Countershading in zebrafish results from an Asip1 controlled dorsoventral gradient of pigment cell differentiation

Dorso-ventral (DV) countershading is a highly-conserved pigmentary adaptation in vertebrates. In mammals, spatially regulated expression of agouti-signaling protein (ASIP) generates the difference in shading by driving a switch between the production of chemically-distinct melanins in melanocytes in dorsal and ventral regions. In contrast, fish countershading seemed to result from a patterned DV distribution of differently-coloured cell-types (chromatophores). Despite the cellular differences in the basis for counter-shading, previous observations suggested that Agouti signaling likely played a role in this patterning process in fish. To test the hypotheses that Agouti regulated counter-shading in fish, and that this depended upon spatial regulation of the numbers of each chromatophore type, we engineered asip1 homozygous knockout mutant zebrafish. We show that loss-of-function asip1 mutants lose DV countershading, and that this results from changed numbers of multiple pigment cell-types in the skin and on scales. Our findings identify asip1 as key in the establishment of DV countershading in fish, but show that the cellular mechanism for translating a conserved signaling gradient into a conserved pigmentary phenotype has been radically altered in the course of evolution.


INTRODUCTION 48
Most vertebrates exhibit a dorso-ventral pigment pattern characterized by a light ventrum 49 and darkly colored dorsal regions. This countershading confers UV protection against 50 solar radiation, but also is proposed to provide anti-predator cryptic pigmentation. In 51 mammals, hair color results from biochemical differences in the melanin produced by 52 melanocytes, the only neural-crest derived pigment cell-type in this taxon. Best studied 53 in mice, the local expression of agouti-signaling protein (ASIP) in the ventral skin drives 54 the dorso-ventral pigment polarization 1,2 . ASIP is mainly produced by dermal papillae 55 cells where it controls the switch between production of eumelanin (black/brown 56 pigment) to pheomelanin (yellow/red pigment) by antagonizing -melanocyte-57 stimulating hormone (-MSH) stimulation of the melanocortin 1 receptor (MC1R) 1 . 58 Temporal control of Asip expression as a pulse midway during the hair growth cycle 59 generates a pale band of pheomelanin in an otherwise dark (eumelanin) hair ('agouti' 60 pattern). In contrast, in the ventral region, constitutive expression of Asip at high levels 61 represses eumelanin production, resulting in pale hair colour. 62 Most other groups of vertebrates share the dorso-ventral countershading pattern, but in 63 ray-finned fishes it results from a patterned distribution of light-reflecting (iridophores 64 and leucophores) and light-absorbing (melanophores, xanthophores, erythrophores, and 65 cyanophores) chromatophores 3,4 . Zebrafish, a teleost fish model for pigment studies, 66 obtains its striped pigmentation by the patterned distribution of three types of 67 chromatophores: melanophores, iridophores and xanthophores 5,6 . Furthermore, it is 68 widely accepted that fish melanophores only produce dark eumelanin, but not 69 pheomelanin 7 . Our recent experiments using overexpression systems have demonstrated 70 that zebrafish utilizes two distinct adult pigment-patterning mechanisms, the striped 71 patterning mechanism and the dorso-ventral patterning mechanism 8 . Both patterning 72 mechanisms function largely independently, with the resultant patterns superimposed to 73 give the full pattern 8 . The zebrafish striping mechanism has received much attention and 74 is based on a cell-cell interaction mechanism 9,10 . In contrast, dorso-ventral patterning has 75 been largely neglected, but we have recently provided evidence that it depends on asip1 76 expression, and furthermore that this is expressed in a dorso-ventral gradient in the skin 77 directly comparable to that in mammals 8,11,12 . This potential conservation of agouti 78 signaling protein function is fascinating, since it opens up the possibility of a very 79 different cellular mechanism of action in mammals and fish 8,13 . Specifically, we have 80 proposed that Asip1 activity in the ventral skin in zebrafish alters the balance of pigment 81 cell differentiation, through repressing melanophore differentiation 8 . 82 Studies of Asip1 function in fish to date have relied on gene overexpression approaches, 83 but loss-of-function experiments provide a complementary approach to test the 84 interpretation of those overexpression data. Here, we investigate the in vivo functional 85 role of asip1 in zebrafish by generating asip1 knockout mutants using clustered regularly 86 interspaced short palindromic repeats (CRISPR)-associated protein-9 nuclease (Cas9) 87 genome engineering tools 14 . We demonstrate that asip1 knockout mutant zebrafish 88 display a disrupted dorso-ventral pigment pattern characterized, in the ventral region, by 89 asip1 iim02 , asip1 iim07 and asip1 iim08 encode 71, 38 and 31 amino acid mutant proteins, 115 respectively (Fig 1.C). All mutated proteins have lost most of their basic central domain  116 and, most significanctly, the C-terminal poly-cysteine domain, which is the crucial region 117 for protein activity 15-17 . All asip1 knockout mutant zebrafish lines examined resulted in 118 a similar dorso-ventral pigment phenotype as described below. 119 120 asip1 function in dorsal-ventral pigment patterning 121 All three asip1-CRISPR knockout lines exhibited a loss of dorso-ventral countershading. 122 Because we did not find any difference in the pigment pattern across the three-knockout 123 mutants' lines, we focused on the study of line CRISPR1-asip1.iim08, here referred to as 124 asip1 K.O. . In asip1 K.O. fish, melanophores and xanthophores were more numerous in all 125 ventral regions ( Fig. 2A-2D), including the ventral head (Figs. 2 E,F). In WT fish, 126 melanophores and xanthophores were very limited in the ventral region, and mainly 127 located on the jaw and the posterior belly regions, near the pelvic fins ( Fig 2G). The WT 128 phenotype shows a low number of melanophores in the ventral head region and high 129 number of iridophores around the branchiostegals and operculum (Fig. 2E). In contrast, 130 asip1 K.O. mutants show melanophores spread throughout the jaws, branchiostegal and 131 opercular regions (Fig. 2F). On the belly, the ventral skin of WT fish showed almost a 132 total absence of melanophores, so that the bright whitish-reflective iridophore sheet of 133 the internal abdominal wall is prominent (Fig. 2G). Conversely, asip1 K.O. fish displayed 134 a strong increase in melanophore and xanthophore number in the ventral skin, as well as 135 many extra cells that transform the incipient 3V of the WT into a prominent 3V reaching 136 to the head in the asip1 K.O. (Fig. 2A-D). We note that the consistent increase in 137 melanophore numbers in the 2V and 3V stripes can also be considered a dorsalisation 138 phenomenon, since our counts show them to now resemble their more dorsal counterparts 139 (Figs 3 and 4). In addition, the abdominal wall exhibits an obvious decrease in the number 140 of iridophores, resulting in an apparent breakup of the iridophore sheet into smaller 141 fragments, thus conferring a darker color to the ventral region of asip1 K.O. fish (Fig. 2H). 142 The Sanger-generated mutant, asip1 sa13993 , showed only a subtle and partial phenotype 143 compared to asip1 K.O. fish, ((e.g. hyperpigmentation in the belly was not obvious; Supp. 144 and Methods and Fig. 3 and 4 for details). No differences in melanophore numbers were 157 found at larval stages (5dpf, SL 3 mm) (data not shown). In contrast, the dorsal-ventral 158 pigment abnormalities began to be visible from the earliest stages of metamorphosis 159 (15dpf, SL 6.3 mm). Although at 15 dpf there were no differences in melanophore number 160 in the belly between asip1 K.O. and WT fish, melanophore number in the ventral head was 161 68.7% higher in asip1 K.O. fish than in WT fish (P<0.05) (Fig. 3A). At 30 dpf, pigment 162 abnormalities also appear in the belly: melanophore number in the ventral head was 63% 163 higher in the asip1 K.O. than in WT fish (P<0.05), while in the belly melanophore numbers 164 were 41% higher in asip1 K.O. than WT belly (P<0.05) (Fig. 3B). 165 The asip1 K.O. fish at 60 and 210 dpf showed significant pigment pattern alterations, 166 particularly in the ventral region compared to WT fish (Fig. 4B). At 60 dpf, the number 167 of skin melanophores of asip1 K.O. fish was 47% higher (P<0.001) in dark stripe 2V, 86% 168 higher (P<0.001) in the ventral head, and 98% higher (P<0.001) in the belly than in 169 equivalent positions of WT fish. No differences were found in dorsal regions or in other 170 dark stripes (Fig. 4C). Furthermore, we found that the number of xanthophores was also 171 affected in ventral regions. At 60 dpf, the distribution of xanthophores in anterior area of 172 the belly was 98% higher (P<0.05) than in WT. No differences were found in dorsal 173 regions ( Fig. 4D). At 210 dpf, the same pattern of an increased number of melanophores 174 in the ventral region was found. The number of melanophores in asip1 K.O. fish was 38% 175 higher (P<0.001) in dark stripe 2V, 78.6% higher (P<0.001) in dark stripe 3V, 84% higher 176 (P<0.001) in the ventral head, and 99% higher (P<0.001) in the belly compared to the 177 equivalent region of WT siblings. Just as in 60 dpf fish, the pigment defects were 178 restricted to ventral regions (Fig. 4E). At 210 dpf, the number of xanthophores in the belly 179 region was 96% higher (P<0.001) compared to WT siblings, while no differences were 180 found in dorsal regions (Fig. 4F).
If Asip1 functioned in fish by a homologous cellular mechanism to that in mammals, we 182 would predict the presence of unpigmented melanophores in the ventral skin. To test this, 183 and to supplement the analysis of pigment cells using their autonomous pigmentation, we 184 also compared the distribution of transgenic markers of melanophores and iridophores in 185 arrows). Thus, scales isolated from the belly of asip1 mutants displayed a "dorsalized" 207 color pattern (i.e., ventral scales become nearly as dark colored as dorsal scales due to an 208 increased number of pigment cells) (Fig. 6C, D). 209 210

Rescue of CRISPR mediated mutations 211
Finally, as a key test of our model, we assess the effect of combining the knockout (KO) 212 mutant with our previously-described asip1-Tg zebrafish line overexpressing asip1 in the 213 entire body. In our model, a graded distribution of Asip1 controls the ratio of ventrally characteristically repressing melanocyte and stimulating iridophore 216 differentiation; in the dorsum, where Asip1 levels are lowest, melanophores differentiate 217 and iridophores are suppressed. We have shown that our asip1-Tg line shows a strongly 218 ventralised pigment pattern in the dorsum (Fig. 7D-F; reference), suggesting that the 219 ubiquitous Asip1 levels generated are equivalent to those in the belly region of a WT fish. 220 We predict therefore that in the background of our new asip1 KO which lacks the 221 endogenous gradient of Asip1, the pigment pattern should also be ventralised, but might, 222 if anything, show a slightly weaker phenotype due to the absence of endogenous Asip1 223 'supplementing' the transgenic Asip1 expression. This is indeed what we observed (Fig.  224   7). WT fish show the typical striped pattern (Fig. 7A), combined with a darker dorsum 225 ( Fig. 7B), and a light ventrum (Fig. 7C). The asip1-Tg zebrafish phenotype presents a 226 striped pattern that shows a mild reduction in melanophore number in the 1D and 2D 227 stripes ( Fig 7D), a light belly similar to WT fish (Fig. 7F), but a drastic reduction of dorsal 228 melanophores ( Fig. 7E) due to the ectopic overexpression of asip1 8 . In asip1 K.O. mutants 229 (Fig. 7G) the striped pattern is enhanced, with a prominent 3V stripe reaching to the head 230 ( Fig. 7F), the belly is considerably darker (dorsalised) than in WT (Fig. 7I), while the 231 dorsum remains similar to that of WT (Fig. 7 H). In the asip1 K.O. ; asip1-Tg, the asip1 K.O. 232 phenotype is suppressed and the asip1-Tg . phenotype prevails (Fig. 7J). The asip1 K.O. ; 233 asip1-Tg zebrafish do not show enhancement of the 3V stripe, but instead show a stripe 234 pattern similar to the asip1-Tg . , except that the ?2D stripe is somewhat more prominent, 235 due to a more WT melanophore count (Fig. 7 J), a light dorsum with a drastic reduction 236 of dorsal melanophore as the asip1-Tg . fish (Fig. 7K), but a light belly similar to both 237 asip1-Tg and WT fish (Fig. 7L). These observations are fully consistent with our 238 hypothesis that the graded expression of asip1 along the dorso-ventral axis is crucial to 239 establish the dorso-ventral pigment pattern and that this results from changed numbers of 240 multiple pigment cell-types. Using quantitation of expression of the xanthophore and iridophore markers, xanthine 281 dehydrogenase (xdh) and leucocyte tyrosinase kinase (ltk) respectively 27,28 , we were 282 unable to demonstrate clearly an effect on xanthophore and iridophore differentiation in 283 transgenic asip1 overexpressing fish 8 . However, these Asip1 transgenic zebrafish did 284 show an extra iridophore interstripe over D1 that we initially interpreted as simply due to 285 the enhanced visibility of underlying iridophores resulting from the lack of melanized 286 cells in the dorsal region 8 . Our new loss-of-function mutants and the rescue of CRISPR 287 induced Asip1 mutations data clearly demonstrates that Asip1 also plays a key role in 288 regulating both iridophore and xanthophore differentiation in the adult skin, suggesting 289 that the extra dorsal iridophore interstripe in Asip1 transgenic fish may, in fact, result 290 from ectopic production of iridophores as well as the absence of melanophores. 291 Our new loss-of-function data provide independent support for our suggestion 8 that Asip1 292 has no role in embryonic pigment cell development nor in larval (pre-metamorphic) 293 pigment pattern formation. However, Asip1-dependent effects on pigment pattern 294 become visible from the very earliest stages of metamorphosis (15 dpf), and then 295 progressively spread to all ventral pattern elements as they are formed during 296 metamorphic growth. We note that the timing of initiation of these effects corresponds to 297 the period when asip1 expression reaches maximum levels (at 15 dpf) and when 298 significant dorso-ventral differences in asip1 expression appear (30 dpf; 8 ). Thus, asip1 299 has a role exclusively in metamorphic and post-metamorphic pigment pattern formation. 300 Early experimental data in amphibian and fish species identified a diffusible melanization 301 inhibition factor (MIF), mainly produced by cells in the ventral skin, that inhibits 302 melanoblast differentiation, but also stimulates or supports iridophore proliferation in the 303 ventrum 29-31 . Our demonstration that absence of Asip1 results in a severe impairment of 304 ventral iridophore development strongly supports the identification of Asip1 as the 305 elusive MIF. 306 Zebrafish iridophores contribute to silver-or white-colored regions. They are classified 307 into two different types according to the size and number of guanine platelets. Type S 308 iridophores contain smaller uniform-sized platelets, but in larger numbers, than type L 309 iridophores. The abdominal wall is covered by a dense internal sheet of type S iridophore 310 5,6 . By analyzing Tg(TDL358:GFP)/asip1 K.O mutant zebrafish lines, we show that Asip1 311 loss-of-function strongly disrupts this abdominal wall iridophore sheet in the ventral 312 trunk. Our previous studies showed asip1 expression in the iridophores of the zebrafish 313 abdominal wall by in situ hybridization 8 and promoter-directed reporter expression 13 ; 314 our new data suggests that asip1 is necessary for the normal development of this 315 abdominal iridophore sheet.
It will be important to determine where, and on what cell-type, Asip1 acts to regulate 317 numbers of each pigment cell-type. Melanocyte stem cells identified in the dorsal root 318 ganglia (DRG) have been shown to generate all three pigment cell-types in the post-319 metamorphic skin of zebrafish, supporting the idea of a common pigment progenitor 32 . 320 These multipotent progenitors have been proposed under a progressive fate restriction 321 model to subsequently segregate bipotent progenitors (melanophore-iridophore, 322 melanophore-xanthophores and xanthophore-iridophore) from which individual pigment 323 cell fates become specified 32 . We propose that Asip1 levels in the skin may control the 324 fate specification of these progenitors when they arrive at the skin. Thus, high ventral 325 levels of Asip1 repress melanophore and xanthophore specification and promote 326 iridophore specification from these progenitors. In contrast, those progenitors choosing 327 the dorsal migratory route from DRG enter a low Asip1 environment and more frequently 328 become melanophores and xanthophores (Fig. 8). it has been shown that the effect of Asip1 over iridophores seems to be different in scales 340 and in the skin 29,30,31 , our data together demonstrate that Asip1 is strongly inhibitory to 341 chromatophore differentiation in the scales. Accordingly, it has been demonstrated that 342 goldfish Asip1 conditioned medium represses medaka scale pigmentation 11. Scale 343 pigmentation has been less-well studied in zebrafish, but it is thought that multipotent 344 pigment cell progenitors that populate the skin also populate the scales 32 . Further work 345 will be necessary to understand the different responses to Asip1 of these progenitors in 346 scales versus the skin, but we suggest that these reflect an evolutionarily ancestral dorsal 347 countershading mechanism that functions in association with scales, and an evolutionarily 348 derived secondary striping mechanism in deeper layers of the skin.
In conclusion, our loss-of-function experiments support and extend the results from our 350 overexpression analysis showing that the graded expression of asip1 along the dorso-351 ventral axis is crucial to establish the dorso-ventral pigment pattern in ray-finned fish. 352 Asip1 has a dramatic effect on the ancestral dorso-ventral pigment patterning process, but 353 also a more subtle control of the striping mechanism. We propose that the Asip1 gradient 354 is an environmental cue that uses the melanocortin-signaling system to bias the adoption 355 of pigment cell fates from progenitors that migrate into the skin (Fig. 8) Initial study of asip1 (sa13992), a randomly induced point mutation predicted to affect 378 splicing, failed to reveal a clear pigment pattern defect (Supp. Fig. 1 and 2). The 379 asip1 sa13992 allele was generated by random mutagenesis during a large-scale mutagenesis 380 project at the Sanger Institute 35 , and obtained from the European Zebrafish Resource 381

Center. 382
Due to uncertainties about the likely effect of compensatory mechanisms limiting the 383 impact of the predicted change in splicing in asip1 sa13992 , we to used CRISPR/Cas9 384 genome editing to engineer a likely null allele. To this end, an asip1 loss-of-function 385 mutation was generated using a CRISPR-Cas9 protocol originally adapted from Bassett 386 et al. 14  We thank Christiane Nüsslein-Volhard from Max-Planck Institute (Germany) for 573 providing the TDL358:GFP and Kita:GalTA4;UAS:mCherry transgenic lines. Also, we 574 would also like to thank Inés Pazos Garridos (CACTI, University of Vigo, Spain) for her 575 assistance with confocal imaging. This work was funded by the Spanish Economy and 576 Competitiveness Ministry projects AGL2011-23581, AGL2014-52473R, AGL2017-577 89648P to JR, and by a BBSRC SWBio DTP Studentship to JO. Partial funding was 578 obtained from AGL2016-74857-C3-3-R to JMCR. L. Cal was supported by pre-doctoral 579 fellowship FPI funded by Spanish Economy and Competitiveness Ministry (AGL2011-580