Inexhaustible hair-cell regeneration in young and aged zebrafish

ABSTRACT Animals have evolved two general strategies to counter injury and maintain physiological function. The most prevalent is protection by isolating vital organs into body cavities. However, protection is not optimal for sensory systems because their external components need to be exposed to the environment to fulfill their receptive function. Thus, a common strategy to maintain sensory abilities against persistent environmental insult involves repair and regeneration. However, whether age or frequent injuries affect the regenerative capacity of sensory organs remains unknown. We have found that neuromasts of the zebrafish lateral line regenerate mechanosensory hair cells after recurrent severe injuries and in adulthood. Moreover, neuromasts can reverse transient imbalances of Notch signaling that result in defective organ proportions during repair. Our results reveal inextinguishable hair-cell regeneration in the lateral line, and suggest that the neuromast epithelium is formed by plastic territories that are maintained by continuous intercellular communication.


INTRODUCTION
Sensory receptors are the interphase between the environment and the nervous system. Vertebrates are generally able to repair sensory organs, but mammals cannot regenerate the hair cells of their inner ear (Burns and Corwin, 2013;Groves et al., 2013;Rubel et al., 2013). Consequently, ototoxic antibiotics, anti-neoplastic therapies or high levels of sound cause irreversible hearing loss and balance disorders (Chen and Fechter, 2003;Huth et al., 2011;Langer et al., 2013). Hair cells in fishes and amphibians occurring in the ear and the lateral line are also susceptible to damage by overstimulation or by drug-induced ototoxicity (Harris et al., 2003;Jiang et al., 2014;López-Schier and Hudspeth, 2006;Ma et al., 2008;Millimaki et al., 2010;Schuck and Smith, 2009;Steiner et al., 2014). However, these aquatic vertebrates can rapidly regenerate hair cells (Behra et al., 2009;Hernández et al., 2007;Williams and Holder, 2000). The superficial lateral line of zebrafish is composed by a collection of neuromasts that are formed by a simple circular epithelium of approximately 60 cells.
Some 16-20 mechanosensory hair cells occupy the center of the organ, whereas the remainder are two types of non-sensory supporting cells: around 30 sustentacular cells intermingle with hair cell, and fewer than 10 mantle cells outline the organ (Fig. 1A). Interneuromast cells connect each organ. Lateralis hair cells are born in pairs from the terminal mitotic division of unipotent hair-cell progenitors (UHCPs). The neuromast has a stereotypical symmetry formed by one axis of planar cell polarity and by the position of functionally distinct non-sensory equatorial and polar areas (Fig. 1A). Equatorial cells are under sustained Notch signaling, which restricts the development of UHCPs to permissive polar areas with low Notch (Wibowo et al., 2011). Stem cells have not been described in neuromasts, and whether the regenerative capacity of neuromasts diminishes with age or after recurrent damage remains unknown.

RESULTS AND DISCUSSION
The ET(krt4:EGFP)sqgw57A transgenic line highlights

Hair cells regenerate efficiently in larval, juvenile and adult zebrafish
A single treatment with the ototoxic aminoglycoside antibiotic neomycin readily ablates every functional hair cell in the superficial lateral line of the zebrafish larva (Harris et al., 2003;López-Schier and Hudspeth, 2006;Pinto-Teixeira et al., 2013). Subsequently, neuromasts enter a regenerative process that is largely complete 72 hours post (neomycin) treatment (hpt) (Ma et al., 2008;Wibowo et al., 2011). To assess hair-cell regeneration in older animals, we treated three different transgenic lines at juvenile (3-month old) and adult (1-and 2-year old) stages with neomycin. In all cases, hair-cell regeneration occurred within 72 hpt ( Fig. 2A-C, and data not shown). Using 1-year old adult fish in which the Tg[myo6b:actb1-EGFP] transgene reveals the apical hair bundle of the hair cells ( Fig. 2D-F), and 6-month old Tg[Alpl:mCherry ; ET(krt4:EGFP) sqet20] that shows neuromast geometry ( Fig. 2G-H), we found that cell polarity and epithelial architecture were comparable between controls and neomycin-treated samples 72 hpt. Thus, neuromasts are endowed with invariant and enduring regenerative capacity, which may have evolved for fish to maintain life-long sensory ability despite persistent environmental insult (Ciba-Foundation, 1991).
Hair-cell regeneration is unaffected by the recurrent and frequent loss of hair cells After treatment with neomycin, the first hair cells appear around 8 hpt, with a sequential production of pairs of hair cells up to 20 by 72 hpt (Ma et al., 2008;Wibowo et al., 2011;Lush and Piotrowski, 2014). We hypothesized that the fast onset of regeneration can be explained by the presence of a subpopulation of "primed" cells that are quickly routed towards a UHCP fate. Therefore, a frequent and long sequence of hair-cell ablations should deplete the epithelium from primed cells, leading to regenerative decline. To test this idea we subjected 2-year old transgenics Tg[ET(krt4:EGFP)sqet4] to six consecutive neomycin treatments with intervening 24-h periods of rest between treatments to allow partial regeneration. Hair cells regenerated efficiently after the sixth injury cycle ( Fig. 3A-B). To assess the temporal profile of regeneration, we subjected Tg[ET (krt4:EGFP)sqet4] larvae to six consecutive hair-cell ablations with neomycin and counted hair cells 24 h after each treatment. Neuromasts showed invariable hair-cell regeneration after each ablation ( Fig. 3C-D). Thus, to maintain this inexhaustible regenerative capacity at constant kinetics, each organ must persistently produce at least 4 UHCPs per day, representing over 10% of the originating Sox-2 + cells.

Atoh1a is transiently and broadly expressed in the epithelium
Hair cells require Atoh1a for their development. Ma and colleagues showed that Atoh1a is expressed in broad areas of the epithelium during the first 24 h of regeneration (Ma et al., 2008). Using the double-transgenic line Tg[Atoh1a:dTomato ; ET(krt4:EGFP)sqet4] to simultaneously visualize Atoh1a expression and hair cells, we extended these previous observations by finding a broad but weak Atoh1a in supporting cells, which stabilizes and becomes stronger in UHCPs and in young hair cells (Fig. 3E). This pattern of Atoh1a expression substantiates the prediction that neuromasts may use a large pool of supporting cells to generate UHCPs to sustain regeneration upon recurrent severe damage, and suggests that the expression of Atoh1a occurs by default in Sox-2 + cells. To further test this hypothesis we abrogated Notch signaling in the entire regenerating neuromast by treating zebrafish larvae with the γ-secretase inhibitor DAPT (Ma et al., 2008;Wibowo et al., 2011). The uniform loss of Notch signaling stabilized Atoh1a in a broader area of the neuromast (Fig. 3F), and generated supernumerary hair cells ectopically ( Fig. 3F-G, supplementary material Fig. S2A-B). The spatial profile of Sox-2 expression and of Notch activity in neuromasts suggests that they do not regulate each other (Hernández et al., 2007;Ma et al., 2008;Wibowo et al., 2011). Supporting this conclusion, Sox-2 expression was not qualitatively affected upon a constitutive activation of Notch signaling by expressing the intracellular domain of Notch (N ICD ) by neuromast- specific chemical induction with tamoxifen using the double transgenic line Tg[Cldnb:Gal4 ERT2 ;5xUAS-E1b:6xMYC-notch1a] (Fig. 4A-G). However, we observed a modestly significant increase of the number of Sox-2 + cells that was accompanied by a small but non-significant increase of the total number of cells (Fig. 4G).
During otic development, Sox-2 establishes pro-sensory competence in the neuroepithelium for the generation of neurons and hair cells by: (1) the direct activation of Atoh1 expression and, (2) the activation of negative regulators of Atoh1 function (Neves et al., 2013). In neuromasts, however, two parallel direct inputs appear to control the spatiotemporal mode of Atoh1a expression: a constitutive and un-patterned activation by Sox-2, and a patterned inhibition by Notch. In contrast to the ear, the neuromast epithelium does not produce neurons, which may account for the differences between the two organs. The profile of Atoh1a expression in normal circumstances and in conditions of high and low Notch signaling provides a mechanistic explanation for the sequence and spatiotemporal pattern of sustentacular-cell conversion into UHCP: the mitotic division of Sox-2 + cells is symmetric with regard to the UHCP potential of the two daughter cells. However, the independent movement of the daughter cells within the epithelium places them in different signaling environments: the cell remaining in an equatorial zone will continue to be exposed to high levels of Notch signaling and maintain a sustentacular-cell character, whereas those entering a polar zone with low Notch activity will stabilize Atoh1a expression and become UHCP. Although we do not know if the movement of each daughter cell is stochastic or deterministic, a default broad expression of Atoh1a provides a robust mechanism for UHCP production regardless of how cells move in the tissue. These results reinforce the conclusion that potentially all Sox-2 + cells may quickly generate UHCPs upon reduction of Notch activity, and may explain the inexhaustible production of hair cells upon recurrent severe damages.
Neuromast organization is reversibly affected by imbalances of Notch signaling It has been proposed that a feedback inhibition of supporting-cell proliferation via Notch signaling maintains a constant cellular population in neuromasts (Ma et al., 2008). Thus, to test a functional link between intercellular communication and organ proportions, we again examined neuromasts with defective Notch activity. We first expressed N ICD constitutively by heat-shock in Tg[hsp70l: Gal4; 5xUAS-E1b:6xMYC-notch1a], and by chemical induction with tamoxifen in the transgenic line Tg[Cldnb:Gal4 ERT2 ;5xUAS-E1b:6xMYC-notch1a], which halted hair-cell regeneration almost immediately (Fig. 4H, supplementary material Fig. S2C), indicating that neuromasts do not contain "engaged" UHCPs that are refractory to Notch inhibition. In a converse experiment, we abrogated Notch signaling by treating zebrafish with DAPT, which increased haircell production (Figs 3G, 4I, supplementary material Fig. S2A-B). Total cell counts varied only marginally in DAPT-treated neuromasts and matched the excess of hair cells (on average, circa 16% of the cells in control neuromasts were hair cells, versus 27% in DAPTtreated samples), revealing an imbalance of cell types and epithelial proportions (Fig. 4G,I-L). Control and DAPT-treated fish were then subject to a final incubation in neomycin to ablate hair cells, and subsequently returned to normal conditions. The neuromasts that had been exposed to DAPT continued to produce supernumerary hair cells one day after the second neomycin treatment (Fig. 4L). This result suggests that after lifting Notch inhibition, a larger than normal pool of sustentacular cells produces UHCPs, possibly by a lasting reorganization of the spatial influence of Notch. Remarkably, however, these neuromasts recovered almost normal proportions one week later (Fig. 3G). Taken together, these results strongly suggest that the neuromast epithelium has self-organizing properties, which may have evolved to allow fish to maintain organ proportions during growth and repair (Fig. 5).
Collectively, the results shown above suggest that compartmentalized Notch signaling determines the geometric order of hair-cell regeneration. Thus, it will be important to discover what controls the spatiotemporal activation of the Notch receptor. One possibility is that polarized activating signals emanate from the hair cells. This could also explain the seemingly contradictory observations that although the expression of a Notch receptor is upregulated in broad areas of the neuromast epithelium during the first 24 h after hair-cell ablation, it becomes co-expressed with Atoh1a and does not repress hair-cell production (Ma et al., 2008). Analyzing neuromasts chronically devoid of hair cells will clarify this issue.

CONCLUSIONS
During organ repair, the number and spatial distribution of different cells must be tightly coordinated. Similarly to the Drosophila ovary and the mammalian skin (Blanpain and Fuchs, 2014;Xie and Spradling, 2000), neuromast regions that allow progenitor cells to progress into terminal differentiation are spatially fixed, resulting in the preferential production of hair cells in permissive polar areas. However, a single change in intercellular signaling can route a large pool of sustentacular cells into acquiring a progenitor fate, resulting in the ectopic production of supernumerary hair cells. The resulting imbalance of tissue proportions is reversible, however, supporting the notion that distinct neuromast regions are territories, as opposed to compartments, because they contain plastic cellular identities that are dynamically maintained by continuous intercellular signaling. In that, the neuromast is more reminiscent to the Drosophila intestine, where Notch signaling generates plastic stem-cell niches that promote localized regeneration (Mathur et al., 2010). In sum, here we present evidence that epithelial self-organization underlies the non-deterministic acquisition of UHCP fate, which governs the inexhaustible hair-cell regeneration of the lateral line.

DNA constructs
The plasmid Cldnb::ERT2-GAL4 that was used to generate the stable transgenic line was placed in a miniTol2 backbone and was a gift of  D. Wilkinson (NIRM, UK). To generate Tg[Cldnb:Gal4 ERT2 ] line, 20 pg of the Tol2-expression clone and 20 pg of the transposase synthetic RNA were simultaneously injected into one-cell stage wild-type eggs. The resulting embryos were raised to adulthood and out-crossed for visual screening of germ-line transmission of the transgene.

Heat-shock induction of gene expression
Transgenic fish carrying the hsp70 promoter were incubated at the desired developmental stage at 39°C for 30 min in groups of 30-40 embryos in 2 ml Eppendorf tubes containing a total volume of 1 ml embryo medium (Danieau 30%). Upon incubation embryos were returned to petri dishes containing embryo medium fish medium at 28.5°C where they recovered until the required time point.

Pharmacology
Neomycin treatments were performed on 5-6 days post fertilization (dpf ) transgenic zebrafish larvae or 6-to 2-year old adult fish. Neomycin sulfate at 10 mg/ml stock solution in dH 2 O (Sigma, St. Louis, MO USA), was diluted in system water to a final concentration of 250 µM. Larvae were incubated with neomycin solution for 1 h at RT, rinsed and allowed to recover in system water at 28.5°C. Sibling fish in system water at RT for 1 h and then transferred at 28.5°C served as control. Notch signaling was inhibited using the γ-secretase inhibitor DAPT (N-[N-(3,5difluorophenacetyl)-L-alanyl]-S-phenylglycine-t-butyl ester) (Sigma). DAPT was reconstituted in 10% DMSO to a stock concentration of 150 mM and then diluted to a final concentration of 50 μM solution of DAPT in system water with 1% DMSO. Water with 1% DMSO served as control. For chemical induction of Gal4, 4-hydroxy-tamoxifen (H7904, Sigma) was dissolved at 12.5 mg/ml in 100% ethanol and stored at −20°C. Subsequent dilutions in embryo medium were made prior to use to a final concentration of 10 μM. Control larvae were incubated with an equivalent amount of ethanol diluted in embryo medium.

Vital imaging of zebrafish larva
For vital imaging, zebrafish larvae at 6 dpf were anesthetized with a 610 µM solution of the anaesthetic 3-aminobenzoic acid ethyl ester (MS-222). Mechanoreceptive hair cells were identified by labeling with the vital dye Di-2-ASP (Molecular Probes, Eugene, OR USA) or by EGFP expression in transgenic fish. Samples were mounted onto a glass-bottom 3 cm Petri dish (MatTek, Ashland, MA USA) and covered with 1% low-melting-point agarose with diluted anaesthetic. Images were acquired with an inverted confocal microscope with a 40× air or 63× water-immersion objective lenses.

Imaging adult caudal fins
For imaging caudal-fin neuromasts, adult zebrafish were anesthetized with a 610 µM solution of MS-222 and caudal fins were cut with a sharp scalpel. Fish were quickly returned to a tank with system water and allowed to fully recover before being transferred to a running-water system. Cut fins were fixed at 4°C in a solution of 4% paraformaldehyde (PFA) in phosphatebuffered saline (PBS) containing 0.2% Tween-20 (PBST). After fixation, the fins were washed in the same solution without fixative and transferred to a solution of Vectashield with DAPI (Vector Laboratories, Peterborough, UK) for at least one hour before imaging.

Immunohistochemistry
Immunohistochemistry was done as described previously (Pujol-Martí et al., 2012). In brief, samples were fixed overnight at 4°C in a solution of 4% paraformaldehyde (PFA) in phosphate-buffered saline (PBS) containing 0.2% Tween-20. After fixation, the samples were washed with PBS containing 1% Tween-20 and permeabilized by incubation in acetone at −20°C for 8 min. Samples were washed with PBS 1% tween-20 and blocked at room temperature with 10% bovine serum albumin. Primary antibody was incubated for 48 h at 4°C in PBS with 0.2% Tween-20 and secondary antibody was incubated overnight at 4°C in PBS with 0.2% Tween-20. Antibodies were used at the following dilutions: rabbit anti-Sox-2 (AbCam, Cambridge, UK) at 1:200. Texas Red-labeled donkey anti-rabbit immunoglobin secondary antibody (Molecular Probes, Life Technologies, Paisley, UK) and Alexa Fluor ® 488 goat anti-rabbit immunoglobin secondary antibodies (A5040, Sigma) at 1:500. Images were obtained using a confocal microscope (LSM 510; Carl Zeiss).

Quantification of hair cells
To quantify hair cells, Tg[ET(krt4:EGFP)sqet4] zebrafish larvae were used. Confocal stacks of neuromasts were acquired using an LSM 510 Zeiss microscope. Hair cells were manually identified by expression of cytoplasmic GFP. All data was processed and analyzed using a nonparametric test Mann-Whitney U-test (P<0.05, two-tailed) with GraphPad Prism version 6.04 for Windows (GraphPad Software, La Jolla, CA USA, www.graphpad.com).