Somites are transient, segmental structures in vertebrate and cephalochordate embryos, derived from paraxial mesoderm (Saga and Takeda, 2001). They are formed on both sides by budding off anteriorly, at regular intervals (90 min in the chick, 2 hours in the mouse), from the presomitic mesoderm (PSM). Mesoderm segmentation and the polarity of the resulting somites play an essential role in the patterning of other structures, such as nerves, vertebrae, muscles and blood vessels (Saga and Takeda, 2001, Keynes and Stern, 1988).
A molecular "segmentation clock", or "somitogenesis clock", was recently identified by Palmeirim et al. (1997), and involves oscillations in both mRNA and protein levels of c-hairy1, a chick homologue of a gene first identified in Drosophila. c-hairy1 expression is not synchronous throughout the PSM: a wave, originating from a large region of posterior PSM, spreads anteriorly while shrinking, and stabilises at the anterior border of the PSM (see below for a more detailed description). A new somite is formed every time a wave reaches the border.
Subsequently, other cycling genes were identified: c-hairy2 (chick) / HES1 (mammals) cycles in both chick and mouse PSM (Jouve et al., 2000), and Lunatic fringe (L-fng), an important regulator of the Notch pathway (Blair, 2000), involved in boundary-formation in insect development, cycles in the chick (McGrew et al., 1998) and mouse (Forsberg et al., 1998) (but not zebrafish, Prince et al., 2001). A Notch ligand, DeltaC, was also identified as cycling in zebrafish PSM (Jiang et al., 2000). The period of oscillation is 90 minutes for the chick, 2 hours for the mouse, and 20-30 minutes for the zebrafish (Stickney et al., 2000). In the chick, PSM cells experience about 2 cell-cycles (Primmett et al., 1989) and 12 clock-pulses before being incorporated into a somite (Palmeirim et al., 1997).
The oscillatory expression pattern has been classified with 3 stages; in the first stage, cells in the caudal half of the PSM (youngest cells) express the genes in synchrony (or at least with much smaller phase-lags than in the rest of the PSM). In the second stage, the expression pattern moves rostrally and narrows; this narrowing was supposed by Palmeirim et al. (1997) and others (Kaern et al., 2000, Jaeger and Goodwin, 2001,2002) to stem from a progressive increase of the clock period, but it has also been proposed that it results from shorter synthesis bursts (Cooke, 1998). In the third stage, expression becomes stabilised in one half of a prospective somite, and this somite forms shortly thereafter. Cells undergo a relative movement at constant speed in the PSM, and they oscillate about 8 times while moving from the posterior end to the middle of the PSM. When cells reach the middle of the PSM, the intensity of the oscillations in c-hairy1 increases sharply (O. Pourquié, personal communication). For other genes such as c-hairy2, the intensity of oscillations is on the contrary down-regulated. L-fng and c-hairy1 expression patterns are synchronous in most of the PSM, and diverge when they reach the boundary of the forming somite (McGrew et al., 1998); a stripe of L-fng expression stabilises in the anterior part of the forming somite in the chick, but not in the mouse.
Normal functioning of the clock seems to be essential for segmentation, as its disruption by mutations affecting the Notch pathway (Barrantes et al., 1999, Jouve et al., 2000, Jiang et al., 2000, Dunwoodie et al., 2002), or enforcement of a non-zero baseline of l-fng expression (Dale et al., 2003, Serth et al., 2003), result in severe segmentation defects. Different models provide a link between clock oscillations and actual somite segmentation, but our purpose here is to address the mechanism for the oscillations, rather than the way they are read out. Somitogenesis is a process which shows great autonomy; for example, PSM explanted from an embryo still undergoes partial segmentation (even though the process requires ectoderm to go all the way, the segmental pattern is observable, Packard et al., 1993; Lash and Ostrovsky, 1986). Based on one type of experiment, the segmentation clock is generally assumed to be cell-autonomous in chick and mice (see section 3.2). However, there is also evidence that there could be some intercellular coupling in segmentation-clock oscillations (see section 2.1), as first suggested by Jiang et al., 2000 in the case of the zebrafish. The two aspects seem difficult to reconcile. In the following, we propose however a model for segmentation-clock oscillations which allows for coupling, but can also behave as if oscillations were cell-autonomous, providing a new possible explanation for experiments previously interpreted as ruling out intercellular coupling. The model could easily be extended to account for the very first segmentation clock oscillations in the primitive streak, which occur at a much earlier stage of embryonic development than PSM segmentation (Jouve et al., 2002), and which remain unexplained by current models. Oscillations in the proposed model partially rely on a positive feedback mechanism, but the model is compatible with experimental data which has been interpreted as supporting a negative-feedback mechanism, and is also compatible with experimental data which contradicts a purely negative-feedback model.