The somite segmentation clock: It takes a licking and keeps on ticking



This primer describes molecular mechanisms critical for somite segmentation clock function. In addition, two investigators of the segmentation clock, Olivier Pourquié, and Yun-Jin Jiang, give their perspectives on current research and on the future of the field. Developmental Dynamics 232:519–523, 2005. © 2004 Wiley-Liss, Inc.


Somites are transient, segmentally organized blocks of mesoderm that lie bilateral to the midline axis in the trunk and tail of vertebrate embryos. In chick embryos, somites bud once every 90 min from presomitic (paraxial) mesoderm (PSM), bands of unsegmented tissue posterior to somites, until approximately 50 somite pairs are formed. Somitic cells will give rise to multiple tissues, including ribs and vertebrae, skeletal muscle, and dermis of the back. The positioning of somites at regular intervals is important not only for patterning paraxial mesoderm, but also for migration paths of neural crest cells and spinal nerve axons (Christ and Ordahl, 1995). Therefore, the periodicity of somite formation is critical for basic patterning and development of the vertebrate embryo.

The somite segmentation clock, and a complex network of molecular pathways that function downstream of it, drive somite budding at regular intervals. The clock controls cyclic gene expression that moves wavelike from the posterior to anterior PSM, such that PSM cells at a given axial level turn cyclic gene expression “on” and “off” as each somite buds (Fig. 1A; Palmeirim et al., 1997). When cyclic gene expression crosses a specific position in the PSM, called the maturation wavefront (Fig. 1B; Dubrulle et al., 2001; Sawada et al., 2001), PSM cells anterior to the wavefront begin to mature, and shortly thereafter, genes that pattern somite polarity and boundary formation are expressed (Fig. 1C; Sawada et al., 2000; Takahashi et al., 2000). The wavefront moves posterior during elongation of the embryonic axis but remains at a fixed point relative to the position of the most recently formed somite, thus ensuring that somites are a consistent size. Consequently, the clock controls cyclic gene expression, but it is dependent on the position of the wavefront and maintenance of cyclic gene expression for periodic somite formation.

Figure 1.

A: Somites bud from the anterior presomitic mesoderm (PSM) such that one pair buds every 90 min in chick embryos. Notch-related cyclic genes (red) are expressed in a posterior to anterior wave of expression coordinated with the budding of each somite pair. Cyclic gene expression is dependent on the somite segmentation clock. B: The maturation wavefront (arrow) is positioned by opposing gradients of fibroblast growth factor-8 (yellow) from the posterior PSM and retinoic acid (RA, blue) from the anterior. The wavefront travels posteriorly during axis elongation but remains at a fixed position relative to the most recently formed somite. C: When cyclic gene expression crosses the wavefront, PSM cells begin to mature and shortly thereafter, Mesp2/thylacine 1(green), one of the earliest indications of segmental organization, is expressed in the anterior compartment of the future somite. Arrows in each panel point to the approximate location of the wavefront. Anterior is up.

The primer below describes molecular mechanisms critical for somite segmentation clock function. This includes mechanisms that synchronize cyclic gene expression in adjacent PSM cells and also between the PSM and the clock; molecular feedback loops that generate and maintain the pace of cyclic gene expression; the molecular basis of the wavefront; and mechanisms behind an unexpected role for the clock: regulation of axial specification. Finally, two investigators of the somite segmentation clock, Olivier Pourquié and Yun-Jin Jiang, give their perspectives on current research and the future of the field.


Notch-Related Genes Coordinate Cyclic Gene Expression

Notch-related genes were the first identified molecular components of the segmentation clock. Their critical role in this biological process is supportd by many lines of evidence. First, several components of the Notch pathway exhibit cycling gene expression in the PSM (Fig. 1A). These include Lfng (lunatic fringe), a glycosyltransferase that directly inhibits the Notch receptor, and Notch pathway transcriptional targets from the Hairy/E(spl) (Hairy Enhancer of Split) family (Bessho et al., 2001; Dale et al., 2003). Furthermore, mice mutant for any one of multiple components of the Notch pathway, including Delta-like Notch ligands, Lfng, Hairy/E(spl) transcription factors, and an intracellular Notch effector, RBP Jk, have disorganized somite boundaries and vertebral and rib malformations (Conlon et al., 1995; Hrabe de Angelis et al., 1997; Zhang and Gridley, 1998; Leimeister et al., 2000; Bessho et al., 2001). Finally, Notch-related cycling gene expression is abolished in some of these mutants (Holley et al., 2000; Jouve et al., 2000; Dale et al., 2003; Serth et al., 2003). These experiments demonstrate that Notch is important for segmentation clock function.

One way to explain these findings is that Notch is an oscillator that synchronizes cyclic gene expression in adjacent cells. This model would predict that, if the oscillator fails, the molecular clock and cyclic gene expression would slowly drift out of phase. This model agrees with the finding that, in Notch pathway mutants, anterior somites develop normally but posterior somites are adversely affected (Conlon et al., 1995). In addition, somites in zebrafish Notch pathway mutants have an unorganized mixture of cells expressing high and low levels of cyclic genes, as if cell–cell synchronization of gene expression were lost (Jiang et al., 2000). These findings suggest that Notch is not an essential component of the clock, but rather coordinates cyclic gene expression locally among PSM cells in addition to synchronizing cycling gene expression to the molecular clock.

Negative Feedback Loops Maintain and Set the Tempo of Cyclic Gene Expression

Maintenance of cycling gene expression is regulated by negative feedback loops within the notch pathway. Whereas the exact composition of such negative feedback loops are not known, they seem to be an integral part of the Notch pathway (Giudicelli and Lewis, 2004). For example, Notch signaling activates transcription of Lfng, and subsequently translated Lfng protein inhibits Notch activity. Reduced Notch signaling results in decreased production of Lfng, thereby effectively relieving repression of Notch, thus allowing the cycle to start again (Dale et al., 2003). Similarly zebrafish Her1 and mouse Hes7 (orthologs of Hairy Enhancer of Split genes) transcription are each inhibited by their own protein product (Holley et al., 2002; Bessho et al., 2003). Importantly, when the half-life of Hes7 protein is increased, oscillating gene expression and normal somite segmentation is disrupted (Hirata et al., 2004). Hence, whereas negative feedback loops maintain cycling gene expression, the endogenous rate of transcription and translation sets the pace.


Recently, it was found that Wnt signaling is another important player in the somite segmentation clock. A negative regulator of Wnt signaling, Axin2, displays cyclic gene expression in the PSM that is out of phase with Notch-related cycling genes (Aulehla et al., 2003). Surprisingly, whereas Axin2 expression is abolished in Wnt3a-deficient mice, its oscillation is unaffected in mice null for the Notch ligand, Dll1. This finding suggests that Axin2 is regulated by Wnt3a signaling but not by Notch signaling. Furthermore, oscillation of the Notch modifier Lfng is abolished in Wnt3a-deficient mice, demonstrating that Wnt3a is upstream of all identified components of the clock and, therefore, is a candidate initiator of the segmentation clock.


The Maturation Wavefront Positions Somite Boundaries

A fundamental problem facing the clock is, how can cyclic gene expression activate spatially restricted gene expression patterns in somites? The maturation wavefront coordinates periodic boundary formation, although how the wavefront prompts expression of boundary genes is just beginning to be understood. The maturation wavefront is positioned by opposing gradients of fibroblast growth factor-8 (FGF8) activity from the posterior PSM and retinoic acid (RA) activity from the anterior (Dubrulle et al., 2001; Diez del Corral et al., 2003; Fig. 1B). RA signaling helps position the wavefront by inducing expression of an FGF inhibitor that counteracts FGF8 activity (Moreno and Kintner, 2004). As the wavefront moves posterior, concomitant with body axis elongation, PSM cells anterior to the wavefront escape a relatively high threshold of FGF8 signaling and subsequently initiates somite maturation. At this stage, they are competent to express genes, such as mesp2/thylacine1, that compartmentalize somites and position somite boundaries. mesp2/thylacine1 expression is localized to the anterior compartment of the future somite and is regulated by Notch signaling (Jen et al., 1999; Sawada et al., 2000; Takahashi et al., 2000; Fig. 1C). Thus, the timing of somite maturation and boundary formation correlates with the periodic passage of Notch-related cyclic gene expression as it passes anteriorly through the wavefront.

The Clock Regulates Hox Expression and Specification of Axial Position

In addition to regulating the spatial patterning of somite boundary genes, the segmentation clock also influences axial specification by means of Hox expression. Hox proteins are a family of homeobox transcription factors expressed in nested anterior/posterior patterns along the axis of many tissues, and accordingly are required for their axial specification. When the maturation wavefront is experimentally shifted anteriorly by placing an FGF8-coated bead in the mid-PSM, somite boundaries form prematurely, thus resulting in smaller somites. Surprisingly, Hox are expressed in the (abnormally small) correct numbered somite, but its overall expression is shifted anterior relative to the embryonic axis. This finding demonstrates that the segmentation clock fine tunes the position of PSM Hox expression.

Surprisingly, Hoxd1 and Hoxd3 cycles in the mouse PSM, and PSM expression is dependent on the Notch effector, RBP Jk (Zakany et al., 2001). This is substantiated by the finding that mice bearing a null mutation in Lfng, or dominant negative Dll1, display both shifts in somite Hox expression and phenotypes consistent with homeotic shifts in vertebrate identity (Cordes et al., 2004). By regulating Hox expression, the segmentation clock ensures that each somite acquires the correct identity appropriate for its final axial location.


What is the future of somite segmentation clock research? Below, Developmental Dynamics discusses current research and the future of the somite segmentation clock field with two experts (Fig. 2): Olivier Pourquié, Associate Investigator, Stowers Institute, Kansas City, Missouri, and Yun-Jin Jiang, Senior Principal Investigator, Institute of Molecular and Cell Biology, Singapore. The complete version of this discussion can be viewed at

Figure 2.

Two experts in the somite segmentation clock field.

Developmental Dynamics: What is the molecular readout of the segmentation clock? Do you think anterior/posterior (AP) patterning (i.e., by means of Hox genes) and/or somite boundary formation are directly regulated by Notch signaling?

Yun-Jin Jiang: I think both are directly regulated by Notch signaling, as shown by Duboule's and Saga's labs in mouse and chick systems, respectively (Takahashi et al., 2000; Zakany et al., 2001). However, Notch signaling is unlikely to be [acting] alone; instead, it works in cooperation with other genes. For example, mesp in somite AP polarity. The readout of the clock is not so clear at this moment. Regulation of Hox [gene expression] could be one readout of the clock.

Olivier Pourquié: The first molecular readout of the segmentation clock is the periodic expression of cyclic gene mRNAs. The clock probably ultimately controls somite boundary positioning, but there is no real compelling evidence for this thus far. One would need to change its periodicity and show that boundary position is affected accordingly. I believe the clock only controls the fine tuning of AP patterning by means of Hox genes, i.e., not the establishment of the colinear expression but rather the precise positioning of Hox boundaries at defined somitic levels. Both the definitive positioning of Hox anterior boundaries and of somitic boundaries depend on Notch signaling.

Dev. Dyn.: Evidence from Aulehla et al. (2003), indicate that Wnt3a is upstream of Notch and may function as the initiator of the segmentation clock. Yet Wnt3a null mice still form anterior somites (Takada et al., 1994), suggesting that Wnt3a is not required for early somitogenesis. Is there a way to explain these apparently conflicting data?

Y-J. J.: The most straightforward explanation will be redundancy.

O.P.: The role of Wnt3a in the system is still unclear. The fact is that anterior somite formation could be attributed to redundancy, as several other Wnts are expressed in a pattern similar to Wnt3a in the tail bud.

Dev. Dyn.: Because anterior somites that form during gastrulation develop normally but those that develop during body axis elongation do not, one possibility is that there is both a “gastrulation clock” and an “elongation clock.”

Y-J.J.: Of course it is possible. But I think it is just an idea and not substantiated by experimental data.

O.P.: The dynamics of the signaling during axis extension have been poorly studied, and it is clear that the most anterior somites (the occipital which are included in the skull) are quite different from the others. While they appear to be produced by a similar mechanism, regional differences might be seen along the AP axis.

There are two different oscillatory behaviors depending on whether the cells are in the streak/tail bud or in the PSM. What is the difference between anterior and posterior somites that explains this different sensitivity to mutations is still unclear.

Y-J.J.: There are some observations, e.g., gene expression patterns and zebrafish mutants, to support the anterior and posterior differences. But it is not clear yet whether it has anything to do with the ideas of “gastrulation/elongation clocks.”

Dev. Dyn.: Does the role of Notch in the segmentation clock fit with its classic role in lateral inhibition (Rooke and Xu, 1998)?

O.P.: Aspects of Notch signaling in the clock are similar to lateral inhibition. However, there is probably some specificity in the PSM. For instance, Saga provided evidence for an unconventional presenilin-independent Notch activity in Mesp2 activation (Takahashi et al., 2000). Data from the Kopan lab on a Notch molecule mutated at its cleavage site also suggest that the pathway might work differently in the nervous system (Selkoe and Kopan, 2003).

Y-J.J.: In a macroscopic scale, it may not be the case, because the stripe expression of the clock genes in the posterior PSM is relatively homogenous within the stripes. Furthermore, it has been shown in the chick that in the anterior PSM, Notch can induce boundary formation. For example, the average amplitude of the stripe may be set by another signal (Wnt, FGF or others) or even Notch signaling itself but through different subcircuits. And the level in each individual cell in that stripe could still be controlled by local lateral inhibition. If we have a real-time reporter transgenic line, this question may be answered in great depth.

O.P.: I am not convinced that the segmentation clock is a simple Notch or Notch and Wnt based oscillator. I think there might be a distinct pacemaker still to be found. That more and more cyclic genes belonging to different pathways argues in favor of more complexity, and I believe that our view of this mechanism might change quite drastically in the next few years. An interesting challenge will be to see how this mechanism has been conserved in evolution. Recent data suggest that spiders might also use dynamic Notch expression to make their segments (Stollewerk et al., 2003), and it will be interesting to see whether what we see in vertebrates in fact better illustrates the ancestral segmentation mechanism than the fly segmentation cascade.

Y-J.J.: I agree with Olivier. A crazy idea about this pacemaker or universal timekeeper is that the temporal periodicity is created by a “different gear.” The embryos use different molecular linkers to achieve the periodicity they need. If this is true, Notch signaling could be the linker to the yet unknown “universal clock.” Other periodic processes may use different linkers.

Dev. Dyn.: What are some important questions that remain to be answered?

O.P.: An important future question is evolutionary conservation. Also I think that much remains to be learned about the clock mechanism. Recent experiments from Kageyama's lab showing oscillations of Hes1 in cultured cells (Hirata et al., 2002) suggest that this oscillatory mechanism might in fact be more widespread than currently thought. It would not be visible elsewhere due to the lack of synchronization. This is certainly an interesting question. Other exciting questions relating to the control of the body plan size, segment number, and interaction with Hox gene expression will be also of great interest.

Y-J.J.: The evolutionary comparison of different species will shed light on our understanding about the clock. For example, define what are core components and what are peripheral components. The other interesting field will be the involvement of physical and mathematical sciences, helping us to better understand this complex process.