eagle is required for the specification of serotonin neurons and other neuroblast 7-3 progeny in the Drosophila CNS

We are investigating the mechanism that specifies serotonin neurons in the ventral nerve cord of the fruit fly, Drosophila melanogaster. Serotonin is an evolutionarily conserved neurotransmitter that has roles in locomotion and behavior in both invertebrates (Carew, 1996; O’Gara et al., 1991; Segalat et al., 1995) and vertebrates (Grillner et al., 1995; Kandel et al., 1991). The general pattern of serotonin cells in segmented invertebrates is rather well conserved in animals from lobsters to insects, implying a conserved developmental pathway for these neurons. These cells are derived from a relatively simple lineage (Taghert and Goodman, 1984; Higashijima et al., 1996; Lundell et al., 1996; Dittrich et al., 1997) and are one of only a few neuronal types where the terminal differentiated phenotype is easy to assay. These characteristics make this lineage an excellent model for examining molecular mechanisms of cell specification.

Neuronal diversity in the insect CNS arises through invariant division of neuroblasts (NBs) and through environmental cues from neighboring neuronal and nonneuronal cells (Chu-LaGraff and Doe, 1993; Huff et al., 1989). 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. The Drosophila ventral nerve cord develops from stereotyped divisions of 30 NBs in each hemisegment, which form a two-dimensional array in which each NB has a unique position (Goodman and Doe, 1993). Recently we have shown that the serotonin neurons arise from NB 7-3 (Lundell et al., 1996). Within most segments of the Drosophila ventral cord are two bilaterally symmetric pairs of serotonin neurons (Fig. 1A).

The lineage of insect serotonin neurons was first examined in grasshopper by injecting NB 7-3 with the vital dye lucifer yellow (Taghert and Goodman, 1984). NB 7-3 in grasshopper produces four GMCs, the last of which degenerates. The paired serotonin neurons are derived from division of the first GMC. In the first thoracic segment, there are three serotonin neurons per hemisegment; this third cell originates from division of the second GMC. The fate of the other cells in this lineage are unknown. The axons of the serotonin neurons project anteriorly and cross to the contralateral side via the posterior commisure. Although the two sister serotonin cells are morphologically very similar they have slightly different growth rates and distinguishable projections. Given the similarities between the grasshopper and Drosophila NB 7-3 lineage, it is most likely that the serotonin neurons in Drosophila are also mitotic sisters.

Since direct injection of neuroblasts in Drosophila is more difficult, initial studies on the NB 7-3 lineage have been done using molecular markers. NB 7-3 expresses several genes including: engrailed (en), huckebein (hkb), seven-up (svp), pdm1 (also known as nubbin) and eagle (eg) (Broadus et al., 1995; Dick et al., 1991; Higashijima et al., 1996; W. Chia, personal communication). Most of these genes are expressed in many other NBs as well and, since NB 7-3 is one of the last neuroblasts to delaminate from the neuroepithelium, the overall CNS pattern of expression for these genes at the time that the NB7-3 lineage initiates is very complex. eg, however, is expressed in only four NBs, including NB 7-3 (Higashijima et al., 1996), making it a relatively simple marker for the NB 7-3 lineage.

Higashijima et al. (1996) show, with eg-lacZ enhancer traps and eg RNA in situ, that NB 7-3 produces four neurons in the abdominal region of the ventral cord. Three of the four neurons, the EW neurons, project anteriorly to the posterior commisure, and one, the GW neuron, projects ipsilateral and posteriorly. Recently these results have been confirmed with an eg antibody and DiI-labeled NB 7-3 clones (Dittrich et al., 1997). A direct relationship between the EW neurons and the serotonin neurons is demonstrated by Dittrich et al. (1997), which appeared while this manuscript was under review. Mutant alleles of eg show the correct number of eg-expressing NB 7-3 progeny but the neurons have abnormal projections (Higashijima et al., 1996; Dittrich et al., 1997). This suggests that eg has a role in neuronal specification rather than the lineage divisions of NB 7-3. The restriction of eg expression to just four neuronal lineages produces an eg mutant phenotype where the overall morphology of the CNS is well preserved.

The name eagle describes an adult phenotype of wings held out at right angles to the body (Fig. 1C), which was initially characterized in 1930 by Thomas Hunt Morgan. eg encodes a zinc finger protein with homology to the steroid receptor family (Higashijima et al., 1996; Rothe et al., 1989). Besides its limited expression in the CNS, the only other reported expression of eg is transient expression in the embryonic gonad (Rothe et al., 1989). Recently eg was isolated in a screen for suppressors of a rough eye phenotype caused by the overexpression of Ras1, suggesting that eg may be involved in the Ras1 signal transduction pathway (Hay et al., 1997).

Our initial interest in eg was to use an eg driven lacZ reporter construct as a simple marker of the serotonin neurons late in embryogenesis. These experiments turned out to be difficult but interesting because P-element insertions in eg produce an unusual differential effect in the two sister serotonin neurons that alters the expression pattern of several genes and disrupts serotonin synthesis. These changes in cell fate enable us to gain further insight into the developmental pathway of serotonin neurons and to present a model that uniquely identifies all progeny in the NB 7-3 lineage.

All wild-type flies are Oregon R. The eg^T6 and eg²⁸⁹ alleles contain lacZ P-element insertions 5’ of eg. eg^mz360 contains a gal-4 P-element insertion 5’ of eg. eg^225A and eg^18B are P-element excision of eg^mz360. eg^18B is a deficiency that removes the first two exons and the translational start of eg (Dittrich et al., 1997). All eg alleles were generously provided by J. Urban and G. Technau. eg^T6 was generated by R. White and eg²⁸⁹ was generated by B. Genisch and G. Korge.

Dissected CNSs were fixed in 4% paraformaldehyde and incubated with primary and secondary antisera as described previously (Lundell and Hirsh, 1994).

Rabbit and rat Ddc antiserum (Beall and Hirsh, 1987) were used at 20× dilution. Rat serotonin monoclonal antibody (Accurate Chemical) was used at 100× dilution. Rabbit Tyrosine Hydroxylase antiserum (Pel Freeze) was used at 40× dilution. Rabbit anti-β-galactosidase (Cappel) and mouse anti-β-galactosidase (Promega) were used at 1000× dilution. en immunoreactivity was detected with a mouse invected monoclonal antiserum 4D9 used at a 2× dilution (a gift from C. Doe). Rabbit pdm1 antiserum was used at 1000× dilution (a gift from W. Chia). Rat zfh-2 antiserum was used at 100× dilution (a gift from A. Tomlinson). All FITC secondary antibodies (Jackson) and all L-RITC antibodies (Tago) were used at a 200× dilution. All antibodies were preabsorbed against first-instar larval tissue.

Dissected larval CNS were incubated in PBS with 0.1 mM 5-hydroxytryptophan and 1.0 mM CaCl² for 30 minutes at room temperature prior to fixation.

All confocal projections were generated on a Molecular Dynamics-2000 or a Biorad MRC 1024 confocal laser scanning microscope. In Figs 1B and 4, the images were assembled from multiple projections along the anterior/posterior CNS axis using Photoshop (Adobe). In the black and white images shown, the contrast has been inverted for clarity.

To directly demonstrate the relationship between eg-expressing neurons and serotonin-synthesizing neurons, we used the allele eg²⁸⁹, which contains a lac-Z P-element inserted 5’ of eg. eg RNA is not detectable after 11 hours of development (Higashijima et al., 1996), and serotonin synthesis is not detectable until 16 hours (Lundell and Hirsh, 1994). In spite of this temporal gap, the perdurance of β-galactosidase from the lacZ reporter allows one to assess which cells expressed eg at an earlier time.

Serotonin neurons can be detected with either DOPA decarboxylase (Ddc) or serotonin immunoreactivity. Since Ddc catalyzes the last step in the biosynthesis of both serotonin and dopamine, a Ddc antibody will detect both cell types (Beall and Hirsh, 1987; Konrad and Marsh, 1987; Lundell and Hirsh, 1994; Valles and White, 1986). In the wild-type larval nerve cord, depicted in Fig. 1A, three types of Ddc-immunoreactive cells are distinguished in each segment: the paired ventral lateral serotonin cells (VL), a single midline dopamine cell (M) and two single dorsal lateral dopamine cells (DL). In the most posterior abdominal segment A8, the serotonin neurons are single cells instead of pairs and in the most anterior thoracic segment, T1, the serotonin neurons are triplets and there are a triplet of medial dopamine neurons instead of a single cell.

Double labeling for eg-lacZ and Ddc demonstrates that eg-lacZ is expressed equally in both serotonin neurons. This result is shown in Fig. 1D, which shows the abdominal segments of a first instar larval CNS heterozygous for eg²⁸⁹, labeled with nuclear localized β-galactosidase in red and Ddc in green. The expression of eg-lacZ in the entire larval CNS is shown in Fig. 1B. In addition to the serotonin cells, eg-lacZ is also expressed in far lateral clusters in the abdominal region, in midline cells and in a large number of cells in the thoracic and subesophageal region. There is no expression of eg-lacZ in the brain lobes. A similar pattern of expression is seen with another enhancer trap, eg^mz360 (data not shown).

The coexpression of eg and serotonin in descendants of NB 7-3 makes a direct connection between the serotonin cells and the three previously identified EW neurons (Higashijima et al., 1996), all of which send their projections contralaterally through the posterior commisure. In Fig. 1D most hemisegments show a third eg-expressing red cell in close proximity to the serotonin neurons (arrowheads), which we hypothesize is the third EW neuron based on its lateral position to the serotonin neurons. This third neuron can also be seen in Fig. 1B in the hemisegments marked with yellow arrowheads. Occasionally three Ddc-immunoreactive cells can be detected in abdominal hemisegments (unpublished results) and, in segment T1 and the most posterior subesophageal segment, the serotonin cells normally occur as triplets, all of which send projections through the posterior commissure. Apparently, in early development, the three EW neurons show similar characteristics in gene expression and growth patterns, even though later in development only two of these cells will synthesize serotonin in the abdominal segments.

In eg loss-of-function mutants, we observe an unusual serotonin cell phenotype that varies in severity with different eg alleles. Fig. 2 shows five eg alleles immunostained for Ddc; eg^mz360, eg^T6 and eg²⁸⁹ are all P-element insertions 5’ of the eg coding region and eg^225A and eg^18B are excisions of the eg^mz360 allele (Dittrich et al., 1997). eg^18B is a null allele (Dittrich et al., 1997) and it is a late embryonic/early larval lethal. All the images are third instar larval CNS except eg^18B, which is a late embryonic CNS. Fig. 2 demonstrates that, although all segments show Ddc immunoreactivity in the midline dopamine cells, many hemisegments do not show Ddc immunoreactivity in the serotonin cells and, strikingly, those hemisegments that do show Ddc-immunoreactive serotonin cells often have a single cell rather than the normal doublet.

The projections in Fig. 2 are arranged with increasing severity of the serotonin phenotype. The eg^mz360 allele shows an almost wild-type pattern of serotonin cells in the abdominal region but only single serotonin cells in the thoracic segments. eg^225A shows a few abdominal segments with doublets and a large number of segments with single cells, but no serotonin cells in thoracic segments. eg^T6 and eg²⁸⁹ show a severe reduction in the number of serotonin cells in both the thoracic and abdominal segments. Even in the null allele, eg^18B, a few Ddc-immunoreactive serotonin cells persist. This suggests that there are redundant mechanisms that allow the formation of serotonin cells in the absence of Eg protein. Most often this redundant function only allows for development of a single serotonin cell even though eg is expressed in both serotonin cells (Fig. 1D). This allelic variation in the serotonin cell phenotype is summarized in Table 1, which quantifies the number of hemisegments that show zero, one or two serotonin cells for each allele. An average value for each allele is calculated to show the reproducibility of the phenotype. Note that the phenotypes of the hypomorphic alleles eg^T6 and eg²⁸⁹ are very similar to the phenotype of the null allele eg^18B.

Confirmation that the remaining single cells in eg mutants can synthesize serotonin is shown by direct examination of serotonin immunoreactivity in Fig. 3E,F. Although these cells can synthesize serotonin, they often have abnormal axonal projections. The normal axonal pathway for the serotonin neurons is to project anteriorly, cross to the contralateral side via the posterior commisure and then project In the eg mutants, we find a mixture of normal and abnormal projections: some of the remaining serotonin cells appear to cross to the contralateral side Fig. 3E (closed arrow) but others are defective in that their projections remain ipsilateral Fig. 3F (open arrow). Abnormal projections of NB 7-3 progeny in eg mutants are also observed during early development (Higashijima et al., 1996; Dittrich et al., 1997). The serotonin neurons in the brain lobes are unaffected by loss of eg function (data not shown), as expected, since eg expression is restricted to the nerve cord (Fig. 1B).

The similarity of the grasshopper and Drosophila NB 7-3 lineages suggests that the serotonin cells of the ventral nerve cord are mitotic sisters. Mutant alleles of eg show the correct number of NB 7-3 progeny early in development (Higashijima et al., 1996); therefore, the undetectable sister serotonin cell must adopt an alternative fate that no longer expresses normal levels of Ddc and serotonin. In an attempt to detect the second cell, we incubated eg mutant CNS in exogenous 5-hydroxytryptophan (5HTP), the metabolic precursor of serotonin that is decarboxylated by Ddc. Both serotonin and dopamine neurons have the ability to uptake exogenous 5HTP, which can be readily detected by serotonin immunoreactivity in cells that contain even very low levels of Ddc (unpublished data). Fig. 2G shows a larval eg^T6 CNS that was first incubated in 5HTP and then immunostained for serotonin. The midline dopamine cells are now detectable, but the missing serotonin cells are still undetectable with this assay. If the undetectable cells are present, they are defective in their uptake mechanism and/or are totally deficient for Ddc, suggesting a dramatic alteration in cell fate.

Although most hypomorphic eg^T6 and eg²⁸⁹ flies die as pupae, 15% survive to viable adults. Given the stochastic variation in development of serotonin cells in the eg mutants (Fig. 2), we suspected that those eg mutant flies that survive to adulthood would have a more normal complement of serotonin-immunoreactive cells. Fig. 4 demonstrates that this is not the case. This figure compares the adult ventral cords from wild type and eg^T6, immunostained with serotonin in green and tyrosine hydroxylase in red to specifically detect dopamine cells. After metamorphosis, the abdominal serotonin cells are located in the posterior tip of the ventral cord. Although the number of dopamine neurons in the eg^T6 mutant is almost equivalent to wild type, the number of serotonin neurons is reduced sixfold to sevenfold, reflecting the phenotype seen previously in the larval CNS (Figs 2, 3).

Serotonin has been implicated as a stimulatory modulator of locomotion in invertebrates (O’Gara et al., 1991; Segalat et al., 1995). Since the eg mutants show a reduced number of neurons synthesizing serotonin, we observed both larvae and adult flies for locomotor and other behaviors. Both eg^T6 and eg²⁸⁹ larvae and flies appear to be normal for phototaxis and geotaxis but, by simple inspection, the adults are much less active in locomotion. Although flight is impossible due to the extended wings this feature is not responsible for the loss of locomotion since the same phenotype is readily visible in flies with clipped wings. We also find that eg mutants have difficulty holding onto glass surfaces when climbing vertically or when rotated upside down but, have no obvious morphological defects in the legs or tarsi. Whether these observed phenotypes are a direct consequence of reduced serotonin levels is uncertain since eg is expressed in three other neuronal lineages. Given the many physiological effects of serotonin that have been observed in other organisms, we are surprised that Drosophila can survive as viable fertile adults with so few serotonin neurons.

Although eg is expressed in both serotonin cells, the eg loss-of-function mutants often affect the development of only one serotonin cell of each pair (Figs 2, 3). We questioned whether the development of one specific cell in each pair is consistently affected. In grasshopper, the two sister cells have slightly different growth patterns and projections (Taghert and Goodman, 1984), but these are impossible to discern in Drosophila. We previously found that the two cells can be distinguished from each other by differential expression of zfh-2 (Lundell and Hirsh, 1994). Here we show that the differential expression of zfh-2 and of another antigen, pdm-1, can be used to determine that the remaining single cell in eg mutants expresses markers characteristic of the more lateral serotonin cell.

In a wild-type CNS, both zfh-2 and pdm1 are selectively expressed in the more lateral serotonin cell but not in the more medial cell (Fig. 5A,B). Both antigens are also expressed in the midline dopamine cells as shown by the single dopamine cell at the top of Fig. 5A,B, and by the entire midline shown in Fig. 5C,D. The remaining single serotonin cells in eg mutants consistently expresses both zfh-2 and pdm-1 (Fig. 5C-F), characteristic of the more lateral serotonin cell. This result is seen in eg^T6 (Fig. 5C,D) where the larval CNS has a high level of Ddc expression, and also in the embryonic CNS of the null allele, eg^18B (Fig. 5E,F). Therefore, even though eg is normally expressed in both serotonin cells, the absence of Eg protein has a more dramatic effect on the fate of the more medial neuron.

To determine the fate of the medial serotonin cell, we examined eg-lacZ expression late in embryonic development using the eg²⁸⁹ allele. In a wild-type line heterozygous for eg²⁸⁹, two to three NB7-3 descendants can be identified in each hemisegment with eg-lacZ (Figs 6A, 1B,D). These three neurons are the EW neurons that generally lie in a row perpendicular to the midline and which project anteriorly to the posterior commisure (Higashijima et al., 1996). For the purpose of discussion, we suggest naming these neurons EW1, EW2 and EW3 from the midline out. Examination of a larval CNS double labeled for eg-lacZ and Ddc shows that EW1 and EW2 are the serotonin neurons and that EW3 is more variable in its position and detectability (Fig. 1D).

In the mutant line homozygous for eg²⁸⁹, which dramatically reduces the number of serotonin cells detectable with Ddc (Fig. 2, Table 1), there are actually more cells expressing eg-lacZ rather than less (compare Fig. 6A to B). This result is summarized in Table 2A, which shows that 26% of the hemisegments in a eg²⁸⁹ homozygote have four or more eg-lacZ-positive cells, versus none in the eg²⁸⁹ /+ heterozygote. This result confirms the work of Higashijima et al. (1996) showing that eg mutations affect cell fates rather than the number of NB 7-3 progeny. Clearly the cells that are not detectable in eg mutants with Ddc or serotonin are present but their specification as serotonin neurons is dramatically altered.

The linear arrangement of the eg-lacZ cells in wild-type CNS (Figs 1D, 6A) suggests that they are the three EW neurons. We presume that the fourth cell that appears in the eg homozygote is the GW neuron, which projects ipsilateral and posteriorly. In a wild-type CNS, the GW neuron must not express as much eg-lacZ as the EW neurons or the eg promoter turns off earlier. Consistent with this latter proposal, eg transcripts in all four NB 7-3 progeny are seen in a wild-type stage 13 embryo, several hours earlier than the stage 17 CNS shown in Fig. 6A (Higashijima et al., 1996; Dittrich et al., 1997). This suggests a negative autoregulatory function of eg that is uniquely active in the GW neuron, a function that could be responsible for its unique properties relative to the EW neurons.

To investigate how eg specifies cell fates in the NB 7-3 lineage, we addressed whether eg²⁸⁹ shows any specific changes in gene expression. In particular, we examined whether the differential expression of pdm1 and zfh-2 is maintained and whether there is any effect on the expression of engrailed (en). We have shown previously that en is required for development of the serotonin neurons (Lundell et al., 1996). Fig. 6 shows the coexpression of these three genes with eg-lacZ in wild-type eg²⁸⁹/+ and mutant eg²⁸⁹/eg²⁸⁹ embryos. The combinatorial use of these four markers allows us to uniquely identify all progeny in the NB 7-3 lineage.

Unique characteristics of EW1, EW2 and EW3 are demonstrated by the immunostained wild-type CNS in Fig. 6C,E,G. In the eg-lacZ/Pdm1 double-labeled wild-type CNS (Fig. 6C), the only eg-lacZ cell to express pdm1 in the NB 7-3 lineage is a single cell that we infer from Fig. 5B to be the more lateral serotonin cell (EW2). In the eg-lacZ/Zfh-2 double-labeled wild-type CNS (Fig. 6E), all eg-lacZ cells of the NB 7-3 lineage show zfh-2 expression except the most medial cell, which we infer from Fig.5A must be the medial serotonin cell (EW1). These results are surprising since pdm1, which is expressed in NB 7-3 (W. Chia personal communication) is restricted to just one progeny neuron late in embryogenesis and zfh-2, which is detectable only after NB formation (Lai et al., 1991), must be independently expressed in the progeny of more than one NB 7-3 GMC. In the eg-lacZ/en double-labeled wild-type CNS, all eg-lacZ cells of the NB 7-3 lineage show en expression (Fig. 6G). In the ventral lateral region of a stage 17 nerve cord, there is cluster of five en-expressing cells. We have shown previously (Lundell et al., 1996) that the two most medial cells in the en cluster are the serotonin cells, EW1 and EW2. In Fig. 6G, all hemisegments show coexpression of en and eg-lacZ in the serotonin neurons but, in some hemisegments, the presence of three coexpressing cells indicates that EW3 can also expresses en. Therefore the three EW neurons can be uniquely identified by their position and expression of pdm1, zfh-2 and en (Fig. 6C,E,G and summarized in Fig. 7). EW1, the medial serotonin neuron, does not express pdm1 or zfh-2 but does express en; EW2, the lateral serotonin neuron, expresses all three gene products; and EW3, which is sometimes detectable with the Ddc antibody (data not shown), expresses zfh-2 and en but not pdm1.

Having established the wild-type pattern of expression, we next asked how the expression of pdm1, zfh-2 and en change in the mutant allele eg²⁸⁹. In summary, we find that expression of zfh-2 and en is dependent on eg function but expression of pdm1 is independent of eg function. The results for each of the three genes are discussed below.

Loss of eg function appears to have no affect on the expression of pdm1 (Fig. 6D; Table 2B). Although there is an increase in the number of eg-lacZ-expressing cells in eg²⁸⁹, expression of pdm1 is still restricted to EW2 (Fig. 6D). Table 2 shows that 89% of the hemisegments in eg²⁸⁹ retain a single coexpressing cell. Since 89% of the hemisegments still show pdm1 expression but 66% of the hemisegments develop with no detectable serotonin cells (Table 1), then clearly the serotonin cell phenotype in eg mutants is not directly related to the expression of pdm1.

Loss of eg function affects the expression of zfh-2 in EW2 (Fig. 6F; Table 2C). Comparison of the eg-lacZ/Zfh-2 images for wild type (Fig. 6E) and eg²⁸⁹ (Fig. 6F) shows that the proportion of cells per hemisegment that express eg-lacZ but not Zfh-2 increases in eg²⁸⁹. In wild-type nerve cords, almost all (99%) hemisegments with three eg-lacZ cells show two cells that also express zfh-2 (EW2 and EW3), and one cell that expresses only eg-lacZ (EW1) (Fig. 6E; Table 2C). In eg²⁸⁹ only 40% of hemisegments show this pattern; instead the remaining hemisegments show two or more cells that express only eg-lacZ. This increase in cells that only express eg-lacZ could be explained either by the appearance of the GW neuron, which does not express zfh-2 or by the loss of zfh-2 expression from EW2 or EW3. We favor the later hypothesis because the cells expressing only eg-lacZ consistently occur as medially located doublets (yellow arrowheads, Fig. 6F), suggesting that expression of zfh-2 is absent in EW2 (Fig. 6F). In addition, some eg²⁸⁹ hemisegments where four eg-lacZ neurons are clearly detectable (asterisk, Fig. 6F), show two medial cells that lack zfh-2 expression and two lateral cells that coexpress zfh-2, suggesting that the GW neuron can express zfh-2 in eg289.

Loss of eg function affects the expression of en in both serotonin neurons (Fig. 6H; Table 2D). Comparing the eg-lacZ/En images of wild type (Fig. 6G) with eg²⁸⁹ (Fig. 6H) the most obvious difference is the appearance of one or two cells in eg²⁸⁹ that express eg-lacZ but not en, per hemisegment. These cells tend to be the cells of the en cluster closest to the midline, EW1 and EW2. Table 2D indicates that, in eg²⁸⁹, 7% of the hemisegments show only coexpressing eg-lacZ/en cells like wild type, 32% show one cell that lacks en expression and 61% show two or more cells that lack en expression. This distribution is similar to the distribution of detectable serotonin cells in eg²⁸⁹ (Table 1); 4% show two cells, 30% show one cell and 66% show no cells. Therefore we propose that hemisegments with two cells that only express eg-lacZ (closed arrowheads in Fig. 6H) would likely correspond to hemisegments with no detectable serotonin cells and hemisegments containing one cell that only expresses eg-lacZ (open arrowheads in Fig. 6H) would likely correspond to hemisegments, where only the single lateral serotonin cell (EW2) is detectable. In some eg²⁸⁹ hemisegments, where four eg-lacZ neurons are clearly observed (asterisk, Fig. 6H), there are two medial cells that lack en expression and two lateral cells that coexpress en, suggesting that the GW and EW3 neurons express en in eg²⁸⁹.

Our previous work demonstrated that the serotonin neurons are descendants of the NB 7-3 lineage (Lundell et al., 1996). This manuscript assigns a unique identity to each of the NB 7-3 progeny neurons based on the expression of specific molecular markers, position within the nerve cord and the effects of eg loss-of-function mutations. We show that eg mutations alter the specification of cell fates in this lineage. Our results are summarized with a model presented in Fig. 7.

During mid-embryogenesis, stages 11-13, the thoracic and abdominal segments of the ventral cord show four neurons that are derived from NB 7-3 (Higashijima et al., 1996; Dittrich et al., 1997). The four NB 7-3 progeny comprise: (a) the three EW neurons that initially lie in a lateral row, project anteriorly to the posterior commisure and cross to the contralateral side and (b) the GW neuron, which is initially located slightly posterior to the EW neurons and projects posteriorly on the ipsilateral side (Higashijima et al., 1996). Here we demonstrate that the two EW neurons closest to the midline, EW1 and EW2, become serotonin-producing neurons (Fig. 1D). By analogy to grasshopper (Taghert and Goodman, 1984), we assume that EW1 and EW2 are mitotic sisters of the first ganglion mother cell, GMC 1. Also by analogy to grasshopper, the EW3 neuron is most likely a progeny cell of GMC 2. The GW neuron is either a mitotic sister of EW3 or represents a single progeny of GMC 3.

During late embryogenesis, stage 17, only three eg-lacZ-expressing descendants of NB 7-3 are detectable in a wild-type CNS (Figs 1D, 6A, 7); two are the serotonin cells and the third, EW3, we identify by its lateral position. Additionally, this third cell infrequently expresses Ddc and projects to the anterior commisure, similar to the EW1 and EW2 neurons (data not shown). The combinatorial use of three molecular markers, pdm1, zfh-2 and en has allowed us to uniquely identify all three EW neurons. In a wild-type nerve cord EW1 expresses only en, EW2 expresses all three markers and EW3 expresses en and zfh-2 (Figs 6C,E,G, 7). The GW neuron is no longer detectable with eg-lacZ during late embryogenesis so we do not know the expression of these genes in a wild-type GW neuron.

Our experiments with eg mutants confirm and extend the conclusions of Higashijima et al. (1996) showing that loss of eg function does not alter the number of NB 7-3 progeny, but rather alters the specification of cell identity in this lineage. Although eg mutations reduce the number of cells detectable with DDC and serotonin (Figs 2, 3) all progeny of this lineage are detectable with eg-lacZ (Figs 6B, 7). Therefore, the specification of serotonin cell fate is altered in eg mutants. In the absence of Eg protein, the mitotic sister serotonin neurons become quite distinct. EW1 is much more sensitive to reduced Eg, resulting in single serotonin cells in many hemisegments (Figs 2, 3). In contrast, serotonin synthesis in EW2 can be completely independent of eg function since we observe these cells as pdm1- and zfh-2-expressing serotonin cells even in the null allele eg^18B (Fig. 5E,F).

The simplest explanation for the difference between EW1 and EW2 is that EW2 selectively contains a redundant mechanism that allows continued synthesis of serotonin in the absence of Eg protein. This redundant mechanism is not 100% efficient since not all segments in an eg mutant CNS contain serotonin cells. What specific factors are present in EW2, but not in EW1, that might account for this redundancy? We have shown that pdm1 and zfh-2 are differentially expressed in EW2 (Fig. 5A,B). Both proteins are potential transcription factors: Pdm-1 is a POU-protein (Billin et al., 1991; Dick et al., 1991; Lloyd and Sakonju, 1991) required for appropriate development of the first ganglion mother cell in the RP2 lineage (Bhat et al., 1995; Yeo et al., 1995) and Zfh-2 is a homeodomain zinc-finger protein (Fortini and Rubin, 1990), which binds to a serotonin cell-specific regulatory element of the Ddc promoter (Lundell and Hirsh, 1992). Our results show that zfh-2, but not pdm1, expression is affected in eg mutants (Figs 6, 7; Table 2). Therefore zfh-2 may be a potential factor for this redundant pathway that establishes eg-independent serotonin synthesis. The model in Fig. 7 presents the possibility that loss of zfh-2 expression in EW2 is correlated with the loss of serotonin synthesis in these cells. Definitive proof of this hypothesis will require direct examination of zfh-2 mutants.

In this manuscript, we demonstrate that, in an eg loss-of-function mutant, the loss of Ddc expression is always accompanied by a loss of en expression, but can occur independently in the two serotonin cells: in a hemisegment where both cells fail to express Ddc, neither cells show en expression, in a hemisegment where only EW2 continues to express Ddc, EW2 shows en expression but EW1 does not (Figs 6H, 7). Previously we have shown that en is required for development of the serotonin cells, with both serotonin cells affected equally in an en loss-of-function mutant (Lundell et al., 1996). Recently it has been shown that in the same en mutant eg expression is absent in NB 7-3 (Matsuzaki and Saigo, 1996). Therefore en and eg show different relationships during the development of this lineage: at the neuroblast stage, en is required to maintain eg expression, whereas after division of GMC1, eg is required to maintain en expression in each cell. Continued expression of en and Ddc does not require eg expression, since eg expression ends at stage 13 (Higashijima et al., 1996).

The results presented in this manuscript are establishing a hierarchy of genetic interactions that leads to the specification of serotonin cell fate. Other genes that have been shown to affect development of the NB 7-3 lineage include wingless, hedgehog, patched, gooseberry and huckebein (Higashijima et al., 1996; Lundell et al., 1996; Matsuzaki and Saigo, 1996; Patel et al., 1989), but their specific contribution to the differentiation of serotonin neurons is unknown. The expression of eg in only a small subset of neurons results in CNS where the overall organization is well preserved in eg mutants. This is not only beneficial for the study of cell specification but leads to the recovery of hypomorphic adults. Our interesting finding that adult flies can survive with low levels of serotonin, albeit with reduced locomotor activity, begs for further analysis into the function of the ventral cord serotonin neurons. Our identification of unique properties for each cell in the NB 7-3 lineage provides new tools for further investigation into both the function and mechanisms that lead to the specification of serotonin neurons during neurogenesis.

The work was supported by NIH grant GM 27318 to JH. We thank J. Urban and G. Technau for the eagle stocks, C. Doe, W. Chia and A. Tomlinson for antisera, S. Britt and A. Cassill for laboratory space and reagents at UTSA, and Claire Cronmiller and the members of the Hirsh laboratory for helpful discussions and comments on the manuscript.