Neurogenesis is a dynamic process that produces a diverse number of glial and neural cell types from a limited number of neural stem cells throughout development and into adulthood. After an initial period of amplification through symmetric division, neural stem cells rely on asymmetric modes of division to self-renew while producing more committed progeny. Understanding the molecular mechanisms regulating the choice between symmetric and asymmetric modes of division is essential to understand human brain development and pathologies, and to explain the increasing cortical complexity observed in evolution. A popular model states the existence of a causal relationship between the orientation of the axis of division of stem cells and the fate of their progeny in many different tissues, but the validity of the model in neural stem cells is not clear. In this review, we briefly present the diversity of neural stem cells and intermediate progenitors in the developing central nervous system. We then draw a historic overview of the assumed causal relationship between spindle orientation and fate determination. We show how this prompted a search for regulators of spindle orientation, and present the current state of knowledge on the mechanism. Finally, we review data on the effect of defective spindle orientation and try to integrate conflicting observations by presenting alternative mechanisms that may regulate the choice between symmetric and asymmetric outcomes.
Vertebrate neurogenesis is a highly dynamic process during which a limited number of neural stem cells produce a highly diverse number of glial and neural cell types. Neural stem cells are initially found as a single layer of neuroepithelial cells forming the neural tube. These cells first undergo a phase of massive proliferation, during which their number increases exponentially through symmetric divisions. Later on, the system switches to a neurogenic mode during which neuroepithelial progenitors start to divide in an asymmetric manner to self-renew and produce a more committed progeny. Depending on the organisms and on the region of the CNS, the timing of the switch as well as the outcome of asymmetric divisions can differ. The question of the regulation of the choice between symmetric and asymmetric divisions has been the focus of intense efforts in the last two decades, not the least because defects in this control may be a causing factor for a number of human developmental diseases. In particular, it is thought that some types of microcephaly are the result of an early exhaustion of the pool of neural stem cells through a premature switch to neurogenic divisions at the expense of stem cell proliferation (Fish et al. 2006; Thornton & Woods 2009). The question has also been addressed from the perspective of vertebrate cortex evolution, as regulating the mode of division of neural progenitors appears as a key process in regulating the size and surface area of the vertebrate brain (Fish et al. 2008; Lui et al. 2011).
Despite this strong interest, how this balance is regulated is still poorly understood at the molecular level, and as we will see in this review, it is likely a complex, multifactorial process that integrates a number of cell autonomous and cell extrinsic information. Early on, the axis of division of neural progenitors in the germinal zone has been proposed to be a key regulator of the outcome of the division. This hypothesis remains very popular today as a number of studies have lent support to the model. However, it has been regularly debated and reformulated; over the last decade, the discovery of new types of intermediate progenitors, the identification of factors regulating spindle orientation and the use of clonal and live analysis of cell divisions have shed new light on the question, and suggest that it is time to re-examine the hypothesis. In this review, we briefly present the diversity of neural stem cells and intermediate progenitors in the developing CNS. We then draw a historic overview of the assumed causal relationship between spindle orientation and fate determination. We show how this prompted a search for regulators of spindle orientation, and present the current state of knowledge on the mechanism. Finally, we review data on the effect of defective spindle orientation and try to integrate conflicting observations by presenting alternative mechanisms that may regulate the choice between symmetric and asymmetric outcomes.
The developing vertebrate CNS contains different types of progenitor cells
The neural plate is initially formed by a single layer of progenitor cells called neuroepithelial cells (NE), which form the neuroepithelium. NE cells have epithelial characteristics and are highly polarized along their apico-basal axis with apical attachments to the ventricular surface and a basal fiber connecting the pial (basal) surface. The neuroepithelium is pseudostratified, with NE cell nuclei found all along the apico-basal axis. These nuclei undergo interkinetic nuclear movement, moving back and forth between the apical and basal sides of the tissue during their cell cycle, and undergoing mitosis at the ventricular surface. NE cells have a small apical domain, forming the ventricular surface and a large baso-lateral domain. These two membrane domains are separated by tight junctions preventing lateral diffusion of apical proteins. Before the onset of neurogenesis, NE cells undergo a phase of massive proliferation by symmetric division allowing the expansion of the neural progenitor cell population (McConnell 1995; Rakic 1995). At the onset of neurogenesis, they switch to an asymmetric mode of divisions allowing self-renewal and neuron generation (Gotz & Huttner 2005).
Concomitant with the beginning of neurogenesis, NE cells in rodent and higher vertebrates start to acquire characteristics associated with glial cells and are then called radial glial cells (RG). They express astroglial markers such as GLAST (Glutamatergic Astrocyte Specific Transporter), BLBP (Brain Lipid Binding Protein), Nestin (Hockfield & Mckay 1985) and Vimentin (Pixley & De Vellis 1984). Newly-born neurons use RG basal fibers as a migration scaffold toward cortical layers (Rakic 1971; Misson et al. 1991). This results in progressive cortical thickening and basal fiber elongation. Tight junctions lose the expression of Occludin and evolve into adherens junctions delimited by a ZO-1 (Zona Occludens 1) domain and higher amount of N-cadherin (Aaku-Saraste et al. 1996). These junctional complexes anchor the RG cells to each other and to the ventricular surface, they also allow the recruitment of cytoplasmic proteins such as Par3, aPKC (Atypical Protein Kinase C), Par6 (Joberty et al. 2000; Lin et al. 2000), as well as, β-catenin and δ-catenin (Zhadanov et al. 1999; Ho et al. 2000). It is important to note that neural progenitor cells form a continuum from early development (NE) through neurogenesis (RG) and into adulthood (adult stem cells) (Merkle et al. 2004). As they maintain an apical attachment and divide apically, NE and RG cells are more generally called apical progenitor cells (AP).
Neurons derive from asymmetric division of RG cells either directly in a process called direct neurogenesis, or indirectly, through the production of intermediate progenitors. Indirect neurogenesis was first revealed in the neocortex by the identification of monopolar neural progenitor cells migrating basally in the subventricular zone (SVZ) to divide (Haubensak et al. 2004; Miyata et al. 2004; Noctor et al. 2004). These cells are called basal progenitors (BP) or intermediate progenitors (IP) and have been observed in the neocortex of all mammals studied so far (Fietz et al. 2010; Hansen et al. 2010). IP differ from RG at the morphological and molecular levels. When produced by asymmetric division of a RG, IP retract their apical attachment and basal extension, do not exhibit hallmarks of apico-basal polarity and migrate basally before they undergo mitosis (Fig. 1b) (Attardo et al. 2008; Noctor et al. 2008; Pontious et al. 2008). IP cells express characteristic molecular markers such as Tbr2 (Englund et al. 2005) or Insm1 (Insulinoma associated 1) (Farkas et al. 2008). In their majority, RG-derived IP cells divide once symmetrically to give rise to two neurons (Noctor et al. 2004). This population of IP cells is a transit amplification compartment that allows indirect neurogenesis to increase the number of neurons produced by surface/time unit (Lui et al. 2011).
Recently a new type of progenitor, called outer Radial Glia (oRG) or basal Radial Glia (bRG), has been described in the cortex of gyrencephalic mammals (Fietz et al. 2010; Hansen et al. 2010). oRG are thought to arise from the division of RG cells; they delaminate from the apical surface and translocate their nuclei in the outer portion of the SVZ where they start dividing (Fig. 1b). While these cells are monopolar and do not express the apical markers found in RG cells, they maintain a long basal fiber connecting the basal lamina and molecular characteristics of RG cells such as expression of Pax6, nestin, GLAST, GFAP (Fietz et al. 2010; Hansen et al. 2010). Moreover, in contrast to IP, oRG can either divide symmetrically to expand their number (Hansen et al. 2010; Reillo et al. 2011), or divide asymmetrically to self-renew and produce intermediate progenitors which then contribute to the production of neurons (Hansen et al. 2010). In this respect, they are very similar to apical RG cells. However they appear to have a distinct behavior and different proliferative potentialities. Just prior to division, oRG nuclei undergo a basal translocation movement along their basal fiber. oRG usually divide in an apico-basal fashion, and the daughter cell that inherits the basal fiber remains an oRG while the other daughter cell generally goes through a transit amplification phase. Contrary to RG-derived IP cells, which express Tbr2 and typically divide only once to produce two neurons, oRG-derived IP cells do not start expressing Tbr2 immediately, and live imaging shows that they go through several rounds of division before their progeny differentiates as neurons (Hansen et al. 2010). It was initially proposed that oRG are specific to gyrencephalic brains and that they could be an essential step in the evolution from lissencephalic to gyrencephalic brains. However, they have also recently been identified in the rodent brain (Shitamukai et al. 2011; Wang et al. 2011b) and in lissencephalic primate brains (Garcia-Moreno et al. 2011; Kelava et al. 2011).
Other types of intermediate cells have been described, such as short neural progenitors (Gal et al. 2006; Stancik et al. 2010), but their position in the lineage between RG and neurons is not well understood (Lui et al. 2011).
A historical overview of the relationship between spindle orientation and progenitor modes of division
At the onset of neurogenesis, progenitor cells switch from symmetric to asymmetric divisions in order to produce committed neural cells. A central question in the field is what regulates the mode of division of these progenitor cells? In the last two decades, the orientation of the axis of division has been shown to be correlated with the choice between symmetric and asymmetric modes of cell division in a number of systems (see Box 1 for two representative examples, and Morin & Bellaïche 2011 for a review).
Cleavage orientation and intrinsic versus extrinsic regulation of binary fate choices
The orientation of the mitotic spindle can be linked to asymmetric modes of division in different ways. Spindle orientation can be correlated with the polarized distribution of intracellular cell fate determinants in the mother cell before division, so that daughter cells will inherit unequal amounts of fate determinants and acquire a different identity, through a process of “intrinsic” asymmetric cell division. This mechanism is frequently used in systems where lineage choices are invariant. The division of drosophila neuroblasts is a canonical example of intrinsic asymmetry (reviewed in Morin & Bellaïche 2011).
Alternatively, spindle orientation can be influential on cell fate choices in an “extrinsic” manner, as it determines the cleavage plane of the dividing cell, and the position of the two daughter cells relative to each other and to their environment. Hence, sister cells may acquire a different identity through differential exposure to signaling molecules.
During development of the mammalian skin, the orientation of the cleavage plane of basal progenitors is correlated with their mode of division: early in development, the majority of divisions is planar and symmetric proliferative. They then switch to asymmetric divisions with the spindle oriented perpendicular to the basal lamina to generate a basal progenitor cell and a supra-basal differentiated cell, leading to the stratification of the epithelium. The molecular mechanisms regulating apico-basal orientation are remarkably similar between mammalian skin progenitors and fly neuroblasts (Lechler & Fuchs 2005; Poulson & Lechler 2010; Williams et al. 2011). However, whereas fly NBs clearly rely on the intrinsic distribution of fate determinants for their asymmetric division, there is currently no evidence that it is the case for skin progenitors. Instead, it appears that apico-basal divisions result in differential Notch activity in the progeny through differential exposure to environmental Notch ligands between the sister cells residing in the basal and supra-basal cell layers (Lechler & Fuchs 2005; Poulson & Lechler 2010; Morin & Bellaïche 2011; Williams et al. 2011).
As early as the mid-60s, it was proposed that the mode of division of vertebrate neural progenitors could be predicted by the orientation of their cleavage plane. Planar divisions (see Box 2 for the definition of planar, horizontal, apico-basal, vertical, and oblique divisions) would produce two progenitor cells remaining near the ventricular surface thanks to equal inheritance of attachment sites, whereas apico-basal divisions would produce a basal daughter that would migrate and differentiate (Langman et al. 1966; Martin 1967). However, analysis of the occurrence of planar, oblique and apico-basal divisions in vertebrate neural progenitors in the mouse neocortex (Smart 1973) indicated a major discrepancy between the low frequency of apico-basal divisions and the high number of differentiating neurons produced during the peak of neurogenesis. The initial hypothesis was reinforced with a landmark paper based on live observation of dividing progenitors in ferret neocortex slices (Chenn & Mcconnell 1995). The authors observed planar divisions, which split equally the apical membrane between the two daughter cells and resulted in symmetric outcomes, while apico-basal divisions produced two behaviorally distinct daughter cells by asymmetric division. They also observed that Notch receptor immunoreactivity accumulated basally and was inherited unequally between daughter cells upon apico-basal divisions and proposed its involvement in resolving binary fate choice between sister cells. These data found a powerful echo in the description of the asymmetric inheritance of the cell fate determinants Numb and Prospero by the committed daughter cell of dividing drosophila embryonic neuroblasts (NB) (Box 1 and Rhyu et al. 1994; Hirata et al. 1995; Knoblich et al. 1995; Spana & Doe 1995; Spana et al. 1995). Chenn and McConnell proposed that the asymmetric division of vertebrate neural progenitors could be regulated by the unequal inheritance of cell fate determinants, such as Notch, in addition to apical attachment sites. This hypothesis was subsequently reinforced by the observation that Numb antagonizes Notch signaling in drosophila (Spana & Doe 1996).
The use of horizontal and vertical terms in the literature on spindle orientation can be confusing. The ventricular surface is usually represented as a horizontal line, so that planar divisions (divisions that occur in the plane of the epithelial surface), which harbor a vertical cleavage plane, are called horizontal divisions. Conversely, divisions along the apico-basal axis are vertical divisions, with a horizontal cleavage plane. For simplicity, we will use the terms planar for horizontal divisions, and apico-basal for vertical divisions. Divisions with an intermediate orientation will be called oblique.
However, apico-basal divisions are too unfrequent to account for all the asymmetric divisions taking place during neurogenesis (Huttner & Brand 1997). Kosodo et al. suggested that asymmetric segregation of apically localized cell fate determinants could still occur without apico-basal division, thanks to the small size of the apical domain. Unlike drosophila neuroblast where cell fate determinants are distributed as large basal crescents, vertebrate neural progenitor cells are very elongated and their apical domain represents only 1–2% of the total membrane surface. Minor changes in spindle orientation may therefore decide whether the cleavage plane would bisect or bypass the small apical domain and result in equal or unequal repartition of the apical domain and its putative cell fate determinants between the daughter cells (Kosodo et al. 2004; Marthiens & Ffrench-Constant 2009). In this model, near vertical and oblique cleavage planes that occur frequently are sufficient to generate asymmetric divisions and to account for the large number of cells engaged in neural differentiation. The model was supported by the observation that cells expressing the Tis21 transcription factor, associated with neurogenic divisions, or a transgene carrying green fluorescent protein (GFP) under the Tis21 promoter, show more frequent “bypassing” cleavage planes than Tis21 negative cells (Kosodo et al. 2004). It also predicts that perturbations of spindle orientation that result in more oblique and apico-basal divisions should initially favor neurogenesis at the expense of stem cell pool expansion and may ultimately lead to a reduced brain size.
Mechanisms of mitotic spindle orientation
One way to test the model is to induce defects in spindle orientation and evaluate their effect on fate determination and neurogenesis. This prompted a search for regulators of spindle orientation. Mechanisms of mitotic spindle orientation have been studied extensively in models of asymmetric division such as the Caenorhabditis elegans zygote and drosophila embryonic and larval neuroblasts as well as larval sensory organ precursors. These cells share common principles and molecular mechanisms to orient their axis of division: a polarity cue is used to position cortical force generators that pull on astral microtubules to orient the mitotic spindle along a specific axis (for a detailed review of these mechanisms see Morin & Bellaïche 2011). In drosophila NB, an apically located complex composed of Par3, Par6 and aPKC, recruits the Inscuteable (Insc) protein to the cell cortex during division (Schober et al. 1999; Wodarz et al. 1999) (Box 1). Insc in turn recruits the Pins (Partner of Inscuteable) molecule, which is itself anchored to the cell cortex through interaction with the cortically localized Gαi subunit of heterotrimeric G proteins (Schaefer et al. 2000; Yu et al. 2000). Pins can simultaneously bind to Mud, which directly interacts with members of the dynein/dynactin motor complex. Hence, the minus end directed motor activity of cortically anchored dynein is thought to produce pulling forces on astral microtubules, leading to the positioning of one spindle pole underneath the apically concentrated complex and result in the alignment of the spindle with the apico-basal axis of polarity of the cell. Absence of any member of this complex results in defects in spindle orientation in these cells (for review see Gonczy 2008).
Homologues of the molecules involved in spindle orientation in invertebrate models have been identified in vertebrates and they are all expressed in neuroepithelial and radial glial cells. One major discrepancy between drosophila NB and vertebrate neural progenitors is that, while NBs divide along the apico-basal axis of the cell, the main axis of division of neural progenitors is planar, that is, perpendicular to the apico-basal axis of polarity. This raises the question of the extent of the conservation of the mechanisms of spindle orientation between models.
The LGN complex: Kingpin of force transmission through the dynein motor complex?
Live imaging in the chick neuroepithelium has shown that the mitotic spindle initially forms with a random orientation and that planar orientation is achieved through directed spindle rotation during early metaphase (Roszko et al. 2006; Peyre et al. 2011) (Fig. 2a). The Pins vertebrate homologue LGN is enriched in the lateral membrane of both mouse and chick dividing neural progenitors (Konno et al. 2008; Peyre et al. 2011) and this lateral distribution is necessary for the rotation and for planar orientation of the spindle (Morin et al. 2007; Peyre et al. 2011; Konno et al. 2008) (Fig. 2a). LGN is recruited to the cell cortex by GDP-bound Gαi subunits and recruits the Mud homologue NuMA to the lateral cortex (Fig. 2b). Removing LGN, NuMA, or Gαi, as well as interfering with the LGN/Gαi interaction, suppress spindle rotation during metaphase and result in defects in final spindle orientation at anaphase. Conversely, homogenization of the complex at the cell cortex by overexpression of Gαi subunits results in erratic spindle movements and random orientation. NuMA interacts directly with members of the dynein complex (Merdes et al. 1996, 2000), and it is postulated that the restricted localization of the LGN complex concentrates pulling forces at the lateral cortex to dictate the plane of cell division, although neither the lateral enrichment nor the requirement of dynein motor activity have been formally demonstrated in dividing vertebrate neuroepithelial or radial glial cells in vivo. This model is supported by the fact that dynein cortical recruitment depends on the LGN complex in the C. elegans zygote and in Hela Cell (Nguyen-Ngoc et al. 2007; Kiyomitsu et al. 2011), and that a lateral cortical distribution of dynein has been described in cultured MDCK epithelial cells during mitosis (Busson et al. 1998). In this case, lateral distribution is also regulated by the dynein complex binding partner lissencephaly 1 (Lis1) (Faulkner et al. 2000), and studies in the mouse cortex have shown that Lis1 is involved in planar spindle orientation (Yingling et al. 2008; Gauthier-Fisher et al. 2009). Remarkably, both loss of LGN and of Lis1 result in almost complete spindle randomization in apical progenitors (Morin et al. 2007; Konno et al. 2008; Yingling et al. 2008; Peyre et al. 2011), suggesting that the cortically anchored LGN complex, together with the dynein/dynactin complex, is the kingpin in the transmission of forces from the cell cortex to astral microtubules and the spindle.
While the core effectors of spindle orientation are evolutionarily conserved, their distribution can be very different between cell types and models, leading to different modes of division. The upstream regulatory events that regulate the precise localization of these effectors are poorly understood. Surprisingly, Gαi subunits are distributed homogeneously at the cell cortex, raising the question of how LGN and NuMA are restricted to a lateral ring in neural progenitors (Peyre et al. 2011). It may result from an active recruitment to the lateral cell cortex, or exclusion from the apical and basal regions of the cell. LGN has been shown to interact directly with a number of molecules implicated in epithelial polarity, but the relevance of these interactions remains to be tested in neural progenitors.
As Inscuteable is a central player in drosophila NB apico-basal spindle orientation, the role of mammalian Insc was also explored in different mammalian tissues, where it was shown to regulate apico-basal versus planar spindle orientation in a binary way both in the rat retina and in mouse skin basal progenitors (Zigman et al. 2005; Poulson & Lechler 2010; Williams et al. 2011). In the mouse developing neocortex, where low levels of Insc are expressed, apico-basal and oblique divisions are rare, and most cells divide in a planar fashion. Remarkably, in these cells, loss of Insc function increases the number of planar divisions, while its overexpression favors oblique and apico-basal divisions (Konno et al. 2008; Postiglione et al. 2011). Clearly, physiological levels of Insc expression in the cortex are not sufficient to instruct a binary choice between planar and apico-basal spindle orientation. Rather, Postiglione et al. propose that basal levels of Insc may induce a background level of imprecision that could control the balance between divisions in which the cleavage plane bisects or bypasses the apical domain in a stochastic manner (see below). The mode of action of Insc is not yet understood. Insc is enriched apically in dividing cells (Konno et al. 2008; Postiglione et al. 2011), possibly in a Par3 dependent manner (Izaki et al. 2006). Insc may actively orient the spindle by recruiting the LGN complex apically through the conserved Insc/LGN interaction (Izaki et al. 2006). Alternatively, it was also shown recently that Insc and NuMA compete for LGN interaction, with Insc having a much stronger affinity for LGN (Culurgioni et al. 2011; Zhu et al. 2011b). Insc may therefore inhibit the formation of the Gαi/LGN/NuMA/Dynein force generator and lower the precision of the planar orientation machinery. Further work on the dynamics of LGN distribution and interaction with Insc, NuMA, and other partners, will be needed to decipher the mechanism.
Polarity cues: The roles of regulators and markers of epithelial polarity
Horizontal and vertical axes of division are defined in relation with the tissue's epithelial organization and cell's apico-basal polarity. As a consequence, any molecule that controls epithelial organization is likely to impact the axis of division, but whether this reflects a specific function directly related to spindle orientation or an indirect effect might be difficult to decide. There are, however, a number of polarity regulators that may have a direct impact on spindle orientation in NE and RG cells. In in vitro models of epithelial cyst formation, it was shown that the key regulators of epithelial polarity cdc42, Par3, Par6 and aPKC, are all involved in regulating planar spindle orientation (Jaffe et al. 2008; Hao et al. 2010; Zheng et al. 2010; Durgan et al. 2011). In particular, in MDCK cells, direct phosphorylation of LGN by aPKC is required for the exclusion of LGN from the apical cortex and its lateral accumulation; moreover aPKC's apical localization requires Par3 (Hao et al. 2010). Remarkably, this mechanism does not seem to be essential in the chick neuroepithelium as aPKC is not necessary for the apical exclusion of LGN, and is not sufficient for its cortical exclusion (Peyre et al. 2011). Likewise, specific defects in spindle orientation have not been reported in vertebrate neuroepithelial or radial glial cells lacking Par complex members, although gain and loss of function of Par3 and Par6 affect the mode of division (proliferative versus neurogenic) of mouse cortical progenitors (Costa et al. 2008; Bultje et al. 2009).
In polarized epithelial cells, adherens junctions (AJs) and their associated proteins are the proposed sites of interaction between the cell cortex and astral microtubules, based on their subapical location and the observation of microtubule plus ends ending in the AJ region in interphase cells (reviewed in Pease & Tirnauer 2011). However, in dividing neuroepithelial cells, once the spindle has reached a planar orientation, both poles face the lateral cortex and are centered in the middle of the rings of LGN and NuMA, several micrometers below the subapical AJs (Peyre et al. 2011). Although astral microtubules are difficult to image in neuroepithelial cells, this suggests that they contact the lateral cortex rather than subapical AJs. Nevertheless, adhesion complexes may help position LGN and NuMA as they do in skin basal progenitors (Lechler & Fuchs 2005). Surprisingly, in mouse radial glial cells, inhibition of β1-integrin at the apical cell surface using a specific blocking antibody, as well as knockout of its interaction partner α2-laminin, resulted in slightly more planar divisions than in the control situation (Loulier et al. 2009). However, this phenotype is accompanied by delamination of interphase cells from the apical surface and it is not clear whether the spindle orientation phenotype is due to defective adhesion or signaling.
Other polarity cues could play a role in mitotic spindle orientation: the tumor suppressors dlg (discs large) and l(2)gl are essential regulators of cell polarity with conserved functions between invertebrates and vertebrates. In epithelial cells, they are typically localized on the basolateral cell membrane. However, in fly NB, dlg is slightly enriched at the apical cortex where it interacts with Pins. Through the khc-73 kinesin motor, dlg captures astral microtubules+ ends and favors the cortical accumulation and stabilization of Pins over spindle poles, allowing the Gαi/Pins/Mud complex to exert pulling forces. This function is normally masked in neuroblasts by the dominant role of the apical Par complex and Insc in the apical recruitment of Gαi/Pins/Mud and apical-basal orientation of the spindle (Siegrist & Doe 2005). The DLG/LGN and DLG/khc-73 interactions are conserved in vertebrates (Sans et al. 2005; Zhu et al. 2011a), and it will be interesting to investigate the involvement of DLG in LGN lateral localization and in planar spindle orientation in neuroepithelial and RG cells. The lgl2 vertebrate homologue of l(2)gl also interacts with LGN, and it may therefore be involved in spindle orientation through the lateral localization of LGN (Yasumi et al. 2005).
Microtubules, spindle poles and the mitotic spindle
Once polarity cues are interpreted to position force generators at the cell cortex, the second essential player in the orientation is the spindle itself. Obviously, any major defect in spindle assembly, such as the formation of mono or multipolar spindles is likely to affect its orientation in a non-specific manner, and this aspect will not be reviewed here. However, subtle defects in the molecular composition of spindle poles and in the production of astral microtubules may have more relevant effects. In addition to positioning the dynein complex at the plus end of astral microtubules, Lis1 is also localized at spindle poles, and loss of Lis1 reduces the density of astral microtubules in cultured mouse embryonic fibroblasts (Yingling et al. 2008). Lis1 interacts with the centrosomal protein Nde1 and loss of Nde1 also results in significant defects in planar spindle orientation in radial glia (Feng & Walsh 2004). This phenotype is slightly enhanced by the removal of one copy of Lis1 (Pawlisz et al. 2008). Interestingly, mutations in these genes are associated with brain size reduction: Lis1 takes its name from its association with lissencephaly in humans, and a homozygous mutation in Nde1, which abrogates its interaction with Lis1, causing severe microlissencephaly (Alkuraya et al. 2011; Bakircioglu et al. 2011). Human Primary Microcephaly (MCPH) is a condition in which patients show a severe brain size reduction and mild mental retardation, despite overall normal brain organization. One possible cause of MCPH is the premature production of neurons at the expense of progenitors, which would ultimately result in the production of fewer neurons. It was proposed that MCPH might result from defects in spindle orientation in dividing neural progenitors. Seven genes have been associated with MCPH: MCPH1 (microcephalin), WDR62, CDK5RAP2, CEP152, ASPM, CENPJ and STIL (Thornton & Woods 2009). Interestingly, all these genes encode proteins located on the mitotic spindle poles, and are involved in centrosomal biogenesis and/or spindle assembly. A potential involvement in mitotic spindle orientation has been investigated for several primary microcephaly genes in the mouse embryonic cortex. ASPM was found to be expressed at lower levels on the spindle poles of cells undergoing neurogenic divisions, and its loss of function by RNAi resulted in mild defects in spindle orientation (Fish et al. 2006). How ASPM controls spindle orientation is not clear, but it was shown that its C. elegans homologue ASPM-1 directly interacts with the NuMA homologue lin-5, and that this interaction is important for spindle positioning and orientation in meiotic cells. Interestingly, this function is independent of GPR1/2 (the nematode homologue of LGN) and Gα (Van der Voet, 2009). Similarly, both a conditional knockout of MCPH1 and a spontaneous mutation in CDK5RAP2 reduce the precision of planar divisions in the mouse embryonic cortex (Lizarraga et al. 2010; Gruber et al. 2011). Loss of CDK5RAP2 in cultured cells reduces the accumulation of the γ-Tubulin Ring Complex (γ-TuRC) at spindle poles and results in anastral spindles, which should therefore be unable to read polarity cues. How MCPH1 may control spindle orientation is less clear. It should be noted that both MCPH1 and CDK5RAP2 mutants display a high frequency of abnormal bipolar and multipolar spindles. It will be interesting to investigate whether other primary microcephaly genes control spindle orientation.
Mutations in the huntingtin gene (Htt) are responsible for human Huntington's disease. Htt interacts with the dynein/dynactin complex and is detected on the spindle poles of dividing cells. Knockdown of Htt in cultured cells results in a smaller mitotic spindle and reduces p150glued (a Dynactin subunit), Dynein and NuMA accumulation on spindle poles. This correlates with mitotic spindle misorientation both in cultured cells and in the mouse embryonic cortex (Godin et al. 2010). Similarly, knockdown of Lfc (Arhgef2), a Rho-specific guanine nucleotide exchange factor that decorates spindle microtubules, and of its negative regulator Tctex-1, both modulate the occurrence of vertical cleavage planes in the ventricular zone of the mouse brain and perturb neurogenesis (Gauthier-Fisher et al. 2009). Tctex-1 is also known as a component of the dynein light chain and it is not clear whether its role on spindle orientation is through Rho signaling or dynein regulation.
Cell shape and cortical tension
In the C. elegans embryo, the efficient translation of forces from the cortex to astral microtubules requires optimal rigidity of the cortex to prevent its collapse (Redemann et al. 2010). This is achieved by cell rounding and stiffening during mitosis. In embryonic mouse skin progenitors, cell rounding is accompanied by, and relies on accumulation of myosin IIa and phosphorylated ERM proteins at the cell cortex (Luxenburg et al. 2011). Treating these cells with inhibitors of actin dynamics disrupts cell rounding. Remarkably, it also affects the distribution of LGN and NuMA and results in spindle misorientation (Luxenburg et al. 2011). It will be interesting to study whether planar divisions and LGN and NuMA rings are similarly controlled by actin dynamics in neural apical progenitors. These cells also round up for mitosis, and evolve from a slightly elongated shape along the apicobasal axis at metaphase onset to adopt a spherical shape by the time of anaphase (notwithstanding their long basal attachment, which remains present as a very thin structure during division). Interestingly, expression of a dominant negative form of the small G protein RhoA results in slightly elongated mitotic cells in the chick neuroepithelium, and in random spindle orientation (Roszko et al. 2006). As RhoA is an upstream regulator of myosin activity, it will be fascinating to study how cell shape, cortical stiffness, and spindle regulation by the LGN complex are interconnected.
Much remains to be done to understand how spindle orientation is controlled in apical progenitors. Although many candidates have been identified in in vitro systems or in other epithelial or polarized cell types in vivo, their function is not always clear in neural progenitors. This highlights the fact that mechanisms may not be simply transposed from one epithelial type to another: modalities are very different, in particular due to the pseudo stratified nature of the neuroepithelial tissue. This section also shows that spindle orientation is controlled at many different levels and integrated with many other cellular events, such as cell polarity, centrosome biogenesis and spindle formation. As a result, specific functions might be difficult to uncover through classical loss of function experiments, and targeting specific protein–protein interactions may prove a useful approach to finely dissect the mechanisms of spindle orientation.
Orientation and cell fate
The diversity of cortical progenitors discussed above had not been described when the hypothesis of a link between spindle orientation and fate determination was first formulated. In particular, radial glia had yet to be identified as the apically dividing neural stem cells. With the identification of different progenitor types and the discovery of indirect neurogenesis in cortical development, the question of fate determination during neurogenesis has gained in complexity: dividing cells not only have to choose between symmetric (proliferative) and asymmetric (neurogenic) divisions, but they also need to make decisions between direct and indirect neurogenesis, and between an apical or a basal localization. This complexity adds a level of difficulty in the interpretation of the effect of defects in spindle orientation in neurogenesis.
Studies reporting the implication of some of the molecules discussed in the previous section in spindle orientation have also assessed their effect on cell fate determination in cortical and/or neural tube development. In many cases, a correlation between the control of mitotic spindle orientation and the balance between symmetric or asymmetric outputs has been observed. The loss of function of ASPM (Abnormal Spindle-Like Microcephaly Associated Protein) by RNA interference in the mouse developing telencephalon leads to the premature appearance of cells with neuronal characteristics in the cortical layers (Fish et al. 2006). Similarly, knockdown of Htt1 (Godin et al. 2010), Tctex-1 (Gauthier-Fisher et al. 2009), conditional KO of MCHP1 (Gruber et al. 2011) and mutations or downregulation of CDK5RAP2 (Buchman et al. 2010; Lizarraga et al. 2010) all result in early neuronal differentiation. Since all these conditions also reduce the accuracy of vertical cleavage planes, the premature differentiation observed may result from more frequent unequal partitioning of the apical membrane domain, resulting in more cells leaving the ventricular zone and engaging in early neuronal differentiation. While these correlative data seem to support the “bisect or bypass” model (Fig. 3a), it is important to note that they do not show a direct causal relationship between inheritance of the apical domain and stem cell identity or lack of apical domain and neuronal differentiation.
Over the past few years, new data have led us to reconsider the original model. First, using recently identified markers of apical RG cells versus IP cells, Noctor and colleagues have shown that the vast majority of RG divide with a vertical cleavage plane throughout cortical development (Noctor et al. 2008)(Fig. 3b). RG horizontal or oblique cleavage planes are seldom and equally frequent between the symmetric proliferation phase and the neurogenic phase. However, during the neurogenic phase, a population of Tbr2 positive IP cells divides close to the ventricular surface and preferentially harbors oblique or horizontal cleavage planes. These cells may have been mistaken for apical progenitors in earlier studies in which these specific markers were not used (Chenn & Mcconnell 1995; Haydar et al. 2003).
Second, in rat cortical slice cultures, live observations of asymmetrically dividing GFP-labelled RG cells with vertical cleavage plane (i.e. planar division) led to the conclusion that orientation of the cleavage plane could be independent from the acquisition of cell fate in the progeny (Noctor et al. 2008)(Fig. 3b). Similarly, in the chick neural tube cultured slices, Wilcock et al. have observed that neurons are often produced through the delamination and differentiation of cells that initially retained an apical attachment at the time of division (Wilcock et al. 2007). In the mouse cortex, Tbr2 positive IP cells are also produced by delamination of apical cells after they downregulate cdc42 and retract their apical attachments, and a conditional cdc42 knockout increases IP production at the expense of apical RG cells, without any effect on spindle orientation in the RG (Cappello et al. 2006). In the mouse embryonic cortex, en-face time lapse imaging of sub-apical membrane labeled with an EGFP-ZO1 reporter revealed that equal apical surface inheritance at the time of cytokinesis is frequently followed by retraction of one of the apical attachments, suggesting subsequent asymmetric fates for the sister cells (Shitamukai et al. 2011). Thus, these observations strongly suggest that asymmetric cell fate acquisition is not correlated with orientation of the mitotic spindle and apical membrane inheritance.
Moreover, in the mouse cortex as well as the chick spinal cord, disruption of the cortical effectors of spindle orientation (Gαi, LGN, or NuMA) resulted in randomization of spindle orientation, but contrary to the prediction of the model, this did not result in accelerated neurogenesis. Rather, it caused the scattering of progenitors in the subventricular zone (Morin et al. 2007; Konno et al. 2008; Peyre et al. 2011) (Fig. 3b). In vivo clonal cell fate analysis indicated that these ectopic progenitors result from the unequal inheritance of apical attachment sites upon division, but that they retain the molecular signature of apical progenitors (Sox2, Notch1, Hes5 expression in the chick neural tube; Pax6, Hes1 in the mouse cortex) (Morin et al. 2007; Konno et al. 2008). Indeed, in the mouse embryonic cortex, these ectopic progenitors are virtually identical to oRG cells, as they also retain a basal process (Shitamukai et al. 2011). Altogether, these data suggest that the primary role of planar spindle orientation in apical divisions is to maintain daughter cells attached to the ventricular surface, but that it does not directly influence the choice between symmetric and asymmetric outcomes. Note that the ectopic position of progenitors may secondarily influence their fate: for example, in the chick neural tube, ectopic progenitors were shown to subsequently proliferate in excess compared with their ventricular counterparts (Morin et al. 2007).
Interestingly, the recent description of the phenotypes associated with loss or gain of function of Insc in the mouse cortex brings a new level of complexity to the field. Data from Postiglione et al. suggest that spindle orientation does not control symmetric (proliferative) versus asymmetric (neurogenic) choices, but that it may indeed influence cell fate decisions within the neurogenic population (Postiglione et al. 2011). Overexpression of Insc, which increases the number of oblique and apico-basal divisions, does not reduce the number of Pax6 positive apical RG cells. However, it increases the number of Tbr2 positive IP cells. Conversely loss of Insc increases planar divisions and reduces the number of IP cells without affecting apical RG cell number. The authors propose that during the neurogenic phase, within the population of asymmetrically dividing apical RG cells, the axis of division may control the choice between direct and indirect neurogenesis: planar divisions would produce an apical RG and a postmitotic neuron, while oblique or apico-basal divisions would produce an apical RG and an IP cell. However, as underlined in an accompanying commentary (Wang et al. 2011a), the direct production of IP cells from oblique/apico-basal divisions of apical progenitors upon Insc overexpression has not been demonstrated. It remains possible that these divisions instead generate oRG (as described upon LGN misregulation by Shitamukai et al. 2011), which would then rapidly divide and produce IP-like cells expressing Tbr2.
The question of the relationship between mitotic spindle orientation and cell fate has also been addressed in the developing vertebrate retina, which contains specific types of progenitors and shows a different organization as compared with the cortex and spinal cord. In the rat retina, Cayouette and Raff first showed that the balance between planar and apico-basal divisions varies during development and observed that the fate determinant Numb is asymmetrically inherited by the apical daughter cell, suggesting that apico-basal divisions are asymmetric (Cayouette et al. 2001; Cayouette M 2003). It should be noted that a similar apical distribution of Numb had initially been described in dividing RG of the mouse cortex (Zhong et al. 1996), but later data suggested that this may actually be essentially accounted for by the apical enrichment of Numb in the endfeet of neighbouring interphase cells (Rasin et al. 2007). Nevertheless, time-lapse imaging of rat retinal progenitor divisions and differentiation suggested that apico-basal divisions produce progenies with a different cell more frequently than planar divisions (Cayouette M 2003). Besides, downregulation of Inscuteable in the rat retina reduced the frequency of apico-basal divisions and enhanced proliferation, suggesting an increase in symmetric proliferative divisions (Zigman et al. 2005). In contrast to the rat model, in the zebrafish retina, progenitors essentially divide in a planar fashion. However, the orientation of divisions within the plane of the retina (radial versus circumferential with respect to the position of the lens) was shown to correlate with symmetric versus asymmetric fate decisions (Das et al. 2003), and this orientation as well as the fate decisions appear to be controlled at the population level by environmental signals to regulate the balance between cell types in the retina (Poggi et al. 2005). Hence, in the retina, fate decisions appear to be correlated with spindle orientation, although the modalities may differ between models.
If not spindle orientation, then what? Mechanisms of cell fate acquisition
It is currently difficult to interpret and reconcile the conflicting data presented in the previous section and the exact role of spindle orientation during neurogenesis remains controversial. Clearly, further work, using live observation and clonal analysis of RG lineage trees in vivo, both in wild type and mutant conditions, will be needed to distinguish unambiguously between the different scenarios. In addition, it will be essential to identify the determinants that regulate fate choices and their distribution during division. The following section explores a number of candidate molecules and cellular structures that have emerged as potential fate regulators over the last few years.
Notch signaling is a common pathway used in binary fate decisions. Notch downstream genes Hes1 and Hes5 antagonize pro neural genes such as the neurogenin2 (Ngn2) transcription factor or the Notch ligand DLL1, allowing the maintenance of the stem cell identity (Ishibashi et al. 1995; Gaiano & Fishell 2002; Yoon K 2005). Although asymmetric Notch protein has been observed in early work in dividing cells of the ferret cortex (Chenn & McConnell 1995), this has not been described in mouse progenitors. Newborn sister cells actually have low but similar amounts of Notch activity (Vilas-Boas et al. 2011), although the level of Notch signaling is oscillating in the neural progenitors during the different phases of the cell cycle in a ligand independent fashion (Shimojo et al. 2008). Soon after birth one of the daughter cells exhibits increasing amounts of Ngn2 and Tbr2 consistent with Notch repression and differentiation (Ochiai et al. 2009). It is proposed that differential Notch signaling between sister cells may be regulated by a combination of cell intrinsic and extrinsic mechanisms. It was observed that Par3 is released from the subapical cortex, becomes cytoplasmic during mitosis and is inherited asymmetrically by one daughter cell in about 50% of RG cell divisions during neurogenesis (Bultje et al. 2009) (Fig. 4). Based on different combinations of gain and loss of function for Par3 and for the Notch inhibitor Numb, Bultje et al. conclude that Par3 inhibits Numb's inhibitory activity on Notch signaling. The cell receiving more Par3 may sequester Numb after divisions, favor Notch activation and retain the progenitor identity, whereas the sister cell with low Par3 would have high Numb activity, low Notch signaling and would therefore differentiate. In addition to this autonomous intracellular modulation, Notch signaling may also be regulated between sister cells by their differential exposure to its ligands, controlled by their asymmetric inheritance of the basal fiber (see below) (Fig. 4).
An intriguing feature of apical and outer RG cells is the persistence of the basal fiber. Upon division, the basal fiber is generally inherited by only one daughter cell (Miyata et al. 2001; Noctor et al. 2001; Das et al. 2003). It was initially thought to serve solely as a support structure for radially migrating neurons, and was not regarded as a potential identity determinant as stem cell determinants were thought to be present at the apical membrane. However, several recent observations suggest its importance for stem cell identities. The newly described oRG population of neural stem cells does not possess apical attachment sites but has a basal fiber extending toward the pial surface (Fietz et al. 2010; Hansen et al. 2010). Moreover, in apical neural progenitor cells in the mouse cortex and the zebrafish anterior spinal cord, lineage analyses have shown that upon oblique cleavage planes, the apical cell that does not inherit a basal attachment differentiates (Konno et al. 2008; Alexandre et al. 2010). The basal cell that inherits the basal fiber reattaches to the apical surface and resumes its asymmetric divisions in the zebrafish. In the mouse brain this cell often delaminates from the ventricular zone (VZ) and acquires characteristics of oRGs (Shitamukai et al. 2011). Thus, the basal fiber seems to be an essential feature of the neural stem cells and its inheritance could play a role in maintaining the RG identity. For example, it could transmit Notch signaling as the Delta ligand is expressed by neurally committed cells migrating along the fiber toward cortical outer layers (Yoon et al. 2008) (Fig. 4). The presence of the basal fiber could then be a determinant for the maintenance of the progenitor identity as it could transmit larger amounts of Notch signaling from the migrating neural cells and repress more effectively pro-neural genes. Other signals originating from the pial surface where the basal fiber attaches could also be at play, such as Integrins and growth factor signaling pathways.
Interestingly, it has been reported that the basal fiber could be bisected during mitosis in the majority of neuroepithelial cells of the mouse embryonic cortex at E10.5 (before the appearance of RG markers) and zebrafish neural tube (Kosodo et al. 2008). It is not clear whether splitting of the basal process correlates with symmetric divisions: it was observed at equal frequency between cells positive or not for the neurogenic reporter Tis21-GFP, but in many cases both processes were inherited by only one daughter cell. Further exploration of the phenomenon is needed to determine whether equal inheritance of the split basal process correlates with symmetric outcome in early stages.
Other intrinsic asymmetric determinants
Many mechanisms could influence cell fate acquisition independently of mitotic spindle orientation. For example, differential degradation rate of cellular components is a way that introduces asymmetry between the daughter cells and leads to different fates. Irrespective of their axis of divisions, the ubiquitin ligase Trim32 shows a basal accumulation in the cell body in two-thirds of dividing neural progenitors, and is preferentially co-inherited with the basal attachment (Fig. 4). This asymmetric repartition of Trim32 results in differential target protein degradation between daughter cells (Schwamborn et al. 2009). Overexpression of Trim32 induces neural differentiation and its loss of function leads to an increase of symmetric proliferative divisions both in vivo and in vitro, a rather surprising result given the preferential maintenance of RG identity by cells that inherit the basal attachment (Konno et al. 2008; Shitamukai et al. 2011). An EGF receptor was also proposed to segregate asymmetrically and maintain the radial glial identity during neurogenesis (Sun et al. 2005). This asymmetry does not correlate with the orientation of the spindle relative to apical-basal cell polarity and was also observed in cultured progenitors in vitro.
The centrosome is one of the building blocks of the mitotic spindle and it gives rise to an inherent asymmetry in the cell. This organelle is composed of two centrioles which replicate in a semi-conservative fashion during interphase. Hence, in a dividing cell, one spindle pole contains a “young” centrosome composed of a newly synthetized and a one-cycle-old centriole, while the other “old” centrosome is composed of a newly synthetized and a more mature centriole that is at least two cell cycles old and typically carries more pericentriolar material, including specific proteins required for microtubule anchoring and ciliogenesis (for review see Nigg & Stearns 2011).
A link between cell fate acquisition and centrosome asymmetry was first established in the drosophila male germline where the mother centrosome maintains a developed aster close to the hub cell (Yamashita et al. 2007). During mitosis the old centrosome is systematically inherited by the stem cell, while the younger centrosome is inherited by the differentiating cell. In the drosophila neuroblast, centrosome asymmetry also correlates with asymmetric fate determination, although in that case the stem cell identity is correlated with inheritance of the younger centrosome (Januschke et al. 2011). In vertebrates, centrosome asymmetry also seems to play an important role during asymmetric division of neural progenitors. In the mouse neocortex, Wang and colleagues used a photo-convertible fluorescent protein fused to Centrin-1 to differentially tag the old and young centrosomes in RG cells. They observed that the older centrosome is preferentially inherited by the daughter cell retaining the stem cell fate (Wang et al. 2009). They also showed that maturation asymmetry seems to influence the daughter cell fate: upon RNAi knockdown of Ninein, a mature centriole-specific protein, asymmetric distribution of the centrosomes was perturbed and the neural progenitor population was reduced. Similarly, loss of function of Hook3 and pericentriolar material 1 (Pcm1), two pericentriolar material assembly proteins, also led to reduction of the neural progenitor population and overproduction of neurons (Ge et al. 2010). The mechanisms allowing mature centrioles to grant stem cell identity are still unknown; centrosome maturation proteins may provide a network that facilitates accumulation of an as yet unidentified stemness determinant over one spindle pole (Fig. 4). With regard to the putative role of centrosome asymmetry in cell fate determination, it is interesting to note that a large number of proteins implicated in spindle orientation are located at centrosomes or are linked to centrosome function. The Dynein/Dynactin complex, through its interacting partner Lis1, plays an important role in orienting the mitotic spindle via anchorage of cortical Dynein and microtubule stabilization (Yingling et al. 2008), but also regulates centrosome separation during mitosis (Kardon & Vale 2009). Defects in cell fate acquisition associated with Lis1 loss of function could therefore result from defective centrosome regulation rather than abnormal spindle orientation. Human autosomal recessive primary microcephaly is essentially caused by mutations in genes that code for centrosomal proteins and regulate the number of centrosomes (Cox et al. 2006; Boehlke et al. 2010; Mahmood et al. 2011). A specific role of centrosomes in cell fate acquisition could then explain why some mutations that affect spindle orientation also affect fate determination (such as ASPM or Htt1), while others do not (e.g Gαi/LGN/NuMA). In support of this idea, although LGN and NuMA are also detected on or near spindle poles, a dominant approach targeting the LGN/Gαi interaction at the cell cortex disrupts spindle orientation and does not affect fate determination (Morin et al. 2007). Interestingly, a role of centrosomes in the asymmetric segregation of fate determinants in the pericentriolar material would de facto eliminate the need of a specific spindle orientation to resolve fate choices, as cytokinesis automatically separates centrosomal components into sister cells.
Primary cilia: The cell's antenna
One way through which centrosome maturation may affect fate determination is regrowth of the primary cilia after division. Primary cilia are small sensor structures on which cells localize receptors for a number of extracellular signals, including Shh and platelet-derived growth factor (PDGF). As cilia are dismantled during cell division, the timing of their regrowth after cytokinesis could affect their sensing and response to these signals and influence their fate. Indeed, in cultured mouse fibroblasts, the cell inheriting the mother centrosome regrows a primary cilium faster after mitosis, so that sister cells respond asymmetrically to SHH signaling and differentially express PDGF receptors (Anderson & Stearns 2009). The difference in maturity between centrosomes could also intrinsically influence their ability to nucleate microtubules, modulating the anchoring to the apical surface and the apical localization of the new cilium (Wang et al. 2009). In this respect, it was recently described that during neurogenesis, a number of cells produce their cilium on their basolateral side, rather than on the apical cell surface, after division (Wilsch-Brauninger et al. 2012). Using a Tbr2-GFP transgene, Wilsch-Bräuninger and colleagues identify cells with a basolateral primary cilium as future Tbr2 positive IP cells that delaminate to accumulate in the SVZ. They propose that the production of a basolateral cilium exposes the cell “antenna” to signals that are different between the basolateral intercellular space and the ventricle surface and thus alters cell identity and promotes delamination. It is tempting to speculate with them that this specific behavior is linked to differential centrosome maturation (Fig. 4). It is not clear how these observations can be reconciled with the reported effect of Insc overexpression on oblique spindle orientation and Tbr2+ cell production.
Over the last decade, the description of the diversity of neural progenitors in the developing nervous system and particularly in the mammalian neocortex has made tremendous progress. In parallel, our understanding of the mechanisms regulating the fate choices that control the balance between these different populations has also improved, but it remains fragmentary. A number of fate determinants have been proposed, and it is not yet possible to reconcile all the observations in one single coherent model of neurogenesis progression. This is in part due to the fact that different studies address similar questions at different stages, in different models and use different analytical tools to interpret the data. It is also very likely that fate determination consists of many superimposed layers of molecular decisions, none of which result in binary choices, and whose particular combination increases the chances of a cell to go along a particular path. Regarding the specific question of spindle orientation, there is currently little convincing evidence for a direct implication in fate decisions based on a classical mechanism of intrinsic asymmetric division. We propose that its main role is to regulate whether two sister cells will retain apical attachments at the time of division. This process may participate in fate choices by controlling the overall organization of the tissue and by regulating cellular interactions and extracellular signal reception. Clearly, identifying these signals and understanding how they are integrated in space and time, will require many years of exciting research. A key aspect to this research will be the ability to distinguish between cell autonomous and non-autonomous events and to follow cellular decisions in cell lineages in vivo, using combinations of clonal analyses and live imaging.
We would like to thank members of the Morin lab for helpful comments on the manuscript. Work in X.M.'s laboratory is supported by an INSERM Avenir Grant (R08221JS), the Fondation pour la Recherche Medicale (FRM implantation nouvelle équipe), the Association pour la Recherche contre le Cancer (projet ARC 2011-LIVESPIN) and institutional grants from INSERM, CNRS, and the ENS. E. Peyre was the recipient of a pre-doctoral grant from the French Ministry for Research (MRT) and of the Association pour la Recherche contre le Cancer (ARC).