A requirement for Notch in the genesis of a subset of glial cells in the Drosophila embryonic central nervous system which arise through asymmetric divisions

Drosophila is an excellent species to use to determine how the vast cellular diversity and complexity is created in the central nervous system (CNS). The generation of the embryonic CNS is a lineage-based process in which neural progenitors, called neuroblasts (NBs), give rise to largely invariant lineages of neural/glial cells. NBs divide in an asymmetric manner to bud off a set of ganglion mother cells (GMC), which in turn divide once to produce two postmitotic daughters. Cell lineage analysis techniques have been used to analyse most of the embryonic NB lineages at the histological level. These studies have elucidated the cellular composition and the specific nature of each NB lineage, as well as the morphologies of identified cells (Bossing et al., 1996; Schmidt et al., 1997; Schmid et al., 1999).

Neural progenitor divisions are asymmetric. Both intrinsic cues, e.g. basally localised and asymmetrically segregated cell fate determinants like Numb and Prospero (Pros) (Rhyu et al., 1994; Hirata et al., 1995; Knoblich et al., 1995; Spana et al., 1995), as well as extrinsic cues (Guo et al., 1996; Spana and Doe, 1996) (mediated by Notch and Delta and cell-cell interaction), are involved in the specification of daughter cell fates. NBs and some GMCs divide with their mitotic spindles oriented along the apical-basal axis, enabling the preferential segregation of basally localised cell fate determinants to just one daughter (Kraut et al., 1996; Buescher et al., 1998; Schuldt et al., 1998). Recent studies have shown that this polarity of neural progenitors is mediated by several molecules, Inscuteable (Insc) (Kraut and Campos-Ortega, 1996; Kraut et al., 1996), Bazooka (Baz) (Schober et al., 1999; Wodarz et al., 1999), and Partner of Inscuteable (Pins) (Raps – FlyBase; Yu et al., 2000; Parmentier et al., 2000; Schaefer et al., 2000) which form a complex that localises to the apical cortex. Correct asymmetric Numb distribution has been shown to be crucial for the specification of distinct fates of sibling neurones in several contexts (e.g. Guo et al., 1996; Spana et al., 1995; Buescher et al., 1998). Numb acts to downregulate Notch signalling, thereby providing a direct link between intrinsic and extrinsic mechanisms for the specification of distinct fates of sibling neurones. However, this model is based upon only a handful of known CNS sibling neurones, which represent early born cells from their respective NB lineages. Therefore, it is still an enigma how later born cells, including glial cells, are specified in the context of NB/neuroglioblast (NGB) lineage development.

In terms of cell numbers, glia comprise ∼10% of the cells in the embryonic nervous system (Ito et al., 1995). Glial cells can derive from the mesectoderm as well as from the neuroectoderm. A common characteristic of all neuroectodermally derived glia is the strict requirement for the gene, glial cells missing (gcm), which serves as a binary switch between neuronal and glial cell fates (Hosoya et al., 1995; Jones et al., 1995; Vincent et al., 1996). However, mesectodermally derived glia do not require gcm, suggesting a different mechanism for their development. Moreover, different types of glial progenitors derived from the neuroectoderm have been identified. Glioblasts (GBs) give rise to only glial cells, whereas NGBs produce mixed glial/neuronal lineages. For at least two NGB lineages, the thoracic NGB6-4T and NGB5-6, it has been demonstrated that an early bifurcation of the glial versus the neuronal sublineages takes place during the first division of the parental NB. This process involves the asymmetric expression of gcm in the glial sublineage (Akiyama-Oda et al., 1999; Bernardoni et al., 1999). However, it is unclear whether this mechanism is generally applicable for other NGB lineages.

In order to better understand how a complete lineage of a specific NGB with all its progeny, including its glial cells, might be created we chose NB1-1 for a detailed analysis. NB1-1 has been extensively used for cell fate specification studies and a sound basis of information about this NB lineage is available (Udolph et al., 1993; Broadus et al., 1995; Skeath and Doe, 1998). NB1-1 is a NB that develops differential lineages in the thoracic versus the abdominal segments (Udolph et al., 1993; Prokop and Technau, 1994; Bossing et al., 1996). In our study, we focused on the abdominal NB1-1A because only these abdominal NB1-1 lineages contain glia. In addition to the aCC/pCC sibling neurones which are the progeny of the first GMC produced from this lineage, NB1-1A generates 2 to 3 glial cells and 4 to 5 clustered interneurones (cN), yielding a total of 9 to 10 cells. The three glial cells belong to the group of subperineurial glia (SPG) that lie at the periphery of the nerve cord and enwrap the entire ventral nervous system (Ito et al., 1995). Two of the glia, the A- and B-SPGs (Klämbt and Goodman, 1991), can be found in dorsal positions, with a third glia, the LV-SPG, located at ventral positions of the nerve cord. All SPGs, including the A- and B-SPG and LV-SPG of NB1-1A, are specifically labelled by two enhancer trap lines, M84 and P101 (Klämbt and Goodman, 1991).

In this study, we used available mutants, genetic markers and clonal analyses to investigate the development and sibling cell fate decisions of later components of the NB1-1A lineage. We present data to indicate that the glial cells within the NB1-1A lineage are direct siblings to neurones and each glia/neurone pair arise through asymmetric divisions. We provide evidence which indicate that Notch is required to resolve distinct neuronal/glia sibling fates. In this context, Notch acts upstream of gcm to specify SPG fate. Our findings reveal a novel mechanism by which glial cells can be generated in the CNS.

The following fly stocks were used: N^55e11, N^ts1, pros¹⁷ (from Bloomington Stock Center); mam^GA345, nb⁷⁹⁶ (from G. Tear, London); M84 (2^nd chromosome), P101 (3^rd chromosome), M84/P101 (from O. Vef); pnt-lacZ (from C. Klämbt); gcm^N7.4, gcm-lacZ^rA87 (from A. Giangrande); UAS-N^μ242 (from Y-N. Jan); and 1407-Gal4 (from J. Urban). 1407-Gal4 is a neural specific driver and we know it is expressed later than sca-Gal4. Unlike sca-Gal4, in combination with uas-N-intra, 1407-Gal4 does not drastically affect neuroblast delamination.

The transplantation procedure was performed as described previously (Prokop and Technau, 1993). The heterogenetic transplantations were generally performed as follows: cells were removed from mutant donors at 0-20% ventrodorsal diameter. Single cells were subsequently implanted into an equivalent position in wild-type host embryos. Homozygous mutant donor embryos were identified by staining each donor separately for the presence of a Blue-balancer chromosome (see also Udolph et al., 1998).

Embryos, both N^ts1 donors and wild-type hosts, were kept at the permissive temperature of 20°C until 1-2 hours after transplantation when the host embryos with the implanted N^ts1 cells were subjected to the non-permissive temperature of 29°C until late stages of embryogenesis. This procedure assured that Notch was selectively removed only during its requirement for sibling cell fate specification and earlier functions during lateral inhibition were largely unaffected. Subsequently embryos were subjected to the standard protocol after transplantation (Prokop and Technau, 1993).

An activated form of Notch (UAS-N^μ242, a homozygous viable insertion on the second chromosome) was crossed to the sca-Gal4 driver, which is also homozygous viable. All F¹ progeny possessing one copy each of both the driver and the transgene were used as donors. Cells were taken from these donors and implanted into wild-type hosts according to standard procedures.

Females carrying the Actin5c promoter/FRT cassette/taulacZ construct (Buenzow and Holmgren, 1995) were crossed to males carrying a HS-FLP construct. From this cross eggs were collected on standard medium for 3 hours on 25°C, embryos were aged for another 4 hours at 25°C. The sequence separating the constitutive actin5c promoter from the taulacZ-coding sequence is flanked by FRT sites (/) and contains a transcriptional termination signal. When this sequence is excised from a cell after activation of FLP, taulacZ is expressed and marks the lineage derived from this cell. After ageing, embryos were dechorionated by bleach and incubated at 34°C in a water bath for 20 minutes. Heat-shocked embryos were raised overnight at 18°C, and were subsequently subjected to a standard antibody staining protocol using an anti-β-gal antibody. After staining, embryos were examined and two cell clones were analysed. Our conditions generated on average one clone every two hemisegments.

Antibody staining was as described previously (Schmidt-Ott and Technau, 1992). Following primary antibodies were used: anti-Even skipped (Frasch et al., 1986); anti-Repo (Halter et al., 1995); anti-β-gal (Promega). Secondary antibodies used throughout this study were coupled either to FITC, to Cy3 or to HRP (Jackson).

The expression patterns of the enhancer-trap lines M84 and P101 (Klämbt and Goodman, 1991) are indistinguishable in abdominal segments: both are expressed in the SPGs including the A- and B-SPG and LV-SPG cells produced from NB1-1A. To investigate the relationship between the three glial cells derived from NB1-1A, we stained M84/P101 (a stock double homozygous for both the M84 and P101 insertions, obtained from O. Vef, Mainz) embryos with anti-β-gal. In stage 12/13 embryos, a single M84/P101⁺ cell appeared in dorsal positions of each hemi-neuromere (Fig. 1A). Embryos double labelled with anti-Even-skipped (anti-Eve) indicated that this cell is located posteriorly to the Eve⁺ cluster of NB1-1 (aCC/pCC) and NB7-1 progeny (CQ-neurones; data not shown). Slightly later a second M84/P101⁺ cell appeared anterior to the first cell (Fig. 1B). According to the positioning within the developing nerve cord, the posterior cell represents the A-SPG, whereas the anterior cell is the B-SPG. A third glial cell, representing the LV-SPG (Ito et al., 1995), could be detected in ventral positions only after the A- and B-SPG were already present, suggesting that this cell is the last born glia within the lineage (Fig. 1C). The temporal order of the M84/P101 expression pattern is summarised schematically in Fig. 1D.

As a first step towards elucidating the origin of the glial cells of the NB1-1A lineage, we examined the effects of loss of function mutants in several genes, Notch, mastermind (mam) and numb, which are known to affect the resolution of distinct sibling cell fates, for their effect on the development of A-, B- and LV-SPGs. Embryos hemizygous/homozygous for a conditional Notch allele, N^ts1, and also carrying one copy each of M84 and P101 (N^ts1/M84/P101) were subjected to the non-permissive temperature of 29°C after 6 hours of development. This regime allows Notch to function during the singling out of NBs and removes Notch during the crucial period when it is required for sibling cell fate resolution. Double staining with anti-Eve and anti-β-gal was performed (Fig. 2). As expected, in most hemisegments, N^ts1/M84/P101 embryos duplicated the RP2 neurone at the expense of its sibling cell. Moreover, in 96% of the hemisegments (n=144) M84/P101⁺ cells could not be found in typical dorsal or ventral positions (Fig. 2B). We conclude that Notch function is required for the specification of the M84/P101⁺ A-, B- and LV-SPGs. But we cannot rule out that these SPGs derived from NB1-1A are still physically present and the removal of Notch only downregulated M84/P101 expression. However, transplantation experiments with Notch mutant cells revealed that cells with glial morphology could not be detected in the mutant NB1-1A lineages (see below). In wild-type embryos, M84/P101 is expressed in about eight SPGs per hemisegment, including the A- and B-SPGs and the LV-SPG (Fig. 2A; Udolph et al., 1993). Removing Notch function results in the near complete abolishment of all M84/P101 expression, indicating a more general function for Notch in SPG specification (Fig. 2B).

These findings prompted us to investigate whether Notch is required in the specification of CNS glia in general. We performed experiments in which glia-specific markers, either enhancer trap lines (gcm-lacZ^rA87 and pnt-lacZ) or an antibody (anti-Repo; Halter et al., 1995) were used. gcm-lacZ^rA87 is an insertion into the gcm gene and expresses β-gal in the pattern of gcm. pointed (pnt) is specifically expressed in glial cells (Klämbt, 1993; Klaes et al., 1994) and has been reported to act downstream of gcm (Giesen et al., 1997). N^ts1 embryos were shifted to the non-permissive temperature after 6 hours of development, and subsequently stained with anti-Eve and anti-β-gal at late stage 16; taking into account the missing SPGs, both enhancer traps were expressed in a pattern reminiscent of wild-type embryos (Fig. 3A-D). However, an increase in the number of β-gal-positive cells was observed, which could be in part due to a mild neurogenic phenotype caused by N^ts1. Furthermore, we found that the glial-specific protein Repo (Campell et al., 1994; Xiong et al., 1994; Halter et al., 1995) was widely expressed in the CNS of N^55e11 (an amorphic N allele) embryos (Fig. 3F). Thus, general glial specific markers like gcm, pnt and repo are expressed in embryos that lack or have strongly reduced Notch function, indicating that although Notch is required for the formation of SPGs, there is not a global requirement for Notch in the specification of all CNS glia.

We also tested another neurogenic gene, mastermind (mam), which has been linked to the Notch signalling pathway by its genetic interactions with Notch and its strikingly similar phenotype in early and late neurogenesis. It has been shown that mam acts downstream of Notch during sibling cell fate specification in the embryonic nervous system (Schuldt and Brand, 1999). The hypomorphic mam³⁴⁵ allele used in our study shows only a mild hypertrophy of the nervous system but clearly has an effect on sibling cell fate specification (Buescher et al., 1998). We observed a severe reduction (94%; n=56 hemisegments) of P101⁺ cells in mam³⁴⁵; P101 embryos similar to that seen with N^ts1/M84/P101 embryos (Fig. 2C).

These data suggest that both genes are strictly required for the specification of SPGs, most likely in a linear pathway.

However, it is unclear how Notch acts in the specification of the SPGs. We consider the possibility that SPG glial cells could arise from a series of asymmetric cell divisions, with Notch being required to specify the glial daughters of these divisions. Based on its function as a negative regulator of Notch signalling, the expected numb phenotype is opposite to Notch in terms of sibling cell fate transformation. We tested the P101 expression pattern in the background of a strong numb mutation (nb⁷⁹⁶) (Buescher et al., 1998; Skeath and Doe, 1998). In contrast to Notch and mam, we found additional P101⁺ cells in the vicinity of the aCC/pCC position. In most of the examined hemi-neuromeres, we detected up to four β-gal-positive cells in dorsal positions close to aCC/pCC (Fig. 2D; 67%; n=125 hemisegments). This is indicative of a duplication of the A- and B-SPGs. We also found additional P101⁺ cells with glial morphology in lateral and ventral positions of the nerve cord, presumably duplications of other SPGs (data not shown). These findings are consistent with our asymmetric cell division model for the genesis of the SPGs.

Our analyses of SPG formation using molecular markers in various mutant backgrounds do not provide any information pertaining to their relationship with other cells of the NB1-1A lineage. To overcome this limitation, we used the transplantation technique to create defined mutant NB clones in an otherwise wild-type background to enable the effects of a mutant gene to be assessed in the context of complete NB lineages. HRP-marked cells were taken from the neuroectoderm of hemi- or homozygous Notch mutants (N^55e11 or N^ts1), or from transgenic donor embryos misexpressing a constitutively active form of the Notch receptor fused to the yeast upstream activating sequence (UAS-N^μ242) driven by the scabrous Gal4 driver (sca-Gal4). These mutant donor cells were implanted into wild-type hosts (see Materials and Methods). After implantation, the embryos were raised until late stages of embryogenesis (stage 16 according to Campos-Ortega and Hartenstein, 1985). Resulting clones of NB1-1A lineages were screened and analysed.

Thirteen NB1-1A clones derived from N^55e11 mutant donors were obtained. In all 13 mutant clones, the pCC neurone was transformed into its sibling aCC, as revealed by the loss of the pCC projection and the duplication of the aCC motorprojection (Fig. 4A-D; Table 1; data not shown). In addition, the correct muscle DO1 was innervated by the duplicated aCC (data not shown). These results confirm and extend on previous findings based on molecular markers for aCC/pCC (Buescher et al., 1998; Skeath and Doe, 1998) and indicate that the transformed pCC exhibits the characteristic axon morphology of its sibling aCC. Furthermore, the later components of the Notch mutant NB1-1A lineages show additional abnormalities that have not been previously described; with the exception of one clone, the glial components of these lineages were totally absent (compare Fig. 4A,B with Fig. 4C,D) based on two criteria: first, we could not detect cells in the expected glial positions and second, there were no cells present that exhibited typical glial morphologies, i.e. extensive glial protrusions. In addition, we found that either the number of cluster neurones was increased above the wild-type number, or additional cells with neuronal-like morphologies appeared in close vicinity of the cluster located ventrally to the duplicated aCC neurones (Table 1).

Our clonal analysis demonstrates that the loss of Notch function results in the loss of SPGs and a concomitant gain of neuronal cells within the NB1-1A lineage, consistent with the notion that a sibling cell fate relationship exists between cluster neurones and glial cells in this lineage and that Notch is required for the asymmetric divisions that generate these postulated neurone/glia sibling pairs. One prediction of this model would be that ectopic Notch function achieved by the overexpression of a constitutively active form of Notch should increase the number of glial cells and decrease the number of cluster neurones within the NB1-1A lineage. To test this prediction, we transplanted cells from donor embryos, which were genotypically UAS-N^μ242/sca-Gal4. This combination of driver and transgene results in ectopic expression of a constitutively activated form of Notch, exclusively in the implanted cell and the resulting lineages. Seven such NB1-1A clones were obtained and as predicted, in all seven NB1-1A clones we found additional glial cells exceeding the number of SPGs that are normally found in wild-type clones and a concomitant decrease in the number of cluster neurones (Fig. 4E; Table 1). However, the locations of the additional glial cells were somewhat variable, occupying dorsal/lateral to ventral positions, possibly due to an effect on cell migration. Furthermore, in all these clones in which additional glial cells were found, we also detect a clear reduction in the number of cluster neurones (Table 1). Therefore, it seemed that additional glial cells were produced at the expense of cluster neurones. These data underline our hypothesis that glial cells and cluster neurones share a sibling relationship within the NB1-1A lineage.

However, these results do not rule out the possibility that (after its first division to produce GMC-1 and the aCC and pCC neurones) the second or a later division of NB1-1A results in a bifurcation, producing a glioblast and a neuroblast and that Notch might be require for this asymmetric division. In this scenario, increases or decreases in the number of glial cells in this lineage caused by the various mutants would also result in a reciprocal change in the number of neurones. If this were the case, however, one would expect that in any given (mutant) NB1-1A clone, glial cells would either be all present or all absent, depending on whether the glioblast is correctly specified. cells in these clones. This is in contrast to the N^55e11 NB1-1A clones, where essentially all SPGs cells of the NB1-1A lineage were absent. This might indicate that for the conditions we used, the N^ts1 cells retain some residual Notch function, and that the specific threshold of Notch activity required to resolve aCC/pCC fates might be greater than that required to specify SPG fate. More importantly, the fact that partial loss of SPGs appears to be the predominant effect in N^ts1 clones argues against the late bifurcation model. Taken together, our data favour a model in which a series of three GMCs produced from NB1-1A can each divide asymmetrically to produce a neurone and a glia, with Notch signalling required to specify the glia fate, and Numb and the absence of Notch signalling required for the neuronal fate.

To directly demonstrate that SPG-glia and neurone form sibling pairs, we used the FRT/FLP tau-bGal ‘flip-out’ cassettes method (Buenzow and Holmgren, 1995) (see Materials and Methods) to generate a series of two cell clones in the wild-type CNS. Sixty hemisegments were obtained containing scoreable two cell clones. In all six cases where the clone contained one SPG on the basis of position and morphology, its sibling had a neuronal morphology (rounded cell body) with clearly visible axonal projections into the neuropil (Fig. 4F). These results further substantiate that SPGs, in general, derive from an asymmetric cell division, suggesting that not only are the SPGs derived from NB1-1A but probably all SPG in the CNS share sibling relationship with neurones.

As shown above, development of the SPGs requires Notch signalling. It is also known that gcm acts as a master regulator in gliogenesis because it has been described as a binary switch between glia and neurones (Hosoya et al., 1995; Jones et al., 1995; Vincent et al., 1996). As expected, gcm appears to be required for the formation of SPGs. P101 marker gene expression was abolished in embryos homozygous for gcm^N7.4, an amorphic gcm allele (Fig. 5B). The similarity of gcm and Notch phenotypes suggests the possibility that both genes share a common pathway required for SPG specification.

We reasoned that if Notch acts downstream of gcm, then overexpression of N^μ242 in a gcm minus genetic background should result in additional SPGs; however, if Notch acts upstream of gcm, then in a gcm minus background the overexpression of N should not result in the production of additional P101⁺ cells. We misexpressed UAS-N^μ242 in a P101 genetic background by using the 1407-Gal4 driver (Jo Urban, Mainz; see Materials and Methods). 1407-Gal4 driven misexpression of UAS-N^μ242 leads to an increase above the number of normal wild-type P101⁺ SPG glial cells (data not shown), which confirms the UAS-N^μ242 transplantation results (see Table 1). Using the same combination of markers and transgenes (P101/1407-Gal4/UAS-N^μ242), but in a gcm minus background, we observed a complete loss of the P101 staining pattern within the nervous system (Fig. 5C; n=25 embryos). In contrast, the ectodermal component of the P101 expression pattern was retained. Thus, in a genetic background lacking gcm function, N^μ242 expression was not able to induce SPG development as indicated by the loss of marker gene expression.

To address this possibility, we transplanted donor cells taken from N^ts1 embryos. In this experiment, the transplantation procedure was performed under the permissive temperature (20°C) and host embryos were subject to the non-permissive temperature 1 hour after transplantation. Analysis of 13 such clones yielded results that were qualitatively similar to those of transplanting N^55e11 cells; we observed pCC to aCC transformation, loss of SPG and gain of neurones. However, the expressivity was incomplete for all of these cell fate transformations (Table 1). In the majority of the N^ts1 NB1-1A lineages, where pCC was transformed into aCC, loss of SPG was incomplete; we could detect between zero and three glial These findings are consistent with the notion that Notch functions upstream of gcm in the context of SPG development.

Two types of neuroectodermally derived glial progenitors in the embryonic nervous system of Drosophila have been described. Glioblasts (GB) generate only glial progeny, and NGBs produce both neurones and glial cells within the same lineage (Udolph et al., 1993; Bossing et al., 1996; Schmidt et al., 1997). A mechanism by which both neurones and glial cells can arise within the lineage of a thoracic NGB, NGB6-4T, has recently been described. NGB6-4T represents a thoracic-specific NGB in which glial and neuronal sublineages bifurcate from each other during the first division of the parental NGB (Akiyama-Oda et al., 1999; Bernardoni et al., 1999). Only one of the two daughters expresses Gcm, a master regulator of glial cell fate that is involved in regulating the expression of other glia-specific target genes (Hosoya et al., 1995; Jones et al., 1995; Vincent et al., 1996). In the NGB6-4T lineage, the asymmetric distribution of Gcm results in a cell that is specified as a GB within a neuroglioblast lineage. The GB will exclusively generate the glial components, whereas its sibling will exclusively give rise to the neuronal components of the lineage.

In this study, we report a novel mechanism by which glia can be generated during CNS development. Our data suggest relationship with neurones and does not involve the generation of a lineage internal GB at all. The asymmetric origin of glia described here provides a novel mechanistic framework of glial origin that might be used as a model system to gain further insights into the regulatory networks involved in glial cell fate specification in the CNS. We conclude that the mechanisms leading to glial cell fate specification are complex and that multiple developmental mechanisms lead to glial cell fate specification. These mechanisms will be likely to require that NB1-1A gives rise to a set of three glial cells through a A series of three GMC asymmetric divisions. Several lines of evidence support the notion that within a NGB lineage GMCs can produce both glial and neuronal cells. First, ^GB immunohistochemical analyses indicate that the A-, B- and LV-SPG arise at different times in development, and their non-simultaneous birth, in conjunction with the fact that their formation can be differentially affected by inactivation of Notch, suggests that they do not derive from a common precursor. Second, we showed that mutations in genes involved in specifying alternative sibling cell fates affect SPG development in a fashion that suggest these cells are siblings with non-glial components of the lineage. Third, transplantation experiments indicate that the 3 SPGs share sibling relationships with the neuronal components of the NB1-1A lineage, the cluster neurones (cN). Finally, analysis of two cell FLP-clones demonstrate that SPG glia and neurones share a direct sibling relationships.

Our proposed mechanism for the genesis of SPGs (Fig. 6) is fundamentally different from the ones described for GBs, e.g. anterior glioblast (Jacobs et al., 1989), and for NGBs, e.g. NGB6-4T and NGB5-6A (Akiyama-Oda et al., 1999; Bernardoni et al., 1999). The first division of NB1-1A produces a neurogenic GMC that gives rise to a pair of sibling neurones (aCC/pCC), and at the level of the first division of the parental NB no GB sublineage is bifurcated. In addition, the NB1-1A derived glial cells share a direct sibling different molecular machinery that involve distinct sets of genes or genetic hierarchies; for example, Notch being required specifically for the SPGs but apparently not for other types of glia. However, most of the glial cells (except midline glia) strictly require the gcm gene. Thus, there seems to be a common molecular basis that is context and cell specifically regulated during development.

It is unlikely that NB1-1A is the only lineage that generates glial cells in an asymmetric manner. Our studies using the SPG glial specific marker M84/P101 showed that Notch, mam and numb mutants affect the formation of all SPGs, not just those derived from NB1-1A. All SPGs derive from NGB because GBs do not give rise to SPGs (Schmidt et al., 1997). Lineage analysis techniques have revealed eight NGBs (NB1-1A, NB1-3, NB2-2T, NB2-5, NB5-6T, NB5-6A, NB7-4) that produce both glial and neuronal histotypes, most of them also generate SPGs, either exclusively or with other glial cell types (NB1-1A, NB2-2T, NB5-6T, NB5-6A, NB7-4). The similarity of their genealogy, and their common response to the loss and gain of Notch function, as well as analyses of two cell clones, indicate that SPG glial cells in general might derive from asymmetric cell divisions that generate one glial and one neuronal progeny.

Our study provides the first evidence that Notch is a crucial component in the specification of a subclass of CNS glial cells, the subperineurial glia (SPG). As revealed by the transplantation experiments, Notch is not only required for the expression of the M84/P101 marker, but loss of Notch function leads to the loss of the SPG cell fate per se. In the case of NB1-1A, our data indicate that the loss of glia is accompanied by the conversion of these cells into neurones. In contrast, Notch gain of function results in the overproduction of SPG-like cells at the expense of the cluster neurones. Hence, Notch signalling is sufficient for the specification of these cells. A second gene, gcm, is also crucially required for the specification of the SPGs; in a gcm mutant background M84/P101 expression is completely absent. If an activated form of the Notch protein is ectopically expressed in the embryonic CNS of animals lacking gcm function, SPGs do not form; in addition, M84/P101 expression cannot be detected in the CNS of these embryos. This indicates that gcm function is strictly required for Notch-mediated SPG specification, suggesting that Notch functions upstream of gcm. It is interesting to note that Notch also appears to play an instructive role in gliogenesis for mammalian neural crest stem cells in culture (Morrison et al., 2000; reviewed by Wang and Barres, 2000).

In this context, gcm acts as an effector of Notch signalling during sibling cell fate specification. Our results provide a framework of a signal transduction paradigm in which a receptor (Notch) is involved in receiving external signals by cell-cell interactions, leading to the activation of gcm, which ultimate results in the specification of the glial fate for a SPG-glia/neurone sibling pair. However, the components of this proposed pathway other than gcm remain to be defined.

With respect to the requirement for Notch, CNS glial cells can be classified into two distinct groups. First, there are the glial cells (SPGs) that arise through asymmetric cell divisions that depend on Notch signalling; the specification of a second group of glial cells is independent of Notch function and these glia are probably born in a non-asymmetric fashion. We also suggest the existence of at least two different classes of NGBs, referred to as type I and type II NGBs. Type I NGBs bifurcate a glial sublineage during their early development; whereas the type II NGBs give rise to glial cells by producing GMC(s) that then undergo asymmetric division to produce a Notch-dependent glial daughter and a neurone (Fig. 6).

Our results, in conjunction with what is already known about the origin of the aCC and pCC neurones, as well as the terminal lineage of NB1-1A, allow us to propose the following for the NB1-1A lineage: in total, NB1-1A gives rise to five pairs of sibling cells. The first pair comprises the aCC/pCC neurones resulting from GMC1-1a. In addition, three later born GMCs each give rise to one cluster neurone and one SPG-glia. A fifth GMC gives birth to two additional cluster neurones. The order of appearance and the position of the MP84/P101-labelled cells would suggest that, of the three glia-producing GMCs, the earliest born GMC gives rise to the A-SPG; the latest born GMC gives rise to the LV-SPG; and the GMC born in the middle gives rise to the B-SPG. We cannot place the time of birth of the fifth GMC in this lineage, which we postulate produces two cluster neurones. Clearly, refinement and proof for this proposal will require experiments that provide direct temporal information.

In summary, we provide evidence for a novel mechanism of glial cell fate specification where a subclass of glial cells, the SPG, derive from a series of GMCs, each of which divide asymmetrically to produce one neurone and one glia. Notch and gcm are both crucial for this event, and in this context, it is likely that Notch acts upstream of gcm.

We thank O. Vef, A. Giangrande, Y.N. Yan, M. Frasch, J, Urban, R. Holmgren, C. Klämbt, G. Tear and the Bloomington Stock Centre for flies and antibodies. This work was supported by the Institute of Molecular and Cell Biology (IMCB), Singapore.