Oesophageal and sternohyal muscle fibres are novel Pax3-dependent migratory somite derivatives essential for ingestion

Energy acquisition in jawed vertebrates is accompanied by two processes: opening of the mouth and subsequent passage of food to the gastrointestinal tract. Two groups of striated muscles facilitate these processes. First, the hyoid musculature, including sternohyoid muscle (SHM), is essential for mouth opening. Second, oesophageal striated muscle (OSM) permits voluntary swallowing. The requirements for both SHM and OSM for efficient feeding are demonstrated by the inability to suckle after SHM fatigue (van Lunteren and Moyer, 2003) and by malnutrition in humans with oesophageal dysphagia (Perry, 2001). Given the physiological importance of OSM and SHM, it is surprising that their cellular origin and development are ill-defined.

Unlike the rest of the gut, the oesophagus of fish and mammals acquires a striated muscle phenotype correlating with voluntary feeding (Domeneghini et al., 1999; García Hernández et al., 2001; Wörl and Neuhuber, 2005; Carrassón et al., 2006; Elworthy et al., 2008; Katori et al., 2010). In mice, two layers of OSM appear sequentially during the foetal period among earlier-formed smooth muscle cells at the anterior end of the lengthening oesophagus (Fig. 1J). The source of OSM is controversial, with multiple hypotheses put forward, including a smooth-to-striated muscle transition (Patapoutian et al., 1995; Kablar et al., 2000; Stratton et al., 2000) or distinct smooth and striated muscle lineages (Rishniw et al., 2003; Wörl and Neuhuber, 2005). In fish and mammals, the SHM attaches to the pectoral girdle (clavicle in mammals, cleithrum in teleosts) and the hyoid bone, thus enabling complex movements associated with food acquisition, processing and swallowing (Cubbage and Mabee, 1996; Diogo et al., 2008; Konow et al., 2010). Although Noden (Noden, 1983) suggests that muscles in the SHM region derive from somites and gene expression in zebrafish supports this view (Schilling and Kimmel, 1997; Neyt et al., 2000; Lin et al., 2006), the embryological origin of SHM has not been proven.

Undifferentiated, hypaxial somite-derived, migratory muscle precursors (MMPs) form all limb, tongue and diaphragm striated muscles (Noden, 1983; Dietrich et al., 1998; Dietrich, 1999). Pax3/7 transcription factors regulate MMP-derived muscle fate (Buckingham and Relaix, 2007). Klein-Waardenburg syndrome (OMIM #148820) is caused by PAX3 mutation and is characterised by reduced MMP-derived striated muscle (Goodman et al., 1982). Similarly, the Pax3 mouse mutant splotch (Sp), has greatly reduced MMP-derived striated muscle (Bober et al., 1994; Tajbakhsh et al., 1997; Tremblay et al., 1998; Borycki et al., 1999; Buckingham and Relaix, 2007; Zhou et al., 2008). pax3/7 genes are expressed in zebrafish somites (Groves et al., 2005; Devoto et al., 2006; Hammond et al., 2007; Minchin and Hughes, 2008) and MMP-derived myogenesis is morphologically and molecularly conserved with amniotes (Bladt et al., 1995; Brohmann et al., 2000; Gross et al., 2000; Neyt et al., 2000; Haines et al., 2004; Ochi and Westerfield, 2009). Although Pax7 regulates myoD expression in Xenopus somites (Maczkowiak et al., 2010), a role for Pax3 during myogenesis has not been elucidated beyond mammals, so the ancestral role of Pax3 in myogenesis is unclear. Here, we determine that Pax3-expressing somite cells give rise to OSM and SHM, a process that is Pax3b dependent in zebrafish and is required for efficient ingestion and swallowing.

Pax3^Cre [Pax3^tm1(cre)Joe] (Engleka et al., 2005), inducible Pax7^CE [Pax7^{tm2.1(cre/ERT2)Fan}] (Lepper et al., 2009), Wnt1^Cre [Tg(Wnt1-cre)11Rth] (Danielian et al., 1998), R26R^EYFP [Gt(ROSA)26Sortm1^(EYFP)Cos] (Srinivas et al., 2001), R26R^lacZ [Gt(ROSA)26Sor^tm1Sor] (Soriano, 1999) and Z/AP [Tg(CAG-Bgeo/ALPP)1Lbe] (Lobe et al., 1999) on a mixed genetic background were bred, genotyped by PCR and staged according to embryonic day. Cre^ERT2 activation was achieved using tamoxifen (Sigma) injection prenatally (50 μl of 20 mg/ml tamoxifen intraperitoneally) (Lepper et al., 2009) or postnatally [20 μl of 2 mM 4OH-tamoxifen (Millipore) subcutaneously].

Wild-type zebrafish, mutant lines myf5^hu2022 (Hinits et al., 2009) and myod^fh261 (Hinits et al., 2011), and transgenic lines Tg(actc1b:egfp)^zf13 (Higashijima et al., 1997), Tg(mylz2:egfp) (Ju et al., 2003) and Tg(kdrl:egfp)^s843 (Jin et al., 2005) were kept on King’s College wild-type or Tupfel long fin backgrounds. Zebrafish maintenance, staging and husbandry were as previously described (Westerfield, 2000).

Morpholinos (MOs, Gene-Tools) were injected into one-cell stage zebrafish embryos. Data showing MO1 and MO2 pairs behave similarly are given in supplementary material Table S1. MO sequences are included as supplementary material Table S2. pax3b MO knockdown efficacy was demonstrated by cloning pax3b 5′UTR and partial coding sequence in frame with GFP (supplementary material Fig. S9). pax3b MO1 induces a subtle small head phenotype that is rescued by co-injection with p53 MO (supplementary material Fig. S10) and was used unless otherwise stated.

Zebrafish in situ mRNA hybridisation was performed as previously described (Coutelle et al., 2001) with probes referenced in supplementary material Table S3. Immunodetection was performed as previously described (Hinits et al., 2009; Anderson et al., 2012). Alkaline phosphatase was revealed after antibody staining by washing into PBS and 2 mM MgCl₂ at 65°C for 30 minutes, transferring into 100 mM NaCl, 100 mM Tris-Cl (pH 9.5), 50 mM MgCl₂, 0.1% Tween and 2 mM levamisole for 20 minutes, and developing in NBT/BCIP (Roche). Antibodies used were against skeletal muscle myosin heavy chain (MyHC; MF20 or A4.1025, DSHB), smooth muscle myosin (Myh11, Biomedical Technologies) (Wallace et al., 2005), GFP (also binds EYFP; Torrey Pines), β-galactosidase (Cappel), Pax3 (DSHB), Pax7 (DSHB), paired-class homeodomain-containing proteins, including Pax3/7 (DP312) (Davis et al., 2005), phospho-Histone H3 (H3P, Millipore), β tubulin (βtub, Covance) and neurofilament (2H3, DSHB). H₂O₂ treatment during staining was used to diminish pigment in older larvae, but phenylthiourea was generally not used owing to potential toxicity (Bohnsack et al., 2011; Li et al., 2012). Phenotype quantification is included as supplementary material Table S1.

OSM and gastrointestinal tract smooth muscle length were measured in projected z-stacks. Smooth muscle was defined according to Myh11 immunofluorescence and expressed as a fraction of total zebrafish length (standard length, SL).

Fluoresbrite plain YG 1.0 μm microspheres (Polysciences) were fed to 5 dpf larvae incubated in sterile GZM media (Pham et al., 2008) as previously described (Farber et al., 2001). Swimming activity was assessed over 5 minutes. Total swimming distance was measured using the Manual Tracking ImageJ plug-in. Time active was expressed as percentage of ‘movement frames’ (adjoining frames that contain movement), relative to total movie frames. Quantitative real-time PCR (qRT-PCR) was conducted as previously described (Kanther et al., 2011) with primers referenced in supplementary material Table S4.

Kaede mRNA was transcribed using mMessage Machine (Ambion) and injected into one-cell stage embryos. Embryos with high levels of widespread Kaede^green were selected and mounted as previously described (Minchin and Rawls, 2011). Somite cells were marked by photoconverting Kaede from green (Kaede^green) to red (Kaede^red) with 405 nm light (Hatta et al., 2006). Photoconversion from two directions, and at different stages of development, was undertaken to diminish the chance of non-somite photoconversion confounding interpretation. First, somite photoconversion from dorsal at 3-5 somite stage (ss) definitively excluded lateral tissue, including lateral mesoderm (Fig. 2A), as determined by expression of lateral mesoderm genes (Dooley et al., 2005). Second, photoconversion of a discrete region of the anteriormost somites from a lateral angle at 14-18 ss excluded labelling of both neural crest and endoderm (supplementary material Fig. S3). Both photoconversion strategies also labelled overlying epidermis and neural tube (supplementary material Fig. S3). After photoconversion, embryos were checked to ensure adjacent somites and surrounding tissue (excluding overlying ectoderm) remained green. Photoconversion and imaging was undertaken on a Zeiss LSM5 Exciter confocal with a Zeiss W Plan-Apochromat 20×/1.0 DIC (UV) VIS-IR objective. Images were processed using ImageJ.

Data were compared by one-way ANOVA followed by Tukey post hoc tests. In histograms, columns with distinct letters (A-C) differ significantly at the P-value shown in the legend. Columns with the same letter do not differ significantly (P>0.05). Error bars are s.e.m. Numbers of embryos are indicated on columns.

To test the hypothesis that SHM and OSM derive from migratory somite cells, we genetically marked somite cells using Pax3/7 Cre-mediated lineage tracing in mouse. Pax3^Cre labelled SHM, longus colli and many other muscles in the neck region (Fig. 1A; supplementary material Fig. S1). Cre driven from the Pax3 locus also labelled some, but not all, OSM cells (Fig. 1B,C). Similarly, treatment of Pax7^CE mice with tamoxifen at E11.5-E12.5 marked OSM (Fig. 1D,F) and longus colli (supplementary material Fig. S2A-D). Thus, both Pax3- and Pax7-expressing cells contribute to OSM and SHM (supplementary material Table S5). The extent of labelling by Pax3^Cre correlated with the level of origin of motor innervation (Fig. 1B,C; supplementary material Fig. S1). Muscles with innervation from C4-C6, such as the infraspinatus, deltoid and biceps were generally well labelled with alkaline phosphatase; like OSM, muscles innervated from C1-C3, such as longus colli, in the neck region of Pax3^Cre;Z/AP mice, had mosaic labelling (supplementary material Fig. S1C-E; Fig. 1B,C). Nevertheless, longus colli was mostly absent in Pax3^Cre/Cre mutants (Fig. 1I). Ventral to the oesophagus, SHM (innervated from C1-C3) were generally well labelled (Fig. 1A), whereas adjacent sternomastoid myofibres and also trapezius (innervated from the more anterior XIth cranial nerve) were entirely unlabelled (supplementary material Fig. S1B). By contrast, Wnt1^Cre did not label OSM, SHM, longus colli or other neck muscles, indicating that neural crest did not contribute to OSM or other striated muscle (Fig. 1E,G; supplementary material Fig. S2A; data not shown). As Pax3 and Pax7 are only expressed in neural crest, non-migratory neuroepithelium and somites (Jostes et al., 1990; Daston et al., 1996; Borycki et al., 1999), these observations suggest that Pax3/7-expressing OSM, SHM and other neck muscle cells arise from the somite.

Temporal analysis of Pax7^CE embryos suggested that Pax7 expression accumulates gradually in OSM precursors. Exposure to tamoxifen at successively later embryonic stages marked increasing numbers of OSM fibres at E16.5, particularly in the outer longitudinal layer, where OSM first arises (supplementary material Fig. S2). Tamoxifen administration at E9.5 or E10.5, marked only a limited number of cells in the outer muscle layer at E16.5, suggesting that, around E10.5, most OSM precursors do not express Pax7 (supplementary material Fig. S2A,B). Nevertheless, tamoxifen administration at E11.5 and onwards marked increasing numbers of OSM fibres in the outer muscle layer at E16.5, indicating that a significant fraction of OSM precursors express Pax7 at E12.5, prior to OSM differentiation (Fig. 1D,F; supplementary material Fig. S2C,D; Table S6). Neonatally, most OSM fibres were marked 3 days after tamoxifen administration, even in Pax7^CE/CE mutants (Fig. 1H; supplementary material Fig. S2E). These findings suggest that Pax7 expression in OSM precursors increases during foetal life, becoming more uniform around birth, as occurs in myogenic cell populations in somite and limb.

To ascertain which somite(s) contribute to OSM, we used Kaede lineage tracing in zebrafish. Single anterior somites were photoconverted to Kaede^red at 3-5 ss (somite stage), or groups of somites were converted at 14-18 ss (Fig. 2A; supplementary material Fig. S3; Table 1). Consistent with Wood and Thorogood (Wood and Thorogood, 1994), early conversion of somite 1 (S1) marked, at 84 hpf, the anteriormost epaxial muscle fibres that extend from the posterior end of the otic vesicle to the cleithrum, having lost a morphological anterior somite border (Fig. 2D). S2 conversion marked the second somite, immediately behind the cleithrum (Table 1). Strikingly, both conversion strategies labelled Kaede^red cells in two stereotypical clusters of migratory cells rostral to the marked somite between 36 and 45 hpf (Fig. 2B; Table 1). By 45 hpf, a medial subpopulation of rostrally migrating cells was associated with the oesophagus (Fig. 2C). Subsequently, robust Kaede^red labelling was observed in the oesophagus at 84 hpf, suggesting somite cells migrate rostrally then medially to the oesophagus (Fig. 2D; Table 1). We conclude that anterior somite-derived cells migrate to the oesophagus and most likely correspond to OSM.

Kaede^green photoconversion within S1 or S2 at 3-5 ss, or S1-S3 at 14-18 ss, yielded migrating mesenchymal Kaede^red cells at 42 hpf that resemble SHM marked by an actin:GFP transgene (Fig. 2E,F; Table 1). Moreover, analysis of Kaede^red cells at 72-96 hpf, revealed strong labelling of the terminally differentiated SHM fibres, demonstrating that SHM derives from S1-S3 (Fig. 2G,H,J; Table 1). Interestingly, S1 marked more anterior regions of SHM compared with S2, suggesting anterior-posterior positioning is retained in the migrating SHM primordium (data not shown). Later photoconversion of anterior somites failed to label SHM, indicating that emigration of SHM precursors occurs before 35 hpf (P<0.0001; Table 1). Pectoral fin was routinely labelled, consistent with a S2-S3 origin for pectoral fin muscle (PFM; Fig. 2E,H) (Neyt et al., 2000; Haines et al., 2004). Somite photoconversion resulted in Kaede^red within blood vessels of the common cardinal vein (CCV) and pectoral fin vasculature (PFV; Fig. 2G; supplementary material Fig. S4B,C; Table 1). Finally, S1-derived marked cells were also associated with the surface of the notochord, but were always located posterior to the marked somite (Fig. 2B; Table 1), presumably reflecting sclerotome fate and posteriorward movement of the notochord (Wood and Thorogood, 1994). Taken together, our lineage tracing analyses identified both presumptive OSM and SHM as novel derivatives of somites in zebrafish.

Lineage tracing suggested that OSM precursors migrate rostrally from S1-S3 prior to 35 hpf before cascading medially onto the oesophagus (Fig. 2B; Table 1). To investigate the temporal and spatial development of zebrafish OSM, we analysed expression dynamics of genes important for striated myogenesis. At 30 hpf, myf5 was expressed in bilateral cell clusters immediately rostral to the anterior-most somites (Fig. 3A). These myf5⁺ striated muscle myoblasts were located within mesenchymal tissue, did not express other MRFs (data not shown) and appeared to move medially at 38 hpf (Fig. 3B,C). By 48 hpf, myf5⁺ myoblasts converged at the midline, at the level of S1-S3, forming a tube-like structure in the region of the developing oesophagus (Fig. 3D,G). Some myf5⁺ cells remained lateral at 48 hpf, thus forming a distinctive seagull shape immediately rostral to S1, and suggesting that migration towards the oesophagus is ongoing in an ‘inverse fountain’ (Fig. 3D,Q). Based on anatomical location during 30-48 hpf, we suggest these myf5⁺ cells are OSM precursors.

Expression of myod (myod1 - Zebrafish Information Network) first became evident in OSM precursors at the midline towards the end of the myf5⁺ fountain at 48 hpf, indicating maturation of OSM precursors in this region (Fig. 3E,H). myog mRNA was not detected in OSM at 48 hpf (Fig. 3F,I). By 72 hpf, levels of myf5 mRNA were significantly reduced (Fig. 3J). However, OSM now strongly expressed myod (Fig. 3K,N), myog (Fig. 3L,O) and mef2ca (data not shown), suggesting that differentiation of OSM was occurring. Indeed, weak expression of both slow (smyhc1) and fast (mylz2) myosin mRNA at 72 hpf suggested differentiated fibres were present in OSM (Fig. 3M; data not shown) (Elworthy et al., 2008). By 78 hpf, robust MyHC immunoreactivity was observed in the oesophagus, which extended caudally to the level of somite 4 (Fig. 3P). Fibres in rostral OSM had elongated morphology, whereas those in caudal OSM appeared less mature (Fig. 3P). To conclude, MRF expression analysis revealed a progression of OSM precursors, initially located in bilateral myf5⁺ clusters immediately rostral to anterior somites, which then appear to converge, express other MRFs and move posteriorly to localise at the oesophagus, where they undergo further differentiation (Fig. 3Q).

MRFs, including myf5, myod and myog, are essential for striated myogenesis in both mouse and zebrafish (Rudnicki et al., 1993; Hinits et al., 2009; Hinits et al., 2011). In mouse, OSM development is Myf5/Mrf4 (Myf6) dependent, but Myod1 independent (Kablar et al., 2000). The zebrafish myf5^hu2022 and myod^fh261 mutants are predicted nulls (Hinits et al., 2009; Hinits et al., 2011). At 72 and 120 hpf, myf5^hu2022 revealed no obvious defect in OSM myod, myog or myosin mRNA or MyHC protein accumulation (Fig. 4A,B; data not shown). However, in myod^fh261 mutants and myod morphants at 72 hpf, myog mRNA within OSM precursors was dramatically reduced, and substantially reduced in other cranial and somite-derived migratory muscles (Fig. 4C,D). Therefore, unlike mouse, Myod is required for myog expression within OSM in zebrafish. At 5 dpf, MyHC accumulation is strongly reduced in OSM of both myod^fh261 mutants and myod morphants, demonstrating that Myod is required for OSM differentiation (Fig. 4B; data not shown). Double myf5^hu2022;myod^fh261 mutants entirely lack OSM and all other skeletal muscle, although both striated cardiac and gut smooth muscle remain intact (Fig. 4B). Thus, in contrast to mouse, development of the OSM in zebrafish appears primarily Myod dependent, with some compensatory role of Myf5.

Little is known regarding genetic factors involved in OSM development prior to MRF expression. We hypothesised that development of somite-derived OSM is dependent on Pax3/7 genes. Mutation of Pax7 in mouse does not affect early OSM formation (Fig. 1H) (Wörl et al., 2009). At E14.5, Pax3^Cre/Cre mutants fail to form many neck muscles (Fig. 1I). Oesophageal smooth muscle was present but disorganised in Pax3^Cre/Cre mutants and no OSM was observed (Fig. 1I). However, as such mutants appear developmentally delayed prior to foetal death from heart defects, it was not possible to determine whether OSM was formed at the equivalent of E15.5, as observed in wild type.

We turned to zebrafish to assess the role of pax3/7 in OSM. Duplicated teleost pax3 genes form discrete pax3a and pax3b clades (Fig. 5A; supplementary material Figs S5, S6). Strikingly, following early neural expression (supplementary material Fig. S7A-G), somitic pax3b mRNA was restricted to anterior somites at 15 ss, in contrast to more widespread pax3a somite expression (Fig. 5B,D). Furthermore, by 18 ss, pax3b mRNA became restricted to the ventrolateral region of anteriormost somites, the putative site of MMPs (Fig. 5C), and persisted in this region until 55 hpf (Fig. 5F-H; supplementary material Fig. S7H). pax3a mRNA was detected widely in somite, but persisted in the ventrolateral region containing MMPs at 25 hpf (Fig. 5D,E,I). However, neither pax7 mRNA was detected in MMPs (Minchin and Hughes, 2008). In conclusion, pax3b mRNA is restricted to regions containing MMPs, suggesting a function for pax3b in the development of MMPs.

To investigate the role of the pax3 genes during OSM formation, previously verified MOs targeting pax3a were used to reduce specifically Pax3a protein (supplementary material Fig. S8) (Minchin and Hughes, 2008). Injection of a chimaeric 5′UTRpax3b:GFP RNA revealed strong efficacy of translation blocking by either of two non-overlapping pax3b MOs (supplementary material Fig. S9). In pax3a morphants, residual Pax3/7 immunoreactivity remained in the location of pax3b mRNA in rhombomere 4, and could be ablated by either pax3b MO (supplementary material Fig. S8H,H′; data not shown). Thus, we conclude that pax3b MOs reduce Pax3b protein.

In mouse, Pax3 is required for Lbx1 expression in MMPs, and both genes are essential for normal MMP migration into the limb (Mennerich et al., 1998; Gross et al., 2000). In zebrafish, lbx2 marks MMPs from 22 ss (Neyt et al., 2000; Ochi and Westerfield, 2009). At 24 hpf, a stream of lbx2⁺ cells projected rostrally from S1, and presumably corresponds to OSM and SHM progenitors (Fig. 5K). Comparison with dlx2a expression within cranial neural crest showed that the anteriormost lbx2⁺ cells had migrated to the hyoid arch in controls (Fig. 5K), but were substantially reduced and had progressed less far anteriorly in pax3b morphants (Fig. 5M,O-Q). Injection of pax3a MO alone had little effect on early migrating lbx2⁺ myoblasts and double pax3a+pax3b morphants had a similar phenotype to pax3b single morphants (Fig. 5L-Q). Therefore, Pax3b is required for the rostral projection of lbx2⁺ presumptive MMPs.

Knockdown of Pax3b also resulted in a severe reduction in MRF mRNAs in OSM myoblasts around 50 hpf (Fig. 6A-H′; supplementary material Fig. S8J) and significantly reduced OSM at 96 hpf (Fig. 6I-L,R). Depletion of Pax3a alone did not affect OSM specification or differentiation, whereas pax3a+pax3b double morphants showed disruption of OSM differentiation similar to single pax3b morphants (Fig. 6). This suggests that reduced MMP migration in pax3b morphants leads to defective OSM differentiation. Other myogenic processes in pax3b morphants appeared unperturbed. These included early somitic myogenesis (supplementary material Fig. S11A-F), presence of Pax3/7⁺ cells within the myotome at 120 hpf (supplementary material Fig. S11G-W) and proliferation of Pax3/7⁺ dermomyotome cells at 53 hpf (supplementary material Fig. S11X,Y). Thus, knockdown of Pax3b preferentially affects myogenesis in muscle near to, but outside of, rostral somites.

It has been suggested that OSM arises from smooth muscle (Patapoutian et al., 1995). We asked whether reduction of OSM in pax3b morphants led to increased oesophageal smooth muscle. Oesophageal smooth muscle, as assayed by either transgelin mRNA at 72 hpf (supplementary material Fig. S8I) or Myh11 accumulation in the oesophagus at 96 hpf (Fig. 6S-V), was not overtly affected by pax3 knockdown. Furthermore, the length of the gastrointestinal tract smooth muscle was not increased, showing that smooth muscle is not augmented in the absence of OSM (Fig. 6W). Thus, Pax3b depletion does not influence oesophageal smooth muscle formation.

Our lineage analysis identified SHM as an anterior somite derivative (Fig. 2E,G,H,J). Knockdown of Pax3b, but not of Pax3a, caused a significant reduction of myog mRNA in all migratory hypaxial muscles at 53 hpf, including SHM, PFM and posterior hypaxial muscle (PHM, Fig. 6G). We next used the Tg(actc1b:egfp) transgenic line to examine the consequence of Pax3b knockdown in hypaxial myogenesis. At 120 hpf, terminally differentiated hypaxial muscle was dose-dependently reduced and disorganised upon injection of pax3b MO (Fig. 7A,B). By contrast, pax3a morphants did not exhibit a severe hypaxial muscle phenotype and pax3a+pax3b double morphants were similar to pax3b single morphants (Fig. 7A). Thus, pax3b appears the primary Pax3 required for efficient hypaxial muscle differentiation in zebrafish. Curiously, formation of the common cardinal vein, which also received cells from anterior somites, was delayed in pax3b, but not in pax3a, morphants at 30 hpf (supplementary material Fig. S12). Knockdown of both Pax7a and Pax7b proteins by MO injection, which ablated Pax7 immunoreactivity (supplementary material Fig. S13A), had no discernible effect either upon MRF expression in MMPs or the later differentiation of OSM or SHM (supplementary material Fig. S13B-D). Together, these data demonstrate that Pax3b is required for migration of normal hypaxial muscle precursors.

To test whether Pax3b-dependent migratory muscles are required for effective feeding, we assessed internalisation of fluorescent microspheres (Farber et al., 2001). At 5 dpf, pax3b morphants exhibited equivalent levels of physical activity in comparison with controls (Fig. 7C-F), and unchanged accumulation of mRNAs associated with lipid and carbohydrate metabolism (Fig. 7G). Thus, pax3b morphants appear physiologically normal at 5 dpf. Strikingly, however, pax3b single and pax3a+pax3b double morphants showed significant reduction in microsphere internalisation, indicating that ingestion is impaired after Pax3b knockdown (Fig. 7H,I). In control larvae and pax3a morphants, microspheres were mainly in the intestine, showing that internalised microspheres quickly pass through the oesophagus (Fig. 7J,K). In pax3b or double morphants a large increase in pharynx-localised microspheres was evident, suggesting that reduced OSM leads to defective oesophageal swallowing (Fig. 7J,K). pax3b morphants did not consistently show defects in neural crest (supplementary material Fig. S14), head or pharyngeal arch development (Fig. 5K′-N′,Q; Fig. 6), suggesting that defects in feeding behaviour of pax3b morphants were not related to neural crest problems. Clearly, zebrafish Pax3b is required for effective microsphere internalisation and swallowing.

The findings in this study demonstrate four major points. First, cells contributing to OSM and SHM come from somites in both zebrafish and mouse. Second, OSM formation is Myod dependent in zebrafish. Third, normal formation of OSM and SHM require Pax3b function. Fourth, loss of Pax3 function leads to defective ingestion and swallowing. As human PAX3 is abundantly expressed in oesophagus (Tsukamoto et al., 1994), and Waardenburg Syndrome, which is caused by mutations in PAX3 (Goodman et al., 1982), can be associated with oesophageal abnormalities (Nutman et al., 1986), our findings raise the possibility that PAX3 may regulate oesophageal myogenesis in humans.

Our work addresses the long-standing issue of the origin of OSM. In mouse, Pax3 and Pax7-expressing cells contribute to OSM, suggesting a somitic origin. Kaede lineage tracing in zebrafish allowed us to gain a more precise view of this event: bilateral streams of lbx2⁺ OSM and SHM precursors emanate from pax3a/b⁺ hypaxial regions of the most anterior somites (S1-3) and migrate rostrally along a shared path. These streams bifurcate as myf5 expression commences and OSM precursors then move medially towards the oesophagus in an ‘inverse fountain’ where they undergo striated muscle differentiation (supplementary material Fig. S15). As myosin is not detected in OSM until after 72 hpf, cell migration appears to occur via MMP intermediates. In mouse, Pax3 is expressed in presomitic mesoderm, dermomyotome and during MMP migration (Relaix et al., 2004; Relaix et al., 2005; Horst et al., 2006). By contrast, Pax7 only becomes evident in limb MMPs at ∼E12, after migration (Relaix et al., 2004; Relaix et al., 2005). We propose that mouse OSM progenitors undergo canonical MMP development, expressing Pax3 within the somite and during migration, with some expressing Pax7 at the site of differentiation. Zebrafish in situ hybridisation did not detect mesodermal pax3b mRNA outside the somite. Furthermore, mouse Pax3 regulates MMP emanation from somites (Epstein et al., 1996; Dietrich, 1999). Therefore, we speculate that Pax3 functions within somites at early stages of OSM development.

It remains possible that OSM has multiple developmental origins. Our lineage analyses marked small numbers of cells in zebrafish oesophagus, suggesting either additional, non-somitic, origins for OSM or limited tracer detection. Several factors conspire to limit detectability of the Kaede lineage tracer: unilateral photoconversion, MMP proliferation, rapid increase in cell volume after muscle terminal differentiation, fusion with unmarked cells, and the flattened monolayer topology and deep location of OSM. Despite these limitations, genetic lineage tracing in mice produced a similar picture: only some OSM fibres were marked by Pax3-driven Cre. Therefore, our data suggest multiple developmental origins for OSM. The residual presence of OSM in Pax3^Sp mice (Kablar et al., 2000) and after pax3b knockdown in zebrafish supports this view. Other potential non-somitic sources of OSM include lateral cranial mesoderm, as loss of tbx1 function also reduces larval oesophageal muscle (Piotrowski and Nüsslein-Volhard, 2000). Perhaps OSM forms from multiple cell populations, as has been suggested for limb muscle (van Swearingen and Lance-Jones, 1993). We find no evidence for smooth muscle transdifferentiation into OSM. First, Pax3^Cre did not mark oesophageal smooth muscle. Second, reduced OSM in Pax3b-deficient zebrafish was not accompanied by noticeably increased smooth muscle, although our methods may not detect small increases in smooth muscle. Thus, our data support existing evidence against smooth muscle transdifferentiation into OSM (Zhao and Dhoot, 2000; Rishniw et al., 2003; Rishniw et al., 2009).

Our study demonstrates a role for pax3b, uniquely among pax3/7 genes, during initial development of zebrafish MMPs, SHM and OSM. Pax3^Cre/Cre and Pax3^Sp mutants die prior to OSM formation with severe neck muscle defects (Tajbakhsh et al., 1997). Pax3^Sp mutants on a non-C57BL background survive longer, however, and have some OSM (Kablar et al., 2000). Interestingly, an 80% reduction of Pax3 in Pax3^neo/neo mice leads to defective tongue but functional diaphragm muscle, suggesting that loss of Pax3 has stronger effects on specific MMP derivatives (Zhou et al., 2008). Pax3a appears to have little role in OSM development, in contrast to its essential role in neural crest (Minchin and Hughes, 2008). However, pax3a+pax3b double morphants had a more severe reduction of hindbrain lbx2 mRNA, increased H3P⁺ cells within the neural tube, developmental retardation and more severely reduced microsphere ingestion than pax3b single morphants. Thus, knockdown of both Pax3a and Pax3b elicits unique phenotypes that warrant further investigation.

In zebrafish, some somite-derived muscles are myod dependent (Hinits et al., 2009; Hinits et al., 2011). In accordance, zebrafish OSM is also myod dependent. In mice, OSM is Myf5;Mrf4-dependent but Myod independent (Kablar et al., 2000). It is currently unclear whether double myf5;mrf4 mutant zebrafish have OSM defects. However, the diminished role for Myod in murine OSM development probably reflects shifting MRF use during vertebrate evolution, as discussed elsewhere (Hinits et al., 2009).

Our lineage analysis demonstrates an anterior somite origin for zebrafish SHM, thus confirming suppositions based on gene expression (Schilling and Kimmel, 1997; Lin et al., 2006). Zebrafish SHM structure retains vestiges of its multi-segmental origin: two separate compartments have aligned fibre ends reminiscent of myotome segment borders (Hinits et al., 2011). No other cranial muscles in zebrafish had a somitic contribution or depended on Pax3. By contrast, Pax3^Cre lineage tracing in mouse indicates a somitic contribution to many neck muscles, including SHM and longus colli muscles which have similar innervation, flank the oesophagus and are mostly absent in Pax3 mutants. Other neck muscles, such as sternomastoid and trapezius, are not marked by Pax3^Cre. Although fish do not have necks, SHM and OSM are the only larval muscles that can be homologised with the more complex neck region of mammals (Diogo et al., 2008). We speculate that the evolutionary descendant muscles of SHM in primitive fish are defective after Pax3 knockout in mouse.

Like MRFs, Pax3/7 function was differentially partitioned in early amniotes (Franz et al., 1993; Monsoro-Burq et al., 2005; Sato et al., 2005; Basch et al., 2006). As OSM and SHM occur in chondrichthyes, but not agnathans (Kusakabe et al., 2004; Chatchavalvanich et al., 2006), it seems that in early bony fish the Pax3/7 and MRF gene families retained substantial evolutionary flexibility, which may have allowed migratory hypaxial muscle to co-evolve with the vertebrate jaw. We speculate that evolution of the jaw occurred concomitant with somite production of OSM/SHM, facilitating voluntary control of swallowing.

We thank D. Stemple and C. Moens for mutant fish, J. Epstein and M. Logan for mouse lines, and F. Conlon for useful discussions.