Although the vertebrate brain commonly stems from the neuroepithelial tube, the size and complexity of the pseudostratified organization of the brain have drastically expanded during mammalian evolution, resulting in the formation of a highly folded cortex. Developmental controls of neural progenitor divisions underlie these events. In this review, we introduce recent progress in understanding the control of proliferation and differentiation of neural progenitors from a structural point of view. We particularly shed light on the roles of epithelial structure and mitotic spindle orientation in the generation of various types of neural progenitors.
The vertebrate brain develops from a single epithelial layer, the neural tube, and the neuroepithelial (NE) cells that initially constitute the neural tube proliferate to expand their pool. Subsequently, these cells enter a neurogenic mode and begin to generate cell populations that are committed to differentiate into neurons via asymmetric cell division. Two different classes of neuronal progenitors in the developing mammalian neocortex exist in this neurogenic stage (Fig. 1). Self-renewing progenitor cells, called the radial glial (RG) cells, constitute one major class of progenitors. RG cells undergo asymmetric cell divisions to give rise to two daughter cells: one cell is equivalent to the mother cell, and the other is committed to differentiate. Thus, identifying the transition from the proliferative to the neurogenic mode of division is critical in determining the overall number of self-renewing progenitors and, thereby, in determining brain size.
Intermediate progenitors (IP) constitute another class of neural progenitors. These cells are transiently generated from self-renewing progenitor cells and are committed to differentiate. Thus, several progenitor subtypes have been observed in the developing neocortex, but how such different progenitors (and differentiated cells) result from the asymmetric cell division of RG cells remains to be elucidated. In this review, we introduce models that have been proposed to explain the asymmetric cell division of mammalian neural progenitors and discuss recent studies that have revised those models from a structural point of view, giving rise to novel information on the role of epithelial structure and spindle orientation in controlling the self-renewal of neural progenitors. We then discuss the remaining problems to be answered in the future.
Self-renewing progenitor cells
The RG (as well as NE) cells have highly polarized epithelial structures with the apico-basal polarity, and extend the process from the ventricular surface to the pial surface. The nuclei of the RG cells undergo dynamic interkinetic nuclear migration (INM) within the ventricular zone (VZ) and divide at the ventricular surface. In the neurogenic stage, the NE cells give rise to the so-called RG cells, which express characteristic markers such as the glutamate transporter GLAST and the radial cell antigen RC2 (Campbell & Gotz 2002). The RG cells divide asymmetrically not only to self-renew but also to produce more committed daughter cells, IP cells or post-mitotic neurons.
It has been known that primates develop an additional germinal zone (outer subventricular zone [SVZ]) outside of the SVZ during brain development. Recently, these progenitors were identified as a new subtype of self-renewing progenitors named outer RG (oRG) cells (Hansen et al. 2010). These cells maintain some RG cell properties, such as the ability to self-renew, RG marker expression and the basal process. However, they do not possess an apical process or apico-basal polarity, nor do they undergo INM (Fietz et al. 2010; Hansen et al. 2010). Similar to RG cells, oRG cells divide asymmetrically to self-renew and produce the more committed cells (Hansen et al. 2010). oRG-like cells also exist in other species, including ferret (intermediate radial glial cells, IRGC) and mouse (outer VZ progenitor cells, OVZ cells), during the middle to late neurogenic stage (Fietz et al. 2010; Reillo et al. 2011; Shitamukai et al. 2011; Wang et al. 2009). The number of oRG cells is correlated with brain size, which is larger in animals with bigger brains (e.g., humans, monkeys, and ferrets) and smaller in the animals with smaller brains (e.g., mice and rats). Therefore, oRG cells are thought to be the key to the evolution of brain size (Lui et al. 2011; Reillo et al. 2011). However, the mechanism of their production and maintenance is largely unknown.
Intermediate progenitor cells
Intermediate progenitor cells are transiently generated from self-renewing progenitor cells. Typical IPs in the developing neocortex (also known as basal progenitor [BP] cells) usually undergo one terminal symmetric division that produces two neurons, while some cells undergo two rounds of symmetric divisions at low frequency (Haubensak et al. 2004; Miyata et al. 2004; Noctor et al. 2004). BP cells divide at the SVZ, which is the second major germinal zone site and are multipolar without the same epithelial structure as daughter cells of RG cells (Haubensak et al. 2004; Miyata et al. 2004; Noctor et al. 2004, 2008). In addition, other groups have reported that morphologically different IP-like cells, short neural precursors (SNP), maintain apical connection and undergo terminal symmetric division at the ventricular surface (Gal et al. 2006; Stancik et al. 2010). Whether the SNPs are a distinct progenitor cell subtype or a transition step for IP cells that have not yet been lost from the VZ remains to be determined. In addition, a subpopulation of IP cells inherit the basal process from RG cells, which is transiently retained for several hours after RG division; however, these cells lose the basal process before the next cell division (Tabata et al. 2009; Shitamukai et al. 2011).
Models for asymmetric cell division of mammalian neural progenitors
Typically, asymmetric cell division is achieved via two distinct mechanisms that are best understood by invertebrate model systems, such as Drosophila and Caenorhabditis elegans (Knoblich 2008). Determinants of cell fate are asymmetrically distributed at mitosis in the population of progenitor cells. This intracellular polarized distribution of determinants is often achieved cell intrinsically, as observed in Drosophila neuroblast and C. elegans fertilized eggs (Knoblich 2008). Alternatively, cell extrinsic factors may act on a progenitor cell asymmetrically, generating a local difference in intracellular responses that results in cell polarization, as observed in Drosophila germline stem cells and C. elegans blastomere and T hypodermal cells (Yamashita et al. 2011; Sawa 2010). In both cases, the coupling of a division axis and cell polarity is critical to maintain the proper determination of cell fate. When the cell polarity axis is orthogonal to the cell division axis, two daughter cells receive equivalent determinants or extracellular signals (Fig. 2A,B). Therefore, the relationship between the cell polarity axis and the cell division axis is important in determining whether a polarized cell divides symmetrically or asymmetrically. This general scheme is thought to be common for RG cells in the developing neocortex.
Based on this principle, two plausible models for the asymmetric cell division of mammalian neural progenitors have been proposed. Chenn & McConnell (1995) identified a change in the division axis during the neocortical development of ferret brain using time-lapse imaging of fluorescent dye-labeled chromosomes (Chenn & McConnell 1995). In this model, planer division is symmetric and proliferative, whereas vertical division is asymmetric and neurogenic (Fig. 2C). However, the majority of RG cell divisions are not vertical but more planar to the ventricular surface (Kosodo et al. 2004; Konno et al. 2008; Noctor et al. 2008). Therefore, an obvious question arises with respect to their model: how are differential fates conferred to the daughters from planer divisions? Huttner and colleagues have proposed a possible mechanism to answer this question (Huttner & Brand 1997; Kosodo et al. 2004). The apical domain facing the ventricle represents a small area in the dividing RG cells, and a slight tilt of the cell division axis is sufficient to ensure asymmetric partition of this domain (Fig. 2D). This model suggests that the asymmetric inheritance of the apical membrane, in addition to spindle orientation, is a key feature in the different daughter cells' fate. Basically, both models share the idea that the differential inheritance of the apical or basal epithelial structure is important for a daughter cell to self-renew or differentiate. Thus, precise observations of the inheritance of epithelial structures and daughter cell fate during RG cell divisions in vivo are crucial to evaluate these models.
Time-lapse imaging analysis has greatly contributed to our understanding of the sequential changes in cell morphology, which is one good marker for cell polarization and neural differentiation status. Recently, several groups have reported direct observations of the inheritance of epithelial structure and daughter cell fate during RG cell divisions (Konno et al. 2008; Alexandre et al. 2010; Shitamukai et al. 2011). In the following sections, recent progress in in vivo observations regarding the inheritance of epithelial structures during RG cell division is discussed. In addition, unequally segregated molecules have been identified by live imaging in vivo (Reugels et al. 2006; Wakamatsu et al. 2007). Therefore, determining the precise correlation between the inheritance of apical and basal epithelial structures and cell fate in vivo would aid our understanding of the mechanism of asymmetric cell division of RG. Moreover, genetic manipulation is useful to understand what the perturbation of spindle orientations causes. Finally, several reports have suggested that prior models are unlikely to explain the asymmetry of daughter cell fate but give rise to a new role for spindle orientation in the production of various types of neural progenitors (Fig. 3).
RG cell divisions in the neurogenic stage
Radial glial cells form adherens junctions at the ventricular surface and at a very small area of the apical membrane domain facing the ventricle. It has been technically difficult to accurately detect the segregation of the apical domain during RG cell divisions in the coronal view of brain slices. However, a new technique has been developed to overcome this difficulty: the expression of the ZO1–enhanced green fluorescent protein (EGFP) fusion protein enables us to directly monitor the partition of the apical junction from the apical side of the slice (Konno et al. 2008). This technique demonstrated that segregation of the apical membrane exclusively to one daughter cell normally occurs in no more than 10–20% of apical divisions during the neurogenic stage (Konno et al. 2008; Asami et al. 2011; Shitamukai et al. 2011), which is less frequent than previously estimated (Kosodo et al. 2004). Instead, the apical junction is partitioned between both daughter cells in the majority of RG cell divisions, and the fate of the daughter cell is mostly asymmetric (Fig. 3). Moreover, no correlation was detected between the inherited size of the apical domain and RG cell fate (Shitamukai et al. 2011). In contrast to the inheritance of the apical junction, the basal process partition is highly asymmetric during the neurogenic stage (Fig. 3), with exceptions in the early neurogenic stages (Kosodo et al. 2004; see below). Because both daughters inherit the apical membrane during most RG cell divisions, structural asymmetry is instead generated by the asymmetric inheritance of the basal process in the majority of RG cell divisions (Miyata et al. 2004; Shitamukai et al. 2011). RG cell fate has also been shown to be tightly correlated with inheritance of the basal process, whereby the daughter cell that does not inherit the basal process mostly differentiates into neurons or IP cells (Shitamukai et al. 2011). In the majority of RG cell divisions, one daughter cell inherits the entire epithelial structure and maintains the capability to acquire RG cell identity, whereas the other cell inherits only the apical epithelial domain and is thus committed to differentiate into a neuron or IP cell (Fig. 3).
NE cell divisions at proliferative stages
The pattern of inheritance of the epithelial structure and plasticity greatly changes in the developing stage. Whereas in the neurogenic stage, most RG divisions are asymmetric in terms of the cell fate and cell morphology of the daughters, in the proliferative stage, daughter cell fate is symmetric. This difference between the stages gives rise to a question: how is epithelial structure segregated during the proliferative stage? At present, there are no reports of live imaging of long-term slice cultures from early-stage mouse brain due to technical difficulties. The live imaging of zebrafish provides some clues to understanding NE cell divisions during the proliferative stages (Alexandre et al. 2010). In zebrafish development, the hindbrain is more proliferative than is the developing mouse neocortex, and both apical and basal processes are able to re-establish themselves, even after the epithelial structure is asymmetrically segregated. Re-extension of the basal process has also been observed in the early neurogenic stage in mice, whereas it is rarely observed in the middle-to-late stage (Miyata et al. 2004; Shitamukai et al. 2011). The regeneration of a missing process during RG cell divisions explains why no obvious phenotype is observed under the spindle-misorientation in the early proliferative stages (Konno et al. 2008; Postiglione et al. 2011). Thus, the integrity of the epithelial structure seems to be more robust in the early stage than it is in the neurogenic stage, and is associated with undifferentiated status, especially with the activation of Notch signaling (Fig. 3). The extension of the basal process is promoted by Notch signal activation (Gaiano et al. 2000; Shitamukai et al. 2011).
The basal process becomes very thin during RG division in the neurogenic stage, and cleavage furrow avoids the basal process (Miyata et al. 2004; Shitamukai et al. 2011). Recently, Kosodo & Huttner (2009) reported that the basal process of NE cells splits before cytokinesis begins. These split processes are occasionally inherited by both daughter cells (−25%). The split basal process has been difficult to detect in RG cell divisions in the mid neurogenic stage due to technical difficulties.
Spindle orientation controls the inheritance of the epithelial structure
The previously discussed models predict that the induction of vertical or oblique divisions causes disruption of the epithelial structure and promotes asymmetric cell divisions, thereby leading to premature neural differentiation when it occurs during the proliferative stages. Indeed, several studies have indicated a correlation between spindle orientation defects and premature neuronal differentiation (Feng & Walsh 2004; Fish et al. 2006; Gauthier-Fisher et al. 2009; Godin et al. 2010). However, these observations need to be carefully interpreted to judge the role of tilted spindle orientation in RG cells' asymmetric divisions because the genes that affect spindle orientation may be directly involved in the regulation of microtubules dynamics. For example, components of the centrosome and the dynein motor complex affect many other processes that can influence cell fates, such as cell cycle progression and vesicle transport. In contrast, in the mouse cortex and in the chick spinal cord, the effects of the LGN and NuMA genes have been studied (Morin et al. 2007; Konno et al. 2008; Peyre et al. 2011). The loss of function of these genes caused spindle randomization with little influence on the net neurogenesis and unexpectedly resulted in the formation of ectopic progenitors outside of the VZ. These ectopic progenitors are generated by the loss of the apical junction during RG divisions and sequential translocation of the cell body to the basal side. Time-lapse imaging and in vivo clonal fate analysis showed that these ectopic progenitors repeat neurogenic asymmetric divisions, thus sharing cell properties with human and ferret oRG cells (Fig. 3B; Shitamukai et al. 2011). Interestingly, a low frequency of oblique RG cell divisions was observed in the normal developing mouse neocortex, which produced a small number of oRG-like cells (Shitamukai et al. 2011; Wang et al. 2009), strongly suggesting that the spindle misorientation induced the oRG-like cells in the mouse neocortex. Recently, analysis of the loss and gain of function analysis of the mouse Inscuteable (mInsc) gene indicated that mInsc functions by perturbing the horizontal spindle orientation of RG cell divisions and by increasing the number of neurons (especially in the upper layer of the cortex), although asymmetric neurogenic divisions occur in its absence. This study suggests that the oblique spindle orientation plays a role in the production of basal progenitors and thereby in the increase in final brain size (Postiglione et al. 2011). This situation differs from the case of the loss of LGN function, which causes both spindle randomization and the production of ectopic progenitors outside of the VZ without significantly affecting the final brain size (Konno et al. 2008; Shioi et al. 2009). Further studies are needed to clarify how spindle orientation controls the brain size as well as the production of basal and outer subventricular zone (OSVZ) progenitors. In the proliferative stages, the perturbation of spindle orientation has been studied, but its effects on progenitors have depended on which genes are compromised (Konno et al. 2008; Yingling et al. 2003; Postiglione et al. 2011). The effect of spindle orientations on RG cell divisions might be dependent of on the developmental stage (Fig. 3B).
Taken together, recent imaging and spindle orientation studies in neural progenitors give rise to a possible understanding of the asymmetric cell division of RG cells in mammalian cortical development. The vertical (apical-basal) or oblique RG divisions create a cleavage furrow that bypasses the apical junction, thereby splitting the apical and basal epithelial structures into two daughters. As such, the daughter cells inheriting only the basal process maintain the ability to self-renew. In contrast, the daughter cells inheriting only the apical part are less proliferative and mostly differentiate into neurons or IP cells. This process is consistent with observations in zebrafish indicating that the more apical daughter adopts the differentiation fate, whereas the more basal daughter self-renews (Alexandre et al. 2010). Thus, the different neurogenic and proliferative activities are clearly associated with the epithelial structure. A recent study using zebrafish indicated that the apical part of the neuroepithelial cells is essential for efficient Notch activation (Ohata et al. 2011). In the majority of mammalian RG cell divisions, both daughters inherit apical junctions. Thus, the differential inheritance of the basal process also seems to play an important role. In the next section, we review the intrinsic and extrinsic potential factors of cell fate determination that may be associated with the asymmetric inheritance of apical and basal processes in RG cell divisions.
The role of epithelial structure in the fate of daughter cells
Intrinsic factors that asymmetrically determine cell fate during RG cell division have frequently been reviewed (Fietz & Huttner 2011; Lui et al. 2011). Here, we briefly discuss the potential cell fate determinants that are associated with the apical part of RG cells and those that are associated with the segregation of the basal process. Numb is a conserved negative regulator of Notch signaling that has been thought to be critical for the determination of daughter cell fate at RG cell divisions. Although Numb is a cell fate determinant that asymmetrically segregates in Drosophila sensory precursor cells, clear evidence has been absent in the case of RG cell divisions. Par3 has been demonstrated to directly bind Numb, resulting in direct phosphorylation of Numb by aPKC, which inhibits Numb function (Nishimura & Kaibuchi 2007). Furthermore, a recent study discovered that Par3 spreads into both the cell cortex and cytoplasm during mitosis and segregates asymmetrically, whereas Par3 is distributed to the apical junction in interphase RG cells (Bultje et al. 2009). These observations led to the creation of a model of Par3 that is asymmetrically segregated and leads to differing fates for the two RG daughters via asymmetric suppression of Numb (Bultje et al. 2009). The mechanism underlying the asymmetric segregation of cytoplasmic Par3 protein remains to be studied. Recently, the correlation between daughter cell fate and asymmetric segregation of centrioles with different age was observed during RG cell division in the developing mouse neocortex (Wang et al. 2007). During cell divisions, two daughter cells unequally inherit centrioles that duplicate into new and old centrioles in a semi-conservative manner (Nigg & Raff 2009). This process also occurs for the pericentriolar materials that constitute the centrosomes. The older centrosome (i.e., the one that inherits the older centriole) is usually more mature and recruits more pericentriolar materials. An attractive hypothesis is that this centrosome asymmetry leads to the asymmetric inheritance of cytoplasmic materials, including cell fate determinants, thereby causing differential cell fates during RG cell divisions. Centrosome dysfunction results in various phenotypes in the progenitor cell, including cell death, changes in spindle orientation, cell cycle defects and the promotion of differentiation (Feng & Walsh 2004; Buchman et al. 2010; Pulvers et al. 2010; Gruber et al. 2011). Thus, careful investigations are necessary to test this hypothesis.
In contrast to apically localized molecules, there are only a few reports of molecules that were specifically localized in the basal process and that affect cell fate at RG cell divisions. Beta 3-integrin is located in the basal process. Inhibition of integrin function (either chemically or with antibodies) in ferret slice culture causes a loss of oRG cell marker expression (Fietz et al. 2010), although controversial results were also reported based on the analysis of integrin knockout mice (Haubst et al. 2006). The cyclin D2 mRNA and protein, a regulator of cell cycle initiation, is also localized in the basal end feet (Glickstein et al. 2007; Tsunakawa et al. 2012). In fact, the cyclin D2 knockout mouse shows a reduction in brain size (Glickstein et al. 2009), whereas the cyclin D2/Cdk4 complex acts in the nucleus. It was recently shown that Cyclin D2 played an important role in promoting the self-renewal of the RG daughter (Tsunekawa et al., 2012). Further characterization of the precise requirements for intrinsic factors in the basal process is necessary to reveal the role of those basally localized factors in differential cell fate during RG cell divisions.
Extrinsic factors such as Notch–Delta pathway are crucial for the maintenance of RG cell identity (Yoon & Gaiano 2008). Notch signaling is activated by the cell–cell interaction between Notch-expressing cells and Delta-expressing cells. The Notch receptor is mainly expressed in the RG cells and is distributed along the entire cell membrane (Gaiano et al. 2000). A Notch ligand, Delta-like 1, is specifically and transiently expressed in the differentiating cells, particularly in the VZ (some cells located in the SVZ; Kawaguchi et al. 2008). As shown in the zebrafish hindbrain (Ohata et al. 2011), the Crumbs complex in the subapical region directly interacts with Notch and plays a key role in its activation and in the maintenance of the epithelial structure of RG cells. In addition, apical adhesion facilitates the interaction and activation of Notch–Delta signaling (Mizuhara et al. 2005). Therefore, it is most likely that Notch–Delta interaction efficiently occurs at the apical side of the VZ. Migrating neurons along the basal process of RG cells are also thought to contribute Notch–Delta interactions (Lui et al. 2011; Fig. 4). However, oRG cells that are located outside of the VZ require Delta from their sibling cell, which appears to interact with the sister oRG cell body (Shitamukai et al. 2011).
Several secreted molecules from the basal and apical sides play important roles in neural differentiation and progenitor proliferation (Fig. 4). Retinoic acid secreted from the meninges, an epithelial sheet covering the brain, is required for the transition from NE to RG cells (Siegenthaler et al. 2009). Sonic hedgehog from the tangentially migrating interneurons and endothelin B from blood vessels affects the proliferation of neural progenitors (Shinohara et al. 2004; Komada et al. 2008). The Reelin protein from Cajal-Retzius cells in the outermost layer of the brain functions in RG cell maintenance by promoting the Notch–Delta pathway (Lakoma et al. 2011). FGF18, FGF9, and neurotrophin-3 are expressed in neural layers and regulate the temporal change of RG cells during neocortical development (Hasegawa et al. 2004; Seuntjens et al. 2009). A possible role of the basal process may be to transmit signals from the basal side. On the apical side, RG cells face the cerebrospinal fluid in the ventricle, which contains several growth factors from the choroid plexus and other brain regions (Fig. 4). In particular, insulin-like growth factor 2 greatly contributes to progenitor proliferation in the middle-to-late stages, and its receptor is specifically localized to the apical surface (Lehtinen et al. 2011). FGF15, a hormone-like FGF secreted from the ventral brain, also affects neocortical neurogenesis (Borello et al. 2008; Itoh 2010). Thus, various extrinsic factors are provided to RG cells in the developing brain. The elongated epithelial structure might contribute to the net receipt of theses extrinsic signals, and/or its asymmetric inheritance might affect the differential cell fate, depending on the distribution of these extrinsic factors in the neocortex and the localization of their receptors in RG cells.
Conclusions and open questions
In conclusion, recent studies demonstrate that proliferation and neuronal differentiation activities differ along the basal to apical axis of RG cells. It is likely that the asymmetric inheritance of epithelial structures during RG cell divisions gives rise to a bias with regard to the reception of intrinsic and extrinsic signals that affect the self-renewal and differentiation of RG cells. The inheritance of the epithelial structure has also been shown to have a direct instructive role on cell fate specification. Furthermore, novel views on the roles of spindle orientation have evolved in generating basal progenitors as well as ectopic neural progenitors outside of the VZ. In the near future, the integration of our understanding of how intrinsic fate determinants and extrinsic signals are associated with epithelial structure will resolve a longstanding question in cortical development: how do RG cells generate neurons via asymmetric cell divisions?