Regulation of melanocyte development by ligand-dependent BMP signaling underlies oncogenic BMP signaling in melanoma

Preventing terminal differentiation is important in the development and progression of many cancers including melanoma. Recent identification of the BMP ligand GDF6 as a novel melanoma oncogene showed GDF6-activated BMP signaling suppresses differentiation of melanoma cells. Previous studies have identified roles for GDF6 orthologs during early embryonic and neural crest development, but have not identified direct regulation of melanocyte development by GDF6. Here, we investigate the BMP ligand gdf6a, a zebrafish ortholog of human GDF6, during the development of melanocytes from the neural crest. We establish that the loss of gdf6a or inhibition of BMP signaling during neural crest development disrupts normal pigment cell development, leading to an increase in the number of melanocytes and a corresponding decrease in iridophores, another neural crest-derived pigment cell type in zebrafish. This shift occurs as pigment cells arise from the neural crest and depends on mitfa, an ortholog of MITF, a key regulator of melanocyte development that is also targeted by oncogenic BMP signaling. Together, these results indicate that the oncogenic role ligand-dependent BMP signaling plays in suppressing differentiation in melanoma is a reiteration of its physiological roles during melanocyte development.

BMP signaling during neural crest development disrupts normal pigment cell 23 development, leading to an increase in the number of melanocytes and a corresponding 24 decrease in iridophores, another neural crest-derived pigment cell type in zebrafish. 25 This shift occurs as pigment cells arise from the neural crest and depends on mitfa, an 26 ortholog of MITF, a key regulator of melanocyte development that is also targeted by 27 oncogenic BMP signaling. Together, these results indicate that the oncogenic role 28

Introduction 35
Tumor differentiation status is often an important prognostic factor in cancer. For many 36 cancer types, tumors that are less differentiated are associated with a higher grade and 37 worse prognosis compared to more differentiated tumors, which often follow indolent 38 courses (Hoek et al., 2006;Rosai & Ackerman, 1979). In order to adopt a less 39 In these studies, we used the gdf6a s327 allele, hereafter referred to as gdf6a(lf), which 104 encodes an early stop codon and has previously been shown to cause a complete loss 105 of gdf6a function (Gosse & Baier, 2009). Previous studies have identified early roles for 106 gdf6a during initial embryonic patterning, including dorsoventral patterning immediately 107 following fertilization, thus gdf6a(lf) mutants have significantly decreased viability during 108 the first 5 days post fertilization (Sidi, Goutel, Peyrieras, & Rosa, 2003). However, we 109 found that a small proportion of gdf6a(lf) animals are able to survive early development 110 and progress to adulthood. These gdf6a(lf) adult zebrafish had increased pigmentation 111 when compared to wild-type zebrafish ( Figure 1A). Furthermore, gdf6a(lf) adult 112 zebrafish had qualitative disruption of the normal pigment pattern of both stripe and 113 scale-associated melanocytes, and a significant increase in the number of scale-114 associated melanocytes as well as the overall scale area covered by melanin (Figure 115 1A,1B). These results indicate that gdf6a(lf) mutants have melanocyte defects. 116 117 Loss of gdf6a or inhibition of BMP signaling leads to an increase in embryonic 118 melanocytes 119 Since zebrafish develop their adult pigment pattern during metamorphosis, it is possible 120 gdf6a acts during this stage to change adult pigmentation, and not during initial pigment 121 were BMP-dependent. We crossed gdf6a(lf) heterozygotes and, in randomly selected 125 progeny, quantified the number of melanocytes that developed by 5 days post-126 fertilization (DPF). Following melanocyte quantification, we determined the genotype of 127 each embryo. In parallel, we treated wild-type zebrafish during the period of neural crest 128 induction and melanocyte specification (12 to 24 hours post fertilization) with a small 129 molecule BMP inhibitor, DMH1, hereafter referred to as BMPi, and performed the same 130 quantification of embryonic melanocytes (Hao et al., 2010). gdf6a(lf) homozygous 131 animals developed approximately 40% more dorsal melanocytes by 5 DPF, when 132 compared to sibling wild-type animals and gdf6a(lf) heterozygotes ( Figures 1C,1D and 133 S1A). gdf6a(lf) animals also showed increased expression of tyrp1b, a marker of 134 differentiated melanocytes, which is consistent with an increase in melanocyte number 135 ( Figure 1E). Furthermore, treatment with BMPi phenocopied the melanocyte changes 136 observed in gdf6a(lf) mutants, coupled with a similar increase in expression of tyrp1b 137 ( Figures 1D and 1E). We observed a similar increase in total body melanocytes, 138 indicating that there is an overall increase in melanocyte development instead of a 139 failure of migration leading to a specific increase in dorsal melanocytes ( Figure S1B). 140 These results indicate gdf6a-activated BMP signaling normally acts in embryos to limit 141 melanocyte development.   Figure S1D, S1E, S1F and S1G). We generated double mutants for both 155 gdf6a(lf) and gdf6b(lf) to assess whether these paralogs functioned redundantly or could 156 compensate for the loss of one another. Unfortunately, gdf6a(lf); gdf6b(lf) double 157 mutants had significant morphologic defects and decreased viability such that we could 158 not adequately compare melanocyte numbers in these animals ( Figure S1H and S1I). 159 However, because there were no pigmentation defects in gdf6b(lf) mutants and gdf6a(lf) 160 pigmentation defects were the same severity as observed in animals treated with a pan-161   Kimelman, 2005). We injected mitfa(lf) animals with miniCoopR-dnBMPR, miniCoopR-255 SMAD1-DVD, or control miniCoopR-eGFP ( Figure 4B). At 5 DPF, we scored animals for 256 rescue of melanocytes. Animals injected with miniCoopR-dnBMPR showed a rescue 257 rate of 79% as compared to 29% of miniCoopR-eGFP-injected animals. Furthermore, 258 animals injected with miniCoopR-SMAD1-DVD showed a 15% rescue rate ( Figure 4C). 259 Together these results suggest BMP signaling is active in mitfa-expressing cells and 260 modulating BMP signaling can alter the fate of these mitfa-expressing cells during 261 development. Thus, gdf6a-driven BMP signaling can both limit the number of mitfa-262 expressing cells arising from the neural crest but also act in mitfa-expressing pigment 263 progenitor cells to influence their development into melanocytes.  iridophores that developed in gdf6a(lf) embryos ( Figure 5B) and embryos treated with 304 Embryos developed 32% and 27% fewer iridophores with gdf6a(lf) or BMPi treatment, 306 respectively. Together, these results indicate that gdf6a-driven BMP signaling promotes 307 iridophore development.

Declaration of Interests 509
The authors declare no competing interests.

Methods 512
Zebrafish 513 Zebrafish were handled in accordance with protocols approved by the University of 514 Massachusetts Medical School IACUC. Fish stocks were maintained in an animal 515 facility at 28.5ºC on a 14-hour/10-hour Light/Dark cycle (Westerfield, 1995). The wild-516 type strain used was AB. Published strains used in this study include gdf6a(lf) 517 Epinephrine, 10mg/mL in embryo media. Embryos were dechorionated by incubating in 567 Pronase (Roche) for 10 minutes with gentle shaking. Dechorionated embryos were 568 transferred to 6-well plates coated in 1.5% agarose in embryo media. Embryo media 569 with appropriate drug concentration or vehicle control was added to each well. For BMPi 570 and 4-OHT treatments, embryos were treated from 12 HPF (6ss) to 24 HPF (Prim-5). 571 Embryos were incubated at 28.5ºC for the duration of the drug treatment. Following 572 drug treatment, embryos were thoroughly washed in fresh embryo medium and returned 573 to incubator in new embryo medium until analysis. 574 575

Lineage Tracing 576
To trace the lineage of embryonic pigment cells, Tg(ubi:switch) embryos were injected 577 with 25 pg of pDestTol2pA2-mitfa:Cre-ERT2:pA and 25 pg of Tol2 transposase RNA at 578 the single-cell stage. At 12 HPF, injected embryos were treated with BMPi and 4-OHT 579 as described above. Following treatment, embryos were thoroughly washed and 580 allowed to mature at 28.5ºC to 5 DPF. Embryos were treated with 1 mg/mL epinephrine 581 to contract melanosomes, anesthetized using 0.17mg/mL tricaine in embryo media, 582 mounted in 1% low-melt agarose on a plastic dish, and submerged in embryo media for 583 imaging.     condition. P-value was calculated using Fisher's exact test, P < 0.05. 1095