Early tagma-specific commitment of Drosophila CNS progenitor NB1-1

The ventral nerve cord (vNC) of Drosophila derives from neural stem cells, the neuroblasts (NB) (Poulson, 1950), which delaminate from the ventral neurogenic region of the ectoderm (vNR) in a reproducible spatiotemporal segmental pattern (Hartenstein and Campos-Ortega, 1984; Hartenstein et al., 1987; Doe, 1992b). In the trunk, this pattern of segregating NB does not show any obvious differences between segments (Doe, 1992b). However, significant segmental-specificities in the vNC arise gradually during further development: in the course of embryogenesis segment-specific expression of several markers and genes arises (e.g. Jiménez and Campos-Ortega, 1981; Gould et al., 1990; Bourgoin et al., 1992; Graba et al., 1992; Mellerick et al., 1992; Hoshino et al., 1993), and proliferation patterns in the thorax and abdomen diverge gradually (Prokop and Technau, 1991) and become even more distinct during postembryonic stages (Truman and Bate, 1988). Eventually, in the imago, the volume of the thoracic neu-romeres clearly predominates over that of the abdominal ones (Hertweck, 1931).

In the early embryo a cascade of patterning genes provides the blastoderm with positional cues along the anterior-posterior as well as the dorsal-ventral axis. These coordinates finally define polarized metameric units, the segments (for review see e.g. Pankratz and Jäckle, 1993). Specification of these segments according to their location in the anterior-posterior axis is mediated by the homeotic selector genes (for reviews see e.g. Lewis, 1978; Duncan, 1987; Peifer et al., 1987; Beachy, 1990; McGinnis and Krumlauf, 1992; Morata, 1993). For the vNC of Drosophila it has been demonstrated that mutations in the bithorax-complex alter segment-specific patterns of markers and gene expression (Jiménez and Campos-Ortega, 1981; Graba et al., 1992; Gould et al., 1990) and eventually interfere with neuromere-specific morphology (reviewed by Doe and Scott, 1988). However, the mechanisms of segment-specific regulation in the CNS are not yet under-stood.

In order to address such mechanisms at the cellular level we made use of the previously described embryonic lineage of Drosophila neuroblast NB1-1 (Udolph et al., 1993). NB1-1 delaminates from abdominal and thoracic segmental domains of the neuroectoderm and gives rise to the aCC and pCC neurons in all of these segments (Doe et al., 1988a and b; Doe, 1992a). However, further components of the NB1-1 lineage express specific differences between thoracic segments (referred to as thoracic tagma) and abdominal ones (abdominal tagma) (Udolph et al., 1993). We show that tagma specificity of this lineage is determined at the early gastrula stage and is autonomously expressed by the neuroectodermal progenitors upon heterotopic transplantation. Activity of the homeotic genes Ubx or abd-A is autonomously required for the abdominal pathway of the NB1-1 lineage. By ectopic induction of Ubx-or abd-A expression until several hours after gastrula-tion or by homeotic derepression in Polycomb mutants, thoracic determination of NB1-1 can be overridden. About the same time, during normal development, UBX and ABD-A are detectable in NB1-1. We suggest that tagma-specific commitment of NB1-1 is induced by an early impact of segmentation genes on later homeotic gene expression.

For genetic analysis we used the following mutants: pb⁵ is a protein null allele of the proboscipedia gene (Pultz et al., 1988; kindly provided by T. Kaufman, Bloomington). Homozygous mutants were identified by the absence of a TM3 blue balancer chromosome carrying the lacZ gene as a marker (courtesy of M. González-Gaitán, Göttingen). Antp^W10 (Wakimoto et al., 1984; kindly provided by J. Castelli-Gair, Cambridge UK and W. Gehring, Basel) is a protein null mutation of the Antennapedia gene. Embryos mutant for Antp could be detected by the lack of the first midgut constriction (Tremml and Bienz, 1989). Ubx^6.28 (Kerridge and Morata, 1982; kindly provided by R. Reuter, Tübingen) is a deletion in the 5′exon region of the Ultra-bithorax gene (Beachy et al., 1985). Ubx mutants could be identified by the lack of the second midgut constriction (Tremml and Bienz, 1989) and by altered patterns of exit glia cells labeled in M84 (Klämbt and Goodman, 1991). The abd-A^MX1 mutation (Sánchez-Herrero et al., 1985; kindly provided by R. Reuter) is a deletion in the 5′ exon region of the abdominal-A gene (abd-A; Karch et al., 1985). Embryos mutant for abd-A could be identified by the lack of the second and third midgut constrictions (Tremml and Bienz, 1989) and by the lack of gonads (Cumberledge et al., 1992). Df(3R)Ubx¹⁰⁹ (Lewis, 1978; courtesy of R. Paro, Heidelberg; here referred to as Df109) deletes the Ubx and abd-A genes (Karch et al., 1985; Sánchez-Herrero et al., 1985). Df109 embryos could clearly be identified by the lack of gonads (see above), even following heat induction of UBX. To identify Df109 homozygotes in transplantation experiments the stock was balanced with a TM3 blue balancer chromosome (see above). Pc³ (Lewis, 1978; kindly provided by R. Paro) is an antimorphic allele of Polycomb, that produces a phenotype stronger than the null (Duncan and Lewis, 1982; Haynie, 1983). Homozygous embryos were detected by the use of a blue balancer chromosome (see above). For heatshock induction of Ubx we used the HSU-42 transformant line in a wild-type background or recombined into Df109 (González-Reyes et al., 1990; González-Reyes and Morata, 1990; both kindly provided by R. Reuter). This hsp⁷⁰-Ubx construct is based on the Ubx Ia splice variant. For induction of ABD-A we made use of a hsp⁷⁰-abd-A stock (González-Reyes et al., 1992; courtesy of E. Sánchez-Herrero, Madrid). Heat induced HSU-42 and hsp⁷⁰-abd-A embryos were clearly distinguishable by different effects on the thoracic cuticle according to previous descriptions. For labeling of the A- and B-glia cells we used the enhancer trap lines M84 (P-element insertion on the 2nd chromosome) and P101 (3rd chromosome) (Klämbt et al., 1991; Klämbt and Goodman, 1991; both kindly provided by C. Klämbt, Köln). To aid in NB identification we made use of the enhancer trap line 5953 (Doe, 1992b; courtesy of C. Doe, Urbana).

Transplantations were carried out according to Technau (1986) and Prokop and Technau (1993a). Donor and host embryos were of the same stage, approximately 10 minutes after the onset of gastrulation (stage 7). Using a pulled and ground glass capillary, with an inner diameter of about 10 μm, cells were removed from the vNR of the donors at about 0-20% ventrodorsal diameter and at either 30% egg length for abdominal or at 60% egg length for thoracic transplanta-tions (according to the early gastrula fate map; Technau and Campos-Ortega, 1985; see Fig. 1B in Technau, 1987). In about one third of the cases the cells were washed in a drop of Ringer’s or 0.2 M KCl before implantation into unlabeled hosts (see also Meise and Janning, 1993). This treatment was to clean cells from attached material of their microenvironment and to obtain a more stringent control on cell autonomy. In all cases single cell implantation into hosts was inspected with Nomarski optics. Hosts were reared into the embryonic stage 16/17.

Stage 16/17 embryos were fixed with glutaraldehyde saturated heptane according to Zalokar and Erk (1977). During fixation, embryos remained stuck to the cover-slip. After several washes with phosphate buffer (PB) the embryos were mechanically removed from the vitelline membrane with a pulled and broken glass capillary. Embryos were stained in a solution of 0.1% diaminobenzidine in PB for about 5 minutes. For double labeling experiments we used the enhancer-trap line P101 as donors and hosts. Following diaminoben-zidine treatment donors were stained for β-galactosidase activity in X-Gal solution (adding one part of 20% X-Gal in DMSO to 99 parts of a warm (approx. 60°C) solution of 150 mM NaCl, 1 mM MgCl², 3.3 mM K³[Fe(CN)⁶] in PB) at 37°C for several hours. For determi-nation of the genotype of donors carrying a blue balancer chromo-some, the embryos were raised to stage 14/15, fixed and washed as mentioned above (still stuck to the cover-slip), the vitelline membrane was torn with a broken capillary and the whole cover slip covered with X-Gal solution. Embryos that did not stain blue were homozy-gous for the mutation. Hosts that received cells from such embryos were treated separately. Embryos were dehydrated in alcohol, cleared in xylene, embedded in Araldite (Serva) and sucked into glass capil-laries of 0.2 mm inner diameter (Hilgenberg), to allow inspection from all sides (for details see Prokop and Technau, 1993a).

Staged embryos were stained with antibodies as described elsewhere (e.g. Schmidt-Ott and Technau, 1992). The final dilutions of primary antibodies were 1:1500 for anti-abd-A (Macías et al., 1990; kindly provided by E. Sánchez-Herrero, Madrid), 1:3 for anti-eve (Frasch et al., 1987; kindly provided by S. Romani, Cambridge), 1:5000 for anti-β-galalactosidase (Cappel) and 1:20 for anti-Ubx (White and Wilcox, 1984 and 1985; kindly provided by R. White, Cambridge). Following the diaminobenzidine reaction embryos were washed in PB. Specimens were dehydrated in ethanol, cleared in xylene and embedded in Araldite (Serva). Single embryos were selected and trans-ferred to a slide with a small drop of Araldite. Whole ventral nerve cords were dissected out using forceps and sucked into borosilicate capillaries with an outer diameter of less than 0.2 mm (Hilgenberg).

Flies were allowed to lay eggs for 30 minutes onto a filter paper soaked with yeast suspension. Egg laying and further development was carried out at 25°C. For heat shocks, the filter paper was tran-siently transferred to a 37°C wet chamber, so that the embryos directly contacted the warm water. After a heatshock period of 30, 45 or 60 minutes the filter paper with the embryos was kept in a humid chamber at 18°C. In combined transplantation- and heatshock-exper-iments the transplantations were carried out at room temperature throughout a period of approx. 1 hour, the coverslip with the hosts was transferred for 30 or 45 minutes to a humid chamber at 25°C and subsequently for 30 or 45 minutes to a 37°C weighing dish with 3S fluorocarbonoil. Further incubation took place at 18°C. In most cases donors were heat-shocked in parallel, subsequently incubated at 25°C and analyzed for the cuticular phenotype under the dissecting micro-scope prior to preparation of the hosts.

In all experiments, the genotype of embryos was verified by cuticle preparations (Wieschaus and Nüsslein-Volhard, 1986). Embryos were kept at 25°C until viable larvae hatched. The remaining eggs were collected with a needle, directly transferred to Hoyer’s medium, incubated at about 60°C for several hours and analyzed with Nomarski optics.

During embryogenesis Drosophila NB1-1 gives rise to a unique lineage. The sibling neurons aCC and pCC represent its first progeny in the thorax as well as in the abdomen (Doe and Scott, 1988). However, among further components of the NB1-1 lineage there are characteristic differences between these tagmata (Udolph et al., 1993): 2–3 subperineural glia cells (SuPn-cells) are part of the lineage only in the abdomen; the thoracic lineage, however, has a higher number of neurons, among them 1–2 motoneurons, that are not present in the abdomen (Fig. 1). The composition of thoracic and abdominal NB1-1 lineages was disclosed by DiI labeling, HRP-injections and isotopic transplantations of HRP-labeled neuroectodermal progenitor cells at the onset of gastrulation (stage 7; stages according to Campos-Ortega and Hartenstein, 1985). We then transplanted neuroectodermal cells heterotopically to test for determination of tagma specificity at the early gastrula stage. Those transplants that resulted in NB1-1 lineages were analyzed at stage 16/17.

(1) Cells were randomly taken from ventral sites of the thoracic vNR and transplanted into the ventral vNR of the abdomen. 8 NB1-1 lineages were obtained from these experi-ments. All of them showed the typical thoracic characteristics (Table 1, Fig. 2), indicating that the neuroectodermal progen-itors developed according to their origin. The ectopic motoneu-ronal projection of the cluster cells (see Fig. 2), which does not normally exist in the abdomen, projected into the region between the ventral oblique muscles 14.1 and 14.2 and the ventral tip of the lateral transverse muscles 21 to 23 (description of muscles according to Bate, 1993). These regions are comparable to the target sites of these motoneurons in thoracic segments (Udolph et al., 1993). In the single case where this motoneuronal projection was missing (No. 21 in Table 1), further criteria (10 cluster cells, no glia) pointed towards a thoracic identity of that lineage.

(2) Transplantations from the ventral abdominal vNR into the ventral thoracic vNR resulted in 12 NB1-1 lineages. All of them expressed the abdominal characteristics (Table 1; Fig. 2). One of these clones derived from a cell taken from the enhancer-trap line P101 (Klämbt et al., 1991; see Materials and methods), which expresses β-galactosidase (β-gal) in the A- and B-glia cells (Klämbt and Goodman, 1991). These cells cor-respond to the dorsal SuPn glia cells in the abdominal NB1-1 clone (Udolph et al., 1993). Although these cells are normally not found in the thorax of P101, at least the anterior dorsal glia cell in the transplant was double labeled for β-gal and HRP, demonstrating its identity as B-glia (not shown). The fact that we could not detect β-gal in the A-glia of this clone is most likely due to the limits of the double labeling technique (see Udolph et al., 1993). Taken together, these results indicate that neuroectodermal progenitors are committed with respect to thoracic versus abdominal specificity at the onset of gastrula-tion (stage 7), 40–50 minutes before the delamination of NB1-1 (see Doe, 1992b; Udolph et al., 1993). Cells at this stage autonomously express the thoracic or the abdominal NB1-1 lineage pathway.

Homeotic selector genes of the Antennapedia- and bithorax-complex (Lewis, 1978; Kaufmann et al., 1980) have been shown to coordinate segment specificity in the CNS of Drosophila (for review see Doe and Scott, 1988). In order to test their influence on the tagma specificity of the NB1-1 lineage, we analyzed the pattern of NB1-1 lineage components in embryos mutant for those homeotic genes that are expressed in most or all thoracic and abdominal neuromeres. As a marker we used the enhancer-trap line M84 (Klämbt et al., 1991; see Materials and methods) which, like P101, expresses β-gal in the A- and B-glia, arising from abdominal NB1-1 only. In addition, some of the mutants were stained with anti even-skipped antibodies (eve; see Materials and methods), which at stage 16 detect the aCC/pCC neurons in the neuromeres of T1 to at least A7 (Doe et al., 1988a and b; Doe and Scott, 1988; Patel et al., 1989).

The homeotic gene proboscipedia (pb) is expressed in all thoracic and abdominal segments, but is most strongly expressed in head segments (Pultz et al., 1988; Mahaffey et al., 1989). Embryos mutant for the null allele pb5 (see Materials and methods) did not reveal any pattern defects in the vNC when tested for M84-or eve-expression (data not shown).

Antennapedia (Antp) is expressed in all thoracic and abdominal segments with highest concentrations in the thorax (e.g. Levine et al., 1983; Martínez Arias, 1986; Bermingham et al., 1990). Inspection of embryos mutant for the null allele AntpW10 (see Materials and methods) did not reveal any pattern defects in M84- and eve-expression.

Ultrabithorax (Ubx) is expressed from T2p (p, posterior) to A9 with a peak in T3p/A1a (a, anterior) (reviewed by Beachy, 1990). Embryos mutant for the Ubx^6.28 allele (see Materials and methods) lacked the A- and B-cells in A1, where instead thorax-specific A/B-like glia cells could be detected; these do not derive from NB1-1 (Fig.3; Udolph et al., 1993). Thus, normal function of Ubx is required for the expression of abdominal characteristics of the NB1-1 lineage in A1.

The abdominal-A gene (abd-A) is expressed from A1p to A8 (Harding et al., 1985; Karch et al., 1990; Macías et al., 1990). Embryos mutant for the null allele abd-AMX1 (see Materials and methods) did not show any defect of the A- and B-glia pattern. However, in embryos homozygous for Df109 (a deletion covering the genes Ubx and abd-A; see Materials and methods), the A- and B-cells were clearly substituted with A/B-like glia in A1-7, whereas the pattern appeared unaffected in A8 (Fig. 3). These observations suggest that abd-A regulates abdominal development of the NB1-1 lineage in A2-7. However, this function seems to be replacable by Ubx (see Discussion).

To test whether homeotic genes regulate further tagma-specific components of the NB1-1 lineage, we made use of the significant phenotype of Df109. We transplanted cells at the early gastrula stage from the ventral abdominal vNR of Df109 into the same region of wild-type embryos. 7 NB1-1 lineages were obtained from donors homozygous for Df109 (as identified by the absence of a blue balancer, see Materials and methods). All of them expressed the thoracic characteristics: no glial cells, additional motoneuronal projection and 10-15 cells in the ventral cluster (Table 1). All NB1-1 from blue balancer-carrying donors expressed abdominal characteristics. These heterogenetic transplantations show that, independent from the genotype of surrounding tissue, the absence of Ubx and abd-A is sufficient to allow thoracic development of the complete NB1-1 lineage.

Taken together, the expression of Ubx or abd-A is not required for NB1-1 development per se but is essential for the expression of the abdominal variant of the NB1-1 lineage, whereas Antp and pb seem to play no role in NB1-1 develop-ment.

The heterotopic transplantations revealed that NB1-1 progen-itors are committed to thoracic or abdominal development at the early gastrula stage (about 3 hours AEL, stage 7). To determine the temporal profile in which Ubx expression exerts regulatory effects on the NB1-1 lineage, we crossed flies, carrying an hsp70-Ubx construct (HSU-42; González-Reyes et al., 1990; see Materials and methods), into the enhancer-trap line P101. Ubx was induced at different time periods and the A- and B-glia pattern was analyzed in thoracic neuromeres. P101 control embryos without the HSU-42 construct were insensitive to heat induction and no glial cells near the dorsal midline of the thorax were labeled (data not shown; see Udolph et al., 1993). However, P101;HSU-42 embryos heat shocked 4 hours AEL showed a complete or almost complete pattern of ectopic A- and B-glia cells in T1–3 (Fig. 3). Also Df109, HSU-42 embryos (carrying the M84 insertion; see Materials and methods), which at 25°C lacked A- and B-glia cells in A1–7 (see above), expressed these cells throughout the vNC when Ubx was heat induced 4 hours AEL (33 out of 37 analyzed Df109 embryos). Induction of Ubx later than 4 hours AEL led to a gradually decreasing number of ectopically induced A- and B-glia cells in the thorax (Fig. 4), although there was still an effect when heat induction started at 7 hours AEL (stage 11).

Similar results were obtained upon heat induction of abd-A (see Materials and methods): 60 minute heat shock starting at 4 hours AEL led to ectopic A- and B-glia cells in the thorax, but almost no effect was seen when the heat shock started at 7 hours AEL. Thus, the competence of NB1-1 to respond to these proteins follows a comparable time course.

To test whether heat induction of Ubx influences other segment-specific features of the NB1-1 lineage, HRP-labeled cells were transplanted from the presumptive thoracic region of HSU-42 embryos into wild-type hosts. 30–105 minutes after the transplantation, hosts were heat induced for 30 or 45 minutes (see Materials and methods) and then raised to embryonic stage 16/17 (Table 1). Two of the eight resulting NB1-1 clones expressed all abdominal characteristics (glial cells, no additional motoaxon, a small cell cluster), and two further lineages expressed abdominal features (glial cells, small cell cluster) as well as thoracic ones (an additional motoaxon). However, four of the clones were completely thoracic, likely due to the fact that Ubx was induced too late, suggesting that regulation of other segment-specific compo-nents of the NB1-1 lineage follow a time course similar to the A- and B-glia (compare variable effects in Fig. 4). Taken together, these experiments show that the induction of Ubx or abd-A expression about 4-6 hours AEL (stage 10/11) is suffi-cient to induce abdominal development of the complete NB1-1 lineage and to override thoracic commitment.

As ectopic expression of homeotic genes is genetically induced in mutants of Polycomb group genes (reviewed by Paro, 1990), we also analyzed the effect of Pc³ (see Materials and methods) on the NB1-1 lineage by looking at M84 and eve expression. In the abdomen, Pc³ embryos showed quite normal patterns of eve- and M84-positive cells, although cells were sometimes dislocated, and single cells, including A- and B-glia and aCC/pCC, were missing at random. For the thoracic vNC of Pc³ tagma-specific defects had been described before (Jiménez and Campos-Ortega, 1981) and were also detectable in our preparations: the number of anti-eve positive EL neurons (Patel et al., 1989; normally about 4 in the thorax) was increased to the abdominal numbers of 5-7 in all thoracic neuromeres. Also ectopic A- and B-glia cells were detected in most embryos, however, they were always restricted to T3, whereas T1 and T2 contained the normal set of A/B-like glia (not shown). Thus, although Pc³ in principle interferes with all thoracic neu-romeres, it affects the regulation of tagma-specific NB1-1 com-

ponents only in part of the thorax (see also below).

UBX and ABD-A are localized in NB1-1 during normal development

Referring to existing NB maps (Doe, 1992) we examined NB1-1 for the presence of UBX and ABD-A in the wild type. During stage 10, antibody staining for both proteins can be detected in the NB-layer. However, in our preparations strong staining of either protein in the nucleus of NB1-1 was first detected at early to mid stage 11 (Fig. 5). UBX was clearly detected in A1, but not in T3. ABD-A in NB1-1 was found in variable numbers of neuromeres within the abd-A expression domain A1p to A8a. Thus, UBX and ABD-A are first detectable in NB1-1 of the wild type at least 2-3 hours after its segment-specific commitment, at a time when heat induction still revealed regulative response of the NB1-1 lineage (see Discussion). Similar staining can also be detected in neighbouring NB such as NB6-1 or NB7-1

In Pc³ mutant embryos, at stage 16, the UBX and ABD-A pattern is similar in all neuromeres of the vNC and matches more or less the wild-type pattern of A7/8 (Beachy et al., 1985; Simon et al., 1992; own observations). However, these patterns are built up gradually (Simon et al., 1992; Struhl and Akam, 1985). Ectopic expression of both proteins in the NB layer starts at stage 10/11. UBX expressing NB are clearly more abundant in T3 than in more anterior locations (Fig. 6), although we did not detect nuclear UBX in cells likely to be thoracic NB1-1. Ectopic ABD-A could be clearly detected in NB near the border of T2/T3, but not, or only very weakly, in more anterior positions. In some specimens NB1-1 appeared to be among the ABD-A-positive NB in T3 as judged from morphological markers (see Fig. 6). Thus the staining pattern of UBX and ABD-A is consistent with the fact that A- and B-glia are ectopically induced in Pc3 only in T3 (see above).

Taken together, effects of heat induced homeotic gene expression, and of ectopic expression in Pc³, as well as the time when the proteins become detectable in NB1-1 suggest that Ubx and abd-A mediate tagma-specific differentiation of the NB1–1 lineage at stage 11, i.e. 2 –3 hours after the progenitor becomes committed for tagma specificity.

NB1-1 provides a good model system for the study of mecha-nisms underlying tagma speci-ficity in the developing CNS of Drosophila. It expresses tagma-specific lineage differences during embryogenesis and is readily amenable to detailed analysis by independent markers for specific lineage components and by experimental manipula-tions like cell transplantation (Udolph et al., 1993).

The transplantation experi-ments described here revealed that early gastrula (stage 7) CNS progenitor cells develop according to the tagma of their origin and autonomously express the thoracic or abdominal variant of the NB1-1 lineage. One could argue that the autonomous behavior of the cells is an artefact due to the transplanta-tion procedure, which might render the cells insensitive to external cues. This, however, is very unlikely, since studies on NB1-1 and other lineages revealed that transplanted prog-enitors react to and interact with the host tissue. For example, the isotopic implantation of a cell is regulated in the neuroectoderm of the host, as it does not inde-pendently produce an additional NB1-1 lineage (Udolph et al., 1993). Furthermore, single prog-enitor cells transplanted from lateral to ventral sites of the neu-roectoderm are amenable to signals inducing ventral cell fates. (Doe and Technau, 1993; Udolph and Technau, unpub-lished results; for further examples see Technau and Campos-Ortega, 1986; Stüttem and Campos-Ortega, 1991). Therefore, our transplantations strongly suggest that NB1-1 progenitors are normally already committed to thoracic or abdominal development in the neu-roectoderm, about 40 –50 minutes before they delaminate. It is important to note that these experiments do not address the mechanisms involved in the regulation of the intrasegmental specification of progenitors along the a-p axis.

As shown in other sets of experiments, neuroectodermal cells turned out to be committed also with respect to tagma-specific proliferation behaviour of NB during embryonic and postembryonic development (Prokop, 1993; Prokop and Technau, 1993b). Thus, early commitment may be a common feature of tagma-or even segment-specificity during develop-ment of the CNS. The same might also be true for other ecto-dermal derivatives (Simcox and Sang, 1983; Chan and Gehring, 1971).

Screening homeotic mutants with independent cellular markers revealed that the bithorax-complex genes, Ubx and abd-A, affect segment-specific features of the NB1-1 lineage. In the Ubx^6.28 null allele the A- and B-glia cells were missing in A1. This observation parallels former descriptions on various ecto-dermal derivatives where T2p to A1a express characteristics of T1p/T2a (CNS: e.g. Jiménez and Campos-Ortega, 1981; Ghysen and Lewis, 1986; Truman et al., 1993; PNS: Harten-stein, 1987; Heuer and Kaufman, 1992; cuticula: Lewis, 1978; Hayes et al., 1984). As shown by M84 expression and trans-plantations, Df109 embryos express the thoracic NB1-1 variant in all abdominal neuromeres except A8. Again, this phenotype is in accord with defects in other ectodermal tissues (CNS: Jiménez and Campos-Ortega, 1981; PNS: Heuer and Kaufman, 1992; cuticula: Lewis, 1978). The presence of A- and B-cells in A8 of Df109, suggests NB1-1 in A8 to be regulated by Abd-B (Casanova et al., 1986). Since Df109 deletes Ubx as well as abd-A (Karch et al., 1985; Sánchez-Herrero et al., 1985), and since Ubx only affects the A- and B-pattern in A1, we conclude that abd-A regulates segment-specific expression of the NB1-1 lineage in segments A2-7. The lack of any phenotype in abd-AMX1 (see also Graba et al., 1992) is likely to be due to the derepression of UBX in more posterior segments (Struhl and White, 1985; Macías et al., 1990), resulting in T3p/A1a-like differentiation in ectodermal tissues of A1p to A7/8 (CNS: Jijakli and Ghysen, 1992; PNS: Karch et al., 1990; Heuer and Kaufman, 1992; cuticula: Sánchez-Herrero et al., 1985). Further examples of substitution of abd-A function by Ubx have been reported previously (Gould et al., 1990; Graba et al., 1992; Röder et al., 1992; Vachon et al., 1992; Appel and Sakonju, 1993; see also Rowe and Akam, 1988).

Taken together, the absence of Ubx and abd-A is sufficient to allow for lineage autonomous thoracic development of NB1-1 (see also Palka et al., 1979; Morata and García-Bellído, 1985; Teugels and Ghysen, 1983). Conversely, ectopic induction of Ubx and abd-A leads to abdominal development of NB1-1.

Besides NB1-1, the NB that generates the thoracic A/B-like glia seems to be regulated by homeotic genes. Furthermore, homeotic mutations have been shown to affect other segment-specific features in the Drosophila vNC, such as expression of various genes and markers (Jiménez and Campos-Ortega, 1981; Graba et al., 1992; Gould et al., 1990), proliferation patterns in the embryonic and postembryonic vNC (Prokop, 1993; Truman et al., 1993; see also Booker and Truman, 1989; Miles and Booker, 1993), neuropile structures (e.g. Ghysen and Lewis, 1986; Jijakli and Ghysen, 1992), projection patterns (e.g. Palka et al., 1979; Strausfeld and Singh, 1980; Green, 1981; Thomas and Wyman, 1984) and function (Schneiderman et al., 1993). In most of these cases homeotic defects in the CNS more or less paralleled the epidermal phenotypes (for an exception see Ghysen, 1978).

The aCC/pCC neurons as the earliest components of the NB1-1 lineage derive from a ganglion mother cell born at about stage 9 (Doe et al., 1988; Doe, 1992b). They show no obvious segment specificity and do not suggest the need for early activity of homeotic genes, and it remains uncertain whether previously reported Ubx expression in abdominal aCC and pCC cells later on is of functional relevance (Doe et al., 1988a). Accordingly, antibody staining revealed UBX and ABD-A in NB1-1 not before stage 10/11. As the anti-UBX antibody is directed against all UBX-variants (White and Wilcox, 1984, 1985), positive staining does not tell us whether the detected protein is really functional, with reference to its splice variant and modifications (O’Connor et al., 1988; Kornfeld et al., 1989; Gavis and Hogness, 1991; López and Hogness, 1991). The same might be true for ABD-A. However, a functional relevance of the observed expression is suggested by the heat shock experiments, which still led to induction of A- and B-glia cells at about stage 11 (see also Michelson, 1994). The corre-lation between the heat shock results and the expression data becomes even more convincing if the delay in UBX-expression is considered. According to González-Reyes and Morata (1990) it takes up to 1 hour following the end of the heat shock for UBX expression to reach its highest level. However, the effects of early induced heat shocks can be easily explained by the fact that the level of ectopically induced UBX will remain high for at least 4 hours (González-Reyes and Morata, 1990). Also the spatial distribution of UBX at stage 11 supports its functional significance: UBX is located in the nuclei of NB1-1 in A1 but is clearly lacking in NB1-1 in the T2p/T3a domain of Ubx expression (reviewed by Beachy, 1990), and is thus in accord with the restriction of mutant effects to A1. Similarly, both expression and mutant defects of abd-A (as uncovered by Df109) coincided in A2-7. Taken together, the expression data argue for a requirement of UBX and ABD-A in NB1-1 starting at stage 10/11. However, the fading effects of later induced heat shocks suggest a requirement for UBX or ABD-A not later than stage 11/12 (see Fig. 4). A requirement for both proteins during stage 11 by the A- and B-glia is supported by the similarity of their transient expression patterns in Pc3 with the spatially restricted effect on the A- and B-glia pattern to T3 only. In mice also, the type of transformation of vertebrae resulting from ectopically induced Hox gene products is dependent on the time of manipulation (Kessel and Gruss, 1991).

As discussed above, UBX and ABD-A seem to be required at about stage 11 for abdominal development of NB1-1. However, as revealed by the heterotopic transplantation exper-iments, tagma-specific commitment of the NB1-1 progenitors is already achieved in the early gastrula (stage 7). Which factors mediate this early commitment ?

At the early gastrula stage only Ubx and abd-A transcripts are present and they are not translated before stage 8/9 (Akam and Martínez Arias, 1985; Beachy et al., 1985; Harding et al., 1985; White and Lehmann, 1986; Karch et al., 1990; Macías et al., 1990). Thus, early commitment of the cells might be achieved by the presence or absence of Ubx or abd-A tran-scripts at the blastoderm stage mediating homeotic gene expression at later stages e.g. via autoregulation (Christen and Bienz, 1992; Irvine et al., 1993). However, derepression of homeotic genes in Polycomb group mutants contradicts this view: although in Pc³ there were no Ubx and abd-A transcripts present in most of the thoracic region of the early gastrula stage, during later stages these mutants gradually build up a proper pattern of Ubx and abd-A expression even in the thorax, including expression in NB1-1 in T3 at stage 11 (Beachy et al., 1985; Struhl and Akam, 1985; Struhl and White, 1985; Wedeen et al., 1986; Simon et al., 1992 and 1993; own results). This suggests the presence of trans-activating mechanisms for Ubx and abd-A also in the thorax, although during wild-type development both genes are not or only partly activated in that region. Such late trans-activating mechanisms could involve CNS expression of segmentation genes like ftz (Doe et al., 1988a) and other gene classes like the trithorax-group genes (reviewed by Kennison, 1993). These activators in the thorax might also allow for an induction of Ubx or abd-A in hetero-topic transplants derived from the abdomen.

We suggest that early commitment of thoracic specificity is due to repression of Ubx and abd-A. A known trans-repressor for Ubx in the anterior region of the embryo is the hunchback gene (hb) (e.g. White and Lehmann, 1986; Quian et al. 1993; Zhang and Bienz, 1992). hb represses Ubx during early stages and the repressed state of Ubx seems to be maintained later on by Polycomb group genes (Paro, 1990). Consequently, the Ubx promoter is not ‘open for business’ (Peifer et al., 1987), although activators for Ubx are present throughout the vNC. This repressed state of the Ubx promoter in the anterior vNC (which can only be overridden by the Ubx enhancer element in some cells of parasegment 5, excluding NB1-1; Simon et al, 1990; Müller and Bienz, 1991) is cell autonomously expressed upon heterotopic transplantation (see also Paro, 1990) and could reflect the thoracic commitment from the early gastrula stage on.

We thank James Castelli-Gair (Cambridge, UK), Chris Doe (Urbana), Walter Gehring (Basel), Thomas Kaufman (Indiana), Christian Klämbt (Köln), Renato Paro (Heidelberg), Rolf Reuter (Tübingen), Suzanna Romani (Cambridge, UK), Ernesto Sánchez-Herrero (Madrid) and Rob White (Cambridge, UK) for providing fly stocks and antibodies, Michael Akam (Cambridge, UK), Chris Doe and James Castelli-Gair for comments on the manuscript, Chris Doe for advice on NB1 –1 identification, and Ursula Böll (Mainz) for help with the photography. This work was supported by a grant from the Deutsche Forschungsgemeinschaft to G. M. T. (Te 130/1 –4).