Involvement of Delta/Notch signaling in zebrafish adult pigment stripe patterning

How sequential and periodic structures form autonomously in organisms is a fundamental issue in developmental biology. One example of periodic structures is animal pigment patterns, which have long been studied both theoretically (Kondo and Miura, 2010; Meinhardt, 1982; Murray, 2002) and experimentally (Johnson et al., 1995; Parichy, 2007). The reaction-diffusion mechanism presented by A. M. Turing more than 60 years ago is the most common theoretical model for explaining various examples of biological pattern formation, including pigment patterning (Kondo and Miura, 2010; Meinhardt, 1982; Murray, 2002; Turing, 1952). This theory suggests that such patterns result from a stationary wave (Turing pattern or reaction-diffusion pattern) made by a combination of putative reactive and diffusible substances. Although the theory was much doubted initially, simulation studies have since demonstrated that this mechanism can lead to various pigment patterns commonly seen in nature (Kondo and Miura, 2010; Meinhardt, 1982; Murray, 2002) and also can predict dynamic pattern changes that occur during animal growth (Kondo and Asai, 1995; Yamaguchi et al., 2007). Molecular evidence supporting the hypothesis, however, has been sparse.

For testing the Turing model empirically, the zebrafish (Danio rerio) is especially useful. The stripe pattern of zebrafish is composed of three types of pigment cells: melanophores, xanthophores and iridophores (Johnson et al., 1995; Odenthal et al., 1996). Experiments using mutant zebrafish with altered pigment patterns have suggested that pigment patterns result from interactions between these three types of pigment cells (Maderspacher and Nüsslein-Volhard, 2003; Nakamasu et al., 2009; Parichy, 2003; Patterson and Parichy, 2013). Interactions specifically between melanophores and xanthophores were first demonstrated by genetic mosaic analyses using a xanthophore-deficient mutant for colony stimulating factor 1 receptor (csf1ra, formerly fms); analyses of a temperature-sensitive allele, csf1ra^utr4e174A, further demonstrated that ablation of xanthophores resulted in the death of melanophores in the adult fish (Parichy and Turner, 2003). Subsequent experiments using laser ablation showed that at least two kinds of interactions between melanophores and xanthophores contribute to pattern formation (Nakamasu et al., 2009) (supplementary material Fig. S1C). In the first case, transient losses of a large number of xanthophores lead to the death of melanophores (supplementary material Fig. S1A), confirming the findings of Parichy and Turner. The functional distance of this interaction should be long because melanophores even in the middle of a black stripe, far from the nearest xanthophores, still died after xanthophore ablation. In the second case, when single melanophores or xanthophores are surrounded by cells of the other type, the isolated cell dies (supplementary material Fig. S1B). In this scenario, the two classes of pigment cells, melanophores and xanthophores, compete with each other for survival over short-range distances.

The network of short-range and long-range interactions deduced from these in vivo experiments is consistent with Turing’s mathematical model (Nakamasu et al., 2009). Short-range repulsive interactions, in which xanthophore dendrites contact melanophores directly, have now been observed in vitro (Inaba et al., 2012). There is, however, little information concerning the long-range interaction. According to Turing’s theory, long-range interactions are more important because they determine the width of stripes. Identifying the molecular bases of these interactions should, therefore, elucidate actual mechanisms underlying Turing pattern formation in a biological system. In the present article, we aimed to identify the molecular and cellular mechanism of long-range interaction.

Previous in vivo experiments suggested that a survival signal is transmitted by xanthophores to melanophores (supplementary material Fig. S1A) (Nakamasu et al., 2009; Parichy and Turner, 2003). To identify candidate mediators of this interaction, we purified xanthophores and melanophores from fins and analyzed gene expression in these cells by microarray. Of the 15,500 genes surveyed, 482 were expressed at 2.5-fold higher levels (at least) in xanthophores than in melanophores. Among these, we selected seven genes encoding ligands or transmembrane proteins for further analysis. Using a reverse transcription-PCR (RT-PCR), we confirmed that six of these seven genes were expressed exclusively in xanthophores.

To screen these candidates functionally, we reasoned that an excess amount of any signal protein should alter the resulting patterns. Therefore, we expressed each of the six candidate genes ectopically in the melanophore lineage using the mitfa promoter (Dorsky et al., 2000; Inaba et al., 2012). In initial screening, we injected embryos with each transgene, and reared at least ten injected (mosaic) fish to adulthood (∼3 months) for each candidate gene. For deltaC, two of 15 fish exhibited pattern alterations (supplementary material Fig. S2). For other candidate genes, no pattern differences were observed.

We performed in situ hybridization on late larval zebrafish and found that the deltaC is expressed in presumptive xanthophores, but not in melanophores (Fig. 1A). Confirming this idea, deltaC expression was not observed in the xanthophore-deficient, csf1ra^j4e1 mutant (Parichy et al., 2000) (Fig. 1B). Next, by RT-PCR, we investigated whether genes encoding other Notch family members and their ligands are also expressed in pigment cells. Among the Notch ligands, we found that deltaC and delta-like 4 were expressed robustly in xanthophores but not in melanophores. deltaD was expressed more weakly in xanthophores than deltaC or delta-like 4 so we did not analyze it further. deltaA and deltaB were not expressed in either cell type (Fig. 1C). Among the Notch family genes, notch1a was expressed abundantly in melanophores. We also detected relatively low levels of notch2 expression in both melanophores and xanthophores (Fig. 1D). These results suggested the possibility that a Delta/Notch signal (Kopan, 2010) is transmitted from xanthophores (deltaC, delta-like 4 or both) to melanophores (notch1a, notch2 or both) to promote melanophore survival in zebrafish skin.

To determine whether melanophore survival depends on the Delta/Notch signal, we incubated 1- to 1.5-month-old adult, wild-type fish with DAPT, which blocks cleavage of the Notch intracellular domain (NICD) fragment of the Notch receptor (Dovey et al., 2001; Geling et al., 2002). As shown in Fig. 2A,F, treatment with DAPT for 15 days decreased the number of melanophores to 60% of that before the treatment (xanthophore numbers were unaffected; supplementary material Fig. S3). In DAPT-treated fish, we observed melanin-containing debris, a hallmark of melanophore death (Lang et al., 2009; Parichy et al., 1999; Parichy and Turner, 2003) in the region from which melanophores had disappeared (Fig. 2D,E). Death of melanophores was not restricted to a specific position of the stripe, but seemed to occur randomly, with subsequent rearrangements of the remaining melanophores generating narrower black stripes (Fig. 2A). Upon removing DAPT, however, melanophore numbers recovered quickly (Fig. 2F; supplementary material Fig. S4), suggesting that melanophore precursors were unperturbed. These observations of melanophore death induced by DAPT are very similar to those following xanthophore-ablation experiments (Nakamasu et al., 2009; Parichy and Turner, 2003).

To confirm that DAPT-induced death of melanophores reflected an inhibition of Notch signaling, we generated stable transgenic lines expressing in the melanophore lineage either DeltaC [Tg(mitfa:deltaC); derived from the fish shown in supplementary material Fig. S1] or the Notch1a intracellular domain [NICD1a; Tg(mitfa:NICD1a)]. We then tested for continued susceptibility to DAPT-induced melanophore death. We reasoned that ectopic expression of DeltaC should allow a Notch signal to be exchanged among melanophores, whereas expression of NICD1a should result in constitutive, autonomous activation of the pathway (Del Monte et al., 2007; Schroeter et al., 1998). In either case, we expected alterations in melanophore patterning. But, we also anticipated that if DAPT specifically blocks Notch signaling, then melanophores of Tg(mitfa:deltaC) fish would remain susceptible to DAPT-induced death despite Notch pathway activation, because such signals would still require NICD cleavage. By contrast, melanophores of Tg(mitfa:NICD1a) fish would be resistant to DAPT effects because NICD1a would be present in melanophores constitutively. As shown in Fig. 3, the stripes of melanophores in both transgenic lines were wider than those of the wild type, further supporting a role for Notch signaling in pattern formation. Moreover, we found that melanophores of Tg(mitfa:NICD1a) persisted even in the presence of DAPT, whereas melanophores of Tg(mitfa:deltaC) fish were lost during DAPT treatment (Fig. 2B,C,F), consistent with our predictions. These results confirm that melanophore death occurred specifically in response to Notch signaling inhibition.

Because xanthophores expressed Notch ligands and elimination of xanthophores caused the death of melanophores, xanthophores are the likely source of the Notch signal. To test this idea, we compared the effects of transient xanthophore loss between wild-type and Tg(mitfa:NICD1a) fish. Following laser ablation of xanthophores (Nakamasu et al., 2009), we found that 28% of melanophores died in wild-type fish (Fig. 4A), whereas no melanophores died in fish expressing NICD1a (Fig. 4B,C). Thus, Notch pathway activation autonomous to melanophores can substitute for xanthophores, consistent with the hypothesis that xanthophores promote melanophore survival through Delta/Notch signaling.

Delta and Notch belong to families of membrane-bound proteins. Therefore, if a Delta/Notch survival signal is provided by xanthophores to melanophores, these cells must contact one another directly. As even melanophores distant from xanthophores died in response to xanthophore ablation, we anticipated that melanophores might contact xanthophores by cellular processes not normally apparent (Fig. 5A). To visualize even thin cellular projections, we used yohimbine (an alpha 2-adrenoceptor antagonist), which induces the peripheral dispersion of melanin-containing melanosomes. As shown in Fig. 5B, yohimbine caused the movement of melanosomes into long projections that extend towards xanthophores. These projections were visible at stripe edges but could not be discerned clearly within the interior of black stripes, because the region is totally filled with black pigment. To visualize the projection from single cells, we used a melanin pigment-lacking mutant, brass^b2, and injected them with a transgene in which mitfa drives expression of membrane-targeted enhanced green fluorescent protein, EGFP-CAAX (supplementary material Fig. S5D) (Gomes et al., 2003; Wright and Philips, 2006). In F0 fish expressing this transgene mosaically, processes of single melanophores even within the stripes are visible. For example, Fig. 5D shows three melanophores in a black stripe each of which extend long projections towards xanthophores. In Fig. 5E, a cluster of melanophores express EGFP-CAAX and most projections are oriented dorsoventrally, towards xanthophores above and below the stripe. Most of the melanophores extend projections and the direction is vertical to the stripe. Fig. 5F shows the cell projections of the melanophores at the region where the stripe terminates (illustrated in Fig. 5C). Even in wild-type fish, such a stripe gap occurs in ∼5% of fish). In this case, the projection appeared to extend radially towards the xanthophore region. In csf1ra mutant fish, which lack xanthophores, long projections were rare (supplementary material Fig. S5A). We also checked the projection of xanthophores in wild-type zebrafish. Although there were some xanthophores that extended long projections towards melanophores, they were rare. Many xanthophores did not have long projections (supplementary material Fig. S5B,C). These projections are often greater than three times the length of the cell body and nearly half of the width of stripes. We further confirmed by electron microscopy that the melanophore projections directly touched xanthophores passing through the layer of iridophores (supplementary material Fig. S6).

In this article, we showed that xanthophores express Delta ligands (deltaC and delta-like 4), a family of membrane-bound ligands, and that melanophores express Notch receptors (notch1a and notch2). Next, we showed that Delta/Notch signaling promotes melanophore survival, and that this signal originates with xanthophores. Finally, we found that melanophores in the black stripes extend long projections towards xanthophores that may enable pigment cells to transfer these long-range signals, despite association of Notch and Delta with the cell membrane. Although we have not confirmed that the Delta/Notch signal is transduced specifically at the sites of such projections, we propose this as a working model for future exploration.

There are at present few data pertaining to cellular mechanisms that control melanophore survival. Although our observations suggest the involvement of Delta/Notch signaling, this could be just part of a more complex mechanism; indeed, melanoblast survival depends on signaling through the kit receptor a, endothelin receptor B1a and the immunoglobulin superfamily member igsf11 (Budi et al., 2011; Cooper et al., 2009; Eom et al., 2012). It is possible that Delta/Notch synergizes with such pathways to promote melanophore survival.

The maximum length of melanophore projections is about half the width of the adult stripe. This suggests an interesting hypothesis to explain how the width of a stripe could be formed and stably maintained (supplementary material Fig. S7). When the growth of a fish makes the stripes wider than twice the projection length, projections of melanophores in the middle of a stripe would no longer reach xanthophores and the melanophores would die. This effect could divide the transiently wider stripe, returning it to its original width. Indeed, bifurcation of stripes has been observed in a tropical fish (Pomacanthus imperator) (Kondo and Asai, 1995), in zebrafish expressing overexpressing a mutant form of the gap junction protein connexin 41.8 (Watanabe and Kondo, 2012) and in very old and big wild-type zebrafish (Asai et al., 1999). This hypothesis also explains the altered pigment patterns of deltaC and NICD1a transgenic fish. In these fish, melanophores should be less sensitive to losing a survival signal from xanthophores, thereby allowing for persistence at sites more distant from xanthophores in the yellow stripes, and resulting in the wider melanophore stripes we observed. Although our hypothesis is simple and useful to explain patterning, it remains to be confirmed by further experiments. Particularly important will be to test if Delta/Notch signaling occurs specifically at the tips of melanophore projections and the extent to which such interactions occur not only in the adult, but also during larval stages when the pattern is first developing.

It has been proposed that a reaction-diffusion mechanism underlies pigment pattern formation because such a mechanism can predict pigment pattern changes in growing fish (Kondo and Asai, 1995; Yamaguchi et al., 2007). However, experimental investigations have indicated that the true mechanisms are considerably different. We showed previously that short-range signals are transmitted to melanophores not by diffusion of a diffusible molecule, but rather by relatively short processes of xanthophores contacting melanophores directly (Inaba et al., 2012). Here, we presented the possibility that a long-range signal is likewise transmitted by direct contact between melanophores and distant xanthophores. There is, so far, no diffusible molecule demonstrated to be involved in pigment pattern formation. If diffusion of chemicals really does not occur during pattern formation, how could a simulation based on a reaction-diffusion mechanism generate and predict the pigment pattern correctly? We guess this is because signaling via long projections can mimic the effect of diffusion in the conventional Turing (reaction-diffusion) model. Gierer and Meinhardt noted that essential conditions for autonomous pattern formation are the combination of short-range activation and long-range inhibition (Gierer and Meinhardt, 1972). ‘Diffusion’ itself is dispensable and can be replaced by any other means that can transmit signals over the appropriate spatial distance. Therefore, even using a mechanism very different from diffusion, a pattern can emerge by a similar dynamic.

Recent studies have shown that signal transduction via cellular projections is important in several developmental contexts (e.g. Sanders et al., 2013), and include intercellular communication via the Delta/Notch pathway (e.g. Cohen et al., 2010; Fitzgerald and Greenwald, 1995; Renaud and Simpson, 2001). We guess that the projection mechanism may participate in other morphogenetic events to which the involvement of a reaction-diffusion mechanism is suspected.

Studies used wild-type and brass^b2 mutant (Link et al., 2004) zebrafish supplied from the Zebrafish Stock Center at the Max Planck Institute (Tübingen, Germany) as well as csf1ra^j4e1 (Parichy et al., 2000). Breeding and the standard methods for fish manipulation were as described (Westerfield, 2000).

We cut caudal and anal fins from 1000 adult fish, and dissociated the pigment cells. Fins were incubated in a solution containing 2.5 mg/ml trypsin, 1.2 mg/ml bovine serum albumin and 1 mM EDTA for 12 hours at 4°C. After incubation, fins were washed five times with PBS for 5 minutes with shaking at 1000 rpm at 28°C, and incubated with 0.1% collagenase I for 60 minutes at 28°C. Then, cells were collected by the method described by Inaba et al. (Inaba et al., 2012).

Total RNA was isolated from the purified pigment cells using an RNeasy Mini Kit (Qiagen). We compared gene expression between melanophores and xanthophores with the Gene-Chip Zebrafish Genome Array (Affymetrix). All microarray data are available in the NCBI Gene Expression Omnibus under accession number GSE42129.

In situ hybridization of late larval [∼3.5 weeks post fertilization; ∼10 mm standardized standard length (Parichy et al., 2009)] fish was performed according to Larson et al. (Larson et al., 2010). Riboprobe was made from full-length 2635-bp cDNA for deltaC, and hydrolyzed to ∼500 nt before use. Hybridizations were performed at 68°C on wild-type and csf1ra^j4e1 mutant larvae.

Assessment of mRNA expression by pigment cells was carried out as described previously (Eom et al., 2012) using primers that spanned the intron. dct and aox3 were used as markers of melanophores and xanthophores, respectively (Kelsh et al., 2000; Parichy et al., 2000). PCR primers and the parameters used in this experiment are shown in supplementary material Table S1.

A stock solution (10 mM) of the γ-secretase inhibiter N-[N-(3,5-difluorophenacetyl)-l-alanyl]-S-phenylglycine-t-butyl ester (DAPT; Wako) was prepared in DMSO. The stock solution was dissolved in 50 ml fish water and adjusted to a final concentration of 10 μM. Fish water containing DAPT was changed every 3 days. After 15 days of incubation in DAPT, fish were moved to normal water. We counted the number of melanophores in the region of the middle trunk shown in supplementary material Fig. S2G and calculated the ratio of surviving melanophores (cell number at Day X/cell number at Day 0).

To express deltaC and NICD1a ectopically, we used the Tol2 transposon-based transgenesis system (Urasaki et al., 2006). Total RNA was isolated from fins using an RNeasy kit (Qiagen) and reverse transcription-PCR was performed using the Super Script III First-Strand Synthesis System (Invitrogen). Primer sets used to generate deltaC and NICD1a cDNA clone are as follows: deltaC-F, 5′-CCCAAGCTTATGGCTCGTGTTTTATTA-3′ and deltaC-R, 5′-CGCCATGGCTATACCTCAGTAGCAAA-3′ (this primer set amplifies whole deltaC open reading frame); NICD1a-F, 5′-ATGCAAGCTTATGGTGGTCGTCTCCAGGAAGAGG-3′ and NICD1a-R, 5′-GCATAGTCGACCTACTTGAAGGCTTCTGGAA-3′ [this primer set amplifies the intracellular domain (1745V∼2437K) of notch1a gene (Del Monte et al., 2007)].

Vector construction and insert position of the transgene are as described (Inaba et al., 2012). We used a 1.3-kb region upstream of the melanophore-specification gene microphtalmia-a (mitfa) to drive expression in melanophores (Lister et al., 1999). In our previous reports (Inaba et al., 2012; Watanabe and Kondo, 2012), we showed that identical promoter could rescue the melanophore defect. In Fig. 5 of this article, we observed the strong EGFP signal only in melanophores, further confirming the specificity of the promoter in adult fish. The same construct was used for both F0 screening and for further analysis with F1 fish.

The CAAX motif (Casey and Seabra, 1996) was fused with EGFP (Gomes et al., 2003; Wright and Philips, 2006) and introduced into pTol2-mitfa plasmid as illustrated in supplementary material Fig. S5D.

Plasmid DNA plasmid (30 ng/μl) and Tol2 mRNA (25 ng/μl) were co-injected to wild-type (deltaC/NICD1a) or brass^b2, csf1ra^j4e1 and wild-type (EGFP-CAAX) fertilized eggs at the one-cell stage. To confirm transgene insertions, we performed genomic PCR using specific primer sets.

Details of the laser ablation method have been described (Nakamasu et al., 2009). Before ablation, fish were anesthetized with 0.01% methyl-methanesulfonate (Westerfield, 1995) and mounted on a dish with water. Ablation was performed by 440-nm multiple light-pulse lasers from the MicroPoint pulse laser system (Photonic Instruments).

To disperse melanin-containing melanosomes into peripheral dendrites of melanophores, fish were placed in a black tank containing 100 μM yohimbine (Sigma-Aldrich) for 30 minutes before analysis.

Zebrafish (adult; ∼25 mm standard length) were anesthetized with MS-222 and cut into small pieces. The specimens were then fixed with 2.5% glutaraldehyde and 2% paraformaldehyde in 0.1 M cacodylate buffer for 60 minutes at room temperature. The fish pieces were then treated with 2% osmium in 0.1 M cacodylate buffer containing 4% sucrose for 60 minutes at room temperature. Then, samples were dehydrated through a graded ethanol series and embedded in Quetol-812 resin. Ultrathin sections were obtained with an Ultramicrotome (Reichert-Jung Ultracut E), counterstained with uranyl acetate and lead citrate, and examined on a transmission electron microscope (JEM-1200 EX, JEOL, Japan) with an acceleration voltage of 80 or 90 kV. Micrographs were recorded with a CCD camera (Morada) using the iTEM software (Olympus Soft Imaging Solutions).

We are grateful to K. Kawakami for providing the Tol2 vector.