A Complex Oscillating Network of Signaling Genes Underlies the Mouse Segmentation Clock

See allHide authors and affiliations

Science  08 Dec 2006:
Vol. 314, Issue 5805, pp. 1595-1598
DOI: 10.1126/science.1133141


The segmental pattern of the spine is established early in development, when the vertebral precursors, the somites, are rhythmically produced from the presomitic mesoderm. Microarray studies of the mouse presomitic mesoderm transcriptome reveal that the oscillator associated with this process, the segmentation clock, drives the periodic expression of a large network of cyclic genes involved in cell signaling. Mutually exclusive activation of the notch–fibroblast growth factor and Wnt pathways during each cycle suggests that coordinated regulation of these three pathways underlies the clock oscillator.

The segmentation clock drives the expression of a very limited number of genes whose mRNA shows a dynamic expression sequence that is repeated in the presomitic mesoderm (PSM) each time a new somite forms (1). Most of the known cyclic genes are components of the notch pathway oscillating in phase with each other. In chick and mouse, they include genes coding for transcription factors of the hairy and enhancer of split (Hes) family and the lunatic fringe (Lfng) glycosyltransferase (1). Additionally, in mouse, a single component of the Wnt pathway, Axin2, oscillates out of phase with the notch pathway cyclic genes (2).

We have used gene expression arrays systematically to explore the cyclic transcription program associated with the segmentation clock in the mouse PSM. During the formation of each somite, Lfng is expressed in the PSM as a wave that sweeps across the tissue in a posterior-to-anterior direction (1). Therefore, by visually comparing the anteroposterior position of the Lfng expression stripes in the PSM in stained embryos, it is possible to define an approximate chronological order of the embryos along the segmentation clock oscillation cycle (3, 4). We collected PSM samples from 40 mouse embryos ranging from 19 to 23 somites and used their Lfng expression patterns as a proxy to select 17 samples covering an entire oscillation cycle (fig. S1 and Figs. 1 and 2, A and B). Probes were produced from RNA extracted from the dissected PSMs by using a two-step amplification protocol and were hybridized to Affymetrix GeneChip MOE430A (Affymetrix, Santa Clara, CA) (3, 4). The transcription profiles of known cyclic genes displayed pronounced patterns of oscillation (Fig. 2B). For example, the temporal pattern of Hes1 expression detected on the arrays was in phase with the pattern of Lfng mRNA expression detected by in situ hybridization in the contralateral PSM (Fig. 2, A and B). This is expected because the mRNA of the notch pathway cyclic genes have been shown by in situ hybridization to oscillate synchronously in the PSM (1).

Fig. 1.

Generation of a microarray time series of PSM samples along one period of the segmentation clock oscillation. (Top left image) Lateral view of a 9.0-day-postcoitus (dpc) mouse embryo labeled with Uncx4.1. Yellow box delimits the tail region, which contains the PSM schematized to the right. Schemes represent dorsal views of the tail region. The right posterior half PSM was dissected for the microarray analysis (red rectangle); the rest of the embryo including the intact left PSM was saved for in situ hybridization with Lfng. On the basis of the position of the Lfng stripes in the left PSM (purple), each sample could be positioned retrospectively along one period of the segmentation clock cycle.

Fig. 2.

Identification of cyclic genes based on the PSM microarray time series. (A) Left side of the 17 mouse embryos, whose right posterior PSMs (below red hatched line) were dissected for microarray analysis. Embryos were ordered along one segmentation clock cycle according to the position of Lfng stripes in their left PSM as revealed by in situ hybridization (fig. S1). (B) Log2 ratios of the expression levels of the Hes1 (blue) and Axin2 (red) cyclic genes in each microarray of the time series. (C) Phaseogram of the cyclic genes identified by microarray and L-S analysis. Blue, decrease in gene expression; yellow, increase in gene expression; pink squares, genes validated by in situ hybridization; and orange circles, nonvalidated genes, that is, not evidently cyclical as detected by in situ hybridization.

To identify genes displaying a periodic expression pattern in the PSM, we applied a recently developed modification of the Lomb-Scargle (L-S) algorithm (3, 4) to the filtered data set. This allowed us to detect cyclic patterns characterized by different periods and to compute statistics that assess the significance of each periodic pattern. We operated under the assumption that the 17 samples were evenly spaced in time along the 2-hour segmentation clock cycle, resulting in a 7-min time interval between two consecutive time points (even though in reality the time points may not have been evenly spaced). This procedure allowed us to identify statistically significant periodic patterns along with their corresponding periods. The period with the most significant P value was selected for each profile. Six of the eight known mouse cyclic genes—Hes1, Hes5, Hey1, Lfng, Axin2, and Nkd1—were identified with periods of 94, 102, 112, 81, 102, and 112 min, respectively. These known cyclic genes were used as true positives to refine filtering parameters that minimize the number of candidate cyclic genes while maximizing the number of highly significant true positives among them. The most specific parameter settings retained 36 genes, including four of the known cyclic genes (table S2).

When ordered by their time of maximum expression in the segmentation clock cycle, identified cyclic genes segregate into two main clusters with opposite phase (Figs. 2C and 3, A to C). One of the clusters contains the known cyclic genes of the notch pathway–Hes1, Hes5, and Hey1–detected in this analysis (Fig. 2C). The basic helix-loop-helix (bHLH) transcription factor Id1 that dimerizes with Hes1 also belongs to this cluster (Fig. 2C) and exhibits a dynamic expression in the PSM (Fig. 3D). Lfng is also detected as periodic and in phase with the other notch cyclic genes, if slightly less stringent filtration parameters are used (Fig. 3A) (3, 4). In addition, this cluster contains Nrarp, a direct target of notch signaling (Figs. 2C and 3A) (5). A clear Nrarp cyclic expression, reminiscent of Lfng, was observed after in situ hybridization in mouse embryos (Fig. 3E). A connection to Wnt signaling is provided by the vertebrate homolog of legless, Bcl9L, which shuttles β-catenin to the nucleus (Figs. 2C and 3A). Also, by using slightly less stringent thresholds (3, 4), we find in this group Nkd1 (Fig. 3A), an inhibitor of Wnt signaling acting downstream of notch in the segmentation clock (6).

Fig. 3.

Mutually exclusive activation of the notch-FGF and Wnt clusters during one segmentation clock oscillation. Expression profiles of cyclic genes of (A) notch, (B) FGF, and (C) Wnt pathways along the microarray time series. (D to O) Lateral view of the right caudal part of 9.0-dpc mouse embryos hybridized with the probes indicated in black boxes. For each probe, two representative images illustrating the dynamic expression of the gene in the PSM are shown.

The same cluster also contains genes coding for proteins involved in the fibroblast growth factor (FGF)–mitogen-activated protein kinase (MAPK) pathway, for example, the FGF pathway inhibitor Spry2 (Figs. 2C and 3B) (7). With slightly less stringent filtration parameters, we also identified in this cluster another inhibitor of the FGF pathway, Dusp6 (also called Mkp3), coding for an extracellular signal–regulated kinase (ERK) phosphatase (8) and the phosphatase Shp2 (also called Ptpn11) required to activate the pathway (3, 4) (Fig. 3B). Cyclic expression of Spry2 and Dusp6 was confirmed by in situ hybridization in the PSM (Fig. 3, F and G). Because Spry2 and Dusp6 expression was shown to be downstream of the FGF-MAPK pathway in the PSM of chick and mouse embryos (8, 9), our data suggest that the FGF pathway is activated periodically in synchrony with the notch pathway in this tissue. Periodic expression of notch-related cyclic genes in the PSM is independent of FGF signaling (9, 10). To test the possibility that the cyclic expression of FGF targets is imposed by periodic notch activation, we examined expression of Spry2 in mice homozygous for a null allele of the Rbpjk gene, which abolishes notch signaling (3, 4). In these mutants, Spry2 expression remained dynamic (fig. S2, A and B, n = 10), suggesting that periodic expression of genes from notch and FGF pathways is activated in parallel. Other genes involved in FGF-MAPK signaling in this cluster include Hspg2 (also called Perlecan) (Figs. 2C and 3B) (a coreceptor for FGF); the Bcl2-family member Bcl2l11 (also called Bim) (Figs. 2C and 3, B and H) and the Zn finger transcription factor Egr1 (Figs. 2C and 3B), which act downstream of the MAPK pathway; and EphrinA1 (efna1) (Figs. 2C and 3, B and I). Because it has been proposed that FGF and Wnt signaling establish a dynamic gradient that controls the competence of PSM cells to respond to the segmentation clock (2, 10), our data point to a further degree of complexity regulating FGF signaling in the PSM.

The second cluster of periodic genes contains genes cycling in opposite phase to the notch-FGF cluster (Figs. 2C and 3C). In this cluster, we found the known cyclic gene Axin2 and a majority of the cyclic genes associated with Wnt signaling (Fig. 3C). These include the soluble Wnt inhibitor Dkk1 and the intracellular Wnt inhibitor binding to dishevelled, Dact1 (also known as Dpr or Frodo). Other genes in this cluster, such as those coding for the transcription factors Sp5 and c-Myc and the transmembrane receptor Tnfrsf19 (also called Troy), are downstream targets of the Wnt pathway (1113). Two other genes in this cluster, the hyaluronan synthase Has2 (14) and the Phlda1 gene involved in Fas signaling (15), have no known association with the Wnt pathway. Their expression is strongly down-regulated in Wnt3a hypomorph mouse mutants vestigial tail (vt), suggesting that these genes are also targets of Wnt signaling (fig. S2, C to F). All of these genes show a dynamic expression pattern in the PSM with the use of in situ hybridization (Fig. 3, J to O, and fig. S3). Inactivation of Dkk1 (16), Sp5 (17), c-Myc (18), and has2 (14) has been reported to produce segmentation defects. Dynamic expression of Dkk1 was observed by in situ hybridization only when we used intronic probes that recognize nascent nuclear transcripts (Fig. 3J). This suggests that Dkk1 is periodically transcribed in the PSM but that its mRNA is too stable to allow visualization of its oscillations by in situ detection using a probe recognizing the cytoplasmic mRNA transcripts. This could essentially reflect different sensitivity ranges between the microarray and the in situ hybridization methods.

The majority of the cyclic genes from the Wnt cluster, including Dkk1, c-Myc, Axin2, Sp5, and Tnfrsf19, are direct downstream targets of the Wnt pathway (1113, 19, 20), suggesting that the Wnt pathway is rhythmically activated in the PSM. Recent gene expression analysis in the mouse embryo identified a much larger number of downstream targets of Wnt signaling than the set of genes coregulated with Axin2 (21, 22). The only common gene between these embryonic Wnt targets and the cyclic genes identified by L-S analysis was Axin2. Thus, Wnt pathway genes periodically transcribed in the PSM appear to involve a restricted subset of the Wnt target genes.

Genes in the notch–FGF, and Wnt clusters identified by our approach are expressed in opposite phases during each segmentation clock cycle (Fig. 4), suggesting that, whereas notch and FGF might act synergistically, their activation is mutually exclusive to that of the Wnt pathway in the PSM. This is consistent with the idea that reciprocal inhibition of notch-FGF and Wnt pathways might play a role in the implementation of the clock oscillations.

Fig. 4.

A network of cyclic genes of the notch, FGF, and Wnt pathways underlies the mouse segmentation clock. Notch and FGF-MAPK cyclic genes (green) oscillate in opposite phase to Wnt cyclic genes (red). The other components (black and white) belong to the canonical notch, FGF-MAPK, and Wnt pathways. A large number of identified cyclic genes are involved in negative feedback loops.

Cyclic expression of 20 out of 29 tested genes (69%) was validated by in situ hybridization, demonstrating the high efficiency of our strategy to identify cyclic genes (Fig. 2C). This number is most likely an underestimation of the total number of bona fide cyclic genes, which is expected to increase with the use of arrays covering a larger fraction of mouse genes, with improved amplification techniques, and with better sampling allowing more robust signal detection. Therefore, we expect a minimum number of cyclic genes between 50 and 100. Most of the validated cyclic genes we identified are involved in signal transduction or transcription and belong to the notch, Wnt, and FGF pathways, suggesting that the oscillator mechanism largely relies on these three pathways. In the yeast cell cycle, periodic transcriptional regulation is restricted to selected (perhaps limiting) subunits of multiprotein complexes that control the cycle (23). Similarly in the segmentation clock, only a subset of the components of the notch, FGF, and Wnt pathways are expressed in a periodic fashion, at least at the mRNA level. The half-life of proteins coded by the cyclic genes Hes1, Hes7, and Lfng has been shown to be very short (1), hence leading to protein oscillations with the same period as their mRNAs. It is expected that an important number of the identified cyclic genes encode cyclic proteins that could act in the oscillator mechanism. Thus, a model of dynamic complex assembly may also control the periodic signaling associated with the segmentation clock network.

Current models of the segmentation clock have their basis in a very limited number of inhibitory components that establish negative feedback loops involved in the generation of oscillations (1). Our analysis identified several additional inhibitors of the notch, FGF, and Wnt pathways that could, in principle, participate in similar negative feedback loops. Thus, our data suggest that the oscillator relies on the periodic regulation of a network of such inhibitors rather than on a few key components. Such a network might account for the robustness of the segmentation process.

Supporting Online Material

Materials and Methods

Figs. S1 to S4

Tables S1 to S5


References and Notes

View Abstract

Navigate This Article