Normal Function of Myf5 During Gastrulation Is Required for Pharyngeal Arch Cartilage Development in Zebrafish Embryos

Abstract

Myf5, a myogenic regulatory factor, plays a key role in regulating muscle differentiation. However, it is not known if Myf5 has a regulatory role during early embryogenesis. Here, we used myf5-morpholino oligonucleotides [MO] to knock down myf5 expression and demonstrated a series of results pointing to the functional roles of Myf5 during early embryogenesis: (1) reduced head size resulting from abnormal morphology in the cranial skeleton; (2) decreased expressions of the cranial neural crest (CNC) markers foxd3, sox9a, dlx2, and col2a1; (3) defect in the chondrogenic neural crest similar to that of fgf3 morphants; (4) reduced fgf3/fgf8 transcripts in the cephalic mesoderm rescued by co-injection of myf5 wobble-mismatched mRNA together with myf5-MO1 during 12 h postfertilization; (5) abnormal patterns of axial and non-axial mesoderm causing expansion of the dorsal organizer, and (6) increased bmp4 gradient, but reduced fgf3/fgf8 marginal gradient, during gastrulation. Interestingly, overexpression of fgf3 could rescue the cranial cartilage defects caused by myf5-MO1, suggesting that Myf5 modulates craniofacial cartilage development through the fgf3 signaling pathway. Together, the loss of Myf5 function results in a cascade effect that begins with abnormal formation of the dorsal organizer during gastrulation, causing, in turn, defects in the CNC and cranial cartilage of myf5-knockdown embryos.

Introduction

A vertebrate embryo must undergo multiple developmental stages to form a multicellular organism, including gastrulation, neurulation, and organogenesis. During gastrulation, the nodal and Wnt pathways are responsible for inducing two initial mesodermal domains: the axial mesoderm that constitutes the dorsal organizer and the remaining non-axial mesoderm.^1,2 The non-axial mesoderm can, in turn, be divided into three regions: paraxial, intermediate, and lateral mesoderm. The mechanisms that control axial and non-axial mesoderm identity during gastrulation are well defined.^3,4 Moreover, it is well known that the determination and differentiation of skeletal muscles from paraxial mesoderm are controlled by myogenic regulatory factors (MRFs), including Myf5, Myod, Myogenin, and MRF4.⁵ The roles these four MRFs play in myogenesis are well studied. However, we noticed that zebrafish myf5 is expressed in the paraxial mesoderm during 75% epiboly to bud stage when myogenesis is still not processed in embryos,^6,7 whereas myod is only expressed in the adaxial cells during 50% epiboly to bud stage.⁸ Thus, it is unclear whether MRFs play roles during early gastrulation.

It has been demonstrated that most vertebrate cranial and pharyngeal (viscerocranial) bones, and all pharyngeal cartilages, are derived from the neural crest.^9,10 Cranial neural crest (CNC) cells originate adjacent to the neural ectoderm and migrate toward the mandibular, hyoid, and branchial arches to form seven pharyngeal arches. Segmentation of the CNC cells into distinct streams is coupled with the segmentation of the hindbrain into rhombomeres (r1–7).^10–12 Craniofacial development requires an orchestrated communication between progenitor cells derived from the cranial paraxial mesoderm and CNC cells derived from the neural ectoderm.¹³ In frog and chick embryos, CNC cells can reorganize skeletal and muscle patterns when they are grafted to ectopic anterior–posterior levels.¹⁴ Rinon et al.¹⁵ proposed that CNC cells control craniofacial development by regulating positional interactions with mesoderm-derived muscle progenitors. TGF-β-mediated fibroblast growth factor 10 (FGF10) signaling in CNC cells can control the development of myogenic progenitor cells during tongue morphogenesis.¹⁶ In Pax3^−/− (splotch-delayed [Spd]) or Myf5^−/−:Myod^−/− double mutant mice, the morphology of the cranial cartilage, ribs, and joint was abnormal.^17–23 In contrast to zebrafish Myod^−/− single mutants, Myf5^−/−:Myod^−/− double mutants displayed more severe defects in cranial cartilage development where almost all ventral pharyngeal skeleton is lost.²⁴ Taken together, this evidence strongly suggested that (1) communication between CNC cells and mesoderm is required for normal development and (2) myf5 and myod might also play roles beyond their functions during myogenesis.²⁴ We previously reported that inhibition of myf5 transcription dramatically reduced krox20 and pax2.1 expression in hindbrain. Moreover, myf5 morphants have a severely reduced head size,²⁵ indicating that Myf5 is a crucial factor involved in head formation and hindbrain development. However, it is still unknown whether abnormal head size is caused by abnormal development of CNC cells when Myf5 is lost. Therefore, the molecular mechanism that underlies myf5 in the context of craniofacial development needs to be investigated in more detail.

To address these questions, we used myf5-morpholino oligonucleotides (MO) to knock down myf5 expression and demonstrated a cascade effect in which the absence of Myf5 leads to the loss of dorsal mesoderm organization, that is, axial and non-axial patterning, resulting in the failure of cell fate determination, followed by a series of developmental defects, including reduced head size resulting from abnormal morphology in the cranial skeleton and decreased expressions of foxd3, sox9a, dlx2, and col2a1, all markers of CNC cells. Additionally, expression of fgf3/fgf8 was greatly reduced in the cephalic mesoderm of myf5 morphants at 12 hours postfertilization (hpf ), eventually affecting patterning of pharyngeal pouches and resulting in the defective development of cranial cartilage.

Materials and Methods

Ethics statement

The animal use protocol listed below has been reviewed and approved by the National Taiwan University Institutional Animal Care and Use Committee (IACUC) with ethics approval number NTU-99-72. No specific ethics approval was required for this project, as all zebrafish (Danio rerio) used in this study were between 0 to 5 days old. Given the age of the embryos, pain perception has not yet developed at these earlier stages and so this is not considered as a painful procedure.

Transgenic zebrafish lines

The zebrafish embryos and larvae of AB strain, and transgenic lines Tg(α-actin:RFP),²⁶ Tg(fli1a:EGFP)²⁷, and Tg(-80Kmyf5:GFP),²⁸ were obtained and raised according to standard procedures.²⁹ Fluorescence was visualized with a fluorescent stereomicroscope (MZ FLIII; Leica) or a confocal spectral microscope (TCS SP5; Leica).

Fish embryos and whole mount in situ hybridization

The procedures for whole mount in situ hybridization were described previously.³⁰ The designations of zebrafish head muscle and cartilage follow the scheme of Schilling and Kimmel.³¹

Knockdown microinjection of zebrafish embryos

All MOs (Gene-Tools) were prepared at a stock concentration of 1 mM and were diluted to the desired concentrations, including 4, 2, or 1 ng, right before injection into each one- to two-cell stage embryos. MO sequences were as follows: myf5-MO1, TACGTCCATGATTGGTTTGGTGTTG³²; myf5-MO2, GATCTGGGATGTGGAGAATACGTCC²⁴; myod-MO, ATATCCGACAACTCCATCTTTTTTG²⁶; fgf3-MO, CATTGTGGCATGGCGGGATGTCGGC³³; and p53-MO, GCGCCATTGCTTTGCAAGAATTG ³⁴.

Detection of apoptotic cell death

Apoptosis assay was performed using the DeadEnd™ Colorimetric TUNEL System (Promega). Embryos were fixed in 4% paraformaldehyde (PFA) overnight at 4°C, and chorions were then removed and transferred into 100% methanol overnight at −20°C. The embryos were rehydrated thrice for 5 min with phosphate-buffered saline (PBS)+0.1% Tween 20 solution and covered with 100 μL of equilibration buffer for 10 min. Embryos were then incubated in 98 μL of equilibration buffer mixed with 1 μL of biotinylated nucleotide and 1 μL of TdT enzyme for 1 h at 37°C. These embryos were then washed thrice with 2× standard saline citrate solution for 5 min each and then with 2× PBS solution for 5 min each. The endogenous peroxidase was blocked by immersing the embryos in 0.3% hydrogen peroxide for 5 min. Then, the embryos were incubated in 100 μL of streptavidin horseradish peroxidase solution for 30 min. The embryos were washed thrice in PBS for 5 min each and then mixed with diaminobenzidine components in 100 μL of staining solution just before observation.

RNA injection

Capped mRNAs of fgf3, myf5, myod, and gfp were synthesized according to the protocol of the manufacturer (Epicentre). The myf5 mRNA used to carry out rescue experiments was designed to mutate at the wobble position of myf5 mRNA, in which seven nucleotides were mismatched to the corresponding nucleotides of myf5-MO1 without altering the amino acid residues. These mutated nucleotides within myf5 mRNA caused no effect of translational inhibition by myf5-MO1. The synthesized mRNAs were diluted to 0.44 ng/μL fgf3 mRNA, 44 ng/μL myf5 mRNA, 11 ng/μL myod mRNA, and 44 ng/μL gfp mRNA with distilled water. About 2.3 nL of mRNA was injected into each one-cell stage embryo.

Plasmid constructs

Heat-shock inducible expression plasmid pHsp-fgf3 was constructed, in which the production of fgf3 mRNA was controlled by heat-shock promoter (Hsp70). After microinjection of plasmid pHsp-fgf3 into the one-cell stage embryos, fgf3 mRNA was conditionally overexpressed starting at 6 hpf when embryos were treated at 37 °C for 10 min. Heat-shock induction was processed every 12 h until embryos developed at 36 or 120 hpf.

Cartilage staining

The Alcian blue staining protocol was modified from Neuhauss et al.³⁵ Embryos at 5 days postfertilization (dpf ) were anesthetized using 0.02% buffered tricaine (Sigma) and fixed overnight in 4% PFA at 4°C. After washing with PBS, embryos were stained overnight in 0.1% Alcian blue, which was dissolved in acidic ethanol (70% ethanol and 5% concentrated hydrochloric acid), washed extensively in acidic ethanol, dehydrated, and then stored in 80% glycerol. For better exposure of cartilage elements, embryos were digested with 0.02% trypsin.

Results

Effect of myf5 knockdown on head size, cranial cartilage formation, and cranial muscle development

The zebrafish transgenic line Tg(α-actin:RFP) is ideal for studying the dynamic changes of cranial muscle in embryos since all muscle of this line is tagged with RFP. Head muscle defects in myf5 and myod morphants were observed with RFP signal (Fig. 1A–C). In myf5 morphants, almost all head muscles were lost at 72 hpf (Fig. 1B). However, some of the cranial muscles were still present in the myod morphants (Fig. 1C), including extraocular muscles, such as superior oblique and inferior oblique muscles; dorsal pharyngeal muscles, such as levator arcus palatini, dilator operculi, adductor hyoideus, and adductor operculi; and trunk migratory head sternohyoideus muscle. By comparing the heads of MO1-treated embryos with those of wild-type (WT) embryos (Fig. 1D), we noticed that the myf5 and myod morphants had abnormal craniofacial structures (Fig. 1E, F). The head size of the Myf5 knockdown morphants was greatly reduced. Further, we used Alcian blue staining to detect the cranial cartilage for the WT (Fig. 1G, J), myf5 morphants (Fig. 1H, K), and myod morphants (Fig. 1I, L) at 5 dpf. We found that myf5 morphants had severe chondrodysplasia in craniofacial cartilage. Specifically, the derivatives of the branchial arches, such as basibranchial, ceratobranchial, and hypobranchial, were completely lost in myf5 morphants (Fig. 1H, K). Morphologically abnormal cartilage, such as Meckel's cartilage, and palatoquadrate, basihyal, and ceratohyal cartilage, were also observed (Fig. 1H, K). However, cranial cartilage development of the Myod knockdown embryos was only mildly affected. For example, Meckel's and palatoquadrate cartilages were developed abnormally, but they were not lost since all branchial arch-derived cartilages could be detected (Fig. 1I, L). Such results suggest that myf5 plays a more substantial role in cranial cartilage development than myod. We also examined the expression of col2a1, another chondrocyte marker, and found that it was correctly transcribed in the neurocranium of myf5 morphants; however, compared with WT (Fig. 1M, O), col2a1 was greatly reduced in the branchial arches, especially in ceratobranchial cartilages, at 72 and at 120 hpf (Fig. 1N, P). Thus, we concluded that the reduced head size in myf5 morphants was caused by the disruption of cranial cartilage development.

FIG. 1.

Inhibition of myf5 translation leads to abnormal cranial muscle formation and cartilage differentiation and reduced head size. WT embryos, morphants (myf5-MO1 and myod-MO), and stages (hpf ) are as indicated in each figure. (A–C) Viewed under fluorescent microscope; (D–P) viewed under a light microscope; (G–L) Alcian blue staining; and (M–P) using whole mount in situ hybridization to detect the transcripts of col2a1. (A) Tracing of head muscles at a late stage by using the zebrafish line Tg(α-actin:RFP). (B, C) Head muscles were dramatically reduced in both myf5 and myod morphants. Head size (D–F) and cartilage differentiation (G–L) were reduced in myf5-MO1-injected embryos (E, H, and K), but only mildly affected in myod morphants (F, I, and L). Five pairs of ceratobranchial (cb) arches (J and L; black arrows) were completely absent in the myf5 morphants (K). ah, adductor hyoideus; am, adductor mandibulae; ao, adductor operculi; cb, ceratobranchial; ch, ceratohyal; do, dilator operculi; dpw1–5, dorsal pharyngeal wall 1–5; hh, hyohyoideus; ih, interhyoideus; ima, intermandibularis anterior; imp, intermandibularis posterior; io, inferior oblique; ir, inferior rectus; lap, levator arcus palatini; m, Meckel's cartilage; mr, medial rectus; pq, palatoquadrate; sh, sternohyoideus; so, superior oblique; tv 1–5, transversus ventralis 1–5. hpf, hours postfertilization; WT, wild-type; MO, morpholino oligonucleotides.

The extent of the defective formation of cranial muscle and cartilage in the myf5 morphants was dependent on the amount of the myf5-MO1 injected. As shown in Table 1, of the embryos injected with 4 ng of myf5-MO1, 83% of the surviving embryos lost cranial muscle, and 82% of the embryos lost ceratobranchial cartilage, while for embryos injected with 0.5 ng of myf5-MO1, the percentage of embryos losing cranial muscle and ceratobranchial cartilage was 3% and 52%, respectively. Similar phenotypes were observed for the embryos injected with myf5-MO2, which was still capable of effectively knocking down myf5 translation (data not shown). We also examined the target specificities of myf5-MO1 and myf5-MO2. In the control group, we observed the GFP signal at 28 hpf when gfp mRNA was co-injected with either myf5-MO1 or myf5-MO2 at the one-cell stage (Supplementary Fig. S1A, C, respectively; Supplementary Data are available online at www.liebertpub.com/zeb). However, the GFP signal was absent when gfp mRNA containing a complementary site for myf5-MO1 was co-injected with myf5-MO1 (Supplementary Fig. S1B). Similarly, the GFP signal was also absent when gfp mRNA containing a complementary site for myf5-MO2 was co-injected with myf5-MO2 (Supplementary Fig. S1D). Thus, the MOs we used were target-specific. Additionally, when embryos were injected with myf5-MO1, the phenotypes were observed at developmental stages as early as 12–14 hpf and as late as 120 hpf. However, we found that these head defects could be rescued by injection of myf5 wobble-mismatched mRNA (Supplementary Figs. S2 and S3), suggesting that the defects during head development are indeed caused by the absence of Myf5. Importantly, cranial muscles and cartilages were still not observed as late as 8 dpf, indicating that the defects of myf5 and myod morphants did not result from developmental delay. Thus, these results suggest that Myf5, previously known as a regulator in muscle development, is also involved in the development of cranial cartilage.

Table 1.

The Percentages of Defective Phenotype Occurring in Cranial Muscle and Cartilage When myf5 Was Knocked Down

	Cranial muscle defect (%)a			Ceratobranchial cartilage defect (%)
myf5-MO1 injected concentration	Loss	Reduced	Normal	Loss	Reduced	Normal
Uninjected	0 (0/16)	0 (0/16)	100 (16/16)	0 (0/16)	0 (0/16)	100 (16/16)
0.1 ng	0 (0/35)	22 (8/37)	73 (29/37)	0 (0/37)	22 (8/37)	73 (29/37)
0.5 ng	3 (2/62)	76 (47/62)	21 (13/62)	52 (32/62)	34 (21/62)	14 (9/62)
2 ng	19 (8/43)	81 (35/43)	0 (0/43)	79 (34/43)	21 (9/43)	0 (0/43)
4 ng	83 (55/66)	17 (11/66)	0 (0/66)	86 (57/66)	14 (9/66)	0 (0/66)

^aCranial muscle defect was observed at 72 hpf, except medial rectus, inferior rectus, and superior rectus muscles, which was normal in myf5-MO1-injected embryos. Cartilage defect was observed after embryos were fixed and staining was performed.

hpf, hours postfertilization.

Myf5 is required for activation of genetic programs that define the chondrogenic neural crest cells

Neural crest cells are known to pattern the muscles of the pharyngeal region¹⁴ The pharyngeal cartilages are derived from migrating CNC cells. The appearance of pharyngeal arches is a characteristically morphogenic event during the pharyngeal period. To determine whether Myf5 knockdown affects pharyngeal cartilage development, we used (1) Tg(fli1a:EGFP), in which the pharyngeal arch neural crest expressed GFP and (2) Zn8 antibody to label the endoderm pouch³⁶. When compared with WT embryos at 32 hpf (Fig. 2A–C), only the first and second arches were formed in myf5 morphants, while the third to fifth pharyngeal arches were lost (Fig. 2D–F). We then examined the neural crest marker dlx2 in myf5 morphants. In 24 hpf WT embryos, the expression of dlx2 was observed in three distinct streams of CNCs that migrate to the mandibular (I; 1), hyoid (II; 2), and two branchial arches (III; 3, 4; Fig. 2G, I). In myf5-MO1-injected embryos, dlx2 was normally expressed in the mandibular and the hyoid arches, but it was slightly reduced in the first and the second branchial arches (Fig. 2H, J). However, compared with WT (Fig. 2L, N), dlx2 expression was greatly reduced in the branchial arches of the myf5-MO1-injected embryos at 36 hpf, resulting in the development of only four pairs of postmigratory crest (Fig. 2M, O), instead of six pairs of crest that had already developed in WT embryos at this stage of development.

FIG. 2.

Myf5 knockdown embryos failed to develop pharyngeal arch structure and exhibited reduced prechondrogenic gene expression in cranial neural crest cells. Tg( fli1a:EGFP) indicates the neural crest (green) at the pharyngeal arch region, and Zn8 indicates the endoderm pouch (red). In 32 hpf WT embryos, five pharyngeal arches can be observed (A–C). In myf5-morphants, only the first and second arches were detected. The third to fifth arches were lost in myf5 morphants (D–F). At 24 and 36 hpf, the prechondrogenic gene dlx2 was expressed at the third arch stream in WT embryos. However, dlx2 was reduced significantly at the third arch stream in myf5 morphants (G–J, L–O). Co-injection of myf5-MO1 and p53-MO did not rescue the reduced expression of dlx2 at the third arch stream (K, P). I, mandibular arch; II, hyoid arch; III, branchial arch; 1–6, pharyngeal arch; arrows in (D) indicate the lost of neural crest; brackets in (G), (H), (L) and (M) indicate the branchial arch.

To rule out the possibility that the craniofacial defects we observed were caused by nonspecific MO toxicity,³⁷ we co-injected p53-MO with myf5-MO1 and found that the reduced dlx2 expression was not rescued in myf5 morphants, either at 24 or at 36 hpf (Fig. 2K, P). These results indicate that the reduced dlx2 signals in myf5 morphants were not caused by nonspecific toxicity. Therefore, our results suggest that Myf5 is essential for cartilage formation derived from migrating chondrogenic crest cells, especially the third CNC stream: branchial arches.

fgf3 knockdown results in defects similar to those of myf5 knockdown

Fgfs belong to a family of extracellular signaling molecules involved in diverse facets of craniofacial development.^33,36 Interestingly, myf5 morphants displayed defects similar to those of the fgf3 morphants, such as lost branchial cartilage.³³ To determine whether Myf5 is involved in Fgf3 signaling during pharyngeal cartilage development, we compared the defects found in myf5 morphants with those found in fgf3 morphants. After Alcian blue staining of cartilage, we found that the defects of myf5 and fgf3 morphants were similar. Specifically, compared with WT (Fig. 3A), the cartilage for pharyngeal arches 3–7 was lost at 120 hpf (Fig. 3B, C). We then detected expression of ERK-inducible transcription factors pea3 and erm, which are completely Fgf-dependent during the early stages of zebrafish development.^38,39 We found that (1) pea3 transcripts were detected at the precursor cells of CNC located at pharyngeal arches from two to four at 24 hpf (Fig. 3D) and (2) erm transcripts were expressed at the precursor cells of CNC located at pharyngeal arches from one to four at 24 hpf (Fig. 3G). However, pea3 and erm transcripts were hardly detectable in the posterior part of pharyngeal arch four in either myf5 or fgf3 morphants (Figs. 3E, F, H, I). Therefore, it is likely that Myf5 modulates craniofacial cartilage development via the fgf3 signaling pathway.

FIG. 3.

The defects in the chondrogenic neural crest cells of Myf5 knockdown embryos were similar to those of fgf3 morphants. WT embryos, morphants (myf5-MO1 and fgf3-MO), embryonic stages (hpf ), and in situ hybridization probes are as indicated in each figure. (A–C) Ventral view of the cranial cartilage after embryos at 120 hpf were stained with Alcian blue. Compared with WT (A), the branchial cartilage did not develop normally in myf5 and fgf3 morphants, and abnormal morphology can be observed in ceratohyal cartilage (ch), Meckel's cartilage (m), and palatoquadrate cartilage (pq) (B, C). At 24 hpf, pharyngeal arch 4 of pea3- and erm-positive neural crest cells was reduced in myf5 and fgf3 morphants (D–F, G–I). Pharyngeal arches 1–4 are labeled in (D–I). FGF, fibroblast growth factor.

Knockdown of myf5 leads to the loss of fgf3/8 expression at the cephalic mesoderm, but fgf3 mRNA can rescue the craniofacial cartilage defects caused by myf5 knockdown

Because the defects in myf5 morphants are similar to those of fgf3 morphants, we hypothesized that Myf5 and Fgf3 may function in the same signaling pathway. To test this hypothesis, we investigated the molecular interaction between Myf5 and Fgf3. Our results showed fgf3/fgf8 expression in the cephalic mesoderm at 12 hpf (Fig. 4A, E), but it was lost in embryos either injected with myf5-MO1 alone or injected with myf5-MO1 plus p53-MO (Fig. 4B, C, F, G). However, the defective expression of fgf3/fgf8 in the cephalic mesoderm of myf5 morphants could be rescued by injection of wobble-altered mRNA of myf5 (Fig. 4D, H). These data suggest that the expression of fgf3/fgf8 at the cephalic mesoderm is dependent on the normal expression of myf5 during early embryogenesis. It has been shown that Fgf signaling from the cephalic mesoderm is required for the induction of pharyngeal pouches and the specification of epibranchial placodes.^36,40 Thus, we further proposed that myf5 deficiency could lead to craniofacial cartilage defect via reduced fgf3/8 signaling from the cephalic mesoderm.

FIG. 4.

The defective expressions of fgf3, fgf8, foxd3, and sox9a in myf5-knockdown embryos could be rescued by injection of myf5-MO1 plus either myf5 mRNA or fgf3 mRNA. Dorsal view of the expression patterns of fgf3 (A–D), fgf8 (E–H), foxd3 (I–M), and sox9a (N–R) in head region of embryos at 12 hpf. In WT embryos, fgf3 (A) and fgf8 (E) were used to label the cephalic mesoderm (brackets); foxd3 was used to label the premigratory neural crest (I); and sox9a was used to label the premigratory neural crest and optic region (N, brackets). In myf5 morphants, fgf3 and fgf8 in the cephalic mesoderm were absent (B, F); foxd3 (J) and sox9a (O) were greatly reduced. Co-injection of p53-MO could not rescue the absent expressions of fgf3 (C) and fgf8 (G) at the cephalic mesoderm or the reduced expressions of foxd3 (K) and sox9a (P) in the neural crest caused by myf5-knockdown. However, injection of wobble-mismatched mRNA of myf5 could restore the expressions of fgf3 (D), fgf8 (H), foxd3 (L), and sox9a (Q) in myf5-morphants. Embryos were injected with myf5-MO1 plus plasmid pHsp-fgf3 and then treated with heat shock to induce fgf3 expression at 6 hpf, followed by analysis of foxd3 and sox9a expressions at 12 hpf. Similar to WT embryos, the expressions of foxd3 (M) and sox9a (R) in the neural crest and optic region were normal. The numbers shown in the lower-right corner of panels (A–R) indicate the number of phenotypes out of the total number of embryos examined. Brackets in (A–H) indicate the cephric mesoderm; brackets in (N–R) indicate the otic vesicle.

Next, we analyzed the patterns of such CNC markers as foxd3 and sox9a during early embryogenesis. Compared with WT (Fig. 4I, N), the expressions of foxd3 and sox9a at the premigratory neural crest and optic region were greatly reduced in the embryos injected either with myf5-MO1 alone or with myf5-MO1 plus p53-MO (Fig. 4J, K, O, P). However, when embryos were injected with either myf5-MO1 plus myf5 wobble-mismatched mRNA or myf5-MO1 plus plasmid pHsp-fgf3 and then heat-shocked to induce fgf3 expression at 6 hpf, we found that the expressions of foxd3 and sox9a were restored at 12 hpf (Fig. 4L, M, Q, R), indicating that the reduced expression of CNC markers in the myf5 morphant is dependent on fgf3 signaling.

Previous studies revealed that the migrating CNC cells undergo apoptosis in the absence of Fgf signaling.³³ Compared to untreated embryos, the apoptotic signals were increased throughout the head region in the myf5-MO1-injected embryos at 30 hpf (Fig. 5A vs. B and 5G vs. H). In contrast, when we co-injected myf5 mRNA with myf5-MO1, the number of apoptotic signals decreased significantly compared with the occurrence of apoptosis in embryos injected with myf5-MO1 only (Fig. 5C, I). To understand whether the apoptotic cells in the head of myf5-morphants were CNC cells, we employed embryos derived from transgenic line Tg(−4.9 sox10:EGFP), in which CNC cells were labeled in green. After myf5-MO1 was injected into embryos, the TUNEL assay was performed to detect the amount of apoptotic signals labeled in red. We found a high frequency of yellow signals detected in the myf5-MO1-injected embryos, indicating that CNC cells undergo apoptosis in the absence of myf5 (Supplementary Fig. S4). These findings suggest that apoptosis plays a significant role in the reduction of head size and that myf5 inactivation results in the loss of pharyngeal cartilage formation.

FIG. 5.

Overexpression of Fgf3 can rescue the craniofacial defects caused by myf5 knockdown. Dorsal (A–F) and lateral (G–L) views of WT embryos and embryos injected either with myf5-MO1 alone or with myf5-MO1 plus myf5 mRNA, fgf3 mRNA, gfp mRNA, or myod mRNA. The TUNEL assay was evaluated at 30 hpf. Compared with WT embryos (A, G), apoptotic signals significantly increased in myf5 morphants (B, H). However, the apoptotic signals were significantly reduced in the embryos injected with myf5-MO1 plus either myf5 mRNA or fgf3 mRNA, compared with apoptotic signaling in the myf5-MO1-injected embryos (C, D, I, J). Embryos injected with myf5-MO1 plus either gfp mRNA or myod mRNA did not exhibit any reduction of apoptotic signaling (E, F, K, L). Additionally, compared with WT (M), the expression domain of dlx2 in the embryos injected myf5-MO1 plus p53-MO was reduced (N), but this defect could be rescued by injection of myf5-MO1 plus either myf5 mRNA or plasmid pHsp-fgf3 (O, P). Similarly, compared with WT (Q), ceratobranchial cartilage in the embryos injected with myf5-MO1 plus p53-MO was absent (R). This defect could be rescued by injection of myf5-MO1 plus either myf5 mRNA or pHsp-fgf3 (S, T). (U) The diagram illustrates the percentages of embryos with defective ceratobranchial cartilage among all the surviving embryos injected with myf5-MO1 (4 ng) alone or injected with p53-MO (8 ng), fgf3 mRNA (0.1-, 0.5-, 1.0-, and 2.5-pg), myf5 mRNA (100 pg), or pHsp-fgf3 (23- and 58-pg) at 120 hpf. Loss: ceratobranchial cartilage was all lost; Reduced: at least one of five pairs, but fewer than five pairs of ceratobranchial cartilage were present; WT-like: all five pairs of ceratobranchial cartilage were present. The numbers shown at upper-right corner of panels (A–T) indicate the number of phenotypes out of the total number of embryos examined. Pharyngeal arches 1–6 are labeled in (M–P). Arrows in (Q), (S) and (T) indicate the presence of ceratobranchial cartilages.

Next, to understand whether reduced Fgf signaling in myf5-knockdown embryos causes CNC cells to undergo apoptosis, we injected myf5-MO1 plus fgf3 mRNA. Results showed that the apoptotic signaling was reduced (Fig. 5D, J). However, the number of apoptotic cells remained unchanged in the embryos injected with myf5-MO1 plus either gfp mRNA (Fig. 5E, K) or myod mRNA (Fig. 5F, L). Additionally, even though overexpression of fgf3 mRNA in myf5 morphants resulted in reduced apoptotic signals, it also rescued the expression of dlx2 (Fig. 5M vs. P). Similar to WT embryos, the six pharyngeal arches were observed in the embryos injected with myf5-MO1 plus plasmid pHsp-fgf3 (Fig. 5M, P). However, only four dlx2-positive pharyngeal arches were observed in embryos injected with myf5-MO1 plus p53-MO (Fig. 5N). Nevertheless, this defective expression of dlx2 could be restored by co-injection of myf5-mRNA (Fig. 5O). At late stage, the cranial cartilage defects caused by myf5-knockdown were rescued by overexpression of myf5 and fgf3, while in embryos injected with myf5-MO1 plus p53-MO, these cartilage defects remained unchanged (Fig. 5Q–T). The percentage of each group of embryos with rescued ceratobranchial cartilage was shown in Figure 5U. We noted that the occurrence of ceratobranchial cartilage defects caused by injection of 4 ng myf5-MO1 was decreased when embryos were co-injected with 0.1 to 2.5 pg of fgf3 mRNA, 100 pg myf5 mRNA, or 23 to 58 pg of pHsp-fgf3. On the basis of these findings, we concluded that Myf5 modulates craniofacial development and that the fgf3 signaling pathway is involved.

Myf5 is required for Spemann organizer patterning during gastrulation

While the above data clearly indicate that Myf5 specifically modulates craniofacial development, myf5 transcripts were not detected in the head during the early segmentation to Prim-5 stage. To determine whether myf5 exists prior to segmentation, we analyzed the expression of myf5 during early to late gastrulation. Our results demonstrated that myf5 was expressed in the non-axial mesoderm at 50–60% epiboly, the shield stage when involution of the first prospective mesendodermal cells leads to the formation of axial and non-axial mesendodermal tissue (Fig. 6A, B). Interestingly, myf5 was only expressed on both sides of the Spemann organizer. Double in situ hybridization of myf5 and floating head ( flh) revealed that the strip pattern of myf5 bordered on the Spemann organizer (Fig. 6F). At 7–9 hpf, myf5 was expressed in the paraxial mesoderm segmental plate and adaxial mesoderm (Fig. 6C, D). Nevertheless, at shield stage, myod was only expressed in a few cells on either side of the Spemann organizer (Fig. 6G). At 7–9 hpf, myod was expressed in the adaxial cells, but not in the lateral plate mesoderm (Fig. 6H, I). At 13 hpf, both myf5 and myod were strongly expressed in the somites and adaxial cells (Fig. 6E, J).

FIG. 6.

Non-axial mesoderm cells that express myf5 during early gastrulation contributed to cranial mesodermal cells at later developmental stages. Myf5 (A–E) and myod transcripts (G–J), as evaluated by whole mount in situ hybridization. The stages of development (hpf ) are as indicated in the upper right of each figure. Images are vegetal pole view (A) or dorsal view (B–D, F–I) and posterior view (E, J). Myf5 expression was first observed in the non-axial mesoderm at the shield stage (A, B), followed by the paraxial mesoderm segmental plate and adaxial mesoderm at 7–9 hpf (C, D) and then in the somites, adaxial cells, and presomitic mesoderm at 13 hpf (E). Double in situ hybridization of myf5 and flh transcripts was also performed (F), revealing that the strip pattern of myf5 bordered on the Spemann organizer. Meanwhile, myod was first expressed in a few cells on both sides of the organizer at the shield stage (G), followed by the adaxial cells at 7–9 hpf (H, I) and then in the somites and adaxial cells at 13 hpf (J).

Next, we investigated whether knockdown of myf5 would cause abnormal gene expression during early gastrulation. Tbx1, which is required for interactions between the cranial mesoderm and the neural crest cells that form the pharyngeal cartilages,⁴¹ was expressed at the paraxial mesoderm in WT during shield stage (Fig. 7A, B). When myf5 was knocked down, tbx1 transcripts at the paraxial mesoderm were reduced (Fig. 7C). Interestingly, with the knockdown of tbx1, the myf5 transcript was also reduced (Fig. 7D, E), suggesting that the interaction between myf5 and tbx1 is likely a positive feedback mechanism. Additionally, knockdown of myf5 resulted in an ectopic expansion of flh-positive cells in the notochord (Fig. 7F–I). This expansion of axial mesoderm was also demonstrated by the expansion of gsc expression, which marks the prechordal plate mesoderm (Fig. 7K, L). In addition, the broadened dorsal clearing of her1 expression was also observed (Fig. 7M, N). Knockdown of myod did not affect the patterning of the dorsal organizer (Fig. 7J). Further, co-injection of myf5 mRNA led to the rescue of the phenotype caused by myf5-MO1 injection, suggesting that the myf5-MO1-induced phenotype was specific (Fig. 7O).

FIG. 7.

Myf5 knockdown affected the expression domain of axial and non-axial mesoderm during gastrulation. Whole mount in situ hybridization of tbx1 (A–C), myf5 (D, E), flh (F–J, O), gsc (K, L), and her1 (M, N) was performed in WT, tbx1 morphant, myf5 morphant (myf5-MO1), and the myod morphant (myod-MO) at shield stage (A–C, F–G, K–O) and 8 hpf (D, E, H–J), as seen in these dorsal views. Tbx1 expression in the paraxial mesoderm was decreased in myf5 morphants (A–C); myf5 expression was also decreased in the paraxial mesoderm when tbx1 was knocked down (D, E). Noticeable expansion of flh-expressing (G, I; arrow) and gsc-expressing (L; arrow) cells was observed in myf5 morphants. The expression of her1 displayed a wider dorsal clearing (N; arrow) when compared with the WT embryos (M vs. N). In contrast, the expression of flh appeared normally in the myod morphant (J). The expression of flh became normal in embryos injected with myf5 mRNA plus myf5-MO1 (O).

It is well known that the neural crest is induced by a combination of secreted signals and that neural crest induction is achieved by combining the same classes of molecules as those required for neural plate patterning: Bmp, Wnts, and Fgfs.^42–45 Therefore, we analyzed whether the signaling of Bmp, Wnts, and Fgfs was affected in the absence of myf5 during early embryogenesis. Compared to WT embryos (Fig. 8A), the expression domain of bmp4 was expanded from the ventral domain to the dorsal domain when myf5 was knocked down (Fig. 8F). While fgf3 was reduced at the dorsal domain (Fig. 8B vs. G), the fgf3 marginal gradient was lost at the ventral region in myf5 morphants (Fig. 8C vs. H). The expanded expression of fgf8 at the dorsal organizer might have resulted from a change in the border of the dorsal organizer, and apparent decrease in the fgf8 marginal gradient (Fig. 8D vs. I and 8E vs. J). Since the Fgf signals antagonize ventralizing Bmp signals, our data showed that myf5 function may be involved in balancing the gradient of Fgf and Bmp signals during gastrulation. Taken together, our data revealed that Myf5 plays an important role in mesoderm patterning and that normal expression of myf5 from non-axial mesoderm is required for Spemann organizer patterning during gastrulation.

FIG. 8.

Signaling of both fgf and bmp was affected in the myf5-knockdown embryos during gastrulation. Using whole mount in situ hybridization to examine the expression patterns of bmp4 (A, F), fgf3 (B, C, G, H), and fgf8 (D, E, I, J) in WT embryos (A–E) and myf5 morphants (myf5-MO1; F–J) at shield stage. Compared to WT embryos (A), the expression of the ventral marker bmp4 was upregulated at the ventral region (v) and extended to the dorsal part in myf5-knockdown embryos at shield stage (F). In contrast, compared with WT embryos (B, C), the expression of fgf3 in the dorsal organizer (do) was reduced (G, H), and the expression of fgf8 in the dorsal region (do) was also reduced (D, E vs. I, J). The borders of fgf3 and fgf8 at the dorsal organizer were significantly extended (arrows mark the size of dorsal organizer).

Discussion

Distinct functions of Myf5 and Myod in craniofacial development in zebrafish

Myf5 and Myod have been intensively studied in various species and are known to play different roles during trunk myogenesis. The regulation of Myf5 is more complex and is markedly different from that of Myod.^46–49 Recently, we demonstrated that Myf5 and Myod exhibit distinct functions during craniofacial myogenesis in zebrafish.^26,32 Other than myogenesis, however, the roles that myf5 and myod play during embryogenesis have thus far been unclear. In this study, we demonstrated that Myf5 is necessary for cranial cartilage development in zebrafish. Specifically, severe chondrodysplasia in craniofacial cartilage and greatly expanded Spemann organizer at the shield stage were observed in myf5 morphants, but not in myod morphants. However, overexpression of fgf3 could rescue the cranial cartilage defects caused by myf5-MO1, suggesting that Myf5 modulates craniofacial cartilage development through the fgf3 signaling pathway. Thus, we propose that Myf5 plays roles in craniofacial development during zebrafish embryogenesis in a manner distinct from that of Myod.

Previous studies revealed that the lack of myogenesis in head muscles could lead to skeletal defects in mouse and chick.^17–22 However, our present data demonstrated that almost all the cranial muscles are lost in myod morphants, while cranial cartilage is still formed, although they displayed abnormal shapes. This evidence is consistent with the result obtained from myod mutants.²⁴ It is also true that defective cranial skeletogenesis, which is observed in myod and myf5 double mutants,²⁴ is very similar to the myf5-knockdown phenotype reported in this study. Therefore, our data indicate that Myf5, not Myod, is the key factor mediating cranial cartilage development. Additionally, we found when Myf5 is lost during gastrulation, the cranial mesoderm cells fail to secrete the Fgf3/8, resulting in abnormal craniofacial development. Similar mechanism is also observed in mouse and chick tendon formation.^50,51 Taken together, we highlight an alternative mechanism that musculature is not the key factor for cranial skeletogenesis. Instead, Myf5 is the key factor for cranial skeletogenesis.

Adequate expression of myf5 during gastrulation determines precisely distributed patterning of axial and non-axial mesoderm

Zebrafish gastrulation starts with the involution/ingression of prospective mesendodermal cells.⁵² At the shield stage, involution of the first prospective mesendodermal cells leads to the formation of axial and non-axial mesendodermal tissue. In this study, we detected myf5 transcripts in the prospective non-axial mesoderm (Fig. 6) at the shield stage. Knockdown of Myf5 resulted in a significant expansion of flh and gsc expression domains in the dorsal organizer (Fig. 7). This evidence strongly supports the hypothesis that Myf5 plays important roles during gastrulation.

In zebrafish, the paraxial mesoderm is derived from two symmetrical territories of the marginal zone located on either side of the Spemann organizer. Kodjabachian et al.⁵³ proposed a model representing the interaction network between axial and paraxial cells. In this model, ntl blocks spt expression in notochord precursors by inducing flh, which consequently represses spt. Dal-Pra et al.⁷ demonstrated that the knockdown of FoxA2 and FoxA3 results in dorsal expansion of myf5 and that the expression area of myf5 could even be fused along the axis. In the present study, we found that Myf5 knockdown resulted in the expansion of flh expression in axial mesoderm (Fig. 7). We also noticed that the expression of ntl increased in the myf5-MO1-injected embryos (data not shown). The expansion of flh and the widened dorsal clearing of her1 expression indicate that the non-axial mesoderm was pushed away by axial mesoderm, suggesting that the non-axial mesoderm might have lost its ability to restrict the distribution of the axial mesoderm, resulting in expansion of the Spemann organizer. Therefore, we concluded that an adequate spatiotemporal expression of myf5 in the myf5-positive non-axial mesoderm is required to maintain normal patterns of the axial and non-axial mesoderms during gastrulation.

Additionally, it is known that Fgf signals may antagonize ventralizing Bmp signaling during gastrulation.⁵⁴ While our data also showed increased bmp4 gradient from the ventral site to the dorsal site, we also found a reduced fgf3/fgf8 marginal gradient during gastrulation in the myf5-knockdown embryos. These results suggest that Myf5 in the non-axial mesoderm is important for mediating the balance between the gradient distribution of Fgf and Bmp signaling during gastrulation. Thus, in non-axial mesoderm during gastrulation, these observations, when taken together, suggest that myf5 functions to both restrict the expansion of axial mesoderm and mediate the balance of Fgf and Bmp gradients.

Normal patterns of axial and non-axial mesoderms determine cell fate of embryos

Multiple nodal-related genes are necessary for induction of mesoderm and endoderm.^55–57 In zebrafish, two nodal-related genes, squint (sqt) and cyclops (cyc), have been defined.^58–62 In sqt^–/– and cyc^–/–double mutants, almost all mesendodermal derivatives are lost, including notochord, trunk somites, pronephros, heart, blood, and gut.⁵⁹ In sqt^–/– and cyc^+/–embryos, widened dorsal clearing of sqt expression is observed. Fate map experiments revealed that the dorsal marginal cells in sqt^–/– and cyc^+/–mutants adopt neural fates, including spinal cord, hindbrain, and midbrain.⁶³ This evidence suggests that precisely distributed patterning of the axial and non-axial mesoderms during gastrulation is required for embryonic cell fate determination. Interestingly, in this study, we found that the absence of myf5 resulted in abnormal development of both axial and non-axial mesoderms. Thus, the failure in proper patterning initiated a cascade that interferes with cell fate determination and causes a host of embryonic defects, as detailed above. We therefore propose that myf5 is an important gene for maintaining correct mesoderm patterning in the non-axial mesoderm during early gastrulation.

Knockdown of Myf5 causes the reduced expression of fgf3 in cephalic mesoderm, resulting in defective development of craniofacial cartilage in embryos

The pharyngeal arches are unique transient structures that give rise to much of the head and neck skeleton.¹⁴ They are populated early in development by CNC cells originating from the posterior midbrain–hindbrain region in a precise rostral–caudal pattern.¹⁰ Considerable evidence indicates that Fgf signals from the cranial mesoderm are important for pharyngeal endoderm formation. Crump et al.³⁶ revealed that Fgf8 and Fgf3 are required in the neural keel and cranial mesoderm during early somite stages (10–14 somite stage) to promote first pouch formation. Nechiporuk et al.⁴⁰ also reported that mesoderm-derived Fgf3 and Fgf8 signals establish both the epibranchial placodes and development of the pharyngeal endoderm.

The zebrafish fgf3 mutant limabsent (lia) shows reduction of almost all cartilage of the gill arches (pharyngeal arches 3–7) and dlx2-positive chondrogenic neural crest cells of arches 3–7.⁶⁴ Analysis of the fgf3 knockdown demonstrated that the pharyngeal endodermal expression of fgf3 is specifically required for survival of skeletogenic neural crest cells to form the posterior pharyngeal arches.^65,66 In this study, we demonstrate that Myf5 is required for pharyngeal cartilage formation. Inhibition of Myf5 function results in the defective development of craniofacial cartilage, especially in the branchial arch derivatives. Interestingly, we found that fgf3 and fgf8 at the cephalic mesoderm were significantly reduced in myf5 knockdown embryos during 12 hpf. Additionally, cranial defects in the myf5 morphant were similar to those observed in the fgf3 morphants. Also, similar to fgf knockdown embryos, endoderm pouch formation in pharyngeal arches was lost in myf5 morphants, as revealed by immunostaining.³⁶ Whole mount in situ hybridization also revealed that the expressions of fgf3, erm, and pea3 are all reduced in myf5 morphants. Further, although results from the TUNEL assay indicate that apoptotic signals are abundant in myf5, but not myod, morphants (Fig. 5), co-injection of fgf3 mRNA with myf5-MO1 is sufficient to rescue the ceratobranchial cartilage defects and reduce the apoptosis. Taken together, we proposed that the defects in cranial cartilage formation in myf5 morphants were caused by reduced Fgf signaling from the cephalic mesoderm.

Hinits et al.²⁴ reported that the cranial muscles of myod-null mutant zebrafish (myod^fh261) were greatly reduced or entirely absent. Nonetheless, all major cranial cartilages were presented at 6 dpf in the myod^fh261 mutant, albeit in the absence of normal cranial cartilage morphogenesis. As shown in Figure 1I and L, these results are consistent with those presented in this study because we found that all branchial arch-derived cartilages could be detected in myod-knockdown embryos, although Meckel's and palatoquadrate cartilages were developed abnormally. However, it should be noted that Hinits et al.⁽²⁴⁾ also demonstrated that myf5^hu2022 and myod^fh261 double mutant fish exhibited severe cranial skeletal defect, suggesting that Myf5 is necessary for cranial cartilage formation. Thus, the myf5-knockdown phenotypes, as presented here by injection of MO1 and MO2 into embryos, are plausible outcomes when Myf5 is absent in zebrafish embryos. While our MO-knockdown approach resulted in an observable defective phenotype, Hinits et al.⁶⁷ reported no phenotype in myf5-knockdown embryos. We might attribute the apparent differences to the effectiveness and concentration of MO they used. Additionally, although Hinits et al.⁶⁷ concluded that the myf5-null mutant also had no observed defects, these mutants were all lethal and could not grow to adulthood. Similarly, Giraldez et al.⁶⁸ previously reported that Zdicer mutants had no obvious defects at 7 to 10 dpf and that these mutants were lethal after 10 dpf. Interestingly, Giraldez et al.⁶⁸ also generated maternal-zygotic dicer (MZdicer) mutants that displayed morphogenesis defects, indicating that maternal mRNAs might play a role in development. Based on this evidence, reasoning arises that might explain why the myf5-null mutant had no observed defects in the Hinits et al.⁶⁷ model. Specifically, mutant embryos contain maternal myf5 mRNA to sufficiently promote normal embryonic growth at least until 5 dpf. However, when we used the MO-knockdown approach, myf5-MO inhibited maternal myf5 mRNA translation, resulting in an observable phenotype.

In summary, this article is the first report showing that Myf5, a well-known MRF involved in myogenesis, functions in gastrulation and plays roles in head mesoderm patterning, CNC survival, and cranial cartilage formation. This study opens up avenues to further investigate the possible regulatory network underlying the involvement of Myf5 in the cranial development of zebrafish.

Acknowledgments

We are grateful to Ms. Yi-Chun Chuang and Ms. Ya-Chan Yang, College of Life Science, NTU, for helping with the confocal laser scanning microscopy. This study was supported by the National Science Council of the Republic of China under NSC grant 100-2313-B-002-043-MY3 (HJ Tsai) and 100-2321-B-002-067-MY2 (CY Lin).

Disclosure Statement

No competing financial interests exist.