Kif1a and intact microtubules maintain synaptic-vesicle populations at ribbon synapses in zebrafish hair cells

Sensory hair cells of the inner ear utilize specialized ribbon synapses to transmit sensory stimuli to the central nervous system. This transmission necessitates rapid and sustained neurotransmitter release, which depends on a large pool of synaptic vesicles at the hair-cell presynapse. While previous work in neurons has shown that kinesin motor proteins traffic synaptic material along microtubules to the presynapse, the mechanisms of this process in hair cells remain unclear. Our study demonstrates that the kinesin motor protein Kif1a, along with an intact microtubule network, is essential for enriching synaptic vesicles at the presynapse in hair cells. Through genetic and pharmacological approaches, we disrupt Kif1a function and impair microtubule networks in hair cells of the zebrafish lateral-line system. These manipulations led to a significant reduction in synaptic-vesicle populations at the presynapse in hair cells. Using electron microscopy, in vivo calcium imaging, and electrophysiology, we show that a diminished supply of synaptic vesicles adversely affects ribbon-synapse function. Kif1a mutants exhibit dramatic reductions in spontaneous vesicle release and evoked postsynaptic calcium responses. Furthermore, kif1a mutants exhibit impaired rheotaxis, a behavior reliant on the ability of hair cells in the lateral line to respond to sustained flow stimuli. Overall, our results demonstrate that Kif1a-mediated microtubule transport is critical to enrich synaptic vesicles at the active zone, a process that is vital for proper ribbon-synapse function in hair cells.

Methods
Nocodazole treatment

To destabilize microtubules, Tg(myo6b:YFP-Hsa.TUBA)idc16Tg larvae at 5 dpf were incubated in 250 nM nocodazole (Sigma-Aldrich, SML1665) for 2 hours. Nocodazole was diluted in E3 media for a final concentration of 250 nM nocodazole and 0.1 % DMSO. After effective nocodazole treatment YFP-Hsa.TUBA labeling was visually disrupted. For controls, larvae were incubated in media containing 0.1 % DMSO. After 2 hours, larvae were fixed for immunohistochemistry or prepared for LysoTracker labeling (see below).

Lysotracker labeling, and imaging

After 2 hours of nocodazole treatment, 100 nM LysoTracker Red DND-99 (ThermoFisher, L7528) was added to the media for 15 min. Larvae were then embedded in 1 % low melt agarose prepared in E3 media containing 0.03 % tricaine (Sigma-Aldrich, A5040, ethyl 3-aminobenzoate methanesulfonate salt), 100 nM LysoTracker and either 250 nM nocodazole or 0.1 % DMSO. A similar labeling approach was used for Lysotracker labeling in kif1aa mutants. LysoTracker Red DND-99 was used to label Tg(myo6b:YFP-Hsa.TUBA)idc16Tg larvae while LysoTracker Green DND-26 (ThermoFisher, L7526) was used to label Tg(myo6b:ctbp2a-TagRFP)idc11Tg larvae. Transgenic larvae were incubated in 100 nM LysoTracker dye in E3 media for 15 min and mounted in 1 % low melt agarose prepared in E3 media containing 0.03 % tricaine and 100 nM Lysotracker dye.

To image LysoTracker label, samples were imaged live on a Nikon A1R upright confocal microscope using a 60x 1 NA water objective lens. Denoised images were acquired using NIS Elements AR 5.20.02 with an 0.425 µm z-interval. Z-stacks of whole or partial neuromasts were acquired in a top-down configuration using 488 and 561 nm lasers.

Immunohistochemistry

Immunohistochemistry was performed on whole larvae. Zebrafish larvae were fixed with 4 % paraformaldehyde (Thermo Scientific, 28906) in PBS for 3-4 hrs at 4 °C. After fixation samples were washed 5 × 5 min in PBS + 0.01 % Tween (PBST), followed by a 5-min wash in H2O. Larvae were then permeabilized with ice cold acetone (at -20 °C) for 5 min. Larvae were then washed again in H2O for 5 min, followed by a 5 × 5-min washes in PBST, and then blocked overnight at 4 °C with PBST containing 2 % goat serum, 2 % fish skin gelatin and 1 % bovine serum albumin (BSA). Primary antibodies were diluted in PBST containing 1 % BSA. Larvae were incubated in primary antibodies overnight at 4 °C. After 5 × 5 min washes in PBST to remove the primary antibodies, larvae were incubated in diluted secondary antibodies (1:1000) in PBST containing 1 % BSA for 3 hrs at room temperature. After 5 × 5 min washes in PBST to remove the secondary antibodies, larvae were rinsed in H2O and mounted in ProLong Gold Antifade (ThermoFisher, P36930).

RNA-FISH to detect kif1aa and kif1ab mRNA in hair cells

To detect mRNA for kif1aa and kif1ab we followed the Molecular Instrument RNA-FISH Zebrafish protocol, Revision Number 10 (https://files.molecularinstruments.com/MI-Protocol-RNAFISH-Zebrafish-Rev10.pdf), with a few minor changes to the preparation of fixed whole-mount larvae as follows. For our dehydration steps we dehydrated using the following methanol series: 25, 50, 75, 100, 100 % methanol, with 5 min for each step in the series. To permeabilize we treated larvae with 10 µg/mL proteinase K for 20 min. RNA-FISH probes were designed to bind after the motor domain of zebrafish kif1aa and kif1ab (Figure 1F, Molecular Instrument Probe lot # RTD364, RTD365, using B2 and B3 amplifiers respectively). Tg(myo6b:Cr.ChR2-EYFP)ahc1Tg larvae were labeled using these probes; the strong EYFP label is retained after  RNA-FISH and allows for delineation of hair cells within the whole-mount larvae. After the RNA-FISH protocol the larvae were mounted in ProLong Gold Antifade (ThermoFisher, P36930).

Confocal imaging of RNA-FISH and immunolabels

Fixed immunostained and RNA-FISH samples were imaged on an inverted LSM 780 laser-scanning confocal microscope with an Airyscan attachment using Zen Blue 3.4 (Carl Zeiss) and an 63x 1.4 NA Plan Apo oil immersion objective lens. Neuromast and inner ear z-stacks were acquired every 0.15 µm. The Airyscan z-stacks were auto processed in 3D. Experiments were imaged with the same acquisition settings to maintain consistency between comparisons. For presentation in figures, images were further processed using Fiji (RRID:SCR_002285).

Quantification of RNA-FISH and synaptic components

Z-stack image acquisitions from zebrafish confocal images were processed in FIJI. Researchers were blinded to genotype during analyses. Hair-cell numbers were counted manually based on Myo7a, Cr.ChR2-EYFP, YFP-Hsa.TUBA or Acetylated tubulin label. Each channel was background subtracted using the rolling-ball radius method. Then each z-stack was max-intensity projected. A mask was generated by manually outlining the region or interest in the reference channel (ex: hair cells via Myo7a, Acetylated tubulin or Tg(myo6b:Cr.ChR2-EYFP)). This mask was then applied to the z-projection of each synaptic component or RNA-FISH channel.                

We then used automated quantification to quantify puncta using a customized Fiji-based macro. In this macro, each masked image was thresholded using an adaptive thresholding plugin by Qingzong TSENG  (https://sites.google.com/site/qingzongtseng/adaptivethreshold) to generate a binary image of the puncta (presynaptic, postsynaptic, CaV1.3 cluster or RNA-FISH puncta). Individual synaptic or RNA-FISH puncta were then segmented using the particles analysis function in Fiji. A watershed was applied to the particle analysis result to break apart overlapping puncta. After the watershed, the particle analysis was rerun with size and circularity thresholds to generate ROIs and measurements of each punctum. For particle analysis, the minimum size thresholds of 0.025 μm2 (Rib b, CaV1.3, kif1aa and kif1ab RNA-FISH particles) and 0.04 μm2 (Maguk) were applied. A circularity factor of 0.1 was also used for the particle analysis. The new ROIs were applied to the original z-projection to get the average intensity and area of each punctum.

To identify paired synaptic components, images were further processed. Here, the overlap and proximity of ROIs from different channels (ex: pre- and post-synaptic puncta) were calculated. ROIs with positive overlap or ROIs within 2 pixels were counted as paired components. The ROIs and synaptic component measurements (average intensity, area) and pairing results were then saved as Fiji ROIs, jpg images, and csv files.

Synaptic vesicle quantification of live and fixed labels

Z-stack image acquisitions from live or fixed confocal images were processed and quantified in FIJI. A minimum of 3-6 hair cells per neuromast analyzed. Hair cells with a clear side view were used for base to apex analyses. For base to apex, two rectangular ROIs with an area of 2.25 x 0.65 µm (L x W), were placed above and below the nucleus of each hair cell, in 1 z-slice. Nuclei were identified based on Acetylated tubulin or YFP-Hsa.TUBA labeling. The mean values were measured in all ROIs. The ratio of the base to apex was measured per hair cell to assess the enrichment of synaptic vesicles. Ratio values for hair cells were then averaged for a single neuromast base to apex fluorescence value. To quantify Lysotracker fluorescence at ribbons, a circular ROI was drawn around the ribbon indicated by Ctbp2a-tagRFP fluorescence, and the mean value was measured in FIJI. In the cristae, Vglut3 label enrichment at the cell base was measured the same manner as for neuromasts, but Vglut3 enrichment was only quantified in tall cells. To quantify the Vglut3 label in the anterior macula, a maximum projection of the stack imaged was generated in FIJI. An ROI was drawn outlining the macula, and the mean values were measured.

Startle behavior

A Zantiks MWP behavioral system (https://zantiks.com)  was used to assess acoustic startle responses in larvae at 5 dpf. The Zantiks system tracks and monitors behavioral responses using an infrared camera at 30 frames per second. During the tracking and stimulation, a Cisco router connected to the Zantiks system was used to relay x, y coordinates of each larva in every frame. A 12-well plate was used for behavioral analyses. Each well was filled with E3 and 1 larva. All fish were acclimated in the plate within the Zantiks chamber in the dark for 15 min before each test. A vibrational stimulus that triggered a maximal proportion of animals startling in control animals without any tracking artifacts (due to the vibration), was used for our strongest stimuli. We used 4 different levels of intensity (1-4, increasing in intensity), with level 4 as the highest intensity stimulus. To deliver the acoustic vibrational stimulus, the solenoid motor in the Zantiks system was set to move by 7.2° (level 4: 4 full steps), 3.6° (level 3: 2 full steps), 1.8° (level 2: 1 full step), 0.9° (level 1: 1/2 step) with a 4 x 4.25 ms motor speed moving in clockwise and anticlockwise movements. For our initial startle assay, each larva was presented with stimuli from intensity levels 1-3, 5 times, with 100 s between trials to avoid habituation. For each animal, the proportion of startle responses out of the 5 trials was plotted. For our habituation and recovery assay, a non-habituating stimulus, followed by a habituating stimulus train, and lastly recovery stimulus was performed. Our non-habituating stimulus was presented 3 times (intensity level 4), with 100 s between trials. This was followed by a habituating train of 30 stimuli (same stimulus intensity), presented with 5 s in between each stimulus. We then presented recovery stimuli once each at 20 s, 40 s, 1 min, and 2 min after the last stimulus in the habituating train. The proportion of startle responses out of the initial 3 non-habituating stimuli and the proportion of responses at each habituation block and recovery stimulus were plotted. To qualify as a startle response, a distance above 4 pixels or ∼1.9 mm was required within 2 frames after stimulus onset. Larvae were excluded from our analysis if no tracking data was recorded. Startle behavior was performed on at least three independent days.

Rheotaxis behavior

A custom microflume was used as the experimental apparatus for rheotaxis behavior. Laminar water flow of a constant velocity was provided by a 6V bow thruster motor (#108-01, Raboesch) inserted into the flume. An Arduino (UNO R3, Osepp) was programmed with custom scripts to coordinate the timing of the flow and video recording. An array of 196 LEDs emitting infrared light (850 nm) provided illumination for video capture through a layer of diffusion material (several Kimwipes sealed in plastic) and the translucent bottom of the flume. A monochromatic high-speed camera (SC1 without IR filter, Edgertronic.com) with a 60 mm manual focus macro lens (Nikon) was used to record behavioral trials at 60 fps. The flume was filled with EM media (28 °C) and the arena placed within the flume. Due to the heat generated by the IR lights, the temperature was monitored and miniature ice packs (2 × 2 cm; −20 °C) were used to maintain a consistent temperature range of 27–29 °C. For each rheotaxis behavior test, individual 6 dpf larval zebrafish were transferred by pipette to the arena within the flume and their swimming activity was monitored for ~10 s to ensure that it exhibited burst-and-glide swimming behavior; larva that did not exhibit normal swimming behavior during the pre-trial period were not included in the analysis. Each larva was genotyped after behavior acquisition and analysis. A total of 43 wild-type siblings and 30 kif1aa mutant larvae were analyzed. Rheotaxis behavior was assessed blind, prior to genotyping. Data was collected from 6 experimental sessions on separate days.

Larval fish were tracked using DeepLabCut. In brief, video files were downsampled to 1000 x 1000 px and cropped. A previously created and trained single animal maDLC project was used to annotate seven unique body parts (left and right eyes, swim bladder, four points along the tail) on each larva. Videos were analyzed with the maDLC project, and the detections were assembled into tracklets using the box method. The original videos were overlaid with the newly labeled body parts to check for trackelet accuracy. Misaligned tracklets were manually adjusted. Rheotaxis behavior was annotated and analyzed using a previously created custom Python feature extraction script (SimBA) that defined positive rheotaxis events as when the larvae swam into the oncoming water flow at an angle of 0 degrees ± 45 degrees for at least 100 ms.  Videos processed through DeepLabCut analysis were converted to AVI format using the SimBA video editor function and imported into SimBA as previously described.

Calcium imaging and electrophysiology

Larvae for electrophysiology recordings were either in a Tg(myo6b:Cr.ChR2-EYFP) transgenic background or a nontransgenic background. For calcium imaging either Tg(myo6b:memGCaMP6s)idc1Tg or Tg(en.sill,hsp70l:GCaMP6s)idc8Tg  transgenic larvae were used. To prepare larvae for calcium imaging and electrophysiology, 3-6 dpf larvae were anesthetized in 0.03 % Tricaine-S (SYNCAINE/MS-222, Syndel), pinned to a Sylgard-filled perfusion chamber at the head and tail. Then larvae were paralyzed by injection of 125 µM a-bungarotoxin (Tocris, 2133) into the heart cavity, as previously described (Lukasz & Kindt, 2018). Larvae were then rinsed once in E3 embryo media to remove the tricaine. Next, larvae were rinsed three times with extracellular imaging solution (in mM: 140 NaCl, 2 KCl, 2 CaCl2, 1 MgCl2, and 10 HEPES, pH 7.3, OSM 310±10) and allowed to recover prior to calcium imaging or electrophysiology. Researchers were blind to genotype during the acquisition.

Calcium responses in the hair cells and afferent process were acquired on a Swept-field confocal system built on a Nikon FN1 upright microscope (Bruker) with a 60x 1.0 NA CFI Fluor water-immersion objective. The microscope was equipped with a Rolera EM-C2 EMCCD camera (QImaging), controlled using Prairie view 5.4 (Bruker). GCaMP6s was excited using a 488 nm solid state laser. We used a dual band-pass 488/561 nm filter set (59904-ET, Chroma). For evoked measurements, stimulation was achieved by a fluid jet, which consisted of a pressure clamp (HSPC-1, ALA Scientific) and a glass pipette (pulled and broken to achieve an inner diameter of ~50 µm) filled with extracellular imaging solution. A 500-ms pulse of positive or negative pressure was used to deflect the hair bundles of mechanosensitive hair cells along the anterior-posterior axis of the fish.  For GCaMP6s imaging in hair bundles or at the presynapse the Tg(myo6b:memGCaMP6s)idc1Tg line was used. For GCaMP6s imaging in the afferent terminal beneath lateral line hair cell, the Tg(en.sill,hsp70l:GCaMP6s)idc8Tg  line was used. GCaMP6s measurements were performed on larvae at 4 and 5 dpf. Each neuromast (L1-L4) was stimulated four times with an inter-stimulus interval of ~2 min. To acquire GCaMP6s evoked responses, 5 z-slices (0.5 µm step for mechanosensation, 1.5 µm step for presynaptic and 2.0 µm step for the afferent process) were collected per timepoint for 80 timepoints at a frame rate of 20 ms for a total of ~100 ms per z-stack and a total acquisition time of ~8 sec. Stimulation began at timepoint 31; timing of the stimulus was triggered by an outgoing voltage signal from Prairie view.

GCaMP6s z-stacks were average projected, registered, and spatially smoothed with a Gaussian filter (size = 3, sigma = 2) in custom-written MatLab software as described previously (Zhang et al., 2018). The first 10 timepoints (~1 sec) were removed to reduce the effect of initial photobleaching. Registered average projections analyzed in Fiji to make intensity measurements using the Time Series Analyzer V3 plugin. Here circular ROIs were placed on hair bundles or synaptic sites; average intensity measurements over time were measured for each ROI. GCaMP6s data was excluded in the case of excessive motion artifacts. Presynaptic responses were defined as >20% ∆F/F0. Hair-bundle responses were defined as >20% ∆F/F0. Postsynaptic responses were defined as >5% ∆F/F0 and a minimum duration of 500 ms. Calcium imaging data then plotted in Prism 10 (Graphpad). The first 20 timepoints were averaged to generate an F0 value, and all responses were calculated as ∆F/F0. The Area under the curve (AUC) function of Prism was used to determine the peak value for each response. Responses presented in figures represent average responses within a neuromast. The max ∆F/F0 was compared between sibling and kif1aa mutants.

Extracellular postsynaptic current recordings from afferent cell bodies of the posterior lateral-line ganglion (pLLg) of zebrafish at 3-6 dpf were performed. Briefly, borosilicate glass pipettes (Sutter Instruments, BF150-86-10 glass with filament) were pulled with a long taper, with resistances between 5 and 15 MW. The pLLg was visualized using an Olympus BX51WI fixed stage microscope equipped with a LumPlanFl/IR 60x 1.4 NA water dipping objective (N2667800, Olympus). An Axopatch 200B amplifier, a Digidata 1400A data acquisition system, and pClamp 10 software (Molecular Devices, LLC) were used to collect signals. To record spontaneous extracellular currents, afferent cell bodies were recorded using a loose-patch configuration with seal resistances ranging from 20 to 80 MW. Recordings were done in voltage-clamp mode, and signals were sampled at 50 μs/point and filtered at 1 kHz.  The number of spontaneous events from one neuron per min was quantified from a 3-5-min recording window using Igor Pro (Wavemetrics).

Transmission electron microscopy

Larvae were genotyped at 2 dpf using a larval fin clip method to identify kif1aa and wild-type siblings to prepare for TEM. Larval tail DNA was genotyped. At 5 dpf kif1aa and wild-type siblings were fixed in freshly prepared solution containing 1.6 % paraformaldehyde and 2.5 % glutaraldehyde in 0.1 M cacodylate buffer supplemented with 3.4% sucrose and 2 µM CaCl2 for 2 h at room temperature, followed by a 24-h incubation at 4 °C in a fresh portion of the same fixative. After fixation, larvae were washed with 0.1 M cacodylate buffer with supplements, and post-fixed in 1 % osmium tetroxide for 30 min, and then washed with distilled water. Larvae were, dehydrated in 30 – 100 % ethanol series, which included overnight incubation in 70 % ethanol containing 2 % uranyl acetate, and in propylene oxide, and then embedded in Epon. Transverse serial sections (60-70 nm thin sections) were used to section through neuromasts. Sections were placed on single slot grids coated with carbon and formvar, and then sections were stained with uranyl acetate and lead citrate. All reagents and supplies for TEM were from (Electron Microscopy Sciences). Samples were imaged on a JEOL JEM-2100 electron microscope (JEOL Inc.). Whenever possible, serial sections were used to restrict our analysis to central sections of ribbons adjacent to the plasma membrane and a well-defined postsynaptic density.            

To quantify ribbon area, ROIs were drawn in FIJI outlining the electron-dense ribbon, excluding the filamentous “halo” surrounding the ribbon. Vesicles with a diameter of 30–50 nm and adjacent (within 60 nm of the ribbon) to the “halo” were counted as tethered vesicles. Readily releasable vesicles were defined as tethered vesicles between the ribbon and the plasma membrane. To quantify reserve vesicles, we counted vesicles that were not tethered to the ribbon but were within 200 nm of the edge of the ribbon. All distances and perimeters were measured in FIJI.