The Synchrony and Cyclicity of Developmental Events

1.1. Derivation of the Paraxial Mesoderm during Embryogenesis

The origin of the somites is the paraxial mesoderm, which is derived from the epiblast during gastrulation. The paraxial mesoderm is bilaterally located on both sides of the midline structures, neural tube and axial mesoderm. After formation of the paraxial mesoderm via ingression of the epiblast, the most posterior parts of the tail region maintain their undifferentiated state and retain the potency to generate either the neural tube or paraxial mesoderm. The Wnt signaling molecules and a T-box transcription factor are implicated in these fate decision processes (Fig. 1A). In the absence of the Wnts (Wnt1 and Wnt3a) or of Tbx6, cells choose the neural tube fate over the paraxial mesoderm, which results in a triple neural tube formation (Chapman and Papaioannou 1998; Yamaguchi et al. 1999). Recently, it has been reported that the Tbx6-dependent regulation of Sox2 determines the fate of the axial stem cells (Takemoto et al. 2011). Paraxial mesoderm-specific misexpression of a Sox2 transgene in wild-type mouse embryos results in ectopic neural tube development. Thus, Tbx6 may repress Sox2 and thereby inhibit neural development, an essential step for the specification of the paraxial mesoderm from the axial stem cells. Hence, the tail bud region acts as a signaling center for normal mesodermal tissues including somites.

1.2. The Derivation of the PSM and Somite Properties Provide a Base for Metameric Structures

The paraxial mesoderm derived from the tail bud shows mesenchymal and immature characteristics as compared with the mature somite, and is thus referred to as the PSM. The tail bud provides a stem cell-like population supplying emerging new cells at the posterior end of the PSM, while cells are pushed anteriorly and finally bud off as a somite from the anterior part of the PSM. This is accompanied by a mesenchymal-epithelial transformation. The somites therefore consist of spherical epithelial cell masses that differentiate into dermis, muscle, and skeleton according to their position within the somite (Fig. 1B). The most striking feature of somite formation is the periodic generation of metameric structures. The human body is composed of segmental units, although this is not obvious from its external structure. Internally, these are easily recognizable. The most prominent structures in this regard are the vertebra and ribs, which are directly derived from somites. The nervous system and vasculature networks also exist in tandem and the major networks are segmental. These structures are established through the behavior of precursor cells. Because the neural crest cells derived from the dorsal part of the neural tube migrate laterally through the rostral compartment of the somite only and never enter the caudal half, the metameric arrangement of the peripheral nervous system is necessarily determined by the metameric nature and arrangement of the somites (Bronner-Fraser 1986; Teillet et al. 1987; Keynes and Stern 1988). This differential cell behavior reflects the different gene expression patterns between the rostral and caudal halves of the somites (Fig. 2). The initial vascular network is also dependent on the somite structures because intersomitic vessels are formed and established in conjunction with somitogenesis.

1.3. Lineage of Somitic Cells and Cell Differentiation (Dorso–Ventral Patterning)

Once epithelial somites are generated, the somites are patterned in response to signals from the surrounding tissues into distinct compartments giving rise to different cell lineages. The ventromedial somite cells lose their epithelial properties to form the mesenchymal sclerotome, expressing transcriptional regulators such as Pax1 and Pax9 under the influence of sonic hedgehog and noggin from the midline notochord and floor plate (Fan and Tessier-Lavigne 1994; McMahon et al. 1998; Furumoto et al. 1999). The sclerotome, together with the notochord, subsequently differentiates into the vertebral column and ribs. The dorsal somite, known as the dermomyotome, remains an epithelial sheet under the control of Wnts expressed from the dorsal neural tube and surface ectoderm. Slightly later, cells delaminate from the dermomyotome edges and migrate underneath it to form a third compartment, the myotome, located between the dermomyotome and sclerotome. As the somite matures, the somitic compartments are further divided into subdomains with unique fates (Fig. 1C). At the dorsomedial edge or lip (DML) of the dermomyotome, cells migrate underneath to generate the epaxial myotome, which then differentiates rapidly into back muscle. Central dermomyotome cells de-epithelialize to form the dorsal dermis, and at the limb bud levels, cells delaminate from the ventrolateral lip (VLL) of the dermomyotome to migrate into the lateral plate mesoderm, where they develop into limb and limb girdle muscle. At the interlimb levels, the cells from the VLL of the dermomyotome translocate underneath, producing the hypaxial myotome. The ventrolateral dermomyotome and hypaxial myotome invade the lateral plate mesoderm together as a somitic bud, from which the body wall and abdominal muscle emerge. Finally, within the sclerotome, the ventromedial cells give rise to the vertebral bodies, intervertebral discs and proximal ribs; the lateral cells, to the neural arches and distal ribs; and the dorsomedial cells, to the spinous processes (Brand-Saberi and Christ 2000; Brent and Tabin 2002).

1.4. Regionalization along the AP Axis

The vertebrate body is composed of domains (cervical, thoracic, lumbar, sacral, and caudal regions) that show different identities. These identities are defined in the tail bud region, which is controlled by Hox gene activation (Wellik 2007). It has been shown that Hox genes are involved in the timing of the ingression of paraxial mesoderm precursors into the primitive streak (Iimura and Pourquie 2006). Cells expressing a greater number of 5′ Hox genes ingress at a later stage than those expressing more 3′ Hox genes and would thus be positioned more posteriorly along the AP axis (Fig. 3). This mechanism results in a collinear Hox gene expression profile along the AP axis (Iimura et al. 2007). The expression of particular sets of Hox genes has been suggested to establish each somite identity and thus define later vertebral identities (Carapuco et al. 2005; Deschamps and van Nes 2005; Wellik 2007). The phenotypes that arise from loss- and gain-of-function mutations in mouse Hox genes have revealed that the spatiotemporally controlled expression of these genes is critical for the correct morphogenesis of embryonic axial structures. For example, simultaneous inactivation of all three Hox paralogous group 10 genes results in mice in which the prospective lumbo-sacral region acquires thoraciclike characteristics whereby these mutant vertebrae display associated ribs all along the thoracolumbar region (Wellik and Capecchi 2003).

2.1. The Clock and Wavefront Model (Classical Model)

Because of the requirement that a specific number of somites forms in a given time period, it is reasonable to predict that somite formation is under tight temporal control. This enigmatic system has been the focus of much speculation among biological theorists and is partly why the so-called “clock and wavefront model” was proposed some time ago in the absence of any molecular information (Cooke and Zeeman 1976). This theoretical model encompasses two components: an oscillator (clock) that measures time and determines the period for somite formation (Fig. 4), and spatial information (wavefront), which governs the maturation of the PSM into somites and is defined at a constant distance from the posterior end of the tail bud. The wavefront regresses in an anterior to posterior direction and sweeps along the body axis in concert with the extension of the trunk and tail. The activity of the wavefront is gated by the clock. Hence a somite forms only when the wavefront encounters a group of cells in which the clock is in an “on” state (Fig. 4). In this model also, the size of each somite is determined by the speed of the wavefront regression, whereas the rate of somite formation is controlled by the frequency of the oscillator.

2.2. Molecular Mechanism Underlying Clock Systems

The first evidence of the existence of an oscillator that was coupled to somitogenesis was provided by the periodic expression of the mRNA encoding the transcription factor c-hairy (an ortholog of Hes1) in the chicken PSM (Palmeirim et al. 1997). During the formation of each somite, the PSM is swept by a dynamic wave of c-hairy messenger RNA (mRNA) expression. Because these transcriptional oscillations were observed to have the same periodicity as somitogenesis, this phenomenon was proposed to represent the segmentation clock functioning in PSM cells. Subsequently, several other genes showing such a cyclic pattern were identified in fish, frog, and mouse, indicating that the clock mechanism might be conserved in vertebrates. These genes are now referred to as cyclic genes and the vast majority belong to the Notch, Wnt, and fibroblast growth factor (FGF) signaling pathways (Dequeant et al. 2006). The requirement of the Notch-signaling pathway for clock behavior has been documented by mutant mouse analysis. Many Notch pathway genes including the receptors and ligands, modifiers of Notch receptors and ligands, and their target genes have been shown to be expressed during the critical steps of somitogenesis, and mutations show defects in different aspects of the process (Fig. 5; Table 1).

After the identification of clock genes, many theoretical models were proposed for the operation of cyclic gene oscillations during somitogenesis. Beginning with the work of Lewis (2003), much attention was directed toward models involving the negative regulation of mRNA transcription by protein products, with delays in transcription and translation resulting in oscillations with periods defined, for the most part, by the total delay in the system. Such models have assumed the following form:

where m denotes the mRNA concentration and p the protein concentration, and a, b, and c are positive constants representing protein translation, protein decay, and mRNA decay rates, respectively. Tm and Tp represent the delays in mRNA transcription and protein translation, and the function f represents the negative regulation of mRNA transcription by promoter binding (Lewis 2003). From this basic model of Lewis, many subsequent models were devised to explain several aspects of clock oscillation and to predict mutant phenotypes. Thus far, the Notch-signaling pathway has been shown to be a common mechanism used in several model animals, and a number of proposed models of how this mechanism works in zebrafish and mouse are outlined in the following sections.

2.3. Synchronized Oscillation

The oscillations of clock genes are regulated by the segmentation clock and the underlying molecular mechanism is well established. However, individual PSM cells will behave as noisy autonomous oscillators with a range of different periods if the embryo is dissociated and the gene expression is monitored in individual cells. Even within the embryo, the oscillation phase of individual oscillators must fluctuate owing to developmental noise imposed by stochastic gene expression and active cell proliferation. Hence, there must be a mechanism that synchronizes the oscillation of active genes in the embryo, by which clear waves of gene expression sweep along the AP axis. How this collective behavior is established has been a key question in furthering our understanding of segmentation clocks. It is now known in zebrafish that the segmentation clock behaves as a coupled oscillator, via Notch-dependent intercellular communication (Jiang et al. 2000; Horikawa et al. 2006). Because DeltaC expression is also under the control of the Her1/7 transcriptional repressor and also oscillates, individual cells can communicate with neighboring cells by transmitting Notch signals, which enable a coupling of the oscillations in individual cells. This coupling process plays a crucial role in minimizing the effects of developmental noise to maintain coherent oscillation. In addition, it has been shown that Delta-Notch coupling influences the collective period of the segmentation clock because disruption of this mechanism results in an extended period of zebrafish somitogenesis. It has thus been suggested that collective control of this period via delayed coupling may be a general feature of biological clocks. However, the mechanism by which synchronized oscillation is regulated in the mouse PSM remains to be elucidated.

3.1. Translating Oscillations into Segments in the Wavefront

The mechanism involved in the transcriptional activation of Mesp2 is directly involved in the mechanism by which temporal clock information is translated into spatial patterning. The analyses of factors involved in the dynamic changes to the Mesp2 expression pattern have revealed that the timing of Mesp2 activation depends on Notch activity, whereas the positioning of the expression domain is defined by both Tbx6 and FGF signaling (Yasuhiko et al. 2006; Oginuma et al. 2008). The anterior border of the Mesp2 expression domain accords with the anterior border of Tbx6, where Mesp2 transcription is triggered on Notch activation. However, once Mesp2 is activated and translated into its protein product, it destabilizes Tbx6 protein expression (Fig. 7B). The downstream target genes Ripply1/2 may be responsible for the degradation of Tbx6 protein because knocking out these genes results in the anterior expansion of Tbx6 (Takahashi et al. 2010). Thus, once Mesp2 is expressed, the next anterior border of Tbx6 expression is defined, which becomes the next somite border. Mesp2 is expressed as a stripe in the anterior part of the Tbx6 expression domain of about one somite in length. The question then arises as to how the posterior border of Mesp2 expression is defined. Intriguingly, the posterior limit accords with the limit of the FGF signaling gradient, which shows its highest concentration at the tail bud. In the absence of FGF signaling, the Mesp2 expression domain shifts into the posterior PSM, indicating that Mesp2 expression is suppressed in the posterior PSM via FGF. In addition, Wnt signaling may also work in conjunction with FGF to define the Mesp2 expression domain. In a mouse embryo that shows constitutive Wnt activity (activated β-catenin), the PSM region is extended anteriorly and the oscillation of clock genes such as L-fng persists for longer in the extended PSM, which is accompanied by the additional formation of Mesp2 stripes (Aulehla et al. 2008).

The evidence shows therefore that both FGF and Wnt are factors responsible for defining the maturation wavefront that facilitates a proper segmentation process. However, the molecular identity of the wavefront has long been a controversial issue. Initially, Fgf8 was proposed to encode wavefront activity because experimental manipulation of FGF8 levels caused corresponding shifts in the position of the determination front in cultured chick and zebrafish embryos. However, mouse embryos lacking Fgf8 in the PSM still undergo somitogenesis (Perantoni et al. 2005). Mice homozygous for null mutations in other FGF ligand genes (Fgf3, Fgf5, Fgf15, Fgf17, and Fgf18) also show no early somitogenesis defects, or die before somitogenesis initiates (Fgf4) (Itoh and Ornitz 2008). Wnt has also been proposed to contribute to wavefront activity, because the manipulation of canonical Wnt signaling also causes corresponding shifts in the determination front (Aulehla et al. 2003; Aulehla et al. 2008). However, in Wnt loss-of-function embryos, FGF signaling is also affected. Thus, the observed somitogenesis defects may be primarily owing to the loss of Wnt signaling or to secondary effects on FGF signaling. To better understand the actual role of FGF signaling in somitogenesis, a conditional strategy was used to inactivate both Fgf4 and Fgf8 in the PSM. This resulted in dramatic shifts in the determination front and the premature differentiation of the PSM (Naiche et al. 2011). Moreover, the restoration of Wnt signaling to these mutants did not restore the determination front, demonstrating that these two FGFs constitute the proposed wavefront activity that maintains the PSM in an undifferentiated state.

3.2. The Mechanism of Segmental Border Formation

The segmental border is positioned through the cooperative function of Tbx6 and Mesp2. However, the nature of how the morphological segregation of the somite is regulated was a crucial question that remained. The anterior border of the segmented somite is perfectly matched with the Mesp2 expression domain. Several target genes of Mesp2 have now been identified, including L-fng, EphA4, and Ripply2 (Morimoto et al. 2005; Nakajima et al. 2006; Morimoto et al. 2007). The transcription of L-fng is under the strict control of Notch signaling in the posterior PSM, but subsequently comes under the control of Mesp2 in the anterior PSM (Morimoto et al. 2005). Because Notch activity can be repressed by L-fng function, it was reasonable to predict that L-fng represses Notch activity in the Mesp2-expressing domain and forms a boundary for the next presumptive somite. This idea was challenged, however, by the impact on somite boundary formation by the uniform expression of NICD in the entire PSM and somitic region (Feller et al. 2008). These results indicated that somite boundaries are generated even in the absence of an NICD border, although they are irregularly shaped and of variable size, and the resulting somites are completely caudalized. This result is further supported by the results of rescue experiments using transgenic mice that express L-fng under the control of Hes7 in the absence of endogenous L-fng expression (L-fng-null mouse) in which the NICD boundary was not generated in the presumptive segmental border (Oginuma et al. 2010). Nevertheless, the segmented somites were found to be normally generated, indicating that L-fng under the control of Mesp2 is dispensable for somite border formation. This being the case, the identity of the event that is actually required for segmental border formation downstream from Mesp2 was a key issue to be resolved. One candidate protein is EphA4, a transmembrane protein kinase, implicated in the formation of the repulsive response between cells via its interaction with ephrin expressed in juxtaposed cells (Fig. 8). Although loss-of-function experiments in the mouse have thus far shown no direct evidence for positive involvement of any particular Eph or ephrin molecule in segmental border formation, the ephrin reverse signaling pathway that activates on EphA binding was implicated in this process in chick (Watanabe et al. 2009).

3.3. The Mechanism of Somite Patterning (Rostro–Caudal Polarity)

One of major features of the segmented somite is its compartmentalization into rostral and caudal parts, which is accompanied by differential gene expression patterns and lineage restriction on differentiation. This compartmentalization is required for the segmental patterning of spinal cord development because the somite compartment contributes to the pathway selection of neural crest cells. How this intrasomitic patterning is established during somitogenesis then became an important question. It is now clear that Notch-signaling activity is a required determinant of the caudal identity of the somite and underlies why embryos that ectopically express active Notch in the entire PSM show completely caudalized somite properties (Feller et al. 2008). A similar phenotype is also observed in Mesp2-null embryos (Saga et al. 1997). Based on the fact that Mesp2 expression is finally restricted to the rostral compartment of a somite, it appears that Mesp2 suppresses Notch activity in the rostral compartment (Takahashi et al. 2000). Initially it was thought that L-fng induced by Mesp2 is involved in the suppression of NICD formation in the rostral compartment. However, as mentioned above, the functional significance of L-fng in this regard has been discounted. Recently, the instructive role of Mesp2 in the suppression of Notch signaling was revealed in analyses of a knockin mouse containing a dominant-negative form of RBP-jk, the mediator of canonical Notch signaling in the Mesp2 locus (Sasaki et al. 2011). The resulting phenotype was quite surprising because almost all of the defects observed in the Mesp2-null embryo were completely rescued in the DN-RBP-jk knockin mouse. A subsequent study indicated that Mesp2 suppresses Notch activity by destabilizing Mastermind, one of the core components of the nuclear NICD complex (Fig. 8) (Sasaki et al. 2011).