Innervation regulates synaptic ribbons in lateral line mechanosensory hair cells

Ribbons (also known as dense bodies) are electron-dense structures that tether glutamate-containing synaptic vesicles and are found at synapses of sensory cells that respond to a broad range of stimulus intensities, such as mechanosensory hair cells of the auditory and lateral line system (in fish and amphibians), retinal photoreceptor and bipolar cells, and pineal cells (reviewed in Moser et al., 2006; Nouvian et al., 2006; Matthews and Fuchs, 2010; Yu and Goodrich, 2014; Nicolson, 2015). They are thought to store a ready-releasable pool of synaptic vesicles and coordinate synchronous vesicle-release at the synapse (Khimich et al., 2005; Buran et al., 2010; Matthews and Fuchs, 2010; Maxeiner et al., 2016). Voltage gated L-type Ca²⁺ channels are found to drive exocytosis at ribbon synapses and are localized at the base of ribbons (Platzer et al., 2000; Brandt et al., 2003; Dou et al., 2004; Sidi et al., 2004; Sheets et al., 2012; Wong et al., 2014). In zebrafish, these channels are also found to play a role in regulating ribbon size, number and maintenance of ribbons at the hair cell synapse (Sheets et al., 2012). Failure to properly form and localize ribbons at the synapse leads to impaired synaptic transmission and sensory input propagation. Mice defective for bassoon, a cytomatrix protein thought to anchor ribbons at the synapse, show reduced exocytosis in auditory mechanosensory hair cells, reduced reliability of spiking at the auditory stimulus onset, abnormal auditory brain responses and lower amplitudes of b-wave responses during electroretinographic recordings in the eye (Brandstatter et al., 1999; Dick et al., 2003; Khimich et al., 2005; tom Dieck et al., 2005; Buran et al., 2010). Zebrafish larvae in which ribeye b, a gene coding for a major scaffolding protein in synaptic ribbons (Schmitz et al., 2000; Zenisek et al., 2004; Wan et al., 2005; Magupalli et al., 2008), is knocked down show reduction in the number of spikes of afferent neurons of the lateral line mechanosensory hair cells when the hair cells are stimulated (Sheets et al., 2011). Similarly, knockout mice for the ribeye gene (also known as Ctbp2) show impaired neurotransmitter release at retinal bipolar cells (Maxeiner et al., 2016).

Electron microscopy and immunohistochemistry data shows that ribbon size and localization changes during cell development and the establishment of functional synapses. During early development of auditory hair cells and retinal photoreceptor cells, electron-dense ribbon precursors are found apically in the cytoplasm (Blanks et al., 1974; Sobkowicz et al., 1982,, 1986; Regus-Leidig et al., 2009; Sheets et al., 2011; Wong et al., 2014). These cytoplasmic ribbon precursors in retinal photoreceptor cells colocalize with the cytomatrix proteins bassoon and piccolo and are thought to be shuttled together to the active zone (Regus-Leidig et al., 2009). As synaptogenesis proceeds, the total number of ribbons decreases, ribbons become localized at the active zone and fewer ribbon precursors are found in the cytoplasm. Whereas multiple ribbons are found at the site of immature hair cell synapses, mature synapses have single ribbons (Pujol et al., 1979; Sobkowicz et al., 1986; Pujol et al., 1997; Wong et al., 2014).

To date, few studies have looked at the role that innervation has on ribbon placement and maintenance at the active zone. Sobkowicz and colleagues have shown by electron microscopy that when compared to their innervated counterparts, inner hair cells in denervated cultured mouse cochlea have a higher percentage of ribbons misplaced from the synapse, even though they maintain their engagement with the hair cell membrane (Sobkowicz et al., 1986). Similarly, ribbons are found in positions other than the synapse in cochlear hair cells of guinea pigs after the post-synaptic fibers are damaged following intracochlear perfusion with 200 μM AMPA (Puel et al., 1995). Furthermore, the number of hair cell ribbons, as assessed by immunohistochemistry, is decreased in brain-derived neurotrophic factor (BDNF)- or Mafb-knockout postnatal mice, where innervation is reduced or delayed (Zuccotti et al., 2012; Yu et al., 2013), and in adult mice when ouabain is applied in the round window causing de-afferentation of cochlear hair cells (Yuan and Chi, 2014). Although these observations were made either with in vitro preparations, such as the cultured cochlea, or in conditions which might affect both the nerve and the hair cells, such as in BDNF mutants and in treated cochlea, these studies suggest that innervation is required for localizing ribbons at the synapse and regulating their number in mechanosensory hair cells.

To directly investigate how innervation affects ribbon formation in an in vivo model, we used the zebrafish mechanosensory lateral line system. Because of the ability to generate mutants through forward genetic screens in zebrafish and the accessibility of lateral line hair cells, this system has become valuable in discovering the molecular mechanisms that oversee hair cell development and ribbon synapse formation (Nicolson et al., 1998; Nicolson, 2005, 2015).

The mechanosensory lateral line system is an additional sensory system found in fish and amphibians (Dijkgraaf, 1963, 1989). It measures constant input from water vibrations and is used in part to detect prey and predators, and orient fish in water currents (Hoekstra and Janssen, 1985; Blaxter and Fuiman, 1989; Hassan, 1989; Montgomery et al., 1997; Montgomery and Hamilton, 1997; Baker and Montgomery, 1999; Kanter and Coombs, 2003; Suli et al., 2012). In zebrafish, the sensory organs, called neuromasts, develop in stereotypical positions along the body and are located exclusively on the body surface of developing larvae; in adults they are found both on the body surface and within bony canals (Raible and Kruse, 2000; Webb and Shirey, 2003; Ghysen and Dambly-Chaudiere, 2007). Neuromasts are comprised of support cells and mechanosensory hair cells, which are similar in characteristics and function to auditory and vestibular hair cells (Kalmijn, 1989). To determine the role of innervation in hair cell ribbon synapses, we analyzed mutants that lack innervation, removed hair cell innervation by cutting innervating fibers after synaptogenesis and looked at ribbon formation during hair cell regeneration.

The shape and size of ribbons differ across cell types, developmental stages and species (reviewed in Sterling and Matthews, 2005; Moser et al., 2006; Nouvian et al., 2006). In all cell types, ribbon precursors initially appear as spheres; however, as ribbons mature, in some cell types, their shape changes. For example, in mature auditory hair cells ribbons are ellipsoid, whereas in mature photoreceptor cells they appear as sheets (Sterling and Matthews, 2005; Nouvian et al., 2006). Transmission electron microscopy (TEM) shows that in mature zebrafish mechanosensory lateral line hair cells, ribbons at the synapse are spherical with a diameter of about 300 nm (Sidi et al., 2004; Obholzer et al., 2008; Nicolson, 2015) (Fig. 1A). To determine whether ribbon size at the synapse differs with hair cell maturation, we fixed and stained 2 days post fertilization (dpf) embryos to 7 dpf larvae for TEM and imaged ribbons in hair cells. We found that ribbons at the synapse remained spherical during development. Their diameter as calculated by averaging measurements obtained in two planes, one perpendicular to the active zone and the other horizontal to the active zone, was ∼300 nm (Fig. 1B). There is a slight size difference between ribbons at 3 dpf and 4 dpf, which might be indicative of ribbon maturation during hair cell development.

Mechanosensory hair cells of the lateral line are innervated by lateral line ganglion neurons that project afferent fibers to hair cells and octavolateralis efferent neurons that provide cholinergic input to hair cells. To initially address the requirement of innervation in ribbon formation, we analyzed ribbons in lateral line hair cells of neurogenin1 mutants (neurog1) (Golling et al., 2002; McGraw et al., 2008), given that neurog1 knockdown results in a lack of afferent innervation of hair cells (Grant et al., 2005). Antibody staining for acetylated tubulin confirmed that neurog1 mutants lack afferent innervation of the posterior lateral line neuromasts (Fig. 2A,B). However, we further observed that the afferent innervation of the anterior lateral line neuromasts was not affected in these mutants, suggesting that neurog1 is only required in the development of the posterior lateral line ganglion and not the anterior lateral line ganglion. When we stained with the SV2 antibody, which recognizes efferent fibers, we similarly found that neurog1 mutants lack efferent innervation in the posterior lateral line (data not shown) but not the anterior lateral line. Therefore, hair cells in posterior lateral line neuromasts cannot send input to the central nervous system through afferent fibers, and are devoid of efferent input from the octavolateralis efferent neurons. We assessed the role of innervation in ribbon formation by looking at ribbons in these non-innervated posterior lateral line hair cells (Fig. 2D,D′,F,F′) and comparing them to their wild-type siblings (Fig. 2C,C′,E,E′). Confocal microscopy on neurog1 mutants stained with an antibody against RibeyeB (Sheets et al., 2011), a scaffolding protein that is part of ribbons, showed that ribbons are present in neurog1^−/− posterior lateral line hair cells (Fig. 2D′,F′) suggesting that ribbon formation is an intrinsic function of the hair cells rather than a response to innervation. Even though ribbons formed in neurog1^−/− hair cells at 5 dpf, their placement in the cell was different from that in wild-type animals. In mutant cells, devoid of afferent and efferent innervation, ribbons were both adjacent to the basal membrane and distal from it (Fig. 2F′). Although we were not able to quantify this effect, the difference is qualitatively evident in the images in Movie 1 (neurog1^−/−) and Movie 2 (neurog1 sibling). When we measured the distance between individual ribbons within a cell (inter-ribbon distance), we found that it was increased in neurog1^−/− posterior lateral line hair cells (Fig. 2J,K). Additionally, we noticed that ribbon number was increased in neurog1^−/− hair cells and that ribbon size distribution was different from in their siblings, with neurog1^−/− hair cells showing an increase in smaller-size ribbons (Fig. 2G,I). We also calculated the average size of ribbons, which appeared reduced in neurog1^−/− hair cells (Fig. 2H). We recognize that we are at the limit of resolution with light microscopy, and thus ribbon sizes are semi-quantitative and might not reflect the true size of ribbons. By contrast, the anterior lateral line hair cells in neurog1^−/−, which retain innervation, showed no difference in ribbon number (Fig. 2L) and a similar ribbon size distribution (Fig. 2N), although the average ribbon size was smaller when compared to their siblings (Fig. 2M). When assessed by TEM, ribbons in 5 dpf posterior lateral line hair cells of neurog1^−/− mutants were not tethered at the membrane (Fig. 3A′,B). From our observations of neuromasts in four different larvae, we found five cases of such ribbons. In addition, in mutant hair cells there appeared to be many more vesicles than in wild-type cells (Fig. 3C). Although we did not perform true serial sectioning (on formvar-coated grids), we followed the same ribbon across three to five ∼90 nm sections to determine the presence of attachments at the membrane. Representative micrographs are shown in Fig. 3A′,B. Taken together, our data demonstrate that innervation is not important for initial generation of ribbons, but suggest that it is crucial for their correct localization at the plasma membrane and for the regulation of ribbon number and size.

Although, initial ribbon formation appears to be an intrinsic property of the hair cell, the question arises as to whether innervation is necessary for subsequent ribbon maintenance at the membrane. To address this issue, we transected the posterior lateral line nerve anterior to the L1 neuromast on one side of 5 dpf larvae (a stage after synapses are established) by applying three iterations of 5-ms 532-nm laser pulses using a spinning disc confocal microscope (Fig. 4A,B). For these experiments, we used Tg(neuroD:GFP) larvae (Obholzer et al., 2008), whose GFP-labeled lateral line ganglion neurons allow visualization of afferent fibers. Because efferent fibers use afferents to pathfind during development and maintain a close proximity at later stages (data not shown), we were unable to affect afferent or efferent fibers separately; therefore, hair cells in the transected side lacked both afferent and efferent innervation. At 24 h after neuronal fiber transection, we immunostained with the anti-RibeyeB antibody and compared the ribbons in denervated posterior lateral line hair cells (Fig. 4B) to the ribbons in the non-transected control side (Fig. 4A). In the control side, ribbons remained adjacent to the membrane (Fig. 4E′), whereas in the denervated hair cells, ribbons were found both adjacent to the basolateral membrane and distal to it (Fig. 4F′), similar to those in neurog1 mutants. Furthermore, the number of ribbons in denervated hair cells was higher than in controls (Fig. 4G), and there was a higher percentage of smaller-size ribbons (Fig. 4H,I). These data demonstrate that innervation is not only important for regulating ribbon localization, size and number, but that continuous innervation influences ribbon maintenance and stability.

To follow ribbon dynamics as hair cells mature and establish synapses, we looked at ribbons during hair cell regeneration. Because they are localized on the surface of zebrafish larvae, lateral line hair cells are susceptible to mechanically or pharmacologically induced cell death, but can overcome such insults by promptly regenerating (Harris et al., 2003). Following hair cell death, the afferent innervating fibers remained proximal to the neuromasts (Fig. S1, Movie 3, wild-type; Movie 4, after hair cell death). We incubated 5 dpf zebrafish larvae for 1 h with 400 μM neomycin to kill hair cells, fixed the larvae at different time points during regeneration, and processed samples both for immunohistochemistry and TEM. Looking at surface renderings of confocal images and rotating the renderings in 360° (Movies 5 and 6), we noticed that at earlier time points during regeneration [12–36 h post treatment (hpt)], ribbons, as detected by anti-RibeyeB antibody, were both distal from the basal membrane and adjacent to it (Fig. 5B–B″,D), although we do not have the resolution to distinguish between anchored ribbons and floating ribbons in proximity to the basal cell membrane. This ribbon phenotype was much like the one we observed in hair cells that lacked innervation. As hair cells matured, the number of ribbons decreased, particularly the number of the ribbons found distal to the basolateral membrane. By 48 h post treatment (hpt), membrane-adjacent ribbons predominated (Fig. 5C–C″,D; Movies 7 and 8), similar to the untreated controls (Fig. 5A–A"), again confirming that innervation is required for ribbon localization at the basolateral membrane.

TEM showed that early during regeneration, electron-dense structures were present in the cytoplasm (Fig. 6A–B′). We interpret these structures to be ribbon precursors, although immunoelectron microscopy for RibeyeB protein would be necessary to definitively prove their identity. Synaptic vesicles were often found near cytoplasmic ribbon precursors (Fig. 6A′). At different points during regeneration, we additionally found cases of double ribbons (Fig. 6C), ‘floating’ ribbons (i.e. ribbons not near a synapse) (Fig. 6E), ribbons next to what appear to be growth cones (Fig. 6G) and mature ribbons at synapses (Fig. 6D,F,H). Double ribbons and floating ribbons were rarely found at 72 hpt, after regenerated hair cells have matured (data not shown). Our observations of ribbon types at different time points during regeneration are summarized in Table S1.

Given that we show innervation to be important for ribbon regulation, we reasoned that mutants that lack hair cell mechanotransduction activity might also show defects in ribbon number, size and localization. To determine this, we assayed ribbons in sputnik mutants (spu) (Söllner et al., 2004), which lack mechanotransduction due to a mutation in cadherin 23 that is required for tip-link formation between the stereocilia in hair cell bundles. By looking at ribbons in anterior lateral line hair cells (Fig. 7A–D′), we found that spu mutants had one fewer ribbon per hair cell when compared to their siblings (Fig. 7E), but the distribution of ribbon size did not seem different (Fig. 7B′,D′,F,G). Furthermore, ribbons appeared to be adjacent to the plasma membrane and not distal to it. This phenotype is very different from hair cells lacking innervation. We conclude that physical presence of afferent fibers is sufficient to regulate ribbon formation and stabilization.

Establishment of ribbon synapses in mechanosensory hair cells is imperative for their proper function. Our in vivo data of ribbons in lateral line hair cells of neurog1 mutant zebrafish that lack innervation show that ribbon generation is an intrinsic property of hair cells and occurs despite the absence of innervation. By contrast, analysis of ribbons in regenerating hair cells or ribbons in larvae where the innervating fibers have been transected shows that although innervation is not important for the initial ribbon formation, it influences ribbon number, size, localization and maintenance at the plasma membrane. These observations are consistent with prior observations of Puel and colleagues, who noticed ectopic or mislocalized synaptic bodies after the nerve endings were damaged by AMPA application in the cochlea (Puel et al., 1995) and with those of Sobkowicz and colleagues who showed mislocalization of ribbons at the plasma membrane in denervated cochlear cultures when compared to their innervated counterparts (Sobkowicz et al., 1986). How might innervating fibers regulate ribbon localization? Our hypothesis is that adhesion molecules that keep hair cells and their post-synaptic partners together also localize ribbons at the active zone juxtaposed to afferent fibers. This is perhaps accomplished indirectly through binding the cytomatrix protein Bassoon, because ribbons in bassoon mutants are found to be floating (Khimich et al., 2005; tom Dieck et al., 2005). As hair cells mature and synapses are formed, any ribbons that do not get localized at the membrane are likely degraded; therefore, we see a decrease in total ribbon number in mature hair cells during regeneration. Alternatively, the decrease of total ribbon number in mature hair cells might be due to the aggregation of several ribbon precursors, which also leads to bigger ribbons. When hair cells lose their innervation, ribbons lose their connection with the active zone, and perhaps even dissociate into smaller aggregates. These events would lead to an increase in ribbon number and decrease in overall ribbon size, as observed in specimens where the innervating fibers have been transected. Our hypothesis, therefore, is that ribbon localization at the synapse depends on contact of nerve fibers with hair cells and not on hair cell activity. This theory is supported by our data in spu mutants that lack mechanotransduction, which show only a slight decrease in ribbon number and no changes in ribbon size. Similarly, another study shows that ribbon numbers do not change in vglut3 zebrafish mutants, which fail to load glutamate in their synaptic vesicles (Obholzer et al., 2008). Interestingly, the presynaptic ribbon conversely plays a role on its post-synaptic partner, given that knocking down ribeye in zebrafish larvae leads to a decrease of afferent innervation (Sheets et al., 2011).

In conclusion, our work provides strong evidence that innervating fibers are crucial for regulating ribbon number, size and localization. Future studies will hopefully identify molecules that are important in this process.

Zebrafish adults were maintained in our facility under standard conditions. The following transgenic and mutant zebrafish lines were used in our experiments: Tg(brn3c:GFP) (Xiao et al., 2005), sputnik (Söllner et al., 2004), Tg(neurog1:EGFP)^w61 (McGraw et al., 2008), neurog1(neuroD3)^hi1089 mutant (Golling et al., 2002) and Tg(neuroD:EGFP) (Obholzer et al., 2008). To generate neurog1 mutants and siblings for experiments in Fig. 2, double heterozygotes for neurog1(neuroD3)^hi1089; Tg(neurog1:EGFP)^w61 were in-crossed. All mutants and siblings from this cross were selected for expression of the neurog1:EGFP transgene and subsequently processed for antibody staining. All animal experiments were performed according to approved IACUC protocols.

Standard immunohistochemistry techniques were used for antibody labeling (Suli et al., 2014). The following is the list of antibodies used for visualization: hair-cells, mouse anti-parvalbumin IgG1 (1:400, Millipore, MAB1572) with goat anti-mouse-IgG1 Alexa-Fluor-488-conjugated secondary (1:400, Invitrogen, A-21121); ribbons, rabbit anti-RibeyeB IgG (1:250) (Sheets et al., 2011) with goat anti-rabbit-IgG Alexa-Fluor-568-conjugated secondary antibody (1:400, Invitrogen, A-11011); afferent neurons, mouse anti-acetylated tubulin IgG2b (1:500, Sigma, T7451) with goat anti-mouse-IgG2b Alexa-Fluor-488-conjugated or goat anti-mouse-IgG2b Alexa-Fluor-633-conjugated secondary antibody (1:400, Invitrogen, A-21146, A-21141); GFP, rabbit anti-GFP IgG (1:400, Invitrogen, A-11122) with goat anti-rabbit-IgG Alexa-Fluor-488-conjugated or goat anti-rabbit-IgG Alexa-Fluor-633 secondary antibody (1:400, Invitrogen, A-11008, A-21071).

Confocal fluorescent imaging stacks were acquired using an Olympus FV1000 confocal microscope. We used a UPLSAPO 60× water lens (NA:1.2), 6× zoom, 0.4–0.6 µm step size and resolution of 640×640 pixels/frame or 800×800 pixels/frame for image acquisition in Figs 2C–F′, 4C–F′, 5 and 7. For image acquisition in Figs 2A,B and 4A,B, we used a UPLSAPO20X air lens, NA 0.75, 1× zoom, 1.1 µm step size, resolution of 800×640 pixels/frame.

Transmission electron microscopy (TEM) was performed in 2–7 days post fertilization (dpf) embryos and larvae as previously described (Owens et al., 2007). Briefly, zebrafish were euthanized in ice-cold embryo medium, fixed in ice-cold 4% glutaraldehyde in 0.1 M sodium cacodylate with 0.001% CaCl₂ (pH 7.4, 583 mOsm), washed with 0.1 M sodium cacodylate (pH 7.4) with 0.001% CaCl₂, post fixed in 1% osmium tetroxide in 0.1 M sodium cacodylate (pH 7.4) with CaCl₂, dehydrated in an ethanol series, washed in propylene oxide and then embedded in Spurr's epoxy resin in silicone rubber molds (Ted Pella, Redding, catalog no. 10504). After being baked at 60°C, blocks were cut to obtain sections from the anterior or posterior line neuromasts. Two series of ∼90-nm ultrathin sections separated by a 2-μm semi thin section were collected on 200 mesh Athene thin-gar grids (Ted Pella). The tissue was incubated in 5% uranyl acetate in 50% methanol, rinsed with 50% methanol and counterstained with 0.3% lead citrate in 0.1 N NaOH. The 2-μm semi-thin section between the series of ∼90-nm sections allowed observation of different hair cells within a neuromast. TEM was performed using a JEOL 1200EXII and a JEOL JEM 1400 microscopes. Although true serial sectioning was not performed, we followed the same ribbon in three to five sections on the same grid. For Figs 1 and 6, images were taken from the anterior lateral line neuromasts. For Fig. 3, ribbons were taken from the posterior lateral line neuromasts. Ribbon sizes in Fig. 1 were measured using ImageJ (see below).

Tricaine (Sigma, E10521) anesthetized 5 dpf larvae were placed laterally in a Lab-Tek Chamber (Electron Microscopy Sciences, 70378-11) and overlaid with a fine nylon mesh and stainless steel harp for stabilization (Warner Instruments, 64-0253 SHD-26GH/10). Lateral line nerves anterior or at the level of the L1 neuromast were cut by exposure to three iterations of 5-ms 532-nm laser pulses (Ablate! System, Intelligent Imaging Innovations). Larvae were subsequently placed in fresh embryo medium and incubated at 28.5°C for 24 h before being processed for immunohistochemistry.

4 dpf Tg(neuroD:EGFP) transgenic larvae (Obholzer et al., 2008) were embedded sideways in 1% low-melting-point agarose (Sigma, A9414-10G) on glass bottom culture dishes (MatTek, P35G-0-10-C). 350×350 pixel/frame images were acquired using a Zeiss spinning disc microscope, with a 20× lens, 4× zoom, every 5 min for a total of 8 h. Using the same embedding and imaging set up, 5 dpf Tg(neuroD:EGFP) transgenic larvae were imaged every 5 min while being incubated in 400 μM neomycin for 1 h. After the incubation, neomycin was washed out. The larvae were re-embedded in fresh agarose and imaged every 5 min for another 12 h.

In Fig. 1, ImageJ was used to measure the diameter of ribbons in TEM sections from two or three larvae. Two measurements, one of the diameter perpendicular to the active zone and one of the diameter horizontal to the active zone, were collected for each ribbon. The two measurements were then averaged, and the averages for each ribbon were graphed using GraphPad.

SVI Huygens software was used to analyze confocal image stacks. The images were deconvolved, cropped to 480×480 pixels and the advanced object analyzer was used to generate surface renderings and measure the size and number of each ribbon. In an effort to standardize the image processing, fixed threshold settings were applied to all the images, but that failed to sometimes capture all the ribbons. As a result, for each image the best fitting threshold was set to capture all the ribbons. To find whether experimentalist-determined threshold settings gave reproducible results, the same image was processed two different times, which included setting thresholds and measuring the objects. The two different datasets obtained this way were not statistically significantly different from one another; hence, we manually set the thresholds each time. To determine the number of ribbons per cell in neurog1 and spu mutant and sibling larvae (Figs 2 and 7) confocal images of hair cells in three neuromasts of three different larvae were deconvolved and surface renderings were generated using Huygens. A region-of-interest (ROI) sphere was fitted to the base of each hair cell (green channel) in a given neuromast. Following that, the number of ribbons, size and ribbon x-y coordinates within each sphere (hair cell) were collected in the red channel. The inter-ribbon distance was then calculated as the Euclidian distance between coordinates using the Matlab pdist function. The average ribbon size and number was calculated for each neuromast using Microsoft Excel, and the data was subsequently entered in GraphPad software to generate graphs and for statistical analysis.

In Fig. 4, images of hair cells of three neuromasts in two different larvae were acquired for each of the uncut and cut conditions. Using Huygens software, the images were deconvolved, surface renderings were generated and the ribbon size and number per neuromast was collected. In this experiment, we did not use the ROI method to obtain the number of ribbons per hair cell. Instead ribbon number and size was obtained at the neuromast level and the number of ribbons per hair cell was calculated as a ratio of total ribbons to the number of hair cells in the neuromast. The average ribbon size and number was calculated in Microsoft Excel. The data was then graphed and statistical analysis was performed using GraphPad.

In Fig. 5, images of hair cells of three neuromasts in three different larvae were acquired using Olympus FV1000, deconvolved using Huygens and the ribbon number per each hair cell in each neuromast was visually counted from images in 3D surface renderings. Any ribbon that in 3D surface renderings was right next to the cell membrane was designated as basal. The average ribbon number per neuromast was generated in Microsoft Excel and the data was graphed in GraphPad.

We thank Steve MacFarlane in the Electron Microscopy Resource Center at University of Washington for sectioning electron microscopy samples, David White in the Zebrafish Facility at University of Washington for fish care and maintenance, Teresa Nicolson at OHSU (Portland, OR) for providing us with anti-RibeyeB antibody and sputnik mutants, Heather Brignull and Michael Stark for editing the manuscript.