Somites are transient metameric structures that later give rise to vertebrae, ribs, skeletal muscles, and subcutaneous tissues. Somites are formed by periodic segmentation of the anterior ends of the presomitic mesoderm (PSM; Fig. 1A, asterisk; Saga and Takeda,2001; Dubrulle and Pourquié,2004; Rida et al.,2004). In mouse embryos, this segmentation occurs every 2 hr, suggesting that this process is regulated by a biological clock with the periodicity of 2 hr. The first molecular evidence for this clock, called the segmentation clock, was reported by Palmeirim et al. (1997) in chick embryos, demonstrating that a wave of expression of the basic helix–loop–helix (bHLH) gene chairy1 propagates from the posterior to the anterior region of the PSM and that each wave leads to segmentation of a pair of somites. These dynamic changes in chairy1 expression are not due to cell movement but the result of oscillatory expression in each PSM cell (Palmeirim et al.,1997). Later, it was found that, in mouse embryos, Hes1, Hes5, and Hes7, homologues of chairy1, are expressed in a similar manner in the PSM (Jouve et al.,2000; Bessho et al.,2001a; Dunwoodie et al.,2002). Analysis of knockout mice revealed that Hes7 is the most important among these bHLH genes for somite segmentation. In the absence of Hes7, somites and somite-derived vertebrae and ribs fuse severely (Bessho et al.,2001b). Furthermore, persistent expression of Hes7 in the PSM also leads to somite fusion (Hirata et al.,2004). Thus, dynamic changes in Hes7 expression are essential for periodic segmentation of somites. Whereas Hes7 expression is specific to the PSM, Hes1 is expressed by many cell types, such as neuroblasts. Interestingly, Hes1 expression also oscillates in non-PSM cells, suggesting that this clock, which was originally found in the PSM, is widely distributed and regulates the timing of many biological events (Hirata et al.,2002). In this review, we describe the mechanism of the segmentation clock, mainly focusing on mouse Hes oscillations. We also discuss roles of Hes1 oscillation in non-PSM cells.
Hes7 OSCILLATION IS REGULATED BY NEGATIVE FEEDBACK
Hes7 is a transcriptional repressor and represses its own expression by directly binding to the promoter (Bessho et al.,2001a,2003; Chen et al.,2005). It has been shown that this negative feedback plays an essential role in Hes7 oscillation (Bessho et al.,2003). Hes7 transcription, which can be detected by in situ hybridization with the intron probe, initially occurs in the posterior PSM (phase I, Fig. 1A). The region of Hes7 transcription moves anteriorly (phase II) and reaches the S-1 region of the PSM (phase III). Then, Hes7 transcription ceases when the S-1 region becomes the new S0, the next forming somite, while new transcription starts again in the posterior PSM (returns to phase I). This dynamic expression is repeated every 2 hr in the mouse PSM. Interestingly, Hes7 protein expression and Hes7 transcription oscillate out of phase with each other (Fig. 1B), and domains for Hes7 protein expression and those for Hes7 transcription are mutually exclusive in all three phases (Fig. 1A; Bessho et al.,2003). These results suggest that Hes7 transcription is repressed when Hes7 protein is expressed. In agreement with this suggestion, when degradation of Hes7 protein is inhibited, Hes7 transcription is constitutively repressed, whereas Hes7 transcription occurs constitutively when synthesis of Hes7 protein is inhibited (Bessho et al.,2001b,2003). Thus, Hes7 oscillation is regulated by negative feedback (Fig. 2). In zebrafish, the Hes7 homologues her1 and her7 display oscillatory expression by a negative feedback mechanism and regulate somite segmentation (Holley et al.,2000,2002; Henry et al.,2002; Oates and Ho,2002).
Hes7 and her1/7 oscillations are mathematically simulated by differential equations based on the negative autoregulation model (Lewis,2003; Hirata et al.,2004). This model predicts that the half-life of Hes7 protein is very important for sustained oscillation. The normal half-life of Hes7 protein is approximately 20 min, which is good for sustained oscillation. However, according to this model, if the half-life becomes 30 min, Hes7 oscillation should be damped after several cycles. In agreement with this prediction, mice with Lysine 14 to Arginine mutation in the Hes7 locus, which makes the half-life of Hes7 protein approximately 30 min without changing the transcriptional repressor activity, display damped oscillation after several cycles (Hirata et al.,2004). Thus, the prediction from the simulation model agrees well with the defects observed in embryos.
COUPLED OSCILLATIONS IN NOTCH ACTIVITY AND Hes7 EXPRESSION
Lunatic fringe (Lfng), a glycosyltransferase, is known to regulate Notch activity by glycosylation (Brückner et al.,2000; Moloney et al.,2000). Lfng expression oscillates in phase with Hes7 in the PSM, and this oscillation is also essential for somite segmentation, because both loss of and persistent expression of Lfng lead to somite fusion (Evrard et al.,1998; Zhang and Gridley,1998; Serth et al.,2003). It has been shown that an interface between the on and off regions for Lfng activity leads to formation of somite borders (Sato et al.,2002). In the absence of Hes7, however, Lfng is constitutively expressed in the PSM, whereas Lfng transcription is constitutively repressed, when degradation of Hes7 protein is inhibited (Bessho et al.,2001a,2003). Thus, Hes7 protein periodically represses not only its own but also Lfng transcription, making Hes7 and Lfng oscillations in phase (Fig. 2; Bessho et al.,2003).
Both Hes7 and Lfng expression is induced by Notch signaling (Fig. 2; Bessho et al.,2001a; Cole et al.,2002; Morales et al.,2002). The transmembrane receptor Notch1 and its ligand Delta-like1 (Dll1) are expressed in the PSM, and it is likely that Notch signaling is mutually activated between PSM cells. Upon activation of Notch, the intracellular domain (ICD) of Notch is released from the membrane portion and transferred into the nucleus, where the ICD forms a complex with the transcriptional repressor RBP-J. RBP-J represses Hes7 and Lfng expression by directly binding to their promoters, but the complex of RBP-J and the ICD acts as a transcriptional activator and induces both Hes7 and Lfng expression (Fig. 2). Thus, Dll1-Notch1-Hes7-Lfng is the major pathway that regulates the segmentation clock.
It has been shown that Lfng inhibits Notch activity in the PSM (Fig. 2; Dale et al.,2003; Morimoto et al.,2005). It is, therefore, likely that Lfng oscillation periodically inhibits Notch, leading to oscillation in formation of the ICD (Fig. 2; Morimoto et al.,2005; Huppert et al.,2005), which could influence Hes7 expression. Thus, oscillations in Notch activity and Hes7 expression are coupled by the Hes7 and Lfng negative feedback loops. It is thought that this coupling is important for stable and synchronized oscillations in the somite segmentation clock. However, in mouse embryos, sustained exogenous expression of Lfng affects but does not abolish endogenous Lfng and Hes7 oscillations, suggesting that Notch signaling is still active even when Lfng is constitutively expressed (Serth et al.,2003). Thus, it remains to be determined whether or not Lfng periodically inhibits Notch activity in the mouse PSM. Furthermore, the turnover rate of Lfng-modified Notch and unmodified Notch on the cell surface is not known, and it would be important to clarify these issues to understand the roles and mechanism of the Notch-Lfng feedback loop in the segmentation clock. Interestingly, Lfng expression does not oscillate in zebrafish (Prince et al.,2001), suggesting that the roles of Lfng in the segmentation clock is different between mice and zebrafish.
INDIVIDUAL PSM CELLS HAVE AN UNSTABLE OSCILLATOR
In the PSM, Hes1 is cyclically expressed in phase with Hes7 oscillation. Like Hes7, Hes1 oscillation is regulated by negative feedback (Hirata et al.,2002). Of interest, Hes1 oscillation occurs in many cell types, such as fibroblasts, myoblasts, and neuroblasts, suggesting that Hes1 regulates the timing in many biological systems (Hirata et al.,2002). While Hes1 oscillation continues in a sustained manner in the PSM (Fig. 3A), it is damped in non-PSM cells after three to six cycles by Northern or Western blot analyses, suggesting that the features of Hes1 oscillation are different between the PSM and non-PSM cells. However, real-time imaging analysis revealed that Hes1 oscillation in non-PSM cells continues even after nearly 2 days at the individual cell level, but because the periodicity is variable from cycle to cycle, oscillations easily become out of synchrony between cells (Masamizu et al.,2006). Thus, the damping of Hes1 oscillation in non-PSM cells is not due to damped oscillations in all individual cells but due to desynchronization between cycling cells. Strikingly, the periodicity of Hes1 oscillations in the dissociated individual PSM cells was also found to be variable, suggesting that each PSM cell has an unstable oscillator, like non-PSM cells (Fig. 3B). A similar result was observed for Lfng oscillation in dissociated chick PSM cells (Maroto et al.,2005). It was shown that Notch signaling is involved in synchronization between cycling cells in zebrafish (Jiang et al.,2000; Horikawa et al.,2006). In the absence of Notch signaling, oscillatory expression becomes out of synchrony between cells (Jiang et al.,2000). Thus, the Notch-mediated cell–cell communication may be important for stabilization and synchronization of Hes1 oscillation. These data also suggest that the periodicity is cell-intrinsically variable and susceptible to extrinsic signals.
The precise mechanism of how Notch signaling regulates synchronization of cycling cells is not known. In zebrafish, expression of the Notch ligand deltaC oscillates (Jiang et al.,2000), suggesting that deltaC oscillation is involved in synchronization by activating Notch signaling in neighboring cells. It was recently found that Dll1 expression also oscillates in the mouse PSM (Maruhashi et al.,2005), although the mechanism of Dll1 oscillation is not known. Because Lfng functions in a cell-autonomous manner, while Dll1 acts on neighboring cells, Dll1 oscillation rather than Lfng oscillation could be responsible for synchronization of Hes7 oscillation in the mouse PSM.
DIFFERENT MODES OF OSCILLATOR EXPRESSION IN THE ANTERIOR AND POSTERIOR PSM
Real-time imaging of Hes1 expression revealed its spatiotemporal expression profile in the PSM (Fig. 4A; Masamizu et al.,2006). This profile showed that the wave of propagation of Hes1 oscillation is different between the anterior and posterior PSM: the oscillation is slowed down, and the periodicity becomes longer, as cells move in their relative position from the posterior to the anterior PSM (Fig. 4B). It is likely that each PSM cell experiences approximately five cycles of Hes1 oscillation after getting out of the tail bud and before entering the S0 region at embryonic day (E) 10.5 (Fig. 4B), although this number may be changed at different developmental stages, when the length of the PSM is different. The periodicity of the first three cycles seems to be approximately 2 hr, whereas that of the last two cycles is longer. These data suggest that Hes1 oscillation is differentially regulated between the anterior and the posterior PSM. Because Hes1 and Hes7 expression oscillate in phase, these results suggest that the periodicity of Hes7 oscillation also becomes longer in the anterior PSM. The mechanism for differential regulation of the anterior and posterior waves of oscillation has been analyzed in zebrafish (van Eeden et al.,1998; Holley et al.,2000), and it was shown that different regions of the her1 promoter are involved in this differential regulation (Gajewski et al.,2003). However, the precise mechanism of how her1 oscillation is differentially regulated in the anterior and posterior PSM remains to be determined.
In the PSM, the Fgf8 level is high in the posterior and low in the anterior PSM, thus forming a gradient (Dubrulle et al.,2001). It is possible that this gradient is involved in different periodicities of Hes1 oscillation between the anterior and posterior PSM. In zebrafish, it was found that her13.2 expression is regulated by Fgf signaling, forming an Fgf8-like gradient (Kawamura et al.,2005). Her13.2 forms a heterodimer with Her1, and this heterodimer has a stronger activity for negative autoregulation than the Her1 homodimer. It is possible that the ratio of Her1-Her13.2 heterodimer to Her1 homodimer is different between the anterior and the posterior PSM and may affect the periodicity of her oscillations. However, such functions are not known for the mouse counterpart of Her13.2, Hes6, which inhibits Hes1 activity by forming a heterodimer (Bae et al.,2000). Thus, the mechanisms for the different periodicity between the anterior and the posterior PSM and the link between the Hes oscillations and the Fgf gradient remain to be determined.
ROLES OF Hes1 OSCILLATION IN NON-PSM CELLS
Hes1 oscillation is observed in many cell types, such as neuroblasts, although the significance of Hes1 oscillation in non-PSM cells is not well understood. In the developing nervous system, there seem to be two different modes of Hes1 expression (Baek et al.,2006; Kageyama et al.,2007). The neural tube is partitioned into many compartments by boundaries. In compartments, embryonic neural stem cells (called radial glial cells) undergo asymmetric cell division, producing another radial glial cell and a neuron or a neuronal precursor. In contrast, boundaries are formed by specialized radial glial cells, which do not give rise to neurons or proliferate extensively. These boundary cells function as the organizing center by secreting morphogens and thereby endowing the neighboring compartmental cells with the regional specificity. Hes1 is persistently expressed at high levels by many boundary cells (Hirata et al.,2001; Baek et al.,2006), while Hes1 expression oscillates in compartment cells (our unpublished data). Although Hes1 is essential for maintenance of radial glial cells in compartments (Ishibashi et al.,1995; Hatakeyama et al.,2004), it was found that forced expression of Hes1 in compartmental cells leads to reduction of cell proliferation and inhibition of cell differentiation, which are reminiscent of two important features of boundary cells. Persistent and high levels of Hes1 not only repress proneural bHLH gene expression constitutively but also retard G1 phase progression (Baek et al.,2006). Thus, Hes1 oscillation may be important for both the normal timing of neuronal differentiation and efficient proliferation of compartmental cells.
Individual PSM cells have an unstable Hes1 oscillator; however, as a tissue, PSM has a stable Hes1 oscillator, suggesting that cell–cell communication is required for stable and sustained oscillation. It is suggested that Notch signaling is involved in both synchronization and stabilization of Hes oscillations. However, while the same genes are expressed in the developing nervous system, such synchronized oscillations are not observed in neuroblasts. Thus, the exact mechanism for synchronized and stabilized oscillations remains to be determined. Furthermore, in the PSM, there is a phase delay of oscillatory expression in the posterior, compared with the anterior PSM and the periodicity becomes longer or the oscillation becomes slower, as cells move from the posterior to the anterior PSM. It was found that, in addition to Notch signaling, downstream factors of Wnt signaling and Fgf8 signaling are also cyclically expressed in the PSM (Aulehla et al.,2003; Dequéant et al.,2006), suggesting that Wnt and Fgf8 oscillators could be involved in the phase delay in Hes oscillation. Regarding the periodicity, the difference among species is another important issue: the period of mouse Hes1 and Hes7 oscillations is approximately 2 hr, whereas that of zebrafish her1 and her7 is approximately 30 min. It is currently unknown what is responsible for the difference in periods of these oscillations.
The biological significance of Hes1 oscillation in non-PSM cells also remains to be analyzed. Hes1 oscillation seems to be required for neuronal differentiation and efficient cell proliferation, but the exact mechanisms of how Hes1 oscillation regulates the timing of cell differentiation and the process of cell cycle remain to be analyzed. It was recently found that more factors show oscillatory response to stimuli in non-PSM cells (Lahav et al.,2004; Nelson et al.,2004), suggesting that oscillatory expression could be a general feature for many cellular events. Understanding of these issues will reveal more precise dynamics of gene networks in many cellular events, including somite segmentation.
We thank the anonymous reviewers for critical comments on our manuscript. Research in our laboratory was supported by the Genome Network Project; the Grants-in-Aid from the Ministry of Education, Culture, Sports, Science and Technology of Japan; and the Uehara Memorial Foundation. Y.M. was supported by Research Fellowships from the Japan Society for the Promotion of Science for Young Scientists. Y.N. was supported by the 21st Century COE Program of the Ministry of Education, Culture, Sports, Science and Technology of Japan and the Iwatare Scholarship.