Hes5 regulates the transition timing of neurogenesis and gliogenesis in mammalian neocortical development

Neural stem/progenitor cells (NSCs) sequentially give rise to deep layer neurons and later to superficial layer neurons; at late embryonic stages, NSCs terminate neurogenesis and shift to gliogenesis. As such, NSCs temporally alter their characteristics, and the timing of the generation of a variety of neurons and glial cells is strictly regulated (McConnell, 1989; Temple, 2001; Ohtsuka et al., 2011). Growing evidence implicates the involvement of epigenetic regulatory systems in the regulation of this transition timing. The polycomb group (PcG) complex of transcriptional repressors has been found to govern the developmental timing of neurogenesis and gliogenesis by modulating histones and chromatin structure during corticogenesis (Vogel et al., 2006; Hirabayashi et al., 2009; Schwartz and Pirrotta, 2013; Pereira et al., 2010; Morimoto-Suzki et al., 2014; Corley and Kroll, 2015). Moreover, it has been reported that the high mobility group AT-hook (Hgma) genes regulate gene expression by modulating chromatin structure (Ozturk et al., 2014), maintain neurogenic NSCs, and inhibit gliogenesis during early- to mid-embryonic stages through global opening of the chromatin state (Kishi et al., 2012). However, the mechanism by which the expression of these epigenetic factors is controlled remains to be analyzed.

Here, we found that Hes5, a transcriptional repressor acting as an effector of Notch signaling, regulates the timing of neurogenesis and gliogenesis via alteration in the expression levels of epigenetic factors. Notch signaling contributes to the elaboration of cellular diversity during the development of various tissues (Kopan et al., 1994; Robey et al., 1996; Weinmaster, 1997; Gridley, 1997; Artavanis-Tsakonas et al., 1999). In addition, in the central nervous system (CNS), Notch signaling governs various developmental processes, such as maintenance of NSCs, neurite outgrowth of cortical neurons, and neuronal versus glial fate choice (de la Pompa et al., 1997; Berezovska et al., 1999; Tanigaki et al., 2001). We previously found that some members of the Hes gene family of basic helix-loop-helix (bHLH) transcriptional repressors (Hes1/3/5) function downstream of Notch signaling and inhibit neuronal differentiation (Ishibashi et al., 1994; Ohtsuka et al., 1999, 2001; Kageyama and Ohtsuka, 1999; Hatakeyama et al., 2004). NSCs initially expand as neuroepithelial cells by symmetric proliferative divisions and then transform to radial glial cells (RGCs), a neurogenic form of NSCs, after switching to asymmetric neurogenic divisions. After this transition, differentiating neurons express Notch ligands such as Delta-like (Dll) and Jagged (Jag), and trigger a boost of Notch-Hes signaling in NSCs. Expression of Hes5 is upregulated in RGCs after the onset of neurogenesis, when NSCs start to receive strong intercellular Delta-Notch signals from differentiating and mature neurons (Hatakeyama et al., 2004). Thus, Hes5 is a key regulator of the maintenance of NSCs after the transition to the asymmetric neurogenic division mode.

To more precisely uncover the diverse role of Hes5 throughout the course of neocortical development, we generated transgenic (Tg) mouse lines in which Hes5 expression in NSCs could be manipulated by the Tet-On system. In the Hes5-overexpressing Tg mice, neuronal differentiation from NSCs was strongly inhibited, while the switching from deep to superficial layer neurogenesis was shifted earlier, and gliogenesis was also accelerated and enhanced. By contrast, the transition of neurogenesis and gliogenesis was delayed in the Hes5 knockout (KO) mice. We found that expression of Hmga1/2 was suppressed in the neocortical regions of Tg brain; conversely, their expression was upregulated in the Hes5 KO brain. Furthermore, we found that Hes5 expression led to suppression of the promoter activity of Hmga1/2 in reporter assays. These results suggest that Hes5 regulates the switching of both neurogenesis and gliogenesis, accompanied by alteration in the expression levels of Hgma genes.

To manipulate Hes5 expression in NSCs, two Tg mouse lines were generated (Fig. 1A). rtTA was driven in NSCs by the promoter and second intron of the nestin (Nes) gene (pNes-rtTA Tg). In the presence of doxycycline (Dox), rtTA binds to the Tet-responsive element (TRE) and bidirectionally activates expression of Hes5 and d2EGFP (TRE-Hes5/d2EGFP Tg). By crossing both lines, Hes5-overexpressing Tg mice were generated, and Dox (2.0 mg/ml in drinking water) was continuously administered to pregnant mice from embryonic day (E) 9.5 until sacrifice. The whole-body and brain sizes were smaller in the Tg mice than in the wild-type (WT) mice, and GFP was highly expressed in the central nervous system, including the brain, retina and spinal cord (Fig. 1B). Notably, olfactory bulbs were frequently missing in the Tg brains in which Hes5 and GFP were highly expressed (Fig. 1B, arrows). We confirmed by in situ hybridization that Hes5 expression was enhanced in the ventricular zone (VZ) and subventricular zone (SVZ), in which GFP was highly expressed (Fig. 1C). Real-time RT-PCR revealed that expression of Hes5 mRNA increased by E10.5 when Dox was administered starting at E9.5 (Fig. 1D). GFP overlapped with Pax6, a marker for NSCs, and partially overlapped with Tbr2, a marker for intermediate progenitor cells (IPs), and Tuj1, a neuronal marker, in the dorsolateral telencephalon (neocortical regions) (Fig. 1E). Noticeably, the lateral ventricles and the VZ (Pax6⁺) were dilated, and the thicknesses of neuronal layers (Tuj1⁺) were markedly thinner in the Tg brain compared to the WT brain (Fig. 1F). Tbr2⁺ IPs in the SVZ were reduced in number, and the thickness of the cortical plate (CP), which consists of NeuN⁺ differentiated neurons, was also reduced in the Tg brain at all developmental stages examined (Fig. 1G-J), indicating that Hes5 overexpression led to maintenance and expansion of NSCs by inhibiting neuronal differentiation. However, analysis of neurogenesis/gliogenesis at postnatal stages was difficult because the Tg mice could not survive after birth, although the cause of death is unknown.

Although neuronal differentiation was inhibited and NSCs were maintained longer in the Tg brain, the brain size was rather reduced (Fig. 1B). Therefore, we analyzed cell death and cell proliferation activity in neural progenitors in the neocortical regions. Immunostaining with anti-cleaved caspase-3 antibodies revealed that cell death was slightly enhanced in the neocortical regions of Tg brain (Fig. S1A). Next, we assessed cell proliferation by conducting 5-bromo-2′-deoxyuridine (BrdU) incorporation experiments and immunostaining with antibodies against Ki67, a marker of proliferating cells, and phospho-histone H3 (pH3), a marker of dividing cells in M phase. BrdU was administered intraperitoneally to pregnant mice 30 min before sacrifice at E13.5 or E15.5 to mark cells that incorporated BrdU in S phase. Although the number of Ki67⁺ cells was rather increased in the Tg brain at E13.5, BrdU⁺ and pH3⁺ cells were decreased in number in the neocortical regions of Tg brain compared to WT brain (Fig. S1B). Later, at E15.5, the numbers of BrdU⁺ and pH3⁺ cells were markedly reduced in the neocortical regions of Tg brain compared to WT brain (Fig. S1C), indicating that proliferating cells and dividing cells were reduced in number in the Tg brain. These results suggested that cell proliferation activity was suppressed by Hes5 overexpression, thus leading to unexpectedly smaller brain size in the Tg mice. Next, we analyzed self-renewal versus cell cycle exit of NSCs in the WT and Tg brains. We administered 5-ethynyl-2′-deoxyuridine (EdU) intraperitoneally to pregnant mice at E14, and characterized EdU⁺ cells by immunostaining with anti-Pax6, Tbr2, Tuj1 and Ki67 antibodies 12 h later at E14.5. A higher proportion of EdU-incorporated cells remained as Pax6⁺ NSCs in the Tg brain compared to the WT brain (Fig. S2A,B), whereas more EdU⁺ cells were colabeled with Tbr2 and Tuj1 in the WT brain than in the Tg brain (Fig. S2C-F). In line with these observations, the proportion of Ki67⁺;EdU⁺/EdU⁺ (cell cycle re-entry) was higher in the Tg cortex, whereas the proportion of Ki67⁻;EdU⁺/EdU⁺ (cell cycle exit) was higher in the WT cortex at E14.5 (Fig. S2G,H). Taken together, these results indicated that the frequency of self-renewal of NSCs was higher in the Tg cortex than in the WT cortex.

Next, we analyzed the expression of Hes1 and neurogenic bHLH genes such as Neurog2 and Ascl1. Expression levels of Hes1 (protein and mRNA) and Neurog2 mRNA were downregulated in the VZ of neocortical regions of Tg brain (Fig. 2A,B), while Ascl1 expression in the neocortical regions was not significantly altered, indicating that Hes5 overexpression could repress Neurog2 and inhibit neuronal differentiation even when Hes1 activity was attenuated. This result is consistent with the notion that Hes1 and Hes5 show compensatory expression patterns in the developing nervous system through mutual repression of their transcription (Hatakeyama et al., 2004). In addition, we found that the thicknesses of all cortical layers were reduced in the Tg brain by using markers for layers II-IV (Cux1), layer V (Ctip2) and layer VI (Tbr1) neurons at E15.5 and postnatal day 0 (P0) (Fig. 2C-E). Furthermore, reduction in the expression level of Fezf2, a fate determinant of Ctip2⁺ neurons, which correspond to subcerebral projection neurons (SCPNs), was demonstrated by in situ hybridization (Fig. 2F). Real-time RT-PCR using total RNAs prepared from the neocortical regions revealed that expression levels of Tbr1, Ctip2 (Bcl11b – Mouse Genome Informatics) and Fezf2 were significantly downregulated in the Tg brain at E14.5 and E16.5 (Fig. 2G). However, Cux1 expression was not significantly changed, probably because it was maintained in the VZ/SVZ (Fig. 2C). It is likely that the suppressed proliferation of neural progenitors via Hes5 overexpression led to the reduced thicknesses of all cortical layers and smaller brain size in the Tg mice.

Next, we performed birth-date analysis by administering EdU intraperitoneally to pregnant mice. When EdU was injected at E10.5, more EdU⁺ cells, which were born during the period of EdU exposure, were found as layer VI neurons (Tbr1⁺) in the Tg cortex at E18.5, suggesting an earlier onset of neurogenesis in the Tg brain compared to the WT brain (Fig. S3A,B). When EdU was administered at E11.5, most EdU⁺ cells were located in layer VI (Tbr1⁺) of the WT brain, while more EdU⁺ cells were observed in layer V (Ctip2⁺) of the Tg brain compared to the WT brain (Fig. S3C-F). When EdU was injected at E12.5, most EdU⁺ neurons were still located in layer VI (Tbr1⁺) but not in layer V (Ctip2⁺) of the WT cortex at E18.5 (Fig. 3A,B). By contrast, EdU⁺ cells were distributed in layer VI (Tbr1⁺), layer V (Ctip2⁺), and further upper layers of the Tg cortex, suggesting that Ctip2⁺ layer V and further upper layer neurons were born prematurely in the Tg brain. In the WT brain, when EdU was injected at E13.5, the majority of EdU⁺ cells were located in layer V (Ctip2⁺), while some were in layers II-IV (Cux1⁺) (Fig. 3C,D). By contrast, in the Tg brain, the majority of EdU⁺ cells were distributed in layers II-IV (Cux1⁺), while some were in layer V (Ctip2⁺), suggesting that Cux1⁺ layer II-IV neurons were born prematurely in the Tg brain. When EdU was injected at E14.5, most EdU⁺ cells were settled in layers II-IV (Cux1⁺) of both WT and Tg brains (Fig. 3E,F). These results indicated that Hes5 overexpression accelerated the switching from deep to superficial layer neurogenesis.

We next addressed whether the onset of gliogenesis was likewise accelerated in the neocortical regions by Hes5 overexpression. Although expression of GFAP (a marker for astrocytes) was not observed in the neocortical regions of WT mice at E17.5, a considerable number of GFAP⁺ astrocytes precociously appeared in the neocortical regions of Tg mice (Fig. 3G). Whereas GFAP signal was present in the neocortical regions of WT mice at P0, the number of GFAP⁺ astrocytes was markedly higher in the VZ/SVZ and throughout the cortex of Tg mice. Most GFAP⁺ cells coexpressed GFP in the VZ/SVZ, but many GFAP⁺ cells in the outer region were GFP⁻, suggesting that the outer region GFAP⁺ cells lost Hes5/d2EGFP expression over the course of glial differentiation. Real-time RT-PCR demonstrated that GFAP expression was significantly higher in the neocortical regions of Tg brain at E16.5 and E18.5 (Fig. 3H). These results indicated that astrogenesis was accelerated and enhanced by Hes5 overexpression. In addition, we analyzed the generation of oligodendrocyte lineage cells with antibodies against Olig2, a marker for oligodendrocytes, and found that the numbers of Olig2⁺ cells were reduced in the neocortical regions of Tg brain compared to WT brain at E17.5 and E18.5 (Fig. S4). This might be attributed to the hypoplastic ventral telencephalon (Fig. 1F), which is the primary source of oligodendrocyte progenitor cells.

Subsequently, we performed birth-date analysis in the neocortical regions of Hes5 KO mice (Ohtsuka et al., 1999; Hatakeyama et al., 2004). When EdU was injected at E10.5 or E11.5, we did not find any significant difference in early neurogenesis between the WT and KO mice (data not shown). However, when EdU was injected at E14.5, most EdU⁺ cells were located in layers II-IV (Cux1⁺), but not in layer V (Ctip2⁺), of the WT cortex, whereas EdU⁺ cells were predominantly distributed in layer V (Ctip2⁺), and fewer EdU⁺ cells were located in layers II-IV (Cux1⁺) of the Hes5 KO cortex, at E18.5 (Fig. 4A,B), indicating that the switching from deep to superficial layer neurogenesis was delayed in the Hes5 KO brain. In addition, immunohistochemistry and real-time RT-PCR for GFAP expression revealed the delayed onset of astrogenesis in the neocortical regions of Hes5 KO brain compared to WT brain at E18.5 (Fig. 4C,D). These results exhibited striking contrast to the accelerated neurogenesis and astrogenesis observed in the Hes5-overexpressing Tg mice.

Because it has been found that PcG complex and Hmga proteins govern the developmental timing of neurogenesis and gliogenesis, we assessed their expression levels in the Hes5-overexpressing Tg brain. Real-time RT-PCR demonstrated upregulation of Ezh2 and Eed, components of polycomb repressive complex 2 (PRC2), in the neocortical regions of Tg brain at E12.5 (Fig. 5A), whereas the expression of Ring1B (Rnf2 – Mouse Genome Informatics), a component of PRC1, was not significantly affected. In situ hybridization also demonstrated a subtle increase in the expression levels of both genes in the neocortical regions of Tg brain at E12.5 (Fig. S5A,B). Notably, expression levels of Hmga1 and Hmga2 were significantly reduced in the Tg brain compared to the WT brain (Fig. 5B). We confirmed by in situ hybridization that the expression of both Hmga1 and Hmga2 was remarkably downregulated in the neocortical regions of Tg brain compared to WT brain (Fig. 5C). Conversely, expression levels of Hmga1 and Hmga2 were upregulated in the neocortical regions of Hes5 KO brain compared to WT brain (Fig. 5D,E). Expression levels of Hmga1/2 in the ventral telencephalon, diencephalon, midbrain and hindbrain were not much altered in the Tg brain, although the brain shape was apparently deformed compared to that of the WT (Fig. S5C,D). To reveal that the phenotype of Hes5 KO brain is mediated by the upregulation of Hmga1/2, we performed knockdown experiments by introducing a combination of expression vectors of shRNA targeting Hmga1 (shHmga1) and Hmga2 (shHmga2), in addition to scrambled shRNA control vectors (shScrambled) (Kishi et al., 2012), in the VZ of the KO brain by in utero electroporation at E13.5. When we performed birth date analysis of Hmga1/2 knockdown cells by administering EdU at E14.5, the switching delay was partially recovered by Hmga1/2 knockdown in the Hes5 KO brain (Fig. 6). Taken together, these results suggest that the transition timing of layer-specific neurogenesis and astrogenesis was regulated at least partly by the expression levels of Hmga1/2 genes.

We next conducted reporter analysis using the Hmga1 (−5000 to +28) and Hmga2 (−3111 to +56) promoters to test whether Hes5 downregulates the transcription of these genes (Fig. 7A). Cotransfection of Hes5 expression vectors (pEF-HA-Hes5) with reporter vectors (pHmga1-luc or pHmga2-luc) in NIH3T3 cells or mouse neural stem (NS) cells significantly downregulated the promoter activity of both Hmga1 and Hmga2 in a dose-dependent manner (Fig. 7B), suggesting that Hes5 functions as a negative regulator for transcription of Hmga1/2. Because the promoter activity of Hmga1 was more effectively downregulated by Hes5 expression compared to that of Hmga2, we focused on the Hmga1 promoter for further analysis. We narrowed down the promoter sequence to the regions responsible for the Hes5-mediated transcriptional repression. Intriguingly, the promoter regions 2.5 kb and 3.9 kb upstream of the ATG codon exhibited no apparent repression by Hes5, whereas the 4.6 kb promoter region exhibited Hes5-mediated transcriptional repression (Fig. 7C,D), suggesting that the specified region (between −4596 and −3910) and/or surrounding region contains critical regulatory sites. It has been established that the Hes family of bHLH factors repress transcription by binding to the N box (CACNAG) and the class C site [CACG(C/A)G], and we found one N box (−4547 to −4542) within the specified region. Therefore, we constructed Hmga1 reporter vectors with various modifications in the N box, such as substitutions of 1, 2 or 5 nucleotides, or a deletion of all six nucleotides in the N box sequence, and performed the luciferase assay. However, these modifications resulted in only mild effects on Hmga1 promoter activity. Given that there are many N box/class C sites in the Hmga1 promoter, as shown in Fig. 7C, it is possible that modifications in a few N box/class C sites cannot induce an effective de-repression. In addition, we found several E box sequences in the Hmga1 promoter region, suggesting that the Hmga1 transcriptional activity could be indirectly regulated by Hes5 through other regulatory sequences including E box.

In the present study, we showed that the transition timing from deep to superficial layer neurogenesis was shifted earlier and gliogenesis was also accelerated with an earlier termination of the neurogenic period via overexpression of Hes5 (Fig. 3); conversely, the transition from deep to superficial layer neurogenesis and the onset of gliogenesis were delayed in Hes5 KO mice (Fig. 4). These results suggest that these transitions are not governed by the number of cell divisions, given that the cell cycle progression was slower in NSCs in the neocortical regions of the Tg brain (Fig. S1), which is in line with a previous report that the transition in temporal identity of NSCs can be controlled independently of cell cycle (Okamoto et al., 2016). Therefore, we next addressed whether the machinery that regulates the transition timing of neurogenesis and gliogenesis was affected by Hes5 overexpression.

Temporal modification of the chromatin structure by the PcG complex is known to be one of the key mechanisms regulating the transition timing of neocortical neurogenesis and gliogenesis (Vogel et al., 2006; Hirabayashi et al., 2009; Pereira et al., 2010; Morimoto-Suzki et al., 2014; Corley and Kroll, 2015). It is also known that there are two classes of PRC; PRC1 contains a ubiquitin ligase Ring1A/B and PcG RING finger (PCGF) proteins, and PRC2 is composed of Eed, Suz12 and a methyltransferase Ezh1/2 (Schwartz and Pirrotta, 2013). Morimoto-Suzki et al. reported that Ring1B suppresses Fezf2, a fate determinant of Ctip2⁺ SCPNs [a class of deep layer (layer V) neurons], and has a role in terminating the generation of Ctip2⁺ neurons (Morimoto-Suzki et al., 2014). These findings suggested that Ring1B mediates the timed termination of Fezf2 expression and switching from deep to superficial layer neurogenesis. Hirabayashi et al. reported that the level of H3K27me3 at the Neurog1 promoter region gradually increases over time, and that PcG proteins suppress the Neurog1 locus during the astrogenic phase and mediate the neurogenic-to-gliogenic fate switching in the developing cortex (Hirabayashi et al., 2009). In the present study, expression levels of Ezh2 and Eed were upregulated by Hes5 overexpression in the neocortical regions only during the early neurogenic stages (Fig. 5A). By contrast, expression levels of Hmga1/2 were markedly downregulated in the neocortical regions of Hes5-overexpressing Tg brain, but upregulated in the Hes5 KO brain (Fig. 5C-E).

Hmga proteins belong to the Hmg protein family, which consists of nonhistone chromatin-associated proteins that regulate gene expression by modulating chromatin structure (Ozturk et al., 2014). Among their diverse functions, previous reports have uncovered roles in self-renewal of NSCs (Nishino et al., 2008) and gliogenesis during brain development (Sanosaka et al., 2008). It was revealed that Hmga2 promotes self-renewal of NSCs by decreasing p16^Ink4a/p19^Arf expression, but its expression declines with age, partly due to the increasing expression of let-7b microRNA (Nishino et al., 2008). Furthermore, it has been reported that Hmga proteins are essential for the open chromatin state in early-stage NSCs to maintain their neurogenic potential at early developmental stages (Kishi et al., 2012). Overexpression of Hmga proteins can retrieve the neurogenic potential even in late-stage NSCs, and knockdown of Hmga1/2 promotes astrogenesis. In addition, Fujii et al. reported that insulin-like growth factor 2 mRNA-binding protein 2 (Igf2bp2) is one of the key mediators of Hmga function that regulates the neurogenic potential of early-stage NSCs (Fujii et al., 2013). Thus, it is likely that the downregulation of Hmga1/2 accelerated the transition from deep to superficial layer neurogenesis and the onset of astrogenesis in the Hes5-overexpressing Tg cortex. This accelerated transition might be partly due to the higher activity of PcG complexes (Fig. 5A), which might have caused the reduction in Fezf2 expression (Fig. 2F,G), resulting in earlier termination of the generation of Ctip2⁺ deep layer neurons with precocious switching to production of Cux1⁺ superficial layer neurons (Fig. 3) in the Tg cortex. In the present study, we demonstrated that Hes5 expression leads to suppression of the promoter activity of Hmga1/2 in reporter analyses (Fig. 7), although we could not obtain evidence for the direct regulation of Hmga1/2 by Hes5. Likewise, we could not demonstrate the direct regulation of PcG genes, such as Ezh2, Eed and Ring1B, by Hes5 (data not shown), in agreement with the modest effects of Hes5 overexpression on expression levels of such PcG genes (Fig. 5A). However, it is possible that the activity of the PcG complex can be modulated by Hmga or other factors that are regulated by Hes5.

During neocortical development, expression of Hes5 is upregulated after the onset of neurogenesis when NSCs start to receive strong intercellular Delta-Notch signals from differentiating and mature neurons (Hatakeyama et al., 2004). Our results indicate that the timing and levels of Hes5 expression must be properly regulated in order to maintain the precise temporal control of deep and superficial layer neurogenesis and switching from neurogenesis to gliogenesis in normal brain development. Otherwise, brain size and neocortical organization become disturbed. It has been revealed that Hes5 exhibits oscillatory expression in NSCs by a negative autoregulation mechanism (Imayoshi et al., 2013), similarly to Hes1 (Hirata et al., 2002), indicating the importance of the control of Hes5 expression within appropriate levels to avoid its excessive expression. Consistent with the above notion, we found that expression levels of Hmga1/2 successively declined in NSCs of mouse neocortical regions during the mid-gestation period between E11.5 and E17.5, whereas Hes5 expression was gradually upregulated during this period (Ohtsuka et al., 2011). Other research groups have also demonstrated that HMGA2 expression successively decreased, while HES5 expression gradually increased, in the developing human brain during the mid-gestation period (Patterson et al., 2014). Interestingly, Patterson et al. reported a link between the let-7/HMGA2 circuit and Notch signaling. They found that HMGA2 regulated fate decisions between neurogenesis and gliogenesis in human NSCs via HES5, and that the let-7 family of miRNAs downregulated HMGA2 expression. In addition, they demonstrated that knockdown of HMGA2 by small interfering RNA led to a dramatic downregulation of HES5, probably due to the blockade of access of NICD to the HES5 promoter, while siRNA for HES5 (40% knockdown of HES5 expression) caused only a subtle increase in the HMGA2 expression level. These results and the present findings together indicate a crucial correlation between downstream effectors of the Notch signaling pathway and epigenetic regulatory factors, although the entire mechanism and biological significance of the correlation between the two systems remain to be elucidated.

For the pNes-rtTA transgene, a rtTA-Advanced fragment excised from pTet-On Advanced vector (Clontech) was subcloned between the Nes promoter (5.8 kb) and polyadenylation sequence of SV40 with the second intron (1.7 kb) of the Nes gene. For the TRE-Hes5/d2EGFP transgene, mouse Hes5 and d2EGFP (Clontech) cDNAs were inserted into pTRE-Tight-BI vector (Clontech) (Fig. 1A). Two Tg mouse lines were generated using each transgene and maintained on the ICR background. Both lines were crossed and doxycycline hyclate (Sigma-Aldrich; 2.0 mg/ml) in drinking water with 5% sucrose was continuously administered to pregnant mice from E9.5 until sacrifice. Images of whole bodies with GFP fluorescence were obtained with a MZ16FA fluorescence stereo microscope equipped with a DFC300 FX digital camera (Leica). Animal experiments were carried out according to the guidelines for animal experiments at Kyoto University.

Immunohistochemistry was performed as described previously (Ohtsuka et al., 2011). Brains were fixed in 4% paraformaldehyde, incubated overnight with 20% (w/v) sucrose in PBS at 4°C, embedded in OCT compound and cryosectioned at 16 μm thickness. Sections were blocked with 5% normal goat serum in 0.1% Triton X-100 in PBS for 1 h at room temperature. Primary antibodies diluted in 1% normal goat serum/0.1% Triton X-100 in PBS were applied overnight at 4°C. Primary antibodies used in this study are listed in Table S1. Alexa Fluor-conjugated secondary antibodies (Molecular Probes; 1:200) were applied for 2 h at room temperature to detect primary antibodies. DAPI (4′,6-diamidino-2-phenylindole) (Sigma-Aldrich) was used for nuclear staining. Images were analyzed using a LSM510 confocal microscope (Zeiss).

Preparation of digoxygenin-labeled RNA probes and in situ hybridization were performed as described previously (Ohtsuka et al., 2011). The coding sequences of Hes5 (NM_010419.4), Hes1 (NM_008235.2), Neurog2 (NM_009718.2), Ascl1 (NM_008553.4), Fezf2 (NM_080433.3), Hmga1 (NM_016660.3), Hmga2 (NM_010441.2), Ezh2 (NM_007971.2) and Eed (NM_021876.3) were used as templates for the RNA probes.

Total RNA was prepared from the neocortical regions of mouse brains (Ohtsuka et al., 2011). Reverse transcription was performed using total RNA as described previously (Tan et al., 2012). Gapdh was used as an internal control. PCR primers are listed in Table S2.

BrdU (Sigma-Aldrich; 50 μg BrdU/g of body weight) was injected intraperitoneally to pregnant mice 30 min before sacrifice. The BrdU⁺ cells were detected by immunohistochemistry as described previously (Tan et al., 2012).

EdU (Molecular Probes; 12.5 μg EdU/g of body weight) was injected intraperitoneally to pregnant mice at E12.5, E13.5 or E14.5, and the offspring were analyzed at E18.5. EdU-labeled cells were detected by a fluorogenic click reaction (Salic and Mitchison, 2008). The numbers of EdU-labeled cells positive and negative for the layer marker were manually counted in the immunostained sections, and the proportions of EdU-labeled cells positive for the layer marker among total EdU-labeled cells were calculated. At least three sections of three mice were used for quantification.

Knockdown experiments were performed by introducing a combination of expression vectors of shRNA targeting Hmga1 (pSiren-EGFP-shHmga1) and Hmga2 (pSiren-EGFP-shHmga2), in addition to scrambled shRNA control vectors (pSiren-EGFP-shScrambled) (Kishi et al., 2012), in the VZ of Hes5 KO brain by in utero electroporation at E13.5. pSiren-EGFP-shHmga1, pSiren-EGFP-shHmga2 and pSiren-EGFP-shScrambled vectors were kindly provided by Dr Yusuke Kishi and Dr Yukiko Gotoh (The University of Tokyo). In utero electroporation was performed as described previously (Ohtsuka et al., 2011).

For the Hes5 expression vector (pEF-HA-Hes5), a hemagglutinin (HA)-tagged coding sequence of Hes5 (NM_010419.4) was cloned into pEF-MM vector. For the luciferase reporter constructs (pHmga1-luc and pHmga2-luc), a 5.0-kb fragment of the Hmga1 (variant 1) (NM_016660.3) promoter region (from −5000 to +28, relative to the ATG codon in exon 2) or a 3.2-kb fragment of the Hmga2 (NM_010441.2) promoter region (from −3111 to +56, relative to the ATG codon in exon 1) was inserted into the multiple cloning site of the pGL4.10 luciferase vector (Promega). For the truncated versions of Hmga1 reporters, −4596 bp, −3910 bp, −2508 bp, −2000 bp, −1506 bp and −1003 bp upstream promoter regions relative to the ATG codon were cloned into the pGL4.10 luciferase vector.

NIH3T3 cells were cultured in DMEM and 10% FBS. Neural stem (NS) cells were originally established from basal forebrain regions of perinatal mice (Imayoshi et al., 2013) and maintained using a NS cell culture method as described previously (Conti et al., 2005) with minor modifications. Briefly, NS cells were plated in serum-free medium (DMEM/F12) supplemented with 1% N2-MAX medium (R&D Systems), 1% Penicillin/Streptomycin (Nacalai), 10 ng/μl of both EGF and basic FGF (Invitrogen), and 4 μg/ml laminin (Sigma-Aldrich). For transfection, NIH3T3 cells or NS cells were cultured at subconfluence and transfected using ViaFect transfection reagent (Promega) according to the manufacturer's instructions.

The luciferase reporters of Hmga1 or Hmga2 (pHmga1-luc or pHmga2-luc) (0.1 μg) and the Hes5 expression plasmids (pEF-HA-Hes5) (0.05-0.4 μg) were transfected into NIH3T3 cells or mouse NS cells along with pRL-SV40 (Promega) to normalize the transfection efficiency. The luciferase assay was performed 24 h post-transfection as described previously (Sakamoto et al., 2003).

Each value was obtained from at least three independent samples. Statistical significance was evaluated by Student's t-test. Data are presented as mean±s.e.m.

We thank Hitoshi Miyachi for help with the generation of pNes-rtTA and TRE-Hes5/d2EGFP transgenic mice; Yusuke Kishi and Yukiko Gotoh for pSiren-EGFP-shHmga1, pSiren-EGFP-shHmga2 and pSiren-EGFP-shScrambled vectors; and Hiromi Shimojo for technical instruction and critical discussion.