Asymmetric Prospero localization is required to generate mixed neuronal/glial lineages in the Drosophila CNS

Neurons and glia have dramatically different morphology and function, yet these two cell types frequently arise from common neural precursors. In vertebrates, in vivo clonal analyses and neural stem cell culture experiments have provided strong evidence that single neural stem cells can generate both neurons and glia (Davis and Temple, 1994; Golden and Cepko, 1996; Johe et al., 1996; Qian et al., 1998; Reid et al., 1997; Stemple and Anderson, 1992; Temple, 1989; Walsh and Cepko, 1992). Similarly, a subset of Drosophila neural stem cells (here called neuroglioblasts; NGBs) are known to generate mixed neuronal/glial lineages that contribute ∼75% of all embryonic CNS glia (Akiyama-Oda et al., 1999; Bossing et al., 1996; Schmid et al., 1999; Schmidt et al., 1997).

What are the molecular mechanisms that govern the production of neurons versus glia within a single stem cell lineage? In Drosophila, a critical regulator of glial identity is the glial cells missing (gcm) gene. gcm encodes a novel transcription factor that is expressed in all Drosophila embryonic glia except for midline/mesectoderm-derived glia (Akiyama et al., 1996; Hosoya et al., 1995; Jones et al., 1995; Schreiber et al., 1997; Vincent et al., 1996). gcm loss-of-function mutant embryos show a transformation of CNS glia into neurons (Hosoya et al., 1995; Jones et al., 1995; Vincent et al., 1996). Conversely, overexpression of gcm within the CNS causes nearly every cell to express glial-specific genes, and gcm misexpression in identified sensory neurons transforms them into glia based on both morphological and molecular features (Hosoya et al., 1995; Jones et al., 1995). Thus, within the nervous system gcm appears to act as a genetic switch, with cells that are Gcm positive developing as glia, and cells that are Gcm negative developing as neurons. gcm is capable of positively regulating its own expression (Miller et al., 1998), which highlights the importance of precisely regulating gcm expression patterns and levels.

How is gcm expression regulated in neural stem cell lineages that give rise to both neurons and glia? Two recent studies have investigated the cell division profile and patterns of gcm expression in an identified precursor, the thoracic NGB 6-4 (NGB 6-4T) (Akiyama-Oda et al., 1999; Bernardoni et al., 1999), which gives rise to both neurons and glia. They report that the first division of NGB 6-4T is along the mediolateral axis and generates a larger lateral cell (here called the ‘post-divisional NGB’) and a smaller medial daughter cell (here called the ‘G daughter cell’). The G daughter cell expresses gcm and produces three glia; the post-divisional NGB lacks gcm expression, switches to an apical-basal division axis, and begins generating neurons. Interestingly, both reports propose that gcm mRNA is asymmetrically localized into the G daughter cell during the first division of NGB 6-4T, and that this asymmetric localization of gcm triggers glial commitment in G, whereas lack of gcm mRNA in the post-divisional NGB allows the production of neural progeny, thereby bifurcating neuron and glial cell fates (Akiyama-Oda et al., 1999; Bernardoni et al., 1999).

This model raises several important questions: how is gcm mRNA localized? What orients the spindle along the mediolateral axis during glia-producing divisions? How is it reoriented along the apical-basal axis for neuron production? Recent work has identified several proteins that may be involved in these processes. The Inscuteable (Insc) protein is apically localized in mitotic neuroblasts, where it is required for the apical-basal orientation of the spindle; indeed, misexpression of Insc in epithelial cells that normally divide perpendicular to the apical-basal axis is sufficient to re-orient their spindles along the apical-basal axis (Kraut et al., 1996). Miranda, Prospero, Staufen and Numb proteins are all localized to the basal cortex of mitotic NBs, and subsequently partitioned into the basal daughter cell (Fuerstenberg et al., 1998a; Matsuzaki, 2000). Miranda is an adapter protein necessary for basal localization of Prospero and Staufen (Ikeshima-Kataoka et al., 1997). Prospero is a transcription factor required for daughter cell-specific gene expression and triggering exit from the cell cycle (Chu-Lagraff et al., 1991; Doe et al., 1991; Li and Vaessin, 2000; Vaessin et al., 1991), while Staufen is an RNA-binding protein necessary for basal localization of prospero mRNA (Broadus et al., 1998). Numb is a membrane-associated protein that can affect cell fate by inhibiting Notch signaling (Frise et al., 1996; Guo et al., 1996; Spana and Doe, 1996). Each of these proteins is a candidate for regulating gcm expression or mRNA localization in the NGB 6-4T lineage.

We initially set out to investigate whether Insc expression or function regulates spindle orientation in NGB 6-4T, and whether Staufen is required for the asymmetric localization of gcm mRNA in NGB 6-4T. We report, in contrast to previous studies, that all divisions of NGB 6-4T are along the apical-basal axis, and that there is no evidence for the asymmetric localization of gcm mRNA or protein. We show that there is transient low level gcm expression in NGB 6-4T and its progeny during its phase of glial production. We show that the Prospero transcription factor is asymmetrically localized into the daughter cells of at least two NGBs (6-4T and 7-4), where it is required to upregulate gcm expression and induce the glial developmental program; failure to localize Prospero leads to ectopic gcm expression in these NGBs and the production of extra glia. We propose a novel model for glial production in mixed neuron/glia lineages that requires the asymmetric localization of the Prospero transcription factor.

All Drosophila stocks were grown at 25°C on standard cornmeal-molasses agar medium. Control animals were from a y w genetic background. Mutant alleles used in this study were as follows: pros¹⁷, pros¹⁴ (Doe et al., 1991), pros^I13 (Srinivasan et al., 1998), mir^ZZ176 (Ikeshima-Katoaka et al., 1997), stau^ry9 (St Johnston et al., 1991) and numb² (Frise et al., 1996). All prospero mutant data presented in figures is from our analysis of pros¹⁷; however, similar results were obtained using the pros¹⁴ and pros^I13 alleles. Mutant stocks were maintained over CyO, ftz-lacZ or TM3, ftz-lacZ balancer chromosomes.

Standard methods were used for collection and fixation of Drosophila embryos. Immunohistochemistry was performed as described elsewhere (Spana and Doe, 1995). Primary antibodies were used at the following dilutions: mouse anti-Engrailed, 1:5; rabbit anti-Miranda 1:500 (Fuerstenberg et al., 1998b); mouse anti-Prospero, 1:5 (Spana and Doe, 1995); rabbit anti-Inscuteable, 1:1,000 (Kraut and Campos-Ortega, 1996); mouse anti-Staufen, 1:100 (Broadus et al., 1998); rabbit anti-Numb, 1:250 (Rhyu et al., 1994); rabbit anti-Eagle, 1:1000 (Higashijima et al., 1996); rat anti-Glial cells missing, 1:1000 (Jones et al., 1995); and rat anti-Repo, 1:1000 (Campbell et al., 1994). Actively mitotic cells were identified using rabbit anti-phosphohistone H3 antibodies at 1:2500 (Upstate Biotechnology). Secondary antibodies used for immunofluorescence were conjugated to DTAF, LRSC, or Cy5 (Jackson ImmunoResearch) and were used at 1:400. Embryos were mounted in anti-fade reagent (BioRad) and viewed on a BioRad MRC-1024 confocal microscope. For immunohistochemistry we used the Vectastain HRP detection kit (Vector labs). β-Galactosidase expressed from balancer chromosomes was detected using a rabbit anti-β-galactosidase primary antibody at 1:1000 (Cappel), followed by anti-rabbit secondary antibodies conjugated to alkaline phosphatase, and detected using standard procedures (Boehringer Mannheim). After staining, embryos were mounted in glycerol, viewed on a Zeiss Axioplan microscope, and imaged with a Sony DK-5000 digital camera.

Digoxigenin-labeled RNA probes were generated according to the manufacturer’s instructions (Boehringer Mannheim). RNA in situ hybridization to embryos were carried out as described previously (Broadus et al., 1998). Immunofluorescent detection was accomplished using fluorescein-conjugated anti-digoxigenin antibodies (Boehringer Mannheim).

NGBs 6-4 and 7-4 are the most lateral Engrailed positive cells in the neuroblast array (Broadus et al., 1995). The most lateral Eagle-positive/Engrailed-positive cell is NGB 6-4T. Eagle is detectable in NGB 6-4T before its first division and subsequently in all NGB 6-4T progeny (Higashijima et al., 1996). Glial progeny of NGBs 6-4 and 7-4 were identified by staining with antibodies to either Gcm or Repo.

For RNA in situ hybridization with gcm mRNA, we used the following criteria to identify NGB 6-4T before, and after, its first division: (1) gcm expression in the lateral glial precursor (GP) was used as a positional landmark – the GP is slightly posterior and lateral to NGB 6-4. Prior to the first division of 6-4T the GP is the only cell in the lateral-most region of the CNS to express gcm. (2) The first division of NGB 6-4T is tightly correlated temporally with the production of two glial cells by NGB 7-4. Three or more Gcm-positive cells at the position of NGB 7-4 are only observed after NGB 6-4T has undergone its first division. (3) NGB 6-4T consistently divides before the first division of GB 6-4A. (4) In experiments where it was not directly labeled, we identified NGB 6-4T by its position in the neuroblast array relative to NB 6-2, NB 7-2, NGB 7-4 and the lateral GP using either Nomarski optics or background immunofluorescence.

In thoracic segments the neural precursor 6-4 generates both glia and neurons, and is referred to as NGB 6-4T; in abdominal segments the 6-4 precursor produces only glia, and so we call it GB 6-4A (Akiyama-Oda et al., 1999; Bossing et al., 1996; Schmid et al., 1999). The neural precursor 7-4 generates a lineage composed of both neurons and glia in all segments, and we refer to it as NGB 7-4.

NGB 6-4T and GB 6-4A form at early embryonic stage 10 as part of the S3 wave of neuroblasts (Broadus et al., 1995). We find that the first division of NGB 6-4T is oriented along the apical-basal axis, producing a large apical post-divisional precursor (NGB 6-4T) and a smaller basal daughter cell (G) (Fig. 1A; Fig. 2). Gcm is expressed before this first division, both daughter cells inherit Gcm protein, which enters the nucleus in these cells immediately after NGB division (Fig. 1A). Interestingly, shortly after NGB 6-4T completes this division, the G daughter cell migrates from its basal position to a position just medial to the post-divisional NGB 6-4T (Fig. 1A; Fig. 2), which may explain why this division was previously scored as mediolateral (Akiyama-Oda et al., 1999; Bernardoni et al., 1998). By the end of the G cell medial migration, Gcm protein is downregulated in NGB 6-4T (Fig. 1A) and maintained only in the G daughter cell. 6-4T continues to divide along the apical-basal axis and subsequently produces neuronal progeny (Fig. 1A). The G daughter cell maintains high levels of Gcm protein and produces three glial cells (Fig. 1A). These glia continue to express Gcm protein, activate the glial specific gene reversed polarity (repo) (Fig. 1A) (Campbell et al., 1994; Xiong et al., 1994), migrate medially, and differentiate into cell body glia (CBGs) (Schmid et al., 1999; Schmidt et al., 1997).

These results raise the question of how Gcm protein becomes asymmetrically restricted to the G daughter cell after the first division of NGB 6-4T. It has been proposed that gcm mRNA is asymmetrically partitioned into the G daughter cell during mitosis of NGB 6-4T (Akiyama-Oda et al., 1999; Bernardoni et al., 1999). To test this model, we scored gcm mRNA localization by fluorescent in situ hybridization. We observe low levels of gcm mRNA in the predivisional NGB 6-4T, followed by uniform localization in the mitotic NGB 6-4T, and equal distribution to both NGB 6-4T and G sibling cells (Fig. 2). After G cell migration we observe gcm mRNA in the G cell and down-regulation of gcm mRNA in the post-divisional NGB 6-4T (Fig. 2). We have also scored gcm-expressing cells throughout the CNS, and have never observed asymmetric localization of gcm mRNA in any mitotic precursor cell (n=31). We conclude that gcm mRNA is not asymmetrically localized in the mitotic NGB 6-4T, but rather that it becomes transcriptionally upregulated in the G daughter cell soon after it is born.

Similar to NGB 6-4T, GB 6-4A divides along the apical-basal axis and its basal daughter cell rapidly migrates to a position medial to its apical sibling (Fig. 1B). In contrast to NGB 6-4T, GB 6-4A expresses high levels of Gcm protein and mRNA before its first division, and both daughter cells maintain gcm expression (Fig. 1B). These two cells subsequently express repo, migrate medially, and differentiate into cell body glia (CBGs; Fig. 1B).

NGB 7-4 forms at late stage 8 as the most lateral En-positive S1 neuroblast (Broadus et al., 1995). The first progeny from NGB 7-4 are Prospero positive and Gcm negative, and differentiate into neurons. At stage 10 (just before the formation of 6-4 neural precursors) NGB 7-4 begins producing several Prospero-positive, Gcm-positive daughter cells that make a total of six to seven glia (Fig. 1C). At stage 12, NGB 7-4 switches back to making Prospero-positive Gcm-negative daughter cells that develop into neurons. All divisions of NGB 7-4 are along the apico-basal axis (Fig. 3A); Gcm-positive progeny are budded off the basal surface of NGB 7-4 but then migrate extensively to their final positions; ultimately, two glia migrate along the ventral surface of the CNS and differentiate as a pair of En-positive midline channel glia, three remain on the ventral surface of the CNS near NGB 7-4 and develop into CBGs posterior to the En-positive stripe; and one or two migrate to a position slightly dorsal and lateral to NGB 7-4 and differentiate as lateral subperineurial glia (Fig. 1D) (Bossing et al., 1996; Schmid et al., 1999). In contrast to the 6-4 neural precursors, we do not detect gcm mRNA or protein in NGB 7-4, only in its glial progeny.

In summary, we find that glia-producing divisions of NGB 6-4T and GB 6-4A occur along the apicobasal axis, and that these divisions are followed by medial migrations of glial progeny. Gcm mRNA and protein are present in the predivisional NGB 6-4T, and Gcm protein enters the NGB and daughter cell nuclei immediately after the first division of this NGB. In addition, we find no evidence for asymmetric localization of gcm mRNA in any glial lineage, including NGB 6-4T.

If gcm mRNA and protein are equally distributed into NGB 6-4T and its first-born G daughter cell, how are gcm mRNA and protein levels upregulated in G but not NGB 6-4T? To address this issue, we assayed mitotic NGB 6-4T for proteins known to be asymmetrically localized along the apical-basal axis of neuroblasts. Our goal was to identify candidate genes that could differentially regulate gcm expression in the NGB 6-4T lineage. Insc protein marks the apical side of most or all mitotic neuroblasts and is necessary and sufficient for apical-basal spindle orientation. In NGB 6-4T, Insc is localized as an apical crescent at all stages of mitosis (Fig. 3A) and is partitioned into the apically-positioned NGB 6-4T following cytokinesis. The mitotic GB 6-4A also shows apical Insc localization (Fig. 3A). Because Insc is sufficient to orient the mitotic spindle in all neuroblasts and epithelial cells assayed (Kraut et al., 1996), the apical localization of Insc in NGB 6-4T and GB 6-4A provides strong confirmation that both cells divide along their apical-basal axis.

Miranda, Prospero, Staufen, and Numb proteins mark the basal side of many or all mitotic neuroblasts and regulate the fate of daughter cells or their neuronal progeny. In NGB 6-4T, Miranda, Prospero, Staufen and Numb all form basal crescents from metaphase through telophase (Fig. 3A; data not shown for Staufen), and are partitioned into the basally positioned G daughter cell of NGB 6-4T after cytokinesis. The mitotic GB 6-4A also shows basal localization of Miranda, Prospero, Staufen and Numb (Fig. 3A; data not shown for Staufen). These results further confirm the apical-basal division axis of NGB 6-4T and GB 6-4A during glial producing divisions, and show that all of the above proteins are candidates for regulating gcm expression in the basal G daughter cell of NGB 6-4T (Fig. 3B).

To determine if miranda, prospero, staufen or numb are involved in the development of glia in the NGB 6-4T lineage, we scored embryos mutant for each gene for the number and position of mature glia derived from NGB 6-4T. The three glia from NGB 6-4T express repo and have distinctive positions within the CNS: two near the midline and one between NGB 6-4T and the midline (Fig. 4A). These are the only Repo-positive glia adjacent to the midline at the ventral surface of the CNS, and thus are easy to identify unambiguously. We find that mutations in staufen and numb have no effect on glial development in the NGB 6-4T lineage (Fig. 4B,C; n>20 for each). By contrast, prospero mutant embryos show striking loss of NGB 6-4T-derived glia (Fig. 4D), while miranda mutant embryos have a similar but weaker phenotype (Fig. 4E, see below). We conclude that prospero and miranda, but not staufen or numb, are required for normal glial development in the NGB 6-4T lineage.

To determine earliest aspect of the prospero mutant phenotype in the NGB 6-4T lineage, we assayed whether gcm is expressed normally in the G daughter cell. In wild-type embryos, Gcm protein is detectable in the predivisional NGB 6-4T and in the sibling NGB 6-4T/G cells immediately after cytokinesis; subsequently, Gcm disappears from NGB 6-4T and is upregulated in the G cell and its progeny (e.g. Fig. 1A), which proceed to migrate medially and express repo (Fig. 1A). In prospero mutant embryos, gcm expression is activated normally in the predivisional NGB 6-4T (Fig. 5A) and is detectable in the immediately post-mitotic NGB 6-4T and G cell. Therefore the early induction of gcm expression in these cells is clearly prospero-independent. However, Gcm protein levels subsequently decline in the G cell and its progeny and these cells fail to migrate to the midline (Fig. 5A) or express repo. These data indicate that Prospero is required in the G daughter cell to maintain or upregulate gcm expression levels, induce medial migration, and activate repo expression. Surprisingly, these Gcm negative, Repo negative cells do not express the neuron-specific elav gene (data not shown), and thus they appear unable to differentiate as glia or neurons.

In prospero mutant embryos, gcm expression is also greatly reduced in the progeny of NGB 7-4 (Fig. 5B). Low level expression of gcm is detectable in many NGB 7-4 progeny shortly after their birth, indicating that in this lineage (as in NGB 6-4T) the induction of gcm expression can occur in the absence of prospero function. However, gcm expression fades rapidly and these cells never express repo. Thus, prospero is essential for the maintenance of gcm expression and normal glial cell fate induction in both the NGB 6-4T and 7-4 lineages.

prospero is clearly necessary for upregulation of gcm and glial cell fate induction in the 6-4T and 7-4 lineages, but is it sufficient to induce gcm expression in these lineages? In miranda mutant embryos, prospero mRNA and protein are delocalized during neural precursor cell division, resulting in similar concentrations of Prospero segregating to both NGBs and their daughter cells (Fuerstenberg et al., 1998a; Ikeshima-Kataoka et al., 1997). Interestingly, in miranda mutants we find ectopic gcm expression in NGB6-4T at stage 13 (Fig. 6A), a time when this NGB is normally making neuronal progeny. miranda mutants also show ectopic expression of gcm in NGB 7-4 lineage during its window of glial production (Fig. 6A). Thus, mislocalization of Prospero to the NGB by removal of miranda function is sufficient to induce ectopic gcm expression in these NGBs.

Does the upregulation of gcm in NGBs drive the production of extra glial progeny? To address this question, we assayed Repo expression in the NGB6-4T lineage in miranda mutants as we can identify the entire NGB 6-4T lineage (see Materials and Methods for details). In miranda mutants we typically find only four Repo positive cells in the entire NGB 6-4T lineage, but neuronal progeny are completely absent (n=45; 46% of hemisegments scored) (Fig. 6B). We interpret this phenotype to indicate that the G cell produces three glia as usual, but that its sibling NGB differentiates directly into a Repo-positive glia cell, resulting in a termination of the lineage (see Discussion). Thus, it appears that Prospero mislocalized to the NGB can potently activate gcm expression in the NGB and transform it into a glial cell.

We also find two additional phenotypic classes in miranda mutants: (1) a variable number of Repo-positive glia are produced (between two and four) and subsequent neuronal progeny are generated normally (12% of hemisegments); or (2) the wild-type pattern of three Repo-positive glia and neuronal progeny are produced (42% of hemisegments scored; e.g. Fig. 6B). These phenotypes indicate that low level Prospero in the NGB is not always sufficient to induce a glial fate, and that reduced Prospero in the G cell may lead to fewer glial progeny.

Mislocalization of Prospero to NGB 7-4 by removal of miranda function can also induce Repo expression in this NGB (25% of hemisegments), showing that NGB 7-4 can also be partially transformed towards a glial fate. We do not know if this NGB differentiates as a glial cell (like the Pros-positive NGB 6-4T), generates extra glial progeny, or if it can eventually produce neurons. miranda, prospero double mutants do not show upregulation of gcm in NGBs 6-4T or 7-4 (data not shown), demonstrating that the upregulation of gcm in these NGBs in miranda mutant embryos is due to Prospero protein that is delocalized into the NGB. Our results indicate that Miranda, by asymmetrically localizing Prospero to NGB daughter cells, restricts gcm upregulation and induction of the glial developmental program to the progeny of NGBs 6-4T and 7-4 during their phases of glial production (Fig. 7).

The Drosophila NGB 6-4T has been reported to divide along the mediolateral axis when producing glia but along the apical-basal axis when producing neurons (Akiyama-Oda et al., 1999; Bernardoni et al., 1999). GB 6-4A was also reported to divide along the mediolateral axis (Akiyama-Oda et al., 1999). From these observations it was suggested that mediolateral cell divisions correlate with glial production while apical-basal division correlate with neuron production. The above studies relied upon the use of general DNA stains and cell position to document the mediolateral division; however, the use of general DNA stains make it hard to distinguish each phase of the cell cycle, and cell position can change rapidly due to cell migration. In this study, we stained for phosphorylated histone to specifically label mitotic cells, and we used antibodies that exclusively mark apical or basal membrane domains of neural precursors. In contrast to previous reports, we find that NGB 6-4T and GB 6-4A always divide along the apical-basal axis, and that in each case the basal daughter cell rapidly migrates medially. We suspect that previous workers missed this apical-basal cell division, and only scored the post-migration mediolateral cell arrangement. From our data we conclude that the glia-producing divisions of NGBs 6-4T and 7-4, and GB 6-4A occur along the apical-basal axis, similar to the division axis of all other Drosophila embryonic neural stem cells.

Cell lineage studies show that the grasshopper median neuroblast (MNB) is a multipotent neuronal stem cell that generates both neurons and glia; it too is reported to have a mediolateral division axis when producing glia and an apical-basal division axis when generating neurons (Condron and Zinn, 1994). In this report, a careful analysis of the MNB at late telophase clearly showed a mediolateral division axis during the time glial progeny are produced. It is not clear if the MNB always, or only occasionally, divides mediolaterally when producing glia; whether there are differences between an unpaired medial precursor (MNB) and bilateral precursors (NGBs 6-4T and 7-4); or whether there are differences between grasshopper and Drosophila in the mechanisms regulating spindle orientation.

The phenomena of apical-basal division followed by medial migration also occurs in the GB 6-4A and lateral GP lineages, with the difference being that all of the cells in these lineages ultimately form glia and migrate medially. Because the medial migration is correlated with gcm expression, and does not occur in NGB progeny that lack gcm expression, we suggest that high levels of gcm expression triggers migration of these glia towards the midline of the CNS. The nature of the cue that orients glial migrations is unknown.

Previous reports have suggested that gcm mRNA is asymmetrically localized in a medial crescent in the mitotic NGB 6-4T, resulting in the selective partitioning of gcm mRNA into the medial glia-producing progeny of NGB 6-4T (Akiyama-Oda et al., 1999; Bernardoni et al., 1999). We believe these conclusions to be in error for three main reasons. First, we show that the mitotic NGB 6-4T contains evenly distributed gcm mRNA, and that this mRNA is partitioned equally between NGB 6-4T and its glial-producing G daughter cell after cytokinesis. Second, previous studies did not use a mitosis-specific marker together with probes for gcm mRNA localization to prove that the localization was being scored in mitotic NGBs. Third, we show that NGB 6-4T always divides along the apical-basal axis, therefore a medial localization of gcm mRNA would not result in it being partitioned unequally into one daughter cell. Indeed, in a recent study that made use of more specific markers for mitotic stage and cell orientation it was found that the first division of NGB 6-4T is not along the mediolateral axis, and that gcm mRNA is inherited by both the NGB and the G daughter cell (Ragone et al., 2001). We therefore conclude that the asymmetric localization of gcm mRNA is not the mechanism by which neuronal and glial lineages are separated in the NGB 6-4T lineage.

Gcm protein has been reported to be absent from the predivisional NGB 6-4T, perhaps owing to translational repression of gcm mRNA prior to its first division (Akiyama-Oda et al., 1999). We show that this is not the case, as Gcm protein is clearly present in the predivisional NGB 6-4T (Fig. 1A) (Bernardoni et al., 1999). In addition, it has been reported that Gcm protein is excluded from the nucleus in the postdivisional NGB 6-4T, and that an unidentified mechanism regulates nuclear entry of Gcm specifically in the G daughter cell (Bernardoni et al., 1999). We have shown robust Gcm protein expression in the nucleus of NGB 6-4T (Fig. 1A), arguing strongly against the existence of such a mechanism.

We observe asymmetric localization of Prospero, Miranda, Staufen and Numb proteins into the glial-producing daughter cells during NGB divisions, but only mutations in prospero and miranda affect the production of glia from NGBs 6-4T and 7-4. These results are consistent with two recent reports of a requirement for prospero in the production of glia from selected NGBs (Akiyama-Oda et al., 2000; Ragone et al., 2001). In prospero mutant embryos, NGB 6-4T progeny do not migrate significantly towards the midline or express the glial-specific repo gene. These phenotypes are probably due to lack of gcm expression, as loss of prospero does not affect migration or repo expression in GB 6-4A (which has high gcm levels even in prospero mutant embryos).

We have extend these findings to show that Gcm expression is induced properly in prospero mutant embryos, but that both the G daughter cell and the post-divisional NGB downregulate gcm expression with a similar timecourse. Thus, the induction of gcm expression in NGB 6-4T is prospero-independent. Interestingly, prospero mutant embryos also fail to upregulate gcm to high levels in NGB 7-4 daughter cells. This is consistent with a previous report that NGB 7-4-derived Repo-positive glia are absent in prospero mutants (Akiyama-Oda, 2000). Unlike NGB 6-4T, we are unable to detect gcm expression in NGB 7-4, only in its new-born (pre-migration) daughter cells. Low level Gcm expression is still present in prospero mutant embryos; thus, the induction of gcm expression in this lineage also appears to be prospero-independent. We propose that prospero functions to upregulate low levels of gcm in these lineages, but is not sufficient to induce gcm expression on its own. Such a mechanism would explain why all neural stem cell progeny in the CNS express high levels of prospero but most never induce gcm and the glial developmental program. In agreement with this model, we have found that misexpression of low levels of gcm thoughout the CNS requires Prospero to induce glia in a subset of lineages (M. R. F. and C. Q. D., unpublished).

How might the Prospero transcription factor upregulate low levels of gcm expression? Prospero is known to act as a co-factor to stimulate transcriptional activity of several DNA-binding proteins (Hassan et al., 1997). Recent studies show that Gcm is can positively autoregulate its own expression in neural tissues (Miller et al., 1998). It is possible that Prospero may act together with Gcm to stimulate expression levels of the gcm gene until they become sufficiently high for Gcm to positively autoregulate its own expression. However, embryos homozygous for the gcm^N7–4 allele produce non-functional Gcm protein (Vincent et al., 1996) that is upregulated with normal kinetics in the NGB 6-4T and 7-4 lineages (M. R. F. and C. Q. D., unpublished), indicating that Gcm function is dispensable for its own upregulation. We propose that a lineage-specific co-factor or extrinsic signal converges with Prospero function in these lineages to upregulate gcm expression.

How does a NGB know when to make neurons or glia? For example, the first born daughter cell from NGB 6-4T gives rise to glia while all subsequent progeny are neuronal (Akiyama-Oda et al., 1999). By contrast, NGB 7-4 first produces several neuronal progeny, then switches to making glia, and finally switches back to making neurons (M. R. F. and C. Q. D., unpublished). Interestingly, the window of developmental time during which these two neural precursors are making glia are strikingly similar: NBG 7-4 begins making glia at stage 10, shortly after this NGB 6-4T is born (late stage 10) and begins making glia; at stage 11 both precursors terminate glial production and switch to making neurons. The coordinate timing of glial production from these lineages may indicate that a temporally regulated extrinsic cue induces gcm expression in these NGBs or their newly born progeny.

In prospero mutant embryos, NGB 6-4T and its progeny only transiently express low levels of gcm. What is the fate of these cells? They never express the glial-specific repo gene, and we have found that they also fail to express the neuron-specific elav gene, indicating that neither the glial or neuronal developmental program has been initiated. gcm is thought to transcriptionally activate genes that promote glial fate or repress neuronal fate (Klambt et al., 1999). We propose that NGB 6-4T in prospero mutants produces enough Gcm protein to repress neuron-specific genes, yet insufficient amounts to robustly induce glial-specific genes. This in turn suggests that there may be different gcm thresholds for activating glial development (high threshold) and for repressing neuronal development (low threshold).

In miranda mutant embryos, Prospero protein is delocalized at mitosis, allowing NGB/daughter cell siblings to inherit equal concentrations of Prospero. In these embryos, we frequently observe ectopically upregulated gcm in NGBs 6-4T and 7-4, and extra glia derived from NGB 6-4T. Akiyama-Oda et al. (Akiyama-Oda et al., 2000) report that in miranda mutants all progeny of NGB 6-4T take on a glial fate as determined by Repo staining. By contrast, we find that this transformation takes place in approximately 46% of hemisegments. This difference is not likely to be accounted for by the use of different miranda alleles as both appear to completely delocalize Prospero protein in NBs (Fuerstenberg et al., 1998b). Regardless of these differences, these data indicate that prospero is a potent activator of gcm expression in the NGB 6-4T and 7-4 lineages. The extra glia we observe could come from an extension of the glial portion of the NGB 6-4T lineage, or from a transformation of this NGB into a purely glial progenitor. We favor the latter model, because neurons are never observed in the NGB 6-4T lineage when we observe extra glia. Moreover, high levels of Gcm are correlated with pure glial lineages such as GB 6-4A and the GP, and gcm is known to positively autoregulate which may commit precursors with high Gcm to a glial-producing fate.

We also find that in miranda mutant embryos, the ectopic expression of gcm in NGB 6-4T and 7-4 is in fact due to delocalization of Prospero and not simply the absence of Miranda, as miranda, prospero double mutants fail to upregulate gcm in NGBs. In both the NGB 6-4T and 7-4 lineage, the delocalization of Prospero has relatively little effect on glial production by the daughter cells, presumably because there is sufficient Prospero protein in these daughter cells to upregulate gcm expression. Thus, with respect to glial cell fate induction, the asymmetric localization of Prospero may be more important for removing Prospero from the NGB than for enriching Prospero in the daughter cell.

We thank Brad Jones for Gcm antibodies and a gcm cDNA, and Joachim Urban for Eagle antibodies. We thank Chian-Yu Peng for the miranda, prospero double mutant line. We thank Bruce Bowerman, Barry Condron, Laura Breshears and Takako Isshiki for critical reading of the manuscript, and members of the Doe laboratory for many helpful discussions. This work was supported by the NIH (GM58899) and the HHMI of which C. Q. D. is an Associate Investigator.