Colour patterns are a prominent feature of many animals and are of high evolutionary relevance. In zebrafish, the adult pigment pattern comprises alternating stripes of two pigment cell types, melanophores and xanthophores. How the stripes are defined and a straight boundary is formed remains elusive. We find that mutants lacking one pigment cell type lack a striped pattern. Instead, cells of one type form characteristic patterns by homotypic interactions. Using mosaic analysis, we show that juxtaposition of melanophores and xanthophores suffices to restore stripe formation locally. Based on this, we have analysed the pigment pattern of two adult specific mutants: leopard and obelix. We demonstrate that obelix is required in melanophores to promote their aggregation and controls boundary integrity. By contrast, leopard regulates homotypic interaction within both melanophores and xanthophores, and interaction between the two, thus controlling boundary shape. These findings support a view in which cell-cell interactions among pigment cells are the major driving force for adult pigment pattern formation.
Many animal species employ colour patterns for interacting with their biotic environment. They are involved in diverse behaviours such as mate choice, camouflage and deterrence, and thus are probably subject to extensive selection. Additionally, colour patterns can vary dramatically even between closely related taxa, and hence seem to undergo rapid evolutionary change. Studying the evolutionary transformation of colour patterns should ultimately allow for understanding how diversity in adult morphology is generated. In recent years, some basic principles of pigment pattern formation have been elucidated in several organisms. For example, the eyespots on butterfly wings(reviewed by Beldade and Brakefield,2002) appear to be generated by readout of positional information provided by an organiser, the eyespot focus(Nijhout, 1980), that expresses morphogens such as Hedgehog(Keys et al., 1999). Cells in the wing epidermis presumably develop their type of pigmentation differently according to their distance from the focus(Brunetti et al., 2001). Another mechanism is exemplified in Drosophila wings where local deposition of a pigment-synthesizing enzyme and the supply of precursors generate a pigment pattern (True et al.,1999).
The pigment patterns of many fish and amphibians are comprised of alternating arrays of different neural crest-derived pigment cell types. Often, these cells are already differentiated as the pattern forms, making a mechanism as described above unlikely. Characterisation of these patterns mainly carried out in amphibian larvae has implicated a variety of cell behaviours such as migration, cell-substrate and cell-cell interactions in the generation of pigment patterns (e.g. Epperlein and Claviez, 1982; Epperlein and Löfberg,1990; Macmillan,1976; Parichy,1996a; Parichy,1996c; Tucker and Erickson,1986; Twitty,1945). However, the identification of specific factors involved in pigment pattern formation is difficult in non-genetic model systems.
The emergence of zebrafish as a model system has resulted in a collection of mutants affecting various aspects of pigment cell development and physiology. One major class of these mutants, such as colourless(Dutton et al., 2001), nacre (Lister et al.,1999), sparse(Parichy et al., 1999), rose (Parichy et al.,2000a) and fms(Parichy et al., 2000b), shows a pigment cell phenotype already during early larval stages. Molecular analysis of some of these mutants has revealed a high degree of functional conservation with genes previously implicated in melanocyte development in mice (Lister, 2002). Another class of mutants, such as asterix, obelix(Haffter et al., 1996b), leopard (Kirschbaum,1975), puma (Parichy and Turner, 2003b; Parichy et al., 2003) and hagoromo(Kawakami et al., 2000), shows a late phenotype during formation of the adult pigment pattern. So far, only one of these genes, hagoromo (hag), has been cloned. hag encodes a dactylin homologue(Kawakami et al., 2000) and perturbs the pattern locally upon mutation.
In zebrafish, the adult pigment pattern(Goodrich and Nichols, 1931)(reviewed by Quigley and Parichy,2002) consists of alternating stripes formed by melanin-bearing melanophores and pteridine containing xanthophores(Bagnara, 1998). The third major class of chromatophores in zebrafish, silvery iridiophores, is likely to be irrelevant for stripe formation, as tissues devoid of iridiophores also display a striped pattern. At the onset of adult development, this pattern evolves through the alignment of newly differentiating pigment cells with a lateral melanophore stripe that persists during larval stages(McClure, 1999; Milos et al., 1983). One major question is therefore how the cells become organised into these domains. In principle, this might be accomplished by filling in a prepattern that is set up independent of the pigment cells. Alternatively, mutual interactions between the pigment cells might define the striped domains. Other crucial characteristics of this pattern are the strict separation between regions occupied by either cell type and the straight boundary between them. How these features are generated and which cell behaviour underlies their formation remains largely obscure. In this study we show, using mutants that abolish formation of either melanophores or xanthophores, that presence and juxtaposition of both pigment cell types is necessary and sufficient for stripe formation. Thus, the domains appear to be largely defined by short-range interactions among pigment cells. Based on the analysis of patterns formed in the presence of only one cell type, we classify the cell behaviours during stripe formation into homo- and heterotypic interactions. In light of this distinction we analyse the phenotypes of two mutants, leopard (leo) and obelix (obe), that specifically alter the adult pigment pattern in a qualitative way and assay their cell type specific requirements by mosaic analyses. leo and obe affect different subsets of these cell behaviours and constitute central components of the stripe-forming system.
MATERIALS AND METHODS
Zebrafish maintenance and genetics
Zebrafish were bred and maintained under standard laboratory conditions(Brand et al., 2002). Double mutants between adult specific phenotypes, such as leo and obe, and mutants with an early larval phenotype such as nacand fms were generated by intercrossing homozygous adult carriers. Subsequently, F2 individuals homozygous for the adult phenotype were incrossed and larvae displaying the larval phenotype were raised. The following alleles were used in this study: nacrew2(Lister et al., 1999), leopardtw28, leopardtq270, obelixtd15 (Haffter et al., 1996b), goldenb1(Streisinger et al., 1986). Mutants in salztl41a and pfeffertm36b,previously assigned to two complementation groups,(Odenthal et al., 1996) were used in this study. Both, salztl41a and pfeffertm36b, fail to complement the pantherj4blue allele, carry mutations in the RTK fms (Parichy et al.,2000b) and will, therefore, be referred to as fms in the following. In salztl41a, codon 187 within the second IG domain is changed from TGC(Cys) to TGA. In pfeffertm36b, a conserved Phe (TTT) at position 579 is changed to a Ser (TCT). The transgenic line expressing EGFP under the control of a 5.3 kb fragment of the zebrafishβ-Actin promoter (bpeGFP) was a gift of K. Poss.
For all transplantations, donor embryos were homozygous for theβ-Actin GFP transgene (bpeGFP) in combination with various mutant genotypes. For leo and obe, the two strongest alleles, tq270 and td15, were used. Cell transplantations were performed essentially as described previously(Kane and Kishimoto, 2002). Surviving larvae were sorted on day 2 according to presence of pigment cell clones and raised to adulthood for analysis.
Image acquisition and analysis
For images of the developmental series, juvenile fish of defined genotype,age and size were maintained individually. Fish were anaesthetised in 0.04%Mesab and mounted on a glass slide in 3% methylcellulose. At 1 day intervals,body size was measured and pictures of the same region of the flank posterior to the dorsal fin were captured using a Zeiss AxioCam mounted on a dissecting microscope. Images of xanthophore autofluorescence and GFP fluorescence were taken with the same setup, using UV epiluminescence and a GFP filter. For images of the adults, male fish were sacrificed and fixed overnight in ice-cold 4% PFA. Scales and pectoral fins were removed and fish were mounted in 0.5% Agarose. All images were processed using Adobe Photoshop 6.0. Direct comparisons of cell positions on at 24-hour intervals were made by overlaying images manually to a maximum overlap of cell positions and marking cells that had changed position.
Formation of the striped pattern requires short-range interaction between melanophores and xanthophores
As the pigment pattern of adult zebrafish(Fig. 1A,D) consists of alternating stripes of two cell types, melanophores and xanthophores, one intriguing issue is whether either of these is capable of forming stripes in the absence of the other. This could indicate a prepattern provided by underlying tissues that is filled in independently by the pigment cells. Alternatively, the striped domains might not be predefined but rather generated by mutual interactions among pigment cells. Mutants that specifically abolish formation of one pigment cell type provide an ideal tool to address this point.
We first analysed the distribution of melanophores in the absence of xanthophores. Homozygous mutants for the kit-related RTK fms(Parichy et al., 2000b)display a strong reduction of xanthophores during larval stages and completely lack this cell-type in the adult (Fig. 1B,E). At the onset of adult pigment pattern formation in fms mutants (Fig. 1I),the initial pattern of melanophore differentiation is indistinguishable from wild type (Fig. 1G,H), with melanophores scattered throughout the flank. At around 25 days of age,melanophores start to aggregate in a similar fashion as in the presence of xanthophores. Later, melanophores fail to be cleared from the region around the horizontal myoseptum, which corresponds to the position of the first xanthophore stripe in wild type. In addition, melanophores do not clear from the more ventral regions, suggesting that the absence of xanthophores negatively influences their tendency to migrate towards the horizontal myoseptum region (Parichy et al.,2000b). In fms mutant adults(Fig. 1B,E), melanophores are found in circular clusters of about 10-20 cells. A substantial fraction of melanophores is interspersed individually between the clusters and no larger regions completely devoid of melanophores can be observed.
To study the pattern formed by xanthophores in the absence of melanophores,we used mutants for nacre (nac; mitfa —Zebrafish Information Network) (Lister et al., 1999), a zebrafish homologue of the mammalian microphthalmia gene (Hemesath et al., 1994; Hodgkinson et al., 1993). nac mutants are homozygous viable and completely lack melanophores in larval as well as adult stages. During larva to adult transition, xanthophores in nac mutants (Fig. 1J) first appear adjacent to the horizontal myoseptum in essentially the same position as in wild type. However, their appearance is strongly delayed, and a fraction (4/9) of the individuals analysed showed no xanthophores by day 30. Xanthophores appear in small clusters of only a few cells and seem to expand their domain much slower than in wild type. In nac adults (Fig. 1C,F), xanthophores occupy a coherent longitudinal field of cells straddling the horizontal myoseptum and irregular patches further ventrally. Between these domains, large areas are completely devoid of xanthophores.
We next asked whether the altered positioning of pigment cells in nac and fms mutants is solely due to the absence of one cell type and thus reflects their intrinsic positioning behaviour. To this end, we generated genetic mosaics by transplanting wild-type blastula cells marked by constitutive GFP expression (bpeGFP) into homozygous nac and fms mutants. Embryos resulting from these experiments were raised to adulthood and patches (`clones') containing the pigment cell type normally absent from the respective mutant were analysed.
Upon transplantation of bpeGFP blastula cells into homozygous nacmutants, all melanophores observed in transplanted fish are of donor origin,as nac has been shown to be autonomously required in melanophores(Lister et al., 1999), whereas xanthophores are both donor as well as host derived. In all animals analysed(n=26) the melanophores within such clones formed stripes that alternate with xanthophore stripes, resulting in a locally rescued pattern that is indistinguishable from wild type(Fig. 2C). The xanthophore stripes within such clones contain donor-derived, GFP-positive xanthophores as well as host derived ones. This shows that nac mutant xanthophores retain the ability to organise into stripes. Only xanthophores that are in contact or close proximity to melanophores will contribute to the striped pattern, whereas donor as well as host derived xanthophores further away from the melanophore clone assume a distribution as seen in untransplanted nac mutants (Fig. 2D). This indicates that indeed the absence of melanophores causes the altered xanthophore distribution in nac mutants. Moreover, the fact that xanthophores will form stripes only in the vicinity of melanophores suggests a short-range mode of interaction between melanophores and xanthophores during stripe formation. The cells transplanted at blastula stages will give rise to a variety of tissues in the adult. However, we found that striped patterns always occur where xanthophores and melanophores are in proximity,irrespective of the origin of the underlying tissue.
In the complementary experiment, we introduced bpeGFP cells into homozygous fms mutant embryos. To allow for distinction of donor derived melanophores from host-derived ones, the donor embryos were additionally homozygous for the golden (gol) mutation(Streisinger et al., 1986),which results in hypopigmented melanophores. Adults carrying clones(n=24) displayed a normally striped pattern within the clone. Most xanthophores were GFP positive, i.e. donor derived, whereas most of the melanophores were normally pigmented (gol+), i.e. host derived(Fig. 2A,B). This indicates that melanophores lacking fms function still can form stripes and thus the melanophore pattern in fms mutants is exclusively caused by the lack of xanthophores. In about one-fifth of the mosaic fish a completely rescued pattern in the entire flank was obtained, a situation we never observed with nac mutant mosaics. This indicates a strong proliferative capability of xanthophores. Notably, in such clones also GFP-negative host-derived xanthophores were present (data not shown), which suggests a partial non-autonomous function of fms. In cases where the striped clone bordered a region devoid of xanthophores, the xanthophores within the stripes were forming one coherent domain, and melanophores outside the clones assumed the distribution normally seen in fms mutant adults.
Taken together, these findings show that juxtaposition of melanophores and xanthophores is necessary and sufficient for stripe formation.
obeandleoaffect stripe integrity and shape respectively
The experiments outlined above indicate that stripe formation requires interactions between the two pigment cell types. To look for factors that might govern these interactions we investigated two mutants, leopard(leo) (Kirschbaum,1975) and obelix (obe)(Haffter et al., 1996b), that alter the adult pigment pattern. Four dominant alleles of obe were isolated in a large-scale mutagenesis screen(Haffter et al., 1996a) (F.M. and C.N.-V., unpublished). Heterozygous adults display only two to three melanophore stripes that are wider than in wild type and frequently interrupted by xanthophores (Fig. 3B,L). This is most likely a dosage-dependent effect of obe, as it is also observed in heterozygotes for the c7 deletion that covers the obe locus (data not shown). In obe homozygous adults (Fig. 3A,E) melanophores are found in two broad longitudinal domains flanking the horizontal myoseptum. Strikingly, the melanophore domains always contain interspersed xanthophores,indicating that obe mutants lack the strict separation between the two cell types seen in wild type.
The four alleles of leo(Haffter et al., 1996b) can be ordered into a series of increasing phenotypic strength. Heterozygotes for weak alleles, such as tw28, show a rather normal striped pattern, but the boundaries between melanophores and xanthophores are undulating. In leotw28 homozygotes(Fig. 3M) the stripes are always undulating and sometimes break up into spots. Mutants for the strongest allele leotq270 show undulating stripes and spots already in the heterozygous condition and in homozygous individuals(Fig. 3C,H) the stripes are entirely transformed into a spotted pattern with melanophore spots of about 10-20 cells surrounded by xanthophores. Unlike in obe, the two cell types still occupy spatially separated domains. The spotting phenotype and the undulating stripes in weaker alleles can be interpreted as decreased leo function that results in an increased boundary length between the melanophore and xanthophore populations. Double homozygotes for leoand obe (Fig. 1F,D)display single melanophores surrounded by xanthophores, suggesting that leo and obe act in parallel to control length and integrity of the boundary between melanophores and xanthophores.
We next examined the ontogeny of the pigment pattern in individual leo and obe mutant fish over the course of 14 days during larva to adult transition. At the onset of adult pigment pattern formation in wild type (Fig. 1G,H),melanophores are differentiating in broad regions of the flank dorsally and ventrally of the horizontal myoseptum, which bears melanophores of the larval lateral stripe. Subsequently, xanthophores appear around the horizontal myoseptum. Concomitantly, the melanophores in the flank begin to converge towards the horizontal myoseptum by migration(Parichy et al., 2000b),whereas melanophores within the xanthophore stripe are eliminated. Thus, a xanthophore stripe flanked by two melanophore stripes has formed. As melanophores are cleared from dorsal and ventral regions, new xanthophores appear, continuing the alternation of the pattern.
In obe mutants, the larval pattern (not shown) and the initial appearance of melanophores at 21 days are similar to wild type(Fig. 4A). During appearance of the xanthophores, melanophores fail to cluster and remain scattered instead. Furthermore, melanophores do not converge towards the horizontal myoseptum and persist in more ventral areas as well as within the xanthophore stripe. The xanthophores in obe mutants (Fig. 4B) appear initially in the same position as in wild type, but later start to emerge between the scattered melanophores, giving rise to the mixing pattern.
In addition, leo mutants display a normal larval pattern (not shown) and a normal initial distribution of early adult melanophores(Fig. 4C). Similar to obe, melanophores fail to undergo clustering and convergence towards the horizontal myoseptum and remain scattered in the flank as well as within the xanthophore stripe. Initially, leo mutant xanthophores appear in a stripe straddling the horizontal myoseptum(Fig. 4D), as in wild type, and later start to populate the space left by the scattered melanophores. These observations suggested that one prime cause of the altered pigment patterns in leo and obe might be the failure of melanophores to aggregate into stripe-like domains. To investigate if this defect is caused by the inability of mutant melanophores to change position, we directly compared images captured at 24 hour intervals during emergence of the adult pattern from the same region of the trunk (Fig. 5). This comparison revealed that both obe and leo mutant melanophores are still able to translocate, albeit in the case of obe, to a lesser extent than in wild type. Therefore, the obe and leo mutant patterns are not caused by immobility of melanophores.
obeandleohave similar effects on melanophore behaviour but affect xanthophores differently
As melanophores in fms and xanthophores in nac mutants retain their ability to form stripes, their distribution most probably reflects interactions also employed normally during stripe formation. Therefore, we used these mutants to study the effects of leo and obe on melanophore and xanthophore patterns. In double homozygotes for fms and obe (Fig. 3F), melanophore clusters are lost and all melanophores seem more or less evenly scattered instead. Strikingly, fms;leo double homozygotes (Fig. 3I) also show the same effect of loss of melanophore clustering and are basically indistinguishable from fms;obe double mutants. We next characterised the effect of leo and obe on the behaviour of xanthophores in nac mutants. The distribution of xanthophores in nac;obe double mutants(Fig. 3G) is essentially the same as in nac single mutants, with xanthophore free regions interspersed. Conversely, in nac;leo double mutants(Fig. 3J), xanthophores tend to fill up the entire flank and leave no patches devoid of xanthophores, an effect that is already discernible upon visual inspection of double mutant adults. The spacing and density of the cells however appear similar as in wild type or nac single mutants.
Taken together, these data indicate that both leo and obehave a similar effect on melanophore behaviour, whereas only leo also affects xanthophore behaviour.
obefunction is required by melanophores
To assay the cell-type-specific requirements for leo and obe, we juxtaposed wild-type melanophores with mutant xanthophores and vice versa. This was accomplished by transplanting mutant or wild-type cells into nac single or double mutants, because in nacmosaics all melanophores will be donor derived. These mosaics could then be scored for formation of a wild-type or mutant pattern. As only one of the cell types will be mutant in a given mosaic condition, the type of pattern generated in that condition indicates the cell types in which the formation of the respective gene is required. Additionally, this experiment should also allow to determine whether the genotype of the pigment cells themselves or rather that of the underlying tissue is decisive for the type of pattern formed. In order to confirm that obe or leo mutant patterns could be reconstituted in mosaic experiments, we transplanted obe or leo mutant cells into nac;obe or nac;leo double mutant hosts, respectively, and found that in both cases the mutant patterns were faithfully reproduced (data not shown). Furthermore, all mosaic conditions for leo and obe generated consistent patterns within the clones, irrespective of whether the underlying tissue was wild type or mutant for obe, thus establishing a requirement for both genes within the pigment cells.
We analysed the cell type specific requirements for obe by transplanting bpeGFP cells that are wild type for obe into nac;obe double mutants(Fig. 6A,B). To distinguish host and donor cells, the donor cells were marked by the gol mutation and GFP expression. In the mosaic adults analysed (n=24) melanophore clones always formed a wild-type-like stripe pattern with clear separation between melanophores and xanthophores (Fig. 6A). This was also the case when the vast majority of xanthophores within a clone was host derived and thus mutant for obe(Fig. 6B), showing that wild-type melanophores can form a regular boundary with obe mutant xanthophores. In the complementary experiment, obe mutant bpeGFP cells were transplanted into nac single mutant hosts, melanophore clones in all mosaics analysed (n=18) formed broad irregular domains,similar to the obe mutant pattern(Fig. 6D,E). In cases where most of the xanthophores were wild type, they mingled with obe- melanophores. Thus, both mosaic conditions indicate that the genotype of the melanophores determines whether a wild type or an obe mutant pattern is formed. Hence, obe function is required within the melanophores for spatial separation between the pigment cell types.
leois required within both melanophores and xanthophores
We performed the analogous experiments for the leo mutant. First,bpeGFP cells that are leo+ were transplanted into a nac;leo host (Fig. 6G,H). None of the adults bearing wild-type melanophore clones(n=22) showed an entirely wild-type melanophore pattern within the clone. Inspection of melanophore clones mainly surrounded by GFP-negative (and hence leo mutant xanthophores) showed that leo mutant xanthophores form roundish, irregular boundaries with wild-type melanophores. This indicates that wild-type melanophores are unable to impose a wild-type striped pattern with straight boundaries onto leo mutant xanthophores. Notably, even though the strongest leo allele leotq270 was used in all mosaic experiments, the clones never consisted entirely of spots, but rather showed an intermediary phenotype, with undulating stripes and spots.
With the complementary mosaic condition, transplantation of leo;bpeGFP cells into a nac single mutant host, a similar result was obtained (Fig. 6J,K). Again,none of the nac adults containing melanophore clones (n=18)showed a clear wild-type pattern in the clone. Instead, similar undulating stripes and spots were observed as for the inverse mosaic condition. In addition, leo mutant melanophores predominantly surrounded by wild-type xanthophores were forming spots, and hence a leo-like pattern. Thus, leo mutant melanophores can organise wild-type xanthophores into a leo-like spotted pattern. These findings indicate that leo, unlike obe, is required by both pigment cell types, and that compromising leo function in either melanophores or xanthophores alone is sufficient to generate a leo like pattern.
Formation of the striped pattern requires interaction between pigment cell types
Our phenotypic and mosaic analysis of the mutants lacking one pigment cell type shows that juxtaposition of both cell types is necessary and sufficient for formation of a striped pattern. The mosaic analysis corroborates recent findings (Parichy and Turner,2003a), which were obtained under slightly different experimental conditions. Thus, mutual interactions between the pigment cells play a predominant role during establishment of the striped domains, even though few features of the pattern can be formed independently (see below). A strict prepattern of stripes that would be formed independently of the pigment cells and populated differentially by them is therefore unlikely. Such prepatterns have been shown to dominate the generation of larval pigment patterns in amphibians (reviewed by Epperlein et al.,1996; Parichy,1996b). In these systems, the underlying tissue plays an instructive role in defining the territories occupied by the different pigment cell types, whereas cell-cell interactions are likely to play a minor role. This apparent difference might indicate different mechanisms operating in the formation of larval versus adult patterns. Indeed, also in zebrafish the position of the larval melanophore pattern coincides with anatomical structures such as the horizontal myoseptum(Milos et al., 1983). Moreover, different mechanisms might be operating in different taxa.
In addition, the analysis of nac and fms mutants also reveals which aspects of cell behaviour in particular depend on the presence of the other cell types. The delayed appearance of xanthophores in nac mutants could indicate an requirement of melanophores during adult development for correct differentiation and/or proliferation of xanthophores. Alternatively, this could be due to a direct requirement for nac in xanthophore development, as in the embryo some xanthophore precursors have been shown also to express mitf(Parichy et al., 2000b). Melanophores can aggregate independently of xanthophores, but they fail to converge towards the horizontal myoseptum and also persist in areas from which they are normally cleared (Parichy et al.,2000b). This indicates an attractive effect of xanthophores on distant melanophores, which promotes their alignment on either side of the xanthophore stripe. Conversely, within their domain, xanthophores might repel melanophores.
The exact cellular mechanisms that underlie these interactions are still unclear. It can be assumed that it involves an ability of the cells to recognise each other and to aggregate. As pigment cells outside clones cannot be organised into stripes at a distance, stripe formation appears to involve mainly short-range interaction or even direct cell-cell contact. Indeed,histological examination of adult dermis revealed intimate cell contact both among and between cells of one type (F.M. and C.N.-V., unpublished). One intriguing possibility is that differential affinities(McNeill, 2000; Steinberg, 1970) between pigment cells account for the spatially distinct domains of the pattern. Clearly, cells in Drosophila wing imaginal disks display differential affinities (Dahmann and Basler,1999; Dahmann and Basler,2000; Milan et al.,2001) during establishment of compartment boundaries. Given the intimate contact and the aggregative behaviour of pigment cells in the adult,one could envisage that affinities of melanophores and xanthophores for cells of the same type could establish separation, whereas affinity for cells of the other type might generate contact at the boundary.
Early features of the adult pattern can be formed in the absence of one cell type
Apart from interacting with cells of the other cell type (heterotypic interaction), pigment cells of one type also interact with each other(homotypic interaction). This is evident from the fact that neither xanthophores nor melanophores are randomly distributed in the absence of the other cell type. Rather, cell-type-specific patterns are formed that can be seen as the outcome of homotypic interactions reflecting the intrinsic positioning behaviour of the cells. One key feature of these single cell type patterns is the cells tendency to aggregate. The clustering of a fraction of melanophores involves translocation of cells from an initially scattered distribution towards the clusters formed during the early adult period(Parichy et al., 2000b). Even though xanthophores appear to change their position very little, if at all,they also remain in coherent areas during expansion of the xanthophore domain. Thus, both cell types must be able to establish and maintain cell-cell contact either directly or indirectly.
Although formation of the hallmarks of the adult pigment pattern depends on presence and interaction of melanophores and xanthophores, early features of the wild-type pattern can be generated independently. The xanthophores in nac mutants appear first close to the horizontal myoseptum, which corresponds to the position of the first xanthophore stripe in wild type. In addition, the initial position in which melanophores appear in fmsmutants is similar to wild type. This indicates that at least the initial positioning of pigment cells during adult pattern formation is largely independent of the presence of the other cell type. Instead, initial positioning might be defined by anatomical landmarks, such as the horizontal myoseptum. Notably, mutants in fss/Tbx24(Nikaido et al., 2002; van Eeden et al., 1996)disrupt horizontal myoseptum integrity, and display locally interrupted stripes as adults (F.M. and C.N.-V., unpublished). The function of the horizontal myoseptum as a starting point for stripe formation may be indirect,as it also governs migration of the lateral line primordium, which has been implicated in pigment pattern formation in salamanders(Parichy, 1996a; Parichy, 1996c).
obecontrols boundary integrity by regulating melanophore aggregation
Among the few mutants known to affect stripe formation in adult zebrafish, leo and obe are probably central regulators of this process,as they perturb stripe formation in all parts of the body. Furthermore, they disrupt qualitative aspects of the pattern: the spatial separation of melanophores and xanthophores (obe), and boundary shape(leo). We believe that theses effects are not due to altered pigment cell numbers, because mutants in genes known to affect the numbers of pigment cells, such as sparse or heterozygosity for some semi-dominant alleles of fms (Haffter et al.,1996b; Odenthal et al.,1996) do not affect boundary integrity or shape in such a dramatic way as leo or obe.
For obe, our analysis indicates that this gene is required in the melanophores to control their homotypic clustering. As xanthophores mingle with melanophores in obe mutants, melanophore clustering appears to be a major prerequisite for establishing spatial separation between melanophore and xanthophore stripes. The fact that in obexanthophores also appear within the melanophore domains indicates an intrinsic affinity of xanthophores for the proximity of melanophores. This can overcome the tendency of xanthophores to remain clustered and to leave substantial regions of the flanks unoccupied, which occurs in the absence of melanophores. Maintaining melanophore clustering may not only be a way to establish separation, but might also regulate stripe width, as obeheterozygotes display wider melanophore stripes. The reduced melanophore clustering in this condition might result in a larger area occupied by more loosely clustered melanophores, whose clustering is just strong enough to prevent xanthophores from invading the melanophore domain.
At present, we can only speculate about which cell-biological aspect of clustering behaviour might exactly be controlled by obe. It is apparent from the ontogeny of wild-type pigment patterns(McClure, 1999; Parichy et al., 2000b) that melanophore clustering involves extensive translocation of melanophores and thus requires an ability of these cells to migrate, to recognise each other and to maintain contact. Notably, obe mutant melanophores neither cluster nor converge individually towards the horizontal myoseptum, nor do they disappear from within the xanthophore domain, suggesting that all of these processes are dependent on obe. This defect might either be caused by a reduced ability to recognise an attractive signal or by impaired motility. One obvious candidate for an attractive signal guiding melanophore aggregation is the kit ligand Steel(Copeland et al., 1990; Huang et al., 1990; Zsebo et al., 1990), which stimulates melanocyte migration in the mouse(Jordan and Jackson, 2000; Kunisada et al., 1998). Although a Steel homologue has not yet been identified in zebrafish, sparse adults that are mutant for the Steel-receptor still display melanophores aggregated in stripes(Parichy et al., 1999) (F.M. and C.N.-V., unpublished), thus making a role for Steel-kit as essential regulators of melanophore aggregation unlikely. Molecular analysis of the obe gene will certainly shed light on its exact function.
leoaffects multiple aspects of cell behaviour
In contrast to the relatively restricted role of obe, leo affects homotypic interactions of both melanophores and xanthophores, as revealed by double mutant and mosaic analysis. The expansion of the xanthophore domain in leo; nac double mutants is unlikely to be caused by hyperproliferation of xanthophores, because in this condition the differentiation of xanthophores is also strongly delayed (data not shown). Instead, loss of leo function might weaken the cohesion between xanthophores and allow for insertion of newly differentiating xanthophores,thus expanding the domain. Interestingly, the disruptive effect of leo on melanophore clustering is highly similar to obe. Nevertheless, melanophores and xanthophores are still spatially separated in adults, raising the question of why melanophores and xanthophores do not mingle in leo in a similar fashion as in obe. This could be explained by the concomitant alteration of homotypic xanthophore behaviour,which might counteract the loss of melanophore clustering and re-establish separation, albeit in a spotted pattern. However, when confronted with leo mutant melanophores in mosaic experiments, xanthophores that are wild type for leo do not mingle. Therefore, the effect of leo on melanophore clustering must be quantitatively or qualitatively different from that of obe.
Additionally, leo function is required in both melanophores and xanthophores for formation of a straight boundary, as loss of leofunction in one cell type can impose a leo pattern on the other. This notion is corroborated by the fact that if only one cell type is mutant for leo, the pattern formed by mosaics displays a weaker leophenotype (undulating stripes and spots) than the leo mutant. These findings strongly suggest that leo also affects heterotypic interactions between both cell types. In light of this idea, it is important to note that the straight interface between melanophore and xanthophore stripes can be interpreted as minimised contact and that reducing leofunction gradually increases contact between melanophores and xanthophores by rounding the boundary between them. The leo gene is thus a central component of the stripe-forming system in that it controls both homotypic and heterotypic interactions of melanophores and xanthophores. This pleiotropy might be achieved by leo regulating several downstream components that control subsets of cell interactions. Alternatively, leo could favour homotypic interaction among melanophores and xanthophores at the expense of heterotypic ones between the two. Loss of leo function could then result in an increased tendency of either cell type to contact the other and hence to increase the boundary length between the two.
A model of cell interactions during stripe formation
Based on our findings, a model of interactions the adult pigment pattern in zebrafish can be formulated (Fig. 7). The process is initiated by the differentiation of melanophores at about 3 weeks of development(Fig. 7A) and influenced by size and age of the fish. Depending on the presence of melanophores,xanthophores start differentiating in a relatively restricted domain straddling the horizontal myoseptum (Fig. 7B). Positioning of this initial xanthophore stripe is presumably directed by anatomical landmarks such as the horizontal myoseptum. Concomitantly, melanophores start aggregating in an obe- and leo-dependent fashion, and align with the xanthophore stripe. Melanophore aggregation prevents further differentiation of xanthophores within the melanophore domain, thus establishing spatial separation and also regulates melanophore stripe width. Additionally(Fig. 7C), xanthophores exert a repulsive effect on melanophores present within their domain. The straight interface between melanophores and xanthophores is subsequently formed by leo-dependent regulation of both homo- and heterotypic cell interactions (Fig. 7D).
Clearly, obe and leo will not be the only factors in this complex process, but the detection of adult pattern mutants is currently limited by the types of screens conducted so far. Novel screening approaches should therefore result in a more complete collection. Even though the exact cellular mechanisms guiding formation of the adult pigment pattern remain elusive, it will be interesting to see to what extent the findings presented here are applicable to other species (e.g. Parichy and Johnson, 2001),especially as stripes, either longitudinal or vertical, are a central feature of the pigment patterns of many animals.
We thank Joëlle Deplazes and Anja Wehner for technical assistance; D. Raible, D. Parichy, K. Poss, Chi-Bin Chien and M. Halpern for fish stocks and reagents; and C. Wolff, H.-G. Frohnhöfer, I. Sütterlin, D. Gilmour,J. Srinivasan and H. Knaut for comments on the manuscript.