Master control genes in the regeneration of rod photoreceptors from endogenous progenitor cells in zebrafish retina

Reviewer #1 (Public review):

Summary:

Shihabeddin et al. used bioinformatic and molecular biology tools to study the unique regeneration of rod photoreceptors in a zebrafish model. The authors identified a few transcription factors that seem to play an important role in this process.

Strengths:

This manuscript is well prepared. The topic of this study is an interesting and important one. Bioinformatics clues are interesting.

Weaknesses:

Considering the importance of the mechanism, the knockdown experiments require further validation. The authors over-emphasized this study's relevance to RP disease (i.e. patients and mammals are not capable of regeneration like zebrafish). They under-explained this regeneration's relevance or difference to normal developmental process, which is pretty much conserved in evolution.

Reviewer #2 (Public review):

This is an interesting and important work from Shihabeddin et al, to identify master regulators for rod photoreceptor regenerations in a zebrafish model of Retinitis Pigmentosa. Building on their scRNA-seq data, Shihabeddin et al dissected the progenitor cell types and performed trajectory analyses to predict transcription factors that apparently drive the progenitor proliferation and differentiation into rod photoreceptors. Their analyses predicted e2f1, e2f2, and e2f3 as critical drivers of progenitor proliferation, Prdm1a as a driver of rod photoreceptor differentiation, and SP1 as a driver of rod photoreceptor maturation. Genetic experiments provide clear support for the roles of e2fs in progenitor proliferation. It's also apparent from Figure 8 that prdm1 knockdown appears to cause a decrease in rhodopsin expression. By colocalizing BrdU and Retp1, the authors inferred that the apparent "new rods" (which exhibit mixed BrdU and Retp1 signal) are decreased with prdm1, providing further support. Overall I found the work to be interesting, rigorous, and informative for the community.

I have a few suggestions for the authors to consider:

(1) Perhaps the authors can consider explaining why the Prdm1a knock-down cells would have a higher Retp1 signal per cell in Fig 9B. Is this a representative picture? This appears to contradict Figure 8's conclusion, although I could tell that the number of Retp1+ cells in the ONL appears to be lower.

(2) The authors noted "Surprisingly, the knockdown of prdm1a resulted in a significantly higher number of rhodopsin-positive cells in the INL (p=0.0293)", while it appears in Figure 9B, 9C that the difference is 2 cells vs 0 in a rightly broader field. It seems to be too strong of a statement for this effect.

(3) It appears to this reviewer that the proteomic data didn't reveal much in line with the overall hypothesis or the mechanism, and it's unclear why the authors went for proteomics rather than bulk RNA-seq or ChIP-seq for a transcription factor knock-down experiment. Overall this is a minor point.

Reviewer #3 (Public review):

Summary:

This study uses a combination of single-cell RNA-Seq to globally profile changes in gene expression in adult P23H transgenic zebrafish, which show progressive rod photoreceptor degeneration, along with age-matched controls. As expected, mitotically active retinal progenitors are identified in both conditions, the increased number of both progenitors and immature rods are observed. DrivAER-mediated gene regulatory network analysis in retinal progenitors, photoreceptor precursors, and mature rod photoreceptors respectively identified e2f1-3, prdm1a, and sp1 as top predicted transcriptional regulators of gene expression specific to these cell types. Finally, morpholino-mediated knockdown of these transcription factors led to expected defects in proliferation and rod differentiation.

Strengths:

Overall, this is a rigorous study that is convincingly executed and well-written. The data presented here will be a useful addition to existing single-cell RNA-Seq datasets obtained from regenerating zebrafish retina.

Weaknesses:

Multiple similar studies have been published and it is something of a missed opportunity in terms of identifying novel mechanisms of rod photoreceptor regeneration. Several other recent studies have used both single-cell RNA and ATAC-Seq to analyze gene regulatory networks that regulate neurogenesis in zebrafish retina following acute photoreceptor damage (Hoang, et al. 2020; Celloto, et al. 2023; Lyu, et al. 2023; Veen, et al 2023) or in other genetic models of progressive photoreceptor dystrophy such cep290 mutants (Fogerty, et al. 2022).

The gene regulatory network analysis here would also benefit from the addition of matched scATAC-Seq data, which would allow the use of more powerful tools such as Scenic+ (Bravo and de Winter, et al. 2023). It would also benefit from integration with single-cell multiome data from developing retinas (Lyu, et al. 2023). The genes selected for functional analysis here are all either robustly expressed in retinal progenitor cells (ef1-3 and aurka) or in developing rods (prdm1a), so it is not really surprising that defects are observed. Identification of factors that selectively regulate rod photoreceptor regeneration, rather than those that regulate both development and regeneration, would provide additional novelty. This would also potentially allow the use of animal mutants for candidate genes, rather than exclusively relying on morphant analysis, which may have off-target effects.

The description of the time points analyzed is vague, stating only that "fish from 6 to 12 months of age were analyzed". Since photoreceptor degeneration is progressive, it is unclear how progenitor behavior changes over time, or how the gene expression profile of other cell types such as microglia, cones, or surviving rods is altered by disease progression. Most similar studies address this by analyzing multiple time points from specific ages or times post-injury.

Author response:

Reviewer 1: “The authors over-emphasized this study's relevance to RP disease (i.e. patients and mammals are not capable of regeneration like zebrafish).”

It is true that humans and other mammals are not capable of regeneration. This is why we and many other groups study zebrafish to identify mechanisms of regeneration that successfully form new rods. That said, our previous paper on the molecular basis or retinal remodeling in this zebrafish model system (Santhanam et al., 2023; Cell Mol Life Sci. 2023;80(12):362) revealed remarkable similarities in the stress and physiological responses of rods, cones, RPE and inner retinal neurons to those in mammalian RP models. Thus, we believe this zebrafish is an adequate model of RP and an excellent model to study rod regeneration.

Reviewer 1: “They under-explained this regeneration's relevance or difference to normal developmental process, which is pretty much conserved in evolution.” and:

Reviewer 3: “It would also benefit from integration with single-cell multiome data from developing retinas (Lyu, et al. 2023).”

It is an excellent suggestion to compare the regenerative response we have studied in a chronic degeneration/regeneration model to the trajectory of developmental rod formation. In Lyu, et at. 2023, it was found that while retinal regeneration has similarities to retinal development, it does not precisely recapitulate the same transcription factors and processes. Any differences between this trajectory and that revealed in developmental studies would be enlightening. We intend to do such analyses to add to a revised manuscript in the future.

Reviewer 2: “Perhaps the authors can consider explaining why the Prdm1a knock-down cells would have a higher Retp1 signal per cell in Fig 9B. Is this a representative picture? This appears to contradict Figure 8's conclusion, although I could tell that the number of Retp1+ cells in the ONL appears to be lower.”

These are different experimental paradigms. Figure 8 shows knockdown 48 hours after injection, at which time prdm1a knockdown is affecting rhodopsin expression directly. That experiment investigated whether prdm1a knockdown affected progenitor proliferation. Figure 9 shows a time point 6 days after injection, at which time we were asking if prdm1a knockdown affected differentiation of progenitors into rods.

Reviewer 2: “The authors noted "Surprisingly, the knockdown of prdm1a resulted in a significantly higher number of rhodopsin-positive cells in the INL (p=0.0293)", while it appears in Figure 9B, 9C that the difference is 2 cells vs 0 in a rightly broader field. It seems to be too strong of a statement for this effect.”

This was a very unexpected finding. We included statistics (Figure 9D) to support the finding, so we don’t think it is too strong a statement to make. Speculation as to what might cause this is fascinating. Are Muller cells producing progenitors that fail to migrate to the ONL before differentiating into rods? The lack of BrdU labeling does not support this idea. Do neurogenic progenitor cells in the INL differentiate towards rods via a pathway that does not require prdm1a? Perhaps. Perhaps there are other explanations.

Reviewer 2: “It appears to this reviewer that the proteomic data didn't reveal much in line with the overall hypothesis or the mechanism, and it's unclear why the authors went for proteomics rather than bulk RNA-seq or ChIP-seq for a transcription factor knock-down experiment. Overall this is a minor point.”

We agree that bulk RNA sequencing would provide a similar answer, possibly with greater sensitivity. We chose proteomics for two reasons: 1) We wanted an independent assessment of the knockdown effects that could evaluate whether the knockdowns worked and what pathways were affected. Since our pathway comparison is to single cell RNAseq data, bulk RNA seq did not seem to be fully independent. 2) Because we used translation-blocking antisense oligos for most knockdown experiments, we did not expect the transcript abundance of the targeted gene to be affected, although these oligos can lead to target transcript degradation. Thus, we were not likely to be able to validate that our knockdown worked with this technique.

Reviewer 3: “The gene regulatory network analysis here would also benefit from the addition of matched scATAC-Seq data, …”

This is certainly true, and the reviewer points to several studies that have made excellent use of this strategy. Given the 1-2 year timeline to obtain and analyze such data, it is unlikely that we will be able to incorporate such data in our revised manuscript, but we hope to do so for follow-up studies.

Reviewer 3: “The description of the time points analyzed is vague, stating only that "fish from 6 to 12 months of age were analyzed". Since photoreceptor degeneration is progressive, it is unclear how progenitor behavior changes over time, or how the gene expression profile of other cell types such as microglia, cones, or surviving rods is altered by disease progression.”

We have shown in a previous study (Santhanam et al. Cells. 2020;9(10)) that rod degeneration and regeneration are in a steady state from at least 4 to 8 months of age, and in other experiments in the lab at least to 12 months of age. In this age range, regeneration keeps up with the pace of degeneration, both of which are very fast. This encompasses the cell types that we specifically study in this manuscript. The reviewer is right that other cell types could undergo changes. This is a separate topic of study in the lab.