1. Stem Cells and Regenerative Medicine
Download icon

Live cell-lineage tracing and machine learning reveal patterns of organ regeneration

  1. Oriol Viader-Llargués
  2. Valerio Lupperger
  3. Laura Pola-Morell
  4. Carsten Marr  Is a corresponding author
  5. Hernán López-Schier  Is a corresponding author
  1. Helmholtz Zentrum München, Germany
  2. Centre for Genomic Regulation, Spain
  3. Institute of Computational Biology, Helmholtz Zentrum München, Germany
Tools and Resources
  • Cited 3
  • Views 3,162
  • Annotations
Cite this article as: eLife 2018;7:e30823 doi: 10.7554/eLife.30823

Abstract

Despite the intrinsically stochastic nature of damage, sensory organs recapitulate normal architecture during repair to maintain function. Here we present a quantitative approach that combines live cell-lineage tracing and multifactorial classification by machine learning to reveal how cell identity and localization are coordinated during organ regeneration. We use the superficial neuromasts in larval zebrafish, which contain three cell classes organized in radial symmetry and a single planar-polarity axis. Visualization of cell-fate transitions at high temporal resolution shows that neuromasts regenerate isotropically to recover geometric order, proportions and polarity with exceptional accuracy. We identify mediolateral position within the growing tissue as the best predictor of cell-fate acquisition. We propose a self-regulatory mechanism that guides the regenerative process to identical outcome with minimal extrinsic information. The integrated approach that we have developed is simple and broadly applicable, and should help define predictive signatures of cellular behavior during the construction of complex tissues.

https://doi.org/10.7554/eLife.30823.001

Introduction

Understanding organogenesis, organ morphostasis and regeneration is crucial to many areas of biology and medicine, including controlled organ engineering for clinical applications (Lancaster et al., 2013; Boj et al., 2015; Sato and Clevers, 2015; Willyard, 2015). External tissues sustain continuous injury and must recurrently repair to maintain physiological function during the life of the organism (Levin, 2009). Structural reproducibility depends on the re-establishment of cell identity, number, localization and polarization. Two aspects of organ regeneration are the current focus of intense attention. First, how multiple cells interact to recapitulate organ architecture. Second, what is the mechanism that controls the correct reproduction of cell number and localization. Here we use the neuromasts of the superficial lateral line in larval zebrafish to gain a global perspective on sensory-organ regeneration. The neuromasts are ideally suited for this purpose because they are small and external, facilitating physical access and three-dimensional high-resolution videomicroscopy of every cell during extended periods. We have combined live single-cell tracking, cell-lineage tracing, pharmacological and microsurgical manipulations, and multidimensional data analysis by machine learning to identify features that predict cell-fate decisions during neuromast repair. Our comprehensive approach is simple and model independent, which should facilitate its application to other organs or experimental systems that are accessible to videomicroscopy. It should help reveal the basic rules that underlie how complex structures emerge from the collective behavior of cells.

Results

Complete neuromast ablation is irreversible in larval zebrafish

The neuromasts of the superficial lateral line in zebrafish are formed by a circular cuboidal epithelium of 60–70 cells (López-Schier and Hudspeth, 2006; Ghysen and Dambly-Chaudière, 2007; Norden, 2017). Mechanoreceptive hair cells occupy the center of the organ, whereas non-sensory sustentacular supporting cells are found around and between the hair cells (Figure 1A). A second class of supporting cell called mantle cells forms the outer rim of the organ. The invariant spatial distribution of these three cell classes generates a radial symmetry (Figure 1B) (Pinto-Teixeira et al., 2015). Neuromasts also have an axis of planar polarity defined by the orientation of the hair-cells’ apical hair bundle (Figure 1C) (Ghysen and Dambly-Chaudière, 2007; Wibowo et al., 2011). In addition to this geometric organization, cell-class number and proportions are largely constant, with around 40 sustentacular, 8–10 mantle, and 14–16 hair cells. Non-sensory cells can proliferate, whereas the sensory hair cells are postmitotic (López-Schier and Hudspeth, 2006; Ma et al., 2008; Cruz et al., 2015; Pinto-Teixeira et al., 2015). Finally, a string of interneuromast cells connects each neuromast along the entire lateral-line system (Figure 1A) (Ghysen and Dambly-Chaudière, 2007). Previous studies have extensively characterized the regeneration of the mechanosensory hair cells (Williams and Holder, 2000; Harris et al., 2003; López-Schier and Hudspeth, 2006; Hernández et al., 2006; Ma et al., 2008; Behra et al., 2009; Faucherre et al., 2009; Wibowo et al., 2011; Namdaran et al., 2012; Steiner et al., 2014; Jiang et al., 2014). However, the regeneration of non-sensory cells remains largely unexplored. To obtain quantitative data of whole sensory-organ regeneration we developed an experimental assay that combines controllable neuromast damage, long-term videomicroscopy at cellular resolution, and live cell-lineage tracing. We used combinations of transgenic lines expressing genetically encoded fluorescent proteins that allow the precise quantification and localization of each cell class in neuromasts, and which also serve as a direct and dynamic readout of tissue organization. This is important because it enables the visualization of cell-fate transitions in living specimens within the growing tissue at high temporal resolution. Specifically, the Tg[alpl:mCherry] line expresses cytosolic mCherry in the mantle and interneuromast cells (Figure 1D). The Et(krt4:EGFP)sqgw57A (hereafter SqGw57A) expresses cytosolic GFP in sustentacular cells (Figure 1E). The Tg[−8.0cldnb:LY-EGFP] (Cldnb:lynGFP) express a plasma-membrane targeted EGFP in the entire neuromast epithelium and in the interneuromast cells (Figure 1F), and the Tg[Sox2-2a-sfGFP] (Sox2:GFP) expresses cytosolic GFP in all the supporting cells and the interneuromast cells (Figure 1G). For hair cells, we use Et(krt4:EGFP)sqet4(SqEt4) that expresses cytosolic GFP (Figure 1H), or the Tg(myo6b:actb1-EGFP)(Myo6b:actin-GFP) that labels filamentous actin (Figure 1I). These transgenic lines have been previously published, but are reproduced here for clarity and self-containment of this work (López-Schier and Hudspeth, 2006; Kondrychyn et al., 2011; Kindt et al., 2012; Shin et al., 2014; Steiner et al., 2014; Pinto-Teixeira et al., 2015).

Geometric organization of the neuromast.

(A–C) Schematic representation of a neuromast depicting (A) cell classes identifiable by expression of transgenic markers. Grey arrows indicate, respectively, (B) radial symmetry and (C) epithelial planar polarity. (D–I) Confocal images of cell-specific transgenic markers. (D) Alpl:mCherry marks mantle and interneuromast cells, (E) SqGw57A shows all supporting cells, (F) Cldnb:lynGFP marks all neuromast cells, (G) Sox2-GFP marks supporting and interneuromast cells, (H) SqET4 labels hair cells, and (I) Myo6b:actin-GFP highlights the planar polarization of the hair cells by decorating their apical stereocilia. Scale bars: 10 µm. (J) Images of dorsal (top) and lateral (bottom) views of a SqGw57A transgenic zebrafish larva, revealing the full complement of superficial neuromasts and their stereotypic position. (K) A single confocal section of the lateral view of a neuromast expressing GFP in supporting cells (Sox2-GFP) and a RFP in all nuclei (H2B-RFP). (L) Same neuromast in K showing RFP-marked nuclei. The white arrow indicates 4 cells (circled), which are target of the laser beam for ablation. (M–P) Four still images of the neuromast in L over a period of five minutes, in which the laser-targeted cells are eliminated from the epithelium (white arrow).

https://doi.org/10.7554/eLife.30823.002

To induce tissue damage in a controllable and reproducible manner, we used a nanosecond ultraviolet laser beam that was delivered to individual cells through a high numerical-aperture objective, which was also used for imaging. The stereotypic localization of the neuromasts along the zebrafish larva varies only marginally between individuals and during larval growth (Figure 1J) (Ledent, 2002; López-Schier et al., 2004). This permits the unambiguous identification of the manipulated neuromast throughout the experiment, and the comparison between corresponding organs in different animals. Using Sox2:GFP 5 day-old zebrafish larvæ that ubiquitously express a nucleus-targeted red-fluorescent protein (H2B-RFP) (Figure 1K–L), we certified that laser-targeted cells are rapidly eliminated from the neuromast epithelium with no detectable collateral damage (Figure 1M–P and Video 1). Having established a well-controlled injury protocol, we decided to probe the limits of neuromast regeneration. We first used specimens co-expressing Alpl:mCherry and Cldnb:lynGFP, which reveal all neuromast cells in green and the mantle cells in red (Figure 2A). We began by ablating entire neuromasts and assessed regeneration for 7 days (Figure 2B–E). Specifically, we looked at the response of flanking interneuromast cells because it has been demonstrated that they can proliferate and generate additional neuromasts, particularly upon loss of ErbB2 signaling (López-Schier and Hudspeth, 2005; Grant et al., 2005; Sánchez et al., 2016). Four hours post-injury (4 hpi) a wound remains evident at the target area (Figure 2B). One day post-injury (1 dpi), the damaged area was occupied by a thread of Alpl:mCherry(+) cells, which based on marker expression are likely interneuromast cells (Figure 2C). None of the removed neuromasts regenerated after 7 days (n = 22) (Figure 2D–E). We obtained an identical outcome using the independent pan-supporting cell marker Sox2:GFP (n = 9) (Figure 2F–J). Finally, incubation of Alpl:mCherry specimens with Bromodeoxy-Uridine (BrdU) to reveal the DNA synthesis that occurs prior to mitosis showed that interneuromast cells do not proliferate after neuromast ablation (Figure 2K–N) (Gratzner, 1982). These data indicate that in contrast to what occurs in embryos (Sánchez et al., 2016), the complete elimination of a neuromast is irreversible in larval zebrafish.

Zebrafish larvæ do not regenerate completely-ablated neuromasts.

(A–E) Confocal images of a 7 day follow-up of the complete ablation of a neuromast in the double transgenic line Tg[Cldnb:lynGFP; Alpl:mCherry]. (A) The site of damage was identified over subsequent days by the position of an intact reference neuromast (white asterisk). (B) Laser-mediated cell ablation produced a wound 4 hours-post-injury (hpi). (C–E) This wound was replaced by a thread of mCherry(+) cells (white arrow) 1 day-post-injury (dpi), which did not change over the subsequent 6 days. (F–J) Confocal images over a 7 day time course after the ablation of a neuromast in the double transgenic line Tg[Sox2:GFP; Alpl:mCherry]. Identically to A-E, the complete ablation of the target neuromast results in a thin trail of interneuromast cells (white arrowheads) covering the damaged area (K–N). Scale bars: 10 µm.

https://doi.org/10.7554/eLife.30823.003
Video 1
A 20 min videomicroscopic recording of a neuromast after laser-mediated ablation of supporting cells.

Four laser-targeted cells (showing a dark spot in the nuclei from focal fluorescent-protein bleaching) are eliminated from the epithelium, which closes the wound. There is no noticeable collateral damage. Time resolution is one image per 30 s.

https://doi.org/10.7554/eLife.30823.004

Neuromasts have isotropic regenerative capacity

To further explore neuromast repair we decided to use milder injury regimes. We systematically produced controlled damage of well-defined scale and location in double transgenic specimens that combine the supporting cell marker Cldnb:lynGFP and the mantle-cell marker Alpl:mCherry (Figure 3A–O). We found that ablation of the posterior half of the neuromast was followed by closure of the wound within 24 hr (Figure 3A–C). At 3 dpi, target neuromasts regained normal cell-class spatial distribution (n = 6) (Figure 3D). At 7 dpi, neuromasts recovered approximately 70% of the normal cell number (Figure 3E,Z). We found no difference in speed and extent of regeneration after concurrently ablating the posterior half of neuromasts and flanking interneuromast cells (n = 5) (Figure 3F–J,Z). The ablation of the posterior or the dorsal half of the epithelium resulted in identical outcome, suggesting that neuromasts are symmetric in their regenerative capacity (n = 6) (Figure 3K–O,Z). Next, we assessed mantle-cell regeneration using a double transgenic line expressing Sox2:GFP and Alpl:mCherry, which reveal mantle cells in red and sustentacular cells in green (Figure 3P–Y). The complete elimination of mantle cells was followed by their re-emergence 3 dpi (Figure 3Q–S), and the reconstitution of the outer rim of the neuromast 7 dpi (n = 15) (Figure 3T,Z). The simultaneous ablation of the mantle cells and the adjacent interneuromast cells led to identical outcome (n = 6) (Figure 3U–Z). The ablation of the interneuromast cells in fish co-expressing Sox2:GFP and Alpl:mCherry on one side of a neuromast (n = 12), or between two adjacent organs (n = 8) did not trigger the proliferation of the remaining interneuromast cells over a period of 7 days (Figure 3—figure supplement 1A–J). Because the complete ablation of mantle cells leaves intact the sustentacular-cell population, and the hair cells are postmitotic, these results yield three important and novel findings: (1) interneuromast cells are not essential for neuromast regeneration in larval zebrafish, although they may contribute to mantle cell regeneration; (2) neuromasts have isotropic regenerative capacity; (3) sustentacular cells are tri-potent progenitors able to self-renew and to generate mantle and hair cells.

Figure 3 with 1 supplement see all
Neuromasts have isotropic regenerative capacity.

(A) Ablation of the posterior half of a neuromast. (B–C) The damage is resolved by cellular movement from the undamaged site 1dpi. (D) Neuromasts recover geometric order after 3 days and (J) return to homeostasis by 7dpi. Dashed lines in A,F,K,P,U delineate the ablated area. (F–J) Simultaneous ablation of the posterior half of a neuromast and the interneuromast cells flanking its anterior and posterior sides (n = 5) led to a regeneration outcome identical to that of the experiment in (A–E). Arrowheads in (F) point the location normally occupied by the interneuromast cells. (K–O) Neuromasts depleted from their dorsal half (n = 6) also recover epithelial size, proportions and geometry in a manner indistinguishable from equatorial-side ablation after 7 days. (P–T) 7 days after their complete laser-mediated ablation, mantle cells regenerated for neuromasts to recover the mantle. (U–Y) The ablation of interneuromast cells flanking both sides of neuromasts that were depleted of mantle cells resulted in the same outcome (n = 6). (Z) Quantification of the number of cells in regenerated neuromasts at 7 dpi. Number of neuromast cells was no statistically significant between groups of different damage regimes as determined by one-way ANOVA (F(4,27)=1.013, p=0.4183). Scatter plot shows mean ±s.e.m.ns: non-significant. Scale bars: 10 µm.

https://doi.org/10.7554/eLife.30823.005

Neuromast architecture recovers after severe loss of tissue integrity

To test the limits of neuromast regeneration we systematically ablated increasing numbers of cells. Extreme injuries that eliminated all except 1 to 3 cells almost always led to neuromast loss (not shown), whereas ablations that left between 4 and 10 cells, reducing the organ to a combination of 2–3 mantle and 2–7 sustentacular cells, allowed regeneration (Figure 4A–E,K). We found that after losing over 95% of their cellular content, neuromasts recover an average of 45 cells at 7 dpi (or approximately 70% of the normal cell count), with exceptional cases reaching 60 cells (equivalent to over 90% of a normal organ) (n = 15) (Figure 4K). Regenerating neuromasts became radial-symmetric as early as 3 dpi (Figure 4D), and had normal cell-class composition and proportions 7 dpi (Figure 4L–M). Next, we concurrently ablated 95% of the neuromast and the flanking interneuromast cells (Figure 4F–G). This intervention was followed by a similar regeneration process, but lead to smaller organs (n = 6) (Figure 4H–J,N–P). These observations reinforce our previous suggestion that interneuromast cells have a non-essential, yet appreciable contribution to regeneration. Timed quantification of cell-class number and localization showed a reproducible pattern of tissue growth and morphogenesis. During the first 24 hpi, the intact cells rebuilt a circular epithelium (Figure 4B). From 1 dpi to 3 dpi, cell number increases rapidly and proportion is restored (Figure 4C,K–M). After 3 dpi, cell number increases at a slower pace (Figure 4K–M). Importantly, each cell class assumes an appropriate position despite a much reduced cell number (Figure 4E,J,L–P).

Recovery of organ architecture after loss of tissue integrity.

(A–E) Confocal images of a neuromast regenerating from 4 to 10 cells during a period of 7 days. Neuromasts recover radial symmetry 3 dpi (D), and original organ proportions at 7 dpi (E). (F–J) Neuromasts reduced to 4–10 cells that were previously deprived from adjacent interneuromast cells (INCs) (arrowheads in F), regenerated and reformed radial symmetry (H–I) and proportions 7 dpi, despite maintaining a reduced size (J). Dashed circles in (A,F) illustrate damaged areas. Scale bars: 10 µm. (K,N) Total cell numbers in regenerating neuromasts over 7 days in the two conditions depicted in (A–J). (L,O) In the first 2 dpi neuromast consist almost exclusively of supporting cells (green and red). Hair cells (blue) begin to appear between at 2dpi. (M,P) Percentages of cell classes during a 7 day regeneration period. Right after damage, neuromast experience an imbalance of cell proportions that is re-established over the course of 3 days. Afterwards the neuromasts continues to slowly increase total cell number at similar rates. The final proportion of cell classes recapitulates that of the starting condition. Time points show mean ± s.e.m. [All except 4–10 cells] n = 15, [All except 4–10 cells + INC] n = 6. (Q) Top and (R) lateral views of a triple-transgenic Tg[Ncad: Ncad-EGFP; Alp:mCherry; H2A:H2A-EGFP] neuromast before injury. (S) Top and (T) lateral views of a regenerated neuromast 7 days post injury (n = 4). Basal location of nuclei and apical N-cadherin enrichment evidence the apicobasal polarization of the organ. The accumulation of N-cadherin (white arrowheads) in the regenerated neuromast shows that apical constrictions are properly re-established during the process. (U–V) Maximal intensity projection of a neuromast in the triple transgenic line Tg[Cldnb:lynGFP; SqGw57A; Alpl:mCherry] prior to injury that eliminates all except 4 to 10 cells (U), and the same neuromast 7 days after damage (V). (W) Hair-bundle staining with rhodamine-phalloidin (colored in pink) reveals the coherent planar polarization of the hair cells in the regenerated neuromast shown in (V). (X) Confocal projection of a neuromast before the removal of flanking interneuromast cells. (Y) Maximal projection of a neuromast 48 hr after interneuromast-cell ablation and 24 hr after neomycin treatment. (Z) Phalloidin staining of hair bundles of hair cells regenerated in the absence of interneuromast cells, showing recovery of coherent epithelial planar polarity. Scale bars: 10 µm.

https://doi.org/10.7554/eLife.30823.007

Next, we examined if the orthogonal polarity axes of the epithelium are re-established after the severest of injuries. To assess tissue apicobasal polarity we used a combination of transgenic lines that allows the observation of the invariant basal position of the nucleus and the apical adherens junctions (Figure 4Q–R) (Ernst et al., 2012; Harding and Nechiporuk, 2012; Hava et al., 2009). We found correct positioning of these markers in the regenerated epithelium (n = 4), including the typical apicobasal constriction of the hair cells (Figure 4S–T). To assess epithelial planar polarity, we looked at hair-bundle orientation using fluorescent phalloidin, which revealed that 7 dpi the regenerated neuromasts were plane-polarized in a manner indistinguishable from unperturbed organs, with half of the hair cells coherently oriented in opposition to the other half (n = 10) (Figure 4U–W). To test if plane-polarizing cues derive from an isotropic forces exerted by the interneuromast cells that are always aligned to the axis of planar polarity of the neuromast epithelium, we ablated these cells flanking an identified neuromast, and concurrently killed the hair cells with the antibiotic neomycin (Figure 4X–Y). In the absence of interneuromast cells regenerating hair cells recovered normal coherent planar polarity (n = 16), suggesting the existence of alternative sources of polarizing cues (Figure 4Z). Collectively, these findings reveal that as few as four supporting cells can initiate and sustain integral organ regeneration.

Sustentacular and mantle cells have different regenerative potential

Injury in the wild is intrinsically stochastic. Thus, we hypothesized that the regenerative response must vary according to damage severity and location, but progress in a predictable manner. To test this assumption and unveil the underlying cellular mechanism, we systematically quantified the behavior of individual cells by high-resolution videomicroscopy. We conducted 15 independent three-dimensional time-lapse recordings of the regenerative process using a triple-transgenic line co-expressing Cldnb:lynGFP, SqGw57A and Alpl:mCherry (Figure 5A–B), ranging from 65 to 100 hr of continuous imaging (each time point 15 min apart). Starting immediately after the ablation of all except 4–10 cells, we tracked every intact original cell (called founder cell) and their progeny (cellular clones) (Figure 5A and Video 2). We followed a total 106 founder cells (76 sustentacular cells and 30 mantle cells). We tracked individual cells manually in space and time, recording divisions and identity until the end of the recording, resulting in 763 tracks and 104,863 spatiotemporal cell coordinates (Figure 5A–B). Each clone was represented as a tree to visualize the contribution of each founder cell to the resulting clones (Figure 5C). We found that the majority of the founder sustentacular cells underwent three divisions and that some divided up to five times (Figure 5D). 14 out of 30 founder mantle cells did not divide at all, and the rest divided once or, rarely, twice. Founder sustentacular cells required on average 19 ± 6 hr (mean ±s.d., n = 76) to divide, whereas the founder mantle cells that divided required on average 27 ± 5 hr, (mean ±s.d., n = 30) (Figure 5E). Clones from founder sustentacular and founder mantle cells were markedly different: founder sustentacular cells produced all three cell classes (sustentacular, mantle and hair cells), whereas founder mantle cells produced clones containing only mantle cells (Figure 5F). We categorized all cell divisions according to the fate of the two daughter cells at the time of the following division, or at the end of the time-lapse recording (Figure 5G). This analysis revealed that 97% of the sustentacular-cell divisions were symmetric: 78% produced two sustentacular cells (SS), 16% produced a pair of hair cells (HH), and 3% generated two mantle cells (MM). Only 3% of the divisions were asymmetric, generating one sustentacular and one mantle cell (SM) (n = 307). All mantle-cell divisions were symmetric (MM) (n = 20). These observations further support the conclusion that sustentacular cells are tri-potent progenitors.

Long-term whole-organ single-cell tracking reveals cell-clone formation during neuromast regeneration.

(A) Still images showing a representative 100 hr time-lapse recording of a regenerating neuromast in Tg[Clndb:lynGFP; SqGw57A; Alpl:mCherry] larva (left and middle panels). Cellular clones that share a common founder cell are clustered and color-coded. Cell trajectories reveal a concentric growth pattern (right panel). (B) Cell trackings at the last recorded timepoints for 10 out of the total of 15 regenerated neuromasts. (C) Cell-lineage tracing from time-lapse movie shown in (A). Branching points symbolize cell divisions. The division of a founder cell generates two cells of the 1 st generation. Subsequent divisions produce cells of the 2nd, 3rd and 4th generation. Cell classes are indicated with green (sustentacular), blue (hair) and red (mantle) colors. (D) Sustentacular founder cells undergo significantly more (p=3.59e-06, Mann-Whitney test) division rounds than mantle founder cells during 100 hr of neuromast regeneration. (E) The first division of sustentacular founder cells (n = 76) occurs significantly earlier (p=1.13e-5, Mann-Whitney test) than that of mantle founder cells (n = 16). (F) Sustentacular founder cells (n = 76) generate all three neuromast cell classes whereas mantle founder cells (n = 30) produce only mantle cells. (G) Out of 307 sustentacular cell divisions, 78% were self-renewing, 16% produced a pair of hair cells, 3% produced sustentacular cells that both became mantle cells within the next generation and 3% generated two sustentacular cells of which only one transited to mantle cell fate within the next generation. All 20 observed mantle cell divisions were self-renewing.

https://doi.org/10.7554/eLife.30823.008
Video 2
100 hr time-lapse recording of a regenerating neuromast after severe ablation.

A neuromast regenerates its original architecture from as few as six founder cells. Founder cells are identified by 1–6 (n) and their daughter cells receive 2 n and 2n + 1 identities. Recording starts 4 hr post injury (hpi) and shows single focal planes. Time is in hours post injury.

https://doi.org/10.7554/eLife.30823.009

Previous studies have firmly established that hair-cell regeneration is strongly anisotropic because hair-cell progenitors develop almost exclusively in the polar areas of horizontal neuromasts, elongating the macula in the dorsoventral direction (Wibowo et al., 2011; Romero-Carvajal et al., 2015). Although our static images suggest that neuromasts have isotropic regenerative capacity, we nevertheless wondered whether regeneration of non-sensory cells is directional. To this end, we fractioned the epithelium of horizontal neuromasts in four quarters of equal dimension (dorsal, ventral, anterior and posterior) (Figure 6A–B), which reflects the known functional territorialization of the neuromast epithelium based on the expression of transgenic markers and Notch signaling (Ma et al., 2008; Wibowo et al., 2011). We first assessed the spatial distribution of cell divisions during the first 60 hr of regeneration and found no pattern that would suggest regeneration anisotropy (Figure 6A). However, 60 hpi, most divisions (74%) took place in the dorsal and ventral (polar) quarters (Figure 6B). This is expected because later divisions mainly produce hair cells from polar progenitors (Figure 4L,M). Thus, the regenerating epithelium is initially homogeneous and becomes territorialized 60 hpi. We reasoned that epithelial territorialization could occur either by the migration of similar cells that are scattered throughout the tissue, or by position-adaptive differentiation of an initially equivalent population of cells. To test these possibilities, we generated a virtual Cartesian coordinate system at the center of the neuromast to fit all founder cells at the beginning of regeneration (4hpi). Next, we analyzed the localization of their progeny 60 hpi (Figure 6C–H). We found that 60% of the progeny of anterior-localized founder cells were located in the anterior side of the resulting epithelium, whereas 64% of the progeny of posterior-located founder cells were found in the posterior side (Figure 6C–E). We also found that 72% of cells derived from dorsal founder cells and 74% of cells from ventral founder cells were located on the same side of the virtual dorsal/ventral midline (Figure 6F–H). Therefore, most of the clones remain ipsilateral to the founder cell. These results indicate that neuromasts have isotropic regenerative capacity and their territorialization occurs by location-adaptive cellular differentiation.

Neuromast regeneration is not stereotypic and reveals different clone type compositions.

(A) Proliferation is markedly isotropic during the first 60 hr of neuromast regeneration (n = 348). (B) Homeostatic, dorso-ventral (DV) proliferative territories are restored after 60hpi (n = 27). (C) 40% and 36% of the progeny from anterior (n = 202) and posterior (n = 173) founder cells crossed to the contralateral side (light grey) after 60 hr of regeneration. (D) Only 28% and 26% of the progeny from dorsal (n = 199) and ventral (n = 176) founder cells crossed to the contralateral side (light grey) during the same period of time. (E) Representative examples of different clone types extracted from time-lapse data. Sustentacular cells give rise to S, SM, SH, and SHM clones (color coded respectively with green, pink, cyan and orange) whereas mantle cells produce only pure mantle cell clones. (F) The clone composition of the 15 regenerated neuromasts is not stereotypic. The length of each bar represents the proportion of neuromast cells that belong to each clone. Neuromast eight has been shown in Figure 5A,B. (G) The most frequent clones contain sustentacular and hair cells (SH, n = 37 clones), followed by those with only sustentacular cells (S, n = 21 clones). The third most frequent are composed by sustentacular and mantle cells (SM, n = 12 clones). Clones containing all three cell classes were rare (SHM, n = 2 clones).

https://doi.org/10.7554/eLife.30823.010

The sustentacular-cell population is tri-potent and plastic

To answer the long-standing question of whether the sustentacular-cell population is homogeneous and approach the problem of what determines symmetric versus asymmetric modes of division, we characterized the composition of all 72 clones from founder sustentacular cells. We found four types of clones: containing only sustentacular cells (S), sustentacular and mantle cells (SM), sustentacular and hair cells (SH), and all three cell classes (SHM) (Figure 6I). Of note, founder mantle cells produced clones containing only mantle cells (M) (Figures 5G and 6I). We observed that 37/72 of the clones from founder sustentacular cells were SH, 21/72 were S, 12/72 were SM, and 2/72 were SHM (Figure 6I–K). The proportion of each clone type suggests that either the sustentacular-cell population is heterogeneous, or that it is homogeneous but plastic. In searching for potential sources of clone heterogeneity, we noted that in some developmental contexts cell-cycle length or proliferative potential can determine the fate of the daughter cells (Calegari et al., 2005; Rossi et al., 2017). Therefore, we quantified the kinetics of proliferation of founder sustentacular cells and of their daughters and compared them to clone composition. We found three clear waves of cell divisions, each spaced by 8–10 hr (Figure 7A), respectively peaking at 20 hr, 28 hr and 38 hr (Figure 7B–C), suggesting that cell-cycle length is strictly regulated. Cell-cycle length in the 1 st generation peaks around 10 hr (9.8 ± 3.3 hr, median ±interquartile range (iqr)) (Figure 7C), but it begins to increase and to vary in the 2nd generation (11.5 ± 7.3 hr, median ±iqr), and more so in the 3rd generation (18.8 ± 20.3 hr, median ±iqr). To identify transition points in cycle lengths, we tested the goodness of fit of a two-segment regression model with variable change points. We found that the length of cell cycles is initially around 11 ± 3 hr (mean ±s.d.) and slowly increases up to 47 hpi. Afterwards, cell-cycle length increases more rapidly and is more variable (Figure 7D). To test if cell number influences cell-cycle length we used a similar two-segment regression model to define when cell-cycle length loosens, and discovered that the vast majority of the cell cycles (76%) span 7–13 hr below a threshold of 24 cells (Figure 7E). Above this threshold, cell-cycles lengths show large variation. With these data, we plotted proliferation kinetics against clone type, and found no significant difference between clones (Figure 7F–G). Thus, the length of the cell cycle or the proliferative potential of founder sustentacular cells cannot explain clone composition.

Quantification of cell divisions during neuromast regeneration.

(A–B) Equally spaced waves of coordinated sustentacular cell divisions (green) underlie the recovery of neuromast cell size. Mantle cell divisions (red) occur occasionally and do not follow the pattern of sustentacular cells. Proliferative waves correspond to the coordinated divisions of cells from independent generations. (C) Cells from the 1 st and 2nd generation divide on average after cell cycles of 11 ± 5 and 14 ± 9 hr respectively (mean ±s.d.). Coordination is lost at the 3rd generation when cell cycles start to lengthen (26 ± 18 hr, mean ±s.d.). (D) Cell cycle length (11 ± 3 hr, mean ±s.d.) is marginally influenced by regeneration time until 47 hr after injury, when cycle length starts increasing proportionally with regeneration time. (E) Cell cycle lengths (12 ± 6 hr, mean ±s.d.) do not correlate directly with neuromast size until 24 neuromast cells. (F) S, SM and SH clones produce similar number of cells (p=0.68, Kruskal Wallis test). In the box plots, the boundary of the box indicates the 25th and 75th percentile, respectively the black line within the box marks the median. Whiskers above and below the box include points that are not outliers. (G) Sustentacular founder cells of S, SM, and SH clones divide similarly early (p=0.42, Kruskal Wallis test) after approximately 18 hr after neuromast injury. (H) Sustentacular founder cells that produce SH (cyan) and S clones (green) are distributed similarly around the center of the organ (at x = y = 0). Those that generate SM clones (pink) are localized further away from the center and are biased towards the posterior side.

https://doi.org/10.7554/eLife.30823.011

Machine learning identifies predictive features for cell-fate acquisition

Multiple extrinsic factors that vary in space and time could determine cell-fate choices. Because manual analysis of such multidimensional data might be biased or neglect certain factors we implemented a quantitative and unbiased computational approach based on machine learning to identify variables (features) that correlate with clone composition. The first step of the workflow is the extraction of spatiotemporal coordinates and cell-lineage information from the manual tracks of the videomicroscopic data sets (n = 15) (Figure 8A). For each cell-track coordinate, we extracted 32 quantifiable features (Table 1), which were used to train the machine-learning algorithm. In a pre-analysis, we compared the performance of 20 algorithms (support vector machines, decision trees and nearest neighbor classifiers) in terms of accuracy and area under the curve (AUC) and chose the ensemble bagged tree random forest algorithm (Breiman, 2001) as the best performing method (Figure 8—figure supplement 1). To avoid overfitting, we trained the random forest using 14 samples to predict clone composition in the remaining sample in a round robin fashion. We evaluated the quality of predictions using Matthews correlation coefficient (MCC) to compensate for imbalances of clone frequencies (Figure 6K)

Table 1
List of prediction features with description.

We used 32 mainly spatial and neighborhood specific features for the classification. Features are explained in the description column.

https://doi.org/10.7554/eLife.30823.012
Feature nameDescription
Absolut timeHours post induction (hpi)
Absolute distance to centerEuclidean distance to the neuromast center
Average distance to H cell-
Average distance to M cell-
Average distance to S cell-
Cell generationNumber of divisions that the cell has undergone
Founder Cell Type-
Minimum distance to H cell-
Minimum distance to M cell--
Minimum distance to S cell-
Movement angle to last divisionAngle between current cell location, neuromast center and location of last cell division (or start of the movie in case of founder cell division)
Movement direction compared to centerRadial distance between current cell location and location of last cell division (or start of the movie in case of founder cell division). If the current location is nearer to the center the value is (+) in case it is further away the value is (-)
Movement distance since last divisionEuclidean distance between current cell location and last cell division (or start of the movie in case of founder cell division)
Normalized distance to centerRadial distance of current cell location divided by the radial distance of the current furthest cell (to approximate the neuromast size)
Number of founder cells-
Number of H cells-
Number of H cells in 10 µm radius-
Number of H cells in 20 µm radius-
Number of H cells in 30 µm radius-
Number of M cells-
Number of M cells in 10 µm radius-
Number of M cells in 20 µm radius-
Number of M cells in 30 µm radius-
Number of S cells-
Number of S cells in 10 µm radius-
Number of S cells in 20 µm radius-
Number of S cells in 30 µm radius-
Number of total cells-
Polar anglePolar angle is the counterclockwise angle between the x-axis, the neuromast center and the current cell location
Time to last divisionTime to last division (or start of the movie in case of founder cell division)
X coordinate-
Y coordinate-
Figure 8 with 3 supplements see all
Implementation of predictive machine-learning analysis.

(A) Overview from experiments to prediction. Movies of neuromast regeneration allow us to track every single cell over 100hpi and to generate a cell lineage from these track points. Information covered in all tracks and lineages can be extracted as features with which we train our random forest machine-learning classifier to predict division or cell lineage fate. (B) Sustentacular founder cell choices between SH vs. SM clones can be predicted with high accuracy (MCC = 0.63 ± 0.09, mean ± s.d., n = 15 bootstrapped samples) whilst choices between S and SH or SM clones are highly inaccurate (MCC = 0.19 ± 0.11 and 0.15 ± 0.10, mean ± s.d., respectively, n = 15 bootstrapped samples), based on 32 calculated features. (C) Features relative to the position of the founder cells and their nearest cellular environment can discriminate between SM and SH clone types. (D) Choices between SM/MM and HH divisions can be predicted with high accuracy (MCC = 0.91 ± 0.07, mean ± s.d., n = 15 bootstrapped samples) while those between SS and HH or SM/MM have low accuracy (MCC = 0.50 ± 0.05 and 0.38 ± 0.15, respectively, mean ± s.d., n = 15 bootstrapped samples) (E) Features describing the cell’s position in relation to the neuromast center and their proximity to other mantle cells have the highest influence on the cell fate choices of a sustentacular cell. (F) SM/MM divisions (red) appear predominantly at the periphery of the organ whereas HH divisions (blue) appear proximal to the center. Sustentacular cell self-renewing divisions (SS, green) occur mostly around the neuromast center, generating a ring-like pattern.

https://doi.org/10.7554/eLife.30823.013

Using machine learning, we were able to predict the occurrence of SH vs. SM clones from a founder sustentacular cell with high accuracy (42 out of 49 correctly predicted clones, MCC = 0.63 ± 0.09, mean ± s.d., n = 15 bootstrapped samples), while neither SH nor SM clones could be discriminated when compared to S clones (Figure 8B). Of the 32 features that we used, those that best discriminated SH vs SM clones were the sustentacular cells’ distance to the center of the epithelium, and the distance to the mantle cells (Figure 8C and Figure 8—figure supplement 2). Next, we focused on the decision-making process of individual sustentacular cells at the time of their division. We trained a random forest to discriminate between SS, HH and SM/MM divisions in a pairwise fashion. The HH and SM/MM divisions were highly predictable (63 out of 66 divisions correctly predicted, MCC = 0.91 ± 0.07, mean ± s.d., n = 15 bootstrapped samples), while the discrimination between SS and HH or SM/MM divisions was much less accurate (Figure 8D). Again, the most informative features were the distance to the neuromast center and the distance to the mantle cells (Figure 8E, Figure 8—figure supplement 3). SM/MM divisions occur consistently at the outer perimeter of the neuromast (Figure 8F), whereas HH divisions take place near the center. Self-renewing SS divisions occupy the area between HH and SM/MM divisions. Interestingly, SM/MM divisions were never seen in the anterior-most region of the organ, suggesting that progenitor sustentacular cells are routed into generating mantle cells specifically in the perimetral areas that lack mantle cells but not elsewhere. Therefore, regenerating neuromasts appear to sense cell-class composition and route cellular differentiation in a spatially regulated manner to regain cell-class proportion and distribution.

Discussion

One long-standing goal of biological research is to understand the regeneration of tissues that are exposed to persistent environmental abrasion. Here we address this problem by developing a quantitative approach based on videomicroscopic cell tracking, cell-lineage tracing, and machine learning to identify features that predict cell-fate choices during organ regeneration. Using the superficial neuromasts in zebrafish, we demonstrate that a remarkably small group of resident cells suffices to rebuild a functional organ following severe disruption of tissue integrity. Our findings reveal that the sustentacular-cell population is tri-potent, and suggest that integral organ recovery emerges from multicellular organization employing minimal extrinsic information. Below, we discuss the evidence that supports these conclusions.

By systematically analyzing cellular behavior, we reveal a hierarchical regenerative process that begins immediately after injury. First, surviving founder cells reconstitute an epithelium. Second, sustentacular cells become proliferative and restore organ size. Cellular intercalation is rare. Third, daughter cells differentiate in a position-appropriate manner to recreate cell-class proportions and organ geometric order. Fourth, the epithelium returns to a homeostatic state that is characterized by low mitotic rate. The milder damage regimes that eliminated one half of the epithelium show that neuromasts are symmetric in their regenerative capacity, and that they preferentially regenerate the cells that have been eliminated. Importantly, these findings, which rely on the quantitative spatiotemporal analysis of regeneration data, could not have been predicted from previous studies using static and largely qualitative information (Williams and Holder, 2000; López-Schier and Hudspeth, 2005; Dufourcq et al., 2006; López-Schier and Hudspeth, 2006; Ma et al., 2008; Wibowo et al., 2011; Wada et al., 2013; Steiner et al., 2014; Romero-Carvajal et al., 2015; Cruz et al., 2015; Pinto-Teixeira et al., 2015). An important corollary of these results is that neuromasts do not contain specialized cells that contribute dominantly to repair. We propose that progenitor behavior is a facultative status that every sustentacular cell can acquire or abandon during regeneration. We did not observe regenerative overshoot of any cell class (Agarwala et al., 2015), suggesting the existence of a mechanism that senses the total number of cells and the cell-class balance during tissue repair (Simon et al., 2009). Together with previous work, our results support the possibility that such mechanism is based on the interplay between Fgf, Notch and Wnt signaling (Ma et al., 2008; Wibowo et al., 2011; Wada et al., 2013; Romero-Carvajal et al., 2015; Dalle Nogare and Chitnis, 2017). Our combination of machine learning and quantitative videomicroscopy shows clear differences between sustentacular and mantle cells, but does not indicate heterogeneity within the sustentacular-cell population. However, further application of this integrated approach and new transgenic markers may reveal uncharacterized cells in the neuromast. This may be expected given recent work that showed the existence of a new cell class in neuromasts of medaka fish (Seleit et al., 2017). It is technically challenging to consistently maintain fewer than 4 cells in toto without eliminating the entire neuromast. Thus, we cannot rule out the possibility that a single founder cell may be able to regenerate a neuromast. We show that the complete elimination of a neuromast is irreversible in larval zebrafish. However, Sánchez and colleagues have previously reported that interneuromast cells can generate new neuromasts (Sánchez et al., 2016). By assaying DNA synthesis prior to mitosis, we show that interneuromast cells do not proliferate after neuromast ablation. These differences may be explained by differences in ablation protocols (electroablation versus laser-mediated cell killing), the age of the specimens (embryos versus early larva) or the markers used to assess cellular elimination.

We find that interneuromast cells are not essential for neuromast regeneration because severely damaged organs recover all cell classes in the appropriate localization in the absence of interneuromast cells. However, we systematically observed smaller organs when interneuromast cells where ablated. These observations suggest that these peripheral cells may yet help regeneration, either directly by contributing progeny, or by producing mitogenic signals to neuromast-resident cells.

The behavior of the mantle cells is especially intriguing. Complete elimination of parts of the lateral line by tail-fin amputation have revealed that mantle cells are able to proliferate and generate a new primordium that migrates into the regenerated fin to produce new neuromasts (Dufourcq et al., 2006). This observation can be interpreted as suggesting that under some injury conditions, mantle cells are capable of producing all the cell classes of a neuromast. Transcriptomic profiling of mantle cells following neuromast injury revealed that these cells up-regulate the expression of multiple genes (Steiner et al., 2014). Furthermore, a recent study has revealed that mantle cells constitute a quiescent pool of cells that re-enters cell cycle only in response to severe depletion of sustentacular cells (Romero-Carvajal et al., 2015), suggesting that these cells may conform a stem-cell niche for proliferation of sustentacular cells. Thus, the collective evidence indicates that the mantle cells respond to damage and contribute to the regenerative processes, and may drive the regeneration of an entire organ if every other cell class is lost.

One outstanding question is how regeneration is controlled spatially. The epithelium may respond to damage via dynamic formation of an injured-intact axis at the onset of repair. Our results support this scenario by unveiling the adaptability of the neuromast epithelium to the localization and scale of damage. We suggest a model in which the invariant radial symmetry of the neuromast serves as a rheostat to identify the site of damage to guide regeneration spatially (Figure 9). A polarized axis along structurally intact and injured areas underlies this process. However, the complete reconstruction of a neuromast by as few as 4 cells suggests that a partial maintenance of radial symmetry is not essential for organ regeneration. Therefore, radial-symmetry maintenance cannot have a deterministic impact on the recovery of geometric order. Yet, partial structural maintenance and polarized tissue responses may optimize repair, respectively, by preventing superfluous cellular production in undamaged areas and by biasing the production of lost cells in the damaged areas. For organs that have evolved under the pressure of persistent damage, compliance to the extent of the injury may be an advantage because the regenerative responses can be scalable and localized, allowing faster and more economical regeneration.

Schematic model of neuromast regeneration.

The top diagram exemplifies the architecture of an intact neuromast. A, B and C indicate three types of injury: A when mantle cells are lost, B when hair cells are ablated, and C when a localized combination of all three cell classes is lost. Under the model that we present, radial symmetry serves to localize damage and canalize regeneration spatially. If central hair cells are lost (A), radial symmetry is maintained for sustentacular progenitors to regenerate hair cells centripetally (grey arrows in A). If outer cells are lost (B), radial symmetry is also maintained for the generation of progeny that will acquire mantle fate and propagate centrifugally to reform the outer rim of the neuromast (grey arrows in B). Upon asymmetric damage, however, the radial symmetry is partially broken (C). The neuroepithelium repolarizes along an injured-intact axis, which canalizes regeneration towards the damaged areas (grey arrows in C). Individual cells are color-coded (mantle cells in red, sustentacular cells in light blue, and hair cells in green), and in each case we indicate the type of division that the intact cells undergo: symmetric (S) when they produce two equivalent cells or self-renew, and asymmetric (A) when their division generates sibling cells that differentiate into different classes.

https://doi.org/10.7554/eLife.30823.017

After the severest of injuries, regenerated neuromasts were plane polarized in a manner indistinguishable from unperturbed organs. This startling result indicates that as few as four founder supporting cells can re-organize the local coherent planar polarity of the epithelium during neuromast repair. An alternative explanation is that founder cells have access to external polarizing cues. One source of this information is an isotropic mechanical forces exerted by the interneuromast cells that flank a neuromast. This is possible because interneuromast cells are always aligned to the neuromast’s axis of planar polarity. Yet, the concurrent ablation of resident hair cells and the interneuromast cells around an identified neuromast led to regenerated hair cells whose local orientation was coherent. Interestingly, recent studies have identified a transcription factor called Emx2 that regulates the orientation of hair cells in neuromasts of the zebrafish (Jiang et al., 2017). Emx2 is expressed in one half of the hair cells of the neuromast (those oriented towards the tail) and absent in the other half (which are coherently oriented towards the head). Loss- and gain-of-function of Emx2 alter planar cell polarity in a predictable manner: loss of Emx2 leads to neuromasts with every hair cells pointing towards the head of the animal, and Emx2 broad expression orients hair cells towards its tail. Because the coherent local axis of polarity is not affected by these genetic perturbations, Emx2 may act in hair cells as a decoder of global polarity cues. This evidence, together with our results, suggests that during neuromast regeneration founder cells autonomously organize the variegated expression of Emx2 in the regrowing epithelium with consequent recovery of a coherent axis of planar polarity and with one half of the hair cells pointing opposite to the other half. The future development of live markers of Emx2 expression will be able to test this prediction. We would like to highlight that we do not currently understand the global polarization of the neuromast epithelium relative to the main body axes of the animal. External sources of polarity may impinge in the recovery of these global axes during neuromast regeneration. Previous work has demonstrated that local and global polarization occur independently of innervation (López-Schier and Hudspeth, 2006), but other potential polarizing cues remain untested. Therefore, at present we can only support the notion that local coherent polarity is self-organizing, whereas global orientation may be controlled externally.

Our results beg the question of whether neuromast cells self-organize. Our operational definition of self-organization is an ‘autonomous increase in order by the sole interaction of the elements of the system’ (Haken, 1983), implying that a cellular collective organizes a complex structure without the influence of external morphogenetic landmarks, patterning cues, or pre-existent differential gene-expression profiles. If these conditions are not met, cellular groups may nevertheless form a complex structure through a process of ‘self-assembly’ (Sasai, 2013; Turner et al., 2016). The reduction of neuromasts to around 5% of their original size shows that intact resident cells can rapidly recreate their original microenvironment to rebuild a neuromast with normal organization, proportions and polarity. Although these observations suggest autonomy, extrinsic sources of information including the extracellular matrix that remains intact after cell loss may serve as a blueprint for epithelial organization. Yet, unless such patterns are rebuilt together with the organ, neuromasts architecture and proportions would depend on the area occupied by the regrowing epithelium. In other words, cell-fate acquisition and cell-class distribution must be tissue-size dependent. However, we show that neuromast regain geometric order as early as 2 days after injury, when their cellular content is less than 60% of the original. Although our results do not irrefutably demonstrate self-organization during neuromast regeneration, they strongly support this idea. We argue that self-organization is an optimal morphogenetic process to govern organ repair because (i) it requires the least amount of previous information and (ii) it is robust to run-off signals that could lead to catastrophic failure.

Conclusions

Understanding how tissues respond to the inherently random nature of injury to recapitulate their architecture requires the identification of cues and signals that determine cell-fate acquisition, localization and three-dimensional organization. Here we reveal an archetypal sensory organ endowed with isotropic regenerative ability and responses that comply to damage severity, nature and localization. An important corollary of our findings is that progenitor behavior is a facultative status that every sustentacular cell can acquire or abandon during regeneration (Blanpain and Fuchs, 2014; Wymeersch et al., 2016). Importantly, we illustrate a machine learning implementation to identify features that predict cell-fate choices during tissue growth and morphogenesis. This quantitative approach is simple and model-independent, which facilitates its application to other organs or experimental systems to understand how multiple cells interact dynamically during organogenesis and organ regeneration in the natural context of the whole animal, and to identify how divergences from the normal regenerative processes lead to failed tissue repair.

Materials and methods

Zebrafish strains and husbandry

Zebrafish were maintained under standard conditions, and experiments were performed in accordance with protocols approved by the PRBB Ethical Committee of Animal Experimentation of the PRBB Barcelona, Spain. Eggs were collected from natural spawning and maintained at 28.5°C in Petri dishes at a density of up to 50 per dish. Transgenic lines used were ET(krt4:EGFP)SqGw57A (referred to in the text as SqGw57A) (Kondrychyn et al., 2011), ET(krt4:EGFP)SqET4 (SqET4) (Parinov et al., 2004), Tg[Myo6b:actb1-EGFP] (Kindt et al., 2012), Tg[−8.0cldnb:Lyn-EGFP] (Cldnb:lynGFP) (Haas and Gilmour, 2006), Tg[Alpl:mCherry] (Steiner et al., 2014), Tg[Sox2-2a-sfGFPstl84] (referred to as Sox2:GFP) (Shin et al., 2014). To label cell nuclei, in vitro transcribed capped RNA coding for histone 2B-mCherry was injected in 1–4 cell embryos at a concentration of 100 ng/μl (Rosen et al., 2009). Throughout the study, zebrafish larvæ were anesthetized with a 610 µM solution of the anesthetic 3-aminobenzoic acid ethyl ester (MS-222).

Laser-mediated cell ablations

For in toto cell ablation, we used the iLasPulse laser system (Roper Scientific SAS, Evry, France) mounted on a Zeiss Axio Observer inverted microscope equipped with a 63X water-immersion objective (N.A. = 1.2) (Xiao et al., 2015). The same ablation protocol was used for all experiments using five dpf larvæ. Briefly, zebrafish larvæ were anesthetized, mounted on a glass-bottom dish and embedded in 1% low-melting point agarose. Three laser pulses (355 nm, 400 ps/2.5 μJ per pulse) were applied to each target cell. After beam delivery, larvæ were removed from the agarose and placed in anesthesia-free embryo medium. All ablations were systematically performed on the L2 or L3 posterior lateral-line neuromasts, except for those in Figure 6F, for which we targeted the LII.2 neuromast.

Phalloidin staining

Samples were fixed in 4% PFA overnight at 4°C, washed several times in 0.1% PBSTw and incubated in phalloidin-Alexa 568 or Alexa 488 (Invitrogen) diluted 1:20 in 0.1% PBSTw overnight at 4°C. Samples were washed several times in 0.1% PBSTw and mounted in 0.1% PBSTw with Vectashield (1/100, Vector Labs, Burlingame, CA, USA).

Regeneration analysis and quantification

For quantification of cell numbers during neuromast regeneration, Tg[Cldnb:lynGFP; SqGw57A; Alpl:mCherry] zebrafish larvæ were anesthetized, mounted on a glass-bottom dish and embedded in 1% low-melting point agarose. All samples were imaged before injury, 4 hpi and every 24 hr up to 7 dpi with an inverted spinning-disc confocal microscope (Zeiss by Visitron), under a 63X water-immersion objective. After imaging, larvæ were quickly transferred to anesthetic-free medium. Cells were manually counted using the FIJI multi-point tool by scrolling throughout the entire volume of the neuromast. Cell classes were identified by the following criteria: Interneuromast cells: Cldnb:lynGFP(+), SqGw57A(-), Alpl:mCherry(+). Mantle cells: Cldnb:lynGFP(+), SqGw57A(+), Alpl:mCherry(+). Sustentacular cells: Cldnb:lynGFP(+), SqGw57A(+), Alpl:mCherry(-). Hair cells: Cldnb:lynGFP(+), SqGw57A(-), Alpl:mCherry(-). Hair cell identity was verified by the concomitant observation of the correct transgene expression pattern, central-apical location and the presence of a hair-cell bundle. Data was processed and analyzed using GraphPad Prism version 6.04 for Windows (GraphPad Software, La Jolla, CA, USA, www.graphpad.com). In the box plots, the boundary of the box closest to zero indicates the 25th percentile (q1), a black line within the box marks the median, and the boundary of the box farthest from zero indicates the 75th percentile (q3). Whiskers above and below the box include points that are not outliers. Points are considered as outliers if they are bigger than q3 + 1.5(q3 – q1) or smaller than q1 – 1.5(q3 – q1).

Videomicroscopy, cell tracking and lineage tracing

Larvæ were anesthetized, mounted onto a glass-bottom 3 cm Petri dish (MatTek) and covered with 1% low-melting point agarose with diluted anesthetic. Z-stack series were acquired every 15 min at 28.5°C using a 63X water-immersion objective. Cells were tracked overtime using volumetric Z-stack images with FIJI plugin MTrackJ (Meijering et al., 2012). Movies were registered two times for image stabilization and centered upon the centroid of the surviving group of cells and the subsequent regenerating organs. Founder cells are identified from 1 to 6 (n) and their daughter cells receive 2 n and 2n + 1 identities. All images were processed with the FIJI software package.

Pharmacology

All pharmacological treatments were performed as described previously (López-Schier and Hudspeth, 2006; Wibowo et al., 2011; Pinto-Teixeira et al., 2015). Briefly, the following concentrations and timings used were: Neomycin sulfate (Sigma, St. Louis, MO) 250 µM for 45 min; N-[N-(3,5-difluorophenacetyl)-L-alanyl]-S-phenylglycine-t-butyl ester (DAPT) (Sigma) 100 µM for 24–48 hr. Equal amounts of DMSO were diluted in embryo medium for control specimens.

Random forest prediction

Random forest algorithms use the majority vote of numerous decision trees based on selected features to predict choices between given outcomes (Murphy, 2012). We used a list of spatial, movement and neighborhood features (see Suppl. Table 1) to perform the random forest prediction of fate choice. We trained the random forest on 14 experiments and tested our prediction on one left-out experiment in a round robin fashion, leading to 15 test sets overall. To evaluate our prediction, we calculated Matthews correlation coefficient (MCC) (Matthews, 1975), which accounts for imbalance in our data (e.g. 78% of all divisions are SS divisions). The MCC is calculated by:

MCC=TP × TN  FP × FN(TP + FP)(TP + FN)(TN + FP)(TN + FN)

where TP denotes true positive, TN true negative, FP true positive and FN false negative predictions. The MCC can have values between −1 and +1, where −1 is a completely incorrect, 0 a random and +1 a perfect prediction. To evaluate the variance of the MCC on the 15 test sets we used a bootstrapping approach, where we draw 15 samples from all test sets with replacement 15 times. From this resampled data we calculated the mean MCC and the standard deviation as shown in Figure 8B and D. All machine-learning analyses were performed using MATLAB (Version 2015b on a Windows 7 machine)

References

  1. 1
  2. 2
  3. 3
  4. 4
  5. 5
  6. 6
  7. 7
  8. 8
  9. 9
  10. 10
  11. 11
  12. 12
  13. 13
  14. 14
  15. 15
  16. 16
  17. 17
  18. 18
  19. 19
  20. 20
  21. 21
  22. 22
  23. 23
  24. 24
  25. 25
  26. 26
  27. 27
  28. 28
  29. 29
  30. 30
  31. 31
  32. 32
  33. 33
  34. 34
    Machine Learning: A Probabilistic Perspective
    1. KP Murphy
    (2012)
    Cambridge: MIT press.
  35. 35
  36. 36
  37. 37
  38. 38
  39. 39
  40. 40
  41. 41
  42. 42
  43. 43
  44. 44
  45. 45
  46. 46
  47. 47
  48. 48
  49. 49
  50. 50
  51. 51
  52. 52
  53. 53
  54. 54
  55. 55

Decision letter

  1. Tanya T Whitfield
    Reviewing Editor; University of Sheffield, United Kingdom

In the interests of transparency, eLife includes the editorial decision letter and accompanying author responses. A lightly edited version of the letter sent to the authors after peer review is shown, indicating the most substantive concerns; minor comments are not usually included.

[Editors’ note: this article was originally rejected after discussions between the reviewers, but the authors were invited to resubmit after an appeal against the decision.]

Thank you for submitting your work entitled "Multicellular self-organization underlies the accuracy of sensory-organ regeneration" for consideration by eLife. Your article has been reviewed by three peer reviewers, and the evaluation has been overseen by Tanya Whitfield as Reviewing Editor and a Senior Editor. The reviewers have opted to remain anonymous.

Our decision has been reached after consultation between the reviewers. Based on these discussions and the individual reviews below, we regret to inform you that your work will not be considered further for publication in eLife.

There was praise among the reviewers for the rigorous and careful quantitative analysis of cell behaviour during neuromast regeneration. However, the overall consensus was that the manuscript tackled many disparate aspects, without focussing in depth on any one of them; it therefore lacked a clear 'big message'. In their discussions, the reviewers highlighted the lack of a significant conceptual advance, together with missing citation or discussion of some of the previous literature that would help to explain or support the results. These included Jiang, Kindt and Wu; 2017; Kempfle et al., Sci Rep, 2016 and Seleit et al., Development, 2017.

Suggestions for improving the manuscript included a cut of earlier sections of the manuscript and a deeper analysis of the later ones; or, placing the results in the context of known molecular pathways of neuromast regeneration. The full reviews, which are appended below, have a number of more specific concerns and suggestions.

Reviewer #1:

The authors of this study proposed a "self-regulatory mechanism that guides the regenerative process (of neuromasts) to identical outcome despite the intrinsically stochastic nature of damage".

While their data largely support the authors’ conclusion, their analysis of potential mechanisms and discussion of how this study enriches what is already known about hair-cell organ regeneration are lacking. Specifically, there were a number of points in the Results and Discussion where conclusions were stated, yet references to recent studies that would support those conclusions and provide potential mechanisms for what the authors observed were notably absent. In addition, there were a few experiments where I was unsure about the authors’ controls.

Point by point concerns are addressed below:

- Results: "laser targeted cells undergo rapid apoptosis" needs a citation or supporting data.

- Results: "ablated neuromasts and their neighboring Schwann cells without affecting the adjacent interneuromast cells"; both the interneuromast cells and Schwann cells contain the mcherry fluorophore, correct? How did the authors verify that the interneuromast cells were not damaged when nearby Schwann cells were ablated?

- Results: is there anything in the literature that would indicate that inhibition of ErB kinase would initiate production of additional neuromasts at 5 dpf i.e. following the migration and establishment of the posterior lateral-line organs?

- Results: The result observed after treating ablated neuromasts with AG1478 is interesting, but I'm puzzled as to why the authors didn't take it one logical step further and address whether wnt/bcatenin signaling is absent i.e. could they initiate interneuromast cell proliferation with exposure to BIO?

- Figure 3K-O: By eye, dorsal mantle cell recovery doesn't look very robust. Is this a representative example? The total cell numbers in Figure 3Z don't really address if certain classes of cells fully recover.

- Figure 4—figure supplement 1 and its related Discussion: I'm not quite sure what the authors claim regarding amount and timing of genetically encoded marker expression means. Are they using GFP reporter expression as a proxy for actual gene expression? Is so, it needs to be supported by endogenous expression data (e.g. insitus) or references to previous studies that demonstrate exogenous reporter expression reflects endogenous gene expression in these reporter lines.

- Results: When describing the combination of reporter lines used, it would be helpful to the reader to put the name of the line next to the cellular structures they label in the text.

- Figure 4U-W; X-Z: By eye, the polarity of phalloidin labeling looks mixed, and X-Z is further confounded by the presence of immature hair-cells. To better make the case for fully reformed radial symmetry 1) a control, non-ablated NM should be shown for comparison 2) the axis of symmetry should be indicated with a hashed line 3) relative orientation of HC per NM should be quantified.

- The observation that a few regenerating hair cells recovered planar polarity could be explained by Emx2 expression in 1 of 2 sibling hair cells (Jiang, Kindt and Wu, 2017). It's important to discuss this potential mechanism either here or in the Discussion section even if the authors didn't test for Emx2 expression in this study.

- Subsection “Sustentacular and mantle cells have different regenerative potential”, the conclusion that Sox2(+) sustentacular cells are tripotent is supported by evidence from a previous study; Sox2 has been shown to both expand progenitor cells and initiate their differentiation in the cochlea (Kemfle et al., Sci Rep, March 2016).

- In the same section, "predecing". Typo.

- "Assessed spatial distribution…60 hrs of regeneration" which ablation protocol did they use here?

- Figure 5E: Both supporting cells and hair cell are labeled with GFP. What criteria did the authors use to differentiate supporting cells from hair cells?

- Figure 6 K-L: Cell cycle length appears strongly correlated to regeneration time (K), but not so strongly correlated to neuromast size overall. The legend and the text states in L that the cell cycle length doesn't correlate well until 24 cells, but that doesn't seem directly apparent from the graph or the regression line.

- Subsection “Sox2(+) sustentacular cells are equipotent and plastic”: There is no mention of Figure 7C in the results, and it's unclear to me from the figure legend how this data was generated.

- Discussion section; "suggesting that Sox2 itself is not essential" is overreaching. The observation made from interneuromast cells may show that Sox2 is not sufficient to drive regeneration, but it does not demonstrate it's not essential.

- Discussion section, second paragraph: Not mentioning the Emx2 study here is a glaring omission.

- The Discussion overall would benefit from a more thorough incorporation of molecular mechanisms reported in previous studies examining supporting cell proliferation, differentiation, and repolarization in hair-cell organs to understand how this study's results enriches our understanding of the self-regulatory processes that guide lateral-line organ restoration.

Reviewer #2:

This study focus on one of the most intriguing and fascinating aspects or regenerative biology: the accuracy with which organs respond to injuries to restore proper size and cellular organization. The authors use neuromasts of fish to explore this phenomenon and fully exploit the advantages of the system. The videos and figures are of very high quality and aesthetic value, and the authors extract from them an exhaustive quantification of cellular responses during regeneration. I find the manuscript nicely written and in most sections easy to follow.

I failed, however, to understand what is the main new finding that the authors are presenting. It is known from several models that tissues (epidermis), organs (liver, skin, the fish heart), limbs (axolotls and newts, caudal fin in fish), systems (hematopoietic) and even organisms (planarians) can re-generate accurately to their pre-injury state. The authors state clearly some of the open questions in the field, but it feels they fall short at addressing them thoroughly when using the neuromast. The major achievement of the manuscript is the acquisition of an extensive dataset on individual cellular behavior during the regeneration response. The two main points I see are that neuromasts regenerate what is lost, and cell identity is acquired along a radial axis. Unfortunately and despite the extensive analysis performed and the quality of the data, it is hard to understand the conceptual novelty of this study.

Major comments:

During the Introduction and the initial section of Results, it feels that the authors neglect several findings from a previous publication of another group, Sánchez et al., 2016, on the regeneration of neuromasts upon ablation of the entire organ. The results presented in the submitted paper do not match a proportion of the results previously published (neuromast regeneration upon ablation or pharmacological inactivation of Schwann cells). The authors should explain this discrepancy and acknowledge that organ regeneration after severe injuries can indeed happen from other sources as well.

The authors state that inter neuromast cells are not necessary for the regeneration response. They show however that in their absence, organs regenerate to a smaller size. What is then the role of interneuromast cells?

In reference to the autonomous regeneration of epithelial polarity the authors show that neuromasts regenerate the proper polarisation after severe injuries in the presence of interneuromast cells, or after ablation of hair cells in the absence of interneuromast cells. To claim an autonomous process, they should look at the polarisation plane after severe injuries and in the absence of interneuromast cells – experiments done for a previous section but not analyzed for epithelial polarity. In addition, the authors should consider that anisotropic forces might originate not only from interneuromast cells and therefore evaluate whether this is the valid conclusion from their experimental data.

The authors emphasize the accuracy of the regenerative response, but this seems to be restricted to the distribution of cell types. The number of cells, and therefore the size of the organ, is restored only to a 70% after 7dpi. This needs to be stressed, taken in consideration for the conclusions, and contemplated in the Discussion.

It is still not clear whether neuromast cells acquire a proper position during regeneration or whether they acquire their fate based on the position they occupy. The authors state both claims along the manuscript although they are mutually exclusive.

The claim that sustentacular cells revert to an embryonic (undifferentiated) state immediately after tissue injury is based in the transient down-regulation of a single fluorescent reporter. The authors should present data for some of the many additional markers or tone down their conclusions. On a more general view, I disagree with the claim that undifferentiated equals embryonic.

Reviewer #3:

The work by Oriol Viader-Llargués et al. examines how zebrafish mechanosensory organs, neuromasts (NMs), respond to injury. The authors ablate portions of various sizes from different regions of the organ and then examine organ regeneration using live imaging of transgenes that mark distinct cell lineages within NMs. A number of previous studies examined regeneration of a single NM cell type, hair cells; however, regeneration capacity of other cell types has not been studied. While carefully done, the study is largely descriptive and does not build on the previous knowledge of molecular mechanisms that drive NM development and regeneration. Nevertheless, this work presents a number of novel findings, related to the cellular mechanisms of organ regeneration that should contribute significantly to the field of organ regeneration: 1) NMs can fully regenerate up 70 to 90% of their original size regardless of injury site; 2) regeneration process restores relative proportion and proper polarity of NM cell types; 3) regeneration is based on a resident population of Sox2-positive cells (not on stem cells) that restore all cell types within NMs; 4) even severely damaged organs (as few as 4 cells left) can fully regenerate; 5) cell fate during regeneration can be predicted based on the distance of dividing progenitors from the center. I have a few specific comments related to data interpretation and conclusions.

As the authors point out, previous studies looking at hair cell regeneration found that the source of hair cell progenitors resides in NM poles. Present study found that Sox2+ progenitors throughout the NM contribute to hair cells. How do they reconcile the differences? This should be discussed.

Ablation of inter-NM cells led to smaller regenerating NM after injury. Again, this issue is not discussed.

The authors conclude that Sox2+ cells are resident progenitors that potentially dedifferentiate following injury and then contribute to all three distinct cells types in the NM. Have they tried eliminating Sox2+ cells to support this assertion?

Related to the last point, it would certainly add to the impact of the paper if the authors can show that Sox2+ cells de-differentiate, based on molecular markers.

[Editors’ note: what now follows is the decision letter after the authors submitted for further consideration.]

Thank you for resubmitting your work entitled "Live cell-phylogeny tracing and machine learning reveal patterns of organ regeneration" for further consideration at eLife. Your revised article has been favorably evaluated by Naama Barkai as Senior editor, Tanya Whitfield as Reviewing editor, and three reviewers.

The three reviewers are all very positive about the revisions to the manuscript, and appreciate the care that has gone into the revisions. However, all three have some additional comments and queries that should be addressed. These mostly involve requests for further discussion and context, together with correction of typos, and should be quick to complete. The full reviews are appended here for your information.

Reviewer #1:

The revised version of the manuscript by Oriol Viader-Llargués et al. is significantly improved from the previous version. Its message is more focused, its conclusions better supported by the data, and references to relevant work were incorporated.

With that said, I have a few major questions that don't necessarily need to be addressed with additional experiments:

1) The mantle cell linage data show that the mantle cells do not give rise to any cell class other than their own, leading to the conclusion that mantle cells are not essential contributors to neuromast repair. Yet none of their injury paradigms leave only the mantle cells intact i.e. every injury with mantle cells present also has sustentacular cells present. It could be the case that mantle cells are essential to regenerating a neuromast if the sustentacular cells are absent.

The authors refer to a study in their rebuttal that suggests mantle cells are capable of producing all neuromast cell classes in a regenerating fin. Do the authors predict this would be the case, if they ablated everything but the mantle cells? If not, why?

2) The authors discussed the potential influence of interneuromast cells on planar polarity, but did not elaborate influence of INCs on neuromast size (Figure 4K-P). The result is striking and should be discussed further than the statement in the Results-"non-essential, yet appreciable contribution to regeneration".

Reviewer #2:

In the new version of this manuscript, Viader-Llargues, Lopez-Schier, and colleagues present their results with a more clear aim, focusing on quantitative aspects of the regeneration response and new tools to investigate it. My main points were properly addressed, namely toning down certain conclusions, including alternative scenarios and focusing on defined topics rather than superficially following many. I appreciate particularly the scientific quality and dedication of the authors' replies to many of my previous concerns. The lack of overshooting of cell types during the regeneration response was a very nice addition that contrasts to other regenerative systems, and illustrates how controlled the regenerative response is in neuromasts.

These are my comments on the current version:

In the title and along the manuscript the authors use the term "cell-phylogeny tracing". Is there any reason not to use the widely accepted term "lineage" instead? I feel that "lineage" works better in this context since it reflects more accurately the continuum of the data that the authors have acquired.

The issue of self-regulation of the regenerative response is dealt more carefully than in the previous version. Still, I would like the authors to be more explicit about their interpretation of the data, mainly on the Discussion. One of the aspects the authors focus on is the re-establishment of a polarity axis, and this is a very interesting aspect of the regeneration response since polarity is heterogeneous among different neuromasts of the posterior lateral line – unlike cell types and their distribution. In other words: neuromasts have one solution for the cell-type problem, but two solutions for the polarity problem. Previous work from the group (Lopez-Schier et al., 2004 and Lopez-Schier and Hudspeth, 2006) has clearly stated that the polarity axis is related to the final migration of the neuromasts during organogenesis. I find troubles imagining how a self-renewing organ (or 4-10 cells, to put it bluntly) will always choose the original polarity in the absence of external cues. I agree with the authors that external cues have minor roles in the establishment of cells types or position of cells, but extending self-organization to the re-establishment of polarity seems inappropriate to me. Am I missing something? Maybe the authors could speculate about the role of the afferent axons (I guess they should remain to some extent under the injury paradigms used in this study), which they have shown to display a high accuracy recognizing hair cells of a given polarity (Pujol-Marti, Current Biology 2014).

I particularly enjoyed the reply of the authors regarding the differences and similarities in mantle cell behavior during homeostasis and regeneration. I think it would be an added value if they incorporate these concepts in the Discussion.

Results section, paragraph two; I believe that the authors should include a reference to Grant et al., 2005. In fact, the authors do so in their response to Reviewer #1.

Subsection “Sustentacular cells are equipotent and plastic”. I found this part difficult to understand. Either the written numbers do not match what the authors show between brackets, or there is some mistake in the annotation. What is "peak at 8h, mean+-s.d at 14+-9 hours"? Overall, I find this part the less clearly written of the manuscript. Also, is it valid to call a peak sharp with such a big s.d.?

Discussion paragraph one: The authors state that sustentacular cells are equipotent, which I think is an overstating.

The data presented in the manuscript reveals that 4 or more of them are enough to regenerate the entire organ. The use of equipotency, in my view, states that all of them do the same during the process. In their accurate data in Figure 7, they show that some of them generate M cells, some other S cells, and some other combinations of H and other cell types. I understand that equipotent members of a population could behave in different manners based on either stochastic internal programs or external cues, but to prove which is the case demands a more extensive dataset and a deep mathematic analysis, which is not the scope of the present study – although this is a fantastic system to tackle it! I feel that the use of equipotent in this context assumes features that were not tested experimentally. I suggest the authors stay with "tri-potent", a term that they have used along Results.

Reviewer #3:

This is a revised version of a manuscript by Oriol Viader-Llargués et al. that deals with cellular mechanisms of neuromast regeneration. The previous version was criticized by its descriptive nature, lack of focus and omission of some references. The authors largely remedied these issues, although I still feel that some of their findings are not sufficiently discussed in the context of previous studies.

Comments:

The authors mention that their findings contrast those of Sanchez et al. study, but do now offer any discussion as to why that is the case. I think it is important to offer at least some explanation.

It seem that examples of severe ablation following full regeneration always leave at least one mantle cell (“Neuromast architecture recovers after severe loss of tissue integrity”). If this is indeed the case, it may indicate that a combinatorial signal(s) is required to initiate organ regeneration.

There are only two cases when sustentacular cell progenitors gave rise to all three cell types. Based on these small numbers, I think the authors need to be careful concluding that a sustentacular cell is a multipotent progenitor for all cell types in the neuromast.

The finding that at least 4 cells are needed to reconstitute an organ is intriguing. This is reminiscent of planarian organ regeneration where 1/300th part but not less can reconstitute full animal. It was later discovered that this is a minimum fraction of the animal roughly containing at least one stem cell (neoblast) necessary to regenerate all tissue lineages. Again, some discussion, as to why the authors think this is case (4 cells but not fewer are required) is warranted

Discussion paragraph two, final sentence: Seleit et al., 2017 showed that "new cell type" exists in zebrafish. They also showed that mantle cell are neuromasts stem cells. Thus, it is important to discuss how homeostatic cell renewal differs from cell renewal during NM regeneration (i.e. mantle cell as stem cell during homeostasis vs. sustentacular cell as a multipotent progenitor during regeneration). This is an interesting question, as there are examples where cells can be driven to change their fate by extreme injury paradigm.

https://doi.org/10.7554/eLife.30823.020

Author response

[Editors’ note: the author responses to the first round of peer review follow.]

Reviewer #1:

The authors of this study proposed a "self-regulatory mechanism that guides the regenerative process (of neuromasts) to identical outcome despite the intrinsically stochastic nature of damage".

While their data largely support the authors’ conclusion, their analysis of potential mechanisms and discussion of how this study enriches what is already known about hair-cell organ regeneration are lacking. Specifically, there were a number of points in the Results and Discussion where conclusions were stated, yet references to recent studies that would support those conclusions and provide potential mechanisms for what the authors observed were notably absent. In addition, there were a few experiments where I was unsure about the authors’ controls.

We thank the reviewer for these comments, who indicates that some data were sub-optimally discussed. We believe that this led to the critique that the work lacks novelty. Consequently, we (i) have thoroughly revised the discussion and the references to include important previous work, (ii) emphasize the novel aspects of our work, and how our conclusions could not have been predicted by previous results, and (iii) have reduced parts of the paper that were not central to the main message, and expanded others by including control experiments and further quantitative analyses. Specifically, we include new panels in Figure 1 that reveal the accuracy and specificity of our laser-mediated cell-ablation protocol, and an accompanying supplemental video. We have also rearranged Figures 6 and 7 and generated a new figure to explain the machine learning approach.

Point by point concerns are addressed below:

- Results: "laser targeted cells undergo rapid apoptosis" needs a citation or supporting data.

We thank the reviewer for raising this point. Indeed, we do not demonstrate that laser ablation triggers apoptotic cell death. Thus, we now state that: “we certified that laser-targeted cells are rapidly eliminated from the neuromast epithelium with no detectable collateral damage”. We include new data showing target-cell elimination in Figure 1K-P and Video 1.

- Results: "ablated neuromasts and their neighboring Schwann cells without affecting the adjacent interneuromast cells"; both the interneuromast cells and Schwann cells contain the mcherry fluorophore, correct? How did the authors verify that the interneuromast cells were not damaged when nearby Schwann cells were ablated?

In neuromasts of the transgenic lines that we have used, interneuromast cells and Schwann cells express different fluorescent proteins, which can be easily distinguished. Precisely to be able to distinguish between these cell classes is the reason behind our generation of a quintuple transgenic line for these experiments. We understand that the size of the images may have prevented the reviewer to appreciate these differences. However, because of the refocus of the work on the technical approach, these data have become superfluous and were eliminated. Please, note that these changes better define the message of the paper without altering its original conclusion.

- Results: is there anything in the literature that would indicate that inhibition of ErB kinase would initiate production of additional neuromasts at 5 dpf i.e. following the migration and establishment of the posterior lateral-line organs?

Yes, there is, and we have cited the relevant publications for the pharmacological inhibition of ErbB (Sánchez et al., 2016), as well as the genetic ablation of the receptor controlling the signal (Grant et al., 2005; López-Schier and Hudspeth, 2005). Yet, as stated above, the new focus of the paper has made this part of the results superfluous.

- Results: The result observed after treating ablated neuromasts with AG1478 is interesting, but I'm puzzled as to why the authors didn't take it one logical step further and address whether wnt/bcatenin signaling is absent i.e. could they initiate interneuromast cell proliferation with exposure to BIO?

This is a valid critique and a very interesting question that has not escaped our attention. However, a thorough investigation of Wnt signaling within the context of organ-size control would result in a longer and less focused paper. Nevertheless, we have added a paragraph that discusses cell proliferation during neuromast formation. Specifically, we state in the Discussion that “we did not observe regenerative overshoot of any cell class (Agarwala et al., 2015), suggesting the existence of a mechanism that senses the total number of cells and the cell-class balance during tissue repair (Simon et al., 2009). Previous work indicates that such mechanims may be based on interplay between FGF, Notch and Wnt signaling (Ma et al., 2008; Wibowo et al., 2011; Wada et al., 2013; Lush and Piotrowski, 2014; Romero-Carvajal et l., 2015; Kozlovskaja-Gumbrienė et al., 2017; Dalle Nogare and Chitnis, 2017).”

- Figure 3K-O: By eye, dorsal mantle cell recovery doesn't look very robust. Is this a representative example? The total cell numbers in Figure 3Z don't really address if certain classes of cells fully recover.

The images shown throughout the paper are representative of the data collected from every sample. A variable number and distribution of mantle cells also occurs in unperturbed organs, and representative examples are shown in Figure 3K-O, as well as in Figure 3P-Y.

- Figure 4—figure supplement 1 and its related Discussion: I'm not quite sure what the authors claim regarding amount and timing of genetically encoded marker expression means. Are they using GFP reporter expression as a proxy for actual gene expression? Is so, it needs to be supported by endogenous expression data (e.g. insitus) or references to previous studies that demonstrate exogenous reporter expression reflects endogenous gene expression in these reporter lines.

We thank the reviewer for pointing out this ambiguty. To clarify, we did not intend to relate the behavior of the GFP reporter lines to changes in expression of endogenous genes, other than that of the Gateway57A transgene. We simply highlight the correlation between the behavior of Sox2:GFP and Gateway57A (also GFP) as suggestive of sustentacular-cell reversion to a primordial status during regeneration. The reviewer, however, makes us realize that this suggestion may be interpreted as a statement that is supported by additional data, which is not the case. Therefore, we have rephrased the entire section. Specifically, we eliminated the data that was used to indicate that SqGw57A represents a live sensor of supporting-cell maturity.

- Results: When describing the combination of reporter lines used, it would be helpful to the reader to put the name of the line next to the cellular structures they label in the text.

We have used Figure 1 for this purpose. We now specify when preseting the transgenic tools the color variants of the fluorescent makers and the identity of the cell types that are marked. When presenting Figure 1, we spell-out that “Specifically, the Tg[alpl:mCherry] line expresses cytosolic mCherry in the mantle and interneuromast cells (Figure 1D). The Et(krt4:EGFP)sqgw57A (hereafter SqGw57A) expresses cytosolic GFP in sustentacular cells (Figure 1E). The Tg[-8.0cldnb:LY-EGFP] (Cldnb:lynGFP) express a plasma-membrane targeted EGFP in the entire neuromast epithelium (Figure 1F), and the Tg[Sox2-2a-sfGFP] (Sox2:GFP) expresses cytosolic GFP in all the supporting cells (Figure 1G). For hair cells, we use Et(krt4:EGFP)sqet4 (SqEt4) that expresses cytosolic GFP (Figure 1H), or the Tg(myo6b:actb1-EGFP) (Myo6b:actin-GFP) that labels filamentous actin (Figure 1I).”

- Figure 4U-W; X-Z: By eye, the polarity of phalloidin labeling looks mixed, and X-Z is further confounded by the presence of immature hair-cells. To better make the case for fully reformed radial symmetry 1) a control, non-ablated NM should be shown for comparison 2) the axis of symmetry should be indicated with a hashed line 3) relative orientation of HC per NM should be quantified.

We are not sure what the reviewer refers to when writing “mixed”. A neuromast always carries one half of the hair cells pointing in the opposite direction of the other half. This is stated in the Introduction of the paper using Figure 1C, and exemplified in Figure 1I. We have introduced changes in Figure 1I and Figure 4W and 4Z to address this comment, by including a double-head arrow to indicate the axis of epithelial planar polarity, and the dual orientation of the hair cells along this axis.

- The observation that a few regenerating hair cells recovered planar polarity could be explained by Emx2 expression in 1 of 2 sibling hair cells (Jiang, Kindt and Wu, 2017). It's important to discuss this potential mechanism either here or in the Discussion section even if the authors didn't test for Emx2 expression in this study.

The reviewer correctly points to a recent paper by Jiang and collaborators about the role of Emx2 in hair-cell orientation. Our manuscript does not address planar cell polarity in particular, and thus did not originally contain a reference or a discussion of the relevance of Emx2. Specifically, we believe that Emx2 instructs the hair cells to implement the planar polarity cues to decide in which direction to polarize, without affecting coherent local polarity. Although the role of Emx2 in planar polarization is certainly fascinating, discussing it at length is beyond the scope of our work. Nevertheless, we recognize that many readers would find it relevant. Thus, we now include a paragraph contextualizing Emx2 with our results. Specifically, we state that: “…recent studies have identified a transcription factor called Emx2 that regulates the orientation of hair cells in neuromasts of the zebrafish. Emx2 is expressed in one half of the hair cells of the neuromast (those oriented towards the tail) and absent in the other half (which are coherently oriented towards the head). Loss- and gain-of-function of Emx2 alter planar cell polarity in a predictable manner because loss of Emx2 renders neuromasts with every hair cells pointing towards the head of the animal, and misexpression orients hair cells towards its tail. Because the coherent local axis of polarity is not affected by these genetic perturbations, Emx2 may act in hair cells as a decoder of global polarity cues. This evidence, together with our results, suggests that during neuromast regeneration founder cells autonomously organize the variegated expression of Emx2 in the regrowing eithelium with consequent recovery of a coherent axis of planar polarity and with one half of the hair cells pointing opposite to the other half”.

- Subsection “Sustentacular and mantle cells have different regenerative potential”, the conclusion that Sox2(+) sustentacular cells are tripotent is supported by evidence from a previous study; Sox2 has been shown to both expand progenitor cells and initiate their differentiation in the cochlea (Kemfle et al., Sci Rep, March 2016).

The reviewer points to a paper that analyzed Sox2 in the mouse, whose inner ear does not possess the regenerative capacity of the neuromast. Specifically, Kemfle et al., (together with Millimaki, Sweet, and Riley, Dev. Biol. (2010) 338(2): 262) analyzed the role of Sox2 in hair-cell embryonic development, forced production in post-embryonic murine ears, and their natural regeneration in zebrafish. The main conclusion of Kemfle et al., is that Sox2 expands the pool of hair-cell progenitors, being more relevant to the specification of sensory hair cells, rather than that of non-sensory cells. We would like to note that we have not attempted to address hair-cell regeneration in our study, but rather the repair of the entire neuromast. Therefore, a direct extrapolation of the cited work to our results is not straightforward. Based on expression patterns of the endogenous Sox2 in neuromasts and the expression of the transgenic Sox2 sensor, we claim that Sox2 is unlikely to play a role in specifying a sub-set of cells that would serve as a pool of progenitors dedicated to regeneration of sensory or non-sensory cells. Moreover, we would like to cite the work of Millimaki, Sweet, and Riley, Dev. Biol. (2010) 338(2): 262, in which it was shown that forced expression of Sox2 in zebrafish expanded the epithelial domain that generates hair cells in the ear, but not so in neuromasts. This is now discussed in our manuscript. We now state that: “We propose that progenitor behavior is a facultative status that every sustentacular cell can acquire or abandon during regeneration.”

- In the same section, "predecing". Typo.

We thank the reviewer for identifying this error. It has been corrected by rephrasing the sentence.

- "Assessed spatial distribution… 60 hrs of regeneration" which ablation protocol did they use here?

We have used the same ablation protocol through the paper, which is now specifically stated in the Materials and methods section.

- Figure 5E: Both supporting cells and hair cell are labeled with GFP. What criteria did the authors use to differentiate supporting cells from hair cells?

The figure legend makes reference to hair cells, but these cells are not labeled in the accompanying images, which may have generated some confusion. In fact, it is precisely the lack of cytosolic green fluorescence in hair cells that we have used to quantify their number and pinpoint their localization during the lineage-tracing experiments. We now express this more clearly in the caption of the Figure 5.

- Figure 6K-L: Cell cycle length appears strongly correlated to regeneration time (K), but not so strongly correlated to neuromast size overall. The legend and the text states in L that the cell cycle length doesn't correlate well until 24 cells, but that doesn't seem directly apparent from the graph or the regression line.

We thank the reviewer for this comment. We specifically checked if the data is better described by a single regression, or by two regression lines with a ‘change point’ in between. Our model comparison accounts for the fact that the latter is more complex. Still, this model (weak correlation until 24 cells, strong correlation afterwards) explains the data significantly better than the single regression line.

- Subsection “Sox2(+) sustentacular cells are equipotent and plastic”: There is no mention of Figure 7C in the Results, and it's unclear to me from the figure legend how this data was generated.

We thank the reviewer for having identified this oversight on our part. The previous Figure 7C was mistakenly included because it showed data that were not used to support the conclusion of the paper. We have thus eliminated it.

- Discussion section; "suggesting that Sox2 itself is not essential" is overreaching. The observation made from interneuromast cells may show that Sox2 is not sufficient to drive regeneration, but it does not demonstrate it's not essential.

We agree with the reviewer. As stated above, the new version of our manuscript now says “We propose that progenitor behavior is a facultative status that every Sox2(+) sustentacular cell can acquire or abandon during regeneration. The Sox2 transcription factor marks many progenitor cells and often drives stem-cell dependent regenerative processes in a variety of animals (Millimaki et al., 2010; Neves et al., 2013; Reinhardt et al., 2015; Sweet et al., 2011). Sox2 is expressed widely in the neuromast, suggesting that Sox2(+) sustentacular cells are tri-potent progenitors.”

- Discussion section, second paragraph: Not mentioning the Emx2 study here is a glaring omission.

We have expanded the Discussion and include the references to Emx2, and we now write “Interestingly, recent studies have identified a transcription factor called Emx2 that regulates the orientation of hair cells in neuromasts of the zebrafish. Emx2 is expressed in one half of the hair cells of the neuromast (those oriented towards the tail) and absent in the other half (which are coherently oriented towards the head). Loss- and gain-of-function of Emx2 alter planar cell polarity in a predictable manner because loss of Emx2 renders neuromasts with every hair cells pointing towards the head of the animal, and misexpression orients hair cells towards its head. Because the coherent local axis of polarity is not affected by these genetic perturbations, Emx2 may act in hair cells as a decoder of global polarity cues. This evidence, together with our results, suggests that during neuromast regeneration founder cells autonomously organize the variegated expression of Emx2 in the regrowing eithelium with consequent recovery of a coherent axis of planar polarity and with one half of the hair cells pointing opposite to the other half. The future development of live markers of Emx2 expression will be able to test this prediction.”

- The Discussion overall would benefit from a more thorough incorporation of molecular mechanisms reported in previous studies examining supporting cell proliferation, differentiation, and repolarization in hair-cell organs to understand how this study's results enriches our understanding of the self-regulatory processes that guide lateral-line organ restoration.

Following the reviewer’s advice, we have substantially revised the Discussion and now include a sentence, fully referenced, about intercellular signaling pathways involved in neuromast formation and the regeneration of hair cells.

Reviewer #2:

This study focus on one of the most intriguing and fascinating aspects or regenerative biology: the accuracy with which organs respond to injuries to restore proper size and cellular organization. The authors use neuromasts of fish to explore this phenomenon and fully exploit the advantages of the system. The videos and figures are of very high quality and aesthetic value, and the authors extract from them an exhaustive quantification of cellular responses during regeneration. I find the manuscript nicely written and in most sections easy to follow.

I failed, however, to understand what is the main new finding that the authors are presenting. It is known from several models that tissues (epidermis), organs (liver, skin, the fish heart), limbs (axolotls and newts, caudal fin in fish), systems (hematopoietic) and even organisms (planarians) can re-generate accurately to their pre-injury state. The authors state clearly some of the open questions in the field, but it feels they fall short at addressing them thoroughly when using the neuromast. The major achievement of the manuscript is the acquisition of an extensive dataset on individual cellular behavior during the regeneration response. The two main points I see are that neuromasts regenerate what is lost, and cell identity is acquired along a radial axis. Unfortunately and despite the extensive analysis performed and the quality of the data, it is hard to understand the conceptual novelty of this study.

We share the reviewer’s enthusiasm about the biological problem that we investigate and the significance of our results, and are grateful for the encouraging words.

Major comments:

During the Introduction and the initial section of Results, it feels that the authors neglect several findings from a previous publication of another group, Sánchez et al., 2016, on the regeneration of neuromasts upon ablation of the entire organ. The results presented in the submitted paper do not match a proportion of the results previously published (neuromast regeneration upon ablation or pharmacological inactivation of Schwann cells). The authors should explain this discrepancy and acknowledge that organ regeneration after severe injuries can indeed happen from other sources as well.

We have cited Sánchez et al., 2016 in the Results section of our initial submittal. Because we could not replicate the findings of Sánchez et al., our data led us to a different conclusion. One possibility that may explain this discrepancy is the differences in the cell-ablation methods and/or the reporters used to identify Schwann cells. Nevertheless, the new focus of our work has made this part of the original paper no longer necessary, and was therefore suppressed. Please, note that this change does not alter any of the main conclusions of our work.

The authors state that inter neuromast cells are not necessary for the regeneration response. They show however that in their absence, organs regenerate to a smaller size. What is then the role of interneuromast cells?

We state that the interneuromast cells are not necessary for regeneration, but did not try to imply that they play no role in the process. We are more specific in the revised version of our manuscript, where we now write: “However, Sox2 is expressed widely in the neuromast, including the interneuromast cells that are not essential to neuromast repair. This observation suggests that Sox2 itself may not be essential for regeneration.” Additionally, we now indicate, on pp8., that “Interneuromast cells are not essential for neuromast regeneration in larval zebrafish, although they may contribute to mantle cell re-emergence”, and on pp9 that “These observations reinforce our previous suggestion that interneuromast cells have a non-essential but appreciable contribution to regeneration.”

In reference to the autonomous regeneration of epithelial polarity the authors show that neuromasts regenerate the proper polarisation after severe injuries in the presence of interneuromast cells, or after ablation of hair cells in the absence of interneuromast cells. To claim an autonomous process, they should look at the polarisation plane after severe injuries and in the absence of interneuromast cells – experiments done for a previous section but not analyzed for epithelial polarity.

We agree with the reviewer, and we no longer make this claim.

In addition, the authors should consider that anisotropic forces might originate not only from interneuromast cells and therefore evaluate whether this is the valid conclusion from their experimental data.

The reviewer is correct in pointing that we have neglected potential additional sources of mechanical forces. In the case of planar polarity in particular, forces by the interneuromast cells are obvious candidates because of the position of these cells relative to the polarity axis. Yet, it is true that other forces may play an instructive role, and this is something that we have not analyzed because neither our own work, nor the literature, suggests candidates that we can test experimentally. Therefore, we now state “To test if plane-polarizing cues derive from anisotropic forces exerted by the interneuromast cells that are always aligned to the axis of planar polarity of the neuromast epithelium, we ablated these cells flanking an identified neuromast, and concurrently killed the hair cells with the antibiotic neomycin (Figure 4X-Y). In the absence of interneuromast cells regenerating hair cells recovered normal coherent planar polarity (n=16), suggesting the existence of alternative sources of polarizing cues (Figure 4Z). Collectively, these findings reveal that as few as 4 supporting cells can initiate and sustain integral organ regeneration.”

The authors emphasize the accuracy of the regenerative response, but this seems to be restricted to the distribution of cell types. The number of cells, and therefore the size of the organ, is restored only to a 70% after 7dpi. This needs to be stressed, taken in consideration for the conclusions, and contemplated in the discussion.

The reviewer recognizes that the focus of our work has been directed towards the recovery of cell classes, their spatial distribution, their relative numbers and of hair-cell polarity, and that we have not focused on organ size. The reason behind this emphasis is that the recovery of organ proportions and geometry remain far lesser understood biological problems. We show that cell-fate acquisition and cell-class distribution are not tissue-size dependent. We also reflect upon our data and clearly indicate that: “Interneuromast cells are not essential for neuromast regeneration in larval zebrafish, although they may contribute to mantle cell re-emergence”.

Additionally, the new version of the manuscript highlights and strengthen this focus, but also briefly touches-upon organ size in the discussion, as stated above: “Moreover, we did not observe regenerative overshoot of any cell class (Agarwala et al., 2015), suggesting the existence of a mechanism that senses the total number of cells and the cell-class balance during tissue repair (Simon et al., 2009). Previous work indicates that such mechanisms may be based on interplay between FGF, Notch and Wnt signaling (Ma et al., 2008; Wibowo et al., 2011; Wada et al., 2013; Romero-Carvajal et l., 2015; Dalle Nogare and Chitnis, 2017).”

It is still not clear whether neuromast cells acquire a proper position during regeneration or whether they acquire their fate based on the position they occupy. The authors state both claims along the manuscript although they are mutually exclusive.

We thank the reviewer for this comment because it is central to the problem of regeneration of a complex structure in vivo. We understand that the reviewer finds some ambiguity in our discussion, specifically whether cells acquire their fate based on position, or position based on fate. Put in other words, whether there is a hierarchical relationship between these two features and, if so, which is their relationship. Our careful tracing of cell fate and machine-learning based analyses of cellular behavior are meant to specifically address this important point. By evaluating patterns of cellular behavior, we find no evidence of directional cell movement or intercalation to support the notion that cells acquire position based on a pre-defined fate. We also found no evidence of incorrectly-localized cell classes being extruded from the epithelium or eliminated by apoptosis. Importantly, by tracking clone-growth trajectories we find that the cells acquire fate based on their localization in the epithelium and, specifically, their position along the mediolateral axis of the epithelium.

The claim that sustentacular cells revert to an embryonic (undifferentiated) state immediately after tissue injury is based in the transient down-regulation of a single fluorescent reporter. The authors should present data for some of the many additional markers or tone down their conclusions. On a more general view, I disagree with the claim that undifferentiated equals embryonic.

Reviewers 1 and 2 have an identical concern about this point. We did not intend to relate the behavior of the GFP reporter lines to changes in endogenous gene expression, other than the transgenes themselves. Our highlighting the correlation between the expression of Sox2:GFP and Gateway57A (also GFP) during early development of the lateral line was done to indicate a possible reversion of sustentacular cells to a primordial status during regeneration. The current version of the manuscript, with its new focus, no longer makes this claim.

Reviewer #3:

The work by Oriol Viader-Llargués et al. examines how zebrafish mechanosensory organs, neuromasts (NMs), respond to injury. The authors ablate portions of various sizes from different regions of the organ and then examine organ regeneration using live imaging of transgenes that mark distinct cell lineages within NMs. A number of previous studies examined regeneration of a single NM cell type, hair cells; however, regeneration capacity of other cell types has not been studied. While carefully done, the study is largely descriptive and does not build on the previous knowledge of molecular mechanisms that drive NM development and regeneration. Nevertheless, this work presents a number of novel findings, related to the cellular mechanisms of organ regeneration that should contribute significantly to the field of organ regeneration: 1) NMs can fully regenerate up 70 to 90% of their original size regardless of injury site; 2) regeneration process restores relative proportion and proper polarity of NM cell types; 3) regeneration is based on a resident population of Sox2-positive cells (not on stem cells) that restore all cell types within NMs; 4) even severely damaged organs (as few as 4 cells left) can fully regenerate; 5) cell fate during regeneration can be predicted based on the distance of dividing progenitors from the center. I have a few specific comments related to data interpretation and conclusions.

We are happy to learn that the reviewer finds that our work presents a number of novel findings related to the cellular mechanisms of organ regeneration, and that it will contribute significantly to the field of organ regeneration.

As the authors point out, previous studies looking at hair cell regeneration found that the source of hair cell progenitors resides in NM poles. Present study found that Sox2+ progenitors throughout the NM contribute to hair cells. How do they reconcile the differences? This should be discussed.

We do not find any contradiction between our findings and those that have led to the conclusion that there are regional differences in supporting-cell behavior, specifically because early studies have focused on hair-cell regeneration rather than whole-neuromast repair. Our finding that sustentacular cells are tripotent progenitors and that the neuromast epithelium is symmetric in its regenerative ability do not contradict previous conclusions.

Ablation of inter-NM cells led to smaller regenerating NM after injury. Again, this issue is not discussed.

This is indeed possible. Thus, we now state that: “Interneuromast cells are not essential for neuromast regeneration in larval zebrafish, although they may contribute to mantle cell re-emergence”.

The authors conclude that Sox2+ cells are resident progenitors that potentially dedifferentiate following injury and then contribute to all three distinct cells types in the NM. Have they tried eliminating Sox2+ cells to support this assertion?

This reviewer shares a concern with Reviewer 2 about the role of Sox2 cells during neuromast regeneration. Every cell, other than the hair cells, express Sox2. The hair-cells are post-mitotic and likely need supporting cells to remain in the epithelium. Therefore, eliminating all Sox2-positive cells will result in the elimination of the entire neuromast, which we have done and shown to lead to permanent organ loss. Instead, we support our assertion by tracking the fate of Sox2(+) cells using clonal analysis, live imaging and machine learning, which reveal that Sox2(+) supporting cells, specifically sustentacular cells, proliferate and differentiate in all three cells classes during neuromast repair.

Related to the last point, it would certainly add to the impact of the paper if the authors can show that Sox2+ cells de-differentiate, based on molecular markers.

We agree with the reviewer, but this is a difficult problem to address. The reason is that it is not clear what markers can be used to examine supporting-cell de-differentiation. In many sensory organs that include the ear and the lateral line, the transcription factor Atoh1 is a marker of committed but not terminally differentiated pro-sensory cells. We have reported previously that in neuromasts Atoh1 is only transiently expressed by supporting cells under low Notch signaling, and that it eventually becomes stabilized in hair-cell progenitors. During neuromast embryonic development, the chemokine receptor CXCR4b is expressed in the lateral line primordium, specifically at its front where uncommitted neuromast progenitor cells are located. Its expression ceases once neuromasts mature despite the continuous production of cells. Thus, CXCR4b cannot be assumed as a de-differentiation marker. Therefore, because of this lack of appropriate markers, the new version of the manuscript no longer makes the original claim about the de-differentiation of Sox2(+) cells during regeneration.

[Editors’ note: the author responses to the re-review follow.]

Reviewer #1:

The revised version of the manuscript by Oriol Viader-Llargués et al. is significantly improved from the previous version. Its message is more focused, its conclusions better supported by the data, and references to relevant work were incorporated.

With that said, I have a few major questions that don't necessarily need to be addressed with additional experiments:

1) The mantle cell linage data show that the mantle cells do not give rise to any cell class other than their own, leading to the conclusion that mantle cells are not essential contributors to neuromast repair. Yet none of their injury paradigms leave only the mantle cells intact i.e. every injury with mantle cells present also has sustentacular cells present. It could be the case that mantle cells are essential to regenerating a neuromast if the sustentacular cells are absent.

The reviewer is correct in arguing that we have not formally ruled out a contribution of mantle cells to neuromast regeneration, specifically when every other cell class is gone. Although such experiment will definitely be revealing, we have been unable to do it in a controlled manner for technical reasons. In light of this, we now include a slightly modified statement about the behavior of the mantle cells under our experiments, as well as their possible role under the condition identified by the reviewer. We now state in the Discussion:

“The behavior of the mantle cells is especially intriguing. Complete elimination of parts of the lateral line by tail-fin amputation have revealed that mantle cells are able to proliferate and generate a new primordium that migrates into the regenerated fin to produce new neuromasts (Dufourcq et al., 2006). This observation can be interpreted as suggesting that under some injury conditions, mantle cells are capable of producing all the cell classes of a neuromast. Transcriptomic profiling of mantle cells following neuromast injury revealed that these cells up-regulate the expression of multiple genes (Steiner et al., 2014). Furthermore, a recent study has revealed that mantle cells constitute a quiescent pool of cells that re-enters cell cycle only in response to severe depletion of sustentacular cells (Romero-Carvajal et al., 2015), suggesting that these cells may conform a stem-cell niche for proliferation of sustentacular cells. Thus, the collective evidence indicates that the mantle cells respond to damage and contribute to the regenerative processes, and may drive the regeneration of an entire organ if every other cell class is lost.”

The authors refer to a study in their rebuttal that suggests mantle cells are capable of producing all neuromast cell classes in a regenerating fin. Do the authors predict this would be the case, if they ablated everything but the mantle cells? If not, why?

As stated above, we believe that this may well be the case, and have added the above paragraph in the Discussion to specifically address this point. Because this possibility is largely based on previous work, we have included the relevant references: Dufourcq et al., 2006; Steiner et al., 2014; Romero-Carvajal et al., 2015.

2) The authors discussed the potential influence of interneuromast cells on planar polarity, but did not elaborate influence of INCs on neuromast size (Figure 4K-P). The result is striking and should be discussed further than the statement in the Results-"non-essential, yet appreciable contribution to regeneration".

Please, note that this experiment equates size with cell count. We have added language to clarify this point and address the reviewer’s comment:

“We find that interneuromast cells are not essential for neuromast regeneration because severely damaged organs recover all cell classes in the appropriate localization in the absence of interneuromast cells. However, we systematically observed smaller organs when interneuromast cells where ablated. These observations suggest that these peripheral cells may yet help regeneration, either directly by contributing progeny, or by producing mitogenic signals to neuromast-resident cells”.

Reviewer #2:

In the new version of this manuscript, Viader-Llargues, Lopez-Schier, and colleagues present their results with a more clear aim, focusing on quantitative aspects of the regeneration response and new tools to investigate it. My main points were properly addressed, namely toning down certain conclusions, including alternative scenarios and focusing on defined topics rather than superficially following many. I appreciate particularly the scientific quality and dedication of the authors' replies to many of my previous concerns. The lack of overshooting of cell types during the regeneration response was a very nice addition that contrasts to other regenerative systems, and illustrates how controlled the regenerative response is in neuromasts.

These are my comments on the current version:

In the title and along the manuscript the authors use the term "cell-phylogeny tracing". Is there any reason not to use the widely accepted term "lineage" instead? I feel that "lineage" works better in this context since it reflects more accurately the continuum of the data that the authors have acquired.

Under a standard definition of lineage and phylogeny, it is true that what we have traced is cellular lineages: “A lineage is a series of cells that can be connected directly by a continuous line of descent from a primordial progenitor”.

We agree with the reviewer that lineage is an optimal word given its widely accepted meaning in the field. Yet, we chose the word phylogeny because we have also shown the relationship between the lineages of several initial primordial progenitors (founder cells). Having said this, however, we understand that “phylogeny” may generate confusion among the readership and accept the critique. Thus, we have replaced phylogeny with lineage throughout the text, including the title.

The issue of self-regulation of the regenerative response is dealt more carefully than in the previous version. Still, I would like the authors to be more explicit about their interpretation of the data, mainly on the Discussion. One of the aspects the authors focus on is the re-establishment of a polarity axis, and this is a very interesting aspect of the regeneration response since polarity is heterogeneous among different neuromasts of the posterior lateral line – unlike cell types and their distribution. In other words: neuromasts have one solution for the cell-type problem, but two solutions for the polarity problem. Previous work from the group (Lopez-Schier et al., 2004 and Lopez-Schier and Hudspeth, 2006) has clearly stated that the polarity axis is related to the final migration of the neuromasts during organogenesis. I find troubles imagining how a self-renewing organ (or 4-10 cells, to put it bluntly) will always choose the original polarity in the absence of external cues. I agree with the authors that external cues have minor roles in the establishment of cells types or position of cells, but extending self-organization to the re-establishment of polarity seems inappropriate to me. Am I missing something? Maybe the authors could speculate about the role of the afferent axons (I guess they should remain to some extent under the injury paradigms used in this study), which they have shown to display a high accuracy recognizing hair cells of a given polarity (Pujol-Marti, Current Biology 2014).

The reviewer has identified one of the most mysterious and, in our opinion, fascinating problem of architectural recovery during organ regeneration. The collective evidence from previous work of our group and others, and evidence that we present here indicates that planar polarity has two components: the local coherent orientation of hair cells along a single axis (in horizontal neuromasts either rostrally or caudally), and the global orientation of this axis relative the main body axes of the fish. The local coherent orientation involves the core planar polarity pathway (López-Schier et al., 2004), as well as a segregated activity of the Emx2 transcription factor (Jiang et al., 2017). We show that the local coherent orientation of hair cells is self-regulatory. We do not know for certain what controls the global orientation of the hair cells, other than that it is initially determined by the direction of movement of the lateral-line primordium (López-Schier et al., 2004). We have always stated that architectural repair (including planar polarity) is achieved with “minimal” extrinsic information (Abstract). We do argue that self-organization is an optimal morphogenetic process to govern organ repair (Discussion), but discuss where and when self-organization may occur, make a clear distinction between self-organization and self-assembly, and do not indicate that a purely self-organizing process is at play. Therefore, we now clarify that when referring to planar polarity we exclusively focus on the local coherent polarization of the hair cells, and further state that:

“We would like to highlight that we do not currently understand the global polarization of the neuromast epithelium relative to the main body axes of the animal. External sources of polarity may impinge in the recovery of these global axes during neuromast regeneration. Previous work has demonstrated that local and global polarization occur independently of innervation López-Schier and Hudspeth, 2006), but other potential polarizing cues remain untested. Therefore, at present we can only support the notion that local coherent polarity is self-organizing, whereas global orientation may be controlled externally.”

I particularly enjoyed the reply of the authors regarding the differences and similarities in mantle cell behavior during homeostasis and regeneration. I think it would be an added value if they incorporate these concepts in the Discussion.

We are glad to read this. Reviewers 1 and 2 have expressed a similar feeling. Thus, as stated above, we now write in the Discussion:

“The behavior of the mantle cells is especially intriguing. Complete elimination of parts of the lateral line by tail-fin amputation have revealed that mantle cells are able to proliferate and generate a new primordium that migrates into the regenerated fin to produce new neuromasts (Dufourcq et al., 2006). This observation can be interpreted as suggesting that under some injury conditions, mantle cells are capable of producing all the cell classes of a neuromast. Transcriptomic profiling of mantle cells following neuromast injury revealed that these cells up-regulate the expression of multiple genes (Steiner et al., 2014). Furthermore, a recent study has revealed that mantle cells constitute a quiescent pool of cells that re-enters cell cycle only in response to severe depletion of sustentacular cells (Romero-Carvajal et al., 2015), suggesting that these cells may conform a stem-cell niche for proliferation of sustentacular cells. Thus, the collective evidence indicates that the mantle cells respond to damage and contribute to the regenerative processes, and may drive the regeneration of an entire organ if every other cell class is lost.”

Results section, paragraph two; I believe that the authors should include a reference to Grant et al., 2005. In fact, the authors do so in their response to Reviewer #1.

This is correct. We have added the reference.

Subsection “Sustentacular cells are equipotent and plastic”. I found this part difficult to understand. Either the written numbers do not match what the authors show between brackets, or there is some mistake in the annotation. What is "peak at 8h, mean+-s.d at 14+-9 hours"? Overall, I find this part the less clearly written of the manuscript. Also, is it valid to call a peak sharp with such a big s.d.?

We see how the previous expression of this result may be difficult to understand. We now state it more clearly as:

“Cell-cycle length in the 1st generation peaks around 10 hours (9.8 ± 3.3h, median ± interquartile range (iqr)) (Figure 7C), but it begins to increase and to vary in the 2nd generation (11.5 ± 7.3h, median ± iqr), and more so in the 3rd generation (18.8 ± 20.3h, median ± iqr)”.

Discussion paragraph one: The authors state that sustentacular cells are equipotent, which I think is an overstating.

The data presented in the manuscript reveals that 4 or more of them are enough to regenerate the entire organ. The use of equipotency, in my view, states that all of them do the same during the process. In their accurate data in Figure 7, they show that some of them generate M cells, some other S cells, and some other combinations of H and other cell types. I understand that equipotent members of a population could behave in different manners based on either stochastic internal programs or external cues, but to prove which is the case demands a more extensive dataset and a deep mathematic analysis, which is not the scope of the present study – although this is a fantastic system to tackle it! I feel that the use of equipotent in this context assumes features that were not tested experimentally. I suggest the authors stay with "tri-potent", a term that they have used along Results.

This is correct. We have deleted the term equipotent and use now tri-potent.

Reviewer #3:

This is a revised version of a manuscript by Oriol Viader-Llargués et al. that deals with cellular mechanisms of neuromast regeneration. The previous version was criticized by its descriptive nature, lack of focus and omission of some references. The authors largely remedied these issues, although I still feel that some of their findings are not sufficiently discussed in the context of previous studies.

Comments:

The authors mention that their findings contrast those of Sanchez et al. study, but do now offer any discussion as to why that is the case. I think it is important to offer at least some explanation.

We believe that these discrepancies may be a result of differences in the experimental approaches and/or the markers used to visualize cells. Another possible explanation is the age of the animals that were used in either study. Sánchez et al., used 3dpf larvae, whereas all our ablations were performed in older animals. These differences, alone or combined, may account for the differences in the outcome or interpretation of the two studies. Recognizing this, on pp.7 we do mention that differences in age may explain the differences between both studies. However, we now state more explicitly in the Discussion:

“We show that the complete elimination of a neuromast is irreversible in larval zebrafish. However, Sánchez and colleagues have previously reported that interneuromast cells can generate new neuromasts (Sánchez, 2016). By assaying DNA synthesis prior to mitosis, we show that interneuromast cells do not proliferate after neuromast ablation. These differences may be explained by differences in ablation protocols (electroablation versus laser-mediated cell killing), the age of the specimens (embryos versus early larva) or the markers used to assess cellular elimination.”

It seem that examples of severe ablation following full regeneration always leave at least one mantle cell (“Neuromast architecture recovers after severe loss of tissue integrity”). If this is indeed the case, it may indicate that a combinatorial signal(s) is required to initiate organ regeneration.

This is correct. Yet, by analyzing neuromasts completely devoid of mantle cells (Figure 3), we show that neuromast regenerate to a normal status. Therefore, we do not think that combinatorial signaling between mantle and sustentacular cells is essential for proper repair.

There are only two cases when sustentacular cell progenitors gave rise to all three cell types. Based on these small numbers, I think the authors need to be careful concluding that a sustentacular cell is a multipotent progenitor for all cell types in the neuromast.

This is correct. Our current and previous work show that sustentacular cells can re-generate and also generate hair cells. We show that they can also give rise to mantle cells. Thus, the sustentacular-cell population is tri-potent. The question is whether each individual sustentacular cell is also tri-potent. Our clonal analysis reveals that this is the case, and makes us believe that the fact that some events are rare does not invalidate the tri-potentiality idea. We have been careful to restrict our conclusions to the available data and trust that we had not overstated our claims. Nevertheless, we have now changed a subtitle to “The sustentacular-cell population is tri-potent and plastic”.

The finding that at least 4 cells are needed to reconstitute an organ is intriguing. This is reminiscent of planarian organ regeneration where 1/300th part but not less can reconstitute full animal. It was later discovered that this is a minimum fraction of the animal roughly containing at least one stem cell (neoblast) necessary to regenerate all tissue lineages. Again, some discussion, as to why the authors think this is case (4 cells but not fewer are required) is warranted

This is indeed very interesting. We have found technically challenging to leave less than 4 cells in a consistent manner without eliminating the entire neuromast. Therefore, we cannot rule out the possibility that as little as a single founder cells may be able to regenerate an entire organ. We have included a sentence in the Discussion about this point

“It is technically challenging to consistently maintain fewer than 4 cells in toto without eliminating the entire neuromast. Thus, we cannot rule out the possibility that a single founder cell may be able to regenerate a neuromast”.

Discussion paragraph two, final sentence: Seleit et al., 2017 showed that "new cell type" exists in zebrafish. They also showed that mantle cell are neuromasts stem cells. Thus, it is important to discuss how homeostatic cell renewal differs from cell renewal during NM regeneration (i.e. mantle cell as stem cell during homeostasis vs. sustentacular cell as a multipotent progenitor during regeneration). This is an interesting question, as there are examples where cells can be driven to change their fate by extreme injury paradigm.

This is indeed very interesting. Seleit et al., 2017 have revealed a "new cell type" in Medaka and zebrafish neuromasts. As stated for reviewers 1 and 2, we have now added a section in the Discussion with a thorough explanation of mantle cell behaviors reported in the literature and how they compare with our findings.

https://doi.org/10.7554/eLife.30823.021

Article and author information

Author details

  1. Oriol Viader-Llargués

    1. Unit Sensory Biology & Organogenesis, Helmholtz Zentrum München, Neuherberg, Germany
    2. Laboratory of Sensory Cell Biology, Centre for Genomic Regulation, Barcelona, Spain
    Contribution
    Data curation, Formal analysis, Validation, Investigation, Visualization, Methodology, Writing—original draft, Writing—review and editing
    Competing interests
    No competing interests declared
  2. Valerio Lupperger

    Institute of Computational Biology, Helmholtz Zentrum München, Neuherberg, Germany
    Contribution
    Resources, Software, Methodology, Writing—review and editing
    Competing interests
    No competing interests declared
  3. Laura Pola-Morell

    1. Unit Sensory Biology & Organogenesis, Helmholtz Zentrum München, Neuherberg, Germany
    2. Laboratory of Sensory Cell Biology, Centre for Genomic Regulation, Barcelona, Spain
    Contribution
    Resources, Investigation, Writing—review and editing
    Competing interests
    No competing interests declared
  4. Carsten Marr

    Institute of Computational Biology, Helmholtz Zentrum München, Neuherberg, Germany
    Contribution
    Software, Formal analysis, Supervision, Methodology, Writing—review and editing
    For correspondence
    carsten.marr@helmholtz-muenchen.de
    Competing interests
    No competing interests declared
  5. Hernán López-Schier

    1. Unit Sensory Biology & Organogenesis, Helmholtz Zentrum München, Neuherberg, Germany
    2. Laboratory of Sensory Cell Biology, Centre for Genomic Regulation, Barcelona, Spain
    Contribution
    Conceptualization, Supervision, Funding acquisition, Visualization, Writing—original draft, Project administration, Writing—review and editing
    For correspondence
    hernan.lopez-schier@helmholtz-muenchen.de
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0001-7925-7439

Funding

European Research Council (2007_205095)

  • Hernán López-Schier

Agència de Gestió d’Ajuts Universitaris i de Recerca (2009-SGR-305)

  • Hernán López-Schier

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

We thank A Steiner, T Nicolson and L Solnica-Krezel for transgenic zebrafish, the animal facility personnel at the CRG of Barcelona and the HMGU for animal care, and Kirill Smirnov for statistical support. Funding was provided by the European Research Council Grant 2007_205095, by the ESF Research Networking Programme ‘QuanTissue’, and by the AGAUR Grant 2009-SGR-305 of Spain.

Ethics

Animal experimentation: Zebrafish were maintained under standard conditions. Experiments with wild-type, mutant and transgenic embryos of undetermined sex were conducted in accordance with institutional guidelines and under a protocol approved by the Ethical Committee of Animal Experimentation of the Parc de Recerca Biomedica de Barcelona, Spain, and protocol number Gz.:55.2-1-54-2532-202-2014 by the "Regierung von Oberbayern", Germany.

Reviewing Editor

  1. Tanya T Whitfield, University of Sheffield, United Kingdom

Publication history

  1. Received: July 27, 2017
  2. Accepted: March 28, 2018
  3. Accepted Manuscript published: March 29, 2018 (version 1)
  4. Version of Record published: April 17, 2018 (version 2)
  5. Version of Record updated: April 18, 2018 (version 3)

Copyright

© 2018, Viader-Llargués et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 3,162
    Page views
  • 560
    Downloads
  • 3
    Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Download citations (links to download the citations from this article in formats compatible with various reference manager tools)

Open citations (links to open the citations from this article in various online reference manager services)

Further reading

    1. Genetics and Genomics
    2. Stem Cells and Regenerative Medicine
    Cheen Euong Ang et al.
    Research Article Updated
    1. Developmental Biology
    2. Stem Cells and Regenerative Medicine
    Kynan T Lawlor et al.
    Research Article Updated