Hes1 overexpression leads to expansion of embryonic neural stem cell pool and stem cell reservoir in the postnatal brain

At early stages of mammalian brain development, neuroepithelial cells, a primary form of neural stem cells (NSCs), proliferate to expand the NSC pool by repeated symmetric divisions, and the neural tube expands to form brain vesicles. Subsequently, neuroepithelial cells start neurogenesis and transform into radial glial cells (RGCs), another form of NSC. During the neurogenic period, RGCs asymmetrically divide in the innermost layer of the neural tube, termed the ventricular zone (VZ). One daughter cell remains as an NSC, and the other differentiates into a neuron or neuronal precursor cell such as an intermediate progenitor cell (IPC). Thus, brain vesicles/ventricles and the VZ enlarge during the symmetric proliferative division mode of NSCs but stop expanding once most NSCs shift to the asymmetric neurogenic division mode, because asymmetric divisions no longer contribute to an increase in NSC numbers (Takahashi et al., 1995, 1999). During the neurogenic period, NSCs in neocortical regions sequentially give rise to deep layer neurons and then superficial layer neurons. At late embryonic stages, NSCs finish neurogenesis and shift to the gliogenic phase, in which astrocytes are mainly produced. Only a subset of NSCs remain in specific regions of the postnatal and adult brain as astrocyte-like cells that retain stemness. These NSC populations in the two major neurogenic niches in the adult brain – the subventricular zone (SVZ) of the rostral part of lateral ventricles and the dentate gyrus of the hippocampus – continue neurogenesis. Indeed, adult neurogenesis is important for the maintenance of cognitive functions such as spatial memory (Imayoshi et al., 2008) and innate olfactory responses (Sakamoto et al., 2011). However, the mechanisms underpinning the sorting and maintenance of a fraction of embryonic NSCs as adult NSCs, and how the number of NSCs in the adult brain is governed, remain elusive.

The Hes family of basic helix-loop-helix (bHLH) transcriptional repressors (Hes1, Hes3 and Hes5) function downstream of Notch signaling (Kageyama and Ohtsuka, 1999), strongly inhibit neuronal differentiation, and maintain NSCs in the developing mammalian brain by repressing the proneural (neurogenic) bHLH factors such as Ascl1 and Neurog1/2 (Ishibashi et al., 1994; Ohtsuka et al., 1999, 2001; Hirata et al., 2000; Hatakeyama et al., 2004). Hes1 single mutant mice demonstrate severe phenotypes caused by accelerated neuronal differentiation, whereas Hes3 or Hes5 mutant mice exhibit only mild phenotypes owing to compensation by upregulated Hes1 (Ishibashi et al., 1995; Ohtsuka et al., 1999; Cau et al., 2000; Hirata et al., 2001). It is thus likely that Hes1 is an essential and dominant factor among the Hes family of repressors that maintains NSCs in an undifferentiated state.

It has been reported that certain alterations occurred in the cellular composition of the developing neocortex during mammalian brain evolution; the secondary germinal layer known as the outer subventricular zone (OSVZ) containing basal RGCs (bRGCs) emerged and distinctively developed during the course of evolution from lissencephalic to gyrencephalic mammals (Hansen et al., 2010; Molnár et al., 2011; Wang et al., 2011; Lui et al., 2011; Reillo et al., 2011; Hevner and Haydar, 2012; LaMonica et al., 2012). These bRGCs behave as an additional NSC pool that continues proliferation and neurogenesis during mid to late embryonic stages and contributes to the expansion of neuronal numbers and cortical surface areas in gyrencephalic mammals. bRGCs have similar molecular characteristics as those of apical RGCs (aRGCs); they are positive for Hes1, Pax6 and Sox2 (markers of NSCs), and negative for Tbr2 (also known as Eomes), a marker of IPCs.

To address how the manipulation of Hes1 expression in NSCs impacts morphology and cell composition of the developing and postnatal brain, we generated transgenic (Tg) mice in which Hes1 expression is controllable by using the Tet-On system. Here, we confirmed that Hes1 overexpression led to maintenance and expansion of NSCs in the embryonic brain, and found that later-born neurons (superficial layer neurons) were slowly produced and neurogenesis was prolonged, whereas gliogenesis was paradoxically accelerated in the Tg cortex. Notably, we observed that bRGC-like cells were increased in the SVZ of Tg cortex at late embryonic stages, suggesting their contribution to the prolonged neurogenesis observed in the Tg cortex. Furthermore, we observed that the number of slowly-dividing Pax6⁺ cells was increased and neurogenesis was enhanced, particularly when Hes1 expression was downregulated, in the dorsal and dorsolateral SVZ of the adult Tg brain, implicating an expanded reservoir of NSCs in the Hes1-overexpressing Tg brain.

We generated and crossed two Tg mouse lines (Fig. 1A); the reverse tetracycline-controlled transactivator (rtTA) is driven by the promoter and second intron of the nestin gene and expressed in NSCs and neural progenitors (pNestin-rtTA Tg) and, in the presence of doxycycline (Dox), rtTA binds to the tetracycline (Tet)-responsive element (TRE) and bi-directionally activates expression of Hes1 and d2EGFP, a destabilized variant of enhanced green fluorescent protein (EGFP) with a half-life of ∼2 h (TRE-Hes1/d2EGFP Tg). Thus, Hes1 and d2EGFP were overexpressed in NSCs and neural progenitors in double Tg mice when Dox was administered. GFP expression was detected in the brain, spinal cord, eyes and parts of the skin of these Hes1-overexpressing Tg mice (Fig. 1B; Fig. S1A,B). The size of embryonic and postnatal brains of Tg mice was slightly smaller than that of wild-type (WT) mice (Fig. S1B,C), and the difference in body size between Tg and WT pups became larger during postnatal stages (Fig. S1D,E). Immunohistochemistry on coronal brain sections using antibodies against GFP and Hes1 revealed that expression of d2EGFP was mostly restricted to the VZ, and high levels of Hes1 were co-expressed in GFP⁺ cells (Fig. 1C). Expression of d2EGFP and high levels of Hes1 were detected by 2 days after the start of continuous Dox (1.0 mg/ml in drinking water) administration to pregnant mice. Notably, the lateral ventricles were remarkably enlarged and the VZ (demarcated by immunostaining for Pax6, a marker of NSCs in developmental stages) was thicker in the Tg brain (Fig. 1D,E). In contrast, outer neuronal layers (indicated by immunostaining for the neuronal marker Tuj1), including the cortical plate (CP), intermediate zone (IZ) and SVZ, and the total cortical thickness were thinner compared with the WT brain. The ventral telencephalon including the ganglionic eminences was also thinner and hypoplastic in the Tg brain.

In Hes1-overexpressing Tg embryos, the VZ became thicker and expanded in tangential directions, whereas outer layers (including CP, IZ, and SVZ) and the total cortical thickness became thinner compared with the WT brain after embryonic day (E) 15.5 (Fig. 2A,B), suggesting that the NSC population was expanded and neurogenesis was suppressed by Hes1 overexpression. The length of ventricular surface of the dorsolateral telencephalon, measured on the coronal section, was markedly longer in the Tg brain, although the length of cortical surface was comparable with that of the WT brain (Fig. 2A,B). The expression of d2EGFP and Hes1 essentially overlapped with Pax6 staining. Notably, a higher number of Pax6⁺ cells remained in the VZ of the dorsolateral telencephalon (neocortical regions) of Tg mice until postnatal stages, whereas their number was substantially attenuated by postnatal day (P) 0 in the WT brain (Fig. 2A,B). It appeared that Pax6 expression was enhanced in some ventricular cells in the ganglionic eminences. It is possible that overexpression of Hes1 ectopically induced Pax6 expression in neural progenitors in the ventral telencephalon. Quantitative real-time RT-PCR revealed that Hes1 expression levels in the dorsolateral telencephalon in the Tg brain were 60 times higher than those in the WT brain at E15.5 (Fig. 2C). High levels of Hes1 and d2EGFP expression persisted in the VZ and SVZ of the Tg brain at later developmental stages and even after birth (Fig. 2D).

We performed in situ hybridization on coronal brain sections at E13.5 and estimated expression levels of Hes1 and the proneural (neurogenic) bHLH gene Neurog2 that is repressed by Hes1. Hes1 expression levels were upregulated but, in contrast, Neurog2 expression levels were markedly downregulated in the VZ of the Tg brain (Fig. 2E). Neurog2 expression levels were still lower at E17.5, and Neurog2-expressing cells were mainly distributed outside the VZ in the Tg brain at P0, whereas Neurog2 expression was mostly restricted to the VZ in the WT cortex. Immunohistochemistry using antibodies against Neurog2 revealed a significant decrease in Neurog2 expression levels in the VZ of the Tg brain at P0 (Fig. 2F). The majority of Neurog2⁺ cells were located outside the VZ, and Neurog2 expression was segregated from d2EGFP expression. These results indicated that overexpression of Hes1 inhibited neuronal differentiation and maintained NSCs in an undifferentiated state by repressing expression of proneural (neurogenic) genes such as Neurog2.

It has been reported that overexpression of Hes1 suppressed cell proliferation, and Hes1 represses the expression of cell cycle-related genes such as cyclin D1 (Ccnd1), cyclin E2 (Ccne2) and Gadd45g (Baek et al., 2006; Shimojo et al., 2008). Thus, we analyzed the rate of cell proliferation by 5-bromo-2′-deoxyuridine (BrdU) incorporation experiments, in which BrdU was intraperitoneally administered to pregnant mice 30 min before sacrifice, and the numbers of BrdU⁺ cells that were also positive for Pax6 or Tbr2, a marker of IPCs, most of which are located in the SVZ, were counted within a radial column of constant width (200 µm) in neocortical regions. We observed a decrease in the number of BrdU⁺ cells at E15.5 (Fig. 3A) and a significant decline in the proportions of BrdU⁺ cells of Pax6⁺ cells in neocortical regions of Tg brain at E13.5 and E15.5 (Fig. S2A,C). We performed immunohistochemistry using antibodies against phospho-histone H3 (pH3), a marker of dividing cells in the M phase; and Ki67 (Mki67), a marker of cycling cells. We observed that the numbers of pH3⁺ cells and Ki67⁺ cells in Tg mice were slightly less at E13.5 but significantly lower at E15.5 compared with those in WT mice (Fig. 3B,C). We then estimated the cell cycle length of Pax6⁺ NSCs and observed that both the length of S phase (Ts) and the total cell cycle length (Tc) were significantly elongated in the Tg brain compared with the WT brain at E15.5. This indicated that NSC proliferation was suppressed by Hes1 overexpression (Fig. 3D). Given that the VZ was expanded in tangential directions and neurogenesis was suppressed in the Tg brain, it was suggested that NSCs continued symmetric proliferative divisions until later embryonic stages. We thus analyzed the division mode of NSCs and observed that NSCs continued expansion of Pax6⁺ cells by symmetric proliferative divisions in the Tg brain, whereas significantly more NSCs produced Tbr2⁺ IPCs by asymmetric neurogenic divisions in the WT brain (Fig. 3E).

We next examined the expression levels of Ccnd1 in the Tg brain using antibodies against Ccnd1. At E13.5 and E15.5, expression levels of Ccnd1 and the number of Ccnd1⁺ cells were markedly reduced in the VZ of neocortical regions of Tg brain compared with the WT brain (Fig. 3F). At E17.5, expression levels of Ccnd1 were still lower in the VZ of Tg brain than in the VZ of the WT brain, although many Ccnd1⁺ cells were observed around the SVZ of the Tg brain. The reduced expression levels of Ccnd1 in the dorsolateral telencephalon of E15.5 Tg mice were confirmed by quantitative real-time RT-PCR (Fig. 3G). These results indicated that high levels of Hes1 expression maintained NSCs but inhibited cell proliferation, partly due to the downregulation of cell cycle-related genes including Ccnd1. We analyzed cell death using anti-cleaved-caspase 3 (Casp3) antibody and observed that cell death was not significantly enhanced in neocortical regions of the Tg brain at E13.5 and E15.5 (Fig. S3).

We have previously found that the transition timing from deep to superficial layer neurogenesis shifted earlier in the developing neocortex of Hes5-overexpressing Tg mice (Bansod et al., 2017). Therefore, we examined whether the timing of neurogenesis was disturbed in the Hes1-overexpressing Tg cortex by immunohistochemical analysis using markers for layer-specific neurons such as Tbr1 (layer VI), Ctip2 (Bcl11b; layer V), and Cux1 (layers II-IV). We observed that the onset of generation of early-born neurons (Tbr1⁺ layer VI neurons and Ctip2⁺ layer V neurons) was roughly comparable in Tg and WT mice (Fig. S4A). We subsequently performed birth date analyses by pulse labeling of DNA synthesis using BrdU. BrdU was intraperitoneally administered to pregnant mice at E13.5 or E14.5, and the fates and locations of BrdU-incorporated cells, which were born during the period of BrdU exposure, were analyzed by immunohistochemistry at P14. Most BrdU⁺ cells differentiated into neurons, migrated out of the VZ and settled in the cortical layers at P14 (Fig. 4A,B). In the dorsal part of the cortex, many cells that incorporated BrdU at E13.5 were Ctip2⁺ and were located in the deep layers (layer V), and some cells were Cux1⁺ and were distributed in the superficial layers (layers II-IV). There was no significant difference in the distribution of BrdU⁺ cells between the WT and Tg cortices (Fig. 4A). Although many cells that incorporated BrdU at E14.5 remained Ctip2⁺ and were located in the deep layers in the WT cortex, the vast majority of BrdU⁺ cells were Cux1⁺ and were distributed in the superficial layers in the Tg cortex (Fig. 4B). These results indicated that the switching from deep to superficial layer neurogenesis occurred earlier in the Tg cortex compared with the WT cortex (Fig. 4C).

We further performed birth date analysis by double pulse labeling using BrdU and 5-iodo-2′-deoxyuridine (IdU) that were intraperitoneally administered to pregnant mice at different embryonic stages. Cells that incorporated IdU at E13.5 were mostly located in the deep layers (layer V) and the deeper part of superficial layers (layer IV) in both WT and Tg mice at P14 (Fig. 4D,G). Cells that incorporated IdU or BrdU at E15.5 were located in the superficial layers (layers II-III) in both WT and Tg mice, although the number of IdU⁺ or BrdU⁺ cells was significantly less in the Tg cortex. In contrast, the number of cells that incorporated IdU or BrdU at E17.5 and settled in the most superficial layer of the cortex was increased in Tg mice compared with WT mice (Fig. 4D-G). To further characterize BrdU⁺ cells, sections were stained with layer markers such as Ctip2 and Cux1. We confirmed that most of the cells that incorporated BrdU at E15.5 and E17.5 (mostly neurons generated at these stages) differentiated into Cux1⁺ neurons (Fig. 4F). We analyzed the neuronal and glial fates of cells generated at E17.5 and found that a majority of cells that incorporated BrdU differentiated into NeuN⁺ (Rbfox3⁺) neurons in the Tg cortex at P14. In contrast, the number of BrdU⁺;GFAP⁺ cells was only slightly increased in the cortex and the SVZ (Fig. 4H). These results indicated that neurogenesis was decelerated around E15.5, but active neurogenesis persisted until later stages in the Tg cortex compared with the WT cortex, as illustrated by the schematic in Fig. 4I. Thus, the timing of neurogenesis was disrupted by Hes1 overexpression, especially during later developmental stages.

However, the cortical layer structure was not significantly disorganized in the postnatal brain, although all layers were thinner in the Tg cortex than in the WT cortex (Fig. S4B). The cortical thickness is generally thinner in the Tg brain during embryonic stages, but the difference became less after postnatal stages, probably because mice with the severe phenotype could not survive and only those with the mild phenotype survived after postnatal stages.

As neurogenesis persisted until later stages in the Tg brain, we investigated whether the onset of gliogenesis was shifted in the Tg brain, using glial markers such as GFAP, a marker of astrocytes; Olig2, a marker of oligodendrocytes; and PDGFRα, a marker of oligodendrocyte precursor cells. We found that high GFAP signals were focally detected in the VZ of neocortical regions of Tg brain at P0 (Fig. 5A), whereas only faint GFAP expression was observed in the same regions of WT brain. This observation indicated that astrogenesis prematurely occurred in the Hes1-overexpressing Tg brain, agreeing with the results in Hes5-overexpressing Tg mice (Bansod et al., 2017). Notably, many GFAP⁺ cells appeared outside the VZ/SVZ and were scattered throughout the Tg cortex at P7 (Fig. 5B), whereas GFAP⁺ cells were localized in the VZ/SVZ and superficial cortical layers of the WT brain. Similarly, many GFP⁺ cells were scattered throughout the Tg cortex at P7 (Fig. 5B), although expression of d2EGFP was mostly restricted to the VZ/SVZ until P0 (Fig. 5A). Notably, those GFP⁺ cells were negative for GFAP, and the expression of d2EGFP and GFAP was virtually non-overlapping in the Tg cortex (Fig. 5B). We observed that most GFP⁺ cells were Pax6⁺ but negative for neuronal markers such as NeuN (Fig. 5B). These results suggest that downregulation of Hes1 is necessary for terminal differentiation into mature (GFAP⁺) astrocytes and that those GFP⁺;Pax6⁺ cells were immature astrocytes in which terminal differentiation/maturation was prevented. The ectopic distribution of immature astrocytes (Pax6⁺) might have enhanced the production of GFAP⁺ astrocytes throughout the cortex. In contrast, we did not detect any notable differences in the oligodendrocyte lineage development between the WT and Tg brains (Fig. S5).

It has been shown that the majority of superficial layer neurons are generated by repeated proliferation of Tbr2⁺ IPCs in the SVZ (Tarabykin et al., 2001; Noctor et al., 2004; Arnold et al., 2008). Thus, we analyzed the number and proliferation rate of Tbr2⁺ cells in the WT and Tg brains. The number of Tbr2⁺ cells was not significantly different between WT and Tg mice until E13.5 (Fig. S6A). However, at later stages (E15.5 and E17.5), the number of Tbr2⁺ IPCs in neocortical regions of the Tg brain was significantly less than that in the WT brain (Fig. 6A; Fig. S6A). pH3 staining also demonstrated that the number of dividing cells in both the VZ and SVZ of the Tg brain was significantly lower than that in the WT brain at E15.5 as shown in Fig. 3B, in agreement with the above observation that the number of Tbr2⁺ IPCs was markedly diminished in the SVZ (Fig. 6A; Fig. S6A), in addition to the observation that the proliferation rate of Pax6⁺ NSCs was suppressed in the VZ (Fig. S2A,C). However, double labeling using antibodies against Tbr2 and BrdU (administered 30 min before sacrifice) revealed that the proliferation rate of Tbr2⁺ cells was not suppressed in the Tg brain and was comparable to that in the WT brain at E13.5 and E15.5 (Fig. S2B,D). These results suggest that the production of Tbr2⁺ IPCs from NSCs was inhibited by Hes1 overexpression, potentially leading to the decelerated generation of later-born neurons.

Meanwhile, we noticed that Pax6⁺ cells were remarkably increased in place of Tbr2⁺ cells outside the VZ in the Tg brain at E17.5 (Fig. 6A). Double labeling for pH3 and Pax6 revealed significantly more pH3⁺;Pax6⁺ cells outside the VZ in the Tg brain at E15.5 and E17.5 (Fig. 6B, arrows), whereas in the WT brain, most pH3⁺;Pax6⁺ cells were located at the apical (ventricular) surface in the VZ where aRGCs are dividing in the M phase. We thus proceeded to characterize these pH3⁺;Pax6⁺ cells outside the VZ in more detail.

It has been reported that bRGCs only retain basal radial processes, lack apical processes and are located in the SVZ (Wang et al., 2011). bRGCs manifest similar characteristics to those of aRGCs; they are positive for Hes1, Pax6 and Sox2, but negative for Tbr2. Thus, we examined whether the pH3⁺;Pax6⁺ cells outside the VZ in the Tg brain were bRGCs. We performed immunostaining with antibodies against phosphorylated vimentin (pVim), a specific marker of RGCs in the M phase (Kamei et al., 1998). Although pVim only labels a subset of RGCs that are dividing in the M phase, it is considered a useful marker to demonstrate increments in delaminated RGCs such as bRGCs (Pilz et al., 2013). Whereas pVim⁺ cells were mostly restricted to the apical (ventricular) surface in the VZ where aRGCs are dividing, and only a few cells were aberrantly detected outside the VZ in the WT brain, pVim⁺ cells markedly increased in number outside the VZ in the Tg brain at E17.5 (Fig. 6C,D). We confirmed that these aberrant pVim⁺ cells were also positive for Pax6 (Fig. 6C) and pH3 (Fig. 6D), similar to dividing aRGCs.

We then attempted immunostaining with the RGC marker GLAST (Slc1a3), which labels radial fibers of RGCs. However, it was difficult to determine whether these pH3⁺ cells in the SVZ were bRGCs that indeed lacked apical processes due to the exuberant radial fibers, although many pH3⁺ cells in the SVZ were co-labeled with GLAST (Fig. S6B). Therefore, we introduced mCherry expression vectors (pEF-mCherry) into the VZ cells by in utero electroporation at E15.5 and analyzed the morphology of transfected cells at E17.5. Although the majority of transfected cells differentiated into neurons or remained as aRGCs in the VZ, a subset of mCherry⁺ cells adopted a bRGC-like morphology; they were positive for Pax6 and located outside the VZ, retaining only basal radial processes but lacking apical processes (Fig. 6E; Fig. S6C,D; Movie 1). This suggested that a fraction of transfected cells transformed from aRGCs to bRGCs in the two days.

We further revealed that a majority of Pax6⁺ cells outside the VZ in the Tg brain were positive for Neurog2 at E17.5, suggesting that they retained neurogenic potentials (Fig. 6F). These results collectively demonstrated that the number of bRGC-like cells increased at later embryonic stages following Hes1 overexpression, instead of the suppressed production of Tbr2⁺ IPCs, in agreement with the localization patterns of Neurog2⁺ cells (Fig. 2E,F) and Ccnd1⁺ cells (Fig. 3F) in the SVZ. These findings suggested that, although switching from deep to superficial layer neurogenesis and onset of gliogenesis shifted earlier in neocortical regions of Tg brain, generation of superficial layer neurons was prolonged by neurogenesis from those bRGC-like cells in the SVZ, as illustrated by the schematic in Fig. 6G.

Next, we addressed whether NSCs were retained more abundantly and whether neurogenesis was enhanced in the postnatal and adult brains of Hes1-overexpressing Tg mice. Higher numbers of Hes1⁺/d2EGFP⁺ and Pax6⁺ cells were observed in the SVZ of the rostral lateral ventricle, from which adult neurogenesis continues to supply newly born neurons to the olfactory bulbs, in postnatal (P14) and young adult (4 weeks after birth) Tg brains compared with the WT brains (Fig. 7A). We also observed that the expression of both d2EGFP and Hes1 was considerably diminished within 3 days after withdrawal of Dox in the Tg brain (4 weeks and 3 days) (Fig. 7A). Next, we estimated proliferation activity in the WT and Tg brains at 6 weeks after birth. When mice were administered with Dox from E11.5 to P35 (5 weeks) and sacrificed at P43 after 1-day administration of BrdU in drinking water from P42 (6 weeks), markedly more BrdU⁺ cells were observed in the SVZ of the lateral ventricle in the Tg brains compared with the WT brains (Fig. 7B). We then performed immunostaining with antibodies against doublecortin (Dcx), a marker of newly born neurons. When mice were treated with Dox from E7.5 to P28 (4 weeks) and sacrificed at P31 after 3-day administration of BrdU in drinking water from P28, the number of Dcx⁺ cells was markedly higher in the SVZ of the lateral ventricle in the Tg brains compared with the WT brains (Fig. 7C). Most of the Dcx⁺ cells were BrdU⁺, indicating that these cells incorporated BrdU and differentiated into neurons during the 3 days before sacrifice. These newly born neurons were broadly distributed along the extended ventricular surface. These results collectively implied the presence of an expanded NSC pool in the adult Tg brain, which retained the competence to continue adult neurogenesis and was enhanced by the attenuation of Hes1 expression (Fig. 7A,B).

Furutachi et al. have reported the existence of a subpopulation of NSCs (that would be the origin of postnatal and adult NSCs) in the embryonic mouse brain from early developmental stages (Furutachi et al., 2015). Those NSCs were slowly-dividing and thus retained fluorescent marker proteins (GFP) without dilution until later stages. We therefore investigated whether the suppressed proliferation rate caused by Hes1 overexpression led to the expansion of postnatal and adult NSCs. First, we administered EdU at E11.5 and estimated the retention of EdU in postnatal and adult SVZ cells. Many EdU label-retaining cells were observed in the dorsal and dorsolateral SVZ of the anterior lateral ventricle in the Hes1-overexpressing Tg brain at P0 (Fig. 7D, arrows), whereas only a few EdU label-retaining cells were observed in the SVZ of the WT brain. EdU label-retaining cells in the SVZ of the Tg brain were mostly Pax6⁺, suggesting that these slowly-dividing cells remained as NSCs. Pax6⁺ clusters sometimes exhibited rosette-like structures in the Tg brain, as shown in Fig. 7D, probably owing to the heterogeneous organization within the VZ/SVZ, although this phenotype was not typical nor reproducible. These results suggested that NSCs that strongly expressed Hes1 existed as slowly-dividing cells in the embryonic brain and remained as NSCs in postnatal and adult brains. Therefore, we next evaluated Hes1 expression levels in EdU label-retaining cells in the WT brain using anti-Hes1 antibody to verify whether NSCs that expressed higher levels of endogenous Hes1 remained as slowly-dividing NSCs in postnatal and adult brains. However, we did not detect significantly higher levels of Hes1 expression in EdU label-retaining cells in the SVZ of embryonic and postnatal brains (Fig. S7). These results suggested that, although Hes1 can expand the NSC reservoir in the adult brain, it is not involved in formation of endogenous adult NSCs.

In the embryonic brains of Hes1-overexpressing Tg mice, neuronal differentiation was suppressed and NSCs were maintained in the enlarged VZ. As NSCs continued symmetric proliferative divisions (Fig. 3E), the VZ extended tangentially, and the ventricular surface and ventricles were expanded (Fig. 2A,B). Furthermore, we observed that the number of Pax6⁺ NSCs was increased in the Tg brain (Fig. 2B), although the proliferation rate of NSCs was reduced, as shown in Fig. 3. This result indicated that the increase of Pax6⁺ NSCs owing to the elongation of the period of symmetric proliferative divisions prevailed over their low proliferation rate.

Regardless of the expansion of NSCs in embryonic brains, Hes1-overexpressing Tg mice exhibited smaller brains and bodies (Fig. S1). This is partly because inhibition of neuronal differentiation impaired brain maturation. Actually, both the cortical wall and ventral part of the telencephalon were thinner, and some brain structures such as the hippocampus and corpus callosum appeared to be hypoplastic in Tg mice compared with WT mice. In addition, suppression of cell proliferation by Hes1 overexpression may have contributed. It has been reported that the appropriate levels of Hes1 promote cell proliferation (Murata et al., 2005) but, on the other hand, Hes1 overexpression inhibits cell proliferation (Baek et al., 2006; Shi et al., 2015). Several cell cycle-related genes were found to be repressed by Hes1 (Georgia et al., 2006; Monahan et al., 2009), and we have previously demonstrated using cDNA microarrays that Ccnd1 expression was downregulated by Hes1 overexpression (Shimojo et al., 2008). Here, we confirmed that Ccnd1 expression was indeed suppressed in the VZ of Hes1-overexpressing Tg brain compared with the WT brain. We hypothesized that abolishing exogenous Hes1 expression by withdrawal of Dox would induces active neurogenesis and increases neuronal number and brain size, as NSC numbers were increased at ∼E15.5 and thereafter. We attempted the withdrawal of Dox at various developmental stages but failed to observe larger brains in Tg mice. Thus, it would be a forthcoming challenge to test whether it is possible to achieve brain expansion by rescuing cell cycle-related genes such as Ccnd1 to release the blockade of cell cycle progression.

In certain mammalian species such as dogs, sheep and primates, the brain surface is convoluted (gyrencephalic), in contrast to the smooth brain surface of lissencephalic mammals such as mice, and it has been thought that mammals have developed gyrencephalic brains during mammalian evolution. It has been revealed that the development of OSVZ and bRGCs led to the expansion of the stem cell pool, increased production of superficial layer neurons, and promoted enlargement of the brain surface.

In the Hes1-overexpressing Tg brains at late embryonic stages, many pH3⁺;Pax6⁺ cells were scattered in the SVZ, whereas Tbr2⁺ IPCs were decreased in number. We attempted to characterize these Pax6⁺ cells outside the VZ to elucidate whether they have morphology and properties corresponding to bRGCs, but it was highly challenging to identify what proportion of these cells were bRGCs owing to the lack of specific markers for bRGCs. However, based on complementary experiments, a subset of these cells appeared to exhibit bRGC-like morphology and properties, and pVim staining revealed a marked increment in the number of mitotic RGCs in the SVZ of neocortical regions in the Tg brain. Although the mechanisms underlying the enhanced production of bRGC-like cells by Hes1 overexpression remain unclear, it is possible that high levels of Hes1 forcibly arrested neuronal differentiation of prospective neuronal daughters after asymmetric divisions of aRGCs, which normally produce a stem cell daughter and neuronal daughter, and retained them as bRGC-like NSCs in the SVZ. This prevention of the production of neuronal daughters may account for the reduction in Tbr2⁺ IPCs in the SVZ. Growing evidence revealed the involvement of various genes and signaling pathways in the generation and expansion of bRGCs (Penisson et al., 2019). However, there have been few studies that demonstrated the direct contribution of Notch signaling to enhanced generation of bRGCs, although Shitamukai et al. reported the implication of Notch signaling in the retainment of basal processes of NSCs and the maintenance of newly born bRGC-like cells (Shitamukai et al., 2011). Our findings provided further supportive evidence implicating the contribution of Notch signaling to the expansion of bRGCs during mammalian brain evolution.

Expansion of both basal progenitors (bRGCs and IPCs) appears to be necessary to achieve neocortical expansion as seen in gyrencephalic mammals. Nevertheless, in the Hes1-overexpressing Tg brain, only bRGC-like cells increased at the expense of IPCs, leading to failure of neocortical expansion. To achieve a bigger brain with a larger neocortical surface, alternative strategies for promoting the production and proliferation of IPCs while maintaining NSCs (aRGCs and bRGCs) would be required.

It is well established that neurogenesis continues in specific regions of the adult brain: the SVZ of the rostral lateral ventricle and the dentate gyrus of the hippocampus. NSCs in these neurogenic regions gradually decline in number, and neurogenesis is attenuated with aging. We observed substantially more Pax6⁺ cells in the SVZ of postnatal and adult brains of Hes1-overexpressing Tg mice, especially in the dorsal telencephalon. Moreover, we observed an enhanced generation of Dcx⁺ cells in the dorsal and lateral part of the enlarged SVZ in the adult Tg brain, when Dox was withdrawn several days before sacrifice, thereby attenuating Hes1 expression and releasing subsequent inhibition of neuronal differentiation. These results suggest that elevation of Hes1 expression levels in NSCs results in expansion of the NSC reservoir with neurogenic potential. Conventional therapeutic strategies are aimed at compensating for the declined neurogenesis in aged brains or brains of neurodegenerative disorders, such as Alzheimer's disease, by maintaining adult NSCs and enhancing adult neurogenesis by activating pre-existing quiescent NSCs. In contrast, our study sheds light on the unexploited aspect of neuroregenerative therapy by aiming to expand the NSC reservoir with the potential to generate neurons to replenish impaired brain regions, leading to improved regenerative capacity against brain damage or neurodegenerative disorders.

Another candidate resource of newly born neurons is the GFP⁺;Pax6⁺ cell population observed throughout the neocortex of postnatal Tg brains (Fig. 5B). It is possible that these cells are immature glial cells in which further differentiation/maturation is suppressed by Hes1 overexpression. In other words, this suggests that downregulation of Hes1 is necessary for terminal differentiation into mature (GFAP⁺) astrocytes. These cells may retain stemness or have the potential to dedifferentiate and commence neurogenesis. Thus, investigating whether these cells produce new neurons when brain tissues are damaged by physical injury, ischemia or neurodegeneration will be of interest. Unfortunately, rtTA expression in the hippocampus was very weak, and we could not detect GFP expression in the hippocampus of the postnatal Tg brain. Therefore, it was difficult to analyze the effect of Hes1 overexpression on NSCs in the hippocampus, including the dentate gyrus.

We observed that many NSCs with low proliferation rate due to Hes1 overexpression were maintained as EdU label-retaining Pax6⁺ cells in the dorsal and dorsolateral SVZ of the postnatal Tg brain. This result suggests that slowly-dividing NSCs more likely remain as NSCs in postnatal and adult brains, in agreement with a previous report (Furutachi et al., 2015). Therefore, we expected expression levels of endogenous Hes1 to be higher in slowly-dividing NSCs (prospective adult NSCs) during normal brain development. However, we observed that Hes1 expression levels were not significantly higher in EdU label-retaining cells in the postnatal WT brain. This result indicates that high levels of Hes1 expression are sufficient to maintain slowly-dividing NSCs as postnatal NSCs but are not essential for the retention of slowly-dividing NSCs during normal brain development. This suggests that the proliferation rate of NSCs is regulated by other factors with activities similar to those of Hes1, such as other Hes family, Hes-related factors, Id protein family, or other cell cycle regulatory proteins. In this context, it was reported that inactivation of Hes1, Hes3, Hes5 and Hey1 depleted virtually all NSCs in the adult brain (Sueda et al., 2019). Uncovering the mechanisms underlying the maintenance of adult NSCs and establishing effective measures to expand them will enable brain modifications to promote higher regenerative capacity.

For pNestin-rtTA transgene, an 8-kb SmaI fragment containing the ZGF cassette, which has a 5.8-kb fragment of the Nestin promoter region, polyadenylation sequence of SV40, and 1.8-kb fragment of the second intron of Nestin gene (Mignone et al., 2004), was subcloned into the SalI site of pBluescript SK- vector using NotI linkers (Takara). A 0.8-kb rtTA-Advanced fragment was excised from pTet-On Advanced vector (Clontech Laboratories) by EcoRI-BamHI digestion and subcloned into the SalI site downstream of the Nestin promoter using SalI linkers (Takara). For microinjection, the 8.8-kb NotI fragment containing the transgene was separated by agarose gel electrophoresis, purified using QIAEXII Gel Extraction Kit (Qiagen), and eluted with TE. For TRE-Hes1/d2EGFP transgene, a 0.9-kb fragment of mouse Hes1 cDNA with ClaI linkers and a 0.9-kb BamHI-NotI fragment of d2EGFP from the pd2EGFP-1 vector (Clontech Laboratories) with EcoRI linkers were inserted into MCS-I and MCS-II of the pTRE-Tight-BI vector (Clontech Laboratories), respectively (Fig. 1A). For microinjection, a 2.9-kb fragment containing the bilateral expression cassette from polyA for MCS-I to polyA for MCS-II was excised using NspI-BspHI, separated by agarose gel electrophoresis, purified using QIAEXII Gel Extraction Kit, and eluted with TE.

Transgenic mice were generated by pronuclear microinjection using fertilized eggs from ICR mice. Transgenic founders and progeny were identified by PCR analysis of tail DNA using primers that amplified the full-length of rtTA, Hes1, or d2EGFP. Both lines were crossed and doxycycline hyclate (Sigma-Aldrich; 1.0 mg/ml) in drinking water with 5% sucrose was administered to pregnant mice. Double transgenic mice were identified by PCR analysis and judged by GFP fluorescence in the brain and eyes. Images of whole bodies and brains 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.

For immunohistochemical analysis, brains were excised, fixed in 4% paraformaldehyde (PFA), cryoprotected, embedded in OCT compound, and cryosectioned at 16 µm. Fixed cryosections were washed with phosphate-buffered saline (PBS), preincubated in PBS containing 5% normal goat serum and 0.1% Triton X-100, then incubated in 1% normal goat serum and 0.1% Triton X-100 with primary antibodies overnight at 4°C followed by secondary antibodies for 1–3 h at room temperature. Primary antibodies used in this study are listed in Table S1. Primary antibodies were detected with Alexa Fluor-conjugated secondary antibodies (1:200; Molecular Probes), and cell nuclei were visualized with 4′,6-diamidino-2-phenylindole (DAPI; Sigma-Aldrich). For Hes1 staining, antigen retrieval was performed in 0.1% Tween 20/0.01 M citrate buffer (pH 6.0) using an autoclave for 15 min at 105°C, and sections were incubated with primary antibodies in Can Get Signal Solution B (Toyobo) overnight at 4°C followed by secondary antibodies (HRP-conjugated donkey anti-rabbit IgG, 1:500; GE Healthcare) for 90 min at room temperature. Color development was enhanced by the TSA amplification system (PerkinElmer) according to the manufacturer's instruction. Fluorescent sections were imaged using a Zeiss LSM510 confocal microscope or a Keyence BZ-X700 fluorescence microscope.

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

In situ hybridization was performed as previously described (Ohtsuka et al., 2011). Digoxigenin-labeled antisense RNA probes were synthesized in vitro using the full-length cDNAs of Hes1 (NM_008235.2) and Neurog2 (NM_009718.2) as templates, and hybridized to brain cryosections. Labeled preparations were imaged using a Zeiss Axiophot microscope equipped with an AxioCam color CCD camera.

To analyze the cell proliferation rate, 5-bromo-2′-deoxyuridine (BrdU; Sigma-Aldrich) (50 μg BrdU/g of body weight) was injected intraperitoneally into pregnant mice 30 min before sacrifice, and embryonic brains were fixed with 4% PFA. For the birth date analysis, we conducted a double pulse labeling of DNA synthesis using BrdU and 5-iodo-2′-deoxyuridine (IdU; Sigma-Aldrich). BrdU (50 μg BrdU/g of body weight) or IdU (50 μg IdU/g of body weight) was injected intraperitoneally into pregnant mice at different embryonic stages, and brains were fixed by transcardial perfusion with 4% PFA at postnatal day 14 (P14). The fixed cryosections were incubated in 2 N HCl solution for 30 min at 37°C, followed by neutralization in 0.1 M sodium tetraborate buffer. Sections were incubated with mouse anti-BrdU antibody (1:100; Becton Dickinson) and rat anti-BrdU antibody (1:500; Serotec) overnight at 4°C, and then with Alexa594-conjugated goat anti-mouse IgG and Alexa488-conjugated goat anti-rat IgG (1:200; Molecular Probes). As mouse anti-BrdU antibody reacts with both BrdU and IdU whereas rat anti-BrdU antibody reacts only with BrdU, cells that incorporated IdU were stained in red whereas cells that incorporated BrdU were depicted in yellow (red+green).

Estimation of the cell cycle length was performed as previously described (Watanabe et al., 2015). We conducted a dual pulse labeling of DNA synthesis using BrdU and EdU (5-ethynyl-2′-deoxyuridine) (Molecular Probes), referring to the previous methods (Martynoga et al., 2005; Mairet-Coello et al., 2012). BrdU (50 μg BrdU/g of body weight) and EdU (12.5 μg EdU/g of body weight) were injected intraperitoneally into pregnant mice 2 h and 30 min before sacrifice, respectively, and the ratios of Pax6⁺ cells that incorporated either or both BrdU and EdU were analyzed to estimate the cell cycle length.

For the pEF-mCherry vector, the coding sequence of mCherry was cloned into the pEF (human elongation factor 1α promoter)-MM expression vector, which was modified from the pEF-BOS vector (Mizushima and Nagata, 1990). To label the basal radial glia-like cells, in utero electroporation was performed with E15.5 pregnant mice using methods described previously (Ohtsuka et al., 2011). Embryos were harvested 2 days after electroporation at E17.5. Brains were excised, fixed in 4% PFA, cryoprotected, embedded in OCT, and cryosectioned at 16-70 µm.

To estimate the division mode (symmetric vs asymmetric division) of NSCs, we first performed in utero electroporation using the pEF-mCherry vector to only label the VZ cells (apical NSCs) at E13.5. EdU (12.5 μg EdU/g of body weight) was injected intraperitoneally into pregnant mice 6 h after in utero electroporation, and mice were sacrificed 18 h after EdU injection at E14.5. The numbers of Pax6⁺;EdU⁺/EdU⁺ cells and Tbr2⁺;EdU⁺/EdU⁺ cells were counted within a radial column of 200 μm width on the coronal sections of neocortical regions.

EdU (12.5 μg EdU/g of body weight) was injected intraperitoneally into pregnant mice at E11.5, and the detection of EdU-labeled cells was performed based on a fluorogenic click reaction (Salic and Mitchison, 2008) at P0.

Fixed and cryoprotected brains were cryosectioned at 70 μm. Fixed cryosections were incubated with anti-mCherry antibody 2-3 overnight at 4°C followed by secondary antibody for 2-3 h at room temperature. Fluorescent images were taken using a Zeiss LSM510 confocal microscope, combined to create z-stack images including full-thickness of brain sections, and converted to 3D movies using ImageJ/Fiji (NIH, USA).

Cell numbers within a radial column of 200 μm width (Figs 2B, 3A-C,E,F, 4A,B and 6A, Figs S2C,D, and S6A), 500 μm width (Fig. 6B,C), or 1 mm width (Fig. 4G) on the coronal sections of neocortical regions were counted. In Fig. 4H, cell numbers in a 450 μm×450 μm square in the cortex and those within a radial column of 450 μm width in the SVZ of cortical area were counted. In Fig. 5B, cell numbers in a 500 μm×500 μm square in the cortex were counted. The measurement of the length of ventricular and cortical surface (from the dorsal ridge to the dorsolateral border) was performed by using ImageJ/Fiji. Hes1 expression level in the postnatal SVZ was analyzed by measuring the signal intensity of Hes1 immunostaining in a 500 μm×100 μm strip in the SVZ along the lateral ventricular surface by using ImageJ/Fiji.

Each experiment was performed with at least three independent samples. Results are shown as mean±s.e.m. Statistical differences were examined with unpaired two-tailed Student's t-test.

We thank Hitoshi Miyachi and Satsuki Kitano for their help in the generation of pNestin-rtTA transgenic mice and TRE-Hes1/d2EGFP transgenic mice.

The peer review history is available online at https://dev.biologists.org/lookup/doi/10.1242/dev.189191.reviewer-comments.pdf