Hes7 3′UTR is required for somite segmentation function

A set of genes in the posterior end of developing mouse embryos shows oscillatory expression, thereby regulating periodic somite segmentation. Although the mechanism for generating oscillation has extensively been clarified, what regulates the oscillation period is still unclear. We attempted to elongate the oscillation period by increasing the time to transcribe Hes7 in this research. We generated knock-in mice, in which a large intron was inserted into Hes7 3′UTR. The exogenous intron was unexpectedly not properly spliced out and the transcripts were prematurely terminated. Consequently, Hes7 mRNA lost its 3′UTR, thereby reducing the amount of Hes7 protein. Oscillation was damped in the knock-in embryos and periodic somite segmentation does not occur properly. Thus, we demonstrated that Hes7 3′UTR is essential to accumulate adequate amounts of Hes7 protein for the somite segmentation clock that orchestrates periodic somite formation.

A set of genes in the posterior end of developing mouse embryos shows oscillatory expression, thereby regulating periodic somite segmentation. Although the mechanism for generating oscillation has extensively been clarified, what regulates the oscillation period is still unclear. We attempted to elongate the oscillation period by increasing the time to transcribe Hes7 in this research. We generated knock-in mice, in which a large intron was inserted into Hes7 39UTR. The exogenous intron was unexpectedly not properly spliced out and the transcripts were prematurely terminated. Consequently, Hes7 mRNA lost its 39UTR, thereby reducing the amount of Hes7 protein. Oscillation was damped in the knock-in embryos and periodic somite segmentation does not occur properly. Thus, we demonstrated that Hes7 39UTR is essential to accumulate adequate amounts of Hes7 protein for the somite segmentation clock that orchestrates periodic somite formation. S omitogenesis is the most prominent cyclic event in vertebrate development. Somites are sequentially formed in the anterior-posterior direction. A pair of somites buds off from the anterior part of the presomitic mesoderm (PSM), i.e., the unsegmented mesenchymal tissue in the posterior embryo of vertebrates, in a rhythmic fashion. Due to this temporal periodicity, the array of somites is formed as a spatial periodic pattern. Somites give rise to vertebrae, ribs, skeletal muscles, and skin, and thus, somites are the earliest segmental units of the vertebrate body 1,2 . The period of somite formation is characteristic of the species, e.g., 2 h in mice, 90 min in chicks, and 20-30 min in zebrafish 3 . Because the timing for the beginning and end of somite formation is strictly determined during development, the number of somites and the number of resulting vertebrae may be dependent on the period of somitogenesis.
The periodicity of somitogenesis is instructed by the synchronous oscillation of gene expression in PSM, which is termed the somite segmentation clock 1 . The expression of a set of genes oscillates in a 2-h cycle in mouse PSM, where expression demonstrates cyclic wave-like propagation from the posterior end of the embryo to the anterior PSM. A pair of somites is generated during each cycles 3 . The oscillating genes in mouse include components of Notch signaling, fibroblast growth factor (FGF) signaling, and Wnt signaling 4 . One of these genes is an effector gene of Notch signaling, Hes7, which encodes a basic helix-loop-helix type transcription factor 5 . Hes7, as a transcriptional repressor, binds to its own promoter, thereby inhibiting its own transcription 6,7 . Thus, Hes7 forms a negative feedback loop, which is the major mechanism for oscillatory gene expression. Hes7 also inhibits the transcription of several target genes; therefore, the expression of these genes oscillates in a synchronized manner 7 . These target genes of Hes7 include lunatic fringe (Lfng), which encodes a glycosyl-transferase that modulates the activity of Notch signaling 8 . The activity of Notch signaling also oscillates in the PSM due to the oscillation of Lfng, and this dynamic Notch activity contributes to the mechanism for somite segmentation 9 .
Mathematical models have been proposed based on this mechanism for the transcriptional regulation that generates oscillatory gene expression in the somite segmentation clock [10][11][12][13] . Of these, a mathematical model that directed the negative feedback loop of a transcription factor with its time delay successfully reproduced oscillatory gene expression 12,13 . The predicted period of oscillation in this model mainly depended on the time delay of the negative feedback loop. The time delay included the time taken to transcribe and process mRNAs, synthesize and modulate proteins, and traffick mRNAs and proteins. One acceptable way to test and verify the model, in which the feedback loop of Hes7 plays a central role, is to manipulate parameters in the model and to detect its resulting phenotype in somitogenesis. For instance, increasing the time for mRNA transcription should increase the time delay, thereby extending the period of gene oscillation and somite segmentation in the model.
We tried to increase the period of gene oscillation and somite segmentation in mouse somitogenesis in this research by increasing the time taken to transcribe Hes7. We inserted a large intron derived from human dystrophin into the 39UTR region of mouse Hes7 in ES cells, and generated knock-in mice. Previous researches have attempted to increase the time delay in the somite segmentation clock by using similar strategies. Stauber et al. inserted large intron sequences derived from human dystrophin into the third intron of mouse Hes7 14 , and Hanisch et al. introduced a her1 transgene that they inserted a large sequence into the second intron into a her1/her7 mutant zebrafish 15 . The large introns were unexpectedly not correctly spliced out in both studies, and both groups failed to increase the time delay in the somite segmentation clock. The exogenous intron was coincidentally not properly spliced out in this study, and thus we also failed to increase the time taken to transcribe Hes7. The transcripts of the knock-in Hes7 allele were prematurely terminated within the intron sequence in the mouse PSM cells. Consequently, Hes7 mRNA lost its 39UTR. Oscillatory gene expression in PSM was lost in the embryos of the knock-in mice and no adequate periodic somite segmentation occurred. Accordingly, the metameric pattern of somites was severely affected, thereby severely disrupting the axial skeletons. This phenotype is similar to that in Hes7 null mutants. In addition, the amount of Hes7 protein was severely reduced in the PSM of the knock-in mice. Thus, Hes7 39UTR was essential to generate sufficient amounts of Hes7 protein.
Altogether, our results demonstrated that Hes7 39UTR is required for the somite segmentation function.

Results
Generation of Knock-in mice. Hes7 protein inhibits its own transcription to form a negative feedback loop, which is supposed to be the core mechanism for gene oscillation 7 . Previous research using mathematical modeling has proposed that the time delay from Hes7 transcription to the accumulation of Hes7 protein is the critical factor in determining the oscillation period 13 . We tried to lengthen the time delay by increasing the size of the Hes7 gene to verify the prediction, and observed whether the oscillation period was extended. We inserted 5, 10, and 20 kb exogenous intron sequences derived from human dystrophin into embryonic stem (ES) cells at the 39UTR of the Hes7 gene by homologous recombination (Fig. 1a: see Methods). The recombinants were confirmed by Southern blotting and polymerase chain reaction (PCR) analyses (Fig. 1b,c, data not shown). The transcription elongation rate of RNA polymerase II was previously estimated by several groups, and they worked out different values from 1.1-4.8 kb/min [15][16][17][18][19][20] . If we had taken 1.1 kb/min as the velocity of the RNA polymerase on the Hes7 allele on the one hand, we would have expected that 5, 10, and 20 kb exogenous intron sequences would respectively increase the time delay by 4.5, 9, and 18 min. If we had assumed it to be 4.8 kb/min on the other hand, we would have expected that 5, 10, and 20 kb exogenous intron sequences would respectively increase the time delay by 1, 2.1, and 4.2 min. We carried out a simulation according to our mathematical model to estimate the expected oscillation period with these alterations 21 (Supplementary note). Our model predicted that the increment of time delay led to sustained oscillations, and that 2.1-9 min additional time delay brought by the 10 kb exogenous intron sequence increased the oscillation period by 4.9-20.2 min. The exogenous intron sequences were expected to be spliced out, and consequently only 89-base insertion derived from loxP and the human dystrophin sequence should remain in the 39UTR of the mature transcripts from the mutant Hes7 allele. The resulting Hes7 protein should be exactly the same as that of the wild type.
Mutant mice were generated from the recombinant ES cells. We found that Hes7 5k/5k mice, Hes7 10k/10k mice, and Hes7 20k/20k mice had short trunks and tails (Fig. 1d, data not shown). The short trunks and tails were unexpected phenotype because our mathematical simulation predicted that increasing the time delay would lead to the oscillation period being altered, but that sustained oscillation would still be maintained.
Because the segmental patterns of vertebrae and ribs were derived from the metameric pattern of somites in the embryonic stage, we next examined somite patterning in Hes7 10k/10k embryos. A homeobox gene, Uncx4.1, was exclusively expressed in the posterior half of each somite, thereby displaying an ordered ladder pattern in the wild-type embryos (Fig. 3a). The expression of Uncx4.1 was severely disrupted in the absence of Hes7 (Fig. 3c), as has previously been reported 5 . The signals of Uncx4.1 were fused in Hes7 10k/10k embryos and the ladder pattern was disrupted (Fig. 3b). Another somite marker, myogenin, which was expressed in the myotome in each somite, also exhibited an ordered ladder pattern in wild-type embryos (Fig. 3d). However, the signals of myogenin were fused in some parts in Hes7 10k/10k embryos (Fig. 3e) as same as in Hes7 -/embryos (Fig. 3f). These results indicate that somites are not sufficiently segmented in Hes7 10k/10k embryos, and this anomaly in somites probably leads to segmental defects in neonates.
Amount of Hes7 protein is reduced in Hes7 10k/10k embryos. We detected Hes7 mRNA with a probe derived from the Hes7 coding sequence in wild-type and Hes7 10k/10k embryos at E 10.5 to check the expression patterns of cyclic genes in Hes7 10k/10k embryos. The expression of Hes7 revealed wave patterns in the PSM of wild-type embryo, as has previously been reported 5 (Fig. 4a). In contrast, the expression of Hes7 mRNA spread uniformly in PSM of Hes7 10k/10k embryos, and did not exhibit wave patterns (Fig. 4b).
We speculated that the amount of Hes7 mRNA and/or Hes7 protein in Hes7 10k/10k embryos would be insufficient to maintain sustained oscillation because the transcription of Hes7 is activated through the PSM cells in Hes7 null embryos 7 . Thus, we assessed the amount of Hes7 mRNA and Hes7 protein derived from the Hes7 10k allele in Hes7 10k/10k embryos. We found from quantitative PCR that the amount of Hes7 mRNA was reduced by 30% in the PSM of Hes7 10k/10k embryos in comparison with wild-type embryos (Fig. 5a). We then carried out immunohistochemistry with an anti-Hes7 antibody. The Hes7 protein in wild-type embryos revealed a wave pattern in the PSM (Fig. 4c). In contrast, we could hardly detect Hes7 protein in the PSM of Hes7 10k/10k embryos (Fig. 4d). These results suggest that the amount of Hes7 protein derived from the Hes7 10k allele was grossly reduced in PSM, and that Hes7 protein was not effectively produced in Hes7 10k/10k embryos. This is consistent with the phenotypes of Hes7 10k/10k embryos, which are similar to Hes7 null mutants.
Transcriptional activities of cyclic genes up-regulated by reduced amount of Hes7 protein in Hes7 10k/10k embryos. Because Hes7 cyclically represses its own transcription and transcription of cyclic genes including Lfng, thereby generating synchronized gene oscillation in PSM 7 , we assumed the transcriptional activities of Hes7 and Lfng would be enhanced in knock-in mice. Thus, we first examined the transcriptional activity of the Hes7 10k allele. The transcriptional activity can be evaluated by detecting the intron  fragments in transcripts 22,23 because the intron sequences in transcripts are spliced out immediately after transcription and the intron fragments are degraded immediately after splicing. We performed in situ hybridization with a probe derived from the first intron of Hes7 to detect regions active in the transcription of Hes7. The regions active in the transcription of Hes7 displayed various patterns in the PSM of wild-type embryos, which is consistent with the expression pattern of Hes7 mRNA (Fig. 4e). However, the entire PSM of Hes7 10k/10k embryos was active in the transcription of the Hes7 10k allele (Fig. 4f). We carried out a quantitative reverse transcription-polymerase chain reaction (RT-PCR) to quantitatively measure transcriptional activity. We detected the sequence of the first intron of Hes7 in the transcripts, and found that the transcriptional activity of the Hes7 10k allele in the PSM of Hes7 10k/10k embryos was substantially up-regulated in comparison with that of the Hes7 allele in the PSM of wild-type embryos (Fig. 5b). These results suggest that the transcriptional activity of the Hes7 10k allele is up-regulated in the PSM of Hes7 10k/10k embryos, probably because of reduced amount of Hes7 protein.
We next examined the expression of Lfng, which is one of the target genes of Hes7. As expected, Lfng was uniformly expressed throughout the whole PSM of Hes7 10k/10k embryos (Fig. 4h). We carried out quantitative RT-PCR analyses to quantitatively evaluate Lfng expression. We found that the transcriptional activity of Lfng was clearly up-regulated in the PSM of Hes7 10k/10k embryos in comparison with that in the PSM of wild-type embryos (Fig. 5d). The amount of Lfng mRNA was also slightly increased in the PSM of Hes7 10k/10k embryos (Fig. 5c). These results suggest that the reduced amount of Hes7 enhanced the expression of Lfng, and that sufficient amounts of Hes7 are essential for oscillatory gene expression and somite segmentation.
Hes7 39UTR is essential for somite segmentation clock. The 10 kb intron derived from human dystrophin in the Hes7 10k allele was inserted immediately downstream of the stop codon of Hes7 (Fig. 6a). We carried out RT-PCR with RNAs extracted from the PSMs of wild-type and Hes7 10k/10k embryos at E 10.5 to check whether this exogenous intron was spliced out to form appropriate mRNAs. We designed a forward primer, F1, in the 4th exon, a reverse primer, R1, in 39UTR, and two reverse primers, R2 and R3, in the intron derived from dystrophin (Fig. 6a-c, 7a). We detected Hes7 mRNA in wild-type embryos, whereas we failed to obtain a PCR product in Hes7 10k/10k embryos with the F1 and R1 primers (Fig. 7b). Although a PCR product was detected in Hes7 10k/10k PSMs by combining the F1 and R2 primers, no products were detected by combining the F1 and R3 primers (Fig. 7b). These results suggest that the human dystrophin intron inserted into Hes7 39UTR was not spliced out in Hes7 10k/10k PSMs, and that the transcripts of the Hes7 10k allele terminated between the positions of the R2 and R3 primers.
We carried out in situ hybridization with a probe derived from Hes7 39UTR to further examine the structure of Hes7 mRNA in Hes7 10k/10k embryos. We detected Hes7 mRNA in wild-type embryos, whereas we failed to detect Hes7 mRNA in Hes7 10k/10k embryos (Fig. 8a,b). In contrast, we uniformly detected mRNAs in the PSM of Hes7 10k/10k embryos by using a probe derived from the 59 end of the dystrophin intron (Fig. 8d), and the pattern was similar to the pattern obtained with the probe derived from the coding of Hes7 (Fig. 4b).
No signal was detected with the dystrophin-intron probe in wild-type embryos because this intron was adopted by the human dystrophin (Fig. 8c). These results support our interpretation that the exogenous intron was not spliced out in the Hes7 10k transcripts, which were prematurely terminated between the positions of the R2 and R3 primers.
We carried out 39-rapid amplification of cDNA ends (39-RACE) to identify the positions of the 39 end of the transcripts of Hes7 10k . We detected several PCR products, and two of them were derived from the Hes7 gene (Fig. 8e). One was terminated at a site 2.4 kb downstream of the stop codon. The other was terminated at a site 3.3 kb downstream of the stop codon. Because the 39 ends of both products were located between the positions of the R2 and R3 primers, this was consistent with the results obtained from RT-PCR analysis (Fig. 6a,b). We found polyadenylation site-like sequences (AAT-AAA) just upstream of the 39 ends of both transcripts (Fig. 6c,8e). Thus, these results suggest that sequences similar to the polyadenylation site in the exogenous human dystrophin intron were misidentified, thereby leading to premature termination and adding polyadenylation tails to the Hes7 10k transcripts. Hes7 39UTR was essential to generate sufficient amounts of Hes7 protein to maintain the somite segmentation clock, taking all results into account.

Discussion
We manipulated the mouse Hes7 gene in ES cells, with which we generated lines of knock-in mice. The proper 39UTR of Hes7 mRNA was replaced by an exogenous sequence derived from the human dystrophin intron by modifying the Hes7 gene. This alteration of 39UTR reduced the amount of Hes7 mRNA and Hes7 protein, leading to gene oscillation in PSM being damped and to the failure of periodic somite formation. We therefore concluded that Hes7 played the major role in generating gene oscillation in PSM, where Hes7 39UTR was essential to generate sufficient amounts of Hes7 protein.
We originally attempted to increase the time for mRNA transcription by increasing the gene size of Hes7, thereby expanding the oscillation period. We expected that the extended oscillation period would decrease the number of somites and vertebrae and/or increase the size of somites. However, the knock-in mice did not exhibit the expected phenotypes because the inserted introns decreased the amount of Hes7 mRNA and Hes7 protein in PSM, thereby leading to damped oscillation.
Stauber et al. recently independently attempted to elongate time delay in the segmentation clock 14 . They generated knock-in mice, where the size of the third intron of Hes7 was increased. However, this modification caused incorrect splicing, thereby drastically reducing the activity of Hes7. Consequently, the homozygous knock-in mutant embryos demonstrated similar phenotypes in somites and axial skeletons as the Hes7 knock out embryos. Thus, they failed to increase the time delay in knock-in mice. Hanisch et al., tried to rescue the mutant of her1/her7 in zebrafish, which were the orthlogs of Hes7, by introducing a her1 transgene that inserted a sequence derived from human dystrophin into its second intron 15 . Unfortunately, they failed to rescue the mutants with the transgene because the modified intron was not correctly spliced out.
Nevertheless, they measured the velocity of RNA polymerase II on the allele by using the transgene, and demonstrated that it was 4.8 kb/min, which was much faster than they expected. According to the results, they discussed that the contribution of time taken for transcription to the time delay in the segmentation clock was much shorter than that of time taken for other steps including the splicing and export of mRNA. Hoyle and Ish-Horowicz measured the time to splice and export the mRNA of cyclic genes in mouse, chick and zebrafish PSM 24 . Although they tried to estimate the time to transcribe the genes in vivo, they failed because the velocity of transcription was too fast. The time delay in the feedback loop of Hes7 contains the time for several steps including transcription, mRNA processing, protein translation, protein modification, and mRNA and protein trafficking. These reports concluded that the contribution of time in these steps taken for transcription to the time delay was much smaller than that of time taken for other steps 15,24 . However, it is quantitatively difficult to alter the time for these steps because their kinetics is not well understood. Thus, although three   [15][16][17][18][19][20] . Thus, the time to transcribe introns should be less than 2 min. Most of the time delay created by introns is probably due to the process of splicing. They found that gene oscillation in mutant mice was damped, and that segmentation was severely affected. Thus, they concluded that the time delay derived from introns was essential for gene oscillation in the segmentation clock 25 . The same group also found that deleting two introns from Hes7, which contains three introns, reduced the time delay by 5 min, and that the deletion increased the number of somites and vertebrae in the rostral part, whereas somite segmentation was affected in the caudal part 26 . They also confirmed a shorter period of gene oscillation and somite segmentation in vivo. Their results are consistent with predictions by the mathematical model in which severe reduction in time delay abolishes gene oscillation, whereas mild reduction in time delay leads to more rapid but dampened oscillation 25,26 . Thus, it is likely that the oscillation period depends on the time delay that is generated by splicing, and that it is the mechanism that controls the oscillation period. However, the possibility that splicing or introns per se are essential for a proper oscillation period cannot be denied. Thus, increasing the delay time is essential to clarify the mechanism that controls the period of the segmentation clock.
We could not clarify why the exogenous intron was not properly spliced out and why it prematurely terminated. A possible reason is that the exogenous intron derived from human dystrophin is too long to be spliced out. However, this is not likely because this intron is properly spliced out in the dystrophin gene in humans 17 , and because the shorter intron (5 kb) also failed to be spliced out. Using other introns instead of those derived from human dystrophin may solve the problem.
Nonsense-mediated mRNA decay (NMD) is a well-studied mechanism that selectively eliminates abnormal mRNAs 27 . An intron located downstream of the stop codon is detected in the NMD mechanism, thereby the transcript is rapidly degraded. However, if the distance between the stop codon and the intron is less than 50-55 bp, the transcripts are not degraded by NMD 28 . Therefore, we placed the exogenous intron just downstream of the stop codon of Hes7 to avoid NMD. The transcripts were not subjected to NMD because the exogenous intron in the mutant Hes7 transcript was not spliced out in the Hes7 10k/10k embryo. Thus, the small reduction of Hes7 mRNA in the Hes7 10k/10k embryo is not likely due to NMD. Total RNA was extracted from the PSMs of wild-type or Hes7 10k/10k embryos, and was analyzed by RT-PCR with the combinations of indicated primers (panels on left). Genomic DNA was extracted from wild-type or Hes7 10k/10k mice. PCR analyses were carried out with combinations of indicated primers (panels on right). Gapdh mRNA was detected as a positive control. The combination of F1 and R1 primers failed to detect the 10 kb band with genome DNA or the mRNA of Hes7 10k/10k mice. The amount of Hes7 protein in the PSM of Hes7 10k/10k embryo was reduced, and this was probably responsible for the phenotype in the mutant embryos. Thus, we concluded that Hes7 39UTR is essential for maintaining the amount of Hes7 protein. We recently demonstrated that the 39UTRs of Hes7 and Lfng were responsible for the differential distribution patterns of their mRNAs in PSM 29 . This is consistent with our findings in this report, which has accumulated evidence on the importance of 39UTR in gene oscillation. We demonstrated in this research that the amount of Hes7 mRNA was reduced by 30% in the PSM of Hes7 10k/10k embryos, whereas the transcriptional activity of Hes7 was slightly up-regulated (Fig. 4). This suggests that Hes7 39UTR maintains the proper stability of Hes7 mRNA. It is not likely that the rate at which Hes7 is degraded is increased in Hes7 10k/10k embryo because the primary structure of Hes7 protein is not altered by mutation in the Hes7 gene. Thus, the efficacy of export from the nucleus to the cytoplasm or translation of Hes7 mRNA may be reduced by abnormal 39UTR. Further studies are needed to clarify the role of Hes7 39UTR.

Methods
Generation of Hes7 10k/10k knock-in mice. We used mutated loxp sequences of lox71, lox2272, and lox66. lox71 and lox2272 are exclusively recombined with lox66 and lox2272, respectively, in the presence of Cre recombinase 30,31 . The targeting vector was constructed by inserting lox71, PGK-neo, PGK-tk, and lox2272 just after the stop codon of Hes7, and it was transfected to embryonic stem (ES) cells (TT2) 32 . Hes7 neo/1 clones were selected by using 250 mg/ml G418 (Nacalai tesque, Japan), and their genotype were checked with Southern blotting. We placed lox66 at the 59 end and lox2272 at the 39 end of the fragments of the human dystrophin gene in the intron vector. We took 11 bp of the 39 end of exon 74, the consequent intron 74 (20 kb), and 10 bp of the 59 end of the exon 75 of the human dystrophin gene 33 . Vectors for the 10kb and 5-kb introns were generated by deleting the middle part of the human dystrophin intron. The intron vectors were transfected into the Hes7 neo/1 clone with the cre expression vector, and the Hes7 20k/1 , Hes7 10k/1 , and Hes7 5k/1 clones were selected by using 0.2 mM of FIAU (Wako, Japan). The genotypes were checked with Southern blotting and PCR. We confirmed Hes7 coding, introns, and the junction of the Hes7 gene and the exogenous fragment by sequencing. We inserted the Hes7 20k/1 , Hes7 10k/1 , and Hes7 5k/1 ES cells into the blastocysts of CD1 mice, and generated chimeric mice, which were crossed with CD1 mice to generate Hes7 20k/1 , Hes7 10k/1 , and Hes7 5k/1 mice. Hes7 20k/20k , Hes7 10k/10k , and Hes7 5k/5k mice were generated by intercrossing. Genotyping was checked with the following primers of Hes7_F GAGCAATGGTCACCCGGGAGCG, Hes7_R TCTGTAAGGCGGTGGCG-GTGGC, and Dys-intron_R GAGCAATGGTCACCCGGGAGCG (Fig. 1b).
Our experiments were approved by the Animal Care Committee of Nara Institute of Science and Technology and conducted in accordance with guidelines that were established by the Science Council of Japan.
Whole-mount immunostaining and whole-mount in situ hybridization. Embryos at E 10.5 were fixed with 4% paraformaldehyde for 3 h for whole-mount immunostaining and they were treated with 1% Triton-X100 (Sigma) and 0.1% H 2 O 2 for 30 min. We incubated the embryos with guinea pig antibodies against Hes7 (diluted 15200) for 12 h at 4uC and then with peroxidase-conjugated antibodies against guinea pig IgG for 12 h at 4uC. We visualized the peroxidase activity by using TSA kit#2 (alexa fluor488). Hoechst33258 was used for nuclear staining, and fluorescent images were collected with LSM710 (Carl Zeiss). Whole-mount in situ hybridization was carried out as previously described 5 . Hes7 nascent transcripts and Hes7 10k nascent transcripts were detected with a probe that corresponds to the first intron of Hes7. Mature transcripts of Hes7 were detected with probes that correspond to the coding or 39UTR of Hes7. Hes7 10k mature transcripts were detected with probes corresponds to the 59 part of intron 74 of human dystrophin. The images were collected with a digital microscope VHX-5000 (Keyence) (Fig. 3) or a stereoscopic microscope SZX16 (Olympus) (Fig. 4,8).
RNA extraction. We extracted total RNA from the PSMs of embryos by using a Nucleo Spin RNA XS kit (Macherey-Nagel). cDNAs were synthesized from total RNAs by using SuperScriptII Reverse Transcriptase (Invitrogen) with random primers according to the manufacturer's instructions.
39-rapid amplification of cDNA ends (39RACE). Total RNAs were extracted from individual Hes7 -/embryos at E 10.5. cDNAs were synthesized from total RNAs by using SuperScriptII Reverse Transcriptase (Invitrogen) with Oligo dT-3sites Adaptor Primer (Invitrogen) according to the manufacturer's instructions. The 39 fragments were amplified with KOD plus (Toyobo) under conditions of 95uC for 5 min, and 32 cycles at 95uC for 30 sec, 60uC for 30 sec, and 72uC for 5 min. We used 1 ml of solution of the RT reaction as a template for the first amplification. We used 1 ml of solution of the first reaction as a template for the first nested reaction, and used 0.5 mM of each primer (39RACE_R1/Adaptor). A second nested reaction was carried out using 1 ml of solution of the first nested reaction (15100 dilution) with 0.5 mM of each primer (39RACE_R2/Adaptor) under the same conditions to ensure specificity. The sequences of the primers were: Oligo dT-3sites Adaptor Primer GGCCAC-GCGTCGACTAGTACT TTTTTTTTTTTTTTTT, Adaptor GGCCACGCGTCG-ACTAGTAC, 39RACE_R1 GATCATTTGTCAGTGAAGACCATGGTAGAG, and 39RACE_R2 TGAGCTCATTGGT CAAGAGCCCTG.
Skeletal preparation. The cartilages of newborn mice were stained with alcian blue and their bones were stained with alizarin red, after being fixed in 95% ethanol 5 .