Volume 236, Issue 9 p. 2397-2409
Special Issue Reviews–A Peer Reviewed Forum
Free Access

Extrinsic versus intrinsic cues in avian paraxial mesoderm patterning and differentiation

Ingo Bothe

Ingo Bothe

Department of Craniofacial Development, King's College London, Guy's Hospital, London, United Kingdom

Search for more papers by this author
Mohi U. Ahmed

Mohi U. Ahmed

Department of Craniofacial Development, King's College London, Guy's Hospital, London, United Kingdom

Search for more papers by this author
Farrah L. Winterbottom

Farrah L. Winterbottom

Department of Craniofacial Development, King's College London, Guy's Hospital, London, United Kingdom

Search for more papers by this author
Gudrun von Scheven

Gudrun von Scheven

Department of Craniofacial Development, King's College London, Guy's Hospital, London, United Kingdom

Search for more papers by this author
Susanne Dietrich

Corresponding Author

Susanne Dietrich

Department of Craniofacial Development, King's College London, Guy's Hospital, London, United Kingdom

Department of Craniofacial Development, King's College London, Floor 27 Guy's Tower, Guy's Hospital, London Bridge, London SE1 9RT, UKSearch for more papers by this author
First published: 24 July 2007
Citations: 45


Somitic and head mesoderm contribute to cartilage and bone and deliver the entire skeletal musculature. Studies on avian somite patterning and cell differentiation led to the view that these processes depend solely on cues from surrounding tissues. However, evidence is accumulating that some developmental decisions depend on information within the somitic tissue itself. Moreover, recent studies established that head and somitic mesoderm, though delivering the same tissue types, are set up to follow their own, distinct developmental programmes. With a particular focus on the chicken embryo, we review the current understanding of how extrinsic signalling, operating in a framework of intrinsically regulated constraints, controls paraxial mesoderm patterning and cell differentiation. Developmental Dynamics 236:2397–2409, 2007. © 2007 Wiley-Liss, Inc.


The vertebral column is the central skeletal structure of modern vertebrates and name-giving to the clade (Goodrich, 1958). In the 19th century, it became clear that the reiterated pattern of vertebrae develops from somites, the segmented postotic paraxial mesoderm that was probably first noted in the chicken embryo by Malpighi in the 17th century (Gilbert, 2000). Ever since, researchers investigated somite patterning and differentiation. The aim was to understand how these processes are being controlled, and how deviations from normal somite pattern may lead to birth defects of the axial skeleton.

Somites are present not only in avian embryos, but in embryos of all agnathan and gnathostome vertebrates, and in embryos of cephalochordates (Goodrich, 1958). Also larvae of urochordates, now thought to be more closely related to vertebrates than cephalochordates (Delsuc et al., 2006), possess reiterated mesodermal cells or cell groups lined up on either side of the notochord (Goodrich, 1958). Thus, segmented paraxial mesoderm is a primitive feature of chordates. In turn, this infers that the generation of vertebrae is not the original function of somites. In fact, the common denominators of the chordate segmented paraxial mesoderm are that (1) it is confined to trunk/tail and (2) it produces skeletal muscle for locomotion (Goodrich, 1958). Therefore, to understand the development and evolution of locomotion in chordates, extensive research was dedicated to the problem of muscle development.

In the 1980s to 1990s, fate-mapping experiments in the chick embryo established that there is a second source of skeletal muscle in the body, namely the head mesoderm, which encompasses the para- and pre-otic paraxial mesoderm and the prechordal (axial) mesoderm (Noden, 1983a; Wachtler et al., 1984; Couly et al., 1992; Evans and Noden, 2006). These tissues deliver the majority of muscles in the head (Noden, 1983a; Wachtler et al., 1984; Couly et al., 1992; Evans and Noden, 2006). Notably, the head mesoderm is not organised into somites (Noden, 1983a; Wachtler et al., 1984; Couly et al., 1992; Evans and Noden, 2006). Moreover, head and body muscles are differentially affected in muscular dystrophies, with Duchenne's and Becker's, Emery-Dreifuss's, limb girdle, and distal muscular dystrophies sparing the head muscles and Facioscapulohumeral and Oculopharyngeal muscular dystrophies predominantly targeting head muscles (Emery, 2002). Thus, the questions arose whether there is more than one program for myogenesis, whether the programs of unaffected muscles could selectively be activated and bring about some protection to muscle affected by the disease, and whether muscle stem cells developed for therapy should be engineered to carry out particular programs of myogenesis.

Ever since Malpighi's discovery of avian somites, and owing to its accessibility in the egg, the chicken embryo has been used as a model to study vertebrate somite development. This work led to a detailed picture of how signals from surrounding tissues may control the formation of somitic derivatives and showed that these developmental processes are largely conserved in amniotes (Brent and Tabin, 2002; Buckingham et al., 2003; Buckingham, 2006). Yet, the focus on the avian model and on the somite as source of muscle has led to the view that vertebrate somite patterning depends exclusively on surrounding tissues. Moreover, although the head mesoderm is not organised into somites, it was assumed that head muscle formation is a variation on the theme of somitic myogenesis. Recent studies suggest that these views have to be revised.

The aim of this review is to summarise the current understanding of avian somite and head mesoderm patterning and differentiation, to discuss this in the context of chordate somite and muscle development, and to identify the challenges that lie ahead.


Avian Somites Are Dorsoventrally Patterned Into Dermomyotome, Myotome and Sclerotome

Avian as well as mammalian somites are epithelial balls of cells with few mesenchymal cells in their lumen, the somitocoel (Schoenwolf, 2001). Somites are generated at regular intervals from the mesenchymal presomitic mesoderm, in the chicken also named segmental plate (Dubrulle and Pourquié, 2004). Soon thereafter, somites become dorsoventrally patterned into the dorsal dermomyotome, which remains epithelially organised for a prolonged period of time and delivers muscle and dorsal dermis. Ventrally, the somite de-epithelialises and forms the sclerotome, which gives rise to cartilage and bone of the axial skeleton and ribs (Schoenwolf, 2001; Brent and Tabin, 2002; Buckingham, 2006), and see the article by Christ et al., this issue, pages 2382–2396). The morphological differentiation of the somite is preceded by the onset of marker gene expression, with Pax3 and Pax7 labelling the prospective dermomyotome, Pax1 labelling the medial part of the sclerotome, and Sim1 labelling the lateral aspect of both dermomyotome and sclerotome (Fig. 1, and data not shown).

Details are in the caption following the image

Marker gene expression anticipating avian somite patterning and schematic representation of the signals controlling this process. AD: Cross sections of stage III epithelial somites in the flank of HH15 chick embryos, prior to the segregation into dermomyotome and sclerotome; dorsal to the top, medial to the left. Note that Pax3 expression (red in A, blue in B) demarcates the prospective dermomyotome. Pax1 (red in B–D) is expressed in the medial aspect of the prospective sclerotome. MyoD signals (blue in A,C) label the first population of differentiating muscle precursors, thought to lay a scaffold for the subsequent stages of myotome development. Sim1 (blue in D) is expressed in lateral/hypaxial cells encompassing the precursors for the lateral sclerotome (Sim1+/Pax3− cells) and the hypaxial musculature (Sim1+/Pax3+ cells). Epaxial muscle develops from Sim1−/Pax3+ cells. E: Key signals from surrounding tissues that pattern the somite. Dermomyotomal Pax3 expression depends on canonical Wnt signals from the dorsal neural tube and surface ectoderm (E, blue arrows). These signals suppress sclerotome formation. Sclerotomal Pax1 expression is induced by Shh (and Noggin) from the notochord and the floor plate of the neural tube (E, yellow arrow). Shh suppresses dermomyotome formation. Bmp4 from the lateral mesoderm (red arrow) controls lateral/hypaxial Sim1 expression. Shh and Wnt cooperate to specify prospective muscle precursors as epaxial (green shading in somite). Wnt and Bmp cooperate to specify prospective muscle precursors as hypaxial (pink shading in somite). ect, surface ectoderm; fp/not, floor plate and notochord; i/lat mes, intermediate and lateral mesoderm; nt, neural tube.

The dermomyotome is a pool of precursor cells and delivers muscle in waves (reviewed in Buckingham, 2006). Myogenesis is initiated when cells begin to express members of the MyoD family of transcription factors (Buckingham, 2006). This happens first in the medial wall of the still epithelial somite, from where Myf5- and MyoD-expressing, postmitotic muscle precursors spread underneath the developing dermomyotome (Kahane et al., 1998b), and Fig. 1A,C). This scaffold of myogenic cells is readily populated by postmitotic cells from the edges or lips of the dermomyotome (Denetclaw et al., 1997; Kahane et al., 1998a; Cinnamon et al., 1999; Denetclaw and Ordahl, 2000). Together, these cells form the primary myotome, the third morphologically defined derivative of the somite. Now, the dermomyotome proper de-epithelialises and releases mitotically active muscle precursors into the myotome underneath, generating the secondary myotome (Ben-Yair and Kalcheim, 2005; Gros et al., 2005; Kassar-Duchossoy et al., 2005; Relaix et al., 2005; Schienda et al., 2006). These cells do not immediately withdraw from the cell cycle and differentiate, and hence are able to drive muscle growth in fetal phases of development (Ben-Yair and Kalcheim, 2005; Gros et al., 2005; Kassar-Duchossoy et al., 2005; Relaix et al., 2005; Schienda et al., 2006). Some of these cells even retain their precursor cell characteristics until adulthood, serving as muscle satellite cells (Ben-Yair and Kalcheim, 2005; Gros et al., 2005; Kassar-Duchossoy et al., 2005; Relaix et al., 2005; Schienda et al., 2006). Upon de-epithelialisation, the dermomyotome also releases cells that accumulate underneath the surface ectoderm (Ben-Yair and Kalcheim, 2005). These cells are often referred to as dermatome as they give rise to the dermis of the back (the remaining dermis stems from lateral mesoderm, and in the head from neural crest cells (Christ et al., 1983; Noden, 1983b; Couly et al., 1992, 1993). Thus, eventually the dermomyotome is turned into segregated muscle and dorsal dermis anlagen; markers indicative of this process are shown in Figure 2.

Details are in the caption following the image

Morphology and marker gene expression in somites with de-epithelialising dermomyotome. Cross-sections of chicken flank somites at HH24, dorsal to the top, medial to the left with (A) haematoxylin-eosin staining and (BG) in situ hybridisation for the markers indicated on top of the panel. Note that, morphologically, the somite is dorsoventrally subdivided into dermomyotome, myotome, and sclerotome. However, due to the de-epithelialisation of the dermomyotomal sheet and the influx of mitotically active muscle precursors from the dermomyotome into the myotome, the border between the two is ill defined. As the mitotically active muscle precursors maintain Pax3 and Pax7 expression, the staining for these markers spreads from the dermomyotome into the myotome, whose differentiating cells express Myf5. De-epithelialising cells collecting underneath the surface ectoderm to form the dermatome maintain Alx4 expression. The abutting expression domains of En1 and Sim1 demarcate the medial/epaxial-lateral/hypaxial boundary of the somite. For details see Ahmed et al. (2006). d, dermatome; dm, dermomyotome; dml, dorsomedial lip of the dermomyotome; m, myotome; scl, sclerotome; vll, ventrolateral lip of the dermomyotome.

The Amniote Somite Is Also Mediolaterally Subdivided

Fate-mapping studies showed that cells in the medial sclerotome contribute to vertebral column and intervertebral discs, while the lateral sclerotome produces the ribs (Huang et al., 2000; Evans, 2003; see the article by Christ et al., this issue, pages 2382–2396). Similarly, also the dermomyotome and myotome give rise to distinct mediolateral derivatives (Ordahl and Le Douarin, 1992; Huang and Christ, 2000). The medial dermomyotome/myotome generates the deep muscles of the back known as epaxial muscles, which eventually will be innervated by the dorsal ramus of the spinal nerve and are kept separate by the thoracolumbar fascia (Fetcho, 1987). In contrast, the lateral aspect of the somite contributes to the ventral, lateral, and superficial muscles known as hypaxials and innervated by the ventral ramus of the spinal nerve (Fetcho, 1987). Conceptually, muscles can also be mediolaterally subdivided into primaxials (they develop within the context of the vertebral column and outside the lateral mesoderm) and abaxials (the precursors enter the lateral mesoderm and develop within this environment) (Burke and Nowicki, 2003; see the article by Winslow et al., this issue, pages 2371–2381). However, the allocation of cells to epaxial-hypaxial and primaxial-abaxial lineages is morphologically concealed since dermomyotome, myotome, and sclerotome are mediolaterally continuous structures (Schoenwolf, 2001). Yet, first the dermomyotome and, when mitotically active muscle precursors arrive in the myotome, also the myotome are molecularly subdivided, with medial cells expressing En1 and lateral cells expressing Sim1 (Cheng et al., 2004; Ahmed et al., 2006) (Fig. 2F,G). Moreover, En1- and Sim1-expressing cells sort in cell aggregation assays, suggesting that they may form a compartment boundary in vivo (Cheng et al., 2004). The significance of the segregated expression of En1 and Sim1 is currently under investigation.

At Occipital, Neck, and Limb Levels, Somites Generate Migratory Hypaxial Muscle Precursors (MMP)

The developmental anatomy of somites described so far applies, strictly speaking, only to flank levels. At occipital and limb levels, cells from the ventrolateral lips of the dermomyotome do not contribute to the myotome, but de-epithelialise and migrate into the periphery to give rise to tongue (occipital somites) and limb muscles (somites at forelimb and hindlimb levels) reviewed by Buckingham et al., 2003; Buckingham, 2006). Migratory muscle precursors (MMP) proliferate and differentiate at the target site and, hence, are representatives of abaxial muscles (Burke and Nowicki, 2003; see the article by Winslow et al., this issue, pages 2371–2381, in this issue). The onset of MMP formation is characterised by the expression of the Lbx1 gene, which also demarcates the ventrolateral dermomyotomal lips of neck-level somites (Dietrich et al., 1998) (Fig. 3A). In mammals, these somites release MMP into the septum transversum to give rise to diaphragm muscles (Dietrich et al., 1999). In birds, possibly due to the elongation of the neck, the septum transversum has moved out of reach for these cells and they remain on site.

Details are in the caption following the image

Markers and regulation of hypaxial myogenesis. A: HH18 chick embryo, Pax3 expression in red, Lbx1 expression in black. Lbx1 is expressed in the ventrolateral lips of the dermomyotome at occipital, neck, and limb levels (arrows). Expression is maintained when cells leave these lips and migrate into the periphery to form tongue and limb muscles. Pax3 is upregulated in the ventrolateral lips of the dermomyotome at all axial levels, including flank levels (open arrowheads). Expression is also found in migratory muscle precursors, but is masked here by the strong Lbx1 staining. B: Schematic representation of the cascades regulating amniote hypaxial muscle formation. Genetic experiments in the mouse established the cooperating Eya and Six transcription factors upstream of Pax3 (Grifone et al., 2005, 2007), which in turn controls its own upregulation (Bober et al., 1994), the expression of the SF/HGF receptor cMet (Daston et al., 1996; Epstein et al., 1996; Yang et al., 1996), and at the occipital, neck, and limb levels, the expression of the transcription factor Lbx1 (Mennerich et al., 1998; Dietrich et al., 1999). Also in the mouse it was shown that Lbx1 acts upstream of CXCR4, the receptor for the chemokine Sdf1, and that the SF/HGF-cMet and the Sdf1-CXCR4 signalling systems interact genetically (Odemis et al., 2005; Vasyutina et al., 2005). Experiments in the chicken and using mouse explants established the role of the extrinsic signals (Pourquié et al., 1996; Fan et al., 1997; Dietrich et al., 1998; Tajbakhsh et al., 1998; Scaal et al., 1999; Schubert et al., 2002; Alvares et al., 2003; Geetha-Loganathan et al., 2005; Linker et al., 2005). The role of the Hox system was demonstrated in the chick (Alvares et al., 2003). Note that surface ectoderm/Wnt and lateral mesoderm/Bmp4 operate at all axial levels, as do Eya/Six and Pax3. At occipital, neck, and limb levels, positional values/Hox genes within the somite ensure that the hypaxial muscle precursors switch on Lbx1 and begin the programme of migratory muscle precursor formation. At flank levels, hypaxial muscle precursors develop as Lbx1-negative, non-migratory cells. This programme can be overridden by Fgf signals from the apical ectodermal ridge of the limb. SF/HGF and Sdf1 signals from the limb mesenchyme control the de-epithelialisation, targeted migration, and survival of the migrating cells.


Avian Somite Patterning Relies Heavily on Extrinsic Cues

Ground-breaking embryological studies in the chick embryo established that the dorsoventral patterning of the somite into dermomyotome and sclerotome, but also the mediolateral subdivision into epaxial and hypaxial territories, is regulated by tissues surrounding the somite. When epithelial somites were inverted in a dorsoventral or mediolateral plane (not in a rostrocaudal direction; these values are established during segmentation and epithelial somite formation), the somite became patterned normally, indicating that dorsoventral and mediolateral values were re-set (Aoyoma and Asamoto, 1988; Ordahl and Le Douarin, 1992). When epithelial somites were cultured in vitro without the surrounding tissues, neither somite patterning nor differentiation into muscle or cartilage occurred (Kenny-Mobbs and Thorogood, 1987; Buffinger and Stockdale, 1994; Münsterberg and Lassar, 1995; Stern and Hauschka, 1995). Ablation experiments indicated that the dorsal tissues, namely dorsal neural tube and surface ectoderm, are required for dermomyotome development, while ventral tissues, namely notochord and floor plate of the neural tube, are needed for sclerotome formation (Brand-Saberi et al., 1993; Dietrich et al., 1997). The medially located neural tube plus notochord are required for the establishment of medial/epaxial fates, while the lateral mesoderm controls lateral/hypaxial fates (Pourquié et al., 1995, 1996; Dietrich et al., 1998). Finally, neural tube and notochord cooperate to induce epaxial myogenesis, yet hypaxial myogenesis depends on signals from surface ectoderm and lateral mesoderm (Münsterberg and Lassar, 1995; Pownall et al., 1996; Dietrich et al., 1997).

Key Regulators of Avian Somite Patterning are Shh, Wnt, and Bmp Signalling Molecules

The identification of vertebrate homologues of signalling molecules that in the fruit fly control the formation of segment boundaries or wings paved the way to unravelling avian somite patterning on a molecular level (Gerhart, 1999; Gilbert, 2000). The hedgehog, wingless, and decapentaplegic homologues, Shh, Wnt, and BMP, respectively, were found in the very tissues that controlled somite patterning and differentiation (Brent and Tabin, 2002). A plethora of gain-of-function studies using gene misexpression, implantation of protein-loaded beads or of cells releasing the factors, and loss-of-function studies using inhibitors of these molecules, accompanied by knock-out experiments in the mouse, established that in amniotes, Shh accounts for much of the function of notochord and floor plate. Also in amniotes, Wnt3a, Wnt1, and the closely related Wnt6 were found to carry out the function of the dorsal neural tube and surface ectoderm, while BMP4 was established as a key signal from the lateral mesoderm (reviewed in Gilbert, 2000; Brent and Tabin, 2002; Buckingham et al., 2003; Buckingham, 2006); a schematic representation of these signals converging on the somite is shown in Figure 1E.

Shh, Wnt, and Bmp Signals Act in Complex Molecular Cascades

Extensive research into the spread, uptake, transduction, and mediation of signals from the tissues surrounding the somites has provided substantial insight into the downstream regulatory cascades. At the same time, the picture became increasingly more complex:
  • 1

    Signals may serve more than one function Shh, for example, is involved in somite patterning, but also acts as a cell survival factor for the sclerotome and as a mitogen (Fan and Tessier-Lavigne, 1994; Fan et al., 1995; Teillet et al., 1998; Borycki et al., 1999; Marcelle et al., 1999). BMP4 patterns the lateral/hypaxial part of the somite (Pourquié et al., 1996; Dietrich et al., 1998), at the same time preventing myogenic differentiation (Reshef et al., 1998).

  • 2

    Signals act as agonists For example, the BMP antagonist Noggin accompanies Shh in notochord-to-sclerotome signalling (McMahon et al., 1998), while BMP4 expressed in the roof plate of the neural tube accompanies the Wnt signals that regulate dermomyotome development (Marcelle et al., 1997; Sela-Donenfeld and Kalcheim, 2002).

  • 3

    Signals act as synergists Shh and Wnt signals cooperate to induce epaxial myogenesis (Münsterberg et al., 1995), while Wnt and Bmp signals cooperate to induce hypaxial myogenesis (Fan et al., 1997; Dietrich et al., 1998; Tajbakhsh et al., 1998). Similarly, Shh and Bmp molecules synergise to trigger sclerotomal chondrogenesis (Murtaugh et al., 1999). On a molecular level, this synergy is achieved, for example, through Shh plus Wnt upregulating β-Catenin (Schmidt et al., 2000), the key mediator of canonical Wnt signalling (Logan and Nusse, 2004), through Shh controlling the expression of the sulfatase QSulf, whose activity may lead to a local boost of Wnt signalling in the epaxial myotome (Dhoot et al., 2001), through Wnt controlling the expression of the regulator of Shh target genes, Gli2 and Gli3 (Borycki et al., 2000), through Gli1 plus Tcf-Lef/β-Catenin regulating the early epaxial enhancer of Myf5 (Borello et al., 2006), and through Shh-induced Nkx3.2/Bapx1 promoting BMP-regulated cartilage formation (Murtaugh et al., 2001).

  • 4

    Signals act as antagonists Shh, while stimulating sclerotome development, suppresses the development of the dermomyotome (Fan and Tessier-Lavigne, 1994; Johnson et al., 1994). Wnt on the other hand stimulates dermomyotome and suppresses sclerotome formation (Fan et al., 1997; Capdevila et al., 1998). Molecularly, this is achieved, for example, through Wnt signals from the dorsal neural tube upregulating expression of Gas1 in the dermomyotome, which is thought to inhibit Shh function (Lee et al., 2001). Similarly, ectodermal/ Wnt signalling stimulates the expression of Gli3, which in the absence of active Shh signalling acts to suppress sclerotomal and myotomal Shh targets within the dermomyotome (Mo et al., 1997; Borycki et al., 2000; Buttitta et al., 2003; McDermott et al., 2005). The sclerotome on the other hand is protected from Wnt signalling, as Shh triggers the expression of the Wnt antagonist Sfrp2 (Lee et al., 2000).

  • 5

    Signals are fine-tuned via negative feedback and by feed-forward mechanisms For the myotome, levels of Wnt and Shh signalling are critical: both signalling cascades must converge, yet the balance must not be tipped towards dermomyotome/dermis or sclerotome formation. The myotome seems to have solved this problem by expressing the negative regulator of Shh signalling, Hip, in response to Shh signalling (Chuang and McMahon, 1999). Similarly, the Tcf/Lef corepressors of the Groucho family are expressed in the myotome (Daniels and Weis, 2005; Van Hateren et al., 2005), possibly upon Wnt signalling (Houghton et al., 2003). Thus, signalling molecules regulate the levels of signalling via negative feedback loops. On the other hand, local reinforcement of signals may be achieved by feed-forward mechanisms, for example by Shh-mediated conversion of Gli repressors into activators and the subsequent upregulation of the solely activating Gli1 gene in the sclerotome (Mo et al., 1997; Buttitta et al., 2003).

Owing to this complexity, the molecular network controlling amniote somite development is still under intensive investigation, and more and more players are being discovered. However, none of those studies undermined the view that the somite is naïve and entirely depends on extrinsic cues for its development. Yet, the comparison of amniote and non-amniote somite patterning and the unravelling of MMP formation evoked doubts.

The Extrinsic Cues Patterning the Avian Somite May Have Been Co-Opted During Amniote Evolution

Somites and their ability to generate skeletal muscle for locomotion are a common feature of chordates. However, while somite patterning is a time-consuming and complex process in amniotes, this is not the case in non-amniotes. In the zebrafish, for example, only a limited set of extrinsic cues seem to influence mesoderm patterning and differentiation, with Shh accounting for the development of a specialised cell population, the adaxial cells, which initially reside next to the notochord. These cells become incorporated into the somite to give rise to slow-twitch muscle and to organise the epaxial-hypaxial boundary (Blagden et al., 1997). A recent study showed that Shh also regulates the formation of a subpopulation of fast-twitch muscle cells lined up along the epaxial hypaxial boundary (Wolff et al., 2003). Interestingly, the role of Hip in fine-tuning Shh signalling is conserved in the fish and in amniotes. (Ochi et al., 2006). The zebrafish somite proper, however, seems to be able to differentiate into (fast-twitch) muscle on its own account (reviewed by Hollway and Currie, 2005; Ochi and Westerfield, 2007; see the article by Stellabotte and Devoto, this issue, pages 2432–2443).

It has to be considered that non-amniotes develop via free-feeding juvenile stages, which have been abandoned in directly developing amniotes (Goodrich, 1958). It is, therefore, possible that the change in the mode of development allowed amniotes to delay mesodermal cell differentiation and to co-opt signals from the environment for elaborate somite patterning. However, the stage was set already in the fish: the zebrafish embryo, even agnathan hagfish, set aside some Pax3-expressing precursor cells in the somite, which, in a dermomyotome-like fashion, may provide cells for later stages of myogenesis (Hollway et al., 2007; Ota et al., 2007; Stellabotte et al., 2007; see also the articles by Stellabotte and Devoto [zebrafish, pages 2432–2443], Elinson [frogs, pages 2444–2453], and Kusakabe and Kuratami [lamprey, pages 2410–2420] in this issue).

The Choice Between the Two Hypaxial Programmes for Myogenesis Is Regulated Intrinsically

The amniote embryo generates two types of hypaxial muscle precursors: at flank levels those that lack Lbx1 and are incorporated into the myotome; and at occipital, neck, and limb levels, those that express Lbx1 and can migrate to target sites in the periphery (Dietrich et al., 1998; Buckingham, 2006). Embryological studies in the chick showed that when ectopic limb buds are generated in the flank, they recruit MMP to form normal limb muscles (Chevallier et al., 1977; Christ et al., 1977; Hayashi and Ozawa, 1995). Likewise, somites grafted to limb levels switch on MMP markers and eventually form normal limb muscles (Chevallier et al., 1977; Christ et al., 1977; Hayashi and Ozawa, 1995). These findings were interpreted as yet another piece of evidence for the naïve nature of somitic cells, which receive instruction from the environment.

A key molecule controlling the de-epithelialisation of MMP is SF/HGF, which is expressed at limb levels, while the receptor cMet is expressed in the dorsomedial and ventrolateral lips of all somites (Bladt et al., 1995; reviewed in Buckingham et al., 2003; Buckingham, 2006). However, the assumption that SF would be the signal for the localised production of MMP proved incorrect: in mice devoid of SF/HGF, MMP are specified correctly but fail to de-epithelialise and emigrate (Dietrich et al., 1999). Thus, localised signals to trigger MMP formation remained elusive. We showed that Fgf molecules from the apical ectodermal ridge of the limb are capable of converting flank somites into MMP-producing ones (Alvares et al., 2003). However, at occipital and neck levels, these signals are not present. Hence, another mechanism must account for MMP formation.

In 1991, Murakami made the observation that somites exchanged between flank and neck levels would not properly adapt to the new environment (Murakami and Nakamura, 1991). Likewise, when we grafted MMP-producing and non-MMP-producing somites to neck or flank levels, somites expressed or lacked Lbx1 expression according to their original location (Alvares et al., 2003). Thus, unless challenged with a developing limb, somites are predisposed towards a particular programme of hypaxial muscle formation, in tune with their axial location. Indeed, when the positional information of flank somites was changed into that of hindlimb level somites by misexpressing posterior Hox genes, then the somites would switch from the non-MMP to the MMP programme of muscle formation (Alvares et al., 2003). Thus, the signals generally stimulating hypaxial muscle development are permissive, and the decision between MMP and non-MMP programmes is intrinsic to the somite (Fig. 3B). Recent studies showed that hypaxial muscles precursors, when they enter the lateral mesoderm, adapt their Hox gene expression to that of their new environment (Nowicki and Burke, 2000; see the article by Winslow et al., this issue, pages 2371–2381). However, when chick-quail somites were exchanged, muscle precursors gave rise to fast or slow twitch limb muscles in a species-specific manner, lending further support to the idea that intrinsic cues determine how extrinsic signals are being interpreted (Nikovits et al., 2001).

Tracing the muscles for the paired fins of the zebrafish, it became apparent that they develop from MMP (Neyt et al., 2000; Haines and Currie, 2001). In cartilaginous fishes such as sharks, however, the paired fins receive muscle in the form of lateral extensions of the myotomes (Goodrich, 1958; Neyt et al., 2000; Haines and Currie, 2001). Therefore, it was thought that MMP are an innovation of bony fishes and tetrapods, and may have evolved in conjunction with more mobile paired fins/limbs (Clack, 2002). Yet, molecular studies have shown that the pectoral fins of both bony and cartilaginous fish already harbour many of the molecules required for tetrapod limb development (Tanaka et al., 2002; Mercader et al., 2006; Tickle, 2006). Thus, it is conceivable that early in vertebrate evolution, the paired appendages were able to recruit muscle and innervation from the axial midline. Specifically, during the evolution of bony fish and tetrapods, the neural tube must have gained the ability to generate specialised motor neurons that are organised in a novel, namely the lateral, motor column (Landmesser, 1978; Hollyday, 1980a, b; Tosney and Landmesser, 1985a, b). Likewise, the mesoderm must have recruited Lbx1, previously only expressed in dorsal interneurons of the neural tube, into the somite to participate in a molecular network that delays myogenic cell proliferation and differentiation for the sake of de-epithelialisation and migration (Schubert et al., 2001; Gross et al., 2002; Muller et al., 2002). We can speculate that this initially took place at occipital levels, the site of pectoral fins in the fish (Goodrich, 1958; Clack, 2002). During tetrapod evolution, the paired fins/limbs lost connection to the skull and translocated to a more posterior position, where they better supported the centre of gravity (Clack, 2002; Lours and Dietrich, 2005). Somites along the way kept the newly acquired MMP programme, such that today, traces of the MMP programme (Lbx1 expression) or, in fact, active MMP production (cell emigration) are still found at occipital and cervical levels (Dietrich et al., 1998, 1999). Interestingly, in Xenopus laevis, MMP are also generated at flank levels (Martin and Harland, 2006). Whether this is associated with a change in somite axial identities, an uncoupling of the MMP system from the Hox system, or indicates a more widespread use of MMP in vertebrates than previously anticipated, is not known.


Head and Somitic Mesoderm Are Set Up Intrinsically to Follow the Head or the Trunk Programmes for Myogenesis

Besides the musculature used for locomotion, vertebrates harbour skeletal muscles in the head, used for the movement of the eyes (extraocular muscles) and the movement of the gills and their derivatives (hypobranchial and branchiomeric muscles; (Goodrich, 1958). In amniotes, the branchiomeric muscles close the jaw (mandibular or 1st branchial arch muscles), control the cranial openings, the opening of the jaw and facial expression (hyoid or 2nd arch branchial arch muscles), and facilitate the movement of the pharynx and larynx (posterior branchial arches). They are crucial for food uptake and, hence, survival.

Head skeletal muscle derives from two sources, the occipital somites and the head mesoderm, the mesenchymal strip of tissue encompassing the para- and preotic paraxial and the prechordal mesoderm (Noden, 1983a; Wachtler et al., 1984; Wachtler and Jacob, 1986; Couly et al., 1992, 1993). However, the occipital somites are a trunk derivative; they were secondarily incorporated into the head during vertebrate evolution, their vertebrae fusing into the basioccipital bone supporting the enlarged brain (Gans and Northcutt, 1983). Moreover, occipital somites are organised as somites in the trunk (Hamilton and Hinsch, 1956; Huang et al., 1997) and express markers typical for trunk somite patterning and cell differentiation (Bothe and Dietrich, 2006; S. Dietrich, unpublished observations). Yet, the properties of the head mesoderm have been the subject of controversy for 200 years.

Ever since Goethe (1749–1832) and Oken (1779–1851) proposed the vertebral theory of the skull, the holy grail of mesoderm research was to find proof for head somites (Oken, 1807; Goethe, 1820; reviewed in Kuratani, 2005). When Balfour 1878 discovered head cavities, vesicular structures buried in the craniofacial mesenchyme of the shark embryo, they were readily celebrated as head somites (Balfour, 1878; reviewed in Goodrich, 1958; Kuratani, 2005). However, different numbers of head cavities form in different species, they arise late in development, they do not share the organisation of somites, and, importantly, they are not found in agnathans, suggesting they are a shared-derived character of gnathostomes (Kuratani, 2005). In the meantime, studies in the chick embryo showed that the head mesoderm, when transplanted into the trunk, is unable to participate in trunk programmes of myogenesis (Mootoosamy and Dietrich, 2002). Thus, the idea that the head mesoderm is intrinsically different from the somitic mesoderm had to be considered. Indeed, evidence is accumulating that the head mesoderm has its unique molecular setup.

The head mesoderm lacks expression of genes required for mesoderm segmentation and epithelial somite formation, indicating that it is not controlled by the segmentation machinery in the trunk (Bothe and Dietrich, 2006). In contrast to somites, the head mesoderm also lacks enzymes involved in Retinoic Acid (RA) synthesis and, instead, expresses factors involved in RA degradation (Bothe and Dietrich, 2006). This suggests that the head mesoderm is set up as RA-free territory while the somite participates in the RA- and Hox-dependent system that determines positional values in the trunk (Burke, 2000; Kmita and Duboule, 2003). Significantly, the head mesoderm never harbours expression of Pax3 (Hacker and Guthrie, 1998; Mootoosamy and Dietrich, 2002; Bothe and Dietrich, 2006), the gene that is transiently expressed in all somitic muscle precursors and that serves as a key upstream regulator of somitic myogenesis (Maroto et al., 1997; Tajbakhsh et al., 1997; Relaix et al., 2004, 2005; reviewed in Buckingham, 2006). Instead, the early head mesoderm expresses its own, distinct set of transcription factors, including Pitx2, Alx4, MyoR, Tbx1, and Twist (Bothe and Dietrich, 2006). Knock-out studies in the mouse have implied some of these in craniofacial, but not somitic myogenesis (Gage et al., 1999; Kitamura et al., 1999; Lu et al., 2002; Kelly et al., 2004; Diehl et al., 2006; Dong et al., 2006; Shih et al., 2007). Significantly, the head mesoderm markers label this tissue prior to the onset of myogenic differentiation, indicated by the start of Myf5 and subsequently MyoD expression (Noden et al., 1999; Bothe and Dietrich, 2006) (Fig. 4). Moreover, the head mesoderm genes have the ability to promote proliferation and suppress differentiation, and, hence, may serve a similar purpose as Pax genes in the dermomyotome (Hebrok et al., 1994, 1997; Lu et al., 1999; Kioussi et al., 2002; Xu et al., 2005; Martinez-Fernandez et al., 2006). Yet, the molecular cascades that ultimately lead to myogenic differentiation are distinct. Studies in the mouse and chick have lent further support to this idea: signalling molecules that are required for somitic myogenesis, namely Shh and Wnt, repress myogenesis from the head mesoderm (Tzahor and Lassar, 2001; Tzahor et al., 2003). Moreover, in the mouse, the MyoD family member Myf5 employs distinct enhancers during somitic and head myogenesis (Hadchouel et al., 2000; Summerbell et al., 2000). Within the MyoD family, Mrf4 can compensate for the absence of Myf5 or MyoD in somitic but not head muscle formation (Kassar-Duchossoy et al., 2004). The challenge is now to unravel the molecular network regulating the differentiation of the head mesoderm.

Details are in the caption following the image

Markers for the undifferentiated head mesoderm. AG: Heads of HH10 chicken embryos; dorsal views, anterior to the top. AiGi: Cross-sections through the anterior hindbrain of the embryos shown in A–G, dorsal to the top. The marker genes are indicated on the top of the panel. Note that Pitx2, Alx4, and MyoR label the anterior head mesoderm, with overlapping but distinct mediolateral expression domains. Similarly, overlapping expression, predominantly in the prechordal plate and posterior head mesoderm, is found for Tbx1 and Twist. Pax3 is expressed in emigrating neural crest cells, not in the head mesoderm. Myf5 is also not expressed in the early head mesoderm, but expression will commence at later stages. In contrast to the head mesoderm, the somites are characterised by strong Pax3 expression in the dermomyotome, and by Myf5 expression in the developing myotome. Twist is also expressed, labelling the sclerotome. For details, see Bothe and Dietrich (2006). as, prospective aortic sac; end, pharyngeal endoderm; fg, foregut; hm, head mesoderm; ncc, neural crest cells; not, notochord; pc, pericardial cavity; pp, prechordal plate; r1/2, rhombomere 1 and 2; som1, 1st somite; splpl, splanchnopleure.

Extrinsic Cues Are Required to Control the Patterning and Differentiation of the Head Mesoderm

While recent studies suggested that the head mesoderm is a distinct type of mesoderm that follows its own differentiation programmes, this is not to say that head mesoderm development is regulated by intrinsic cues only. Indeed, when avian head mesoderm with surrounding tissues is cultured in an explant culture system, it differentiates into muscle (Tzahor and Lassar, 2001; Tzahor et al., 2003). If the head mesoderm is cultured alone, neither the early head mesoderm markers nor muscle differentiation markers are expressed (I.B. and S.D., unpublished observations). Thus, also head mesoderm development relies on extrinsic cues. Co-culture experiments suggested that pharyngeal endoderm supports the expression of some of the head mesoderm markers, for example Pitx2 (I.B. and S.D., unpublished observations). However, the signalling molecules that account for this effect have proved difficult to identify. Expression analyses suggest that the same signalling molecule may be expressed at various sites (Chapman et al., 2002; Karabagli et al., 2002; Chapman et al., 2004; I.B. and S.D., unpublished observations). Moreover, the topology of these sites changes over time. Nevertheless, gain-of-function experiments suggested that both Fgf and Bmp molecules serve as negative regulators of myogenic differentiation in the head (Tzahor et al., 2003; Tirosh-Finkel et al., 2006; von Scheven et al., 2006). However, the significance of these findings is not fully understood: Fgf upregulates expression of the early head mesodermal markers at the expense of MyoD family members (von Scheven et al., 2006). At the same time, however, Fgf upregulates expression of genes typical for the jaw muscle precursors in the first pharyngeal arch and suppresses markers for the lateral rectus eye muscle, which develops outside the 1st arch (von Scheven et al., 2006). Whether and how muscle differentiation and extraocular versus branchiomeric muscle patterning are linked needs to be clarified. Similarly, BMP molecules are strongly expressed in the primary heart field (Schultheiss et al., 1997; reviewed in Brand, 2003), and can convert head mesoderm into cardiac mesoderm (Tzahor et al., 2003; Tirosh-Finkel et al., 2006). Thus, the abolition of myogenic differentiation under the influence of Bmp may well result from head mesoderm re-specification (Tzahor et al., 2003; Tirosh-Finkel et al., 2006).


Although far from being complete, a picture is emerging by which at least in amniotes, various signals from surrounding tissues converge onto the somitic and head mesoderm to control their patterning and cell differentiation. Amazingly, the signalling molecules provided by these tissues are largely the same both in the head and in the trunk. However, how these signals are being interpreted depends on the intrinsic cues, the molecular setup, of the receiving tissue; hence the outcome of, for example, Wnt and Bmp signalling to somites at different axial levels, or of Wnt signalling to the head versus somitic mesoderm, is different. Thus, the challenge ahead of us will be to gain insight into the molecular setup of mesodermal tissues and its significance as this will lead to the understanding of what is commonly called “context-dependent signalling.”

Inevitably, the molecular setup of the somitic and head mesoderm reflects the evolutionary history of these tissues. Thus, we have to reflect whether the perception of the head muscles as just another skeletal muscle has been deceiving. Maybe it is relevant that the muscles in the head were never meant to serve locomotion? We will have to clarify the evolutionary origin of the head mesoderm. For this, comparative analyses of the established vertebrate model organisms, but also the non-model organisms discussed in this issue of Developmental Dynamics, will be required. Yet, the rapid progress in molecular techniques, combined with the opportunity to carry out embryological manipulations, renders the chicken an invaluable model to unravel regulatory cascades in mesoderm patterning and differentiation.

Table 1. Development of Somitic Versus Head Mesoderm
somitic mesoderm = post-otic paraxial mesoderm head mesoderm (hm) = para-and pre-otic paraxial plus prechordal mesoderm
Fate of cells
Dorsal dermis, skeletal muscle, cartilage/ bone, endothelial cells Skeletal muscle, cartilage/ bone, endothelial cells; no contribution to dermis
Organisation of the tissue
Segmented Unsegmented
Epithelial balls of cells, mesenchymal core (somitocoel), later morphological differentiation into sclerotome, dermomyotome, myotome Mesenchymal tissue, Temporary emergence of epithelially organised head cavities in gnathostomes, but morphological different from somites; no differentiation into sclerotome, dermomyotome, myotome
Acquisition of axial identity
Under the control of retinoic acid (RA) and Hox/HOM genes Possibly set up as RA-free territory
Marker gene expression
Onset of expression in whole somite at the time of clock arrest, mesoderm condensation and epithelial somite formation, later confined to dermomyotome: Pax3 In differentiating muscle precursors: Myf5, MyoD In the sclerotome: Pax1, Pax9 No Pax3
Sequential activation in the early head mesoderm of Pitx2, Alx4, MyoR, Tbx1, Twist; these genes positively regulate cell proliferation/ negatively regulate myogenic differentiation Expression precedes that of Myf5 and MyoD
Initiation of myogenic differentiation
Somites transplanted into the head carry on with somitic myogenesis, hence signals for somitic myogenesis are present in the head Hm transplanted into the trunk does not participate in somitic myogenesis, hence hm requires its own, head specific signals to differentiate
Positive regulators of myogenic differentiation: canonical Wnt signalling, Shh, Bmp inhibitors Positive regulators of myogenic differentiation: antagonists of canonical Wnt signalling and of Bmp signalling
Negative regulators of myogenic differentiation Bmp, Fgf Negative regulators of differentiation canonical Wnt, Shh, Bmp, Fgf
Further differences:
members of the MyoD family of bHLH transcription factors are required for skeletal myogenesis, but in the mouse:
  Distinct enhancer elements regulate the expression of Myf5 in the head and in somites
  In the absence of Myf5 and MyoD, Mrf4 rescues embryonic muscle formation from somites, but not from hm
Control of muscle patterning
Shh + Wnt1/3a − epaxial Wnt6/7a + Bmp4 − hypaxial (intrinsic control via Hox genes over the formation of migratory versus non-migratory hypaxial muscle precursors) Neural tube plus further, unidentified signals— extraocular muscles
Branchial arches/ Fgf8—branchial arch muscles


We thank A. Basson, A.-G. Borycki, A. Burke, P. Currie, S. Devoto, A. Graham, S. Hughes, R. Kelly, R. Knight, P. Maire, A. Münsterberg, M. Meredith-Smith, F. Schubert, A. Schwienhorst, and S. Vainio for inspiring discussions. We are indebted to F. Schubert for critically reading the manuscript. Gudrun von Scheven is an EU Marie Curie Early Stage Fellow.