Radial glia diversity: A matter of cell fate


  • Arnold R. Kriegstein,

    Corresponding author
    1. Department of Neurology, Columbia University College of Physicians and Surgeons, New York, New York
    • Department of Neurology, Columbia University College of Physicians and Surgeons, P&S Building, Rm 4-408, 630 W. 168th Street, New York, NY 10032
    Search for more papers by this author
  • Magdalena Götz

    Corresponding author
    1. Max-Planck Institute of Neurobiology, Planegg-Martinsried, Germany
    • Max-Planck Institute of Neurobiology, Planegg-Martinsried, Germany
    Search for more papers by this author


Early in development of the central nervous system, radial glial cells arise from the neuroepithelial cells lining the ventricles around the time that neurons begin to appear. The transition of neuroepithelial cells to radial glia is accompanied by a series of structural and functional changes, including the appearance of “glial” features, as well as the appearance of new signaling molecules and junctional proteins. However, not all radial glia are alike. Radial glial lineages appear to be heterogeneous both within and across different brain regions. Subtypes of neurogenic radial glia within the cortex, for example, may have restricted potential in terms of the cell types they are able to generate. Radial glia located in different brain regions also differ in their expression of growth factors, a diverse number of transcription factors, and the cell types they generate, suggesting that they are involved in regionalization of the developing nervous system in several aspects. These findings highlight the important but complex role of radial glia as participants in key steps of brain development. GLIA 43:37–43, 2003. © 2003 Wiley-Liss, Inc.


Radial glia represent a remarkable class of cells that are present transiently in most brain regions during the periods when neurons are generated. Radial glia are mitotically active cells, with each nucleus located in the ventricular zone (VZ) and undergoing interkinetic nuclear migration with cell cycle progression (Misson et al., 1988). In the mammalian telencephalon, radial glia exhibit a characteristic bipolar morphology with one endfoot on the ventricular surface and a radial process that extends to the pia. Deriving from neuroepithelial (NE) cells that surround the neural tube, radial glial cells in most regions of the mammalian brain disappear or transform into astrocytes when neuronal generation and migration are complete (Schmechel and Rakic, 1979a; Voigt, 1989; Misson et al., 1991). Exceptions include some radial glia that persist in the adult CNS, such as Bergmann glia, found in the developing and adult cerebellum, Müller glia present in the retina, and radial glia that persist in the dentate gyrus of the adult hippocampus (Eckenhoff and Rakic, 1984; Rickmann et al., 1987; Cameron et al., 1993).

All neurons and glial cells of the vertebrate CNS originally derive from the NE cells of the neural plate. Around the time that neurogenesis begins, NE cells appear to transform into radial glia. Recent evidence suggests that the radial glia, in turn, begin to generate postmitotic neurons (Malatesta et al., 2000; Noctor et al., 2001), and possibly additional precursor cells. Interestingly, radial glia share their bipolar morphology with the NE cells. If radial glia truly represent a cell type distinct from the NE cell, the transition of NE to radial glia should be accompanied by the expression of distinct molecular markers. It is therefore important to examine specifically how radial glial cells and NE cells may differ. The main differences between these two types of radial cells lie in the “glial” properties of radial glia, which are not apparent in NE cells. For example, ultrastructural features of radial glia include the presence of 24-nm microtubules and 9-nm intermediate filaments within the radial fiber and cytoplasmic glycogen granules, which are particularly prominent in the subpial endfeet (Choi and Lapham, 1978; Bruckner and Biesold, 1981; Gadisseux and Evrard, 1985). These features led to the early classification of these cells as a form of specialized glia. Indeed, glycogen granules, a hallmark of astrocytes in the adult, appear around the onset of neurogenesis in the telencephalon (Gadisseux and Evrard, 1985), coincident with the differentiation of further astroglial characteristics. The expression of a number of molecules characteristic of astrocytes in the adult CNS, including glial fibrillary acidic protein (GFAP), the astrocyte-specific glutamate transporter (GLAST), brain lipid-binding protein (BLBP), and Tenascin C (TN-C) (for review, see Campbell and Götz, 2002), also appear around the onset of neurogenesis, at embryonic day (E) 13 in the mouse telencephalon, in a defined subset of precursor cells located within the VZ (Hartfuss et al., 2001; Heins et al., 2002; Noctor et al., 2002). These molecular similarities with astrocytes distinguish radial glia from NE cells, the earliest precursors in the CNS present before neurogenesis. However, radial glial cells also share several molecular characteristics with early NE cells, such as nestin and the antigens recognized by the monoclonal antibodies RC1 and RC2, of which the latter most likely recognizes a posttranslational modification of nestin (Chanas-Sacre et al., 2000, and references therein). Nestin and RC1-/RC2-immunoreactivity are already present at E9 in the mouse telencephalon, i.e., in NE cells, about 3–4 days before the onset of the astroglial markers GLAST and BLBP (Hartfuss et al., 2001, and references therein). The GLAST- and BLBP-positive radial glial cells continue to express nestin throughout their postnatal transformation into astrocytes, when nestin expression is down-regulated (Hartfuss et al., 2001). Taken together, a number of astroglial properties that develop around the onset of neurogenesis appear to characterize radial glial identity and distinguish radial glial cells from their NE cell precursors.

Importantly, the differentiation of radial glia, indicated by the expression of astroglial traits, is accompanied by further cell biological changes, such as the transformation of tight junctional complexes into adherens junctions and the upregulation of several adhesion and extracellular matrix molecules, including R-cadherin and TN-C (Aaku-Saraste et al., 1997; Stoykova et al., 1997). These molecular changes might be important for the regulation of neuronal migration, a process that also begins at the onset of neurogenesis (Morest, 1970; Rakic, 1972; Hatten, 1999).


Recently, several genes encoding molecules involved in signaling cascades have been identified that are transiently expressed by neurogenic precursor cells, and therefore might function as early markers of neurogenic radial glial cells. The Minibrain gene (Mnb) encodes a family of protein kinases and is involved in neurogenesis in Drosophila (Tejedor et al., 1995). The Mnb chick orthologue was recently cloned and found to be expressed in the NE of the neural tube in both mouse and chick embryos (Hammerle et al., 2002). Mnb is expressed in all dorsoventral levels of the neural tube, but only transiently between G1 and S-phase during a single cell cycle. Moreover, Mnb mRNA is asymmetrically localized and inherited by only one daughter cell during the Mnb-expressing cell division. Expression moves in a rostrocaudal wave that parallels the onset of neurogenic divisions, but Mnb-expressing cells are sparsely distributed within any given region, suggesting that some mitotic NE cells may not express Mnb, and therefore that expression may not occur in all cell lineages. Mnb expression precedes and overlaps with the expression of Tis21, an antiproliferative gene that is also expressed at the onset of neurogenic divisions and which may also identify neurogenic radial glia. Like Mnb, Tis 21 mRNA is expressed in the mouse VZ at the onset of neurogenesis in a subpopulation of NE cells that has been suggested to generate neurons at their next division (Iacopetti et al., 1999). It has been proposed that Mnb defines a transition step between proliferating and neurogenic divisions of NE cells, although it remains unknown whether Mnb and Tis21 are expressed at the first neurogenic division, or during an earlier division. Furthermore, it is unclear what the fate of the non-Mnb expressing daughter cell might be. Assuming that radial glial cells become distinct from NE cells around the beginning of neurogenesis, molecules such as Mnb and Tis21, expressed by some NE cells just before neurogenesis, may in fact be early markers of some neurogenic radial glia.

Another candidate marker of neurogenic radial glia may be Ephrin B1. Ephrins are receptor tyrosine kinases that are important mediators of cell-cell communication. Ephrins and their ligands are both membrane-bound and, when activated at points of cell-cell contact, downstream signals are transduced bidirectionally by the class B ephrin receptors. Ephrin B1 is expressed in NE cells beginning at E9.5–10.5 coincident with the onset of neurogenesis and thereafter in all cortical NE cells until expression is downregulated at E14.5–P0 (Stuckmann et al., 2001). Interestingly, Ephrin B1 is only expressed in the anterior and lateral telencephalon, not in the ventral telencephalon (ganglionic eminences), consistent with the presence of neurogenic radial glia only in the former, but not the latter region (Malatesta et al, 2003; and see below). Despite this heterogeneity along the neuroaxis, within the embryonic cortex ephrin B1 is expressed in most, if not all, VZ cells and radial glia. Furthermore, consistent with an expression pattern in neuronal precursor cells, ephrin B1 is colocalized with nestin in radial glia, and expression persists throughout cortical neurogenesis but is subsequently downregulated. The first cells positive for ephrin B1 coexpress Tis21, consistent with their presumed pattern of expression in neuron-generating VZ cells. However, while Tis21 is only expressed in a subset of VZ cells, most if not all VZ cells at E10.5–12.5 are ephrin B1-positive. Because there is a radial gradient of ephrin B1 expression, and ephrin acts as a repulsive signal in cell migration in other systems, it has been proposed that ephrin may be involved in neuronal migration or differentiation (Stuckmann et al., 2001; Zhou et al., 2001).

Like ephrin B1, extra-large G protein α-subunit (XLαs) is another signaling molecule that is expressed in a subset of neurogenic NE cells. XLαs is a splice variant of the α-subunit of the heterotrimeric G protein, Gs (Kehlenbach et al., 1994), which serves to bind GTP and activate adenylyl cyclase (Pasolli et al., 2000). XLαs is expressed in mitotic NE cells of the neural tube beginning at the onset of neurogenesis (Pasolli and Huttner, 2001); however, only a subset of mitotic VZ cells are XLαs-positive. Similarly, XLαs was found in some, but not all, young neurons. Expression in mitotic NE cells and young neurons, albeit in only a subset of these cells, suggests a possible role in neurogenesis and neuronal differentiation. However, mutations rendering the XLαs gene defective have not been associated with any severe phenotype, suggesting that XLαs function is either nonessential or compensated for during development. In addition, shortly after the onset of neurogenesis, cortical radial glia express functional receptors for the neurotransmitters γ-aminobutyric acid (GABA) and glutamate (LoTurco and Kriegstein, 1991; LoTurco et al., 1995), activation of which appears to influence cortical proliferation (LoTurco et al., 1995; Haydar et al., 2000). In summary, Mnb, Tis21, ephrin B1, XLαs, and other molecules whose expression is timed to the onset of neurogenesis, may therefore identify an early step in the transformation of some NE cells to neuronal precursor cells, and therefore may represent early markers of neurogenic radial glia.


Before and after closure of the neural tube, NE cell divisions are thought to be primarily symmetric, serving to expand the NE population. Subsequently, at the time of neurogenesis, particularly in forebrain regions, many divisions are thought to become asymmetric, each leading to both renewal of a precursor cell and the generation of a daughter neuron. Radial glial cells are present at this time, most likely participating in neurogenic divisions. Lateral cues for spindle orientation and thus symmetric divisions may be provided by tight junctions, or adherens junctions, or associated cytoskeletal components, as suggested by Huttner and Brand (1997). Indeed, several proteins associated with tight junctions, such as occludin and epithelial (E)-cadherin, are expressed by NE cells before neural tube closure (Redies, 1995; Takeichi, 1995; Aaku-Saraste et al., 1996), when most NE cells are undergoing symmetrical cell divisions. These proteins are downregulated when NE cells lose their tight junctions before neurogenesis (at around E9 in the mouse) (Aaku-Saraste et al., 1996). Based on the hypothesis that cleavage plane orientation may be involved in determining symmetrical divisions (Chenn and McConnell, 1995), changes in tight junction complexes may influence the symmetry of cell division since they could act to anchor mitotic spindles, thereby influencing spindle orientation and cleavage plane (Huttner and Brand, 1997). Support for the concept that tight junctions help to regulate NE cell divisions comes from recent experiments with mice engineered to overexpress β-catenin, an integral component of adherens junctions (Chenn and Walsh, 2002). This manipulation had the effect of promoting symmetric NE divisions and resulted in a significant expansion of the neuronal precursor pool, ultimately expanding overall cortical surface area. Other membrane proteins, such as zonula occludens 1 (ZO-1), are upregulated around the time of neural tube closure and the onset of neurogenic divisions (Redies, 1995; Takeichi, 1995; Aaku-Saraste et al., 1996). In addition, cadherins play a role in linking membrane receptors and gene expression and are also expressed at tight junctions between NE cells. Interestingly, expression of (E)-cadherin, a protein associated with formation of epithelia, is downregulated around the time of neurogenesis, while (N)-cadherin, a protein associated with differentiation of epithelia and neurons, is upregulated (Larue et al., 1996). These changes in gene expression are temporally related to the onset of neurogenesis and may represent steps along the pathway of transformation of NE cells into neurogenic radial glia.


Radial glia generate most of the pyramidal neurons in the cortex (Tamamaki et al., 2001; Noctor et al., 2002; Malatesta et al., 2003). However, radial glia can be subtyped in terms of molecular expression and are probably heterogeneous in terms of the cell types they generate. Moreover, the output of at least some individual radial glia may be determined very early and may already be established around the time radial glia arise from NE cells. At E10 in the mouse, the primordial cerebral cortex consists of a single layer of proliferating NE cells. Clonal studies indicate that approximately 15% of these are multipotent stem cells that generate both neurons and glia, while the remainder are restricted neuroblasts that generate only neurons (Qian et al., 1998, 2000). If most radial glia generate neurons and then transform into astrocytes, one might expect that most postnatal clones generated by labeled precursors would consist of mixed neuronal and glial cells. But studies based on retroviral labeling have found that relatively few clones appear to be mixed neuronal and glial clones. McCarthy et al. (2001) infected individual precursors in the telencephalon at E9.5 and examined 3 weeks postnatally. These investigators found that 34% of clones contained neurons only, 47% glia only, and 18% contained both neurons and glia (McCarthy et al., 2001). Similarly, Luskin et al. (1988) infected precursors in the embryonic telencephalon at early to mid-neurogenesis and found that most clones were homogeneous, including most neuron-only clones (39/47), and some glia-only clones (7/47), but very few mixed astrocyte and neuron clones (1/47). Similar results were reported in other studies using retroviral lineage markers (Price and Thurlow, 1988; Parnavelas et al., 1991; Grove et al., 1993; Luskin et al., 1993). It is possible, however, that migration of clonally related astrocytes and neurons away from each other may have confounded identification of mixed clones. Since it has been shown that most precursors in the ventricular zone are radial glial cells (Hartfuss et al., 2001; Noctor et al., 2002), these lineage data apply directly to radial glia and suggest that they consist of different sublineages, some of which may be specialized to generate neurons, and others to generate glial cells.

While pyramidal neurons, oligodendrocytes, and astrocytes in the cortex may derive from radial glia, this does not mean that all radial glia generate pyramidal neurons, or astrocytes, or both. There may be subsets of radial glia restricted in terms of the cell types that they can generate. For example, some radial glia, perhaps the majority in some regions, may generate neurons throughout neurogenesis and then deplete themselves possibly by symmetrical terminal division. These clones would be all neuronal. Indeed, in vitro analysis of isolated radial glial cells showed that during neurogenesis, most radial glial cells are restricted to generate only neurons, while just a minor proportion of radial glia generated both neurons and glia (Malatesta et al., 2000, 2003). Radial glia are therefore not only heterogeneous in the expression characteristics described above, but in their lineage potential as well. Other radial glia may not be mitotically active during the period of neurogenesis and may simply transform or generate astrocytes postnatally, thereby generating glia-only clones. Evidence in favor of this concept comes from the finding that not all radial glia in primates are actively dividing during periods of neurogenesis (Schmechel and Rakic, 1979b), in contrast to rodents, in which no quiescent radial glia could be detected (Hartfuss et al., 2001). Interestingly, overexpression of activated Notch appears to slow down cell division as well as trap radial glia in a gliogenic fate (Gaiano et al., 2000). Such non-neurogenic radial glia might nonetheless serve as migrational guides for radially migrating neurons during the embryonic period. Evidence from clonal analysis at different stages of neurogenesis in the mouse suggests that symmetrical progenitor divisions, asymmetrical neurogenic divisions, and symmetrical terminal divisions occur at all stages of cortical neurogenesis, but that the proportions change over time (Cai et al., 2002). Thus, given the correlation of birth order with laminar fate, individual neuronal precursor cells could generate neurons either destined to occupy a single layer (horizontally dispersed) through symmetrical divisions, or that migrate to different layers (vertically dispersed) by asymmetrical divisions. Retroviral clones have been observed in both radial and horizontal configurations in primate cortex (Kornack and Rakic, 1995), suggesting that at least during a restricted period of neurogenesis, some precursors may be committed to divide either symmetrically or asymmetrically.


Notably, however, radial glial cells differ not only within a brain region, such as described above for the cerebral cortex, but also between regions of the developing CNS. This regional heterogeneity of radial glial cells is apparent at the molecular level, and translates into pronounced differences in cell fate. For example, radial glial cells of the lateral ganglionic eminence (LGE), the anlage of the striatum (Olsson et al., 1998), contain the retinol-binding protein 1 (RBP-1), which is not detectable in the MGE or cortex (Toresson et al., 1999). Indeed, retinoid signaling seems important for the development of neurons in the striatum, but not in adjacent regions (Toresson et al., 1999; see Campbell, this issue). Interestingly, ectopic expression of the transcription factor Gsh2 which is normally restricted to the ganglionic eminence, in the cortex seems to result in ectopic expression of RBP-1 (Corbin et al., 2000). Unfortunately, the effects of ectopic RBP-1 on neuronal differentiation in the cortex have not yet been examined. A region-specific influence on neuronal differentiation, however, has most clearly been demonstrated for glial cells from the developing midbrain (see Hall and Arenas, this issue). Nurr-1 transduction of CNS precursors does not induce a dopaminergic phenotype; however, when the transduced cells are cultured with “astrocytes” derived from embryonic midbrain, a notable increase in dopaminergic neurons was observed (Wagner et al., 1999). Because these astrocytes are derived from embryonic midbrain, they are most likely radial glia releasing an as yet unknown factor required in cooperation with Nurr-1 to induce dopaminergic neuron differentiation. The release of this factor is obviously region-specific, since radial glia isolated from other CNS regions do not cooperate with Nurr-1 in dopaminergic neuron differentiation (see Hall and Arenas, this issue). Another crucial factor instructing the differentiation of specific cell fates, particularly in ventral regions of the developing CNS, is sonic hedgehog (SHH), which is contained specifically in radial glial cells in the zona limitans and the hypothalamic region (Campbell and Götz, 2002). Because SHH has been shown to instruct particular neuronal subtypes as well as oligodendrocyte fate (for recent reviews, see Rallu et al., 2002; Campbell, 2003), its release by radial glia could instruct specific neuronal and glial fate. Taken together, these observations raise the suggestion that radial glial cells could be involved in the local region-specific differentiation of particular cell and neuronal subtypes by supplying differentiation factors in a local context.

Radial glial cells also express distinct transcription factors, depending on their location. In the dorsal telencephalon, for example, radial glial cells contain the transcription factor Pax6 (Götz et al., 1998), while radial glia in the LGE contain Gsh2, and those in the medial ganglionic eminence (MGE) contain Olig2 (Malatesta et al., 2003). All these transcription factors have been implicated as potent cues in patterning the developing neural tube (Stoykova et al., 1996; Corbin et al., 2000; Stoykova et al., 2000; Toresson et al., 2000; Toresson and Campbell, 2001; Yun et al., 2001). These data suggest that radial glial cells are not only exquisitely suited to pattern the CNS because of their anchored position at the apical and basal surface (Götz, 1995; Campbell and Götz, 2002), but are indeed functionally involved in patterning, since they are the cells containing the cell-autonomous and non-cell-autonomous cues instructing regionalization of the developing CNS.


The final read-out and implementation of brain regionalization is the differentiation of distinct types of neurons in specific brain regions. Accordingly, the transcription factors involved in patterning simultaneously affect cell fate decisions. For example, the loss of Pax6 reduces the number of neurogenic radial glia cells in the cerebral cortex, resulting in a prominent reduction of neurons in a region-specific manner (Heins et al., 2002). Notably, Pax6-deficient radial glia of the cortex assume several properties characteristic of radial glia in the ventral telencephalon (Götz et al., 1998), although they do not acquire RBP-1 expression (Toresson et al., 2000). Intriguingly, Pax6-mutant cortical radial glia also seem to acquire the fate of their counterparts in the ventral telencephalon, predominantly generating glial cells and only few neurons (Malatesta et al., 2003). Recent fate mapping of the entire progeny of radial glial cells demonstrated pronounced differences in the ventral and dorsal telencephalon: while cortical radial glia generate the vast majority of neurons (more than 90%) in the cerebral cortex, radial glia in the ventral telencephalon generate very few neurons (Malatesta et al., 2003) (Fig. 1). Most surprisingly, the few neurons generated by radial glia in the ventral telencephalon are specific types of interneurons, primarily those migrating to the olfactory bulb (Fig. 1). This observation fits very well with the fact that precursor cells in the LGE appear to give rise to olfactory bulb interneurons (Wichterle et al., 2001). In fact, it has been proposed that these cells arise in the dorsalmost portion of the LGE, the same region where Gsh2-positive radial glia are concentrated and the specific population of Er81-positive, Dlx5/6-positive, cells migrating to the olfactory bulb appears to arise (Stenman et al., 2003; Malatesta et al., 2003). The Olig2-positive radial glial cells located in the MGE are the most likely source of oligodendrocytes derived from radial glial cells (Malatesta et al., 2003). Indeed, the small proportion of GABAergic interneurons in the cerebral cortex that were found to derive from radial glial cells in this fate mapping study might derive from a bi- or multipotent precursor located in the MGE (He et al., 2001; Yang et al., 2003; Fig. 1). Yung et al. (2003) recently suggested that the local influence of SHH induces the expression of the transcription factor Olig2 and thereby instructs precursors to generate GABAergic neurons and oligodendrocytes. These data are further in line with the suggestion that the Olig2-positive cells in the MGE generate GABAergic neurons and oligodendrocytes. Notably, however, Mash1 seems to be additionally required for the lineage of GABAergic neurons (Yang et al., 2003), but few radial glial cells in the GE contain Mash1 (Hartfuss et al., 2001; Malatesta et al., 2003). Taken together, these data would support a view that most MGE radial glial cells generate oligodendrocytes. A small subpopulation of Olig2-positive radial glia that also contain Mash1 could still be bipotent, however, generating some GABAergic neurons migrating into the cortex. It will be important to determine whether these are specific subtypes of GABAergic interneurons. These data lend further support to the remarkable ability of the developing neural tube to generate distinct neuronal subtypes at specific dorsoventral positions, as depicted in Figure 1. Radial glia appear to be important in mediating the direct translation of patterning into neuronal fates at a number of levels, thus contributing significantly to neuronal diversity.

Figure 1.

Micrograph depicting a section of the embryonic telencephalon at mid-neurogenesis with a schematic drawing of different radial glial populations. Note that this section is rostral, and the subdivision of the ventral telencephalon, the ganglionic eminence (GE), into lateral and medial parts is not visible. Olig2-positive radial glia are, however, located primarily in the medial ganglionic eminence (MGE), and Gsh2-positive radial glia are enriched in the dorsalmost part of the lateral ganglionic eminence (LGE). The remarkable region-specific fate specification of radial glial cells that generate different progeny at different positions along the dorsoventral axis of the telencephalon is schematized.

These data also show that radial glial cells located in different brain regions differ not only in the neuronal subtypes that they generate, but also in the cell classes they produce. Radial glial cells contribute to different lineages (i.e., oligodendrocytes, neurons, and astrocytes), but do so in apparently restricted lineages from distinct regions of the developing CNS. Several lines of evidence suggest that these fate differences are mediated by cell-autonomous differences. For example, in vitro experiments that combined different cell types showed that the environment was not sufficient to alter radial glia fate. Radial glia from the ventral telencephalon did not generate neurons when cultured within a cortical environment, nor did cortical radial glia from the cortex or the spinal cord lose their neurogenic properties when cultured in an environment from the ventral telencephalon (Malatesta et al., 2003). While these in vitro experiments could still be explained by the lack of the relevant extrinsic cues in vitro, the notion of intrinsic fate differences of radial glial cells from different regions is further substantiated by the presence of distinct transcription factors implicated in cell-autonomous regulation of cell fate. Indeed, the loss of Olig2 in (radial glial) cells of the ventral telencephalon results in a lack of oligodendrocytes in the forebrain (Lu et al., 2002; Zhou and Anderson, 2002), the loss of Gsh2 in radial glia of the LGE leads to defects in olfactory bulb and striatal neurogenesis (Corbin et al., 2000; Toresson and Campbell, 2001; Yun et al., 2001), and the loss of Pax6 in cortical radial glia results in a reduction of cortical neurons (Heins et al., 2002). These results therefore underline the role of radial glial cells as the pivotal cell type in mediating the region-specific differences of the developing CNS, not only by instructing cell differentiation in a region-specific manner by the release of factors, but also by direct implementation of cell-autonomous fate differences in their role as precursor cells.