Author Response:
The following is the authors’ response to the original reviews.
Reviewer #1 (Public Review):
In their study, Aman et al. utilized single cell transcriptome analysis to investigate wild-type and mutant zebrafish skin tissues during the post-embryonic growth period. They identified new epidermal cell types, such as ameloblasts, and shed light on the effects of TH on skin morphogenesis. Additionally, they revealed the important role of the hypodermis in supporting pigment cells and adult stripe formation. Overall, I find their figures to be of high quality, their analyses to be appropriate and compelling, and their major claims to be wellsupported by additional experiments. Therefore, this study will be an important contribution to the field of vertebrate skin research. Although I have no major concerns, I would like to offer a few minor comments for the authors to consider.
- The discovery of ameloblasts in the zebrafish skin is a fascinating finding that could potentially provide a new research model for understanding the development and regeneration of vertebrate teeth. It would be beneficial if the authors could provide further elaboration on this aspect and discuss how the zebrafish scale model could be utilized by researchers to better understand the morphogenesis of vertebrate teeth and/or hair.
We have provided additional discussion points regarding epidermal EMP+ cells with ameloblast-like transcriptional profiles. We believe that further studies of scale matrix composition and the material properties endowed by various collagenous and non-collagen matrix proteins will be useful for understanding fundamental mechanisms of biomineralization. This section of the discussion now reads:
“We systematically assessed the expression of genes encoding non-collagen calcified matrix proteins throughout the skin during squamation, leading to the discovery of a transcriptionally distinct population of basal epidermal cells that express EMP transcripts, likely corresponding to epidermal secretory cells proposed to participate in scale matrix formation based on ultrastructure (Sire et al., 1997). These cells also express dlx3a, dlx4a, runx2b and msx2a but not sp7, a transcription factor suite that is shared with ameloblasts that form tooth enamel. While these transcription factors are not exclusive to ameloblasts and have been reported in osteoblasts and odontoblasts, in addition to cell types that do not produce calcified matrix, such as neurons, their co-expression along with EMP-encoding transcripts in basal epidermal cells is consistent with a common origin of ameloblasts and the EMP+ epidermal cells reported here. One alternative hypothesis is that co-expression of these gene products arose convergently and can be explained by mechanistic linkages among them. Future work aimed at functionally dissecting the regulatory mechanisms that govern EMP gene expression in a variety of organisms may clarify these issues either by providing evidence of additional commonalities, supporting a shared ancestor, or by revealing diverse, lineage-specific regulatory architectures, supporting convergent evolution of superficial enamel deposition in teeth and fish skin appendages.”
- While the overexpression-rescue experiments (i.e., fgf20a and pdafaa) provide crucial evidence to support the author's conclusions, it is important to note that overexpression driven by the heat-shock promoter is not spatially regulated. Therefore, it should be acknowledged that the rescue effects may not be cell-autonomous, as suggested in the current version.
The reviewer is correct that hsp70l promotor is not spatially regulated and F0 transgenics have random mosaic expression. Importantly, since we were testing specific hypotheses regarding signaling interactions between basal epidermal cells and dermal cells, we applied stringent selection and only analyzed individuals with transgene expression in basal epidermal cells. This approach enabled us to assay the results of basal cell expression of signaling ligands in eda mutant and hypo-thyroid backgrounds. The original manuscript omitted this crucial aspect of our experimental design, and we thank the reviewer for noticing this omission. We have revised the following parts of the results section.
“Indeed, heatshock-driven expression in F0 mosaics stringently selected for basal epidermal expression of Fgf20a in the skin of Eda mutants led to localized rescue of scales where transgene expression was detectable (Figure 5D).”
“When we forced expression of Pdgfaa in basal cells of epidermis by heatshock induction and stringent selection of basal epidermal expression in F0 mosaics, we found, as predicted, a recruitment of dermal cells in hypoTH skin, leading to a locally stratified dermis (Figure 6E) similar to that of the wild-type (Figure 4C).”
We additionally revised the legends for Figure 5 and Figure 6 to mention stringent selection of basal epidermal expression of fgf20a and pdgfaa, respectively.
- Figure 7D. The authors used the ET37:EGFP lines to visualize hypodermis. Based on the absence of EGFP signal in the deep dermis of bnc2 mutants, the authors concluded that the hypodermis may be missing, suggesting the importance of the hypodermis in pigment cell formation. However, since the EGFP evidence is indirect, it is crucial to confirm the absence of the hypodermis structure with histology.
It is indeed conceivable that hypodermal cells physically persist in bnc2 mutants yet have sufficiently altered gene expression that they neither cluster with wild-type hypodermal cells in single cell RNA-seq analyses nor initiate or maintain the broadly expressed dermal reporter ET37:GFP that we used to assess the presence or absence of such cells in a defined anatomical position. Though we believe this to be somewhat unlikely (hence our original interpretation), we have added a caveat referencing this formal possibility in the revised manuscript:
“It is possible that hypodermal cells physically persist in bnc2 mutants but have sufficiently altered transcriptional profiles such that they no longer cluster together with wild-type hypodermal cells or express the ET37:EGFP transgene. Nevertheless, these analyses suggest that ET37:EGFP+ hypodermal cells likely play a role in pigment pattern development.”
We believe this issue raises interesting philosophical questions about the definition of a “cell-type.” If cells constituting the deep surface of the dermis physically persist, but have a profoundly altered transcriptional profile and no longer perform the biological functions of their wild-type counterparts, are they still the original cell type, or was the wild-type cell type lost? As researchers continue to discover new cell types and deepen our understanding of cell-state plasticity in normal and pathological conditions, the community will need to articulate new rubrics of categorization to ensure that “cell-type” remains a rigorous and useful concept (if, indeed, it has been one).
- As the dataset is expected to be a valuable asset to the field, please provide Excel tables summarizing the key genes and their corresponding expression levels for each major cluster that has been identified.
This table has been provided in the revised manuscript (Supplementary file 2 – Table 5.)
Reviewer #2 (Public Review):
The authors used single cell transcriptome analysis of zebrafish skin cells and characterized various types of cells that are involved in scale formation and stripe patterning. The methods employed in this study is highly powerful to provide mechanistic explanation of these fundamental biological issues and will be a good example for many researchers studying other biological issues. Furthermore, the results characterizing differences in gene expression patterns among various types of cells will be informative for other researchers in the field.
For scale formation, it is known that mineralized tissues may significantly differ in rayfins and lobefins since sox9, col2a1, and col10a1 are all expressed in osteoblasts, in addition to chondrocytes, in zebrafish and gar (Eames et al., 2012, BMC Evol. Biol.). Furthermore, in mammals, Col10 is expressed in chondrocytes in mature cartilage that undergoes ossification. Thus, unlike the authors argue, col10a1 expression is not apparently relevant to the elasticity of scales.
The authors also state that the expression of dlx4a, msx2a, and runx2b characterize cells homologous to mammalian ameloblasts. However, dlx4, runx2, and msx2 are all duplicated in zebrafish, and the function of duplicated genes in teleost fishes may differ from that of single ancestral gene. Moreover, none of Dlx4, Msx2, and Runx2 is expressed specifically by ameloblasts in mammals. Indeed, both Msx2 and Runx2 are expressed in osteoblasts, while the expression of Dlx4 in ameloblasts is not reported. These results, together with the expression of an enamel gene, enam, in dermal cells (SFC), do not appear to support the homology of the surface tissue of mammalian teeth and zebrafish scales.
We appreciate the reviewers’ comments and have provided caveats to our interpretation in the revised manuscript (see our response to Reviewer #1, item 1, above). In the revised manuscript, we also display results for an additional Dlx gene, dlx3b, that is coexpressed in EMP+ basal epidermal cells (Figure 3C), although dlx4 has been reported in mammalian tooth germs and elasmobranch tooth and odontode epithelia (Pemberton et al., 2007; Debiais-Thibaud et al., 2011 ; Woodruff et al., 2022).
More generally, expression of specific genes can be useful characters for testing hypotheses of homology. The operant inference depends on a parsimony assumption: if a transcriptional profile is shared between celltypes in disparate organisms, one explanation is that this transcriptional profile was inherited from a common ancestor. This inference is not impacted by the teleost whole genome duplication. If the common ancestor had one ortholog and a subset of modern animals have two, the homology hypothesis predicts that at least one ortholog will be expressed in common in the tissue that descended from the common ancestor. This interpretation is entirely compatible with our understanding of the mechanisms that underlie retention of duplicated genes in animal genomes. Additionally, exclusivity is not necessarily predicted by homology hypotheses. Indeed, all the transcription factors used here as characters for evaluating homology have pleiotropic roles in many cell types.
In this specific case, we found two EMP genes, ambn and enam, co-expressed with a complement of transcription factors that is also co-expressed in ameloblasts. These findings are consistent with a model in which both ameloblasts and EMP+ epidermal cells associated with zebrafish scales inherited this transcriptional profile from a common ancestral cell type. Given the temporal and phylogenetic continuity of superficial enameling in the fossil record of skin appendages, and the dual origin of mineralized matrices in extant skin appendages and teeth, we continue to favor the model where these traits are shared and conserved among vertebrates. Nevertheless, we have acknowledged in the revised manuscript the limitations of homology testing by analyses of gene expression and the possibility that these traits might have evolved convergently; we suggest additional research avenues for testing this hypothesis further (response to Reviewer #1, item 1, above).
Reviewer #3 (Public Review):
This work describes transcriptome profiling of dissected skin of zebrafish at post-embryonic stages, at a time when adult structures and patterns are forming. The authors have used the state-of-the-art combinatorial indexing RNA-seq approach to generate single cell (nucleus) resolution. The data appears robust and is coherent across the four different genotypes used by the authors.
The authors present the data in a logical and accessible manner, with appropriate reference to the anatomy. They include helpful images of the biology and schematics to illustrate their interpretations.
The datasets are then interrogated to define cell and signalling relationships between skin compartments in six diverse contexts. The hypotheses generated from the datasets are then tested experimentally. Overall, the experiments are appropriate and rigorously performed. They ask very interesting questions of interactions in the skin and identify novel and specific mechanisms. They validate these well.
The authors use their datasets to define lineage relationships in the dermal scales and also in the epidermis. They show that circumferential pre-scale forming cells are precursors of focal scale forming cells while there appeared a more discontinuous relationship between lineages in the epidermis.
The authors present transcriptome evidence for enamel deposition function in epidermal subdomains. This is convincingly confirmed with an ameloblastin in situ. They further demonstrate distinct expression of SCPP and collagen genes in the SFC regions.
The authors then demonstrate that Eda and TH signalling to the basal epidermal cells generates FGF and PDGF ligands to signal to surrounding mesenchyme, regulating SFC differentiation and dermal stratification respectively.
Finally they exploit RNA-seq data performed in parallel in the bnc2 mutants to identify the hypodermal cells as critical regulators of pigment patterning and define the signalling systems used.
Whilst these six interactions in the skin are disparate, the stories are unified by use of the sci-RNA-seq data to define interactions. Overall, it's an assembly of work which identifies novel and interesting cell interactions and cross-talk mechanisms. There are some aspects that require clarification:
With respect to the discontinuous relationship noted in Figure 2I in the epidermis, the authors did not make mention of the fact that there are in fact two independent sources of periderm in the zebrafish. The first periderm derives from the EVL, is segregated a gastrulation, and gradually replaced from the basal epidermis during post-embryonic stages. Could this residual EVL-derived periderm have reduced sensitivity of the trajectory mapping from basal to periderm? The authors should comment whether their transcriptome dataset likely had residual EVL-derived periderm and if this might have impacted their trajectory continuity interpretation.
While dual origin of periderm may impact the single cell analysis, this should not be an issue for suprabasal cells, which also show no continuity with their basal cell progenitors in UMAP space. We thank the reviewer for bringing this issue up and comment on the dual origin of periderm in the revised manuscript.
“During this stage of development, basal epidermal cells are the stem cell population that differentiate into both suprabasal and periderm cells, and each of the three major epidermal cell types are well represented in our dataset (Figure 2H,I; Figure 1—figure supplement 3)(Guzman et al., 2013; Lee et al., 2014). While periderm cells at the sampled stage are likely of dual origin, representing a mixture of early embryonic and stem cell derived cells, suprabasal cells are entirely derived from basal cells (Kimmel et al., 1990; Guzman et al., 2013; Lee et al., 2014).”
During this stage of development, basal epidermal cells are the stem cell population that differentiate into both suprabasal and periderm cells, and each of the three major epidermal cell types are well represented in our dataset (Figure 2H,I; Figure 1—figure supplement 3)(Guzman et al., 2013; Lee et al., 2014). While periderm cells at the sampled stage are likely of dual origin, representing a mixture of early embryonic and stem cell derived cells, suprabasal cells are entirely derived from basal cells (Kimmel et al., 1990; Guzman et al., 2013; Lee et al., 2014).
The authors ask if dermal SFCs express proteins associated with cartilage formation and use Col10a1 orthologues as markers (Fig 3B, I). I wonder if these are the best transcripts to answer this question as this has also been described to label osteoblasts in certain contexts in the fish and the authors might want to refer to Li et al 2009 or Avaron et al 2005. Were other markers of cartilage formation present such as collagen2 genes? These may be more definitive. The authors might want to reinterrogate their datasets for true cartilage markers or reframe their question.
In the revised manuscript, we have clarified and moderated inferences from col10a1 ortholog expression. Col2 genes were not detected robustly in our dataset. This section now reads:
“Scale elasmoidin is a flexible, collagenous ECM, material properties that are similar to cartilage (Quan et al., 2020). We therefore wondered whether dermal SFCs express matrix proteins associated with cartilage formation. Col10a1 is a major structural molecule in collagen, although its expression has also been documented in osteoblasts (Gu et al., 2014; Yang et al., 2014; Kawasaki et al., 2021). The zebrafish genome harbors genes encoding two Col10a1 orthologs (col10a1a and col10a1b) and we found both transcripts in SFCs representing distinct steps of maturation (Figure 3B,I; Figure 2—figure supplement 1F,I).”
Finally, of interest, were there any clear clusters on the UMAP plots (Fig 1 Supp3A) of unassigned identity? Even comment on these and how they were dealt with would be of significant interest to the field, as it is highly unlikely all cell types in the skin have been defined. This dataset promises to be a critical reference for defining these in the future.
Thanks for raising this issue. We provide a new figure (Figure 1 – supplement 4) displaying the unsupervised clustering of the wild-type dataset and a new table (Supplementary file 2 – table 5) with gene expression information for these clusters.
Minor clarification:
Fig 2E top. The authors interpret that fate-mapped SFCs at the posterior margin are progressively displaced towards the scale focus. This is confusing as the margin SFC in Fig 2E seems to show them staying largely at the margin. Please clarify if this is what you meant.
In Figure 2E, a new row of newly differentiated, non-photoconverted SFC were added, displacing the existing row of cells towards the scale focus. Since these cells are all very thin, the net displacement was not as dramatic as the displacement found for sub-marginal SFCs. This point has been clarified in the figure legend in the revised manuscript. This figure legend now reads:
“Figure 2. Postembryonic skin cell lineage relationships are not reflected in UMAP space. (A) UMAP visualization showing distribution of differentiated SFC expressing sp7 and pre-SFC progenitors expressing runx2b. (B) In-situ hybridization of sp7 and runx2b shows that a halo of pre-SFC progenitors surround the growing scale (arrows). (C) sp7:nEOS expressing differentiated SFC (magenta), were labelled by photoconversion on Day 1. Over the following two days, newly differentiated, un-photoconverted SFC appeared at the scale margin (arrows; n = 5 fish). (D) Schematic representation of differentiated SFC (purple) and the associated halo of pre-SFC (blue). (E) Photoconversion of small groups of SFC in the scale margin and sub-margin; and single-cell photoconversion of focus SFCs (arrows) showed that SFC are progressively displaced toward the scale focus and that SFC in all these regions are capable of cell division (arrows, n ≥ 4 fish for each region tested). Margin SFCs were displaced towards the posterior by newly differentiated, un-photoconverted SFCs (arrowheads). (F) SFCs in UMAP space colored by “pseudotime” rooted in the SFCs. (G) SFCs in UMAP space colored by the ratio of a mesenchymal (migratory) signature to an epithelial signature (Supplementary file 2—Table 3). (H) Schematic representation of epidermis with major substrata. (I) UMAP visualization of wild-type epidermis, subclustered independently of other cell types and displaying expression of the epidermal basal cell marker tp63 (blue) and the periderm marker krt4 (red). Scale bars, 50 μm (B,C,E); 25 μm, (C, lower). (J) The fraction of cells from panel H that pass a minimum threshold for expression of tp63, krt4 or both genes. .”
References
Debiais-Thibaud M, Oulion S, Bourrat F, Laurenti P, Casane D, Borday-Birraux V. 2011. The homology of odontodes in gnathostomes: insights from Dlx gene expression in the dogfish, Scyliorhinus canicula. BMC Evolutionary Biology 11:307. doi: 10.1186/1471-2148-11-307,
Pemberton TJ, Li FY, Oka S, Mendoza-Fandino GA, Hsu YH, Bringas P, Jr., Chai Y, Snead ML, Mehrian-Shai R, Patel PI. 2007. Identification of novel genes expressed during mouse tooth development by microarray gene expression analysis. Dev Dyn 236:2245-57. doi: 10.1002/dvdy.21226, PMID: 17626284
Woodruff ED, Kircher BK, Armfield BA, Levy JK, Bloch JI, Cohn MJ. 2022. Domestic cat embryos reveal unique transcriptomes of developing incisor, canine, and premolar teeth. Journal of Experimental Zoology Part B: Molecular and Developmental Evolution 338:516-31. doi: https://doi.org/10.1002/jez.b.23168
