In vertebrates, successive waves of myoblasts contribute in building the skeletal striated muscle tissue (Kahane et al., 1998a and b; Kahane et al., 2001). In amniote somites, two myogenic waves essentially derived from the dermomyotome have been identified (Figure 2A). The first embryonic wave is at the origin of muscle fibers that form the primary myotome. The second wave participates in myotome growth (Gros et al., 2004 and 2005). In Xenopus and zebrafish, the first signs of myogenesis appear early, during the blastula/gastrula transition, long before somite formation (Hopwood et al., 1992; Weinberg et al., 1996; Coutelle et al., 2001; Polli and Amaya, 2002). The myogenesis of skeletal muscle is orchestrated in vertebrates by the four bHLH transcription factors of the Myod family (Comai et al., 2014; Hernández-Hernández et al., 2017). They were classified as master genes since the four members are able to convert fibroblasts into skeletal muscle cells (Davis et al., 1987; Braun et al., 1989; Rhodes and Konieczny, 1989).

Based on the expression of the four myogenic regulatory factors (MRFs) of the Myod family (Myod1, Myf5, Myf6, and Myogenin), three myogenic waves have been identified in Xenopus (Grimaldi et al., 2004; Della Gaspera et al., 2012a) (Figure 2B). The first wave at the origin of the primitive myotome appears before dermomyotome formation and is composed of two myogenic populations differentiating in different places and times (Figure 2B). Indeed, from stage 9.5, Myf5 and Myod1 start to be expressed in the dorso-lateral marginal zone and initiate medial myogenesis, which gives rise to the first differentiated fibers located near the notochord at the gastrulation/neurulation transition (Gurdon et al., 1984; Hopwood et al., 1989; Frank and Harland, 1991; Hopwood et al., 1991; Scales et al., 1991; Kato and Gurdon, 1993; Fisher et al., 2002; Polli and Amaya, 2002; Della Gaspera et al., 2012a). During gastrulation, Myod1 expressed in the marginal zone also participates in the activation of genes involved in somitogenesis and seems to link myogenesis to somite formation (Maguire et al., 2012). At the beginning of neurulation from stage 13, a hallmark of lateral myogenesis is the strong expression of Myod1 mRNA in the lateral paraxial mesoderm. These myogenic cells differentiate during somitogenesis (Frank and Harland, 1991; Della Gaspera et al., 2012a). The muscle fibers corresponding to these two myogenic populations are distributed separately at tailbud stage. The dorso-lateral cells of the marginal zone (medial myogenesis) remain associated with the notochord in the head and trunk regions, whereas most of the ventral cells (lateral myogenesis) give rise to muscle fibers that envelop the medial ones during neurulation (Krneta-Stankic et al., 2010). The formation of the first muscle fibers presents a peculiarity in Xenopus since it is initiated in the head somites (w, x, y, and z) of the preotic region (Chung et al., 1989; Della Gaspera et al., 2012a). These singular somites gradually disappear during the next embryonic phases (Chung et al., 1989). In Xenopus, most of the identified genes of the muscle development program are implicated in the formation of the primitive myotome and somites suggesting that both are interconnected. Among them, three genes, Hes6.1 (Hes6), Egr1, and Mef2d act downstream of Fgf signaling, illustrating the important function of Fgf in the formation of the paraxial mesoderm as well as the primitive myotome (Murai et al., 2007; Nentwich et al., 2009; Della Gaspera et al., 2012b). Other transcription factors are also involved in the formation of the primitive myotome in Xenopus. Ebf2 and 3 play a role in muscle specification and in some aspects of differentiation (Green and Vetter, 2011). Sox5 enhances indirectly myogenic transcription through transrepression (Della Gaspera et al., 2018). Other factors involved in the RNA metabolism [Rbm24 (Seb4), Trab2, and Mir-206] and cell cycle decision [Cdknx (p27Xic-1)] also promote primitive myotome formation (Vernon and Philpott, 2003; Li et al., 2010; Dichmann et al., 2015; Vergara et al., 2018).

In zebrafish, the expression of Myf5 and Myod1 is initiated precociously, during gastrulation in the medial region of paraxial mesoderm (Weinberg et al., 1996; Coutelle et al., 2001). During the following segmentation period (there is not strict distinction between segmentation and neurulation in zebrafish), the lateral expression of Myf5 and Myod1 is also visible in the posterior region of each somite (Kimmel et al., 1995; Groves et al., 2005; Stellabotte et al., 2007). The primitive myotome in zebrafish is also made up of different muscle fiber populations (see Muscle Fiber Population of the Primitive Myotome) and is also the first somitic compartment to form before the dermomyotome (Stellabote et al., 2007).

The dermomyotome is defined as the part of a somite capable of generating both the dorsal dermis and the myotome (Christ and Ordahl, 1995). In amniotes, the four edges of the dermomyotome are the source of myoblasts, which will form the primary myotome, whereas the central region of the dermomyotome contains progenitor cells common to the dermis and the muscle (Gros et al., 2004; Ben-Yair and Kalcheim, 2005) (Figure 2A). These progenitors proliferate in the plane of the dermomyotome, and when they divide perpendicular to the plane of the dermomyotome, dermal and myogenic progenitors are generated dorsally and ventrally, respectively (Ben-Yair et al., 2011). Pax3 and Pax7, which play a role upstream of the Myod family of transcription factors, have been identified as dermomyotome marker in amniotes (Tajbakhsh et al., 1997; Relaix et al., 2005; Bajard et al., 2006). The Six family of transcription factors have also been identified as upstream factors during dermomyotome formation and skeletal myogenesis in mice (Grifone et al., 2005).

The existence of a dermomyotome in anamniote embryos has long been discussed, until Devoto et al. (2006) suggested that somites of all vertebrate embryos have a dermomyotome compartment. In cephalochordates, histological analyses show that the somitic dorso-lateral external cells are not equivalent to the vertebrate dermomyotome (Holland, 1996; Holland et al., 1999; Mansfield et al., 2015). However, pax3/7 has been detected in the dorso-lateral region of amphioxus somites indicating that these external cells possess some features of vertebrate dermomyotome (Yong et al., 2021). Moreover, the lateral somitic domain of amphioxus can be subdivided at mid/late neurula stage into three subdomains with specific expression of Pax3/7, Pax1/9, and Hand genes suggesting that somitic compartmentalization already exists in the ancestral chordates (Figure 3A). Pax3 and Pax7 orthologs are expressed in the dorso-lateral cells of somites in lampreys (Kusakabe and Kuratani, 2005), zebrafish (Groves et al., 2005), and sturgeons (Devoto et al., 2006), suggesting that the thin sheet of dorso-lateral cells of anamniotes is homologous with the amniote dermomyotome. Although dermal and myogenic bipotent progenitors have not been identified in anamniotes, the expression of collagen genes in the dorso-lateral cells of somites in teleosts and Xenopus argues that the dorsal dermal lineage originates from the dermomyotome (Grimaldi et al., 2004; Le Guellec et al., 2004; Rescan et al., 2005).

In Xenopus, a thin dorso-lateral tissue distinct from the underlying myotome has been first assimilated to a dermatome, at the origin of the dorsal dermis (Ryke, 1953; Hamilton, 1969). Moreover, Grimaldi et al. (2004) showed that this region is similar to the amniote dermomyotome since it expresses Pax3 and gives rise to myogenic cells at epaxial and hypaxial levels (Martin and Harland, 2001). The myogenesis at epaxial and hypaxial regions is in addition to that of the primitive myotome and has therefore been described as a second myogenic wave in Xenopus (Grimaldi et al., 2004; Della Gaspera et al., 2012a) (Figure 2B). Hence, it can be considered that the first wave of primitive myotome and the second wave of dermomyotome participate both to the formation of the primary myotome, which is initially made up of mononucleated fibers extending across a somite in an antero-posterior direction as is the case in amniote vertebrates (Gros et al., 2004) (Figure 2B). In Xenopus, there is a third myogenic wave at stage 36–37 characterized by the presence of Myf5-expressing cells within the myotome, which could participate in the growth of the myotome and the formation of multinucleated fibers from stage 45 onward (Grimaldi et al., 2004; Della Gaspera et al., 2012b) (Figure 2B). From stage 48, myogenin has also been identified in some isolated mononucleated cells located at the periphery of large larval myofibers (Nicolas et al., 1998; Chanoine and Hardy, 2003). This third wave undoubtedly corresponds with previously histologically identified mesenchymal cells that participate to the secondary myogenesis during the larval phase (Boudjelida and Muntz, 1987; Radice et al., 1989; Kiełbówna and Daczewska, 2005). Similarly, from stage 45–46 onward, satellite cells expressing Pax7 and positioned under the basal lamina are also observed within the myotome (Chen et al., 2006; Daughters et al., 2011). These satellite cells originate at the early neurula stage from the dorso-lateral region of the mesoderm, later identified as the LSF at the origin of both dermomyotome and sclerotome (Daughters et al., 2011; Della Gaspera et al., 2012b; Della Gaspera et al., 2019). Although this has not been directly demonstrated by lineage studies, these results suggest that satellite cells and the third-wave myoblasts are also derived from the dermomyotome in Xenopus like in amniotes and zebrafish (Gros et al., 2005; Relaix et al., 2005; Hollyway et al., 2007; Seger et al., 2011). On the other hand, it can be also noted that the dermomyotome (and/or sclerotome) contributes to the mesenchyme of dorsal fin in amphibians (Sobkow et al., 2006; Garriock and Krieg, 2007). In conclusion, the dermomyotome in anamniotes and amniotes possesses similar functions.

The function of the four myogenic regulatory factors (MRFs) of the Myod family (Myod1, Myf5, Myf6, and Myogenin) in the specification, determination, and differentiation of muscle fibers has been studied in numerous single, double, and triple knockout experiments in mice (Bismuth and Relaix, 2010; Yamamoto et al., 2018). Single Myf5 and Myod1 knockout mice survive whereas the double knockout mice die due to the absence of myoblasts (Rudnicki et al., 1993; Kassar-Duchossoy et al., 2004) (Figures 4A–C). Thus, Myf5 and Myod1 were defined as determination factors that could partly compensate each other. For instance, the existence of two myogenic lineages that could compensate each other has previously been proposed, one that is Myf5-dependent and the other being Myf5-independent, and probably driven by Myod1 (Haldar et al., 2008). Experiments based on the conditional expression of cell-killer gene, the diphtheria toxin, activated by the Cre recombinase under the control of the Myf5 promoter indicate that Myod1 is unable to compensate the lost Myf5 lineage suggesting that compensation between Myf5 and Myod1 is due to a functional redundancy of the two proteins (Comai et al., 2014). Surprisingly, the morphant and mutant studies in zebrafish have come to slightly different conclusions (Figure 4D). In zebrafish, double mutants for Myf5 and Myod1 have any myoblasts but Myod1 mutants also die since Myod1 drives lateral fast fibers myogenesis in somites and is essential for cranial myogenesis (Hinits et al., 2009; Hinits et al., 2011). Another study with morpholino oligos have found that both Myf5 and Myod1 are each necessary to the development of some muscle anlagen in the head (Lin et al., 2006). In Xenopus, analysis of the role of the three MRFs (Myod1, Myf5, and Myf6) during primitive myotome formation using gene knockdown experiments revealed that Myod1 is necessary for lateral myogenesis (Della gaspera et al., 2012b) as previously observed in zebrafish (Hinits et al., 2009). Hence, in anamniotes, Myf5 cannot fully compensate the absence of Myod1. During Xenopus head myogenesis, each of these two determination factors is weakly expressed or not expressed with the other one in some anlagen (Della Gaspera et al., 2012a). It is also the case for Myf5 in zebrafish that is absent from superior, medial, and lateral rectus muscles anlagen (Lin et al., 2006). Indeed, Myf5 and Myod1 have not the same abilities to initiate the differentiation. Myod1 is the better inducer, suggesting that variation of expression of each MRF inside the same anlage could have an effect on differentiation timing (Ishibashi et al., 2005; Della Gaspera et al., 2012a; Conerly et al., 2016).

Regarding the in vivo function of the differentiation factor Myogenin, the knockout mice die due to the absence of differentiated muscle fibers whereas in zebrafish, the Myogenin mutants survive with defects in myocyte fusion (Hasty et al., 1993; Ganassi et al., 2018). In mice, it was also shown that Myogenin was not necessary for the formation of the primary myotome (Venuti et al., 1995) (Figures 2A and 4A). In Xenopus, Myogenin is weakly expressed during the first myogenic wave and strongly expressed at the hypaxial and epaxial edges during the second myogenic wave at stage 32–34 suggesting that Myogenin may be necessary for the formation of the multinucleated fibers from stage 45 in Xenopus (Kielbowna and Dacszewska, 2005; Della Gaspera et al., 2012a) (Figure 2B).

Distinct functions have also been highlighted for Myf6 between mice and anamniotes. In mice, Myf6 acts as both a determination and a differentiation factor. Myf6 is able to initiate myogenesis in the absence of Myf5 and Myod1 during myotome formation but not during fetal myogenesis (Figures 4A–C). Furthermore, Myf6 acts upstream of Myod1 during extraocular myogenesis (Kassar-Duchossoy et al., 2004; Sambasivan et al., 2009). In contrast, in zebrafish, Myf6 is unable to initiate myogenesis in the double Myf5/Myod1 mutants, and in Xenopus, Myf6 is always the last MRF expressed in the head and in the myotome (Schnapp et al., 2009; Hinits et al., 2011; Della Gaspera et al., 2012a) (Figure 4D). Until now, the results obtained in anamniote species suggest that the role of Myf6 as a determination factor is not an ancestral function.

In any case, it seems that the transition from anamniotes to amniotes impinges on the MRF functions not just in somites but also in the head. The reorganization of MRF core networks could be an indication of the evolution of the compartmentalization process in somites as we describe it, and could mean that head myogenesis has also dramatically evolved. Recently, the phylogenic tree of Myod family has been redefined (Aase-Remedios et al., 2020) (Figure 4E). The ancestral vertebrate could possess only two clustered MRF genes, one acting early like a determination factor and the other acting late like a differentiation factor. The cyclostome vertebrate species, lamprey and hagfish, each have only two MRF genes, both derived from the same early MRF gene. The increase in gene number of the Myod family with four members derived from the early and late ancestral MRFs in gnathostome vertebrates has probably contributed to the higher complexity of muscle formation and composition.

In vertebrates, the sclerotome is mainly at the origin of the vertebrae and ribs but vascular and tendon cells also derive from specialized sclerotome parts (Scaal, 2016; Tani et al., 2020). In amniotes, the newly formed somite subdivides into two easily discernable compartments, the sclerotome that represents about half of the somite size and the dermomyotome. The dermomyotome consists of cells that remain in epithelial state, whereas the sclerotome cells undergo EMT (epithelial–mesenchymal transition) and migrate around the notochord and neural tube. While the sclerotome is a relatively large compartment in amniotes, it is reduced to a thin sheet of cells on the ventro-medial side of the huge myotome in anamniotes (Figure 1A).

In Xenopus, initial morphological and histological works reported the identification of some polymorphic cells at the late tailbud stage in the ventro-medial edge of the somites that could constitute the sclerotome compartment (Ryke, 1953; Youn and Malacinski, 1981b). At the later stages, in both urodele and anuran species, sclerotomal cells migrate into the perinotochordal and perineural space and give rise to the axial skeleton (Mookerjee, 1930; Mookerjee, 1931). The late formation and migration of the sclerotome cells in amphibian is supposed to be due to delayed vertebral development at the end of the larval stage (Scaal and Wiegreffe, 2006). The ventro-medial location of sclerotome cells was confirmed more recently by in situ hybridization analysis with the main specific markers of the sclerotome which are not expressed in the dermomyotome and the myotome such as, Pax1 and Pax9 but also with other markers such as Twist1, Uncx, Foxc1, and Foxc2 mRNAs (El-Hodiri et al., 2001; Sánchez and Sánchez, 2013; Della Gaspera et al., 2019). Two or three domains in the sclerotome (dorsal, notochord-associated, and ventral) have been characterized in zebrafish and Xenopus, on the basis of differential expression of markers (Sánchez and Sánchez, 2013; Ma et al., 2018; Sánchez and Sánchez, 2021). The dorsal domain could also contribute to dorsal fin at least in fish (Freitas et al., 2006; Ma et al., 2018). Until now, little is known about the gene functions involved in the sclerotome formation in anamniotes. Knockdown experiments of Pax1, Pax9, and Twist1 in medaka have confirmed the essential function of these genes in sclerotome formation with a pronounced subfunctionalization for Pax1 and Pax9 (Yasutake et al., 2004; Mise et al., 2008). Any functional study of the sclerotome genes has yet to be done in Xenopus. Due to the late formation of sclerotome, long after the dermomyotome and myotome, its developmental origin in anamniotes has been questioned; it could derive from myotome cells or from a separate population still unidentified (Keller, 2000). Indeed, we recently identified Twist1 as a marker of migrating sclerotome progenitors in two amphibians, Xenopus and axolotl, and showed that both sclerotome and dermomyotome cells originate from a cell population located at LSF revealing the ancestral location of MSCs (Della Gaspera et al., 2012b; Della Gaspera et al., 2019).

In amniotes, the syndetome is the somitic compartment at the origin of the dorsal tendons (Brent et al., 2003) (Figure 1A). Syndetome is a dorsal compartment of the sclerotome, between two neighboring myotomes, induced by fgf8 secreted by muscle cells (Brent et al., 2003; Brent and Tabin, 2004). It could be noted that the acquisition of tendon fate is highly dependent of local cellular environment. Double knockout mice for Sox5 and Sox6, two transcription factors involved in chondrocyte differentiation saw an expansion of the syndetome at the expense of cartilage differentiation only in the somites, whereas limb tendons form normally (Brent et al., 2005). Similarly, muscle tissue and fgf signaling are necessary for tendon progenitor specification in the somites only, whereas muscle tissue remains still essential for tendon differentiation in the limb and head (Kardon, 1998; Bonnin et al., 2005; Brent et al., 2005; Grenier et al., 2009).

The earliest and specific marker of tendons and ligaments is Scleraxis, a bHLH transcription factor of the Twist family (Schweitzer et al., 2001) that regulates genes involved in tendon differentiation such as Tenomodulin and Col1a1 encoding two extracellular matrix proteins (Shukunami et al., 2006 and, 2018; Léjard et al., 2007). Scleraxis is also necessary for the recruitment of some tendon progenitors at the elongation sites of the longest tendons (Huang et al., 2019). However, Scleraxis is not the only master gene of tendinogenesis since the tendon progenitors appeared normal in Scleraxis knockout mice (Murchison et al., 2007). Nevertheless, in these mice, some of the tendons present a defect in differentiation with the extracellular matrix less organized, tenomodulin expression lost, and collagen I expression reduced. A somitic scleraxis-positive compartment, giving rise to tendons, was also identified in Xenopus (Della Gaspera et al., 2009) and in the following years in fish species (Bricard et al., 2014; Ma et al., 2018). In zebrafish and probably in Xenopus, tenocytes adopt a particular cellular morphology with cell bodies positioned at the sclerotome edge and cytoplasmic extensions of tree-like processes slip into intersomitic space at myotendinous junctions. In zebrafish, as well as in trout and Xenopus likely, tenocytes originate from sclerotome indicating that the generation of tenocytes from the sclerotome dates back at least to the last common ancestor of the sarcopterygians and the actinopterygians (Bricard et al., 2014; Ma et al., 2018). Tendon development studies in zebrafish have confirmed that muscle tissue is only necessary for the formation of dorsal tendons, but it is still important for the differentiation of the fin and the cranial tendons (Chen and Galloway, 2014). More recently, crispr/cas9 has permitted the generation of Scleraxis mutants where both genes coding for Scleraxis (a and b) are mutated, leading to tendon differentiation defects particularly in the head and to deficiencies in rib mineralization and muscle growth. These results illustrate the co-development of the musculoskeletal system (Kague et al., 2019).

To date, Scleraxis is the earliest and most specific tendon marker. However, we have observed Mef2c expression in Xenopus syndetome before Scleraxis (Della Gaspera et al., 2009). Mef2c belongs to the MEF2 family of transcription factors with Mef2A, B, and D in gnathostome vertebrates. They are involved in the development of numerous mesoderm derivatives like smooth, cardiac, and skeletal muscles, and also in neuron differentiation (Potthoff and Olson, 2007). In Xenopus, Mef2c mRNA is detected in muscle-associated connective tissue not only in somites but also in the hypaxial and cranial muscles. In addition, Mef2c mRNA colocalizes with scleraxis mRNA at later stages (Della Gaspera et al., 2009). These results suggest that Mef2c could be involved precociously in larval tendon development. When in situ hybridization results are compared between various species, the Mef2c expression profile appears to be conserved in anamniote and amniote species (Figure 5). Mef2c has also been identified by RNA-seq screening in mouse tendon progenitors (Havis et al., 2014). Moreover, it appears in zebrafish that Mef2c expression is visible in intersomitic space at 24 hpf and Scleraxis at 36 hpf suggesting that Mef2c precedes Scleraxis expression in this region (Ganassi et al., 2014; Ma et al., 2018). Interestingly, Mef2c expression in chick and in mice remains closely associated with muscle tissue in a subdomain of Scleraxis-expressing cells suggesting a specific role in this part of connective tissue associated with muscle (Figure 5). Gain-of-function experiments in Xenopus have also shown a synergistic effect between the two transcription factors, Mef2c and Scleraxis, on the expression of two genes known to be expressed in tendon cells, Tgfβi, and Tenascin c (Della Gaspera et al., 2009). Nevertheless, to date, any experiment of loss of function has demonstrated the implication of Mef2c in muscle-associated connective tissue, tendon, or myotendinous junction development. Indeed, Mef2c knockout mice die at E10.5 as a result of a cardiac and vascular malformation making it difficult to study tendon formation. Nevertheless, it has been shown that Mef2c is involved in the formation of the heart and endothelial cells and has a role in cranial neural crest development and chondrocyte hypertrophy (Lin et al., 1997, 1998; De Val et al., 2004; Arnold et al., 2007; Verzi et al., 2007). However, when one of the two Mef2c genes in zebrafish, Mef2ca, is mutated, a minor defect in the formation of some head ligaments is observed (Nichols et al., 2016). Ideally, to study the function of Mef2c in muscle-associated connective tissue, loss-of-function experiments should be performed specifically in these cells. The cre/lox technique could be used to direct the Cre recombinase activity in the sclerotome with Pax1 promoter, but this could also affect chondrocyte development. In order to direct Cre activity in tendon progenitors, Scleraxis promoter could be used but Mef2c could act earlier. An alternative strategy could be to mutate a potential enhancer that could control specifically the expression of Mef2c in these particular cells of connective tissue.

Somites also give rise to populations of endothelial cells (ECs) and vascular smooth muscle cells (vSMCs) that make up blood vessels (Pouget et al., 2006; Wilting and Becker, 2006; Ben-Yair and Kalcheim, 2008; Hirst and Marcelle, 2015). We should also add that somites give rise to another cell type, the adipocyte at least in mice (Sanchez-Gurmaches and Guertin, 2014; Sebo et al., 2018). In chick, somitic ECs give rise to trunk, abdominal wall, limb vessels, and also to lymphatic ones (Pardanaud et al., 1996; Ambler et al., 2001; Wilting and Becker, 2006; Sato, 2013). The contribution of somitic vSMCs is limited to the aorta in the trunk region and to the vessels in the limb and abdominal wall (Pouget et al., 2008; Wang et al., 2015). These results obtained in chick have only been partly extended to mice indicating particularly that some limb ECs emanate from somites (Mayeuf-Louchart et al., 2014). In mice and chick, the trunk aorta is first bilaterally paired before fusing at the midline. It is formed initially from the hemangioblastic splanchnic mesoderm, then two waves of somitic ECs from lateral epithelial somite/hypaxial region of the dermomyotome renew the aorta walls (Pouget et al., 2006; Jaffredo et al., 2013) (Figure 6). Next, somitic vSMCs cover the ECs layer of the aorta but it is not clear whether vSMCs originate from sclerotome or dermomyotome (Esner et al., 2006; Wiegreffe et al., 2007; Pouget et al., 2008; Ben-Yair and Kalcheim, 2008; Mayeuf-Louchart et al., 2014). Interestingly, Ben-Yair and Kalcheim, (2008) suggest that the lateral somite region is the source of multipotent progenitors which give rise to skeletal muscle, smooth muscle and endothelial cells.

In anamniotes, both in zebrafish and Xenopus, a single aorta is formed at the midline below the notochord (Figure 6). The hemangioblasts of the bilateral dorsal lateral plate mesoderm (posterior lateral plate mesoderm (PLM) in zebrafish and dorsal lateral plate mesoderm (DLP) in Xenopus) migrate to the ventral side of somites in medialward direction and coalesce into the aorta (Cleaver and Krieg, 1998; Zhong et al., 2001; Ciau-Uitz et al., 2010; Kohli et al., 2013; Ciau-Uitz and Patient, 2016). In addition, a new somite compartment, marked by cxcl12 and called the endotome, is at the origin of somitic ECs in zebrafish (Nguyen et al., 2014; Keenan and Currie, 2019). (Figures 3B and 6). In zebrafish and in chick, the origin of vSMCs and pericytes of the aortic wall has been identified in the sclerotome (Stratman et al., 2017; Rajan et al., 2020) (Figure 6).

In Xenopus, neither EC nor vSMCs originating from somites have yet been identified. Interestingly, the formation of the posterior cardinal vein (PCV) in Xenopus is closer to the amniotes than to the zebrafish (Figure 6). In zebrafish, the single cardinal vein results from migration and aggregation at the midline of ECs from the PLM, but it has been suggested that somitic ECs could also contribute to this (Isogai et al., 2001; Kohli et al., 2013; Nguyen et al., 2014; Hogan and Schulte-Merker, 2017). In Xenopus and amniotes, a pair of nascent cardinal vein appears bilaterally before fusing medially (Cleaver et al., 1997; Levine et al., 2003). In chicks, they are formed from somitic ECs, and in Xenopus, the question on whether somitic ECs could participate in the formation of the PCV and aorta is raised (Pardanaud et al., 1996; Pouget et al., 2006).

Fgf signaling acts at multiple levels in somite formation in anamniotes. In both zebrafish and Xenopus, Fgf favors the formation of dorsal structure like the paraxial mesoderm, the notochord, and the neural tube (Fürthauer et al., 2004; Fletcher and Harland, 2008). More precisely, in Xenopus, the use of an Fgf receptor antagonist (SU5402) has shown that Fgf signaling is necessary for the specification of the presumptive paraxial mesoderm and the maintenance of gene expression in the Spemann’s organizer, but not for the mesoderm induction (Fletcher and Harland, 2008). In animal cap assays, Fgf is able to induce Tbxt and Myod1 expression, two genes important for paraxial mesoderm specification and primitive myotome construction (Fisher et al., 2002; Fletcher et al., 2006; Fletcher and Harland, 2008) (Figure 7). Inversely, Fgf signaling inhibits ventral mesoderm specification (Kumano and Smith, 2000; Walmsley et al., 2008). Later, Fgf signaling counteracts BMP and favors the paraxial mesoderm fate at the expense of lateral mesoderm as shown in zebrafish (Row et al., 2018). Inside the paraxial mesoderm, Fgf8 also drives the lateral myogenesis and Myod1 expression in both Xenopus and zebrafish (Groves et al., 2005; Della Gaspera et al., 2012b). In addition, this is made at the expense of dermomyotome formation in zebrafish (Groves et al., 2005). The role of Fgf in somite formation can be traced back to the chordate ancestor as Fgf signaling plays a role in anterior somite formation in Amphioxus but seems to exert a more specific role in the vertebrate “clock and wavefront” system (Bertrand et al., 2011 and 2015).

Fgf signaling also acts at multiple levels on somite formation in mice. For instance, it acts early at the primitive streak and is necessary for morphogenetic movement and specification of paraxial mesoderm since without Fgf receptor 1, expression of Tbxt and Tbx6 is decreased (Ciruna and Rossant, 2001). Moreover, Fgf favors paraxial mesoderm fate at the expense of lateral plate mesoderm since SU5402 rescues the expansion of paraxial domain in mutant mice partially deficient in BMP signaling (Miura et al., 2006). The role of Fgf in paraxial mesoderm specification seems to be conserved between amniotes and anamniotes. However, Fgf cannot induce early myogenic program in mouse but plays a role later in pax3 progenitors of the dermomyotome where it favors the triggering of the myogenic program (Lagha et al., 2008b).

In anamniotes, Fgf plays a peculiar role in the early myogenic program and primitive myotome construction (Figure 7). It is necessary early in medial myogenesis and later in lateral myogenesis both in zebrafish and Xenopus (Fürthauer et al., 2004; Groves et al., 2005; Fletcher and Harland, 2008; Della Gaspera et al., 2012b; Osborn et al., 2020). These results suggest that these signaling pathways contribute toward coordinating the myogenic program to other cell behaviors that take place during embryogenesis. In zebrafish, Fgf and Hh signaling cooperate to couple cell movements to muscle differentiation during the apparent somite rotation (Yin et al., 2018). In Xenopus, Fgf4 is involved in a community effect that triggers the myogenic program only if several cells interact with each other (Standley et al., 2001).

The Wnt signaling pathway also plays multiple roles in the dorso-mesoderm and the paraxial mesoderm specification in anamniotes. In Xenopus, stabilization of β-catenin in the dorsal region of the early embryos is a key event involved in the formation of Nieuwkoop center and Spemann organizer (Kimelman, 2006; Hikasa and Sokol, 2013). As such, β-catenin favors the formation of the entire dorsal region including the paraxial mesoderm. Next, after Xenopus midblastula transition, wnt8 has been identified as a ventralizing factor that inhibits the formation of dorsal mesoderm and favors the formation of the presumptive paraxial mesoderm (Christian and Moon, 1993; Hoppler et al., 1996). A similar two-step function has also been identified in the neuro-mesodermal progenitors of the caudal region in zebrafish. First, Wnt promotes the mesoderm fate at the expense of neural fate, and second, Wnt favors the paraxial mesoderm fate at the expense of lateral plate mesoderm (Martin and Kimelman, 2012). Like Fgf, Wnt is essential for the specification of the presumptive paraxial mesoderm in Xenopus, but its action on primitive myotome construction is more complex than Fgf (Hoppler et al., 1996; Leyns et al., 1997; Kazanskaya et al., 2004; Fletcher and Harland, 2008). The animal cap assay shows that Wnt is unable to induce mesoderm and muscle cells in these pluripotent cells; nevertheless, during gastrulation, β-catenin directly promotes early Myf5 expression (Christian et al., 1992; Guger and Gumbiner, 1995; Shi et al., 2002). It has also been shown in Xenopus embryos that the expression of Wnt target genes could be dependent on other signaling pathways (Christian and Moon, 1993; Nakamura et al., 2016; Kjolby et al., 2019). Therefore, these results suggest that Wnt could be a permissive signal for the instructive Fgf signaling during primitive myotome formation, first in the medial and probably later in the lateral paraxial mesoderm (Shi et al., 2002; Della Gaspera et al., 2012b) (Figure 7).

In amniotes, Wnt/β-catenin also acts early as an organizer inducer and next plays a role in the mesoderm and the paraxial mesoderm specification but without inducing precociously the myogenic program (Yoshikawa et al., 1997; Morkel et al., 2003; Engert et al., 2013; Houston, 2017). Next, during somite formation, the dorsal neural tube and the surface ectoderm express different Wnt genes. Thus, Wnt signaling induces the epithelial state of the somites (Capdevila et al., 1998; Schmidt et al., 2004; Geetha-Loganathan et al., 2006). Wnt signaling is also involved in the dorso-ventral patterning of the somites by promoting the dermomyotome at the expense of the sclerotome (Figure 7). It maintains the epithelial organization of the dermomyotome via the transcriptional activation of Tcf15 gene (Wagner et al., 2000; Linker et al., 2005). Moreover, Wnt in cooperation with Hh signaling favors myotome formation in the epaxial somite region (Ikeya and Takada, 1998; Borello et al., 2006). This later function of Wnt could correspond in Xenopus to the initiation of myogenesis from dermomyotome (stages 28–30) where the second myogenic wave is initiated at the epaxial and the hypaxial levels of the somite (Della Gaspera et al., 2012a). Although Wnt signaling has been less studied at these stages in Xenopus, Wnt11 is one of the Wnt ligands that seems to fulfill a function both in Xenopus and amniotes dermomyotome at the same period of development (Garriock and Krieg, 2007; Geetha-Loganathan et al., 2008).

In vertebrates, Bmp4 is expressed in lateral plate mesoderm, in surface ectoderm and in the floor plate. In amniotes, Bmp4 acts on somitic fate at least in two distinct ways: In mice and chick, high Bmp4 concentration specifies the lateral plate mesoderm at the expense of somites (Tonegawa et al., 1997; Miura et al., 2006), at lower concentration, it leads to lateralization of the somites, increasing the expression of Sim1, a lateral marker of somites (Pourquié et al., 1995; Pourquié et al., 1996; Tonegawa et al., 1997; Wijgerde et al., 2005). Bmp4 extends also Pax3 expression of the hypaxial dermomyotome region at the expense of Myod1, keeping cells in an undifferentiated state and inhibiting the differentiation into skeletal striated muscle (Pourquié et al., 1996; Amthor et al., 1999; Kahane et al., 2007) (Figure 7). In the lateral/hypaxial region of somites, Bmp4 also favors the endothelial cell fate at the expense of skeletal striated muscle (Nimmagadda et al., 2005; Ben-Yair and Kalcheim, 2008). In ventro-medial somite region, Bmp4 inhibition by Noggin and Gremlin antagonists is necessary for sclerotome specification (Stafford et al., 2011). In addition, Bmp4 from the dorsal neural tube favors the dorsal sclerotome development or blood vessels cells formation in the medial sclerotome (Monsoro-Burq et al., 1996; Christ et al., 2004; Nimmagadda et al., 2005).

In anamniotes, Bmp from the roof plate and from the hypochord can also act on somites and limits the specification of muscle pioneer cells by Shh in zebrafish (Nguyen-Chi et al., 2012; Keenan and Currie, 2019). Bmp4 has been mainly identified as a lateralizing/ventralizing factor during embryogenesis (Kondo, 2007; Bier and De Robertis, 2015; Zinski et al., 2018). Indeed, in Xenopus, the morphants for Bmp4 antagonists (Chordin, Noggin, and Follistatin) are ventralized and the development of all the dorsal structures, i.e., the neural tube, the notochord, and somites is strongly affected (Khokha et al., 2005). In zebrafish, Bmp4 also favors the formation of lateral plate mesoderm at the expense of somites by inducing the expression of Id-HLH genes, which antagonize somitic bHLH such as Mesogenin or Myod1 (Row et al., 2018). Therefore, the main Bmp function in favor of lateral plate mesoderm appears to be conserved between amniotes and anamniotes. Bmp action on dermomyotome could also be an ancestral function since BMP favors dermomyotome development at the expense of the myotome both in urodela species, axolotl, and in zebrafish. (Epperlein et al., 2007; Patterson et al., 2010) (Figure 7).

In vertebrates, sonic hedgehog (Shh) is expressed by the notochord and the floor plate and will favor the development of the medial part of somites (Kremnyov et al., 2018). Thus, in amniotes, KO mice for Shh have a major defect in the development of vertebrae, involving Shh in the formation and differentiation of the sclerotome. Pax1, a sclerotome marker, is only transiently expressed in KO mice, whereas the expression of Pax3, a dermomyotome marker, is expanded (Chiang et al., 1996). In chicks and mice, Shh is also necessary for the epaxial myogenesis and Myf5 expression (Teillet et al., 1998; Borycki et al., 1999; Applebaum and Kalcheim, 2015) (Figure 7). In chicks, the transition from the primary myotome formation to the growth phase appears to be related to the ability of the dermomyotome muscle progenitors to respond to hh. In the first phase, they are sensitive and differentiate into muscle fibers, and in the second one, they are insensitive and continue to proliferate (Kahane et al., 2013; Applebaum and Kalcheim, 2015).

In anamniotes, the myotome is the first medial compartment to be formed. In zebrafish, Hh allows the specification of at least three medial cell types in the myotome (Keenan and Currie, 2019). The slow fiber program is induced in adaxial cells by Hh, the pioneer cells, one of the two sub-types of slow fibers, are further specified by Hh, and later, the fast medial fiber fate depends also on Hh (Coutelle et al., 2001; Wolff et al., 2003; Hirsinger et al., 2004). Hh also downregulates the expression of Pax3 and Pax7 in the myogenic progenitors of the dermomyotome and induces myogenic differentiation as in chick (Feng et al., 2006). Moreover, Hh is also necessary for sclerotome development, particularly for the migration of sclerotome cells from the ventral part of somites around notochord, (Ma et al., 2018).

In Xenopus, at least two populations of myotome fibers are under the control of Hh. Hh is necessary for the formation of the “adaxial” cells, which give rise to the superficial slow fibers in the caudal region of embryos, and Hh also promotes the formation of the fast muscle fibers in the trunk region (Grimaldi et al., 2004; Martin et al., 2007). Hh also increases the size of the myotome at the expense of Pax3 expression in both epaxial and hypaxial regions of the dermomyotome at stage 28 in Xenopus (Martin et al., 2007). The role of Hh on myotome and sclerotome formation seems conserved between amniotes and anamniotes (Figure 7).

In amniotes, Hh is considered as a medializing factor generally opposed to the lateralizing factor Bmp4. In chick, Bmp4 favors the expression of Pax3 in the lateral region at the expense of myogenic differentiation, while Shh has the opposite effect (Pourquié et al., 1996; Amthor et al., 1999). It has been shown that Wnt and Shh pathways can activate Noggin in somites, a secreted protein that neutralizes Bmp4 (Hirsinger et al., 1997). Interestingly, in Xenopus, Hh effect begins before stage 20, i.e., before the dermomyotome formation when the somite compartmentalization in Xenopus is truly medio-lateral with myotome cells in medial position and MSCs in lateral position (Della Gaspera et al., 2019). Hence, at least two questions could be raised regarding the effect of Shh on medio-lateral patterning in Xenopus: 1) Could Shh inhibit the formation of the lateral MSCs population which gives rise to dermomyotome later? 2) Could Bmp4 counteract Shh activity on this cell population? In Xenopus, early inhibition of Bmp4 signaling by Noggin decreases satellite cell number at larval stage 45, suggesting that Bmp4 acts precociously on the satellite cell lineage. BMP4 could favor laterally the development of MSCs and/or the dermomyotome at the expense of the primitive myotome (Daughters et al., 2011).

Among the other signaling pathways involved in somite compartmentalization and cell fate decisions, the Notch pathway appears as one of the main way to inhibit skeletal myogenesis or maintain myogenic cells in undifferentiated state. Premature somitic myoblasts differentiation is observed in KO mice for Notch ligand Delta1 that causes a deficit in myogenic progenitors and severe muscle hypotrophy (Schuster-Gossler et al., 2007). Furthermore, in chick and mouse somites, Notch pathway plays a role in the acquisition of smooth muscle and endothelial cell fate at the expense of skeletal muscle (Ben-Yair and Kalcheim, 2008; Mayeuf-Louchart et al., 2014). Interestingly, in Xenopus, Myod1 activates the Notch pathway during gastrulation, linking thus myogenesis to somitogenesis and/or somite compartmentalization through a potential feedback inhibitory loop (Wittenberger et al., 1999; Maguire et al., 2012).

Retinoic acid, another key signaling factor of somitogenesis, could also play a role in Xenopus somite compartmentalization. Indeed, one of its receptors, RARγ, promotes the formation of the primitive myotome during gastrulation, whereas another one, RARβ2 is necessary for the formation of the hypaxial region, which is derived from MSC territory (Janesick et al., 2017 and 2018). However, like other signaling pathways involved in primitive myotome formation, to what extent this function is retained in amniotes remains to be determined. Interestingly, another unidentified signal from the neural plate can extend the primitive myotome domain in Xenopus during neurulation, but it is not known if this signal is used to favor the primitive myotome at the expense of MSCs or to favor the paraxial mesoderm at the expense of the lateral plate mesoderm (Mariani et al., 2001).

It appears that the gain and/or redeployment of genetic programs in MSCs and/or MSC-daughter cells seem to be a key mechanism involved in changes of MSCs cell potency and so, in the somite evolution. The essential role of the LSF as the source of multipotent somitic cells giving rise to different somitic lineages, necessary for the formation of the dorsal musculoskeletal system, has been highlighted in Xenopus. The LSF has been defined by Burke and Nowicki (2003) as the changing interface between somites and lateral plate mesoderm which separates the primaxial domain (the musculoskeletal structures comprising somitic cells only) and the abaxial domain (containing muscle or bone of somitic origin associated with connective tissue derived from lateral plate mesoderm). This boundary zone changes during development and has long been recognized as a region where critical signals are exchanged during somite patterning (Nowicki et al., 2003; Durland et al., 2008; Shearman and Burke, 2009). We identified an early developmental and evolutionary event, taking place at the LSF, with the appearance of MSCs. The evolutionary origin of the lateral plate mesoderm can be also traced back to the appearance of chordates as it could be the case for MSCs (Prummel et al., 2019). Moreover, lateral somitic domain in amphioxus also gives rise to the lateral plate mesoderm, raising the question about the causal evolutionary link between the advent of both LSF and MSCs in vertebrates (Mansfield et al., 2015; Yong et al., 2021). In addition, any changes in MSCs cell potency and lineage that occur at the LSF during vertebrate evolution could potentially affect the development of abaxial region. Recently, it has been shown that endothelial cells of the somitic hypaxial region are necessary for the migration of myogenic progenitors into the limb abaxial domain in mice (Yvernogeauet et al, 2012; Mayeuf-Louchart et al., 2016). Colonization by myogenic cells of the vertebrate appendages is observed in gnathostomes but the colonization mode differs between species. In basal gnathostomes, extension of epithelial somites is supposed to be the primitive mode, whereas migration of myogenic progenitors is adopted before the sarcoptegyrian radiation (Neyt et al., 2000; Wotton et al., 2015). So, endothelial cells lineage formation inside somites during vertebrate evolution should be analyzed in parallel to their function in the migration of myogenic progenitors to the abaxial region, in order to explore the potential relationship between MSCs lineage at LSF and the development of the abaxial region.