A single-cell transcriptomic atlas of inner ear morphogenesis in zebrafish

Reviewer #1 (Public review):

Summary:

The authors dissected the ears with some surrounding tissue from 600 embryos at 4 developmental time points of wild-type larvae, as well as from an lmx1bb mutant, performed scRNA-seq analyses, and subclustered the ear/neuromast clusters. They identified cluster markers and performed PAGA pseudotime analyses to build developmental timelines of lineages. They validated some of the cluster markers with HCRs. Many of the clusters are not annotated in detail, but the data sets are still valuable for the community.

Strengths:

Using scRNA-Seq, the authors identified cluster markers for tissues of the developing zebrafish ear and validated some of them with HCRs. The data they compiled and submitted to public databases is a valuable resource for the community.

Weaknesses:

Many of the clusters have not been annotated or rely on published data. For the ones for which no HCRs or UMAPs are shown, it is therefore difficult to estimate which of the markers are indeed the most cell type/state-specific ones.

Major comments:

(1) It would be very useful if the cluster numbers in the Excel files also had the associated cell type annotations as a second column (at least for the ones that are known). E.g., in Supplemental Table 2, the text states which clusters represent which neuromast and ear cell type, but these are not mentioned in the Excel table.

(2) Many of the clusters have not been annotated or rely on published data. For the ones for which no HCRs or UMAPs are shown, it is therefore difficult to estimate which of the markers are indeed the most cell-type/state-specific ones.

(3) Uploading the data to gEAR (https://umgear.org/dataset_explorer.html), a web-based, publicly available ear database, would further increase the usefulness of this study to the broader community.

Method:

The authors should provide the details about how many cells were sequenced for each ear developmental stage, how many cells were present per cluster (page 8), and how many cells were present in each subcluster of ear and lateral line clusters (page 10).

Reviewer #2 (Public review):

Summary:

Munjal and colleagues present a single-cell RNAseq atlas of otic tissue at 4 developmental stages, generate coarse-grained PAGA graphs to describe the development of various otic cell types, rigorously validate their scRNAseq annotations using fluorescent in situ hybridization, and identify changes in epcam expression in lmx1bb mutants that potentially cause the dramatic defects in otic vesicle formation in these mutants.

Strengths:

The data set is very nice, and the annotations are extremely rigorous and more in-depth than other datasets that include these tissues, since these investigators have enriched significantly for this tissue of interest. Their use of PAGA to identify potential developmental relationships within the data is rigorous. I also would like to specifically point out how incredibly gorgeous the microscopy of the lmx1bb phenotype is in Figure 7. Wow.

Weaknesses:

A missed opportunity is that the authors describe creating an additional scRNAseq dataset from lmx1bb mutants, but do not show any comparative scRNAseq analyses that would identify broader sets of differentially expressed genes. It seems almost as if a key element of the study was removed at the last minute, and as a result, the discussion of changes in epcam expression in lmx1bb mutants in Figure 7 seems somewhat tacked onto the end of the study and not motivated by the analyses presented in the manuscript.

Overall, I do not think this study requires any major revisions to be appropriate and useful to the community. This study would be potentially stronger with a more formal analysis of what gene expression changes occurred in otic tissue in lmx1bb mutants, but it is also useful without this. I did have a couple of minor suggestions for the presentation of some aspects that would have made it easier for me as a reader.

Reviewer #3 (Public review):

Summary:

The authors use single-cell transcriptomic analysis to identify distinct cell types in the zebrafish inner ear. They identify markers of hair cells and supporting cells associated with sensory patches, cells that generate the semicircular canals, endolymphatic duct and sac, and periotic mesenchymal cells.

Strengths:

The computational analysis is thorough, and the findings are clearly described. In situ hybridization provides corroboration of cell identities in many cases. This resource atlas will be of particular interest for studies of inner ear morphogenesis. Indeed, the identification of a smooth muscle marker in the endolymphatic sac suggests future analysis of the degree to which this structure undergoes contraction. Identification of cell signaling components in BMP, Wnt, FGF, and other signaling pathways will also provide a resource for understanding signals coordinating ear development.

Weaknesses:

The manuscript is incomplete. Important details that would allow replicable analysis are not provided, with notebooks not available on the referenced GitHub site, and additional files are missing.

The authors make a detailed description of hair cells and supporting cells that are consistent with previous findings (Figures 2 and 3). By contrast, the analysis of distinct cell types that have not been previously well characterized in zebrafish is somewhat incomplete. Markers are described for cells forming the semicircular canals, including ccn1l1 (Figure 4). The authors report an intriguing pattern of its expression before overt bud formation; however, they provide no detailed expression analysis to support this assertion.

The authors also identify new markers for subsets of periotic mesenchyme (Figure 6). These include epyc and otos, which mark distinct populations within the mammalian inner ear - cochlea supporting cells, spiral limbus, and ligament, respectively. Identification of the equivalent of the spiral ligament would be of particular interest. However, the expression analysis is not of sufficient resolution to identify which cell types these represent in the zebrafish inner ear.

Differences in gene expression are reported for lmx1bb mutants. However, none of the single-cell data for mutants is provided, and the table (S8) of differential gene expression is missing. Significantly more detail would be needed to interpret these findings.

Author response:

We thank the editors and reviewers for their careful consideration of our manuscript and for their constructive feedback, which we will address in detail in our revised version. We value that Reviewer 1 considered that “data they compiled and submitted to public databases is a valuable resource for the community.” We are also encouraged by Reviewer #2 when they stated that “The data set is very nice, and the annotations are extremely rigorous and more in-depth than other datasets that include these tissues, since these investigators have enriched significantly for this tissue of interest. Their use of PAGA to identify potential developmental relationships within the data is rigorous. I also would like to specifically point out how incredibly gorgeous the microscopy of the lmx1bb phenotype is in Figure 7. Wow.” We were encouraged by Reviewer #3’s comments that “The computational analysis is thorough, and the findings are clearly described. In situ hybridization provides corroboration of cell identities in many cases. This resource atlas will be of particular interest for studies of inner ear morphogenesis.”

We spent a significant effort and time considering and addressing the reviewers’ public criticisms.

Below we address the criticisms of the reviewers’ Public Reviews individually.

Public Reviews:

Reviewer #1 (Public review):

Weaknesses:

Many of the clusters have not been annotated or rely on published data. For the ones for which no HCRs or UMAPs are shown, it is therefore difficult to estimate which of the markers are indeed the most cell type/state-specific ones.

Major comments:

(1) It would be very useful if the cluster numbers in the Excel files also had the associated cell type annotations as a second column (at least for the ones that are known). E.g., in Supplemental Table 2, the text states which clusters represent which neuromast and ear cell type, but these are not mentioned in the Excel table.

Thank you for the suggestion, we will include additional annotations in the revised version.

(2) Many of the clusters have not been annotated or rely on published data. For the ones for which no HCRs or UMAPs are shown, it is therefore difficult to estimate which of the markers are indeed the most cell-type/state-specific ones.

We recognize the need to evaluate potential new markers, we will include a heat map of markers and clusters to assess cell-type/state specificity in the revised version.

(3) Uploading the data to gEAR (https://umgear.org/dataset_explorer.html), a web-based, publicly available ear database, would further increase the usefulness of this study to the broader community.

We appreciate the suggestion to upload to gEAR and will upload to the database in the near future.

Method:

The authors should provide the details about how many cells were sequenced for each ear developmental stage, how many cells were present per cluster (page 8), and how many cells were present in each subcluster of ear and lateral line clusters (page 10).

We will add cell numbers for each cluster in the revised version as an additional column in the supplemental tables.

Reviewer #2 (Public review):

Weaknesses:

A missed opportunity is that the authors describe creating an additional scRNAseq dataset from lmx1bb mutants, but do not show any comparative scRNAseq analyses that would identify broader sets of differentially expressed genes. It seems almost as if a key element of the study was removed at the last minute, and as a result, the discussion of changes in epcam expression in lmx1bb mutants in Figure 7 seems somewhat tacked onto the end of the study and not motivated by the analyses presented in the manuscript.

Overall, I do not think this study requires any major revisions to be appropriate and useful to the community. This study would be potentially stronger with a more formal analysis of what gene expression changes occurred in otic tissue in lmx1bb mutants, but it is also useful without this. I did have a couple of minor suggestions for the presentation of some aspects that would have made it easier for me as a reader.

We will include analysis of the lmx1bb mutant data in the revised version and value the suggestions for improved presentation. We will work on irmpoving presentation of the mutant data, including a UMAP with the WT cells in one color and the mutant cells in another color.

Reviewer #3 (Public review):

Weaknesses:

The manuscript is incomplete. Important details that would allow replicable analysis are not provided, with notebooks not available on the referenced GitHub site, and additional files are missing.

Python notebooks will be added shortly, and files for mapping in Drops data will be provided at the GitHub site.

The authors make a detailed description of hair cells and supporting cells that are consistent with previous findings (Figures 2 and 3). By contrast, the analysis of distinct cell types that have not been previously well characterized in zebrafish is somewhat incomplete. Markers are described for cells forming the semicircular canals, including ccn1l1 (Figure 4). The authors report an intriguing pattern of its expression before overt bud formation; however, they provide no detailed expression analysis to support this assertion.

The authors also identify new markers for subsets of periotic mesenchyme (Figure 6). These include epyc and otos, which mark distinct populations within the mammalian inner ear - cochlea supporting cells, spiral limbus, and ligament, respectively. Identification of the equivalent of the spiral ligament would be of particular interest. However, the expression analysis is not of sufficient resolution to identify which cell types these represent in the zebrafish inner ear.

Thank you for your input regarding the analysis of the periotic mesenchyme. In the revised version, we will attempt to improve resolution of different populations, first by comparing epyc and otos expression by HCR. It is unclear how to correlate any patterns with structures that have yet to evolve, but we will look for similarities and differences to studies performed in mice (PMID: 37720106).

Differences in gene expression are reported for lmx1bb mutants. However, none of the single-cell data for mutants is provided, and the table (S8) of differential gene expression is missing. Significantly more detail would be needed to interpret these findings.

We will include analysis of the lmx1bb mutant data in the revised version and value the suggestions for improved presentation.