Asymmetric inheritance of Cyclin D2 maintains proliferative neural stem/progenitor cells: A critical event in brain development and evolution

Authors

  • Yuji Tsunekawa,

    1. Gene Expression Laboratory, Salk Institute for Biological Studies, La Jolla, California, USA
    Search for more papers by this author
  • Takako Kikkawa,

    1. Department of Developmental Neuroscience, United Centers for Advanced Research and Translational Medicine (ART), Tohoku University School of Medicine, Aoba-ku, Sendai, Japan
    Search for more papers by this author
  • Noriko Osumi

    Corresponding author
    1. Department of Developmental Neuroscience, United Centers for Advanced Research and Translational Medicine (ART), Tohoku University School of Medicine, Aoba-ku, Sendai, Japan
    Search for more papers by this author

Abstract

Asymmetric cell division and cell cycle regulation are fundamental mechanisms of mammalian brain development and evolution. Cyclin D2, a positive regulator of G1 progression, shows a unique localization within radial glial (RG) cells (i.e., the neural progenitor in the developing neocortex). Cyclin D2 accumulates at the very basal tip of the RG cell (i.e., the basal endfoot) via a unique cis-regulatory sequence found in the 3′ untranslated region (3′UTR) of its mRNA. During RG division, Cyclin D2 protein is asymmetrically distributed to two daughter cells following mitosis. The daughter cell that inherits Cyclin D2 mRNA maintains its self-renewal capability, while its sibling undergoes differentiation. A similar localization pattern of Cyclin D2 protein has been observed in the human fetal cortical primordium, suggesting a common mechanism of maintenance of neural progenitors that may be evolutionarily conserved across higher mammals such as primates. Here, we discuss our findings and the Cyclin D2 function in mammalian brain development and evolution.

Introduction

During cortical development in mammals, the expansion of the cortical wall relies on the generation of large numbers of neurons by neural stem/progenitor cells (NSPCs) situated in the inner wall of the neural tube termed the ventricular zone (VZ). NSPCs are highly polarized, stretching their apical and basal processes towards the ventricular (apical) surface and the pial (basal) surface of the cortical primordium. At early stages of corticogenesis, NSPCs divide symmetrically to increase their numbers (embryonic day 9 [E9] to E11 in mice) (Haubensak et al. 2004; Radakovits et al. 2009). As development proceeds (starting around E11 in mice), NSPCs become longer and thinner and are called radial glia (RG) because they support the radial migration of cortical neurons (Noctor et al. 2001; Gotz & Huttner 2005). The RG cells divide asymmetrically and produce one apical progenitor cell (AP) and one neuronal cell or intermediate progenitor cell (IP) (Gotz & Huttner 2005; Huttner & Kosodo 2005). APs continue to divide asymmetrically, thereby increasing the number of neuronal cells while maintaining the number of APs. Newly produced neurons migrate out of the VZ to form the cortical plate (CP), while IPs divide symmetrically in the upper region of the VZ (i.e., the subventricular zone; SVZ) and generate a pair of IPs or neurons (Haubensak et al. 2004; Miyata et al. 2004; Noctor et al. 2004) (Fig. 1).

Figure 1.

Schematic depiction of cortical development in the mouse. At the early stage of corticogenesis (~E9.5–11.5), neuroepithelial cells divide symmetrically to yield more progenitors, resulting in a thickened pseudo-stratified sheet of mitotic cells concentrated mainly on the apical side of the epithelium (the proliferation period). Later in corticogenesis (~E12.5–15.5), neuroepithelial cells become long and thin radial glial cell (RG) and start to divide asymmetrically (the neurogenic period). RGs produce self-renewing apical progenitors (APs), together with terminally differentiated neurons (blue), intermediate progenitors (IPs, green), or outer radial glial cells (oRGs, yellow).

One of the specific characteristics that distinguish RG cells from the other types of progenitors present during neurogenesis is their highly polarized morphology. Various molecules have unique distribution patterns within RG cells (Fig. 2). It is well known that cell adhesion-related molecules such as cadherins and ZO-1 as well as polarity proteins including Par complexes are located at the apical most area of the RG (Aaku-Saraste et al. 1996; Kadowaki et al. 2007; Kosodo et al. 2008). Par3, a key component of the Par protein complex, regulates the asymmetric cell division of RG cells (Bultje et al. 2009). Notably, the centrosome of RG cells remains near the apical surface during most phases of the cell cycle, and centrosomal proteins such as gamma tubulin, pericentrin, and ninein are thus accumulated at the apical side (Shinohara et al. 2013). The apically located centrosome seems to be important for the proper movement of the nucleus during the cell cycle (i.e., interkinetic nuclear movement) (Tamai et al. 2007; Shinohara et al. 2013). Another important molecule located in the apical plasma membrane is prominin-1 (CD133), which is a marker for NSPCs (Kosodo et al. 2004). However, this molecule seems to be involved in the symmetric cell division that is typical of early NSPCs because a Prominin-1-enriched apical membrane is one characteristic of neuroepithelial cells undergoing symmetric division (Kosodo et al. 2004; Dubreuil et al., 2007). Within the nucleus, various transcription factors (TFs) activate or repress suites of downstream genes that determine the cell fate of RG cells. Sox2, Pax6, and Hes1 are essential TFs for RG progenitor identity and potential (Gotz et al. 1998; Ellis et al. 2004; Hatakeyama et al. 2004), whereas Neurog2 is a “proneural” bHLH TF that promotes neuronal fate commitment (Ma et al. 1996). Pax6 is also involved in neuronal differentiation through the direct regulation of Neurog2 and indirectly via the TF Dmrta1 (Scardigli et al. 2003; Kikkawa et al. 2013). In the cytoplasm of RG cells, some molecules seem to function to maintain the character of RG cells. For example, fatty acid binding protein 7 (Fabp7/BLBP), a widely used marker for NSPCs, maintains the proliferation of RG cells (Arai et al. 2005). Notably, another sugar-related molecule, Lewis X, which is generated by fucosyltranscription factor FucT9, is specifically expressed by RG cells (Shimoda et al. 2002). At the basal-most area, localized integrins receive signals from the extracellular matrix (ECM) situated above the RG cells. Mice lacking the integrin family of molecules exhibit disrupted interaction between RG cells and the basal lamina in the cortical marginal zone, indicating that ECM signals are important for the maintenance of RG cells (Graus-Porta et al. 2001; Kowalczyk et al. 2009). In this review, we focus on another basally located molecule, Cyclin D2.

Figure 2.

Unique distribution pattern of various molecules within and outside of the radial glial (RG) cell. For details, see the text.

Mammalian brain development and asymmetric cell division

Cell division is a pivotal aspect of organism development. After numerous cell divisions, a single fertilized egg will give rise to every cell type in the body, which has an estimated 1013 cells in the case of humans. During this process, asymmetric cell division plays a role in the generation of the huge variety of cell types while also increasing the number of cells in the body as a whole (Knoblich 2010). During asymmetric cell division, the two daughter cells unequally inherit many components such as protein, RNA and organelles, which eventually gives them a different fate (Knoblich 2008, 2010). For example, in Drosophila melanogaster neuroblast (i.e., neural progenitor cell), Pros, Numb, and Miranda are localized at the basal cortex of the neuroblast, whereas Baz, Par6, and aPKC localize apically to be segregated asymmetrically into one of each of the daughter cells (Gonczy 2008). The asymmetric cell division of mammalian NSPCs is also critical for the establishment of the cerebral cortex architecture by regulating the balance between proliferative and neurogenic populations. This balance is achieved by producing daughter cells that become postmitotic neurons together with daughter cells that are self-renewing, thereby increasing the number of neurons while maintaining the number of APs (Fishell & Kriegstein 2003; Gotz & Huttner 2005). Even though some of the important feature is conserved from invertebrates (Delaunay et al. 2014), the asymmetric cell division of mammalian APs in the developing brain is more complicated than that of Drosophila melanogaster neuroblasts. The cleavage plane of a dividing cell is not always vertical or horizontal, as in a neuroblast (however, a certain percentage of horizontal cell division seems to occur) (Chenn & Mcconnell 1995; Paridaen et al. 2013; Delaunay et al. 2014), but is rather oblique with various division planes (Shitamukai et al. 2011). In early stages of cortical development, neuroepithelial cells divide with a cleavage plane that is vertical to the apical membrane and symmetrically inherit all cell components, such as the basal process and apical membrane (Costa et al. 2008). Later, in the neurogenic stage when neurons begin to be produced by asymmetric cell division, the basal process and apical membrane are reported to be inherited asymmetrically by one of the two daughter cells, and it has been proposed that these components function as cell fate determinants (Miyata et al. 2001; Noctor et al. 2001; Kosodo et al. 2004; LaMonica et al. 2013). As mentioned above, the molecules localized in the apical region (e.g., Numb, Prominin1, and Par complex proteins) have been well characterized (Kosodo et al. 2004, 2008; Bultje et al. 2009). However, relatively little information is available for molecules on the basal side. We have thus previously set out to explore the role of the basal region, in particular the contribution resulting from the polarized distribution of Cyclin D2, a key regulator of cell division.

Mammalian brain development and cell cycle regulators

Many reports have indicated that cell cycle regulators such as cyclins and cyclin-dependent kinases and their inhibitors have multiple functions during several stages of neuronal development. One of the most important observations to date is the correlation between G1-phase lengthening and AP cell fate determination. Classical cumulative labeling using bromodeoxyuridine (BrdU) and time lapse imaging studies of brain slices indicated that the percentage of G1-phase AP cells increases during mammalian neuronal development (Takahashi et al. 1994, 1995; Calegari & Huttner 2003; Dehay & Kennedy 2007; Lange & Calegari 2010; Salomoni & Calegari 2010; Betizeau et al. 2013). It has also been reported that the overexpression of Cyclin D1 and cyclin E1, cyclins known to have essential functions during the G1-S phase transition, along with their partner kinase CDK4 shortens the G1-phase, leading to downregulation of neurogenesis and increased populations of APs and IPs (Lange et al. 2009; Pilaz et al. 2009; Lim & Kaldis 2012; Tsunekawa & Osumi 2012; Tsunekawa et al. 2012). Moreover, genetic ablation of CDK2 and CDK4 increases the G1-phase, which eventually increases neurogenic cell division during mammalian cortical development (Lim & Kaldis 2012). However, data that contradict this scenario have also been reported (Arai et al. 2011). Interestingly, cell cycle regulators function not only to determine cell fates by controlling the cell cycle but also to modulate the radial migration of post mitotic neurons (Nguyen et al. 2006a; Frank & Tsai 2009; Kawauchi et al. 2013).

Mammalian brain development and Cyclin D2

Mouse Cyclin D2 was first identified in a screen for delayed early response genes induced by colony-stimulating factor 1 and recognized as a member of a family that includes at least two other related genes, Cyclin D1 and D3 (Matsushime et al. 1991). Cyclin D2 protein forms a complex with cyclin-dependent kinases (Cdk) 4 or 6 and translocates to the nucleus, where the tumor suppressor protein Rb is phosphorylated to activate transcription factor E2F. This cascade of events leads to the progression of the cell cycle from the G1- to S-phase (Sherr 1994; Johnson & Walker 1999).

In the developing central nervous system (CNS), the mRNA of Cyclin D2 has a unique localization at the surface of the neural tube that is not observed for the other cyclins (Ross & Risken 1994; Ross et al. 1996). Because of this unique localization pattern, Cyclin D2 was initially thought to be expressed in post-mitotic neurons (Ross & Risken 1994; Ross et al. 1996), but recent work, including ours, has identified that the Cyclin D2 mRNA and protein is actually localized at the tip of the AP (i.e., endfoot) (Glickstein et al. 2007; Tsunekawa et al. 2012). As with the other cyclins, Cyclin D2 is also localized in the nucleus of mitotic cells found in the VZ and SVZ and was assumed to have a function in cell cycle progression (Glickstein et al. 2007). In Cyclin D2 knockout mice, the brain size is smaller and adult neurogenesis is dramatically impaired, as expected (Kowalczyk et al. 2004; Glickstein et al. 2009; Matsumoto et al. 2011; Jedynak et al. 2012).

Cyclin D2 is essential for the expansion of NSPCs in both embryonic and adult brains, but the significance of the biased localization of Cyclin D2 in the basal endfoot of the APs is unclear. We have recently shown that overexpression of Cyclin D2 increases the population of Aps and that the loss of Cyclin D2 function increases the neuronal population (Tsunekawa et al. 2012). This finding indicates that Cyclin D2 localization at the endfoot of APs is an example of a “basal fate determinant”. This is unique in that the mechanism for fate determination of APs is at the subcellular level (Fig. 2). Cyclin D2 mRNA is continuously transferred toward the basal side up to the endfoot via its unique 50-bp cis-element (Step 1 in Fig. 3) and is locally translated into the protein (Step 2 in Fig. 3). During asymmetric cell division, one of the daughter cells inherits the basal process, which automatically leads to the asymmetrical inheritance of Cyclin D2 protein between the daughter cells (Step 3 in Fig. 3). The daughter cell with Cyclin D2 will become an AP, while the daughter without Cyclin D2 will become a neuronal cell or an IP (Step 4 in Fig. 3). As we described above, G1-phase of RGs elongate during the neurogenesis (Takahashi et al. 1994) Simultaneously, basal processes of RGs become longer due to accumulation of radially migrated post-mitotic neurons and IPs in the cortical plate. This co-relationship lead to an interesting idea that transportation distance of Cyclin D2 from the basal endfoot to the nucleus may control the length of G1-phase. More detailed studies are required to examine this hypothesis.

Figure 3.

Schematic depiction of Cyclin D2 mRNA and protein dynamics during the cell cycle and its putative role as a fate determinant. Pink in the nucleus indicates Cyclin D2 protein. In Step 1, Cyclin D2 mRNA is transported to the basal endfoot during G1 and the S- to G2-phase transition due to the cis-transport element that resides in the 3′UTR region of Cyclin D2 mRNA (blue box in mRNA) together with the transportation machinery that recognizes the cis-element (red circle). In Step 2, transported mRNA is locally translated into protein via ribosomes localized at the basal endfoot. In Step 3, during mitosis, Cyclin D2 protein is inherited by one of the daughter cells along with the basal process. In early G1-phase, the inheritance of Cyclin D2 creates a clear asymmetry between Cyclin D2 protein levels in the two daughter cells. In Step 4, the daughter cell that has inherited Cyclin D2 in the basal process remains a progenitor, whereas the other daughter without the basal process proceeds to differentiate.

We have shown that Cyclin D2 affects the fate of APs; however, the exact molecular mechanism is largely unknown. As mentioned above, the correlation between G1-phase lengthening and neurogenesis has been noted (Takahashi et al. 1994, 1995; Calegari & Huttner 2003; Dehay & Kennedy 2007; Lange & Calegari 2010; Salomoni & Calegari 2010). If the lengthening of G1-phase causes neuronal differentiation, the biased localization of Cyclin D2 will provide a shorter G1-phase to the daughter cell that inherits the basal process, which in turn biases the fate of that daughter cell to a progenitor. Although this is an intriguing possibility, time-lapse studies using slice culture suggest that inheritance of the basal process does not always lengthen the total cell cycle compared to the other daughter cell (Shitamukai et al. 2011; Wang et al. 2011). Another possibility is that Cyclin D2 controls cell fate in a manner other than by affecting the cell cycle itself. For example, Cyclin D2 is known to have a function in exporting the Cdk inhibitor p27(Kip1) out of the nucleus, thereby promoting its degradation (Susaki & Nakayama 2007; Susaki et al. 2007). Because p27(Kip1) promotes neurogenesis and the radial migration of postmitotic neurons (Nguyen et al. 2006a,b), inherited Cyclin D2 may inhibit neurogenesis and promote cell proliferation (Tsunekawa et al. 2012) via a p27(Kip1)-dependent mechanism. Many other reports have shown that cell cycle regulators may function as cell fate determinants by a role independent of cell cycle regulation (Ratineau et al. 2002; Nguyen et al. 2006a; Bienvenu et al. 2010; Ali et al. 2011). Furthermore, another detailed analysis suggested that both G1-phase and S-phase are correlated to the differentiation state of NSPCs (Arai et al. 2011). Thus, the exact physiological roles of Cyclin D2 in terms of fate determination during in vivo neurogenesis still remain to be elucidated.

Cyclin D2 and brain evolution

As described above, we have reported a new physiological function of Cyclin D2 in the neuronal development of the mouse. We next questioned whether this mechanism is conserved among mammalian species. In humans, we found an accumulation of Cyclin D2 protein at the basal side of the cortical primordium at gestation week 16 (Tsunekawa et al. 2012). We also noted that the cis-acting element identified in mice that promotes basal transportation is highly conserved in human (74% match in the National Center for Biotechnology Information [NCBI] database). Therefore, it is tempting to speculate that in the human cortical primordium, Cyclin D2 mRNA is similarly transported within the basal process toward the basal endfoot and locally translated into protein. Notably, the basal transport cis-element that we have identified appears to be unique to mammals, as similar sequences are not found in avians or amphibians (NCBI database). Similarly, no accumulation of Cyclin D2 mRNA in the basal side of the chick forebrain was observed (Tsunekawa and Osumi, unpubl. data). The acquisition of the genomic DNA sequences that correspond to the basal transportation regulatory element in the 3′UTR of Cyclin D2 mRNA might have been a critical diversification point in vertebrate brain evolution.

Recent progress in live imaging studies has revealed a new population of proliferative progenitors that have basal processes but no apical processes. These neural progenitor cells are located in the outer subventricular zone (OSVZ) of the fetal cortex of human and ferret and are thus called OSVZ radial glia-like cells (oRG or bRG stand for basal radial glia) (Fietz et al. 2010; Hansen et al. 2010; Lui et al. 2011). Although they were originally believed to exist only in primates or gyrencephalic mammals, several groups have recently reported that a population of oRGs can also be found in non-gyrencephalic mammals, including mice and marmosets (Garcia-Moreno et al. 2011; Shitamukai et al. 2011; Wang et al. 2011). According to observations made in the fetal human brain, the human CP expands via the proliferation of transit amplifying daughter progenitors (TAPs), seems to have no basal process progenitor cells derived from oRGs, and goes through multiple rounds of symmetric division to produce TAPs and finally produce multiple types of neurons (Fietz et al. 2010). Notably, there seems to be a correlation between basal process possession and asymmetric cell division capacity. This may be partially explained by the observation that oRGs show a clear correlation between Hes1 expression and basal process inheritance, indicating that the basal process may be required for receiving the Notch signal, a pivotal mechanism for maintaining the progenitor state (Hansen et al. 2010; Lui et al. 2011). From our observations, Cyclin D2-positive cells exist in the OSVZ, and punctate staining of Cyclin D2 is frequently observed in the basal side but not in the apical side (Tsunekawa et al. 2012) making it likely that these cells are oRGs. Interestingly, Betizeau et al. (2013) reported three additional new subtypes of proliferating neuronal progenitor cells in the OSVZ of the fetal monkey brain: apical process-bearing bRG (bRG-apical-P) cells, apical and basal process bearing bRG cells (bRG-both-P), and bRG cells that change shape with or without processes, named transient bRG (tbRG) cells. They also showed that the progression of oRG to TAP to neuron fate suggested in previous studies (Hansen et al. 2010) is not always accurate; instead, all five types of basal progenitors in the OSZV can produce neurons or other types of progenitors or undergo self-renewal, thus indicating a far more complex fate flow (Betizeau et al. 2013). It is noteworthy that G1-phase elongation of RGs during neurogenesis may not be conserved in primates where a shortening of G1-phase has been observed during the later stages of corticogenesis, at the time of generation of supragranular layers (Betizeau et al. 2013). It would be interesting to know the situation in corticogenesis in humans. Thus, it seems that the proliferation and differentiation of NPCs in the OSVZ is not simple. It is obvious that there is a clear relationship between brain size and the percentage of oRGs among all proliferating progenitors (Reillo et al. 2011), and more studies are required to understand the precise fate-mapping and function of oRGs in brain development. Moreover, it would be interesting to determine the functions of Cyclin D2 in this context.

In addition to polarized cells such as RGs and oRGs, non-polarized IPs are another type of NPC situated in the SVZ. During mouse development, Cyclin D2 has been shown to be also expressed in IPs and required for their proliferation (Glickstein et al. 2007). This has been further confirmed by our group; Cyclin D2 gain- or loss-of-function experiments drastically increased or decreased, respectively, the population of Tbr2-positive IPs in the SVZ (Tsunekawa et al. 2012). Therefore, Cyclin D2 is also essential for the proliferation not only of APs but also of IPs. This has prompted us to imagine the same might be true for oRGs. Data from recent papers indicate that the key component of gyrencephalic brain generation is the proliferation of IPs and oRG cells (Stahl et al. 2013). Thus, it would be interesting to gain a mechanistic understanding of the roles of Cyclin D2 (including post-transcriptional regulation) in controlling the proliferation and differentiation of all three types of main NPCs (APs, oRG and IPs) to address the question of how the brain became bigger in higher mammals such as humans.

Conclusion

Asymmetric inheritance of Cyclin D2 in dividing daughter cells of APs is the first description of a post-transcriptional regulatory mechanism in the developing vertebrate CNS. This unique mechanism is a result of the shape of the AP, which is highly polarized and has a long basal process. The basal process of APs was first identified as a guide cable for the radial migration of post mitotic neurons in the developing monkey brain (Rakic 1972). Forty years after the first report on basal process function, many reports have shed light on novel functions of the basal process. A recent report has revealed that the basal process of APs is necessary for the proper movement of the cell body in the basal direction during interkinetic nuclear migration (apico-basally directed, cell cycle-dependent movement of the cell body) to avoid over clustering of APs in the apical side of the cortex, which would potentially lead to the malformation of brain architecture (Okamoto et al. 2013). It is also believed that basal processes function to receive signals such as retinoic acid (Siegenthaler et al. 2009), Notch signals (Shitamukai & Matsuzaki 2012; Lee & Song 2013), Reelin (Lakoma et al. 2011), Sonic hedgehog (Shinohara et al. 2004; Komada et al. 2008) and FGFs (Hasegawa et al. 2004; Seuntjens et al. 2009) from cells located at the basal side of the cortex, while the cell body of APs is always situated at the apical side of the developing cortex (Shitamukai & Matsuzaki 2012). Additionally, as described by us and another group, the basal process can function as an asymmetrically inherited fate determinant (Fietz et al. 2010; Tsunekawa et al. 2012). Notably, we have identified mRNAs of several types of genes encoding transcription factors, motor proteins, mitochondrial associated factors and kinases that are localized within the basal endfoot of APs (our unpubl. data). These mRNAs and/or proteins could also be asymmetrically inherited by one of the daughter cells during cell division, giving different fates to the daughter cells. We believe that within the basal process of RGs there are further interesting post-transcriptional molecular mechanisms including transportation of mRNAs that ensure proper mammalian brain development.

Acknowledgment

We thank Dr Jun Wu for critical reading of the manuscript.

Ancillary