During somitogenesis, a pair of somites buds off from the presomitic mesoderm every 2 hours in mouse embryos, suggesting that somite segmentation is controlled by a biological clock with a 2-hour cycle1,2,3. Expression of the basic helix-loop-helix factor Hes7, an effector of Notch signaling, follows a 2-hour oscillatory cycle controlled by negative feedback; this is proposed to be the molecular basis for the somite segmentation clock4,5,6. If the proposal is correct, this clock should depend crucially on the short lifetime of Hes7. To address the biological importance of Hes7 instability, we generated mice expressing mutant Hes7 with a longer half-life (∼30 min compared with ∼22 min for wild-type Hes7) but normal repressor activity. In these mice, somite segmentation and oscillatory expression became severely disorganized after a few normal cycles of segmentation. We simulated this effect mathematically using a direct autorepression model. Thus, instability of Hes7 is essential for sustained oscillation and for its function as a segmentation clock.
Hes7 is expressed in a cyclical fashion, and each cycle of Hes7 expression coincides with the generation of each pair of somites. Related basic helix-loop-helix genes and another Notch signaling component, lunatic fringe (Lfng), show the same dynamic expression in vertebrates7,8,9,10,11,12,13,14,15,16,17,18. We previously found that the cyclic expression of Hes7 in the presomitic mesoderm (PSM) could be mimicked by Hes1 in cultured cells19. Serum stimulation triggers an oscillation in Hes1 mRNA and Hes1 protein levels governed by negative feedback: Hes1 protein represses Hes1 transcription but is rapidly degraded by the proteasome system, thereby allowing the next bout of transcription. This negative feedback loop is the molecular basis for the 2-hour oscillation in Hes1 levels in cultured cells19. We recently found that the same mechanism works for Hes7 oscillation in the PSM5. The dynamic expression of Hes7 is thought to depend on its short lifespan: the half-life of Hes7 protein is only ∼22 min, and this short half-life is expected to be required for genesis of the 2-hour periodicity5,20,21,22. To address the importance of the instability of Hes7, we manipulated its half-life.
We first screened for Hes7 mutants with long half-lives and normal repressor activity. Because lysine residues are the typical ubiquitination targets, we focused on seven lysine residues in Hes7 (Fig. 1a). We introduced a lysine-to-arginine point mutation at each lysine residue and created seven mutants (Hes7 K14R, Hes7 K17R, Hes7 K22R, Hes7 K52R, Hes7 K55R, Hes7 K129R and Hes7 K211R). To analyze the transcriptional activity of these Hes7 mutants, we carried out a transient transfection assay with the luciferase reporter gene under the control of the N box–containing promoter. Wild-type Hes7 repressed luciferase gene expression (Fig. 1b). Hes7 K14R, Hes7 K129R and Hes7 K211R also repressed transcription, whereas the other mutants, which have a mutation in the basic helix-loop-helix domain (Supplementary Fig. 1 online), did not repress transcription and were not studied further. We next determined the half-lives of the three Hes7 mutants with normal repressor activity. The protein half-life of wild-type Hes7 was as short as 22.3 min, whereas those of Hes7 K14R, Hes7 K129R and Hes7 K211R were 30.3 min, 21.1 min and 15.6 min, respectively (Fig. 1c). Because of its longer half-life (30.3 min) and normal repressor activity, we chose to study the effect of mutant Hes7 K14R on the mouse segmentation clock.
To examine the role of Hes7 half-life in somitogenesis, we generated knock-in mice expressing Hes7 K14R instead of wild-type Hes7. This was done by introducing an A→G point mutation at nucleotide residue 118 of Hes7 in embryonic stem (ES) cells (Fig. 2a). The neomycin resistance gene was excised by transient expression of Cre recombinase. From these ES cells, we generated Hes7118G/118G knock-in mice, which express Hes7 K14R (Fig. 2b). To examine cyclic gene expression in the PSM, we carried out in situ hybridization with probes for Uncx4.1, Hes7 intron and Lfng and immunostaining with antiserum to Hes7. At the two- or three-somite stage, expression of the homeobox gene Uncx4.1 showed the normal striped pattern in both Hes7118G/118G and Hes7+/+ embryos but was disorganized in Hes7−/− embryos (Fig. 3a–c). Because intron regions are present only in nascent transcripts, in situ hybridization with intron probes identifies only the regions where the gene is actively transcribed5,23. In situ hybridization at the two- or three-somite stage with a Hes7 intron probe showed that Hes7 was dynamically transcribed in Hes7118G/118G and Hes7+/+ embryos, whereas Hes7 transcription was constantly upregulated in Hes7−/− embryos (Fig. 3d–f). In agreement with this observation, we noted oscillatory expression of Hes7 protein in Hes7118G/118G and Hes7+/+ embryos but not in Hes7−/− embryos (Fig. 3g–i). Similarly, we observed cyclic expression of Lfng in Hes7118G/118G and Hes7+/+ embryos, whereas Lfng mRNA was constantly upregulated in Hes7−/− embryos (Fig. 3j–l). Thus, periodic gene expression in Hes7118G/118G embryos occurs normally, indicating that the longer half-life of Hes7 K14R does not disrupt early somite segmentation.
To examine somite segmentation at later stages, we carried out in situ hybridization and immunostaining with embryos at the eight-somite stage. In situ hybridization with the Uncx4.1 probe showed that, in contrast to Hes7+/+ embryos, somite segmentation in Hes7118G/118G embryos was disrupted after three to four somites had been generated, whereas in Hes7−/− embryos, we observed fused somites throughout the entire anterior-posterior axis (Fig. 4a–c). Hes7 transcription was not periodic in either Hes7118G/118G or Hes7−/− embryos, whereas Hes7 was transcribed dynamically in Hes7+/+ embryos (Fig. 4d–f). At the protein level, Hes7 expression was cyclical in Hes7+/+ embryos, continuous at a low level in Hes7118GA118G embryos and absent in Hes7−/− embryos (Fig. 4g–i). In addition, Lfng was expressed at a steady level in both Hes7118G/118G and Hes7−/− embryos but was expressed periodically in Hes7+/+ embryos (Fig. 4j–l). In Hes7118G/118G embryos, the peaks and troughs of expression became ill-defined at the four- to five-somite stage (data not shown). These results show that in Hes7118G/118G embryos, periodic gene expression was maintained in early somitogenesis (one to four somites) but not at later stages. This suggests that the increase of just 8 min in Hes7 half-life disabled the oscillator, although it is also possible that the K14R mutation is hypomorphic and thereby leads to damped oscillation. Hes7+/118G mice had a normal trunk but a kinked tail, indicating that somite segmentation was damped in these embryos, but less rapidly than in Hes7118G/118G embryos (data not shown).
We next examined the relationship between Hes7 and Axin2, a Wnt-driven oscillator24 (Fig. 5a–d). We found that Axin2 expression remained dynamic in Hes7 mutant embryos (Fig. 5e–h). On the other hand, somite segmentation seemed to occur normally in Axin2-null mice (W. Birchmeier, personal communication). These results suggest that Axin2 oscillation and Hes7 oscillation are each capable of occurring independently, but that the Hes7 oscillation is essential for segmentation. Alternatively, these oscillations could be interdependent and the damped oscillation in Hes7118G/118G embryos could be due to uncoupling of Notch- and Wnt-driven oscillators.
To see whether an increase in Hes7 half-life would be expected to disrupt the segmentation clock in the way we observed, we simulated the Hes7 oscillation. The simulation is based on a recent model proposed for the zebrafish somitogenesis oscillator20. In this model, oscillations are generated by direct autorepression by its own protein product, and the transcriptional (Tm) and translational (Tp) delays (the times from initiation to completion of synthesis of an individual mRNA or protein molecule) have crucial roles. For sustained oscillation, the half-lives τm and τp of the mRNA and the protein, respectively, must be short compared with the sum of these delays, Tm + Tp. The period of oscillation is then 2(Tm + Tp + (τm/ln2) + (τp/ln2)), approximately 2(Tm + Tp + 1.4τm + 1.4τp). These conditions impose quite narrow constraints on the delays and lifetimes required to give sustained oscillations with the observed 120-min period. If either of the half-lives is longer than a certain threshold value of somewhat less than Tm + Tp, sustained oscillation becomes impossible, but the system will show damped oscillations if it is started far from its steady state. Assuming that the protein half-life τP = 20 min, we obtained sustained oscillations with a period of ∼120 min (Fig. 6a). To model the lengthening of protein half-life from 20 min to 30 min, we repeated the computation substituting τP = 30 min but without changing other parameters. This, in essence, simulates the effect of the K14R mutation. Oscillations were no longer sustained but were damped and became severely attenuated after three to four cycles (Fig. 6b). The damped oscillations of the model thus mimic well the experimental phenomena that result from slight stabilization of Hes7 protein. Mathematical simulation with the mixed half-lives of 20 min and 30 min for Hes7+/118G also showed damped oscillation but less rapidly than for Hes7118G/118G, mimicking our results in heterozygous embryos (data not shown). Taken together, these simulation results and the Hes7 K14R observations suggest that the half-life of Hes7 protein is crucial for sustained oscillation as a vertebrate segmentation clock.
Generation of Hes7 mutants.
We created the Hes7 mutants by introducing point mutations in hemagglutinin (HA)-tagged Hes7 cDNA to change the lysine residue (AAA or AAG) to the arginine residue (AGA or AGG, respectively) by the Quick Change method (Stratagene). We subcloned HA-tagged wild-type or mutated Hes7 cDNAs into the eukaryotic expression vector containing the cytomegalovirus promoter, pCI-vector (Promega).
We cotransfected C3H10T1/2 cells with the luciferase reporter (0.2 μg) under the control of β-actin promoter with six repeats of N boxes with or without the expression vector for HA-tagged wild-type or mutant Hes7 (1 μg) and plated them in 6-multiwell plates. We cotransfected them with the vector for Renilla luciferase under the control of the SV40 promoter (5 ng, Promega) as an internal standard to normalize the transfection efficiency. After 24–36 h, we collected the cells and measured luciferase activities.
Measurement of protein half-lives.
We transfected C3H10T1/2 cells with expression vectors for HA-tagged wild-type or mutant Hes7. The next day, we exchanged the culture medium for one containing cycloheximide (20 μM, Nacalai Tesque). Thirty minutes after the medium change (time = 0) and every 20 min after that, we collected the cells and subjected them to western blotting with rat monoclonal antibody to HA-peroxidase (3F10, Roche; diluted 1:5,000). We visualized the immunoreactive bands with ECL Plus system (Amersham) and measured their intensities by Typhoon9410 (Amersham).
Generation of Hes7118G/118G knock-in mice.
All animals used in this study were maintained and handled according to protocols approved by Kyoto University. We constructed the targeting vector by replacing exon 1 with mutated exon 1 and by inserting loxP-neo-loxP into intron 1. We identified ES cell lines with the neo allele by PCR and Southern-blot analysis. We then transiently expressed the Cre expression vector by electroporation and identified ES cell lines with the Hes7118G allele. We confirmed the 200-bp promoter and all the exons and introns of the Hes7118G allele in ES cells by sequencing. We generated chimeric mice and bred them with ICR mice. Primer sequences used for PCR are available on request. After 35 cycles of the reaction (94 °C for 30 s, 68 °C for 30 s and 72 °C for 1 min), we separated the products by electrophoresis on 1.5% agarose gels.
Whole-mount immunostaining and whole-mount in situ hybridization.
We carried out whole-mount immunostaining as described previously25. We fixed embryos at embryonic day 8.0 or 8.5 with 4% paraformaldehyde in phosphate-buffered saline at 4 °C for 3 h and treated them with 0.1% H2O2 overnight. We then incubated the embryos with antibody to Hes7 (diluted 1:100) for 5 d at 4 °C and then with peroxidase-conjugated antibody to guinea pig IgG (Chemicon) overnight at 4 °C. We visualized the peroxidase deposits with 4-chloro-1-naphthol. We carried out whole-mount in situ hybridization as described previously4. We detected Hes7 nascent transcripts with the Hes7 intron probe that corresponds to the entire first intron (1.0 kb).
We computed the behavior of the Hes7 oscillator by numerically solving the following equations:
in which p(t) and m(t) are the quantities of functional Hes7 protein and mRNA, respectively, per cell at time t, and f(p) is the rate of initiation of transcription, which depends on the amount of the protein, p, present at the time of initiation. a is the rate constant for translation, and b and c are the degradation rate constants for protein and mRNA, respectively, which are simply related to the half-lives of the molecules:
To represent the case where transcription is inhibited by Hes7 protein, which acts as a dimer, we assume that
where k is the number of molecules of Hes7 mRNA synthesized per unit time in the absence of inhibition and pcrit is the amount of protein that gives half-maximal inhibition. As in the zebrafish model20, we set a = 4.5 protein molecules per mRNA molecule per min, pcrit = 40 molecules per cell, k = 33 mRNA molecules per cell per min, τm = 3 min. We assume that Tm + TP = 37 min and that the protein half-life τP = 20 min for wild-type Hes7 and 30 min for Hes7 K14R.
Note: Supplementary information is available on the Nature Genetics website.
We thank D. Ish-Horowicz and L. Saint-Amant for critical reading of the manuscript. This work was supported by research grants from the Ministry of Education, Culture, Sports, Science and Technology of Japan and the Japan Society for the Promotion of Science and by Cancer Research UK.
Sequence comparison of the bHLH domain of Hes proteins.