Oscillatory expression of Hes1 regulates cell proliferation and neuronal differentiation in the embryonic brain

Many gene activities are oscillatory, and oscillatory expression regulates cellular activities (Levine et al., 2013; Purvis and Lahav, 2013; Isomura and Kageyama, 2014; Johnson and Toettcher, 2019). For example, the transcription factor NF-κB exhibits nuclear-cytoplasmic shuttling upon activation of this pathway and induces downstream gene expression differently depending on the shuttling frequencies (Hoffmann et al., 2002; Nelson et al., 2004; Ashall et al., 2009). Similarly, phosphorylated ERK (pERK) levels are pulsatile upon activation, and pulsatile but not sustained induction of pERK is important for cell proliferation rates (Albeck et al., 2013; Aoki et al., 2013). Another example is the somite segmentation clock gene Hes7, which exhibits oscillatory expression with ∼2-h periodicity in the mouse presomitic mesoderm; each cycle of Hes7 oscillation leads to formation of a bilateral pair of somites (Bessho et al., 2001; Pourquié, 2011; Oates et al., 2012). Hes7 represses its own expression by directly binding to the Hes7 promoter, resulting in its oscillatory expression (Bessho et al., 2003). When Hes7 expression is absent or sustained, all somites are severely fused (Bessho et al., 2001; Takashima et al., 2011), and an increase in the frequency of Hes7 oscillations accelerates somite formation (Harima et al., 2013), indicating that Hes7 oscillations regulate the pace of somitogenesis. However, the significance of oscillatory expression in the development of other tissues still remains to be analyzed.

The transcriptional repressor Hes1 is expressed in an oscillatory manner by negative feedback, similar to Hes7 oscillations, in many cell types, including neural progenitor cells (NPCs) (Jouve et al., 2000; Hirata et al., 2002; Shimojo et al., 2008; Bonev et al., 2012; Imayoshi et al., 2013; Chen et al., 2017). In the absence of Hes1 and its related genes, NPCs are not properly maintained and prematurely differentiate into neurons without sufficient cell proliferation, resulting in microcephaly or anencephaly (Ishibashi et al., 1995; Hatakeyama et al., 2004; Sueda et al., 2019). However, sustained overexpression of Hes1 also inhibits proliferation of NPCs and enhances their quiescence (Baek et al., 2006), raising the possibility that oscillatory Hes1 expression is important for proliferation of NPCs. Hes1 oscillations periodically repress the expression of proneural genes, such as Ascl1 and Neurog2, thereby driving their oscillatory expression in NPCs (Shimojo et al., 2008; Imayoshi et al., 2013). By contrast, in differentiating neurons, Hes1 expression disappears, and Ascl1 or Neurog2 expression becomes sustained, indicating that the expression dynamics of proneural genes are different between NPCs and differentiating neurons (Shimojo et al., 2008; Imayoshi et al., 2013). A previous study demonstrated that Ascl1 has dual opposing functions (Castro et al., 2011), and optogenetic analysis showed that Ascl1 induces neuronal differentiation when its expression is sustained, but activates the proliferation of NPCs when its expression is oscillatory (Imayoshi et al., 2013). These results suggest that the oscillatory expression of proneural factors, which normally depends on Hes1 oscillations, is important for efficient proliferation of NPCs (Imayoshi and Kageyama, 2014).

Despite the above findings, the significance of Hes1 oscillations in development is still obscure. Although sustained overexpression of Hes1 inhibits proliferation of NPCs, it is not clear which is more responsible for the inhibition of NPC proliferation – sustained (non-oscillatory) or high levels of Hes1 expression. To address this issue, it is important to examine whether dampened Hes1 oscillations, in the absence of increased expression levels, affect developmental processes. Mathematical modeling suggested that oscillatory expression is controlled by negative feedback with a delayed timing, which depends on transcriptional delays (the time required for production of mature mRNAs), and that changing such delays alters the oscillatory dynamics in the absence of increased expression levels (Lewis, 2003; Monk, 2003). We previously found that an important part of transcriptional delays is intronic delays (the time required for transcription and splicing of intronic sequences) (Takashima et al., 2011). Deletion of all three introns from the Hes7 locus shortens the intronic delays, and under such a condition, Hes7 oscillation is dampened without an increase in its expression levels (Takashima et al., 2011). Similarly, removal of all intronic sequences from the Delta-like1 (Dll1) gene produced a shorter Dll1 gene and decreased its transcriptional delay. By contrast, insertion of Dll1 cDNA into the first Dll1 exon while maintaining all exon and intron sequences produced a longer Dll1 gene and increased its transcriptional delay (Shimojo et al., 2016), suggesting that the gene length also affects the transcriptional delay.

Therefore, to change the transcriptional delays, we decided to generate two types of Hes1 mutant mice: a shorter version of Hes1 with all three introns removed (type-1 mutant), and a longer version of Hes1, in which Hes1 cDNA was inserted between the 5′UTR and the coding sequence (type-2 mutant). We then examined Hes1 expression dynamics and developmental defects in these mutant mice, focusing on neural development.

To examine the significance of Hes1 oscillation in development, we tried to alter the Hes1 expression dynamics. To this end, we first performed numerical simulations, using the same equations and parameter values to those of Hes7 oscillations, because Hes1 has similar oscillation dynamics to Hes7 (Hirata et al., 2004; Takashima et al., 2011). This simulation suggested that altering the transcriptional delay (T_m; see Materials and Methods) would change the dynamics of Hes1 oscillations (Fig. S1). When the transcriptional delay was shortened compared with that of the wild type (Fig. S1A), oscillations were accelerated (faster oscillations) but dampened (Fig. S1B,C). On the other hand, when the transcriptional delay was lengthened, oscillations with a longer period continued (slower oscillations; Fig. S1D). We previously showed that the intronic delay and the gene length affect the transcriptional delay (Takashima et al., 2011; Harima et al., 2013; Shimojo et al., 2016). Therefore, we generated two types of Hes1 mutants, a shorter version of Hes1 with all three introns removed (type-1 mutant; Fig. 1A) and a longer version of Hes1 in which the coding sequence with a stop codon of Hes1 cDNA was inserted between the 5′UTR and the coding sequence in the first exon (type-2 mutant; Fig. 1B).

To determine whether type-1 and type-2 mutations change the transcriptional delay of Hes1 expression compared with the wild-type control, we generated three reporters – the wild-type Hes1 gene (control), the intronless Hes1 gene (type-1 mutant) and the Hes1 gene with Hes1 cDNA knocked-in into the first exon (type-2 mutant) – each of which contained Luc2 cDNA fused in-frame at the 5′ end of the coding region (Fig. S2A). We introduced each reporter into C2C12 myoblast cells and neural stem cells (NSCs) and activated the Hes1 promoter with serum treatment to measure the expression kinetics of the Hes1 reporters. Expression of the type-1 mutant reporter occurred ∼13.5 min faster than the control in C2C12 cells and 3-12 min faster than the control in NSCs (Fig. S2B,C). By contrast, expression of the type-2 mutant reporter was ∼5 min slower in C2C12 cells than the control but did not exhibit a significant difference in NSCs compared with the control (Fig. S2B,C), implying that the transcriptional delay was shorter in the Hes1 type-1 mutant but longer or the same in the type-2 mutant cells compared with the control.

To generate Hes1 type-1 mutant mice, a targeting vector containing Hes1 cDNA and a neomycin expression cassette (pPGK-Neo-pA) was prepared (Fig. 1A). To generate Hes1 type-2 mutant mice, Hes1 cDNA and pPGK-Neo-pA were inserted between the 5′UTR and the coding region of the first exon (Fig. 1B). Each targeting vector was introduced into mouse embryonic stem cells (ESCs) and homologous recombinants were obtained. These recombinants were used to make chimeric mice, which were then crossed with mice ubiquitously expressing flippase (FLP) to remove the neomycin expression cassette (Fig. 1A,B). Heterozygous mutant mice derived from the recombinant ESCs were born normally, and from these mice we generated homozygous mice (Hes1 type-1 and type-2 mutant mice). Both Hes1 type-1 and type-2 mutant mice were born at the Mendelian ratio and grew to fertile adults. We also confirmed that the expression continued in the downstream region of Hes1 type-2 mutant NPCs (Fig. 1C).

To examine the Hes1 expression dynamics of Hes1 type-1 and type-2 mutants, we analyzed time-lapse imaging of the Hes1 promoter-driven destabilized luciferase reporter (pHes1-Ub-luc; Fig. 2A), which monitors the endogenous Hes1 expression (Fig. S3) (Masamizu et al., 2006). As previously described, the Hes1 reporter exhibited oscillatory expression with a period of 173.5±4.4 min in NPCs derived from wild-type mice (Fig. 2B,E,H,K,O, Fig. S4A). Hes1 expression also oscillated in both Hes1 type-1 and type-2 mutant NPCs (Fig. 2C,D,F,G,I,J,L,M, Fig. S4B,C). However, in Hes1 type-1 mutant NPCs, the amplitude of the oscillation was much smaller (about 37% of the WT; Fig. 2N), and the period was slightly shorter (159.9±2.6 min; Fig. 2O). In Hes1 type-2 mutant NPCs, the oscillation amplitude was smaller (about 62% of the wild type; Fig. 2N), but the period was slightly longer (187.0±4.3 min; Fig. 2O). These results indicated that Hes1 oscillations were severely dampened in type-1 mutant NPCs, whereas they were less affected in type-2 mutant NPCs. We previously found that Hes1 oscillations in NPC cultures are very similar to those in brain slices (Shimojo et al., 2008), suggesting that Hes1 oscillations are also affected in the intact brain of both types of mutants. In the type-2 mutant embryos, we did not find any significant defects in the developing nervous system (Fig. S5), but the type-1 mutant embryos displayed defects in neural development. Thus, we decided to examine the type-1 mutant mice in more detail.

Analyses of anti-Hes1 immunohistochemical staining in the developing nervous system demonstrated that Hes1 protein expression was variable and displayed a ‘salt-and-pepper’ pattern in wild-type NPCs (Fig. 3A-C), which might reflect oscillatory expression. In individual cells, there was a wide range of Hes1 protein levels in the wild-type cortex (Fig. 3B,C). By contrast, in Hes1 type-1 homozygous mutant NPCs, Hes1 protein levels were less variable (Fig. 3A-C) and average levels were slightly reduced (Fig. 3D). This reduction in the average level was also confirmed by western blot analysis (Fig. 3E). Thus, Hes1 type-1 mutant NPCs, in which the transcriptional delay of Hes1 expression was shortened, exhibited dampened oscillations without an increase in expression levels.

In situ hybridization analysis indicated that Hes1 mRNA was slightly downregulated in Hes1 type-1 homozygous mutant embryos, compared with the wild type (Fig. 3F). Quantification analysis revealed that Hes1 mRNA levels were ∼40% lower at embryonic day (E) 12.5 in the Hes1 type-1 mutant brain than in the wild type (Fig. 3H). By contrast, Hes3 and Hes5 mRNAs were slightly upregulated in the Hes1 type-1 homozygous mutant brain at E12.5, compared with the wild type (Fig. 3G,H), suggesting that Hes3 and Hes5 upregulation may compensate for Hes1 downregulation in the Hes1 type-1 mutant brain at E12.5. However, at E14.5, although the Notch ligand gene Dll1 was slightly upregulated, Hes1, Hes3 and Hes5 mRNA levels were not significantly different in Hes1 type-1 mutant compared with wild-type embryos (Fig. 3H), suggesting that Hes gene expression became normalized at later stages of development.

To explore the possible effects of altered Hes1 expression dynamics on development, we first examined the neural development of Hes1 type-1 mutant mice. Hes1 type-1 mutant mice had smaller whole bodies than wild type at E12.5 but recovered afterwards (Fig. 4A,B). The brains of Hes1 type-1 mutant mice were also smaller than those of the wild type at E12.5 and E16.5 (Fig. 4A,C). Immunohistological and in situ hybridization analyses showed that the proportions of the nestin⁺ and Pax6⁺ cells in the ventricular zone, Tbr2 (Eomes)⁺ intermediate progenitor cells, and Tuj1 (Tubb3)⁺ neurons were not significantly different between Hes1 type-1 homozygous mutant and wild-type embryos (Fig. 4D-I). Although expression of the proneural gene Neurog2 was not significantly different, the proneural gene Ascl1 was slightly upregulated in Hes1 type-1 homozygous mutant embryos compared with the wild type at E12.5 (Fig. 4J-L). There were no significant differences in the proportions of Cux1⁺ cells in cortical layers 2-4, Ctip2 (Bcl11b)⁺ cells in layer 5, or Tbr1⁺ cells in layer 6 between Hes1 type-1 homozygous mutant and wild-type embryos (Fig. 4M-O). Furthermore, differentiation of inhibitory neurons [GAD65 (GAD2)⁺, GABA⁺] was not affected in the lateral and medial ganglionic eminence (Fig. 4P,Q). We also compared the proliferation and neuronal differentiation of NPC cultures prepared from E12.5 wild-type and Hes1 type-1 homozygous mutant embryos but did not find any significant differences (Fig. S6). Together, these results indicated that despite a smaller brain size and upregulation of Ascl1, neural development proceeds almost normally in Hes1 type-1 homozygous mutant embryos.

We next examined whether cell proliferation and death are affected in Hes1 type-1 homozygous mutant embryos. The number of phospho-histone H3 (pH3)-positive mitotic cells in the apical (ventricular) but not basal region of the Hes1 type-1 homozygous mutant brain was slightly lower than in the wild type at E12.5, suggesting that proliferation of NPCs decreased in the mutants (Fig. 5A-D). Furthermore, the number of cleaved caspase3⁺ apoptotic cells was slightly higher in the cortical layers (Fig. 5A,E) but not significantly different in the ganglionic eminence (Fig. 5F) of Hes1 type-1 homozygous mutant brain at E14.5 compared with the wild type. These results suggested that decreased cell proliferation and increased apoptosis led to the smaller brains of Hes1 type-1 homozygous mutant embryos.

In Hes1 type-1 homozygous mutant mice, both Hes1 protein and Hes1 mRNA levels were lower at E10.5-E12.5 (Fig. 3D,E,H), and therefore the defects observed in these mutant mice might be due to lower levels of Hes1 expression. To test this possibility, we also examined Hes1^+/− mice, which had ∼60% and ∼30% lower levels of Hes1 mRNA (Fig. S7C) and Hes1 protein (Fig. S7F), respectively. These mice expressed a lower level of Hes1 mRNA and a similar level of Hes1 protein compared with Hes1 type-1 homozygous mutant mice (see Fig. 3D,E,H). The analysis of Hes1 expression dynamics monitored with pHes1-Ub-luc revealed a similar period but a significant reduction in the amplitude of Hes1 oscillations (Fig. S7A-C,E). However, this reduction in the amplitude is proportional to the reduction in the mean expression value, and the ratio of the amplitude to the mean of Hes1 oscillations in Hes1^+/− NPCs was similar to that in Hes1 type-2 homozygous mutant but higher than in Hes1 type-1 homozygous mutant NPCs (Fig. S7D). In Hes1^+/− mice, we did not find any significant defects in body and brain sizes, proliferation of NPCs, apoptosis, or neuronal differentiation (Fig. S8A,B,E). These results suggest that the defects observed in Hes1 type-1 homozygous mutant mice were due to dampened Hes1 oscillations rather than to lower levels of Hes1 expression.

The defects observed in the Hes1 type-1 homozygous mutant brains were mild, suggesting that Hes1-related genes such as Hes5, levels of which are also fluctuating (Imayoshi et al., 2013; Manning et al., 2019), may compensate for the Hes1 abnormality. Therefore, we next examined Hes1 type-1-mutant;Hes5-null embryos. Although neural development is slightly delayed in Hes5-null embryos compared with the wild type (Bansod et al., 2017), the size and the final morphologies of Hes5-null mice are normal (Ohtsuka et al., 1999); therefore, we used Hes5-null mice as controls. The brains of Hes1 type-1-mutant;Hes5-null embryos were significantly smaller than those of Hes5-null embryos at E12.5 (Fig. 6A,B). They also tended to be smaller than Hes5-null embryos at later stages but this was not statistically significant (Fig. 6A,B). Immunochemical analyses showed that there were more Tuj1⁺ neurons at E14.5 (Fig. 6C,D), and the number of Tbr2⁺ intermediate progenitors increased at E12.5 in Hes1 type-1-mutant;Hes5-null embryos, compared with Hes5-null embryos (Fig. 6E,F). Furthermore, there were more Ctip2⁺ layer 5 neurons at E14.5 in Hes1 type-1-mutant;Hes5-null embryos compared with Hes5-null embryos (Fig. 6G,H), although at postnatal day (P) 0 Ctip2 expression was not significantly different (Fig. 6I,J). In addition, there were more Cux1⁺ layers 2-4 neurons in Hes1 type-1-mutant;Hes5-null embryos than in Hes5-null embryos at P0 (Fig. 6I,J). Together, these results suggested that neurogenesis was accelerated in Hes1 type-1-mutant;Hes5-null embryos compared with Hes5-null embryos.

We also examined Hes1^+/−;Hes5^−/− mice. There were no significant differences in body and brain sizes (Fig. S8C,D), neuronal differentiation (Fig. S8F-I), proliferation (Fig. S8J-M) or apoptosis (Fig. S8J,N) between Hes1^+/−;Hes5^−/− and Hes5^−/− mice, suggesting that neural development proceeds almost normally in Hes1^+/−;Hes5^−/− mice, and that the defects observed in Hes1 type-1-mutant;Hes5-null embryos are due to dampened Hes1 oscillations and not to decreased Hes1 expression levels.

Because the defects observed in Hes1 type-1-mutant;Hes5-null embryos could be still compensated by Hes3, we next examined the Hes1 type-1 mutant in the Hes3^−/−;Hes5^−/− background. We previously showed that Hes3^−/−;Hes5^−/− mice were mostly normal (Hatakeyama et al., 2004) and therefore used them as controls. The size of the telencephalon was significantly smaller in Hes1 type-1-mutant;Hes3;Hes5-double-null embryos than in the controls at E12.5 (Fig. 7A-D). Furthermore, neurogenesis was increased in the ventral telencephalon of Hes1 type-1-mutant;Hes3;Hes5-double-null embryos compared with the controls (Fig. 7C,E), but apoptosis was not significantly affected (Fig. 7F,G). Thus, Hes1 type-1 mutation led to microcephaly in the Hes3^−/−;Hes5^−/− background, indicating that robust Hes1 oscillations are required for maintenance and proliferation of NPCs and the normal brain morphogenesis.

In this study, to examine the significance of Hes1 oscillations in NPCs, we made two types of Hes1-mutant mice that altered Hes1 expression dynamics. In Hes1 type-1 homozygous mutant embryos, Hes1 oscillations were dampened, and the maintenance and proliferation of NPCs were reduced. The average levels of Hes1 expression in these mutant NPCs were about 60-70% of those in wild-type NPCs but slightly higher than (mRNA level) or similar to (protein level) those in Hes1 heterozygous mutant mice, which exhibit no apparent defects. Together, these findings suggest that the defects observed in Hes1 type-1 mutant mice are due to dampened Hes1 oscillations rather than to decreased levels of Hes1. It was previously shown that Hes1 oscillations periodically repress the expression of the proneural gene Ascl1, driving Ascl1 oscillations, and that optogenetically induced Ascl1 oscillations promote proliferation of NPCs (Imayoshi et al., 2013). Furthermore, it was shown that sustained, high levels of Hes1, which suppress Ascl1 expression, promote quiescence in NPCs (Sueda et al., 2019). These results indicated that robust Hes1 oscillations and the resultant proneural gene oscillations are important for efficient proliferation of NPCs. The defects in Hes1 type-1 homozygous mutant embryos were mild, because Hes1-related genes, such as Hes3 and Hes5, compensate for the Hes1 abnormality. Indeed, when Hes3 and Hes5 were additionally deleted, neurogenesis was enhanced in Hes1 type-1 homozygous mutant embryos. These findings suggest that dampened Hes1 oscillations affect proneural gene oscillations, which may affect the proliferation and differentiation of NPCs.

In Hes1 type-2 homozygous mutant mice, Hes1 oscillations were dampened but less altered than in Hes1 type-1 homozygous mutant mice, and Hes1 type-2 homozygous mutant mice were mostly normal. Thus, developmental processes may be somewhat resistant to dampened Hes1 oscillations. This is partly due to compensation by Hes1-related genes such as Hes3 and Hes5, and it remains to be analyzed whether Hes1 type-2 homozygous mutant embryos exhibit any abnormality in the Hes3- and/or Hes5-null background.

According to mathematical modeling, negative feedback with a delayed timing is required for oscillatory expression, and negative feedback at shorter than optimal delays dampens the oscillations (Fig. S1; Takashima et al., 2011). It has been shown that such delays depend on intronic delays, the time required for transcription and splicing of intronic sequences (Swinburne et al., 2008; Swinburne and Silver, 2008). Indeed, it was previously shown that deletion of all three introns from the Hes7 gene accelerates the timing of negative feedback (faster Hes7 mRNA/protein formation), leading to steady (non-oscillatory) Hes7 expression and severe somite fusion (Takashima et al., 2011). In Hes1 type-1 homozygous mutant embryos, deletion of all three introns led to dampened Hes1 oscillations, but Hes1 expression still oscillated unstably with shorter periodicity and smaller amplitudes. In the developing nervous system, gene expression oscillates out of phase with that of neighboring cells (Shimojo et al., 2016), and it has been mathematically suggested that whereas in-phase oscillations observed in the presomitic mesoderm are robust and stable, out-of-phase oscillations are unstable (Yoshioka-Kobayashi et al., 2020). Thus, unstable Hes1 oscillations in NPCs may be susceptible to interfering noises, and in Hes1 type-1 homozygous mutant embryos noise produced by cell movements and/or cell division may still induce fluctuations in Hes1 expression. Another issue is that Hes1 expression decreased in Hes1 type-1 mutant NPCs. This decrease could be partly because introns are important for both efficient transport and translation of mRNA. Further analyses are required to understand how the dynamics of in-phase and out-of-phase oscillations are controlled.

As mathematically modeled, the Hes1 type-2 mutation delays the timing of negative feedback (slower Hes1 mRNA formation) and leads to Hes1 oscillations with longer periodicity. However, although modeling suggests that slower negative feedback generates oscillations with higher amplitudes, the actual amplitudes of Hes1 oscillations in Hes1 type-2 homozygous mutant embryos were smaller than those in wild-type embryos. This might be partly due to nonsense-mediated mRNA decay (He and Jacobson, 2015), which is often observed when the stop codon is present in the first exon, but the exact mechanism remains to be analyzed.

Our data suggest that robust Hes1 and proneural gene oscillations are necessary for efficient proliferation of NPCs. This notion agrees well with previous data demonstrating that dampened Hes1 oscillations induced by accelerated or delayed Dll1 expression resulted in inhibition of NPC proliferation and acceleration of neurogenesis (Shimojo et al., 2016). Similarly, it was recently shown that Hes1 oscillations drive oscillatory expression of the muscle determination factor MyoD in activated muscle progenitors, and that Hes1 and MyoD oscillations may be important for proliferation and differentiation of muscle progenitors (Lahmann et al., 2019). Furthermore, it has been shown that high and sustained levels of Hes1 repress the expression of cell cycle regulators and lead to G1 arrest in cultured fibroblasts, NPCs and hematopoietic stem cells (Baek et al. 2006; Yu et al., 2006; Sang et al., 2008), suggesting that high and sustained Hes1 is a general feature of maintaining quiescence in many cell types. However, the exact mechanism by which oscillations activate cell proliferation remains to be analyzed. Genes upregulated by oscillatory and downregulated by sustained levels of Hes1, or vice versa, should be determined to understand whether genes, particularly those involved in cell cycle progression, are differentially controlled by the two modes of Hes1 expression.

Another important issue is how the timing of neuronal differentiation is regulated. When Hes1 oscillations were dampened in Hes1 type-1-mutant;Hes5-null and Hes1 type-1-mutant;Hes3;Hes5-double-null embryos, neurogenesis was enhanced compared with the controls, suggesting that Hes1 oscillations play an important role in the normal timing of neurogenesis. Dampened Hes1 oscillations may also dampen proneural gene oscillations, and sustained proneural gene expression might lead to enhanced neurogenesis. It was previously shown that Dll1 oscillations, which are controlled by Hes1 and proneural gene oscillations, are important for the normal timing of neurogenesis (Shimojo et al., 2016). In Hes1 type-1 mutant embryos, Dll1 oscillations might be dampened by dampened Hes1 and proneural gene oscillations, and sustained Dll1 expression may affect the timing of neurogenesis, because sustained Dll1 overexpression induces neuronal differentiation in a non-cell-autonomous manner (Kawaguchi et al., 2008). Taken together, our results suggest that robust Hes1 oscillations play an important role in efficient proliferation and maintenance of NPCs and normal brain morphogenesis.

All animal experiments were approved by the Institutional Animal Care and Use Committee at Kyoto University. Hes3- and Hes5-null mice and Hes1 reporter mice were previously described (Hatakeyama et al., 2004; Shimojo et al., 2008).

For generation of Hes1 type-1 mutant mice, the Hes1 gene from ATG to the stop codon were replaced with HA sequence-fused Hes1 cDNA by BAC recombineering. In the case of Hes1 type-2 mutant mice, HA sequence-fused Hes1 cDNA followed by the 3′UTR of the Hes1 gene were inserted between the 5′UTR and the initiation codon of Hes1 by BAC recombineering. For preparation of the targeting vector of the Hes1 type-1 mutant, an FRT-Neo^R cassette was inserted into the downstream region of the 3′UTR. The region from 2.5 kb upstream of the HA sequence to 8.0 kb downstream of the FRT-Neo^R cassette was retrieved to the pMCS-DTA vector. For preparation of the targeting vector of the Hes1 type-2 mutant, an FRT-Neo^R cassette was inserted into the downstream region of the 3′UTR following the Hes1 cDNA sequence, and the region from 1.7 kb upstream of the HA sequence to 6.0 kb downstream of the FRT-Neo^R cassette was retrieved to the pMCS-DTA vector. The linearized targeting vectors were electroporated into TT2 ESCs, and G418-resistant clones were screened by PCR. The homologous recombinant ESC clones were injected into eight-cell stage mouse embryos to obtain chimeric mice. Chimeric mice were crossed with pCAG-FLP mice to delete the Neo sequence.

Hes1 type-1 and type-2 mutant mice were crossed with Hes1 reporter mice. NPCs were prepared from the cortex of E12.5 and E14.5 wild-type embryos and Hes1 type-1 and type-2 mutant embryos that carried the Hes1 reporter. The protocol of dissociation culture using the Papain Dissociation System (Worthington) was described previously (Shimojo et al., 2008, 2016). Dissociated NPCs (2.0-3.0×10⁶ cells/ml) were plated into φ12-mm or φ27-mm glass-bottom dishes (IWAKI) with 1 mM luciferin (Nacalai Tesque) in N2/B27 medium [DMEM/F12 supplemented with 1× N2 (Thermo Fisher Scientific), 1× B27 (Thermo Fisher Scientific), 1 mM N-acetyl-cysteine, 10 ng/ml bFGF (Thermo Fisher Scientific)]. After a 1-h pre-incubation, cells were cultured in medium containing luciferin for measurement of bioluminescence. The dish was placed on the stage of an inverted microscope (Olympus IX81) and was maintained at 37°C. Bioluminescence was measured using an Olympus objective lens (UPLFLN 40 O) and was transmitted directly to a CCD camera (Princeton Instruments, VersArray 1 kb), as previously described (Shimojo et al., 2008, 2016).

Images were processed using Fiji image analysis software. Stack images were processed using the Spike Noise Filter to remove signals from cosmic rays and then the Savitzky Golay Temporal Filter to get clear dynamic expression. Period and amplitude of oscillatory expression were measured by single cell tracking (Webb et al., 2016). Detrended fluctuation analysis (DFA) was used for determining self-affinity of a signal, and the time series was detrended by subtracting the moving average within a 240 min window. To calculate moving averages near the edges (initial and late 120 min frames in the time series), data-padding was performed by mirroring. Extreme values were extracted manually, and periods shorter than 60 min and longer than 250 min were discarded. The amplitude was calculated as the average of peak heights relative to the mean (in arbitrary units).

Embryos and brains were dissected and fixed in 4% paraformaldehyde overnight at 4°C. After being washed in PBS, equilibrated in 25% sucrose/PBS overnight at 4°C and then embedded in OCT compound, tissues were sectioned at 16 μm using a cryostat (CM1950, Leica). NPC cultures were fixed in 4% paraformaldehyde for 20 min at room temperature. For immunostaining of Hes1, antigen retrieval was performed in 0.1% Tween20/0.01 M citrate buffer (pH 6.0) using an autoclave (15 min at 105°C). Sections were incubated with primary antibody (rabbit anti-Hes1; Kobayashi et al., 2009) overnight at 4°C. After being washed in PBS, sections were incubated with secondary antibody (HRP-conjugated anti-rabbit IgG) for 90 min at room temperature. After washes in PBS and PBST (PBS with 0.3% Triton X-100), color development was enhanced using the TSA amplification system (Perkin Elmer) according to the manufacturer's instructions. For immunostaining of Pax6, Tbr2, Tbr1 and Cux1, antigen retrieval was performed in 0.1% Tween20/0.01 M citrate buffer (pH 6.0) using an autoclave (15 min at 105°C). Primary antibodies used were as follows: mouse anti-Pax6, rabbit anti-Tbr2, rabbit anti-Tbr1, rabbit anti-Cux1, mouse anti-nestin, rabbit anti-Tuj1, mouse anti-pH3, rabbit anti-cleaved-casp3, mouse anti-GAD65 and rabbit anti-GABA. Sections were incubated with primary antibodies overnight at 4°C. The secondary antibodies used were as follows: Alexa 488-conjugated anti-mouse IgG, Alexa 594-conjugated anti-rabbit IgG, Alexa 488-conjugated anti-rabbit IgG and Alexa 594-conjugated anti-rat IgG. Antibody details are listed in Table S1.

Preparation of DIG-labeled antisense RNA probes and in situ hybridization using NBT/BCIP (Roche) detection were performed as described previously (Bessho et al., 2001; Shimojo et al., 2008).

Samples were collected from the telencephalon of each embryo, total RNA was extracted, and then reverse transcription (RT) reaction was performed. Real-time PCR (qPCR) was performed as described previously (Kobayashi et al., 2009). Quantified values of RNA were normalized with those of Gapdh. Primers are listed in Table S2. At least three embryos of each genotype were examined.

The intensity of Hes1 immunostaining of individual NPCs in the cortex of each genotype was measured using Fiji software, as previously described (Baek et al., 2006).

Western blot analysis of Hes1 was performed as previously described (Hirata et al., 2002).

Images of embryos were obtained using a stereomicroscope (MZ16 FA, Leica). The length of the long axis and the area of the whole body and of the telencephalon were measured using Fiji software. Quantified values of mutants were compared with those of control mice. At least three embryos of each genotype were examined.

The area or the number of cells expressing each neuronal marker (Tuj1, Tbr2, Cux1, Ctip2, Tbr1) in the cortex of each embryo was measured using Fiji software. For the quantification of neuronal layers, serial 16-µm sections were cut, and every 20th section was used. At least three embryos of each genotype were examined.

pH3⁺ mitotic cells in the cortex were classified into apical or basal mitoses (Cárdenas et al., 2018). The numbers of pH3⁺ mitotic and cleaved caspase3⁺ apoptotic cells were counted. At least three embryos of each genotype were examined.

NPC and neurosphere cultures were performed, as previously described (Ohtsuka et al., 2006; Imayoshi et al., 2013). For cell cycle exit analysis, 10 µM 5-bromo-2′-deoxyuridine (BrdU) was added to NPCs, which were then fixed 24 h later and subjected to immunostaining for BrdU and Ki67 (Mki67). The proportion of BrdU⁺Ki67⁻ cell number over BrdU⁺ cell number was calculated.

Luc2-fused control, Hes1 type-1 and Hes1 type-2 vectors were transfected into C2C12 cells and NSCs. Synchronized Hes1 expression in transfected cells was induced by application of 5% fetal bovine serum (Harima et al., 2013). The kinetics of Hes1 expression in these cells were measured using a photomultiplier tube (CL24B-LIC/B, Churitsu Electric Corporation).

where k is the number of molecules of Hes1 mRNA synthesized per unit time in the absence of inhibition and p_crit is the amount of protein that gives half-maximal inhibition. We set a=4.5 protein molecules per mRNA molecule per min, p_crit=40 molecules per cell, k=33 mRNA molecules per cell per min, and τ_m=3 min. We assume that the Hes1 protein half-life τ_p=20 min, the translational delay T_p=8 min, and the transcriptional delay T_m=29 min. Under these conditions, oscillatory expression continues. When T_m=34 min (5 min longer), oscillations are maintained. In contrast, when T_m=10 min (19 min shorter) and T_m=15.5 min (13.5 min shorter), oscillations are dampened. The figure of Hes1 dynamics (Fig. S1) was exported from the mathematical model using XPPAUT.

We thank Dai Watanabe, Satoshi Yawata and Chika Nishimura for technical help.