Vertebrate craniofacial muscles are marvelously paradoxical. Their basic organization has been highly conserved during evolution, yet among their members are some of the most diverse musculoskeletal structures present in vertebrates. While the developmental bauplan established during emergence of gnathostomes (vertebrates with jaws) has in large part been retained, an enormous number of specializations have arisen to accommodate or permit a wide range of craniofacial adaptations.
Skeletal muscles throughout the body originate within paraxial mesoderm. In the head of amniote embryos, this tissue forms as and remains a mesenchymal population, devoid of either epithelial compartments (somites) or overt segmentation. Following gastrulation, cephalic paraxial mesoderm is overlaid by neural and surface ectodermal epithelia and underlain by pharyngeal endoderm. Neural crest cells subsequently spread over its superficial surface (Fig. 1). Identifiable compartments such as myotomes and sclerotomes are not evident histologically nor by most molecular criteria, yet the cell types generated within head paraxial mesoderm are identical to those present in trunk somites.
At all axial levels, the initiation and continued differentiation of skeletal muscle require a consortium of signals, whose positive and negative regulatory influences act sequentially to establish spatially and temporally diverse myogenic populations (Emerson and Hauschka,2004). In the trunk, these signals are produced by surrounding tissues, e.g., surface ectoderm, neural tube, and notochord, as well as by neighboring paraxial mesodermal cells (Brand-Saberi,2005). Many tissues adjacent to craniofacial myogenic mesoderm are either unique to the head, e.g., pharyngeal endoderm and neural crest–derived mesenchymal cells, or express an extensive set of region- and segment-specific features, e.g., midbrain-hindbrain boundary, rhombomeres (Rhinn and Brand,2001). Each of these neighbors needs to be vetted for its role(s) in head myoblast differentiation and muscle morphogenesis. In addition, having neighboring mesodermal cells being mesenchymal rather than epithelial creates opportunities for different signaling and signal transduction strategies.
Most craniofacial muscle primordia move from their sites of initial differentiation within paraxial mesoderm into peripheral locations, whose ontogenetic histories are different. A few, e.g., muscles that elevate or rotate the skull, arise within somites and move in conjunction with their connective tissue precursors, but most venture unadorned into novel surroundings. During these excursions, which may precede, parallel, or follow overt differentiation of myocytes (post-mitotic mononucleated cells expressing skeletal muscle–specific markers), most head muscles become surrounded by and then infused with neural crest cells (Fig. 2), which form endo-, peri-, and epimysial compartments in addition to fascia and tendons. The integration of crest and myogenic mesoderm populations coincides with the elongation and alignment of myocytes and immediately precedes both the segregation of individual muscles from common progenitor populations and the formation of intramuscular vascular channels.
This review will first expand upon each of the preceding features, using specific examples to document key cellular and molecular similarities and differences both among different craniofacial myogenic populations and between head and trunk systems. Emphasis is on avian and murine systems, in which a greater number of descriptive and more extensive experimental analyses have been conducted. Most analyses focus on primary myogenesis, during which the scaffolding necessary for secondary myofiber differentiation is established.
This largely descriptive tapestry provides contexts to examine known and suggest likely additional tissue interactions and signal/response networks that operate during head muscle development. By emphasizing the challenges associated with this heterogeneous system and highlighting the many areas in which our understanding of cellular processes and molecular mechanisms is rudimentary, we hope to provoke new explorations of craniofacial muscle induction, differentiation, and morphogenesis.
WHAT ARE HEAD MUSCLES?
Traditionally, craniofacial skeletal muscles are catalogued as four distinct populations: extra-ocular, branchial1, laryngoglossal, and axial, which includes epaxial and hypaxial muscle groups that elevate, depress, and rotate the head (Fig. 3). Many terrestrial vertebrates have additional superficial muscles. Some are grouped as part of a constrictor colli system; these circumscribe the caudal part of the head and may extend beneath the jaw, and generally function to provide ventral muscle tone and aid in swallowing. Mammals are additionally endowed with an extensive set of superficial facial muscles that allow fine movements of lips, eyelids, and cheeks, and with specialized pharyngeal constrictors.
Extra-Ocular Muscles (EOMs)
Extra-ocular muscles move and maintain the rotational stability of the eye, with additional accessory ocular muscles involved in protecting the cornea, e.g., the retractor bulbi in mammals and pyramidalis/quadratus in birds. The basic pattern of six EOMs (Fig. 3) is shared among all vertebrate classes, but there have been many evolutionary modifications and adaptations. In some early gnathosomes, for example, the lateral rectus has a vertical orientation instead of the horizontal alignment present in most vertebrates. In caecilian amphibians, the eyes are reduced in size and several EOMs have become modified as protractors and retractors of an external appendage called the tentacle (Billo and Wake,1987). A striking functional EOM adaptation is present in marlins, sailfish, and swordfish. Here the dorsal rectus is greatly enlarged and modified as a local heat-generating tissue capable of warming the brain and eye 14°C above ambient temperatures (Block,1994).
Amniote EOMs are compartmentalized into deep (global) and superficial (orbital) domains, which are established during primary myogenesis (Noden et al.,1999). The distal attachments of each domain are different, which allows the global domain to initiate eye movements while the orbital acts via connective tissue pulleys as an elastic loading system. This maintains rotational stability as the eyes move along horizontal and vertical planes (Demer,2002).
EOMs, especially in mammals, have metabolic and fiber type compositions distinct from most branchial and trunk muscles (Porter et al.,2001; Hoh,2002; Wigmore and Evans,2002; Cheng et al.,2004). Each of the two EOM domains also has distinctive fiber types (Khanna et al.,2004; Rubenstein et al.,2004) and their innervation varies along their lengths (Buttner-Ennever and Horn,2002; Briggs and Schachat,2002), which enables fine control of slight movements. In some fibers, the unique superfast myosin isoforms that facilitate rapid contraction are synthesized only at the level of the neuromuscular junctions (Schachat and Briggs,2002).
In addition to the EOMs, most vertebrates have accessory ocular muscles involved in protecting the cornea. Typically, these assist in retracting the globe deeper into the orbit, but in some species skeletal muscles participate in moving primary and auxillary eyelids. Most of these muscles are innervated by neurons located in a distinct accessory abducens subnucleus, whose peripheral axons travel with the abducens (VIth) cranial nerve.
The ancestry of accessory ocular muscles is uncertain. In many fishes, the lateral rectus shares a common innervation with a longitudinally elongated basicranial muscle that elevates the upper jaw, and Northcutt and Bemis (1993) have proposed that this muscle may have been the evolutionary precursor of accessory ocular muscles. As with EOMs, these accessory ocular muscles have undergone many adaptations. Some ranid amphibians, for example, have co-opted the retractor bulbi complex to aid in swallowing (Levine et al.,2004).
Branchial arch muscles are those associated with jaw, hyoid, and caudal branchial skeletal structures and their homologues. These muscles historically were thought of as having evolved from an iterative set of serially homologous gill muscles, which together with gill skeletal elements constituted the branchiomeric apparatus (reviewed by Neal,1918; Kuratani et al.,1999). Indeed, the branchial musculoskeletal system has often been touted as a novel and defining feature of vertebrates (reviewed in Gans and Northcutt,1983; Hall,2005; Northcutt,2005). However, current evidence suggests that the first (mandibular) branchial arch may have arisen with unique structural and molecular attributes and not evolved from a gill-supporting ancestor (Kuratani,2004; Cerny et al.,2004).
Axons innervating branchial muscles project from motor neurons that become segregated in lateral regions of the brainstem, distinct from EOM and somatic motor nuclei (reviewed by Szekely and Matesz,1993; Northcutt,1990). These neurons also have molecular signatures not shared with somatic motor neurons (Jacob et al.,2001).
With the emergence of terrestrial vertebrates and loss of gill-supporting skeletal elements, the muscles in each arch underwent substantial modifications. Many muscles associated with the more caudal arches have disappeared, and the few that remain would not at first examination reflect their branchial heritage (discussed later). Muscles associated with the branchial arches 1 and 2 are highly diverse, reflecting the wide range of evolutionary changes associated with sites of jaw articulation and modes of masticatory movement.
The muscles that elevate (move rostrally and dorsally) the larynx and root of the tongue in mammals are functionally grouped as the suprahyoid musculature. However, their embryonic origins are diverse, including branchial arches 1 (mylohyoid) and 2 (digastricus, stylohyoid) plus occipital somites (geniohyoid). Unique to mammals are muscles associated with external ear movement and facial expression. In most situations, these are derived from late-differentiating second branchial arch myoblasts that disperse to periauricular, periorbital, and perioral locations (Fig. 4; Gasser,1967; Carvajal et al.,2001). An exception is the retractor for the cheek pouch in cricetids (hamsters, gerbils), which has shifted rostrally from a 3rd arch location (Ryan,1986). An analogous muscle in squirrels is of 2nd arch origin, which illustrates the dynamic, and in this case convergent, range of variation among these branchial muscles.
Tongue and Intrinsic Laryngeal Muscles
Tongue (glossal) and laryngeal structures are relatively recent evolutionary adaptations of the craniofacial musculoskeleton, appearing coincident with terrestrial amphibian species (Edgeworth,1935; Iwasaki,2002). However, the muscles associated with these structures have a more extended ancestry. Analyses of pax3/7 gene expression in lamprey embryos identify a population of hypobranchial myoblasts that originate from the lateral borders of rostral somites and migrate beneath pharyngeal structures (Kusakabe and Kuratani,2005). Comparable myoblasts in fishes attach ventral gill elements to the coracoid bone of the pectoral girdle (reviewed by Jarvik,1980; Hildebrand and Goslow,2001). In amphibians and amniotes, migratory myoblasts arising from occipital somites form a condensed mesenchymal band, the hypoglossal cord, that elongates and similarly brings myoblasts ventral to pharynx (Hammond,1965; Hazelton,1972), where they form both intrinsic and most extrinsic tongue muscles and intrinsic laryngeal muscles.
Because they arise at the boundary of trunk and head, there has been much conjecture about which category best suits these wandering myoblasts. Early hypotheses that tongue muscles are modified gill (branchial arch) muscles are mitigated by the finding that their innervation (cranial nerve XII) is from neurons located in somatic rather than branchial motor nuclei (reviewed by Northcutt,1990; Cordes,2001). Also, some tongue muscles are spared in transgenic mice null for Tbx1, a transcription factor that is required for differentiation of all branchial muscles (Kelly et al.,2004). They are probably best described as hybrids, originating from somites but completing their development within craniofacial microenvironments.
Head-Moving, Postural, and Infrahyoid Muscles
Muscles associated with postural stabilization and movements of the head are multisegmental and all arise from medio-dorsal and latero-ventral domains of occipital and cervical somites (Noden,1983b; Huang et al., 2001; Couly et al.,1992,1993; Matsuoka et al.,2005). These muscles become organized as epaxial (dorsal to the transverse process of each vertebra) and hypaxial (ventral to the transverse process) similar to other trunk muscles. Among the latter are the infrahyoid muscles, which in mammals include the omo-, sterno-, and thryo-hyoid muscles; in aves, these are grouped as caudal external laryngoglossal muscles. These function to lower (move caudally) the larynx and root of the tongue, and are innervated by both hypoglossal (XII) and cervical nerves.
EMBRYONIC ORIGINS AND ORGANIZATION OF HEAD MESODERM
Head Cavities and Pre-Otic “Somites”
Both the phylogenetic and ontogenetic histories of head muscles were contentious and controversial topics for much of the 20th century. Beginning with Goethe's hypothesis that the head (skull) arises from a segmentally organized set of progenitors, as does the vertebral musculoskeletal apparatus, arguments about the existence and number of head segments and their structural basis have abounded (reviewed by Kingsbury,1926; Goodrich,1930; Neal and Rand,1946; Jarvik,1980; Northcutt,1990; Kuratani,2005). A segmental origin of the neurocranium (the part of the skull that surrounds the brain) has, except for the occipital region, been largely discredited. However, the possibility that cranial nerves, muscles, and the viscerocranium (the part of the skull associated with the jaw and branchial skeleton) arise from metameric primordia remains actively debated.
Balfour (1877) beautifully described in shark embryos a series of three segmentally arranged, epithelial-lined condensations rostral to the otic vesicle (reviewed by Jarvik,1980; Kuratani,2003, 2005). At early stages, these retain connections with the coelomic cavity, prompting Balfour to name them head cavities, although their histological appearance more closely resembles somites. These subsequently become isolated during expansion of pharyngeal pouches and visceral clefts (indentations of the surface ectoderm between each branchial arch). The dorsomedial layer surrounding each head cavity forms a myotome-like epithelium that becomes associated with an EOM-innervating cranial nerve (III, oculomotor; IV, trochlear; and VI, abducens). Comparable head cavities/somites have since been described in several fish species (reviewed by Neal and Rand, 1940; Jarvik,1980; Gilland and Baker,1993), in Xenopus (Chung et al.,1989) but not other amphibians (Kuratani et al.,1999), and in Chelydra (snapping turtle, Johnson,1913).
Gilbert (1947,1957) identified mesenchymal condensations in early (25–30 somites) cat and human embryos that included primordia of extraocular muscles and within which small epithelial clusters formed (Gilbert,1952). Jacob and colleagues (Jacob et al.,1984; Wachtler and Jacob,1986) described small, paired premandibular cavities in chick embryos. In some other avian species these cavities are substantially more robust in their appearance (e.g., heron, Wedin,1953), and are bounded by cells that resemble squamous epithelium but lack a basement membrane. However, the location of these cavities does not correspond to sites where molecular mapping data indicate that EOM primordial are present (Fig. 3; Noden et al.,1999), and their structural resemblance to early endothelial aggregates raises the likelihood that head cavities in amniotes are involved in vasculogenesis.
Although not present in all vertebrates, head cavities and pre-otic somites were incorporated during the first decades of the 20th century as support for the hypothesis that the developing brain, pharynx, and the full length of paraxial mesoderm are integrated, segmentally organized tissues, with each segment ancestrally including a sensory ganglion, motor nerve, and musculoskeletal assembly. Advocates were not deterred by the fact that no extant vertebrate retained the full complement of head segments (discussed by Romer,1972). Clarification of the organizing principles of amniote head development required mapping of all mesenchymal populations and analyses of the interactions among them (see below).
Head mesoderm is generated beginning at the onset of gastrulation and is in place by mid-gastrulation stages. In the median plane, it includes prechordal mesoderm, a sparse population of mesenchymal cells located beneath the rostral neural plate, followed caudally by the notochord. Prechordal mesoderm is contiguous laterally with paraxial mesoderm (Meier,1979; Seifert et al.,1993), which extends caudally without interruption to the first somite.
Prechordal mesoderm is generated from progenitors located in the rostral margin of the elongating primitive streak (avian) or embryonic organizer center (mouse), and is often indistinguishable from the overlying neuroepithelial and underlying endodermal layers (Lawson et al.,1991; Barteczko and Jacob,1999; Vesque et al.,2000). Hedgehog and nodal signals from these mesodermal cells, acting alone or in conjunction with endodermal signals, are necessary for the regional specification of the rostral neural plate and also for the bilateralization of prosencephalic structures (Patten et al.,2003; Zhang and Yang,2001).
Mapping studies of avian prechordal mesoderm reveal that while some of these cells contribute to the rostral elongation of the notochord, most shift laterally concomitant with the lateral expansion of the rostral neural plate. In some mammals, a prominent transmedian bridge of condensed prechordal mesenchyme persists for several stages (Gilbert,1957). As the neural plate folds dorsally to form a closed neural tube, laterally located prechordal cells together with adjacent rostral paraxial mesoderm cells remain closely associated with the optic primordium. Evagination of the optic vesicle displaces both mesenchymal populations caudally (Fig. 5) and eliminates any observable spatial distinction between them. Transplantation (Wachtler et al.,1984) and retroviral mapping (Evans and Noden,2006) studies show that these cells contribute to the formation of the dorsal, ventral, and medial rectus and ventral oblique muscles, all of which are innervated by the oculomotor (IIIrd) nerve. Whether prechordal cells are the exclusive source of myoblasts for these EOMs or share this fate with paraxial mesodermal populations is unresolved.
Paraxial mesoderm is also generated from the primitive node and streak beginning at the onset of gastrulation, and most head mesoderm has formed by the time the streak reaches its maximum length. Unlike trunk paraxial mesoderm, in which the medial and lateral parts derive from spatially and temporally distinct progenitors within or beside the primitive streak (Schoenwolf et al.,1992; Tam et al.,2000; Sporle et al.,2001; Lopez-Sanchez et al.,2001), head paraxial mesoderm appears as a rapidly generated population without spatially separable progenitors (Freitas et al.,2001).
Most head paraxial mesoderm remains mesenchymal, and the population appears homogeneous from the optic vesicles to the hindbrain, where the first somite forms either close to (birds) or several rhombomere lengths (some mammals) caudal to the otic primordium. The progressive formation of somites is coordinated by varying levels of Fgf8, retinoic acid, and Wnt signaling (reviewed by Deschamps and van Nes,2005) integrated with periodic waves of Notch pathway-related genes such as hairy1 and 2 and lunatic fringe boundaries (reviewed by Pourquie,2003; Dubrulle and Pourquie,2004). The periodicity of these transcription factors corresponds to and determines the rate of new somite formation and the locations of inter-somitic boundaries. Two waves of these Notch-driven expressions occur during the formation of unsegmented head mesoderm (Jouve et al.,2002), but the full array of possible early signals has not been assessed in this region. No epithelialization follows these waves, and to date no subsequent events within head paraxial mesoderm have been associated with this 2-cycle history (Noden and Trainor,2005).
In a series of elegant stereo SEM examinations in amphibians (Jacobson and Meier,1984), reptiles (Meier and Packard,1984), chicks (Meier,1979, 1982), and mammals (Meier and Tam,1982), the superficial and deep (ventral) surfaces of head mesoderm were found to be demarcated by a series of shallow transverse grooves, between which the mesenchymal cells showed a biased circular orientation. Meier proposed that these cryptic segmental units, named somitomeres, represented vestiges of an evolutionarily earlier condition in which head mesoderm was fully segmented. However, unlike somites in the trunk, they arrest before epithelialization and thus lack any apparent separations.
This finding was enthusiastically embraced by comparative morphologists, some of whom believed that somitomeres represented the long-sought evidence for a segmental organization within head paraxial mesoderm (Thompson,1993). However, other researchers have been unable to visualize somitomeres (Wachtler and Jacob,1986) and, thus far, no molecular evidence of a segmental organization within this mesoderm has been found (Freund et al.,1996). Moreover, the initial sites of specific muscle, cartilage, and bone determination do not correspond to individual somitomeres (Noden,1983b; Couly et al., 1992; Evans and Noden,2006). Comparative analyses show that the position of specific somitomeres relative to other axial structures, specifically mid- and hindbrain segments, is not constant among embryos of the different vertebrate classes thus far examined.
While neural crest cells require the presence of an underlying substratum, there is no evidence that the somitomeric topography provides specific guidance cues. Indeed, following focal surgical lesions to the crest population, adjacent crest cells readily alter their normal excursions relative to underlying paraxial mesoderm (Noden,1980; Grim and Wachtler,1991; Saldivar et al.,1997). Localized signals from head mesoderm are known to affect adjacent brain, neural crest, and placodal cells (Anderson and Meier,1981; Ladher et al.,2000; Trainor et al.,2002; Litsiou et al.,2005), but to date there is no evidence that a metameric organization underlies the positional specificity of these interactions.
The diversity of cell lineages that arise in head paraxial mesoderm has been well documented, especially in avian embryos, by transplantation studies (Hazelton,1972; Noden,1983b; Couly et al.,1992,1993; Koentges and Lumsden,1996) and, more recently, by the microinjection of replication-incompetent retroviruses carrying a β-galactosidase reporter gene (Figs. 6, 7; see also Fig. 10; Wahl et al.,1994; Evans and Noden,2006). These studies identified progenitors for endothelial cells, smooth muscles, and a wide range of connective tissues in addition to skeletal muscles within paraxial mesoderm.
An important early distinction among the progenitors for each of these lineages is their pattern of movement (Fig. 6). Chondrogenic mesoderm cells are essentially stationary, changing location by differential expansion and growth of prechondrogenic and early cartilage cell populations, analogous to the processes underlying vertebra and rib formation (Nowicki et al.,2003). Osteogenic precursors of intramembranous bone originate deep and medially within head paraxial mesoderm. These cells move dorsally around the brain, occupying sites where in the chick embryo the roofing part of the frontal and the parietal bones will form. In mice, the entire frontal is of neural crest origin (Jiang et al.,2002).
Angioblasts are ubiquitous throughout paraxial mesoderm. These endothelial precursors are the most aggressively motile cells in the embryo, moving invasively in all directions (Noden,1989). Angioblasts penetrate crest cell populations while the latter are en route to branchial and periocular sites, allowing for rapid vascularization of all peripheral mesenchymal tissues. Trunk paraxial mesoderm contains a small number of bipotential myoangioblasts, capable of invading peripheral tissues and, in response to as yet unknown cues, contributing to either endothelial or muscle differentiation (Kardon et al.,2003; Borue and Noden,2004). A comparable population has not been identified in the head.
Early accounts of head muscle origins in a wide range of vertebrate embryos described the initial locations of many myogenic condensations within branchial arches, and concluded that myoblasts originate within this lateral mesoderm (Adelmann,1927; Romer,1972; Jarvik,1980). With the availability of progenitor cell–labeling methods, especially in avian (Noden,1983b; Couly et al.,1992,1993; Evans and Noden,2006) and murine embryos (Trainor and Tam,1995; Gage et al.,2005), it was discovered that all head myoblasts arise in paraxial mesoderm, often far removed from the locations at which overt condensation and terminal differentiation will occur (Fig. 7).
Progenitors of head skeletal muscles are organized either as individual muscle primordia, e.g., most EOMs, or as cohorts of myoblasts that will populate specific branchial arches and later segregate into several muscles (Fig. 8). Branchial muscle primordia are located superficially within head paraxial mesoderm. As such they are initially in contact with overlying surface ectoderm and are more distant from the notochord and pharyngeal endoderm. EOM primordia tend to be closer to the midbrain and hindbrain than are branchial myoblasts (Couly et al.,1992). The precursors of the lateral rectus, especially, are located ventromedially, tucked between the trigeminal ganglion and the ventrolateral margin of rhombomere 2 (Mootoosamy and Dietrich,2002).
Occipital Somites and the Somite–Presomitic Mesoderm Boundary
Intrinsic laryngeal and tongue muscles are derived, respectively, from rostral and caudal occipital somites (somites 1–5) in amphibians (Piatt,1938), birds (Hammond,1965; Hazelton,1972; Noden,1983b; Huang et al.,1997,1999; Couly et al.,1992), and mammals (Yamane,2005). Their precursors break away from the lateral myotomes of these somites, then coalesce ventrolateral to somite 4 where they form the hypoglossal cord (Figs. 3 and 9). This cord expands ventrally, concordant with the morphogenetic movements of adjacent structures (Hilfer and Brown,1971).
Beneath the caudal pharynx, intrinsic laryngeal muscle precursors separate from the hypoglossal cord and move into lateral mesoderm adjacent to the laryngotracheal diverticulum (Mackenzie et al.,1998). The remaining myogenic cells within the cord continue shifting rostrally, moving into neural crest populations associated with the ventral aspects of branchial arches 1 through 3, wherein they form the extrinsic and intrinsic tongue muscles.
In transgenic mice lacking the c-Met receptor, which is necessary for hypoglossal cord formation and extension (Prunotto et al.,2004), some proximal tongue muscles are nonetheless present. This raises the possibility that other tongue myogenic populations may be present in mammals. Mapping data from chick embryos are not applicable to resolve this issue because, while most avian species have a robust and highly derived extrinsic tongue musculature, they lack intrinsic tongue muscles.
The boundary between unsegmented head paraxial and somitic mesoderm demarcates strikingly different tissue organizations and patterns of gene expression in amniotes. As expected, this includes genes involved in segmentation, e.g., paraxis (Fig. 9; Johnson et al.,2001), and epithelialization, e.g., eph/ephrins (Barrios et al.,2003). In addition, many genes associated with lineage commitment, e.g., scleraxis, pax3, pax1, and members of the groucho and lef/tcfI families are expressed in somites (Brown et al.,1999; Schmidt et al.,2004; Van Hateren et al.,2005), but not in paraxial mesoderm located immediately rostrally. This is surprising because all the cell types that arise within occipital somites also are generated in unsegmented head paraxial mesoderm. At later stages, most lineage-specific differentiation markers are expressed within tissues on both sides of this boundary, e.g., myf5 and myoD for skeletal muscles, Sox9 and Runx for cartilage and bone, but the upstream regulatory antecedents to these lineage markers are very different.
It could be argued that these differences are simply an exaggerated example of variations that have been documented within paraxial mesoderm at different levels of the body axis in other species. Occipital somites, for example, do not give rise to segmented skeletal structures, but instead fuse to form the occipital bone. This process depends upon proper expression of Hox genes (Lufkin et al.,1992; but see Kant and Goldstein,1999) within presomitic mesoderm (Carapuco et al.,2005) or possibly earlier. Some members of the first and second paralogous groups of Hox genes are transiently expressed in caudal head mesoderm, but their roles have not been investigated. Furthermore, while many studies have shown causal relations between boundaries of Hox gene expression and axial skeletogenesis, it is not clear whether early Hox-based positional specification extends directly to myogenic populations (Alvares et al.,2003) or affects these progenitors secondarily through their interactions with connective tissues.
The importance of this boundary goes beyond the issue of muscle origins. Somites impose many spatial constraints upon the pathways available to migrating neural crest cells (Loring and Erickson,1987; Kalcheim and Tiellet,1989; Ferguson and Graham,2004) and preclude the dissemination of large connective tissue-forming crest populations. This becomes especially problematic in unraveling the homologies in amniotes of muscles associated with ancestral caudal gill arches, which formed at the same axial levels as the rostral somites but (presumably) derived their connective tissues from the neural crest.
For example, the cranial part of the trapezius muscle is innervated by cranial nerve XI. This led some comparative anatomists to propose a branchial arch ancestry for this muscle, while others placed it in a category that includes superficial muscles of the head and neck such as the cucullaris and constrictor colli, which are considered axial muscles (reviewed by Kasakabe and Kuratani,2005). The trapezius spans from cervical vertebrae and in some species the dorsomedial margin of the occipital bone (nuchal ridge) to the spine of the scapula. In avian embryos, trapezius myoblasts originate from occipital and cranial cervical somites (Fig. 10; Huang et al., 2001).
Recent mapping studies in mice confirm a similar origin, but reveal that connective tissues associated with the trapezius are derived from the neural crest (Matsuoka et al.,2005). These accounts, in conjunction with molecular analyses (reviewed by Kasakabe and Kuratani,2005) indicate that the trapezius is similar to tongue and laryngeal muscles. Precursors for these muscles emigrate from lateral margins of occipital somites and become associated with neural crest–derived connective tissues, which ancestrally were likely associated with caudal gill elements. Thus, while the somite:presomitic boundary demarcates many anatomical and molecular differences, it does not fully delineate trunk from head muscles.
Intermediate and Lateral Mesoderms
Located lateral to occipital somites is a thin band of intermediate mesenchyme that bridges and demarcates paraxial from lateral mesoderms. Beginning close to somite 6, this band forms epithelial cords with distinct molecular signatures including the forkhead/winged helix family of transcription factors (reviewed by Jones,2003). These markers do not define comparable populations in the head of amniote embryos (Wilm et al.,2004). Curiously, cephalochordates activate these markers in a longitudinal band extending well into the head; however, this is not accompanied by overt changes in morphology (Holland et al.,2001).
In the head, lateral mesoderm loosely fills the space between surface ectoderm and pharyngeal endoderm (Fig. 1). It is continuous with and indistinguishable from paraxial mesoderm, medially, and with cardiogenic mesoderm in the ventral midline. During head folding, lateral mesoderm initially located rostral to the pharynx becomes displaced outside the body of the embryo except for precursors of the septum transversum and caudal heart tube, which are brought to the midline below the pharynx. At the same time, the cephalic extension of the coelomic cavity is obliterated, remaining only in association with cardiac structures. As such, no distinction between visceral (splanchnic) and parietal (somatic) layers is possible in the head.
Most lateral mesoderm cells in the future branchial region are angiogenic. These, together with paraxial angioblasts, infiltrate the expanding branchial and periocular neural crest populations, enabling the formation of patent aortic arches as soon as crest cells reach their terminal locations (Noden,1989).
Some of these lateral mesoderm cells also contribute to the myocardium of the outflow tract. These cells, part of the secondary heart field, become progressively incorporated into the cranial aspect of the heart tube as it moves caudally from its origin ventral to the first branchial arch (Waldo et al.,2001). The outflow tract is also invaded by angioblasts derived from head paraxial mesoderm (Noden,1991).
MOLECULAR BIOGRAPHIES OF DEVELOPING HEAD MUSCLES
The patterns of gene expression in embryonic head muscles show substantial heterogeneity, both amongst each other and in comparison with trunk populations. All head muscles express the myogenic determination transcription factors, myf5, then myoD. However, the initial expression of these genes is delayed compared with most somitic muscles. The expression of both is prolonged, often extending for several days before differentiation proceeds, as characterized by the synthesis of desmin (Hacker and Guthrie,1998) and myosins (Noden, et al.,1999).
Myf5 expression in the cervical and occipital somites is regulated by a unique promoter, and loss of gene function (e.g., six1/4, pax3, CREB, or shh) has been shown to differentially affect myogenic differentiation in this region (Dietrich et al.,1999; Chen et al.,2005; Grifone et al.,2005). These differences may be the result of the combinatorial Hox code along the body axis, changes in Notch signaling, and/or the retinoic acid gradient in the mesoderm underlying the developing hindbrain (Burke,2000; Maden et al.,2000; Cordes et al.,2004; Vermot and Pourquie,2005).
Heterogeneity between the different head and trunk muscles is further exemplified by analyses of the myf5 regulatory sites (Fig. 11), in which distinct enhancers regulate myf5 expression both temporally and spatially within the developing embryo in a combinatorial and co-operative fashion. There are at least 6 enhancer sites that direct myf5 expression in somite compartments, three for the limb, and two for the hypoglossal cord (Summerbell et al.,2000; Carvajal et al.,2001; Gustafsson et al.,2002; Teboul et al.,2002; Buchberger et al.,2003; Hadchouel et al.,2003). Myf5 expression within the craniofacial musculature is also differentially controlled (Summerbell et al.,2000). Each of the several branchial arch muscle enhancers will drive myf5 expression, but loss of function studies suggest these regulatory sites operate as part of an integrated modular controlling network rather than as isolated elements (Rigby, personal communication).
The myoD enhancer is much simpler and can be broken down to the 258-bp core enhancer, which controls the onset of myoD expression, and a distal regulatory region, which regulates myoD expression in differentiating myoblasts (Goldhamer et al.,1995; Kucharczuk et al.,1999; Chen et al.,2001). Even within the short core enhancer, myoD expression in the somites versus the limb and branchial arches is differentially regulated, again indicating that the signaling networks regulating the onset of myogenesis are not mimicked throughout the embryo (Kucharczuk et al.,1999).
Other regulatory genes common to subsets of both head and trunk muscles are the transcription factors, myoR, meox1, six1 and -4, barx2, the homeobox transcription factor pitx2, and the c-Met ligand HGF (hepatocyte growth factor) (Fig. 12; Candia et al.,1992; Lu et al.,2002; Esteve and Bovolenta,1999; Smith and Tabin,1999; Ozaki et al.,2001, Meech et al., 2003; Bessarab et al.,2004; Noden and Francis-West, unpublished data). The order of expression of these genes is often dissimilar between head and trunk.
Loss of function studies of these genes typically result in less severe disruption of head myogenesis compared with the trunk. Inactivation of the homeobox gene, six1, affects the diaphragm and abdominal wall muscles together with a subset of limb and tongue muscles but leaves epaxial and other head muscles relatively unaffected (Laclef et al.,2003; Li et al.,2003). In the double Six1/Six4 knockouts, these defects are exacerbated. Moreover, the defects now extend to the head muscles, delaying extraocular but not branchial arch myogenesis (Grifone et al.,2005).
Myogenic lesions following loss of pitx2 function in mice are found only in the extraocular muscles, which fail to condense (Kitamura et al.,1999). In this initial study, it was not evident whether the muscle dysgenesis resulted from an intrinsic defect in the developing myoblasts or was secondary to loss of pitx2 expression in the periocular mesenchyme. Our own studies have suggested that pitx2 regulates myogenic cell number cell autonomously, consistent with its ability to regulate cell proliferation in the satellite cell line, C2C12 (Kioussi et al.,2002; Francis-West, unpublished data). This has been confirmed recently by the demonstration that gene-inactivation of pitx2 in the periocular cranial neural crest cell does not affect the early differentiation of eye muscles (Evans and Gage,2005).
Many genes essential to the formation of subsets of trunk muscles are either not expressed in head muscles or only in very limited circumstances. Pax3, for example, is an essential upstream element for appendicular and some hypaxial muscle differentiation (Relaix et al.,2003), but is not activated in head muscle precursors. Transplantation data suggest that the signals necessary to activate or maintain pax3 expression are lacking in the head (Hacker and Guthrie,1998). Yet, curiously some of the down-stream targets of pax3 such as lbx1 and paraxis, characteristic of subsets of somitic muscle cells, are expressed in the lateral rectus and, for lbx1, the dorsal oblique muscle (Mootoosamy and Dietrich,2002; Borue and Noden,2004).
Conversely, some head muscles or their precursors express several regulatory genes, including tbx1 (Fig. 13), engrailed, and capsulin (epicardin, POD1) that are not typically part of the axial myogenic repertoire (Robb et al.,1998). Inactivation of Tbx1 in mice or zebrafish (van gogh) causes severe reduction of branchial arch muscles (Piotrowski et al.,2000; Kelly et al.,2004), sparing only a small number of myoR and capsulin-expressing muscle cells in the mandibular prominence. These survivors indicate the presence of heterogeneity among branchial arches either in the myogenic populations or the surrounding signaling environments. Microarray analysis of E9.5 murine tbx1 mutants reveals a significant upregulation of myoR in the absence of tbx1 function (Ivins et al., 2005). MyoR, was originally identified as a myogenic repressor and it is possible that this upregulation is linked to the first arch myogenic defects in the tbx1 mutants (Lu et al.,1999).
Tbx1 expression is widespread through head mesoderm prior to the onset of myogenesis, and also is present in pharyngeal endoderm, subsets of neural crest cells, and surface ectoderm (Garg et al.,2001; Robson et al., unpublished data). Mesodermal tbx1 expression is modulated bimodally by levels of retinoic acid and by Shh signaling via Fox transcription factors. Tbx1 in turn regulates expression of fgf10 and, additionally, in the secondary heart field, fgf8 (Garg et al.,2001; Yamagishi et al.,2003; Roberts et al.,2005). Functional T-box binding sites have been identified in the Xenopusmyf5 promoter, suggesting a cell-autonomous role (Lin et al.,2003).
Similarly, T-boxes are also thought to regulate skeletal muscle differentiation cell autonomously in sea urchin and ascidians (Mitani et al.,2001; Croce et al.,2001). Our analyses using in vitro micromass cultures and of the mesoderm-specific tbx1 knockout mouse support this autonomy, although to date mutational analyses of the most highly conserved T-box binding site in the mouse myf5 branchial promoter indicate that this site alone is not essential for myf5 expression (Kelly et al.,2004; Robson et al., unpublished data). However, it should be noted that even if T-box regulation of myf5 expression is ultimately demonstrated, tbx1 alone cannot specify myogenic fate but must act in conjunction with additional transcription factors because it is expressed in several other mesodermal derivatives (Rodriguez-Esteban et al.,1999).
HGF (hepatocyte growth factor), the ligand for the c-Met receptor, and SDF1, the ligand for the chemokine CXCR4 receptor, are expressed in appendicular and caudal pharyngeal regions, and are necessary for the delamination and migration of myoblasts from the lateral myotomes at these levels (Sonnenberg et al., 1993; Dietrich et al.,1999; Vasyutina et al.,2005). Both act as trophic factors for migratory myoblasts (Brand-Saberi et al.,1996; Hayashi and Ozawa,1995).
During early head muscle morphogenesis stages, HGF is expressed in first branchial arch muscle populations (Caton et al.,2000) and also in and around the dorsal oblique condensation (Fig. 12). In these contexts, however, HGF acts as a chemoattractant for outgrowing motor axons (Irving et al.,2002). C-Met loss-of-function studies reveal a key role for this signaling pathway in the formation of the muscles of facial expression, e.g., the orbicularis, buccinator, and platysma (Prunotto et al.,2004), in addition to most (but not all) tongue muscles. Neither the mode of action of HGF nor the actual cellular activities by which facial myoblasts disperse from the second branchial arch are known.
MECHANISMS OF CRANIOFACIAL MYOGENESIS
The preceding sections have emphasized unique attributes of developing craniofacial muscles, both individually and as groups, and revealed significant disparities between head and trunk myogenic populations. Relatively little is known of the mechanisms underlying this diversity. Extrapolating from analyses of trunk myogenesis, tissue interactions that control development of primary head myofibers will involve multiple surrounding tissues, including other members of the paraxial mesoderm community (Fig. 14). These interactions will require a consortium of molecular signals to produce a combination of positive inductive and inhibitory signals (reviewed in Pownall et al.,2002; Emerson and Hauschka,2004; Brand-Saberi,2005).
Overview of Axial Muscle Differentiation
In chick and Xenopus embryos, initial induction of trunk myogenesis as measured by the onset of myf5 and myoD expression occurs broadly within unsegmented presomitic mesoderm in a cell-autonomous fashion several hours in advance of somite formation. In chicks, wnt5b expression in presomitic mesoderm initiates myoD expression, but levels of expression of both myf5 and myoD are kept low through the action of bone morphogenetic proteins (Bmps) (Kopan et al.,1994; Cossu et al.,1996; Dosch et al.,1997; Gerhart et al.,2000; Kiefer and Hauschka,2001; Linker et al.,2003; Galli et al.,2004).
Surface ectoderm, the dorsal neural tube, and lateral mesoderm are all sources of Bmp (Pourquie et al.,1996; Amthor et al.,1999). With the onset of somite formation myogenic repression is overridden by the production of BMP antagonists such as noggin, follistatin, and flik within the developing somite (Amthor et al.,1999; Zimmerman et al.,1996). This is evidenced first in the medial margin of the somite with elevated myf5 expression and later in lateral and boundary cells where both myf5 and myoD are activated (Montarras et al.,1991; Sassoon et al.,1989; Hirsinger et al.,2001). Noggin secretion is in response to Wnt signalling from the neural tube (Hirsinger et al.,1997; Marcelle et al.,1997; Reshef et al.,1998).
Wnts continue to have a role in somite differentiation, and several tissues adjacent to somites produce members of the Wnt family of growth factors, e.g., Wnt4, -6, -7a from surface ectoderm and Wnt1, -3a, and 4 from the neural tube (Fig. 14; Münsterberg et al.,1995; Fan et al., 1997; Tajbakhsh et al.,1998; Wagner et al., 2000). These Wnt signals continue to promote and co-ordinate myogenesis, but the mechanisms vary between the medial and lateral myotomes. Some Wnts induce the expression of the intrasomitic transcriptional modulators of Shh signaling, e.g., Gli 2 and -3, in the medial somite compartment. This allows amplification/maintenance of the committed myoblasts (Borycki and Emerson,2000).
Shh is produced by the notochord and ventral spinal cord (Münsterberg et al.,1995; Fan et al.,1997; Tajbakhsh et al.,1998; Wagner et al., 2005). Wnt3a in conjunction with Shh has been shown to have positive mitogenic/survival effects on myoblasts (Teillet et al.,1998; Marcelle et al.,1999; Krüger et al.,2001; McDermott et al.,2005). In the lateral myotome, Wnt signaling acts independent of Shh, signaling instead via a novel pathway involving the transcription factor CREB (Chen et al.,2005). The FGF family is also involved. FGFs acting via FGFR1 maintain proliferation and retard differentiation, while FGF signaling via the FGFR4 receptor has the converse effect, promoting myogenic differentiation (Itoh et al.,1996; Marics et al.,2002).
Induction and Differentiation
The many apparent differences in organization and nearest neighbor relations between head and trunk myogenic populations would suggest that underlying regulatory interactions might be qualitatively different along the body axis. However, muscle precursor transplantation experiments have shown that paraxial mesoderm from all parts of the body axis has broadly shared response capabilities. Presomitic (segmental plate) trunk mesoderm cells grafted alone or with accompanying surface ectoderm into head mesoderm are able to respond to local cues and form normal EOMs or branchial arch muscles (Fig. 15; Noden,1986; Borue and Noden,2004). This response includes expression of appropriate head but not trunk muscle transcription factors according to the normal head myogenic timetable. Preliminary results of converse grafts reveal that head mesoderm is similarly capable of contributing to epaxial and hypaxial trunk muscles, but not to appendicular muscles (Westesson and Noden, unpublished data). The latter may reflect an inability to activate c-Met.
In these trunk-into-head experiments, however, some grafted cells fail to become assimilated into the head myogenic program, but rather mimic the trunk myogenic scenario. Many form epithelial condensations containing cells expressing myf5, myoD, paraxis, and pax3. These subsequently differentiate as large ectopic muscles close to the site of implantation (Fig. 16; Hacker and Guthrie,1998; Mootoosamy and Dietrich,2002; Borue and Noden,2004). Other grafted cells disperse and form scattered islands and clusters of individual myocytes; these are primarily located along the routes normally traversed by EOMs (discussed later).
There are two nonexclusive explanations for these multiple outcomes. One is that the grafted tissue was heterogeneous, containing some myoblasts already committed and awaiting release from BMP inhibition to execute their prior decision, a second population poised to migrate towards HGF signals, and a third that remains responsive to novel myogenic environments. Alternatively, the dual outcome may reflect different local inductive niches in the head that override or modulate prior commitments within grafted cells.
Cranial myogenic mesoderm is variably proximate to a number of tissues (Fig. 13), e.g., the neural tube, notochord, pharyngeal endoderm, surface ectoderm, and neural crest cells, each of which harbors cells simultaneously making localized commitments to different lineages. For example, within surface ectoderm overlying several head muscle precursors are sites at which neurogenic placodes for trigeminal, facial and glossopharyngeal sensory ganglia are established. The initial specification of these placodal fields are determined by combinatorial signaling involving members of the Wnt, Fgf, and Bmp families that are produced by underlying paraxial mesoderm: Bmps and Wnts are inhibitory for placodal specification while FGFs are inductive (Litsiou et al.,2005).
Additional region-specific signaling scenarios have been identified within the paraxial mesoderm underlying the otic placode. Here the mesoderm produces Fgf3, Fgf19, and Tbx1. During earlier premyogenic stages, cranial mesoderm expresses Wnt and Bmp antagonists (Ogita et al.,2001; Chapman et al.,2002). It is likely that these early gradients and foci of intra-mesodermal activity persist or leave an imprint that subsequently creates different response biases within myogenic mesoderm and surrounding tissues.
In vitro analyses of head paraxial mesoderm cocultured with individual or combinations of surrounding tissues reveal that factors released by the hindbrain and surface ectoderm depress the ability of mesoderm cells to initiate (or elevate) expression of myf5 (Tzahor et al.,2004). This repression is mediated by Bmps and Wnts from the neural tube and ectoderm. Removal of the ectoderm promotes the precocial onset of myogenesis within head mesoderm in culture (Tzahor et al.,2004).
Normally, the neural crest population expands over the precursors of 1st branchial arch muscles, effectively separating myogenic cells from this overlying source of inhibition. Furthermore, crest cells express the Wnt antagonist Frzb, and the Bmp antagonists noggin and gremlin. The prompts for these paracrine secretions are not known, but influences from underlying paraxial mesoderm need be considered (Farlie et al.,1999).
Wnt expression in the head is not uniform along either the rostro-caudal or ventro-dorsal axes of the brain. Wnt3a expression is confined to the caudal hindbrain (myelencephalon), while wnt5a transcripts are present in the rostral hindbrain (metencephalon); wnt7a is absent from the ventral midbrain. Wnt1 exhibits a dynamic expression pattern in the midbrain, and except for the dorsal midline is not expressed in the metencephalon (Parr et al.,1993). It can be envisaged that local areas of lesser Wnt release might correspond to specific myogenic foci within paraxial mesoderm. Unfortunately, only 1st arch muscle precursors have been assayed for their responses to Wnt and Bmp signals.
Repression of myogenesis by Bmps and its override by noggin is an attribute common to both head and trunk myogenic populations. In contrast, the negative effect of Wnts upon head myogenic cells is opposite to their enhancing effect upon myotome cells, revealing a totally unsuspected novelty of craniofacial myogenic differentiation. Shh is released by the notochord and ventral neural tube in the head, and also by bands of pharyngeal endoderm cells. Despite the key role of Shh in the maintenance of epaxial trunk muscles, cranial mesoderm explants failed to demonstrate any physiological role for Shh in the initial regulation of myogenic differentiation (Tzahor et al.,2004), consistent with an absence of the patched receptor in head paraxial mesoderm at stages immediately antecedent to the onset of myogenesis (Noden, unpublished data). This lack of Shh responsiveness may be augmented by the co-expression of Bmp7 from the floor of the diencephalon and midbrain and also the notochord (Dale et al.,1999). Bmps can modulate and inhibit Shh activity (Dale et al.,1999; Sterneckert et al., 2005; Rios et al.,2004). In this respect, cranial mesoderm resembles lateral myotomes, which are Shh independent.
Again, however, caution should be exercised as different head myogenic populations were not assayed in these studies. The lateral rectus, for example, becomes determined and initiates differentiation very close to the floor of the hindbrain and notochord at the level of rhombomere 2. This location is proximate to Shh signaling, and corresponds to the axial level at which Bmp7 expression is waning and netrin-1, a target of Shh signaling, is activated (Dale et al., 1999). This muscle anlagen also expresses Tbx1, a down-stream mediator of Shh signaling (Garg et al., 2001).
Pharyngeal endoderm plays key roles in patterning facial structures and is required for head mesenchymal cell survival. For example, Fgf19 expression within deep paraxial mesoderm is regulated by Fgf8 released by underlying endoderm (Ladher et al., 2005). This Fgf subsequently regulates the expression of wnt8c in the hindbrain, initiating a cascade of signals that are necessary to specify and induce otic vesicle development (Ladher et al.,2000). Similarly, TGF signaling from the anterior endoderm determines prechordal plate character (Vesque et al.,2000). The possible effects of endodermal signaling upon head myogenesis, either directly or secondary to actions upon neural crest cells, have not been examined.
Finally, we must consider intra-mesodermal interactions, which in somites are essential to specify both the medial and lateral myotomal regions. Thus far, the roles of Bmp, Wnt, and Fgf signals produced within head mesoderm have not been analyzed, nor have essential signal transduction and processing capabilities analogous to those involved in Shh/gli pathway responses within somites (Borycki and Emerson,2000) been examined. Here, again, it is especially difficult to extrapolate from epithelial (somitic) to mesenchymal populations.
Morphogenesis of Head Muscles
Muscle morphogenesis encompasses three integrated processes: the relocation of muscle precursors from their site of origin to the location of their terminal differentiation, the assembly and alignment of primary myotubes, and the segregation of individual muscles with appropriate attachments to surrounding connective tissues. The diversity of relocation strategies is exemplified by (1) the migration of individual myoblasts into the limbs and facial regions, (2) expansions of sheets or cords of myogenic precursors (e.g., muscle buds, Christ and Ordahl,1995; Burke and Nowicki,2003) to form hypaxial body wall and branchial arch muscles, and (3) translocations of condensed cohorts of differentiating muscle cells as seen during EOM development (Wahl et al.,1994). In some cases, e.g., tongue muscles and the diaphragm, a combination of these strategies is likely employed.
Migrating myoblasts emigrate from the lateral myotomes at specific, restricted sites (occipital, cervical, brachial, lumbosacral). Most (all?) of these cells are pax3-dependent (Brown et al.,2005), and express the transcription factors lbx1, paraxis, and six1 (Delfini and Duprez,2000; Gross et al.,2000; Uchiyama et al.,2000; Williams and Ordahl,2000). Lbx1 and both the c-Met and CXCR4 receptors are essential for their emigrations. Migrators from each of the several occipital somites contribute to all intrinsic and extrinsic tongue muscles; therefore, the spatial organization for individual muscles occurs either en route or upon the arrival of myoblasts at their final destination.
The formation of hypaxial musculature, which includes axial flexor and most body wall muscles, is often compared with appendicular myogenesis, as both populations share several markers (e.g., lbx1, pax3, paraxis, Tbx1) and can be driven by the same promoter sites (e.g., pax3, Brown et al.,2005; reviewed by Dietrich et al.,1998; Buckingham,2003). However, body wall and limb muscles are derived from separate myotomal compartments. The former do not express the c-met receptor and do not migrate as single cells into the periphery. Instead, hypaxial myoblast populations, together with their accompanying connective tissue partners, expand ventrally from each somite (Nowicki et al.,2003; Burke and Nowicki,2003).
All head muscles emigrate from their sites of origin, but except for some hypoglossal and superficial facial muscles, migration of individual myoblasts is not part of this process. Neural crest cells en route to each arch overlie presumptive branchial muscle primordia, and both of these mesenchymal populations shift ventrally in concert (Fig. 2). During this process, crest cells circumscribe but do not penetrate the condensed cord of muscle precursors. Scattered angioblasts are present within the cord, but blood vessels are lacking (Ruberte et al.,2003). Shortly thereafter, motor axons contact and enter the myogenic cord (Song and Boord,1993; Kuratani and Tanaka,1990).
EOM primordia achieve their terminal locations by a combination of population expansion, extended, multidirectional movements, and changes in the relative position and orientation of the eye due to cranial flexure (Fig. 17). These populations of myogenic cells, often already including multinucleated myotubes (Wahl et al.,1994), move en masse into neural crest–occupied periocular territories (Wahl and Noden,1997; Creuzet et al.,2005). Migration of single cells does not occur and these myoblasts do not express the c-Met receptor. Bringing EOMs into the periocular microenvironment is essential not only for their subsequent biomechanical roles. Reciprocal interactions between the eye and EOMs are necessary for both differentiation of the retina (Kablar,2003) and formation of secondary motubes (Amprino, 1952; Twitty,1932).
From its origin lateral to the midbrain, the dorsal oblique primordium shifts rostrally more than 90° around the equatorial margin of the optic cup. These movements distance it from its initial neighbor, the lateral rectus precursor, and, remarkably, cause it to pass the ventral oblique primordium that is moving in the opposite direction (Noden,1983a; Noden et al., 1999; Marcucio and Noden,1999). Analyses of mouse embryos expressing LacZ linked to promoter elements of myogenin (Cheng et al.,1992) or myoD (Goldhamer et al.,1995) reveal similar patterns of EOM movement, confirming classical descriptions by Gilbert (1947,1957) in cat and human embryos. In these species, the primordia of all muscles that will become innervated by nerve III coalesce into a common supraorbital mesenchymal condensation (premandibular condensation). Comparable descriptive studies in Zebrafish also indicate that EOM precursors move from their sites of origin, based on studies of myoD expression and myosin synthesis (Easter and Nicola,1996; Schilling and Kimmel,1997), although detailed lineage tracing of their progenitors has not been reported.
These translocations bring each EOM primordium into periocular mesenchyme derived from neural crest cells. The crest-mesoderm interface is undetectable in normal embryos, but is easily identified in chimeric quail-chick/quail-duck embryos (Noden,1978, 1983a; Helms and Schneider, 2003; Eames and Schneider,2005) or mouse embryos in which crest cells are labeled (Jiang et al.,2002; Matsuoka et al.,2005; O'Gormon,2005). The processes driving movements of tightly condensed EOM primordia remain enigmatic. Models based on active “migration” of these cohorts of myotubes and myocytes (Fig. 18A) seem improbable and lack precedence. Models that propose shifts in the location of the interface (Fig. 18B) are inconsistent with the mapping data, which show that these EOM primordia dramatically change their positions relative to many other nearby structures.
Borue and Noden, (2004) have proposed a passive displacement model based on changes in the contour of the neural crest:mesoderm interface. Following transplantation of trunk paraxial mesoderm into sites normally occupied by lateral rectus and some dorsal oblique progenitors, these authors found a discrete band of small muscles and individual myotubes distributed along the pathway traversed by the dorsal oblique muscle (Fig. 16). This suggests that the dorsal oblique is not a moving island, but rather is at the tip of a finger-like projection of mesoderm that penetrates and interdigitates with comparable but oppositely oriented projections of neural crest cells (Fig. 18C). Normally, periocular crest cells would encircle and isolate the myogenic tip, perhaps similar to the process by which branchial crest cells circumscribe the myogenic core within each arch. However, this final step is blocked when ectopic trunk mesoderm cells are present, as has been previously described (Noden,1986).
In this “deformed interface” model, there is no requirement for active migration of cells or myogenic condensations (Noden and Trainor,2005). Deformation of the crest-mesoderm interface could be driven by a combination of underlying processes. These include matrix-driven translocations as described by Newman (2003; Newman and Comper,1990), differential cell adhesivity, possibly involving members of the eph/ephrin family that mediate repulsive and attractive cell interactions (Poliakov et al.,2004; Twigg et al.,2004), and local differences in the rates of mesenchymal cell proliferation or apoptosis.
The next phase of muscle morphogenesis, segregation of individual muscles from common progenitor populations, is foreshadowed by different alignments of elongated myocytes and primary myotubes within mesodermal condensations. Within the 1st branchial arch muscle plate, for example, the orientation of myotubes varies at different locations (Fig. 19; McClearn and Noden, 1989; Noden et al.,1999). As in the limb, these morphological asymmetries identify the precursors of each specific muscle that subsequently will break away from the common progenitor pool (Robson et al.,1994; Kardon, 1998). Concomitant with these alignments, connective tissue elements derived from the neural crest penetrate the muscle column. This occurs at the interface between eccentrically aligned myotubes, and marks the sites of subsequent separation of individual muscles (Bogusch,1986). At this stage, endothelial cells stream into the previously avascular muscle primordium (Ruberte et al., 2002). While a few participating molecules are known (Sheela et al.,2005), the molecular basis for these alignments and separations is not known.
In the limb, the transcription factors Tcf4 and scleraxis are expressed in multiple foci within limb mesenchyme (Kardon et al.,2003; Schweitzer et al,2001; Perez et al.,2003). Tcf4 expression precedes muscle formation and is found in the pre-tendinous and fascial mesenchyme surrounding the developing limb myoblasts. Disruption of Tcf4 function in chick embryonic limb buds alters muscle patterning (Kardon et al.,2003). To date, neither Tcf4 nor scleraxis expression has been detected within head mesoderm or neural crest populations.
The separation of individual muscles and their formation of attachments with branchial skeletal and connective tissues is dependent upon positional cues provided by surrounding connective tissues, in this case largely derived from the neural crest (Francis-West et al.,2003). Manipulation of branchial arch crest populations, either by transplantation (Noden,1983a; Gross and Hanken, 2005) or genetic alteration (Trainor and Krumlauf,2001; Santagati et al.,2005; Baltzinger et al.,2005), results in a patterned alteration in the organization of muscles that mimics the changes in skeletal organization (reviewed in Noden and Trainor,2005). Despite the regional differences in pre-tendon signals, there must be extensive overlap in the consortium of signals controlling muscle morphogenesis in the head and trunk, as transplanted mesoderms cells are fully able to respond to cues in either location.
The morphogenesis of many head muscles is a prolonged process during which some muscle cells form attachments far from their original branchial confinements. In a nicely detailed analysis of these movements, Koentges and Lumsden (1996) showed that neural crest and myogenic populations often maintain their original close registration even when they exit the arch in which they originated. Together, these establish attachments with skeletal elements in adjacent branchial arches and also with the mesodermal cartilages in laryngeal, neurocranial, and shoulder regions (Matsuoka et al.,2005). This congruency between myogenic and crest-derived connective tissue precursors is not established until after crest cells have fully occupied the branchial arches and circumscribed each muscle plate (Evans and Noden,2006).
Some later remodeling of branchial muscle attachment sites occurs independent of the neural crest. The avian mandibular depressor, for example, undergoes a series of changes in its proximal attachment site, from early contacts with the quadrate and squamosal elements to intermediate association with otic and petrosal structures and finally to attachments upon the roof of the occipital bone. The mechanisms by which individual muscles detach and then reattach to separate skeletal structures are not known. The rule of congruence between myoblasts and neural crest cells also does not apply to the mimetic muscles (muscles associated with facial expression). O'Gorman (2005) examined these in mouse embryos in which all 2nd branchial arch neural crest cells were labeled, and found that these myoblasts move into midfacial and jaw territories populated only by frontonasal and 1st branchial arch crest cells.
While it is presumed that periocular neural crest cells play a similar role in EOM morphogenesis, direct evidence is lacking. Gross disruptions of either eye development (e.g., Cahn,1958; Tanaka et al.,1987) or periocular neural crest (Evans and Gage,2005) result in dysmorphogenesis of EOMs but data showing a more specific role in muscle patterning are not available.
SUMMARY AND PERSPECTIVES
Commitment to the myogenic lineage within head mesoderm occurs in synchrony with the spatial determination and differentiation of many other cell types, both within mesoderm (e.g., angioblasts, connective tissues) and among neighboring tissues (e.g., optic vesicle, midbrain, rhombomeres, placodal epithelium, neural crest, pharyngeal endoderm). How each lineage within head mesoderm becomes specified and integrated with similar processes ongoing in neighboring tissues is largely unknown, and it is unclear whether comparable events occurring in epithelial somites provide helpful models. The continuous movements of crest and mesoderm relative to each other and nearby epithelia make dissecting primary muscle-inducing events in the head especially challenging.
Craniofacial and trunk muscles exhibit similar progressive stages of myogenesis, and have the ability to partially substitute for one another. Certainly, knowledge available concerning the manifold cell and molecular pathways operating during trunk myogenesis provides an important foundation for studies of craniofacial skeletal myogenesis. However, a growing body of data reveals significant differences in both signaling environments and response capabilities among myogenic mesoderm populations. These differences occur between the head and trunk (somitic) cells and also amongst craniofacial muscle populations, e.g., branchial versus extraocular muscles. Therefore, extrapolating from trunk myogenic scenarios to the heterogeneity seen during early branchial and EOM differentiation is very problematic.
In the trunk, myogenic activation/competence, as assessed by the expression of myf5 transcripts, is found broadly in the presomitic mesoderm and subsequently becomes localized to specific regions of the somite by a combination of repressive and inductive/maintenance signals. In the head, it is not known whether myoblast induction initially occurs broadly throughout the unsegmented paraxial mesoderm, or instead occurs focally at sites where individual muscle primordia form; both scenarios could be occurring. Certainly, the segmental organization of adjacent tissues, particularly the brain and pharynx with their differing signaling capacities, allows for such region-specific influences, and may indeed be necessary owing to the absence of mesodermal segmentation.
The application of transgenic mouse models provides wonderful opportunities to augment data generated largely through experimental analyses in the chick embryo. Frustratingly, many potentially useful experimental models have been generated but muscle development has not been examined. Examples include several elegant analyses of signaling gradients, e.g., retinoic acid and the isthmic organizer (Sato et al.,2004; Nakamura and Watanabe,2005), and gene functions, e.g., hox, krox, otx, fgf, all of which change the spatial identity of brain regions and may have consequences for the differentiation and morphogenesis of the craniofacial muscles.
Similarly, analyses of signaling pathways and activities typically do not extend into the head. When and where, for example, are Fgf/Wnt/Bmp pathways active in head mesoderm and how do these change across the trunk/head interface? Examples include analysis of the Shh and Wnt signaling pathways, including assays for activation of target genes such as ptc1 and axin2, respectively. The grand diversity of cell behaviors and transcription factors expressed among different head muscle groups needs to be viewed as a wonderful opportunity for comparative analyses.
In somites, distinctions among medial versus lateral versus intermediate myotome regions reflect differences in signaling environments, response capabilities, and subsequent cell behavior. Dissection of these differences has taken several decades, and has required a combination of critical mapping studies (e.g., Ordahl and LeDouarin,1992), lesioning experiments, and the identification/manipulation of unique molecular markers and genetic requirements, such as pax3, for the development of the hypaxial versus epaxial musculature. However, it is unknown whether analogous spatial or temporal differences based on either ancestry or local neighborhoods exist within head paraxial mesoderm. The identification of genes that label all (e.g., myf5, myoD, myogenin, barx2, capsulin, pax2, myoR) or mark the various subsets of craniofacial muscles (e.g., tbx1, engrailed 1, lbx1, paraxis, hgf) provides exciting entrees into targeted misexpression analyses. Frustratingly, we still lack both the elusive EOM-specific marker, if it exists, and do not yet know how myf5/myoD expression is activated in these cells. Embryological and molecular data strongly suggest that EOMs are a heterogeneous group, with the lateral rectus having an especially unusual molecular repertoire. Yet, transplantation studies indicate that a convergence of trunk and head myogenic signal-response capabilities can occur.
Our understanding of the several steps during head muscle morphogenesis is particularly anemic. The cellular basis for translocation of condensed muscle masses from mesodermal to neural crest environments is unexplored. So, too, are the signals by which myocytes of each subpopulation within a common muscle mass establish their initial axis of elongation. This alignment sets the stage for multiple asymmetric intrusions by surrounding neural crest and angiogenic populations, and presages the separation of individual muscles. Transplantation studies indicate that these processes are largely dependent upon signals from surrounding tissues, and reveal a commonality regardless of whether the pattern-imposing population is neural crest (head), sclerotome (axial), or lateral mesoderm (appendicular). Boundaries of somitic epithelium align primary myofibers in zebrafish (Henry et al.,2005), but how similar events occur in mesenchymal contexts is not known.
Our goal in preparing this review has been to highlight the many opportunities for comparative and analytical investigations of branchial and EOM myogenesis. The induction, differentiation, and morphogenesis of craniofacial muscles need not be viewed as an unapproachable morass of complex tissue relations and intractable anatomy, but rather as a fertile and relatively untouched area for multifaceted experimentation, using available tools to dissect the signaling cascades and regulatory genes that specify each group of craniofacial muscles. We cannot predict whether future research will unify head and trunk myogenesis as a few variations upon a common theme, or instead will reveal fundamental differences owing to their long evolutionary separation and absence of epithelial origins. If this review provokes some to engage in the adventure, we will happily consider it a worthwhile investment.
The authors thank Andrea Münsterberg, Chris Wahl, Natali Krekeler, and the reviewers for Developmental Dynamics for constructive suggestions on the manuscript. Much of the research included in this review has been generously supported by grants from the NIH (DE06632, EY15917) to D.M.N. and BBSRC to P.F.W.
Current nomenclature recommends using the term pharyngeal arch to describe each of the transverse swellings, including the lower jaw, that form on the lateral side of the embryonic head, and restricts branchial arch to those that form gill structures in anamniotes. However, in mammals pharyngeal muscles are a specialized subset of muscles. Thus, to avoid confusion we will use the term branchial arch in the traditional manner to include all embryonic circumpharyngeal and perioral swellings regardless of their subsequent fates.