A Systematic Survey of Expression and Function of Zebrafish frizzled Genes

Wnt signaling is crucial for the regulation of numerous processes in development. Consistent with this, the gene families for both the ligands (Wnts) and receptors (Frizzleds) are very large. Surprisingly, while we have a reasonable understanding of the Wnt ligands likely to mediate specific Wnt-dependent processes, the corresponding receptors usually remain to be elucidated. Taking advantage of the zebrafish model's excellent genomic and genetic properties, we undertook a comprehensive analysis of the expression patterns of frizzled (fzd) genes in zebrafish. To explore their functions, we focused on testing their requirement in several developmental events known to be regulated by Wnt signaling, convergent extension movements of gastrulation, neural crest induction, and melanocyte specification. We found fourteen distinct fzd genes in the zebrafish genome. Systematic analysis of their expression patterns between 1-somite and 30 hours post-fertilization revealed complex, dynamic and overlapping expression patterns. This analysis demonstrated that only fzd3a, fzd9b, and fzd10 are expressed in the dorsal neural tube at stages corresponding to the timing of melanocyte specification. Surprisingly, however, morpholino knockdown of these, alone or in combination, gave no indication of reduction of melanocytes, suggesting the important involvement of untested fzds or another type of Wnt receptor in this process. Likewise, we found only fzd7b and fzd10 expressed at the border of the neural plate at stages appropriate for neural crest induction. However, neural crest markers were not reduced by knockdown of these receptors. Instead, these morpholino knockdown studies showed that fzd7a and fzd7b work co-operatively to regulate convergent extension movement during gastrulation. Furthermore, we show that the two fzd7 genes function together with fzd10 to regulate epiboly movements and mesoderm differentiation.


Introduction
During development a small number of types of signaling factors are utilized repeatedly to allow this complex and beautiful process to unfold. These signals are usually transduced by members of an equally small number of receptor molecule families, raising major questions as to how distinct responses can be generated by the same limited number of signals, and in what ways signals generated by individual members of the receptor families may be different. The Wnt family of signaling proteins comprises a key example of these signals, playing diverse roles throughout all stages of vertebrate and invertebrate embryonic development [1,2]. The Wnt family proteins are cysteine-rich secreted glycoproteins, and they have been shown to be crucial for a variety of developmental events, such as formation of ventral mesoderm during gastrulation [3,4], posteriorisation of neural plate [5,6], induction of neural crest cells (NCCs) from ectoderm, and then melanocyte specification from these NCCs [7]. Wnt signaling is generally thought to be transduced principally by seven-path transmembrane proteins of the Frizzled (Fzd) family and their co-receptors, low-density lipoprotein receptor-related protein (LRP) 5, 6, with the Wnt ligands interacting with a cysteine rich domain (CRD) in the extracellular part of the Fzd protein [2,8]. Thus it is vital that we characterize the specific Wnt and Fzd proteins associated with each of the developmental processes associated with Wnt signaling, as a crucial starting point from which to begin to understand the basis for the specific responses triggered in each case. Surprisingly, however, our knowledge of the expression patterns and functions of the fzd gene family remains limited to only a small subset of genes and of developmental roles [9,10]. Therefore, we wanted to carry out a comprehensive analysis of fzd gene expression patterns with respect to some of the early developmental roles for Wnt signaling. To allow comprehensive assessment of the entire complement of fzd genes we chose to use the zebrafish model system: The centralized repository of gene and genomic information (ZFIN, http://zfin.org/; Ensembl, http://www.ensembl.org/ Danio_rerio/Info/Index) allowed rapid identification of the full complement of fzd genes. After examining the expression patterns, the well-established method of single or multiple gene knockdown using morpholino oligos [11] allowed us to explore gene functions, including allowing for the expected functional redundancies. Such a comprehensive survey of expression and functional analysis would be expected to provide crucial information allowing more focused investigations of Wnt signaling for specific functions.
For this approach, we chose to focus functionally on three important developmental roles of Wnt signaling, which we explored as test cases; melanocyte specification, NCC induction and convergent extension (CE) movements during gastrulation. NCCs form a transient embryonic structure, and are remarkable for producing many distinct cell types, including pigment cells, peripheral neurons, glial cells, and jaw cartilage [12]. Among NCC-derivatives, melanocytes have been extensively studied because their characteristic color and morphology makes them highly amenable to genetic screening [13,14]. Previous work has identified a conserved role for Wnt signaling in specification of melanocytes from NCCs [15]. This system thus provides a paradigm for understanding the role of Wnt signals in fate specification from multipotent progenitor (stem) cells [16,17]. Activation of canonical Wnt signaling by injection into zebrafish embryos of mRNA encoding an activated form of ß-Catenin results in NCCs adopting melanocyte fates at the expense of the neuronal and glial lineages [17,18]. Conversely, conditional knockout of ß-Catenin in mice also showed the requirement of Wnt signaling for formation of melanocytes and their precursors [19,20]. Analysis of the expression patterns of various Wnt ligands suggests Wnt1, Wnt3, Wnt3a and Wnt4 are all expressed in largely overlapping regions in the dorsal midline of neural tube [18,21]. The lack of melanoblast marker gene expression after combinatorial knockout of Wnt1 and Wnt3a in mice [22] supports the idea that Wnt1 and Wnt3a are primary ligands promoting melanocyte specification. In contrast, the relevant Wnt receptor genes remain unknown.
In addition, Wnt signaling has a crucial role in NCC induction [7,8]. NCCs, like placodal cells, are formed at the border of neuroectoderm and non-neuroectoderm early in neural tube formation. Analysis using several model organisms demonstrates that activation of Wnt signaling combined with attenuation of bone morphogenetic protein (BMP) signaling is important to induce NCCs instead of placodal cells [23,24,25]. In the zebrafish, Wnt8 has been shown to be a major ligand required for NCC induction [26]. fzd7 has been identified as a key receptor required for NCC induction in Xenopus [27]. However, the fzd7 gene is also required for cell movement and polarity in vertebrate development [28], suggesting the possibility that other fzd genes may be involved in NCC development and may provide the specificity.
Finally, non-canonical Wnt signaling plays a pivotal role in CE movements [29,30,31]. During gastrulation, cells that give rise to the ectoderm, mesoderm and endoderm undergo coordinated movements towards the dorsal side (convergence), and those cells intercalate around the dorsal midline resulting in elongation of the embryo along its antero-posterior body axis (extension). In contrast to our detailed understanding of the genes required for specification of cell types along the dorsoventral and anteroposterior axes, those controlling this coordinated movement remain poorly understood. Furthermore, as Wnt components required for CE movements are thought to contribute to the invasiveness of cancer tissues [32], identification of the molecular mechanisms underlying Wnt signaling in this process may well be informative for understanding the molecular mechanisms underlying cancer cell invasion and metastasis.
Comprehensive expression analysis of fzd gene expression would be expected to identify candidate fzd genes relevant to each of these developmental roles of Wnt signaling. Here, we carried out in situ hybridization analysis for thirteen fzd genes at multiple developmental stages relevant to the three developmental processes selected, and identified fzd genes that were candidates for mediating each of them. Morpholino knockdown of these identify Fzd7a, Fzd7b and Fzd10 as key receptors mediating Wnt signaling in CE movements and mesoderm differentiation. Surprisingly, morpholino-mediated knockdown, alone or in combination, of all the fzds considered to be candidates for melanocyte specification and NCC induction, failed to generate the expected phenotypes, suggesting the importance of ubiquitously or low level expressed fzd genes in mediating these processes.

Fish husbandry
Wild type (AB) strain and TOP:dGFP transgenic line (kindly provided by Dr. Steve Wilson) were maintained under a light and dark cycle of 14 and 10 hours, respectively. Embryos collected from natural spawnings were cultured at 28.5uC, and staged according to Kimmel et al [33]. Embryos for in situ hybridization were treated with 0.003% 1-phenyl-2-thiourea from 10 hpf (hours post-fertilization) to prevent pigmentation. The sox10 m618 strain was maintained as heterozygous carriers [14].

Cloning of fzd genes
The primer sequences used for cloning are shown in Table 1. cDNAs for cloning were prepared from 24 hpf wild type AB embryos using TRI reagent (Sigma) and superscript III (Invitrogen). To amplify specific fragments by polymerase chain reaction (PCR), we used GoTaq polymerase (Promega); amplified fragments were cloned into pGEM-T easy vector (Promega) by standard TA-cloning. Cloned fragments were identified by restriction digestion and partial sequencing. All cloned fzd genes showed .98% nucleotide identity to the reported sequences, although fzd2 showed a 38 bp deletion in the 39UTR and is likely to represent a splicing variant. We further validated our clones by comparison of expression patterns with those reported for zebrafish and/or other vertebrates.

Probe synthesis and whole mount in situ hybridization
The enzymes used for making probes are also in Table 1. fzd7b cDNA in pSPORTvector was kindly provided by Dr. Vladimir Korzh [28]. It was linearized and transcribed with Sal I and SP6, respectively. Whole mount in situ hybridization was carried out as described before [34]. All embryos were cleared by transferring through a series of glycerol buffered by PBS (phosphate buffered saline). Images were taken under compound microscope (NIKON Eclipse E800) using a Digital Sight DS-U1 camera (Nikon). Image processing was carried out using Photoshop (Adobe).

Morpholino injection
Sequences of all morpholinos used in these experiments are in Table 2. Morpholinos against fzd9b and fzd10 were designed for this work (GeneTools LLC.). For fzd3a-MO, we used the previously published sequence [35], while for fzd7a and fzd7b we obtained sequence information for effective MOs from Dr. Vladimir Korzh (pers. comm.). As negative controls, we used either standard morpholino (CCTCTTACCTCAGTTACAATT-TATA) or random control oligo 25-N provided by GeneTools LLC as noted. Morpholinos were suspended in 1xDanieau's solution (58 mM NaCl, 0.7 mM KCl, 0.4 mM MgSO 4 , 0.6 mM Ca(NO 3 ), 2.5 mM HEPES, pH 7.6), and injected into yolk at 1-4 cell stages. Embryos were cultured until appropriate stages at 28.5uC, and fixed for in situ analysis. The fzd3a-MO used here was designed to disrupt the splicing out the first intron (16 kb) of the fzd3a transcript, which allowed us to assess the efficacy of this morpholino simply by reverse transcription (RT)-PCR assay as previously reported [35]. Briefly, embryos injected with fzd3a-MO were collected at 24 hpf, and total RNAs were extracted with TRI reagent (Sigma) for reverse transcription using Superscript III (Invitrogen) and following primers designed against the two fzd3a exons flanking the first intron as reported [35]; forward: 59-ATGCTGACTGTATGCATGGCC-39, reverse: 59-GCCA-TAATCCCGTTGAACTGC-39. The size of the normal fzd3a transcript amplified with these primers should be 465 bp. Under our PCR conditions, this PCR fragment cannot be detected in the experimental group due to the length of the first intron (16 kb), but is readily seen when splicing occurs normally. As a positive control for RT-PCR reaction, we used primers to amplify b-actin cDNA (Forward: 59-TACCCCATTGAGCACGGTAT, reverse: 59-GTTCCCATCTCCTGCTCAAA).

Ethics statement
This study was performed with the approval of the University of Bath ethics committee and in full accordance with the Animals (Scientific Procedures) Act 1986 under Home Office license PPL30/2415.

Identification of the full complement of zebrafish fzd genes
To allow a comprehensive analysis of expression patterns of fzd genes, we utilized two databases, ZFIN (http://zfin.org/) and Ensembl (http://www.ensembl.org/Danio_rerio/Info/Index). At the start of this project, twelve fzd genes were already registered with their cDNA sequence in ZFIN; these fzd genes were as follows; fzd1, fzd2, fzd3a, fzd3b, fzd4, fzd5, fzd7a, fzd7b, fzd8a, fzd8b, fzd9b, and fzd10 (See Table 1 for accession numbers of sequences used here). Furthermore, using the full coding sequence of reported fzds or cDNA sequences corresponding to their cysteine rich domain (CRD) when full coding sequence was not available, we screened for fzd-related genes in the zebrafish genomic DNA database in Ensembl (Zv8); a repeat of this procedure in the Zv9 release of the zebrafish genome identified the same genes. Most of those we identified were already reported in ZFIN, but fzd9a was newly found (see Table 1 for accession number.). Also, the fzd6 gene was published during the course of this survey [36]. Thus, we identified orthologues for all of fzd1-fzd10, representing the full complement of fzds known from other vertebrates. We cloned subregions suitable for making probes for all these genes to examine their expression patterns in embryos, with the one exception of fzd8b which has been clearly shown to have expression in a very tightly localized pattern in the forebrain at 1-somite stage and later in the neurohypophysis [37]). Primers for cloning cDNAs were designed to amplify as long a region as possible of the 39UTR to ensure gene specificity of probes (See Table 1 for sequences of primers used). When information for the 39UTR was limited or unavailable, we designed primers to clone a 1-1.5 kb fragment encompassing the 39 region of the cDNAs (See Figure S1 for location of each probe.). We prepared antisense riboprobes from these cloned partial cDNAs, and performed whole mount in situ hybridization analysis on wild-type embryos at three different stages encompassing the key stages for NCC induction and melanocyte specification. Firstly, we examined 1-somite stage embryos since this is when early NCC markers are first seen mainly in the lateral border of the neural plate [38,39]. We expected that fzd genes relevant to the induction of NCC would be expressed in part or all of the ectoderm in the neural plate, prospective epidermis or both. Secondly, melanocyte specification is first detected as induction of mitfa expression, and begins around 18 hpf [40]. At this stage, sox10, which plays a key role in specification of melanocytes by direct activation of mitfa [41] is expressed uniformly in NCCs after delamination (premigratory NCCs) and, more posteriorly (i.e. in developmentally younger cells), in dorsal neural tube cells (NCCs prior to delamination) [38]. Consequently, we expected fzd genes driving melanocyte specification would be expressed in dorsal neural tube or premigratory NCCs at 18 hpf. We also examined embryos at 30 hpf when melanoblasts are readily identified among migrating NCCs on both the medial and lateral migration pathways.
All fzd genes showed specific expression patterns at one or more of the stages examined ( Fig. 1 and Fig. 2). fzd3a, fzd7b, fzd9b and fzd10 were all expressed in regions consistent with possible roles in NCC induction or melanocyte specification; three candidates for melanocyte specification (fzd3a, fzd9b, fzd10) will be described in detail in the next section. fzd7b is expressed in a pattern consistent with potential involvement in NCC induction, and will be described later in this section. Based on the expression patterns we observed, the other genes were considered highly unlikely to have roles in the processes of development that form the focus of this paper, nevertheless the expression patterns of fzd1, fzd2, fzd3b, fzd4, fzd5 and fzd8a will be described briefly (Fig. 1A, 1B, 1C, 1D, 1E, 1I, respectively). We found ubiquitous expression of some fzd genes at 1-somite (fzd6, Fig. 1Fa; fzd7a, Fig. 1Ga) and 18 hpf (fzd6, Fig. 1Fb; fzd7a, Fig. 1Gb, fzd9a, Fig. 1Jb) stages, but no sign of specific expression of any of these genes in early melanoblasts in the 30 hpf stage (Fig. 1Fc, 1Gc, 1Jc); again, these were considered unlikely to be important in NCC formation or melanocyte specification (see Discussion).
fzd1 did not show expression at 1-som. stage (Fig. 1Aa, 1Aa') whilst expression in the central nervous system (CNS) started to appear from 18 hpf (Fig. 1Ab, 1Ac). As has been described before [42], fzd2 was expressed in the forming somite at 1-somite stage (white arrowhead in Fig. 1Ba). In addition, we found fzd2 expression in the fin and dorsal neural tube at 30 hpf (Fig. 1Bc), although there was no expression in the trunk consistent with migrating NCCs. fzd3b showed significant expression in a dorsal part of telencephalon and diencephalon at 18 and 30 hpf (arrowheads in Fig. 1Cb, Cc). fzd4 was expressed strongly as bilateral stripes located around the border of neural plate (arrowheads in Fig. 1Da''). Based on the expression pattern at 18 hpf (Fig. 1Db'), we interpreted these as marking future cranial placodes instead of NCCs. To test this idea, we performed two color in situ hybridization using probes for fzd4 and foxd3, an early marker for NCC, at 1-somite stage as shown in Fig. 1Da'''. fzd4 expression (purple) was clearly expressed in a region lateral to the foxd3 expression domain (red), confirming that fzd4 is not expressed in NCC, and is indeed likely to be expressed in the cranial placodes. fzd5 expression is highly restricted to eye (Fig. 1Eb, 1Ec), forebrain ( Fig. 1Ea-c) and gut (arrowheads in Fig. 1Eb, 1Ec) during development. fzd8a expression was seen in forebrain ( Fig. 1Ia-c) and gut (arrrowheads in Fig. 1Ib, 1Ic), with strong expression becoming more widespread in the ventral CNS by 30 hpf (Fig. 1Ic).
fzd7b showed faint expression around the border of the neural plate in addition to strong expression in the future mid-and hindbrain at 1-somite stage (Fig. 1Ha, Ha'). It is possible that this bilateral expression pattern simply marks the future cranial placode area like fzd4. However, we chose the fzd7b gene as a candidate for further analysis of NCC induction because it is reported that the fzd7 gene is required for NCC induction in Xenopus [27], and our two color in situ hybridization showed that fzd7b expression in the mid-and hindbrain region overlaps the foxd3 expression domain, marking NCC ( Figure S2). Taken together, we concluded that fzd7b was a good candidate for a role in NCC induction. Therefore, we explored this hypothesis in more detail (see later).
Expression patterns of fzd3a, fzd9b, fzd10 at NCC induction and pigment cell formation stages Unexpectedly, we were unable to detect expression of any of the fzd genes in trunk NCCs, including differentiating melanocytes and melanoblasts, at 30 hpf ( Fig. 1 and Fig. 2). Nevertheless, three genes, fzd3a, fzd9b and fzd10, showed expression patterns that included some or all of the dorsal neural tube at 18 hpf (Figs. 2B, 2E, 2I, 2b, 2e, 2i), a place where melanocyte specification is likely to occur. These genes showed distinct patterns (Fig. 2).
fzd3a expression in the neural plate is first detected in an anterior domain at 1-somite stage ( Fig. 2A). Optical sections at the level of the posterior hindbrain (level labeled with ''a'' in Fig. 2A) shows that expression extends throughout almost the entire neural plate (Fig. 2a). By 18 hpf, fzd3a expression has expanded to encompass the entire length of the neural tube, although expression remained strongest in the cranial regions (Fig. 2B). Optical sections again revealed that the fzd3a expression domain extends throughout the dorsal part of the neural tube (Fig. 2b). However, we found no evidence of fzd3a expression in migrating NCCs in trunk or tail at 30 hpf, although expression persists strongly throughout the CNS (Fig. 2C).
fzd9b expression is much weaker than that of the other two genes at 1-somite stage (Fig. 2D), although it includes most of the neural plate in anterior half of the embryos. A pair of expression domains outside of the neural plate (arrowheads in Fig. 2d) were considered not to include NC because they lie internal to the ectoderm. By 18 hpf, fzd9b expression is readily detected in a complex pattern within the fore-and midbrain; levels in the hindbrain and spinal cord are also much higher, and posteriorly the expression domain has extended to the tail (Fig. 2E).
Interestingly, the dorsal part of the anterior CNS (marked by bracket) lacks fzd9b transcripts whilst the dorsal neural tube in the spinal cord clearly expresses fzd9b (Fig. 2e). As with fzd3a, we saw no evidence of fzd9b expression in migrating NCCs at 30 hpf, although expression is seen in sites including most of the CNS, the tail tip, and at low levels in muscle (Fig. 2F).
fzd10 is expressed strongly in the neural plate at 1-somite stage, but expression is absent from the anteriormost region, likely corresponding to some or all of the forebrain, and is strongest in the anterior part (Fig. 2G) and lateral border (Fig. 2H) of the neural plate expression zone, consistent with possible involvement in NCC induction. In addition there is very strong expression in the tail bud, which also persists at 18 hpf (Fig. 2I). At later stages (18 hpf, 30 hpf), fzd10 is strongly expressed and tightly restricted to the dorsal neural tube along the entire axis of the embryo (Fig. 2I, 2i, 2J). Once again, expression in migrating NCCs was not detected (Fig. 2J).

Knockdown of fzd3a, fzd9b and fzd10
Given their expression patterns in the dorsal neural tube, fzd3a, fzd9b and fzd10 were strong candidates for encoding the Wnt receptor(s) associated with melanocyte specification from the NCC. To test this hypothesis, we performed morpholino-mediated knockdown analysis of each gene; in addition, to allow for likely redundancy of function, we took advantage of the ease with which multiple gene knockdown can be performed using this technique in zebrafish, knocking down all three genes simultaneously. A splice site morpholino targeting fzd3a has already been reported [35], so we used this same morpholino for our studies. Our analysis of the cDNA and genomic data (Ensembl) for fzd9b and fzd10 indicated that there were no intron-exon boundaries within the known coding regions of these two genes; consequently, we designed morpholinos against the predicted translation start (ATG) codon. Sequence information for these morpholinos is listed in Table 2. To examine the efficacy of fzd3a-MO we used reverse transcription-polymerase chain reaction (RT-PCR) to confirm the changed size of transcript due to mis-splicing as reported [35] (See Materials and Methods for detail). As shown in Fig. S3A, 10 ng of the fzd3a-MO was sufficient to substantially suppress normal splicing of fzd3a in our hands. To show that our fzd9band fzd10-MO can suppress translation of their target sequences in vivo, we designed fusion constructs containing target sites of these morpholinos and an enhanced green fluorescent protein (egfp) reporter gene (Schematic drawing in Fig. S3B). Co-injection of morpholinos and synthesized RNA encoding fusion constructs containing the target sequence demonstrated that even 1 ng of fzd9bor fzd10-morpholino was sufficient to suppress EGFP fluorescence derived from 100 pg of fusion construct mRNA (EGFP-positive embryos, 0/31 for fzd9b-MO injected; 0/23 for fzd10-MO injected) (See Fig. S3D, F, S3D', F'). In contrast, EGFP protein was detected in a significant proportion of embryos injected with even 5 ng of random control morpholino in both cases (EGFP-positive embryos, 73/93 for fzd9b-MO experiment; 23/28 for fzd10-MO experiment) (See Fig. S3C, E, 3C', E'). Although these data indicate the efficacy of our morpholinos against their chosen targets, morpholino knockdown of each of fzd3a, fzd9b or fzd10 alone showed no noticeable reduction of melanocytes ( Figure S4 and Table S1).
We tested for functional redundancy of these three genes in this process by co-injecting all three morpholinos simultaneously to achieve multiple gene knockdown. In initial experiments, we experimented with differing amounts of MOs for each gene. Injected embryos were examined at 1 day post-fertilization (dpf) (Fig. 3A-F and Table 3) and 3 dpf (Fig. 3G-3L), noting the presence of and pattern of melanocytes, as well as morphological abnormalities. Since we knew that 10 ng/embryo of fzd3a-MO efficiently inhibited fzd3a splicing, we fixed the amount of fzd3a-MO to 10 ng/embryo, and varied the amounts of the others. The combined injection of 10 ng of fzd3a-MO and 5 ng of fzd10-MO resulted in substantial necrosis, so we used only doses of 1 or 2 ng/ embryo of fzd10-MO in this experiment. Although combinatorial injection of these MOs caused a percentage of embryos to show severe necrosis and in some cases a severely truncated or bended body axis, in all cases at least c. 50% of injected embryos showed an elongated body axis and no or only minimal necrosis restricted to the brain. Surprisingly, we saw no reduction in melanocytes in those embryos with normal morphology, even in the embryos injected with the highest amounts of morpholino (i.e. fzd3a, fzd9b, fzd10-MOs at 10, 5, 2 ng/embryo, respectively; compare Fig. 3B, 3H with Fig. 3A, 3G.). We found less pigmented embryos in some cases, but all of these showed very severe morphological defects (Table 3; data not shown). Therefore, we discounted these pigmentation defects as secondary defects of disruption of early development. The embryos injected with lower doses of morpholinos also showed no reduction of pigmentation ( Fig. 3C-F, 3I-L). The significant developmental abnormalities seen in the doses used here precluded the further increase in MO doses.
One possible explanation for our failure to see the expected melanocyte specification phenotype was that the MOs failed to inhibit Fzd function sufficiently. Our efficacy controls (Fig. S3) argue strongly against this caveat, but we also performed a direct test of the MOs' efficacy in reducing canonical Wnt signaling in vivo at the expected time of Wnt signaling. We injected embryos generated by intercrossing hemizygous TopdGFP transgenic fish with fzd3a, fzd9b, fzd10-MOs, and examined gfp expression by in situ hybridization analysis. This strain contains a transgene combining multiple Lef1-binding sites followed by a destabilized gfp cDNA, and has been shown to be a good in vivo reporter of Wnt activity [43]. In these crosses, we expected embryos to be homozygotes, hemizygotes and non-transgenic; consistent with this, we were able to categorize the signal intensity of GFP expression into three categories, strong, weak and no expression. Figure S5 shows typical examples of embryos showing ''strong'' (Fig. S5A) and ''weak'' (Fig. S5B, C) expression, and Table S2 describes the distribution of embryos categorized into these three categories. As shown in this table, the number of embryos classified as ''strong'' expression is reduced by the co-injection of fzd3a, fzd9b and fzd10 MOs (2 out of 37) compared with the controls (11 out of 37). Statistical testing confirmed that the proportion of the three GFP expression patterns deviated from the expected Mendelian ratio (1:2:1), and the one observed in fzd-3, fzd9b, fzd10morphant was significantly different from the prediction (Chi-squared, p = 0.022), whilst the one for control (standard morpholino-injected) was not (p = 0.868). We conclude that co-injection of the three fzd MOs substantially reduces Wnt signaling in vivo.
Whilst our data clearly indicates that we achieved very substantial knockdown of Wnt signaling, we considered the possibility that nevertheless the remaining residual activity (See Fig. S5C) was sufficient to drive normal melanocyte specification. Current experimental data indicates that melanocyte specification depends upon a combination of Wnt signaling [44] and Sox10 activity [34,38]. We thus reasoned that co-injection of the fzd MOs into sox10 mutant heterozygotes might provide a sensitized background in which the roles for these fzd genes in melanocyte specification would be revealed. We crossed identified sox10 mutant heterozygotes with wild types, and injected the resultant embryos (expected to be 50% sox10 heterozygotes; 50% wild-types) with fzd3a, fzd9b, fzd10-MOs at various concentrations, alone or in combination. We quantitated the numbers of embryos showing reduced melanisation, expecting at most 50% of the injected embryos show reduced melanisation. Surprisingly, we again saw that almost all of the injected embryos showed normal melanisation at 1 dpf (Fig. 4B-F) and 3 dpf (Fig. 4H-L)(See also Table 4).
Taken together, we conclude that our data argue against a major role for fzd3a, fzd9b or fzd10 in melanocyte specification (or subsequent differentiation). It seems likely that ubiquitously expressed fzd genes (e.g. fdz6, fzd7a, fzd9a) and/or low level expression of other fzd genes, or other types of receptors, transduce sufficient Wnt signaling for melanocyte development (see Discussion).

Examining the role of fzd genes in NCC induction
In our analysis of the expression patterns of fzd genes, only fzd7b and fzd10 gene showed expression in the lateral border of neural keel at 1-somite stage (Fig. 1Ga' and Fig. 2H). We thus hypothesized that these genes might mediate Wnt signaling during NC induction, and tested this idea by using morpholino-mediated knockdown. In addition, given the possibility of functional redundancy between fzd7a and fzd7b, we knocked down both fzd7a and fzd7b genes simultaneously. Morpholinos against the two fzd7 genes have already been reported [28], and some other morpholinos against these were also examined by the same group. So we chose morpholino sequences based on their observations [28] (V. Korzh, pers. comm.) (for sequences, see Table 1). As marker genes for early NCCs, we used foxd3, pax3 and sox10. These three genes encode transcription factors of the winged helix, paired box containing and HMG group transcription factor families respectively, and all have been shown to be expressed in premigratory NCC and to be critical factors for development of neural crest derivatives [34,38,45,46,47]. Importantly, previous reports suggested that their expression was dependent upon Wnt signaling [26,48]. Fig. 5 shows the expression pattern of these three genes in control MO-treated embryos at 1-somite stage ( Fig. 5A-C). Simultaneous inactivation of fzd7a, fzd7b and fzd10 has no noticeable effect on the intensity of marker gene expression. However, these experiments clearly showed that the lateral spacing between the bilateral marker domains was increased in fzd MO-treated embryos, suggesting that CE movements during gastrulation might be disrupted [49,50].
We then considered the possibility that the lack of a NCC induction phenotype might be due to other fzd genes. In particular we noted that fzd3a and fzd9b are also expressed in the neural plate, though these two are not restricted to the border between neural plate and epidermis ( Fig. 2A, 2a, 2D, 2d). Consequently, we examined the effect of co-injection of fzd3a, fzd7a, fzd7b, fzd9b and fzd10-MO on NCC marker expression at 1-somite stage (Fig. 5G-I), but again saw no defect in NCC marker expression pattern or intensity, although the wider spacing of the lateral marker domains was seen. Interestingly, subsequent careful examination of the expression pattern of fzd10 in comparison with the early NCC marker foxd3 by double in situ hybridization, showed that the fzd10 expression domain does not in fact overlap foxd3 at this stage (See Figure S2). From these observations, we conclude that knockdown of fzd7a, fzd7b and fzd10 is insufficient to affect NCC induction (See Discussion), but might be important for CE movements.  Table 3 are presented here for each injection. G-L) 3 dpf embryos. All are left side views with dorsal oriented to the top. A, G) Control standard-MO, 17 ng/embryo. B, H) fzd3a, fzd9b, fzd10-MO = 10, 5, 2, ng/embryos, respectively. C, I) fzd3a, fzd9b, fzd10-MO = 10, 5, 1, ng/embryos, respectively. D, J) fzd3a, fzd9b, fzd10-MO = 10, 2, 2, ng/embryos, respectively. E, K) fzd3a, fzd9b, fzd10-MO = 10, 2, 1, ng/embryos, respectively. F, L) fzd3a, fzd9b, fzd10-MO = 10, 1, 1, ng/embryos, respectively. Scale bars: 200 mm. doi:10.1371/journal.pone.0054833.g003 Table 3. Melanisation occurs in embryos injected with fzd3a, 9b, 10-MO at 1 dpf. Notes. Numbers in parenthesis are of significantly melanised embryos. We did not inject 5 ng/embryo or more of fzd10 MO with other morpholinos because we found combinatorial injection of 10 ng of fzd3a-MO and 5 ng of fzd10-MO caused severe necrosis (data not shown). Here, ''severe'' necrosis means we found necrotic cell death everywhere in the body, whilst ''slight'' means necrosis was observed mainly in the head. Most embryos show elongated body axis, and those are categorised into ''elongated axis''; others, categorised as ''shortened axis'', lack tail and sometimes have severely bent or reduced trunk. doi:10.1371/journal.pone.0054833.t003 Co-operative involvement of fzd7a and fzd7b on CE movement and requirement for fzd7a, fzd7b and fzd10 in mesoderm formation To investigate the requirement of these three fzd genes in CE, we first examined their expression at early gastrula stages when CE movements occur [51,52]. All three fzd genes were expressed in gastrula stages, but each had a unique pattern (Fig. 6). At 6 hpf, fzd7a and fzd10 are expressed in the blastoderm margin (arrowheads in Fig. 6A for fzd7a and in Fig. 6G for fzd10), whilst fzd7b expression was throughout the blastoderm but excluded from the marginal area (Fig. 6C). Like fzd7b (Fig. 6C), fzd10 expression is reduced or lacking from the embryonic shield (Fig. 6E, G). At the 80% epiboly stage, the expression domains of fzd7a and fzd7b expand to become widespread throughout the entire embryo, except for the marginal region of the dorsal blastoderm (arrows in Fig. 6B, D). In contrast, fzd10 expression remains restricted to the ventro-lateral marginal zone of blastoderm (arrowheads in Fig. 6H). These in situ analyses clearly demonstrate expression patterns of these three fzd genes that are fully consistent with their possible involvement in CE movement.
The expression patterns described in Fig. 6 suggest that, in addition to the CE defects we have observed in the ectoderm (Fig. 5), we might expect CE defects in the mesoderm too. To test this, we examined the standard markers used to assess mesodermal CE movements, no tail (ntl) [53] and paraxial protocadherin (papc) [54]. ntl is expressed in the blastoderm margin and the presumptive notochord from early gastrula to 1-somite stage, whilst papc is strongly expressed in paraxial mesoderm and excluded from the dorsal midline. Previous studies have shown that the ntl notochord expression domain becomes shortened along the antero-posterior axis and widened mediolaterally when CE is inhibited [55,56]. Conversely, the spacing between the two papc expression domains becomes wider and the expression domains shorter where CE is  Table 4   disrupted [49]. At first, we knocked down fzd7a, fzd7b and fzd10 separately by injecting with 5 ng of morpholino for each. Figs. 7A-7D and 7E-7H show the expression pattern of ntl (7A-7D) and papc (7E-7H) in each case at 1-somite stage. As shown here, fzd7a and fzd7b-MO have no significant effect on the length and width of these expression patterns (compare panel 7A or 7E with 7B, 7C or 7F, 7G), though papc expression seems slightly shortened in fzd10morphants (Fig. 7H).
Since we expected functional redundancy between fzd7a and fzd7b, we next tested experimentally the effects of double knockdown of fzd7 function, and also co-injection with a series of doses of fzd10-MO with the fzd7a, fzd7b MOs in order to assess the co-operative action of these three genes. We injected embryos with standard-MO as control (Fig. 7I), or fzd7a-MO and fzd7b-MO together (Fig. 7J), or fzd7a-MO, fzd7b-MO and fzd10-MO together (Fig. 7K, 7L) and examined the expression of ntl in embryos fixed at 1-somite stage. The combined injection of only 2 ng each of fzd7 morpholinos resulted in dramatic shortening of the ntl domain (compare Fig. 7I and 7J), consistent with a CE defect. To quantify this phenotype, we measured the length of the notochord domain of ntl in these embryos (Fig. 7M). Statistical analysis by one-way ANOVA and further Tukey's multiple comparison analysis confirms the statistically significant reduction of notochord length in the fzd7a, fzd7b double morphants compared with control (p,0.001). These data strongly suggests that fzd7a and fzd7b act in a functionally redundant manner to mediate Wnt signals important for CE movement. Next, we assessed whether fzd10 might act in addition to the two fzd7 genes to contribute towards this role in CE movement. When compared with the control embryo (Fig. 7I), combined injection of fzd7a, fzd7b and fzd10-MOs resulted in a severely truncated and broadened ntl expression domain, although this was only seen with 5 ng of fzd10-MO (Fig. 7L, 7M) and not with 1 or 2 ng doses (compare Fig. 7J and 7K; quantitation in Fig. 7M). Further, we also observed an effect on epiboly, with 5 ng of fzd10-MO causing a delay in progression of epiboly, and reduction in ntl expression in the marginal zone (compare Fig. 7I and 7L). From these observations, we conclude that fzd7a and fzd7b are the primary Wnt receptors driving proper CE movement during gastrulation, and that fzd10 does not have a major role. However, reduction of ntl expression in embryos injected with higher amounts of fzd10-MO as well as both fzd7a and fzd7b-MOs suggests that all act together to regulate epiboly movements and early mesoderm differentiation.

Discussion
In this study, we have comprehensively surveyed the expression patterns of the fzd genes in zebrafish development up to 30 hpf. In doing so, we identify discrete and diverse expression domains for most of these genes, although there is much overlap at specific times and in specific organs, and three genes show apparently ubiquitous expression at at least one stage of the three we have focused on. These data are consistent with the diverse and repeated uses of Wnt signaling in development, and the challenge now is to relate individual fzd gene expression patterns to specific roles of Wnt signaling. In an attempt to begin to address this challenge, we focused initially on two well-characterized processes known to be dependent upon Wnt signaling, NCC induction and melanocyte specification from the NC.
We identified fzd3a, fzd9b and fzd10 as being expressed in the dorsal neural tube, where melanocyte specification likely occurs. Unexpectedly, we were unable to identify conditions of single or combinatorial morpholino-mediated knockdown of these Fzds that resulted in phenotypes consistent with an inhibition of melanocyte specification. Likewise, we identified fzd7b and fzd10 as being expressed in the neuro-ectodermal border at 1-somite stage, and tested these for a role in NCC induction by morpholino-mediated knockdown. Again, unexpectedly we were unable to identify a clear role for either of these genes in NC induction. However, these experiments identified a clear phenotype in CE movement, suggesting that fzd7a, fzd7b and fzd10 act in a partial redundant manner to mediate this aspect of Wnt signaling.
Are fzd genes required for NCC induction and melanocyte specification?
Our experiments using morpholino combinations to knockdown up to five (fzd3a, fzd7a, fzd7b, fzd9b, and fzd10) or up to three (fzd3a, fzd9b and fzd10) Fzd proteins failed to result in detectable reduction of either NCC or melanocyte formation, respectively. Nevertheless, we have evidence that our morpholinos were generating substantial knockdown of the targets. Firstly, the absence of normal length transcripts for fzd3a after injection of fzd3a-MO suggests this morpholino was successfully suppressing normal splicing. Secondly we observed a CE defect and reduction of mesodermal markers when fzd7a, fzd7b, fzd10-MO were injected into embryos. Thirdly, we showed that our fzd9b-MO could bind to the target sequence and suppressed the translation of an egfp reporter gene at doses of only 1 ng per embryo, which is much less than the amount injected for analysis. Fourthly, reduction of GFP reporter expression in TopdGFP transgenic analysis injected with fzd3a, fzd9b, fzd10-MO indicated that we were achieving substantial suppression of the canonical Wnt pathway. Finally, in fzd3a, fzd9b, fzd10-morphants, we found frequent tail fin defects (see Fig. 3, 4), resembling those seen after disruption of Wnt signaling by double knockdown of lef1 and tcf7 [57]. Taken together, these observations strongly suggest that our morpholinos were effective, and thus that the lack melanocyte and NCC induction phenotypes provides strong evidence indicating that these fzd genes are unlikely to play a major role in these processes.
Given the robust evidence for Wnt signaling involvement in NC induction and melanocyte specification, this forces the question of whether other fzd genes are involved in these developmental aspects. Mammals have 10 fzd genes and since an extra wholegenome duplication event occurred in the teleost lineage compared with that of mammals, it is conceivable that zebrafish may have 20 (or more) fzd genes in total. Our comprehensive BLAST search used the previous (Zv8) version of the zebrafish genome database; repeating our search using the latest version (Zv9) did not find any new fzd genes. However, we cannot rule out there being fzd genes still to be identified. In particular, other paralogues may be identified as the genome sequence becomes refined. It is known that duplicated genes may have (partially) divergent expression patterns in zebrafish, so that only one of the duplicated genes retains specific ancestral functions [58]. Any novel fzd genes should be directly assessed for candidacy as receptors mediating NCC induction and melanocyte specification, by expression and functional analysis.
Our assessment criteria relied on specific expression at specific locations at specific times; it is conceivable that we did not detect the crucial fzd genes, for example, due to lower levels of expression in these locations. We did not perform comprehensive assessment of Fzd1, Fzd2, Fzd4, Fzd5, Fzd6 and Fzd8 for roles in melanocyte development or NCC induction. Where the expression patterns of other vertebrate fzd genes correspond to the time and place of NC induction or melanocyte specification, this might indicate which other fzd genes should be assessed more carefully. In mice, for example, Sox10 is reported to be expressed in dorsal neural tube from E8.5 to E10.5 [59] so that fzd genes expressed during these stages will be potential receptors required for melanocyte specification. The data from mouse does not currently provide evidence that any of Fzd1, Fzd2, Fzd4, Fzd5, Fzd6 or Fzd8 show appropriate expression patterns [60], although studies to date have not focused on NC development. In the case of NC induction, fzds expressed in the neuro-ectodermal border around E8 stage might be strong candidates, since this is when expression of Msx and Zic, two early markers of the NC, are seen [61,62,63,64,65]. However, again the mice data do not currently implicate any of the fzd genes which were not knocked down here in NC induction, since the described patterns encompass mesodermal lineages (Fzd1, Fzd6, Fzd8) [60], future ventral neural tube (Fzd1) [60] or telencephalon (Fzd5) [9,60]. Fzd2 alone has been reported to be expressed in the entire neural plate (including NCC contributing to the cardiac outflow tract [66]). We found no sign of fzd2 expression in the zebrafish neural tube at 1-somite nor 18 hpf stage; thus, a duplicated fzd2 gene might be a strong candidate gene for zebrafish melanocyte specification. Our assumption that the key receptors would be specifically expressed in the lateral neural plate or in the early NCCs may also be mistaken; we dismissed fzd genes showing ubiquitous expression from further study. Thus, fzd6, fzd7a and fzd9a might be worthy of further assessment in this context, although it seems likely that if they are crucial they are likely to be at least partially redundant amongst themselves.
Our data at least hint at the possibility that Wnt signals mediating NCC induction and melanocyte specification might depend primarily on receptors belonging to families other than the Frizzled family. As mentioned before, LRP5, 6 also function as Wnt receptors, although they appear to function as co-receptors for Fzd receptors, rather than functioning as an independent receptor [2,67]. Other studies have recently revealed two other Wnt receptors, Derailed/Ryk and Ror, which can transduce Wnt signal in a Fzd-independent manner [2,68]. The Derailed/Ryk family transmembrane proteins are atypical receptor type tyrosine kinases, and have been studied so far in D. melanogaster (Derailed), C. elegans  and vertebrates (Ryk). Phenotypic analysis in Drosophila and C. elegans [69,70] and biochemical analyses [70,71] provided evidence of genetic and physical interaction between Wnt and Ryk protein in a Fzd-independent manner. Functional analyses have revealed that Derailed/Ryk were involved in axon guidance in Drosophila [70] and mouse [71], in vulval formation in C. elegans [69], and in CE movement in Xenopus and zebrafish embryo [72,73]. However, the evidence published to date suggests no compelling argument for the requirement of Ryk protein in the development of NCCs and their derivatives. Embryonic expression patterns of ryk genes in Xenopus and zebrafish show their strong enrichment in somite and central nervous system; expression is not The amount of morpholino in each case were as follows. Control-MO, 9 ng; fzd7a, fzd7b-MOs, 2 ng each; fzd7a, fzd7b, fzd10-MOs, 2, 2, 1 ng, respectively; fzd7a, fzd7b, fzd10-MOs, 2, 2, 2 ng, respectively; fzd7a, fzd7b, fzd10-MOs, 2, 2, 5 ng, respectively. P values estimated by Tukey's multiple comparison test are indicated for each pair. Error bars are s.e.m. for each group. doi:10.1371/journal.pone.0054833.g007 enriched in dorsal neural tube at the times examined in this study, although ryk seems to be weakly and ubiquitously expressed in zebrafish at 19 hpf [73,74]. Another Wnt receptor, Ror, is also a transmembrane tyrosine kinase, which contains Fzd-like cysteine-rich domain in the extracellular domain. Two structurally related proteins, Ror1 and Ror2, have been identified so far, and recent analyses have demonstrated evidence of involvement of Ror family kinases in Wnt signaling pathways, such as the physical interaction of Ror2 protein with Wnt5a [75], consistent with the similar phenotypes of Ror2 and Wnt5a deficient mice [76], and genetic interaction between cam-1 (C.elegans homolog of Ror ) and Wnt family gene egl-20 [77]. Functional analysis further revealed that Ror mediated the changes in cell polarity and cell migration in CE movement that result from Wnt5a-and Wnt11-triggered non-canonical Wnt signaling [76,78]. However, again, these studies provide no evidence for the involvement of Ror in NCC development. Rather, studies using cultured cells demonstrate Wnt5a-Ror2 signaling actually inhibits the canonical Wnt pathway, whilst Wnt5a can also activate the canonical pathway when the appropriate receptor (Fzd4) is available [75]. Taken together, we conclude that it is unlikely that zebrafish homologs of Ror genes mediate Wnt signaling in NC induction or melanocyte specification, although their expression patterns have not been reported yet. In zebrafish, MuSK (muscle-specific receptor kinase), a related protein to Ror [79], is mutated in unplugged mutants [80]; whilst it plays a role in controlling segmental migration of trunk NCCs, there is no report of defective NC induction or melanocyte specification in these mutants [81,82].
In summary, knockdown of the subset of fzd genes on which we focused here is apparently insufficient to effect the reduction of melanocytes and NCCs, although the morpholinos themselves seem to be effective. We propose that further exploration, focused on ubiquitously expressed fzds or fzd gene expression at low levels, is now required to identify the crucial fzd receptor(s) mediating melanocyte development and NCC induction. Completion of the zebrafish genome sequencing project will clarify whether further fzd genes remain still to be considered in this context.

Other roles of fzd genes
In addition to illuminating the role of fzd genes in CE movements, NC induction and melanocyte specification, our comprehensive analysis of fzd gene expression during somitogenesis and prim stages suggests further roles for specific genes. Here, we highlight specific aspects of the expression patterns of some fzds. fzd4 shows very localised expression in the region of the otic and lateral line placodes (Fig. 1Da, Da', Da''). Cranial sensory placodes are, like NCCs, formed at the boundary of the neural plate and prospective epidermis. Development of the placodes is thought to start with induction of the preplacodal region, a ground state for all cranial sensory placodes. Subsequently, the preplacodal domain becomes subdivided into individual placodes, such as olfactory, lens, otic and lateral line placodes, by specific inducing signals [83]. Several lines of evidence suggested the involvement of Wnt signaling in this subdivision process to specify otic placode [84].
In contrast to otic development, the role of Wnts in the initial development of the posterior lateral line placode has been little studied in zebrafish, whilst neuromast formation in later stages has been intensively studied [85]. The fzd4 expression pattern demonstrated here suggests the involvement of Wnt signaling in the initial formation of the lateral line placode. Intriguingly, fzd4 seems to be expressed in progenitors for both the future otic and lateral line placodes. It is not clear whether otic and lateral line placodes are each induced from a shared set of progenitor cells in the preplacodal region, but this is one possible interpretation of the fzd4 expression patterns we have observed at 1-somite stage, well before the emergence of morphologically-distinct otic and lateral line placodes. Knockdown studies of fzd4 will be of interest for a test of this hypothesis.
Other interesting expression domains revealed here are those of fzd5 and fzd8a in the ventral ( = subpalial) telencephalon. The nascent telencephalon is known first to be divided into dorsal and ventral parts, which then form the palium and subpalium, respectively [86]. This dorso-ventral patterning is followed by further development of different types of neurons along that axis [87]. Wnt/BMP (bone morphogenetic protein) and Hedgehog/ FGF (fibroblast growth factor) signals are thought to play major roles in patterning of the dorsal and ventral regions respectively [86]. Our data showing expression of fzd3b and fzd4 is consistent with this notion. On the other hand, fzd5 (Fig. 1Ec) and fzd8a (Fig. 1Ic) are both expressed in the ventral part of the telencephalon. These observations clearly imply a role for Wnt signaling in the development of ventral telencephalon, too. Indeed, conditional knockout ofß-Catenin causes severe impairment of growth of the medial ganglionic eminence in the subpalium [88], so that this is at least one candidate role for fzd5 and fzd8a in this region.

Identification of new fzds required for CE movement
Here, we have demonstrated the co-operative requirement of fzd7a and fzd7b in CE movement. Since the important role of fzd7 in CE movement has already been reported in Xenopus and mice [89,90] and it is reasonable that duplicated genes share the function in a certain developmental event, this result reveals strong conservation of fzd7 function in CE movements. In addition, we here identify fzd10 as another Wnt receptor contributing to epiboly movement and mesodermal differentiation, working together with fzd7a and fzd7b. Although a recent publication reported the redundant function of fzd2 and fzd7 in CE movement in mice [90], co-operation of fzd10 with fzd7s during gastrulation has not been reported so far. Furthermore, previous reports on zebrafish fzd10 suggest this receptor activates the canonical Wnt pathway during early development [91], rather than the noncanonical Wnt pathway usually associated with cell movement. Therefore, co-operation of fzd7a and fzd7b with fzd10 in regulating epiboly movements was unexpected. Nevertheless, analysis in human cancer cells demonstrates the promotion of cell motility through activation of the non-canonical pathway by FZD10 [92]. These observations support the idea that fzd10 can transduce Wnt signaling through both canonical and non-canonical pathway dependent on the developmental context. As noted previously, switching between canonical and non-canonical Wnt signaling is, at least in certain cases, dependent upon the type of (co)receptor available [75]. Characterizing the co-receptors involved in these two examples may help with understanding the molecular nature of dual properties of fzd10 in signal transduction.
In summary, our data define the roles of specific fzd genes in CE movements during gastrulation, and limit the candidates for Wnt receptors mediating NC induction and melanocyte specification. Furthermore, our expression data provides the basis for a targeted dissection of the role of Fzds in the embryonic development of numerous structures, some of which we have highlighted here. Figure S1 Probe regions in fzd cDNAs used in this study. Solid black lines represent the entire cDNA reported (For accession numbers for these fzds, see Table 1). Figures on the right ends of each cDNA are total base pair numbers. White bars represent the cloned location to make probes. ATG and stop codon sites are shown here as predicted in registered information. Figures under each black line are base pair number from the first of the cDNAs. (TIF) Figure S2 fzd7b expression domain overlaps future NCC marked by foxd3 domain, whilst fzd10 expression domain does not. Here, eya1 is used as placodal marker, and foxd3 is used as NCC marker. Aa-Ac) Staining with fzd7b (purple) and eya1 (red). Ba-Bc) fzd10 (purple) and eya1 (red). Ca-Cc) fzd7b (purple) and foxd3 (red). Da-Dc) fzd10 (purple) and foxd3 (red). All images show dorsal views of right half of flat-mounted 1 -somite stage embryos. Anterior to the top. a, b, c for each data set represent bright field, dark field, and merged images, respectively. Note that there is a gap between fzd10 expression domain and eya1 expression domain (B), whilst fzd7b and eya1 expression domains contact each other directly (A), suggesting fzd7b expression domain expands more laterally than fzd10. In C, fzd7b and foxd3 are coexpressed (arrowhead). In contrast, fzd10 and foxd3 expression domains abut, but do not overlap (D). Scale bar: 100 mm. (TIF) Figure S3 Testing efficacy of fzd gene morpholinos. A) 10 ng of fzd3a-MO can substantially suppress the normal splicing of fzd3a mRNA. RT-PCR analysis was carried out as previously reported (See Materials and Methods). We injected 10 ng of morpholinos into wild-type embryos, and extracted total RNA at 24 hpf. Whilst control random oligo-injected (lane 2) and uninjected (lane 3) samples showed a clear fzd3a transcript band of the expected size (465 bp, arrowhead)., this band was barely detectable after 10 ng of fzd3a-MO injection due to the inhibition of correct splicing (lane 1). Primers against b-actin cDNA were used as a positive control for RNA extraction and RT-PCR (lanes 4, 5, 6.). B) Schematic drawing of fusion constructs to test ability of fzd9b and fzd10 morpholinos to bind to the target sites and suppress translation in vivo. Target sites for fzd9band fzd10-MO (red) are fused to cDNA of egfp (green). Actual sequences of target site and the first 3 bps of egfp sequence are shown below the scheme. Original ATG for egfp gene is modified into ATC, which is next to GTG in green, in order to avoid translation from this. C-F, C'-F') Our fzd9band fzd10-MOs can efficiently suppress GFP fluorescence derived from injected fusion constructs. In all cases, 100 pg of mRNA of the respective fusion construct was injected, and embryos were observed at 10 hpf stage. GFP fluorescence was clearly detected in embryos injected with 5 ng of control random morpholinos (C, C', E, E'). In contrast, injection of even just 1 ng of the respective experimental morpholinos suppressed fluorescence completely (D, D', F, F'). C-F are bright field images of C'-F', respectively. Numbers in C'-F' are ratios of GFP-positive embryos out of surviving injected embryos. Scale bar: 500 mm.  Table S2 for quantitation. Scale bar: 100 mm.