In general, the function of the canonical Notch pathway is to coordinate gene expression in contiguous cells. It does this in a particularly direct way. The signal-sending cell expresses a Notch ligand (belonging to either the Delta or the Serrate/Jagged subfamily) on its surface; this binds to the receptor, Notch, in the membrane of the signal-receiving cell and thereby triggers cleavage of Notch, releasing an intracellular fragment, the Notch intracellular domain (NICD); NICD translocates to the nucleus, where it acts as a transcriptional regulator 17 (Figure 1). The main – or at least, the best-studied – targets of direct regulation by NICD are the members of the Hairy/E(spl) family (Hes genes in mammals, her genes in zebrafish) 89; these code for inhibitory basic helix-loop-helix (bHLH) transcriptional regulators, which can control many different secondary targets, including Notch ligand genes and the Hes/her genes themselves. The role of Notch signaling in pattern formation depends on the ways in which these components – and others that modulate their activity – are functionally connected into regulatory feedback loops 10. Mathematical modeling highlights several possibilities. Thus, one type of linkage, where Notch activation leads to down-regulation of Notch ligand expression in the signal-receiving cell, can lead to lateral inhibition, forcing neighboring cells to become different from one another 11 (Figure 2). An opposite linkage, whereby Notch activation stimulates ligand expression, can have an opposite effect, inducing contiguous cells to be similar 12. Still other types of circuitry built from the same components can perform yet other tricks, including the production of temporal oscillations of gene expression 1314. And this brings us back to somitogenesis, where such oscillations are in fact seen.

Somites derive from the unsegmented presomitic mesoderm (PSM) at the tail end of the embryo. PSM cells are specified by the combined action of Wnt and fibroblast growth factor (FGF) signaling molecules, which are produced at the tail end of the PSM and spread anteriorly to generate a morphogen gradient. At the point where the level of Wnt and FGF falls below a threshold value, somites form. Thus, as the PSM grows caudally, extending the embryo, one pair of somites after another is budded off from the anterior end of the PSM in a regular head-to-tail sequence. Each species generates its characteristic number of somites at its own pace, ranging from one new somite pair approximately every 30 minutes in zebrafish to one pair every 2 hours in mice. This rhythmic process involves coordinated patterns of cell behavior not only in space but also in time: it depends on an underlying gene expression oscillator – the segmentation clock – that ticks in the cells of the PSM and dictates the rhythm of somite formation, with each oscillator cycle corresponding to the production of one additional somite 15. The genes that were first found to oscillate in the PSM and that show this cyclic expression in all vertebrates belong to the Notch signaling pathway; these oscillatory genes include, specifically, certain members of the Hairy/E(spl) gene family of bHLH transcriptional regulators – in particular Hes1 and Hes7 in mice, her1 and her7 in zebrafish, and hairy1 and hairy2 in chick 1516171819202122 – and (in zebrafish) the Notch ligand DeltaC, whose expression is controlled by them. These, and certain other oscillatory genes, display a characteristic pattern of expression that can be seen in fixed specimens stained by in situ hybridization. In the posterior part of the PSM, the level of expression may be high or low, depending on the phase of the oscillation cycle at the moment when the embryo was fixed. In the anterior part of the PSM, meanwhile, one sees a stripy pattern, in which bands of cells that express the oscillatory gene strongly alternate with bands of cells that do not (Figure 3). This pattern reflects the gradual slowing of the oscillations as cells approach the point of exit from the PSM, beyond which oscillation is halted: cells in more anterior positions are thus delayed in phase relative to more posterior cells, with the consequence that one sees laid out along the antero-posterior axis of the PSM an ordered array of cells in different phases of the oscillator cycle 1523. Disturbances of oscillator behavior are thus clearly displayed in a disturbed spatial pattern of gene expression in the anterior PSM – a great convenience for experimental analysis.

Since, as we noted earlier, any mutation that blocks Notch signaling leads to disrupted somite segmentation, an obvious suggestion is that the oscillation depends on Notch signaling and fails to occur when Notch signaling fails. However, the detailed consequences of mutations in the Notch pathway do not quite fit this simple explanation. A different interpretation is instead suggested by a closer examination of the behavior of one of the oscillatory genes, coding for the Notch ligand DeltaC, in zebrafish with mutations in the Notch pathway 24. The individual PSM cells in these mutants still express DeltaC, but in an uncoordinated way: tissue fixed for analysis by in situ hybridization shows a pepper-and-salt mixture of cells expressing DeltaC at different levels, as though the cells are still oscillating individually, but no longer in synchrony with their neighbors (Figure 4). Moreover, both in zebrafish and in mice, the first few somites of embryos with Notch pathway mutations develop almost normally 252627, implying that Notch signaling is not absolutely necessary for somite segmentation and that the consequences of failure of Notch signaling make themselves felt only gradually, after the onset of somitogenesis. These findings led to the suggestion that the primary function of Notch signaling is not to drive the oscillations of individual cells, but only to coordinate them and keep them synchronized; and that the cells begin oscillation in synchrony at the start of somitogenesis, and take several cycles to drift out of synchrony when Notch signaling is defective 24. This proposal – that Notch signaling from cell to cell in the PSM serves to maintain synchrony but is not necessary for oscillation of individual cells – has been supported by several subsequent experiments. For example, zebrafish embryos can be treated at different stages of somitogenesis with the inhibitor DAPT, which inhibits the enzyme that releases NICD from the membrane (Figure 1) and thus blocks Notch signal transmission. When Notch signaling is prevented in this way, somite defects ensue, but always with a delay that corresponds to a gradual disordering of the pattern of oscillator gene expression 2829. Other evidence comes from experiments where PSM cells are transplanted into a wild-type zebrafish embryo from an embryo in which the expression of the oscillatory her genes is defective. The transplanted cells then cause abnormal segmentation behavior in their neighbors; but they fail to exert this effect if they are prevented from expressing the Notch ligand DeltaC 30. The oscillatory behavior of individual PSM cells and the influence of Notch signaling can also be demonstrated through study of cells from the PSM of a transgenic mouse embryo containing a luminescent Hes1 reporter. These cells show oscillating expression of the reporter gene even when they are dissociated and thus unable to communicate via Notch 31, but in that condition the oscillations are much less regular than in the intact tissue.

All these findings support the view that Notch is needed to maintain synchrony between the oscillations of the individual cells, which are somewhat noisy and imperfect timekeepers when left to their own devices. But what is generating the cell-intrinsic oscillations? According to one view, the core oscillator – the pacemaker of the whole process – is a delayed negative feedback loop in the auto-regulation of the oscillatory Hes/her genes – Hes7 in mammals, her1 and her7 in the zebrafish 1332 (Figure 5). Loss of Hes7 in the mouse, or of her1 and her7 in the zebrafish, disrupts segmentation all along the body axis; and it has been shown experimentally that these genes are indeed subject to negative regulation by their own products 22233233. The idea that this Hes/her negative feedback loop is the core oscillator has been articulated in quantitative mathematical terms and is supported by many pieces of evidence, but it still lacks firm proof 34. In mouse and chick, the PSM cells also show oscillating expression of various other genes, including (in the mouse) genes in the Wnt and Fgf pathways 353637, some of which appear to continue their oscillation even when the Hes7 oscillations fail 35. Thus, the nature of the ultimate generator and pacemaker of the oscillations is still under debate, especially for mouse and chick 383940.

Failure of synchronization is sufficient to explain the disruption of segmentation in Notch pathway mutants. But that is not necessarily the end of the story. To acknowledge that Notch signaling has this critical function, and that that is enough to explain the mutant phenotypes, is not the same as saying that synchronization is the only function of whose period is determined by the total delay in the feedback loop.

Notch signaling in somitogenesis. At least two additional functions have been proposed. One is in the final step at which a segment boundary is created by physical separation of one nascent somite from the next; the other is in creating or maintaining the difference between anterior and posterior halves of each somite. Each of these possible further roles for Notch signaling – in boundary formation and in segment polarity – seems attractive on the basis of analogies with other systems. Thus, in the Drosophila wing disc, Notch signaling plays a critical part in organizing the dorso-ventral compartment boundary 41; and in the vertebrate hindbrain, likewise, it is involved in organizing the boundaries between rhombomeres 42. As for segment polarity, the creation of a difference between the cells of the anterior and posterior parts of each somite could be seen as similar to the creation of differences between adjacent cells through lateral inhibition – a well known function of Notch signaling in many different systems 1.

It is in the anterior part of the PSM, where the oscillation of cyclic genes slows down and then halts, that cells are assigned to anterior or posterior somite compartments and clefts form, finally demarcating one somite from the next. Thus, the formation of the segment boundary and the specification of antero-posterior polarity are both processes that occur relatively late in the history of each somite, after its precursor cells have graduated to the anterior part of the PSM from the posterior as the embryo grows and extends. If the early function of Notch signaling in maintaining synchrony in the posterior PSM is disrupted, any failure in these later functions is likely to be imperceptible amid the general chaos. One can, however, test for the later functions by imposing a block of Notch signaling part way through somitogenesis. For example, one can take a zebrafish that has already formed five somites and immerse it in a DAPT solution to block Notch signaling from that time point onwards. The result is striking: the next approximately 12 somites proceed to form in the normal way, with regularly spaced boundaries, and only after that does one begin to see segmentation defects 2829. This shows that Notch signaling is not needed, in the zebrafish at least, for the creation of somite boundaries, and it quantitatively matches predictions based on the proposition that the only function of Notch signaling is to maintain synchrony in the posterior PSM 29.

Findings in the mouse, however, are not so clear, and there are differing schools of thought. In a series of papers 43444546474849, Saga and colleagues have argued that Notch signaling is indeed needed to create a sharp boundary of gene expression that is necessary to mark the future cleft between one nascent somite and the next 4344. Their conclusions emerge from study of a pair of transcriptional regulators – Mesp2, and the less well characterized Mesp1 – that are expressed in the anterior PSM. They seem to operate as orchestrators of the process by which the output of the somite oscillator is translated into the spatially repeating pattern of the somites 45 – a process that is disrupted in Mesp2 mutants 46. Mesp2 is expressed dynamically in each forming somite, beginning as a one-somite-wide stripe, rapidly narrowing to a half-somite-wide stripe (which marks the future anterior compartment of the somite), then disappearing completely as the somite buds off from the PSM. In the brief window during which it is expressed, Mesp2 seems to be responsible for allocating anterior or posterior identity to the cells of the somite through activation or repression of various targets that distinguish the anterior from the posterior cells, and for regulating some of the genes required for border formation 4748. In particular, somite boundaries form at interfaces where cells with high expression of Mesp2 but low Notch activation confront cells in an opposite state, with high Notch activation but no expression of Mesp2. These observations strongly suggest that some sort of feedback loop involving Mesp2 and Notch signaling organizes the formation of an interface between cells with high Notch activation and cells with low Notch activation, and that this interface is necessary to define the segment boundary. Moreover, the same studies suggest that Notch signaling is involved in the restriction of the Mesp2 expression domain from the whole presumptive somite to just its anterior half 4849, and thus essential for the establishment of the anterior-posterior polarity of each new somite. However, these observations do not amount to firm proof: correlation need not imply causation, and Mesp2, acting independently of Notch activity, could be the critical factor. The pattern of Mesp2 expression is indeed altered in Notch pathway mutants 43, but it is hard to be sure whether this reflects a function of Notch signaling in the anterior PSM where Mesp2 is expressed, or merely the aftermath of the disorder created by prior failure of Notch signaling in the posterior PSM.

Feller et al. 50 tested the role of Notch signaling in the mouse PSM in a different way and came to a somewhat different view. When they artificially expressed NICD, the intracellular transcriptional regulator domain of Notch, throughout the entire PSM, they found that many somite boundaries still formed, despite the absence of any interface between cells with differing levels of Notch activation; these boundaries, however, were irregularly spaced, and the resulting irregular blocks of somite tissue lacked the normal antero-posterior polarity. The same was seen when Notch signaling, instead of being artificially activated, was inactivated by mutations in Notch1, or Dll1 (Delta1), or Pofut1 (coding for an enzyme that fucosylates Notch and is required for Notch function). In fact, a similar outcome is seen in zebrafish Notch pathway mutants – clefts eventually appear in the mesoderm, dividing it up into somites, but these clefts form later than normal and are crooked and irregularly spaced. The somitic mesoderm, it seems, has a propensity to split up into tissue blocks and will do so even if the segmentation clock is broken and Notch signaling defective. The role of the clock is to control the pattern of this splitting, ensuring that the clefts are regularly spaced, and to confer on each somite a regular antero-posterior polarity. For this last step, it seems that Notch signaling is required directly and not merely to keep the segmentation clocks of the individual cells ticking synchronously in the run-up to overt segmentation; for in the mice where NICD is expressed throughout the tissue, each somite has a double-posterior character, whereas when Notch fails each somite has a double-anterior character 50.

The formation of the somites is not the end of the involvement of Notch signaling in the development of the somitic cell lineage. For example, skeletal muscle tissue, which arises from the somites, also depends on this pathway to control the differentiation of myoblasts and satellite cells and their incorporation into multinucleate muscle fibers 51525354. Like that other ubiquitous communication device, the mobile phone network, the Notch signaling pathway has been recruited for many different purposes – for the simple delivery of instructions from one individual to another, for competitions and collaborations, for the synchronization of individual actions, and for the playing of the tunes to which cells dance.