In early mouse embryos, the Sip1 (Smad- interacting protein 1) gene, encoding a member of the ZFHX1 family of two-handed zinc finger transcriptional factors (Remacle et al.,1999), is expressed mainly in neuroectoderm, paraxial mesoderm, and neural crest cells. To investigate the function of the Sip1 gene during embryo development, we generated Sip1-null mutant mice by gene targeting (Higashi et al.,2002). Using mutant embryos, we have previously shown that Sip1 is essential for the development of vagal neural crest and migration of cranial neural crest cells (Van de Putte et al.,2003). Homozygous mutant (Sip1−/−) embryos die around E9.5 accompanied by cardio-vascular dysfunction (Van de Putte et al.,2003).
We describe here another important phenotype, namely development of defective somites displaying shortened spatial periodicity. Somite segmentation occurs periodically in embryonic development and produces regularly arranged metameric structures. A generally accepted model of somite segmentation assumes interaction between the two systems, i.e., intracellular oscillation of expression of a set of genes in the paraxial mesoderm that provides the basis for periodical somite segmentation, and a caudally-regressing front of gene activities that are involved in maintaining the unsegmented state of paraxial mesoderm. Evidence indicates that Notch signaling, involving Lunatic Fringe (L-fng) and HES7, steers the former oscillation mechanism, which operates in 2-h cycles in mouse embryos. The rostrally shallower and caudally regressing FGF/WNT signal gradient in the paraxial mesoderm underlies the latter mechanism (Dubrulle et al.,2001; Sawada et al.,2001; Aulehla et al.,2003). Here we show that the rostro-caudal truncation of somites in Sip1−/− embryos is presumably caused by persistent and slow-regressing FGF/WNT signals, which we interpret as slower caudal displacement of the “wavefront” in the current “clock and wavefront” model (Dubrulle et al.,2001; Aulehla et al.,2003).
Sip1 Expression in the Neural Plate and The Paraxial Mesoderm
In gastrulating mouse embryos, Sip1 is expressed in a pair of paraxial zones in the dorsal view (Fig. 1A). This expression pattern is a composite of signals in the neural plate and paraxial mesoderm. The signal in the neural plate is strong rostral to the node, and that in the paraxial mesoderm occurs with spatial periodicity in the rostral boundaries of each of already formed somites and is strongest in the next-forming somite (Fig. 1B, C, note that Fig. 1C shows the same embryo as B after removal of neural plate anterior to the broken line).
Generation of Short Somites in Sip1−/− Embryos
We examined somitogenesis in Sip1−/− mutant embryos during gastrulation at stages earlier than E9, when mutant embryos were still healthy. In Sip1−/− embryos of up to the 7-somite stage, the number of somites was the same as in wild-type embryos. However, the somites were shorter in the rostro-caudal dimension and irregular in the shape of their boundary. This is illustrated by in situ hybridization of Uncx4.1, marking the caudal half of already segmented and the next-forming somites (Fig. 2A, B), and was confirmed in parasagittal sections (Fig. 2C–F). In contrast, the size of somites and unsegmented paraxial mesoderm in the lateral dimension was normal (Fig. 2G–J). Analysis of cell proliferation by BrdU labeling and of apoptosis using TUNEL technique indicated no significant difference with wild-type embryos (data not shown). The length of trunk paraxial mesoderm, measured between the rostral margin of the second somite and the caudal tip of the embryo, was not significantly different among embryos regardless of the Sip1 genotype and of the developmental stages (Fig. 2K), indicating that the caudal extension of the unsegmented paraxial mesoderm was unaffected even after the 7-somite stage. Thus, in Sip1−/− mutant embryos, only the length of the area occupied by the segmented somites was reduced, and this was compensated for by the extension of the unsegmented, more caudal region (Fig. 2L).
Although somite segmentation proceeded up to the 7-somite stage in Sip1−/− embryos, no somite cleavage occurred beyond this stage (Fig. 2M–O, and the schematic presentation in Fig. 2P).
Rostral Expansion of the Region of Gene Oscillation
Lunatic fringe (L-fng) and Hes7 are representative genes that display oscillatory expression in the unsegmented paraxial mesoderm, the most rostral stripe halting at the next-forming somite boundary in mouse embryos (Fig. 3Aa and Ac), and have been documented to participate in somite segmentation (Bessho et al.,2001a; Cole et al.,2002; Morales et al.,2002; Dale et al.,2003).
Dll1 expression has not been considered to oscillate (Hrabe de Angelis et al.,1997), presumably after over-development of the hybridization signals. However, we found that its expression oscillates synchronously with L-fng, as shown by hybridization of the same embryo specimens using L-fng and Dll1 probes (Fig. 3B). This observation is consistent with the oscillatory expression of deltaC, a zebrafish homologue of mouse Dll1 (Jiang et al.,2000).
It is also known that Axin2 is strongly expressed in the caudal end of the embryo and its expression continues to the anterior paraxial mesoderm, rostrally ending with one or two stripes of Axin2 expression in the rostral margins of presumptive somites (Fig. 3Ag) (Aulehla et al.,2003).
In Sip1−/− embryos, the domain of paraxial mesoderm expressing these genes is rostrally expanded. One or two additional stripes of L-fng and Hes7 expression occurred in the unsegmented paraxial mesoderm, with the most rostral stripe reaching nearly to the last-formed segmental boundary (Fig. 3Ab and Ad). In addition to L-fng and Hes7, the striped expression domain of Dll1 also showed anterior extension (Fig. 3Af). Axin2 expression was also rostrally expanded, similar to other oscillatory genes, and a stripe of Axin2 expression occurred just caudal to the most recently formed segmentation boundary (Fig. 3Ah).
Rostral Expansion of Fgf8 Expression in the Sip1−/− Embryos
The above observations suggested that the unsegmented region of paraxial mesoderm was expanded rostrally. Several genes, including Fgf8, are known to be involved in the maintenance of unsegmented state. At various stages of somitogenesis, Fgf8 mRNA was detected in the unsegmented region more rostrally extended in Sip1−/− embryos than normal, as shown using an exon-specific probe (Fig. 4A–D). Interestingly, when an intron-specific probe was used to detect nascent transcripts, its expression domain was identical between Sip1−/− and wild-type embryos (Fig. 4E,F). This suggests that rostral expansion of Fgf8 expression is not caused by expansion of the cell population actively transcribing Fgf8 but rather results from extended stabilization of Fgf8 transcripts in the paraxial mesoderm of Sip1−/− embryos. In addition, the expression of Wnt3a, Dll3, and Tbx6 was also extended rostrally (Fig. 4G–L). Therefore, Sip1 deficiency caused general rostral expansion of the expression domain of the genes involved in the maintenance of the unsegmented state of paraxial mesoderm.
Raldh2 Expression in Sip1−/− Embryos
It is known that Fgf8 and retinoic acid synthesis mediated by Raldh2 (retinaldehyde dehydrogenase 2) gene are mutually repressive, and retinoic acid affects Fgf8 mRNA distribution in the paraxial mesoderm (Diez del Corral et al.,2003). We thus examined whether the Raldh2 expression in the paraxial mesoderm is altered in Sip1−/− embryos, using in situ hybridization of Raldh2 transcripts. In wild-type embryos, the domain of strong Raldh2 expression in paraxial mesoderm starts from the rostral region and terminates at the position one somite length more caudal to the last somite cleavage, namely in the position rostral to the block of the strongest Sip1 expression (Figs. 5A and 6A). In the Sip1−/− embryos, the boundary of strong Raldh2 expression was shifted rostrally, in a way complementary to rostral extension of Fgf8 mRNA distribution (Figs. 5B and 6A). This raises the possibility that Sip1 activity is involved in modulating the mutually repressive Raldh2-Fgf8 interaction.
Based on our observations in this study, we find it likely that the rostral expression of Fgf8 regresses more slowly in a caudal direction in Sip1−/− mutants than in normal embryos. According to the “clock and wavefront” model (Cooke and Zeeman,1976), a slower caudal regression of the “wavefront” indeed results in shorter inter-segmental distance. We find that (1) the rostral expansion of expression of Fgf8, which is involved in the maintenance of unsegmented state of paraxial mesoderm (Fig. 4), (2) the rostral extension of the domain of oscillatory gene expression (Fig. 3A), and (3) the shorter somites produced in Sip1−/− embryo (Fig. 2) are consistent with this model.
As Sip1 is expressed both in the neural and mesodermal tissue (Fig. 1), it is formally possible that Sip1 expressed in the neural plate determines the rostral limit of Fgf8 mRNA distribution in the paraxial mesoderm. However, since the somite segmentation proceeds normally in the absence of the neural plate, as demonstrated using chicken embryos (Chapman et al.,1996), it is more likely that Sip1 expressed in the paraxial mesoderm determines the rostral limit of Fgf8 mRNA distribution, which is indicative of “wavefront” activity in the paraxial mesoderm (Fig. 6B).
It is known that expression of Raldh2, leading to retinoic acid synthesis, and distribution of Fgf8 mRNA are mutually exclusive (Diez del Corral et al.,2003). In Sip1−/− mutant embryos the boundary of strong Raldh2 expression and Fgf8 mRNA was rostrally shifted (Fig. 5), in support of this interaction. Thus, two alternative models for involvement of SIP1 function in regulation of somite segmentation can be considered, either SIP1 regulating Fgf8 mRNA or SIP1 primarily regulating Raldh2 expression (Fig. 6B). In either model, Fgf8 mRNA is regulated for its turnover, as Sip1−/− genotype does not affect Fgf8 transcription per se (Fig. 4E, F), consistent with the previous notion that steepness of Fgf8 mRNA gradient in the paraxial mesoderm is determined by regulation of mRNA stability (Dubrulle and Pourquie,2004). It is possible that SIP1 augments the activity of machinery for Fgf8 mRNA degradation through its transcriptional regulation (Model 1, Fig. 6B). Alternatively, SIP1 may augment Raldh2 expression at the Raldh2/Fgf8 boundary, leading to promotion of Fgf8 mRNA degradation (Model 2, Fig. 6B). Overall, our observations on the defects of somitogenesis in Sip1−/− embryos suggest that SIP1 activity is involved in the positioning of somite boundaries by modulating the Raldh2/Fgf8 interaction boundary in the paraxial mesoderm.
Sip1 +/− heterozygous mice (Higashi et al.,2002) were maintained in ICR background and mated to obtain homozygous (−/−) embryos. The mice and embryos were handled according to the guidelines of the Committee for Animal Experiments, Osaka University, in the research project approved by the committee (approval number 05-025).
Plasmid and In Situ Hybridization
Whole-mount in situ hybridization was done using digoxigenin-labeled probes and color development using NBT (Nitro blue tetrazolium chloride)/BCIP (5-Bromo-4-chloro-3-indolyl phosphate, toluidine salt) as described by Uchikawa et al. (1999), except that 1% SDS was used instead of CHAPS. A full-length Sip1 cDNA sequence was used for preparing the antisense probe. Other probes were originally gifts from A. Mansouri (Uncx4.1; Mansouri et al.,1997), R. Johnston (L-fng; Johnston et al.,1997), R. Kageyama (Hes7; Bessho et al.,2001b), A. Gossler (Dll1; Bettenhausen et al.,1995), P. H. Crossley (Fgf8; Crossley and Martin,1995), S. Takada (Wnt3a; Takada et al.,1994), R. Beddington (Dll3; Dunwoodie et al.,1997), and D. Chapman (Tbx6; Chapman et al.,1996). Fgf8 intron probe was prepared according to Dubrulle and Pourquie (2004).
For detection of Dll1 mRNA, color development with solution of NBT/BCIP was performed at 4°C overnight, rather than at room temperature, in order to avoid over-development of the hybridization signal. For sequential detection of L-fng and Dll1 in the same embryo specimens, the probes were labeled by fluorescein and digoxigenin, respectively, and color development using alkaline phosphatase-conjugated anti-fulorescein and INT (2-[4-lodophenyl]-3- [4-nitrophenyl]-5-phenyl-tetrazolium chloride)/BCIP was done first, followed by photographic recording and decolorization in methanol. Then Dll1 mRNA was detected as described above.
Measurement of the Length of Paraxial Mesoderm
The length from the rostral end of the second somites to the caudal end of embryo, as well as that of segmented domain, was determined as a trajectory through the dorso-ventral midpoints of paraxial mesoderm, using printout of lateral embryonic images. More than eight embryos were examined for each somite stage.
We appreciate stimulating discussions with Drs. O. Pourquie and Y. Takahashi and members of the Kondoh and Huylebroeck laboratories. We thank Drs. H. Hamada and Y. Saga for provision of in situ hybridization probes. M.M. was a Research Fellow of the Japan Society for the Promotion of Sciences.