Synaptically silent sensory hair cells in zebrafish are recruited after damage

Abstract

Analysis of mechanotransduction among ensembles of sensory hair cells in vivo is challenging in many species. To overcome this challenge, we used optical indicators to investigate mechanotransduction among collections of hair cells in intact zebrafish. Our imaging reveals a previously undiscovered disconnect between hair-cell mechanosensation and synaptic transmission. We show that saturating mechanical stimuli able to open mechanically gated channels are unexpectedly insufficient to evoke vesicle fusion in the majority of hair cells. Although synaptically silent, latent hair cells can be rapidly recruited after damage, demonstrating that they are synaptically competent. Therefore synaptically silent hair cells may be an important reserve that acts to maintain sensory function. Our results demonstrate a previously unidentified level of complexity in sculpting sensory transmission from the periphery.

Fig. 1

Hair-cell mechanosensation and presynaptic Ca²⁺ activities. a Cartoon neuromast organ illustrating hair cells (cyan) with apical mechanosensory bundles, basal presynaptic ribbons (magenta), supporting cells (gray), and the associated afferent processes (blue). Bundles made up of stereocilia (cyan) and kinocilia (gray) are oriented to respond to either anterior or posterior stimuli. b, c Images of a double transgenic neuromast expressing GCaMP6s-CAAX (labels apical and basal hair-cell membranes) and Ribeye b-mCherry (labels presynaptic ribbons) in hair cells. A top-down cross-section reveals mechanosensory bundles (b), or the synaptic plane (c). d, f Functional Ca²⁺ imaging acquired from the neuromast in b, c during a 2-s 5 Hz (anterior–posterior directed square wave) mechanical stimulus that stimulates all hair cells. Spatial patterns of GCaMP6s Ca²⁺ signals during stimulation (right panels) are colorized according to the ∆F/F heat maps and superimposed onto a baseline GCaMP6s images (left panels). At the apex, robust mechanosensitive Ca²⁺ signals are detected in nearly all bundles (d), while only a subset of hair cells shows robust presynaptic Ca²⁺ signals at their base (f). e, g Temporal curves of mechanosensitive (1.5 μm ROIs) or presynaptic (3 μm ROIs) Ca²⁺ responses of all 16 hair cells in this example, highlighting four cells with robust presynaptic Ca²⁺ signals (cells 1, 3, 4, 5, labeled in d, f). h Enlarged image of the dashed box in c and f highlights an active (cell 1) and a silent (cell 2) cell. i Using 1.5 μm ROIs (h), in cell 1, the Ca²⁺ signals are larger on ribbons compared to off ribbon (blue traces) while in cell 2, Ca²⁺ signals off and on the ribbon are similar (red traces). j In response to an unsaturated stimulus, there is a slight correlation between presynaptic and mechanosensitive Ca²⁺ signals, R² = 0.26, n = 94 cells. k Using a threshold of ∆F/F = 20%, the active and silent cells in j were separated. Mechanosensitive responses were on average larger in active cells (115.1% ± 11.79, n = 35) versus inactive cells (54.71 ± 4.44, n = 59). A Wilcoxon test was used in k; ***p < 0.0001. Scale bars = 5 μm

Fig. 2

Presynaptic vesicle fusion occurs in a subset of hair cells. a Neuromast hair cells expressing SypHy (vesicle fusion, cyan) and Ribeye b-mCherry (ribbons, magenta). b, c Spatial patterns of SypHy signals during a 2-s 5 Hz stimulation (right panels) are colorized according to the ∆F/F heat map and superimposed onto a baseline SypHy image (b, left panel) or onto an image of ribbons (c, left panel). Only a subset of the hair cells display vesicle fusion. d Dashed circles indicate ROIs (3 µm diameter) used to detect SypHy signals from example in a–c. e Plot of SypHy signals (mean with upper and lower limits) from cells in d from five cells with (white) and eight cells without (orange) vesicle fusion. f The % of hair cells per neuromast with vesicle fusion (naïve blue, 34.04% ± 3.55, n = 5 neuromasts) decreases after 10 µM isradipine treatment (red, 5.20% ± 2.16, p = 0.007), but is not altered (naïve blue, 23.50% ± 3.20, n = 6 neuromasts) after 10 µM Bay K treatment (green, 30.33% ± 3.48, p = 0.10). g The % of hair cells per neuromast with vesicle fusion in response to different stimuli, n = 6 neuromasts per stimulus. h The % of hair cells per neuromast with vesicle fusion does not vary during development despite the increase in hair-cell number, n = 6 neuromasts per age. The % of cells with a SypHy response at day 2 (28.57% ± 3.01) is not different from day 13 (45.47% ± 5.46), p = 0.11. i–n The same subset of hair cells has robust presynaptic Ca²⁺ signals in response to a variety of stimuli, n = 5 neuromasts. Similar to g, in j, fewer active cells are observed in response to a 2-s 50 Hz stimulus. A paired t-test was used for comparisons in f. A one-way ANOVA, df = 30 or 2-way ANOVA, df = 25, with post-hoc Tukey’s test to correct for multiple comparisons were used in g and h, respectively; *p < 0.05, **p < 0.01. Scale bars = 5 μm

Fig. 3

Robust hair-cell Ca²⁺ influx corresponds with postsynaptic Ca²⁺ responses. a A double transgenic line with afferent neurons expressing GCaMP6s-CAAX to detect postsynaptic Ca²⁺ activity (cyan) and hair cells expressing Ribeye b-mCherry to label ribbons (magenta). b, c Representative neuromast demonstrating spatial patterns of afferent GCaMP6s signals during a 200-ms step stimulation (right panels). GCaMP6s signals correspond to the ∆F/F heat map and are superimposed onto the baseline GCaMP6s image (b, left panel) or relative to presynaptic ribbons (c, left panel). Only a subset of hair cells is associated with postsynaptic Ca²⁺ activity. d Dashed circles indicate ROIs (diameter of 3 µm) used to detect afferent Ca²⁺ signals. e Plot of the postsynaptic Ca²⁺ signals in the six cells with (white) and 10 cells without (orange) postsynaptic Ca²⁺ activities (mean with upper and lower limits plotted). f The percentage of hair cells associated with afferent Ca²⁺ activity (naive blue, 35% ± 2.76) is decreased after 10 µM DNQX treatment (purple, 5.50 ± 2.24) to block postsynaptic AMPA receptors, n = 8 neuromasts, p < 0.0001. g Double transgenic with hair cells expressing RGECO1 (magenta) and afferent neurons expressing GCaMP6s-CAAX (cyan). h–l RGECO1 and GCaMP6s responses acquired from the same neuromast organ during a 2-s 5 Hz (anterior–posterior directed square wave) stimulus that activates all hair cells. Hair-cell RGECO1 (h, right panel) or afferent GCaMP6s (i, right panel) responses during stimulation are superimposed onto the baseline grayscale images (h, i, left panels). j–l 10 ROIs (3 µm) outlined in j were used to generate plots of the hair-cell RGECO1 Ca²⁺ (k) or GCaMP6s afferent Ca²⁺ (l) signals. Hair cells with strong Ca²⁺ influx (cells 1–3, k) also have afferent Ca²⁺ signals (l). m RGECO1 hair-cell Ca²⁺ signals associated with afferent Ca²⁺ signals (19.84% ± 2.38, n = 31 cells) were larger than those without (6.04% ± 0.62, n = 50 cells), p> 0.0001. A paired t-test was used in f, a Mann–Whitney test was used in m; ***p < 0.001, ****p < 0.0001. Scale bars = 5 μm

Fig. 4

The proportion of active cells is not dependent on innervation. a–c Presynaptic Ca²⁺ profiles of a representative neuromast in response to a 2-s 5 Hz (anterior–posterior directed square wave) stimulus that activates all hair cells, before (b) and after decapitation to eliminate efferent activity (c). Spatial patterns of GCaMP6s Ca²⁺ activities during stimulation are colorized according to the ∆F/F heat map and superimposed onto the baseline GCaMP6s image (a). d Presynaptic Ca²⁺ responses before (49.90% ± 7.24) and after decapitation (32.31% ± 4.70), n = 50 cells, p < 0.0001. Immunostaining of a wild-type (WT, e) and neurog1a morphant (MO, f) neuromast. Ribeye b labels presynaptic ribbons (yellow), Acetylated Tubulin (AceTub, magenta) labels afferent neurons, and Vamp2 (cyan) labels efferent neurons. The neurog1 morphants lack afferent (magenta) and efferent (cyan) innervation. g There is no significant difference in the magnitude of the presynaptic Ca²⁺ responses between the WT (32.64% ± 4.36) and neurog1a morphants (37.73% ± 4.93), n = 59 hair cells, p = 0.79. h There is no significant difference in the percentage of hair cells with vesicle fusion per neuromast between WT (38.94% ± 5.97) and neurog1a morphants (32.53% ± 5.75), n = 7 neuromasts, p = 0.45. A Wilcoxon test was used in d, a Mann–Whitney test in g, and an unpaired t-test in h; ****p < 0.0001. Scale bars = 5 μm

Fig. 5

Low [K⁺] and gap junctions may facilitate presynaptic activity. a Presynaptic Ca²⁺ jRCaMP1a Ca²⁺ signals during a 2-s 5 Hz stimulus. Spatial patterns of jRCaMp1a Ca²⁺ signals during stimulation are colorized according to the ∆F/F heat maps and superimposed onto a pre-stimulus baseline jRCaMP1a image (a, left panel). b Image of the same neuromast as a after labeling with the K⁺ indicator, APG-2. The active cells in a and b are marked with asterisks. c Quantification of APG-2 intensity shows that active cells (341.1 a.u. ± 33.17, n = 14 cells) have lower resting [K⁺]_in levels than silent cells (520.1 a.u. ± 28.37, n = 46 cells) p = 0.001. d APG-2 dye labeling before (d, left panel) and after (d, right panel) FFA treatment to block gap junctions. e, f Quantification of APG-2 intensity shows 25 µm FFA significantly increases [K⁺]_in levels in hair cells (naïve: 950.2 a.u. ± 67.9; after FFA: 1105 a.u. ± 99.7, n = 48 cells, p = 0.01) and in supporting cells (naïve: 767.1 a.u. ± 34.4; after FFA: 845.1 a.u. ± 50.3, n = 48 cells, p = 0.015). g–i, k–m Mechanosensitive and presynaptic GCaMP6s Ca²⁺ signals within the same cells before and after application of 25 μM FFA. Mechanosensative (h) and presynaptic (l) Ca²⁺ signals during a 2-s 5 Hz stimulus prior to drug treatment; 25 μM FFA does not alter mechanotransduction (i) but decreases presynaptic Ca²⁺ responses (m). Spatial patterns of GCaMP6s Ca²⁺ signals during stimulation are colorized according to the ∆F/F heat maps and superimposed onto baseline GCaMP6s images (g, k). j Quantifications of the mechanosensitive Ca²⁺ signals show no significant differences before (86.10% ± 7.55) and after 25 μM FFA (85.18% ± 6.64), n = 60 hair cells, p = 0.76. n In the same cells as j, presynaptic Ca²⁺ signals (41.55% ± 4.85) are significantly decreased after FFA application (20.27% ± 3.07), n = 60 cells, p < 0.0001. A Mann–Whitney test was used in c, a Wilcoxon test was used in e, f, j, and n; *p < 0.05, **p < 0.01, ****p < 0.0001. Scale bars = 5 μm

Fig. 6

Depolarization does not activate Ca_V1.3 channels in all cells. a–c Example neuromast with immunostain labeling presynaptic ribbons (Ribeye b, magenta) in a, and presynaptic Ca_V1.3a channels (cyan) in b. Images in a and b are merged in c. d Quantification of the percent of Ribeye b positive ribbons per neuromast that colocalize with Ca_V1.3a (93.00% ± 1.93, n = 10 neuromasts). e–h Representative neuromast demonstrating spatial patterns of hair-cell presynaptic GCaMP6s Ca²⁺ signals during mechanical and high K⁺ stimulation. Signals are colorized according to the ∆F/F heat map and superimposed onto baseline GCaMP6s image (e). f A subset of cells shows presynaptic Ca²⁺ signals in response to a 2-s 5 Hz mechanical stimulus. g All mechanical responses are eliminated after BAPTA treatment. h After BAPTA treatment, to depolarize hair cells, high K⁺ was applied (via the fluid jet). High K⁺ application activates the same cells as the original mechanical stimulus (f). i, l Example neuromasts demonstrating spatial patterns of hair-cell Bongwoori voltage signals during high K⁺ stimulation. Spatial patterns of Bongwoori voltage signals during high K⁺ stimulation are colorized according to the ∆F/F heat map and superimposed onto baseline Bongwoori image (i, l, left panels). k, n Using 5 µm ROIs shown in j, m, all cells show depolarization in response to high K⁺ application. Inset in k demonstrates that changes in Bongwoori signals in the background skin pigment (ROIs a, b) do no correlate with the stimulus. Scale bars = 5 μm

Fig. 7

Laser damage to an active cell activates silent cells. a–f Presynaptic GCaMP6s signals from two representative neuromast organs before and after laser damage. b, e Presynaptic Ca²⁺ signals during a 2-s step stimulus (anterior and posterior responses are merged in the profile) prior to laser damage. Spatial patterns of GCaMP6s Ca²⁺ signals during stimulation are colorized according to the ∆F/F heat map and superimposed onto baseline GCaMP6s images (a, d). c, f Presynaptic Ca²⁺ signals after high power laser-induced damage to one active cell in each neuromast (green lightning bolt, cell 1 in b and e). After laser damage, the damaged cell (cell 1) no longer displays a detectable presynaptic Ca²⁺ signal, but new cells now display robust presynaptic Ca²⁺ signals (cell 2 in c and cells 2–4 in f). Scale bars = 5 μm

Fig. 8

Model of presynaptic silencing in neuromast organs. Left-side, Model 1: low [K⁺]_in is causative for Ca_V1.3 channel (red) activation. In active cells, (1) similar to glia, supporting cells may use Kir4.1 channels (green) to take up [K⁺]_ex, gap junctions (Cx, blue) to spatially buffer K⁺ among syncytia of supporting cells, and a K⁺-Cl⁻ cotransporter such as KCC3 (yellow) to clear K⁺ from supporting cells. This results in (2) lower [K⁺]_in in active cells. Cells with lower [K⁺]_in may be at sufficiently depolarized resting membrane potentials where (3) mechanosensation and depolarization (∆V) is able to (4) activate Ca_V1.3 channels. In this model, despite K⁺ buffering by supporting cells, not all hair cells are able to be maintained with [K⁺]_in levels low enough to facilitate presynaptic function. Blocking gap junctions elevates hair-cell [K⁺]_in and presynaptic function is lost in all cells. After laser damage (green lightning bolt), a signaling cascade could lower [K⁺]_in levels in silent cells, and raise resting membrane potentials into a range suitable for Ca_V1.3 channel activation. Right-side, Model 2: low [K⁺]_in is a consequence of presynaptic activity. This schematic demonstrates that (1) mechanosensation and depolarization (∆V), lead to (2) Ca_V1.3 channel activation in active cells. Presynaptic activity results in (3) lower [K⁺]_in. In this model, Ca_V1.3 channels may be inactive due to a lack of biochemical modification, such as phosphorylation (P), as shown in this example. Alternatively, Ca_V1.3 channels could also be rendered inactive due to a missing interaction partner or improper assembly at the plasma membrane (these examples not shown). In this example, after laser damage (green lightning bolt), a signaling cascade could promote phosphorylation and lead to Ca_V1.3 channel activation