CRISPR knockouts of pmela and pmelb engineered a golden tilapia by regulating relative pigment cell abundance

Premelanosome protein (pmel) is a key gene for melanogenesis in vertebrates. Mutations in this gene are responsible for white plumage in chicken, but its role in pigmentation of fish remains to be demonstrated. In this study we found that most fishes have two pmel genes arising from the teleost-specific whole genome duplication. Both pmela and pmelb were expressed at high levels in the eyes and skin of Nile tilapia. We mutated both genes in tilapia using CRISPR/Cas9 gene editing. Homozygous mutation of pmela resulted in yellowish body color with weak vertical bars and a hypo-pigmented retinal pigment epithelium (RPE) due to significantly reduced number and size of melanophores. In contrast, we observed an increased number and size of xanthophores in mutants compared to wild-type fish. Homozygous mutation of pmelb resulted in a similar, but milder phenotype than pmela-/- mutants, without effects on RPE pigmentation. Double mutation of pmela and pmelb resulted in loss of additional melanophores compared to the pmela-/- mutants, and also an increase in the number and size of xanthophores, producing a strong golden body color without bars in the trunk. The RPE pigmentation of pmela-/-;pmelb-/- was similar to pmela-/- mutants, with much less pigmentation than pmelb-/- mutants and wild-type fish. Taken together, our results indicate that, while both pmel genes are important for the formation of body color in tilapia, pmela plays a more important role than pmelb. To our knowledge, this is the first report on mutation of pmelb or both pmela;pmelb in fish. Studies on these mutants suggest new strategies for breeding golden tilapia, and also provide a new model for studies of pmel function in vertebrates. Author Summary Melanophores, the most common pigment cell type, have been studied for nearly 150 years. Many genes are involved in melanoblast migration, melanophore differentiation, and melanin biosynthesis. Pmel is fundamental for melanosome development by directing melanin biosynthesis and melanosome phase transition. Specifically, PMEL can form a fibrillar structure within the melanosome upon which melanin is deposited. We identified two pmel genes in Nile tilapia arising from the teleost-specific whole genome duplication. Disruption of either pmela or pmelb in tilapia leads to significant hypo-pigmentation. PMEL disrupted fish showed not only a reduction in melanin and tiny melanophores, but also a significant increase in the number of xanthophores, and even guanine-filled melanophores, which led to a golden tilapia with hypo-pigmented RPE. Our study confirmed the role of pmel in melanin biosynthesis and maturation, and also highlighted its effects on melanophore number and size. These results provide new insights into pigment cell biology and will help us better understand the mechanisms of color patterning in teleosts. Knockout of both pmela and pmelb provide a new strategy for engineering a golden tilapia, which might provide a foundation for developing new strains in the tilapia industry.

number and size of xanthophores in mutants compared to wild-type fish. Homozygous 23 mutation of pmelb resulted in a similar, but milder phenotype than pmela -/mutants, without 24 effects on RPE pigmentation. Double mutation of pmela and pmelb resulted in loss of 25 additional melanophores compared to the pmela -/mutants, and also an increase in the number 26 and size of xanthophores, producing a strong golden body color without bars in the trunk. The 27 RPE pigmentation of pmela -/-;pmelb -/was similar to pmela -/mutants, with much less 28 pigmentation than pmelb -/mutants and wild-type fish. Taken together, our results indicate 29 that, while both pmel genes are important for the formation of body color in tilapia, pmela 30 plays a more important role than pmelb. To our knowledge, this is the first report on mutation 31 of pmelb or both pmela;pmelb in fish. Studies on these mutants suggest new strategies for 32 breeding golden tilapia, and also provide a new model for studies of pmel function in 33 vertebrates.

Melanophore size, number and melanin biosynthesis were decreased in pmel mutants 186
Several waves of increase in melanophore number were detected in wild-type tilapia, 187 during which melanophore populations increased to pattern the whole fish [33]. Like the 188 tetrapods, melanophores in teleosts were also able to release melanosomes extracellularly (Fig  189   S7). We chose to analyze the melanophores of wild-type and pmel mutant fish at 7 dpf, 190 shortly after melanophores first appear on the embryo and spread on both the yolk sac and the 191 trunk. The wild-type fish had many well-pigmented macro-melanophores and normal sized 192 melanophores on the head from 7 dpf (Fig 3A and 3A'). However, pigmented melanophores 193 were greatly reduced in the pmela -/- (Fig 3B' and 3B') and pmela -/-;pmelb -/mutants ( Fig 3D  194 and 3D'). The remaining melanophores were unable to synthesis mature melanosomes, and 195 most of the melanosomes were in phase II. The remaining macro-melanophores were light 196 grey, a hypo-pigmented phenotype (Fig 3A, 3A', 3D and 3D'). In contrast, although the 197 pmelb -/mutants also had fewer melanophores, some melanophores were still able to develop 198 dark, mature melanosomes ( Fig 3B and 3B'). The number of melanophores in pmela -/-;pmelb /-and pmela -/-;pmelb -/mutants was significantly lower than that of the wild-type fish. Although 200 pmelb -/mutants retained more melanophores, most of the pigmented melanophores were 201 fragmented or found in a state of pigment-aggregation. The pmela -/-;pmelb -/double mutants 202 had significantly reduced melanophore numbers and sizes compared with both wild-type fish 203 and the single gene mutants (Fig 3E and 3F). Additionally, xanthophores were detected in 204 significantly higher numbers and larger sizes than the wild-type fish at 12 dpf, which gave the 205 fish a yellowish head (Fig 4A-4C). This phenomenon probably could explain the complete 206 golden color of the whole fish at later stages. 207 Xanthophore number and size were increased in pmel mutants 208 In our previous study, we found that in wild-type fish xanthophores arose at 6 dpf, and 209 sharply increased in number by 12 dpf [33]. We used 12 dpf wild-type, pmela -/-, pmelb -/and 210 pmela -/-;pmelb -/mutants to analysis xanthophore number and size in PMEL disrupted fish. 211 The pmela -/and pmela -/-;pmelb -/mutants showed significantly more and larger sized 212 xanthophores, which was a major reason for the yellowish color of those mutants (Fig 4A-4C). 213 In wild-type fish, the xanthophores were often in a pigment-aggregated state, with a much 214 smaller size than the expanded melanophores [33]. However, in the pmela -/and pmela -/-215 ;pmelb -/mutants, the lateral top of the heads showed lots of yellow pigmentation, the 216 xanthophores were filled with pteridines/carotenoids, and the sizes were much larger than 217 wild-type fish and pmelb -/mutants. The remaining melanophores were small and hypo-218 pigmented with small sizes, similar to the 7 dpf mutants (Fig 4A). These results indicated that 219 loss of pigment biosynthesis function in melanophores led to decreased melanophore number 220 and size, and a corresponding increase in xanthophore size and number. 221 RPE, iris pigmentation and eye development were affected in pmela mutants, while only 222 iris pigmentation was affected in pmelb mutants Reduced eye pigmentation was observed in all developmental stages of pmel mutants, 224 and the pmela -/and pmela -/-;pmelb -/mutants were free of melanin in the eyes at early larvae 225 stages (before 60 dpf). Thus fish at 60 dpf and 90 dpf were used as materials for studying eye 226 pigmentation and development. Pigmentation in the eyes (including both iris and RPE) 227 displayed significant differences between the mutants and wild-type fish. The pmela -/and 228 pmela -/-;pmelb -/mutants showed significant hypo-pigmentation in eyes, the RPE was dark red, 229 and the iris was golden (Fig 5B and 5D). In contrast, the wild-type fish and pmelb -/mutants 230 had a black RPE, and a normally pigmented iris (Fig 5A and 5C). The development of eyes 231 was also altered in pmela -/and pmela -/-;pmelb -/mutants. Around 1/3 of both pmela -/and 232 pmela -/-;pmelb -/mutants had both serious reduction of iris iridescent-white pigmentation and 233 significant eye abnormal development. Pigment deposition onto the inner surface of the 234 cornea causes the appearance of the Krukenberg spindle. The area of iris covered by melanin 235 stuck to the cornea was around 1/3 compared with the central black RPE area (Fig 5B', 5D' 236 and 5E). Even though slow restoration of RPE pigmentation was detected at 90 dpf, part of 237 the RPE area still had a blood red color indicative of insufficient melanin, and the Krukenberg 238 spindle phenotype was durable ( Fig 5E). In contrast, the wild-type fish and pmelb -/mutants 239 showed no significant abnormalities of eye development (e.g. pigment stuck to the cornea), 240 even though the pmelb -/mutants hypopigmentation of the eyes. The results of global 241 pigmentation in eyes were in consistent with the phenotype we showed above (Fig 5F). 242

;pmelb -/mutants 244
Phenotypic analysis of the caudal fins suggested that pigmented melanophores increased 245 with age in the mutants. The pmela -/and pmela -/-;pmelb -/mutants had only a few pigmented 246 melanophores before 60 dpf. However, after further development, small pigmented 247 melanophores were observed in the pmel mutants. The caudal fins in wild-type, pmela -/-, 248 pmelb -/and pmela -/-;pmelb -/mutants were used as samples to study the re-pigmentation and re-patterning progress of melanophores. In wild-type fish at 60 dpf, hyper-pigmented bars 250 separated by light colored inter-bars were observed in the caudal fin ( Fig 6A). Melanophores 251 were detected in large numbers in the hyper-pigmented bar regions (Fig 6A'). In pmela -/-252 mutants at 60 dpf, yellowish caudal fins without any bars were detected (Fig 6B). No 253 pigmented melanophores were detected at this time period. What was striking was that 254 aggregated iridescent purple guanine was detected in cells with branching-like-clusters 255 sharing a shape and size similar to melanophore , both in caudal fins and even scales (Fig 6B'  256 and Fig S8). We suggest that melanophores under specific conditions (melanin-free or 257 melanin biosynthesis deficient) were able to synthesize or accumulate guanines. In pmelb -/-258 mutants at 60 dpf, no obvious bars but some pigmented melanophores were detected in the 259 caudal fins (Fig 6C and 6C'). In pmela -/-;pmelb -/mutants at 60 dpf, the whole fins were 260 yellowish without bars, and many purple-guanine-pigmented melanophores were detected 261 randomly spread across the whole fins ( Fig 6D and D'). At 90 dpf, a significant increase in 262 pigmented melanophores was detected in the caudal fin ( Fig 6E). Additionally, many red 263 erythrophores/xanthophores were detected in the fins (Fig 6E'). In pmela -/mutants at 90 dpf, 264 a significant increase in small melanophores was detected across the whole fins, and they 265 spread evenly across the fin (Fig 6F). Large numbers of evenly distributed white iridophores 266 were also detected ( Fig 6F'). In pmelb -/mutants at 90 dpf, increased numbers of 267 melanophores were detected in the caudal fins, with larger size and significant higher number 268 than the pmela -/mutants at the same developmental stage ( Fig 6G). Additionally, the white 269 iridophores also increased compared with the wild-type fish, similar to the situation revealed 270 in pmela -/mutants. In pmela -/-;pmelb -/double mutants at 90 dpf, melanophores were also 271 more abundant compared to the 60 dpf pmela -/-;pmelb -/mutants. The melanophores in the 272 double mutants were in spot-like small size, and fewer in number than in wild-type, pmela -/-273 and pmelb -/mutants at the same stage. Interestingly, the number of red 274 erythrophores/xanthophores increased in the double mutants, which contributes to a more 275 yellowish or reddish body color ( Fig 6H, H'), and the whole fish were seriously hypo-276 pigmented compared to the wild-type fish ( Fig S9). Statistical analysis of the melanophores of 277 the pmel mutants and wild-type fish at 60 and 90 dpf was consistent with the descriptions 278 above (Fig 6I and 6J). The restoration of melanin biosynthesis in pmela -/and pmela -/-;pmelb -/-279 mutants at older ages indicated that additional pathways might be involved in melanin 280

Body color of pmel mutants at different developmental stages 282
We characterized the phenotype of the mutant fish in their early developmental stages. 283 At 12 dpf, wild-type fish showed no bars, and melanophores were spread evenly across the 284 whole trunk (Fig 7A). At 60 dpf, wild-type fish showed black bars separated by light colored 285 inter-bars. A higher density of melanophores was detected in the bar regions. Iridophore 286 number sharply increased in the inter-bar regions (Fig 7B). At 90 dpf, wild-type fish showed 287 strong black bars separated by iridescent inter-bars ( Fig 7C and 7C'), much like the 150 dpf 288 and adult fish ( Fig 7M). 289 At 12 dpf, the pmela -/mutant fish showed serious hypo-pigmentation across the whole 290 fish (including RPE). Melanophore numbers were reduced on the trunk surface, and the 291 remaining melanophores were light grey, indicative of a reduction in melanin synthesis (Fig  292   7D). Almost no pigmented melanophores were observed in the mutants at 60 dpf ( Fig 7E). 293 Hypo-pigmented patterns on the dorsal fin and a red-black RPE were observed in the lateral 294 view (Fig S10). At 90 dpf, the body color was yellowish, and the iris showed obvious hypo-295 pigmentation ( Fig 7F and 7F'). 296 At 12 dpf, pmelb -/mutant fish showed hypo-pigmentation but with more pigmented 297 melanophores than the pmela -/mutant fish (Fig 7G). At 60 dpf, melanophores were present in 298 reduced numbers on the trunk surface, and although regular bars were not seen, we occasionally detected irregular patches of pigmentation (Fig 7H). At 90 dpf, the fish showed 300 yellowish body color with slightly pigmented dorsal patches compared with the pmela -/-301 mutant fish (Fig 7I and 7I'). 302 At 12 dpf, pmela -/-;pmelb -/double mutants showed a phenotype similar to the pmela -/-303 mutants. The RPE and the trunk were hypo-pigmented. Macro-melanophores with reduced 304 melanin content were observed, but no pigmented melanophores were observed on the trunk 305 ( Fig 7J). At 60 dpf, no bars or pigmented melanophores were detected in the double mutants. 306 They showed even more significant hypo-pigmentation compared with the pmela -/or pmelb -/-307 mutants, which gave the whole fish a golden color (Fig 7K). At 90 dpf, significant hypo-308 pigmentation was detected in the double mutants. No bars or black/grey patterns were 309 detected in the trunk and all the fish showing completely golden color. The iris was hypo-310 pigmented, but it was difficult to distinguish the differences of RPE pigmentation between the 311 mutants and the wild-type fish from the lateral view ( Fig 7L and 7L'). At 150 dpf, the double 312 mutants showed a golden color with reddish fins, and the yellow pigmentation and remaining 313 melanophores were spread evenly across the whole fish. The RPE was as black as the wild-314 type fish (Fig 7N). 315

Pigment cell number, morphology and location were greatly changed in pmel mutants 316
The thin and partially transparent scales as skin appendages provided us excellent 317 materials to investigate the cell number, morphology and location between each type of 318 pigment cells. The 60 dpf (before melanin biosynthesis partially restored) and 150 dpf 319 (melanin gradually accumulate over many weeks) scales of wild-type fish and pmela -/-;pmelb -320 /double mutants were used as materials to reveal the relative pigment cell number, 321 morphology and location between the wild-type fish and PMEL-free golden fish. Just as 322 predicted, the number and sizes of melanophores and xanthophores were consistent with the 323 results we showed above in different developmental stages. In wild-type fish, many melanophores were healthy and heavily pigmented with mature melanin, the dendrites were 325 active and fully spread, which made a large branched cell, especially in the older 150 dpf 326 melanophores. Additionally, iridophores were spread mainly on the surface of the mainly 327 body of melanophores. However, the xanthophores were detected with a smaller, round shape. 328 They located near the dendrites of melanophores, or even further away, probably restricted 329 and rejected by melanophores (Fig 8A and 8C). As a contrast, in scales of 60 dpf PMEL-free 330 hypo-pigmented fish, the melanophores with insufficient melanin biosynthesis were often 331 detected with melanin-free melanophores filled with iridescent purple, blue or even red 332 guanine, in both the main body and the dendrites of melanophores (Fig 8B). In the scales of 333 older 150 dpf PMEL-free golden fish, the small number of melanophores detected had a tiny-334 spot appearance, and the gradually accumulated melanin gathered as a weak spots, with a 335 much smaller size than that of the wild-type melanophores. Additionally, the xanthophores 336 were obviously enlarged with "airenemes" in the mutants compared with those of the wild-337 type. The larger sizes of xanthophores allowed the mutants to produce more yellow 338 pteridines/carotenoids. Besides, the xanthophores located near the melanophores, and some of 339 them even clung on the main body of the tiny-spot-like melanophores, indicating that the 340 nearest-neighboring distances between the melanophores and xanthophore were probably 341 reduced and the interactions of them were heavily influenced. The remaining iridophores 342 were arranged in a line between the weak melanophores, and eventually combined into a fish 343 net structure in the scales (Fig. 8D). These results suggeste that the cell-cell interactions 344 between the pigment cells (both the same type and different types of pigment cells) were 345 altered in the PMEL-free golden fish. The relative pigment cell number, morphology and 346 location (mainly between the melanophores and xanthophores) were completely different 347 between the wild-type and mutants, which was the fundamental reason for the formation of 348 golden body color in tilapia.

Pmela and pmelb showed similar expression patterns in Nile tilapia 351
Pmel is a key gene involved in melanin biosynthesis and the development of body color 352 in many fish species, such as carp [17,36], and cichlids [24,37,38]. In those studies, the 353 expression levels of pmela in hyper-pigmented skin were always significantly higher than in 354 hypo-pigmented skin. In the skin transcriptome of Malaysian red tilapia, pmela was found to 355 be expressed significant higher in black skin than in red/pink skin [37], suggesting that pmel 356 might be involved in skin color differentiation between black and red tilapia. Additionally, 357 this study suggested that pmela and pmelb were most highly expressed in eyes, indicating that 358 the two genes might be involved in eye pigmentation and development. In zebrafish, pmela 359 has been shown to be involved in eye pigmentation and anterior segment size maintenance in 360 early juvenile stages, through loss of function studies [18]. In different developmental stages 361 of wild-type Nile tilapia, pmela and pmelb were detected with similar expression patterns. 362 They were both highly expressed in early developmental stages (from 6-12 dpf), which was in 363 consistent with the several waves of melanophore proliferation identified in our previous 364 studies of color patterning in Nile tilapia [33]. However, no expression of pmela and pmelb 365 was detected at 2-4 dpf, probably because these genes are downstream in the melanogenesis 366 pathway, and follow the expression of NCCs-melanophores specification genes like mitf and 367 kita/kitlga. Moderate expression of pmela and pmelb was also detected in fish at 30-90 dpf 368 and the adult stage, suggesting that they are probably involved in melanin biosynthesis or 369 melanophore survival at later stages. The pmelb was detected with some expression in dorsal 370 fins in adult wild-type fish, which was in accordance with the results of our loss of function 371 studies. The pmela -/mutants were still observed with banding in dorsal fin at 60 dpf, while 372 pmelb -/mutants presented a more serious hypo-pigmentation of this fin. This was similar to 373 the expression analysis in the cichlid, Neolamprologus meeli, in which higher pmel expression was detected in adult dorsal fin than in the ventral part of the anal fin or the caudal 375 fin [39]. 376

Homozygous mutation of pmel genes resulted in golden color in tilapia 377
Using CRISPR/Cas9 gene editing, we successfully disrupted the expression of PMEL. 378 Different levels of hypo-pigmentation were detected in the F0 mutants [33]. As predicted, the 379 homozygous mutants of pmela, pmelb and pmela;pmelb all showed obvious hypo-380 pigmentation. All three mutants had yellowish/golden body color, especially in the double 381 mutants. The pmela -/-;pmelb -/double mutants had completely golden bodies, without any 382 black/grey patches or spots, which made them quite attractive (Fig 7N and Fig S9). However,

The effects of PMEL mutation on pigment cells relative abundance and melanin 390 biosynthesis in Nile tilapia 391
As mentioned above, loss of function of both pmela and pmelb in Nile tilapia led to a 392 complete golden skin color without any black bars or patches, but with hypo-pigmented eyes, 393 due to the increased numbers and sizes of xanthophores, and decreased numbers and sizes of 394 melanophores, as well as hypo-pigmentation reflected by deficient in melanin biosynthesis 395 (Fig 8, 9A and 9B). 396 Some of the melanin-free melanophores were able to accumulate guanine, especially in 397 early developmental stages (stages before 60 dpf). During the time period the pigmentary 398 basis for the continued presence of banding in the fins of mutant fish were bands of guanine.
We did not detect red erythrophores until at least 50 dpf in our previous studies [33], and the 400 sizes and shapes were most similar to melanophores. The 60 dpf wild-type fish also showed 401 many melanophores with different levels of black/grey melanin. Previous studies on zebrafish 402 suggested that melanophores were still detected in tyr homozygous mutants, but they did not 403 contain melanin pigment [43]. Our results are similar to those from zebrafish tyr mutants, but 404 the tilapia PMEL-free mutants displayed milder defects compared to the zebrafish tyr mutants. 405 Melanophores size was also affected by mutation of pmel, as melanophores were 406 smaller in the mutants, not only in the whole trunk and fins, but also in the scales (Fig 8D and  407   9B). This is likely to be a result of defects in forming a fibrillar structure within the 408 melanosome upon which melanin is deposited [44,45], which finally led to the appearance of 409 small melanophores with insufficient melanosome development. 410 At the same time, mutation of pmel genes caused increases in the numbers and sizes of 411 xanthophores, which plays a decisive role in the complete golden phenotype. In wild-type fish, 412 xanthophores are small and round in shape, while the neighboring melanophores are dendritic. 413 In pmela -/-;pmelb -/mutants, xanthophores were larger than in wild-type fish, while 414 melanophores were detected with spot-like shapes, not only in the trunk surface and fins, but 415 also in the scales (Fig 8C, 8D and 9B). 416 Mitf is a key transcription factor for melanin biosynthesis, and is necessary for all the 417 processes from the development of melanosomes to mature melanin release. Both pmel genes 418 have promoter sequences that suggest they are regulated by mitf. Probably pmel was critical 419 for melanosome transition from phase I to III, as we detected hypo-pigmented melanophores 420 with light grey melanin in PMEL-disrupted fish. However, we detected slow restoration of 421 melanin biosynthesis during development (Fig 9C), possibly because some melanin synthesis 422 can occur in the absence of the fibrillar substrate provided by PMEL. Even though melanin 423 could accumulate gradually over many weeks, the mutants did not produce a melanin-424 pigmented phenotype as the melanophores were small. 425 Pmela was more important than pmelb in RPE pigmentation and eye development 426 In this study, both pmela and pmelb were detected with the highest expression levels in 427 the eyes of wild-type fish. They were further proved to be involved in eye pigmentation by Body color is an important economic trait. In aquaculture, the fish with pleasant body 449 color is always preferred by consumers. For example, the koi carp and golden fish have been 450 raised as pet fish all over the world, due to their attractive body color or amazing specific 451 color patterns. In tilapia, the red tilapia is also preferred by consumers globally because of the 452 lack of black pigmentation in the trunk and peritoneum [46,47]. The establishment of a 453 golden tilapia (pmela -/-;pmelb -/mutants) would be of great significance to the tilapia 454 aquaculture industry, and might also be of interest in the pet fish market. 455

456
Fish 457 The founder strain of Nile tilapia was obtained from Prof. Nagahama (Laboratory of 458 Reproductive Biology, National Institute for Basic Biology, Okazaki, Japan). This strain has 459 been in laboratory culture for more than 20 years, and is thus largely homozygous. The

RT-PCR validation of pmela and pmelb expression
The pmela and pmelb specific primers used for reverse transcription PCR amplification 495 were designed using Primer Premier 6. The sequences are listed in Supplemental Table 1. 496 Triplicate samples were collected at 2, 4, 6, 8, 10 dpf (whole fish) and 20, 30, 60, 90 dpf (skin) 497 and adult stage (different tissues, including skin). Reverse transcription was conducted in a 498 total reaction volume of 20 μl, which included 2 μg total RNA and 2μl RT reaction mixture. 499 For PCR amplification, pmel specific primers or β-actin primers were added into the reaction 500 at the beginning of PCR and each PCR run for 34 cycles. The PCR products were separated 501 by agarose gel electrophoresis and photographed under UV illumination. 502

Establishment of pmela -/-, pmelb -/and pmela -/-;pmelb -/mutants by CRISPR/Cas9 503
The pmela and pmelb mutant fish with the highest indel frequency were used as G0 504 founders. They were raised to sexual maturity and mated with wild-type tilapia. F1 larvae 505 were collected at 10 dah and genotyped by PCR amplification and subsequent FspEI and 506 Tsp45I digestion. 507 CRISPR/Cas9 was performed to knockout pmela and pmelb in tilapia as described 508 previously [33]. Briefly, the guide RNA and Cas9 mRNA were co-injected into one-cell-stage 509 embryos at a concentration of 150 and 500 ng/µL, respectively. Twenty injected embryos 510 were collected 72 h after injection. Genomic DNA was extracted from pooled control and 511 injected embryos and used to access the mutations. DNA fragments spanning the target site 512 was amplified. The mutated sequences were analyzed by restriction enzyme digestion with 513 FspEI and Tsp45I and Sanger sequencing. 514 Heterozygous F1 offspring were obtained by F0 XY male founders mated with WT XX 515 females. The F1 fish were genotyped by fin clip assay and the individuals with frame-shift 516 mutations were selected. XY male and XX female siblings of F1 generation, carrying the 517 same mutation, were mated to generate homozygous F2 mutants. The pmela -/-, pmelb -/and 518 pmela -/-;pmelb -/mutants were screened using restriction enzyme digestion and Sanger sequencing. The genetic sex of each fish was determined by genotyping using sex-linked 520 marker (marker 5) as described previously [50]. 521

Image recording and pigment cell observation at different developmental stages 522
Larvae fish at 5, 7, 12 and 30 dpf and early juvenile stage at 60 dpf were shifted to an 523 observation dish with clean water, photographed by Olympus SZX16 stereomicroscope 524 (Olympus, Japan) under bright or transparent field with different magnification. The 90 dpf 525 wild-type and mutant fish were shifted to the same 15 × 5 × 15 cm 3 glass water tanks 526 separately, before being photographed with a Nikon D7000 digital camera (Nikon , Japan) 527 against a blue background. ACDSee Official Edition software (ACDSystems, Canada) and 528 Adobe Illustrator CS6 (Adobe Inc. USA) were used to format the pictures. 529

Pigment cell analysis of the pmela -/-, pmelb -/and pmela -/-;pmelb -/mutants 530
Larvae submerged in clean water at 7 and 12 dpf were photographed from the lateral 531 view by Olympus SZX16 stereomicroscope (Olympus, Japan) under bright field. Caudal fin 532 from fish at 60 and 90 dpf were photographed with a Nikon D7000 digital camera (Nikon, 533 Japan) against a white background. The caudal fins were removed with medical scissors, 534 soaked in 0.65% Ringers' solution and directly observed with Olympus SZX16 535 stereomicroscope (Olympus, Japan) without cover slip under transparent or bright field. 536 Scales from fish at 60 dpf and 90 dpf were soaked in 0.65% Ringers' solution under cover 537 slip, and were observed under microscope (Germany, Leica EM UC7). Image recording of 538 pigment cells was conducted as quickly as possible after putting them in the Ringers' solution 539 as the preparations are not stable. ACDSee Official Edition software and Adobe Illustrator 540 CS6 were used to format the pictures. To analyze the number of melanophores, nine fish per 541 group were anesthetized with tricaine methasulfonate (MS-222, Sigma-Aldrich, USA) and 542 immersed in 10 mg/ml epinephrine (Sigma, USA) solution for 15 min to contract melanin. 543 The sizes of the pigment cells and global pigmentation were measured using Image J software 544 [51]. GraphPad Prism 5.01 software (Graphpad, USA) was used to analyze the differences in 545 the numbers and sizes of melanophores in pmela -/-, pmelb -/and pmela -/-;pmelb -/mutants and 546 wild type fish. Data values (mean ± SD) were statistically evaluated by one-way ANOVA 547 with Duncan's post-hoc test and Student's t-test. P<0.05 was considered to be statistically 548 significant, as indicated by different letters above the error bar. ;pmelb -/mutants were yellowish with very serious hypo-pigmentation across the whole fish. 594

Data Availability Statement 627
The required links or identifiers for our data are present in the manuscript as described.