Regulation of zebrafish melanocyte development by ligand-dependent BMP signaling

Essential revisions:

1) Figure 3A: In situs for gdf6a in the neural crest, and the absence from mitfa expression are not convincing. Figure 3B: It is not clear what the phopho-SMAD signal is being counted in the bottom embryo. What regions of the animal are defined as leading edge and rostral?

We recognize that our initial description of gdf6a and mitfa in situ expression data should have been more thorough and precise. In the revised manuscript, we cover the domains and sites of expression for gdf6a and mitfa and note that these are consistent with previously described expression data (Lister et al., 1999; Reichert et al., 2013; Rissi et al., 1995). Neural crest expression of gdf6a is not detected by in situs at later embryonic time points. Since expression patterns for gdf6a and mitfa have not previously been juxtaposed, we simply use the in situ data to show that the domains of expression are ‘mostly, if not completely, non-overlapping’.

In addition to the clarification above, we have specifically examined gdf6a expression in neural crest cells. We isolated eGFP-positive neural crest cells by FACS from Tg(crestin:eGFP) embryos and performed qRT-PCR to measure gdf6a and mitfa expression (new Figure 3A). In this experiment, neural crest cells from early (14ss, 16hpf) embryos and late (26ss, 22hpf) embryos were isolated. We confirmed gdf6a expression in the neural crest and found that it was relatively higher at 16hpf as compared to the 22 hpf neural crest. By contrast, mitfa expression was relatively higher at 22hpf compared to 16hpf. These data complement the in situ studies, showing early gdf6a neural crest expression as well as an inverse correlation between gdf6a and mitfa expression, at least in this cell type.

To clarify pSMAD positivity, we have redrawn the zebrafish diagram to show rostral regions and the leading edge on a single embryo. The leading edge is categorized as the 5 distal-most mitfa:GFP positive cells. Rostral mitfa:GFP cells include other mitfa:GFP cells not in the 5 distal-most GFP positive cells. We have included this description in both the figure legend and within the body text.

2) The authors present an adult phenotype for the gdf6a mutant and clearly state that "zebrafish develop their adult pigment pattern during metamorphosis, it is possible gdf6a acts during this stage to change adult pigmentation", although nothing is mentioned about the neural crest-derived cells established during somitogenesis at the level of the DRG and give rise to the adult melanocytes (Singh et al., 2015; Dooley et al., 2013; Hultman et al., 2009). The authors should explore this further.

To address the reviewer comment, we have performed Hu C/D antibody staining in Tg(mitfa:eGFP) embryos at 3 DPF treated with BMPi or vehicle control (new Figure 5—figure supplement 1F). We observed an increase in the number of DRG-associated mitfa:EGFP-positive cells in BMPi-treated embryos. This suggests BMP inhibition not only increases embryonic melanocytes, but also impacts neural crest-derived melanocyte stem cells and their role in adult pigment pattern formation. This raises several interesting questions (e.g. what is the source of these extra cells – altered specification, proliferation, migration?; what are the fates that these cells adopt?) that are to be the subject of extensive further studies.

3) In order to determine the proliferation activity of neural crest cells and pigment progenitor cells the authors measured the DNA content by fluorescence activated cell sorting. Although this gives an indication of the cell cycle phase in which cells are found at a certain moment, it doesn't tell us whether cells have higher proliferation rate which could be assessed using thymidine analogue incorporation. This would provide additional data that would support their conclusions, given the modest effects seen in Figure 2C and 2E.

We have performed EdU incorporation during neural crest development of Tg(crestin:eGFP) and Tg(mitfa:eGFP) embryos to determine if any changes in proliferation rate are evident. During each pulse of EdU (12-14hpf, 16-18hpf, 20-22hpf) we observed similar proportions of EdU and eGFP double-positive cells in BMPi- and vehicle-treated groups, indicating no significant change in proliferation rate upon BMP inhibition. These results are included in Figure 2—figure supplement 1 and discussed in the Results section of the main text.

4) One interesting aspect of this study and the previous study of GDF6 regulation in melanoma is the repression of MITF expression. Here, repression of mitfa expression by BMP signaling is proposed to play a key role in the choice of melanocyte versus iridophore fate. The authors should better address the mechanism of MITF repression. They previously showed binding of phospho-SMAD to a region upstream of the MITF-M promoter in human melanoma cells. Is this element conserved in zebrafish? How does phospho-SMAD act as transcriptional repressor under these conditions? It has been shown that foxd3, like BMP signaling regulates the ratio of melanophores to iridophores. Curran et al., 2009, argue that foxd3 acts upstream of mitfa expression. In this study, and in the Venkatesan et al., 2018 paper, BMP signaling stimulated phospho-SMAD acts upstream of mitfa expression. Do they act in parallel or in sequence? Is there a genetic interaction study that would help establish the epistasis? For instance, does gdf6(lf) still increase melanocytes in foxd3(lf) mutants? These experiments would better help to distinguish between a direct repression by SMAD binding to a cis-regulatory element of MITF or an indirect mechanism.

In our previous work, we used ChIP-seq to identify a pSMAD binding region at the MITF locus, within the intron of long MITF isoforms, and upstream of MITF-M (Venkatesan et al., 2018). Within this binding region we found a known pSMAD binding motif (GC-SBM) (Morikawa et al., 2011). To assess whether there is a potential for pSMAD regulation of the mitfa locus in zebrafish, we first investigated whether the MITF and mitfa loci reside in syntenic regions of the genome (new Figure 7—figure supplement 1A). We found multiple orthologous genes flanking MITF and mitfa, indicating synteny at this region of the genome. We next examined whether the pSMAD binding region identified in human cells was conserved in zebrafish. Using BLAST and BLAT searches as well as pairwise DNA matrices between the pSMAD binding region and mitfa locus, we did not identify extensive tracts of DNA sequence conservation. However, within the mitfa locus in zebrafish, we identified multiple intronic and exonic pSMAD binding motifs, as well as multiple motifs within the mitfa promoter (new Figure 7—figure supplement 1B). These include the GC-SBM motif described above as well as additional SBMs described previously (Jonk et al., 1998). Given the synteny between the MITF and mitfa loci, the presence of SMAD binding motifs, and similar transcriptional changes under BMP inhibition, we believe it is possible mitfa is directly regulated by BMP-activated pSMAD.

To address the potential interaction of foxd3 and gdf6a signaling in regulating the development of melanocytes, we first performed qPCR for foxd3 expression in both gdf6a-mutant and BMPi-treated animals. Compared to controls, neither experimental group showed significant changes in foxd3 expression (Figure 5—figure supplement 1). To assess if there is functional relationship between BMP signaling and foxd3 in melanocyte development, we treated foxd3(zdf10)-mutant zebrafish with BMPi or vehicle, and assessed development of the melanocyte lineage. As previously described (Stewart et al., 2006), foxd3(zdf10) mutants have deficiencies in specific neural crest derived populations, but maintain a normal number of melanocytes during early development. In our experiment, we observed a comparable increase in the number of melanocytes between mutant, heterozygous, and wild-type foxd3 embryos treated with BMPi. Since inhibition of BMP signaling (more melanocytes) is different than loss of foxd3 (melanocytes unchanged), this indicates that BMP signaling is not acting exclusively with foxd3 to regulate melanocyte fate. Consequently, the simplest interpretation is that BMP signaling can act independently of foxd3 to regulate melanocyte development. We have included these results in Figure 5—figure supplement 1.

5) The study contains little description of how BMP signaling affects the expression of key melanocyte genes other than mitfa and key neural crest genes aside those shown in Figure 5A. The authors should investigate for example whether MITF target genes are down-regulated along with mitfa and if and how sox10 and foxd3 expression is affected.

We have performed qPCR for additional genes associated with melanocyte differentiation. We observed an upregulation of mitfa targets tyr and mc1r when gdf6a is lost or BMP signaling inhibited (now included in Figure 1E), supporting our previous findings for mitfa and tyrp1b. We have also performed qPCR for both foxd3 and sox10 (Figure 5—figure supplement 1). As discussed above, foxd3 expression was unaltered in gdf6a-mutant or BMPi-treated embryos, and functional evaluation suggested BMP signaling can act independently of foxd3 in melanocyte development. We observed a slight downregulation of sox10 when gdf6a is lost or BMP signaling is inhibited, which is in line with our previous findings in melanoma (Venkatesan et al., 2018). sox10 normally promotes mitfa expression. Since we still observed an increase of mitfa expression in gdf6a(lf) and BMPi-treated embryos, we speculate that the reduced sox10 expression was not enough to impede its activation of mitfa or reduced sox10 had some negative impact on mitfa expression but this impact was counteracted by loss of BMP-mediated mitfa repression.