The developmental bases for species differences in adult phenotypes remain largely unknown. An emerging system for studying such variation is the adult pigment pattern expressed by Danio fishes. These patterns result from several classes of pigment cells including black melanophores and yellow xanthophores, which differentiate during metamorphosis from latent stem cells of presumptive neural crest origin. In the zebrafish D. rerio,alternating light and dark horizontal stripes develop, in part, owing to interactions between melanophores and cells of the xanthophore lineage that depend on the fms receptor tyrosine kinase; zebrafish fmsmutants lack xanthophores and have disrupted melanophore stripes. By contrast,the closely related species D. albolineatus exhibits a uniform pattern of melanophores, and previous interspecific complementation tests identified fms as a potential contributor to this difference between species. Here, we survey additional species and demonstrate marked variation in the fms-dependence of hybrid pigment patterns, suggesting interspecific variation in the fms pathway or fmsrequirements during pigment pattern formation. We next examine the cellular bases for the evolutionary loss of stripes in D. albolineatus and test the simplest model to explain this transformation, a loss of fmsactivity in D. albolineatus relative to D. rerio. Within D. albolineatus, we demonstrate increased rates of melanophore death and decreased melanophore migration, different from wild-type D. rerio but similar to fms mutant D. rerio. Yet, we also find persistent fms expression in D. albolineatus and enhanced xanthophore development compared with wild-type D. rerio,and in stark contrast to fms mutant D. rerio. These findings exclude the simplest model in which stripe loss in D. albolineatusresults from a loss of fms-dependent xanthophores and their interactions with melanophores. Rather, our results suggest an alternative model in which evolutionary changes in pigment cell interactions themselves have contributed to stripe loss, and we test this model by manipulating melanophore numbers in interspecific hybrids. Together, these data suggest evolutionary changes in the fms pathway or fms requirements,and identify changes in cellular interactions as a likely mechanism of evolutionary change in Danio pigment patterns.
An outstanding challenge for developmental biology is to elucidate the mechanisms underlying adult form, and how changes in these mechanisms produce variation within and between species. Over the last few years, substantial progress has been made in identifying genes associated with major morphological differences (Galant and Carroll, 2002; Tanaka et al.,2002; Shapiro et al.,2004). Nevertheless, determining how genetic differences are translated into morphological differences will require a thorough understanding of how cellular behaviors are altered in different genetic backgrounds. Of particular interest are the mechanisms underlying variation in traits that have relevance to human health and disease, clear adaptive significance in nature, or both.
One useful system for studying the genetic and cellular bases for variation in adult form is the pigment pattern expressed by Danio fishes(Parichy, 2003; Kelsh, 2004; Quigley et al., 2004). These patterns differ dramatically across species, and include horizontal stripes,vertical bars, spots, and uniform patterns resulting from the arrangements of several classes of pigment cells, including black melanophores, yellow-orange xanthophores and reflective iridophores. Pigment cells in teleosts and other vertebrates are derived from neural crest cells, which also contribute to neurons and glia of the peripheral nervous system, bone and cartilage of the craniofacial skeleton, adrenal chromaffin cells, endocardial cushion cells,and other tissues (Hörstadius,1950; Smith et al.,1994; Le Douarin,1999). Neural crest-derived lineages are associated with a variety of human disease syndromes (Matthay,1997; Amiel and Lyonnet,2001; Ahlgren et al.,2002; Widlund and Fisher,2003; Farlie et al.,2004) and have had major roles in the diversification of vertebrates (Gans and Northcutt,1983; Hall, 1999). Besides serving as a potential model for development and evolution of other neural crest-derived traits, pigment patterns are especially interesting because of their ecological and behavioral significance, with teleost pigment patterns having roles in shoaling, mate recognition, mate choice and predator avoidance (Endler, 1983; Houde, 1997; Couldridge and Alexander,2002; Allender et al.,2003; Engeszer et al.,2004).
One approach to identifying the genetic and cellular bases for pigment pattern diversity in danios has used hybrids between zebrafish, D. rerio, and other danio species(Parichy and Johnson, 2001; Quigley et al., 2004). Wild-type D. rerio exhibit four to five melanophore stripes(Fig. 1A,D). When crossed with other danios, hybrid offspring develop pigment patterns that typically resemble D. rerio more closely than the heterospecific danio. This finding suggested that complementation tests could be used to screen loci identified as recessive D. rerio pigment pattern mutants for contributions to pigment pattern differences between species: mutants for which hybrids have pigment patterns different from controls identify genes that may differ between species and thus identify candidates for further analyses.
A previous study used interspecific hybrids to investigate the genetic bases for the evolutionary loss of stripes in D. albolineatus, in which pigment cells are nearly uniformly dispersed(Fig. 1B,E)(Parichy and Johnson, 2001). Hybrids between wild-type D. rerio and D. albolineatusdevelop stripes similar to D. rerio but unlike D. albolineatus. By contrast, one of ten mutant loci tested yielded a non-complementation phenotype in which hybrids lacked stripes like D. albolineatus. This locus was identified as fms, which encodes a type III receptor tyrosine kinase (Parichy et al., 2000b) known previously for roles in hematopoiesis and osteoclast development (Yoshida et al.,1990; Stanley et al.,1997; Dai et al.,2002; Scheijen and Griffin,2002; Barreda et al.,2004). Analyses of D. rerio fms mutants(Fig. 1C) demonstrate that fms promotes the development of a late-appearing population of adult melanophores that differentiates from latent stem cells during the larval-to-adult transformation, or metamorphosis(Parichy et al., 2000b). fms also is essential for melanophore survival and migration into stripes, although melanophores themselves do not detectably express fms. Rather, fms is expressed by cells of the xanthophore lineage and is essential for recruiting xanthophores from latent precursors. In turn, interactions between melanophores and fms-dependent cells of the xanthophore lineage are required for melanophore stripe formation(Parichy and Turner,2003a).
In this study, we test whether changes in fms or fms-dependent cell lineages underlie pigment pattern differences between D. rerio and D. albolineatus, as well as other danios (Parichy and Johnson,2001). We first identify additional species for which pigment patterns of hybrids depend on fms, and show that stripe loss in D. albolineatus hybrids depends on fms and other modifier loci. We next ask whether pigment pattern development in D. albolineatus resembles that of fms mutant D. rerio, as would be predicted by the simplest model in which a loss of fmsactivity has contributed to the evolutionary loss of stripes in D. albolineatus. We find that melanophore deficits and behaviors in D. albolineatus are similar to fms mutant D. rerio, yet D. albolineatus exhibit a dramatic increase - rather than a decrease- in xanthophore numbers. These findings reject the simplest model in which stripe loss in D. albolineatus depends on a loss of fmsactivity and a corresponding loss of the xanthophore lineage. Finally, we use interspecific hybrids to test an alternative model in which evolutionary changes in pigment cell interactions are responsible for stripe loss. Together these results identify interspecific variation in the fms pathway or cellular requirements for fms activity, and support a model in which evolutionary changes in pigment patterns depend in part on alterations in melanophore interactions.
Materials and methods
Fish stocks and rearing conditions
Fish were reared at 28.5°C (14 hour light: 10 hour dark). Danio rerio were the inbred mapping strain ABut or a mixed genetic background comprising ABut, wikut, ekkwill and other stocks. Danio aff. albolineatus, D. choprae, D. aff. dangila, D. `hikari' and D. aff. kyathit were from Transship Discounts (Jamaica, NY, USA). The precise taxonomic status of the stocks is uncertain. Danio albolineatus were derived from stocks provided by M. McClure (Cornell University).
Complementation tests between D. rerio and heterospecific danios were performed as described (Parichy and Johnson, 2001). When mutant loci were mapped, we examined hybrid phenotypes in crosses segregating the mutant allele to control for allelic variation at other unlinked loci, and we genotyped hybrid offspring by PCR. Danio rerio mutants used for interspecific complementation tests have been described: sox10 (colourless)(Dutton et al., 2001); endothelin receptor b1 (ednrb1, roseb140)(Parichy et al., 2000a); tfap2a (lockjawts213)(Knight et al., 2003; Knight et al., 2004); mitfa (nacrew2)(Lister et al., 1999); jaguarc7 (Fisher et al., 1997); and pumaj115e1(Parichy and Turner, 2003b; Parichy et al., 2003). Additional mutants were derived from on-going mutagenesis screens (D.M.P.,unpublished).
Nomenclature for pigment pattern elements
Previous studies defined elements of the adult pigment pattern in D. rerio (Parichy and Johnson,2001; Parichy and Turner,2003b), including the first developing or `primary' melanophore stripes (1D, 1V) that develop dorsal and ventral to the horizontal myoseptum,as well as later-developing `secondary' dorsal and ventral melanophore stripes(2D, 2V). We refer to the xanthophore-rich areas between melanophore stripes as `interstripe' regions.
fms genotyping was accomplished by primer extension assays using conditions described (Parichy and Turner,2003a). A 100 bp product was amplified from genomic DNA using forward and reverse primers flanking the fmsj4e1 mutant lesion (fms-j1f, ACT CTT GGT GCT GGT GCG TTT G; fms-j1r, CTT TGA GCA TTT TCA CAG CC) (Parichy et al.,2000b). Wild-type D. rerio or D. albolineatus fms alleles result in the addition of two nucleotides (ddCA), whereas the D. rerio fmsj4e1 allele results in addition of four nucleotides (ddCTTA) to the extension primer (fms-j1r). Genotyping methods for other loci used in interspecific complementation tests are available on request.
In situ hybridization and histology
Methods for in situ hybridization, as well as tyrosinase assays and controls followed those described previously(Quigley et al., 2004).
Imaging and quantitative analysis
We examined melanophore behaviors by imaging individual larvae once-daily or twice-daily beginning when melanophores first appear outside of early larval melanophore stripes (∼14 days post-fertilization, dpf)(Parichy et al., 2000b; Parichy and Turner, 2003b),through development of the adult pigment pattern (46 dpf; once-daily series)or middle stages of pigment pattern metamorphosis (35 dpf; twice-daily series). Images were acquired with a Zeiss Axiocam HRc digital camera mounted on an Olympus SZX12 stereozoom microscope then transferred to Adobe Photoshop for analysis with FoveaPro 3.0 (Reindeer Graphics).
To quantify melanophore numbers, three regions of the anterior flank in both D. rerio and D. albolineatus were defined to represent the location of the ventral primary melanophore stripe, the primary interstripe, and the region populated by dorsal and scale-associated melanophores in D. rerio. Regions were defined by measuring the height (h) of the flank at the anterior margin of the anal fin; the measurement areas were then placed 0.5 h anterior from this location, and extending 0.25 h further anteriorly. Regions were located dorsoventrally as functions of h, according to preliminary analyses of D. rerio and dorsoventral margins of each region were 0.1 h. All melanophores were counted within these regions for each day of imaging using semi-automated feature recognition. Melanophore densities were calculated according to the areas of each region. We examined three to six individuals of each species per image series.
To examine melanophore behaviors, we followed individual melanophores throughout pigment pattern metamorphosis(Parichy et al., 2000b; Parichy and Turner, 2003b). This approach allows quantification of new melanophores that arise by differentiation or proliferation ('births') and loss of melanophores by death or de-differentiation; we refer to losses as `deaths' based on additional histological evidence, although we cannot formally exclude the possibility that some melanophores disappear by de-differentiation.
We assessed melanophore movements in twice-daily image series by determining the relative dorsal-ventral position of each melanophore followed,with the dorsal edge of the flank receiving a value of 0, and the ventral edge of the flank receiving a value of 1. We then examined the distances moved by melanophores relative to flank height, and calculated net dorsal-ventral changes in melanophore position as the difference between final and initial positions. Negative dorsal-ventral changes reflect dorsal movements, whereas positive dorsal-ventral changes reflect ventral movements. Total movements were calculated as the absolute values of these displacements. In once-daily image series, we overlayed sequential images that had been rescaled to correct for growth and aligned to minimize overall melanophore displacements and we calculated changes in melanophore position in any direction as proportions of flank height. In both approaches, using relative as opposed to absolute distances controls to some degree for passive movements due to growth, but cannot control entirely for potential differences in growth pattern between species. Thus, we further verified the magnitude of melanophore movements between species by examining relative changes in melanophore position that cannot be accounted for simply by passive movements. These rearrangements are consistent with quantitative analyses and are most easily viewed in animations(see below).
Statistical analyses were performed with JMP 5.0.1a for Macintosh (SAS Institute, Cary NC, USA). Residuals were examined for normality and homoscedasticity (Sokal and Rohlf,1994). Total melanophore numbers, melanophore births and melanophore deaths were examined by nested analyses of variance, in which individuals were nested within species and day of development was treated as a categorical variable and main effect. Births and deaths were square-root transformed prior to analyses to normalize residuals. Melanophore movements were examined by nested analyses of variance or covariance in which species differences were tested after controlling for variation among individuals(nested within species). To assess species differences in absolute movements of melanophores, we calculated absolute values for net directional movements. To assess species differences in directional melanophore movements, we controlled additionally for variation among anteroposterior regions (anterior,middle, posterior; nested within individuals), and we treated melanophore starting position as a covariate; dorsal and ventral regions of the flank were analyzed separately owing to differences suggested by preliminary analyses. Absolute and directional movements were arcsine-transformed prior to analyses. Least squares means from these analyses are reported below. Alternative parameterizations of statistical models yielded qualitatively similar results.
Phylogenetic relationships were reconstructed from mitochondrial 12S and 16S rDNA sequences (12S: H1478, 5′-TGA CTG CAG AGG GTG ACG GGC GGT GTG T-3′; L1091, 5′-AAA AAG CTT CAA ACT GGG ATT AGA TAC CCC ACT AT-3′; 16S: 16Sar-L, 5′-CGC CTG TTT ATC AAA AAC AT-3′;16Sbr-H, 5′-CCG GTC TGA ACT CAG ATC ACG T-3′)(Kocher et al., 1989; Palumbi et al., 1991). Analyses were performed as described(Quigley et al., 2004) using PAUP* 4.0b10 and MrBayes(Huelsenbeck and Ronquist,2001; Swofford,2002).
Comparative fms dependence of hybrid pigment patterns
Hybrids between wild-type D. rerio and D. albolineatusdevelop stripes, whereas hybrids between fms mutant D. rerioand D. albolineatus lack stripes(Parichy and Johnson, 2001). Given this association between fms and hybrid stripe loss in D. albolineatus, and the critical role for fms in melanophore stripe formation in D. rerio, we examined the fms dependence of hybrid pigment patterns for other danios. For each species, melanophore patterns of control hybrids with wild-type D. rerio resembled D. rerio more closely than the heterospecific danio(Fig. 2).
Our analyses reveal fms-dependent hybrid pigment patterns for additional taxa (Fig. 2A). For both D. albolineatus and D. aff. albolineatus,fmsj4e1 hybrids lacked stripes(Fig. 2B-G). For D.`hikari', fmsj4e1 hybrids developed normal melanophore stripes but fewer xanthophores than controls(Fig. 2B-J, Fig. 3A,B). Within the clade that includes D. rerio, D. kyathit and D. nigrofasciatus,fmsj4e1 hybrids did not differ consistently from controls(Fig. 2K-M)(Parichy and Johnson, 2001)(data not shown). Finally, D. choprae fmsj4e1 hybrids exhibited disrupted adult stripes and a severe xanthophore deficit(Fig. 2N-P, Fig. 3C-H), but D. dangila fmsj4e1 hybrids were not discernibly different from controls(Fig. 2Q-S).
Thus, four out of seven danios exhibited noncomplementation phenotypes with fmsj4e1 mutant D. rerio(Fig. 2A); one of these was mild (D. `hikari') and three were severe (D. albolineatus,D. aff. albolineatus, D. choprae). Whether mild non-complementation for D. `hikari' represents a unique derived change for this species, or a basal change within the D.`hikari'-D. albolineatus clade is uncertain without deeper taxonomic sampling. Given the close phylogenetic relationship of D. albolineatus and D. aff. albolineatus, these results imply at least two evolutionary changes resulting in severe noncomplementation phenotypes, one separating D. albolineatus-D. aff. albolineatus from other danios, and another separating D. choprae from other danios.
fms-dependent hybrid stripe disruption in D. albolineatus
Given the broader variation in fms-dependence across danios, we sought to further test evolutionary roles for fms and fms-dependent pathways, focusing on D. albolineatus because of the simplicity of its pattern. Previous analyses tested D. albolineatus hybrids for non-complementation of fmsj4e1, fmsj4e3 and fmsj4blue, all of which are recessive in D. rerioand exhibit presumptive null phenotypes(Parichy et al., 2000b; Parichy and Johnson, 2001). Hybrids for fmsj4e1 and fmsj4e3 lacked stripes, whereas hybrids for fmsj4blue developed stripes. Since fmsj4e1 and fmsj4e3 were maintained in the inbred AB* (ABut) genetic background,whereas fmsj4blue was maintained in a different background, the formal possibility exists that other loci in the ABut background were responsible. Alternatively, modifier loci affecting the penetrance of a fms effect could differ across backgrounds. Thus, we asked whether pigment pattern variation in hybrids with D. albolineatus segregates with alleles at the fmslocus.
Our analyses support a model in which stripe disruption in hybrids depends on fms, with the magnitude of this effect determined by additional modifier loci. We generated heterozygous fms mutant D. rerioby crossing fmsj4e1, maintained in the inbred background ABut, with another inbred mapping strain, wikut. We then crossed these fmsj4e1(AB)/fms+(wik)D. rerio to D. albolineatus. Hybrid offspring segregated two phenotypes in ∼1:1 ratios: either well-organized, `strong' melanophore stripes, or a poorly organized, `weak' stripe pattern, with significantly fewer melanophores and xanthophores (Fig. 4A-F). We categorized fish into alternative `strong' and `weak'stripe classes, then asked whether individuals carried the fms+(wik) wild-type D. rerio allele or the fmsj4e1(AB) mutant D. rerio allele. Primer extension genotyping for the fmsj4e1 lesion demonstrates that, in every instance, hybrids with `strong' stripes carried the wild-type allele whereas hybrids with `weak' stripes carried the mutant allele (n=105; Fig. 4G,H).
These data confirm that a hybrid non-complementation phenotype segregates with fms. Yet, this phenotype is less severe than that of fmsj4e1 (Fig. 2D) and fmsj4e3 in the AButbackground, and more severe than that for fmsj4blue in a different background (Parichy and Johnson,2001). This suggests that modifier loci contribute to the phenotype, and that these modifiers differ across genetic backgrounds.
Segregation analyses thus place the non-complementing locus in the vicinity of fms, and suggest roles for modifier loci in determining hybrid pigment patterns.
Temperature-sensitive fmsallele confirms role in hybrid stripe loss
We used a temperature-sensitive fms allele to further confirm the requirement for fms in D. albolineatus hybrid pigment pattern development. Segregation analyses placed the non-complementing locus within ∼1 cM of fms, a region likely to include several other genes. We reasoned that a fms allele demonstrated previously to exhibit temperature sensitivity could be used to exclude roles for these neighboring loci: a fms-specific effect should be manifested as a complementation phenotype at the permissive temperature, and a noncomplementation phenotype at the restrictive temperature. Thus, we used the temperature-sensitive allele fmsut.r4e174A(fms174), which exhibits a wild-type phenotype at 24°C and a fms null phenotype at 33°C(Parichy and Turner, 2003a). We crossed homozygous fms174 mutant D. rerio to D. albolineatus and reared hybrid siblings at either 24°C or 33°C. Tester fms174 hybrids reared at 24°C were indistinguishable from control hybrids(Fig. 5A,B), as were wild-type hybrids reared at 33°C (data not shown). By contrast, fms174 hybrids reared at 33°C developed poorly organized melanophore stripes and fewer xanthophores(Fig. 5C,D). These results provide compelling additional evidence that hybrid non-complementation phenotypes depend on fms, rather than on other closely linked loci.
Finally, D. albolineatus hybrids did not reveal noncomplementation phenotypes for any of 16 other recessive D. rerio mutants affecting pigment cell numbers and arrangements (ednrb1b140, kitb5, mitfaw2, sox10ut.r13e1, tfap2ats213, cezanneut.r17e1, degasut.r18e1, jaguarc7, leopardt1, oberonj198e1, pissarrout.r8e1, picassout.r2e1, primrosej199, pumaj115e1, or seuratut.r15e1),including bonaparteut.r16e1, which has melanophore and xanthophore defects similar to fms mutants (D.M.P. and E.L.M.,unpublished). Thus, D. albolineatus hybrid pigment patterns are uniquely fms-dependent within this broader collection of loci required for pigment pattern development.
Altered melanophore lineage development during D. albolineatus adult pigment pattern formation
Genetic analyses above reveal a strong fms dependence of hybrid pigment pattern development for D. albolineatus (as well as D. choprae) but do not indicate how this dependence reflects natural variation between species. Given the noncomplementation phenotype of fms hybrids, we reasoned initially that D. albolineatusmight exhibit a loss of fms activity relative to D. rerio. This simple model predicts that pigment pattern metamorphosis in D. albolineatus should resemble that of fms mutant D. rerio. By comparison to wild-type D. rerio, fms mutants have fewer metamorphic melanophores, increased melanophore death, decreased melanophore movement into stripes, and an absence of xanthophores(Parichy et al., 2000b; Parichy and Turner, 2003a). We thus asked whether D. albolineatus pigment pattern metamorphosis entails some or all of these differences relative to wild-type D. rerio. Our analyses reveal dramatic differences between wild-type D. rerio and D. albolineatus(Fig. 6; see Movies 1-4 in supplementary material). Although some similarities are seen between D. albolineatus and fms mutant D. rerio, there are major differences as well (next section).
We first examined the temporal accumulation of melanophores. Daily image series show that D. albolineatus exhibited only ∼67% as many melanophores as D. rerio (F1,142=245.56, P<0.0001) across the entire image series. The deficit was most evident where melanophore stripes form in D. rerio, and became increasingly severe at later stages (Fig. 6, Fig. 7A). Only where the primary interstripe forms in D. rerio were melanophore numbers greater in D. albolineatus, reflecting the more uniform pattern of melanophores (Fig. 7B).
Fewer metamorphic melanophores in D. albolineatus could reflect a change in melanophore specification, increased rates of melanophore or melanoblast death, or both. To distinguish among these possibilities, we first quantified the appearance and disappearance of melanophores in a temporally higher resolution series of images taken at 12-hour intervals. Averaged over entire series, the D. albolineatus larvae exhibited ∼91% as many melanophore `births' as the D. rerio larvae, but 376% as many`deaths' (Fig. 7C); suggesting that fewer melanophores in D. albolineatus does not reflect a failure to recruit stem cells into the melanophore lineage, but to some extent a reduction in the subsequent survival of these cells. Consistent with this inference, numbers of melanophore precursors were not obviously reduced in D. albolineatus, as revealed by molecular markers(Fig. 8A-H) and tyrosinase activity (Fig. 8I-L). Moreover, D. albolineatus melanophores frequently appeared, then disappeared,over short time intervals (Fig. 8M-O) and these cells, as well as l-dopastained,tyrosinase+ melanophore precursors, were common within the epidermis and in `extrusion bodies' at the epidermal surface, characteristic of teleost melanophore death (Parichy et al., 1999; Sugimoto,2002; Parichy and Turner,2003a) (Fig. 8P,R-T); only ∼10% as many epidermal, melanized cells occurred in D. rerio larvae (Fig. 8Q). These results indicate that fewer melanophores in D. albolineatus result, at least in part, from the death of these cells and their immediate precursors.
Finally, we examined melanophore migration. Total distances moved by melanophores were reduced in D. albolineatus larvae compared with D. rerio larvae (Fig. 7D). Moreover, stripes in D. rerio develop in part through the directional migration of initially more dispersed melanophores to sites of stripe formation, and such movements were significantly reduced in D. albolineatus (Fig. 7E,F). Animations of developing larvae further support the interpretation that D. albolineatus melanophores move less than D. rerio melanophores (see Movies 5, 6 in supplementary material).
Melanophore morphogenesis in D. albolineatus thus resembles melanophore morphogenesis in fms mutant D. rerio, with fewer melanophores, increased death of cells in the melanophore lineage, and reduced melanophore migration as compared with wild-type D. rerio.
Enhanced xanthophore development in D. albolineatus
Melanophore morphogenesis in D. albolineatus is consistent with a model in which this species has evolved a loss of fms activity relative to wild-type D. rerio. This model also predicts that D. albolineatus should have fewer xanthophores, consistent with the reported absence of xanthophores in adult D. albolineatus(McClure, 1999). Yet, our analyses reject this notion: instead, we find that D. albolineatusactually have many more xanthophores than wild-type D. rerio.
During pigment pattern metamorphosis, D. albolineatus had greater numbers of xanthophores and these cells were distributed more widely than in D. rerio, in which xanthophores initially occur only near the horizontal myoseptum, and the dorsal and ventral margins of the flank(Fig. 9A-D,F,G). Xanthophores persist in older larvae and adult D. albolineatus, and are interspersed with melanophores (see Fig. S1 in supplementary material). Moreover, control hybrids between D. albolineatus and wild-type D. rerio had an intermediate number of xanthophores relative to parental species (Fig. 9E,H),in contrast to the severe xanthophore deficiency of fmsj4e1 mutant hybrids(Fig. 4D). Thus, xanthophore development is enhanced in D. albolineatus and this trait is dominant in hybrids but highly sensitive to reduced fms activity. Interestingly, D. choprae similarly exhibit enhanced xanthophore development and a strong fms noncomplementation phenotype(Fig. 2P; see Fig. S2 in supplementary material).
To further assess xanthophore development in D. albolineatus, we examined the distributions of cells expressing molecular markers of the xanthophore lineage. Between the epidermis and myotomes, where the adult pigment pattern develops, precursor distributions were similar between species(Fig. 9I,K). These data suggest that species differences in xanthophore development reflect differences in terminal differentiation of widely distributed precursors, rather than differences in the abundance or patterning of precursors themselves. In medial locations, however, precursors were more abundant in D. albolineatusthan D. rerio (Fig. 9J,L).
The many xanthophores and xanthophore precursors in D. albolineatus suggest that fms continues to be functional in this species. Consistent with this possibility, we could not detect differences in fms expression between D. rerio and D. albolineatusduring or after pigment pattern metamorphosis(Fig. 9M,N), and gross lesions are not apparent in the fms coding sequence(Parichy and Johnson,2001).
These results show that D. albolineatus develop xanthophores in greater numbers and over a broader area than D. rerio, tending to exclude a model in which the evolutionary loss of stripes in D. albolineatus results simply from a loss of fms activity.
Evolutionary changes in cell-cell interactions during pigment pattern formation
The persistence of xanthophores in D. albolineatus led us to seek other explanations for the similarity of melanophore behaviors between this species and fms mutant D. rerio. In wild-type D. rerio, melanophore survival and organization into stripes depends on interactions between melanophores and fms-dependent cells of the xanthophore lineage (Parichy and Turner,2003a), as well as interactions among melanophores. For example,the D. rerio leopard gene mediates both heterotypic interactions between melanophores and xanthophores, and homotypic interactions between melanophores (Maderspacher and Nusslein-Volhard, 2003), which we refer to collectively as`melanophore interactions'. The nature of these interactions is not yet known,but could include direct contacts between melanophores, xanthophores, or their precursors; alternatively, interactions could be indirect, involving secreted signaling molecules, trophic factors, or even intermediary cell types. Whatever their mechanism(s), the nearly uniform pigment pattern of D. albolineaneatus with interspersed melanophores and xanthophores(Fig. 1D; see Fig. S1C in supplementary material) and the irregular stripes of wild-type D. rerio ×D. albolineatus hybrids (compared with other danios, Fig. 2) resemble different D. rerio mutant alleles of leopard(Asai et al., 1999), as well as jaguar (obelix), which contribute to homotypic interactions among melanophores (Maderspacher and Nusslein-Volhard, 2003). Thus, we hypothesized that instead of a loss of xanthophores, stripe absence in D. albolineatus might reflect changes in melanophore interactions. In principle, a species difference in melanophore interactions could be revealed with genetic mosaics(Parichy and Turner, 2003a; Quigley et al., 2004), but incompatibilities during early embryogenesis have so far precluded cell transplantations between D. albolineatus and D. rerio(D.M.P., unpublished). Thus, we used an alternative approach.
We reasoned that variation in melanophore interactions would be revealed if melanophore numbers were reduced (by analogy with reduced xanthophores in fms mutant D. rerio and hybrids with D. albolineatus): with fewer melanophores, strong interactions should allow the emergence of an organized pattern of stripes or spots, whereas weak interactions should result in a failure to organize such pattern elements. To achieve this, we used the D. rerio mutant, duchamput.r19e1. A single mutant allele for duchamp reduces melanophores in heterozygous D. rerio to∼45% that of wild-type, yet the remaining melanophores form well-organized spots (Fig. 10A,B); D. rerio homozygous for duchamp exhibit more dispersed melanophores(see Fig. S3 in supplementary material). We predicted that for species with melanophore interactions equivalent to D. rerio, duchamp hybrids should develop spots similar to heterozygous duchamp mutant D. rerio. For species with weaker melanophore interactions than D. rerio, tester duchamp hybrids should fail to generate organized pattern elements and could exhibit more severe melanophore deficiencies.
Phenotypes of tester duchamp hybrids support a model in which variation in melanophore interactions contributes to pigment pattern differences among danios. Within the clade that includes D. rerio,duchamp hybrids for both D. kyathit and D. nigrofasciatus developed spots or stripes of melanophores(Fig. 10C-F). By contrast, duchamp hybrids for D. albolineatus failed to develop organized spots or stripes and also had a reduction in melanophore numbers that became increasingly severe as fish grew(Fig. 10G,H, Fig. 11C,D). duchamphybrids with D. `hikari' developed an intermediate phenotype between D. albolineatus and other danio hybrids, in which melanophores failed to organize into spots and remained either dispersed or in a reticulated pattern (Fig. 10I,J). Finally, duchamp hybrids for D. choprae and D. dangiladeveloped clusters of melanophores similar to D. rerio(Fig. 10K-N, Fig. 11A,B).
These analyses demonstrate that hybrid pigment patterns exhibit differential sensitivity across species to the duchamp mutant defect,with the greatest sensitivity in D. albolineatus. These findings are consistent with a model in which variation in melanophore interactions contribute to pigment pattern variation among danios.
Danio pigment patterns are emerging as a useful system for understanding both the development and evolution of adult form in vertebrates. Our study suggests a model for how evolutionary changes in pigment cell development have generated naturally occurring variation in this ecologically and behaviorally significant trait.
Evolution of melanophore patterning in Danio
Melanophore patterns vary markedly among danios, and our analyses demonstrate dramatic differences in melanophore morphogenesis underlying the uniform pattern of D. albolineatus and the striped pattern of D. rerio. Fewer melanophores accumulate during pigment pattern metamorphosis in D. albolineatus, largely due to increased death of melanophores and their immediate precursors. This late block in melanophore development resembles that seen in Astyanax cavefish(McCauley et al., 2004), but contrasts with an early block affecting melanophore specification in D. nigrofasciatus, which accounts for an equivalent total melanophore deficit as compared with D. rerio(Quigley et al., 2004). Interestingly, the mode of melanophore loss in D. albolineatusresembles that of kit mutant D. rerio(Parichy et al., 1999),raising the possibility of a difference in kit signaling between species. Finally, we also demonstrate that D. albolineatus melanophores move little and thus do not coalesce into distinctive stripes as in D. rerio. In these respects, pigment pattern metamorphosis of D. albolineatus resembles that of fms mutant D. rerio. However, this is where the similarities end, as D. albolineatusretain large numbers of xanthophores, in stark contrast to fms mutant D. rerio.
We propose a model in which changes in melanophore interactions underlie the evolutionary loss of stripes in D. albolineatus. In D. rerio, stripe formation depends on interactions between melanophores and fms-dependent cells of the xanthophore lineage, as well as on interactions among melanophores; in the absence of such interactions,initially dispersed melanophores fail to migrate into stripes and melanophore death is increased (Maderspacher and Nusslein-Volhard, 2003; Parichy and Turner, 2003a). Genetic analyses of D. albolineatus initially suggested that changes in melanophore behaviors might result from a fms-dependent loss of xanthophores. Yet, the persistence of xanthophores excludes this model. Rather, we favor an alternative scenario involving changes in interactions between melanophores and xanthophores, or between melanophores themselves. This model does not exclude the possibility that differences outside of pigment cell lineages also influence species differences in melanophore morphogenesis, either directly, or by modulating the competence of pigment cells to interact with one another.
Several lines of evidence support a model in which the loss of stripes in D. albolineatus results at least partly from changes in melanophore interactions. First, we find an interspersed arrangement of D. albolineatus melanophores and xanthophores, which resembles leopard mutant D. rerio that are defective for melanophore-melanophore and melanophorexanthophore interactions, as well as jaguar (obelix) mutant D. rerio that are defective for melanophore-melanophore interactions(Maderspacher and Nusslein-Volhard,2003). Moreover, hybrids between D. albolineatus and semi-dominant jaguarc5 mutant D. rerio develop uniform pigment patterns like D. albolineatus that are qualitatively more severe than the heterozygous jaguarc5 pigment pattern(Parichy and Johnson, 2001),suggesting a difference between species in the jaguar pathway or in jaguar-dependent cellular interactions. However, hybrids between D. rerio carrying a recessive jaguarc7 deficiency and D. albolineatus are indistinguishable from control hybrids,suggesting the jaguar gene of D. albolineatus is not grossly hypomorphic compared to wild-type D. rerio (this study, data not shown).
Second, phenotypes of hybrids between fms mutant D. rerioand D. albolineatus are consistent with evolutionary changes in melanophore interactions. In these hybrids, melanophore patterns more closely resemble the uniform pattern of D. albolineatus. fms acts autonomously to the xanthophore lineage in promoting melanophore stripe organization in D. rerio (Parichy and Turner, 2003a). Thus, the dramatically fewer xanthophores in hybrids between D. albolineatus and fms mutant D. rerio may phenocopy the actual difference between species; i.e. melanophore interactions likely to have been lost evolutionarily are restored in the control hybrid, but abrogated by the reduced xanthophore number of the fms mutant hybrid (14% of control; Fig. 4).
Third, phenotypes of hybrids between D. albolineatus and duchamp mutant D. rerio are consistent with interspecific changes in melanophore interactions. When we use the duchamp mutant allele of D. rerio to reduce melanophore numbers in hybrids, clusters of melanophores form on the flank of D. rerio and hybrids of four other species, but not D.albolineatus. The failure of D. albolineatus hybrids to organize melanophore spots does not reflect phylogenetic distance from D. rerio, as spots were formed in hybrids of the more distantly related D. dangila and D. choprae. Nor does the absence of spots result merely from a low starting number of melanophores in D. albolineatus, as the adult melanophore number is indistinguishable between this species and D. nigrofasciatus (Quigley et al.,2004). Finally, the lack of spots does not result from a higher growth rate that might carry melanophores passively away from one another, as D. dangila hybrids develop spots, despite growing more rapidly than D. albolineatus hybrids (D.M.P., unpublished). Our finding that duchamp hybrids for D. `hikari' have a phenotype intermediate to D. albolineatus and the other danios suggests that changes in melanophore interactions may have contributed to overall differences between the albolineatus-'hikari' clade and other danios,with a more extreme difference in D. albolineatus conceivably responsible for the absence of stripes in this species. Identification of the duchamp gene product will allow a fuller analysis of roles for this gene and pathway in melanophore interactions and their evolution.
Our analyses and previous studies of D. rerio highlight the potentially important role that intercellular interactions are likely to play in the development and evolution of neural crest derivatives. Such interactions have started to be characterized within melanocyte, neurogenic and rhombomeric lineages as well(Aubin-Houzelstein et al.,1998; Hagedorn et al.,1999; Trainor and Krumlauf,2000; Paratore et al.,2002; Hou et al.,2004). In principle, melanophore interactions could involve factors secreted or presented at the cell surface(Wehrle-Haller, 2003; Hou et al., 2004), or adhesive or junctional contacts among pigment cells or their precursors(Twitty, 1945; Tucker and Erickson, 1986; Parichy, 1996), although increased stratification of skin and pigment cell locations may preclude some direct contacts in adults (Hirata et al.,2003). By extension, evolutionary changes could reflect modifications to the interactions, if the competence to provide or receive signals is altered. Or, the same outcome could be effected by changes to the cellular context in which these interactions occur. For instance, simply the increased number of xanthophores in D. albolineatus may interfere with melanophore-melanophore interactions. In support of this notion,melanophores that become isolated within xanthophore-rich interstripe regions typically are lost in D. rerio(Goodrich et al., 1954; Parichy and Turner, 2003b),whereas in the anal fin of D. albolineatus, melanophores and xanthophores appear in temporally and spatially distinct waves, and a narrow melanophore stripe develops (Goodrich and Greene, 1959). Whatever their mechanisms, interactions within and among pigment cell classes suggest a rich source of variation for the evolutionary diversification of pigment patterns, without the necessity of correlated changes in other cell and tissue types. Such an independence of pigment pattern variation from other traits may in turn explain rapid and extensive pigment pattern evolution across species that are otherwise relatively similar in form.
Evolution of fms activity and function during xanthophore development
Interspecific complementation tests initially suggested that D. albolineatus could have reduced fms activity compared with D. rerio, as hybrids between wild-type D. rerio develop well-organized stripes, whereas hybrids with fmsj4e1mutant D. rerio either lack stripes or have poorly formed stripes,depending on genetic background (this study)(Parichy and Johnson, 2001). Despite this non-complementation phenotype, our analyses do not support a model in which D. albolineatus have evolved a loss of fmsactivity. Rather, these data suggest that species differ in their dosage sensitivity for fms during xanthophore development, or that evolutionary changes have occurred that affect molecular interactions within the fms pathway.
Despite a reported absence of xanthophores in adult D. albolineatus (McClure,1999), we find that larvae develop more xanthophores, these cells and their precursors arise in a broader range of locations than in D. rerio, and also persist into the adult. These observations contrast with the simplest loss-of-function model, in which xanthophores should be absent or reduced.
The excess of xanthophores in wild-type control D. rerio× D. albolineatus hybrids, and the reduction of xanthophores in tester fms mutant D. rerio ×D. albolineatushybrids reveals a genetic interaction not predicted by the recessive fms mutant phenotype in D. rerio. We envisage at least two complementary explanations for this interaction. First, xanthophore development may differ between species in its sensitivity to changes in Fms signaling. Genetic background effects are well known for mouse melanocyte mutants, including the structurally and functionally similar Kitlocus, and variable degrees of haploinsufficiency have been associated with modifier loci for mutants affecting neural crest derivatives more generally(Lamoreux, 1999; Ingram et al., 2000; Rhim et al., 2000; Nadeau, 2001; Cantrell et al., 2004). Conceivably, the more rapid and extensive development of xanthophores in D. albolineatus (and D. choprae) could entail a greater,continuous requirement for fms, such that any deficit results in pigment pattern defects.
A second explanation for fms hybrid non-complementation phenotypes lies in interspecific structural differences in fms, its ligand, csf1, or both. As fms and csf1 each act as dimers(Li and Stanley, 1991; Carlberg and Rohrschneider,1994; Ingram et al.,2000), signaling could be reduced if interspecific receptors or ligands dimerize less efficiently, or if structures of receptors and ligands co-evolve so that mismatched receptor-ligand pairings function less efficiently. To illustrate this point, we can imagine an extreme model in which any species mismatch in a receptor-ligand pairing ablates signaling. In a wild-type hybrid, functional receptor-ligand pairings would drop to one-eighth that of parental species (i.e. each species' pairing would comprise one-sixteenth of all combinations in the hybrid individual). In the tester fms mutant hybrid, functional receptor-ligand interactions would drop to one-sixteenth that of parental species. Given a fixed threshold of dosage sensitivity (here, between one-eighth and one-sixteenth of maximal), this model can easily account for the noncomplementation phenotype of tester fms mutant hybrids.
By extension, our analyses suggest evolution of cellular requirements for fms or rapid evolution of genes within the fms pathway. In fms hybrids, we observed strong noncomplementation phenotypes for D. albolineatus, D. aff. albolineatus and D. choprae, a weak non-complementation phenotype for D. `hikari',but complementation indistinguishable from wild-type D. rerio for D. kyathit, D.nigrofasciatus and D. dangila. These findings suggest most parsimoniously that: (1) changes have occurred that differentiate the D. albolineatus-D. `hikari' clade from the D. rerio-D. kyathit-D. nigrofasciatus clade; and (2)additional changes have occurred in the lineages leading either to D. choprae or D. dangila. This interspecific variation is striking,as similar tests across danios have failed to reveal noncomplementation for more than a dozen other D. rerio pigment pattern mutants, including the structurally and functionally similar kit locus(Parichy and Johnson, 2001)(D.M.P., unpublished). Thus, the fms pathway may be particularly useful for investigating the evolution of genetic dominance and developmental robustness, as well as the co-evolution of gene products within molecular pathways (Meir et al., 2002; Nijhout, 2002; Kondrashov and Koonin,2004).
Finally, this study reveals enhanced xanthophore development in both D. albolineatus and D. choprae compared with wild-type D. rerio. These findings raise the possibilities of differences in the distribution or abundance of csf1, or quantitative (as distinct from constitutive) gains of fms function in D. albolineatus and D. choprae compared with other danios. The latter possibility is opposite to initial predictions of allelic strengths(Parichy and Johnson, 2001),but is not inconsistent with the models proposed above. Immunohistochemical,transgenic, and other approaches should allow distinguishing between these models.
Thanks to C. Lee for help rearing fish, as well as M. Reedy, J. Wallingford and two anonymous reviewers for insightful discussions or comments on the manuscript. T. Wilcox provided valuable advice on phylogenetic methods. Mutant stocks were generously provided by M. Halpern, R. Knight, J. Lister, D. Raible, T. Schilling, S. Johnson, K. Poss and the Zebrafish International Resources Center at the University of Oregon. Supported by NIH R01 GM62182 to D.M.P.