Hunchback is required for the specification of the early sublineage of neuroblast 7-3 in the Drosophila central nervous system

The central nervous system (CNS) of all higher organisms consists of a large variety of different neuronal and glial cell types that build up a highly complex but reproducible cellular network. Thus, one challenging question in developmental neurobiology is how these many different neural cell types are specified in a correct spatiotemporal pattern during development. While several reports exist on neural fate specification with respect to position (reviewed by Bhat, 1999), virtually nothing is known about how neuronal diversity is organized with respect to the temporal axis. Therefore, it is of high interest to unravel genetic interactions that lead to the differential specification of neuronal cell fates in specific time windows. Owing to its accessibility and relative simplicity, the Drosophila ventral nerve cord (VNC) provides a very good model system with which to answer such questions.

The VNC of Drosophila develops from neural precursor cells, the so called neuroblasts (NBs), which delaminate in five successive waves. Each NB is characterized by its time of delamination, its position and gene expression (Broadus et al., 1995; Doe, 1992). Most NBs divide asymmetrically several times and are believed to generate one ganglion mother cell (GMC) at each division, which subsequently produces two neurons and/or glia cells (Hartenstein et al., 1987). At the end of embryogenesis, the VNC consists of ∼320 neurons and ∼30 glia cells per hemineuromere. Previous work indicates that NBs give rise to characteristic reproducible cell lineages (Bossing et al., 1996; Schmidt et al., 1997). As specific neurons have always been shown to derive from the first born GMC of the parental NB, it has been suggested that NB lineages are generated in a fixed temporal sequence (Brody and Odenwald, 2000). This has highlighted Drosophila NB lineages as being a very good model system for studying sequential neural cell fate specification.

Recently, the involvement of a set of genes encoding transcription factors that are expressed sequentially within most, if not all, NB lineages, has been detected (Kambadur et al., 1998; Brody and Odenwald, 2000). This ‘gene cassette’ consists of the genes hunchback (hb), the redundantly acting POU domain protein 1 (pdm1; nub – FlyBase) and POU domain protein 2 (pdm2), castor (cas), and grainyhead (grh) (Brody and Odenwald, 2000), which are expressed in early-, middle- and late-born neurons of NB lineages. Furthermore, klumpfuss (klu) has to be added to this list, as it is also expressed in NB sublineages (Yang et al., 1997). The successive time periods of hb, pdm1 and pdm2 and cas expression is partially established through the inhibition of pdm1 and pdm2 expression via hb and cas in early and late sublineages, respectively. The expression patterns of these genes clearly suggest a function in sequential determination of neural cell types. Indeed, pdm1 and pdm2 (Johnson and Hirsh, 1990; Billin et al., 1991; Dick et al., 1991; Lloyd and Sakonju, 1991; Prakash et al., 1992; Yang et al., 1993; Bhat and Schedl, 1994; Ng et al., 1995; Yeo et al., 1995), cas (Mellerick et al., 1992; Cui and Doe, 1992; Cui and Doe, 1995) and klu (Yang et al., 1997) have a role in neural cell fate specification. However, a function of hb and grh in CNS development has not yet been shown.

In this study, we have analyzed the function of hb within NB lineages by focusing on the NB7-3. We chose this lineage for two reasons: first, it is small, consisting of only four neurons (Bossing et al., 1996; Schmid et al., 1999); and second, all NB7-3 neurons can be distinguished by their characteristic position, expression of marker genes and/or their transmitter phenotype (Bossing et al., 1996; Higashijima et al., 1996; Dittrich et al., 1997; Lundell and Hirsh, 1998) (this paper). Using BrdU incorporation, we show that the NB7-3 neurons derive from three GMCs. We found that hb is expressed in the early NB7-3, its first GMC (GMC7-3a) and its progeny, and that this gene is indeed necessary for the specification and/or maintenance of the early-born neurons within this lineage. Ectopic expression of hb in the later born sublineage is sufficient to transform these cells into neurons with the same fate as the GMC7-3a progeny. Moreover, additional neurons of early GMC7-3a fate are observed. We also found that hb might play a similar role in at least one other NB lineage – NB 7-1. Thus, our data support the hypothesis that hb is an important early cell fate determinant in the process of sequential cell fate specification within the CNS development.

The following fly strains were used in this work: Oregon R (wild type); eagleGal4 (Mz360) (Dittrich et al., 1997); scabrousGal4 (Budnik et al., 1996); engrailedGal4 (Tabata et al., 1995); UAS-hb (second and third chromosomal insertion) (Wimmer et al., 2000); CycA³/TM3, sb, Ubx-lacZ (Lehner and O’Farrell, 1989); Df(3L)H99/TM6b, abdA-lacZ; hb¹²/TM3,Sb; and hb⁹/TM3,Sb (Bloomington stock center).

BrdU injection was as described previously (Prokop and Technau, 1991) with the following changes: the embryos were chemically dechorionized with 7.5% bleach shortly before injection. After injection they were allowed to develop until stage 15 and subsequently fixed with heptane saturated with 37% formaldehyde for 20 minutes.

Embryos and L1 larval brains were fixed, stained and mounted as described previously (Dittrich et al., 1997). Primary antibodies used were rat anti-Hunchback (1:100), rat anti-Sal (1:250), mouse anti-Zfh-1 (1:300), mouse anti-Zfh-2 (1:200), rabbit anti-Eagle (1:1000), mouse anti-Eagle (1:200), rabbit anti-Pdm-1 (1:500), rabbit anti-Serotonin (1:5000, Sigma), rabbit anti-Corazonin (1:2000), mouse anti-Eve (1:2, Developmental Studies Hybridoma Bank), rabbit anti-Eve (1:1000), rabbit anti-Castor (1:500) and anti-DIG-AP (1:1000, Roche). The secondary antibodies anti-rabbit-TRITC, anti-mouse-FITC and anti-rat-FITC from goat (Jackson Laboratories) were used at a 1:250 dilution. The mounted embryos were analyzed with a confocal laser scanning microscope (Leica TCS SPII). Scanning images were processed with Adobe Photoshop Macintosh Version 5.5.

DIG-labeled RNA probe was synthesized with SP6 RNA Polymerase and pGEM-dSERT as a template according to the manufacturer’s protocol (Roche). The hybridization on embryos was performed as described previously (Tautz and Pfeifle, 1989; Plickert et al., 1997).

To investigate mechanisms involved in the specification of cell fates within the Drosophila CNS, we have chosen the lineage of the abdominal NB7-3 as a model system. The NB7-3 and all of its progeny express the transcription factor eagle (eg) (Higashijima et al., 1996; Dittrich et al., 1997), allowing a reliable identification of the NB7-3 derived neurons. The abdominal NB7-3 generates three contralaterally projecting interneurons (EW1, EW2, EW3) and one motoneuron [GW; nomenclature according to Higashijima et al. (Higashijima et al., 1996)] each of which can be identified by a combination of specific markers (see Fig. 1). As the GW motorneuron is always the most posterior cell within the NB7-3-derived cell cluster, it can easily be distinguished from the interneurons by its position (Higashijima et al., 1996; Dittrich et al., 1997; Lundell and Hirsh, 1998). Using this as a criterion we found that GW is the only NB7-3 derived neuron which expresses the transcription factor Zfh-1 (see Fig. 1E, Fig. 4D). Previous work has shown that from the second thoracic (T2) to the seventh abdominal segment (A7), the two most medially lying interneurons, EW1 and EW2 (see Fig. 1B), express the neurotransmitter Serotonin (5-HT) (Fig. 1D,E) (Lundell et al., 1996; Dittrich et al., 1997; Lundell and Hirsh, 1998). Furthermore, we found that the most laterally lying EW3 synthesizes the neuropeptide Corazonin (Crz; see Fig. 1C,D) (Veenstra, 1989; Cantera et al., 1994). As Serotonin as well as Corazonin antibody staining worked earliest in brains of freshly hatched L1 larvae, we used dSerT as a marker. The transcript for the Drosophila Serotonin transporter (dSerT) is specifically expressed in the serotonergic cells from stage 15 onwards (see Fig. 4J). Thus, we performed dSerT mRNA in situ hybridization to differentiate between the serotonergic and the corazonergic cells in the embryo. Additionally, anti-Zfh-2 antibody was used to distinguish between the two serotonergic neurons, as Zfh-2 is expressed only in the serotonergic EW2 and the corazonergic EW3 neuron.

To determine the sibling relationships of the individual neurons of the NB7-3 lineage, BrdU incorporation experiments were performed. BrdU is a thymidine analog that is incorporated into DNA during replication (S-phase). Thus, all cells going through an S-phase after BrdU application are labeled (Bodmer et al., 1989; Prokop and Technau, 1991). According to existing models of NB division, NBs produce a chain of GMCs, which subsequently divide once to generate a pair of neurons (Hartenstein et al., 1987). Based on this assumption, we expected the four progeny neurons of NB7-3 to be generated by two GMCs, which would mean that the two early- and the two late-born neurons should always co-label with BrdU. Conversely, if two neurons do not co-label, they must derive from different GMCs.

NB7-3 delamination takes place at about 6.5 hours after egg laying (AEL). Therefore BrdU was injected into embryos at two time points: between 6.5 and 7 hours AEL and between 7 and 7.5 hours AEL. After injection these embryos were allowed to develop until stage 15 and stained with anti-BrdU and anti-Eg antibody. At stage 15 all NB7-3 derived cells are postmitotic as they express the neuronal marker Elav (Robinow and White, 1991) (data not shown). Out of 111 abdominal NB7-3 lineages derived from injection between 6.5 and 7 hours AEL, 38% were BrdU positive in all four neurons, 55% in three neurons, 5% in two and 2% in one neuron. Injection at a later time point (7 to 7.5 hours AEL) resulted in an increase of NB7-3 lineages with less than three BrdU-positive neurons: 20% of hemisegments showed BrdU labeling in all four neurons, 28% in three neurons, 14% in two neurons and 39% in one neuron (n=153). Injection at around 8.5 hours AEL did not yield any BrdU staining in the NB7-3 cluster, indicating that after that time point there is no more S-phase in this lineage (Fig. 2E). By evaluating these results with respect to the position of the BrdU labeled neurons (see Fig. 2A-E,A′-E′), we found that GW and EW1 must be siblings as they are always labeled together (Fig. 2A,A′,B,B′). This is in contrast to EW2 and EW3, which are very often separately labeled (63%, n=264; Fig. 2B,B′,D,D′). Thus, according to our results, EW1 and GW are siblings, while EW2 and EW3 are generated separately.

We postulate that EW1 and GW are produced by the first GMC of NB7-3 (GMC7-3a) because both cells express hb, which has been shown to be expressed by the first cells generated from a neuroblast (Kambadur et al., 1998; Brody and Odenwald, 2000). This assumption is supported by the fact that NB7-3 already expresses hb before its first cell division and hb expression is subsequently maintained in its first progeny (Fig. 3A,B). This is also supported by the analysis of stage 15 embryos that are mutant for Cyclin A (CycA). In these mutants, cell division is blocked in the nervous system (Wai et al., 1999) and as a consequence only one, two or three cells were found in the position of the NB7-3 cluster (8%, 75% and 17%, respectively; n=63). When two cells were detected, one was Hb positive in 98% of these cases (n=47; Fig. 2H). This cell was often smaller and lay dorsal to the other cell, showing that it is indeed the first born GMC that stays Hb positive. In wild-type embryos, EW1 and GW are mostly found in the dorsal-most position that is typical for the first-born neurons of an NB lineage (Kambadur et al., 1998). Additionally, we found several NB7-3-derived clusters where EW2 and EW3 were BrdU labeled, while GW and EW1 were not (10%, n=264; Fig. 2C,C′). This indicates that the GMC giving rise to GW and EW1 has finished its S-phase before the precursors of the other two neurons in these cases (Fig. 2C′′). However, the fact that a high number of clusters show EW2 to be the only unlabeled cell suggests that the S-phase of the first GMC is often delayed with respect to the S-phase of the EW2 precursor (39%, n=264; Fig. 2B′′). This suggests that EW2 is generated by the second GMC (GMC7-3b), while EW3 derives from GMC7-3c supported by the fact that EW3 is always the last cell labeled by BrdU (Fig. 2D,D′). This means that two of the three GMCs produce only one neuron each, which would be unusual as it is believed that GMCs in the insect nervous system normally divide once to produce two sibling neurons (Hartenstein et al., 1987). Possible explanations for this are that either one of the resulting siblings is dying by programmed cell death (PCD) or the GMCs are able to differentiate directly without further cell division or there is a mixture of both. To test whether PCD occurs at all in this lineage, we stained cell death deficient mutant embryos at stage 14 (Df(3L)H99) (White et al., 1994) with anti-Eg antibody and found in 68% of hemisegments between six and ten cells at the position of the NB7-3 cluster when compared with four cells in wild type (n=38; Fig. 2F,G). Thus, PCD clearly occurs in this lineage raising the possibility that the siblings of EW2 and/or EW3 are indeed eliminated during wild-type development.

After the detailed characterization of the NB7-3 lineage, we next analyzed candidate genes that could be involved in the specification of cell fate during lineage development. A recent publication (Kambadur et al., 1998) suggested that the sequential expression of the transcription factors Hb, Pdm-1, Pdm-2 and Cas might be involved in the specification of different cell fates within sublineages. Antibodies against Hb, Pdm-1 and Cas were used to test whether these proteins are also present in the NB7-3 lineage. As a result, we found hb and pdm-1 but not cas to be expressed within the lineage (Fig. 3). At stage 15, Pdm-1 could be detected in all NB7-3-derived neurons (Fig. 3C), although often not all cells of a particular cluster were labeled (not shown). Hb was found only in the early born NB7-3 sublineage, i.e. GW and EW1 (Fig. 3B). Analyses of hb expression at earlier stages of development showed that Hb was already present in the NB7-3 shortly after delamination (Fig. 3A, left hemisegment). After the first division of NB7-3, the NB itself and GMC7-3a were Hb positive (Fig. 3A, right hemisegment). In NB7-3 clusters consisting of three cells, all cells expressed hb (Fig. 3A, middle hemisegment). Because at later stages only two neurons were Hb positive, hb must be downregulated in the neuroblast after its first division, so that EW2 and EW3 stay Hb negative (Fig. 3B).

To test whether hb plays a role in the specification of GMC7-3a and its progeny we analyzed the NB7-3 lineage of embryos mutant for hb. To make sure that the obtained phenotypes are indeed related to Hb function we analyzed two independent mutations [the hypomorpic allele hb⁹ (Lehmann and Nüsslein-Volhard, 1987) and the null allele hb¹² (Tautz et al., 1987; Hulskamp et al., 1994)] in parallel in most experiments. As similar results were obtained with both alleles, we will present the results found with hb¹², unless stated otherwise. Because in hb mutant embryos the thoracic segments and the last abdominal segment are missing, and segments 7 and 8 are fused (Lehmann and Nüsslein-Volhard, 1987), analyses were carried out only in the abdominal segments A1 to A6.

hb¹² mutant NB7-3 lineages consisted of two (20%) or three (80%) Eg-positive neurons at stage 13 (Fig. 4E) (n=66) but only two remaining Eg-positive cells at stage 15 (98%, n=50; Fig. 4B,E′′). To test which of the cells are missing, we performed double staining with antibodies against cell specific markers. Double staining against Zfh-1 and Eg at stage 13 showed that one posterior lying cell is mostly Zfh-1 positive (98%, n=53; Fig. 4E′) suggesting that the GW neuron is initially made, while at the later stage 15 only two Zfh-1-negative cells can be detected (n=49; Fig. 4E′′). Thus, the GW neuron either does not survive or does not maintain zfh1 expression at this later time point. The two remaining cells always express zfh-2 (n=72; Fig. 4H), the marker for the late sublineage neurons (Fig. 1E, Fig. 4G). This strongly suggests that in hb¹² mutant embryos, the neurons of the late NB7-3 sublineage (EW2 and EW3) are still present. These data were supported by RNA in situ hybridization against dSerT mRNA in hb mutant embryos, showing that only one neuron per hemisegment is labeled (n=56; Fig. 4K), which is probably the Zfh2-positive EW2 neuron. In addition, we tested for the presence of the neurotransmitters Serotonin and Corazonin in hb⁹ mutant early larval VNCs (Fig. 5). In accordance with the decrease in dSerT-positive neurons, a significant reduction of Serotonin levels in the VNC was detected, with about 20% of hemisegments showing only one serotonergic cell (n=66; Fig. 5E). However, although the EW3 neuron seemed to be correctly specified as judged by zfh-2 expression, the Corazonin staining is missing in 39% of the hemineuromeres (n=112; Fig. 5B). This suggests that, although present, the EW3 neuron is affected with respect to its transmitter phenotype.

We next tested whether hb is able to transform the fates of late-born cells into those of early-born cells, by expressing hb ectopically in the late sublineage using the UAS/Gal4 system (Brand and Perrimon, 1993). By staining egGal4::hb embryos with anti-Eg, we found mostly five to eight NB7-3 derived cells in abdominal segments (82%, n=132; Fig. 4C). A subsequent analysis of the expression of marker genes revealed that 67% of the clusters have more than one Zfh1 positive cell (n=70; Fig. 4F), whereas 9% of these hemisegments show at least 3 Zfh1 labeled NB7-3 neurons. In addition, all cells of the NB7-3 cluster showed no zfh2 expression (n=68; Fig. 4I). Upon ectopic hb expression, the number of dSerT-positive neurons was increased by up to three or four cells in 74% of hemisegments (n=111; Fig. 4L). Correspondingly, the number of serotonergic neurons was increased in 90% of hemisegments (n=70; Fig. 5F) and Corazonin could not be detected (Fig. 5C). Taken together these results indicate that ectopic hb expression has two effects: first, the total number of NB7-3-derived neurons is increased; and second, all neurons express markers of the early sublineage, while those of the late sublineage are missing.

Having found that Hb is an important determinant of early cell fate within the NB 7-3 lineage, we next tested whether this might also be true for other lineages of the Drosophila VNC. To do this we chose the eve (even skipped)-expressing cells, because a known subset of these are early-born progeny of identified NBs (see Fig. 6). Indeed, using antibodies against Eve and Hb in stage 15 embryos, we found co-expression in a dorsal subset of eve-expressing cells that consists of aCC, pCC (from GMC1-1a), RP2 (from GMC4-2a) (Fig. 6A′), fpCC and the dorsalmost CQ neuron (most likely the first born neurons from NB7-1; Fig. 6A′′) (Bossing et al., 1996). Hb was not found in the NB 3-3-derived EL cells. We subsequently analyzed the function of Hb within these lineages using spalt (sal) as an additional marker for aCC, pCC and RP2 (Fig. 6B′). One of the EL neurons is also Sal positive (Fig. 6B′′) but this cell can be reliably identified by its much more ventral position. We could show that Hb affected the NB7-1-derived fpCC and CQ neurons in a similar way as the GW and EW1 neurons in the NB7-3 lineage. In hb¹² mutant embryos, the dorsal subset of the NB 7-1 derived neurons was missing in 97% of hemineuromeres (n=72): 80% showing a loss of two, 10% of one and 7% of three cells (Fig. 6C′′). The missing cells are probably fpCC and/or the dorsalmost CQ neuron, which normally express hb. We assume that the missing third cell might be shifted into the EL cluster where it can no longer be identified (Fig. 6C′′). Conversely, we observed an increase of the number of Eve-positive neurons in CQ position when we ectopically expressed hb via scaGAL4 or engrailedGAL4 (Fig. 6D′′,E′′) (Tabata et al., 1995). enGAL4 drives hb expression in all row 6 and 7 NBs. Among these neuroblasts, only NB7-1 gives rise to Eve-positive neurons (Broadus et al., 1995; Bossing et al., 1996). Therefore, we conclude that these additional neurons are most likely additional fpCCs and/or ‘dorsal’ CQ neurons. However, owing to the lack of markers for the Eve-negative cells of the NB7-1 lineage, we could not check whether these additional neurons appear at the cost of other NB7-1-derived neurons.

In addition, other Eve-positive neurons are affected in hb loss of function mutant embryos, because either aCC or pCC were missing in ∼39%, aCC and pCC in ∼11%, and RP2 in ∼67% of hemisegments (Fig. 6C′; n=60). In scaGal4::hb embryos, rarely one or two additional Sal-positive neurons were observed in the vicinity of aCC and pCC (7%, n=104; Fig. 6D′). However, we could not assign identities to these cells, because the additional neurons could be due to a duplication of aCC and/or pCC or to additional fpCCs, as on rare occasions (less than 1%) fpCC is Sal positive too (Fig. 6E′).

In this work, we have analyzed the role of hb in CNS development. Using the NB7-3 lineage as a model system, we have found that Hb is an important factor in the determination and/or specification of the hb-expressing neurons in the NB7-3 lineage. Additionally, we have shown that other early born neurons are dependent on Hb as well, a result that is independently supported by Isshiki et al. (Isshiki et al., 2001).

By injecting BrdU at different time windows during development, we determined the sibling relationships of the progeny of NB7-3, which consists of two serotonergic interneurons (EW1 and EW2), one interneuron that expresses the neuropeptide Corazonin (EW3) and one motorneuron (GW) (Fig. 1) (Higashijima et al., 1996; Lundell et al., 1996; Dittrich et al., 1997; Lundell and Hirsh, 1998) (this work). According to our results, EW1 and GW are siblings, while EW2 and EW3 are generated sequentially from the NB7-3.

We have strong evidence that EW1 and GW are produced by the first GMC of NB7-3 (GMC7-3a). As EW3 is nearly always the last neuron labeled by BrdU, we hypothesize that this neuron is most likely to be generated by the third GMC, while GMC7-3b gives rise to EW2. The fact that we do not find any siblings for the later-born neurons can be explained in two ways: either the missing sibling is dying by PCD or the GMCs are able to differentiate directly without further cell division. We currently cannot conclude which of these possibilities is true but investigations in apoptosis-deficient mutant embryos clearly show that PCD occurs within the NB7-3 lineage. The frequent occurrence of more than six NB7-3 derived cells in these mutant embryos might be due to additional cell divisions of the surviving cells.

The restricted expression of hb in the early NB7-3 (before and shortly after its first division) and the first-born GMC, including its progeny suggest that hb is involved in the specification of these cells. Our analysis strongly supports this assumption: in stage 13 hb loss-of-function mutant embryos, we found either a complete loss of the GMC7-3a-derived EW1 and GW neurons or of the EW1 neuron alone. At stage 15 the GW neuron could also no longer be detected. Therefore, we conclude that the loss of Hb cannot transform GMC7-3a into GMC7-3b or GMC7-3c (Fig. 7B). This is different from the findings of Isshiki et al. (Isshiki et al., 2001); they found a duplication of EW2 neurons in a low number of cases. We do not know why this difference occurs, but it might be due to the use of different hb alleles. Our results suggest that either additional factors might be necessary for determining the fates of GMC7-3b and/or c, which are not present in the Hb positive GMC7-3a, or that GMC7-3a harbors additional factor(s) that suppress other GMC fates. However, ectopic expression of hb in all cells of the lineage leads to additional neurons of the EW1 and GW type at the cost of neurons normally generated by the later-born GMCs (Fig. 7C). This shows clearly that hb is sufficient to induce fates of the first GMC and to suppress those of the later-born GMCs. This can be interpreted in two ways. One possibility is that overexpression of hb in the progeny of GMC7-3a is sufficient to reiteratively cause both of these cells to adopt the fate of the parental GMC (GMC7-3a). But as hb is strongly expressed in the differentiating progeny of the first GMC anyway, it is very unlikely that a higher dose of Hb has such an effect. Therefore we favor the idea that ectopic Hb in the other GMCs and/or the late NB7-3 (after the first division) is able to transform the fate of these GMCs into the fate of the normally Hb-positive GMC (GMC7-3a). However, the fact that we found a high number of two to three additional EW1-like neurons (as judged by the presence of dSerT and absence of zfh2 expression), but mostly only one additional Zfh1-positive GW neuron suggests that only one additional GMC is fully transformed in most cases. This hints at an additional factor being important for the generation of GW, which is present mainly in one additional GMC beside GMC-1, most probably in the second GMC (Fig. 7C). A good candidate for such a factor appears to be Krüppel (Kr) (Isshiki et al., 2001).

Interestingly, these overexpression experiments give rise to a maximum number of eight Eg-positive cells in the NB7-3 cluster, instead of six, as it would be expected when all three GMCs generate two cells. This could mean that hb is able to protect the neuroblast from apoptosis. The supernumerary cell(s) could then be additional progeny generated by the surviving NB or the NB itself.

Our analysis of the NB7-3 lineage revealed that, in hb loss-of-function mutants, the neurons that are normally Hb positive are affected. However, using the hypomorphic allele hb⁹, which develops into L1 larval stage we could additionally show that the Hb negative EW3 is also affected and fails to express Corazonin in a significant number of cases. Because not only this cell but also the immediate precursor cell(s) must have had a reduced Hb function, we can imagine two scenarios. It is possible that Corazonin expression can be switched on only in the last generated neuron when the NB receives Hb activity during the early phase of lineage development. This would favor the idea that at least some aspects of neuronal fate are dependent on transient gene activity within the stem cell long before the neuron is generated. Alternatively, initiation and/or maintenance of Corazonin expression might depend on cellular interactions such as synaptic inputs or the presence of extrinsic factors that could be altered in the hb mutant CNS. This could also explain the difference in our results and those of Isshiki et al. (Isshiki et al., 2001), who report no change in Crz expression in hb mutant embryos at a much earlier stage than we have analyzed. That extrinsic factors indeed can play a prominent role in specifying and maintaining neurons, including the transmitter phenotype, has been extensively shown in vertebrates (for a review, see Edlund and Jessell, 1999).

hb-positive cells of other lineages are also affected in the hb mutant background: the Eve-expressing neurons RP2 (derived from GMC4-2a) (Doe, 1992) and aCC and/or pCC (derived from GMC1-1a) (Doe, 1992) are frequently undetectable. Additionally, fpCC and/or the dorsalmost of the four CQ neurons (derived from NB7-1) (Bossing et al., 1996) are not found in 97% of the hemineuromeres. These cells are most probably identical to the U1 and U2 neurons described by Isshiki et al. (Isshiki et al., 2001). The lower penetrance of the phenotype in RP2 and aCC/pCC (67% and 50%, respectively) could be due to other cell or lineage specific factors that are co-expressed with hb and are able to act redundantly in these cases. In particular, RP2 might be rescued by the presence of Pdm1 and Pdm2 activity in its parental GMC, because it has been shown that these two proteins are necessary for acquiring a GMC4-2a fate (Yang et al., 1993; Yeo et al., 1995). Maternally contributed hb (Lehmann and Nüsslein-Volhard, 1987) might also be able to partially rescue the phenotype in the lineages of the early delaminating NB4-2 and NB1-1.

In any case, the phenotype of the loss-of-function mutant embryos suggests that hb is, as in NB7-3, involved in the specification or at least maintenance of certain NB progeny. In particular, the role of hb with respect to the NB7-1 derived fpCC and dorsal CQ neuron seems to be comparable with what we found in the NB7-3 lineage, as pan-neural hb overexpression leads to additional Eve-positive fpCCs and/or CQ neurons. Also, additional Sal-positive neurons were observed in the position of aCC/pCC. This is in agreement with the results of Isshiki et al. (Isshiki et al., 2001) who found a duplication of aCC/pCC.

Our results indicate that hb must be added as a functional member of the group of CNS sublineage genes that are involved in the specification of GMCs. The other three genes of this group, which have been investigated on the NB lineage level are the redundantly acting pdm1 and pdm2 genes and klumpfuss (klu). As mentioned above, pdm1 and pdm2 have been shown to be involved in the specification of the first GMC (GMC-1) in the NB4-2 lineage (Yang et al., 1993; Yeo et al., 1995). However, ectopic pdm2 expression did not lead to a transformation of the later-born GMCs as we postulate for hb in the NB7-3 lineage, but transformed the progeny of GMC4-2a into the fate of the parental cell (Yang et al., 1993). Thus, although pdm1 and pdm2 are able to induce a NB4-2 specific first GMC fate, they can do this only in cells that derive from the first born GMC. This suggests that the ‘quality of being GMC-1’ (of NB4-2) is independent of Pdm1 and Pdm2 function.

However, in klu loss-of-function mutant embryos, the normally Klu-positive GMC-4-2b transforms into the fate of GMC-4-2a (Yang et al., 1997). Thus, in the absence of Klu, GMC4-2b has also a ‘GMC-1 quality’ and Klu is necessary for the switch to a ‘GMC-2’ fate. What determines this ‘GMC-1 quality’ in the NB4-2 lineage is not known, but our results imply that Hb might have such a function during NB7-3 lineage development, as it is able to induce ectopic GMC-1 fate(s) at the cost of the other GMC fates. Preliminary experiments show that, unlike in the NB4-2 lineage, Klu is not necessary for the fate of the second GMC of NB7-3 (data not shown). In accordance with this, Isshiki et al. (Isshiki et al., 2001) have shown that Kr is the factor that specifies the GMC 7-3b fate.

Taken together, our data suggest that there are principle mechanisms leading to the sequential cell specification in NB lineages, but different lineages seem to use either different genes to determine specific sublineages (e.g. klu) or use the same genes with different contributions (e.g. hb).

In any case, it is clear that an exact temporal regulation of their expression is crucial for the correct specification of neurons along the time axis. Therefore, an important aspect in understanding the sequential sublineage specification within NB lineages, is to know how sublineage genes like hb are switched on and off at the right time during lineage development. There are strong hints that the cell cycle plays an important role in this respect (Cui and Doe, 1995; Weigmann and Lehner, 1995). Owing to its accessibility, the NB7-3 lineage provides a powerful tool to investigate this and other related questions in the future.

We are very grateful to the following people who provided us with various antibodies, plasmids and fly strains: Bill Chia and Gerald Udolph (anti-Pdm1), Frank Hirth and Jürg Müller (anti-Hb), Rosa Barrio (anti-Sal), Manfred Frasch (anti-Eve), Ward Odenwald (anti-Cas), Jan Veenstra (anti-Crz), Marc Freeman (anti-eg, mouse), Zhi-Chun Lai (mouse anti-Zfh-1 and anti-Zfh-2) Michael Quick and Norman Davidson (pGEM-dSERT), Ernst Wimmer (UAS-hb fly strain), Christian Lehner (CycA fly strain), and Kathy Matthews of the Bloomington stock center. We especially thank R. Cantera who made us aware of Corazonin as a potential marker. In addition, we thank Gerd Technau, Andreas Prokop, Natalia Sanchez-Soriano, Tor Erik Rusten, Christian Berger, Christof Rickert, Benjamin Altenhein, Nirupama Deshpande and Ana Rogulja for their valuable comments on the manuscript. We especially thank Gerd Technau for his strong and continuous support during our studies. This work was supported by a grant of the Deutsche Forschungsgemeinschaft to J. U. (UR42/3-2).