The regulation of apoptosis by Numb/Notch signaling in the serotonin lineage of Drosophila

The Drosophila central nervous system (CNS) develops from a small number of precursor cells that divide to produce a variety of neurons and glial cells. During the specification of cell fates, subsets of cells are induced to undergo programmed cell death or apoptosis(Abrams et al., 1993). Elucidating both the intrinsic and extrinsic signals that establish cell fates and trigger apoptosis is fundamental to understanding CNS development. This paper examines the regulation of apoptosis in the ventral nerve cord of Drosophila, specifically within in the cell lineage that produces the serotonergic neurons. The serotonin lineage is a relatively simple lineage,producing only four neurons. However, additional cells are observed when this lineage is examined in the apoptosis deficient line H99(Isshiki et al., 2001; Novotny et al., 2002). This suggests that the mitotic divisions of the serotonin lineage also produce several cells that undergo apoptosis.

The segmented Drosophila nerve cord develops from stereotyped division of 30 neuroblasts (NB) in each hemisegment(Doe, 1992; Goodman and Doe, 1993). A pair of serotonergic neurons in each hemisegment arise from NB7-3(Lundell et al., 1996). NBs are stem cells that undergo several asymmetric divisions producing a specific number of ganglion mother cells (GMCs). Each GMC divides once to form two neuronal or glial progeny (Hartenstein et al., 1987). The divisions of the NB7-3 lineage have recently been determined using a combination of molecular markers and clonal analysis(Isshiki et al., 2001; Novotny et al., 2002). NB7-3 produces three GMCs. GMC-1 produces two neurons: GW, a motoneuron, and EW1,the more medial serotonergic neuron. GMC-2 produces EW2, the more lateral serotonergic neuron. GMC-3 produces EW3, a neuron that synthesizes the neuropeptide corazonin. The GW neuron projects an axon ipsilateral and posteriorly, and the three EW interneurons all project axons anteriorly to the posterior commissure (Fig. 1)(Bossing et al., 1996; Higashijima et al., 1996; Dittrich et al., 1997; Schmid et al., 1999).

Several genes have been shown to be essential in the differentiation of the NB7-3 lineage. Sequential expression of the segmentation transcription factors, Hunchback→ Krüppel→Pdm1, within neuroblasts has been shown to be important in the development of several lineages, including NB7-3(Brody and Odenwald, 2002). The subsequent GMCs and neuronal progeny maintain the expression of the transcription factors that are present in the NB at their birth. In the case of NB7-3, Hunchback (Hb) is expressed only in GMC-1 and its progeny, and is both necessary and sufficient to define the fates of these cells(Isshiki et al., 2001; Novotny et al., 2002). Krüpple (Kr) is expressed in both GMC-1 and GMC-2 and is necessary and sufficient to establish the fate of the EW2 serotonergic neuron(Isshiki et al., 2001). Pdm1 is expressed primarily in EW2 (Fig. 1) (Lundell and Hirsh,1998). Differentiation of the NB7-3 lineage is also affected by mutations in wingless (wg) and other members of the Wingless signaling pathway such as, engrailed (en), hedgehog(hh) and patched (ptc)(Patel et al., 1989; Lundell et al., 1996). In addition, mutations in the transcription factors eagle (eg)(Dittrich et al., 1997; Lundell and Hirsh, 1998) and huckebein (hkb) (Lundell et al., 1996) also disrupt NB7-3 differentiation. eg has been shown to suppress a rough eye phenotype caused by the overexpression of Ras1, suggesting that eg may be involved in Ras signaling(Hay et al., 1997). hkb is regulated by both en and hh in the NB7-3 lineage (Lundell et al., 1996; McDonald and Doe, 1997). The exact relationship between these genes and signal transduction pathways within the NB7-3 lineage remains to be determined.

In this study, we examine the role of Numb/Notch signaling, in the differentiation of the NB7-3 lineage. Numb is a membrane-associated protein that has been shown to partition asymmetrically with one daughter cell during binary mitotic divisions and to be responsible for establishing alternative cell fates in the Drosophila CNS, PNS (reviewed by Lu et al., 2000; Cayouette and Raff, 2002; Skeath and Thor, 2003) and myogenic precursor cells (Ward and Skeath,2000). Numb has also been suggested to have a role in the differentiation of CNS lineages in vertebrates (reviewed by Lu et al., 2000; Cayouette and Raff, 2002; Shen et al., 2002; Sommer and Rao, 2002). Numb affects cell fate by inhibiting the intercellular Notch signaling pathway. During cell division, the cell that inherits Numb will lack Notch signaling,while the sibling cell that does not receive Numb will maintain Notch signaling (reviewed by Lu et al.,2000; Cayouette and Raff,2002; Sommer and Rao,2002; Skeath and Thor,2003). The mechanism for how Numb inhibits Notch signaling is still under investigation, but Numb binds to the intercellular domain of Notch(Frise et al., 1996) and may regulate α-Adaptin, a protein involved in receptor-mediated endocytosis(Berdnik et al., 2002).

We present evidence that Numb/Notch signaling regulates apoptosis within the NB7-3 lineage. Notch signaling has previously been associated with apoptosis in an array of both negative and positive effects. Notch has been shown to inhibit apoptosis in T and B lymphocytes(Pear et al., 1996; Deftos et al., 1998; Jehn et al., 1999; Shelly et al., 1999; Morimura et al., 2000; Jundt et al., 2002). Constitutive Notch activity has been associated with cell transformation and cancer, presumably by inhibiting apoptosis (reviewed by Artavanis-Tsakonas et al.,1999; Miele and Osborne,1999; Mumm and Kopan,2000). Notch has also been implicated in inhibiting apoptosis in the brain. In the cerebellum of mice, Notch1 mediates the onset of neurogenesis and prevents apoptosis of neuroepithelium(Lütolf et al., 2002). A number of investigations have linked neurodegeneration in Alzheimer's disease to a reduction in Notch signaling (reviewed by Miele and Osborne, 1999). Presenilins, which are often mutated in individuals with Alzheimer's disease,are also necessary for Notch processing. It has been proposed that a loss in presenilin function may lead to a decrease in Notch signaling and a subsequent increase in cell death. Furthermore it has been shown in Drosophilathat constitutive expression of Notch can inhibit presenilin-induced apoptosis(Ye and Fortini, 1999). Thus Notch signaling may protect neurons from apoptotic cell death.

By contrast, Notch has also been shown to induce apoptosis during development of the Drosophila retina(Cagan and Ready, 1989; Wolff and Ready, 1991; Rusconi et al., 2000; Yu et al., 2002), wing(Milan et al., 2002) and PNS(Orgogozo et al., 2002). Constitutive expression of Notch in zebrafish causes cells of the developing retina to enter apoptosis (Scheer et al.,2001). Notch has also been shown to induce apoptosis in vertebrate neural crest cells (Maynard et al.,2000). Recently, several studies have found a reduction in Notch activity to be coincident with cell proliferation and cancer(Verdi et al., 1999; Wakamatsu et al., 1999; Talora et al., 2002),suggesting that Notch-induced apoptosis may be an important mechanism in regulating cell growth. In this study, we demonstrate that during terminal mitotic divisions of the Drosophila CNS Notch signaling can induce an apoptotic cell fate.

eg²⁸⁹ and eg^mz360 were both provided by G. Technau. eg^mz360 contains a gal-4 P-element insertion 5′ of eg. eg²⁸⁹ contains a lacZP-element insertion 5′ of eg and was generated by B. Genisch and G. Korge. numb¹, spdo^G104 and numb¹: spdo^G104 were provided by C. Doe. UAS-Notch^ACT was provided by Y. N. Jan. Df(3L)H99,en-gal4 and UAS-p35 were obtained from the Bloomington Stock Center.

Dissected CNS were fixed in 4% paraformaldehyde and incubated with primary and secondary antisera as previously described(Lundell and Hirsh, 1994). Primary antibodies used were: rabbit anti-eg (1:2000, G. Technau), guinea pig anti-Hb (1:600, C. Doe), rabbit anti-Ddc (1:50, M. Lundell), rat anti-Ddc(1:30, J. Hirsh), rabbit anti-corazonin (1:200, C. Doe), rabbit anti-zfh-1(1:400, R. Lehman), rat anti-zfh-2 (1:200, A. Tomlinson), rabbit anti-pdm1(1:1000, T. Dick), guinea pig anti-lacZ (1:1000, M. Lundell) and rabbit anti-lacZ (1:1000, Cappel). All Texas Red, FITC and Cy5 secondary antibodies were from Jackson Laboratories and used at 1:200. Images were obtained using a BioRad 1024 laser-scanning microscope and processed with Adobe Photoshop.

Apoptosis was assayed by TUNEL using a TACS 2 TDT-Fluor In Situ Kit from Trevigen. To reduce background the Strep-Fluor solution was preabsorbed against embryonic tissue and used at a 1:400 dilution. After the TUNEL procedure, tissue was washed with an Avidin/Biotin blocking kit from Vector.

All four neuronal progeny of the NB7-3 lineage can be uniquely identified with the specific molecular markers shown in Fig. 1. Several previous investigations have contributed to this knowledge(Lundell and Hirsh, 1992; Bossing et al., 1996; Higashijima et al., 1996; Dittrich et al., 1997; Lundell and Hirsh, 1998; Schmid et al., 1999; Isshiki et al., 2001; Novotny et al., 2002). NB7-3 begins division in late stage 11 and produces four progeny that are all positive for Eg expression (Fig. 1B,C). GW and EW1, the progeny of GMC-1, are distinguishable by their expression of Hb (purple cells in Fig. 1B). The GW motoneuron is usually the most posterior cell in the cluster, often reduced in size and shows expression of Zfh-1 (purple cells in Fig. 1C). EW2 and EW3, the progeny of GMC-2 and GMC-3, express Zfh-2 (orange cells in Fig. 1B,C). EW1 and EW2 produce serotonin and are detectable with antibodies against either Dopa decarboxylase(Ddc) (Fig. 1D-F) or serotonin(data not shown). The two serotonergic cells are distinguishable from each other owing to their birth order, EW1 is more medial and expresses Hb, whereas EW2 is more lateral and expresses Zfh-2(Fig. 1D). EW3 is usually the most lateral cell and is unique in its production of the neuropeptide corazonin (Fig. 1E). At stage 15, Pdm1 has been reported to be expressed in several different progeny(Isshiki et al., 2001; Novotny et al., 2002), but by stage 17, its expression is limited to EW2(Fig. 1F)(Lundell and Hirsh, 1998). Not all of these markers are expressed simultaneously during development. Eg expression begins to decline just as Ddc and corazonin become detectable. Eg-lacZ, Hb, Zfh-2 and Pdm1 expression persists to stage 17 and into larval stages. By stage 17 the GW neuron becomes undetectable with Eg-lacZ (Fig. 1E,F).

As the GMC divisions of NB7-3 appear to be asymmetric, we asked whether Numb might have a role in specifying cell fates within this lineage. We found that in the mutant allele numb¹, the development of EW2 and EW3 was dramatically altered, but there was only a mild effect on the development of EW1 (Fig. 2). Comparison of stage 17 CNS immunoassayed for Ddc/Zfh-2/Hb from wild-type(Fig. 2A,A′) and numb¹ (Fig. 2B,B′) embryos showed that the number of immunoreactive Ddc cells decreased from two per hemisegment in a wild-type CNS to one per hemisegment in a numb¹ CNS(Table 1C). The Ddc cells that remained in numb¹ mutants were immunoreactive for Hb (95%, n=65), indicating they were EW1 cells(Fig. 2B). The EW1 cell was detectable in most hemisegments, but in 7% of the numb¹mutant hemisegments there were no detectable Ddc cells(Fig. 2D, Fig. 4D and Table 1C). Comparison of stage 17 CNS immunoassayed for Ddc/Eg-lacZ/corazonin from wild-type(Fig. 2C) and numb¹ (Fig. 2D) embryos showed that no corazonin-positive EW3 cells were detectable in the numb¹ mutant(Table 1D). Not only were EW2 and EW3 undetectable with Ddc and corazonin in a numb¹mutant, but they also failed to express Eg-lacZ at stage 17. In a wild-type CNS three cells per hemisegment were detected with Eg-lacZ,whereas in a numb¹ mutant CNS only one cell per hemisegment was detected with Eg-lacZ(Fig. 2C,D and Table 1E). These data indicate that in a numb¹ mutant development of the EW1 neuron proceeds normally, but the development of the EW2 and EW3 neurons is significantly altered.

Previous studies have shown that numb mutants alter cell fates but do not usually affect cell divisions. To examine whether NB7-3 divides normally in a numb¹ mutant we examined the lineage in stage 15 embryos (Fig. 3). In a wild-type stage 15 hemisegment there were four Eg-positive cells(Fig. 3A). In a numb¹ mutant 66% of the hemisegments had only two Eg-positive progeny that were both Hb positive, the EW1 and GW neurons(Fig. 3B and Table 1A). In 22% of the hemisegments there was a third Eg positive cell that expressed reduced levels of Zfh-2 (Fig. 3C and Table 1A). This cell must be a derivative of GMC-2 or GMC-3, however, by stage 17 only 3% of hemisegments had an EW2 cell (Table 1C) and 0%of hemisegments had an EW3 cell (Table 1D). These results suggest that in the numb¹mutant at least some or possibly all divisions of GMC-2 and GMC-3 do occur,but the progeny of these divisions quickly take an alternate fate that renders them undetectable by several different molecular markers.

As mentioned previously, 7% of numb¹ hemisegments had no detectable Ddc cells at stage 17 (Table 1C). Interestingly, at stage 15, 10% of numb¹hemisegments show two Eg cells that both express Zfh-1(Fig. 3J and Table 1B). Wild-type hemisegments and most numb¹ hemisegments only have one Zfh-1-positive cell, the GW neuron (Fig. 3H,I and Table 1B).

This suggests that in a numb¹ mutant, division of GMC-1 may sometimes produce two cells with the GW fate, rather than a single GW and EW1 cell. This 10% transformation in cell fate to two GW cells, may account for the 7% of hemisegments that showed no detectable EW1 cell with Ddc immunoreactivity. Taken together, these results indicate that the differentiation of GMC-1 progeny is mostly independent of Numb function,whereas Numb is essential for the differentiation of GMC-2 and GMC-3 progeny.

In a wild-type CNS, there is a pair of serotonergic neurons within each hemisegment, with two exceptions: in A8, the most posterior abdominal segment,there is a single serotonergic cell; and in T1, the most anterior thoracic segment, there is a triplet of serotonergic cells(Fig. 4A)(Valles and White, 1988). We found that in a numb¹ mutant there is a change in the identity of the A8 serotonergic cell. In a wild-type animal the single cell in A8 expressed Zfh-2, indicating that it has characteristics similar to EW2 cells (Fig. 4B). In a numb¹ mutant, there was also a single cell in A8, but it expressed Hb similar to an EW1 cell (Fig. 4E) and similar to the more anterior segments of a numb¹ mutant (Fig. 2B).

Examination of A8 in a wild-type embryo, at stage 15, showed three Eg immunoreactive cells. Two of the cells expressed Hb and one expressed Zfh-2(Fig. 4C). The Zfh-2 expressing cell presumably develops into the Ddc-expressing neuron. EW3 cells were not detected with corazonin in either segments 7 or 8(Fig. 5A), suggesting that GMC-3 does not form in these segments. These results suggest that although GMC-1 and GMC-2 both produce progeny in A8, only EW2 develops into a serotonergic neuron and the development of EW1 into a serotonergic neuron is suppressed.

Examination of A8 in a numb¹ embryo, at stage 15, also had three Eg immunoreactive cells. Hb expression was unaltered in two of the cells, but Zfh-2 expression was dramatically reduced in the third(Fig. 4F). These results suggest that the loss of Numb function in A8 prevents EW2 from developing into a mature serotonergic cell, just as it does in the more anterior segments. Furthermore, the loss of Numb function apparently induces GMC-1 to produce a Ddc expressing cell in A8 when it normally would not. This suggests that the EW1 fate is a default state in all hemisegments of a numb¹mutant.

As Numb has previously been shown to inhibit Notch signaling, we next asked whether ectopic expression of Notch within the NB7-3 lineage would produce a similar phenotype to the numb¹ mutant. Ectopic expression of Notch was achieved using the UAS/Gal4 system(Brand and Perrimon, 1993),where the expression of a UAS-Notch^ACT transgene was driven by an eg-gal4 transgene. The extracellular domain of Notch is absent in the Notch^ACT allele, thus this allele provides constitutive Notch activity (Doherty et al., 1996). In UAS-Notch^ACT embryos and larvae, very few Ddc- and corazonin-containing cells were detected(Fig. 5C and Table 1C,D). This phenotype is extremely penetrant, most CNS show no serotonergic or corazonin-containing cells, the example presented in Fig. 5C is an exception. Some of these flies actually eclosed as adults, but died within several days. This phenotype is similar to numb¹ in that EW2 and EW3 are undetectable, but is different in that EW1 is also undetectable.

The number of Eg immunoreactive cells detectable in the UAS-Notch^ACT allele at stage 15 was variable(Fig. 3D,K; Table 1A). Four percent of hemisegments showed four Eg-positive cells, with the normal distribution of Hb and Zfh-2 expression (Fig. 3D). Fifteen percent of hemisegments showed three Eg-positive cells, but most hemisegments showed two Eg-positive cells(Fig. 3D). This variable number of Eg immunoreactive cells suggests that the NB7-3 lineage divisions occur when Notch is ectopically expressed, but that the cells quickly follow an alternative fate that makes them undetectable, similar to the numb¹ phenotype. Interestingly, 89% of the UAS-Notch^ACT hemisegments had two cells that were immunoreactive for Zfh-1 (Fig. 3K and Table 1B). This suggests that most EW1 cells were converted to a GW cell fate in the UAS-Notch^ACT genotype. This would explain the inability to detect EW1 cells with Ddc immunoreactivity(Fig. 5C). This transformation in cell fate is similar to the 10% of hemisegments in the numb¹ mutant that had two Zfh-1-positive cells. Therefore in both the numb¹ and UAS-Notch^ACTgenotypes, the number of hemisegments that have no detectable Ddc immunoreactivity is proportional to the number of hemisegments that show two Zfh-1 immunoreactive cells (Table 1B-C).

These results suggest that Notch signaling must be inactivated within the NB7-3 lineage for normal development of EW1, EW2 and EW3. As the previous results show that Numb is required for the development of EW2 and EW3, Numb is most probably responsible for repressing Notch signaling in EW2 and EW3. However, as the effect of a numb¹ mutation on EW1 development is minor, some additional mechanism must be primarily responsible for repressing Notch signaling in EW1.

Because the ectopic expression of Notch produced a phenotype that inhibited development of the NB7-3 lineage, we next investigated the phenotype associated with a loss of Notch function in the NB7-3 lineage. To produce a loss of Notch function we used a mutant allele of sanpodo (spdo). spdo has been reported to be a tropomodulin-like molecule (Dye et al.,1998) but its functional capacity has not yet been shown. Previous experiments indicate that spdo mimics loss-of-function Notchmutations (Buescher et al.,1998; Dye et al.,1998; Park et al.,1998; Skeath and Doe,1998; Ward and Skeath,2000). It is advantageous to use a spdo mutation, instead of a Notch mutation, because spdo is not involved in Notch-mediated lateral inhibition during early neurogenesis and thus does not have the severe morphological defects of a Notch embryo(Salzberg et al., 1997).

In a homozygous spdo^G104 mutant embryo at stage 15, 50%of the hemisegments had one or two ectopic cells immunoreactive for Eg(Fig. 3E and Table 1A). The additional cells were also immunoreactive for Zfh-2 suggesting that they are derivatives of GMC-2 or GMC-3. After stage 16 the homozygous spdo^G104mutant CNS became quite fragile and difficult to dissect, nevertheless we were able to isolate fragments of stage 17 CNS from spdo^G104mutants. These CNS showed random ectopic expression of Ddc cells and altered morphology. In the position of the NB7-3 lineage the thoracic clusters showed numerous Ddc immunoreactive cells (Fig. 5D), and there were three to five Ddc immunoreactive cells in each abdominal cluster (Fig. 5E and Table 1C). These clusters of Ddc-expressing cells showed two cells that were immunoreactive for Hb, with the remaining cells expressing Zfh-2 (Fig. 5F). A wild-type CNS has only one Ddc cell that expresses Hb(Fig. 1D), suggesting that in the spdo^G104 mutant there is a transformation in cell fate such that, GMC-1 produces two EW1 cells, rather than a single EW1 cell and the GW cell. Corazonin immunoreactivity was poorly developed in spdo^G104 mutants, only 50% of the hemisegments had a corazonin-immunoreactive cell (Table 1D) and there was no evidence of ectopic corazonin-containing cells. This phenotype of ectopic Ddc cells in spdo^G104 is opposite to the absence of Ddc cells observed in numb¹ and UAS-Notch^ACT phenotypes.

We then asked whether the spdo^G104 allele could rescue the numb¹ mutant phenotype. Because the spdo^G104 homozygous phenotype is so severe, we attempted the rescue with just one copy of the spdo^G104 allele. The spdo^G104 heterozygote alone showed a phenotype that was similar to wild-type at stage 17 (Fig. 5G). When the spdo^G104 allele was combined with the homozygous numb¹ mutation we found evidence that EW2 and EW3 were partially rescued. 23% of hemisegments showed both EW1 and EW2 Ddc neurons and 7% of hemisegments showed an EW3 corazonin-containing neuron (Fig. 5H; Table 2). This is a substantial increase over the number of hemisegments in the numb¹mutant that have a wild-type pattern of EW neurons. The ability of spdo to rescue the numb¹ mutant phenotype demonstrates that spdo is epistatic to numb, which has been demonstrated in other lineages (Dye et al., 1998; Skeath and Doe,1998; Ward and Skeath,2000). These results imply that the function of Numb in the NB7-3 lineage is to inactivate Notch signaling during differentiation of EW2 and EW3 neurons.

We next asked whether ectopic expression of Numb would be sufficient to inhibit Notch signaling and produce a phenotype similar to the spdo^G104 mutation. Ectopic expression of Numb was achieved using a UAS-Numb transgene driven by an eg-gal4 transgene. At stage 15, UAS-Numb:eg-gal4 embryos had one or two ectopic cells immunoreactive for Eg in 42% of the hemisegments, similar to the spdo^G104 phenotype(Fig. 3F and Table 1A). However, at stage 17, we did not detect any ectopic Ddc or corazonin-containing cells using several different gal4 drivers to activate UAS-Numb (data not shown). Although UAS-Numb is sufficient to produce ectopic NB7-3 cells at stage 15, it is apparently insufficient to induce terminal differentiation of Ddc and corazonin-containing neurons. Taken together, the results presented thus far show that inhibiting Notch signaling with either spdo^G104 or UAS-Numb produces ectopic cells in the NB7-3 lineage, whereas enhancing Notch signaling with either numb¹ or UAS-Notch^ACT alters the fate of NB7-3 progeny so that they become undetectable with several different molecular markers.

Previous studies with the apoptosis deficient line H99 have shown that supernumerary Eg cells were produced within the NB7-3 lineage(Isshiki et al., 2001; Novotny et al., 2002). This result suggests that there are cells within the NB7-3 lineage that undergo apoptosis. The ectopic NB7-3 progeny we observe by inhibiting Notch signaling with spdo^G104 and UAS-Numb might be due to an inhibition of apoptosis and the rescue of cells in the NB7-3 lineage that normally undergo apoptosis. Conversely, the loss of NB7-3 progeny we observe by enhancing Notch signaling with numb¹ or UAS-Notch^ACT, might be due to Notch-mediated apoptosis.

To test this hypothesis, we first examined the NB7-3 lineage for apoptosis using TUNEL analysis, which measures DNA fragmentation, a hallmark of programmed cell death. In wild-type embryos we examined the occasional hemisegments that had five immunoreactive Eg cells rather than the standard four Eg cells. In these hemisegments we found that the extra Eg-positive cell undergoes apoptosis (Fig. 6A,A′). This indicates that during the mitotic divisions of the NB7-3 lineage some cells are eliminated by apoptosis. In numb¹ mutant embryos, we examined the hemisegments that had three immunoreactive Eg cells and found that the cell, which we previously showed to be Zfh-2 positive (Fig. 3C), underwent apoptosis (Fig. 6B). We also found a small fraction of numb¹hemisegments where the EW1 and GW neurons were apoptotic(Fig. 6C). It is likely that these hemisegments correspond to the 7% of numb¹hemisegments that have no detectable Ddc neurons at stage 17(Table 1C and Fig. 2D, Fig. 4D). Thus, a loss of numb function initiates novel apoptosis within the NB7-3 lineage.

The UAS-Notch^ACT:eg-gal4 mutant also induced apoptosis within the NB7-3 lineage. Fig. 6D,D′ show a hemisegment in which two Eg cells, that were not GMC-1 progeny, were TUNEL positive. Fig. 6E,E′ shows a hemisegment that had the two GMC-1 progeny and four closely associated TUNEL-positive cells. The four TUNEL cells are arranged in distinct pairs,suggesting they are mitotic siblings. These cells are most likely the GMC-2 and GMC-3 progeny. One of the four TUNEL cells is marked with Eg immunoreactivity, presumably apoptosis in the other three cells has either prevented or eliminated Eg expression. We did not find evidence of apoptosis in GMC-1 progeny of UAS-Notch^ACT embryos. Interestingly,we also found apoptosis in UAS-Numb:eg-gal4 embryos. Fig. 6F,F′ show a UAS-Numb hemisegment that had six cells immunoreactive for Eg, where one of the GMC-2/3 progeny is apoptotic. Apparently, expression of UAS-Numb initially allows the recovery of ectopic Eg cells, but these cells eventually undergo apoptosis. This would explain the lack of ectopic Ddc and corazonin-containing cells in stage 17 UAS-Numb embryos. Taken together, the results of our TUNEL analysis shows that apoptosis occurs in the wild-type NB7-3 lineage and that novel apoptosis can be induced by either a loss of numb function or the ectopic expression of Notch.

We next addressed whether inhibition of apoptosis could produce ectopic Ddc and corazonin-containing neurons. We first examined the apoptosis deficient line H99 (White et al.,1994), which had previously been shown to produce ectopic NB7-3 cells at stage 15 (Isshiki et al.,2001; Novotny et al.,2002). We found that H99 mutant embryos showed a wild-type pattern of expression for Ddc and corazonin at stage 17 (data not shown). Apparently, the extra cells produced at stage 15 in H99mutants are unable to mature into Ddc- and corazonin-containing neurons. Therefore, we tried another approach and examined the effect of ectopic expression of the apoptosis inhibitor, p35. Ectopic expression of p35 was achieved using a UAS-p35 transgene driven by an eg-gal4transgene. At stage 15, UAS-p35/eg-gal4 embryos produced ectopic Eg cells in 56% of all hemisegments(Fig. 3G; Table 1A). This phenotype is similar to H99 (Isshiki et al.,2001; Novotny et al.,2002), spdo^G104 and UAS-Numb embryos(Fig. 3E-G; Table 1A). At stage 17, UAS-p35/eg-gal4 embryos produced a pattern of Ddc and corazonin immunoreactivity that was mostly wild-type with occasional ectopic cells(Fig. 7A,D; Table 1C,D). Two corazonin-containing cells were found in 9% of the hemisegments instead of the normal one per hemisegment (Fig. 7A). Corazonin-containing cells were also found in A7, where corazonin is normally not expressed (data not shown). Triplets of Ddc-expressing cells were observed in 6% of the hemisegments(Fig. 7D). These ectopic cells suggest that p35 can, to a limited degree, rescue cells of the NB7-3 lineage that would normally undergo apoptosis.

We next examined whether ectopic expression of p35 could rescue the numb¹ mutant phenotype. We found partial rescue of the numb¹ phenotype (Fig. 7B); 28% of hemisegments showed at least two Ddc neurons and 34%of hemisegments showed at least one corazonin-containing neuron(Table 2). Interestingly, A8 showed two Ddc cells (Fig. 7B),presumably the EW1 cell was present due to the numb¹mutation and the EW2 cell was rescued with p35. As the numb¹ phenotype can be rescued both by the inhibition of apoptosis with p35 and by the inhibition of Notch signaling with spdo^G104, we hypothesize that in a wild-type animal Numb inhibits both Notch signaling and apoptosis in the NB7-3 lineage.

To test directly if Notch signaling induces apoptosis within the NB7-3 lineage, we examined whether inhibiting apoptosis with ectopic expression of p35 would rescue the phenotype produced by the ectopic expression of Notch. When both UAS-p35 and UAS-Notch^ACT were expressed simultaneously with eg-gal4, p35 was able to protect cells from Notch-induced apoptosis(Fig. 7C). However, only 18%(n=146) of larval CNS showed the rescued phenotype depicted in Fig. 7C, the majority of CNS appeared identical to the UAS-Notch^ACT phenotype alone(Fig. 5C). Of the larval CNS that showed rescue there were no patterns of partial expression(Table 2), all hemisegments showed the wild-type number of Ddc- and corazonin-expressing cells with occasional ectopic cells. This all-or-nothing phenotype suggests that there is some global trigger that determines whether UAS-Notch^ACTor UAS-p35 will have the dominant effect on development of the NB7-3 lineage. When en-gal4 was used to drive expression of both transgenes, 100% (n=14) of larval CNS showed the rescued wild-type pattern. eg expression fades during late embryogenesis, whereas en expression persists into larval stages. Therefore the extended expression of en-gal4 may maintain sufficient levels of p35 to prevent apoptosis, whereas declining expression of eg-gal4 may cause levels of p35 to drop such that Notch-induced apoptosis predominates.

Our results suggest that Notch-induced apoptosis is an essential mechanism for regulating cell fates within the CNS. In support of this hypothesis, we observed in heterozygous spdo^G104 larval CNS, tumor-like overproliferation of Ddc-expressing cells that protrude from the surface(Fig. 7E,E′). These areas of cell growth appeared in random positions throughout the CNS and varied in size and frequency. Ten percent of the CNS showed regions of cell proliferation as large as those in Fig. 7F and another 30% of the CNS showed smaller foci of cell proliferation. We do not know the origin of these cells, but because we have seen overproliferation in the brain lobes they cannot originate exclusively from the NB7-3 lineages. They may not even originate from Ddc-expressing lineages, as induction of Ddc expression is a characteristic of human small-cell lung carcinomas (Bepler et al.,1987). This observation that a reduction in Notch signaling can promote abnormal cell growth, suggests that Notch-induced apoptosis may have an important role in preventing overproliferation of cells during CNS development.

The results of this study demonstrate that the intercellular Notch signaling pathway can be modulated during terminal divisions of the CNS to direct a choice between neuronal development and programmed cell death. Fig. 8 summarizes our results and presents a model of how Numb/Notch signaling may be used to specify cell fates in the NB7-3 lineage.

The division of GMC-1 produces two distinct neuronal cell fates: the EW1 interneuron and the GW motoneuron. In this division, genetic alteration in the expression of Notch leads to switching between these two cell fates. A loss of Notch activity in spdo mutants leads to two Ddc/Hb-expressing EW1 cells and the overexpression of Notch leads to two Zfh-1 expressing GW cells. Therefore, Notch signaling must be inactivated during development of the EW1 neuron. Numb appears to have a minor role in this inactivation. In a numb¹ mutant, 7% of the hemisegments do not develop an EW1 neuron, and a similar number of numb¹hemisegments show two Zfh-1-expressing GW cells. This transformation from an EW1 cell fate to a GW cell fate is what one would expect if Numb were inhibiting Notch. However, most EW1 neurons develop normally in a numb¹ mutant and do not convert to the GW cell fate. Therefore, inactivation of Notch signaling in EW1 is mostly independent of Numb function. One possible explanation is that EW1 has a factor that is redundant for Numb function, which can inhibit Notch signaling and is capable of masking the effect of a numb¹ mutation in most hemisegments. The unique expression of Hb in GMC-1 progeny could be responsible for establishing this redundancy. However, if a redundant Numb-like factor does exist, it is insufficient to protect EW1 during expression of the UAS-Notch^ACT transgene.

The A8 segment is unique in that it only has a single serotonergic neuron instead of the pair of serotonergic neurons found in the more anterior segments. In a wild-type fly, this cell appears to be a derivative of GMC-2,because it expresses Zfh-2, but in a numb¹ mutant, this cell appears to be a derivative of GMC-1, because it expresses Hb. A single Hb/Ddc-expressing cell in A8 is identical to the phenotype in the more anterior segments of a numb¹ mutant. This suggests that in a numb¹ mutant an EW1 cell is the default developmental pathway for this lineage. One possibility is that the redundant Notch-inactivating mechanism we proposed for EW1 is only induced in the presence of a numb mutation. This would explain why the A8 EW1 cell is only seen in the numb mutant and not in a wild-type animal. If this were true, then the preservation of EW1 cells in a numb mutant would be due to the mutation itself. Until a putative redundant factor is identified it is impossible to determine whether it is expressed normally in wild-type animals or is expressed only in numb mutant animals.

Like GMC-1, most GMCs divide producing two progeny cells. However, GMC-2 and GMC-3 of the NB7-3 lineage only produce one neuron. It has previously been suggested that the mitotic sisters of EW2 and EW3 may undergo apoptosis(Isshiki et al., 2001; Novotny et al., 2002). This idea is supported by our detection of apoptotic cells with TUNEL in the wild-type NB7-3 lineage and our experiments with the apoptosis inhibitor p35,which can produce ectopic Ddc and corazonin-containing cells. The origin of the ectopic cells within NB 7-3 has not been formally determined by lineage tracing; however, the hypothesis that they are mitotic sisters of EW2 and EW3 is supported by the observations that GMC-2 and GMC-3 progeny often appear as mitotic pairs (Fig. 6E,F) and that ectopic NB7-3 cells are immunoreactive for Zfh-2(Fig. 3E-G).

During the divisions of GMC-2 and GMC-3, genetic alterations in the expression of Notch leads to a switching between a neuronal cell fate and apoptosis. A reduction of Notch signaling with either spdo^G104 or UAS-Numb embryos produces ectopic NB7-3 cells that express Zfh-2. Conversely, the overexpression of Notch in either UAS-Notch^ACT or numb¹ embryos led to an increase in TUNEL labeling of GMC-2 and GMC-3 progeny. Additionally,inhibiting apoptosis with UAS-p35 or reducing Notch activity with spdo^G104 can rescue the numb¹phenotype. We hypothesize that during the divisions of GMC-2 and GMC-3, Numb partitions asymmetrically into EW2 and EW3 where it inactivates Notch signaling and leads to neuronal development. The mitotic sisters of EW2 and EW3 do not receive Numb, maintain Notch signaling and undergo apoptosis. The difficulty in detecting wild-type hemisegments that have more than four immunoreactive Eg cells, suggests that any other cells produced during divisions of the NB7-3 lineage quickly undergo apoptosis.

Ectopic Eg cells in the NB7-3 lineage can be induced at stage 15 by H99 (Isshiki et al.,2001; Novotny et al.,2002), UAS-Numb, spdo^G104 and UAS-p35(Fig. 3E-G). However, the ability of these alleles to produce ectopic Ddc and corazonin-containing neurons at later stages is variable. We were unable to detect significant ectopic Ddc or corazonin-containing cells in either H99 or UAS-Numb CNS. We have shown for UAS-Numb that the ectopic Eg cells detected at stage 15 can undergo apoptosis. spdo^G104mutants produce only ectopic Ddc cells, but the reduction in the number of corazonin-containing cells in general suggests that either GMC-3 does not consistently form in these mutants or that GMC-3 progeny may convert from a corazonin-containing cell fate to a serotonergic cell fate. UAS-p35mutants produce both ectopic Ddc and corazonin-containing cells at low frequency, but the allele is much more efficient at rescuing the EW neurons in numb¹ and UAS-Notch mutants. Therefore, apoptosis is harder to reverse in cells that normally undergo apoptosis, than in the cells genetically induced to undergo apoptosis. The ability of these various alleles to produce ectopic Ddc- and corazonin-containing cells could be influenced by mutant affects they cause outside the NB7-3 lineage or may reflect different roles they have in the apoptotic pathway. The mechanism by which Notch induces apoptosis in the NB7-3 lineage remains to be determined,but the apoptotic genes reaper, grim and hid may be involved because all three of these genes are deleted in the H99 allele(White et al., 1994).

Notch-induced apoptosis in the NB7-3 lineage will probably be regulated by other factors in addition to Numb. The Ras signaling pathway has been shown to inhibit Notch-induced apoptosis in the Drosophila pupal retina(Cagan and Ready, 1989). Wingless has been shown to mediate Notch signaling(Axelrod et al., 1996; Brennan et al., 1999; Wesley, 1999; Lawrence et al., 2001; Ramain et al., 2001) and mutations in the Wingless pathway can lead to ectopic serotonergic cells(Patel et al., 1989). It will be a challenge to determine how these different signaling pathways interact to specify apoptosis within the NB7-3 lineage.

The tumor-like expansion of Ddc-expressing cells we observe in heterozygous spdo^G104 larvae suggests that Notch-induced apoptosis may be essential for regulating cell proliferation. This spdo phenotype is reminiscent of three tumor-suppressor genes; discs large(dlg), lethal giant larvae (lgl) and scribble (scrib), which produce tumors in the CNS and imaginal disks (Gateff, 1978; Manfruelli et al., 1996; Woods et al., 1996; Bilder et al., 2000). Interestingly, these three genes work in a common pathway that regulates cell polarity, and lgl and dlg have been shown to be essential for the distribution of Numb and other asymmetric determinants(Ohshiro et al., 2000; Peng et al., 2000). Further investigation will be necessary to determine if spdo is part of this same mechanism and exactly how spdo mutants inhibit Notch signaling. Spdo expression is ubiquitous throughout embryogenesis and persists through the larval stages and into adults (Dye et al., 1998). If a spdo mutation can alter the response of the Notch receptor to environmental cues that induce apoptosis, one would expect to see overproliferation in additional tissues.

Finally, although not discussed explicitly, our figures demonstrate that Numb/Notch signaling also affects development of the midline dopaminergic cells. The expression of Ddc is essential to the biosynthesis of both serotonin and dopamine. In the images presented, anti-Ddc antibody detects not only the serotonergic neurons, but also midline dopamine neurons. As a consequence of using Ddc as a marker for the serotonin lineage, a number of observations could be made about the development of midline dopamine cells. We found that in a numb¹ mutant very few midline dopamine cells were detectable with Ddc (Fig. 2, Fig. 4D, Fig. 5B). spdo^G104 mutants produce ectopic dopamine cells (data not shown) and can rescue dopamine cells in the numb¹ mutant phenotype (Fig. 5H). Thus,Numb/Notch signaling also has a role in the development of midline dopamine cells, but further investigation into the significance and whether apoptosis is involved in this lineage will require lineage analysis to determine the origin of the midline dopamine cells. Neither eg(Fig. 2C) nor en(Lundell et al., 1996) is expressed in midline dopamine cells. As the UAS alleles in this paper were induced with either eg-gal4 or en-gal4, the phenotype of dopamine cells in these experiments was not affected by these genetic backgrounds.

This work was supported by NIH grant RR 13646 to M.J.L. and a NIH MBRS Rise Fellowship GM 60655 to E.P. We thank J. Urban, G. Technau, C. Doe, Y. N. Jan and K. Matthews of the Bloomington stock center for provision of fly stocks. We thank J. Urban, G. Technau, C. Doe, A. Thomlinson, J. Hirsh and R. Lehman for provision of antibodies. We thank J. Sunter for helping with the DIC image. We thank R. Karcavich and C. Doe for communicating unpublished results,and E. Wheeler and C. Cronmiller for their comments on the manuscript.