RARγ is required for mesodermal gene expression prior to gastrulation in Xenopus

Retinoic acid (RA) is a signaling molecule present at discrete concentrations in precise locations throughout the developing embryo, activating gene expression where RA receptors (RARs) are expressed. The RARs are a family of three genes that encode receptor subtypes RARα, β and γ, which are ligand-modulated transcription factors in the nuclear hormone receptor superfamily (reviewed by Maden, 2007; Evans and Mangelsdorf, 2014). Where RA is absent, RARs can also function as unliganded repressors (Janesick et al., 2014). RARs possess a range of activities and developmental functions that depend on the local concentration of RA, the spatial distribution of each RAR subtype (α, β or γ) and the availability of co-activator and co-repressor proteins (Rochette-Egly and Germain, 2009).

RAR signaling is implicated in numerous developmental processes, including patterning and boundary-setting, proliferation/differentiation and organogenesis (Niederreither et al., 2001; Maden, 2007; Rhinn and Dolle, 2012; Janesick et al., 2013; Cunningham and Duester, 2015). RA is essential for the development of many non-axial mesodermal derivatives, in particular the heart, somites and kidney (Niederreither et al., 2001; Moreno and Kintner, 2004; Naylor et al., 2016). We previously showed that liganded RARβ2 in the trunk is required for somite patterning and number, hypaxial muscle migration, and restriction of the presomitic mesoderm, whereas unliganded RARγ maintains the size of the presomitic mesoderm and the pool of bipotential caudal progenitor cells (Janesick et al., 2014, 2017).

Early studies revealed that RA is unable to instruct naive ectoderm to become mesoderm. Animal caps (ectodermal explants) fail to elongate in the presence of RA (Ruiz i Altaba and Jessell, 1991; Sive and Cheng, 1991); therefore, RA signaling was not considered to have an instructive role in mesoderm formation. Furthermore, the RA-synthesizing enzyme Aldh1a2 and RA receptors are not vegetally enriched (De Domenico et al., 2015) where mesoderm inducers are located; therefore, these are unlikely to be endogenous mesoderm-inducing factors. Instead, it has been shown that RA influences mesoderm patterning after gastrulation, acting to suppress anterior and dorsal characteristics (Ruiz i Altaba and Jessell, 1991; Sive and Cheng, 1991). Nevertheless, the expression of RARγ in the circumblastoporal region of the early gastrula embryo (Pfeffer and De Robertis, 1994) led us to ask whether RA signaling is functioning earlier than initially thought.

The importance and existence of competence or maintenance factors located in the responding marginal zone has become increasingly recognized (Fletcher and Harland, 2008). It is conceivable that RARγ signaling could be such a factor, given its expression in the marginal zone, although this has not previously been addressed. Here, we provide evidence supporting a novel role for RARγ signaling in mesoderm formation. We show that the function of RARγ in mesoderm patterning after gastrulation is preceded by an earlier phase in which RARγ modulates the expression of wnt8, brachyury (tbxt; also known as T), gdf3 and fgf8 in the gastrula-stage embryo. Using RNA-seq, we identify several novel genes expressed in the circumblastoporal region that are upregulated by RAR signaling. We show that RARγ signaling is required for the maintenance of Myod protein in the presomitic mesoderm (PSM) domain and also find that RARγ knockdown stabilizes the muscle progenitor state, producing ectopic Myod⁺ cells in the trunk and impeding somite maturation.

In 2002, Carrasco and colleagues first demonstrated that loss of retinoic acid receptor gamma 1 (RARγ1) abolished brachyury (T) and foxa4 (pintallavis) expression on the injected side (Fig. S1; Franco, 2002). Hence, we hypothesized that the presence of RARγ is required for the mesodermal competence or maintenance. RARγ is expressed in the circumblastoporal domain during gastrula stages, as are the genes encoding both the RA-synthesizing enzyme Aldh1a2 and the degradation enzyme Cyp26a1 (Fig. S2). Therefore, components of RA signaling are found in the correct time and place to regulate mesoderm development. We designed a second morpholino (MO) to Xenopus laevis RARγ1 in addition to the original MO used in Fig. S1, and validated specificity of the phenotype on T expression (Fig. S3). Loss of RARγ1 does not completely derail gastrulation (except at high MO doses), but strongly downregulates expression of genes wnt8, gdf3 and fgf8 in signaling pathways (Wnt, Nodal, Fgf) that are crucial for regulating mesoderm formation (Fig. 1A-D). This effect is selective for the RARγ subtype, as RARα MOs do not affect these markers (Fig. 1E-G), and rarb is not detectable at gastrula stages (Janesick et al., 2017). To be certain that the RARα morpholino was functional, we verified in the same clutch of embryos that RARα knockdown strongly inhibited expression of N-tubulin (tubb2b) at stage 14 (Fig. 1H) as we previously showed (Janesick et al., 2013).

We conducted transcriptomal analysis (Fig. S4) of gastrula-stage embryos that had been treated at early blastula (stage 6/7) with the RAR-selective agonist TTNPB, the RAR-selective antagonist AGN193109, or 0.1% ethanol vehicle control. We identified 135 differentially regulated genes (after reducing .L or .S homeolog pairs to one gene symbol) upregulated by TTNPB compared with control (Benjamini–Hochberg false discovery rate q<0.05, FPKM_VEH> 0.1, fold change>1.5) (Table S1). We identified four genes significantly downregulated by TTNPB [orthodenticle homeobox 2 and 5 (also known as cone-rod homeobox), ADAM metallopeptidase with thrombospondin type 1 motif 1 and betacellulin], and no genes significantly regulated by AGN193109. If we employ a less-stringent P-value, the number of differentially expressed genes is substantially increased. However, we aimed for 100% validation by in situ hybridization, and therefore selected the more stringent q-value cut-off in these studies.

We used gene ontology (GO) analysis to assess the types of genes regulated by TTNPB in gastrula embryos and identified terms highly enriched for genes involved in known developmental roles of RA, such as regulation of homeobox proteins, embryonic morphogenesis, regionalization and patterning (Fig. 2A). Antennapedia-type homeobox genes are not typically expressed in control embryos at gastrulation, but were turned on prematurely by TTNPB (Fig. 2A). A number of Wnt signaling genes are upregulated by TTNPB, including some that promote signaling (wnt5b, frizzled class receptor 4 and 10) as well as some that antagonize signaling [dishevelled-binding antagonist of beta-catenin 1 (dact1), kringle containing transmembrane protein 2 (kremen2) and zinc finger protein 703 (znf703)]. GO terms for neural crest, ear and hindbrain development are also highly enriched (Fig. 2A).

We selected genes for further study based on three criteria: (1) novelty (RA targets not previously characterized), (2) potential role in mesoderm development (e.g. connection to Nodal, Wnt, Fgf or Bmp pathways), or (3) regulation of mesodermal derivatives at later stages. Fig. 2B shows the FPKM values for selected genes, and we note that the expression of the .L and .S homeologs is relatively concordant. We validated the RNA-seq data by whole-mount in situ hybridization (WISH) (Fig. 3). Although WISH is not quantitative, we observed a positive correspondence between FPKM values and WISH signal intensity, with the exception of MAM domain containing 2 (mamdc2), which we detected despite low FPKM (Fig. 3D). Probes not detected at stage 10.5 [kielin/chordin-like protein (kcp), hyaluronan and proteoglycan link protein 3 (hapln3) and follistatin like 1 (fstl1)] were validated at neurula or tailbud stages to show correct, published expression when this information was available on Xenbase (Karpinka et al., 2015).

znf703, znf503, kremen2 and mamdc2 are expressed in the circumblastoporal ring, but absent from the organizer (Fig. 3A-D), reminiscent of the expression patterns of ventx2 or wnt8 (Karpinka et al., 2015). TTNPB intensifies and expands expression of znf703, znf503, kremen2 and mamdc2 into the dorsal domain (Fig. 3E-H). This result is concordant with zebrafish data showing that RA causes the ortholog Znf703 to be ectopically expressed in the embryonic shield (Andreazzoli et al., 2001). skida1, which normally is barely detectable, is intensified in TTNPB-treated embryos, although its domain is unchanged (Fig. 3I,M). Nkx6-2 is found in the dorsal lip, and TTNPB widens and ventralizes its expression (Fig. 3J,N). dhh was not previously characterized at gastrula stages; we show that it is expressed in a punctate pattern surrounding the blastopore, and is greatly expanded by TTNPB (Fig. 3K,O). cyp26a1 is a known direct RAR target and serves as a positive control for the effects of modulating RA signaling. cyp26a1 expression is expanded into the animal domain accompanied by punctate, apparently nuclear staining in the vegetal tissue (Fig. 3L,P).

Considering the overlapping expression of the genes encoding the RA-synthesizing enzyme Aldh1a2 and the RA-degrading enzyme cyp26a1 with rarg mRNA expression (Fig. S2), we tested whether RARγ1 was functioning as an activator or repressor in modulating the expression of one of its targets, znf703. RARs are capable of both direct activation or repression of target genes depending on their proximity to ligand during development (Janesick et al., 2014). We possess expression constructs for wild-type RARγ1 (WT RARγ1), constitutively active RARγ1 (VP16-RARγ1, which mimics the liganded, active form of the receptor), and dominant-negative RARγ1 (DN-RARγ1, which mimics the unliganded, repressive form of the receptor), which were made in a similar manner to the RARα constructs previously described by Blumberg et al. (1997). Microinjection of these constructs alone led to expansion of znf703 in VP16-RARγ1-injected embryos (Fig. S5A,B) and knockdown of znf703 by DN-RARγ1 (Fig. S5C,D). WT RARγ1 rescues znf703 expression as expected (Fig. 4C,G), and further confirms the specificity of our morpholino. VP16-RARγ1 rescues znf703, and expands its domain dorsally and animally (Fig. 4B,F; Fig. S5E) whereas DN-RARγ1 does not rescue znf703 expression (Fig. 4D,H). Therefore, we infer that RARγ1 is functioning as a transcriptional activator during mesoderm development, and that retinoids are likely available to the embryo at this stage, earlier than previously recognized.

Xenopus animal caps can be induced to become dorsal mesoderm at high concentrations of activin (Sokol et al., 1990; Ruiz i Altaba and Jessell, 1991). We isolated animal caps from RARγ1 MO-injected embryos using a low-dose MO that does not elicit non-specific effects on gastrulation. We treated the caps with activin and found that, instead of elongating, they dissociated within 12 h irrespective of the activin concentration, compared with control MO (Fig. 5A,B). ncam1 expression was upregulated by activin in control embryos as previously published (Bolce et al., 1992), but was significantly decreased in RARγ1 MO-injected caps at 8-h post-activin treatment (Fig. 5C), with little change in expression of the housekeeping gene histone H4 (hist1h4d). This indicates that RARγ1 is required for cellular adhesion and the ability of explants to elongate. Furthermore, adhesion is also crucial for intercellular signaling required for mesodermal competence (Gurdon, 1988; Kato and Gurdon, 1993), as well as in boundary formation, where RAR is known to play a major role (e.g. in somite and rhombomere border delineation).

One of the earliest derivatives of mesodermal induction is muscle (Gurdon et al., 1993). Soon after gastrulation begins, myod expression occurs in presumptive muscle in response to mesoderm induction (Hopwood et al., 1989; Harvey, 1992). Consistent with our results in early mesoderm (Fig. 1), we found that expression of myod mRNA was restricted and diminished on the injected side of RARγ1 MO-embryos (Fig. 6A; also see figure S13 in Janesick et al., 2014). Generally speaking, myod WISH staining was mostly diffuse and reduced in intensity such that the only conclusion it is possible to make confidently is that somite boundaries have been blurred or lost. Therefore, we employed an immunohistochemical approach to detect Myod protein (Hopwood et al., 1992) in coronal sections. As expected, we observed loss of unsegmented PSM expression on the injected side (Fig. 6B), which agrees with our previous finding that RARγ is required for the maintenance of the PSM (Janesick et al., 2014).

Despite the fact that Myod is frequently used as a somite marker, its presence is insufficient to assure terminal myogenic differentiation (Hopwood et al., 1991; Kato and Gurdon, 1993). Rather, continued expression of Myod inhibits terminal differentiation of skeletal muscle. We found ectopic Myod-positive nuclei throughout the trunk of the tailbud-stage embryo (Fig. 6A). Accordingly, differentiated skeletal muscle marked by 12/101 staining was significantly reduced on the injected side (Fig. 6B). We conclude that loss of RARγ1 prevents the full myogenesis program from occurring and that this phenotype is likely to be linked to deficiencies in early mesoderm formation (see Discussion).

RA is not considered to be an instructive signal in mesoderm induction, in contrast to Nodal and Wnt signaling. RA inhibits expression of dorsal mesoderm markers in animal caps exposed to high doses of activin, but is unable to promote elongation of caps alone (Ruiz i Altaba and Jessell, 1991; Sive and Cheng, 1991). Previous research showed that RA acts during mesoderm formation by transforming the anteroposterior value of mesoderm toward the posterior. Nevertheless, factors that serve as permissive signals or contribute to the competence of responding tissue are very important for embryonic inductions, including mesoderm induction (Fletcher and Harland, 2008). RA, in combination with BMP4, can induce the production of mesodermal precursors from embryonic stem cells (Torres et al., 2012), and many mesodermal derivatives (e.g. somites, heart, kidney, gonads) are affected when RA signaling is perturbed (Cartry et al., 2006; Hasegawa and Saga, 2012; Rhinn and Dolle, 2012, and reviewed by Cunningham and Duester, 2015; Naylor et al., 2016). Here, we propose a new role for RARγ signaling and show that it is required much earlier than had been previously thought for the expression of mesodermal markers, in vivo.

Retinoids are present in Xenopus eggs and embryos, but their availability in early development has been controversial (Pijnappel et al., 1993; Creech Kraft et al., 1994; Blumberg et al., 1996). Reporter gene studies have suggested that retinoids are not active in the embryo until neurula stages (Ang and Duester, 1999); however, 4-oxo-retinaldehyde (Blumberg et al., 1996) and 4-oxo-RA (Pijnappel et al., 1993) can activate RAR and are present in gastrula-stage embryos. Chicken embryos also possess high levels of RA in Hensen's node during gastrulation (Chen et al., 1992). Although cyp26a1 is highly expressed in the Xenopus circumblastoporal domain, it has not been determined whether this enzymatic pathway is responsible for generating 4-oxo-retinoids, in comparison with its known activity in producing 4-OH-retinoids (Shimshoni et al., 2012; Topletz et al., 2015). We showed that only the wild-type and constitutively active forms of RARγ1 rescued mesodermal markers (e.g. znf703) in RARγ1-knockdown embryos. As the constitutively active form of RARγ1 mimics liganded RARγ1, our results suggest that retinoids are available to the embryo (at least locally) at gastrula stages, unless the receptor functions as an unliganded activator, which would be unprecedented for an RAR.

We identified several novel targets expressed in the circumblastoporal region that are upregulated by RAR signaling in the early gastrula. znf703, znf503, kremen2 and mamdc2 are normally absent from the dorsal organizer, but are ectopically expanded there in the presence of RA. Hence, RA ventralizes dorsal tissue during mesoderm development, a phenomenon already recognized in later stages, as RA-treated embryos display a bauchstuck (ventralized) phenotype (Durston et al., 1989; Sive et al., 1990). RA also expanded the expression domain of the ventral patterning gene nkx6-2. This is concordant with the observation that Nkx6.2 is reduced in vitamin A-deficient quail as well as RAR antagonist-treated chicken embryos (Diez del Corral et al., 2003; Bayha et al., 2009).

Kremen2 is the receptor for Dickkopf, a Wnt inhibitor that mediates mesendoderm internalization and promotes head mesoderm and cardiac fate (Schneider and Mercola, 2001). skida1, a DACH/dachshund-related gene of unknown function, was more ubiquitously expressed than other markers, but highly responsive to RA. Its expression in the pronephric kidney at later stages (Seufert et al., 2005) could indicate a potential role of skida1 in responding to RA signaling in pronephric precursors. dhh is a hedgehog gene important for mammalian Leydig cell differentiation (reviewed by Franco and Yao, 2012), and is highly expressed in Xenopus testis (Haselman et al., 2015). These findings could provide molecular mechanisms linking RA signaling with intermediate mesoderm and gonadal development.

Our results show that RARγ1 MO inhibits cell adhesion in animal caps which has important implications for numerous developmental processes, such as the morphogenic movements of gastrulation and the formation of somite or rhombomere boundaries. Two of the targets identified in our RNA-seq dataset regulate adhesion. Although the function of mamdc2 is not well characterized, the MAM domain is found in proteins that regulate cell adhesion and motility, such as zonadhesin and nephronectin (Beckmann and Bork, 1993). znf703 is implicated in epithelial-mesenchymal transitions (EMTs), downregulates E-cadherin (cdh1), and increases cell invasion/motility (Holland et al., 2011; Slorach et al., 2011; Shahi et al., 2015). Although mesoderm formation in Xenopus is not considered to be a classical EMT (Nakaya and Sheng, 2008), the transformation of epithelioid cells to mobile mesenchymal cells during Xenopus gastrulation undoubtedly requires genes that regulate cell adhesion and morphogenesis, such as znf703. znf703 is also expressed in somites and the neural plate border (Hong and Saint-Jeannet, 2017). Therefore, regulation of znf703 by RA might be important for the epithelialization of somites as well as neural crest migration.

RA can act late on myoblast, cardiac, renal or gonadal progenitor populations to affect specification or differentiation, but our results open the possibility that RARγ acts earlier to promote mesodermal fates. Loss of adhesion and subsequent dissociation in explants from RARγ1 MO-injected embryos is likely to disrupt the ‘community effect’ (intercellular signaling), which has been shown to be important for myogenic differentiation and the early stages of somitogenesis (Gurdon, 1988; Cossu et al., 1995). In these early studies, myogenic commitment was not solely determined by inducing signals, but additional signals (e.g. embryonic FGF ligand) within the responding tissue (Symes et al., 1988; Standley et al., 2001). One of the earliest mesodermal genes upregulated after induction is Myod, but expression of Myod alone doth not a muscle make (Hopwood et al., 1991). The fact that the injected side of RARγ1 MO tailbud embryos still expresses Myod without 12/101 indicates that cells have not received the appropriate signals (even if only from each other) to fully stabilize myogenic commitment and terminal differentiation.

In line with our previous findings (Janesick et al., 2014), injection of RARγ1 MO led to an absence of Myod protein in nuclei of the unsegmented PSM. This is expected because RARγ1 is known to play an important role in the maintenance of chordoneural hinge and unsegmented PSM (Janesick et al., 2014). However, we also observed that in RARγ1 MO-injected embryos, Myod⁺ cells were found ectopically in the trunk, which could indicate that PSM identity is transformed rostrally. Alternatively, the presence of Myod⁺ cells without somite maturation (as indicated by 12/101 staining) suggests that the muscle differentiation program commences, but cannot finish. In this scenario, RARγ1 loss of function might actually stabilize the muscle progenitor state, thus hindering somite maturation, or Myod might require RARγ1 for its degradation.

Unlike other manipulations of RA signaling (chemical treatment, RARβ MO injection, overexpression of constitutively active or dominant-negative RARα or RARγ), which shift somitomere boundaries or alter somitomere size (Moreno and Kintner, 2004; Janesick et al., 2014), or chevron morphology (Janesick et al., 2017), loss of RARγ is the only manipulation of RA signaling that completely precludes any somite boundaries from forming. Furthermore, genes that normally regulate boundary formation, such as ripply2 and thyl2 (mespa) are barely detectable in RARγ MO-injected embryos (Janesick et al., 2014). The RARγ MO-induced dissociation of explants is a clue that cellular adhesion, an important component of boundary formation, is perturbed. Furthermore, somite maturation requires Myod degradation for proper somite epithelialization and boundary formation. Continued ectopic Myod expression observed on the injected side of RARγ1 MO-injected tailbud embryos is likely to completely thwart a successful myogenic program.

Our finding that RARγ is required for the expression of early mesodermal markers opens the possibility that RA signaling provides a more instructive role, as opposed to simply providing positional cues to already committed tissue. Current guidance protocols for making mesoderm or muscle from stem cells in culture currently do not utilize RA (Chal et al., 2016; Loh et al., 2016). Our results suggest that RA signaling could be exploited in mesoderm formation, but with careful attention to specificity of receptor subtypes. As new receptor-selective ligands are developed (Tsang et al. 2003; Shimono et al., 2011), it will be interesting to consider whether mesodermal lineages can be further manipulated, stabilized or specialized based on the results provided here and an increased understanding of how RA receptors actually function in development.

Eggs from one clutch of female X. laevis were fertilized in vitro and embryos were staged as described (Janesick et al., 2014). Stage 6-7 embryos were treated in groups of 25 in 60-mm glass Petri dishes with 10 ml of 0.1× MBS containing 1 μM RAR-selective agonist TTNPB (Tocris), 1 μM RAR-selective inverse agonist (antagonist) AGN193109 (a gift from Dr Rosh Chandraratna, IO Therapeutics; Arima et al., 2005) or vehicle control (0.1% ethanol). Each dish was harvested at gastrula stage 10.5 in five-embryo aliquots; each aliquot was designated as n=1 biological replicate. Replicates were taken from two different clutches, and five replicates were analyzed within each clutch (ten replicates in total). Approximately ten embryos from each chemical treatment were reserved and aged until stage 40 to verify that the chemicals resulted in strong retinoid phenotypes as previously published (Koide et al., 2001).

Each five-embryo pool was homogenized in 200 μl of TriPure (Roche). Total RNA was DNAse treated, LiCl precipitated, and 100 ng was reverse transcribed into cDNA using Ovation RNA Amplification System V2 (NuGEN). Deep-sequencing libraries with barcode indices were synthesized using ABI SOLiD 5500 Fragment Library Core Kit and subjected to 50-nucleotide, single-read multiplex sequencing with ABI SOLiD 5500XL deep sequencers (Aceto, et al., 2014). The Bowtie, TopHat and Cufflinks pipeline was used (Trapnell et al., 2012), and expression data deposited in Gene Expression Omnibus under accession number GSE119124. Differential gene expression analysis (TTNPB or AGN193109 versus control) was conducted using CyberT (Kayala and Baldi, 2012) to facilitate comparison with our previously published microarray datasets (Arima et al., 2005; Janesick et al., 2014). We employed VSN normalization and performed unpaired two-conditions data analysis on FPKM (fragments per kilobase per million mapped reads) values for 1 μM TTNPB versus vehicle control, and 1 μM AGN193109 versus vehicle control. q-values from differential expression analysis are reported as Benjamini–Hochberg-corrected P-values. The full dataset of differentially expressed genes is available in Tables S1 (1 μM TTNPB) and S2 (1 μM AGN193109). We conducted GO analysis using DAVID (Database for Annotation, Visualization and Integrated Discovery; Huang da et al., 2009a, b).

Xenopus eggs were fertilized in vitro and embryos were staged as described (Janesick et al., 2012). Embryos were microinjected bilaterally or unilaterally at the two- or four-cell stage with gene-specific MOs (Table S3) and/or mRNA together with 100 pg/embryo β-galactosidase (β-gal) mRNA lineage tracer. Both morpholinos against RARγ1 yield the same knockdown phenotype on znf703 (Fig. S3). Embryos were maintained in 0.1× MBS until appropriate stages. Embryos processed for WISH were fixed in MEMFA (10% 10× MEMFA salt, 10% formaldehyde, 80% DEPC water), stained with magenta-GAL (Biosynth), and then stored in 100% ethanol (Janesick et al., 2012).

WISH was performed on microinjected embryos as previously described (Janesick et al., 2012). fgf8 probe (courtesy of Nancy Papalopulu, University of Manchester, UK) was linearized with BamHI. All other probes were prepared by PCR amplification of protein-coding regions (∼500-800 bp) from either cDNA or library clones with a bacteriophage T7 promoter at the 3′ end. Relevant primers are listed in Table S4. Probes were transcribed with MEGAscript T7 (Life Technologies) in the presence of digoxigenin-11-UTP (Roche) as previously described (Janesick et al., 2012).

Microinjected animal caps were explanted at stage 8-9 and collected into 1% agarose, 0.4× MBS Petri dishes. Each cap was transferred to individual depressions created in a 1% agarose-coated 6-well plate, in 0.4× MBS plus activin treatment (R&D Systems recombinant Activin A) or control vehicle. After overnight incubation (∼10 h), animal caps were homogenized in 200 μl TriPure (Roche). Total RNA was DNAse treated, LiCl precipitated, reverse transcribed into cDNA, and quantified in a Light Cycler 480 System (Roche) using ncam1 and histone H4 primer sets listed in Table S5 and SYBR green detection. Quantitative PCR data were analyzed by 2^−ΔCt relative to histone H4 (Schmittgen and Livak, 2008).

Embryos were embedded in 4% low-melt agarose (Bio-Rad, 1613111) in 1× PBS under a dissection microscope in disposable molds (VWR, 15160-215). The embryos were sectioned coronally in cold 1× PBS using a Leica VT1200 vibratome (50 µm thickness, 1 mm amplitude, 0.5 mm/s speed). Free-floating vibratome sections were immunostained in mesh-bottomed baskets. Sections were permeabilized (0.5% Triton X-100) for 30 min then blocked (1% bovine serum albumin, 0.2% Triton X-100) in 1× PBS at room temperature. Incubation in primary antibody against either Myod (1:3) or 12/101 (1:10) (Developmental Studies Hybridoma Bank) was carried out in blocking buffer overnight at 4°C. Sections were washed three times for 15 min each wash in 0.2% Triton X-100 in 1× PBS then incubated in secondary antibody (donkey anti-mouse Alexa Fluor 647 1:200; Thermo Fisher Scientific) along with DAPI nuclear stain (1:2000). After washing again, sections were mounted on glass slides in antifade medium with 0.12 mm spacer (Thermo Fisher Scientific, S24735). Sections were imaged at 1.0 zoom on the Zeiss LSM880 confocal microscope at 20× magnification (Plan-Apochromat 1.3 numerical aperture) using a tiling stage and Zeiss Zen Black acquisition software.

We thank Dr Stefan Heller (Stanford University) for generous use of his lab's Leica VT1200 vibratome and the Stanford Otolaryngology Imaging Core (Lars Becker) for use of the Zeiss LSM700 confocal microscope, which were essential for Fig. 6 of this manuscript.