Ectopic expression of either the Drosophila gooseberry-distal or proximal gene causes alterations of cell fate in the epidermis and central nervous system

In Drosophila, the anteroposterior body axis is divided into a series of repeating units called segments. The process of segmentation is initiated at the beginning of embryogenesis by maternally deposited morphogens in the fertilized egg. Subsequently, a hierarchy of three groups of zygotic segmentation genes are sequentially activated to refine the maternal field of positional information (for review see Ingham and Martinez Arias, 1992).

In addition to patterning the epidermis, the segmentation genes are also required for the development of the Drosophila central nervous system (CNS). In gap and pair-rule gene mutants, lost epidermal segments are associated with elimination of the corresponding segmental ganglia. While these segmentation defects were expected, a second role for gap and pair-rule genes in the specification of particular neurons was not. A number of gap and pair-rule genes are reexpressed in segmentally repeating patterns in the developing CNS. From the analysis of conditional mutants, the expression of the pairrule genes, fushi tarazu (ftz) and even-skipped (eve) (Doe et al., 1988a,b), has been shown to be critical to the proper specification of certain neuronal identities.

The segment polarity genes occupy an intriguing position within the segmentation hierarchy. They are required in the epidermis to specify cell fate within each segment and are active both before and during the time that neuroblasts (NBs) are delaminating to generate the CNS. In a systematic study of CNS alterations in different segment polarity mutant embryos, Patel et al. (1989) have shown that five of the segment polarity genes (gooseberry, patched, Cell, wingless and hedgehog) cause specific defects in neurogenesis, while two other segment polarity mutants (fused and armadillo) have relatively normal CNS patterns. Only mutations in the patched and gooseberry genes cause CNS defects that appear to mirror the defects observed in the epidermis. These results suggest that segmentation of the CNS is not controlled by epidermal segmentation per se. Some of the segment polarity genes (patched and gooseberry) may play homologous roles in epidermis and CNS segmentation, perhaps by functioning in common ectodermal precursor cells, while others (wingless and Cell) function during neurogenesis in a second more restricted fashion. This suggestion is further supported by a more recent report showing that the repeated expression pattern of achaete (ac) in NBs is affected only in naked and gooseberry mutants and not in the other segment polarity mutants (Skeath et al., 1992).

Here we report our studies on the segment polarity gene, gooseberry (gsb), which plays a critical role in specifying cell fates in the epidermis as well as of NBs and their descendants. The gsb locus has been found to contain two highly homologous transcripts, gooseberry-proximal (gsbp) and gooseberry-distal (gsbd) (Baumgartner et al., 1987; Cöté et al., 1987). Both gsbd and gsbp encode putative transcription factors, each containing a homeodomain and a paired domain. In mutants deficient for both gsb genes, the naked cuticle of each epidermal segment is eliminated and replaced with a mirror-image duplication of the denticle belt (Fig. 8B; Nüsslein-Volhard and Wieschaus, 1980). The alterations in the CNS include the deletion of the posterior commissure and the CQ neurons as well as the duplication of aCC, pCC and RP2 neurons (Fig. 8H; Patel et al., 1989). The regions expressing the gsb genes closely correspond to the positions where structures are eliminated in gsb mutants. In the epidermis, gsbd has been localized to the region bracketing the parasegmental boundary of each segment and, in the CNS, gsbd and gsbp are expressed in a subset of NBs and neurons in the same region (Baumgartner et al., 1987; Ouellette et al., 1992; Gutjahr et al., 1993b).

In this study, we generated hs-gsbd and hs-gsbp fly lines to misexpress the gsb genes throughout the embryo. We have found that ectopic expression of either gsb gene causes cuticle and CNS alterations that are nearly reciprocal to that of a gsb mutant. Our results suggest that, in the epidermis, gsbd organizes pattern elements around the parasegmental boundary by regulating the expression of wingless (wg) and engrailed (en). In the CNS, gsbd may function as a selector gene to specify the cell fate of a subset of NBs. gsbp is expressed in the descendants of the gsbd-positive NBs and may provide continued gsb function in these cells.

The OrR, rosy⁵⁰⁶, Df(2R)SB1(gsb,Kr), Df(2R)IIX62(zip,gsb) are described in Lindsley and Zimm (1992). The A31 enhancer trap line was obtained from A. Ghysen and the wg enhancer trap line CyO/wg^17en40 was obtained from N. Perrimon. gsbp/lacZ is a P element transformant in which lacZ coding sequences have been fused in frame to gsbp genomic sequences at an EcoRI site in the third exon. The expression of β-galactosidase in gsbp/lacZ embryos precisely coincides with gsbp expression in the CNS (Fresquez, unpublished data).

To make gsbd protein, a 1 kb HincII fragment, which contains most of the gsbd coding region, was cloned into the T7 expression vector pET8c at the BamHI site. The protein was induced in pBL21(DE3) bacteria cells with IPTG and partially purified from inclusion bodies (Studier et al., 1990).

A full-length gsbp cDNA clone was isolated by screening a 0to 16-hour cDNA library (a gift from B. Hovemann) with an appropriate probe and confirmed by partial sequence analysis. An NcoI-NsiI fragment, which contains the entire gsbp coding region, was cloned into the T7 expression vector pET8c between the NcoI and XbaI sites. The gsbp protein was subsequently induced and partially purified from inclusion bodies.

Standard techniques for generating antibodies (Harlow and Lane, 1988) were used to make both rat polyclonal and monoclonal antibodies specific to gsbd and gsbp, respectively. In both cases, the obtained antisera also cross react with the other gsb protein. To remove the cross reactivity, the antisera were passed through a sepharose column coupled with the other gsb protein (for example, the anti-gsbd antiserum was passed through a gsbp coupled column and vice versa). The elimination of the cross reactivity in preabsorbed antisera was confirmed by western blots and embryo stainings. For monoclonal antibodies, immunized rat spleen cells were fused to mouse NSO-1 cells. Two monoclonal lines, 10E10 and 16F12, were isolated which specifically recognize gsbd. One monoclonal line, 9A1, was found to be specific to gsbp. Preabsorbed antisera and monoclonal antibodies give the same results in all the embryo stainings described in this paper; we refer to them generally as gsbd and gsbp stainings.

The antibody staining patterns were visualized using indirect immunofluorescence or horseradish peroxidase (HRP) techniques as previously described (Patel et al., 1989). All the fluorescent stainings were examined by confocal microscopy using a BioRad MRC-600 confocal microscope. For double labeling experiments with HRP coupled secondary antibodies, NiSO₄ and CoCl₂ were added to one of the reactions to give a blue-gray color (Lawrence et al., 1987). The rabbit anti-β-gal antiserum was generated in our laboratory (Holmgren, unpublished), the rabbit anti-eve antiserum was a gift from M. Frasch, the rabbit anti-wg antiserum was a gift from R. Nusse and the anti-ac antibody was a gift from S. Carroll.

Cuticles were mounted in a 1:1 mixture of Hoyer’s solution and lactic acid (van der Meer, 1977) and photographed under phase contrast.

A genomic clone containing the entire gsbd coding sequence and a cDNA clone containing the entire gsbp coding sequence were independently inserted into modified Carnegie 20 P element vectors (Rubin and Spradling, 1983) containing a 0.4 kb hsp70 promoter (from S. Lindquist). Each construct was then coinjected with the wings clipped plasmid into rosy⁵⁰⁶ embryos as described in Roberts (1986). Multiple insertion sites for each line were examined and gave similar phenotypes after heat shock.

Embryos were collected for 30 minutes and allowed to develop at room temperature for a specified time period as described in the text. The heat-shock treatment was performed by placing embryos onto prewarmed molasses-agar egg collection plates floating in a water bath at 36.5°C for 30 minutes. Multiple heat shocks were performed with a 1 hour recovery period following every 30 minute heat shock. To generate mutant phenotypes in the CNS and cuticles, the optimal time for heat shocking is 2.5-5 hours. Both heat shocked hs-gsbp and hs-gsbd embryos give similar phenotypes, though the hs-gsbp lines respond better than the hs-gsbd lines to late heat shocks. Whether this reflects differences in the intrinsic activity of the gsbp and gsbd proteins or more efficient expression from the intronless gsbp construct is not known. Heat shocks before 2.5 hours disrupt the segmentation of embryos, presumably by interfering with the normal function of the pair-rule gene, paired, since the gsbp protein and the paired protein can bind to the same DNA target sites (Treisman et al., 1991). Later heat shocks (up to 7 hours) can cause small regions within denticle belts to be transformed into naked cuticle.

In experiments correlating patterns of gene expression with cuticle phenotypes, each collection of heat shocked embryos was split: one half was used for antibody staining and the second half was used for cuticle preparation. In these experiments, at least 50% of the embryos showed cuticle transformations.

In order to express gsbd or gsbp ectopically in fly embryos, we generated hs-gsbd and hs-gsbp fly lines each of which has a gsb coding sequence under the control of an hsp70 promoter. Following the heat-shock treatment, there is a ubiquitous expression of gsbd in hs-gsbd embryos and gsbp in hs-gsbp embryos, which fades away within the first 60 minutes after heat shock (data not shown). In our study, the effect of this ubiquitous expression of gsb on cell fate specifications was assessed in the development of both the ventral epidermis and the CNS.

In gsb mutants, the naked cuticle of each segment is eliminated and replaced by a mirror-image duplication of the denticle belts (Figs 1B, 8B; Nüsslein-Volhard and Wieschaus, 1980). In the cuticles of heat shocked hs-gsbd and hs-gsbp embryos, groups of denticles or entire denticle belts are replaced with naked cuticle (Fig. 1C-F). The severity of the phenotype correlates with the timing of the heat shock. Heat shocks from 2.75 to 3.25 hours transform complete segments while those after 4 hours can transform sections of segments. Multiple heat shocks generate embryos with completely naked cuticles. To verify that cell fates are being transformed, we examined the Keilin’s organs, which are located at the parasegmental border in each thoracic segment. In gsb mutants Keilin’s organs are deleted, while in heat shocked hs-gsbd and hs-gsbp embryos, ectopic Keilin’s organs occasionally develop (data not shown). These results indicate that the cells in the anterior portion of the segment have been respecified and that the cuticle phenotype of heat shocked hs-gsbd and hs-gsbp embryos is nearly reciprocal to that of gsb mutants (Fig. 8C). It was somewhat surprising that heat shocked hs-gsbp was able to give pattern transformations similar to heat shocked hsgsbd since gsbp is not normally expressed in the epidermis at the times that the heat shocks were administered (Ouellette et al., 1992; Gutjahr et al., 1993b). Therefore, we examined the ability of heat shocked hs-gsbd and hs-gsbp to rescue the phenotype of Df(2R)IIX62 embryos which are deleted for the gsb locus. With a series of three heat shocks both hs-gsbd and hs-gsbp are able to rescue the formation of naked cuticle (Fig. 1G-H).

To determine whether changes at the molecular level correlate with the cuticle phenotype of heat shocked hs-gsb (hs-gsbd or hs-gsbp) embryos, we examined the expression of gsbd and two other segment polarity genes wg and en. We have found that hs-gsb embryos given heat shocks between 2.5 and 4 hours after egg-laying, activate ectopic expression of the endogenous gsbd gene (Fig. 2A,C). The ectopic stripe of gsbd expression is maintained 2 hours after heat shock and is dependent on the presence of a wild-type copy of the gsb locus (data not shown). Thus the ectopic gsbd expression must be due to the activation of the endogenous gsbd gene. The activation of ectopic gsbd is accompanied by the formation of an ectopic parasegmental groove (data not shown) flanked by inverted ectopic stripes of wg and en (Fig. 2A,C). Thus, it appears that the ubiquitous expression of gsb induces a mirror-image duplication of the region bracketing the parasegmental boundary. Heat shocks after 4 hours also transform the denticle belts into naked cuticle. These late heat shocks do not lead to ectopic en expression but correlate with a low level of wg expression throughout the segment (Fig. 2B,D).

In addition to epidermal expression, gsbd is expressed in a subset of NBs and their progeny in the posterior portion of each segment (Gutjahr et al., 1993b). gsbp is first expressed in NBs and is found at high levels in the progeny of these NBs (Ouellette et al., 1992). Later as the germ band starts to retract, low level of gsbp becomes detectable in the epidermis. Our analysis of the expression patterns of gsbd and gsbp generally agrees with the results by Gutjahr et al. (1993b), who found that in the CNS gsbd is expressed in row 5, row 6 NBs and transiently in NB7-1. One exception is that we find gsbd to be continually expressed in NB7-1 (as in Fig. 3A). gsbp is also expressed in the NBs of row 5, row 6 and 7–1 (Fig. 3B), and thus gsbp appears to follow gsbd expression in the CNS.

To confirm that gsbp is indeed expressed in all gsbd-positive NBs, we used a gsbp/lacZ line in which a lacZ gene is under the control of a gsbp promoter. Using double label experiments and confocal microscopy, we have found that the gsbp/lacZ construct is expressed in the exact same set of NBs and neurons as is the gsbp protein (data not shown). By double labeling with anti-β-galactosidase (β-gal) and anti-gsbd antibodies, it can be seen that β-gal (representing gsbp) is expressed in gsbdpositive NBs and their progeny (Fig. 3C-E).

It should be noted that the level of gsbd expression in GMCs and neurons is quite low and fades away during germ band retraction, whereas gsbp retains high levels of expression in GMCs and neurons as neurogenesis progresses. This suggests that, in the CNS, gsbd functions primarily in NBs and gsbp in GMCs and progeny neurons.

To study the effects of heat-shock-induced ectopic gsbd expression on NB cell fate, we used the anti-ac antibody. In stage 9 wild-type embryos, four rows of NBs are present within each CNS hemisegment and the anti-ac antibody recognizes NBs in row 3 and 7 (Figs 4A, 8D; Skeath and Carroll, 1992). In gsb mutants, Skeath et al. (1992) have shown that the ac protein is additionally expressed in the two NBs at the position of row 5 (Fig. 8E). In heat shocked hs-gsbd embryos, acpositive NBs are restricted to the NBs of row 7 in each CNS hemisegment and the staining of the row 3 NBs is eliminated (Figs 4B, 8F). These reciprocal transformations suggest that, in gsb mutants, row 5 NBs are transformed into row 3 and that, in heat shocked hs-gsbd embryos, row 3 NBs are transformed into row 5.

This conclusion was strengthened by analyzing wg gene expression in heat shocked hs-gsbd embryos. To monitor wg expression, we used a wg enhancer trap line CyO/wg^17en40 in which β-gal expression is regulated by the wg promoter (Perrimon et al., 1991). In wild-type embryos, β-gal from the wg enhancer trap is expressed only in row 5 NBs (Fig. 4C; Doe, 1992). In heat shocked hs-gsbd embryos, additional NBs anterior to row 5 NBs express β-gal (Fig. 4D), suggesting the transformation of row 3 NBs into row 5 NBs.

The effect of ectopic gsbd or gsbp on axonal pattern in the mature embryonic CNS was first examined by anti-HRP antibody staining. We have found that heat shocks at 2.75 to 3.25 hours of development for hs-gsbd embryos and 4.5–5.0 hours for hs-gsbp embryos give the most consistent alterations to the pattern of axons (Fig. 5A, wild type; Fig. 5B, hs-gsbd and 5C, hs-gsbp). The defects in heat shocked hs-gsbd and hsgsbp embryos are very similar, typified by the fusion of the two commissures within each segment and the elimination of the longitudinal connectives.

In order to follow cell fate changes in heat shocked hs-gsb embryos, we chose to examine anti-eve antibody staining, because anti-eve antibodies recognize subsets of neurons with different behaviors in gsb mutants: CQ neurons are deleted, aCC, pCC and RP2 neurons are duplicated, and EL neurons appear to be unaffected (Fig. 8H; Patel et al., 1989). Anti-eve antibody staining of heat shocked hs-gsbp embryos reveals pattern alterations nearly reciprocal to that of gsb mutants. Fig. 6B and 6D show that the eve expression characteristic of the RP2 and EL neurons is eliminated in most segments. In Fig. 6D, the eve expression expected in aCC and pCC neurons also appears to be eliminated in some segments, and there are putative duplications of the CQ neurons.

Using anti-eve antibody stainings, it is difficult to distinguish unambiguously the aCC and pCC neurons from the adjacent CQ neurons in heat shocked hs-gsbp or gsb mutant embryos. To resolve this problem, we used the A31 enhancer trap line. The A31 enhancer trap line has a P[lacZ] element inserted into the fasciclin II gene (Ghysen and O’Kane, 1989; Grenningloh et al., 1991). In the CNS of A31 embryos, β-gal is only expressed in the aCC and pCC neurons at early stage12.The A31 enhancer trap line was crossed into the hs-gsbp, hs-gsbd and gsb mutant backgrounds and the behavior of aCC and pCC neurons was examined by anti-β-gal antibody stainings. Two pairs of aCC and pCC neurons are in each CNS segment of wild-type embryos (Fig. 6E), and apparent duplication of aCC and pCC neurons occurs in gsb mutant embryos (Fig. 6F). Elimination of β-gal-expressing aCC and pCC neurons occurs occasionally in heat shocked hs-gsbp embryos (heat shocking at 4.5±0.25 hours) (Fig. 6G) and occurs much more frequently in heat shocked hs-gsbd embryos (heat shocking at 3.0±0.25 hours) (Fig. 6H). Since the aCC and pCC neurons are never duplicated in heat shocked hs-gsbp embryos, the duplicated neurons (Fig. 6D) are most likely CQ neurons.

The neuronal phenotypes observed in gsb mutant and heat shocked hs-gsbp embryos suggest that gsb function is both necessary and sufficient for the cell fate specification of a subset of posterior neurons, like the CQ neurons. It is important to examine whether gsbp is indeed expressed in these neurons but not in neurons which are duplicated in gsb mutants and deleted in heat shocked hs-gsbp embryos. Again, we used eve expression as our neuronal marker.

Double stainings of anti-gsbp and anti-eve antibodies in wild-type embryos show that anti-eve-stained CQ neurons also stain with anti-gsbp antibody, while the EL neuron cluster is outside the gsbp staining region (Fig. 7A). At a more dorsal focal plane of the same embryo (Fig. 7B), the aCC and pCC neurons are stained by the anti-eve antibody and appear to be above the gsbp-expressing neurons, while the RP2 neurons lie just anterior to the gsbp-expressing region.

It is known that the aCC and pCC neurons are progeny from NB1-1 (Goodman et al., 1984; Udolph et al., 1993). During neurogenesis, they migrate anteriorly to the region just dorsal to the gsbp-expressing neurons. To confirm that the aCC and pCC neurons are not labeled by anti-gsbp antibodies, a double labeling experiment was done with the A31 enhancer trap line. In the lateral view of an A31 embryo double labeled by antigsbp and anti-β-gal antibodies (Fig. 7C-E), it is clear that the aCC and pCC neurons are just outside the gsbp expression region. Thus, it is confirmed that gsbp is normally expressed in CQ neurons, but not in aCC, pCC, RP2 or EL neurons.

It is known that the lack of the gsb (gsbd and gsbp) activity results in the cell fate respecification of both epidermal and neural cells in gsb mutants. In this study, we set out to ask the converse question: does the ectopic expression of gsb genes commit cells to a particular cell fate during embryonic development? To this end, we have generated the hs-gsbd and hsgsbp fly lines in which ubiquitous gsb (gsbd or gsbp) expression can be induced by heat-shock treatment. Our results show that ectopic expression of either gsbd or gsbp causes pattern alterations which are approximately reciprocal to that of gsb mutants. In the epidermis, this results in the duplication of structures surrounding the parasegmental boundary and, in the CNS, neuroblasts which do not normally express gsb appear to be respecified to fates associated with the gsbexpressing NBs.

Previous results have shown that within each segment, gsbd is expressed in a region of the ventral ectoderm bracketing the parasegmental boundary (Gutjahr et al., 1993b). Eventually all wg-expressing cells express gsbd and a subset of the anterior en-expressing cells express gsbd. Epidermal cells in this region normally give rise to naked cuticle, but develop into denticle belts in gsb mutants. This transformation is correlated with the loss of wg expression during stage 11 (Hidalgo, 1991).

With heat shocks from 2.5 to 4.0 hours after egg laying, both hs-gsbd and hs-gsbp cause a transformation of the denticle belts into naked cuticle. Associated with this transformation is the generation of a new mirror-image duplicated parasegmental boundary with ectopic expression of wg, en and the endogenous gsbd genes. Our explanation of this result is that the gsb expression can activate inappropriate wg transcription that establishes the feedback loop normally found between the wgand en-expressing cells (DiNardo et al., 1988; Bejsovec and Martinez Arias, 1991). Establishing this ectopic feedback loop is dependent upon the endogenous gsbd gene since heat shocks with hs-gsbd or hs-gsbp in a gsb mutant background do not lead to ectopic expression of either wg or en and the anterior denticle belts are never transformed into naked cuticle.

Heat shocks after 4 hours of development are ineffective in establishing the wg-en cell feedback loop but still cause transformation of the denticle belt into naked cuticle. In these embryos, high level ectopic expression of wg, en and the endogenous gsbd is not observed but low level wg expression is seen throughout the segment. Experiments with wg^st mutants and HS-wg flies have shown that wg function is required through 11 hours of development for the formation of naked cuticle (Baker, 1988; Noordermeer et al., 1992). With later heat shocks, gsbd and gsbp are able to induce low levels of ectopic wg and this presumably leads to the formation of naked cuticle.

Both hs-gsbd and hs-gsbp are able to transform denticle belts into naked cuticle in wild-type animals and rescue the development of naked cuticle in gsb mutants. We believe that these functions are normally carried out by the gsbd gene since we cannot detect the gsbp protein in the epidermis of wild-type animals at the times that heat shocks are effective. Therefore, gsbp protein can partially substitute for the function of the gsbd protein. This might be expected given that gsbp can bind to the same DNA target sites as the related prd protein (Treisman et al., 1991). Redundancy between the prd and gsbd proteins may in part explain the continued expression of wg through stage 10 in gsb mutants since the prd protein is present until the end of germ band extension and is expressed in the same cells that express gsbd (Gutjahr et al., 1993a). We have examined wg expression in prd gsb double mutants (data not shown) and wg expression is reduced in the remaining even-numbered parasegments but not eliminated. Thus, it seems that another factor contributes to the maintenance of wg expression through stage 10 in gsb mutants.

In the CNS analysis, we focused on the cell fate transformations of a subset of NBs and neurons. Our assignment of neuronal and NB transformations is based on changes in the expression patterns of various proteins. It is possible that these changes do not reflect complete transformations of one cell type into another. This will be known only after neuronal phenotypes have been carefully analyzed. We believe that we are following cell fate transformations, as two neuronal markers (eve and A31 β-gal expressions) and two NB markers (ac and wg expressions) used in this study all behave appropriately.

At both neuronal and NB levels, we have found that gsb (both gsbd and gsbp) is normally expressed in those cells that are deleted in gsb mutants but not expressed in cells that are duplicated in gsb mutants. Ectopic expression of gsb leads to cell fate transformations that are nearly reciprocal to those of gsb mutants. These results suggest that the gsb locus may function as a classical selector gene, similar to the homeotic genes, and choose between different neuronal fates (Garcia-Bellido, 1975). The simplest model would be one in which there was a one-to-one correspondence between gsb-expressing and non-expressing NBs and their neurons. If this were the case, inappropriate expression of the gsb genes would lead to pattern transformations reciprocal to those in gsb mutants. This prediction is approximated in the initial patterning of NBs.

It is known that wild-type stage 9 embryos express the ac protein in row 3 and row 7 NBs but not in row 2 and 5 (Skeath and Carroll, 1992; Doe, 1992). In gsb mutants, ac expression is also found in row 5 (Skeath et al., 1992) and may represent the transformation of row 5 into row 3. (It cannot be a transformation of row 5 into row 7 since the NBs in row 5 of gsb mutants do not express en, while row 7 NBs normally do express en.) In heat shocked hs-gsbd embryos, ac expression is eliminated in row 3, suggesting a reciprocal transformation of row 3 into row 5. This conclusion is supported by examining the expression of β-gal from a wg enhancer trap line in hs-gsbd background. wg is normally expressed only in row 5 NBs (Doe, 1992). In heat shocked hs-gsbd embryos, the wg enhancer trap is also detected in some row 3 NBs (Fig. 5B). The role of gsbd in the specification of NB 7-1 is less clear. This NB also expresses en and it continues to express ac in gsb mutants.

Later in nervous system development, additional NBs are formed and a complex pattern of neurons is generated. With the anti-eve antibody, we can follow the progeny from four different NBs (Doe, 1992). The aCC and pCC neurons are derived from NB 1-1, an en-expressing NB located just posterior to row 7 (Fig. 7A). The RP2 neuron is derived from NB 4-2, which does not express gsb and is located just anterior to the gsb expression region (Fig. 7A). The lineage of the EL neurons is not known, but is outside the region of gsb expression. The CQ neurons are derived from NB 6-2 which expresses gsb (Fig. 7A). The effects of heat shock hs-gsbp on aCC, pCC, RP2 (deletion) and CQ (duplication) are the reciprocal to that in gsb mutants. This result again suggests the possibility that there are NBs and neuronal lineages that differ from one another only in their expression of the gsb genes. This correlation is not perfect, as the EL neurons do not fit this pattern. The EL neurons are present in gsb mutants, but do not appear to be duplicated. In heat shocked hs-gsbp embryos these neurons are consistently eliminated.

The roles of several segmentation genes in CNS development have been studied in detail. Some pair-rule genes, such as ftz and eve, have two independent roles, one in epidermal segmentation and a second in neurogenesis (Doe et al., 1988a,b). gsb is distinct from this paradigm since the pattern of gsbd expression in the CNS is always found in segmentally repeating stripes which begin in the neuroectoderm and are maintained as NBs delaminate. This may reflect similar independent functions in the epidermis and NBs or a single function in the common ectodermal precursor cells.

We thank Cindy Motzny, Darlene Buenzow, Jennifer Kennedy and Andrea Brand for stimulating discussions, and Linda Orlofsky for generating the hs-gsbp line. We are grateful to Lin Gu for her expertise and assistance in producing the monoclonals, Jim Skeath for providing his anti-ac antibody and sharing his expertise in the staining procedures, A. Ghysen for providing the A31 enhancer trap line, N. Perrimon for providing the wg enhancer trap line, M. Frasch for the anti-eve antiserum, R. Nusse for the anti-wg antiserum and C. Doe for sharing unpublished results of NB maps. This work was supported by an NSF Graduate Fellowship (A. U.) and an NIH grant no. NS28472 (R. H.).