An Fgf-Shh signaling hierarchy regulates early specification of the zebrafish skull.

The neurocranium generates most of the craniofacial skeleton and consists of prechordal and postchordal regions. Although development of the prechordal is well studied, little is known of the postchordal region. Here we characterize a signaling hierarchy necessary for postchordal neurocranial development involving Fibroblast growth factor (Fgf) signaling for early specification of mesodermally-derived progenitor cells. The expression of hyaluron synthetase 2 (has2) in the cephalic mesoderm requires Fgf signaling and Has2 function, in turn, is required for postchordal neurocranial development. While Hedgehog (Hh)-deficient embryos also lack a postchordal neurocranium, this appears primarily due to a later defect in chondrocyte differentiation. Inhibitor studies demonstrate that postchordal neurocranial development requires early Fgf and later Hh signaling. Collectively, our results provide a mechanistic understanding of early postchordal neurocranial development and demonstrate a hierarchy of signaling between Fgf and Hh in the development of this structure.

1. Introduction

The neurocranium is an embryonic structure that generates essential craniofacial structures including the skull vault, skull base, and palate. The palate and skull base are connected and demarcate the prechordal, or anterior, and postchordal, or posterior, regions of the neurocranium, respectively. Extensive fate mapping in tetrapod species demonstrate that the neural crest and mesoderm contribute to the prechordal and postchordal neurocranium, respectively (Couly et al., 1992, 1993; Crump et al., 2004; Eberhart et al., 2006; Gross et al., 2008; Köntges and Lumsden, 1996; McBratney-Owen et al., 2008; Wada et al., 2011). Despite our knowledge of the tissue origins of the neurocranium, mechanistic studies of development have overwhelmingly focused on the prechordal neurocranium, while the postchordal region has been left largely neglected.

For proper formation, the neurocranium requires the orchestration of numerous signaling and morphogenetic events, and is dependent on interactions with surrounding neural as well as non-neural ectoderm, mesoderm, and endoderm (Alexander et al., 2011; Kimmel et al., 2001; Marcucio et al., 2011; Noden and Trainor, 2005; Richtsmeier and Flaherty, 2013). The complexity of these interactions implicates the involvement of multiple signaling molecules in neurocranial development. Much progress has been made in elucidating those factors that induce craniofacial formation and patterning. These factors include many signaling pathways, including Shh, Fgf, Bmp, and Wnt (Alexander et al., 2011; Marcucio, 2005; Richtsmeier et al., 2013; Wada et al., 2005; Wilson and Tucker, 2004). No one single factor can direct craniofacial formation; instead, they interact in spatial and temporal hierarchies that coordinate craniofacial growth and development.

While we know of many of the signaling hierarchies in the neural-crest derived-portions of the craniofacial skeleton, little is known of the mesoderm-derived portions. Hosokawa and colleagues showed that development of the posterior skull vault, a mesoderm-derived structure, involves a signaling hierarchy between TGF-B and Msx2 (Hosokawa et al., 2007). In mouse, chick, and zebrafish, Shh is a critical midline signal necessary for chondrocyte differentiation in the postchordal neurocranium, as well as other regions of the skull (Balczerski et al., 2012a; Eberhart et al., 2006; Wada et al., 2005). Specification of the early cephalic mesoderm, however, is refractory to sonic hedgehog signaling (Balczerski et al., 2012a), suggesting that other signals are necessary for early cephalic mesoderm specification. Due to their proximity to the postchordal neurocranium and their importance in numerous aspects of craniofacial development, the Fibroblast growth factor (Fgf) family is a prime candidate for this function.

Fgfs are part of a large family of intercellular signaling molecules (Itoh, 2007) that emanate from multiple tissue sources in the head. They are also crucial in numerous aspects of craniofacial development, including the proper migration, survival, and patterning of the neural crest (Creuzet et al., 2004; Crump et al., 2006; Hu et al., 2009; Wilson and Tucker, 2004) as well as cranial suture formation (Nie et al., 2006; Rice et al., 2000). Furthermore, Fgfs are implicated in a number of congenital craniofacial disorders with reported skull base defects including Apert and Crouzon syndrome (Aggarwal et al., 2006; Tokumaru et al., 1996). However, the role that Fgfs play in early postchordal neurocranial development is unknown.

Here, we characterize a signaling hierarchy required for proper postchordal neurocranial development involving Fgf and Shh signaling. Loss of function of both fgf3 and fgf8a lead to a striking loss of the postchordal neurocranium that can be rescued by restoring Fgf3 and Fgf8a signaling centers in the brain and mesoderm. We go on to precisely describe, for the first time, the dual tissue origins of the zebrafish neurocranium. The zebrafish postchordal neurocranium has small pockets of neural crest-derived areas occurring in a mostly mesodermally-derived structure. In situ analysis reveals that both the early cephalic mesoderm marker hyaluron synthetase 2 (has2) and markers for chondrocytes are lost in fgf3; fgf8a knockdown embryos, and that has2 is required for postchordal neurocranial development in an Fgf-dependent manner. Examination of Hh loss-of-function embryos reveals that Hh signaling is mostly dispensable for specification of head mesoderm. These results provide evidence of an early signaling interaction required for proper postchordal neurocranial development.

2. Results

2.1. The postchordal neurocranium requires fgf8a and fgf3

The neurocranium can be split into anterior and posterior halves (dashed line in Fig. 1A). While much research has been directed at the development of the neural crest-derived prechordal neurocranium (Fig. 1A, prech.), little is known of the development of the postchordal neurocranium (Fig. 1A, postch.). The postchordal neurocranium includes the parachordal cartilages (pc), which abut the notochord (n); as well as the anterior and posterior basicapsular commissures (abc and pbc, respectively) which encircle the developing ear, the lateral commissures (lc), and the occipital arches (oc) (Fig. 1A; de Beer, 1937). Due to its development adjacent to the notochord and the hindbrain, we reasoned that signals from one or both of these structures could influence posterior neurocranial development.

Fig. 1

The postchordal neurocranium requires Fgf signaling. (A-E) Wholemounted zebrafish neurocrania with the viscerocrania removed at 5 days post-fertilization. Anterior is to the left. (A) Wildtype, showing the prechordal versus postchordal regions of the neurocranium demarcated by a dashed line, (B) fgf8a mutant, (C) and fgf3 mutant embryos display normal neurocrania. (D) Variable neurocranial defects occur in 63% (n=5/8) of fgf8a−/− ; fgf3+/− embryos (arrowheads). (E) The postchordal neurocranium in fgf8a−/− ; fgf3−/− embryos is almost completely absent in 83% (n=5/6) of embryos, including the parachordals, anterior and posterior basicapsular commissure and the lateral commissures (arrowhead). abc-anterior basicapsular commissure, lc-lateral commissure, n-notochord, oc-occipital arch, pbc-posterior basicapsular commissure, pc-parachordals, prech.=prechordal neurocranium, postch.=postchordal neurocranium. Scale bar=20 μm.

Fgf8 and Fgf3 signal cooperatively to pattern the hindbrain, making these two Fgfs prime candidates for our analyses. We used a combination of genetic and morpholino-based loss of function of these Fgf ligands, each of which resulted in similar phenotypes (Fig. 1, Supplemental Fig. S1). Whereas neither fgf8a nor fgf3 single-mutants show any profound postchordal neurocranial defects (Fig. 1B and C compared to Fig. 1A), the vast majority of the postchordal neurocranium is absent in double fgf8a;fgf3 mutants (Fig.1E, arrowhead), with only the prechordal neurocranium and occipital arches remaining. Furthermore, even the loss of a single allele of fgf3 in an fgf8a mutant background gives rise to variable postchordal neurocranial defects, including anterior basicapsular and parachordal cartilage loss (Fig. 1D, arrowheads). These mutants also have severe viscerocranial defects (not shown, Crump et al., 2004). Because, at least, fgf8a is maternally expressed (Reifers et al., 1998), we extended our analyses using the well-characterized morpholinos against each of these Fgfs (Crump et al., 2004; Liu et al., 2003; Maves et al., 2002).

Morpholinos targeting either fgf8a or fgf3 injected into fgf3 or fgf8a mutants, respectively, recapitulated postchordal neurocranial defects observed in the double mutants (Supplemental Fig. S1). Embryos injected with fgf3 morpholinos display infrequent loss of the anterior basicapsular commissure (24%, n=6/24, Supplemental Fig. S1), however the remainder of the neurocranium in these embryos is well developed. Embryos injected with fgf3 morpholinos also consistently have fused otoliths (Supplemental Fig. S1) as well as variable viscerocranial defects, including fusions of Meckel's cartilage and the palatoquadrate, and loss of the ceratobranchial cartilages (data not shown; Crump et al., 2004). As in our mutant analysis, we found strong synergy between fgf8a and fgf3 in development of the postchordal neurocranium.

In contrast to phenotypes in single mutants, the postchordal neurocranium is lost in fgf3 morpholino-injected fgf8a mutants (100%, n=21/21, Supplemental Fig. S1). The ethmoid plate and trabeculae are retained, albeit reduced, in these embryos (n=21/21). Viscerocranial defects also occur in these embryos, including variable hyosymplectic loss, severe reductions and fusions of Meckel's cartilage and the palatoquadrate (data not shown; Crump et al., 2004). Lastly, we recapitulated the postchordal neurocranial loss in fgf8a morpholino-injected fgf3 mutants (Supplemental Fig. S1). Together, these data show fgf3 and fgf8a synergize during postchordal neurocranial development.

2.2. The mesoderm and neural ectoderm are Fgf sources required for postchordal neurocranium formation

Fgf signaling is required throughout development, forming numerous centers of activity including the mesoderm and neural ectoderm early, and the endoderm and otic placode later (Crump et al., 2004; Sivak et al., 2005; Thisse and Thisse, 2005). To investigate the required Fgf signaling source for postchordal formation, we generated genetic chimeras to re-introduce Fgf signaling from wild-type donor tissues in fgf3;fgf8 knockdown hosts. We tested four tissue sources: the mesoderm, neural ectoderm, endoderm, and otic placode (Fig. 2, Supplemental Fig. S2). While neither endoderm nor otic placode transplants restored the neurocranium (n=0/12 for endoderm, n=0/9 for otic placode, Fig. 2A–A” and B–B”), we observed a partial rescue of the postchordal neurocranium in embryos receiving either mesoderm or neural ectoderm transplants (n=4/7 for mesoderm, n=8/16 for neural ectoderm, Fig. 2C–C”, D–D” and H). In these transplants most of the parachordal cartilage and anterior basicapsular commissure were present, but rescue was incomplete, suggesting that both tissues might be required. To test this dual requirement, double mesoderm and neural ectoderm transplants were performed, and indeed, complete rescue of the postchordal neurocrania on the transplanted side in fgf3;fgf8a loss-of-function hosts was attained (n=4/12 for full rescue, n=3/12 for partial rescue, Figs. 2E–E”, 2H). Together, these data suggest that Fgf signaling from the mesoderm and neural ectoderm cooperate during formation of the postchordal neurocranium.

Fig. 2

The mesoderm and neural ectoderm are Fgf sources required for postchordal neurocranial formation (A–E, A’–E’) 24 h post-fertilization confocal images of transplanted wildtype tissues into fgf3;fgf8 morpholino-injected hosts with and without DIC, respectively; (A”–E”) corresponding 4 days post-fertilization (dpf) neurocranial wholemounts with viscerocrania removed. Transplanted side of neurocrania is marked with a T, and control side with a C. (F and G) 4 dpf postchordal neurocrania of uninjected and fgf3;fgf8 morpholino injected,embryos respectively. Anterior is to the left. (A, A’ and A”) Endoderm transplants, (B, B’ and B”) otic placode transplants, (C, C’ and C”) mesoderm only, (D, D’ and D”) neural ectoderm only, and (E, E’ and E”) mesoderm and neural ectoderm double-transplants. Endoderm and otic placode transplants fail to rescue the posterior neurocranium (compare A” and B” to F and G, n=0/12 for endoderm and n=0/9 for otic placode). Transplantation of mesoderm or neural ectoderm alone partially rescues the postchordal neurocranium (compare C” and D” to F and G, n=4/7 for mesoderm and n=8/16 for neural ectoderm). Transplantation of mesoderm and neural ectoderm together can fully rescue the postchordal neurocranium on the side of the embryo receiving the transplant (compare E” to F and G, n=4/12 for full rescue and n=4/12 for partial rescue). Quantification of partial or full rescue is shown in graph H. scale bars=100 μm in A–E and scale bar=20 μm in A”.

Fgf signaling from the neural ectoderm induces the otic placode (LÈger and Brand, 2002; Sai and Ladher, 2015). Thus, the partial rescue by neural ectoderm could be indirect, due to a signal relayed from the placode. If this indirect signaling were responsible for the rescue, then loss of the otic placode should phenocopy fgf8a;fgf3 double loss-of-function embryos. The otic placode is lost in embryos co-injected, but not singly injected, with dlx3b and foxi1 morpholinos (Solomon et al., 2002, 2003). We found that, as expected, injection of both dlx3b and foxi1 caused complete loss of the otic placode at 24 hpf (Solomon et al., 2004; Supplemental Fig. S3). However, in these embryos only the anterior basicapsular commissure was missing or reduced (Supplemental Fig. S3). In fact, the parachordal cartilages may be expanded. Thus, the otic placode may provide signals regulating postchordal neurocranial development, and future experiments analyzing single versus double loss-of-function embryos would help characterize this role. However, defects to the otic placode itself do not explain the extensive loss of the postchordal neurocranium found in fgf3;fgf8 loss-of-function embryos.

2.3. The postchordal neurocranium is primarily mesoderm-derived

Deeper analysis of the role of Fgf signaling in postchordal development requires a detailed characterization of the precursors to this structure. However, the origins of the postchordal neurocranium are unknown in zebrafish. To fully characterize any neural crest contribution to the postchordal neurocranium, we utilized three neural-crest labeling transgenic lines, sox10:KikGR, sox10:Kaede, and sox10:Cre;ubi:RSG (Balczerski et al., 2012b; Dougherty et al., 2013; Kague et al., 2012). The sox10:KikGR and sox10:Kaede embryos were photoconverted at 24 hpf when only neural crest cells express the transgene (Balczerski et al., 2012b; Dougherty et al., 2013), while the sox10:Cre;ubi:RSG line genetically labels neural crest cell descendants using a promoter active in neural crest cells (Kague et al., 2012). Results in all 3 lines were identical, with extensive labeling in the ethmoid plate (ep) and trabeculae (tr) of the prechordal neurocranium (Fig. 3A–D and G, data not shown; Eberhart et al., 2006; Kague et al., 2012; Wada et al., 2005). We also observed highly localized labeling within postchordal structures. Labeled areas included the lateral auditory capsule as well as the lateral and anterior-most region of the anterior basicapsular commissure (Figs. 3A-D, 3G). It is of interest that this precise region articulates with the second arch crest-derived hyosympletic of the viscerocranium (Crump et al., 2004).

Fig. 3

The zebrafish postchordal neurocranium is derived from both mesoderm and neural crest tissues. (A–F) Confocal images of 5 days post-fertilization flatmounted zebrafish neurocrania of (A and B) sox10:KikGR, (C and D) sox10:Cre;ubiRSG, and (E and F) drl:CreERT2;ubi:switch. Anterior is to the left. In C and D the red and green channels have been switched for clarity in comparisons to A and B. (B, D and F) Lineage-traced cells only, neurocrania outlined in white. (A–D) Neural crest contributes to the prechordal elements the ethmoid plate and trabeculae, as well as the lateral and anterior regions of the anterior basicapsular commissures and the lateral auditory capsules. (E and F) Mesoderm contributes the parachordals, anterior and posterior basicapsular commissure, the occipital arch, and the lateral commissures. Schematic of results shown in G (n ≥ 10 for each transgenic line). abc-anterior basicapsular commissure, ac-auditory capsule, ep-ethmoid plate, lc-lateral commissure, n-notochord, oc-occipital arch, pbc-posterior basicapsular commissure, pc-parachordals, tr-trabeculae. scale bar=50 μm.

The first and second arches undergo dynamic morphogenetic movements that involve numerous neural crest-specific cellular rearrangements (Crump et al., 2004, 2006; Eberhart et al., 2006). To elucidate which pharyngeal arch contributes neural crest cells to the postchordal neurocranium, we performed fate mapping with sox10:Kaede embryos. We photoconverted either the first or second arch at 26 hpf (Supplemental Fig. S4; Crump et al., 2006), and found that the second arch crest contributed to the lateral auditory capsule (Supplemental Fig. S4), while the first arch crest contributed to the lateral and anterior most region of the anterior basicapsular commissure (Supplemental Fig. S4). Together, these results show that neural crest of the first and second arch contribute to adjacent regions of the postchordal neurocranium.

In both mouse and chicken, mesoderm contributes significantly to the posterior neurocranium. We used the mesoderm-labeling drl:CreERT2;ubi:switch line to track mesoderm derivatives in zebrafish (Supplemental Fig. S5; Mosimann et al., 2011). Tamoxifen was added from 6 hpf to 24 hpf to induce Cre- excision. We find a pattern of labeling perfectly complimentary to that generated by the neural crest (Fig. 3E–G). The parachordal cartilages, posterior basicapsular commissures, and occipital arches appear to be completely of mesoderm origin (Fig. 3E–G). The regions of the anterior basicapsular commissure that were not labeled by neural crest are also of mesoderm origin (Fig. 3E–G). The lateral commissure was the only structure where neural crest and mesoderm cells appear to mix (Fig. 3). Collectively, these data provide a high-resolution fate map of neural crest and mesoderm contribution to the neurocranium and suggest that, even in regions of dual origin, there is little mixing between these cell types.

In order to characterize what happens to this mesoderm population in Fgf loss-of-function embryos, it is necessary to understand their location in the early embryo. We used Kaede photoconversion to label and track groups of cells from 24 hpf to 4 days post-fertilization (dpf), when the postchordal neurocranium is well formed in zebrafish (Fig. 4A–D). We injected fli1:EGFP embryos, which labels neural crest early in development but all cartilage later, with Kaede mRNA and photoconverted cells just dorsal to the pharyngeal arches for our analyses, a location that should contain mesoderm (Kimmel et al., 1990). In the images, the intense EGFP fluorescence masks the much dimmer green Kaede fluorescence so only the red, photoconverted, Kaede is apparent. We found that at 24 hpf, cells that contribute to the postchordal neurocranium are positioned within the head in relative accord to their location at 4 dpf (Fig. 4E and E’, see Supplemental Fig. S6 for individual data). Based on our analyses in Fig. 3 these cells are of mesodermal origin. Overall, no relative change in position of these labeled cells occurred between 24 hpf and 4 dpf. Cells labeled anterior of the otic capsule (o.c.) maintain this position at 4 dpf (Fig. 4E and E’, dark gray shade). Those cells labeled more medial and posterior to the otic capsule at 24 hpf populate areas medial and posterior 4 dpf (Fig. 4E and E’, gray and light gray shades). During this same time period, neural crest cells appear to undergo extensive cell rearrangements in forming skeletal structures (Crump et al., 2006; Le Pabic et al., 2014). We find little dispersion of labeled mesoderm cells, suggesting a lack of similar rearrangements in the mesoderm-derived skeleton. Together, these data show that the progenitors of the postchordal neurocranium are appropriately positioned along the anterior to posterior axis at 24 hpf.

Fig. 4

The anterior/posterior organization of postchordal cells is set by 24 h post-fertilization (hpf). (A–C) Confocal images of a Kaede-injected embryo at (A) 24 hpf and (B and C) 4 days post fertilization (dpf). Anterior is to the left. (A) At 24 hpf Kaede was photoconverted, shown in red (the remaining green Kaede is not evident due to the intense green fluorescence from the fli1:EGFP transgene), and (B) at 4 dpf, this same embryo shows labeling in the postchordal neurocranium (inset shows relative position in the neurocranium of magnified view). (C) Red channel only, the neurocrania is outlined in white. (D) Schematic of panel C, with red cells depicting the photoconverted region. (E and E’) Graphical representation of Kaede-mediated fate mapping at 24 hpf and 4 dpf. Inset in E’ shows relative position of magnified neurocrania. Question mark denotes a region that remained unlabeled in our analyses. ac- auditory capsule, o.c.-otic capsule. Scale bars=10 μm in A and 20 μm in C.

2.4. Proper specification of the head paraxial mesoderm requires Fgf signaling

Our fate map of the postchordal neurocranium shows that these cartilage precursors are in place by 24 hpf. To investigate whether these precursors are present in embryos lacking fgf3 and fgf8a, we analyzed the expression of the prechondrogenic marker sox9a and the cartilage marker col2a1a. Compared to un-injected and control morpholino-injected fgf8a mutants, fgf3;fgf8a double loss-of-function embryos display a loss of col2a1a in the cephalic mesoderm (Fig. 5D, compared to A–C, asterisks; Piotrowski and Nusslein-Volhard, 2000). The notochord staining for col2a1a is retained in double loss-of-function embryos, demonstrating specificity of the effect (Fig. 5D, arrowheads denote anterior notochord). Furthermore, the pre-cartilage marker sox9a is also lost in these fgf3 morpholino-injected fgf8a mutants (Fig. 5H, compared to E–G). Together, these data suggest that the head paraxial mesoderm that will generate these cartilages is either mislocalized or improperly specified when Fgf signaling is attenuated.

Fig. 5

Cartilage and pre-cartilage markers col2a1a and sox9a are absent at 24 h post-fertilization (hpf) in fgf3;fgf8a knockdown embryos. (A–H) Images of 24 hpf embryos labeled with (A–D) col2a1a and (E–H) sox9a riboprobe. Dorsal view with anterior to the left, arrowhead denotes the anterior limit of the notochord. (A and E) Wildtype, (B and F) fgf8a mutant, and (C and G) fgf3 morpholino embryos display normal expression of both col2a1a and sox9a. However, (D and H) fgf3 morpholino-injected fgf8a mutants lose the expression of both col2a1a and sox9a (asterisk). scale bar=20 μm.

To directly address if head paraxial mesoderm is mislocalized in Fgf loss-of-function embryos, we tracked endomesoderm progenitor cells from the initiation of gastrulation (6 hpf) to the end of gastrulation (10 hpf) via Kaede photoconversion. We elaborated on established fate-maps of the mesoderm to label and track postchordal progenitor cells (Kimmel et al., 1990). We injected embryos with Kaede mRNA, photoconverted endomesoderm cells adjacent to the shield at 6 hpf (Fig. 6A-B), and imaged their progression adjacent to the notochord at 10 hpf (Fig. 6C). These labeled cells were then confirmed to contribute to the postchordal neurocranium at 4 dpf (Fig. 6D). Similar to control and Fgf single loss-of-function embryos, mesoderm cells in fgf3;fgf8a morpholino-injected embryos migrated to the notochord by 10hpf (Fig. 6H compared to E–G). Furthermore, these cells remain adjacent to the notochord at 24 and 48 hpf (Supplemental Fig. S7). These results demonstrate that head paraxial mesoderm does migrate to its location lateral to the notochord in Fgf loss-of-function embryos and suggest that the postchordal neurocranial defects may be due to a failure in specification of this population of head mesoderm cells.

Fig. 6

Kaede photoconverted endomesoderm postchordal-progenitor cells migrate appropriately in fgf3;fgf8a knockdown embryos. (A) Schematic of 6 h post-fertilization (hpf) zebrafish embryo showing the region of the embryo photoconverted in all subsequent experiments, dorsal to the right. (B–D) DIC confocal images showing the photoconverted area at 6 hpf (B, magnified region outlined in A), 10 hpf (C), and 4 days post-fertilization (dpf) (D). Cells have migrated adjacent to the notochord by 10 hpf and contributed to the postchordal neurocranium at 4 dpf. At 10 hpf, cells in (E) control, (F) fgf8, (G) fgf3, and (H) fgf3;fgf8 double morpholino-injected embryos are appropriately positioned (compared to C). scale bar=20 μm.

Because the head paraxial mesoderm expresses sox9a and col2a1a relatively early in development, we hypothesized that some genes important in chondrocyte maturation would be early markers of this population of head paraxial mesoderm. Hyaluronan synthetase 2 (has2) is required for hyaluronic acid synthesis (Bakkers et al., 2003; Moffatt et al., 2011; Necas et al., 2008; Weigel et al., 1997; Yoshida et al., 2000). Hyaluronic acid is an important extracellular glycosaminoglycan involved in numerous cellular processes including chondrocyte maturation (Moffatt et al., 2011; Necas et al., 2008). Mouse Has2 knockouts display cephalic mesoderm defects (Camenisch et al., 2000) and in chick limb mesodermal cells, fetal bovine chondrocytes, mouse ear placodal cells, and breast cancer cells, Has2 is positively regulated by Fgf signaling (Bohrer et al., 2014; Hamerman et al., 1986; Munaim et al., 1991; Urness et al., 2010). These data suggest that has2 might not only be a marker of head paraxial mesoderm, but may also be involved in the Fgf loss-of-function phenotype.

Our earlier analysis revealed that localization of the head paraxial mesoderm was unperturbed in fgf3;fgf8 knockdown embryos at 10 hpf. However, by 10 hpf, has2-positive cells abutting the notochord in the most anterior region, where the postchordal neurocranium is developing, are lost in fgf3;fgf8 knockdown embryos (Fig. 7D–D’, compared to A–C, A’–C’). Expression of has2 at 24 hpf is also lost in double fgf8a; fgf3 mutant embryos (Fig. 7E and F). No gross alterations in earlier has2 expression, at 6 hpf, could be ascertained in fgf3;fgf8 knockdown embryos (Supplemental Fig. S8). This suggests that early cephalic mesoderm is not properly specified when Fgf signaling is attenuated.

Fig. 7

Loss of has2 in Fgf knockdown embryos contributes to the postchordal neurocranial loss phenotype. (A–D, A’–D’) Images of 10 h post-fertilization, and (E–F) 24 h post-fertilization embryos labeled with has2 riboprobe. Dorsal view and anterior to the left in all images. (G–J) Wholemounted zebrafish neurocrania with the viscerocrania removed at 5 days post fertilization. Anterior is to the left. (A’–D’) Magnified areas of region outlined in (A–D). (A and A’) Uninjected wildtype, (B and B’) fgf8a mutants, and (C and C’) fgf3 morpholino-injected wildtype display normal expression of has2, however, (D and D’) fgf3 morpholino-injected fgf8a mutants exhibit loss of expression of has2 in the presumptive postchordal neurocrania (D and D’). (E) At 24 hpf, some has2 expression remains along the notochord, however, (F) expression is lost in fgf8a;fgf3 double mutant embryos. Arrowheads denote end of the notochord. (G) Control morpholino-injected wildtype and (H) fgf8a mutants, as well as (I) has2 morpholino-injected wildtypes retain the postchordal neurocranium. However, (J) has2 morpholino-injected fgf8a mutants have a substantial loss of the postchordal neurocranium (arrowheads). n=notochord. scale bar=50 μm in A, 10 μm in A’ and E, and 20 μm in G.

Chondrocyte maturation has been shown to rely on has2 function, however, it was unclear whether loss of has2 contributed directly to the postchordal neurocranial defects in fgf3;fgf8 knockdown embryos. We injected fgf8a mutants with a suboptimal dose of a previously published morpholino targeting has2 (Bakkers et al., 2004). Wildtype siblings are largely unaffected in the posterior neurocranium (Fig. 7I compared to G), as are control morpholino-injected fgf8a mutants (Fig. 7H compared to G). However, has2 morpholino-injected fgf8a mutants display partial loss of the postchordal neurocranium (Fig. 7J, arrowheads). This treatment also causes reductions to the neural crest-derived portion of the neurocranium, which may be explained by the role of has2 in the neural crest (Casini et al., 2012). These data suggest that the exacerbated postchordal neurocranial defects observed in fgf3; fgf8 knockdown embryos is, at least, partially due to loss of has2.

The major function of Has2 is in the synthesis of hyaluronic acid (Bakkers et al., 2003; Moffatt et al., 2011; Necas et al., 2008; Weigel et al., 1997; Yoshida et al., 2000), which is known to regulate cartilage development (Matsiko et al., 2012; Matsumoto et al., 2009; Sato et al., 2014). To test the requirement of this function of has2, we treated fgf8a mutants with a suboptimal concentration of 4-methylumbelliferon (4-MU; Sigma-Aldrich), which is known to inhibit hyaluronic acid production (García-Vilas et al., 2013). Treating fgf8a mutants with 4-MU between 6 and 10 hpf did not result in appreciable postchordal neurocranial defects (not shown). However, when treated between 10 and 30 hpf, 4 MU-treated fgf8a mutants display loss of col2a1a expression at 30 hpf and, at 5 dpf, disruption of mesodermal-derived portions of the postchordal neurocranium, as compared to control-treated fgf8a mutants and 4 MU-treated wildtype siblings (Supplemental Fig. S9). Collectively, these data suggest that the postchordal neurocranial defects found in Fgf-compromised embryos is dependent on an early requirement for Fgf signaling on hyaluronic acid production in postchordal neurocranial precursors.

Due to the loss of has2 expression at 10 hpf in fgf3;fgf8a knockdown embryos, we reasoned that Fgf signaling would be required early for postchordal neurocranial development. Using a suboptimal dose of SU5402 (Tocris Biosciences) from 6 to 10 hpf on fgf8a mutants and siblings, which only partially reduces Fgf-dependent etv4 expression at 10 hpf in wildtype and fgf8a-treated embryos (Supplemental Fig. S10), we found that, while SU5402-treated wildtypes showed normal neurocranial development (Fig. 8B compared to A and E), both heterozygote and mutant fgf8a zebrafish showed postchordal neurocranial defects, affecting the anterior basicapsular commissure in heterozygotes, and the entirety of the postchordal neurocranium in mutants (Fig. 8D compared to A–C and E). Furthermore, has2 expression at 10 hpf in SU5402-treated fgf8a mutants is completely absent in the region of the developing postchordal neurocranium (compare Fig. 8I to F and G). Likewise, at 24 hpf, both col2a1a and sox9a are reduced in SU5402-treated fgf8a mutants (Supplemental Fig. S11). These data strongly suggest that Fgf signaling is required during gastrulation, between 6 and 10 hpf, for postchordal neurocranial formation and specification.

Fig. 8

Postchordal neurocranial development requires Fgf signaling during gastrulation. (A–D) Wholemount zebrafish neurocrania, anterior is to the left. (A and B) Wildtype neurocrania treated with DMSO or SU5402 develop normally. (C) DMSO-treated fgf8a mutants are also unaffected, however, (D) those treated with SU5402 from 6 to 10 hpf develop severe postchordal neurocranial loss. (E) Quantification of postchordal neurocranial defects including none (see A), ABC loss, or complete postchordal neurocranial loss (p. nc. loss) in DMSO and SU5402 treated wildtype (fgf8a+/+), heterozygous (fgf8a+/−) and mutant (fgf8a−/−) embryos. (F–H) DMSO-treated wildtype and fgf8a mutants and SU5402-treated wildtype express has2 appropriately; however, (I) SU5402-treated fgf8a mutants display a loss of expression of has2 in postchordal neurocranial precursors (arrowhead denotes most anterior expression). abc- anterior basicapsular commissure, lc-lateral commissure, n-notochord, oc-occipital arch, pc-parachordals, pbc-posterior basicapsular commissure. scale bar=20 μm in A and scale 10 μm in F.

2.5. The notochord is dispensable in the formation of the postchordal neurocranium

Our data show that Fgf signaling, originating from the mesoderm and neural ectoderm, is required in the development of the postchordal neurocranium. Previous reports have shown the importance of Shh signaling from the notochord in this process (Balczerski et al., 2012a). Considering our data, Fgf and Shh together may form a signaling hierarchy required for postchordal neurocranial development. Thus, we investigated the function of the notochord and Shh in the formation of the postchordal neurocranium.

To directly ask whether the maintenance of a notochord is necessary for cephalic mesoderm induction and postchordal neurocranial formation, we analyzed brachyury mutants (previously known as no tail), which transfate the notochord early in development (Amacher et al., 2002). However, brachyury mutants express has2 at 10 hpf (Fig. 9B–B’, compared to A–A’), col2a1a at 24 hpf (Fig. 9D), and retain a postchordal neurocranium (Fig. 9F, compared to E). These data show that notochord maintenance is dispensable for the formation of the postchordal neurocranium.

Fig. 9

Postchordal neurocranial development does not require brachury function or a notochord. Anterior is to the left in all panels. (A–B, A’–B’) hyaluronan synthetase 2 (has2) is expressed in cephalic regions at 10 h post-fertilization (hpf) in both wildtypes and brachury mutants. (C and D) The late chondrogenic marker col2a1a at 24 hpf appears in both brachury mutants and siblings in the forming postchordal area of the developing embryo (compare D to C, arrowhead in C denotes anterior notochord). (E and F) At 5 days post-fertilization, postchordal neurocrania of brachury mutants are normal, sans notochord, compared to siblings (compare F to E). scale bar=50 μm in A and 10 μm in A’ and C and 20 μm in E.

In brachyury mutants, midline sources of Hh remain. To directly test the involvement of Hh signaling in posterior neurocranial development, we analyzed smo mutants, which lack all Hh signaling and a postchordal neurocranium (Eberhart et al., 2006). In smo mutants, we find that has 2 expression in the region of the postchordal neurocranium is present (Fig. 10B–B’ compared to A–A’), but there are severe reductions in cells expressing the chondrocyte marker col2a1a (Fig. 10D compared to C, Eberhart et al., 2006; Wada et al., 2005). Our smo data suggests that Hh signaling is required for differentiation of chondrocytes in the mesoderm-derived postchordal neurocranium after 10 hpf.

Fig. 10

Chondrocyte differentiation is particularly sensitive to disruption of Hh signaling. All panels are anterior to the left. (A–B, A’–B’) At 10 h post-fertilization (hpf), the early mesoderm marker has2 is expressed in the anterior region of smoothened mutants and wildtypes (Compare B to A, B’ to A’). (C and D) However, smoothened mutants display a marked reduction in col2a1a expression in the forming postchordal neurocranium at 28 hpf compared to siblings (compare D to C, arrowhead denotes anterior notochord). Scale bar=50 μm in A and 10 μm in A’.

To directly test the temporal requirement of Hh in the formation of the posterior neurocranium, we utilized the pan-Hh inhibitor cyclopamine (Toronto Research Chemicals; Hirsinger et al., 2004). Treating wild-type embryos from 6 to 10 hpf with cyclopamine did not abolish has2 expression at 10 hpf in the anterior region of the embryo (Fig. 11B and B’ compared to A and A’). In these same embryos, sox9a and col2a1a expression at 24 hpf remained as well (Fig. 11D and G compared to C and F, respectively), albeit to a potentially lesser degree. However, blocking Hh signaling between 10 and 24 hpf resulted in the complete absence of sox9a and col2a1a expression (Fig. 11E and H). Together, these data suggest that, while Hh signaling is essential for cartilage differentiation, it plays less of a role in early specification of mesoderm-derived postchordal neurocranial progenitors (Fig. 12). Fgf signaling, on the other hand, is important for this early specification step, via activation of has 2 (Fig. 12).

Fig. 11

Hh signaling is required after Fgf signaling for postchordal neurocranial development. The expression of has2 is retained following either (A–A’) DMSO or (B–B’) cyclopamine treatment from 6 to 10 h post-fertilization (hpf). (C and D and F–H) DMSO- and cyclopamine-treated embryos from 6 to 10 hpf show expression of both sox9a and col2a1a at 24 hpf, however, (E and H) cyclopamine-treated embryos from 10 to 24 hpf show a loss of sox9a and col2a1a expression in the postchordal neurocranium at 24 hpf. Arrowheads denote the anterior limit of the notochord. scale bar=50 μm in A and 10 μm in A’.

Fig. 12

Cephalic mesoderm specification and differentiation utilizes Fgf and Hh signaling pathways. (A) During early postchordal mesoderm specification, Fgf signaling from both the overlying neuroectoderm and mesoderm is necessary for has2 expression in mesodermal progenitors of the postchordal neurocranium. (B) During differentiation, Hh signaling from the overlying neuroectoderm promotes the expression of sox9a and col2a1a in these mesodermal cells. (C) Loss of fgf8a and fgf3 results in a loss of has2 expression in mesodermal cells, while (D) reduced Hh signaling results in loss of proper differentiation of these cells.

3. Discussion

Here we describe a hierarchy of genetic signaling required for the specification and differentiation of the postchordal neurocranium in zebrafish. The postchordal neurocranium is a structure primarily derived from mesoderm, and is lost in embryos with attenuated Fgf or Shh signaling. Fgf signaling plays an early role in the specification of head mesoderm via has2. The loss of has2 and subsequent hyaluronic acid production is at least partly causative of the postchordal neurocranial defects. Shh signaling is then required for the later differentiation of sox9a and col2a1a-expressing chondrocytes in the postchordal neurocranium. Together, these results reveal a previously unknown genetic signaling hierarchy required in the development of the postchordal neurocranium.

3.1. The dual origin of the zebrafish neurocranium

The zebrafish neurocranium originates from the neural crest and mesoderm. Frog, mouse, and chick fate-maps show similar neurocranial contributions (Couley et al., 1993; Gross and Hanken, 2008; Jiang et al., 2002; Kontges and Lumsden, 1996; McBratney-Owen et al., 2008). Our results strongly suggest that the ancestral pattern of neurocranial contribution is neural crest being largely restricted to prechordal regions, and mesoderm only providing contributions to postchordal regions. The prechordal neurocranium is exclusively of neural-crest origin. This region of the neurocranium has received a good deal of characterization in zebrafish (Eberhart et al., 2006; Kague et al., 2012; Mongera et al., 2013; Wada et al., 2005), therefore, here we will focus on the postchordal neurocranium.

Using a pan-mesodermal Cre-transgenic line driven by draculin, we found that the majority of the postchordal neurocranium is mesoderm-derived. In mouse and chick, the postchordal neurocranium is also primarily mesoderm-derived (Couly et al., 1992, 1993; Kontges and Lumsden, 1996; McBratney-Owen et al., 2008). These data are lacking in frog, however, in neural-crest labeling fate-map studies, non-neural crest derived cartilages are all positioned in the postchordal neurocranium (Gross and Hanken, 2008).

More effort has been spent in mapping the neural crest-derived portions of the prechordal skull (Couly et al., 1993; Gross and Hanken, 2008; Kague et al., 2012; Kontges and Lumsden, 1996; Jiang et al., 2002; McBratney-Owen et al., 2008). Our fate mapping has defined the precise postchordal structures that neural crest cells contribute to in zebrafish, including the most lateral regions of the basicapsular commissures, the lateral auditory capsule, and the parts of the lateral commissures. These areas are important for articulations with the jaw support element the hyosymplectic, as well as muscle attachment sites (Köntges and Lumsden, 1996). Furthermore, first and second arch neural crest cells contribute to distinct regions of the postchordal neurocranium. Along with results in chick (Köntges and Lumsden, 1996), this finding suggests an evolutionarily conserved function of neural crest in forming attachment sites in the postchordal neurocranium with the neural crest-derived jaw, jaw supports, and muscle attachment sites (Köntges and Lumsden, 1996).

3.2. Fibroblast growth factor signaling in the zebrafish neurocranium

The vertebrate neurocranium is the product of the mesoderm and the cranial neural crest, yet requires interactions between multiple tissues. An important regulator of the development of the neural crest-derived portion of the neurocranium is Fibroblast growth factor signaling (Creuzet et al., 2004; Crump et al., 2006; Monsoro-Burq et al., 2003). Our data reveal a second, and much earlier, role for Fgf signaling in the mesoderm-derived postchordal neurocranium.

We show that loss-of-function of both fgf3 and fgf8a cause severe defects in postchordal neurocranial development. The root of this defect is likely the misspecification of cephalic mesoderm at the end of gastrulation. Notably there is a loss of expression of the chondrocyte-regulator hyaluronan-synthetase 2 (has2), which is essential for hyaluronic acid production, positioning has2 downstream of Fgf. In fgf8a mutants, loss of has2 or hyaluronic acid, led to perturbed postchordal neurocranial defects similar to fgf3;fgf8 loss-of-function embryos. These data suggest that has2 is necessary for proper postchordal neurocranial formation in fgf8a-dependent fashion. It will be of interest to determine if transgenic activation of has2 is sufficient to rescue the Fgf loss-of-function defects. How has2 expression is activated downstream of Fgf signaling and if Fgf signals directly to the head paraxial mesoderm remain to be elucidated. Fgf signaling could maintain has2 expression directly, via STAT3 activation (Saavalainen et al., 2005). Indeed, Fgf receptors have been shown to activate HAS2 function in breast cancer cells via STAT3 (Bohrer et al., 2014). Understanding the requirement of Fgf receptors in the postchordal neurocranium is of ongoing interest and will be important to understand the exact mechanism of Fgf-dependent has2 function in the postchordal neurocranium.

3.3. A signaling hierarchy orchestrates the specification and differentiation of the postchordal neurocranium

Proper craniofacial development relies on complex hierarchies of signaling pathways. Our analysis indicates Fgf signaling from the mesoderm and neuroectoderm is a major regulator of postchordal neurocranial development. Otic placode development is dependent upon signals from the neuroectoderm (LÈger and Brand, 2002; Sai and Ladher, 2015), which could confound our otic placode transplantation results. However, loss of the otic placode itself only results in variable anterior basicapsular commissure loss, not the severe postchordal neurocranial loss phenotype observed in fgf3;fgf8 knockdown embryos. Thus, the mesoderm and neuroectoderm appear to be the most important sources of Fgf for postchordal development.

Other studies have purported that the notochord is also vital in postchordal neurocranial formation and that shh emanating from this structure mediates the formation of the postchordal neurocranium via chondrogenesis of the paraxial mesoderm (Balczerski et al., 2012a). Consistent with this report, we find a loss of differentiated chondrocytes in the postchordal neurocranium. In contrast, has2 expression was retained, albeit potentially reduced, in smoothened mutants, which completely lack Hh signaling (Varga et al., 2001), and in embryos treated with cyclopamine from 6 to 10 hpf compared to controls. This shows that, unlike Fgf signaling, Hh signaling may be dispensable in early specification of postchordal neurocranium precursors, but is certainly important in their terminal differentiation.

The notochord has been thought to be a critical source of Shh for posterior neurocranial development (Balczerski et al., 2012a, 2012b). Analysis of brachyury mutants, which transfate the notochord during gastrulation (Amacher et al., 2002; Halpern et al., 1993), showed that the postchordal neurocranium was fully developed and included a chondrocytic region expanded into the area where the notochord develops. The expression of shh is maintained in brachyury mutants (Amacher et al., 2002). Collectively, these findings suggest that the notochord is not a required shh source. However, the possibility remains that, in wild-type embryos, shh from the notochord does serve a signaling function to the developing postchordal neurocranium.

Our findings now place Fgf signaling prior to Hh in the formation of the postchordal neurocranium. Cephalic mesoderm is first specified in an Fgf-dependent manner beginning at 10 hpf via activation of has2. Temporal-loss of Fgf signaling via SU5402 treatment also shows that Fgf signaling is required early, from 6 to 10 hpf, for has2 activation. Loss of has2 ultimately results in the loss of the chondrogenic program resulting in postchordal neurocranial defects in Fgf loss-of-function embryos. Shh, on the other hand, is not essential for this early activation of has2, but supports proper chondrogenic differentiation of this group of cells. Together, these results clarify the temporal and genetic control required for proper postchordal neurocranial development in zebrafish.

4. Materials and methods

4.1. Fish husbandry and care

All embryos were raised and cared for using established protocols (Westerfield, 1993) with IACUC approval from the University of Texas at Austin. The fgf8ati282a (Brand et al., 1996), fgf3t24149 (Herzog et al., 2004), brachyury (Schulte-Merker et al., 1994), smoothened(Varga et al., 2001), sox10:Cre (Kague et al., 2012), sox10:KikGR (Balczerski et al., 2012b), sox10:kaede (Dougherty et al., 2013), ubi:switch (Mosimann et al., 2011), and ubi:RSG (Kikuchi et al., 2010) alleles have all been described previously. The drl:CreERT2 line was generated using a 3.8 kb draculin promoter upstream of the ATG start site. Primers used to amplify this promoter were: for1: ATTGCGGCCGCTTCAATTGTGGTTGAGCAGTC. rev1:ATTACTAGTCCAAGTGTGAATTGGGATCG. The 3.8KB fragment was amplified with iProof polymerase (BioRad), then cloned into TOPO-Blunt (Invitrogen). After verification, the promoter was sub cloned into the Tol2 vector (Kawakami et al., 2004) upstream of the CRE-ERT2 (Feil et al., 1996, PNAS). The Tol2-drl-creert2 vector was co-injected with Tol2 mRNA into AB* in order to establish a founder line. Tamoxifen was added from 10 to 24 h post-fertilization on the drl:CreERT2 line, and the sox10:KikGR and sox10:kaede lines were UV-activated at 24 h post-fertilization using DAPI-filter attached to a Zeiss LSM 710. Embryos were treated in embryo media with 5 μM SU5402 (Tocris Biosciences) from 6 to 10 hpf, and 100 μM cyclopamine (Toronto Research Chemicals) or 1 mM 4-methylumbelliferone (Sigma-Aldrich) from 6 to 10 and 10–24 hpf.

4.2. Morpholino and RNA injection

Approximately 5 nl of morpholinos (Gene Tools), working concentrations of 5 mg/ml of a combination of fgf3b and fgf3c morpholinos with sequences 5′GGTCCCATCAAAGAAGTATCATTTG3′ and 5′TCTGCTGGAATAGAAAGAGCTGGC3′, respectively (Maves et al., 2002), of a combination of fgf8aE212 and fgf8aE313 with sequences 5′TAGGATGCTCTTACCATGAACGTCG3′ and 5′CACATACCTTGCCAATCAGTTTCCC3′ (Draper et al., 2001), and control morpholinos with sequence 5′CCTCTTACCTCAGTTACAATTTATA3′ were injected into one- or two-cell stage embryos of fgf8a and fgf3 lines. Approximately 3 nl of a working concentration of 3 mg/ml of fgfr3 morpholino with sequence 5′AAATGAGGTGTAATGTCTGACCTGT3′ was injected into fgf8a mutants. This dose was suboptimal, as it did not cause any defects to wildtype-injected embryos. This is a splice-blocking morpholino targeting the first exon-intron boundary of fgfr3. To validate the targeting of this morpholino, whole embryo RNA extracts were isolated from un-injected and fgfr3 morpholino-injected zebrafish at 24 hpf using Trizol extraction (Invitrogen). cDNA pools were then synthesized using Superscript First-Strand Synthesis System (Invitrogen). To detect changes in mRNA transcripts in morphant embryos, PCR was performed on cDNA pools using the gene-specific primers spanning the morpholino-targeted exon-intron boundary For-TACAGTGCACACCTGCTGTC and revAGCCAATGGATACTGGGCG giving a final size of 491 bp in wildtype. Other morpholinos used that were previously described: dlx3b: 5′-ATATGTCGGTCCACTCATCCTTTAAT-3′ (Solomon et al., 2002); foxi1: 5′-TAATCCGCTCTCCCTCCAGAAACAT-3′ (Solomon et al., 2003) ; and, has2, which required dual injection of two morpholinos: 5′-AGCAGCTCTTTGGAGATGTCCCGTT-3′ and 5′-CGTTAGTTGAACAGGGATGCTGTCC-3′ (Bakkers et al., 2004).

Kaede mRNA was injected into one-cell stage embryos, with or without morpholinos, and UV activated at either 6 or 24 hpf using a Zeiss LSM 710 Confocal microscope.

4.3. Cartilage and bone staining

Five and four day postfertilization (dpf) zebrafish embryos were stained with Alcian blue and Alizarin Red (Walker and Kimmel, 2007), and then were either flat mounted (Kimmel et al., 1998) or had the viscerocrania removed for imaging. Images were taken with a Zeiss Axio Imager-AI microscope. Graphs were made in Microsoft Excel 2011.

4.4. Confocal microscopy and figure processing

Confocal z-stacks were collected on a Zeiss LSM 710 using Zen software. Images were processed in Adobe Photoshop CS. Kaede and tissue fate maps were generated in Adobe Photoshop CS by overlaying images gathered on the Zeiss confocal. Graphs were generated using Microsoft Excel 2011.

4.5. In situ hybridization

RNA in situ hybridization was performed as reported in Miller et al. (2000). AB wildtype, fgf8a, fgf3;fgf8a morpholino, brachyury, and smoothened embryos were treated with 0.0015% PTU (1-phenyl 2-thiourea) to inhibit the production of melanin (Westerfield, 1993). Probes used were sox9a (Yan et al., 2002), col2a1a (Yan et al., 1995), and the has2 probe was generated using primers For:ACAAGTCACTGGCCCTATGC and Rev:GGTAGGTAATGGGCGTCTCG (NCBI ref# NM_153650.2). DIC images of in situ hybridizations were collected on a Zeiss Axioimager.

4.6. Cell transplants

Genetic mosaics were generated as described elsewhere (Crump et al., 2004; Maves et al., 2002; Stafford et al., 2006). For neural and otic placode tissue transplants, embryos were injected at the 1-2 cell stage with 2.5% Rhodamine Alexa 568 dextran. At shield stage (6 hpf), donor cells were removed and placed into corresponding areas in fgf3;fgf8 morpholino-injected hosts using previously described fate maps (Kimmel et al., 1990 and Woo and Fraser, 1995). Mesoderm transplants were performed at shield stage (4 hpf) using donor tissue cells located at the margin and moved to the margins of fgf3;fgf8 morpholino-injected hosts (Kimmel et al., 1990). For endoderm transplants, donor 1 cell stage embryos were injected with a mixture of 2.5% Rhodamine Alexa 568 dextran and sox32 mRNA and donor cells located at the margin at shield were transplanted into the margin of fgf3;fgf8 morpholino-injected hosts (Stafford et al., 2006). For double mesoderm and neural transplants, donor tissue from 2.5% Rhodamine Alexa 568 dextran injected hosts was transplanted at both sphere (taking cells from the margin) and shield (taking cells from the neural ectoderm-forming region) into fgf3;fgf8 morpholino-injected hosts. At 24 hpf, all hosts were screened using a LeicaM216F fluorescence stereomicroscope for substantial and tissue-specific contributions of donor tissue for subsequent analysis.