Single cell RNA sequencing unravels the transcriptional network underlying zebrafish retina regeneration

Reviewer #1 (Public Review):

Muller glia function as retinal stem cells in the adult zebrafish retina. Following retinal injury, Muller glia are reprogramned (reactive Muller glia), and then divide to produce a progenitor that amplifies and differentiates into retinal neurons. Previous scRNAseq analysis used total retinal RNA from uninjured and injured retinas isolated at time points when Muller glia are quiescent, being reprogrammed, and proliferating to reveal genes and gene regulatory networks underlying these events (Hoang et al., 2020). The manuscript by Celotto et al., used double transgenic zebrafish that allow them to purify by FACS quiescent and reactive Muller glia, Muller glia-derived progenitors, and their differentiating progeny at different times post retinal damage. RNA from these cell populations was used in scRNAseq studies to identify the transcriptomes associated with these cell populations. Importantly, they report two quiescent and two reactive Muller glia populations. These results raise the interesting possibility that Muller glia are a heterogenous population whose members may exhibit different regenerative responses to retinal injury. However, without further experimentation, the validity and significance of this result remain unclear. In addition to putative Muller cell heterogeneity, Celotto et al., identified multiple progenitor classes, some of which are specified to regenerate specific retinal neuron types. Because of its focus on Muller glia and Muller glia-derived progenitors at mid to late stages of retina regeneration, this new scRNAseq data will be a useful resource to the research community for further interrogation of gene expression changes underlying retina regeneration.

Major concerns:

1. The identification of multiple populations of Muller glia, reactive Muller glia, and progenitors is interesting, but beyond a few in situ hybridization studies to validate injury-dependent gene inductions, there are no experiments that confirm that multiple cell populations exist in vivo, and no experiments examining the significance of these different populations in the regenerative process. It would be helpful to discuss how the peripheral to the central gradient of Muller cell maturation influences the scRNAseq-based cell clustering results.

2. While the reliance on transient GFP and mCherry expression may be sufficient, the final population used for the scRNAseq analysis is only partial in nature. Permanently marking the MG through a Cre-Lox system is more ideal. The authors mention the possibility of missing highly proliferative populations of MG/RPC through the dilution of fluorescent proteins; a transgenic system that allows for true lineage tracing may then capture more appropriate MG/RPC populations. The lack of gating for a pure GFP population also confounds this problem which the authors do mention in the discussions; this oversight was not explained.

3. Much time was taken to identify each cell cluster and to list the differentially expressed genes, but no functional significance for these genes was probed. While a lot of work has gone into the analysis shown, altering some of the MG/RPC trajectories through differentially expressed genes would go a long way to making this study more impactful.

4. The data presented in this paper has significant overlap with scRNAseq data presented by Hoang et al., 2020 in Science where Muller glia, reactive Muller glia, and Muller glia-derived progenitors were carefully analyzed. How does their data fit with the data presented here? The authors could have used that paper as a jumping-off point and offered more time points for comparison, especially as progenitors differentiate.

5. A major conclusion of the paper is that neurogenic progenitors in the injured retina differentiate into neurons with a similar order as that taking place during development. This analysis is based on two time points, and while the trends stay true to the authors' model, two time points are too few to make such a conclusion. In addition, because of the time points chosen for this analysis, many mature neuronal markers are lacking. Including additional time points so mature neuronal markers are detected in the dataset would enhance the trajectory proposed.

Reviewer #2 (Public Review):

In this publication, the authors provide a comprehensive trajectory of transcriptional changes in Müller glia cells (MG) in the regenerating retina of zebrafish. These resident glia cells of the retina can differentiate into all neural cell classes following injury, providing full regenerative capabilities of the zebrafish retina. The authors achieved this by using single-cell RNA sequencing of Müller glia, progenitors, and regenerated progeny, comparing uninjured and light-lesioned retinae.

The isolation strategy involves using two transgenic strains, one labelling dividing cells and their immediate progeny, and the other Müller glia cells. This allowed them to separate injury-induced proliferating and non-reactive Müller glia cells. Subsequent single-cell transcriptomics showed that MG could be non-reactive under both uninjured and lesioned conditions and reactive MG give rise to a cell population that both replenishes the pool of MG and replenish neurogenic retinal precursor cells. These precursor cells produce regenerated neurons in a developmental time series with ganglion cells being born first and bipolar cells being born last. Interestingly hybrid populations have been detected that co-share characteristics of photoreceptor precursors and reactive glia.

This is the first study of its kind following the dynamic changes of transcriptional changes during retinal regeneration, providing a rich data source of genes involved in regeneration. Their finding of transcriptionally separable MG populations is intriguing.

This study focuses on the light-lesioned retina and leaves open the question if the observed transcriptional trajectories of regenerating neurons are generalizable to other lesion models (e.g. chemical or mutational lesions) or are specific to the light-damaged retina.