Divergent early mesoderm specification underlies distinct head and trunk muscle programmes in vertebrates

In vertebrates, head muscles serve functions such as sensory perception (by moving the eyes), feeding (by opening and closing the jaw) and breathing, as well as vocal expression (by controlling the larynx). Their function is strikingly different from those of trunk muscles, which primarily aid in locomotion. Ontogenetically, trunk muscles develop from the segmented trunk paraxial mesoderm, i.e. somites, whereas head muscles derive from the unsegmented head/cranial mesoderm (Evans and Noden, 2006). The extraocular head muscle group derives from the premandibular head mesoderm subtype. The branchiomeric group, which develops from the pharyngeal arches, and includes jaw, laryngeal and facial muscles, shares a common pool of progenitors with the heart known as the cardiopharyngeal mesoderm (CPM; Diogo et al., 2015; Grifone and Kelly, 2007; Kelly et al., 2001, 2004; Lescroart et al., 2010, 2014, 2015; Stolfi et al., 2010; Tirosh-Finkel et al., 2006; Tzahor and Evans, 2011). The specialized head muscles, along with the chambered heart, are considered crucial for the behavioural transition from filter-feeding invertebrate ancestors to active predation among vertebrates (Diogo et al., 2015; Gans and Northcutt, 1983). Despite this enormous evolutionary importance, little is known about the early mechanisms governing specification of head mesoderm.

Molecular differences between the head and somitic trunk muscle programmes are well documented (Bothe and Dietrich, 2006; Buckingham, 2017; Hacker and Guthrie, 1998; Harel et al., 2009; Lu et al., 2002; Sambasivan et al., 2009, 2011; Tajbakhsh et al., 1997; Tzahor et al., 2003). However, the source of this divergence is obscure. Mutant mouse studies have revealed the central role of the feedback loop involving Wnt/β-catenin and FGF4/8 signals, as well as T (brachyury) and the Tbx6 T-box transcription factor network, in the specification of trunk paraxial mesoderm upstream of trunk muscles (Chapman and Papaioannou, 1998; Ciruna et al., 1997; Deng et al., 1997; Galceran et al., 1999; Herrmann, 1991; Naiche et al., 2011; Yamaguchi et al., 1999). However, the early mesoderm specification programme of premandibular mesoderm and CPM is unclear.

Mesoderm emerges from the primitive streak at the posterior pole of the embryo, induced by the concerted action of BMP4, Nodal (TGFβ), Wnt/β-catenin and FGF signals along with T (Arnold and Robertson, 2009; Ramkumar and Anderson, 2011). The late-streak population develops at the posterior pole, where sustained Wnt/β-catenin and FGF4/8 signals induce Tbx6 to specify trunk paraxial mesoderm (Arnold and Robertson, 2009; Aulehla and Pourquie, 2010; Martin and Kimelman, 2012; Mendjan et al., 2014; Ramkumar and Anderson, 2011; Turner et al., 2014). However, mesoderm from the early streak migrates to the anterior pole (Lawson et al., 1991; Tam and Beddington, 1987), wherein it downregulates T and begins to express key regulators of heart and head muscle lineages, such as Tbx1, Pitx2, Isl1, Nkx2.5, Tcf21 and Msc (Diogo et al., 2015; Kelly et al., 2004; Lu et al., 2002; Nathan et al., 2008; Sambasivan et al., 2009, 2011; Shih et al., 2008). The cues driving this developmental progression and, thus, specifying head mesoderm are unclear.

Our work reveals that the mechanism diversifying the head and trunk muscle pathways occurs at an early step in mesoderm fate commitment. Here, we show that the development of head mesoderm and head muscles is independent of T and Tbx6, unlike that of trunk paraxial mesoderm. Furthermore, in contrast to the Wnt and FGF-driven trunk pathway, we show that the CPM subtype of the head mesoderm is specified by Wnt/β-catenin and Nodal inhibitory cues of the embryonic anterior pole. Moreover, by simulating the developmental sequence specific for CPM development with external cues, we have generated progenitors with skeletal muscle and cardiac twin potential from pluripotent embryonic stem cells. Notably, by uncovering and recapitulating the ontogenetic trajectory, we have generated progenitors specific to branchiomeric muscles. This marks a significant advance towards modelling and treating diseases selectively involving head muscles.

To dissect the specification of mesoderm that generates head muscles, we assessed the requirement of T/Tbx6 pathway, which is crucial for the somitic programme of trunk muscles. T is an early marker and is expressed in all mesoderm. However, Tbx6 RNA is detected in the primitive streak around embryonic day E7 in mouse embryos (Fig. S1A; Chapman et al., 1996), suggesting induction at least in the late emerging subset of head mesoderm. Our lineage-tracing analysis using Tbx6-Cre mouse line and Pax7^GPL, a Cre-responsive tissue-specific reporter, showed robust reporter expression in all developing muscles in the embryo, including those in head (Fig. 1A). Consistently, analysis employing the ubiquitous ROSA reporter revealed expression in extraocular muscle progenitors derived from premandibular mesoderm, the core of pharyngeal arches, which generate branchiomeric muscles, and in heart (Fig. S1A-C). Thus, in addition to the documented expression in trunk paraxial mesoderm, our genetic tracing shows that the Tbx6 locus is at least transcriptionally induced in all of mesoderm that generates head muscles.

Next, we addressed the functional requirement of T and Tbx6 in muscle development from head mesoderm. Tbx6 mutants carrying the muscle reporter Myf5^nlacZ displayed no observable head muscle phenotype at E11.5 (Fig. 1B). Muscle anlage in the pharyngeal arches and in the extraocular region developed normally, as in heterozygote littermates. In stark contrast, somites below the forelimb were completely missing, whereas the cervical somites display mispatterned myotome (Fig. 1A; Chapman et al., 1996). The patterning of the forelimb muscle anlage and the tongue progenitors revealed that the migratory progenitors from cervical somites were unaffected (Fig. 1B). These results, supported by analyses of other marker transcripts such as Myf5 and Myod (both muscle markers), Twist1 (a mesoderm and head neural crest marker) and Sox10 (a proxy marker for mesoderm development) (Fig. 1C, Fig. S2A,B) confirmed unaffected head muscle mesoderm in Tbx6 mutants, as opposed to severe disruption in trunk. Analysis of head mesoderm in T null (T^Wis) mouse mutants using Myf5 reporter also revealed unperturbed development of the pharyngeal arch muscle core (Fig. 1E, Fig. S2C). Thus, the individual knockouts revealed no apparent role for T and Tbx6 in myogenesis from head mesoderm.

The core myogenic programme includes the bHLH muscle regulatory factors Myf5 and Myod. Notably, the crucial requirement for Pax3 and Tbx1, genes upstream of the core myogenic programme, has been revealed only in the Myf5 mutant background (Sambasivan et al., 2009; Tajbakhsh et al., 1997). Therefore, as a rigorous test of Tbx6 function, we analysed Tbx6:Myf5 double mutant embryos and control littermates using the Myf5^nlacZ reporter. Strikingly, X-gal staining showed unperturbed branchiomeric and extraocular muscle development (Fig. 2A). Although the extraocular muscle loss is reported in Myf5^nlacZ mutants, the phenotype is manifested only at a later stage (Sambasivan et al., 2009). Muscle progenitors from the cervical somites, which form the myotome, migratory tongue progenitors and forelimb muscle anlage displayed a phenotype similar to that of Myf5 mutants. Because deficiency downstream of Myf5 is possible, we analysed the progression in the myogenic lineage by Myod RNA in situ hybridization. Myod expression appeared unaffected in all of the six Tbx6:Myf5 double mutants analysed, with no significant reduction in the pharyngeal arches or in the cervical somites (Fig. 2B, Fig. S2D). The absence of phenotype specifically in the 1st and 2nd arches was consistent with unaffected Tbx1 expression, a key myogenic factor for jaw and facial muscles, in Tbx6 mutants (Fig. S2E). The weaker forelimb signal with both the Myf5 reporter and Myod RNA is likely due to an overall developmental delay in double mutants relative to littermates, because marker expression in age-matched comparisons at E11.5 were comparable (Fig. S2D). In summary, muscle development from head mesoderm and cervical somites was unaffected in the Tbx6:Myf5 double knockouts. In essence, the analyses of mutants underscore a broad divergence in anterior head and neck versus posterior trunk muscles at an early stage in the mesoderm specification programme.

Although there is a differential requirement for T/Tbx6, the progenitors of both neck and trunk somites experience sustained Wnt/β-catenin and FGF signalling, and T/Tbx6 expression in the posterior pole around the primitive streak. However, head mesoderm develops in the anterior end of the embryo, the signalling environment of which contrasts with the Wnt3A- and FGF4/8-secreting posterior domain. Instead, the anterior signalling centre secretes inhibitors of Wnt and Nodal, such as Dkk1, Cer1 and Lefty1 in order to establish the anterior pole of the embryonic body axis (Arkell and Tam, 2012; Lewis et al., 2008; Perea-gomez et al., 2002; Yamamoto et al., 2004). We hypothesized that Wnt and Nodal inhibitory cues instruct the distinct head mesoderm identity upstream of the divergent head muscle pathway. To address this, we first assessed whether the anterior polarity cues and head mesoderm development coincide spatiotemporally. In tune with the inhibition of Wnt signal, Axin2-d2EGFP, a faithful reporter of Wnt/β-catenin activity (Jho et al., 2002), was completely absent in the anterior domain in the head-fold stage mouse gastrulae (Fig. 3A). Cer1, which encodes a secreted inhibitor of Nodal is also expressed in this spatiotemporal domain (Fig. 3A; Biben et al., 1998). We found that the expression of head mesoderm markers Tbx1 and Pitx2 map on to the domain under the influence of the inhibitory cues of the anterior signalling centre (Fig. 3A,B).

Next, we addressed whether the anterior signalling center plays any role in CPM development in vivo. To achieve this, we employed mouse embryo ex utero culture. This technique is challenging. Therefore, we focused on Wnt inhibitory cues, as Axin2-d2EGFP mouse strain, a facile tool with which to assess modulation of the Wnt pathway, was available. Mouse embryos were collected at embryonic day (E) 7, at which point a significant proportion of head mesoderm progenitors have emigrated. In order to mask the anterior Wnt inhibitory cues, the embryos were cultured in the presence of a Wnt agonist (CHIR) for ∼24 h up to late head-fold stage and then phenotypically analysed. Embryos treated with CHIR showed expansion of the Wnt-reporter expression in Axin2-d2EGFP embryos to the anterior end of the embryo compared with vehicle control (DMSO; Fig. 3C). Even though the induction of the reporter is weak in the anterior end, the result revealed at least a partial masking of Wnt inhibition in the domain of head mesoderm development. We then analysed induction of Tbx1 as a measure of head mesoderm developmental progression in CHIR-treated wild-type embryos. Conspicuously, CHIR treatment resulted in truncation of the head region (Fig. 3C) reminiscent of the abrogation of the anterior signalling centre (Lewis et al., 2008; Thomas and Beddington, 1996). This serves as a phenotypic corroboration for the modulation of Wnt inhibition. Notably, we found substantially reduced Tbx1 expression in CHIR-treated embryos (Fig. 3C). These findings suggest a role for Wnt inhibition in either the induction of Tbx1 expression in head mesoderm or expansion/migration of head mesoderm. The technical limitation in the duration of ex utero embryo culture precluded the analysis of head myogenesis in these experiments. Nevertheless, our observations hint at a role for anterior polarity cues in head mesoderm development.

Our data indicate that the progenitors of head mesoderm are exposed to anterior polarity cues and that the cues may play a role in head mesoderm development. We addressed whether these cues confer head mesoderm identity to the nascent mesoderm population by employing a mouse embryonic stem cell (mESC) differentiation approach. Initially, embryoid bodies (EBs) were generated from mESCs and then exposed to the early streak signals BMP4 and the Nodal substitute activin A (‘mesoderm’; Fig. 4, Fig. S3A; Mendjan et al., 2014; Yang et al., 2008) for mesoderm induction. We obtained spheres of uniform size, which is key for efficient EB-mediated differentiation protocols (Fig. S3A; Behringer et al., 2016). This resulted in strong induction of T, Tbx6 and other early streak markers (Eomes, Gsc and Mesp1), whereas Tbx1 was not upregulated (Fig. 4B,C; Fig. S3B). Induction of Sox17 and Foxa2 (Fig. S3B) without upregulation of Sox2 indicated that the cells were nascent mesendoderm like, consistent with early streak identity. Subsequently, the EB-derived ‘mesoderm’ spheres were cultured with either the Wnt agonist CHIR 99021 and FGF2 (in place of FGF4/8) or with antagonists of Wnt (Dkk1 or Xav 939) and Nodal (SB431542). The former combination is reported to generate late streak mesoderm, the precursor of trunk muscle mesoderm (Mendjan et al., 2014); we refer to this as ‘posterior’ condition. The dual inhibition of Wnt and Nodal is the test that mimics cues from the anterior signalling centre. After 48 h of treatment, the cultures were assayed by RT-qPCR. Reduced phospho-Smad2/3 levels confirmed Nodal inhibition by SB431542 (Fig. S4B). Axin2 downregulation and reduction in total β-catenin levels verified effective Wnt inhibition by Xav as well as by Dkk1 (Fig. 4A, Fig. S4C). Remarkably, we found robust and selective induction of Tbx1, Isl1 and Nkx2.5, as well as Msc and Tcf21, which are key cardiopharyngeal mesoderm (CPM)/branchiomeric muscle regulators (Lu et al., 2002) upon dual inhibition (Fig. 4A; for Pitx2, see Fig. S3F). In contrast, posterior markers such as Cdx2, Cdx4 and Msgn1 were upregulated or maintained in ‘posterior’, while suppressed upon dual inhibition (Fig. 4B). Immunostaining showed Tbx1, Isl1 and Nkx2.5 expression uniquely in dual inhibition (Fig. 4B,C; see Fig. S3D for low magnification and Fig. S3E for individual channels). On the other hand, uniform induction of Cdx2, initially induced in a small proportion of ‘mesoderm’ cells, and expression of Tbx6 protein is unique in the ‘posterior’ condition (Fig. 4B,C).

Pax3 is a key somitic muscle factor, which is neither expressed nor required in head mesoderm (Hacker and Guthrie, 1998; Tajbakhsh et al., 1997). Although Pax3 RNA was detected in Xav+SB431542 (Fig. S3F), Pax3 protein was undetectable by immunostaining (not shown). We suspect that RNA of Pax3, which is also a neural marker, comes from the occasional Tbx1− Sox2+ neural progenitor clusters observed in dual inhibition cultures (Fig. S3F). In summary, based on selective induction of key markers upon dual inhibition, we conclude that Wnt and Nodal inhibitory cues specify CPM identity and possibly a pan-cranial mesoderm identity.

Our results show that inhibition of Wnt and Nodal confer CPM identity in a murine model. We tested whether these anterior polarity cues could guide similar differentiation of human embryonic stem cells (H9; hESCs; Fig. 5A). An RT-qPCR assay as well as immunostaining analyses revealed that the CPM markers were upregulated, whereas trunk mesoderm markers were repressed (Fig. 5B,C). These experiments also attest that the effect of anterior polarity cues is cell line-independent and suggests that this mechanism underlying head mesoderm/CPM identity is likely conserved across mammals.

We next tested whether inhibition of both Wnt and Nodal pathways was necessary or whether inhibition of either is sufficient. We compared Xav+SB431542 dual inhibition with treatments with either Xav alone or SB431542 alone. Treatment with inhibitors of either pathway individually resulted in significant and selective induction of the head mesoderm markers comparable with dual inhibition (Fig. S4A). Remarkably, the treatment with Xav alone resulted in significant reduction in pSmad2/3 levels (Fig. S4B). On the other hand, treatment with SB431542 alone led to a dramatic dip in total β-catenin levels (Fig. S3C). Thus, inhibition of either pathway appears to indirectly attenuate the other as well, suggesting that Nodal and Wnt pathways are tightly interconnected in early mesoderm. In fact, extensive crosstalk between Wnt/β-cat and Nodal-TGFβ pathways during development is known (Skromne and Stern, 2001; Guo and Wang, 2009) and Nodal induces Wnt to initiate gastrulation and mesoderm formation (Arnold and Robertson, 2009; Ramkumar and Anderson, 2011). In particular, TGFβ activates several Wnt ligands and elevates β-catenin levels (DiRenzo et al., 2016). Similarly, Wnt/β-catenin signal induces the expression of Nodal as well as its co-receptor Cripto (Morkel et al., 2003; Rodríguez-Esteban et al., 2001). In the context of CPM, the mechanism interlinking the inhibition of both the pathways remains to be explored.

Using the small molecules iCRT3 and iCRT5, which specifically inhibit β-catenin responsive transcription by blocking β-catenin – Tcf interaction (Gonsalves et al., 2011), we distinguished between the transcription function of β-catenin and its function in the cell membrane. The treatment with iCRTs also resulted in selective induction of most CPM markers assayed (Fig. S4A; iCRT3 data not shown). Thus, with regard to Wnt signal, inhibition of β-catenin-mediated transcription regulatory function may be key to induce CPM.

We have demonstrated that Wnt and Nodal inhibition impinge on the regulatory programme of CPM. To investigate further, we tested the potential of dual inhibition-derived cultures to differentiate into skeletal muscle and cardiomyocytes. This is a decisive test for CPM identity (Diogo et al., 2015; Grifone and Kelly, 2007; Tzahor and Evans, 2011). To achieve this, we treated dual inhibition-derived mesoderm with previously published muscle-induction regimen (Chal et al., 2015). Briefly, post dual inhibition, the cells were cultured for 2 days in media containing LDN193189 (LDN; a BMP inhibitor), HGF, FGF2 and IGF1. Subsequently, the cultures were maintained in the same growth factor cocktail minus LDN for 12 days. During this differentiation, cells in the culture began aligning and fusing with each other to differentiate into long multi-nucleated myotubes (Fig. 6A). Immunofluorescence for muscle markers Myod, Myog and desmin revealed large clusters of cells differentiating into skeletal muscle myotubes (Fig. 6A). This downstream skeletal muscle differentiation in Xav+SB431542-treated cultures is comparable with that of ‘posterior’ cells (Fig. 6A). Thus, the dual inhibition-specified mesoderm responds to myogenic cues to make skeletal muscle.

Next, we assessed the cardiogenic potential of Wnt- and Nodal-inhibited mesoderm. In fact, Wnt inhibition has been documented to be key in driving heart differentiation (Marvin et al., 2001; Schneider and Mercola, 2001; Yang et al., 2008). To address cardiogenic potential, dual inhibition cells were treated with ascorbic acid (AA) and VEGF (Cao et al., 2012; Kokkinopoulos et al., 2016; Takahashi et al., 2003), along with FGF2 and FGF10. The embryoid body-derived spheres differentiated readily and efficiently into beating cardiomyocyte clusters, which expressed cardiac markers such as cardiac troponin T (cTnT) and α-actinin (Fig. 6B). In contrast, mesodermal cells from ‘posterior’ condition never generated beating cardiomyocyte clusters. A few cTnT/α-actinin positive clusters, very small compared with the dual-inhibition condition, were observed in ‘posterior’, which were scored as positive (Fig. 6B). Notably, in this ‘cardiogenic’ milieu, the dual-inhibition cultures also generated clusters of myotubes positive for Myod, Myog (Fig. S5A) and desmin (not shown). Together, these results attest to the twin cardiac/skeletal muscle potential of mesoderm treated with Xav+SB431542. Moreover, the twin potential conferred by dual inhibition is not cell line specific, as we have replicated the differentiation assay using a different mouse ES line (Fig. S5B).

As previously reported for a similar in vitro cell population, poor survival of the cells precluded clonal analysis (Chan et al., 2016; Swartz et al., 2016) and, thus, the test of bipotency in Xav+SB431542 culture. Instead, we assessed expression of PDGFRα and Flk1 to assess the proportion of cells that may contain bipotent cells. Flk1 is a marker of lateral plate mesoderm (Kataoka et al., 1997) and PDGFRα+ Flk1+ cells represent cardiac fated population (Kattman et al., 2011), whereas a PDGFRα+ Flk1− subset may contain bipotent cells (Chan et al., 2016). We find that around 30% of the Xav+SB431542 cultures are PDGFRα+ Flk1−, and that uniquely Flk1+ and the double-positive subsets are minor (Fig. 6C). Although, these experiments do not provide direct evidence, they suggest that the dual inhibition-derived CPM-like population may contain bipotent cells. In summary, our data highlight the twin potential of Wnt and Nodal inhibited mesoderm, which attests to its identity as the progenitor of branchiomeric muscles.

Head muscles are functionally and ontogenetically distinct from somite-derived muscles and are a vertebrate novelty. Recent discovery that both branchiomeric head muscles and heart emerge from the unique ‘cardiopharyngeal’ developmental unit has triggered great interest in head muscle development and evolution. We reveal that head and trunk muscle programmes are inherently different by demonstrating that mesoderm patterning and specification by body axis polarity cues underlie the divergence (Fig. 7). This advance in understanding the development of head muscles offers insight into their evolutionary origin. Furthermore, we have uncovered the developmental cues specifying CPM with skeletal muscle/heart twin potential. This has allowed us to generate skeletal muscle of CPM lineage from pluripotent stem cells in vitro, which has the potential to enable efforts in modelling and treating muscular dystrophies differentially involving head muscles.

Emergence of specialized head muscles is key for the switch from filter feeding to active predation associated with vertebrate evolution (Gans and Northcutt, 1983). Whether this novel group arose from modification and elaboration of trunk programme or evolved independently is a key issue (Diogo et al., 2015; Onai et al., 2015; Sambasivan et al., 2011; Stolfi et al., 2010; Wachtler and Jacob, 1986). Our findings that Wnt inhibition confers CPM identity, upstream of the documented role in inducing branchiomeric myogenesis (Tzahor et al., 2003), indicates that Wnt inhibition is a recurrent theme in head muscles. This contrasts with the persistent role of Wnt/β-catenin pathway in the somitic muscle programme. In this context, the emergent theory that the opposing cues of Wnt have a deeply conserved role in establishing anterior-posterior body axis polarity (Niehrs, 2010; Petersen and Reddien, 2009) is significant. We suggest that CPM specification by Wnt inhibition, the conserved anterior polarity cue, reflects a deeper evolutionary origin of CPM and branchiomeric muscles independent of the Wnt-driven somitic programme.

Differences in the transcription factor regulatory network of head and somitic myogenesis are well documented. Whereas the somitic muscles are Pax3 driven (Tajbakhsh et al., 1997), head muscles are Pax3 independent and are governed by Tbx1, Isl1 and Pitx2 (reviewed by Sambasivan et al., 2011). Our Tbx6 and T mutant studies provide compelling evidence for the independence of the Tbx1/Isl1 head regulatory programme from T/Tbx6 mesodermal network. These findings point to either redundancy between T and Tbx6 or to a pathway parallel to T/Tbx6, upstream of the head mesoderm programme. These possibilities remain to be addressed. In addition, dissecting the mechanistic link between Wnt and Nodal dual inhibition and the Tbx1/Isl1 network is important to fully understand head muscle programme.

The early primitive streak signals BMP4 and Nodal pattern ESCs into mesoderm with cardiogenic potential (Mendjan et al., 2014). However, a variety of downstream cues, such as BMP2, retinoic acid and ascorbic acid in combination with VEGF and Wnt inhibition, have been employed for cardiac differentiation (Burridge et al., 2013; Keller, 2005). Whether these cues drive uniquely the cardiac lineage or broadly specify the bipotent CPM remained obscure. Investigating the inductive role for surrounding tissues on cardiac mesoderm in chick embryos, a pioneering study revealed that Wnt inhibition from anterior endoderm drives cardiogenesis (Marvin et al., 2001). Taking a similar approach to investigate the role of niche signals, we show that the body axis cues establishing anterior polarity, Wnt and Nodal antagonists, specify dual potential CPM. Remarkably, these antagonists are also secreted by anterior endoderm in mouse embryos (Stower and Srinivas, 2014). Taking these pieces of evidence together, we suggest that the BMP/Nodal primitive streak signals represent the early ‘permissive’ cue, whereas Wnt and Nodal dual inhibition at the anterior pole constitutes the ‘instructive’ cue that acts in tandem to progressively specify CPM. Unravelling this developmental mechanism of CPM identity specification has allowed us to guide ESCs towards CPM-like progenitors with dual potential.

The ability to derive progenitors specific to muscle groups with distinct developmental history is pivotal to address disease mechanisms of muscular dystrophies. Directed ESC differentiation into skeletal muscle based on the trunk muscle programme has been reported only recently (Chal et al., 2015; Jiwlawat et al., 2018). To date, generation of cranial mesoderm-derived muscles from pluripotent cells has been elusive. Notably, forced expression of the bHLH factor Mesp1 programmes ESCs into cardiac/skeletal muscle bipotent CPM lineage (Chan et al., 2016). However, guided differentiation without transgene introduction is key for obtaining a regenerative cell type for therapy. Developmental signals specifying CPM identified here allowed us to simulate the developmental trajectory of the head muscle group. Such a strategy to recapitulate the ontogenetic programme appears crucial for efficient directed muscle differentiation (Chal et al., 2015) and may allow generation of progenitors specific to various muscle groups. We anticipate that our method will have a major impact by enabling generation of in vitro models of muscular dystrophies affecting head muscles as well as generating stem cells for therapy of dystrophies.

Animals were handled according to Committee for the Purpose of Control And Supervision of Experiments on Animal (national) guidelines and the institutional animal ethics committee approved all animal experiments. Wild type refers to B6D2 (F1 of C57Bl/6J crossed to DBA/2) animals. Strains used are listed in Table S1.

The embryos were fixed for 40 min to 1 h in 4% PFA, washed in PBS and stained overnight at room temperature in a standard X-gal staining solution.

Whole-mount RNA in situ hybridization was performed following an established protocol (Henrique et al., 1995). The incubation period for proteinase K treatment was empirically determined for different stages of the mouse embryos. The antisense riboprobes were generated by in vitro transcription incorporated with digoxigenin-11_UTP (Roche). Washes following the incubation with alkaline phosphatase (AP)-conjugated anti-digoxigenin-11 antibodies were extended for 48 h and BM Purple (Roche) was used as the substrate for AP. For reliable comparison of in situ hybridization signals between mutants and control littermates or age-matched controls, each set of embryos was hybridized in the same tube. In the case of ex utero cultures, the vehicle-treated and CHIR 99021-treated embryos were hybridized in different tubes, as the embryos from the two groups cannot be distinguished. However, both sets were hybridized under the same conditions in parallel.

Mouse ES cells (E14TG2a, E14TG2a T-GFP) were maintained in GMEM media with 10% serum and leukaemia inhibitory factor (LIF) on gelatin-coated dishes. Differentiation assays were performed in DMEM/F12 and Neurobasal (1:1) medium with N2 and B27 (minus vitamin A, Invitrogen, 12587-010) supplements. Differentiation was initiated by embryoid body (EB) formation; 3×10⁵ cells/ml were plated in individual wells of a six-well low-attachment (Corning/Sigma, CLS 3471) dish for 48 h. This was then followed by growth factor and small molecule treatments of EBs in suspension for differentiation assays. For muscle and heart differentiation, EB-derived spheres from the Xav+SB431542 or posterior culture conditions were plated on a dish with StemPro34SF media (for cardiac fate; Invitrogen, 10639011) and in DMEM with KnockOut Serum Replacement (ThermoFisher) supplemented with bovine serum albumin (for skeletal muscle). The concentrations of factors used are listed in Table S2.

Human ES cell differentiation: H9 cells (WiCell, SLA# 16-W0026) were cultured in mTESR1 (Stem Cell Technologies). Differentiation was induced using a protocol similar to mouse ESC differentiation that excludes the formation of embryoid bodies prior to small molecule/growth factor induction. Differentiation was carried out in chemically defined media (Vallier and Pedersen, 2008).

Wild-type embryos were collected at E7 and dissected in M2 media (Sigma). They were cultured on roller tubes in media (DMEM/F12 with Glutamax and penicillin-streptomycin) containing 50% rat serum (home-made) at 37°C, 5% O₂, 5% CO₂ for 22-24 h ex utero in the presence of DMSO (control) or CHIR99021 (8 µM).

For immunostaining, EB-derived ‘mesoderm spheres’ were stained in suspension. Quantitation of marker-positive nuclei was performed on optical sections through the spheres. For differentiation into cardiomyocytes and muscle, the spheres were plated. The attached cells originating from individual spheres aggregated into spheres and differentiated into big clusters of cardiomyocytes. Skeletal muscle differentiation also occurred in clusters and not uniformly in the dish. This precluded quantitation as the percentage marker-positive nuclei/cells in the population. We plated defined number of spheres per dish and each of these could be distinguished, as part of each sphere remained as a dome due to partial dispersal following attachment. Therefore, quantitative analysis measured the percentage of spheres associated with marker positive clusters of cells. For cultured cells, samples were fixed in 4% paraformaldehyde and washed with PBS. Permeabilization was carried out with 0.1% Triton-X in PBS followed by blocking in 5% normal donkey serum. Primary antibody was incubated overnight. This was followed by washes with 0.1% Tween-20 in PBS and incubation with secondary antibody. Cells were imaged following further washes with 0.1% Tween-20 in PBS. For whole-mount and EB-derived sphere immunostaining, samples were fixed in 4% paraformaldehyde and washed with 0.1% Tween-20 in PBS. Permeabilization was carried out using 0.3% Triton-X in PBS followed by blocking in 10% normal donkey serum. Primary antibody was incubated overnight. This was followed by washes with 0.1% Tween-20 in PBS and incubation with secondary antibody. Antibodies and dilutions are listed in Table S3. Olympus FluoView FV1000 confocal scanning microscope and software was used to acquire images. Optical sections (taken at 1-3 µm intervals) were z-stacked using ImageJ (NIH; Schneider et al., 2012). Statistical analysis for quantitation of marker-positive nuclei used one-way ANOVA followed by Tukey's post hoc test for multiple pairwise comparisons among differentiation conditions.

Total RNA (500 ng-1 µg) extracted using the Qiagen RNeasy Micropurification Kit was used to prepare cDNA (Super-Script III, Invitrogen). Oligo dT primers were used for cDNA synthesis from total RNA. Real-time PCR reactions were performed using PowerSYBR Green Universal Master Mix (Thermo Scientific) and analysed using ViiA7 Real Time PCR system (Applied Biosystems). Data were normalized using GAPDH expression. For all targets, primer efficiencies were calculated with serial dilutions of total cDNA and only those primer sets, which matched the efficiency of amplification of GAPDH internal reference primer set were used for the study. The specificity of the primers was also tested by melt curve. The primers are listed in Tables S4 and S5. Each condition was tested using two technical replicates and at least three biological replicates was analysed for relative gene expression using 2^−ΔΔCt method (Schmittgen and Livak, 2008). Error bars on all qPCR graphs represent s.e.m. Unpaired Student's t-tests (two-tailed) or one-way ANOVA followed by Dunnett's post hoc test for comparing individual inhibitor treatments to ‘posterior’ (many-to-one comparison) were performed.

For flow analysis, the cells were stained with PE-conjugated anti-Flk1 and Superbright 436-conjugated anti-PDGFRα antibodies. Briefly, EBs derived from E14Tg2a T-GFP cells (a gift from P. Chandra Shekar, CCMB, Hyderabad) were attached and differentiated by dual inhibition as described above. Following treatment, cells were dissociated with Accutase. The cell suspension was washed with 1× PBS+0.5% BSA. 1×10⁶ cells in 100 µl suspension were stained with 0.5 µg Flk1 and 1 µg PDGFRα antibodies for 1 h, washed with 1×PBS+0.5% BSA three times. Live/death discrimination was carried out using 7-AAD (10 µg/ml). Analysis was performed using FACSVERSE (BD Scientific) and data analysed using FlowJo.

Protein (20 µg) samples were fractionated by electrophoresis on 10% SDS-PAGE gel, transferred on a nitrocellulose membrane (GE10600016, GE Healthcare), and incubated overnight at 4°C with primary antibodies and then with appropriate HRP-conjugated secondary antibodies (Table S3). ECL kit (32209, ThermoFisher Scientific) was used for visualization. Quantification of the bands was carried out using Fiji software. One-way ANOVA followed by Tukey's post hoc test was performed to analyse statistical significance.

We thank the animal facility (ACRC-NCBS/inStem), the Central Imaging and Flow Cytometry Facility, NCBS and the inStem Stem Cell Facility. We thank Upnishad Sharma and Payel Das for technical help, Dr Raj Ladher for critical reading of the manuscript and Profs Shahragim Tajbakhsh, Jyotsna Dhawan and Maneesha Inamdar, and Dr Michael Schubert for discussion.