Asymmetric segregation of Numb: a mechanism for neural specification from Drosophila to mammals
Michel Cayouette & Martin Raff
Nature Neuroscience 5,1265-1269 (2002)
doi:10.1038/nn1202-1265
doi:10.1038/nn1202-1265
December 2002 Volume 5 Number 12 pp1265-1269
Asymmetric segregation of Numb: a mechanism for neural specification from Drosophila to mammals
Michel Cayouette1, 2 & Martin Raff1
1. MRC Laboratory for Molecular Cell Biology and Cell Biology Unit, University College London, Gower Street, London WC1E 6BT, UK
2. Present address: Stanford University, Department of Neurobiology, Fairchild Building, 299 Campus Drive, D231, Stanford, California 94305-5125, USA
Correspondence should be addressed to MCayouette.e-mail:
It is a major challenge to understand how the neuroepithelial cells of the developing CNS choose between alternative cell fates to generate cell diversity. In invertebrates such as Drosophila melanogaster and Caenorhabditis elegans, asymmetric segregation of cell-fate determining proteins or mRNAs to the two daughter cells during precursor cell division plays a crucial part in cell diversification. There is increasing evidence that this mechanism also operates in vertebrate neural development and that Numb proteins, which function as cell-fate determinants during Drosophila development, may also function in this way in vertebrates. Recent studies on mouse cortical progenitor cells have provided the strongest evidence yet that this is the case. Here, we review these and other findings that suggest an important role for the asymmetric segregation of Numb proteins in vertebrate neural development.
The mammalian brain contains hundreds of cell types that develop from neuroepithelial cells that initially all look alike. It is still uncertain to what extent asymmetric cell division of neuroepithelial cells contributes to this cell diversification process. The term 'asymmetric cell division' is often used to describe any division in which the two daughter cells adopt different fates1. Because identical daughter cells of a symmetric division could be induced to adopt different fates as a result of their exposure to different extracellular signals, we use the term asymmetric cell division in its more restricted sense, to mean a division in which the two daughter cells inherit different cell-fate determinants from the mother cell.
We begin by briefly reviewing the key findings in Drosophila, where the asymmetric segregation of cell-fate determinants is known to be crucial for neural cell diversification2, 3. These findings provide a framework for considering the evidence that similar mechanisms operate in vertebrates.
Asymmetric segregation of Numb in Drosophila
Our understanding of the molecular basis of asymmetric cell divisions in animal development has come mainly from studies in C. elegans and Drosophila. In Drosophila, asymmetric segregation of the Numb protein influences cell-fate choice in both neural and non-neural development4-9. Numb is a plasma membrane–associated cytoplasmic protein that contains a phosphotyrosine-binding (PTB) domain and antagonizes Notch signaling10-12. Numb is thought to become associated with the plasma membrane through phosphotyrosine binding, and in mammalian cells, there is evidence that different PTB domain isoforms may regulate the subcellular localization of Numb proteins13. In the Drosophila peripheral nervous system, each external sensory organ develops from a single sensory organ precursor (SOP) cell that produces the four cells that form the organ. In the first division of the SOP cell, Numb localizes asymmetrically at one pole of the mitotic cell cortex. It overlies one pole of the mitotic spindle so that only one daughter cell inherits the protein. As a result, this daughter becomes a IIb cell, whereas the other becomes a IIa cell6. These two cells then divide to produce the different cell types of the sensory organ (Fig. 1a). In Numb-deficient mutants, the first SOP division produces two IIa cells (Fig. 1b), whereas if Numb is overexpressed, this division produces two IIb cells (Fig. 1c). These results suggest that Numb promotes the IIb cell fate.
A good deal is known about the molecular machinery that localizes and segregates Numb and how Numb influences cell fate in Drosophila (for review, see refs. 2,3,14). In Drosophila neural precursors, for example, the Inscuteable protein localizes asymmetrically to the apical cortex before mitosis and directs mitotic spindle orientation and the segregation of Numb to the basal daughter cell. In SOP cells, Numb influences cell fate by repressing Notch signaling, although in other cases its mechanism of action is unclear. In general, it seems that that the primary function of Numb is not to specify a particular cell fate, but rather to create two daughter cells that can respond differently to environmental cues.
Plane of cell division and cell-fate choice in vertebrates
The neuroepithelial cells of the developing vertebrate CNS function as neural progenitor cells. They divide extensively and produce a large variety of cell types. The type of cells they produce and the rate at which they produce them change as development proceeds. Early on, cell divisions are largely proliferative, each producing two neuroepithelial cells. Later, most cell divisions are differentiative, each producing either one daughter that differentiates and one that remains a neuroepithelial cell, or two daughters that differentiate. The early differentiative divisions produce neurons, and the later ones produce glial cells or their precursors.
Although in Drosophila the orientation of cell division plays a critical part in segregating cell-fate determinants, and thereby in cell-fate choices, it is less clear if the plane of neuroepithelial cell division influences cell fate choices in vertebrates. The first direct evidence that the plane of neuroepithelial cell division changes with development and may influence cell fate came from pioneering studies by Chenn and McConnell15. They used videomicroscopy to follow cell divisions in the ventricular zone of explants of developing ferret cortex. They found that early in development, most cells divide with their mitotic spindle aligned horizontally to the plane of the neuroepithelium, and later in development, more cells divide with their mitotic spindle aligned vertically to the plane of the neuroepithelium. As their name suggests, neuroepithelial cells have some epithelial cell characteristics, including apical–basal polarity and molecularly distinct apical and basolateral plasma membrane domains. Thus, neuroepithelial cells that divide with a horizontal spindle are likely to distribute both apical and basolateral components equally to their daughter cells, whereas the cells that divide with a vertical spindle are likely to distribute apical components preferentially to the apical daughter cell (Fig. 2). Indeed, there is molecular evidence that the divisions with a vertical spindle are asymmetric, in that an unidentified antigen recognized by anti-Notch1 antibodies segregates exclusively to the basal daughter cell15.
The Chenn and McConnell study15 also suggests that the plane of neuroepithelial cell division can influence the fate of the two daughter cells, raising the possibility that asymmetric segregation of cell-fate determinants might contribute to cell diversification. In the developing cortical neuroepithelium, most daughter cells either stay in the epithelium and divide again, or migrate away and differentiate. Using videomicroscopy of cortical slices, Chenn and McConnell found that in divisions with a horizontal spindle, both daughter cells tend to stay in the neuroepithelium, suggesting that they remain neuroepithelial cells. In divisions with a vertical spindle, by contrast, the basal daughter cell tends to migrate away, suggesting that it committed to differentiation; the apical daughter of such divisions tends to stay in the epithelium, suggesting that it remains a neuroepithelial cell. Somewhat later in development, it is likely that many neuroepithelial divisions produce two daughter cells that commit to differentiation, but such divisions were not included in that study, and it remains unknown whether they occur with a horizontal or vertical spindle.
Some neuroepithelial cells in the developing rat retina also divide with a vertical spindle16. In recent timelapse videomicroscopy experiments in newborn rat retinal explants, we found that the two daughter cells of divisions with a horizontal spindle tend to become the same cell type, whereas the two daughter cells of divisions with a vertical spindle tend to produce daughters that become different (M.C. and M.R., unpub. data). Thus, the plane of cell division can apparently influence cell-fate choice in the retina, as well as in the cortex. In both cases, it is likely to do so through the asymmetric segregation of cell-fate determinants (see below).
In Drosophila, cell-fate determinants such as Numb can be asymmetrically segregated either along the apico–basal axis of a dividing epithelial cell (in divisions with a vertical spindle) or perpendicular to this axis—in the plane of the epithelium (in divisions with a horizontal spindle)17. In principle, therefore, divisions with a horizontal spindle could also segregate cell-fate determinants asymmetrically in vertebrate neuroepithelia, as suggested by others18, but this remains to be shown.
Asymmetric segregation of mammalian Numb
Two groups independently identified a mammalian homolog of Drosophila Numb (m-Numb)19, 20. The m-Numb protein is located at the apical cortex of mitotic neuroepithelial cells in both the developing mouse cortex and rat retina16, 20. Thus, when a neuroepithelial cell divides with a vertical spindle, only the apical daughter cell inherits m-Numb, whereas when it divides with a horizontal spindle, both daughter cells inherit m-Numb.
As m-Numb is localized apically in mitotic mouse cortical neuroepithelial cells, and the apical daughter of divisions with a vertical spindle tends to remain a neuroepithelial cell15, it has been suggested that early in development, m-Numb acts to inhibit the differentiation of the neuroepithelial cells21. At later stages, however, it seems to be permissive for neuronal differentiation. As we discuss later, another mouse homologue of Numb called Numblike seems to act redundantly with Numb in mouse neurogenesis. Although Numblike is a cytoplasmic protein in Drosophila and distributes symmetrically in neuroblasts, the level of Numblike in vertebrate neuroepithelial cells is too low to be sure of its subcellular localization22.
A Numb homologue has also been identified in chickens23. Surprisingly, it localizes basally rather than apically in dividing avian neuroepithelial cells. Thus, chicken Numb would be segregated to the basal daughter rather than the apical daughter when a neuroepithelial cell divides with a vertical spindle. The significance of this difference between mammalian and chicken Numb distribution remains unknown.
Effects of altering Numb expression
To help determine the function of Numb proteins in vertebrate neural development, both gain-of-function and loss-of-function experiments have been done. Expression of an m-numb transgene in mammalian neural cell lines was found to be permissive for neuronal differentiation24, whereas in similar experiments in chick neuroepithelial cells in vivo, it was found to allow neuronal differentiation in some cells but to inhibit differentiation and promote proliferation in others23. The simplest interpretation of these findings is that Numb overexpression has different effects on neuroepithelial cells at different stages of maturation, as we discuss later.
Two groups independently generated m-numb knockout mice, which die around E11.5 with CNS abnormalities. One group found that cells expressing neuronal markers appear earlier in the m-numb mutant forebrain than in wild-type forebrain21, consistent with the possibility that m-numb can normally suppress neuronal differentiation. A second group25, by contrast, found a delay in neuronal differentiation in the hindbrain of m-numb mutants compared to wild type, consistent with the possibility that m-numb can be permissive for neuronal differentiation. The significance of these contradicting results remains unexplained, but recent double-knockout experiments provide some clarification about the role of m-Numb in early CNS development22. Although numblike knockout mice appear normal, conditional inactivation of m-numb in neuroepithelial cells in numblike-deficient mice gives a much more severe CNS phenotype than that seen in m-numb knockout mice, with a dramatic decrease in neuroepithelial cells22 (Y. N. Jan, pers. comm.). These findings suggest that m-Numb and Numblike may cooperate to promote neuroepithelial cell self-renewal in early neurogenesis.
Taken together, these findings indicate that Numb proteins are essential in vertebrate neural development, but they do not indicate whether the asymmetric segregation of these proteins at mitosis influences the fate of the two daughter cells.
Segregation of Numb and vertebrate cell fate
It has recently been shown that the asymmetric segregation of m-Numb can influence the developmental fate of daughter cells26. This study used a powerful clonal culture system in which the development of isolated embryonic mouse cortical neuroepithelial cells in microwells can be followed by videomicroscopy. Previously, by recording each cell division and then staining the differentiated cells with antibodies to identify the cell types to which the divisions gave rise, these investigators were able to reconstruct the cell lineage trees that revealed some common patterns of cell division and differentiation27, 28. As found in vivo, neurons developed before glial cells, and some divisions produced daughter cells that developed similarly, whereas others produced daughters that developed differently. Because the environment was relatively uniform in these cultures and the cells could not contact cells outside their own clone, it seems likely that these differences in behavior reflect intrinsic differences between the neuroepithelial cells.
In more recent experiments, these investigators focused on pairs of daughter cells that originated from a single neuroepithelial cell in culture26. Using immunostaining to detect m-Numb, they found that about 30–40% of such daughter cells segregate m-Numb asymmetrically into one of the two daughter cells. Thus, as previously reported in retinal neuroepithelial cells16, dissociated cortical neuroepithelial cells can segregate m-Numb asymmetrically when they divide in culture. Most importantly, they found that the asymmetric distribution of Numb correlates with the fate of the daughter cells. Using antibodies to identify neurons, they observed three types of cell divisions: those that produced two neuroepithelial (progenitor) cells (P/P), those that produced two neurons (N/N) and those that produced a neuroepithelial cell and a neuron (P/N). Using double-staining experiments, in which pairs of daughter cells were stained for both m-Numb and a neuronal marker, they found that in more than 80% of the P/N divisions, m-Numb is distributed asymmetrically, whereas in 80–90% of the N/N divisions, m-Numb is distributed symmetrically. Thus, when m-Numb is distributed asymmetrically between daughter cells, the daughters usually have different fates (P/N); when m-Numb is distributed symmetrically, the daughters usually have the same fate (N/N).
These findings raise the possibility that m-Numb directly influences cell-fate choice. To test this possibility further, this group compared the behavior of cortical progenitor cells from m-numb knockout mice with those from wild-type littermates26. Although Numblike was presumably still present in these cells, they found that in the m-Numb-deficient cells, the proportion of P/N divisions was decreased and the proportion of P/P divisions was increased. The proportion of N/N divisions, however, remained unchanged. This finding suggests that, without m-Numb, neuroepithelial cells are less able to produce two different daughter cells, consistent with the possibility that m-Numb influences asymmetric cell-fate decisions.
Is the daughter cell that inherits m-Numb in an asymmetric cell division more or less likely to become a neuron? In E10 cells, the m-Numb-containing daughter cell in P/N cell pairs is about equally likely to be a neuroepithelial cell or a neuron. By contrast, in E14 cells, the m-Numb-containing daughter cell in a P/N cell pair is usually the neuron. This is another example of Numb proteins having different effects at different stages of development. We return to this point below.
The influence of m-Numb in asymmetric cell divisions that produce two neurons has also been studied26. In 92% of the N/N pairs with asymmetric m-Numb distribution, the neurons had different morphologies after 4 days in culture: the neuron that inherited m-Numb usually had longer processes, raising the possibility that m-Numb may also influence the subtype or morphology of neurons in N/N divisions.
Studies in the newborn rat retina also suggest that the asymmetric segregation of m-Numb influences cell fate. As mentioned earlier, in retinal neuroepithelial cells that divide with a vertical spindle, m-Numb is inherited preferentially by the apical daughter cell15, and these divisions produce daughter cells that tend to become different cell types; in divisions with a horizontal spindle, both daughters inherit m-Numb16 and tend to become the same cell type, usually photoreceptor cells (M.C. & M.R., unpub. data). Moreover, overexpression of m-Numb in dividing retinal neuroepithelial cells increases the development of photoreceptor cells and decreases the number of non-photoreceptor cells (M.C. & M.R., unpub. data). Thus, the asymmetric segregation of m-Numb may contribute to cell diversification in a number of regions of the CNS.
How do Numb proteins influence cell fate?
Numb proteins have been shown to inhibit Notch signaling in both Drosophila and vertebrates, possibly by binding to the cytosolic tail of transmembrane Notch10-12. Recent work by Knoblich and colleagues suggests that Drosophila Numb inhibits Notch signaling by recruiting alpha-Adaptin, which would be expected to stimulate the endocytosis of Notch29. The same mechanism may operate in vertebrate cells, as m-Numb is associated with clathrin-coated pits, vesicles and endosomes30. These results also raise the interesting possibility that Numb proteins might regulate the internalization of various receptors involved in cell-fate decisions. Numb proteins also bind to other intracellular proteins such as Siah-1 and LNX. These proteins can regulate Numb function, but it is still unclear what roles they have in development31-34.
It is well known that Notch signaling has different effects at different times in development35. Early in development, for example, Notch mediates lateral inhibition, in which a differentiating cell inhibits its neighbors from differentiating as well. Later in development, Notch signaling may promote the differentiation of some cell types (for example, refs. 36–38). Thus, even if Numb proteins act exclusively by inhibiting Notch signaling, they would be expected to have different effects at different times of development and in different circumstances. In Drosophila, Numb acts at multiple stages in both the SOP and neuroblast lineages: it initially helps daughter progenitor cells to adopt different fates, and it later helps terminally differentiating cells to adopt different fates. As mentioned earlier, overexpression of Numb in chick neuroepithelial cells causes some cells to undergo premature differentiation and others to delay differentiation and continue to proliferate23. It is possible that these different responses reflect differences in the maturation of the developing chick neuroepithelial cells. In light of the results discussed above26, it seems that the effect of Numb on mouse neuroepithelial cells also changes as development progresses.