A β-catenin gradient links the clock and wavefront systems in mouse embryo segmentation

Somites are transient epithelial structures that give rise to the axial skeleton, and are the first overt sign of a metameric body plan in vertebrate embryos. They are produced by segmentation of the paraxial mesoderm, a periodic process that has been associated with a molecular oscillator or segmentation clock identified in studies demonstrating periodic transcription of cyclic genes in the PSM¹. A large network of signalling genes that are involved in the Notch, Wnt and fibroblast growth factor (FGF) pathways was shown previously to be rhythmically expressed in the posterior PSM with a periodicity matching that of somite production^1–3. Although the molecular nature of the oscillator’s pacemaker remains unclear, it has been shown in mouse embryos that intact Wnt signalling is required for oscillations to occur^2,4,5. In addition, gradients of Wnt, FGF and retinoic acid signalling translate the signalling pulse generated by the clock into the spatial periodic pattern of segments^2,6,7. Activity of the Wnt/FGF gradient is highest in the posterior embryo; however, the antagonistic retinoic acid-signalling gradient peaks towards the anterior part of the embryo. The Wnt/FGF gradient has been shown to set up a threshold defining a PSM level — called the determination front — at which cells become responsive to the clock signal. At the determination front, mRNA oscillations cease and cells exhibit striped expression of key factors, such as Mesoderm posterior 2 (Mesp2), which specify somite polarity and future segment boundaries^8,9. Here, we examine the role of the Wnt gradient in somitogenesis and determine how it interconnects with the function of the Wnt pathway in clock oscillations of the PSM.

Using immunohistochemistry, we observed a clear posterior-to-anterior nuclear gradient of β-catenin (a key intracellular mediator of the Wnt pathway) in the PSM of mouse embryos (Fig. 1a–c). A posterior gradient for β-catenin was observed in all embryos analysed (n = 19), irrespective of their segmentation clock phase, as determined by the mRNA expression pattern of the cyclic Wnt target gene Axin2 (refs ^{2, 10, 11}; Supplementary Information, Fig. S1). To examine the role of the nuclear β-catenin gradient in somitogenesis, we used a conditional strategy to selectively delete or stabilize β-catenin in the PSM of mouse embryos. First, to selectively delete β-catenin in the PSM, mice carrying a conditional β-catenin^floxed allele¹² were crossed with the T–Cre mouse line in which Cre recombinase is driven by the T (Brachyury) promoter. In this transgenic mouse line, Cre is expressed in precursors of the paraxial mesoderm located in the primitive streak. In embryos homozygous for the conditional-null β-catenin^floxdel allele in primitive streak descendents, a few abnormal somites formed anteriorly and a severe axial truncation was observed (Supplementary Information, Fig. S2 and data not shown). This confirms the requirement of Wnt/β-catenin signalling in antero-posterior axis formation and gastrulation; however, further analysis of the role of β-catenin during somitogenesis was not feasible with this phenotype.

Next, we used T–Cre mice carrying the conditional gain-of-function allele β-catenin^lox(ex3) to disrupt the β-catenin gradient by constitutively stabilizing β-catenin in the PSM. β-catenin^lox(ex3) contains loxP sites flanking exon 3, which codes for a crucial sequence recognized by the destruction complex that targets β-catenin for degradation¹³. This strategy resulted in accumulation of mutant β-catenin^del(ex3) protein specifically in mesodermal lineages, including the PSM. Nuclear staining of β-catenin was elevated throughout the entire PSM of β-catenin^del(ex3) mutants (Fig. 1d–f), without an appreciable posterior-to-anterior protein gradient, as seen in wild-type embryos (Fig. 1a–c). No significant difference in the proliferative state of the posterior PSM or the tail bud was observed between mutant and wild-type embryos (Supplementary Information, Fig. S1). In mutant embryos carrying the β-catenin^del(ex3) allele, no sign of morphological boundary formation or somites could be seen along the body axis (Fig. 1g, h), except for up to four irregular somites seen occasionally in the most anterior aspect of mutant paraxial mesoderm. Mutant embryos could not complete the turning process during gastrulation, exhibited a rounded allantois that failed to fuse with the chorion and died at approximately day 10.5 of development. The accumulation of nuclear β-catenin throughout the PSM was accompanied by upregulation of Wnt signalling, as determined by using the Wnt-reporter mice BAT–Gal¹⁴. In β-catenin^del(ex3)–BAT–Gal embryos, significant anterior extension of lacZ expression was observed in the PSM (Supplementary Information, Fig. S3; n = 8). The Wnt-responsive gene Axin2 (refs ^{10, 11}) was markedly upregulated in the PSM (Fig. 2a, á; n = 4), consistent with sustained activation of canonical Wnt signalling. Other direct targets of the Wnt pathway involved in PSM patterning were upregulated and shifted anteriorly in mutant β-catenin^del(ex3)/+ embryos, compared with control littermates. These include the genes T-box6 (Tbx6; ref. 15), mesogenin1 (Msgn1; ref. 16) and Delta-like 1 (Dll1; refs ^{17, 18}; Fig. 2b, b´; n = 5; c, c´; n = 5; d, d´; n = 2). The expression domain of Fgf8 was slightly extended anteriorly (Fig. 2e, é; n = 4). In addition, the FGF-signalling target Pea3 (ref. ¹⁹) was clearly upregulated and shifted anteriorly in the PSM, suggesting that the strength of the FGF-signalling gradient is increased (Fig. 2f, f´; n = 5). The anterior PSM of β-catenin^del(ex3)/+ mutants, nevertheless, showed signs of maturation. Thus, markers for the anterior PSM and somites, such as Paraxis²⁰ (Fig. 2g, g´; n = 2), and Raldh2 (Aldh1a2; ref. 21; Fig. 2h, h´; n = 2) were present, but their expression was only transient or reduced, respectively, compared with wild-type embryos. We found that Mesp2 expression, which correlates with the level of the determination front in wild-type embryos (Fig. 3a), was significantly shifted anteriorly in mutant embryos (Fig. 3b, n = 16 and Supplementary Information, Fig. S3). Downstream targets of Mesp2 (Epha4 and Cer1; ref. 22) were expressed in stripes in the anterior PSM of controls but absent from anterior PSM in mutant littermates (Fig. 3c–f; n = 5 for both). Furthermore, the expression of a posterior somite marker Uncx4.1 was clearly downregulated in mutant embryos (Fig. 3g, h; n = 3). Thus, overexpression of β-catenin causes an anterior shift of the determination front and delays the activation of Mesp2 expression (Fig. 3i). In wild-type embryos, the position of the Mesp2 stripe was found to lie outside the β-catenin gradient (Supplementary Information, Fig. S1). These data indicate that downregulation of nuclear β-catenin is required for activation of Mesp2 downstream targets, and explains the absence of morphological boundary formation in mutant embryos.

Oscillations of the Wnt pathway have been postulated to rely on the function of a group of negative feedback inhibitors of the Wnt pathway, such as Axin2, dickkopf 1 (Dkk1) or Dact1 (refs ^{2, 3}). Periodic expression of these direct Wnt targets is expected to rhythmically alter β-catenin levels in PSM cells. We did not detect visible oscillations of β-catenin protein in the PSM (Supplementary Information, Fig. S1), but fluctuations of small amplitude may, in principle, account for the oscillatory transcription of bona fide targets of the canonical Wnt pathway, such as c-myc (ref. ³) or Axin2 (ref. ²) in the PSM. If this were the case, then oscillations of the Wnt pathway should be disrupted in β-catenin^del(ex3)/+ embryos in which a high level of nuclear β-catenin is constantly maintained in PSM cells. To test this hypothesis, we analysed the expression of a cyclic direct Wnt target, Dkk1, using an intronic probe to detect only nascent, pre-mRNA³ (Fig.4a–d). In β-catenin^del(ex3)/+mutant embryos, Dkk1 expression levels were higher overall when compared with wild-type embryos, and expression was always detected in the posterior PSM (Fig.4c, d). When mutant littermates were compared after identical staining procedures, the intensity of Dkk1 pre-mRNA expression varied in the posterior PSM, ranging from faint (16 out of 32 embryos) to strong (16 out of 32 embryos) (Fig. 4c, d; arrows). In addition, a clear, striped pattern of active transcription was observed in the PSM of mutant embryos (Fig. 4c, d; arrowheads). Therefore, constitutive expression of high nuclear β-catenin levels does not block dynamic expression of the Wnt cyclic gene, Dkk1, suggesting that Wnt-signalling oscillations still occur in β-catenin^del(ex3)/+ mutants.

To confirm the status of Notch oscillations in β-catenin^del(ex3)/+ mutants, expression of the Notch cyclic genes lunatic fringe (Lfng; Fig. 4e–h) and Hairy and enhancer of split 7 (Hes7; ref. 23) was examined. Posterior expression of Lfng in mutant embryos varied considerably in intensity between littermates (Fig. 4g, h; arrows; 14/22 strong expression, 8/22 weak expression), suggesting that in the posterior PSM, Notch oscillations still occur. Up to five stripes of variable width and inter-stripe distance were found in the middle and anterior PSM, both for Lfng (Fig. 4g, h; arrowheads) and for Hes7 mRNA (Supplementary Information, Fig. S4). During development, the total number of stripes increased slowly in the PSM of mutant embryos, paralleling the progressive increase in size of the PSM (Supplementary Information, Fig. S5). To rule out the possibility that the remaining wild-type allele in β-catenin^del(ex3)/+ mutant embryos controls the dynamic expression of Lfng, we generated β-catenin^{del(ex3)/floxdel} mutants in which PSM cells express only the gain-of-function allele. The phenotype of these mutants was very similar to that of β-catenin^del(ex3)/+ mutants, with several stripes of Lfng in the PSM (Supplementary Information, Fig. S4), indicating that the remaining wild-type β-catenin allele does not contribute to this phenotype.

To test directly whether the multiple stripes of Lfng expression in the mutant PSM correspond to oscillating transcription domains, we developed a real-time imaging strategy to visualize the activity of the segmentation clock in living mouse embryos (Fig. 5 a–f). To this end, we generated transgenic animals that expressed a highly destabilized Venus reporter (a variant of yellow fluorescent protein, YFP) under the control of the cyclic Lfng promoter^24,25. Real-time imaging of control reporter mouse embryos with two-photon, time-lapse microscopy revealed periodic waves of Lfng expression, which are initiated in the posterior PSM approximately every 2 h (Fig. 5a, c; Supplementary Information, Movie S1; n = 3). These waves of transcriptional activity traversed the PSM from posterior to anterior and stopped in the anterior PSM where they briefly stabilized as a striped expression domain before disappearing (Fig. 5a, arrows; Fig. 5c; Supplementary Information, Movie S1). Cells in the anterior PSM stopped oscillating once they reached a certain maturation state and due to somite formation, there was a posterior regression of the domain showing oscillations (Fig. 5c, e). To examine Lfng oscillations in the mutant background, mice carrying the β-catenin^del(ex3) allele were crossed with the Lfng reporter line. Analysis of the movies generated from these mutant embryos showed that the striped expression of Lfng corresponds to multiple, oscillating expression domains that sweep through the expanded anterior PSM (Fig. 5b, arrows; 5d, f; Supplementary Information, Movie S2; n = 4). According to their axial location, these cells should have stopped oscillating, but instead they continued to do so for an extended period of time under the influence of elevated β-catenin levels (Fig. 5b, d, f). Consistently, no regression of the oscillatory domain was observed during time-lapse recordings (Fig. 5d). The signal intensity and the amplitude of the oscillations were, however, reduced in mutants, compared with wild-type embryos (Fig. 5). These results demonstrate that oscillations of the segmentation clock occur even in the presence of high and steady nuclear β-catenin levels. In addition, this shows that β-catenin is able to maintain clock oscillations and suggests that the arrest of oscillations under normal conditions is linked to the level of β-catenin protein in the PSM.

In this study, we have shown that by deleting β-catenin in mesoderm precursors in the primitive streak, formation of the paraxial mesoderm is blocked, causing axis truncation. In contrast, expression of a stable form of β-catenin throughout the paraxial mesoderm led to a Wnt-signalling gain of function, causing a delayed and defective maturation process and a subsequent increase in the size of the PSM along the antero-posterior axis. This suggests that the function of the Wnt-signalling gradient in the PSM is carried out by the β-catenin protein gradient.

The delay in maturation was accompanied by sustained oscillations of the Wnt and Notch pathway in PSM cells. Thus, Wnt activation in the posterior PSM provides a permissive environment for oscillations of the segmentation clock to occur. This also suggests that the decision to stop oscillations in the anterior PSM under normal conditions is not based on a limited ability of cells to oscillate (that is, a counting mechanism) but requires the downregulation of Wnt signalling.

However, the PSM of mutant embryos retained signs of maturation and Mesp2 became consistently activated at the level where oscillations eventually stopped in the extended anterior PSM of the β-catenin^del(ex3) mutants. Thus, maintaining a high level of nuclear β-catenin was not sufficient to maintain the posterior PSM identity indefinitely, indicating that β-catenin interacts with other signalling pathways to perform this function. The Wnt/nuclear β-catenin gradient overlaps with an FGF-signalling gradient, which has also been shown to have an important role in the maintenance of posterior PSM identity^6,26,27. Wnt3a/β-catenin was shown in loss-of-function experiments to function upstream of Fgf8 (refs ^{2, 28}). Here, we show that in a gain-of-function situation, β-catenin accumulation can increase Fgf8 expression, and FGF signalling extends more anteriorly in the expanded PSM. Thus, some of the observed effects in β-catenin^del(ex3) mutants could be mediated indirectly by the FGF gradient. To test this possibility directly, we generated compound β-catenin^lox(ex3)/+–Fgfr1^f/ko–T–Cre mutant embryos to block FGF signalling in the PSM (Fig. 4i–q). In the PSM of compound mutants, expression of the FGF target Dusp4 (ref. ²⁹) was downregulated (Fig. 4k; n = 8). The anterior shift of the Mesp2 stripe was still observed in the compound mutants (Fig. 4n; n = 5), but was not as pronounced as that in the β-catenin^lox(ex3)/+–Fgfr1^f/+–T–Cre single mutants (Fig. 4m). Therefore, part of the effect of β-catenin stabilization in the PSM seems to be mediated indirectly through FGF signalling. This also suggests that the β-catenin gradient can influence PSM patterning and maturation directly, possibly through the activation of targets such as Msgn1 (ref. ¹⁶). In addition, although no oscillations of Lfng were detected in the PSM of conditional homozygous Fgfr1–T–Cre single mutants³⁰, the presence of a β-catenin^del(ex3) allele led to the formation of several Lfng stripes in the extended PSM of compound mutants (Fig. 4q; n = 5), indicating that Lfng oscillations were rescued. This is consistent with the proposed role of Wnt signalling downstream of FGF signalling in the control of Lfng oscillations³⁰. Thus, these experiments demonstrate the lack of a simple epistatic relationship between Wnt and FGF signalling in the PSM. Rather, these pathways are tightly interconnected and seem to synergize in controlling PSM maturation, thereby defining the permissive environment for oscillations of the segmentation clock.

Although accumulation of nuclear β-catenin markedly altered PSM maturation, activity of the segmentation clock was not disrupted by these experimental conditions. We conclude that oscillations of Wnt and Notch targets are not achieved by oscillating (nuclear) β-catenin protein levels. Regulation of β-catenin protein levels has normally been regarded as essential in defining canonical Wnt-signalling activity. However, there is growing evidence that nuclear β-catenin is not the sole determinant of transcriptional activity downstream of canonical Wnt signalling, but relies also on the regulated interaction with additional (nuclear) cofactors³¹. Our data suggest that oscillations of Wnt targets in the PSM are probably caused by the cyclic activity of a β-catenin cofactor. Interestingly, cyclic recruitment of several nuclear cofactors of β-catenin to the genomic enhancer/promoter locus of Wnt targets (such as c-myc, a Wnt cyclic gene in mice)³ has indeed been shown to occur³². This cyclic recruitment occurred even in the presence of a steady and high nuclear β-catenin level. The present study illustrates two distinct strategies of Wnt-signalling regulation that operate simultaneously in embryonic cells. Clarifying the molecular details of this versatile regulation in the developing embryo might reveal some underlying principles of signalling regulation in biological systems. □