The differentiation of the serotonergic neurons in the Drosophila ventral nerve cord depends on the combined function of the zinc finger proteins Eagle and Huckebein

The insect central nervous system (CNS) develops from neuroblasts (NBs) which delaminate from the neuroectoderm following gastrulation. NBs divide asymmetrically to produce a chain of smaller ganglion mother cells (GMCs); each GMC divides once to give rise to a pair of neurons and/or glia. The mechanisms regulating NB formation are quite well understood (for review, see Campos-Ortega, 1993), but the mechanisms leading to the specification of the individual NBs and the differentiation of their lineages are less well known. Each NB can be uniquely identified by its time of formation, position of delamination and pattern of gene expression (Doe, 1992; Broadus et al., 1995). In addition each NB produces a characteristic set of neuronal and/or glial progeny (Udolph et al., 1993; Bossing et al., 1996b). Laser ablation experiments (Doe and Goodman, 1985), in vitro NB cell culture (Huff et al., 1989; Lüer and Technau, 1992) and cell transplantation studies (Udolph et al., 1995) suggest that the lineages are largely controlled by intrinsic properties of the respective NBs. To clarify processes governing cell fates in the CNS, it is necessary to identify the genes employed and to understand their functional relationships.

This paper focusses on the NB 7-3 lineage in the Drosophila ventral nerve cord. Single cell labelings with DiI show that it is a very small lineage consisting of only 4 to 5 neurons at the end of embryogenesis (Bossing et al., 1996b), and 2 of these cells are likely to represent the ventral nerve cord serotonin cells (Lundell et al., 1996), comparable to the situation in the grasshopper (Taghert and Goodman, 1984).

To identify genes that might be involved in the specification of these cells, we screened for enhancer trap lines and found two, Mz360 and P289, with specific expression of the reporter gene in the NB 7-3 lineage. Both lines were found to be insertions at the eagle (eg) locus. eg encodes a putative zinc finger protein and the eg RNA is expressed in NB 7-3 as well as in three other NB lineages (Higashijima et al., 1996). Analysis of eg mutant embryos revealed impaired axogenesis in the putative progeny of NB 7-3 and NB 2-4, suggesting that eg plays a critical role in the development of these cells (Higashijima et al., 1996).

We find that Eagle (Eg) protein is expressed in NB 7-3 just after delamination and is present in all NB 7-3 progeny, including the serotonin neurons, until late stage 17. Mutant phenotypes of eg demonstrate that this gene is required for correct axonal pathfinding, for the expression of serotonin and, in the thoracic NB 7-3 lineage, for the generation of the correct cell number. Furthermore, eg is not only necessary for correct serotonin expression but is also able to promote ectopic serotonergic differentiation of specific neurons at defined positions in the Drosophila CNS. Loss-of-function alleles of huckebein (hkb) and engrailed (en) lead to a nearly complete loss of serotonin immunoreactivity and suggest an important role for these two genes in NB 7-3 development (Lundell et al., 1996). We show that en function in NB 7-3 development is upstream of eg and is necessary for eg expression in this lineage. On the contrary, the ectopic serotonergic neurons do not derive from en-expressing NB(s) but do depend on hkb. This led us to conclude that Eg and Hkb are part of a ‘combinatorial code’ of transcription factors acting in the NB 7-3 lineage to control serotonergic differentiation while en function is required earlier for the activation of eg expression.

The Gal4 P-element construct of Brand and Perrimon (1993) was used to generate more than 2000 independent enhancer-trap lines, including Mz360 (Ito et al., 1995), a homozygous viable fly strain, which shows an eg phenotype and failed to complement eg¹, eg² and eg^spy obtained from Indiana stock center. Therefore, we refer to this line as eg^Mz360. eg^P289, an enhancer-trap line bearing a P-larB P-element construct (Wilson et al., 1989) in the eg locus, was kindly provided by B. Genisch and G. Korge, Berlin.

In order to obtain an eg deficiency, we remobilised the P-element of eg^Mz360 by using the P[ry, Δ 2-3] chromosome (Robertson et al., 1980). We tested over 1000 excision events for lethality and non-complementation with eg^Mz360 and obtained one line, eg^18B, that fullfilled both criteria.

The genomic DNA fragments flanking the Mz360 and P289 insertions were cloned by the plasmid rescue technique (Bier et al., 1989). Southern hybridisation (Sambrook et al., 1989) was used for mapping the flanking DNA of the P-element insertions and to characterise the deletion of eg^18B.

DiI labeling of individual progenitors, in vivo identification of NBs and photoconversion of labelled clones was done as described elsewhere (Bossing and Technau, 1994; Bossing et al., 1996b).

For protein expression and preparation, we made use of the QIAexpress system (Quiagen). A SacI fragment from Exon C of the eg coding sequence (Higashijima et. al., 1996) was cloned into the SacI site of pQE 31. After induction, the expressed protein was prepared under non-denaturating conditions as described in the QIAexpress manual (Quiagen). Immunisation was done in rabbits by the Eurogentec company, using a standard protocol. The obtained serum was tested in embryos for specific eg staining and used without further purification.

An eg RACE-type 1 clone (Higashijima et al., 1996), containing the A1 and B exon of the eg gene was cut with HindIII and NaeI and the fragment consisting of exon A1 and B was cloned into HindIII/NaeI-restricted H 3.7. This genomic clone contains exon B and exon C of the eg gene (Higashijima et al., 1996). The resulting construct was cut with KpnI and HindIII and the fragment containing the entire eg coding sequence was inserted into a KpnI/HindIII-restricted pBlue-script II KS-vector. In a final step, this construct was cut with XbaI and the eg sequences were cloned into the XbaI site of p-UAST (Brand and Perrimon, 1993). Transgenic flies were generated, using standard techniques (Spradling and Rubin, 1982). Several insertions were isolated and tested for eg expression when crossed to sca-Gal4. The lines UAS-eg18.1 and UAS-eg43.1 were chosen for further experiments.

For sca-GAL4/UAS-eg expression in the hkb²-mutant background, we crossed the lines UAS-eg43.1; hkb²/TM3, Sb and sca-GAL4; hkb²/TM6b, Tb. The hkb mutants were identified by their head involution defect.

Whole-mount in situ hybridization was done according to Tautz and Pfeifle (1989). Antibody stainings in Drosophila embryos against β-galactosidase and En were performed as previously described (Schmidt-Ott and Technau, 1992). For anti-Eg antibody staining, embryos were incubated overnight at room temperature in a 1:500 dilution of preabsorbed anti-Eg antiserum in PBT (0.3% Triton X-100, 10% calf serum in PBS), washed with PBT and then incubated for 2 hours at room temperature with the HRP- or rhodamine-conjugated goat anti-rabbit antibody (Dianova) diluted 1:400 in PBT. The HRP was detected using 0.5 mg/ml DAB and 0.03% hydrogen peroxide. Embryos were mounted as flat preparations on coverslips in 90% glycerol in PBS. HRP-stained embryos were viewed under a Zeiss Axiophot microscope.

DDC, TH, β-galactosidase and serotonin staining of first instar larval nervous systems was done according to Beall and Hirsh (1987). The antibodies were used in the following dilutions: polyclonal rabbit anti-DDC (Scholnick et al., 1991) 1:100, polyclonal rabbit anti-TH (Pel Freeze) 1:50, monoclonal mouse anti-β-galactosidase (Promega) 1:1000 and monoclonal rat anti-serotonin (Accurate) 1:50. As secondary antibodies, we used FITC-conjugated goat anti-rat, FITC-conjugated goat anti-mouse and RITC-conjugated goat-anti rabbit (Boehringer) in a 1:400 dilution. Immunofluorescence was imaged on a Leica DM confocal microscope.

In Figs 1C-G, 3-7 we combined different focal planes by using Photoshop 3.0 (Adobe).

We identified the P-GawB (Brand and Perrimon, 1993) enhancer-trap line Mz 360 (Ito et al., 1995) and the lacZ enhancer-trap line P289 as alleles of eg and therefore refer to them as eg^Mz360 and eg^P289. The embryonic reporter gene expression patterns (Gal4, lacZ) of the two lines are identical. In abdominal segments, double labeling with anti-Invected antibodies revealed a reporter gene expression in four NBs and their putative progeny: NB 2-4, NB 3-3, NB 6-4 and NB 7-3 (Fig. 1A,B). This expression pattern is also described for the enhancer trap line P179 by Higashijima et al. (1996) which also represents an insertion in the eg locus.

In order to determine the distribution of the Eg protein, we generated a polyclonal antibody against Eg. Here we describe the Eg expression pattern in the abdominal neuromeres as revealed by this antibody. At early stage 11 (embryo staging according to Campos-Ortega and Hartenstein, 1985), when the S4 NBs have formed, Eg expression first appears in NB 2-4 and NB 3-3 (Fig. 1C). At late stage 11, NB 2-4, NB 3-3, NB 6-4 and NB 7-3 show Eg expression; NB 6-4 has already divided and there are several small Eg-positive cells located in the dorsal vicinity of NBs 2-4 and 3-3, representing their putative progeny (Fig. 1D). By the end of stage 12, Eg expression in the glial descendants of NB 6-4, the MM-CBG and M-CBG (terminology of glial cells according to Ito et al., 1995), is no longer detectable. At this stage, 3 to 4 Eg-expressing cells are visible at a more medial position in addition to 10 to 15 laterally located Eg-positive cells (Fig. 1E). By the end of stage 15, the lateral Eg-positive cells are hardly detectable, while the medial cells (called EW and GW neurons by Higashijima et al., 1996) still show strong Eg expression. This cell cluster shows a typical morphology with the GW neuron lying just posterior to the EW cells (Fig. 1F). Eg expression fades from the entire CNS at late stage 17 (Fig. 1G).

The lineage of NB 7-3 is described in Bossing et al. (1996b; see also Fig. 4A). At stage 17, most NB 7-3 clones consist of 4 mediodorsal neurons: three interneurons, called 7-3I neurons, which project contralaterally across the posterior commissure and one ipsilaterally projecting motoneuron, called 7-3M. Differentiation of 7-3M starts rather late (stage 15/16). At stage 17 the projection of 7-3M reaches the posterior root of the intersegmental nerve. The same projection pattern is described for the EW and GW neurons (Higashijima et al., 1996). Based on cell position, cell number and projection pattern, we conclude that EW neurons and 7-3I neurons, on the one hand, and the GW neuron and the 7-3M cell, on the other hand, are identical. Therefore, combining lineage data and the Eg expression pattern, we show that Eg is expressed in NB 7-3 just after delamination and is present in all NB 7-3 progeny (7-3I; 7-3M) until late stage 17.

In the grasshopper, the serotonergic neurons represent contralaterally projecting interneurons which are progeny of NB 7-3, resulting from the division of the first ganglion mother cell (GMC) (Taghert and Goodman, 1984). In Drosophila, the serotonergic neurons of the ventral nerve cord also project contralaterally across the posterior commissure (White and Vallés, 1985; Vallés and White, 1986; Lundell and Hirsh, 1994), and it has recently been shown that the hkb and en genes are coexpressed uniquely in the serotonin cells and in NB 7-3. This suggests that, as in the grashopper, NB 7-3 is the progenitor of the serotonergic neurons (Lundell et al., 1996). Double labeling for β-gal and serotonin in the CNS of eg^Mz360-GAL4/UAS-lacZ first instar larvae reveals colocalisation in two of the NB 7-3 progeny (data not shown). This confirms that the serotonin neurons are indeed among the progeny of the eg-positive NB 7-3.

In order to investigate the function of eg within the NB 7-3 lineage, we analysed fly strains carrying mutations at the eg locus. The P-element insertions of eg^Mz360 and eg^P289 are homozygous viable and cause, as shown by P-element remobilisation, a ‘wings held out’ phenotype, which is allelic to classical eg mutations (Lindsley and Zimm, 1992). We cloned the flanking regions of the P-elements by plasmid rescue and mapped the insertions at the molecular level. The P-elements are inserted only about 100 bp apart from each other and approximately 200 bp upstream of 5′-Exon A2 (Higashijima et al., 1996) of the eg gene (Fig. 2). After remobilisation of the P-element in eg^Mz360, we isolated eg^18B, a homozygous lethal line that is allelic to eg^Mz360 and other eg mutations. Southern hybridisation indicated that in this line at least 7 kb of genomic DNA of the eg locus is deleted, including sequences of exons A1, A2 and B bearing the translational start site (Higashijima et al., 1996; Fig. 2). This suggests that eg^18B is a null mutation for eg. In support of this, the analysis of the homozygous mutant embryos of eg^18B revealed no detectable expression of Eg protein. This was also true for eg^P289, suggesting that the P-element insertion in eg^P289 alone causes a severe embryonic loss-of-function mutation.

In order to analyse in detail the NB 7-3 development in an eg mutant background, we used the lipophilic fluorescent tracer DiI (Bossing and Technau, 1994) to label NB 7-3 and its progeny in eg^P289. As already mentioned, the wild-type NB 7-3 produces three interneurons (7-3I), projecting contralaterally across the posterior commissure and one motoneuron (7-3M) that develops an ipsilateral posterior axon (Bossing et al., 1996b; Fig. 3A,B). We generated three abdominal and two thoracic NB 7-3 clones and examined their cell numbers, positions and axonal projections (Fig. 3C-E). The abdominal clones correspond to wild type in cell number and positions and all show an ipsilateral posterior axon, suggesting normal 7-3M neurons. However, all mutant clones show severe pathfinding defects of the interneuronal projections. These projections no longer cross the midline, but stay ipsilaterally where the orientation of these projections is quite variable. In one case, the mutant cells project anteriorly and posteriorly (Fig. 3E) and, in the remaining cases, they project out of the CNS (Fig. 3C,D; thick arrows). Thus, NB 7-3 in eg^P289 embryos produces interneurons (7-3I) with significant axon pathfinding defects, whereas 7-3M seems to have a normal axonal projection. In contrast, Higashijima et al. (1996) described a lack of axonal outgrowth of NB7-3-derived neurons in the eg mutant using kinesin-lacZ driven by the eg enhancer as a marker. This discrepancy might be explained by the higher resolution of the DiI method.

In addition to the interneuronal pathfinding defects mentioned above, eg^P289 embryos have a thorax-specific abnormality in the NB 7-3 clones (Fig. 3C). The mutant clones consist of 6 to 8 cells, as compared to 4 to 5 cells in wild type. This suggests that eg is necessary to restrict the cell number in the thoracic NB 7-3 lineage.

Since the serotonergic neurons of the ventral nerve cord are among the progeny of NB 7-3, we further investigated serotonin expression within the eg mutant. In abdominal segments of a wild-type first instar larval CNS, there are two serotonergic neurons per hemineuromere, except for A8, where there is only one (White and Vallés, 1985; Vallés and White, 1986; Lundell and Hirsh, 1994; Fig. 4A). In eg^P289 and eg^18B mutants of the same stage, truncal serotonergic cells are almost entirely lacking (Fig. 4B,C). Only 2 to 4 escaper serotonergic cells are generally present and occur mostly as single cells in a random distribution. eg is therefore an important factor for serotonin expression in the NB 7-3 lineage, yet other genes must exist that are able to promote serotonin expression at a low frequency in the absence of eg function.

eg is an important factor for the serotonergic differentiation in the NB 7-3 lineage. However, is it able to promote an ectopic serotonergic cell fate? To answer this question, we induced ectopic eg expression using the GAL4 system (Brand and Perrimon, 1993). sca-GAL4 was used to drive GAL4 expression throughout the developing embryonic nervous system, reflecting the scaborous (sca) expression pattern (Mlodzik et al., 1990; Klaes et al., 1994). Consistent with this, eg expression, as revealed with the anti-Eg antiserum, was first detected as early as stage 10 in neuroectodermal cells and delaminating NBs. At later stages, it was present in most, if not all, NBs and their progeny until late stage 16.

First, the expression of various markers in the vNC of sca-GAL4/UAS-eg embryos was analysed. As revealed with the axon markers anti-Fasciclin II (Grenningloh et al., 1991), mAb22C10 (Fujita et al., 1982) and mAbBP102 (Patel, 1994; Seeger et al., 1993), the neuropile of these embryos is disorganised, indicating various pathfinding defects. The expression of en appears completely wild type (data not shown).

Despite these defects, most of the sca-GAL4/UAS-eg embryos hatch and therefore the CNS of the first larval instar could be analysed with an anti-serotonin antibody. In about 70% (n=100) of the hemineuromeres, we observe 1 to 2 ectopic serotonin-expressing cells (Fig. 5B). These es-cells (ectopic serotonergic cells) show neuronal cell morphology and occur at a reproducible, dorsolateral position. This suggests that ectopic serotonin expression does not occur randomly but in specific neurons of the CNS.

The position and number of the es-cells suggested that they could possibly be identical to the dorsolateral dopamine cells (DL-cells). In a wild-type larval CNS, there are 1 to 2 DL-cells expressing the enzymes DOPA decarboxylase (DDC) and tyrosine hydroxylase (TH) (Lundell and Hirsh, 1994). TH catalyses the first step of dopamine biosynthesis and is, within the Drosophila larval vNC, expressed only in the midline dopamine cell and the DL-cells (Levitt et al., 1965; Lundell and Hirsh, 1994). We examined serotonin, DDC and TH expression in the first larval instar CNS of sca-GAL4/UAS-eg animals either by double labeling with anti-serotonin and anti-DDC or with anti-serotonin and anti-TH antibodies (Fig. 5C-F). In both cases, we observed 2 serotonin cells and 1 to 2 escells per hemisegment expressing serotonin as well as DDC or TH. TH expression in serotonergic neurons is never observed in wild type. In both experiments, we found 1 to 2 dorsally located cells expressing only DDC or TH in addition to the serotonin and es-cells. Due to their position and number, these cells are most likely the DL-cells. Thus, although the ability of eg to induce the expression of both DDC and TH obscured the results of the double labeling experiments, these observations show that the es-cells are not identical to the DL-cells. The fact that we get ectopic TH expression in the serotonergic neurons must be due to higher levels of Eg protein in the NB 7-3 lineage, as the same result is observed by overexpression of eg using eg^Mz360-GAL4/UAS-eg (data not shown).

We asked further whether specification of es-neurons occurs at the level of the progenitor cells or at the level of the mature, differentiating cells. To induce ectopic eg expression only in postmitotic neurons, we used the elav-GAL4line to reflect the elav expression pattern (Robinow and White, 1988). Though eg expression could be evoked efficiently in elav-GAL4/UAS-eg embryos, we never observed ectopic serotonergic cells in the ventral nerve cord of elav-GAL4/UAS-eg first instar larvae (data not shown). This suggests that ectopic serotonergic expression in the sca-GAL4/UAS-eg line is due to the presence of Eg in NBs and/or GMCs.

At late embryonic stages, the serotonergic neurons are the only cells expressing en and hkb simultaneously, leading to the hypothesis that these genes act in combination to determine the serotonergic cell fate. In support of this, loss-of-function mutations in these genes lead to loss of serotonin immunoreactivity (Lundell et al., 1996). Thus, it is of interest to investigate the functional relationship between en, hkb and eg in order to clarify the serotonergic differentiation pathway within the NB 7-3 lineage.

To examine the functional relationship between eg and en a loss-of-function mutation, en^IM99 (Heemskerk et al., 1991), was stained with the anti-Eg antiserum (Fig. 6B). In about 90% of mutant hemineuromeres (n=100), we observed a complete loss of eg expression at the position of the NB 7-3 neurons. This is in agreement with results of Matsuzaki and Saigo (1996) who found a loss of Eg expression in 81% of hemineuromeres. In the remaining hemineuromeres, we find eg expression in all NB 7-3 progeny (7-3I; 7-3M). Due to their position and their strong Eg expression, these cells are clearly distinguishable from the lateral Eg-expressing neurons. Thus, en function in NB 7-3 development must be upstream of eg and is therefore necessary for correct eg expression within this lineage.

To address whether en could have a second function in NB 7-3 development, which is more directly related to the serotonin expression, we examined whether en expression is necessary for promoting the es-cell fate. Using the UAS/GAL4 system, eg was ectopically induced in the en expression domain. Although Eg expression in en-GAL4/UAS-eg embryos was strong, it failed to promote an es-cell fate (data not shown). Since the onset, intensity and perdurance of en-GAL4-driven UAS-eg expression is comparable to that in sca-GAL4, we conclude that en is not necessary to establish eg-dependent ectopic serotonergic differentiation. Consistent with this, invected (inv), which in the embryo is coexpressed with en, is not expressed in the es-neurons of sca-GAL4/UAS-eg first instar larvae (Fig. 7A). This is in contrast to the wild-type serotonergic cells, as they are inv positive at this developmental stage (Fig. 9A).

Taken together, these data indicate that en is not directly involved in the process of serotonergic differentiation, but acts further upstream by activating eg expression.

Another transcription factor important for NB 7-3 development is hkb. In hkb mutants, serotonin differentiation is severely disturbed (Lundell et al., 1996). Analogous to en, we wanted to elucidate the functional relationship between hkb and eg. We therefore stained hkb², a strong hypomorphic allele (Weigel et al., 1990), with anti-Eg antiserum. eg expression is observed in all NB 7-3 progeny (7-3I; 7-3M) in approximately 80% (n=100) of hkb² hemineuromeres. A complete loss of eg expression at the position of NB 7-3 neurons occurs in only 20% (n=100) of mutant hemisegments (Fig. 6C). This loss of eg expression cannot exclusively account for the observed serotonin phenotype in hkb mutants, since earlier studies revealed a loss of 80% of the ventral ganglion serotonin cells in hkb⁶², another hypomorphic hkb allele (Lundell et al., 1996). Thus, the majority of 7-3I cells in hkb embryos express eg but fail to exhibit serotonin expression.

To rule out the possibility that eg regulates hkb expression in the NB 7-3 lineage, we analysed hkb RNA expression in eg^P289. In early stage 11 wild-type embryos, hkb is expressed, among other cells, in a neuroectodermal cluster at the position where NB 7-3 will delaminate (Chu-LaGraff et al., 1995). At late stage 11, when the S5 NBs have formed, NB 7-3 is the posteriormost NB-expressing hkb (Chu-La Graff et al., 1995). This is also the case in eg^P289 embryos (data not shown), indicating that eg is not necessary for correct hkb expression in NB 7-3, but rather acts in parallel with hkb.

To investigate whether a combinatorial function of eg and hkb is generally necessary to promote serotonergic differentiation, ectopic eg expression was induced by sca-Gal4 in a hkb²-mutant background. This was designed to test whether the differentiation of es-neurons is also dependent on the function of hkb. For the analysis of the first instar larval CNS, we took advantage of the fact that the wild-type and ectopic serotonin neurons can be distinguished by their position, and that the es-neurons show no inv expression (Fig. 7A; see also previous section). This experiment revealed that approximately 80% of the wild-type and ectopic serotonin cells are absent (Fig. 7B). Previous work on hkb indicates a cell autonomous function of this gene (Chu-La Graff et al., 1995; Bossing et al., 1996a). Therefore this finding strongly suggests that the ectopic serotonergic neurons are generated by a hkb-positive NB and, furthermore, together with the data regarding the eg and hkb mutant phenotypes of the NB 7-3 lineage (see above), implies that the Eg and Hkb transcription factors are part of a combinatorial code of regulatory proteins necessary for promoting the serotonergic cell fate.

To investigate mechanisms controlling the development of a specific NB lineage, we have chosen the lineage of NB 7-3 as a model system. NB 7-3 has the advantage of generating a small number of progeny (4 to 5), two of which representing serotonergic neurons. Here we analysed the role of eg, a gene coding for a zinc finger protein (Higashijima et al., 1996), which is expressed in NB 7-3 and all its progeny.

Although the cell number of the NB 7-3 lineage is not tagma specific in wild-type embryos (Bossing et al., 1996b), we find a thorax-specific abnormality in NB 7-3 clones of eg embryos. The mutant clones in the thorax consist of 6 to 8 cells, as compared to only 4 to 5 cells in wild type. This suggests that there are differences in the proliferative capabilities within the thoracic and abdominal NB 7-3 lineages and the function of eg might be necessary to restrict these in thoracic neuromeres.

Besides this segment-specific phenotype, we observe striking defects in the differentiation of the 7-3I neurons in all truncal segments. These are pathfinding defects and lack of serotonin expression.

How does the lack of the eg gene lead to these differentiation defects? Since eg codes for a DNA-binding protein, one possible mechanism is that it directly regulates the genes whose products are involved in serotonin synthesis, pathfinding of the 7-3I fibres and/or restriction of cell proliferation in the thorax. This direct regulation is possible with respect to axonal guidance and NB proliferation, since Eg is expressed at the appropriate time, ie. during axogenesis and in the neuroblast after proliferation has ceased. However, it is not compatible with a direct function of eg in serotonin expression, since Eg protein fades in 7-3I neurons at the onset of DDC expression (this paper; Lundell and Hirsh, 1994) which catalyses the second step of serotonin biosynthesis. Thus, a more likely explanation for the differentiation defects observed in the mutant NB 7-3 lineage is that loss of eg leads to cell fate changes at an early stage. This is also supported by the observation that eg is able to promote ectopic serotonergic differentiation only after misexpression in progenitor cells.

Despite this abnormal development of the NB 7-3 lineage in eg embryos, certain features of the abdominal clones, like cell number, the average cell positions and the 7-3M projection appear normal. Together with the information about time and position of the NB delamination, these wild-type characteristics allow us to unambigously identify the mutant clones and indicate that loss of eg does not transform NB 7-3 into a different NB identity; instead, eg seems to be autonomously required for the execution of distinct differentiation features within the lineage. Similar observations have been made for the function of hkb in the NB 4-2 (Chu-LaGraff et al., 1995), NB 1-1 and NB 2-2 lineages (Bossing et al., 1996a).

en, eg and hkb are all expressed in the NB 7-3 lineage, and loss-of-function mutations in any of these genes lead to a nearly complete absence of serotonin differentiation (this paper; Lundell et al., 1996). What is the regulatory and/or functional relationship between these three genes in the differentiation of the serotonergic neurons?

In 90% of en mutant hemisegments, we observed a complete loss of eg expression at the position of the NB 7-3 neurons and this is the case throughout neurogenesis. Taking into account that three additional NB 7-3 lineage markers (Inv, hkb-lacZ and serotonin) are lacking in en embryos (Lundell et al., 1996), we conclude that, in most neuromeres of these embryos, NB 7-3 is unable to develop properly or is not formed. Recent results indicate that genes belonging to the segment polarity group have a function in neuroblast formation (Chu-LaGraff and Doe, 1993; Hartenstein et al., 1994; Skeath et al., 1995; Bhat, 1996; Matsuzaki and Saigo, 1996). Thus, it would not be surprising if en has similar functions in the generation of NB 7-3. However, it is also possible that NB 7-3 still delaminates in en embryos but fails to express any marker tested so far. In any case, loss of serotonin differentiation in en embryos is preceded by the loss of eg and probably also hkb expression within the NB 7-3 lineage.

Although en mutant embryos show a loss of eg-positive NB 7-3 in most hemisegments, it is still present in approximately 10% of the hemisegments, even in en/inv double mutants (see also Matsuzaki and Saigo, 1996). Therefore, there must be additional genes turning on eg expression in the absence of en/inv function. One of these genes might be hkb. We observe a complete loss of Eg expression at the position of NB 7-3 and its progeny in 20% of hkb mutant hemisegments. This loss of eg expression indicates an early function of hkb in NB 7-3 development consistent with the fact that hkb is already expressed in the neuroectodermal cluster at the position where NB 7-3 will delaminate (Chu-LaGraff et al., 1995). Additionally, a combined function of wingless (wg) and hedgehog (hh) has been shown to be necessary for the formation of NB 7-3 (Matsuzaki and Saigo, 1996) and this may also include turning on eg expression in NB 7-3.

The nearly complete loss of serotonin immunoreactivity in hkb embryos (Lundell et al., 1996), while having eg expression in NB 7-3 progeny in most hemisegments, shows that in the absence of hkb the majority of 7-3I cells are present but fail to express serotonin. This indicates that hkb also has a late function in NB 7-3 development and is, like eg, necessary for establishing serotonergic cell fate within its lineage (Fig. 8).

Upon early expression of Eg in the entire developing nervous system, 1 to 2 es-neurons were found in most hemisegments. This shows that eg is not only necessary but in some cases also sufficient for the specification of these cells. The fact that only a few cells are able to switch towards a serotonergic pathway suggests that Eg needs the function of additional factors to promote serotonergic differentiation. One of these factors is Hkb since severe reduction of es-neurons caused by eg overexpression is seen in the absence of hkb function. However, since there are nine NBs per hemineuromere expressing hkb, the coexpression of eg and hkb cannot be sufficient for triggering serotonergic cell fate. This is consistent with the fact that two of the eg-expressing NBs, NB 2-4 and NB 7-3, coexpress hkb (Doe, 1992) but only NB 7-3 generates the serotonergic neurons. Since the NB 7-3 lineage is the only lineage expressing en in addition to eg and hkb, En might be another factor necessary for es-neuron development. However, we have circumstantial evidence that this is not the case: unlike the endogenous serotonergic cells, the es-cells do not show anti-Inv immunoreactivity, an antibody that labels all en-expressing cells. In addition, ectopic expression of eg using en-GAL4 did not reveal any es-neurons.

Thus, although necessary for turning on eg expression and possibly also for NB formation, en function seems to be dispensable with respect to serotonergic differentiation, once the expression of the essential combination of transcription factors is established. Among these essential factors are the Eg and Hkb proteins. Currently we do not know how and at which level these two zinc finger proteins interact, but judging by our ectopic expression studies, it is very likely that both proteins are needed early in lineage development.

Therefore, we propose that eg and hkb represent a subset of NB and/or GMC identity genes necessary for the generation of serotonergic neurons. Upon overexpression of eg the ectopic combination of eg and hkb together with other unknown factors confers a ‘partial NB 7-3 lineage identity’ to another unidentified NB lineage. The function of these genes is downstream of segment polarity genes that are involved in NB determination in the neuroectoderm (Chu-La Graff et al., 1993; Skeath et al., 1995; Bhat, 1996; Matsuzaki and Saigo, 1996; see Fig. 8). As already mentioned, in case of NB 7-3, it was recently shown that wg and hh are redundantly required for the formation of this NB (Matsuzaki and Saigo, 1996). Another player in NB 7-3 formation is gooseberry (gsb). gsb is expressed in the row 6 neuroectoderm and seems to repress NB 7-3 cell fate in this region (Patel et al., 1989; Matsuzaki and Saigo, 1996).

The combined function of hh, wg and en leads to the expression of eg and probably other NB 7-3 identity genes like hkb, necessary for the development of serotonergic neurons. At present, we do not know which genes act further downstream to finally execute the expression of serotonin. However, good candidates exist that may be directly involved in the regulation of DDC, including two POU domain proteins (for review see: Herr et al., 1988; Ruvkun and Finney, 1991) and a homeodomain protein with a zinc finger motif. One of the POU domain proteins, Cf1a (Johnson and Hirsh, 1990), also called Drifter (Drf) or Ventral veinlet (Vvl) (Anderson et al., 1995; deCelis et al., 1995) was originally identified by its ability to bind to a DDC upstream regulating element. Consistent with its possible role in DDC regulation, Cf1a is expressed in a subset of specific neurons including the serotonergic neurons (Treacy et al., 1991). The second gene, coding for the alternatively spliced transcripts I-POU and twin of I-POU (Treacy et al., 1992), is expressed in a highly overlapping pattern within the nervous system again including the serotonin cells. This is also true for the third gene, zfh-2, which codes for a DNA-binding protein possessing three homeodomains and sixteen zinc fingers, and which was also shown to bind to the DDC promoter region (Lundell and Hirsh, 1992). Further work needs to be carried out to elucidate if these three genes are really involved in DDC regulation and if they are downstream of eg and/or hkb.

Besides DDC the biosynthesis of serotonin is carried out by a second enzyme: Tryptophan Hydroxylase (TPH) (Wright, 1987). Both enzymes are expressed in serotonergic as well as dopaminergic cells, although TPH is not needed for dopamine synthesis (Vallés and White, 1986; Beall and Hirsh, 1987; Konrad and Marsh, 1987; Neckameyer and White, 1992). In contrast to this, only dopaminergic cells express TH. Upon ectopic eg expression, we found colocalisation of serotonin and TH in some of the es-cells and in the cells identified as the wild-type serotonergic cells. This ectopic TH expression is unexpected, since coexpression of serotonin and TH is not obtained in wild type. Here the serotonergic neurons are not TH positive although they express eg and, in eg mutants, the TH expression is not affected. The fact that the expression of high levels of Eg protein within the eagle lineages using eg^Mz360-GAL4/UAS-eg also leads to ectopic TH expression in the serotonergic cells shows that this relies on a cell autonomous mechanism. It suggests that TH, DDC and probably also TPH share common regulatory mechanisms that will be activated as a consequence of high eg expression. Interestingly, the expression patterns of TH, TPH and DDC in the vNC are identical, with the exception of TH not being expressed in the serotonin cells (Beall and Hirsh, 1987; Neckamayer and White, 1992; Lundell and Hirsh, 1994). We therefore postulate a mechanism within the developing NB 7-3 lineage that leads to a selective block of TH expression in the serotonergic neurons and which can be overridden by a high level of Eg protein. According to this hypothesis eg would be able to promote the decision towards a serotonergic and aspects of dopaminergic cell fate, which would be further specified through the action of additional genes. However, the fact that the dopaminergic cells are unaffected in the eg mutant shows that there must be at least one other pathway leading to dopaminergic differentiation.

We are very grateful to R. White and T. Weaver for their help in the early stages of the project, to B. Genisch and G. Korge, who gave us the eg^P289 line, to K. Saigo, who sent us the eg cDNA prior to publication, and to J. Hirsh for discussions. We also would like to thank A. Brandt, U. Hinz, and K. White for the en-Gal4, sca-Gal4 and elav-Gal4 lines, S. di Nardo for the en^IM99/CyO, C.Q. Doe for the hkb²/TM3 stock, G. Brönner for the hkb cDNA, S. Scholnick (antiDDC) and C. Klämbt (anti-Inv, mAb22C10, mAbBP 102) for the antibodies, and G. Udolph, C. Rickert, L. Meadows and N. Deshpande for critical reading of the manuscript. This work was supported by a grant from the Deutsche Forschungsgemeinschaft to G. M. T and J. U. (Te 130/4-3).