Functional development and regeneration of hair cells in the zebrafish lateral line

Hair cells are mechanosensory receptors responsible for transducing auditory and vestibular information into electrical signals, which are then transmitted with remarkable precision to afferent neurons. Different from mammals, the hair cells of lower vertebrates, including those present in the neuromasts of the zebrafish lateral line, regenerate following environmental or chemical insults. Here we investigate the time-course of regeneration of hair cells in vivo using electrophysiology, 2-photon imaging and immunostaining applied to wild-type and genetically-encoded fluorescent indicator zebrafish lines. Functional hair cells drive spontaneous action potentials in the posterior lateral line afferent fibres, the frequency of which progressively increases over the first 10-days post-fertilization (dpf). Higher firing-rate fibres are only observed from ~6 dpf. Following copper treatment, newly formed hair cells become functional and are able to drive APs in the afferent fibres within 48 hours in both early-larval (≤8 dpf) and late-larval (12-17 dpf) zebrafish. However, the complete functional regeneration of the entire neuromast is delayed in late-larval compared to early-larval zebrafish. We propose that while individual regenerating hair cells can rapidly become active, the acquisition of fully functional neuromasts progresses faster at early-larval stages, a time when hair cells are still under development. At both ages, the afferent terminals in the regenerating neuromast appear to make initial contact with supporting cells. The ablation of the lateral line afferent neurons prevents the timely regeneration of supporting cells and hair cells. These findings indicate that the afferent system is likely to facilitate or promote the neuromast regeneration process.


Introduction
Damage to the auditory hair cells can result from a wide range of factors including environmental and chemical insults, genetic predisposition and ageing. While humans and other mammals have limited or no hair cell regenerative abilities, lower vertebrates such as the zebrafish can fully regenerate them following damage (e.g. Corwin & Oberholtzer, 1997;Kelley, 2003;Rubel et al. 2013). In the zebrafish lateral line, the hair cell regeneration process is rapid (Harris et al. 2003;López-Schier & Hudspeth, 2006;Hernandez et al. 2007), with different populations of supporting cells serving as progenitors for the newly formed hair cells (Thomas & Raible, 2019). Although some of the mechanisms leading to hair cell regeneration in the lateral line have been identified (e.g. Atkinson et al. 2015;Monroe et al. 2015;Thomas et al. 2015), little is known about the temporal profile defining the ability of hair cells and their afferent innervation to transduce spontaneous and mechanically induced responses during regeneration.
Zebrafish use the lateral line to signal hydrodynamic variations and water currents around the body (Dijkgraaf, 1963;Kroese & Schellart 1992). The lateral line includes the anterior (head) and posterior (tail) regions that, by sensing low-frequency signals (up to 200 Hz: Ghysen & Dambly-Chaudiere, 2004), mediate several behavioural tasks such as schooling, prey capture, predator avoidance and rheotaxis (e.g. Montgomery et al. 1997;Engelmann et al. 2000;Bleckmann & Zelick, 2009). The functional unit of the lateral line is the neuromast, which consists of hair cells clustered in a rosette-like structure on the surface of the fish (MetCalfe et al. 1985;Sheets et al. 2012;Olt et al. 2016). Each neuromast contains two sets of hair cells responding to opposite directional stimuli, a feature which is determined by the specular orientation of their mechanosensitive stereociliary bundle (Flock & Wersäll, 1962;López-Schier et al. 2004). Hair cells are innervated by bipolar afferent neurons that merge in cephalic ganglia (MetCalfe et al. 1985;Ghysen & Dambly-Chaudiere, 2004). Each afferent neuron receives information from hair cells present in one or multiple neuromasts, but of the same polarity (Nagiel et al. 2008;Obholzer et al. 2008;Faucherre et al. 2009). Primary neuromasts are initially deposited by a primordium along the length of the fish during the first 2 days post-fertilisation (dpf) (Pujol-Martí & López-Schier, 2013). Although hair cells are already functional at 3 dpf, the majority of them exhibit an immature biophysical profile (Olt et al. 2014). The proportion of mature-like hair cells largely increases over the following several days, and seems to mainly cluster at the centre of the neuromast (Olt et al. 2014). During this initial developmental period, afferent fibres are also under the influence of intrinsic developmental cues and undergo extensive reorganisation . Considering that most of the hair-cell regeneration studies in the zebrafish lateral line are performed at around 3-5 dpf, it is unclear whether the reported rapid time course of regeneration is also a characteristic present at more functionally mature ages.
Here we performed electrophysiological, two-photon imaging and immunostaining experiments to investigate the functional recovery of hair cells from larval zebrafish after chemical insult. We found that the functional and morphological regeneration of the hair cells in the lateral line occurred faster in early-than late-larval zebrafish. afferent terminals appear to innervate the regenerating neuromast devoid of hair cells, and instead make initial contact with the supporting cells. In zebrafish that underwent ablation of the posterior lateral line ganglion (PLLg), the normal rapid regeneration of the supporting cells and hair cells was prevented. This indicates that the afferent system plays a key role in the timely response of the regeneration process.

Ethics statement and zebrafish lines
All animal work was performed at the University of Sheffield (UK), licensed by the Home Office under the Animals (Scientific Procedures) Act 1986 (PPL_P0FBE066C) and approved by the University of Sheffield Ethical Review Committee (180 626_Mar). The experimental work also conforms to the principles and regulations as described in the editorial by Grundy (2015). Zebrafish (Danio rerio) were raised in petri dishes at ∼28°C in an incubator until 5.2 days post fertilisation (dpf) when they were transferred to a light/dark incubator and fed powdered food. Zebrafish aged >10 dpf were maintained in tanks in the aquarium facility on a 14/10-h light/dark cycle and fed artemia twice daily. At the end of each procedure, zebrafish were culled by immersion in a solution containing 0.1-0.4% of the anaesthetic tricaine methanesulfonate (MS-222; Henry Schein, Inc., Dumfries, UK) until unresponsive and then decapitated.
Experiments were performed using the following zebrafish lines: wild-type AB; Tg(NeuroD:EGFP), which express the green fluorescent protein EGFP in the afferent neurons to visualise the afferent terminals and posterior lateral line ganglion (Obholzer et al. 2008); Tg(brn3c:GAL4);Tg(UAS:iGluSnFR), which express the glutamate sensor iGluSnFR (Marvin et al. 2013) in the hair cells to monitor vesicle release at the ribbon synapses; Tg(Myosin6b:R-GECO), which express the red fluorescent Ca 2+ indicator R-GECO in the hair cells (Maeda et al. 2014); Tg(NBT:GCaMP3), which express the green fluorescent Ca 2+ indicator GCaMP3 in the afferent neurons (Bergmann et al. 2018) and allowed us to measure Ca 2+ entry at their terminals (Sebe et al. 2017).
Under the husbandry conditions in place at the University of Sheffield and using information from previous studies (Kimmel et al. 1995;Parichy et al. 2009), larval zebrafish are those from ∼3 dpf to ∼2 weeks post-fertilisation (2 wpf), juvenile zebrafish are those from ∼2 wpf to the point at which they become sexually mature (3-6 months) (see also Olt et al. 2014). Because most of the work in this study was done within the larval stage of development (2-17 dpf), we classified 'early-larval' zebrafish as those aged between 2 and 10 dpf and 'late-larval' zebrafish as those from 11 to 17 dpf. The average lengths of zebrafish maintained in our facility, and over the age range investigated, are shown in Fig. 1A.

Patch-clamp electrophysiology
Afferent neuron recordings were obtained from the cell bodies of the posterior lateral line ganglion (PLLg). Zebrafish aged 2-5.2 dpf were briefly treated with MS-222 before being paralysed by injecting 125 μM α-bungarotoxin (Tocris, Bristol, UK) into the heart cavity (Olt et al. 2014(Olt et al. , 2016. The zebrafish (<5.2 dpf) were then transferred to a microscope chamber and immobilised onto a thin layer of Sylgard using fine tungsten wire (0.015 mm diameter, Advent Research Materials, Oxford, UK). Because α-bungarotoxin injections could not be performed at >5.2 dpf (zebrafish then become protected animals in the UK), older zebrafish were restrained using 2% low melting-point agarose dissolved in E3. The agarose was also used to fix the restrained zebrafish to the bottom of the recording chamber on their side. In order for the patch electrode to access the PLLg, the agarose posterior to the head of the fish was gently removed. Restrained zebrafish (either using tungsten wire or agarose) were continuously superfused by a peristaltic pump with extracellular solution containing (in mM): 140 NaCl, 2.0 CaCl 2 , 2.0 KCl, 1.0 MgCl 2 , 25 NaHCO 3 , pH adjusted to 7.8 with NaOH. For zebrafish restrained in agarose at >5.2 dpf, extracellular solution was continuously bubbled with carboxygen (95% O 2 and 5% CO 2 ).
The PLLg cell bodies were viewed using an upright microscope (Olympus BX51) equipped with Nomarski Differential Interference Contrast (DIC) optics (60× water immersion objective) and 15× eyepieces. Patch pipettes were made from borosilicate capillaries (12-14 M ). The loose-patch configuration (35-120 MΩ) was used to record spontaneous action potential activity from the PLLg (Trapani & Nicolson, 2010;Olt et al. 2016). The patch pipette contained (in mM): 140 NaCl, 2.0 KCl, 2.0 CaCl 2 , 1.0 MgCl 2 , 10 HEPES-NaOH (pH 7.8). Action potentials from the PLLg were obtained at room temperature (20-24°C) using a Multiclamp 700B (Molecular Devices, USA) amplifier. Data acquisition was controlled by pClamp software using a Digidata 1440A board (Molecular Devices, USA). Recordings were low-pass filtered at 2.5 kHz (8-pole Bessel), sampled at 5 kHz and stored on computer for off-line analysis (Origin: OriginLab, USA). The Mini Analysis Program (Synaptosoft Inc. NJ, USA) was used to detect spike events in the loose-patch configuration to calculate their frequency and to analyse interspike intervals (ISIs). The firing rate of the action potentials was calculated as the reciprocal of the mean ISI for each cell and an indication of the spread of ISI values about the mean was obtained by calculating the coefficient of J Physiol 599.16 variation (CV), equal to the standard deviation (SD) divided by the mean.

Application of the ototoxic compound copper sulphate
The ablation of hair cells and afferent terminals was achieved by the application of different concentrations of copper (II) sulphate (cat no. 451657, Sigma-Aldrich, UK), which was dissolved in plain embryonic medium E3 (in mM: 5 NaCl, 0.17 KCl, 0.33 CaCl 2 , 0.3 MgSO 4 ). The different zebrafish lines (see above) at 3 dpf (early-larval stage) or 12 dpf (late-larval stage) were bath-treated with copper for 2 h at 28.5°C in the dark. Following this, the fish were washed three times in E3 and left to recover in Petri dishes (<5.2 dpf) or in a non-circulating system tank (>5.2 dpf) prior to being processed for the different experiments. The use of copper was preferred to the ototoxic aminoglycoside antibiotics (e.g. neomycin and streptomycin), since under our experimental conditions streptomycin was only able to reduce the number of hair cells within a neuromast by about 70%, even at very high concentrations (2 mM: data not shown but available upon request). In addition, high concentrations of copper, but not aminoglycosides, also affected the survival of the afferent fibres and supporting cells (see Fig. 10), which allowed us to compare the time course of regeneration of different cell populations within the neuromast.

Ablation of the afferent fibres or PLLg
For ablation of the PLL afferent nerve or the PLL ganglion, we immobilized copper-treated NeuroD zebrafish (3 dpf) in low melting-point agarose (2%) at the bottom of a microscope chamber, which was then transferred to the stage of a dual-laser two-photon microscope (see below). The PLL was visualized using the fluorescence emission of EGFP/NeuroD under two-photon excitation (wavelength: 925 nm) using a 60× objective (Olympus). The ablation of the PLLg or the severing of the PLL afferent nerve was performed using a second un-attenuated laser beam (wavelength: 716 nm, Mai Tai BB, Spectra-Physics, USA), which was focused through the imaging objective and stirred by a pair of galvanometric mirrors across the focal plane. This second laser beam was applied in a spiral pattern on a small area (afferent severing: 425 μm 2 , 50-100 repetitions; PLLg ablation: 1500 μm 2 , 50-100 repetitions) and the ablation of the PLLg or the severing of the PLL afferent nerve was visually confirmed through the EGFP fluorescence emission. Zebrafish were then gently freed from the agarose using a pair of fine forceps and left to recover in a Petri dish. Zebrafish used as the experimental control (copper-untreated) were subjected to the same experimental procedure as those treated, but omitting the copper.

Hair bundle stimulation
The mechanical displacement of the cupula containing the stereociliary bundles of the hair cells in the neuromast was achieved by using a fluid jet from a pipette driven by a 25 mm diameter piezoelectric disc (Erickson et al. 2017;Carlton et al. 2021;Jeng et al. 2021). The fluid jet pipette tip had a diameter of ∼30 μm and was positioned at ∼100 μm from the cupula. The distance of the pipette tip from the cupula was adjusted to elicit a maximal bundle displacement. Mechanical stimuli were applied as pressure steps (filtered at 1 kHz, 8-pole Bessel).
The quantification of the number of hair cells, supporting cells, CtBP puncta and afferent terminals was performed by using z-stack images captured with either a Nikon A1 confocal microscope or a Zeiss LSM 880 with AiryScan for super-resolution confocal microscopy from the Wolfson Light Microscope Facility at the University of Sheffield. Afferent terminals were identified by the presence of well-defined enlargements innervating the neuromast. Images were taken from the first 5 posterior lateral line neuromasts for zebrafish up to 5 dpf, and neuromasts 1-8 for older ages. Images and z-stacks were processed using ImageJ analysis software.

Two-photon imaging and displacement of the hair cell stereociliary bundles
Hair cell recordings were performed from the primary neuromasts (L2-L4) originating from the first primordium (primI) (Pujol-Martí & López-Schier, 2013). Calcium and glutamate signals in hair cells and afferent terminals were recorded using a two-photon laser-scanning microscope (Bergamo II System B232, Thorlabs Inc., USA) based on a mode-locked laser system operating at 925 nm, 80 MHz pulse repetition rate, <100 fs pulse width (Mai Tai HP DeepSee, Spectra-Physics, USA). Images were captured with a 60× objective (LUMFLN60XW, Olympus, Japan) using a GaAsp photomultiplier (Hamamatsu) coupled with a 525/40 bandbass filter (FF02-525/40-25, Semrock). Images were analysed offline using custom-built software routines written in Python (Python 3.7, Python Software Foundation) and ImageJ (NIH). Ca 2+ signals were measured as relative changes of fluorescence emission intensity (F/F 0 ). Images were acquired at 15 (512 × 512 pixels, Ca 2+ imaging) or 396 (256 × 32 pixels, when using the iGluSnFR zebrafish) frames per second. Signals from individual ROIs were smoothed offline using a Savitzki-Golay filter (windows size:11; polynomial order: 1). Traces were normalised to baseline fluorescence (F 0 ) which was calculated as the 5th percentile of the fluorescence values of the entire trace.

Statistical analysis
Statistical comparisons of means were made by Student's two-tailed t test or, for multiple comparisons, analysis of variance (one-way or two-way ANOVA followed by an appropriate post hoc test: Tukey's or Sidak's post hoc test) was applied. P < 0.05 was selected as the criterion for statistical significance. Mean values are quoted in the text and figures as means ± SD. Statistical analysis was performed using Prism 8.0 (GraphPad Prism Software, Inc., San Diego, CA, USA) and R (R foundation for Statistical Computing, R Core Team 2020).

Spontaneous action potential activity in the afferent fibres
Action potential (AP) activity has been shown to occur in the lateral line afferent fibres at around 5 days post-fertilisation (dpf) (e.g. Trapani & Nicolson, 2011;Levi et al. 2015), a time when the zebrafish is swimming and thus receiving sensory inputs from the surrounding environment. Therefore, we investigated whether swimming behaviour is required for the onset of spontaneous AP activity by performing experiments at 2 dpf, which is a time when zebrafish normally hatch from the chorion (Kimmel et al. 1995). Extracellular loose-patch recordings from the cell body of the neurons of the posterior lateral line ganglion (PLLg; Trapani & Nicolson, 2010Olt et al. 2016) were performed from prematurely manually hatched and swimming zebrafish. We found that PLLg neurons from manually hatched zebrafish, with no prior swimming experience, already exhibited spontaneous AP activity with a mean frequency (1.63 ± 1.45 Hz, N = 6, Fig. 1B) that was not significantly different to that recorded from aged-matched swimming fish that hatched naturally (1.89 ± 0.95 Hz, N = 9, Fig. 1C, P = 0.6812, t test). These data indicate that the onset of spontaneous firing activity does not require sensory inputs due to the swimming behaviour. This AP activity is likely to be driven by the spontaneous release of glutamate from hair cells ( Fig. 1D; supplementary Video 1 in Supporting Information).
In the UK zebrafish become protected from the day of independent feeding at larval stage, which at 28.5°C holding temperature corresponds to 5.2 dpf. Therefore, while in vivo experiments from zebrafish up to 5.2 dpf were done using pinned down zebrafish injected with the neuromuscular blocker α-bungarotoxin (Trapani & Nicolson, 2011;Levi et al. 2015;Olt et al. 2014), at older ages zebrafish were restrained in 2% agarose. However, we found that the different restraining method had no influence on the afferent AP firing activity, since the frequency from pinned-down 4 dpf zebrafish (4.6 ± 3.5 Hz, N = 20) was not significantly different to that recorded in aged-matched agarose-restrained zebrafish (6.6 ± 2.6 Hz, N = 6, P = 0.2050, t test).
The neuromasts and afferent fibres of the zebrafish lateral line undergo extensive growth and reorganisation during larval stages Haehnel et al. 2012;Olt et al. 2014), which may be essential for the fine tuning of the sensitivity to movement direction. Therefore, we investigated whether the characteristics of afferent spontaneous activity would also change with age.  Note that the hair cells in the top image refer to the recordings depicted in the blue and light blue traces; the image of the second hair cell (black and grey traces) is not shown. Note the high frequency signals present in the synaptic traces denoting spontaneous glutamate release (see also supplementary Video 1 in Supporting Information). The light traces are plotted as a visual reference of the intrinsic noise level of the recordings. E, average frequency of spontaneous APs as a function of dpf. Overall 598,004 spikes from 133 PLLg (8.1 ± 2.9 min/PLLg). F, the AP frequency as a function of dpf was subdivided into low-(black circles) and high-frequency (red circles) clusters using k-means clustering. The data were fitted with a Boltzman equation, with a half-peak value of 5.2 dpf. G, coefficient of variation (CV) of the spontaneous APs plotted as a function of dpf. The red dashed line indicates the CV value for a random Poisson process, which is 1. H, histogram of CV values (bin size, 0.1) for the two clusters highlighted in black and red in panel F. I, histogram of individual AP inter-spike intervals of the two clusters identified in panel F. J, CV from each PPLg recording against their firing rate for the two groups. K, example of spontaneous APs from a PPLg of a 10 dpf zebrafish showing a bursting firing pattern (CV: 2.7; frequency: 6.9 Hz). In this and the following figures, data are presented as means ± SD.
We found that the mean frequency of spontaneous APs significantly increased with age (P < 0.0001, one-way ANOVA, Fig. 1E). We performed unbiased classifications of afferent neurons based on their mean firing rate using k-means clustering with k = 2 (Fong et al. 2014). Based on the mean frequency rate, PLLg neurons were subdivided into a low-frequency cluster (black circles: 5.12 ± 3.78 Hz, N = 106, 2-17 dpf), which was present throughout the age-range tested, and a high-frequency cluster that only appeared from around 6 dpf (red circles: 24.63 ± 8.19 Hz, N = 38, P < 0.0001, t test, Fig. 1F). We then used the coefficient of variation as a quantitative measure of regularity for spontaneous spike firing (Sonntag et al. 2009;Jones & Jones, 2000). The coefficient of variation (CV) for a random Poisson process is 1; values < 1 indicate more regular activity and values > 1 indicate irregular or bursting activity. The average CV across the entire age-range was 1.30 ± 0.69 (N = 144) and did not change significantly over the age-range tested (P = 0.8735, one-way ANOVA, Fig. 1G). When the CV values were divided into the two clusters identified with the k-mean analysis (Fig. 1F), we found that the firing activity of the high-frequency PLLg neurons was more regular (1.00 ± 0.49 Hz, N = 38) compared to low-frequency neurons (1.41 ± 0.72, N = 106, P = 0.0016, t test, Fig. 1H-J). Some high-frequency neurons displayed a burst-like firing pattern (Fig. 1K).

Hair cell regeneration following chemical-induced insult
Previous work investigating hair-cell regeneration in the zebrafish lateral line has mainly been done using morphological approaches applied to transgenic fish with hair cells tagged with fluorescent reporters (e.g. GFP-or tdTomato: Pinto-Teixeira et al. 2013; Monroe et al. 2015) or fluorescent vital dyes such as DASPEI, 4-Di-2-ASP and FM1-43FX (e.g. Harris et al.2003;Ma et al. 2008;Nagiel et al. 2008). Although these studies have provided valuable information regarding the temporal profile of hair cell regeneration, very little is currently known about their functional regeneration.
We initially established the experimental protocol required to ablate the hair cells with copper, which is an ototoxic drug known to cause their death (Hernández et al. 2006;Mackenzie & Raible, 2012). Three-day-old zebrafish were incubated for 2 h in copper at increasing concentrations of between 0.1 and 250 μM ( Fig. 2A). The number of hair cells per neuromast, which was measured by counting the remaining labelled cells from the transgenic zebrafish Tg(Myosin6b:R-GECO) (see Methods), was plotted as a function of copper concentration (Fig. 2B). The concentration of copper that caused a 50% reduction in hair cell number was obtained by fitting the data with a Hill equation (IC 50 : 2.6 μM, Fig. 2C). Because 10 μM was the highest concentration of copper that, while ablating ∼95% of hair cells, did not cause afferent fibre retraction (see Fig. 10A), it was used for most of the following regeneration experiments. Following copper-induced hair cell death, their number per neuromast gradually increased reaching a mean value comparable to that present in untreated control zebrafish after 48 h post-treatment (hpt) ( Fig. 2D and E, P < 0.0001, two-way ANOVA).

Functional regeneration of spontaneous and induced activity in the lateral line of early-larval zebrafish
Functional regeneration was initially assessed by investigating the ability of newly formed hair cells to drive spontaneous AP activity in the PLLg neurons. Action potentials in the PLLg were absent in 10 out of 11 neurons from zebrafish between 0.5 and 4.5 hpt ( Fig. 3A-C), but both the frequency and CV became comparable to those recorded in untreated zebrafish towards the end of the first day post-treatment ( Fig. 3B and C, respectively). We then used the iGluSnFR zebrafish line to investigate whether the regenerating hair cells were capable of neurotransmitter release. Considering that in our zebrafish line iGluSnFR is only expressed in a few hair cells within a neuromast, we found that all 10 hair cells expressing the fluorescent reporter from four untreated zebrafish showed spontaneous glutamate release (Fig. 3D), which was visible as high frequency spiking in the iGluSnFR fluorescence trace. In the copper-treated zebrafish, spontaneous glutamate release was detected in only 12 out of 20 hair cells expressing iGluSnFR (36-48 hpt, Fig. 3E), indicating the presence of hair cells at different regenerating stages at this time point.
The ability of regenerating hair cells to faithfully respond to sensory-induced stimuli was investigated by deflecting the neuromast cupula containing the stereociliary bundles towards the excitatory direction with a fluid-jet (Fig. 4A), a procedure that gate-opens the mechanoelectrical transducer channels located at the tip of the transducing stereocilia. Excitatory cupula displacements caused the frequency of the spiking activity of the PLLg neurons of 4-5 dpf zebrafish to increase rapidly at the stimulus onset from the baseline (Fig. 4B  and D). A similar increase in firing rate was also observed in aged-matched zebrafish that underwent regeneration for 48 hpt (Fig. 4C and D). The peak firing rate (Fig. 4E) and first spike latency (Fig. 4F) of the AP activity were not significantly different in the two experimental conditions.
The displacement of the cupula in untreated early-larval zebrafish with a fluid-jet elicited Ca 2+ responses in hair cells (red: Fig. 5B and C) and afferent terminals (green: Fig. 5B and D; see also supplementary Video 2 in Supporting Information). After treating 3 dpf Tg(Myosin6b:R-GECO);Tg(NBT:GCaMP3) zebrafish with 10 μM copper, we found that the percentage of regenerating hair cells per neuromast responding with Ca 2+ signals during the deflection of the cupula was significantly lower than that from age-matched control zebrafish (Fig. 5E, P < 0.0001, two-way ANOVA). The same was also found for the number of postsynaptic terminals displaying Ca 2+ elevations during fluid-jet stimulation (Fig. 5F, P = 0.0100, two-way ANOVA). However, by 48 hpt, the fractions of responding hair cells and synaptic contacts were no longer significantly different from controls (P > 0.9999, Sidak's post hoc test, for both Fig. 5E and F). Interestingly, we found that only ∼50% of the postsynaptic afferent terminals within a neuromast showed Ca 2+ signal changes during excitatory displacement of the cupula (Fig. 5F). This indicates that a large number of hair cells did not release any glutamate despite the fact that the large majority of them responded to the displacement of their hair bundles (Fig. 5E). This finding is comparable to that obtained when hair cell synaptic activity was measured in zebrafish expressing SypHy (∼30%: Zhang et al. 2018), which is an indicator of vesicle fusion (Odermatt et al. 2012). The functional recovery of synaptic function following hair cell ablation (Figs 4 and 5) was then correlated with morphological regeneration. Tg(Myosin6b:R-GECO);Tg(NBT:GCaMP3) zebrafish (3 dpf) were treated with 10 μM copper and then immunostained using antibodies to label the ribbon synapses (CtBP1 and CtBP2; see Methods) in hair cells of the lateral line (Sheets et al. 2014;Lv et al. 2016;Kindt & Sheets, 2018). A few CtBP puncta were evident over the few hours following copper treatment, but by 24 hpt their number was not significantly different compared to that found in untreated hair cells (Fig. 6A-C). However, during the 24-48 hpt period, the number of CtBP puncta in regenerating hair cells became very variable (Fig. 6B). A larger number of ribbons during the initial stages of hair cell regeneration has also been reported in 5 dpf zebrafish treated with the aminoglycoside neomycin (Suli et al. 2016), indicating the presence of some degree of synaptic refinement during regeneration. The number of CtBP puncta that colocalised with the afferent terminals was also comparable between the two experimental conditions by 24 hpt (Fig. 6C). We also observed the presence of several afferent terminals within the regenerating neuromasts that were still devoid of hair cells (Fig. 6D). The absence of visible hair cells was not due to the inability of the Tg(Myosin6b:R-GECO) reporter line to identify newly formed hair cells since Myosin6b has been shown to be expressed in hair cell precursors (Seiler et al. 2004), and R-GECO labels nascent hair cells well before they can be detected by antibody staining (data not shown but available upon request).
The above data show that functional regeneration of hair cells in early-larval zebrafish can occur within 1 day of copper-induced cell death. However, due to the continuing addition of hair cells and the extensive synaptic refinement in the regenerating lateral line, approximately 2 days are required for a neuromast to reacquire normal functionality.

Functional regeneration of hair cells in late-larval zebrafish requires a longer timescale than early-larval zebrafish
We sought to determine whether the timescale of regeneration in the early-larval lateral line was mirrored in the late-larval zebrafish, when neuromasts contain a higher proportion of mature hair cells (Olt et al. 2014). Because of the additional ethical implications and The image shows a PLL neuromast containing hair cells of opposite polarity (light and dark grey). The hair bundles of the hair cells are displaced by saturating stimuli applied using a piezo driven fluid-jet (see Methods) while recording firing activity in the PLLg neuron connected to the neuromast. B and C, representative AP recordings from the PLLg afferent neurons while deflecting the cupula of a connected neuromast with the fluid jet (see Methods) from a control (B) and copper-treated (C) zebrafish. The top trace represents the 3 s saturating driving voltage step to the piezoelectric actuator. Note the increase in firing rate at the onset of the stimulus and the subsequent adaptation. D, raster plot of individual afferent neuron activity during the application of the stimuli in the excitatory direction (top panels: driving voltage). E and F, the peak firing rate (E) and the latency between the stimulus onset and the generation of the first spike (F) were not significantly different between control and copper-treated zebrafish (P = 0.4956 and P = 0.5243, t test, respectively). experimental complexities when working with late-larval zebrafish, only a subset of the experiments performed in the larvae were carried out in older zebrafish. As for larvae, the treatment of late-larval zebrafish (12 dpf) with 10 μM copper for 2 h caused the loss of hair cells in the lateral line neuromasts. Although the number of regenerating hair cells per neuromast gradually increased between 24 and 120 hpt, it was significantly reduced compared to that present in control neuromasts over the same time period (Fig. 7A and B, P < 0.0001, two-way ANOVA). This finding indicates that neuromasts from late-larval zebrafish require more than 3 additional days to fully regenerate all the hair cells compared to larvae (Fig. 2E). This discrepancy is not due to the different number of hair cells present in a fully functional neuromast after copper treatment, since it was comparable between early-larval (72-120 hpt or 6-8 dpf: 15.0 ± 2.3, N = 23, Fig. 2E) and late-larval zebrafish (0-120 hpt or 12-16 dpf: 16.1 ± 4.4, N = 71, P = 0.2547, Fig. 7B). While at 0 hpt firing activity was absent or sporadic in PLLg neurons, at 24 hpt a more sustained activity was already evident (Fig. 7C-F), although significantly reduced in frequency compared to control zebrafish (Fig. 7E, P < 0.0001, two-way ANOVA). The large variability in the firing activity characteristics of control PLLg neurons (see also Fig. 1E), which is primarily due to the acquisition of high-frequency firing neurons in older zebrafish (Fig. 1F), was largely absent at least during the first few days post-copper treatment in late-larval zebrafish. A similar trend of recovery to that observed for the number of hair cells (Fig. 7B), was also present for the number of CtBP puncta per neuromast and for those that colocalised with the afferent terminals ( Fig. 8A-C), which were significantly reduced compared to those found in control zebrafish over the entire range tested (P < 0.0001, two-way ANOVA, for both comparisons). As shown for early-larval zebrafish (Fig. 6D), we observed the formation Figure 5. Synaptic activity in regenerating early-larval hair cells A, schematic of the experimental set-up used to displace the mechanoelectrical transducer apparatus of the hair cells. The image shows a PLL neuromast containing hair cells of opposite polarity (light and dark grey) viewed by the microscope objective (above the fish). The hair bundles of the hair cells are displaced by saturating stimuli applied using a piezo-driven fluid-jet (see Methods). Hair cells are excited by stimuli which displace their hair bundle towards the kinocilium, opening the MET channels and depolarizing the membrane potential. The increase in intracellular Ca 2+ ensues from the opening of voltage-gated Ca 2+ channels in the plasma membrane. B, image of a 3 dpf zebrafish (Tg(Myosin6b:R-GECO);Tg(NBT:GCaMP3)) neuromast, in which hair cells are labelled red and the afferent fibres/terminals in green (see also supplementary Video 2 in Supporting Information). C, R-GECO responses of four hair cells to two saturating bundle displacement stimuli showed by the driver voltage above the recordings. D, calcium responses (GCaMP3) in the afferent terminals during the same stimulation protocol that elicited hair cell responses (C). Note that the traces in C and D are from the same recording, but were not obtained simultaneously, due to the difficulty of having several hair cells and their afferent terminals in the same focal plane. E, percentage of hair cells showing presynaptic Ca 2+ responses during hair bundle stimulation in both control (black) and copper-treated (red) zebrafish as a function of hpt. F, percentage of afferent synaptic terminals showing GCaMP3 responses during bundle stimulation in control (black) and copper-treated (red) zebrafish as a function of hpt. Number of neuromasts tested is shown above the data; 2-3 neuromasts per zebrafish. Statistical significance from left to right: E: * * P < 0.0001; P < 0.0001; * P = 0.0204; n.s. P > 0.9999; F: * * P < 0.0001; * * P = 0.0003; * P = 0.0036; n.s. P > 0.9999, Sidak's post hoc test two-way ANOVA.

Figure 6. Ribbon synapses and afferent fibres in regenerating early-larval hair cells
A, confocal images of hair cells (blue) and afferent fibres (magenta) within the neuromasts of Tg(Myosin6b:R-GECO);Tg(NBT:GCaMP3) zebrafish at 0, 48 and 120 hpt in control and copper-treated zebrafish. Experiments were done using 3 dpf zebrafish. Ribbon synapses were visualised with an antibody against the presynaptic ribbon protein RIBEYE (CtBP: green). Scale bar: 10 μm. Note that the punctate-like labelling (column: Tg(Myosin6b:R-GECO)) in hair cells, which co-localises with the CtBP labelling (columns: CtBP and Merge), is a characteristic of the Tg(Myosin6b:R-GECO) zebrafish since it was also present in the absence of the anti-CtBP antibody (see Fig. 2). B, number of CtBP puncta present in hair cells from untreated control zebrafish (black) and regenerating hair cells following copper treatment (red) as a function of hpt. Statistical significance from left to right: * * P < 0.0001; * * P < 0.0001; n.s. of several afferent projections within the regenerating neuromasts without hair cells (Fig. 8D). Despite the longer time course required to regenerate the lateral line in the late-larval zebrafish, the newly formed hair cells and their postsynaptic afferent terminals displayed Ca 2+ responses when stimulated using the fluid-jet. Moreover, the proportion of active hair cells and afferent terminals was comparable to those in untreated zebrafish at 48 hpt (Fig. 9A-E), as also found in early-larval zebrafish (Fig. 5E and F). This suggests that the newly regenerated hair cells can rapidly become functional even in older-larval zebrafish. Compared to early-larval zebrafish (Fig. 5F), the percentage of active postsynaptic afferent terminals within a neuromast in 14-17 dpf zebrafish increased to ∼70% (Fig. 9E), which is comparable to previous measurements of synaptic activity in hair cells at 13 dpf (∼50%: Zhang et al. 2018).
The above findings indicate that the complete functional and morphological regeneration of neuromasts in late-larval zebrafish requires 5 or more days post-treatment, which is substantially longer than that required for early-larval zebrafish.

Afferent input is required for the regeneration of the neuromast
In the regenerating early-and late-larval neuromasts we found several newly formed afferent terminals in regions devoid of hair cells, as judged by the absence of the early hair cell marker myosin 6 (Figs 6D and 8D, respectively). Therefore, we investigated whether the regenerating afferent terminals were initially targeting the supporting cells. Because the afferent fibres appear more resistant than hair cells to copper exposure, we first established the concentration required to fully abolish their protrusions. We found that 2 h treatment with 30 or 50 μM copper removed the large majority of afferent terminals (Fig. 10A and B) and hair cells ( Fig. 10A and C) with comparable effects (P = 0.5169 and P = 0.7612, respectively, one-way ANOVA, Tukey's post hoc test). This range of copper concentrations also ablated around 50% of the supporting cells within each neuromast ( Fig. 10A and D). Therefore, we treated Tg(NeuroD:EGFP;Tg(Myosin6b:R-GECO) early-larval zebrafish with 30 μM copper for 2 h, and subsequently fixed them at different time points during the regeneration of the lateral line. We performed immunostaining with an antibody against Sox2, which is a transcription factor expressed in the supporting cells of the neuromast (Hernández et al. 2007). Although the large majority of neuromasts were deprived of afferent terminals immediately after the copper treatment, by 10 hpt all neuromasts investigated displayed afferent terminals as found in control zebrafish ( Fig. 11A and  B). Indeed, the regeneration of the PLLg nerve and terminals within a neuromast has been shown to be independent of the presence of hair cells (Villegas et al. 2012). The number of Sox2-positive cells within each neuromast also became comparable to that of untreated zebrafish by 10 hpt (Fig. 11A and C). Additionally, several newly formed afferent terminals within the regenerating neuromasts were in close proximity to the supporting cells, especially within the first few hours after copper treatment, when only very few hair cells were present ( Fig. 11D; see also supplementary Videos 3 and 4 in Supporting Information). We next investigated whether the presence of the afferent terminals within the regenerating neuromast Figure 9. Synaptic activity in regenerating late-larval zebrafish hair cells A, images of a neuromast in a control (upper panel) and copper-treated (lower panel) 12 dpf zebrafish (Tg(Myosin6b:R-GECO);Tg(NBT:GCaMP3)), in which hair cells are labelled red and the afferent fibres/terminals green. Scale bar: 10 μm. B and C, Ca 2+ responses in hair cells (R-GECO, B) and afferent terminals (GCaMP3, C). Saturating driver voltage displacing the cupula is shown above the recordings. Note that some hair cells respond to either the excitatory or inhibitory cupula displacement depending on their polarity sensitivity. D, percentage of hair cells showing presynaptic Ca 2+ responses during hair bundle stimulation in control (black) and copper-treated (red) 12 dpf zebrafish as a function of hpt (P = 0.7617, two-way ANOVA). The x-axis shows four time points (hpt) following the application of copper at 12 dpf zebrafish. E, percentage of afferent synaptic terminals showing GCaMP3 responses during bundle stimulation as a function of hpt in control (black) and copper-treated (red) zebrafish (P = 0.6531, two-way ANOVA). In panels E and F, the number of neuromasts tested is shown above the data; 2-3 neuromasts per zebrafish. J Physiol 599.16 was required for the regeneration of new hair cells (Figs 12 and 13). Initially, we severed the afferent nerve of copper-treated zebrafish (30 μM copper for 2 h) downstream to the PLLg, but before the first neuromast L1, (Fig. 12A, see Methods). Immediately after copper treatment (with or without severed nerve: 0 hpt), the number of hair cells and supporting cells within each neuromast followed a comparable regeneration time course (Fig. 12B and C; see also supplementary Fig. S1 in Supporting Information). However, severed afferent nerves were able to regenerate relatively rapidly ( Fig. 12D: see also Grant et al. 2005;Villegas et al. 2012), thus preventing a reliable assessment of the possible role of the afferent input on the reappearance of both hair cells and supporting cells within the neuromast. We next ablated the entire PLLg (Fig. 13A), which completely prevented the regeneration of the afferent nerve, leaving the neuromasts devoid of afferent input ( Fig. 13B-D). In copper-treated zebrafish with the PLLg ablated, there was very little or no regeneration of hair cells (Fig. 13E, P < 0.0001 for copper vs. copper-ablated, Sidak's post hoc test, two-way ANOVA) and supporting cells (Fig. 13F, P < 0.0001) even at 72 hpt. Interestingly, control zebrafish with the ablated PLLg exhibited a progressive reduction in the number of hair cells (Fig. 13E, P < 0.0001, for control vs. control-ablated, Sidak's post hoc test, two-way ANOVA) and supporting cells (Fig. 13F, P < 0.0001) over the investigated time window. These findings indicate that afferent input is required not only for the normal regeneration of the neuromast but also for its maintenance.

Discussion
This study has shown that the firing rate of the spontaneous action potential activity in the PLLg of   (4); afferent fibres and supporting cells (5); and afferent fibres, hair cells and supporting cells (6). The two images in the 'expanded' column (7) highlight the presence of afferent terminals in close proximity to supporting cells. Experiments were done using 3 dpf Tg(NeuroD:EGFP);Tg(Myosin6b:R-GECO) zebrafish. Scale bar: 10 μm. B, percentage of neuromasts showing afferent terminals as a function of hpt. Control neuromasts were only analysed at 10 and 24 hpt. C, number of supporting cells per neuromast obtained in control (black) and copper-treated (red) zebrafish has a function of hpt. At 10 hpt, control vs. copper: P = 0.1668; At 24 hpt, control vs. copper: P = 0.9998, one-way ANOVA, Tukey's post hoc test. D, confocal images as described in panel A, for 5, 7 and 10 hpt neuromasts showing some additional examples of afferent terminals in the proximity of the supporting cells. Scale bar: 10 μm. the zebrafish progressively increases over the first 6-8 dpf, with the afferent fibres displaying a higher firing rate from around 6 dpf (Fig. 1). The treatment of 3 dpf zebrafish with 10 μM copper ablates all hair cells, which then regenerate to a number comparable to untreated control zebrafish by 48-72 hpt (5-6 dpf, Fig. 2). A comparable time course of recovery was seen in the ability of the neuromasts to respond to activity-dependent stimuli induced by the displacement of the mechanoelectrical transducer (MET) apparatus. Within each neuromast, the induced Ca 2+ signals in both the hair cells and postsynaptic afferent terminals were comparable between copper-treated and control zebrafish by ∼48 hpt (Fig. 5), a time when the induced APs in the PLLg neurons were also indistinguishable among the two experimental conditions (Fig. 4). Despite requiring 2-3 days to regenerate the full complement of hair cells within a neuromast, newly formed hair cells showed spontaneous glutamate release at their synapses and elicited action potential activity in the PLLg indistinguishable from those measured in untreated zebrafish by ∼24 hpt (Fig. 3). This rapid acquisition of synaptic function was also consistent with the time course of regeneration of their synaptic machinery, defined by the colocalization of the CtBP puncta with the  and supporting cells (cyan: anti-Sox2 antibody) within a neuromast from control zebrafish (top panels, no-copper treated and no-PLLg ablated zebrafish), control-ablated zebrafish (second row of panels), copper-treated zebrafish (third row of panels) and copper-treated zebrafish with ablated PLLg (bottom panels) at 0 hpt (B) and 72 hpt (C). Scale bar: 10 μm. Zebrafish (3 dpf: Tg(NeuroD:EGFP);Tg(Myosin6b:R-GECO)) were treated for 2 h with 30 μM copper sulphate. The last column represents the merged images showing afferent fibres, supporting cells and hair cells. D, percentage of neuromasts showing afferent protrusions/total neuromast investigated. Zebrafish subjected to the ablation of the PLLg are indicated as: Control (PLLg ablation) and Copper (PLLg ablation). E, number of hair cells per neuromast as a function of hpt. F, number of supporting cells per neuromast obtained under the same experimental conditions mentioned in panel E. For additional statistical analysis see the Statistical Summary Document in Supporting Information. In panels E and F, the number of neuromasts (zebrafish) tested is shown above the data points. afferent projections (Fig. 6). At older ages (12-18 dpf), when the hair cells (Olt et al. 2014) and afferent fibres firing characteristics become more mature (Fig. 1), we found that the functional regeneration of the neuromast required longer than in the >5 dpf zebrafish (Figs 7-9). We also found that in both early-and late-larval zebrafish the afferent terminals in the regenerating neuromast appeared to make an initial contact with supporting cells (Figs 8 and 11). Moreover, the removal of the afferent input to the neuromast substantially altered the normal regeneration of supporting cells and hair cells (Figs 12  and 13).

Spontaneous firing activity in the lateral line system
Spontaneous electrical activity in the hair cell postsynaptic afferent fibres has previously been shown to occur not only in lower vertebrates (e.g. zebrafish: Weeg & Bass, 2002;Trapani & Nicolson, 2010;bullfrog: Keen & Hudspeth, 2006) but also in the mammalian system (e.g. guinea-pig: Manley & Robertson, 1976;mice: Sontag et al. 2009). This activity is known to be dependent upon the spontaneous release of glutamate-containing vesicles from the hair cells (Obholzer et al. 2008;Seal et al. 2008) via the activation of Ca V 1.3 Ca 2+ currents (Platzer et al. 2000;Brandt et al. 2003;Trapani & Nicolson, 2010;Jeng et al. 2020). Ca V 1.3 Ca 2+ channels have a relatively hyperpolarised activation (about −70 mV: Zampini et al. 2010;, which is at or near to the resting membrane potential (V m ) of hair cells in the zebrafish lateral line (Olt et al. 2014). Because the MET channels in the hair cells of the lateral line are partially open in the absence of external stimulation (Erickson et al. 2017), as also shown in other lower vertebrates and in mammals (e.g. Crawford et al. 1989;Holt et al. 1997;Corns et al. 2014), the standing inward depolarising MET current directly contributes to the generation of the spontaneous action potential activity in the PLLg neurons (Trapani & Nicolson, 2011). Indeed, we found that the PLLg neurons, which carry the sensory input transduced by the hair cells in the neuromast, already exhibit spontaneous action potential activity at 2 dpf, which is when the resting MET current was first detected in these cells .
The biophysical characteristics of this spontaneous firing activity in the afferent fibres of the lateral line has extensively been investigated below ∼6 dpf (e.g. Trapani & Nicolson, 2011;Levi et al. 2015) and in the adult zebrafish (Weeg & Bass, 2002). One distinct property of this electrical activity is the large variability in its firing rate at all age-ranges investigated, which is likely to be due to several reasons, including: the size and age of the neurons , the number of neuromasts and hair cells of the same polarity that the neurons are contacting (Nagiel et al. 2008) and the possible differential modulation of the efferent system (Lunsford et al. 2019). Despite the relative similarity in the firing rate of the afferent neurons between larval and adult zebrafish, the presence of multiple patterns of APs appears to be only a characteristic of adult afferent neurons (Weeg & Bass, 2002). Here we demonstrated that a second neuronal population showing a higher firing rate and a more regular pattern is appearing from ∼6 dpf, becoming more prominent during the following days (Fig. 1), indicating that the lateral line is still maturing during the first week after fertilisation. This is also consistent with previous data showing that the afferent fibres of the lateral line undergo extensive growth and reorganisation  and that the biophysical properties of hair cells within a neuromast are still changing (Olt et al. 2014) during early-larval stages, which are essential for the fine-tuning of and increased sensitivity to movement of adult zebrafish.

Time course of functional regeneration in hair cells and afferent fibres
Hair cell regeneration in the zebrafish lateral line has primarily been investigated during the first few days post-hatching using morphological approaches. The application of ototoxic compounds, such as aminoglycosides, or noise-induced damage (Holmgren et al. 2020), has been shown to cause the rapid loss of hair cells in the early-larval neuromasts, which is followed by their regeneration from surrounding supporting cells within ∼48 hpt (e.g. Harris et al. 2003;Ma et al. 2008;Wibowo et al. 2011;Mackenzie & Raible, 2012). Our functional data from the early-larval neuromast further support these previous studies. The rapid recovery can also be visualized by the ability of larval zebrafish (< 8 dpf) to regain normal escape and rheotaxis responses following aminoglycoside-induced hair cell death (within 48 hpt: McHenry et al. 2009;Suli et al. 2012). We also showed that despite the need for ∼2 days to regain functional neuromasts following their damage, newly formed hair cells and their ribbon synapses are already functional by ∼24 hpt, since they are able to drive spontaneous action potential activity in the PLLg neurons with a frequency and pattern comparable between treated and untreated zebrafish. Despite the similar number of hair cells present in the fully functional neuromasts after copper treatment at 6-8 dpf (∼15 cells, Fig. 2E) and 12-18 dpf (∼16 cells, Fig. 7B), the older zebrafish require >120 hpt to functionally regenerate a neuromast, which is almost double that needed in early-larval zebrafish. The regeneration of hair cells from the lateral line of early-larval zebrafish has been shown to be mediated by supporting cells with distinct progenitor identities (Thomas & Raible, 2019). The possible change in the progenitor populations or mechanisms underpinning regeneration in older fish could explain their delayed regeneration capacity. However, a similar regenerative time window to that identified in late-larval zebrafish (> 120 hpt) is required to fully regenerate the much larger number of hair cells present in the neuromast of a 3-year old fish (∼148 hpt, Cruz et al. 2015). Therefore, the most likely explanation for the faster regeneration capability of the neuromasts in early-larval zebrafish is that during the first week post-fertilisation, hair cells are still undergoing major developmental changes ( Fig. 1; see also Olt et al. 2014), which occur in parallel to regeneration.
Our morphological data show that the afferent fibres innervating the regenerating early-and late-larval neuromasts were seen to form enlarged terminals away from the newly formed hair cells, but in close proximity to supporting cells. It is possible that these supporting cells are targeted during the transdifferentiation into hair cells. Supporting cells divide symmetrically to give rise to two daughter hair cells (López-Schier & Hudspeth, 2006;Mackenzie & Raible, 2012). Since the regenerative capacity of the neuromast is an ongoing process, (e.g. Pinto-Teixeira et al. 2015) supporting cells have to be continuously replaced. Indeed, supporting cells from the neuromast of larval zebrafish can regenerate after copper-induced damage very rapidly and faster than hair cells (< 24 hpt, Fig. 11). Although previous studies have shown that the initial development of the zebrafish lateral line neuromast also occurs in the absence of the PLLg nerve (Andermann et al. 2002;Grant et al. 2005;López-Schier & Hudspeth, 2005), our data indicate that the timely regeneration of the supporting cells and hair cells requires the presence of the afferent input. Since the afferent terminals seem to make initial contacts with supporting cells within a regenerating neuromast (Figs 6,8 and 11), it is possible that they are promoting the proliferation or transdifferentiation of supporting cells into hair cells. Once the hair cells have regenerated, the presence of the afferent innervation has also been shown to be required for the correct localization of the ribbons at their presynaptic site (Suli et al. 2016). Interestingly, we also found that untreated zebrafish with ablated PLLg neurons undergo a progressive loss of hair cells and supporting cells, indicating that the afferent input is not only required for the timely regeneration of the neuromasts, but also for their maintenance, as shown in the adult zebrafish (Wada et al. 2013). A comparable role for the afferent input has also been identified in the maintenance and regeneration of the olfactory system of the Xenopus laevis larvae (Yoshino & Tochinai, 2006) and adult zebrafish (Byrd, 2000;Villanueva & Byrd-Jacobs, 2009;Paskin et al. 2011).