During Drosophila neurogenesis, glial differentiation depends on the expression of glial cells missing (gcm). Understanding how glial fate is achieved thus requires knowledge of the temporal and spatial control mechanisms directing gcm expression. A recent report showed that in the adult bristle lineage, gcm expression is negatively regulated by Notch signaling (
INTRODUCTION
The two major cell types that compose the nervous system, neurons and glia, are in close contact with each other in all animals. Glial cells provide neurons with survival and axonal guidance cues, electrically shield axons by ensheathing them, and function as macrophages upon neuronal death. Such intimate functional relationships suggest a logical link of gliogenesis with neurogenesis through induction or by lineage. Indeed neurons and glia are known to arise from common multipotent precursors in both vertebrates and invertebrates (Turner and Cepko, 1987; Luskin et al., 1988; Udolph et al., 1993; Condron and Zinn, 1994). Signals that separate glial and neuronal lineages are largely unknown.
In Drosophila, development of CNS as well as PNS glial cells is dependent on the expression of glial cells missing (gcm) (also called glide) (reviewed by Anderson, 1995). gcm encodes a transcription factor that is sufficient to activate the glial fate through regulation of the expression of its downstream target genes (Hosoya et al., 1995; Jones et al., 1995; Schreiber et al., 1997). reversed polarity (repo), a glia-specific homeobox gene (Campbell et al., 1994; Xiong et al., 1994; Halter et al., 1995), is a good candidate for a direct target of gcm, as it contains multiple GCM-binding sites in the 5′ upstream region (Akiyama et al., 1996). In gcm mutants, presumptive glial cells fail to differentiate, and are often transformed toward neurons. Thus, gcm acts as a binary switch between glial and neuronal cell fates and its transcriptional regulation plays a crucial role in their binary decisions.
Within the glial determination pathway, the gcm gene currently occupies the most upstream position (Hosoya et al., 1995; Jones et al., 1995; Vincent et al., 1996; Schreiber et al., 1997). This suggests that gcm transcription is regulated by a combination of factors that themselves are not specific to glia. The gcm promoter may integrate a set of developmental signals that are identical in all gcm-positive glia. Alternatively, each glial subtype may have its own regulatory system, using various developmental cues differently depending on their context. Distinguishing these possibilities requires a comparative analysis of gcm regulation in multiple glial subtypes.
The transmembrane receptor Notch is used in many developmental contexts for determination of binary cell fates, such as asymmetric cell divisions (reviewed by Artavanis-Tsakonas et al., 1999). Dividing cells can receive a cell-intrinsic cue by Numb, a membrane protein containing a phosphotyrosine-binding domain, that binds Notch and represses Notch signaling in one of the daughter cells (Uemura et al., 1989; Posakony, 1994; Guo et al., 1995; Jan and Jan, 1995). Recently, a role for Notch signaling in glial differentiation was found in the adult PNS; Notch signaling negatively regulates gcm expression and glial cell differentiation during asymmetric division in the bristle lineage (Van De Bor and Giangrande, 2001). Given the ability of Notch to function in many developmental decisions, Notch activation may be an obligatory signal in repressing gcm transcription in asymmetric divisions that generate glia. However, a previous report suggested an opposite role for Notch signaling in glial differentiation in the embryonic PNS. The dorsal bipolar dendritic (dbd) lineage consists of one neuron and a glial cell, where glial differentiation depends on gcm activity (Bodmer et al., 1989; Brewster and Bodmer, 1995; Jones et al., 1995). In this lineage the numb mutation shows a double-glia phenotype at the expense of the neuron (Brewster and Bodmer, 1995). If Numb acts by repressing Notch activity in the dbd lineage, this result would imply that Notch promotes, rather than represses, glial development in this lineage.
In this study, we present direct evidence that Notch positively regulates glial differentiation in the dbd lineage. Our data indicate that Notch signaling activates gcm expression in one of the two sibling cells and acts postmitotically to specify the glial fate. Thus, the effect of Notch signaling on gcm transcription is reversed in the bristle lineage compared to the dbd lineage. We identified an additional lineage where Notch promotes glial fate, and have shown that a molecular context similar to that of the dbd lineage works during asymmetric cell division. Our data indicates that one of the factors that provides Notch-dependent gliogenic context is likely the POU-domain protein Nubbin/PDM-1.
MATERIALS AND METHODS
Fly stocks
Canton-S was used as the normal strain. The following mutant stocks were used in this study: Notch55e11 (Simpson, 1994), sanpodoC55 (Dye et al., 1998), numb1 (Uemura et al., 1989), scB57 (Lindsley and Zimm, 1992) and gcme1 (Hosoya et al., 1995). UAS-NΔB2a2 (=UAS-Notchact) (Doherty et al., 1996), UAS-gcm (Hosoya et al., 1995) and UAS-nub (=UAS-pdm-1) (Neumann and Cohen, 1998) strains were used for ectopic expression by crossing these lines with the C155-GAL4 effector line that directs expression in all embryonic neurons (Lin and Goodman, 1994). UAS-nub UAS-NΔB2a2 strain was produced by chromosome recombination. p12xSu(H)bs-lacZ strain (Go et al., 1998) was used to visualize the Su(H)-dependent Notch activity, and is referred to as Su(H)-reporter in the text.
Embryo staining
Antibody staining was carried out as described previously (Ito et al., 1995). The following primary antibodies were used: rabbit anti-REPO (Halter et al., 1995) (a gift from G. M. Technau) at 1:500; rat anti-REPO (Yuasa et al., personal communication) (a gift from H. Okano) at 1:500; rabbit anti-Nubbin/PDM-1 (Yeo et al., 1995) (a gift from W. Chia) at 1:500; mouse anti-ELAV (9F8A9; Developmental Studies Hybridoma Bank) at 1:100; mouse mAb 22C10 (Fujita et al., 1982) at 1:100-200 and mouse mAb anti-β-galactosidase (40-1a; Developmental Studies Hybridoma Bank) at 1:100-200. Secondary antibodies used were biotinylated goat anti-rat, anti-rabbit, anti-mouse (Vector Laboratories), FITC-conjugated goat anti-rat and Cy3-conjugated goat anti-rabbit (Jackson) antibodies, all at 1:200. Biotinylated secondary antibodies were detected using the ABC elite kit (Vector Laboratories). Double labeling involving horseradish peroxidase (HRP) histochemistry was performed using diaminobenzidine as a substrate, with NiCl for the first staining, and without NiCl for the second staining. Immunofluorescence was viewed with a BioRad MRC 1024 confocal microscope. Mutant embryos were identified by the lack of anti-β-galactosidase staining from the balancer chromosome or by their typical phenotypes reported previously. Whole-mount in situ hybridization was performed essentially as described previously (Lehmann and Tautz, 1994). Digoxigenin-labeled RNA probes were generated from full-length gcm cDNA (Hosoya et al., 1995) and from full-length repo cDNA (Xiong et al., 1994) (a gift from H. Okano). Fluorescence-labeled RNA probe was generated from 1.2 kb PCR fragment derived from nubbin/pdm-1 cDNA. TSA system (NEN Life Science Product) was used for the fluorescence-labeled RNA detection. Images were processed using Photoshop software (Adobe).
RESULTS
Postmitotic activation of gcm determines glial fate in the dbd lineage
Three types of REPO-positive glial cells are present in each abdominal hemisegment of the embryonic PNS; one dbd support glial cell (DBDG), one PG3 glial cell and five ligament cells of the lateral chordotonal organ (Campbell et al., 1994; Xiong et al., 1994; Halter et al., 1995) (Fig. 1). DBDG and the dbd neuron can be reliably identified by their location and characteristic cell morphologies at stage 16 (Bodmer et al., 1989; Brewster and Bodmer, 1995) (Fig. 1C). However, the pattern of cell division in the dbd lineage remained unclear because of the lack of useful markers. We found that the expression of a POU domain protein Nubbin (also called PDM-1) enables a developmental analysis of the dbd lineage (Fig. 2A). The sensory organ precursor (SOP) of the dbd lineage could be identified at the beginning of stage 12 as a large, weakly Nubbin-expressing cell in the anterior-dorsal region of the abdominal segments (data not shown). This cell soon divided asymmetrically, producing two Nubbin-positive daughters, the larger one located basal to the smaller one. The larger cell expressed neuronal marker protein ELAV and differentiated as the dbd neuron, whereas the smaller daughter expressed REPO, migrated dorsally and became DBDG. No further cell division was observed in this lineage.
While the dbd neuron continued to express Nubbin, its sibling glial cell underwent a rapid change in Nubbin expression during gliogenesis (Fig. 2A,C). Shortly after mitosis, Nubbin was temporally down-regulated in the presumptive glial cell, and gcm mRNA expression was initiated. Upon establishment of high gcm activation, Nubbin was re-expressed in the glial daughter cell. gcm expression remained at high levels during stage 12 and then disappeared rapidly. Subsequently, Nubbin was again down-regulated in DBDG, resulting in expression specific to the dbd neuron at stage 16 (Dick et al., 1991; Lloyd and Sakonju, 1991) (Fig. 2B). The final down-regulation of Nubbin in DBDG was dependent on gcm function, whereas the preceding modulation occurred normally in gcm mutant embryos (data not shown).
Notch activity is restricted to the glia in the dbd lineage
To assess the role of Notch in the binary fate decision of the dbd lineage, we first analyzed Notch activity within this lineage. It is known that activation of Notch results in the nuclear translocation of an intracellular domain of Notch together with a transcription factor Suppressor of Hairless [Su(H)] (reviewed by Honjo, 1996; Weinmaster, 1997; Bray, 1998). We used transgenic lines containing a lacZ reporter construct driven by the E(spl)mγ promoter fused to multimerized Su(H)-binding sites, which we refer to as the Su(H)-reporter. This line allows effective visualization of a direct response to the Su(H)-dependent Notch activity in vivo (Go et al., 1998); in the sanpodo mutant, the phenotype of which mimicks the loss of Notch activity (Park et al., 1998; Skeath and Doe, 1998), the Su(H)-reporter activity was dramatically reduced throughout the embryo, as it was in Notch mutant embryos (Fig. 3A-C). During normal development of the dbd lineage, strong Su(H)-reporter activity was observed in the glial cell but not in the dbd neuron (Fig. 3D). When we expressed a constitutively active form of Notch (UAS-Notchact) in all neurons, the Su(H)-reporter became activated in the presumptive dbd neuron (Fig. 3E). Conversely, in the sanpodo mutation, the reporter expression was undetectable in the presumptive DBDG (Fig. 3F). Since the sanpodo mutation produces a double-neuron phenotype at the expense of the glial cell (Dye et al., 1998), the reporter expression correlates with the glial fate. In contrast, forced gcm expression in the presumptive neuron failed to activate the Su(H)-reporter even though neuron-to-glia transformation took place (Jones et al., 1995) (Fig. 3G). Thus Notch activity is likely to be upstream of gcm expression and glial differentiation.
dbd glial differentiation fails in Notch mutant embryos
To test the requirement of Notch in embryonic PNS glial development, we examined REPO expression in Notch mutant embryos. Since Notch is required for lateral inhibition, its removal leads to production of excess neuronal precursors in the PNS as well as in the CNS (reviewed by Artavanis-Tsakonas et al., 1995; Campos-Ortega, 1995). Indeed, at the earliest stage of the dbd SOP division, excess Nubbin-positive cells were located at the correct position, suggesting that supernumerary dbd SOPs were produced (data not shown). However, at stage 12 no REPO-positive cells were associated with Nubbin-positive neurons, indicating the absence of DBDG (Fig. 4B). This requirement for Notch in gliogenesis was specific to the dbd lineage, as supernumerary REPO-positive PG3 and ligament cells were present in Notch mutant embryos (Fig. 4B).
To test whether the defects in glial development within the dbd lineage arose through the misregulation of gcm, we analyzed gcm mRNA expression during stage 12, when its expression level is highest. No gcm expression was found within the dbd lineage of Notch mutant embryos, whereas PG3 and ligament cells were gcm positive (Fig. 4D). These results indicate that there is a specific requirement of Notch activity for gcm expression in the dbd lineage.
Notch pathway activates glial fate
If Notch activity plays an instructive role in gliogenesis, artificial activation of Notch in the presumptive neuron may cause a neuron-to-glia transformation. When constitutively active Notch was expressed in all neurons, the dbd neuron was replaced by an extra REPO-positive cell that was associated with and resembled a glial cell (83% of hemisegments; n=42, Fig. 5B). In contrast, expression levels of neuronal markers ELAV (data not shown) and Nubbin were dramatically reduced in the dbd lineage (Fig. 5F). Accompanying glial transformation of the presumptive dbd neuron to a glial cell, gcm was ectopically expressed in this cell. gcm responded quickly to Notch activation; at stage 12, gcm mRNA was already detectable in the presumptive neuron, coinciding with the expression of the driver construct (Fig. 5J). Although normal expression of gcm in the glial cell was transient, ectopic gcm that was induced by constitutively active Notch continued to at least stage 16 (Fig. 5L). This is unlikely to be due to autoregulation, because forced expression of gcm in the dbd neuron did not activate transcription of a lacZ reporter gene inserted into the gcm locus (data not shown). These data indicate that within the dbd lineage Notch activation is sufficient for inducing gcm expression.
Our misexpression experiment also offers an explanation for the double-glia phenotype of numb mutants within the dbd lineage (Brewster and Bodmer, 1995) (Fig. 5D). We found that in numb mutant embryos, gcm was misexpressed in the presumptive dbd neuron (Fig. 5K). In contrast, Nubbin expression was absent in the presumptive dbd neuron (84% of hemisegments; n=49, Fig. 5G). Since this phenotype is identical to the result of an artificial activation of Notch in neurons, it is likely that Numb represses Notch activity in the neuronal daughter upon asymmetric division of the dbd SOP. Thus both in the adult bristle lineage (Van De Bor and Giangrande, 2001) and in the dbd lineage, Numb causes asymmetric activation of Notch in two daughter cells. However, the effect of Notch activation is context dependent; in the bristle lineage Notch activity results in the repression of gcm expression, whereas in the dbd lineage Notch induces gcm transcription.
The Notch pathway can activate gcm transcription outside the dbd lineage
While the dbd lineage is the only place in the embryonic PNS where Notch activity is necessary for glial development, we discovered that artificial activation of Notch can initiate gliogenesis outside this lineage. When constitutively active Notch was expressed in all neurons, often two REPO-positive cells formed in the dorsal cluster of sensory organs (Fig. 6A). One of these cells is the transformed dbd neuron, as discussed above. We identified that another cell was derived from the dorsal dendritic arbor (dda) organ (Lloyd and Sakonju, 1991; Brewster et al., 2001) (Fig. 6A), and focused our analysis on this dda lineage.
The dda lineage has many similarities with the dbd lineage. While most embryonic sensory organs form as a result of the proneural activities of the bHLH proteins encoded by the ASC (Achaete-scute Complex) genes, both dbd and dda organs form independently of ASC genes, and require the proneural activity of absent md neurons and olfactory sensilla (amos) (Huang et al., 2000) (Fig. 6B). In addition, both dbd and dda SOPs express Nubbin. Just as in the dbd lineage, the dda SOP was seen to divide asymmetrically, and both daughters initially had high levels of Nubbin (Fig. 6B). One of the sibling cells continued to express Nubbin and differentiated into a dda neuron (Fig. 5E). In the other sibling cell, gcm and REPO expression was induced and Nubbin expression was down-regulated (Fig. 6B). Unlike the dbd glial cell, however, REPO expression was not maintained in the dda lineage and became undetectable by stage 16 (Fig. 1). The final fate of the sibling of the dda neuron is not known.
To analyze the role of Notch signaling within the dda lineage, we examined the effects of removing sanpodo and numb activities. In normal embryos a single dda neuron can be found dorsal to the dbd organ in each abdominal hemisegment at stage 16. In sanpodo mutant embryos, duplication of the dda neuron was frequently observed (71% of hemisegments; n=49, Fig. 5H), whereas in numb mutants they were absent (63% of hemisegments; n=49, Fig. 5G). Thus, as in the dbd lineage, Notch activity represses neuronal development in the dda lineage. In numb mutant embryos, we detected ectopic gcm and REPO expression in the presumptive dda neurons (34% of hemisegments; n=41, Fig. 5D and data not shown), a phenotype mimicking the artificial activation of Notch in neurons (Fig. 5L, Fig. 6A). We conclude that Notch signaling can activate gcm transcription not only in the dbd lineage, but also in the dda lineage.
The specific response of dbd and dda lineages to Notch activation suggests that factors specifically expressed in these lineages may provide the developmental contexts that allow Notch activation to be interpreted as a gliogenic signal. Although the proneural gene amos is specifically expressed in these two lineages, its expression is transient and is absent by the time of the SOP division (Huang et al., 2000). Furthermore, coexpression of AMOS and constitutively active Notch in all neurons did not generate any additional glial cells compared to Notch activation alone (data not shown). Another candidate is the POU-domain protein Nubbin. In both dbd and dda lineages, Nubbin expression is initiated in the SOP and high levels of Nubbin protein are found in both daughter cells after the SOP division (Fig. 2, Fig. 6B). To test whether Nubbin can modify the effect of Notch activation, we coexpressed Nubbin and constitutively active Notch in all neurons. Upon such treatment, a few extra REPO-positive glial cells appeared dorsally to the dda lineage (Fig. 6C). These cells likely correspond to the extra glial cells that have been reported to form upon ectopic expression of GCM in neurons (Jones et al., 1995). This suggests that some of the presumptive neurons in the dorsal cluster were redirected to the glial differentiation pathway upon Notch activation. Such a phenotype was never observed when constitutively active Notch alone or Nubbin alone was expressed (data not shown). We propose that Nubbin is one of the factors that provide a developmental context for Notch-dependent gcm expression and glial differentiation.
DISCUSSION
Notch plays context-dependent roles in gliogenesis
We demonstrated that Notch signaling promotes glial fate during asymmetric division in the embryonic dbd lineage. Notch is specifically activated in the presumptive DBDG owing to the negative regulation by Numb in the sibling cell, and provides instructive information to induce gcm transcription and glial development. Expression of gcm occurs quickly after the artificial activation of Notch, even in cells that have initiated neuronal development. In gcm mutants, DBDG are transformed into neurons, although the activation of Notch, visualized by the Su(H)-reporter, is normal in the presumptive glia (data not shown). Likewise, ectopic expression of gcm in presumptive dbd neurons caused neuron-to-glia transformation without affecting Notch activity. These findings suggest that gcm expression appears to be the sole target of Notch activation in establishing glial fate in the dbd lineage. Within the 3.5 kb region upstream of the gcm gene, we identified two sequences perfectly matching the consensus core sites for Su(H) (Bailey and Posakony, 1995; Lecourtois and Schweisguth, 1995). Thus, gcm could be a direct target of Su(H), downstream of the Notch signaling pathway.
While our present data demonstrate a positive role for Notch in gliogenesis in the dbd lineage, other embryonic PNS glial cells do not require Notch activity for their formation. For example, in the adult bristle lineage Notch has an opposite function on gliogenesis; that of repressing gcm expression and glial development (Van De Bor and Giangrande, 2001). Thus the role of Notch in the regulation of gcm expression is context-dependent (Fig. 7). Notch has recently been shown to be a component of combinatorial signaling in cell fate determination in the Drosophila eye (Flores et al., 2000). It is possible that Notch signaling has different consequences depending on other factors that act on the same regulatory element.
A common developmental context allows Notch to promote gliogenesis
The context-dependent effect of Notch suggests that the gcm promoter may have a modular structure where each unit integrates different developmental signals. However, given the large diversity of glial subtypes in the nervous system, it is unlikely that each glial subtype has its own regulatory sequences and a unique mode of regulation. We favor a model in which gcm has a limited number of regulatory elements that respond to developmental signals that are present in multiple environments. Indeed we have shown that a subset of glial subtypes respond in a similar way to Notch signals: in addition to the dbd lineage, the dda lineage can also induce gcm transcription upon Notch activation. Comparison of these two lineages offers hints on the nature of the developmental context in which Notch activation causes gcm transcription.
One common feature that distinguishes dbd and dda lineages from other PNS lineages is the cell division pattern of their SOP. In dbd and dda lineages, SOPs divide to generate a neuron and a glial cell through an asymmetric division. In other gliogenic PNS lineages, the sibling cells of glial cells are not postmitotic neurons, but tertiary precursors that undergo further division to generate neurons and associated cells (Fig. 7). These observations suggest that an interaction with the neuronal sibling may play a crucial part in promoting the Notch-dependent gcm activation during asymmetric cell division. Recently, Notch was shown to positively regulate gcm expression in the Neuroblast 1-1A lineage of the CNS, where the sibling pattern is identical to that of the dbd lineage (Udolph et al., 2001) (Fig. 7C). This also supports the idea that the cell division pattern provides a context that determines the effect of Notch activity.
We showed that coexpression of constitutively active Notch with Nubbin also generates ectopic glia outside dbd and dda lineages. This raises the possibility that Nubbin may be a part of the developmental context that allows Notch to promote gliogenesis. Within the embryonic PNS, dbd and dda neurons are the only two neurons that express Nubbin. In both lineages, Nubbin is present in both SOP daughter cells, at the time of glia versus neuron cell fate choice. Furthermore, we detected temporal activation of Nubbin in presumptive glial cells derived from the NB1-1A lineage (data not shown). Nubbin thus might create a permissive environment for the activation of gcm expression by the Notch signal (Fig. 7A-C). Since coexpression of Nubbin and constitutively active Notch does not cause glial transformation of all neurons, additional factors must exist that create a Notch-dependent gliogenic context.
Nubbin is a POU-domain transcription factor with sequence-specific DNA-binding activity (Neumann and Cohen, 1998). The contextual role of Nubbin in Notch-dependent expression of gcm could employ a similar mechanism to the modulation of Notch activity in wing development, where Nubbin and Su(H) bind on the same enhancer element of Notch target genes (Neumann and Cohen, 1998). It will be interesting to further analyze the role of Nubbin in gliogenic lineages.
Acknowledgements
We thank Y. Akiyama-Oda, S. M. Cohen, D. Eastman, K. Matsuno and M. Okabe for fly stocks, and W. Chia, G. M. Technau, Y. Yuasa, and H. Okano for cDNAs and antibodies. The 40-1a monoclonal antibody generated by J. R. Sanes was obtained from the Developmental Studies Hybridoma Bank at The University of Iowa. We also thank K. Agata, L. Tsuda, M. Okabe and the members of our laboratory for helpful discussions, and Ms. M. Sakai, Ms. C. Asaka and Ms. M. Aono for technical assistance. This work was supported by the CREST program of Japan Science and Technology Corporation (Y. Hotta ) and grants from the Ministry of Education, Culture, Sports, Science and Technology (Y. Hiromi). Y. Hotta is a CREST researcher and Y. U. is a CREST fellow of Japan Science and Technology Corporation.