Report

Retinoic Acid Controls the Bilateral Symmetry of Somite Formation in the Mouse Embryo

See allHide authors and affiliations

Science  22 Apr 2005:
Vol. 308, Issue 5721, pp. 563-566
DOI: 10.1126/science.1108363

Abstract

A striking characteristic of vertebrate embryos is their bilaterally symmetric body plan, which is particularly obvious at the level of the somites and their derivatives such as the vertebral column. Segmentation of the presomitic mesoderm must therefore be tightly coordinated along the left and right embryonic sides. We show that mutant mice defective for retinoic acid synthesis exhibit delayed somite formation on the right side. Asymmetric somite formation correlates with a left-right desynchronization of the segmentation clock oscillations. These data implicate retinoic acid as an endogenous signal that maintains the bilateral synchrony of mesoderm segmentation, and therefore controls bilateral symmetry, in vertebrate embryos.

The body plan of vertebrate embryos is overtly symmetric; only later in development do internal organs move into asymmetric positions. Among the most obviously symmetric embryonic structures are the left and right somitic columns, in which paired epithelial structures arise by segmentation of the paraxial mesoderm. Somite development relies on a “clock and wavefront” mechanism (1), in which a molecular oscillator that depends on Notch and Wnt pathways (the “segmentation clock”) generates cyclic waves of gene expression along the presomitic mesoderm (PSM) (2, 3). In addition, a caudal-to-rostral Fgf8 (fibroblast growth factor 8) mRNA gradient acts as a moving wavefront (the “determination front”), triggering somite differentiation and setting the intersomitic boundaries (4). Retinoic acid (RA) plays multiple roles during patterning of the vertebrate anteroposterior axis. Altered RA signaling affects patterning of the vertebrae, generating homeotic transformations and/or segmentation defects (5). Experiments in chick and amphibian embryos have shown that RA signaling can counteract the Fgf8 gradient in both the neural tube and PSM (6, 7).

To define the role of RA signaling during mouse somitogenesis, we first used lacZ in situ hybridization (ISH) to characterize the pattern of activation of the RARE_hsp68_lacZ transgene, a sensitive reporter for the presence of endogenous RA (8). At somite stages 0 to 2, the transgene was active in both the somitic and PSM but not within the primitive streak (Fig. 1A) (9). From somite stages 4 to 10, RA activity was observed throughout the formed somites and within the rostral PSM up to a sharp, symmetric posterior boundary. Because this boundary was always located at the same distance from the last formed somite, the RA response appears to progress as a symmetric moving front during formation of the first somite pairs (Fig. 1, B to D).

Fig. 1.

Embryonic RA signaling is dynamically regulated during somitogenesis. (A to D) Whole-mount lacZ detection in RARE_hsp68_lacZ transgenic mouse embryos at successive somite stages (ventral views). lpm, lateral plate mesoderm; ps, primitive streak. (E to G) Combined detection of β-Gal (green) and Mesp2 transcripts [black in (E) and (F), whole mounts; red in (G), confocal section] in RARE_hsp68_lacZ embryos. (H to J) lacZ signal in a transgenic embryo at somite stage 8 after increasing staining times (in minutes). Signal in the PSM demarcates a fixed boundary (red line). (K and L) Combined detection of Cyp26A1 and lacZ transcripts in transgenic embryos (n, node). (M to O) RALDH2 protein immunodetection in wild-type embryos. Somite stages are as indicated (0s to 12s). White arrowheads point to the last formed intersomitic (SI/S0) boundary. Whatever the somite stage, SI and S0 designate the last formed mature (epithelialized) somite and the next forming somite, respectively [see (10)].

To map the location of RA-responsive cells in embryos at somite stages 6 to 10, we combined immunofluorescent β-galactosidase (β-Gal) detection with ISH for Mesp2, whose rostral transcript boundary marks the border between presomites S-I and S-II [(10) for presomite nomenclature; (11)]. β-Gal+ cells reached the Mesp2 domain (Fig. 1, E to G), which shows that the RA response occurs in S-I. To determine whether the RA response is gradual, we examined embryos after increasing times of lacZ staining. No change was observed in the extent of the signal (Fig. 1, H to J); this result implies that cells respond to RA according to a sharp front rather than a gradient.

We mapped the location of the RA-responsive front with respect to the expression of the RA-metabolizing enzyme Cyp26A1 (12). Both domains were consistently separated by nonlabeled PSM cells (Fig. 1, K and L), indicating that the RA-responsive front is not dictated by CYP26A1 activity. On the other hand, expression of the RA-synthesizing enzyme RALDH2 (retinaldehyde-specific dehydrogenase type 2) matched that of the reporter transgene in the rostral PSM (Fig. 1M). RALDH2 protein was no longer detected in the PSM from somite stages 10 to 12 (Fig. 1, N and O), although expression was later found in the mature somites (9). A similar drop in the activity of the RARE_lacZ transgene was seen in the PSM at these stages (fig. S1). Thus, a transition in RA biosynthesis occurs in the PSM during the formation of somites 11 and 12, which corresponds to the future cervicothoracic transition in the mouse.

To investigate the function of RA during somitogenesis, we analyzed Raldh2–/– embryos, which are known to exhibit abnormally small somites (13). The dynamics of somite formation was analyzed using Uncx4.1 as a marker of mature somites. Until somite stage 7 or 8, somites were smaller in mutants but were evenly aligned along the left and right sides (Fig. 2, A and B). However, at later somite stages, some of the Raldh2–/– embryos exhibited asymmetric Uncx4.1 patterns, such that at least one additional expression stripe was present on the left side (Fig. 2, C and D). Asymmetric patterns were also observed for Meox1, a pansomitic marker induced during somite formation in S-I, which indicates that the defect occurs within the somite-forming region (fig. S2).

Fig. 2.

Asymmetric somite development in Raldh2–/– embryos. (A to E) Whole-mount ISH of wild-type (A) and Raldh2–/– [(B) to (E)] embryos with an Uncx4.1 probe (dorsal views). The numbers of formed somites on left and right sides are indicated below. (F) Summary of the numbers of Raldh2–/– embryos with equal (diagonal, green) or unequal somite numbers. The most acute phase of the asymmetry defect is shown in red. Few embryos exhibited an additional somite on the right side (blue). A “symmetric” (sym*) score at late somite stages (16 and above) refers to the location of the last formed somites, as more rostral somites were often misaligned [bracket in (E)].

Analysis of numerous Uncx4.1-hybridized Raldh2–/– embryos confirmed that asymmetric somite development first appears by somite stage 8 to 9 (Fig. 2F). The frequency of asymmetric patterns increased during the next cycles of somitogenesis, such that only 2 of 18 mutants exhibited a symmetric 11-somite pattern. In most mutants, one or more supernumerary somites were formed on the left side (Fig. 2F). Up to three additional somites could thus be seen (Fig. 2D). Eventually, a realignment of the last formed somites was seen in most mutants after formation of 15 or 16 somite pairs, although the adjacent somites were not properly aligned (Fig. 2E, bracket). One mutant showed arrested somite formation on the right side, with at least six somites fewer than on the left side (9).

We investigated the molecular basis of asymmetric somite development in Raldh2–/– embryos. The pace of somitogenesis is controlled by a molecular oscillator generating cyclic waves of gene expression within the PSM, in particular for genes of the Notch pathway (3, 4). Among these, we analyzed Lunatic fringe (Lfng), whose cycling expression (14) can be visualized as three successive phases in which expression (i) is confined to the caudal PSM, (ii) “sweeps” along the PSM (Fig. 3A), and (iii) becomes localized to the forming somites (Fig. 3B). Asymmetric Lfng patterns were first observed in some Raldh2–/– embryos at somite stages 7 to 9, ranging from a subtle posterior shift on the right side (9)(n = 4 of 14) to a complete change in the oscillating phase patterns (Fig. 3C; n = 2 of 14). In slightly older mutants, “canonical” Lfng expression patterns were usually observed in the left PSM, whereas a supernumerary band of expression could be seen in the right PSM (Fig. 3D; n = 4 of 6). The right-side rostral bands were misaligned with the contralateral band, and a “salt and pepper” pattern of expressing and nonexpressing cells was seen between them (Fig. 3D) (fig. S3). At somite stages 15 and 16, Lfng expression was symmetric in the majority of Raldh2 mutants (9) (n = 6 of 8). These results show that the progression of the waves of Lfng expression is no longer coordinated along the left and right PSM, during a temporal window correlating with asymmetric somite development, in the RA-deficient embryos.

Fig. 3.

Whole-mount ISH of wild-type (A, B, and E) and Raldh2–/– (C, D, and F) embryos with Lfng [(A) to (D)] and Hes7 [(E) and (F)] probes (dorsal views), showing abnormal progression of the molecular oscillations in Raldh2–/– embryos. PSM, presomitic mesoderm; CPSM, caudal presomitic mesoderm; FS, forming somite.

We analyzed whether other “oscillating” genes have their expression patterns altered in Raldh2–/– mutants. The basic helix-loop-helix factor HES7 is a pivotal component of the molecular clock, as it represses Lfng in the PSM during each oscillation (15). Several Raldh2–/– embryos exhibited asymmetric Hes7 expression patterns (Fig. 3, E and F; n = 4 of 10) that further suggested a delay in the progression of the oscillatory waves along the right-side PSM (Fig. 3F). Similar results were obtained for Hes1 (n = 2 of 7) and for Axin2 (n = 4 of 9), which encodes a repressor of Wnt signaling (3, 9). Downstream genes such as Mesp2 also exhibited asymmetric patterns (16). Thus, RA deficiency leads to a lack of left-right coordination of the waves of expression of various oscillating genes. Such a molecular defect has never been described in a mouse (or another vertebrate) mutant. It is reminiscent, however, of the misphased oscillatory patterns observed in wild-type zebrafish embryos exposed to a left-right temperature gradient, which increases the rate of somitogenesis on the warmer side (17).

The position of somite boundaries is controlled by a caudal-to-rostral Fgf8 gradient that regresses posteriorly in concert with axis extension (4, 18). We analyzed Fgf8 expression in the Raldh2–/– embryos. At somite stages 4 to 8, the Fgf8 expression domain was extended anteriorly in the mutant PSM (Fig. 4, A and B; n = 7 of 7). It therefore appears that an abnormal distribution of the Fgf8 gradient may be linked to the compacted somite phenotype of Raldh2 mutants, as reported for RA-deficient quail embryos (6). At somite stages 9 and 10, an ectopic distribution of Fgf8 transcripts in the right PSM was also seen in some mutants (Fig. 4, C and D; n = 5 of 12). In the chick embryo, altering the Fgf8 gradient symmetry by grafting an FGF8 bead on the right side results in an asymmetric cyclic gene expression pattern and positioning of somite boundaries (4). This result was interpreted as a slowing of the FGF8 front posterior regression on the grafted side. Similarly, the unequal left-right Fgf8 levels may reflect a delayed regression of the FGF8 front on the right side, contributing to the asymmetric positioning of somite boundaries in severely affected Raldh2–/– mutants. We analyzed whether expression of other Fgfs may also be altered. Fgf18, whose expression is normally restricted to SI and S0 (19) (Fig. 4E), was undetectable in Raldh2–/– embryos (Fig. 4F; n = 6 of 6). Hence, lack of FGF18 signaling may contribute to the somitic abnormalities and make the RA-deficient embryos particularly sensitive to changes in the Fgf8 gradient.

Fig. 4.

Whole-mount ISH of wild-type (A, C, and E) and Raldh2–/– (B, D, and F) embryos with Fgf8 + Uncx4.1 [(A) to (D), dorsal views] and Fgf18 [(E) and (F), lateral views] probes, showing altered Fgf expression in the somite-forming region of Raldh2–/– embryos. Red brackets in (A) and (B) are landmarks to compare Fgf8 mRNA distribution (n, node; so, somites). Red brackets in (D) show unequal Fgf8 levels along left and right PSM.

Our results show that endogenous RA deficiency leads to a lack of coordination of somite formation between the left and right side of the mouse embryo, apparently due to a progressive delay of the oscillatory waves of gene expression in the right PSM. We have thus uncovered a genetically controlled mechanism that actively maintains the bilateral synchrony of mesoderm segmentation. Selective pressure for coordinated left-right somitogenesis is likely to occur throughout vertebrates, given that even a subtle lack of coordination would lead to defects in the bilateral symmetry of somitic derivatives (including the axial skeleton). Interestingly, asymmetric somite development occurs naturally in the cephalochordate Amphioxus (20), which suggests that the role of RA in the control of somitogenesis synchrony is an evolutionary acquisition of vertebrates.

The first waves of Lfng expression have been shown to be asymmetric when they reach Hensen's node in chick embryos (21). This biased expression is necessary for the left side–specific induction of Nodal, an early left-right determinant directly regulated by Notch signaling (22, 23). Our data suggest that RA could act to synchronize the segmentation clock oscillations after the early phase of “natural,” transient left-right asymmetry around the node region. Such an interaction with the left-right machinery is suggested by the lateralization of the somitogenesis defect, which almost always occurs on the same side of the RA-deficient embryos (16). Although the mechanisms involved in the breaking of left-right symmetry may differ between chick and mammal embryos (24), a common output is the generation and amplification of asymmetric signaling cascades in the mesoderm. As the RA signal progresses symmetrically along the left and right PSM of wild-type embryos (Fig. 1), it is likely to counteract the effect of left-right asymmetric signals in order to stabilize the progression of the molecular oscillations and thereby protect the bilateral synchrony of somite formation.

Supporting Online Material

www.sciencemag.org/cgi/content/full/1108363/DC1

Materials and Methods

SOM Text

Figs. S1 to S5

References and Notes

View Abstract

Navigate This Article