Single cell RNA-Seq reveals distinct stem cell populations that drive sensory hair cell regeneration in response to loss of Fgf and Notch signaling

Mark E. Lush, Daniel C. Diaz, Nina Koenecke, Sungmin Baek, Helena Boldt, Madeleine K. St. Peter, Tatiana Gaitan-Escudero, Andres Romero-Carvajal, Elisabeth Busch-Nentwich, Anoja Perera, Kate Hall, Allison Peak, Jeffrey S. Haug, Tatjana Piotrowski

Preprint posted on December 14, 2018


Distinct progenitor populations mediate regeneration in the zebrafish lateral line.

Eric D Thomas, David Raible

Preprint posted on November 29, 2018

Article now published in eLife at

How to keep an organ ticking? Two preprints delve into the molecular and functional heterogeneity within the progenitor cell populations that maintain the zebrafish sensory hair cells, and forms the basis of hair cell regeneration.

Selected by Rudra Nayan Das


Fishes have been long recognized for their regenerative abilities, and zebrafish has emerged as a highly desirable model for vertebrate regeneration. Among the several organ systems that show regenerative capabilities, the lateral line sense organ is one of the well-characterized systems that is experimentally amenable because of its superficial location and a relatively simple architecture.

The lateral line system is comprised of discrete mechanosensory organs for the fish to sense changes in water current. Each organ contains a cluster of centrally located sensory hair cells, surrounded by support cells and peripheral mantle cells. While the zebrafish sensory hair cells show regeneration upon injury, their mammalian counterpart in the inner ear is incapable of regeneration, and in many cases, hair cell loss in mammals leads to loss of hearing and vestibular function.

Prior work in lateral line organ has established that the sensory hair cells are derived from the surrounding non-sensory support cells. However, the support cells display non-uniform regenerative potential which indicates the existence of support cell sub-populations. To understand support cell heterogeneity, Lush et al. and Thomas and Raible took two very different experimental approaches. While Lush et al. used scRNA-Seq for a deeper dive into the molecular architecture of the different sense organ populations, Thomas and Raible focussed on functional demonstration of differential regenerative capabilities of three different support cell populations.

Important results

Using scRNA-Seq, Lush et al. could show 14 cell clusters in the 5-day old larval sense organs. The restricted expression of some of the cluster gene markers were validated using in situ hybridization. The 14 clusters included two clusters of sensory hair cells, two of mantle cells and ten of support cells.

Under normal homeostatic conditions, at any given point, the sense organ is comprised of cells at various stages of differentiation. The scRNA-Seq analysis allowed identification and clustering of cells at various stages of fate/state transition. Alignment of cells in the predicted path of progressive differential state allowed for the identification of gene expression dynamics along this path.

Lush et. al. also predicted the existence of two major lineages of cycling cells within the sense organs. The peripherally located cells, adjacent to the mantle cells, are more important in mediating amplifying divisions that maintain the support cell populations. The more centrally located support cells showed features of differentiating divisions that give rise to the sensory cells. Furthermore, the authors demonstrated the importance of FGF signaling in the support cell populations and how FGF and Notch pathways, independently, regulate Wnt signaling to control differentiating divisions of the central support cells.

(Left) t-SNE plot depicting the 14 cell clusters identified through scRNA-seq. (Figure 1D from Lust et. al)
(Right) Schematic summarizing the roles of different signaling pathway in the homeostatic neuromast. (Figure S10 from Lush et al.)


Thomas and Raible took advantage of a CRISPR-mediated insertional screen that labels different population of cells based on the expression of the targeted gene (the screen was not described in this manuscript). Three genes were identified – sfrp1a, tnfsf10l3 and sost, that are differentially expressed within subsets of the support cells.

By generating transgenic lines for lineage tracing or cell-specific ablation, the authors found that the sost-expressing ‘Dorsal and ventral support cells’ (DV cells) were more likely to contribute towards the replacement of sensory cells both under homeostatic conditions and during regeneration. Selective ablation of these sost-expressing cells significantly decreased hair cell regeneration. Interestingly, DV cell ablation also drove other support cells to replace the lost DV cells. This also included the sfrp1a-expressing peripheral cells that are generally non-responsive to the sensory cell ablation.

Furthermore, the extent of regulation of support cells’ regeneration capabilities through Notch signaling was found to be different for different sub-population of support cells.


Why I chose these papers

The two highlighted preprints represent different approaches to define sub-populations within the progenitor populations of the lateral line sense organs.

Utilizing the entire transcriptome, Lush et al. has given us a very comprehensive definition of sense organ sub-population. Their scRNA-Seq dataset is expected to be beneficial for the entire community interested in related questions. Thomas and Raible’s work neatly provides functional demonstration of the regenerative capabilities of different progenitor populations. Furthermore, both the groups have shown evidence of how different population within the progenitors are differentially regulated by different signaling systems. These principles of progenitor population regulation are likely to be of general value for researchers interested in other model systems of regeneration and maintenance.

Future Directions

The support cell populations and their transcriptome has been defined for 5-day old zebrafish larvae. As the fish grows, several changes occur in its physiology. It would be interesting to investigate whether the sub-populations and transcriptome of the organ remain comparable with the developmental stage described in these papers. It is possible that some of the sub-population/gene expression signatures may be features of immature sense organs that may cease to persist in juvenile and the adult animal. This might also have consequences in the regenerative potential of the mature animal.

Another exciting line of query was proposed by Lush et al. Comparative analysis of the mammalian support cell transcriptome with its fish counterpart can reveal important differences that restrict mammalian hair cell regeneration.

Since there are different regenerative potentials and parallel pathways operating in the support cell populations, it might be interesting to figure out whether different external cues or contexts utilize distinct sub-populations or signaling pathways for hair cell proliferation.

Questions to the authors

  1. Lush et al. report a number of lineage relationship among the clusters through the scRNA-Seq dataset analysis, which may not always be conclusive. Do the authors feel it would be productive to embark upon lineage analysis experiments to solidify their conclusions?
  2. Thomas and Raible use sost to define DV cells, which perhaps corresponds to the cluster 5, 6, 11 and 3 of the Lush et al. study (Fig. 3B). The sost-expressing cells in Thomas and Raible’s study do display differentiating division, while sost is mainly shown as a feature of amplifying lineage by Lush et al. Can both groups, in the light of each other’s findings, comment on this?


Posted on: 18th January 2019 , updated on: 21st February 2019

(1 votes)

  • Author's response

    Tatjana Piotrowski shared about Single cell RNA-Seq reveals distinct stem cell populations that drive sensory hair cell regeneration in response to loss of Fgf and Notch signaling

    Dear Rudra,

    Thank you very much for highlighting our work! I am happy to answer the questions that you posed. As you mentioned, scRNASeq data does not provide evidence for cell lineages but it generates hypotheses that need to be experimentally confirmed. In the case of our study, we already possessed lineage information from time lapse analyses (Romero-Carvajal et al., 2015), as well as lineage analyses from other groups (Seleit et al., 2017; Viader-Llargues et al., 2018).

    For example, lineage tracing of mantle cells in medaka revealed that mantle cells give rise to support and hair cells and constitute long term stem cells (Seleit et al., 2017). Even though killing hair cells with neomycin treatment does not trigger proliferation in mantle cells more severe injury does, suggesting supporting the hypothesis that mantle cells are quiescent stem cells (Romero-Carvajal, 2015).

    Our time lapse analyses determined that central support cells give rise to hair cells (Romero-Carvajal et al., 2015, also see Figure 3K, Fig. S4, Suppl. Movie 1) but also that A-P support cells give rise to hair cells (Fig. 2G). We also found that the vast majority of amplifying cells is located immediately next to mantle cells, divides symmetrically and gives rise to two more support cells (Romero-Carvajal, 2015). Only when these daughter cells are displaced centrally they differentiate into hair cells (Fig. 2E).

    Based on these data, I believe that we have a fairly good understanding of the lineage relationships in regenerating neuromasts, however additional lineage experiments are warranted to determine how mutations, or experimental manipulations affect these lineages, as reported by Thomas and Raible.


    In your second question you suggest that we consider sost expression as a marker for amplifying cells. However, that interpretation is not quite correct. In the manuscript we describe that sost and other polar genes are expressed in more cells than just the ones immediately adjacent to mantle cells and that we did not identify genes specifically expressed in amplifying cells. sost labels the D/V compartments of neuromasts (see pink cells in the schematic that you posted above), including amplifying support cells that reside next to mantle cells (in red). Thus, sost labels the amplifying cells, as well as more centrally located support cells that differentiate into hair cells. It is therefore not a marker of amplifying cells.


    Eric Thomas shared about Distinct progenitor populations mediate regeneration in the zebrafish lateral line.

    Hello Rudra,

    Thank you for taking an interest in our work, and for the excellent write-up! Regarding your second question, while our sost:nlsEos cells do generate the majority of new hair cells, they are also capable of generating new sost:NTR-GFP cells following their ablation with metronidazole (Fig. 11). Thus, these cells can serve as both hair cell progenitors as well as support cell progenitors (performing what Lush et al refer to as differentiating and amplifying divisions, respectively). As Dr. Piotrowski noted in her response, sost expression is not limited to the amplifying lineages, so these data are consistent with ours.

    It is also important to note that our sost lines may be labeling multiple cell types. We demonstrated that the sost:NTR-GFP and sost:nlsEos populations do not completely overlap, with a number of cells expressing only nlsEos as the neuromast matures (Fig. 5). Our ablation data suggest that it is this subset of the sost:nlsEos population that serves as hair cell progenitors (Fig. 6). These sost:nlsEos only cells are located more centrally within the neuromast than sost:NTR-GFP cells. It is possible that this subset of sost:nlsEos cells are not actually expressing sost; because nlsEos persists for a much longer time than GFP, these cells could still retain nlsEos even after downregulating sost. If this is true, then the sost:NTR-GFP cells would more accurately reflect sost gene expression. We have classified these cells only expressing sost:nlsEos as “mature DV hair cell progenitors” and the cells expressing both sost:nlsEos and sost:NTR-GFP as “immature DV hair cell progenitors” (Fig. 12). This transition from immature to mature progenitors suggests that support cells may need to downregulate sost before serving as hair cell progenitors (thus turning off NTR-GFP but retaining nlsEos). These data are consistent with previous work from Viader-Llargués et al. (eLife 2018;7:e30823) suggesting that radial position is predictive of whether cells will give rise to hair cells or support cells Lush et al have also hypothesized this based on their lineage studies, so our data are once again consistent.




    Have your say

    Your email address will not be published. Required fields are marked *

    This site uses Akismet to reduce spam. Learn how your comment data is processed.

    Sign up to customise the site to your preferences and to receive alerts

    Register here