in each cluster form 108 rapidly between 2 to 3 days post-fertilization (dpf) but by 5-6 dpf, the majority of hair cells are 109 mature, and the system is functional McHenry et al., 2009;Metcalfe, 1985;110 Murakami et al., 2003;Santos et al., 2006). Thus, these two ages (2-3 dpf and 5-6 dpf) can be 111 used to study mito-Ca 2+ and NAD(H) redox in developing and mature hair cells respectively. 112 Using this sensory system, we find that presynaptic Ca 2+ influx drives mito-Ca 2+ uptake. 113 In mature hair cells, mito-Ca 2+ uptake occurs during evoked stimulation and is required to 114 sustain presynaptic function and ultimately synapse integrity. In developing hair cells, mito-Ca 2+ 115 uptake coincides with spontaneous rises in presynaptic Ca 2+ . Blocking these spontaneous 116 changes in Ca 2+ leads to the formation of larger ribbons. Using a redox biosensor, we 117 demonstrate that specifically in developing hair cells, decreasing mito-Ca 2+ levels increases the 118 NAD + /NADH redox ratio. Furthermore, we show that application of NAD + or NADH can increase 119 or decrease ribbon formation respectively. Overall our results suggest that in hair cells 120 presynaptic Ca 2+ influx and mito-Ca 2+ uptake couple to impact ribbon formation and function. 121

Mitochondria are located near presynaptic ribbons 125
In neurons, synaptic mitochondria have been shown to influence synapse formation, 126 plasticity and function (Flippo and Strack, 2017;Todorova and Blokland, 2017). Based on this 127 work, we hypothesized that mitochondria may impact synapses in hair cells. Therefore, we 128 examined the proximity of mitochondria relative to presynaptic ribbons in zebrafish lateral-line 129 hair cells. We visualized mitochondria and ribbons using transmission electron microscopy 130 (TEM) and in live hair cells using Airyscan confocal microscopy. 131 Using TEM, we examined sections that clearly captured ribbons (Example, Figure 1C). 132 We were able to observe a mitochondrion in close proximity (< 1 µm) to ribbons in 74 % of the 133 sections ( Figure 1D, median ribbon-to-mitochondria distance = 174 nm, n = 17 out of 21 134 sections). To obtain a more comprehensive understanding of the 3D morphology and location 135 of mitochondria relative to ribbons in live cells, we used Airyscan confocal microscopy. To 136 visualize these structures in living cells, we used transgenic zebrafish expressing MitoGCaMP3 137 (Esterberg et al., 2014) and Ribeye a-tagRFP  in hair cells to visualize 138 mitochondria and ribbons respectively. Using this approach, we observed tubular networks of 139 mitochondria extending from apex to base ( Figure 1A

Mito-Ca 2+ uptake at ribbons is MCU and CaV1.3 dependent 145
In zebrafish hair cells, robust rises in mito-Ca 2+ have be reported during mechanical 146 stimulation (Pickett et al., 2018). Due to the proximity of the mitochondria to the ribbon, we 147 predicted that rises in mito-Ca 2+ levels during mechanical stimulation are related to presynapse-148 associated rises in Ca 2+ . 149 To test this prediction, we used a fluid-jet to mechanically stimulate hair cells and evoke 150 presynaptic activity. During stimulation, we used MitoGCaMP3 to monitor mito-Ca 2+ in hair cells. As previously reported, we observed robust mito-Ca 2+ uptake during stimulation (Figure 152 1E-F). We examined the subcellular distribution of MitoGCaMP3 signals over time and found 153 that the signals initiated near ribbons ( Figure 1E). During the latter part of the stimulus, and 154 even after the stimulus terminated, the MitoGCaMP3 signals propagated apically within the 155 mitochondria, away from the ribbons (Example, Figure 1E-E'', regions 1-3). We characterized 156 the time course of MitoGCaMP3 signals with regards to onset kinetics and return to baseline. 157 During a 2-s stimulus, we detected a significant rise in MitoGCaMP3 signals 0.6 s after stimulus 158 onset ( Figure S1B). Interestingly, after the stimulus terminated, MitoGCaMP3 levels took 159 approximately 5 min to return to baseline ( Figure S1C To verify that MitoGCaMP3 signals reflect Ca 2+ entry into mitochondria, we applied 168 Ru360, an antagonist of the mito-Ca 2+ uniporter (MCU). The MCU is the main pathway for rapid 169 Ca 2+ entry into the mitochondria (Matlib et al., 1998). We found that stimulus-evoked 170 MitoGCaMP3 signals were blocked in a dose-dependent manner after treatment with Ru360 171 ( Figure 1F). Due to the initiation of mito-Ca 2+ near ribbons, we examined whether presynaptic 172 Ca 2+ influx through CaV1.3 channels was the main source of Ca 2+ entering the mitochondria. To

Mito-Ca 2+ uptake occurs in cells with presynaptic Ca 2+ influx
Interestingly, we observed that mito-Ca 2+ uptake was only present in ~40 % of cells 181 (Example, Figure 2A'; n = 10 neuromasts, 146 cells). This observation is consistent with previous 182 work demonstrating that only ~30 % of hair cells within each neuromast cluster have 183 presynaptic Ca 2+ signals and are synaptically active (Zhang et al., 2018b). Because presynaptic 184 Ca 2+ signals initiate near mitochondria, it is probable that mito-Ca 2+ uptake may occur 185 specifically in hair cells with synaptic activity. 186 To test whether evoked mito-Ca 2+ uptake occurred exclusively in cells with presynaptic 187 Ca 2+ influx, we performed two-color functional imaging. We used a double transgenic approach 188 that utilized a membrane-localized GCaMP6s (GCaMP6sCAAX; green) to measure presynaptic 189 Although we observed mito-Ca 2+ uptake specifically in hair cells with active Ca 2+ 199 channels, the impact of mito-Ca 2+ uptake on the function of hair-cell synapses was unclear. To determine if mito-Ca 2+ uptake impacted presynaptic function, we assayed evoked 204 presynaptic-Ca 2+ signals by monitoring GCaMP6sCAAX signals adjacent to ribbons as described 205 previously (Example, Figure S2 To quantify ribbon-synapse morphology after MCU block, we immunostained mature-224 hair cells (5 dpf) with Ribeye b and MAGUK antibodies to label presynaptic ribbons and 225 postsynaptic densities (MAGUK) respectively. We first applied 2 μM Ru360 for 1 hr, a 226 concentration that partially reduces evoked mito-Ca 2+ uptake (See Figure 1F') yet is effective at 227 reducing sustained presynaptic Ca 2+ influx (See Figure 2D-D'). At this dose, Ru360 had no impact 228 on hair cell or synapse number ( Figure 3E). In addition, we observed no morphological change 229 in ribbon or postsynapse size ( Figure 3F, Figure S3A). These findings indicate that partial MCU 230 block can impair presynaptic function without any observable pathology. 231 We also tested a higher dose of Ru360 (10 µM) that completely blocks evoked mito-Ca 2+ 232 uptake (See Figure 1F). Interestingly, a 30-min or 1-hr 10 µM Ru360 treatment had a 233 progressive impact on synapse and cellular integrity. After a 30-min treatment with 10 µM 234 Ru360 we observed significantly fewer complete synapses per hair cell, but not fewer hair cells  Similar to 30-min treatments with Ru360, after 1 hr, ribbons were also significantly larger 241 ( Figure 3F). Neither 30-min nor 1-hr 10 µM Ru360 treatment altered postsynapse size (Figure 242 S3A). Overall, our results indicate that in mature hair cells, partial block of mito-Ca 2+ uptake can 243 impair presynaptic function without altering presynaptic morphology or synapse integrity. 244 Complete block of mito-Ca 2+ uptake is pathological; it impairs presynaptic function, alters 245 presynaptic morphology, and results in a loss of synapses and hair-cells. 246 247

Spontaneous presynaptic and mito-Ca 2+ influx pair in developing hair cells 248
In addition to evoked presynaptic-and mito-Ca 2+ signals in hair cells, we also observed 249 instances of spontaneous presynaptic-and mito-Ca 2+ signals (Example, Figure 4A . Therefore, we predicted that similar to mammals, spontaneous presynaptic-253 Ca 2+ uptake may be a feature of development. Furthermore, we predicted that spontaneous 254 mito-Ca 2+ uptake may correlate with instances of spontaneous presynaptic-Ca 2+ influx. 255 First we tested whether spontaneous presynaptic-Ca 2+ signals were a feature of 256 development. In zebrafish neuromasts, hair cells are rapidly added between 2-3 dpf, but by 5-6 257 dpf relatively fewer cells are added and the hair cells and the organs are largely mature (Kindt 258  Therefore, we examined the magnitude and frequency of spontaneous, presynaptic 260 GCaMP6sCAAX signals in developing (3 dpf) and mature hair cells (5 dpf). We found that in 261 developing hair cells, spontaneous GCaMP6sCAAX signals occurred with larger magnitudes and 262 more frequency compared to those in mature hair cells ( Figure 4B-C). Our spontaneous 263 GCaMP6sCAAX imaging demonstrates that similar to mammals, spontaneous presynaptic Ca 2+ 264 activity is a feature of developing zebrafish hair cells. 265 Next, we tested whether spontaneous mito-Ca 2+ uptake and presynaptic-Ca 2+ influx 266 were correlated. For this analysis we concurrently imaged GCaMP6sCAAX and MitoRGECO1 267 signals in the same cells for 15 mins to measure presynaptic and mito-Ca 2+ responses 268 respectively. We found that spontaneous presynaptic-Ca 2+ influx was often associated with 269 spontaneous mito-Ca 2+ uptake (Example, Figure 4A

cells. 282
To characterize the role of MCU function and spontaneous mito-Ca 2+ uptake on ribbon 283 formation, we applied the MCU antagonist Ru360 to developing hair cells (3 dpf). After this 284 treatment, we quantified ribbon synapse morphology by immunostaining hair cells to label 285 presynaptic ribbons and postsynaptic densities. After a 1-hr application of 2 μM Ru360 to block 286 the MCU, we observed a significant increase in ribbon size in developing hair cells ( Figure 5A-B, 287 E). In contrast, this same treatment did not impact ribbon size in mature hair cells ( Figure 3F). 288 We also applied a higher concentration of Ru360 (10 µM) to developing hair cells for 1 hr. In 289 developing hair cells, after a 1-hr 10 µM Ru360 treatment, we also observed a significant In addition to larger ribbons, at higher concentrations of Ru360 (10 µM) we also 296 observed an increase in cytoplasmic, non-synaptic Ribeye aggregates ( Figure 5F, G Our results indicate that spontaneous Ca 2+ influx through CaV1.3 channels and 306 subsequent loading of Ca 2+ into mitochondria regulates ribbon formation in developing hair 307 cells. But how do these two Ca 2+ signals converge to regulate ribbon formation? It is possible 308 that mitochondria could buffer Ca 2+ during spontaneous presynaptic activity and function to 309 decrease resting levels of cytosolic Ca 2+ (cyto-Ca 2+ ); cyto-Ca 2+ levels could be a signal that 310 regulates ribbon formation. To examine resting cyto-Ca 2+ levels in hair cells, we examined the 311 fluorescence signal change of the cytosolic Ca 2+ indicator RGECO1 (CytoRGECO1) before and 312 after a 30-min pharmacological manipulation of CaV1.3 or MCU channels ( Figure 6A-C). 313 We observed that treatment with the CaV1.3 channel antagonist isradipine and agonist 314 Bay K8644 decreased and increased resting CytoRGECO1 fluorescence respectively ( Figure 6B). 315 However, treatment with MCU blocker Ru360 did not significantly shift resting CytoRGECO1 316 fluorescence levels ( Figure 6B). Similar results with Ru360 were observed in developing and 317 mature hair cells ( Figure 6B-C). These data suggest that, unlike CaV1.3 channel function, MCU 318 function and associated mito-Ca 2+ uptake does not play a critical role in buffering steady state 319 cyto-Ca 2+ levels. 320 Alternatively, it is possible that rather than impacting cyto-Ca 2+ levels, both CaV1.3 and 321 MCU activity are required to load and maintain Ca 2+ levels within the mitochondria. In this 322 scenario, mito-Ca 2+ levels could be a signal that regulates ribbon formation. To test this 323 possibility, we used MitoGCaMP3 to examine resting mito-Ca 2+ levels before and after 324 modulating CaV1.3 or MCU channel function ( Figure 6D-F). We observed that blocking CaV1.3 325 channels with isradipine or the MCU with Ru360 decreased resting MitoGCaMP3 fluorescence 326 ( Figure 6E-F). Conversely, CaV1.3 channel agonist Bay K8644 increased resting MitoGCaMP3 327 fluorescence ( Figure 6E). These results were consistent in developing and mature hair cells 328 ( Figure 6E-F). Our resting MitoGCaMP3 measurements indicate that the effects of CaV1.3 329 channel and MCU activity converge in to regulate mito-Ca 2+ levels. When either of these 330 channels are blocked, the resting levels of mito-Ca 2+ are decreased. Therefore, if presynaptic 331 Ca 2+ influx and mito-Ca 2+ regulate ribbon formation through a similar mechanism, they may act 332 through mito-rather than cyto-Ca 2+ homeostasis. 333 334

Mito-Ca 2+ levels regulate NAD(H) redox in developing hair cells 335
If mito-Ca 2+ levels signal to regulate ribbon formation, how is this signal transmitted 336 from the mitochondria to the ribbon? An ideal candidate is via NAD(H) homeostasis. Ribeye  To examine NAD(H) redox, we created a stable transgenic line expressing Rex-YFP, a 341 fluorescent NAD + /NADH ratio biosensor in hair cells ( Figure 6G). We verified the function of the 342 Rex-YFP biosensor in our in vivo system by exogenously applying NAD + or NADH for 30 min. We 343 found that incubations with 100 µM NAD + increased while 5 mM NADH decreased Rex-YFP 344 fluorescence; these intensity changes are consistent with an increase and decrease in the 345 NAD + /NADH ratio respectively ( Figure 6H). Next, we examined if CaV1.3 and MCU channel 346 activities impact the NAD + /NADH ratio. We found that 30-min treatments with either CaV1.3 or 347 MCU channel antagonist increased the NAD + /NADH ratio (increased Rex-YFP fluorescence) in 348 developing hair cells ( Figure 6H). Interestingly, similar 30-min treatments did not alter Rex-YFP 349 fluorescence in mature hair cells ( Figure 6I). Together, our baseline MitoGCaMP3 and Rex-YFP 350 measurements indicate that during development, CaV1.3 and MCU channel activities normally 351 function to increase mito-Ca 2+ and decrease the NAD + /NADH ratio. Overall, this work provides 352 strong evidence that links NAD(H) redox and mito-Ca 2+ with ribbon formation.

NAD + and NADH directly influence ribbon formation 355
Our Rex-YFP measurements suggest that spontaneous CaV1.3 and MCU Ca 2+ activities 356 normally function to decrease the NAD + /NADH ratio; furthermore, this activity may function to 357 restrict ribbon formation. Conversely, blocking these activities increases the NAD + /NADH ratio 358 and may increase ribbon formation. If the NAD + /NADH ratio is an intermediate step between 359 CaV1.3 and MCU channel activities and ribbon formation, we predicted that more NAD + or 360 NADH would increase or decrease ribbon formation respectively. To test this prediction, we 361 treated developing hair cells with exogenous NAD + or NADH. 362 After a 1-hr treatment with 100 µM NAD + , we found that the ribbons in developing hair 363 cells were significantly larger compared to controls ( Figure 7A-B, E). In contrast, after a 1-hr 364 treatment with 5 mM NADH, ribbons were significantly smaller compared to controls (Figure 365 7A, C, E). Neither exogenous NAD + nor NADH were able to alter ribbon size in mature hair cells 366 In this study, we determined in a physiological setting how mito-Ca 2+ influences hair-cell 375 presynapse function and formation. In mature hair cells, evoked CaV1.3-channel Ca 2+ influx 376 drives Ca 2+ into mitochondria. Evoked mito-Ca 2+ uptake is important to sustain presynaptic Ca 2+ 377 responses and maintain synapse integrity ( Figure 8B). During development, spontaneous CaV1.3 378 channel Ca 2+ influx also drives Ca 2+ into mitochondria. Elevated mito-Ca 2+ levels rapidly lower 379 the NAD + /NADH ratio and downregulate ribbon formation ( Figure 8A). Furthermore, during 380 development, NAD + and NADH can directly increase and decrease ribbon formation 381 respectively. Our study reveals an intriguing mechanism that couples presynaptic activity with 382 mito-Ca 2+ to regulate the function and formation of a presynaptic structure. 383 384

Functional significance of ribbon size 385
Our work outlines how presynaptic activity controls the formation and ultimately the 386 size of ribbons. When either presynaptic Ca 2+ influx or mito-Ca 2+ uptake was perturbed, ribbons 387 were significantly larger ( Figure 5A Because we also saw an increase in cytoplasmic Ribeye aggregates after MCU block ( Figure 5F-429 G) it is alternatively possible that NAD + and NADH could impact interactions between A and B 430 domains to more broadly impact Ribeye interactions and accumulation. 431 Regardless of the exact mechanism, the effect of presynaptic activity and related 432 changes in NAD(H) redox homeostasis may extend beyond the sensory ribbon synapse. Ribeye 433 is a splice variant of the transcriptional co-repressor CtBP2 (Schmitz et al., 2000b). While the A 434 domain is unique to Ribeye, the B domain is nearly identical to CtBP2 minus the nuclear 435 localization sequence (NLS) (Hübler et al., 2012). In vertebrates, the CtBP family also includes 436 CtBP1 (Chinnadurai, 2007). CtBP proteins are expressed in both hair cells and the nervous 437 system, and there is evidence that both CtBP1 and CtBP2 may act as scaffolds at neuronal 438 synapses (Hübler et al., 2012;tom Dieck et al., 2005). Interestingly, in cultured neurons, it has been shown that synaptic activity is associated with both an increase in CtBP1 localization at 440 the presynapse, as well as a decrease in the NAD + /NADH ratio (Ivanova et al., 2015). In our in 441 vivo study, we also found that the NAD + /NADH ratio was lower in developing hair cells with 442 presynaptic activity ( Figure 6H). But in contrast to in vitro work on CtBP1 in cultured neurons, 443 we found that Ribeye localization to the presynapse and ribbon size was reduced when the 444 NAD + /NADH ratio was lowered ( Figure 7A-C). It is unclear why presynaptic activity regulates 445 Ribeye localization differently from that of CtBP1. Ribeye and CtBP1 behavior may differ due to does mito-Ca 2+ uptake impact presynaptic Ca 2+ activity? Although mito-Ca 2+ uptake could 459 function to buffer cyto-Ca 2+ to maintain presynaptic function, our current work indicates that 460 blocking mito-Ca 2+ uptake does not raise cytosolic Ca 2+ levels ( Figure 6A-C). Therefore mito-Ca 2+ 461 uptake may not be required to buffer or clear Ca 2+ from the cytosol during steady-state. 462 Alternatively, mito-Ca 2+ uptake could buffer Ca 2+ locally during presynaptic activity to prevent 463  In further support of this idea, recent work in mice has investigated the role of the MCU 492 in noise-related hearing loss . This work demonstrated that pharmacological 493 block or a loss of function mutation in MCU protected against synapse loss in auditory inner 494 hair cells after noise exposure. Although this result is counter to our observed results where 495 complete MCU block reduces synapse number (Figure 3E), it highlights an association between 496 mito-Ca 2+ , noise exposure and synapse integrity. It is possible that these differences can be explained by transitory versus chronic alterations in mito-Ca 2+ homeostasis. These differences 498 may be resolved by studying the hair cells in a zebrafish MCU knock out. In the future it will be 499 interesting to examine both mito-Ca 2+ uptake and ribbon morphology during other pathological 500 conditions that enlarge ribbons such as noise exposure, ototoxicity and aging. 501 502

Role of spontaneous mito-Ca 2+ uptake in developing hair cells 503
Although mitochondria have been studied in the context of cellular function and cell 504 death, relatively few studies have examined the role mitochondria play in development. We 505 found that mitochondria spontaneously take up Ca 2+ during hair-cell development ( Figure 4B as a downstream signaling organelle that couples presynaptic-Ca 2+ influx to ribbon formation 510 ( Figure 8A). In the future, zebrafish will be a useful model to further explore the origin and role 511 of these spontaneous Ca 2+ signals. 512 In our study, we also found that altering baseline mito-Ca 2+ levels rapidly influenced the 513 NAD + /NADH ratio and altered ribbon size in developing hair cells (Figure 5, 6, 7). However, in 514 mature hair cells, while alterations to mito-Ca 2+ levels increased ribbon size they did not 515 influence NAD(H) redox ( Figure 6I). One reason why NAD(H) redox does not change in mature 516 hair cells is that ribbon enlargement could be occurring through a different mechanism. For 517 example, ribbon enlargement could be a pathological byproduct of synapse loss ( Figure 3E). In 518 mature hair cells, after MCU block it is possible that individual ribbons are not enlarging, but 519 instead ribbons are merging together as synapses are lost. In the future live imaging studies will 520 help resolve whether there are different mechanisms underlying ribbon enlargement in mature 521 and developing hair cells. 522 Overall this study has demonstrated the zebrafish-lateral line is a valuable system to 523 study the interplay between the mitochondria, and synapse function, development and 524 integrity. In the future it will be exciting to expand this research to explore how evoked and 525 spontaneous mito-Ca 2+ influx are impacted by pathological treatments such as age, noise and 526 ototoxins. For immunohistological studies, zebrafish larvae were exposed to compounds diluted in 555 E3 with 0.1% DMSO (isradipine, Bay K8644, NAD + (Sigma-Aldrich, St. Louis, MO), Ru360 556 (Millipore, Burlington, MA)) or Tris-HCl (NADH (Cayman Chemical, Ann Arbor, MI)) for 30 min or 557 1 hr at the concentrations indicated. E3 with 0.1% DMSO or Tris-HCl were used as control 558 solutions. In solution at pH 7.0-7.3, NADH oxidizes into NAD + by exposure to dissolved oxygen. 559 To mitigate this, NADH was dissolved immediately before use, and was exchanged with a 560 freshly dissolved NADH solution every half hour. Dosages of isradipine, Ru360, NAD + and NADH 561 did not confer excessive hair-cell death or synapse loss unless stated. After exposure to the 562 compounds, larvae were quickly sedated on ice and transferred to fixative. 563 564

In vivo imaging of baseline Ca 2+ and NAD(H) redox 565
To prepare larvae for imaging, larvae were immobilized as previously described (Kindt et  appropriate. Then larvae were exposed to pharmacological agents for 30 minutes and a second 584 acquisition was taken. Any neuromasts with cell death after pharmacological treatment were 585 excluded from our analyses. 586

587
In vivo imaging of evoked Ca 2+ signals 588 To measure evoked Ca 2+ signals in hair-cells, larvae were prepared in a similar manner 589 as described for baseline measurements. After α-bungarotoxin paralysis, larvae were immersed 590 in neuronal buffer solution (in mM: 140 NaCl, 2 KCl, 2 CaCl2, 1 MgCl2 and 10 HEPES, pH 7.3). 591 Evoked Ca 2+ measurements were acquired using the Bruker Swept-field confocal system 592 described above. To stimulate lateral-line hair cells, a fluid-jet was used as previously described 593 to deliver a saturating stimulus (Lukasz and Kindt, 2018). 594 To measure presynaptic GCaMP6sCAAX signals at ribbons, images were acquired with 1 595 x 1 binning with a 35 µm slit at 50 Hz in a single plane containing presynaptic ribbons. Ribbons 596 were marked in live hair cells using the Tg(myo6b:ribeye a-tagRFP) idc11Tg transgenic line (Figure 597 S2). Ribbons were located relative to GCaMP6s signals by acquiring a Z-stack of 5 planes 1 µm. 598 To correlate presynaptic GCaMP6sCAAX signals with mitoRGECO1 signals in hair cells, 2-color 599 imaging was performed. Images were acquired in a single plane with 2 x 2 binning at 10 Hz. 600 MitoGCaMP3 signals were acquired in Z-stacks of 5 planes 1 µm apart at 2 x 2 binning. High 601 speed imaging along the Z-axis was accomplished by using a piezoelectric motor (PICMA P-602 882.11-888.11 series, Physik Instrumente GmbH, Karlsruhe, Germany) attached to the objective 603 to allow rapid imaging at a 50 Hz frame rate yielding a 10 Hz volume rate. For pharmacological 604 treatment, acquisitions were made prior to drug treatment and after a 20-min incubation in the 605 pharmacological agent. Any neuromasts with cell death after pharmacological treatment were 606 excluded from our analyses. 607 608

In vivo imaging of spontaneous Ca 2+ signals 609
To measure spontaneous Ca 2+ signals in hair-cells, larvae were prepared in a similar 610 manner as described for evoked Ca 2+ measurements. Spontaneous Ca 2+ measurements were 611 acquired using the Bruker Swept-field confocal system described above. To measure spontaneous presynaptic GCaMP6sCAAX signals, images were acquired with 2 x 2 binning with 613 a 70 µm slit at 0.33 Hz in a single plane for 900 s. For acquisition of two-color spontaneous 614 presynaptic GCaMP6sCAAX and mitoRGECO1 signals images were acquired with 2 x 2 binning 615 with a 70 µm slit at 0.2 Hz in a single plane for 900 s. 616 617

Electron microscopy 618
Larvae were prepared for electron microscopy as described previously (Sheets, 2017). Zeiss AG, Oberkochen, Germany) using an 63x 1.4 NA oil objective lens. The median (± median 650 absolute deviation) lateral and axial resolution of the system was measured at 198 ± 7.5 nm 651 and 913 ± 50 nm (full-width at half-maximum), respectively. The acquisition parameters were 652 adjusted using the control sample such that pixels for each channel reach at least 1/10 of the 653 dynamic range. The Airyscan Z-stacks were processed with Zeiss Zen Black software v2.1 using 654 3D filter setting of 7.0. Experiments were imaged with the same acquisition settings to maintain 655 consistency between comparisons. 656 657

Analysis of Ca 2+ and NAD(H) signals, processing, and quantification 659
To quantify changes in baseline Ca 2+ and NAD(H) homeostasis, images were processed in 660 FIJI. For our measurements we quantified the fluorescence in the basal-most 8 µm (4 planes) to 661 avoid overlap between cells. The basal planes were max Z-projected, and a 24.0µm (Rex-YFP 662 and RGECO1) or 26.8 µm (MitoGCaMP3) circular region of interest (ROI) was drawn over the 663 neuromast to make an intensity measurement. To correct for photobleaching, a set of mock-664 treated control neuromasts were imaged during every trial. These mock treatments were used 665 to normalize the post-treatment intensity values. 666 To quantify the magnitude of evoked changes in Ca 2+ , fluorescent images were 667 processed in FIJI. Images in each time series were aligned using Stackreg (Thevenaz et al., 668 1998). For evoked MitoRGECO1, MitoGCaMP3, CytoGCaMP3 and two-color GCaMP6sCAAX and drawn over each hair cell. For ribbon-localized measurements, GCaMP6sCAAX signals were 671 measured within a 1.34 µm round ROIs at individual ribbons, and intensity of multiple ROI 672 within a cell were averaged. Cells with presynaptic Ca 2+ activity is defined by max DF/F of > 0.05 673 for MitoRGECO1 and MitoGCaMP3, and max DF/F > 0.25 for GCaMP6sCAAX for a 2-s 674

stimulation. 675
To quantify the average magnitude and frequency of spontaneous Ca 2+ changes in 676 GCaMP6sCAAX signals, images were processed in Matlab R2014b (Mathworks, Natick, MA) and 677 FIJI. First, images in each time series were aligned using Stackreg (Thevenaz et al., 1998). To 678 measure the average magnitude during the 900 s GCaMP6sCAAX image acquisition, a 5 µm 679 diameter circular ROI was drawn over each hair cell and a raw intensity value was obtained 680 ImageJ, each Airyscan Z-stack was background subtracted using rolling-ball subtraction. Z-stacks 695 containing the MAGUK channel were further bandpass filtered to remove details smaller than 6 696 px and larger than 20 px. A duplicate of the Z-stack was normalized for intensity. This duplicated 697 Z-stack was used to identify individual ribbon and MAGUK using the Simple 3D Segmentation of 698 ImageJ 3D Suite (Ollion et al., 2013). Local intensity maxima, identified with 3D Fast Filter, and 3D watershed were used to separate close-by structures. The centroid for each identified