|Home | About | Journals | Submit | Contact Us | Français|
The neural crest is a stem cell-like population exclusive to vertebrates that gives rise to many different cell types including chondrocytes, neurons and melanocytes. Arising from the neural plate border at the intersection of Wnt and Bmp signaling pathways, the complexity of neural crest gene regulatory networks has made the earliest steps of induction difficult to elucidate. Here, we report that tfap2a and foxd3 participate in neural crest induction and are necessary and sufficient for this process to proceed. Double mutant tfap2a (mont blanc, mob) and foxd3 (mother superior, mos) mob;mos zebrafish embryos completely lack all neural crest-derived tissues. Moreover, tfap2a and foxd3 are expressed during gastrulation prior to neural crest induction in distinct, complementary, domains; tfap2a is expressed in the ventral non-neural ectoderm and foxd3 in the dorsal mesendoderm and ectoderm. We further show that Bmp signaling is expanded in mob;mos embryos while expression of dkk1, a Wnt signaling inhibitor, is increased and canonical Wnt targets are suppressed. These changes in Bmp and Wnt signaling result in specific perturbations of neural crest induction rather than general defects in neural plate border or dorso-ventral patterning. foxd3 overexpression, on the other hand, enhances the ability of tfap2a to ectopically induce neural crest around the neural plate, overriding the normal neural plate border limit of the early neural crest territory. Although loss of either tfap2 or foxd2 alters Bmp and Wnt signaling patterns, only their combined inactivation sufficiently alters these signaling gradients to abort neural crest induction. Collectively, our results indicate that tfap2a and foxd3, in addition to their respective roles in the differentiation of neural crest derivatives, also jointly maintain the balance of Bmp and Wnt signaling in order to delineate the neural crest induction domain.
The neural crest is a migratory cell population unique to vertebrates that played a fundamental role in vertebrate evolution (Gans and Northcutt, 1983). The neural crest arises at the border between the neural plate and non-neural ectoderm at the end of gastrulation. Later in development, neural crest cells delaminate from the dorsal part of the neural tube and start migrating throughout the body to their final destinations where they differentiate to many different cell lineages, including craniofacial cartilage and bone, peripheral and enteric neurons, and smooth muscle and pigment cells (Le Douarin et al., 2004).
Neural crest induction takes place early during gastrulation when signals from surrounding tissues start to define the neural plate border. Bone morphogenetic protein (Bmp), Wnt, fibroblast growth factor (Fgf), and Notch/Delta signaling pathways have been implicated in neural crest induction in various species (Huang and Saint-Jeannet, 2004). Among these pathways, Bmp and Wnt signaling play critical roles in neural crest induction (Aybar and Mayor, 2002; Nieto, 2001; Raible and Ragland, 2005; Wu et al., 2003).
Bmp signaling is established in a concentration gradient with high levels in the ventral ectoderm and progressively lower levels towards the dorsal midline, where the neural plate is formed. This distribution is shaped by the expression pattern of Bmp factors in ectoderm and Bmp antagonists such as chordin in dorsal mesoderm. These Bmp antagonists are also involved in neural plate induction (Barth et al., 1999). It has been proposed that intermediate levels of Bmp signaling at the neural plate border are necessary for neural crest induction (LaBonne and Bronner-Fraser, 1998). In Xenopus, overexpression of Bmp4 antagonists enlarges the neural crest domain, whereas Bmp overexpression causes neural crest reduction (LaBonne and Bronner-Fraser, 1998; Liem et al., 1995). Similarly in zebrafish, early neural crest is very sensitive to Bmp signaling levels. For example, bmp2b/swirl mutants exhibit a loss of Bmp signaling and a decrease in neural crest progenitors, while bmp7/snailhouse or smad5/somitabun mutants have moderate or low Bmp activity and expanded neural crest domain (Le Douarin and Kalcheim, 1999; Nguyen et al., 1998; Schmid et al., 2000). These data suggest that Bmp signaling controls the position and size of the neural crest induction domain (Taneyhill and Bronner-Fraser, 2005).
The involvement of Wnt signaling in neural crest induction has also been well documented. Inhibition of canonical Wnt/β-catenin signaling pathway prevents neural crest formation in mouse, chick, xenopus, and zebrafish (Brault et al., 2001; Garcia-Castro et al., 2002; Hong et al., 2008; Lewis et al., 2004; Vallin et al., 2001; Yanfeng et al., 2003). In Xenopus, ectopic expression of Wnt family members increases the number of neural crest cell progenitors. Moreover, addition of soluble Wnts to the intermediate neural plate promotes de novo neural crest induction in the chick, suggesting that Wnt signaling is sufficient for neural crest induction (Garcia-Castro et al., 2002). Wnt signaling is modulated by a number of antagonists, including Dkk1 and Kremen, which are also required for neural crest induction (Carmona-Fontaine et al., 2007; Hassler et al., 2007; He, 2003). Interestingly, several studies have shown that repressing Bmp and Wnt signaling is also required for neural crest induction (LaBonne and Bronner-Fraser, 1998; Patthey et al., 2008; Steventon et al., 2009), suggesting that fine tuning of Bmp and Wnt signaling is necessary for proper neural crest formation. However, the mechanisms underlying these processes remain unclear.
The transcription factor activator protein 2 alpha (Tfap2a) plays crucial roles in development (Hilger-Eversheim et al., 2000) and melanoma progression (Bar-Eli, 2001). Tfap2a knockout mice exhibit defective neural tube closure and loss of the craniofacial skeleton and peripheral nervous system, tissues that are derived from cranial neural crest (Schorle et al., 1996; Zhang et al., 1996). In humans, mutations in TFAP2A result in the branchio-oculo-facial syndrome, which is characterized by severe craniofacial malformations (Milunsky et al., 2008). In zebrafish, loss of tfap2a results in apoptosis of neural crest progenitors, leading to defects in neural crest derivatives, including posterior pharyngeal arches, pigment cells, enteric neurons, and dorsal root ganglia (Barrallo-Gimeno et al., 2004; Knight et al., 2003; O'Brien et al., 2004). A role for tfap2a in neural crest induction has also been proposed in Xenopus (de Crozé et al., 2011; Luo et al., 2003), and in Lamprey, ap2a was shown to act in the induction of the neural plate border (Nikitina et al., 2008).
Foxd3, a member of the Forkhead transcription factor family, has multiple functions during embryonic development. Mouse embryos lacking Foxd3 die at peri-implantation stages due to failure in maintaining self-renewing embryonic stem cells (Hanna et al., 2002; Nelms et al., 2011). Tissue-specific deletion of Foxd3 in the neural crest shows that it is required for neural crest maintenance in mice (Teng et al., 2008). Foxd3 has also been linked to neural crest development in Xenopus and neural crest migration in chick embryos (Kos et al., 2001; Pohl and Knochel, 2001). In addition, a variant in the human FOXD3 promoter sequence results in autosomal dominant vitiligo, a pigmentation disorder caused by abnormalities in the melanoblast lineage (Alkhateeb et al., 2005). Finally, foxd3 mutations in zebrafish lead to reduction of neural crest derivatives, resulting in defects in posterior craniofacial skeleton, peripheral nervous system, and pigmentation (Montero-Balaguer et al., 2006; Stewart et al., 2006).
Consistent with the well-established role of Foxd3 and Tfap2a in neural crest differentiation, a double mutant of tfap2alow;foxd3sym1 was shown to prevent the specification of developmentally distinct neural crest sub-lineages, yet in this combination of mutant alleles, there were no neural crest induction defects observed (Arduini et al., 2009). Interestingly, the defects caused by the foxd3 mutations show striking similarities to those of tfap2a mutants, suggesting that the two genes act in the same or parallel pathways that are critical for neural crest development.
Here, we report that tfap2amob and foxd3mos act within the same genetic pathway, and their combined loss results in an almost complete absence of neural crest-derived tissues. Conversely, foxd3 overexpression enhances the ability of the tfap2a to ectopically induce neural crest around the neural plate, overriding the normal neural plate border limit of early neural crest territory. We provide evidence that the effects of tfap2 and foxd3 on neural crest are due to modulation of the Bmp and Wnt signaling pathways, suggesting that the two transcription factors jointly maintain the balance of Bmp and Wnt signaling in order to delineate the neural crest induction domain. Our findings indicate that tfap2a and foxd3 are not only necessary for neural crest specification, but also play an important role in the earliest steps of the neural crest precursor cell development, delineating a powerful paradigm for understanding the genesis of this complex stem cell population.
We have previously shown that zebrafish mutations in transcription factors tfap2a (mobm610) and foxd3 (mosm188) display similar phenotypes during embryonic development with a reduction of neural crest derivatives such as pigment cells, craniofacial chondrocytes and peripheral neurons (Barrallo-Gimeno et al., 2004; Montero-Balaguer et al., 2006). To investigate the potential genetic interactions between tfap2a and foxd3 in neural crest development, we generated mob;mos double mutant embryos. We found that mob;mos embryos exhibit more severe phenotypes than single mutant embryos (Figure 1). For example, while mob and mos mutant embryos have reduced pigment cells at 5 dpf, the most prominent feature of mob;mos embryos is the almost complete lack of pigmentation, already evident at 36 hpf (Figure 1A–D). In some embryos, isolated melanocytes are present over the hindbrain region at 4–5 dpf. Retina pigmentation is normal, arguing against a general defect in melanin synthesis.
mob;mos embryos also lack lower jaw structures. In mob or mos single mutant embryos, the Meckel´s cartilage is pointed ventrally leading to prominently protruding and gaping jaws. In mob;mos embryos, however, the ventral part of the head is concave and devoid of pharyngeal skeleton (Figure 1D). In order to better assess the structure of the head skeleton, we stained 5 dpf embryos with alcian blue (Figure 1E–H). In agreement with earlier studies (Barrallo-Gimeno et al., 2004; Montero-Balaguer et al., 2006), mob and mos embryos display well-shaped cartilages of the first pharyngeal arch, Meckel's (m) and palatoquadrate (pq), whereas the posterior pharyngeal arches are severely reduced (Figure 1F, G). In mob;mos embryos, the cartilage elements of the entire pharyngeal skeleton, and the trabeculae (tr) and ethmoid plate (ep) of the neurocranium are absent. The mesodermally derived neurocranium and pectoral fin structures are unaltered, further indicating that the mob;mos mutations specifically affect tissues of neural crest origin (Figure 1H). We observed similar defects by injecting morpholino (MO) against foxd3 in mob embryos or against both tfap2a and foxd3 in wild-type embryos, supporting the idea that the observed phenotypes are due to the combined loss of Tfap2a and Foxd3 function (Supplementary Fig. S1 and data not shown).
To examine the deficits of neural crest-derived peripheral neurons in mob;mos embryos, we examined dorsal root ganglia (DRG) and enteric neurons using an antibody against the pan-neuronal protein Hu. In wild-type embryos at 3 dpf, enteric neurons are present in the gut and trunk DRG sensory neurons are bilaterally distributed along the anteroposterior axis and positioned at the level of the ventral spinal cord with one pair of DRG in each somitic segment (Figure 1I) (Kelsh et al., 2000). Single mob and mos mutant embryos have few scattered neurons, as previously described (Barrallo-Gimeno et al., 2004; Montero-Balaguer et al., 2006). In contrast, all peripheral neurons are absent in mob;mos embryos (Figure 1J–L). However, Rohon-Beard neurons are present in mob;mos embryos, suggesting that placode derivatives are not affected (Figure 1L).
The consistent lack of all neural crest-derived tissues in mob;mos embryos indicated that Foxd3 and Tfpa2a might be required at migratory or premigratory stages of neural crest development. To address this question, we analyzed the expression of genes expressed in migratory neural crest such as crestin, dlx2a, and sox10 at 24 hpf (Figure 1M–R). We found no expression of the pan-neural crest marker crestin in mob;mos embryos (Figure 1M,N). dlx2a expression in migrating neural crest streams in the head region is absent, although appears normal in the telencephalon and diencephalon (Figure 1O,P). sox10, which is expressed in non-ectomesemchymal neural crest precursors including pigment and peripheral nervous system (Dutton et al., 2001) is absent in mob;mos embryos, although it is normally expressed in the otic vesicle (Figure 1Q,R).
We next asked whether the lack of migratory neural crest cells in mob;mos embryos was due to impaired formation of premigratory neural crest progenitors in the lateral neural plate by examining the expression of the transcription factor snail1b that labels premigratory neural crest from 11 hpf. Consistent with our previous studies (Barrallo-Gimeno et al., 2004; Montero-Balaguer et al., 2006), expression of snail1b is slightly reduced in mob or mos embryos as compared to wild types (Figure 1S–U). However, snail1b expression is nearly absent in mob;mos embryos (Figure 1V). Similarly, expression of transcription factors sox9b and sox10 is reduced in the premigratory neural crest territory in mob;mos embryos (Supplementary Fig. S2).
Taken together, our results indicate that the combined tfap2a and foxd3 loss of function mutations leads to more severe defects than either mutation alone, suggesting that the two regulatory factors genetically interact. Moreover, they are required for the development of neural crest derived cell populations at a stage preceding neural crest specification.
We and others have shown that loss of Tfap2a leads to cell death during neural crest migration in the head region (Barrallo-Gimeno et al., 2004; Knight et al., 2003). To test whether the mob;mos phenotype is a consequence of neural crest cell death at premigratory stages, we used acridine orange staining to detect apoptotic cells at the end of gastrulation. At 10 hpf, mob and mos embryos have more dying cells in the neural plate territory than wild types (Supplementary Fig. S3). However, mob;mos embryos exhibit a dramatic increase of cell death in this region (Figure 2A–B). Furthermore, acridine orange staining showed that dying cells are already present in mob;mos embryos at 8 hpf, while wild-type, mob, and mos embryos have very few dying cells at this stage, with their number rising in all three mutant phenotypes at 10 hpf (Supplementary Fig. S3).
We further confirmed and quantified these results using the TUNEL (terminal transferase-mediated dUTP nick end-labeling) assay at 10 hpf. Consistent with the acridine orange results, we found very few dying cells in wild-type embryos, while the number of apoptotic cells is increased by approximately 4- and 6-fold in the neural plate region in embryos lacking tfap2a or foxd3 respectively (7.8 ± 5.1 in wild-type embryos; 34.0 ± 14.1 in mob embryos; 48.6 ± 11.5 in foxd3-knockdown embryos, Figure 2C–E). In double mutants, apoptotic cell death is increased by ~17-fold compared to wild types (247 ± 44.8 in foxd3-knockdown mob embryos; figure 2D,E). Histological sections of the TUNEL-stained embryos revealed that the majority of dying cell are localized to the surface of the embryo and only sporadically found inside the neural plate (Supplementary Fig. S3).
In order to assess whether cell death is responsible for the observed lack of neural crest derivatives, we injected mob;mos embryos with a morpholino to block p53 translation, which mediates apoptosis during early embryogenesis in zebrafish (Plaster et al., 2006). Using TUNEL assay, we found that apoptosis is suppressed in mob;mos embryos injected with the p53-MO at 10 hpf (Supplementary Fig. S4). We then analyzed the formation of pharyngeal arch cartilage and the development of peripheral nervous system in p53-knock-down mob;mos embryos. The results show that prevention of cell death does not rescue the head cartilage and peripheral nervous system defects (Figure 2F–J). We conclude that the p53-dependent cell death is not the cause of neural crest cell loss in mob;mos embryos, but rather is the consequence of an earlier developmental abnormality. This is consistent with the histological analysis of TUNEL-stained embryos, which showed that cells undergoing apoptosis are rarely present in the neural crest domain.
The results described above suggest that Tfap2a and Foxd3 are required before 10 hpf, possibly during neural crest induction, which in zebrafish takes place at mid-gastrulation between 60% and 75% epiboly, approximately 7–8 hpf (Lewis et al., 2004). Although the role of tfap2a and foxd3 in neural crest progenitors is well established (Barrallo-Gimeno et al., 2004; Knight et al., 2003; Montero-Balaguer et al., 2006; Odenthal and Nusslein-Volhard, 1998; Stewart et al., 2006), their function during neural crest induction has not been investigated in zebrafish.
To determine whether foxd3 and tfap2a are expressed during these stages of neural crest induction, we examined their expression patterns in early development by in situ hybridization. We found that tfap2a first appears in the non-neural ectoderm from the shield (6 hpf) to bud stages (10 hpf) (Figure 3A–C). In contrast, foxd3 is maternally deposited (data not shown). At the shield stage, foxd3 is expressed in the organizer and the adjacent bilateral margins (figure 3D,G). As development progresses, the foxd3 expression domain in the mesoderm extends along the anteroposterior axis, while its expression in the lateral margin remains until the end of gastrulation (Figure 3D–I). At 10 hpf, foxd3 expression is almost absent in the axis, but is first detected in neural crest progenitors at the neural plate border (Montero-Balaguer et al., 2006). This analysis and double staining experiments with foxd3 and tfap2a riboprobes show that both tfap2a and foxd3 are expressed during the early stages of neural crest induction in distinct, but complementary domains throughout gastrulation (Figure 3J–L).
The non-overlapping expression domains of tfap2a and foxd3 during gastrulation suggest that their cooperative induction of neural crest does not involve physical interaction and subsequent regulation of a common set of target genes. Instead, it appears likely that two transcription factors regulate pathways in different parts of the embryo, i.e., the ventral non-neural ectoderm by Tfap2a and the dorsal mesendoderm and ectoderm by Foxd3 (Figure 3L). Previous studies have shown that the balance of Bmp and Wnt signaling is critical for the regulation of neural crest induction (Jin et al., 2001; Kleber et al., 2005; Raible and Ragland, 2005; Sakai et al., 2005). Similarly, tfap2a and members of the Bmp family are expressed in the ventral non-neural ectoderm during gastrulation, whereas foxd3 and members of the wnt family are expressed in the margin and dorsal territory. Therefore, we hypothesized that Tfap2a and Foxd3 may alter the balance of Bmp and Wnt signaling, thus influencing neural crest cell induction.
To test this hypothesis, we examined bmp2b and bmp4 expression in wild-type and mutant embryos. Compared to wild-type embryos at 8 hpf, the expression domains of bmp2b and bmp4 in the non-neural ectoderm are expanded in mob and mos mutants, and further increase in mob;mos embryos (Figure 4A–H, Supplementary Fig. S5). Because Bmp activity from the ventral non-neural ectoderm during gastrulation is limited in the neural plate by a number of Bmp antagonists, including Chordin, Noggin and Follistatin (Blader et al., 1997; Fainsod et al., 1997; Warren et al., 2003), we examined chordin expression at 8 hpf (Figure 4I–P). In wild-type embryos, chordin is expressed in the organizer and the dorsal margin, similar to foxd3 expression. However, in individual mob and mos mutants, the chordin expression domain is reduced in the lateral margin and further diminishes in mob;mos embryos.
To test whether changes in bmp2b, bmp4 and chordin expression domains lead to abnormal Bmp signaling, we examined the levels of phosphorylated Smad1/5/8 proteins, which act downstream of activated Bmp receptors (Attisano and Lee-Hoeflich, 2001; von Bubnoff and Cho, 2001). To this end, we stained wild-type and mutant embryos at 8 hpf with an antibody recognizing phospho-Smad1/5/8. The immunostaining results show that the domain of active Bmp signaling is expanded in mob and mos embryos and further increases in mob;mos embryos as compared to wild types (Figure Q–X and Supplementary Fig. S5).
Taken together, our results show that Tfap2a and Foxd3 together regulate the extent of Bmp signaling during late gastrulation. Interestingly, although Bmp signaling is known to influence the general dorso-ventral patterning of the embryo, we observed no defects in mesoderm induction in mob;mos embryos as evaluated by fgf8, pea3, no tail and papc expression at the end of gastrulation in axial and paraxial mesoderm (Supplementary Fig. S6 and data not shown) (Tucker et al., 2008). Similarly, there are no defects in neural plate border induction (evaluated by pax3a expression, data not shown). Moreover, we observed reduced foxd3 expression at the margin, adjacent to where neural crest is induced, but not in the organizer or the axial mesendoderm, which would argue against a mesoderm defect in mos embryos and subsequently in mob;mos double mutant embryos (Supplementary Fig. S7). These results are consistent with the lack of defects in mesodermal derivatives and Rohon-Beard neurons in mob;mos embryos later in development (Figure 1L and data not shown).
To test whether Wnt signaling is also affected by the loss of Tfap2a and Foxd3, we examined the expression of several wnt family members at 8 hpf. Expression of wnt1, wnt3a, wnt5a, and wnt8 was indistinguishable among wild-type and mutant embryos (data not shown). We then tested the expression of Dkk1, a secreted negative regulator of Wnt signaling that interacts with the co-receptor low-density lipoprotein receptor-related protein (Lrp)-5/6 to hinder the interaction between Wnts and their Frizzled receptors, thus blocking canonical Wnt signaling (Mao et al., 2001; Zorn, 2001). Dkk1 has been shown to inhibit neural crest formation at the anterior neural fold (Carmona-Fontaine et al., 2007). Our results show that at 8 hpf dkk1 is expressed in the axial mesendoderm of wild-type embryos (Figure 5A). In mob and mos embryos, dkk1 expression does not change significantly, however its expression is dramatically expanded in mob;mos embryos (Figure 5B–D).
The elevated levels of dkk1 suggest that canonical Wnt signaling in axial mesendoderm is suppressed in mob;mos embryos. To confirm this possibility, we analyzed the expression of two direct targets of canonical Wnt signaling, axin2 and sp5l (Jho et al., 2002; Weidinger et al., 2005) by in situ hybridization and quantitative RT-PCR. In situ hybridization results showed that axin2 and sp5l are expressed at the lateral margin in wild-type embryos at 8 hpf (Supplementary Fig. S8). The expression levels of axin2 and sp5l are not significantly altered in mob or mos embryos (data not shown), but are substantially reduced in tfap2a and foxd3 double knockdown morphants. Quantitative analysis in 8-hpf embryos corroborated this result (Figure 5E).
To test whether attenuation of the high dkk1 expression levels in mob;mos suppresses the neural crest phenotype in double mutants, we injected embryos with three different doses (1, 2, and 4 ng) of dkk1-translation blocking MOs. We found that 2 and 4 ng of dkk1-MO cause smaller heads and bent bodies, consistent with the previously described dkk1 mutant phenotypes, indicating that the MOs interfered with Dkk1 function (data not shown). In contrast, 1 ng of dkk1-MO did not affect the development of pharyngeal arch cartilage or pigmentation in wild-type embryos, allowing us to test whether downregulation of excess dkk1 in double mutants restores neural crest derivatives. Therefore, we injected mob;mos embryos with 1 ng of dkk1-MO. We found that about 40% of dkk1-MO knockdown mob;mos morphants have partially rescued pharyngeal arch cartilages and trabeculae compared to mob;mos embryos, as revealed by alcian blue staining at 5 dpf (Figures 5F–J). Similarly, significant rescue was observed in the melanophore lineage, ranging from partial to almost complete restoration of pigmentation (Supplementary Fig. S9). This finding indicates that the increase in dkk1 and the corresponding decrease in canonical wnt signaling is partially responsible for the defects in mob;mos mutants.
Our results so far support the idea that loss of Foxd3 and Tfap2a function disrupts the normal patterns of Bmp and Wnt signaling during late gastrulation, suggesting that the two factors are necessary for neural crest induction. To test whether the functional interaction between Tfap2a and Foxd3 is sufficient to induce neural crest, we overexpressed tfap2a, foxd3, and tfap2a together with foxd3 by mRNA injection in wild-type embryos and examined the formation of neural crest progenitors by sox10 expression at 12 hpf. Embryos injected with tfap2a mRNA exhibit increased sox10 expression in the lateral neural plate compared to wild-type embryos (Figure 6A–B), indicating that tfap2a alone is sufficient to increase neural crest formation. foxd3 overexpression, however, decreased sox10 expression (Figure 6C), suggesting that excessive foxd3 has a negative effect on neural crest formation. Strikingly, when tfap2a is overexpressed together with foxd3, sox10 expression is dramatically increased, not only at the lateral neural plate, but also in the most medial part of it, indicating that the neural crest competent territory has significantly expanded (Figure 6D). To address the overall patterning of the injected embryos we tested the expression of mesodermal markers fgf8 and pea3 at 8 and 10 hpf, respectively, and found them largely unaltered (Figure 6E,F). The expression of axial markers shh and hgg1 was also largely normal, as were the neural plate markers egr2b, six3b and dlx3b (Figure 6G,H). Embryos injected with both foxd3 and tfap2a mRNAs were consistently smaller than uninjected controls.
Taken together, the gain-of-function data further support the idea that, in addition to their previously defined roles in maintaining neural crest, Tfap2a and Foxd3 cooperate in neural crest induction.
Biological processes are complex events that are controlled by gene regulatory networks composed of multiple transcription factors that work together to induce the appropriate target genes at the proper stage and territory (Li and Davidson, 2009). Here, we describe the cooperative role of transcription factors Tfap2a and Foxd3 on neural crest development since simultaneous loss of both factors leads to more severe defects than loss of either factor alone. While single tfpa2a or foxd3 mutations negatively affect neural crest survival or maintenance at the onset of their migration, neural crest specification appears normal. As a result, neural crest-derived cell types are severely depleted, but still present in reduced numbers (Barrallo-Gimeno et al., 2004; Knight et al., 2003; Montero-Balaguer et al., 2006; O'Brien et al., 2004; Stewart et al., 2006). Conversely, combined loss of both transcription factors leads to complete loss of neural crest, best exemplified by the absence of all skin pigmentation. This defect can be found not only at migratory stages, but also before neural crest cells leave the neural tube. In this respect, our results are consistent with those recently obtained by Arduini et al. using different zebrafish mutations of tfap2a/low and foxd3/sym1 (Arduini et al., 2009). Tracing back the origin of the double mutant phenotype, we have uncovered novel information about the role of these two factors in maintaining proper patterns of Wnt and Bmp signaling during the initial stages of neural crest development.
Our results show that tfap2a and foxd3 are expressed in zebrafish embryos during gastrulation in distinct, complementary domains: tfap2a in the ventral non-neural ectoderm and foxd3 in the dorsal mesendoderm and ectoderm. The tfap2a expression territory overlaps with bmp2b and bmp4 expression, whereas the foxd3 expression domain coincides with that of chordin, a Bmp antagonist. It appears that Tfap2a and Foxd3 directly or indirectly regulate these genes since expression of Bmp ligands is enlarged in double mob;mos mutant embryos, while chordin expression is significantly reduced. Together, these changes result in expanded Bmp signaling territories towards the dorsal side of the embryo, as marked by phosphorylated Smad1/5/8. At the same time, canonical Wnt activity, as monitored by the expression levels of target genes, is decreased in mob;mos embryos due to higher expression of the Wnt-antagonist dkk1 in axial mesendoderm. Thus, the combined inactivation of tfap2a and foxd3 genes perturbs two of the key signaling pathways involved in neural crest induction. We postulate that this disruption is at least partially responsible for the loss of neural crest induction at the neural plate border as suggested by the lack of snail1b expression at 12 hpf.
Previous studies proposed that Tfap2a and Foxd3 are part of the group of “neural crest specifier genes” acting in the neural crest regulatory network under “neural crest inducer genes” such as msx and pax, which give the neural plate border its identity (Sauka-Spengler and Bronner-Fraser, 2008). However, tfap2a and foxd3 are expressed during gastrulation prior to neural crest induction, suggesting that the two genes have distinct roles prior to neural crest specification. Most importantly, the dramatic changes in Bmp and Wnt signaling pathways observed in mob;mos mutants and the absence of premigratory neural crest indicate that Tfap2a and Foxd3 cooperate in proper neural crest induction during gastrulation at the non-overlapping territory between non-neural ectoderm and prospective neural plate.
The cooperative function of Tfap2a and Foxd3 in neural crest development is further supported from overexpression experiments, which show that foxd3 enhances the ability of the tfap2a to ectopically induce neural crest around the neural plate, overriding the normal neural plate border limit of early neural crest. In contrast, foxd3 overexpression alone reduces the expression of early neural crest markers. These results support earlier studies in Xenopus embryos, which showed that tfap2a overexpression induces premigratory neural crest (Luo et al., 2003), whereas foxd3 overexpression prevents its formation (Pohl and Knochel, 2001), although this later effect remains controversial (Sasai et al., 2001). More recent work in zebrafish by Kwon and colleagues on the induction of preplacodal ectoderm corroborate our findings (Kwon et al., 2010).
Foxd3 has been recently reported to be essential for mesoderm development in zebrafish (Chang and Kessler 2010) and mesoderm is required for neural crest development in some species (Monsoro-Burq et al., 2003; Ragland and Raible, 2004). Therefore, it is possible that the neural crest induction defects observed in mob;mos are secondary to deficits in mesoderm patterning. However, we believe this is unlikely for the following reasons. First, mesoderm formation in mob;mos embryos is normal, as double mutant embryos have regular somites and notochord at 3 dpf. Second, we observed normal no tail expression in axial mesoderm and protocadherin8 (pcdh8, also known as papc) in paraxial mesoderm at the end of gastrulation. Instead, it appears that neural crest formation is more sensitive than mesoderm to foxd3 levels, as foxd3 morpholino doses needed to mimic the neural crest phenotype of mos or sym1 mutations are lower by almost one order of magnitude than those causing a mesodermal defect (Chang and Kessler, 2010). Moreover, since the mos mutation suppresses foxd3 expression in the neural crest at premigratory stages (Montero-Balaguer et al., 2006), we analyzed if this also occurs during gastrulation. In addition, there is reduced foxd3 expression at the lateral margin, adjacent to the area where neural crest is induced, but not in the organizer or the axial mesendoderm, which argues against a mesodermal defect in mos and mob;mos double mutant embryos. However, it is conceivable that although patterning of mesendoderm is not altered in mob;mos, signals emanating from the dorsal mesoderm (including Wnt inhibitors) are changed by downstream effects of tfap2a and foxd3 deficiency, thus resulting in upregulation of dkk1 that contributes to failure of neural crest induction.
A Bmp signaling gradient across the ectoderm has been proposed to be crucial for neural crest development. This gradient is mediated by the secretion of Bmps in the ventral part of the embryo and Bmp antagonists in the dorsal part. High levels of Bmp signaling drive differentiation to epidermis, while low levels allow the formation of the neural plate, and intermediate levels are necessary for neural crest induction. In this light, it is not surprising that the expansion of the Bmp activity territory in mob;mos mutants causes neural crest defects. However, although abnormal Bmp signaling may also affect the overall dorso-ventral patterning of the embryo, mob;mos embryos do not show dorso-ventral phenotypes. One possible explanation is that despite the Bmp signaling domain expansion towards the dorsal side, there is still a source of chordin, which may prevent the complete ventralization of the embryo. Consistent with this idea, overexpression of both bmp2b and bmp4 has little effect in the general dorso-ventral pattering of zebrafish embryos (Schmid et al., 2000).
The expansion of the Bmp signaling domain in mob;mos also does not appear to cause an overall defect in the neural plate border. For example, the dorso-ventral patterning of the neural tube is not affected, as pax3a is normally expressed in the dorsal neural tube of double mutant embryos (data not shown). Moreover, mob;mos embryos show no defects in Rohon-Beard neurons as revealed by anti-HuC immunostaining. Thus, it is likely that neural plate border and neural crest formations are independent events as previously proposed (Basch et al., 2006). Alternatively, neural crest induction may be more sensitive than neuronal development to subtle changes in Bmp and Wnt signaling caused by mutations in tfap2a and foxd3.
It is also possible that the mob;mos phenotype is restricted to neural crest because of concurrent changes in Wnt signaling caused by strong dkk1 upregulation. Identified as an inducer of head structures, Dkk1 is a secreted Wnt antagonist. In zebrafish, dkk1 is expressed in the embryonic shield and later in the anterior axial mesendoderm and prospective prechordal mesoderm during gastrulation (Hashimoto et al., 2000). dkk1 overexpression affects forebrain and axial mesendoderm development, whereas blocking dkk1 expression accelerates cell movements during gastrulation (Caneparo et al., 2007; Hashimoto et al., 2000). Furthermore, Dkk1 is secreted from prechordal mesoderm to block canonical Wnt signaling and inhibit neural crest formation in the anterior neural plate (Carmona-Fontaine et al., 2007). Increased dkk1 expression in the axial mesendoderm of mob;mos embryos, in combination with abnormal Bmp signaling, may be responsible for the lack of neural crest. In support of this notion, morpholino-mediated dkk1 knockdown partially rescues the mob;mos neural crest phenotype.
In 1998, LaBonne and Bronner-Fraser first proposed the “two-signal model”, which suggests that Bmp and Wnt signaling coordinate to regulate neural crest induction (LaBonne and Bronner-Fraser, 1998). In Xenopus embryos and explants, Bmp inhibition by chordin overexpression is not sufficient to induce the neural crest markers snail1 and snail2; however, combined chordin and wnt overexpression significantly promotes neural crest formation. Similar results have also been reported from studies using chick embryos (Patthey et al., 2008). Furthermore, coordination of Bmp and Wnt signaling also plays a role in the maintenance of neural crest (Kleber et al., 2005). More recently, Bmp inhibition and Wnt activation has been shown to be required for neural crest induction (Steventon et al., 2009). Our results corroborate the two-signal model, since neural crest induction is severely disturbed only in tfap2a and foxd3 double mutant embryos, which have an imbalance of both Bmp and Wnt signaling. One possible link between the Bmp and Wnt pathways is that Bmp4 can induce dkk1 expression to inhibit Wnt signaling and, in turn, induce cell apoptosis to sculpt the shape of limbs during mouse embryo development (Grotewold and Ruther, 2002). Therefore, it is possible that Tfap2a and Foxd3 regulate dkk1 expression indirectly via their cooperative modulation of Bmp signaling.
Our results are also significant from an evolutionary point of view, since the neural crest is an exclusive feature of vertebrates. Many of the genes expressed in the neural crest are already present in basal chordates, which are devoid of neural crest, but they are not expressed within the appropriate territories (Barrallo-Gimeno and Nieto, 2006). Consistent with this scenario, in the basal chordate amphioxus (Branchiostoma floridae) amphiAP2 is expressed in the dorsal non-neural ectoderm (Meulemans and Bronner-Fraser, 2002) and amphiFoxD, the only orthologue for the FoxD group in amphioxus, in the mesoderm underlying the neural tube (Yu et al., 2002). These expression patterns are similar to those of tfap2a and foxd3 during zebrafish gastrulation. It has been proposed that in the course of evolution, some of those genes may have been co-opted into the neural plate border in order to allow the formation of the neural crest (Yu et al., 2002). Our results, which indicate a fundamental role for Tfap2a and Foxd3 in the induction of neural crest markers from adjacent territories, argue against the co-option hypothesis. Because the early amphiAP2 and amphiFoxD expression patterns are similar to those of tfap2a and foxd3 in zebrafish, this suggests that the gene regulatory network to induce the expression of neural crest genes in the neural plate border exists in amphioxus. Consistent with this idea, amphiSnail, the amphioxus orthologue of one of the genes considered essential for neural crest migration, is expressed in the neural plate border (Langeland et al., 1998). Therefore, the absence of neural crest cells in amphioxus may be due to the lack of necessary effectors to promote further differentiation or migration.
In summary, our results indicate that the transcription factors Tfap2a and Foxd3 are required for the proper establishment of Bmp and Wnt signaling gradients during gastrulation when neural crest induction occurs. Although loss of either factor alters Bmp and Wnt signaling patterns, only their combined inactivation appears to abort neural crest induction and deplete neural crest derivatives. The early functions of Tfap2a and Foxd3 appear distinct from their role in neural crest differentiation, when both factors are expressed in neural crest cells. Instead, it seems likely that at early developmental stages, the two factors control the expression boundaries of key morphogens in complementary territories to establish the proper Bmp and Wnt signaling balance needed for neural crest induction.
Zebrafish were raised and kept under standard laboratory conditions (Westerfield, 1993). Embryos were staged and fixed as described (Kimmel et al., 1995). 0.2 mM 1-phenyl-2-thiourea (Sigma) was used in some experiments to inhibit pigmentation.
The mobm610 (mob) allele is a mutation in the acceptor splice site of the sixth intron of the tfap2a gene, which results in a 14-bp deletion and a premature stop codon in the DNA binding domain (Barrallo-Gimeno et al., 2004). The mosm188 (mos) allele is a mutation in the non-coding region of the foxd3 gene locus (Montero-Balaguer et al., 2006). The mob;mos double mutant line was generated by crossing mos and mob heterozygote fish. The mos, mob, and mob;mos mutant embryos used for the experiments in this study were obtained by individually crossing heterozygous carriers. The ratio of embryos with altered gene expression in each clutch followed the expected Mendelian pattern. All results were recapitulated using morpholinos for tfap2a and foxd3, individually or together.
Embryos were obtained by natural mating and injected at the one- to two-cell stage. The sequence and amount of antisense morpholino oligonucleotides (Gene Tools) used in this study were as follows: tfap2a-MO (5 ng): 5'-GCGCCATTGCTTTGCAAGAATTG-3' (Knight et al., 2003); two foxd3 morpholinos were combined to generate mos-like embryos foxd3-MO5´UTR (1.5 ng): 5’-CACCGCGCACTTTGCTG CTGGAGCA-3’, and foxd3-MOAUG (1.5 ng): 5’-CACTGGTGCCTCCAG ACAGGGTCAT-3’ (Montero-Balaguer et al., 2006); p53-MO (5 ng): 5'-GCGCCATTGCTTTGCAAGAATTG-3' (Plaster et al., 2006); and dkk1-MO (1 ng): 5'-AATTGTAGGATGTATTCCCTGGGTG-3' (Caneparo et al., 2007).
Full-length tfap2a and foxd3 cDNAs were generated by PCR amplification from cDNA of wild-type embryos (primers used are available upon request) and cloned into the pCS2+ vector. After plasmid linearization, synthetic capped mRNA was generated using the mMessage mMachine kit (Ambion) and mRNA was injected into one-cell stage embryos. The following amounts were used: 150 pg of tfap2a mRNA and/or 15 pg of foxd3 mRNA.
In situ hybridization was performed as described (Müller et al., 2006). Antisense probes labeled with digoxigenin-UTP (Roche) were synthesized using cDNA templates encoding axin2 (Shimizu et al., 2000), bmp2b and bmp4 (Martinez-Barbera et al., 1997), chordin (Miller-Bertoglio et al., 1997), crestin (Rubinstein et al., 2000), dkk1 (Shinya et al., 2000), dlx2a (Akimenko et al., 1994), foxd3 (Kelsh et al., 2000), hgg1 (Vogel and Gerster, 1997), snail1b (Thisse et al., 1995), sox9b (Li et al., 2002), sox10 (Dutton et al., 2001), sp5l and tfap2a (Barrallo-Gimeno et al., 2004). Double in situ hybridization was performed using digoxigenin-and fluorescein-labeled riboprobes, and developed sequentially with NBT/BCIP and FastRed. Immunostaining was modified from a previously described protocol (Barrallo-Gimeno et al., 2004). In brief, fixed embryos were bleached with H2O2 as for cartilage staining. After washing, embryos were digested with 50 µg/ml proteinase K and post-fixed in 4% PFA for an additional 20 minutes. Embryos were then washed with PTD (1% DMSO, 0.3% Triton X100 in PBS) and placed in PTDNB (3% normal goat serum, 2 mg/ml BSA in PTD) for 2 hours. Embryos were then incubated with anti-Hu (1:20) (Invitrogen) or anti-phospho-Smad1/5/8 (1:200) (Cell Signaling Technology) antibodies overnight at 4°C. After washing, embryos were exposed to HRP-conjugated secondary antibody (1:5,000) for another hour at room temperature and washed again. Signals were developed using DAB as a substrate (Vector Laboratories).
TUNEL assay, acridine orange labeling and alcian blue staining were performed essentially as described (Barrallo-Gimeno et al., 2004).
Specimens were analyzed and photographed using a Zeiss Axioscope microscope, and composite images were prepared with Adobe Photoshop.
Total RNA was extracted using Trizol reagent (Invitrogen). We utilized 3 µg of total RNA for cDNA synthesis. Quantitative PCR reactions were performed with iQ™ SYBR Green Supermix (BioRad). Quantitative RT-PCR analysis was performed using the Bio-Rad Iq5 System and gene expression levels for each individual sample were normalized to actin. Results were analyzed by a previously described formula (Livak and Schmittgen, 2001). The following primers were used: actin: 5’-GACTCAGGATGCGGAAACTG -3’, 5’-GAAGTCCTGCAAGATCTTCAC-3’; axin2: 5’-ATTACCCAGCACTCGAAACTAA -3’, 5’-GCGAATTGTAGTCCAAATAAGC -3’; and sp5l: 5’-GATGCCTTATGTTGAAATCCTG-3’, 5’-GCACAGTCGATCGTGTTTATTA-3’.
We thank C. Guthrie, W. Rybski and K. Zavalin for excellent animal care, Alejandro Barrallo Gimeno for helpful comments and suggestions on the manuscript, W. Rybski for help with the experiments, and the Knapik lab for helpful discussions. The work has been supported in part by the NIH grants R01 DE018477 (E.W.K), HL083958 and HL100398 (A.K.H) and the Zebrafish Initiative of the Vanderbilt University Academic Venture Capital Fund (E.W.K).
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
The authors declare that they do not have any competing or financial interests.
Author ContributionsConceived and designed the experiments: WDW, DBM, MMB, AKH, EWK. Performed the experiments: WDW, DBM, MMB. Analyzed the data: WDW, DBM, MMB, AKH, EWK. Wrote the paper: WDW, AKH, EWK. Corrected manuscript drafts: EWK, DBM, AKH.